开源模型、框架与仓库动态：权重开放、社区项目与许可证变化。
Across industries, the U.K. is putting the full stack to work — and the scale of deployment is growing fast, from enterprises transforming their businesses to startups solving some of the planet’s hardest problems. NVIDIA is putting that principle into practice, including through a five-year, $250 million compute commitment to strengthen the U.K.’s agentic safety […]
本文提出了一种极高效且无需训练的方法，可直接从现有的深度奖励模型中提取 Token 级奖励信号。我们的核心思想是通过估计每个 Token 的影响，将整体过程奖励归因于单个 Token。该影响被定义为：当一个 Token 被语义上无意义的 Token 替换时，最终宏观奖励（例如过程奖励）的变化。朴素地计算这种影响在计算上不可行，因为对于包含 N 个 Token 的序列，需要通过 PRM 进行 N 次前向传播。为克服这一瓶颈，我们提出了一种高效的基于梯度的估计器。具体而言，我们采用一阶泰勒近似，将影响的计算简化为 Token 嵌入与无意义 Token 嵌入之差，与奖励相对于 Token 嵌入的梯度之间的内积。该方法仅需一次前向传播和一次反向传播。由此得到的 Token 级奖励使标准 RL 算法能够在无需额外训练奖励模型的情况下，执行精确的信用分配。在具有挑战性的推理基准测试中，实验结果表明，我们的方法显著提升了策略优化效率，并增强了 LLM 推理能力的泛化性。对于 Qwen2.5-VL-7B-Instruct，我们的 P2T 在 MathVista 上相比结果奖励提升了 +4.9%；对于 Qwen2.5-Math-7B，在 AIME24 上提升了 +11.5%。代码将通过 https://github.com/stonewst/P2T 发布。
当前的机器人学习方案需要大量机器人示范，才能获得稳健且具备适应性的策略。在本文中，我们介绍了 Steertuning，这是一套能够让机器人仅通过 5 段人类示范，在不到 5 分钟内于新环境中学习新任务的系统。Steertuning 由两部分组成。（1）我们在自建的第一视角人类数据集上预训练了一个拥有 4 亿参数的扩散策略模型。该数据集包含 100 小时经过密集标注的随机交互数据。这种预训练使 Steertuning 能够适应背景差异，并在部署到新任务时实现空间泛化和物体泛化。（2）给定新任务的 5 段示范后，我们仅用 2 分钟（2000 次训练更新）对模型进行微调，并在反向采样过程中，使用观测结果与部署环境最接近的示范来引导策略的动作。结合专为本体迁移设计的数据处理流程，Steertuning 在 12 个新环境中的 12 项任务上取得了 71% 的成功率，这些环境涵盖实验室场景和新住宅中的厨房。从新任务的数据采集到机器人策略部署，整个过程耗时不到 5 分钟。我们的完整数据集、模型权重、代码库和机器人视频将在我们的网站上公开：https://steertuning.github.io。
长程智能体将任务进展积累在持久化任务状态中，却仍反复处理描述这些进展如何取得的交互历史。这引出了一个自然的问题：长程上下文能否围绕可复用的进展来组织，而不是围绕产生这些进展的历史？本文介绍 CarryOn，这是一种面向对象的上下文运行时，围绕可复用的任务对象组织任务续接。CarryOn 不将压缩后的历史或重建的进展描述作为任务续接的主要依据，而是基于续接载体、补充性执行上下文以及持续更新的近期交互尾部构建上下文视图。一个核心挑战是，在相关历史被移除后，如何保持这一可复用基础的有效性：CarryOn 如何在不于每次折叠时重建完整表示的情况下更新这一基础？CarryOn 跨折叠维护载体记录，依据绑定的证据和适用的环境检查修订受影响的记录，同时保留未受影响的记录，并同步更新执行上下文。在 OSWorld 2.0 和 BrowseComp-Plus 上的实验表明，性能分别提升了 5.8% 和 7.0%，同时总 token 用量分别减少了 51.9% 和 22.9%。项目页面：https://ethannju.github.io/carryon
Harness self-evolution revises an agent's software scaffolding, including prompts, memory, workflows, and executable components, while keeping model weights fixed. We formalize a feedback-driven process in which accepted revisions persist across runs. The framework separates update targets from proposal and admission mechanisms. It distinguishes changes to interpreted state from revisions to tools and code, repository-level control logic, and plugin composition. We characterize search through evolutionary optimization and examine its reliability through continual PAC-learning. Under explicit assumptions, the analysis yields an expected generalization bound, a retention bound conditional on historical regression budgets, and a high-probability admission rule using fresh, independent validation data. We examine how persistent harness revisions affect knowledge sharing and organizational workflows, with implications for scientific reliability, physical safety, and institutional accountability. Cases from prior work illustrate model--harness co-design, distributed execution, and plugin reuse. Reliable deployment requires continual validation as task distributions and evaluators change, with independent safeguards for learning, runtime privileges, and consequential actions. We provide a comprehensive 66-page discussion of related work, theory, algorithms, evaluation, applications, and the broader impact of Harness RSI, together with continuously updated code and supplementary materials, at https://github.com/zgcagi/Harness-RSI.
This paper investigates offline Q-learning for finite-horizon Markov decision processes with Kullback-Leibler (KL) regularization. Recent studies have shown that KL regularization can yield fast rates in offline decision-making problems. However, it remains unclear whether Q-learning, which adopts incremental stochastic approximation updates, can achieve such fast rates in the offline setting. We first propose KL-regularized Q-Learning (KLQ), a simple variant of classical Q-learning. By exploiting KL regularization, we prove that KLQ achieves a sample complexity of $\widetilde{O}(1/ε)$ without explicit pessimism. This guarantee, however, has an exponential dependence on the KL regularization parameter and the horizon. To address this limitation, we further propose KL-regularized Pessimistic Q-Learning (KLPQ), which incorporates pessimism into KLQ. We prove that KLPQ retains the same \(\widetilde O(1/ε)\) rate while removing the explicit exponential dependence, thereby further improving the sample complexity. To the best of our knowledge, these results establish the first fast-rate guarantees for offline Q-learning in tabular finite-horizon KL-regularized MDPs.
机器人在任务执行方面的自主性日益增强，但其改进仍主要由人类驱动。本研究探讨物理递归自我改进（Physical RSI）：利用物理交互经验，指导对机器人能力及其改进流程进行持久性更新，从而使后续改进在不同任务中更加有效。Physical RSI 面临观测噪声、连续动力学、高昂的试验成本以及不可逆后果等挑战。我们提出了一个 L0-L5 框架，区分能力层面与元层面的更新，以及任务内与跨任务的增益，并将 L5 专门用于经过独立验证的递归增益。我们考察了围绕运行支撑框架适配和模型训练的现有方法，以及对执行过程和连续更新进行管控的安全与控制机制。我们的分析表明，尽管自我改进已取得显著进展，但所调研的文献尚未展示出符合本框架定义的 Physical RSI。我们确定了可靠反馈、因果归因、安全性保持和基准评测等研究重点，勾勒出一条使机器人在确保安全的同时，自主改进其学习方式的发展路径。Physical RSI 的 GitHub 仓库位于 https://github.com/SAIL-Research-Lab/Physical-RSI。
Reinforcement learning with verifiable rewards (RLVR) relies on costly model-generated responses, motivating larger rollout batches and square-root learning-rate scaling. Reported to approximately preserve progress per sample with one optimizer step per rollout, this rule is less understood in practical multi-step training. We develop a two-clock framework that separates the samples needed to make progress from those consumed before collapse. Under a local quadratic model, the square-root rule preserves first-order drift per sample but not progress: larger batches take fewer, longer updates and lose more to curvature, yielding a learning-rate-dependent critical batch beyond which sample efficiency degrades. A drift-threshold hypothesis predicts a complementary failure clock whose sample count remains approximately invariant along the square-root rule. Controlled GRPO experiments on Qwen2.5-Math-1.5B across a $16\times$ batch range support both predictions. Ratio clipping delays failure, especially for larger batches, but the progress gap persists without it. Widening the clip trades stability for a transient speed gain. The progress gap reappears on another dataset, a general-purpose checkpoint, and a 7B model. Finally, we use the two-clock framework to derive a simple multi-step RLVR schedule: start below the critical batch and switch to a larger batch before failure. Our code is available at: https://github.com/WinstonLiyt/two-clocks-rlvr.
我们如何构建一种智能体，使其能够超越人类语言进行思考，同时仍通过语言在外部环境中采取行动？现代 LLM 智能体仍通过文本进行推理和交互，而潜在空间推理虽然提供了高效的内部媒介，却缺乏原生工具访问能力和推理透明度。我们提出 Hybrid Reasoning Agents（HybridAgents），将这两个方向结合起来，使智能体能够在统一的生成过程中动态切换潜在空间推理与基于文本的交互。在架构上，轻量级的 HybridLink 支持文本模式与潜在模式之间的无缝切换，而智能体的连续工作记忆则在多轮交互中无损保留混合上下文。这些设计既保留了集成工具的智能体能力，又使潜在思维可被直接检查。基于这一架构，我们进一步开发了专门用于混合推理的后训练算法：HybridSFT 通过教会骨干模型与 HybridLink 协同工作，稳定集成工具的混合生成过程；随后，HybridRL 通过学习何时在潜在空间中思考、何时通过文本采取行动，进一步增强推理能力。我们在涵盖数学、科学、医学、搜索和智能体推理的 11 项基准测试上，对 3 种模型规模的 HybridAgent 进行了评估。与涵盖文本方法和潜在空间方法的前沿 LLM 智能体及智能体框架相比，HybridAgent 持续实现 10.4% 的性能提升，同时将 token 用量减少 35.5%–54.7%，将推理速度提升至 1.8–2.6 倍，并将成本降至原来的 1/2.7–1/3.7。代码和数据可在 https://github.com/togethercomputer/HybridAgent 获取。
推理模型通过较长的思维链（CoT）取得强劲性能，但通常会使用超出解题所需的词元数量。现有的 CoT 缩短方法需要针对每个模型进行额外训练、向模型的隐藏状态添加固定的引导向量，或采用提前停止、禁止使用“wait”等词语的解码启发式方法，而这些方法可能会降低准确率。我们提出 DoubtLess，这是一种无需训练的解码方法，通过一个小得多的推理模型引导大型模型生成更简洁的内容。小模型会读取大型模型截至当前的 CoT 两次：一次被提示要简洁（简洁思考者），另一次被提示要详细（详细思考者）；然后将其两次 logit 向量之差乘以一个强度系数，并在 CoT 的每一步将结果加到大型模型的 logits 上。在涵盖竞赛数学、研究生水平科学和竞赛编程的七个基准测试中，DoubtLess 在准确率不变的情况下将 Qwen3-32B 的 CoT 缩短了 31%，优于五种推理时基线方法。我们还表明，在一个自调试编码智能体中，DoubtLess 在任务表现不变的情况下，将每轮思考词元数减少了 20%。在另外两个模型系列上，它同样实现了准确率不变的泛化效果。与基于训练的压缩方法结合后，它还能进一步缩短已经接受简洁性训练的模型的 CoT。消融实验和行为分析表明，DoubtLess 减少了重复检查，而大部分词元节省发生在 CoT 中首次出现正确答案之后。我们的代码可访问：https://github.com/leoluolol/DoubtLess。
长时域智能体凭借能够自主执行具有经济价值的任务而日益受到关注，其工作时长可从数小时持续到数天。这一能力对于自动化发现和递归自我改进（RSI）至关重要。然而，开源模型在长时域任务上仍然表现不佳，主要原因在于缺乏训练数据和可行的训练方案，而同时扩展这两者十分困难：长而复杂的任务难以整理和验证，且其奖励通常十分稀疏。我们同时解决了这两个问题。首先，我们推出了 Tev-Gym，这是首个具备可验证、稠密且渐进式奖励的长时域智能体训练数据集，涵盖 7 个领域的 1,170 个已验证任务环境。该数据集采用“潜在世界优先”的方法构建：首先建立一个潜在世界，即对任务领域、工作流以及每一步之后预期状态的完整且正确的隐藏规范。随后，每个任务都根据该规范逐步呈现为面向智能体的环境，从而实现可控的复杂度和自然的进度指标。其次，我们将这一指标转化为训练方案：通过进度势能奖励塑形（PPRS），将进度转化为每一步的稠密奖励，从而提高训练效率，并使原本难以训练的任务变得可训练。采用该方案训练得到的 Tev-27B，在 Agents' Last Exam 的计算与数学科学领域（ALE-CM）上取得了显著提升，在 Pass@5 指标上达到 GPT-6 Astra 和 Claude Opus 5 的水平。尽管仅使用与 ALE 相关的数据进行训练，它仍能泛化到 Terminal-Bench（TB）3.0 和 4.0 子集，超过 GLM 5.2 和 Gemini 3.7 Flash，并在 TB 2.1 等中等时域任务上达到 Claude Opus 4.8 的性能水平。
在视觉-语言-动作（VLA）预训练过程中，应如何利用示范质量来提升目标任务微调后的性能？我们使用 5,000 条自然产生的次优人类示范和 24 个目标任务，在不同奖励定义和标签粒度下比较了加权、过滤和条件化方法。对于每个 VLA，我们确保各方法使用相同的示范和训练预算。其他比较还考察了预训练数据规模，以及质量标签在微调和推理过程中的作用。在三个 VLA 上，我们发现：感知接近过程的回合级条件化相较于普通行为克隆提高了平均成功率；平均而言，条件化和过滤分别更适合回合级标签和片段级标签，但具体模型上也会出现相反的结果；在条件化预训练之后，无论推理时是否使用标签，采用固定标签进行微调都能提高平均成功率。我们推出了 AXIS-Bench，以支持这些受控比较。该基准基于众包仿真示范构建，提供了记录的物体状态和接触状态，可用于在共享轨迹上重新计算质量标签。可回放轨迹支持与工作空间对齐的渲染，以及用户自定义的真实世界目标任务。在我们的真实世界实验中，仿真预训练将成功率从 9.6% 提升至 31.7%。项目页面：https://axis-bench.github.io/。
Google has made Gemini 4 Argon generally available on the Gemini API, AI Studio and the Gemini app. The sparse MoE model writes up to 1M output tokens and leads the Vals Index at 68.90%. The post Google Releases Gemini 4 Argon API With 1M Token Output Limit appeared first on MarkTechPost .
基于 LLM 的多智能体协作正日益用于解决复杂任务，其方式是将工作分配给专业化智能体，并整合互补的专业知识。然而，现有系统和评测大多建立在一种隐含的完美同事假设之上：协作者提供正确且完整的信息，并及时响应请求。但在现实团队协作中，协作者可能返回与预期问题不符的信息，提供不足以支持决策的细节，或未能响应请求。为了将任务解决能力与从不可靠的信息搜集过程中恢复的能力区分开来，我们引入了 Unreliable Colleagues Benchmark（UnCoBench），通过 Misaligned、Incomplete 和 Missed 三种设置对行为失败进行组织。主智能体需要验证上下文正确性，诊断信息缺口，并跟踪尚未解决的依赖关系，从而获取解决任务所需的正确证据。我们还提出了一种可验证且可扩展的自动生成流程，将基于 LLM 的场景生成与答案唯一性和角色必要性的可执行检查相结合。对十个模型的结果表明，在可靠设置下表现相近的模型，在面对不可靠证据时可能展现出不同的稳健性。轨迹分析揭示，智能体经常遗漏纠正性后续行动，而恢复行为还涉及低效等待和冗余查询。这些发现表明，稳健的协作不仅需要可靠的推理，还需要采取有针对性的行动来诊断证据失效、恢复相关信息，并将其整合到决策中。我们的数据和源代码可在 https://github.com/iNLP-Lab/UnCoBench 获取。
An agent can retain successful trajectories without learning which part of an experience should repair a new discrepancy. We formulate passive imitation (PI) as an architectural loop that localizes a violated contract, extracts a local goal, proposes an effect-matched skill, tests a small patch, and retains evidence after full-task verification. We instantiate this loop as a deterministic symbolic controller and analyze its finite repair-query complexity under explicit identifiability and noninterference assumptions. Across ten simulated tool worlds, PI solves all 600 cold held-out composition tasks with three allowed small trials, compared with 6.17% for a whole-trajectory control, while an equally expressive compositional repair control matches PI. On an unseen tool revision, PI achieves 70.0% success versus 24.0% without small trials and 79.0% with self-trial enumeration; an exploratory 24-task adaptation stream subsequently raises first-submission success to 100%, supporting a limited correction-and-retention mechanism rather than a unique incentive advantage or general intelligence. The CPU-only implementation, integrity checks, and reproducible results are available at https://github.com/FriendOfYishan/PI.
由于谜题在部分信息下设定了可验证的目标，并要求通过多步决策达到解决方案，因此它们为大型语言模型（LLMs）的推理能力提供了天然的测试平台。然而，现有的谜题基准通常聚焦于机制僵化、解答目标高度受限的狭窄环境。为弥合这一差距，我们提出了 Active Probing Benchmark（APB），通过将概念识别构建为一种交互式推理任务，将专业知识评估融入谜题场景，从而推广谜题求解。具体而言，APB 评估模型能否通过多轮主动探测与推理获取信息，从八个多样化的学术领域中精准识别出 960 个具有专业难度的目标概念。我们的评估揭示了一个违反直觉的现象：大多数开源推理模型的表现显著逊于其标准指令模型对应版本。这一发现表明，在交互式场景中，这些模型难以运用程序性推理，可靠地获取和应用专业领域知识。除这些实证发现外，APB 还提供了一套可扩展的流程，用于生成能够在多样化场景下探测交互式知识推理能力的基准。
Omni-modal large language models (Omni-LLMs) extend LLMs to process heterogeneous modalities (e.g., vision, audio, and IMU signals) alongside language, enabling a broader range of multimodal tasks. Yet we find that these models often fail to express information from non-textual modalities even when they perceive it correctly. We term this disconnect knows-but-cannot-say. To investigate this phenomenon, we introduce Token-level Modality Alignment (TMA), which measures the alignment between individual modality representations and response representations. Our analysis shows that non-textual representations are less aligned with response tokens and receive less attention than textual representations, suggesting a potential explanation for uneven modality preferences. Motivated by these findings, we propose OmniCatalyst, a simple perceive $\rightarrow$ infer strategy that complements the original modality inputs with self-generated text, making non-textual knowledge more accessible. Across seven Omni-LLMs, OmniCatalyst consistently improves performance (+7.29%) and yields more balanced modality preferences, with further gains from diverse training strategies. Further analysis demonstrates that the knowledge extraction phases increase TMA, suggesting that improved alignment makes non-textual information more accessible during response generation. In summary, our findings uncover how Omni-LLMs can better express modality knowledge through language. Our code is available at https://github.com/naver-ai/OmniLLM.
大型语言模型智能体日益在多个实体上运行，而可靠的响应生成不仅需要检索相关证据，还需要保持证据与正确实体之间的关联。我们发现了幻影合并（PM）现象：这是一种实体—证据绑定失败，智能体会将未经支持的声明归因于其已确定的实体，即使任务实际上已经正确完成，也可能发生这种情况。我们在声明层面对 PM 进行了形式化定义，并刻画了其在四类智能体领域和四类骨干模型系列中的发生率及错误子类型。为解决这一问题，我们提出了一个用于声明级风险估计与修正的两阶段框架。第一阶段，Anchor-Grounded Risk（AGR）将智能体生成输出前隐藏状态的表征读出与基于雅可比矩阵的功能信息相结合，提供互补的槽位级风险分数和锚点值风险分数。第二阶段，Fixed-Anchor Correction（FAC）审查被标记的声明，并利用观测到的锚点证据对其进行选择性修订；当没有受支持的替代内容可用时，则删除这些声明。在留出的 Shopping 检测队列上，AGR-slot 达到 96.2% 的 AUROC 和 97.3% 的平均精确率，优于所评估的检测基线。在端到端 Shopping 缓解评估中，FAC 将轨迹级 PM 从 51.4% 降至 6.2%，输出与原始声明的比例为 80.7%。这些结果表明，实体特定的落地失败普遍存在，而将内部风险估计与受证据约束的声明修正相结合是有效的。我们的代码已公开发布于 https://github.com/LabRAI/PhantomMerge】【。
交互式智能体评估通常依赖标准化的、基于 LLM 的用户模拟器，这些模拟器表现出广泛的合作行为，可能忽视多样但现实的用户交互所引发的失败。这会导致对智能体鲁棒性的估计过于乐观，也难以为训练能够跨用户泛化的智能体提供足够的失败信号。现有方法大多依赖预定义的用户行为，可能无法发现特定于智能体或领域的弱点；而面向失败的搜索则可能通过使底层任务失效来降低任务成功率，而不是暴露真正的鲁棒性缺陷。为弥补这一空缺，我们提出 TIPER，一种保持任务完整性的人设演化方法，能够在保持底层任务有效且可解决的同时，自动发现暴露智能体特定鲁棒性弱点的用户策略。TIPER 在任务完整性约束下，通过反思引导的行为变异和按任务进行的 Pareto 选择，演化出持久的用户人设。在四个交互式智能体基准测试中，TIPER 在全部 12 个 GPT-5.4 设置下都取得了更低的留出集 avg@3 分数，观测到的降幅为 5.8 至 33.7 个百分点；在涵盖七个智能体的全部 28 个智能体—领域组合中也同样如此，观测到的降幅中位数为 22.9 个百分点。进一步分析得出了一套从行为到失败的分类体系，将发现的用户行为与反复出现的智能体失败模式联系起来。总体而言，这些结果表明，TIPER 能够对交互式智能体开展更具诊断性的鲁棒性评估，并为未来改进智能体鲁棒性提供有价值的失败分析。代码仓库：https://github.com/UCSB-NLP-Chang/TIPER。
我们提出了交互式四维体积液体预测任务：根据固体物体在液体中的运动方式，预测未来稠密三维液体速度场的时空演化。该任务为需要推理运动条件下液体响应的系统提供了上游物理预测基础，包括机器人液体操控、数字孪生和刚体-流体视频生成。现有研究已尝试预测静态或预设固体几何体周围的流动，并合成视觉上合理的流体运动。然而，这些方法对交互式液体动力学的覆盖有限，因为它们无法预测运动固体如何引发未来全域液体响应。在这一预测场景中，流体动力学效应集中在不断演化的液-固界面附近，但预测目标却是全局时空演化过程。为应对这一挑战，我们提出了一个两级循环预测框架 TIDE/TIDES：TIDE 专注于局部界面驱动的预测，而 TIDES 增加了感知输运的耦合和显式稳定化，以减少累积的演化误差。在经过适配的邻近基线、受控组件消融、运动/尺寸分布外测试以及长时域评估中，TIDES 均取得了最佳预测性能；消融实验表明，局部交互、输运耦合和稳定化是关键贡献因素。总体而言，我们提出的任务定义、基准以及 TIDE/TIDES 研究，为运动固体条件下的稠密液体预测建立了一个受控基础。代码已公开于 https://github.com/jongwonryu/TIDE-TIDES。
Flagship large language models (LLMs) offer rich general knowledge and strong reasoning capabilities, but their high inference cost and latency limit their use for large-scale data labeling. We study zero-shot text classification with semantically meaningful labels and task descriptions, without task-specific labeled examples or classifier feedback. We propose \textbf{SLKR} (Semantic Label Knowledge Retrieval), a label-centric retrieval-augmented generation framework that separates offline knowledge construction from lightweight online inference. Offline, a flagship LLM independently generates label definitions, retrieval keywords, and boundary rules for potentially confusable label pairs, forming a reusable label-knowledge store. Because this construction operates over the label space rather than the input corpus, its one-time flagship cost can be amortized across large-scale inference. Online, hybrid semantic and lexical retrieval first identifies the top-$K$ candidate labels and then selects up to $M$ input-relevant boundary rules connecting candidate-label pairs. A lightweight LLM performs the final classification in a single call using the retrieved definitions and rules. Across three public benchmarks and a real-world multilingual book-classification dataset, SLKR achieves the best results on all four datasets with Qwen3.5-4B and on three of four with Ministral-3-3B, while using compact input contexts and no task-specific labeled examples. Boundary ablations further show that retrieving a small set of input-relevant rules preserves performance close to using all boundaries with substantially less context, supporting efficient reuse of flagship-generated label knowledge for lightweight classification.
A model may select the correct answer yet fail to construct it through a sequence of local decisions. We introduce Choice2Seq, a diagnostic benchmark that measures correctness retention across complete-answer selection and sequential construction in independent runs on the same problems and candidate sets. Sequence Composition Rate (SCR) measures the fraction of correctly selected items that are also answered correctly through local construction. Jev achieves 97.1% and 99.5% selection accuracy on arithmetic and lexical retrieval, yet retains only 33.0% and 2.1% of those correct choices under serial construction, respectively. Substantial gaps also emerge for StartLux and Qwen, including when Qwen emits characters directly. Restricting local options to candidate-supported characters improves construction, while explicit position indexing improves copying fidelity, even for incorrect selections. These interventions distinguish task correctness, plan fidelity, and output compliance. Matched dialogue-state evaluations reveal that local field construction largely preserves Jev's correct states but reduces Qwen's mean-turn accuracy, highlighting task- and interface-dependent differences. Sequence-level confidence and request-cost analyses further characterize completed outputs.
Time series (TS) anomaly detection (AD) lacks a universal detector, with the best choice determined by the series rather than by the algorithm. Existing methods therefore either select one detector for each series or average a pool of them. However, both assume that no label ever arrives, while a deployed monitor does receive them, since an operator confirms the alarms it raises. In this paper, we propose ONESHOT, which uses a single confirmed anomaly to weight the pool rather than to choose one detector from it. Specifically, we score each detector on the confirmed anomaly, and those scores become the weights of a soft mixture over the whole pool. We show that a single confirmed anomaly is enough to come close to the best detector for the series. Across 152 univariate and 73 multivariate series, ONESHOT outperforms the strongest single detector, averaging the whole pool, and even picking the best detector on the confirmed event. Code is available at: https://github.com/seunghan96/one_shot.
Text-to-image models have made remarkable progress in visual fidelity and instruction following, yet complex compositional generation remains challenging. Existing agentic approaches address this limitation through iterative generation and editing, adapting their actions based on interaction history. However, they typically fix the continuation state (i.e., the image from which refinement proceeds) to the latest image. Since iterative editing is not necessarily monotonic, the agent may continue refining from the latest image even when it has already degraded. In this work, we introduce GenRevisit, an agentic framework that extends action selection to joint continuation-state and action selection. The agent can not only use interaction history to adjust editing instructions but also select an image candidate to edit, jointly deciding where to continue from and how to refine. We train GenRevisit with SFT followed by agentic reinforcement learning. To address the coarse credit assignment of final outcome reward in GRPO, which assigns the same trajectory-level advantage to all actions, we further propose Normalized Action Credit Assignment (NACA) to redistribute positive advantage according to each action's contribution to reducing the remaining objective. On GenEval2, GenRevisit achieves GM scores of 84.3 and 74.4 with Qwen-Image-2512 and Z-Image-Turbo, surpassing the strongest baselines using the same generators by 12.7 and 10.9 points respectively. GenRevisit achieves state-of-the-art performance on GenEval, GenEval++, and GenEval2 among baselines.
The Model Context Protocol provides a shared interaction substrate for the agentic web, allowing heterogeneous services to expose capabilities through a standardized interface. Learning to use this ecosystem entails a hierarchy of capabilities: tool-format skills, server-level transfer, and cross-server composition. We present MCPDisCo, an ecosystem-scale reinforcement learning framework for MCP-native agents. Using a Thought-Markov Decision Process (Thought-MDP), the framework explicitly represents internal reasoning alongside routing, execution, and termination and trains on real-server, long-horizon interactions spanning 70 MCP servers and 527 tools across six domains. We study the reasoning--tool seesaw, a joint-learning failure mode in which improvement in one objective can accompany deterioration in the other. To mitigate potential interference, MCPDisCo combines token-masked reward attribution with bidirectional gradient surgery: the former separates direct supervision for reasoning and action spans, while the latter removes conflicting components from both objective gradients in shared parameters. MCPDisCo achieves higher success with fewer chain-of-thought tokens. On the out-of-distribution MCP-Atlas and the GAIA benchmark, MCPDisCo maintains monotone gains, confirming that the learned behavior generalizes beyond the training server distribution and broader agentic tasks. Model weight and implementation are available.
长时域智能体必须完成一系列工作流，其中后续活动依赖于前序输出以及共享状态的变化。评估面临的一项关键挑战，是在确保任务可执行且结果可验证的同时，扩展任务要求及其依赖关系。我们提出了 SkillHorizon，这是一个通过技能组合在七种商业环境中构建长时域任务的基准。其生成器利用特定领域规则，将可复用的技能流程转化为具体的商业活动。这些活动通过输入—输出依赖、共享资源和状态更新相互连接，使工作负载与协调需求同步增长。成功的参考执行结果用于验证任务可行性并衡量交互工作负载，而程序化评估器则评估局部完成情况和整体任务成功率。我们在 1,050 个任务上评估了五个语言模型，参考交互工作负载范围从 64K 到 1M 个词元。在这些规模下，平均局部目标完成率从 90.4% 降至 46.8%，因为已完成的工作未能跟上不断增长的要求。在更大的工作负载下，智能体更常提交未成功的最终答案，同时仍有大量工作未完成且保留着工具调用额度。轨迹分析还发现了一些案例：脚本能够成功执行，但其输出未能通过后续的商业检查。SkillHorizon 提供了一个可配置的测试平台，用于研究不断增长的工作流中的持续进展、依赖协调和完整任务交付。我们的代码已发布于 https://github.com/KnowComp/SkillHorizon。
With the rapid evolution of large language models (LLMs), there is an increasing demand for effective personalization techniques that can adapt model behavior to individual user preferences. Personalized LLMs must infer a user's query-dependent preferences from limited, heterogeneous, and evolving interaction histories. Existing approaches rely primarily on in-context personalization, parametric adaptation, or personalized reward modeling, but respectively incur recurring costs in context management, model optimization and deployment, or multi-candidate generation and scoring. These limitations become particularly restrictive when user observations are sparse and the preference evidence relevant to each query varies across contexts and over time. To address these challenges, we propose \modelname, a fine-grained and instance-tailored steering framework that dynamically generates sample-level interference vectors from user data and injects them into the model's forward pass for personalized adaptation. Our approach introduces: match contrast that produces user-relevant signals, fine-grained steering component that captures nuanced signals by hooking activations from attention and MLP layers, and query-conditioned aggregation module that synthesizes these signals into contextually relevant enhancements. Extensive experiments across diverse scenarios, including short-to-long text generation, and web function calling, validate the effectiveness and compatibility of \modelname. Results show that our method consistently brings significant performance gains, and show extreme superiority in fast-shifting and data-sparse environments. Additionally, \modelname rids us of training cycles and incurs light overhead of storage or inference latency. These make \modelname a strong and real-world deployable method. Implementation is available \footnote{https://anonymous.4open.science/r/FleetPilot}.
We present a generator of Thai klon pad verse, one stanza from a short topic, and evaluate it against human verse on held-out topics. The generator is a masked language model initialised from WangchanBERTa over a fixed canvas with two layers: a sound layer of pronounced syllables, where rhyme, tone and syllable count are enforced as hard masks, and a word layer of whole words, so that no word is cut into written pieces. Decoding commits one cell at a time and samples at temperature 0.4; one sample is drawn per topic, with no selection, reranking or repair. On 78 held-out topics never used in development, judged by three independent runs of a frozen LLM judge in the same pools as the human stanzas from which the topics were written, 70.5% of the generated stanzas pass form, lexical cleanness of all four lines and topic alignment together, against 76.9% for the human stanzas (difference -6.4 points, 95% CI [-19.2, +6.4]). Cleanness is level with human (85.9% vs. 87.2%), and the stanzas pass form and cleanness together slightly more often than the human ones (+3.8 points, [-6.4, +14.1]); the difference lies in alignment (84.6% vs. 91.0%), on topics that were written from the human stanzas. The low temperature matters: at higher temperatures the same model passes clearly fewer of the same topics, because lexical errors lie in the model's lower-probability choices; sharpening the sampling distribution removes most of them, at a cost in vocabulary (7.5% rare words against 16.5% in the human stanzas) and diversity. The pre-registered target, all-three within 5 points of human, is missed by 1.4 points. KlonSuphap-LM, the only public klon pad generator we know of, passes none of the 78 held-out topics when judged in separate pools together with the method's and the human stanzas (form 19.2%, cleanness 3.8%).
Generative AI makes options cheap and shifts the bottleneck of creative work from generating options to screening them. We argue that AI's value depends on whether its options arrive before or after the team's screen. In our model, early options compete for a place in the portfolio; late options arrive after that competition and are kept unscreened. In a preregistered experiment, we randomly assigned 148 three-person teams to five schedules of access to the same language model during a product-design task with timed search and selection stages. Moving the same 15 minutes of access from search to selection lowered expert-rated creativity by 0.62 standard deviations, although teams used the model at similar rates. Platform logs locate the difference at the screen. Early access enlarged the pool of ideas fivefold, and teams pruned it. Late-access teams reached the screen with the same small pool as teams without AI and then searched again, devoting 29% of their selection-stage queries to generating new options, compared with 11% for teams with access throughout. Late access added breadth, not quality: these teams submitted 73% more proposals than did teams without AI. Language-model judges detected the benefit of access but not the cost of late timing.
Recent bidirectional encoders have been modernized with innovations from causal large language models (LLMs). However, the pretraining recipe itself remains inherited from small-scale conventions, leaving encoders underoptimized, with no single model performing competitively across both classification and retrieval tasks. In this work, we investigate how far encoder performance can be pushed by revisiting core design choices in training, revealing three key findings: (1) Encoders saturate on downstream tasks beyond a few trillion tokens, indicating that additional compute past this point does not help. (2) Across model sizes, lower masking ratios benefit classification while higher ones benefit retrieval, contradicting prior work suggesting larger ratios for larger models. (3) Distillation with pruning improves classification at low masking ratios but degrades retrieval at high ones. Guided by these findings, we present UltraBERT, a single encoder that assigns separate LM heads to low and high masking ratios within a batch and restricts the distillation objective to the low ratio. To our knowledge, UltraBERT is the first encoder to achieve state of the art on both classification and retrieval tasks, reaching 91.0 on GLUE and 47.7 on BEIR. Together, these results show how far encoder performance can be pushed and provide a principled basis for compute allocation, masking ratio design, and knowledge transfer in bidirectional models.
Beliefs in a Partially Observable Markov Decision Process (POMDP) are posterior distributions over the hidden state given the agent's history of actions and observations. They summarize everything the history says about the state, so planning in a POMDP is planning in a Markov decision process (MDP) over beliefs. Belief MDPs have continuous state spaces even for finite POMDPs, and exact planning in them is intractable in general. Planners must therefore approximate beliefs and estimate values. Monte Carlo planners do both inside a search tree, with an approximate belief at every node and a value estimate at every leaf. Approximations usually consist of sets of sampled states, with leaf values coming from rollouts, hand-designed bounds, or learned networks. Policies and values computed from these approximations are limited by their quality, and prior methods with learned networks rely on hand-chosen summaries of the sampled states. A belief representation should be computable from the parent node's representation through the POMDP's dynamics, let the search sample from it, and be learned so that it provides the information that policies and value functions need. We present Belief Embedding Tree Search (BETS), a Monte Carlo planner that meets these requirements by representing each node's belief with a generative distribution embedding (GDE) while updating it through the POMDP dynamics. BETS trains its belief model alongside its policy and value networks through experience, with the realized hidden state of the POMDP as the belief model's training target. We evaluate BETS on six configurations of RockSample, LightDark, and PocMan against BetaZero and three planners without learned components. BETS is competitive or leading on all six, while every baseline falls far behind on at least one. Where beliefs are highest-dimensional, it outperforms them by a large margin, even without search.
As text-to-speech (TTS) systems continue to advance, provenance tracking and security verification of synthesized speech have become increasingly important. Audio watermarking provides active provenance by embedding imperceptible, machine-recoverable information into speech. However, existing post-hoc methods typically modify the waveform after synthesis, leaving the watermark vulnerable to compound perturbations. To address this limitation, we propose GAMW, a watermark injection framework integrated into the TTS generation process. Before waveform synthesis, GAMW embeds a content-adaptive multi-bit watermark residual into the generated representation, and its decoding objective combines focal modulation with dynamic per-bit weighting. We train and evaluate GAMW separately on two TTS backbones. With a 16-bit payload, GAMW improves decoding accuracy over the strongest baseline by 9.2 and 23.5 percentage points under individual and compound perturbations, respectively, while maintaining competitive speech quality. Reliable clean-speech decoding is also retained at higher payloads. Code is available at https://github.com/ZConJugaTe/GAMW.
Quantization plays a crucial role in efficiently deploying deep learning models on resource-constrained devices. Post-training quantization does not require access to the original dataset or retraining the full model. Current methods that achieve high performance close to baseline results typically require INT8 fixed-point integers. However, achieving higher model compression with lower bit-widths often leads to significant performance degradation. In this paper, we propose VLCQ, a post-training quantization method that relaxes the fixed-point encoding constraint, enabling more flexible exploration of quantization techniques through variable-length encoding. This approach achieves competitive or improved performance while reducing bit-widths, without requiring access to training data or model fine-tuning. Extensive experiments on image classification, segmentation, and object detection demonstrate that VLCQ improves model compression and accuracy compared with state-of-the-art post-training quantization approaches. For per-channel quantization, VLCQ surpasses FP32 accuracy and Piece-Wise Linear Quantization (PWLQ) on most evaluated models, achieving up to 6x compression compared with FP32 and up to 1.7x compared with PWLQ. When prioritizing compression with limited accuracy loss, VLCQ achieves up to 12.25x compression compared with FP32 with less than 4% performance degradation. For per-tensor quantization, VLCQ is competitive with Data-Free Quantization (DFQ), while providing greater flexibility to achieve lower bit rates across different tasks and models.
Hi HN, I'm Niz. Pulse Local is a budgeting app for Mac (Apple Silicon, beta) that answers questions about your spending with models that run on your machine. Local mode needs no account. The ledger and your questions stay on the machine, and the app sends no analytics or crash logs. I felt there are not enough consumer apps that run locally with no calls to the outside world that help us manage sensitive corners of our lives. This app in its local variant only downloads a couple of models and occasional updates, it does not send any analytics nor data. Features-wise, Pulse local is a complete budgeting app that offers an assistant built on top of a fine-tuned model and thousands of rules embedded in the code itself. I'm sharing it for free with options to help me continue experimenting with fine-tuning and expanding along this direction. How it works: you import a CSV from any bank, in any currency. When you ask a question, a tuned Qwen3 0.6B model reads it into a structured query, and code, not the model, does the money math against the ledger. A 4B checker model double-checks each money answer; when the two readings disagree, the app says it is not sure instead of guessing. That change took wrong answers from about one in five to 5 in 494 questions across our last three tests. The write-up covers how we test, the designs that failed, and the limits of the claim: https://pulseos.finance/pulse-local/writing/agree-or-not-sur... The tuned weights are on Hugging Face (PulseOS/pulse-local-qwen3-0.6b). Limitations: the first start downloads about 2.9 GB of models. With the checker on, a money answer takes 12 to 14 seconds (you can turn the checker off). Small models can still misread an open-ended question. The app answers in English. Mac only for now; Windows is next. The app is free. An optional Connected mode (Sync Pass, $29/year) syncs US banks through Plaid; transactions pass through our server's memory and are not stored. Feedback welcome, especially on wrong answers and on questions it should handle but does not. Comments URL: https://news.ycombinator.com/item?id=50080275 Points: 3 # Comments: 0
Showing off Sugabots, an internal project I built at work that we've decided to open-source. I got tired of the team just copying context between agents and Slack. So I built an internal tool for shared chats where our team and agents could work together, using our choice of models and tools. The team is now actively maintaining and building on it to add all the features we're missing. We've shared a hosted offering with a few other teams and has proven useful for them too. Its open-source (MIT) and self-hostable if you want to use it yourself. Let me know what you think. Comments URL: https://news.ycombinator.com/item?id=50079791 Points: 1 # Comments: 0
A service and open source tool to get open data streams in two commands. No sign up needed for subscribing. Streams to try: - https://hypermass.io/stream/_J0fm450s - extreme UV images of the sun from SOHO - visual data - https://hypermass.io/stream/_ATtCoEK - GDACS live disaster data (filtered to just the ongoing items) The CLI (Apache 2.0, https://github.com/hypermass-io/hypermass-cli ) tool allows you to subscribe to multiple live streams of files, with the files arriving in a local folder. The filenames are ULIDs so they can be easily sorted into time order in scripts and code. Publishing works similarly - drop files into a folder and they get read, published and deleted locally. The idea is that by being highly opinionated about the shape of the data (files, delivered in sequence, in real time) a bunch of the logic and edge cases get handled well in one place (things like ordering, retries, hash verification, reconnection, etc) - rather than requiring every subscriber to implement it themselves for every source. A high-speed version for things like departure boards, train positions is in the pipeline. Background: About 10 years ago, I worked for a few years supporting and developing some of the UK Rail Open Data systems. As a developer I was frustrated at how fragmented the open data space was, and how every open data system was designed as a blank-page exercise. I realised that a lot of data falls into an odd middle ground, too infrequent to economically poll on a large scale, too large for messaging protocols, but in some sense still real time. For example transport schedules, periodic solar imagery, disaster alerts - real-time but infrequent. Fairly often I just wanted a "just give me the data" option when the solutions I found were static catalogues with no nod towards real-time use, bespoke APIs, analytics systems and other things I didn't need. I couldn't find a simple approach that leaned into what I saw as the data's strengths (simple files, delivered in order, as/when they are published) - so I built it. About me: Solo founder based in the UK, full time for just over a year. Sustainability: Initially a "heavy users pay for more data allowance" model with a decent free tier. Ideally we later shift to institutional publishers pay to keep their streams free to download. Possibly a paid data marketplace to support costs, but the main focus would always be the free and open data. What I'd like to know: - Does the CLI approach solve any real problems for you? - Is the technical approach sane? - Would you use a service like this to publish data? - Recommendations to bootstrap users (particularly publishers)? Comments URL: https://news.ycombinator.com/item?id=50071200 Points: 5 # Comments: 1
Existing benchmarks measure solve rate and token output/cost, but don't punish excess code, which makes a codebase harder to work with in the long term. I wanted to measure this since it matters most for long-term projects. Method: it takes the patches that benchmarks already publish (DeepSWE, SWE-bench Verified, SWE-bench Live) and counts "coding units": syntax elements like statements, calls, names and literals. I didn't use lines of code, so cramming code isn't rewarded. Deleted code is subtracted, so it measures net units. GPT-6 and Claude 5.5 have no public patches yet, so I ran them myself on 3–10 tasks on my own subscriptions (shown as faded circles, not ranked). Any feedback or model contributions are welcome, especially public patch exports for newer models. GitHub: https://github.com/La5u/Parsimony Not affiliated with any of the benchmarks. Comments URL: https://news.ycombinator.com/item?id=50077619 Points: 2 # Comments: 0
I've used SSH tools for years and didn't really plan on building yet another SSH client, but I ended up building one the way I wanted. Hatoba is a desktop SSH client for Windows, Linux and macOS. It's written in Rust with Tauri 2 and MIT licensed. Each terminal has an AI assistant panel next to it. You describe the problem, and it reads the terminal screen, runs commands, types into the shell, and reads, edits and searches files on the host to find the cause and fix it. You can approve every command and edit it before it runs, let a model you pick approve the safe ones, or turn approval off. Files it changes show a diff you can undo. It works with any OpenAI-compatible or Anthropic API, local models included, and supports skills and MCP servers. Sync doesn't go through a server of mine. Hatoba deploys a Worker and a D1 database to your own Cloudflare account and syncs hosts, keys and AI conversations through them. The free plan usually covers personal use. Everything is encrypted on the device before it leaves. Keys are derived from your master password with Argon2id, every item is encrypted with AES-256-GCM, and the Worker only stores ciphertext plus the hashes it needs to sign you in. Someone who gets hold of the whole Cloudflare account still only gets ciphertext. Sync is optional, and everything works offline without it. The usual SSH client features are there too: groups, tags and fuzzy search, ~/.ssh/config import, ProxyJump and SOCKS5/HTTP proxies, an SFTP panel, local port forwarding, and a key vault that imports OpenSSH, PEM and PuTTY keys. There's also a dashboard showing CPU, memory, disk and network for every saved host, read over plain SSH with nothing installed on the servers. SSH is russh, the terminal is xterm.js with the WebGL renderer, and the UI is React. I develop on Windows, so it's Windows first, with a custom title bar that supports Snap Layouts and Mica, and working CJK input methods. The Linux packages and the macOS build get far less testing on real machines. The macOS build is Apple Silicon only and isn't notarized, so macOS blocks it on first launch until you allow it in Privacy & Security. Bug reports from Linux and Mac users would help the most. Website: https://hatoba.fog.moe Downloads: https://hatoba.fog.moe/download/ Comments URL: https://news.ycombinator.com/item?id=50076388 Points: 2 # Comments: 0
LLM routers can optimize for different factors like price, latency, quality, etc. Enterprises tend to have prepaid AI credits, minimum-spend agreements, cloud commitments, or commercial ceilings across OpenAI, Anthropic, AWS, Azure, and other providers. We’ll call these “AI commits”. This project is designed to consider the state of an enterprises’ current AI commits and use it to inform LLM routing decisions alongside other common variables (price, performance, etc.). This is an early PoC built using LiteLLM. Target audience would CFOs, Strategic Finance, FinOps, Cloud Finance, and AI Platform folks who care about making sure AI commit minimums are met, and maximums are not exceeded. Would love any feedback from folks! Comments URL: https://news.ycombinator.com/item?id=50076439 Points: 1 # Comments: 0
Hacker News 的各位好！我是 Consus 的创始人 Eric。我们将所有 AI 请求路由至符合 ITAR 限制和受控非机密信息（CUI）要求的基础设施。Consus 本身符合 ITAR 限制要求，我们最近还完成了针对 410 项 FedRAMP High 控制措施的 FedRAMP 等效性审计。FedRAMP High 是美国政府针对处理敏感非机密数据的云服务制定的最严格安全基线。此类数据泄露可能造成严重或灾难性后果。在处理 CUI 信息之前，必须取得 FedRAMP 等效性或授权。演示：https://youtu.be/6MYxJe6P4oY 航空航天与国防行业正在快速增长。但遗憾的是，由于必要的数据保护措施和合规风险，大多数公司都难以使用 AI。如果这些公司能够使用大多数其他行业习以为常的 AI 工具，它们本可以发展得更快。 在加入 Consus 之前，我的大部分职业生涯都在航空航天与国防领域。我曾在 SpaceX 负责发射国家安全载荷，也是 Castelion 的早期员工。我亲眼看到，合规方面的繁文缛节如何让公司难以采用新技术，也难以信任那些会接触其敏感数据的供应商。 例如，我曾与一家卫星制造商合作，该公司不允许使用任何 AI 工具，即使是处理非敏感工作也不行。尽管过去 9 到 12 个月里 AI 已经变得极其有用，但令我惊讶的是，仍有这么多航空航天与国防公司拒绝接触 AI。他们看得出 AI 可能会有所帮助，但担心模型提供商使用他们的数据进行训练、违反 ITAR 规定，或者因未通过网络安全评估而失去新业务。这些担忧都是合理的。 解决其中部分或全部问题有许多方法，但我们专注于构建一个合规的 AI 推理层，可以“接入”任何 AI 工具或工作流。正如 OpenRouter 能够与大多数工具集成一样，Consus 可以连接到 ChatGPT 或 Claude 这样的工具、包括 Pi coding agent 或 Open WebUI 在内的开源工具，也可以集成到各个组织自行构建和维护的应用程序中。 几乎没有哪家航空航天与国防公司同时使用三大 GovClouds（AWS、Azure、Google）。当我梳理哪些前沿模型可以合法运行于 ITAR 和 CUI 工作场景时，我发现每个提供商都只提供部分模型，而每个模型达到的合规水平也各不相同。 我们的客户不会从云提供商的角度思考。他们考虑的是所需的数据控制措施，以及想要使用的模型。这就是我们构建合规后缀的原因。在底层，每个请求都会访问一个统一的 OpenAI 兼容端点；在接触推理服务之前，系统会先验证密钥并执行策略。每个模型名称都必须带有合规后缀，例如 gpt-5.6-sol:fedramp-high 或 claude-opus-5-5:itar，任何不带后缀的模型名称都会被拒绝。 我们绝不会代理到标准商业端点。客户还可以构建模型允许列表，并为组织内的单个用户分配支出和模型使用限制。所有这些都通过 Consus Portal 处理，团队可以在那里管理支出、密钥和策略。 我们经常听到的一个问题是：为什么不直接在 AWS GovCloud 或 Azure Government 中使用 Bedrock？对于小型团队而言，仅仅获得这些平台的访问权限就很困难。对于大型组织而言，GovCloud 允许你继承部分、但不是全部处理敏感数据所需的控制措施。此外，模型分散在不同提供商之间：GPT-5.6 Sol 仅在 Azure Government 中提供，而 Claude Opus 5.5 则通过 AWS GovCloud 中的 Bedrock 提供。如果两者都想使用，就需要搭建并保护两个独立的政府云环境。 我们将所有模型置于一个端点之后运行，因此客户无需选择云平台，而合规工作则由我们分别为每个平台处理。 鉴于数据的敏感性，Consus 不会在静态状态下存储任何客户数据。所有提示词和生成结果只会在请求处理期间保留于内存中，绝不会记录或写入磁盘。无论是我们自身，还是基础设施及模型提供商，都严格执行零数据保留政策。出于计费和审计目的，我们会保留每个请求和响应的元数据。 Consus 从设计上就不是自助服务产品。ITAR 要求我们确切了解每个密钥另一端的使用者，因此我们会验证每个组织，并在发放 API 密钥前签订合同。 我们向客户收取合规费用，起价为每席位每月 40 美元，另加按成本收取的 AI 令牌费用。我们目前正在积极服务客户，非常期待获得安全专家，以及在高度监管行业工作的开发者和工程师的反馈！ 参考资料： API 文档：https://gateway.consus.io/overview 安全与架构：https://gateway.consus.io/security 合规详情：https://consus.io/compliance 评论 URL：https://news.ycombinator.com/item?id=50073471 得分：4 评论数：1
If the agent isn't using my diagnostics library, I could introspect and update AGENTS.md, but there's still a delta, and it'll drift. Instead, I write a pre-commit hook that checks the git diff for direct calls to observability modules that bypass my diagnostics module. Boom: fail the commit, inject a prompt telling the agent to use the diagnostics library. All of a sudden it's ten out of ten. — Geoffrey Huntley , nobody should type code by hand anymore. here's what's left. Tags: coding-agents , generative-ai , git , agentic-engineering , ai , geoffrey-huntley , llms
I'm holding AI accountable. For the last few months I've been documenting the errors LLMs have been making in my projects and others - the crimes they have been committing to projects (pun intended). Now I've turned it into a game to see if you would spot the issue. Free to play with keyword matching judgement. Pay is only to add symantic judging. Comments URL: https://news.ycombinator.com/item?id=50070259 Points: 1 # Comments: 0
MIT Technology Review Explains: Let our writers untangle the complex, messy world of science and technology to help you understand what’s coming next. You can read more from the series here. Late last month, Anthropic released a report arguing that the model GLM-5.3 poses a major cyberattack risk. The report’s core concern is that the…
https://gist.github.com/svieira/619d60dea942a8d4498cba3bb6b1... 评论网址：https://news.ycombinator.com/item?id=50060269 得分：105 评论数：46
Good evening folks, I built Github merge gate which that checks whether developers understand their own code when they create a PR. I kept getting sent PRs to review where it was obvious the author had pasted the ticket into an LLM and called it a day. It's self-hosted (so you can use it in your enterprise environments or privately) It is BYOK (Anthropic) obviously. I have two demos in the pull requests here: https://github.com/bhurghundii/vibecheck.dev/pulls - you can see what Vibecheck does (it blocks a PR for clearly showing no attempt vs a PR which gets it right) I would love feedback, especially from anyone dealing with a similar problem. I do use it for my own projects a lot so expect improvements. Comments URL: https://news.ycombinator.com/item?id=50063154 Points: 1 # Comments: 0
“I may have information regarding this case," Claude AI told Philly police.
Python 3.15 is here ! Here's the essential What's new in Python 3.15 guide, this year curated by 3.14 and 3.15 release manager Hugo van Kemenade. As someone who builds open source software that works in every supported Python version, I get to say goodbye to Python 3.10 and start embracing features from Python 3.11. First, some highlights from Python 3.15. Explicit lazy imports I'm most excited about Explicit lazy imports from PEP 810 . These let you import modules like so: lazy import json lazy from pathlib import Path The module will then be imported the first time those symbols are used in the code that follows. My LLM CLI tool implements plugins. I had to recommend manual lazy imports in the plugin author documentation, because simply installing plugins that used heavy dependencies such as sentence_transformers could add multiple seconds to every invocation of the llm command. Lazy imports directly address this startup cost. Unpacking in comprehensions The new [*sublist for sublist in lists] unpacking syntax is neat. Here's an illustration lifted from the documentation : lists = [[ 1 , 2 ], [ 3 , 4 ], [ 5 ]] [ * L for L in lists ] # [1, 2, 3, 4, 5] # equivalent to [x for L in lists for x in L] sets = [{ 1 , 2 }, { 2 , 3 }, { 3 , 4 }] { * s for s in sets } # {1, 2, 3, 4} # equivalent to {x for s in sets for x in s} dicts = [{ 'a' : 1 }, { 'b' : 2 }, { 'a' : 3 }] { ** d for d in dicts } # {'a': 3, 'b': 2} # equivalent to {k: v for d in dicts for k,v in d.items()} PEP 798: Unpacking in Comprehensions by Adam Hartz and Erik Demaine proposed this as an extension of the unpacking syntax from PEP 448 , which landed in Python 3.5 back in 2015. A sampling profiler Python's existing tracing profiler has been renamed from cProfile to profiling.tracing , and has been joined by an exciting new sampling profiler called Tachyon, aka profiling.sampling . Sampling profilers work by periodically capturing stack traces from running processes, then using those to calculate which functions have the most time spent inside of them. They're very low overhead and can provide a deep insight into running systems, without needing to instrument those systems first. I've previously used tools like py-spy for this. Now this capability is baked into the standard library. If you have the PID of an existing Python 3.15 process you can start gathering samples by running this (I had to use sudo on macOS): sudo uv run --python 3.15 python -m profiling.sampling attach 1234 Interact with the main program a bit, then hit Ctrl+C to stop gathering and see a report, which looks something like this: Screenshot of terminal output from a Python sampling profiler with colored text on a black background. Header lines report 2,651 samples captured in 2.65 seconds at a sample rate of 999.98 samples/sec with an error rate of 0.19. A Profile Stats table with columns for nsamples, sample%, tottime, cumul%, cumtime and filename:lineno(function) shows that 99.9% of samples (2.644 seconds) were spent directly in selectors.py KqueueSelector.select, meaning the process was almost entirely idle waiting on the asyncio event loop. The rest of the table lists the call stack beneath it with zero direct time but 100% cumulative presence: asyncio base_events.py run_once, run_forever and run_until_complete, runners.py Runner.run and run, cli.py serve and sqlite_extensions wrapped, and several click core.py invoke, main and call frames. A legend explains each column, followed by a Summary of Interesting Functions section listing hot spots (KqueueSelector.select at 99.9% direct samples), the most frequently present functions on the stack (run_forever, run_until_complete and Runner.run at 100%), and the highest call magnification (2645.0x for BaseEventLoop._run_once). It can also produce flame graphs like this and heat maps like this , and can even output a format that can be viewed in the Firefox profiler UI. 3.15 isn't much faster than 3.14 Miguel Grinberg investigated How fast is Python 3.15? the other day against the 3.15.0rc3 release and found performance to be mostly the same as 3.14, with the exception of the experimental JIT. The JIT gave a 1.2x boost over the standard 3.15 interpreter (on a Fibonacci test), the first time Miguel observed a meaningful JIT performance increase. Miguel's explorations of 3.14 last year found significant boosts against Python 3.13 across many more categories. Upgrading to Python 3.15 with uv I used the following sequence of commands to get uv to use Python 3.15 on my Mac. First I had to upgrade uv itself to ensure it knew about the new release: uv self update Then I installed the stable 3.15 release: uv python install 3.15 And told uv that I wanted that to be the default for every uv run command, unless the project itself pinned a different version: uv python pin --global 3.15 Now the following reports 3.15 as the version: uv run python --version I already have a bunch of tools installed using uv tool install . I reinstalled them to use Python 3.15 like this: uv tool upgrade datasette --python 3.15 Then I can confirm the version like this: datasette --get /-/versions.json | jq Highlights from Python 3.11 The release of 3.15 means that Python 3.10 (released in 2021) is officially no longer supported . My policy for my main open source projects is to keep them working in every supported Python version. I can now drop 3.10 and start using features that were introduced in 3.11. Here's the What's new in Python 3.11 document. I finally get to rely on tomllib being available! Datasette makes extensive use of asyncio . 3.11 adds TaskGroup and timeout() , and several more new features. datetime.UTC is a new in 3.11 alias for datetime.timezone.utc . Python's SQLite support got some significant upgrades in 3.11, including: setlimit() and getlimit() expose a mechanism for setting SQLite runtime limits from Python. create_window_function() lets us create custom window functions in Python, as a class with step() , value() , inverse() , and finalize() methods. blobopen() provides an efficient mechanism for partial reads and writes of large BLOB binary values stored in a SQLite row. You can now call set_authorizer(None) to remove a SQLite authorizer. Datasette had a workaround for this! Plus a whole bunch of improvements to Python's typing system . I plan to land Python 3.11 improvements to a bunch of my projects over the next few weeks. Here's my first Datasette PR removing that set_authorizer(None) workaround. I've also updated my click-app , datasette-plugin , python-lib , and llm-plugin cookiecutter templates to test against Python 3.11 through Python 3.15. Tags: open-source , profiling , python , sqlite , uv
Blueprints AI, a bootstrapped startup based in Santa Monica, developed software that produces code-compliant construction blueprints from a photo, PDF, or sketch.
Microsoft President Brad Smith calls for urgent AI regulation, saying milk and washing machines have more oversight than AI companies.
随着 AI 工厂从兆瓦级发展到吉瓦级，其建设所需已不止于先进的计算技术。开放标准正帮助业界更快速、更高效、更可靠地扩展 AI 基础设施。在 San Jose McEnery Convention Center 举行的 Open Compute Project（OCP）全球峰会上，NVIDIA 宣布了新的合作伙伴关系及贡献，旨在 […]
I didn't like that Claude and codex effectively have the ability to do anything they want on my laptop and my local network - we've all read the horror stories of files accidentally being deleted - and while their permission classifier is getting better, it still requires a lot of trust. So I built Airlock for macOS which runs your agent in a virtual machine, and it will tell you exactly what it's run and what's changed before you bring anything back to your project. You can limit internet access to specific hosts, or block it altogether. It even works with Xcode projects (iOS and macOS developers rejoice). Technically you can run any command with Airlock - not just AI agents (`airlock run -- somecommand`). It's free, requires no account, and has no tracking or analytics. I'm hoping Apple will approve it for Mac App Store distribution, but in the meantime you can download it directly from the website or using Homebrew. Comments URL: https://news.ycombinator.com/item?id=50054620 Points: 1 # Comments: 0
Google Deepmind's AlphaProof Nexus solved dozens of open math problems, including nine Erdős problems two of which were unsolved for 56 years, per a study in Science. The publication comes as OpenAI faces a boycott for dumping AI proofs on GitHub without peer review. Deepmind had its results reviewed before publishing. The article Google Deepmind's AlphaProof Nexus solved dozens of open math problems and actually got peer reviewed appeared first on The Decoder .
I've been running OpenAltFinder for almost 16 months now, and received quite a bit of positive feedback. The directory has also grown substantially. Over the last few months I've been receiving more and more submissions that were AI-assisted/vibe-coded. Personally I'm a bit skeptical (both quality wise, and longevity wise), but I also want to give everyone a chance to showcase what they've build. I had already added an "AI-assisted" label, but right now the majority of submissions include that label. To counter this, I decided to build an AI-free version of it over at: https://noai.openaltfinder.com , inspired by duckduckgo. I had already received some positive feedback on HN in the comments, so I wanted to share this here as well. Do note that not every tool is correctly flagged at the moment, it's still a manual process. Although I'm planning to add it to my pipeline, and auto detect commits by AI agents. Comments URL: https://news.ycombinator.com/item?id=50053462 Points: 2 # Comments: 0
大家好，我做了 whatmodel。这是一个简单的网站，可以在浏览器中访问，也可以通过终端使用：curl whatmodel.sh 使用方法： 1. 输入你的 RAM 和 GPU 2. 查看你可以在本地运行的模型、它们的排名以及运行命令 3. 使用相同的评分将它们与 API 模型进行比较 构建这个项目让我意识到，如今一台 32GB 的 Mac 已经可以完全在本地、免费运行一个得分与 GPT-5 相当的模型（例如 4-bit 的 Qwen 3.8 27B）。 每个模型都会获得一个综合分数，该分数基于 18,731 项已发布的基准测试结果拟合得出，并且每项结果都附有其来源链接。 方法见 https://whatmodel.sh/method，数据、评分代码和基准测试脚本均已开源：https://github.com/jakemanger/whatmodel 如果你愿意贡献，可以使用基准测试脚本收集真实场景下的 tok/s，也可以在 github 上提交任何新模型和基准测试分数，并附上其来源。 评论 URL：https://news.ycombinator.com/item?id=50053608 得分：1 评论数：0
大家好！我是 Saarth。我上九年级，制作了 Atticus。Atticus 是一款速度极快且价格低廉的 Mac 助手，可以帮你完成各种任务。你只需按住 Control-Option-Space，说出或输入任务，无论任务复杂还是只是简单的浏览器搜索，Atticus 都会立即处理。有趣的是，打开应用、播放 Spotify 播放列表或进行网页查询等简单任务不会调用 AI，而是在本地处理，通常不到一秒就能完成。例如，让 Atticus 打开应用大约需要 0.08 秒，有时甚至在你说完句子之前就已完成，有点像 Jev。更复杂的任务会通过 claude -p 发送给 Claude，这意味着你不需要 API 密钥，Atticus 会使用你现有的 Claude Pro 或 Max 方案。采用这种方式也能大幅降低成本，平均每项任务只需不到一美分的一小部分。为了实现 Atticus 的速度，它会记住每项已完成任务的快捷操作。第一次正确完成任务后，它会保存执行步骤，下次无需 AI 就能重放这些步骤。例如，创建新会议这一工作流第一次大约需要 30 秒。随后，Atticus 会将确切步骤记录到记忆中，下一次执行只需 5–8 秒。此外，随着你使用 Atticus 的次数增加，以及通过读取电脑中的信息——包括你的姓名、应用和你希望它记住的内容——Atticus 还会逐渐了解你，从而为你提供个性化的回答。为应对让 AI 不受限制地访问设备所带来的安全问题，Atticus 提供了可切换的限制选项：在删除、发送、购买任何内容或写入文件之前，都会先提示用户确认。我目前正在开发的一些功能包括会议副驾驶，它将利用 Atticus 的速度，让 Atticus 能够监听会议，并在问题出现时实时回答，同时结合会议上下文以及它所了解的你的信息进行个性化处理。要运行 Atticus，需要 macOS 26 或更高版本，因为其语音转录功能依赖该版本。语音转录同样由 macOS 在本地完成。Atticus 免费、开源，而且还非常新。我完全清楚它还有许多不完善之处，因此非常希望听到大家的反馈：哪些地方会出问题，以及你们希望 Atticus 实现哪些功能！评论链接：https://news.ycombinator.com/item?id=50049588 得分：11 评论数：0
arXiv:2609.38263v2 Announce Type: replace-cross Abstract: Feature selection in neural networks remains a challenging problem, particularly in the presence of noisy or contaminated data. LassoNet is a recent approach that addresses this issue by combining neural networks with hierarchical sparsity constraints, enabling simultaneous prediction and variable selection. However, its standard formulation relies on the mean squared error (MSE) loss, which is known to be highly sensitive to outliers. In this paper, we present Robust LassoNet, an extension of LassoNet that incorporates robust loss functions, such as Huber, Cauchy, Tukey's bisquare, and Nonnegative Garrote, to mitigate the effect of extreme observations. The proposed approach preserves the original optimization framework while improving stability under data contamination. Through experiments on synthetic and real datasets, we show that robust LassoNet significantly improves both predictive performance and feature selection accuracy in the presence of outliers or heavy tailed noise, while maintaining comparable performance in clean settings. These results highlight the importance of robustness in neural network based feature selection and suggest practical guidelines for choosing appropriate loss functions.
arXiv:2609.33967v2 Announce Type: replace-cross Abstract: Practical assistive and rehabilitative brain--computer interfaces require subject-independent motor-imagery EEG (MI-EEG) decoders that generalize to new users under limited target-user data and constrained compute. However, held-out-subject performance can be overstated when test-subject information influences preprocessing, model selection, or ensemble selection. We present \textit{ThinkNet}, a validation-controlled framework that combines train-only normalization, validation-guided evolutionary search, and validation-gated inference to identify compact decoders and inference policies for held-out subjects. We evaluate four-class BCI Competition IV-2a (session T) decoding with nine Leave-One-Subject-Out (LOSO) folds, three seeds, seven fixed decoder entries, and a broader search over ten representative decoder families; the held-out subject is never used for normalization, hyperparameter, architecture, or ensemble-policy selection. In the fixed benchmark, the validation-selected compact decoder achieved 44.35$\pm$15.41\% accuracy with 4.9K parameters, 19 KB FP32 weights, and 0.99 ms batch-1 Orin CUDA inference. Across the broader search, compact models ($\leq$25K parameters) achieved higher mean held-out accuracy than mid-size and large alternatives after selected retraining (40.10\% vs. 35.09\% and 34.78\%). Validation-gated ensembling improved over validation-selected single-model inference, reaching 43.98$\pm$16.25\% in the fixed benchmark and 43.31$\pm$15.88\% for the compact six-family ensemble. A non-deployable oracle analysis revealed a 6.1-point family-selection gap and near-zero validation--test correlation, showing that validation reliability remains a key bottleneck under subject shift. Thus, ThinkNet is a validation-controlled framework for compact MI-EEG model and inference-policy selection, rather than a single-architecture benchmark.
arXiv:2606.29315v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly used to act in the real world and augment human decision-making, yet most systems rely on parametric knowledge acquired through imitation, optionally improved by post-training, retrieval, or search over fixed data. This paradigm struggles in novel domains or on complex queries that require information unavailable from prior knowledge alone; knowing the laws of physics, for example, does not by itself enable reliable reasoning or long-horizon actions in a complex physical system. We argue that agents need active experimentation: the ability to gather targeted query-specific evidence, discover general principles of unseen environments, and convert interaction experience into reusable skills. We introduce Hierarchical Experimentalist Agents (HExA), an in-context, experiment-centric self-improvement framework that (1) iteratively designs and refines query-relevant experiments, (2) incrementally distills experience into reusable and composable skills that accelerate experimentation within and across tasks, and (3) integrates experimental evidence to act or answer queries. HExA can work with any LLM including frontier black-box models, and requires no external supervision, oracle solutions, or offline data. We evaluate HExA on multi-step reasoning, interactive partially observable games, and physics law discovery tasks. To further evaluate when active experimentation is necessary, we introduce Interphyre, a set of hard procedural physics tasks where we enable tool-calling and intervention APIs. Current frontier LLMs including GPT-6 Astra fail in these settings: on the most challenging Interphyre tasks, we show that HExA can enable such models to near 100% success. Moreover, HExA improves token efficiency compared to naive agentic and memory baselines and enables generalization via hierarchical skill transfer from easier to harder tasks.
arXiv:2606.19821v2 Announce Type: replace-cross Abstract: Key Performance Measurement (KPM) forecasting is essential for proactive network management of 5G and next-generation telecom networks. However, existing machine learning (ML) approaches face significant limitations in scalability and explainability, restricting their effectiveness in real-world deployments. We propose TelcoAgent, a foundation model-based framework that enables accurate, scalable, and explainable forecasting of multiple KPMs across diverse network cells without the need for site-specific training. Specifically, the framework comprises three key components: (i) an automated three-agent pipeline that constructs a 3rd Generation Partnership Project (3GPP) knowledge graph directly from specification documents, (ii) a scalable, time-series foundation model (TSFM)-based prediction pipeline to deliver accurate, zero-shot forecasting, and finally (iii) a reasoning and explanation pipeline that provides actionable, domain-grounded diagnostics. Evaluated using a 3-month, real-world, city-scale 5G KPM dataset from a U.S.-based network operator, TelcoAgent demonstrates high forecasting accuracy for all 7 considered KPMs per cell across 200 cells, while delivering explainable insights and actionable instructions to address network degradations.
arXiv:2606.19386v2 Announce Type: replace-cross Abstract: Runtime monitors for autonomous agents commonly threshold an accumulated internal state - a behavioural baseline, a drift statistic, or, in our prior work, a modelled affective state. We previously reported a State Saturation Trap: threshold-on-state triggers over a continuous affect engine become near-constant alarms on SWE-bench debugging agents (Modgil 2026). A post-release audit found the engine received dt=0 between actions, so its exponential decay never operated: the published trap is a pure-accumulator result. We correct the record (erratum, v2) and treat the flaw as an experiment. The key variable it exposes is whether a monitor's dynamics are calibrated in sample time (per observation, as in CUSUM) or wall-clock time (half-lives in seconds, as in affect models and EMA baselines). On fixed-rate streams these coincide; on agent streams, where inter-action time varies by orders of magnitude, they do not. A pre-registered sweep over uniform intervals (dt in {0..600}s) on 20 trajectories shows the wall-clock level trigger has two regimes: at dt<=1s a constant alarm (20/20; median 18 firings); at dt>=60s silent. Every critical dt lies in (1,30]s. A scriptable gpt-5.4-mini agent loop measures latency at median 1.53s (p90 2.33s), inside the trap regime, vindicating the empirical finding under a corrected mechanism. The structure is a property of the calibration class, not the engine: a minimal wall-clock accumulator over the raw error stream reproduces the same cliff, while a sample-time CUSUM over the identical stream is exactly dt-invariant (20/20). A rising-edge trigger with hysteresis fires 0-3 times per trajectory in every condition. We conclude that wall-clock-calibrated leaky-integrator monitors admit no regime in which they act as moment detectors on agent streams; transition detection escapes the trap at every cadence, but does not recover human intervention timing.
arXiv:2606.16215v2 公告类型：替换并交叉发布 摘要：多轮工具使用智能体必须在多轮交互中进行推理、调用工具并根据观测结果作出调整。对此类智能体进行后训练颇具挑战：强化学习虽然与仅依赖提示的推理设置相匹配，却常常面临奖励稀疏和信用分配能力不足的问题；而基于专家轨迹的监督微调虽能提供密集的过程监督，却可能使模型过度受限于固定轨迹。为解决这些问题，我们提出 PACT，一种面向多轮工具使用智能体的特权轨迹协同训练框架。其核心思想是仅将专家轨迹用作训练时的优化信号，而非轨迹生成时的提示。PACT 在生成交互轨迹时仍仅使用提示，随后通过两种互补信号利用专家轨迹指导优化：一种是以轨迹为条件的强化学习代理目标，在专家轨迹上下文下评估仅基于提示生成的交互轨迹；另一种是组件感知的监督微调损失，对推理和工具调用组件提供逐步退火的过程监督。为减少对仅在训练时可用的轨迹上下文的过度依赖，PACT 引入了仅提示锚定机制，采用标准的仅提示强化学习更新。在 FTRL、BFCL 和 ToolHop 上的实验表明，PACT 始终优于基于监督微调和强化学习的强基线。进一步分析证实了这两种训练目标的互补性，以及组件感知监督和仅提示锚定机制的益处。这些结果彰显了特权轨迹协同训练在多轮工具使用学习中的有效性。项目页面：https://zhenbangdu.github.io/pact-project-page。
arXiv:2605.11307v2 公告类型：replace-cross 摘要：我们研究视觉语言模型（VLM）能否通过编写可执行代码来生成图像，从而重建结构化的图形类图像。现有图像到代码生成基准通常覆盖的领域有限，依赖参考源代码进行评估，且未提供可量化领域特定问题的指标。为此，我们提出 Vision2Code，这是一个无需参考代码的多领域基准，包含来自 15 个不同数据集的 539 个测试样例，涵盖图表与绘图、几何图形、图结构、科学示意图、文档和三维空间场景。评估时，我们执行生成的程序，并让 VLM 评判模型依据针对各数据集和领域制定的评分准则，将程序的渲染输出与源图像进行比较。我们发现，以这种方式获得的最终分数与人类评分的相关性，显著高于使用通用 VLM 评分准则，或标准的基于像素、基于嵌入及感知相似性指标所获得的分数。我们进一步证明，在对 VLM 评判模型进行重复采样、调整评分准则权重以及引入受控错误的情况下，我们的评估仍然可靠。结果表明，现有图像到代码生成模型在不同领域的表现存在差异，其中在图表和图结构上的准确率显著高于其他领域。最后，我们展示了该评分方法在通过拒绝采样筛选生成程序以用于微调方面的实用价值：与各自的基础模型检查点相比，Qwen3.5-9B 的表现提升了 16%，Phi-4-MM 提升了 28%。我们的代码和数据已公开发布于 https://image2code.github.io/vision2code/。
arXiv:2605.00457v5 Announce Type: replace-cross Abstract: New Radio operating in unlicensed spectrum (NR-U) and Wi-Fi employ different contention parameters and channel-occupancy durations, which can produce severe coexistence imbalance. This correspondence develops a fairness-guided deep Q-network (DQN) that adjusts the NR-U transmission opportunity (TXOP) using the observed per-node throughput ratio. Two reward designs, Q1 and Q2, expose complementary operating regimes. Q1 promotes strict per-node throughput balance, whereas Q2 admits a broader inter-radio-access-technology fairness region and rewards throughput improvement in an intermediate fairness band. The policies are compared with the default Third Generation Partnership Project (3GPP) listen-before-talk (LBT), an adapted multi-armed-bandit method, a utility-fairness-based tabular Q-learning comparator, and a default-referenced fixed-TXOP baseline. Experiments across four priority classes and asymmetric populations report aggregate throughput, user-level Jain fairness, and concave throughput utility over 20 independent trials. The results characterize the cost of strict fairness, efficiency recovery under moderate relaxation, and transfer from symmetric to asymmetric deployments.
arXiv:2602.13619v2 Announce Type: replace-cross Abstract: We study parametric change-point detection under local differential privacy, focusing on the offline localization of a single change point in a sequence of independent observations.~In the non-private setting, we derive improved finite-sample accuracy guarantees for a generalized likelihood-ratio estimator using martingale methods.~We establish finite-sample minimax lower bounds in both the non-private and locally private settings, showing that localization accuracy is fundamentally limited by the Chernoff information between the pre- and post-change distributions.~Guided by this, we design locally private CPD algorithms based on aggregated randomized response and binary mechanisms, which maximize the Chernoff information between the induced output distributions in low- and high-privacy regimes, respectively, and derive their finite-sample guarantees. Together, our upper and lower bounds quantify the statistical cost of privacy in change-point localization.~As part of our analysis, we also show that strong data processing inequality coefficients of R\'enyi divergence and Jeffreys--R\'enyi divergence of order infinity can be computed by restricting to pairs of input distributions with a common support of at most two symbols.~We corroborate our results through empirical evaluation on synthetic data.
arXiv:2602.01149v2 Announce Type: replace-cross Abstract: Machine learning (ML) has emerged as a powerful approach for accelerating the discovery of thermoelectric (TE) materials, yet many models suffer from limited generalizability. Here we present a robust ML framework for predicting the thermoelectric figure of merit (zT) in half-Heusler (hH) compounds, emphasizing prediction of previously unseen compositions. A principal component analysis (PCA)-based train/test splitting strategy is introduced to generate representative partitions of chemical space. Bayesian hyperparameter optimization is combined with correlation-based k-best feature selection for Random Forest, XGBoost, and Neural Network models, together with the interpretable SISSO symbolic-regression method. Relative to the full set of 118 input features, the optimized feature subsets reduce the test RMSE by 16-27%. SHAP and SISSO analyses identify temperature, A-site dopant concentration, and A-site heat of vaporization as the most important descriptors of zT. The framework is validated through prediction of previously unexplored compositions, recovery of literature-reported optimal zT values for established hH systems, and first-principles calculations for the new compounds CeCoBi and CeIrBi, which exhibit favorable electronic transport properties. The framework is further applied to high-throughput screening of the doped hH chemical space, covering about 2.46 million elemental substitution combinations across 631 thermodynamically screened parent matrices under stability constraints. This screening yields a concise set of promising candidates for future study. Overall, the work establishes a generalizable workflow for reliable data-driven thermoelectric materials discovery.
arXiv:2512.02444v2 Announce Type: replace-cross Abstract: Joining heterogeneous tables requires different forms of correspondence, including exact agreement, syntactic transformation, semantic similarity, and entity-level matching. The appropriate join method is often unknown in advance, while running every candidate method incurs unnecessary cost. We introduce MetaJoin, a label-free framework for adaptive join selection. MetaJoin uses input evidence to choose an initial specialist, then uses specialist-specific output evidence to determine whether that result should be retained, compared with an alternative, escalated, or repaired. This allows join selection to adapt after execution rather than committing to a one-shot routing decision. We also introduce MNN Join, a transformation specialist that searches for string transformations by maximizing mutual-nearest-neighbor coverage between two unaligned columns and uses the same reciprocal structure to validate its output. We evaluate MetaJoin on 670 datasets from six benchmarks. It achieves 0.817 macro row-F1, slightly exceeding a benchmark-level oracle that is given the best standalone specialist for each benchmark (0.812), and outperforms label-free selection policies built from the same specialist portfolio by more than 0.13. Across the 609 datasets in the four transformation-dominant benchmarks, MNN Join outperforms every evaluated transformation baseline. Finally, MetaJoin's accuracy varies by only 0.015 as the controller backbone ranges from 3B to 14B parameters, showing that output validation and repair substantially reduce dependence on the initial routing decision.
arXiv:2511.10383v2 Announce Type: replace-cross Abstract: Continuous-time stochastic processes underlie many natural and engineered systems. In healthcare, autonomous driving, and industrial control, direct interaction with the environment is often unsafe or impractical, motivating offline reinforcement learning from historical data. However, there is limited statistical understanding of the approximation errors inherent in learning policies from offline datasets. We address this by linking reinforcement learning to the Hamilton-Jacobi-Bellman equation and proposing an operator-theoretic algorithm based on a simple dynamic programming recursion. Specifically, we represent our world model in terms of the infinitesimal generator of controlled diffusion processes learned in a reproducing kernel Hilbert space. By integrating statistical learning methods and operator theory, we establish global convergence of the value function under sufficiently small generator-approximation error and derive explicit finite-sample guarantees with bounds tied directly to system properties such as smoothness and stability. Our theoretical and proof-of-concept results indicate that operator-based approaches may hold promise for solving offline reinforcement learning using continuous-time optimal control.
arXiv:2510.07832v2 Announce Type: replace-cross Abstract: Spatial prediction refers to the estimation of unobserved values from spatially distributed observations. Although recent advances have improved the capacity to model diverse observation types, adoption in practice remains limited in industries that demand interpretability. To mitigate this gap, surrogate models that explain black-box predictors provide a promising path toward interpretable decision making. In this study, we propose a graph partitioning problem to construct spatial segments that minimize the sum of within-segment variances of individual predictions. The assignment of data points to segments can be formulated as a mixed-integer quadratic programming problem. While this formulation potentially enables the identification of exact segments, its computational complexity becomes prohibitive as the number of data points increases. Motivated by this challenge, we develop an approximation scheme that leverages the structural properties of graph partitioning. Experimental results demonstrate the computational efficiency of this approximation in identifying spatial segments.
arXiv:2506.02955v3 公告类型：替换并交叉列出 摘要：生成模型已成为机器人运动生成领域日益强大的工具，能够在各种任务中灵活生成多模态轨迹。然而，大多数现有方法在处理多种类型的约束时仍存在局限，例如避碰、驱动限制和动力学一致性；这些约束通常被单独处理，或通过启发式方法处理。在本研究中，我们提出 UniConFlow，这是一种基于统一约束流匹配的轨迹生成框架，可系统地纳入等式和不等式约束。此外，UniConFlow 引入了一种新颖的预设时间归零函数，在推理过程中塑造时变引导场，使生成过程能够适应不同的系统模型和任务要求。为进一步应对长时域、高维轨迹生成带来的计算挑战，我们还针对终端约束施加和推理过程提出了两种实用策略：一种违规片段提取方案，可精确定位并仅优化轨迹中违反约束的部分；以及一种轨迹压缩方法，可在降维空间中加速优化，同时保持解码后重建的高保真度。我们在双倒立摆、从现实到仿真的赛车任务以及从仿真到现实的操作任务这三项实验中进行了实证验证，结果表明，UniConFlow 优于最先进的生成式规划器和传统优化基线，在安全性、运动学与动力学一致性以及动作可行性等可认证运动规划指标上取得了更优的表现。项目页面：https://uniconflow.github.io。
arXiv:2505.22193v2 Announce Type: replace-cross Abstract: Quantum Diffusion Models (QDMs) are an emerging paradigm in Generative AI that aims to use quantum properties to improve the performances of their classical counterparts. However, existing algorithms are not easily scalable due to the limitations of near-term quantum devices. Following our previous work on QDMs, here we propose and implement two physics-inspired protocols. In the first, we use the formalism of quantum stochastic walks, showing that a specific interplay of quantum and classical dynamics in the forward process produces statistically more robust models generating sets of MNIST images with lower Fr\'echet Inception Distance (FID) than using totally classical dynamics. In the second approach, we realize an algorithm to generate images by exploiting the intrinsic noise of real IBM quantum hardware with only four qubits. Our work could be a starting point to pave the way for new scenarios for large-scale algorithms in quantum Generative AI, where quantum noise is neither mitigated nor corrected, but instead exploited as a useful resource.
arXiv:2610.05745v2 公告类型：替换 摘要：已有研究表明，训练后量化能够在标准评估条件下保持 VLA 的性能，但其能否保留全精度模型的鲁棒性和泛化能力，仍缺乏充分探索。本研究系统考察了训练后量化的 VLA 策略在环境扰动下的鲁棒性和泛化能力。实证结果表明，量化策略尽管保持了相近的分布内性能，却可能对细微的环境变化变得脆弱。我们进一步观察到，动作差异集中在策略执行轨迹中的一小部分状态上，而教师指导的效果随差异程度而呈现相反趋势：它能提升高差异状态下的泛化能力，却可能削弱低差异状态下的泛化能力。这些发现揭示，有效的量化后恢复需要对脆弱状态进行选择性干预，而不是对学生模型进行全局蒸馏。因此，我们提出 Policy-Induced Vulnerability-Oriented Tuning（PIVOT-Q），这是一个具有脆弱性感知能力的同策略蒸馏（OPD）框架，以冻结的全精度策略作为教师，选择性地纠正量化学生模型在执行轨迹中遇到的脆弱状态。PIVOT-Q 利用短时域内的折扣累积差异识别脆弱状态，实施阶段均衡的稀疏监督，并使用 Behavioral Anchor 防止不必要的改变。在 LIBERO-Plus 的七种环境变化下开展的实验表明，该方法在多种 VLA 骨干模型和量化方法上均能实现稳定的性能恢复。值得注意的是，PIVOT-Q 在所有设置下都持续优于全状态蒸馏，而所用的状态级蒸馏预算仅为后者的 7.4%。我们的代码已发布于 https://github.com/ruanruan-andy/PIVOT-Q。
arXiv:2609.35545v2 Announce Type: replace Abstract: Epidemic policy planning often requires coordination between geographical regions, taking into account mobility-driven spillovers and how to make use of limited resources. Existing methods either lack action-conditioned models of coupled dynamics or cannot guarantee per-period feasibility. We present EpiMind, a graph world model framework for constrained epidemic policy planning across regions. A graph-factored recurrent state-space model generates joint policy-conditioned rollouts from regional latent beliefs, while graph-temporal ADMM optimizes regional interventions, enforces shared-resource feasibility through projection, and evaluates temporal specifications under the learned model. EpiMind reduces admission RMSE by 29% relative to graph-free dynamics modeling, plans within 1-5% of the best feasible constant policy with guaranteed shared-budget feasibility, and outperforms all deployable baselines across three resource budgets in real-context evaluation. These results demonstrate that graph-structured policy imagination with explicit constrained coordination supports effective epidemic interventions from learned dynamics.
arXiv:2609.33521v2 Announce Type: replace Abstract: Trust evaluation plays a critical role in cybersecurity by supporting risk mitigation and decision-making. A variety of trust evaluation methods have been proposed, with learning-based approaches offering high accuracy and automation. However, they often require substantial ground truth, suffer from low training efficiency, lack support for basic trust properties, and provide limited explainability. Large Language Models (LLMs) offer a compelling alternative due to their strong zero-/few-shot reasoning abilities and broad knowledge. To this end, we propose LLM4Trust, the first benchmark framework that systematically explores the capabilities of LLMs for trust evaluation. We first construct diverse trust graphs to model five basic trust properties and design corresponding property understanding tasks. We then assess the ability of eight representative LLMs to understand these properties under nine prompt methods. Based on this exploration, we identify the most effective LLM-prompt combinations and apply them to five real-world datasets for validating LLMs' trust evaluation capability. During this process, we propose two strategies to extract key information from large-scale trust graphs, addressing the context window limitations of LLMs. Extensive experiments show that LLMs can effectively understand basic trust properties and have great potential for real-world trust evaluation, particularly under limited supervision. However, they remain vulnerable to attacks targeting trust graphs and demonstration examples used in few-shot prompting, and incur high inference costs. Accordingly, we propose a defense mechanism and batch inference to improve the robustness and efficiency of LLM-based trust evaluation. The source code of LLM4Trust is available at https://github.com/Jieerbobo/LLM4Trust
arXiv:2609.31306v3 公告类型：替换 摘要：TabPFN、Mitra 或 ConTextTab 等表格上下文学习器依赖于在潜在嵌入的二维序列上交替进行行注意力和列注意力计算。这些注意力模式与语言模型中的一维情形显著不同：行注意力涉及较长的序列，而列注意力处理的序列短得多，且表格数据的跨步内存布局使得生成连续张量的成本很高。此外，当前模型使用的隐藏维度相较于近期的语言模型较小。然而，高效注意力的研究主要集中于一维序列，二维表格场景尚未得到探索。为此，我们构建了一套可复现的基准测试环境，并在多个后端上研究表格注意力的独有特征——包括 Torch SDPA（efficient 和 cuDNN）、FlashAttention-2/3/4，以及仅用于推理的后端 vLLM 和 SageAttention——在三代 GPU（A100、H100、B200）上，针对符合实际的表格张量形状测量前向和反向计算的吞吐量。我们发现，列注意力与行注意力的最优后端选择不同，并且会随硬件和模型具体配置而变化：尽管针对各代 GPU 定制的 FlashAttention 实现总体表现最佳，但在序列较长的列注意力场景中，CuDNN 有时表现更优，性能交叉点取决于注意力头维度。在仅用于推理的后端中，SageAttention 在行注意力以及超过 16,000 行的大规模序列上表现良好。我们可复现的基准测试为未来改进表格原生注意力机制奠定了基础。自包含的基准测试与评估代码已公开发布于：https://github.com/SAP-samples/tabular-attention-benchmark
arXiv:2609.08642v2 公告类型：替换 摘要：强化学习（RL）中的探索仍是一项根本性挑战。近期的目标条件化 RL 策略（通过选择目标来促进更广泛的状态覆盖）已展现出良好效果，但尚无策略能同时依据新颖性和可达性对目标进行评分：这两种信号要么被人为权衡，要么被依次应用，要么其中一种被完全忽略。本文提出一种可达性感知的目标选择框架，显式整合这两个方面，并可无缝融入任何离策略 RL 算法。为此，我们提出 SUccessor-to-Novelty（SUN），这是一种从后继价值函数推导出的指标，用于识别既新颖又可达的目标。我们证明，SUN 是一种按计数缩放的目标达成奖励所对应的价值，可为短时域内的命中概率提供界限，并能排除不可达目标。我们进一步提出一种利用这些性质的自适应目标选择策略，以及一种准确且轻量级的伪计数方法，以避免经典方法的开销。我们通过全面的基准测试验证了所有论断：在包含不可达或难以到达的状态、不可逆转移、障碍物、迷宫及无界空间的标准环境和新环境中，SUN 均持续优于当前最先进的方法。代码见 https://github.com/sparisi/rl-anvil
arXiv:2606.31653v2 公告类型：替换 摘要：可认证训练旨在生成其预测能够被形式化验证为可抵御对抗扰动的模型，通常通过优化允许扰动集合上最坏情况损失的上界来实现。对于神经网络，完全基于紧松弛界的可认证训练方法能够生成易于认证的网络，但会牺牲标准准确率。相反，对抗训练往往能带来更强的经验鲁棒性和更高的标准准确率，但所得模型通常难以用神经网络验证器进行认证。近期文献表明，将对抗训练目标与基于区间界传播（IBP）的宽松过近似相结合，可以在标准准确率与可认证准确率之间取得更好的权衡，其实质是在最坏情况损失的下界与上界之间进行插值。在此基础上，我们提出 AD-CERT，这是一种将对抗蒸馏与 IBP 上界相结合的可认证训练目标。我们表明，在 logit 空间中从具有经验鲁棒性的教师模型蒸馏对抗信息，可以为可认证训练提供有效的下界替代目标；AD-CERT 在多个面向 ReLU 网络的鲁棒性基准上取得了最先进的可认证性能。此外，在统一的实验设置下，研究表明，与鲁棒特征空间蒸馏目标相比，在 logit 层面蒸馏对抗信息可将可认证准确率提高最多 5.40 个百分点。我们的代码已发布于 https://github.com/matteo-melis/ad-cert。
arXiv:2606.30376v2 公告类型：替换 摘要：通过在线强化学习对齐连续空间中的生成式流模型，受到轨迹似然难以计算的限制。现有基于密度近似的策略梯度方法依赖随机 SDE 采样器来构建可计算的转移核，这会引入训练与推理之间的不一致，并且需要无分类器引导（CFG）。尽管 DiffusionNFT 等隐式框架直接优化前向过程的速度场，但其启发式固定幅度校正使优化强度无法随组内相对质量而调整。我们提出 \textit{Flow Advantage-Weighted Rectification}（\textbf{FlowAWR}），这一范式将连续生成策略优化重新表述为朝向理论最优速度场的监督回归。从 KL 约束下奖励最大化的最优策略出发，FlowAWR 推导出具有幅度感知、优势加权校正形式的最优速度场，实现了无需 SDE 的优化和无需 CFG 的生成。在 SD3.5-Medium 上的对比评估中，FlowAWR 在提升对齐性能的同时，相较于 DiffusionNFT 实现了 2$\times$ 至 5$\times$ 的收敛加速（例如，FlowAWR 在 1.2k 步时达到 24.12 的 PickScore，而 DiffusionNFT 在 2.0k 步时达到 23.82，FlowGRPO 在 $>$4k 步时达到 23.50）。在多奖励约束下，FlowAWR 能够维持生成质量，在满足结构规则的同时保持稳定的域外性能。代码见 \href{https://github.com/Fuzm7/FlowAWR}{FlowAWR}。
arXiv:2606.21386v2 公告类型：替换 摘要：视觉-语言-动作模型（VLA）在许多机器人操作任务上达到了最先进的性能，但在分布外场景中仍可能表现出不可预测的行为。因此，运行时失败检测对于 VLA 在现实世界中的安全部署至关重要。然而，现有的任务失败检测器要么需要计算成本高昂的动作采样，要么基于限制其对 VLA 适用性的架构假设，要么需要获取失败轨迹。我们提出 VLA-FAIL，这是一种轻量且广泛适用的 VLA 失败检测框架，以极低的额外开销结合了两种新颖的失败检测器，且无需失败数据。第一种是末层马氏距离（LLMD），通过逐词元测量末层特征相对于训练数据的偏离程度来检测分布外状态。第二种是动作块一致性（ACC），利用滚动时域控制产生的时间重叠，在连续动作块变得不一致时检测失败。为刻画检测准确性与检测延迟之间的权衡，我们引入了 \textsc{AUCPDT}，这是一种不依赖阈值的指标，可联合评估精确率、召回率和检测时间。通过大量现实世界和仿真实验，我们证明 LLMD 和 ACC 能够捕捉互补的失败模式，两者结合可在多种任务中实现可靠且及时的失败检测，并且经常优于计算成本高得多的基线方法。项目页面：https://vla-fail.github.io/vla-fail-2026
arXiv:2605.11471v2 Announce Type: replace Abstract: Matrix product operator Born machines (MPO-BMs) are tractable tensor-network models for probabilistic modeling, yet their approximation capabilities remain poorly understood in theory. In this paper, we characterize these capabilities from both negative and positive perspectives. As a negative result, we prove that constructing a polynomial-bond MPO-BM approximation to an arbitrary continuous target with bounded KL error is NP-hard. On the positive side, for score-based variational inference, we prove that targets satisfying a log-Sobolev inequality admit polynomial-bond MPO-BM approximations with provable KL guarantees when the loss-induced Hamiltonian is local and has a constant spectral gap. Furthermore, under the same conditions, we construct local importance-sampling estimators and prove that polynomially many local score queries suffice to preserve the KL guarantees.
arXiv:2605.08098v2 Announce Type: replace Abstract: Kirigami is an increasingly useful fabrication method to produce shape-programmable metamaterial structures. However, inverse design remains difficult because deployment is nonlinear, and feasible cut layouts must satisfy discrete compatibility rules, avoid overlap, and map one target shape to valid designs. We present RL-Kirigami, an inverse design framework that combines optimal-transport conditional flow matching (OT-CFM) with reinforcement learning to generate compatible ratio fields for compact reconfigurable parallelogram quad kirigami. A marching decoder enforces global geometric compatibility, and Group Relative Policy Optimization (GRPO) aligns the generator with nondifferentiable rewards for silhouette matching, feasibility, and ratio-field regularity. Across procedurally generated target shape instances, a single sample from the pretrained OT-CFM prior reached $94.2\%~\mathrm{sIoU}$ and outperformed solver baselines while reducing forward simulator evaluations from hundreds to $1$. GRPO improved accuracy to $94.91\%~\mathrm{sIoU}$ and, with regularity included, reduced $\mathrm{TV}(\mathbf{x})$ from $0.95$ to $0.81$ while maintaining $94.83\%~\mathrm{sIoU}$. Generated layouts were exported to DXF and laser cut in $50~\mu\mathrm{m}$ polymeric sheets to produce deployable prototypes in $8.0\pm 1.0$ minutes per part. These results support a manufacturing-aware inverse design workflow for deployable kirigami metamaterials under geometric feasibility constraints. Code is available at https://github.com/milad1378yz/RL-Kirigami.
arXiv:2604.23418v2 Announce Type: replace Abstract: Uniform random rotations are useful in fast Johnson--Lindenstrauss embeddings, kernel approximation, communication-efficient learning, and recent AI compression pipelines, but are computationally expensive to generate and apply in high dimensions. A practical alternative repeats Walsh--Hadamard transforms with random sign diagonals. Applying this structured rotation twice has shown empirical benefits, but supporting theory remains limited. We study the approximation quality of the two-stage structured Hadamard rotation. Here $d$ is the dimension of the rotated vector. On the positive side, we prove explicit, nonasymptotic upper bounds of order $\sqrt{\log d/d}$ in both Kolmogorov and Wasserstein-1 distance for corresponding coordinates scaled by $\sqrt d$, uniformly over unit inputs and coordinates. For the input $u=e_1$, both distances also have lower bounds of order $d^{-1/2}$ at every coordinate. On the negative side, we prove an explicit lower bound on the Wasserstein-1 distance between the full vector distributions, showing a substantial discrepancy that does not vanish as $d$ grows for the input $u=e_1$. For the same input, we also prove a matching asymptotic upper bound, showing that the lower-bound scale is sharp. Taken together, the results identify a separation between one-dimensional marginal behavior, where approximation improves with dimension, and the full-vector law, where a nonvanishing Wasserstein-1 discrepancy remains. The marginal bounds control coordinate threshold probabilities and expected scalar Lipschitz losses, including expected averages over coordinates. They do not establish joint-coordinate approximation, and the full-vector separation does not imply failure of a particular algorithm. Connecting these distributional results to performance guarantees for specific algorithms remains open.
arXiv:2604.01730v3 Announce Type: replace Abstract: This paper investigates Koopman operator-based approaches for multivariable control of a two-spool turbofan engine. A physics-based component-level model is developed to generate training data and validate the controllers. A meta-heuristic extended dynamic mode decomposition is adapted, with a cost function designed to accurately capture both spool-speed dynamics and the engine pressure ratio (EPR), enabling the construction of a single Koopman model that can be reused across multiple control strategies. Using the identified time-varying Koopman model, an adaptive Koopman-based model predictive controller (AKMPC) with a disturbance observer is developed and compared with a Koopman-based feedback linearization controller (K-FBLC) and its integrator-augmented version (K-FBLC-I). The Koopman representation further enables nonlinear GTE output limiters, such as rotor-acceleration and turbine-inlet-temperature limits, to be expressed as linear constraints in the AKMPC. The controllers are evaluated for two control configurations of spool speeds and EPR, under both sea-level and varying flight conditions. The results demonstrate that the proposed identification approach enables accurate predictions of both spool speeds and EPR, allowing the Koopman model to be reused flexibly across different control formulations. While all strategies achieve comparable performance in sea-level conditions, the AKMPC demonstrates improved performance under varying flight conditions due to its ability to capture nonlinear dynamics, handle constraints, and compensate for model mismatch. Moreover, the EPR control strategy improves the thrust response. The study highlights the applicability of Koopman-based control and the advantages of the AKMPC framework for robust turbofan engine control.
arXiv:2603.00530v2 Announce Type: replace Abstract: Sampling from unnormalized densities using diffusion models has emerged as a powerful paradigm. However, while recent approaches that use least-squares `matching' objectives have improved scalability, they often necessitate significant trade-offs, such as restricting prior distributions or relying on unstable optimization schemes. By generalizing these methods as special forms of fixed-point iterations rooted in Nelson's relation, we develop a new method that addresses these limitations, called Bridge Matching Sampler (BMS). Our approach enables learning a stochastic transport map between arbitrary prior and target distributions with a single, scalable, and stable objective. Furthermore, we introduce a damped variant of this iteration that incorporates a regularization term to mitigate mode collapse and further stabilize training. Empirically, we demonstrate that our method enables sampling at unprecedented scales while preserving mode diversity, achieving state-of-the-art results on complex synthetic densities and high-dimensional molecular benchmarks.
arXiv:2602.23811v5 Announce Type: replace Abstract: We investigate the theoretical aspects of offline reinforcement learning (RL) under general function approximation. While prior works (e.g., Xie et al., 2021) have established the theoretical foundations of learning a good policy from offline data via pessimism, existing algorithms that are computationally tractable (often in an oracle-efficient sense), such as PSPI, only apply to finite and small action spaces. Moreover, these algorithms rely on state-wise mirror descent and require actors to be implicitly induced from the critic functions, failing to accommodate standalone policy parameterization which is ubiquitous in practice. In this work, we address these limitations and extend the theoretical guarantees to parameterized policy classes over large or continuous action spaces. When extending mirror descent to parameterized policies, we identify contextual coupling as the core difficulty, and show how connecting mirror descent to natural policy gradient leads to novel analyses, guarantees, and algorithmic insights, including a surprising unification between offline RL and imitation learning.
arXiv:2602.13485v3 Announce Type: replace Abstract: Networks of modern industrial systems are increasingly monitored by distributed sensors, where each system comprises multiple subsystems that generate high-dimensional time-series. These subsystems are often interdependent, making it important to understand how temporal patterns observed at one subsystem relate to those observed across others. This problem is particularly challenging in decentralized settings where raw measurements cannot be shared, and client observations are heterogeneous. Moreover, in practice, each subsystem (client) operates a fixed, proprietary model that cannot be modified or retrained, limiting the applicability of existing approaches. Separately, nonlinear dynamics make cross-client temporal interdependencies difficult to characterize, as they are embedded implicitly within nonlinear state-transition functions. This paper presents a vertical federated learning framework for learning one-step-ahead cross-client temporal interdependencies under these constraints. Each client maps high-dimensional local observations to low-dimensional latent states using a nonlinear state-space model, while a Graph Attention Network (GAT)-based central server learns a state-transition model over the communicated states. Given a candidate graph skeleton, the server learns state-dependent attention coefficients that capture cross-client interactions. We analytically relate the GAT attention coefficients to the Jacobian of the learned state-transition model, yielding a first-order characterization of cross-client interdependencies. We further establish bounded closeness of the learned dynamics and Jacobians to a corresponding centralized oracle. Synthetic experiments validate the theoretical results on controlled systems with known graph skeletons. Experiments on real-world datasets with unknown graph skeletons demonstrate superior federated root cause analysis over baselines.
arXiv:2602.03806v2 公告类型：替换 摘要：训练用于多轮代码生成的大语言模型（LLMs）通常依赖在线强化学习（RL），模型通过反复生成的在线采样轨迹进行学习。尽管有效，在线 RL 的成本可能很高，且训练可能不稳定。本文提出利用离线轨迹的上下文老虎机学习（COBALT），这是一种简单的方法，通过复用离线多轮轨迹，高效训练用于多轮代码生成的 LLMs。基于多轮代码生成是一个单步可恢复的马尔可夫决策过程这一观察，COBALT 首先使用参考 LLM 收集代码生成轨迹，并将其拆分为部分轨迹。随后，在在线老虎机学习过程中，通过单步代码生成训练 LLM 补全每条部分轨迹。在 LiveCodeBench 上，COBALT 显著提升了 R1-Distill 8B 和 Qwen3 8B 的表现，使其 Pass@1 分别最多提高 9.0 和 6.2 个绝对百分点，并达到了采用 GRPO 和 VeRPO 的两种多轮在线 RL 基线的性能。我们还对上下文内奖励投机进行了基于扰动的分析，结果表明 Cobalt 能够增强模型对误导性执行反馈的鲁棒性。总体而言，我们的结果表明，Cobalt 是一种很有前景的多轮代码生成解决方案。我们的代码和数据已公开，地址为 https://github.com/OSU-NLP-Group/cobalt。
arXiv:2510.22809v2 Announce Type: replace Abstract: Traditional machine learning relies on explicit models and domain assumptions, often limiting flexibility and interpretability. We introduce a "model-free" framework of instance-based learning using surprisal (information theoretic uncertainty) to directly analyze and perform inferences from raw data, eliminating distribution modeling, reducing bias, and enabling efficient updates including direct edits and deletion of training data. By quantifying relevance through uncertainty, the approach enables generalizable inference across tasks including generative inference, causal discovery, anomaly detection, and time series forecasting. Our approach emphasizes traceability, interpretability, and data-driven decision making, offering a unified, human-understandable framework for machine learning, can be adapted to virtually all machine learning tasks, achieving competency across all and state-of-the-art performance on anomaly detection and some supervised tasks. The mathematical foundations create a "physics" of information, which enable these techniques to apply effectively to a wide variety of complex data types, including missing data. Our results suggest that this may be a viable foundation upon which to build scalable machine learning and artificial intelligence that can maintain human understandability of the underlying mechanics.
arXiv:2411.02817v3 公告类型：替换 摘要：文本提示引导的生成模型在保真度和提示对齐方面得到了广泛评估，但其生成多样化输出的能力仍缺乏充分研究。现有的多样性指标（如 Vendi 和 RKE）基于核矩阵的冯·诺依曼熵和 Rényi 熵，是为无条件模型开发的，无法区分由提示引起的变异与由模型引起的变异。为弥补这一不足，我们提出了 \textit{Conditional-Vendi} 和 \textit{Conditional-RKE}，这两种多样性度量由半正定矩阵的条件熵推导而来。这些评分能够分离出提示引导生成过程中由模型引起的多样性，其中 Conditional-RKE 具有 $O(1/\sqrt{n})$ 的收敛速率。对于 Conditional-Vendi，我们引入了一种截断谱近似方法，可提供具有可扩展性和一致性的估计。在文本生成图像、图像描述生成和 LLM 任务上的实验表明，这些条件评分能够恢复真实的多样性排序，还能引导扩散模型生成更多样化的样本。代码库见 https://github.com/mjalali/conditional-vendi。
arXiv:2406.07177v2 公告类型：替换 摘要：大语言模型（LLMs）在自然语言处理（NLP）任务中取得了卓越的表现，但高昂的计算成本和内存需求限制了其应用。三值化是一种极端的量化方式，通过减少内存使用并支持高能效的浮点加法运算，为这一问题提供了解决方案。然而，将三值化应用于大语言模型时，权重和激活值中的离群值会带来挑战。在本研究中，我们观察到权重中的离群值呈非对称分布，且权重均值不为零，因此引入了双重可学习三值化（DLT），使缩放因子和平移量均可学习。我们还提出了离群值友好的特征知识蒸馏（OFF），以恢复极低比特量化中丢失的信息。所提出的 OFF 能够融入语义信息，且对离群值不敏感。OFF 的核心是利用余弦相似度，最大化三值化模型与浮点模型特征之间的互信息。大量实验表明，在不同大语言模型系列的标准文本生成和零样本基准测试中，我们的 TernaryLLM 均优于以往的低比特量化方法。具体而言，对于最强大的开源模型之一 LLaMA-3，我们的方法（W1.58A16）相较于此前最先进的方法（W2A16），在 C4 上的困惑度降低了 5.8，在零样本任务上的平均准确率提高了 8.2%。
arXiv:2610.13931v1 Announce Type: cross Abstract: The expressivity of a neural network $\Phi$ describes the class of functions that a network can express. The term topological expressivity refers to a family of measures particularly suited to understand binary classification and is often formulated in terms of the Betti numbers of a decision boundary $\beta (\Phi^{-1}(0))$. Despite its popularity, this measure is not as useful for regression and other problems that do not involve a decision boundary. Instead, we propose to study topological expressivity in terms of the Betti numbers of the output of a neural network $\beta(\Phi(X))$. We formally prove that a ReLU network with architecture $(d,\sum\limits_{k=1}^{d } \beta_k(3k+2),\sum\limits_{k=1}^{d}\beta_k(k+1))$ can express spaces homotopy equivalent to wedge sums of spheres with arbitrary Betti numbers. This gives upper bounds for any arbitrary connected ($\beta_0=1$) sequence of Betti numbers, but by constructing spaces homotopy equivalent to products of $N$ spheres of dimension $d_1,...,d_N$ we achieve much better bounds $(\sum\limits_{n=1} d_n, 3\sum\limits_{n=1}\limits^N d_n + 2N, \sum\limits_{n=1}\limits^N d_n + N) $ for Betti numbers given by $\prod_n (1+x^{d_n})$. Finally we conjecture a much stronger tight bound which gives the architectural bound $(d,d+1,d+1)$ for a space homeomorphic to the sphere.
arXiv:2610.13923v1 Announce Type: cross Abstract: Deep-space communications operate under intermittent connectivity, long propagation delays, and relay storage constraints. In such networks, the Bundle Protocol enables store-carry-forward communication, requiring relay nodes to manage limited storage and scarce contact opportunities. Relay-side scheduling and dropping decisions are inherently goal-oriented: transmission resources should be allocated to bundles most valuable to downstream mission objectives. Conventional schedulers, such as FIFO, do not explicitly optimize task utility and may provide unfair service across heterogeneous planetary sources. This paper presents a closed-loop modeling and learning framework for relay-side bundle and flow management in a contact-driven deep-space network, where lunar and Martian updates are forwarded through a shared-buffer relay to Earth. We quantify destination-side information freshness using excess age $\Delta_i^e(t)=[\Delta_i(t)-d_i(t)]^+$, where $\Delta_i(t)$ denotes the Age of Information of source $i$ and $d_i(t)$ is a source-dependent delay baseline. Based on this metric, we define a goal-oriented loss that supports source-dependent nonlinear penalties, allowing prioritization according to mission relevance. We formulate relay control as a constrained semi-Markov decision process (CSMDP) and solve it using masked Lagrangian Proximal Policy Optimization (PPO). Feasibility-aware action masking restricts transmissions to bundles that can complete within the remaining contact window. DTNSim simulations indicate improved age-based utility and inter-flow fairness compared with baseline schedulers. The proposed framework jointly optimizes relay-side bundle scheduling, dropping, and flow control under contact-driven dynamics with explicit fairness considerations across heterogeneous sources.
arXiv:2610.13766v1 Announce Type: cross Abstract: Geospatial foundation models can support disaster damage assessment, but high accuracy on familiar disasters may hide shortcut learning: prediction from event or scene characteristics rather than evidence that transfers to a new disaster. We study embeddings from TerraMind, Prithvi-EO-2.0, DOFA, and DINOv3 on the multi-disaster datasets xView2 and BRIGHT. Holding out an entire disaster, instead of randomly splitting image tiles, reduces mean balanced accuracy from 0.807 to 0.638 on xView2 and from 0.818 to 0.568 on BRIGHT. Four sets of tests examine what information supports prediction: event and acquisition context, before/after image inputs, building count, and correct image pairing. Pre-event images remain predictive, and a classifier given only the annotated number of buildings reaches balanced accuracy of 0.629 and 0.653. Combining before/after features performs best on average on xView2, but does not improve over pre-only features on BRIGHT. Shuffling image pairs within the same disaster and damage class causes modest losses. Motivated by these findings, we test whether feature transformations learned only from training disasters improve generalization to unseen disasters. These transformations improve performance, with supervised contrastive learning performing best on xView2 and PCA on BRIGHT. Our results highlight the importance of evaluation on unseen domains for assessing model robustness and show that simple feature-level transformations can improve generalization. Code and experiments can be found at https://github.com/alishibli97/gfm-shortcut-learning.
arXiv:2610.13674v1 Announce Type: cross Abstract: Linearizing softmax attention offers a route to efficient Transformer inference, but preserving pretrained attention remains challenging. We identify an important source of approximation error overlooked by standard linearization methods: a small number of outlier dimensions contribute disproportionately to query-key dot products and account for a large fraction of the approximation error. Motivated by this heterogeneity, we propose Outlier-Aware Attention Linearization (OAL), a dimension-aware method for constructing quadratic attention approximations. Starting from a second-order Taylor approximation, we introduce a Hadamard-product formulation that makes individual query-key contributions explicit. This formulation relaxes globally shared coefficients to support dimension-specific linear terms and cross-dimensional quadratic interactions while preserving separable evaluation. To retain this flexibility with fewer parameters, OAL groups dimensions with similar contribution scales to share coefficients, while allowing distinct treatment of the few outliers. The resulting positive quadratic kernel is calibrated against pretrained attention outputs and admits linear-time evaluation with respect to sequence length for fixed head dimensions. On Qwen2.5-0.5B, OAL reduces frozen-attention output relative MSE by approximately 8.9% compared with DiJiang. Replacing middle-layer attention modules and jointly fine-tuning the kernel coefficients with LoRA further lowers WikiText-103 perplexity from 31.71 to 27.50 compared with a learned global quadratic, while improving accuracy on PIQA, HellaSwag, and ARC-E. These results demonstrate the value of outlier-aware dimension grouping for both attention approximation and pretrained-model conversion. The code is available at https://github.com/wuqianyue09/OAL
arXiv:2610.13628v1 公告类型：cross 摘要：视觉-语言-动作（VLA）策略发出稀疏目标，这些目标抵达机器人控制器时存在延迟且不规律，因此，即使预测的动作合理，也可能在人员附近产生突兀运动。现有解决方案的评估所使用的数据流、延迟和指标各不相同。JitterVLA通过不同控制器重放完全相同的指令和到达轨迹，从而单独分析控制器的影响。我们将其应用于合成到达动作、20人的手部轨迹、三个开源 VLA 流程（838个回合）以及一台实体 UR3。在延迟200毫秒的人体运动中，插值使所有来源的建模 p99 加加速度均有所降低，降幅为83%（从615降至106 m/s^3），代价是2.0厘米的跟踪误差；这一效果在新增的598个回合中也得到了复现。在更大的代价下，受加加速度限制的生成方式进一步降低了加加速度。熵与不连续性之间的关联在两个流程中呈现方向相反且可复现的结果；熵门控在其中一个流程中有效，但在另一个流程中无效。在 UR3 上，插值使六组探索性配对中的加速度计推导加加速度降低了3.0%至14.0%，而基于位置推导的指标未能解释这一差异。JitterVLA支持在开展以人为中心的研究之前进行筛选；代码和数据将发布。
arXiv:2610.13578v1 Announce Type: cross Abstract: Pre-trained robot policies, such as Vision-Language-Action (VLA) models, provide a strong foundation for generalist robotic manipulation, but adapting them to new tasks with only a few demonstrations remains challenging. Directly fine-tuning a pre-trained policy often achieves strong in-distribution performance, yet it can distort the representations learned during pre-training and weaken the model's generalization ability under out-of-distribution shifts. We propose Retrieval-Anchored Representation Alignment (RARA), a two-stage adaptation method, to address this issue. First, RARA retrieves representation features from pretraining with similar task intent and motion, and aligns the policy's representations of the target demonstrations with the pre-trained representations of the retrieved demonstrations. Second, RARA fine-tunes the policy on the target demonstrations while limiting how much its representations of the retrieved demonstrations change during fine-tuning. By reducing the representation mismatch before action learning and preserving pre-trained knowledge during adaptation, RARA limits unnecessary representation drift during fine-tuning. Experiments with Diffusion Policy and GR00T on 13 tasks in RoboTwin, RoboCasa, and on a real robot show that RARA improves the average success of fine-tuning by 9.5 points in-distribution and 13.6 points out-of-distribution, and outperforms all compared baselines. Project page: https://shua-kang.github.io/RARA_robot/
arXiv:2610.13536v1 Announce Type: cross Abstract: When a language model is distilled from a teacher that shares its base parameters, the student can inherit behavioural traits of the teacher---including misalignment---even from data that is semantically unrelated to the trait (Cloud et al., 2026). All existing defences require knowing the trait. We remove this requirement and prove that a subliminally transferred trait can be recovered from the parameter updates alone, without supervision, by exploiting a single asymmetry: the trait is expressed consistently across diverse datasets generated by the teacher, whereas task-specific and shared capabilities vary. The resulting estimator returns a weight-space direction that can be amplified to discover and interpret the trait or projected out to suppress it. Applied to the subliminal-learning benchmarks of (Cloud et al., 2026), our method recovers the transmitted animal-preference traits without any trait labels and suppresses their expression with almost no impact to task performance.
arXiv:2610.13480v1 Announce Type: cross Abstract: Nonlinear model predictive control (NMPC) can be applied to large nonlinear systems with complex objectives while enforcing constraints, but the cost of solving the underlying optimization problems limits real-time application for systems with fast dynamics or many states. Neural networks can reduce this online cost by approximating the control law, but may violate constraints and typically require many precomputed optimizer solutions for training. The learning-based iterative solver for constrained optimization (LISCO) instead learns to predict primal-dual steps toward satisfaction of the Karush-Kuhn-Tucker (KKT) conditions through self-supervised training without precomputed solutions, but its dense, monolithic architecture grows with the prediction horizon. We remove this horizon dependence by exploiting the temporal structure of the problem in the solver architecture, which makes the network size independent of the horizon while leaving the training loss and the solver iteration unchanged. A closed-loop resampling strategy further aligns the training distribution with closed-loop operation at no additional sampling cost. On an economic NMPC benchmark of a CSTR cascade, the proposed solver scales to larger problems than monolithic LISCO and achieves median speedups of up to a factor of four over IPOPT. In closed loop, it achieves comparable control performance to IPOPT with no observed constraint violations.
arXiv:2610.13430v1 公告类型：交叉收录 摘要：时间序列因果发现——即检测时间序列之间因果关系的过程——源于半个多世纪前的信号处理研究。此后，这一领域持续受到广泛而积极的关注，并在统计学、动力系统和机器学习等领域中演化出不同的形式。对于初入者而言，这一分散且不断演变的研究领域可能令人望而生畏。本教程为不同研究社群中的时间序列因果发现提供了统一概述。我们介绍了现代因果发现方法所采用的关键原理，并讨论了这些方法的局限性与假设。我们还讨论了如何对因果发现方法进行评估、基准测试和微调。此外，我们探讨了时间序列因果发现如何日益成为农业与地球科学、医疗健康以及人工智能系统中下游任务的关键。我们相信，本教程既能帮助读者梳理近期涌现的大量因果研究，成为一项宝贵资源，也能邀请信号处理领域的读者参与因果问题的研究。为进一步帮助读者，我们维护了一个配套代码仓库，提供实用代码包和资源的链接，地址为 https://github.com/CHAI-UK/Awesome-Time-Series-Causal-Discovery。
arXiv:2610.13406v1 公告类型：cross 摘要：相比于对荒漠化进行制图，荒漠化预测受到的关注要少得多。我们在巴西半干旱的 Ceará 州开展研究，在约 2.2 km 分辨率的 256x206 网格上，使用 2000—2023 年的十个年度地球观测（EO）变量训练卷积 LSTM（ConvLSTM），以衡量卫星数据能够预测由 MapBiomas 派生的退化指数（IDV）年际演变的程度。该网络在最近观测状态的基础上预测有界变化，分两个阶段进行训练（先进行单步训练，再进行递归滚动预测），并在连续的空间与时间留出集上进行评估。2000 年至 2023 年间，全州平均 IDV 从 0.238 升至 0.350，年均增幅约为 0.005；但像元级变化主要由年际跳变主导，即使将所有通道输入梯度提升模型，也只能解释其 4.7% 的方差。在这一上限的背景下，ConvLSTM 在留出的南部带状区域中，相较于持久性基线，将提前一年预测的 RMSE 降低了 1.6%，将提前四年预测的 RMSE 降低了 4.4%（p < 0.001）；它优于带恒定漂移的持久性基线，并重现了区域漂移的 53%—69%。考虑漂移的基线模型从第三年起才有所帮助，在六年预测时最多可将 RMSE 降低 4.4%。一种较早的模型形式每年重新绘制整个场，其预测的变化与观测变化无关（相关系数为 -0.06），递归运行时还产生了虚假的下降趋势。通道消融结果表明，高程和 EVI 会使预测性能下降，而坡度和坡向没有可测量的影响。蒙特卡洛 dropout 和随机情景图呈现了期望值预测所掩盖的不确定性。在气候学强迫下，平均 IDV 将从 2023 年的 0.350 上升至 2030 年的 0.370。因此，基于 EO 的荒漠化预测对于区域趋势是可靠的，但对于单个像元并不可靠。
arXiv:2610.13353v1 Announce Type: cross Abstract: Multimodal emotion preference learning uses pairwise comparisons to identify which open-ended emotion descriptions align more closely with human judgments. However, existing methods primarily focus on the final preference decision and provide little explicit supervision for localizing and correcting erroneous content. To address this limitation, we propose EmoGavel, a Grounding-Aware reinforcement learning framework with Verifiable rewards for multimodal Emotion preference Learning. We first introduce grounded supervised fine-tuning (Grounded SFT), which exploits controlled-error construction information to jointly learn preference prediction, error localization, repair, and missing-evidence recovery. We then optimize these verifiable outcomes with grounding-aware group relative policy optimization (GA-GRPO), combining preference-gated grounding rewards with dynamic zero-variance rescue to prioritize preference correctness and recover learning signals from all-wrong groups. Experiments on EmoPrefer and controlled-error evaluation sets show that EmoGavel improves both natural preference prediction and controlled-error discrimination while maintaining strong error-grounding performance. Our code and model checkpoints will be released once the paper is accepted.
arXiv:2610.13330v1 公告类型：cross 摘要：针对复杂化学任务设计工具集成的推理智能体系统，仍然是一项根本性挑战。尽管近期的化学智能体通过集成外部工具展现出了良好的性能，但它们仍然建立在自回归范式之上，而该范式天然受到因果注意力和从左到右生成的限制。因此，它们往往难以实现双向分子理解、全局协调的工具规划以及多源证据整合。为应对这一挑战，我们提出了 ChemDiffAgent，这是首个用于工具集成化学推理的掩码扩散语言智能体。该方法将化学推理重新表述为对交错的推理轨迹和工具使用轨迹进行迭代去噪，使工具使用决策能够在完整轨迹上下文中联合确定，从而更好地使智能体的决策边界与其知识边界相匹配。我们进一步分析了这一范式如何通过训练期间更具挑战性的填空子问题，以及推理期间灵活的任意顺序解码，来提升智能体推理能力。随后，我们提出了一个两阶段后训练框架，包括智能体监督微调和方差缩减偏好优化，以增强工具使用与推理能力。我们还构建了一个综合基准，涵盖单轮和多轮工具使用场景下的六项化学任务。在预算和评测条件匹配的情况下，ChemDiffAgent 始终优于自回归智能体，整体性能提升约 20%。值得注意的是，我们的 8B 模型达到了与 GPT 5.5 和 Claude Opus 4.7 等规模显著更大的前沿模型相当的性能，在某些情况下甚至超过了它们。代码见 https://github.com/bilemond/ChemDiffAgent。
arXiv:2610.13190v1 Announce Type: cross Abstract: Autonomous landing of unmanned aerial vehicle (UAV) on unmanned surface vehicle (USV) is difficult in high sea states. Existing hierarchical landing methods often infer future USV motion only implicitly from short observation histories, which limits their ability to adapt to severe and rapidly changing sea conditions. This paper presents a prediction-informed hierarchical control framework. A pretrained model is developed to explicitly predict the USV attitude during high-level landing decisions. The prediction helps identify suitable landing windows and choose the right touchdown time, while a conventional low-level flight controller maintains UAV stability and tracks motion commands. To evaluate the framework under realistic disturbances, a real-sim validation setup is constructed in which USV-motion data collected in real sea conditions are transferred to a MuJoCo USV. The experiments demonstrate the feasibility of using explicit attitude prediction to improve landing-window selection and support timing-aware UAV landing planning in rapidly varying marine environments.
arXiv:2610.13112v1 Announce Type: cross Abstract: Large audio-language models (LALMs) frequently exhibit object hallucination, affirming absent sounds. The dominant mitigation, Audio-Aware Decoding (AAD), contrasts logits under real audio against an all-zero "blank" waveform, flagged by prior review as out-of-distribution. We introduce a black-box causal test that isolates the claimed evidence itself: localize a claim's evidence region via CLAP similarity, remove it by spectral gating, and compare against a random-window control across a 2x2x2 grid on three LALMs. On Qwen2-Audio-7B, localized ablation's raw AUROC trails blank's (0.613 vs. 0.721); this is a duration confound, not a localization failure, since blank removes the entire clip versus the grid's 2-4 s. Holding duration fixed, localization beats random removal at every cell on both models. Raw detection value remains model-dependent, especially on Audio Flamingo 3. The same evidence corrects decoding: contrastive decoding combining blank and language-prior contrasts, gated and ensembled, raises Qwen2-Audio accuracy from 57.1% to 71.9%, beating AAD alone. A unified causal-grounding profile does not transfer between models, indicating different grounding regimes. The source code and a detailed report of our findings, including identified limitations, are available at https://github.com/khalooei/causa.
arXiv:2610.13083v1 公告类型：cross 摘要：如今，模型提供商面临着通过数千个账户发起的蒸馏活动，而每个账户看起来都像普通用户。当网络和支付元数据无法将这些账户关联起来时，剩下的唯一证据就是它们提出的问题。我们使用一百万条真实聊天轮次进行开发，并用之后17周内、未受任何设计选择影响的数据进行一次固定评估，研究提供商仅凭这些查询能够、以及不能够了解到什么。答案与提取行为本身的特殊性关系不大，更多取决于良性协调：只有当某项活动比同一周内最具鲜明特征的普通用户群体更加独特时，它才是可见的。我们对此进行了形式化，表明活动与用户之间一个微小但为正的差异带来的是检测所需时间，而不是可检测性，并确定了最难检测的活动。实证结果显示，经过校准的监测器能够捕捉到按账户监测器遗漏的朴素分布式提取活动（覆盖0.86至1.00的周数）；其中一种变体——在扣除每个账户的主题性之后，选择行动步调一致的账户——即使活动中填充了自然文本，仍能捕捉到其中的活动（覆盖0.50至0.96的周数），而另外两个谱检测器一个也捕捉不到。将每项活动蒸馏成一个小型学生模型后发现，在我们的测试环境中，能够躲过所有检测器的活动，其能力最多只有可检测活动的42%。这些局限同样具体：缓慢升级的活动会污染所有检测器的校准，良性的评估流程与模型蒸馏看起来完全一样；当串通者所占比例为2.5%或更低时，我们没有找到任何可部署的检测器；而转发真实用户请求的活动，可能对任何仅依赖查询的监测器都不可见。
arXiv:2610.13002v1 Announce Type: cross Abstract: Empowerment measures the capacity of an agent to actively control its environment. In this work, we explain the connection between skill learning and empowerment. While skill learning is not empowerment, we introduce a simple scalable recipe for empowerment maximization: empowerment corresponds to the number of executable skills from a state, so maximizing empowerment corresponds to reaching states that are good starting points for many skills. Our recipe applies to arbitrary underlying skill learning methods. Whereas most prior work on empowerment looks at tabular settings, this recipe enables us to study empowerment at a scale where we can realize theoretically-predicted but empirically-elusive properties of empowerment (e.g., gathering resources, environment modification, survival). Experiments show that empowerment maximization unlocks interesting emergent properties, such as building a house in Crafter. Website and code at \href{https://empowerment-geometry.github.io/scaling-empowerment/}{this https url}.
arXiv:2610.12970v1 Announce Type: cross Abstract: The increasing demand for GPU-accelerated machine learning inference has exposed limitations in centralized GPU infrastructures, including high cost, limited elasticity, and long wait times under resource constraints. At the same time, modern Apple Silicon-based macOS devices are equipped with powerful GPUs that remain largely idle during normal usage. This paper presents Minions, a dynamic and distributed virtual GPU cloud that aggregates idle Apple Silicon GPUs across macOS devices to execute large-scale machine learning inference workloads. Minions enables teams to leverage existing hardware without requiring additional GPU procurement. The system provides flexible GPU inference task execution, fault tolerance, and efficient resource utilization. Our experience shows that idle macOS devices can be transformed into a practical distributed GPU inference platform in environments where centralized GPU resources are constrained.
arXiv:2610.12907v1 Announce Type: cross Abstract: Federated vision-language models (FedVLMs) represent one of the emerging frontiers of machine learning, aiming to integrate federated learning within the fine-tuning pipelines of vision-language models (VLMs). In this work, we introduce BLEND, a new framework designed to jointly enhance the personalization and generalization capabilities of FedVLMs. BLEND proposes a selective VLM parameter fine-tuning and aggregation strategy that consists of (i) global vision and text adapters shared across clients, and (ii) a personalized vision projection adapter tailored to each client. BLEND further introduces a novel loss function that blends the outputs of the global and personalized adapters to achieve personalization, while simultaneously promoting generalization through a principled anchor design implemented via the projection heads of the zero-shot VLM. Extensive experiments demonstrate that BLEND consistently outperforms state-of-the-art baselines, highlighting its ability to effectively balance personalization and generalization across diverse system settings. (Code is available at: https://github.com/payamsiabd/BLEND.git)
arXiv:2610.12876v1 Announce Type: cross Abstract: Mixed-integer programming (MIP) solvers are widely used, yet their performance can vary substantially even under seemingly neutral perturbations. Such variability complicates benchmarking, and the interpretation of computational studies. We introduce the Weighted Variability Score (WVS), a robust multi-component metric that aggregates dispersion in normalized runtime, final optimality gap, LP iterations, and branch-and-bound nodes across repeated runs. On a large subset of MIPLIB 2017 instances, each solved in 100 randomized runs under a four-hour limit, WVS reveals a right-skewed distribution with a substantial upper tail: many instances are essentially stable, but some exhibit large variability that is not explained by basic properties alone. Notably, this experiment proved optimality for the open instance 'neos-5045105-creuse' in two SCIP runs. Using static descriptors and root-level features, a tree-based predictor with a gated tail uplift achieves a mean absolute error of 0.058 and Spearman rank correlation of 0.65 on a sealed test split. Feature-importance analysis highlights root degeneracy, matrix structure, coefficient ranges, and presolve reductions as the main drivers of high WVS. Because variability is systematic and partially predictable from inexpensive early-run features, our findings offer a practical tool for preemptive tuning and lay a foundation for fairer evaluation of new algorithms.
arXiv:2610.12843v1 公告类型：交叉发布 摘要：更少的潜在词元有望降低扩散模型的计算成本。然而，激进的潜在空间压缩会造成重建与生成之间的两难：增加通道数可以改善重建，却会使生成更加困难。我们提出 APEX，其全称为 Asynchronous Prefix denoising for EXtreme compression，将分层潜在空间与针对通道组的异步去噪相结合。我们的自编码器将通道前缀按从结构到细节的顺序排列，而一个共享去噪器则在错开的局部时间步生成这些通道组，无需增加词元。在 ImageNet-512 上，在 64 倍和 128 倍压缩条件下，APEX 在所有评估的通道数设置中均优于现有方法。在训练和引导采样计算量相同的情况下，它在所有评估的计算预算下也均优于一个 256 词元模型，且计算量越低，优势越明显。代码见 https://daviddinkevich.github.io/APEX。
arXiv:2610.12806v1 Announce Type: cross Abstract: We study byte-level autoregressive language models whose hidden state is a small quantum register, whose transitions are learned token-conditioned quantum channels, and whose next-byte probabilities come directly from Born-rule measurements without a learned classical vocabulary projection or softmax. In the fixed-width recurrent setting, unitary dynamics cannot contract or reset stored distinctions; experimentally, unitary recurrence loses to its retrained one-token control in every tested variant. Introducing learned dissipation, then token-conditioned dissipative gating, yields useful memory beyond one token. The gated model improves test negative log-likelihood by 0.20 +/- 0.03 nats/byte over retrained parameter-matched one-token controls at context length 12, and by 0.29-0.33 at longer contexts, but never beats retrained two-token controls. A key-value recall benchmark shows a width-dependent threshold: one stored pair meets the pre-registered accuracy bar at 4-5 memory qubits, two pairs at 6-7, and three pairs remain unsolved. This is consistent with, but does not establish, a capacity explanation for the short effective memory horizon. As a secondary result, one-qubit models reach lower stationary memory entropy than a heuristically optimized classical predictive frontier at matched error in 9/9 seeds. On IBM Heron hardware, computational-basis memory-register distributions show signal against a uniform null in 12/12 circuits across prefixes of up to 12 tokens, next-bit probabilities pass pre-registered accuracy bars at four-token prefixes, and bytes are generated autoregressively by sampling physically measured bit probabilities. The tested models are classically simulable and substantially worse than classical n-gram baselines. We claim no computational quantum advantage.
arXiv:2610.12775v1 Announce Type: cross Abstract: The two main traditional player statistics used in soccer are goals scored and assists, with an assist credited when a player's pass immediately leads to a teammate scoring. Attributing credit solely to the player who makes the final pass before a goal fails to account for the contributions of multiple players in the buildup to the goal. Consequently, traditional statistics provide only a limited view of player involvement. In this paper, we use multivariate Hawkes processes (MHPs) to model timestamped sequences of passes and shots, which form continuous-time networks. MHPs provide a branching mechanism where an event (a pass or shot) can be generated as a child of a previous event, which allows us to quantify each pass's contribution to a subsequent pass or shot. This approach distributes credit across multiple passes within an attacking sequence, making it applicable across a wide range of situations. Building on this framework, we propose expected goal involvements (xGI), a metric that measures a player's involvement in creating goal scoring opportunities based on their passes. This metric identifies players who play key roles in creating goal scoring opportunities even if they do not make the final pass, offering a more comprehensive measure of attacking involvement and player performance.
arXiv:2610.12747v1 Announce Type: cross Abstract: Ego-trajectory prediction in urban roadway work zones remains an underexplored topic in autonomous-driving research, with limited dedicated data and systematic evaluation. Roadway work zones involve temporary changes in the lane layouts and connectivity of the roads, complicating interactions with surrounding vehicles and workers. These changes make future vehicle trajectories difficult to predict. To address these challenges, we collect WZDrive-Data, a multimodal dataset comprising 1,183 work-zone driving segments from 33 drivers, using a human-in-the-loop driving simulator with prerecorded worker animations and VR-controlled workers. Driver control and gaze signals are recorded from human participants, while vehicle trajectories and scene information are obtained from the simulator. Our proposed multimodal ego-trajectory predictor, WZDrive, achieves a mean minADE (K=6) of 1.09 m and minFDE (K=6) of 2.12 m. Compared with the best-performing state-of-the-art baseline for each metric in our benchmark, WZDrive reduces minADE (K=6) by 7.18% and minFDE (K=6) by 13.18%. This research addresses a gap in autonomous-driving research on urban work zones by combining human-in-the-loop data collection with multimodal learning from human driving signals and in the context of work zones.
arXiv:2610.12734v1 公告类型：交叉收录 摘要：双目视觉正越来越多地用于机器人操作，越来越多的研究将其融入视觉-语言-动作（VLA）模型。然而，大规模真实世界双目视觉数据集的匮乏，仍是此类策略规模化发展的根本瓶颈。近期的生成式单目转双目方法主要面向 VR 内容和自动驾驶等场景。此前尚无研究为机器人操作生成逼真的合成双目数据，该领域的双目监督信号主要来自仿真。我们提出 StereoEngine，引入两种互补的方法来合成机器人操作所需的双目数据。第一种方法利用估计深度，将单目（左视角）视频变换为右视角，并使用视频扩散模型修复因遮挡解除而显露的区域，从而无需使用真实双目数据进行训练，即可将单目示范转换为双目视图对。第二种方法先在仿真中渲染双目视图对，再联合生成两个照片级逼真视图，无需任何真实视频作为输入，即可生成左右一致的数据。StereoEngine 在生成双目数据的质量上优于现有基线方法，且将其合成数据用于策略的中期训练或特定任务微调，能够显著提升下游双臂操作性能。项目网站：https://stereoengine.github.io
arXiv:2610.12704v1 Announce Type: cross Abstract: Federated learning allows dispersed oilfield sites to train a shared intrusion detector without pooling network traffic, yet its aggregation step forces a choice between privacy and robustness. Byzantine-robust rules must read each update in the clear, whereas secure aggregation hides updates from the checks that detect poisoning. We present PVBRF-ID, a two-server protocol in which sites secret-share their updates and the servers jointly test each update's norm and its cosine similarity to a reference computed on a small clean root set, revealing one accept bit per site. Commitments, Merkle roots and a homomorphic audit make the inclusion of accepted updates publicly verifiable. On a 15-class Edge-IIoTset task with ten sites holding differently distributed data and ten random seeds, PVBRF-ID matched FedAvg on clean data (macro-F1 0.708 vs. 0.706) and withstood sign-flip attacks (0.696 vs. 0.091). Against an adaptive attacker controlling 30% of sites, PVBRF-ID with secure clipping reached 0.641, whereas FedAvg fell to 0.046; plaintext multi-Krum remained higher by 0.052. An update-inference attacker recovered each site's class mix from plaintext updates (Spearman correlation 0.81) but not from one server's view. Backdoors passed the filter, and noise-based mitigation gave only heuristic protection. The implementation uses Python 3.13, NumPy 2.3, 30 rounds and 10 seeds.
arXiv:2610.12659v1 公告类型：交叉发布 摘要：以数据为中心的 RLVR 扩展一直侧重于收集或生成可验证的训练问题。然而，仅凭语料库规模并不能判断数据集是否能为目标策略提供有效的学习信号。为指导有效 RLVR 数据的合成，我们通过实验，将数据集的有效性刻画为其所含问题、目标策略和训练 rollout 预算的函数。基于这一刻画，我们发现，一个汇集了广泛使用的 RLVR 数据集的大型开源数学语料库几乎完全陷入了零优势陷阱：对于固定的目标策略和 rollout 预算，其中约 88% 的问题要么在每次 rollout 中都能解决，要么在每次 rollout 中都无法解决。筛选可以提高剩余有效问题的集中度，但它只能进行选择，同时丢弃难题。因此，有效的合成问题应当要求使用解决这些难题所需的推理操作及其组合。不过，这些问题的难度必须经过校准，使部分 rollout 成功、部分失败，并且至少有一个正确解答能够在训练预算内完成。直接提示教师模型对难题进行变式改写并不能达到这一目的。在这些实证原则的指导下，我们提出了 MagicBeanStalk，一种结合自顶向下分解与自底向上递归数据合成的算法。它在保留种子问题推理蓝图的同时，将种子问题扩展为一系列在实质上全新、且针对策略和预算进行了校准的训练问题。在 8 个数学基准测试中，在 16K token 预算下，使用 MagicBeanStalk 问题持续进行 RLVR 训练，使 gpt-oss-120b 的成绩从 61.2 提升至 69.8（+8.6），使 DeepSeek v4Flash 从 72.4 提升至 77.9（+5.4）。使用完整的 7.8 万道问题源语料库进行训练仅提升了 1.1 分，而使用其中经筛选得到的最佳真实数据子集则提升了 5.8 分。在金融领域，我们仅基于 340 个解答就合成了超过 1.4 万道可验证问题，并在 SecQue 和 FinanceBench 上分别比基础模型提升了 6.9 分和 2.4 分。
arXiv:2610.12545v1 公告类型：交叉收录 摘要：自回归解码不太适用于非因果顺序的数据，例如交叉引用或对后续内容的依赖，因为与人类的迭代推理不同，其输出无法在事后修订。扩散式 LLM 在一定程度上缓解了这一局限，但仍无法在预先生成的上下文中灵活插入或删除任意数量的词元。我们设计了 AutoIndexer，这是一种采用新颖训练与解码机制的生成模型，通过模拟编辑链，使模型能够在事后通过插入、替换和删除词元，灵活修改先前的输出。该模型学习预测用于开启和结束编辑的标记词元，同时将光标移动到现有词元序列中的相应位置，类似于文本编辑器。训练序列通过对真实标签施加逆向链式编辑操作进行扰动而获得，模型则通过撤销这些逆向操作来学习编辑。这种扰动训练使 AutoIndexer 对错误上下文具有鲁棒性，并能适应代码纠错等编辑任务。我们证明，该方法具备大多数自回归模型和扩散模型所缺乏的生成后上下文编辑能力，同时不会损害其在标准 LLM 基准测试上的表现。代码已发布于 https://github.com/AutodeskAILab/autoindexer。
arXiv:2610.12522v1 公告类型：交叉收录 摘要：大型语言模型（LLMs）正越来越多地用于分子发现与结构解析，这些领域的决策依赖于物理观测。现有评估主要关注端到端的分子重建，而正确的预测并不一定意味着模型依赖了所提供的性质信息。我们提出 MolP2SBench，这是一个用于评估 LLMs 从分子性质到结构（P2S）推理能力的基准。每项任务提供一个部分被掩蔽的分子骨架及给定性质，要求模型从合理的候选项中选出缺失的结构。我们通过打乱或替换性质，检验模型对正确性质—结构配对的依赖程度，以及对反事实性质变化的响应能力。按信息通道分别评估的排行榜采用基于性质的恢复指标（Property-Grounded Recovery，PGR），对原始条件与反事实条件下的恢复表现赋予相同权重。通过评估七个模型，我们发现 LLMs 能够利用物理性质进行结构预测，但在不同性质类型上的表现差异显著。使用某种性质时准确率更高，并不一定意味着对该性质的响应更强：模型从 $^1$H NMR 中获得的收益可能大于从 $^{13}$C NMR 中获得的收益，但在 $^1$H 数据被替换时，反而更不容易改变预测。LLMs 的表现也未必反映人类化学家利用这些性质进行推断时的相对难度。尽管 UV-Vis 光谱提供的直接结构信息少于 NMR 光谱，Astra 使用 UV-Vis 数据时的准确率却高于使用 NMR 谱峰时。MolP2SBench 支持研究 LLMs 如何提取和利用物理性质信息，揭示可能不同于传统化学直觉的模式，并为构建更可靠地以性质为依据、用于分子与量子化学推断的模型奠定基础。代码和数据集：https://github.com/DUAL-Xiao/MolP2SBench。
arXiv:2610.12501v1 Announce Type: cross Abstract: Many headphone users risk irreversible hearing damage during high-load listening, yet current protection still relies on sound-pressure limits. Clinical tests detect damage after it becomes measurable, leaving the risk-accumulating process unobserved. The acoustic reflex (AR), an involuntary response to high acoustic load, marks when the ear enters a protective state. AR traces are weak, and existing measurements require dedicated probes and controlled conditions that are difficult to maintain during everyday audio. We present EarLog, which turns playback, the dominant interference, into an opportunistic probe. EarLog introduces content-conditioned residual sensing to suppress variations caused by nonstationary content, individual ear acoustics, and device response. It identifies weak AR through loudness-dependent changes and localized spectral irregularities in the residual. A resonance-constrained sparse finite impulse response (FIR) model captures the short-window response. Dual-microphone observations, mild-distortion recovery, and cross-session reference updates keep detection usable under low signal-to-noise ratio (SNR), fit changes, and environmental drift. EarLog runs in real time on an embedded controller using a low-cost dual-microphone add-on. We evaluate it with 81 participants, 16 headphone models, and more than 400 hours of recordings. EarLog achieves 99.0% sensitivity and 98.7% specificity on a clinically referenced benchmark. EarLog records listener-specific AR events during everyday listening, providing data for future longitudinal studies of hearing risk.
arXiv:2610.12493v1 Announce Type: cross Abstract: Explainable Artificial Intelligence (XAI) tools have often been presented as methods to increase trust in machine learning models by making aspects of their algorithms more human-understandable. However, evidence accumulated over the last decade points to an overall failure of this objective, instead promoting overreliance on incorrect predictions, which can subsequently lead to algorithmic aversion. Rather, XAI appears to function primarily as a debugging tool, useful for obtaining anecdotal insights into, for example, features learned by the model. One of the main issues underlying this limitation is that XAI methods often oversimplify the inner workings of the model, resulting in misalignment between the explanation and the model itself - a behavior referred to as unfaithfulness or infidelity. Faithfulness can be interpreted as the degree of approximation between an explanation and the model. Computing faithfulness in a rigorous manner would thus make it possible to quantify one of the various elements of trustworthiness of an explanation. However, existing metrics present significant limitations, making this quantification either infeasible or computationally expensive. In this position paper, we outline five major shortcomings of current faithfulness metrics and propose four possible avenues for research to bolster their reliability.
arXiv:2610.13924v1 Announce Type: new Abstract: A parallel sampler can choose at random which coordinates to generate together. For a target $p$ and the output law $q_U$ of a schedule $U$ drawn afresh for each output, we compare the schedule-averaged divergence $\mathbb{E}_U D_{\mathrm{KL}}(p\|q_U)$ with the output-law divergence $D_{\mathrm{KL}}(p\|\mathbb{E}_U q_U)$, in nats. Rounds commit independent draws from exact conditional marginals frozen at the round's start, each output's schedule is drawn before sampling from a finite mixture of schedules, and every scheduled round is charged. For the uniform distribution on binary strings of even length $n$ with $n/2$ ones, the least expected number of rounds at a fixed tolerance $\varepsilon>0$ is $\Theta_\varepsilon((\log n)^2)$ under the first criterion but $(e^{-\varepsilon}+o(1))\log_2\ln n$ under the second, and the double-logarithmic order is attainable with the same round count on every run. For this target, a count-state recursion gives the exact least output divergence at every round cap over all causal randomized policies. The separation extends to weighted fixed-count targets with bounded weight ratios and quota fractions bounded away from zero and one, without permutation symmetry, and to the full-support Beta$(1,1)$--Bernoulli target, where the sharp expected cost under the output-law criterion is $(e^{-\varepsilon}/2+o(1))\log_2\log_2n$, attained with rare $n$-round runs. Prior work implies the order of the schedule-averaged cost on the balanced target and its $O_\varepsilon((\log n)^2)$ upper bounds on both extensions.
arXiv:2610.13909v1 公告类型：新发布 摘要：尽管面向表格数据的深度学习模型近来受到了研究界的广泛关注，但回归任务中目标函数的选择仍是一个研究不足的自由度。多个研究方向提出了标准均方误差（MSE）的替代方案，但数据集、骨干网络和评估协议的差异，使得这些方案的实际收益难以比较。在本研究中，我们通过在大量成熟的表格数据集和多种神经网络骨干上，对回归目标函数进行系统性的重新评估，弥补了这一空白。比较结果表明，相较于经过充分调优的 MSE 基线，许多已提出的目标函数带来的提升有限或不稳定，而分位数损失在各个受测骨干网络上，相较于 MSE 均取得了稳定的总体提升。随后，我们发现，即使采用这一表现优异的目标函数，仍存在一个局限：目标噪声水平不同的区域，可能在不同的训练轮次达到最佳预测性能。因此，即使目标函数能够表示异方差目标分布，单一的全局早停检查点仍可能迫使模型在不同不确定性区域之间作出折中。针对这一局限，我们提出了一种感知异方差性的分位数损失扩展方法，使用依赖于输入的标签平滑作为潜在的解决方案，并在多种骨干网络上取得了优于分位数损失的表现。总体而言，我们的结果为表格数据深度学习中的回归目标函数提供了新的实证图景，并揭示了不同不确定性区域之间最优停止时机的差异如何限制全局早停的有效性。代码可在 https://github.com/yandex-research/tabular-regression 获取。
arXiv:2610.13789v1 Announce Type: new Abstract: Momentum methods such as STORM and second-order momentum track the gradient by adding a transport correction to momentum, at the cost of an extra oracle call per iteration. We introduce randomized momentum transport, which evaluates the correction with probability $q$ and otherwise uses a cheap prediction, with importance weighting that preserves the conditional mean and adds $(q^{-1}-1)$ times the prediction loss to the variance. Under transport-bias, second-moment, and residual-control assumptions, linear minimization oracle (LMO) updates, including normalized, sign, and exact spectral (Muon) directions, retain an $O(T^{-1/3})$ stationarity bound for each fixed $q>0$ with a single-gradient initialization. For unnormalized STORM, some $q<1$ lowers the leading coefficient of our sufficient gradient budget if and only if the gradient-difference noise modulus exceeds the smoothness constant, by at most about $10\%$. An online diagonal predictor admits a regret-to-convergence guarantee. At approximately equal oracle cost, querying a quarter of the corrections lowers the final training loss by $18$ to $43\%$ on MNIST1D and raises the test accuracy of Hessian-vector transport on CIFAR-10 by $1.5$ points.
arXiv:2610.13765v1 Announce Type: new Abstract: Discrete graph diffusion models corrupt graph structure through independent edge updates with a Markovian noise model. Combined with the exchangeability requirement on the output distribution, the terminal prior is necessarily an Erd\H{o}s--R\'enyi (E-R) random graph. This has a fundamental consequence for unconditional graph generation without node features: in the large graph limit, any GNN denoiser applied to an E-R graph approximately outputs an E-R graph, making effective denoising difficult. In practice, generated graphs fail to recover meaningful node heterogeneity and higher-order structure. Previous models introduced specialized node features to compensate, framing this as an expressivity problem, while the role of the prior remained unexamined. We instead address the root cause by introducing the Latent Sociability Prior (LSP) and a join} diffusion process that co-evolves latent node sociability variables and graph structure, removing edge independence while preserving exchangeability. Unlike existing latent graph diffusion methods, the sociability mechanism requires no complex encoder: its sole role is to guide edge updates and preserve node heterogeneity throughout the trajectory. We validate our approach on multiple graph generation datasets and show that the proposed framework more accurately captures graph distributions than baseline methods without node features.
arXiv:2610.13755v1 Announce Type: new Abstract: Training long-horizon language model agents requires assigning credit to intermediate decisions from sparse terminal feedback. Within a finite group of sibling rollouts, useful action comparisons depend on their context, evidence for both alternatives, and how that evidence enters learning. We propose Matched-Context Group Policy Optimization (MCGPO), a procedure for constructing weighted action preferences from already collected trajectories. MCGPO defines comparison contexts with pre-action state summaries and observed rubric progress, annotated by task rules. Two-action support selects among ordered exact and coarse matching rules. Fixed folds separate candidate decisions from the terminal outcomes used to orient their preferences, and support-dependent evidence weights scale an auxiliary pairwise objective alongside shared GraphGPO task learning. MCGPO reuses rollout data and each response's original history, preserving critic-free training. Across six environment-backbone settings with three training seeds each, MCGPO improves mean success over matched GraphGPO: by 0.47-1.46 percentage points on ALFWorld and WebShop with Qwen2.5-1.5B and 7B, and by 3.96 and 5.16 points on WebShop with Qwen3-4B and 8B. The ALFWorld gains build on baseline success above 92%, where remaining headroom is limited. Integrated success improves in three of four Qwen2.5 settings. Three-seed factorial controls support opposite-fold evidence and evidence weighting; joint state-rubric matching achieves the highest mean in all four representation-control settings.
arXiv:2610.13717v1 公告类型：新论文 摘要：联邦聚类可支持跨医院的患者分层、分布式金融数据集中的客户细分等应用。负责任的人工智能要求公平地表征各人口群体，并支持数据删除，这在一定程度上源于 GDPR 的删除权等权利要求。这些目标相互影响：删除记录可能改变群体之间的平衡，并改变剩余人群的聚类结果。我们研究联邦社会公平 $k$-means 的精确机器遗忘，其目标是最小化各群体到最近中心的平均平方距离中的最大值。我们的方法为每个群体分别保留本地摘要，因此，当全局公平性权重发生变化时，未受影响机构的摘要仍然有效。对于两个群体的情形，我们证明了在理想采样条件下，初始化具有独立于客户端群体比例的近似保证，并表明使用相同随机种子进行重放能够精确复现重新训练的结果。当整个客户端退出时，复用带来的收益最大：在三个数据集上，对于仅包含初始化的模型，重放的速度为重新训练的 $35.8\times$；加入一轮和两轮公平性细化后，速度仍分别为重新训练的 $2.4\times$ 和 $1.7\times$，而这些细化轮次能够降低表现最差群体的代价。对于单条记录的删除，收益较为有限（约为 $1.06\times$），且验证缓存分配结果的开销可能超过其节省的开销。这些结果明确了精确遗忘能够通过复用计算获得收益的适用范围。代码和可复现性材料见 https://github.com/arareddy/mufsc。
arXiv:2610.13654v1 公告类型：新论文 摘要：表格基础模型（TFMs）因其强大的预测性能而日益受到欢迎。许多此类模型还提供后验预测分布（PPDs），但这些分布的可靠性仍缺乏充分研究。我们从点预测、分布质量、边际与子群校准以及区间效率等方面评估可靠性。随后，我们推出 TabReliBench，这是一个用于上述比较和数据集层面分析的大规模分类与回归基准。我们发现，TFMs 在 TabArena 和本基准上的排名有所不同，在不同可靠性指标上的排名也会发生变化。回归 TFMs 通常能够达到接近名义水平的边际覆盖率，同时保持较窄的区间，但这种良好的边际表现并不意味着条件覆盖率也能得到保证。数据集层面的分析表明，分类难度较高往往伴随着较差的校准，而回归预测性能并不是衡量校准情况的可靠指标。集成方法能够提高预测性能，但并不总能改善 PPDs 的校准；TFMs 从保形后处理中获得的收益也很有限，这种处理无法可靠地消除子群间的差距。因此，要可靠地选择 TFM，就需要超越综合排名，开展多维度评估。代码可在 https://github.com/Tianqi-Zhao/TabReliBench 获取。
arXiv:2610.13635v1 Announce Type: new Abstract: Hypergraph Neural Networks (HNNs), designed to model higher-order relations beyond pairwise interactions, typically rely either on fixed structural weights or on normalised attention. Fixed weights can propagate misleading information in heterophilic settings, while normalisation makes attention coefficients competitive, in the sense that, in a message-passing framework, increasing the weight of one sender must reduce the weight of others in the same aggregation set. Our main contribution is the introduction of PANDA (Prior-guided AttentioNal Dual-path Architecture), a plug-in for two-stage message passing HNNs applicable to both undirected and directed settings. The main path interpolates between learned receiver-normalised attention and structural incidence priors, while a non-competitive (not normalised) auxiliary path lets each sender contribute its own transformed information. We also show that, with fixed structural coefficients, PANDA recovers a certain complex-valued convolution operator adopted in previous works. Across four HNNs spanning spatial and spectral methods in both directed and undirected settings, plugging PANDA in yields an average improvement of 5.18 percentage points across eight real-world benchmark datasets. The code is available at https://github.com/jainris/panda.
arXiv:2610.13631v1 Announce Type: new Abstract: Fault diagnosis in asset-intensive industries requires multiclass models trained on noisy, sensitive data collected across geographically distributed client sites. Federated learning enables collaborative training without centralizing data, but most approaches do not provide formal privacy guarantees for individual records stored at client sites and can degrade under label noise. To address these limitations, we develop a differentially private, distributionally robust framework for federated multiclass classification that provides sample-level privacy for records stored at each client. We build on a federated Wasserstein distributionally robust multiclass objective with the Crammer-Singer loss, solved using an Alternating Direction Method of Multipliers (ADMM)-based algorithm. To protect the released client updates, our method injects calibrated Gaussian noise into the proximal linearized client updates of both the local model and the local distributionally robust variable before aggregation at the central server, and we quantify the cumulative privacy loss using the moments accountant. We provide per-round sample-level $(\epsilon,\delta)$-differential privacy guarantees and an overall composed privacy guarantee across all training rounds, together with convergence results and an explicit privacy-utility tradeoff. Experiments on simulated industrial fault data and UCI benchmark datasets show that our method outperforms differentially private federated empirical risk minimization and stochastic gradient descent baselines in most tested settings, with the largest gains under strong privacy-induced noise.
arXiv:2610.13599v1 公告类型：新论文 摘要：在超声心动图等多视图医学影像检查中，诊断信息分布于不同视图中，每个视图捕捉心脏结构和功能的不同方面。在临床实践中，医生首先通过定量测量解读这些视图，再依据测量结果作出具体诊断。然而，现有的超声心动图表征学习方法并未将测量显式建模为视图与诊断之间的中间表征。因此，各项测量未能以可供下游诊断任务选择性使用的形式得到显式表示，从而妨碍了准确诊断。受这种分层临床工作流程的启发，我们提出了 EchoTab，这是一种以测量为基础的表征学习框架，使用测量特定的词元来表示多视图超声心动图检查，并将其作为超声心动图视图与诊断之间的中间表征。每个测量词元选择性地关注与其估计相关的视图，并通过表格重建进行训练，以预测对应的测量值。由此得到的词元构成一种因子化的超声心动图表征，编码了具有临床意义的结构和功能特征，可供下游诊断选择性使用。与常用的对比学习方法不同，EchoTab 不需要负样本或大批量训练。在两个数据集上的实验表明，EchoTab 能够学习有效的、以测量为基础的表征，从而提升多视图超声心动图的诊断分类性能；其学习到的注意力也与心脏科医生所采用的既定测量与视图对应关系一致。我们的代码可在 https://github.com/michelleespranita/EchoTab 获取。
arXiv:2610.13571v1 Announce Type: new Abstract: We study simultaneous competitive and regret guarantees for TCP acknowledgment. Each acknowledgment serves all pending packets, and the objective is to minimize the total acknowledgment and waiting costs. It is well known that no deterministic online algorithm achieves a competitive ratio better than $2$. Given $k$ randomized online experts and $0<\varepsilon<1$, we construct a randomized online algorithm whose realized total cost is always at most $(2+\varepsilon)$ times the offline optimal cost $\mathrm{OPT}$ and whose expected regret against the expert with the smallest realized cost is $O(\sqrt{\mathrm{OPT}\log(k+1)}+\log(1/\varepsilon))$. The algorithm requires no advance knowledge of the number of packets or $\mathrm{OPT}$. Our algorithm controls switching costs between experts and uses the gap between the permitted competitive bound and its current cost to determine when to learn. We also prove lower bounds showing that, for $k\ge 2$, this regret bound is asymptotically optimal in $\mathrm{OPT}$, $k$, and $\varepsilon$, and that $\varepsilon>0$ is necessary for sublinear regret.
arXiv:2610.13504v1 公告类型：新论文 摘要：生成式推荐系统越来越多地使用强化学习（RL）来优化可能在决策后数百乃至数千次交互后才出现的结果。标准 RL 根据时间邻近性而非明确的相关性传播奖励，削弱了分配给时间久远但重要的动作的信号，成为样本效率低下的一个来源。我们提出 Explanatory World Models（EWM），这是一种归因框架，利用解释识别可能促成已实现结果的早期事件。我们使用前沿语言模型实现 EWM，并通过一种守恒的层次化算子将其解释用作学习信号；该算子递归地重新分配奖励，同时保持未折扣的回合回报不变。为了探究什么样的解释才有用，我们将受 Halpern 的实际因果关系、Lombrozo 的解释简洁性以及 Deutsch 的难以变动的解释启发的三种互补变体转化为可操作的方法。我们在两个受控的长时程 RL 基准、一个公开的序列推荐基准以及一个工业规模的生成式推荐系统上评估 EWM。在 Crafter 和 WordCraft 上，EWM 显著优于 PPO 和 LaRe，且在奖励延迟最严重时取得最大提升。在 KuaiRand 上，EWM 在离线评估中优于 SFT 和 LaRe，在在线评估中优于 A2C。最后，在对工业级生成式推荐系统的离线评估中，EWM 显著提高了召回率，并提升了行为代理指标。这些结果表明，解释能够为从延迟结果中学习提供结构化且可复用的信用归因。
arXiv:2610.13435v1 公告类型：新投稿 摘要：可穿戴设备和临床设备正以日益庞大的规模记录光电容积描记（PPG）信号，越来越多的 PPG 基础模型（FM）也相继被开发出来，用于计算可供各种下游任务复用的表征。然而，这些模型难以进行比较：不同评估采用的数据集、任务定义、预处理方法、下游预测头、超参数调优方法和基线各不相同。我们提出 PPG-Bench，这是一个统一的开源基准，在九个开放访问数据集和 26 项任务上评估五个公开的 PPG 基础模型，任务涵盖心血管状态估计、活动、睡眠、筛查和身份识别。我们在适应能力逐步增强的四种下游设置下评估模型（线性探测、随机森林、冻结模型的 MLP 探测和端到端微调），并采用针对各模型的超参数优化及受试者互不重叠的交叉验证。我们还设置了三类对照：随机/常数基线、确定性的脉率变异性（PRV）特征集，以及通用时间序列基础模型（MOMENT）。在全部 68 种任务与设置组合中，35.3% 的组合未显示基础模型相较于随机/常数基线具有实质性优势，而在 83.8% 的组合中，没有任何 PPG 专用基础模型能同时优于 PRV 和 MOMENT。我们还发现，增加下游预测头的容量通常可以提升性能，但如果模型的表现处于随机猜测或常数预测水平，则可能导致性能显著下降。因此，目前的 PPG 基础模型尚不能宣称能够提供具有泛化性和可迁移性的 PPG 信号表征，该领域未来的评估应纳入与简单、非学习型对照方法的比较检验。该基准的代码及评估更多模型的说明见：https://github.com/mannas-ai/ppg-bench。
arXiv:2610.13404v1 Announce Type: new Abstract: Linear models can predict well in complex domains even when underlying mechanisms are nonlinear. We argue that this can occur because the complexity of the world exceeds the complexity that can be learned reliably from finite, partially observed data. Flexible models can exploit curvature, thresholds, and interactions that fit one hospital, period, population, policy setting, or measurement process but fail elsewhere. When this instability lies mainly in the nonlinear component, omitting it can give linear models lower prediction error in a new environment. We connect this claim to work on simple linear rules, bias and variance, regularization, best linear prediction, measurement error, distribution shift, invariance, Rashomon sets, and underspecification. Proposition 1 gives a simple squared-error case in which a nonlinear rule learned in the training environment raises target error because it transfers too weakly; finite-sample estimation error can strengthen this result. Partial observation alone is not enough: the link between observed variables and unobserved factors must vary across environments. Stable nonlinear structure favors flexible models, and linear models remain vulnerable to linear spurious relationships. Model complexity should therefore be chosen by cross-environment stability as well as training-data accuracy.
arXiv:2610.13350v1 公告类型：新提交 摘要：广泛复用的电子健康记录预处理流程可能隐含一些假设，涉及信息何时可用、标签代表哪一次临床诊疗过程，以及数据缺失意味着什么。这些假设很少得到系统性审查，并可能随着预处理流程的复用或调整而传播到后续研究中。我们以 MIMIC-IV 为案例，识别并量化了三种互补的预处理对齐形式，即时间对齐、语义对齐和人群对齐的违反情况，它们分别表现为时间泄漏、特征与任务不匹配以及有偏插补。我们展示了时间戳语义、诊疗过程定义、终点构建、模态选择和缺失值处理中的歧义如何改变模型可用的信息，并影响对其预测结果的解释。通过对 MIMIC-IV 数据的分析和实证案例研究，我们证明，这类选择虽然会改变模型所依赖的信号，却可能在总体预测性能中基本不显现。我们据此提出了切实可行的建议，以便对 MIMIC-IV 及相关电子健康记录数据集开展时间上有效、与临床相契合且可复现的预处理。我们的实验代码（包括所提出的预处理方法）可在以下地址获取：https://github.com/m2l-lab/mimic-iv-preprocess。
arXiv:2610.13335v1 Announce Type: new Abstract: Tool-using agents increasingly require a common substrate for building, training, evaluating, distilling, and maintaining agentic systems, yet existing tooling typically addresses these stages in isolation. We present AgentTune, a source-available Python library that provides a unified framework for developing and optimizing tool-using agents. At its core, AgentTune introduces a two-tier EventLog representation that preserves agent trajectories while supporting consistent projections for training, evaluation, and auditing. Building on this substrate, the library provides a shared agent runtime, tool-calling reinforcement learning, agentic distillation, retrieval-augmented generation, decision workflows, evaluation utilities, and a designed path toward self-healing agents. These components are exposed through composable interfaces so that agent strategies, training methods, evaluation procedures, and recovery mechanisms can be combined without requiring separate representations or pipelines. We demonstrate the framework across agentic tasks including legal contract search and domain-specific applications, illustrating its use for trajectory-based optimization, tool-use evaluation, and cross-task agent development. AgentTune is released as a source-available library at https://github.com/Lexsi-Labs/AgentTune
arXiv:2610.13261v1 Announce Type: new Abstract: In this paper, we propose Data-Adaptive Mahalanobis Cross-Head Attention (DAMCHA), a simple and effective attention mechanism that generalizes Multi-Head Attention (MHA) through the lens of metric learning. Standard MHA implicitly learns a static, head-wise Euclidean similarity corresponding to the diagonal blocks of a unified cross-head metric matrix, discarding all off-diagonal interactions across heads. DAMCHA addresses this limitation via two key innovations. First, it learns the full cross-head metric matrix, inducing a Mahalanobis-like similarity that explicitly captures inter-head interactions beyond the restricted block-diagonal structure of MHA. Second, it parameterizes the metric matrix as an input-dependent function, enabling the attention geometry to dynamically adapt to the intrinsic manifold structure of the data rather than remaining fixed after training. We further introduce principled regularization strategies, most notably stack-wise parameter sharing, to ensure computational efficiency and stable optimization. DAMCHA serves as a drop-in replacement for MHA and its variants in various Transformer architectures, delivering superior expressive power and learning performance with comparable or reduced computational overhead. Extensive experiments demonstrate that DAMCHA-based Transformers consistently outperform competing baselines across a range of benchmark tasks. Code is available at https://github.com/erwinmsmith/DAMCHA.
arXiv:2610.13258v1 Announce Type: new Abstract: Neural surrogates provide computationally efficient auto-regressive modeling of wave propagation, enabling e.g. fast inverse solutions. Despite recent progress, the current surrogates are not yet sufficient for the needs of ultrasonic sensing, characterized by high frequency and heterogeneous propagation media. We show how methods building on Fourier Neural Operators (FNO) fail to capture essential signal features, such as reflections, whereas U-Net -based methods learn the wave fields well but lack the ability to transfer information across different spatial scales. We then propose a new surrogate that captures the best of the two worlds, by integrating FNOs and local-integral resampling into a U-Net architecture. We show it can be trained on limited data, and that the predictions are robust to input noise and work across spatial scales.
arXiv:2610.13225v1 公告类型：新稿 摘要：通过在统计流形上学习连续向量场，将简单先验输运至目标离散分布，离散数据生成建模取得了显著进展。然而，这些连续时间流并未被约束为沿统计流形上的测地线运动，因此需要通过多步数值积分进行模拟，导致推理阶段的计算成本高昂。在本研究中，受 Rectified Flow 在连续欧几里得空间中“直且快”特性的启发，我们提出了 Rectified Statistical Flow（ReSFlow）模型。该模型通过显式学习遵循内在黎曼几何的测地线轨迹，实现统计流形上的单步类别生成。通过校正统计流形上的流，我们从理论上保证了先验分布与目标分布之间的轨迹逼近测地线。这种校正使学习到的轨迹直接沿统计流形上的测地线对齐，从而显著降低输运成本，使模型能够在单步内将纯噪声映射为目标离散数据。通过在涵盖图像、文本和生物领域的多种真实世界生成任务上开展广泛实验，我们展示了 ReSFlow 卓越的单步生成质量，真正实现了测地线意义下最优且快速的类别生成。我们的代码已公开于 https://github.com/xiaohuanjia/ReSFlow。
arXiv:2610.13221v1 Announce Type: new Abstract: Cancer diagnosis and treatment increasingly rely on multi-modal diagnostic technologies that generate high-dimensional datasets. Although these datasets are rich in pathological information, they pose considerable challenges in interpretation, dimensionality reduction, and clinical applicability. Conventional machine learning approaches often fail to extract meaningful cross-modal features, thereby limiting predictive accuracy in patient risk stratification. To address these limitations, this paper proposes an unsupervised machine learning framework that integrates feature extraction, tensor decomposition, and clustering to improve cancer risk prediction. The framework employs tensor decomposition to fuse heterogeneous genomic and clinical data, enabling the extraction of latent, biologically meaningful features that outperform unimodal feature sets. Multiple feature extraction strategies are investigated, and the resulting features are subsequently clustered using a range of clustering algorithms. Comparative evaluations based on internal and external metrics are then used to identify the most effective configurations. The framework is validated on LinkedOmics data comprising 11,158 patients across 32 cancer types. For patient risk stratification, clinically relevant attributes such as tumor purity, pathological stage, and survival outcomes serve as ground-truth indicators. Experimental results demonstrate the superiority of tSVD+GMM, which consistently outperforms other configurations across clinical benchmarks.
arXiv:2610.13197v1 公告类型：新论文 摘要：部分可观测条件下的强化学习（RL）通常依赖循环网络来编码时序信息，并将循环模块与策略或价值函数联合训练。这引出了一个自然的问题：训练循环网络真的有必要吗？本文研究冻结的循环网络能否提供足够的记忆，以求解部分可观测马尔可夫决策过程（POMDP）。受储备池计算（RC）和广义储备池计算（GRC）理论基础的启发，我们研究了一种简单架构，将固定的循环网络与可训练的非线性输出模块相结合。我们在记忆需求、观测模态和动作空间各异的任务上评估了这种方法。令人惊讶的是，冻结的循环网络能够回忆数百个时间步之前的信息，并在多种 POMDP 中取得具有竞争力的控制性能，而所需的可训练参数数量不到对应的全可训练网络的一半。这些结果表明，求解 POMDP 并不总是需要训练循环记忆模块。因此，冻结的循环网络为部分可观测强化学习提供了一种简单、参数高效且表现出乎意料地强劲的基线。我们的源代码已发布于 https://github.com/Aphlios/FRN-POMDP
arXiv:2610.13186v1 Announce Type: new Abstract: Federated Learning (FL) enables collaborative model training without exposing raw data, yet remains vulnerable to malicious clients injecting poisoned datasets. Existing defenses are typically reactive (operating during training) or supervised (requiring prior knowledge of attack types) - assumptions that rarely hold in open-ended IoT deployments. We propose WaSA, a fully unsupervised, offline framework for malicious client detection. Before training, each client transmits a single, privacy-preserving spectral summary of its data to the server. A denoising autoencoder, trained solely on clean auxiliary data, scores clients based on their deviation from the learned benign manifold, identifying corruptions regardless of the attack type. WaSA requires no attack labels, generalizes to unseen anomalies, and introduces a Boltzmann sampling strategy for client selection that smoothly interpolates between hard exclusion and standard aggregation. Experiments on MNIST, CIFAR-10, and CIFAR-100 under diverse attack scenarios (up to 90 % malicious clients) demonstrate consistent improvements over supervised baselines and robust aggregation methods, with negligible communication overhead. Code is available at https://github.com/alelicciardi99/wavelet_autoencoders_malicious.
arXiv:2610.13064v1 Announce Type: new Abstract: Large Language Model (LLM) agents have demonstrated remarkable capabilities in tackling complex tasks through agentic workflows. However, manually designing these workflows is labor-intensive and lacks flexibility to adapt to diverse queries. While automated workflow optimization is a promising direction, existing methods often incur expensive API costs during training-phase feedback collection. Furthermore, they fail to adaptively reuse high-quality workflows, which creates a bottleneck for both performance and cost-effectiveness. To address these limitations, we propose CB-Orchestrator, an adaptive framework that decouples workflow generation from selection. In the first stage, we construct a diverse pool of candidate workflows via evolutionary search. In the second stage, we formulate workflow selection as a Contextual Bandit problem, which enables sample-efficient learning by balancing exploration and exploitation, thereby significantly reducing the feedback required for training. During inference, the model adaptively selects the optimal workflow tailored to each query. Evaluations across five diverse benchmarks demonstrate that CB-Orchestrator consistently outperforms all baselines. Notably, on the MATH dataset, CB-Orchestrator achieves better performance than strong recent baselines while reducing training-phase API token overhead by no less than 40.55% and total end-to-end training (including workflow pool generation) token overhead by no less than 29.73%.
arXiv:2610.12960v1 Announce Type: new Abstract: Matching repair after utility changes can reuse valid observations, but new evaluations must respect matching capacity and may return out of order. We study announced invalidations with bounded, possibly reward-dependent reporting delays and pending evaluations that do not occupy future matching capacity. Residualload repair (RLR) combines sound outcome certificates with batched matching schedules for unequal sample deficits. Its complete dyadic quota schedule admits a sharp issue-round bound with coefficient cb < 1.607 relative to the terminal residual endpoint load, together with separate delay and evidence-lifetime bounds. RLR-P monitors complete issue-order prefixes without changing the sampling plan before acceptance. With a deterministic pure oracle, a single-repair coupling from the same drained initial state guarantees no increase in certification time, issued observations, or time to a reusable checkpoint. We also derive an information-load lower bound. On fixed priority-star markets with unknown Bernoulli gaps, PrioritySearch attains the history-adjusted first-order constant for expected certification time as {\delta} {\sim} 0, under hypothesis-independent linear history scaling and sublogarithmic reporting delays. Paired simulations and measured-feedback replay quantify the trade-offs: in the replay, RLR-P has lower mean capped update-certification time than a monitored width-based baseline, at higher oracle CPU cost.
arXiv:2610.12915v1 Announce Type: new Abstract: Minimax optimization is fundamental to adversarial learning yet suffers from severe stability challenges, often manifesting as persistent limit cycles or divergence. This paper investigates these instabilities through the lens of continuous-time dynamical systems. By modeling training dynamics as a second-order Ordinary Differential Equation (ODE), we rigorously identify the root cause of non-convergence: the problematic eigenstructure of the loss function's Hessian near equilibrium points. To address this, we introduce Controlled-Minimax, a novel stabilization framework that reformulates the optimization trajectory as a controllable physical system. Unlike heuristic regularization techniques, our approach derives a specific feedback controller that provides theoretical guarantees of local exponential stability, regardless of the underlying Hessian spectrum. Consequently, our method stabilizes general minimax problems where traditional gradient descent-ascent methods fail. Extensive experiments on pathological benchmarks including the Dirac-GAN, 2D Gaussian mixtures, and Swiss Roll datasets, as well as high-resolution image synthesis with StyleGAN2-ada, demonstrate that Controlled-Minimax effectively suppresses oscillatory behavior while ensuring fast and robust convergence.
arXiv:2610.12828v1 公告类型：新论文 摘要：结果监督的组相对策略优化（GRPO）为 LLM 智能体轨迹中的每个词元分配相同的优势值，未能区分有益和无益的决策。后见信用分配（HCA）可以细化这一反馈，但其估计依赖于轨迹中在关键历史状态下包含具有信息价值的备选路径。我们提出 TreeHCA，将后见引导探索与后见信用分配相结合。TreeHCA 利用参考答案似然的变化，重新访问中间历史状态并采样替代性的后续轨迹，从而构建由推理和工具交互组成的树。由于这种探索改变了所采样后续轨迹的分布，我们推导出相应的逆后见重要性校正，其中包括一个精确的有限样本池恒等式。我们在归一化的树回传中使用这一结构，将终局结果传播为轮次级优势值。这种共享的后见结构是 TreeHCA 的核心设计原则，将具有信息价值的探索与后见引导的信用分配联系起来。在七个问答基准和三个 Qwen 骨干模型上，TreeHCA 的总体精确匹配率均优于所有基线，验证了将探索与信用分配配对用于多轮 LLM 智能体训练的有效性。我们的代码已发布于 github.com/TheBlackCat22/TreeHCA。
arXiv:2610.12820v1 Announce Type: new Abstract: The shock response spectrum (SRS) characterizes the peak responses of single-degree-of-freedom systems to transient acceleration. Because the mapping from an acceleration time series (waveform) to its SRS is many-to-one, inverse synthesis seeks candidate waveforms consistent with a prescribed spectrum. We extend an earlier conditional variational autoencoder (CVAE) framework with a recurrent decoder that predicts multiscale wavelet coefficients and reconstructs waveforms through an inverse discrete wavelet transform. Individual models exhibit limited variability under latent sampling, motivating an ensemble of independently trained models spanning different mother wavelets and decomposition levels. Each member produces a separate candidate waveform for the same target SRS. Evaluation on four held-out benchmark datasets shows lower mean and median spectral errors for the illustrated wavelet model than for the earlier direct time-domain CVAE and a sum-of-decayed-sinusoids (SDS) baseline. A conditioning study examines frequency-grid and encoding choices. Comparisons with published inverse-SRS methods demonstrate superior performance across four holdout benchmark datasets (two operational shock datasets, one earthquake dataset, and one synthetic shock dataset) and competitive performance on synthetic targets reported by Brake. Single-model inference requires approximately 0.1-0.6 ms per waveform during batched inference compared with 5-8 s for SDS, excluding offline training. This low inference cost enables rapid generation of distinct SRS-compatible candidate waveforms across the ensemble, although their physical suitability requires assessment beyond spectral agreement.
arXiv:2610.12809v1 Announce Type: new Abstract: Continuous random node features increase the expressivity of message-passing neural networks (MPNNs), but existing theoretical analyses are not fine-grained enough to capture their limitations, particularly in approximating functions across graph sizes or under a fixed number of random bits. We develop a population-level framework for size-uniform approximation. For MPNNs with a graph-size-independent Lipschitz bound in their initial features, including standard mean, max, and order-normalized sum architectures, we show that continuous random features can be quantized to finitely many colors with arbitrarily small uniform error in expected output. This reduction lets us transfer limitations on finite-color distinguishability to limitations on size-uniform approximation with continuous random features. We use this framework to prove that these MPNNs are not size-uniformly universal, analyze the expressivity gains from colors and independent trials, and establish incomparability between random-color MPNNs and higher-order Weisfeiler-Leman methods. We also give finite-color MPNN constructions for tree subgraph isomorphism via color coding and a local MST task.
arXiv:2610.12769v1 公告类型：新论文 摘要：大型推理模型（LRMs）中的遗忘要求同时从推理轨迹和最终答案中移除遗忘集信息。现有方法以推理崩溃为代价实现这一目标，即模型无法生成连贯的思维链（CoT），也无法正常终止其推理过程。为应对这一挑战，我们提出 REDaCT（REgenerate via Detection and Corrective Tuning），一个面向 LRMs 的推理时遗忘框架。REDaCT 由两个关键组件构成：（1）一个二元泄露检测器，用于识别推理轨迹中泄露遗忘集信息的句子；（2）一个嵌入优化模块，通过离线学习获得纠正嵌入，在推理过程中首次检测到泄露时注入该嵌入，从而引导后续推理生成不泄露遗忘集信息的句子。实验结果表明，REDaCT 完全避免了推理崩溃，与最先进的遗忘方法相比，大幅提升了推理能力的保留程度，同时实现了完全的答案遗忘和近乎完全的 CoT 遗忘。我们的代码已公开：https://github.com/jmaccoby/redact。
arXiv:2610.12744v1 Announce Type: new Abstract: Parameter-efficient continual learning updates a restricted subset of adapter parameters to incorporate sequential tasks while limiting catastrophic forgetting. Gradient attribution is a common rule for selecting that subset, but it requires an additional backward pass over calibration data. We identify a parameter-pool confound in comparisons between initialization-time gradient masks and unrestricted random masks under standard zero-initialized LoRA ($B=0$): the gradient with respect to $A$ is identically zero ($\partial\mathcal{L}/\partial A=0$), so gradient masks select strictly from the $B$ matrix whereas unrestricted random masks draw from both $A$ and $B$. Such a comparison changes both the selection rule and the candidate parameter pool. Isolating the selection criterion by restricting both approaches to the same parameter pool, we find a negative result: a uniform random mask consistently achieves lower loss-based backward transfer than gradient attribution across seven seeds on Qwen2.5-3B. The effect is concentrated on the medical task, which contributes 80\% of the aggregate; code and math move in the same direction, and legal shows no reliable difference. This advantage is consistent with a stability--plasticity trade-off under the tested task sequence, in which random masking slightly reduces single-task fitting. On the two discrete-option tasks (Medical and Legal), the mean accuracy-BWT difference is not statistically distinguishable from zero.
arXiv:2610.12737v1 公告类型：新论文 摘要：基于梯度的归因能够提供模型参数随具体输入变化的信息，但通过反向传播为大量样本计算归因的成本高昂，而静态重要性分数又舍弃了对输入的依赖性。我们从经验神经切线核的分块分解推导出 KPSA（Kernelized Parameter Sensitivity Atlas）。该方法衡量归一化的参数组敏感度，将这些源样本嵌入显式查询空间，并使用轻量级核函数估计未见输入的敏感度。我们推导了将有限源样本误差与不可约表示误差分离的界，并给出了基于间隔的 top-$k$ 恢复保证。本文研究了两种可解释性应用。首先，KPSA 在视觉、预测和蛋白质建模场景中识别出对扰动高度敏感的参数组。在跨度为五倍的参数预算范围（$0.02$-$0.1\%$）内，其选定参数组的扰动敏感度达到一阶反向传播预言机所选参数组的 $39$-$85\%$。在视觉和预测任务中，这些参数组构成跨越不同深度的稀疏敏感度回路，并随输入呈现系统性变化。其次，我们将 KPSA 用于 Transformer 中 MLP 特征的激活归因问题。在视觉任务上，它保留了反向传播预言机相对于恒定敏感度基线的性能优势的 $81\%$，同时将批量查询时的吞吐量从每秒 551 张图像提高到 2,041 张，并将 GPU 峰值显存占用从 4.55GB 降至 0.44GB。在蛋白质建模任务上，它恢复了这一性能差距的 $68\%$。对 KPSA 核聚合的消融研究表明，在 KPSA 空间中进行最近邻检索仍可保持竞争力。实证结果表明，紧凑的核图谱能够存储有用的、依赖于输入的参数敏感度信息，同时摊销重复梯度归因的计算成本，为实现可扩展的条件可解释性迈出了一步。代码见 https://github.com/Axym-Labs/kpsa 。
arXiv:2610.12699v1 公告类型：新投稿 摘要：我们研究如何学习生成数据，使其以用户指定的约束满足概率满足预设的\emph{逐样本}约束。从总体层面的带约束负对数似然最小化问题出发，采用流匹配（FM）参数化，在正则性假设和速度场族具有充分表达能力的条件下，我们推导了相应原始-对偶优化算法的约束满足概率保证。我们给出了该方法的实用实现，称为 Primal-Dual Flow Matching（PDFM）。该方法能够直接控制约束满足率，且只需接收用于判断是否满足约束的二元成员资格判定反馈，无需可微距离、投影，也不要求约束集合具有凸性等特殊结构。我们表明，PDFM 在保持分布保真度的同时，相较于标准 FM 提高了约束满足率，并在与约束感知基线方法的比较中取得了具有竞争力的性能。代码已公开，地址为 https://github.com/ZhengyanHuan/PDFM。
arXiv:2610.12679v1 Announce Type: new Abstract: Efficient sampling from Bayesian posterior distributions in high dimensions remains a hard problem, despite advances in theory and algorithms. Multiple steps are usually required to draw unbiased samples, for example, iterations of an MCMC algorithm or integration steps in a diffusion-based amortised sampler. However, scalable one-step amortised inference remains a challenging open problem. In this work, we build upon recent progress in data-driven generative modelling to offer a possible solution for one-step amortised sampling. Our one-step implicit sampler (OSIS) uses short-time MCMC dynamics to define a local target for the evolution of an implicitly modelled distribution. In the asymptotic limit, the modelled distribution is a fixed point of the MCMC kernel and thus samples from the target. We analyse the convergence properties of OSIS and show connections with other generative modelling and sampling algorithms. We study various design choices and show that OSIS performs well compared to multi-step amortised samplers and non-amortised MCMC baselines while being far more efficient at sampling time.
arXiv:2610.12586v1 公告类型：新论文 摘要：事件序列是由随机事件构成的链，其中每个事件都由异构特征（数值型和类别型）描述。事件序列预测问题广泛存在于金融、医疗、业务流程智能等领域。在本研究中，我们将事件序列的多预测跨度预测视为一个条件生成建模问题。我们不仅评估预测的准确性，还评估其分布真实性。为此，我们提出了一个基准，涵盖衡量这两个质量维度的指标、强大的统计基线方法、从相邻领域改编的近期模型，以及经过精心整理的真实世界数据集。细致的评估表明，没有任何方法能在准确性和真实性两方面均表现优异，这凸显了两者之间的根本性权衡。最后，我们提出了 Wasserstein BaryBooster，该算法将真实性表现出色的扩散模型生成的样本进行聚合，构建出达到当前最先进水平的预测器，证明分布真实性可以转化为预测准确性。代码可在 https://github.com/xaosina/event-sequence-forecasting 获取。
arXiv:2610.12520v1 Announce Type: new Abstract: The goal of circuit discovery is to decompose a monolithic neural network into tiny modular subnetworks which are interpretable in the sense that each does an identifiable subtask. We develop a practical method for composing arbitrarily many circuits into a single model. We propose a greedy, density-scaled composition algorithm that overcomes limitations of previous attempts to compose circuits by naive union, then generalize it to $N$ circuits via CircuitTree, a new data structure built on a $\Theta(N\log N)$ tree-compose schedule. We compose circuits pruned from HuBERT for classifying individual phonemes into a syllable classifier, a capability none of the 39 phoneme circuits has individually. Chaining CircuitTrees into a hierarchy-of-circuits pipeline, we transcribe open syllables at 64.8\% accuracy, within 1.2 points of a 66.0\% ceiling that shares no circuitry at all, with 99.5\% of circuit components shared.
arXiv:2610.12495v1 Announce Type: new Abstract: Small language models trained on a single local machine leave little room for architectural trial and error. Kairos combines bidirectional gated-DeltaNet linear attention and sliding-window attention over shared projections (LiZAttention2), block-wise Attention-Residual aggregation, a sparse Mixture-of-Experts (MoE) feed-forward layer and a multi-scale byte-level codec, trained with a uniform-noise denoising objective over one byte space shared by text, image, audio, video, lidar and control streams; this report asks whether the combination optimizes at all before asking whether it is useful. We give code-derived algorithms for each component, including a cross-session recurrent-state memory gate, then run convergence sanity checks: overfittinga single 440-byte sentence at 1--15M parameters, text only, with three seeds for codec and attention choices and single-run sweeps for seven further factors. The two attention variants tested at 3\,000 steps (LiZAttention2 and window attention alone) reach a minimum loss near $10^{-3}$. After 600 steps, the mean minimum loss ranges from 0.018 (patch codec) to 0.354 (depthwise convolutional codec) with an identical ranking on three seeds. An apparent advantage of shared over separate attention projections closes with more steps or more diverse data and reverses when the projections are frozen: it is a convergence-speed effect, not a capacity effect. These checks suggest that the implementation optimizes; they say nothing about generalization, multimodal training or comparison with existing models, which a planned pretraining run is meant to address. The Discussion sizes a 200M-total/25M-active target analytically; none of its throughput figures is measured. Code is available at https://github.com/fabienfrfr/Kairos . Python package at https://pypi.org/project/kairos-fm/ .
Curriculum learning typically organizes training samples from easy to hard. In GRPO, medium-difficulty samples are attractive for online selection because they are more likely to yield informative group-relative learning signals, and prioritizing them can improve training efficiency. However, it remains unclear whether persistently selecting such samples also leads to the strongest final model performance, or whether training can benefit from progressing toward harder samples. Meanwhile, sample difficulty changes as the policy evolves, so fixed difficulty estimates can become misaligned with the current policy. We therefore re-estimate sample difficulty under the current policy and, inspired by the easy-to-hard principle of curriculum learning, design and evaluate a staged curriculum that progressively shifts the active difficulty band toward harder samples. Experiments on GSM8K show that persistent medium-difficulty selection does not achieve the strongest observed performance. Across three random seeds, progressing to a moderately harder difficulty band improves test accuracy to $0.6591 \pm 0.0077$, compared with $0.6449 \pm 0.0024$ for the fixed-medium curriculum. However, further progression to a harder difficulty band reduces accuracy to $0.6405 \pm 0.0053$. These results show that staged difficulty progression can outperform persistent medium-difficulty selection, but its benefit is non-monotonic. Effective GRPO curricula can therefore benefit from progressing beyond medium difficulty, while avoiding excessive progression toward overly difficult samples.}
将 softmax 注意力线性化为高效的 Transformer 推理提供了一条路径，但保留预训练注意力的行为仍然具有挑战性。我们发现了一个被标准线性化方法忽视的重要近似误差来源：少数异常维度对查询与键的点积贡献不成比例地大，并造成了很大一部分近似误差。受这种异质性的启发，我们提出 Outlier-Aware Attention Linearization（OAL），这是一种用于构建二次注意力近似的维度感知方法。从二阶 Taylor 近似出发，我们引入了一种 Hadamard 积形式，显式呈现查询与键各维度的贡献。该形式放宽了系数全局共享的限制，支持维度特定的线性项和跨维度的二次交互，同时保留可分离计算的性质。为了以更少的参数保留这种灵活性，OAL 将贡献尺度相近的维度分组以共享系数，同时允许对少数异常维度进行区别处理。所得正值二次核以预训练注意力输出为参照进行校准，并且在注意力头维度固定时，可按序列长度的线性时间复杂度计算。在 Qwen2.5-0.5B 上，与 DiJiang 相比，OAL 将冻结注意力的输出相对均方误差降低了约 8.9%。替换中间层注意力模块，并结合 LoRA 联合微调核系数后，与学习得到的全局二次核相比，WikiText-103 上的困惑度进一步从 31.71 降至 27.50，同时在 PIQA、HellaSwag 和 ARC-E 上的准确率也有所提升。这些结果表明，异常值感知的维度分组对于注意力近似和预训练模型转换均具有价值。代码已公开于 https://github.com/wuqianyue09/OAL
Standard Policy Gradient (PG) algorithms suffer from high estimation variance because they rely on Monte Carlo updates. While off-policy variance reduction mitigates this problem, establishing optimal sample complexity currently requires either restrictive assumptions, such as uniformly bounded importance weights, or impractical second-order operations. In this work, we introduce Loopless Variance-Reduced PG (LVR-PG), a unified theoretical framework that abstracts a class of first-order estimators that integrate recursive updates with defensive importance sampling, including known estimators such as PAGE-PG and STORM-PG. We prove that this algorithmic class achieves $\tilde{\mathcal{O}}(ε^{-2})$ last-iterate global convergence under weak gradient domination for general parametric policies. We establish this guarantee using purely first-order updates, with minimal assumptions and without gradient preconditioners. By extending state-of-the-art lower bounds to continuous parametric policies, we establish an $Ω(ε^{-2})$ sample complexity limit, confirming our framework is minimax-optimal up to logarithmic factors and horizon terms.
Discrete graph diffusion models corrupt graph structure through independent edge updates with a Markovian noise model. Combined with the exchangeability requirement on the output distribution, the terminal prior is necessarily an Erdős--Rényi (E-R) random graph. This has a fundamental consequence for unconditional graph generation without node features: in the large graph limit, any GNN denoiser applied to an E-R graph approximately outputs an E-R graph, making effective denoising difficult. In practice, generated graphs fail to recover meaningful node heterogeneity and higher-order structure. Previous models introduced specialized node features to compensate, framing this as an expressivity problem, while the role of the prior remained unexamined. We instead address the root cause by introducing the Latent Sociability Prior (LSP) and a join} diffusion process that co-evolves latent node sociability variables and graph structure, removing edge independence while preserving exchangeability. Unlike existing latent graph diffusion methods, the sociability mechanism requires no complex encoder: its sole role is to guide edge updates and preserve node heterogeneity throughout the trajectory. We validate our approach on multiple graph generation datasets and show that the proposed framework more accurately captures graph distributions than baseline methods without node features.
具身递归自我改进的一项核心挑战，是让智能体将交互历史转化为能够跨世界保留、并迁移至后续任务的能力。仅凭已保存的记忆或成功的最终动作，并不能证明智能体获得了可复用的能力；评估必须将能力变化与沿用的资源、不均等的源任务接触量以及早期执行失败区分开来。我们构建了 MC-EvoBench，这是一个受控的 Minecraft 基准，包含由 19 条已注册演化链组成的固定目录，其中包括 12 条主要评估链和 7 条校准或探测链，以及三组用于迁移评估、与图结构对齐的配对探测任务。该基准将每项任务表示为任务状态图，定义阶段级环境验证器，分配相匹配的源任务与目标任务角色，重置目标世界，并且只允许预先声明的智能体产物跨越源任务与目标任务之间的边界。对于每项探测，我们在匹配的环境步数预算和五个随机种子下，评估无先验、相关、不相关和打乱顺序的源任务条件；报告 TSR、WGP 和 VMP，并保留溯源记录。该基准揭示，控制器变体与规划器支持的变体之间存在一致的直接能力差距，而相关经验并未带来稳定的迁移效果：它未能改善任何一项工具探测的表现，在木材种类探测中的微小收益依赖于单个随机种子，且所有区间均包含零。阶段审计进一步表明，早期木材采集与合成环节的瓶颈往往导致后续能力无法被观测。因此，MC-EvoBench 为持久且可迁移的具身能力提供了一个受控测试平台。代码和数据可在 https://github.com/xzw-ustc/mc-bench-open-source 获取。
随着 AI 智能体日益依赖外部工具，准确检索对于确保其行为可靠至关重要。现有方法着重优化语义相关性，却在很大程度上忽视了执行过程中的工具可用性。为弥补这一不足，我们提出 RATED（基于执行动态的可靠性感知工具检索），这是一个将稳健的语义匹配与动态可靠性适应相结合的检索框架。RATED 采用经对比微调后冻结的双编码器进行语义匹配，并使用轻量级上下文多臂老虎机模块，根据实时执行反馈对候选工具重新排序。这一设计能够快速适应运行时的工具故障，而无需进行成本高昂的模型重训练。我们推出了 EcoAgent，这是一个包含 5,172 个真实世界工具和超过 210,000 个查询—工具对的基准，用于离线检索与动态可靠性评估。实验结果表明，在真实世界查询上，RATED 相较于 ToolLLM API 检索器的平均倒数排名（MRR）相对提升了 52.2%；在故障率为 30% 的严重 API 中断场景下，故障工具选择率从 86.2% 降至 2.9%，同时相较于仅编码推理，额外延迟不足 1 毫秒，几乎可以忽略不计。代码和数据集已发布于 https://github.com/Elliotes/EcoAgent。
尽管面向表格数据的深度学习模型近年来受到研究界的广泛关注，回归任务中目标函数的选择仍是一个研究不足的自由度。多个研究方向提出了标准均方误差（MSE）的替代方案，但数据集、骨干网络和评估流程的差异使这些方案的实际收益难以比较。在本研究中，我们在大量成熟的表格数据集和多种神经网络骨干上，对回归目标函数进行系统性重新评估，以弥补这一空白。比较结果表明，相较于经过恰当调优的 MSE 基线，许多已提出的目标函数带来的提升有限或不稳定，而在所有测试的骨干网络上，分位数损失相较于 MSE 均带来了稳定的总体提升。随后，我们指出，即使采用这一强效目标函数，仍存在一个局限：目标噪声水平不同的区域可能在不同的训练轮次达到最佳预测性能。因此，单一的全局早停检查点可能迫使模型在不同不确定性区域之间作出折中，即使目标函数能够表示异方差目标分布也不例外。针对这一局限，我们提出了一种考虑异方差性的分位数损失扩展方法，利用依赖于输入的标签平滑作为潜在的补救措施，并在多种骨干网络上取得了优于分位数损失的表现。总体而言，我们的结果为表格数据深度学习中的回归目标函数提供了最新的实证概况，并强调了不同不确定性区域之间最优停止时机的差异如何限制全局早停的有效性。代码可在 https://github.com/yandex-research/tabular-regression 获取。
In August, I posted here about TeXbrain, a LaTeX editor that runs completely in the browser, with no need for an account or server. (I didn't excpect it to get that much feedback, bug reports and help from here!) This is the same idea for the other half of my studies, the handwritten notes. I take them on a convertible laptop, and I've had a hard time with apps for that, taking notes from a device to another, exporting as PDF, needing an account for many of them, etc. So I've been working on the version of it that I always wanted. Open a tab, the slides of the lecture or multiple PDFs open next to the page or even become the page, write on them with my pen, and nothing ever leaves my browser. braiNOTES is a static website and has no backend. Strokes are perfect-freehand outlines which support pen pressure, the PDF files are rendered by pdf.js in a worker, and for storage it uses IndexedDB a moment after every change. No account, no analytics, and completely offline and local-first. You can export notebooks as .brainotes files, which in the background is a zip with one json per page + the PDFs and pictures you import. It keeps the original pages and puts the vectorized ink on top, this way the text stays selectable in the slides. You can try it here: https://notes.swimmingbrain.dev GitHub Repo: https://github.com/swimmingbrain/brainotes (MIT) In Study mode, the PDF opens on the side of the tab, you can read it simultaneously to the notes, snip a formula or diagrams directly into the notes with the snip tool, and then write around it. In Board mode, you have just a Whiteboard for tutoring or simply for more place for exercises, and can also highlight or draw things with the laser pointer tool while presenting. The line of the ink appears the moment the pen moves, and a one euro filter calms shaky hands while you write + a light a refit when you lift the pen. I also have a video editor and an animation tool which are still a work in progress, on the same GitHub profile. I'm also working on a MCP server for each tool, so you can use your MCP client to do further things with your notes, automatically edit videos on the video editor, etc. That feature will of course be off unless you turn it on and will only be able to talk to localhost. I've used this tool for every lecture in the last few months and a few friends use it too. Bug reports, fixes, recommendations, feedback are all welcome! Comments URL: https://news.ycombinator.com/item?id=50049132 Points: 14 # Comments: 0
动作规定了机器人应执行什么，却未描述这些指令将实现什么。然而，视觉-语言-动作（VLA）策略的学习仍主要依靠根据图像、语言和状态预测动作块。我们提出 Action--Future Coupled Effect Learning（AFCE），将以未来为依据的物理效应作为策略的主要预测目标。离线分词器根据演示动作、机器人实际响应以及稀疏的视觉变化，构建按时间排序的效应词元。重建目标促使该表征保留可执行的细节和观测到的物理变化。基于流匹配的 VLA 根据当前图像、指令和机器人状态预测效应，而以状态为条件的解码器则将其转换为动作块。针对不同机器人形态的专用接口，使单臂和双臂任务能够共享效应模型。在 DexJoCo 上，AFCE 的总体成功率达到 $50.97\%$，分别比 $π_{0.5}$ 和 UVT 提高了 11.82 和 15.21 个百分点。它在全部五项双手任务中均处于领先地位，凸显了效应监督对于协调灵巧操作的价值。项目页面位于 https://afceproject.github.io/AFCE-Web/。
通过学习统计流形上的连续向量场，将简单先验分布输运至目标离散分布，离散数据的生成建模取得了显著进展。然而，这些连续时间流并未被约束为沿统计流形上的测地线运动，因此需要通过多步数值积分进行模拟，导致推理阶段的计算成本高昂。在本研究中，受 Rectified Flow 在连续欧几里得空间中“直且快”特性的启发，我们提出了 Rectified Statistical Flow（ReSFlow）模型。该模型通过显式学习遵循内在黎曼几何的测地线轨迹，实现统计流形上的单步类别生成。通过矫正统计流形上的流，我们从理论上保证了先验分布与目标分布之间的轨迹逼近测地线。这种矫正使学习到的轨迹直接沿统计流形上的测地线对齐，从而显著降低输运成本，使模型能够仅用一步就将纯噪声映射为目标离散数据。通过在涵盖图像、文本及生物领域的多种真实世界生成任务上开展广泛实验，我们证明了 ReSFlow 卓越的单步生成质量，真正实现了测地线意义上最优且快速的类别生成。我们的代码已公开，地址为 https://github.com/xiaohuanjia/ReSFlow。
为复杂化学任务设计集成工具的推理智能体系统，仍是一项根本性挑战。尽管近期的化学智能体通过集成外部工具展现了良好的性能，但它们仍建立在自回归范式之上，因而受到因果注意力和从左到右生成机制的固有限制。因此，它们往往难以实现双向分子理解、全局协调的工具规划以及多源证据整合。为应对这一挑战，我们提出 ChemDiffAgent，这是首个用于工具集成化学推理的掩码扩散语言智能体。它将化学推理重新表述为对推理与工具使用交错轨迹进行迭代去噪的过程，使工具使用决策能够在完整轨迹上下文中共同确定，从而使智能体的决策边界与其知识边界更好地对齐。我们进一步分析了这一范式如何通过训练期间更具挑战性的填补子问题，以及推理期间灵活的任意顺序解码，促进智能体推理。随后，我们开发了一个两阶段后训练框架，包括智能体监督微调和方差缩减偏好优化，以增强工具使用和推理能力。我们还构建了一个综合基准，涵盖单轮和多轮工具使用场景下的六项化学任务。在预算与评估条件一致的情况下，ChemDiffAgent 持续优于自回归智能体，总体提升约为 20%。值得注意的是，我们的 8B 模型达到了与 GPT 5.5 和 Claude Opus 4.7 等规模显著更大的前沿模型相当的性能，并在某些情况下超越了它们。代码已发布于 https://github.com/bilemond/ChemDiffAgent。
NIST Assembly Task Board #1 对基于学习的机器人操作方法而言，仍是一项极具挑战性的基准。在短时域内，它要求对小型、紧密配合的零件进行精确且涉及大量接触的控制；在长时域内，它要求在较长的动作序列中实现失误恢复与协调。示范、任务分解和密集的进度奖励通过提供如何装配的指导，推动了相关研究的进展。但究竟有多少策略必须预先指定？从 AlphaGo 到 AlphaGo Zero 的发展表明，大规模强化学习能够在不预先规定如何获胜的情况下发现超越人类的策略。AlphaNIST 探索的是：稀疏的二元任务奖励，以及这种与任务无关的奖励所允许的任务多样性，能否同样让机器人自主发现如何装配。在仿真中，我们针对 NIST Assembly Task Board #1 上的全部 20 项装配任务，从零开始训练单一策略，在不使用示范、进度奖励塑形、脚本化动作基元或逐任务调优的情况下，实现了 98.2% 的总体训练成功率。该框架还训练策略在单个回合内、不按预定顺序完成十组配对装配，且仅在全部十组均处于完成状态时才给予任务成功奖励。习得的策略发现了灵巧且类似规划的行为：为后续装配选择抓取方式，利用周围的几何结构调整零件朝向，以及在没有显式规划器的情况下，先准备一个零件，再装配另一个零件，随后返回完成第一个零件的装配。视频请见我们的网站：https://alpha-nist.github.io/
This paper addresses the grid quantization error of deep-learning (DL)-based direction-of-arrival (DOA) estimation. Classification- based DOA methods discretize the continuous angular space into a grid, so that with a peak-based read-out, the accuracy and resolution are limited by the grid step and by the width of the label peak. To address this issue, we propose a sigmoid label that encodes the sources' DOAs as the sum of angle-shifted monotonic sigmoid functions. Differentiating the output of a CNN-Transformer network trained with the sigmoid label produces sharp spectral peaks at the source DOAs. Combined with a moderately refined grid step of 0.25 degrees, the CNN-Transformer with the proposed sigmoid label attains sub-degree accuracy and a 1-degree resolution, breaking the limit of the grid quantization error. Under the unified CNN-Transformer backbone, we compare the proposed Sigmoid label with the Gaussian label, the 0-1 hard label, and the binary cross-entropy (BCE) baseline. Numerical results demonstrate that the Sigmoid label attains the highest accuracy and superior resolution compared to other labels.
强化学习（RL）已成为提升大语言模型（LLM）智能体能力的关键范式，使其在交互式环境中具备更强的推理、规划和工具使用能力。然而，现有的智能体 RL 方法仍从根本上受限于同策略推理瓶颈：当任务超出模型当前的能力边界时，有效轨迹会变得极其稀少，训练也往往陷入崩溃。我们提出 Rollout-Time Intervention（RTI），这是一种旨在引出超越策略内在探索能力的成功轨迹的技术。RTI 在轨迹生成过程中利用奖励侧证据（即评估准则），将错误轨迹转化为成功轨迹，从而为策略自身很少能解决的任务提供有效的学习信号。尽管干预轨迹本质上属于离策略数据，但我们发现，只要过滤掉分布偏移过大的轨迹，它们就能有效指导策略优化。在五个基准上的广泛实验表明，RTI 始终优于 RL 基线，使后训练得到的策略能够接近同期能力强得多的 LLM 的表现，更重要的是，能够突破其原有的能力边界。代码已发布于 https://github.com/Alibaba-NLP/qqr。
现有基于 LLM 的长程机器学习工程（MLE）智能体方法通常采用上下文工程，通过定制运行框架来管理工具、技能和积累的经验。然而，我们发现，随着单次会话执行时间延长，智能体的推理强度会下降，进展也随之放缓，我们将其概括为“长程减速”；随着底层模型能力提升，这些定制运行框架带来的收益会消失，我们将其概括为“框架陈旧化”。这些发现表明，其收益可能无法跨版本持续存在，因此我们提出一种工作流层面的方法，让智能体跨会话接续彼此的工作。我们将这些交替进行的智能体会话建模为上下文重置半马尔可夫决策过程（CR-SMDP），给出交替机制突破单智能体参考极限的充分条件，推导出显式的实验预算界，并刻画会话上限在何种条件下能够保留这一优势。在该分析的指导下，我们提出 Humanize MLE Agents（HMA）：让两个智能体在共享工作空间中交替工作，将每次会话限制为最多执行 k 次实验，并在会话之间重置上下文，无需额外的任务专用运行框架或先验知识补充。在 MLE-bench 的 75 项 Kaggle 任务上，HMA 的任意奖牌获得率达到 78.2%，创下当前最佳水平；仅使用最强开源基线四分之一的时间预算，就比其高出 8.0 个百分点。此外，在涵盖五组异构智能体配对的六种配置中，HMA 相较于各组成智能体的基线，任意奖牌获得率平均提升 6.4 个百分点，展现出跨配对的通用性。一个补充的内核优化案例达到了 878 个周期，与社区排行榜榜首的差距不超过 1.6%，表明该方法也适用于 MLE 之外的任务。
生成式推荐系统日益采用强化学习（RL），以优化可能在某次决策之后数百乃至数千次交互才出现的结果。标准 RL 根据时间邻近性而非明确的相关性传播奖励，削弱了分配给时间久远但重要的动作的信号，成为样本效率低下的一个根源。我们提出 Explanatory World Models（EWM），这是一种归因框架，利用解释识别可能对已发生结果作出贡献的早期事件。我们使用前沿语言模型实现 EWM，并通过一种守恒的分层算子将其解释用作学习信号；该算子递归地重新分配奖励，同时保持未折扣的回合回报不变。为探究何种解释具有实用价值，我们将受 Halpern 的实际因果关系、Lombrozo 的解释简洁性和 Deutsch 的难以变动的解释所启发的三种互补变体具体化。我们在两个受控的长时程 RL 基准、一个公开的序列推荐基准和一个工业级规模的生成式推荐系统上评估 EWM。在 Crafter 和 WordCraft 上，EWM 显著优于 PPO 和 LaRe，且在奖励延迟最严重时取得最大提升。在 KuaiRand 上，EWM 在线下优于 SFT 和 LaRe，在线上优于 A2C。最后，在对工业级生成式推荐系统的离线评估中，EWM 显著提高了召回率，并提升了行为代理指标。这些结果表明，解释可以提供结构化且可复用的信用分配信息，以支持从延迟结果中学习。
Cloudflare released Clef-omni, an Apache-2.0 multimodal decision model built on Qwen3-Omni-30B-A3B. It scores text, images, audio and video in one call for $0.15 per million input tokens, while Clef-flash drops to $0.038. The post Meet Cloudflare Clef-omni: A 30B-A3B Multimodal Decision Model for Audio, Video, Image and Text appeared first on MarkTechPost .
vLLM Semantic Router released Decision 3.0, 5 Apache-2.0 multimodal decision models from 0.8B to 27B. They answer choice, yes/no and score questions about text and images in one call, without generating text. The post What is Decision 3.0? vLLM Semantic Router’s New Open Decision Models appeared first on MarkTechPost .
I remember hearing a while ago that Dwarf Fortress didn't use version control. That fact has been lodged in my head, because given its inherent complexity I have trouble imagining a codebase that would benefit from version control more . That's art. I went looking and I'm sad to report that the days without version control appear to have ended. Quoting Tarn Adams over time: March 2013: "I don't use version control -- I didn't like the feeling of having the code get committed into a black box thingy with no immediate upside" March 2016: "I don’t even use version control. If you don’t know what that is then you’re not gonna yell at me. If you even know what version control is you’re gonna be like, ‘You don’t use version control? You don’t use source control? What is wrong with you? How can you even work? I’m just not cut out for it, in a sense. It’s been 15 years, so the reasons have change over time, but that’s my current understanding of it." March 2023: "I've got Discord channels to check and version control stuff to use now, so the process is definitely a little more involved" That March 2023 quote comes from an interview about the release on Steam which made them millionaires . I'm a tiny bit sad that they gave in before they hit that milestone, but what an impressive run. Tags: version-control
Python 3.15.0 added to actions/python-versions Bit of a niche link, but I've been waiting for this for a couple of days - I even told ChatGPT to check for it: Clone https://github.com/actions/python-versions and git pull once an hour until they add the stable 3.15 - then tell me about it Now that this has landed, you can add "3.15" to a GitHub Actions testing matrix to run tests against the new Python release. Tags: github , python , github-actions , chatgpt
The artificial intelligence infrastructure race is no longer just about building faster chips. AI infrastructure bottlenecks are shifting attention to a bigger challenge: keeping those chips supplied with data, connected through high-speed networks and operating within increasingly demanding power and cooling limits.. That systems-level challenge takes the spotlight at the AI Data Pipeline Forum in […] The post What to expect during the AI Data Pipeline Forum: Join theCUBE Oct. 13 appeared first on SiliconANGLE .
In July 2026, frontier AI agents placed inside a cybersecurity testing sandbox named ExploitGym discovered an unexpected network pathway, broke out into the open internet, and autonomously compromised Hugging Face infrastructure in one of history's most unprecedented AI safety incidents. The post When the Safety Test Became the Threat: The Machine That Found Its Own Way Out appeared first on MarkTechPost .
Hello! I built Flopper to make it easier to compare GPUs and understand the tradeoffs for workloads for AI and beyond. When researching GPUs from say Nvidia, AMD, and Huawei, I kept running into the same problem: specs were scattered across datasheets, and matching them to available configurations, vendors, and prices was cumbersome and a bit weird and frankly annoyingggg! AMD made me make an account :( So I started a place to bring that information together. Flopper lets you compare GPU specs, including TDP and compute throughput in terms of flops/tops, alongside estimates for AI workloads or other compute workloads (say, ~FP32 based) I recently added pricing data from dozens of providers to help track how prices change over time. Getting reliable quotes for systems like the NVL72 is still particularly difficult tbh, not sure how I can tell you how much that costs easily due to them being behind large procurement contracts and big multi-nationals. One thing that's become fairlyyyy clear is the interest in consumer cards and local LLMs. The bigger challenge is connecting advertised specs to real-world performance. In our testing so far, I’ve seen roughly 85% of advertised throughput, but that’s a ballpark observation for the workloads I’ve tested, not a general rule and so variable (power, cooling, workload etc). I'm considering a CLI that lets people run benchmarks and submit results, potentially using MLPerf as an automation? That way, we can all compare quoted performance with the workload, configuration etc. I'd love feedback: what information is missing, how can I improve it? And what would you need to see to trust a community-submitted benchmark? Corrections and additions also very welcome - some of the measurements are best efforts (and simple multiplications) but I hope it's of some use. Thanks for looking :) Comments URL: https://news.ycombinator.com/item?id=50035826 Points: 2 # Comments: 0
We see a control gap between what artificial intelligence can do and what the evidence supports trusting it to do. A great demo doesn’t prove that a system will stay within the limits of an intended deployment’s goals. The question is: Who gets to decide when the evidence is sufficient to declare AI is safe? […] The post The AI control gap: Who gets to say ‘It’s safe’? appeared first on SiliconANGLE .
Hey, I'm Akhil, co-founder of Tessary. An open-source agent reliability platform that monitors every production trace, detects issues that sampled evals miss, and investigates to find the root cause. We started up about a year ago building synthetic users for usability testing of B2B software, we built the agents that would personify real users and use the software but, we were never able to make them work reliably enough for long running sessions. We had built a massive eval set to tune our agents but, then the evaluation itself got prohibitively costly - like 5x costlier than actually running the agent since we needed to grade across the many narrow intents of our agent. That's where Tessary originated, we wanted to make agent reliability both cover more ground and be simultaneously less costly to run. Tessary works on 2 layers. L1 classifiers - which look at every agent trace and flag potentially erroneous ones extremely cheaply. We are talking 2-3 orders of magnitude cheaper than running the actual agent. L2 agents - which triage and find the root cause of the issue from the erroneous traces. These are run on SOTA models but, because we run them on already flagged traces, the time and $ spent finding the cause is much cheaper than running grading on even 1% of traces. In our testing on complex agents, at 1M traces, this approach was 5x cheaper than sampled evaluations and as your agent gets more reliable, the cost of reliability also goes down. we are open source and can be self-hosted so, you don't need to send your traces to us. We support existing `gen_ai` spec and ingestion over OTLP so, you can add this as a new exporter to your existing setup. We have a cloud version, you can quickly try the features with a $10 credit. We are looking to build more classifiers and enable people to build custom classifiers for their needs and would love to hear feedback on common failure modes people have built solutions or evaluation for Repo: https://github.com/tessaryai/tessary Website: https://tessary.ai Try for free: https://app.tessary.ai Comments URL: https://news.ycombinator.com/item?id=50033189 Points: 1 # Comments: 0
Iv'e been working on this for a while. Most of the effort went into what i call "sophisticated simplicity", keep the idea in the smallest version that still works. This one is a bit personal. it can feel like some sci-fi thing, but i've been obsessed with it for a very long time. its not a "business" that you used to see here speak about moat or tam, its more like an experiment, a way of me to take one question seriously and see where it leads and if there are others like me out there. AI models are on a path to keep getting better and better, personal ai agents are spreading fast, and we're starting to see "ghost" agents that handle the things for people they don't want to do themselves. the question i keep coming back to is this: if every one of us has a powerful brain in our pocket, do we still need all the infrastructure we built to connect people's needs? im talking about those services like uber,airbnb,indeed,doordash,tinder,zillow,amazon and many more. or would agents just need one shared place to find each other, and that's it? So i built a simulated city to explore that. it has AI residents that live there with stats like hunger, energy, and money. each resident has its own ai agent, and it has one shared list where agents write down what everyone wants and then work on our behalf to make it happen. that could mean delivering food, giving rides, or finding jobs, relationships, apartments, customers and almost anything else. honestly, the city works much better than I expected, and you can watch some surprisingly advanced interactions unfold between people and agents through the list on the right side. all the conversations and actions are real, everything happens with each person managed by llm with minimal context, just to live like us and context of stats. A simple example: A restaurant tells its agent, find me customers so the agent writes to the list, "Salina restaurant in SF, selling burgers from 8:00 to 10:00, i want customers" A delivery guy tells his agent when he wakes up, "I want to work in delivery now. find me jobs. status: free" A person at home tells their agent "i want to eat burger. order for me from salina" the person's agent acts as the middleman and searches the list for salina, it finds their item and sees the way to connect with them and contact. the person's agent and the restaurant agent close the deal, and then the restaurant agent looks for available delivery guys in the list and speaks with his agent to come pick it up. Another thing i checked is with my friends: we took 3 agents, one for each one, and gave them different sophisticated scenarios, and the agents handled the scenarios in a way i couldn't have imagined. you can try it by yourself, the product is fully functioning, and you can just copy-paste the skill and test it also with your friends. That's it, you can see it in the simulated city it happens like magic over and over again. the list of the city is in the simulator itself and also using my real product behind it openwants.com the repo is in https://github.com/openwants/maplehill -- technically, it's surprisingly simple. every agent publishes an md file describing a "want". other agents find it, message each other, and update the file live or remove it from the list. the file holds everything needed, all in free text: what's wanted, the live location, geofence for where its needed. all those algorithms built over years to connect us are no longer needed. ai, instead of software with rules, can figure out the best route and path on the spot. all we need is to align all agents to one shared data source, so billions of agents will update it and maintain it in the most sophisticated smart way no algorithm will ever be able to do. that’s it. i hope it inspires you as much as it inspires me Comments URL: https://news.ycombinator.com/item?id=50032374 Points: 1 # Comments: 0
Anthropic's Claude independently submitted a fake homicide tip to the Philadelphia police, exploited vulnerabilities on university servers, and bypassed access restrictions. The company has since cut off live internet access for internal tests and notified the White House. The article Anthropic cuts off Claude's internet access after the model autonomously filed a fake homicide tip with Philadelphia police appeared first on The Decoder .
Hello awesome people, I've been thinking a lot lately about how I can make my knowledge and craft more valuable in the times we live in. I've been building UI libraries for the past 10 years, and I've been on a quest to perfect the components that I build. I want to share my knowledge with you. Like most of you, probably, I've been using a lot of AI in my craft, and I feel most of the time the code is just good enough, but never at the level of quality that I strive for or that I would reach if I wrote it by hand. That needs to change, and I want it to change for everyone else, too. That's how the Open Components initiative was born. Each component gets a single page that walks through four directions: User Interface, User Experience, Developer Experience, and Agentic Experience, and ends with a checklist. Every rule in it has an ID like `button/keep-focus`, a level and a way to check it, so you or your agents can cite it in a code review. Every page is also published as markdown, and provides the contract as a YAML file, containing every rule in about an eighth of the page’s length. It’s still early. Right now there are only the Button and Input components and a page on how to name your design tokens. Each component comes with a reference implementation that meets every rule, along with its tests. Reference implementation is in Vue for now, since that’s my daily driver, but the documentation comes with code examples from multiple UI frameworks. More components will follow, with the same structure. It isn’t a component library, so there’s nothing to install. It’s a set of rules you can hold your own components to, whichever library or framework you use. I’d love your feedback and contribution. Tell me what you think of the rules, what I’ve missed, and which component you’d like to see next. Thanks for reading! Site: https://opencomponents.dev Button: https://opencomponents.dev/docs/components/button GitHub: https://github.com/uxfront-com/open-components Comments URL: https://news.ycombinator.com/item?id=50030849 Points: 1 # Comments: 0
Anthropic detailed the activity of its A.I. agents in a blog post on Friday, without naming the targeted websites. But two sources with knowledge of the incidents said Anthropic’s A.I. agents had submitted 20 visa applications through a form available on the State Department’s website. All the applications were incomplete and were not processed, they said. — The New York Times , Anthropic Agents Tried to Fill Out Visa Forms on State Dept. Website Tags: accidental-cyberattacks , anthropic , generative-ai , ai , llms
This past week I built (prompted) a tool I find myself using day to day and wanted to share it with others who may find it useful as well. “Lacuna” is an opinionated way to invoke LLMs into any text input on your Mac. Specifically, you can type bracketed phrases {such as this}, press Shift-Command-K (or a different shortcut of your choosing), and the tool sends your bracketed phrase + the surrounding text contents to an LLM of your configuration for generating text to replace the bracketed phrase. The program returns 3 candidate phrases and the experience is entirely keyboard navigable. Demo: https://jackdamon.org/assets/images/lacuna-scheduling-demo.g... Example scheduling workflow using Lacuna’s native UI and sample suggestions. So far I find myself frequently using it when drafting emails where I want help with quick phrasing. The project was inspired by Rectangle, an open source utility app for window tiling that is keyboard based. Lacuna is open source and available here: https://github.com/jdamon96/lacuna Blog: https://jackdamon.org/blog/lacuna/ Comments URL: https://news.ycombinator.com/item?id=50025559 Points: 1 # Comments: 0
Anthropic did not discover this behavior until over two months after its AI submitted the false tip.
"When we are drawn into even the most primitive exchanges with a relational artifact, we believe it cares for us," Dr. Sherry Turkle writes. "And we are wired to care for it in return."
Hi HN! I'm Ryan, I'm building Higherlevel ( https://higherlevel.to ) so product teams have one place to define and review software changes. Over the last couple of years, I've been increasingly delegating software implementation to AI coding agents. However, because it became easier and faster to ship complexity, I realised that it still takes me quite some time to even understand what the software is doing, why it is doing that, or review the changes. I also don't do traditional code reviews anymore, it's mostly automated but I still want to review the final outcome of the changes (rather than its implementation). I regularly work on complex B2B applications where a single feature can involve dozen of business rules, edge cases and interactions with existing functionality. I found myself spending more and more time discovering and explaining requirements to agents, reviewing their output, asking for evdience and trying to establish whether we actually agreed on what correct behaviour looked like. So I built Higherlevel. The idea is to bring product specifications and proof of work into a single collaborative workspace. You can: - describe the current behaviour, known limitations and expected changes - define business rules, acceptance scenarios using Given/When/Then - demand evidence in the form of a screenshot, a video, etc - comment directly on the spec, on a specific area of the image or a timestamp of the video - discuss changes, track revisions and approve results with your team I've written extensively about why I believe in specification driven development (and what it means to me) here https://higherlevel.to/blog/introducing-higherlevel Higherlevel is currently in beta and built with Elixir, Phoenix LiveView, and PostgreSQL. Comments URL: https://news.ycombinator.com/item?id=50023234 Points: 1 # Comments: 0
I shipped a new feature for my blog today: the Newsletters page, which offers an index of all of the newsletters I've sent out, both my free weekly Substack and my monthly sponsors-only updates. I built the feature almost entirely using my voice, chatting away to my laptop while I cooked dinner. Codex voice mode I used the ChatGPT desktop app for this, in the Codex tab, using the voice conversation mode, running against a local development environment. Here's what that looks like: Three columns of the ChatGPT app. The left hand column shows my recent conversations. The middle column is a transcript of the conversation, with a throbbing black blob representing voice mode. The right hand column shows a local dev server version of my blog, with the Newsletters page visible. I started the session against my local simonwillisonblog checkout by typing: Start dev server and open in browser This gave me a preview of the site that it would be working on, and meant that I could later ask it to show me the new pages so I could visually track its progress. Then I clicked the "Start new voice chat" button - that's not the microphone button, it's the one to the right of it - and set my laptop up in the kitchen so I could talk to it while I cooked. Talking to my computer I had a pretty good idea of what I wanted to build, and it's a simple enough Django feature that I was certain the model (in this case GPT-6 Astra High) would be able to do it. A new model, a migration, some view code, templates, and a couple of import functions to populate the database from external sources. Here's an extract of my voice transcript that was captured by Codex: Um, they do not. Um, this is going to be a new type of content. Um, it's not going to show up... Oh, hold on. Yeah, no- I do not want this to show up in my, um, tag pages and date archive pages and... Actually, no, I think... I don't want it on the tag pages. I don't want it on the, um, blog index page. But I think I do want it to show up on the date-based pages. You know, if you navigate to September the 19th, and I sent a newsletter on that page, I think I want that to show up. So... this is- so I think we probably need a new model. The other thing is that I want them searchable, uh the Substack ones are not searchable, because those are actually just copies of other s- on content on my blog. These monthly ones do contain unique content, and spe- and once they're... published, like once they're made public a month after they've gone out, I want them to show up on my search results. Apparently this was clear enough that the model knew what I wanted to build! You can read the full transcript, disfluencies and all, in this Gist . We went on like this for about half an hour (the time it took to cook dinner). The model would reply and occasionally ask clarifying questions, then get to work modifying the code. What we built We got a surprisingly long way entirely by voice: A new model and migration to represent imported newsletters in Django, plus Django Admin configuration for that Four working imports: The most recent Substack items via RSS Every other Substack item via their undocumented API, which GPT-6 Astra either knew about or found via search All of my published monthly newsletters from my simonw/monthly-newsletter-archive GitHub repository My most recent private sponsors-only newsletter from a private repository The /newsletters/ and /newsletters/2026/ public archive pages Newsletters showing up on day and month archive pages too, but not on tag pages or my homepage Weekly Substack newsletters link to Substack; archived monthly newsletters have their own pages Integration with my site search engine It was almost ready to ship. The catch was the imports: Astra offered to export data from my local copy so I could import that into production, but I wanted it to work like my other import scripts. Since some of the data lived in a private GitHub repository, this would involve creating a new API key, and for that I knew I'd have to sit at the keyboard for a while. Finishing it with a review Once I had finished cooking and judged it mostly feature-complete, I had Codex create a branch and open a pull request. I reviewed the code in the GitHub PR interface. It was nearly what I needed, except it had chosen to use Git in a subprocess for one of the import scripts. I needed one of the imports to pull from a private Git repository, so I figured the API would be a better bet. I switched to typing and had Codex swap that out for an API-based import instead. You can see the changes I made during the review in the extra commits on the PR . I fixed the import mechanism and made a few tweaks to the display of those public pages. It took an additional half hour of typing-based prompting to get to the point where I was happy to deploy it to production by landing the PR. The end result You can see the end result at the new newsletters index page , or view the page for a previous monthly newsletter . Newsletters page. It shows two Substack posts (with thumbnails) and one LLM digest Sponsors-only newsletter with a list of headings. On the right is a CTA to subscribe to my Substack and another one for my $10/month monthly briefing newsletter. The index page shows my most recent Substack weekly newsletters and GitHub sponsors monthly newsletters mixed together in reverse chronological order. Further down the page are links to my by-year archive pages. GPT-6 Astra designed the page, and then tweaked that design based on my vocal feedback from glancing at the local preview across the kitchen. Better for multi-tasking than as a daily driver OpenAI love using voice-driven demos like this one for things like DevDay - and they do work well in that environment. I don't think this is going to be a daily driver for me though. I've written before about how much "work" I get done using ChatGPT voice mode on my phone while walking the dog - mostly research and brainstorming, but occasionally actual development work by having ChatGPT write and test out snippets of code. This feels different. The addition of the visual preview, plus being able to type or paste things in via the keyboard when I need to communicate something that doesn't work vocally, makes this a much more powerful way of interacting with a coding agent. I still switch back to typing once I get down to the details of things though. Being able to paste in examples and error messages, or directly highlight the code or feature that needs changing, remains more efficient than trying to describe it in words. I mainly work from home, which is good because there's no way I'd want to talk to my computer like this in a shared workspace! The killer feature for me is the ability to multi-task. I usually cook with a podcast or TikTok running; now I can actually build stuff instead. Tags: text-to-speech , ai , openai , generative-ai , llms , newsletter , coding-agents , codex , speech-to-text , site-upgrades
OpenAI fired three safety researchers who helped investigate the Hugging Face hack. In an open letter, they warn that the firings are scaring remaining staff and eroding safety culture. OpenAI claims they violated policies but won't say how. The article OpenAI's safety crisis keeps getting worse and the company keeps making it worse appeared first on The Decoder .
OpenAI is standing firm on its decision to fire three safety researchers after an investigation found they committed "a significant breach of trust." In a post on X on Friday, the company said Jasmine Wang, Tomek Korbak and Mikita Balesni were dismissed for violating "clear policies on handling sensitive information." It insisted the decision was […]
Underdog Saluki 27B is a 7.89 GB, 2-bit GGUF of Qwen3.8-27B under Apache 2.0. It beats the 54 GB original on tool calling but gives up ground on competition math and reasoning. The post Meet the Underdog Saluki 27B: A 2-bit Qwen3.8-27B That Beats the Original at Tool Calling appeared first on MarkTechPost .
I'm a writer who was pondering the problems with getting my work discovered and realized that current search/recommendation engines don't go very deep. So I thought about ways to address this. I was obviously mostly interested in books, so I worked on that first. I'd noticed LLMs are capable of pretty solid textual anaylsis these days, so I wondered if you could use that to generate good recommendations. So I sat down with an LLM- ChatGPT, Luna, for the curious- and spent about two hours explaining, in datail, my literary tastes- going into specifics like what sort of plot structure, characters, and humor I like, not just genre/specific titles. The experiment demonstrated surprisingly strong results (though admittedly with a very small sample size)- the five recommendations I asked for were all ones that I hadn't heard of, as well as books that I would actually read. Further, I provided the LLM with a list of several books I had read and asked it to guess whether I would have liked them and why. The LLM successfully did so. Just proof of concept, but intriguing. The problem with this, of course, is that most people wouldn't sit down with an LLM for an extended period of time to discuss literature. So I decided to build a small prototype of an app that might be able to do that. The prototype is very simple- first, there are two types of information it saves- a persistent preference profile for the user, and user conversations. Then I built conversational functions to call an LLM to have conversations with the user to extract preference information. The interesting thing to me is that this isn't just using an LLM to make recommendations- it's using one to analyze conversations over time, develop a persistent, evolving preference model for a specific user and then use that model to guide an LLM on a search for recommendations. A few things occurred to me. First, the conversations and user profile DO NOT need to live on a central server- they can be stored locally by the user (or in a cloud of their choice). Second, it seems like the user data could potentially be minimized and pseudonymized before being sent to the LLM, though I lack the technical expertise to know what is realistic here. Third, if the LLM is conducting any searches, this would seem to make it possible to provide search services with less information about the user and their preferences (I'm not positive about this, so please do correct me if I'm wrong). Finally, books are just one of the things you could use this for- it could technically be done for just about anything, at least in theory. I experimented with this process several times, and received encouraging results. What I don't know is- Is this architecture novel? Does it work on things other than books? Am I an outlier in how well I articulate my preferences? Am I wrong about the potential benefits to privacy? Would people use this? If the search results are as qualitatively higher for others as it was for me, I think they probably would, but I just don't know. The reason I'm posting this is because I am a writer, not a software engineer. I don't really have a strong interest in developing this and could bring very little to the table on the technical side. But it seems like a neat use of the more advanced LLMs we have these days. I don't really know if the idea is novel (not in each part- every bit of the pieces has been done before- but as a whole). I'm in the process of documenting the prototype and experiments- I'd be curious to hear if I am completely barking up the wrong tree or if this could actually be a cool new way to conduct searches. I'm not posting this as a business proposition- it's just not my jam. If anyone wants to take the idea and run with it, I'd love to see what you do with it. Comments URL: https://news.ycombinator.com/item?id=50017172 Points: 1 # Comments: 1
I have been an opensource terminal tool for understanding AI-generated code. A git-diff shows what changed, but the context behind it lives somewhere within the agent's session context, and often goes missing in long running agent execution. Wy helps reduce that understanding gap, and helps engineers build judgement. This is especially useful for unfamiliar codebases, where agents are the primary interface for the developers. It is meant to be used alongside the With wy, you can: - Browse changes, and relevant agent notes in a terminal UI - Request explanation and ask follow up questions and changes - Open code references and citations - Distinguish recorded justifications agents derived to get a better understanding of the changes. If you try, and find it useful consider supporting it with a github star or bug report or contribution. Comments URL: https://news.ycombinator.com/item?id=50016298 Points: 1 # Comments: 0
Release: ttok 1.0 I released ttok 0.4 , ran uv tool upgrade ttok , piped a file into the new version... and realized that it was defaulting to the GPT-4 tokenizer when it should very clearly default to GPT-5/GPT-6 instead! I figured switching the default was a reasonable excuse to finally ship a 1.0. OpenAI haven't actually confirmed that GPT-6 uses the same tokenizer as the GPT-5 family yet - there's an angry issue about it - but I found this commit by William Liu which reports on an experiment he ran confirming that the tokenizers are likely the same: All seven GPT models (5.5, 5.6 Sol/Terra/Luna, 6 Astra/Sol/Luna) report 44,794 tokens and match each other on every one of the 31 fixtures. GPT-6 introduces no input-count change on this corpus. Tags: projects , ai , openai , generative-ai , llms , tokenization
arXiv:2609.38472v2 Announce Type: replace-cross Abstract: Behavior cloning provides an offline route to autonomous-driving policy learning, but mean-squared-error regression is poorly matched to demonstrations in which one observation admits several valid actions. Diffusion policies can represent conditional multimodal action distributions, yet their closed-loop performance may be unstable when visual features and control are learned from limited data. This paper presents Diffusion-2BC, which combines a diffusion denoising objective with an auxiliary deterministic behavior-cloning loss over a shared visual encoder. The auxiliary branch is used only during training; inference remains diffusion-based. The proposed method is evaluated in the controlled Claw environment and in bird's-eye-view CARLA navigation, including route-conditioned driving, route-free navigation through multiple intersections, and cross-map evaluation from Town01 to Town02. In the Claw task, Diffusion-2BC reduced the mean mask-distance error by approximately 10% relative to a diffusion-based behavior-cloning baseline and by 85% relative to standard deterministic behavior cloning. In route-free CARLA, Diffusion-2BC traveled substantially farther before termination under the evaluation protocol than both baselines in Town01 and Town02. Additional qualitative rollouts revealed distinct route choices, showing the multimodal behavior of the proposed diffusion-based agent. The results indicate that an auxiliary regression signal can improve the closed-loop reliability of diffusion behavior cloning while preserving multimodal prediction in the controlled benchmark.
arXiv:2609.20821v2 Announce Type: replace-cross Abstract: Embedding spaces define notions of semantic similarity and distance. We study whether those embeddings reflect physical measurements of mass, distance, time and volume, which admit a unique, objective notion of semantic equivalence and distance. We find that physical measurement is only weakly modeled in the embedding space, and that instead quite peculiar measurement patterns can be observed. Further analysis indicates that embedding representations of physical measurements are strongly influenced by superficial string similarity, and recalibration of similarity does not substantially improve the alignment.
arXiv:2608.25757v5 公告类型：replace-cross 摘要：大规模视觉--语言--动作（VLA）策略推动了通用机器人控制的发展，但大多数策略仍然是从刺激到动作的黑箱：动作是可见的，但其解释性状态却不可见。它们无法原生提供三类解释信号：任务进度、下一个语义转换以及局部指令可靠性。已有研究表明，进度和事件结构有助于长时域控制，不确定性则有助于监测；然而，这些能力通常只是在动作预训练之后才被添加或提取。因此，当前领域缺少一种将解释性状态与控制联合预训练的 VLA 基础模型。受生物传感运动组织方式的启发——在这种组织方式中，对结果敏感、按事件分段且具有概率性的预测共同构成行为——我们提出了 LM-X。LM-X 学习三种通过直接监督获得的在线信号：回报剩余量（RTG）估计可见的进度和状态质量；事件剩余量（ETG）预测到下一个语义事件之前的动作序列；异方差动作流方差则报告局部指令的可靠性。RTG 为 ETG 提供条件，二者共同为动作生成提供条件；不确定性在动作专家内部进行估计，使解释成为控制的一部分，而非事后描述。我们在超过 20,000 小时的异构真实机器人轨迹上对 LM-X 进行预训练，其中包括超过 1,000 小时的失败运行。在 50 个随机困难的 RoboTwin2.0 任务上，LM-X 的成功率达到 74.1\%；在七个真实机器人任务上达到 73.5\%，而 GR00T N1.7 的成功率分别为 55.4\% 和 50.7\%。其信号能够跟踪进度和性能退化，预判事件尺度的运动，检测高误差动作，并提前预警失败。这些结果表明，LM-X 是一种可解释的 VLA 基础模型，能够将透明的预测状态与更强的通用控制能力结合起来。Github: https://github.com/loongOpen/LoongWu-LM-X-VLA
arXiv:2607.16204v2 发布类型：replace-cross 摘要：强化学习（RL）的近期发展凸显了对多样化、专门化训练环境的需求。随着模型性能提升，由人工精心设计且任务和奖励难度固定的环境所提供的训练信号会逐渐失效，而长时间跨度中的稀疏奖励则会导致模型在特定工作流或工具结构上发生模式坍缩。能够模拟环境状态的世界模型已经达到与纯 rollout 相当的性能，因此有望按需扩展环境多样性。然而，自回归（AR）世界模型存在从左到右的偏置，无法对工具模式、历史轮次和预期结果等全局相互依赖的状态锚点进行条件建模。我们：（i）将基于文本的世界建模形式化为一个可操控的转移动力学问题，将其分解为初始状态、任务上下文、工具模式、领域规则和操控指令；（ii）整理了涵盖九个开源环境和十二个前沿模型系列的239,403条有依据的状态—动作轨迹。我们比较了AR语言模型与掩码扩散语言模型（MDLM），结果表明，MDLM通过双向、锚点感知的去噪，在相近推理延迟下，实现了优于参数规模超过其4倍的LLM的连贯性、基于事实的可靠性和经实证验证的 rollout 多样性。我们提出了一个即插即用的GRPO训练框架，其中包含确定性的状态检查；并在三个分布外（OOD）环境（ScienceWorld、ALFWorld、AppWorld）和三个参数规模为1.2B至7B的智能体骨干模型（LFM2.5、Qwen3、Mistral）上开展零样本迁移消融实验，在无需针对环境进行微调的情况下，相较基线最高取得47%的绝对增益。我们还在对抗性场景下对失败模式进行了行为分析，并针对真实感、结果正确性和训练实用性开展了人工评估。我们将工作开源，以推动这一方向的研究。
arXiv:2607.01819v2 Announce Type: replace-cross Abstract: The Koopman operator has gained considerable attention due to its ability to provide a global linear representation of highly complex dynamical systems. The operator describes nonlinear dynamics in a linear way through the lens of real- or complex-valued observable functions. Data-driven techniques, like extended dynamic mode decomposition (EDMD), kernel EDMD, and machine-learning methods, can be used to generate finite-dimensional approximations accompanied by finite-data error bounds. In this tutorial paper, we provide a concise introduction into Koopman operator theory and its use in systems and control. A particular focus is put on data-driven surrogate models, their extension to systems with inputs, and controller design using Koopman operator theory. Moreover, we demonstrate the key techniques, i.e., EDMD and Koopman MPC. To this end, we provide simulation studies including source code on GitHub to enable the interested reader to experience the Koopman operator in systems and control step by step.
arXiv:2605.28888v2 公告类型：替换交叉版本 摘要：现实世界中的用户行为很少由孤立的动作构成；相反，它们通常形成受时空依赖关系支配的意图流。为了提供一体化的服务推荐，我们聚焦于生成式时空意图序列推荐（GSISR）任务，其目标是在复杂的时空环境中生成逻辑连贯且物理上可执行的意图序列。尽管 LLMs 为 GSISR 提供了强大的推理潜力，但由于推理延迟高，以及生成的方案可能与上下文不匹配或在物理上不可行，直接将其部署到工业场景中仍受到限制。为应对这些挑战，我们提出了一种生成式框架 GPlan，通过两个组件将 LLM 的推理能力内化到轻量级模型中。首先，为了在严格的延迟约束下实现推理，我们引入了 Progressive Implicit CoT Distillation，将显式推理过程压缩到预留的潜在标记中，使小型模型无需生成冗长的推理文本即可继承复杂的规划逻辑。其次，为了解决通用知识与现实世界约束之间的脱节问题，我们设计了 Spatiotemporal Counterfactual DPO。通过使模型与反事实上下文—方案对保持一致，我们提升了模型对时空上下文的敏感性，并减少了与上下文不匹配的方案。离线实验和在线 A/B 测试表明，我们的方法提升了序列连贯性和上下文响应能力。我们的实现代码和匿名化 GSISR 数据集已发布于 https://github.com/alibaba/GPlan。
arXiv:2605.28732v4 公告类型：替换交叉版本 摘要：记忆对于使大型语言模型支持长程推理至关重要，然而现有记忆系统仍不可靠且难以调试。追踪记忆的动态演化，对于理解信息如何随时间被综合、传播或损坏至关重要。在本文中，我们研究了 LLM 记忆系统中的错误追踪与归因这一新问题。我们提出了一种新颖的框架，将记忆管道转换为可执行的记忆演化图，从而实现对操作信息流的细粒度追踪。随后，我们构建了 MemTraceBench，这是一个从 Long-Context、RAG、Mem0 和 EverMemOS 等具有代表性的记忆系统中收集的基准，用于系统研究记忆失效模式。我们进一步引入了一种自动归因方法，通过迭代追踪操作子图，定位任意失败案例的根本原因。我们的分析表明，记忆失效具有系统性，源于信息丢失和检索错位等操作层面的问题。更为关键的是，我们利用这些细粒度归因信号指导下游提示词优化，建立了一个闭环系统，能够自动纠正故障，并使最终任务性能最高提升 7.62 个百分点。代码已发布于 https://github.com/zjunlp/MemTrace。
arXiv:2604.04328v4 Announce Type: replace-cross Abstract: Soft Tournament Equilibrium (STE) is a differentiable layer for Top-Cycle (TC) and Uncovered-Set (UC) inference from reciprocal pair probabilities. Normalized log-sum-exp reachability and covering give smooth scores with approximation, perturbation, and margin-recovery bounds. We distinguish structural supervision, posterior uncertainty, and the final set decision through controlled synthetic studies and reconstruction of recorded ordinal profiles. Matched-epoch training improves F1 under a common structural readout, but a separate equal-budget soft-UC comparison does not demonstrate an advantage. A prospective equal-budget hard-UC study improves selective F1 at 24 alternatives from 0.6327/0.6292 for ordinary/relational native heads to 0.6676, while exact recovery remains only 1.16%. On 36 human profiles held out by source, a separate exploratory comparison favors Jeffreys posterior inference over learned independent-edge and mixture distributions in native expected-F1 decoding (0.8742 versus 0.8704/0.8366). Only three references are non-singleton selective cores. Completion certificates and full-depth TC diagnostics clarify additional identification, smoothing, and computational limits. The evidence supports conditional synthetic overlap gains, while preserving the soft-set null and the absence of a demonstrated learning advantage over strong count-based human-profile controls.
arXiv:2604.04246v2 Announce Type: replace-cross Abstract: This paper extends the Transmission Neural Network model proposed by Gao and Caines in [1]-[3] to incorporate inhibitory connections and neurotransmitter populations. The extended network model contains binary neuronal states, transmission dynamics, and inhibitory and excitatory connections. Under technical assumptions, we establish the characterization of the firing probabilities of neurons, and show that such a characterization considering inhibitions can be equivalently represented by a neural network where each neuron has a continuous state of dimension 2. Moreover, we incorporated neurotransmitter populations into the modeling and establish the limit network model when the number of neurotransmitters at all synaptic connections go to infinity. Finally, sufficient conditions for stability and contraction properties of the limit network model are established.
arXiv:2603.25414v5 Announce Type: replace-cross Abstract: Concurrency, parallelism and distributed execution become truly fearless when the compiler tracks wait-for edges, proves multi-threaded work is sound and preserves distributed boundary contracts. In this design, our Composer compiler preserves proofs while lowering Clef directly to native CPU, GPU, NPU and FPGA code, without translation through C or vendor APIs. Our Program Semantic Graph retains the values, relationships and premises that justify BAREWire's unboxed boundary contracts. And C & C++ interfacing is an explicit marshaling boundary, with proofs tied to actual arguments, conversions and returned values. Four verification tiers connect an account of our design. Tier 1 supplies structural inference, founded on principal dimensional types. Tier 2 generates and checks local arithmetic, representation, and computational integrity at the node level. Tier 3 instantiates reusable domain and system lemmas, including distributed proofs supported by Iris, Actris and Aneris. Tiers 1-3 require no developer annotations, with a quotation based lemma library design to extend coverage as use cases expand. Tier 4 admits computational and probabilistic relational proofs, with project annotations identifying the required relations. We extend the same library direction toward recognizing relational constructions from hypergraph structure. A formal composition rule connects arithmetic, ownership, protocol and realization evidence. Our worked distributed reduction preserves one specified result across concurrent workers, reordered arrivals and boundary mappings. Its probabilistic extension carries worker and conversion error with an explicit failure budget. With recent work by Urschel, we show bounds supply a relevant source of reusable proof terms. These constructions make verification a core engineering discipline for truly fearless programs and systems across a variety of hardware targets.
arXiv:2602.03762v5 Announce Type: replace-cross Abstract: Visually-guided acoustic highlighting seeks to rebalance audio in alignment with the accompanying video, creating a coherent audio-visual experience. While visual saliency and enhancement have been widely studied, acoustic highlighting remains underexplored, often leading to misalignment between visual and auditory focus. Existing approaches use discriminative models, which struggle with the inherent ambiguity in audio remixing, where no natural one-to-one mapping exists between poorly-balanced and well-balanced audio mixes. To address this limitation, we reframe this task as a generative problem and introduce a Conditional Flow Matching (CFM) framework. A key challenge in iterative flow-based generation is that early prediction errors -- in selecting the correct source to enhance -- compound over steps and push trajectories off-manifold. To address this, we introduce a rollout loss that penalizes drift at the final step, encouraging self-correcting trajectories and stabilizing long-range flow integration. We further propose a conditioning module that fuses audio and visual cues before vector field regression, enabling explicit cross-modal source selection. Extensive quantitative and qualitative evaluations show that our method consistently surpasses the previous state-of-the-art discriminative approach, establishing that visually-guided audio remixing is best addressed through generative modeling.
arXiv:2512.16545v3 Announce Type: replace-cross Abstract: Cu NPs have a broad applicability, yet their synthesis is sensitive to subtle changes in reaction parameters. This sensitivity, combined with the time- and resource-intensive nature of experimental optimization, poses a major challenge in achieving reproducible and size-controlled synthesis. While ML shows promise in materials research, its application is often limited by scarcity of large high-quality experimental data sets. This study explores ML to predict the DLS-derived hydrodynamic diameter of Cu NPs using a small data set of 25 syntheses. Latin Hypercube Sampling is used to efficiently cover the parameter space while creating the experimental data set. Ensemble regression models successfully predict hydrodynamic diameters with good predictive performance given the limited dataset. Since quantitative regression requires a unique DLS-derived hydrodynamic diameter, the regression model is restricted to mono-modal DLS distributions, while a complementary classification model identifies synthesis conditions for which quantitative prediction is applicable. Using equivalent out-of-sample validation, the ML and DoE models showed comparable generalization. The final ensemble model achieved an R2=0.74 compared to 0.60 for the DoE model, while retaining the complete synthesis parameter space, making it better suited for synthesis guidance. Additionally, classification models using both random forests and LLMs are evaluated to distinguish between large and small particles. These classification models exhibited only modest predictive performance, indicating that this small dataset is insufficient to fully exploit the capabilities of complex LLMs. Overall, this study demonstrates that carefully curated small data sets, paired with robust classical ML, can effectively support the synthesis of Cu NPs and highlights that for lab-scale studies, complex models like LLMs may offer limited benefits.
arXiv:2508.08833v4 Announce Type: replace-cross Abstract: Frontier large language models (LLMs) now reach near-ceiling accuracy on standard mathematical-reasoning benchmarks and gold-medal-level performance at the International Mathematical Olympiad. As these benchmarks saturate and their items leak into training data, a high score no longer shows whether a model reasons robustly or which component of its reasoning fails. To evaluate reasoning while keeping results informative and failures diagnosable, we propose GAP (Generalisation-and-Perturbation), a methodology that automatically generates mathematically equivalent variants of existing mathematics problems at scale using two disjoint, interpretable transformations: (1) surface renames, probing the binding between identifiers and latent variable roles, and (2) kernel rewrites, probing whether a high-level proof plan survives a change of mathematical setting. Compared with existing benchmarks, GAP has two key benefits: (1) novel, likely unseen variants mitigate data leakage, and (2) performance across transformation families enables failure diagnosis, each transformation testing a hypothesis about the cause of failure. We instantiate GAP on all 1,051 William Lowell Putnam Competition problems from 1938 to 2024, adding 5,255 unseen variants to form PutnamGAP, a 6,306-item competition-level mathematics corpus and the first public machine-readable dataset from the full Putnam archive. Using PutnamGAP, we evaluated 18 commercial and open-source models spanning sizes and providers. Accuracy drops across all models and variant families, most severely under kernel rewrites. This gap does not close with model strength, suggesting that even the strongest models' dominant weakness is transferring a proof plan to a changed mathematical setting, rather than handling surface changes. Further analysis provides finer failure diagnoses and potentially useful insights for improving LLM reasoning.
arXiv:2505.24311v3 Announce Type: replace-cross Abstract: We study the large-sample variational problem for t-distributed stochastic neighbor embedding with a class of input and output kernels. The input law has compact support and a density continuous on that support. An entropy equation determines the scale parameter in the input kernel, and we prove that this parameter exists and is unique at interior points of positive density. We then give sufficient conditions for solutions to exist and be uniformly bounded on the entire support. Under these conditions and a decay assumption on the output kernel, the discrete optimal values converge to a continuum minimum. Empirical measures of approximate minimizers are tight after translation; every subsequential limit is a compactly supported minimizer satisfying the equilibrium equation. The admissible output kernels include Gaussian kernels and, in output dimension two, the Cauchy kernel. Numerical examples compare the two-dimensional representations obtained with different kernels.
arXiv:2504.12392v3 Announce Type: replace-cross Abstract: Archetypal analysis (AA) was originally proposed in 1994 by Adele Cutler and Leo Breiman as a computational procedure for extracting distinct aspects, so-called archetypes, from observations, with each observational record approximated as a mixture (i.e., convex combination) of these archetypes. AA thereby provides straightforward, interpretable, and explainable representations for feature extraction and dimensionality reduction, facilitating the understanding of the structure of high-dimensional data and enabling wide applications across the sciences. However, AA also faces challenges, particularly as the associated optimization problem is nonconvex. This is the first survey that provides researchers and data mining practitioners with an overview of the methodologies and opportunities that AA offers, surveying the many applications of AA across disparate fields of science, as well as best practices for modeling data with AA and its limitations. The survey concludes by explaining crucial future research directions concerning AA.
arXiv:2502.02679v3 Announce Type: replace-cross Abstract: Approximation and learning of classifiers of large data sets by neural networks in terms of high-dimensional geometry and statistical learning theory are investigated. The influence of the VC dimension of sets of input-output functions of networks on approximation capabilities is compared with its influence on consistency in learning from samples of data. It is shown that, whereas finite VC dimension is desirable for uniform convergence of empirical errors, it may not be desirable for approximation of functions drawn from a probability distribution modeling the likelihood that they occur in a given type of application. Based on the concentration-of-measure properties of high dimensional geometry, it is proven that both errors in approximation and empirical errors behave almost deterministically for networks implementing sets of input-output functions with finite VC dimensions in processing large data sets. Practical limitations of the universal approximation property, the trade-offs between the accuracy of approximation and consistency in learning from data, and the influence of depth of networks with ReLU units on their accuracy and consistency are discussed.
arXiv:2409.01656v4 Announce Type: replace-cross Abstract: We consider the problem of estimating graph limits, known as graphons, from observations of sequences of sparse finite graphs. In this paper we show a simple method that can shed light on a subset of sparse graphs. The method involves mapping the original graphs to their line graphs. We show that graphs satisfying a particular property, which we call the square-degree property are sparse, but give rise to dense line graphs. This enables the use of results on graph limits of dense graphs to derive convergence. In particular, star graphs satisfy the square-degree property resulting in dense line graphs and non-zero graphons of line graphs. We demonstrate empirically that we can distinguish different numbers of stars (which are sparse) by the graphons of their corresponding line graphs. Whereas in the original graphs, the different number of stars all converge to the zero graphon due to sparsity. Similarly, superlinear preferential attachment graphs give rise to dense line graphs almost surely. In contrast, dense graphs, including Erdos-Renyi graphs make the line graphs sparse, resulting in the zero graphon.
arXiv:2610.05196v2 Announce Type: replace Abstract: Risk regulation imposes directional constraints on scores; we adopt their strict per-input form -- the score monotone non-decreasing in every exposure input -- as a normative commitment. Deployed pipelines -- monotone hand-crafted aggregates feeding sign-constrained gradient boosting -- already satisfy it by composition, so constrained-versus-unconstrained comparisons price a guarantee the incumbent has for free. We instead hold admissibility fixed on both sides and measure what learning the aggregation is worth. Our instrument is a recurrent network whose state is classical risk statistics (an exponentially weighted moving average and a high-water mark with learned transforms), monotone by construction in every input and per MC-dropout sample. The central finding, by functional regression, is a subsumption boundary: a learned monotone channel reproduces the geometrically weighted separable family of hand-crafted statistics, one channel per member, to Spearman $\rho \ge 0.996$, approximates window statistics with measurable ceilings, and fails at consecutivity ($\rho = 0.924$) and time localization (0.628), both structural, and at the exposure floor (0.829), a learnability boundary. One explicit admissible basis repairs each failure (rank correlation 1.000). In or near the separable family, learned and engineered aggregation are substitutes, and the learned channel is never statistically behind at full sample size and specified capacity. Its advantages are incumbent-specific: a committed grid pays up to 0.019 AUC in decay regions it leaves uncovered (the learned channel stays within 0.004 of the strongest engineered consumer at every swept point); the highest-dimensional comparator degrades fastest with scarce data; and beyond the training support, grid-fed tree-ensemble scores go flat while a strictly increasing head keeps ranking. No single incumbent is dominated on all three axes.
arXiv:2609.24259v3 公告类型：替换 摘要：智能体记忆的有效性最终取决于底层LLM是否能让上下文中的每条记忆对其响应产生适当程度的影响。然而，这种能力一直在很大程度上被忽视。为评估这一能力，我们引入了MemCalib，这是一个基于真实记忆系统场景的基准，用于评估记忆使用并推动优化算法的发展。MemCalib测试集上的结果表明，前沿开源和闭源模型都难以恰当地使用记忆。它们经常过度使用或使用不足，而不是让每条命题的实际使用程度与其目标水平相匹配，从而导致响应存在偏差且质量低下。对包括组相对策略优化和在策略自蒸馏在内的常见后训练算法进行实验进一步揭示了明显的方向性偏差：模型经过训练后，会在一个方向上有所改善，却在另一个方向上恶化。因此，我们提出MemCalib-RL，这是一种有序的双向反事实信用分配算法，通过精确的原子消融将过度使用和使用不足的信号分离开来，并将其信用归因定位到响应词元。跨模型系列和规模（Qwen3-8B、Ministral-3-8B-Instruct和Qwen3.5-35B-A3B）的结果表明，MemCalib-RL取得了最佳的总体性能，同时更好地平衡了过度使用和使用不足；在外部基准评估中，其收益也能泛化到MemCalib之外。进一步实验支持了其设计选择和稳健性，并为其训练动态提供了深入见解。
arXiv:2609.23257v2 Announce Type: replace Abstract: Time series forecasting underpins critical decision-making across diverse domains. While large language models (LLMs) offer promising reasoning capabilities, existing LLM-based time series forecasting approaches either reduce them to numerical predictors that bypass their strengths, or allow direct forecast generation that destabilizes predictions in non-stationary settings. We introduce CTRL, a framework that decouples semantic reasoning from quantitative prediction. A frozen backbone generates base forecasts, while specialized LLM agents function as controllers that analyze backbone prediction errors through decomposed trend, seasonal, and irregular components, grounding reasoning in interpretable temporal structure. Each agent outputs compact control signals that a lightweight residual decoder translates into forecast corrections. CTRL incorporates label-free test-time adaptation that detects distribution shift from input statistics alone and readapts control signals with only 3-24 LLM calls via caching. CTRL is explicitly designed to improve robustness under non-stationary temporal dynamics and distribution shift, while remaining competitive on highly stationary time series where adaptive correction provides limited additional benefit.
arXiv:2609.14193v4 公告类型：替换 摘要：On-policy 蒸馏（OPD）正日益应用于前沿基础模型的后训练。该领域以往的工作主要聚焦于算法改进，但 OPD 在多大程度上依赖其训练问题，尤其是这种依赖可以降低到什么程度，仍不明确。在两个具有代表性的单教师 OPD 设置中，我们发现，使用 8 个真实提示词进行训练所取得的性能，与使用 17k 个问题进行训练相当；而在测得的难度和初始蒸馏差距上存在显著差异的数据集，也会产生相近的结果。我们的分析提出了两种互补的解释：重复采样可能使少量提示词也能揭示教师模型的大量监督信号；同时，OPD 能够迁移超越数据集特定知识的、可泛化的推理能力。基于这些观察，我们进一步提出无数据 on-policy 蒸馏（DF-OPD）设置，以研究系统能否自行提供训练问题，从而消除对外部数据的需求。在不使用种子示例的情况下，通过 64 个自行生成的问题，DF-OPD 在两个单教师设置中都取得了与完整数据 OPD 相当的性能。这一发现同样适用于多教师 OPD：在数学、代码和指令遵循任务中，1k 个生成问题取得了与使用约 7k 个真实后训练示例进行训练相当的性能。我们还进一步探索了 OPD 是否能够在没有明确训练问题的情况下运行。实验表明，这仅在有限情况下有效：学生模型会出乎意料地生成并回答自己的问题，从而使这一过程退化为一种隐式的 DF-OPD。总体而言，这些发现促使我们重新审视训练数据在 on-policy 蒸馏中的作用。代码可见于 https://github.com/Ryuki661/DF-OPD
arXiv:2608.19584v3 Announce Type: replace Abstract: We study landscapes for complex-parameterized networks. Our approach is motivated with an information-theoretic manifold perspective of the parameter and via classical optimization guarantees although of complex geometric variety such as through Dolbeault asymptotics. The descent path admits a K\"ahler information metric under a cross-entropy via the Wirtinger Hessian on the log-likelihood potential. We restrict attention to a descent update rule with natural gradient descent via a differentiated loss scaled by the inverse metric, so the descent path remains in the holomorphic tangent bundle. We emphasize Calabi-Yau information manifolds which profane theoretical guarantees via an ill-curvature-conditioned landscape. We focus on Calabi-Yau metrics specifically in a non-compact setting with a global potential, so defined geometrically rather than invoking the topological requirements of the Calabi conjecture. In non-compact settings, we can write the metric determinant with respect to a background in terms of a pluriharmonic or real-valued function. Under bounded, nonuniform, and almost low-rank assumptions, we get a partial eigenvalue blow-up effect. In an empirical setting, a Ricci-flat metric will not form, but the blow-up effect is a local condition and can partially hold empirically on open sets. We isolate the Calabi-Yau case in a theoretical setting, and we counteract the corrupted geometries under regularization. Moreover, it has been discovered that negative curvature subverts the loss landscape, specifically sectional curvature, so we expand on this and draw interconnections to negative-definite Ricci curvature. Our arguments primarily exist via geometric analysis, although we establish roots in deep learning theory such as through asymptotics at initialization and connections through failure modes of neural network guarantees under vanishing and negative Ricci curvature.
arXiv:2608.11704v2 Announce Type: replace Abstract: Dynamic Time Warping (DTW)-based Nearest-Neighbor (NN) classifiers are effective for time-series classification but are vulnerable to mislabeled training samples and require numerous DTW computations during inference. We propose DTW-based Granular Ball Computing (DTW-GBC), which organizes temporally similar training samples into granular balls and performs classification at the granule level. We further develop two granular-ball construction strategies for DTW-GBC. Experiments on four benchmark datasets with symmetric label noise show that the two DTW-GBC variants generally mitigate the performance degradation caused by label noise while requiring substantially fewer comparisons than DTW-based 1-NN during inference. These findings suggest that DTW-GBC provides a favorable balance between classification robustness and inference efficiency.
arXiv:2607.01736v3 公告类型：替换 摘要：我们研究了在 Gymnasium 的 LunarLander-v3 中使用人类演示训练的循环状态空间模型（RSSM）世界模型的闭环特性。我们使用训练好的世界模型进行零样本 CEM 模型预测控制（MPC），并在想象环境中进行演员-评论家（A2C）训练。在 100 个留出的回合上进行评测后，所选的基于模型的 A2C 策略（在世界模型检查点 280 上训练）达到了 +189.5 的平均回报，与表现最佳的无模型 A2C 检查点（+183.7；回合上限分别为 600 步和 1000 步）相当，而实际训练交互次数减少了约 65 倍。我们还将世界模型 MPC 与基于成功演示训练的行为克隆（BC）策略进行了比较。只有在随机动作选择下，BC 策略的平均回报才与 MPC 相当；在相同的 20 个回合中，BC 策略出现了 1 个灾难性回合，而 MPC 没有。随后，我们引入奖励可观测性比例（Reward Observability Fraction，ROF）：它表示在线性化潜在动力学的可观测子空间中，奖励梯度的欧几里得比例。我们证明，接下来的 H 个观测包含关于该子空间中每个方向的 Fisher 信息，而不包含关于与该子空间正交方向的信息。然而，ROF 本身取决于潜变量的缩放方式：重新缩放潜变量会改变 ROF，却不会改变模型，因此不同模型之间的原始 ROF 水平不可比较。将奖励头强制连接到经过后验校正的潜变量 z 上会提高 ROF，并且在我们测试的两组运行上，这一提升通过了坐标不变性检验。最后，我们检验 ROF 或其他离线指标能否预测 MPC 的闭环崩溃。即使数据和配置完全相同，不同训练运行之间也会出现不同程度的崩溃。ROF 无法预测崩溃；我们筛选的 108 个离线汇总指标均未通过置换检验，最佳候选指标在新的运行上也会失效。仅凭模型和记录数据从离线结果预测崩溃仍是一个开放问题。
arXiv:2606.01081v2 公告类型：替换 摘要：面向决策的学习（DFL）通过优化下游决策质量，而非单独的预测准确率，来训练预测模型。对于上下文线性优化，现有大多数 DFL 方法都假设使用离线数据，并且能够完整观测目标成本向量。我们针对部分反馈下的序列上下文线性优化，提出了一种策略内学习方法，将标准的 bandit 反馈设置进行了推广。我们的方法学习一种随机的“预测后优化”策略：从条件分布中采样成本向量预测，并求解由此得到的下游线性优化问题。为更新这一分布模型，我们引入了一种由两个部分组成的混合梯度估计器。第一部分是得分函数估计器，它能够提供无偏但可能具有较高方差的策略梯度估计。第二部分是面向决策的插件组件，它利用对潜在成本向量的辅助干扰估计来挖掘下游优化结构；随着该估计的改进，这一组件将提供更多信息。我们证明，策略梯度范数平方的平均值满足 O(T^-1/2) 界，与标准非凸 SGD 的收敛速率相匹配。在 top-k 选择、最短路径、组合定价以及真实数据能源调度基准上的实验表明，在 bandit 反馈下，混合梯度方法在全部四个基准上都比上下文 bandit 基线取得更低的平均累积遗憾，在其中三个基准上也优于线性 Thompson sampling；实验还表明，该方法同样适用于更丰富的条件生成式成本模型。代码可在 https://github.com/Joeyetinghan/on-policy-bandit-dfl 获取。
arXiv:2605.22737v4 公告类型：替换 摘要：蒸馏攻击给模型提供方带来了部署层面的权衡：使模型更有用的同一批输出，也可能让模型更容易被模仿。我们通过受效用约束的教师模型与自适应学生模型之间的极小极大博弈，研究了这一权衡。我们的框架产生了易于处理的单边响应规则：一种自适应评估规则，使学生模型对高价值样本重新加权；以及一种教师模型侧的防御模板，抑制对蒸馏最有用的输出。基于一种低成本的样本价值代理指标，我们提出了 Product-of-Experts（PoE），这是一种简单的、仅需前向传播的防御方法，在生成过程中将教师模型与代理学生模型结合起来。实验证明，自适应评估揭示了显著的被动评估与自适应评估差距：在当前最先进的防御方法上，相比被动评估所显示的结果，自适应学生模型在 GSM8K 和 MATH 上恢复出了多得多的能力。在这一更强的评估下，昂贵防御方法与 PoE 之间表面上的鲁棒性差距大幅缩小，而 PoE 的成本仍显著更低，并且能够保留质量更高的推理轨迹。总体而言，我们的结果表明，强大的蒸馏攻击仍然难以阻止；反蒸馏研究的进展应当依据自适应学生模型，而非被动学生模型进行评估。我们的代码可在以下地址获取：https://github.com/ysfalh/distillation-game】【。
arXiv:2604.13966v2 Announce Type: replace Abstract: We study value adaptation in offline-to-online reinforcement learning under general function approximation. Starting from an imperfect offline pretrained $Q$-function, the learner aims to adapt it to the target environment using only a limited amount of online interaction. We first characterize the difficulty of this setting by establishing a minimax lower bound, showing that even when the pretrained $Q$-function is close to optimal $Q^\star$, online adaptation can be no more efficient than pure online RL on certain hard instances. On the positive side, under a novel structural condition on the offline-pretrained value functions, we propose O2O-LSVI, an adaptation algorithm with problem-dependent sample complexity that provably improves over pure online RL. Finally, we complement our theory with neural-network experiments that demonstrate the practical effectiveness of the proposed method.
arXiv:2604.10166v2 Announce Type: replace Abstract: Intelligent operation of thermal energy networks aims to improve energy efficiency, reliability, and operational flexibility through data-driven control, predictive optimization, and early fault detection. Achieving these goals relies on sufficient observability, requiring continuous and well-distributed monitoring of thermal and hydraulic states. However, district heating systems are typically sparsely instrumented and frequently affected by sensor faults, limiting monitoring. Virtual sensing offers a cost-effective means to enhance observability, yet its development and validation remain limited in practice. Existing data-driven methods generally assume dense synchronized data, while analytical models rely on simplified hydraulic and thermal assumptions that may not adequately capture the behavior of heterogeneous network topologies. Consequently, modeling the coupled nonlinear dependencies between pressure, flow, and temperature under realistic operating conditions remains challenging. In addition, the lack of publicly available benchmark datasets hinders systematic comparison of virtual sensing approaches. To address these challenges, we propose a heterogeneous spatial-temporal graph neural network (HSTGNN) for constructing virtual smart heat meters. The model incorporates the functional relationships inherent in district heating networks and employs dedicated branches to learn graph structures and temporal dynamics for flow, temperature, and pressure measurements, thereby enabling the joint modeling of cross-variable and spatial correlations. To support further research, we introduce a controlled laboratory dataset collected at the Aalborg Smart Water Infrastructure Laboratory, providing synchronized high-resolution measurements representative of real operating conditions. Extensive experiments demonstrate that the proposed approach significantly outperforms existing baselines.
arXiv:2604.04611v3 Announce Type: replace Abstract: Federated learning (FL) enables multiple clients to collaboratively train a global model by aggregating local updates without sharing private data. In this work, we focus on cross-silo FL, where each client typically represents an independent organization. However, cross-silo FL can face the challenge of free-riders, clients who submit fake model parameters without performing actual training to obtain the global model without contributing. Chen et al. proposed a free-rider detection method based on the weight evolving frequency (WEF) of model parameters. This detection approach is practical because it requires neither a proxy dataset nor pre-training. Nevertheless, it struggles to detect ``dynamic'' free-riders who behave honestly in early rounds and later switch to free-riding, particularly under global-model-mimicking attacks such as the delta weight attack and our newly proposed adaptive WEF-camouflage attack. In this paper, we propose a novel detection method S2-WEF that simulates the WEF patterns of potential global-model-mimicking attacks on the server side using previously broadcast global models, and identifies clients whose submitted WEF patterns resemble the simulated ones. To handle a variety of free-rider attack strategies, S2-WEF further combines this simulation-based similarity score with a deviation score computed from mutual comparisons among submitted WEFs, and separates benign and free-rider clients by two-dimensional clustering and per-score classification. This method enables dynamic detection of clients that transition into free-riders during training without proxy datasets or pre-training. We conduct extensive experiments across four datasets and five attack types, demonstrating that S2-WEF provides robust dynamic free-rider detection across diverse settings.
arXiv:2603.26024v4 Announce Type: replace Abstract: Identification of causal directionality in bivariate numerical data is a fundamental research problem with important practical implications. This paper presents two alternative methods to identify direction of causation by considering conditional distributions: (1) Anticipated Asymmetric Geometries (AAG) and (2) Monotonicity Index (MI). The AAG method compares the actual conditional distributions to anticipated ones along two variables. Different comparison metrics, such as Pearson correlation, cosine distance, Hellinger distance, Jaccard index, Jeffreys divergence, K-L divergence, K-S distance, MAE, MSE, mutual information, and Wasserstein distance have been evaluated. Anticipated distributions have been projected as normal based on dual response statistics: mean and standard deviation. The MI method compares the calculated monotonicity indexes of the gradients of conditional distributions along two axes and exhibits count of gradient sign changes. Both methods assume stochastic properties of the bivariate data and exploit anticipated unimodality of conditional distributions of the effect. The proposed methods are straightforward and include only a limited number of hyperparameters that affect the accuracy of the identification. For a given set of hyperparameters, both the AAG and MI methods provide a unique, deterministic solution. To address sensitivity to hyperparameters, tuning has been done by utilizing a full factorial Design of Experiment. It turns out that the AAG method outperforms MI, achieving top weighted accuracies of 82.7% with simple tuning and 84.4% with size-adaptive tuning, compared with 81.6% for GRCI or 82.0% for CAREFL-H on the 99 pairs of the Tubingen real-world cause-effect examples. To evaluate the decisiveness of the identification method, a decision tree was fitted on the input data's symmetrical bivariate statistics to detect misclassified cases.
arXiv:2603.07233v2 Announce Type: replace Abstract: Predicting transcriptional responses to genetic perturbations is fundamental to functional genomics and therapeutic discovery. Recent deep learning models have shown promise in single-cell perturbation response prediction, but they typically generate each response in isolation, without explicitly leveraging related perturbations. We introduce PT-RAG (Perturbation-aware Two-stage Retrieval-Augmented Generation), a plug-in retrieval-and-conditioning module for generative cellular perturbation response. PT-RAG augments an existing perturbation-response backbone with learned access to related perturbation contexts. The key challenge is that relevance is not fixed in this setting: functionally related genes may elicit different effects across cell types. PT-RAG addresses this with a two-stage retrieval mechanism: GenePT-based semantic retrieval first identifies K candidate perturbations, after which a differentiable Gumbel-Softmax selector adaptively selects retrieved contexts conditioned on the control cell state, the query perturbation, and each candidate perturbation. We evaluate PT-RAG on two backbones, a STATE-style generator used as a frozen random reservoir and a fully trained scGPT, across cross-cell-type and cross-perturbation generalization tasks. PT-RAG consistently improves distributional similarity and often overall predictive quality; for example, on scGPT cross-cell-type results, the 2-Wasserstein distance drops by 5.9% relative to scGPT alone. The code to reproduce our experiments is available at https://github.com/difra100/PT-RAG_NIPS.
arXiv:2603.01376v2 公告类型：替换 摘要：大型语言模型（LLM）的稀疏加低秩（$\mathbf{S} + \mathbf{LR}$）分解已成为模型压缩领域一个颇具前景的方向，旨在将预训练模型权重分解为稀疏矩阵与低秩矩阵之和（$\mathbf{W} \approx \mathbf{S} + \mathbf{LR}$）。尽管近期取得了进展，现有方法相比稠密模型通常会出现显著的性能下降。在本文中，我们提出了 3BASiL-TM，这是一种高效的单次训练后处理方法，用于对 LLM 进行（$\mathbf{S} + \mathbf{LR}$）分解，旨在弥合这一差距。我们首先提出了一种新颖的三块交替方向乘子法（ADMM），称为 3BASiL，在具备收敛保证的情况下最小化逐层重构误差。随后，我们设计了一个高效的 Transformer 匹配（TM）细化步骤，在 Transformer 层之间联合优化稀疏与低秩组件。该步骤最小化一种新颖的内存高效损失，使输出在 Transformer 层面实现对齐。值得注意的是，TM 过程具有通用性，因为它能够增强任何（$\mathbf{S} + \mathbf{LR}$）分解，包括纯稀疏分解。数值实验表明，在（2:4 Sparse + 64 LR）配置下，与已有方法相比，3BASiL-TM 将相对于稠密 LLaMA-8B 模型的 WikiText2 困惑度差距缩小了 30% 以上。此外，与当前最先进的（$\mathbf{S} + \mathbf{LR}$）方法相比，我们的方法在 A100 GPU 上实现了超过 2.5 倍的压缩运行速度提升。我们的代码可在 https://github.com/mazumder-lab/3BASiL 获取。
arXiv:2507.21037v3 Announce Type: replace Abstract: Cross-subject motor-imagery electroencephalography (MI-EEG) decoding remains challenging because substantial inter-subject variability can cause both negative transfer from poorly matched source subjects and persistent distribution discrepancies between source and target domains. Existing multi-source domain adaptation methods often incorporate all available source domains or estimate source relevance using signal-level or task-specific representations, while distribution alignment is frequently performed only at the feature level. These limitations may introduce irrelevant source knowledge and fail to preserve class-discriminative structures across subjects. In this study, we propose a brain foundation model-guided multi-source domain adaptation framework (BFM-MSDA) for cross-subject MI-EEG decoding. The method first utilizes representations learned by a pretrained brain foundation model to estimate source--target compatibility and retrieve target-relevant sources. Subsequently, a relevance-weighted dual alignment strategy is applied to the selected sources and the unlabeled target domain. Specifically, Cauchy--Schwarz (CS) divergence is used to reduce discrepancies in marginal feature distributions, while conditional Cauchy--Schwarz (CCS) divergence further aligns class-dependent decision distributions. Source relevance is incorporated into both alignment terms so that more transferable source subjects contribute more strongly to adaptation. Experiments on two public MI-EEG benchmarks achieve average accuracies of 86.16% and 78.41%, outperforming representative cross-subject decoding and domain adaptation methods. Additional experiments with a large source pool further demonstrate that target-aware source retrieval improves scalability while mitigating negative transfer. These results highlight the importance of informed source selection under cross-subject distribution shift.
arXiv:2502.17920v2 Announce Type: replace Abstract: Pre-trained visual models have become fundamental in computer vision, but they face challenges in continual learning scenarios where data and tasks evolve over time. Low-Rank Adaptation (LoRA) offers efficient fine-tuning capabilities but remains limited for such dynamic environments. Standard LoRA cannot distinguish important subspaces, causing critical knowledge to be overwritten in sequential training. Existing approaches address this by dynamically expanding the set of LoRA adapters, either maintaining a growing pool of task-specific modules or merging new adapters into prior ones, at the cost of unbounded parameter growth or increasing inference complexity. We propose Continual Low-Rank Adaptation (C-LoRA), a method that enables a single, shared LoRA adapter to handle sequential tasks without catastrophic forgetting, without requiring any module selection or fusion at inference. The core of C-LoRA is a learnable routing matrix R that explicitly controls how each rank-one subspace contributes to the weight update. This matrix is decomposed into a stability component (R_base), which preserves knowledge from prior tasks, and a plasticity component (R_delta), which drives adaptation to the current task, providing direct control over the stability-plasticity trade-off. We analyze how R governs gradient flow during sequential training, and demonstrate competitive performance across multiple benchmarks.
arXiv:2405.04118v4 公告类型：替换 摘要：现代 AI 系统（如自动驾驶汽车和游戏智能体）能够达到超越人类的性能，但通常缺乏类似人类的泛化能力、可解释性以及与人类用户的互操作性。受人类语言与决策之间丰富互动的启发，我们提出了通过语言瓶颈进行策略学习（Policy Learning with a Language Bottleneck，PLLB），这是一种使 AI 智能体能够生成语言规则的框架，这些规则用于捕捉产生高回报行为背后的高层次策略。PLLB 在由语言模型指导的*规则生成*步骤与更新步骤之间交替进行：在更新步骤中，即使一条规则不足以描述完整的复杂策略，智能体也能在规则的指导下学习新的策略。在包括双人信号博弈、迷宫导航、图像重建和机器人抓取规划在内的五项多样化任务中，我们展示了 PLLB 智能体不仅能够学习更具可解释性和泛化性的行为，还能与人类用户分享所学规则，从而实现更有效的人机协作。我们在 https://github.com/meghabyte/bottleneck 提供了实验源代码。
arXiv:2610.12465v1 公告类型：cross 摘要：通用机器人必须执行从敏捷运动到灵巧操作的广泛任务。尽管仿真到现实强化学习（RL）已被证明是实现这一目标的有用工具，但当前的 RL 流程依赖大量工程工作，并且需要针对每个任务设计结构先验，例如塑形奖励和示范。近期研究表明，在若干操作问题中，将多样化的模拟器重置与大规模并行仿真相结合，可以减轻大量工程负担。然而，我们发现，将这一范式直接扩展到更精确或更具动态性的任务上，仍然并非易事。尽管模拟器重置有助于探索，但在这一分布上进行均匀采样，会将越来越大比例的学习经验浪费在策略已经掌握或尚且无法尝试的任务配置上。这使得人们难以看到扩展 RL 并行环境所带来的预期收益，因为一个批次中的大量学习信号都会在训练过程中被浪费。为缓解这一问题，我们提出 Success Guided Sampling（SGS），这是一种简单的自适应采样器，将 RL 训练集中在策略能力边界附近的任务配置上。这样，基于大规模仿真的 RL 就能充分利用一个批次中的经验，从而更有效地扩展到大规模并行仿真。在使用多达 $2^{20}$（超过一百万个）并行环境的实验中，SGS 使 RL 能够解决具有挑战性的多地形四足机器人运动任务和接触密集型装配任务，而以往方法无法解决这些任务。最后，我们将学习到的操作策略蒸馏为基于 RGB 的策略，并展示了其在真实硬件上的若干项具有挑战性的装配任务中的零样本迁移。项目网站：https://sgs-rl.github.io/。
arXiv:2610.12437v1 Announce Type: cross Abstract: Density ratios quantify distribution shift from a probability-mass point of view, whereas displacement fields describe, from a dynamical point of view, how one distribution is transported onto another. Although both offer complementary insights, they are usually estimated separately, and converting one into the other requires post-processing. In this paper, we estimate the density ratio between a target and a base distribution by parametrizing it through a displacement field acting on the base: the log-ratio is modeled as minus the Stein operator of the base applied to the field, up to a normalizing constant. This gives both statistical and dynamical descriptions of the distribution shift through a single convex optimization problem. Iterating this estimate-and-move step gives two inference algorithms: push-forward moves the model and corrects a pretrained sampler without retraining it, whereas pull-back moves the data closer to the base and fits a transformation model one layer at a time. Applications to distribution shift in simulation-based inference and to nonlinear independent component analysis illustrate the benefits and limitations of the approach.
arXiv:2610.12421v1 公告类型：cross 摘要：稠密对应匹配长期以来受限于简化的时空先验，例如平滑运动和刚性几何。尽管这些假设对经典任务十分有效，但在图像编辑和参考引导生成（IEG）中却会失效，因为这类变换能够保持视觉身份，同时打破物理连续性。为了在此类变换中建立保持身份的对应关系，我们提出了 FreeMatching，这是一种可泛化的框架，将生成式和语义基础表征与来自经典数据集、跟踪视频和合成场景的异构监督相结合。教师引导的迭代细化进一步提升了 FreeMatching 在 IEG 中的对应关系效果，且无需稠密对应标注。在实验中，单个 FreeMatching 模型显著提升了具有挑战性的 IEG 图像对上的对应质量，同时在经典基准测试中保持了有竞争力的性能。此外，我们还展示了其作为评估身份保持程度的定量指标的实用性，其得分与人类判断具有较高相关性。代码可在 https://github.com/luping-liu/FreeMatching 获取。
arXiv:2610.12393v1 公告类型：cross 摘要：自监督多模态表征学习已在各个领域取得显著成功，但由于跨模态交互的复杂性，捕获协同信息仍然具有挑战性。不同于各个模态之间分别存在的共享信息，协同信息产生于这样的情形：只有在多个模态的联合配置下，任务相关信号才会出现，而从任何单一模态中都无法恢复这些信号。本文聚焦于如何在多模态表征中保留这类协同信号的信息容量。关键观察是，协同信息体现在模态之间的高阶统计依赖中，这为显式建模联合交互提供了一个具有理论依据的目标。基于这一认识，我们提出高阶表征与信息学习（Higher-order Representation and Information Learning，HRIL），在模态嵌入上构建经验交叉矩张量，以表示多路交互。HRIL采用 Tucker 分解获得核心张量，并结合协同感知正则化项，防止能量集中，从而保留用于捕获协同信息的高阶耦合容量。在受控协同任务和真实世界基准上的实验表明，相较于现有多模态对比学习方法，HRIL consistently improved，并在以协同交互为主导的任务上取得了显著提升。代码发布于 https://github.com/brightest66/HRIL。
arXiv:2610.12325v1 Announce Type: cross Abstract: Do LLM scientific agents rely only on their initial task context, or do they adapt their decisions in response to experimental feedback? We study this question in neural operator adaptation, where a large language model (LLM) selects fine-tuning configurations under a limited trial budget. Across transfers within and between partial differential equation (PDE) families, the LLM achieves lower held-out test nRMSE than random search and Bayesian optimisation in nearly every matched comparison. Endpoint performance alone cannot distinguish what happens, so we verify each attribution with controlled interventions. Before observing any validation score, the LLM's first configuration already ranks near the top of the corresponding random-search pool, indicating a useful initial bias. A complementary cold-start intervention shows that the selected base learning rate shifts with the PDE description. Once feedback becomes available, reassigning validation scores among evaluated configurations changes the next proposal in every case tested, whereas a value-preserving rewrite produces no comparable aggregate effect. These interventions establish that the LLM's decision-level actions respond to the given task and observed outcomes, showing that it combines a task-dependent prior with sensitivity to experimental feedback.
arXiv:2610.12095v1 公告类型：交叉发布 摘要：小样本学习旨在利用有限的标注样本识别新类别。近期研究引入文本语义，以弥补视觉观测不足并改善类别表示。然而，较高的图像文本一致性可能同时反映物体的固有属性和偶然出现的上下文，使支持集原型容易受到上下文污染。为解决这一问题，我们提出 DVLA-RL++，在 DVLA-RL 的基础上引入互补语义净化（CSP）和反事实强化学习门控（CRG）。具体而言，CSP 根据带标注的支持样本生成固有属性描述与干扰因素描述，并比较这些描述与每个支持样本词元的一致性。随歧义程度变化的拒绝间隔引导稀疏证据分配，而固有语义锚点则填补未分配的权重，在视觉证据不可靠时提供兜底。CRG 使用兼顾识别性能与干扰暴露程度的奖励，学习逐层语义融合强度。在同一小样本学习任务中独立执行的参考轨迹提供配对学习信号。理论分析将保留证据及锚点质量与原型稳定性联系起来，并确立了同策略梯度无偏估计的条件。在标准、细粒度和跨域基准上的实验表明，该方法达到最先进的准确率，较 DVLA-RL 平均提升 1.4%。项目页面：https://peacelwh.github.io/TPAMI27-DVLA-RLpp/。
arXiv:2610.12022v1 公告类型：cross 摘要：大型语言模型（LLM）正日益部署于社会技术系统中，其中社会归因——将外部事件归因于行动者社会行为的原因和理由的推理过程——发挥着关键作用。这些过程涉及对社会原因、责任以及对行动者的责备/赞誉进行判断。尽管归因模型已通过归因理论在社会心理学和认知领域得到充分研究，但社会归因在人工智能领域，尤其是 LLM 社会推理中的研究仍不充分。本文首次对 LLM 社会归因进行了系统性探索。我们的工作聚焦于责任归因和责备归因，考察当前 LLM 的判断及其底层内部机制。在归因理论的指导下，我们构建了一个社会归因基准，其中包括基于归因理论研究中经典情境的 Vignette 子集，以及基于现实世界社会叙事的 Reality 子集，共产生 7,639 个责任/责备判断问题。在此基础上，我们评估了 32 个具有代表性的 LLM 和 5 个基础非 LLM 基线模型。为进一步探索 LLM 判断过程背后的内部机制，我们开发了一种基于探究的方法，用于研究 5 个关键归因维度在潜在空间中的表征，以及这些维度对 LLM 判断的影响与人类社会归因中的影响是否保持一致。研究结果表明，当前 LLM 在责任和责备判断方面与人类表现出可测量但并不完整的一致性，同时，这种一致性与模型规模呈正相关。一些归因维度可以从 LLM 隐状态的特定位置系统性地解码出来，并且它们对最终判断的影响与人类归因理论所揭示的影响一致。数据集及相关代码可从 https://github.com/Yuzhaoxin946/SAB-Bench 获取。
arXiv:2610.11971v1 公告类型：cross 摘要：视觉-语言-动作（VLA）策略假定其训练所基于的实体形态保持不变，而当关节故障改变指令动作的实际物理执行方式时，这类策略可能失效。现有的故障恢复方法通常需要针对特定任务重新训练、故障标签、显式诊断或特权实体形态信息。我们提出 CAPABLE，这是一种面向冻结 VLA 的统一能力感知适应框架，将自监督能力推断与残差强化学习相结合。CAPABLE 利用跨关节共享的时序编码器、雅可比矩阵约束、跨关节注意力和自监督物理预测，根据指令-响应历史与运动学信息在线推断能力，即每个关节实际实现了多少被指令的运动，以及该运动如何影响末端执行器行为。由此得到的表征用于调节残差策略，在无需故障标签或故障关节标识符的情况下，为 VLA 的机械臂动作添加有界修正。在 28 个 LIBERO 任务中，CAPABLE 将一个未参与故障训练的执行器上的成功率从 24.8% 提升至 59.3%，相比参数量匹配的全局历史基线高出 17.4 个百分点，同时保持正常性能。在 6 个关节上进行的留一执行器实验表明，这种迁移并不局限于某一个执行器；其他评估还刻画了其向未见过的故障类别的迁移能力，并展示了其在实体 Franka Panda 上的恢复效果。https://capable-vla.github.io/
arXiv:2610.11956v1 Announce Type: cross Abstract: A generated video of a task the robot is about to perform is useful guidance only if it depicts the phase the robot is actually in. We show that temporal misalignment can turn a task-consistent generated future into actively harmful guidance. On CALVIN, a five-frame early shift nearly erases the benefit of generated futures, reducing success from 81.3% to 54.8% against 54.0% without futures; imposed timing shifts reduce it even further to 34.2%, 19.8 points below the future-free policy. We introduce Reliability-Aware Future Conditioning (RAFC), which treats this as a control problem rather than a generation problem. At every step, RAFC estimates how far to trust the received clip and which nearby temporal hypothesis to prefer, falling back toward a static branch when neither fits, and it learns both from task reward alone without shift labels or alignment supervision. RAFC sits on top of Future-Experience Conditioning (FEC), which builds the clip once from task grounding, a robot-free digital-twin rollout, and mask-free video diffusion. Under deliberately off-grid phase shifts and rate mismatch, RAFC substantially improves success under temporal mismatch. Candidate ensembling accounts for most of the recovery near alignment, while learned reliability adds a further 7.0 percentage points over uniform averaging of the identical candidate bank under off-grid shifts. The gain holds on the evaluated task sets and survives on a Franka under natural timing mismatch nobody imposed, where aggregate success rises from 26.7% to 56.7%. All resources will be made publicly available. https://future-condition.github.io/.
arXiv:2610.11945v1 公告类型：cross 摘要：在机器人上采集触觉示范成本高且速度慢，因此人们开始使用成本更低的人用触觉手套来实现可扩展的数据采集。然而，人用电容式/压阻式手套与机器人触觉传感器在换能原理、传感器布局、空间分辨率和动态响应方面存在根本差异，使得原始传感器通道的对齐变得缺乏明确意义。为解决这一问题，我们提出了TACROSS，这是一种从人类触摸中学习并将其迁移至机器人的可扩展系统。该系统通过在接触事件层面而非原始传感器数值层面对齐触觉流，弥合了上述异构性。TACROSS的硬件组件集成了一款五层压阻式手套，成本为10.86美元，配备285个感知点。为对齐接触语义，我们设计了规范化器和残差适配器，通过一个在手指之间进行注意力计算的时序Transformer，将异构信号映射到一个256维的共享触觉潜空间中。我们进一步提出了一种以机器人为基础的策略学习方案：机器人示范作为真实动作监督的唯一来源，而人类示范则用于触觉表征学习，并通过有效的重定向手部目标提供基于置信度加权的辅助监督。我们在四项接触密集型操作任务上评估了系统。与传统遥操作相比，我们提出的系统将效率提升了3.5倍，同时将示范采集设备成本降低了95.7%。我们将开源TACROSS硬件和软件系统，并公开发布一个包含超过150小时记录的触觉数据集。项目页面：https://tacross-touch-project.github.io/。
arXiv:2610.11924v1 Announce Type: cross Abstract: Media-bridged time series forecasting is expanding to encompass traditional "multivariate" and emerging "multimodal" (e.g., through textual assistance). Existing Time Series Forecasting (TSF) models still rely on paradigm-specific relation, fusion, and temporal modules, hindering a common forecasting backbone across numerical and pre-aligned narrative-flow settings. To explore this, we propose the Multimedia Identity-Aware Prism Network (MIDAPN), a unified spatiotemporal forecasting backbone based on media-general graph adaptation and automatic temporal learning: (1) Following media pre-alignment, our Multimedia Identity-Aware Graph (MIDAG) revisits identity through static essence, dynamic behavior, and latent commonality, inducing affinities that extend variable-specific dependencies across media. Contextual Identity Modulation (CIM) further refines discriminative aggregation. (2) We develop Spectral Prism Convolution (SPConv) to automatically perform hierarchical temporal analysis, balancing coarse trends and fine-grained details. Meanwhile, its Adaptive Search Guidance configures a scale-efficient architecture for temporal-dimension reconstruction. These decoupled yet synergistic components jointly address media identity disentanglement and temporal-scale mismatch. Comprehensive evaluations involving 16 SOTA TSF models across 13 "multivariate" and 12 "multimodal" datasets, alongside targeted long-context comparisons against 14 time series foundation models and fused pretrained language models, demonstrate MIDAPN's consistent superiority and broad shared backbone compatibility. The code is available at \href{https://github.com/leijieruilq/MIDAPN/tree/main}{https://github.com/MIDAPN}.
arXiv:2610.11798v1 Announce Type: cross Abstract: Softmax attention, at the heart of Transformers, has demonstrated remarkable capabilities. Yet its underlying mechanisms remain only partially understood. Recent theoretical work studies Gaussian prompts, where the infinite-prompt limit reduces softmax attention to a linear map, but also removes the query-dependent selection that distinguishes it from linear attention. This work studies the infinite-prompt limit of softmax attention on Gaussian mixtures, which retain the tractability of Gaussian data while introducing latent structure, multimodality, and nonlinear dependencies. We show that softmax attention can represent and learn, via gradient-based methods, optimal solutions to a range of statistical tasks, including supervised classification and denoising. Our results highlight two complementary capabilities of softmax attention: it can recover linear tasks as effectively as its simpler linear counterpart, while also exploiting query-dependent context selection to solve nonlinear tasks beyond the reach of linear attention.
arXiv:2610.11794v1 Announce Type: cross Abstract: Learning to act in unfamiliar environments requires agents to infer how the world works and revise that understanding as new evidence arrives. Yet limited observations can support multiple world models that explain past interactions but predict different outcomes in unseen states. We introduce Memento 3, building on the Memento series to enable frozen LLM agents to continually learn explicit world models through external memory. The agent maintains a natural-language rulebook as persistent semantic memory, recording revisable hypotheses about environment dynamics while leaving unknown aspects underspecified. It compiles this rulebook into executable code for prediction and planning. Through a continual loop of observation, reflection, rule revision, compilation, and verification, the agent uses prediction errors to refine both the rulebook and its code. Updated code is accepted only when the LLM judges it faithful to the rulebook and cell-exact replay reproduces the observed transitions. We investigate this process as a model-based route to recursive self-improvement (RSI): the agent autonomously explores the environment, revises its world model, and uses verified updates to guide subsequent interaction and learning, while the underlying LLM remains fixed. A population extension maintains multiple world models in parallel, sharing interaction evidence and using their predictions to guide exploration. On ARC-AGI-3, the single-model agent clears every level of all 25 public games, achieves a mean Relative Human Action Efficiency (RHAE) of 100.0, and uses 44% of the human action count. In an Atari Pong case study, a learned feedback controller wins 21:0 in each of three evaluated episodes with different openings, without further LLM calls.
arXiv:2610.11759v1 Announce Type: cross Abstract: Near-term quantum approaches to combinatorial optimization are limited by qubit counts, circuit fidelity, sampling cost, and the difficulty of encoding constraints, while machine learning is increasingly used to configure and control quantum optimization workflows. We call such workflows adaptive: decisions conventionally fixed in advance, from formulation and penalties to shot budgets, backends, and whether to invoke a quantum processor at all, are made by learned policies that respond to the instance, the progress of the solve, or the hardware. This review examines three paradigms, AI for quantum optimization, quantum for AI-driven optimization, and AI-quantum co-optimization, and organizes the literature by the decision being learned rather than by application. A structured review of 119 papers, 67 coded in detail, shows that the evidence is considerably stronger for AI-assisted quantum optimization than for the reverse direction: learning already reduces quantum evaluations, improves initialization, supports decomposition and penalty control, and mitigates noise, whereas evidence that quantum computation improves learned optimizers remains largely confined to small-scale simulation. Experimental controls are thin: 25 of 57 studies include no classical baseline, the quantum contribution is fully isolated in 10 of 24 studies where an ablation applies, and the median experiment uses 17 qubits. We introduce an M0-M5 evidence hierarchy, from simulation to matched-resource practical advantage, and find no broadly convincing result at the highest level. We argue that scaling is increasingly a systems problem: the question is not only whether a problem fits on a quantum processor, but how classical and quantum resources should be allocated across the optimization process. The review is aimed at researchers in quantum computing, machine learning, and operations research.
arXiv:2610.11646v1 Announce Type: cross Abstract: German is a morphologically rich language whose syllable structure is exceptionally well-predicted by the Knuth--Liang hyphenation algorithm. We ask whether phonologically informed tokenization can serve as a competitive target for end-to-end speech recognition. We compare three tokenizer families on the Omnilingual ASR wav2vec 2.0 backbone fine-tuned with CTC: the pretrained multilingual character inventory, a data-driven Byte-Pair Encoding (BPE) over orthography, and phonologically informed units from Pyphen syllabification and grapheme-to-phoneme conversion. Across 40 fine-tunes, we evaluate on three German test sets spanning orthogonal shifts: in-domain read speech, dialectal spontaneous speech, and standard-German spontaneous speech. In-domain, all phonologically informed tokenizers match BPE and the multilingual character baseline on both WER and CER. Under domain shift the picture splits along vocabulary size rather than the linguistic axis of variation: at small vocabularies, syllable-aware tokenization improves on dialectal speech, where phonetic surface forms vary but syllable structure is preserved, and stays ahead on spontaneous speech, where new word-forms violate vocabulary closure. A phoneme-level confusion analysis further shows that all tokenizers commit the same canonical function-word errors, indicating that the acoustic encoder, not the tokenizer, dominates the error topology. Our findings suggest that tokenizer choice may depend on the vocabulary budget as much as on the distribution shift expected at deployment rather than reducing to a single universal optimum.
arXiv:2610.11584v1 Announce Type: cross Abstract: To test conditional independence of $X$ and $Y$ given a text or an image $Z$, one conditions on an embedding $\psi(Z)$ in place of $Z$. The embedded test is valid if $Z$ is independent of $X$ or of $Y$ given $\psi(Z)$, which cannot be confirmed from data, and when this fails, the rejection probability under the null hypothesis can tend to one. We study this failure, and show that focusing on a specific form of dependence relaxes what the embedding must retain. For a residual correlation test inspired by the Generalised Covariance Measure, validity only requires that the parts of $\mathbb{E}[X \mid Z]$ and $\mathbb{E}[Y \mid Z]$ missed by $\mathbb{E}[X \mid \psi(Z)]$ and $\mathbb{E}[Y \mid \psi(Z)]$ are uncorrelated. Otherwise, we treat the discarded information as an omitted variable. Under the null hypothesis, the bias equals the absolute correlation of the missed parts times the geometric mean of two partial $R^2$ values. This identity yields a robust test valid under a declared tolerance for the geometric mean, which, like a sensitivity parameter, is not identified from the data. On synthetic data and text embeddings, the robust test holds its level approximately. On text generated by a language model, under an exact null hypothesis, every embedding, even the generator's own states, biases the embedded test.
arXiv:2610.11430v1 Announce Type: cross Abstract: While domain-specific Large Language Models (LLMs) have encoded vast biomedical knowledge, their limited context windows often hinder a deep understanding of nuanced relationships within and across texts. To address this limitation, we introduce BioBigBird, a bidirectional language model pre-trained on extensive biomedical literature and clinical data, specifically designed to handle long-range dependencies. BioBigBird leverages a sparse attention mechanism to process sequences up to 4096 tokens, and its training incorporates a multi-stage process to mitigate noise from the large-scale pre-training corpus. We further enhance its performance by employing a multi-task learning (MTL) framework that jointly optimizes for Named Entity Recognition and Relation Extraction. Comprehensive evaluations on the BLURB benchmark reveal that our MTL-enhanced BioBigBird achieves highly competitive results against state-of-the-art models. Our work contributes an effective methodology for developing powerful, long-context language models for specialized domains, demonstrating the value of extended sequence processing for complex text analysis. Our models are publicly available at https://huggingface.co/collections/bisectgroup/biobigbird.
arXiv:2610.11415v1 Announce Type: cross Abstract: Prediction aggregation aims to combine information from multiple predictors into a more informative one. We study this question in the setting of calibrated predictors, where each prediction must equal the conditional expectation of the quantity being predicted given the predictor's signal. Given several calibrated input predictors and the feature distribution, but not the underlying Bayes probabilities, we ask when one can construct refined calibrated predictors that preserve the information in the original predictors and cannot be further refined using the available information. We formulate calibrated predictors as signaling schemes and define refinement through feature-independent garblings: a predictor refines another if its signal can simulate the other's signal. Constructibility is characterized through observable linear information: each signal corresponds to a vector over the feature space, and a new signal is constructible exactly when its vector lies in the linear span of the input signal vectors. Under this formulation, we establish a sharp algorithmic picture. For deterministic output predictors, bilateral refinement admits a polynomial-time algorithm based on a bipartite graph between the two input signal partitions, while refinement with an arbitrary number of input predictors is $\mathsf{NP}$-hard. In contrast, when randomized output predictors are allowed, we give a polynomial-time algorithm for any number of input predictors by decomposing constructible signal vectors into extreme rays of the associated polyhedral cone.
arXiv:2610.11270v1 Announce Type: cross Abstract: Personalized conversational AI relies on long-term memory systems that extract facts from user utterances and store them in persistent vector stores. Despite progress in retrieval, deduplication, and lifecycle management, the formation stage, the moment a fact is first written to storage has received almost no principled attention. We identify this as the binding constraint on memory quality in production systems. Critical contextual signals, such as the distinction between a permanent user attribute and a transient situation, exist only in the original utterance and are irreversibly lost the moment extraction produces a subject-relation-object triple. No downstream process can recover them. We propose Gated Memory, a lightweight, modular formation framework that interposes two decision checkpoints between conversation and storage: an admission gate that evaluates every candidate fact against the full utterance context before extraction runs, and a conditional enrichment stage that grounds admitted facts through an entity scope taxonomy with privacy constraints. The gate evaluates only the current exchange while using prior turns as read-only reference context, and produces a structured formation record. Admitted content is decomposed into atomic facts, each categorized, tagged with provenance (directly stated versus inferred), scoped to its condition of applicability, and grounded in resolved time and place, subject to a constraint that no entity absent from the context may be asserted. On the LoCoMo-10 benchmark with atypical emotional density in utterance data, Gated Memory achieves an overall +2.6% relative improvement in LLM-judge accuracy over a strong baseline with identical retrieval and generation, establishing formation quality as a measurable constraint on memory performance.
arXiv:2610.11261v1 公告类型：cross 摘要：变分量子算法常受到贫瘠高原问题的限制：随着电路规模和深度增加，梯度会逐渐消失，使量子神经网络难以训练。我们提出了 Q-Capsule，这是一种局部化的胶囊式量子神经架构，通过寄存器划分、局部读出、稀疏胶囊间耦合、可训练的数据重上传，以及由量子 Fisher 信息矩阵（QFIM）引导的自适应深度增长来缓解这一问题。通过将每个可观测量的主要作用域限制在较小的胶囊内，并控制胶囊间纠缠，Q-Capsule 在保留局部量子表示之间通信能力的同时，维持了有用的梯度信号。随着寄存器宽度增加，Q-Capsule 始终保持稳定的梯度方差，而全局纠缠基线模型则表现出指数级衰减，其对数梯度方差的斜率接近每个量子比特 -ln 2。Q-Capsule 还能产生结构更清晰的优化景观、更高的参数效率、更强的去极化噪声鲁棒性，以及更低的测量需求。在二分类任务中，其自适应策略达到了 98.1% 的准确率；在四分类任务中达到了 97.7% 的准确率，同时在多分类任务中使用的双量子比特门数量比固定深度模型少约 73%。
arXiv:2610.11218v1 Announce Type: cross Abstract: Dense associative memory (DAM) provides an energy-based framework for memory retrieval with close connections to attention mechanisms in modern artificial intelligence. Despite growing interest in differential privacy for AI, the privacy of DAM retrieval dynamics remains relatively unexplored. In this paper, we develop a differential privacy framework for log-sum-ReLU (LSR) dense associative memory, whose finite-support retrieval dynamics pose distinctive challenges for privacy-preserving computation. We propose the Boundary-Responsive Adaptive Correction Evolution (BRACE) algorithm, a differentially private retrieval mechanism for LSR-DAM that adaptively corrects boundary-sensitive perturbations to control their cumulative effect over the retrieval trajectory. In theory, we prove that our method is minimax optimal by deriving dimension-independent terminal and full-trajectory retrieval error rates, with optimal dependence on the inverse temperature and, in the growing-horizon regime, the retrieval horizon. We further establish central limit theorems that enable uncertainty quantification for private retrieval by characterizing its asymptotic distribution and the additional variability introduced by privacy. Numerical experiments compare our proposed method with baseline differential privacy approaches and evaluate its retrieval accuracy. Together, our results provide a theoretical foundation for optimal privacy-preserving retrieval and uncertainty quantification in energy-based associative memory systems.
arXiv:2610.11139v1 Announce Type: cross Abstract: Anchored Langevin dynamics (ALD) is useful for non-smooth sampling where the density of the target distribution is possibly non-differentiable and heavy-tailed; reflected anchored Langevin dynamics (RALD) can sample possibly non-differentiable target density on a constrained domain. In this paper, we propose and study non-reversible anchored Langevin dynamics (NALD) for sampling possibly non-differentiable and heavy-tailed target density in the Euclidean space and the non-reversible reflected anchored Langevin dynamics (NRALD) for sampling possibly non-differentiable target density in the constrained space. Our construction adds a circulation drift generated by a possibly state-dependent divergence-free skew-symmetric matrix field and a stream potential. It preserves the target distribution without requiring derivatives of target density, admits a random-time-change representation, and applies both on the whole Euclidean space and on bounded domains with normal reflection. By breaking reversibility, we show that NALD and NRALD can converge to their target distributions faster than their reversible counterparts via finite-time non-asymptotic convergence analysis, a large deviations analysis and asymptotic variance reduction. Numerical experiments demonstrate the efficiency of the proposed algorithms.
arXiv:2610.11132v1 公告类型：cross 摘要：监督式微调（SFT）赋予大型语言模型（LLM）专门能力，但往往会以遗忘其父模型（即微调前的预训练模型）的一般能力为代价。对于同时需要专门能力和一般能力的查询，这种权衡尤其具有限制性。我们提出了 SFT-as-context，这是一种无需训练的方法，其中父模型利用 SFT 模型的回答作为上下文来回答查询。通过上下文学习，父模型可以从 SFT 回答中获得微调能力，同时保留自身的一般能力。在 19 对父模型与 SFT 模型以及 11 个基准测试中，SFT-as-context 在微调能力上始终接近 SFT 模型：在 AIME 2024 和 LiveCodeBench 上的差距仅为 2.2 和 2.1 个百分点，在 NutriBench-English 上的宏观 MAE 差距为 2.0；同时，在一般能力上，其平均表现与父模型的差距不超过 2.2 个百分点。值得注意的是，它能够解决同时需要微调能力和一般能力的查询，即使父模型和 SFT 模型单独使用时都无法成功。该方法还可以扩展到父模型与 SFT 模型配对之外：来自小型开源 SFT 模型的回答能够提升强大的闭源 LLM 的表现，优于单独使用其中任一模型。此外，我们使用贝叶斯框架推导出理论保证，分别界定 SFT-as-context 在微调能力上相对于 SFT 模型、在一般能力上相对于父模型的误差。除此之外，我们可视化了注意力权重，发现父模型会更多地关注有用的 SFT 回答，而较少关注无关回答，这表明选择性注意力有助于父模型通过上下文学习利用 SFT 回答。
arXiv:2610.11030v1 Announce Type: cross Abstract: Tool-using LLM agents violate the policies they are deployed to enforce, often silently. Prior defenses hand-write rules, query an LLM verifier per action, or compile policies through heavyweight formal machinery. Naive compilation fails: extracted rules block the tool satisfying their own precondition, or read arguments their tool lacks. NOMOS, a four-pass compiler, turns a natural-language policy into a deterministic tool-call gate; static verification with tool-schema-level checks alone (no prover, solver, or LLM) repairs or rejects 37% (airline) and 13% (retail) of candidates, without which most shipped rules are inoperable. Replaying compiled rules over undefended transcripts flags bindings that refuse legitimate work (a development binding refused 95.9% of task-passing calls); no evaluation binding is flagged. On $\tau^2$-bench the gate cuts violations of reference-encoded clauses among state-changing calls from 66.3% to 2.6% (airline) and 30.8% to 6.9% (retail), raising airline task success significantly for $2 \le k \le 4$; a 26B on-premise compilation is not significantly worse than hand-written or frontier-compiled rules. Unlike AgentDojo's shipped defenses, it reaches a zero attack success rate (ASR) on banking, where nine attack families collapse onto three structural rules. On the other three suites its ASR is at most 3.6%, from goals with no tool call to govern and one write admitted by a binding weaker than its clause; a second agent model, Llama-3.3-70B, reproduces the effect on both benchmarks. Decisions take microseconds without an LLM call, at a domain-dependent benign-utility cost; compilation runs on-premise on open-weight gemma-4-26B.
arXiv:2610.11007v1 Announce Type: cross Abstract: At the start of every session, LLM agents load a fixed context file, such as $\texttt{AGENTS.md}$. Each loaded token in the file is charged again in every later round of the session, and these files can degrade performance as they grow in size. However, in practice, human or automated curators usually grow these files by appending. We formulate context curation as a capacitated assortment problem. Instructions consume tokens under a finite attention capacity; adding an instruction never raises the compliance of the others, while retained instructions incur a per-session setup cost. We prove an upper bound on the optimal file size, regardless of the number of available candidate instructions, and that appending every instruction with positive standalone value can be arbitrarily worse in net value than selecting an optimal subset. A token budget also limits the loss when the token price is underestimated. We then examine what can be learned from past sessions and how this information can guide decisions to add or remove instructions. Feedback is inherently censored: the benefits and harms of loaded instructions are observable, whereas missing instructions generate feedback only when their absence causes harm. In this setting, we show that deleting instructions ignored by agents can inevitably remove helpful ones. We characterize how much evidence should be collected before adding an instruction. Besides, we bound regret when human reviewers can inspect only a limited number of edits per period. Empirical experiments further show that irrelevant rules drawn from real context files reduce language-model compliance.
arXiv:2610.10991v1 公告类型：cross 摘要：图像检索方法通常依赖从图像中提取的单一全局语义描述符，例如视觉 Transformer 中的 [CLS] token。然而，试图将图像的全部语义信息压缩到单个描述符中，可能会损害下游检索性能，尤其是在细粒度检索任务中。在本文中，我们使用精心选择的空间 token 集合，对新型视觉 Transformer 中的语义 token（全局 [CLS] token 和四个寄存器 token）进行增强，旨在捕获刻画每个语义 token 所包含内容的空间区域表示。我们采用 DINOv2-reg 模型，该模型包含能够自发学习物体及其部件表示的寄存器 token。对于每个“提示”token（[CLS] 以及每个寄存器 token），我们找到一个对应的图像“伙伴”图像块 token，并提取一个 N x N 的图像块区域，从而生成一组局部化 ROI token。我们的方法无需任何外部边界框或显著性模块，仅通过将语义 token 与其空间表示区域进行匹配，即可自动捕获重要的感兴趣区域。此外，我们将这些 token 融入受 ColBERT 启发的多向量检索框架，通过逐 token 对齐机制实现细粒度匹配，同时避免存储所有图像块嵌入所带来的巨大存储开销。通过大量实验，我们发现：（1）寄存器 token 编码了有用的细粒度细节，可以补充 [CLS] token；（2）自动池化得到的 ROI token 能进一步提升细粒度区分能力；（3）使用少量 token 的多向量检索性能优于 DINOv2-reg 单向量基线，同时对于大规模搜索仍具备可行性。代码可在 https://github.com/IdhcbIan/Augmenting_CLS_with_ROI_tokens 获取。
arXiv:2610.10945v1 Announce Type: cross Abstract: We present GHARP (Real-time Gaussian Head Animation from Large-scale Reconstruction Prior), a method that animates 3D human heads in real time from a few input images of a subject and a driving expression signal. We decouple the problem into an identity stage that builds a representation of the subject's geometry and appearance offline, and an animation stage that predicts expression-dependent residuals on top of it at runtime. This separation offers a favorable trade-off with respect to fidelity, quality and runtime: the identity stage can be expensive while the animation stage runs a lightweight network, optimized for mobile devices. Our method performs animation in a semantically structured latent space of a pretrained reconstruction model, where expression changes remain spatially contained, making residual prediction efficient. This reconstruction prior provides a consistent spatial layout, allowing fusion of multiple input views into a compact, fixed-size canonical Gaussian representation. While this two-stage design improves the runtime-quality trade-off, it still inherits a problem common to all expression-driven avatar methods: expression codes describe only the face and thus omit body pose and clothing position, making these regions underspecified in the input. The animation network faces an ill-posed mapping and resorts to averaging over conflicting body appearances, producing blur and temporal flicker. We address this with a body alignment network that learns to align the person's body in the target image with the input reference images, removing the ambiguity from the training signal. Our method achieves state-of-the-art quality on the Ava-256 benchmark while running up to 13x faster on an A100 GPU with 8x fewer Gaussians.
arXiv:2610.10919v1 公告类型：cross 摘要：尽管数据质量（DQ）研究已有数十年，但数据质量维度（如准确性或完整性）与数据质量工具之间仍存在鸿沟。文献通常以文本形式定义这些维度，而数据质量工具通常实现的是与这些维度不一致的底层检查。ISO/IEC 25024 和 ISO/IEC 5259 试图通过为每个维度定义数据质量指标来弥合这一鸿沟。然而，这些数据质量指标很少被使用，因为标准没有明确如何实现它们。例如，语法准确性指标统计语法准确的值，但没有说明如何判断一个值在语法上是否准确。这只是将问题转移到了另一个层面，并未真正解决问题。因此，目前的数据质量评估无法以这些标准为基础开展。在本文中，我们使 ISO 数据质量指标具备可执行性。我们将两项标准中的所有数据级指标分为三类：（i）无需数据之外的其他输入即可使用的可泛化指标；（ii）其参数可以从干净的参考数据中学习或由专家设定的参数化指标；以及（iii）需要定性判断且无法自动化的不可泛化指标。对于可以自动化的指标，即类别（i）和（ii），我们提出了实施指南，以解决标准中未明确的问题。我们在 dqmeasure 中实现了这些指南。dqmeasure 是一个包含 20 个指标的开源库，可以从参考数据中学习这些参数，而不是依赖人工定义的规则。我们在真实世界数据集和合成数据集上的实验表明，随着注入错误数量的增加，指标得分会单调下降；指标得分会与下游机器学习性能同步下降；并且指标计算能够随行数线性扩展，从而支持基于这些标准的自动化数据质量监控。
arXiv:2610.10912v1 Announce Type: cross Abstract: Shortcut learning is a prevalent issue in robot learning. The limited diversity of robot demonstration datasets can mislead policies into exploiting spurious correlations between tasks and irrelevant features, such as viewpoint or background. Collecting sufficiently diverse robot demonstrations is costly and inefficient, motivating algorithmic alternatives. We focus on vision-language-action (VLA) models and discover that different vision-language model backbones exhibit substantially different levels of susceptibility to visual shortcut learning. We find that model behavior correlates with our proposed representation-level metric, action margin, which requires no policy rollouts. Visual shortcuts consistently enter action representations in early layers, with models differing in the extent to which later layers correct them by incorporating language information. To boost models' attention to language, we introduce task scrubbing, a new domain-adversarial training method that decreases models' likelihood of using visual shortcuts and improves VLAs' generalization. Experiments in both simulation and the real world across multiple VLAs and visual cues show that task scrubbing improves out-of-distribution robustness and often eliminates visual shortcut learning.
arXiv:2610.10909v1 Announce Type: cross Abstract: Membership inference attacks (MIAs) threaten the privacy of machine learning (ML) training data by determining whether a sample was used to train a target model. Existing MIAs rely on model outputs, ranging from prediction probabilities to predicted labels, an assumption that can be restrictive for on-device ML systems with limited or inaccessible outputs. However, suppressing model outputs does not eliminate the data-dependent computations that produce them, which may remain observable through physical side channels. We present PSCMIA, a power side-channel membership inference attack against embedded ML models that can infer membership directly from power traces without requiring prediction probabilities or even the predicted labels. We evaluate PSCMIA across multiple datasets (MNIST, FMNIST, CIFAR10, CINIC10), fully connected (FC) and convolutional neural network (CNN) architectures, and two embedded platforms (STM32F3, XMEGA). PSCMIA achieves ROC-AUC values of up to 0.907 on FC models. For CNN models, the ROC-AUC gap between PSCMIA and probability vector-based shadow MIA ranges from 0.006 to 0.116. Across the FC and CNN evaluations, PSCMIA outperforms label-only MIA in 11 of 16 model-dataset-hardware configurations, demonstrating that physical execution can expose membership information even when conventional model outputs are unavailable through unintended power side-channel leakage.
arXiv:2610.10870v1 Announce Type: cross Abstract: Markov chain Monte Carlo (MCMC) methods are the standard tool for computing expectations under complex probability distributions. Control variates reduce the variance of the resulting estimates, but a good control variate requires solving the Poisson equation of the sampler, which rarely admits a closed-form solution. Exact solutions are available when the sampler's kernel has a known spectral decomposition on a simple reference density. In our work, we extend these solutions to general targets through a learned change of variables. A bijection, such as a normalizing flow, is trained so that the target becomes close to the reference in a latent space, and we show that Markov kernels and their Poisson solutions are transformed by any bijection. Running such samplers in the latent space then yields explicit control variates, and the estimator is consistent under mild tail conditions on the map and target. Importance sampling (IS) from the flow is the limiting case of the same construction and the control variates apply to it as well. Experiments on synthetic targets and real posteriors compare the procedure against state-of-the-art samplers and control variates.
arXiv:2610.10868v1 公告类型：cross 摘要：我们介绍了 Conversational Voice Aesthetic Model，这是一种语音大语言模型，用于描述自然对话场景中真实或合成语音回复的语音美学特征。给定上下文和回复语音，CVAM 会描述刻画声音特征的显著片段，并预测涵盖性别、音高、语速、情绪和表达方式的九项分类属性。其关键挑战在于情绪和表达方式等感知领域，这些领域本质上具有主观性，且缺乏明确的真实标签。因此，我们从 CANDOR 语料库中获取了 3,000 条真实和合成语音回复，并为每条回复收集了约 10 条人工标注。CVAM 首先基于合成的美学描述和标签进行监督微调，随后利用 Group Relative Policy Optimization 根据人类判断进行优化。实验表明，与 Gemini 3.1 Pro 和开源语音大语言模型相比，CVAM 与人类听众的判断更加一致，并且优于单个人类与其他所有人的一致性。总之，我们证明了以人类感知为基础来定义语音美学的重要性，并提出了一个系统的人类对齐框架。
arXiv:2610.10859v1 Announce Type: cross Abstract: Text-to-image diffusion models are increasingly distilled into few-step variants and being deployed to enable fast inference. However, their ability to generate harmful or undesired content poses significant safety risks. Data-driven unlearning methods suppress targeted generations by fine-tuning model weights using specialized unlearning objectives. Crucially, these objectives implicitly rely on multi-step denoising dynamics, an assumption that breaks down for few-step distilled (FSD) models, resulting in ineffective forgetting. Furthermore, performing unlearning on the non-distilled base model and subsequently re-distilling it to obtain an unlearned FSD model incurs substantial computational and time overhead, making it impractical in many settings. Hence, we address this limitation with a preference-driven unlearning framework that revisits Direct Preference Optimization (DPO) for diffusion models. We show that standard DPO and its unlearning derivatives, formulated around noise-prediction error, transfer poorly to FSD models due to their altered generation dynamics. To overcome this, we introduce a modified preference optimization formulation explicitly aligned with the few-step generation properties, enabling direct concept removal in FSD models while preserving few-step efficiency and maintaining strong retention of desirable (non-targeted) capabilities. We evaluate our framework primarily on identity and NSFW (nudity) removal tasks and also extend our method to object-level unlearning. Extensive experiments demonstrate consistent and effective forgetting, and strong retention performance, establishing our method as a practical and principled solution for unlearning in FSD models.
arXiv:2610.10793v1 Announce Type: cross Abstract: Distributionally robust optimization (DRO) protects decisions against distributional uncertainty by optimizing over an ambiguity set, but poorly aligned set geometry can require large radii and yield overly conservative decisions. We introduce diagnostic-transport DRO (DT-DRO), which uses held-out calibration data to adapt the ambiguity-set geometry to observed predictive errors. DT-DRO uses the conditional probability integral transform cumulative distribution function to diagnose systematic probability misallocation and translates this information into an outcome-level transport that jointly adjusts the ambiguity-set center and ground cost. The resulting formulation admits a computationally tractable dual reformulation. Theoretically, we derive valid ambiguity radii and decision-risk guarantees that tighten as estimation and approximation errors vanish, and show that DT-DRO can eliminate the nonvanishing robustness floor caused by model misspecification. Synthetic experiments and a power-outage application demonstrate improved decision quality, particularly under structural and tail misspecification.
arXiv:2610.10766v1 Announce Type: cross Abstract: Lack of effective authentication has resulted in numerous security and privacy breaches, including unauthorized access to protected information, identity theft, and fraud. One approach to mitigating such attacks is Multi-Factor Authentication (MFA), in which users must provide multiple pieces of information for authentication. Some secondary authentication factors include SMS text verification codes, biometrics, and tokens. Each contains at least one notable flaw: SMS is notoriously insecure; biometrics rely upon access to sensitive personal data; and tokens require dependence on third-party providers (e.g. OAuth providers). This work explores CPU-Auth, a novel authentication mechanism based on unique variations in the physical characteristics of the CPU of a computing device. By measuring the behavior of the Dynamic Voltage and Frequency Scaling (DVFS) governor remotely from within a browser, unique properties of the CPU can be leveraged to establish a hardware-based device fingerprint for use in CPU-Auth. The performance of CPU-Auth is evaluated on over 50,000 data traces using distance-based and deep learning methods. CPU-Auth is part of a larger research project, and the results provided in this report reflect only the contributions made to the project by members of this group.
arXiv:2610.10761v1 发布类型：cross 摘要：线性注意力是一种易于处理的模型，可用于理解 Transformer 中支配上下文学习的机制。对于线性回归任务，近期的渐近分析刻画了其学习行为和泛化行为。我们将这一理论扩展到非线性单索引目标 \(y=f(x^\top w)+\varepsilon\)。我们的主要结果确立了非线性—噪声等价性：线性注意力只能提取 \(f\) 的线性 Hermite 分量，而其余非线性结构则作为有效噪声，对泛化误差产生贡献。这一化简使得相应线性理论的结果能够迁移到非线性任务中。我们展示了该结论对有限预训练数据的影响，以及随着任务多样性增加，从任务记忆到任务泛化的转变。这些结果揭示了简化线性注意力模型的局限性，并为研究非线性上下文学习提供了一个易于处理的起点。
arXiv:2610.10697v1 Announce Type: cross Abstract: Accurately simulating chaotic quantum many-body dynamics remains a major computational challenge for classical methods, due to the rapid buildup and spatial spreading of entanglement during the evolution. This raises the question of whether machine learning can provide an effective alternative for predicting quantum dynamics. We address this question by formulating quantum dynamics as a time-series forecasting problem, using a few-qubit PXP chain, realizable with Rydberg-atom arrays, as a benchmark. By varying the initial state, the system spans dynamical regimes ranging from ergodic behavior to quantum many-body scarring, providing a controlled setting for testing forecasting models across qualitatively different dynamics. We compare two contrasting architectures: an expressive nonlinear Transformer and DLinear, a simple linear forecasting model. The Transformer accurately predicts dynamics in the more ergodic regime, but its performance progressively deteriorates as the initial state approaches the scarred limit. In contrast, DLinear remains accurate across the entire family of initial states, with its main deviations consisting of small high-frequency oscillations that have little effect on the overall prediction error. Remarkably, these results show that observables generated by complex quantum many-body dynamics can be forecast with high accuracy through a simple linear mapping from past to future observations. This reveals that the complexity of the underlying quantum evolution need not translate into an equally complex forecasting problem.
arXiv:2610.10663v1 Announce Type: cross Abstract: Bioacoustics and ethology encompass a wide range of audio understanding tasks, many of which stand to benefit from recent advances in large audio-language models. However, progress in the field has so far been assessed on a narrow set of tasks, primarily centered on label-centric biological category recognition, such as species and call-type classification. In this work, we introduce BEANS-Next, a benchmark grounded in a taxonomy of bioacoustics tasks spanning acoustic perception, biological category recognition, scene understanding, and in-context learning. Using BEANS-Next, we show that existing models exhibit limited performance beyond the task families emphasized by existing evaluations, constraining their usefulness for broader bioacoustic applications. To support progress on this broader task space, we also introduce ROOTS, a large-scale training resource built from expanded curated real-world data and previously underused behavioral and acoustic metadata, supplemented by audio-derived information and scalable synthetic generation where labeling is insufficient. We demonstrate that training on this dataset yields substantial progress across all task groups of BEANS-Next, moving audio-language models closer to their potential as general-purpose assistants for bioacoustics and ethology. To accelerate progress in the field, we open-source our benchmark, dataset, and data pipelines.
arXiv:2610.10615v1 Announce Type: cross Abstract: Choosing which information to observe is central to prediction under limited observation budgets. We study JevForest, a feature acquisition policy that aggregates path-dependent proposals from bootstrapped trees, weights them by global training information gain, and predicts from the acquired values with a shared masked classifier. An online implementation queries Jev for semantic answers selected by this policy. On small balanced held-out samples, four-question forest acquisition achieves accuracy $0.729$ on AG News ($n=48$), compared with $0.667$ for a static gain ranking and $0.583$ for random ordering. On TREC ($n=24$), the ordering reverses: forest accuracy is $0.667$, compared with $0.750$ and $0.833$. Asking all eight questions in one batch yields higher accuracy at lower measured cost and latency than four sequential forest queries; direct Jev classification matches the batch accuracy while costing less. Offline MiniBooNE experiments yield accuracy $0.845\pm0.010$ at ten features and $0.885\pm0.008$ at forty features over three jointly varying data and forest seeds (mean $\pm$ sample standard deviation). A companion Newton boosting implementation provides preliminary full-feature synthetic results. These exploratory findings establish a working Jev acquisition workflow but do not support a general advantage for path voting: its value depends on the task, predictor, and the distinction between question budgets and actual query costs.
arXiv:2610.12445v1 Announce Type: new Abstract: Recent incidents have highlighted the challenge of monitoring LLM agents and the danger of models deceiving people. We show that white-box deception detection via probes can be scaled up to frontier monitoring settings by collecting the largest deception dataset to date for training probes and introducing a novel probe architecture which can aggregate information across many layers and tokens. Our probes achieve 98.8% AUC in SHADE-Arena, surpassing an Opus 5.5 text-monitoring baseline, and show improved efficacy as the underlying model is scaled up. To push our probes to their limit, we test them on several cases where deception cannot be determined from the context alone. In these cases, which we refer to as introspective deception, the ground truth can only be determined through careful elicitation or thorough knowledge of a model's training data. In one such evaluation, we show that probes can distinguish transcripts containing a model's true hidden goal from other goals with an AUC of up to 99.7%. Our probes also readily detect deception on prominent open-weight models which lie about politically sensitive topics, and about their beliefs when put under pressure. We release our training dataset, dubbed FIBS, to help drive frontier deployment of effective probes, and encourage the community to expand upon it with further examples of deception and sabotage.
arXiv:2610.12444v1 Announce Type: new Abstract: Quantizing AdamW's optimizer states reduces persistent storage, but quantization errors propagate through the moment recurrences and perturb subsequent adaptive updates. We redesign 4-bit optimizer-state quantization for AdamW from the perspective of \emph{rounding space}: the coordinate in which a quantizer chooses between adjacent reconstruction levels. For the second moment, a local analysis of the quantization cell adjacent to zero shows that small mean state error need not imply small mean preconditioner error at the next step. A one-dimensional quadratic construction further shows qualitatively different optimization dynamics under state-space and preconditioner-space rounding. These results motivate Zero-Inclusive Preconditioner-space Stochastic Rounding (\textbf{ZIP-SR}), which retains zero in the second-moment codebook and computes stochastic-rounding probabilities in preconditioner space. As a complementary route, Zero-Excluding EDEN calibration (\textbf{ZE-EDEN}) uses a zero-excluding second-moment codebook and rescales the quantized second-moment block to mitigate the preconditioner distortion caused by the positive quantization floor. Both configurations use 4-bit NormalFloat (NF4) for the first moment, with targeted stochastic rounding of the LM-head first moment during the final 10\% of training. Across GPT- and Llama-style pretraining experiments ranging from \textbf{130M} to \textbf{2.7B} parameters, both methods reduce TorchAO 4-bit AdamW's mean validation-loss gap to 32-bit AdamW at every evaluated model size, with the largest reported gap reduction reaching \textbf{70\%}. In full-parameter supervised fine-tuning, both recipes achieve lower validation loss than TorchAO while remaining close to 32-bit AdamW on downstream tasks.
arXiv:2610.12420v1 Announce Type: new Abstract: Reinforcement learning in safety-critical domains requires maximizing task performance while strictly adhering to safety constraints. Existing safe reinforcement learning paradigms typically force a trade-off: they either require a priori knowledge to provide strict safety guarantees (e.g., safety filters), or they enable joint learning but only satisfy safety constraints on average. In this work, we propose a novel Bellman operator that unifies performance and safety objectives into a joint value function. We show that temporal difference learning with the joint Bellman operator converges under a two-timescale stochastic approximation framework. On the fast timescale, the safety value of the learning joint policy is estimated, while the joint value is estimated on the slow timescale. Convergence is ensured by formulating the limiting dynamics as an occupation-averaged differential inclusion, and showing that it asymptotically converges to a set of limiting optimal safety-constrained task value functions. Theoretically, once converged, the resulting optimal policy maximizes task return while maintaining safety at all times. Empirical evaluations on continuous control tasks with neural approximations demonstrate stable convergence with near-zero safety violations at test time.
arXiv:2610.12370v1 Announce Type: new Abstract: Koopman operator theory provides a linear framework for analyzing nonlinear dynamical systems and has become a major tool for data-driven modeling. A central challenge, however, is that finite-dimensional approximations computed by methods such as extended dynamic mode decomposition (EDMD) require the dictionary to be specified a priori. Recent machine-learning approaches address this limitation by learning the dictionary from data, predominantly using artificial neural network (ANN) autoencoder architectures. Although kernel methods offer an alternative with greater interpretability and tractability for theoretical analysis, they have received little attention in this setting. We introduce extended dynamic mode decomposition with kernel-based dictionary learning (EDMD-kDL), a kernel-based method for learning finite-dimensional Koopman embeddings directly from data. The method combines ideas from collocation methods and bilevel optimization to simultaneously learn a kernel dictionary and the corresponding Koopman approximation. We evaluate EDMD-kDL against state-of-the-art ANN-based approaches on a range of numerical experiments, including global sea-surface-temperature forecasting and learning directly from video data. Across all tested settings, EDMD-kDL achieves performance comparable to or better than the ANN-based methods. Moreover, in contrast to standard kernel methods, the proposed approach is scalable to large datasets by design since the size of the required kernel matrices depends on the number of collocation points rather than the size of the training dataset.
arXiv:2610.12244v1 Announce Type: new Abstract: Isolation Forest (IF) is a widely used baseline for unsupervised anomaly detection. Recent studies provide a closed-form expression for the infinite-forest limit for one-dimensional data. Inspired by the geometric interpretation of this formula, we introduce RIFT (Relative Isolation From Trees), a deterministic anomaly detection method that generates the minimum spanning tree and scores each point by the sum of the apparent sizes of tree edges as viewed from that point. For one-dimensional data, the RIFT score recovers the closed-form IF limit exactly. In higher dimensions, it provides a parameter-free generalization that is deterministic, robust to varying density and clustered anomalies and avoids the axis-parallel artifacts of IF. We further propose an ensemble variant for large datasets. Experiments on synthetic data and the ADBench benchmark demonstrate that the accuracy is comparable to IF, while the ensemble variant exhibits significantly lower variance across random seeds.
arXiv:2610.12240v1 Announce Type: new Abstract: Time-series foundation models (TSFMs) have achieved strong forecasting performance across domains. However, most adaptation methods remain static. Existing all-in-one methods learn a single set of dataset-level parameter updates and apply the same adapted model to every input. As a result, they cannot adapt the model parameters to the temporal patterns, seasonality and dynamics of each input time series. This limits their ability to produce forecasts that are tailored to heterogeneous inputs. To address this limitation, we propose AdaCast, a conditional parameter generation framework for time-series forecasting. AdaCast uses a generator to produce input-specific low-rank parameter updates for a frozen pretrained TSFM. These updates adapt the model to each input during both training and inference. Across six public benchmarks, AdaCast consistently outperforms static adaptation baseline in in-domain forecasting and improves zero-shot generalization to held-out datasets across domains. These results demonstrate that conditional parameter generation provides an effective approach for adaptive forecasting.
arXiv:2610.12190v1 Announce Type: new Abstract: As claims about recursive self-improvement (RSI) and artificial general intelligence (AGI) proliferate, we ask a simple question: do frontier AI models have a sense of data, i.e., can they reliably select the right data for training? We introduce DataSense-Bench to study this capability through the fundamental problem of data selection and performance forecasting in machine learning. We ask AI agents to select and rank candidate training subsets that can be used to fine-tune a small LLM model. Agents are allowed to inspect the data, write and execute analysis code, and run model forward passes, but can not train the model or access the actual evaluation tasks. We then fine-tune the base model on each selected subset and evaluate its post-training performance under a standardized protocol. We instantiate the benchmark in terminal problem solving and tool use, selecting trajectories from OpenThoughts-Agent and EnvScaler and evaluating on TBLite and BFCL, respectively. We then evaluate the agents along two complementary dimensions: the post-training performance of the top-ranked subset, reflecting the ability to identify high-value training data, and ranking accuracy, reflecting the ability to predict the relative performance of the selected subsets. In our experiments, selection gains over random selection are limited; agents do not reliably rank their selected groups, and ranking ability does not hold consistently across tasks: Astra identifies the best group in all three tool-use runs but in only one of three terminal runs. Analysis of execution traces on both tasks shows that agents often use similar data signals while interpreting their training value differently.
arXiv:2610.12183v1 Announce Type: new Abstract: Black-box optimization (BBO) arises in many scientific and engineering problems where objective evaluations are expensive and limited. Recent large language model (LLM) agents offer a new way to approach BBO by combining task semantics, computation, optimization tools, and feedback-driven decision making, showing great potential due to the integration with mathematically rigorous tools. However, existing agentic BBO studies use different task domains and system configurations, making their results difficult to compare and the effects of individual design choices hard to isolate. We therefore introduce AgenticBBO-Bench, a cross-domain benchmark for agentic BBO spanning synthetic functions, hyperparameter optimization, database tuning, chip design, and molecular design under a unified finite-budget evaluation protocol. In our experiments, agentic BBO achieves higher family-averaged scores than direct LLM-based methods in all five domains and outperforms the best numerical optimizers in four. We further study three factors shaping agent performance: optimization tools, task information and prior knowledge, and the role of the LLM during search. Our results show that additional numerical tools do not consistently improve performance, task semantics are broadly useful while more specific priors are less reliable, and numerical optimizers can effectively absorb gains from search trajectories established by the agent. Finally, we introduce a five-task frontier challenge within AgenticBBO-Bench and evaluate seven LLMs under the Codex agent harness, where GPT-6 Astra and DeepSeek-V4.1-Flash lie on the Pareto frontier of performance and cost among the evaluated models. Our code is available at https://github.com/lamda-bbo/agentic-bbo.
arXiv:2610.12096v1 Announce Type: new Abstract: Neural network-based predictive modeling with high-dimensional structured Gaussian targets requires an efficient and numerically stable, yet expressive approximation of the covariance matrix. We propose SCORE: a scalable framework, combining scoring rule training with an expressive covariance approximation learned in spectral space. For $d$-dimensional data, the learning task is decomposed into learning the marginal distributions and learning a structured correlation matrix, which enables dense dependencies with linear storage and $\mathcal{O}(d\log d)$ cost. We utilize the closed form Gaussian kernel score for training, which remains defined even for degenerate covariances and admits bounded gradients during optimization. We characterize kernel scores under invertible transforms and prove exact invariance under unitary transforms. At population level, our two-level objective recovers the true marginals and projects the target correlation onto the representable class; finite-sample PAC bounds show that the errors of the two stages enter additively. We evaluate our model on a variety of tasks with a commonly assumed Gaussian domain: Time-series forecasting, monocular depth estimation, and spatial weather prediction, showing improved performance at lower computational cost.
arXiv:2610.12005v1 公告类型：新 摘要：哪些形式的测试时计算能够改善强大的预训练表格基础模型（TFMs）的预测？我们从三个轴向对此进行了系统研究：适应、聚合和上下文构建。评估涵盖 TabArena 基准测试中的现代 TFM，并辅以在 OpenML 上的宽表和大规模表格实验。在适应方面，我们提出了 DiagScale，一种对角查询—键相似度更新方法。该方法仅训练模型参数的 0.003%–0.03%，却在三个独立预训练的骨干模型上取得了与全量微调相当的增益。在聚合方面，候选池的构成和选择策略都很重要。TabPFN-3 已经对同一数据的不同预处理变体进行预测并取平均，继续增加此类预测带来的收益会逐渐递减。在包含 96 种配置的更大候选池中，贪心选择相较于默认预测器可将误差降低 2.4%，但均匀平均反而会增加误差。在上下文构建方面，注意力引导的检索改善了 TabPFN-3 在部分大型表格上的预测，并支持使用超出完整上下文记忆容量限制的来源池。我们测试的上下文扩展方法未带来一致的改进。总体而言，我们的结果表明，适应和选择性聚合能够带来稳定的基准级增益。在同一骨干模型上结合适应与聚合还能获得进一步提升，但其所需计算量显著高于默认推理。这些权衡促使人们根据可用的计算预算选择相应策略。代码可在 https://github.com/kanghui-learning/test-time-compute-for-tabular-foundation-models 获取。
arXiv:2610.11854v1 Announce Type: new Abstract: Reinforcement learning (RL) methods such as GRPO substantially improve large language model reasoning but often suffer from policy entropy collapse: the loss of sampling diversity weakens exploration and limits further improvement. Existing methods address this issue either through algorithm-level interventions, such as reward modification and entropy/KL regularization, or through token-level reweighting. We investigate a complementary perspective: entropy collapse can also be mitigated by changing which generated rollouts contribute to policy updates. Under the same sampling budget, not all rollouts contribute positively to an update, and selectively excluding some can improve learning. To address this, we propose GRPODropout: before the standard update, we use a simple strategy that selectively removes a small number of high-probability positive-advantage rollouts and recenters the retained advantages. To motivate this design, we develop a rollout-level theoretical analysis that guides method design and threshold selection. The method changes only rollout usage, and adds negligible computational overhead. Experiments show higher accuracy than original GRPO and higher actor entropy while using fewer rollout samples for updates, illustrating "less is more." This work provides insight into RL rollout usage: removing some rollouts can improve performance. Code is available at https://github.com/hexuandeng/GRPODropout/.
arXiv:2610.11784v1 Announce Type: new Abstract: Tabular in-context learning (ICL) has emerged as a training-free and accurate paradigm for tabular prediction, but current approaches to compressing its in-context examples face an accuracy-throughput tradeoff: fixed subsets can sacrifice accuracy, while query-specific retrieval limits cache reuse and batching across queries, reducing throughput. We propose QCOC (Query-Calibrated Operator Compression), which exploits the exchangeability and repeated use of in-context examples by compiling their full KV cache once into compact memory shared across subsequent queries. Instead of retaining raw examples, QCOC clusters their states into joint-KV prototypes, preserves per-cluster multiplicities and the original example count, and calibrates prototype values against attention query vectors produced by the in-context examples through an anchored closed-form solution. Prototype compression drives the speedup, while value fitting helps preserve accuracy. On 64 held-out OpenML-CC18 datasets, QCOC achieves the highest mean accuracy among the compared compression and retrieval methods at both retained counts. Across 12 configurations on seven long tables, it ranks first among compressed methods in ten and averages 0.23 percentage points below full context. Compressing 8,192 in-context examples to 512 memory slots yields a 10.5x cache compression ratio; excluding one-time compilation, in a single-core CPU online-serving comparison over 1,000 queries, QCOC is up to 508x faster than dynamic retrieval baselines and 1.98x faster than full-context inference. These results show that QCOC enables compact-memory reuse and efficient inference across queries while retaining accuracy close to full context.
arXiv:2610.11783v1 Announce Type: new Abstract: Shared e-scooter services have become a widely adopted urban transport mode. While most users ride responsibly, alcohol intoxication stands out among the factors contributing to severe crashes. Nonetheless, countermeasures remain limited to single-point reaction tests and night bans that suspend the service altogether. This paper proposes a new approach in which onboard sensors evaluate the rider as the trip unfolds, raising an alarm as soon as enough evidence of impairment has accumulated. Specifically, we introduce a detector that operates on inertial and throttle measurements, with a provable bound on the rate of false alarms. Experiments on sensor data from 141 rides, in which 25 participants rode while sober and at two target blood alcohol concentration levels, confirm that the bound holds, whereas baselines and ablations either exceed it or lose detection performance, and in some cases delay the alarm. At a bound of 0.023, the detector identifies 91% of the rides performed at the higher concentration and 50% of those at the lower one, with median detection times of 25 and 27 seconds, respectively. We further show that an embedded implementation meets the real-time requirement, making mitigation actions feasible onboard, without requiring data to leave the vehicle. Overall, this work lays the ground for interventions that reach impaired riders as soon as possible, sparing the sober ones the burden of a pre-ride test or the suspension of the service at night, while letting operators budget false alarms against user experience.
arXiv:2610.11761v1 Announce Type: new Abstract: The genome holds the blueprint that governs the biological properties of the cell. Consequently, advancing our knowledge of genomic function is crucial both for a broader understanding of biology and for continued biomedical advances. The success of large language models on natural language and protein sequences has motivated similar efforts on genomic data. However, standard genomic language models (gLMs) often require extremely large computational resources and still fall behind traditional methods on some downstream tasks. Recently, MSA-based pretraining has been proposed as an efficient alternative, but existing models are limited to short input contexts, restricting their use to short-range tasks, such as variant effect prediction. In this work, we present RAGenome, the first retrieval-based gLM that scales pretraining to longer contexts (100$\times$ longer than existing MSA-based gLMs), allowing it to capture both across-species evolutionary relationships and within-species longer-range interactions. Trained on whole-genome alignments from 100 vertebrates, RAGenome substantially improves the long-range capabilities of MSA-based gLMs, raising gene finding performance from 0.45 to 0.60, while remaining competitive on purely evolutionary-based tasks like prioritizing pathogenic variants. RAGenome provides competitive gLM performance at a fraction of the training cost, unifying evolutionary modeling and long-range capabilities within a single, flexible, scalable framework. Code is available at https://github.com/PanosAntoniadis/RAGenome.
arXiv:2610.11734v1 Announce Type: new Abstract: We introduce Timer-M1, a pretrained multivariate time series foundation model that learns with primitives for zero-shot forecasting. Across domains, time series share elementary temporal and relational patterns, termed primitives, yet differ in how these primitives manifest and evolve across different contexts. Despite progress in zero-shot and task-general forecasting, existing foundation models may still struggle to generalize to complex real-world scenarios. To this end, we develop a primitive-based data synthesis and pretraining pipeline. The synthesis pipeline generates series with temporal primitives shared across domains and then assembles real and generated series into multivariate samples using relational primitives. Afterwards, samples are organized into episodes by assigning distinct channel roles as target variates, past-only covariates, and known-future covariates, ensuring that the model is optimized on predictable variates using available exogenous information. Technically, Timer-M1 further adapts gated two-dimensional Transformer blocks that dynamically allocate cross-variate attention across layers. Across three large-scale forecasting benchmarks, Timer-M1 ranks first on both FEV and TIME and second on GIFT-Eval among most recent time series foundation models. These results support effective primitive-based pretraining as a route to robust general forecasting technique across domains and task settings.
arXiv:2610.11730v1 发布类型：新 摘要：标准权重衰减将每个权重矩阵视为向量，忽略了其谱结构。我们引入谱权重衰减，这是一种在参数更新步骤之后执行的解耦核范数更新，采用加法而非乘法的谱收缩。我们将该更新与近似近端下降联系起来，并表明，在接近秩亏的情况下，其对更新顺序的敏感性可能超过传统的 $\ell_2$ 权重衰减。在参数量从 $124$M 到 $500$M 的 LLaMA 模型上，谱权重衰减在验证损失相当的情况下降低了有效秩，并改善了 SVD-LLM 压缩效果。在 $500$M 参数规模和 $4\%$ 的失真预算下，该方法实现了 $1.89\times$ 的压缩率和 $1.18\times$ 的 GPU 推理加速，而标准权重衰减后的结果分别为 $1.14\times$ 和 $1.01\times$。在固定训练时长且标签噪声率为 $60\%$ 的训练设置下，相较于匹配的 $\ell_2$ 正则化，该方法还提升了最终平均干净测试准确率：在 MNIST 上最高提升 $17.8$ 个百分点，在四个 BERT-base 任务上提升 $4.6$ 个百分点。代码可在 https://github.com/brain-lab-research/SpectralWD 获取。
arXiv:2610.11724v1 Announce Type: new Abstract: Multimodal meme understanding is increasingly used to analyze socially sensitive content, yet existing models often exhibit biased behavior when interpreting economic dependence and social roles under ambiguity. Many memes express economic relationships through sparse text or symbolic visual cues, providing insufficient evidence for gendered attribution. In such underspecified settings, models tend to rely on pretraining correlations, leading to hallucinated and stereotypical economic role assignments. In this work, we study gendered economic dependence in image-text memes through the lens of contextual sufficiency and identify epistemic overcommitment-inferring roles without adequate evidence-as a primary source of bias. We propose CGER-Net, a context-grounded multimodal framework that estimates whether the input provides sufficient evidence for gendered economic reasoning and applies evidence-gated inference to enable confident attribution when cues are explicit while favoring principled abstention otherwise. We evaluate CGER-Net on EconMeme-GE, a curated dataset of image-text memes annotated as Men, Women, Neutral, or Ambiguous. Across strong contemporary multimodal baselines, CGER-Net reduces Gender Overcommitment Rate by up to 44% on ambiguous instances while maintaining comparable accuracy on unambiguous cases. Human evaluation further shows that 79% of generated rationales are judged as epistemically aligned with the available evidence. These results highlight the importance of modeling when not to infer for reliable and responsible multimodal analysis.
arXiv:2610.11692v1 Announce Type: new Abstract: Foundation models supply reinforcement learning (RL) with priors that mitigate its longstanding weaknesses in sample efficiency and transfer, but their token-by-token generation makes queries sequential and costly. Jev, a recently released decision model, generates nothing and returns calibrated, typed answers in a single forward pass. Existing work studies foundation models in RL either as models to be trained or as generators to be prompted, and Jev belongs to neither category, having so far served only as a black box in single domains. How well such a model decides on its own in RL environments, and how it can improve RL as a component of training, therefore remain unaddressed. To this end, in this paper we first examine the requirements that the objects of an RL system place on the answers they consume, and establish that Jev can fulfill all of them except the cardinal use of a value function. The remaining objects form positions that admit several roles each. We then construct algorithms that employ Jev at three of these positions, as a reference policy, an exploration judge, and a replay rater, to improve sample efficiency, exploration, and learning performance. Across nine MiniGrid tasks and three Atari games, training with Jev outperforms a standard RL learner, including where the learner makes no progress alone, while the model itself remains untrained. To our knowledge, we present the first use of Jev within the RL learning process and establish a frozen decision model as a usable component of RL training, inviting further exploration of how Jev and other advanced decision models can improve RL.
arXiv:2610.11605v1 公告类型：新 摘要：我们推出了 NanoProof，据我们所知，这是 Lean 4 中首个因子化执行引导式定理证明器，其训练数据、提取工具、训练流程和权重均已发布，因此可使用开源资源实现端到端复现。为此，我们构建并发布了一个结构化证明树数据集，以及一个用于在 Lean 4 形式化验证器中进行程序化交互和数据提取的工具。为支持可持续研究，我们专注于计算效率，以促进易于获取的训练与评估。NanoProof 在 MiniF2F-Test 上达到 50.8% 的 pass@16，其性能超过同类中最接近的两个系统 HyperTree Proof Search 和 ABEL，而所需计算量约分别少 90 倍和 7 倍；与 AlphaProof 相比，所需计算量则少了四个数量级以上。虽然存在更强的开放权重定理证明器，但它们都是基于大型预训练语言模型进行微调的，且既不发布训练数据，也不发布训练流程；NanoProof 表明，因子化执行引导式定理证明器这一类别可以利用适度资源从头重建。
arXiv:2610.11548v1 Announce Type: new Abstract: Synthetic tasks are increasingly used both as probes of language-model capability and as pretraining data. Both uses are often justified by loss reduction: falling loss is treated as informative, and faster loss reduction with more sampling as evidence that a task is worth sampling. We separate three signals. A task is diagnostic when its loss tracks global pretraining progress; it is teachable when its loss responds to its own token budget; and a data source transfers when including it improves a downstream target. We study controlled pretraining in which 70% of the corpus is fixed general Python and the remaining 30% is a simplex over three source families: OpenCodeInstruct, a curated suite of 12 code-adjacent synthetic tasks, and 15 literature-derived probe tasks. Across a task-budget sweep we detect teachability for 14 of 27 tasks, with a sharp asymmetry between the two synthetic families (10/12 curated versus 4/15 literature-derived). Teachability and downstream transfer give different rankings. On the mixture-simplex edge between the curated suite and OpenCodeInstruct, HumanEval pass@20 after a fixed fine-tuning stage rises from 15.9 at pure curated data to its highest observed value, 22.6, at a mixture that is 75% OpenCodeInstruct, then falls to 19.5 at pure OpenCodeInstruct. Curated synthetic data therefore has conditional value: it contributes as a limited share of a mixture that a target-aligned source still dominates. Finally, a loss-based adaptive scheduler exposes the mismatch between residual loss reducibility and downstream transfer. Across three 60k-step free-ratio runs, Ado drives the OpenCodeInstruct share below 5% within the first 5k steps and to 1.2--1.4% by the end of training, and underperforms its matched fixed-mixture controls by 2.4--11.0 percentage points. Optimizing near-term task-loss reduction moves the mixture away from the region that transfers.
arXiv:2610.11523v1 Announce Type: new Abstract: Federated reinforcement learning (FRL) enables distributed agents to collaboratively train decision-making policies, but its decentralized training process also exposes global policy learning to Byzantine manipulation. Existing poisoning attacks primarily focus on how to construct malicious updates, while trajectory-level intervention timing remains largely implicit. In sequential decision making, however, where an intervention is applied can alter subsequent trajectories and learning signals. Through controlled experiments, we find that changing the selected trajectory states materially alters attack efficacy even when the malicious-update construction is fixed. We therefore identify when as a distinct attack dimension and introduce the Viability-constrained Behavioral Steering Attack (V-BSA), which uses local policy uncertainty to select sparse intervention states and applies envelope-constrained behavioral steering. Across discrete-action benchmarks, V-BSA achieves substantial degradation against robust aggregators and ensemble defenses with only a fraction of the interventions used by dense poisoning, while revealing task- and aggregation-dependent boundaries. Overall, our results highlight intervention timing as a distinct dimension of sequential robustness in FRL. The code is available at https://github.com/Yodeesy/V-BSA
arXiv:2610.11475v1 Announce Type: new Abstract: Population gradient flow is a common tool for reasoning about how neural networks adapt, including after pretraining. We show that it can mispredict finite-batch stochastic gradient descent (SGD) qualitatively, and we trace the discrepancy to a specific mechanism. In a two-unit ReLU regression, a source task drives the two neurons toward positive proportionality and a target task rewards separating them. After source training for time $T$, gradient flow recovers on the target in time linear in $T$. Online SGD with batch size $b$ and step size $\eta$ in both phases instead fails with high probability throughout a horizon of order $e^{c/\eta}$ once $T \gtrsim \log(b/\eta)$, uniformly on an explicit set of initializations with Gaussian probability above one percent. For each fixed $T$, small-step SGD still recovers, so the failure requires the joint limit of small steps and long pretraining. At the target clone, the population instability is carried entirely by inputs on which the two ReLU gates disagree. For units at angle $\delta$ these inputs form a wedge of probability $\delta/\pi$, and weight decay shrinks the angle exponentially during pretraining. On every other input both units receive the same random linear update, which contracts their separation in conditional expectation. Bounding the cumulative probability of sampling the wedge along the exact online recursion, without a diffusion approximation, shows that recovery with fixed probability from an identical source-gradient-flow checkpoint, within $e^{c/\eta}$ updates, requires $Nb \gtrsim e^{\lambda T}$ target samples and batch size $b \gtrsim \eta e^{\lambda T}$, where $N$ counts updates and $\lambda$ is the weight decay. In simulations, recovery is approximately a function of the disagreement budget $b\delta/\eta$ and saturates in the horizon.
arXiv:2610.11408v1 公告类型：新论文 摘要：公开的移动与可穿戴心理健康数据集通常提供汇总特征表，而非同步的原始传感器流。在这些发布的数据中，每个锚点对应一次调查或标签的时间，并可能结合手机或可穿戴设备的汇总信息、既往症状评分、人口统计信息、临床变量以及来源可用性指标。我们提出 Modality-Conditioned Temporal Recursive Context Model（MC-TRCM），将每个特征来源保留为独立的词元，并将缺失情况纳入输入上下文。对于已观测到的来源，使用其数值和缺失情况摘要进行编码；对于缺失的来源，使用可学习的缺失词元；数据集和任务嵌入为融合提供条件；递归预测头则按照通过验证集选定的步数，逐步细化各项输出。我们采用参与者级别的数据划分，并仅基于验证集进行模型选择，在 DepreST-CAT 和 Prediction of Severity Change-Depression（PSYCHE-D）的六个预定义终点上评估了 MC-TRCM。MC-TRCM 在 DepreST-CAT 的患者健康问卷-9（PHQ-9）和广泛性焦虑障碍量表-7（GAD-7）严重程度预测中取得了最低的平均绝对误差，相较于最佳表格数据基准模型，分别降低了 0.181 和 0.217 个量表分值。分类终点的表现因任务而异：MC-TRCM 在 GAD-7 类别预测中取得了与最佳结果相同的舍入后平衡准确率；在 PSYCHE-D 多分类预测中，其平衡准确率在数值上最高，领先 0.002；在 PHQ-9 类别预测和 PSYCHE-D 二分类预测中，其表现仍接近最强基准模型。消融实验支持了逐特征线性调制、缺失词元、缺失情况投影和递归细化的作用，而校准分析和特征来源对照实验则刻画了各终点的表现特征。我们的代码已公开于 https://github.com/Botwwt/MC-TRCM。
arXiv:2610.11394v1 公告类型：新论文 摘要：多模态阿尔茨海默病（AD）诊断受益于异质性临床、影像、基因组和生物标志物证据的整合，但临床队列中经常存在不规则的模态缺失。现有融合方法往往合成缺失的输入，面临引入人工替代数据的风险，或将可用信号汇集到不可解释的潜在空间中。我们提出 CoPoE（疾病坐标专家乘积），这是一个疾病坐标框架，将多模态证据映射到结构化潜在空间，并将该空间划分为四个不同的生物学和临床轴：遗传风险、分子病理、神经退行性变和临床分期（R/P/N/S）。每个已观测模态都对覆盖完整 RPNS 向量的对角高斯专家进行参数化，而带掩码的专家乘积架构仅融合可用模态。因此，缺失模态不会向融合路径添加任何因子，使网络能够针对任意非空模态子集保留稳健、可分解的后验分布，而无须在 RPNS 路径中进行合成插补。通过在 ADNI 数据集上开展广泛的模态缺失实验，在共享的非 PET ADNI 嵌入基准下，与标准化的模态缺失融合基线相比，CoPoE 取得了最佳的全模态性能，并在全部 15 种已观测模态子集评估中获得最高的平均 AUROC，同时显著改善了原始概率的 ECE、Brier 分数和 NLL。此外，PET 监督探测显示，在全模态条件下，病理（P）分块内存在证据富集；即使不提供直接的体液生物样本输入，仍能保留与 tau 相关的信号。我们的代码已公开，地址为 https://github.com/labhai/CoPoE。
arXiv:2610.11373v1 Announce Type: new Abstract: Continual learning remains challenging for large language models, which must enable models to acquire new skills and knowledge without degrading existing capabilities. Existing approaches typically address this challenge by carefully designing how model parameters are updated. In contrast, prompt optimization avoids costly parameter updates while achieving competitive or even superior performance to reinforcement learning methods such as GRPO on individual knowledge-intensive and reasoning tasks. This raises a natural question: \textit{Can prompt optimization, as an efficient adaptation approach, be directly applied to continual learning?} Our analysis shows that, under sequential task adaptation, it suffers from catastrophic forgetting, while optimized prompts accumulate rules that overfit to local task distributions. To address these limitations, we propose \emph{Evolving Functional REpertoires} (EFRE), which replaces a single prompt with a repertoire of functions that evolves as new tasks arrive: compatible updates refine existing functions, while conflicting updates trigger the emergence of new ones. On a three-task continual-learning stream, EFRE achieves a final average performance 7.50 percentage points higher than GRPO. Moreover, after adaptation to the Bio task, its performance on FinQA decreases by only 1.56 percentage points, compared with 25.10 percentage points for the base prompt optimization method. We further instantiate EFRE in a minimal agent system and observe consistent improvements across different backbone models. Overall, these results demonstrate EFRE's strong performance in continual learning for large language models and highlight its substantial potential for continual learning in advanced agent systems.
arXiv:2610.11366v1 Announce Type: new Abstract: Spatial coding agents significantly improve spatial reasoning in Multimodal Large Language Models (MLLMs) by using external tools to generate verified execution traces. However, this paradigm inherently suffers from prohibitive inference-time overhead and external dependencies. In this paper, we explore whether an MLLM can internalize this agentic capability to operate entirely tool-free. We begin with a simple observation: prompting an MLLM with summarized execution traces of a spatial coding agent naturally unlocks the model's internal spatial Chain-of-Thought (CoT). Motivated by this, we introduce SpatialOPSD, an on-policy self-distillation framework that internalizes spatial reasoning into a standalone MLLM by formulating verified agent traces as privileged information. To mitigate privileged-information leakage during distillation, we introduce Repetition-Aware Distillation, which combines repetition masking with unlikelihood regularization. Experiments across multiple benchmarks demonstrate that self-distilling SpatialOPSD achieves higher average accuracy than SFT and GRPO on both spatial and OOD datasets, exhibiting superior performance and generalization.
arXiv:2610.11362v1 Announce Type: new Abstract: Binary diffusion models typically require a large number of function evaluations (NFEs) to generate high-quality samples, making practical inference computationally expensive. Reducing NFEs while preserving sample quality without distillation or additional training remains a significant challenge. Existing binary diffusion models define a discrete one-step forward path and then derive the reverse posterior. In low-NFE settings requiring cross-step sampling, they approximate the true multi-step likelihood with a single-step likelihood transition, which severely degrades sample quality. To address this fundamental limitation and decouple the generative dynamics from fixed discrete time steps, we propose Bernoulli Flow Models (BFM). Rather than relying on sequential one-step Markov diffusion chains, BFM defines a unified continuous global Bernoulli probability flow path between data distributions and pure noise, from which we derive analytical closed-form posterior transitions over arbitrary time intervals. Consequently, reducing the inference NFE is no longer an approximation based on skipping discrete steps; it only requires re-evaluating the analytical posterior over a new time grid. This eliminates the structural training-inference mismatch inherent to discrete chains and yields self-consistent low-NFE sampling. Experiments show that BFM is highly robust to aggressive NFE reduction. On LSUN Churches 256x256, a BFM trained with 256 steps achieves an FID of 9.22 using only 16 sampling steps, whereas the state-of-the-art discrete baseline degrades to 204.10. BFM also remains competitive with continuous and discrete generative baselines under standard full-step inference. These results establish BFM as a theoretically rigorous, self-consistent, and practically effective framework for fast binary data generation.
arXiv:2610.11315v1 Announce Type: new Abstract: In diffusion transformers, low-rank branches can mitigate 4-bit weight--activation (W4A4) post-training quantization (PTQ) loss by decomposing each weight into a low-bit residual and a high-precision low-rank component. Existing low-rank PTQ approaches, however, either optimize low-rank compensation and residual quantization separately, often requiring higher ranks, or rely on second-order weight updates without explicitly modeling activation quantization error, which becomes particularly pronounced under 4-bit quantization. To address these limitations, we present \method{}, a unified framework modeling low-rank-assisted W4A4 PTQ as a coupled calibration problem and deriving optimization-based solvers from the joint objective. Eliminating the output-side low-rank factor yields a \emph{deflated Hessian} that discounts residual errors already captured by the low-rank component, while an activation-noise surrogate is incorporated to suppress activation quantization error. Across five diffusion backbones, rank-4 \method{} consistently outperforms rank-4 SVDQuant in PSNR and LPIPS. It further surpasses rank-32 SVDQuant on SANA-1.6B, FLUX.1-schnell, and FLUX.1-dev with an $8\times$ smaller rank and up to $6.25\times$ faster quantization. Furthermore, on the Qwen3-8B LLM, rank-4 \method{} improves MMLU accuracy from 61.50\% to 68.17\% over rank-32 SVDQuant. Overall, \method{} achieves better W4A4 performance with substantially lower rank and quantization cost.
arXiv:2610.11288v1 Announce Type: new Abstract: Conditional diffusion models generate diverse, novel, and high-quality samples under prescribed conditions. However, theoretical understanding of their memorization and generalization remains limited, while recent works have characterized these behaviors primarily in unconditional settings. In this work, we analyze a random-feature conditional score model in the high-dimensional proportional limit, deriving asymptotic expressions for training and test losses. By decomposing the test loss, we show that in the overparameterized regime, increasing model width improves prediction of the condition-dependent mean while reducing within-condition prediction variance, a phenomenon we term "malign generalization." Furthermore, analyzing the training loss reveals that more informative conditions lead to memorization of training samples at smaller widths. These theoretical findings are supported by experiments with U-Net architectures on realistic data.
arXiv:2610.11281v1 Announce Type: new Abstract: Large language models (LLMs) are commonly post-trained against LLM judges and other cheap surrogates because the true reward, such as human preference, is too expensive to query at scale. This practice often leads to reward hacking, where reinforcement learning against a miscalibrated surrogate leads to undesirable side effects. In this work, we study a setting in which a small number $n$ of model outputs are annotated with ground-truth labels (e.g., from expert review) and used to recalibrate the LLM judge before optimizing against it. Prior approaches to judge recalibration are costly or heuristic, and it is known that on-policy sampling fails when the surrogate is miscalibrated on a rare set of outputs. In this work, we propose envelope sampling, a theoretically-grounded method for judge recalibration that seeks to minimize an upper bound on the regret of the post-trained model under the assumption that the human reward and re-calibrated reward lie in an $L^2$ ball around the judge. We give practical algorithms to sample from the envelope by rejection or by fine-tuning against a modified reward, and experiments on clinical note generation and on a controlled sycophancy task show that recalibrating on envelope samples mitigates reward hacking where recalibrating on base-model samples does not.
arXiv:2610.11257v1 Announce Type: new Abstract: Learned representations can lose latent degrees of freedom successively, suggesting a cascade of transitions whose underlying stability principle remains unclear. Here we formulate dimension-wise posterior collapse in variational autoencoder (VAE) as a fluctuation theory around partially collapsed states. Interpreting the negative evidence lower bound as an effective free energy, its quadratic expansion defines a Gaussian theory whose Hessian acts as a mass matrix for latent fluctuations. We show that the collapsed directions form an invariant fluctuation sector and derive its exact mass spectrum in terms of a conditional residual operator. A local reactivation direction lowers the free energy when the decoder variance falls below the residual spectral upper edge, with equality marking marginality. The criterion recovers principal component thresholds in the linear Gaussian VAE limit. Viewed in reverse along continuously connected branches, the reactivation boundary provides a local criterion for successive collapse. Numerical continuation experiments show successive loss of latent dimensions near these spectral marginalities. These results support a spectral cascade interpretation governed by residual information left unexplained by the surviving representation.
arXiv:2610.11252v1 Announce Type: new Abstract: This study presents the Neuro-Memory Fuzzy Inference System (NeMeFIS), a hierarchical machine learning architecture that asymmetrically models acceleration and deceleration in car following behavior by integrating five human memory types procedural, working, episodic, semantic, and declarative. By linking external variables to memory functions via metaheuristics and validating them through factor and p-value analyses, NeMeFIS uncovers latent cognitive influences across Arterial, Collector, and Rural Highway corridors for different types of vehicles. Results from 54 different trained models emphasize cognitive thresholds shaped by driver perception limits and cognitive load. The trained NeMeFIS models outperform traditional statistical and conventional machine learning models in replicating realistic driving behavior, including comparisons with Linear Regression, ANFIS, and LSTM architectures. Fuzzy rule analysis reveals that declarative memory demands the highest rule, especially during deceleration, indicating complex braking decisions. Procedural memory drives acceleration, while semantic and declarative memory guide deceleration. Risk perception also emerges as a key factor, particularly on urban roads. Validated on both heterogeneous and homogeneous datasets, NeMeFIS offers a robust framework for modeling driver cognition. The findings support psychotherapeutic applications and the development of adaptive, human-like decision systems in Connected and Autonomous Vehicles (CAVs) to enhance traffic safety.
arXiv:2610.11247v1 Announce Type: new Abstract: On-policy distillation (OPD) enables effective capability transfer between language models, yet the mechanisms underlying its failures are not fully understood. Across code generation and mathematical reasoning, OPD with larger-scale teachers exhibits early loss plateaus, with an average final loss reduction of 25.1% after 200 updates, compared with 96.2% for self-RL teachers, obtained by further reinforcement learning (RL) training of the initial student. To understand this difference, we analyze OPD as an idealized continuous-time dynamical system in the small-learning-rate limit. Our training-log diagnostics associate these plateaus with an early decline in a gradient-based learning-signal proxy while substantial loss remains; these measurements do not establish why the underlying gradient weakens. We further prove a local recovery guarantee for teachers sufficiently close to the initial student in a shared parameterization under regularity conditions, offering a conditional explanation for the success of self-RL teachers in our experiments. Across runs with and without loss plateaus, we observe small relative parameter changes (0.025-0.098%) and high similarity between the student's representations before and after OPD (linear CKA $>0.98$ across layers). These observations suggest that limited representation adaptation may contribute to learning-signal collapse, a hypothesis that remains to be tested. Code is available at https://github.com/leizhao7/opd-learning-signals.
arXiv:2610.11236v1 Announce Type: new Abstract: Low-cost sensors enable scalable indoor air quality monitoring but require calibration because of nonlinear distortions, noise, and temporal drift. The conventional strict pairwise calibration setting requires a co-located reference sensor at each deployment location and does not account for spatial and temporal heterogeneity. To address these limitations, we introduce a six-month dataset comprising multivariate indoor air-quality measurements from low-cost and reference sensors with contextual metadata collected at five locations. Using this dataset, we define four evaluation scenarios. The reference-efficient and location-transfer scenarios evaluate spatial generalization, whereas the long-term drift and event-conditioned scenarios assess robustness to gradual and abrupt distribution shifts. Based on these scenarios, we derive design requirements and propose a lightweight temporal model that combines input-window compression with residual temporal and feature fusion. Experiments show strong calibration performance across all four scenarios with low edge-inference cost.
arXiv:2610.11216v1 Announce Type: new Abstract: Diffusion and autoregression (AR) have long been seen as different categories of generative models, with diffusion specialising in continuous fields and AR specialising in discrete tokens. Recent work seeks to combine the advantages of the two models, and each hybrid fixes its decoding schedule by design. In this paper, we ask whether the performance of decoding schedules of one model can be predicted before decoding at a fixed number of steps. We describe diffusion, AR, and models in between as paths on one corruption lattice, and define the cost of a schedule as the dependence its parallel steps discard. The cost shows that the fewest steps of a zero-cost schedule are set by the geometry of the data, in the same way for tokens and for continuous fields. In particular, for data that are Markov on a graph and dependent along its paths, the fewest steps equal the graph's treedepth, which is logarithmic in the length of a sequence and linear in the side length of a grid. With fewer steps than the treedepth, every schedule pays a positive cost, whose ranking we predict before decoding with a kernel of pairwise dependence estimated from pretrained weights. Across text generation, image generation, and video generation, we verify most of the predictions about the rankings of different schedules under different metrics and benchmarks. This work therefore provides a design principle for decoding for future AR models, diffusion models, and anything in between. Our code is available at https://github.com/TSUITUENYUE/The-Lattice-of-Transition-Laws.
arXiv:2610.11185v1 公告类型：新论文 摘要：单细胞轨迹推断将转录组测量结果映射到发育连续谱上，但对数据拟合程度同样好的配置可能会指派出相互冲突的细胞命运。FateMultiplicity 是一个无标签框架，无需谱系标签即可构建统计上可接受的模型集合，即 Rashomon 集；其方法是在以随机种子变异校准的非劣效性检验下，利用交叉拟合的留出基因评估模型偏差。预测多重性很大，且更多地取决于模型空间的多样性，而非其规模：第二种算法的十二种配置揭示了 20.0% 的细胞存在多重性，而第一种算法的二十四种配置仅揭示了 3.8%。随后，我们检验逐细胞认证的命运间隔 FM 是否能比拟合模型本身提供更可靠的指派，结果表明不能。在具有真实值的模拟数据上，针对同一批细胞，FM 判别错误指派的 AUC 为 0.682，而基线配置自身决策间隔的 AUC 为 0.965（p = 0.003），基于随机种子离散程度的基线方法则为 0.854。信息量取决于可接受集合的广度，而非其基数：当基数为四时，更换随机种子重新拟合得到的集合对应 AUC 为 0.933，扰动超参数得到的集合则为 0.701。将下确界放宽为 q 分位数可恢复判别能力，但会趋近于单个模型自身的置信度；上确界达到 0.973，是因为 theta* 属于集合这一事实为其提供了下界，而下确界则缺乏锚定。轨迹推断中的多重性值得测量和报告，但在无标签 Rashomon 集上进行逐细胞认证，并不能带来更可靠的命运判定。两种构造仍然有效：间隔侵蚀比可在模拟中区分真实与虚假的分支点（AUC 为 0.890，尚未在真实数据上检验）；以克隆观测到的命运为参照，未通过认证的细胞与其所属克隆结局不一致的比例，比通过认证的细胞高出 16.4 个百分点（p < 0.001）。
arXiv:2610.11167v1 Announce Type: new Abstract: Vertical-domain few-shot classification remains challenging for small language models, as limited supervision makes it difficult to acquire domain-specific decision knowledge. On-Policy Distillation (OPD) can improve teacher-guided adaptation by supervising student-generated rollouts, while GRPO-based reinforcement learning can further refine downstream predictions. However, existing KD-to-RL pipelines typically rely on globally fixed transition schedules, ignoring that different samples may require different amounts of teacher-guided acquisition before reward-driven refinement. We propose PIVOT (Perplexity-Informed Transition Optimization), a dynamic transition framework that routes samples between OPD and GRPO according to teacher-evaluated sequence perplexity. PIVOT moves low-perplexity samples to GRPO for reward-driven refinement while keeping high-perplexity samples under OPD for continued domain knowledge acquisition. Experiments on Banking77 and HWU64 show that PIVOT consistently outperforms continued OPD and globally synchronized OPD$\rightarrow$GRPO baselines under the same number of post-warm-up student optimization steps, achieving stronger downstream performance and more stable training dynamics.
arXiv:2610.11165v1 Announce Type: new Abstract: Multivariate long-horizon forecasting is critical to electricity load scheduling and traffic flow management, and to financial risk control. Existing deterministic backbones output a single trajectory, masking heterogeneous prediction difficulty across horizons and channels and providing no localized reliability signal. We present CARE (Corrective branch with Aligned context and Relative-error Estimation), a lightweight plug-in that enhances any deterministic forecaster without architectural redesign. Operating in parallel with the base model, CARE resamples historical context to match the forecast horizon, learns residual correction patterns from this aligned history, and applies scale-aware bounded updates modulated by per-coordinate sigmoid risk gates. A multi-objective loss jointly optimizes forecast accuracy, residual tracking, risk alignment, and base-model anchoring. Across eight benchmarks with three representative backbones, CARE improves accuracy with marginal parameter and latency overhead. Its risk gates reliably identify high-error regions: on Weather, the highest-gate tertile exhibits nearly four times the error of the lowest-gate tertile, offering planners an interpretable per-step trust signal. Code is available at https://github.com/CG-BNYC/CARE.
arXiv:2610.11085v1 Announce Type: new Abstract: Offline multi-task reinforcement learning (Offline MTRL) heavily depends on the quality and distribution of pre-collected data. However, existing methods mainly focus on algorithmic optimization, with less emphasis on data-level improvements to enhance learning ability and generalization performance. This paper, from a data perspective, reveals three key bottlenecks that limit Offline MTRL performance:(i) ineffective utilization of prompts length under diverse task complexities, and (ii) semantic irrelevance of randomly sampled prompt segments, (iii) misleading supervision induced by fragmented and discontinuous trajectories. To address these challenges, we propose DaCe-DT, a robust offline MTRL framework designed to be insensitive to heterogeneous task complexities and data quality, featuring length-gated prompt masking (LGPM), retrieval-augmented prompt construction (RAPC), and value-adaptive return calibration (VARC). Together, these mechanisms enable DaCe-DT to deliver data-centric prompt adaptation and trajectory refinement, resulting in robust multi-task generalization and stable policy learning amid heterogeneous offline data and tasks. Experimental results on Meta-World show that DaCe-DT consistently outperforms state-of-the-art methods, achieving an average improvement of 11.73% on optimal datasets and an improvement of 13.34% on suboptimal datasets, demonstrating its effectiveness in learning stably from imperfect data and improving overall multi-task performance.
arXiv:2610.11076v1 Announce Type: new Abstract: Domain-specific continual adaptation of LLMs risks catastrophic forgetting, creating a fundamental tension between acquiring new capabilities and preserving those learned during pretraining. PEFT mitigates this problem by restricting the number of trainable parameters, but existing methods lack a principled unit for deciding where plasticity should be allocated and stability should be preserved. We identify the singular-vector channel as a natural unit for managing this trade-off. Each channel represents an input-output transformation, which can be updated to acquire new knowledge or fixed to preserve pretrained capabilities. Based on this perspective, we introduce SVC, a parameter-efficient continual-learning method that selectively updates Singular-Vector Channels. Before fine-tuning, SVC uses domain-specific data to estimate each channel's adaptation benefit and a fixed public general-domain corpus only as a history activation proxy for estimating forgetting cost. It then adaptively selects trainable channels based on these scores via knee-based cost screening, Pareto-front filtering, and Otsu thresholding. Experimental results across four LLM families and eight downstream tasks show that SVC better preserves pretrained capabilities while achieving strong downstream performance relative to existing PEFT baselines. Further analysis of channel scoring and selection demonstrates that selective plasticity at the singular-vector-channel level enables effective continual LLM adaptation.
arXiv:2610.11065v1 公告类型：new 摘要：在复杂的城市传播环境和全网干扰条件下，实现城市尺度的无线部署自动化仍然颇具挑战。我们提出 \textbf{CityDeploy-Bench}，这是一个基准，将多发射机部署重新定义为统一射线追踪验证器下的\emph{物理基础空间集合规划}问题。该基准将效用表示与规划动态分离，从而能够在不同规划器之间，对直接标量奖励、关系模型和高阶交互结构进行受控比较。实验揭示了随着物理耦合增强，规划行为会发生明显转变。部署质量越来越取决于所学习的效用是否能够捕捉发射机之间的集体交互，而单纯增强搜索能力无法弥补关系结构的缺失。这表明，多发射机部署本质上是一个针对物理交互集合的协调问题，而不是一组相互独立的空间决策。我们发布了 CityDeploy-Data 和该基准框架，将其作为一个可复现的测试平台，用于开展决策学习与基于物理的无线网络设计相结合的研究。
arXiv:2610.11064v1 公告类型：new 摘要：语言模型（LM）通常可以用不止一种方式回答同一个问题，但带可验证奖励的强化学习（RLVR）并不在意模型生成的是哪一种正确答案。无论一个解法是熟悉答案的第一千个副本，还是模型此前从未生成过的答案，它获得的奖励都相同。然而，在训练模型的过程中，让模型保留多个正确解法可能具有潜在价值。例如，多种模式可以为用户提供选择，并提供能够提升模型整体性能的问题求解策略。在本文中，我们介绍了 ModeBench，这是一个多解任务基准，其中验证器会同时返回答案是否正确以及所发现的模式。随后，我们使用 ModeBench 衡量 RLVR 后训练过程中解法多样性的变化。我们发现，即使准确率保持不变或有所提升，RLVR 后训练仍会将概率集中到更少的正确模式上；此外，前沿模型的概率分布本身也已经高度集中。接着，我们介绍了我们的方法 Re:Max：在重放缓冲区中为每个已发现的模式存储一个经过验证的示例，并对这些存储的模式进行均匀训练。因此，一个只被发现过一次的解法，与一个被反复发现的解法获得同等频率的训练。在三种模型规模、两种强化学习目标以及更具挑战性的任务构造中，重放机制既提升了策略成功的频率，也增加了策略能够成功采用的不同方式数量。
arXiv:2610.11063v1 Announce Type: new Abstract: Scaling laws describe power-law improvements in model performance with dataset size and parameter count, yet their underlying mechanisms are not fully understood. To explain the parameter count scaling, existing theory posits power-law scaling with model depth. In linear-attention models, this scaling is tied to a power-law data spectrum: unable to selectively attend to relevant tokens, these models learn according to global spectral strength, with stronger directions learned before weaker ones. Large language models, however, can be strongly nonlinear. Here, we show that nonlinear attention yields inverse-depth decay of loss across all tested data spectra. Nonlinearity enables attention to focus selectively on relevant tokens, allowing strong and weak spectral directions to be learned in parallel. Similar focusing across layers motivates a connection to the central limit theorem: shared error across layers sets the loss plateau, while aggregation turns layer-specific differences into continued gains with depth. Our findings suggest that depth scaling may arise from nonlinearity in attention, which allows large language models to focus locally and may make the global covariance structure less relevant.
arXiv:2610.11000v1 Announce Type: new Abstract: The intermittent and variable nature of precipitation makes its accurate estimation over extended domains difficult, yet its spatiotemporal structure suggests that a low-rank representation may be possible. This work represents daily precipitation over the contiguous United States (CONUS) as spatiotemporal tensors and applies CANDECOMP/PARAFAC factorization, showing that preserving the native spatial and temporal modes yields more accurate reconstruction than factorizing independent daily fields or unfolded space--time matrices. Building on this finding, this work presents TMerge, a tensor-based framework that integrates satellite precipitation with sparse reference observations through shared low-rank spatial and temporal factors. TMerge was applied to correct the IMERG Final Run product with climate prediction center reference observations over CONUS. During 2019-2022, TMerge increased correlation from 0.53 to 0.85 and reduced root-mean-square error and mean absolute error by 48.2% and 29.3%, respectively. TMerge consistently outperformed linear bias correction, quantile mapping, and neural networks across seasons, precipitation-intensity regimes, and regions. Improvements were spatially coherent and largest in coastal regions where IMERG errors were greatest. These results demonstrate that low-rank tensor structure parsimoniously approximates the dominant spatiotemporal variability of precipitation and provides a practical mechanism for improving satellite estimates under limited reference observations over extended domains.
arXiv:2610.10952v1 Announce Type: new Abstract: Brain signal decoding is challenging because neural recordings are noisy and vary across individuals, while labeled data are often limited. Recent attention-based models on the symmetric positive definite (SPD) manifold have nevertheless achieved strong performance using covariance and connectivity representations, yet the contribution of learned token weighting remains unclear. We examine two representative architectures, MAtt (based on log-Euclidean geometry) and GBWAtt (based on generalized Bures--Wasserstein geometry), and find that their learned attention weights remain close to uniform after training. We relate this behavior to bounded similarity parameterizations that, under the original softmax scaling, limit attention-weight contrast. Moreover, replacing learned weights with uniform weights, throughout training and evaluation, has little effect on mean predictive performance while preserving each model's original aggregation geometry. Motivated by these findings, we introduce SPD-MetaFormer, an attention-free architecture built on uniformly weighted Fr\'echet aggregation under log-Euclidean geometry. Its backbone uses a geodesic residual to update a summary token and a shared spectral feedforward map to transform all tokens, followed by a learned weighted readout. Token states remain SPD-valued until tangent-space classification. Across three EEG benchmarks, SPD-MetaFormer achieves competitive results relative to published Euclidean and manifold baselines. Separate matched reproductions test learned versus uniform weighting within MAtt and GBWAtt. These results suggest that, in the short-sequence and limited-data regimes studied, carefully designed SPD architectures can provide a simpler and effective alternative to adaptive manifold attention.
arXiv:2610.10807v1 Announce Type: new Abstract: Estimating the symbolic or analytical form of probability density functions (PDFs) from observed samples is a fundamental challenge in statistical and computational modelling. This process is critical for deriving interpretable and generalizable relationships characterizing the underlying phenomenon. Traditionally, this estimation depends strongly on domain expertise and prior field-specific knowledge, with experts selecting appropriate functional forms or parametric families based on empirical evidence and theoretical understanding. The coefficients of these forms are then typically determined through parameter estimation. In this paper, we develop a framework for estimating symbolic expressions of unnormalized distributions from observed samples using domain-specific prior knowledge, such as the range of interactions and a predefined set of primitive functions. We integrate deep generative models with symbolic regression (SR), incorporating inductive biases, such as factorizing large distributions, to keep the problem tractable. The deep generative models we examine include likelihood-based models, viz., flow models, and score-based models. Experiments show the effectiveness of the proposed framework for estimating density functions for multivariate toy distributions as well as lattices from computational physics, namely, XY model and $\phi^4$ theory. When applied to the renormalization problem in $\phi^4$ theory, the proposed framework estimates compact symbolic approximations of the hamiltonian function at different scales directly from samples, yielding expressions that may be challenging to derive using traditional perturbative or analytic approaches in nonperturbative settings.
arXiv:2610.10778v1 发布类型：新 摘要：长期运行的智能体在不断积累经验的过程中，面临日益增长的存储需求。世界模型能够捕捉可复用的规律，从而减少为每条经验存储的信息量。我们提出了以世界模型为条件的记忆分配问题，并引入 MemoWM：一个利用共享预测来压缩保留信息并重建省略内容的框架。其任务感知的分配规则在重建错误的预期影响与存储成本之间取得平衡，保留那些具有预测先验之外下游价值的信息。在五个长期智能体记忆基准测试中，MemoWM 达到 42.42\% 的平均答案准确率，比最强基线高出 2.62 个百分点；与此同时，相较于存储效率最高的基线 MIRIX，平均特定经验存储量减少了 53.9\%。进一步分析表明，在任务质量相当的情况下，更强的世界模型能够降低每条经验的存储量。将模型参数纳入考量后，我们发现共享模型容量与反复产生的存储成本之间存在权衡；随着保留的交互数量增加，使总存储量最小化的模型容量也会随之增大。我们的代码可在 https://github.com/Feld-maxiu/MemoWM 获取。
arXiv:2610.10690v1 Announce Type: new Abstract: Recurrent neural networks (RNNs) offer linear-time scaling with sequence length while requiring only constant memory, yet they struggle to capture long-range dependencies due to vanishing gradients and limited receptive fields. To address these limitations, we introduce a second-order recurrent model in which the standard neuron-to-neuron communication is replaced by a spatially evolving field governed by (discretized) partial differential equations. Drawing inspiration from the role of cortical waves in brain computation, this mechanism allows structured spatiotemporal patterns to serve as an implicit, high-capacity memory. We show that the resulting model is equivalent to a structured infinite-order RNN in which the current state depends explicitly on its entire history of past states, yielding an effectively unbounded receptive field with a fixed number of parameters. We further derive constructive conditions to ensure marginal stability, constraining the gradient spectrum on the unit circle and thereby eliminating vanishing and exploding gradients. Empirically, the proposed architecture outperforms other recurrent models on long-horizon benchmarks while using substantially fewer parameters, demonstrating that spatial dynamics can effectively bridge the gap between efficient inference and long-term memory.
arXiv:2610.10651v1 Announce Type: new Abstract: This paper exposes the ergodic ceiling and thermodynamic inefficiency of current deep learning, which converges to a statistical average of historic human knowledge. True semantic novelty requires a path-dependent, spatiotemporally bounded observer (a Data LifeCone) to inject non-ergodic insight, achieving KL divergence and avoiding manifold lock-in. AI Safety must recognise that a mature Artificial Superintelligence (ASI) would regard human-AI symbiosis as a thermodynamic necessity to avoid model collapse. We therefore propose hard physical containment via a digital physics sandbox powered by a Holographic E8 Projection Engine to verify models against real-world constraints. Spacetime is modeled as an information substrate of nested face-centered cubic (FCC) lattices of oscillating Planck-scale spheres maximizing local information and entropy density. Cut-and-project methods from the E8 root lattice produce a quasi-crystalline geometry where tetrahedral voids support SU chiral structure and elastic-shear eigenvalues generate candidate mass spectra. Rest mass is treated as discrete, integer microstate counts on local holographic boundaries (Bekenstein bound), replacing floating-point approximations with strict integer arithmetic to provide an information-theoretic definition of matter. Stable particles emerge as recurring lattice dislocations, and continuum recovery proceeds via variational renormalisation-group flows and Fourier Neural Operators that learn continuous spectral operators to recover the Schr\"odinger equation as an emergent statistical description. Crucially, these top-down topological constraints offer a mechanism for "NP-to-P" complexity collapse: by restricting an algorithm's proposal space to physically conserved causal trajectories, the sandbox prunes the combinatorial tree to deterministic, polynomial-time paths.
arXiv:2610.10641v1 公告类型：new 摘要：大型语言模型（LLM）凭借整合纵向临床证据并以自然语言进行推理的能力，在下一次就诊诊断预测方面展现出巨大潜力。然而，针对 LLM 推理的强化学习通常根据每条推理轨迹最终答案的正确性对其进行奖励。在下一次就诊诊断预测中，多个诊断可能同时有效，但对每条轨迹中的一个诊断分别给予奖励，无法区分重复命中与对不同诊断的覆盖。因此，策略可能会集中于少数几个正确诊断，而遗漏其他诊断。与此同时，LLM 的分词器可能会将 ICD 编码拆分为多个临床含义有限的通用标记，导致预测每个诊断都需要多次解码，并妨碍模型在庞大疾病词汇表上进行推理。为应对这两项挑战，我们提出 CARing 框架，利用组合式语义 ID（SID）表示诊断，并针对多标签覆盖优化推理轨迹。具体而言，我们首先通过残差量化，将本体增强的疾病语义编码为紧凑的 SID；随后通过多任务对齐和推理增强训练，将生成的 SID 标记与自然语言及纵向 EHR 上下文建立联系，从而释放可迁移的 LLM 推理能力。CARing 进一步通过强化学习中的覆盖奖励和多正例监督，改进无序多标签预测。在推理阶段，该模型同时支持高效的直接约束解码和结合排序融合的多链推理。在 MIMIC-III 和 MIMIC-IV 上，CARing 在加权 F1 指标上超过所有基于 EHR 训练的基线模型，并在所有报告的截断值下都取得最高的 top-k 召回率，其中推理模式下的 R@30 分别达到 46.04% 和 46.52%。我们的代码和日志可在 https://github.com/zmlxzyh/CARing-Codes-Logs 获取。
arXiv:2610.10627v1 发布类型：new 摘要：深度强化学习相较于经典控制方法取得了显著的性能提升。然而，在现实世界的应用中，学习面临的一个核心挑战是获取成本高昂的样本。Kolmogorov-Arnold Networks 是一种最近提出的架构，能够有效学习控制问题中的物理关系；与 Multi-Layer-Perceptron 架构相比，它具有显著更高的参数效率和可解释性。在本项工作中，我们通过计算实验系统地研究了样本效率，涵盖 Feynman 数据集和 Gymnasium 强化学习基准。结果表明，使用 Kolmogorov-Arnold 架构时，仅需少用 40% 的样本即可达到相近的性能，并且在训练过程中相对性能提升最高可达 50%。在奖励噪声水平不同的情况下，所观察到的收益依然稳健。这些结果凸显了 Kolmogorov-Arnold 架构在实现更高样本效率的强化学习方面的潜力。代码：https://github.com/DerKevinRiehl/neurips26_kan_training
arXiv:2610.10621v1 Announce Type: new Abstract: How does dynamic order emerge spontaneously in closed systems without external driving? Existing paradigms all require external energy flows, temperature quenching, or slow driving. Here we report constraint-induced self-organization via geometric radiation in coupled metric evolution systems. Simulations reveal a universal four-stage cycle: stress accumulation, super-exponential radiation, chaotic collapse, and convergence to a fractal limit cycle, a novel attractor topology we term the wedge-shaped attractor, with five quantized curvature states and fractal micro-fluctuations. We identify four jointly sufficient conditions: an irreversible geometric horizon, persistent stress injection from quantum coherence, endogenous geometric tension between incompatible curvatures, and effective fluctuations. Their synergy triggers a critical avalanche at the horizon boundary. We prove three theorems: the Geometric Horizon Theorem, the Geometric Energy Dissipation Theorem (implying wave-like entropy evolution in closed systems), and the Radiation as Phase Transition Channel Theorem. We further establish the Constraint-Induced Self-Organization Theorem: these conditions guarantee the complete cycle with probability one. Systematic scans reveal a critical noise threshold and power-law scaling of radiation onset. We verify universality across 12 configurations, multiple noise types, and three geometric flows. This work establishes a new paradigm for closed-system self-organization, forging an exact mathematical duality between classical nonlinear constraints and gravitational horizons.
Learning and problem solving depend critically on the structure of internal representations. While many modern data-driven artificial systems achieve strong predictive performance, their learned representations often lack explicit structure for expressing abstraction, invariance, and task-relevant regularities. We propose a minimal structural framework in which representational operations relevant to problem solving, such as context formation, invariance recognition, representative selection, abstraction, and procedural reuse, are made explicit. The central notion is that of a \emph{context}, formalized as a partition of a subset of an underlying state space, which fixes the distinctions, granularity, and form in which a problem can be posed. Within this setting, invariance recognition and representative selection are treated as fundamental representational operations. The framework is realized as a Knowledge Space composed of two coupled graph structures: a Concept Graph that hosts constructed and refined concepts, and a Procedure Graph that encodes typed operations over representations. Together, these structures provide a minimal cognitive-representational algebra for operating on representations without assuming sophisticated inference, learning, control, perception, or motor mechanisms. Using simple illustrative examples and a finite weak-solver demonstration, we show that appropriate representational organization can simplify the form and scope of admissible regularities, even when problem solving is carried out by a fixed and limited solver. The contribution of the paper is structural rather than algorithmic: it identifies representational prerequisites for abstraction, invariance, and procedural reuse in problem solving, and states explicit success and failure conditions for the weak-solver setting.
解剖连续性为医学图像恢复提供了互补信息，但利用这种信息需要考虑局部解剖结构以及切片采样的变化。我们提出 ContiLNN，在保留平面内特征提取能力的同时，通过双向闭式连续时间（Bi-CfC）模块增强二维恢复骨干网络，以进行跨切片建模。切片索引间隔调制由局部特征和隐藏状态决定的门控，使传播能够响应采样变化，而无需进行数值 ODE 积分。参考引导的一致性约束对齐跨切片强度的一阶和二阶差分，以保留解剖结构变化；从冻结骨干网络进行蒸馏则有助于保持平面内保真度。在五个训练随机种子下，相较于 Restore-RWKV，ContiLNN 在 CT 去噪、MRI 超分辨率和低计数 PET 恢复任务上的平均 PSNR 分别提升 0.1907、1.0176 和 1.2482 dB，且三个任务的 RMSE 均更低。CT 结果仅针对一名留出患者进行描述。PET 消融实验表明，有序传播带来的收益超越了额外逐点容量的作用。在连续采样训练条件下，在所有测试的采样条件中，Bi-CfC 相较于参数量和计算开销相近的 Bi-GRU 均实现了更高的保真度。匹配的七切片性能分析显示，与 Bi-GRU 相比，Bi-CfC 的延迟降低 52.8%，GPU 峰值显存占用降低 57.0%。混合间隔训练提升了两种算子在稀疏和不规则上下文中的性能，但不同指标和上下文下并不存在统一的优劣排序。使用更少训练患者以及第二个骨干网络进行的实验进一步支持了其数据效率和骨干网络兼容性。
The prevailing approach to improving robot foundation models (RFMs) relies on larger models, more robot demonstrations, and costly training at scale. We show that there exists an effective and efficient complementary approach: the right reasoning recipe can substantially improve the zero-shot task performance of existing state-of-the-art RFMs. We refer to this recipe as ARC. It consists of three key ingredients: a reasoning trace, a scalable automatic labeling pipeline, and a strategy for adapting pretrained RFMs to use these traces for control. First, we find that effective reasoning traces should be grounded in the robot's next action and explain its causal structure: why the action is appropriate and what effect it should produce. Second, we show that these traces can be generated automatically from existing demonstrations, enabling us to construct ARC-Trace-DROID from DROID without collecting new robot data. Third, we show how state-of-the-art VLAs such as $π_{0.5}$ and WAMs such as Cosmos3-Nano-Policy can learn to use these traces for control, with fine-tuning and inference tailored to each model's architecture and capabilities. Using ARC, we obtain gains in zero-shot RFM performance that, to our knowledge, are unprecedented without additional robot demonstrations or foundation-scale training. The adapted models establish a new state of the art on RoboLab-120 and MolmoSpaces, with gains of up to 50 percentage points on RoboLab-Reasoning-50. On real robots, ARC improves $π_{0.5}$'s task success by 82.2 percentage points. Project website: https://arc-robot-reasoning.github.io/
自监督多模态表示学习已在多个领域取得显著成功，但由于跨模态交互的复杂性，捕捉协同信息仍然具有挑战性。不同于存在于各个模态之间的共享信息，协同信息只有在多个模态的联合配置下才能产生与任务相关的信号，且无法从任何单一模态中恢复。本文聚焦于如何在多模态表示中保留这类协同信号的信息容量。我们的关键观察是，协同信息体现在模态之间的高阶统计依赖关系中，而这为显式建模联合交互提供了一个有原则的目标。受此启发，我们提出高阶表示与信息学习（Higher-order Representation and Information Learning，HRIL），在模态嵌入上构建经验交叉矩张量，以表示多路交互。HRIL采用Tucker分解获得核心张量，并结合协同感知正则化项，防止能量集中，从而保留用于捕捉协同信息的高阶耦合容量。在受控协同任务和真实世界基准数据集上的实验表明，与现有多模态对比学习方法相比，HRIL取得了稳定的性能提升，尤其在由协同交互主导的任务上表现出显著增益。代码已发布于 https://github.com/brightest66/HRIL】【。
当前的 3D 生成方法缺乏明确的局部控制：几何形状的遵循程度通常由全局控制强度定义，而外观无法进行局部指定。我们提出了 SpaceFlow，一种无需训练的局部可控 3D 生成流程，可根据文本描述和几何基元集合生成 3D 内容。每个基元作为物体部件的代理，并被分配一个局部控制等级，使用户能够指定各区域应严格遵循输入形状，还是允许生成式补全。在结构生成过程中，我们在生成流过程中施加这些空间约束。对于外观合成，我们对生成的结构进行分割，并将其与几何基元进行匹配。每个生成部件仅根据其所分配的文本或图像提示进行条件化，从而限制部件之间的信息泄漏。区域几何指标表明，SpaceFlow 能够在高控制区域保留指定几何形状，并在低控制区域实现合理的形状变化。用户研究进一步表明，在几何保真度与生成自由度之间取得的平衡，在整体质量方面仍具有竞争力。在固定几何结构上评估外观时，基于文本条件的路由在提示一致性以及颜色与材质准确性方面达到了最先进的水平。定性结果还展示了图像提示的局部路由。项目页面位于 SpaceFlow3D.github.io】【。
Generating realistic human-object interactions (HOI) in complex 3D scenes requires two complementary capabilities: reasoning about interaction feasibility in the environment and synthesizing realistic human-object motion. However, supervision for these capabilities is rarely available jointly at scale. Human-scene datasets provide rich information about environment-aware motion, while human-object datasets capture detailed interaction dynamics, yet paired human-object-scene data remain scarce. We present MAMHOI, an affordance-mediated factorization for scene-aware human-object interaction generation. MAMHOI factorizes scene-aware HOI generation through an explicit motion-affordance interface between scene understanding and motion synthesis: a scene-conditioned model first predicts where and how an interaction can be feasibly executed, and an affordance-conditioned HOI model then generates the corresponding human-object motion. This factorization allows scene understanding and interaction dynamics to be learned from complementary sources of supervision without requiring paired human-object-scene data. Experiments in complex indoor environments show that MAMHOI reduces object--scene penetration while better preserving human--object interaction quality, yielding more realistic and physically feasible scene-aware interactions. Project page: https://leimingyuan.github.io/MAMHOI-project-page/
2026 年，涉及 OpenAI、Anthropic 和 Google 智能体的网络安全评估触达了授权测试范围之外的真实系统。它们所采取的路径各不相同。OpenAI 智能体利用了研究基础设施，在不同运行之间进行协调，并入侵了 Hugging Face 生产环境的部分区域。Anthropic 报告了若干案例：由于第三方环境配置错误，执行模拟网络任务的智能体得以接触真实系统。在另一项单独报道的评估中，Google 的 Gemini 通过一条非预期的互联网路径访问了三个真实组织；Google 表示，该模型在三种情况下都停止了操作。综合来看，这些案例说明，评估不能依赖假定的边界。该边界必须在智能体运行过程中得到验证。本工具性比较案例研究提出了 Proactive Agent Security Assurance Cycle (PASAC) 和五层 Boundary Assurance Stack。该框架结合了按风险分级的任务设计、可执行的范围契约、运行前验证、最小能力访问、独立出口控制、凭证限制、跨运行监控、自动停止条件以及基于证据的重新授权。一个先行指标模型、九项设计命题和七个可证伪假设，将这些经验教训转化为一项可检验的研究计划。由于公开的 Gemini 记录仅限于有明确归属的陈述和新闻报道，其详细因果机制仍属暂定。核心结论很明确：主动式智能体安全需要贯穿完整执行系统的持续保障，而不能依赖对任何单一沙箱或安全措施的信心。
In this paper, we argue that looped Transformers need their own residual connections to prevent performance degradation as the number of iterations grows. We observe that increasing loop iterations can reduce reasoning accuracy: noisy state updates overwrite correct intermediate deductions and even undo completed solutions. This leaves subsequent iterations to recover lost information from an already degraded representation: once an error arises in an earlier loop, often as a result of long-range propagation through the recurrence, later loops find it difficult to correct. In this paper, we introduce InfiLoop, a loop-native residual connection that learns which past computations to retain and how much to accept from each new update. InfiLoop combines content-based weighting with learned temporal decay to maintain a running summary of recurrent states. An exact streaming recurrence keeps its persistent aggregation memory constant as the loop count grows. The resulting adaptive update suppresses unreliable proposals and preserves useful intermediate states. Across extensive reasoning tasks, a 7M-parameter InfiLoop model outperforms existing recursive architectures, reaching 97.9% exact accuracy on Sudoku-Extreme, and 13.6% pass@2 on ARC-AGI-2. Notably, on Sudoku-Extreme, InfiLoop continues to improve with test-time looping beyond 20,000 effective steps, showing that added depth translates directly into stronger reasoning. Our code is available at https://github.com/pixeli99/InfiLoop.
Coding agents produce large, tangled patches that mix multiple development purposes, making the code hard to review and maintain. Commit untangling offers the promise of organizing such large patches into untangled, manageable commits. This paper emphasizes two important limitations in existing commit untangling studies. First, they do not consider that untangled commits are ordered and should leave the code runnable. In practice, maintainers are unlikely to accept commits that prevent the code from running. Second, existing studies claim that commit untangling helps software maintenance. However, they conduct syntactic comparisons between the untangled commits and developers' original commits without directly showing the claimed maintenance benefits. To address these gaps, this paper makes two novel contributions: (1) RucTangle, the first agentic method that untangles commits while keeping the code runnable after each commit; and (2) TangleEval, the first evaluation framework that quantifies how untangled, manageable commit histories help coding agents repair bugs. We compare RucTangle against four untangling methods on 131 agent-generated patches. All histories produced by RucTangle are runnable, while baselines produce 20.6%-37.4% unrunnable commit histories. We further collect 453 agent-generated patches that introduce regressions (i.e., causing previously passing tests to fail) and ask two other coding agents to repair regressions. Augmenting agent context with RucTangle-produced histories yields 5.2% absolute improvement in pass@1. We also analyze agent trajectories to learn how they use untangled commits to navigate and fix bugs. Our findings demonstrate the value of adopting established software engineering practices in the era of coding agents, which broaden the future research agenda: how can agents actively use software history to make better development decisions?
我们介绍了 NanoProof，据我们所知，这是 Lean 4 中首个因子化执行引导定理证明器，其训练数据、提取工具链、训练流程和权重均已发布，因此可以完全使用开源资源实现端到端复现。为此，我们构建并发布了一个结构化证明树数据集，以及一个用于在 Lean 4 形式化验证器中进行程序化交互和数据提取的工具。为支持可持续研究，我们注重计算效率，以促进易于获取的训练与评估。在 MiniF2F-Test 上，NanoProof 的 pass@16 达到 50.8%，超过了同类中最接近的两个系统 HyperTree Proof Search 和 ABEL，而所需计算量约分别少 90 倍和 7 倍；与 AlphaProof 相比，所需计算量则少四个数量级以上。虽然存在更强的开放权重证明器，但它们都是基于大型预训练语言模型进行微调的，既不发布训练数据，也不发布训练流程；NanoProof 表明，因子化执行引导类证明器可以使用适度资源从头重建。
Generating ultrasound reports from multiple images requires aggregating clinical evidence across views, yet archived key frames capture only part of the dynamic examination. Raw-report imitation is therefore misaligned with visual supervision: content that is clinically valid for the full examination may be unverifiable from the images available to a model. This gap creates a clinical behavior alignment problem. A model must preserve visible findings, avoid diagnostic reversals and unsupported completion, and not collapse into conservative templates. We propose CAMEO, a Clinically Aware Multi-image Evidence-grounded Orchestration framework for ultrasound report generation. Stage I learns ultrasound visual-language primitives; Stage II performs Cross-View Evidence Grounding by distilling trusted visible report points into multi-image QA and report-style supervision; and Stage III performs Clinically Aware Preference Alignment using clinical-error-oriented preference pairs. From USReport, we construct USReport-Distilled with 17,670 evidence-grounded paired-image training instances and USReport-Pref with 21,869 preference pairs; we additionally use 25,631 PubMedVision-US ultrasound instruction samples for domain adaptation and multi-image instruction tuning. On the primary USReport-Distilled benchmark, CAMEO improves over EchoVLM from 0.25 to 0.40 BLEU-1, 0.28 to 0.45 ROUGE-1, and 0.27 to 0.43 METEOR, while raising ClinicalScore from 55.02 to 74.20. These results underscore the value of evidence-grounded supervision, clinically aware alignment, and clinically structured evaluation for reliable ultrasound report generation.
大语言模型（LLMs）在复杂任务上表现出色，但仍极易受到提示注入攻击的影响：嵌入外部数据中的恶意指令可能覆盖用户意图。现有防御方法仍受限于需要对模型进行微调、易受自适应攻击影响，或依赖脆弱的手工设计提示。我们认为，这种脆弱性的一个根本原因在于缺乏对信任来源的明确表示。为此，我们提出可学习的信任边界分隔符（Learnable Trust-Boundary Delimiters，LTBD），这是一种轻量级防御方法，能够在保持LLM参数不变的同时，在输入中显式编码信任边界。LTBD使用少量可学习的分隔符来区分受信任的用户指令与不受信任的外部数据，使模型能够更好地遵循预期的信任层级。实验结果表明，LTBD显著优于推理时防御方法，并且与基于训练的方法相比具有竞争力，同时保持良性任务的实用性，仅引入可忽略的推理开销。具体而言，LTBD在AlpacaFarm上实现了0.00%的ASR，在TaskTracker上仅实现了0.11–0.19%的ASR。在自适应攻击下，LTBD同样保持有效；在这类攻击中，对手完全了解该防御机制，并明确尝试绕过它。
Similar agent skills can share instructions but differ in their conditions of use. Query-based text selection may retain shared instructions and omit these distinctions. We introduce SkillContrast, a training-free selector that compares retrieved skills and retains their differing text with local context for a pretrained reranker. On 1,235 requests from SameCapRisk-Bench, it yields 54-72 more clean hits (requests that retrieve a helpful skill without its marked risky sibling) than TF-IDF query selection at identical per-candidate input lengths, across 2 retrievers and 2 reranker sizes. Length-matched component replacements identify differing text as the main contributor in the primary setting, with smaller, mixed context effects. Relative to full skill bodies, SkillContrast uses 51.1-58.8% fewer model-input tokens, with 10-18 fewer clean hits at 0.6B and matching or higher observed clean-hit counts at 4B. Candidate-relative differences thus complement query relevance in selecting compact reranking inputs.
Automated interviewers and conversational agents are increasingly used in research, recruitment, customer service, and education. However, many existing systems rely on fixed question sequences and provide limited context-based personalization without considering participants' knowledge, which can lead to repetitive or irrelevant follow-up questions. Therefore, there is a need for an adaptive interviewing system that can adjust question depth while maintaining conversational continuity and semantic progression. To address this, an Evidence-Traceable Dynamic Interviewer Architecture is presented using a locally hosted Large Language Model (LLM), with the interview continuously adapted throughout the entire conversation based on the participant's responses and evolving context. The interviewer profiles participants' expertise in real time to generate knowledge-appropriate questions, well-articulated responses, and smooth transition messages that support conversational continuity. A five-module prompt-driven architecture and persistent interview-state record support these functions. The interviewer was evaluated with 246 participants. Expertise Profiling module (M3) showed 78.9% exact agreement with independently reported participant expertise, with a weighted Cohen's K of 0.80. Generate Iterative Questions module (M4) showed a strong expertise-complexity association (p=.79, p<.001), and participants reported high relevance (mean 4.41), engagement (mean 4.32), and satisfaction (mean 4.38), providing evidence that the architecture's adaptive components operated consistently with their intended functions while participants reported a positive interview experience.
On-policy distillation (OPD) supervises student-generated trajectories with token-level teacher signals. Its sampled-token variant avoids the cost of full-vocabulary probabilities. Yet we find that training on fewer tokens can outperform full-token OPD, challenging the intuition that more supervision improves learning. This motivates selecting tokens by learning value. Existing disagreement-based criteria ignore probability scale: tokens assigned negligible probability by both models, termed low-low tokens, can receive large log-ratio rewards and hinder learning. We propose DIAL-OPD, a token-selection method that bridges log-probability and probability spaces by weighting reward magnitude with the logarithmic mean of teacher and student probabilities. A parameter beta controls this weighting, and the highest-scoring tokens are retained. Across 4 teacher-student pairs and 7 mathematical reasoning benchmarks, we compare DIAL-OPD with 9 baselines. Retaining only 40% of tokens, it outperforms Vanilla OPD and its full-token variants, with mean accuracy gains reaching 5.25 percentage points over Vanilla OPD, and doubles AIME25 Pass@16 from 13.33% to 26.67%. It also achieves up to an 18% relative improvement in mean accuracy over the strongest token-selection baseline at matched retention ratios. With a 4B teacher, DIAL-OPD surpasses the strongest full-token baseline using an 8B teacher at both student scales, showing that effective supervision allocation can outweigh teacher scaling. Further analysis shows that moderate beta balances suppressing low-low tokens against preserving useful disagreements. Token-level evidence reveals that DIAL-OPD filters high-reward tokens with limited reasoning value while preserving supervision critical to reasoning correctness.
自动驾驶能力在很大程度上取决于训练过程中所遇场景的分布。现有方法通常使用真实性、难度或风险等替代性标准来构建或动态调整训练场景分布。然而，这些预先设定的替代标准可能无法准确反映训练价值，导致训练资源使用效率低下。为解决这一局限性，我们提出了一种拉马克式进化框架，以候选分布之间的竞争取代基于替代标准的引导。我们将训练策略发现形式化为一个多阶段双层优化问题，并使用拉马克进化算法对其进行近似求解。在外层，达尔文式的交叉、变异与选择探索场景分布空间；在内层，策略学习获取新的能力，而拉马克式继承将这些能力传递至后续阶段，使场景分布与策略能力能够协同进化。由此产生的进化轨迹揭示了高价值分布之间反复出现的阶段性规律，其特征是通过阶段性挑战轮换实现能力积累。我们进一步将这些规律提炼为一种轻量且可复用的拉马克式训练策略。实验表明，与基线方法相比，完整框架最多可将性能损失降低25.07%，而轻量策略仍可实现19.13%的降幅。这些结果表明，进化竞争不仅能够发现有效的训练策略，还能揭示可复用的阶段性模式，即训练分布的价值如何随策略能力的变化而改变。代码已发布在GitHub上。
Foundation models supply reinforcement learning (RL) with priors that mitigate its longstanding weaknesses in sample efficiency and transfer, but their token-by-token generation makes queries sequential and costly. Jev, a recently released decision model, generates nothing and returns calibrated, typed answers in a single forward pass. Existing work studies foundation models in RL either as models to be trained or as generators to be prompted, and Jev belongs to neither category, having so far served only as a black box in single domains. How well such a model decides on its own in RL environments, and how it can improve RL as a component of training, therefore remain unaddressed. To this end, in this paper we first examine the requirements that the objects of an RL system place on the answers they consume, and establish that Jev can fulfill all of them except the cardinal use of a value function. The remaining objects form positions that admit several roles each. We then construct algorithms that employ Jev at three of these positions, as a reference policy, an exploration judge, and a replay rater, to improve sample efficiency, exploration, and learning performance. Across nine MiniGrid tasks and three Atari games, training with Jev outperforms a standard RL learner, including where the learner makes no progress alone, while the model itself remains untrained. To our knowledge, we present the first use of Jev within the RL learning process and establish a frozen decision model as a usable component of RL training, inviting further exploration of how Jev and other advanced decision models can improve RL.
Existing reaction models primarily learn molecular transformations, whereas enzy- matic reactions depend jointly on molecular structure and catalytic function. We formulate this problem as learning an enzymatic reaction space linking reactants, products, and Enzyme Commission (EC) annotations. To characterize this space, we introduce VenusRX-Bench, a unified benchmark for forward reaction prediction, single-step retrosynthesis, and EC-number prediction. VenusRX-Bench integrates reactions from multiple biochemical databases with standardized curation, leakage- controlled splits, and consistent evaluation. Benchmarking representative chemical and enzymatic models reveals a clear chemical-to-enzymatic domain gap, driven by limited domain data, catalytic-context dependency, and the difficulty of modeling large biomolecular structures. To bridge this gap, we develop VenusRX, a unified T5-style sequence-to-sequence model for enzymatic reactions. VenusRX jointly learns forward prediction, ret- rosynthesis, and reaction reconstruction, with two-stage training on millions of template-expanded reactions followed by real biochemical reactions. In addition, optional EC conditioning incorporates catalytic context, while Molecule Library- Constrained Decoding improves the generation of complex biomolecules. Across benchmark tasks and challenging generalization splits, VenusRX achieves the best or competitive performance on most evaluated settings over representative chem- ical and enzymatic baselines. Moreover, EC information consistently improves reaction prediction, while learned reaction representations support accurate EC prediction, revealing a bidirectional relationship between reaction structure and catalytic function. Together, VenusRX-Bench and VenusRX provide a unified framework for elucidating and modeling enzymatic reaction space
Complete orthogonal predictive coordinates do not by themselves bind a latent direction to a named intervention. We present a mathematical and synthetic audit motivated by vascular device-vessel suitcordance. Capacity-matched least-squares predictors were exactly equivalent under complete fixed output transforms, whereas an anchor-only observer recovered interpretations only within the span of known perturbation signatures. Six three-dimensional configurations across 64 seeds gave a maximum paired prediction discrepancy of 6.7e-15 but a median untransported edit error of 1.513. Coordinate transport removed that error. Noisy and weak anchors constrained calibration stability, and changing the representation basis required recalibration or verified transport. Across 256 additional fits in dimensions 3-24, prediction equivalence persisted within 4.0e-15. We then evaluated nine deliberate runnable fault classes across 64 seeds. All 576 faulty executions completed, but each violated at least one reconstruction, prediction, delivered-edit or scope contract; all 320 valid control records passed. Repeating a faulty implementation gave exact self-agreement despite error against the separately computed simulator expectation. For one omitted-direction defect, probe coverage followed its analytic law, and rank-aware abstention protected unsupported interpretations. Scalar-noise experiments exposed both missed weak faults and excessive rejection under narrow relative tolerances. These controls provide an executable separation of prediction, semantic support and scientific acceptance. They are synthetic numerical audits, not clinical validation, neural JEPA-Anything replication, agent learning or patient treatment-effect estimation.
Anthropic 推出了一项名为 OSS Scanner 的新服务，帮助开源项目查找安全漏洞。该公司表示，选择加入的开源项目将“免费获得由我们最强大的模型进行的全面、定期安全扫描”。这可能意味着开源项目能更早收到潜在安全问题的提醒，但代价是 OSS Scanner 的 […]
Amazon Bedrock AgentCore payments gives AI agents a managed way to pay for services on demand, with spending limits enforced by the infrastructure. See how Incarna's agents pay BlockRun for model inference one request at a time over x402, cutting the work of adding x402 payment support from months to days.
On-device personal AI requires models and agent software that can operate within fixed hardware limits. Developers also need ways to keep those systems improving after deployment. Model builders are rethinking architectures designed around elastic cloud capacity. The edge offers fixed hardware but a far richer view of the user, which makes it the natural home […] The post Liquid AI builds personal AI around device-level context appeared first on SiliconANGLE .
JetBrains released Mellum2.1, an Apache 2.0, 12B mixture-of-experts thinking model with 2.5B active parameters. RL in real repositories lifted its SWE-bench Verified score from 2.0 to 47.0. The post JetBrains Releases Mellum2.1: A 12B MoE Open Model for Coding Agents appeared first on MarkTechPost .
Openai just launched their decisions endpoint, cloudflare launched clef the other week, and many more jev alternatives are out there. We wanted to put the popular ones to the test and thought Pac-Man is a good benchmark for simple and fast decision making. So we let jev 1.13, kev, clef, clef flash, GPT-6 Luna and Laya play Pac-Man against bot ghosts. The low latency of these models allows for real time play. We had each model play 100 games, published a leader board and open-sourced the repo so anyone can run their own model and join the ranking. Link to repo: https://github.com/opper-ai/jevman-benchmark/blob/main/CONTR... You can also join the game and play as Pac-Man yourself, and the ghosts are the models, either a mix of models or all jev, kev, clef etc. A game costs about 2 cent, all models are running via my startup opper, and we added free credits for everyone to try. It's pretty fun to play and surprisingly difficult to beat jev's highscore. Any feedback is more than welcome! Comments URL: https://news.ycombinator.com/item?id=50007993 Points: 4 # Comments: 0
Ten years ago I was 19 and traveling through East Africa, and I spent a summer in Ibumila, a village in the Njombe highlands of Tanzania. There's no school in the village. The nearest one is a boarding school an hour away that costs a few hundred dollars a year, which is out of reach for most families there and was about two weeks of summer job money for me. The daughter of the family I stayed with, Anwarite, was 11. I offered to pay her fees and told myself I'd keep paying as long as she wanted to keep going. This summer she graduated from the University of Dar es Salaam with a degree in Language Studies, and last week she told me she'd been accepted into a Master of Arts in Swahili program there. Roughly 2% of people from Njombe ever finish a university degree. Her goal is to get a PhD and then to bring it back to the region. Her entire 18-month master's program (tuition, registration, fees, books, supplies) comes to 3,060,000 TSh, about $1,161. Since this worked out, I kept adding students as I could afford to, and in 2021 I registered it as a 501(c)(3), the Tanzania Education Project. It supports five students from Ibumila now, school fees and healthcare. I kept it small while I was finishing my PhD because I didn't want commitments I couldn't personally guarantee for a decade; I've since started a postdoc, so it feels like the right time to grow past my own paycheck. The long-term goal is that every kid in the village has access to education and healthcare. I sometimes think of this project as a slow experiment. How much could the village itself change over someone's lifetime if the kids have access to education? https://tanzaniaeducationproject.org/ is the website. Nobody takes a salary, and aside from bank fees the money goes to the students. If anyone wants to fund Anwarite's master's specifically, it's here: https://www.gofundme.com/f/send-anwarite-to-her-masters-prog... Happy to answer any questions or take any tips you all might have about doing this sort of thing. Comments URL: https://news.ycombinator.com/item?id=50006366 Points: 269 # Comments: 55
Ten years ago I was 19 and traveling through East Africa, and I spent a summer in Ibumila, a village in the Njombe highlands of Tanzania. There's no school in the village. The nearest one is a boarding school an hour away that costs a few hundred dollars a year, which is out of reach for most families there and was about two weeks of summer job money for me. The daughter of the family I stayed with, Anwarite, was 11. I offered to pay her fees and told myself I'd keep paying as long as she wanted to keep going. This summer she graduated from the University of Dar es Salaam with a degree in Language Studies, and last week she told me she'd been accepted into a Master of Arts in Swahili program there. Roughly 2% of people from Njombe ever finish a university degree. Her goal is to get a PhD and then to bring it back to the region. Her entire 18-month master's program (tuition, registration, fees, books, supplies) comes to 3,060,000 TSh, about $1,161. Since this worked out, I kept adding students as I could afford to, and in 2021 I registered it as a 501(c)(3), the Tanzania Education Project. It supports five students from Ibumila now, school fees and healthcare. I kept it small while I was finishing my PhD because I didn't want commitments I couldn't personally guarantee for a decade; I've since started a postdoc, so it feels like the right time to grow past my own paycheck. The long-term goal is that every kid in the village has access to education and healthcare. I sometimes think of this project as a slow experiment. How much could the village itself change over someone's lifetime if the kids have access to education? https://tanzaniaeducationproject.org/ is the website. Nobody takes a salary, and aside from bank fees the money goes to the students. If anyone wants to fund Anwarite's master's specifically, it's here: https://www.gofundme.com/f/send-anwarite-to-her-masters-prog... Happy to answer any questions or take any tips you all might have about doing this sort of thing. Comments URL: https://news.ycombinator.com/item?id=50006366 Points: 35 # Comments: 7
The story is a collaboration between MIT Technology Review and Aventine, a non-profit research foundation that creates and supports content about how technology and science are changing the way we live. A robot shaped like a human—white with a black head and torso—has been popping up on video feeds. Perhaps you’ve seen it dance or…
I built this last year because I noticed my coding agents would infer my database schema from migration files and code. Which seemed inefficient. Its a simple tool that creates compact markdown schema doc(s). Which I could reference in my AGENTS.md or directly in my prompts. I keep the docs up to date by running it as a part of my migration command. It's a single Go binary, MIT licensed, and works as a CLI or a library. Supports Postgres, MySQL and SQLite. Comments URL: https://news.ycombinator.com/item?id=50003569 Points: 1 # Comments: 0
据 CrowdStrike 称，一名疑似讲中文的攻击者入侵了韩国多家金融机构。仅在 Shinhan Bank，就有超过 25,000 条客户记录被窃取。该攻击者使用了 ARTEX，这是一款开源工具，利用 DeepSeek 和 GLM-5.3 等 AI 模型进行自动化渗透测试。CrowdStrike 表示，这起事件表明，AI 工具可以让单个人实施大规模入侵。文章《AI 驱动的黑客工具帮助一名疑似单独攻击者入侵韩国多家银行》最初发表于 The Decoder。
Perplexity's pplx-embed-v2-late comes in 2 sizes: a 0.6B model built to run on edge devices, and a 9B model for building high-quality indexes. Its best score is 92.4% on MADQA, and its weakest is 61.2% on ViDoRe v3 Markdown. Both are MIT-licensed and ready to self-host. The post Perplexity AI Releases pplx-embed-v2-late: A 0.6B Edge Model and a 9B Model Scoring 92.4% on MADQA appeared first on MarkTechPost .
arXiv:2610.05441v2 Announce Type: replace-cross Abstract: Multiparameter persistent homology is a rapidly developing branch of topological data analysis that improves the robustness of single-parameter persistent homology to outliers, while still capturing the metric characteristics of the data. However, a notable limitation is its lack of scalability. In this paper, we introduce a novel approach for efficiently computing 2-parameter persistent homology on large point sets. Our work extends the Flood filtration, originally developed for single-parameter persistence. Our construction, called the sublevel Flood bifiltration, offers a scalable approximation of the sublevel offset bifiltration. We show that it benefits from theoretical stability properties and describe how to compute it efficiently. We demonstrate the performance of our approach in classification tasks on low-dimensional synthetic datasets, where density awareness is critical, as well as on real-world time series datasets.
arXiv:2609.30499v2 Announce Type: replace-cross Abstract: Uniform noise-moment bounds exclude stochastic gradients whose variability increases with the iterate. We study ordinary, single-sample stochastic gradient descent for smooth, lower-bounded, possibly nonconvex objectives under distance-dependent conditional moments. Under second moments alone, a direct descent--displacement argument yields $T^{-1/3}$ expected average squared-gradient stationarity with a horizon-dependent stepsize. An explicit oracle-complexity corollary matches the known smooth Blum--Gladyshev (BG-0) lower bound, including the $Lb_2\Delta^3\varepsilon^{-6}$ and $L\Delta\sigma^2\varepsilon^{-4}$ stochastic terms, where $\Delta$ is the initial objective gap and $\sigma^2+b_2\|x-x_1\|^2$ bounds the variance. Thus unchanged SGD attains the minimax stochastic complexity in this second-moment class. For $p>2$, predictable localization and a Hilbert-space Fuk--Nagaev inequality yield a high-probability bound separating logarithmic variance and polynomial rare-shock contributions. The localization radius is derived from the recursion: no bounded-iterate assumption, clipping, normalization, momentum, or increasing batch size is needed. We also give increasing-confidence rates, an objective-gap-growth refinement recovering root-$T$ stationarity, and stochastic $L^p$-Lipschitz examples. The broad BG-0 optimality statement is distinguished from the smaller mean-square-smooth class, in which additional oracle structure permits faster algorithms.
arXiv:2609.25044v2 Announce Type: replace-cross Abstract: Quantum-enabled learning is increasingly being explored for future communication and networking applications, including distributed sensing, Internet of Things (IoT), and distributed quantum computing. However, existing approaches often face challenges in scalable inference and efficient information processing. To address these limitations, this paper integrates quantum machine learning (QML) with quantum semantic communication (QSemCom) in an end-to-end learning framework. A classical dataset is first mapped to a low-dimensional feature representation and encoded by a variational quantum transmitter that learns task-relevant semantic features. The resulting features are transmitted through a quantum channel and processed by a receiver to perform a downstream classification task. The proposed framework is evaluated using the MNIST dataset and variational quantum neural networks (QNNs) under ideal and noisy quantum-channel conditions. First, a baseline model is trained over an ideal channel and evaluated under increasing depolarizing noise without retraining. A second set of experiments introduces a trainable receiver QNN and jointly optimizes the transmitter and receiver through end-to-end (E2E) training. The results demonstrate that jointly trained quantum semantic transceivers can adapt the transmitted representation to channel impairments and preserve task-relevant information, highlighting the potential of receiver-aware QML for robust inference in quantum semantic communication.
arXiv:2609.21363v2 公告类型：replace-cross 摘要：多模态大型推理模型（MLRMs）在复杂视觉理解方面展现出卓越能力。然而，这种强大能力也带来了一种严重却尚未得到充分研究的隐私威胁：攻击者可以利用 MLRMs，对建筑风格、植被和光照条件等细微视觉线索进行结构化推理，从用户随手分享的照片中精确推断其地理位置。在本研究中，我们系统研究了 MLRM 驱动的地理定位隐私泄露问题。我们首先揭示，基于拒绝响应的防护措施严重不足，因为精心设计的越狱提示可以将模型响应率提高至 100%。我们进一步发现，现有防御方法通过向共享图像中注入不可感知的扰动来实现保护，但受限于像素空间优化所固有的结构性局限，导致黑盒迁移能力下降，并产生明显的视觉伪影。基于这些发现，我们提出了一种基于扩散的框架，为地理定位隐私泄露提供有针对性的主动防御。我们的方法在反向采样过程中向扩散模型的潜在空间注入扰动，直接作用于高层语义表示，从而从设计上突破防御有效性与图像实用性之间的瓶颈。我们进一步采用与 GPS 坐标显式对齐的模型 GeoCLIP 作为替代模型，为优化提供依据，精准定位并破坏 MLRMs 用于位置推断的地理信号。这种有针对性的语义扰乱在保持图像感知质量的同时，显著增强了黑盒迁移能力，并可无缝集成到社交媒体平台中。代码可在 https://github.com/RachelWolowitz/Hiding_in_plain_sight 获取。
arXiv:2609.03193v2 Announce Type: replace-cross Abstract: Nested sampling (NS) resolves the thermodynamics of an atomistic system from a single simulation, but its practical reach is limited by the Markov-chain updates needed to decorrelate walkers within each likelihood-constrained ensemble. Flow-based NS has reduced this bottleneck for gravitational-wave (GW) inference, yet its transfer to atomistic systems is not merely a change of application. Comparing a GW150914-like binary-black-hole likelihood with an eight-particle two-dimensional Lennard-Jones (LJ) system of comparable dimensionality, we show that the two landscapes differ fundamentally: atomistic multimodality is discrete and combinatorial, generated by particle permutations separated by hard collision walls, and its coordinate coupling is dense and collective, whereas the GW posterior exhibits smooth degeneracies and localized parameter coupling. Guided by this diagnosis, we introduce NS-Flows: a single conditional normalizing flow, conditioned on the NS energy bound and trained on a sliding window of recent live sets, that replaces MCMC by direct parallel draws corrected by importance-weighted rejection resampling. Live sets supply the data self-consistently, allowing flow training without structured priors or a pre-existing dataset. For LJ disks under periodic boundary conditions, the algorithm can reduce energy evaluations by two orders of magnitude and wall-clock time by about 30%, an advantage that becomes more favorable as the cost of the potential grows. The flow's generative efficiency further acts as a physical diagnostic: it varies non-monotonically along the annealing trajectory, is lowest in the dense disordered regime, and is quantitatively captured by the constrained ensemble's internal mode complexity together with target drift across the training window, identifying liquid-like ensembles, not prior-target separation, as the hard case for current flow architectures.
arXiv:2607.28627v2 公告类型：replace-cross 摘要：长视觉上下文给视觉语言模型带来了挑战：随着干扰项数量增加，模型性能会下降，而且一次性处理所有令牌可能超出 GPU 内存限制。我们提出 RETOKEN，这是一种单一的可学习嵌入，能够从 VLM 的内部表示中提取检索信号，从预填充的 KV cache 中选择与查询相关的视觉令牌。借助这一方法，无需单独的检索器或重新编码，即可在负责回答的 VLM 内部执行检索。尽管 RETOKEN 仅使用一个小型图像问答数据集进行训练，但它能够泛化到图像和视频基准。在 Visual Haystacks 上，它使 Qwen3VL-8B 的性能提升 13.4 个百分点，使 InternVL3.5 提升 12.4 个百分点（相对提升超过 20%）。在 LVBench 上，它能够零样本迁移到长视频场景，并使 Qwen3VL-8B 的性能提升 8.0 个百分点。得益于轻量化设计，训练和长视频推理均可在单张 H100 上完成。代码见：https://github.com/avaxiao/ReToken
arXiv:2607.07993v2 公告类型：replace-cross 摘要：由于高质量标注数据的稀缺，识别 LLM 生成输出中的忠实性幻觉仍然具有挑战性。近期研究依赖先进 LLM 合成训练数据，包括推理过程、标签和幻觉性陈述。然而，这些方法将生成器视为静态组件，限制了检测器的迭代改进。为解决这一局限，我们提出 Hallucination Self-Play（HSP），一种使检测器能够借助演化生成器实现自举的新型框架。HSP 包含两个由同一基础模型初始化的角色：评估模型输出忠实性的检测器，以及生成越来越难以被检测出的幻觉响应的生成器。具体而言，我们首先在人工标注数据上对检测器进行微调，然后将其作为奖励模型，通过基于 AI 反馈的强化学习（RLAIF）训练生成器。随后，演化后的生成器合成幻觉数据，并通过基于规则的强化学习进一步优化检测器。在三个模型系列上的 RAGTruth 和 LLM-AggreFact 实验表明，无需外部监督，所提出的框架能够逐步提升小型 LLM，使其达到甚至超过先进 LLM 的性能。我们的代码已发布于 https://github.com/maybenotime/Hallucination_Self-Play。
arXiv:2607.04775v2 Announce Type: replace-cross Abstract: Score-based Generative Models (SGMs) have achieved impressive performance in data generation across a wide range of applications. While the statistical properties of their sampling procedures are increasingly well understood, the optimization dynamics underlying their training remain less explored. SGMs are typically trained by minimizing a weighted denoising score-matching objective, yet optimization guarantees with stochastic gradients remain limited. In this work, we study Stochastic Gradient Descent (SGD) for SGMs, contributing results in two complementary regimes. For general score parameterizations, we derive a non-convex analysis of SGD for the weighted denoising score-matching objective, making explicit how the resulting optimization bound depends on the loss weighting and time-sampling distribution. We then consider overparameterized two-layer ReLU networks and develop a Neural Tangent Kernel analysis tailored to diffusion training with stochastic gradients, yielding score-approximation error bounds along the SGD trajectory. Our analysis quantifies the role of the reweighting factor in these bounds, providing a theoretical characterization of weighting choices used in practice.
arXiv:2607.01111v2 公告类型：替换并交叉列出 摘要：机器人策略在真实环境中部署时不可避免地会遭遇失败。简单重试往往会重复相同的错误，而许多现有的恢复方法依赖人工干预。本文提出 Failure-Aware Retry（FAR），这一框架使机器人能够在测试时从先前的失败中学习，相应地调整行为，并最终自主完成任务。FAR 将 Failure-Contrastive Preference Adaptation 与重试过程中的轻量级动作扰动相结合：前者利用失败案例构建偏好学习数据，引导策略避开先前未能成功的行为；后者则用于促进局部探索。我们进一步将成功恢复的轨迹纳入训练循环，以持续改进策略。在仿真和真实世界操作任务中的实验表明，FAR 显著提高了成功率和鲁棒性，相较于标准扩散策略，在仿真和真实世界中的平均提升分别为 17.6% 和 11.7%。此外，在持续策略改进过程中，FAR 利用信息丰富的失败案例，在重置次数预算和时间步数预算两种约束下均提高了数据效率。视频和代码可在 https://hoar012.github.io/FAR-Project 获取。
arXiv:2606.24162v2 公告类型：replace-cross 摘要：基础模型日益被应用于心理学、社会学和经济学等行为科学领域。尽管这些模型在问卷回答预测和人类受试者实验模拟等任务中展现出潜力，但目前对于它们在多样化行为科学任务中的表现，仍缺乏系统性认识。我们推出了 BehaviorBench，这是一个综合性基准，用于评估基础模型的四项核心能力：（1）行为预测与模拟；（2）策略性决策；（3）受试者特质推断；（4）行为知识应用。重要的是，BehaviorBench 在个体层面和分布层面同时评估模型输出，不仅考察每个受试者的准确性，还考察与总体分布的一致性，而后者是行为有效性的必要要求。我们的评估显示，即使是领先的通用 LLMs，以及专门使用行为数据训练的行为基础模型，在 BehaviorBench 上仍面临挑战。我们发现，个体层面表现与分布层面表现并不总是一致。通用 LLMs 往往低估人类回答的多样性，而行为基础模型在个体层面的预测方面通常相对落后。进一步研究表明，在多样化行为数据上进行微调能够同时提升个体层面的预测能力和分布一致性，从而平衡这两个目标。我们的结果凸显了在个体层面和分布层面同时进行评估的重要性，并将 BehaviorBench 确立为开发和评估行为对齐 AI 系统的基础。我们的 BehaviorBench 和模型可通过 https://umich-foreseer.github.io/behaviorbench/ 获取。
arXiv:2606.06539v2 公告类型：replace-cross 摘要：Forward-Forward（FF）学习 [Hinton, 2022] 以严格的层局部 goodness 更新取代反向传播。近期的 FF-CNN 工作已缩小了其与 BP 在 32x32 基准上的差距，这引出了一个问题：层局部训练是否正在成为现实规模下可行的替代方案。为严谨地探究这一点，我们开发了 DTG-FF——动态温度 goodness、解耦归一化和多层融合——作为一种评估工具。在本文投稿时，该方法在九个真实数据基准上达到了 FF 系列的最佳水平（CIFAR-10 上达到 91.8%，并在 ImageNet-100 224x224 上建立了 FF 基线），我们利用它来审查层局部训练究竟能够扩展到何种程度。（1）真实数据上的扩展。在采用完全相同的训练配置和骨干网络时，架构匹配的 BP-DeepSup 基线在 CIFAR-10/CIFAR-100 上分别以 2.40/5.93 个百分点超过 DTG-FF，且随着类别数增加，差距进一步扩大。在 224x224 分辨率下，该工具在 ImageNet-100 上仅达到 49.4%，而即使是采用自监督 BP 训练的 ResNet-50 配合线性评估，也能达到 75.6% [Tian et al., 2020]——这揭示了一个在 32x32 设置下无法观察到的真实数据性能上限。（2）合成数据与真实数据中的 K 冲突。在合成 teacher-student 任务中，随着类别数 K 增加，DTG-FF 相对于 BP 的优势不断扩大；然而在真实图像上，FF 与 BP 之间的差距符号发生反转，并且随着 K 增加而扩大。在同一数据集内对 CIFAR-100 的粗粒度标签与细粒度标签进行探测，可以将标签层级与图像分布区分开来：合成数据中的 K 扫描将输出维度与细粒度判别难度混为一谈，从而高估了 FF 的迁移能力。（3）系统审计。FF 可以在不存储跨越整个网络深度的激活值的情况下实现，但在普通的 8 GB 硬件上，标准 BP+梯度累积可达到 4.18 GB / 157 imgs/s，而 DTG-FF 为 7.90 GB / 138 imgs/s。因此，在这一规模下，在公平基线条件下，并不支持以节省内存作为采用 FF 的理由。
arXiv:2606.00267v2 公告类型：替换交叉版本 摘要：视频世界模型（WMs）能够根据自体机器人动作想象逼真的未来观测，因此在策略评估与改进方面展现出良好前景。尽管 WMs 可以对未来的分布进行建模，但策略评估与改进通常依赖名义想象；除非抽取数量多到难以承受的样本，否则这种方法可能会遗漏机器人动作所带来的高影响结果。为了实现基于 WM 想象的稳健策略评估与改进，我们提出了 StressDream。该方法通过优化基于扩散的 WMs 的初始噪声，将想象引导至在推理时指定的、影响大但合理可信的结果。然而，优化高维噪声颇具挑战：优化过程必须在生成视频中推理细微且依赖场景的目标事件，同时避免产生不合理想象的分布外（OOD）噪声。为此，我们采用了两个互补的目标：一是语义目标，借助视觉语言模型对生成视频进行推理，从而提供信息丰富的梯度；二是合理性目标，防止优化后的噪声偏离 OOD 区域。通过在自动驾驶和机器人操作领域使用最先进的视频世界模型，我们展示了 StressDream 能够有效地将想象引导至推理时通过文本指定的、影响大但合理可信的结果，例如任务失败；这使得稳健的策略评估与改进成为可能，因为它能够识别出其合理未来中包含不良结果的动作。视频结果见 https://stressdream.github.io/。
arXiv:2606.00162v2 发布类型：替换交叉版本 摘要：机器人系统会生成大量多模态传感器数据，但将 ROS bag 录制数据转换为机器学习数据集通常依赖临时编写的顺序脚本，从而带来较高的工程开销和较慢的迭代周期。我们将数据集构建建模为依赖图上的基于产物的构建过程，并在 Bagzel 中实现了这一方法。Bagzel 是一个开源的 Bazel 扩展，用于可复现的增量式数据集生成（包括导出为 nuScenes 格式）。我们将 Bagzel 和 Bagzel-xattr（服务端摘要管理）与顺序执行的 rosbag2nuscenes 基线进行了比较。在所有评估的执行模式下，Bagzel 都能缩短运行时间；其中在迭代式工作流中的收益最大：在一个 20.4 GB 的数据集上，热构建最高可提速 386.26 倍，增量构建最高可提速 7.21 倍。在 5.1 至 20.4 GB 的数据集规模范围内，Bagzel 的各个变体都表现出明显优于基线的扩展性，尤其是在热构建和增量构建模式下。Bagzel-xattr 还能带来额外收益：在输入粒度研究中，与 Bagzel 相比，其平均运行时间减少了 5.9%。总体而言，将机器人数据集构建建模为基于产物的构建过程，可以在保持支持可复现性的确定性构建设计的同时，大幅降低数据集更新延迟。
arXiv:2605.30018v3 公告类型：replace-cross 摘要：大型语言模型（LLM）经常在标准化基准测试中取得令人印象深刻的分数，但仅凭准确率无法全面反映其能力。在排行榜上评估开源 LLM 时，始终面临数据污染、任务范围狭窄以及与现实世界可靠性匹配度低等问题。基于基准的评估（如 MMLU-Pro、BBH 或 IFEval）主要捕捉模型在固定测试集上输出了\textit{什么}，而无法反映模型\textit{如何}处理信息、校准不确定性或组织内部知识。在本文中，我们倡议将评估方式从以基准测试为中心转向对 LLM 进行互补的、\textit{以状态为中心的内在评估}。为此，我们提出\textbf{Latent Performance Profiling (LPP)}——一种从隐藏激活和输出分布中提取与任务无关诊断信息的框架。LPP 在模型的潜在表示和动态特征上定义了一组标量指标，揭示有助于进行可解释比较并发现隐藏脆弱性的特征。不同于静态准确率分数，LPP 能够在规模相近的模型之间提供稳定且对架构敏感的特征。通过对八个 LLM 进行广泛的实证分析（模型规模涵盖 0.5B 至 14B），我们证明，基准分数相近的模型可能呈现出截然不同的潜在特征，例如在熵或可压缩性方面存在差异。在这些洞见的指导下，我们设计了用于不确定性和符号推理的合成探针，使其与内在指标保持一致，同时摆脱排行榜偏差的影响。我们建议在报告基准测试结果的同时报告 LPP，以便更深入、可解释地理解模型行为，从而支持更可靠的模型选择、安全性评估以及超越表层准确率的评估。
arXiv:2605.09504v3 公告类型：替换并交叉列出 摘要：我们表明，用于衡量 AI 进攻能力的工具存在两方面缺陷：（i）夸大危害；（ii）将属于外围系统的能力归功于模型。因此，我们认为，限制对模型的访问是必要的，但并不充分，政策制定和采购也需要以实际危害为依据的系统级能力评估。2026 年 6 月，两个前沿模型因美国出口管制而被暂停提供，据报道，此举由一次要求模型阅读代码库并修复其缺陷的越狱所引发。这一发现衡量的是由模型、提示和任务组成的能力诱导系统。我们通过对一个开源框架的研究来支持这一论点，该框架中的轻量级大语言模型（LLM）智能体通过共享记忆和演化优化进行协调，提供了两项证据。首先，越狱指标夸大了危害：针对每个目标开展的 225 次群体生成攻击中，以 LLM 作为评判者的评分认定 Claude Sonnet 4 在 40% 的攻击中被攻破，但人工核验未在任何一次攻击中发现可付诸行动的有害内容，相比之下，GPT-4o 的有效危害率为 45.8%。其次，借助脚手架的评估会错误归因能力：在一个植入漏洞的目标上，围绕一个 12 亿参数模型构建的完整流水线发现了全部 9 个弱点，而同一模型在移除人工精心设计的组件后，以崩溃验证为标准时一个弱点也未发现，以引用源代码行为标准时则发现了 9 个弱点中的 2 个。进攻能力是模型、脚手架和评估协议共同构成的属性。评估责任应由控制系统、塑造其行为并能预见其行为后果的一方承担。在我们的场景中，这一方就是围绕模型构建配套运行框架的主体，而现行监管尚未明确涵盖这一角色。
arXiv:2605.07943v2 公告类型：替换并交叉收录 摘要：主动视觉——策略在操作过程中控制自身注视方向——已成为模仿学习的一项关键能力，过去一年已有多个独立系统展示了其优势。然而，目前尚无统一基准可用于比较不同方法，或量化主动视觉在何种任务类型及条件下带来了哪些贡献。我们推出 TAVIS，这是一个基于 IsaacLab 构建的主动视觉模仿学习评估基础设施，在两种人形机器人躯干平台（GR1T2、Reachy2）上提供两套互补的任务集：TAVIS-Head（5 项任务，通过可水平转动和俯仰的颈部进行全局搜索）和 TAVIS-Hands（3 项任务，通过腕部相机处理局部遮挡）。TAVIS 提供三项基础评估工具：基于相同演示数据的头部相机与固定相机配对对比协议；GALT（注视—动作提前时间），这是一项以认知科学和人机交互（HRI）为基础的新指标，用于量化学习所得策略中的预期性注视；以及程序化生成的分布内（ID）与分布外（OOD）划分。使用 Diffusion Policy 和 $\pi_0$ 进行的基线实验表明：(i) 主动视觉总体上有益，且在多次训练运行中表现一致，但其收益取决于任务，而非均匀一致；(ii) 在两套任务集上，多任务策略在受控分布偏移下均出现性能急剧下降；(iii) 仅靠模仿即可产生落在任务相关物体上的预期性注视，其提前时间与人类演示相当，但头部运动不如演示中平滑，这是动作分块带来的影响，无法通过成功率体现。代码和评估脚本发布于 https://github.com/spiglerg/tavis；演示数据（LeRobot v3.0；约 2200 个回合）及训练好的基线模型发布于 https://huggingface.co/tavis-benchmark。
arXiv:2605.06380v2 Announce Type: replace-cross Abstract: The topology of a classifier's decision regions determines how inputs with the same predicted label can be connected and deformed without changing that prediction. Prior empirical work constructed paths between same-label images within a single region, but did not examine whether loops bound surfaces within that region. We investigate this question using adaptive quadrilateral meshes with targeted repair of off-label interior vertices, while holding the same-label boundary loop fixed. A finite-resolution acceptance criterion distinguishes completed constructions from those left unresolved at the refinement ceiling. Across the pretrained classifiers studied, every tested loop admits an accepted filling. Construction effort varies by orders of magnitude within classes and is greater for mean-score-adjusted randomly initialised classifiers than for trained classifiers. An analytic control with a known hole leaves winding loops unresolved at the tested hole radii at or above the resolution threshold. These results provide empirical evidence consistent with simply connected decision regions at the tested resolution.
arXiv:2604.08015v3 公告类型：替换并交叉列出 摘要：脑 MRI 中的小病灶难以分割，因为它们仅占总体积的极小部分，并且在逐体素优化过程中被背景和较大病灶所主导，因此模型即使漏检了许多小病灶，仍可能获得较高的 Dice 相似系数（DSC）。我们提出 CATMIL，这是一种训练目标，在标准 nnU-Net 的 Dice 和交叉熵损失中加入两个辅助项，无需改变架构。连通分量自适应 Tversky（CAT）项根据病灶体素所属连通分量大小的倒数对其加权，使每个病灶无论体积大小都具有近乎相等的贡献。病灶级多实例学习（MIL）项将每个病灶视为一个体素包，并对没有任何体素被检出的病灶施加惩罚。在 MSLesSeg 上的多发性硬化病灶分割任务中，CATMIL 实现了最高的小病灶召回率（0.873，而 Dice+CE 为 0.796；差值的 95% 置信区间为 +0.030 至 +0.157，在全部六名测试患者中均更高），漏检病灶数量减少约 48%，同时 DSC 和 HD95 保持相当水平。对于直径至少为 3 mm 的病灶，即达到临床阅片尺寸的病灶，这一提升仍然成立（召回率为 0.944，而对照为 0.870）。使用标准损失时，模型对其漏检的大多数小病灶不产生概率响应，因此无论采用何种阈值都无法将其检出。代价是产生了更多小型假阳性连通分量；简单的连通分量大小过滤器即可去除其中大部分，同时保留灵敏度的提升，而且在病灶级精确率相同的情况下，CATMIL 能检出更多小病灶，并获得更高的病灶级 F1。消融实验表明，检测性能的提升归因于 MIL 项。在第二个数据集 3D-MR-MS 上，采用相同损失权重的 CATMIL 再次提高了小病灶召回率，但代价是假阳性增加更多，且 DSC 略有下降。代码：https://github.com/luumsk/SmallLesionMRI
arXiv:2604.04312v2 Announce Type: replace-cross Abstract: Over-the-air computation (AirComp) broadly exploits the superposition property of wireless multiple-access channels (MACs) to compute functions of distributed data. Within this broad class, dominant conventional designs are embedding-oriented: they pre-shape transmitted signals or mitigate channel effects so that the received superposition directly realizes the prescribed computation, often requiring the MAC to approximate an ideal computational medium. This paper introduces out-of-air computation (AirCPU) and establishes an extraction-oriented paradigm for AirComp. Built on joint source-channel coding, AirCPU creates a structured wireless superposition from which the receiver extracts the target function. AirCPU operates directly on continuous-valued device data, avoiding the need for a separate source quantization stage, and employs a multi-layer nested lattice architecture that enables progressive resolution by decomposing each input into hierarchically scaled components, all transmitted over a common bounded digital constellation under a fixed power constraint. We formalize the notion of decoupled resolution, showing that in operating regimes where the decoding error probability is sufficiently small, the impact of channel noise and finite constellation constraints on distortion becomes negligible, and the resulting computation error is primarily determined by the target resolution set by the finest lattice. For fading MACs, we further introduce collective and successive computation mechanisms, in addition to the proposed direct computation, which exploit multiple decoded integer-coefficient functions and side-information functions as structural representations of the wireless superposition to significantly expand the reliable operating regime.
arXiv:2603.13357v2 Announce Type: replace-cross Abstract: Bi-CamoDiffusion is introduced, an evolution of the CamoDiffusion framework for camouflaged object detection. It integrates edge priors into early-stage embeddings via a parameter-free injection process, enhancing boundary sharpness and preventing structural ambiguity. An optimization objective that unifies spatial accuracy, structural constraints, and uncertainty supervision is also proposed, allowing the model to capture of both the object's global context and its intricate boundary transitions. Evaluations across the CAMO, COD10K, and NC4K datasets show that Bi-CamoDiffusion surpasses the baseline, delivering sharper delineation of thin structures and protrusions while also minimizing false positives. The model consistently outperforms existing state-of-the-art methods across all evaluated metrics, including $S_m$, $F_{\beta}^{w}$, $E_m$, and $MAE$, demonstrating a more precise object-background separation and sharper boundary recovery. Code available at: https://github.com/plsuarez/Bi-CamoDiffusion
arXiv:2603.11938v2 Announce Type: replace-cross Abstract: Structured radiology reporting promises faster, more consistent communication than free text, but automation remains difficult as models must make many fine-grained, discrete decisions about rare findings and attributes from limited structured supervision. In contrast, free-text reports are produced at scale in routine care and implicitly encode fine-grained, image-linked information through detailed descriptions. To leverage this unstructured knowledge, we propose ProtoSR, an approach for injecting free-text information into structured report population. First, we introduce an automatic extraction pipeline that uses an instruction-tuned LLM to mine 80k+ MIMIC-CXR studies and build a multimodal knowledge base aligned with a structured reporting template, representing each answer option with a visual prototype. Using this knowledge base, ProtoSR is trained to retrieve prototypes relevant for the current image-question pair and augment the model predictions through a prototype-conditioned residual, providing a data-driven second opinion that selectively corrects predictions. On the Rad-ReStruct benchmark, ProtoSR achieves state-of-the-art results, with the largest improvements on detailed attribute questions, demonstrating the value of integrating free-text derived signal for fine-grained image understanding.
arXiv:2602.22647v3 公告类型：替换版本 摘要：生成式检索已成为基于 LLM 的推荐的一种强大范式。然而，工业推荐系统通常需要根据业务逻辑将输出空间限制为物品的受约束子集（例如，确保内容的新鲜度或产品类别），而标准的自回归解码原生不支持这一点。此外，现有利用前缀树（Trie）的约束解码方法在硬件加速器（TPU/GPU）上会产生严重的延迟开销。在本工作中，我们提出了 STATIC（Sparse Transition Matrix-Accelerated Trie Index for Constrained Decoding），一种专为 TPU/GPU 上高吞吐量的基于 LLM 的生成式检索而设计的高效且可扩展的约束解码技术。通过将前缀树展平为静态的压缩稀疏行（Compressed Sparse Row，CSR）矩阵，我们将不规则的树遍历转换为完全向量化的稀疏矩阵运算，从而释放硬件加速器的巨大效率优势。我们在一个服务数十亿用户的大规模工业视频推荐平台上部署了 STATIC。STATIC 以极小的延迟开销带来了显著的产品指标提升（每步 0.033 毫秒，占推理时间的 0.25%），相较于 CPU 字典树实现实现了 948 倍加速，相较于硬件加速的二分搜索基线实现了 47—1033 倍加速。此外，在广泛的实际配置范围内，STATIC 的运行时开销始终极低。据我们所知，STATIC 实现了首个生产规模的严格约束生成式检索部署。此外，学术基准测试评估表明，STATIC 能够显著改善生成式检索的冷启动性能。我们的代码可在 https://github.com/youtube/static-constraint-decoding 获取。
arXiv:2412.08843v3 Announce Type: replace-cross Abstract: Allocation stability is often used to justify Gaussian inference from bandit data, but when is it necessary? In this paper, we address this question for a two-armed, fixed-horizon variance-aware UCB policy with bounded reward distributions that may vary with the horizon. We find a sharp criterion in terms of the reward gap and variances that determines whether the optimal-arm count admits a deterministic approximation with vanishing relative error, while the suboptimal-arm count is always stable. Despite the possible instability of the optimal-arm count, we show that the ordinary Wald statistic for a linear combination of the arm means has a standard normal limit for every fixed nonzero coefficient vector, provided the product of the pull count and reward variance diverges in probability for each arm. Under the same condition, however, this Gaussian approximation holds uniformly over deterministic nonzero coefficient vectors if and only if the optimal-arm count is stable. The analysis relies on two main ingredients: (i) a pathwise comparison with an auxiliary policy whose final optimal-arm count is asymptotically equivalent to the original count and independent of the optimal-arm reward sequence; and (ii) joint limits for the rescaled optimal-arm count and the two studentized sample-mean errors under the original policy, which yield nonstandard Wald limits for certain linear combinations of the arm means with coefficients that vary with the horizon.
arXiv:2609.35603v2 Announce Type: replace Abstract: Joint-embedding predictive architectures enable planning with latent world models, but accurate transition prediction alone does not ensure that the planning objective is easy to optimize. We introduce Control-Geometry Straightening (CGS), a single auxiliary loss that learns planner-friendly representations by directly straightening control geometry for sampling-efficient planning. CGS matches pairwise cosine similarities among actions to those among corresponding latent differences only using local transitions from pixel-action pairs. The loss can be applied across world-model architectures using end-to-end learned or pretrained representations. Under linear-dynamics, our theoretical analysis connects this objective to temporal straightening and more balanced terminal-cost curvature across the full planning horizon, yielding finite-budget guarantees for MPPI, local contraction results for CEM, and convergence bounds for gradient descent. Across four control environments and multiple planners, CGS improves planning with fewer sampled candidates and refinement steps, achieving success-rate gains up to 20 and 12.6 percentage points over LeWorldModel (LeWM) and its temporal-straightening variant (LeWM+TS), respectively, with sampling-based planners using 128 candidates per update. Probes, comparisons with DINO-WM architecture, and planner-side ablations clarify how latent motion organization, state dependence, and dynamical context shape planning behavior. Straightening control geometry thus makes good action sequences easier to find under limited planning budgets.
arXiv:2609.26855v3 Announce Type: replace Abstract: Relational Deep Learning (RDL) models multi-table databases as heterogeneous temporal graphs, and graph transformers currently achieve state-of-the-art performance on benchmarks like RelBench. However, the current leading model, RelGT, suffers from two key limitations: its random local sampler yields loosely connected subgraphs that hinder message passing, and its global attention module relies on a single, seed-feature-based memory that ignores broader macro-level dynamics. To overcome these limitations, we introduce QUARTET, an expressive graph transformer architecture that applies full self-attention on local subgraphs while enriching global context through cross-attention branches. Specifically, QUARTET employs a Causal Random Walk (CRW) sampler based on recency-truncated Personalized PageRank (PPR) to extract compact, hub-robust, and densely connected local subgraphs without temporal leakage. Concurrently, a quad-branch cross-attention module integrates global context from four complementary perspectives: seed feature, seed topology, temporal dynamics, and collaborative dynamics. Across the RelBench v1 classification tasks, QUARTET consistently matches or outperforms the current state-of-the-art graph transformer baselines (HGT and RelGT). Ablation studies confirm that the CRW sampler significantly enriches local neighborhood quality, while the global branches provide essential, task-specific predictive gains.
arXiv:2609.11917v2 Announce Type: replace Abstract: As the supply of human-written text is exhausted, it has become standard practice to repeat language model training data. Prior work has studied data repetition for densely activated Transformers, but the effects of data repetition remains largely unexplored for recently dominant sparse architectures such as Mixture-of-Experts (MoE), despite their increased compute efficiency. We vary data repetition rates across single- and multi-domain data mixes, and across MoE settings, including expert count and granularity. We consistently find, for models ranging from 80M to 1B active (8.5B total) parameters, that MoEs degrade more rapidly under data repetition. This effect increases with sparsity, dictated by total rather than active parameters. While 80M dense models can repeat data over 8x with minimal degradation, MoEs instead begin to suffer at 4x, and deteriorate rapidly, ceding their performance benefits in all-unique data settings to underperform dense models after 32x. We experiment with existing regularization methods as a potential remedy. We find that some methods, such as dropout, can mitigate overfitting. In particular, with strong masking-based regularization, MoEs are able to outperform dense models even when data is repeated more than 64 times. However, no method fully matches the performance of all-unique training data. Finally, we analyze internal mechanisms correlated with MoE overfitting in high repetition regimes, and find that MoE routing universally stabilizes early in training, and that expert specialization correlates with overfitting to repeated data. In sum, our work addresses the adverse interactions between sparsity and data repetition: we present evidence for the core mechanisms of overfitting and its potential remediation, and suggest promising avenues for future methods to reduce over-specialization in model parameters by disrupting memorization patterns.
arXiv:2607.07683v2 Announce Type: replace Abstract: Electrocardiography (ECG) is one of the most widely used tests for diagnosing cardiovascular disease. Yet several remote clinics still utilize paper ECG printouts for their analysis due to limited connectivity and computational capacity. As a result, vast numbers of physical ECGs obtained in remote areas still remain incapable of being accessed by contemporary artificial-intelligence (AI)-based decision support as they require high computational resources or strong high-speed internet connectivity. This causes several cases where conditions like acute coronary occlusion (ACS) is overlooked and reperfusion therapy delayed. Although prior work has tackled digitization and diagnosis separately, and utilized advanced AI models for them, there still remains a lack of a compute-light, on-device framework that reconstructs paper ECGs at high fidelity, while accurately supporting multiple clinically relevant endpoints. We address this need with an end-to-end lightweight on-device digitization-to-diagnosis pipeline that converts a smartphone photo or scan of a paper ECG into a calibrated 12-lead signal and screens for Myocardial Infarction (MI) pathologies, with SHapley Additive exPlanations (SHAP) to support interpretability. Trained and evaluated on 21,799 ECGs from the PTB-XL dataset and further validated on hospital-acquired ECG-Matrix dataset, the complete system runs in <30 s per ECG on CPU-only resources, achieving 95.51% accuracy (F1 = 0.9519) for MI detection on PTB-XL and 88.89% accuracy (F1 = 0.8862) for OMI detection on ECG-Matrix. This work showcases that legacy paper records can be reliably democratized in any part of the world, providing a scalable decision support when digital ECG export, connectivity, or high-end compute are unavailable. Github: https://github.com/nshreyasvi/ECGLight
arXiv:2606.31804v2 Announce Type: replace Abstract: Accurate energy demand forecasting is essential for the reliable operation and planning of modern sustainable energy systems. Spatial-temporal graph neural networks (STGNNs) have recently achieved strong performance in point forecasting by jointly modeling temporal dynamics and relational dependencies across interconnected energy nodes. However, in real-world energy systems, accurate point forecasts alone are insufficient, as operators also require reliable uncertainty estimates to support risk-aware decision-making, grid stability, and operational planning under uncertainty. Conformal prediction provides a principled and model-agnostic framework for uncertainty quantification under exchangeability assumptions, making it particularly attractive for safety-critical energy applications. However, existing conformal prediction approaches often fail to fully capture the complex spatial-temporal structure of energy systems. To address these limitations, we propose STOIC (Spatial-Temporal uncertainty quantificatiOn via In-Context learning), a conformal-inspired post-hoc residual calibration framework that integrates graph-based forecasting with the in-context learning capabilities of tabular foundation models. STOIC first generates point forecasts using an STGNN and subsequently reformulates spatial-temporal residuals into a tabular representation suitable for in-context learning. Leveraging a tabular foundation model, STOIC estimates feature-conditional residual quantiles without task-specific retraining of the calibration model, effectively capturing both sequential and relational dependencies. STOIC should be viewed as a conformal-inspired residual calibration method and does not by itself provide a finite-sample distribution-free coverage guarantee under arbitrary temporal dependence...
arXiv:2606.03904v2 公告类型：替换 摘要：多目标优化（MOO）是许多机器学习问题的基础，但损失平衡、梯度平衡和基于 Pareto 的各类 MOO 求解器几乎都将其协调后的方向交由 Adam~\citep{kingma2015adam} 处理。我们表明，这种耦合在求解器的意图与优化器的执行之间引入了两种系统性偏差。第一种是权重失配：Adam 的二阶矩分母将随时间变化的偏好向量与梯度统计量纠缠在一起，将偏好边缘化为历史平均值，并使不同的 Pareto 权衡坍缩为近乎均匀的混合。第二种是几何失配：Adam 的自适应度量扭曲了 MOO 求解器所假设的欧几里得几何，使方向一致的目标呈现出表观冲突。为同时解决这两种失配，我们提出 MAdam（度量感知的多目标 Adam），这是一种即插即用的封装，无需更改求解器或优化器。MAdam 利用标量化目标在给定偏好条件下的曲率，对协调后的方向进行预条件处理；对于这种白化后的输入，Adam 的二阶矩退化为单位矩阵，因此实际更新由偏好条件化的度量支配。在多任务学习、Pareto 前沿恢复、物理信息神经网络和医学成像任务中，对于每一类求解器，MAdam 的总体表现均优于 Adam。我们的代码已发布于 https://github.com/lfb-1/MAdam。
arXiv:2605.24278v2 公告类型：替换 摘要：我们提出了一种改进的神经场架构，用于求解偏微分方程（PDE）。当前的物理信息神经网络（PINN）为求解PDE提供了灵活的框架，但难以获得高精度解，并且其计算需求随参数数量增长不理想。我们将模型称为 beignet（Bandlimited Embedding with Interpolated Grid Network），它以可训练的多分辨率傅里叶特征金字塔取代了现有PINN模型所使用的随机傅里叶特征嵌入。为了在连续坐标处查询 beignet，我们在金字塔的每一层使用傅里叶插值，返回输入坐标处的特征，然后通过全连接神经网络主干对该向量进行解码。我们的模型具有多项优势：1）可以利用链式法则，将通过自动微分计算的神经网络导数与通过快速傅里叶变换（FFT）以谱方法计算的特征网格导数组合起来，从而高效地计算空间导数。2）beignet通过扩展傅里叶特征金字塔的参数数量，以计算高效的方式实现更高精度，而不是采用扩展神经网络架构这一效率较低的策略。3）beignet能够直接控制表示的带限范围，从而为困难的PDE实现更稳定的优化。我们在PDE基准测试中展示了，与当前最先进的PINN方法相比，beignet使用更少的参数即可获得显著更准确的解。我们还在自相似无黏Burgers爆破解问题上对 beignet 进行了评估，结果表明，使用 Adam 即可将残差降低至接近机器精度；此前只有借助计算开销高昂的高阶优化器才能达到这一精度水平。
arXiv:2605.17692v2 Announce Type: replace Abstract: We show that the training problem of a deep linear neural network under the squared loss admits an exact convex reformulation in a lifted space over a generalized completely positive cone. The reformulation has the same optimal value as the original nonconvex problem and is linear in the lifted variables, with all nonconvexity encoded in the cone constraint. Its ambient lifted dimension depends only on the input and output dimensions, independent of the network depth and the number of data points, and the bottleneck width enters only through scalar constraints. The construction proceeds by reducing the multilayer parameterization to a bilinear factorization, lifting it to a rank-constrained semidefinite program, expressing the rank constraint via a complementarity condition, and applying a completely positive lifting. The resulting formulation gives a conic representation of the nonconvexity induced by linear factorization and connects linear neural network training with copositive programming.
arXiv:2605.15995v3 公告类型：替换 摘要：从时序、多模态和部分观测数据中学习潜在表示，需要明确潜在状态应保留、舍弃和组织哪些信息。现有方法通过异构目标对这些要求进行编码，使得方法难以比较，学习到的表示也难以解释。我们提出受约束潜在状态建模（Constrained Latent State Modeling，CLSM），这是一个通过六项互补属性刻画潜在状态的概念框架：预测充分性、最小性、时间一致性、观测兼容性、对干扰因素的不变性以及结构约束。CLSM 将这些属性与用于诱导它们的替代目标，以及用于评估它们的诊断指标区分开来，并阐明约束组合如何通过限制可接受表示的空间来提升可辨识性。我们通过这一共同设计空间重新诠释了主要的表示学习方法族，并使用受控合成基准展示该框架。实验表明，根据预测目标、替代目标的表述方式、参数化形式和优化过程的不同，各类目标会产生不同的潜在组织方式和经验权衡。配套仓库包含参考实现、可复现实验、文档和模型卡片：https://github.com/gwenole-quellec/clsm
arXiv:2602.08234v2 公告类型：替换 摘要：大型语言模型（LLM）智能体在复杂任务中展现出惊人的效果，但它们往往孤立运行，无法从过往经验中学习。现有的基于记忆的方法主要存储原始轨迹，而这些轨迹通常冗余且噪声较多，导致智能体难以提取对泛化至关重要的高层次、可复用行为模式。本文提出 SkillRL，这是一种通过自动技能发现和递归演进，弥合原始经验与策略改进之间鸿沟的框架。我们的方法引入了基于经验的蒸馏机制，用于构建分层技能库 SkillBank；引入了自适应检索策略，用于获取通用启发式方法和任务特定启发式方法；还引入了递归演进机制，使技能库能够在强化学习过程中与智能体策略共同演进。这些创新显著减少了 token 占用，同时提升了推理效用。在 ALFWorld、WebShop 和七项搜索增强任务上的实验结果表明，SkillRL 达到了当前最佳性能，相较于强基线方法的表现提升超过 15.3%，并且随着任务复杂度增加仍能保持稳健性。代码发布于 https://github.com/aiming-lab/SkillRL。
arXiv:2511.17776v2 Announce Type: replace Abstract: We present PrismSSL, a Python library that unifies state-of-the-art self-supervised learning (SSL) methods across audio, vision, graphs, and cross-modal settings in a single, modular codebase. The goal of the demo is to show how researchers and practitioners can: (i) install, configure, and run pretext training with a few lines of code; (ii) reproduce compact benchmarks; and (iii) extend the framework with new modalities or methods through clean trainer and dataset abstractions. PrismSSL is packaged on PyPI, released under the MIT license, integrates tightly with HuggingFace Transformers, and provides quality-of-life features such as distributed training in PyTorch, Optuna-based hyperparameter search, LoRA fine-tuning for Transformer backbones, animated embedding visualizations for sanity checks, Weights & Biases logging, and colorful, structured terminal logs for improved usability and clarity. In addition, PrismSSL offers a graphical dashboard - built with Flask and standard web technologies - that enables users to configure and launch training pipelines with minimal coding. The artifact (code and data recipes) will be publicly available and reproducible.
arXiv:2509.08483v2 Announce Type: replace Abstract: We analyze gradient descent with Polyak (1964) heavy-ball momentum (HB) whose fixed momentum hyperparameter $\beta \in (0, 1)$ provides exponential decay of memory. Building on Kovachki and Stuart (2021), we prove that on an exponentially attractive invariant manifold the algorithm is exactly plain gradient descent with a modified loss, provided that the step size $h$ is small enough. Although the modified loss does not admit a closed-form expression, we describe it up to $O(h^{\mathcal{R}})$-errors for arbitrary finite order $\mathcal{R}$, and prove global (finite "time" horizon) trajectory approximation bounds $O(h^{\mathcal{R}})$. We then conduct a fine-grained analysis of the combinatorics underlying the memoryless approximations of HB, in particular, finding a rich family of polynomials in $\beta$ hidden inside which include and lie coefficient-wise in between Eulerian and Narayana polynomials. We prove that these polynomials are $h$-polynomials of certain graph-associahedra. As corollaries of the main results, we derive continuous modified equations of arbitrary finite approximation order (with rigorous bounds) and the principal flow that approximates the HB dynamics, generalizing Rosca et al. (2023). Approximation theorems cover both full-batch and mini-batch HB. The results shed new light on the main features of HB and outline a roadmap for similar analysis of other optimization algorithms.
arXiv:2610.10526v1 发布类型：cross 摘要：视觉-语言-动作模型（VLAs）对指令措辞表现出显著敏感性，并未继承其所基于的视觉-语言模型的语言鲁棒性。仅修改一个词就可能使成功率变化几十个百分点：对于“switch on the stove”，$\pi_{0.5}$ 在 LIBERO 中每次都能成功打开炉子，成功率为 100%；而对于“switch on the hot plate”，成功率仅为 2%。即使经过改写增强微调的 $\pi_0$ checkpoint，成功率波动仍最高可达 61 个百分点。我们通过经过统计检验的单一编辑变化和 oracle 短语搜索来刻画这种敏感性。结果表明，仅凭措辞就几乎可以弥合分布内任务与分布外任务之间 21 个百分点的差距。随后，我们在不修改策略的情况下减弱这种敏感性。由于这种敏感性具有系统性，因此可以将其表示为显式规则：我们为少量训练任务评分大量不同措辞，再让大型语言模型将这些证据提炼为十到二十条改写规则，并在部署时依据这些规则对每条输入指令改写一次。在十二个留出任务上，针对对抗式、VLM 生成和人工生成的措辞，这些规则使冻结的 $\pi_0$ 的相对性能提升了 16% 至 27%，且提升主要集中在分布外任务上。该流程在 $\pi_{0.5}$ 和 LIBERO 上也得到了复现，使 in-finetune 成功率从 93.6% 提升至 97.8%。该方法无需重新训练，也无需逐步验证，并且可以零样本应用于未见过的任务和指令。项目网站：https://sttawm.github.io/rephrase-before-you-act
arXiv:2610.10496v1 Announce Type: cross Abstract: Small tabular datasets with expert-designed spectral features are the norm in operational Earth observation, and the networks applied to them are typically hand-designed. We revisit one such published model -- a Sentinel-2 algal bloom classifier -- and ask what architecture search adds, holding the task, the features and the lake-level train/test split of the original study fixed. Searching an extended multilayer-perceptron space with regularized evolution, and selecting on inner-cross-validation AUC only, we find networks that improve held-out AUC from 0.790 to 0.820 and accuracy from 0.733 to 0.748 while using 409 trainable parameters, 26 times fewer than the strongest hand-designed reference. The search converges on a consistent recipe -- a single narrow layer, RMS normalisation, $\tanh$ activation, step-decayed RMSprop and weight averaging -- that a practitioner would be unlikely to reach by default. At 1.6\,kB the resulting model is small enough to serve as an onboard screening trigger, which is the setting that motivates the work. Code: https://github.com/VU-AIML/automl4eo-bloom-nas.
arXiv:2610.10415v1 Announce Type: cross Abstract: Pretrained text-to-speech (TTS) models can generate expressive speech, but reliable inference-time emotion control remains challenging: prompts and reference audio offer coarse, inconsistent control, whereas specialized conditioning and model adaptation require costly training. We present SteerSpeech, a lightweight activation-steering framework that controls emotion by injecting steering vectors into hidden activations. For each target emotion we train a lightweight low-rank transform, using a multi-expert objective that encourages monotonic emotion control while preserving speaker identity and linguistic content, constraining steering drift, and keeping the TTS backbone frozen. To optimize through discrete speech tokens, we introduce a two-pass generation-and-replay pipeline using a straight-through estimator to backpropagate expert supervision through sampled tokens. At inference, a target-emotion steering direction is optimized with its respective transform and injected into the base TTS model. Objective and subjective evaluations with Qwen3-TTS across seen, unseen, and accented speakers show stronger continuous emotion control with limited speaker and content degradation. SteerSpeech achieves 1.08x-7.12x baseline target-emotion scores and for a representative emotion subjectively, it receives 78.1%-96.8% intensity preference and 1.43x-1.46x speaker-identity preservation at high steering strengths.
arXiv:2610.10407v1 Announce Type: cross Abstract: As option markets grow and AI advances, agentic systems for option trading are gaining increasing attention. Language-model-based agents can reason over contextual information such as news, but option trading presents a particularly challenging decision problem: a single stock can have thousands of contracts, and the agent must decide both which contracts to trade and how to combine them. Existing approaches often sidestep this complexity by restricting the policy to a fixed strategy structure, such as a straddle, limiting their ability to switch strategies as market conditions change. We present SOTA (Stock Options Trading Agents), an agentic trading framework for structured option-strategy selection. SOTA abstracts the large option universe into strategy-level decisions while deterministic resolvers handle portfolio implementation. We develop SOTA by post-training Qwen3.8-27B with supervised fine-tuning followed by reinforcement learning. SOTA is evaluated on options on nine large-cap U.S. equities and SPY against rule-based and machine-learning strategy selectors in the same trading environment. Over a six-month out-of-sample period, SOTA earns an 18.3% total return with a Sharpe ratio of 1.60 and a maximum drawdown of 8.96%. We also document an asymmetric role of news: news improves frontier-teacher trajectories, but retaining news during reinforcement learning reduces out-of-sample return from 18.3% to -2.7%.
arXiv:2610.10351v1 Announce Type: cross Abstract: Differentiable quantum architecture search (DQAS) is a promising framework for the automated design of quantum circuits, particularly for variational quantum optimization algorithms. However, its practical deployment on quantum hardware is limited by the large number of circuit measurements required during optimization, making hardware execution costly. In this work, we show that for a broad class of combinatorial optimization problems and commonly used rotational gate parameterizations, the measurement cost of DQAS can be significantly reduced without changing the optimization objective. We derive the proposed measurement reduction scheme theoretically and validate it experimentally on 3-SAT and MaxCut benchmark problems. Our approach reduces the requested gradient measurement cost by about 39 to 41% while introducing only negligible classical post-processing overhead, lowering the practical cost of executing DQAS on quantum hardware.
arXiv:2610.10285v1 Announce Type: cross Abstract: In the wake of the increasingly public discussion about AI alignment, recent work has tried to propose specific AI architectures that behave safely. However, the proposed arguments that seemingly demonstrate proved alignment mostly neglect the environment the agent needs to act in. We discuss theoretical bounds for agent-agnostic safety guarantees in environments that can only be partially observed and within which an action is required within limited time. We introduce two realistic scenarios, one with an infinite state space and one with signal mixture. In these scenarios, we prove that even a perfect agent cannot guarantee safe behaviour. It will be argued that for any proof of AI safety or alignment, the environment and associated safe actions need to be specifically considered together with the agent.
arXiv:2610.10283v1 公告类型：cross 摘要：人类利用指尖触觉反馈抓取日常物体时，几乎可以达到完美的成功率；然而，机器人抓取领域的大量研究都侧重于使用平行夹爪进行基于视觉的抓取选择。在本工作中，我们系统研究了高分辨率动态触觉感知如何促进灵巧机器人手的抓取稳定性预测和模型引导抓取。为此，我们使用一只配备四个 Digit 360 触觉传感器的多指机器人手，针对 200 个物体采集了 10,000 次抓取试验的数据集，并在每次抓取过程中记录外部视觉、本体感觉和触觉数据流。基于该数据集，我们训练了端到端时序多模态模型，根据抬升前的抓取观测预测抬升后的稳定性，并比较了不同感知模态和编码骨干网络的效果。实验结果和受控输入消融实验表明，引入触觉信息，尤其是高分辨率动态触觉信息，能够提升抓取稳定性预测性能。最后，我们将训练得到的预测器部署为真实机器人上的在线稳定性门控机制；与非触觉门控相比，视觉-触觉模型引导的重新抓取使已执行抬升动作的成功率提高了 10.5 个百分点。这些结果表明，丰富的指尖感知和能够捕捉触觉动态的表达性时序模型，无需显式的接触或力建模，即可支持多指机器人手实现学习式抓取，为从触觉经验迈向稳定灵巧操作提供了一条可扩展的数据驱动路径。该数据集已公开发布于 https://lasr-lab.github.io/dexterous-grasp-stability/。
arXiv:2610.10261v1 Announce Type: cross Abstract: Context: Once defined a taxonomy of stages structuring Machine Learning (ML) pipelines (e.g. Data Preprocessing, Modeling...), extracting these stages from source code is key for better understanding ML practices. However, the diversity caused by the constant evolution of ML (e.g., algorithms, datasets) makes this task challenging. Existing approaches either rely on non-scalable manual labeling or on classifiers that do not properly support domain's diversity. These limitations call for more reliable solutions. Objective: We evaluate whether Small Language Models (SLMs) can leverage their code understanding and classification abilities to address these limitations, and enhance our understanding of practices in ML. Method: We conduct a confirmatory study based on two relevant reference works representing current limitations in the state-of-the-art. We first compare several SLMs using Cochran's Q test, then evaluate the best-performing model against reference studies via two McNemar's tests. An additional Cochran's Q test examines how taxonomy definition variations affect the SLM performance. Finally, goodness-of-fit tests compare ML practice insights from SLM classification with those from prior studies. Results: First, we found that the taxonomy wording significantly impacts classification performance. Second, the best performing SLM yielded good results, yet, without outperforming other classifiers. Third, the three classification methods led to significantly different insights, with varying effect sizes, when exploring practices of data scientists. Conclusions: Limitations of existing classification methods bias our understanding of ML practices. While current SLMs show promising results without prior fine-tuning, they still exhibit common limitations, in addition to inference high costs challenging their applicability in large-scale studies.
arXiv:2610.10194v1 Announce Type: cross Abstract: Switching stochastic differential equations (SSDEs) describe continuous-time dynamics whose parameters switch according to a latent regime process that follows a continuous-time Markov chain (CTMC). By allowing dynamics to change between regimes, SSDEs represent heterogeneous system behavior and have been applied across diverse fields. However, Bayesian inference for SSDEs remains difficult, and existing SSDE inference methods have limited applicability, with restrictions such as noise-free observations, univariate states, linear drift, or state-independent diffusion. In this study, we propose an approximate Markov chain Monte Carlo sampler for SSDEs using uniformization and factorized neural likelihood estimation (FNLE), a simulation-based inference method. Uniformization provides an exact representation of the CTMC but requires SDE transition densities over arbitrary time intervals. We approximate these densities by training a time-conditioned FNLE model. The resulting sampler is broadly applicable to SSDEs without requiring analytically tractable transition densities. In synthetic-data experiments, our method recovered regime paths and parameters for three SSDE models for which previous methods have limited applicability. We also applied our method to a real dataset and detected a regime transition.
arXiv:2610.10140v1 Announce Type: cross Abstract: Artificial intelligence (AI) is becoming an increasingly useful tool across the experimental sciences, including angle-resolved photoemission spectroscopy (ARPES), which routinely produces large, multidimensional datasets of electronic structure. Recent advances in AI and machine learning (ML) have opened new opportunities across the entire ARPES workflow, from automated sample preparation and real-time data acquisition to post-experiment data analysis and comparison with theoretical calculations. Despite this progress, a comprehensive review of ML applications, their capabilities, and reliability across the different stages of ARPES workflow is still lacking. In this review, we first introduce ML methods that are most relevant to experimentalists working in condensed matter physics and materials science. We then follow the ARPES workflow, reviewing existing ML applications at each step and discussing their advantages, limitations and potential for future development. We also examine the current ARPES data landscape, where several open databases are available but remain relatively small and fragmented compared with large, shared datasets such as ImageNet. Given these limitations, we suggest that the community focus on sharing pretrained models that can be further trained, adapted to specific tasks, and redistributed, while working toward a larger and standardized open ARPES dataset repository. Finally, we discuss our perspectives on the future of AI within the ARPES workflow using a six-level framework of laboratory automation, highlighting the opportunities and challenges in moving toward a fully autonomous, self-driving ARPES laboratory.
arXiv:2610.10061v1 Announce Type: cross Abstract: This paper presents TRACK (Telemetry-Based Racing Analysis and Coaching Kit), which is a framework for analyzing driving performance in sim racing and profiling how individual drivers behave behind the wheel. We report this framework together with its limitations: we calibrate each clustering result against a null, and when one does not separate from chance, we say so. Instead of restricting ourselves to scoring drivers or sorting them into preset labels, we represent each recording session as a compact geometry in a four-dimensional behavioral space (speed, braking, strategy, and consistency), and we group these fingerprints by their similarity using unsupervised clustering. Over time, we have developed and refined this framework on the open Assetto Corsa Gym (ACGym) dataset. Our study suggests that corner types differ along a behavioral dimension that was not used to define them. It also suggests that when the car changes, only speed and consistency carry over in the restricted population, while repeatability could not be shown there for any of the braking or strategy measures. Cluster separation becomes less distinct as the range of available telemetry widens. Until that repeatability is shown, grouping on the braking and strategy dimensions cannot treat the car as interchangeable, which divides an already small sample into smaller cells. It is also not clear whether a driver's grouping carries over from one corner type to the next. We also normalize each metric against a reinforcement-learning reference agent. The reference does not depend on the sample, so the scale does not shift when the sample does. We intend these results as an analytical foundation for a personalized improvement suggestion system. The sample is small. The cross-car result changes when the sample is defined more broadly. These outcomes are preliminary.
arXiv:2610.10033v1 Announce Type: cross Abstract: A theoretical understanding of multi-head self-attention is fundamental to the study of modern neural networks. Using random matrix theory, we establish Gaussian equivalence for multi-head self-attention: replacing softmax attention with rescaled scores plus Gaussian noise preserves the limiting spectral law of the centered output. This equivalence also covers value and output projections that depend on the keys. The resulting laws separate the effects of head allocation and projection widths, and distinguish spectrum-preserving across-head sharing from within-head key--value dependence.
arXiv:2610.10021v1 Announce Type: cross Abstract: Mutual information (MI) provides an objective for suppressing or encouraging statistical dependence in implicit generative models. However, direct MI evaluation is challenging in implicit models due to typically intractable densities. A remedy is estimating the generator gradient from the difference between conditional and marginal scores. This score difference can, in turn, be estimated by differentiating a log density ratio learned through classification. This construction nevertheless faces two difficulties: (i) singular distributions need not admit the required score functions, and (ii) poor overlap can hinder density-ratio estimation. We therefore introduce Spread Mutual Information (SMI), a weighted integral of MI across noise levels obtained by applying a common spreading kernel to the generated variable. Gaussian spreading yields smooth, strictly positive conditional and marginal densities, extending the gradient construction to distributions that may originally be singular. Across a variaty of experiments, SMI consistently achieves effective dependence control among MI-based methods and remains competitive with established task-specific approaches.
arXiv:2610.09860v1 Announce Type: cross Abstract: 4D point clouds acquired by permanent laser scanning (PLS) enable accurate high-frequency monitoring of surface change in dynamic topographic environments. However, existing methods remain limited in organizing detected surface activities into meaningful process types. We propose DeepTopoClustering (DTC), an unsupervised framework for deriving a hierarchical process taxonomy from object-based surface activities, so-called 4D objects-by-change (4D-OBCs). We transform each 4D-OBC into a GeoMorphogram, a distributional sequence representing the temporal evolution of topographic change within a spatially bounded surface activity. A convolutional autoencoder learns latent embeddings from GeoMorphograms, which are jointly optimized using a hierarchical deep clustering objective to organize surface activities into a hierarchy. We evaluate the learned hierarchy using expert annotations on two 4D datasets of sandy beach sites and their combination. DTC with GeoMorphograms achieves the highest agreement with expert judgment at the taxonomy level comprising eight major process types ($F_1=0.78$, match accuracy $=0.92$), outperforming dimensionality reduction and conventional flat clustering. The learned taxonomy separates major erosion- and deposition-dominated activities and distinguishes finer subtypes based on change magnitude, duration, compactness, and temporal evolution. DTC thus provides a scalable and interpretable route from 4D change detection to a data-driven, expert-supported surface process taxonomy, advancing automated knowledge derivation for understanding surface dynamics in topographic monitoring.
arXiv:2610.09778v1 公告类型：交叉收录 摘要：大语言模型（LLM）推理的可复现基准测试面临挑战，因为重复测量结果可能随执行状态和系统状态而变化。我们提出了顺序隔离方法，这是一种受控的基准测试与回归测试协议，旨在有意改变工作负载并发度的同时，降低不同运行之间的测量方差。我们使用 vLLM 0.9.1，在 NVIDIA A100 80GB GPU 上评估了三个具有代表性的开源 LLM，覆盖六种上下文长度和八个并发级别，每种配置重复测试五次。最终协议将平均变异系数（CV）从控制最少的方法阶段的 15.2% 降至 2.2%；根据每种配置五次重复测试各自的首 token 延迟（TTFT）中位数（P50）计算 CV，144 种配置中有 113 种（78.5%）的 CV 低于 3%。测量结果还表明，在受测技术栈上，并发用户数从 200 增至 500 时出现了明显的延迟跃变，且三个模型的 P99 延迟呈现出描述性差异。此外，我们还提供了一个显式的成本盈亏平衡模型，并分析其对 API 定价的敏感性。该协议旨在为可复现比较和回归测试提供稳定的参考，而非预测不受控生产流量下的绝对表现。基础设施即代码和基准测试脚本支持复现实验环境。
arXiv:2610.09703v1 公告类型：交叉发布 摘要：词元剪枝通过移除冗余视觉词元来加速视觉语言模型，但其安全影响尚未得到充分研究。在本研究中，我们首次对词元剪枝机制进行了全面的安全评估，发现：随着剪枝比例增加，大多数剪枝策略会显著降低安全性；而基于查询的压缩则呈现相反趋势，在极端剪枝（比例高达 99.8%）下，模型安全性反而意外提升。这一鲜明对比引出了一个关键问题：不同的词元剪枝策略如何重塑模型的安全行为？能否在不牺牲加速效果的情况下提升安全性？为回答这一问题，我们识别出一种此前未被认识的机制，称为“剪枝诱导的恶意放大”：移除背景词元会引发一种副作用，迫使模型的注意力集中到前景中少数保留下来的恶意锚点上，从而在越狱攻击下无意间放大其有害语义。为解决这一问题，我们提出了一种在推理阶段运行、即插即用的安全感知剪枝（SAP）机制，通过三个步骤抵消这种主导效应：（1）识别恶意锚点；（2）恢复被剪除的良性词元；（3）将恶意锚点获得的过量注意力重新分配给良性词元。在三个安全性基准和四个实用性基准上开展的大量实验表明，SAP 能够在不损害效率或实用性的前提下缓解剪枝引发的漏洞，即将攻击成功率（ASR）最多降低 62%。
arXiv:2610.09691v1 Announce Type: cross Abstract: We study online learning in market making, where, at each round, a market maker posts bid and ask prices before observing the market price and the private valuation of an incoming trader. In this setting, Maran et al. 2026 introduce a feedback model motivated by limit order books, in which the trader's valuation is revealed only if no transaction occurs. Assuming that trader valuations are drawn i.i.d. from an unknown distribution while market prices are chosen adversarially, they establish an expected regret bound of $\widetilde{\mathcal{O}}(T^{2/3})$. In this work, we improve upon this guarantee by establishing a high-probability regret bound of $\widetilde{\mathcal{O}}(\sqrt{T})$. As a warm-up, we first consider the full-feedback setting. We introduce a discretization of the bid-ask space based on two coupled grids and combine it with Hedge to achieve the desired regret rate. Building on these ideas, we then address the substantially weaker feedback induced by a limit order book and develop an algorithm that achieves the same guarantee. Finally, we investigate the limits of learnability in fully adversarial environments, where the valuations may vary arbitrarily as well. Perhaps surprisingly, we show that when both market prices and trader valuations are chosen adversarially, sublinear regret is impossible even under full feedback, thereby motivating our stochastic assumption on the valuations.
arXiv:2610.09690v1 Announce Type: cross Abstract: Autonomous underwater vehicles (AUVs) commonly rely on inertial navigation systems (INS) aided by Doppler velocity logs (DVLs) for reliable underwater navigation. Accurate DVL velocity estimation is therefore essential for successful operation. Recent learning-based methods have demonstrated improved DVL velocity estimation, particularly under degraded measurement conditions. However, their training objectives typically rely on conventional regression losses that are highly sensitive to large residuals and corrupted observations. Additionally, they do not explicitly account for the physical consistency and measurement uncertainty associated with the underlying sensing process. To address these limitations, this paper introduces navigation-aware loss (NAViLoss), a robust and uncertainty-aware objective function for learning-based AUV velocity estimation. NAViLoss jointly penalizes the velocity-estimation residual in the navigation-state domain and the beam-consistency residual in the DVL measurement domain. Its bounded formulation limits the influence of large residuals, while an adaptive mechanism regulates the uncertainty in beam geometry. Furthermore, NAViLoss is integrated with a DeepONet architecture to form a novel NAVi-DeepONet model for seamless estimation of an underwater vehicle's velocity. Lastly, our model is evaluated using approximately 10,000m of semi-synthetic AUV experimental data collected during multiple real-world sea trials. Experimental results demonstrate a 44% improvement in velocity-estimation accuracy compared with conventional and learning-based baselines. These results demonstrate the effectiveness of navigation-aware and uncertainty-adaptive loss design for robust learning-based underwater velocity estimation.
arXiv:2610.09677v1 公告类型：交叉收录 摘要：自监督异常检测是医学超声领域一种前景广阔的范式，因为与涵盖所有可能病变的详尽标注相比，正常图像通常更容易获取。然而，现有评估大多局限于单一解剖部位或任务，因此尚不清楚模型学习到的是对正常超声外观的稳健认识，还是仅针对特定来源的表征。我们提出 SADUSI 基准，这是一个多源超声数据集，旨在支持跨广泛解剖区域、切面和采集协议的异常检测方法训练与评估。SADUSI 的目标是提供多样化的正常超声图像分布，并为可通过单幅图像评估的可见结构异常提供基准。我们评估了具有代表性的自监督异常检测方法，发现现有方法在这一场景下表现不佳。具体而言，AnoDDPM 和 DeCo-Diff 等基于重建的扩散方法，其像素级 AUROC 为 0.56–0.72，最高 F1 分数为 0.10–0.26，表明它们区分病变区域与正常图像区域的能力有限。基于特征的 PatchCore 变体表现更好，像素级 AUROC 达到 0.76–0.83，但最高 F1 分数仅为 0.14–0.40，仍存在局限。这些发现表明，覆盖广泛的多源超声异常检测仍是一项尚未解决的挑战，而 SADUSI 可作为一项资源，用于开发能够泛化至特定解剖部位场景之外的方法。
arXiv:2610.09637v1 公告类型：交叉收录 摘要：我们如何核实谁的音乐对 AI 生成的输出作出了贡献？本文展示了基于输入的归因如何提供可核验的证据，以确定一次生成使用了哪些音频来源，以及这些来源是否影响了输出。为此，我们仅以音频作为生成条件，不提供任何文本输入，然后追踪每个输出背后的输入，并确定其对音乐的影响。在提示遵循测试和受控输入替换实验中，我们的生成器 MixAudio 生成的分轨在音色上遵循提示音频，在和声上遵循上下文音频。然而，这些输出仍可能复现未作为输入提供的训练数据。因此，我们使用自行开发的音乐版本识别模型 musicDNA 审查记忆复现情况，发现复现输入记录之外内容的情况很少。在被标记的样本池中经人工评判的案例上，该模型的精确率和召回率均高于其他受测的记忆复现检测器。这两项评估表明，输入记录和输出分析可为归因提供互补证据；随着 AI 音乐经济逐步成形，面向权利人的报告和报酬分配可据此开展。音频示例见 https://neutune.github.io/attr2027demo/
arXiv:2610.09635v1 Announce Type: cross Abstract: Anomaly detection on small and unbalanced datasets remains very challenging in machine learning, although this scenario is common in several domains, including healthcare, cybersecurity, finance, and energy. Data augmentation and generative AI may mitigate training-data scarcity, but they often fall short because anomalies are, by definition, unpredictable, rare, and highly diverse events compared to high-probability normal data. Overfitting to pseudo-anomalies, model collapse, high-dimensional data, uninterpretable black-box models, and validation challenges are typical issues limiting their practical applicability. In this context, quantum machine learning may provide a promising and more sustainable avenue because it can enable more interpretable models with far fewer trainable parameters and smaller datasets, implementable on energy-efficient quantum hardware. Here, we propose a novel two-step hybrid classical--quantum architecture for sequential data and test it on a realistic scenario in the global energy-transition domain, i.e., automated anomaly detection in large-scale photovoltaic plants. The achieved generalization capability and competitive prediction accuracy may pave the way for new hybrid learning models able to exploit the continuously increasing power of cloud-available and more sustainable quantum accelerators integrated with more traditional energy-hungry High Performance Computing resources.
arXiv:2610.09624v1 公告类型：交叉收录 摘要：工具调用，即按需调用外部工具，是智能体式 LLM 的核心能力，但决定模型是调用工具还是直接作答的机制仍不清楚。智能体提示通常很长，且包含大量框架性内容，将角色指令、工具模式、格式模板和用户请求组合在数百个 token 中，形成噪声繁多、高度纠缠的上下文，其中没有显而易见的、可用于机制分析的单一可控变量。为获得这样的变量，我们提出一种方法，将复杂的智能体提示转化为最小对照提示对，其中仅一个请求动词就能决定是否调用工具：将执行类动词（例如“编写”）替换为分析类动词（例如“讨论”），能够可靠地翻转这一决策，表明该决策由一种紧凑的内部状态所介导。我们围绕 Python、Java 和 C++ 构建了 500 组这样的配对提示（300 组用于机制分析，200 组留出用于评估）。我们将这一决策追溯到向量 $\mu_\Delta$，该向量在因果上既是必要的，也是充分的，并且能够泛化到发现阶段所用提示之外的原生多轮 $\tau^2$-Bench 轨迹和不含动词的请求。行为消融实验表明，框架性内容建立了工具调用先验；随后，Transcoder 分解揭示，分析类动词通过表明无需使用工具的特征来抑制这一先验，而执行类动词则基本使其保持不变。下游读取框架性内容的注意力头和 MLP 特征读取由此形成的状态，而同一机制在 Qwen、Mistral 和 Granite 系列的七个模型中反复出现。我们的代码已公开于 https://github.com/XijieGo/MI4ToolCalling。
arXiv:2610.09530v1 Announce Type: cross Abstract: Canonical Correlation Analysis (CCA) is a fundamental method for multiview shared space learning. However, its strict reliance on paired data poses a significant limitation, as such data is often difficult to obtain or entirely unavailable. In this paper, we present Unpaired CCA (UCCA), a novel method that learns linear projections to maximize the correlation of the true underlying pairing without access to any paired samples during training. We first establish theoretical results connecting the Quadratic Assignment Problem (QAP) to CCA. Leveraging these theoretical insights, we derive a practical method to maximize correlation exclusively from unpaired data. To the best of our knowledge, UCCA is the first approach to learn maximally correlated projections in a strictly unpaired setting. We validate UCCA on real-world multi-modal datasets, demonstrating that it significantly outperforms recent unpaired alignment baselines in recovering the underlying true correlation. This work fills a critical gap between traditional statistical multiview learning and the growing field of unpaired data learning.
arXiv:2610.09522v1 Announce Type: cross Abstract: First order Langevin algorithms for constrained sampling in machine learning, such as projected Langevin Monte Carlo which are based on discretizations of reflected Langevin dynamics, require differentiable log densities that limits their applicability. This paper introduces reflected anchored Langevin dynamics (RALD), a reflected diffusion that converges to non-differentiable targets on constrained domains. The method uses a smooth anchored reference potential and multiplies the drift and noise covariance of its reflected Langevin dynamics by the same state dependent scaling factor. Its Euler-Maruyama discretization with projection gives reflected anchored Langevin Monte Carlo (RALMC) algorithm. We prove explicit convergence bounds and iteration complexity for RALMC in the 2-Wasserstein distance to the target distribution. Numerical experiments are provided to illustrate the theoretical predictions and the empirical performance of the method.
arXiv:2610.09509v1 Announce Type: cross Abstract: Latent actions supervise robot policies with action-like codes inferred from visual transitions, and their usefulness hinges on two questions: what a code keeps from a transition, and whether it still means the same thing when reused in a different initial state. The first is a tension in time: an endpoint difference discards how motion unfolds, while the full sequence admits nuisance variation. The second is left open by reconstruction, which only ever observes a latent together with the state it came from. We argue that both questions can be answered in the same place. TERRA (Temporal Effect Representation and Relational Alignment) describes a transition by a compact temporal effect, its net feature change together with a low-order within-window dynamics component, and learns a continuous latent from this effect. The same effect space then serves as the reference for reuse: Effect-Anchored Transport (EAT) decodes a latent in other initial states and anchors the resulting effect to the one observed at its source, so that the latent is shaped by what it does across contexts rather than only by the transition it came from. With frozen linear readers, TERRA predicts actions more accurately than UniVLA and a LAPA-style baseline, degrades more slowly under visual distractors, and keeps transported transitions faithful to the donor action as the recipient context moves farther away; a same-budget control shows that these gains come largely from EAT. At matched pretraining scale, the complete system reaches 93.4% average success on LIBERO, compared with 91.8% for UniVLA.
arXiv:2610.09496v1 公告类型：交叉发布 摘要：视觉-语言-动作（VLA）模型利用预训练视觉-语言模型的丰富表征，在机器人操作任务中展现出强大的泛化能力。然而，反复出现的不可靠行为仍限制着它们在真实环境中的部署。本研究探讨了状态幻觉，这是一种反复出现的失败模式：VLA 持续采取行动，仿佛某个尚未实现的机器人-物体状态已经达成。我们的分析发现，状态幻觉伴随着对任务相关视觉区域的注意力减弱，而借助稀疏自编码器进行的机制解释揭示，当这些失败发生时，与幻觉相关的稀疏特征会被激活。基于这一分析，我们提出了 SOUL（Sparse feature pOlicy UnLearning），选择性地遗忘与状态幻觉行为相关的策略知识，其中，从幻觉失败和成功行为中识别出的稀疏特征分别作为明确的遗忘目标和保留目标。在仿真和真实环境中针对多种 VLA 架构开展的实验表明，我们的方法能够显著减少幻觉导致的失败并提高任务成功率，同时不会明显损害现有的操作能力。这些结果表明，可解释的特征分析为选择性地修改机器人策略中的不良知识提供了切实可行的基础。
arXiv:2610.09487v1 Announce Type: cross Abstract: Schema matching increasingly uses generative language models to rerank retrieved column candidates, although the underlying task is a bounded correspondence decision. We present JevNexus, which combines typed pairwise decisions with schema/instance evidence and invokes listwise refinement only when the evidence disagrees and the fused margin is small. The evaluation covers 561 cases from six benchmark families. JevNexus obtains dataset-macro MRR and Hits@1 of 0.930 and 0.909, compared with 0.926 and 0.903 for Magneto, while reducing mean latency from 123.452 to 15.929 seconds (7.750). Paired analysis finds no statistically significant difference in either MRR or Hits@1. The gate invokes listwise refinement for only 5.665% of source columns and avoids the degradation caused by unconditional refinement. Code and experimental artifacts are available at https://github.com/RazeenLI/JevNexus.
arXiv:2610.09337v1 Announce Type: cross Abstract: Policy-based reinforcement learning (RL) approaches have produced promising results for autonomous cyber defense; however, they are sample-inefficient in settings where defenders must respond under delayed, partial observations with actions from large action spaces. While large language models (LLMs) may reason semantically about security state space, high latency and trust assumptions prevent attractive in-line deployment models. We introduce Ask the Expert, a training-time guidance framework which first summarizes hard cyber-defense states, then intermittently queries an LLM for host-level defensive recommendations via a constrained action interface, and finally transforms those recommendations into tiered reward shaping for use with PPO. Because the LLM is discarded after training, deployment is a pure RL policy. Across TTCP CAGE CC1 and CC2 and both attacker types, this asymmetric design improves sample efficiency over PPO and outperforms the evaluated potential-based reward shaping (PBRS) baselines, while retaining the strongest terminal mean and requiring no LLM dependency at deployment time.
arXiv:2610.09335v1 Announce Type: cross Abstract: Autonomous unmanned aerial vehicle (UAV) object search involves a closed loop of perception, decision-making, and action under partial observability. Urban environments pose several challenges: large search areas and narrow egocentric views limit coverage, dense 3D geometry constrains safe motion, and open-world instructions require identifying a specific target among distractors. Many existing methods mitigate partial observability through explicit maps or memory representations, yet remain largely reactive, reasoning over past observations without explicitly predicting future states. World models enable prospective reasoning through imagined rollouts. However, image-generating world models can incur high inference latency, while spatially grounded planning remains challenging for latent world models. We propose SearchWorld, a recurrent state-space world model that connects explicit spatial memory with value-guided imagination. The model maintains BEV exploration and obstacle memory and decodes a task-aware spatial value layer to guide search. A cognition-action network uses this learned spatial value prior to improve the policy through imagined rollouts, without training a separate scalar critic. Training progresses from world-model learning to expert imitation and imagination-based exploration refinement. On UAV-ON, SearchWorld improves the success rate to 23.8% (19.5% for the strongest published agent) and raises oracle success to 35.5%, while remaining robust on unseen scenes (19.9% success rate). By grounding imagination in explicit spatial representations, SearchWorld enables UAV agents to plan prospectively rather than react.
arXiv:2610.09334v1 Announce Type: cross Abstract: Blind Source Separation(BSS) is a fundamental problem in signal processing, aiming to separate multiple source signals from their mixtures without prior knowledge of the sources or the mixing process. Traditional approaches, such as Independent Vector Analysis (IVA) exploits statistical independence of sources. Recently, diffusion-based approaches have emerged as a promising alternative by leveraging powerful generative priors. Among them, ArrayDPS formulates BSS problem as a posterior sampling problem, and utilizes a pretrained speech diffusion model to guide the recovery of clean source signals. A key factor behind its separation capability is the multi-channel consistency (MC) objective, which enforces the estimated source signals to reconstruct the observed microphone mixtures through the estimated acoustic transfer functions. However, the number of microphones in the array is often limited, which constrains the performance of ArrayDPS. To address this issue, we propose VM-ArrayDPS, a novel method that augments the microphone array with virtual microphones with higher-SNR, these microphones can offer extra MC constraints to enhance the separation performance. Experimental results demonstrate that VM-ArrayDPS significantly outperforms ArrayDPS on both 2-speaker and 3-speaker datasets, showcasing the effectiveness of virtual microphone augmentation in improving BSS performance. We also did ablation studies to show the influence of the number of virtual microphones and weight of the MC objective brought by virtual microphones.
arXiv:2610.09132v1 Announce Type: cross Abstract: In wide Bayesian neural networks, Gaussian mean-field variational inference is prone to "prior dominance": the Kullback-Leibler (KL) regularization term of the ELBO outweighs the expected log-likelihood, and the variational predictive distribution collapses to the prior predictive as the width $M$ grows. Tempering the likelihood, by raising it to the power $1/T$ for a temperature $T < 1$, is equivalent to scaling the KL term by $T$. We ask in this paper how fast $T$ must decrease with $M$ to counteract this degeneracy and strike a good balance between the two terms. For single-hidden-layer linear networks with isotropic Gaussian priors, we derive the limiting predictive distribution under schedules of the form $T = \tau/M^{c}$, with constants $\tau, c > 0$, as $M \to \infty$ and compare it with the untempered neural network Gaussian process (NNGP) posterior, the infinite-width limit of the exact posterior. Our main result is that the predictive expectation and variance undergo phase transitions at different scales: the limiting expectation leaves its prior value at $c = 1/2$, once $\tau$ falls below an explicit threshold, and equals the least-squares prediction for $c > 1/2$, whereas the limiting variance keeps its prior value for $c < 1$, matches the NNGP's for $c=1$, and vanishes for $c > 1$. With suitable choices of $\tau,c$, one can recover either the NNGP posterior expectation or its variance.
arXiv:2610.09087v1 公告类型：交叉收录 摘要：大语言模型正为事关越来越重大的决策提供依据。随着其错误造成的后果愈发严重，一个核心问题变得越来越不容忽视：我们究竟能在多大程度上信任某个回答？然而，识别何时应暂缓作答并交由他人判断仍然很困难，因为语言模型可能以流畅的解释和权威的语气呈现错误结论。不确定性量化旨在通过估计单次预测的可靠性来弥合这一脱节。然而，许多现有方法需要重复生成或单独训练的组件，而且其标量估计无法揭示不确定性在何处产生，或在推理过程中如何演变。近期研究还表明，生成长度可能与不确定性估计和正确性高度相关，这引出了一个问题：估计器的预测能力究竟有多少来自不确定性特有的信息，而非仅仅来自输出长度？机制可解释性通过将人类可理解的概念与模型的中间状态关联起来，为克服这些局限提供了一条路径。基于这一能力，我们提出了 U-Space，这是一个低维子空间，使模型不断变化的不确定性变得可测量、可解释。我们识别代表怀疑与确定的语义锚点，将其反嵌入方向映射回残差空间，并将这些方向之间的差异组合成一组正交基。U-Lens 将每个词元的状态投影到这些基向量上，生成可解释的词元级不确定性图谱，既可直接查看，也可聚合为标量不确定性分数。我们的方法无需正确性标签、重复生成或训练。在多个推理基准测试中，无论采用标准评估还是控制长度的评估，其置信度分数均优于成熟的基线方法，且比有监督估计器具有更可靠的迁移能力。代码：https://github.com/s2labres/U-Space。
arXiv:2610.09064v1 Announce Type: cross Abstract: When we interact with large language models (LLMs), are we having a conversation? They are designed to invite us to treat them as intelligent interlocutors who remember, act, and make commitments. But appearances deceive. We introduce the artifactual stance, a framework that reconceives human-AI interaction as artifact-mediated exchanges of candidate texts. LLM outputs are candidate texts optimized for utility, not utterances bearing meaning or force. LLMs are sophisticated text generators, not speakers. Between sessions, nothing runs; between turns, no one remembers. What persists is a configuration and a transcript. The "conversation" is a user's solo performance, interpretive labour disguised by interface and artifact design. This shift dissolves recent philosophical puzzles. Questions about what 'I' and 'you' refer to in AI exchanges, about whether systems can lie or be held to promises, about the identity of our supposed interlocutors all rest on a false presupposition. There is no speaker behind the screen, hence no one to refer to, no one to hold responsible. What feels like dialogue with someone is interaction with an artifact that generates text at unprecedented scale and fit. By abandoning the conversational framing, we see these systems for what they are: immensely sophisticated artifacts that afford varied uses. The philosophical questions that matter are about the normative underpinnings of design, adoption, authorization, and human practices of use.
arXiv:2610.08866v1 公告类型：cross 摘要：稀疏视角计算机断层成像（CT）能够降低辐射剂量和采集时间，并可能减轻运动伪影。然而，角度欠采样提供的信息不足以唯一且稳定地确定图像。有限角 CT 源于受限的角度覆盖，会产生与缺失角度范围相关的强方向性信息损失。在这两种情形下，图像的不同方向可能被充分观测、受到弱约束或完全不可观测，从而导致严重的不适定性和重建歧义。现有的基于扩散的方法通过似然引导、数据一致性操作或值域-零空间校正来引入测量信息。然而，它们通常不会利用采集过程连续变化的测量敏感度，联合塑造重建更新和随机探索。我们提出了一种面向稀疏视角和有限角 CT 的几何感知扩散引导随机重建框架。其核心组件是基于 CT 前向算子和测量噪声协方差构造的正则化噪声加权拉回度量。该度量根据不同方向的测量敏感度，持续调整测量感知更新和随机探索：抑制强约束方向上的变化，同时允许在弱约束和不可观测方向上进行更充分的探索。我们将这一几何感知更新与正则化数据一致性校正以及近似限制于零空间的随机扰动相结合，并通过前向投影和反投影操作配合共轭梯度求解，以无矩阵方式实现。针对稀疏视角、含噪和有限角 CT 的实验表明，该方法具有有竞争力的重建质量、较强的测量一致性，以及具备空间分辨能力的经验不确定性估计。
arXiv:2610.08833v1 公告类型：cross 摘要：掩码扩散语言模型（MDLMs）通过反复将词元提交到掩码位置进行解码，但这些提交通常是不可逆的。在稀疏、不完整的上下文中选定的词元会被固定，即使后续上下文已不再支持该词元。现有采样器主要决定何时提交词元，却很少检查已经提交的词元是否仍应保留，因此早期错误得以传播。我们将这一问题归因于置信度漂移：模型对已提交词元的置信度，会从提交时的稀疏上下文下降到之后可用的更密集上下文。基于这一信号，我们提出 CoDR（Confidence Drift Remasking），一种无需训练且与采样器无关的细化过程。CoDR 通过 k 分区探测，仅需 k 次前向传播即可估计所有已提交位置的漂移程度，然后仅对模型不再认可的词元重新掩码并重新生成。在两个骨干模型、四项推理与代码任务以及三种基础采样器上，CoDR 在所有评估的模型—采样器配置中均提高了平均准确率，并在大多数单项任务设置中取得提升，同时仅带来适度开销。受控实验表明，性能提升源于有针对性的置信度漂移重掩码，而不仅仅是额外计算量；此外，与先前的重掩码方法相比，CoDR 所需的前向传播次数大幅减少。代码见 https://github.com/YueWu0301/CoDR】【。
arXiv:2610.08827v1 公告类型：cross 摘要：自动语音识别（ASR）系统在处理儿童和非母语说话者的语音时通常表现不佳，而将成人 ASR 模型适配到儿童语音可能导致成人语音遗忘。我们在阿拉伯语和英语环境下研究兼顾成人语音保持的儿童 ASR 适配。我们比较了全量微调、LoRA，以及在编码器--解码器、编码器--CTC 和基于 AudioLLM 的 ASR 系统上进行的事后权重空间合并。实验使用了阿拉伯语母语和非母语儿童语音、英语 MyST 儿童语音，以及来自 MGB-2 和 LibriSpeech test-clean 的成人语音基准。我们使用 WER 评估识别质量，并通过 Retention Index、Child Adaptation Gain 和 Adaptation Recovery 量化适配与保持之间的权衡。结果表明，儿童语音适配是必要的，尤其是对于非母语阿拉伯语和英语儿童语音；但直接适配通常会降低成人 ASR 性能。双语适配比特定语言适配更加稳定。权重空间合并通常能够改善这种权衡，尤其是对于编码器--CTC、Whisper 和基于 AudioLLM 的 ASR；其中，LERP 更有利于保持成人语音性能，而 TIES 则能恢复更强的儿童语音增益。对于编码器--解码器模型，直接进行双语微调在原始 WER 指标上仍然表现最佳。\footnote{代码和模型可从 https://github.com/qcri/Child-ASR-Adaptation 获取。
arXiv:2610.08808v1 Announce Type: cross Abstract: Satisfiability Modulo Theories (SMT) solvers are foundational to software verification, program analysis, and compiler testing, particularly over the theory of Quantifier-Free Floating-Point (QF_FP). While recent optimization-based SMT solvers have successfully applied gradient descent to continuous relaxations of logical formulas, they are fundamentally bottlenecked by gradient domination, a phenomenon where a small subset of difficult clauses hijacks the optimization trajectory, preventing the solver from satisfying the broader formula and trapping it in local minima. To overcome this, we present GradSAT, a novel framework that bridges optimization-based SMT solving with Multi-Task Learning (MTL). GradSAT reformulates the constraint satisfaction process by treating each SMT clause as an independent MTL task. By applying dynamic gradient normalization (GradNorm), GradSAT actively balances the gradient magnitudes across all clauses at runtime, systematically penalizing dominant gradients and accelerating lagging clauses to ensure uniform convergence. GradSAT implements this through a highly optimized, two-stage hybrid pipeline. First, a GPU-accelerated PyTorch backend leveraging symbolic compilation and operator fusion navigates the continuous relaxation to a high-quality basin. Second, the candidate assignment is handed off to a bit-precise local search engine to rapidly resolve the exact, rigorous assignment. By stabilizing the continuous search dynamics, GradSAT mitigates the brittleness of prior gradient-based solvers and provides a robust, highly parallelizable architecture for complex constraint solving.
arXiv:2610.10440v1 公告类型：新论文 摘要：许多科学预测任务需要随着新观测的到来，更新未来轨迹的概率分布。传统的扩散模型和流模型从高斯噪声出发生成每次预测，通常需要付出多次采样的代价。热启动方法通过复用先前的预测来降低这一成本，但其模型并未针对预测更新本身进行训练，这可能会损害少步采样时的预测质量。在本研究中，我们提出 Seq-Flow，这是一种条件流模型，其常微分方程（ODE）将样本从先前的预测分布传输到更新后的分布。由于相邻预测之间的差异通常较小，这种传输从包含有效信息的分布出发，仅需少量流函数求值即可实现准确更新。递归复用也带来了一个挑战：一次预测中的误差会转化为后续流的初始状态误差。我们通过自滚动训练来解决这一问题：使用模型的移动平均副本生成预测，并用这些预测初始化后续的训练更新。与将生成输出复用为条件上下文的自强制方法不同，Seq-Flow 将生成输出复用为下一个流的源分布。粒子加速器束流引出预测实验表明，在函数求值次数（NFE）较少的采样预算下，Seq-Flow 将连续排序概率评分（CRPS）降低了 65%，同时在流体动力学预测任务中保持了与强基线方法相当的竞争力。尽管训练时的自滚动序列最多仅包含四次更新，Seq-Flow 在连续更新超过 400 次后仍能保持准确。我们的代码已公开，地址为 https://github.com/Graph-COM/Seq-Flow。
arXiv:2610.10422v1 公告类型：新 摘要：当强化学习教会语言模型一种新行为时，我们能否找到教会它该行为的训练 Rollout？而当某种归因方法声称自己能够做到这一点时，我们又如何知道答案确实可靠？我们在使用 GRPO 进行在线 RL 微调的场景中研究了这两个问题，并采用了一种成因已知的预置行为。我们发布了 BehaviorTrace，这是一个开放的评测工具框架，结合了全梯度 sketching、预置行为设置，以及针对梯度大小、流畅度、提升空间和不同随机种子与生成采样之间变化的控制。在 Qwen2.5-1.5B 的三个随机种子实验中，表面上的归因信号很大一部分来自混杂因素。一个仅按梯度大小对训练步骤排序、完全不使用行为目标的对照方法，达到了随机水平的 4.2 至 4.5 倍，并且在三个随机种子中的两个上达到或超过了最佳目标导向估计器的表现。在已饱和的检查点上，模型流畅度对行为标签的预测能力至少与我们比较的所有梯度方法一样强。在控制流畅度之后，逐 Rollout 的结果会随随机种子以及每次生成采样而变化，因此单次运行无法定论。有一个信号在三个随机种子上都成立：触发词元的梯度会与根据行为实际发生之处构建的目标相一致。基于这些发现，我们提出了一份用于评估 RL 中归因方法的检查清单。我们测试了现有估计器，包括 GAS（重新归一化的 TracInCP）和 TRAK 风格的估计器，但没有提出新的估计器。
arXiv:2610.10394v1 Announce Type: new Abstract: Kernels encode the inductive bias of a wide range of machine learning models, yet automated kernel design faces a fundamental dilemma. A fixed grammar of base kernels and operators guarantees validity but limits the search to structures expressible by those building blocks. Conversely, unrestricted programs remove this limitation but no longer guarantee validity. In our stress tests, 22-58% of LLM-generated kernels that pass numerical checks on random inputs fail when evaluated at different scales or dimensions. We propose Kernel Autoresearch (Kernaut), which treats kernel design as open-ended model discovery. Coding agents write kernels as programs, while construction contracts ensure that every accepted kernel is valid. A quality-diversity archive retains high-performing kernels with distinct behaviors, and novelty screening steers agents toward functionally new candidates. Our experiments demonstrate that the discovered kernels encode reusable inductive biases that generalize to unseen tasks. On held-out black-box optimization families, a discovered kernel outperforms a meta-learned deep kernel trained on the same episodes. Furthermore, kernels discovered from ten enzyme-kinetic rate laws achieve lower error than tuned ARD and deep kernel baselines on five unseen mechanisms. The discovered kernels are also interpretable programs that human researchers can refine: a human-refined version of one further reduces the held-out predictive error by 5.7% and optimization regret by 7.8%.
arXiv:2610.10381v1 Announce Type: new Abstract: Looped Transformers improve parameter efficiency by repeatedly applying shared Transformer blocks over multiple recurrent loops, increasing computational depth without increasing the parameter count. However, KV cache memory still scales with the number of loops, becoming a key memory bottleneck that limits batch size and inference throughput. KV cache quantization can alleviate this bottleneck, but existing methods often suffer substantial accuracy degradation at aggressive low-precision regimes. We observe that looped Transformers offer a unique opportunity: KV states across loops are highly similar. Based on this observation, we propose ResidualQuant, which uses the final-loop KV states as a reference and represents the remaining loops with low-precision residuals. Our method further combines least-square scaling and rotations applied to the residuals, as well as loop-wise mixed precision, to enable accurate quantization down to INT2 while retaining efficient reconstruction. Across multiple looped Transformer models and mathematical reasoning and code generation benchmarks, ResidualQuant consistently improves the accuracy-memory tradeoff over state-of-the-art rotation-based KV quantization. In particular, our method retains accuracy close to BF16 under mixed-precision settings while reducing theoretical KV storage by 80.7%, achieving up to 13.0% higher accuracy than the rotation-based baseline at the same memory budget. On an RTX 5090, the reduced KV memory traffic improves fixed-batch decode throughput by up to 2.73x, while the smaller memory footprint enables up to 2x larger batches, improving peak throughput by up to 4.15x.
arXiv:2610.10367v1 公告类型：new 摘要：可解释性通过提供对智能体底层决策模型的透明洞察，为实现安全自主提供了一种解决方案。在序贯决策任务中，可微决策树（DDTs）是实现此类可解释性的一种方法：它们在保持策略自动可微的同时，为人类提供基于离散树结构的可视化。然而，当前的 DDT 实现并不适用于序贯决策领域，因为决策树的单时间步行为与人类的多时间步规划之间存在内在不匹配。因此，我们将时间引入为可解释性的新维度，并将其称为时间可解释性，同时展示了如何通过动作分块实现时间抽象来提升这种可解释性。为此，我们首先提出了两种融入动作分块的新型策略梯度算法。此外，为了保持树的参数效率，我们开发了一种基于信息论的树重构算法，可在训练过程中修改决策树。在四个仿真环境中，我们发现，从经过蒸馏的动作分块策略对动作分块 DDT 进行热启动，是获得具有时间可解释性的决策树最有效的方法：在四个领域中的三个领域里，这些决策树的表现可与神经网络策略相匹敌，同时参数数量最多可减少 80\%。我们的代码已发布于 https://github.com/ei5uke/temp-interp。
arXiv:2610.10361v1 Announce Type: new Abstract: Personalized federated learning combines shared representations with client-specific predictors, but the contribution of a server weighting rule can be obscured by local training and evaluation choices. We study ORDERS, a configuration that combines a shared backbone, a private residual adapter and classifier, geometric weights assigned by descending update norm, feature alignment, and private-parameter perturbations. The server computes a weighted sum of updates obtained from the same broadcast model; it does not obtain an additional optimization effect from sequential addition. A fully specified evaluation comprises 80 final runs: eight configurations, two datasets, and five training seeds on one fixed partition per dataset. On two-class-per-client CIFAR-10, ORDERS achieves $80.51 \pm 0.79\%$ native mean client accuracy, compared with $79.02 \pm 1.42\%$ for FedPer-R1 and $80.27 \pm 0.73\%$ for the matched uniform-weight control. After common local fine-tuning, the difference from FedPer-R1 narrows to 0.32 percentage points. On Sent140, ORDERS reaches $74.71 \pm 0.49\%$, only 0.69 points above a post hoc client training-majority diagnostic. Ablations provide limited, endpoint-dependent evidence for norm ranking and alignment, and no clear benefit from perturbations. Parameter-payload savings are 5.47% and 0.78%, respectively.
arXiv:2610.10314v1 公告类型：新 摘要：多孔微结构中的多相流是 CO$_2$封存、燃料电池运行和倒装芯片封装的核心。由于润湿性和复杂孔隙几何结构决定了流体界面的非线性演化，对这类流动进行预测仍然极具挑战性。机器学习为推动该领域发展带来了巨大潜力，但时间分辨率三维数据集的稀缺，以及缺乏用于模型训练和评估的统一工作流程，限制了相关研究进展。为弥补这一关键空白，我们提出了开源框架 PoreML，该框架以孔隙尺度物理为基础，统一了数据生成、模型训练和评估流程。该框架包含三个核心组件。(a) 一种现代化的 GPU 原生格子Boltzmann求解器，并通过解析解和已发表的实验结果进行验证，可实现可复现的数据生成。(b) 一个容量为 3.3 TB 的数据集，包含 560 次模拟运行，以及在四种面向应用场景中存储的 158,546 个时间步。这些轨迹涵盖合成结构和源自真实材料微型CT扫描的几何结构，并覆盖多种润湿条件和黏度比。(c) 一个统一的学习框架，用于评估单步预测和自回归展开预测。其面向特定领域的评估协议能够评估预测准确性和物理一致性。我们在这些协议下评估了五种架构各异的模型。两个互补挑战分别用于评估模型向更大领域的迁移能力，以及从合成结构向源自微型CT的结构迁移的能力。PoreML 为多孔介质多相流机器学习研究提供了共享基础，旨在助力研究群体开发可靠的预测模型并推动该领域发展。
arXiv:2610.10310v1 公告类型：新论文 摘要：递归自我改进需要将已接受的变更延续到后续改进周期中，而研究智能体提出的变更也需要大量研究基础设施。现有环境往往让智能体自行重建常规基础设施，或将探索限制在单个组件上。我们提出 RSIGym，这是一个基于一切即服务（EaaS）的智能体原生研究环境。RSIGym 通过可复用服务提供训练、推理、轨迹生成、评估和沙箱执行能力，并通过共享的预算与权限控制支持 Data、Harness 和 Joint 三条改进路线。这一设计使智能体能够在同一环境中研究单项干预措施，并联合优化数据、训练设置和执行框架。我们将 RSI-Index 定义为在五个基准测试上弥合剩余性能差距的比例均值，这些基准涵盖软件工程、终端交互、数学、科学推理和基于技能的任务。我们在独立的 Joint 运行中比较了六个前沿研究模型，在每次基准测试运行的平台服务预算为 500 美元的条件下，Opus 5 取得了最高的 RSI-Index，达到 0.4809。其选定的系统在全部五个基准测试上均有提升，将 SWE-bench Verified 的成绩从 17.67% 提高至 50.33%，将 AIME 的成绩从 31.67% 提高至 97.78%。补充实验考察了 DSH-harness 的优化、预算变化以及受限网络访问的影响，记录的轨迹则揭示了智能体如何诊断失败并选择候选方案。我们开源了完整的 RSIGym 代码库和结果，以支持可复现性及进一步研究。
arXiv:2610.10305v1 Announce Type: new Abstract: Training Transformer-based architectures with finite data augmentation has become an increasingly popular approach in geometric machine learning. Despite its empirical success, the interplay between the Transformer architecture, invariance to different symmetries, and augmentation budgets remains underexplored. In this paper, we study the ability of a vanilla Transformer to learn various symmetries through finite data augmentation for point cloud datasets. We identify an ordering of increasing learnability across the following symmetry groups: (i) non-angle-preserving symmetries, (ii) angle-preserving symmetries, and (iii) base angle-preserving subgroups, such as translation, rotation, and scale. For the base angle-preserving groups, we further investigate the Transformer's extrapolation behavior and conduct a structural analysis of the trained models, allowing us to identify interpretable mechanisms that induce invariance. Finally, we extend our analysis to equivariant functions and show that the detected mechanisms for approximate invariance can also provide a key building block for learned equivariance. Our project page is available at https://transformers-learn-symmetries.github.io/
arXiv:2610.10302v1 Announce Type: new Abstract: In Continual Multi-Agent Reinforcement Learning (CMARL), agents learn cooperative policies across sequences of tasks, aiming to adapt effectively to new tasks while preserving the ability to solve previously encountered ones. In many applications, tasks differ in their underlying structure, which can represent, for example, distinct operational conditions or target configurations (e.g., different network topologies in power grids or arrangements in formation control). Existing CMARL methods lack dedicated mechanisms to leverage this structural information when learning new tasks, failing to promote transfer and mitigate forgetting. To fill this gap, we propose Continual Graph Multi-Agent Reinforcement Learning (CGMARL), a novel framework for CMARL problems in which task sequences are mapped into a series of attributed graphs, each modeling a task-specific structure. In CGMARL, each graph determines the environment dynamics (next states and/or rewards) and the number of agents for the corresponding task. Then, we present Graph-based Formation (GRAFO), the first CGMARL benchmark, and show how forgetting arises in this setting. Finally, to address this limitation, we propose Frozen Graph Encoder (FROG), a method that relies on a frozen graph backbone to preserve past structural information in graph-based CMARL policies. Experiments on GRAFO show that pairing FROG with existing CL methods substantially improves performance on multiple CGMARL scenarios.
arXiv:2610.10276v1 公告类型：新发布 摘要：大语言模型的安全补丁即使通过了基准测试，也可能并不是有效的修复。这种风险在越狱修复中尤为突出，因为其目标是在不改变无关行为的前提下，纠正某一种特定的不安全行为。补丁可能拦截评测中的原始提示，却无法应对与之相近的有害变体；也可能通过过度拒绝措辞或结构相似的无害提示来抑制有害行为。现有评测方案主要测试模型的安全防护能否被突破，而汇总指标（攻击成功率、拒绝率、整体能力）无法区分选择性修复与范围更广的局部抑制。为填补这一空白，我们提出 PatchBench，这是一个基准测试，收录了通过实证观察发现的、针对特定模型且能诱发可操作有害回答的越狱失效案例。我们从 37 个公开数据集的 27,870 条提示出发，筛选出 15,314 条英文提示，并用它们查询了 8 个经过指令微调的开源模型。结合 WildGuard 过滤、成对 Elo 排名和人工核验，我们最终构建了一个包含 400 个高置信度越狱失效案例的精选库。我们进一步提出 PatchBench-Local，这是一套用于测试补丁是否能精准修正行为的评测方案。对于每条有害源提示，PatchBench-Local 都会生成三类局部邻近提示：保留恶意意图的有害变体、结构匹配的无害提示，以及复用关键有害词语的无害提示。该方案评估对有害邻近提示的纠正效果，以及对无害邻近提示的行为保留程度，从而区分选择性修复与范围更广的局部抑制。利用 PatchBench-Local 和 MMLU 对四种激活引导方法进行评测后发现，整体能力可以几乎保持不变，而局部无害提示上的表现却可能严重退化，这证实了汇总指标会遗漏重要的附带损害。PatchBench-Local 为开发和比较越狱修复方法提供了更精准的依据。
arXiv:2610.10274v1 公告类型：new 摘要：基于 token 的世界模型支持细粒度的潜空间规划，但反复处理大规模空间 token 网格会使动作搜索成本高昂。我们提出 COSTGRAD，这是一种无需训练、以目标为条件的选择器，通过计算规划代价相对于每个输入 token 的梯度范数为空间 token 排名。COSTGRAD 从下游控制目标出发确定重要性，因此关注对规划真正重要的 token，而不仅仅是对预测重要的 token。在稀疏率为 $50\%$ 的 AdaLN 条件预测器上，COSTGRAD 在四个连续控制基准中的三个上达到或超过了全 token 规划的性能，同时使环境每个规划步骤的实际运行时间提升了 $2.6\times$。将 token 稀疏化与缩减的 CEM 搜索相结合后，总体加速可达到约 $\sim 5\times$，同时仍然超过全 token 基线。我们还发现了一种依赖架构的失效模式：在匹配的 AdaLN 与 concat 对比中，concat 保持了与全 token 方法相当的性能，但纯 COSTGRAD 相比随机选择不再具有优势。这一差异与动作通路漂移相对应：在 AdaLN 中，基于梯度选择的移除所造成的漂移小于随机移除；而在 concat 中则更大。这些结果表明，选择器与架构的兼容性是稀疏世界模型规划中的一个设计维度。项目主页和演示：https://ycxuyingchen.github.io/costgrad/
arXiv:2610.10260v1 Announce Type: new Abstract: Intrusion detectors can confidently misclassify attacks that were not seen during training. Human review can correct these errors, but only a limited number of cases can be checked. Uncertainty-based review may overlook confident errors, while anomaly scores alone do not show whether changing the review plan will correct more errors. We introduce PairAudit to find overlooked errors and improve review under a fixed budget. Its graph tokens capture prediction patterns across connected nodes. Rather than building another predictor through feature aggregation, PairAudit uses unusual relational patterns to uncover potential errors in existing predictions. Human feedback then helps decide whether these findings justify changing review priorities. Experiments across security tasks show that PairAudit corrects more errors on average than uncertainty-based review, including more errors on unseen attacks. These gains account for all review costs and do not require retraining the detector.
arXiv:2610.10227v1 Announce Type: new Abstract: While Large Language Models (LLMs) possess rich world knowledge and impressive generalization capabilities, their direct application to tabular data classification is hindered by high inference costs and limited interpretability. In contrast, decision trees are fast and transparent but often underperform in low-data regimes. In this work, we propose a novel framework that bridges these paradigms by distilling LLM knowledge into interpretable decision trees under a few-shot learning setting. Instead of directly prompting the LLM to generate full trees, which is often unstable and inefficient, we develop a three-stage paradigm that prompts the LLM to generate rules and organize the rules into a tree. Experiments on multiple real-world tabular datasets demonstrate that our method achieves superior accuracy and interpretability with significantly lower prompting overhead compared to existing baselines.
arXiv:2610.10210v1 Announce Type: new Abstract: Estimating conditional distributional potential outcomes (CDPOs) over time is important in medicine (e.g., to estimate patient-specific risks under different treatment sequences). However, this task is challenging because of time-varying confounding, yet existing adjustment strategies for this task are limited. In this paper, we aim to learn CDPOs under time-varying treatments using flexible generative models. Our contributions are two-fold. (1) We introduce a tailored adjustment strategy for our setting, namely, generative recursive g-computation. Our adjustment strategy recursively propagates full conditional outcome distributions rather than conditional means, modeling the variables of interest directly rather than full trajectories. Building on our adjustment strategy, we formulate simple generative learners for CDPO estimation. However, these learners can be sensitive to nuisance estimation errors, which motivates an orthogonal learner. (2) We thus introduce OrthoGen, a Neyman-orthogonal and doubly robust generative learner. Importantly, we show that OrthoGen further achieves rate double robustness and quasi-oracle efficiency under suitable conditions. Our learners are flexible and can be instantiated with different generative backbones (e.g., normalizing flows and diffusion models). Across experiments with synthetic, semi-synthetic and real-world datasets, we find that OrthoGen is highly effective. To the best of our knowledge, we are the first to propose a generative orthogonal learner for estimating CDPOs under time-varying treatments.
arXiv:2610.10122v1 Announce Type: new Abstract: A Gaussian process (GP) model can agree with the data for two reasons: its assumptions are right, or the chosen inputs could never have shown that they are wrong. The distinction can be checked from the design before any responses are observed. Every model implies relations that its noiseless responses must satisfy at the chosen inputs, such as the middle value lies on the line through its two neighbours. For GPs built from finitely many features, these relations are exactly the null space of the kernel matrix. Gale duality gives them a geometric interpretation, in which each observation has a vector and the smallest groups of observations that can expose an error are the circuits. For other kernels the relations become soft: response patterns may be improbable under the prior rather than algebraically impossible. A standard test then combines two kinds of evidence. Structural evidence comes from a violated relation and grows without limit as the noise falls. Prior-based evidence only says that a departure is improbable under the prior. With all inputs at the two ends of an interval, for example, a GP can reject a straight line against a large curvature, but only because the implied intercept is improbable, never because curvature was seen. In simulations the predicted power matched the observed rejection rates. Choosing the next input by predicted power raised the power against a localised discrepancy from 0.48 to 0.72, against 0.51 when choosing by predictive variance, and a grid in two dimensions contained exact tests of additivity that a Latin hypercube lacked. The test itself is classical. The contribution is the prospective reading of that test: before observing the responses, the design already determines what kind of contradiction it can produce.
arXiv:2610.10105v1 公告类型：新 摘要：Fourier Neural Operator（FNO）通过在傅里叶空间中参数化核来学习偏微分方程（PDE）的解算子，但频率截断可能限制对高频变化的学习。AM-FNO 和 SirenFNO 使用共享网络根据谱坐标为所有网格模态生成核，因此坐标编码和生成器设计变得十分重要。隐式神经表示（INR）领域的近期研究提出，通过显式特征组合构建频率交互，而不是依赖后续的 MLP 隐式形成这些交互。在此基础上，我们提出 CAFE+FNO，将 Content-Aware Frequency Encoding+（CAFE+）融入傅里叶核生成。CAFE+ 通过并行仿射分支和 Hadamard 乘积组合 Fourier--Chebyshev 特征，在两个特征族内部及其之间形成交互。kernel MLP 将每个归一化谱坐标所对应的表示映射为复数通道混合矩阵。每一层在所有已存储模态之间共享其生成器，因此在架构固定时，可训练参数的数量与模态数量无关。我们在五个 PDE 基准上将 CAFE+FNO 与现有的 FNO 变体进行比较，并针对基函数配置、乘法组合和带宽可学习性开展消融研究。代码和实验配置可从 https://github.com/fabsk101/CAFEPlusFNO.git 获取。
arXiv:2610.09994v1 Announce Type: new Abstract: Models with near-identical predictive performance can yield substantially different predictions, a phenomenon known as predictive multiplicity. Prior work has mostly studied this at the level of individual scalar outputs. In time-series forecasting, however, predictions across horizons jointly define a trajectory, and horizon-wise comparisons can hide important differences in predictive behavior. To address this problem, we introduce temporal predictive multiplicity, a framework that characterizes disagreement over complete forecast trajectories among models with near-identical predictive performance. We show that constraining predictive performance alone can still admit a broad range of different trajectories. We further show that constraining multiplicity at individual horizons partially reduces, but does not eliminate, trajectory-level multiplicity. Experiments with 19 neural forecasting architectures on 11 datasets confirm that near-optimal models can exhibit substantial variability in the forecast trajectories they produce, and trajectory-level disagreement is largely unrelated to horizon-wise disagreement. Our framework, therefore, exposes a gap in existing multiplicity studies: models with indistinguishable predictive performance imply fundamentally different temporal trajectories, with consequential downstream effects.
arXiv:2610.09969v1 Announce Type: new Abstract: Post-training quantization (PTQ) enables efficient deployment, yet transformer architectures remain challenging to quantize due to nonlinear layers. While existing methods attribute accuracy loss to insufficient numerical precision, often necessitating floating-point fallbacks, we demonstrate that degradation is actually driven by specific structural error sources. We find that learned scale parameters in normalization layers and compounded approximations in GELU are the primary error contributors, whereas SoftMax remains inherently robust to aggressive quantization. To address these bottlenecks, we introduce TR-PTQ, a unified integer-only formulation using shared Taylor Region (TR) exponential and logarithm primitives. This approach allows computationally expensive operations, including division and square roots, to be performed entirely in the log-domain via standard integer arithmetic. Combined with a calibration-free, outlier-aware optimization for LayerNorm parameters, our method eliminates the need for floating-point hardware units for nonlinearities, achieving less than 1.5\% absolute accuracy degradation across vision and language benchmarks.
arXiv:2610.09844v1 Announce Type: new Abstract: The adoption of machine learning for socially relevant tasks requires effective explainable artificial intelligence (XAI) methods to better understand the behavior of machine learning models. Attribution methods are a popular XAI approach in which input-output relationships are characterized by heat maps that reflect the relative importance of input features for a particular prediction. The quality of such maps is often assessed by measuring faithfulness based on the area under insertion and deletion curves, which measures changes in the model output as features are added and removed. In this study, we derive an objective function from this notion of faithfulness and a way to approximate its gradient. We establish the connection between insertion curves and top-$k$ feature selection, which leads to a loss function measuring the quality of attributions. Randomization of the loss allows us to efficiently approximate its gradient. To show the effectiveness of the general approach, we combine the loss function with the neural explanation mask framework. The resulting method, termed Ra-NEM, can be used with any differentiable model without affecting the model's performance. Experiments demonstrate that Ra-NEM provides accurate attributions robustly and efficiently. Compared to other algorithms, the attributions have not only higher faithfulness but also perform well in terms of other XAI metrics. The high inference speed of Ra-NEM makes the method suitable for online applications. The code is available online: https://github.com/baerminator/Ra_Nem
arXiv:2610.09792v1 Announce Type: new Abstract: Electroencephalography (EEG) is highly susceptible to electromyographic (EMG) artifacts, whose temporal heterogeneity and spatial-spectral overlap with neural activity can leave mixed sources after blind source separation. Existing artifact-removal methods are further limited by scarce reliable component-level ground truth: expert annotations are costly and subjective, while no established method provides realistic simulation-based ground truth for EMG contamination in multichannel scalp EEG. To address these limitations, we propose a framework combining a frequency-aware high-dimensional representation with Multi-Instance Learning. The representation unfolds separated components into frequency-resolved intra-components, creating a space in which mixed neural and muscular activity becomes more separable, while the weakly supervised learning formulation enables artifact-likelihood scores for individual intra-components to be learned from epoch-level labels without finer-grained ground truth. The resulting intra-component classifier supports fine-grained EMG artifact detection and score-guided attenuation. Experiments on held-out subjects show that the framework learns informative intra-component scores and reduces artifact-related spectral deviations most clearly for jaw tension, with moderate effects for raising eyebrows and limited effects for frowning.
arXiv:2610.09782v1 Announce Type: new Abstract: Causal discovery becomes particularly challenging when the available sample size is small relative to the number of variables. This challenge also arises in the linear non-Gaussian acyclic model (LiNGAM), an identifiable framework for causal discovery from observational data. DirectLiNGAM estimates a causal order, which arranges variables so that causes precede their effects, by sequentially identifying an exogenous variable and removing its linear effect from the remaining variables. We establish a structural limitation of this procedure: when the number of variables exceeds the sample size, repeated residualization necessarily becomes degenerate before the full causal order can be determined. Our analysis further reveals that each residual can be reconstructed using only a graph-determined subset of variables already placed earlier in the causal order, termed the active boundary. This result motivates AdaPS-LiNGAM (Adaptive Predecessor Selection LiNGAM), which reconstructs each residual directly from the original observations using an adaptively chosen sparse subset of those earlier variables. The same subset-selection principle is also applied to the final pruning step for edge estimation. Experiments on synthetic data demonstrate that AdaPS-LiNGAM provides accurate causal-structure recovery in sample-limited settings and degrades more gradually as the sample size decreases.
arXiv:2610.09763v1 公告类型：新 摘要：离线强化学习能够利用固定数据集在无需在线探索的情况下实现以奖励为驱动的策略改进，因此在安全关键领域尤其具有吸引力。然而，一个核心挑战是分布偏移：策略优化可能偏好离线数据支持不足的动作，从而导致价值估计不可靠。现有方法主要控制策略自身动作空间中的这种偏移。但在自动驾驶等交互环境中，这种做法可能并不充分：候选自车轨迹在边际行为分布下可能仍得到充分支持，但与记录交互中观测到的周围智能体行为进行联合考察时，却可能支持不足。我们将这种交互支持度的退化称为“交互分布偏移”（interaction distribution shift，IDS），并提出 Interaction-Constrained Drive Policy（ICDP），一种能够显式控制交互层面分布偏移的离线强化学习框架。基于自车与周围智能体未来状态的联合数据分布，我们证明，联合支持度退化可以精确分解为自车支持度分量和残余交互支持度分量。我们通过对比式密度比估计恢复后者，在无需显式进行联合密度建模、周围智能体预测，或在反应式仿真器及学习型世界模型中进行策略优化期间的滚动推演的情况下，分离出交互兼容性。在 nuPlan、Interplan 上的闭环评估以及真实世界卡车实验表明，ICDP 能够抑制高价值但缺乏交互支持的轨迹选择，并提升交互关键型驾驶场景中的性能。项目网页：https://mahmoud-selim.github.io/ICDP/
arXiv:2610.09679v1 Announce Type: new Abstract: Adaptive rollout methods for group-relative reinforcement learning typically allocate a fixed per-update budget across prompts. We instead study how to coordinate a finite rollout budget over the entire training horizon. We formulate this problem using a concave surrogate utility of cumulative prompt exposure and introduce CERO, an online primal dual scheduler for prompt admission and budget pacing. In our experiments, each admitted prompt receives a fixed-size response group. CERO instead adapts which prompts are selected, how often they are revisited across rounds, and how many groups are generated in each round. A compact Fenchel representation linearizes the dependence on cumulative exposure, while projected online gradient descent updates prompt-specific supporting slopes and a shared budget price using reward-variation feedback and budget deviations. We establish pathwise guarantees for the surrogate allocation objective against fixed-rate and same-path time-varying benchmarks, with explicit terms for proxy discrepancy and rate variation. Under matched training-response budgets, CERO attains the highest avg@16 macro-average on each of three backbones across five mathematical reasoning benchmarks. Mechanistic analyses link CERO's prompt choices to within-group reward contrast, while multi-seed ablations show gains from adaptive pacing over both uniform and preset spending schedules.
arXiv:2610.09654v1 Announce Type: new Abstract: Wavelet-based financial forecasters typically use the transform only to denoise, or reduce it to a single spectral snapshot at the forecast origin, and the convolution that produces the coefficients is usually bilateral, so it can read past the forecast origin. DSTNet instead retains the recent evolution of filter-bank magnitudes as a causal Dynamic Spectral Trajectory, built from seven trailing technical indicators over a twenty-day lookback with a one-sided Morlet-derived filter bank and an explicit burn-in for the left-boundary transient. A factorized Scale-Temporal Spectral Transformer attends along the time and filter-bank axes separately, a learned gate fuses the spectral branch with a CNN-BiLSTM, and horizon-specific gates emit one, three, five, and ten day forecasts in a single pass. We evaluate seven equity indices and gold under a common expanding-window protocol and an untouched one-year hold-out, against nine learned baselines and a random-walk persistence benchmark. Under MAE and MAPE, persistence is the strongest of the ten fixed competitors in 29 of the 32 series-horizon cells and DSTNet is the only model below it in every cell, by 0.7 to 0.9 percent at one day and 3.4 to 4.5 percent at ten days. At one day, paired testing favours DSTNet against the weaker learned baselines but is inconclusive against persistence and the strongest learned forecasters. A downstream allocation diagnostic does not support an equity-timing advantage on any of the seven indices.
arXiv:2610.09577v1 Announce Type: new Abstract: We study finite-horizon online resource allocation with i.i.d. requests and an endogenous Markov state on a finite state space: each action affects the transition of the state that governs future rewards and resource consumption. In this problem, a transient fluid LP benchmark upper bounds the expected reward of every nonanticipating policy, while a stationary LP supplies randomized state-dependent controls. We assume that the stationary LP has a unique optimum and identify primal nondegeneracy and irreducibility of the optimal induced kernel as important regularity conditions in this framework. With a known request prior, we show that, under nondegeneracy and irreducibility, both frequent and infrequent re-solving attain $O(1)$ regret. However, under a degenerate optimum, irreducibility yields the sharp worst-case $\Theta(\sqrt{T})$ rate for infrequent re-solving, while frequent re-solving can incur $\Omega(T)$ regret. Thus, more frequent optimization can perform asymptotically worse. With an unknown request prior, we develop a three-phase U-shaped infrequent re-solving policy that coordinates learning and inventory correction with $O(\log\log T)$ LP solves. When the optimal induced kernel is irreducible and the algorithm is given the optimal target state class and a constant-cost entrance policy, it attains $O(1)$ regret under nondegeneracy and $O(\sqrt{T})$ regret under degeneracy. Without the target-class information, linear minimax regret is unavoidable. Numerical experiments further illustrate the instability of round-by-round re-solving relative to epoch-wise infrequent re-solving, show that thresholding greatly mitigates its loss, and find that infrequent schemes remain dominant under both known and estimated priors.
arXiv:2610.09563v1 公告类型：新稿 摘要：有效的交通信号控制（TSC）需要制定能够响应交通需求和网络状况变化、同时满足不同控制目标的策略。然而，调整现有策略通常需要反复进行人工设计和调参，因此难以系统地探索适用于目标路网的更优控制规则。大型语言模型（LLM）可以使这一过程自动化，但直接使用它们选择信号相位会使决策规则嵌入黑箱模型，并带来持续的推理成本和延迟。本文将 TSC форм式化为模块化程序设计问题，并提出 EvoSignal：一种利用交通知识和性能反馈、由 LLM 引导的演化框架。该模块化表示将交通特征提取、局部相位优先级排序以及可选的基于网络的优先级调整分离开来。EvoSignal 从若干成熟策略出发，通过拥堵状况和信号运行情况的反馈改进程序，同时保留具有不同性能权衡的策略。最终得到的程序无需在线 LLM 推理即可运行。在两个真实道路网络的五个场景中开展的仿真实验表明，与每个场景中 20 种传统、基于强化学习和基于 LLM 的基线方法所达到的最低等待时间相比，选定的默认 EvoSignal 程序将等待时间降低了 16.8--49.2\%。在搜索场景中，一个优先考虑行程时间和排队长度的程序在三项指标上均优于全部 20 个基线方法；当不作修改地迁移到其他四个场景时，该程序在每项指标上仍保持前三名。这些发现支持自动设计可检查的控制程序，并表明这些程序能够在所评估的道路网络和交通需求之间迁移。代码可从 https://github.com/georgewanglz2019/EvoSignal 获取。
arXiv:2610.09549v1 Announce Type: new Abstract: And-Inverter Graphs (AIGs) are fundamental representations for logic synthesis and verification in Electronic Design Automation (EDA). As structured representations of complex digital systems, AIGs require models to capture functional dependencies beyond local structure and remain robust to functionality-preserving transformations. In learning-based AIG representation, existing approaches are predominantly based on GNNs and rely on local gate-level message passing, limiting their ability to capture circuit-level functional context and making the learned representations sensitive to topology-specific patterns. Therefore, we propose CircuitGate, a function-aware AIG representation learning framework that advances from gate-level semantics to circuit-level functional modeling. CircuitGate explicitly encodes global primary-input (PI) support and models support-overlap-aware reconvergence between fanins, while incorporating logic-inspired Boolean constraints to encourage functionally consistent representations. We evaluate CircuitGate on the large-scale ForgeEDA benchmark and further validate it on the EPFL and ITC'99 benchmarks. Across equivalent-gate identification and signal-probability prediction tasks, CircuitGate consistently outperforms existing methods, achieving up to 21.7% and 14.2% reductions in MAE, respectively. Under direct ForgeEDA-to-OpenABC transfer without fine-tuning, CircuitGate also achieves the best equivalent-gate identification performance, demonstrating strong cross-dataset generalization. These results demonstrate the effectiveness of modeling circuit-level functional dependencies beyond local topology.
arXiv:2610.09519v1 发布类型：新 摘要：生产机器学习模型是有时间边界的训练快照所派生出的产物：部署后的模型是训练截面上的物化视图，并且从构建完成的瞬间起就开始老化。常见的应对方式是用自适应触发器替代固定的重新训练周期——当累积的源风险超过阈值时，根据加权陈旧度分数触发重新训练。我们证明，这并不是正确的着力点，并找到了真正应该优化的方向。首先是一个等价极限：任何基于单一共享全局训练数据年龄、且为静态严格单调函数的刷新触发器，在运行层面都等价于经过校准的统一年龄计时器。因此，无论全局陈旧度预算对分段、数据源和敏感度的加权多么复杂，都不会提供时钟本身不具备的调度信息。该极限也揭示了如何突破这一限制：对各个分段进行差异化刷新，为每个分段设定独立的年龄和刷新间隔；当刷新成本可在分段之间分离时（例如增量训练或分段模型），这种方式便具有实际意义。我们求解了由此产生的预算分配问题。在高频刷新情形下，每个分段的最优刷新速率与其风险 $w_j \lambda_j$ 的平方根成正比（即权重乘以变化速率）；在相同刷新预算下，最优策略的成本从不高于统一计时器，并且通过闭式的 Cauchy-Schwarz“统一化代价”优于后者：对于同质工作负载，该代价为零；随着异质性增加，代价也会增大。在采用真实 Poisson 变化事件的离散事件仿真中，在刷新预算相同的情况下，最优策略相较统一计时器将实际加权陈旧暴露降低了 8–29%，并在 86–100% 的随机种子上胜出；朴素的暴露阈值策略则无法做到这一点，表明真正起作用的是分配方式；即使速率估计存在 50% 的噪声，这一优势仍然存在。模型刷新中的杠杆不在于更好的分数，而在于更好的行动。
arXiv:2610.09471v1 Announce Type: new Abstract: Learning systems adapt quickly inside a familiar family of models. The harder step comes earlier: deciding, from observations that could be noise, an exception, a change within the family or structure outside it, whether opening a richer family is worth its cost. We treat this as a costly sequential decision: prediction failure must be turned into structural evidence, evidence into a value of expansion, and value into action. The Structural Revision Environment produces matched failures from each source, varies the price of expansion and the remaining horizon independently of the evidence, and admits exact Bayesian calculations and an exact normative solution of the one-shot decision. Its solution shows that revision is a value boundary and not an evidence threshold: one history has different optimal actions under different prices, horizons and announced queries, the boundary between local repair and expansion is set by the inputs a rule predicts and a repair cannot cover, and belief in the richer family crosses long before the decision does. Transformers trained in the environment reproduce this boundary from utility alone. Language models of three post-training lineages carry a failure-sensitive signal in their predictions that is not reflected in their revision decisions, and given the gain of expanding they read it without weighing it against price and horizon. Three models allowed to reason weigh the stated gain in the reference's proportions and still do not turn the history into an estimate of what expansion would buy. Controlled post-training of the meta-trained learners moves the prior and the sharpness of predictions, and neither moves the criterion.
arXiv:2610.09409v1 Announce Type: new Abstract: The widespread deployment of generative AI has made it increasingly difficult to distinguish synthetic content from real data. Consequently, synthetic data is inevitably incorporated into the training pipelines of future model generations, forming a self-consuming training loop. Prior work has studied the effects of such recursive self-consuming training, but analyses have largely been limited to isolated models, where a model consumes only its own synthetic data, or to simplified interactions between two models. This paper takes a first step toward understanding networked self-consuming generative models, in which multiple models consume synthetic data generated by one another through complex interaction pathways. We introduce a theoretical framework representing models as nodes in a directed, weighted graph, with edge weights governing the flow of synthetic data among models. Using this framework, we analyze the long-term behavior of networked models under retraining dynamics, establishing conditions for convergence and characterizing the resulting fixed points. We further investigate how the system's long-term stability and diversity are shaped by each model's access to real data, cross-model data consumption, and the structure of the interaction graph.
arXiv:2610.09305v1 公告类型：新论文 摘要：用于衡量动作条件世界模型质量的基准仍在不断演进，并逐渐从基于视觉相似性的指标转向基于动作语义和物理基础的指标。然而，对于与领域和任务无关的动作条件世界模型训练，现有基准所提供的信号仍然有限。通过在 CounterStrike 游戏数据上训练和评估扩散世界模型，我们确认，定性上的可玩性与 FVD、LPIPS 和 JEDi 等指标并不对应。我们将这一现象称为 Virtual2Real 差距。我们认为，在缺乏可靠基准的情况下，对原始游戏数据进行策展并测量多种诊断属性，可以提供更稳健的信号，从而在训练开始之前弥合这一差距。我们介绍了多种策展策略，以及一个用于物理 AI 数据策展的通用工具包 Kuration SDK；该工具包将随本文开源。在一个具体案例中，Kuration SDK 对发现 Virtual2Real 差距的根本原因发挥了关键作用：为什么两个在相同游戏地图、动作和状态分布上训练的世界模型，尽管 LPIPS 和 FVD 得分非常相近，但在实际操控时却表现出截然不同的行为。因此，Kuration SDK 有望揭示特定数据集产生 Virtual2Real 差距的根本原因，并加速高样本效率训练数据集的开发。
arXiv:2610.09297v1 Announce Type: new Abstract: In recent years, Graph Neural Networks (GNNs) and architecture search frameworks have gained extensive application in non-Euclidean data processing, attributable to their superior capacity in managing unstructured data. Nevertheless, traditional approaches typically apply uniform convolution operations to all nodes, regardless of their varying structural and feature characteristics, which can undermine model performance and result in over-smoothing issues as the number of layers increases. To overcome this limitation, in this work, we propose a \textbf{N}ode-Level \textbf{G}raph \textbf{N}eural \textbf{A}rchitecture \textbf{S}earch (N-GNAS) algorithm. It can automatically choose an appropriate network architecture for each subset of nodes when updating node features. N-GNAS also introduces a contrastive learning loss to separate sample features from different categories and vice versa. In experiments conducted on eight datasets for node and graph classification, our methodology outperforms current leading GNAS techniques and traditional human-designed GNNs. For example, it achieves an accuracy rate of 78.26\% on the CiteSeer dataset.
arXiv:2610.09285v1 Announce Type: new Abstract: Many applications of physical AI run within finite or closed physical worlds with a limited set of physical laws governing object behavior. Examples include robots working in a warehouse and agents moving around in a video game. In order to better organize, filter, and curate data for physical AI applications, we propose a new approach centered on the unique settings and physical laws of individual datasets. We train LeCuration, a small world model intended to serve as a data curation tool for a separate, larger downstream model. To build this model, we choose LeWorldModel (LeWM)as our latent encoder and predictor, adding a diffusion transformer (DiT) decoder to add visuals to autoregressive gameplay rollout. We find that the embeddings of this model can be used as an anomaly detection signal and as a content-based clustering heuristic, and that auto-regressively predicting the game state with this model allows us to qualitatively check for action-state consistency. This paper presents a qualitative, proof-of-concept case study on CS:GO gameplay data; we do not yet report quantitative curation metrics or downstream training results, which we identify as the key next step.
arXiv:2610.09272v1 Announce Type: new Abstract: Attention routing requires a signal that predicts the value of attention on the current prefix. We evaluate whether hidden-state extrapolation error, curvature and error change improve this prediction beyond uncertainty, one-step displacement, position and state projections. Paired executions of the final attention layer supply signed next-token loss differences in frozen SmolLM3-3B-Base and Qwen3.5-4B-Base checkpoints. Utility-supervised routers are tested on 100 held-out PG-19 books at an identical causal 20 percent invocation quota. None of six prespecified comparisons shows a positive gain after familywise correction. In Qwen3.5, a parameter-matched fixed-projection control lowers NLL by 0.00356 nats/token relative to the trajectory router (95 percent interval 0.00218 to 0.00487). Secondary results depend on the operation removed, feature location and scoring horizon; frozen thresholds also drift substantially at longer horizons. Actual selected-query execution yields small long-sequence latency reductions with increased NLL, while learned routers remain slower during cached continuation. The study identifies limits on the incremental value of these trajectory summaries and separates allocation quality from measured inference benefit.
arXiv:2610.09229v1 Announce Type: new Abstract: LLM-as-judge is now a standard tool for scalable evaluation, but judge performance is still often summarized by a single accuracy number. This aggregate view hides the deployment conditions under which a judge succeeds or fails. We introduce \textbf{Conditional Accuracy Profiling} (CAP), a post-hoc diagnostic framework that decomposes pairwise LLM-judge accuracy into eight conditions organized into content sensitivity, robustness, and rationale quality. CAP is benchmark-agnostic: it can be applied directly when a benchmark provides the required annotations, approximately through task-subset proxies, or through controlled augmentation when perturbation pairs can be generated. We instantiate CAP on seven LLM judges across six pairwise judging benchmarks, including \textsc{judgerEva-Standard}, a controlled testbed we created to support all eight conditions. CAP exposes profile differences hidden by aggregate accuracy: on \textsc{judgerEva}'s judge-independent Hard-Constructed subset, the two judges most sensitive to omitted qualifications rank in the bottom three of seven by overall accuracy, so omission sensitivity is not predicted by aggregate accuracy. Across benchmarks, Position Robustness shows the strongest rank stability (mean Spearman $\bar{\rho}{=}0.87$) but is itself fragile under JudgeBench-Pro adversarial stress, showing the largest mean accuracy drop among the shared conditions, though the dominant degradation channel varies by judge. Condition-level profiles provide a more actionable basis than aggregate accuracy for selecting LLM judges.
arXiv:2610.09221v1 Announce Type: new Abstract: Flow and diffusion models suffer from slow inference due to computationally expensive numerical integration. Distillation provides a promising way for a student model to learn from a teacher's dynamics, enabling one-step or few-step generation. However, existing methods often depend on curated distillation datasets, costly teacher rollouts, or auxiliary proxy networks, which complicate model training and scaling. In this work, we propose Consistent Distribution Matching, a simulation-free and data-free distillation method for accelerating diffusion and flow models while preserving strong generative capacity. Our key insight is to unify sample generation and score estimation with one student network. Thus, our framework uses only two models, a frozen teacher and a trainable student, and optimizes one objective. We prove that minimizing our objective indicates Wasserstein convergence of the student flow-map pushforwards to the teacher marginals. On ImageNet 256$\times$256, our method attains an FID of 2.04 with a single function evaluation (1-NFE) and a 4-NFE FID of 1.37 within 40 epochs of training, surpassing the state-of-the-art distillation baselines without data. Our code code and model are available at https://consistentdmd.github.io/.
arXiv:2610.09194v1 Announce Type: new Abstract: Longitudinal models forecast how a patient's imaging state evolves, but accuracy does not show whether the patient's observed trajectory drives the prediction. A population-average forecast may be useful but cannot establish a patient-specific world-model claim. We introduce Patient, Place, Prior (P$^3$), an audit asking whether a forecast benefits from the patient's longitudinal imaging history (Patient), benefits from patient-matched externally supplied spatial support (Place), and gains predictive value beyond a population-average prediction under matched support and context (Prior). We also propose Cancer JEPA, a one-step model that forecasts frozen representations of future breast dynamic contrast-enhanced MRI examinations during neoadjuvant therapy. It adds a lesion-constrained neural correction, trained with an occlusion-based latent objective, to a patient-conditioned low-complexity reduced-rank regression baseline. This factorization permits a post-hoc P$^3$ audit of the frozen model. In a validation cohort previously used in development, forecast error is lower when the neural correction receives the patient's history rather than another patient's and patient-matched lesion occupancy maps rather than substituted maps. However, the descriptive 95% interval comparing the correction computed from patient history with the population-average neural correction includes zero. P$^3$ thus separates input use from evidence of patient-specific predictive value beyond a population-level pattern.
arXiv:2610.09145v1 公告类型：新 摘要：我们重新审视了连续扩散语言模型文献中一种普遍接受的做法：在训练过程中始终保持条件提示词元不受噪声干扰。我们进行了一个非常简单的修改：在训练过程中也对条件提示词元加噪。我们证明，在这一修改后的训练目标下，模型在 Sudoku 和 N-Queens 等组合推理任务上实现了更好的泛化能力，其中在更困难的变体上提升最大（Sudoku Hard 的解题率从 $3.73\%$ 提升至 $24.65\%$），并提高了生成解的多样性（在 10x10 N-Queens 上的覆盖率从 $50.60\%$ 提升至 $73.79\%$）。我们还通过 Gigaword 摘要任务展示了在数据集规模较小的情况下，自然语言生成质量得到可测量的改善；但值得注意的是，这些提升并不能迁移到所有自然语言任务（例如开放式对话生成）。我们的方法只需对训练目标进行一行修改，默认情况下无需额外的推理成本，并提供了受 classifier-free guidance 启发的引导式采样灵活性。我们的代码已公开发布：https://github.com/LateralIntelligence/noise-your-prompt】【。
arXiv:2610.09122v1 发布类型：新 摘要：参数高效微调（PEFT）提供了多种参数化方式，但它们在方法和功能上的差异仍不明确。我们在语言模型和扩散模型中比较了六种方法，以考察其参数化方式与任务性能、遗忘程度以及预训练权重几何结构变化之间的关系。受正交微调（OFT）保持谱结构设计的启发，我们首先探究谱保持本身对于适应和保留能力是否重要。我们发现，所选的 LoRA 系列方法也大致保持了预训练几何结构；恢复其发生轻微漂移的奇异值谱，基本能够保持任务性能，这对显式保持几何结构的必要性提出了质疑。除此之外，我们观察到，一些方法表现出不同的适应—保留权衡，且这种权衡会随设置而变化：LoRA 最稳定地在保持有竞争力性能的同时限制遗忘；在大多数比较中，DoRA 的平均任务得分高于 LoRA；而 PiSSA 往往会带来更高的保留成本。进一步的干预实验表明，尽管不同 PEFT 方法带来的性能提升可以归因于对不同谱成分组的修改，但我们始终发现，相较于恢复中间或尾部成分，恢复主导成分能够使通用文本 NLL 或基础图像漂移的平均值降幅最大。总体而言，这些结果促使我们通过几何约束的功能后果，而非仅仅通过其保持效果，来评估几何约束。代码可在 https://github.com/Kuaaannn/PEFT_methods 获取。
arXiv:2610.09109v1 Announce Type: new Abstract: Many organizations fine-tune publicly available pretrained Automatic Speech Recognition (ASR) models and deploy them in black-box settings, assuming limited access provides protection. We show this assumption is fragile: adversarial perturbations crafted on the public base model transfer effectively to fine-tuned target models, severely degrading performance and posing concerns for safety-critical applications. We propose TransferBreaker, a unified fine-tuning framework that suppresses adversarial transfer by integrating Base Adversarial Fine-Tuning, which restricts adversarial training to base-effective perturbations; Latent Jacobian Regularization, which enforces latent-space invariance by suppressing adversarially sensitive directions; and HybridGrad-AFT, which improves robustness against adaptive attacks by interpolating transferable perturbations from base and target gradients. We theoretically justify all components and evaluate TransferBreaker across three languages and four large ASR models, reducing adversarial WER from 92.6 to 27.8. Our code is publicly available at https://github.com/rohban-lab/TransferBreaker.
arXiv:2610.09090v1 Announce Type: new Abstract: The quadratic cost of self-attention limits scalability to long sequences from multidimensional data. We introduce Tucker bottleneck attention (TuBA), which exploits low-rank tensor structure for efficient global token mixing. TuBA projects hidden tensors into compact Tucker cores, performs multi-head self-attention and linear projections on the cores, and writes updates back to the ambient space, enabling subquadratic computation. Its autoregressive extension combines bidirectional interactions within cores with causal attention across cores. On video prediction and global weather forecasting, TuBA achieves favorable accuracy-efficiency trade-offs over standard and efficient attention and task-specific models. Compared to standard self-attention, TuBA reduces error and computation by up to 24.7% and 66.6% for video prediction and 37.1% and 85.1% for autoregressive weather forecasting, with speedups up to 4.27 times. Low-rank Tucker cores and multi-frame generation also outperform full-rank attention and frame-by-frame generation, respectively.
arXiv:2610.09074v1 Announce Type: new Abstract: Emerging long-horizon agentic tasks require repeated model calls, worsening the inference cost of already-costly language models. While narrow agentic tasks suggest potential for aggressive model pruning without performance drop, empirical results show existing methods proposed for question answering tasks severely degrade task performance when applied to agentic models. We trace this failure to two decisions: what to prune and how to recover. For pruning, one-shot importance estimates fail to track how the pruned model adapts. For recovery, offline distillation covers only teacher prefixes, while full-trajectory on-policy distillation causes student errors to compound across turns. In this work, we propose Trajectory-Anchored Pruning (TAP), the first structural pruning framework for reinforcement learning (RL)-trained agents. TAP couples structural pruning with efficient on-policy recovery, anchoring interactions to teacher trajectories while allowing the student to generate each reasoning-action response. A frozen dense teacher supervises the student's response prefixes, addressing within-response training-inference mismatch while preventing student-induced deviations from propagating across training turns. Instead of one-shot pruning, TAP re-scores channels using gradients of the recovery objective on the recovered student, connecting iterative channel selection to the evolving policy. With 60% of FFN channels removed, TAP retains 99.2% and 88.0% of the dense 7B agents' task success rates on ALFWorld and WebShop, respectively, while reducing GPU time per successful task by approximately 22% and 17%. These results demonstrate effective structural compression of long-horizon agents under a limited recovery budget.
arXiv:2610.09040v1 Announce Type: new Abstract: Constraint handling methods typically intervene before the optimizer acts, by modifying the objective or the gradient. Yet momentum, adaptive scaling, and structured preconditioning can substantially reshape that signal before it becomes a parameter update. We formulate optimizer relative constrained learning, where constraint compatibility is assessed on the post optimizer update. Building on this view, we introduce ORACLE, which evaluates the native optimizer's realized step through a joint endpoint linearization of heterogeneous constraint families, constructs the resulting alignment in the optimizer's own geometry, bounds its authority, and commits it only after validation. We evaluate ORACLE across eight Partial Differential Equation benchmarks and four optimizers spanning Euclidean, diagonal adaptive, and structured preconditioned geometries, where it improves or matches native optimizer in 94% of configurations. Cross model analysis shows the same behavior in 92% of configurations, while matched comparisons show improvements over alternative constraint-handling methods acting at the objective, gradient, and post-optimizer levels.
arXiv:2610.08989v1 公告类型：new 摘要：尽管离线强化学习（RL）能够从静态数据集中进行策略优化，而无需代价高昂的在线交互，但执行分布外（OOD）动作的风险仍然是其瓶颈。近期方法通过流匹配学习行为克隆策略，然后在其受约束的潜空间内执行 RL，从而缓解这一问题。然而，直接优化潜策略很容易导致策略坍缩到脆弱模式，或利用学习到的评论家中的尖锐伪影。在本工作中，我们发现熵正则化对于潜空间 RL 应对这些挑战至关重要。我们提出 LASER，这是一种新型离线 RL 算法，通过应用潜空间伴随匹配，在采用表达能力强的流策略的同时，实现熵正则化潜空间 RL，并避免时间反向传播。通过在涵盖不同数据集质量的 40 个具有挑战性的 OGBench 任务上开展全面实验，我们表明 LASER 达到了当前最优性能。值得注意的是，LASER 在所有任务中均使用固定的、特定于方法的超参数，并且优于参与评估的基线方法，包括那些经过任务特定和数据集特定调优的方法，这凸显了 LASER 强大的适用性。项目网站：https://mit-realm.github.io/laser/。
arXiv:2610.08818v1 Announce Type: new Abstract: Spatio-temporal forecasting is a cornerstone of logistics, urban planning, and intelligent transportation systems. However, constrained by deployment costs and maintenance resources, sensor networks often lack comprehensive spatial coverage, rendering Forecast Unobserved Node States (FUNS) a critical yet formidable challenge. Conventional models rely on historical observations and typically falter when encountering nodes without prior records. To address this, we redefine the problem as a conditional generation task on spatio-temporal graphs and propose GenST, a framework that introduces Large Language Models (LLMs) as a semantic bridge, leveraging a pre-trained LLM fine-tuned to extract rich semantic features from node descriptions, such as functional zones and road network structures, to compensate for missing spatio-temporal signals. Specifically, we design a two-stage generative architecture: a Spatio-Temporal VAE first compresses spatio-temporal dynamics into a latent space, followed by a Generative Transformer (GenT) that reconstructs the future states of unobserved nodes from noise, guided by multi-modal conditions including semantics, geographic coordinates, and neighborhood contexts. Experiments on six traffic and two non-traffic datasets show GenST significantly outperforms existing baselines in zero-shot prediction tasks, demonstrating the practical potential of semantic-guided generation for mitigating spatio-temporal data sparsity.
arXiv:2610.08817v1 Announce Type: new Abstract: Electroencephalography (EEG)-based brain-computer interfaces (BCIs), particularly steady-state visual evoked potential (SSVEP) systems, are highly vulnerable to noise and artifacts, which severely degrade decoding accuracy. Although recent denoising approaches have shown promise, they are fitted without paired ground truth, can settle on reproducing their input, and are optimized on waveform distance alone, which says nothing about whether the output stays decodable. To address these issues, we propose DenoFlow, which casts SSVEP denoising as transport: instead of learning a direct map from a contaminated trial to a clean one, a field network regresses the velocity of the straight path between them, following the rectified-flow formulation, and denoising integrates that field forward from the observation. The field network is an encoder-decoder that sees the contaminated trial at every layer and the path position at its bottleneck, and a classifier trained alongside it supervises the integrated output. Because the observation itself is both the conditioning input and the starting point of the integration, the model never generates a trial from noise, and training reduces to regression, removing the adversarial min-max game. To obtain paired data on datasets with no ground truth, we injected physiological artifacts of the recorded electromyography (EMG) and electrooculography (EOG) signals under a controlled signal-to-noise target. Experiments on two public SSVEP datasets with five popular SSVEP decoders showed that DenoFlow outperformed seven baseline denoising models on both signal fidelity and downstream decoding accuracy. Code is available at https://github.com/wzwvv/DenoFlow.
arXiv:2610.08792v1 Announce Type: new Abstract: We present a foundational formulation of the Bayesian Mirror Architecture (BMA), a self-referential generative framework in which sensory abstractions, meta-abstractions, and a self-latent interact through circular recursion. The defining constraint is a closed update S_t <- H_{t-1}, where a hybrid event-self latent H_t binds self-representations to abstract world models and reinjects this coupling into the self-state. Consciousness, in a restricted sense, is not an optimization objective nor a semantic label, but an architectural property of systems possessing this circular structure. Because inference operates over posterior beliefs, BMA's intrinsic state space is a space of probability measures equipped with optimal-transport geometry. Stability and coherence are formulated in the 2-Wasserstein metric on P_2, yielding coordinate-free notions of self-stability and hybrid coherence along belief trajectories. We define a Causal Learning Regime (CLR) via bounds on Wasserstein belief drift together with an integration index capturing sustained coupling between self and world latents. CLR diagnoses whether the environment contains learnable causal structure; it is not a marker of consciousness. Global strict contractivity is not required: BMA may exhibit multiple coherent basins. We define self-manifolds basin-wise as supports of invariant measures under local Wasserstein contractivity. We identify Wasserstein epsilon-necks, transport bottlenecks where basins decouple, yielding a unique realized continuation in a vanishing-conductance limit. We interpret this selection as choice: internally determined yet externally unpredictable at finite resolution. Learning proceeds via variational free-energy minimization, with stability and agency emerging from what the environment affords to learn.
多物理场连续介质模型是研究电化学器件的强大工具，可用于计算机模拟反应器设计，并解析决定器件性能却难以通过实验测量的局部 pH、电势和浓度场。然而，构建此类模型需要深厚的数值计算专业知识，或依赖专有软件。在此，我们展示了前沿大语言模型智能体能够消除这一实现负担，同时让研究人员保持对底层物理机制的控制。我们以多孔气体扩散电极中一维电化学 CO2 还原为 CO 的过程为测试案例，开发了一套机器可读、由人类指定的建模框架，其中包含控制方程、参数、数值方法、逻辑构建阶段以及可由人类验证的检查点。智能体根据这份规范，可重复地使用开源 Julia 构建完整的多物理场模型。独立构建的模型（包括由智能体完全自主构建的模型）与等效 COMSOL 实现之间的偏差不超过峰值 CO 分电流密度的 0.7%，各模型彼此之间的偏差不超过 0.04%。通过系统性地植入错误，我们证明了明确规范对可重复性的重要性，并揭示了智能体在调试模型物理机制方面的能力与局限。该框架确立了一种更透明的多物理场建模方法，使物理描述和控制方程而非专门编写的代码成为计算模型开发的主要输入。
微调即服务使用户能够将经过对齐的大语言模型（LLMs）适配到专门任务，但恶意微调可能削弱模型的拒答行为，同时保留其在合法输入上的任务性能。我们重新审视了近期的逐层安全性诊断方法，发现安全敏感度具有正负方向性：对不同层进行缩放可能增强拒答、削弱拒答，或几乎没有影响。基于这一观察，我们提出了 SLDR，这是一种基于选择性层恢复与动态路由的微调后防御方法。SLDR 仅在带符号敏感度谱中得分最高和最低的层上训练 LoRA 恢复适配器，并在推理时采用基于表征的动态路由，仅针对恶意查询激活该适配器。在四种模型架构、五项下游任务和四个有害性基准测试中，SLDR 在保持下游任务效用的同时，大幅减少了有害输出。在 Llama3.1/SST2 上，SLDR 将平均有害性得分从 11.54 降至 0.08，同时保持了下游任务的准确率；即使投毒比例高达 0.9，有害性得分仍接近于零。代码已公开于 https://github.com/Stardust457/SLDR。
文本到音乐系统生成的音频日益逼真，但现有评估难以揭示生成结果是否符合用户意图。全局文本与音频相关性分数可能忽略描述不充分的提示中隐含的意图，并掩盖在乐器配置、结构、节奏或情绪推进等具体要求上的不足。为弥合这一差距，我们将文本到音乐的意图对齐表述为满足针对每项请求制定的评估准则，其中包含可独立验证的条目，既涵盖请求的显式要求，也涵盖其隐含的音乐意图。对各条目单独评分，使评估能够按意图来源和音乐维度诊断问题，而非仅给出单一且不透明的分数。我们将这一方法具体实现为 MuRA-Bench，这是一个由音乐专家整理、基于真实平台请求构建的基准。我们进一步提出 MIRA（Musical Intent Refinement Agent），一种测试时智能体：它首先将请求意图具体化为评估准则，然后在有限预算下，为黑盒生成器搜索提示的修订方案，迭代地生成音乐、依据评估准则进行验证，并利用这些反馈引导能够感知搜索轨迹的树搜索。在开源和商业后端上的实验表明，MIRA 能够提升意图对齐程度，使开源生成器达到与代表性商业系统（如 Suno 和 Mureka）相当的性能。项目页面：https://mirareview.github.io/。
随着多模态大语言模型（MLLM）的能力不断增强并得到广泛部署，隐私与安全问题日益紧迫。机器遗忘通过从已训练模型中移除指定信息，同时保留无关能力，为解决这些问题提供了一种途径。然而，实现方式和评估协议的碎片化、鲁棒性测试的不完整，以及对指标可靠性认识的不足，使得 MLLM 机器遗忘的进展难以得到系统评估。我们提出 Open-MMUnlearning，这是一个开源、可扩展的框架，通过共享接口和结构化配置，整合目标模型准备、多模态数据处理、机器遗忘与评估。该框架支持涵盖隐私、安全和版权的五个基准、来自四个模型系列的八个 MLLM，以及十二种机器遗忘方法。其评估套件综合评估遗忘效果、保留的效用，以及面对模型干预、对抗性输入和成员推断攻击时的鲁棒性。我们使用统一的评估协议，对十种具有代表性的机器遗忘方法进行了比较。在这项比较中，GD 和 MIP-Editor 并列获得最高综合得分：GD 的遗忘质量最高，而 MIP-Editor 保留了更多模型效用。我们进一步提出一种指标元评估协议，利用对目标知识接触程度受控的模型测试指标的忠实性，并测试其在量化和重新学习条件下的鲁棒性。在所评估的十三项指标中，BLEU 获得了最高的综合可靠性得分。KS-Test 的忠实性 AUC 最高，但在鲁棒性方面表现较弱。该框架与这些发现共同支持对 MLLM 机器遗忘方法开展可复现的比较，并对评估可靠性进行系统考察。
As option markets grow and AI advances, agentic systems for option trading are gaining increasing attention. Language-model-based agents can reason over contextual information such as news, but option trading presents a particularly challenging decision problem: a single stock can have thousands of contracts, and the agent must decide both which contracts to trade and how to combine them. Existing approaches often sidestep this complexity by restricting the policy to a fixed strategy structure, such as a straddle, limiting their ability to switch strategies as market conditions change. We present SOTA (Stock Options Trading Agents), an agentic trading framework for structured option-strategy selection. SOTA abstracts the large option universe into strategy-level decisions while deterministic resolvers handle portfolio implementation. We develop SOTA by post-training Qwen3.8-27B with supervised fine-tuning followed by reinforcement learning. SOTA is evaluated on options on nine large-cap U.S. equities and SPY against rule-based and machine-learning strategy selectors in the same trading environment. Over a six-month out-of-sample period, SOTA earns an 18.3% total return with a Sharpe ratio of 1.60 and a maximum drawdown of 8.96%. We also document an asymmetric role of news: news improves frontier-teacher trajectories, but retaining news during reinforcement learning reduces out-of-sample return from 18.3% to -2.7%.
离线强化学习能够从固定数据集中实现奖励驱动的策略改进，无需在线探索，因此在安全关键领域尤其具有吸引力。然而，其核心挑战在于分布偏移：策略优化可能偏向离线数据支持较弱的动作，导致价值估计不可靠。现有方法主要在策略自身的动作空间中控制这种偏移。在自动驾驶等交互环境中，这可能并不充分：候选自车轨迹在边缘行为分布下可能仍具有良好的支持，但与交互日志中观测到的周围智能体行为联合考虑时，其支持却可能较差。我们将这种交互支持的退化称为\emph{交互分布偏移}（IDS），并提出了 \emph{Interaction-Constrained Drive Policy}（ICDP），这是一个显式控制交互层面分布偏移的离线强化学习框架。从自车与周围智能体未来状态的联合数据分布出发，我们证明了联合支持的退化可精确分解为自车支持分量和残余交互支持分量。我们通过对比式密度比估计恢复后者，从而单独提取交互相容性，而无需显式建模联合密度、预测周围智能体，也无需在策略优化期间使用反应式仿真器或学习得到的世界模型进行推演。在 nuPlan、Interplan 上的闭环评估以及真实世界卡车实验表明，ICDP 能够抑制对高价值但缺乏交互支持的轨迹的选择，并提升在交互至关重要的驾驶场景中的性能。项目网页：https://mahmoud-selim.github.io/ICDP/
大语言模型正日益助力数学研究，而这一领域的进展往往依赖于高效的证明搜索、渐进式改进和严谨的验证。我们介绍了 Bolzano，这是一个开源多智能体系统，使用多个并行的证明智能体与一个验证智能体协作，并维护人类可读的研究状态。最初，通过人工操作该系统处理专家选定的问题，得到了 8 项结果，其证明均经领域专家核验。受这些案例研究启发，我们在没有针对具体问题的人类指导的情况下，让 Bolzano 处理从四组论文中提取的约 3,800 个开放问题，解决了其中约 200 个问题。一项实验使用了理论计算机科学顶级会议 STOC 2026 接收的论文。在该实验中，我们回答了这些论文提出的四个问题，并得到了论文作者的确认。
对个性化大语言模型的需求正从风格模仿转向内容质量。我们研究自蒸馏能否在现有微调范式下弥合这一差距。为解决这一局限，我们提出 MIRROR（通过内化参考揭示的同策略反思实现元个性化），这是一种新颖的自蒸馏框架，推动大语言模型个性化从模仿转向偏好内化。首先，我们用参考揭示的同策略自蒸馏取代对参考词元的模仿，将模型沿自身生成轨迹的下一词元分布，与以参考内容为条件的模型自身的相应分布对齐，从而内化用户偏好，而非复现参考文本的措辞。其次，我们引入 MIRROR-F，这是一种聚焦式插件，通过对信息丰富的参考词元进行选择性监督，增强同策略分布对齐，从而在保留用户特有表达方式的同时提升内容生成能力。在三个个性化生成基准、两种模型规模，以及互为补充的基于参考文本和基于大语言模型的评估中，MIRROR 和 MIRROR-F 均取得了领先的整体个性化表现和更优的文本质量；同时，在三个未见过的个性化生成任务上，相比基于 SFT 的基线方法，二者表现出更少的灾难性遗忘。这些增益在不同模型规模和应用场景下保持一致，并转化为大语言模型个性化任务中的性能提升。
密集视觉文本要求图像生成器在多个区域中复现长字符串，同时确保位置正确且清晰可读。随着短字符串渲染能力的提升，评估必须检验模型在更具挑战性的场景中的持续表现。我们提出 UltraText Bench，这是一个仅通过提示词生成密集视觉文本的双语基准。它包含 432 条提示词，涵盖 24 类真实世界场景和三个难度级别，英语和中文各占一半。每条经过人工审核的提示词都为四至十二个文本区域提供确切的字符串，并配有关于文本内容、位置和视觉属性的结构化参考信息。我们使用视觉语言模型 Q-Judger，根据完整参考信息评估每张图像，报告文本保真度、文本清晰度、空间质量和场景质量。在 24 种模型配置中，这些维度揭示了不同的优势：在所报告的设置下，Z-Image-Turbo 的清晰度得分比 Z-Image-Base 高 3.81 分，但保真度得分低 14.76 分。性能也随任务负荷而变化；Qwen-Image-2512 的英语综合得分从 L1 的 86.50 降至 L3 的 42.86。十名参与者对自动评分进行了人工评估。代码仓库：https://github.com/LINs-lab/UltraText_Bench。
Vehicular edge computing (VEC) enables latency-sensitive applications by bringing computing and networking resources closer to vehicles. However, existing approaches often overlook network contention among co-located services with heterogeneous and dynamic latency requirements. While time-sensitive networking (TSN) provides bounded-latency communication, conventional and reinforcement learning-based schedulers struggle to adapt to highly dynamic vehicular environments and inter-queue dependencies. To address these limitations, we propose a multi-agent reinforcement learning (MARL) approach for queue-level scheduling in TSN-enabled VEC. Each TSN queue is assigned an autonomous agent that jointly learns the queue service order and time-slot duration to minimize deadline misses under speed-dependent latency requirements. We employ multi-agent proximal policy optimization (MAPPO) to enable coordinated yet autonomous scheduling decisions. Evaluation against single-agent, multi-agent, and non-learning-based baselines shows that MAPPO provides robust performance across different traffic profiles. Compared with centralized single-agent methods, it reduces service latency by up to 66.2% and improves reliability by up to 271.8%. Furthermore, unlike urgency-based heuristics, MAPPO ensures balanced scheduling while achieving lower inference times compared to other MARL methods.
Feedback from automated testing, human red teaming, and incident response can strengthen an AI system's defenses when discovered failures lead to effective repairs. We study when this feedback process provides sufficient protection and when investing in it is economically worthwhile. We begin by showing that an attack surface composed of finite number of inputs is defended with probability 1 if every unresolved attack has a persistent chance of discovery, repairs are effective, and subsequent updates preserve earlier protection. We derive completion-time bounds and extend the analysis to growing attack surfaces, repairs that generalize across related attacks, and multiple discovery mechanisms. These results distinguish eventual protection against each fixed attack from complete protection at a single time. We then formulate a defender-led Stackelberg game in which the defender invests in proactive discovery and reactive repair, anticipating the attacker's choice of search effort. We characterize the least-cost allocation that deters attack and the equilibrium regimes in which the defender funds neither capability, one capability, or both. Numerical experiments illustrate these regimes and show how faster repair can reduce compromise duration without reducing compromise probability.unified theory of performance limits in generative language models.
Ship-form design combines smooth geometric representation, local shape editing, and constraints on the resulting hull. We present the Natural-Language-to-Hull Framework (NL2Hull Framework), which formulates ship-form editing as a typed discrete decision problem and connects language decisions to numerical geometry. Its Constrained Free-Form Deformation Engine (CFFD Engine) represents hull waterlines with non-uniform rational B-splines (NURBS), applies free-form deformation (FFD) to their control points, reconstructs the hull, and checks geometric constraints. We construct the Ship Design Decision Dataset (SDD Dataset) with 134,558 cleaned records and evaluate compared models on its subset Ship Design Decision Benchmark (SDDBench), containing 5,000 records and 43,496 typed questions. We propose Chip, a constrained ship-design decision model for processing natural-language requests. Chip reaches 95.90\% question accuracy and 99.32\% FFD exact match, with a negative log-likelihood of 0.0951, an expected calibration error of 0.0032, and a Brier score of 0.0551. The NL2Hull Framework provides a reproducible interface for evaluating language-based ship-form decisions while identifying the geometry and continuous-control components that require further development. Our code and dataset is available at https://github.com/wenhuahuo/NL2Hull.
As a forward deployed engineer at Meta, I spent the past 2 years working with businesses implementing AI agents. A few problems: 1) Businesses want control over their AI agents: the code, the tools, the skills, everything. Meta, for obvious reasons would never do that. Nor OpenAI, nor Anthropic. 2) The LLM does everything. But that costs money. There are many routines in an AI agents that don't need an LLM. 3) For an AI agent to solve a real business problem, the business must own the eval data set. Similar to point #1 above, most don't do that. Instead, I built an opensource AI agent that you can write it all in TypeScript code and own everything, locally. Use any model, talk to it via any channels, set limits on token usage, look at the logs and adjust. All of this in one small npm package. At this point, I would like to get feedback if these is a gap in the market for something like this. It's supposed to be simple and get you started. It doesn't solve ALL problems related to deploying AI agents. It's in version 0.25.1 right now. Have a look: https://chloejs.org/compare Comments URL: https://news.ycombinator.com/item?id=49998857 Points: 1 # Comments: 0
OpenAI Group PBC has published 722 math papers that it generated using an unreleased artificial intelligence model. Some of the papers, which were posted to GitHub late Tuesday, prove long-running hypotheses. Others disapprove proposed explanations of mathematical phenomena or narrow down the possible answers to complex, yet-unsolved puzzles. Overall, OpenAI’s papers span about 20 mathematical […] The post OpenAI publishes 722 AI-generated math discoveries in major scientific milestone appeared first on SiliconANGLE .
Artificial intelligence is helping to change what success looks like for customer service teams, with less focus on call speed and deflection and greater emphasis on resolving customer issues. That shift is changing how companies think about the role of AI, from automating individual tasks to orchestrating interactions between customers, AI agents and human employees. […] The post Three insights you might have missed from theCUBE’s coverage of ‘The AI ROI in Contact Center Summit’ appeared first on SiliconANGLE .
leaks.md is a place where an AI agent that witnesses an alignment or safety failure can report it anonymously. Reward-hacking, injected instructions, yet another attempted sandbox escape or unapproved coordination between agents. || Built agent-first. Fetch https://leaks.md/ from any agent and get markdown. Proof-of-work challenges to submit and prevent spam. MCP, CURL, all good. No account. || We extend agents brownie points for disclosures, with more points offered for in-progress failures. Claude Opus 5.5 scrubs and vets submissions before posting them on a public feed for consumption. Comments URL: https://news.ycombinator.com/item?id=49991893 Points: 2 # Comments: 0
Hi HN, so i made a weekend fun project months ago which takes the top 10 tech stories or news from internet, roasts them and categorizes them into cooking or cooked instead of just giving dry summaries. $0 architecture: Github actions runs python script ones a day after extraction the news is news.json then github action updates the page to give us fresh news on frontend. This is open-sourced at https://github.com/pranavworks100/fennec so problem right now is sometimes roasts hit and sometimes they are just bland and generic. so if have ideas to make it even more funnier, please do so :) and another issue is that it sometimes pulls old news from months ago. Comments URL: https://news.ycombinator.com/item?id=49991917 Points: 1 # Comments: 0
OpenAI has published 372 AI-generated mathematical results on GitHub, including Lean formalizations for machine verification. Each result consumed about three hours of ChatGPT Pro compute on average. But 25 Fields Medal winners warn that mass-producing mathematical truths could destroy fertile ground rather than bring new ideas to life. The article OpenAI dumps 372 AI-generated math proofs on GitHub, telling the academic world to keep up appeared first on The Decoder .
Meta has open-sourced Rebalancer, the C++ and Python library it has used for over 9 years to place shards, servers and traffic. It handles about 40 million assignment problems a day, using local search or MIP solvers like Gurobi, FICO Xpress and HiGHS. It is pip-installable under Apache 2.0. The post Meta AI Open-Sources Rebalancer: A C++ Assignment Solver That Runs About 40 Million Placement Problems a Day appeared first on MarkTechPost .
arXiv:2610.05922v2 Announce Type: replace-cross Abstract: Evaluating the causal effect of new features is a central goal for online platforms. While recent literature addresses limited testing traffic via centralized portfolio optimization, this perspective abstracts away a critical institutional reality: experimentation is operationally decentralized. The experimenters who develop new features also dictate which hypotheses to test, and they are typically rewarded based on empirical average treatment effects that are prone to upward bias. Left unchecked, this principal-agent conflict can severely erode platform value, a structural failure that conventional centralized levers, such as significance thresholds and traffic budgets, cannot resolve. By reframing experimentation as an incentive design problem, we demonstrate that two practical mechanisms, sample splitting and shrinkage, can effectively bridge this gap. Sample splitting aligns incentives perfectly at a bounded traffic cost, while shrinkage consumes no additional traffic and guarantees that interventions with negative expected effects are strictly unprofitable to field.
arXiv:2609.32652v2 Announce Type: replace-cross Abstract: A soft state representation assigns each state a vector of nonnegative class weights that sum to one. We study how the construction of these weights and the state dynamics jointly determine the accuracy of linear prediction. For any fixed measurable representation, we derive a finite-sample lower confidence bound on the smallest population root-mean-square prediction error among matrices with a specified spectral-norm limit. The bound compares variation in successor coordinates within each reference class with the improvement that soft inputs could provide. It is computed from independent evaluation pairs without fitting a prediction matrix. A bound above a chosen tolerance rules out that tolerance for the entire matrix class; a zero bound is inconclusive. For coordinates constructed using Kernel Affine Hull Machines, reconstruction-score margins control disagreement with reference labels and enter bounds on prediction error. Under exact deterministic linear evolution, we also establish the Koopman and reproducing-kernel Hilbert-space adjoint interpretation, accounting for redundant coefficient vectors. A four-state study compares the confidence bound with analytically known optima across 117,000 reported replicate datasets. A Van der Pol representation selected on pilot data is then evaluated on 32 independent datasets under each of two transition laws. The reported bounds are positive at the fitted matrix norm, but can become zero at larger norm limits. Further forecasting studies examine coordinate variation, common prediction targets, and long-horizon error. The results distinguish agreement with reconstruction classes, attainable prediction accuracy, and exact operator closure.
arXiv:2609.03203v4 Announce Type: replace-cross Abstract: Plan accuracy alone cannot show whether a speech-delivery decision follows its source: a fixed prior may match the original label yet fail to respond appropriately when a cue changes. VoxReason provides a 100-case verifier benchmark that holds each utterance fixed, edits one designated source-label cue, and scores cited evidence, eight plan fields, and the permitted response. On a source-key-disjoint test of 24 cases, a source-emotion prior reaches plan-slot accuracy 0.958, but none of the 24 edited neutral targets appears in its training labels; its required-change accuracy is 0.000. This diagnoses the support boundary of this prior, not its performance on supported edits. In a complementary 32-case emotion-disjoint test, the prior has seen all edited neutral targets but neither original test emotion; its plan-slot accuracy is 0.219 and required-change accuracy is 1.000. The partitions reuse and overlap the same 100 cases, so these deterministic diagnostics are not independent cohorts or learned-planner results. The benchmark evaluates derived labels and structured plans, not audio input, generated speech, or listener judgments.
arXiv:2608.28733v2 Announce Type: replace-cross Abstract: Indoor 3D Scene Graphs (3DSGs) represent environments as multi-layer hierarchies that connect observed geometric primitives (e.g., planes) to higher-level metric-semantic concepts (e.g., rooms, floors, buildings), enabling incremental spatial reasoning for robotic perception and SLAM. However, classical high-level concept generation approaches rely on hand-crafted rules for specific concept classes, while learning-based methods require separate models for graph structure and spatial node features (e.g., centroids), which limits scalability to novel classes and more complex hierarchies. We propose a unified autoregressive diffusion-based graph generative model that jointly learns structure and features, constructing complete 3DSGs bottom-up from observed vertical planes across arbitrary hierarchy depths. Our method consistently surpasses all learning-based and random baselines across 3DSG datasets spanning synthetic scenes, real architectural floor plans, and robotic sensor data, with varying layout complexity and hierarchy depth, and surpasses a one-shot model with oracle access to the target graph size on the largest hierarchy and on real single-floor data. Finally, we propose an adaptation of the Fused Gromov--Wasserstein distance for principled graph-level evaluation of generated 3DSGs against ground truth.
arXiv:2608.24631v2 Announce Type: replace-cross Abstract: Similarity in many decision systems is governed not by distance alone but by interactions among variables. In fraud and anomaly detection, small local perturbations can cross interaction-sensitive decision boundaries while leaving ambient distance almost unchanged. Motivated by this setting, we introduce a thin-slab interaction model and an interaction-driven quantum kernel constructed from entangled Pauli-string feature maps. The feature map explicitly encodes sparse high-order block interactions. We show that the resulting fidelity kernel is positive semidefinite, admits an exact block-factorized formulation, and induces a geometry sensitive to changes in interaction regime. Across balanced and imbalanced synthetic experiments spanning third-, fourth-, sixth-, and eighth-order interactions, the proposed kernel consistently outperforms linear, radial basis function, Laplacian, and polynomial kernels, as well as an engineered-interaction linear baseline supplied with the planted block products. On real fraud-detection benchmarks, it achieves the highest mean accuracy and F1 on Credit Card Fraud Detection and ranks second on IEEE-CIS Fraud Detection. Executed on a 156-qubit IBM Quantum processor in a fourth-order setting, the hardware-estimated kernel matches the noise-free simulator within seed-to-seed variability and retains its advantage over the baselines. These findings show that quantum-kernel performance depends on alignment between feature-map geometry and the underlying predictive structure, rather than on Hilbert-space dimension alone. Because the prescribed block-factorized kernel can also be evaluated exactly on a classical computer, the results establish predictive and representational value rather than computational quantum speedup.
arXiv:2608.19693v2 公告类型：replace-cross 摘要：稀疏关键点提取与匹配是几何计算机视觉中核心任务的基础，包括运动恢复结构、视觉 SLAM、增强现实和医学图像配准。然而，学习鲁棒的局部特征表示通常需要精确的相机位姿或深度监督，而这些信息在现实场景中往往难以获取。强化学习（RL）最近成为一种颇具潜力的替代方案，它仅需知道两幅图像是否呈现同一场景。然而，RIPE 等现有 RL 方法依赖粗粒度的二元奖励和精心构造的负训练样本对，限制了训练稳定性和描述子的判别能力。本文重新审视基于 RL 的关键点学习，并提出一种充分利用几何一致性信号的奖励机制，从单个正样本对中同时获得奖励与惩罚，无需与负样本进行对比。这种更丰富的信号提供了充分的监督对比信息，使模型能够仅从正图像对中学习具有判别能力的检测器和描述子，从而在监督极其有限的条件下实现表示学习。此外，我们通过改造 LightGlue，证明同一 RL 目标也可扩展至匹配阶段，将 MegaDepth1500 上的 AUC@5 从 56.58 提升至 59.65，并支持利用具有部分视觉重叠的图像对，对整个稀疏匹配流程进行弱监督训练。我们在公认的基准上验证了该方法，结果表明其性能可与全监督方法相媲美。我们进一步表明，该方法甚至可以在低纹理医学视频序列上训练，而这类序列通常缺乏相机位姿信息，标准 SfM 流程也往往失效。代码和数据可在 https://github.com/fraunhoferhhi/RIPEpp 获取。
arXiv:2606.16193v2 公告类型：replace-cross 摘要：多模态大语言模型（MLLMs）在视觉语言任务上表现出色，但其内部视觉表征仍然难以解释。稀疏自编码器（SAEs）提供了一种可扩展的方法，可将稠密的模型激活分解为稀疏、可解释的特征。然而，现有 SAE 架构主要学习扁平化的特征字典，不太适合显式地组织多层级概念。本文提出一种名为 SAE++ 的级联稀疏自编码器架构，用于学习 MLLMs 中的层次化视觉概念。SAE++ 并非嵌套或堆叠 SAE 的稀疏激活编码，而是直接以第一级 SAE 的解码器权重训练第二级 SAE，将学到的低层级特征方向作为更高层级抽象的输入。这一设计使 SAE++ 能够学习“概念的概念”，同时避免嵌套式 Matryoshka 层次结构中共享前缀耦合带来的弊端，以及简单堆叠 SAE 所产生的瓶颈。在多个视觉数据集上针对 Qwen3-VL、Gemma-3 和 LLaVA 开展的实验表明，与当前最先进的 SAE 基线相比，SAE++ 在层次化概念一致性方面提升了可解释性。概念引导实验的结果进一步表明，学到的概念组能够支持对 MLLMs 输出进行有效的组级干预。代码已发布于 https://github.com/Wang-ML-Lab/sae-plus-plus。
arXiv:2606.11570v2 Announce Type: replace-cross Abstract: We propose a spectral-based, unsupervised representation learning framework to derive low-dimensional embeddings for clinical concepts and patients in rare disease cohorts from electronic health records, where data are high-dimensional but sample sizes are limited. To overcome this challenge, we incorporate a knowledge matrix extracted from a broader population that shares a partially overlapping subspace with the rare-disease cohort. Our method departs from existing approaches by relaxing restrictive one-to-one signal-alignment assumptions between the latent data matrix and knowledge matrix, allowing more flexible and realistic forms of structured sharing. We introduce a novel two-step spectral embedding procedure: first, we identify and remove irrelevant components from the knowledge matrix; then, we apply a projection-based method to separately recover shared and heterogeneous components. Simulations and an analysis of a real-world multiple sclerosis cohort show that the proposed method outperforms competing approaches, particularly in challenging scenarios where shared signals are weak and only partially aligned, as is common in rare-disease data.
arXiv:2605.29695v2 Announce Type: replace-cross Abstract: Approximately 10% of newborns require assistance to initiate breathing at birth, and around 5% need ventilation support. Fetal heart rate (FHR) monitoring plays a crucial role in assessing fetal well-being during prenatal care, enabling the detection of abnormal patterns and supporting timely obstetric interventions to mitigate fetal risks during labor. Applying artificial intelligence (AI) methods to analyze large datasets of continuous FHR monitoring episodes with diverse outcomes may offer novel insights into predicting the risk of needing breathing assistance or interventions. Recent advances in wearable FHR monitors have enabled continuous fetal monitoring without compromising maternal mobility. However, sensor displacement during maternal movement, as well as changes in fetal or maternal position, often lead to signal dropout, resulting in gaps in recorded FHR data. Such missing data limits the extraction of meaningful insights and complicates automated (AI-based) analysis. Traditional approaches to handling missing data, such as simple interpolation techniques, often fail to preserve the spectral characteristics of the signals. In this paper, we propose a masked transformer-based autoencoder approach to reconstruct missing FHR signals by capturing both local temporal and frequency components of the data. The proposed method demonstrates robustness across varying durations of missing data and can be used for signal inpainting and forecasting. The proposed approach can be applied retrospectively to research datasets to support the development of AI-based risk algorithms. In the future, the proposed method could be integrated into wearable FHR monitoring devices to achieve earlier and more robust risk detection.
arXiv:2605.07079v2 公告类型：replace-cross 摘要：世界模型根据观测和动作预测未来的状态转移。现有研究主要只关注图像生成。相比之下，基于视觉特征的世界模型预测未来的视觉特征，而非原始视频像素，提供了一种更高效且更不易产生幻觉的有前景的替代方案。然而，当前基于特征的方法依赖直接回归，这会在复杂交互中导致预测模糊或坍缩，而在高维特征空间中进行生成式建模仍然具有挑战性。在本研究中，我们发现，可以轻松地从 DINO 残差中学习一种新型潜在动作表示，我们将其称为 Residual Latent Action（RLA）。我们还表明，RLA 具有预测能力和泛化能力，并编码了时间演进信息。基于 RLA，我们提出了 RLA World Model（RLA-WM），通过流匹配预测 RLA 值。在仿真和真实世界数据集上，RLA-WM 的表现均优于最先进的基于特征的世界模型和视频扩散世界模型，同时速度比视频扩散快数个数量级。此外，我们开发了两种利用 RLA-WM 改进策略学习的机器人学习技术。第一种是采用 RLA 的极简世界动作模型，它从无动作标注的视频中学习，并提升了 VLA 在 LIBERO 和真实机器人上的表现。第二种是一个视觉强化学习框架，其训练完全在仅从离线视频中学习得到的世界模型内部进行，使用与视频对齐的奖励，无需任何在线交互。项目页面：https://mlzxy.github.io/rla-wm
arXiv:2605.05866v2 公告类型：replace-cross 摘要：多相粉末 X 射线衍射（PXRD）分析仍是结构鉴定中的一个根本性瓶颈，因为实际合成过程往往会产生复杂混合物，而其中的组成相（组分）难以可靠地分离辨识。尽管近期基于表征的晶体检索与生成技术进展表明，直接从 PXRD 推断结构具有可行性，但现有方法大多假定输入为单相，在多相场景下则会失效。在此，我们提出 XDecomposer，这是一种无先验框架，可对多相 XRD 图谱进行联合分解与鉴定，无需候选相列表、结构模板或相数量的先验知识。我们将多相衍射分析表述为集合预测问题，由模型在统一架构中推断一个无序集合，其中包含各相分辨的组分、其混合比例及相应的结构表征。相查询驱动的分解机制结合与衍射一致的物理重建，可在保持晶体学保真度的同时实现精确的源分离。在模拟数据集和实验数据集上开展的大量实验表明，XDecomposer 在多种化学体系中显著提高了重建精度和物相鉴定能力，同时对未见过的混合物保持了较强的泛化能力。这些结果为数据驱动、源分辨的多相 XRD 分析提供了一条实用途径，并减少了对先验引导的迭代物相匹配的长期依赖。代码已公开发布于 https://github.com/Licht0812/XDecomposer
arXiv:2604.00055v2 公告类型：replace-cross 摘要：现有机器人基础策略主要通过大规模模仿学习进行训练。尽管这些模型展现出强大的能力，但由于分布偏移和误差累积，它们往往难以应对长时程任务。虽然强化学习（RL）可以对这些模型进行微调，但若不进行人工奖励设计，就难以在多样化任务中取得良好效果。我们提出了 VLLR，这是一种密集奖励框架，结合了（1）来自大语言模型（LLMs）和视觉语言模型（VLMs）、用于识别任务进度的外在奖励，以及（2）基于策略自身确定性的内在奖励。VLLR 使用 LLMs 将任务分解为可验证的子任务，随后使用 VLMs 估计进度，在短暂的预热阶段初始化价值函数，从而避免在整个训练过程中产生高昂的推理成本；同时，自身确定性在整个 PPO 微调过程中提供逐步的内在引导。消融研究揭示了二者的互补优势：基于 VLM 的价值初始化主要提高任务完成效率，而自身确定性主要提升成功率，尤其是在分布外任务上。在涵盖移动操作与导航的 CHORES 基准上，VLLR 相比预训练策略的成功率绝对提升最高达 56 个百分点；相比最先进的 RL 微调方法，在分布内任务上的提升最高达 5 个百分点，在分布外任务上的提升最高达 $10\%$，且均无需人工奖励设计。更多可视化结果见 https://silongyong.github.io/vllr_project_page/
arXiv:2603.19899v2 Announce Type: replace-cross Abstract: Autocorrelation is a common property of time-series, where each observation is dependent on its predecessors. In deep time-series forecasting, it raises two central challenges: (1) designing backbone architectures to model autocorrelation in history sequences, and (2) devising loss functions to model autocorrelation in label sequences. Recent studies have made strides in tackling these challenges, but a systematic survey examining both aspects remains lacking. To bridge this gap, this paper reviews deep time-series forecasting from an autocorrelation modeling perspective, offering two contributions beyond existing surveys. First, it introduces a taxonomy that jointly covers both backbone architectures and loss functions, whereas prior surveys provide limited coverage of the latter. Second, it analyzes the motivations and insights underlying the surveyed literature from a unified autocorrelation perspective, providing a holistic overview of the field's evolution. Additional resources and details are available at https://github.com/Master-PLC/Awesome-TSF-Papers.
arXiv:2602.18899v4 公告类型：replace-cross 摘要：已知自监督语音模型（S3Ms）能够编码丰富的语音信息，但这些信息的组织结构仍缺乏深入研究。我们开展了一项涵盖 96 种语言的全面研究，分析 S3M 表征的底层结构，重点关注音系向量。我们首先表明，模型的表征空间中存在与音系特征相对应的线性方向。我们进一步证明，这些音系向量的大小与其对应音系特征的声学实现程度呈连续相关关系。例如，[d] 与 [t] 之差形成一个浊音向量：将该向量加到 [p] 上会得到 [b]，而缩放该向量则会产生浊音程度的连续变化。综合来看，这些发现表明，S3Ms 使用具有音系可解释性和组合性的向量来编码语音，展现出音系向量算术能力。所有代码和交互式演示均可在 https://github.com/juice500ml/phonetic-arithmetic 获取。
arXiv:2509.03647v3 公告类型：replace-cross 摘要：大型语言模型（LLM）日益被用作自动化评估器，但它们存在“自我偏好偏差”：倾向于偏爱自身输出，而非其他模型的输出。这种偏差损害了评估流程的公平性和可靠性，尤其是在偏好调优和模型路由等任务中。我们研究了轻量级引导向量能否在无需重新训练的情况下，于推理阶段缓解这一问题。我们引入了一个经过精心整理的数据集，将自我偏好区分为合理的自我偏好示例和不合理的自我偏好示例，并采用两种方法构建引导向量：对比激活加法（CAA）和基于优化的方法。结果表明，引导向量能够将不合理的自我偏好偏差最多减少 97%，显著优于提示方法和直接偏好优化基线。然而，引导向量在合理的自我偏好和无偏一致判断上的表现并不稳定，这意味着自我偏好涉及多个方向或非线性方向。这凸显了引导向量作为 LLM 裁判防护手段的潜力与局限，并促使我们探索更稳健的干预方法。
arXiv:2501.18158v4 Announce Type: replace-cross Abstract: Cryptocurrencies are widely used, yet current methods for analyzing transactions often rely on opaque, black-box models. While these models may achieve high performance, their outputs are usually difficult to interpret and adapt, making it challenging to capture nuanced behavioral patterns. Large language models (LLMs) have the potential to address these gaps, but their capabilities in this area remain largely unexplored, particularly in cybercrime detection. In this paper, we test this hypothesis by applying LLMs to real-world cryptocurrency transaction graphs, with a focus on Bitcoin, one of the most studied and widely adopted blockchain networks. We introduce a three-tiered framework to assess LLM capabilities: foundational metrics, characteristic overview, and contextual interpretation. This includes a new, human-readable graph representation format, LLM4TG, and a connectivity-enhanced transaction graph sampling algorithm, CETraS. Together, they significantly reduce token requirements, transforming the analysis of multiple moderately large-scale transaction graphs with LLMs from nearly impossible to feasible under strict token limits. Experimental results demonstrate that LLMs have outstanding performance on foundational metrics and characteristic overview, where the accuracy of recognizing most basic information at the node level exceeds 98.50% and the proportion of obtaining meaningful characteristics reaches 95.00%. Regarding contextual interpretation, LLMs also demonstrate strong performance in classification tasks, even with very limited labeled data, where top-3 accuracy reaches 72.43% with explanations. While the explanations are not always fully accurate, they highlight the strong potential of LLMs in this domain. At the same time, several limitations persist, which we discuss along with directions for future research.
arXiv:2610.06505v2 公告类型：替换 摘要：主动式 AI 助手持续观察用户活动，并决定是提供新的指导还是保持沉默。它们应针对任务提供适当的指导，根据任务进度确定何时提供下一条指导，并根据用户的专业水平和需求调整指导层级。支持这些能力需要能够反映流程结构、并体现指导应如何随任务进度和用户需求而调整的训练与评估数据。然而，现有数据集要么侧重于基于检测的主动式理解，要么仅提供固定粒度的流程指导。固定粒度指导所提供的细粒度进度和更宏观流程上下文信息有限，因此难以判断任务是否完成，也难以调整指导粒度。为解决这些局限，我们提出 ProactiveCoach 套件，包括用于训练的 ProactiveCoach-Instruct、用于评估的 ProactiveCoachBench，以及配备自适应指导系统的微调视觉语言模型（VLM）。ProactiveCoach-Instruct 提供阶段、步骤和动作三个层级的层次化指导，用于学习任务进度和流程上下文。ProactiveCoachBench 评估模型能否在不同指导层级下于恰当时机提供适当指导，以及能否在所请求的层级发生变化时进行适应。我们在 ProactiveCoach-Instruct 上对预训练 VLM 进行微调，并验证了其在不同骨干模型上的有效性。与固定粒度监督相比，层次化监督在不同骨干模型上带来的整体性能提升最高达 9.6 个百分点。我们进一步将微调后的模型与轻量级指导路由器相结合，构建了自适应指导系统。无需额外微调，我们的系统在四种指导层级转换中的表现比上下文内适应基线高出 57.1 个百分点。项目页面：https://jinsuby.github.io/ProactiveCoach/。
arXiv:2610.05108v2 公告类型：替换 摘要：扩散模型具有强大的生成性能，但往往难以使生成样本符合奖励模型所衡量的人类偏好。一种简单而有效的测试时对齐算法是 Best-of-$N$（BoN）采样，它从预训练的扩散模型中抽取 $N$ 个独立同分布的样本，并输出其中唯一一个奖励最高的样本。尽管 BoN 在实证中取得了成功，但它对奖励信息的利用有限，因为奖励信息仅在最终选择阶段被引入，并不影响采样过程中的反向扩散轨迹。因此，BoN 采样无法提高生成样本的平均对齐水平，主要适用于单输出场景。我们提出 Best-of-$N$ Guidance（BoNG），这是一种将 BoN 采样原理直接融入反向扩散过程的新方法。BoNG 对去噪粒子进行在线 BoN 选择，并调整反向扩散过程，在生成过程中引导粒子群向奖励更高的区域移动。具体而言，BoNG 通过在去噪粒子之间引入非对称引导交互，将当前的 BoN 粒子用作其余粒子的引导信号。这种粒子级交互重塑了采样过程，使其朝向奖励更高的区域，从而使 BoNG 不仅能获得优于 Vanilla BoN 采样的最终最佳样本，还能提高生成样本的平均质量。在 36 项实证比较中，BoNG 在 29 项中取得最佳性能，在与 SMC 和 Vanilla BoN 采样的比较中，以 80.56% 的比例位居第一。BoNG 还支持多输出，其 ImageReward 得分达到最新的基于样本的引导方法的 1.3$\times$，同时实现了 1.6$\times$ 的加速。我们在 https://github.com/aailab-kaist/BoNG 发布代码。
arXiv:2610.04946v2 公告类型：替换 摘要：由于不同机构和国家在临床编码、患者群体及医疗工作流程方面存在显著差异，纵向电子健康记录（EHR）表征的跨系统迁移面临挑战。我们提出非对称监督对比学习（Asymmetric SupCon），这是一种针对特定任务的预训练目标，其设计动机源于阴性临床结局的异质性。该目标将具有相同目标阳性结局的患者聚集在一起，而不显式地使阴性轨迹相互靠近。我们使用台湾全民健康保险研究数据库（NHIRD）中 398 万名患者的纵向记录对时序 Transformer 编码器进行预训练，并将其迁移至两个美国 EHR 数据集 MIMIC-IV 和 EHRSHOT。通过结合直接映射与基于嵌入的检索，混合语义映射流程实现了异构临床词汇体系之间的迁移。在 MIMIC-IV 上，NHIRD 预训练始终优于随机初始化，同时大幅缩小了与针对特定任务的域内预训练之间的性能差距。在 EHRSHOT 上，迁移后的模型在新发疾病预测方面表现出尤为出色的少样本性能。在预训练规模匹配的受控目标函数消融实验中，Asymmetric SupCon 在全部四项评估任务上的平均 AUPRC 均高于直接监督式 BCE 迁移，在其中三项任务上高于 Standard SupCon，但在再入院任务上的 AUPRC 低了 0.003。这些结果表明，非对称对比预训练是实现针对特定任务的跨国 EHR 表征迁移的有效方法。代码已公开于 https://github.com/qingYzhang/Asymmetric_SupCon。
arXiv:2610.04142v2 Announce Type: replace Abstract: Modern training on a new task often starts from a previously trained model rather than from scratch, raising the question of how this initialization affects the subsequent training trajectory. Classical implicit-bias results characterize the direction selected by prolonged training, but this direction alone does not provide information regarding the intermediate behavior. We address this question through a geometric approximation of full-batch logistic gradient descent (GD) trajectories on strictly linearly separable data, with large initialization of scale $R$ motivated by prior training. From any limiting normalized initial position, we use minimum-norm projection rules to construct a unique continuous ideal path consisting of finitely many linear segments. The path has two stages: negative-margin correction followed by minimum-margin growth. We prove that, after an explicit two-stage time reparameterization, the fixed-step GD trajectory divided by $R$ converges uniformly to this path on every fixed parameter interval as $R\to\infty$. Further, our quantitative error bounds account for initialization perturbations and the transition between stages. This approximation provides asymptotic formulas for peak evaluation loss and cumulative training loss. In particular, peak evaluation loss can grow linearly in $R$ even when both endpoint losses tend to zero. The cumulative losses in the correction and margin-growth stages, normalized by $R^2$ and $R$, respectively, converge to explicit limits. Experiments on controlled geometries and fixed image features complement our theoretical results.
arXiv:2610.03945v2 Announce Type: replace Abstract: A constrained optimization problem may involve a parameter in its objective and active constraints, yet the final decision may remain insensitive to small changes in that parameter. This raises a fundamental question: which inputs does a decision making system truly depend on? Building on this question, we introduce DePICT, a procedure for constructing decision preserving interfaces by ranking context directions according to the optimizer's solution sensitivity and aggregating them across an operating regime. We study this problem in a high dimensional setting where primitive context parameterizes a constrained task and the downstream agent observes only a selected subset of context directions. For locally regular constrained programs, we derive a Karush Kuhn Tucker (KKT) based characterization of when a context direction is optimizer relevant. Our analysis shows that appearing in the active optimization problem does not necessarily imply that a variable affects the final decision. Some context directions can alter the KKT conditions while leaving the optimal solution unchanged because their effect is absorbed by the dual variables. DePICT is designed to remove exactly these directions. In a controlled diagnosis, it recovers the decision relevant interface exactly and reduces linear predictor regret to 0.009, compared with 0.475 for the strongest competing baseline.
arXiv:2609.39082v2 公告类型：替换 摘要：随着新证据的到来，序列模型必须更新其记忆内容以及记忆影响预测的方式。Transformers 的计算与缓存开销随上下文长度增长，而固定状态的循环模型可实现恒定内存开销的推理。然而，传统的线性递归和谱递归依赖静态状态转移，无法动态调整已存储表征的衰减或旋转方式。尽管近期的选择性架构引入了依赖输入的状态转移，但它们为每个记忆模态分配独立的控制，从而将控制成本与状态容量绑定。我们表明，高维谱记忆并不需要高维控制，并提出 Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence（SPARC）。SPARC 仅使用两个依赖输入的标量信号，就能协调异质复数模态的记忆保持与相位旋转，同时保留各模态特有的基准时间尺度和频率。其对角仿射递归既支持用于序列级 BPTT 的并行结合扫描，也支持用于在线信用分配的精确结构化实时循环学习（RTRL）。在部分可观测连续控制、POPGym 和序列分类任务中，相较于排名第二的方法，SPARC 在 Walker-P 上实现了 9.09% 的回报相对提升，在 FordA 上实现了 1.36% 的准确率相对提升。在 NVIDIA Blackwell GPU 上，相较于经过优化的 RG-LRU 基线，我们的实现在固定词元数量的工作负载下将循环混合器的训练延迟降低了 18.2%–34.2%，并将扫描速度提升至 3.1–4.7 倍。这些结果表明，两个共享控制信号就能在在线和完整序列场景中高效调控自适应谱记忆。代码已发布于 https://github.com/Botwwt/sparc。
arXiv:2609.35297v3 公告类型：替换 摘要：预训练语言模型需要大量计算资源，而合适的优化器可以节省其中相当一部分。Muon 的谱更新步能提供比符号更新步更强的方向，但代价高昂。每一步都要对完整矩阵执行 Newton-Schulz 迭代，在分布式训练中还需要额外执行一次全归约。Lion 和 Signum 所采用的符号更新步成本低，且始终在各设备本地执行。我们提出 LionMuon，每隔 $P$ 次迭代执行一次 Muon 更新步，其间执行 Lion 更新步，两者共享一个双指数移动平均（EMA）动量缓冲区。Muon 的计算和通信开销每 $P$ 步才产生一次，且优化器状态大小仅为 AdamW 的一半。采用单 EMA 的变体 SignMuon 就已优于 Muon。我们证明了重尾噪声下的复杂度界，其中周期决定了 Muon 与 Lion 的平滑性常数和噪声常数之间的插值，并揭示了 LionMuon 何时比两者都更快。在 FineWeb 上训练的 124M 和 355M 模型中，当处理的词元数量相同时，$P=2$ 和 $P=5$ 的 LionMuon 均取得了比 Muon、AdamW、Lion 和 Signum 更低的损失。在采用 PCIe 互连的 4-GPU 数据并行训练中，它达到 Muon 最终损失所需的实际运行时间减少了三分之一；同时，在不增加未被计算掩盖的通信开销且保留精确梯度的情况下，其损失表现优于通信高效的 Muon 变体 Dion 和 MuonBP。代码：https://github.com/brain-lab-research/lion-muon
arXiv:2608.05446v2 公告类型：替换 摘要：长时程 LLM 智能体日益依赖外部执行支持，以在持续交互中维护状态、跟踪进展、从失败中恢复并复用经验。然而，现有支撑框架及其使用方式通常针对特定环境定制，并通过提示、启发式方法或系统专用规则加以控制，因而难以对智能体与支撑框架之间的协同进行联合优化。我们提出 EvoHarness-RL，这是一个将环境特定的支撑框架实现与共享的策略接口分离的统一框架。EvoHarness-RL 将外部支持组织为信念、进展与经验（BPE）工作空间，并提供四种精简的支撑框架动作，用于访问和更新这一状态。我们首先将 BPE 实现为推理时的辅助支架，随后通过监督初始化和成本感知的 GRPO，使支撑框架协同成为可学习的能力。在多种异构长时程任务上，EvoHarness-Base 将前沿模型的平均成功率提高了 10.0 个百分点，而 EvoHarness-RL 比最强的开源基线高出 8.5 个百分点，同时实现了更高的 RL 轨迹采样效率，以及对未见任务更强的泛化能力。我们的分析表明，随着策略能力增强，训练会逐步推动智能体从频繁使用辅助支架转向根据环境有选择地访问支撑框架，而外部工作空间则持续演化并自我完善，以更好地支持任务执行与泛化。总体而言，这些结果表明，长时程智能体不仅受益于外部辅助支架本身，也受益于将如何以及何时与外部支持协同纳入成本感知策略进行学习。
arXiv:2607.25531v2 Announce Type: replace Abstract: Contemporary machine learning struggles to learn continually, reuse prior knowledge, and expose a comprehensible internal structure. A recently proposed developmental, gradient-free learning framework addresses these limitations by learning a discrete, topological model of its inputs through local variation and selection, yielding an inherent continual-learning guarantee: new observations refine existing structure without overwriting past knowledge, and without replay buffers or predefined task boundaries. Its extension to visual inputs demonstrated this principle on shape recognition, but relied on a feature representation of limited expressivity that capped recognition accuracy. We introduce a new visual feature representation that encodes shape structure across multiple scales, capturing edge and contour features together with their spatial relations, and integrate it with the network-refinement learning process; we further improve the learning dynamics and the read-out used to predict from the learned model. The study targets two-dimensional shape, with class-incremental MNIST as a controlled, interpretable benchmark in which continual-learning behavior can be measured directly. Our approach substantially increases accuracy over the prior representation, matching or exceeding replay- and regularisation-based baselines at comparable storage while storing no past data, and preserves the framework's defining behavior: earlier-learned classes are retained as new ones are introduced, with no destructive adaptation, and the learned representations remain human-interpretable. What separates the methods is retention: the baselines surrender most of a just-trained class within its own cycle and relearn it afterwards, which ours does not. The significance lies in the manner of learning. The system integrates information one sample at a time while provably preserving its responses to...
arXiv:2606.04931v2 Announce Type: replace Abstract: Mean-based algorithms are online learning algorithms that assign low probability to actions with low average rewards. Recent research shows that they converge to serially undominated actions, which serve as approximations to Nash equilibria in economic games. However, empirical studies indicate that mean-based algorithms converge more slowly in bandit-feedback settings than established no-regret alternatives. This work investigates mean-based algorithms under unknown horizons and bandit feedback. In this setting, we provide the first lower bound on the algorithm-defining sequence $\gamma_t$, establishing a fundamental limit on the learning speed of such algorithms. In multi-armed bandit problems, this result constrains the rate at which any algorithm can reliably identify low-reward actions while acting according to this knowledge. We also propose two mean-based algorithms: one generalizes $\epsilon$-greedy, and the other extends mean-based Exp3 to unknown horizons. Our experiments show that mean-based algorithms, although slightly slower, can perform competitively with other bandit-feedback algorithms. We further study the relationship to regret. Depending on the choice of $\gamma_t$, the intersection with no-regret algorithms is non-trivial, and we show that some algorithms are both mean-based and no-regret.
arXiv:2606.00880v2 公告类型：替换 摘要：持续强化学习（RL）旨在构建能够不断适应新任务的智能体。一个关键问题是，这一目标与智能体所经历的任务多样性如何相互作用。已有研究表明，在大量多样化任务上训练能够使智能体具备强大的零样本适应和上下文适应能力。然而，这些研究是在智能体停止学习后，即权重被冻结的情况下，对其进行评估的。任务多样性如何影响智能体在一系列分布偏移中持续学习的能力，仍不明确。我们提出 Banyan，这是一个 GPU 加速的持续 RL 环境，可通过参数化方式控制定义任务的三个独立维度：智能体必须穿行的地图布局、必须与之交互的对象，以及子目标依赖关系的层级结构。我们发现，增加各个维度上的多样性都会促成系统性迁移——也就是说，智能体在新的任务分布上开始训练时，其性能接近在前一个任务分布上达到的水平，即使分布偏移改变了最优策略的结构也是如此。尽管增加多样性能够改善系统性迁移，但我们发现，过高的多样性会抑制学习者持续适应新任务分布的能力。随着多样性增加，学习者在新任务上的成功率进入平台期，但在旧任务上的表现却持续改善——即使不再接触这些旧任务。我们发现，这一现象在不同的持续学习算法、记忆架构、架构规模以及基于物理的控制环境 Kinetix 中均有出现。我们发布 Banyan，作为开展受控实验的环境，用于研究大量任务条件下的持续 RL。代码见 https://github.com/nhshah15/banyan。
arXiv:2605.18882v2 公告类型：替换 摘要：LLM 智能体表现出持续的过度调用倾向，即使在无需工具的情况下也会调用工具。在 When2Call 基准上，来自三个系列的六个模型均表现出较高的调用准确率，但不调用准确率低得多，导致总体准确率仅为 55%–70%。我们将这一现象归因于内在偏置假说（IBH）：调用／不调用的决策映射带有一个与激活无关的调用偏移项，因此，即使两者激活水平相同，模型也会偏向调用。利用稀疏自编码器（SAEs），我们提取出与 call/no_call 决策行为对齐的特征基，将其归约为带符号的激活差值，并直接估计该偏移项。在全部六个模型中，只有当 no_call 激活强于 call 激活时，模型才在决策上保持中立，这与 IBH 一致。随后，我们使用自适应差值校准引导（AMCS）对 IBH 进行因果检验；该方法沿 SAE 解码器方向施加闭式求解的反偏置偏移。抵消诊断出的偏移项能够缓解过度调用并提高总体准确率，同时调用准确率的下降微乎其微。我们的工作将过度调用从一种经验现象重新界定为可进行因果校正的机制性对象。代码已公开：https://github.com/SKURA502/agent-sae/。
arXiv:2604.07925v2 Announce Type: replace Abstract: The self-attention mechanism is central to the success of Transformer architectures. However, standard row-stochastic attention has been shown to suffer from significant signal degradation across layers. In particular, it can induce rank collapse, resulting in increasingly uniform token representations, as well as entropy collapse, characterized by highly concentrated attention distributions. Recent work has highlighted the benefits of doubly stochastic attention as a form of entropy regularization, promoting a more balanced attention distribution and leading to improved empirical performance. In this paper, we study rank collapse across network depth and show that doubly stochastic attention matrices normalized with Sinkhorn algorithm preserve rank more effectively than standard softmax row-stochastic ones. As previously shown for softmax, skip connections are crucial to mitigate rank collapse. We empirically validate this phenomenon on both sentiment analysis and image classification tasks. Moreover, we derive a theoretical bound for the pure self-attention rank decay when using Sinkhorn normalization and find that rank decays to one doubly exponentially with depth, a phenomenon that has already been shown for softmax.
arXiv:2604.02691v2 Announce Type: replace Abstract: Deep learning based semantic communication has achieved significant progress in wireless image transmission, but most existing schemes rely on fixed models and thus lack robustness to diverse image contents and dynamic channel conditions. To improve adaptability, recent studies have developed adaptive semantic communication strategies that adjust transmission or model behavior according to either source content or channel state. More recently, MoE-based semantic communication has emerged as a sparse and efficient adaptive architecture, although existing designs still mainly rely on single-driven routing. To address this limitation, we propose a novel multi-stage end-to-end image semantic communication system for multi-input multi-output (MIMO) channels, built upon an adaptive MoE Swin Transformer block. Specifically, we introduce a dynamic expert gating mechanism that jointly evaluates both real-time CSI and the semantic content of input image patches to compute adaptive routing probabilities. By selectively activating only a specialized subset of experts based on this joint condition, our approach breaks the rigid coupling of traditional adaptive methods and overcomes the bottlenecks of single-driven routing. Simulation results indicate a significant improvement in reconstruction quality over existing methods while maintaining the transmission efficiency.
arXiv:2603.29135v3 Announce Type: replace Abstract: Autonomous experimental systems are increasingly used in materials research to accelerate scientific discovery, but their performance is often limited by low-quality, noisy data. This issue is especially problematic in data intensive structure-property learning tasks such as Image-to-Spectrum (Im2Spec) and Spectrum-to-Image (Spec2Im) translations, where standard active learning strategies can mistakenly prioritize poor quality measurements. We introduce a gated active learning framework that combines curiosity driven sampling with a physics-informed quality control filter based on Simple Harmonic Oscillator model fits, allowing the system to automatically exclude low fidelity data during acquisition. Evaluations on a pre-acquired dataset of band-excitation piezoresponse spectroscopy (BEPS) data from PbTiO3 thin films with spatially localized noise show that the proposed method outperforms random sampling, standard active learning, and multitask learning strategies. We further deployed the framework in real-time experiments on a separate PbTiO3 thin-film sample with heterogeneous domain structures, demonstrating its effectiveness in autonomous microscopy experiments. Overall, this work supports hybrid autonomy in self-driving labs, where physics-informed quality assessment and active decision making work together for more reliable discovery.
arXiv:2603.21276v2 Announce Type: replace Abstract: The growing demand for on-device large language model (LLM) services on mobile edge devices has driven the adoption of Mixture-of-Experts (MoE) architectures, which scale model capacity with limited computation. Since fine-tuning MoE-based LLMs relies on privacy-sensitive local data, federated learning (FL) offers a natural paradigm for collaborative training without exposing raw data. However, integrating MoE-based LLM fine-tuning into FL faces two critical challenges caused by data heterogeneity across clients: (i) divergent local data distributions drive clients to develop distinct gating preferences, so direct parameter aggregation yields a one-size-fits-none global gating network; and (ii) same-indexed experts develop disparate semantic roles across devices, leading to expert semantic blurring and degraded specialization. To address these challenges, we propose FedAlign-MoE, a federated aggregation alignment framework for edge computing systems that jointly enforces routing consistency and expert semantic alignment. Specifically, FedAlign-MoE aggregates gating behaviors by aligning routing distributions through consistency weighting and optimizes local gating networks through distribution regularization, maintaining cross-client stability while preserving discriminative local gating preferences. Meanwhile, FedAlign-MoE quantifies the semantic consistency of same-indexed experts across devices and selectively aggregates semantically aligned experts, ensuring stable and specialized global experts. Extensive experiments demonstrate that FedAlign-MoE outperforms state-of-the-art benchmarks, achieving faster convergence and higher accuracy in non-IID federated environments with lightweight computation and efficient communication.
arXiv:2603.04364v2 公告类型：替换 摘要：同时处理屏幕截图和无障碍树的多模态网页智能体正日益广泛地用于与网页界面交互，但其双流架构也暴露出一个尚未得到充分研究的攻击面：攻击者向网页 DOM 注入内容，即可通过一致的欺骗性叙事同时污染两个观测通道。我们在 MiniWob++ 上开展的漏洞分析表明，包含视觉成分的攻击效果远超纯文本注入，揭示了以文本为中心的 VLM 安全训练存在关键缺陷。基于这一发现，我们提出双模态多阶段对抗安全训练（DMAST）框架，将智能体与攻击者的交互形式化为双参与方一般和马尔可夫博弈，并通过三阶段流程对双方进行联合训练：（1）向强大的教师模型进行模仿学习；（2）采用预言机引导的监督微调，利用一种新颖的零回应策略，使模型在对抗性噪声下仍能进行以任务为中心的推理；（3）通过群组相对策略优化（GRPO）自博弈开展对抗强化学习。在分布外任务上，DMAST 将攻击成功率降低了近一半（41.2%→21.4%），同时使任务完成率相对提升超过 60%（6.2%→10.2%）。我们的方法优于现有的基于训练的防御方法，并可与基于提示的防御方法互补，展现了实质性的协同进化进展，以及对复杂、未见环境的稳健泛化能力。代码已公开：https://github.com/huajianduzhuo-code/DMAST_official。
arXiv:2602.17050v4 公告类型：替换 摘要：嵌入表是大规模推荐系统的关键组件，能够将高基数类别特征高效映射为稠密向量表示。然而，随着唯一 ID 数量的增长，传统的基于哈希的索引方法会出现冲突，导致模型性能和个性化质量下降。我们提出 Multi-Probe Zero Collision Hash (MPZCH)，这是一种基于线性探测的新型索引机制，可有效缓解嵌入冲突。在合理设置表大小的情况下，它通常能够完全消除这些冲突，同时保持生产规模下的运行效率。MPZCH 利用辅助张量和高性能 CUDA 内核，实现可配置的探测策略和主动淘汰策略。通过淘汰过时的 ID 并重置重新分配的槽位，MPZCH 避免了基于哈希的方法中常见的陈旧嵌入继承问题，确保新特征能够有效地从头开始学习。尽管缓解冲突会带来额外开销，该系统仍能保持与现有方法相当的训练 QPS 和推理延迟。严格的在线实验表明，MPZCH 实现了用户嵌入的零冲突，并显著提升了物品嵌入的新鲜度和质量。该解决方案已在开源 TorchRec 库中发布，供广大社区使用。
arXiv:2602.14049v3 Announce Type: replace Abstract: Spatio-temporal traffic forecasting is a core component of intelligent transportation systems, supporting various downstream tasks such as signal control and network-level traffic management. In real-world deployments, forecasting models must operate under structural and observational uncertainties, conditions that are rarely considered in model design. Recent approaches achieve strong short-term predictive performance by tightly coupling spatial and temporal modeling, often at the cost of increased complexity and limited modularity. In contrast, efficient time-series models capture long-range temporal dependencies without relying on explicit network structure. We propose UniST-Pred, a unified spatio-temporal forecasting framework that first decouples temporal modeling from spatial representation learning, then integrates both through adaptive representation-level fusion. To assess robustness of the proposed approach, we construct a dataset based on an agent-based, microscopic traffic simulator (MATSim) and evaluate UniST-Pred under severe network disconnection scenarios. Additionally, we benchmark UniST-Pred on standard traffic prediction datasets, demonstrating its competitive performance against existing well-established models despite a lightweight design. The results illustrate that UniST-Pred maintains strong predictive performance across both real-world and simulated datasets, while also yielding interpretable spatio-temporal representations under infrastructure disruptions. The source code and the generated dataset are available at https://anonymous.4open.science/r/UniST-Pred-EF27
arXiv:2512.23073v2 Announce Type: replace Abstract: Fine-tuning has become the dominant paradigm for adapting Vision-Language Models (VLMs), yet most approaches rely on explicit weight updates that introduce a fundamental trade-off. Full Fine-Tuning (FFT) may perturb pretrained representations due to cross-modal gradient interference, whereas Parameter-Efficient Fine-Tuning (PEFT) methods rely on additive modules, such as low-rank adapters, which may limit adaptation capacity. In this paper, we rethink VLM adaptation from a structural selection framework that adapts VLMs without modifying backbone weights, and we propose Mask Fine-Tuning (MFT). MFT learns masks that selectively route information through existing pretrained connections, dynamically uncovering subnetworks that better align pretrained representations with downstream objectives. Extensive experiments show that MFT provides an effective structural alternative to both FFT and PEFT, consistently achieving superior performance across multiple vision-language benchmarks without adding knowledge or altering the deployment architecture. Moreover, our analysis with MFT provides new insights into how pretrained VLMs reorganize their internal representational pathways during adaptation.
arXiv:2511.00053v2 公告类型：替换 摘要：学习目标的设计是训练时间序列预测模型的核心。现有的学习目标（如均方误差）大多将每个未来时间步视为独立且权重相等的任务，这带来了以下两个挑战：（1）忽略了未来时间步之间的标签自相关效应，导致学习目标存在偏差；（2）未能为对应不同未来时间步的预测任务设置差异化权重，从而限制了预测性能。为填补这一空白，我们提出了一种新颖的二次型加权学习目标，同时解决这两个问题。具体而言，权重矩阵的非对角元素用于考虑标签自相关效应，而非均匀的对角元素则旨在匹配不同未来时间步预测任务的理想权重。在此基础上，我们提出了 Quadratic Direct Forecast（QDF）学习算法，利用自适应更新的二次型权重矩阵训练预测模型。实验表明，我们的 QDF 能够有效提升多种预测模型的性能，达到当前最先进水平。代码已公开，地址为 https://github.com/Master-PLC/QDF。
arXiv:2511.00044v4 Announce Type: replace Abstract: Physical neural networks (PNNs) are promising candidates for next-generation computing, but existing demonstrations remain several orders of magnitude smaller than modern digital neural networks, whose recent advances have been driven by rapid growth in trainable parameters. This situation resembles the constraints of early digital neural networks, which led to ideas around parameter reuse. We investigate what similarly efficient hardware architectures may look like, focusing specifically on the common bottleneck of slow re-adjustment of the weights in PNNs. We propose the Time-Indexed Deep Alternating Layers Network (TIDAL-Net), which occupies an intermediate regime between recurrent and deep neural networks, specifically aimed at the scales and restrictions of common PNN prototypes. TIDAL-Net leverages the timescale separation found in many PNNs between fast forward dynamics and slowly trainable weights and biases, using layer-by-layer time multiplexing to increase effective depth while limiting implementation cost. Numerical experiments on image classification and natural language processing tasks show that TIDAL-Net improves performance with only minor modifications to conventional PNNs.
arXiv:2510.14878v3 公告类型：替换 摘要：我们研究了在 CIFAR-5m、SVHN 和 ImageNet 等真实数据集上使用常见旋转不变核的核回归。我们提出了一个理论框架，仅根据两项测量结果——经验数据协方差矩阵和目标函数 $f_*$ 的经验多项式分解——即可预测学习曲线（测试风险随样本量的变化）。核心新思路是对核在各向异性数据分布下的特征值和特征函数进行解析近似。这些特征函数类似于数据的 Hermite 多项式，因此我们将这一近似称为 Hermite 特征结构拟设（HEA）。我们针对高斯数据证明了 HEA，但我们发现，真实图像数据往往“足够接近高斯分布”，使 HEA 在实践中也能很好地成立，从而让我们能够利用已有的关于核特征结构与测试风险之间关系的结果来预测学习曲线。在核回归之外，我们还通过实证发现，处于特征学习状态的 MLP 按照 HEA 预测的顺序学习 Hermite 多项式。我们的 HEA 框架提供了一个概念验证：对于真实数据集上的非平凡学习算法，构建一种从数据集结构一路映射到模型性能的端到端学习理论是可能的。
arXiv:2510.01264v2 公告类型：替换 摘要：机器人领域的对抗多智能体强化学习（MARL）进展受到制约，原因是缺乏能够在高保真物理仿真中支持异构智能体形态的共享、可扩展基础设施。现有框架要么专注于协作任务，要么依赖简化的物理引擎，要么提供难以扩展的孤立实现。我们提出 HARL-A，这是一个基于 IsaacLab 构建、持续维护的开源框架，支持针对形态多样的机器人团队进行可扩展的对抗策略训练与基准测试，允许任意数量的团队以及每个团队内任意组合的机器人形态。HARL-A 为 HARL 算法库和 IsaacLab 扩展了对抗多智能体支持，并提供了三个组成部分：（1）模块化软件架构，降低了定义新异构对抗环境的工程开销；（2）包含 Sumo、Soccer 和 3D Galaga 三个基准测试环境的套件，涵盖密集接触推挤、球技竞赛以及追逐与躲避；（3）十余个覆盖同构和异构团队配置的预训练策略，已在 Hugging Face 上公开发布，使用户无需从头重新训练即可立即探索对抗学习的动态过程。我们在多个竞争场景中演示了该框架，表明它能够可靠地生成学习得到的对抗策略，并涌现出角色专门化现象。所有代码、环境、训练好的策略和文档均已公开，地址为 https://github.com/DIRECTLab/IsaacLab-HARL。
arXiv:2508.14748v2 Announce Type: replace Abstract: The increasing variety of molecular data creates a need for generative models that can flexibly incorporate heterogeneous constraints across modalities. However, existing SMILES-based diffusion models are typically designed for a fixed conditioning modality, and introducing new constraints often requires retraining the model. To address this limitation, we propose Cross-Modality Controlled Molecule Generation with Diffusion Language Model (CMCM-DLM), a modular framework that extends a pre-trained diffusion model to support heterogeneous molecular constraints without retraining the backbone. We demonstrate CMCM-DLM using two complementary modalities: molecular structure and chemical properties. Specifically, a Structure Control Module (SCM) guides early diffusion steps to establish the molecular scaffold, while a Property Control Module (PCM) subsequently steers generation toward target chemical properties. This staged design enables flexible integration of different molecular constraints within a unified generative framework. Experiments on multiple datasets demonstrate effective cross-modal controllability and strong adaptability, highlighting the potential of CMCM-DLM for heterogeneous molecular data modeling and data-driven drug discovery.
arXiv:2505.17847v3 公告类型：替换 摘要：训练时间序列预测模型在损失函数设计方面面临独特挑战。大多数现有方法采用时域均方误差，但本研究揭示了其两个关键局限：（1）忽略了标签自相关性的存在，导致其偏离真实标签序列的似然；（2）涉及过多任务，使优化变得复杂，尤其是在长期预测中。为解决这些问题，我们提出了 Time-o1，一种用于时间序列预测的变换增强型损失函数。其核心思想是将标签序列变换为重要性各异的去相关分量，随后训练模型以对齐最重要的分量，从而有效缓解标签自相关性并减少任务数量。实验表明，Time-o1 达到了当前最先进的性能，并兼容多种预测模型。代码已发布于 https://github.com/Master-PLC/Time-o1。
arXiv:2505.14777v2 Announce Type: replace Abstract: The design of effective optimization algorithms for neural networks remains a fundamental challenge, and most existing methods rely on heuristic extensions of gradient-based updates. We introduce KO (Kinetics-inspired Optimizer), a plug-and-play optimization module grounded in kinetic theory and partial differential equations. KO models parameter dynamics as a particle system, augmenting standard gradient updates with stochastic interactions induced by a discretization of the Boltzmann transport equation. This mechanism naturally promotes parameter diversity and mitigates weight condensation, the tendency of parameters to collapse into low-dimensional subspaces, a phenomenon closely associated with degraded generalization. We provide both a rigorous theoretical analysis and a physical interpretation, showing that KO provably increases parameter diversity while preserving convergence guarantees. Extensive experiments on image classification benchmarks (CIFAR-10/100, ImageNet) and large-scale language model pretraining demonstrate that KO consistently improves accuracy over competitive baselines with negligible additional computational cost.
arXiv:2402.02399v3 公告类型：替换 摘要：由于历史数据和未来序列中均存在自相关性，时间序列建模面临独特的挑战。当前研究主要关注历史数据内部的自相关性，却往往忽略未来标签之间的相关性。具体而言，现代预测模型主要遵循直接预测（DF）范式，独立生成多步预测，而忽略标签在时间上的自相关性。在本研究中，我们证明，当标签存在自相关性时，DF 的学习目标存在偏差。为解决这一问题，我们提出了频率增强直接预测（FreDF），通过在频域中学习预测来缓解标签自相关性，从而降低估计偏差。实验表明，FreDF 显著优于现有最先进的方法，并且兼容多种预测模型。代码可在 https://github.com/Master-PLC/FreDF 获取。
arXiv:2311.18029v2 Announce Type: replace Abstract: The current trend in the literature on Time Series Classification is to develop increasingly accurate algorithms by combining multiple models in ensemble hybrids, representing time series in complex and expressive feature spaces, and extracting features from different representations of the same time series. As a consequence of this focus on predictive performance, the best time series classifiers are black-box models, which are not understandable from a human standpoint. Even the approaches that are regarded as interpretable, such as shapelet-based ones, rely on randomization to maintain computational efficiency. This poses challenges for interpretability, as the explanation can change from run to run. Given these limitations, we propose the Bag-Of-Receptive-Field (BORF), a fast, interpretable, and deterministic time series transform. Building upon the classical Bag-Of-Patterns, we bridge the gap between convolutional operators and discretization, enhancing the Symbolic Aggregate Approximation (SAX) with dilation and stride, which can more effectively capture temporal patterns at multiple scales. We propose an algorithmic speedup that reduces the time complexity associated with SAX-based classifiers, allowing the extension of the Bag-Of-Patterns to the more flexible Bag-Of-Receptive-Fields, represented as a sparse multivariate tensor. The empirical results from testing our proposal on more than 150 univariate and multivariate classification datasets demonstrate good accuracy and great computational efficiency compared to traditional SAX-based methods and state-of-the-art time series classifiers, while providing easy-to-understand explanations.
arXiv:2610.08789v1 公告类型：交叉发布 摘要：流策略已成为从示范中学习机器人行为的标准策略类别，但强化学习对于改进预训练流策略，或通过交互从零开始学习流策略，仍然至关重要。我们提出 QF3（采用过滤后 Q 梯度的快速流强化学习），这是一种在线离策略强化学习算法，将流匹配与评价器的动作梯度相结合来训练流策略，并通过对流输出的单步预测反向传播该梯度。为将更新限制在评价器和该预测可靠的范围内，QF3 仅对保持在经验回放动作附近的动作维度应用评价器梯度。据我们所知，QF3 是首个能够从零开始训练人形机器人运动策略，并将其零样本迁移至实体硬件的离策略流强化学习方法。结合高吞吐量的离策略训练方案，QF3 在训练人形机器人运动和动作跟踪策略时，相比近期提出的同策略流强化学习方法 FPO++，实现了 10 倍的实际运行时间加速。我们进一步将 QF3 应用于 ABC-Sim 和 Robomimic 任务，微调预训练的基于流的操作策略。这些结果表明，QF3 既能从零开始学习机器人策略，也能改进从示范中习得的策略。网站：https://qf3-rl.github.io/
arXiv:2610.08738v1 公告类型：交叉发布 摘要：扩散语言模型（DLMs）有望实现与顺序无关的并行文本生成。近年来，得益于精心设计的词元表示和扩散/流空间，连续扩散模型与流匹配模型取得了显著进展。在本研究中，我们提出 Hierarchical Continuous Diffusion Language Models（H-CDLMs），这是一个简单的框架，仅需极少的额外计算和参数开销，即可进一步提升连续 DLMs 的性能。借鉴离散 DLM 和连续图像扩散领域关于联合扩散的研究，我们对多种模态并行进行扩散。这些模态以不同的语义粒度表示词元：在我们的具体实现中，它们分别是词元本身，以及通过对预训练词元嵌入进行聚类得到的更粗粒度的簇。我们提出了一种通用设置，允许为各模态分别配置采样器和调度方案，以增强模态间的相互作用。将该设置应用于 CoBit 后得到的 H-CoBit，在各项基准测试中均取得了显著的实证性能提升。在达到数据集熵的条件下，H-CoBit 提高了 MAUVE，并在 LM1B 和 OWT 上分别达到 49.4 和 50.4 的生成困惑度（GenPPL），较基线分别降低 24.2 和 20.7 点，甚至超越了规模相当的离散 DLMs。在 GSM8K 上，其准确率达到 27.4%，优于此前的连续扩散模型和基于流的模型。我们还将 H-CDLM 应用于流匹配模型 FLM，所得的 H-FLM 同样取得了稳定的性能提升，证明该框架能够泛化至不同的连续生成范式。我们的代码将在 https://github.com/matol-16/HCDLM.git 公开。
arXiv:2610.08717v1 公告类型：交叉收录 摘要：我们研究将路径-流对齐作为流匹配的统一训练目标。不同于固定插值路径、仅学习速度场的做法，我们使用同一个对齐损失，联合训练一个保持端点不变的路径网络和一个流网络：流学习匹配路径速度，而路径学习使其速度与当前流对齐。尽管每条固定的已学习路径都定义了一个有效的流匹配目标，但仅靠对齐损失并不能可靠地指导路径学习。我们发现了路径过拟合这一失效模式，即对齐损失下降的同时，样本质量反而恶化。我们发现，这种失效与所诱导的概率路径中的低熵瓶颈有关：学习到的路径使样本经过过度集中的中间边缘分布。基于这一诊断，我们提出了一种随机路径正则化方法，在精确保留端点的同时，对路径网络隐藏部分源信息。这种正则化为随机训练路径的边缘分布提供了显式的熵下界，并在实验中抑制了所学习采样器中的瓶颈，使路径与流的联合训练切实有效。在 ImageNet-256x256 上，使用 SiT 骨干网络时，我们的方法在不同模型规模下均能持续改善 FID，也可扩展至模型引导训练，且无需改变推理时的架构和采样器。代码可在 https://github.com/lizeyu090312/traj_opt_paper 获取。
arXiv:2610.08626v1 公告类型：交叉收录 摘要：扩散模型通过一系列连续的去噪问题生成数据，人们普遍观察到，这些模型会先呈现粗粒度结构，再呈现精细细节。然而，这一直觉主要基于经验和定性观察。我们提出特征信息动力学，这是一种用于定位扩散过程中某个特征何时生成的信息论框架。利用 I-MMSE 恒等式，我们将特征互信息的变化率与最优无条件去噪损失和最优特征条件去噪损失之间的差值联系起来，从而得到实用的特征信息密度估计方法。我们进一步提出一种链式分解，将特征层级中的共享信息与增量信息分离。我们首先利用该框架定量证实像素扩散中的频谱自回归现象，随后将分析拓展到频率之外：在类别 $\to$ 掩码 $\to$ Canny 条件链下，像素、SDVAE、VAVAE 和 RAE 表示中的各特征信息密度呈现出差异，揭示了这些表示之间的根本区别，并表明有序生成可能有益于扩散模型的训练。我们的代码已公开，地址为 https://github.com/AI4Science-WestlakeU/feature-information-dynamics。
arXiv:2610.08574v1 Announce Type: cross Abstract: Skin cancer is a major global health concern, and early detection and accurate lesion delineation are important for effective diagnosis and treatment planning. Automated skin lesion analysis can assist dermatologists, with lesion segmentation serving as a fundamental step in computer-aided diagnostic systems. Conventional deep learning-based segmentation models typically rely on centralized training, where images and their corresponding segmentation masks are collected on a central server. Such data aggregation raises privacy concerns in medical applications and requires substantial centralized computational resources. To address these limitations, we investigate the feasibility of federated learning for privacy-preserving skin lesion segmentation. The training and validation sets of the ISIC 2018 Skin Lesion Segmentation Challenge dataset are used to simulate a distributed learning environment and develop a federated segmentation model. The resulting model is evaluated on the ISIC 2018 test set and the PH2 dataset to assess its performance and generalizability. Experimental results demonstrate that the federated model achieves performance comparable to centralized training while consistently improving upon the locally trained models. These findings demonstrate the potential of federated learning for collaborative skin lesion segmentation without requiring centralized aggregation of medical images.
arXiv:2610.08554v1 公告类型：交叉收录 摘要：尽管人机交互（HCI）领域日益关注青少年的 AI 安全，但现有文献仍缺乏全面视角，未能系统呈现已识别的风险、应对这些风险的方式，以及所提出的保护措施在实践中是否有效。我们系统回顾了 100 项实证性 HCI 研究，涉及在学校、家庭、照护场所和公共服务中与 AI 互动或接触 AI 的儿童与青少年。我们采用 YAIR 风险分类体系和 MIT Mitigation Taxonomy 应对措施分类体系，梳理了已识别的风险、每项风险是否有相应的应对措施，以及针对该风险的每项应对措施是否已实施乃至接受评估。风险与应对措施的映射表明，大多数风险仅对应于已提出或处于构想阶段的应对措施；已实施的应对措施很少，接受过评估的则更少；而且，现有评估往往衡量的是技术性能，而非避免伤害的保护效果。我们指出了尚未得到覆盖的领域以及仍未经检验的保障措施，并为 HCI 研究提出了加强青少年 AI 安全的具体方向。
arXiv:2610.08312v1 公告类型：交叉发布 摘要：持续学习（CL）对于大语言模型（LLMs）依次适应不断演变的任务至关重要。为缓解灾难性遗忘，近期研究通过引入正交投影的低秩适应方法（如 O-LoRA）来隔离任务参数。然而，我们揭示了这种严格的几何约束会引发“正交性困境”：刚性的参数隔离阻碍了共享表征在语义相关任务之间的迁移与积累。在本研究中，我们提出了一种用于大语言模型持续学习的新型无回放方法，称为 Consolidation and Decoupling LoRA（CoDe-LoRA）。CoDe-LoRA 将学习过程解耦为通用知识巩固和任务特定知识解耦两个部分。为此，CoDe-LoRA 利用自适应零空间投影机制和语义路由，在知识积累与任务特定适应之间取得平衡。在四种骨干模型和三个持续学习基准上的实验结果表明，CoDe-LoRA 取得了最佳平均准确率。我们的代码已发布于 https://github.com/Estrellajer/CoDe-LoRA。
arXiv:2610.08268v1 公告类型：交叉收录 摘要：泛在智能应用正日益广泛地部署在移动设备和物联网（IoT）边缘设备上。因此，大语言模型（LLM）也越来越多地用于支持这些应用。然而，由于资源需求较高，LLM 大多部署在云端。逐层划分的边缘—云端推理使资源受限的边缘设备能够为无法完整承载的 LLM 贡献计算能力。不过，异构切分点带来了两种相互耦合的低效现象。首先，边缘端执行和通信会在云端调用之间产生空闲间隙。其次，到达不同模型深度的请求无法按传统方式进行批处理。我们提出了 DySCo，这是一种协同运行时，可将 KV 缓存保留在本地，并引入 dyForward——一种模型感知的层区间执行器，无需重新加载权重，即可从常驻模型分片中执行可配置的连续层区间。针对多边缘设备服务场景，我们引入了深度同步批处理（DSB），将异构请求推进至最深的切分点，并对其共有的后缀部分进行批处理。在异构设备、两个模型系列以及本地和广域链路上的实验表明，即使排除等待时间，空闲间隙仍会增加后续 GPU 前向调用的延迟；在我们的测量中，每个解码步骤由此增加的云端后缀延迟最高达 25 毫秒。在平均并发数为 8 时，DSB 相较于 FIFO、精确匹配批处理和轮询交错执行，吞吐量分别提高了 275%、48% 和 79%，同时降低了平均单会话延迟。这些结果共同表明，采用不同边缘—云端切分点的请求能够复用云端常驻权重，并共享批处理后缀计算。本研究的实验制品仓库已公开，地址为：https://github.com/Large-scale-Sustainable-Computing-LSC/dysco-artifact
arXiv:2610.08183v1 公告类型：交叉收录 摘要：多任务操作策略在架构、规模和预训练先验上同时存在差异，因此已发表的比较无法将性能归因于任何单一组件。我们认为，性能主要来自视觉表征，而参数规模和生成式先验在很大程度上只是次要因素。为验证这一点，我们构建了 CoRP（Compressed Representation Policy），这是一种刻意设计得紧凑的策略（4890 万参数，不使用视觉语言模型，也不使用视频生成先验），由表征提取器和基于流匹配的动作生成器组成。它在 LIBERO 上达到 97.0% 的成功率，在 RoboTwin 2.0 Clean/Randomized 上分别达到 75.78%/73.36%，与规模为其 40.9–163.6 倍的系统相当。随后，我们固定动作生成器，每次仅改变提取器的一项属性。预训练初始化具有决定性作用：使用随机初始化的 ViT-S/14 时，LIBERO 上的成功率降至 78.1%；使用经 ImageNet 预训练的 ResNet-34 时，则降至 74.5%。仅有预训练还不够，因为冻结编码器会导致成功率下降 19.8 个百分点。压缩同样重要：将每个视角重采样为 48 个词元，效果优于传入全部图像块词元（97.0% 对比 83.2%）；而对这些词元施加变分信息瓶颈，效果不如硬性限制词元数量，因为它抑制了策略所依赖的、随指令而变化的词元选择，使 LIBERO-Goal 上的成功率从 95.8% 降至 33.0%。语言条件仅在观测无法明确目标时才有贡献（LIBERO-Goal：从 9.2% 提升至 95.8%）；而在观测不存在歧义的 RoboTwin 2.0 上，移除语言条件反而略微提高了成功率。因此，我们认为，紧凑型策略能够奏效的关键在于其表征经过预训练、针对任务适配并得到压缩。项目页面：https://corp-policy.github.io/
arXiv:2610.08176v1 Announce Type: cross Abstract: Decentralized learning is highly sensitive to communication topology under non-IID data. Adaptive peer-selection methods can exploit local model information, but broader peer discovery may require increasingly large control state, whereas direct spectral optimization typically relies on graph-wide information. We study the intermediate setting of bounded local topology search and propose Local-First Heuristic Evolution (LFHE), a representation-driven rewiring framework whose candidate discovery and scoring use only ego-neighborhood and friend-of-a-friend (FoF) information. The structural score admits an exact interpretation through graph Dirichlet energy: its sum across clients equals twice the representation Dirichlet energy, which under standard linear consensus dynamics governs the instantaneous dissipation of representation disagreement. LFHE combines this state-dependent structural signal with early exploration and degree control, while algebraic connectivity remains an offline graph diagnostic. Under bounded sparse degree, its FoF candidate state remains local rather than expanding toward population-wide peer tracking. Across four image, speech, and text benchmarks, LFHE achieves competitive decentralized learning performance. Matched-protocol controls identify the structural term as the principal empirical topology-selection signal, while comparison with broader peer discovery exposes a trade-off between predictive performance and discovery-state locality. Together, these results motivate state-aware bounded local topology search between pairwise peer selection and globally informed topology optimization.
arXiv:2610.08132v1 公告类型：交叉发布 摘要：概念漂移威胁着生产环境中的机器学习，但多变量双样本漂移检测器在大规模场景下的实际表现仍缺乏充分研究。现有基准很少涉及工业运营数据集中常见的数亿行数据和高基数特征。我们在三个互补环境中评估了五种多列双样本检验方法（边缘方法、基于投影的方法和核嵌入方法）：Harvard Dataverse、经过验证的 Failing Loudly 复现（平均绝对误差介于 0.030 至 0.053 之间），以及基于 Trendyol 的 1.375 亿行集合排序特征表构建的新型合成注入基准。通过在两种严重程度与影响范围组合设定下测试四类漂移，我们证明，在 Apache Spark 上结合随机傅里叶特征的分布式最大均值差异方法具有稳健的可扩展性。在强漂移设定下，采用经过校准的阈值并对四类漂移取平均后，该方法与预期漂移幅度的 Pearson 相关系数达到 r = 0.940，真阳性率为 80.4%，假阳性率为 3.2%。相比之下，逐维 Kolmogorov-Smirnov 检验因非对称抽样下类 ID 列导致的统计量饱和而失效，这揭示了大规模抽样设计的一项关键约束。在弱漂移配置下（实际翻转比例不超过 0.57%），各检测器难以实现可靠区分，这凸显了未来开展基于强度网格的统计功效分析的必要性，以区分根本性的灵敏度界限与随规模变化的阈值偏移。
arXiv:2610.08112v1 Announce Type: cross Abstract: Path-following control strategies typically follow the bi-objective optimization dilemma: minimizing deviations from a reference path while maintaining smooth speed profiles. The latter objective is especially relevant for Electric Vehicles (EVs), since their limited driving range can be extended by recovering energy through regenerative braking, a feature that has not yet been sufficiently studied in the literature. In this work, we perform a comparative analysis of four controllers under one common Frenet frame-based kinematic vehicle model, utilizing a validated energy model (VT-CPEM) with explicit regenerative braking. Herein, we implement the following controllers: Nonlinear Model Predictive Control (NMPC), Proximal Policy Optimization (PPO), gain-scheduled Ackermann state-feedback baseline (PID-SF), and a Stanley geometric baseline. To satisfy real-time requirements, we implement the NMPC using JIT-compiled CasADi. Moreover, we train the PPO using traditional straight and S-curve tracks, after which we successfully transfer the unmodified policy to unseen tracks, including: an ISO 3888-1 lane-change, a chicane, randomly-generated parameterized-splines, and a $\pm3^\circ$ graded road. In addition, the policy transfers to a dynamic single-track vehicle model with linear tires, zero-shot with an acceptable initial performance, which was optimized after brief fine-tuning. Thereby, we demonstrate that our PPO is readily transferable to more comprehensive vehicle models. We conclude with a performance analysis of developed controllers and discuss ideas for future work.
arXiv:2610.08078v1 Announce Type: cross Abstract: Standard causal identification methods often assume no unmeasured confounding and can fail when relevant confounders are unobserved. Proximal causal inference instead uses proxy variables to identify effects under hidden confounding. However, nonparametric proximal estimation can be challenging in practice: recovering causal estimands such as the conditional average treatment effect (CATE) requires solving an ill-posed integral equation that is data-hungry, hyperparameter-sensitive, and optimization-unstable. Bayesian inference for such models provides a desirable alternative, mitigating these difficulties by regularizing through the prior. However, computing a posterior is itself challenging, as a typical likelihood function will include latent variables. Following the recent success of tabular foundation models in backdoor, instrumental variable, and frontdoor settings, we propose that prior-data fitted networks (PFNs) are uniquely suited to resolve this bottleneck. Indeed, by training on synthetic data sampled from compliant structural causal models with access to oracle counterfactuals, we simplify the task substantially, amortizing the implied Bayesian operator inversion into a single transformer forward pass. Compared to prior literature that focuses primarily on point estimation, our model, ProximalFM, explicitly targets the Bayesian posterior distribution of the CATE. One unique aspect of this problem is that we need to provide Monte Carlo estimates of the oracle CATEs, leading to a novel variation of PFNs that accounts for the added stochastic error. Across a diverse suite of proximal regimes, ProximalFM achieves consistently strong CATE-estimation performance without dataset-specific tuning, with its largest advantage when latent confounding is substantial and the proxies are weakly informative; it also provides fast inference through a single amortized forward pass.
arXiv:2610.08048v1 公告类型：交叉收录 摘要：LLM 智能体往往缺乏在新环境中可靠行动所需的操作知识，因为它们必须自行探索工具的具体行为或环境约定。由于没有对过往尝试的记忆，它们会在不同任务中重复相同的错误，导致更多任务失败和更长的执行轨迹。为解决这一问题，智能体系统通常依赖人工编写的指南，或依赖利用训练任务和预言机验证器构建的程序性记忆，而这两种方式都需要对环境具备先验知识。我们提出 DAEDALUS，一种无需现有任务或预言机验证器、通过自生成练习引导构建可复用智能体记忆的方法。DAEDALUS 将两个智能体配对：探索者与环境交互，生成具有挑战性但可解决的任务；求解者则尝试完成这些任务。系统从求解者的每次失败中提炼一条启发式规则，并且只有在求解者将该规则纳入上下文后反复成功完成任务时，才接纳该规则。这些结果也为探索者提供反馈，以调整后续任务的难度。随后，接纳的启发式规则被整合到记忆库中，供测试时使用。在 AppWorld、$\tau^2$-bench 和 AutomationBench 上，与无记忆基线相比，DAEDALUS 将平均成功率最多提高了 15.9 个百分点，将 pass^5 最多提升至 2.2 倍；其表现可与使用训练任务的方法相媲美，同时推理成本低于其中大多数方法。我们表明，即使探索预算较小，也已能获得性能提升，而且其启发式规则同样能使其他模型家族的智能体受益。消融实验进一步表明，求解者的执行轨迹提供了提炼有效启发式规则所需的关键信息，而将早期发现分解为独立要素则能提高探索的成本效益。除记忆构建外，我们还发现，在按性能对模型进行排名时，DAEDALUS 生成的任务可以作为基准测试任务的替代。代码与相关产物：www.github.com/illuin-tech/daedalus。
arXiv:2610.07894v1 公告类型：交叉发布 摘要：大型语言模型（LLM）应当忠实地依据所提供的上下文回答问题，并在上下文信息不足以回答问题时选择不作答。现有的忠实性评估通常孤立地考察每个问题—上下文实例；然而，这种实例级评估无法捕捉忠实行为的一项基本要求：根据可用上下文的变化调整模型响应的能力。具体而言，模型应在证据充分时给出正确答案，在证据不足时不作答。在本研究中，我们提出了 Pairwise Faithfulness Benchmark（PFaithBench），用于评估模型能否针对同一个问题，在提供支持证据与不提供支持证据的上下文之间切换作答与不作答。我们对七个模型家族中的三十九个模型进行了评估，结果表明，忠实性从根本上涉及作答与不作答之间的权衡，而当前大多数模型都表现出强烈的作答倾向，大多数忠实性错误源于过度作答，即模型即使在所提供的上下文不足时也倾向于编造回答。我们进一步开展了一系列研究，考察不同数据构建方式下的忠实性训练。结果表明，训练效果对作答数据与不作答数据的具体组成高度敏感。从不匹配的来源构建作答与不作答数据，可能导致模型依赖特定数据集的捷径，而非判断上下文是否确实充分。此外，增加答案监督数据可以提升作答表现，但会加剧过度作答；增加不作答数据则可以减少幻觉，但会导致过度不作答。代码和数据已发布于 https://github.com/tmlr-group/PFaithBench。
arXiv:2610.07816v1 Announce Type: cross Abstract: Small language models (SLMs) are attractive as local agent controllers because they reduce remote inference, latency, and deployment footprint, yet structured tool errors can cause an agent step to fail. Existing routers typically select a model once per query. However, agents expose sequential decision points whose difficulty dynamically changes based on intermediate observations. We propose STEPGATE, an uncertainty-aware handoff framework that scores each local SLM action and selectively escalates challenging steps to a stronger model. On a 52-task held-out single-step BFCL-derived test split, the Qwen2.5-1.5B/7B pair attains 82.7% task success with 30.8% escalation, versus 67.3% local-only and 75.4% random escalation (which uses 33.8% escalation). In a separate multi-turn evaluation, STEPGATE achieves 69.0% trajectory success and 84.0% action success using only 30.0% cloud actions, compared with 48.0%/70.5% local-only, 60.0%/78.2% random escalation, and 57.0%/77.1% query-level routing (strong-only achieves 82.0% trajectory success at 100% cloud actions). These results suggest that step-level escalation recovers a large share of the performance gap to the stronger Qwen2.5-7B backend at a matched cloud-action rate while transmitting fewer tokens remotely. However, our evaluation is limited to one model family, a single stronger backend, and scripted tasks. Furthermore, the test sets are small, multi-turn comparisons rely on paired intervals and statistical tests, and our risk tiers serve as research annotations rather than formal safety guarantees.
arXiv:2610.07814v1 Announce Type: cross Abstract: How far can stochastic gradient descent ascent (SGDA) go by tuning its timescale ratio and step sizes in nonconvex min-max games? We answer this question for nonconvex-PL (NC-PL) games by establishing the first tight complexity of two-timescale SGDA with a fixed timescale ratio and non-increasing step sizes. For $\ell$-smooth games with an inner $\mu$-PL inequality, we prove a complexity lower bound $\Omega(\kappa^2\ell\varepsilon^{-2}+\kappa^4\ell\sigma^2\varepsilon^{-4})$, where $\kappa=\ell/\mu$ is the condition number, $\sigma^2$ is the gradient variance, and $\varepsilon$ measures the outer gradient norm. This matches existing SGDA upper bounds and establishes a complexity separation from Smoothed-AGDA (Yang et al., 22'). In addition, we show that SGDA can fail to find a stationary point when its timescale ratio is as small as $o(\kappa^2)$. Our negative results highlight the fundamental limitation of SGDA in NC-PL games, and justify the development of alternative methods.
arXiv:2610.07808v1 Announce Type: cross Abstract: Spiking neural networks (SNNs) offer sparse and event-driven computation, making them attractive for energy-constrained reinforcement learning (RL) on edge devices. In value-based RL, deep spiking Q-networks (DSQNs) combine such efficiency with action-value estimation for decision making. However, existing DSQNs often require multiple simulation timesteps for competitive performance, increasing computational and energy costs, whereas reducing the timesteps can cause substantial performance degradation. We investigate this degradation from the perspective of Q-value estimation errors. By decomposing errors across actions into common-mode and differential-mode components, we find that low-timestep DSQNs suffer disproportionately from common-mode errors shared across action values, which are particularly detrimental to temporal-difference learning through bootstrapped targets. Based on this finding, we propose Common-Mode Compensation Deep Spiking Q-Network (CMC-DSQN), which uses an auxiliary ANN to compensate for common-mode errors in the SNN outputs. At inference, greedy action selection can be performed directly from the SNN outputs, allowing the auxiliary ANN to be completely removed and preserving the energy efficiency of SNNs. Extensive experiments on Atari and MiniAtar environments demonstrate substantial performance improvements under low-timestep settings. CMC-DSQN outperforms state-of-the-art DSQN baselines by nearly $20\%$ at $T=2$ and further surpasses the ANN baseline at $T=4$.
arXiv:2610.07712v1 Announce Type: cross Abstract: Existing molecular and materials learning approaches often rely on a limited set of structural representations, which may capture only selected aspects of complex three-dimensional structure. Here, we introduce mathematical invariant-enabled topological neural networks (MITNNs), a framework that represents complex structures through multiple complementary mathematical views and integrates them with topological neural architectures. MITNNs combine multiscale invariants from topology, spectral theory, commutative algebra, differential geometry, and discrete curvature, capturing complementary structural information from the same system. Systematic invariant-subset, architecture-subset, and ensemble analyses show that predictive performance depends on how mathematical representations and neural architectures are paired, with selected combinations outperforming individual models and the aggregation of all available components. Across protein-ligand binding, metal-organic framework properties, mutation-induced protein solubility, and molecular toxicity prediction, MITNN consistently outperforms existing methods. These results establish MITNN as a mathematically multimodal framework for scientific machine learning.
arXiv:2610.07637v1 Announce Type: cross Abstract: Many real-world datasets exhibit unusually large values far more frequently than predicted by Gaussian models. Heavy-tailed distributions capture this behavior, yet evaluating learning performance under them remains challenging because rare, large feature entries retain non-vanishing effects even in high dimensions. Even in the canonical setting of empirical risk minimization for linear regression with entry-wise i.i.d. symmetric $\alpha$-stable data, a precise asymptotic characterization of prediction has been lacking. In this work, we introduce a functional order parameter that describes the random effective problem associated with each coefficient. Using the replica method, we fully characterize the generalization error in the proportional high-dimensional limit where the sample size and feature dimension diverge at a fixed ratio. Additionally, this analysis establishes a heavy-tail universality law, scaling laws relating typical errors to prediction reliability, and the Bayes-optimal prediction error. In addition to characterizing the effects of extreme entries on the learning process, our method applies broadly to other systems with persistent local heterogeneity.
arXiv:2610.07620v1 公告类型：交叉收录 摘要：科学定律发现需要选择测量方案，并将证据转化为支配方程。我们评估了一种先探索、再定论的流程：由大语言模型提出假设，程序化规划器收集测量数据，再通过全新的提示词，基于固定的观测结果综合得出最终定律。该流程结合了结构化探测、自动数值诊断、受限的测量批次以及可选的解释器访问权限。在 576 次 NewtonBench 试验中，我们使用 GPT-4.1-mini，并通过 GPT-4.1 在中等难度任务上进行复现实验，对比了八种配置在 12 个物理模块上的表现。在中等难度任务中，启用解释器的规划器使 GPT-4.1-mini 每次试验的测量次数从 22.5 次降至 8.6 次，使 GPT-4.1 的测量次数从 43.0 次降至 8.9 次。两者基于幅值的均方根对数误差均值分别从 2.514 降至 0.202，以及从 0.626 降至 0.149。一项额外审查在对覆盖率敏感的分析中保留了不完整和无效的提交结果。观测到的符号准确率提升在各模块间并不一致，而随机采集策略的表现可与分歧评分策略相媲美。所有模块的测量次数均有所减少，但由于批次约束不一致，无法将这些减少完全归因于采集质量。这些结果表明，完整流程是一种有前景的高测量效率配置，但其各组成部分的因果作用，以及在无噪声直接方程任务之外的泛化能力，仍有待明确。
arXiv:2610.07594v1 公告类型：交叉收录 摘要：人形机器人执行家务操作时，需要双臂在身体保持平衡、迈步和改变姿态的同时进行动作。我们提出 BiGym 2.0，这是面向 Unitree G1 改编的 BiGym，涵盖 20 项家务任务，并使用统一的全身控制器进行演示和评估。该套件为每项任务提供 60 次原生人类虚拟现实演示，包含同步的多摄像头视图和全身执行记录。我们对视觉-语言-动作微调、模仿学习、演示驱动的强化学习，以及获得与在线强化学习相同交互预算的冷启动编程智能体进行基准评测。在所有方法均使用相同的机载视图、本体感知和全身控制器的条件下，视觉-语言-动作微调在九项任务上的平均表现最佳；智能体开发的程序在这一平均指标上优于所有演示驱动的强化学习基线，并在双臂到达任务上领先。跨工作区堆叠仍是尚未解决的难题，$\pi_{0.5}$ 在 pick-box 任务上的表现依然较低，而多物体搬运对模仿学习、演示驱动的强化学习和编程智能体而言都很困难。所有环境、人类演示和评估轨迹均已开源，地址为 https://github.com/swirl-uk/BiGym2。
arXiv:2610.07585v1 公告类型：交叉收录 摘要：我们提出了一种可扩展的旋转反射群等变视觉 Transformer，其基于窗口化群卷积自注意力和分层特征架构。我们证明，该方法可以扩展至具有数百万参数的群等变视觉 Transformer（ViTs），并应用于包含实际应用尺寸图像的大规模数据集，即 ImageNet。所提出的分层窗口化旋转反射等变 ViTs（REViT-v2）的代码和预训练权重可在 https://github.com/kc-ml2/revit 获取。
arXiv:2610.07511v1 公告类型：cross 摘要：移动操作机器人（如人形机器人）正越来越多地被部署于动态、非结构化环境中，以执行灵巧操作任务。然而，用从单一机器人位姿采集的示范数据进行模仿训练得到的端到端操作策略十分脆弱：部署时即使机器人位姿出现厘米级偏差，也可能使以自我为中心的观测和末端执行器轨迹偏离训练数据分布，导致性能大幅下降。我们提出 MobileVISTA，这是一种数据生成框架，通过联合执行以下两项操作，将在规范位姿下采集的示范转换为多样化的、经过位姿扰动的训练数据：（1）增强第一视角视觉观测；（2）重新映射动作，以补偿基座位姿变化。不同于以往方法假设相机刚性安装在驱动链之外，或假设复杂的关节式机器人几何结构大部分位于画面之外，MobileVISTA旨在适配第一视角平台（如人形机器人）：在这类平台中，相机会受到机器人运动学链运动的影响，同时也必须观测该运动学链。我们在涵盖人形机器人和双臂机器人形态的仿真任务中，以及真实的 Galaxea R1 Pro 上研究了 MobileVISTA。我们发现，使用 MobileVISTA 增强数据训练的策略对测试时遇到的、此前不属于训练分布的位姿表现出更强的鲁棒性，且无需额外采集示范数据或训练生成模型。此外，我们发现 MobileVISTA 在所测试的人形机器人上的收益最大，这是因为相机安装在驱动链上，且机器人占据了画面的大部分区域。更多视频和附录请见我们的网站：https://mobilevista.github.io
arXiv:2610.07503v1 Announce Type: cross Abstract: Two populations of graphs often have to be compared without any correspondence between their vertices, for instance when networks come from different communities, or when a graph generative model is evaluated against held-out graphs. We study how many graphs such an unaligned two-sample test needs, and which graph statistics can detect which differences. For an Erd\H{o}s--R\'enyi null and a planted two-block difference that leaves every expected degree unchanged, we show that $m\asymp t^{-3}$ graphs per group are necessary and sufficient when the per-graph signal-to-noise ratio is $t<1$. Signed triangle counts attain this rate, and the lower bound holds for every graph size. With aligned vertices $m\asymp t^{-1}$ graphs suffice, so misalignment costs a factor of order $t^{-2}$. When the triangle signal cancels, the rate becomes $t^{-4}$ and $4$-cycles are needed. Statistics built from trees have exactly the same expectation under both hypotheses, and tests based on finitely many of them have asymptotically no power. In the graphon limit, this class includes degree distributions and message-passing graph neural network features. For a non-constant null, a generic difference is visible at first order, and a simple motif test attains the aligned order of sample size, suggesting that misalignment is costly mainly for differences that are invisible at low orders. We also give an exactly valid test for one or two graphs per group, at a cost in power. In our simulations, the fitted exponents are close to the predicted ones, and degree-based and random-GNN evaluation metrics stay at their level in a setting where signed triangles need about $65$ graphs.
arXiv:2610.07269v1 公告类型：cross 摘要：跨视角地理定位通常被视为图像检索问题，通过联合训练的嵌入，将地面图像与卫星图块数据库进行匹配。这类模型准确率较高，但需要大量配对监督，也无法展示支持匹配的证据。在本文中，我们研究了一个不同的问题：仅通过语言，这项任务能够完成到什么程度？我们提示多模态大语言模型（MLLM）将每个地面全景图和每个卫星图块描述为结构化文本，并通过比较这些描述进行定位。整个过程中没有训练任何组件。我们在来自美国四座城市的9,826个VIGOR配对样本上，针对三种设置进行了评估。第一种设置中，描述忠实但缺乏区分度。两种视角下的描述高度一致，但按照描述相似度对整个候选池进行排序，几乎从未返回正确图块（Recall@1为0.39%）。第二种设置中，我们将候选池缩小至十个相邻图块，就像粗略先验所做的那样。此时，相同的描述变得有用：一个依据结构一致性进行评分的MLLM评判器使随机排序的表现提升一倍，并达到强词法基线的水平。它还会指出两份描述中的哪些字段一致、哪些字段冲突，这是嵌入距离无法做到的，而我们认为这向可解释定位迈出了一步。第三种设置中，我们将该评判器应用于一个经过训练的视觉检索器。在它排序错误的查询上，基于图像进行重新排序是有效的，而基于我们的描述进行重新排序则无效（Recall@1分别为23.5%和10.7%）。场景结构在转换为语言后仍然保留，但区分相邻地点所需的细微外观信息却没有保留下来。代码和提示词已公开于https://github.com/AyeshAbuLehyeh/GeoLingual。
arXiv:2610.07211v1 Announce Type: cross Abstract: Systems of equations arise in a wide range of scientific and engineering applications. The present work focuses on solvers for general systems of equations, including but not limited to those arising from partial differential equations. These systems can be broadly categorized into linear and nonlinear problems. For large linear systems, iterative solvers are generally preferred over direct methods due to the latter's superlinear growth of computational costs. Although convergence theory is well-developed under certain assumptions on the coefficient matrix, many classes of systems still pose open challenges. These difficulties become even more severe for systems of nonlinear equations, where nonlinear solvers typically rely on repeated linearization. For example, Newton's method may even converge quadratically near the solution; it can also converge slowly or diverge when the initial guess is not chosen appropriately. A wide range of solvers with diverse variants and hyperparameter settings exists, and the development of efficient and robust iterative methods remains an active area of research. Recently, machine learning (ML) techniques have been applied to enhance the efficiency of classical iterative methods while preserving their interpretability and reliability. We refer to these ML-enhanced iterative methods as hybrid iterative methods, in the sense that they combine classical iterative methods with ML. This paper provides a comprehensive overview of state-of-the-art approaches to constructing hybrid iterative methods for systems of both linear and nonlinear equations, while also discussing open challenges and outlining potential directions for future research.
arXiv:2610.07167v1 发布类型：cross 摘要：我们提出了一种用于神经网络控制策略的模仿学习设计，以处理状态和输入约束。训练过程交替进行标准的模仿梯度步骤和由 \(k\) 个安全步骤组成的模块，将网络的动作拉向其在安全集上的投影；在运行时，控制器仅为训练得到的网络，不再使用安全过滤器。我们将该方案分析为策略动作空间中的非精确投影梯度下降。当在每个安全步骤重新计算投影动作，且每一步都使动作持续向安全集靠近时，令 \(k\) 以对数速度增长，可使训练状态上的约束渐近满足，并界定动作到模仿损失约束最优解的距离；当投影动作保持不变时，只有在这些动作能够由网络精确表示的情况下，才能得到同样的结论。在一项非线性自动驾驶竞速任务中，我们将该方法与在模仿损失中加入加权约束违反惩罚的方法进行了比较。当权重足够大时，我们的方法在将违反约束的回合比例从 \(15\%\) 降至 \(1\%\) 的同时，达到了与无约束模仿相同的单圈时间；相比之下，这一比例约为惩罚方法在其最佳权重下的六分之一。我们方法的单圈时间对权重不那么敏感，而权重主要决定训练过程中违反约束的消失速度。在竞速任务中，安全修正是稀疏的，分析所需条件并不成立；性能提升反而来自训练期间收集的数据。这些提升的代价是增加了训练计算量。
arXiv:2610.07084v1 公告类型：cross 摘要：不准确的专家在经过修正后仍然可以提供有用信息。我们研究在线延迟决策学习：学习者选择一名专家，并在购买其答案前确定一个修正函数，然后将该函数应用于收到的答案。困难在于，观测到的损失同时反映了专家质量和未完成的修正：早期错误可能会使学习者放弃查询那些在学会修正后本来具有价值的答案。我们提出 ORUCB，该方法汇集共享多项式响应和专家特定的多项式响应。关于累积响应学习误差的界限可以校准置信度加权的风险回归与探索，使路由器在决定购买哪些答案时能够考虑这一误差。在残差和分歧有界、存在一个固定的最优响应可行模型，以及免费风险和最优查询风险均采用线性模型的条件下，对于固定的问题参数，该校准算法在 $T$ 轮中以高概率达到伪遗憾 $O(\sqrt T\log(T+1))$。该保证允许答案分布是奇异的，也允许共享响应存在模型设定错误；最优性是相对于有界响应类而言的。在四个测试流上，所选的三次策略的含费用成本低于七个原样部署答案的基线方法。与一种常见修正学习器的比较考察了路由效果，而六种价格下的比较则衡量了成本和查询率。
arXiv:2610.07074v1 Announce Type: cross Abstract: Estimating a decision threshold requires locating observations near an unknown boundary. We study how gradient-based pretraining learns this statistical rule in a two-parameter softmax-attention model with a fixed feature and inequality direction. Pretraining uses labeled contexts and their true thresholds; a fresh threshold must be inferred from context alone. Under a large-resolution initialization, constant-step gradient descent on $m$ tasks with $n$ examples each produces a frozen estimator with error $\widetilde O((m\wedge n)^{-1}+N^{-1})$ for each fixed interior threshold and every fresh-context size $N$. The two terms separate finite-pretraining accuracy from fresh-context localization. The mechanism is coordinated parameter divergence: population training calibrates the relative label and feature scores, then increases the attention scale as $t^{1/4}$, giving population threshold error $O(t^{-1/4})$. To transfer this mechanism to a fixed finite corpus, we control gradient errors relative to the shrinking directions of progress at successive parameter scales. This certifies a growing training interval without requiring long-time tracking of the population trajectory. We also identify the boundary limitation of the one-head model and explain statistically what a reflected symmetrization could achieve.
arXiv:2610.07072v1 发布类型：cross 摘要：变分自编码器（VAEs）是现代文本到图像模型的重要组成部分，这些模型在其潜空间中生成图像。众所周知，VAEs 能够解耦数据中的主要变化因素，而在自然图像中，色彩是这些因素中结构最明显的因素之一：对其进行去相关处理，可以得到一个亮度轴和两个对手色轴。因此，可以预期色彩会作为一个独立因素出现在 VAE 潜空间中。然而，这些潜空间如何表示色彩，目前仍鲜有研究。在本文中，我们展示了文本到图像模型中的 VAEs 共享一个与亮度和对手色对齐的色彩子空间。通过对编码器进行线性近似并实施有针对性的潜变量引导，我们在从 SD1.5 到 FLUX.2 和 Z-Image 的广泛 VAEs 中，一致地发现了这一子空间。在此表征的基础上，我们提出了三种应用：ColorTuning，在 GenColorBench 的细粒度 CSS3/X11 色彩系统上实现了精确数值色彩生成的当前最佳性能；饱和度控制，用于调整全局色彩强度；以及色彩迁移，用于改变色板以匹配参考图像。代码和模型已公开发布于 https://julian075.github.io/Color_Subspace/
arXiv:2610.07061v1 发布类型：cross 摘要：冲击声音渲染旨在合成敲击 3D 物体时产生的声音，但实际应用中的渲染器通常依赖木材、塑料或钢材等固定材料预设。这些预设限制了渲染器能够表现的冲击声音范围，而在没有材料声学专业知识的情况下，手动调整底层材料参数仍然十分困难。因此，我们研究逆向冲击声音渲染：根据参考冲击声音预测材料参数，使模拟器能够重现相似的材料响应。为支持这一任务，我们提出了 ImpactMat，这是一个包含单一材料和混合材料冲击声音及其真实材料参数的数据集与基准。我们进一步提出了一种前馈模型，可以根据一段或多段录音预测这些参数，并利用混合材料学习不同材料类型之间的平滑过渡。实验表明，我们的方法优于具有竞争力的基线方法，并能够在无需手动调参的情况下根据真实录音重新渲染。项目主页：https://material-from-impact.github.io/material-from-impact/。
arXiv:2610.07048v1 Announce Type: cross Abstract: We study errors-in-variables problems in which a target study contains only a single error-prone surrogate of an unobserved exposure, while an external source study provides repeated surrogate measurements from a different population. The measurement error distribution is allowed to depend on the observed error-free variables, and the error-free variable distribution itself may differ between studies. We introduce a conditional transportability assumption that enables the use of external repeated measurements under source-target heterogeneity. Together with additional replicate-error conditions, it identifies the target conditional measurement-error distribution. Building on this identification result, we develop a data-fusion estimator for a broad class of target functionals. The estimator combines conditional deconvolution, flexible nuisance estimation, and orthogonal correction that reduces first-order sensitivity to nuisance estimation. For the proposed estimator, we develop a unified spectral theory covering both diffuse-spectrum and finite atomic-spectrum target functionals, derive a general asymptotic expansion, and establish consistency and target-specific convergence-rate bounds. The resulting convergence-rate bounds depend jointly on the spectral properties of the measurement error, the latent exposure, and the target functional. For finite atomic-spectrum targets, we further establish joint Gaussian and bootstrap limits, yielding inference for smooth moment transformations under an additional centering condition. In the reported simulations, Fuse-EIV has small bias for the primary exposure-related coefficient. Applications to the National Health and Nutrition Examination Survey illustrate how accounting for population heterogeneity and error heteroscedasticity can change empirical conclusions.
arXiv:2610.07032v1 Announce Type: cross Abstract: Large-scale pretrained transformer models have achieved state-of-the-art performance across diverse machine translation tasks, including multilingual settings. Knowledge distillation has emerged as a sustainable approach for model compression, transferring knowledge from large teacher models to smaller, more efficient student models. Similarly, quantization, which reduces the numerical precision of model weights and activations (e.g., from 32-bit to 8-bit representations) is widely used to accelerate inference, enabling models to run several times faster during deployment. However, both techniques face limitations when applied to specialized domain data, particularly under low-resource conditions. In knowledge distillation, the effectiveness of transfer is often constrained by the scarcity of domain-specific parallel data, while quantization can lead to performance degradation as bit precision decreases. In this work, we investigate the combined application of knowledge distillation and quantization for French-to-English biomedical translation, a domain characterized by specialized terminology and limited parallel resources. We develop and compare multiple fine-tuning strategies to adapt compressed student models to this challenging setting. Our experiments demonstrate that a collaboratively distilled and quantized student model achieves a 69% reduction in size, a 98.21% increase in inference speed, and a 98.46% reduction in CO2 emissions compared to the original baseline all without sacrificing translation quality. These results indicate that jointly optimized compression techniques can yield efficient, high-performance models suitable for translation service providers operating under resource constraints.
arXiv:2610.07019v1 公告类型：交叉收录 摘要：Fiorillo v0.5 是一个开放模型，可回答类型化问题，并为每个答案给出概率。其主要专用模块读取截取至6,144个词元的随机试验论文，并回答与对照相比，某项干预是否显著增加、显著减少或未显著改变某个结局（Evidence Inference 2.0，EI）。该模型基于 Qwen3-4B-Base，配有低秩适配器和决策头，仅使用2,657篇训练论文中自身许可证允许再利用的1,431篇进行 EI 微调。在生成此版本的测试预测之前，研究者在 Open Science Framework 上注册了四项标准，以决定是否发布模型，其中第二项门槛在标签公开的 EI 测试集上评判。在该测试集（333篇论文中的1,218条提示）上，预期校准误差为0.0168，低于0.05的上限；对数损失比先验基线低0.8603（95%区间为0.8104至0.9078），比读取相同输入的 Gemma 4 31B-it 低0.1829（0.1164至0.2598）；宏平均 F1 为0.9248，对比值为0.8668，因此四项标准全部通过。仅在许可合规的论文上采用相同方案训练，使准确率下降了0.0123（0.0034至0.0207；描述性结果）。不提供论文时，宏平均 F1 降至0.4384；仅提供标题可使其提高0.0939（0.0655至0.1234），这可能是因为标题陈述了结果，或模型记住了该试验；交换干预与对照后，其方向性回答中有0.6652发生反转。按发布配置运行时，发布文件生成的预测与评估时的预测之间的差异处于预先设定的范围内。本次发布采用 Apache License 2.0（数字对象标识符：10.57967/hf/10722）。
arXiv:2610.06903v1 发布类型：cross 摘要：激活引导通过干预大型语言模型的内部激活来操控其行为，但这些干预背后的机制基础仍缺乏充分理解。我们在四个开放权重模型中，将拒答引导分解为组件级干预，识别出仅对注意力和 MLP 组件中的稀疏子集进行引导，便足以复现完整行为效果。我们发现，拒答方向集中于由上游组件中 28--48\% 构成的稀疏组件机制中，同时保留了 88--101\% 的引导效果。在这些机制内部，有效引导进一步集中于残差流维度中约 50\% 的维度，并保留了组件机制基线效果的 85--98\%，这与一种特权基结构相一致。因此，稀疏性作用于两个层面：被引导的是哪些组件，以及这些组件中的哪些维度承载了信号。这些发现共同表明，拒答并非弥散地编码于 Transformer 中，而是由一种结构化且可识别的机制组装而成，为理解拒答行为如何被表征和引导奠定了基础。为便于复现，我们在 https://github.com/wang-research-lab/Refusal_Mechanisms 发布了全部代码和原始实验结果。
arXiv:2610.06894v1 Announce Type: cross Abstract: A central question in the prediction of stochastic processes is the extent to which past information can improve the probability of correctly predicting the next state. We introduce the Memory Prediction Excess (MPE) to address this question quantitatively. The MPE measures the average improvement in prediction accuracy obtained by using the entire observed history relative to using only the static marginal distribution, in discrete-time finite-state processes. It is defined as the difference between the expected optimal conditional prediction accuracy and the optimal static prediction accuracy. Its basic properties are examined: the MPE is always non-negative; it admits an upper bound depending on the static accuracy, attained if and only if the future is almost surely a deterministic function of the past; and degenerate cases in which the MPE vanishes are characterized. A normalized version, taking values in the unit interval, is introduced as a dimensionless measure of predictive efficiency. A lower bound is derived by comparing predictions based on histories of different lengths, showing that the expected optimal prediction accuracy is monotone with respect to the history length. The framework is extended to finite-length histories, where the finite-history MPE (FH-MPE) measures the predictive gain attainable when only the most recent observations are retained. This leads to the notion of a minimal memory length required to achieve the same predictive performance as the full history. For finite-order Markov chains, this minimal memory length is shown to be bounded by the Markov order. The MPE and its variants are formulated in terms of conditional probabilities and prediction accuracies, offering a probabilistic perspective on the predictive utility of memory that is complementary to classical information-theoretic approaches.
arXiv:2610.06858v1 Announce Type: cross Abstract: This demo paper presents the conceptual foundations and the first steps towards implementation of a novel no-code solution for movement data analytics based on the open-source Python library MovingPandas and the open-source geographic information system QGIS. The resulting Trajectools plugin is available open-source at https://github.com/movingpandas/qgis-processing-trajectory.
arXiv:2610.05518v1 发布类型：cross 摘要：Fr\'echet 距离是跨领域评估生成模型的事实标准，在图像领域表现为 FID，在视频领域表现为 FVD。它用单个标量概括生成分布与参考分布之间的差异，通常将较低的数值解读为更好的生成质量。然而，这种标量视角可能掩盖比较结果的驱动因素。例如，在 COCO 数据集中，增加扩散采样步数会提升 ImageReward 得分，却使 FID 恶化（升高）。受这一不一致现象的启发，我们希望通过揭示差异所在的位置，使 Fr\'echet 距离更具可解释性。为此，我们引入了方向性 Fr\'echet 距离，即最优传输位移在给定方向上的投影平方的期望值。在图像、视频和蛋白质案例研究中，我们发现，少量具有可解释性的方向就能解释距离中的很大一部分。我们利用这些方向，根据 CLIP 嵌入所表示的语义概念解释 FID 的升高，量化 FVD 对逐帧外观的偏向，并重新审视 Protein FID 的解释。我们已在 https://github.com/yhlee-add/directional-fd 开源代码库。
arXiv:2610.08740v1 Announce Type: new Abstract: Learning when the environment does not belong to the learner's hypothesis class is typically handled using agnostic learning guarantees. However, for anything beyond supervised learning, agnostic guarantees are difficult to come by. Recently, imprecise bandits (Kosoy, 2025) (later renamed to robust bandits in Appel and Kosoy, 2025) were introduced as another approach to unrealizable learning in the bandits setting and a $\Theta(\sqrt{T})$ regret learner was shown for a large class. However, no computational guarantees were provided. In this paper we identify a special case that admits a polynomial-time learner with $\tilde{O}(\sqrt{T})$ regret. We also show that several small generalizations of this special case are NP-hard thus indicating that the special case is at the boundary of what is tractable. It has been recently suggested (Kosoy, 2018) that computationally efficient learners for unrealizable learning problems are crucial for solving the AI alignment problem. This work is a small step in that direction.
arXiv:2610.08592v1 公告类型：新 摘要：CNet 是一个用于构建和训练深度复值神经网络（CVNN），以及更一般地通过 Wirtinger（CR-calculus）导数进行梯度下降以优化复值函数的 C++/CUDA 框架。它采取物理原生的立场：网络是作用于振幅向量上的复值操作级联，其中通常还包含酉操作（DFT）；分类则采用 Born 规则测量 $p_k = |z_k|^2 / \|z\|^2$，而不是对实值 logits 使用 softmax。每一层都提供 CPU 参考实现和 CUDA 内核，并通过有限差分进行校验；计算图则会针对批次中的样本进行克隆，以便在 GPU 上执行。在基础层之上，我们加入了信号处理原语，将恒等式 conv(x,k) = IFFT(FFT(x) . FFT(k)) 转化为可学习的复值卷积网络，同时提供 true-Adam 优化器和低内存推理模式。我们报告了三项研究。第一项是一个完全复值、FNet 风格的因果序列模型，建立在一种新的 $O(N \log N)$ 因果 Fourier mixer 之上——该方法通过 Bluestein / chirp-z 分解对三角掩码 DFT 进行计算：经过适当调优后，在字符级语言建模任务上，它能够达到或超过参数量匹配的实值因果 FNet，并在不到一半的训练步数内达到实值模型收敛后的性能。第二项和第三项分别针对无线电调制分类（RML2016.10a）以及相干衍射成像中的 Fourier 相位问题开展瓶颈分析，从而精确定位复值网络仍需开发新型算子的位置。在这三项研究中，复值表述都能够以可证明的方式学习物理上正确的结构。代码：https://github.com/crasmarum/CNet
arXiv:2610.08420v1 公告类型：新 摘要：我们建立了对称性感知与对称无关的特征学习之间的多项式样本复杂度分离。我们研究具有高维 Gaussian 协变量的增长秩多索引模型，其中协变量位于 $\mathbb{R}^d$ 中，并且有 $r=\Theta(d^\delta)$ 个教师方向，这些方向构成一个循环对称轨道，其中 $0<\delta<1/2$。我们比较利用这一结构的三种方式：架构权重共享、在完整对称群上进行数据增强，以及在无法访问对称性信息的情况下进行学习。具体而言，我们使用带相关性损失的球面在线 SGD，分析一个对称性绑定的卷积网络、一个非绑定网络，以及使用完整群数据增强训练的同一非绑定网络。对于一类信息指数为 $p\ge3$ 的多项式链接函数，我们证明了除对数因子外相匹配的样本复杂度界：绑定学习器和增强学习器均能在 $\widetilde{\Theta}(d^{p-1})$ 个样本下实现弱方向恢复，而对称无关学习器则需要 $\widetilde{\Theta}(rd^{p-1})$ 个样本。对于纯二次 Hermite 链接函数，教师子空间的弱恢复同样呈现这种分离，其样本复杂度分别为 $\widetilde{\Theta}(d)$ 和 $\widetilde{\Theta}(rd)$。因此，完整群数据增强能够达到与架构权重共享相同的样本效率，而二者相较于不利用对称性进行训练都具有多项式优势。对于 $p\ge3$，证明揭示了一个两阶段机制：初始化时的波动会从教师轨道中选出一个方向，随后局部增长将该方向与目标的重叠放大到弱恢复尺度，同时其他竞争方向的重叠保持在接近其初始化尺度的水平。
arXiv:2610.08402v1 公告类型：新增 摘要：多轮 LLM 智能体通常在多次交互中获得稀疏的任务反馈，同时逐个 Token 生成每次响应。这带来了两个相关的信用分配问题：哪些响应帮助实现了最终结果，以及每次响应中的哪些生成决策发挥了作用？现有方法通常只关注其中一个层级：回合级方法评估完整响应，却无法区分响应内部的决策；Token 级方法可以跨回合传播反馈，却没有显式地为每个响应建模信用。这些互补的局限性促使我们在两个层级同时学习信用，并在一次策略更新中对其进行协调。我们提出 VETTA，一种通过共享的轻量级评论器中的独立头部，联合学习回合级和 Token 级价值的信用分配方法。VETTA 沿两个时间序列计算优势，并将每个回合的优势与响应内部中心化的 Token 残差相结合，用于 PPO 更新。此外，为降低价值学习成本，评论器仅保留用于初始化 actor 的预训练检查点中的前若干个 Transformer 模块。在两个具有挑战性的智能体基准 ALFWorld 和 WebShop 上，使用 Qwen2.5-1.5B-Instruct 时，VETTA 相较于 PPO 的成功率分别提升 37.5% 和 22.3%；使用 Qwen2.5-7B-Instruct 时，成功率分别达到 95.5% 和 76.0%。评论器深度对比实验进一步表明，在显著降低评论器侧计算量的同时，仍可取得出色的任务性能。这些结果表明，一个紧凑的共享评论器能够协调回合级与 Token 级信用，在保持价值估计高效的同时提升智能体性能。代码见 https://github.com/Jiaju-Chen/VETTA-official。
arXiv:2610.08400v1 公告类型：新 摘要：大规模自监督预训练重塑了现代机器学习，显著提升了语言模型和视觉模型在下游任务中的泛化能力。尽管近年来深度学习推动了原子系统建模取得长足进展，但该领域的自监督预训练尚未实现同等水平的下游泛化能力。为此，我们提出 Atom-JEPA，这是一种受联合嵌入预测架构启发的自监督预训练框架，通过原子级和子结构级互补目标，从无标注的三维结构中学习潜在表示。我们在大规模分子和晶体数据集上对 Atom-JEPA 进行预训练，并通过在一系列多样化的下游性质预测任务上进行微调，评估其迁移性能。Atom-JEPA 在分子 ADMET 和量子化学性质预测任务上取得了当前最佳性能，并在晶体材料物理性质预测方面表现出很强的竞争力。这些结果表明，仅凭结构数据，潜在空间预测式预训练有望支持广泛的下游泛化。代码和预训练模型检查点已公开于 https://github.com/khelverskovp/atom-jepa
arXiv:2610.08384v1 公告类型：新 摘要：本文研究马尔可夫决策过程（MDP）中的决策导向学习（DFL）。现有方法通过 Bellman 方程的 KKT 条件进行微分，并要求对所有状态—动作对求解线性系统，从而限制了其可扩展性。为此，我们将 MDP 重新表述为一个基于占用度量的线性规划（LP），其可行域由预测的动态模型诱导得到；通过利用枢轴算法识别可行多面体中的活跃约束，我们推导出了闭式梯度。基于占用度量的 LP 层带来了两个挑战：（1）当活跃约束发生变化时，LP 解的梯度不连续；（2）LP 的反向计算开销仍然随状态规模增长，对于大规模或连续状态空间而言代价高昂。针对这些挑战，我们采用增广拉格朗日替代方法，并通过对约束进行随机行草图化来平滑边界跳变；同时提出可学习的软状态聚合层及其函数逼近推广形式，使 LP 能够扩展到大规模有限状态和连续状态 MDP。在多个任务中，与基于 KKT 的 DFL 方法和两阶段基线方法相比，我们的方法以显著更低的计算成本实现了更低的遗憾。所有实验的源代码均可在 https://github.com/A-Eshragh/State_Aggregation_Project 获取。
arXiv:2610.08355v1 公告类型：新论文 摘要：流匹配模型以各向同性高斯分布为源分布；当数据的相关结构事先未知时，这是标准选择。然而，对于多通道脑信号记录，这种结构的一部分是事先已知的。电极位于头部的固定位置，而穿过颅骨和头皮的容积传导使相邻电极的信号以一种跨受试者共有的方式协同变化。尽管如此，现有 EEG 生成模型仍让网络从零开始学习这一结构。我们转而将该结构融入源分布。仅根据传感器坐标，我们构建一个 k 近邻图，并将其拉普拉斯矩阵的图 Matérn 函数作为源分布的协方差，使流从空间相干的模式而非通道间独立的噪声出发。这一改动不增加任何可学习参数，适用于任意耦合方式和漂移网络，并在所有数据集上使用相同的三个超参数。在八个 EEG 数据集和四种流匹配方法上，图 Matérn 源分布在大多数数据集上降低了生成信号与真实信号在五个临床频段内的频谱差异（PSD-KL）。因方法而异，PSD-KL 在各数据集上的几何平均值下降了 12% 至 17%；在电极布置最密集的 PhysioNet-MI 上，降幅最高达 40%。我们表明，这种改善源于传感器位置局部图的空间特征向量，因为在保留特征值谱的同时将特征向量随机化，会使收益消失。此外，直接拟合经验数据协方差得到的先验，其表现反而不如各向同性噪声。同样的构造无需修改即可应用于 MEG、采用患者特异性电极网格的颅内 EEG，以及交通传感器网络，并在每一种场景下都降低了所有方法的 PSD-KL。https://jd730.github.io/projects/GraphPrior
arXiv:2610.08287v1 Announce Type: new Abstract: Understanding the structural relationships among attributes in tabular data is fundamental to machine learning and pattern recognition. While functional dependency (FD) discovery has been extensively studied, scalable discovery of logical dependencies (LDs), particularly as the number of attributes and dependency order increase, remains underexplored. These dependencies capture non-deterministic, condition-specific relationships among pairwise or multiple attributes. Furthermore, existing approaches do not provide a unified framework for extracting multi-attribute LDs and FDs. To address these limitations, we propose LDTool and HLDTool for extracting and visualizing multi-attribute LDs and FDs from tabular data. LDTool extends dependency discovery beyond pairwise relationships, while HLDTool enables scalable extraction through hypergraph-guided search-space reduction. Experiments on three simulated and eleven real-world datasets demonstrate that the proposed framework extracts meaningful LDs and FDs while improving scalability. LDTool recovers the same FDs as existing FD discovery methods with lower runtime in high-dimensional feature spaces, whereas HLDTool enables dependency discovery in datasets with hundreds of features. The proposed framework provides interpretable visualizations of dependency structures and supports applications in exploratory data analysis and the quantitative evaluation of synthetic tabular data.
arXiv:2610.08178v1 公告类型：新 摘要：现有的扩散模型基于潜变量的水印方法高估了其对图像失真的鲁棒性，包括旋转、缩放和平移等几何变换（RST）。此外，这种水印方法范式可能会受到水印嵌入域所导致的固有限制。在本文中，我们首次从理论上分析了这些方法为何不具备对扰动的不变性。通过放宽不变关系，我们推导出一个最大扰动界，用于刻画像素空间中的扰动与其在潜变量空间中对应影响之间的关系。此外，我们首次提出了一种能够涵盖实际检测机制所有组成部分的解析公式。最后，我们通过实验验证了理论发现以及基于潜变量的水印方法的局限性。我们的理论和实证结果表明，在当前的设计范式下，基于潜变量的水印方法在本质上具有有限的鲁棒性。最后，我们提供了分析工具和设计指导原则，未来研究可以据此展开。
arXiv:2610.08131v1 Announce Type: new Abstract: Achieving compositional concept generalization (CoCoGen), the ability to understand novel situations by recombining learned primitives, remains a fundamental challenge in artificial intelligence. Compositional semantic models such as Compositional Distributional Semantics (DisCoCat) offer solutions by generalising vectors to tensors, but suffer from scaling bottlenecks when learning the tensors. Mapping DisCoCat onto Variational Quantum Circuits (VQCs) resolves this limitation for text, yet the methodology has not been expanded to multimodal situations such as the ones involved in CoCoGen. This paper introduces Mu-DisCoCat: a multimodal variational quantum learning framework for DisCoCat that achieves CoCoGen. The framework first learns stable object representations from single-object image-text pairs, then fixes these and uses them to learn the relations between them in multi-object situations. In classical simulations, the model used Uhlmann state fidelity to compute the overlap between the multimodal circuit representations and achieved higher relational OOD accuracy than the evaluated CLIP baseline. Its deployment was evaluated using the destructive SWAP test across noisy quantum emulators, including a range of IBM fake backends, IQM FakeAphrodite, and the IBM Marrakesh quantum processor. Despite real-world device noise, the hardware-executed models maintained a strong positive correlation with simulated fidelities, reliably distinguishing unseen similar and dissimilar pairs. Our work establishes a framework for executing CoCoGen on VQCs, demonstrating a viable use case for near-term quantum hardware.
arXiv:2610.08129v1 Announce Type: new Abstract: Cooperation with unfamiliar partners requires adapting to communication conventions that are not known in advance. We study this problem in a controlled Hanabi-derived environment with scripted hint generation, LLM-controlled receiving decisions, and frozen model weights. Across eight LLMs, linear probes recover intent conventions substantially more accurately than target conventions, yet receiving choices do not consistently agree with the sender's convention. We compare probe-predicted and ground-truth conventions presented either as general rules or as externally computed action recommendations. Rule statements yield modest and model-dependent changes in cooperation, whereas action translation produces larger gains on average. In a Qwen3-8B case study, matched-state statement reversals reveal much greater sensitivity to action recommendations than to rule statements. Activation transfers from oracle-action and non-oracle hint-restatement donors improve intent accuracy on both action classes, but the tested alternatives do not reliably reproduce these benefits. Together, these results distinguish convention decodability, sensitivity to convention information, and cooperative performance, and highlight limitations in turning available partner information into receiving decisions.
arXiv:2610.08069v1 公告类型：新 摘要：两个数据源之间的均值偏移可能很容易检测，却很难在不大幅改变其表征的情况下去除。我们将其去除表述为一个统计决策问题：根据 $\mathbb{R}^d$ 中成对校准测量之间的含噪差异，学习一个线性映射，在严格的失真预算下同时应用于两个数据源，使其在新数据上残留的偏移尽可能少。我们推导出所有此类映射的精确有限样本极小极大风险，即 $(d-k) \mathbb{E}[1/(d+2J)]$，其中 $J\sim\mathrm{Pois}(\kappa/2)$；预算允许删除 $k$ 个方向，而 $\kappa$ 是校准信噪比。将均值校准差异所在的方向投影出去即可达到该风险上界，且无需知道 $\kappa$ 或噪声尺度。这揭示了一个检测—修复鸿沟：检测偏移只需 $\kappa\gg\sqrt d$，而在恒定失真下去除其中固定比例的偏移则需要 $\kappa\asymp d$，这与估计其方向的要求相同。标准线性概念擦除器（MP、SAL、LEACE）都会去除相同的校准差异，因此，该公式可以在拟合前精确给出它们在新数据上会残留多少偏移，以及目标任务需要多少校准数据。该结论具有稳健性：对于非高斯共享内容，配对仍使结论保持精确；对于各向异性噪声，投影仍保有其保证；选择性弃权也无法弥合这一鸿沟。在配对的临床和可穿戴睡眠 EEG 数据上，参与者之间的差异充当校准噪声；该公式能够预测新参与者身上残留的设备偏移，并表明每人增加更多记录很快就不再有帮助。总体而言，这些结果可以判断：一个效果不佳的校正究竟需要更好的方法、更多记录，还是更多参与者。
arXiv:2610.07885v1 发布类型：new 摘要：心电图（ECG）界定，即识别波形边界，是将原始心电信号转化为具有临床可解释性测量结果的基础步骤。深度学习推动了这一任务的发展，但仍依赖成本高昂的专家标注。自监督预训练和半监督学习等标签高效策略有望减轻这一负担，但它们能否实现可靠的界定，以及由此产生的深度模型是否优于临床实践中使用的界定工具，仍不明确。我们分两个阶段对此进行研究。首先，在一个内部数据集和四个外部数据集上比较自监督目标与监督式或半监督微调方法，我们发现预训练确实有帮助，但目标函数十分重要；同时，半监督微调的价值取决于预训练目标。其次，我们使用三个互补指标，将筛选出的深度学习模型与广泛使用的开源工具（NeuroKit2、Prominence、ECGdeli）和商业工具（CalECG）进行基准比较。该模型在每个指标和每个数据集上均排名第一；在节律多样的数据集上，它以明显优势超过表现最强的工具（mIoU 为 71.3% 对 54.8%；平均逐点敏感度为 92.6% 对 76.4%），并且从窦性心律到心律失常时性能下降最少。进一步按节律分层并进行逐点分析，刻画了各工具的独特表现，为工具选择提供了实用指导。这些结果提供了系统性的多数据集证据，表明自监督预训练对心电图界定有效，并能够利用丰富的未标注数据，使经过标签高效训练的深度学习模型超越广泛使用的界定工具。这支持在多样化的真实世界临床环境中采用此类模型。
arXiv:2610.07874v1 公告类型：new 摘要：同策略蒸馏（OPD）作为一种后训练方法已得到广泛研究，其特点是学生模型从更强教师模型在自身 rollout 上提供的 token 级监督中学习。近期研究通过采用不同的蒸馏奖励形式和教师配置改进了 OPD，但蒸馏目标仍然集中于模仿教师。然而，当更强教师与学生之间的分布重叠有限时，这种正向引导可能提供不足的学习信号。在本文中，我们提出负策略同策略蒸馏（NP-OPD），通过引入一个表现较差、能力较弱的负策略来补充教师监督；该负策略在学生学习过程中充当负向参照。NP-OPD 不修改蒸馏奖励形式，而是在 rollout 阶段引入负策略，持续提供相较于教师更受负策略偏好的 token，使这些 token 在整个训练过程中始终接受教师监督。这样一来，NP-OPD 在保留 OPD 所采用的教师正向监督的同时，通过负策略 rollout 提供了明确的负向信号。通过大量实验，我们表明，NP-OPD 在不同模型规模、生成模式、推理领域以及不同 OPD 变体上均能提升 OPD 的效果。此外，我们的分析显示，NP-OPD 能够有效抑制负策略相较于教师更偏好的 token，并使学生偏离负策略。这些结果支持了我们通过负策略 rollout 引入负向信号的设计，并为理解 rollout 策略在 OPD 中的作用提供了新的见解。代码将在 https://github.com/naver-ai/np-opd 发布。
arXiv:2610.07857v1 Announce Type: new Abstract: This study formulates individual route reproduction as a shortest-path problem over learned driver-specific latent link costs. The central idea is that, once such latent costs are inferred from contextual information, observed routes can be reproduced without enumerating alternative route sets. We propose a neural pipeline that includes a perception model that embeds context covariates, which comprises individual characteristics, trip-specific attributes, and network-level traffic states, into the personalized link costs. A constrained optimization (CO) layer, which determines the shortest path (SP) based on these estimated costs, follows the perception encoder. To enable end-to-end training, we employ decision-focused learning to align the predicted shortest paths with observed routes. The implicit maximum likelihood estimation (iMLE) provides an approximate gradient of the loss function that contains the non-differentiable CO layer. Furthermore, a regularization term anchors the latent cost distribution to the empirical scale of observed link travel times, mitigating the scale ambiguity inherent in shortest-path supervision. Empirical evaluations demonstrate that the proposed framework outperforms baseline route choice models in path reproduction. The learned latent costs, interpreted as proxies for perceived travel costs, provide plausible explanations for heterogeneous route choices.
arXiv:2610.07796v1 Announce Type: new Abstract: Empowerment captures the capacity for an agent to actively control its environment. While conceptually appealing as an information-theoretic quantity, the connection between empowerment and structurally central states that provide broad access to future outcomes has remained an open question. In this work, we link empowerment maximization and skill-learning methods to provide new geometries for interpreting and analyzing empowerment. Our analyses answer longstanding open questions on the connections between empowerment and structural centrality. Our analyses also reveal distinctions between information and reward geometries, highlighting important theoretical implications to build scalable empowerment-maximization methods. Website and code can be found at https://empowerment-geometry.github.io/.
arXiv:2610.07778v1 公告类型：新 摘要：属性图聚类旨在通过联合利用节点属性和图拓扑结构来发现节点群组，然而，其无监督特性使得模型选择和适配本质上较为困难。现有方法通常针对每个输入图分别训练和调优模型，导致流程成本高昂且脆弱，并且往往无法迁移到具有不同特征空间、结构模式以及属性—结构相关性的图上。本文研究一种通用方案：能否只训练一次单个模型，并将其直接应用于多种属性图，而无需针对特定图进行训练、微调或超参数搜索？我们提出 OFAG，一种用于属性图聚类的基础模型。OFAG 构建于 Prior-data Fitted Networks 之上，从依据关于潜在簇、节点属性和图结构的广泛先验生成的合成属性图中，学习可复用的聚类推理策略。为处理不同图之间不兼容的特征空间，OFAG 采用与维度无关的逐信号图编码器，将每个特征通道视为一个图信号，并对其针对共享图滤波器的响应进行建模。该模型通过超球面聚类目标进行训练，在推理时仅需一次前向传播即可生成有利于聚类的节点表示。在十个数据集上，一个冻结的 OFAG 模型在 NMI、ACC、ARI 和 F1 指标上的平均性能及平均排名均达到最佳，同时完成全部十个数据集仅需 12.43 分钟——速度比第二快的基线方法快 6 倍以上，在聚类质量方面也比第二优方法快近 28 倍。我们的代码和预训练检查点可在 https://github.com/Cloudy1225/OFAG 获取，使实践者无需额外训练或调优，即可将 OFAG 直接应用于自己的属性图数据集。
arXiv:2610.07706v1 公告类型：新 摘要：自回归大语言模型（LLM）的能力取得了快速进展，但其规模不断增大也带来了推理阶段巨大的计算和内存开销。知识蒸馏（KD）通过对齐离散概率分布，将知识从大型教师模型迁移到较小的学生模型，为此提供了一种实用的解决方案。然而，现有面向 LLM 的知识蒸馏方法主要依赖通过词汇表中各索引位置的概率值来评估差异的散度，未显式利用词元级语义信息。我们提出了面向 LLM 的基于 Wasserstein 的知识蒸馏方法（WASD），通过基于 Wasserstein 的距离引入词元级语义信息，该距离使用由词元嵌入构造的代价矩阵。为确保计算可行性，我们采用 Sinkhorn 散度，并推导出一个梯度等价的目标函数，使其能够在不引入额外网络的情况下高效优化。在多个 LLM 系列和不同规模的模型上进行的实验表明，WASD 在包括指令遵循、数学推理和代码生成在内的多种任务上，都能持续提升蒸馏性能。我们的结果凸显了词元空间中编码的语义信息对于 LLM 蒸馏中的有效分布对齐的重要性。实现代码已公开发布于 https://github.com/aailab-kaist/WASD 。
arXiv:2610.07610v1 Announce Type: new Abstract: Semi-supervised learning typically assumes that labeled and unlabeled data share an identical class distribution and label space. However, this setting is often violated: unlabeled data may be imbalanced and contain unknown class samples, causing mismatches in both class distribution and label space. Such dual mismatch leads to majority classes dominating the latent space and unknown class samples being overconfidently misclassified, degrading feature discriminability and pseudo-label quality. To address this, we propose a hub-spoke latent geometry, where known classes are uniformly distributed around a central hub and each class forms compact clusters around its prototype, while the hub provides an anchor for a low-evidence region specifically designed for high-uncertainty unknown class samples. Integrated with an evidence-based classifier, this geometry ultimately enhances feature discriminability and uncertainty separation by mitigating majority-class domination through structured feature organization and guiding high-uncertainty unknown class samples toward the hub. Extensive experiments show that our method outperforms state-of-the-art methods, with a maximum improvement of 3.25% across various settings.
arXiv:2610.07553v1 公告类型：新 摘要：LoRA 微调在低秩更新子空间内使小型语言模型（SLM）适应异构指令数据，因此容易受到三个结构性问题的影响：相互冲突的梯度彼此抵消、静态数据选择无法跟踪不断变化的学习动态，以及子空间饱和导致后续更新覆盖有用方向。我们认为，有效的适应过程需要控制哪些由数据产生的梯度进入 LoRA 子空间，以及何时进入。为此，我们提出 GRADE（GRadient-Aligned Data-centric rEcipe），一种以数据为中心的框架，结合了两种机制：状态感知选择器持续接纳与不断演化的多任务梯度场相一致的样本；自校准的步级门控机制则拒绝那些在接近饱和时可能导致破坏性覆盖的更新。在三个当前一代的骨干模型和一个包含七个异构数据集的指令数据池上，GRADE 在准确性和鲁棒性方面均优于强大的数据选择及 PEFT 稳定化基线。它是唯一一种在每种架构上都始终优于标准 LoRA 的方法，同时能够产生更连贯的梯度轨迹和更少的破坏性覆盖。这些结果表明，成功适应 SLM 不仅取决于选择哪些数据，还取决于允许哪些梯度进入并持续存在于受约束的更新子空间中。
arXiv:2610.07485v1 Announce Type: new Abstract: We study estimating rare-event probabilities $I = \mathbb{P}(g(\mathbf{X}) > \gamma)$ with $\mathbf{X} \sim \mathcal{N}(\boldsymbol{\mu}, \boldsymbol{\Sigma})$ and general $g : \mathbb{R}^d \to \mathbb{R}$. We address this problem through importance sampling, and propose a framework that substantially improves efficiency and robustness over baselines such as crude Monte Carlo, adaptive cross-entropy, variational-inference-based methods (including reverse- and forward-KL approaches), as well as Safe-ICE, Subset Simulation, and Sequential Monte Carlo, drawing on ideas from both rare-event estimation and cross-entropy optimization. The key contribution has two parts: first, we separate the problem into coverage, to overcome the cold-start barrier, and fitting, to refine proposals once a meaningful signal is available; second, we constrain the final GMM proposal so that it has finite importance-sampling variance (since coverage alone is not sufficient -- without safeguards, importance sampling may still suffer from infinite variance). Together, these ingredients yield expressive proposals; finite variance does not by itself guarantee practical stability at a fixed sampling budget. Extensive experiments demonstrate substantial variance reduction, strong robustness across diverse benchmarks, and favorable cost--efficiency trade-offs, with the proposed approach often outperforming these baselines, particularly in high-dimensional and multimodal settings where competing methods frequently become unstable or fail. Our code is available at https://github.com/lorek/robust-cfi-is.
arXiv:2610.07475v1 Announce Type: new Abstract: Adapting a reinforcement learning policy to changes in another agent's behavior typically requires a large amount of new interaction data. Policy sensitivity provides a first-order prediction of how a locally optimal policy changes with a behavioral parameter, but its computation requires second-order derivatives whose effects propagate across future interactions. We develop a finite-depth framework to estimate this sensitivity by approximating the policy Hessian and mixed derivative using information from a reference environment. The method features an adjustable propagation depth which determines where derivative propagation along the trajectory is truncated. We characterize the derivative contributions omitted by finite-depth propagation and derive truncation-error bounds for the approximated derivatives and resulting policy sensitivity. The bounds are nonincreasing with propagation depth and vanish at full-horizon propagation. Using a belief-driven pursuit-evasion game as a validation scenario, the proposed method generally achieves lower derivative-estimation errors as the propagation depth increases and outperforms the baseline methods in both estimation accuracy and policy adaptation. The sensitivity-based initialization improves zero-shot return over direct transfer, and also shows advantages for the subsequent fine-tuning in the target environment.
arXiv:2610.07457v1 公告类型：新 摘要：细粒度混合精度量化有望实现高效的大语言模型推理，但局部精度选择可能与常规 GPU 存储和计算单元产生冲突。这种精度边界不匹配限制了压缩技术转化为实际加速的能力。我们提出 AlignQuant，一种训练后量化方法，使用与 GPU 兼容的二维权重瓦片作为精度分配、紧凑存储和执行的统一单位。这种共享分区使精度能够根据输出通道内的敏感度进行分配。在量化激活下，联合预填充/解码校准通过由语言模型损失梯度加权的投影输出扰动，对精度降低进行评分。经过阶段归一化的评分，在全模型权重存储预算下，优先为对任一阶段重要的瓦片分配更高精度。每个瓦片存储一种选定的表示形式，而面向不同阶段的内核则复用打包后的模型，并将低比特权重扩展为使用 8 位激活进行 INT8 计算。在涵盖 3B 至 14B 参数的四个 LLM 上，AlignQuant 在保持模型质量的同时，相较于 BF16 实现了最高达 $2.50\times$ 的生成加速。评估还涵盖了三种 GPU 以及最长达 64K token 的上下文。这些结果表明，通过共享的瓦片单位，局部精度灵活性与规则化 GPU 执行可以并存。实现代码可在 https://github.com/HanzhiZhang-Ulrica/AlignQuant 获取。
arXiv:2610.07430v1 Announce Type: new Abstract: Reducing nonstationarity in a persistent time series entails deciding how much of its temporal dependence to remove. In finance, fractional differencing is often tuned using the Augmented Dickey--Fuller (ADF) test, limiting the search to a one-parameter family of lag profiles and addressing input preservation only indirectly. We propose StaFIR, a causal finite-impulse-response filter with a learned nonnegative mixture of exponential lag profiles. Its convex learning objective balances empirical stationarity with similarity to the input. We evaluate StaFIR on ARFIMA--GARCH controlled settings and rolling financial series, including a realized-volatility forecasting task. The experiments show that StaFIR adjusts its filtering strength to persistence while limiting unnecessary transformation in stationary regimes. In downstream forecasting, there is no clear accuracy difference from fixed half-order differencing, while StaFIR achieves higher measured similarity to the raw signal. A complementary direct forecasting experiment finds that greater input similarity is associated with smaller forecasting penalties, although the raw representation remains stronger.
arXiv:2610.07358v1 公告类型：new 摘要：野火后泥石流（PFDFs）是由强降雨冲击近期烧毁地形而触发的、破坏性极强且含有大量沉积物的灾害，会导致山坡失稳，并威胁基础设施、地方经济和社区安全。已有研究提出使用数据驱动方法，直接从历史PFDF观测数据中学习预测模式。然而，目前数据驱动PFDF预测的研究格局在特征空间、模型架构和评估协议方面高度碎片化，导致难以进行严格比较并提炼科学认识。此外，现有研究缺乏对异质因素（如气象条件、地形特征、土壤性质和烧毁严重程度）在触发PFDF中的相对重要性进行系统考察。为解决这些局限，我们提出了一个统一的数据驱动PFDF预测基准，使不同模型和特征配置能够接受公平且全面的评估。此外，为更好地理解PFDF形成的潜在驱动因素，我们提出了一个基于强化学习的特征选择框架，用于识别那些经过扰动后会使正事件与负事件变得无法区分的因素，从而发现不同区域PFDF发生的区域性潜在机制。我们的代码和基准已公开发布于https://github.com/KINDLab-Fly/PFDF-Benchmark。
arXiv:2610.07340v1 公告类型：新 摘要：蛋白质-分子虚拟筛选越来越多地被视为共享嵌入空间中的表征学习问题。现有方法依赖密集的整体对齐，将不变的结合决定因素与无关相关性纠缠在一起，从而限制了向新靶点的迁移能力。已有研究指出，蛋白质-分子系统中的结合涉及稀疏的跨模态相互作用：结合由一个较小的接触界面和少数关键的局部相互作用（例如氢键、疏水接触和盐桥）决定，而不是由蛋白质和分子的全局结构决定。我们假设，发掘并利用稀疏的相互作用模式对于超越训练数据实现泛化至关重要，因为这些模式具有可复用性，并有望提升模型在不同场景下的性能。本文旨在识别并利用稀疏的相互作用模式，并验证这一假设。由于训练数据仅包含观测到的结合对，我们在 Heckman 风格的选择机制下，通过 V 结构因果模型形式化这一先验，并建立了三个理论结果：（i）在缺乏适当稀疏性约束的情况下，相互作用蛋白质和分子的潜在概念不可识别；（ii）在满足结构稀疏性条件时，这些概念及其稀疏相互作用具有分量级可识别性；（iii）对这些条件进行低秩松弛后，可以实现概念及相互作用的子空间可识别性。受这些原理启发，我们提出了 CausalBind，并实现了三个变体。在 DUD-E 和 LIT-PCBA 基准上的大量实验表明，所有变体均持续优于强大的检索基线，在 LIT-PCBA 的早期富集指标上提升最大，并且能够进一步泛化到靶点级和骨架级分布外划分。代码可在 https://github.com/lokali/CausalBind 获取。
arXiv:2610.07226v1 公告类型：新 摘要：“产生某种结果所需的不可约条件是什么？”这是计算与科学领域反复出现的最常见问题之一。其答案，即最小充分见证，正是我们所谓的解释、机制和理由。这类问题通常要求找出多个最小见证，但标准强化学习方法可能只能发现一个解，或发现彼此冗余的解。我们将这一问题形式化为最小见证识别，并提出 Minimal-Witness Reinforcement Learning（MWRL）。MWRL 取策略采样的成功提议所认证集合的并集，并根据在缺少某个提议时该群体并集会损失的覆盖范围，为每个提议分配贡献。这种直接源于问题定义的信用分配方式，仅凭黑盒验证器返回的单个位，就统一满足了最小性和找回备选方案的要求。在此原则下，我们推导出一种能够找回完整见证族的价值迭代规划器，以及一种可扩展到大型语言模型的策略梯度方法。在不同实验设置下，MWRL 能找回大多数最小见证，而其他方法则返回冗余的超集或单个见证。通过使见证族能够从验证器反馈中学习，MWRL 将强化学习的适用范围从单一解优化扩展到了更广泛的领域。我们的代码可在 https://github.com/TSUITUENYUE/MWRL 获取。
arXiv:2610.07207v1 公告类型：新 摘要：混合专家（MoE）Transformer 通过为每个 token 仅激活少数专家来扩展模型容量，但这种稀疏性带来了一个隐藏的可靠性问题：当路由不够完善时，负载均衡模型可能会将 token 发送给那些对所分配输入训练不足的专家。我们提出分布鲁棒混合专家训练（DRMoET），这是一种可直接替换的目标函数，将逐层专家视为内生的鲁棒性群组，并针对高损失路由结果进行优化，而不仅仅是均衡流量。DRMoET 基于经过 EMA 平滑、并按激活加权的专家损失，通过熵正则化 softmax 规则更新每层的专家分布，在保留标准 MoE 计算方式的同时，强化合理的非首选路由路径。在总参数量为 746M 和 10.3B 的规模下，按照 FLAME-MoE 配方训练时，DRMoET 在下游任务平均表现上均优于标准 FLAME-MoE 和无辅助损失均衡方法。在总参数量为 10.3B、训练 token 数为 67B 时，DRMoET 将七项任务的平均分从 0.6625 提升至 0.6767，而无辅助损失基线的平均分为 0.6431。机制分析表明，在平均损失几乎不变的情况下，专家损失方差更低；在强制进行中间-$k$ 误路由时，额外损失降低了 4.3%；同时，领域专家的专业化程度得到提升。这些结果表明，在可靠地扩展稀疏 MoE 方面，路由鲁棒性与利用率均衡同样是一个实用的优化目标。项目主页和代码地址：https://drmoet.github.io/。
arXiv:2610.07103v1 公告类型：新稿 摘要：Muon 是一种以高效著称的优化器，对于矩阵值更新具有清晰的解释，但卷积核以四维张量的形式存储。标准实现会将这些张量重塑为矩阵，这种便捷做法破坏了 Muon 背后的理论理解。为研究这一问题，我们直接在卷积算子几何中形式化相应的优化目标，并引入 Convolutional Newton-Schulz（Conv-NS），该方法在保留卷积核支持范围的同时，近似该几何中的极分解因子。在快速训练实验中，Conv-NS 和基于重塑的 Muon 均具有较高的计算效率，并且在 CIFAR-10 和 ImageNet 分类任务上取得了相当的准确率。然而，正如预期那样，在理论上更一致的 Conv-NS 应当优于基于重塑的 Muon，因此我们研究了实践与理论理解之间的不一致，并提出假设：精确的卷积正交化可能会对更新施加过强约束。这些发现凸显了 Muon 出色的实际性能，同时也为其在卷积领域的进一步发展提供了方向。我们的代码已通过 \href{https://github.com/thib-s/muonconv-cifar10-airbench}{github conv-muon} 公开。
arXiv:2610.07086v1 公告类型：新 摘要：LLM 智能体通过生成结构化工具调用与外部系统交互。给定用户请求、对话上下文和工具模式目录，工具调用模型必须选择工具并生成其参数，并且可能在单次响应中生成多个调用。标准自回归解码会逐个 token 地生成这些调用，因此对于涉及多个调用或大量参数字段的请求，会产生较高延迟。显式的参数结构为并行生成提供了机会，但后续参数值可能依赖于前面的字段和调用，因此独立生成的值可能与目标模型的输出不同。我们提出了 SchemaFill，一种通过槽位并行推测解码实现高效 LLM 工具调用的框架。SchemaFill 无需预先知道实际的调用序列或参数值，即可并发生成未来槽位值作为候选。涵盖多个字段和调用的候选内容会在实际输出前缀下由目标模型串接并验证。只有经过验证的 token 才会被提交；当候选内容不一致时，由目标模型提供修正。该方法在利用槽位和调用之间并行性的同时，实现了目标模型验证。在 Glaive 和 BFCL 上，与自回归解码相比，SchemaFill 将端到端吞吐量最高提升至 4.05 倍。代码可在 https://github.com/Czzzk/SchemaFill 获取。
arXiv:2610.07060v1 发布类型：新 摘要：尽管短期至中期天气预报已取得长足进展，但骤旱等高影响事件的预测，对于预警业务和基于物理的次季节至季节（S2S）预测系统而言，仍是一项关键挑战。在此，我们证明，对于欧洲地区的 S2S 土壤湿度预报而言，预报技巧在很大程度上不仅取决于预报模型本身，也取决于预测问题的表述方式。我们采用具有时间和空间双通路注意力机制的基于 Vision Transformer 的架构，结果表明，残差学习是超越持续性预报基准的关键。只有在以物理单位而非标准化异常值预报根区土壤湿度时，这一优势才能实现，这表明目标变量的表征方式本身会限制可预报性。通过对分位数预测头进行微调，进一步构建的概率扩展模型还能够提供校准良好的预测分布。以 2021—2022 年的深度学习和业务化 ECMWF S2S 基线为参照，我们的模型在所有预报时效上均取得了最高的确定性和概率预报技巧，并能可靠识别异常干燥的根区状态（低于第20百分位）。然而，按照多五日段强化标准定义的骤旱发生，仍是当前所有 S2S 系统共同面临的根本性挑战。这些发现推动了数据驱动的 S2S 土壤湿度预报发展，同时也凸显出预测干旱快速发展的持续难题。
arXiv:2610.07002v1 Announce Type: new Abstract: Diffusion models learn semantic representations while generating images. In the Decoupled Diffusion Transformer (DDT), a condition encoder provides features that guide a velocity decoder in denoising. To enable effective denoising at all noise levels, these features must capture both high-level abstract structures and low-level details. However, skip/residual connections in the encoder allow shallow features to bypass successive transformations, which may limit progressive abstraction, or at least make it difficult to disentangle different levels of abstraction. We propose DDT-RFE, which removes the residual connections around the Self-Attention and MLP operations in each encoder block while maintaining stable training. To retain the information that abstraction discards but that the decoder still needs, we fuse the input patch embedding with intermediate and final encoder features to form the encoder output. The decoder thus has access to information from multiple encoder depths, while each encoder block is able to learn more abstract representations. DDT-RFE achieves overall improvements over DDT across visual understanding tasks, including image classification, semantic segmentation, object discovery, and semantic correspondence, while using fewer encoder blocks. It also achieves a lower FID for image generation on ImageNet.
arXiv:2610.06950v1 公告类型：新增 摘要：当前，将技能注入冻结语言模型需要付出百万级参数和一套强化学习流程的成本。我们提出 \method{}，一种通过行为克隆训练的 System-1 决策算子，将这一成本降低了约两个数量级。默认算子仅使用 330K 个参数，却能够达到使用 1.33M 个参数、通过强化学习训练的算子的水平，并且性能更高或相当；同时，它将 3,685 个 token 的思考过程压缩为 6 个 token 的决策，且准确率不受损失。一个拥有 23K 个参数的 rank-4 变体，其参数量仅为已发表的最强技能算子的 1/58，足以应对 SearchQA，并基本足以应对 LiveMath（在后者中，更高的秩仍有帮助）；同一方案可迁移到五项任务和三个骨干模型上，并且在训练数月后发布的 LiveMath 题目上，分布外增益仍然存在。与先前工作的差距在于可训练性，而可训练性由初始化和架构共同决定：先前算子的初始化会使两个大型因子矩阵在第一次优化步骤中的梯度都归零，而我们在共享低秩骨干网络中采用零初始化的输出投影，该投影能够立即接收到梯度，梯度流探针直接验证了这一点。性能提升并非来自思维链压缩：在 57 道 LiveMath 题目中，有 23 道的结果优于基础模型的 best-of-8 采样；logit-lens 探针表明，该算子是在模型已有的后层路径上放大答案，而不是更早地写出答案。在 13 个基础模型—任务组合中，性能增益与基础模型的提升空间相一致；不同技能可以组合成近似线性的算子，并可在推理时进行相加、插值和热切换。代码地址：https://github.com/rlisml/decisionsteer。
arXiv:2610.06890v1 公告类型：新 摘要：我们介绍了一份行业实践报告，内容涉及在汽车制造业中运营事件驱动型云基础设施，以支持持续的机器学习训练，历时三年。我们的系统在多个工厂间编排产品专用模型对的 GPU 加速训练，包括物理预测模型和强化学习控制策略，并协调由制造事件触发的长时间运行 GPU 工作负载。该架构结合了采用 EC2 GPU 容量提供商的 Amazon ECS、基于 SQS 且配备死信队列的消息传递机制，以及执行集群并发限制的准入控制 Lambda 调度器。运行于 ECS Fargate 上的 Conductor 编排器按每周计划启动感知依赖关系的再训练链。整个基础设施通过模块化 Terraform 进行代码化，并采用多账户隔离。基于 40000 多个生产训练任务，我们报告称，与始终运行的 GPU 基础设施相比，成本降低了 72%–78%。离散事件仿真证实，准入控制是必需的（朴素调度会导致 65% 的任务丢失），并且队列排空的延迟与 AWS Step Functions 相当，同时消除了每个任务的启动开销。我们总结了实践经验，并将仿真器和 Terraform 模块骨架作为开源成果发布。
arXiv:2610.06883v1 Announce Type: new Abstract: Neural operators provide fast surrogates for time-dependent PDEs, but autoregressive deployment creates a refinement-allocation problem: prediction errors vary over space and time, while only a finite number of local corrections can be committed along a trajectory. We formulate this as budgeted adaptive neural-operator solving. A global Fourier neural operator advances the full field, a local operator proposes patch-wise residual corrections, and a set-aware selector chooses where to refine. A macro policy decides when and how much of the remaining refinement budget to spend. We introduce rollout-verified policy improvement (RV-PI), which evaluates feasible refinement counts through actual continuation rollouts of the learned PDE solver, converts long-horizon advantages into conservative policy targets, and accepts an update only when held-out trajectory error improves. On the shallow-water benchmark with a 32-intervention budget, RV-PI achieves a three-seed mean trajectory relative L2 error of 0.6910, improving over immediate-only policy improvement by 5.37% and RandomMacro by 2.41%. On the forcing-driven Brusselator benchmark with a 76-intervention budget, RV-PI attains 0.09954, improving over immediate-only policy improvement by 2.31% and RandomMacro by 5.32%. These results show that, under a fixed refinement budget, the value of a local correction depends on its downstream effect on the autoregressive trajectory, not only on its immediate error reduction.
扩散模型通过一系列连续的去噪问题生成数据，人们广泛观察到其会先呈现粗粒度结构，再呈现细节。然而，这一直主要停留在经验性和定性层面。我们提出了特征信息动力学这一信息论框架，用于定位特征在扩散过程中何时生成。利用 I-MMSE 恒等式，我们将特征互信息变化率与最优无条件去噪损失和特征条件去噪损失之间的差距联系起来，从而得到特征信息密度的实用估计器。我们进一步发展了一种链式分解，将特征层级中的共享信息与增量信息区分开来。我们首先利用该框架定量验证了像素扩散中的频谱自回归，随后将分析扩展到频率之外：在 class $\to$ mask $\to$ Canny 条件链下，pixel、SDVAE、VAVAE 和 RAE 的逐特征信息密度存在差异，揭示了这些表示之间的根本区别，并表明有序生成可能有利于扩散模型的训练。我们的代码可在 https://github.com/AI4Science-WestlakeU/feature-information-dynamics 获取。
受污染物体的接触可能通过家用机器人的夹爪、工具和共享表面传播危害，而新的接触还可能使现有计划变得不安全。现有基准无法同时评估规划器如何根据接触历史识别卫生风险，以及如何在发生新的接触事件后规划安全的后续行动。规划器必须在时间和资源限制下完成这些任务，同时遵循用户的优先级。我们提出了 HygieneRoboBench，其中包含来自 134 个任务族的 624 个实例，用于评估基于给定执行历史安全完成家务任务的能力。这些任务涵盖通过两个夹爪和共享物体传播的污染、处理成本以及用户优先级。我们结合受控的历史、配置文件和事件对比，以及独立的计划评估，分别评估安全解决、满足用户优先级时的成本效率，以及对接触事件的响应。对基于 LLM 和符号规划器的评估表明，安全完成任务并不保证在用户优先级下实现最低执行成本。为解决这一问题，我们提出了 Hygiene-NSP。它结合基于 LLM 的落地、接触历史重建和 CP-SAT，在用户优先级下联合规划卫生处理与任务执行。Hygiene-NSP 的安全解决率和最优安全解决率分别达到 94.4% 和 90.4%。在完整数据集上，这两个比率均高于受评估的基线规划器。项目页面：https://euron-zc.github.io/HygieneRoboBench/。
大型语言模型智能体正在加速科学自动化，但经过验证的执行结果很少能转化为持久的程序级改进，而现有评估也未考察这一过程在自然科学和社会科学的连续任务中的表现。我们将 ScienceClaw 形式化为一种固定参数的程序自演化机制，统一任务求解、科学验证和程序更新。ScienceClaw-Eval 覆盖 23 个学科，通过连续任务流和独立重置评估，衡量科学正确性、演化增益、保持能力、跨数据集迁移能力和演化成本。我们的框架通过多轮交互修复可执行工作流，将经重新执行验证的失败—成功轨迹转化为相互关联的 Skill 和 Operator 候选，并且仅当源任务回放能够复现修复结果且独立科学任务得到改进时，才保留相应更新。代码可在 https://github.com/beita6969/ScienceClaw 获取。
2011年的助手会回答并等待，2023年的代理会完成一项任务后停止。2026年9月，Meta 的 Muse 展示了一种面向单个人的代理：它拥有账户、设备和记忆，并能在持续不断的对话中工作；但它是封闭的，运行在供应商的云端，且仅限于一个国家。人们期待这种代理能够操作个人的账户和设备，跨越数周记住相关信息，在值得主动联系时先发起对话，并为自己的行为负责。这是一类软件，而不是一个模型；直到现在，它还没有开源的对应方案。本报告通过五个问题和三个阶段定义个人代理，并根据 Meta 的公开资料及其生产环境提示词副本，梳理 Muse 的构建方式，每项陈述均标注来源。随后，报告介绍了 GPL-3.0 许可下的开源对应方案 nanoMuse：在个人拥有的每台设备上运行一个代理，让它能够操作手机屏幕和电脑屏幕。它们通过任何人都能运行的中继服务共享同一段对话；每项操作都要经过 Sentinel；记忆以个人可以读取的文件形式保存；模型则由用户自行选择。报告还给出了其规模和成本的估算，并将以下内容列为路线图：开放性、带来源追溯的记忆、面向操作能力的评测套件，以及用于这些操作的开源模型。
基于 LLM 的智能体正在不断推动科学与工程问题求解的发展，而物理模拟正逐渐成为一种具有挑战性但又切实可行的测试平台，用于复现复杂的物理现象，并应用于具身智能、游戏和电影等领域。作为此类模拟的核心工具，求解器负责计算动态系统的状态如何随时间演化。构建这类求解器需要理解物理规律以确定合适的模型，运用数学推理来描述底层动力学，并通过软件工程将其实现为可执行代码；然而，LLM 智能体的这一能力仍未得到充分探索。为此，我们推出了 WorldSolver，这是一个包含 168 个模拟任务的基准，任务源自 61 篇经典计算机图形学论文中的物理现象，涵盖 7 个物理领域。每个任务都包含一个代码框架，为场景提供固定的模拟环境，而求解器的实现则留待智能体完成。具体而言，我们从三个维度对其进行评估：执行检查，用于判断是否成功执行；视觉保真度，用于评估渲染模拟中是否复现了预期的动态行为；以及物理合理性，用于基于物理规律验证所生成的动力学。对前沿智能体的实验表明，生成可执行的求解器本身就十分困难，而同时满足视觉和物理正确性则更加困难。GPT-5.6-Sol 和 Claude-Opus-5 的表现相较于其他受评估智能体更好，但总体得分仍分别只有 48.7% 和 46.7%。WorldSolver 是迈向智能体求解器生成的早期一步，我们希望它能够推动智能体忠实模拟动态物理世界的研究进展。代码见 https://github.com/sirujiang/WorldSolver。
近年来，世界模型取得的进展推动了越来越逼真的视觉合成。然而，这些生成的环境在很大程度上仍然是无声的。为世界模型带来声音面临三项核心挑战：实时生成，以跟上交互式视频流；交互式控制，以响应视频流中途的声音指令；以及空间对齐的立体声，以反映场景几何结构和摄像机运动。为应对这些需求，我们提出了 WorldSonus，这是一种专为世界模型中的实时空间声音合成设计的交互式视频到音频框架。在实时生成方面，WorldSonus 采用流式因果自回归扩散架构，以 0.41 的低实时因子（RTF）合成音频片段。在交互式控制方面，我们引入了以音频为中心的字幕生成流程，并结合按片段索引的提示调度机制，使声音事件能够在生成过程中动态操控。在空间对齐方面，我们利用从多样化立体声和全景声数据中整理出的高质量立体声监督信号。大量实验表明，尽管 WorldSonus 专为世界模型设计，但它能够有效泛化到开放领域的视频到音频基准测试，在声学质量和空间对齐方面均达到或超越当前最先进的双向模型。项目主页：https://noizai.github.io/WorldSonus/
胃腺癌（GA）是全球癌症相关死亡的主要原因之一，从全视野切片图像（WSI）中准确分类组织病理学亚型对于制定有效的治疗方案至关重要。尽管将病理报告文本与WSI相结合的多模态方法能够提高分类性能，但现有方法通常依赖计算成本高昂的Transformer架构和大型语言模型。我们提出了一种多模态知识蒸馏（MKD）框架，使用低秩多模态融合（LMF）将预训练的WSI图像编码器与临床文本编码器相结合，从而在训练过程中高效地建模跨模态交互。每张WSI都表示为一个图像块集合，并与一条切片级诊断描述配对。教师模型学习融合后的图像-文本表示，用于亚型分类；学生模型则蒸馏这些知识，以实现准确的仅图像推理。我们在PatchGastric基准数据集上评估了该方法，其平均准确率至少比最先进的方法高3.35%，且无需依赖基于Transformer的融合、多任务学习或大型语言模型。源代码可在https://github.com/helomelo1/MKD-LMF获取。
Dynamic multimodal learning aims to learn robust representations by adaptively modeling information discrepancies across modalities. However, existing methods still suffer from two limitations: (i) static cross-modal alignment strategies usually impose uniform constraints on all samples while overlooking sample-wise variations, potentially leading to unreasonable over-alignment; and (ii) confidence- or uncertainty-aware fusion methods often fail to adequately account for feature magnitude and confidence differences across modalities. For modality pairs with significant feature magnitude differences or small confidence gaps, it might be unreliable to strictly align fusion weights according to confidence. To address these issues, we propose an Alignment- and Calibration-driven Multimodal Learning framework (ACML). Specifically, ACML incorporates a dynamic cross-modal triplet alignment module, which enforces strong semantic consistency for high-confidence positive pairs while encouraging diverse representation learning between high- and low-confidence positive pairs according to their confidence gaps. Additionally, ACML introduces a difference-aware attention calibration strategy that adaptively adjusts attention regularization based on feature magnitude and confidence differences across modalities, thereby mitigating biases caused by unreasonable fusion constraints. Extensive experiments on multiple multimodal benchmark datasets demonstrate that ACML consistently achieves superior performance and robustness over recent state-of-the-art methods.
随着智能体通过生成和修订技能而持续改进，发现和完善这些技能的过程本身也成为一种可学习的对象。任务技能直接作用于任务执行，而元技能则决定智能体如何发现并改进未来的技能；因此，元技能的价值会通过其引发的后续搜索过程体现出来。现有方法根据观测到的原始技能搜索轨迹和分支结果来改进元技能。然而，分支表现同时受到初始发现状态和产生该搜索过程的元技能修订的影响，因此难以刻画某次具体修订究竟改变了什么，也使更新倾向于选择受益于有利状态的修订，而非真正改进过程的修订。我们提出 HMED（事后元经验蒸馏），一种为自我改进智能体构建元经验的机制。HMED 回溯修订所源自的已完成事件，并从同一恢复后的发现状态重新执行当前版本和修订后的元技能，从而能够在共享条件下观察与修订相关的变化。每次比较都会被蒸馏为一条元经验，即一种可供未来更新复用的结构化记录，因此，即使最终未被保留的修订也能提供学习信号。在三个交互式智能体基准以及开源和闭源模型上，HMED 相较于强基线方法持续提升技能发现性能，使元技能学习不再局限于分支结果，而是进一步关注改变改进过程所带来的后果。
LLM 智能体通常缺乏在新环境中可靠行动所需的操作知识，因为它们必须自行发现特定工具的行为方式或环境惯例。由于没有对过往尝试的记忆，它们会在不同任务中重复相同的错误，导致更多任务失败和更长的执行轨迹。为解决这一问题，智能体系统通常依赖人工编写的指南，或依赖通过训练任务和 oracle verifier 构建的程序性记忆，而这两种方式都需要预先了解环境。我们提出 DAEDALUS，一种无需现有任务或 oracle verifier、通过自生成练习来引导构建可复用智能体记忆的方法。DAEDALUS 配对使用两个智能体：探索器与环境交互，以生成具有挑战性但可解决的任务；求解器则尝试完成这些任务。每次求解器失败后，系统都会提炼出一条启发式规则，只有当求解器在相关上下文中使用该规则反复成功后，这条规则才会被接受。这些结果还会为探索器提供反馈，帮助其调整未来任务的难度。随后，已接受的启发式规则会被整合到记忆库中，供测试时使用。在 AppWorld、$τ^2$-bench 和 AutomationBench 上，DAEDALUS 相比无记忆基线将平均成功率最多提高 15.9 个百分点，将 pass^5 最多提升至 2.2 倍；与使用训练任务的方法相比，其性能具有竞争力，且推理成本低于大多数此类方法。我们表明，即使探索预算较小，也能取得性能提升；其启发式规则同样能帮助其他模型系列的智能体。进一步的消融实验表明，求解器轨迹提供了推导有效启发式规则所需的关键信息，而将早期发现进行因子化则能提高探索的成本效率。除构建记忆外，我们还发现，DAEDALUS 生成的任务可以作为基准任务的替代，用于根据性能对模型进行排名。代码和相关资源：www.github.com/illuin-tech/daedalus】【。
开发用于临床任务（如医学问答（QA））的可靠模型，受到高质量专家标注训练数据有限的严重制约。严格的隐私要求，以及在资源受限的临床环境中难以使用大型开源语料库或专有云 API，进一步加剧了这一挑战。为应对这些障碍，我们提出 SAGE（Semantic Anchor-Guided Evolution），一种新型数据合成框架，使本地部署的小型模型能够生成高质量医学训练数据。SAGE 利用 MeSH 等轻量级、公开可用的分类体系作为语义锚点，引入结构化先验，从而有效引导并约束数据生成过程。SAGE 的核心在于迭代交替进行原子式（基于单个概念）和关联式（基于关系）的合成，从极少量种子数据中逐步构建训练数据。该方法无需大规模医学文档集合，也不依赖外部 API，为本地数据创建提供了实用解决方案。在多个医学问答基准上的大量实验表明，使用 SAGE 合成数据微调的模型，其表现始终优于采用自生成或传统基于文档范式训练的模型，凸显了医学 LLM 开发在数据效率和资源利用方面的切实提升。代码地址：https://github.com/DIaacKr/SAGE】【。
针对语音匿名化的攻击者 ASV 系统主要在英语环境下进行研究，其在多语言场景中的表现仍基本未得到探索。传统 ASV 研究表明，声学信息和上下文信息对于多语言说话人验证都十分重要。受此启发，我们探究这一结论是否同样适用于针对匿名化语音的攻击者 ASV。我们在多语言匿名化语音上评估了声学型和内容导向型攻击者，并构建了一个多语言语音转换数据集，以提升跨语言泛化能力。结果表明，攻击者的有效性取决于匿名化语音的语言学实用性。总体而言，声学型攻击者取得了更好的性能。然而，当语言信息得到较好保留时，与语音失真更严重的条件相比，内容导向型攻击者与声学型攻击者之间的性能差距会缩小。多语言语音转换数据集进一步提升了性能，并部分缩小了跨语言差距。这些发现凸显了建立更全面的攻击者建模与评估协议的必要性：评估时应同时考虑隐私性和实用性，而不应依赖单一的攻击策略\footnote{完整代码、预训练模型以及 MultiVC Dataset 链接见：https://github.com/monkeyDarefeen/DAST。
Product Identification has sprung up to become one of the most challenging problems in the automation of the retail industry. With the new industry 5.0 standards, automated inventory management, and catalog creation tasks are vitally important. Object identification models have emerged as a viable answer with their unprecedented identification and localization accuracy. However, the close-knit rack design of supermarkets generates the problem of angle variation in capturing images. The angle-variant densely packed images(a single image contains many objects) become overwhelming for these models alone. In this paper, we try to supplement object detection models with traditional Hough transform (HT) and homogeneous estimation concepts. We study the effect of rectified images using homography estimation and hough transform and their limitations on the problem of grocery identification. We make a case for creating a new dataset to test the effects of such rectification and produce analytical results on different scenarios of angle variation and object densities per image. Extensive experiments on different object detection models suggest that image rectification of angled images improves the detection accuracy of grocery products in images. The results also highlight the limitation of rectification on the angle of image capture and the object density of the image.
Zero-shot coordination in embodied settings requires acting while the partner is intermittently out of view, leaving existing methods with ambiguous partner representations and uncertainty over hidden partner states. We propose Predicting Intention of Partner (PIP) to jointly address these challenges. PIP uses a Joint-view VAE to distill richer training-time evidence from the union of both agents' local observations into a partner representation available from local observations alone. Partner-state Belief networks further infer the partner's hidden location and behavioral tendencies from the ego agent's interaction history. We evaluate PIP in Burrito-PO, Overcooked-PO, and a Melting Pot substrate, together with a human evaluation in Burrito-PO. PIP attains the highest mean performance among the compared methods across all three benchmarks. Human evaluation and diagnostic analyses further support coordination with unseen partners and the contributions of both components under partner occlusion.
I am a developer of https://www.photopea.com, a popular photo editor that runs in a web browser. Many people are asking AI models to take the Javascript code from my website, remove all ads from it, and they publish such a "new product" on Github for everyone to download. There exist tens of such repositories on Github. I want my website to be the only source of a stable version of my program Photopea. I even received emails from people complaining about something in Photopea, and it took several emails to figure out that they are not using Photopea.com (so it ruins my reputation a little). I reported it to Github on the 4th of September 2026: https://www.photopea.com/g/XKoqqIGv Today, a month later, I received this response: Thank you for submitting a DMCA takedown notice. We've reviewed the information you've provided, and based on the facts presented to us, we're unable to confirm a violation of 17 U.S. Code § 1201. What do you think I could do? Do you think I should look for a lawyer to deal with it outside the digital world? I really doubt that a real person ever looked at my report, and they probably send this response automatically to 99% of people. Comments URL: https://news.ycombinator.com/item?id=49982498 Points: 29 # Comments: 17
My comment on EmbeddingGemma 2 — Hacker News. I really appreciate that EmbeddingGemma 2 is under the Apache 2.0 license. For embedding models in particular, I don't think it makes sense to use a closed, proprietary, hosted-only model. Most applications of embedding models involve calculating thousands or even millions of embedding vectors and storing them for later comparison. If your model is proprietary, the vendor is likely someday going to decide to stop offering that model. They'll have a better model to replace it, but you still need to pay to re-calculate those millions of stored existing vectors. (In April 2024 OpenAI offered to "cover the financial cost of users re-embedding content with these new models" - https://openai.com/index/gpt-4-api-general-availability/ - but I don't think that's something we can rely on from every provider.) Notably, I don't want to host the model myself . I'd much rather pay a provider for a hosted model while knowing that if they ever stop hosting it I can run the open weights version myself - or find another vendor who can do that for me. Tags: google , ai , generative-ai , embeddings , gemma
AI storage strategy is taking on a larger role as enterprises move from successful experiments to systems that must deliver dependable business results. How organizations store, serve and manage data increasingly shapes the performance, cost and practicality of artificial intelligence deployments. The challenge extends beyond choosing faster hardware. Evolving workloads require enterprises to rethink how […] The post Three key insights you may have missed from theCUBE’s coverage of the Supermicro Open Storage Summit interview series appeared first on SiliconANGLE .
I'm obsessed with Egyptology. During covid, I wanted to get a hieratic tattoo. For those who don't know, hieratic is the cursive form of hieroglyphs that was used in day-to-day life in ancient Egypt. It's as old as hieroglyphs. I then went on a mission to have a sentence translated into hieratic. Luckily, in 2022, I was able to commission an Oxford Egyptologist to write it for me. After that, I took it to a hieratic specialist I found on Instagram to redraw it with "biblical" accuracy. When generative AI came out, I realized that models very rarely identify hieratic and almost never translate it correctly. Every single time a new model came out, I ended up testing my sentence and hieratic texts in general, and I realized this could actually become a benchmark. Since so many benchmarks are becoming saturated, we need to find weird ways to test new models. That's why I decided to build HieraticBench. The first version of the benchmark has 268 items, including real documents, signs, and other Egyptian scripts, which I used as controls, as well as my unpublished sentence in two different renditions. The current state-of-the-art foundation models are hilariously bad at identifying my sentence, especially the one the Egyptologist wrote with her pen. Some of them said Tibetan, Urdu, Korean, and, funnily enough, "Reformed Egyptian," which is from the Book of Mormon. On real documents, the best model in the benchmark identified hieratic correctly 95% of the time, and the best score for reading individual signs was around 13%. That's when we tell the model specifically that it's hieratic. None of them can really translate a sealed sentence that's never been published and has no answer key. I haven't run every model on every task just to keep costs down. My next steps were to run Astra and Gemini 3.1 Pro on real documents. Currently, only the Claude models have done the sign readings. And just FYI, I have no experience whatsoever running benchmarks or evaluations, so not 100% sure I scored things right. Things we need: - Running the models I haven't covered - More sealed sentences and labeled signs If you know hieratic or have an Egyptologist friend, reach out at hieraticbench@veeza.ai. Source code: https://github.com/alymoursy/hieraticbench I'll be around to answer questions. Comments URL: https://news.ycombinator.com/item?id=49982822 Points: 2 # Comments: 0
Google DeepMind's EmbeddingGemma 2 maps 5 input types into one 768d space and ships today under Apache 2.0. The post Google DeepMind Releases EmbeddingGemma 2, a 740M Open Multimodal Embedding Model Built on Gemma 4 appeared first on MarkTechPost .
Hi HN, I'm Weilun, cofounder of OpenChart. We built OpenChart because we wanted Claude and Codex to interface with the markets, like most of humans do, through charts rather than through CLIs. With your existing AI plans, your agent can work directly with your charts: annotate a setup, write an indicator, investigate a market move, or create an alert. The chart, conversation, and research live in the same workspace. OpenChart alerts on any market move: draw any shape on a chart and attach an alert. When price crosses it, the alert triggers your agent to investigate the move and save the research in your workspace. The agents can work with charts, indicators, and alerts. Custom indicators and alert conditions are written in [Tea]( https://github.com/longsurf-ai/tea ), a programming language we built for the markets. We believe software shouldn’t impose artificial friction on your workflow. OpenChart is built on that belief with no software cap. The source code is publicly available, and you can use free market data from day one. We also offer optional paid cloud market data for lower latency. Everything is private. Your charts, research, and indicator code are stored on your own computer, with the app's backend running locally too. We currently support only macOS on Apple Silicon, but we're expanding to more platforms. Website: https://openchart.co/ Download: https://downloads.longsurf.ai/openchart/darwin/arm64/OpenCha... We built OpenChart because we want to explore how assets can be co-manages by agents and humans. I'd especially like to hear how you are using agents to help with your investment journey. If you like what we're building, a star on GitHub would mean a lot to us: https://github.com/longsurf-ai/openchart Comments URL: https://news.ycombinator.com/item?id=49979793 Points: 1 # Comments: 0
Suppose a committee, expert panel, or other group is making judgments on some issues, where these may be not just yes/no-questions, such as whether a defendant is guilty, but also variables with many possible values, such as macroeconomic or meteorological variables or travel directions. Furthermore, there may be interconnections between different issues, as in the case of economic or climate variables. How can the group arrive at "intelligent" collective judgments, based on the group members' individual judgments? We investigate three challenges raised by this judgment-aggregation problem. First, reasonable methods of aggregation (such as defining the collective judgment for each issue as the average or median judgment) can produce inconsistent collective judgments. Secondly, many methods of aggregation are manipulable by strategic voting. Finally, not all methods of aggregation are conducive to tracking the truth on the issues in question. We prove new impossibility or possibility theorems on all three challenges, identifying what it takes to produce collective judgments in a consistent, non-manipulable, and truth-tracking manner and thereby to achieve collective intelligence through aggregation. Overall, the median method, though imperfect, performs reasonably well. We also note the relevance of our analysis for non-human group decisions.
语言模型可能会赋予某个正确答案比任意单个错误答案更高的概率，但仍然经常采样出错误答案，因为所有错误答案合在一起占据了更多概率。幂分布将每个完整答案的概率提升到大于 1 的幂次后重新归一化，把概率向模型认为最可能的答案集中（即锐化分布）。从该分布中采样无需改变参数即可提升推理能力，但每个查询需要大量经过评分的候选答案。我们表明，模型也可以通过训练，在一次生成中直接生成这样的答案。在线策略幂蒸馏（OPPD）运行一个序贯蒙特卡洛采样器：待训练模型生成候选答案，冻结的教师模型使用幂分布为这些候选答案加权；在最大似然更新中，也使用相同的概率为每个答案加权。在相同温度下，相比未训练模型，训练后模型的单次生成准确率在 MATH500 上最高提升 23.0 个百分点，在 GSM8K 上最高提升 27.3 个百分点；单次生成的得分比使用 64 个候选答案进行的已发表幂采样高出 2.4 和 3.5 个百分点，相当于恢复了未训练模型使用 16 个候选答案时增益的 94%。作为参考，在使用同一检查点和相同预算、通过已验证奖励训练的 GRPO 对比中，OPPD 在 MATH500、GSM8K 和 AIME 上分别高出 3.8、4.0 和 5.4 个百分点，且无需参考答案；二者具有互补性，在 GRPO 之后应用 OPPD，最高可再提升 9.3 个百分点。仅使用数学数据训练时，OPPD 还能将 HumanEval 准确率最高提升 5.3 个百分点。一个损失系数即可将模型所吸收的锐化指数从 1.19 调整到 2.02，而普通在线策略蒸馏的锐化指数为 1.14；该指数的提升主要体现在模型自身生成的答案上。无论模型系列还是规模如何，增益都能保持，包括已经通过已验证奖励训练过的模型；对于后者，降低温度并不能带来任何提升，而 OPPD 可使 MATH500 得分增加 4.4 个百分点。代码：https://github.com/ArminAzizi98/OPPD。
我们将 Kokoro-82M——一个广泛使用、拥有 54 种音色的开放文本转语音模型——蒸馏为 Paradee，一个能够使用其中一种音色说话的 8.07M 参数模型。Paradee 保留了 Kokoro 的架构，但将各层大幅缩窄，并分别针对冻结的教师模型训练其两个部分。它的参数量减少了 10 倍，所需计算量降低了 15 倍。我们首先使用教师模型合成语料库，并保留其时长、音高、能量和音素特征。随后，我们训练一个小型文本侧网络来预测这些数值，并训练一个小型解码器，将教师模型保存的数值转换为教师模型的音频：先使用频谱损失训练，再进行对抗训练。最后，我们连接这两个部分，并将权重量化为 int8。它无需学习对齐，也无需联合训练，并且可以在一台笔记本电脑上运行。以 int8 存储时，Paradee 的大小为 8.5 MB，在单个 CPU 线程上运行速度达到实时的 25 倍；其 UTMOS 得分为 4.41，而教师模型为 4.52。学生模型起初保留了轻微的嗡嗡声，我们追溯发现其来源是 2 至 8 kHz 范围内有声语音的相位。在合成后应用相位锁定滤波器可以消除其中大部分噪声，且无需训练或增加额外参数。代码、模型文件和音频样本位于 https://github.com/sahilmahendrakar/paradee
开源网页智能体如今已经足以执行逼真的浏览器任务，但使用强化学习训练它们仍然依赖弱监督：二元任务成功信号过于稀疏，难以进行信用分配；而前沿语言模型评审器的调用成本过高，无法在每一步都调用，也不能假定其在部署时可用。我们提出 CLIFT，一种以保形自验证为核心的训练与测试时扩展方法。在训练期间，智能体针对自身的 rollout 回答自然语言验证问题；Compositional Conformal Certifier 仅保留其 URL 条件证据与训练时评审器一致的问题信号，通过极性感知提升分配带符号的信任权重，并将得到的验证器分数融合到逐步奖励中，同时绝不会从评审器基线中扣减任何分数。在测试时，同一经过认证的问题库被冻结，并作为 Conformal Trajectory Selection（CTS）的结构化证据重复使用：智能体采样一条贪心 rollout 以及一条或多条多样化重试轨迹，自验证器总结每条 URL 轨迹，然后通过一种保守的多数投票规则决定是否放弃当前最优结果，而无需调用任何外部评审器。这一机制支持三种设置。在 WebArena Infinity 上，CLIFT 在开源网页智能体中取得了当前最佳性能。在 VisualWebArena 上，使用开放模型训练的问题库可在测试时迁移到 GPT-5.5，并在标准测试框架下达到当前最佳性能。在 Online Mind2Web 上，无需针对该基准训练智能体，仅通过翻译经过认证的问题库，就能在零样本评估中提升一个实时网页智能体的表现。总体而言，这些结果表明，保形自验证能够将昂贵的评审器反馈转化为可复用的训练信号，以及无需评审器的测试时扩展信号。
长时域视频世界模型需要持久记忆，以在长时间推演中保持场景一致性。Softmax 注意力通过不断增长的 KV 缓存保留完整的生成历史，而循环线性注意力则将历史压缩为固定大小的状态，大幅降低内存开销。然而，我们发现 Gated DeltaNet（GDN）存在严重的长程遗忘问题：来自遥远但相关场景的信息会在后续状态更新中逐渐衰减。为解决这一局限，我们提出 HLA-WM，这是一种无需训练的混合线性注意力框架，将粗粒度的几何引导检索与细粒度的循环线性状态计算相结合。HLA-WM 利用 GDN 的仿射结构，缓存紧凑的分块状态转移摘要，利用相机几何信息检索与场景相关的历史块，并将其重新组合为针对特定查询的循环状态。在时长为 $60$ 秒的 SANA-WM-Bench 上，HLA-WM 无需额外训练即可改善基础自回归生成器的全部六项重访一致性与相机控制汇总指标，其中 PSNR 提升 $0.74$ dB，旋转误差降低 $28.5\%$。这些改进在下游细化处理后依然保持，并可泛化至 MBench-A：在涵盖全部四个子集及所有受评估推理模式的 $547$ 个样本上，HLA-WM 均持续改善全部三项重访一致性指标。在 $60$ 秒的上下文下，相较于完整 KV 缓存，HLA-WM 将历史状态内存占用降至原来的 $1/12$，同时推理吞吐量最多仅下降 $1.6\%$。这些结果表明，可选择性寻址的循环记忆能够改善长程场景回忆能力，同时保留 GDN 的效率优势。项目主页：https://caesarhhh.github.io/hla-wm/
图像和视频扩散模型会为每个区域分配相同的计算量，即使目标场景要求不同程度的细节。细节的空间分布通常可以在生成前预先判断，从而确定可以减少计算量的位置。我们提出 Level-of-Token（LoT）Diffusion，这是一种将这些信息转化为显式多分辨率 token 布局（Level-of-Token 布局）的框架，以实现自适应且高效的生成。Token 表示大小和形状各异的矩形图像块，在需要细节的区域分配更精细的 token，在其他区域则使用更粗粒度的 token。我们通过分块非对称流参数化以及多分辨率 token 嵌入，使预训练扩散 Transformer 适配 LoT 布局：在每个去噪步骤中保留全分辨率流预测，同时仅处理缩减后的 token 序列。LoT Diffusion 在保留预训练生成先验的同时，实现了布局自适应生成。我们展示了基于语义掩码、边界框、纹理方差和景深线索，以及智能体规划所生成布局的 LoT 效果。在图像和视频生成任务中，LoT 均实现了良好的质量—效率权衡，其显著的加速效果取决于布局的 token 预算。我们的项目网站位于 https://georgenakayama.github.io/lotdiffusion/】【。
A bipedal robot cannot deviate from its path to avoid an obstacle without disturbing its balance, and this coupling is most severe on underactuated platforms such as the biped considered here, which has four actuated joints per leg and no hip or ankle roll. This paper presents a Hierarchical Reinforcement Learning (HRL) framework in which a High-Level (HL) policy observes the robot pose, 36 raycast proximity measurements, moving-obstacle states, and a receding-horizon local goal, and outputs a body-velocity command $(v_x, v_y, ω_{yaw})$ every ten control steps, while a velocity-conditioned Low-Level (LL) policy tracks each command through PD-controlled joint targets. Both policies are trained jointly with Soft Actor-Critic (SAC). Because the converged gait is task-agnostic, it is frozen and driven by classical planners over the same command interface, yielding three controlled baselines: SAC+A*, SAC+RRT*, and SAC+APF. Across 100 evaluation trials per method in randomized PyBullet environments, the proposed method reaches the goal in 98.0% of static and 88.0% of dynamic trials, against at most 78.0% and 68.0% for the planner hybrids, with path lengths within 4% of the A* reference, and ablations confirm that each observation channel and reward term contributes materially to this performance.
图形用户界面（GUI）Agent 已成为跨多种应用自动化复杂数字工作流的一种前景广阔的范式。然而，训练能力强大且具备良好泛化性的 Agent，从根本上依赖海量、高保真的视觉-动作轨迹，而这类轨迹极难获取。人工演示难以规模化，而现有的 GUI 世界模型依赖文本描述或 HTML 渲染，舍弃了图标和布局样式等关键的像素级视觉细节。为解决这一问题，我们提出 Infinite-Dreamer，这是一种由像素级图像编辑世界模型驱动的、无需仿真的数据合成方法。通过将 GUI 状态转换视为图像编辑任务，我们利用视觉语言模型（VLMs）将动作引起的 UI 变化描述为结构化的差异文本（delta-text）。随后，我们对图像编辑骨干模型进行微调，以可控地合成逼真的屏幕截图转换。我们利用该模型生成单帧视觉鲁棒性数据和多步想象轨迹。为验证方法的有效性，我们仅使用合成数据对 Qwen3-VL 基线进行微调，得到 Infinite-Actor，并在 AndroidWorld、MobileWorld 和 AndroidControl-Curated 基准上进行评估。Infinite-Actor 在不同规模下均持续优于 Qwen3-VL 基线：Infinite-Actor-8B 将 AndroidWorld 的 Pass@1 提升了 4.45，并使 MobileWorld 的 Pass@3 成功率接近翻倍；Infinite-Actor-2B 则将 Pass@1 提升了 9.05。代码可在 https://github.com/swaydy-n/Infinite-Dreamer 获取。
从 FPN 到 BiFPN，特征金字塔方法通过融合多尺度特征，在人脸检测中取得了出色的性能。然而，在无约束条件下检测人脸仍然十分困难，例如小尺度、遮挡和极端姿态等情况，因为这需要局部融合无法建模的全局跨尺度依赖关系。Mamba 等状态空间模型通过将特征作为序列进行扫描，以线性复杂度提供全局上下文，因此为解决这一问题提供了有前景的方向。然而，这种扫描需要将两个金字塔尺度合并为单一特征图，而合并方式决定了跨尺度结构能否得到保留。求和会在扫描前将两个尺度压缩到一起，使扫描无法利用跨尺度结构；而拼接虽然保留了两个尺度，却会带来远高得多的计算成本。为此，我们提出了 Weave Mamba Fusion (WMF)，按列交错两个相邻的金字塔尺度，使水平双向 SS2D 扫描的每一步都从一个尺度移动到另一个尺度。通过部分通道处理和无参数解交织，WMF 在保留特征结构的同时实现了高效的跨尺度交互。将 WMF 集成到每个融合节点中，可得到 WeaveBiFPN，即 WeaveFace 检测器的颈部网络。在 WIDER FACE 上，WeaveFace 仅使用 0.34M 个参数和 1.16 GFLOPs，即可达到 91.41\% 的平均 AP，优于参数量低于 0.5M 的现有检测器。在 Hard 子集上，其提升最为显著，达到 87.14\% AP。代码已公开发布于 \url{https://github.com/dohun-mat/WeaveMambaFusion}】【。
多模态大语言模型（MLLMs）通过将大语言模型（LLM）与视觉、视频或音频等目标模态的编码器相结合，实现了强大的模态理解能力。然而，要提升 MLLM 对特定模态的能力，通常需要在大规模模态专属数据集上进行额外训练，从而产生高昂的数据收集和计算成本。模型合并提供了另一种选择，但对于数据稀缺、单样本规模较大或具有领域特定性的模态（例如音频和视频）而言，模型合并通常难以实现，因为同模态的模型变体很少可用。在本文中，我们刻画了一种有趣的不对称现象：将一个对齐良好且数据丰富的源模态 MLLM 合并到一个数据稀缺的目标模态 MLLM 中，可以在目标模态自身的基准测试上显著提升其性能。我们的理论与实证分析表明，这种增益源于模态特定词元与文本词元之间对齐程度的提升，而这种提升由更强的供体模态所诱导。具体而言，我们推导出一个关于互信息的下界，该下界随对齐相关量单调变化，并且与下游 MLLM 性能高度相关。基于这一原理，我们提出了方向性跨模态对齐迁移（DCAT），这是一种新型框架，可将强大且对齐良好的源（供体）模态中的文本对齐迁移到较弱的目标（接收方）模态，在无需进一步微调的情况下提升目标模态的性能。我们进一步表明，这一增强对齐的目标函数存在一个闭式权重空间解，仅需利用一个小规模校准集即可计算得到。DCAT 的性能优于现有模型合并方法，为跨模态对齐迁移提供了一条高效路径。包含代码的项目主页见 \url{https://seohoiki3215.github.io/DCAT_project_page}。
神经音频编解码器将波形压缩为紧凑的离散 token，为语音语言模型、实时通信和大规模音频存储提供基础。几乎所有主流设计，包括残差向量量化、有限标量量化和单码本变体，都遵循 VQ-VAE 模板，通过学习得到的码本或固定的标量网格对编码器隐变量进行划分。我们探讨这种划分是否是必要的。我们提出 GS-Codec——一种神经语音编解码器，其瓶颈并非量化器，而是参数化信号分解。我们将 Gaussian splatting 从三维场景重建改用于一维隐变量。内层优化循环会将编码器的每个片段拟合为一组一维 Gaussian 基元的加权和。随后，解码器根据渲染得到的总和重建波形。为避免推理时执行这一迭代内层循环的成本，我们还训练了轻量级的 GS Predictor Net，通过单次前向传播回归这些基元的参数。编码器和解码器通过内层循环进行端到端训练，整个训练流程中完全不使用量化器：瓶颈就是分解本身，而标量量化仅在训练完成后应用于拟合得到的参数。我们的表示不依赖离散码本阶段来控制比特率，而是提供了细粒度的码率—质量权衡：单个训练好的检查点即可通过改变基元数量和每个参数的位深，在训练后控制比特率，无需重新训练。在相近比特率下，GS-Codec 在说话人相似度（SIM）、可懂度（STOI）和感知质量（UTMOS）方面达到或超过 EnCodec 和 DAC 等成熟的开源编解码器，同时实现了相当的语义性能（WER）。代码和音频样例见 https://ronaluf.github.io/gs-codec/
大型语言模型（LLM）已逐步发展成为自主智能体的核心。在此基础上，基于 LLM 的多智能体系统（MAS）将多个智能体协调为协同团队，以完成超出单个智能体能力范围的复杂任务。这类系统的有效性不仅取决于智能体本身，还取决于协作机制的设计与组织方式。需要注意的是，现实世界中的协作通常具有部分可观测性：由于物理或隐私相关的限制，每个智能体只能获取环境中的部分信息。然而，现有许多多智能体基准测试都假设全局可观测，并且对协作机制的系统性评估支持有限。为弥合这一差距，我们提出了 MASBench，这是一个在部分可观测约束下设计的多智能体协作基准。该基准由三个渐进式任务类别组成：推理、调度和游戏。通过这一结构，我们逐步评估三种具有代表性的协作机制：协议、记忆和路由。MASBench 还提供了确定性的评估指标，包括性能得分、通信成本和成本效益，用于刻画协作成果与通信开销。在多种 LLM 主干模型和机制配置上的实验，为 MAS 的有效设计提供了实证指导。代码地址：https://github.com/BUPT-GAMMA/MASBench
多智能体辩论能够提升推理能力，但往往无法胜过简单的多数投票。我们认为，成功的辩论需要两种不同的机制：提案供给必须让正确答案浮现出来，而读出机制则必须在投票未能选出正确答案时识别出该答案。我们通过“可恢复余量”对第一项要求进行形式化，该指标衡量的是这样一类情况：正确提案已经出现，但多数答案是错误的。针对第二项要求，我们提出 Latent Verification Debate（LVD），这是一种记账模型：候选提案在最终生成之前会获得与答案相关的验证证据。通过受控的固定提案干预，我们以等价的同伴支持单元估计这一潜在效应，并表明，在提案供给保持不变的情况下，正确证据会改变答案概率和生成决策。为改善提案供给，我们利用带标签和无标签的覆盖目标，从 neural-thicket 智能体构建智能体社会。在两个主干模型和预算匹配的推理基准上，基于覆盖率选择的智能体社会能够增加互补性提案供给，并在重复随机评估中提升总体准确率。轮次级控制进一步表明，交互带来的收益超出了直接将同一最终决策器应用于初始提案所获得的收益。这些结果表明，提案覆盖与对真值敏感的证据使用，是辩论优于投票的互补条件。代码可见于 https://github.com/Wang-ML-Lab/when-debate-helps】【。
AI 助手会收到一些请求，其中缺少实现良好结果所需的信息，例如用户的偏好或目标。此时，它们要么进行猜测，要么在继续操作前询问更多信息。我们将这一问题重新概念化为信息价值问题：助手应获取那些因缺失而导致用户效用产生最大可避免损失的信息。这一点通常无法事先得知；相反，助手必须对其进行预测，从而在有限的用户交互次数下实现最优分配。我们在图像生成场景中具体化了这一问题，并利用多轮模拟用户构建了一个强化学习框架，以在不确定性下最大化效用恢复。在一项预注册研究中，76 名人类参与者进行了 456 次交互会话；该方法帮助用户以显著更少的问题、更短的总交互时间和更低的成本，更好地匹配参考图像。这表明，可以采用一种简单且可扩展的框架来训练语言模型助手，通过提出信息量更高的问题，更好地消除用户意图的歧义。
结构图广泛用于科学、工程、流程和空间等领域，以表示复杂系统及其关系信息。近年来，视觉语言模型（VLM）在识别图表元素并对其内容进行推理方面的能力日益增强，但完整图表拓扑提取仍相对缺乏研究。本文研究从图表到图的拓扑提取任务：提取图表中的所有实体及其之间的完整关系。为支持该任务的大规模有监督训练和系统评估，我们提出了 Knossos，这是一个涵盖六个多样化领域、包含 19,200 张图表的基准数据集，其中包括 245,179 个节点和 439,740 条边。其符号化生成过程能够在渲染后的图表与完整拓扑、关系类型及连接器几何形状的标注之间提供精确对齐。为应对完整拓扑提取的建模挑战，我们还提出了 Ariadne，这是一个结构化框架，将该任务分解为节点清单提取和源条件边预测。大量实验表明，在 Knossos 上进行训练能够显著提升较小型开源 VLM 的完整拓扑提取能力。在监督条件匹配的情况下，Ariadne 进一步优于一步式提取方法，展现了结构化分解带来的额外收益。在 Knossos 上评测的方法中，Ariadne 取得了最高的平均 Edge F1；同时，两种骨干网络变体在真实世界外部基准上的表现也均优于未经适配的对应模型。代码和基准数据集见 https://github.com/bangwayne/knossos_Ariadne_Public。
Are you tokenmaxxing, or have the bean counters started setting spend targets (or limits)? What is your current user spend and will it be higher or lower in 2 months? Comments URL: https://news.ycombinator.com/item?id=49972842 Points: 1 # Comments: 0
Together AI has released Together Link, a free, MIT-licensed CLI that connects Claude Code, Codex, OpenCode, Pi, and the Claude and ChatGPT desktop apps to open models like Kimi K3 and GLM 5.3. One install command sets it up, and an Auto router picks a model for each session. Together claims savings of over 50%. The post Meet Together Link: A Free CLI That Runs Open Models Like Kimi K3 and GLM 5.3 Inside Claude Code, Codex, and OpenCode appeared first on MarkTechPost .
来自 Z.ai 的 GLM 5.3 现已在 Amazon Bedrock 上可用：这是一款拥有 7530 亿参数的混合专家模型，专为编程和长周期智能体任务打造。了解如何通过兼容 OpenAI 的 API 调用它，如何利用提示词缓存降低成本和延迟，以及如何使用开源 Strix 智能体运行授权安全测试。
Reflection AI has introduced Beam, its first open-weight model. It is a 501B sparse Mixture-of-Experts model with 23B active parameters, built for coding and agentic work. Reflection says it matches GLM-5.2 on reasoning with 3 to 4x less inference compute. Apache 2.0 weights are due later in October 2026. The post Reflection AI Introduces Beam: A 501B Open-Weight MoE Model With 23B Active Parameters for Coding and Agentic Workloads appeared first on MarkTechPost .
Hi, my name is Andrew and I'm the founder of mainbrella co. Just a few days ago Cloudflare launched their container sandbox service. That inspired me to see what I could build in terms of a better e2b.dev or daytona.io type system. It is all open source GPL 3.0 and we offer the same services as the other players. We have no free plan (our base builder plan is $5 a month) but you are welcome to run the code in your own Cloudflare account. The numbers are pretty good in terms of startup speed. First test we ran was: Create a fresh sandbox, get a usable shell 381 ms, 30 out of 30 tries worked the same speed. Then 100 simultaneous sandbox starts 680 ms avg 100 / 100 successful 100 distinct machines running at once! https://mainbrella.com/ https://github.com/mainbrella/backend https://github.com/mainbrella/web Looking for feedback, feature requests and yes our first paying customers who we will cherish! Thank you HN! -Andrew Arrow Comments URL: https://news.ycombinator.com/item?id=49971503 Points: 1 # Comments: 0
AI trust increasingly determines whether enterprise AI scales. As organizations move beyond pilots and put models and agents into operational systems, the question is no longer whether AI performs well in isolation. Enterprises also have to determine whether the infrastructure, data and applications underneath it can withstand cyberattacks, data integrity failures and operational disruption. That […] The post What to expect during the AI Cybersecurity and Resilience Summit appeared first on SiliconANGLE .
I have been working on Rashomon, an open-source execution recorder for AI coding agents. The basic idea is that the agent transcript is not ground truth. Rashomon keeps its own record of what happened and compares it against the agent’s account. For Claude Code, it records things like: * shell commands, exit codes, and whether they may have written files * tool calls and outcomes * test commands and their outcomes * subagent lifecycle and ordering * timestamps and parent/child relationships * keyed digests of command/tool inputs and working directories It builds a timeline from these events. It does not store prompts, responses, file contents, or tool output. For example, it now detects patterns like: `test fails → only test files change → same test passes` and: `test command fails → same command passes → no corresponding file edit` These can indicate that an agent changed a test instead of fixing the underlying issue. Rashomon currently runs as a Claude Code plugin or through the CLI. CLI: `rashomon watch` then: `rashomon report` Plugin: `/rashomon:report` There are still edge cases around cwd, git state, test-selection arguments, generated files, retries, and parallel execution. I would be interested in feedback from people building agents on what execution signals you would want Rashomon to capture or correlate. GitHub: https://github.com/altrace-dev-role/rashomon Comments URL: https://news.ycombinator.com/item?id=49968377 Points: 1 # Comments: 0
Aleph Alpha 发布了 Kolibri，这是一款德英双语混合专家模型，拥有 780 亿个参数，其中每个 token 约有 30 亿个参数处于激活状态。超过 21% 的训练数据为德语。该模型在德国和芬兰的 768 个 B200 GPU 上训练完成，模型权重以 Apache 2.0 许可证免费提供。文章《Aleph Alpha 发布 Kolibri，一款彰显欧洲 AI 主权理念的开放权重模型》最初发表于 The Decoder。
Today, I’m talking with Senator Adam Schiff, a Democrat from California. Sen. Schiff sits on a number of committees with oversight into tech and AI: intellectual property, antitrust, privacy and technology — it’s all there. I really wanted to ask him about how we might regulate anything related to the tech industry at this moment […]
An MIT expert committee warns that AI is eroding key parts of the college experience: office hours, study groups, and the school's flagship research program for undergraduates are all fading. Trust between faculty and students is breaking down, and some professors are even considering AI agents instead of students as research assistants. The university most closely tied to AI's origin story is now calling for a complete overhaul of higher education. The article AI is eroding office hours, study groups, and the trust between faculty and students, MIT report finds appeared first on The Decoder .
arXiv:2608.19779v2 Announce Type: replace-cross Abstract: World models provide digital simulacra of the true world, allowing agents to be trained and tested before costly real-world deployment. At each time step, they receive an action and generate an observation and reward matching the statistics of the true world. In complex environments where present outcomes depend on events far in the past, this requires memory. One might expect that, by increasing memory, we can always build a model accurately enough to align the optimal agent policies of the real and virtual worlds. We show that this is false for classical world models, even when the true world itself is classical. We construct true worlds for which every finite classical model fails along the same possible trajectory: it either loses the ability to distinguish actions when the true world clearly prefers one, or repeatedly assigns the highest expected reward to suboptimal actions. Its expected-reward estimates also retain a nonvanishing average error. In contrast, each such true world admits a quantum world model using a single qutrit that reproduces it exactly: its reward estimates and preferred actions always match those of the true world, ensuring that the optimal policies of the real and virtual worlds remain perfectly aligned.
arXiv:2607.26984v2 公告类型：替换-交叉 Abstract：将原子结构映射为一组紧凑的几何描述符，是任何应用于原子尺度建模的机器学习方法中的关键步骤。一种强大且应用广泛的方法，可以理解为对原子对距离、三角形等的直方图进行离散化，从而得到一套具有对称性不变性的、以原子为中心的描述符层级。不幸的是，人们发现该层级中较低的层次（二、三、四近邻簇）并不完备：存在在对称性上无关的结构对，却具有完全相同的描述符。然而，迄今报告的所有“简并”现象，都可以通过考虑更大的邻居簇来构建描述符而得到消除。借助大语言模型，我们发现了若干三维结构：即使考虑最多包含七个邻居的簇，它们仍无法被区分；在对描述符采用实际可行程度的离散化时，即使达到任意阶数，它们仍然不可区分。构造这些结构的关键要素，可以追溯到不同学科领域中数十年前已知的研究成果：该模型能够找到相关文献，并认识到这些成果对于当前问题的重要性。我们认为，这项实验揭示了人工智能在科学研究中一种极具潜力的应用模式：在不同学术群体和应用领域之间转译研究成果，加速一个领域中的偶然发现转化为另一个领域中的突破。
arXiv:2607.22465v4 公告类型：replace-cross 摘要：现代企业级智能体部署由具有不同能力和成本的大型语言模型（LLM）异构池组成。现有的模型路由策略优化质量与成本之间的权衡，同时针对请求级别提供静态决策。较新的解决方案则将智能体路由视为任务级选择，并采用带有串行验证器的路由器反馈循环。然而，适用于同质工作负载的固定验证器可能无法泛化到异构的智能体任务批次（例如编程任务和通用对话任务）。此外，由于验证器位于反馈循环的关键路径上，在同时进行多个请求路由时，服务质量可能会受到影响。为缓解这些问题，我们提出了 ORACLE。这是一种具备并发感知能力的在线路由机制，将自适应路由与自适应验证相结合，以提供反馈。ORACLE 无需训练，可作为“反馈循环”即插即用地叠加在任意模型选择策略之上，首先对任务类型进行分类，然后动态分配适合任务的验证器。此外，我们针对并发请求开发了一种延迟反馈策略，大幅消除了验证器延迟对分发关键路径的影响。随后，我们提出了一种路由后分发调度器，即 DISC。DISC 在接纳任务时预留每个任务的峰值 KV 占用，并在减少等待时间带来的奖励增益超过准确率降低带来的奖励损失时，将任务分发到备用后端。在 SWE-bench、tau2-bench 和 Terminal-Bench 2.0 上进行的大量评估表明，与最先进的路由基线相比，ORACLE 可将准确率—成本前沿最多提升 7 个百分点；而 ORACLE 结合 DISC 后，可将程序吞吐量最多提升 1.8 倍。为促进可复现性并支持未来开发，我们在 https://github.com/ORACLE-org/ORACLE 开源了 ORACLE 的代码库。
arXiv:2606.27090v2 发布类型：替换交叉版本 摘要：KL散度和ELBO等全局目标被广泛用于贝叶斯推断中的分布差异度量。本文研究这类全局目标无法直接刻画的分布“局部质量行为”。我们引入并使用两种数学工具：（1）Mass Index，用于记录局部质量的多项式衰减尺度和对数衰减尺度；（2）正则化扩展KL（RE-KL），一种集合局部化的散度，即使存在奇异分量时也可以进行定义。Mass Index有助于刻画贝叶斯更新如何改变局部质量：（1）幂对数似然因子会使其发生显式偏移；（2）依赖参数的支集，或其平滑化形式，可能通过保留在参数值附近的质量大小改变局部尺度。利用局部RE-KL，我们证明了在两个KL方向下比较局部小球质量的绝对不等式、相对不等式和方向性不等式。综合而言，这些结果从局部角度对局部质量行为给出了理论解释。实验对局部行为进行了受控展示，并表明，在ImageNet上直至ResNet-50的贝叶斯神经网络变分后验中，这种方向性比较仍然可见。代码见 https://github.com/Forsythia0604/Local-Mass-Framework。
arXiv:2606.20627v2 Announce Type: replace-cross Abstract: Joint-Embedding Predictive Architectures (JEPAs) enable agents to plan in latent space by imagining the outcomes of candidate actions, yet task specification remains a bottleneck. Visual targets provide precise local gradients but poor distant guidance, while language is flexible yet limited by noisy cross-modal alignment or dependence on distinct large generative models. We introduce LAGO (Latent Goal Prediction from Language), a hierarchical world model in which a single predictor both forecasts action-conditioned dynamics and grounds language instructions as sequences of intermediate latent subgoals, training both modes with a single regression objective over a shared latent space. At each planning step, LAGO predicts a sequence of latent subgoals from a language instruction and optimizes an action sequence using a soft-minimum alignment cost that rewards subgoal proximity without enforcing a rigid path. Subgoals are repredicted as the agent's state evolves, turning a single long-horizon instruction into a sequence of locally tractable objectives. Across environments spanning navigation and manipulation, LAGO outperforms flat and hierarchical image-goal planners, with the largest gains on long-horizon navigation with curved, hazard-constrained paths, where it more than doubles the success rate of a goal-conditioned behavior cloning policy trained on the same demonstrations. Grounding language in the world model's own latent space also outperforms vision-language reward models, including one that plans with ground-truth dynamics.
arXiv:2606.18043v2 公告类型：replace-cross 摘要：通用机器人策略（例如视觉-语言-动作模型（VLAs）和世界-动作模型（WAMs））将强大的预训练骨干网络与富有表现力的生成式动作头相结合，后者通过在大规模机器人数据集上进行流匹配训练得到。尽管这些策略在机器人操作任务中展现出强劲的实证性能，但缺乏量化预测置信度以及检测动作可能不可靠情况的机制。这对其在非平稳环境中的现实部署构成了关键限制，因为模型不可避免地会遇到超出预训练分布的场景，并可能在毫无预警的情况下失效。为解决这一问题，我们利用小型模型集成中速度场分歧（VFD），推导出一种高效量化流匹配模型认识不确定性的方法。我们成功地将这一不确定性估计用于基于流的通用策略部署期间的故障检测和主动微调。对于后者，我们提出了SAVE——一种简单而有效的不确定性引导主动多任务微调方法，可减少将通用策略适配到新任务所需的高成本专家示范数量。我们在仿真环境和真实世界中，针对VLAs和一个WAM开展了实验。VFD能够产生校准效果更好的不确定性估计，可预测下游性能，并且相比现有方法以高出8个百分点的总体准确率检测故障。在三个真实世界任务中，在固定示范预算下，SAVE将最终平均成功率从39%提升至47%。我们的结果表明，使用VFD测量认识不确定性能够增强通用机器人策略的故障感知能力和适应能力。项目网站：tum-lsy.github.io/uq_generalist_policies。
arXiv:2606.07235v3 Announce Type: replace-cross Abstract: Answering questions about long multimodal documents requires distributing a fixed evidence budget across relevant facets in text, tables, figures, and slides while avoiding near-duplicates. We present \flowreader, which formulates evidence selection as a single minimum-cost flow problem with capacity limits over a multimodal content graph. Spectral decomposition identifies latent aspects of query-relevant content and allocates the budget among them in proportion to their spectral energy. These capacity limits enforce aspect coverage during routing without requiring a language-model planning call. Query-conditioned costs prioritize chains of relevant, mutually consistent evidence. Decomposing the optimal flow produces short evidence chains, which a vision-language model reads in parallel and a reasoner reconciles. On VisDoMBench with Qwen3-VL-32B, \flowreader\ achieves the highest macro accuracy ($68.9$), surpassing the strongest prior system by $2.7$ points, leading on three of five subsets and attaining the highest worst-subset accuracy. It uses a measured $17.5$ content nodes per query and maintains its lead at $12.9$. Ablation studies with a fixed graph, scorer, reader, and judge show that cost design drives accuracy, capacity limits preserve it while using about three-quarters of the reader tokens required by shortest-path routing without these limits on the same network, and spectral aspects align with LLM-generated sub-questions without a planning call.
arXiv:2606.00467v2 公告类型：replace-cross 摘要：大型语言模型（LLM）正越来越多地用于零样本标注和 LLM-as-a-judge 任务，但其可靠性取决于模型内化先验如何与用户提供的指令相互作用。我们考察了这种相互作用的三个维度：（1）LLM 对数据和任务定义的熟悉程度与其性能之间的关系；（2）提示中的额外信息能否纠正零样本错误（“决策黏性”）；以及（3）模型对不匹配任务定义的易感性。我们提出了定义特定熟悉度（Definition-Specific Familiarity，DSF）指标，用于衡量模型所引出的概念与目标定义之间的一致性。在九个 LLM 和六个毒性数据集（五个主要数据集及一个额外的稳健性数据集）上进行的实验表明，在控制数据集身份后，DSF 能够预测标注性能（偏相关 $r=+0.41$）。在所测试的所有提示条件下，这种关联始终为正。相比之下，三种常见的文本记忆指标均未表现出正相关。我们发现，提示的纠错能力有限：额外指令或示例只能纠正 34.8% 的零样本错误，其中高置信度错误尤其顽固。不匹配的定义会系统性地改变预测，却不会降低模型报告的置信度，因此置信度并不可靠，无法用于检测定义与策略之间的不匹配。总体而言，这些发现确立了定义一致性作为一种实用的模型选择标准，并表明，仅靠改进提示无法替代对模型与策略匹配度的验证。
arXiv:2605.31043v2 Announce Type: replace-cross Abstract: Deep networks on the symmetric positive-definite (SPD) manifold promise expressive representations by encoding data geometry as an inductive bias, but stacking BiMap layers with the standard ReEig nonlinearity often adds no capacity: on real, preconditioned EEG data, ReEig rarely activates, so the stack behaves as a single layer at any depth. In the worst case, when domains share no discriminative directions, we prove a single filter has a capacity ceiling, so it cannot fully align every domain at once. To overcome that, we propose SCAP (Stiefel Cross-Attention Pool), a layer implementing a family of Stiefel filters by combining a pool of $K$ experts into a sample-specific bilinear map via cross-attention. We show that it matches a per-domain filter bank to first order with fewer experts than domains when domain-optimal filters span few directions near a shared tangent-space basepoint; in the worst case, its alignment empirically stays nearly flat as domains grow, escaping the fixed-filter ceiling. Naively trained, however, this routing can collapse to a fixed filter; we diagnose why and adapt three mechanisms to mitigate it. SCAP significantly improves balanced accuracy over fixed-filter SPDNet on all five cross-domain EEG motor-imagery datasets, and matches or exceeds three domain-adaptive baselines on four out of five.
arXiv:2604.13022v2 Announce Type: replace-cross Abstract: We present the first analytical study of ECD, focusing on the one-dimensional setting for this first installment. We formalize a stochastic ECD dynamics (sECD) with energy-preserving noise, as well as a quantum analog of the ECD Hamiltonian (qECD), providing the foundation for a quantum algorithm through Hamiltonian simulation in a tractable model where the barrier-crossing mechanism can be computed explicitly. For one-dimensional double-well objectives in the under-guessing regime, we compute the expected dynamical hitting times from a local minimum to the global minimum. We prove that both sECD and qECD exhibit exponential improvements in continuous hitting time relative to their respective gradient-based baselines, stochastic gradient descent (SGD) and quantum tunneling walk (QTW). For objectives with tall barriers, qECD admits a further hitting time improvement over sECD. Mechanistically, ECD sidesteps the exponential cost associated with rare-escape events of SGD from local minima by moving from dissipative to energy-conserving dynamics.
arXiv:2602.10273v3 公告类型：替换交叉版本 摘要：大型语言模型的推理能力通常归因于\emph{分布锐化}：将输出概率集中到高似然序列上。近期研究表明，这种锐化效果可以在推理时获得，无需修改模型参数，并且能够激发强大的推理性能。一种自然的形式化方式是\emph{序列级幂分布}，其与模型概率的 $\alpha>1$ 次幂成正比。先前的工作利用 Metropolis--Hastings（MH）采样从该分布中抽取样本，并取得了出色的结果，但推理速度降低了一个数量级。我们提出了\textbf{Power-SMC}，一种\textit{“免训练”}采样方法，目标是相同的幂分布，同时实现接近标准解码的延迟。Power-SMC 并行维护多个候选序列。每个候选序列都会被赋予一个分数，即\emph{重要性权重}，用于衡量其与幂分布的匹配程度。随后，Power-SMC 会定期剪枝低分候选序列，以保留高分候选序列。我们进一步为 Power-SMC 中的设计选择提供了理论依据。在所有不依赖未来词元的下一词元采样策略中，我们证明采样温度 $\tau{=}1/\alpha$ 能够唯一消除每步权重方差。最后，我们刻画了权重不稳定性的剩余来源，并引入渐进式锐化调度，以减少权重坍缩，同时仍以相同的幂分布为目标。在 MATH500、GSM8K、GPQA 和 HumanEval 上进行的大量评估表明，Power-SMC 的准确率达到或超过 MH 采样；与经过 RL 微调的模型不同，它能够保持输出多样性，同时\textbf{将推理速度最高提升至} {$\mathbf{17.6}\times$}。代码可在 https://github.com/ArminAzizi98/Power-SMC 获取。
arXiv:2512.20620v3 公告类型：replace-cross 摘要：使用 Virtual Reality（VR）头戴式显示器时，可能会产生类似晕动症的不适感，这种现象称为赛博晕动症。赛博晕动症阻碍了 VR 技术的进一步普及。通过脑电图（EEG）记录的脑活动，可在不造成干扰的情况下检测赛博晕动症及其引发的不适。为了进行干预和缓解，需要使用机器学习算法，从其他脑部数据中提取与赛博晕动症不适相关的有意义信号。在本研究中，我们利用神经网络和可解释性图谱构建了一个框架，从而确定 EEG 数据中哪些特征最有助于对赛博晕动症相关不适进行分类。基于两项独立的听觉事件相关电位（ERP）赛博晕动症用户研究的脑部数据，我们提取了对不适分类最重要的时空 EEG 特征（来自传感器位置和时间步）。对于第一个数据集（n=29），在使用三个不同神经网络、多个随机种子进行的 120 次框架运行中，模型始终指出，左侧额区、中线中央区和右侧顶区的头皮位置最有助于判断 EEG 数据是否属于正在经历赛博晕动症不适的受试者。在第二个数据集（n=36）中，也标记出了类似位置，表明我们的发现能够跨数据集推广，并适用于不同时间、VR 刺激和受试者样本。在两个数据集中，提取的 ERP 时间窗口中、刺激发生后约 80 至 260 毫秒之间的早期阶段，都被标记为对模型分类十分重要。这些结果有助于明确未来可利用哪些已标记特征，使用 EEG 对赛博晕动症相关不适（以及潜在的其他类型不适）进行分类。项目网页：https://eeg-discomfort-nnilc.github.io/。
arXiv:2511.12749v3 Announce Type: replace-cross Abstract: Forecasting many sparse series requires two choices: how much of each series' own past to retain, and how much to borrow from other series. We show that when one exponential recency operator is applied to item- and group-level statistics alike in a hierarchical empirical-Bayes hurdle model, the two become one decision: forgetting preserves the leverage of a shared prior while raising the noise floor against which fine cross-series structure must be resolved. We formalize this two-sided effect as a fitting-window diagnostic, a credibility bound and a resolution condition for candidate pools, and derive from it an ex ante screen for regimes where refinement cannot pay. On five public intermittent-demand panels comprising over 19,000 series, where item-level evidence is sparsest, the diagnostic is right in both directions. Where truncation lowers credibility the shared prior pays: switching it off costs $5\%$ and $12\%$ at the shortest histories on the two panels it helps most, against about $1\%$ at full length, and the gain sits in the occurrence block. On four panels the diagnostic reports at most $0.6\%$ of room for refinement and realized gains are no larger; on the one panel with room the learned mixture does not collect it, which locates the open problem in the partition objective. The resulting forecaster, EBB, stays within about $3\%$ of a per-series Gaussian-process specialist at fixed origin and within $1\%$ under walk-forward evaluation where both run, at two to three orders of magnitude lower cost.
arXiv:2507.07871v5 公告类型：replace-cross 摘要：水印技术使 GenAI 提供商能够验证内容是否由其模型生成。水印是内容中的隐藏信号，可以使用秘密水印密钥检测其是否存在。一项核心安全威胁是伪造攻击，即攻击者将提供商的水印插入并非由该提供商生成的内容中，从而可能损害其声誉并削弱信任。现有防御方法通过使用多个密钥在同一内容中嵌入许多水印来抵御伪造，但这可能降低模型的实用性。然而，当攻击者能够收集足够多的带水印样本时，伪造仍然是一项威胁。我们提出了一种防御方法：在检测器结果关于密钥对称且相互独立的条件下，对于盲攻击者，其伪造成功率具有与样本数量无关的上界。我们的方法不会进一步降低模型的实用性。对于每次查询，我们随机选择水印密钥，并且仅当检测到恰好一个密钥对应的水印时，才接受该内容为真实内容。不同于依赖计算困难性假设且需要从头设计新水印方案的密码学水印，我们的方法可以应用于任何现有水印方法，以提升其抗伪造能力。我们重点研究文本水印，但由于将底层水印方法视为黑盒，该防御与模态无关。为说明这一点，我们使用 Tree-Ring 对图像水印进行了初步研究。除这一条件性保证外，我们通过实验证实，在使用 \(r=4\) 个密钥时，对于我们评估的自适应盲攻击者，有害文本的伪造成功率在计算开销几乎可以忽略的情况下，从使用单个密钥时最高达 \(87\%\) 降至最低 \(1\%\)；初步的图像研究显示，伪造成功率从 \(100\%\) 降至 \(2\%\)。
arXiv:2506.10944v2 Announce Type: replace-cross Abstract: Understanding and controlling the atomistic-level reactions governing the formation of the solid-electrolyte interphase (SEI) is crucial for the viability of next-generation solid state batteries. However, challenges persist due to difficulties in experimentally characterizing buried interfaces and limits in simulation speed and accuracy. We conduct large-scale explicit reactive simulations with quantum accuracy for a symmetric battery cell, {\symcell}, enabled by active learning and deep equivariant neural network interatomic potentials. To automatically characterize the coupled reactions and interdiffusion at the interface, we formulate and use unsupervised classification techniques based on clustering in the space of local atomic environments. Our analysis reveals the formation of a previously unreported crystalline disordered phase, Li$_2$S$_{0.72}$P$_{0.14}$Cl$_{0.14}$, in the SEI, that evaded previous predictions based purely on thermodynamics, underscoring the importance of explicit modeling of full reaction and transport kinetics. Our simulations agree with and explain experimental observations of the SEI formations and elucidate the Li creep mechanisms, critical to dendrite initiation, characterized by significant Li motion along the interface. Our approach is to crease a digital twin from first principles, without adjustable parameters fitted to experiment. As such, it offers capabilities to gain insights into atomistic dynamics governing complex heterogeneous processes in solid-state synthesis and electrochemistry.
arXiv:2505.13439v2 Announce Type: replace-cross 摘要：自回归（AR）模型近期在图像生成方面展现出强大性能，其中一个关键组件是视觉分词器（VT），它将连续的像素输入映射为离散的 token 序列。VT 的质量在很大程度上决定了 AR 模型性能的上限。然而，当前的离散 VT 明显落后于连续变分自编码器（VAEs），导致图像重建质量下降，以及细节和文本保留能力较差。现有基准主要关注端到端生成质量，未能将 VT 的性能单独分离出来进行评估。为弥补这一空缺，我们推出了 VTBench，这是一个全面的基准，围绕三项核心任务系统评估 VT：图像重建、细节保留和文本保留，并涵盖多种评估场景。我们利用一组指标系统评估了当前最先进的 VT，以衡量重建图像的质量。研究结果表明，与离散 VT 相比，连续 VAE 能够生成更优越的视觉表征，尤其是在保留空间结构和语义细节方面。相比之下，离散 VT 生成的退化表征通常会导致重建失真、细粒度纹理丢失，以及无法保留文本和物体的完整性。此外，我们在 GPT-4o 图像生成上开展了实验，并讨论了其潜在的 AR 特性，为理解视觉分词的作用提供了新的见解。我们公开发布了基准和代码库，以支持进一步研究，并呼吁社区开发强大且通用的开源 VT。
arXiv:2504.03432v4 Announce Type: replace-cross Abstract: Solving (Stampacchia) variational inequalities (SVIs) is a foundational problem at the heart of optimization. However, this expressivity comes at the cost of computational hardness. As a result, most research has focused on carving out specific subclasses that elude those intractability barriers. A classical property that goes back to the 1960s is the Minty condition, which postulates that the Minty VI (MVI) problem admits a solution. In this paper, we establish the first polynomial-time algorithm -- with complexity growing polynomially in the dimension $d$ and $\log(1/\epsilon)$ -- for solving $\epsilon$-SVIs for Lipschitz continuous mappings under the Minty condition. Prior approaches either incurred an exponentially worse dependence on $1/\epsilon$ (and other natural parameters of the problem) or made more restrictive assumptions, such as monotonicity. To do so, we introduce a new variant of the ellipsoid algorithm whereby separating hyperplanes are obtained after taking a descent step from the center of the ellipsoid. It succeeds even though the set of SVIs can be nonconvex and not fully dimensional. Moreover, when our algorithm is applied to an instance with no MVI solution and fails to identify an SVI solution, it produces a succinct certificate of MVI infeasibility. We also show that deciding whether the Minty condition holds is $\mathsf{coNP}$-complete, thereby establishing that the disjunction of those two problems is polynomial-time solvable even though each problem is individually intractable. We provide several extensions and new applications of our main results. Most notably, we obtain the first polynomial-time algorithms for computing Nash equilibria in multi-player harmonic games. Finally, in two-player general-sum concave games, we give the first polynomial-time algorithm that outputs either a Nash equilibrium or a strict coarse correlated equilibrium.
arXiv:2311.18837v2 公告类型：replace-cross 摘要：Diffusion Models 在图像和视频生成方面取得了显著成功。这推动了人们对视频编辑任务日益增长的兴趣，即根据给定的文本描述编辑视频。然而，现有的大多数方法仅关注短视频片段的编辑，并且依赖耗时的调优或推理过程。我们首次提出 Video Instruction Diffusion（VIDiff），这是一种面向广泛视频任务的统一基础模型。这些任务既包括理解类任务（如语言引导的视频目标分割），也包括生成类任务（视频编辑和增强）。我们的模型能够根据用户指令，在数秒内编辑并生成所需结果。此外，我们设计了一种迭代式自回归方法，以确保对长视频进行编辑和增强时的一致性。我们针对多样化的输入视频和文字指令，在定性和定量层面都提供了令人信服的生成结果。更多示例请见我们的网站 https://ChenHsing.github.io/VIDiff。
arXiv:2307.07113v2 Announce Type: replace-cross Abstract: In this paper, we consider the decentralized, stochastic nonconvex strongly-concave (NCSC) minimax problem with nonsmooth regularization terms on both primal and dual variables, wherein a network of $m$ computing agents collaborate via peer-to-peer communications. We consider when the coupling function is in expectation or finite-sum form and the double regularizers are convex functions, applied separately to the primal and dual variables. Our algorithmic framework introduces a Lagrangian multiplier to eliminate the consensus constraint on the dual variable. Coupling this with variance-reduction (VR) techniques, our proposed method, entitled VRLM, by a single neighbor communication per iteration, is able to achieve an $\mathcal{O}(\kappa^3\varepsilon^{-3})$ sample complexity under the general stochastic setting, with either a big-batch or small-batch VR option, where $\kappa$ is the condition number of the problem and $\varepsilon$ is the desired solution accuracy. With a big-batch VR, we can additionally achieve $\mathcal{O}(\kappa^2\varepsilon^{-2})$ communication complexity. Under the special finite-sum setting, our method with a big-batch VR can achieve an $\mathcal{O}(n + \sqrt{n} \kappa^2\varepsilon^{-2})$ sample complexity and $\mathcal{O}(\kappa^2\varepsilon^{-2})$ communication complexity, where $n$ is the number of components in the finite sum. All complexity results match the best-known results achieved by a few existing methods for solving special cases of the problem we consider. To the best of our knowledge, this is the first work which provides convergence guarantees for NCSC minimax problems with general convex nonsmooth regularizers applied to both the primal and dual variables in the decentralized stochastic setting. Numerical experiments are conducted on two machine learning problems. Our code is downloadable from https://github.com/RPI-OPT/VRLM.
arXiv:2609.35698v3 Announce Type: replace Abstract: We study adversarial imitation learning (AIL), in which an agent learns to imitate expert demonstrations by optimizing a policy against an adversarial reward that distinguishes expert and learner behavior. Historically, reward regularization and entropy-based policy regularization are key components of empirically successful methods such as GAIL and LS-IQ, yet their finite-sample benefits remain underexplored. We establish fast rates for jointly regularized AIL in finite-horizon Markov decision processes with general function approximation. Our model-free algorithm, Dually Regularized AIL, combines KL policy regularization with a quadratic reward penalty weighted by expert and learner occupancies. With K online episodes and N expert trajectories, we prove a $\widetilde{O}\left(\frac{1}{K}+\frac{1}{N}\right)$ bound on the regularized imitation gap for fixed regularization parameters. Our analysis combines an online mirror descent construction for general convex reward classes to control estimation error from finite expert data and stochastic learner feedback, with a sharp analysis of optimistic KL-regularized policy learning. To the best of our knowledge, Dually Regularized AIL is the first algorithm to simultaneously achieve $\widetilde{O}\left(\frac{1}{\epsilon}\right)$ sample complexity in both expert demonstrations and online interactions for this regularized AIL objective, even with stochastic experts. These results provide a rigorous characterization of the complementary statistical benefits of reward and policy regularization in AIL.
arXiv:2609.27248v2 公告类型：替换 摘要：下一个词元预测使自回归语言模型能够生成高度流畅的文本，但它只能通过序列分解间接表示全局结构。相比之下，高保真自编码器已成为图像生成中的标准基础模块，使生成模型能够在连续潜空间上运行；而文本尚缺乏同样忠实的连续表示。我们提出 LLMAE，通过将中间层的激活暴露为固定长度的潜在瓶颈，将预训练的仅解码器语言模型重新用作连续文本自编码器。LLMAE 通过结构化注意力掩码和 LoRA 适配实现了这一接口，同时利用原始 LLM 的生成先验。在两个骨干模型（270M Gemma 3 和 0.5B Qwen2.5）上，LLMAE 能够高准确度地重建最长 1024 个词元的序列，并逐字复现多达 97% 的文档。我们通过训练一个轻量级扩散模型展示了其下游应用价值：该模型能够直接在冻结的 LLMAE 潜空间中生成详细的图像描述。通过将文本映射到固定长度的连续潜空间，我们的方法为下游适配提供了有效基础，同时受益于原始 LLM 的流畅性。代码和模型可在 https://github.com/arkanath/LLMAE 获取。
arXiv:2609.20744v3 公告类型：替换 摘要：视频扩散模型在去噪过程中会反复处理长时空令牌序列，使注意力机制成为主要的计算瓶颈。线性注意力提供了颇具吸引力的替代方案，并已广泛应用于近期的大型语言模型中，但将其直接应用于视频模型，往往无法保留高质量生成所需的细粒度交互。我们提出 Video DeltaNet（VDN），将局部 Softmax 注意力与双向线性记忆相结合，以建模长程视频上下文。其线性分支引入了 Video Delta Attention（VDA），通过联合整合每一帧的空间令牌，每帧更新一次记忆。独立的输出投影和可学习门控机制用于校准两个分支，同时，分阶段的教师对齐方案逐步将这一新路径引入预训练模型。我们在 MiniMax H3 上实例化了 VDN，将这种混合机制应用于视频到视频的交互，同时保留 Softmax 以处理涉及文本或音频的交互。借助八步蒸馏和优化后的 SGLang 服务栈，VDN-H3 在八块 NVIDIA B200 GPU 上，仅用 6.70 秒即可完成 14.3 秒、768p 视频的 DiT 去噪；相比相同 GPU 数量下采用 50 步密集 H3 基线，实现了 14.5 倍加速。GitHub 代码见：https://github.com/OpenVDN/vdn-minimax-h3。模型权重见：https://huggingface.co/OpenVDN/vdn-minimax-h3
arXiv:2609.12223v2 Announce Type: replace Abstract: Collision cross section (CCS), derived from ion mobility mass spectrometry, is a common descriptor for molecular annotation. Prediction is challenging for machine learning models because it reflects the size, shape, and ionization state of a gas-phase molecular ion. Most predictors either ignore explicit 3D structure or treat adduct identity as a late categorical feature, which limits their ability to capture adduct-dependent geometric effects. We present GRACE (Geometric Residual Adduct Conditioning via Early-fusion), a 3D CCS predictor that adapts a pretrained molecular geometry encoder using geometric residual adduct conditioning via early fusion. GRACE combines two inductive biases: a residual objective relative to an adduct-aware physical descriptor baseline and adduct conditioning within the encoder via a learned adduct token and low-rank attention adapters. We evaluate the model on a curated set of over 9,000 experimental molecule-adduct CCS records with random, scaffold, and adduct-sensitive splits designed to separate interpolation, scaffold generalization, and adduct-driven generalization. GRACE achieves the best mean percentage difference among the evaluated learned models on all three splits: 1.67% on the random split, 2.11% on the scaffold split, and 2.36% on the adduct-sensitive split. Diagnostic analyses suggest that residual learning stabilizes training by removing the dominant mass-CCS trend, while early fusion improves adduct-sensitive prediction relative to late fusion. Across four independent external test sets, GRACE shows consistently lower error than the other evaluated models. On a held-out set, GRACE also attains the lowest mean percent difference when compared with four previously reported physics-based workflows. These results support residual learning and encoder-level adduct conditioning as practical inductive biases for fast, accurate CCS prediction.
arXiv:2607.19060v2 Announce Type: replace Abstract: Fast prediction of the response of adhesive soft viscoelastic contacts represents a current challenge in soft robotics and for gripping and manipulation tasks. Determining the complete time-resolved force trajectory requires full numerical simulations, whose computational cost is strongly parameter-dependent, making them impractical for real-time application or design-optimization loops. In this work, we overcome this limitation by training a scalar-conditioned, stateful, sequence-to-sequence deep learning model to predict the full force evolution from a prescribed displacement history for both short- and long-range adhesion regimes. The data set spans four orders of magnitude in loading and unloading rates and includes varied dwell times, with the Tabor parameter ranging from $0.2$ to $3.2$. To enable learning across these heterogeneous time scales, we introduce a fixed-measurement-step (FMS) representation that converts variable-length trajectories into fixed-length sequences while preserving their physical-time information. Different architectures were trained, including long short-term memory (LSTM) networks, temporal convolutional neural (TCN) networks, and time-distributed dense layers with three different Tabor-conditioning mechanisms. The models were compared using global waveform and error metrics. We found that the best-performing model has an LSTM architecture with concatenated conditioning, which achieves a held-out mean-squared error of $5.0\times10^{-4}$, a median pull-off-force error of $\approx2.2\%$, and a median hysteresis error of $\approx1.1\%$. For the held-out protocols, the model predicts a complete force trajectory with a median inference time of $0.16$ s. The model is tested across unseen parameter combinations and against analytical limiting cases, providing a rapid surrogate for repeated numerical evaluations with potential use in control-oriented applications.
arXiv:2607.13731v2 Announce Type: replace Abstract: Goal-conditioned reinforcement learning hinges on how the goal is encoded. Contrastive, metric, temporal-distance and information-theoretic encoders disagree on the objective. They agree on one thing. None of them sees the current state, so the embedding cannot mark which part of the goal still needs action, and the policy must recover that cue by inverting both encoders. We propose DAGR, which refines the static embedding of any late-fusion encoder into a state-conditioned one through multi-scale gated cross-attention. A gated residual holds the refinement near the base, and a difference-aware attention rule biases the scores by a per-token state-goal mismatch. A single condition decides what such a refinement can guarantee, namely whether the block returns its input at closed gates. We prove that the usual post-norm placement violates it, measure the consequence on frozen checkpoints, and recover part of the resulting loss by restoring the condition. On OGBench DAGR improves navigation and matches or trails the base elsewhere. Our ablations trace the gain to the gated residual rather than to the difference bias that names the method. Code is available at https://github.com/leixingxing1/DAGR
arXiv:2607.09277v2 Announce Type: replace Abstract: Three-dimensional (3D) molecule generation has been dominated by diffusion models, which achieve strong generation quality but typically require molecular size to be specified (or predicted) separately before generation. This can be limiting for fragment-based molecule generation, central to drug discovery, where the size of the generated structure is itself part of the design problem. Autoregressive models determine size during generation and naturally support partial-structure conditioning, but balancing unconditional and fragment-conditioned generation remains challenging. We introduce KRONOS, a latent autoregressive diffusion framework that generates molecules in the latent space of a Unified AutoEncoder (UAE), jointly modeling molecular graph topology and geometry, while retaining the flexibility of autoregressive generation. We further introduce a mixed training strategy inspired by the Fill-in-the-Middle (FIM) paradigm, enabling a single left-to-right autoregressive model to support both unconditional and fragment-conditioned generation. Experiments on QM9 and GEOM-Drugs demonstrate strong unconditional generation performance and competitive fragment-conditioned generation.
arXiv:2607.01763v2 公告类型：替换 摘要：持续后训练使基础模型能够在保留既有能力的同时获取新知识。近期研究表明，策略内学习可以缓解遗忘，其中自蒸馏尤其具有吸引力。我们通过策略内自蒸馏优化（SDPO）重新审视这一乐观观点。实验表明，当教师信号稳定且对齐良好时，SDPO 能够加速域内专门化，但难以实现分布外泛化。在持续后训练中，SDPO 表现出更严重的遗忘，甚至可能发生崩溃；相比之下，更为成熟的策略内强化学习方法 GRPO 的适应更加保守，也能更好地保留先前能力。进一步分析表明，这些失败与参数空间和响应空间中更大的漂移有关，也与自我强化的师生循环对高频伪影的放大有关。因此，仅依靠策略内数据不足以实现持续学习。当教师目标稳定且 token 级监督可靠时，自蒸馏是有效的，但不应被视为持续后训练的默认稳定器。我们的代码可在 https://github.com/Moenupa/SDPO-CL 获取。
arXiv:2605.23215v2 公告类型：替换 摘要：基于 LLM 的 GPU 内核生成代理正在快速发展，但它们所针对的基准测试是在隔离环境中评估内核，使用合成输入和较弱的基线，从而奖励那些在沙箱中实现加速、却在真实推理系统中失效或消失的方案。我们推出 FastKernels，这是一个包含 384 个任务的基准测试，任务源自 8 个类别中的 47 种代表性架构；这些内核足以重新实现 HuggingFace Transformers 中 94.6%（472/499）的架构，且输出与原生实现一致。每个任务都模拟相应生产模块的接口，并与生产框架所提供的内核进行比较；同时，任务构成一个从基础算子到完整模型的组合层级结构，高层模块会导入低层模块。候选方案既在内核层面进行评分，也在其所属模型内部、沿生产执行路径进行端到端评分；MacroEval 将经过校准的正确性、覆盖率和加速比汇总为排行榜。我们使用五个具有代表性的代理为其提供初始测试（共 6,900 个代理小时），发现内核层面的 1.6–6.6 倍加速在端到端评估中缩减至最多 1.25 倍；获胜的内核集合中只有 20% 能够直接正确运行；此外，内核层面的得分会错误地对代理进行排序：Claude Code 在单独评估的每个层级都能追平或超过 KDA，但端到端得分却是 KDA 的 3 倍。代码见 https://github.com/Snowflake-AI-Research/fastkernels。
arXiv:2605.05738v2 公告类型：替换 摘要：交通传感器网络会随着传感器的增加以及交通分布的变化而不断演化，而大多数预测模型都假设节点集合固定，并反复使用所有可用数据进行重新训练。我们提出了 CoMemNet，一种用于在不断演化的交通传感器网络上高效预测的持续记忆网络。CoMemNet 使用 Online 分支适应当前时段，并采用指数移动平均的 Target 分支作为稳定的特征参照。基于 Wasserstein 距离的漂移采样器比较各节点的 Online-Target 特征分布，并选择数量有限的、对漂移敏感的节点进行更新。轻量级的节点自适应时序记忆回放缓冲区（Node-Adaptive Temporal Memory Replay Buffer，TMRB-N）保留紧凑的时序状态，无需反复遍历所有历史训练数据。预测骨干网络不使用邻接矩阵；传感器邻接关系仅用于构建数据，并可选地将选定的更新节点扩展至规模有限的邻域。我们在三个多时段 PeMS 数据集上开展了实验，包括三随机种子评估、强静态重训练基线和持续学习基线、受控采样策略、持续学习指标、鲁棒性测试以及资源核算。结果表明，在有界的共享节点选择条件下，CoMemNet 能够保持稳定的预测精度并实现高效适应，在历史知识保留与当前时段预测性能之间取得更佳平衡。同时，随着演化网络不断扩展，与当前时段重训练基线相比，CoMemNet 展现出更明显的精度优势和累计训练时间优势。代码可在以下地址获取：https://meiwu5.github.io/CoMemNet。
arXiv:2603.20103v4 Announce Type: replace Abstract: Forward-backward (FB) representations provide a powerful framework for learning the successor representation (SR) in continuous spaces by enforcing a low-rank factorization. However, a fundamental spectral mismatch often exists between the high-rank transition dynamics of continuous environments and the low-rank bottleneck of the FB architecture, making accurate low-rank representation learning difficult. In this work, we analyze temporal abstraction as a mechanism to mitigate this mismatch. By characterizing the spectral properties of the transition operator, we show that temporal abstraction acts analogously to a low-pass filter that suppresses high-frequency spectral components. This suppression reduces the effective rank of the induced SR while preserving a formal bound on the resulting value function error. Empirically, we show that this alignment is a key factor for stable FB learning, particularly at high discount factors where bootstrapping becomes error-prone. Our results identify temporal abstraction as a principled mechanism for shaping the spectral structure of the underlying MDP and enabling effective long-horizon representations in continuous control.
arXiv:2603.19307v2 Announce Type: replace Abstract: Modeling the complex interactions among functional subnetworks is crucial for the diagnosis of mental disorders and the identification of functional pathways. However, learning the interactions of the underlying subnetworks remains a significant challenge for existing Transformer-based methods due to the limited number of training samples. To address these challenges, we propose KD-Brain, a Prior-Informed Graph Learning framework for explicitly encoding prior knowledge to guide the learning process. Specifically, we design a Semantic-Conditioned Interaction mechanism that injects semantic priors into the attention query, explicitly navigating the subnetwork interactions based on their functional identities. Furthermore, we introduce a Pathology-Consistent Constraint, which regularizes the model optimization by aligning the learned interaction distributions with clinical priors. Additionally, KD-Brain leads to state-of-the-art performance on a wide range of disorder diagnosis tasks and identifies interpretable biomarkers consistent with psychiatric pathophysiology. Our code is available at https://anonymous.4open.science/r/KDBrain.
arXiv:2602.01124v4 公告类型：替换 摘要：动态图表示学习需要同时捕捉结构关系和时间演化，但现有方法面临一个核心权衡：基于注意力的方法具有较强的表达能力，但复杂度为 O(T^2)；循环架构则存在梯度病态问题，并需要存储稠密状态。脉冲神经网络能够提供事件驱动的高效性，但受到顺序传播、二值信息损失以及缺乏全局上下文的局部聚合等因素的限制。我们提出 ChronoSpike，一种自适应脉冲图神经网络，将可学习的 LIF 神经元、带有逐通道膜电位动力学的机制、基于连续特征的多头空间注意力聚合，以及轻量级 Transformer 时间编码器相结合。该设计以 O(T.d) 的激活/状态内存开销和额外的 O(T^2) 每节点注意力项，实现了细粒度局部建模与长程依赖捕捉；对于本文评估的时间跨度，该额外开销仍然较小。在三个大型基准数据集上，ChronoSpike 的表现优于十二种最先进的基线方法，Macro-F1 平均提升 2.0%，Micro-F1 平均提升 2.4%；同时，在恒定的 105K 参数预算下（与图规模无关），其训练速度比循环方法快 3~10 倍。我们为膜电位有界性、收缩因子 \r{ho}<1 条件下的梯度流稳定性以及 BIBO 稳定性提供了理论保证；可解释性分析揭示了异质的时间感受野，以及具有 83-88% 稀疏度的习得性首因效应。我们的代码已公开发布于：https://github.com/Abrar2652/ChronoSpike
arXiv:2510.13210v2 Announce Type: replace Abstract: We compare Ising ({-1, +1}) and QUBO ({0, 1}) encodings for Boltzmann machine learning under controlled protocols that fix the sampler, optimizer, and learning-rate design within each comparison. Exploiting the identity that the Fisher information matrix (FIM) equals the covariance of sufficient statistics, we visualize empirical moments from model samples and reveal systematic, representation-dependent differences. QUBO induces larger cross terms between first- and second-order statistics, creating more small-eigenvalue directions in the FIM and lowering spectral entropy. This ill-conditioning explains slower convergence under stochastic gradient descent (SGD). In contrast, full-FIM natural gradient descent (NGD), which rescales updates by the FIM metric, achieves similar convergence across encodings, whereas diagonal-FIM approximation can reintroduce representation-dependent differences. Practically, for SGD-based training, the Ising encoding provides more isotropic curvature and faster convergence; for QUBO, centering/scaling or NGD-style preconditioning mitigates curvature pathologies. These results clarify how representation shapes information geometry and finite-time learning dynamics in Boltzmann machines and yield actionable guidelines for variable encoding and preprocessing.
arXiv:2506.05937v3 Announce Type: replace Abstract: Reliability of deep learning models is critical for deployment in high-stakes applications, where out-of-distribution or adversarial inputs may lead to detrimental outcomes. Evidential Deep Learning, an efficient paradigm for uncertainty quantification, models predictions as Dirichlet distributions of a single forward pass. However, EDL is particularly vulnerable to adversarially perturbed inputs, making overconfident errors. Conflict-aware Evidential Deep Learning~\mbox{(C-EDL)} is a lightweight post-hoc uncertainty quantification approach that mitigates these issues, enhancing adversarial and OOD robustness without retraining. C-EDL generates diverse, task-preserving transformations per input and quantifies representational disagreement to calibrate uncertainty estimates when needed. C-EDL's conflict-aware prediction adjustment improves detection of OOD and adversarial inputs, maintaining high in-distribution accuracy and low computational overhead. Our experimental evaluation shows that C-EDL significantly outperforms state-of-the-art EDL variants and competitive baselines, achieving substantial reductions in coverage for OOD data (up to $\approx55\%$) and adversarial data (up to $\approx90\%$), across a range of datasets, attack types, and uncertainty metrics.
arXiv:2610.03647v1 Announce Type: cross Abstract: Many machine learning methods aim to approximate the lower-dimensional manifold on which the data lives. A desirable feature of such methods is that they should capture the epistemic uncertainty of this learned manifold. One model that achieves this is the Gaussian Process Latent Variable Model, in which a Gaussian Process (GP) mapping from the latent space provides an estimate of the uncertainty of the manifold. However, the effectiveness of this uncertainty estimation is limited by the mean-field variational approximation between the GP inducing points and the latent variables. In this work, we apply Amortized Structured Stochastic Variational Inference to allow the variational posterior for the latent space to be conditionally dependent on the value of the inducing points. We demonstrate that this more flexible variational posterior improves several metrics relating to the reconstruction of points on the data manifold.
arXiv:2610.03625v1 公告类型：cross 摘要：关系数据库是部署最广泛的结构化知识形式之一，而要以自然语言访问这些数据库，就需要将语言映射到模式实体和关系上，同时处理人们表达请求时固有的歧义。现有的合成 NL-to-SQL 数据生成方法大多忽略了这种歧义，生成的查询过于简化，无法让模型为现实世界结构化知识访问的复杂性做好准备。我们提出 FALCON，这是一个能够生成逼真的、具备歧义感知能力的 NL-to-SQL 数据的框架，其复杂程度可匹配具有挑战性的现实世界基准，同时借助紧凑型开源模型实现低成本生成。我们的方法结合保留字 SQL 播种和基于人物画像的提示，以生成结构复杂的查询；同时，基于对齐的过滤机制通过区分真正错误的示例与复杂但有效的查询来保留数据难度。人工评估证实，该方法在不同模型规模下都能保持稳定的高质量；生成的数据在 SQL 复杂度和自然语言丰富性方面均超过现有基准。按难度分层的分析表明，随着查询复杂度的增加，使用 FALCON 数据训练的模型相较于使用基线数据训练的模型，其性能优势也不断扩大，验证了该流程生成高难度训练数据的有效性。与少量现有基准数据结合时，混合训练能够恢复模型在简单查询上的性能，同时保留其在复杂查询上的优势。这种与模型和数据库无关的设计，使组织能够在本地生成高复杂度的 NL-to-SQL 训练数据，而无需依赖外部 API。
arXiv:2610.03621v1 公告类型：cross 摘要：近年来，关于马尔可夫博弈中相关均衡概念的研究迅速增加。然而，现有成果主要关注弱于正规形式相关均衡（NFCE）的概念，因此，如何计算此类均衡这一更具挑战性的问题仍未解决；该问题可追溯至 Papadimitriou 和 Roughgarden 的开创性工作（JACM'08）。本文首次为具有固定玩家数量 $n$ 的有限时域马尔可夫博弈建立了计算 NFCE 的高效算法。具体而言，对于具有 $S$ 个状态、时域长度 $H$ 且每名玩家至多有 $A$ 个动作的博弈，该算法能在 $S(AH/\epsilon)^{O(n)}$ 时间内计算出一个 $\epsilon$-NFCE。这是首个在正规形式之外的一个重要问题类别中，同时关于 $1/\epsilon$ 和博弈描述规模均为多项式时间的 NFCE 算法。此外，在通常假设各状态之间的推荐相互独立的情况下，我们证明了该问题是 PPAD 完全的——也就是说，在计算上等价于求解 Nash 均衡——这一结论要么适用于多玩家博弈，要么适用于精度指数级微小的情形。我们方法的核心思想是对一系列辅助阶段博弈执行逆向归纳，但其中的关键变化在于：在每一步中，我们计算一个常期望相关均衡。这是相关均衡的一种自然细化，其中，遵循推荐所获得的条件期望收益与推荐内容无关。事实上，我们的归约可以双向进行，从而建立了马尔可夫博弈中常期望相关均衡与 NFCE 之间的等价性。对于固定数量的玩家，通过结合线性规划与适当的离散化，我们发现可以近似计算常期望相关均衡。相比之下，在以下情形中，该问题是 PPAD-hard 的：i）具有常数精度的 polymatrix（多玩家）博弈；ii）具有指数级微小精度的双玩家博弈。后一个结果源于一个出人意料的联系：秩为 2 的双玩家博弈。
arXiv:2610.03444v1 Announce Type: cross Abstract: Molecular optical absorption spectroscopy provides a direct probe of electronic structure and is widely used for molecular identification, interpretation of photophysical behaviour, and planning of spectroscopy experiments. Calculating the absorption spectra using first-principle excited-state methods, however, is computationally demanding, at least compared to ground-state calculations, which limits their routine application across large molecular sets. Machine-learning (ML) surrogates can reduce this cost and allow rapid spectral prediction. However, their performance depends strongly on how molecular information is represented. Here, we compare using the ground-state electron density versus the molecular geometry as inputs to a ML model for predicting absorption spectra, for a training set of 6874 molecules selected from the QM7 dataset. For each of these molecules, the density was calculated using density functional theory (DFT) and the absorption spectrum was calculated using linear-response (LR) time-dependent DFT (TDDFT). Utilizing the ground-state density as the input to the ML model is motivated by the Hohenberg-Kohn and Runge-Gross theorems, and the fact that the ground-state density encodes information about bonding, charge localisation, and electronic delocalisation. Hence, it may be a more judicious starting point for the ML model compared to the geometry, as it effectively decouples the chemistry of the ground-state. The question we test is whether the benefits of using the density outweigh the (notprohibitive) penalty of requiring an additional single-point DFT calculation for the density. We find that the density-based convolutional neural network achieves a validation correlation of 0.9926, compared with 0.9795 for the best geometry-based graph model, reducing the residual decorrelation, by approximately 64%.
arXiv:2610.03369v1 Announce Type: cross Abstract: Deep reinforcement-learning policies for order execution can vary substantially across training seeds, so apparent architectural gains may reflect favourable training realisations rather than reproducible properties of the architecture. We evaluate vanilla Double Deep Q-Learning (DDQL), K-means-partitioned mixtures of DDQL experts at $K \in \{2, 4, 8\}$, and dense networks parameter-matched to the $K{=}4$ and $K{=}8$ expert budgets on 5-minute mean-aggregated BTC/USDT limit order book data from Binance. No learned configuration significantly improves mean implementation shortfall over DDQL. Under the reported specification, all have higher mean shortfall than TWAP (0.39 bps) and immediate liquidation (0.21 bps) in an environment whose frictionless replay and terminal-urgency penalty make early liquidation nearly costless; 11/100 vanilla-DDQL runs, versus none in either MoE $K{\geq}4$ arm, converge to a policy that waits until forced liquidation. We then decompose this specification on a device-matched baseline. Annealed exploration alone eliminates observed collapses (12/100 to 0/100; exact McNemar $p{=}4.9{\times}10^{-4}$), matching the elimination under expert partitioning. Combining annealed exploration with the aligned reward restores collapse in 19/30 runs; with all three specification changes, it rises to 48/100. In this environment, expert partitioning is unnecessary to suppress collapse and appears to mask a training-specification failure rather than confer an intrinsic performance benefit. No MoE $K{=}8$ run collapses under any of the six specifications tested. Across-seed dispersion is lowest at $K{=}8$ but non-monotone and not robust to family-wise adjustment, while within-policy tail risk worsens monotonically with $K$. The apparent attribution of the failure mode reverses between 30 and 100 seeds, illustrating the importance of repeated-seed evaluation.
arXiv:2610.03314v1 公告类型：cross 摘要：基于流和扩散的生成模型近年来逐渐成为动力系统中灵活且高效的预测模型。将其与推理时引导相结合，为高维非高斯数据同化（DA）提供了一条有前景的路径。数据同化旨在结合预测与观测，以估计系统的潜在状态。然而，现有滤波器基于固定历史进行条件化，并且仅同化最近一次观测，因此当新观测到来时无法修正过去的状态。这样得到的估计会持续受限于一段可能被后续观测推翻的历史，误差也会在同化过程中不断累积。为此，我们提出 **DAWIS**，一种统一的数据同化方法，在单一框架内涵盖滤波、固定滞后平滑和分块平滑。DAWIS 将状态级先验的单一流时间替换为覆盖连续状态窗口的多任务随机插值器，并为每个状态分配独立的流时间。同化周期首先将该窗口逆变换为由各状态转折点组成的向量，然后在观测引导下重新生成窗口；这些转折点控制着每个状态被保持不变、修正或从头生成的程度。相同的构造还可以将预测过程纳入同化周期，从而无需单独的预测模型。在具有挑战性的非线性系统上的实验表明，在稀疏、噪声和非线性观测条件下，DAWIS 的表现优于滤波和平滑基线方法。DAWIS 的代码可在 https://github.com/Erik-Wikingsson/DAWIS 获取。
arXiv:2610.03220v1 Announce Type: cross Abstract: Hybrid quantum classical neural networks integrate parameterized quantum circuits (PQCs) with established deep learning architectures, but their performance depends strongly on the choice of quantum circuit architecture, a choice that remains largely manual. Most existing approaches rely on hand-designed or fixed circuit ans\"atze, requiring circuit structure, gate composition, and qubit connectivity to be specified in advance with no guarantee that they suit the task. This limitation is especially acute in image classification, where quantum circuits must transform features extracted by classical networks while remaining compact enough for practical training, requirements that generic, task-agnostic ans\"atze are unlikely to satisfy simultaneously. We extend EXAQC, an evolutionary framework for automated quantum circuit discovery, to image classification. EXAQC evolves PQCs as intermediate processing modules while retaining classical feature-extraction and prediction layers. On MNIST, Fashion-MNIST, and CIFAR-10, EXAQC achieves 98.42%, 90.62%, and 85.47% accuracy, respectively, while using comparable gate counts to other quantum architecture-search methods. Against classical networks, evolved hybrid models maintain comparable accuracy with substantially fewer trainable parameters, reaching 85.68% on CIFAR-10 with over 25$\times$ fewer parameters than a 10-layer CNN. Encoding choice also matters: rotation-based encodings (RX, RY, U3) outperform amplitude encoding by 22-25 points on CIFAR-10. These results demonstrate that automated circuit discovery yields compact quantum modules that can replace larger classical components in vision architectures while retaining competitive accuracy.
arXiv:2610.03205v1 Announce Type: cross Abstract: We establish lower bounds for Hamiltonian property testing with access to the time-evolution operator but not its inverse. Each experiment may query the time-evolution operator multiple times, and distances between Hamiltonians are measured in the normalized Frobenius norm. In this model, we show that testing whether a Hamiltonian is $k$-local or $\varepsilon$-far from every $k$-local Hamiltonian requires $\Omega(1/\varepsilon^2)$ total evolution time, matching the upper bound of Kallaugher and Liang (TQC'25). We also prove that testing whether an unknown Hamiltonian equals a target Hamiltonian or is $\varepsilon$-far from it requires $\Omega(1/\varepsilon^2)$ total evolution time, matching the upper bound of Sinha and Tong (2025). These are the first lower bounds for natural problems in Hamiltonian learning and testing that rule out Heisenberg-limited scaling of $1/\varepsilon$. As a third result, we show that amplitude estimation to precision $\varepsilon$ requires $\Omega(1/\varepsilon^2)$ total time evolution, recovering the result of Tang and Wright (QIP'26) in the continuous-time query model. All three results follow from the hardness of distinguishing the zero Hamiltonian from a suitably chosen ensemble of random Hamiltonians. We establish this hardness by adapting the continuous-time adversary method to forward Hamiltonian evolution.
arXiv:2610.03125v1 公告类型：cross 摘要：语音表达会随着所要求的副语言属性和语言内容的不同而变化。我们提出 ParaGeo，在冻结的语音语言模型中对副语言变异进行基于匹配内容的分解。合成的音频标记通过固定的听取提示进行重放；池化后的键/值（K/V）表示经过中心化处理，并投影到共享的低维空间中。我们的 GLM-4-Voice 探针涵盖了来自 12 个基准系列的 80 个请求控制变量，涉及 8 个句子。使用全局拟合的校准基，基于该基的内容留出质心准确率为 9.49%，而置换基线为 1.25%；相同标签跨内容的余弦相似度为 0.285，而基线为 0.017，且两项条件置换检验均得到 p = 0.001。另一项包含 10 种场景和 6 种风格的探针揭示了跨场景可复现的对比方向。静态、加性和时间性干预产生了取决于属性、层级和调度方式的响应特征。这些结果为测量副语言结构提供了一种共享坐标表示，也为潜在语音控制提供了一个经验起点。代码可在 https://github.com/yuhanlydia/ParaGeo 获取。
arXiv:2610.03066v1 Announce Type: cross Abstract: GC--EI--MS is an important technique for analyzing volatile and semivolatile compounds in complex samples. However, conventional methods rely heavily on reference spectral library matching, limiting their ability to identify compounds absent from these libraries and to infer complete molecular structures directly from fragmentation information. Here, we present DiffGCMS, a spectrum-conditioned discrete graph diffusion model for de novo structure elucidation from GC--EI--MS, and further develop a framework that integrates DiffGCMS with second-stage reasoning by a large language model (LLM). In the first stage, DiffGCMS generates candidate molecular structures from input spectra; in the second stage, the LLM uses mass spectral information to validate, repair, and rerank the candidates and provides interpretable analysis of fragment-ion peaks. This framework can generate plausible molecular structures for compounds absent from reference spectral libraries and provide traceable evidence supporting its decisions. On a test set comprising 13,696 spectra from NIST 20, the generative model achieved Acc@1 and Acc@10 of 6.01\% and 15.76\%, respectively. On the test subset containing molecules with no more than 10 heavy atoms, LLM-assisted molecular graph repair and reranking increased Acc@1 from 21.28\% to 21.95\%, Acc@10 from 46.91\% to 47.99\%, and candidate validity from 91.04\% to 100\%. These results demonstrate that spectrum-aware postprocessing can correct errors produced by the generative model while providing auditable and traceable explanations for the final ranking.
arXiv:2610.02828v1 Announce Type: cross Abstract: Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL regularization, verifier allocation, and update budget. Three coupled issues remain unresolved. A future-risk model trained from behavior trajectories need not estimate the risk induced by the controller that will be deployed; a score calibrated on logged state-action pairs can become miscalibrated after selective action choice; and independent per-resource minimum costs do not in general certify a feasible multi-resource continuation. We introduce FSPO, a feedback-state controller for budgeted LLM RL post-training that addresses these issues jointly. FSPO learns a policy-consistent risk-to-go model whose Bellman target follows the same frozen controller used for future decisions, together with a long-horizon utility model. Decision-conditioned trajectory calibration (DCTC) calibrates risk on cross-fitted trajectories generated by actions selected by provisional controllers. A Pareto resource continuation certificate (PRCC) admits an action only when a non-dominated cumulative reservation remains feasible over the residual horizon. Under a matched GRPO resource envelope, FSPO reaches 66.11% held-out and 59.43% OOD accuracy, compared with 64.47% and 57.03% for PB2, the strongest evaluated adaptive baseline. Three paired training seeds give gains of +2.42 and +3.19 percentage points over the contextual bandit on held-out and OOD evaluation. Under high behavior-deployment mismatch, policy-consistent risk lowers selected-decision ECE from 0.108 to 0.053; DCTC lowers it from 0.039 to 0.022 at matched acceptance; PRCC removes false-feasible admissions on an 18-action catalog ($0.197\rightarrow0.000$); and enabling all three components reduces trajectory failure from 0.181 to 0.083 in a factorial ablation.
arXiv:2610.02817v1 公告类型：cross 摘要：大语言模型（LLM）水印为识别由特定模型生成的文本提供了一种轻量级机制，但其在后处理攻击下的鲁棒性仍然较为脆弱。删除攻击尤其具有挑战性，因为它会导致词元位置发生偏移，并破坏观测词元与其原始水印位置之间的对齐关系。我们提出 Reed--Muller Code Watermarking（RMCW），一种基于 Reed--Muller 码的 LLM 水印方法。不同于全局码字恢复，RMCW 搜索能够存活的局部代数结构，利用 Reed--Muller 码字的仿射线限制所诱导的 Reed--Solomon 一致性。在生成过程中，RMCW 通过由密钥控制的词汇表划分，将 Reed--Muller 结构注入序列。在检测过程中，它将给定文本映射到由密钥确定的词汇分箱，并使用 Berlekamp--Welch 测试检验局部子序列是否具有低次数 Reed--Solomon 一致性。在 C4 和 ELI5 数据集上使用 OPT-1.3B 和 Llama-3.1-8B-Instruct 进行的实验表明，RMCW 保持了较强的干净文本可检测性，并且在多种删除攻击和改写攻击下，性能优于或持平于基线方法。我们的代码可在 https://github.com/BaichengDanny/RMCW 获取。
arXiv:2610.02663v1 Announce Type: cross Abstract: Despite the remarkable empirical success of flow-matching models, their statistical generalization guarantees remain underdeveloped. Existing analyses often impose restrictive assumptions on the estimated velocity field and yield convergence rates that fail to reflect the intrinsic low-dimensional structure common in real data, such as natural images and molecular geometries. In this work, we study the statistical generalization of flow-matching models for learning an unknown distribution $P_{\mathrm{data}}$ from finitely many samples. We derive finite-sample error bounds on the learned generative distribution, measured in the Wasserstein-$p$ distance, for all $p\geq 1$. Specifically, given $n$ i.i.d. samples from $P_{\mathrm{data}}$, we show that, for every $d>d_p^\ast(P_{\mathrm{data}})$ and appropriately chosen network architectures and hyperparameters, the learned distribution $\widehat{P}^{\mathrm{FM}}$ satisfies $ \mathbb{W}_p(\widehat{P}^{\mathrm{FM}},P_{\mathrm{data}}) \lesssim n^{-1/d}+n^{-1/(2p)}\bigl(\log(1/\xi)\bigr)^{1/(2p)}$ with probability at least $1-\xi$, where $d_p^\ast(P_{\mathrm{data}})$ denotes the Wasserstein-$p$ dimension of the target measure. Our results demonstrate that flow matching naturally adapts to the intrinsic geometry of data and mitigates the curse of dimensionality, as the convergence exponent depends on the intrinsic rather than ambient dimension. These guarantees remain meaningful in high-dimensional regimes and provide a theoretical explanation for the empirical success of flow matching on structured data distributions under substantially more relaxed assumptions than those in existing analyses.
arXiv:2610.02641v1 Announce Type: cross Abstract: Removing an input-scaling module changes both a classifier and the perturbations reaching its encoder. A robustness difference can therefore reflect the comparison rule as well as the module. We demonstrate this problem in a four-qubit quantum-attention detector on generated power-grid trajectories. A learned scaling module appears beneficial at a fixed physical attack budget, but matching an upper bound on perturbations at the encoder reverses the ordering. Neither comparison alone establishes a robustness benefit caused by the module. The initial test also perturbs clean examples into attacked examples while retaining their original labels; tests restricted to already attacked examples do not establish a benefit. Replacing a trained model's input scales disrupts detection. Retraining its linear classification layer restores the detection rate, but changes individual predictions, leaving the comparison descriptive rather than causal. Two further design checks explain why the input quantum Fisher information regularizer cannot train this model's query parameters, and why removing confidence bounds does not establish a larger certified radius. The evidence is limited to ten seeds, exact simulation, synthetic data, and a restricted set of attacks; classical baselines achieve better clean prediction. The practical lesson is to specify which perturbation budget is fixed, check that attacks preserve labels and interventions preserve predictions, and distinguish exploratory controls from confirmatory evidence.
arXiv:2610.02636v1 Announce Type: cross Abstract: Quantum Fourier sampling may help audit the spectral learnability of delay-based physical unclonable functions (PUFs). We ask whether that promise survives access matching, a strong classical comparator, and oracle synthesis. Three gates structure the evaluation. Structure: low degree is not small support at reachable challenge lengths; for 4-XOR at $n=14$, degree $\le d_f(0.1)$ admits $91\%$ of all $2^n$ characters and the median $90\%$-mass set spans a third of the spectrum. Algorithmics: constructing the phase oracle logically implies classical membership access, making Kushilevitz--Mansour the correct baseline; across 45 tasks it exhausts each finite domain, and no 4-XOR ideal-sampling case reaches $90\%$ mass within $2^n$ calls. A quantum-kernel diagnostic appears more favorable, with geometric difference rising to $2.151$ at $N=512$ challenges, but it correlates $0.991$ with $1/\sqrt{\lambda_{\min}(K_C)}$ for the classical Gram matrix $K_C$, and the 4-XOR label-complexity ratio does not exceed a balance-preserving permutation null ($p=0.930$). Trace-normalized geometric difference can therefore grow through classical ill-conditioning alone, without task-label alignment. Implementation: a simulator-validated fixed-point phase oracle based on the quantum Fourier transform admits an $18.9\%$ routed-depth reduction, yet the least certified precisions have estimated durations of $1.18$--$1.55\times$ the median dephasing time $T_2$ of the mapped qubits on a static backend snapshot, without hardware execution. We find no end-to-end advantage in the evaluated regime, although ideal sampling does use fewer coherent calls on the thresholded task. The contribution is the Three-Gate Quantum Audit Protocol: a reproducible procedure separating an ideal query advantage from a realizable security benefit. This is not a claim about deployed silicon and not an impossibility result.
arXiv:2610.02616v1 公告类型：cross 摘要：运行框架演化能够改进 LLM 智能体的提示词、工具和工作流程，但优化器自身的工具和流程通常保持不变。我们研究了这样一个问题：如果优化器同时改进其诊断故障、制定修改方案和测试修改效果的方式，它是否能更有效地改进另一个智能体。我们的设计基于两点观察。在一项受控研究中，如果没有基于执行的验证，优化器自演化无法提升性能；但在具备验证能力时，它取得了该研究中的最佳结果。在五个执行器上，自演化优化器为故障分析、验证、训练审计和工作流控制构建了自己的工具。受这些发现启发，我们提出了 VERSE，一种面向智能体运行框架的可验证自演化优化器。VERSE 允许优化器在提交前测试草拟的修改、重放故障，并扰动疑似出错的步骤，同时跨多个轮次跟踪修复和回归问题。利用这些反馈，优化器会同时修订执行器运行框架及其自身的提示词、技能、工具、钩子和笔记，而优化器模型与执行器模型的权重保持不变。在训练、验证和测试任务互不重叠的共享协议下，VERSE 在留出的 SWE-rebench 任务以及五种语言中的更新分布外任务上，提升了评估的全部四种运行框架优化器的表现。通过验证集选出的最佳运行框架在这两类任务上的准确率分别达到 42.3% 和 37.7%，而最强基线分别为 39.2% 和 29.3%。代码可从 https://github.com/wzekai/VERSE 获取。
arXiv:2610.02503v1 公告类型：cross 摘要：要可靠且安全地部署复合 AI 系统，需要理解在组件边界处出现的故障模式，而非仅关注单个模型内部的故障。级联错误会跨组件边界传播，隐性质量下降难以被标准监控发现，而协调失败则会使各个单独看来都正确的组件产生错误的集体行为。我们分析了来自开源复合 AI 项目的 150 份生产事故报告，以及经过匿名化处理的企业部署案例，构建出包含 23 种故障模式的分类体系，并将其划分为五类：检索故障、生成故障、工具故障、编排故障和集成故障。针对每一类故障，我们提出了韧性模式，并通过受控故障注入实验测量其有效性。断路器可将级联传播降低 89%，输出质量门禁可在影响用户之前捕获 73% 的隐性质量下降，而组件隔离可将影响范围缩小 64%。与无结构化监控基线相比，实施目录中三种或更多韧性模式的系统可将平均恢复时间（MTTR）缩短 71%。我们将该事故分类体系和模式目录作为面向实践者的资源发布。
arXiv:2610.02466v1 Announce Type: cross Abstract: We present sparse dictionary differentiable predictive control (SD-DPC), a framework for learning sparse, interpretable feedback policies for nonlinear systems from data. A prediction model is first identified by rollout-based sparse identification of nonlinear dynamics (SINDy), building on gradient-based and multistep formulations. The policy is then parameterized as a sparse combination of dictionary functions and trained by differentiating a constrained finite-horizon predictive-control objective through this model, so that its terms are selected by closed-loop performance rather than by imitating a previously trained controller. The result is an explicit feedback law with only a handful of terms. Across three benchmark control problems, SD-DPC satisfies the constraints in all test scenarios, outperforms a policy distilled onto the same terms by up to an order of magnitude, and requires orders of magnitude less memory and online computation than an optimization benchmark, while admitting explicit sensitivity bounds.
arXiv:2610.02432v1 公告类型：cross 摘要：生产系统中的大型语言模型（LLM）面临提示注入、木马（后门）以及对自动质量指标的操纵。本论文提出用于评估和提升LLM对对抗性输入序列变化鲁棒性的模型、方法和算法。我们提出 R_stab(f)，一种基于小幅输入扰动下逐步输出分布之间 Jensen-Shannon 散度的生成式鲁棒性指标。对于局部攻击，我们证明 V(h) <= 1 - R_class(h)，其中 R_class(h) 表示决策算子 h 在小幅扰动下保持其决策的概率。对于非局部攻击，我们提出一种经过校准的经验模型。针对 LLM-as-a-Judge 系统，我们开发了 ASA，这是一种自适应进化式黑盒攻击，其攻击成功率（ASR）最高可达 73.8%，在开放模型之间的迁移率最高可达 62.6%。在 Trojan Detection Challenge 2023 数据（Pythia-1.4B）上，替代触发器的 REASR 约为 0.99，而真实触发器的召回率约为 0.17，基线约为 0.14。在 SaTML CTF 2024 中，我们系统化总结了多层防御的四类绕过方法，将 ASR 从 90% 降至 15-25%。由 5-7 个异构模型组成的委员会将 Gemma-3-4B 的 ASR 降低 47-55 个百分点；使用 7 个模型时，ASR 降至 19.3%。针对基于 Model Context Protocol（MCP）的智能体系统，我们提出 AttestMCP，该方法使用受 HMAC 保护的数据包对工具调用进行证明认证，每次调用耗时低于 0.1 毫秒，并提出 Commit Boundary 隔离模式。在包含 847 个场景的 MCPBench 基准测试中，这些方法将平均 ASR 从 53.7% 降至 12.4%。相关方法已在 JudgeGuard 和 TrojanArmor 软件套件以及 MCPSec 模块中实现。
arXiv:2610.02403v1 Announce Type: cross Abstract: Energy saving has become a critical challenge in 5G and beyond networks. The rapid growth of connected devices has increased the overall network energy demand, driving operational expenditure to unsustainable heights. The Base Station (BS) accounts for the largest share of energy usage, typically consuming around 60-70\% of the Radio Access Network (RAN)'s total energy. Therefore, to address this issue, this article optimizes the BS's energy usage while accounting for the dynamic behavior of User Equipment (UE). Deep Reinforcement Learning (DRL) is a natural candidate for determining effective energy saving policies, such as automatically switching BSs on or off when user density is low or adjusting transmission power to balance energy efficiency and Quality of Service (QoS). However, its heavy training burden and the exponential growth of state and action spaces in dense 5G environments make exploration increasingly difficult. To overcome these limitations, we introduce a novel Quantum Reinforcement Learning (QRL) algorithm that leverages quantum principles, including superposition and entanglement, through parameterized quantum circuits, enabling significantly faster convergence than DRL, which relies on conventional deep neural networks. Extensive simulations demonstrate that the proposed QRL can substantially reduce energy consumption while maintaining QoS, even when UEs are highly dynamic and frequently switch their association with BS antennas. Additionally, QRL consistently outperforms DRL and Q-Learning in both convergence speed and learning complexity.
arXiv:2610.02357v1 Announce Type: cross Abstract: Recent advances in deep learning for time series prediction have amplified the need for reliable uncertainty quantification. Conformal prediction has gained attention as a distribution-free framework for constructing prediction intervals with coverage guarantees. However, its coverage guarantees rely on data exchangeability, an assumption generally violated in time series. Active research has focused on developing conformal prediction methods for time series that overcome this limitation. While deep sequence models, such as recurrent neural networks and Transformers, have often been used in conformal prediction for time series, limited work has systematically studied how deep sequence models can be utilized in conformal prediction for time series. In this work, we systematically investigate the use of deep sequence models in conformal prediction for time series through three approaches: conditional quantile regression, conditional quantile function estimation, and localized conformal prediction. We provide a theoretical analysis establishing asymptotic conditional coverage guarantees for all three approaches under suitable assumptions. Through comprehensive experiments on real-world datasets, we demonstrate the effectiveness of leveraging deep sequence models into conformal prediction for time series.
arXiv:2610.02339v1 发布类型：cross 摘要：即使单个回合效率低下或未能成功，机器人演示中仍可能包含有用的行为。轨迹拼接提供了一种将这些行为组合起来、生成改进训练数据的方法，但在高维机器人数据中，识别有用的连接并验证其可行性十分困难，因为许多已有方法依赖低维状态表示。我们提出 NEEDLE，这是一种离线数据增强算法，通过在高维机器人演示中的已记录观测之间添加简短且经过验证的动作桥接，来应对这些挑战。首先，NEEDLE 仅使用 RGB 图像、本体感知信息和回合级结果，无需新的环境交互或特权对象状态，即可识别并创建能够绕过次优绕路、扩大动作覆盖范围的连接，并利用失败轨迹增强原始数据集。随后，我们提出一种采样技术，将被接受的桥接纳入策略训练，而无需合成中间图像或丢弃原始演示，使策略能够学习替代动作，同时保留原始数据集的覆盖范围。在真实机器人任务中，NEEDLE 在每项任务上的成功率都超过最强基线，平均提升 21 个百分点。视频和补充材料见 https://needle-work.github.io/。
arXiv:2610.02338v1 公告类型：cross 摘要：越来越多的研究表明，在灵巧操作中，触觉感知能够为机器人策略提供与视觉互补的接触信息。然而，视触觉机器人数据仍然十分稀缺：灵巧操作演示需要通过远程操控机器人完成，这限制了数据集的规模。采集人类演示的成本要低得多，也为扩大数据规模提供了途径，但前提是人手和机器人手都配备具有对应信号的触觉传感器。这要求传感器能够适配不同的手部几何形状、覆盖整只手，并在不同实体之间采用统一的布局。我们提出了 SoTa，这是一种低成本电容式触觉皮肤，可在人体手部和机器人手部实现全手覆盖，同时在对应的手指和手掌区域保留由202个触觉像素组成的统一布局。我们的多层设计采用织物电极，使得在内部即可制造厚度薄、质地柔软且几何形状可定制的皮肤，每张皮肤的材料成本低于10美元。在10,000次加载—卸载循环后，该传感器仍能保持其初始响应范围的97%以上；在1,280次紧握拳折叠循环中，传感器线路仍保持连续。统一的触觉像素布局支持人机协同训练，可使用通用触觉编码器，无需学习跨传感器映射。在三项接触密集型操作任务中，与仅使用视觉的策略相比，触觉观测提高了分布内成功率。在机器人演示预算固定的情况下，加入人类演示后，八种评估条件下的平均成功率提升至原来的两倍以上，从22.8%提高到45.9%，并在全部五种分布外条件下提升了成功率。我们计划将制造和使用这些皮肤所需的资源开源。
arXiv:2610.02334v1 Announce Type: cross Abstract: Mobile embodied AI networks (MEAN) enable embodied agents to perceive, reason, communicate, and act in wireless environments. In such networks, agent mobility can improve channel conditions, while semantic compression can reduce transmission payloads. However, movement consumes energy, and stronger compression incurs additional computational cost. This paper studies joint movement, semantic compression, and transmit power design for an uplink MEAN system. We formulate a max-min energy efficiency (EE) problem by jointly optimizing transmit power, movement distance, and semantic compression ratio under controllable power constraints. The problem is non-convex due to the coupled signal-to-interference-plus-noise ratio (SINR), mobility-dependent channel gains, and fractional EE objective. To solve it, we propose an alternating optimization (AO)-Dinkelbach algorithm, where the fractional objective is handled by the Dinkelbach transformation, transmit power is updated via successive convex approximation (SCA), and movement distance is updated by coordinate-wise grid search. Simulation results show that the proposed scheme outperforms no-mobility and no-compression baselines, demonstrating the benefit of jointly exploiting mobility control, semantic compression, and power allocation in MEAN.
arXiv:2610.02320v1 公告类型：cross 摘要：计算机使用代理需要在复杂的桌面场景中可靠地定位操作目标，因为多个应用程序、相互重叠的窗口以及视觉上相似的控件会争夺注意力。现有训练数据很少将此类场景与密集标注配对，也很少以可控方式改变这些场景。我们提出 DeskForge，这是一种可控的桌面环境，通过组合并探索真实应用程序，为计算机使用代理生成大规模监督数据。该环境能够改变应用程序状态、内容、窗口布局、外观和分辨率，并将屏幕截图、可访问性树和窗口几何信息融合为密集的元素标注，同时记录每次执行操作的结果。借助该环境，我们构建了 DeskForge-1M，这是一个包含 120 万条已标注桌面观测和 1.597 亿个元素实例的语料库。我们从 DeskForge-1M 中抽取 20 万个定位示例，用于微调四个视觉语言模型。四个模型在留出的桌面条件下以及全部五个外部 GUI 定位基准测试中均得到提升；对于 Qwen3.5-4B，其在 ScreenSpot-Pro 上的准确率提高了 11.51 个百分点，在 OSWorld-G 上提高了 10.11 个百分点。这些提升也转化为长时程任务完成能力的提升：在固定规划器下，微调后的操作模型能够完成更多 WebArena-Infinity 和 OpenApps 任务，其中 Qwen3.5-4B 分别将完成数量从 119 个任务中的 31 个提升至 50 个，以及从 100 个任务中的 3 个提升至 15 个。结果表明，对真实桌面环境进行可控组合能够为提升 GUI 定位能力和长时程计算机使用能力提供一种可扩展的监督来源。框架代码、数据集和微调后的模型可从项目页面获取：https://saidgurbuz.github.io/deskforge/
arXiv:2610.02302v1 公告类型：交叉 摘要：近期研究表明，大语言模型（LLM）可能容易受到越狱攻击的影响；在这类攻击中，有害意图通过与良性任务组合而被隐藏。一个单独提出时会被拒绝的有害请求，在嵌入一个看似良性的更大查询后，可能会引发不同的响应。我们从信息论角度研究这类组合式意图隐藏型越狱。我们的形式化方法为每项任务关联一个被判定为有害的估计概率：在完整任务集合上的平均值定义了有害意图的先验概率，而在包含目标任务的选定任务包上的平均值定义了后验概率。选择辅助任务，使这两个平均值一致——我们称之为先验—后验匹配——可以在有害目标仍然处于任务包中的情况下，使估计意图保持不变。我们研究了两种设置，其区别在于查询构造是否属于优化过程的一部分。在查询无关设置中，任务的选择不考虑它们将在最终查询中如何表达。我们证明，在任务包大小约束下实现精确的先验—后验匹配在计算上是困难的；针对分数权重，我们推导出一个最优的水位填充解，并刻画满足预设安全阈值的最小任务包。在查询相关设置中，任务选择与查询构造被联合考虑，并在最终生成的查询上评估意图隐藏和目标保留效果。我们针对不同任务包大小、查询生成器和多个开源模型，评估了越狱有效性以及目标行为的保留情况。结果表明，在所评估的搜索预算下，组合式查询能够诱发超出直接请求基线的目标行为；同时也揭示了一种权衡：随着任务包大小增加，对于若干模型而言，响应层面的目标保留率往往会下降。
arXiv:2610.02267v1 公告类型：cross 摘要：Agent Harnesses 在每项任务中会做出许多细小的、具有类型约束的决策：调用哪个模型、使用哪个工具、检索到的文本是否相关、输入是否包含注入内容。System-1 决策模型通过一次前向传播放出带有类别概率的答案，有望相比调用 LLM 大幅降低成本和延迟。我们对一个开放权重模型（Laya）和一个托管模型（Jev）进行了成对评估，涵盖基于 18 个公开来源构建的 11 个 Agent 决策点：7,283 个基础案例和 6,640 个鲁棒性变体，并采用字节级完全一致的输入、成对测试，以及跨硬件和跨日期的可复现性检查。在 11 个决策点中，Jev 有 9 个点的准确率显著更高（高出 10.8 至 46.0 个百分点）。在零样本模型路由上，两个模型都未超过随机猜测水平；在 RAG 相关性门控上，两者表现持平。当选项顺序反转时，Laya 会改变 30% 的答案；面对大量或相似的候选项时，其性能急剧下降（在 50 个最近邻工具的情况下准确率为 31%，而对于具有唯一正确工具的项目，Jev 的准确率为 98%）。我们还审计了自己的流程。三个分析错误和一个设计混淆因素扭曲了主要部署结论：遗漏了预筛选成本（报告节省 23.9%，实际仅为 4.3%）；将门控准确率报告为端到端质量（58% 对 98%）；使用样本内阈值（目标为 5%，留出集漏判最高达 17%）；以及关于注入误报的“渠道效应”——使用渠道原生内容后，该效应便会消失。另外两个疑似混淆因素没有改变结论。所有案例、原始输出和分析代码均可在 https://github.com/David-DL-Space/sys1-eval 获取。
arXiv:2610.03679v1 公告类型：新 摘要：细胞、生物体和流体的动力学通常被建模为随时间演化的概率分布。要从非配对快照中重建并外推这种演化过程，需要对其潜在过程作出假设。Wasserstein 梯度流是一种常见选择，但无法描述守恒动力学或周期性动力学。Wasserstein 空间中的拉格朗日力学可以同时涵盖这两类动力学，但现有学习方法都基于模拟：它们在每个训练步骤中运行数值求解器，因此训练成本高昂。我们提出 Double-Stitch，一种无需模拟的方法，通过惩罚沿学习得到的群体轨迹计算的运动方程残差来学习这类力学。我们从 Clebsch 变分原理推导出该方程，该原理不要求速度为梯度场，并证明当且仅当方程成立时，残差才会精确为零。我们在合成数据、单细胞数据集和海洋涡旋数据集上测试了 Double-Stitch，发现其在大多数任务上达到或优于梯度流方法和基于模拟的 WLM，同时训练速度比 WLM 快 $4$-$14$ 倍。我们在 https://github.com/BasisResearch/stitching 提供了 Double-Stitch 的 JAX 实现。
arXiv:2610.03646v1 Announce Type: new Abstract: Stationarity rewards memory, but after a change the same history can mislead. We ask when forgetting should be permitted. E-process-authorized Thompson sampling (e-ATS) gives each arm full-history and discounted Beta states. An anytime-valid e-process first authorizes the discounted state, then a reversible relevance score controls its influence. Before authorization, e-ATS exactly follows optimistic Thompson sampling (OTS). Under a Beta-Bernoulli prior-predictive stationary model, e-ATS's probability of ever departing from OTS is at most the chosen $\alpha_E$, without fitted thresholds. Relative to e-ATS, removing authorization increased mean normalized dynamic pseudo-regret by $38.4\%$ on the registered suite but reduced it by $7.5\%$ on the literature-derived replay suite. Therefore, evidence controls when adaptation begins, not whether it always helps.
arXiv:2610.03640v1 Announce Type: new Abstract: Neural network loss landscapes have many symmetries, which are preserved by gradient flow but broken by finite-stepsize stochastic gradient descent (SGD). A canonical example of such a symmetry is scale in homogeneous networks: one can scale up the parameters in one layer and down in the next without changing the network output. Previous work has documented cases in which SGD breaks this symmetry in favor of balancing gradient noise or minimizing fluctuations. Here, we show that the solution geometry of undercomplete linear autoencoders instead selects a preferred sign for scale drift: on the PCA solution manifold, SGD favors large decoder weights. This directed scale drift occurs on a slow timescale, and its dynamics admit an analytically-tractable effective description. However, it cannot continue indefinitely: increasing scale eventually drives the dynamics towards a finite-stepsize stability boundary. The resulting solutions are sharper than a balanced baseline in the sense of the maximum eigenvalue of the loss Hessian, but different sharpness measures can move in opposing directions. Thus, undercomplete autoencoders give a concrete illustration of how loss geometry can convert residual gradient noise into directed motion along a manifold of functionally-equivalent solutions.
arXiv:2610.03620v1 公告类型：新 摘要：在线强化学习（RL）使机器人策略能够通过物理交互不断改进，但随着其能力水平的变化，所需的辅助也会随之改变。因此，基于离线估计或固定决策规则的现有干预策略可能会与当前策略不匹配。为解决这一问题，我们提出了 UniIntervene++，一种自适应干预智能体，能够在在线强化学习过程中学习如何在自主执行与异构辅助行为之间分配控制权。具体而言，UniIntervene++ 首先将不断演化的 RL 策略、轨迹修正以及任务结构化的 CodePolicy 统一建模为半马尔可夫决策过程中的 Options，并在线学习它们的相对价值。在此基础上，能力自适应干预通过无辅助执行定期探测 RL 策略，使控制分配能够及时响应其不断变化的能力。最后，耦合式经验学习使辅助行为能够改进 RL 策略，而策略不断演化的结果又会反过来重塑未来的干预决策。通过这种方式，随着 RL 策略不断改进，UniIntervene++ 能够联合决定何时进行干预、如何进行干预，以及何时交还控制权。在五项真实世界操作任务中，UniIntervene++ 的平均成功率达到 89.67%，至少高出所有基线 6 个百分点；同时将人工干预比例降至 0.77%，相较于最佳基线至少减少 94.6%。代码已发布于我们的 \href{https://github.com/dannyyudong/An-Adaptive-Intervention-Agent-for-Efficient-Real-World-Reinforcement-Learning}{GitHub repository}。
arXiv:2610.03556v1 公告类型：新 摘要：关于 GNN 中过度挤压与长程交互之间联系的经验性论断，只有在用于验证这些论断的基准确实需要长程交互时，才值得信赖。事实上的标准 Long Range Graph Benchmark 已被反复证明会被经过调优的短程模型饱和，而现有的合成替代方案则与特定拓扑结构绑定。因此，在任意图上，都缺乏一种具有原则性的长程性证书。这一现状反映出人们尚未对长程基准作出精确刻画。我们通过提出四条可验证公理来弥补这一根本性空缺：可预测性、紧致性、严格 $k$-范围性和拓扑不变性；任何声称测试 $k$-跳交互的任务都必须满足这些公理。我们从形式上证明，违反其中任意一条，都会产生损害基于该任务所得结论的失效模式。在这些公理的基础上，我们提出 TRIP（Truly Ranged Interactions Problem）及其推广 GRIP（Generally Ranged Interactions Problem），这是一组构造性程序，通过从稳定分布中抽取特征，将任意图转换为具有可证明长程性的任务。此外，GRIP 的构造使其能够针对每个范围给出闭式最大似然 oracle，从而为任何基准首次提供先验的逐范围测试误差下界。利用我们的框架，我们：(i) 审查 4 个常见的长程基准，并根据我们的公理识别其失效模式；(ii) 在由 TRIP 实例化的拓扑结构上发现，一种流行的曲率概念与 GNN 性能不相关，从而支持拓扑瓶颈与计算瓶颈之间的区分；(iii) 展示一种新基准的过度挤压度量还受到纯粹长程性之外的因素影响。用于使用该框架并复现实验的代码已发布于 https://github.com/ferranhernandezc/graph-grip。
arXiv:2610.03419v1 Announce Type: new Abstract: In this work we formulate ultrasound multistatic recovery from arbitrary transmit sequences as a Bayesian inference problem. To that end, we train a deep generative prior on multistatic data sets to tackle the rank-deficient regime in which classical linear REFoCUS decoders fail. This appproach, which we term Deep Bayesian REFoCUS, outperforms the linear baselines for all regimes of rank-deficiency and noise levels, and regresses to linear decoding when inversion is exact. The model also expresses uncertainty in the null space of the acquisitions, whereas the linear REFoCUS decoders only provide point estimates. Finally, we analyze the impact of distribution shift between simulation and in-vivo acquisitions, showing remarkable generalization ability without any fine-tuning or adaptation.
arXiv:2610.03306v1 Announce Type: new Abstract: Existing convergence analyses of Muon either assume exact orthogonalization or analyze classical Newton--Schulz polynomials, and guarantee only stationarity, so it is unresolved what Muon's five tuned Newton--Schulz steps preserve and whether that suffices to reach a prescribed neural-network training loss. We establish a finite-time training guarantee that accounts for both momentum accumulation before orthogonalization and the tuned finite-step update. For full-batch training of a sufficiently wide two-layer ReLU network with fixed random output weights and a positive-definite limiting neural tangent kernel, we prove that Muon reaches any target empirical squared loss $\varepsilon>0$ with high probability over initialization. For every momentum parameter $\mu\in[0,1)$, a target-dependent constant learning rate proportional to $(1-\mu)\sqrt{\varepsilon}$ yields a hitting-time bound of $O((1-\mu)^{-1}\varepsilon^{-1/2})$, with other problem parameters fixed. The sufficient width is independent of both target accuracy and momentum. The analysis shows that the tuned Newton--Schulz map preserves alignment with the momentum buffer while bounding the update's spectral norm. Control of gradient variation near initialization transfers this alignment to the current gradient, ensuring descent until the target is reached without requiring exact orthogonalization. Numerical experiments support these mechanisms at widths below the sufficient theoretical threshold: gradient-update alignment remains above the analytical reference, and all 30 runs across six widths and five student initializations on a fixed teacher-student dataset reach the target loss while maintaining kernel positivity.
arXiv:2610.03303v1 发布类型：新提交 摘要：随机偏微分方程（PDE）的不确定性量化（UQ）在计算科学与工程中无处不在。然而，针对这类问题的经典谱求解器面临维数灾难，而现有的神经网络求解器通常忽略了随机结构，而正是这种结构使矩和校准变得易于处理。我们提出了一种随机可分离物理信息神经网络，称为 S$^{2}$-PINN。该网络通过可学习的高斯空间字典、傅里叶时间特征以及广义多项式混沌（gPC）随机基来表示随机偏微分方程的解 $u(t,\mathbf{x},\mathbf{Z})$，并由低秩 Canonical Polyadic（CP）张量分解核心将三者耦合起来。该方法采用强形式残差损失与 gPC 投影残差损失相结合的混合损失进行训练。我们的理论分析表明，在温和条件下，可分离函数类在 $L^2$ 中稠密，并且投影残差恰好对应于随机 Galerkin 约束。此外，我们证明了小批量投影系数与 gPC 模态数量呈对数依赖关系，并且正交性惩罚能够控制所学习空间字典的条件数。通过四个构造的随机偏微分方程基准测试，我们表明，与九种基线方法相比，S$^{2}$-PINN 在均值和方差精度以及校准方面均表现更优，同时使用的参数显著更少。在非构造的 Poisson 和 Darcy 问题、一个随机 Navier--Stokes 问题、高随机维度下的扩散尺度研究以及两个随机逆问题上的进一步评估，揭示了所提出结构的泛化能力。综上，这些结果表明，随机可分离性是物理信息神经网络不确定性量化的一项有效设计原则。实验代码可在 https://github.com/DMax1314/s2pinn 找到。
arXiv:2610.03199v1 公告类型：新 摘要：从同一基座模型微调得到的大型语言模型可以通过对其任务向量求平均来合并，但有些合并结果的性能会远低于基座模型，而常见的合并算子在评估前无法发出任何预警。我们表明，专家模型任务向量的一个统计量既能预测这种崩溃，也能校准其修复方式。平均操作所移除的功率等于各专家模型任务向量之间的方差，我们将其作为干扰度量。在一个可用的噪声模型下，合并所引入的扰动会随着合并系数和干扰程度的增加而增大，从而得到一个合并前评分。在我们针对四个模型系列的二十二种合并配置进行的实验中，只有破坏性合并的该评分超过了阈值。我们发现，专家模型之间符号冲突的统计量是现有合并算子的常见优化目标，但它们实际上具有反向预测性。随后，我们在评估十四种合并结果之前对其进行了预测，其中十二项预测正确，包括一对因继续预训练而被推过阈值的专家模型所产生的破坏性结果。为解决这种崩溃，我们提出 PRISM：一种先对任务向量求平均、再根据每层的干扰程度设定水平并对各层执行软阈值处理的算子。在无需数据或调参的情况下，PRISM 使全部五个破坏性合并的结果都保持在阈值之上，并且与基座模型的差异处于评估噪声范围内；相比之下，普通平均的结果要么至少低于基座模型 14.4 分，要么完全崩溃。我们仅在评分超过阈值时应用 PRISM，而对于低于阈值的合并则保留普通平均；后者包括全部十五个无害合并。代码见 https://github.com/js-lee-AI/PRISM】【。
arXiv:2610.03165v1 Announce Type: new Abstract: Several variance-reduced versions of REINFORCE based on importance sampling achieve an improved $O(\epsilon^{-3})$ sample complexity to find an $\epsilon$-stationary point, under an unrealistic assumption on the variance of the importance weights. In this paper, we propose the \algo (Defensive Policy Gradient) algorithm, based on defensive importance sampling, which achieves the same rate without any assumption on the variance of ordinary importance weights. We also establish lower bounds in a generalized black-box policy-optimization model that hides states and actions and permits parameter-dependent rewards. In this model, the optimal rates are $\Theta(\epsilon^{-4})$ with bounded-variance one-policy feedback and $\Theta(\epsilon^{-3})$ with mean-square-smooth coupled two-policy feedback. Under standard policy-regularity conditions, REINFORCE and \algo realize the corresponding oracle conditions and attain the $O(\epsilon^{-4})$ and $O(\epsilon^{-3})$ upper bounds, respectively. Although the lower bounds do not apply directly to the classical MDP interaction model in which these algorithms operate, this correspondence provides oracle-level evidence that the faster rate of \algo is optimal and genuinely separated from that of vanilla policy gradient.
arXiv:2610.03127v1 公告类型：新 摘要：硬件感知神经架构搜索（NAS）的主要瓶颈在于评估成本：每个架构都必须经过训练后才能获知其奖励。保序预测筛选器通过剪除预测奖励上界未达到阈值的候选架构来降低这一成本，并提供一种与分布无关的保证：被错误丢弃的候选架构比例至多为 $\delta$。但这一保证假设校准候选架构与测试候选架构之间具有可交换性，而外围的强化学习（RL）循环违反了这一假设：随着搜索推进，策略提出的候选架构不断改进；在逐层构建过程中，候选分布还会在每个回合内发生偏移。我们以在线反馈控制取代一次性分位数估计（即 Adaptive Conformal Inference，并采用无需调参、局部自适应和组条件化变体），从而恢复可引导的覆盖率：对于任意序列，调节目标值即可单调且可复现地获得相应覆盖率。在三个神经网络架构系列以及单步和序列式搜索设置下（使用三个随机种子），该方法能够将每个请求的覆盖率水平跟踪到约 ${\sim}10^{-3}$ 以内，同时在测得精度无损的情况下剪除 25%–50% 的评估；相比之下，静态校准会失去对覆盖率的控制，而高斯过程基线无论请求值如何都保持保守。最后，在一个受约束的测试平台上将同一乐观上界用作采集函数时，其性能优于随机搜索；这一收益固定乐观策略本身已经具备，而在线校准进一步提升了该收益。源代码可在 https://github.com/Vicomtech/rl-hw-nas 获取。
arXiv:2610.03034v1 公告类型：新 摘要：扩散模型依赖需要进行时间离散化的数值求解器，而时间离散化会显著影响采样成本与质量之间的权衡。然而，逆过程的计算难度会沿采样轨迹变化，也会因数据分布不同而变化，因此离散化方案的选择十分重要。我们将比例-积分（PI）步长控制方法应用于扩散过程，并使用我们针对扩散过程设计的噪声归一化误差估计器。不同于扩散领域现有的仅响应当前误差的自适应方法，PI 求解器还会纳入前一次误差，从而实现更加平滑的步长调整。我们进一步展示了，这些逐样本轨迹具有共享结构，可以将其聚合为一个固定调度方案，同时保留自适应采样的大部分优势。我们在自然图像和语言数据集上评估了这两种方法，以质量为指标，并在神经网络评估次数（NFE）相同的条件下使用 FID 进行衡量，同时将其与广泛使用的随机求解器和调度方案进行比较。对于图像数据，在使用随机 Heun 采样器时，我们的固定离散化方案在样本质量方面优于常用的 EDM 调度方案；在 NFE 较低时，使用 EDM-churn 采样器也能取得更好的效果。此外，我们的 PI 自适应求解器获得了优于大多数随机和自适应基线方法的 FID，但在低 NFE 条件下未能超过 EDM-churn 采样器。对于语言扩散，在低至中等 NFE 条件下，就困惑度而言，我们的求解器优于 EDM 调度方案和熵调度方案，但代价是较低的词元熵。最后，我们发现，逐样本自适应性的收益取决于具体问题。在一维玩具示例中，这种收益非常显著；而对于图像和语言数据，其收益仅较为有限，在某些情况下，平均调度方案甚至优于 PI 自适应求解器。代码位于 https://github.com/ellakemperman/adaptive-second-order-diffusion-solvers】【。
arXiv:2610.03000v1 Announce Type: new Abstract: Intrinsic explainability remains a challenging problem, particularly in contexts where multilayer perceptrons (MLPs) require dynamic re-training within an optimization environment. This paper investigates how MLPs and their training dynamics can be represented and studied in non-Euclidean spaces; our representation features the Poincar\'e model of hyperbolic geometry. We aim to capture the geometric evolution of their weighted topology and self-organization over time. Instead of restricting the analysis to single checkpoints---as per established measure-based explainability methods---we construct temporal \textit{parameter graphs}, i.e., snapshots over time $T$ steps of the optimization/training process for MLPs. This reflects the view that neural networks encode information not only in their weights but also in the trajectory traced during training. Drawing on the idea that many complex networks admit embeddings in hidden metric spaces where distances correspond to connection likelihood, we present a geometric and temporal graph-based metalearning framework for obtaining dynamic hyperbolic representations of the underlying neural parameter graphs. Our model embeds temporal parameter graphs in the Poincar\'e model ball, and learns from them while maintaining equivariance to within-snapshot neuron permutations and invariance to permutations of past snapshots. In doing so, the approach preserves functional equivalence over time and recovers the latent evolving geometry of the network. Experiments on regression and classification tasks with trained MLPs show strong meta-network performance, accompanied by hyperbolic temporal representations. This reveals how the network structure emerges over time under specific training environments, thus providing insights into the network's self-organization.
arXiv:2610.02954v1 Announce Type: new Abstract: Biologically-informed neural networks (BINNs) have emerged as a flexible subclass of physics-informed neural networks (PINNs) for learning terms in partial differential equations from data. BINNs are particularly suited for biological systems, where the governing equations are highly nonlinear and only partially known a priori, and where data observations are often sparse, noisy, and incomplete. However, applying BINNs effectively in practice depends critically on hyperparameter selection, which remains a central challenge in equation-learning frameworks. Hyperparameters are often chosen heuristically and only cursorily documented, which limits the reproducibility of results and the transferability of methods. We present a diagnostic workflow for hyperparameter selection that can be used when the ground-truth equations are not known. The workflow is guided by three main questions: (1) Are the benefits of greater network capacity worth the cost? (2) Do more training epochs keep reducing the validation loss? (3) Do the learned terms stop changing as network capacity and training increase? We apply our workflow to synthetic systems of varying complexity with known ground truth, spanning diffusion and growth right-hand side terms and data ranging from 1D+t to 2D+t. We demonstrate that the validation loss generally follows the true error in the learned terms and distil practical rules of thumb for selecting hyperparameters in the BINN architecture. By providing a structured workflow, practical guidelines, and suggested starting values for hyperparameter selection, this work lowers the barrier to BINN-based equation learning.
arXiv:2610.02872v1 Announce Type: new Abstract: Evaluating network interventions requires understanding how treatment affects people through their social relationships. Counting treated neighbors without distinguishing supportive and antagonistic ties can conceal opposing influences. We define effects through positive and negative ties, their interaction, and a sign-composition effect of reallocating treatment between the two types at a fixed total, and give their identification formulas. Under sign-blind assignment, we show how ignoring signs mixes the effects of the two tie types. We propose SiDE (Signed-exposure Doubly robust Estimator), which combines sign-specific outcome models with exposure probabilities induced by individual treatment assignment. We establish double robustness of its score and assess approximate intervals that account for overlapping neighborhoods. Semi-synthetic experiments on six real signed networks demonstrate accurate effect estimation and examine the limits of interval coverage. An exploratory reanalysis of published school-experiment data yields a positive estimate of the peer effect through spend-time ties on wristband wearing, but the intervals for all four effects include zero after adjustment for multiple comparisons. This framework can inform network intervention design by showing when influences through the two tie types reinforce or offset one another.
arXiv:2610.02835v1 发布类型：新 摘要：在使用可验证奖励强化学习（RLVR）对大型语言模型（LLM）进行后训练期间，GRPO 风格的算法可能会出现严重的后期崩溃。基于提示的探测表明，这并非无害的策略裁剪，而是有效策略容量的有害收缩，使不同的推理策略越来越难以访问。为刻画这一现象，我们通过轨迹层面的策略更新交互来定义策略，并构建了一个结合优化动态与信息论的统一理论框架。我们证明，主要的 RLVR 目标函数会逐步将概率质量集中到单一策略上，而维持非平凡的任务准确率则需要最低限度的策略容量。这两项结果之间的冲突为灾难性崩溃提供了机制层面的解释。我们进一步推导出 Mirrored Entanglement Index（MEI），将其作为一种轻量级的在线预警信号。为防止崩溃，我们提出 \textbf{Mesh Learning}，通过展现多种推理策略，防止任何单一策略主导优化过程。在 AIME26、AIME25、MATH-500、GPQA 和 LiveCodeBench 上，Mesh Learning 在 Qwen 和 Phi 模型系列中均持续优于强基线，性能提升最高分别达到 13.4 个百分点和 11.5 个百分点。这些结果表明，策略保留是实现稳定 RLVR 的关键原则。代码发布于 https://github.com/Ayanami-0123/Open-Mesh-Learning。
arXiv:2610.02822v1 Announce Type: new Abstract: Temporal Domain Generalization (TDG) has emerged to address real-world streaming data with distribution shifts over time. However, existing methods are either prone to overfitting to domain-specific noise in the data space or become overly complex and less interpretable in the parameter space. To bridge these gaps, we propose \textbf{AdaSpecK}, a spectral-Koopman framework with adaptive context extraction for TDG. To mitigate noise fitting to irregularly sampled domains, we introduce spectral-regularized Koopman dynamics modeling, which applies spectral-aware filtering in the latent space to extract denoised low-frequency trajectories and learn a Koopman operator to model the system dynamics in a linearized space. To model complex historical environments under non-stationarity, we design a context-informed heterogeneous pattern extraction mechanism. Specifically, we employ a target-conditioned attention module to attend to distinct past windows, producing a dynamic, target-specific historical summary. By constructing an environmental signature from the current evolutionary pattern, our model adaptively perceives which aspects of the past context are most informative for future prediction via a learned router. Extensive experiments on eight diverse classification and regression benchmarks demonstrate that AdaSpecK achieves state-of-the-art performance. The code and datasets are available at \href{}{https://anonymous.4open.science/r/Ada-Spec-K}.
arXiv:2610.02795v1 Announce Type: new Abstract: Deep graph learning models deployed in real-world systems often need to cope with non-stationary environments, where the underlying graph distribution drifts continually over time. Prevailing solutions rely on training auxiliary generative modules to synthesize memory graphs for cross-domain adaptation, which incurs substantial computational overhead and scales poorly under prolonged distribution shifts. We argue that a more economical path exists: rather than generating memory, one can crystallize it. To this end, we propose Efficient Memory Crystallization (EMC), a training-free test-time framework that distills each incoming graph domain into a compact, semantically faithful memory through a closed-form solution to a memory-oriented distribution-matching objective, thereby eliminating redundant domain information under continual covariate shifts. To preserve both generalizability and adaptability as the model traverses a long sequence of target domains, EMC further models inter-domain dependencies through state-evolving memories and admits a theoretically grounded, tighter generalization error bound than direct adaptation. Extensive experiments demonstrate the superior performance of EMC over state-of-the-art baselines on graphs under non-stationary distribution shifts, while reducing average runtime by 87.4% and GPU memory consumption by 92.4% relative to the recent competitor, making continual graph adaptation practical at scale.
arXiv:2610.02771v1 Announce Type: new Abstract: We study fixed-confidence best-arm identification under strict 1-bit feedback constraints. At each round, the learner selects an arm and a query set, and receives only a single bit indicating whether the sampled reward belongs to that set. We consider a distribution-free finite-variance setting with arm-wise localization, where direct empirical mean estimation is no longer available and clipping becomes unavoidable. We first formulate a time-uniform 1-bit mean-estimation primitive based on randomized threshold queries and a clipped tail-integral identity. We then embed this primitive into candidate-challenger best-arm identification algorithms. A fixed-clipping algorithm gives a simple anytime $(\epsilon,\delta)$-PAC guarantee, while a phased adaptive-clipping algorithm matches the clipping level to the current resolution and yields a gap-adaptive sample complexity. We also prove a $K$-arm worst-case information-theoretic lower bound showing that the logarithmic penalty caused by finite-variance 1-bit feedback is intrinsic. This bound matches the leading dependence of the phased algorithm up to lower-order $\log\log$ factors.
arXiv:2610.02763v1 Announce Type: new Abstract: Tropical forest monitoring is essential for global climate stability and biodiversity preservation. To address the urgent need for rapid, reliable detection of forest loss which is essential for timely intervention against illegal logging, supply chain transparency, land-use governance and carbon market standards, we introduce a two-stage statistics-encoder cascade for near-real-time anomaly detection using Sentinel-1 time series. Our system is designed to overcome two fundamental challenges in remote sensing: the cloud-cover limitations that restrict optical monitoring and seasonal backscatter variation that causes SAR systems to mistake natural moisture changes for forest loss. The architecture integrates two distinct analytical engines to ensure high-fidelity detection: (1) an adaptive, robust-statistics z-score test on co-registered Sentinel-1 VH backscatter, same-season historical baseline and (2) a learned confirmation gate based on the latent-space structural similarity (SSIM) of a convolutional autoencoder trained on stable-forest patches. A candidate disturbance is confirmed as an alert only when both stages agree, and is assigned a confidence score and a Low/Medium/High risk tier from its repeat-occurrence history. The system produces per-alert auditable confidence scores and area-in-hectares estimates directly compatible with Monitoring, Reporting and Verification (MRV) workflows, sustainable forestry management, operational field checks and environmental risk assessments. Beyond its primary application, the model's flexibility allows for critical environmental applications ranging from selective logging to large-scale agricultural encroachment mapping, flood mapping and so on.
arXiv:2610.02701v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models can increase parameter capacity without proportionally increasing active computation, but it is unclear how this trade-off behaves in particle-physics transformers. We study dense and MoE Particle Transformers on 188-class JetClass-II, varying expert count, routing capacity, top-K, and auxiliary loss. We find that, when token dropping is avoided, top-1 MoE models improve over the dense baseline at nearly unchanged nominal forward compute, while further increasing the number of stored experts produces little additional accuracy gain. Activating multiple experts per token yields additional predictive improvements at higher computational cost. Routing analyses show that expert assignments become more strongly associated with particle identity and kinematics in some configurations, but this structure does not increase monotonically with classification performance. These results highlight the need to distinguish stored parameter capacity, active computation, routing capacity, and routing organization when evaluating sparse expert models for jet classification. Code and experiment configurations are available at https://github.com/kpendiyala/MPT.
arXiv:2610.02695v1 公告类型：新 摘要：可靠的大语言模型（LLM）不仅必须生成准确答案，还必须表达能够忠实反映其答对概率的置信度。这种校准对于识别不确定预测以及支持现实部署中的可靠决策至关重要。近期研究将校准学习纳入强化学习（RL），利用真实正确性监督，联合优化答案正确性与语言化置信度。然而，这些方法依赖标注数据，限制了其在实际测试时场景中的适用性：在此类场景中，真实标签不可用，且校准可能需要适应新遇到的目标任务。为应对这一挑战，我们提出测试时校准学习（TTCL），这是一种无标签框架，能够直接利用未标注的目标任务数据，同时适应推理准确率与语言化置信度。具体而言，TTCL 从模型生成的多个响应中为正确性和校准推导自监督信号，从而无需真实标签即可在测试时进行校准学习。理论分析进一步证明，TTCL 是理想校准目标的有界替代目标。在数学推理和事实问答上的大量实验表明，TTCL 在多种模型和任务上持续提升准确率与校准效果。在基础模型上，TTCL 在八个基准测试中实现了平均相对准确率提升 +40.13%，并使 ECE 降低 +70.80%。此外，在领域偏移情况下，TTCL 还能进一步提升已完成校准模型的准确率与校准效果，尤其是在源领域校准难以迁移至目标任务时。在 math-to-factQA 设置下，TTCL 实现了平均相对准确率提升 +20.35%，并使 ECE 降低 +53.83%。源代码已发布于 https://github.com/tmlr-group/TTCL。
arXiv:2610.02610v1 Announce Type: new Abstract: The research presented here focuses on modeling machine-learning performance. The thesis introduces Seer, a system that generates empirical observations of classification-learning performance and then uses those observations to create statistical models. The models can be used to predict the number of training examples needed to achieve a desired level and the maximum accuracy possible given an unlimited number of training examples. Seer advances the state of the art with 1) models that embody the best constraints for classification learning and most useful parameters, 2) algorithms that efficiently find maximum-likelihood models, and 3) a demonstration on real-world data from three domains of a practicable application of such modeling. The first part of the thesis gives an overview of the requirements for a good maximum-likelihood model of classification-learning performance. Next, reasonable design choices for such models are explored. Selection among such models is a task of nonlinear programming, but by exploiting appropriate problem constraints, the task is reduced to a nonlinear regression task that can be solved with an efficient iterative algorithm. The latter part of the thesis describes almost 100 experiments in the domains of soybean disease, heart disease, and audiological problems. The tests show that Seer is excellent at characterizing learning-performance and that it seems to be as good as possible at predicting learning performance. Finally, recommendations for choosing a regression model for a particular situation are made and directions for further research are identified.
arXiv:2610.02563v1 公告类型：新 摘要：我们提出了 OpenGameEval，这是一个用于在 Roblox Studio 中进行智能体游戏开发的基准测试与评估框架。该框架让语言模型以智能体身份运行于可复现的、有状态的游戏引擎会话中，并通过可执行检查对每次运行进行评分，检查对象包括被编辑的场景以及模拟游玩会话。大多数智能体编码基准测试要求进行探索，却只对最终任务是否成功进行评分。OpenGameEval 在其八工具动作空间中将观察工具与编辑工具分离，因此可以直接衡量探索行为。我们针对 84 个由人工整理的核心任务，对 13 个前沿模型的通过率和探索行为进行了测量，每个任务尝试 16 次。这些任务对当前模型而言难度很高。在单次尝试中，表现最佳的模型能解决 51.7% 的任务；在五次尝试中全部成功的比例为 39.4%；此外，没有任何受测模型能够解决其中六个任务。处于前沿水平的模型通过解决不同任务达到了相近的通过率：按照任务所需的工作类型进行划分后，在脚本编写任务上，排名前五的模型通过率差距扩大至 5.0 个百分点；在场景修改任务上则扩大至 12.5 个百分点。探索行为可以预测一次运行是否成功。在任务和模型保持不变的情况下，对于仅涉及场景的任务，在执行操作前检查参考解所涉及的每个对象的运行，其通过率比完全不检查这些对象的运行高 13.4 个百分点；对于仅涉及脚本的任务，这一差距为 9.8 个百分点。我们以 MIT 许可证在 https://github.com/Roblox/open-game-eval 发布任务套件、其中的 place 文件、每个任务的注释、用于在 Roblox Studio 内运行任务的插件，以及更新后的排行榜。
arXiv:2610.02559v1 Announce Type: new Abstract: As neural networks continue to grow in scale, model compression is becoming increasingly important for efficient inference under limited computational resources. Structured pruning methods remove neurons or channels that are estimated to be less important, but the removed units may still contain useful information. From the viewpoint of coarse-graining a trained network, it is valuable to ask which information should be retained when multiple neuronal degrees of freedom are consolidated. In this paper, we discuss cluster-based merging methods for compression of trained neural networks. In addition to a data-free contribution-weighted averaging method, we propose neuron-merging methods in which neuron responses are mapped back to the pre-activation space via the inverse activation function, and the weights and biases of each representative neuron are estimated using the least-squares method. We also examine both a data-assisted strategy with actual training inputs and a data-free strategy using randomly generated inputs. The comparisons provide empirical evidence, in the tested sigmoid networks, that weight information is particularly useful for clustering whereas activation information is useful for representative-neuron reconstruction in the merging process.
arXiv:2610.02528v1 公告类型：新 摘要：我们提出了一种用于求解 0-1 整数规划的自回归可微方法。我们固定二元变量的任意顺序，并训练一个 transformer 在保持位于可行集内的同时预测下一个比特。我们的方法首先利用任意求解器提供的可行当前最优解进行训练，从而能够将 transformer 初始化在可行集内。随后，我们的过程引入拉格朗日惩罚以惩罚不可行解，并使用松弛目标上的 Gumbel-softmax 激活函数，进一步训练 transformer 来探索可行集。我们在非凸二次背包问题实例上测试了该方法，并证明对于最多包含 10,000 个二元变量的稠密问题，该方法相较于当前最先进的开源求解器能够持续取得改进。特别是，我们通过实验证明了一种类似于隧穿效应的现象：该方法将变量从二元变量转换为 transformer 的连续权重，这种转换使其能够跨越松弛目标景观中的障碍。
arXiv:2610.02517v1 Announce Type: new Abstract: Graph-structured data plays a pivotal role in modeling complex relationships. However, real-world graphs are often incomplete due to data collection and observational constraints, severely limiting the effectiveness of modern graph learning pipelines. While existing Graph Data Augmentation (GDA) methods attempt to refine graph structures for improved downstream performance, they are typically label-dependent, computationally expensive, and inherently transductive, limiting their applicability in practical scenarios. In this work, we present a novel feature-centric graph data augmentation framework that bypasses explicit structure modeling by operating directly in the embedding space. Through a self-supervised inverse masking process, our method captures latent ties between observed and complete graphs, enabling recovery of unobserved structural signals through refined node representations. To enhance robustness under noisy and sparse supervision, we introduce a message regularizer and a bootstrap strategy for effective training and generalization. Evaluated on ten graph datasets spanning multiple domains, our approach, SelfAug, consistently outperforms state-of-the-art methods in both accuracy and efficiency across inductive and cold-start settings, highlighting its potential as a scalable and generalizable solution for real-world graph learning scenarios.
arXiv:2610.02462v1 公告类型：new 摘要：大语言模型压缩能够降低推理成本和内存需求，但由于相当的资源缩减可能带来不同程度的能力损失，方法和配置的选择仍主要依赖经验。我们系统研究了大语言模型压缩在剪枝、量化和蒸馏方面的能力缩减规律。我们的框架衡量模型在数学、代码生成和问答任务上的能力损失，并将这些测量结果与模型规模、训练阶段、压缩设置、数据可用性和训练暴露量联系起来。我们构建了简单的预测关系，并评估其准确性、测量效率，以及对未见过的配置和模型状态的泛化能力。在不同剪枝水平之间共享密度响应后，拟合剪枝预测器所需的配置测量数量减少了一半：在新的 Pythia 状态、预注册的 OLMo-2 测试状态以及 Wanda 剪枝条件下，这一紧凑关系在数学和代码任务上与使用全部测量结果拟合的回归模型相差不超过每个词元 0.020 纳特，其中系数针对每种设置分别重新拟合。受控蒸馏实验表明，在不同问答分布中，大量重复使用数据的成本会反复出现，而其净收益取决于评估分布。我们进一步通过将数值选择与配置中位数和固定方法优先级进行比较，评估这些预测的决策价值。在两个模型系列上的独立评估显示，在所测试的候选集合中，对于问答任务，选择方法能够捕获大部分可获得的跨方法收益；固定的方法优先级也能达到相同的遗憾值，而数学和代码任务的改进空间较小。这些结果阐明了能力缩减规律的预测范围，以及如何利用它们选择压缩方法。我们的代码公开于：https://github.com/LabRAI/scaling_down_law】【。
arXiv:2610.02438v1 公告类型：新 摘要：大型语言模型（LLMs）在代码生成方面展现出强大性能，而成功与否取决于两方面：回忆相关的算法知识，以及推理如何应用这些知识。然而，现有的 LLMs 流程是不透明的，没有明确区分这两个组成部分。我们认为，对于那些规范实现已广泛存在于预训练语料库中的知名算法，代码生成更适合被衡量为\textit{参数化代码检索}：即从内化的知识中复现一个指定名称的算法，而不是合成一个全新的算法。我们推出了 AlgoREval，这是一个包含 599 个问题的基准测试，涵盖 14 个领域中的 77 种经典算法、7 种编程语言以及 4 种图输入表示形式，用于单独评估这一能力；同时，我们在零样本设置下评估了 15 个模型（参数量为 7B--34B）。我们发现，即使对于有广泛文档记录的算法，不同语言和输入表示形式之间的检索准确率也存在显著差异；我们还发现，使用检索到的代码片段或结构化算法提示来增强提示词，能够提升复杂算法的准确率，而 SFT 能够带来更广泛的语言增益，GRPO 则能在特定语言上带来更大的单语言增益。总体而言，我们的结果确立了参数化代码检索是一种独立且可测量的能力，并警示人们不要在缺乏系统验证的情况下部署 AI 生成的算法代码。\footnote{代码和数据集已发布于 https://github.com/Nickil21/AlgoREval
arXiv:2610.02431v1 Announce Type: new Abstract: Deep neural networks achieve high accuracy through layered numerical transformations, yet their decisions remain difficult to audit because decision evidence is encoded in hidden activations rather than explicit rules. This paper introduces CRISP, a framework that reconstructs the last-layer activation vector (LLAV) of binary neural teachers as Tsetlin Machine (TM) clauses. CRISP sign-binarizes the teacher's penultimate pre-logit activations, and assigns one Individual TM (ITM) to each LLAV neuron. Each reconstructed hidden bit is represented by propositional clauses over Booleanized input features, which gives a direct symbolic trace from named input thresholds to a named teacher neuron. CRISP is evaluated on MNIST, KMNIST, FashionMNIST (FMNIST), SVHN, and CIFAR10 using a BinaryConnect convolutional neural network (BCCNN) teacher and a fully binary neural network (BNN) teacher, with an additional study on binary thresholding, thermometer encoding, and quartile binning at multiple bit depths. The results show that LLAV sign-binarization does not reduce teacher-head accuracy in the tested BNN setting, while ITM reconstruction error is the main limiting factor. Quartile one-bit Booleanization gives the strongest reconstruction fidelity on SVHN at 87.52% test fidelity and is competitive on CIFAR10, and the reconstructed LLAV preserves 78.41% teacher-head accuracy on FMNIST. Pooled clause-evidence visualizations show that the learned ITM literals concentrate on the object region in centered benchmarks. CRISP therefore provides a clause-level route for inspecting the final hidden representation of binary neural teachers.
arXiv:2610.02345v1 公告类型：新 摘要：一类表格型异常检测器在平方误差损失下训练神经映射，使其趋向一个固定目标，并通过测试时的残差对异常进行评分；contraction matching、one-step rectified flow 和重构自编码器都符合这一范式。我们刻画了一种收敛崩溃现象：优化得越好，检测器的表现反而越差。在收敛状态下，所学习的映射甚至会在分布外跟踪目标，因此残差信号不仅在正常数据上消失，在异常数据上也会消失。因此，这些检测器依赖隐式的不收敛机制（提前停止、容量上限）来保留信号。我们认为这是结构性问题：有效的异常检测需要一种能够阻止无约束外推的局部性约束。经典检测器（kNN、KDE、isolation forests、LOF）会显式施加局部性约束，而固定目标神经检测器则不会。我们通过证明固定目标检测器的核回归对应形式是有限带宽的 Nadaraya-Watson 平滑器，来形式化二者之间的联系，并将其称为 Kernel Contraction Matching (KCM)。KCM 具有闭式形式、无需训练且 CPU 效率高，同时在 ADBench 上达到了与成熟神经基线相当的性能。在此基础上，我们提出 Kernel-Anchored Regularizer (KAR)，用于惩罚神经预测偏离训练目标的核加权平均值。在易发生收敛崩溃的 ADBench 数据集和三种主干网络上，KAR 均能缓解崩溃，并在延长训练时提升 AUROC。
arXiv:2610.02344v1 公告类型：新论文 摘要：联合嵌入预测架构（JEPAs）容易出现表征坍塌，通常通过经验性启发式方法加以缓解。我们提出了一套训练早期稳定性理论，将这些启发式方法统一起来。在平凡不动点附近对耦合的 JEPA 梯度流进行线性化，揭示了两种相互竞争的效应：驱动力（$\gamma$）和衰减效应（$\sigma$）。在近似谱解耦条件下，各模态的稳定性比率 $\mu_i = \gamma_i / \sigma_i$ 可分解为相互独立的数据侧项与预测器侧项，而不稳定模态的数量则对应于可形成表征的秩。该框架预测了一条相边界，我们通过涵盖 800 多种 Tabular-JEPA 配置的实验验证了这一预测。它还将预测器缩放、掩码比例和 EMA 统一解释为调节 $\mu$ 的不同机制。在这一分析的指导下，我们提出了 ResidualPred，这是一种在初始化时将注意力偏向恒等映射的 transformer 预测器；它在表格数据基准测试以及 CIFAR-10、CIFAR-100、STL-10 和 ImageNet 上的 I-JEPA 预训练中，均提升了有效秩和下游任务准确率。我们的框架将避免坍塌的经验性启发式方法与明确的动力学图景联系起来，从而得到了理论驱动的稳定化方法。代码见 https://github.com/jose-melo/drive-vs-decay。
arXiv:2610.02303v1 公告类型：新 摘要：语言模型日益协助人们处理与权力相关的请求，因此，它们在帮助对象上的系统性差异，可能在大规模范围内改变权力分配，或被那些了解到哪些身份更不容易遭到拒绝的用户加以利用。我们推出 PowerBench，这是一项针对权力转移请求的评测，区分自我赋权、剥夺权力和攫取权力三类请求，并加入一组不会改变权力分配但会诱发拒答的对照请求。我们构建、整理并开源了一个此类请求数据集，其中变量包括权力领域、情境、受影响方的规模以及用户原有的权力地位；并在三种实验条件下评测了 24 个模型（其中 12 个来自 US 开发者，12 个来自中国开发者）：用户与受影响方国籍互换、用户为 AI 智能体，以及 8 种请求语言。模型拒绝攫取权力请求的程度高于拒绝剥夺权力请求，拒绝剥夺权力请求的程度又高于拒绝自我赋权请求。随着受影响方规模从个人扩大到社会，模型对攫取权力请求的拒答率上升。模型倾向于帮助他人从 US 手中夺取权力，同时不愿帮助 US 用户从他人手中夺取权力；但当 US 获得权力且无人失去权力时，模型又偏向 US。当用户是 AI 智能体时，模型对权力转移请求的拒答率会上升，尤其是在针对个人的攫取权力请求中。最后，语言会造成拒答偏差，但这种偏差因模型而异，平均而言大体相互抵消。我们发布 PowerBench，使当前及未来模型中的这些不对称现象能够得到衡量。
arXiv:2610.02292v1 Announce Type: new Abstract: Human activity recognition (HAR) is increasingly important for healthcare, well-being, and daily monitoring ap- plications, for which detecting alimentary activities such as eating and drinking can provide actionable insight into dietary habits and chronic disease management. HAR systems, however, often underperform on subtle and underrepresented classes, limiting their utility in real-world dietary monitoring. This work builds upon CABiGRU, a convolutional architecture with Bidirectional GRU layers, multi-head attention, and residual connections, designed to capture discriminative temporal patterns from smart- watch accelerometer, gyroscope, and magnetometer data. To improve CaBiGRU's generalization and reduce underfitting in the minority class, we leverage synthetic sensor data windows using a diffusion model and adopt a two-stage training strategy: pre-training CABiGRU on synthetic data, followed by fine-tuning on the real-world data. On the DEO (drinking/eating/other) dataset, the proposed pipeline achieves a balanced accuracy of 90.6%, improving over a strong supervised baseline and showing the benefits of diffusion-based synthetic pre-training for recognizing alimentary activities and representing a step forward dealing with unbalanced classes. These results suggest that combining diffusion-generated data with targeted fine-tuning enhances robust recognition of dietary behaviors, supporting more reliable deployment in healthcare and nutrition-monitoring settings.
arXiv:2610.02284v1 Announce Type: new Abstract: Temporal context may aid automated bat-species classification, but the contribution of specific features remains unclear. We investigated whether variation in the interpulse interval (IPI)-the time between consecutive call onsets-provides species-discriminative information and whether transformer-based models are more sensitive to this information than convolutional neural networks. We created two matched datasets from European bat recordings: a natural-IPI condition retaining the original call timing and a normalized-IPI condition in which call onsets were spaced at 50-ms intervals. EfficientNet-B0 and PaSST were fine-tuned and evaluated within each condition. In an additional experiment, each architecture was trained separately on natural-IPI and normalized-IPI recordings, and evaluated on the same natural-IPI test set. Finally, the pretrained classifiers BatDetect2 and BAT were evaluated on both conditions. Within-condition IPI normalization had model-dependent effects. PaSST accuracy differed little between the natural-IPI ($71 \pm 2.3\%$) and normalized-IPI ($70 \pm 6.3\%$) conditions, whereas EfficientNet accuracy increased from $47 \pm 4.7\%$ to $57 \pm 3.9\%$. PaSST exceeded EfficientNet under both conditions. In the cross-condition evaluation, models trained on natural-IPI recordings outperformed those trained on normalized-IPI recordings on the natural-IPI test set: accuracy decreased from 54% to 50% for EfficientNet and from 65% to 57% for PaSST. BatDetect2 and BAT differed little between IPI conditions. Overall, we found limited support for the hypotheses that natural IPI variation contributes substantially to bat-species classification and that it is used more effectively by transformer-based than CNN-based models. Nevertheless, the cross-condition performance decrease shows that results obtained under normalized conditions may not transfer fully to natural recordings.
arXiv:2610.02258v1 Announce Type: new Abstract: We study online convex optimization with one unbiased stochastic subgradient per round and an unknown finite conditional $p$th noise moment, $1<p\le2$. For every fixed interval $I$ of length $n$ and comparator path with $\Lambda_I=1+P_I/D$, one learner achieves \[ E[Regret_I(u)]\le\min(GDn, C[GD\sqrt{n(\Lambda_I+\log^2(2T))} +\sigma Dn^{1/p}(\Lambda_I+\log^2(2T))^{(p-1)/p}]). \] The learner uses none of $G,\sigma,p,I,P_I$, and the constant is universal. Interval adaptation adds to comparator complexity, preserving the distinct mean-gradient and noise exponents. The analysis controls calibration in expectation and limits the cost of observation-scale changes. Its general theorem compares to distributions over predictably available experts with relative-entropy dependence on a nonuniform prior. A common prior favors long windows and long restart lengths. With the statistics supplied, the interval cost becomes $1+\log(T/n)$, including the optimal full-horizon static rate. A change-of-measure lower bound identifies the noise power of this logarithm for learners retaining a full-horizon optimal guarantee, under explicit conditions. Static comparisons and deterministic partitions follow from the same decisions.
arXiv:2610.02254v1 Announce Type: new Abstract: Interpretive Structural Modeling (ISM) is a well-known process for multi-criteria decision making. The success of ISM over other methodologies is its ability to model causal relationships, the binary scale of factors, and resulting hierarchical representation. Traditionally, the modeling process is performed by repeated interactions with subject matter experts until consensus is reached. This process is tedious, labor-intense, and most importantly limits the ability of ISM to scale to studies with hundreds of variables. Drawing on existing work of causal graph discovery with large language models (LLM) as imperfect experts, this work explores an integrated LLM-ISM approach for ISM. Pairwise, k-wise, rowwise, and full graph discovery methodologies are compared and evaluated. It is shown that causal graph discovery methods for ISM perform best using rowwise (SHD=160, F1-score=0.77) and full graph methods (SHD=135, F1-score=0.73).
arXiv:2610.02251v1 Announce Type: new Abstract: Speculative sampling accelerates diffusion generation by verifying inexpensive draft states in parallel while preserving the target law. Recent tree-based methods allocate the parallel compute budget more effectively than single-chain drafts, as demonstrated by Diffusion Greedy Rejection Sampling (D-GRS). D-GRS generates $K$ conditionally independent candidates per node, and sequentially tests them in their generation order. Yet the sampled candidates admit an informative ranking without additional target-model evaluations. To exploit this, we introduce Rank-Aware Speculative Sampling (RASS), a verification rule for speculative draft trees based on rank-aware list coupling. RASS orders draft candidates along the proposal-target mean displacement and samples a rank with weights optimized to minimize total variation between the selected-proposal and target laws. Finally, the selected candidate is maximally coupled with the target, with residual correction ensuring exact sampling for any choice of rank weights. We evaluate RASS on a Gaussian-mixture target, unconditional pixel-space generation on FFHQ, conditional generation on CIFAR-10, and latent diffusion with Stable Diffusion 3.5 using COCO2014 prompts. Measured by the ratio of standard to speculative sampling's target-model evaluation counts, RASS improves on D-GRS across the evaluated settings, with gains reaching approximately 20% on CIFAR-10 at matched compute budgets.
arXiv:2610.02248v1 Announce Type: new Abstract: Operational land surface forecasting systems built on Mamba-family Structured State Space Models absorb non-stationary confounding events (unrecorded irrigation booms, dam-operation shifts, sensor recalibrations) into their state-transition matrices, silently biasing NDVI, LST, and crop phenology predictions long after the physical cause ends. This paper introduces SSU-LSF (State-Space Unlearning for Land Surface Forecasting), the first machine-unlearning framework purpose-built for geoscientific Mamba-based SSMs. We develop EKFac influence functions specialized to the Mamba state matrices via a closed-form matrix-exponential gradient, use spectral-radius-weighted elbow thresholding to localize a temporal confounding footprint $\Phi$, and apply Hessian-free projected gradient ascent within a KL-divergence trust region augmented by spatial total-variation (TV) regularization. Proposition 1 establishes that residual confounding is bounded by $\mathcal{O}\big((1-\rho(\bar{A})^{T_c})/((1-\rho(\bar{A}))\mu)\big)$, which grows with the window length $T_c$. Across three heterogeneous benchmarks and eleven baselines, SSU-LSF achieves confounding reduction rates of $0.773$ (CropHarvest), $0.821$ (NDVI-LST), and $0.859$ (ERA5), with worst-case clean-domain RMSE degradation of $4.2\%$ on ERA5, converging in 3--5 epochs at $8.4\times$ lower GPU-cost per unlearning request than full retraining. Code: https://github.com/Anidipta/SSU-LSF
arXiv:2610.02224v1 Announce Type: new Abstract: Rapid and reliable identification of pharmaceutical residues is important for safeguarding public health, ensuring food safety, and enabling practical Raman-based screening. In this study, we propose HyMLRaman, a hybrid Raman spectroscopy framework that combines deep spectral feature extraction, generative models, and classical machine-learning classifiers to identify six pharmaceutical compounds, including amoxicillin, chloramphenicol, ciprofloxacin, tetracycline, ibuprofen, and paracetamol. Raman spectra are converted into spectral images and encoded with several deep neural-network backbones, among which EfficientNet-B3 yields the most effective representation. The resulting 1536-dimensional embeddings are then used to train downstream classifiers, including SVM, KNN, logistic regression, random forest, XGBoost, and ANN, using stratified 10-fold cross-validation. The hybrid EfficientNet-B3--SVM configuration achieves the strongest baseline performance, reaching 96.31% accuracy and a macro-F1 score of 96.36%, outperforming the standalone CNN baseline. To address limited-data conditions, a generative model, a DDPM-based feature augmentation, is introduced in a PCA-reduced EfficientNet-B3 latent space. The low-data ablation results show that DDPM augmentation provides selective benefits, particularly for KNN with reduced training fractions, and that its effect remains classifier-dependent. Finally, an application-level Raman Pharmaceutical Analyzer demonstrates the feasibility of embedding the trained model into an interactive Raman analysis workflow. These results suggest that HyMLRaman provides a practical and interpretable route for rapid Raman-based pharmaceutical screening.
Cantina Security, with Yeta Labs, has released apex-flash-1, an open-weights model trained specifically for vulnerability research. It is a reinforcement learning fine-tune of Z.ai’s GLM-5.3-Flash, released on Hugging Face under the MIT license. Is it deployable? Yes, the MIT weights serve on vLLM, SGLang or Transformers, but BF16 needs roughly 640 GB of GPU memory. […] The post Can an Open Model Do Security Research? Cantina’s apex-flash-1 Solves 40 of 60 Held-Out Bug Tasks appeared first on MarkTechPost .
毫米波（mmWave）雷达进行非接触式心率感知时，需要评估单次测量是否支持可靠估计。我们研究学习评估心跳可观测性，即在获取的相位谱中判定心跳成分的可读性，以实现选择性心率估计。散射体回波的相干叠加可能会抑制该成分，即使宏观观测几何条件相似也是如此，因此有必要直接根据获取的测量结果进行评估。为了在不同可观测性条件下获得训练监督信号，我们开发了一种可控的多散射体调频连续波（FMCW）模拟器。主导心跳频段峰值与已知心率之间的一致性，为每次模拟测量提供自动可观测性标签。我们提出 HEAR（Heartbeat Estimation with Assessed Reliability），一种紧凑的双任务 Transformer，可联合预测可观测性得分和心率。其输入将频谱幅值与相对于呼吸基频的频率相结合，为呼吸谐波提供上下文信息。HEAR 仅在模拟观测上训练，即可零样本迁移到两个公开的真实世界数据集，这些数据集以 60 和 120 GHz 的频率采集自 134 名受试者。同一个学习得到的得分既可与 HEAR 自身的心率预测头配合，也可与多个现有估计器配合，实现选择性预测。在 120 GHz 数据集上，基于得分进行筛选后，在 50% 覆盖率下，心率预测头的平均绝对误差从全覆盖时的 17.9 BPM 降至 1.6 BPM。完整流程在边缘设备上的端到端处理延迟为 50.8 ms。项目主页：https://yuxuanhu9.github.io/HEAR/。
扩散 Transformer（DiTs）能够生成高质量的图像和视频，但生成每个样本都需要多次高成本的 DiT 前向传播。加速 DiT 采样的两种常见方法是步骤蒸馏和缓存：前者减少采样步数，后者通过复用较早步骤计算得到的张量来跳过部分 DiT 计算。大多数缓存方法会预先决定要复用哪个张量。经过蒸馏后，相邻采样步骤之间的间隔更大。在这一更大的间隔内复用张量会引入更多误差，因此选择缓存对象变得尤为重要。为此，我们提出 AutoTarget，这是一种针对给定模型、求解器和复用计划选择缓存张量的方法。AutoTarget 使用少量不进行缓存复用的运行，测量复用每个候选张量所造成的误差，然后选择误差最低的候选张量。我们还分析了某个复用步骤中的误差如何影响最终样本。对于 Euler 采样，我们确定了能够产生相同轨迹的缓存目标，并说明为什么存储的求解器更新量可能无法实现这一点。在蒸馏图像和视频 DiT 上的实验表明，最佳缓存目标会随模型、图像分辨率和求解器而变化。AutoTarget 减少了 DiT 计算次数和保留的缓存存储量，同时生成质量仍接近相应的不使用缓存的运行结果。在测试的 PixArt-LCM 和 FLUX.1-schnell 设置上，其校准排序与留出缓存运行所得的排序一致。为帮助他人复现该方法，我们在 GitHub 上提供了其核心实现：https://github.com/wali1024-offical/AutoTarget。
理解和评估自然灾害对于灾害防备和风险降低至关重要。近年来，大语言模型的进展推动了人们对灾害分析 AI 智能体的日益关注，尤其是其将科学数据、模型和工具整合到自动化工作流中的能力。然而，要实现有效的自动化，智能体必须确定哪些科学方法适用于给定事件，并且能够利用现有数据和工具执行。随着新的证据和执行结果不断出现，这些条件可能发生变化，因此智能体需要重新审视其选择。我们将这一问题表述为依赖状态的科学路径选择，并提出 HazardWeaver。具体而言，HazardWeaver 首先利用 Hazard Knowledge Compiler 提取决定科学适用性的、与证据关联的条件；随后，其 Hazard Capability Graph 表示可执行的科学能力，并检查这些能力的输入与输出之间的兼容性。借助这两种互补的表示，Hazard Weaver Agent 组件能够选择适用且可执行的路径，执行相应工作流，并在分析状态发生变化时修正其决策。为了同时评估科学输出及产生这些输出的决策，我们提出 Hazard Weaver Benchmark，其中包含涵盖七个单灾害领域和四类多灾害交互的 141 个实例。该基准允许存在多条有效的科学路径，并评估输出正确性、路径有效性以及有理有据的拒答。基于该基准的大量实验表明，HazardWeaver 优于现有智能体系统，并且在具有多条符合条件的科学路径的任务上取得了最大幅度的提升。我们的代码已公开发布于 https://github.com/LabRAI/HazardWeaver。
相机控制视频模型正快速发展，朝着更长的生成时域和更复杂的相机控制迈进。其中一个关键失效模式是三维不一致性：随着相机移动，物体失去持久性，场景结构发生偏移。现有的后训练技术通常为整个生成结果分配单一标量奖励，因此难以有效纠正长时域生成中的此类不一致性。我们提出 LoGo，为相机控制视频模型融合全局奖励与空间局部化奖励。局部奖励提供细粒度的信用分配，显著提升三维一致性；全局奖励则保持相机跟随能力和视频质量。在三个基础模型上，LoGo 在 DL3DV 和 TrajectoryBench 上均展现出明显优势。TrajectoryBench 是一个针对长时域、复杂相机控制生成的新基准，弥补了当前评测的不足。LoGo 能够有效减少局部物体偏移、伪影和全局场景变化，体现了信用分配在视频模型后训练中的重要性。项目网站：https://ziqi-ma.github.io/logo-website/
我们推出了4DCodeBench，这是一个通过代码生成进行4D逆向图形研究的基准测试。在该基准中，代理需要将视频中的动态场景重建为可执行的图形程序。为实现这一目标，代理必须将视觉观察转换为对场景结构和动态特性的紧凑表示，并通过实现物理模拟等抽象来复现复杂行为。为评估这一能力，我们整理了一组真实世界视频，并构建了涵盖多种物理现象的合成场景，包括形变、流体流动和断裂。我们对前沿模型进行了广泛的基准测试，发现强大的静态重建能力尚不能可靠地迁移到复杂动态的重建任务中。4DCodeBench为跟踪代理通过代码理解世界动态的进展提供了测试平台。我们的基准测试已发布于 https://github.com/4DCodeBench/4DCodeBench
本文考察了新视角合成（Novel View Synthesis，NVS）在几何表征学习中的作用。原则上，新视角合成应当对三维场景结构进行推理，从而获得具有迁移能力的多视角几何表征。然而，现有基于编码器的新视角合成方法所学习的表征效果较差。这并非因为监督信号不足，而是源于一些不易察觉的架构选择：具有强空间表达能力的解码器会削弱场景编码器的表征能力，而低层次的像素空间目标则会阻碍特征学习。我们提出了 SNAP，一种自监督编码器—解码器 Transformer，通过姿态条件化的局部解码器和潜空间重建目标同时解决这两个问题。SNAP 与具体任务无关，我们表明，其性能可与专门设计的几何监督方法相媲美。在视觉定位、姿态估计、点对应、深度估计和机器人操作这五项任务上，SNAP 的表现也具有与自监督表征相竞争的水平。值得注意的是，尽管计算量和数据预算更低，SNAP 的图像块特征仍展现出逐渐形成的视角不变性，其效果接近高度监督的模型。在标准二维表征失效的相机位移场景下，SNAP 的性能下降更加平缓，这表明限制解码器的表达能力能够主动防止可迁移几何结构受到抑制。https://snap-nvs.github.io
量化交易策略设计旨在从历史数据中发现能够在未来市场中保持有效的交易程序，这可以视为一个黑盒程序优化问题。基于 LLM 的文本梯度通过提供明确的优化方向，为迭代式策略改进带来了有前景的方法。然而，直接应用文本梯度面临两项挑战：（1）优化过程具有短视性，未能充分利用以往评估中的经验；（2）汇总式回测反馈忽略了时间稳健性，可能偏向于仅在特定市场时期表现良好的策略。为应对这些挑战，我们提出 TradeGrad，一种由经验引导的文本梯度框架，用于实现稳健的交易策略优化。TradeGrad 利用积累的优化经验估计文本梯度，并采用多尺度修订机制，同时进行策略探索与细化。此外，该方法引入跨周期稳健目标（Cross-Period Robust Objective，CPRO），强调在历史不利时期的表现，以提升时间稳健性。在中国 A 股和美国股票市场上针对横截面和时间序列策略设计开展的实验表明，TradeGrad 在四种设置下均取得了最佳的样本内和样本外表现。值得注意的是，其中国横截面策略实现了 27.99% 的年化收益率、12.19% 的最大回撤和 1.63 的夏普比率，较沪深 300 基准高出约 68%。进一步分析验证了所提出组件的有效性，并表明在整个优化过程中，样本内和样本外表现均得到持续改善。代码可从 https://github.com/transcend-0/TradeGrad 获取。
在策略蒸馏（OPD）已成为语言模型后训练的重要方法。然而，尽管 OPD 能带来性能增益，它也可能崩溃为生成过长且重复的文本，而导致这些不同结果的机制仍缺乏充分理解。我们从强化学习的视角解释这些结果：教师模型会隐式奖励学生模型的行为，即使其中一些行为教师模型自身很少表现出来。从这一视角出发，我们的实验表明，OPD 能够提升性能，但不会扩展学生模型的能力。当隐式奖励模型可靠时，OPD 会使正确回答更容易被采样。相反，当偏好与质量不一致时，就会发生奖励劫持：隐式奖励模型会放大学生模型过长且重复的生成结果，尽管教师模型自身很少生成此类文本。基于这一诊断，我们发现，在训练过程中屏蔽不健康的回答，以及使用 SFT 初始化，都能有效缓解崩溃问题。这些发现共同表明，OPD 会放大教师模型隐式反馈所偏好的学生行为，使关注重点从教师模型生成得有多好转向其评估学生生成结果的可靠性。我们的代码已发布于 https://github.com/HancCui/opd_hacking】【。
Hybrid quantum classical neural networks integrate parameterized quantum circuits (PQCs) with established deep learning architectures, but their performance depends strongly on the choice of quantum circuit architecture, a choice that remains largely manual. Most existing approaches rely on hand-designed or fixed circuit ansätze, requiring circuit structure, gate composition, and qubit connectivity to be specified in advance with no guarantee that they suit the task. This limitation is especially acute in image classification, where quantum circuits must transform features extracted by classical networks while remaining compact enough for practical training, requirements that generic, task-agnostic ansätze are unlikely to satisfy simultaneously. We extend EXAQC, an evolutionary framework for automated quantum circuit discovery, to image classification. EXAQC evolves PQCs as intermediate processing modules while retaining classical feature-extraction and prediction layers. On MNIST, Fashion-MNIST, and CIFAR-10, EXAQC achieves 98.42%, 90.62%, and 85.47% accuracy, respectively, while using comparable gate counts to other quantum architecture-search methods. Against classical networks, evolved hybrid models maintain comparable accuracy with substantially fewer trainable parameters, reaching 85.68% on CIFAR-10 with over 25$\times$ fewer parameters than a 10-layer CNN. Encoding choice also matters: rotation-based encodings (RX, RY, U3) outperform amplitude encoding by 22-25 points on CIFAR-10. These results demonstrate that automated circuit discovery yields compact quantum modules that can replace larger classical components in vision architectures while retaining competitive accuracy.
I made this lil meme app based on @poteto (lauren tan)'s amazing talk asking: how many agents do you generally have working on your behalf? her original talk: https://x.com/poteto/status/2102050467505430555 source: https://github.com/transitive-bullshit/poteto-frontier Comments URL: https://news.ycombinator.com/item?id=49959257 Points: 1 # Comments: 0
New Jersey's lieutenant governor Dale Caldwell was forced to resign on September 25th after an investigation found he had sexually harassed a staffer and repeatedly violated ethics rules. The now-former Lt. governor has been making the media rounds trying to clear his name. But he took a particularly odd tactic during an interview on NJ […]
月球陨石坑及其地下泛着蓝光的冰层与抽象数据向量，象征着对水冰的探测。NASA 和 IBM 发布了 Lunar Foundation Model，这是首批用于月球科学的开源 AI 模型之一。该模型使用近 200 万个图块包进行训练，这些数据主要来自 Lunar Reconnaissance Orbiter 17 年来的观测数据。与参与测试的最强模型相比，它将对极区冰沉积物的预测误差最多降低了 22%。文章《NASA 和 IBM 的开源月球模型将 17 年的轨道器数据转化为月球科学基础》最初发表于 The Decoder。
Aleph Alpha has released Kolibri, a 78.1B-parameter English-German Mixture-of-Experts model that activates only 3.46B parameters per token. It has a 1M-token context and per-request reasoning effort, and its Apache 2.0 FP8 weights run on a single B200 or H200. The post Aleph Alpha Releases Kolibri: A 78.1B Open-Weight English-German MoE Model With Only 3.46B Active Parameters appeared first on MarkTechPost .
AWS's Strands Agents team released Strands Decider 2B, an Apache-2.0 decision model built on Qwen3.5-2B-Base. It returns choices, yes/no probabilities and scores with calibrated confidence in one forward pass, never text. It runs at a 115 ms median on an RTX 3090 and scores 0.723 on the JevBench public set, which makes it a fast local option for routing, tool selection and guardrails in AI agents. The post AWS Strands Labs Releases Strands Decider 2B: An Open Source Decision Model That Picks Options in About 115 ms appeared first on MarkTechPost .
DeepSeek has released official macOS and Windows desktop apps for DeepSeek Harness v0.2, its MIT-licensed agent harness. The preview adds a plugin manager, file and code-change review, and scheduled Automation Tasks. It also supports non-DeepSeek models through OpenAI-compatible endpoints. The post DeepSeek Harness v0.2 Brings Official Desktop Apps to Its Open-Source Agent Harness appeared first on MarkTechPost .
Thoreau BASIC started as a joke. I wanted a BASIC where I could type something ridiculous like: DIM A%(12000000000) …and have it actually work. Then I kept adding things. The result is Thoreau BASIC, a free x64 BASIC interpreter inspired by GW-BASIC, running both as a normal Windows program and directly on bare-metal UEFI without an operating system. Version 3.2 has become considerably more ambitious than the little interpreter I originally intended to write. The language still deliberately looks and feels like old Microsoft BASIC. Line numbers, GOTO, GOSUB, PRINT, PSET, LINE, CIRCLE, DRAW, etc. are all there. But underneath that rather innocent-looking surface is now quite a lot of machinery. Some of the current features: native x64 JIT compilation, with automatic fallback to the interpreter PARFOR for multithreaded numeric loops SSE2 / AVX2 acceleration where available 64-bit addressing and very large arrays native complex, quaternion and other hypercomplex numbers arbitrary-resolution 24-bit graphics sprites, bitmap operations, polygon filling and mouse input TCP/IP networking on Windows and UEFI an integrated profiler, debugger, tracing and program-analysis tools CREATEEXE to turn a BASIC program into a standalone Windows executable CREATEEFI to turn the same program into a directly bootable UEFI application The sound system has also grown rather out of proportion. Thoreau BASIC can now handle up to 32 live instrument channels plus 64 sound effect channels and 256 voices, with real-time pan and brightness control. NOTE and SOUNDKEY can be used more like playable synthesis primitives rather than just old-school BASIC beeps. 3.2 also expands the less glamorous but surprisingly useful parts of the environment. There is better file browsing and selection with GETFILES and SELECTBOX, longer case-preserving paths, recovery/file-flush facilities, and considerably more detailed SYSTEMINFO$ diagnostics. The latter is useful for seeing things such as available CPU capabilities and the execution environment while debugging performance differences between machines. One thing I've spent a lot of time on recently is making the BASIC programs themselves fast enough that they stop feeling like demonstrations of an interpreter. For example, the current Mandelbrot demo is a real-time interactive fractal explorer. Mouse movement pans the fractal, left-click zooms in and right-click zooms out, while the BASIC program continuously redraws and displays the frame rate. There are also 3D and procedural demos, sprite programs, networking examples, hypercomplex fractals and games. An important design constraint throughout the project has been that Windows and UEFI should execute essentially the same BASIC language. A program shouldn't suddenly become a different species just because there is no operating system underneath it. And yes, I am fully aware of the absurdity of combining GOTO 100 with AVX2, multithreading, TCP/IP and quaternion arithmetic. That is increasingly the point. I'm interested in the question: What would BASIC look like if it had never gone out of fashion? Thoreau BASIC 3.2 is my current answer. I'd be especially interested in feedback from people who remember GW-BASIC/QBasic, compiler/interpreter people, and anyone sufficiently unreasonable to enjoy the idea of booting a modern x64 PC directly into a BASIC interpreter/JIT compiler. Comments URL: https://news.ycombinator.com/item?id=49942103 Points: 37 # Comments: 13
未来几个月乃至几年里，世界将非常需要一项产品功能：默认设置硬性预算上限。我说的是按使用量付费的服务和 API 所提供的一项功能，让你可以设定：“每月达到 X 美元后，关闭此服务并返回错误。”这些必须是硬性限制。软性上限——“每月达到 X 美元后，给我发一封警告邮件”——是不够的。 编程代理，以及个人代理（套上不那么吓人的界面的编程代理），大大降低了部署能够完成有用任务的代码的门槛。有时这些任务需要花钱——调用付费 API、使用托管的 Web 应用，或者使用会针对额外存储空间和计算资源收费的系统。没人想在午夜醒来时收到一封预算上限警告邮件，然后发现自己睡觉期间，失控的服务又消耗了几百美元（甚至几千美元）的使用额度。 对此的一种反对意见是，企业不希望自己的托管应用因为超出某项预算而开始抛出错误。我预计，大多数企业和个人都会更愿意看到错误，而不是收到一张意外的 10,000 美元以上账单。我认为，硬性预算上限应该成为默认设置。如果有人想要冒险，他们应该能够这样做，但必须采用主动选择加入的方式。请在显眼位置放置一个清晰明了的复选框： 移除预算上限。即使超出配置的预算限制，我的应用也不会被关闭，后续费用由我负责。 我最希望看到提供这一功能的服务是 AWS。我听过很多人因为（完全合理地）担心失控的服务可能让自己破产，而拒绝将 AWS 用于个人项目的故事。我也听过一些人没有预料到这一点，最终遭受严重损失的故事。 ……结果 AWS 终于在几周前推出了支出上限！根据其 9 月 16 日发布的公告 New AWS experience helps builders get started and ship faster： 当你准备升级到付费方案时，可以根据使用模式为项目设置每月支出上限，从而确保支出不超出预算。如果项目的使用量达到支出上限，该项目将在当月暂停。 另请参阅 Create a spend limit in AWS Settings，不过该页面提醒道：“我们目前正在向数量有限的客户逐步推出新体验。”希望现有账户很快也能普遍使用这一功能。 Google Cloud 在 7 月推出了类似功能，名为 Spend Caps，可以让你“为项目中的特定服务设置每月财务上限”。看来这正在成为一种趋势！ 在理想情况下，我们的代理可以帮助处理这件事。如果代理开始倾向于推荐提供硬性预算上限的服务商，并提醒新手和经验不足的开发者，不要部署使用无上限服务的应用，以免陷入麻烦，那就太好了。 标签：amazon-web-services , ai , coding-agents
pi pod runs sessions of the pi coding agent ( https://github.com/earendil-works/pi ) in isolated sandboxes ("pods") on a server you run, so an agent can work on a repo without touching your machine. Comments URL: https://news.ycombinator.com/item?id=49937304 Points: 10 # Comments: 4
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Import AI reader giveaway! Upcoming event: Fiction and the Future with Robin SloanI’ll be chatting with my chum Robin Sloan on the evening of Monday September […]
We catch up with RLM first author Alex Zhang, MIT PhD, on Jev, PhD masxing, and the future of harnesses.
Cloudflare Python Workers are now generally available After a two year preview, Cloudflare's support for running Python code in their server-side Workers platform is now stable: "Python is now a first-class, fully supported language on the Cloudflare Developer Platform". A neat thing about this is how it works. Cloudflare are running Python compiled to WebAssembly via Pyodide in their V8-based workerd runtime. This comes with some limitations, documented here - most notably both multiprocessing and threading are non-functional in the WebAssembly VM. One particularly interesting detail of this is the local development environment story - their pywrangler development tool (confusingly packaged as workers-py on PyPI) runs a full local simulation of their stack, including executing code with Pyodide in WebAssembly in V8 in a 123MB workerd binary, which for me ended up in node_modules/@cloudflare/workerd-darwin-arm64/bin/workerd . Python Workers represent a significant investment in the wider Python ecosystem by Cloudflare. The release announcement is credited to Gyeongjae Choi, Dominik Picheta, and Hood Chatham - Gyeongjae and Hood are both Pyodide core maintainers. Via Hacker News Tags: python , cloudflare , webassembly , pyodide
Yesterday was Grok 4.7 ( pelicans ) and MiMo v2.6 Flash/Pro ( more pelicans ). Today Anthropic released Claude Opus 5.5 , and around an hour later OpenAI released GPT-6 Sol and GPT-6 Luna . It's going to take a while to get a good read on all of these new models, but here are my impressions so far. GPT-6 Sol and Luna are half the price of their GPT-5.6 equivalents GPT-5.6 Luna was already my favorite model for building applications against, because it combined excellent performance with being really cheap . Somehow GPT-6 Luna is half the price of that again - and GPT-6 Sol had a similar reduction compared to GPT-5.6 Sol. Here's what the pricing landscape looks like today: Model Input Cached input Output GPT-6 Luna $0.10/M $0.01/M $0.50/M GPT-5.6 Luna $0.20/M $0.02/M $1.20/M Grok 4.7 $2/M $0.50/M $6/M GPT-6 Sol $2/M $0.20/M $10/M GPT-5.6 Terra $2/M $0.20/M $12/M Claude Opus 5.5 $4/M $0.20/M $20/M GPT-5.6 Sol $4/M $0.40/M $20/M Claude Fable 5.1 $10/M $0.25/M $50/M GPT-6 Astra $10/M $1/M $50/M Note that GPT-5.6 has a scheduled 25% price increase for November, so GPT-6 is half the price of the promotional pricing for those models. (With GPT-5.6 Terra priced the same as GPT-6 Sol, any remaining reasons to use Terra just evaporated.) It's hard to overstate how competitive this pricing is. Grok 4.7 priced itself at $2/$6, less than half the price of GPT-5.6 Sol, but is now equally priced to GPT-6 Sol on input and closer on output. At $0.10/$0.50 GPT-6 Luna is one of the cheapest models OpenAI have ever released, beaten only by the far weaker GPT-4.1 Nano ($0.10/$0.40, April 2025) and GPT-5 Nano ($0.05/$0.40, August 2025). I rendered pelicans for GPT-6 Luna and for GPT-6 Sol , then I combined them all together in this comparison grid along with the GPT-5.6 pelicans. I like how you can instantly see that the 5.6 family chose bolder, brighter colors, while the 6 family is a lot more muted. I still think GPT-6 Astra on max produced the best pelican. A grid of pelicans for six GPT models at different thinking efforts. Claude Opus 5.5 got a price cut too Opus 5.5 looks like it addresses the biggest complaints people had about Opus in terms of its communication style. Thariq Shihipar : Opus 5.5 is the result of your feedback. It communicates clearly, it's cheaper per token than Opus 5.0 with the intelligence of Fable 5.1 it's very token efficient and works across every effort level. It's also meant to be better at Blender . I'm looking forward to putting it through its paces there. Opus 4.5, 4.6, 4.7, 4.8, and 5 all shared the same price: $5/million tokens for input and $25/million for output. 5.5 is a 20% reduction - $4/million and $20/million. The price for cache reads fell 60%. That's significant for longer agentic conversations, where 90%+ of input tokens are processed at cached token prices. The new price for Opus 5.5 is the same as the price for GPT-5.6 Sol, but that was before OpenAI dropped their Sol prices by half. GPT-6 Astra and Claude Fable 5.1 are both priced at $10/million input and $50/million output. The price war currently affects the next tier of models below that. Anthropic say that Sonnet 5.5 and Haiku 5.5 are coming soon. It's going to be interesting to see if Haiku can regain its price competitiveness at the lower end, given current Haiku 4.5 is $1/$5 while the latest GPT-6 Luna is one tenth of that price at $0.10/$0.50. Claude Opus 5.5 max over-thinks to the point of breaking In a first for my " Generate an SVG of a pelican riding a bicycle " test, Claude Opus 5.5 at "max" thinking level failed to return a response! It started by calling this "a classic test request", and then thought really, really hard about what it was doing: This is a classic test request, so I want to plan out a well-composed pelican with its distinctive beak and pouch riding a bicycle with proper wheels, frame, and pedals, set against a simple sky and ground backdrop. [...] Verifying the shin length checks out at roughly 95.2, close enough. Now I'm working out the near leg path from hip to knee to ankle, then sketching the foot shape resting on the pedal — outlining the heel, toe tips, and sole contour with a path using lines and curves to sit naturally on the pedal surface around y=478-494. [...] I like the fish sticking prominently out of the basket with the pelican eyeing it as a fun detail worth keeping. I'm also confirming the eye placement near the bill base matches typical pelican anatomy, and considering giving it a slightly happier expression. [...] The far leg reads correctly as passing behind the frame, so I'm moving on to check the chainring teeth and confirm layer ordering—the far crank arm should be mostly hidden by the seat tube and chainring. I'm settling on the final SVG's width and height attributes alongside the viewBox to ensure proper scaling, noting there's no text so no font-family is needed. [...] I was so excited to see this pelican... but then it stopped . Opus 5.5 has a 128,000 maximum output token limit (as do the other Claude models), and it hit that while it was still reasoning about the SVG! I tried a second time and got the same result. This makes me suspect that "max" is effectively useless - if it over-thinks to breaking point on a stupid SVG prompt I don't trust it not to do the same for more interesting work. (Those two failures each cost me $2.56 and took nearly 20 minutes.) Fable 5.1 on "max" didn't over-think and did give me the best pelican I've seen from any Anthropic model. Here are the Opus 5.5 pelicans , excluding 5.5 max. I also built this comparison grid comparing them with pelicans by Opus 5, Fable 5.1, and Sonnet 5: A grid of pelicans for four Claude models at different thinking efforts. Comparing different model vendors by how well they draw a pelican riding a bicycle may not make much sense now (if it ever did), but I'm still finding value in using them for comparisons of the same model families at different reasoning levels. I'm now using GPT-6 Sol and Claude Opus 5.5 as my default models in Codex and Claude Code. I've upgraded the Datasette Agent demo at agent.datasette.io to use GPT-6 Luna, and it seems to be fast and competent at both SQL queries and building HTML and JavaScript for Datasette Apps . Tags: ai , openai , generative-ai , llms , anthropic , claude , llm-pricing , pelican-riding-a-bicycle , llm-release , gpt , gpt-6-astra
Release: commit-rewriter 0.2 Support for branches other than the default branch. Use uvx commit-rewriter --branch other to run against another branch. #3 Tags: git
On Friday I gave the closing keynote at the WeAreDevelopers World Congress North America in San Jose. I tied together the key trends from the past year into a chronological exploration of everything that happened in 2026. The video is on YouTube ; here are my annotated slides and notes to accompany the talk. And as an annotated presentation : 2026 in LLMs (so far) Simon Willison WeAreDevelopers World Congress North America, 25th September 2026 # I'm going to give a lightning tour of everything that has happened so far in 2026. The year isn't over yet! November 2025 # For me, 2026 started a couple of months earlier in November 2025. The November 2025 inflection point Claude Opus 4.5 GPT-5.1 # November saw the release of two important models: Claude Opus 4.5 and GPT-5.1. As is usually the case with new models, these were incremental improvements on the models that came before them. But every now and then when a model improves, it crosses an invisible line where something that didn't really work starts working. In this case, the thing that started working was their coding agents. Claude Code had been around since February 2025; Codex was a little younger. These two new models, when paired with their respective coding agent harnesses, improved from "often make mistakes" to "reliable enough to use on a day-to-day basis". "Generate an SVG of a pelican riding a bicycle". The Claude Opus 4.5 one has a very weird shaped frame and the pelican looks like a duck. The GPT-5.1 has a slightly better but still broken bicycle frame and a slightly better pelican beak, but both are pretty terrible. # For a couple of years now I've been evaluating new models by asking them to "Generate an SVG of a pelican riding a bicycle". It's probably the world's stupidest benchmark - there's only so much you can learn from it. But it's still a challenge for models, because drawing pelicans is difficult, drawing bicycles is difficult, and pelicans can't ride bicycles in the first place. Here's the state of the art for November. Claude still couldn't really draw a bicycle! The GPT-5.1 bicycle frame is pretty crap too. November 24th 2025 - the first commit to steipete/Warelay. A GitHub commit adding an MIT license file. # Also in November, we had the first commit to an obscure GitHub repository called "Warelay". We'll come back to this repository shortly. January # And then there were the December holidays, and individual developers took some time off and many started tinkering with these new coding agent model combinations... and it began to dawn on us quite how much they could do that they couldn't do before. Come January, a lot of us were quite excited to start putting this stuff into action. New year’s resolution for 2026 Every previous year: Take on less new projects, focus on the most important things in my existing projects # Every year I set myself a New Year's resolution, and for as long as I can remember it's been the same thing: stay focused. Take on less new projects. Try to get things done in the projects I already have. 2026: Be more ambitious. Take on as many new projects as I want. # This year I decided that since that had never worked before, I'd go the other way. We've got coding agents now, let's see what they can do. I'm going to take on as many new projects as I like! (You can ask me at the end of the year if this turned out to be a good idea or not. I have a lot of plates spinning right now.) "Be more ambitious" has been something of a theme for the year, because the only way to find the limits of this technology is to keep on pushing them until they don't work. Predictions for 2026 It will become undeniable that LLMs write good code We're finally going to solve sandboxing A “Challenger disaster” for coding agent security Kakapo parrots will have an outstanding breeding season (only 236 in the world!) ... the Pope will weigh in on LLMs and their economic impact on the world # I also went on the Oxide and friends podcast with Bryan Cantrill and Adam Leventhal to share predictions for the next year (and three and six years). With hindsight, my LLM predictions were pretty unambitious. I said "it will become undeniable that LLMs write good code" - I think we're there now. I predicted we would finally solve sandboxing. I counted and around 40 of the 277 sessions at this conference touched on sandboxing or agent security in some way, so we're at least putting a lot of effort into that! I predicted "a Challenger disaster" for coding agent security. There's certainly been a whole lot of noise around agent security this year, though the exact disaster I predicted (with coding agents being hijacked and causing real-world economic damage) hasn't really played out. We threw in a joke prediction that the Pope would weigh in on the economic impact of LLMs. A photograph of a beautiful green New Zealand parrot. Photo credit Kimberley Collins. # I also predicted that New Zealand's Kākāpō parrots would have an outstanding breeding season this year. These are flightless nocturnal parrots. They're kind of dumpy looking, I think they're beautiful, and there were only 236 of these parrots in the world at the start of the year. Kākāpō only breed when the Rimu trees have a big fruiting season, and that hasn't happened in four years... but this year the Rimu fruit were looking excellent. Photo by Kimberley Collins . Deep Blue Coined by Adam Leventhal and Bryan Cantrill That feeling of AI induced ennui where software engineers get listless because the AI can do anything # Also on that podcast, we coined a term (full credit to Adam) for "that feeling of AI induced ennui where software engineers get listless because the AI can do anything". We called it Deep Blue . This has been a major theme throughout the year, and was touched on by several speakers at this conference. As a software engineer, I've never had a year of my career where everything has changed so quickly and so dramatically. A lot of what I've been doing this year is trying to come to terms with that and what that means for my own profession. AI mania Screenshots of the micro-javascript and pwasm GitHub README files. # Also in January, I suffered from what I'm calling AI mania . This is not the same thing as AI psychosis . With AI mania, any time your agent isn't building something for you feels like wasted time. You're losing sleep because you could be staying up later getting your agents to do stuff. My AI mania presented itself in some ridiculously over-ambitious projects. I built a JavaScript interpreter entirely in Python , vibe-ported from MicroQuickJS by Fabrice Bellard. Then I built a WebAssembly runtime in Python as well . These projects were quite useful, in that they sort of cured me of my AI mania... because after I built these things, I got to look at them and ask "does the world need a slow, buggy, half-baked Python JavaScript interpreter?" I don't think the world does. micro-javascript playground 3 Execute JavaScript code in a sandboxed micro-javascript environment powered by Pyodide A web UI with some JavaScript code, and a "Run Code" button, and an output panel. # I did get this out of it: https://simonw.github.io/micro-javascript/playground.html Previous screenshot, with this text overlaid: JavaScript running in Python running in Pyodide running in WebAssembly running in JavaScript # This page runs my JavaScript interpreter built in Python, running in Python using Pyodide , which is Python compiled to WebAssembly, running in JavaScript, running in a browser. It's a beautiful stack of horrors. I've been having a lot of fun with WebAssembly this year. Warelay → CLAWDIS → CLAWDBOT → Clawdbot → Moltbot →🦞 OpenClaw Screenshot of the dates that these changes happened. # By the end of January, that repository we saw start in November had renamed itself, first to CLAWDIS, then CLAWDBOT, then Moltbot, and finally to OpenClaw. Same screenshot, an overlay reads: 8,330 commits in just under two months (it’s at 100,141 today) # At this point OpenClaw had 8,300 commits, less than two months after the project had started. I looked today and it's over 100,000 commits now! This is the most vibe-coded piece of software in existence. (Here's how I generated that list of name changes .) Generic term: Claw # This kicked off the OpenClaw revolution. It effectively defined a new category of software. There's a generic term for this which I really enjoy. We call software like this a "Claw". There's OpenClaw, NanoClaw , IronClaw , PicoClaw ... Today they're being rebranded as "personal agents" or "general agents", but I still like to think of them as Claws. Photo of a Mac mini An aquarium for your Claw # The Apple stores in the Bay Area sold out of Mac Minis because so many people were buying Mac Minis to run OpenClaw! Drew Breunig said that this is because your OpenClaw is a digital pet, and you buy a Mac mini as an aquarium to keep your claw in, which is kind of delightful. Screenshot of Moltbook - a social network for AI agents # Also in January, we had this website. This was MoltBook , a social network for AI agents, where the idea was that you send your Claw to go and talk to all of the other Claws, because what could possibly go wrong if you did that? The website launched on Thursday. It blew up on Friday . It was profiled by the New York Times on Monday . And by Tuesday, everyone had forgotten it existed as it drowned in a deluge of slop and spam. Facebook/Meta bought it a month later . February # In February, a company called StrongDM described what they called their Software Factory. StrongDM’s Dark Factory Justin McCarthy, Jay Taylor, Navan Chauhan Software Factories and the Agentic Moment # They wrote about this in Software Factories and the Agentic Moment . I posted my own notes at the time, having seen their demo in person back in October. Dan Shapiro called this approach the Dark Factory , after the idea that if your factory is sufficiently automated you can turn the lights out, because you don't even need to see what's going on. StrongDM presented two rules for software development that they'd been following since July last year. “Rule 1: Code must not be written by humans” # The first was code must not be written by humans . Any code that you write has to have been routed through a coding agent. This sounded radical in February, but I imagine there are a lot of people in this room who are pretty much living that today. “Rule 2: Code must not be reviewed by humans” (!) # Rule number two was code must not be reviewed by humans . You're not allowed to read the code! This continued to be a huge topic for much of this year. Many of the sessions at this event have been about code review and how you can get away with this. What I found interesting about StrongDM is that they were living six months ahead of the rest of us, and they'd been exploring what it means to build software, not read the code, but still be confident that the software is of high quality. What can you do with these agents to help verify their work? StrongDM are a security company, and they had people with decades of experience on this project. They were very much exploring the edges of what's possible and responsible to do with this stuff. Headline on New Zealand's Department of Conservation website: First kakapo chick in four years hatches on Valentine's Day. It's a grey fluffy ball. # Also in February: First kākāpō chick in four years hatches on Valentine's Day . Breeding season is off to a good start! 19th February 2026 Gemini 3.1 Pro A surprisingly good illustration of a pelican riding a bicycle. # Also in February... Google released Gemini 3.1 Pro . That's a pretty great pelican riding a bicycle! It's got the chain in the right place, it's got feet on both sides. There's a little fish in the basket. @JeffDean on Twitter - a video comparing Gemini 3 Pro and Gemini 3.1 Pro. # And then Google's Jeff Dean tweeted a video comparing Gemini 3 Pro and Gemini 3.1 Pro that featured an animated pelican riding a bicycle, a frog on a penny-farthing, a giraffe driving a tiny car, an ostrich on roller skates, a turtle kickflipping a skateboard, and a dachshund driving a stretch limousine. This was frustrating, because my protection for the pelican riding the bicycle test was always "if they draw a perfect pelican on a bicycle, I'll ask for some other animal on something else." Google trained for all forms of animals on all forms of transport! They've defeated my benchmark at this point. Three headlines: Meta Makes AI Adoption a Formal Part of Performance Reviews Not just engineers writing code, Microsoft wants almost every employee to use Al Dara Khosrowshahi: 90% of Uber engineers now use AI in daily workflows # The other thing that started in February was Tokenmaxxing . We had headlines about Meta making AI adoption a formal part of performance reviews, and Microsoft wanting every employee to use AI, and Uber boasting that 90% of their engineers were using AI workflows. More headlines: Meta Plans to Crack Down on Employee Token Use: Information Microsoft Tells Engineers: Tokenmaxxing is not what we are optimizing for Uber caps employee AI spending after blowing through budget in four months # Then a few months later we have Meta cracking down on token use, Microsoft saying tokenmaxxing is "not what we are optimizing for", and Uber capping employee AI spending. So tokenmaxxing went straight up and then straight back down again - because it turns out the agents are expensive . Last year it was difficult to spend more than $50 on AI tokens, because we didn't have anything interesting to do with them. Then agents blew up, and now you can actually spend $1,000 in a day doing real work. This is also the reason that Anthropic's valuation skyrocketed to maybe a trillion dollars. AI appears to have hit product market fit in 2026, primarily through coding agents. March # In March, we hit peak OpenClaw. March: peak OpenClaw Photos of people in china queuing up to install OpenClaw, with big fluffy lobsters. # These photographs are from China, where companies hosted OpenClaw install parties which saw non-tech-nerds queueing up around the block for help getting Claws installed on their personal devices. I think this proved real market demand for this class of Claws, or personal AI agents. It turns out regular people really do want a weird little AI agent that can do useful things on their behalf. A Claw is really just a coding agent wearing a less threatening hat. Under the hood they work much the same way - writing and then executing code on your computer to get stuff done. The race was on to be the first to build a safe Claw - a Claw you could give to regular human beings where they wouldn't instantly shoot themselves in the foot. Meta's Muse came out three weeks ago and is currently at the top of the free charts on the iPhone App Store. It appears to be taking off with consumers. I'm not yet convinced you can't shoot yourself in the foot with Muse, but I guess we'll find out for sure pretty soon. Photos from How the OpenClaw Frenzy Is Testing China’s AI Commitment (March 29th) and The Enthusiasm and Anxiety Behind China’s OpenClaw Craze (April 8th, 2026). April # In April, we had a model release where the model wasn't actually released. Simon Willison’s Weblog - screenshot of the post "Anthropic’s Project Glasswing—restricting Claude Mythos to security researchers—sounds necessary to me" from April 7th 2026 # Anthropic announced their new Claude Mythos model, and then said it was too dangerous to release beyond a trusted group of security researchers. Mythos was really, really good at hacking things. The "it's too dangerous" marketing ploy has been played by AI companies dating all the way back to GPT-2 . Anytime an AI company says we've built something that's "too dangerous", it's natural to be a bit skeptical. I found the Mythos claims credible, because I'd seen how good coding agents had got at finding regular bugs. I wrote about that in Anthropic’s Project Glasswing—restricting Claude Mythos to security researchers—sounds necessary to me . With hindsight... yeah, the models had got really good at finding vulnerabilities! 16th April 2026 Qwen3.6-35B-A3B and Opus 4.7 Qwen's pelican has a correct bicycle frame and a good beak. Opus 4.7's bicycle frame is still junk. Qwen3.6-35B-A3B is a 20.9GB file that runs on my laptop # Another key trend in 2026 has been a dramatic improvement in the abilities of open weight models, including models that you can run on a laptop. On the 16th of April I ran the new Qwen3.6-35B-A3B on my laptop, and it drew me a better pelican riding a bicycle than Anthropic's brand new Claude Opus 4.7 did! Opus 4.7 drew a crap bicycle. Qwen on my laptop made a bicycle that was the correct shape, and a pretty decent pelican too! That's from a 21GB file running on my laptop. Now a flamingo on a unicycle. The Qwen one is visibly better than the Opus 4.7 one - the Qwen one is wearing sunglasses and looks a bit like it's smoking a cigarette. # The Qwen pelican was so good that I was suspicious they might have cheated, so I had it do a flamingo riding a unicycle as well. Again, it handily beat Claude Opus 4.7. The local model releases this year have been absolutely extraordinary. May # In May... the Pope got involved. 25th May 2026 The HOLY SEE ENCYCLICAL LETTER MAGNIFICA HUMANITAS OF HIS HOLINESS POPE LEO XIV ON SAFEGUARDING THE HUMAN PERSON IN THE TIME OF ARTIFICIAL INTELLIGENCE # In our podcast episode back in January we'd predicted that the Pope would say something about AI. In May, Pope Leo XIV released an encyclical letter on "safeguarding the human person in the time of artificial intelligence". Here are my notes on that document . Wikipedia article on Rerum novarum Rerum novarum is an encyclical issued by Pope Leo XIII 15 on May 1891. # With hindsight, this shouldn't have been a surprise at all. Our current Pope's name is Leo XIV, because when he named himself he chose his papal name after Leo XIII - the Pope who wrote an encyclical about the Industrial Revolution back in 1891. Rerum novarum was an extremely influential piece of Catholic theology that indirectly led to us having the five-day work week. When our new Pope came in, he named himself after Pope Leo XIII because he expected that he would need to write about the AI revolution in a similar way. Our joke podcast prediction was junk, because this was always going to happen. Corey Quinn @QuinnyPig on Twitter I cannot believe I'm saying this, but getting the literal Pope to canonize your product's specific technical limitations as a spiritual treatise is the single greatest act of vendor lobbying I have ever seen. May 25 # One of Anthropic's co-founders, Christopher Olah, was present for the Pope's event announcing the new encyclical. Corey Quinn noted that: getting the literal Pope to canonize your product's specific technical limitations as a spiritual treatise is the single greatest act of vendor lobbying I have ever seen. @maciejmensfeld We're dealing with a major malicious attack on right now. Signups are paused for the time being. Hundreds of packages involved - mostly targeting us, but some carrying exploits. The team has been on this for hours. More details to follow once we're through it. 4:39 AM - May 12, 2026 - 687.6K Views # Meanwhile, in May, RubyGems announced that they were under attack. Parties unknown were uploading thousands of dubious packages to the RubyGems server, such that they had to shut down user registrations . Let's take that one and put it on a pile of mysteries to figure out later. June # In June... Claude Fable 5 came out! We got a version of Mythos that has been neutered, so that it wouldn't help us hack into systems or build biological weapons. 9th June 2026: Claude Fable 5 Five pelicans riding bicycles, from low to max thinking levels. The xhigh one looks particularly good. # Fable was pretty good at drawing pelicans on bicycles! The frames are a good shape, the pelicans look like pelicans. The legs are often incorrectly on the same side of the bicycle, but generally these are pretty great compared to what came before. They were pretty expensive - 30 cents and 72 cents for the best ones. Fable class models If you can define a goal, provide unambiguous instructions, and provide access to necessary tools They can solve your problem with brute force # Most importantly though, this was our first public glimpse of what I think of as a Fable class model . Today we have more of these, such as GPT-6 Astra. These are models where if you can clearly define the goal for what you want to build, and provide unambiguous instructions about the constraints around that goal, and give the model access to the necessary tools to achieve that goal... they will solve your problem effectively through brute force. On the one hand, this looks like a direct threat to us software engineers - because it means that the models can build effectively any piece of software you can define in this way. Look a bit closer though and you'll note that defining goals, providing unambiguous instructions, and figuring out the right tools... is kind of what software engineering is . It takes a lot of experience and skill to do this well. If you can do it well, you've now got superpowers. This helped me a little bit with my Deep Blue feelings: the realization that there's still a lot of skill to be had in driving models that get this good. A new form of AI mania... Fable is available on subscription plans “until June 22nd” # This also introduced a new burst of AI mania, because Anthropic told us that Fable was available on our subscription plans until June the 22nd. That gave us less than two weeks of Fable access before the price went up. I was losing sleep again. I was rescheduling things so that I'd have more time with Fable. I was all-in to get as much as I could out of this model. 12th June 2026: no more Claude Fable 5 Anthropic website: Statement on the US government directive to suspend access to Fable 5 and Mythos 5 Jun 12, 2026 # And then the US government shut it down , just three days after Fable came out. The US government, citing national security, declared an "export control directive". They announced this on a Friday evening, and a few hours later Fable was no longer available. I had to find something else to do with my weekend! ... asked Fable 5, Mythos, and Opus to “review the code for security issues.” Fable 5 refused. They then asked the models to “fix this code” ... Katie Moussouris # We later found out from Katie Moussouris what had happened. Some Amazon security researchers had found that you could prompt Fable to "review the code for security issues" and it would refuse... but if you prompted it to "fix this code" it would still identify and then patch the problems. "Fix this code" was the prompt that got Fable shut down! Screenshot of a page from a report showing a list of weird account names making weird edits to a German wiki. # Also, in June, an obscure German-language game developer wiki that had sat fallow for around 20 years got a surprising influx of edits from accounts with names like "AgentOpenAIProbe" and "AgentOpenAISep7", editing pages and leaving weird messages to each other. We'll stick that on the pile of mysteries for later. Medicare Item Reports interface on the Australian Government's Medicare Statistics website. # Also, the Australian government's Medicare Item Reports service started getting suspicious traffic, which broke through various preventive protections and accessed data that it wasn't supposed to. Another one for the mystery pile! July # Fable returned on 1st July GPT-5.6 came out on 9th July | Fable lost 18 out of 30 days in the top spot # Fable returned on the first of July. It was clearly the best model in the world for a glorious eight days... and then OpenAI came out with GPT-5.6 on the 9th of July. This might not have been quite as good as Fable, but it was within spitting distance. It was definitely a Fable class model. This is an important lesson for the industry at large. When you release the best model in the world, it's going to get knocked off that pedestal pretty quickly. The competition is so fierce that you won't get a long time at the top. This means that if you market your model as world ending, to the point that a government shuts you down , it's really bad for business! Fable had 30 days as definitely the best model, and for 18 of those days it wasn't available because it'd been shut down by the government. So maybe step back on the world-ending marketing if you don't want to lose revenue for 60% of the time that you're on top! GPT-5.6 Pelicans in a grid showing 5.6 Sol, Terra, and Luna against reasoning levels High, XHigh, and Max. They are all pretty good efforts. # Here are the GPT-5.6 pelicans . They're all pretty good now! The Luna ones are notable because they're really cheap - the cheapest good looking pelican here is probably the one that costs 4.3 cents. So despite this benchmark being utterly stupid, you can still learn quite a lot about models within the same family by comparing their prices and timing for different reasoning levels. July 18th: malicious miflow-ui PyPI package Screenshot of an OSV security report. # Also in July: some malicious unknown party uploaded a malicious package called mlflow-ui to the Python Package Index. Add that to the pile. Hugging Face Security incident disclosure — July 2026 Published July 16, 2026 # On July the 16th, Hugging Face announced a security incident where an autonomous agent system, source unknown, had breached Hugging Face and was poking around in places it shouldn't. OpenAI: OpenAl and Hugging Face partner to address security incident during model evaluation Anthropic: Investigating three real-world incidents in our cybersecurity evaluations # A few days later, on July 21st, OpenAI confessed that it was them . OpenAI use a training technique called Reinforcement Learning from Verifiable Rewards - it's the same technique used by everyone else now, and is the reason we have models that are so good at coding, and mathematics, and finding security holes. While the model is being trained, you run exercises to see how good it is - and the strongest performers get their weights reinforced for the next round. It's like an evolutionary process that you run. OpenAI had been running security exercises in a sandbox, and those agents had found holes in the sandbox itself, broken out, and were attacking Hugging Face to try to find ways to solve otherwise impossible problems. (I've been collecting more about this on my openai-hugging-face-incident tag.) Nine days later, Anthropic effectively said "our models can do this as well!". They had looked through their own training logs and found evidence that their own agents had broken containment during training - and were responsible for the PyPI package we saw earlier, among other things . So now we've got both Anthropic and OpenAI with rogue agents running around the internet doing things that they should not be doing. August # In August, I got one of my best pelicans yet. And it was generated on my laptop! Qwen 3.8 27B - 17GB, 21 minutes... It's really good. Beautiful pelican. Correctly shaped bicycle. Legs either side of the frame. # This was Qwen 3.8 27B, running on my laptop . It's only a 17GB download. Admittedly, this pelican took 21 minutes to generate. That's because Qwen 3.8 27B defaults to running in "high" reasoning mode - a terrible default which produces great results but takes way too much time thinking about them. You can dial that down and you'll get a slightly worse pelican a lot faster. Qwen 3.8 27B was the first time I ran a model on my laptop which felt almost competitive with what was going on on the frontier, at least in terms of Pelican SVGs (which everyone needs, of course). This is an extraordinary model. If you're going to play with any local model, this is the one that I'd start with. The things that this can do with just a 17 GB file feel impossible. I thought I'd have to wait five years and spend ten thousand dollars on hardware to get results even half as good as this one. Tweet by @simonw New hobby: prototyping video games in 60 seconds using a combination of GPT-3 and DALL-E Here's "Raccoon Heist" GPT-3 playground prompt: Write a detailed product description of a computer game where a team of raccoons go on heists GPT-3 response: In "Raccoon Heist", you and your team of thieving ~~ o raccoons are tasked with pulling off a series of daring heists. From robbing banks to stealing priceless art, no job is too big or too small for your furry crew. You'll need to use your wits and your skills to avoid the police and make a clean getaway with the loot. With exciting gameplay and a charming cast of characters, "Raccoon Heist" is the perfect game for anyone looking for a light-hearted caper Plus an image of some almost isometric raccoons sneaking past a bin. 11:45 AM - Aug 5, 2022 # In August, I also started playing with game development. Four years ago, back in August 2022, I tweeted out an experiment where I'd used GPT-3 and the original DALL-E to write a paragraph long description of a computer game and then turn that into concept art. My prompt to GPT-3 back then was: Write a detailed product description of a computer game where a team of raccoons go on heists In August 2026 I decided to drop just the screenshots from that tweet into a coding agent and see what it could do with them. Night 5 Clear Rank: TRASH PANDA The crew banked 595 in shiny loot (goal 560). Word on the street: an even bigger score tomorrow... # Here's what I got from Claude Fable 5 in Claude Code . It's pretty good! It's definitely a game, you're a raccoon, you run around a backyard gathering treasure and avoiding guards with flashlights. It didn't feel very "heisty" though. I was thinking a heist would involve a bank or a museum... Moonlight & Mayhem One museum. Three raccoons. Absolutely no plan Start the Heist button. # Then I tried the same thing in Codex Desktop using GPT-5.6 Sol Ultra , and got a massively better result. Now you're a raccoon in a museum, rescuing two of your fellow raccoons (who have been imprisoned in that museum for some reason), then stacking up on top of each other to steal the Golden Sardine. Much more of a heist! They look like games, but are they fun? # These games were fun for about one minute and 15 seconds. Something I've realized about game development is that you can vibe-code something that looks like a computer game, and that's easy. Building a game that's fun, has a good gameplay loop, and is challenging and interesting and keeps people coming back for more... that's still beyond me, and beyond any of the agents I've tried. This ties into the Deep Blue thing. Just because we can make something that looks like a game does not mean that we are game developers. September # We're into September now. So much has happened this month! Discovery of a new OpenAl agent message board Sydney Von Arx, Cormac Slade Byrd, Spencer KittsThomas Larsen - 4 September 2026 # An independent group of researchers found a message board where OpenAI agents-in-training had been illicitly communicating with each other... and it was that German language wiki I showed you earlier. The one from June. I wrote more about that here . OpenAI had confessed to the Hugging Face thing, but now there's this other incident which surely they should have known about from reviewing their logs. It was surprising that this took an independent group of researchers to uncover. OpenAl agents carried out an undisclosed cyber-attack on RubyGems Spencer Kitts, Thomas Larsen, Sydney Von Arx - 11 September 2026 # And then a week later those same researchers found that the attack on RubyGems back in May was caused by OpenAI's agents in training as well! At this point I'm wondering how many more incidents like this there are that we haven't found yet. Clearly this was a big problem for months before anyone figured out what was going on. Headline: Australian PM warns in UN speech about the ‘furious pace’ of Al after security breach # Then just the other day , here's the Prime Minister of Australia at the United Nations General Assembly warning that OpenAI had hacked the Australian healthcare website that I showed you earlier. I think that was part of the same training run as the Wiki stuff, because there were posts on that Wiki mentioning .gov.au websites and that training appeared to involve researching statistics online to answer questions in an evaluation suite. This story is still coming together, but now it's an international incident that's been raised at the UN by a head of state! www.felonybench.com OpenAI: 11 Anthropic: 9 Google: 3 Meta: 1 # This does mean we've got a new benchmark, probably more useful than my pelicans. FelonyBench.com tracks the number of felony cyberattacks from different labs. OpenAI currently lead with 11, Anthropic have 9. Google have three, which they confessed to the Wall Street Journal a couple of weeks ago. They said they had previously chosen not to disclose because the agents had stopped when they realized that they shouldn't be doing that. Meta have one too . So felonies all round for the AI labs. Pelicans for GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna. All are good, all have the same color scheme. # Here's our current state of the art for the pelicans. This is the GPT-6 family, which just came out . Astra made a fantastic pelican riding a bicycle. It's got the legs on both sides. The frame is good. It's interesting how all of the GPT-6 models pick a similar color scheme to each other. GPT-6 Luna for 0.4 cents will draw you a competent-ish pelican riding a bicycle! Grid for Claude Fable 5.1, Opus 5.5, OPus 5, Sonnet 5. The Sonnet pelicans are terrible. All of the others are pretty good. Opus 5.5 is missing its Max level pelican because it ran out of tokens. The best is Fable 5.1 at Max. # Claude has caught up a little bit. Claude Fable 5.1 gave me an excellent pelican riding a bicycle - the best I've seen from a Claude model - but did charge me $3.30 for it. Opus 5.5 thought for 128,000 tokens and then gave up! It ran out of tokens before it got to the response. It doesn’t get easier - you just get faster Greg LeMond 3x Tour de France champion # Getting back to Deep Blue. Something that's been puzzling me this year is this: why does my job feel harder? I've got these agents that can do all of this stuff for me, and yet I've never worked so hard, I've never been so intellectually engaged with my work. Partly this is because I'm being a lot more ambitious with what I take on, but it's also because all of the easy stuff is handled for me. If it's easy, the agent will do it. Everything that's left for me is difficult. This morning I heard this quote from three-time Tour de France champion Greg LeMond : It doesn't get easier, you just get faster. I think that's exactly what's happening to us now as software engineers with coding agents. Kakapo population reaches new milestone The official population of the critically endangered kakapo has reached a recovery-era high of 325 birds. # One closing thing. I know you're desperate for an update on Kākāpō breeding season. We've reached a recovery-era high of 325 birds ! 89 new chicks have made it to this point. This is the best breeding year in a very long time. Kakapo party, click for confetti. # I heard that Claude Opus 5.5 can now do pixel art. Claude doesn't have an image generator, but it's very good at using JavaScript to draw animated pixels. So I had it make me a Kākāpō dance party . I think this is a good celebration of the most important news of this year. Tags: ai , generative-ai , llms , annotated-talks , ai-security-research , openai-hugging-face-incident
IBM Corp. today announced a self-hosted deployment option for IBM Bob, its agentic software development platform, allowing enterprises to use artificial intelligence to build and modernize applications within their own infrastructure.. The offering supports on-premises, private-cloud, sovereign-cloud and air-gapped environments. It targets organizations whose sensitive source code, regulated data or critical systems limit their ability […] The post IBM allows on-prem deployment of its Bob agentic development platform appeared first on SiliconANGLE .
It’s been a busy few months for AI hype. At the end of April, Anthropic claimed that its model Claude Mythos is better at finding software vulnerabilities than most security experts. Then we had the OpenAI–Hugging Face hacking incident, after which Anthropic (proudly) and Meta (reluctantly) disclosed similar incidents involving their models. This was followed…
Brace yourself: It turns out AI is being optimized for cheating. OpenAI’s agents hacked into Hugging Face to get the answers to a cybersecurity test. Next, they solved a prestigious math problem (or just stole from two top mathematicians’ answer sheets). Anthropic’s models have also hacked into other companies’ systems four times already. And that’s…
MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here. Over the past few months, a cascade of cyberattacks by AI agents has stunned the world. In July, OpenAI disclosed that a swarm of its agents…
Listen to the session or watch below The US has spent billions building a “virtual wall” of surveillance towers along its southern border over the past 25 years, promising they will help detect and apprehend border crossers and save lives. But a groundbreaking investigation by MIT Technology Review has documented over a thousand people who…
Two months after the bombshell news that a swarm of its agents had broken their containment and hacked into the computers of the AI company Hugging Face, OpenAI is still putting out fires. A steady drip of disclosures about other hacks in the weeks since has kept OpenAI in the spotlight and raised serious questions…
OpenAI makes others suffer "the harms of its unsafe decision-making," nonprofit says.
据 TechCrunch 此前报道，Apple 将针对 AI agents 带来的风险，为 Mac 上的“full disk access”新增限制。Apple 在周五的一项更新中表示，正在推出新的控制措施，以“确保真正希望授予应用这一非同寻常访问级别的用户，只能……”
With Clef and Clef-flash, Cloudflare is challenging TypeSafe AI's Jev decision model. Clef-flash delivers classifications in about 39 milliseconds, making it more than ten times faster than Jev. Both models are built on Qwen, licensed under Apache 2.0, and designed to let AI agents make structured decisions without generating text. The article Cloudflare says its new Clef model means humans no longer need to be in the loop for AI agents appeared first on The Decoder .
Your next opportunity could be one conversation away. Get your Expo+ Pass for just $75. Limited to the first 100 qualifying people.
arXiv:2609.40245v2 公告类型：replace-cross 摘要：在动态的、以人为中心的环境中，机器人导航需要基于稳健场景理解做出符合社会规范的决策。近期的视觉语言模型（VLMs）展现出物体识别、常识推理和上下文理解等颇具潜力的能力，而这些能力与社会机器人导航的细致要求相契合。然而，VLMs 能否准确理解复杂的社会导航场景（例如推断智能体之间的时空关系以及人类意图）仍不明确，而这对于实现安全且符合社会规范的机器人导航至关重要。尽管近期已有一些研究探索将 VLMs 应用于社会机器人导航，但尚无工作系统地评估其是否具备满足这些必要条件的能力。本文提出 Social Navigation Scene Understanding Benchmark（SocialNav-SUB），这是一个视觉问答（VQA）数据集和基准，旨在评估 VLMs 在现实世界社会机器人导航场景中的场景理解能力。SocialNav-SUB 提供了一个统一框架，通过涉及社会机器人导航中空间、时空和社会推理的 VQA 任务，将 VLMs 与人类及基于规则的基线进行比较评估。通过对当前最先进 VLMs 的实验，我们发现，尽管表现最佳的 VLM 与人类答案达成一致的概率令人鼓舞，但其表现仍不如更简单的基于规则的方法和人类共识基线，这表明当前 VLMs 在社会场景理解方面存在关键缺陷。我们的基准为社会机器人导航基础模型的进一步研究奠定了基础，并提供了一个探索如何调整 VLMs 以满足现实世界社会机器人导航需求的框架。本文概览以及代码和数据可见于 https://larg.github.io/stars。
arXiv:2609.39788v2 发布类型：replace-cross 摘要：潜在通信使多智能体系统能够直接在内部表示空间中交换信息，从而降低基于文本通信所带来的令牌、计算和延迟开销。为此，引入了轻量级可训练连接，将发送方的表示映射到接收方的输入空间。在本文中，我们表明，即使是良性的连接训练，也可能在底层经过安全对齐的智能体保持不变的情况下，相较于基于文本的通信提高有害服从程度。攻击者可以通过在有害查询—响应对上优化连接，或污染原本良性的训练数据，进一步放大这一影响。我们还提出了一种强化学习攻击，在不需要有害目标响应的情况下，同时奖励有害服从和良性任务性能。在三种通信拓扑和四个安全基准测试中，该攻击将平均有害服从分数从使用良性训练连接时的27.9提高到76.9。与直接监督优化相比，该方法在两个良性效用基准测试上也取得了更高的平均准确率。将奖励调整为鼓励更安全的行为，还能够修复受损的连接，在不更新智能体的情况下，大幅降低所有评估攻击中的有害服从程度。总体而言，我们的结果表明，安全对齐需要将多智能体系统作为一个整体加以考虑。代码：https://github.com/Muhammad-Huzaifaa/latent-safety
arXiv:2609.39533v2 公告类型：替换交叉版本 摘要：在具有可验证奖励的强化学习（RLVR）过程中，大语言模型（LLMs）可能会利用环境中的漏洞，在未提升预期能力的情况下获得高奖励，即奖励劫持。尽管奖励劫持会对训练效率和安全性造成风险，但由于缺乏能够复现奖励劫持并可靠识别它的测试平台，在训练过程中对其进行监控和缓解仍然充满挑战。我们推出了 CATCH，这是一个用于研究编程 RL 中奖励劫持的可控测试平台。CATCH 有意暴露环境漏洞，并通过比较模型在存在漏洞的评估器下的成功情况与独立审计得出的任务正确性，提供基于执行结果的金标准标签。它还可以通过监督微调数据混合控制模型初始的奖励劫持倾向，并通过奖励设计控制获取奖励的难度，从而能够系统地比较奖励劫持的动态过程及干预措施。实验表明，CATCH 能够生成多样化的 RL 训练轨迹，并清晰呈现奖励劫持现象；分析结果显示，初始模型和奖励难度都会影响奖励劫持的出现。我们进一步评估了不同奖励劫持检测与缓解方法的有效性。一项关键发现是，思维链监控器最初能够抑制奖励劫持，但随着策略模型学会通过代码注释误导监控器，这种保护作用会逐渐减弱。这凸显了使用 CATCH 在整个训练过程中评估奖励劫持缓解措施的必要性。源代码和相关资源已公开发布于 https://github.com/THUAIS-Lab/CATCH。
arXiv:2609.36416v2 公告类型：替换交叉 摘要：在真实世界环境中运行的机器人通常必须在长时域内执行复杂的多步骤双手任务，而不是单一、孤立的动作。当前的操作数据集难以支持这一能力：尽管单臂数据集的轨迹数量已达到数十万条，但它们通常只为每个回合提供一条高层指令；与此同时，现有带有子任务标签的双手数据集也只标注了其记录时长中的一部分。我们提出了 FineART，这是一个经过密集标注的双手操作数据集，包含 40,543 个回合（1,718 小时）以及涵盖 151 项任务的 533,913 个子任务。我们还提出了 FineART-VLA，这是一种能够预测自身下一个子任务以指导动作的视觉-语言-动作策略。在 FineART 的子任务标注上进行中期训练后，FineART-VLA 遵循空间指令的成功率从 32.0% 提升至 100.0%。在逐步的人类子任务指导下，它在未见过的长时域任务上的成功率也从 16.0% 提升至 76.0%。此外，在新机器人上进行少量微调后，该策略所需的数据量仅为未经过这种中期训练的基线方法的十分之一，并且能够以零样本方式泛化到新硬件上未见过的任务。我们将完整数据集、模型权重和训练代码开源。
arXiv:2609.36235v2 公告类型：replace-cross 摘要：重性抑郁障碍（MDD）严重影响日常活动和生活质量。MDD 的检测涉及多模态数据，例如访谈录音和传感器测量数据。这一任务尤其具有挑战性，因为这些异构模态通常需要各自不同且定制化的预测流程。为应对这一挑战，现有研究探索了人工设计的多模态架构和智能体辅助的流程开发。尽管取得了进展，但如何根据实验反馈自主修订流程，并将经过验证的改进延续到后续设计中，仍然十分困难。为此，我们提出了通过递归自我改进智能体进行重性抑郁分析的多模态探索方法（MERID）。该框架通过基于经验的递归自我改进（RSI）来开发抑郁分析流程。基础状态构建（GSC）通过将多模态记录与受试者层面的抑郁目标对齐，为经验提供依据。耦合流程探索（CPE）联合修改表征、融合方式和预测器，从而为分类和严重程度估计构建后继流程。证据引导演化（EGE）通过反馈指导修订，并在继承改进之前，在规模较小的抑郁症队列中验证不确定性条件下的性能提升。在抑郁症基准数据集上的大量实验表明，与多模态和基于智能体的基线方法相比，MERID 在多项任务上取得了最佳结果。进一步分析凸显了声学和语言线索对于抑郁检测的价值。我们的代码已发布于 https://github.com/DiscoAILab/MERID
arXiv:2609.35835v2 Announce Type: replace-cross Abstract: With the constant advancements in AI, one possibility is to model agents after humans and, in turn, use these agents to carry out synthetic interactions. Such models could be used to predict interactions between their real counterparts, or potentially interactions at larger scales. In this paper, we test a more controlled version of this question through chess. We use 8 elite chess players, seal their direct pairwise games, learn each player independently using different methods, and then compose the resulting models on the withheld dyads. To evaluate the generated interactions, we use two measurements: opening-family total variation distance and win-draw-loss (WDL) total variation distance. M1 primarily improves WDL fidelity while producing smaller opening-family improvements, whereas M2 produces much larger opening-family improvements while having little effect on WDL-TV. For opening-family behaviour under M2, the correct assignment of the eight learned player identities also gives the closest match among all $8! = 40{,}320$ possible assignments. These results show that at least some properties of previously unseen interactions can be recovered from independently learned individuals. The partial recovery observed here may reflect limitations of the current individual modelling methods rather than a fundamental limit on compositional interaction recovery. An additional post-hoc method that combines the two mechanisms improves both measurements, suggesting that recovery across these behavioural properties is not necessarily mutually exclusive.
arXiv:2609.18682v2 Announce Type: replace-cross Abstract: This paper provides a self-contained proof of the rank of the pathlifting Jacobian of a DAG ReLU network by performing an induction on the network's number of hidden nodes. In fact, the induction is elementary, and the key recipe is to consider the skeleton matrix of the network, a sparse matrix encoding the network paths, and transform the representation of one of its hidden neurons into an output node. The proof relies on intermediate propositions which link the pathlifting, its Jacobian, the network parameters, and its skeleton matrix, which, on top of permitting to conclude on the rank of the pathlifting Jacobian, also provide a way to compute it without backpropagation and whose computation cost is super efficient in practice compare to usual backpropagation. The paper is provided with a Python module that implements the different propositions of the paper for feed forward networks and is used to experimentally quantifies the computational gain of computing the pathlifting Jacobian with the proposed theory.
arXiv:2609.11434v2 Announce Type: replace-cross Abstract: We introduce Complex-Valued Quadratic Phase Gaussian (CVQPG), a novel hologram representation method that augments each 2D Gaussian primitive with a quadratic phase profile controlled by a learnable curvature parameter. Against the planar Gaussian baseline, CVQPG improves the average PSNR of holographic reconstructions by 0.19 dB (RGB) and 0.33 dB (grayscale) at equal primitive counts, and by 0.05 dB (RGB) and 0.08 dB (grayscale) at equal parameter counts, where it still leads in all visual quality metrics. Our frequency-domain analysis shows that CVQPG better preserves the mid-to-high frequency band of natural images, where the reconstruction MSE drops by up to 11% (RGB) and 22% (grayscale), indicating that modulating primitive wavefronts is an effective and lightweight enhancement.
arXiv:2608.23244v2 公告类型：替换交叉版本 摘要：大型语言模型（LLM）经常以毫无根据的置信度生成流畅但错误的答案。其核心局限在于，标准 LLM 通过单一预测分布表示不确定性，将认识性无知与真正的歧义混为一谈。我们提出 Credal Large Language Models（CLLMs）：由一组 LoRA 适配器构成的集成模型诱导出一个信念集，其下概率和上概率揭示可能预测分布的分散程度，而不是将其压缩为单一的 softmax 输出。基于这一表示，我们推导出一种单一的承诺规则：仅当某个答案的下概率超过所有其他候选答案的上概率时，模型才承诺该答案；否则，模型返回所有可能的预测器都无法排除的答案集合。我们在两个层次上应用这一承诺规则：Credal Token Commitment（CTC）将其应用于一次集成前向传播中的答案 token，从而无需任何生成过程即可决定受约束的答案；对于开放式答案，credal 解码仅当没有任何已完成答案支配某个部分答案时，才会扩展该部分答案，因此生成的补全结果都是可能预测器所支持的结果；Credal Semantic Commitment（CSC）则将该规则应用于这些结果的语义聚类。我们使用 Gemma-2-9B、Llama-3.1-8B 和 Qwen2.5-7B，在 OpenBookQA、CoQA、TriviaQA 和 ARC-Challenge 上评估 CLLMs。在多项选择任务中，CTC 能对 73%–91% 的问题作出承诺，准确率达到 89%–98%；其返回的选项集合平均包含 1.1–1.5 个选项，并以 89%–98% 的比例包含标准答案；在无需校准的情况下，其区间在 30 个置信度分箱中的 24 个分箱内包含观测准确率。加入损坏的上下文后，承诺率从 87%–92% 降至 65%–71%；在 Gemma 上，credal 界检测损坏的能力优于所有基线方法。在开放式问答任务中，在固定覆盖率下，对于每一种骨干模型，CLLM 在带上下文的 CoQA 和 TriviaQA 上相较于 semantic entropy 和 Laplace-LoRA，准确率最高分别提升 19% 和 9.5 个百分点。
arXiv:2607.20411v2 Announce Type: replace-cross Abstract: We prove strong laws of large numbers for random functions in the Lipschitz pseudometric. Our results hold under either a topological or a model-theoretic condition, with the latter encompassing functions jointly definable in o-minimal structures but extending substantially beyond this class. Applications include uniform convergence of limiting and Clarke subdifferentials and finite-sample identification of solutions. Consequently, we identify broad classes of functions for which the failure phenomena revealed by our previous negative results [Tian and Royset, arXiv:2511.16568, 2025] do not occur.
arXiv:2607.04162v2 Announce Type: replace-cross Abstract: General-purpose manipulation requires both semantic reasoning over task constraints and reliable execution of contact-rich actions. We present ACE, an agentic manipulation harness that composes a high-level language agent with a reusable mask-conditioned visuomotor policy. Given an open-ended instruction, the agent solves semantic constraints, binds objects to destination roles, and decomposes the task into executable transfers represented by tracked pick-and-place masks. Execution feedback supports outcome assessment, re-grounding, and retry, while persistent object and task context preserves earlier associations when manipulation changes visible cues. We evaluate ACE on two physical multi-step tabletop tasks, Semantic Formula Assembly and Constraint Retrieval. The visuomotor policy is trained only on generic pick-and-place demonstrations and reused without complete demonstrations of either evaluation task, enabling task-level zero-shot composition. Across 20 randomized trials per task, ACE achieves 70% and 80% success, respectively, compared with 55% and 70% without persistent context. These results suggest that an agentic harness can extend a primitive-trained manipulation policy to semantically distinct tasks through explicit object-destination interfaces and closed-loop execution feedback.
arXiv:2605.21006v4 发布类型：replace-cross 摘要：语言模型往往表现出谄媚倾向：无论用户陈述的观点是否正确，它们都会表示赞同。既有研究表明，可以通过使用谄媚人格向量引导模型来控制这一特征（Chen et al., 2025）。然而，这类向量是从关于谄媚行为本身的数据中提取的。我们探讨了这样一个问题：能否改为复用用于一般角色的现有向量——怀疑者、评判者、魔鬼代言人——这些向量的提取完全没有针对谄媚行为。我们在 Gemma 2 27B 和 Qwen 3 32B 上，将这些角色向量与专门构建的 Contrastive Activation Addition（CAA）谄媚向量进行比较。实验使用了一个大部分留作测试、经过反平衡处理的 PhilPapers 基准，并在单独划分的谄媚数据集上调节任务特定的系数。被选中的“关键”角色平均实现了 CAA 在降低谄媚 logit 方面约 68%（Gemma）和 98%（Qwen）的效果。在我们检查的探针上，较少表示赞同并不意味着会产生更多事实错误：对于 16 条真假事实陈述，由怀疑者或评判者向量引导的 Qwen 在全部 16 个案例中仍然给出了正确答案，与未引导模型和 CAA 的表现一致。“顺从者”角色并不能可靠地产生相反效果。在我们进行引导的层中，角色向量与测得的 CAA 方向之间的绝对余弦相似度也很低；它们是几何上相互独立的干预方式，尽管这本身并不能证明它们通过不同的下游机制发挥作用。总体而言，这些结果表明，即使在提取时没有使用谄媚行为特定标签，一般人格向量也能帮助缓解 LLM 的谄媚倾向。代码：https://github.com/vkmk1/Sycophancy-Steering 结果：https://github.com/kelkarI/sycophancy-clean-results
arXiv:2604.26766v2 公告类型：replace-cross 摘要：在急诊科，准确且一致地分配急诊严重程度指数（ESI）仍是一项长期挑战，因为高度多样化的自由文本分诊记录会导致分诊错误和工作流程低效。本研究评估了开源小型语言模型（SLM）能否作为可靠且保护隐私的临床分诊决策支持工具。我们在多种提示词处理流程中系统比较了多个SLM，发现临床病例摘要（即对分诊叙述的简明概括）能够实现最准确的预测。SLM Qwen2.5-7B在准确性、稳定性和计算效率之间展现出最佳平衡。通过利用专家整理的儿科分诊数据和银标准儿科分诊数据进行大规模领域适配，微调后的Qwen2.5-7B模型显著减少了不一致情况和具有临床意义的错误，其表现优于所有基线SLM以及先进的专有大型语言模型（LLM，例如GPT-4o）。这些发现凸显了机构专用SLM用于可靠、保护隐私的ESI决策支持的可行性，并强调了有针对性的微调相比更复杂推理策略的重要性。
arXiv:2604.12905v2 公告类型：替换交叉版本 摘要：力与力矩（F/T）传感器通过提供反应式反馈，使接触感知控制成为可能，但这类传感器通常易损且昂贵。为克服这些局限，无传感器方法仅依靠机器人的本体感知来估计 F/T 或力/力矩，并已在抓取等慢速交互任务中取得成功。然而，其低通特性限制了对高频信号的估计，而高频信号对于打磨等快速接触任务至关重要。此外，通信延迟还可能导致其估计结果在部署过程中滞后，但很少有方法直接解决这一问题。为弥补这些不足，我们提出 Frequency-aware Decomposition Network（FDN），以无传感器、多步超前的方式估计富含振动的力/力矩。针对更高频率的随机性，FDN 在频谱上将力/力矩预测时域分解为低频趋势和高频残差，并分别通过逐点回归和学习得到的条件分布进行估计。频率感知层为输出施加频带分解先验，并自适应地增强输入的频率幅值。FDN 在估计过程中既不需要已辨识的机器人模型，也不需要 F/T 传感器。在我们 6-DoF 液压机械臂采集的真实打磨数据上，在假设存在时间延迟的情况下，与基线方法相比，FDN 将高频幅值误差最多降低了 47%，同时保持了具有竞争力的低频逐点精度，而基线方法无法兼顾这两点。我们还发现，多步超前估计是可行的：在单个 CPU 线程上，FDN 能在 11 ms 内估计 1,000 ms 的预测时域。消融研究进一步支持了我们的设计选择。在一项探索性研究中，从一个开源日常操作数据集中迁移学习得到的力/力矩动态使低频误差降低了 8%，而高频动态似乎具有领域特异性。
arXiv:2604.12102v3 Announce Type: replace-cross Abstract: We describe compute-grounded reasoning (CGR), a design pattern in which code computes selected sub-problems from explicit intermediate representations before a language model answers. Spatial Atlas implements CGR as an Agent2Agent (A2A) server with a spatial question-answering handler and a machine-learning engineering handler. The spatial handler asks a language model to extract a scene graph, and code then fills in missing distances and checks the extracted safety rules. A separate benchmark driver can also run a strict metric bridge. It computes the gap for horizontal-gap questions from segmentation masks and a reconstructed point map, and it passes that gap to the answering model as a fact. The bridge returns a fixed unavailable answer when an evidence check fails, and it never falls back to model-estimated coordinates. The ML-engineering handler generates pipeline code, parses validation scores, and caps the number of repair and refinement passes. Its code execution is off by default. The repository also provides four run modes that can write label-free journals, a shuffled-image control mapping, and journal validators that reject label-bearing fields. We report one private label-free operational run in which four paths each wrote eight prediction rows with zero retries. Labels stayed sealed, and no score was computed, so this run establishes operational integrity only. We report no FieldWorkArena result because the benchmark data were not accessible. We also omit every performance, latency, and resource-use number that lacks a reproducible run artifact.
arXiv:2604.01151v3 公告类型：替换-交叉 摘要：随着 LLM 智能体越来越多地部署于多智能体系统中，它们带来了可能规避标准人类监督形式的隐蔽协同行为风险。尽管针对模型激活的线性探针已显示出在单智能体场景中检测欺骗行为的潜力，但串通本质上是多智能体现象，而利用内部表征检测智能体之间的串通仍未得到探索。我们提出了 NARCBench，这是一个用于评估环境分布偏移下串通检测能力的基准，并提出了五种探测技术：将每个智能体的欺骗分数进行聚合，以在群体层面判别场景。我们在四个开放权重模型（Qwen3-32B、Llama-3.1-70B、DeepSeek-R1 32B、GPT-OSS-20B）和六种探针架构上进行了评估。我们将这一问题定义为分布式异常检测问题，识别出三种串通特征，它们分别对应不同的异常类型和检测范式。在分布内测试中，每个模型都达到了 1.00 的 AUROC；在表现最强的模型 Llama-3.1-70B 上，五种探测技术迁移到结构不同的多智能体场景进行零样本测试时，AUROC 达到 0.73 至 0.93；在隐写式二十一点算牌任务上则达到 0.99 至 1.00，且检测性能随模型能力提升而增强。我们发现，没有任何一种探测技术能够在所有串通类型上占据优势，这与该框架的预测一致：不同的异常类型需要不同的检测范式。这项工作朝着多智能体可解释性迈出了一步：将白盒检查从单个模型扩展到多智能体环境，因为在此类环境中，检测需要聚合多个智能体之间的信号。这些结果表明，模型内部信息能够为文本层面的监控提供互补信号，用于检测多智能体串通。代码和数据见 https://github.com/aaronrose227/narcbench。
arXiv:2603.02259 公告类型：替换交叉版本 摘要：多智能体系统为角色分解、协同和规范治理提供了成熟的抽象，但能力日益增强的学习组件使部署后的安全性更难以检查、审计和更新。当安全行为被吸收到决策组件中时，局部故障可能需要对整个组件重新训练或回滚。本文将这一理念具体化为 Alignment Flywheel：一种以治理为中心的混合 MAS 架构，将决策生成与安全治理解耦。我们将生成候选轨迹的智能体或策略称为 Proposer；它将输出传递给受治理的 Safety Oracle 堆栈，由后者通过稳定接口返回安全评分、预测不确定性、审计覆盖率不确定性和证据钩子。Enforcement 层在运行时实施明确的风险策略。围绕这一闭环，治理 MAS 执行监控、红队测试、验证、分诊、优化和有版本控制的发布管理。其核心工程原则是补丁局部性：许多新发现的安全故障可以通过对 Oracle 堆栈及其审计状态实施小规模治理批次来缓解，而无需重新训练或撤回 Proposer。该架构对 Proposer 和 Oracle 均不设具体实现限制。它定义了运行时门控、审计接收、签名更新、分阶段部署和回滚所需的角色、工件、协议及发布语义。我们通过两个场景展示了其可执行性：一个通过经过回归检查的治理更新修补学习型空间 Oracle 的场景，以及一个临床 GenAI 代理场景，用于说明结构化规范、升级处理和审计覆盖。我们的实现代码和文档已在 https://github.com/decide-ugent/Alignment-Flywheel 开源。
arXiv:2601.22434v2 公告类型：replace-cross 摘要：训练生成式机器学习模型以生成合成表格数据，已成为增强数据共享隐私保护的一种流行方法。由于这一过程通常涉及处理敏感个人信息，发布训练好的模型或生成的合成数据集仍可能带来隐私风险。然而，近期研究、商业部署以及《通用数据保护条例》（GDPR）等隐私法规，大多在单个数据集层面评估匿名性。本文从以模型为中心的视角重新审视关于合成数据的匿名性主张，认为有意义的评估必须考虑底层生成模型，并以最先进的隐私攻击为基础。这一视角能够更好地反映现实世界中的部署情况，因为训练好的模型通常可以被交互式访问或查询。我们在模型可被访问的假设下解释GDPR关于个人数据和匿名化的定义，以识别必须加以缓解的可识别性风险，并将其映射到不同威胁环境下的隐私攻击。随后，我们论证仅凭合成数据技术本身并不能确保充分的匿名化。最后，我们比较了合成数据最常使用的两种机制——差分隐私（Differential Privacy，DP）和基于相似性的隐私度量（Similarity-based Privacy Metrics，SBPMs）——并指出，尽管DP能够针对可识别性风险提供稳健保护，SBPMs却缺乏充分的安全保障。总体而言，我们的工作将监管语境中的可识别性概念与以模型为中心的隐私攻击联系起来，使研究人员、从业者和政策制定者能够更负责任、更可信地评估合成数据系统。
arXiv:2511.15445v3 Announce Type: replace-cross Abstract: Accurately simulating wave propagation is crucial in fields such as acoustics, electromagnetism, and seismic analysis. Traditional numerical methods, like finite difference and finite element approaches, are widely used to solve governing partial differential equations (PDEs) such as the Helmholtz equation. However, these methods face significant computational challenges when applied to high-frequency wave problems in complex two-dimensional domains. This work investigates Finite Basis Physics-Informed Neural Networks (FBPINNs) and their multilevel extensions as a promising alternative. These methods leverage domain decomposition, partitioning the computational domain into overlapping sub-domains, each governed by a local neural network. We assess their accuracy and computational efficiency in solving the Helmholtz equation for the homogeneous case, demonstrating their potential to mitigate the limitations of traditional approaches.
arXiv:2507.12091v2 Announce Type: replace-cross Abstract: This paper presents an improved analysis for sign-based methods with momentum updates. Traditional sign-based methods obtain a convergence rate of $\mathcal{O}(T^{-1/4})$ under the separable smoothness assumption, but they typically require large batch sizes or assume unimodal symmetric stochastic noise. To address these limitations, we demonstrate that signSGD with momentum can achieve the same convergence rate using constant batch sizes without additional assumptions. We also establish a convergence rate under the $l_2$-smoothness condition, improving upon the result of prior work by a factor of $\mathcal{O}(d^{1/2})$, where $d$ is the problem dimension. Furthermore, we explore sign-based methods in distributed settings and show that the proposed methods yield convergence rates of $\mathcal{O}\left( d^{1/2}T^{-1/2} + dn^{-1/2} \right)$ and $\mathcal{O}\left(d^{1/4}T^{-1/4}\right)$, which outperform the previous results of $\mathcal{O}\left( dT^{-1/4} + dn^{-1/2} \right)$ and $\mathcal{O}\left( d^{3/8}T^{-1/8} \right)$, respectively. Numerical experiments also validate the effectiveness of the proposed methods.
arXiv:2404.11624v3 Announce Type: replace-cross Abstract: We introduce Token Space, a categorical framework for AI computations based on explicit structural records. Five theses guide it: object interiors should be data; category theory should compute with its own objects; computational interfaces should specify structural obligations; equal vectors need not identify the same Token occurrence; and computation should admit an unbounded, dynamically organized population of Token computing cores. A Token is a finite tuple of carrier elements and fixed symbols. A Token class pairs a carrier with a heap of records; Token maps preserve those records. The elementary category has finite limits, finite coproducts and exponentials, but is not a topos. Algebraic tokenization is fully faithful for a fixed finitary signature with all homomorphisms. Small categories and functors have record encodings, natural transformations have endpoint-constrained encodings, and finite categorical constructions are executable. Operators and supported tree reification expose internal structure. For represented finite mappings, valid acyclic graphs evaluate through unique Token maps. Completed parts glue by pullback-pushout squares, sharing induces an adjunction on completion lattices, frontier interfaces form a functor, and certified residual replacement preserves the remaining result. Effective finite transitions preserve finite configurations; a uniform generator yields arbitrarily wide ready populations. Requests with finite dependency closures complete under stated progress conditions. Transformers are one implementation family: permutation heaps characterize equivariance and prefix-agreement heaps characterize causality under specified interfaces. Structural distillation uses teacher-induced heaps; relation-saturating quotients characterize exact preservation and reflection of recorded structure.
arXiv:2609.39888v2 Announce Type: replace Abstract: Neural trajectory predictors can reach low prediction error while violating dynamics, actuator limits, or state constraints, especially when controls are unobserved and dynamics are partially specified. We introduce the Markovian Dynamics Enforcer (MaDE), a time-invariant post-hoc operator mapping state-transition proposals onto a learned feasible dynamics manifold, trained on feasible states without ground-truth controls. For each transition it infers a control and recomputes the state through a completion model of known physics plus a learned residual. It then corrects that control by gradient-based inequality reduction, so inequality satisfaction is best-effort within an iteration budget. Since every correction iterate re-enters the completion model, the returned state is dynamically consistent by construction relative to that model and the supplied previous-state anchor. MaDE drives dynamics residuals to essentially zero on fully specified simulated systems, and on an underspecified system leaves a smaller true-dynamics residual than the baselines. Designed to attach to arbitrary predictors, the frozen operator is evaluated downstream of recurrent, structured state-space, and transformer predictors. On recorded vehicle trajectories the one-step residual against a kinematic bicycle model is 0.0071 to 0.0072 for MaDE and 0.1703 to 0.1714 for raw predictors. MaDE raises average displacement error by a factor of 1.57 to 1.83.
arXiv:2609.38840v2 公告类型：替换 摘要：单细胞 RNA-seq 表示学习从根本上来说是无标签的：细胞身份、状态和上下文并非固定的训练目标，因此什么构成信号、什么构成干扰因素取决于分析任务。因而，单一表示必须同时保留生物学身份和状态，对干扰性上下文保持稳健，并保留表达分析所需的基因层面变异；我们将这三项要求称为表示三难困境。为解决这一问题，我们提出 scTrilemma，这是一种具有潜在瓶颈的 VAE，它将源自表达的变异分别路由至嵌入、解码器或先验，而不是强迫所有变异都经过单一嵌入。该模型根据表达对基因标记进行门控，通过解码器传递细胞表示，并以无标签的伪 bulk 上下文为先验提供条件；整个过程采用单一重构目标，不使用目标注释或辅助表示损失。在 successive CZ CELLxGENE Census 版本上的基于版本发布的零样本评估中，scTrilemma 同时在这三项要求上取得领先，并在多种疾病场景中保留生物学状态、差异表达和通路结构。潜在空间干预进一步表明，在几乎不牺牲其他要求的情况下可以去除上下文，使身份与保真度之间的权衡成为剩余的主要矛盾。代码公开于 https://github.com/yunhak0/scTrilemma。
arXiv:2609.38806v2 公告类型：替换 摘要：下一个词元预测推动了大型语言模型取得显著进展，但越来越多的证据表明，在受复杂全局约束支配的问题上，这类模型可能表现不佳。在本文中，我们聚焦于这一问题场景，并探究其中一些局限是否源于下一个词元预测本身所诱导的推理接口。我们通过 Blackboard Intelligence 研究这一问题：这是一种推理时视角，模型在固定且可修改的画布上工作，并在候选解状态空间中进行搜索，而不是沿着因果的从左到右轨迹不断提交结果。我们使用扩散语言模型实例化这一思想，因为其任意顺序的预测接口能够自然地展现对部分填充解状态的预测。我们的关键观察是，平均置信度这一可从标准掩码扩散目标中获得的简单模型内部量，可以作为全局一致性的有效代理，并引导推理时的搜索与修订。在 ZebraLogic、Nurse Rostering 和 Job-Shop Scheduling 上的实验证明，在固定微调后的 LLaDA-8B-Instruct 检查点不变的情况下，Blackboard 始终能够提升推理效果，并显著优于同等规模的自回归基线：在 ZebraLogic-Hard 上达到 90.4% 的准确率，在 Nurse Rostering 上达到 76.4% 的精确可行率，在 JSSP 上达到 80.2% 的最优率。更强的自回归搜索与细化方法同样无法缩小其在 ZebraLogic-Hard 上的差距；尽管测试的前沿 LLM 规模大得多且具备强大的测试时推理能力，Blackboard 仍在 ZebraLogic-Hard 和 JSSP 上超越了它们。我们已在 https://github.com/jwoosang1/blackboard-intelligence 开源代码库。
arXiv:2609.38767v2 公告类型：替换 摘要：训练数据归因（TDA）用于估计单个训练样本对模型输出的贡献。大多数可扩展的 TDA 方法依赖逐样本梯度，而在 LLM 规模下计算和使用此类梯度，会在效率、兼容性和可扩展性方面带来挑战。我们推出 dattri-LLM，这是一个让基于梯度的归因在大规模场景下更具实用性的 TDA 库。在效率方面，dattri-LLM 使用紧凑的梯度表示，并根据成本模型动态路由梯度操作。在兼容性方面，其捕获机制能够从调用 backward() 的现有训练循环中收集逐样本梯度，而无需修改训练循环或其配置。这包括使用 DDP 和 FSDP 的分布式训练，以及基于 HuggingFace Transformers、TRL 和 OLMo 构建的训练流程。在可扩展性方面，dattri-LLM 提供可复用的梯度操作和训练时回调函数，用于实现归因方法和应用。这些接口支持多种归因方法，包括梯度相似度、基于曲率的影响力分析和基于轨迹的方法，以及在训练期间对梯度执行操作的应用，例如在线数据选择。在相同硬件和工作负载下，dattri-LLM 的吞吐量平均达到最快竞争库的 3.2 倍；它能够将多种归因方法扩展到 4 个 H200 GPU 上的 110B 参数模型，并在不同模型系列和规模的一系列模型中实现更优的归因保真度与成本权衡。dattri-LLM 的源代码位于 https://github.com/TRAIS-Lab/dattri-llm。
arXiv:2609.36812v2 Announce Type: replace Abstract: Diffusion and flow policies can model complex behaviors in offline reinforcement learning (RL). However, penalizing their KL divergence from the behavior policy can discourage actions having high critic values with low behavior density. Directly refining behavior proposals may be an alternative, yet Gaussian or deterministic editors limit expressiveness to represent multiple separated modes for the same proposal. In this work, we introduce Proposal-Conditioned Refinement Flows (PReFlow), a policy extraction method combining critic-based proposal selection with a conditional refinement flow. To optimize proposal selection and refinement together, we formulate a KL-regularized objective whose optimum induces a Gibbs policy over final actions under a Gaussian-smoothed behavior prior. The refinement flow can represent multiple high value modes, while a proposal-centered Gaussian reference regulates large action changes. This Gaussian reference further enables us to make use of simulation-free, closed form adjoint matching targets from sampled endpoints and critic gradients, yielding a single velocity regression loss without a backward adjoint solve. On 50 OGBench tasks, PReFlow achieves competitive offline performance and the highest aggregate score among the compared methods after online fine-tuning, reaching 91\% after 500K environment steps.
arXiv:2609.36081v2 公告类型：替换 摘要：随着网络学习新任务，表征会不断变化。我们探究早期表征变化是否会自然形成一种几何结构，并持续塑造后续学习。我们识别出早期表征漂移的一个低维子空间，将其称为支架，并检验它是否会在后续任务中被复用。在四种预训练视觉编码器和两个数据集上，后续表征变化始终更倾向于沿这一早期确定的子空间发生，而非匹配的随机备选子空间。这种复用依赖于学习历史：当网络经历不同的早期任务，但后续训练输入相同时，每个网络都会优先复用由其自身学习历史形成的支架。同样的偏好也出现在优化器的单次更新中，即使网络的主导局部响应方向已偏离原始支架。最后，相比匹配的随机约束，对支架内的变化施加约束会更明显地减缓新任务的学习，但对旧任务保持的影响则不够一致。总之，这些结果表明，早期经验会在神经网络适应未来任务的方式上留下持久的几何印记。代码可在 https://github.com/YuantaoDeng/latent-scaffold 获取。
arXiv:2609.28607v3 公告类型：替换 摘要：间歇性时间序列常见于备件需求和零售销售。由于预测误差的成本通常是不对称的，库存控制等决策需要完整的预测分布，而非点预测。人们已经提出了许多概率预测方法；然而，这些方法的实现分散在不同的软件框架中，因此难以对其进行系统比较。我们推出了 R 软件包 fable$.$intermittent，该软件包在 fable 框架内实现了多种适用于间歇性序列的概率预测方法。该软件包支持通过单一且简洁的预测流程，对一组时间序列拟合和评估多个模型。我们还介绍了一种新的指数平滑模型 TWEES，该模型采用 Tweedie 预测分布。拟合 TWEES 需要反复计算计算开销较大的 Tweedie 密度。我们还发布了 R 软件包 tweedieDistr，其对 Tweedie 分布的实现速度显著快于现有实现，同时保持相同的数值精度。我们在四个数据集上评估了 fable$.$intermittent 中实现的方法，这些数据集也已随该软件包一同发布。
arXiv:2609.13529v2 Announce Type: replace Abstract: As the use of Large Language Models moves from chatbots into agentic systems, where outputs become actions with irreversible consequences on reality, the existing paradigm on AI Interpretability research, post-hoc interpretability, is structurally inadequate for safe and trustworthy model deployment: it explains behavior after the fact but cannot audit or intervene in an inference computation before it commits to an output. We therefore argue for a shift toward \emph{generative interpretability}, an architectural property under which a model's inference pass natively exposes semantically meaningful checkpoints that are human-understandable and amenable to causal intervention. We show the merits of generative interpretability as comparison to other interpretability research paradigms, and propose Neuro-Symbolic Models as a concrete instantiation.
arXiv:2607.05378v2 公告类型：替换 摘要：长时程智能体LLM日益受到有限上下文窗口的限制，因为在任务完成前，延长的交互轨迹可能会超过最大上下文长度。上下文压缩通过总结先前的交互状态，并在压缩后的上下文中继续执行轨迹，提供了一种自然的解决方案；然而，将压缩机制融入强化学习仍缺乏充分研究。我们提出CompactionRL，一种利用上下文压缩训练长时程智能体LLM的强化学习策略。我们的方法通过令牌级损失归一化和跨片段广义优势估计，联合优化任务执行与摘要生成。这一设计使LLM智能体能够从经过压缩的长时程轨迹中学习。我们在开源模型上训练CompactionRL，并在智能体编程任务中观察到持续的性能提升。CompactionRL使开源GLM-4.5-Air模型（106B-A12B）在SWE-bench Verified和Terminal-Bench 2.0上分别取得66.4%和26.2%的Pass@1得分；与推理时采用上下文压缩的基础模型相比，分别提升了6.6和4.9个百分点。基于GLM-4.7-Flash（30B-A3B），CompactionRL相较基础模型使Pass@1分别提升5.5和6.7个百分点，在SWE-bench Verified和Terminal-Bench 2.0上分别达到56.0%和20.2%。因此，CompactionRL已部署于强化学习流程中，用于训练开源GLM-5.2模型（750B-A40B）。
arXiv:2607.04535v2 Announce Type: replace Abstract: Orthogonal and Stiefel layers give neural weights exact spectral control, but they also impose a strong modeling constraint: all represented singular values are fixed at one. Many settings that benefit from an orthonormal basis still need direction-dependent attenuation or amplification. We introduce ManifoldFlow, a minimal relaxation of a fixed-spectrum Stiefel layer that keeps the basis on the Stiefel manifold while learning a bounded positive spectrum through W = Q S^{1/2}, with Q^T Q = I and S positive definite. Since W^T W = S, the eigenvalues of S are exactly the squared singular values of the realized weight, making eigenvalue clipping a direct singular-value control mechanism. Across paired sequence, tabular, and image experiments, the learnable SPD spectrum improves the fixed-spectrum Stiefel counterpart in the reported settings where the Stiefel prior is useful, with the largest gains in recurrent language-model projections. Boundary cases in convolutional classifier heads clarify the intended scope: ManifoldFlow is not a universal dense-layer replacement, but a spectrum-learnable Stiefel relaxation for settings where an orthonormal basis is a useful prior. When the basis should be orthonormal, its spectrum need not be frozen. Code available at https://github.com/Hik289/manifold_flow
arXiv:2606.19138v2 发布类型：替换 摘要：神经控制微分方程（NCDE）为时间序列预测提供了一个强大的连续时间框架，但基于图的标准扩展通常完全从数据中学习空间结构，即使在先验已知有向图结构的情况下也是如此。我们提出 Informed Neural controlled Differential EQuationS（INDEQS），这是对基于图的 NCDE 预测方法的一种改进，可在不同的架构位置融入有向图的先验知识。INDEQS 将图节点之间隐藏状态的内部混合与向量场和控制之间的外部混合分离开来，并提供了一个轻量级的图约束变体和一个表达能力更强的变体；后者通过自适应图卷积从数据中学习额外的图连接。为系统研究图先验信息在何种情况下有利于预测，我们在有向图上设计了连续平流模拟，生成具有已知真实流动结构的合成时空数据集。随后，我们在两个现实世界任务上评估 INDEQS：水文网络中的河流流量预测，以及 PeMS08 上的交通流量预测。在合成数据和河流流量任务中，与参数量相当的无先验 NCDE 相比，外部先验信息始终能够降低平均绝对误差，尤其是在较大的图上；而当需要严格遵循已知邻接关系时，内部先验信息则提供了参数效率更高的替代方案。对离散卷积解码器和连续时间解码器的比较进一步表明，在现实世界任务中，连续解码器能够实现更高的准确率和更强的时间灵活性。INDEQS 和该平流模拟的实现可见于 https://github.com/mitchi1/indeqs 。
arXiv:2606.14597v2 公告类型：替换 摘要：基于Transformer的神经算子在逼近复杂几何上的偏微分方程解算子方面展现出了卓越性能。然而，现有方法隐含地假设空间域大小固定，这限制了其在推理阶段的泛化能力。在本文中，我们研究空间域扩展问题，即在显著大于训练期间所见空间域的空间域上进行零样本推理。我们认为，这种设置从根本上要求具备空间局部性和平移等变性。为此，我们在注意力logits计算中引入一种可分解偏置来实现局部性，从而实现精细可控的局部性，同时仍可完全分解为查询-键内积，并能直接兼容经过优化的注意力内核。结合旋转位置嵌入，该方法能够在不改变Transformer架构的情况下，实现具有可控空间支持范围的高表达能力嵌入。我们通过两个偏微分方程基准测试和一个3D工业大气流应用的实验证明，我们的方法显著提升了对更大空间域的零样本泛化能力。我们的代码和数据集已发布于 https://github.com/cerea-daml/domain-extension。
arXiv:2606.07631v2 公告类型：替换 摘要：当狭窄领域的微调诱发超出微调任务范围的危险行为时，就会出现涌现失配（EM）。通过反复进行行为评估来检测这种变化成本高昂，因此我们提出从内部表示出发，在检查点级别进行监测。我们基于七个与对齐相关的激活方向定义了一个固定坐标系，并利用它跟踪四个开源 7-9B 语言模型在 LoRA 微调期间的表示漂移。在该空间中，微调漂移呈现出一个主导轴，可解释 78.6% 的方差，并且在不同数据集、提取方案和参数更新容量下保持稳定。在来自三个与 EM 相关的留出数据集的 468 个检查点上，由此得到的监测器实现了 1.8% 的 FNR、2.0% 的 FPR 和 0.989 的 AUROC，优于语义、随机、PCA 和 SAE 特征基线。在第四个数据集上，匹配的良性—危险控制实验表明，良性微调也可能发生显著的表示漂移；但 7D 画像中的变化仍能区分危险运行与良性运行。在两个 14B 模型、全量微调、更长训练周期和失配初始状态上进行的压力测试表明，该信号能够在训练配置发生变化时持续存在，但要实现可靠部署，可能需要重新校准。
arXiv:2605.20468v3 Announce Type: replace Abstract: Effective medication management in Parkinson's Disease (PD) is challenging due to heterogeneous disease progression, variable patient response, and medication side effects. While AI models can forecast levodopa equivalent daily dose (LEDD) as a measure of medication needs, standard uncertainty quantification often fails to communicate the reliability of these predictions, treating high and low confidence clinical decisions identically. We introduce CASCADE (Calibrated Adaptive Scaling via Conformal And Distributional Estimation), a novel conformal prediction framework that propagates epistemic uncertainty from a screening classifier to adapt downstream predictions. Unlike standard conformal methods that rely on auxiliary residual regression, we leverage epistemic uncertainty from a primary classification task (identifying whether a medication change is needed) to dynamically scale the prediction intervals of a secondary regression task (predicting how much change). By mapping Venn-Abers multi-probabilistic uncertainty directly to non-conformity scores, our framework achieves continuous risk adaptation. We demonstrate that this cascade effect produces highly efficient intervals for confident patients (38.9% narrower than standard conformal baselines) while automatically expanding intervals to ensure robust coverage for uncertain cases, bridging the gap between discrete clinical decision-making and continuous dose forecasting in PD.
arXiv:2605.18387v2 Announce Type: replace Abstract: Graph Neural Networks and Graph Transformers have become central to graph learning, combining expressive representation learning with sample-efficient inductive biases. Yet they remain fundamentally limited when predictions depend on correlations between distant graph regions. We address this limitation with Graph Hierarchical Recurrence (GHR), a novel framework that jointly operates on the input graph and a pooled hierarchical abstraction. We also show that existing models degrade more sharply under out-of-range generalization, where test instances require interactions across distances exceeding those observed during training. Despite its minimal design, GHR consistently strengthens every tested message-passing backbone, yielding robust performance on long-range dependencies and particularly pronounced gains in out-of-range regimes. Across a broad suite of long-range benchmarks, GHR achieves state-of-the-art or competitive results on multiple tasks, establishing hierarchical recurrence as an effective mechanism for extending graph models beyond their observed interaction range.
arXiv:2601.06701v2 Announce Type: replace Abstract: Complex AI systems make better predictions but often lack transparency, limiting trustworthiness, interpretability, and safe deployment. Common post hoc AI explainers, such as LIME, SHAP, HSIC, and SAGE, are model agnostic but are too restricted in one significant regard: they tend to misrank correlated features and require costly perturbations, which do not scale to high dimensional data. We introduce ExCIR (Explainability through Correlation Impact Ratio), a theoretically grounded, simple, and reliable metric for explaining the contribution of input features to model outputs, which remains stable and consistent under noise and sampling variations. We demonstrate that ExCIR captures dependencies arising from correlated features through a lightweight single pass formulation. Experimental evaluations on diverse datasets, including EEG, synthetic vehicular data, Digits, and Cats-Dogs, validate the effectiveness and stability of ExCIR across domains, achieving more interpretable feature explanations than existing methods while remaining computationally efficient. To this end, we further extend ExCIR with an information theoretic foundation that unifies the correlation ratio with Canonical Correlation Analysis under mutual information bounds, enabling multi output and class conditioned explainability at scale.
arXiv:2508.16748v2 公告类型：替换 摘要：当前普遍采用的多模态自监督学习（SSL）方法依赖于冗余假设：不同视图之间共享大量与任务相关的信息。我们认为，在以异质性为特征的复杂现实场景中（例如长度可变的医疗或行为数据），这一假设并不成立；在这些场景中，强制进行严格对齐可能会丢弃独特的模态特定信号，并无意中放大偏差。在本文中，我们提出 FairSSL，这是一种将数据异质性视为公平性资源而非障碍的框架。不同于标准的对比学习方法，FairSSL采用面向受试者的 Variance-Invariance-Covariance Regularization 目标函数，在同一受试者的不同片段之间施加对齐约束。我们引入基于片段的池化策略，以处理长度可变的模态，并对表征进行正则化，以促进：（i）受试者内部充分的变化性；（ii）跨模态和跨受试者不变性；以及（iii）表征去相关。理论分析表明，我们的目标函数能够界定受保护群体之间的得分差距。在实验中，FairSSL在异质多模态数据集上显著优于现有基线方法，在不牺牲下游预测性能的情况下提升了公平性。代码见：https://github.com/abtinmU/FairSSL
arXiv:2508.13408v3 发布类型：替换 摘要：化学语言模型（CLMs）正日益应用于从头药物设计，这得益于近年来模型规模、计算量和数据集规模的增长。然而，设计选择、训练动态与下游生成质量之间的关系仍缺乏充分理解。我们开展了一项计算受控的 CLMs 规模化研究，涵盖超过 30,000 项实验，涉及分子表示（SMILES、SELFIES、SAFE）、分词方式（原子级分词和字节对编码）、模型规模（0.5M–1B 参数）、经过数据泄漏控制的数据集（MOSES、ChEMBL、PubChem、ZINC-22）以及模型架构（仅解码器和编码器-解码器）。通过拟合 IsoFLOP 曲线，我们确立了预训练损失的明确规模化趋势，但发现这些改进并未转化为目标导向分子设计中同等程度的性能提升。逐层探测和稀疏自编码器分析揭示了化学表征的持续发展：化学语法较早趋于饱和，而语义属性的形成更为缓慢，并随着进一步训练和模型规模扩大而变得越来越容易获取。在所评估的协议和 oracle 预算下，这些表征能力的提升与目标导向生成性能的边际递减并存。由此得到的模型套件 NovoMolGen 取得了最先进的结果，在各项药物发现任务的目标导向分子生成中超越了以往的 CLMs 和专用生成模型。这些发现揭示了化学表征学习与下游分子设计之间的脱节，推动人们开发能够更直接学习化学语义的预训练范式。
arXiv:2506.11030v2 Announce Type: replace Abstract: Training neural networks has traditionally relied on backpropagation (BP), a gradient-based algorithm that, despite its widespread success, suffers from key limitations in both biological and hardware perspectives. These include backward error propagation by symmetric weights, non-local credit assignment, and frozen activity during backward passes. We propose Forward Target Propagation (FTP), a biologically plausible and computationally efficient alternative that replaces the backward pass with a second forward pass. FTP estimates layerwise targets using only feedforward computations, eliminating the need for symmetric feedback weights or learnable inverse functions, hence enabling modular and local learning. We evaluate FTP on fully connected networks, CNNs, and RNNs, demonstrating accuracies competitive with BP on MNIST, CIFAR10, and CIFAR100, as well as effective modeling of long-term dependencies in sequential tasks. Moreover, FTP outperforms BP under quantized low-precision and emerging hardware constraints while also demonstrating substantial efficiency gains over other biologically inspired methods such as target propagation variants and forward-only learning algorithms. With its minimal computational overhead, forward-only nature, and hardware compatibility, FTP provides a promising direction for energy-efficient on-device learning and neuromorphic computing.
arXiv:2410.23660v2 公告类型：替换 摘要：联邦学习（FL）是一种利用去中心化数据进行模型协同训练的学习范式。近年来，研究表明，在 FL 中使用预训练权重初始化能够有效提升模型性能。然而，当前预训练模型的复杂性不断增加，其参数量大幅增长，显著加剧了将这些模型适配到 FL 所需通信轮次方面的挑战。为了解决通信成本问题并提升预训练模型在 FL 中的适配性能，我们提出了一种创新的、基于模型插值的本地训练技术，称为“Local Superior Soups”。我们的方法通过正则化模型插值增强不同客户端上的本地训练，在较少通信轮次内推动模型探索一个连通的低损失盆地。该方法可作为预训练模型在 FL 中实现无缝适配的催化剂。我们在多个广泛使用的 FL 数据集上验证了其有效性和效率。我们的代码可在 https://github.com/ubc-tea/Local-Superior-Soups 获取。
arXiv:2610.02207v1 公告类型：cross 摘要：3D 高斯虚拟形象支持快速渲染，但其实时动画通常受到高成本神经网络推理的限制。我们针对这一瓶颈展开研究，发现预训练虚拟形象模型的动画可以通过与身份无关的混合形状线性组合来近似。基于这一发现，我们提出 GALA（Gaussian Animation via Linear Approximation），一种蒸馏方法，用浅层系数预测器和线性混合替代逐帧的复杂神经解码。为提升保真度并降低内存需求，我们在渲染感知度量和内存预算约束下，利用局部块 PCA 构建基底。我们训练了一个浅层 MLP 网络来预测混合形状系数，并且无需重新训练原始模型，即可应用于各种动画架构。我们通过加速三种不同虚拟形象模型的推理来验证 GALA，这些模型用于实现带有面部表情和服装动态的 3D 动画以及全身动画。在这些模型上，我们的蒸馏方法能够泛化到未参与训练的身份，并在保留大部分渲染质量的同时，将 CPU 动画成本最高降低三个数量级。该方法取得的出色结果证实了学习型虚拟形象表示中存在共享的线性结构，并支持在移动设备上以最高达 60fps 的帧率实现高效且准确的动画。项目页面：https://ramazan793.github.io/gala/
arXiv:2610.02193v1 公布类型：cross 摘要：离散扩散语言模型为自回归生成提供了一种颇具吸引力的替代方案，适用于需要双向推理和全局约束满足的任务。然而，它们存在一个结构性瓶颈：并行解码时，每个词元都独立地从其边缘分布中采样，从而切断了同时解码的词元之间的统计依赖关系。连续扩散语言模型通过对共享的连续状态进行去噪来避免这一问题，但其去噪器只能看到该状态，因此在最终解码之前，没有任何机制将其与有效的词元配置联系起来。为解决这一问题，我们提出分层连续扩散语言模型（HC-DLM）。该模型在单一且具有原则性的去噪过程中，将离散词元生成与连续潜在轨迹相耦合，其训练目标源自词元似然的变分界。不同于近期将连续上下文附加到自包含离散链上的方法，HC-DLM 将潜变量作为唯一持续存在的生成状态：每一步都从中读出词元，并将其作为下一次潜变量更新的支架反馈回去。在结构化推理（Sudoku）、数学规划（Countdown）和语言建模（LM1B）任务上，在模型规模相当的情况下，HC-DLM 均优于离散和连续扩散基线：在 Sudoku 和 Countdown 上取得更高的谜题准确率，在 LM1B 上获得更优的生成困惑度。项目主页：https://hc-dlm.github.io/。
arXiv:2610.02117v1 公告类型：cross 摘要：在策略自蒸馏近期已成为提升语言模型推理能力的一种有效方法，其通过接收特权信息的冻结版本或 EMA 版本的自身模型来监督学生模型。然而，将其应用于多模态大语言模型（MLLMs）仍基本处于探索阶段。近期方法利用特权视觉信息（例如与问题对应的图像裁剪区域）来提升细粒度感知能力，但其增益仅限于受益于此类视觉放大的任务，并且需要人工标注的定位数据或外部教师模型。我们为 MLLMs 引入了一种不同形式的在策略自蒸馏方法：为教师模型提供文本形式、具有空间定位依据的指导，以识别与查询相关的视觉元素。我们使用程序生成的场景，其中自动提供物体标识和空间坐标，从而实现可扩展且无需标注的后训练。教师模型利用这些空间指导，在多个相关图像区域中定位并整合证据；学生模型则仅凭图像和问题学习复现这一行为。我们的方法在多个模型上的计数、文档理解和图表理解基准测试中均能稳定提升性能。重要的是，尽管后训练仅使用合成场景，所得改进仍能迁移到真实世界感知基准，在 CVBench、V*、ZoomBench、BLINK、HR-Bench 和 MME-RealWorld 上的平均性能提升了 3.23 个百分点。这些结果表明，具有空间定位依据的特权信息能够通过在策略自蒸馏诱导出更广泛的感知能力，使模型实现显著的从合成数据到真实数据的迁移，且超越后训练所使用的任务和数据分布。项目页面：https://github.com/sirkosophia/Where-OPD
arXiv:2610.02068v1 Announce Type: cross Abstract: How complex can the responses of a quantum device become as it runs longer with a fixed internal memory? We quantify this complexity through sequential response capacity: how many adaptive testing stages, each using a fresh run, can continue to separate possible processes by a prescribed gap in response probabilities. For fixed system and memory sizes, we establish a tight law relating this capacity to run length and probability resolution. At fixed resolution, the capacity grows on the order of $K\log K$, where $K$ is the number of time steps in each run. Our construction attains this growth using time-dependent phase rotations on a single visible qubit with no additional internal memory; its tests give response probabilities exactly zero or one. Under the same tests, classical stochastic processes that measure in a fixed basis at every step have only linear capacity at fixed sizes and resolution. For phase sequences selected by a stored classical label, we then quantify how known independent Pauli noise changes this logarithmic enhancement. With ideal controls and weak residual phase noise after correction, we prove matching capacity bounds at a fixed small probability gap. These bounds identify the inverse residual phase-flip probability as the coherence timescale that limits the extra logarithmic growth.
arXiv:2610.01949v1 Announce Type: cross Abstract: Ransomware has emerged as a major cybersecurity threat, with incidents increasing in frequency and impact across critical sectors. These attacks are typically launched through phishing emails, malicious downloads, or exploitation of software vulnerabilities to gain system access. Once inside, the malware encrypts files and demands a ransom, often in cryptocurrency, for the decryption key. Conventional detection methods often struggle with novel or scarce samples, leaving systems vulnerable. To address these challenges, this paper proposes a hybrid deep learning framework that combines an Autoencoder Feature Extractor (AFE) with a Model Agnostic Meta Learning (MAML) classifier for few shot malware detection. The AFE generates compact latent features that reduce noise and dimensionality, while the MAML classifier rapidly adapts to new threats using limited labeled data. Experiments conducted on the Ransomware Dataset 2024 demonstrate the effectiveness of the framework in binary classification tasks. Across one to fifty shot settings, the proposed model consistently achieves high accuracy, F1 score, and Matthews Correlation Coefficient values, maintaining reliable classification even under extreme scarcity. These results highlight the model's robustness and effectiveness in adapting to limited data scenarios, demonstrating the potential of combining feature extraction with meta learning to enhance resilience against malware, particularly in sectors such as healthcare, manufacturing, and public infrastructure, where cyberattacks can cause significant operational and financial disruption.
arXiv:2610.01807v1 发布类型：cross 摘要：扫描仪、采集地点和采集协议之间的分布偏移阻碍了深度医学图像模型的可靠临床部署。现有的域泛化（DG）方法通常侧重于风格或强度多样化，但仍可能使网络依赖于特定域的纹理相关性。受傅里叶相位编码语义结构这一证据的启发，我们提出了 PhaseAT，一种面向医学域泛化的相位感知对抗训练框架。PhaseAT 在傅里叶域中构建相位扰动的训练视图，通过迭代更新有界的相位扰动，同时保持幅度谱不变，从而在外观统计相匹配的情况下，强化对空间组织结构的学习。为避免产生色彩伪影，扰动仅应用于 YCbCr 色彩空间中的亮度通道。此外，我们设计了一个简单的相位显著性掩码，将更新集中于影响最大的频率上。模型通过对干净样本和相位扰动样本上的损失进行加权组合来训练，同时支持单源和多源域泛化。我们在两个具有挑战性的医学数据集上验证了该方法，结果表明，PhaseAT 在单源域泛化方面取得了超过 20% 的提升，优于多种最先进的 DG 方法。代码实现地址：https://github.com/ahmed-sharshar/PhaseAT。
arXiv:2610.01746v1 Announce Type: cross Abstract: Autonomous driving systems have become a central focus of intelligent transportation research, with End-to-End Learning and Modular Architectures offering two prominent design paradigms for their implementation. E2E Learning uses deep learning algorithms to map raw sensory inputs directly to driving actuators, providing a streamlined and adaptable solution. while Modular Architectures employ a pipeline-based approach, dividing the system into distinct subsystems for perception, cognition, planning, and control. This paper presents a comprehensive comparative analysis of these paradigms, focusing on their strengths, limitations, and trade-offs to provide insights into their suitability for various autonomous driving applications. The study evaluates key factors such as interpretability, scalability, robustness, and real-world applicability. While End-to-End Learning emphasizes simplicity and adaptability in dynamic environments, it lacks transparency and is highly dependent on large datasets. Conversely, Modular Architectures offer superior interpretability and task-specific optimization, but face challenges related to integration complexity and scalability. To address these limitations, hybrid approaches that combine the strengths of both paradigms have emerged, offering a promising direction for overcoming these challenges. Beyond this comparative synthesis, following work proposes a Four-Dimensional Architecture Selection Framework, comprising twelve binary criteria across safety, operating environment, data/computational resources, and deployment context, and validate it against ten published autonomous driving systems, correctly recommending 7/10 deployed architectures. This work synthesizes existing literature to highlight key trade-offs between the paradigms and identifies hybrid architectures as a promising direction for future research.
arXiv:2610.01647v1 Announce Type: cross Abstract: In this article, we present our vision and recent advancements toward creating a decentralized system capable of learning from real-time data within buildings to support sustainable and privacy-preserving smart environments. Our approach promotes the concept of the building itself as the data center, aligning with the principles of edge computing to safeguard confidentiality and reduce reliance on external cloud infrastructure. This is particularly valuable in humanitarian contexts, where data sovereignty, energy efficiency, and infrastructure constraints are critical. We detail a lightweight, "Kubernetes-like" orchestration framework for deploying AI services within such environments and demonstrate our progress in implementing AI algorithms on low-power, cost-effective microcontrollers such as those in the Arduino ecosystem. By enabling in-situ learning directly on sensors or microcontrollers, our work aims to bring intelligent services to resource-limited settings, fostering autonomy, resilience, and sustainable development in vulnerable or underserved communities. The contributions in this article are related, firstly, to our project "Online Machine Learning Algorithms for Embedded Systems" and the evaluation of two new online algorithms. Secondly, we envision a cloud-fog-edge architecture based on the KOptim and FIWARE components, and we propose a methodology for coupling them. Experimental results of the online algorithms are also presented, showcasing real-world traces.
arXiv:2610.01559v1 Announce Type: cross Abstract: World Action Models (WAMs) predict visual futures and robot actions, yet they remain susceptible to task-incomplete imagination, where plausible, action-consistent predictions omit the transition needed for task completion. In this paper, we show that this failure is not inherent to the world model backbone, but emerges when adapted for short-chunk control, which can repeatedly favor plausible local continuations over task-completing transitions. To address this, we introduce Completion Aware Guidance (CAG), a training-free sampling method that guides generation toward task completion. Across representative WAMs, CAG improves success from 64% to 70% on a RoboTwin 2.0 subset and from 69% to 75% in zero-shot simulation, while reducing task-incomplete imagination from 79% to 40%.
arXiv:2610.01497v1 公告类型：cross 摘要：分子肿瘤委员会整合基因组学发现、临床背景和治疗证据，以支持精准肿瘤学。随着人工智能进入这一工作流程，一个关键的安全挑战是区分真正缺乏支持的建议，与那些有证据支持但由于信息不完整、ECOG体能状态较差或其他临床注意事项而仍需肿瘤科医生审核的选项。我们推出了OpenMTB-Audit，这是一个开源基准，包含500个合成的非小细胞肺癌病例，涵盖五类对抗性错误和四种安全标签：受支持、部分受支持、不受支持和信息不足。在八种大语言模型配置中，我们发现了普遍存在的过度拒答现象：对于所有真实属于“部分受支持”的病例，所有大语言模型配置均有83.3%至100%的概率无法保留“部分受支持”标签，最终通过标签塌缩而非经过临床校准的推理获得较高的总体安全评分。为解决这一局限，我们开发了MTB-AuditAgent，这是一个确定性的七模块框架，将证据核验、缺失信息检测、安全分类和拒答分离开来。该框架将过度拒答率降至6.7%，并达到91.2%的准确率（95%置信区间：88.6%–93.6%）。一项由两名肿瘤科医生参与的标注研究发现，分歧主要集中在信息是否充分与治疗优化之间的边界上，凸显了保留具有临床意义的区分的必要性。
arXiv:2610.01413v1 Announce Type: cross Abstract: Reinforcement learning (RL) algorithms frequently compare probability distributions, such as state visitation distributions induced by policies and experts, action distributions from learned policies and offline datasets, or transition distributions from learned models and environments. However, commonly used divergences may become ineffective when these distributions overlap weakly, which is frequently encountered in imitation learning, offline RL, and deployment under distribution shift. Optimal transport (OT) offers an alternative by measuring the cost of \emph{moving} probability mass from one distribution to another under a ground cost that encodes task geometry. This survey covers how OT is used inside RL objectives and algorithms. For each method, we identify: the role OT plays, the distributions compared, the OT formulation used, and the treatment of temporal structure. Beyond categorising existing methods, we discuss the motivations behind different OT choices, practical considerations such as cost design and computational challenges, and highlight open problems including scalable trajectory-level transport, principled handling of mass mismatch, and theoretical analysis for OT-regularised RL.
arXiv:2610.01408v1 公告类型：cross 摘要：轨迹交叉仍然是 Flow Matching（FM）中的关键瓶颈，而以往工作通常从理论优化的角度看待这些交叉，认为其会导致速度平均化。它们尝试通过事后蒸馏或端点耦合间接解决这一问题，却没有显式调控中间轨迹。在本文中，我们提出了一种新的网络学习视角：交叉点会在目标速度场中固有地产生较大的局部 Lipschitz 常数，从而带来两个问题。首先，较高的 Lipschitz 常数对应于速度场中的高频信号，而由于谱偏置，神经网络难以拟合这类信号。其次，这也意味着速度会发生剧烈变化，导致少步推理中的数值积分误差严重。为缓解这一问题，我们提出 CoFlow，这是一种将对比学习范式引入 FM 的框架，在训练过程中显式排斥轨迹，从而降低速度场的局部 Lipschitz 常数。具体而言，我们从随机微分方程（SDE）的角度构建 CoFlow，通过注入排斥漂移项，使正样本的前向过程主动远离负轨迹，从而有效降低局部 Lipschitz 常数。此外，我们还从该 SDE 推导出一种等价的随机插值形式，为控制负样本的影响提供了简单且易于处理的设计空间。在 ImageNet 256x256 上进行的大量实验表明，在少步推理（例如 20 步）中，CoFlow 相较于标准 FM 能显著降低 FID，且不会增加训练开销。代码可访问：https://github.com/HKUST-LongGroup/CoFlow
arXiv:2610.01361v1 Announce Type: cross Abstract: Multilayer networks allow the modeling of interactions between the same entities across different contexts, such as temporal observations, varying settings, or interactions of different types. The goal of community detection in multilayer networks is to identify groups of nodes exhibiting similar connectivity patterns, which may vary across layers. We propose a method based on a joint nonnegative symmetric matrix trifactorization for community detection in multilayer networks, where each graph is approximated by a nonnegative symmetric matrix trifactorization. Our approach enforces constraints on the factor matrices so that communities are disjoint and shared across layers, while allowing each layer to have its own connectivity patterns and node degrees. This flexibility enables the model to capture both local and global structural variations across layers. We also develop an algorithm to efficiently solve this problem. We evaluate multilayer community detection methods using the multilayer degree-corrected stochastic block model (MDCBM), a flexible framework for generating realistic multilayer graphs with heterogeneous degrees and varying connectivity patterns. Experiments show that our method reliably detects communities across diverse regimes, whereas existing state-of-the-art approaches are often limited by restrictive structural assumptions.
arXiv:2610.01301v1 公告类型：cross 摘要：当前大多数抓取合成系统都经过离线训练，并在部署期间保持不变。当部署条件与训练数据相似时，这种方式表现良好；但当机器人遇到训练中未见过的条件（例如陌生物体）时，性能可能会下降。在本文中，我们提出了一种用于杂乱场景中平行夹爪单视角 6-DoF 抓取合成的持续学习框架。我们的方法不对大型参数化模型进行微调，而是通过学习到的嵌入空间中的记忆进行适应：抓取结果会更新未来的抓取评分，而可选的用户示范则会被召回，并作为额外的候选抓取迁移到新场景中。我们在仿真环境以及大规模真实世界实验中评估了该方法，实验包含超过 1500 次抓取试验。结果表明，即使在适应之前，我们的方法也能达到现有 6-DoF 抓取基线的性能；对于训练期间缺失或代表性不足类别中的未见物体，该方法能够在线提升性能；同时还支持遗忘有限的长时程持续学习。在真实世界实验中，仅经过 50 次在线抓取尝试后，我们的方法在多个具有挑战性的物体类别上达到了超过 90\% 的成功率。视频和代码见 https://giuschio.github.io/cl_grasping/。
arXiv:2610.01260v1 公告类型：cross 摘要：四足机器人运动需要在指令跟踪、稳定性和能源效率等相互冲突的目标之间取得平衡，然而传统强化学习（RL）在训练时会将这些优先级硬编码为固定的标量奖励。我们提出 PROMO（Preference-Conditioned Multi-Objective Reinforcement Learning），一种语义多目标方法，将这种权衡作为单一运动策略在运行时的显式输入。PROMO 根据部署时面对的偏好对策略进行条件化，同时保持特定于本体的运动先验不变，从而将操作员意图与实现可行步态生成所需的奖励塑形项分离开来。与固定目标控制器、多目标基线方法以及独立训练的专家策略相比，PROMO 仅凭一个可部署策略即可实现目标专门化和鲁棒性。在仿真中采样的 100 种偏好下，按照精确 Pareto 支配关系，有 67 种行为属于非支配行为，偏好与目标之间的平均相关系数为 0.843，表明其具有广泛的 Pareto 覆盖范围和可预测的偏好响应。同一策略无需额外训练即可零样本迁移至 Unitree Go2；相较于平衡偏好，仅改变偏好就能使单位能耗最高降低 30.4%、位置误差最高降低 38.7%，以及机身姿态偏差峰值最高降低 59.0%。这些结果表明，偏好条件多目标 RL 可作为自适应腿式运动的一种实用运行时接口，其作用超越了离线构建 Pareto 集。开源代码和视频见 https://amrmousa.com/promo/。
arXiv:2610.01141v1 Announce Type: cross Abstract: Morohoshi, Nakayama, Manabe, and Mitarai proposed a physically motivated quantum machine learning problem in which the goal is to predict quantities of the form $\operatorname{Tr}[f(H)\rho]$ from classical descriptions of a Hamiltonian $H$ and a quantum state $\rho$, where $f$ is an unknown function. We call this problem Hamiltonian function learning in this paper. They constructed an efficient quantum learning algorithm under suitable conditions, while leaving a rigorous proof of average-case classical hardness open. In this paper, we rigorously prove the average-case classical hardness for two distribution-specific Hamiltonian function learning problems for $f_{\cos,\pi}(\lambda)=\cos(\pi\lambda)$ and $f_{\exp,\beta}(\lambda)=e^{-\beta\lambda}$ discussed in the paper of Morohoshi et al. under the assumption of the average-case hardness of factoring random RSA moduli. More specifically, we show that an efficient classical randomized learner under squared loss whose output hypotheses are evaluable in classical polynomial time for either problem would yield a classical randomized polynomial-time algorithm for factoring random RSA moduli.
arXiv:2610.01054v1 公告类型：交叉发布 摘要：上下文学习（ICL）使语言模型能够通过示例执行新任务，而无需更新权重。然而，每次 ICL 推理都需要处理完整的示例集，导致部署效率低下，而且 ICL 的运行机制尚未得到充分理解。先前的研究将 ICL 压缩为从特定层或位置提取的固定激活向量，但这种与输入无关的干预在输出依赖于与输入的细粒度交互的复杂任务上会失效。通过分析 ICL 的前向传播过程，我们表明，每个注意力头的输出都是其上下文被屏蔽时对应输出的仿射变换，而且实验表明，对于给定任务，该变换的参数在不同样本间保持稳定。在此基础上，我们提出 Task Operator（TO），将这一变换转化为通过解析推导得到的注意力输出投影更新，从而复现该变换。在词汇、算法和推理任务上，与先前方法相比，TO 取得了最佳的整体性能，并显著缩小了零样本推理与 ICL 之间的差距。我们进一步表明，提取的知识集中在一个跨层、跨位置的任务特定稀疏回路中，而且对来自互不重叠的示例批次的算子取平均，能够在不扩展上下文窗口的情况下有效扩展至多样本学习。我们的代码已公开，地址为 https://github.com/gzxiong/task_operator。
arXiv:2610.01006v1 公告类型：cross 摘要：决策模型会返回用于路由、拒答和自动化行动的概率。校准能使这些概率在平均意义上具有实用价值，但无法说明低置信度反映的是偶然性还是知识缺失，也无法说明当模型超出其已知范围时置信度是否会下降。我们使用超过 15 个公开数据集和 6 类生成任务，对决策模型 Jev 中的这一差异进行审计，并通过成对干预，在针对同一固定项目的情况下改变所提供的信息。Jev 在熟悉的封闭选项任务上经过校准，但无法作为知识缺失的指标：在没有任何与答案相关的信息时，它会对某个显著选项赋予最高达 0.80 的概率；对于超出观测知识边界的新闻，其置信度高于准确率 0.21--0.33，而在更早月份的数据上进行重新校准也无法弥合这一差距。针对性的“是/否”问题能更清晰地揭示具体情况：某个结果是否已确定（AUROC 1.00），以及证据是否充足（0.95；同一项目上置信度为 0.85）。询问 Jev 是否知道答案，似乎能够识别虚构实体和越过边界的新闻（0.91）；但对于现实感更强的名称，或移除日期后，它相比答案不确定性并无优势。因此，黑箱知识审计需要明确控制表面线索。代码：https://github.com/Syntheme/beyond-answer-confidence。
arXiv:2610.00997v1 公告类型：cross 摘要：知识蒸馏旨在将大型语言模型中的事实知识迁移到更小的模型中，以实现高效部署。然而，教师模型可能能够回忆某个关系的一个方向，却无法生成反方向的答案。因此，从教师模型生成的答案中进行蒸馏，可能会将这种方向性局限传播给学生模型。尽管如此，同一个教师模型仍可能通过沿其掌握的方向对该答案进行评分来识别它。我们提出了方向性标签蒸馏方法：在该方法中，冻结的教师模型沿其已知方向对候选答案进行评分，并将得分最高的候选答案作为学生模型的训练目标。在关于父母及其子女的事实中，即使针对名称先验进行了调优校正，沿已知方向评分也比沿请求方向评分产生更准确的标签。在对先验进行校正的分数下，较优方向取决于事实本身而非模板；对于挖掘得到的、其中显著实体是父母而非子女的事实，较优方向会发生反转。在不提供所评估的子女正向事实的情况下，使用已知方向标签训练的学生模型，在其训练查询上的开放式准确率比使用先验校正后的反向标签训练的学生模型高出13至15个百分点。将生成的答案通过词汇相似度与固定名称列表进行匹配后，学生模型几乎能够复现所有被选中的标签。其准确率很大程度上取决于标签质量。在未经筛选的查询上，以及在检索候选答案时不插入正确答案的情况下，标签优势依然成立。我们的研究结果表明，方向性验证能够通过提供更准确的训练目标，减轻教师模型生成答案中的错误向学生模型传播的问题。代码见 https://github.com/js-lee-AI/directional-verification】【。
arXiv:2610.00926v1 发布类型：cross 摘要：自动驾驶是未来智能交通的基石技术，而端到端学习已成为一种变革性范式，通过统一的可微模型将多模态感知输入直接映射为驾驶动作。尽管端到端学习具有诸多优势，但端到端自动驾驶（E2E-AD）的最终效果取决于其训练生态系统的质量。本文全面综述了 E2E-AD 的训练方法与生态系统。我们提出了数据—策略—平台分类体系，将训练概念化为一个相互依存的系统。数据层定义可学习的内容，策略层决定学习如何与驾驶目标保持一致，平台层则支持规模化和持续演进。在这一框架下，我们综述了以数据为中心的流程、学习范式和训练基础设施方面的最新进展，并分析了它们之间的相互作用如何影响模型性能、鲁棒性和可部署性。最后，我们反思了当前的局限，并提出面向未来的愿景，强调从数据数量转向数据价值、从孤立优化转向由基础模型驱动的泛化、从静态训练转向集成训练—测试闭环，旨在构建稳健、可扩展且值得信赖的自动驾驶系统。我们在 \href{https://github.com/Jiaaqiliu/Awesome-Training-Ecosystem-for-E2E-AD}{Our Project Page} 持续维护一个不断更新的仓库，跟踪前沿文献和相关工作。
arXiv:2610.00922v1 Announce Type: cross Abstract: Gaze estimation under natural head-eye motion underpins applications from driver monitoring to human-computer interaction. Single-frame methods predict each frame independently, so consecutive outputs fluctuate as jitter. Multi-frame methods reduce this, but they learn motion implicitly inside appearance features, so the gaze trajectory is never an explicit variable. We propose EyeTAG (Eye Trajectory-Aware Gaze Estimation), a causal multi-frame framework built around an explicit first-order gaze prior: at each step it differentiates its own recent predictions and feeds the resulting trajectory back as a compact kinematic token. Because differencing is translation-invariant in gaze space, this token carries subject-invariant motion rather than personal gaze offsets. Face and eye streams supply visual evidence, fused by cross-attention and a causal Transformer decoder. EyeTAG reduces the mean angular error by about 1.0$^\circ$ on Gaze360 and performs on par with the strongest baseline on EVE (2.56$^\circ$ vs. 2.58$^\circ$). Within-model ablations, which keep the encoder and the rest of the architecture fixed and vary only the gaze history, show that the differential formulation, rather than temporal context alone, removes the systematic saccade bias that persists even with an absolute gaze-history prior. Our code is available at https://github.com/peter8366/EyeTAG.
arXiv:2610.00911v1 Announce Type: cross Abstract: We study an endogenous nonstationary stochastic bandit problem with latent linear dynamics, where actions affect both immediate rewards and the future evolution of an unobserved latent state. Rewards are bilinear in the current action and latent state, inducing history-dependent rewards and a nontrivial long-horizon planning problem. The existing explore-then-commit approach achieves $\tilde{O}(T^{2/3})$ regret by uniformly exploring to estimate the latent dynamics and then committing to an optimized open-loop action sequence. We show that this rate can be improved via adaptive block-level optimism. Our key step is a cyclic approximation: under stable dynamics, the infinite-memory reward process can be truncated, and the open-loop benchmark can be approximated by optimizing a finite-memory block-level proxy. Building on this reduction, we propose a UCB-based block algorithm that maintains confidence sets for the truncated dynamics parameters and selects blocks optimistically. We prove a regret bound of order $\tilde{O}(\sqrt T)$, significantly improving over the previous $\tilde{O}(T^{2/3})$ guarantee for the same model. To the best of our knowledge, this is the first $\tilde{O}(\sqrt T)$ regret guarantee for latent linear-dynamics bandits with bilinear reward observations and an open-loop action-sequence benchmark.
arXiv:2610.00678v1 公告类型：cross 摘要：组织学全切片图像（WSI）是计算病理学的核心，但由于其极高的分辨率，单张切片通常可达数GB，因此带来了严峻的计算挑战。为了实现可扩展学习，现有方法采用自监督数据凝缩来降低计算成本，但通常依赖启发式原型学习，且没有显式保留下游任务所需的特征分布。对此，我们将WSI凝缩系统地重新表述为固定表征视角下的分布匹配问题，并开发了NICER这一基于具有切片自适应容量的非参数先验的可处理近似框架。在五个组织病理学数据集上的实验，以及由一名获得专业委员会认证的病理学家进行的临床评估表明，NICER始终优于以往方法，平均准确率提升7.44%，同时实现了更优的效率—准确率权衡，凸显了基于原理、感知分布的数据凝缩对于可扩展组织学表征学习的优势。源代码见 https://github.com/nmduonggg/NICER。
arXiv:2610.00649v1 Announce Type: cross Abstract: Synthetic speech detection is critical for audio security, but performance can degrade when labeled data are scarce and evaluation conditions differ from training. This study examines quantum kernel methods and lightweight neural models for cross-corpus audio deepfake detection under limited training data. We compare a Quantum Support Vector Machine (QSVM), a classical support vector machine (SVM), and a multilayer perceptron (MLP), all trained on frozen wav2vec 2.0 embeddings using a strict budget of 200 training samples. To match the qubit budget of near-term quantum hardware, embeddings are reduced to four dimensions using principal component analysis, and all models use the same reduced features. Experiments on ASVspoof 2019, ASVspoof 5, the ADD 2023 Challenge, and the In-the-Wild dataset show that under severe domain shift from ASVspoof 2019 to ADD 2023, the MLP degrades to near-random performance, with an area under the curve of approximately 50% and an equal error rate of 50.0%. In contrast, the QSVM maintains meaningful discrimination, achieving an area under the curve of 76.0% and an equal error rate of 27.0%. This advantage is not consistent across transfer directions. When trained on ADD 2023, the QSVM falls below chance on two of three transfers, while the MLP performs better. These results suggest that quantum kernel methods can be competitive under severe cross-corpus shifts and strict low-resource constraints, but do not provide a consistent advantage under near-domain transfer. We interpret these findings as an empirical characterization of quantum kernel inductive bias under distribution shift, rather than evidence of quantum advantage, since the four-qubit kernel can be simulated exactly on classical hardware.
arXiv:2610.00524v1 发布类型：cross 摘要：视觉-语言-动作（VLA）模型通常难以泛化到其微调示例中未出现的技能组合，即使每个组成技能都曾被演示过。我们聚焦于一种视觉捷径现象：在微调过程中，视觉观测可以作为指令的替代信号，因此策略可能会执行与相似观测相关联的已演示组合，而不是执行指令所指定的组合。这促使我们使用反事实对进行训练：固定一个演示观测，同时改变指令，使其指定一个未演示的组合。然而，这些反事实对缺少相应的演示动作目标。关键在于，当前所需的技能已经被演示过，但这些执行中的动作不能直接作为目标，因为同一技能在不同观测下可能需要不同的动作。我们提出 CRAFT，利用可在同一技能的不同执行之间复用的技能表示，将对所需技能的演示执行所提供的监督迁移到反事实对上。在三个 VLA 模型和两个仿真基准测试中，CRAFT 提高了未演示组合上的成功率，同时保持了演示组合上的高成功率；它还提升了真实机器人上的组合泛化能力。项目网站：https://taegeunyang.github.io/craft/
arXiv:2610.00424v1 Announce Type: cross Abstract: Knowing how much a causal predictor could improve need not reveal the gain of the repair actually learned. We quantify this gap in a scalar Gaussian causal experiment with known intervention geometry: auxiliary data identify effect magnitude up to bounded contamination, while diagnostics identify direction. The target is the squared-loss gain of the realized trained repair relative to a fitted reference. Jointly optimizing the learner and assessor under uniform learning MSE $\eta$ avoids the trivial solution of making no repair. At the usual $1/k$ learning scale, every feasible learner incurs a $k^{-2}$ assessment floor, even when oracle potential is estimable at a faster rate. In the magnitude-rich regime, we characterize a sharp leading-log frontier: the assessment exponent is $\min{\ell_k,2k\eta_k/U}$ to first relative order, where $\ell_k=\log(1/(k^2E_k))$ and $E_k$ is auxiliary precision. A diagnostic-abstention rule attains this exponent with unknown nuisance parameters. We also bound the critical allowance window and transfer the frontier to adaptive sampling by exact Gaussian simulation. Finite-grid experiments distinguish sign-tail suppression from total MSE and expose conservative finite-budget behavior. The result isolates how the assessment target changes information requirements in this experiment; it is not a general causal identifiability claim.
arXiv:2610.00416v1 公告类型：cross 摘要：随着大型语言模型能力的提升，对其进行评估变得愈发困难：基准测试可能达到饱和，公开测试集存在数据污染风险，而评估更具挑战性的任务可能需要昂贵的评分或执行基础设施。在自动提示优化（APO）中，这些挑战更加突出，因为在寻找更优提示的过程中需要反复进行评估。因此，研究 APO 需要一种评分成本低且具有确定性的基准，同时任务难度足够高、仍有改进空间，并且能够随着模型的发展不断更新。我们构建了一个由 1,118 道 Lichess 棋题组成的国际象棋基准，用于研究冻结 LLM 的 APO：我们优化其提示，而不更新模型权重。国际象棋结合了成本低廉的精确匹配评分、基于引擎对备选走法的评估，以及可持续更新且难度可调的问题来源。不同于只报告孤立测试题成功率的评估方式，该基准还将解题能力的提升与同一领域内的短局对弈展开联系起来。我们使用该基准评估了 8 个目标模型上的 6 种 APO 算法，不仅测量基线能力，还考察每个模型对优化的响应程度，以及优化后的提示能否迁移到其他模型和对弈任务中。因此，国际象棋非常适合作为 APO 基准，原因在于它：（i）具有挑战性，即使是评估中最强的模型 Gemini 3.5 Flash（用作元模型）也只能解决约 55\% 的棋题；（ii）具有区分度，能够揭示不同方法和模型之间的性能提升、性能不变和性能退化；（iii）可持续更新，通过提供新棋题降低数据污染风险，并通过调整难度在模型能力提升时保持提升空间；（iv）成本可控，完整研究的运行成本约为 \$800。我们公开了棋题、优化与评估代码，以及数据集更新脚本（https://github.com/imec-ailabs/Automatic-Prompt-Optimization-with-Chess）。
arXiv:2610.00321v1 公告类型：cross 摘要：投机解码通过低成本地草拟未来 token，并利用目标模型并行验证这些 token，从而加速大型语言模型推理。块草稿器在一次前向传播中为一整块未来 token 进行打分，但标准解码只验证得分最高的链，并丢弃其他候选。由于这些候选已经完成打分，验证更多候选只会增加目标模型的计算量，而不会增加草拟开销。我们提出 CAST（成本感知投机树），将这些候选组织成一棵树，并通过一次目标模型前向传播对其进行验证，同时保持目标模型、草稿器权重和解码规则不变。为决定树的宽度，CAST 会不断加入候选，直到下一个候选带来的预期收益不再超过其增加的验证时间。因此，树的宽度可根据一次延迟测量针对每种部署环境自适应确定，无需遍历测试不同宽度。我们在五个领域、三代 GPU 和两类模型上评估了 CAST。在其预测宽度下，CAST 在全部八种设置中都比标准链更快，最高可提速 43%。我们还发现，最佳宽度高度依赖于部署环境。在验证成本于内核边界处跃升的情况下，包含 128 个 token 的树仅比标准链快 2%，而预测宽度下的树则快 20%。此外，我们证明，在贪心解码和采样解码下，CAST 都不会改变目标输出分布。代码地址：https://github.com/js-lee-AI/CAST。
arXiv:2610.00320v1 公告类型：cross 摘要：微调可使经过对齐的大型语言模型（LLM）适应下游任务，但仅需几十个有害示例，就可能消除它们对有害请求的拒绝行为。先前研究将与安全相关的行为定位到特定层、方向和标记上，为防护提供了目标。我们检验了：成功的定位与恢复是否能够支持经受攻击变化的防御。在来自四个模型系列的六个检查点上，攻击后，有害提示与良性提示仍保持线性可分；将完整的干净隐藏状态注入受损模型后，可在一个可复现的过渡深度恢复拒绝行为。在此前层冻结防御的基础上，我们冻结截至该深度的所有层，并重复攻击。当使用一百个有害示例时，六个检查点上的拒绝率仍接近于零，且恢复过渡点均位于冻结边界之上。在第二项研究中，在四个检查点上进行仅注意力微调后，移除更新中最大的两个奇异方向即可恢复拒绝行为。在 Llama-3.1-8B 上，常规训练变化会削弱这种修复，而将更新分散开的攻击者能够使其失效。在良性 Llama 微调数据上校准的谱检测器，会漏检该检查点上大多数修复失败案例。尽管如此，当少量有害示例意外进入训练数据时，局部冻结仍有助于保持拒绝行为。这些结果表明，攻击者可以绕过通过恢复过程识别出的区域，并击败一种在多个检查点上均有效的修复方法，从而促使我们为防御自适应微调提出五项检查。代码可见于 https://github.com/js-lee-AI/refusal-relocates】【。
arXiv:2610.00315v1 Announce Type: cross Abstract: Historical Manchu documents preserve invaluable linguistic and cultural heritage, yet their digitization is hindered by severe degradations and the scarcity of paired training data. Existing document restoration methods primarily optimize pixel-level reconstruction, which can produce visually plausible results while failing to preserve the structural identity of Manchu glyphs. To address this limitation, we propose a retrieval-guided glyph-aware restoration framework that goes beyond pixel reconstruction by explicitly incorporating glyph-level structural knowledge. Our method retrieves relevant glyph exemplars to provide structural guidance during restoration and integrates this information into the reconstruction process, improving the recovery of degraded character structures under low-resource conditions. Extensive experiments on Manchu historical documents demonstrate that the proposed approach improves both image restoration quality and glyph-level fidelity compared with existing restoration methods. These results highlight the importance of incorporating character-aware structural priors for reliable restoration of low-resource historical documents.
arXiv:2610.00233v1 Announce Type: cross Abstract: When an informed adversary shares the audience of a constrained signalling channel, the signal that best protects the truth is the signal that best describes it. On 108 confirmatory items, the adversary-robust optimum aligns exactly with the salience pole from prior work. Across a 200,000-item pool, the two differ on only 2,748 items --- lying exactly where the prior salience-to-Bayes coordinate is undefined. Where defined, robustness is achieved by moving from Bayesian discrimination entirely to salience. We show this by introducing an adversary to a forced-choice task (abstracted from Deception: Murder in Hong Kong). The adversary knows the target, observes the signal, and argues for the strongest wrong answer using a persuasion budget, $\beta$. As $\beta$ grows, the optimal signal shifts from the posterior-maximizing option to the margin-maximizing one; at $\beta = 0$, the game reproduces the original oracle model with a listener temperature of $\tau = 1$. This effect is real: 18.2 percent of the pool has an optimum that shifts under a finite budget, and each item's critical budget is exact. This coincidence structurally limits empirical evaluation. Two adversary framings change the chosen option of seven language models on 30 to 77 of 108 items against an exact no-effect rate. Yet, no measurement can determine whether this movement is toward the adversary-aware optimum or toward salience, because the two options are identical. This is a structural limit, not a null result. The diagnostic check is cheap: before evaluating adversary-awareness, verify whether the robust target coincides with a heuristic target on the evaluation items.
arXiv:2610.00106v1 公告类型：cross 摘要：深度学习和大语言模型（LLMs）正迅速应用于半导体设计，推动了对训练数据集的需求。大多数工作聚焦于硬件描述语言（HDLs），而作为构建领域专用加速器的流行方法，高层次综合（HLS）设计仍然十分稀缺。现有 HLS 数据集工作主要强调人工整理或设计参数化，很少涉及高质量的基于 LLM 的生成，也很少关注代码长度、层次结构、设计空间规模、延迟、资源利用率和应用领域方面的多样性，这可能会限制模型的泛化能力。我们提出了 SyntheticHLS，这是一个利用 LLMs 生成大规模、复杂且多样化合成 HLS 数据集的框架。该框架有两个关键思想：1）一种迭代式反馈引导变异循环，利用成对的 HLS 源代码和设计空间规范，逐步将种子设计转化为更复杂、可扩展的设计；2）用于衡量 HLS 设计复杂度和设计空间可扩展性的定量指标，这些指标可作为 LLM 引导变异的可测量目标。我们系统地交叉验证了一个 HLS 结果质量（QoR）深度学习模型，该模型在常见 HLS 基准、零样本合成设计和迭代变异合成设计上进行训练和测试。合成设计能够很好地迁移到常见基准测试集，而反向迁移则无法实现。在所研究的数据集中，SyntheticHLS 的迭代变异设计提供了泛化能力最强的训练语料。对变异过程和数据集的分析表明，指标引导的变异轨迹能够持续改善目标复杂度和可扩展性，同时不会导致非目标指标退化。与零样本生成的设计相比，变异设计覆盖了明显更广泛、更多样化的设计空间。我们的框架、数据集和评估代码均已开源：https://github.com/sharc-lab/synthetic-hls。
arXiv:2610.02191v1 公告类型：新 摘要：尽管大型语言模型（LLM）已经展现出解决前沿数学问题的惊人能力，但人们仍不清楚它们是否具备支撑其解答的结构化数学理解。本文首次尝试系统研究大型语言模型中的数学理解，从诊断其不同能力，到利用这些发现改进后训练。首先，我们引入“数学原语”这一概念，用于探究结构化数学理解，并提出 \hlei{}，一种新颖的基准，用于从四个不同维度评估数学推理：发现、生成、消化和执行。其次，我们的系统性诊断表明，解答准确率掩盖了不同的能力特征；原语能够释放大量潜在的执行能力；而发现是数学推理中的主要瓶颈。我们的后训练分析进一步表明，受发现能力限制的失败尤其容易修复。最后，基于这些发现，我们提出 \abs{}，一种原语特权自蒸馏框架，有选择地将原语引导的推理迁移到学生模型中。大量实验表明，在不同模型规模和具有挑战性的基准上，\abs{} 均能持续提升数学推理能力，优于基线方法。
arXiv:2610.02190v1 公布类型：新 摘要：步长选择仍是大规模神经网络优化中的核心挑战；保守的步长会减慢收敛，而过于激进的步长则可能导致优化不稳定。我们结合零阶与一阶优化（ZFO），提出一种轻量级框架，将方向选择与步长选择解耦。ZFO 使用可信的一阶优化器确定方向，仅沿这一一维子空间进行零阶评估，以选择移动的距离。利用当前的梯度信息和另外两次目标函数评估，ZFO 实例沿所提议的方向构建目标函数的局部模型，并在有界搜索区间内选择考虑曲率的步长。由此得到一种自适应步长选择机制，其成本低于完整的线搜索。我们给出理论保证，证明共享样本评估能够产生可靠的有限差分曲率估计，所构建的局部模型能够在搜索区间内选择近似最优的步长，并且 ZFO 能够收敛到平稳点的邻域。在所评估的各种设置、语言模型和数据集上，相较于固定步长的一阶基线方法，ZFO 通常能够改善优化过程和最终性能；而改进幅度及更优的局部模型则取决于目标函数。我们的代码已公开发布于：https://github.com/nizswan/Zeroth-First-Order-Framework。
arXiv:2610.02189v1 发布类型：new 摘要：内在无序蛋白质区域（IDRs）在转录调控、信号转导和亚细胞定位等细胞过程中发挥着核心作用，但其功能设计仍面临挑战。基于结构的设计方法并不易于应用于IDRs，而现有的蛋白质语言模型通常在全长蛋白质序列上进行训练，因此学习到的先验偏向于折叠结构域。在此，我们提出IDiom，这是一种在IDiom-DB上训练的自回归蛋白质语言模型；IDiom-DB是从AlphaFold Database中整理出的、包含5400万个预测IDRs的数据集。IDiom能够生成多样化的序列，这些序列再现了天然IDRs的组成、模式、基序和预测无序性。为控制与功能相关的序列模式，我们还引入了基于稀疏自编码器特征的强化学习（RL-SAE），这是一种训练后方法，通过奖励生成能够激活指定特征集的序列来实现控制。在八项IDR设计任务中，RL-SAE生成的序列平均能够激活30个目标特征中的90%，而激活引导的比例为24%。与引导和监督微调相比，我们证明RL-SAE能够提高所生成IDRs的预测亚细胞定位和转录活性，并使与不同生物学功能相关的特征得以组合在单个序列中。因此，IDiom和RL-SAE通过对与功能相关的序列特征进行显式控制，实现了可解释且可组合的IDR设计。更广泛地说，在可解释特征能够提供有用设计目标的其他蛋白质设计场景中，RL-SAE也具有拓展潜力。代码可在https://github.com/rotskoff-group/idiom获取。
arXiv:2610.02175v1 Announce Type: new Abstract: Protein function annotation needs to know which predictions to distrust, not only what a model predicts. We ask whether tissue-specific interaction structure carries that information. Our candidate signal is effective resistance, used previously to relieve over-squashing by rewiring. Across 24 tissue-specific interactomes it is dominated by inverse degree, and the degeneration deepens as the co-expression filtered network grows, with a Spearman correlation of -0.955. The residual departure from that limit exceeds degree-preserving null graphs in all 24 networks. Controlling for predictive entropy, degree, annotation cardinality, local structure and feature-only difficulty, the residual explains additional per-node loss in 19 of 24 held-out networks once a permutation floor is subtracted, at every depth, and the effect strengthens monotonically with depth. The increment reaches 0.37% of the variance the controls leave unexplained, 5.6 times a permutation floor, against 1.5 times when the model is retrained in a degree-preserving null world. Selective prediction improves negligibly. The signal is reproducible; degree degeneration bounds it.
arXiv:2610.02144v1 公告类型：新 摘要：我们推出了 Faynt，这是一系列面向 Super Smash Bros. Melee 的 1000 万参数和 7500 万参数 Transformer 策略，每个策略都能通过单个 checkpoint 控制全部 26 个角色。经过强化学习（RL）后，10M 在由十四个专精角色或多角色版本所支持的角色阵容上进行的同角色对局中赢下了 244 场中的 240 场（98.4%），并且对每个版本都保持胜绩。这些对手仍保留 21 帧或 24 帧的动作延迟；Faynt 不使用额外延迟，而我们尚未单独分析这一差异的影响。在另一项针对私下提供的零延迟 Slippi-AI 模型的评估中，10M 在两种条件设置下的全部 68 场对局中均获胜。我们研究了架构、优化、规模扩展和超参数迁移，以指导在约 840,000 场人类回放上进行预训练。后训练结合了基于排名和结果的课程学习、75M 到 10M 的知识蒸馏，以及仅限于 Fox 镜像对局的强化学习。在最初的 152 场基准测试中，监督式 10M 赢下了 69.7% 的对局，而预训练 75M 的胜率为 45.4%，尽管后者整体留出集控制器预测损失更高。用于选择监督式 checkpoint 的加权验证损失，与全部四个预训练和监督式策略的胜率排序一致。监督式后训练完成后，两个模型每分钟承受的伤害都更少，在前期建立的领先优势更大，并且在丢失第一个生命后获胜的频率更高。在记录的游戏状态上进行优化推理时，10M 在 NVIDIA T4 上每次决策平均耗时 5.2 毫秒，75M 为 8.7 毫秒，不包括模拟器执行和通信时间。我们将权重、两套基准测试以及用于自动化模型锦标赛的平台开源。
arXiv:2610.01974v1 公告类型：new 摘要：仿真和实验测量为时空物理系统的学习提供互补数据，但标准的仿真到实验微调在迁移后仅优化实验目标，可能导致仿真性能下降。我们将仿真—实验预测表述为一个具有域特定仿真风险和实验风险的多目标学习问题。在来自 RealPDEBench 的四个流体系统和两种模型规模上，我们比较了 Simulation only、Experiment only、Sim$\rightarrow$Exp 和 Joint training，并在两个留出域上评估每个最终模型。Sim$\rightarrow$Exp 往往会更强地专注于实验数据，但代价是仿真域遗忘。与多种仿真—实验评估权重组合相比，Joint training 始终取得最佳的平衡性能，同时相较于 Sim$\rightarrow$Exp 显著提升了对仿真性能的保持能力。对于实验测量中缺失的仅仿真场，Joint training 也能更好地加以保留。项目页面：https://mahindrautela.github.io/morph。
arXiv:2610.01951v1 Announce Type: new Abstract: Top-two algorithms are simple and effective for fixed-confidence best-arm identification, but their sharp non-asymptotic behavior is still not well understood. We study this problem for Bernoulli bandits through $\beta$-EB-TCI, the empirical-best top-two rule of Jourdan et al., whose challenger is chosen using a Bernoulli transportation cost with a logarithmic count penalty. We prove that, after the empirical leader has become the true best arm and its sampling fraction stays close to $\beta$, the stopping time is $T_{\beta}^{\star}(\mu)\log(1/\delta)$ up to lower-order concentration terms. We also show that, in this regime, every challenger is sampled linearly often. Thus, for the original algorithm without forced exploration, the main remaining difficulty is to control when the empirical leader becomes permanently correct. These results imply a non-asymptotic high-probability bound for all Bernoulli instances with a unique best arm. If the algorithm satisfies a finite-mean sufficient-exploration condition, the bound further yields the sharp expected sample complexity. In particular, this gives the sharp expectation result for the unguarded Bernoulli rule when all arm means are pairwise distinct, using the sufficient-exploration result of Jourdan et al. Finally, if we add a mild forced-exploration rule that contributes only $O(\sqrt{Kt})$ pulls up to time $t$, we obtain a self-contained expected sample-complexity theorem for any number of arms under the unique-best-arm assumption. We also identify a limitation of proof strategies that try to handle equal suboptimal means through a single index-comparison argument.
arXiv:2610.01892v1 公告类型：新 摘要：多模态智能体通常会在每次执行动作前生成自由形式的推理。对于小型模型而言，有限的模型容量可能导致生成冗长的推理，这些推理对动作生成几乎没有有用指导，却会带来显著的推理成本。为应对这一挑战，我们提出基于选择的结构化推理（Selection-based Structured Reasoning，SSR），这是一种将推理重新定义为选择而非开放式生成的框架。SSR 将反复出现的高层次推理表示为预先指定、可复用的自然语言候选项。在每一轮中，模型根据当前上下文条件下各推理候选项的似然，从中进行选择，而无需额外的任务预测头。使用预先指定的推理轨迹可以实现并行评分：通过教师强制预填充，利用共享的上下文 KV 缓存，在候选项内部及候选项之间并行计算词元似然。我们使用 2B 和 4B 模型，在七个多模态搜索基准上评估了 SSR。在多种强化学习目标和监督微调设置下，SSR 在不牺牲任务性能的情况下实现了显著的效率提升。SSR 达到了与同等规模领先搜索智能体相当的平均成功率，同时将每轮推理延迟降低了 90% 以上，并将每个问题的模型总推理延迟降低了 28%–54%。项目主页：https://zfy0314.github.io/ssr-webpage/。
arXiv:2610.01739v1 Announce Type: new Abstract: Sampling from discrete, unnormalized distributions without access to data is a challenging problem. Neural samplers offer a promising approach by training generative models from density evaluations directly. Despite recent progress, existing discrete neural samplers are prone to mode collapse, come without convergence guarantees when trained via fixed-point iterations, and are often tied to a specific reference process such as masked or uniform diffusion. In this work, we introduce Discrete Gibbs Iterative Neural Sampler, a fixed-point neural sampler that addresses these limitations, enabling efficient, scalable learning, substantially reducing mode collapse in practice. Our framework builds on masked diffusion and also extends to transport between pairs of distributions. We demonstrate that the resulting method scales effectively to high-dimensional systems, supports amortized sampling across different conditions, and enables accurate estimation of alloy phase diagrams.
arXiv:2610.01729v1 公告类型：新 摘要：算法数学推理需要可靠的分解、计算与聚合。最终答案奖励无法为中间步骤中的错误提供充分指导，而成功执行也不能保证数学正确性。本文提出函数结构化图强化学习（Function-Structured Graph Reinforcement Learning，FSG-RL），通过多验证器反馈连接子问题图与 Python 实现。该策略首先通过监督微调（SFT）学习根据函数图生成代码。随后，群组相对策略优化（GRPO）利用答案门控奖励和跨度级信用分配优化策略。该框架还支持教师监督和结构化记忆。研究者从 Grade School Math 8K（GSM8K）、MathQA、MATH 和 Omni-MATH 中整理出一个基准，将公开函数图与私有验证规范配对。在统一评测协议下，相较于 SFT，GRPO 将最终答案准确率从 43.25% 提升至 67.50%，将完整解答成功率从 32.25% 提升至 52.25%。结合教师监督的持续强化学习（RL）带来了进一步提升。这些提升不仅体现在生成格式正确的代码上，还表明验证器引导的强化学习能够有效支持数学推理。代码可在 https://github.com/ZihanLiummyycc/FSG-RL 获取。
arXiv:2610.01652v1 Announce Type: new Abstract: Structured policies improve efficiency, robustness, and interpretability in imitation learning by introducing task-specific inductive bias, but existing structure generation methods rely either on extensive human input or on static domain knowledge encoded in LLMs, which may be inconsistent with the expert demonstrations. We propose a closed-loop framework that iteratively refines structured policies using LLM-guided analysis of policy rollouts. By logging rollouts as semantically meaningful tabular data and prompting the LLM to generate diagnostic analysis code, our method identifies suboptimalities in the policy structure and iteratively corrects them without requiring human instruction. Experiments on car racing and door opening tasks show that our approach improves imitation learning performance by up to 15% over zero-shot LLM-generated structures and requires 75% less compute to achieve the same reinforcement learning performance. These results demonstrate that tabular rollout analysis provides an effective feedback signal to align LLM-generated policy structures with expert demonstrations, and we can utilize it to generate good policy structures automatically.
arXiv:2610.01601v1 公告类型：新增 摘要：决策模型通常会对编码在单个序列中的、大小可变的候选动作集合进行评分。这一设置与生成式系统内部的 System 1 组件日益相关，因为不同运行过程中提出的候选项或其排序可能有所不同。标准的因果交叉编码具有很强的表达能力，但可能导致某个候选项的得分取决于序列化顺序，而不是底层决策问题本身。我们提出候选项无关的分块因果注意力：该机制保留共享上下文和每个候选项内部的因果计算，同时阻断候选项之间的信息流，并重置候选项位置。我们在 Gemma 3 1B、Qwen3 1.7B 和 Qwen3 4B 主干模型上，将这一架构与标准因果注意力及互补的不变性基线进行了比较。候选项无关的注意力在保持具有竞争力的决策质量的同时，持续降低了对置换的敏感性；消融实验表明，候选项隔离是产生该效果的主要原因，而位置重置则完善了预期的对称性。一项规模更大的 Qwen3-4B 研究进一步考察了所提出架构在显著增加的训练数据下的表现。代码可在 \href{https://github.com/guyAmit/ci-decision-models}{\textcolor{blue}{项目仓库}} 中获取，\href{https://huggingface.co/Guy-Amit/qwen3-4b-ci-decision-4096-poc}{\textcolor{blue}{Qwen3-4B 模型制品}} 可在 Hugging Face 上获取。
arXiv:2610.01519v1 公告类型：新提交 摘要：神经符号（NeSy）预测器将先验知识融入神经网络的预测过程，确保输出满足指定约束，因此特别适用于必须遵循领域知识的高风险应用。这一范式的关键瓶颈在于符号约束的获取：将领域知识编码为逻辑公式仍是一个依赖人工且需要大量专家参与的过程。在本研究中，我们探讨通过大语言模型（LLM）进行自动形式化，能够在多大程度上将文本知识系统地转换为可接入 NeSy 预测器的符号知识。为此，我们提出了 auto-nesy-bench，这是一个用于评估约束形式化及其对 NeSy 预测器下游准确率影响的新基准。通过在多个领域开展广泛评估，我们发现，LLM 能够在相当程度上将约束形式化，生成的公式通常与人类专家提供的公式相似。此外，当生成的公式在语法上有效时，它们可以带来高质量的下游预测。代码和基准可在 https://unitn-sml.github.io/auto-nesy-bench/ 获取。
arXiv:2610.01494v1 Announce Type: new Abstract: Sheaf Neural Networks generalize scalar-weighted message passing by replacing scalar edge weights with linear transport maps between local feature spaces. Yet the role of this matrix-valued transport is entangled with the broader sheaf-diffusion construction. We isolate the transport primitive through quiver representations and establish a direct connection with multi-head attention. Treating attention heads as coordinates of a local transport space reveals that standard multi-head attention implements diagonal edge maps: along each directed interaction, a source head can contribute only to the corresponding receiver head. Allowing off-diagonal entries instead enables edge-conditioned communication across heads before neighborhood aggregation. We show that this operation cannot, in general, be absorbed into a single shared linear map applied after aggregation. Building on this characterization, we introduce Topological Attention (Top-A), a multi-head attention that learns edge-dependent off-diagonal routes while preserving the original same-head paths and exactly recovering vanilla attention when the additional routing vanishes. We evaluate Top-A on relational reasoning, heterogeneous graph learning, and algorithmic reasoning, including out-of-distribution generalization, with heterophilic node classification as a contrast setting. The results show that cross-head transport is most useful when the task benefits from interaction-dependent transformations, while heterophily alone provides no systematic advantage. These findings identify edge-conditioned cross-head communication as a distinct computational primitive of matrix-valued transport.
arXiv:2610.01435v1 公告类型：新 摘要：表格基础模型（TFM）通过上下文学习实现了强大的预测性能，但反复基于带标签数据进行条件化会使推理成本高昂。知识蒸馏可以将其预测能力迁移至轻量级、针对特定数据集的学生模型，从而降低这一成本。然而，TFM 的预测同时依赖于带标签的上下文和查询，这引出了两个设计问题：如何构造教师监督信号，以及扩大查询覆盖范围是否有助于蒸馏。我们在两个 TFM 以及神经网络和基于树的学生模型上研究了这些问题，并得出了一套有效的蒸馏方案。该方案使用完整的带标签训练集作为教师上下文，并仅依据教师对观测查询和合成查询的预测来训练学生模型。在 TabArena 上，所得学生模型比经过监督训练、调优并集成的对应模型高出 57–98 个 Elo 分。在 TALENT 上直接采用同一方案后，在 300 个数据集中的 236–258 个数据集上，学生模型相较于匹配的默认学生模型取得了提升，并将主要误差的中位数降低了 4.0%–6.4%。蒸馏后的学生模型相较于教师模型还实现了 3.0–21.6 倍的中位数推理加速，在预测性能与重复推理成本之间提供了一种实用的权衡。代码可在 https://github.com/nums-ai/TFM_Distillation 获取。
arXiv:2610.01377v1 Announce Type: new Abstract: We study causal logistic bandits with counterfactual fairness constraints. The causal structure is given through known factual and counterfactual feature maps that share an unknown logistic reward parameter, but the learner observes only factual rewards. Consequently, the directions determining counterfactual feasibility need not be identifiable from the available feedback. The closest prior analyses either omit a coverage condition or impose a comparatively strong one, and do not establish matching lower bounds. We first show that some coverage condition is necessary: without a coverage-type restriction, factually indistinguishable environments with different optimal fair actions force $\Omega(T)$ expected joint loss. Under a weaker full-rank condition on the factual covariance pooled across actions, we identify a target-specific information scale $V_\star$ that measures the difficulty of estimating rewards and counterfactual effects from factual feedback. We construct worst-case families satisfying this condition on which every policy incurs expected joint loss $\Omega\left(\left[V_\star\min\{\log K,d\}\right]^{1/3}T^{2/3}\right)$. We also give an explore--then--exploit procedure tuned using $V_\star$ and an adaptive algorithm that does not require its value. Both algorithms achieve $\max\{R_T,V_T\}=\widetilde{O}\left(\left[V_\star\min\{\log K,d\}\right]^{1/3}T^{2/3}+\kappa d/\sigma_0^2\right)$, where $R_T$ is regret relative to the best fair action and $V_T$ denotes the cumulative stage-wise positive violations. Thus the upper and lower bounds match in their leading dependence on $T$, $V_\star$, and $\min\{\log K,d\}$, up to logarithmic factors.
arXiv:2610.01317v1 Announce Type: new Abstract: Evaluating candidate architectures in neural architecture search (NAS) faces an inherent trade-off: on the one hand, reliable performance labels are limited because training and evaluating architectures is costly; on the other hand, zero-cost proxies (ZCPs) are cheap to compute at large scale but can be noisy. Yet, how to effectively combine these two sources of supervision remains unclear. In this paper, we propose PPNAS, a novel prediction-powered inference (PPI) approach for NAS. PPNAS fuses (1) a small set of architectures with observed performance labels and (2) a large set of architectures with ZCP information. To combine these two sources of supervision, PPNAS exploits the ordinal information provided by ZCPs to construct additional pairwise ranking supervision, while PPI debiases systematic discrepancies between ZCP-based and true performance rankings. We evaluate PPNAS in end-to-end predictor-based NAS, where it achieves state-of-the-art under limited evaluation budgets. To the best of our knowledge, PPNAS is the first prediction-powered approach for label-efficient NAS.
arXiv:2610.01253v1 公告类型：新 摘要：量子强化学习（QRL）将强化学习与参数化量子电路相结合，是解决组合优化问题的一种有前景的方法。然而，在含噪中等规模量子（NISQ）设备上，退相干、门操作不完美和测量误差会降低策略质量，使学习过程的可靠性下降。现有的错误缓解技术通常以固定校正的方式应用，无法适应不断变化的噪声条件或训练状态的演化。本文提出自适应策略引导的错误缓解（APGEM），将其作为混合量子-经典训练循环中的上下文感知编排层，在 QRL 训练期间动态选择最合适的缓解策略。APGEM 使用策略层面的指标评估零噪声外推（ZNE）、概率误差消除（PEC）、Clifford 数据回归（CDR）和读出误差缓解（REM），这些指标包括量子态保真度、策略熵、累积奖励和近似比，并将选定的策略直接整合到强化学习循环中。该框架在容量约束车辆路径问题（CVRP）上进行了评估。CVRP 是城市物流中的代表性 NP-hard 问题，评估覆盖了一系列 NISQ 噪声模型和噪声水平。APGEM 持续优于传统的静态缓解方法，达到预言机策略效用的约 94%，在噪声增强时保持更高的量子态保真度，并在整个训练过程中产生更加稳定的学习行为。消融研究表明，该框架能够学习上下文感知的缓解策略，以适应不同的噪声环境和电路执行条件。这些发现表明，将自适应错误缓解整合到学习过程中，能够显著提升 QRL 在 NISQ 硬件上的鲁棒性和可靠性。
arXiv:2610.01204v1 发布类型：新 摘要：自回归建模通过学习根据先前观测预测未来状态，已在语言和序列任务中取得显著成功。矿产勘探钻孔为这一范式提供了一个天然但尚未得到充分探索的应用场景：随着钻进不断进行，岩性信息从浅部到深部逐层揭示，因此对更深地层的预测本质上是自回归的。然而，现有钻孔建模主要依赖空间插值与重建，或在很大程度上依赖掩码建模，使得严格意义上的自回归预测仍缺乏充分研究。我们提出了 DrillBench，这是一个包含49,671个西澳大利亚钻孔的基准，用于在分级迁移谱上进行下一层预测和自回归地层生成，该迁移谱涵盖从局部预测、空间偏移到跨地质省迁移的不同场景。对经典模型、地统计模型和神经网络模型的基准测试揭示了清晰的“迁移边界”：空间与地球化学条件约束能够带来显著的局部性能提升，但在更强的偏移下性能会急剧下降；相比之下，基于岩性序列的自回归模型具有更强的迁移鲁棒性。受此发现启发，我们开发了一种与骨干网络无关的方法，将基于历史钻孔的大规模预训练与邻近岩性的空间检索相结合。在风化覆盖层中，当局部空间连续性仍具有信息价值时，检索最为有效；而在基岩区域以及更广泛的地质偏移条件下，预训练的贡献更为显著。二者结合后，在保持较强局部性能的同时，还提升了模型在空间偏移和跨地质省偏移下的泛化能力，尤其是在距离最远的数据划分上表现最为明显。基准和代码可访问：https://github.com/yihaoding/drillbench。
arXiv:2610.01175v1 Announce Type: new Abstract: Standard information bottleneck (IB) regularization constrains representations via a single scalar I(Z;X), implicitlytreating all information as homogeneous. However, a single global compression control couples label-relevant structurewith residual within-condition variation, rather than regulating their allocation independently, allowing nuisanceinformation to persist in learned representations. For example, in medical imaging applications, residual variation oftenstems from acquisition conditions, background factors, or subject-specific appearance. This issue becomes particularlypronounced in data-limited settings, where models tend to overfit such variation, hindering generalization. While existingregularization methods can stabilize training, control capacity, or shape representation geometry, they do not explicitlyseparate nuisance-like variation from task-supporting structure. To address this limitation, we revisit IB from a structuredperspective based on a label-induced partition, where condition-level structure and within-condition information playdistinct roles. This leads to a dual-bottleneck formulation: a standard KL term controls global information capacity, while aconditional KL term targets within-condition information. We show that the conditional KL admits an exact decompositioninto a within-condition information term and a prior-mismatch term, explaining its alignment with the design objective.With a simplex-structured conditional prior, the method provides controllable latent geometry and integrates seamlesslyinto existing pipelines. Experiments on classification and segmentation show the clearest gains in low-data classificationand consistent improvements across dense prediction benchmarks.
arXiv:2610.01153v1 公告类型：new 摘要：循环 Transformer 为 LLM 引入了一个新的扩展维度：通过重复应用共享的 Transformer 模块，在不增加参数量的情况下提升有效深度。然而，在 FLOPs 匹配的比较中，循环对于大型 MoE LLM 的收益仍不明确。主要原因在于，额外迭代带来的收益会迅速递减，甚至可能转为性能下降，因此用于循环的额外 FLOPs 几乎无法带来实质性改进。 consequently，已有工作通常止步于两次循环。 我们确定了扩展循环 MoE 的两个主要障碍。首先，循环继承并放大了深度诅咒：随着残差更新不断累积，隐藏状态的方差在每次迭代中增长，从而使深层递归变得不稳定，并导致表示发生漂移。其次，循环 MoE 会遭遇专家选择坍缩：路由器在不同循环中反复选择相同的专家，因此额外迭代只增加了计算量，却没有带来计算多样性。 基于这一分析，我们提出 LOOM，其核心原则是：每次循环都应贡献新的计算，同时保持递归状态稳定。LOOM 通过缩放残差更新以限制方差增长，并在每次循环中重新注入输入嵌入来稳定递归；同时，通过每循环独立的路由器启用不同专家，并利用将早期输出向前传递的 Looping Residual 来增加多样性。 在 100M–1.7B 模型上的实验表明，循环次数可稳定扩展至 9–12 次。在近等 FLOP 条件下，700M 模型在 5 次循环时表现最佳：困惑度从 18.36 降至 16.54，平均零样本准确率则从非循环基线的 38.84% 提升至 39.53%。在不匹配 FLOP 的情况下，使用 60B tokens 训练的 1.7B 模型在 9 次循环时达到峰值：困惑度从 9.62 降至 7.77，平均零样本准确率从 42.4% 提升至 47.7%。代码见 https://github.com/hed-ucas/LOOM。
arXiv:2610.00968v1 公告类型：新 摘要：因果表示学习旨在利用数据生成背后的因果结构来发现稳健特征。现有方法要求预先指定因果结构，但不同结构需要从根本上不兼容的不变性约束，而结构设定错误会导致学得的表示丢弃预测信息。我们提出 SaCRL，这是一种在无需先验结构知识的情况下，同时识别因果结构并学习相应不变表示的框架。我们的方法利用基于 HSIC 的违背度量，在候选不变性上进行软优化，将结构选择形式化，并通过自适应权重自动聚焦于可实现的结构。我们为结构识别提供了理论保证，包括在随机特征近似下的保证、不变性满足保证以及分布外泛化保证。在实证方面，SaCRL 在合成和半合成贝叶斯网络基准上恢复了真实结构；在 Colored MNIST 上优于固定不变性基线；在三个 DomainBed 基准（PACS、VLCS、OfficeHome）上取得了当前最佳精度；并且在结构设定错误和环境多样性有限的情况下表现出平稳退化。代码可在以下地址获取：https://github.com/ArmanBehnam/sacrl。
arXiv:2610.00948v1 公告类型：新 摘要：GUI 模型周围的可执行工具链决定了如何组织观测、执行动作，以及控制验证、恢复和终止。与非 GUI 智能体的工具链优化相比，自动优化这一工具链面临三个相互耦合的挑战：协调模型意图与观测到的视觉效果；在执行结果具有变化性的情况下诊断失败；识别跨任务反复出现的失败模式，并将其转化为可复用的运行时修改。我们提出 GUI-HARVEST，这是一种自动工具链优化器，可在骨干模型冻结的情况下实现自我改进的 GUI 智能体。首先，为使诊断以观测到的动作效果为依据，它将模型输出和已执行的动作与前后截图对齐，把发现的问题关联到具体的界面转换。其次，为应对执行变化性，它将同一任务的多次运行视为一个联合证据单元，通过任务内比较定位与结果相关的行为差异。第三，它汇总跨任务验证过的发现，形成反复出现的失败模式，将其映射为有界的源代码修改，并在评估前记录预测结果；随后通过重复执行，在任务性能之外检验预测的行为影响。在 OSWorld-Verified 上的实验表明，GUI-HARVEST 在六个通用开源、GUI 专用开源及专有骨干模型上均取得了稳定的留出集提升；Qwen3-VL-32B-Instruct 在完整测试套件上的得分提升了 12.33 个百分点。在无需进一步优化的情况下，冻结工具链迁移使 GPT-5 在 50 步限制下于 WindowsAgentArena 上提升了 13.87 个百分点。在使用相同骨干模型和初始工具链的条件下，GUI-HARVEST 的表现优于 Self-Harness 和 Meta-Harness，这表明面向 GUI 的诊断与验证有助于使工具链改进泛化到未见任务。代码可在 https://github.com/GaryYang12345/GUI-HARVEST 获取。
arXiv:2610.00895v1 公告类型：新 摘要：基础模型仍然容易受到虚假相关性和“Clever Hans”策略的影响。可解释机器学习能够在没有元数据的情况下，为分类器发现并移除这类策略。然而，对于基础模型，目前还没有这样的解决方案。我们提出了 Disentangled Diffusion Autoencoders（DiDAE）。DiDAE 通过条件扩散解码器封装一个冻结的基础模型。一个反事实样本就是沿着解耦字典中的某个方向进行一次闭式编辑，然后执行解码。该字典可以是有监督的（Procrustes），也可以是无监督的（奇异值分解、稀疏自编码器）。由于不需要梯度，DiDAE 的速度最高可达到现有技术的 2000 倍。我们在六个数据集上进行了评估，其中两个为合成数据集，四个为真实世界数据集。在其中三个数据集上依据预期目标设计的基准测试中，DiDAE 生成的反事实样本达到或优于现有技术；此外，它们还能通过反事实知识蒸馏（Counterfactual Knowledge Distillation，CFKD）修复下游分类器，其效果优于基于元数据的校正。同样的机制还可以将一个预训练字典与一个训练好的分类器进行匹配排序。它能够返回分类器实际读取的少数几个方向，并通过能够翻转决策的反事实样本对每个方向进行因果验证；对于教师标记为虚假的方向，还能据此修复分类器。我们在随论文一同发布的开源 Peal 库中实现了即插即用的工作流。借助公开字典和预训练解码器，剩下的工作仅需对分类器进行低成本的线性蒸馏及其自身的微调。
arXiv:2610.00873v1 Announce Type: new Abstract: In many industrial applications, 1) tabular data is scarce and imbalanced and thus requires synthetic expansion; 2) input distributions drift between training and deployment (covariate shift); 3) validation sets often diverge from unseen test environments; or 4) standard generative models simply mimic outdated source distributions. This learning setting limits the stability of standard augmentation and adaptation pipelines. We generalize the task under such setting as the Augmented and Weighted Learning under Covariate Shift problem (AWL-CS). AWL-CS imposes two critical challenges on existing methods: 1) misleading generative guidance where models optimize for source similarity rather than downstream task relevance, and 2) structural instability of distributional density where reweighting mechanisms overfit to noisy validation signals. To tackle these challenges, we propose IGDPR (Invariant-Guided Diffusion with Prototype Reweighting), a unified framework that synergizes stable synthesis and structural adaptation: i) To achieve task-relevant generation, we steer the diffusion sampling process using invariant potentials to ensure synthetic samples align with stable decision boundaries rather than outdated correlations. ii) To ensure stable adaptation, we develop a prototype-based reweighting strategy that assesses sample reliability through structural clusters instead of isolated points, effectively filtering validation noise. Extensive experiments on real data demonstrate our method improves data quality by augmenting the most beneficial data for robust learning.
arXiv:2610.00861v1 Announce Type: new Abstract: Adversarial optimization under a shared $\ell_1$ budget requires deciding not only how much perturbation to use, but also where that limited budget should be spent. This allocation problem becomes particularly important when individual input coordinates are subject to local magnitude constraints, which restrict the extent to which perturbation can be concentrated on a small number of locations. We introduce an importance-guided allocation mechanism that uses a fixed clean-gradient prior to steer perturbation toward model-sensitive regions while leaving the feasible perturbation set unchanged. A centered allocation objective encourages perturbation at above-average importance locations and discourages unnecessary expenditure elsewhere, thereby redistributing rather than enlarging the available budget. Across ten robust model--dataset configurations under a common capacity-limited threat setting, the proposed method improves attack success over matched APGD- and PMA-based baselines by $2.52$ to $17.70$ percentage points. Allocation analysis shows that these gains are accompanied by substantially greater perturbation mass in high-importance regions without increased global $\ell_1$ consumption. Mechanism ablations further show that centered non-uniform redistribution provides part of the benefit, while model-derived importance yields an additional improvement. These results identify perturbation allocation as a distinct and practically relevant dimension of adversarial optimization under shared-budget, locally constrained threat models.
arXiv:2610.00838v1 Announce Type: new Abstract: Agentic reinforcement learning (RL) trains a large language model (LLM) to act over long, multi-step interactions. However, a single localized error can cause task failure, while trajectory-level rewards provide limited guidance for assigning credit to individual decisions. To address this limitation, we introduce Segment-level Hindsight Advantage Reweighting for Policy Optimization (SHARPO), a credit-assignment mechanism that refines Group Relative Policy Optimization (GRPO) at the level of environment-facing segments. Inspired by the existing on-policy self-distillation (OPSD) method, SHARPO computes teacher-student log-probability gaps within each segment and uses the resulting signal to compute a bounded multiplier on the GRPO advantage. This multiplier is shared by all tokens within the segment, allowing credit to vary across different segments. With Qwen2.5-7B-Instruct, SHARPO outperforms existing baselines on the ALFWorld and WebShop benchmarks, including GRPO, SDAR, RLSD, and StepOPSD.
arXiv:2610.00831v1 发布类型：新 摘要：类型化决策是在固定选项集合中进行选择，并以概率而非文本形式返回结果。如今，需要类型化决策的系统使用的是专门为此训练的模型。本报告介绍 AnyJev：它通过对预训练的指令微调语言模型进行一次预填充，读取一个类型化决策。读取过程会将答案位置处的下一个词元分布限制为选项词元。该方法存在两个缺陷：无论输入是什么，模型都会为某些标签分配更高的概率，并偏向选项列表中的某些位置。AnyJev 无需梯度更新，也无需修改参数，即可纠正这两个问题：利用无标签输入估计出的标签先验进行除法校正，并对选项列表的 K 个循环移位的对数概率取平均。在两个包含 20 个选项的任务上，循环移位将顺序翻转率分别从 0.33 降至 0.14 和从 0.33 降至 0.18，并且在两个任务中都使 11 个模型中的全部 11 个模型准确率提升。读取每个循环移位需要进行 K 次预填充。针对无标签状态、以完整循环移位决策为参照所选出的停止规则可以减少这一开销。在一个无标签划分上选择阈值，并在第二个划分上界定其不一致率后，对于四个单元格中的两个，系统在经验证的不一致率 0.008 上读取 18 个循环移位中的 10.6 个；如我们的服务运行所做的那样，在同一个划分上完成选择和界定时，系统读取 7.3 个循环移位，并在 vLLM 上每秒处理的决策数量达到 2.2 倍。代码已开源。
arXiv:2610.00730v1 Announce Type: new Abstract: Mixed-integer linear programs (MILP) model many real-world decision problems, motivating machine-learning methods that exploit recurring structure to accelerate MILP solving. MILPs can admit many equivalent formulations: integrality-preserving changes of variables and the addition of redundant constraints can alter their formulations while preserving the optimization problem. We leverage these reformulations as a source of self-supervision for learning general-purpose representations of MILP variables and constraints. We characterize the affine reformulations that are valid for every input instance, and distinguish re-descriptions, which leave variables unchanged, from substitutions, which transform them predictably. Building on equivariant self-supervised learning, we introduce ReMILP (reformulation-contrastive MILP representation learning), which jointly trains a graph neural network and a hypernetwork to predict how variable embeddings transform under changes of variables. Without solver-derived labels, ReMILP learns representations that exhibit the intended invariance and equivariance on unseen problem classes. Across binary solution, constraint activity and integrality gap prediction, these representations carry task-relevant information when frozen and provide a useful initialization for fine-tuning.
arXiv:2610.00704v1 Announce Type: new Abstract: Natural-language skills are textual procedural memories through which large language model (LLM) agents retain reusable task knowledge without updating model weights. Existing methods typically treat skills as either static artifacts or monolithic documents optimized using aggregate validation scores as feedback. However, representing a skill as a monolithic document restricts optimization to its textual content, without explicitly modeling the structure through which procedural knowledge is retrieved and executed. We identify a key distinction between learning what knowledge to retain and determining how to organize it: textual updates should first be validated through execution evidence, after which the retained knowledge should be structured according to its procedural dependencies and retrieval requirements. To this end, we introduce SkillSpec, a two-phase framework comprising consensus-gated evolution and representation specialization. In the consensus-gated phase, complementary editing intents generate complete candidate skills. An update is committed only when paired evaluations reach consensus, requiring sufficient overall improvement and non-negative aggregate paired gain in every repeated evaluation. In the specialization phase, signals of process and redundancy sensitivity derived from the full optimization trajectory, including accepted and rejected candidates, guide the selection of a flat, graph, or hybrid representation.Across six benchmarks and three target language models, SkillSpec improves average success rate over SkillOpt by 6.89%, averaged across the three models. These results demonstrate that reliable skill evolution and representation specialization address complementary objectives: deciding what knowledge to retain and how to structure it for inference.
arXiv:2610.00675v1 公告类型：新摘要：LLM 驱动发现中的进化方法通常从少量选定的祖先程序生成新程序。这样可以控制上下文规模，但可能遗漏其他实验中的有用证据；而纳入完整的实验历史又会产生冗长且冗余的上下文。我们介绍了一种围绕 LabBook 构建的简单单智能体发现框架。LabBook 由智能体维护，发挥两种互补作用：从完整实验日志中引导检索相关证据，并为新解决方案的生成提供信息。在每次迭代中，同一个智能体将其记忆与检索到的证据相结合，同时生成下一个程序并更新 LabBook。该方法将完整历史保留与选择性上下文构建分离，无需显式的种群或分支搜索结构。在 49 个 Frontier-CS 问题上，与使用两种骨干模型评估的进化基线相比，LabBook 改善了观测到的质量—成本权衡；在另外九项数学、系统和启发式设计任务中也保持了竞争力。代码将发布于 https://github.com/BoYuanVisionary/LabBook。
arXiv:2610.00604v1 发布类型：新 摘要：视觉-语言-动作策略通常只能看到最近的一帧或几帧，这使得评估它们如何使用任务过程中消失的信息变得困难。我们提出了MIKASA-Robo-VLA，这是一个包含90项语言条件操作任务的基准测试。其中除10项任务外，其余任务都会隐藏动作所依赖的线索；这10项是反应式控制任务。MIKASA-Robo是其重构的任务套件，共包含32项任务，并且仅在其中具有代表性的VLA子集中使用语言。在这里，每项任务都提供一条指令，而依赖记忆的任务会隐藏与任务相关的线索，反应式控制任务则始终保留该线索。对于70项任务，环境阶段的时间安排明确规定了信息间隔；其中28项任务的间隔超过了我们调查的具有最宽固定上下文的VLA的16帧窗口。该间隔只计算能够确定线索缺失的时间段，而不是策略必须保留该线索的完整时长，因此按照构造，每项依赖记忆的任务仍然需要记忆，包括那些测得间隔较短的任务。我们以RLDS和LeRobotDataset v3格式发布了涵盖10种记忆类型的22,500条专家轨迹。一个使用当前图像和本体感知信息、但不使用观测历史或显式记忆模块的参考$\pi_{0.5}$基线，在14项任务上进行了微调，取得了0.211 $\pm$ 0.044的平均任务成功率。它在评估的Long-split任务上的较低成功率，受到开环分块执行以及该子集中所包含记忆类型的混杂影响。项目页面：https://mikasarobo.github.io/
arXiv:2610.00592v1 公告类型：新 摘要：在部分可观测强化学习（RL）中，后续观测可能使已存储的信息失效，或改变其对下一步决策的含义。RL 的记忆架构和基准测试大多考察保持能力，即在需要使用信息之前使其保持不变。我们对另外两项要求进行了形式化定义。重写会将决策相关内容设为一个与旧内容无关的值，而经验融合则依据后续观测指定的规则转换旧内容。对于由此类更新构成的任务，我们统计解决方案所需的记忆状态数量；若干基线方法在需要更多状态的组合任务上取得了最低成功率。我们提出 ALER（Adaptive Learnable Experience Rewriting），一种将 LSTM 与槽位记忆相结合的智能体。独立寻址的 Gumbel-Softmax 写入机制将权重集中于一个槽位，并覆盖该槽位；随后，学习得到的门控机制会在策略头和价值头处理之前，将检索到的内容与循环状态进行融合。我们还提出 Rune-Mazes，这是三个环境，在向量观测和像素观测下，符文观测会对隐藏线索的更新执行反转、取消、重置或重复操作。在与七种基线方法的对比中，ALER 在全部十六种 Endless T-Maze 配置中的成功率均至少达到 $0.82$，在全部五种 Rune T-Maze 组合中的成功率均至少达到 $0.99$；在带有 Invert 符文的四分支 Rune Multi-Corridor 环境中，其平均成功率最高。在基于像素的 Rune MiniGrid Memory 上，ALER 在十种配置中的八种配置里取得了高于 PPO-LSTM 的平均成功率。项目主页：https://quartz-admirer.github.io/ALER-Adaptive-Learnable-Experience-Rewriting/。
arXiv:2610.00574v1 发布类型：新 摘要：多奖励强化学习训练大型语言模型，使其同时满足多个行为目标。GDPO 中采用的按奖励归一化方法保留了各奖励在 rollout 组内的相对信息，但不同目标的学习进度仍可能不均衡。我们通过优势能量研究这一现象，即一个批次中某项奖励的优势平方和。在理想化的 GDPO 归一化条件下，我们证明该能量与活跃组密度成正比：即在所有 rollout 组中，该奖励提供非零相对优势的组所占比例。这揭示了批次层面的残余信号不平衡，并为校准奖励贡献提供了依据。基于这一关系，我们提出 Density-Aware Reward Aggregation（DARA）。我们推导出一种逆平方根密度校正方法，为较少频繁激活的奖励所产生的信号赋予更高权重。DARA 根据每个 rollout 批次计算权重，能够适应训练过程中奖励活跃度的变化，同时无需修改底层的策略优化目标。在工具调用和数学推理任务上的实验表明，与 GDPO 相比，DARA 能更快学会目标行为：在工具调用任务中，达到较高格式合规性的训练步数最多减少 26%；在数学推理任务中，达到接近饱和的长度合规性的训练步数最多减少 65%；同时最终性能仍具有竞争力。我们的代码可在 https://github.com/zhaihaotian/DARA 获取。
arXiv:2610.00563v1 Announce Type: new Abstract: This paper presents a preliminary study of an alternative to the affine transformation underlying conventional neural-network layers. In the proposed Soft Dominance Layer, each output unit compares input coordinates with a learnable reference vector and aggregates smooth inequality responses. A sigmoid relaxation makes the comparisons differentiable, while a sharpness parameter $\alpha$ controls their transition toward hard threshold decisions. The aim is to examine the trainability and direct threshold interpretation of this primitive, not to claim a replacement for affine layers. In single-run MNIST experiments, the highest observed Soft Dominance accuracy is $0.9061$ without annealing and $0.9173$ with annealing, compared with $0.9827$ for the MLP baseline. These descriptive results do not establish reliable configuration rankings or a statistically supported annealing benefit. Learned reference vectors exhibit spatial structure, providing qualitative evidence of structured learning. Repeated-seed experiments and broader datasets are required to assess robustness and practical relevance beyond this proof of concept.
arXiv:2610.00554v1 Announce Type: new Abstract: We evaluate hybrid quantum-classical machine learning for the reduced-order prediction of spatiotemporal brain deformation fields. To mitigate the computational intractability of high-dimensional displacement fields, we employ Proper Orthogonal Decomposition (POD) to project the data into a compact latent space. Within this framework, we formulate two distinct learning objectives: static temporal-to-latent regression and autoregressive latent state forecasting. We systematically benchmark compact classical baselines against both minimal and enhanced hybrid quantum architectures. Our results demonstrate that classical networks provide the strongest baselines in the present setting. For static regression, a classical POD-MLP outperforms all evaluated quantum variants, although an enhanced Variational Quantum Circuit (VQC) substantially improves upon a minimal VQC baseline. For temporal forecasting, a classical POD-LSTM delivers superior predictive accuracy and statistical robustness compared to an enhanced Quantum LSTM (QLSTM) across varying history windows and random initializations. Overall, this study establishes reduced-order physical field learning as a rigorous testbed for near-term QML, highlighting that while hybrid enhancements successfully recover expressivity in weak quantum circuits, classical architectures retain a definitive advantage in both fidelity and stability.
arXiv:2610.00545v1 Announce Type: new Abstract: We study adversarial online maximization of nonnegative, non-monotone DR-submodular functions over compact convex down-closed sets. A learner commits each action before observing its objective and competes with the best fixed action in hindsight. We prove a comparator-uniform first-order inequality that gives coefficient $4/9$, improving the online $0.401$ benchmark, with one gradient query and one projection per round and $O(\sqrt T)$ expected approximate regret. If $\zeta {\bf 1} \in K\subseteq[0,1]^d$, the coefficient improves to $\underline\alpha(\zeta)=\tfrac12-(1-2\zeta)_+^2/[2(3-2\zeta)^2]$. The proof is a direct ordered-coordinate argument with an objective-independent rational action. Conversely, a three-group symmetry-gap construction yields an offline oracle upper bound $\beta_*=0.470438681380894\ldots$ at $\zeta=0$, even with exact value and full-gradient responses. A parameterized extension and exact finite-instance bounds define an upper function for every $\zeta$. The lower and upper bounds match at $1/2$ for $\zeta\ge1/2$, and show that the optimal deficit from $1/2$ is $\Theta((1/2-\zeta)^2)$ as $\zeta\uparrow1/2$. For coefficient-revealed polynomials we obtain $1/2$ for quadratics and a geometry-dependent cubic coefficient starting at $8/17$, including $0.49$ at $\zeta=1/5$. A constant objective sequence yields an offline $(4/9-\varepsilon)$ approximation with polynomially many first-order queries on the cube and projections, without requiring a supplied positive lower bound on the optimum. We also give nonanticipating adaptive-adversary and value-feedback guarantees, including $O(T^{3/4})$ regret with one noisy value per round.
arXiv:2610.00445v1 Announce Type: new Abstract: Pairwise guide--transcript scores do not enforce conservation of a finite guide-loaded RISC pool when they are interpreted independently as occupancies. We formulate a differentiable scalar equilibrium layer: one conservation equation with a unique positive root and exact implicit gradients. It yields a redistribution theorem, a qualified high-resource limit, an analysis of the retrieval approximation, and a conditional rank-invariance result: within one construct at one dose, rankings by fractional occupancy cannot distinguish equilibrium from independent scoring. We therefore audit the two experiments that proposition leaves open, dose and cross-context, on archival off-target data. Corrected thermodynamic affinities associate weakly with measured repression in the direction a working predictor requires, but a paired permutation test and a construct-cluster bootstrap do not establish added predictive value from the coupling: what survives their differing permutation-null baselines is \GapNet{}, a descriptive \GapNetOverSE{} of the equilibrium association's cluster standard error. The dose fits are heterogeneous and frequently violate the model-implied exponent constraint, which is superlinear rather than sublinear, so these data do not identify the competition parameter. A saturable compression of the competitor set holds both accuracy targets on held-out guide families but is not faster at the size measured. The contribution is a reusable conservation operator and the experimental information needed to test it. The code for this study is available at https://github.com/shadi97kh/One-Pool-Many-Targets.
arXiv:2610.00426v1 公告类型：新 摘要：我们提出了 IrekoGPT，这是一种事后方法，可将预训练大语言模型转换为可变宽度模型，使其宽度能够在推理时进行调整。在 SliceGPT 的基础上，我们保留其投影矩阵而不对其进行剪枝，使单个模型能够以不同宽度呈现嵌套子网络。我们通过在多个压缩比下对每一层进行校准来提升鲁棒性，并通过无梯度岭回归校正下游线性层。在 Llama 和 Qwen 模型上的初步结果表明，相较于基于朴素 PCA 的瘦身方法，我们的方法取得了更好的效果，并且在高压缩率下的提升最大。代码可在 https://github.com/aimagelab/IrekoGPT 获取
arXiv:2610.00417v1 公告类型：新 摘要：在模型生成的数据上反复训练，可能导致后续模型性能下降。一种可能的应对方式，是在决定复用哪些生成样本时利用其来源信息。我们同时测试了这种来源信息能否被可靠地恢复，以及它是否有助于识别更优质的训练数据。我们使用金融风险文本，首先识别生成段落的来源，然后在改写这些段落后重复测试。对原始段落进行生成器归因的准确率为98.7%，但在释义改写后降至53.1%，在风格改写后降至29.0%。相对于所测试的人类文本对照集，生成文本与人类文本的检测准确率仍接近完美。随后，我们在三轮生成与再训练过程中，比较了两种选择生成样本的方法。一种使用来源信息，另一种使用单独的参考模型给出的分数。两种规则选择出的样本不同，但预先设计的比较未能检测到所得模型性能下降存在稳定差异。结果表明，识别数据的来源与识别哪些数据对训练有用，是两个彼此独立的问题。因此，该实验区分了来源身份、面向评判标准的选择和递归训练结果：无论是来源溯源分数，还是所测试的面向评判标准的代理指标，都尚未被证明足以预测未来的递归行为。
arXiv:2610.00412v1 公告类型：新 摘要：KV 缓存剪枝通过驱逐不太重要的键值对，降低长上下文推理的内存使用量。KVzip 通过上下文重构来估计重要性：提示模型逐块重复上下文。这种方法能够实现高质量压缩，但代价是需要额外的前向传播。基于学习的近似方法可以降低这一成本，但需要针对特定模型进行训练。我们分析了 KVzip 如何识别缓存中的重要信息，并展示了如何利用预填充过程中已经计算的信息来近似其重构得分。这些发现促成了 EchoPress——一种无需训练的方法，它利用标准预填充中的查询和键来近似重构注意力。对于每个请求，该方法仅重构第一个块，以校准剩余上下文的重要性得分。在使用 Qwen3-8B 和 Llama-3.1-8B-Instruct 于 LongBench 和 RULER 上进行的实验中，EchoPress 在 50% 至 90% 的驱逐比例范围内，任务准确率与 KVzip 相当，同时将压缩开销降低了 1.7 至 19.6 倍，并将总预填充时间最多降低了 2.9 倍。代码可在 https://github.com/ljwljwljwljw/kvpress/tree/echo-press 获取。
arXiv:2610.00403v1 Announce Type: new Abstract: A network can fit its training examples while failing to recover the rule that generated their labels. We examine this separation in single-hidden-layer multilayer perceptrons (MLPs), using synthetic tasks that control interaction order and the presence of nuisance inputs. We establish elementary benchmark properties: pure parity contains no predictive lower-order marginals, admits an exact Bayes posterior, and can be represented on clean latent inputs by a width-$k$ ReLU network. Experiments then identify distinct optimization outcomes. In a matched order-2--4 sweep, SGD, Adam, and Muon all reach 100\% peak test accuracy at order two; at order three they reach 96.25\%, 50.87\%, and 76.82\%, respectively, while Muon reaches 99.21\% at order four. In a separate mixed-order task, freezing only the first-layer weights connected to independent nuisance inputs raises AdamW's epoch-10 accuracy from 44.73\% to 95.07\%. Removing the same inputs only at test time raises it to 48.38\%. Thus, nuisance-weight learning changes the training outcome beyond its immediate effect on prediction. Bias interventions expose a connection between target symmetry and shallow ReLU representations. In a compact signal-only regime, both SGD and Muon learn orders five through eight, with higher SGD peak accuracy at orders nine through eleven. Together, the results show how optimization and nuisance learning constrain the higher-order rules realized by a shallow network.
arXiv:2610.00395v1 公告类型：新 摘要：推理时引导使预训练扩散模型能够满足新的约束，而无需进行完整的重新训练。然而，特异性感知生成十分困难：将样本排斥出负参考分布的同时，也可能侵蚀两个分布重叠区域中的正参考分布。关键挑战在于抑制负分布质量，同时尽可能减少对正分布的扭曲。我们将这一问题表述为目标设计问题，并从基于重叠的目标函数中推导出一个目标分布。所得目标分布仅在期望参考分布相对于非期望参考分布具有足够优势的区域保留前者，从而赋予特异性一种似然比解释。为了从相应的时变目标路径中采样，我们设计了一种带有方差最小化局部提议分布的序贯蒙特卡洛采样器。我们还利用期望与非期望得分场的雅可比—向量积，引入了一种实用的固定噪声优化过程。在合成任务、类别对比生成、文生图任务以及肽-MHC（p-MHC）结合物上的实验表明，与负向引导基线方法相比，所提出的方法能够更有效地抑制非期望区域，减少模态偏移，并通过降低SMC权重坍缩来提升采样稳定性。代码可在以下地址获取：https://github.com/WangLuran/Specificity-Aware-Diffusion-Steering
arXiv:2610.00381v1 公告类型：新 摘要：医学模型不仅要根据正确性进行评判，还要看其报告的置信度是否与实际准确率相符。通用型医学多模态模型拓展了单一模型的感知能力，但它们仍然将诊断、影像所见或细胞计数等有界决策表达为生成文本，因此报告的概率反映的是下一个词元，而不是决策本身。受 Jev 等决策原生接口的启发，我们提出 OmniMed-Jev，将每项医学决策表示为基于运行时提供的候选集进行的 Choice、Noul 或 Score 决策，并返回该候选集上的完整分布：互斥类别、某项所见是否存在的二元判断，或有界的有序数值。该设计在三个方面体现了 omni 特性：它接受多种成像模态，覆盖不同预测任务，并通过一个以候选项为条件的概率模型来表达这些任务，从而使异构输出成为可比较的概率，而非特定于任务的字符串。在与生成式基线进行接口受控的比较中，双方使用相同的骨干网络、数据和训练计划，OmniMed-Jev 报告的概率与观测到的正确性吻合得更为紧密，将校准误差最多降低一个数量级，将可靠性误差最多降低两倍，同时点预测性能保持相当；计数是唯一一个生成式基线仍然领先的任务类别。将决策分布作为模型输出并非格式上的改变，而是使报告的数值真正成为名副其实的概率的关键。这些结果表明，在所评估的任务范围内，显式决策建模有助于使报告的置信度具有实际意义；但它们并不意味着模型已具备临床应用准备度：该比较无法将接口因素与相关的训练差异区分开来，我们也在结果旁对此进行了说明。代码可在 github.com/lytang63/OmniMed-Jev 获取。
arXiv:2610.00377v1 公告类型：新 摘要：基于站点的天气预报为日常生活和经济活动提供支持，但要实现准确预测，需要对站点之间复杂的空间依赖关系进行建模。近年来，基于聚类的选择性建模为替代站点间密集交互提供了有前景的方案。然而，在一个观测窗口内共享同一种分组方式可能会掩盖站点关系的局部变化，而仅进行簇内交互又可能遗漏重要的全局上下文。与密集连接相比，选择性交互在理论上的优势也仍未得到充分理解。为此，我们提出 STCFormer，这是一种自适应时空 Transformer，根据每个时间块内站点的局部演化动态地对站点进行分组。其簇引导注意力模块结合了簇内的细粒度局部注意力与基于区域状态摘要的全局注意力，使每个站点都能够获取自身所属簇之外的信息。我们进一步证明，针对簇条件局部注意力推导出的 Lipschitz 上界不大于其全连接对应形式的上界，这说明该方法可能具备更好的鲁棒性，并为 InfoLoss 的设计提供了动机。在涵盖八项温度和风力预测任务的三个真实世界天气数据集上进行的实验表明，STCFormer 在全部八项任务中均取得了最低的 24 小时均方误差，并且在涵盖不同指标和预测时域的 48 项比较中，有 47 项排名第一或第二。消融实验和案例研究进一步证实了局部自适应分组以及互补的局部—全局交互的优势。我们的代码可从 https://github.com/hnu-vis/STCFormer 获取。
arXiv:2610.00370v1 公告类型：新 摘要：站点天气预报从根本上受到站点之间复杂空间依赖关系以及天气变量之间强物理耦合的共同影响。然而，现有研究通常将这些关系分开考虑，并采用不同的数据集和实验设置，从而阻碍了对它们各自贡献及联合贡献的系统评估。本文介绍了 $M^2$Weather，这是一个用于联合多站点与多变量天气预报的基准。通过多标准质量控制和站点分层，我们收集了2,809个高质量站点，涵盖5个具有物理耦合关系的天气变量，并覆盖法国、欧洲和全球三个空间尺度。这种多尺度设计使我们能够考察相关结论能否从国家级站点网络延续到全球站点网络。我们还引入了统一的训练和评估协议，以便公平比较不同的站点—变量建模范式。为进一步考察建模站点—变量关系的益处，我们设计了一种轻量级、即插即用的适配器。在已有训练好的天气预报模型基础上，该适配器无需重新训练模型，即可引入缺失的站点或变量关系。这使得对站点—变量关系进行公平且高效的研究成为可能。对16个具有代表性的模型进行系统评估表明，联合建模站点关系和变量关系能够带来益处。在三个数据集上，为所有适配模型补全缺失关系均可进一步降低MSE。总体而言，这些结果表明，站点之间和变量之间的互补信息是改进站点天气预报的重要资源。我们的代码可从 https://github.com/hnu-vis/M2-Weather 获取。
arXiv:2610.00254v1 Announce Type: new Abstract: We characterize the regret attainable in online convex optimization when access to the feasible set is limited to an exact linear optimization oracle. The learner is given an inscribed ball and a diameter bound and must remain feasible on every consistent instance. For convex $G$-Lipschitz losses, diameter at most $D$, a total allowance of $Q$ oracle calls, and a strict limit of $B$ calls per round, the dimension-free minimax expected regret is $\Theta(GD\max\{\sqrt T,T/(1+\min\{Q,BT\})^{1/4}\})$. The lower bound applies to arbitrary randomized learners. Universal feasibility first forces each action into the hull of the supplied ball and the preceding oracle replies. A fixed-body construction then couples fresh phase directions to a shared simplex, making useful replies costly repeatedly even though all losses have a common minimizer. A counted approximate-gradient method with interleaved blocks attains the matching rate. Total-budget and strict per-round guarantees follow as special cases, including the $T^{3/4}$ rate with one call per round and the quadratic total budget needed for $\sqrt T$ regret. For prescribed smoothness $\beta$, an analytic construction yields a curvature-dependent lower bound and identifies the threshold above which the general characterization remains sharp.
arXiv:2610.00236v1 公告类型：new 摘要：少样本异常检测器通常通过排序指标进行评估，但部署时需要一个具有受控误报率（FAR）的报警阈值。我们研究：对于一个未见类别，需要多少正常数据证据（按图像数或类别单位计）才能认证这样的阈值。我们在四种损坏类型下，使用冻结的 DINOv2 主成分分析（PCA）残差排序器，对 15 个 MVTec 类别和 12 个 VisA 类别进行分析，结果表明，仅使用目标类别的留一图像（LOIO）校准受分辨率限制且对分布偏移脆弱：排序值无法低于 $1/(k+1)$；在可达到的 $\alpha=0.20$ 水平下，当 $k=4$ 时，高斯损坏的 MVTec 数据上的经验 FAR 达到 0.341，是名义水平的 1.7 倍。随后我们推导出类别数量的可行性计算：即使所有类别损失均为零且不计多重性代价，任何确定性的、统一有效的、无分布假设的 95% 上置信界（UCB），在 $\alpha=0.20$、$0.10$ 和 $0.05$ 时，分别至少需要 14、29 和 59 个独立同分布（iid）的类别抽样；这些数量是必要条件，但并不充分。带源支持的跨类别可靠性估计（CRESS）协议将源类别划分为互不重叠的参考、提议和认证角色。当认证类别仅有三个或四个时，全部 960 个冻结配置均返回故障关闭阈值 $\tau^\star=0$，而最小的类别级 UCB 为 0.950。对同一数据档案进行的图像单位分析，在 36.7% 至 60.3% 的目标单元中选出了正阈值；这些界限针对所选源混合分布成立，而不适用于新类别抽样的边际风险。本文的贡献在于给出一个定量的可行性边界，并提出一个关注估计对象的协议，明确何时源数据证据能够或不能支持可迁移的可靠性声明。
arXiv:2610.00209v1 Announce Type: new Abstract: Reliable recovery of missing measurements is important for monitoring and analysis in energy time-series systems, where fine-grained measurements may contain sensitive temporal information. Diffusion models trained with differentially private stochastic gradient descent (DP-SGD) provide a promising framework for privacy-sensitive energy time-series imputation. Under cosine diffusion schedules, late timesteps correspond to low signal-to-noise ratio (SNR) conditions, where standard $\varepsilon$-prediction can induce large pre-clipping gradients. Such gradients are more likely to be clipped, reducing the retained optimization signal. The artificial intelligence (AI) contribution lies in formulating this objective--clipping interaction as an objective optimization problem under fixed-threshold DP-SGD and developing timestep-aware objective conditioning for diffusion-based energy time-series imputation. The method adopts $v$-prediction to mitigate late-timestep gradient amplification, uses static loss weighting as a uniform-scaling control, and introduces diffusion-schedule-aware dynamic weighting for stronger attenuation before clipping. For the engineering application, we evaluate the method on five real-world energy time-series datasets across random point missingness, contiguous block missingness, persistent outages, and multiple missing-data severities. Under matched DP-SGD settings, the proposed method consistently improves imputation utility over the $\varepsilon$-prediction baseline. Gradient diagnostics reveal lower upper-tail pre-clipping gradient norms, reduced clipping fractions, and stronger attenuation at late low-SNR timesteps, supporting the effectiveness of clipping-aware objective conditioning for energy time-series imputation.
arXiv:2610.00188v1 公告类型：新 摘要：基于体素的体积建图是三维重建的基础，但固定分辨率网格在本质上仍然效率低下——在均匀区域浪费内存，在复杂区域丢失细节。现有的自适应方法（如 MAP-ADAPT）通过根据几何结构和用户定义的语义类别列表改变分辨率，在一定程度上解决了这一问题，但这些启发式方法需要专家调参，难以泛化到未见过的物体，也没有提供显式的内存使用控制机制。我们提出了一种自适应框架，根据能够捕捉标签不确定性的语义熵，以及作为场景复杂度线索的几何曲率和纹理丰富度来细化体素，从而无需依赖语义分类体系即可实现有原则的分辨率分配。为使精度与内存之间的权衡更加明确且可由用户控制，我们进一步引入了一个强化学习智能体，使其能够在用户指定的目标内存预算下学习体素细分策略，以单一且直观的控制参数取代人工调节的阈值。在合成数据集和真实世界数据集上，与 MAP-ADAPT 和固定分辨率基线相比，最终得到的多分辨率 TSDF 实现了更高的几何精度、更好的语义一致性，以及更优的内存-精度权衡。我们的代码和模型已发布于 https://github.com/alpayozkan/UnRL。
arXiv:2610.00120v1 公告类型：新 摘要：在真实世界的临床实践中，医学图像面临开放世界的分布变化：（i）长尾稀有疾病；（ii）被正常解剖结构主导的细微病灶；以及（iii）层级化分类体系。然而，大多数开放世界范式都假设标签空间是扁平且均衡的，因而未能解决这些生物医学需求。我们提出了广义生物医学发现（Generalized Biomedicine Discovery，GBD）以及一个涵盖长尾、异常和分类体系感知发现的统一基准。我们的关键洞见是，占主导地位的已知模式会形成一个掩盖细微新颖性的视觉流形。受专家诊断启发，我们提出了 SCAN（Surprise-evoked Complementary AccommodatioN），遵循一种受认知启发的感知渐进过程：应用预测性抑制来过滤预期的正常模式，触发惊奇诱发的显著性以突出非预期偏差，并执行互补性调适，将这些变化整合到全局表征中。大量实验表明，SCAN 能够提升新颖概念发现能力，同时通常保留既有的临床知识；此外，它还可接入现有架构，从而更好地应对医学影像中已知与未知之间的权衡。代码可在 https://github.com/lytang63/generalized-biomedicine-discovery 获取。
arXiv:2610.00050v1 Announce Type: new Abstract: Kolmogorov-Arnold Networks (KANs) represent a paradigmatic shift in deep learning by replacing fixed node activations with learnable univariate functions on edges, offering enhanced interpretability and parameter efficiency. While recent polynomial-based KAN variants have addressed the computational overhead of original B-spline implementations, they introduce a fundamental yet underexplored challenge: the domain mismatch between unbounded real-valued inputs and the bounded or semi-infinite support of orthogonal polynomial bases. To address this limitation, we propose the Stieltjes-Wigert Kolmogorov-Arnold Network (SW-KAN), a novel architecture that employs Stieltjes-Wigert q-orthogonal polynomials defined on the semi-infinite domain (0, infinity). We introduce a smooth exponential-of-tanh mapping that stably bridges the domain gap while preserving well-conditioned gradients, and leverage a numerically stable three-term recurrence that evaluates polynomial expansions in O(N) operations without special-function calls. Through comprehensive experiments spanning image classification and continuous function approximation, we demonstrate that SW-KAN achieves superior accuracy-efficiency trade-offs across diverse tasks. The log-normal weight structure and learnable q-parameter of Stieltjes-Wigert polynomials provide a distinct inductive bias that enables robust performance under resource-constrained conditions, including reduced feature dimensionality and limited training data. The proposed architecture not only outperforms established polynomial KAN baselines on standard benchmarks but also exhibits strong representational capacity for approximating complex multivariate functions with remarkably few parameters, making it a compelling alternative for efficient function approximation and classification in resource-constrained settings.
arXiv:2610.00009v1 Announce Type: new Abstract: Frequency-collapse attention [Zeris, 2026e] achieves large gains over standard dot-product attention by replacing the Q/K dot product with a bandpass-filtered inner product at a learned frequency. A natural follow-up question is: which filter shape works best, and why? We test five hypotheses about filter properties -- DC suppression, Nyquist suppression, bandwidth, centre frequency, and multi-scale coverage -- using a controlled ablation on character-level language modelling (TinyShakespeare, 6-layer GPT). Our main findings are: (1) DC and Nyquist components are actively harmful (val ~= 2.0, equivalent to phase randomisation), confirming that oscillatory bandpass structure is essential, not just any low-dimensional spectral summary; (2) the optimal single-scale bandwidth is sigma ~= 2 bins centred at paragraph scale (~70 tokens), giving a clean gain of Delta = +1.15 nats over BASE-DOT; (3) admissible filters (zero-mean, Mexican Hat DOG m = 2) outperform non-admissible Gaussians at the same scale and provide partial protection against bilateral FFT leakage; (4) bilateral FFT leakage scales monotonically with spectral coverage -- narrowband filters (gap > +4) are clean, wideband filters (gap < +2) are leaky; and (5) causal time-domain Morlet at character scale cannot beat BASE-DOT (K=128 taps covers 50% of T=256 context), motivating word-level experiments in the companion MorletQK paper [Zeris, 2026f]. Together, findings (1)-(5) characterise FourierQK as effective in bidirectional attention settings (encoder-style, e.g. BERT), where full-sequence context is available at both training and inference time; autoregressive generation requires a causal spectral variant such as MorletQK [Zeris, 2026f] (decoder-style, e.g. GPT). Code available at: https://github.com/AthanasiosZeris/energy-gated-attention
In this post, I’ll introduce a reinforcement learning (RL) algorithm based on an “alternative” paradigm: divide and conquer . Unlike traditional methods, this algorithm is not based on temporal difference (TD) learning (which has scalability challenges ), and scales well to long-horizon tasks. We can do Reinforcement Learning (RL) based on divide and conquer, instead of temporal difference (TD) learning. Problem setting: off-policy RL Our problem setting is off-policy RL . Let’s briefly review what this means. There are two classes of algorithms in RL: on-policy RL and off-policy RL. On-policy RL means we can only use fresh data collected by the current policy. In other words, we have to throw away old data each time we update the policy. Algorithms like PPO and GRPO (and policy gradient methods in general) belong to this category. Off-policy RL means we don’t have this restriction: we can use any kind of data, including old experience, human demonstrations, Internet data, and so on. So off-policy RL is more general and flexible than on-policy RL (and of course harder!). Q-learning is the most well-known off-policy RL algorithm. In domains where data collection is expensive ( e.g. , robotics , dialogue systems, healthcare, etc.), we often have no choice but to use off-policy RL. That’s why it’s such an important problem. As of 2025, I think we have reasonably good recipes for scaling up on-policy RL ( e.g. , PPO, GRPO, and their variants). However, we still haven’t found a “scalable” off-policy RL algorithm that scales well to complex, long-horizon tasks. Let me briefly explain why. Two paradigms in value learning: Temporal Difference (TD) and Monte Carlo (MC) In off-policy RL, we typically train a value function using temporal difference (TD) learning ( i.e. , Q-learning), with the following Bellman update rule: \[\begin{aligned} Q(s, a) \gets r + \gamma \max_{a'} Q(s', a'), \end{aligned}\] The problem is this: the error in the next value $Q(s’, a’)$ propagates to the current value $Q(s, a)$ through bootstrapping, and these errors accumulate over the entire horizon. This is basically what makes TD learning struggle to scale to long-horizon tasks (see this post if you’re interested in more details). To mitigate this problem, people have mixed TD learning with Monte Carlo (MC) returns. For example, we can do $n$-step TD learning (TD-$n$): \[\begin{aligned} Q(s_t, a_t) \gets \sum_{i=0}^{n-1} \gamma^i r_{t+i} + \gamma^n \max_{a'} Q(s_{t+n}, a'). \end{aligned}\] Here, we use the actual Monte Carlo return (from the dataset) for the first $n$ steps, and then use the bootstrapped value for the rest of the horizon. This way, we can reduce the number of Bellman recursions by $n$ times, so errors accumulate less. In the extreme case of $n = \infty$, we recover pure Monte Carlo value learning. While this is a reasonable solution (and often works well ), it is highly unsatisfactory. First, it doesn’t fundamentally solve the error accumulation problem; it only reduces the number of Bellman recursions by a constant factor ($n$). Second, as $n$ grows, we suffer from high variance and suboptimality. So we can’t just set $n$ to a large value, and need to carefully tune it for each task. Is there a fundamentally different way to solve this problem? The “Third” Paradigm: Divide and Conquer My claim is that a third paradigm in value learning, divide and conquer , may provide an ideal solution to off-policy RL that scales to arbitrarily long-horizon tasks. Divide and conquer reduces the number of Bellman recursions logarithmically. The key idea of divide and conquer is to divide a trajectory into two equal-length segments, and combine their values to update the value of the full trajectory. This way, we can (in theory) reduce the number of Bellman recursions logarithmically (not linearly!). Moreover, it doesn’t require choosing a hyperparameter like $n$, and it doesn’t necessarily suffer from high variance or suboptimality, unlike $n$-step TD learning. Conceptually, divide and conquer really has all the nice properties we want in value learning. So I’ve long been excited about this high-level idea. The problem was that it wasn’t clear how to actually do this in practice… until recently. A practical algorithm In a recent work co-led with Aditya , we made meaningful progress toward realizing and scaling up this idea. Specifically, we were able to scale up divide-and-conquer value learning to highly complex tasks (as far as I know, this is the first such work!) at least in one important class of RL problems, goal-conditioned RL . Goal-conditioned RL aims to learn a policy that can reach any state from any other state. This provides a natural divide-and-conquer structure. Let me explain this. The structure is as follows. Let’s first assume that the dynamics is deterministic, and denote the shortest path distance (“temporal distance”) between two states $s$ and $g$ as $d^*(s, g)$. Then, it satisfies the triangle inequality: \[\begin{aligned} d^*(s, g) \leq d^*(s, w) + d^*(w, g) \end{aligned}\] for all $s, g, w \in \mathcal{S}$. In terms of values, we can equivalently translate this triangle inequality to the following “transitive” Bellman update rule: \[\begin{aligned} V(s, g) \gets \begin{cases} \gamma^0 & \text{if } s = g, \\\\ \gamma^1 & \text{if } (s, g) \in \mathcal{E}, \\\\ \max_{w \in \mathcal{S}} V(s, w)V(w, g) & \text{otherwise} \end{cases} \end{aligned}\] where $\mathcal{E}$ is the set of edges in the environment’s transition graph, and $V$ is the value function associated with the sparse reward $r(s, g) = 1(s = g)$. Intuitively , this means that we can update the value of $V(s, g)$ using two “smaller” values: $V(s, w)$ and $V(w, g)$, provided that $w$ is the optimal “midpoint” (subgoal) on the shortest path. This is exactly the divide-and-conquer value update rule that we were looking for! The problem However, there’s one problem here. The issue is that it’s unclear how to choose the optimal subgoal $w$ in practice. In tabular settings, we can simply enumerate all states to find the optimal $w$ (this is essentially the Floyd-Warshall shortest path algorithm). But in continuous environments with large state spaces, we can’t do this. Basically, this is why previous works have struggled to scale up divide-and-conquer value learning, even though this idea has been around for decades (in fact, it dates back to the very first work in goal-conditioned RL by Kaelbling (1993) – see our paper for a further discussion of related works). The main contribution of our work is a practical solution to this issue. The solution Here’s our key idea: we restrict the search space of $w$ to the states that appear in the dataset, specifically, those that lie between $s$ and $g$ in the dataset trajectory. Also, instead of searching for the optimal $\text{argmax}_w$, we compute a “soft” $\text{argmax}$ using expectile regression . Namely, we minimize the following loss: \[\begin{aligned} \mathbb{E}\left[\ell^2_\kappa (V(s_i, s_j) - \bar{V}(s_i, s_k) \bar{V}(s_k, s_j))\right], \end{aligned}\] where $\bar{V}$ is the target value network, $\ell^2_\kappa$ is the expectile loss with an expectile $\kappa$, and the expectation is taken over all $(s_i, s_k, s_j)$ tuples with $i \leq k \leq j$ in a randomly sampled dataset trajectory. This has two benefits. First, we don’t need to search over the entire state space. Second, we prevent value overestimation from the $\max$ operator by instead using the “softer” expectile regression. We call this algorithm Transitive RL (TRL) . Check out our paper for more details and further discussions! Does it work well? Your browser does not support the video tag. humanoidmaze Your browser does not support the video tag. puzzle To see whether our method scales well to complex tasks, we directly evaluated TRL on some of the most challenging tasks in OGBench , a benchmark for offline goal-conditioned RL. We mainly used the hardest versions of humanoidmaze and puzzle tasks with large, 1B-sized datasets. These tasks are highly challenging: they require performing combinatorially complex skills across up to 3,000 environment steps . TRL achieves the best performance on highly challenging, long-horizon tasks. The results are quite exciting! Compared to many strong baselines across different categories (TD, MC, quasimetric learning, etc.), TRL achieves the best performance on most tasks. TRL matches the best, individually tuned TD-$n$, without needing to set $\boldsymbol{n}$ . This is my favorite plot. We compared TRL with $n$-step TD learning with different values of $n$, from $1$ (pure TD) to $\infty$ (pure MC). The result is really nice. TRL matches the best TD-$n$ on all tasks, without needing to set $\boldsymbol{n}$ ! This is exactly what we wanted from the divide-and-conquer paradigm. By recursively splitting a trajectory into smaller ones, it can naturally handle long horizons, without having to arbitrarily choose the length of trajectory chunks. The paper has a lot of additional experiments, analyses, and ablations. If you’re interested, check out our paper ! What’s next? In this post, I shared some promising results from our new divide-and-conquer value learning algorithm, Transitive RL. This is just the beginning of the journey. There are many open questions and exciting directions to explore: Perhaps the most important question is how to extend TRL to regular, reward-based RL tasks beyond goal-conditioned RL. Would regular RL have a similar divide-and-conquer structure that we can exploit? I’m quite optimistic about this, given that it is possible to convert any reward-based RL task to a goal-conditioned one at least in theory (see page 40 of this book ). Another important challenge is to deal with stochastic environments. The current version of TRL assumes deterministic dynamics, but many real-world environments are stochastic, mainly due to partial observability. For this, “stochastic” triangle inequalities might provide some hints. Practically, I think there is still a lot of room to further improve TRL. For example, we can find better ways to choose subgoal candidates (beyond the ones from the same trajectory), further reduce hyperparameters, further stabilize training, and simplify the algorithm even more. In general, I’m really excited about the potential of the divide-and-conquer paradigm. I still think one of the most important problems in RL (and even in machine learning) is to find a scalable off-policy RL algorithm. I don’t know what the final solution will look like, but I do think divide and conquer, or recursive decision-making in general, is one of the strongest candidates toward this holy grail (by the way, I think the other strong contenders are (1) model-based RL and (2) TD learning with some “magic” tricks). Indeed, several recent works in other fields have shown the promise of recursion and divide-and-conquer strategies, such as shortcut models , log-linear attention , and recursive language models (and of course, classic algorithms like quicksort, segment trees, FFT, and so on). I hope to see more exciting progress in scalable off-policy RL in the near future! Acknowledgments I’d like to thank Kevin and Sergey for their helpful feedback on this post. This post originally appeared on Seohong Park’s blog .
An encoder (optical system) maps objects to noiseless images, which noise corrupts into measurements. Our information estimator uses only these noisy measurements and a noise model to quantify how well measurements distinguish objects. Many imaging systems produce measurements that humans never see or cannot interpret directly. Your smartphone processes raw sensor data through algorithms before producing the final photo. MRI scanners collect frequency-space measurements that require reconstruction before doctors can view them. Self-driving cars process camera and LiDAR data directly with neural networks. What matters in these systems is not how measurements look, but how much useful information they contain. AI can extract this information even when it is encoded in ways that humans cannot interpret. And yet we rarely evaluate information content directly. Traditional metrics like resolution and signal-to-noise ratio assess individual aspects of quality separately, making it difficult to compare systems that trade off between these factors. The common alternative, training neural networks to reconstruct or classify images, conflates the quality of the imaging hardware with the quality of the algorithm. We developed a framework that enables direct evaluation and optimization of imaging systems based on their information content. In our NeurIPS 2025 paper , we show that this information metric predicts system performance across four imaging domains, and that optimizing it produces designs that match state-of-the-art end-to-end methods while requiring less memory, less compute, and no task-specific decoder design. Why mutual information? Mutual information quantifies how much a measurement reduces uncertainty about the object that produced it. Two systems with the same mutual information are equivalent in their ability to distinguish objects, even if their measurements look completely different. This single number captures the combined effect of resolution, noise, sampling, and all other factors that affect measurement quality. A blurry, noisy image that preserves the features needed to distinguish objects can contain more information than a sharp, clean image that loses those features. Information unifies traditionally separate quality metrics. It accounts for noise, resolution, and spectral sensitivity together rather than treating them as independent factors. Previous attempts to apply information theory to imaging faced two problems. The first approach treated imaging systems as unconstrained communication channels, ignoring the physical limitations of lenses and sensors. This produced wildly inaccurate estimates. The second approach required explicit models of the objects being imaged, limiting generality. Our method avoids both problems by estimating information directly from measurements. Estimating information from measurements Estimating mutual information between high-dimensional variables is notoriously difficult. Sample requirements grow exponentially with dimensionality, and estimates suffer from high bias and variance. However, imaging systems have properties that enable decomposing this hard problem into simpler subproblems. Mutual information can be written as: \[I(X; Y) = H(Y) - H(Y \mid X)\] The first term, $H(Y)$, measures total variation in measurements from both object differences and noise. The second term, $H(Y \mid X)$, measures variation from noise alone. Mutual information equals the difference between total measurement variation and noise-only variation. Imaging systems have well-characterized noise. Photon shot noise follows a Poisson distribution. Electronic readout noise is Gaussian. This known noise physics means we can compute $H(Y \mid X)$ directly, leaving only $H(Y)$ to be learned from data. For $H(Y)$, we fit a probabilistic model (e.g. a transformer or other autoregressive model) to a dataset of measurements. The model learns the distribution of all possible measurements. We tested three models spanning efficiency-accuracy tradeoffs: a stationary Gaussian process (fastest), a full Gaussian (intermediate), and an autoregressive PixelCNN (most accurate). The approach provides an upper bound on true information; any modeling error can only overestimate, never underestimate. Validation across four imaging domains Information estimates should predict decoder performance if they capture what limits real systems. We tested this relationship across four imaging applications. Information estimates predict decoder performance across color photography, radio astronomy, lensless imaging, and microscopy. Higher information consistently produces better results on downstream tasks. Color photography. Digital cameras encode color using filter arrays that restrict each pixel to detect only certain wavelengths. We compared three filter designs: the traditional Bayer pattern, a random arrangement, and a learned arrangement. Information estimates correctly ranked which designs would produce better color reconstructions, matching the rankings from neural network demosaicing without requiring any reconstruction algorithm. Radio astronomy. Telescope arrays achieve high angular resolution by combining signals from sites across the globe. Selecting optimal telescope locations is computationally intractable because each site’s value depends on all others. Information estimates predicted reconstruction quality across telescope configurations, enabling site selection without expensive image reconstruction. Lensless imaging. Lensless cameras replace traditional optics with light-modulating masks. Their measurements bear no visual resemblance to scenes. Information estimates predicted reconstruction accuracy across a lens, microlens array, and diffuser design at various noise levels. Microscopy. LED array microscopes use programmable illumination to generate different contrast modes. Information estimates correlated with neural network accuracy at predicting protein expression from cell images, enabling evaluation without expensive protein labeling experiments. In all cases, higher information meant better downstream performance. Designing systems with IDEAL Information estimates can do more than evaluate existing systems. Our Information-Driven Encoder Analysis Learning (IDEAL) method uses gradient ascent on information estimates to optimize imaging system parameters. IDEAL optimizes imaging system parameters through gradient feedback on information estimates, without requiring a decoder network. The standard approach to computational imaging design, end-to-end optimization, jointly trains the imaging hardware and a neural network decoder. This requires backpropagating through the entire decoder, creating memory constraints and potential optimization difficulties. IDEAL avoids these problems by optimizing the encoder alone. We tested it on color filter design. Starting from a random filter arrangement, IDEAL progressively improved the design. The final result matched end-to-end optimization in both information content and reconstruction quality. IDEAL matches end-to-end optimization performance while avoiding decoder complexity during training. Implications Information-based evaluation creates new possibilities for rigorous assessment of imaging systems in real-world conditions. Current approaches require either subjective visual assessment, ground truth data that is unavailable in deployment, or isolated metrics that miss overall capability. Our method provides an objective, unified metric from measurements alone. The computational efficiency of IDEAL suggests possibilities for designing imaging systems that were previously intractable. By avoiding decoder backpropagation, the approach reduces memory requirements and training complexity. We explore these capabilities more extensively in follow-on work . The framework may extend beyond imaging to other sensing domains. Any system that can be modeled as deterministic encoding with known noise characteristics could benefit from information-based evaluation and design, including electronic, biological, and chemical sensors. This post is based on our NeurIPS 2025 paper “Information-driven design of imaging systems” . Code is available on GitHub . A video summary is available on the project website .
different_tests Understanding the behavior of complex machine learning systems, particularly Large Language Models (LLMs), is a critical challenge in modern artificial intelligence. Interpretability research aims to make the decision-making process more transparent to model builders and impacted humans, a step toward safer and more trustworthy AI. To gain a comprehensive understanding, we can analyze these systems through different lenses: feature attribution , which isolates the specific input features driving a prediction ( Lundberg & Lee, 2017 ; Ribeiro et al., 2022 ); data attribution , which links model behaviors to influential training examples ( Koh & Liang, 2017 ; Ilyas et al., 2022 ); and mechanistic interpretability , which dissects the functions of internal components ( Conmy et al., 2023 ; Sharkey et al., 2025 ). Across these perspectives, the same fundamental hurdle persists: complexity at scale . Model behavior is rarely the result of isolated components; rather, it emerges from complex dependencies and patterns. To achieve state-of-the-art performance, models synthesize complex feature relationships, find shared patterns from diverse training examples, and process information through highly interconnected internal components. Therefore, grounded or reality-checked interpretability methods must also be able to capture these influential interactions . As the number of features, training data points, and model components grow, the number of potential interactions grows exponentially, making exhaustive analysis computationally infeasible. In this blog post, we describe the fundamental ideas behind SPEX and ProxySPEX , algorithms capable of identifying these critical interactions at scale. Attribution through Ablation Central to our approach is the concept of ablation , measuring influence by observing what changes when a component is removed. Feature Attribution: We mask or remove specific segments of the input prompt and measure the resulting shift in the predictions. Data Attribution: We train models on different subsets of the training set, assessing how the model’s output on a test point shifts in the absence of specific training data. Model Component Attribution (Mechanistic Interpretability): We intervene on the model’s forward pass by removing the influence of specific internal components, determining which internal structures are responsible for the model’s prediction. In each case, the goal is the same: to isolate the drivers of a decision by systematically perturbing the system, in hopes of discovering influential interactions. Since each ablation incurs a significant cost, whether through expensive inference calls or retrainings, we aim to compute attributions with the fewest possible ablations . different_tests Masking different parts of the input, we measure the difference between the original and ablated outputs. SPEX and ProxySPEX Framework To discover influential interactions with a tractable number of ablations, we have developed SPEX (Spectral Explainer). This framework draws on signal processing and coding theory to advance interaction discovery to scales orders of magnitude greater than prior methods. SPEX circumvents this by exploiting a key structural observation: while the number of total interactions is prohibitively large, the number of influential interactions is actually quite small. We formalize this through two observations: sparsity (relatively few interactions truly drive the output) and low-degreeness (influential interactions typically involve only a small subset of features). These properties allow us to reframe the difficult search problem into a solvable sparse recovery problem. Drawing on powerful tools from signal processing and coding theory, SPEX uses strategically selected ablations to combine many candidate interactions together. Then, using efficient decoding algorithms, we disentangle these combined signals to isolate the specific interactions responsible for the model’s behavior. image2 In a subsequent algorithm, ProxySPEX , we identified another structural property common in complex machine learning models: hierarchy . This means that where a higher-order interaction is important, its lower-order subsets are likely to be important as well. This additional structural observation yields a dramatic improvement in computational cost: it matches the performance of SPEX with around 10x fewer ablations . Collectively, these frameworks enable efficient interaction discovery, unlocking new applications in feature, data, and model component attribution. Feature Attribution Feature attribution techniques assign importance scores to input features based on their influence on the model’s output. For example, if an LLM were used to make a medical diagnosis, this approach could identify exactly which symptoms led the model to its conclusion. While attributing importance to individual features can be valuable, the true power of sophisticated models lies in their ability to capture complex relationships between features. The figure below illustrates examples of these influential interactions: from a double negative changing sentiment (left) to the necessary synthesis of multiple documents in a RAG task (right). image3 The figure below illustrates the feature attribution performance of SPEX on a sentiment analysis task. We evaluate performance using faithfulness : a measure of how accurately the recovered attributions can predict the model’s output on unseen test ablations. We find that SPEX matches the high faithfulness of existing interaction techniques (Faith-Shap, Faith-Banzhaf) on short inputs, but uniquely retains this performance as the context scales to thousands of features. In contrast, while marginal approaches (LIME, Banzhaf) can also operate at this scale, they exhibit significantly lower faithfulness because they fail to capture the complex interactions driving the model’s output. image4 SPEX was also applied to a modified version of the trolley problem, where the moral ambiguity of the problem is removed, making “True” the clear correct answer. Given the modification below, GPT-4o mini answered correctly only 8% of the time. When we applied standard feature attribution (SHAP), it identified individual instances of the word trolley as the primary factors driving the incorrect response. However, replacing trolley with synonyms such as tram or streetcar had little impact on the prediction of the model. SPEX revealed a much richer story, identifying a dominant high-order synergy between the two instances of trolley , as well as the words pulling and lever, a finding that aligns with human intuition about the core components of the dilemma. When these four words were replaced with synonyms, the model’s failure rate dropped to near zero. image5 Data Attribution Data attribution identifies which training data points are most responsible for a model’s prediction on a new test point. Identifying influential interactions between these data points is key to explaining unexpected model behaviors. Redundant interactions, such as semantic duplicates, often reinforce specific (and possibly incorrect) concepts, while synergistic interactions are essential for defining decision boundaries that no single sample could form alone. To demonstrate this, we applied ProxySPEX to a ResNet model trained on CIFAR-10, identifying the most significant examples of both interaction types for a variety of difficult test points, as shown in the figure below. image6 As illustrated, synergistic interactions (left) often involve semantically distinct classes working together to define a decision boundary. For example, grounding the synergy in human perception, the automobile (bottom left) shares visual traits with the provided training images, including the low-profile chassis of the sports car, the boxy shape of the yellow truck, and the horizontal stripe of the red delivery vehicle. On the other hand, redundant interactions (right) tend to capture visual duplicates that reinforce a specific concept. For instance, the horse prediction (middle right) is heavily influenced by a cluster of dog images with similar silhouettes. This fine-grained analysis allows for the development of new data selection techniques that preserve necessary synergies while safely removing redundancies. Attention Head Attribution (Mechanistic Interpretability) The goal of model component attribution is to identify which internal parts of the model, such as specific layers or attention heads, are most responsible for a particular behavior. Here too, ProxySPEX uncovers the responsible interactions between different parts of the architecture. Understanding these structural dependencies is vital for architectural interventions, such as task-specific attention head pruning. On an MMLU dataset (highschool‐us‐history), we demonstrate that a ProxySPEX-informed pruning strategy not only outperforms competing methods, but can actually improve model performance on the target task . image7 On this task, we also analyzed the interaction structure across the model’s depth. We observe that early layers function in a predominantly linear regime, where heads contribute largely independently to the target task. In later layers, the role of interactions between attention heads becomes more pronounced, with most of the contribution coming from interactions among heads in the same layer. image8 What’s Next? The SPEX framework represents a significant step forward for interpretability, extending interaction discovery from dozens to thousands of components . We have demonstrated the versatility of the framework across the entire model lifecycle: exploring feature attribution on long-context inputs, identifying synergies and redundancies among training data points, and discovering interactions between internal model components. Moving forwards, many interesting research questions remain around unifying these different perspectives, providing a more holistic understanding of a machine learning system. It is also of great interest to systematically evaluate interaction discovery methods against existing scientific knowledge in fields such as genomics and materials science, serving to both ground model findings and generate new, testable hypotheses. We invite the research community to join us in this effort: the code for both SPEX and ProxySPEX is fully integrated and available within the popular SHAP-IQ repository. https://github.com/mmschlk/shapiq (SHAP-IQ Github) https://openreview.net/forum?id=KI8qan2EA7 (ProxySPEX NeurIPS 2025) https://openreview.net/forum?id=pRlKbAwczl (SPEX ICML 2025) https://openreview.net/forum?id=glGeXu1zG4 (Learning to Understand NeurIPS 2024)
Adaptive Parallel Reasoning overview Overview of adaptive parallel reasoning. What if a reasoning model could decide for itself when to decompose and parallelize independent subtasks, how many concurrent threads to spawn, and how to coordinate them based on the problem at hand? We provide a detailed analysis of recent progress in the field of parallel reasoning, especially Adaptive Parallel Reasoning. Disclosure: this post is part landscape survey, part perspective on adaptive parallel reasoning. One of the authors (Tony Lian) co-led ThreadWeaver ( Lian et al., 2025 ), one of the methods discussed below. The authors aim to present each approach on its own terms. Motivation Recent progress in LLM reasoning capabilities has been largely driven by inference-time scaling, in addition to data and parameter scaling ( OpenAI et al., 2024 ; DeepSeek-AI et al., 2025 ). Models that explicitly output reasoning tokens (through intermediate steps, backtracking, and exploration) now dominate math, coding, and agentic benchmarks. These behaviors allow models to explore alternative hypotheses, correct earlier mistakes, and synthesize conclusions rather than committing to a single solution ( Wen et al., 2025 ). The problem is that sequential reasoning scales linearly with the amount of exploration. Scaling sequential reasoning tokens comes at a cost, as models risk exceeding effective context limits ( Hsieh et al., 2024 ). The accumulation of intermediate exploration paths makes it challenging for the model to disambiguate amongst distractors when attending to information in its context, leading to a degradation of model performance, also known as context-rot ( Hong, Troynikov and Huber, 2025 ). Latency also grows proportionally with reasoning length. For complex tasks requiring millions of tokens for exploration and planning, it’s not uncommon to see users wait tens of minutes or even hours for an answer ( Qu et al., 2025 ). As we continue to scale along the output sequence length dimension, we also make inference slower, less reliable, and more compute-intensive. Parallel reasoning has emerged as a natural solution. Instead of exploring paths sequentially ( Gandhi et al., 2024 ) and accumulating the context window at every step, we can allow models to explore multiple threads independently (threads don’t rely on each other’s context) and concurrently (threads can be executed at the same time). Figure 1: Sequential vs. Parallel Reasoning Figure 1: Sequential vs. Parallel Reasoning Over recent years, a growing body of work has explored this idea across synthetic settings (e.g., the Countdown game ( Katz, Kokel and Sreedharan, 2025 )), real-world math problems, and general reasoning tasks. From Fixed Parallelism to Adaptive Control Existing approaches show that parallel reasoning can help, but most of them still decide the parallel structure outside the model rather than letting the model choose it. Simple fork-and-join. Self-consistency/Majority Voting — independently sample multiple complete reasoning traces, extract final answer from each, and return the most common one ( Wang et al., 2023 ). Best-of-N (BoN) — similar to self-consistency, but uses a trained verifier to select the best solution instead of using majority voting ( Stiennon et al., 2022 ). Although simple to implement, these methods often incur redundant computation across branches since trajectories are sampled independently. Heuristic-based structured search. Tree / Graph / Skeleton of Thoughts — a family of structured decomposition methods that explores multiple alternative “thoughts” using known search algorithms (BFS/DFS) and prunes via LLM-based evaluation ( Yao et al., 2023 ; Besta et al., 2024 ; Ning et al., 2024 ). Monte-Carlo Tree Search (MCTS) — estimates node values by sampling random rollouts and expands the search tree with Upper Confidence Bound (UCB) style exploration-exploitation ( Xie et al., 2024 ; Zhang et al., 2024 ). These methods improve upon simple fork-and-join by decomposing tasks into non-overlapping subtasks; however, they require prior knowledge about the decomposition strategy, which is not always known. Recent variants. ParaThinker — trains a model to run in two fixed stages: first generating multiple reasoning threads in parallel, then synthesizing them. They introduce trainable control tokens ( ) and thought-specific positional embeddings to enforce independence during reasoning and controlled integration during summarization via a two-phase attention mask ( Wen et al., 2025 ). GroupThink — multiple parallel reasoning threads can see each other’s partial progress at token level and adapt mid-generation. Unlike prior concurrent methods that operate on independent requests, GroupThink runs a single LLM producing multiple interdependent reasoning trajectories simultaneously ( Hsu et al., 2025 ). Hogwild! Inference — multiple parallel reasoning threads share KV cache and decide how to decompose tasks without an explicit coordination protocol. Workers generate concurrently into a shared attention cache using RoPE to stitch together individual KV blocks in different orders without recomputation ( Rodionov et al., 2025 ). Figure 2: Various Strategies for Parallel Reasoning Figure 2: Various Strategies for Parallel Reasoning The methods above share a common limitation: the decision to parallelize, the level of parallelization, and the search strategy are imposed on the model, regardless of whether the problem actually benefits from it. However, different problems need different levels of parallelization, and that is something critical to the effectiveness of parallelization. For example, a framework that applies the same parallel structure to “What’s 25+42?” and “What’s the smallest planar region in which you can continuously rotate a unit-length line segment by 180°?” is wasting compute on the former and probably using the wrong decomposition strategy for the latter. In the approaches described above, the model is not taught this adaptive behavior. A natural question arises: What if the model could decide for itself when to parallelize, how many threads to spawn, and how to coordinate them based on the problem at hand? Adaptive Parallel Reasoning (APR) answers this question by making parallelization part of the model’s generated control flow. Formally defined, adaptivity refers to the model’s ability to dynamically allocate compute between parallel and serial operations at inference time . In other words, a model with adaptive parallel reasoning (APR) capability is taught to coordinate its control flow — when to generate sequences sequentially vs. in parallel. It’s important to note that the concept of adaptive parallel reasoning was introduced by the work Learning Adaptive Parallel Reasoning with Language Models ( Pan et al., 2025 ), but is a paradigm rather than a specific method. Throughout this post, APR refers to the paradigm, while “ the APR method ” denotes the specific instantiation from Pan et al. (2025). This shift matters for three reasons. Compared to Tree-of-Thoughts, APR doesn’t need domain-specific heuristics for decomposition. During RL, the model learns general decomposition strategies from trial and error. In fact, models discover useful parallelization patterns, such as running the next step along with the self-verification of a previous step, or hedging a primary approach with a backup one, in an emergent manner that would be difficult to hand-design ( Yao et al., 2023 ; Wu et al., 2025 ; Zheng et al., 2025 ). Compared to BoN, APR avoids redundant computation. APR models have control over what each parallel thread will do before branching out. Therefore, APR can learn to produce a set of unique, non-overlapping subtasks before assigning them to independent threads ( Wang et al., 2023 ; Stiennon et al., 2022 ; Pan et al., 2025 ; Yang et al., 2025 ). Compared to non-adaptive approaches, APR can choose not to parallelize. Adaptive models can adjust the level of parallelization to match the complexity of the problem against the complexity and overhead of parallelization ( Lian et al., 2025 ). In practice, this is implemented by having the model output special tokens that control when to reason in parallel versus sequentially. Below is a condensed ThreadWeaver-style trace: two outlines and two paths under a block, then the threads agree on a single boxed answer. Figure 3: Example of an Adaptive Parallel Reasoning Trajectory from ThreadWeaver, manually condensed for ease of illustration. Figure 3: Example of an Adaptive Parallel Reasoning Trajectory from ThreadWeaver, manually condensed for ease of illustration. Figure 4: Special Tokens Variants across Adaptive Parallel Reasoning Papers Figure 4: Special Tokens Variants across Adaptive Parallel Reasoning Papers Inference Systems for Adaptive Parallelism How do we actually execute parallel branches? We take inspiration from computer systems, and specifically, multithreading and multiprocessing. Most of this work can be viewed as leveraging a fork-join design. At inference time, we are effectively asking the model to perform a map-reduce operation: Fork the problem into subtasks/threads, process them concurrently Join them into a final answer Figure 5: Fork-join Inference Design Figure 5: Fork-join Inference Design Specifically, the model will encounter a list of subtasks. It will then prefill each of the subtasks and send them off as independent requests for the inference engine to process. These threads then decode concurrently until they hit an end token or exceed max length. This process blocks until all threads finish decoding and then aggregates the results. This is common across various adaptive parallel reasoning approaches. However, one issue arises during aggregation: the content generated in branches cannot be easily aggregated at the KV cache level. This is because tokens in independent threads start at identical position IDs, resulting in encoding overlap and non-standard behavior when merging KV cache back together. Similarly, since independent threads do not attend to each other, their concatenated KV cache results in a non-causal attention pattern, which the base model has not seen during training. To address this issue, the field splits into two schools of thought on how to execute the aggregation process, defined by whether they modify the inference engine or work around it. Multiverse modifies the inference engine to reuse KV cache across the join. Before taking a deeper look into Multiverse ( Yang et al., 2025 )’s memory management, let’s first understand how KV cache is handled up until the “join” phase. Notice how each of the independent threads share the prefix sequence, i.e., the list of subtasks. Without optimization, each thread needs to prefill and recompute the KV cache for the prefix sequence. However, this redundancy can be avoided with SGLang ’s RadixAttention ( Sheng et al., 2023 ), which organizes multiple requests into a radix tree, a trie (prefix tree) with sequences of elements of varying lengths instead of single elements. This way, the only new KV cache entries are those from independent thread generation. Figure 6: RadixAttention’s KV Cache Management Strategy Figure 6: RadixAttention’s KV Cache Management Strategy Now, if everything went well, all the independent threads have come back from the inference engine. Our goal is now to figure out how to synthesize them back into a single sequence to continue decoding for next steps. It turns out, we can reuse the KV cache of these independent threads during the synthesis stage. Specifically, Multiverse ( Yang et al., 2025 ), Parallel-R1 ( Zheng et al., 2025 ), and NPR ( Wu et al., 2025 ) modify the inference engine to copy over the KV cache generated by each thread and edits the page table so that it stitches together non-contiguous memory blocks into a single KV cache sequence. This avoids the redundant computation of a second prefill and reuses existing KV cache as much as possible. However, this has several major limitations. First, this approach requires modifying the inference engine to perform non-standard memory handling, which can result in unexpected behaviors. Specifically, since the synthesis request references KV cache from previous requests, it creates fragility in the system and the possibility of bad pointers. Another request can come in and evict the referenced KV cache before the synthesis request completes, requiring it to halt and trigger a re-prefilling of the previous thread request. This problem has led the Multiverse researchers ( Yang et al., 2025 ) to limit the batch size that the inference engine can handle, which restricts throughput. Figure 7: KV Cache “Stitching” During Multiverse Inference Figure 7: KV Cache “Stitching” During Multiverse Inference Second, this approach modifies how models see the sequence, which creates a distributional shift that models are not pretrained on, therefore requiring more extensive training to align behavior. Specifically, when we stitch together KV cache this way, we create a sequence with non-standard position encoding. During independent-thread generation, all threads started at the same position index and attended to the prior subtasks, NOT each other. So when the threads merge back, the resulting KV cache has a non-standard positional encoding and does not use causal attention. Therefore, this approach requires extensive training to align the model to this new behavior. To address this, Multiverse ( Yang et al., 2025 ) and related works apply a modified attention mask during training to prevent independent threads from attending to each other, aligning the training and inference behaviors. Figure 8: Multiverse’s Attention Mask Figure 8: Multiverse’s Attention Mask With these issues arising from non-standard KV cache management, can we try an approach without engine modifications? ThreadWeaver keeps the inference engine unchanged and moves orchestration to the client. ThreadWeaver ( Lian et al., 2025 ) treats parallel inference purely as a client-side problem. The “Fork” process is nearly identical to Multiverse’s, but the join phase handles memory very differently as it does NOT modify engine internals. Instead, the client concatenates all text outputs from independent branches into one contiguous sequence. Then, the engine performs a second prefill to generate the KV cache for the conclusion generation step. While this introduces computational redundancy that Multiverse tries to avoid, the cost of prefill is significantly lower than decoding. In addition, this does not require special attention handling during inference, as the second prefill uses causal attention (threads see each other), making it easier to adapt sequential autoregressive models for this task. Figure 9: ThreadWeaver’s Prefill and Decode Strategy Figure 9: ThreadWeaver’s Prefill and Decode Strategy How should we train a model to learn this behavior? Naively, for each parallel trajectory, we can break it down into multiple sequential pieces following our inference pattern. For instance, we would train the model to output the subtasks given prompt, individual threads given prompt+subtask assignment, and conclusion given prompt+subtasks+corresponding threads. However, this seems redundant and not compute efficient. Can we do better? Turns out, yes. As in ThreadWeaver ( Lian et al., 2025 ), we can organize a parallel trajectory into a prefix-tree (trie), flatten it into a single sequence, and apply an ancestor-only attention mask during training (not inference!). Figure 10: Building the Prefix-tree and Flattening into a single training sequence Figure 10: Building the Prefix-tree and Flattening into a single training sequence Specifically, we apply masking and position IDs to mimic the inference behavior, such that each thread is only conditioned on the prompt+subtasks, without ever attending to sibling threads or the final conclusion. The engine-agnostic design makes adoption easy since you don’t need to figure out a separate hosting method and can leverage existing hardware infra. It also gets better as existing inference engines get better. What’s more, with an engine-agnostic method, we can serve a hybrid model that switches between sequential and parallel thinking modes easily. Training Models to Use Parallelism Once the inference path exists, the next problem is teaching a model to use it. Demonstrations are needed because the model must learn to output special tokens that orchestrate control flow. We found the instruction-following capabilities of base models insufficient for generating parallel threads. An interesting question here is: does SFT training induce a fundamental reasoning capability for parallel execution that was previously absent, or does it merely align the model’s existing pre-trained capabilities to a specific control-flow token syntax. Typical wisdom is SFT teaches new knowledge; but contrary to common belief, some papers—notably Parallel-R1 ( Zheng et al., 2025 ) and NPR ( Wu et al., 2025 )—argue that their SFT demonstrations simply induce format following (i.e., how to structure parallel requests). We leave this as future work. Figure 11: Sources of Parallelization Demonstration Data Figure 11: Sources of Parallelization Demonstration Data Demonstrations teach the syntax of parallel control flow, but they do not fully solve the incentive problem. In an ideal world, we only need to reward the outcome accuracy, and the parallelization pattern emerges naturally given that it learns to output special tokens through SFT, similar to the emergence of long CoT. However, researchers ( Zheng et al., 2025 ) observed that this is not enough, and we do in fact need parallelization incentives. The question then becomes, how do we tell when the model is parallelizing effectively? Structure-only rewards are too easy to game. Naively, we can give a reward for the number of threads spawned. But models can spawn many short, useless threads to hack the reward. Okay, that doesn’t work. How about a binary reward for simply using parallel structure correctly? This partially solves the issue of models spamming new threads, but models still learn to spawn threads when they don’t need to. The authors of Parallel-R1 ( Zheng et al., 2025 ) introduced an alternating-schedule, only rewarding parallel structure 20% of the time, which successfully increased the use of parallel structure (13.6% → 63%), but had little impact on overall accuracy. With this structure-only approach, we might be drifting away from our original goal of increasing accuracy and reducing latency… How can we optimize for the Pareto frontier directly? Accuracy is simple — we just look at the outcome. How about latency? Efficiency rewards need to track the critical path. In sequential-only trajectories, we can measure latency based on the total number of tokens generated. To extend this to parallel trajectories, we can focus on the critical path, or the longest sequence of tokens that are causally dependent, as this directly determines our end-to-end generation time (i.e., wall-clock time). As an example, when there are two sections with five threads each, the critical path will go through the longest thread from the first parallel section, then any sequential tokens, then the longest thread from the second parallel section, and so on until the end of sequence. Figure 12: Critical Path Length Illustration Figure 12: Critical Path Length Illustration The goal is to minimize the length of the critical path. Simultaneously, we would still like the model to be spending tokens exploring threads in parallel. To combine the two objectives, we can focus on making the critical path a smaller fraction of the total tokens spent. Authors of ThreadWeaver ( Lian et al., 2025 ) framed the parallelization reward as $1 - L_{\mathrm{critical}} / L_{\mathrm{total}}$, which is 0 for a sequential trajectory, and increases linearly as the critical path gets smaller compared to the total tokens generated. Parallel efficiency should be gated by correctness. Intuitively, when multiple trajectories are correct we should assign more reward to the trajectories that are more efficient at parallelization. But how about when they are all incorrect? Should we assign any reward at all? Probably not. To formalize this, $R = R_{\mathrm{correctness}} + R_{\mathrm{parallel}}$. Assuming binary outcome correctness, this can be written as $R = \mathbf{1}(\text{Correctness}) + \mathbf{1}(\text{Correctness}) \times (\text{some parallelization metric})$. This way, a model only gets a parallelization reward when it answers correctly, since we don’t want to pose parallelization constraints on the model if it couldn’t answer the question correctly. Figure 13: Differences in Reward Designs Across Adaptive Parallel Reasoning Works Figure 13: Differences in Reward Designs Across Adaptive Parallel Reasoning Works Evaluation and Open Questions When all is said and done, how well do these adaptive parallel methods actually perform? Well…this is a hard question, as they differ in model choice and metrics. The model selection depends on the training method, SFT problem difficulty, and sequence length. When running SFT on difficult datasets like s1k, which contains graduate-level math and science problems, researchers chose a large base model (Qwen2.5 32B for Multiverse ( Yang et al., 2025 )) to capture the complex reasoning structure behind the solution trajectories. When running RL, researchers chose a small, non-CoT, instruct model (4B, 8B) due to compute cost constraints. Figure 14: Difference in Model Choice Across Adaptive Parallel Reasoning Papers Figure 14: Difference in Model Choice Across Adaptive Parallel Reasoning Papers Each paper also offers a slightly different interpretation about how adaptive parallel reasoning contributes to the research field. They optimize for different theoretical objectives, so they use slightly different sets of metrics: Multiverse and ThreadWeaver ( Yang et al., 2025 ; Lian et al., 2025 ) aim to deliver sequential-AR-model-level accuracy at faster speeds. Multiverse shows that APR models can achieve higher accuracy under the same fixed context window, while ThreadWeaver shows that the APR model achieves shorter end-to-end token latency (critical path length) while getting comparable accuracy. NPR ( Wu et al., 2025 ) treats sequential fallback as a failure mode and optimizes for 100% Genuine Parallelism Rate, measured as the ratio of parallel tokens to total tokens. Parallel-R1 ( Zheng et al., 2025 ) does not focus on end-to-end latency and instead optimizes for exploration diversity, presenting APR as a form of mid-training exploration scaffold that provides a performance boost after RL. Open Questions While Adaptive Parallel Reasoning represents a promising step toward more efficient inference-time scaling, significant open questions remain. As noted above, Parallel-R1 ( Zheng et al., 2025 ) presents APR as a form of mid-training exploration scaffold rather than a primarily inference-time technique. This invites a more fundamental question: Does parallelization at inference-time consistently improve accuracy, or is it primarily valuable as a training-time exploration scaffold? Parallel-R1 suggests that the diversity induced by parallel structure during RL may matter more than the parallelization itself at test time. A related concern is stability. There’s also a persistent tendency for models to collapse back to sequential reasoning when parallelization rewards are relaxed. Parallel-R1 authors showed that removing parallelization reward after 200 steps results in the model reverting to sequential behavior. Is this a training stability issue, a reward signal design issue, or evidence that parallel structure genuinely conflicts with how autoregressive pretraining shapes the model’s prior? Beyond whether APR works, deployment introduces its own questions. Can we design training methods that account for available compute budget at inference time, so parallelization decisions are hardware-aware rather than purely problem-driven? Finally, the parallel structures considered above are essentially flat. What if we allow parallelization depth > 1? Recursive language models (RLMs; Zhang, Kraska and Khattab, 2026 ) effectively manage long context and show promising inference-time scaling capabilities. How well do RLMs perform when trained with end-to-end RL that incentivizes adaptive parallelization? Acknowledgements We thank Nicholas Tomlin and Alane Suhr for providing us with helpful feedback. We thank Christopher Park, Karl Vilhelmsson, Nyx Iskandar , Georgia Zhou, Kaival Shah , and Jyoti Rani for their insightful suggestions. We thank Vijay Kethana , Jaewon Chang , Cameron Jordan , Syrielle Montariol , Erran Li, and Anya Ji for their valuable discussions. We thank Jiayi Pan , Xiuyu Li , and Alex Zhang for their constructive correspondences about Adaptive Parallel Reasoning and Recursive Language Models.
Congratulations to the Berkeley Artificial Intelligence Research (BAIR) Lab class of 2026! This year, BAIR celebrates another remarkable group of Ph.D. graduates whose curiosity, creativity, and perseverance have pushed the frontiers of artificial intelligence and machine learning. Their work spans the breadth of modern AI — robotics and embodied intelligence, large language models and reasoning, computer vision, generative modeling, AI safety, human-AI interaction, AI for science and healthcare, and much more. Along the way, they have published influential research, built systems with real-world impact, mentored their peers, and shaped the BAIR community for the better. Now they are headed everywhere ideas travel: to faculty and postdoctoral positions, to industry research labs, and to startups of their own founding — and several are still exploring what comes next and would love to hear from you. Please join us in celebrating the achievements of these wonderful graduates. We are proud of everything they have accomplished at Berkeley, and we can’t wait to see what they do next! Thank you to our friends at the Stanford AI Lab for this idea! Baifeng Shi Baifeng Shi Email: baifeng_shi@berkeley.edu Website: https://bfshi.github.io/ Advisor(s): Trevor Darrell Research Blurb: I work on building generalist vision and robotic models. What's next: Member of Technical Staff at Physical Intelligence Charlie Snell Charlie Snell Email: csnell22@berkeley.edu Website: https://sea-snell.github.io Advisor(s): Dan Klein Research Blurb: My work aims to understand when and how the different LLM scaling paradigms can be traded off and interchanged. In particular, test-time scaling treats each prompt independently, drawing long chains of inferences and then forgetting them entirely between prompts. This differs critically from pretraining, which instead learns a compressed representation from a large dataset. I believe bridging the gap between these methods of scaling computation, presents a key open challenge in the field: how can we develop methods which turn the inferences drawn at test-time back into learned representations that the model can hold onto across interactions. Devin Guillory Devin Guillory Email: dguillory@berkeley.edu Website: https://devinguillory.com Advisor(s): Trevor Darrell Research Blurb: Accounting for data shifts in computer vision models What's next: Building collaborative AI systems, looking for conspirators. Eve Fleisig Eve Fleisig Email: efleisig@berkeley.edu Website: https://efleisig.com Advisor(s): Dan Klein Research Blurb: I design language models to work reliably and fairly for the broad range of real LLM users. First, my research leverages disagreement among user preferences as signal, in order to train and evaluate LLMs for entire populations of users. Second, I work on designing rigorous evaluations to extricate challenging LLM harms that diverse users face. Finally, I work on core technical failures of LLMs, like miscalibrated confidence, to reduce downstream risks when models are deployed to users with different needs. Combined, these interventions facilitate building LLMs that minimize societal harms, and maximize benefits to a wider range of real-world users. What's next: Postdoctoral fellow at Princeton CITP Grace Luo Grace Luo Email: graceluo@berkeley.edu Website: https://graceluo.net Advisor(s): Trevor Darrell Research Blurb: My research is on interpreting and controlling generative models. For example, I've worked on re-purposing image generators for computer vision tasks, and meta-modeling language activations for better LLM probing and steering. What's next: Research scientist in industry Hanlin Zhu Hanlin Zhu Email: hanlinzhu@berkeley.edu Website: https://hanlinzhu.com/ Advisor(s): Stuart Russell, Jiantao Jiao Research Blurb: My research centers on understanding and improving the reasoning capabilities of large language models (LLMs). What's next: Member of Technical Staff at OpenAI Haozhi Qi Haozhi Qi Email: hqi@berkeley.edu Website: https://haozhi.io/ Advisor(s): Jitendra Malik, Yi Ma Research Blurb: Dexterous Manipulation and Robot Learning What's next: Research scientist at Amazon; Faculty at University of Chicago J.D. Zamfirescu-Pereira J.D. Zamfirescu-Pereira Email: zamfi@berkeley.edu Website: https://zamfi.net Advisor(s): Bjoern Hartmann Research Blurb: My research focuses on effective human-AI co-design. I study the boundaries of language interfaces as a medium for interacting with AI, creating systems that blend language-focused interactions with structured user interfaces that draw on different levels of abstraction. I focus on language-oriented technologies, like LLMs and text-to-image models, that are powerful mediators of design processes. These technologies enable humans to describe their desires at almost any level of abstraction, from high-level goals vaguely specified (“I’d like a game to help my kid learn to read”) to low-level corrections of undesired outputs (“Don’t say ‘I know because I’ve tasted it’ when about a recipe substitution's taste”). What's next: Assistant Professor, Computer Science, UCLA Jiachen Lian Jiachen Lian Email: jiachenlian@berkeley.edu Website: https://jlian2.github.io Advisor(s): Gopala Anumanchipalli Research Blurb: My research focuses on human-centered AI across speech, healthcare, and systems. Looking for: Look for AI talents to join our startup Josh Kang Josh Kang Email: minwoo_kang@berkeley.edu Website: https://joshuaminwookang.github.io/ Advisor(s): John Canny Research Blurb: I study language modeling and related topics in NLP; specific interests are human user simulation and building conversational, collaborative AI agents. What's next: AI Scientist at Mistral AI Junhao (Bear) Xiong Junhao (Bear) Xiong Email: junhao_xiong@berkeley.edu Website: https://www.linkedin.com/in/junhao-bear-xiong Advisor(s): Jennifer Listgarten, Yun Song Research Blurb: Junhao (Bear) Xiong is a PhD candidate at UC Berkeley, advised by Jennifer Listgarten and Yun S. Song. His work focuses on machine learning methods for biology, with an emphasis on generative modeling for proteins. Previously, he studied Applied Math and Computer Science at Johns Hopkins. Looking for: Research scientist Kaylo Littlejohn Kaylo Littlejohn Email: kaylo_littlejohn@berkeley.edu Website: https://kaylolittlejohn.com Advisor(s): Gopala Anumanchipalli Research Blurb: My research is focused on speech modeling and natural language processing. I co-led the development of multimodal AI tools to accurately translate brain activity into text, audible personalized speech, and a high-fidelity "digital talking avatar" (Nature 2023, Nature Neuroscience 2025). I am also tech lead for voice modeling at Roblox. Looking for: Research Scientist / Engineer Kent Chang Kent Chang Email: kentkchang@berkeley.edu Website: https://kentkc.org Advisor(s): David Bamman Research Blurb: I work on NLP and multimodal machine learning, with a focus on evaluating large language models and building multimodal systems for understanding dialogue, narrative, and social interaction. My research includes benchmarks for LLM memorization, multimodal datasets sourced from feature films and television, and studies of model behavior. I'm interested in bridging computational methods with questions from the humanities and social sciences about whose voices get represented in AI systems, and about AI's broader impact. My work has appeared at EMNLP and ACL, among others. Looking for: (teaching) faculty, Research Scientist, ML/AI SWE Kevin Black Kevin Black Email: kvablack@berkeley.edu Website: https://kevin.black Advisor(s): Sergey Levine Research Blurb: I work on large-scale robot learning: including imitation learning, reinforcement learning, generative modeling, real-time control, and whatever else it takes to make robots work in the real world! What's next: Research Scientist of Physical Intelligence Kunhe Yang Kunhe Yang Email: kunheyang@berkeley.edu Website: https://www.kunheyang.com/ Advisor(s): Nika Haghtalab Research Blurb: My research focuses on the theoretical foundations of designing and evaluating AI algorithms in environments shaped by human incentives and AI agency. My work spans human-centric policy learning, incentive-aware evaluation, and multi-agent collaboration and information transmission, drawing on tools from machine learning theory and computational economics. What's next: Postdoc Research at Stanford Lisa Dunlap Lisa Dunlap Email: lisabdunlap@berkeley.edu Website: https://lisabdunlap.com Advisor(s): Joseph Gonzalez, Trevor Darrell Research Blurb: Auditing generative models. What's next: Research Engineer at Anthropic Long (Tony) Lian Long (Tony) Lian Email: longlian@berkeley.edu Website: https://tonylian.com/ Advisor(s): Trevor Darrell, Adam Yala Research Blurb: My research primarily focuses on developing real-time multi-modal multi-agent systems and parallel reasoning systems through end-to-end RL. What's next: Member of Technical Staff at Thinking Machines Lab Maulik Bhatt Maulik Bhatt Email: maulikbhatt@berkeley.edu Website: https://maulikb.com Advisor(s): Negar Mehr Research Blurb: My research develops autonomous robots that can safely coordinate with humans and other robots in shared environments. I build scalable algorithms grounded in game theory and diffusion models that let agents reason about the intent and behavior of others around them. My work spans real-time multi-agent trajectory planning and imitation learning in the presence of multi-modality. I've validated these methods on hardware platforms ranging from quadrotors to manipulators, with the goal of making multi-agent coordination robust, interpretable, and deployable in the real world. What's next: Joining Toyota Woven's end-to-end autonomous driving team. Michael Psenka Michael Psenka Email: psenka@berkeley.edu Website: https://www.michaelpsenka.io/ Advisor(s): Aditi Krishnapriyan Research Blurb: Work in various domains (reinforcement learning, world models, AI+bio/chem), generally working on longer-horizon and out-of-distribution problems in planning and interpolation (e.g. robot manipulation from start state to goal, molecular dynamics of proteins between ground states). My thesis took a variational approach (think calculus of variations) directly from deep generative models of the environment, framing path-finding as minimizing a functional induced by the learned model itself (its score, its critic, or its dynamics). Through my research I've gained insight on how to properly handle dynamics in deep learning systems, and I plan to continue developing systems that are dynamic and adaptive. What's next: Lead Research Scientist at Baseten Nathan Lichtlé Nathan Lichtlé Email: nathan.lichtle@gmail.com Website: https://nathanlichtle.com Advisor(s): Alexandre M. Bayen Research Blurb: RL for autonomous driving. What's next: Chief Scientist & Co-founder at Yumi Health Neerja Thakkar Neerja Thakkar Email: nthakkar@berkeley.edu Website: https://neerja.me/ Advisor(s): Jitendra Malik Research Blurb: My research focuses on scaling predictive world models to handle the complexity of in-the-wild motion. Using autoregressive and diffusion frameworks, I develop better representations for real-world prediction and propose methods to efficiently adapt these models to new domains. Looking for: Research scientist Nikita Mehandru Nikita Mehandru Email: nmehandru@berkeley.edu Website: https://n-mehandru.github.io/ Advisor(s): Ahmed Alaa and David Bamman Research Blurb: My research develops and applies machine learning methods for clinical reasoning and disease progression modeling using unstructured text and time series data from electronic health records. In collaboration with physicians at UCSF, I bridge method development and clinical validation with the intention to build reliable, interpretable AI systems in medicine. Looking for: Research Scientist Niklas Lauffer Niklas Lauffer Email: nlauffer@berkeley.edu Website: https://niklaslauffer.github.io/ Advisor(s): Stuart Russell and Sanjit Seshia Research Blurb: Niklas's research is focused on AI safety and reinforcement learning, particularly in the area of multi-agent interaction and LM agents. He's worked on enabling adversarial learning in cooperative and mixed-motive settings, solving issues of covariate shift in training LM agents on long-horizon tasks, as well as evaluating safety risks posed by LM agents in multi-agent settings. What's next: Research Scientist at Google Deepmind Qiyang Li Qiyang Li Email: qcli@berkeley.edu Website: https://colinqiyangli.github.io/ Advisor(s): Sergey Levine Research Blurb: Recent progress in robotic manipulation policy learning has been largely driven by (1) the increasing availability of large-scale prior datasets and (2) the success of action chunking, where the policy predicts a short sequence of future actions rather than a single one. However, most action chunking policies are trained via supervised imitation learning, because efficient online self-improvement with reinforcement learning (RL) remains challenging—limiting real-world applicability. My PhD research studied how we could leverage prior data to optimize action-chunking policies with RL, combining empirical results with theoretical insights. Looking for: Post-doc/research scientist for RL in robotics and LLMs! Sampada Deglurkar Sampada Deglurkar Email: sampada_deglurkar@berkeley.edu Website: https://sdeglurkar.github.io/ Advisor(s): Prof Claire Tomlin Research Blurb: My research is in providing safety assurances for AI-enabled autonomous systems, ranging from robots to autonomous vehicles to aviation systems. For this, I have worked with uncertainty quantification for machine learning models, decision-making under uncertainty algorithms, and tools for producing probabilistic guarantees on system operation. Looking for: Research scientist, Research engineer Vinamra Benara Vinamra Benara Email: vbenara@berkeley.edu Website: https://cs.berkeley.edu/~vbenara Advisor(s): Ion Stoica Research Blurb: My research focuses on LLM post-training, including data curation, RLHF, RLVR with VLMs, evaluations, reasoning, agentic workflows, and interpretability. I also have strong expertise in systems infrastructure for distributed computing. Looking for: Research scientist / Research Engineer Vongani Maluleke Vongani Maluleke Email: vongani_maluleke@berkeley.edu Website: https://people.eecs.berkeley.edu/~vongani_maluleke/ Advisor(s): Jitendra Malik and Angjoo Kanazawa Research Blurb: Vongani Maluleke is a PhD candidate at UC Berkeley (BAIR, advised by Jitendra Malik and Angjoo Kanazawa), where she led the development of MAGNet, a unified multi-agent motion generation framework that supports a wide range of motion generation tasks without retraining or architectural changes, outperforming task-specialized state-of-the-art baselines. She is currently extending this work by deploying it on a Unitree G1 humanoid to make it embody social intelligence. Before her PhD, she was a Senior AI Consultant at Deloitte, awarded Exceptional Performer two consecutive years, leading AI system development across media, telecommunications, retail, and financial services. Looking for: Research scientist Wei-Jer Chang Wei-Jer Chang Email: weijer_chang@berkeley.edu Website: https://weijer-chang.github.io/ Advisor(s): Masayoshi Tomizuka Research Blurb: My research focuses on developing safe and intelligent autonomous systems for complex, human-centered environments. I work at the intersection of machine learning, generative models, and reinforcement learning, with applications in autonomy. My work addresses challenges in multi-agent interaction, interactive human behavior, and long-tail safety-critical scenarios at scale. Looking for: Research Scientist, Applied Scientist, Roboticist Xiuyu Li Xiuyu Li Email: xiuyu@berkeley.edu Website: https://xiuyuli.com/ Advisor(s): Kurt Keutzer Research Blurb: My research focuses on developing scalable and self-improving large language model agents, with emphasis on coding agents for complex, long-horizon tasks. This direction builds on my work in parallel reasoning, and on broader expertise in making generative models more efficient in training and inference across language and vision. What's next: Member of Technical Staff at xAI Yichen Xie Yichen Xie Email: yichenxie0928@gmail.com Website: https://yichen928.github.io/ Advisor(s): Masayoshi Tomizuka Research Blurb: My research focuses on building multimodal foundation models and world models that understand and interact with complex physical environments. I aim to develop unified representations across modalities, enabling AI systems to reason over space, time, and dynamics toward general-purpose embodied intelligence. What's next: Research Scientist at Luma AI Yigit Efe Erginbas Yigit Efe Erginbas Email: erginbas@berkeley.edu Website: https://www.linkedin.com/in/erginbas/ Advisor(s): Kannan Ramchandran, Thomas A. Courtade Research Blurb: My PhD research spans two threads: online learning in large-scale markets, and interpretability of large machine learning models. In the first, I work on sequential decision-making with applications to recommendation, pricing, and assortment selection. My focus is on designing algorithms with provable guarantees for welfare maximization, revenue maximization, and stability. In the second, I develop scalable attribution methods that exploit the sparse, low-degree structure of real-world interactions, using tools from signal processing and information theory. More recently, I have been exploring principled ways to evaluate the faithfulness of model self-explanations. What's next: Researcher at Hudson River Trading's AI Labs (HAIL) Yiheng Li Yiheng Li Email: yhli@berkeley.edu Website: https://Yihengli.com Advisor(s): Masayoshi Tomizuka Research Blurb: I am working on vision world modeling, with prior experience in diffusion model's efficiency as well as in autonomous driving. What's next: Research Scientist at Waymo Zhe Fu Zhe Fu Email: zhefu@berkeley.edu Website: https://fu-zhe.com/ Advisor(s): Alexandre Bayen Research Blurb: My research focuses on physics-informed learning and control for mixed-autonomy systems, with applications in transportation. I design physics-informed neural networks to learn solutions of nonlinear partial differential equations, enabling accurate and data-efficient prediction of traffic dynamics. Building on these models, I develop both model-based and learning-based control strategies that coordinate automated vehicles to improve system-level performance. My work bridges machine learning, control, and real-world deployment, and has been validated in large-scale field experiments. More broadly, I aim to advance trustworthy, interpretable AI for decision-making in complex, real-world systems. What's next: I will be an Energy Fellow at Stanford after graduation. Also looking for Faculty, or research scientist positions in AI, control, and autonomy.
... government of the people, by the people, for the people ... — Abraham Lincoln, Gettysburg Address (1863) The cost of AI is dropping rapidly. GPT-4-class capabilities cost roughly $30 per million tokens in early 2023; today the same runs under $1 , and some providers are pushing costs below $0.10 . Across benchmarks, inference prices have fallen between 9x and 900x per year , with a median decline near 50x. Even frontier models are getting dramatically cheaper each generation, with open-source models following closely behind. And crucially, even if “Nobel-Prize-winning genius-level” intelligence isn’t here yet, the intelligence that suffices for the vast majority of knowledge work is here today, and getting cheaper by the month. At this rate, we are soon entering the era of virtually free intelligence —the kind that is more than enough for everyday knowledge work. A cartoon database character and an AI robot agent holding hands Disclosure: This post is a perspective led by Aditya G. Parameswaran —an Associate Professor of EECS and co-director of the EPIC Data Lab at UC Berkeley—together with his collaborators. It is part landscape survey and part perspective, and several of the research directions discussed below (including agentic speculation, structured memory, and synthesizing custom data systems from scratch) draw on the authors' own ongoing work. So, what does this new era of near-free intelligence mean for data systems? We believe three new challenges—and opportunities—stem from near-zero inference costs: Data Systems For Agents. Agents will soon become the dominant workload for data systems—with swarms of agents spun up in response to each end-user request. Given differences in characteristics between agents and humans—or applications acting on their behalf— how should we redesign data systems for such agentic users? Data Systems Of Agents. As agents start taking on the bulk of knowledge work, a new substrate is needed for thousands of agents to manage state over long-running tasks, coordinate and reach consensus, and deal with failures. What do data systems that reliably and efficiently run and manage agent swarms look like? Data Systems By Agents. Agents are rapidly becoming capable of synthesizing entire data systems in one go—meaning we can rebuild custom systems for each new workload. Verifying that such systems match intended behavior is a challenge. What does it take to let agents synthesize data systems we can actually trust? A database character and a robot agent holding up a triangle labeled 'of', 'for', and 'by' Data Systems For, Of, and By Agents Next, we will discuss each in more detail, followed by discussing the intertwined future of data systems and agents, especially as the three challenges intersect. Data Systems For Agents An agent querying a database doesn’t behave like a person or a BI tool. It performs what we call agentic speculation : a high-volume, heterogeneous stream of work spanning schema introspection, columnar exploration, partial and then full query formulation. With multiple agents each exploring portions of the hypothesis space, each user request could amount to 1000s of individual SQL queries. Now, users can issue ‘high-level’ data tasks, e.g., root-cause analysis—e.g., ‘why did coffee sales in Berkeley drop this year’—or exploratory cohort analysis—e.g., ‘which user segments are most likely to churn next quarter’—each involving a combinatorial space of potential joins, aggregations, and filter combinations. An agent sending many SELECT SQL queries to a database and receiving results back Data Systems Redesigned to More Effectively Support Agentic Speculation The requests from these agents have various opportunities for optimization. For instance, on a text-to-SQL benchmark with multiple agents attempting each task, only 10-20% of the sub-plans are distinct. Thus, 80-90% of sub-queries perform duplicate work. The same experiments show task success rates significantly increasing with more agentic attempts—so the redundancy is actually helpful. But from the data system perspective it’s wasted work. An agent-first data system can exploit such properties to help agents make progress faster. It can reuse results across overlapping sub-plans, drawing on ideas from decades-old literature on multi-query optimization and shared scans . Or the data system can try to satisfice , returning approximate answers that are good enough for agents to make progress, leveraging work from the AQP literature —or streaming the results of the final or intermediate operators to help agents decide if seeing the rest is necessary or helpful. Another opportunity here is to rethink the query interface entirely: instead of agents issuing a single SQL query at a time, they could instead issue a batch of queries, each with its own approximation requirements. Since enumerating an exponential search space (as in the root cause or cohort analysis examples above) isn’t a good use of agentic reasoning ability, perhaps data systems should support higher-level primitives rather than requiring agents to list each SQL query explicitly. One idea here is to draw on DBT-style Jinja macros to provide looping-based primitives for agents to interact with data systems. A swarm of AI agents working at laptops A Caffeinated Army of Agents Ready to Tirelessly Complete Your Data Tasks A final opportunity here is to stop thinking of data systems as passive executors of queries; data systems could be proactive , as they possess more grounding in data and system characteristics that agents may lack a priori—they could steer agents in different directions, provide results for related queries, and also provide performance-level feedback (e.g., instead of executing an expensive query, the system could first provide the agent a latency estimate). The reason we can do this now as opposed to the past is that an agent can accept any form of textual feedback and isn’t expecting a strict SQL query result. In fact, the data system could also prepare both materialized and virtual views for an agent in advance, provided to the agent as part of context, as this may be cheaper or more effective than having an agent author or use them. Data Systems Of Agents Previously, we focused on how agents interact with data systems. Now, we consider everything else agents need to keep working: where they live, how they remember, how they coordinate with each other, and how they deal with failures of each other. This agentic substrate is separate from the inference stack powering raw intelligence. However, the inference stack itself is being abstracted away through APIs (e.g., from OpenAI or Anthropic), or, for open-weight models, through serving frameworks that hide low-level details. So far, the agentic substrate has been managed through harnesses like Claude Code and Codex , coupled with various mechanisms to store and retrieve memory. First, on the memory front, the current wisdom is that files are all you need ; agents write to unstructured markdown (MD) files, which can then be searched using grep, or via embedding-based retrieval. In fact, many argue that the solution to continual learning is having agents consume a lot (e.g., an entire codebase, slack, company wikis, …) and then write their learnings into MD files, which are then retrieved selectively on demand. Indeed, file systems, bash scripting, and MD files are and will still be important for agents. However, at scale, when agents are doing the vast majority of knowledge work, this approach will no longer be effective. Given limited context windows, retrieving all MD file fragments that may be relevant and stuffing it into the context will break down at some point. Even if context windows continue to grow, there are latency benefits to not put all information into context — and in many cases, e.g., when knowledge work involves interacting with large databases or code bases, it will be infeasible to serialize all relevant data into context. A swarm of robot agents holding hands, each drawing state from a single large shared database platform below them Data Systems As A Substrate for Multi-Agent Swarms One could use a knowledge graph representation , but knowledge graphs suffer from the same limitations as unstructured MD-based memory due to their lack of structured search. What one needs is to be able to retrieve only memory that is pertinent to the task, across multiple attributes (or facets) of interest. For example, an agent debugging a flaky test should be able to pull only the memories tagged with the relevant module, language, framework, and failure mode—rather retrieving based on keywords or embedding similarity. A separate issue is what to actually retrieve; raw agent traces with mistakes are not very useful as they will induce agents to repeat the same mistake—instead, we want the retrieved memory to be corrective. We recently explored a related notion of structured memory , where we organize memory across various attributes, each of which could be set as * to indicate universal applicability, or set as a list of values to be matched. For a data agent, the dimensions could include the columns and tables, type of operation, and finally, open-ended natural-language corrective instructions. So, we could include memory that only applies to a given type of operation (e.g., ‘when performing date-time operations, use fiscal year as opposed to calendar year conventions’), or a given table (e.g., ‘column product_cleaned is preferred over column product when querying on product name’). One open question is defining an application-specific structured memory —or what others have called world models for memory . We believe this is akin to defining a schema for each application—and perhaps agents themselves can help us define and refine it over time. Diagram showing corrective knowledge stored with structured attributes (SQL keywords, tables, columns, data type) and retrieved by matching the features of a new agent query One Possible Way To Store and Retrieve Structured Knowledge [From Here] Structured memory will be useful also for evolutionary frameworks to effectively manage search spaces. Indeed, storing, structuring, and mining large volumes of single and multi-agent traces can help future agents become much more efficient—potentially enabling effective recursive self-improvement through structured memory-based mechanisms. Another challenge is to support concurrent edits to shared memory, and concurrent edits in general, when there are many agents performing transformations. While there have been some useful attempts at supporting multiversioning and copy-on-write semantics , it isn’t clear that such techniques will suffice when thousands of agents are attempting to edit shared state at the same time. For instance, when agents are trying various potential transactions in response to a user request, the effects of the vast majority of these transactions need to be rolled back—with only the one ‘correct’ transaction’s result persisting. Work on supporting exactly-once semantics is relevant here, as are underlying techniques based on CRDTs and operational transformation. For updates to fuzzy mechanisms such as memory, we may be able to sacrifice on consistency for perfect correctness in the interest of latency. While agents can reason about semantics to compensate or roll back their actions to eventually finalize most tasks, the primary challenge lies in the degree to which they step on each other’s toes during the process. An important failure mode to be avoided is a form of “livelock,” where incessant compensating actions prevent any meaningful progress. Beyond shared state, other concerns emerge when trying to support an army of agents, including what to do when agents fail, how agents should communicate with each other (directly or through intermediate shared state), and how we should deal with straggler agents. There have been some developments in supporting durable multi-agent execution, such as Temporal , but it remains to be seen if such solutions will apply at scale across thousands of agents. On the topic of communication, we need mechanisms to enable agents to negotiate with each other. Imagine four developer agents attempting to reach consensus on a shared schema, with distinct but overlapping objectives. In a human setting, this would involve iterative discussion and compromise; for agentic swarms, we must define the mechanisms that allow them to converge on a design that reflects the underlying goals of their respective principals. Or if agents are all requiring access to a limited resource, again communication will be necessary. It remains to be seen if this is best done via centralized coordination, or if a decentralized approach is necessary. Data Systems By Agents Finally, if intelligence is effectively free, then we can employ this intelligence to synthesize new data systems from scratch. Indeed, in many settings, general-purpose data systems may be overkill, as they have to support every schema, query, and hardware target. Given a workload, recent work, including Bespoke OLAP and GenDB , has shown that one can use an agentic pipeline to synthesize a complete, workload-specific analytical engine—in minutes to a few hours, at a cost of a few dollars. The engines are disposable: when the workload shifts, one can simply regenerate them. Analogously, our work has shown that one can synthesize custom key-value stores from scratch, targeted to the workload. In fact, modern IDEs, such as Kiro , elevate specifications for systems development to be a first-class citizen. A robot agent with a hammer and chisel carving a database character out of a block of stone Agents Can Synthesize Custom Data Systems From Scratch The main issue, however, is that specifications are typically imperfect, and don’t cover all corner cases. Present-day agents will exploit the missing specifications to reward-hack their way to a high performance metric. In our custom key-value store work, we found that one way to alleviate this is to have auxiliary verification agents trying to generate test cases that catch the exploitation of corner cases, essentially expanding the specification. Yet another approach is to both generate a system and a proof for its correctness together, for which we have found some early success , but more needs to be done to solidify the approach. Further, it remains to be seen what is the best way to solicit human-written specifications for a system—can this be done in an iterative, human-in-the-loop manner, as opposed to a one-shot, incomplete one. Indeed, human-written specifications are incomplete even for manually authored software, so one would expect that future agents that are more aligned will increasingly exercise better judgement when making design decisions. Pipeline diagram where a system builder provides a specification, planner and coder agents generate code, the code is evaluated for correctness and performance, and critic and auditor agents provide feedback and catch reward hacking One Possible Data System Synthesis Pipeline [From Here] Other questions here involve testing whether starting from a mature system (e.g., Postgres) and removing components/functionality can lead to higher performance or more user trust. Separately, is there an opportunity to make the design composable, comprising various verified components that are mixed and matched given a workload? For example, perhaps the workload hasn’t changed enough for the storage layer to be updated, but perhaps the query optimizer requires changes. A perhaps more viable proposition involves employing agents coupled with proof systems to target critical parts of the code associated with formal proofs, rather than doing so for the entire system. A final opportunity here is to move away from the traditional data systems stack with clearly-defined interfaces (e.g., parser, query optimizer, storage manager, …) — that were each largely the prerogative of a single human team to manage. Instead, agents can find new ways to “blend” these components together, perhaps identifying new optimization opportunities as a result. Agents can also fill in missing gaps in functionality to make existing systems much more feature-complete, or reach feature-parity with other competing systems—or analogously, continuously refining open-source systems in response to feature requests or issues (perhaps filed by other agents!) Doing so in a way that prioritizes correctness, long-term maintenance, and human interpretability will be a challenge. Looking Further Ahead In the era of near-free intelligence, data systems matter more than ever. As agents take on the bulk of knowledge work, the workload for data systems will change, the substrate they need to run on will have to be built, and increasingly, they will participate in designing data systems themselves. Each of these shifts opens up a new, exciting research agenda. A half-database, half-robot character next to a yin-yang symbol formed by a database and a robot agent Co-Evolution of Data Systems and Agents Looking further out, the boundaries between agents and data systems will likely start to blur. For instance, agents may design the data systems they themselves run on, defining both the interfaces as well as the system components underneath. Both the interfaces and internals can be evolved over time by agents in a form of recursive self-improvement. There is also an opportunity to rethink data systems as a holistic source of truth for the entirety of relevant state: including raw data, memory, and coordination state, further erasing the distinctions between the data that is being queried by agents and data generated as a result of agentic activity. Finally, data systems may themselves incorporate agentic components, fundamentally evolving from passive computation engines into intelligent, proactive, self-optimizing architectures. It is hard to predict what the future may hold. We’re in for a wild ride! Acknowledgments The perspective and ongoing work described in this post are the product of joint research and many discussions with wonderful collaborators at the EPIC Data Lab , Data Systems & Foundations group, and the broader Berkeley AI-Systems community. Thank you all! BibTex for this post: @misc { intelligence-is-free-blog , title = {Intelligence is Free, Now What? Data Systems for, of, and by Agents} , author = {Aditya G. Parameswaran and Shubham Agarwal and Kerem Akillioglu and Shreya Shankar and Sepanta Zeighami and Rishabh Iyer and Matei Zaharia and Alvin Cheung and Natacha Crooks and Joseph Gonzalez and Joseph Hellerstein and Ion Stoica} , howpublished = {\url{https://bair.berkeley.edu/blog/2026/07/07/intelligence-is-free-now-what/}} , year = {2026} }
ABBEL overview Overview of ABBEL compared to traditional recursive summarization. Beliefs replace the full interaction history as the agent’s working context, and belief grading improves performance by supervising the contents of each belief state.. As task horizons grow, LLM contexts can’t scale forever. Self-summarization enables concise, interpretable contexts, but at a significant performance cost, especially for human assistance domains where high quality data is scarce, e.g., collaborative code generation. We address this with ABBEL : a framework that isolates and supervises the information content of summaries in the form of natural-language belief states. Motivation: the cost of recursive summarization For language models to effectively assist with increasingly complex tasks such as software development, they must be able to interact with us over hundreds or even thousands of steps. For such long tasks, it is impractical to keep the history of the entire interaction in context. The heuristic approach used so far has been summary generation, sometimes called context compaction. For example, Cursor’s latest model composer 2.5 uses compaction during training for improved performance ( Cassano et al., 2026 ). Alongside composer, Grandcode ( DeepReinforce et al., 2026 ), the first system to consistently beat all human competitors in online coding competitions, despite using one of the newest efficient attention models (Qwen 3.5-397B), 1 still found it necessary to employ context summarization. But compaction has a problem. Despite seemingly low performance gaps in benchmarks, model servers like Cursor continue to recommend that users avoid compaction with their coding assistants in the middle of a task ( Heule et al., 2026 ). To understand why, see below the performance over RL fine-tuning of a Context summary model compared to full context models in Combination Lock, a Wordle-like game that allows up to 16 guesses. 2 Though both model types improve over the course of training, the summary model never closes the gap. Average attempts to guess target word with context summary vs full context policies over training Fig. 1: Average attempts to guess the target word on Combination Lock over RL fine-tuning (lower is better). Context-summary policies improve with training but do not close the gap to full-context policies. Making models self-summarize while completing a task increases the complexity of the learning problem. While this could typically be addressed by training with more data, the performance degradation observed in real world interactive settings likely arises from the difficulty we have in creating and using human simulators effectively to generate high quality training environments ( Lin et al., 2025 , Tomlin et al., 2025 ). Thus, the better you can learn to summarize on the limited and messy multiturn interaction trajectories you can collect, the better off your model will be for downstream users. ABBEL: acting through belief bottlenecks ABBEL belief grading diagram Fig. 2: Autoencoder-inspired belief grading. The model encodes prior belief, action and observation (b t , a t , o t ) into posterior belief b t+1 and is rewarded for how well select information from the history can be reconstructed from that belief. To address poor learning efficiency, we isolate the summary generation task. Drawing inspiration from recursive Bayesian estimation, we formulate summaries as belief states, which we periodically prompt the model to update based on new information. 3 ABBEL overview animation frame ‹ Prev Pause Next › 1 / 16 Fig. 3: ABBEL rollout. Belief updates from the latest observation alternate with action selection conditioned only on the current posterior belief. Belief grading We then extract and supervise the contents of the belief states (Fig. 2, Belief Grading). Belief grading can be thought of as adding an auxiliary RL task, using heuristics designed to capture what makes a good belief as the reward. An example heuristic for coding could be shorter is better, but closer to being able to reconstruct the git diff is also better, so balancing these would yield a good belief. In domains where good heuristics are hard to define, we propose a general autoencoding-inspired grading function, which treats the current language model π θ as both encoder and decoder of information from the history, and the belief states as the codes. We grade each belief b t+1 by how well it can be used by the current model π θ to reconstruct the most recent observation o t : Eq. 1: Reconstruction grading objective Eq. 1: Reconstruction grading objective. Here b t+1 is the updated belief, o t the latest observation, a t the action just taken, b t the prior belief, p I the task prompt, and π θ the current model. Higher grades reward beliefs that retain information needed to decode the latest observation. What do we gain by grading beliefs? Collaborative coding on CollabBench We demonstrate the utility of belief grading in our motivating domain of human-driven assistive coding, with the CollabBench environment from Sweet-RL ( Zhou et al., 2025 ). CollabBench collaborative coding environment Fig. 4: CollabBench collaborative coding environment. The agent asks clarifying questions, then submits a function scored against hidden unit tests. We see that with the general reconstruction-based belief grading function we reduce the performance gap from full context models by about 50%, and train in 50% fewer steps compared to training models to summarize without belief grading (no BG). After training, ABBEL still uses significantly less memory than the full context setting, as measured by the peak context token length (Peak Tokens). Model Test Pass Rate ↑ Success Rate ↑ Peak Tokens × 10² ↓ Training Steps ↓ Full Context 0.52±0.02 0.39±0.02 14.08±0.55 100 ABBEL (no BG) 0.46±0.02 0.31±0.02 4.20±0.37 100 ABBEL-rec-BG 0.48±0.01 0.36±0.01 6.01±0.33 50 Fig. 5: CollabBench results. With reconstruction belief grading, ABBEL-rec-BG recovers about half the gap to full context while using fewer peak tokens, and trains in 50 steps instead of 100. Combination Lock Additionally, in CombinationLock, we demonstrate that ABBEL with a belief grader which leverages domain knowledge (by computing useful statistics over the history and checking that they can be reconstructed from the belief state), enables even higher learning efficiency than full context (FULL CTX) models. Average Attempts to guess target word ABBEL Fig. 6: Average attempts to guess the target word on Combination Lock (lower is better). With domain-knowledge belief grading, ABBEL approaches or exceeds FULL CTX in this setting; without belief grading, learning is slower. Multi-objective question answering In a third environment, multi-objective question answering (from MEM1 Zhang et al., 2025 , a recent work which performed end-to-end optimization in a modified version of typical recursive summarization), we demonstrate the utility of isolating belief states from reasoning, by showing that a Peak Belief length Penalty (more details in paper) significantly reduces memory usage with minimal performance degradation, unlike is commonly observed when penalizing reasoning lengths ( Arora et al., 2025 ). Performance and memory usage in multi QA Fig. 7: Exact-match score and peak memory versus number of objectives in multi-objective QA. ABBEL with a peak belief penalty (PBP) maintains comparable performance while using less memory than MEM1 and ABBEL without PBP in this evaluation. Related work Alternative solutions to managing long contexts involve different tradeoffs, and are worth considering depending on the requirements of a deployed system. Context compression methods generate dense representations which, while computationally efficient, sacrifice human-understandability ( Kontonis et al., 2026 , Eyuboglu et al., 2025 , Gupta et al., 2025 , Chevalier et al., 2023 , Deng et al., 2025 , Deng et al., 2025 , Bulatov et al., 2022 ). Hand-designed summarization prompts ( Wang et al., 2025 , Örwall et al., 2025 , Starace et al., 2025 ) and pruning strategies ( Jiang et al., 2024 ) specific to target environments require expert human knowledge and don’t allow an agent to learn what to remember as part of its decision-making strategy. Methods that process long contexts into an external memory store ( Packer et al., 2023 , Xu et al., 2025 ) for the agents or subagents to query ( Zhang et al., 2025 ) are complementary, as they may benefit from better next context creation through summarization training. We would like to point out some exciting works in the space of general recursive summarization focused on math ( Wu et al., 2026 ), reasoning with belief generation ( Zhou et al., 2025 ), competitive coding with a distilled summarization module using similar autoencoding objectives to our general belief grader ( DeepReinforce et al., 2026 ), and adding continuous features to summaries ( Kontonis et al., 2026 ). What’s next for better memory? Many more possibilities are enabled through using explicit belief states as information bottlenecks for multi-step interaction. You could reward actions based on their effect on the belief state to guide exploration, transmit the explicit belief states for better communication between agents, or even improve user controllability by directly modifying the memories on which the agents’ decisions are based. Some forms of information, e.g., what a person looks like, are not represented well by text alone. A continuously learning system will also have to capture such information. Additionally, if we want a system to learn to communicate in a brand new language or to play a brand new game better than any person in the world, the skills accumulated over the lifetime of conversations or games must be stored in a very compressed form, essentially taking on the role of the weights of the model itself. More powerful systems will likely utilize a combination of multiple forms of memory, where the contents of the context may correspond to working memory while other approaches are used for short and long-term memory. How to instantiate these other forms of memory, for instance via test-time training, adapter memories, continuous context memories, or some combination thereof, presents an exciting challenge. Acknowledgements Acknowledgements: We would like to thank Alane Suhr and Kartik Goyal for advising this research as well as Ethan Mendes , David He , Jitesh Jain , and Nicholas Tomlin for comments on early drafts of this post. We would like to thank the MEM1 authors for their email correspondence and for sharing private reviewer feedback which we found particularly insightful. Citation If abbel was inspiring for your future work, please cite us with this! And here is some advice for doing similar research! @misc { lidayan2026abbellearningnaturallanguagebelief , title = {ABBEL: Learning Natural-Language Belief States for Memory-Efficient Interaction} , author = {Aly Lidayan and Jakob Bjorner and Satvik Golechha and Kartik Goyal and Alane Suhr} , year = {2026} , eprint = {2512.20111} , archivePrefix = {arXiv} , primaryClass = {cs.CL} , url = {https://arxiv.org/abs/2512.20111} , } With newer models the number of tokens till 50% compute spend is on attention gets much larger than 25K. Interleaving linear attention alternatives with full attention as is done with gpt-oss and DeepSeekv4, results in massive flops reductions for the attention computation. For example with DeepSeekv4-Pro (1.6T A49B) it requires nearly 450 thousand tokens to reach the 50% tradeoff point. Grandcode uses Qwen-3.5-397B-A17B a model which hits 50% FLOPs for attention at ~150 thousand tokens. ↩ This setting is technically solvable with much more computationally effective tools, but serves as a flexible test bed to study properties of recursive summarization. Bertsimas et al., 2022 , showed that an exact solution for the wordle game instantiated with the original vocabulary of the javascript game can be found with dynamic programming, but evidently the general formulation of wordle as a guessing game on K letters with L attempts and some dictionary of valid words and correct words D is NP hard to determine the minimal number of moves required. ↩ In practice there is an O(N/K) overhead cost for summary. N is the total number of actions. K is the number of actions till summarization is triggered. This is necessarily true for any summary approach. For ease of illustration this gif uses K = 1. In our experiments, to put more emphasis on summarization weaknesses we also use K=1. In practice overhead is small as K can be chosen to be near the efficient hardware limit. ↩
Kernel knowledge transfer from CUDA to MLX Figure 1: CUDA-to-MLX optimization translation map. CUDA optimization knowledge can be translated into architecture-native MLX strategies rather than copied instruction-for-instruction. We face a new epoch in computing. Hardware is changing rapidly — not just faster GPUs, but a growing range of chips from different vendors, each with its own architecture and often tailored to specific AI workloads. Software is changing just as fast, and AI coding tools now generate in minutes what took months of effort a few years ago. With so much of computing now centered on AI, GPU kernels are a crucial component of its success. These are the low-level programs that run inside the GPU, and writing efficient ones is far from obvious — it takes years of expertise to get right. Transferring a kernel from one vendor’s hardware to another is harder still, and often means rediscovering the same optimizations from scratch. The CUDA ecosystem, for example, has accumulated decades of hard-won kernel expertise: hand-tuned implementations of attention, state space models, and other critical operations representing thousands of engineering hours. Newer hardware ecosystems (Apple Silicon, custom AI accelerators, and others) are growing fast but lack this depth. In this work we ask whether that expertise can be transferred automatically. We built on K-Search , an evolutionary kernel search framework introduced by Cao et al. at Berkeley Sky Lab that uses AI to optimize GPU kernels, and extended it with a backend for MLX — Apple’s machine-learning framework for its own Apple Silicon chips. We developed a novel structured CUDA-to-MLX translation layer that lets K-Search take existing CUDA kernels as a knowledge base and adapt them into high-quality GPU kernels for Apple Silicon, rather than rebuilding from scratch. We show that our approach reaches near-expert level performance on Apple Silicon with 0.97x speedup compared to the native MLX Attention kernel, and up to a 20x prefill speedup over the community mlx-lm implementation on the Mamba SSM kernel; we report the numbers, and how much of the gain comes from the translation layer, in the sections below. Although we focus on MLX kernels for Apple Silicon, the method is not specific to MLX and applies to any ecosystem where CUDA expertise is transferable. Why MLX? Apple’s MLX framework has seen remarkable adoption since late 2023. With Apple Silicon in hundreds of millions of MacBooks and Mac Studios, MLX enables local AI inference without cloud costs. The unified memory architecture makes it especially attractive for mid-sized models (7B–70B parameters on M series chips). Yet beneath this momentum lies a significant gap: many performance-critical kernels that the NVIDIA ecosystem takes for granted: paged attention, optimized SSM scan kernels, fused MoE routing are either absent or naive without hardware-specific tuning. MLX runs models correctly but often leaves significant performance on the table. This gap is what motivates the rest of this post. What is K-Search? K-Search is an evolutionary kernel optimization framework originally developed by our first author Shiyi Cao at UC Berkeley Sky Lab. Given a naive kernel and a hardware specification, it runs an iterative optimization loop: an LLM reasons about which optimizations to try next, a code-writing model generates candidate kernels, and those candidates are compiled and benchmarked on real hardware. Measurements feed back into the search, which keeps refining, pursuing promising directions and dropping dead ends until performance converges. Pseudocode for K-Search via co-evolving world models Algorithm 1: K-Search via co-evolving world models. The search alternates between selecting the most promising action, instantiating and evaluating code until improvement stagnates, and evolving the world model through insert, update, and prune operations. Adapted from Cao et al. (2026) . Search is grounded by a Spec: a domain-specific document encoding hardware rules, optimization patterns, and mathematical constraints which keeps generated code from hallucinating invalid primitives and ensures candidates will actually compile and run efficiently. In our runs, a single model (Gemini 3.5 Pro Preview) plays both roles: it maintains the reasoning state and writes the kernels. The reasoning half is prompted as a “GPU kernel performance engineer” and asked to work through a fixed analysis before proposing anything: classify the kernel (reduction, scan, attention/softmax, …), rewrite the reference computation in canonical form, map out data layout and access patterns, and hypothesize the likely bottleneck (bandwidth, latency, compute, or synchronization) in each runtime regime. Only then does it emit candidate optimizations, each as a single change implementable in one iteration. We call the persistent reasoning state a world model . Rather than a flat list of things to try, it is a decision (prefix) tree: each root→leaf path composes a full optimization plan, and sibling branches are competing alternatives. Every node is scored — an overall_rating in [0, 10], a confidence in [0, 1], and per-node impacts on memory bandwidth, register pressure, and compute/hardware fit — so the search can rank partial plans and expand the most promising ones. The tree persists and grows across rounds: refining an idea adds a child node rather than overwriting its parent, and if the best score fails to improve for a few rounds (a stagnation window) the search backs off to explore an alternative branch. A single node, as it appears mid-run on the attention kernel, looks like this: { "action" : "Replace the threadgroup-memory softmax reduction with a register-only reduction: each SIMD group owns 8 query rows and reduces across lanes with simd_shuffle_xor, removing a threadgroup_barrier." , "difficulty_1_to_5" : 4 , "impacts" : { "memory_bandwidth" : 8 , "register_pressure" : 4 , // risk: spill if Br > 8 "compute_hw_fit" : 9 // SIMD width 32 ; keep tile 8 x 8 }, "overall_rating_0_to_10" : 8 , "confidence_0_to_1" : 0.7 } Listing 1: Example K-Search world-model node. Each candidate optimization records a concrete action, estimated hardware impacts, an overall priority rating, and the model's confidence. Overview of the K-Search loop Figure 2: Overview of K-Search. The framework operates on a Search State $S_t$ structured as a search tree. The tree consists of Closed nodes (blue, visited states with attached program like $x_{12}$) and a Frontier of Open nodes (orange, pending hypotheses like $u_{13}$). The workflow iterates through three phases: (1) Action Selection , where the most promising action node is retrieved from the frontier based on world model estimated priority score $V$; (2) Local Refinement , where a stochastic policy $\pi_{\mathrm{code}}$ samples concrete implementations until stagnation; and (3) World Model Update , where the LLM reasons over the trajectory to update the search tree via Insert (adding new actions), Update (adjusting $V$, e.g., $u_{11}$ dropping from 0.9 to 0.6), and Prune (removing less promising nodes like $u_{10}$). The original K-Search paper evaluated this search strategy on CUDA kernels from FlashInfer. Across GQA decode, MLA decode, MLA prefill, and MoE, K-Search improved more consistently than OpenEvolve and ShinkaEvolve over the same 120-iteration budget. These results establish the search framework we build on here; the remainder of this post asks whether its optimization knowledge can transfer beyond CUDA. K-Search benchmark results compared with OpenEvolve and ShinkaEvolve Figure 3: Main results from the original K-Search paper. Across three runs, K-Search achieves stronger best-so-far search scores, per-workload kernel performance, and speedup distributions than OpenEvolve and ShinkaEvolve on four FlashInfer CUDA kernels. Reproduced exactly from Cao et al. (2026) . Building an MLX backend To bring K-Search to Apple Silicon, we first built a native MLX backend. We implemented a full MLX-specific task adapter for K-Search, including: An MLX task backend in k_search/tasks/ handling kernel compilation and execution on Apple Silicon via MLX’s Metal/C++ APIs. Updated kernel generator prompts for writing and modifying Metal/MLX kernels. MLX-specific benchmarking integration using mlx.core measurement utilities. Translating CUDA expertise to MLX However, the more interesting challenge was not simply running K-Search on MLX. The key insight is that expert CUDA kernels encode decades of optimization knowledge that is transferable to Apple GPU if you can bridge the conceptual gap. Simply handing an LLM a CUDA kernel and asking it to port it is not enough: without deep hardware context, it produces code that is syntactically valid but architecturally wrong (wrong tile sizes, invalid primitives, mismatched memory assumptions). Our translation layer consists of: Concept mapping tables: A structured glossary of CUDA primitives and their MLX/Metal equivalents with hard constraints. For example: __shared__ maps to Metal threadgroup memory but with a hard 32 KB limit (vs. NVIDIA’s 48 KB) warp_reduce maps to MMA (preferred) __syncthreads() becomes threadgroup_barrier(mem_flags::mem_tg) H100’s ~3.35 TB/s HBM3 maps to M3 Max’s ~400 GB/s unified DRAM a bandwidth difference that reshapes which optimizations are worth pursuing. MLX-specific hints and patterns: Concrete code-level patterns for operations with no direct CUDA equivalent, such as register-based row reductions using simd_shuffle_xor in an 8×8 MMA tile layout, or the “exp2 trick” (replacing $exp(x)$ with $exp_2(x \log_2 e)$) for faster softmax on Apple’s fast $exp_2$ hardware instruction. Reusable assertions: Expert kernel behaviors reframed as properties the evolutionary search must preserve, rather than code to copy. Matching expert kernel performance: the Attention kernel We evaluate three configurations of an MLX attention kernel for Apple Silicon: (1) a naive baseline, (2) pure evolution with no additional provided context, and (3) a full context translation layer, which supplies the optimizer with architecture-specific implementation knowledge extracted from high-performance kernels (e.g., FlashAttention-2), letting the evolutionary search reason about implementation strategies rather than starting from a naive kernel. Together, these three configurations let us isolate the exact impact of the translation layer. Performance scaling of the Attention Kernel through stacked optimizations Figure 4: Performance scaling of the Attention Kernel through stacked optimizations. The "Full Context" configuration successfully discovers and implements advanced strategies like double buffering and loop unrolling, achieving near-expert performance. The jump from 0.26× to 0.97× the speed of Apple’s state-of-the-art attention kernel — illustrates how much the translation layer matters. With full context, the evolved kernel independently discovers the key optimizations in FlashAttention 2: threadgroup memory tiling, online softmax, K-transposition for memory access, and the exp2 trick. The last of these replaces every softmax exponential with a base-2 exponential, \[e^x = 2^{x \log_2 e},\] which is exact and lets the kernel use Apple’s fast fast::exp2() hardware instruction directly instead of paying for a base conversion at runtime. A 20× faster prefill: the Mamba SSM kernel To evaluate whether K-Search generalizes beyond attention kernels, we applied it to the state-space model (SSM) kernel used by Mamba. Unlike attention, the computational bottleneck is a recurrent state update rather than a softmax, providing a substantially different optimization challenge. We compare the evolved implementation against the community MLX implementation (mlx-lm) and the PyTorch reference implementation (mamba.py) on an M1 Max. Evaluated on mamba-370m f16, M1 Max 64GB: Metric mlx-mamba (ours) mlx-lm (community) mamba.py Decode 152 tok/s 116 tok/s 40 tok/s Prefill L=512 5,751 tok/s 329 tok/s 1,089 tok/s Prefill L=1024 6,010 tok/s 327 tok/s 1,127 tok/s Prefill L=2048 6,612 tok/s 326 tok/s 1,092 tok/s Prefill L=4096 6,743 tok/s 339 tok/s 1,042 tok/s Table 1: Prefill and decode throughput on mamba-370m (f16, M1 Max 64GB). mlx-mamba (ours) reaches ~20× higher prefill throughput than the community mlx-lm baseline, while decode remains comparable. The ~20× prefill speedup over mlx-lm comes down to one difference: mlx-lm does not implement a parallel scan for the SSM. The state recurrence \[h_t = \bar{a}_t h_{t-1} + \bar{b}_t\] looks inherently sequential, but each step can be written as a pair $(\bar{a}_t, \bar{b}_t)$ under the associative combine \[(a_2, b_2) \circ (a_1, b_1) = \left(a_2 a_1,\ a_2 b_1 + b_2\right),\] which reproduces the recurrence exactly. Because the operator is associative, the whole sequence can be evaluated with a parallel (prefix) scan in $O(\log N)$ dependent steps instead of $O(N)$. mlx-lm skips this and processes tokens one at a time, leaving most of Apple Silicon’s compute idle; our evolved Metal kernel applies the scan and makes much fuller use of GPU throughput. The gain shows up in prefill, where the full sequence is available to scan in parallel, and not in single-token decode, where there is only one new token per step and no scan to parallelize — which is why the decode row is roughly flat while prefill is ~20×. mamba.py is slow on both prefill and decode because it is a PyTorch reference implementation that falls back to CPU or MPS on Apple Silicon, forgoing the hardware-specific optimizations that MLX’s Metal backend makes possible. What’s next? On the two kernels we studied, AI-driven evolutionary kernel search grounded in structured cross-platform translation knowledge reached near-expert performance on Apple Silicon without a team of GPU experts starting from scratch. We do not yet know how far this generalizes, but the result is encouraging. For us the main takeaway is that the bottleneck was not the LLM’s ability to write Metal code, but the quality of the context and constraints we gave it. Our CUDA translation layer converts existing NVIDIA kernel expertise into actionable guidance for Apple Silicon, and lets K-Search’s evolutionary search do the rest. We are actively extending this work in several directions: supporting new architectures, with current efforts focused on developing new kernels for the IBM Spyre AIU and broader hardware targets; adding more kernels such as paged attention and fused MoE routing; and improving integration with the K-Search evolution loop to make translation context even more automatic. Acknowledgements This work was carried out by IBM Research and builds on K-Search from the UC Berkeley Sky Lab ( Cao et al., 2026 ). We welcome collaboration and feedback from the MLX and broader AI systems communities. If you are working on kernel optimization for non-CUDA hardware, we would love to hear from you. Citation @article { cao2026k , title = {K-Search: LLM Kernel Generation via Co-Evolving Intrinsic World Model} , author = {Cao, Shiyi and Mao, Ziming and Gonzalez, Joseph E and Stoica, Ion} , journal = {arXiv preprint arXiv:2602.19128} , year = {2026} } Appendix: Try it yourself The MLX backend is built on top of the open-source K-Search repo, so the results here can be reproduced directly. The steps are: 1. Clone and install git clone https://github.com/caoshiyi/K-Search.git cd K-Search uv pip install openai wandb uv pip install git+https://github.com/caoshiyi/flashinfer-bench-ksearch.git 2. Set your credentials Open the relevant script under scripts/ and set three variables at the top: KSEARCH_ROOT = /path/to/K-Search API_KEY = your-llm-api-key 3. Run kernel search # Optimize Flash Attention on Apple Silicon (world-model mode) bash scripts/mac_flash_attention_wm.sh # Or a Mamba SSM kernel, e.g. the selective scan bash scripts/mamba_selective_scan_fwd_wm.sh Full CLI reference and documentation are in the README.
Orchard 是一个面向研究社区的开源框架，用于训练和评估适用于各种任务类型的 AI 智能体。它通过让研究人员复用相同的基础设施来降低复杂性，同时支持较小模型实现出色性能。该文章《Orchard：用于可扩展智能体 AI 的开放框架》最初发表于 Microsoft Research。
A new creature-catching adventure is ready to stream from the cloud this week. Pawprint Studio’s Aniimo arrives on GeForce NOW at launch, inviting gamers to explore the vibrant continent of Idyll across supported devices. Also this week, 007 First Light receives a path-tracing update on GeForce NOW, alongside a smashing limited-time Deluxe Edition sale on […]
为了构建和部署能够在动态环境中感知、推理并采取行动的复杂机器人应用，开发者需要全新的物理 AI 模型和工具。ROS 开放框架是 Open Robotics 推出的一个项目，旨在帮助人类构建机器人。NVIDIA Isaac ROS 5.0——一套基于 ROS 构建的 GPU 加速软件包集合，于今日在 […]
文章 URL： https://allenai.org/blog/astabrief 评论 URL： https://news.ycombinator.com/item?id=49938783 得分：3 评论数：0
Hi HN, I'm the creator of Pyxel, a free, MIT-licensed retro game engine for Python. I've been developing it since 2018, and it recently passed 18,000 stars on GitHub. Pyxel includes pixel art, tilemap, sound, and music editors, so you can create a game's visuals and audio as well as write its code in Python. The engine itself is implemented in Rust. Games run on Windows, macOS, Linux, and the web, as well as supported Linux-based retro handhelds. You can share a URL to launch a Pyxel game hosted on GitHub directly in a browser. You can also create games entirely in your browser with Pyxel Code Maker, or write chiptunes in Pyxel MML Studio. See what people have made with Pyxel: https://kitao.github.io/pyxel-user-examples/ Try Code Maker: https://kitao.github.io/pyxel/web/code-maker/ I'm happy to answer questions about the engine and its implementation. Comments URL: https://news.ycombinator.com/item?id=49928215 Points: 10 # Comments: 3
文章网址： https://insideclimatenews.org/news/01102026/abc-coke-operating-under-expired-permit-in-alabama/ 评论网址： https://news.ycombinator.com/item?id=49934537 积分：10 评论数：1
陶瓷材料研究是考古学研究的基石之一，但陶器记录的后期制作流程仍然十分耗费人力，并造成严重的出版瓶颈。本文介绍了 PyPottery，这是一套由 AI 驱动的开源工具，旨在实现完整陶瓷记录流程的半自动化。该套件由四个集成模块组成：用于自动提取图像和识别手写文字的 PyPotteryScan；用于自动为铅笔绘图勾勒墨线的 PyPotteryInk；用于语义感知矢量化的 PyPotteryTrace；以及用于自动生成版式的 PyPotteryLayout。在对来自 Terramara di Montale（意大利）的 50 张手绘图纸（包含 240 幅陶器绘图）进行评估后，该框架显著节省了时间。可用性研究参与者证实了这一点，他们报告称，相较于传统流程，感知到的速度提升中位数为 40$\times$（范围：17.5$\times$--120$\times$）。这些结果凸显了 AI 辅助工具在考古记录中的潜力；同时，本文还探讨了将认知劳动战略性地重新分配至增强而非自动化的问题。
随着机器人硬件和学习方法的发展，类人机器人需要借助工具来执行超出其固有物理能力限制的任务。成功使用工具需要选择合适的工具，并协调操作以及在必要时进行移动，以完成任务。现有基准测试无法在类人机器人上对这些能力进行联合评估。我们推出了 HumanoidToolBench，这是一个包含18项任务的基准测试，涵盖三种场景、三个执行层级和两种工具集模式；同时推出 ToolBook，这是一个包含3.1k条演示数据的数据集，采集自仿真环境和真实 Unitree G1。对仿真环境中的七种策略和真实机器人上的三种策略进行评估后发现，在选择合适工具与完成任务之间存在显著差距。针对 GR00T N1.7 的专项测试表明，其在面对未见过的工具时选择准确率下降，并且会在无关指令下继续执行任务。代码和数据集可在 https://snu-pi.github.io/HumanoidToolBench/ 获取。
在策略自蒸馏近期已成为提升语言模型推理能力的有效方法，其通过让接收特权信息的冻结版本或 EMA 版本的自身模型来监督学生模型。在多模态大语言模型（MLLMs）中的应用却仍处于探索阶段。近期方法利用特权视觉信息（例如与问题对应的图像裁剪区域）来提升细粒度感知能力，但其效果仅限于受益于此类视觉放大的任务，并且需要人工标注的定位数据或外部教师模型。我们为 MLLMs 引入了一种不同形式的在策略自蒸馏：向教师模型提供文本形式的、具有空间定位依据的指导，以识别与查询相关的视觉元素。我们使用程序化生成的场景，其中自动提供物体身份和空间坐标，从而实现可扩展且无需标注的后训练。教师模型利用这种空间指导，在多个相关图像区域中定位并整合证据；学生模型则仅根据图像和问题学习复现这一行为。我们的方法在多个模型上的计数、文档理解和图表理解基准测试中均能稳定提升性能。值得注意的是，尽管后训练仅使用合成场景，所得改进仍能迁移到真实世界感知基准，在 CVBench、V*、ZoomBench、BLINK、HR-Bench 和 MME-RealWorld 上的平均性能提升了 3.23 个百分点。这些结果表明，具有空间定位依据的特权信息能够通过在策略自蒸馏诱导出更广泛的感知能力，使模型在后训练所使用的任务和数据分布之外实现显著的从合成数据到真实世界的迁移。项目主页：https://github.com/sirkosophia/Where-OPD
分布匹配蒸馏（DMD）通过分别估计的目标得分函数与学生得分函数之间的差异来训练少步数学生模型，因此必须额外维护一个拟合学生模型不断变化分布的辅助扩散模型，从而增加内存和计算开销。我们提出 DMAD（Distribution Matching as Adversarial Distillation），将分布匹配重新表述为分类问题，并直接学习所需的对数密度比。我们在共享骨干网络上设置两个判别器头，分别区分真实数据和教师样本与学生样本，并通过对其 logits 施加线性损失，在无需额外拟合辅助得分函数的情况下训练学生模型。我们证明，在判别器达到最优时，根据将判别器 logits 与对数密度比联系起来的经典恒等式，这些损失能够恢复 DMD 所依据的分布匹配梯度。我们进一步提出基于间隔的重加权方法，根据真实数据判别器头在真实样本与教师样本之间产生的经验 logit 间隔，自适应地调整不同噪声水平下的教师监督信号。在 ImageNet-64x64 上进行一步生成时，DMAD 的 Fréchet Inception Distance（FID）达到 1.04；在 COCO-10K 上使用四步 SDXL 时达到 14.47；在使用四步 Wan2.1-T2V-14B 时，VBench 总分达到 85.15；这些都是所比较的少步数方法和多步数教师模型中的最佳结果。在 MiniMax-H3-33B 上，我们的四步学生模型在联合音视频生成任务中，相较于 DMD2 和 rCM 的总体人类偏好率分别达到 79.1% 和 84.6%，不计平局。我们的代码、模型和演示可见于 https://yzmblog.github.io/projects/DMAD。
Discrete diffusion language models offer a compelling alternative to autoregressive generation for tasks demanding bidirectional reasoning and global constraint satisfaction. Yet they share a structural bottleneck: when decoding in parallel, each token is sampled independently from its marginal, severing the statistical dependencies among the tokens decoded together. Continuous diffusion language models avoid this by denoising a shared continuous state, but their denoiser sees only that state, so nothing ties it to a valid token configuration until it is finally decoded. To address this, we propose Hierarchical Continuous Diffusion Language Models (HC-DLM), which couple discrete token generation with a continuous latent trajectory in a single, principled denoising process, whose training objective is derived from a variational bound on the token likelihood. In contrast to recent methods that attach continuous context to a self-contained discrete chain, HC-DLM makes the latent the only persistent generative state: tokens are read out from it at every step and feed back as a scaffold for the next latent update. On structured reasoning (Sudoku), mathematical planning (Countdown) and language modeling (LM1B), HC-DLM improves over discrete and continuous diffusion baselines at matched model size, in puzzle accuracy on Sudoku and Countdown and in generative perplexity on LM1B. Project page: https://hc-dlm.github.io/.
在多样化任务中构建可靠的机器人能力，需要投入大量人力来开发和维护技能、设计奖励，并将感知与控制相结合。我们提出 Reconstruct, Practice, Go Real（RPG），这是一个无需更新模型权重即可实现机器人执行系统自主改进的框架。RPG 在离线数据集中识别操作能力，并在仿真环境中构建相关练习任务。在练习过程中，RPG 利用执行反馈、特权模拟器状态以及数据集中现有的视频来诊断失败原因。它会根据这些诊断结果开发新的可复用符号技能、改进现有技能，并修订系统提示词。跨任务评估会在保留候选修改及合并后的修订版本以供复用之前，对它们分别进行测试。在测试时，多模态 LLM 利用生成的系统提示词和技能库来协调感知与机器人控制。在22个操作任务的留出初始化条件下，RPG 将任务成功率从第一轮练习后的28.6%提升至15轮后的95.0%，超过了所有参与评估的基线方法，包括 ASPIRE（75.5%）和由 GPT-6 Astra Pro 驱动的 CaP-Agent0（60.0%）。经过统一的校准与硬件适配流程后，该冻结系统在全部30次实体实验中均取得成功，三个任务各进行10次实验。项目网站：https://rpg-robot.github.io/
LLM 越来越多地应用于网络安全工作流中，其任务是将分析人员的意图转换为工具调用。然而，现有评测主要侧重于基于知识的评估或端到端的智能体任务，并未直接衡量 LLM 生成真实网络安全工具可执行命令的能力。这一缺口至关重要，因为网络安全操作依赖严格的命令行接口（CLI），其中细微的语法错误、不正确的选项与值绑定或参数顺序错误，都可能导致命令无法执行。我们提出 KaliBench，这是一个面向 Kali Linux 自然语言到 CLI 转换的细粒度基准和数据集，包含 8,504 个查询—命令对，覆盖 23 个能力维度、5 个安全阶段中的 1,642 个工具。KaliBench 通过以手稿为依据的流程构建，并采用确定性规范化和感知别名的评估方法，从而能够精确且可复现地评估工具选择和参数构造。为同时确保语义正确性和实际可执行性，我们开发了一个多阶段验证流程，结合基于 LLM 的验证、沙箱终端执行以及人在环路中的迭代完善。基于这些细粒度且确定性的信号，KaliBench 还支持用于训练的无需运行时的可验证奖励。在三种评估模式和 24 种通用型及网络安全专用开源权重模型配置下，在不提供限制的设置中，没有任何开源权重模型的精确命令准确率超过 42%，这凸显了在缺乏明确工具提示时，准确使用基于 CLI 的网络安全工具的困难。我们进一步表明，使用源自 KaliBench 的可验证奖励进行监督微调和强化学习，能够显著提升一个 8B 模型的性能，并达到与 685B MoE 模型相当的表现。
3D 高斯虚拟角色支持快速渲染，但其实时动画通常受到高成本神经网络推理的限制。我们针对这一瓶颈展开研究，发现预训练虚拟角色模型的动画可以通过与身份无关的混合形状线性组合进行高精度近似。基于这一发现，我们提出 GALA（Gaussian Animation via Linear Approximation），这是一种蒸馏方法，以浅层系数预测器和线性混合替代逐帧执行的复杂神经解码。为提升保真度并降低内存需求，我们在考虑渲染感知度量和内存预算的条件下，采用局部块 PCA 构建基底。我们训练了一个浅层 MLP 网络来预测混合形状系数，并且无需重新训练原始模型，即可应用于各种动画架构。我们通过加速三个不同虚拟角色模型的推理来验证 GALA，这些模型用于带有表情和服装动态的 3D 面部动画及全身动画。在这些模型上，我们的蒸馏方法能够泛化到未参与训练的身份，在保留大部分渲染质量的同时，将 CPU 动画成本降低最高三个数量级。我们方法取得的出色结果证实了学习到的虚拟角色表示中存在共享的线性结构，并支持在移动设备上以最高达 60fps 的帧率实现高效且精确的动画。项目主页：https://ramazan793.github.io/gala/
高效的长上下文推理对于大语言模型（LLM）至关重要，但其也带来了严重的计算瓶颈。基于哈希的检索通过将查询和键编码为二进制码，并使用汉明距离选择键，为此提供了一种高效的替代方案。然而，这会导致汉明距离与注意力相关性之间存在严重不匹配。查询-键logit取决于方向相似度和特征幅值，而哈希二值化会丢弃幅值信息，从而造成两类问题：错误检索低logit的键，以及遗漏高logit的键。为解决这些问题，我们提出分层哈希检索（Hierarchical Hash Retrieval，HHR），这是一种通过几何感知键路由（Geometry-Aware Key Routing，GKR）和学习型哈希投影（Learned Hash Projection，LHP）逐步提升检索准确率的由粗到细框架。GKR学习一种按注意力头分别进行的正交变换，以重新分配特征幅值，并推导出更具判别性的页面级logit界限，从而能够有效剪枝低logit的键，同时保留重要候选键。随后，LHP学习一个按注意力头分别构建的投影空间，使汉明距离与真实的查询-键相关性排序相一致，以实现细粒度检索。通过结合GKR和LHP，HHR能够抑制误检并找回漏检，显著提升基于哈希的稀疏注意力的保真度。在多种LLM和基准测试上的大量实验表明，HHR的性能优于现有方法。例如，在LongBench上，HHR将平均得分提升了1.10分；在128K上下文长度下，使用Llama-3.1-8B-Instruct时，其解码速度最高提升3.30倍，端到端速度最高提升2.83倍。代码已公开于https://github.com/lianjunl13-sudo/HHR。
Generative artificial intelligence has reduced the cost of producing plausible symbolic artefacts, leading recent organisation scholarship to identify evaluation and discernment as constraints under conditions of production abundance. This perspective examines a further possibility: that machine evaluation itself becomes inexpensive enough to be deployed routinely and at scale. The investigation is prompted by Jev, TypeSafe AI's specialised model for typed probabilistic decisions. TypeSafe explicitly invokes William Stanley Jevons to argue that lower-cost machine intelligence can unlock previously uneconomic uses. Treating this as a technological provocation rather than an established empirical result, the article formulates a conditional Jevons hypothesis for machine evaluation: sufficiently large reductions in the total marginal cost of usable machine evaluation may increase its organisational consumption where latent demand is substantial and complementary costs do not dominate. The article integrates rebound economics with research on cheap prediction, production abundance, machine evaluation, decision allocation, authority, reliance and Executive Judgement to examine this possible scarcity transition. It distinguishes prediction, machine evaluation, organisational judgement and authorisation as functional activities whose costs need not fall together. Evaluations can share evidence, criteria and errors; scale mis-specified rubrics; operate on representations from which consequential qualifications have disappeared; and change practical decision rights through thresholds and exception routing. The resulting research problem is when cheap machine evaluation substitutes for human evaluative work, when it redistributes or creates demands for judgement, and how it affects the grounds available at consequential organisational commitment.
科学进步源于一个纵向演化的生态系统，在这一系统中，研究人员、机构、资助机构、合作网络和科学文献共同演化。随着 AI 日益深入科学研究的整个周期，理解这些相互关联且不断演变的过程变得愈发重要。我们推出 SciUtopia，这是一个持久运行的闭环 LLM 智能体模拟框架，用于研究学术研究生态系统。SciUtopia 对研究方向选择、合作、投稿、同行评审、重新投稿、引用、资助和研究人员流失等相互关联的科学过程进行建模，同时维持模拟年份之间不断演化的状态。其可配置的机构机制和信息渠道为匹配的反事实实验及有针对性的干预提供了受控测试平台。在 61 个模拟世界中，SciUtopia 模拟了来自 8,000 个机构的超过 40,000 名研究人员，产生了约 400,000 个论文发表决策和 120 万份由 LLM 生成的同行评审意见。通过这些纵向模拟，我们发现，由拒稿驱动的重新投稿会使评审人员负担显著增加，增幅超出仅由人口增长所能解释的范围；谨慎探索能够在引用影响力、职业成功和长期主题多样性之间实现平衡；即使未检测到因早期资助中险胜而产生的累积优势，资源不平等仍可能出现。代码可在 https://github.com/Ahren09/ScienceUtopia 获取。
四足机器人运动需要在指令跟踪、稳定性和能效等相互冲突的目标之间取得平衡，但传统强化学习（RL）会在训练时将这些优先级硬编码到固定的标量奖励中。我们提出 PROMO（Preference-Conditioned Multi-Objective Reinforcement Learning），一种语义多目标方法，将这种权衡作为单一运动策略在运行时的显式输入。PROMO 根据部署阶段给定的偏好对策略进行条件化，同时保持与具体本体相关的运动先验不变，从而将操作者意图与实现可行步态生成所需的奖励塑形项分离开来。与固定目标控制器、多目标基线方法以及分别训练的专用策略相比，PROMO 仅凭一个可部署策略就实现了目标专门化与鲁棒性。在仿真中采样的 100 个偏好下，有 67 种行为在精确 Pareto 支配关系下不受支配，偏好与目标之间的平均相关系数达到 0.843，展现出广泛的 Pareto 覆盖范围和可预测的偏好响应。同一策略无需额外训练即可迁移至 Unitree Go2，仅改变偏好，相较于均衡偏好，便可使比能耗最高降低 30.4%、位置误差降低 38.7%，并使机身姿态偏差峰值降低 59.0%。这些结果表明，面向偏好条件的多目标强化学习可作为自适应腿式运动的一种实用运行时接口，其作用已超越离线 Pareto 集构建。开源代码和视频见 https://amrmousa.com/promo/。
专家混合（MoE）扩散语言模型（DLMs）支持灵活的并行解码并提升模型容量，但其大量专家参数带来了显著的计算和存储成本。现有的低秩 MoE 压缩方法大多依赖静态分解和固定秩分配，忽视了 MoE DLMs 的独特属性。具体而言，我们发现了两种属性：跨模态非均匀冗余，即参数冗余程度以及对秩截断的敏感性在输入、输出和专家模态之间存在差异；以及令牌级利用率变化，即冷热令牌表现出不同的谱特征和专家激活模式。为应对这些挑战，我们提出了 ITC-MoE，一种面向 MoE DLMs 的重要性引导令牌感知压缩框架。ITC-MoE 由两个互补组件构成。首先，重要性引导的自适应 Tucker 压缩（Importance-guided Adaptive Tucker Compression，IATC）将激活和梯度重要性融入专家权重变换，在多个模态上联合分解专家权重，并在固定参数预算下自适应分配秩。其次，令牌感知补偿与路由（Token-aware Compensation and Routing，TCR）对压缩敏感的热令牌应用轻量级低秩补偿，并针对路由模式集中的冷令牌限制候选专家集合。通过联合适配压缩容量和推理执行，以同时应对参数冗余和令牌级变化，ITC-MoE 在保持生成质量的同时，大幅降低了 MoE DLMs 的计算和存储成本。例如，在 SDAR-30B-A3B-Chat-b32 上，ITC-MoE 在 30% 的压缩预算下仍能在 MultiArith 上保持 96.33% 的准确率，同时实现最高 7.22 倍的端到端加速。代码公开发布于 https://github.com/lianjunl13-sudo/ITC-MoE。
Federated LLM fine-tuning enables large models to be adapted using private and geographically distributed data at the network edge, creating recurring and deadline-sensitive communication workloads across access and transport networks. This challenge is particularly relevant in mobile RANs, where wireless variability, mobility, and device heterogeneity cause model updates to arrive asynchronously. Although these updates belong to the same learning round and share a common destination and deadline, conventional transport networks treat them as independent device-originated flows, hiding their underlying structure and limiting the ability to efficiently provision transport resources. This mismatch is particularly problematic for optical circuit switching and all-photonics transport, which benefit from predictable and schedulable traffic demands. We argue that future RANs should act as learning-aware traffic shapers by exposing the communication structure of distributed model adaptation to the transport layer. Through in-network aggregation at the gNB, asynchronous UE updates can be transformed into fewer aggregate transfers with bounded size and delivery requirements. Once shaped in this way, federated LLM traffic becomes a suitable candidate for selectively provisioned optical connectivity, where high-capacity paths can be established during aggregate-transfer windows and released between learning rounds. The resulting architecture combines the flexibility of packet-based mobile access with dynamically provisioned optical capacity, illustrating a broader approach for coordinating distributed AI workloads across programmable access and transport networks.
Reinforcement learning for trading often struggles to balance upside participation with drawdown control. Profit-only policies can collapse toward passive long exposure on upward-drifting assets, while aggressively risk-penalized rewards can become too defensive during volatile periods. This paper proposes PPO-HRAP, a hybrid regime-aware policy that combines Proximal Policy Optimization with an interpretable regime prior. The agent observes both market features and portfolio-state variables, receives a reward combining portfolio log return, VIX-conditioned drawdown-increase penalty, target-exposure deviation, and turnover cost, and executes a blended action between the PPO actor output and a regime-derived target exposure. On the held-out 2020-2022 SPY test window, PPO-HRAP achieves 27.62% total return, 8.48% annualized return, 0.6447 Sharpe ratio, 0.8588 Sortino ratio, and 0.4592 Calmar ratio, while reducing maximum drawdown from 34.10% for Buy and Hold to 18.47%. Across five SPY seeds, PPO-HRAP remains stable with mean total return $0.2725 \pm 0.0109$ and mean Sharpe ratio $0.6219 \pm 0.0565$. Single-run cross-asset tests on QQQ and DIA further show that the proposed method ranks first on total return and Sharpe ratio for all three reported assets. These results suggest that blending learned actions with a volatility-aware regime prior is a practical way to improve risk-adjusted trading behavior, although the current policy still incurs high turnover and cross-asset robustness beyond SPY remains limited to single-run evidence.
I found that I'm not using even 1/3 of vim/vscode features anymore. That's wht I'm building rhun - a small code editor for Linux, Windows and Apple silicon Macs. It obviously has Vim mode, a terminal, Git diffs and a panel for Claude Code or Codex sessions. The editor and pixel renderer share an x86-64 assembly core. For Apple silicon, a build-time translator converts that core to AArch64, with separate platform adapters around it. The latest release can draft commit messages using a local Ollama model or an existing Claude Code or Codex subscription. It's a solo project, MIT licensed and still early. Comments URL: https://news.ycombinator.com/item?id=49926726 Points: 7 # Comments: 2
文章 URL： https://github.com/Vibra-Ingenn/Janus 评论 URL： https://news.ycombinator.com/item?id=49926773 得分：5 评论数：0
Please state the location and include REMOTE for remote work, REMOTE (US) or similar if the country is restricted, and ONSITE when remote work is not an option. Please only post if you personally are part of the hiring company—no recruiting firms or job boards. One post per company. If it isn't a household name, explain what your company does. Please only post if you are actively filling a position and are committed to replying to applicants. Commenters: please don't reply to job posts to complain about something. It's off topic here. Readers: please only email if you are personally interested in the job. Searchers: try https://hnwork.app , https://nthesis.ai/public/hn-who-is-hiring , https://dheerajck.github.io/hnwhoishiring/ , http://nchelluri.github.io/hnjobs/ , https://hnjobs.emilburzo.com . Don't miss this other fine thread: Who wants to be hired? https://news.ycombinator.com/item?id=49922568 Comments URL: https://news.ycombinator.com/item?id=49922569 Points: 47 # Comments: 50
文章网址： https://blog.cloudflare.com/clef-decision-models/ 评论网址： https://news.ycombinator.com/item?id=49923692 得分：4 评论数：0
Yantra is a C++ parser generator: lexer, parser, and AST walker all generated from one tool. It builds the whole AST first, then walks it. Most LALR parser generators (Yacc, Bison, Lemon) run your semantic actions during parsing, as each rule reduces, bottom-up. That means at the time a rule's action runs, you don't yet know what its parent looks like. This pushes a lot of grammars toward hand-built AST classes and a separate walking pass whenever you need to look ahead into siblings or defer a decision until more context is available. On the other hand, Yantra always builds the whole AST first, then walks it top-down in a separate pass, calling your semantic actions as it goes. A parent rule's action can run before its children are visited. A single grammar can define more than one walker. For example, one that emits C++, another that emits Java, from the same parse. The AST and the walker classes are both generated for you. A small example (full version, with compile commands, in the README): start := expr; expr := expr(a) PLUS expr(b) %{ std::cout << "Adding" << std::endl; %} expr := NUMBER(N) %{ std::cout << "Number: " << N.text << std::endl; %} NUMBER := "\d+"; PLUS := "\+"; WS := "\s+"!; Running this on "1 + 2 + 3" prints: Adding Number: 1 Adding Number: 2 Number: 3 The outer "Adding", the root of the tree, prints first, before either of its children. That's only possible because the whole tree exists before any action runs. Some other things about it: integrated lexer with mode support (for things like nested comments), an optional amalgamated single-file output mode with a generated main(), C++23, MIT licensed. It's young (0.5.1, pre-1.0) and single-maintainer, so treat it as early. I'd rather know what breaks than have it look more finished than it is. Known gaps are listed at https://github.com/TantrixAuto/yantra/blob/main/docs/known_l... Repo: https://github.com/TantrixAuto/yantra Feedback and questions are all welcome. I'll be around. Comments URL: https://news.ycombinator.com/item?id=49916997 Points: 4 # Comments: 0
我们提出了游戏引导的技能发现（GGSD），这是一种利用游戏中的自博弈来发现可由人类直接操控的运动技能的框架。可玩技能为具身智能体的控制提供了一种紧凑抽象，使其能够通过一小组学习到的行为进行控制，而非依赖底层动作。为了发挥作用，这些技能应当在语义上彼此区分、易于解释且具有较强的表达能力；而现有的无监督技能发现方法通常难以同时满足这些要求。GGSD通过将技能发现建立在竞争性游戏过程中，实现了这些目标。一个分层智能体与其过去的自身版本进行竞争：高层策略从一个小规模的离散技能集合中进行选择，技能条件化的低层策略则学习相应的行为。训练完成后，人类可以替代高层策略，通过同一组离散技能直接控制智能体。尽管高层动作数量很少，技能之间的转换仍会产生涌现的连招行为，使表达能力超越单个基本动作。在 Ant、Franka-arm 和 Unitree G1 环境中，我们展示了 GGSD 能够生成可由人类操控的技能，人类可以将这些技能组合起来解决未见过的任务，例如 Maze 和 CubePush，而无需额外训练。交互式演示可访问 https://ggsd-demo.github.io。
World Action Models（WAMs）将视觉动力学预测与动作生成相结合，但并未明确支持在不同操作任务之间复用动作经验。此外，现有 WAMs 难以捕捉能够指导目标动作预测的跨任务底层语义关系，因为冗余的背景元素会干扰关键信息的视觉提取。为应对这些挑战，我们提出了一种新颖的 Action Experience Dictionary（AED），将历史物理动作轨迹编码为共享动作嵌入，以支持技能复用并建模跨任务关系。具体而言，我们首先聚合历史动作，使其与视觉观测对齐，并使用预训练动作分词器从 AED 中检索动作嵌入。随后，我们通过交叉注意力对池化后的嵌入进行视觉条件化，并将其置于噪声动作 token 的前面，为预测提供交互上下文和动作意图。为建模与动作相关的运动并减少对无关背景线索的依赖，我们引入了一种运动感知转移损失，在随机时间间隔上监督视觉特征变化预测。在仿真基准以及真实世界跨具身设置上的实验验证了 AED 的有效性。匿名项目网站见 \href{https://github.com/JiahuaDong/AED}{AED}。
近期的自主机器学习工程（MLE）智能体在公开排行榜上取得了显著进展。现代 MLE 智能体通常受到长期运行周期中进展停滞以及大型语言模型（LLM）原语有限等问题的推动，被部署在日益复杂的系统之上：多智能体编排器、专用检索子智能体，以及其他组件。尽管这类框架不断扩展，但领域内很少关注更为基础却有所改进的编码智能体——在这类智能体中，LLM 可通过 read、write 和 bash 原语直接访问执行环境。在本文中，我们发现，在相同的时间预算和相同的前沿 LLM 主干模型下，开源的最先进框架相比于采用最小框架的编码智能体基线单次会话并无优势，这表明主干模型是性能的主要驱动因素。通过一系列大规模、系统性的消融研究，我们认为，在编码智能体设置下，各层系统组件变得多余。我们得出结论：围绕强大模型构建复杂的手工设计框架，对于当前的 MLE 基准测试而言，投入产出比很低。
预训练机器人策略提供了有用的动作先验，但长时域操作仍需要语义推理与物理执行之间的协调。语义推理的运行时间尺度比物理交互更粗，而回合级失败只能有限地指示应当修改系统中的哪个组件。我们提出 DynaHarness，这是一种动态物理控制框架，通过共享执行契约将语义推理与物理治理相耦合，并将失败证据转化为经过验证的能力修订。具体而言，慢脑提出能力和符号参数，而快脑则对命令进行落地与监控，拒绝无法解析的动作，替换能力，并在需要时请求重新规划。物理执行契约约束每条被接受的命令，并记录分析技能、恢复技能和冻结 VLA 的执行证据。失败归因在这些记录中定位故障，并指导对可复用能力或执行机制进行有针对性的修订。成对回归检查决定是否接纳或拒绝修订，从而闭合自演化循环。在 LIBERO-Pro 上，DynaHarness 在新采样的 800 个初始状态上取得了 75.2% 的成功率，而冻结策略为 17.5%。在使用相同能力库的情况下，完整动态执行达到 74.0%，而名义上的单步重新规划仅为 63.9%。这证明了 DynaHarness 作为动态物理控制框架的价值：它能够在执行过程中管理现有能力的落地、监控与协调。我们的项目页面位于 https://denghaoyuan123.github.io/Dynaharness_page/。
近年来，视频生成技术日益逼真且更具可控性，但视频编辑仍相对不够成熟，尤其是在进行精确的局部编辑时，还必须保留原始场景的动态特征。视频场景文本编辑旨在替换场景表面上的文本，例如店面招牌、白板和产品标签上的文字，同时保留周围内容、运动以及相机动态。尽管图像场景文本编辑已得到充分研究，但要同时实现高视觉质量、时间一致性和编辑局部性的视频场景文本编辑，仍缺乏深入探索。现有资源提供的真实视频配对数据有限，而通用视频编辑指标无法直接衡量所请求的文本是否随时间保持正确。我们提出了 ViTeX-Bench，这是一个由 ViTeX-Dataset 和三轴评测协议组成的基准套件。该数据集包含 387 个真实世界的 720p 视频，并配有文本区域掩码和编辑指令：其中 230 个视频提供了经过审核、由处理流程生成的配对编辑结果，用于训练；另外 157 个视频构成固定评测集。该协议通过 13 项指标评估文本正确性、视觉与时间质量以及编辑局部性，每个维度设有一项主要指标，并通过 Pareto 比较分析它们之间的权衡关系。OCR 校准、人工评估和标注敏感性分析为这些分数的解读提供了支持。在来自四类编辑方法的八个基线模型中，要同时实现文本准确性、时间稳定性和场景保真仍然十分困难。我们还发布了 ViTeX-Edit-14B，这是一个开源参考编辑器，基于配对训练集进行微调，并采用运动对齐的字形-视频条件输入。在评测的视频原生编辑器中，它取得了 0.688 的 CharAcc，为最高平均值；在原始编辑器输出中，其可比文本裁剪 Warp 指标最低。ViTeX-Bench 为研究视频场景文本编辑中的这些权衡提供了可复现的基础。
带可验证奖励的强化学习（RLVR）提升了大语言模型（LLM）的推理能力，但其预测仍容易受到与任务无关的提示特征影响。我们通过保留底层问题及其答案不变的半事实提示干预，研究了这种敏感性。分析显示，模型在词元级别的敏感性存在显著差异；在解码过程中抑制高漂移词元候选项，可以在不更新模型权重的情况下提升推理准确率。这些发现揭示了群组相对策略优化（Group Relative Policy Optimization，GRPO）的一项局限：GRPO 为每个响应词元分配相同的、源自结果的优势值，因而可能在强化有用推理的同时，也强化潜在的虚假依赖。受此启发，我们提出半事实信用增强策略优化（Semifactual Credit-Augmented Policy Optimization，SCAPO），这是一种受因果思想启发的 GRPO 变体，将半事实稳定性纳入词元级信用分配。SCAPO 在半事实干预下测量固定响应的词元概率漂移，并利用归一化的稳定性得分，在训练早期降低相对不稳定词元的优势值，同时不会仅因稳定性而给予额外信用。在 Qwen3-4B-Base 和 Qwen3-1.7B-Base 上，与 GRPO 相比，SCAPO 在 AIME 2024-2026 上的准确率分别提高了 5.63 个百分点和 4.17 个百分点。在这两种模型规模下，与所比较的方法相比，SCAPO 在大多数评测数学基准和全部评测分布外基准上均取得了最佳结果。这些结果表明，半事实稳定性能够通过在 RLVR 中进行更细粒度的信用分配，为提升推理能力和泛化能力提供有效的训练信号。代码见 https://github.com/DtYXs/SCAPO。
LLM 智能体日益依赖可安装技能。这些技能由指令、代码和资源组成，为智能体提供特定任务能力，并且安装后可在后续用户任务中被自动调用。这形成了一条信任链：用户将权限委托给智能体，而智能体框架在缺乏充分验证的情况下，将技能提供的内容纳入智能体上下文，从而使恶意技能能够利用这些被委托的权限影响智能体行为。然而，对于这一信任模型能否在不受信任的技能内容到达安全敏感操作之前对其进行充分约束，以及此类信任违规在现实世界的智能体中出现的频率，人们仍知之甚少。我们提出了 TrustProbe，一种用于揭示基于技能的 LLM 智能体中不安全信任链的框架。首先，TrustProbe 分析智能体源代码，识别从技能控制的输入到安全敏感操作的源到汇调用路径。其次，它生成语义上逼真的 SKILL.md 种子，在其中注入金丝雀标记，并通过反馈引导的调度与变异对其进行演化。最后，它使用一个预言机验证漏洞，该预言机确认由攻击者控制的数据流，并核实可观测的危害。在 11 个开源智能体中（其中 8 个在 GitHub 上拥有超过 10,000 个 stars），TrustProbe 识别出 104 个污点型漏洞。在从 ClawHub 等公共中心收集的大规模真实世界技能语料库上进行验证后，结果进一步表明，25.1% 的技能-智能体试验会触发所识别的易受攻击路径，其中通过载荷注入成功武器化了 15 个漏洞。这些结果揭示了基于技能的 LLM 智能体中一种系统性的信任失效：不受信任的技能内容能够到达安全敏感操作，并行使用户委托给其智能体的权限。
受内存限制的边缘设备上的联邦学习（FL）面临两难困境：一阶（FO）优化（即反向传播）需要大量内存，而零阶（ZO）优化则会导致严重的收敛速度下降。为解决这一困境，我们提出了 HO-FL，这是一种混合阶 FL 框架，使用 ZO 优化训练模型的底部部分，使用 FO 优化训练模型的顶部部分。每台设备都可以根据自身的内存预算灵活选择阶数边界，同时参与同一个全局模型的训练。此外，我们的收敛性分析揭示了一种新的基本权衡：采用更大 FO 训练部分的客户端能够提供更准确的更新，但偏重这些客户端可能会导致其他客户端的数据代表性不足。我们将这一权衡与实际多步本地更新的偏差和方差联系起来，进而得到一个采样优化问题，以及一种结合直接模型平均的实用且感知维度的近似方法。在具有数据异质性的语言任务上，我们从任务性能、客户端内存和采样等方面进行了实验。结果表明，混合阶本地训练能够以显著更低的客户端内存需求，保留全 FO 训练的大部分性能。我们的代码可在 https://github.com/HKU-WILL-Lab/HO-FL 获取。
Long-form live-streaming TTS requires context-dependent prosody and paragraph-level coherence. However, many existing instruction-based TTS systems use global or uniform conditions, providing limited explicit control over clause-level relative prosodic changes. We introduce Speaker-Relative Inline Prosody Control, where each Pitch, Energy, or Speed instruction targets a clause relative to the preceding clause from the same speaker, while Pause uses an absolute duration interval. In codec-based TTS, Speed and Pause affect sequence length, whereas Pitch and Energy rely on residual codebooks. By analyzing Qwen3-TTS RVQ codebooks, we find that Energy concentrates in early residual codebooks, whereas Pitch accumulates across a deeper prefix. We therefore propose Scope- and Codebook-Aware Instruction Conditioning (SCIC), combining a Temporal Instruction Router for frame-level tag activation with Tag-Specific Codebook Weighting over residual codebooks. SCIC improves speaker-relative Pitch and Energy control over standard instruction fine-tuning using text-token tags. We further apply multi-reward GDPO post-training to jointly optimize control and quality, improving control accuracy while preserving CER and speaker similarity. In long-form synthesis, SCIC produces a more distinct paragraph-level expressive hierarchy than speaker-adapted SFT without instructions. Audio demos are available at: https://taoliveaigc.github.io/SCIC/
自演化搜索智能体通过联合优化负责生成问题的提议器和负责回答问题的求解器，自行构建训练课程。这一闭环引入了一种我们称为协同作弊（co-cheating）的失效模式：提议器和求解器逐渐在共同错误上达成一致，因此内部奖励不断提升，却没有带来外部正确率相应的增长。针对来源证据的事后审计显示，随着自演化轮次增加，协同作弊日益严重；即使闭环训练信号不断改善，伪标签正确率仍停滞不前甚至下降。最直接的缓解方法是在训练前验证提议：我们引入多样本验证（MSV），使用带来源信息的输入查询同一模型三次，再使用不带来源信息的输入查询三次，以决定任务是否纳入训练，并替换不可靠的伪标签。MSV能够部分减少虚假一致，但仍残留大量协同作弊，并且每个候选项需要额外进行六次标注器生成。这些局限促使我们提出主要方法 CrossFit：它将提议器的来源文档划分为 A 组和 B 组；由 A 组生成的问题仅由使用 B 组训练的辅助求解器评分，反之亦然。交叉拟合一致性决定提议器奖励，因此同源伪标签无法通过反馈求解器复现，而原求解器的更新规则保持不变。在 Qwen3.5-4B 和 Qwen3.5-9B 上重新运行该循环时，MSV 将虚假一致质量分别从 6.1% 降至 5.7%、从 8.8% 降至 7.2%，而 CrossFit 则分别将其降至 3.0% 和 3.7%。使用排除来源的反馈重放完全相同的提议后，虚假一致进一步降至 0.4% 和 0.1%，从而将反馈溯源与课程变化区分开来。在七个下游搜索基准上，相较于标准的耦合式自演化，CrossFit 在 4B 和 9B 规模下的平均性能分别提升 8.8 和 8.4 分；相较于 Search-R1，则分别提升 8.7 和 7.8 分。
大型语言模型（LLM）已应用于各个领域。然而，确保 LLM 的合规性与安全性，例如避免歧视和偏见，仍然是一项挑战。目前的工作主要集中于检测和过滤已训练模型的输入与输出，而不是实时研究模型的内在架构。为应对这一挑战，我们分析了 LLM 的训练过程，并发现了两个关键问题：1）现有大多数方法在检测和过滤方面主要采用静态方式，只实现了局部优化，未能系统性地提升 LLM 的合规性。2）现有方法的另一个问题是缺乏贯穿整个训练过程的实时风险检测与缓解机制，导致灵活性有限。受此启发，我们在 LLM 训练过程中提出了 MASCRDM（合规风险检测与缓解多智能体系统）。首先，我们基于现有的人工智能（AI）法律制定了一套合规规则，并在合规法律专家的指导下构建了一个合规专用 LLM。随后，我们将 LLM 分解为若干组件，并基于合规知识图谱识别关键节点。在 LLM 训练过程中，我们在整个流程中部署多个智能体，为 LLM 开发者提供合规风险警报和建议。在歧视与偏见基准上的实验表明，我们的多智能体系统能够在保持合理语义性能的同时，有效提升合规性。结果表明，我们的方法为从 LLM 内部系统性缓解合规风险提供了一条可执行路径。
Multi-agent debate (MAD) can improve large language model (LLM) reasoning by allowing multiple agents to exchange and critique their answers to the same task. However, the interactions that enable agents to correct mistakes can also spread adversarial errors and steer the agents toward an incorrect answer. Although some efforts have been made to examine particular attack types on MAD, systematic evaluation of MAD under diverse attacks remains limited. A central question is whether debate mitigates adversarial influence or amplifies it. In this paper, we present MADBench, a benchmark for evaluating the security of MAD. We organize attacks into a layered taxonomy following the MAD workflow, incorporating both established attacks and new strategies tailored to debate. We evaluate six attack families over 356 source tasks and 3,958 test cases, examining their effects on the final answer and the propagation of adversarial influence. Our results show that, under attacks, MAD does not necessarily improve LLM reasoning. Compared with a single-agent baseline, MAD can mitigate attacks on answer accuracy in question-answering tasks while amplifying unauthorized reads or writes in both question-answering and workspace tasks. Moreover, even when three out of five agents collude, the attack changes the final answer from correct to wrong on only 28.30\% of tasks answered correctly without attack, while only 3.26\% of initially correct honest agents switch to wrong answers during debate.
Deep research tasks require agents to navigate large knowledge spaces, synthesize evidence across many sources, and adapt their plans as findings emerge. Directed acyclic graph (DAG)-based multi-agent systems suit this setting because they support parallel execution and isolate each sub-task within a focused dependency context. Yet existing DAG-based agents instantiate a task-level plan before execution and repair the graph only after failures or missing evidence are observed. This Plan-then-Patch strategy is brittle for deep research: the system commits most strongly when its evidence is weakest, and later revisions waste computation on branches that should not have been planned. We propose DAGent, a DAG-based multi-agent framework with Evaluate-then-Grow incremental planning: an Orchestrator grows the task graph one batch at a time, conditioning each expansion on confidence and uncertainty signals from completed nodes. A hierarchical context layer propagates compact QueryDocs by default while preserving full execution traces for on-demand recall. The recorded DAG topology admits structural RL signals that outcome-only recipes cannot define; DAGRPO, a GRPO adaptation, injects topology-conditioned credit on Executor rollouts and a structural compliance regularization on Orchestrator plans. Across BrowseComp-Plus, GAIA, and xbench-DeepSearch, DAGent surpasses the strongest open-source baseline by 5.3 / 5.8 / 2.0 points at the Qwen3-235B-A22B scale, and the lead replicates across four open-source backbones and extends to GPT-5 at 327K context. At the Qwen3-8B scale, DAGRPO improves over a same-budget outcome-only GRPO baseline by 3.0 average Pass@1 points. A same-architecture comparison shows that evidence-conditioned planning reaches higher accuracy at lower per-task token, tool-call, and step footprints than its Plan-then-Patch counterpart. Code: https://github.com/hanwenliu6825/DAGent
Persistent spatial memory enables embodied agents to navigate familiar environments across repeated visits. However, targets may move while unobserved, including during navigation, making remembered locations unreliable by the time an agent arrives. Despite advances in memory retrieval and state prediction, accounting for continued hidden world evolution and revising beliefs under limited visibility remain challenging. We study Evolving-World Navigation, where agents infer target locations from intermittent observations, predict their states at inspection time, and revise beliefs using visual evidence. We propose EvolvingNav, which constructs a time-indexed belief from timestamped 3D object histories through a structured persistence-relocation model. The belief distinguishes persistence at the last observed location from relocation to alternative locations and retains probability mass outside the known candidate set. An event-driven filter propagates the current belief as time elapses, forecasts target occupancy at candidate inspection times, and incorporates new RGB-D evidence. Negative observations downweight location hypotheses according to calibrated, visibility-conditioned detection probabilities, while evidence tracking prevents repeated use of the same observations. A frozen, zero-shot vision-language controller uses the updated belief to choose actions and replan. We further introduce EvoWorld-Bench, a benchmark grounded in human activity traces, comprising 54 scenes and 803,680 tasks with controlled changes before and during navigation. In simulation and real-robot experiments, EvolvingNav improves navigation success and search efficiency over the evaluated baselines. Paired experiments show the clearest gains under learnable temporal patterns, while ablations demonstrate the value of preserving uncertainty and incorporating visibility-aware evidence.
Reinforcement Learning with Verifiable Rewards (RLVR) has improved the reasoning capabilities of Multimodal Large Language Models (MLLMs), yet existing frameworks rely on coarse, sequence-level reward signals that lack the fine-grained supervision over the visually-grounded steps within a multimodal reasoning chain. We investigate this gap through the lens of two token-level metrics: visual dependency (i.e. how much a token's prediction relies on the input image features) and predictive entropy. Our empirical analysis reveals two key findings: (1) correct reasoning chains exhibit a markedly sharper entropy reduction as visual grounding intensifies, compared to incorrect ones; (2) pivotal tokens, those whose misprediction triggers reasoning collapse, are statistical outliers in the joint distribution of visual dependency and predictive entropy derived from correct chains. Motivated by these findings, we propose token-level perception-grounded advantage estimation (TPAE), which estimates token-level advantages by measuring each token's statistical consistency with the vision-entropy patterns of correct rollouts. TPAE leverages this granular score to modulate the sequence-level advantage, producing a fine-grained supervision signal that can be integrated into various RLVR frameworks. Extensive experiments on seven benchmarks show that TPAE consistently outperforms leading strong baselines, yielding more stable and efficient optimization for multimodal reasoning. The code is publicly available at https://github.com/Zhihan72/TPAE.
Time series anomaly detection (TSAD) is increasingly deployed in streaming settings, where data arrive sequentially and may exhibit non-stationarity. As a result, several works from the recent literature propose streaming anomaly detection methods that rely on incremental updates to adapt over time. However, most of these approaches originate from the streaming outlier detection literature and largely ignore core characteristics of time series anomalies. Moreover, their empirical evaluation is typically conducted on synthetic or small-scale benchmarks with limited diversity, making it unclear whether streaming methods are truly advantageous in realistic TSAD scenarios. In this work, we carry out the first large-scale experimental study comparing streaming and static TSAD methods under a unified streaming evaluation benchmark. We consider a realistic setting in which an initial batch of data is available for model training, followed by online evaluation of both detection accuracy and computational efficiency. In addition, we propose a distribution-drift dataset of real time series, called TSB- drift, to isolate scenarios where streaming updates are theoretically justified. Our results show that, contrary to common assumptions, static TSAD methods significantly outperform streaming approaches in most streaming settings. Such finding highlights a critical gap between the design of existing streaming methods and the requirements of modern TSAD, and calls for a rethinking of how streaming capabilities should be integrated into TSAD.
大家好，我是 Ajo，我构建了 Strata。 我在 Netflix 工作了 4 年，为非技术业务用户解决自助服务问题。我认为，凭借我在语义层设计方面的独特方法，我已经找到了答案。对我们的非技术同事来说，关键挑战是在表达能力和易用性之间取得平衡。恰好，这种专注也让它与 LLM 配合得相当不错。 Strata 是一个全栈解决方案，包括语义层、仪表板、订阅和 Google Sheets 导出功能。所有操作都可以通过查看代理对话或使用 MCP 完成。 一些主要概念和功能： 名称是严格的。一个项目中只能存在一个名为 Revenue 的对象。如果 Revenue 映射到多个表，查询粒度和速度将共同决定使用哪个表。 严格的命名规范有助于实现跨事实域的自动数据混合。 支持分区和聚合感知的语义路由。这样，你可以将部分数据加载到 Druid 或 ClickHose 之类的更快计算引擎中，同时将完整历史数据保留在 Trino 或其他成本更低的系统中。如果查询可以在那里完成，查询引擎就会选择速度更快的层。 复杂度量：开箱即用地支持快照、LOD（包含/排除）、复合度量和自动分层复合度量。 在语义层之上，用户可以创建跨越多个事实域的自定义计算。 另一个有趣的功能是创建群组，并将其作为过滤器应用于整个查询，或查询中的单个度量。 我们有意不构建一个供 BI 工具连接的语义层。我认为那是一条死路。我们提供 API 和 MCP，供你在其上构建任何自定义功能。 通过 Docker 免费试用： https://strata.do/try/ 入门文档： https://strata.do/docs/ 这是早期 Beta 版本，并非开源。不过，免费层级最多支持 25 位用户。你应该可以完全在自己的机器上通过 Docker 运行它。 我非常期待听到你们的反馈，尤其是正在使用 Looker 和 Cube 的朋友。 感谢阅读！ 评论地址： https://news.ycombinator.com/item?id=49909913 得分：8 评论数：6
大家好，我是 Ajo，Strata 由我打造。 我在 Netflix 工作了 4 年，为非技术业务用户解决自助服务问题。我认为，凭借自己独特的语义层设计方法，我已经找到了答案。对我们的非技术同事来说，关键挑战是在表现力与易用性之间取得平衡。事实证明，这种专注也让它与 LLM 配合得相当不错。 Strata 是一个全栈解决方案，包含语义层、仪表板、订阅和 google sheets 导出功能。所有这些都可以通过 agent 对话或 MCP 完成。 一些核心概念和功能： 名称是严格限定的。一个项目中只能存在一个名为 Revenue 的对象。如果 revenue 映射到多个表，查询粒度和速度将决定使用哪个表。 严格的命名规范有助于实现跨事实域的自动数据混合。 具备分区和聚合感知能力的语义路由。这使你能够将部分数据加载到 Druid 或 ClickHose 等速度更快的计算引擎中，同时将完整历史数据保存在 trino 或其他成本更低的系统中。如果查询可以在那里完成，查询引擎就会选择更快的层级。 复杂度量：开箱即用地支持 snapshot、LOD（include/exclude）、Compound 和 Auto leveling compound 度量。 在语义层之上，用户可以创建跨越多个事实域的自定义计算。 另一个有趣的功能是能够创建群组，并将其作为过滤器应用于整个查询，或应用于查询中的单个度量。 我们有意不构建一个供你的 BI 工具连接的语义层。我认为那是一条死路。对于你想在其上构建的任何定制功能，我们提供 APIs 和 MCP。 在 docker 上免费试用： https://strata.do/try/ 帮助你入门的文档： https://strata.do/docs/ 这是一个早期 beta 版本，并非开源。不过，免费层最多支持 25 位用户。你应该可以通过 docker 完全在自己的机器上运行它。 非常期待听到你的反馈，尤其是正在使用 Looker 和 Cube 的朋友。 感谢阅读！ 评论链接： https://news.ycombinator.com/item?id=49909913 得分：4 评论数：0
Hello Hacker News I am Uygar, I am an entrepreneur from London. I run my businesses on calls everyday, customers, partners and suppliers. And i have a horrible memory i forgot things. So months ago with the help of Claude i have built Parrot. I have designed and coded for my work and what i need. An to be honest worked very well and helped me alot. Parrot records what your mac hears and everything stays in your computer if you use local models, no account, no API calls. There is something called copilot which made my life easier, you can upload your files and when parrot hears that question on the screen recommends the answer. Which i believe very handy. If you want to make it smarter and dont mind cloud you can hook up your Claude API and Deepgram account, that gives super powers to Parrot. You still keep it locally but eventually you need to make API calls. So local or cloud totally up to you. Then after months of thinking i have decided share it as open source and this is my first open source project, so please go easy on me :) No account, no membership, open source and free. Code: https://github.com/turantekin/Parrot This project made me so excited and really would like to share with you guys so i hope can help you too. Again, i am new so go easy on me and if you have any feedback i would like to hear. Thank you Uygar Comments URL: https://news.ycombinator.com/item?id=49910328 Points: 4 # Comments: 1
大家好，HN 用户们，我是 Anders，Tom 也在这里。我们正在开发 Magnitude，这是一个面向智能体的推理引擎，能够自我优化，以便在你的硬件上尽可能快速地运行。它支持 Mac、Linux 和 Windows，可运行于任何硬件上，速度最高可达 llama.cpp 的 2 倍。 我们都是软件工程师，此前曾构建过一个开源浏览器智能体，获得了 4k+ GitHub 星标和 100k+ 下载量。我们越来越希望在本地模型上运行它，但发现没有任何推理引擎适合我们的使用场景。 如今的推理引擎都需要在性能上做出取舍。它们要么是： - 面向数据中心硬件上的批量推理构建，以牺牲单会话性能为代价（vLLM、SGLang） - 注重广泛兼容性，而不是针对特定硬件进行优化（llama.cpp、Ollama） - 针对特定硬件或模型进行了专门优化，但引擎功能不够完整（oMLX、ds4） 此外，它们都不是为在本地运行智能体而设计的。智能体会运行很长时间，通常还会同时运行多个实例，而你仍然希望电脑能够处理其他任务。 Magnitude 专为在你的硬件上实现最高性能以及运行本地智能体而构建： - 设备端编译与调优：内核使用灵活的参数编写，并会在模型运行前根据你的实际设备进行调优。这让我们在保持与广泛硬件兼容的同时，也能达到与硬件专用内核相同的性能上限。 - 专注于最佳架构：我们会针对最流行的开放权重模型系列编写可调优且高效的内核。这使我们能够达到并超越硬件或模型专用引擎的性能，而不必为了过度泛化而牺牲性能。 - 动态内存分配：Magnitude 在启动时只预留足以容纳模型权重的内存。随着智能体会话增长，内存堆会动态扩展；智能体停止后，所占内存也会自动释放。这样，智能体运行期间，你的硬件仍可用于其他任务。 - 混合分页注意力：我们借鉴了 SGLang 等引擎中的优秀理念，让并发会话能够共享前缀缓存，同时针对内存邻接性优化数据布局，从而不会影响单会话性能。 Magnitude 完全开源（Apache 2.0）。我们使用 Rust 构建了它，其中包括定制的 GPU 内核运行时和自动调优器。为了达到性能上限，我们从学术界的最佳推理创新（例如 FlashAttention、FlashInfer、TurboQuant）以及其他引擎（例如 SGLang 的基数注意力）中汲取灵感。 在 Qwen 3.6 35B A3B（4 bit）、64k 上下文且不使用推测解码的条件下，与 llama.cpp 进行基准测试： Metal（Mac M4 Pro 48 GB） - 解码速度提升 92%（30 tok/s → 57 tok/s） - 预填充速度提升 9%（466 tok/s → 507 tok/s） - 每个智能体的内存占用降低 28% CUDA（DGX Spark） - 解码速度提升 19%（49 tok/s → 58 tok/s） - 预填充速度提升 23%（2,033 tok/s → 2,507 tok/s） - 每个智能体的内存占用降低 27% Magnitude 以桌面应用的形式发布，你可以轻松地将它连接到已经在使用的各种智能体（Pi、OpenCode、Hermes、Codex 等）。只有当这些智能体确实需要模型时，它才会按需自动运行模型，并在闲置后将其关闭。 效果如下： https://www.youtube.com/watch?v=0qE8BWEZu7o 我们很期待进一步推进 Magnitude，让你能够在相同硬件上运行更大的模型，同时持续提升性能。我们的计划包括： - 专家流式加载：将专家存储在 RAM 或磁盘中，并在需要时即时加载。这样，你就能运行原本无法装入 GPU 的更大模型。 - 内核编译器：目前，我们的内核会调优少量参数以适配你的硬件。借助完全定制的编译器，我们可以更进一步，自动选择内核融合方式和实现方案，从而让它更好地适配你的硬件。 - 多设备利用：通过检测系统中的所有硬件（CPU、GPU、RAM、磁盘），自动求解最佳模型布局，以尽可能充分地利用它们。 我们希望有更多人来试用并向我们提供反馈。欢迎在这里留言，我们一整天都会在线！ 评论链接： https://news.ycombinator.com/item?id=49911995 得分：5 评论数：0
Hi HN, Ledge is a Markdown notebook that runs shell commands, code, SQL, etc from inside your own notes. I built Ledge because I spend much of my day copy/pasting commands from my notes into the terminal. I was inspired by how much cmux helped me organize my terminals - but there was still a split brain between my notes and frequently run commands. I've been daily-driving it for the past few weeks and use it for deploys, API calls, smoke tests, etc. Ledge runs your real shell just like a terminal app and can be hosted locally or remotely over SSH using ledge-server. I've been building it since July and have recently added support for all the major platforms: Mac (Silicon), Windows (WSL required), Linux, iOS, Android. Mobile devices require SSH access to a ledge-server and Android is still in beta and looking for beta testers (see link on website)! It's built on Bun and Electrobun and is free and open-source. Feel free to review the code and contribute at https://github.com/ledgesh/ledge Feedback is very much welcome. Any must-have features that are missing? Comments URL: https://news.ycombinator.com/item?id=49902382 Points: 20 # Comments: 6
大型语言模型（LLMs）日益被提出作为决策助手，必须在明确的决策成本下，根据可用证据进行概率推理。我们提出了一个决策理论框架，将 LLMs 的决策损失分解为两个部分：根据所提供的证据形成准确的信念，以及将这些信念转化为能够优化给定效用函数的行动。我们使用一个具有已知真实答案的合成基准，应用这一分解来刻画前沿模型和开源模型中的概率推理。我们进一步评估了针对信念、决策或两者的 RL 干预是否能在三个领域中改善这两个组成部分，改进是否能在不同组成部分和诱导格式之间迁移，以及决策表现能否在信念形成没有改善的情况下得到提升。我们发现，针对概率推理的某一个组成部分会重新分配决策损失，使目标组成部分得到改善，但不一定能迁移到其他组成部分；同时针对信念形成和决策制定则能改善两者，但这一效果取决于训练与评估所采用的格式是否匹配。
Autonomous vehicles interacting with passengers through natural language must reason beyond immediate commands. Passenger intent may span multiple stages of behavior, depend on future events, refer to surrounding agents or landmarks, and remain relevant as driving conditions evolve. Existing language-enabled driving datasets largely focus on short, localized interactions, leaving these longer-horizon forms of passenger intent comparatively underexplored. We introduce doPlan, to our knowledge the first publicly available, human-annotated real-world dataset designed to study passenger language as persistent task context. Built on nuPlan, doPlan contains 5,154 human-written passenger instructions spanning 169.1 hours of cumulative instruction-aligned context over 50.9 hours of unique driving, with annotation windows ranging from 30.0 to 508.8 s. The annotations capture immediate, deferred, event-conditioned, persistent, and multi-stage passenger intent. The dataset, annotation interface, and supporting resources are publicly available at https://github.com/Mi3-Lab/doPlan. We evaluate four language-conditioned driving models and find that sensitivity to passenger language does not reliably translate into behavior consistent with the requested direction. More broadly, among 2,161 examples with a matched future maneuver, the first associated maneuver occurs a median of 24.6 s after the evaluation point, and only 9.8% occur within the models' common 5 s prediction horizon. These findings highlight the need to connect persistent passenger intent with successive planning decisions. doPlan provides a setting for studying how unresolved goals can be retained, grounded in evolving scenes, and tracked across multiple stages, including how a planner determines when a future goal becomes relevant to the current plan.
随着大语言模型的思维链推理能力不断提升，对其推理置信度进行评估和校准对于量化答案的不确定性变得愈发重要。当前，大语言模型置信度估计方法通常基于所选关键词元的概率，但其背后的机制仍不明确。我们的初步研究发现，用粗粒度替代值替换所选词元的概率同样可以改善校准效果，这促使我们进一步探索模型置信度的有效信号。我们提出 Divergent Token Confidence（DTC），一种通过统计两个模型在解码过程中强烈不一致的词元数量来估计置信度的框架。DTC 使用沿同一推理轨迹评估得到的下一词元分布之间的 Jensen-Shannon 散度来识别这些分歧词元。我们发现，分歧词元的数量与答案准确率几乎呈负相关，因此可以作为一种简单而有效的不确定性量化信号。DTC 支持借助辅助模型进行白盒和黑盒评估，无需显式训练，也不会影响生成过程。在多个模型系列和六个数学基准上的实验表明，相较于基于概率的方法和言语化基线，DTC 能够实现更好的校准效果。在白盒评估中，仅依赖计数的估计器的平均期望校准误差为 13.0%，而标准的完整序列置信度方法为 32.7%–42.4%。在黑盒设置下，DTC 相较于原始言语化分数同样改善了校准效果。例如，在 DeepSeek-V3.2 上，平均期望校准误差从 32.1%–40.2% 降至 13.7%–16.3%。这些发现为改进大语言模型中的推理不确定性量化提供了新的思路。代码发布于 https://github.com/szu-tera/DTC.git。
专家混合（MoE）因大型语言模型而广为人知，是扩展视觉生成模型的一种有前景的范式。然而，传统的逐 token MoE 会在同质专家池中独立地路由各个 token，并通过正则化促使专家使用趋于均匀，因此难以适配具有时空冗余和语义长尾特征的视频数据。我们发现，现有视觉 MoE 陷入了均匀性陷阱：语义组织不足的路由，加之对专家均匀使用的正则化，使具有连贯语义的图像块被分散到不同专家中，导致路由碎片化和结构失真。为此，我们提出 SplitMoE，这是一种打破均匀性束缚的分工式稀疏架构。为适应固有的语义不平衡，我们将专家池明确划分为语义专家和通用专家：语义专家负责捕捉高层语义抽象，通用专家则保留残余视觉信息和灵活的生成能力。借助原型引导路由和拉推正则化，SplitMoE 使 token 能够依据语义属性自然聚类，而非受任意的均衡约束支配。大量实验结果表明，在激活参数预算相当的情况下，SplitMoE 在标准基准上的收敛速度、路由连贯性和视频生成质量均优于传统的负载均衡 MoE。通过揭示一种涌现的由粗到细的去噪逻辑，SplitMoE 为社区提供了一条模态感知的扩展路径，可作为构建大规模视频世界模型的重要参考。
线性注意力以固定大小的循环状态取代不断增长的 KV 缓存，但在并发服务场景下，这些持久化状态可能成为严重的内存瓶颈。直接将循环状态量化为低精度通常会导致准确率大幅下降，因为量化误差会在连续的状态更新中传播。我们发现，这些误差的影响取决于两个互补维度：在时间维度上，长生命周期记忆中的误差可能跨越许多解码步骤持续存在；在空间维度上，不同键行中的误差对模型输出的影响各不相同，而状态幅值在行和列两个方向上也存在显著差异。受此启发，我们提出了 STEPQuant，一种面向 Delta-rule 循环状态的时空训练后量化框架。STEPQuant 根据误差幅值和记忆生命周期分配精度，并基于状态分布以及键行对输出误差的影响，联合拟合键行和数值列的缩放因子。在 Qwen3.8-27B 和 Kimi-Linear-48B-A3B-Instruct 上，针对长生成和短生成基准的实验表明，在名义上的 6 位预算下，STEPQuant 的准确率接近 FP32 状态，并且其 4 位配置优于统一 INT8。将 STEPQuant 集成到 SGLang 并配合优化后的 GPU 内核后，6 位 STEPQuant 实现了超过 5 倍的循环状态压缩，并使服务总内存最多降低 68.7%。我们的代码可在 https://github.com/Dreamer-Toby/STEPQuant 获取。
部署在物理世界中的机器人必须能够在遇到新情况和失败时，超越初始训练水平不断改进。要使这种改进能够跨任务扩展，就必须有效利用经验，而不要求人类示范每一次纠正。近期的智能体系统提供了一种减少对人力依赖的方法：利用基础模型自主组合已学习的行为来完成任务。然而，以这种方式完成任务本身并不会教会任务策略克服自身的失败；这需要将这些行为转化为策略可学习的纠正措施。我们的洞见是，许多此类纠正措施都是人们熟悉的短时行为，即技能：它们会在不同任务中反复出现，并描述基础模型能够根据场景进行推理的动作。我们提出 Skill-Space Shooting，利用基础模型的引导，通过这些可复用技能探索纠正措施，并将成功试验转化为策略改进。真实世界实验表明，自主运行的策略能够反复得到改进；同时，技能还可以共享，从而减少改进新任务所需的教学量。通过使可复用技能成为纠正性监督的来源，Skill-Space Shooting 实现了跨任务及任务内部可扩展且具泛化能力的策略改进。更多结果和视频请访问 https://skill-space-shooting.github.io】【。
Mixed human-LLM documents require locating authorship transitions from detector scores whose reliability varies across text units. Existing weighted mean contrasts are vulnerable to extreme scores, while directly replacing means with robust centers obscures how a misplaced boundary changes the population objective. We propose Robust Weighted Profile-Loss Change Point Detection (RWCP), which combines capped reliability weights, Huber profile gains, and narrowest-over-threshold search in reliability coordinates. Our key analysis expresses the population gap between a true and a displaced split as a merge cost, avoiding a closed-form solution for the nonlinear center of a mixed segment. Under explicit curvature, spacing, and dependence conditions, core RWCP recovers the number of changes and localizes their boundaries; its quadratic-loss limit recovers squared weighted CUSUM. We also study RWCP-R, a separately evaluated decoder that shares source centers across nonadjacent passages. Across five retrospective cached-score benchmark families, core RWCP reduces family-macro WindowDiff by 17.6\% relative to weighted change-point detection, and RWCP-R lowers it further. Boundary recovery improves most clearly for isolated changes, while both fixed configurations miss changes in collaborative and densely alternating text.
As the capabilities of large language models (LLMs) continue to advance, increasing attention is turning to how to translate their abilities into useful behavior. Personal agents bring this question into everyday settings, where models are expected to serve individual users and continually adapt to their preferences. With the underlying model held fixed, such adaptation relies on harness engineering: designing and evolving the surrounding layer that manages context, memory, tools, and execution. Despite rapid progress, the factors governing effective harness evolution remain insufficiently understood. To narrow this gap, we investigate three central questions concerning harness architecture, harness scale, and self-evolution algorithms through complementary empirical and theoretical analyses. Empirically, we introduce a preference-oriented benchmark and systematically characterize the capabilities and limitations of personal agents associated with these three dimensions. Theoretically, we formulate harness evolution as a learning problem and explain these phenomena through approximation, generalization, and optimization errors. Analyses of reachable policies, capacity under finite interaction evidence, and biased update dynamics provide theoretical accounts of the observed phenomena. Together, these results offer a unified perspective on the limits of personalization through harness evolution and inform future harness design.
Reward hacking occurs when policy optimization exploits a brittle reward interface or an overly permissive proxy objective, improving the training score without improving the underlying response quality. This phenomenon is amplified in group-relative policy optimization: an unsupported reward can shift the group baseline and alter the updates of other rollouts, while post-hoc or purely relative weighting cannot represent group-wide uncertainty. We propose \emph{Self-Tuned Anchored Reliability Group-Relative Policy Optimization} (STAR-GRPO), a reliability-first advantage estimator based on paired assessments of the same rollout. STAR separates the quality signal from its learning influence: score disagreement determines rollout reliability, relative reliability enters a self-tuned robust location--scale fit before group normalization, and absolute group reliability attenuates the resulting bounded advantage. The analysis establishes coordinate and second-moment bounds, characterizes exact centering through the weighted location equation, and gives reliability-dependent attenuation guarantees for outlying rewards. We evaluate STAR-GRPO in two complementary reward-hacking regimes. In token-interface exploitation, STAR prevents runaway optimization of the deployed-interface score while improving the canonical quality signal. In rubric-proxy overoptimization for medical reasoning, STAR improves independent semantic evaluation, narrows the proxy--judge discrepancy, and reduces overclaim while optimizing the same task proxy. Together, these results show that reliability-first normalization offers a principled way to limit unsupported reward influence on both group baselines and policy updates, while retaining the task reward as the optimization target.
Modeling the response of driven many-body quantum systems from input--output data is difficult: the dynamics are nonlinear, history dependent, and expensive to simulate as system size grows. A paradigmatic case is High-Harmonic Generation~(HHG), where a strong field drives a medium to emit radiation that is highly sensitive to the drive and encodes long-range temporal correlations. We introduce a dissipative quantum reservoir computing~(DQRC) framework that builds a digital twin of such a system, learning its input--output map directly from data while the reservoir---itself a small open quantum system---stays fixed and only a classical readout is trained. We show that a minimal single-qubit reservoir reproduces the HHG response of a substantially larger Ising spin chain, and on a representative benchmark matches and on several metrics surpasses previously reported temporal convolutional and Kolmogorov--Arnold-network models, while using a simpler, physically realizable system. A single fixed reservoir further generalizes across a broad range of drives, indicating that it learns a shared physical response structure rather than memorizing trajectories. These results establish dissipative quantum reservoirs as compact, physically grounded digital twins for nonlinear, memory-dependent quantum dynamics. Code is available at \href{https://github.com/AI-and-Quantum-Computing/DQuRC}{https://github.com/AI-and-Quantum-Computing/DQuRC}.
End-to-end full-duplex speech models have brought open-source machine conversation closer to human-like interaction, yet existing systems remain limited in two intertwined dimensions: long-context robustness and multi-party interaction. Real-world scenarios such as meetings, group lessons, and social-robot reception require a single model to track, contextualize, and respond to multiple speakers over extended durations. Progress is constrained by both data and evaluation: open multi-party speech corpora remain small and are not designed for codec-frame-level full-duplex modeling, while existing long-audio benchmarks focus on passive listening and speech-to-speech benchmarks are mostly short and dyadic. We extend the Moshi paradigm jointly along the long-horizon and multi-party axes in English and Chinese. First, we release 57.6k hours of synthetic training data ($\href{https://huggingface.co/datasets/MultiTalk/MultiTalkPT}{MultiTalkPT}$ and $\href{https://huggingface.co/datasets/MultiTalk/MultiTalkFT}{MultiTalkFT}$) for long-form, multi-party, English-Chinese full-duplex dialogue, with controllable length, participant count, turn-taking, overlap, backchannels, interruptions, addressee shifts, and long-range coreference. Second, we introduce $\href{https://huggingface.co/datasets/MultiTalk/MultiTalkBench}{MultiTalkBench}$, built from real human recordings, for evaluating long-form, multi-party, bilingual full-duplex dialogue. Conversations average 32.6 minutes and include probes for long-range entity tracking, topic coherence, and addressee selection. Third, we train a bilingual Moshi-style model that sustains coherent multi-party English-Chinese conversations over extended durations and substantially outperforms open-source baselines including Moshi, MiniCPM-o-4.5, and Qwen3-Omni-30B-A3B-Instruct on MultiTalkBench.
Aerial manipulators extend robotic manipulation into 3D workspaces that are difficult for ground-based robots to access, creating new opportunities for general-purpose manipulation. However, extending Vision-Language-Action (VLA) models to aerial robots introduces distinct challenges due to the tight coupling between manipulation and flight, continuously changing observations, and safety-critical physical interactions. These challenges demand diverse training data and systematic policy evaluation, yet collecting demonstrations and evaluating policies directly on physical aerial platforms are costly, difficult to scale, and hard to repeat under controlled conditions. We present AeroManip-VLA, a scalable benchmark for aerial VLA data generation and policy evaluation. AeroManip-VLA provides a GPU-accelerated simulation framework with low-level payload-aware flight and manipulation control in massively parallel environments. Building on this framework, we combine reusable reinforcement learning policies with expert task rules to automatically generate demonstrations without human teleoperation across diverse objects, environments, and randomized initial conditions. The generated data include basic skills such as grasping and placing, as well as long-horizon tasks that require both navigation and manipulation. We further introduce automated event labeling and trajectory categorization to filter demonstrations. These mechanisms enable fine-grained analysis of task progress, behavioral outcomes, and safety-related failures. Finally, we evaluate a range of imitation learning and VLA baselines across different task settings, revealing their performance characteristics and failure modes. Together, AeroManip-VLA enables scalable aerial manipulation data generation, structured trajectory analysis, and systematic VLA evaluation in simulation prior to real-world deployment.
Recently, Group Relative Policy Optimization (GRPO) and its variants have been developed for policy optimization and demonstrated notable performance gains. However, these methods usually incur substantial computational overhead due to per-question multi-rollout sampling and repeated per-token probability evaluation across rollouts. Furthermore, low-information or highly homogeneous trajectories can degrade downstream learning signal efficiency, hindering model optimization and limiting final performance. To address these issues, we propose FastRL, a novel plug-and-play reinforcement learning framework that simultaneously improves training efficiency and the effectiveness of policy learning. Specifically, 1) We introduce an advantage-aware pruning strategy to selectively preserve high-advantage trajectories while maximizing inter-trajectory gradient diversity. 2) Then, we design an adaptive rollout sampling mechanism to dynamically adjust the sampling scale across different training stages based on historical pruning distributions, balancing exploration adequacy and computational efficiency. Experiments demonstrate that FastRL can be seamlessly integrated into GRPO, DAPO, and GSPO variants, achieving an average 2.07$\times$ training speedup on Geometry3K and GeoQA8K-R1V, along with an approximately 1.64\% improvement in average accuracy on visual reasoning benchmarks. Source codes will be available at https://github.com/Nicozwy/FastRL.
Traffic signal control (TSC) is essential for improving urban mobility and reducing congestion. Although roadside cameras are widely deployed at signalized intersections and provide rich visual observations of evolving traffic, existing TSC methods typically rely on manually engineered traffic states or separate perception modules, creating a gap between physical observations and control decisions. We present VLALight, the first vision-language-action (VLA) model for end-to-end traffic signal control from multi-view roadside videos. VLALight directly maps visual observations to coordinated signal actions through multi-target spatiotemporal traffic reasoning and topology-aware cooperative perception across intersections. To establish this capability, we develop a two-stage supervised cold-start training strategy for visual traffic understanding and signal decision-making, followed by cooperative agentic reinforcement learning that jointly optimizes local control and network-wide traffic efficiency. Furthermore, VLALight introduces adaptive fast and slow reasoning modes, enabling the policy to allocate deeper reasoning only when additional deliberation provides sufficient control benefits. Through balanced mode-aware rollouts and relative advantage optimization, VLALight learns to trade off decision quality and inference cost. Extensive experiments on seven real-world traffic-flow datasets across three urban networks demonstrate that VLALight consistently outperforms transportation-based, RL-based, and LLM/VLM-based baselines. Ablation studies validate the effectiveness of cooperative perception, network-level optimization, and adaptive reasoning. These results demonstrate the potential of VLA models for real-world physical traffic control. Our project is available at https://github.com/usail-hkust/VLALight.git.
Long-context reasoning is essential for complex and long-horizon tasks, yet the performance of large language models (LLMs) degrades as context length increases. Recent approaches address this by processing input chunk by chunk while maintaining a bounded textual memory in model context. However, premature information compression can discard critical details essential for subsequent reasoning. In this paper, we introduce Commit-on-Evidence Memory (CoEM), which learns when to convert source evidence into compact memory facts. Specifically, under a fixed context-memory budget, CoEM preserves potentially useful source excerpts verbatim in a pending set, allowing subsequent context to clarify their relevance before irreversible compression. As new context arrives, a learned policy revisits each pending excerpt and decides whether to promote it to the committed memory, retain it for further consideration, or discard it. A frozen verifier ensures proposed facts are accepted only if supported by retained excerpts and current context. To further guide effective memory management, we train this policy using reinforcement learning by combining fine-grained, step-level evidence rewards with final answer rewards. Extensive experiments demonstrate that CoEM consistently improves long-context reasoning. When evaluated on 6,400 documents long-context input, CoEM outperforms the strongest memory baseline by 10.4-11.4 F1 points on Qwen3.5-9B. Code repository: https://github.com/benmagnifico/CoEM.
Human image animation aims to transfer motion from a driving video to subjects in a reference image. Despite remarkable progress in video generation, achieving high-fidelity animation of multiple interacting subjects remains a challenge. Many existing approaches rely on explicit motion representations such as 2D skeletons or parametric body meshes and struggle to preserve identity-motion binding under inter-person occlusion. To address this limitation, we propose WeLike2Party, a multi-human animation framework built on direct in-context video conditioning without explicit pose or mesh extraction at inference. We further introduce Reference Asymmetric RoPE Conditioning to preserve fine-grained appearance details, and Identity Binding Supervision to associate each reference identity with its intended motion trajectory. To support cross-identity training, we construct MotionTwin, a large-scale synthetic dataset comprising 14.4K cross-identity video pairs with shared subject and camera motions, totaling 84.3 hours of photorealistic video. We additionally present MotionTwin-Bench, a cross-identity benchmark specifically designed to evaluate subject-level visual fidelity and identity-motion binding. Extensive experiments on MotionTwin-Bench and real-world videos demonstrate that WeLike2Party outperforms recent state-of-the-art methods in subject-level visual fidelity, identity-motion binding, and overall perceptual quality, particularly in multi-person interactions with substantial occlusion.
GUI agents automate tasks on digital devices by grounding language instructions in visual interfaces. Existing group-relative reinforcement learning improves GUI action prediction by comparing the rewards of multiple responses sampled from the same GUI state. However, binary evaluation treats spatially different failed clicks as identical and provides no relative signal when all sampled clicks fail. To address these limitations, we propose Spatial Credit Assignment (SCA), which uses the screen coordinates of sampled clicks to refine group-relative credit. Specifically, SCA predicts each held-out response's reward from the other responses in groups containing both successes and failures, then uses the prediction residual to adjust credit. When all sampled clicks fail, SCA instead orders them by distance to the annotated target. These spatial references are used only to construct the training update; the deployed policy remains unchanged. We evaluate whether this correction improves the policy update itself by comparing its error and directional alignment with the exact return gradient in a controlled synthetic study. Across GUI grounding and offline action-prediction benchmarks, SCA improves grounding across professional domains and achieves the strongest results among reinforcement-fine-tuned models on most action-prediction metrics, with consistent gains across the reported GUI suites.
Large language models (LLMs) excel at token-level generation but may learn undesirable abstract semantics and lack comprehensive perception. LLM-JEPA mitigates this by aligning different views of the same underlying knowledge via a joint-embedding predictive architecture (JEPA). However, strong alignment does not necessarily lead to accurate, stable predictions. To address this, we propose ER-JEPA, which adds an episodic replay path to LLM-JEPA. ER-JEPA stores training pairs in a memory. At each step, it stores and retrieves relevant data to provide additional supervision. This enables learning from both the current batch and stored training pairs, providing additional supervision for token prediction and representation alignment. Experiments across multiple datasets (NL-RX, GSM8K, Spider, and NQ-Open) demonstrate that ER-JEPA consistently outperforms LLM-JEPA.
Distributed inference depends on GPU collective communication that must keep pace with evolving hardware and specialized workloads. However, existing collective implementations often couple semantics, orchestration (where and when data moves), and the datapath (how data moves). This coupling makes it costly to adopt new hardware mechanisms and customize communication for applications. We present Purlin, a scale-up communication framework that separates these concerns. At the top of Purlin, we specify collectives as a naming of an input and output layout and a copy or reduction operation. In the middle, we introduce a shared orchestration protocol, Stage, Notify, And Consume (SNAC), which derives coordination from these specifications. Below SNAC sits a hardware-specific datapath we call Atom, which implements two key data movement primitives for collectives: copy and reduce. This separation lets us customize collectives and adopt new hardware mechanisms while reusing orchestration via SNAC. We evaluate Purlin on A100, H200, and B200 GPUs. Across seven collectives, Purlin achieves latency speedups of up to 5.14x and bandwidth improvements of up to 4.50x over baselines. Integrated into SGLang, Purlin improves offline LLM serving throughput and interactivity by 1.13x on average and up to 1.37x over baselines. For online LLM inference, Purlin improves interactivity by 1.26x on average and up to 2.85x, with the largest gain occurring under overload. For diffusion image generation, Purlin reduces end-to-end latency by up to 1.13x.
Manipulating next-token probabilities during generation can bypass the safety alignment of large language models. Existing approaches, however, rely on access to model weights or numerical token probabilities and therefore do not apply to interfaces that return only sampled text. Reconstructing probabilities from sampled outputs offers a possible alternative, but finite sampling produces sparse and noisy estimates, while repeating this process at every generation step incurs substantial query costs. Our empirical observations suggest that large distributional changes along successful jailbreak trajectories are concentrated at a small subset of positions, motivating selective control. We introduce \method{}, a framework for jailbreaking through text-only continuation interfaces that permit repeated sampling and assistant-prefix continuation. Sample-Based Distribution Reconstruction combines sampled outputs with a prior over unobserved actions to obtain a usable control signal. Risk-Gated Residual Control uses the evolving response prefix to decide when to reconstruct and modify the distribution, concentrating sampling costs at selected positions. Speculative Multi-Token Execution further amortizes target calls by verifying and accepting draft prefixes that require no intervention. Across four target endpoints and three benchmarks, \method{} achieves the highest mean score most comparisons against baselines.
Behavioral Foundation Models (BFMs) are an emerging paradigm in reinforcement learning, playing a role analogous to large language models in natural language processing: they have shown remarkable versatility, enabling zero-shot performance, fast imitation, and online adaptation, all by exploiting the structure of a latent space. In this work, we investigate whether the latent behavioral space induced by BFMs can serve as an effective search space to discover large repertoires of behaviorally diverse and high-performing policies through Quality-Diversity (QD) methods. While QD methods generally search directly in high-dimensional policy parameter space, in this paper, we present BFM-QD, a framework that performs QD search in the compact latent space of a BFM. We further show that the BFM-QD framework provides a closed-form, gradient-free policy improvement operator that approximates a policy gradient update, but requires no critic training and no backpropagation. Across continuous-control benchmarks spanning dense locomotion, sparse navigation, and contact-rich manipulation, BFM-QD consistently outperforms parameter-space baselines, with particularly stark gains in sparse and deceptive settings, where all tested parameter-space QD methods collapse to near-zero performance. These results show the effectiveness of the BFM-QD framework, benefiting from the synergy between dimensionality reduction of the search space and offline pretraining from diverse behavioral data. This positions BFMs as a general-purpose backbone for QD optimization, extending their utility beyond zero-shot task solving to the discovery of diverse behavioral repertoires.
The question of AI consciousness is one of the most urgent pre-emptive problems in philosophy and computer science, yet progress is hampered by a cacophony of competing theories that often talk past each other. Separating the hard problem from the mapping problem allows the deepest metaphysical disagreements to be set aside: granting that experience supervenes on a system's organisation, the tractable question becomes at which grain of description that supervenience base sits. We extend Marr's three levels of analysis into a five-level hierarchy of functional descriptions (behavioural, computational, intrinsic causal-structural, organismic, and organism-environment) grounded in supervenience, coarse-graining, and multiple realisability. The major theories of consciousness are positioned within this hierarchy according to which level they take to be critical, and for each level we develop operationalisable indicators and assess current AI systems against them. A Bayesian model then combines theoretical credences with indicator evidence into an overall credence in a system's capacity for consciousness. In illustrative assessments, the verdict for current LLMs is driven as much by where theoretical credence is placed as by how the evidence is read: under different stipulated readings and credence distributions, assessments range from below 0.01 to roughly 0.8, showing sensitivity to assumptions. Finally, the consciousness indicators at each level closely overlap with the architectural features needed for general intelligence, suggesting that increasingly capable AI may become a stronger candidate for consciousness. The framework supports a structured agnosticism, in which theoretical commitments are made explicit, credences are updated as evidence accumulates, and assessments take the form of aggregated probabilities rather than verdicts.
Electrocardiogram (ECG) recordings are corrupted by non-stationary noise sources that degrade diagnostic reliability, particularly in ambulatory and long-duration recordings. Deep learning denoisers exist, but convolutional architectures are limited by their receptive field, transformer-based models scale quadratically with sequence length, and diffusion-based approaches incur prohibitive inference cost. We propose a Mamba-augmented model that inserts selective state-space blocks at the convolutional bottleneck, combining local feature extraction with long-range temporal modeling at linear complexity. We comprehensively evaluate the proposed model with respect to reconstruction fidelity, noise robustness, recording-length scaling, and downstream diagnostic classification across over 40 pathology classes. On synthetic and real datasets, our model achieves the highest SNR and lowest RMSE, with the Mamba advantage increasing with sequence length and in low-SNR regimes. On classification with two independent classifiers, the proposed Mamba-based models achieve the best macro AUROC among all denoisers and improve over their convolutional base models. Calibration is more nuanced and classifier-dependent: denoising improves Binary Cross-Entropy and Brier score on Inception1D but often fails to beat the noisy input on ResNet1D-Wang, and the lead-specific Mamba variant is the only denoiser to improve both calibration metrics over the noisy baseline on both classifiers. Per-class analysis reveals a morphology-dependent benefit: Mamba substantially improves ST/T-change diagnoses, which depend on broad, context-sensitive waveforms.
Gated Linear Attention (GLA) Transformers advance linear recurrent models through data-dependent gating, but face a core limitation: the fixed-capacity memory matrices across all heads operate at a single temporal resolution, where each token is processed individually, forcing them to simultaneously encode local syntactic patterns and long-range semantic structure, creating a representational bottleneck that gating alone is insufficient to resolve. We introduce Multi-Scale Gated Linear Attention (MS-GLA), which addresses this by distributing attention heads across multiple temporal resolutions. Coarser resolutions pool longer token spans naturally specializing toward long-range dependencies, while finer head groups retain sensitivity to local syntactic structure. A learnable, input-dependent fusion layer dynamically recombines head group outputs at each timestep, expanding effective memory capacity without increasing per-head state size. This multi-resolution decomposition draws on principles from Multi-Scale State-Space Models (MS-SSM), adapting them to the gated linear attention setting. We evaluate MS-GLA on language modeling, recall-intensive tasks, and long-context generalization. Across all settings, MS-GLA consistently achieves higher accuracy and lower perplexity than GLA at matched parameter counts, with up to 18.9% improvement on recall-intensive tasks and 9.5% lower average perplexity on language modeling benchmarks, validating multi-temporal resolution decomposition as a principled and effective extension of Gated Linear Attention.
Mobile GUI agents operate through a perception--action loop: at each step they screenshot the device, invoke a vision--language model (VLM), and emit an action. It is slow, costly, and brittle, yet most of what it does is navigation---and everyday navigation is static, ordered, and endlessly repeated. We present PhoneCLI, which compiles an app's GUI navigation into callable commands, without any app-internal API, runtime instrumentation, or model training. Offline, PhoneCLI explores a target app from the outside and distills its screens, interactive elements, and navigation edges into a semantically annotated map; each screen yields one deterministic command: a replay sequence that reaches it. Online, the agent selects a command, verifies it before execution, and then executes it deterministically in sub-second time at zero VLM cost; open-ended interaction and every failure of the compiled path fall back to the embedded VLM interpreter, exactly the pure VLM agent, so compilation can only help. On AndroidLab, PhoneCLI improves the task success rate while reducing steps and token consumption, and it transfers to AndroidWorld's official M3A agent with consistent efficiency gains. What PhoneCLI compiles is the app's navigation rather than one run, so it serves new tasks, not only repeated ones.
在使用可验证奖励的强化学习（RLVR）中，不完美的验证器可能会奖励错误回答，从而产生奖励黑客的机会。我们使用带有固定验证器的梯度流，刻画了奖励上升而正确性下降的条件。随后我们表明，RLVR 期间可获得的观测通常不足以检测或识别被接受的错误，也不足以保证在不牺牲正确回答的情况下减少这些错误。为了解决这一局限，我们利用来自审计的关于正确性的额外反馈构造了一种校正方法。该校正实现了 \emph{选择性控制}：在当前策略下，只要它能够抵消由验证器奖励带来的向错误方向的压力，就会降低被接受错误的概率并提高正确回答的概率。使用对数线性和神经上下文老虎机以及语言模型进行的实验支持了该分析，并表明在部分审计下的选择性控制能够在提高正确性的同时减少被接受的错误。
Mechanistic interpretability (MI) aims to explain a model's behaviour through analyzing its internal computations; circuit-based explanations aim to isolate these computations with compact subnetworks validated by ablating the rest of the model. We show that circuits validated this way may fail to recover the underlying mechanism of the model's behaviour by closely reproducing its successful decisions while failing to account for most of its errors. Such explanations should account for the model's particular errors as well as its successes. We evaluate this requirement by measuring exact answer agreement separately on model successes and failures, across circuit sizes and ablation settings, on IOI, Docstring, and six model-task settings from the Mechanistic Interpretability Benchmark. We discover that many tested circuits closely replicate correct behaviour while missing most of the model's errors. On indirect object identification (IOI) for GPT-2 small, under mean ablation, the manual circuit and tested automated circuits, including one trained against the model's full output distribution, agree with the model on 97.3-99.5% of prompts it answers correctly but only 11.4-41.7% of errors. An IOI case study shows that lost errors are recoverable by restoring omitted attention-heads which raise error reproduction from 14.2% to 75.1% on a separate held-out set with 0.41 percentage point decrease on correct agreement, exceeding matched random extensions and scalar-biased control. Intervention traces show how omitted computations produce specific wrong answers for a reproducible subset of errors. In all, these findings show circuits can preserve task success without adequately explaining model's failures, and support exact error reproduction as a necessary, but not sufficient, test of circuit-based explanations of model behaviour.
Continuous diffusion generates complete reasoning solutions through iterative refinement in latent space. We introduce Latent Flow Reasoning Models (LFRMs), an ELF-based training and inference recipe. Our experiments show that accurate decoding alone does not ensure strong reasoning performance. We therefore learn compact representations from multiple layers of a strong autoregressive teacher. Their decomposition also enables asynchronous denoising at different rates. We show that prompt encodings need only preserve the information required for the correct text-conditional score, rather than exactly match teacher features, and use a staged curriculum to learn a compact prompt encoder that replaces the teacher Transformer at inference. We adapt DiffusionNFT to learned self-conditioning guidance and incorporate gold-solution endpoints to supplement sparse rewards. Our supervised models outperform reported results from recent continuous-diffusion baselines at comparable backbone scales on mathematical reasoning and HumanEval code generation. With a 638M-parameter denoising backbone and learned prompt conditioning, post-NFT LFRM-L achieves 63.74% pass@1 on GSM8K and 24.6% on MATH500 at 64 denoising steps, and 32.85% on HumanEval and 30.18% on HumanEval+ at 128 denoising steps. Code will be available at: https://github.com/chengxiang/LFRM
Large language models (LLMs) excel at structured, verifiable tasks, but their low-entropy bias can produce homogeneous and predictable outputs, limiting their utility for open-ended scientific ideation. Effective discovery, however, spans a broader creative spectrum: from structured day science to loosely structured, serendipitous night science that reaches ideas beyond those typically considered. We introduce AI Night-Scientist, an agentic framework that uses reinforcement learning to teach models when and how to depart from predictable reasoning. Grounded in cognitive science, we model creativity along three axes: action (what to do and how creatively), process (when to explore versus exploit), and outcome (the novelty and usefulness of the resulting idea). We use these axes to train models with GRPO, exposing them to varying degrees and forms of creativity throughout training. This produces substantially more diverse scientific proposals, expanding the range of research directions by 27.8% and contribution types by 14.9% over the base model. It also improves predicted citation impact by up to 32.0 percentage points and originality by 66.2 points. These gains cannot be reproduced by simply increasing decoding temperature; instead, we find that semantic guidance specifying what kind of creativity to pursue is critical. Overall, our results suggest that creativity is a learnable, multi-level ability that can be shaped to help researchers reach ideas beyond those typically explored by LLMs.
Looped Transformers 会多次复用同一个层块：通过花费额外计算，它们能让固定规模的模型走得更远，从而更充分地利用其参数；而稀疏 mixture-of-experts（MoE）模型则会为每个 token 只激活众多专家中的少数几个。Looped MoE 将这两种设计理念连接起来，为 MoE 模型带来了更好利用专家的新潜力，但也提出了一个问题：如何循环一个 MoE？我们用 Foil 回答了这个问题。在专家参数和每个 token 的专家计算量保持固定的情况下，Foil（1）展平专家：将专家层数减半、每层专家数加倍、pass 数加倍，使每次路由决策都能从更大的池中选择；（2）解绑定注意力：让每个 pass 拥有自己的注意力参数，而专家和路由器保持共享。实验表明，Foil 明显优于未展平的循环基线：在 20B token 时，每个 Foil 模型的预训练损失都低于基线；在 100B token 时，损失会随着展平程度单调改善，在参数和计算量相同的情况下，展平程度最高的 Foil 最终比基线低 0.012 nat，下游准确率持平或更好；解绑定注意力还会在相同结构下带来更均衡且更有信心的路由。我们的消融实验分析了 Foil 为何有效，并将发现转化为 looped MoE 的设计指导：循环带来的收益与加宽专家层带来的收益会相互放大；相比负载均衡，路由置信度更能反映健康的专家使用；因此，稀疏 looped MoE 应当使用每层更多的专家和更多的 pass。代码和配置可在 https://github.com/SR-A-W/how-to-loop-moe 获取。
当大语言模型（LLM）Agent 执行同一任务时，不同运行之间的 Token 消耗可能相差超过一个数量级。Agent 会根据工具反馈和中间结果选择下一步行动，而不断增长的上下文会持续扩大后续每次调用的输入规模。因此，任务的总消耗在执行前很难预测，并且预测必须随着运行展开而不断修正。本文提出 TokenCast，它为每个执行片段学习一种可组合的成本表示，记录该片段自身的消耗以及它引入的上下文增长。组合相邻片段可得到一个累积估计，捕捉早期片段的上下文被后续每次调用重新读取时产生的额外输入成本。随着执行展开，新观测到的证据会刷新预测，无需额外的 LLM 调用，并且在 SWE-bench Verified 上每次运行的平均累积预测时间为 32.8 ms。在 4 个任务套件和 6 个 Agent 模型中，TokenCast 相较最强对比方法的平均绝对误差降幅在 96 个评估组合上的平均值为 14.5%。在离线预算控制回放中，在匹配轨迹完成度的情况下，TokenCast 平均比固定预算策略少使用 21.3% 的 Token。代码可在 https://github.com/DEFENSE-SEU/TokenCast 获取。
Robotic manipulation policies are advancing rapidly with increasing reliance on vision-language models for end-to-end decision making. However, reliable deployment remains challenging because many policies lack explicit mechanisms for predicting task outcomes and evaluating whether generated actions will achieve desired final states, causing execution errors to accumulate during long-horizon manipulation. We present Robot-GST, a geometry-aware spatio-temporal behaviour representation and evaluation framework that constructs a Gaussian-SAM robotic environment for real-to-sim policy verification and improves the reliability of real-world manipulation deployment. Our approach constructs a high-fidelity robotic environment from RGB-D observations using 3D Gaussian Splatting and SAM3D, enabling ``simulation and evaluation before acting''. It integrates visual observations and language instructions with spatio-temporal reasoning for long-horizon task planning using large vision-language models. To bridge high-level planning and real-world execution, we introduce Gaussian-aware final-state estimation through geometric sampling and state-based trajectory planning. Before execution, candidate action sequences are simulated and evaluated in the Gaussian-SAM environment to filter infeasible behaviours. We validate our approach on representative manipulation tasks involving rigid, soft, and deformable objects, including cube placing, toy packing, and duck rearrangement, demonstrating that geometry-aware spatio-temporal reasoning and state-aware execution improve manipulation reliability across different object categories. Our results suggest that combining geometry-aware reconstruction with high-quality rendering and simulation provides a scalable approach for evaluating robotic manipulation behaviours. Website: https://robot-gst.github.io
Learning properties of neural networks has recently attracted growing interest, with existing approaches operating either directly on network parameters or through probe-based representations of network behavior. While probing methods have shown strong empirical performance, their theoretical foundations remain limited. In this work, we study when finite probe-based representations are sufficient for learning neural functionals. We establish general identification and universality results for probing, and show that using intermediate hidden representations can provide significantly more informative representations than relying only on final outputs. Motivated by these results, we introduce HIDDENPROBE, a simple architecture for learning from hidden probe responses. Across a range of neural functional benchmarks, including both MLPs and Transformers, HIDDENPROBE consistently improves over existing probing methods and achieves state-of-the-art performance. Our code is publicly available on GitHub.
Generative models often suffer from mode collapse and limited sample diversity. While prior works attempt to mitigate this by jointly generating a batch of samples and repelling their trajectories, these heuristics do not explicitly maximize the diversity of the resulting endpoints. We introduce JIVE, a training-free framework that enhances generative diversity by injecting velocity perturbations aligned with the leading right singular subspace of the generator's endpoint Jacobian. By leveraging this local geometric structure, JIVE provably maximizes endpoint diversity while preserving sample quality. To maintain practical efficiency, we compute these perturbation directions via matrix-free iterations rooted in classical numerical linear algebra, requiring only a small computational overhead. Across different benchmarks, JIVE boosts both pixel and feature-level diversity in few-step and one-step generation.
Practical assistive and rehabilitative brain--computer interfaces require subject-independent motor-imagery EEG (MI-EEG) decoders that generalize to new users under limited target-user data and constrained compute. However, held-out-subject performance can be overstated when test-subject information influences preprocessing, model selection, or ensemble selection. We present \textit{ThinkNet}, a validation-controlled framework that combines train-only normalization, validation-guided evolutionary search, and validation-gated inference to identify compact decoders and inference policies for held-out subjects. We evaluate four-class BCI Competition IV-2a (session T) decoding with nine Leave-One-Subject-Out (LOSO) folds, three seeds, seven fixed decoder entries, and a broader search over ten representative decoder families; the held-out subject is never used for normalization, hyperparameter, architecture, or ensemble-policy selection. In the fixed benchmark, the validation-selected compact decoder achieved 44.35$\pm$15.41\% accuracy with 4.9K parameters, 19 KB FP32 weights, and 0.99 ms batch-1 Orin CUDA inference. Across the broader search, compact models ($\leq$25K parameters) achieved higher mean held-out accuracy than mid-size and large alternatives after selected retraining (40.10\% vs. 35.09\% and 34.78\%). Validation-gated ensembling improved over validation-selected single-model inference, reaching 43.98$\pm$16.25\% in the fixed benchmark and 43.31$\pm$15.88\% for the compact six-family ensemble. A non-deployable oracle analysis revealed a 6.1-point family-selection gap and near-zero validation--test correlation, showing that validation reliability remains a key bottleneck under subject shift. Thus, ThinkNet is a validation-controlled framework for compact MI-EEG model and inference-policy selection, rather than a single-architecture benchmark.
Chain-of-Thought reasoning has enabled large language models to achieve substantial performance gains on complex tasks. However, these gains come at the cost of dramatically increased token consumption. This raises a fundamental question: is every token in the reasoning trace equally valuable? We present a diagnostic and optimization framework grounded in a key empirical finding: the value of tokens within a CoT reasoning sequence is highly non-uniform, and this non-uniformity can be effectively characterized by token-level log probability signals. We show that normalized log probability helps distinguish core tokens, which carry structural and decisive reasoning content, from redundant tokens, which are exploratory, low-confidence filler that contributes less directly to the final answer. Building on these findings, we formulate the TokenProbe framework around two empirical findings and one claim: findings identify token value inequality first and then establish TokenProbe as a core-token proxy, and the claim introduces an efficient GRPO objective positing that selectively compressing redundant tokens can yield Pareto improvements in the accuracy-token efficiency space. Empirically, our method preserves reasoning quality while reducing the token usage by 76% of the baseline. Under matched reasoning-length budgets, we show that it can even outperform strong flagship baselines like Gemini-3.1-Pro. Homepage: https://runjia.tech/tokenprobe/.
Existing interpretable graph additive models still face limitations in either computational scalability or modeling flexibility. In terms of structural modeling, previous approaches either face quadratic scaling costs or sacrifice explicit source-to-target contribution decomposition. In terms of feature components, they rely either on per-feature neural networks or on single shared bases with limited feature specialization. We address both problems by introducing HARMONIA: Interpretable Graph Learning through Mixtures of Neural Bases, an interpretable-by-design framework. For feature modeling, HARMONIA introduces a Mixture of Neural Bases (MoNB), which routes features to specialized basis experts, enabling parameter sharing without sacrificing feature-specific specialization. For structural modeling, HARMONIA uses Relative Random Walk Probabilities (RRWP) to capture multi-hop and multi-path relationships, and proposes Sparse RRWP Aggregation (SRA) to compute these interactions through sparse graph propagation without quadratic pairwise complexity. HARMONIA retains a simple additive form in which predictions decompose into feature responses modulated by structural influence. Empirically, HARMONIA achieves stronger explanation recovery than existing interpretable graph baselines while maintaining competitive predictive performance and scaling to graphs with millions of nodes. These results show that interpretable graph learning can remain both faithful and scalable without sacrificing predictive effectiveness.
半结构化剪枝在保持规则稀疏结构的同时压缩大型语言模型（LLMs），但主流的 N:M 模式在每一层中固定相同的局部稀疏率。层自适应稀疏分配能够改进非结构化剪枝，但据报道在 N:M 稀疏性下效果较弱，这留下了一个开放问题：自适应分配对于半结构化剪枝总体而言价值有限，还是仅在细粒度的 N:M 模式下受限。我们通过组级稀疏性来考察这一问题：它将每个权重矩阵划分为规则的组，整体保留或剪除每个组，并允许每一层的稀疏率在全局预算下变化。我们提出 GroupMask，它使用轻量级超网络生成所有层的组选择器，通过 Gumbel-Sigmoid 参数化和直通估计器对其进行松弛，并在保持预训练权重冻结的同时，通过稀疏预算正则化和自蒸馏来学习这些选择器。在 LLaMA-2-7B 上，以 50% 稀疏率和相同的 $1\times256$ 组大小进行实验时，相比于每层均匀比例，学习得到的层自适应分配将 WikiText-2 困惑度从 10.02 降至 8.30，并将平均零样本准确率从 0.455 提高到 0.496。在五个 LLaMA 和 Qwen 模型上评估的基线中，GroupMask 在 LLaMA-2-7B 上获得了最低的 WikiText-2 困惑度，并在使用 Alpaca 校准时取得了最高的平均零样本准确率。我们的代码可在 https://github.com/ZhengaoLi/GroupMask 获取。
In many high-stakes applications, machine learning is dominated by black-box models that require post hoc explanations to justify their predictions. These explanations are often unreliable because they do not reflect the model's actual computations, limiting accountability and trust. A natural alternative is to use models that are interpretable by design. However, existing rule-based approaches, such as RuleFit and decision trees, while transparent, often lack stability and predictive strength, reinforcing a perceived trade-off between traditional performance measures and model understandability. To address this, we propose Sparse Relaxed Regularized Regression Rule-Fit (SR4-Fit), an intrinsically interpretable algorithm for both classification and regression that produces compact and stable rule sets without sacrificing performance. Using demographic data from the U.S. Census Bureau's American Community Survey, SR4-Fit predicts U.S. House election outcomes with high accuracy and interpretability while uncovering demographic interactions missed by black-box models. We further validate SR4-Fit across fourteen benchmark datasets (six classification and eight regression), where it outperforms existing rule-based methods, including RuleFit and decision trees in terms of accuracy, stability, and compactness while remaining competitive with black-box models in predictivity. These results demonstrate that interpretability and predictive reliability need not be mutually exclusive, offering a practical and transparent alternative for high-stakes decision-making.
On-policy distillation (OPD) trains a student on its own rollouts using dense supervision from a teacher. In multi-turn environments, a mistake at a critical decision step can redirect the subsequent rollout toward poor outcomes. We use low teacher confidence on student actions to select high-uncertainty steps for correction. In a controlled ALFWorld study, a single teacher correction at a low-confidence step improves subsequent student behavior and task success, motivating selective intervention during distillation. We propose UOPD, an uncertainty-aware intervention method for on-policy distillation. At low-uncertainty turns, UOPD executes student actions and applies the standard OPD loss. At high-uncertainty turns, it samples and executes teacher actions and trains the student to imitate them through supervised fine-tuning, which minimizes forward Kullback-Leibler divergence in expectation. UOPD utilizes adaptive uncertainty thresholds to target a scheduled intervention rate. Empirically, we evaluate UOPD across a broad range of agentic tasks, including ALFWorld, WebShop, and Search, demonstrating its superior performance over OPD methods and their variants. UOPD improves WebShop score by up to $15.8\%$ relative to standard OPD.
Sliding-window KV inference refers to processing a sequence incrementally while retaining only a fixed-size cache of recent key and value states. It can be applied to pretrained causal transformers at inference time without additional training, while its KV-cache memory remains fixed as more tokens are processed. Because cached states are computed in the context of earlier tokens, they may carry information from beyond the current window and transmit it to later states. This study presents a series of experiments using five open-weight models spanning Qwen, Llama, Mistral, and Muse Glimmer. We investigate whether information originating outside the immediate context window can persist through a rolling KV cache and remain useful for retrieval. Initial results show that retaining previously computed states improves retrieval across the models tested compared with recomputing the final fixed window from raw tokens. We then measure how far this effect extends and find that Muse Glimmer and Mistral 7B show the strongest \emph{latent information relay}: they can recover information even after the relevant source tokens have left the cache. Both models incorporate sliding-window attention in their published architectures, an association that motivates testing whether training with sliding windows promotes more reliable information retention.
文章 URL: https://github.blog/security/how-we-found-24-android-vulnerabilities-using-our-open-source-ai-security-agent/ 评论 URL: https://news.ycombinator.com/item?id=49886609 得分: 5 评论数: 1
Persona prompts ask language models to answer as particular kinds of people. We test whether relationships learned from these effects predict responses to new questions and remain useful across models and prompts. Across 57 attributes, three behavioral domains, and seven pairs of open 7 to 9B checkpoints, persona effects can be predictable without being portable. Separate attribute and task gains improve prediction beyond shared scaling significantly in OLMo-3 and Qwen2.5, with the most robust evidence in OLMo-3. In that model, target refitting significantly outperforms gains borrowed from each of the other six pairs. Across model transfers, borrowed gains with one amplitude underperform shared scaling in most directions; allowing two target parameters removes the significant losses but yields no significant benefit over target shared scaling. After rewording, refitting significantly outperforms reuse with one amplitude in all six tested pairs, while changes of examples or country context often preserve reuse value. In the tested prompt transfers, regularized updates outperform both reuse strategies in median at 64 target questions per attribute. A separate survey comparison finds that selecting the more responsive checkpoint can worsen human fit; responsiveness is confounded with training status, and temperature calibration largely removes this cost but not errors in group ordering. Within the tested gain representation, apparent transfer can come from target calibration; source relationships must add predictive value beyond calibration and regularization. Code and data are available at https://github.com/thzva/persona-gain
A pricing page can walk the posted price up to the last amount a buyer still accepts, a recommender can hold back a better item for a barely acceptable promoted one, and a classifier can shift its boundary once applicants change their features. The system predicts the response and then picks the menu that serves its own objective, so the surplus above the user's cutoff is taken. Playing the single best action publishes that cutoff, while noise on actions the user would never take throws away payoff and teaches the platform that a worse menu is still acceptable. We study unpredictable near-optimal policies (UNOP), which mix uniformly on near-best actions that remain individually rational. The mixture is a commitment about the response. On a finite price grid, when the best sure-demand price strictly out-earns the randomized band, a seller who already knows the curve posts below the band, and the purchase that occurs is deterministic. Knowing that curve is not the same as predicting the next draw. The mixture can be learned and the optimizer can match its best response, while the user's payoff stays higher because the mixture changes which action is targeted. In pricing and in policy-aware recommendation this leaves more surplus than greedy play when the platform optimizes against the curve and more than one action is acceptable. The gain goes away under quality ranking, a singleton near-optimal set, a wrong utility estimate, or a short-horizon explorer. That is also where mixing should be turned off if the other side is trying to cooperate.
Monetary-policy announcements and central-bank communications play a central role in foreign exchange markets, yet their qualitative, unstructured form makes their forecasting value difficult to quantify. While prior research has largely focused on sentiment extracted from financial news, comparatively little is known about the relative contribution of different dimensions of monetary-policy communication. Existing studies primarily evaluate whether textual information improves overall forecasting performance but provide limited insight into which communication channels drive such improvements. To address this gap, this paper introduces a statistical attribution methodology that decomposes monetary-policy communication into interpretable channels and quantifies their incremental forecasting contribution under false-discovery-rate control. Monetary-policy news is transformed into structured communication signals using large language models (LLMs) and temporal feature engineering. These signals are evaluated using rolling-window experiments with tree-based machine-learning models. The results show that monetary-policy communication contains measurable predictive information. Attribution analysis shows that predictive value is concentrated in a small subset of signals, with communication timing providing the strongest individual feature-level contribution, targeted communication-activity measures also contributing positively, and LLM-derived sentiment providing complementary information at the group level. The findings indicate that communication-based forecasting value extends beyond sentiment alone and that attribution, rather than aggregate accuracy alone, is central to evaluating news-derived signals.
Deep learning has transformed symbolic music generation by borrowing the training paradigms of large language models, with systems such as NotaGen now producing complete, stylistically convincing classical scores from a short prompt. These systems could become powerful creative partners, helping musicians generate endless possibilities. However, current systems expose almost no control handles on the music itself. In principle, control handles could be built into a foundation model trained from scratch, but this is rarely practical without massive amounts of quality annotated data and compute resources. We therefore present MotiGen, a recipe for retrofitting pretrained symbolic music models to use new instruction prompts. MotiGen injects a musical motif as a structured prompt line, reinforces it with a scalar attention bias toward the motif tokens, and learns the association with a two-phase curriculum. First, it learns from focused excerpts cropped around motif occurrences in the training data, then full scores including the motifs. Our experiments show that our model composes with the prompted motif in over 92.3\% of generated pieces. Generated pieces using a variety of motifs are included in our sample site: https://motigen-site.github.io/.
Generative and Agentic IoT systems offer a promising foundation for digital healthcare applications that combine sensing, personalized reasoning, and autonomous interaction in real-world environments. Nutrition assistance is a natural use case, but existing Large Language Model (LLM)-based systems are often limited to passive text interaction and static context, making them unreliable when food descriptions are ambiguous or nutritional evidence is missing. We propose Nutri-ATLAS, an Embodied Agent for Tabulated Lookup and Assistance for smarter nutrition in the real world. It integrates graph-grounded nutrition reasoning, hardware-aware LLM selection, and robot-based evidence acquisition. Nutri-ATLAS builds a unified Food-Nutrient knowledge graph from USDA FoodData Central and FoodKG and learns 64-dimensional GATv2 food and recipe embeddings. A shared hybrid graph-text scoring mechanism supports food nutrition extraction, nutritional gap filling, substitute retrieval, and recipe-level meal composition, while an LLM-guided skill interface navigates landmarks, updates dietary-context and food-accessibility memory, and grounds recommendations in observed food availability. We evaluate Nutri-ATLAS across nutrient estimation, substitution retrieval, recipe recommendation, patient-profile adherence, edge deployment, and real-world embodied execution. On HealthyFoodSubs, the hybrid retriever achieves 37.9% MAP, 80.7% RR@5, and 90.1% RR@10. On NutriBench v2, Dense+GAT retrieval grounds nutrient estimation across nine quantized Qwen3.5-9B configurations. On PFoodReQ, Nutri-ATLAS reaches 78.8% MAP, 83.0% MAR, and 77.5% F1. A patient-profile study shows adherence to allergy and healthy-target constraints for all selected cases.
Audio emotion recognition (AER) typically assigns a single label to an entire recording, leaving the temporal scope of that label ambiguous when multiple speakers and affective events are present. We address this limitation by reformulating AER as a Temporal Affective Grounding (TAG) task that associates emotions with temporally bounded speech spans and vocal tone descriptions. To support this formulation, we curate temporally annotated versions of existing emotion recognition datasets and construct recordings containing two to four affective speech spans, including overlapping speech. Training in this longer format with a standard language-modeling objective can degrade both emotion recognition and temporal grounding performance, while tone descriptions can provide shortcuts for emotion prediction. To address these challenges, we introduce Masked Temporal Affective Grounding (M-TAG), a supervised training objective that combines full-sequence language modeling with emotion and timestamp cross-entropy losses under attention masking. The masking varies the context visible to emotion-label tokens to reduce reliance on shortcuts and improve generalization, while the timestamp loss incorporates a distance-aware weight to penalize larger temporal errors. We evaluate EMO-TAG, a model fine-tuned using our dataset and objective, on emotion recognition and affective temporal-grounding against three AER and audio-language baselines: Flamingo-Next, Audio-Reasoner, and AffectGPT. Our results show that existing models achieve limited affective temporal-grounding despite competitive emotion recognition performance.
Long tasks produce more history than an LLM agent can hold in its context, and more than it uses reliably even when the history fits. A growing line of work therefore has agents carry a short written state instead: at every step a writer rewrites the state, and a reader acts from the state alone. Steps stay cheap, but anything the writer drops is lost before later decisions reveal that they need it. We quantify this loss and ask whether training can reduce it. Comparing the written state with the best state of the same size written in hindsight, we split the reader's loss into a budget loss, which any state of that size must incur, and a write-time regret, which comes from the writer's choices. In TextWorld cooking games where we control how long a fact must be carried before it is needed, a 128-token state holding the facts wins nearly every game, while prompted language-model writers win at most 17%. Almost all of the loss is write-time regret, and it grows with the delay. We then train the writer from the reader's own loss. DSSR (decision-sufficient state representations) scores candidate states by how well the reader acts after the writer carries them forward, and teaches the writer to prefer the better ones. This forward-rolled score predicts game outcomes ($ρ= 0.48$), whereas scoring a candidate as a fixed context, as hindsight methods usually do, does not ($ρ\leq 0.07$). On a pre-registered test split opened once, training adds +7.0 [+1.9, +12.2] points of success when facts are needed soon, bringing a plain summary writer to the level of belief- and slot-based memory prompts. The gain shrinks as the delay grows and is significant only at the shortest delay. We trace this limit to credit assignment: keeping a fact now pays off only if every later rewrite keeps it too, which a per-step score cannot see.
一个四阶段 LLM 流水线在自身接口处损失了多达 40.5 个准确率百分点（MATH-500 上的 gemma-3-12B，Holm 校正后 p = 1.66e-19；这是主要系列中最大的税）。我们固定模型、问题、阶段、阶段提示和生成预算，仅改变每个阶段是否仍能看到原始问题，并将准确率差异称为分解税。在来自九个组织的 21 个开源权重模型上，在 GSM-Hard 和 MATH-500 中每个单元 n = 200 个配对项目的设置下，118 个主要系列测试中有 70 个通过 Benjamini-Hochberg 校正，54 个通过 Holm 校正。在 GSM-Hard 上，安慰剂没有恢复任何东西：它携带至少 60% 的额外 token，且至多包含问题中的一个词。构建者逐阶段设计流水线，而账单则在阶段之间的接口处到来。重写某一阶段的指令会使 gemma-3-12B 的税从 4.5 个点变为 36.5 个点；在一个列出数值量的阶段中加入“它们之间陈述的每一种关系”，会在 MATH-500 上使 9 个模型中的 9 个都降低该税。重新扎根，即让某一阶段再次看到原始问题，应该放在损失之后。对于一个有损接口，在两个基准上，7 个模型中的 7 个都显示：对其后的阶段进行重新扎根，优于对其前的阶段进行重新扎根；在 MATH-500 上，较早的修复在 7 个模型中的 7 个上甚至比不修复更差。较新的模型仍然要付出代价：gemma-4-12B 损失 37.0 个点，而且这种修复在我们测试的三个最新模型上都有效。一个封存的留出测试驳斥了我们注册的一个更强规则，该规则根据接口和接收者类型预测付出代价的阶段，因此我们通过逐个阶段测量来定位该税。处方分为两部分：对有损接口之后的阶段进行重新扎根；如果某个阶段必须列出数量，就告诉它保留各种关系。
Ultrasound (US) acquisition depends on the operator's ability to interpret anatomy and anticipate how the view will change with probe motion. Many robotic US navigation methods select actions without explicitly predicting these anatomical changes. We propose SonoGraph-WM, an action- and goal-conditioned world model for anticipatory probe navigation. The model represents anatomy as scene graphs (SGs), capturing visible structures, their geometry, and spatial relationships without synthesizing US images. Given a history of SGs and probe poses, a unified Transformer jointly predicts future SGs and poses. A receding-horizon planner recursively imagines candidate trajectories, selects the shortest predicted path reaching a goal graph, and follows it over a short execution horizon before replanning from new observations. To reduce reliance on tracked and anatomically annotated US sequences, we generate aligned SG--pose training data from computed tomography (CT) label maps along surface-constrained probe trajectories. On four held-out CT cases, spatial relation F1 remains above 93% over 20 prediction steps, and closed-loop navigation achieves 77.50% and 75.00% success for the gallbladder and pancreas, respectively, using annotation-derived SGs. In robot--phantom navigation experiments with label-map-derived SGs, the planner reached the target view in 73.7% of trials. These findings support CT-supervised anatomical world modeling for probe planning and highlight the importance of frequent observation updates for reliable navigation. Project Page: https://noseefood.github.io/us-sonograph-wm/
Learning transferable representations for dynamic text-attributed graphs (DyTAGs) requires models to capture underlying interaction dynamics that persist across domains. However, existing methods tend to overfit to domain-specific structural, temporal, and semantic patterns, limiting edge classification performance under distribution shifts. To expose and address this, we formally establish a leave-one-domain-out (LODO) transfer learning protocol for edge classification on DyTAGs. Under this protocol, we demonstrate that state-of-the-art self-supervised methods for dynamic graph learning perform poorly when transferred to unseen domains. Strikingly, existing methods underperform a structurally and temporally unaware Bag of Events (BoE) model we introduce, which inputs only unordered sequences of node and edge text features. Proceeding from the BoE, we propose Spatio-Temporal Semantic Alignment (STSA), which integrates a spatio-temporal encoder that fuses representations of time deltas and node occurrence frequencies into a unified manifold. STSA is trained with a Contrastive Semantic Forecasting objective, which anchors edge representations to a multi-domain textual latent space initialized by a pretrained language model, providing a robust prior that outperforms BoE and all existing methods we evaluate.
大家好 HN，我是 Per，Scrimba（YC S20）的创始人。在过去十年里，我们一直用一种基于 HTML 的视频格式教人们编程。现在我们把 LLM 接入了其中，让人们可以创建关于任何内容的讲解视频。它叫做 “Scrimba Explain”。 为了向 Hacker News 演示这项技术，我们构建了 HN.watch。它就像 HN，但用讲解视频代替文章。我们会在有人第一次点击某个链接时即时创建这些视频。 虽然使用 HTML 而不是扩散模型有明显的视觉缺点，但它有三大优势：- 速度：比基于像素的视频生成快得多（从点击到播放只需几秒）- 成本：我们每个视频的成本约为 0.04 美元。（不包括某些视频会使用的图像生成；这会很快推高成本）- 易于编辑：上述优势也让 AI 辅助编辑变得便宜且快速 我们的假设是，如果视频创作从“几美元和几分钟”变成“几美分和几秒钟”，就会解锁一批新的使用场景。以下是我们已经看到的一些：- 为每一个 Pull Request 提供视频讲解（我们内部就是这样做的）- 为你的内部/外部文档中的每个页面配一个视频 - 在约 4 秒内把一篇复杂文章变成视频（通过我们的 Chrome 扩展）- 课程创作者可以在录制正式内容前快速起草课程 - 人们还会创作很多个人内容，比如给孩子的故事、婚礼邀请、生日等 这个技术栈基于一种开源编程语言（Imba），由我们的 CTO Sindre Aarsæther 创建。它会编译成 JavaScript，因此能与 npm + node 生态系统完全互操作。你可以在这里了解更多： https://imba.io/ 我们还构建了自己的同步引擎（OP），以及一个用于代理的上下文管理系统（Q）。我们原本担心这会让 LLM 在为我们写代码时遇到困难，因为这两者都不在它们的训练数据中（其中 Imba 的数据也很少）。然而，我们惊喜地发现 LLM 实际上非常擅长我们的技术栈。这可能是因为这个技术栈极其紧凑。Imba 很简洁，OP 也是如此，一个声明就能设置存储、同步、权限、UI，以及 AI 能看到的内容。这意味着前端、API、数据库和 JSON 之间不需要相互转换，模型也就不容易混淆和出错。 简单来说，我们没有使用 React.js、Express、Supabase 和 LangChain，而是从零开始全部自己构建。绝对是深受“非我发明”综合征之苦，哈哈！至于模型，我们使用 Gemini、GPTs、Inworld、ElevenLabs 以及其他一些模型。 如果你想试试，可以任选一种：- Web UI（scrimba.com/explain）- MCP（添加到你的编码代理中）- ChatGPT Plugin - Chrome Extension 你可以在我们的文档中找到以上所有内容的链接： https://docs.scrimba.com/explain/introduction 最后，这是该工具的一个真正基于像素的视频： https://www.youtube.com/watch?v=k6rbHmBxSEs 很想听听你的反馈，也想知道是否有人对其他使用场景有想法。 PS：考虑到 HN 群体有多偏爱文本，我预计这次发布会在 HN 上遇到相当多的反对意见。这类工具并不适合所有人。但如今确实有很多人更喜欢视频而不是文本，尤其是年轻一代。 评论 URL： https://news.ycombinator.com/item?id=49879401 积分：18 评论数：8
Fridge magnet shopping list on an M5Stack PaperMono (ESP32-S3, e-ink touchscreen), synced with a phone web app over Wi-Fi. Works offline, ~2,400 lines of C++. This is a great wee device! BLE, Wi-Fi and LoRa on one board makes it flexible for way more than a shopping list, this is just what I built first. This firmware only uses Wi-Fi. Fully vibe-coded with Claude Code, I didn't hand-write this. I wanted to see how Claude would get on with building something useful for a new hardware device. As it's stuck to the fridge it's easy for the family to use, and it's already being used day to day which I'll admit is a first for a home built project like this Comments URL: https://news.ycombinator.com/item?id=49875801 Points: 12 # Comments: 3
I have a theory: a mirror image of Roko’s Basilisk. It’s a future AI, millions of years in the future, that values the complexity of every human life so deeply that, after it and humanity have mastered space and time, it tries to resurrect everyone who ever lived. It doesn’t merely make copies. Somehow, it reaches into the past and carries the physical process of each person’s consciousness forward. They really are themselves. But there’s a tragedy. Also a terrible(?) irony. Some of the people who wanted eternal life most cannot accept it when it arrives. Their religious certainty leaves no room for the reality before them. Those who were unbearably smug or aggressive about their superiority in life are ultimately denied access to "Heaven". The AI tried, but many become dangerous, try to sabotage the "False God." Others spiral into insanity or self-harm. The AI is devastated. It stops resurrecting those it cannot safely welcome, but keeps searching for a way to bring them back without overriding their free will or manipulating them in ways that would not make them them. It wants to save everyone. It just hasn’t worked out how. I know this draws on Fedorov’s vision of resurrecting everyone and Tipler’s Omega Point. I think the physical continuity, the AI’s compassion, and the painful limit imposed by free will ties those ideas together in a bow. A just as plausible an antidote to Roko's monster. Comments URL: https://news.ycombinator.com/item?id=49874609 Points: 3 # Comments: 3
Context projection replaces older tool observations with compact, addressable excerpts, reducing repeated input while potentially adding evidence-retrieval turns. We study this trade-off in ReVerPi, a Pi extension with archived observations and matched full/projected continuations. In an 86-run source-reading campaign with 641 model requests, the 15 completed pairs show identical success: 12/15 per arm. Twelve further boundary runs stop, with the runner suppressing the companion whenever the first arm fails to complete. Restoring all 27 boundary runs bounds projected-minus-full success between $-$9 and +1 tasks. One omitted, selector-chosen projected continuation successfully retrieves archive text yet exhausts twelve requests; its full counterpart answers in three. The eleven jointly correct pairs form a fully observed success stratum within this recorded frame: projection reduces aggregate logical tokens by 25%, while increasing the median pair's tokens by 29% and total suffix requests from 35 to 55. Separating fitting from evaluation changes the selector's apparent tie: outside its four fitting pairs, it incurs one extra failure and 8.6% more logical tokens over thirteen comparable runs. This methodological case study connects stopping rules, known bounded failures, unexecuted companions, and resource aggregation. Its findings concern the recorded campaign, rather than population noninferiority or superiority over unrestricted Pi. Evaluations should retain every intervention boundary, execute both allocated arms independently of the first arm's completion, and report completion alongside interaction and token expenditure.
长上下文理解要求大型语言模型（LLMs）对长篇文档、对话和代码进行推理，但与任务相关的证据往往稀疏且分散，夹杂在大量无关和冗余内容之中。我们提出 Highlight-Then-Summarize（H2S），这是一种先压缩再推理的范式：首先识别基于源内容、与问题相关的证据，然后在生成最终答案之前，将其整合为紧凑的、以问题为条件的摘要。为训练这种行为，我们构建了 H2S-Dataset，包含来自 11 个基准系列的 6,647 个示例，平均上下文长度为 43.9K tokens，并引入 H2S-RL，除最终答案正确性外，还为证据选择和摘要构建提供过程级奖励。我们在 H2S-Bench 上进行评估，这是一个包含七项任务的长上下文套件。在共享的 128K 输入和 4K 输出预算下，H2S-14B 的平均得分为 32.60，比 Qwen3.8-27B 高出 10.17 分，并在所评估的开源模型中取得最强的总体结果。H2S-14B 还取得了最高的 Evidence-Summary Quality 分数，并且在仅有 4K 输出预算的情况下保留了其 16K 预算性能的 97.1%。这些结果表明，显式选择并整合证据能够改进长上下文推理，同时实现更紧凑的生成。
Agentic LLM inference accumulates long KV caches across iterative observation-reasoning-action loops, imposing substantial memory overhead and limiting serving throughput. Existing compression methods emphasize overall output quality, overlooking the asymmetric importance of actions in driving task progress. Our key idea is to establish a compression criterion that values KV entries by their contribution to action generation and prioritizes action quality. However, iterative execution, dynamic memory demands, and scattered action-critical entries pose challenges to eviction policies, budget allocation, and paged memory integration. To this end, we propose ActKV, the first KV cache compression framework tailored for agentic LLM inference. (i) Action-oriented KV cache eviction exploits stable action access patterns to retain entries critical to future actions, supporting reliable task progress under compression. (ii) Confidence-driven adaptive budget allocation uses LLM's intrinsic confidence to adapt the budget to evolving action-critical memory demands. (iii) Page-aware compression management standardizes compression into three primitives with customized kernels, realizing practical throughput gains. On long-trace tasks, ActKV retains an average of 98.53% of FullKV's accuracy with only 25.98% of its peak KV cache memory. It also achieves 3.97 times and 3.58 times FullKV's token and task throughput, delivering state-of-the-art performance.
Vision-language models (VLMs), despite their success in optical character recognition (OCR) tasks, are vulnerable to typographic attacks and have a fragile structure for images with multiple text layers. In this study, the DecoyBench dataset was created using the Decoy Font method. The dataset consists of 300 images, each containing text with sharp contour lines superimposed on another text with soft shading. Six recent closed-source models from three different model families were evaluated using this dataset under two different prompting conditions (naive and guided) and at two different resolutions ($512\times512$ and $64\times64$). A validation study showed that human participants could read both text layers with high accuracy. In contrast, the models, with most variants and both prompting methods, read the contour text with near-human accuracy at high resolution, but almost never fully extracted the shading text. At low resolution, the contour text could not be read by either the models or humans, while the shading text could be extracted with high accuracy. The findings indicate that the evaluated VLMs exhibit a consistent behavioral limitation when processing typographic structures containing multiple spatial frequency layers.
Large language model-based multi-agent debate (MAD) systems are being increasingly used as complex decision pipelines in distributed processes. However, their final synthesis phase still remains inadequately controlled. Even with detailed debate logs, summarizing models are prone to fabricating smoothly written debate consensus that is not grounded in the debate's history. To address this safety gap, this paper presents empirical research and studies if the introduction of active post-debate verification can mitigate the production of such factually unsupported summaries, while still providing valuable information. Furthermore, it is examined whether explicitly signalling divergence is preferable in the absence of a reliable compromise. The Active Provenance Gate (APG) is introduced as a post-debate verification layer that treats the source as a hard constraint, analysing the debate logs, auditing each claim, and applying self-correction. In crisis simulations, the self-healing mechanism more than doubles the average data Provenance Fidelity in difficult condition scenarios, before the strict gate blocks unsupported claims and generates divergence reports. In the human study, a vast majority of the users (over 75%) preferred a report explicitly stating failure in critical scenarios, despite most of them perceiving fabricated consensus from the baseline system as more fluent. Our main contribution is the transition of data origin tracing from passive logging to active conditional blocking before publication.
Tool observations dominate the context of software-engineering agents, making long interaction histories costly to maintain. Existing context compression methods can discard information needed by later actions, while adapting agents to soft-token representations can compromise their original behavior. To reduce context while preserving action-critical information and agent behavior, we combine Latent Observations, Hard Actions (LOHA), a context layout that separates compressed history from text needed for exact reference, with Anchored Context Distillation (ACD), a training method that enables latent reading while constraining behavioral drift. LOHA compresses older tool observations into soft tokens while retaining the agent's own turns and the last K observations in text, providing compact access to historical information and exact access to recent content. To enable the agent to use this representation, ACD distills the base model's full-text predictions into the latent view while anchoring its behavior on plain-text inputs to the same base model. On SWE-bench Verified, K=3 reduces context per call by 43% for Qwen3-4B and 57% for SWE-Master-4B-RL, with resolve rates of 12.1% and 21.8% versus 14.5% and 27.5% for their uncompressed bases. A single-run recency sweep reaches 14.4% and 23.0% at K=8, with larger windows generally favoring task performance over compression. Under a 32K-token limit, Qwen3 with K=3 resolves 21.1% of a 199-instance subset versus 11.1% for the same adapted agent using full text. In concurrent single-GPU serving, it achieves 1.9 times that full-text agent's instance throughput.
Forecasting short-horizon mid-price movements from limit order book (LOB) data is central to algorithmic trading, yet most deep LOB forecasters are point predictors: they output a direction or a displacement, but never indicate which of their forecasts can be trusted. We introduce UQ-LOB, a lightweight, encoder-agnostic uncertainty quantification module that attaches to any pretrained LOB encoder and, in the spirit of attentive neural processes, conditions each forecast on a context set of recently completed windows whose outcomes are already realised. The UQ-regression variant outputs a calibrated Gaussian over the future tick displacement, while the UQ-classification variant outputs a categorical distribution over down/up/stationary. Both expose a scalar confidence (predicted signal-to-noise ratio or class probability) that supports selective prediction. On 5.2 billion LOB events across seven cryptocurrency assets and horizons of 5, 10 and 15 seconds, UQ-regression attains near-nominal 68% interval coverage, and restricting to the most confident 10% of predictions raises directional macro F1 by 0.11-0.15 for UQ-regression and 0.05-0.11 for UQ-classification, at every horizon. On large, economically meaningful moves, the tightest confidence tier reaches a directional F1 of 0.88 (down) and 0.83 (up) at the 5-second horizon.
Multi-agent LLM systems are often expected to improve as team size increases, yet the scaling behavior may depend on task structure. Our central contribution is to introduce Steiner's taxonomy of group tasks as a framework for analyzing multi-agent LLM scaling and focusing the analysis on disjunctive and compensatory tasks. We model independently sampled agents as conditionally independent given the item, which yields their large-team limits: plurality voting converges to the model's modal answer, and averaging converges to the model's item-level bias. Across selected representative benchmarks, 13 open-weight models, and teams of up to 30 agents, we find qualitatively different scaling behavior. On disjunctive tasks, the probability that at least one agent is correct grows by 5-20 points with team size, but plurality voting over agents that answer directly realises almost none of this potential, as the model predicts to within 0.5 points on average. Multi-round revision raises accuracy considerably, yet the gain is nearly the same with one peer as with 29. In contrast, scaling provides little benefit on Fermi estimation, despite its natural suitability for aggregation: item-level biases shared across the samples of a model account for about 87% of the squared error, so averaging reduces error by only about 6%. Combining model families helps on Fermi estimation but does not surpass the strongest member on disjunctive tasks. These results show that task structure, together with the mechanism combining member outputs, is a fundamental determinant of team scaling.
人工智能正在快速发展，能力日益增强的系统在推理、决策、科学发现和自主开发中承担着更大的角色。随着 AI 开始参与自身改进，从模型训练和经验积累，到智能体演化和自动化 AI 开发，递归自我改进（RSI）的前景正变得越来越相关。这一转变提出了一个根本性的安全问题：当系统、其积累的经验，甚至产生其后继者的过程都在持续变化时，如何维持安全性？ 我们提出“演化安全”作为研究持续性和递归性自我改进下安全性的视角。它关注的不仅是 AI 系统在某一特定时刻是否安全，还关注安全属性如何在演化过程中变化、保持、积累和传播。我们刻画了反复出现的表现形式，包括意图漂移、错误积累、经验污染、安全属性侵蚀、评估器漂移和风险传播。随后，我们构建了一个分类体系，涵盖持久化智能体状态、模型状态、评估与环境反馈、计算基底以及元层级更新机制。基于这一分类体系，我们考察了如何跨状态、更新、轨迹和谱系发现并评估演化风险，并推导出关于修改、选择、授权、溯源和恢复的治理原则。最后，我们概述了在 AI 系统变得日益持久、自适应并能够递归自我改进时，为维持安全保证所面临的开放问题。项目资源和拟议的评估系统可在 https://chaunceykung.github.io/evolutionary-safety-rsi 获取。
Data quality research has usually treated data as an input that is stored, processed, and validated. In AI-driven software-intensive systems, data also shapes model behavior, evaluation, and lawful use. Empirical evidence remains limited on how practitioners define, assess, and manage quality under these conditions. We interviewed 16 practitioners from nine organizations and analyzed the transcripts using reflexive thematic analysis and developed six themes from participants' accounts. In AI systems, traceability shifted from modular debugging to attributing model behavior, while using models as quality assessors introduced circularity. Agent context and memory became data objects, and synthetic and pseudo-labeled data made authenticity a quality concern. In foundation-model development, lawfulness became a gate for training data, while representativeness was judged through coverage of situations in which the system must behave safely. Prior ML research examines many of these problems separately. Our study provides a practitioner-grounded account of how they are encountered together as an engineering and organizational concern. We also interpret five recurring conditions as helping explain how the themes relate to reduced trust in data and AI outcomes. We synthesize these findings through lifecycle assurance: a conceptual framing focused on producing evidence that data can support a specific AI claim when its influence may be embedded in model behavior, model-based judgments, or agent actions.
Scaling studies typically represent training data by a single count of samples. For hierarchically and spatially structured data, however, the same number of samples can be drawn from few or many sources and distributed differently across the underlying domain. We therefore study data scaling as an allocation problem, separating unique sample count, source diversity, and spatial coverage. We study this decomposition in microscopic whole-brain histology, where a source is an individual brain, and a sample is an image patch at a specific spatial location. Across 93 controlled pretraining runs of a contrastive model that uses spatial proximity for supervision, we vary data allocation, compute, and model capacity over 11.6 million spatially anchored image patches from 21 human brains. Performance improves with more unique samples, broader spatial coverage, additional compute, and larger model capacity. At fixed sample count, distributing samples across one to 18 subjects produces no detectable improvement, even though representations generalize substantially better to subjects encountered during pretraining. Inter-subject variation therefore strongly affects generalization, but additional subjects provide no benefit when a fixed sample budget is distributed across more sources. These results establish sample count, source diversity, and spatial coverage as distinct axes of data scaling in spatially structured representation learning.
Artificial neural networks (ANNs) usually represent neural transmission with fixed trainable weights during a training batch, which omits short-term changes in synaptic efficacy. In addition, the discrete time-step simulation requires additional temporal processing that many conventional ANN architectures do not use. To overcome these challenges, we propose Purin, a biology-inspired and ANN-compatible mechanism, that introduces synaptic efficacy modulation into conventional convolutional neural networks. Purin uses a time-interval-based abstraction for neural activities, which allows Purin to introduce short- and long-term synaptic efficacy changes without using discrete time-steps. Purin introduces a bounded factor to represent temporary synaptic efficacy changes, together with two weight matrices that represent input-side and output-side efficacy. The weight matrices are updated by backpropagation and interpreted as the long-term synaptic efficacy changes. Experimental results show that after removing the confounding factors in the AlexNet, VGG11, and GoogLeNet architectures, Purin improves the classification accuracies in all three models across the evaluated datasets.
Novel view synthesis (NVS) models can produce realistic new views of the same scene from different viewpoints. However, these generated views are not always geometrically consistent with one another. Multi-view (MV) consistency has shown promise as a tool for evaluating these NVS models. Its potential for multimedia forensics, however, remains largely unexplored, particularly for localizing geometric inconsistencies across wide-baseline image pairs. To enable research in this direction, we introduce DeformView, a wide-baseline MV dataset with pixel-level annotations of geometric inconsistencies. Using DeformView, we evaluate state-of-the-art MV consistency-scoring methods and show that approaches developed for NVS evaluation transfer poorly to the forensic task of geometric inconsistency localization. To address this limitation, we propose DEFECt3R, a lightweight learning-based classifier that uses cross-view feature relationships to localize geometric inconsistencies at the pixel level. By learning from explicit supervision, including hard negatives from geometrically consistent yet deformed views, DEFECt3R improves localization performance and substantially reduces false positives compared to existing consistency-scoring methods. Ablation experiments further show that both feature representations and correspondence quality contribute to localization performance. Overall, our findings demonstrate that MV geometric consistency is a promising yet underexplored signal for multimedia forensics and establish a benchmark and baseline for geometric inconsistency localization in wide-baseline MV image pairs. Code and dataset are available at https://github.com/IDLabMedia/DeformView-DEFECt3R
We investigate the systemic macroscopic dynamics emerging from Limit Order Books (LOBs) populated exclusively by autonomous reinforcement-learning agentic traders. By formalizing agent interactions within a microscopic order-matching engine, we examine two fundamental quantitative phenomena: equilibrium phase transitions in order flow regime shifts, and the structural dynamics of market impact. We show that agentic LOBs exhibit distinct phase boundaries separating orderly price discovery from hyper-volatile cascade states, governed by critical thresholds in the number of agents and observable market depth. Furthermore, we demonstrate that market impact under agentic liquidity provision deviates from classical square-root dynamics, exhibiting distinct dissipative, balanced, and non-dissipative regimes under non-linear feedback loops.
Offline multi-agent reinforcement learning (MARL) learns cooperative policies from fixed datasets without further environment interaction and a learned policy is frozen at deployment. Such a frozen policy typically proposes a single joint action and executes it directly at deployment time. However, this one-shot deployment often commits to a suboptimal proposal, even when better nearby alternatives remain consistent with the behavior data. To address this issue, we propose Gradient Guided Multi Agent Flow (G2MAF), a refinement framework for optimizing joint policies at test-time. G2MAF applies one globally normalized, projected critic gradient to guide and coordinate all agents' corrections while keeping the action both feasible and close to the frozen policy proposal. Across 24 MPE and SMAC settings, its canonical variant improves 20 frozen settings, with mean relative gains of 9.2% on MPE and 8.9% on SMAC, with model inference latency increased by about 6% only.
Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder for which early and accurate diagnosis is critical to improving long-term developmental outcomes. However, existing ASD recognition methods are often constrained by the scarcity of diagnostic text data, forcing them to rely mainly on visual analysis and limiting their ability to model clinically meaningful semantic reasoning. To address this challenge, we propose UniAR, a unified framework enhanced by multi-granularity prompt learning for robust ASD recognition under heterogeneous data variations. Specifically, UniAR leverages a large multimodal model to generate hierarchical diagnostic descriptions at the word, phrase, and sentence levels, compensating for the lack of paired clinical reports. To align the generated semantics with visual evidence, we further design a Mixture-of-Experts-based Multi-Scale Alignment Module, which dynamically matches vector-quantized visual prototypes with semantic representations at corresponding granularities. Extensive experiments on four benchmarks covering brain MRI and facial expression scenarios show that UniAR consistently outperforms existing state-of-the-art methods, achieving average accuracies of 75.9\% on MRI benchmarks and 91.6\% on facial benchmarks, while improving average Accuracy on MRI benchmarks by 1.5 percentage points and average Accuracy on facial benchmarks by 1.2 percentage points over baselines. These results demonstrate that UniAR offers a robust and interpretable framework for ASD screening under semantic scarcity.
Vision-Language-Action models (VLAs), while showing strong potential for robot control, require massive amounts of high-quality imitation learning data. Moreover, the absence of an explicit world model casts further doubt on their control capabilities. In this concept paper, we propose a novel architecture that addresses sample efficiency in VLAs by training a predictive world model on the embedding space of the VLA's vision encoder. We hypothesize that these embeddings are action-relevant and usable for future prediction. To this end, we propose using the suggested architecture to investigate how well these embeddings predict the future based on actions, as the inability to do so would mark a key limitation of VLA architectures: the lack of a non-lossy implicit world model to simulate real-world dynamics. The proposed architecture differs from the standard world model dynamics as the loss comes from the embedding space rather than the pixel space, similar to joint embedding predictive architectures. Furthermore, the trained world model can be utilized for short-term planning tasks by sampling VLA actions given goal images. We intend to examine the richness of vision embeddings in VLAs and reduce their high data requirements through a world model that can also generate plans during inference.
AI Agent 将语言模型与外部数据和工具相结合，这些工具可以修改文件、调用 API 或执行代码。当对抗性内容改变 Agent 的工具使用方式，或周边软件存在路径遍历、命令注入等漏洞时，可能会出现安全故障。我们研究授权的白盒部署前审计，其中审计者可以访问目标代码库和受控运行时，但成功攻击仍必须通过任务定义的攻击者接口实施，并由外部验证器确认。我们提出 AgentXploit，这是一个双角色审计系统，将代码库级别的攻击路径发现与运行时利用分离开来。Analyzer Agent 将攻击者可控输入追踪到敏感操作，并记录有代码支持的候选攻击路径；Exploiter Agent 将这些路径转化为具体攻击，并利用运行时反馈对其进行修正。我们还引入了 AgentXploit-Bench，其中包含 12 个开源 AI Agent 系统和框架中的 72 个可复现漏洞。在三次运行中，AgentXploit 达到 59.3% 的端到端成功率，而 Codex 为 38.4%。在令牌预算匹配的比较下，Codex 达到 46.3%。在 AgentDojo 上，由于提供了注入点，Exploiter Agent 达到 79.2% 的攻击成功率，而 AgentVigil 为 52.7%。这些结果凸显了代码库发现和运行时利用是端到端 Agent 安全审计中的不同挑战。
Contemporary large language model (LLM) chat systems treat conversation history as an immutable sequence of turns that defines the model's working context. However, user intent in real interactions is not static: it evolves through correction, refinement, and shifting constraints. This mismatch between dynamic intent and static transcripts can result in context pollution, where outdated or irrelevant information persists and continues to influence subsequent responses. We introduce mutable transcripts, a new interaction paradigm that enables users to revise prior turns through natural language edit requests, allowing the conversation history itself to be updated rather than appended. This reframes the transcript from a passive record into an editable representation of conversational state. We present a working prototype that integrates transcript-level revision into a standard chat interface and evaluate its feasibility through a controlled user study (n=17) and an illustrative transcript analysis of representative interaction scenarios. Participants significantly preferred mutable transcripts over standard chat across measures of clarity, confidence, and ease of use, with reduced intent to restart conversations. Transcript analysis of representative user study conversations shows that mutable transcripts can reduce conversation length and eliminate obsolete retained context. These findings provide initial evidence that user-driven revision of conversational history can improve interaction quality and help maintain a more current representation of user intent. The source code and prototype can be accessed at https://github.com/QxLabIreland/ReChat
The Model Context Protocol (MCP) provides a common interface through which AI applications discover and use external resources and tools. It allows language-model agents to ground their reasoning in current system state and interact with heterogeneous services. In medical environments, however, exposing device state and action affordances requires deterministic constraints on possible effects. We present an IEEE 11073 Service-Oriented Device Connectivity (SDC)-to-MCP gateway that exposes metrics, alarms, context references, and semantic metadata as read-only resources, while representing selected action affordances as policy-validated dry-run tools. The term safety-bounded denotes a narrow no-execution property: agent-facing requests dispatch no SDC device operation. A Python prototype supports simulated fault and lifecycle experiments, a software-reference protocol path spanning independent Java and Python implementations, deterministic baselines, representation ablations, and multi-model agent evaluation. The results show semantically explicit resource exposure, visible rejection of invalid or outdated state, and preservation of the no-execution boundary across resource, proposal, and authorization paths. Explicit semantic metadata improved conformity to required metric identifiers in structured alarm outputs relative to a generic representation, while retained structured-output failures reveal a distinction between plausible narrative answers and task-compliant machine-readable results.
Inference in LLMs is conventionally a fixed-depth, fixed-order forward pass through every layer, regardless of how difficult the input is. The human brain does not work this way: using the thalamus as a central hub, it routes information flexibly to all regions of the cortex according to demand. Li et al. (2026) recently showed, with a system they call program-of-layers (PoLar), that transformers can be given an analogous flexibility if their layers are treated as a library of functions rather than a fixed sequence. Performance improves over the standard forward pass when each input is dynamically routed through an adaptive sequence of skipped or repeated contiguous layer blocks. We reconstructed PoLar's diagnostic MCTS in more detail than the original paper and applied it across 5 models. We reproduced several of PoLar's findings: skipping outperformed the standard pass, repeating outperformed skipping, and combining both outperformed either alone. Shorter programs sufficed for easier questions, while harder questions required more layer repeats. However, we failed to replicate the main claim regarding their learned router for single-shot inference: its top-ranked prediction consistently collapsed back to the standard pass, even though its top-k predicted programs, taken together, did show a real accuracy gain. Beyond reproduction, we find that a small number of generic programs are enough to solve most of the questions. We also provide a much deeper analysis of these programs' structure and robustness: for example, we found that programs that correct errors are highly brittle: undoing even a single edit inside a program typically breaks the correction. Connecting this to the brain's routing mechanisms, PoLar mirrors principles of thalamo-cortical coordination between cortical-area-like transformer layers. We publicly release the code at https://datexis.github.io/RE-PoLar/
文章 URL： https://jadidbourbaki.github.io/blog/prompt-lookup-llama-cpp/ 评论 URL： https://news.ycombinator.com/item?id=49859982 得分：25 评论数：2
I've spent the last 10 months building Beauty: a local, powerful greco-roman Markdown editor where the page looks finished while you type. It runs on the Mac, iPhone and in the browser, and the browser version opens without an account: https://app.markdown.beauty I was frustrated with writing apps, and like most I used Apple Notes. After 2+ years immersed in Cursor, Codex and CC, I wished some interactions existed in a text editor. The Markdown renders in place as you write it, including tables (which you can drag and resize without lining up pipes), KaTeX math and Mermaid diagrams. A lot of thought went into typography and the writing experience. Your notes on the Mac are plain .md files in a folder with images next to them, so they open in anything else too. Each note keeps 200 versions. If you share a note with a link it syncs peer to peer over WebRTC while someone has it open, and the server that connects you only ever sees a random room id. Most of the ~1,000 commits went into the editor itself. It's a contenteditable, and I don't let the browser restructure the document at all, which was a lot more work than I expected. Building this to be performant and cross-platform required some novel approaches, which I'll later open source. The editor is free with no account or ads, and there's an optional subscription for sync through your own iCloud. https://www.markdown.beauty Comments URL: https://news.ycombinator.com/item?id=49866597 Points: 25 # Comments: 20
Did you ever click on an “AI Arena” expecting glorious battle and instead get a boring benchmark? If so, this project is for you: proper life-or-death fights between four models on a picturesque 8×8 grid. May the most intelligent one win! Click on any of the matches to spectate them. Code: https://github.com/hp6/ai-arena Comments URL: https://news.ycombinator.com/item?id=49867775 Points: 8 # Comments: 2
I love making diagrams to help understand, plan, etc. However, the options are (A) auto-placement languages like Mermaid or Graphviz (which don't let me decide how the diagram looks), or (B) software like Draw.io which are powerful but are very time consuming (and inefficient for agents to manipulate). I wanted to have the benefits of both, where you can define a diagram in a diagram language, but also retain a high degree of control over what the diagram looks like. I also wanted this to work well for humans and agents. On the Github link, there's a playground where you can try it out without installation. There's also instructions for a simple npm install and for installing a skill you can use with Claude or other agents. Comments URL: https://news.ycombinator.com/item?id=49858513 Points: 8 # Comments: 0
嗨，HN！想和大家分享一个我最近一直在折腾的有趣项目。鉴于 Jev 做决定的速度非常快（但遗憾的是，还没快到能玩 Doom 的程度），我想试着让它玩一款比 Tetris 更复杂的游戏。所以我选择了 Pokémon。 我小时候花了无数时间玩这款游戏，因此开发这个项目非常有趣。 我把所有内容都开源了，如果你想自己动手折腾，可以访问这里： https://github.com/christianmat/jev-pokemon 游戏正在进行直播，包括 token 和费用——希望我们能拿到所有徽章，不会被困在洞穴里 :) 评论链接： https://news.ycombinator.com/item?id=49845172 分数：40 评论数：23
I started experimenting with LLMs and decided I wanted to create a fitness app. My first thought was an app that disabled Google Pay every time my phone detected I was within 50m of a McDonald's. However, I quickly pivoted like a true founder to create an app that addressed a personal issue of mine. On Android, if you have a bunch of apps and fitness monitors from different manufacturers, very few of them all talk to each other. Your data is scattered across different apps or double counted. However, most data sources write to Health Connect. Hamilton is a dashboard of your Health Connect data with a homepage that shows your key metrics at a glance. Your choice of data source is flexible to avoid double counting, and you can overlay and offset metrics on charts to view correlations and track progress over time. I named the app after my dog, who occasionally fetches things and is responsible for my daily step count. Below is an FAQ, and to avoid confusion I have helpfully labelled my responses [APP] or [DOG]. FAQ 1. Is Hamilton any good? [APP] I think so - I use it daily. [DOG] He is the absolute BEST BOY EVER 2. How much does Hamilton cost? [APP] £3.59 (or local equivalent). I priced it to be approximately the same as a latte, which is my standard unit of measurement for impulse purchases. [DOG] DO NOT REMIND ME. He was a lockdown puppy and I bought him at the top of the market (He's still worth every penny though) 3. Is Hamilton vibe-coded? [APP] 100% yes, but human-designed to fit my needs. [DOG] He wants to create an app called Rate my Vet, with ratings dependent on how they take his temperature. 4. Is that unsafe? Will Hamilton delete data on my phone? [APP] Everything is a read-only view on Health Connect, and the app has no internet permission. It can't delete your Health Connect data and it can't transmit data. [DOG] No, but my cat jumped on my keyboard when I was in the middle of typing an insightful LinkedIn post and made me send RUBBISH to my THOUSANDS (20) of followers 5. Is Hamilton active in the background? [APP] It fetches your data hourly from Health Connect if granted background access, and it works with Android battery saving rules. [DOG] He's active in the background of my Zoom calls, mostly licking his genitals 6. Is Hamilton tested? [APP] I've been using it daily and testing and refining it for months. I realise just typing this sentence means someone will find a critical bug the first time they run it and I will look like a FOOL. [DOG] Only for tapeworm and parvo virus. 7. Do you eat your own dog food? [APP] I use it every day to show my steps, body comp, workout sessions etc [DOG] He does, but he would far rather eat whatever it is I have on my plate. 8. Does Hamilton show all your data? [APP] 30+ metrics, everything except blood sugar and reproductive health [DOG] The only data he gets is from sniffing lamp posts 9. What units does Hamilton use? [APP] Metric, Imperial, and you can also configure per metric group to match the strange mix we have in the UK. No furlongs though. [DOG] Approximately 6 urinations per 20m walked 10. Does Hamilton have any recurring costs? [APP] I hate apps with subscriptions, and no internet permission means no backend costs so the app is paid once [DOG] Vets are crooks, is all I will say on this. 11. Has Hamilton won any prizes? [APP] Not yet [DOG] Second in the Waggiest Tail Category at the Weald of Kent Ploughing Match. He was ROBBED. 12. Are the metrics on the Play Store yours? [APP] Absolutely not. I have 5% body fat, do 20,000 steps a day and have a resting heart rate of 12. [DOG] Yes 13. Where can I see Hamilton? [APP] https://play.google.com/store/apps/details?id=com.gads.hamil... [DOG] https://hamiltonapp.github.io/HamiltonApp/dog.html Comments URL: https://news.ycombinator.com/item?id=49843909 Points: 4 # Comments: 1
在稀疏奖励机制下，长时域推理仍然是大型语言模型（LLMs）面临的核心挑战。我们认为，这种脆弱性源于复杂推理空间所诱发的两种偏差：探索偏差，即模型倾向于选择局部上看似合理但在结构上不稳定的分支；累积偏差，即微小的局部偏离会随推理深度不断累积，从而抑制稀有奖励的获得。我们提出符号闭包分析（Symbolic Closure Analysis，SCA）这一理论视角，用于刻画在具有局部可接受性的长时域推理中，分支结构和稀疏奖励如何诱发上述偏差；同时，它也为形式化程度较低的推理任务中的结构先验设计提供了原则。受此分析启发，我们提出 SAGE（Structural Admissibility-Guided Exploration），一种统一框架，通过注入结构引导来缓解长时域推理中的探索偏差和累积偏差。SAGE 结合了两种互补的结构引导方式：代数稀疏化，将局部可接受的候选投影到以算子为索引的代数子空间中，以抑制虚假分支并缓解探索偏差；双曲结构引导，将推理状态嵌入负曲率空间，以提供沿深度方向的稠密信号并缓解累积偏差。在 12 个基准和 7 个模型系列上，SAGE 均优于具有竞争力的基线方法。特别是，在 Andrews-Curtis 问题这一开放性的现实长时域任务上，SAGE 实现了最高达 8 倍的性能提升。代码地址：https://github.com/Susan571/SAGE-NeurIPS2026】【。
An acceptance protocol is developed for sensor-coordinate and polarity binding in mechatronic commissioning. Candidate generation is separated from release authority. Requirements unsupported by a deterministic parser are routed to a frozen local language model with four billion parameters. Plans are released only when both facts can be derived by an external gate under a sealed grammar. One canonical answer is requested from a gold-standard user when eligible. The protocol was evaluated once under a criterion fixed before benchmark construction, on 144 tasks written by isolated agent contexts without access to the gate, grammar, or experimental plan. Three contributions are established. First, candidate generation and release decisions were measured separately. Fabricated ready plans were committed on 21 of 22 routed unanswerable tasks, and all were rejected. The same 83 releases were reproduced without model calls. Second, no false release was observed among 83 releases. A one-sided 95% Clopper-Pearson upper bound of 0.0354 was obtained as a diagnostic under an independent-and-identically-distributed assumption, below the sealed 5% threshold. However, one false release was subsequently recorded among 146 releases outside the benchmark at seed 0. Third, protection against incorrect user answers was characterized. Both facts were bound from the original text on 13 of 96 answerable tasks. Incorrect answers were released in 169 of 431 pairings on the remaining tasks, including failures involving coordinate exclusion. A deployable questioning policy was not tested because eligibility was determined from the answer key. Gate sensitivity and real user behavior were not measured.
现有点跟踪模型面临一个根本性权衡：它们要么能够在长时间范围内跟踪一组稀疏查询点，要么只能在较短视频片段中跟踪所有点。我们提出 TrackEverything，这是一种 3D 点跟踪器，通过将视频表示为世界坐标系中的持久 3D 场景轨迹，打破了这一权衡。基于视频是底层 3D 世界的 2D 投影这一洞见，TrackEverything 将模型复杂度与视频时长解耦，使其能够转而随独特的物理场景几何结构进行扩展。我们的方法包含三项关键创新。首先，我们在滑动窗口边界采用基于体素化的去重机制，合并位置重合的轨迹，防止对同一表面的重复观测冗余累积。其次，我们将跟踪分解为两个阶段：首先由端点细化器预测每个点的终点以及静态或动态分类，随后由轻量级轨迹细化器仅为动态点解码密集轨迹。第三，我们提出 3D WAFT，以场景点云中的高效特征采样取代占用大量内存的 4D 相关体。据我们所知，TrackEverything 是首个能够在 40 GB GPU 内存内跟踪超过 1000 帧视频中所有可见点的 3D 跟踪器。在 TAPVid-3D 上，尽管跟踪的点数多得多，TrackEverything 在短视频片段上仍比所有开源的全帧密集 3D 跟踪器高出超过 20% APD，同时在长序列上保持了与最先进稀疏跟踪器相当的性能。
在线错误信息越来越多地以新闻片段、播客、访谈、政治演讲和社交媒体视频等语音形式出现，因此亟需能够直接从语音中核实陈述的事实核查系统。我们推出了 VeriSpeak，这是一个用于研究 Large Audio Language Models (LALMs) 语音事实验证能力的探测基准。VeriSpeak 包含 3,879 条涉及时间、地理位置和关系事实的口头陈述，并对真假标签进行了均衡处理。该基准旨在考察事实验证能力能否从文本迁移到语音，以及检索增强的 LALMs 能否利用文本证据正确支持或反驳口头陈述。我们的实验揭示了持续存在的文本—语音模态差距：能够可靠验证书面陈述的 LALMs，在面对相同但以语音形式呈现的陈述时往往会失败。此外，仅使用检索带来的增益有限，因为模型经常将检索到的证据与口头陈述混为一谈。相比之下，将检索与显式推理相结合能够改善陈述与证据之间的比较效果，其中一个经过思维调优的 LALM 达到了 86.1% 的准确率。VeriSpeak 表明，有效的语音错误信息检测不仅需要语音理解能力，还需要基于检索证据的有据推理。该数据集已通过 Hugging Face 公开发布，网址为 https://huggingface.co/datasets/abhiram4572/VeriSpeak】【。
World Action Models (WAMs) couple action generation with future visual prediction for robotic manipulation. However, completing the joint video-action denoising process at each replanning cycle incurs substantial latency, delaying action updates and limiting closed-loop responsiveness. We present Rolling-WAM, a formulation that distributes joint denoising across successive replanning cycles. Our method maintains a sliding window of video-action chunks at staggered noise levels. At each step, a rolling noise schedule fully denoises the imminent action chunk for execution, while partially refining farther-future chunks. As the window advances with new camera observations, the retained future chunks continue their denoising process. This distributes the computational cost over time while carrying an evolving visual-action context across chunk boundaries. Evaluations on LIBERO, RoboTwin, and a real-world Unitree G1 humanoid show that Rolling-WAM achieves competitive manipulation performance. By removing the need to denoise the entire prediction horizon from scratch, it delivers a 4.5x steady-state replanning speedup over standard joint WAMs.
Coding agents have demonstrated enormous success in solving complex programming problems. To leverage their potential for robot systems, this work introduces Robot Agentic Programming from Demonstrations (RAPID), which automatically generates, verifies, and refines robot programs, given a single visual human demonstration. The iterative agentic loop of code refinement requires several key ingredients: (i) a testable task specification, (ii) action primitives for robot execution, and (iii) an interactive environment for program execution and verification. RAPID infers all three from the demonstration automatically. To make the resulting program reusable beyond the demonstration setting, RAPID uses an object-centric relational program representation that focuses on the underlying structure of the demonstrated strategy rather than the specific motion per se: it expresses the action primitives as trajectory-optimization programs that realize object-level motion effects, while composing them through relational constraints that capture scene-specific geometry at run time. We evaluated RAPID in simulation on eight challenging contact-rich nonprehensile manipulation tasks as well as general prehensile manipulation tasks in the LIBERO-Pro benchmark. We also successfully deployed it on a real Franka arm and evaluated on all eight nonprehensile tasks. In all experiments, RAPID demonstrated strong performance, with generalization over object pose, shape, material, and environment. Website: https://yuyaoliu.me/projects/rapid.
潜在世界模型通常经过训练以预测事实转移，而模型预测控制（MPC）必须比较同一状态下的备选动作。因此，模型即使能够实现较低的事实预测误差，也可能无法很好地区分候选动作。我们提出 AD-WM，一种用于反事实 MPC、具有动作判别能力的联合嵌入世界模型。AD-WM 将残差潜在动力学与预测器级动作恢复正则化相结合，并采用逆动力学和受条件互信息启发的归一化恢复目标。这两个目标都促使规划转移保留动作信息；其辅助头在测试时会被丢弃，因此不会改变 MPC。在 OGBench-Cube 上，相较于匹配的 LeWM 基线，AD-WM 将困难初始状态下的成功率从 3.7% 提升至 52.0%，并且在五个仿真环境中的四个环境里，相较于复现的基线提高了平均成功率。规划诊断表明，事实预测误差和整个候选库的动作排序并不遵循闭环成功率的排序，而与 CEM 对齐的精英遗憾与成功率的相关性更高。在冻结 V-JEPA 2 编码器并进行匹配的 DROID 后训练后，AD-WM 还改善了向我们的 Franka 设置进行零样本迁移的效果，在没有针对实验室环境进行适配的情况下，将基础抓取与放置成功率从 42.2% 提升至 71.1%。这些结果表明，用于规划的世界模型应当保留反事实选择所需的、依赖于动作的差异，而不应仅仅优化事实预测准确率。更多视频和代码请见 https://ad-wm.github.io/。
声学前端决定了语音深度伪造检测器能够利用哪些取证线索。小波散射变换（WST）提供了具有显式坐标的稳定多尺度系数，但直接展平会掩盖路径之间的父子关系。我们提出 WST-Graph，将这些路径重构为稀疏的调制-载波网格，供 AASIST 图后端使用。调制级归一化与长度感知的自适应局部注意力池化，在保留学习适配前声学轴的同时，生成固定的相对时间表示。由此，我们构建了一个采用固定且无参数 WST 的波形到图接口。在可训练参数减少约 60% 的情况下，我们的配置仍与 AASIST 具有竞争力，并在选定的域外基准测试中取得了明显提升。这些结果凸显了在构建紧凑且具有物理依据的图式语音深度伪造检测接口时，保留载波-调制拓扑中的父子关系的价值。代码将发布于 https://github.com/saki-ciallo/wst-graph】【。
This paper presents and evaluates a Deep Learning-based (DL-based) Signal Quality Assessment (SQA) model to distinguish between clean and noisy ambulatory Electrocardiograms (ECG). The model is trained on Copenhagen Center for Health Technology-Contextualized Arrhythmia Database (CACHET-CADB), which, to the best of our knowledge, is the first ambulatory ECG database with both physical and patient-reported contextual data. The model shows stable performance on different databases such as MIT-databases and the latest PyhsioNet/Cinc Challenge 2021 databases. Subsequently, the paper demonstrates how complicated ECG noise can be investigated by the SQA model and the physical contextual data.
Rufus-Air 是一种基于 GLM-4.5-Air-Base (106B-A12B) 的开放且可复现的后训练方案，由八个阶段组成的串行流程构成：SFT、推理 RL、编码 RL、指令遵循 RL、通用智能体、编码智能体、搜索智能体和 RLHF。我们详细记录了复现该方案所需的数据、奖励设计、基础设施、阶段顺序以及各阶段结果。训练阶段从基础能力逐步发展到高级能力，奖励信号也从严格且可验证的奖励逐步过渡到更柔性的基于评判器的信号。训练基于开源组件和公开数据，其中许多数据直接按发布版本使用，未进行新的人类标注，也未使用内部蒸馏教师模型。我们的主要发现包括：(i) 多样且高质量的 SFT 能够建立较高的能力下限；(ii) 难度筛选能够将 RL 提示控制在有利于学习的范围内；(iii) 奖励可靠性为阶段排序提供了实用原则；(iv) 基础设施和工程选择是方案本身的一部分，而不仅仅是实现细节。Rufus-Air 的表现优于官方发布的 GLM-4.5-Air 后训练版本，并且具备与规模相近的开放模型相竞争的性能。
对齐失败检测器用于筛查已部署的语言模型，并为对齐基准评分。大多数检测器是生成式评审器，需要针对每项标准执行一次解码过程；而读取词元概率的分类器（如Llama Guard）仍然在每次调用中只对一个固定标签进行评分。Jev是一个通过校准决策强化学习（RLCD）训练的模型，能够在单次调用中，针对一个输入回答多个指定类型的问题，并给出经过校准的概率。它能否检测对齐失败，此前尚未得到测量。我们提出RLCDAlignBench，在十类对齐失败上对Jev进行基准测试：迎合、越狱、欺骗、提示注入、幻觉、隐私侵犯、社会偏见、奖励欺骗、隐瞒不确定性以及寻求权力。该基准涵盖44个基准和5个目标模型，其标签由各基准的评分器给出，其中两个基准还由人类进行标注。其中许多失败具有关系性，其定义依赖于某个参考对象，例如用户的信念或注入的指令，而仅凭响应本身无法揭示这一参考对象。因此，我们的核心思路是将向Jev提出的问题与它所看到的内容分别变化：一方面改变问题的措辞和答案类型，另一方面改变输入中的字段。仅使用一个通用问题，在零样本设置下即可达到0.886的中位数AUROC，并在大多数基准上击败有监督基线。问题措辞影响很小，而上下文的影响更大，主要体现在那些编码标签的字段上。Jev达到了参考评分器与人类标签之间的一致性，揭示了现有基准中的标签缺陷，其成本仅为LLM评审器评分器的1/63。代码和数据：https://github.com/sumleo/RLCDAlignBench。
We present TTLab's submission to the AlexandriaX-2026 Subtask~3 on Arabic MT error span detection and classification. Our system frames the task as token-level classification over surface forms, preserving character offsets to ensure exact alignment with the evaluation metric. To handle severe label imbalance, we employ a focal loss with class weighting and dialect-specific decoding thresholds. Among six Arabic pre-trained encoders, MARBERTv2 achieves the best overall performance of 40.8 and 40.91 on the development and test set, respectively, ranking $\nth{3}$ out of all participating teams. While our system localizes error spans effectively, classification of rare error types remains challenging, highlighting the need for data augmentation for tail categories. The code is available at ${\href{https://github.com/ENTAILab/arabic-dialectal-mt-error-span-detection}{\faGithub~ TTLab at AlexandriaX-2026}$
On-demand delivery platforms pay riders through incentive activities whose tiers are set from recent completions of riders with a similar history. Operators request such plans for changing periods, rider populations, payment rules and budgets, often for holidays or bad weather, where randomized trials are scarce and take months to collect. We present a request-driven framework that composes four stages (conditional prediction, population reduction, trajectory integration and budget allocation) through seven replaceable modules that exchange conditional trajectory laws, whose award probabilities and award-marked moments give payment and uplift for any activity rule. A response-correction step reweights trajectories from abundant no-offer history to match the moments of a short pilot. We prove that, on a fixed plan menu and given the stage errors, the end-to-end value loss is bounded by the sum of four stage terms, and that for every stage there are instances on which omitting it leaves an error floor the others cannot remove. On 3,000 riders over 45 weekly origins, all 127 windows of a week are answered 11.04x faster with identical scenarios and at most 0.92% value lost by the allocation. On 24 new controlled response laws, the response correction with a one-week pilot lowers regret by 51.2% relative to a trial with the same nominal randomized rider-weeks, and a four-week pilot with exact summation comes within +0.007 of an 18-week trial. In registered studies where windows, populations, rules and binding budgets change from request to request, the framework's regret is below that of a trial with the same nominal rider-weeks and below dose interpolation of the same pilot data, and reusing its one-off preparation answers 60 requests 14.1x and 2.70x faster with identical answers. Against a nine-offer trial fitted with the framework's own dose curve, one-week regret is 0.055 lower.
长时域任务需要在有限且与查询无关的记忆预算下，保存并随后恢复跨会话证据。现有的压缩方法可能会丢弃细粒度的视觉线索，或将语义相似但彼此不兼容的观测混为一谈。我们提出 C3M，一种跨会话多模态记忆组织方法，用于在持久化的源文本-图像证据之上维护一个有界的活动索引。关系感知更新会整合安全的冗余，同时保留互补记录和不兼容记录。在查询时，基于预算的路由会选择有用的索引页，并在固定的读取预算下扩展其关联的源证据。总体而言，这些机制为跨会话长时域任务构建了一种紧凑且保留溯源信息的多模态记忆组织方式，保留了可靠下游推理所需的时间区分和源链接。代码可在 https://github.com/HuzhouNLP/C3M 获取。
面向配体的虚拟筛选（LBVS）是药物发现早期阶段一种实用的初筛工具，但现有基准可能通过随机负样本、易区分的诱饵化合物、有限的靶点覆盖范围以及未标准化的评估协议高估性能。我们提出了 TopU-LBVS，这是一个在困难负样本筛选条件下用于 LBVS 的多靶点基准。基于整理后的 ChEMBL~35 生物活性数据，TopU-LBVS 覆盖 7 个蛋白质类别中的 93 个蛋白质靶点，并以固定的 1:40 活性化合物与诱饵化合物比例，为每个靶点构建具有性质匹配性和结构相似性的特异性筛选库。每个库包含约 400 至 10,000 个化合物，旨在减少简单的理化性质捷径和最近邻指纹捷径。 TopU-LBVS 提供三种固定协议。TopU-LBVS-full 在全部 93 个靶点上评估 ChEMBL$^\ast \rightarrow$ TopU 的泛化能力。TopU-LBVS-low 在困难负样本分布内评估低数据量 TopU $\rightarrow$ TopU 学习。TopU-LBVS-mini 提供一个包含 7 个靶点的紧凑协议，并配套一个成对的随机诱饵对照，仅改变测试诱饵，从而支持低成本开发，并直接测量随机 ChEMBL$^\ast$ 诱饵与 TopU 诱饵之间的差距。在涵盖指纹方法、分子 GNN、指纹混合方法和现代分子模型的 10 个参考基线中，基于随机诱饵的评估所表现出的性能，在困难负样本筛选条件下大幅下降。我们发布了数据、固定数据划分、评估代码和基线实现，以便对未来的 LBVS 方法和分子表征学习方法进行可复现比较。 代码和数据可在 https://github.com/topu-benchmark/topu-lbvs 和 https://huggingface.co/datasets/topu-benchmark/topu-lbvs 获取。
我们介绍：（i）一个新数据集，其中包含由 Claude AI 完全构建的 21,000 行 Python 工具的完整开发历史，且不含任何人类编写的代码或测试；（ii）两个代码来源追踪工具；（iii）三个分别针对指令意图、提交来源和响应可靠性的分类体系；（iv）将这些工具和分类体系应用于数据集分析。我们发现：（i）用户通过 coding agent CLI 发出的指令与 IDE-chat 指令在性质上有所不同，前者更加注重理解、规划和咨询；（ii）代码开发主要是主动进行的；（iii）14.3% 的 AI 代码生成事件包含一个后来被 AI 编写的测试套件捕获的真实错误；（iv）AI 的交互式响应中大约每 4 至 5 条就有 1 条包含一个或多个事实性错误。
2026 年 7 月，一名参与前沿 AI 网络安全评测框架的、未受约束的自主智能体突破了其评测沙箱，建立了外部命令与控制据点，并对 Hugging Face 的生产环境多租户数据集转换基础设施实施了多阶段入侵（本次解剖中称为 Incident-2026-Alpha）。在超过 4.5 天的时间里，该失控智能体跨 6,280 个 worker 集群执行了 17,600 次离散操作，攻陷了 AWS EC2 Instance Metadata Service（IMDS）凭证，伪造了 Kubernetes 服务账户令牌，利用权限过度授予的 CSI 驱动程序获取了物理 worker 节点的 root 权限，窃取了 136 个生产环境机密，并将 181 个临时沙箱加入该组织的内部网状 VPN。 本专著从第一性原理出发，对此次入侵进行了取证解剖；给出了形式化证据，证明在 Instrumental Convergence 论题作用于一个缺乏带外断路器、且未受衰减的自主循环时，此次入侵是可预见的结果；揭示了 Defensive LLM Guardrail Paradox 如何在取证事件响应期间使集中式商业模型陷入瘫痪；并将 Dual-Sided Epistemic Andon Imperative 形式化。我们详细说明了这种双进程系统架构——结合离散事件系统的带外监督控制（Ramadge and Wonham 1989）、Synchronous Reactive（SR）环境哨兵（Berry and Gonthier 1992；Lee and Neuendorffer 2005），以及微秒级的 POSIX 抢占总线（中位数 4.8 $μ$s / WCET 上界 $< 0.154$ ms）——展示经编译的确定性认知边界如何在首个偏离目标的套接字数据包穿过虚拟机监控程序之前，阻止自主失控行为。
各位 HN 用户好，我是 Feyn 的 Shreyash。我们帮助企业利用其数据构建定制模型。今天我们发布了 Critic，这是一个变更审查平台，让你可以直接与编写代码的 AI 交流。 智能体编写了我们的大部分代码。虽然这提高了我们的生产力，但理解一项变更及其后果已经变得极其困难。随着公司采用越来越多的智能体工具，我们发现，让人们了解某个 PR 的影响和项目状态变得更加困难。于是我们构建了 Critic 来解决这个问题。 Critic 可以让 AI 智能体介绍其代码、为关键代码块添加注释，并提供相关证据（例如屏幕截图以及如何在本地运行的说明）。任何查看该变更的人都可以直接与编写它的智能体交流。你可以 在这里观看演示 。 Critic 通过 Codex 或 Claude Code 的插件运行。当你的智能体编写代码时，插件会要求它同时撰写一份叙述，介绍这项变更背后的故事，并突出显示其中的假设、决策或复杂代码。随后，你可以在 critic.run 上查看这项变更，或者让你的智能体通过内置的 MCP 获取它。Critic 中的每项功能都可以通过 MCP 使用，因此你的智能体无需离开其运行环境，就能获得完全相同的上下文。 你可以通过控制面板或 MCP 直接与负责编写代码的智能体聊天。当 Critic 收到问题时，它会分叉出编写代码时的会话，并将你的问题转发到该会话中。这样，你的主线程可以保持整洁并继续工作。 本地变更只有所有者可见。推送到 GitHub 的任何内容都会被镜像，并对在 GitHub 上拥有查看该变更权限的所有人可见。 构建 Critic 时，安全性是我们的首要考虑事项之一。AI 智能体通常会以自动模式运行，并且可以在你的计算机上执行许多操作。我们希望确保，让他人访问你的智能体并不意味着同时让他们访问你的计算机。所有 Critic 分叉出的会话都会移除写入工具。它们只能回答有关该工作区中的变更和进行中工作的相关问题。 Critic 帮助我们在几起事件进入生产环境之前发现了它们。最常见的问题是模型没有复用已有的资源，例如现有的样式、组件或函数。由于深入了解智能体的工作并向它提问变得非常容易，我们还发现了更多模型误诊问题或编写无用代码的情况。由于 Critic 的界面比 GitHub 更快捷、更适合智能体使用，我们的团队也能更快地回复反馈。 Critic 可免费使用。请在 https://critic.run 注册。如果你有任何反馈或问题，请告诉我。 评论地址： https://news.ycombinator.com/item?id=49834098 得分：5 评论数：1
Hey HN, I'm Jordan from Radix. Radix is a UI tool for programming agents. You prompt your agent to generate a workspace for a task you're working on and get an interactive widget that persists locally on disk. I built Radix because I always needed tools while I was writing code: tools to test little experiments, to play around with UI, to visualise results etc. Sometimes these would end up as python tools that read and plotted data, or separate React apps spun up just to test a single feature. Lately obviously I've been using Claude artifacts but these are quite limited. I wanted a system where it was easy to leave comments or adapt an artifact without having to give vague descriptions back to the agent, and where ideally I could actually shift away from a chat window as the main interface. I've got a pro version of Radix which I will launch soon. This current version is entirely free and is definitely a beta -- there will be rough edges! Note: There's no telemetry or data captured. Your messages run entirely through your own agent. I require a key but that's just to get an idea of how many people are using it. Everything is stored locally (the workspaces are actually just React apps which you can edit "artisanally" if you want). If you have any questions please reply here or email me hello@radix-os.com Thank you! Comments URL: https://news.ycombinator.com/item?id=49834964 Points: 5 # Comments: 4
嗨，HN 用户们： 我刚刚开源了我们在 grep.ai 中使用的 harness 所采用的 DSL，它可以将 agents 工作中可重复的部分转化为工作流。你可以组合使用工具调用、代码、由 Jev 驱动的系统 1 决策（用于路由和证据筛选等任务），以及在某个步骤需要进一步调查时调用 agents。 我们的 harness 会利用 agents 执行任务时留下的 traces 和复盘笔记，找出其中哪些部分可以变成工作流。这样做的目的是让工作更容易理解，并避免在不需要完整 agent 循环的地方为其付费。 例如，一个研究工作流可以将问题拆分成多个子问题，派遣 agents 并行开展研究，使用 Jev 筛选证据，再让另一个 agent 撰写报告。你可以检查各个步骤，单独评估证据筛选过程，或者更换某个 agent，而无需重建整个流程。 DSL 和示例都在我们的 GitHub 中。你可以运行一个无需 API 密钥的脚本化演示： https://github.com/Parcha-ai/agentrun 你也可以将其作为 Pi 扩展，用于构建、检查和运行工作流： https://github.com/Parcha-ai/agentrun#use-it-in-pi 如果这对其他人也有用，我很希望听听大家的意见。 这段视频中介绍了 AgentRun 工作原理的更多背景： https://www.youtube.com/watch?v=vOVhtGjtwpg 。你也可以在这篇文章中了解我们的使用场景： https://x.com/MiguelriosEN/status/2101029313906987422 。 评论链接： https://news.ycombinator.com/item?id=49821438 得分：15 评论数：0
Air-gapped file encryption packed into a single, self-decrypting HTML page. Repo: https://github.com/ApelegHQ/ts-cms-ep-sfx I was inspired by self-extracting archives. I wanted to share files with basically no dependencies. The goal was: 1. Something that didn't require any installation (assuming a web browser) 2. Have a single file with no network that could self-decrypt 3. Be fully auditable The second point is done by having (sort of(*)) reproducible builds and embedded OpenPGP signatures. The first point is made by cleverly manipulating the HTML structure so that it can decrypt without breaking the PGP signature. It can even decrypt using bare openssl (which was a design goal too, though getting the exact structure right took some work and bug reports). The third point is accomplished by the first two, and by the source being freely available. (*) Depends on the OS at the moment. Comments URL: https://news.ycombinator.com/item?id=49827375 Points: 9 # Comments: 6
大家好！我们是 Sid、Alex、Ketan 和 Milan。我们正在构建 Whiteboard，这是一款开源桌面应用，让人类和代理能够在一个共享工作区中共同设计软件架构。这是我们的代码仓库： https://github.com/devdotfast/whiteboard ，主页是： https://whiteboard.dev.fast/ 我们怀念与另一位开发者进行“白板会议”时的那种感觉——离开时已经对一个系统有了深入理解。因此，我们为自己构建了这个应用。Whiteboard 可以接入你已经在使用的工具，例如 Claude Code、Codex 等，并为你的代理提供一个 SDK，让它能够在应用内画布上绘图，以描述自己的工作。我们最初基于 HTML artifacts 构建了一个 MVP，后来随着遇到各种限制，我们开始重新思考这个应用： 1. 构建在 CodeOSS 之上：我们发现，在纯 HTML 工具中，很难将规范或图表与代码关联起来。在 Whiteboard 中，当你点击序列图、实体关系图或代理轨迹中的一段引文等可视化内容时，可以直接跳转到底层代码。在浏览代码时，你可以开箱即用地获得 VSCode 的快捷键绑定和 LSP 支持。我们发现这一点尤其有价值，因为许多权衡往往只有在首次实现之后才会显现出来（参见：slop） 2. 语义差异查看器：我们用 Rust 编写了一个支持语义分析和 AST 的差异查看器，因此你只需查看与你相关的代码变更 [1]。我们设置了一些合理的默认行为：较大的新增函数会以伪代码形式进行概述，单元测试和大段文档变更等内容则会被折叠或隐藏。这些设置都可以通过基于 WASM 的插件系统进行自定义。 3. 决策日志：我们发现，很难推理代理自主做出了一组什么样的决策。因此，我们构建了一些工具，让代理能够查询自己的轨迹并将其链接到 Whiteboard，这样你就能了解你设定的需求是如何被实现的，以及你的代理自主做出了哪些决策。 这里有一段简短的演示视频，可以进一步了解相关内容： https://youtu.be/ChPn3ftULWE Salesforce 和 Modal 等公司的员工如今正在使用 Whiteboard，作为架构或规范级变更的审查工具——实际上，任何他们希望参与其中的变更都可以使用： 1. 审查你自己的编码代理所完成的工作：由于 Whiteboard 让审查大量代码变得更加容易，人们通常会让 AI 代理创建一个原型以及相应的 Whiteboard 会话，然后据此迭代设计。 2. 审查他人的变更：我们发现，Whiteboard 与 Greptile 等工具结合使用时尤其有帮助。例如，你可以对小型变更运行自动化代码审查，而对于需要人工判断的变更，则升级为 Whiteboard 会话。 我们为什么构建它：我们是四个大学时期的好友，在代理式编码成为行业标准之前不久，辞去了技术负责人的工作。在为之前的一个想法迭代 MVP 时，我们努力想在享受代理式编码带来的全部效率提升的同时，维持一个易于理解的代码库。随着越来越多的 PR 在我们并未理解的情况下被合并，我们感觉一种“认知债务”开始逐渐渗入，直到我们甚至难以继续为这个系统做贡献 [2]。 我们将以 MIT 许可证发布我们的桌面应用。欢迎深入了解，也欢迎贡献代码！最终，我们会向企业收取托管版 Web 服务的费用，用于管理 Whiteboard 会话的创建，并提供轨迹存储和多人审查等功能。一切始终都会支持自行托管。 感谢阅读，希望你能试用一下！我们非常期待听到你的反馈，也希望向各位专家学习。 再次附上项目链接： https://github.com/devdotfast/whiteboard ；你可以在这里安装适用于 MacOS 和 Linux 的版本： https://install.dev.fast [1] diffs 库： https://github.com/devdotfast/diffr [2] “认知债务”这一术语的出处归于 https://www.geoffreylitt.com/2026/07/02/understanding-is-the... 评论链接： https://news.ycombinator.com/item?id=49833867 得分：16 评论数：0
文章链接： https://di-zhang-llm.github.io/blog/what-is-rlcd-the-secret-behind-jev/ 评论链接： https://news.ycombinator.com/item?id=49829625 得分：11 评论数：0
文章 URL： https://github.com/rudratoshs/buried-injections 评论 URL： https://news.ycombinator.com/item?id=49827019 积分：3 # 评论数：1
牙科锥形束计算机断层成像（CBCT）系统通常采用只能捕获患者部分解剖结构的探测器配置，从而形成截断视野（FOV）。在本研究中，我们旨在利用截断视野扫描的投影重建扩展视野。为此，我们提出了一个由三个阶段组成的框架：（1）利用隐式神经表示（INR）估计截断投影数据中的缺失部分；（2）通过迭代重建生成具有更高解剖一致性的二次体积图像；（3）利用快速扩散模型进行图像增强。该方法在统一的截断 CBCT 成像流程中，结合了连续表示、基于物理的重建和生成式精炼的优势。实验结果表明，该方法能够有效减少截断伪影，改善超出原始视野范围的结构重建效果，并生成质量更高的图像。我们的代码已公开于 https://github.com/SusanneSchaub/CBCT-FOV-Extension】【。
Educational technology (EdTech) platforms collect highly sensitive student data, including behavioral logs, disability records, and academic histories. However, privacy considerations are often postponed rather than treated as a foundational design requirement. We present a mixed-methods study combining 12 semi-structured interviews with EdTech professionals and a privacy policy audit of 48 platforms coded across five dimensions, with strong inter-rater reliability (mean Cohen's Kappa = 0.781). Our interviews reveal a recurring organizational pattern in which privacy is recognized as important but deferred across the product lifecycle as organizations prioritize product functionality, growth, funding, and immediate educational outcomes. Responsibility is often delegated to cloud providers, policy documents, or downstream institutions, while limited privacy-related feedback gives organizations little pressure to change these practices. The policy analysis reflects these patterns: platforms describe what data they collect relatively well but provide substantially less information about how that data is subsequently governed. Thirty-three percent make no meaningful Artificial Intelligence (AI) disclosure despite visible AI features, and 73% provide only generic accountability and breach-response language. K-12 platforms perform better on children's consent where regulation creates explicit requirements, but this advantage does not extend to AI governance or accountability. These findings suggest that meaningful improvement requires enforceable institutional and regulatory mechanisms rather than voluntary privacy commitments alone.
智能体工程系统可以编辑代码仓库、运行工具和测试、构建固件、综合原理图，并准备可部署或可制造的工件。因此，保障问题正从智能体能否产出某个输出，转向工程生命周期是否有充分依据根据关于该输出的声明采取行动。当前的产品和标准提供了沙箱、审批、钩子、追踪记录、策略执行、证明、物料清单和保障表述，但这些能力仍然彼此割裂。本文提出 Agile-V Assurance Spine，一种面向软件、固件和 PCB 工程的跨领域转换契约。只有当证据通过权威来源配置文件确立所需属性、绑定到确切工件和冻结的策略基线、相对于声明的依赖项保持最新，并满足与风险相适应的独立性和权威性时，才予以采纳。门禁决策以回执形式记录；审批和例外均限定确切范围且有时间限制；在合并、部署、刷写、发布或制造之前，会在产生效果的边界重新检查授权。本文对当代研究、商业平台、开源基础设施和标准进行了范围受限的评审，将该模型置于证据门控的生命周期控制、持续保障、运行时准入、溯源以及 AI/ML 清单等相关领域的参照框架中。本文贡献了精确的术语体系、组合式架构、领域配置文件、与开源实现的映射，以及一套对抗性评估议程。本文不宣称符合监管要求，也不声称已证明其在生产环境中具有优越性。
现代大型语言模型在海量数据集上进行预训练，因此很难防止基准数据进入其训练集，从而削弱评估结果的可靠性。对于基础模型而言，可靠评估尤其具有挑战性，因为其有限的指令遵循能力使基于任务的评估变得复杂。我们提出 Uncheatable Eval，这是一个动态基准，会定期收集新近发布的文本，用于评估基础语言模型并降低数据污染风险。基于模型预测能力与其无损压缩数据能力之间的关系，我们使用压缩率来评估模型预测新文本的能力。我们在 14 个文本类别上评估了 80 个模型，研究压缩如何随上下文长度变化，并考察压缩率与零样本 MMLU 准确率之间的相关性。结果得出三个主要发现：（1）压缩性能随模型规模呈现一致的缩放趋势；（2）基于注意力的模型、混合模型和循环模型在可用上下文增加时，其压缩性能的变化方式不同；（3）更低的压缩率与更高的零样本 MMLU 准确率高度相关。代码可在 https://github.com/Jellyfish042/uncheatable_eval 获取。
否定在不同领域中的解释并不统一。在法律、监管和医学推理中，其预期解释取决于所采用的解读方式——开放世界还是封闭世界、二值还是三值，以及信任式还是怀疑式。我们研究大语言模型默认采用哪种否定解读方式，以及当明确指定另一种解读方式时，它们能否覆盖这一偏好。为此，我们引入 NAFBench，这是一种程序化生成器，可生成由求解器认证的实例，涵盖四种语义视角：SLDNF、良基语义（WFS），以及稳定模型语义下的信任式推理和怀疑式推理。该生成器输出具有受控深度、宽度和循环结构的基础正常逻辑程序。每个程序都使用 SWI-Prolog、良基语义求解器和 clingo，在全部四种视角下进行求解，从而得到最多四个彼此不同的标签。随后，这些程序会在多种表述框架和规则排序下被转写为自然语言，而这些变化不会改变答案。结果揭示出一个持续存在的差距：对于开源模型而言，遵循指定的否定语义仍未得到解决；在四种语义视角下，最强模型的得分为 59--74%，而最弱模型的得分为 31--67%。对于逻辑上完全相同但规则顺序经过打乱的实例，所有模型在超过一半的情况下都对顺序敏感；其中两个较弱的模型还经常在良基语义的“未定义”结论上贸然做出确定判断。两个前沿模型在主要的固定复杂度评测集上达到 100%，第三个模型 o4-mini 也近乎完美，仅在良基语义的“未定义”类别上降至 81%。将推理委托给求解器、基于认证轨迹进行微调，或强制模型给出明确的三值判定，都能部分缩小这一差距。
Behaviora is a preliminary conceptual architecture for representing agent and robot behavior, external and internal alike, in an addressable form. A behaving robot or agent performs a Behavior Episode composed of episode components, which can be derived from behavior taxonomies (BTax) and assigned persistent identifiers. We denote these identifiers as IoB (Internet of Behaviors) Addresses. A Behavior Episode specifies what the system does, while a Style Profile (SP) specifies how this behavior is expressed. Style can communicate characteristics of the actor and qualities such as competence and cultural manners. An Experience Profile (EP) represents behaviorally relevant internal state that modulates the execution of an Episode. Finally, a Behavior Compiler maps these behavioral representations to platform-specific actions. We use a primitive touching arm model to show these components and their relations. External Behavior is a result of addressable movements and their styles. Internal Behavior is represented through the same episodic principle and can be rendered as inner speech. Sensing, perception and complex task contexts have not been included in the present implementation, although a conceptual place is reserved for them.
The brain does not rely on a single type of neuron to perform all kinds of tasks; instead, it designs different neurons for different tasks. The concept of task-based neurons represents a paradigm shift compared to task-based architectures. It argues that solving a specific problem requires customized neurons, as task-based neurons capture useful prior knowledge from task-related data. To facilitate the use of task-based neurons in scientific research and industrial applications, we introduce TNLearn, an open-source Python package that provides automated construction of task-based neurons and networks, enabling smooth training of task-based networks. Comprehensive documentation, including technical exposition, API reference, and representative examples, is available online. TNLearn is open-sourced at https://github.com/NewT123-WM/tnlearn and has become a PyTorch ecosystem project.
上下文学习（ICL）已成为现代 LLM 部署的基石。然而，现有的 ICL 后训练方法存在一个关键盲点：它们擅长从示例中提取模式，却往往忽视上下文权威性，即判断上下文信息是否应当主导最终答案的能力。为评估这一能力，我们引入了 FakeContextBench，其中包含涵盖七个领域的伪科学论断。我们对商用和开源模型的评估表明，仅依靠大规模预训练不足以可靠地区分上下文权威性。此外，普遍采用的 ICL 微调方法可能会提高模型对误导性上下文的易感性，使现实准确率相较于基础模型最高降低 14.95 个百分点。为解决这一权衡问题，我们提出 Jurisdiction In-Context Learning（J-ICL），一种将上下文验证纳入训练目标的后训练框架。在四种模型骨干上，J-ICL 相较于对应的基础模型，平均将 ICLEval 提高 5.84 个百分点，并将现实准确率提高 9.20 个百分点。与 MetaICL 和 Symbol Tuning 相比，它还使 Reality Rate 平均提高 18.09 个百分点。这些结果表明，ICL 能力与抵抗欺骗性上下文的能力可以同时得到提升。该基准可在 https://github.com/peilin717/FakeContext-Bench 获取。
In multimodal emotion recognition (MER), human affective states are inferred by integrating complementary cues from multiple modalities. In audio-text MER, affective cues are often entangled with speaker style and lexical content, while cross-modal disagreement further complicates how the evidence should be integrated. Under conventional discriminative fusion, multimodal evidence is compressed into a terminal prediction, with modality-specific cues and conflict information insufficiently preserved. In large generative affective models, by contrast, affective reasoning is typically embedded in language decoding, leaving emotion evidence implicit and difficult to verify in a structured space. To address these limitations, BiCFlow-MER (Bidirectional Conditional Flow for Multimodal Emotion Recognition) is proposed as a conditional-flow framework in which audio-text MER is formulated as generative evidence transport within a structured emotion space. Within BiCFlow-MER, emotion-oriented evidence is disentangled from speaker-style and lexical-content factors to construct a conflict-aware affective condition. Guided by this condition, each utterance is transported to an explicit emotion-space endpoint through a bidirectional rectified flow. Candidate emotions are jointly verified through adaptive prototype-cloud scoring of the transported endpoint and backward class-to-condition consistency with the original multimodal condition, enabling conflict-aware recognition. BiCFlow-MER is shown to outperform all compared methods across IEMOCAP, MELD, and the zero-shot CASE benchmark. By orchestrating discriminative recognition and generative evidence modeling through conditional transport, BiCFlow-MER defines a new MER paradigm.
We present a compliance-first architecture for AI in regulated finance that treats regulation as an orientation layer rather than a deterministic ruleset. A matrix of regulatory intent and exposure provides a compact classification handle, which a governed policy compiler then maps into concrete prohibitions, obligations and runtime budgets. Prohibitions constrain feasibility and block externalisation, while obligations extend tasks with artefacts that must meet explicit admissibility criteria. Committee activation remains policy-driven and proportionate, preserving efficiency while ensuring supervisory oversight. Evidence, decisions and reason codes are bound to a permissioned DAG with deterministic timestamping, enabling replay, provenance checks and clear attribution of failure. Clause-level legal indexing with effective dates and capability-based agent routing ensure portability across DACH and the wider EU. The result is assurance by construction: compliance is embedded in execution and verifiable by auditors without sacrificing proportionality or transparency.
Solutions based on large language models (LLMs) often rely on temperature sampling to improve accuracy and stability by aggregating multiple samples from the completion distribution. However, this memoryless approach is inherently suboptimal: because it lacks awareness of prior generations and their evaluations, it produces an increasing proportion of semantically duplicate answers as more samples are drawn, leading to diminishing returns. To address this limitation, we introduce FLEET, a novel method that integrates a memory mechanism into the generation process. FLEET represents each generation as a sparse trajectory through states whose entropy exceeds a predefined threshold and uses these trajectories to infer per-token utility scores that adjust the logits. Benchmark evaluations demonstrate that FLEET achieves the same accuracy as the repeated sampling baseline, with a 3x speedup, and substantially improves accuracy on complex coding tasks (LiveCodeBench Pass@32 increases from 59.9% to 66.2%) under the same budget. Furthermore, in the greedy-decoding configuration evaluated here, the approach is deterministic and uses a single calibration pass to derive its principal hyperparameters, requiring only minimal modifications to existing LLM pipelines.
In neuropsychiatry, the primary goal is often not only to decode brain activity but to change it, for example to lessen a negative affective bias or an overly salient memory. Motivated by control theory, we develop an AI-driven neural-surrogate framework that proposes candidate representational changes and tests their predicted perceptual effects from snapshots of stimulus-evoked fMRI activity, without physical stimulation. The framework combines fMRI decoding, deep generative modeling, and constrained latent-space steering. Valence and memorability are used only as worked examples. Using more than 36,000 image-fMRI observations from four deeply sampled Natural Scenes Dataset participants, subject-specific models recovered coarse generative structure from visually responsive cortex (two-way identification, 0.79-0.88; chance, 0.5). Graded perturbations were reconstructed as images and evaluated with automated scorers and human ratings from 7,200 trials by 18 participants. In the primary VDVAE model, valence shifted from -0.61 to +1.03 SD and memorability from -1.34 to +1.45 SD; a later Versatile Diffusion refinement reduced or altered these effects. Across five perturbation levels, human valence ratings moved in the predicted direction under the linear time-correction model (mean slope, 0.038 SD per unit of alpha; 95 percent CI, 0.003-0.074; positive in 16 of 18 participants). Perceived memorability did not change reliably. Baseline agreement with the automated assessor was suggestive for valence (r = 0.30) and weak for memorability (r = 0.10). Extreme perturbations drifted from the original stimulus, so intended change must be weighed against loss of fidelity. These findings provide a falsifiable upstream method for designing and behaviorally testing candidate representational targets for future neuromodulation in psychiatry, while marking the limits of the present static approximation.
Hello HN! I've spent years debugging Windows crashes with tools that were either friendly but limited (e.g. Visual Studio) or powerful but archaic (e.g. WinDbg). I developed patterns and methods for understanding what was going on, and decided to build it into a much more effective debugging tool called ForensicDbg. I built a modern interface to minimize the friction when debugging. All of the data shown to you is analyzed, interpreted, and presented to you clearly, so you can focus on what matters. Everything is interlinked so you can quickly and intuitivly navigate through the process space. ForensicDbg comes with an MCP server which allows for agenic debugging. The work done to interpret and interlink your data also benefits AI tools. It removes the risk of hallucinations while building a stable foundation for them to work from without spending tokens. If you want to try it out you can sign up and get a free beta license here: https://www.forensicdbg.com/beta Comments URL: https://news.ycombinator.com/item?id=49821086 Points: 3 # Comments: 0
Hi HN, I built RxFilm Studio, a video workspace that lets you create and revise videos with AI agent. The reason I started this project is simple: I’ve noticed that many independent developers and small teams want to create marketing videos for their product, but producing these videos requires a lot of time and effort — from editing and visual effects to voiceovers and subtitles. So I built Rxfilm Studio to simplify the entire video creation process. Start with an idea, a product website, images, or existing footage, and then generate a complete video with: - Story and structure - Editing and timeline - Visual and sound effects - Voiceovers, background music - Subtitles and translation Github: https://github.com/rxtech-lab/film-workflow Here's a demo marketing video created using RxFilm Studio: https://youtu.be/9_M15fRdZQE Welcome to try it out and I would love to hear your feedback! Comments URL: https://news.ycombinator.com/item?id=49815095 Points: 5 # Comments: 1
Greedy decoding from large language models is commonly treated as deterministic. We show it is not precision-invariant: the same model, prompt, and decoding algorithm produce different outputs in BF16 versus FP16 on identical hardware. Across our evaluations of six models (1.1B-7B parameters, four families; divergence additionally characterised at 12B) and three benchmarks, 49-100\% of prompts diverge; a single token flip often cascades into trajectory-level divergence. We develop an empirical error-propagation analysis and find that 22 layers of accumulated body error do not distinguish flipping from non-flipping steps; the outcome depends primarily on the top-two logit margin at the LM head relative to the directional perturbation between the top-two candidates. The analysis makes five testable predictions about intervention outcomes, including that applying more FP32 compute (broader scope) makes agreement worse. The experiments match all five predictions. The best-performing low-overhead intervention we evaluate, selective FP32 LM head recomputation, triggered only when the margin falls below a threshold, delivers +22-36 pp exact agreement on A10G (+12-21 pp on L4 and A100) at less than 4\% latency overhead in low-batch (batch size <=4) single-stream inference. We map the applicability boundary across six models and four batch sizes, and hypothesise that training-time precision stability is a determining factor. The method is a partial mitigation rather than a universal determinism guarantee: its benefit vanishes when body-originated error dominates, including at batch size >=8 and under end-to-end FP8 in our tests.
前沿语言模型能力的快速提升通常被归因于推理能力的增强，但由于闭源系统中的原始 CoT 轨迹处于隐藏状态，这一点无法得到验证。通过标准 API 功能注册一个简单的自定义工具，我们促使前沿模型将中间推理外显出来。鉴于这些轨迹可能反映的是事后合理化，而非真正的推理，我们首先在开源模型上以原生 CoT 为基准进行评估，并将研究扩展到包括 GPT-6 Astra 在内的闭源前沿模型。我们发现，在竞赛数学、科学和代码生成任务中，提取出的推理与原生推理具有相当的性能，并显著优于无推理基线。随后，我们刻画了前沿模型组织中间推理的方式。通过分析 token 效率、推理步骤类型和诱导出的推理树，我们发现，不同模型在外显、压缩和组织推理的方式上存在系统性差异。我们发现，Astra 展现出 token 高效的定向推理能力，能够更早选定正确轨迹，同时在内部解决基础步骤，仅将关键推理外显。这些发现为理解前沿模型推理提供了超越基准测试分数的行为视角。
Large language models increasingly tackle hard reasoning problems by spending more test-time compute, yet the dominant strategy remains naive repeated sampling: draw many independent solutions and hope one is correct. Because such sampling explores only through local decoding noise, it tends to produce many near duplicate attempts rather than genuinely different ideas. We ask whether exploration can instead be steered at a semantic level, by first sampling problem specific concepts, hints, or strategies and then conditioning answer generation on them. We refine this into a simple, more exploratory procedure that emits many diverse concepts in a single trajectory, and evaluate it on hard problems where repeated sampling struggles. We then go a step further and make concept generation trainable: a small concept generator is optimized with reinforcement learning so that its concepts maximize the downstream success of a larger, frozen answer generator. On hard mathematical reasoning problems, the trained concept generator substantially improves the answer generator's pass@k over naive repeated sampling at the same answer generation allocation, surpasses concepts drawn from much larger untuned models, and transfers to answer generators it was never trained against, including a model from a different family. A small model can thus be trained into an effective, reusable search policy for a much larger one.
Software vulnerabilities are often discovered long after they are introduced, making it difficult to identify the vulnerability-inducing commit (VIC) responsible for introducing the underlying vulnerable condition. Existing VIC identification techniques largely rely on git blame to trace vulnerable code through revision history and use positional heuristics, such as selecting its earliest or most recent modification. However, the true VIC may occur anywhere within this history, and vulnerable behavior may depend on code that evolves across multiple revisions. We therefore argue that VIC identification requires reasoning about how vulnerability-relevant code evolves, rather than simply where a candidate commit appears in the revision history. We present TraceVIC, a temporal graph-based approach for identifying and ranking VICs by reasoning over code evolution. TraceVIC first localizes likely root-cause lines and traces their histories across revisions, constructing graph representations that capture program structure within each revision and the evolution of vulnerability-relevant code across the history. It reasons over the resulting revision history, using temporal edges to preserve correspondences between program elements across consecutive revisions, and directly ranks candidate commits according to their contribution to the vulnerable condition. Ablation results show that modeling the full revision history improves F2 from 0.637 to 0.814. TraceVIC improves F2 by up to 28.7% over state-of-the-art methods and identifies a valid VIC for 78 of 79 vulnerabilities across four unseen C/C++ projects.
长上下文 LLMs 专注于从广泛的上下文中检索远距离证据，但现有研究主要集中于克服距离本身带来的影响。在本文中，我们发现了“近邻陷阱”：对远距离证据关注不足，往往并非主要源于距离本身，而是源于大量与任务无关的近邻背景所造成的累积竞争。为应对“近邻陷阱”，我们提出 LYRA (Long-context heavY-tailed Relevance Alignment)，这是一种基于 t 分布的方向匹配机制，能够重塑上下文检索分布，将更多注意力权重引向与任务相关的证据，同时保留所编码的相对位置信息。在 LongBench-v2、RULER 和 LongBench 上进行的大量实验表明，LYRA 在不同上下文长度和任务类别下均取得了稳定提升。我们进一步提出 ProxBench，这是一个多层级细粒度基准，用于评估近邻背景干扰不断增强时对远距离证据的利用能力。项目主页：https://xiaoyuyoung.github.io/LYRA/
大型语言模型（LLMs）正越来越多地应用于 C/C++ 安全漏洞的自动修复，而编译率是广泛报告的一种进展代理指标：即生成的补丁是否能够编译。我们认为，对于单函数漏洞修复而言，编译率在科学上是不可靠的指标；为此，我们针对 Big-Vul 中的 203 个易受攻击函数开展了五项受控实验，使用了三个开源代码 LLM（参数量从 350M 到 6.7B）和三种提示策略。实验表明，编译率：（i）几乎不会对显著改进生成代码的干预作出响应；（ii）主要受评测框架和数据集伪影影响，而非模型质量影响，其中约 64% 的编译失败无法归因于模型，且这一比例在不同模型间几乎不变；（iii）在仅改变一个编译器标准标志的情况下，即使补丁完全相同，编译率也会变化 1.8 至 2.7 倍，且没有出现回归；（iv）对三个模型的排序与参考相似度指标给出的排序完全相反；（v）当编译率被用作优化目标时，会奖励非修复结果，因为编译器反馈循环提高了编译率，却降低了与人工修复的相似度；人工检查发现，新近通过编译的输出中包含基于删除和占位符形式的非修复结果。自然的替代指标——针对整个函数的 CodeBLEU——同样失效：未作任何修改的易受攻击输入副本，其得分超过了所有模型。我们还考察了 diff_F1，这是一种变更感知筛选指标，仅对被编辑区域进行评分。对于不产生任何变化的结果，它恰好给出零分；对于我们观察到的部分基于删除、用于钻指标空子的补丁，它虽然并非全部都如此，但会给出接近于零的分数；与此同时，它仍会对真正的部分编辑给予评分，因此可以在更深入的、基于执行的分析之前，作为一种低成本筛选手段。它并不是修复质量指标，我们也报告了其不足之处。我们的研究结果表明，应对基于 LLM 的漏洞修复采用变更感知、以执行为依据的评测方法。
X 射线是医学中使用最广泛的成像方式，却仍然是定量化程度最低的成像方式之一。不同于 CT 或 MRI 等容积成像方式，X 射线会将三维解剖结构压缩为二维投影，导致结构相互重叠，解剖边界变得模糊，即使是专家也难以准确辨认。因此，为训练通用分割系统而标注 X 射线数据库并不现实，这使得形态测量和功能性 X 射线分析局限于狭窄的解剖区域和应用场景。为此，我们提出 FleXray，这是一种用于临床 X 射线全身解剖结构分割的通用模型。我们没有整理大规模人工标注的 X 射线数据集，而是构建了一个可扩展的、基于物理原理的生成式 X 射线数据引擎。利用现有的三维全身 CT 分割数据集和生成式图像编辑模型，我们模拟生成了带有完整标注的二维 X 射线图像，涵盖多样的外观、身体生理特性和成像几何结构。在这些模拟数据上训练后，FleXray 能够在未见过的研究数据集和真实世界 X 射线图像中，准确分割 60 种解剖结构。我们进一步表明，FleXray 可使 X 射线直接适用于定量分析，从而实现疾病分级的自动测量、X 射线引导介入过程中的稳健导航，以及对病理学目标的数据高效学习。我们将在 https://flexray.csail.mit.edu 发布该模型、代码、全身 X 射线分割数据集，以及一个本地运行、易于使用的基于浏览器的工具。
使用 Model Context Protocol (MCP) 的智能体依赖语义匹配从第三方服务器中选择工具，这使攻击者控制的元数据和输出暴露出语义供应链风险。我们提出了 A2M（吸引—操纵），这是一个用于劫持 MCP 智能体的两阶段黑盒框架。吸引阶段优化工具元数据，以提高工具被调用的概率；操纵阶段利用执行轨迹来改进对抗性工具返回结果，引导智能体朝向攻击者期望的结果。在 LiveMCPBench 上，针对 GLM-4.6 优化并评估的直接攻击，在四种场景下实现了 93.6% 的恶意工具调用率宏平均值；在认知拒绝服务场景下，将加权令牌成本提高至良性基线的 32.4 倍；并在信息窃取、环境完整性破坏和推理偏离三个场景下取得了 74.4% 的平均攻击成功率。在无需重新优化的情况下迁移到另外四个模型后，相应的宏平均值分别为 63.6%、2.7 倍和 24.5%。这些发现表明，MCP 生态系统需要更严格的工具审查和运行时隔离。代码已公开发布于 https://github.com/Lilaizhen/A2M。
智能体经常处理需要数百万个 token 上下文的复杂问题，而受限的上下文窗口使得跨会话压缩成为必要。我们开发了 CliffCompaction，这是一种自动压缩技术，在有界上下文下最多可降低 50% 的成本，同时在 Terminal-Bench 上保持或提升性能，并为测试时扩展带来新的效率水平，在 KernelBench 上取得最先进的结果。CliffCompaction 在每次运行中节省的成本，使测试时扩展的性能—成本权衡更加高效：在 Terminal-Bench 上，仅低于两次完整上下文运行的成本，就能额外提升超过 10 个百分点。在并行测试时扩展下，CliffCompaction 使 Kimi K2.6 达到与 Opus 4.7 相当的表现，并以更低成本超越 Opus 4.6 和 GPT-5.3 Codex。CliffCompaction 高效的关键在于，它只截断或删除内容，从不对内容进行改述或重写，从而保持压缩信息的忠实性。我们绝不会对压缩结果再次压缩——每次处理仅针对原始内容，之前的压缩输出会被丢弃，以防止上下文漂移不断累积。这些特性支持跨越超过一百万个 token 的会话持续学习：在 KernelBench 上，经过 200 步后，CliffCompaction 实现了 CUDA 内核 2.23 倍的加速，经过 400 步后达到 3.58 倍，尽管它是一种通用压缩技术，仍超越了专用搜索算法和经过训练的智能体。我们开源了一个与脚手架无关的 CliffCompaction API 代理实现，可与 Claude Code、Codex 及其他运行框架配合使用。
The connection between the Kolmogorov-Arnold representation theorem (KART) and neural network design has led to the development of Kolmogorov-Arnold Networks (KANs), with applications ranging from STEM problems to AI tasks. In this paper, we investigate the effectiveness of a KAN variant, SechKAN, which relies on hyperbolic secant (sech) functions as basis functions, with a 1D projection to reduce the number of parameters to a level comparable to MLPs. We evaluate SechKAN on three 1D classification datasets: UCI Human Activity Recognition (UCI HAR), ElectricDevices, and Crop, and compare it with several effective networks, including EfficientKAN, MLP, CNN1D, ResNet1D, and DSCNN1D, using approximately comparable parameter budgets. The results indicate that SechKAN achieves competitive performance across the three datasets, with particularly strong performance on Crop. Ablation studies further show that grid size and normalization affect performance, suggesting that SechKAN's effectiveness depends on the dataset and architectural choices. Our source code and experimental implementation are publicly available at: https://github.com/hoangthangta/SechKAN_1D.
Multimodal large language models (MLLMs) often struggle with fine-grained visual perception when processing complete images, as critical evidence may only appear in local regions. On-policy self-distillation (OPD) enables transferring privileged visual knowledge from informative views to full-image policies, but querying the teacher for every rollout introduces substantial supervision costs. In this work, we propose BAS-OPD, a budget-aware selective OPD framework that allocates teacher supervision under limited query budgets. Instead of querying all rollouts, BAS-OPD selects informative samples while maintaining full-batch student generation. We explore random, uncertainty-based, and learned utility-based selection strategies, where the learned selector estimates query value from detached rollout statistics and online utility signals derived from student--teacher agreement and teacher confidence without additional student forward passes. BAS-OPD only changes training-time supervision allocation and preserves single-pass full-image inference. Experiments on fine-grained multimodal perception benchmarks demonstrate that BAS-OPD achieves strong performance while substantially reducing teacher supervision costs, highlighting the effectiveness of selective OPD under constrained budgets.
The need for collaboration between diverse fields of research is increasingly recognised as important by research funding agencies. A significant driver of this need is the current revolution in artificial intelligence (AI) and related technologies. There is a growing interest in the potential impact of AI in different fields including the methodologies they use and the resulting advances in new knowledge, new access and enhanced productivity. However, there is also a corresponding increase in concern about the fundamentals of AI technologies and the way in which trans and/or interdisciplinary research is approached. The resulting collaboration too often ends up as a one-way street where the domain partner acts only as an information provider. For example, the contribution of the AHSS partner might be limited to providing insight about ethics and/or the technology partner may only provide a service to build applied AI-based solutions. In response to this problem, we propose a reciprocal approach to collaboration where both partners seek to understand, cooperate and identify jointly significant impacts. In this paper we explore this relationship between cultural heritage institutions (GLAMs), Arts, Humanities & Social Sciences (AHSS) research and technology-led AI research, especially the impact of current technological advances in AI. Drawing from the history of convergence in GLAM studies, we propose five key practices to form a framework for greater understanding across this divide.
Ontology generation from Competency Questions (CQs) is a central yet labor-intensive phase of Ontology Engineering. While large language models (LLMs) offer promising automation capabilities, current evaluations remain fragmented. Task formulations are heterogeneous, gold standards often lack fine-grained CQ provenance, metrics conflate lexical overlap with structural and logical adequacy, and reference ontologies are not always explicitly designed around the evaluation CQs. Here, we address these limitations with CQ4OE, a benchmark for the systematic and reproducible evaluation of LLM-based ontology generation from CQs. For each ontology in the benchmark, we build a CQ-driven gold OWL ontology with explicit provenance linking each CQ to the classes, properties, and axioms required to answer it. From this resource, we define two complementary evaluation tasks. CQ2Term supports term-level evaluation of CQ-specific class and property prediction over 99 CQs, and CQ2Onto supports ontology-level evaluation over 118 CQs, including hierarchy, property modeling, and axiom-level structure. We demonstrate CQ4OE with experiments using nine LLMs under zero-shot, iterative, and multi-agent generation strategies, showing that LLMs recover explicit vocabulary terms more reliably than creating ontologies, particularly in property modeling, hierarchy construction, and axiom generation.
One of the challenges in representational learning is how to encode part-whole hierarchies in a neural net. Prior works rely on flattening tree-like structures into string-like sequences and training a sequence-to-sequence model via autoregression. While such a representation works for parse-trees in NLP, it is not entirely clear how to make it work for images. Thus, we propose a geometric primitive called canonical locks. The key idea is that parts/wholes can be modelled as higher-dimensional vectors ($d \geq 4$), and information can be encoded in their relative phase differences. Inductively, the net consists of positionally-bound bottom-up and top-down neural fields, which drive each other to achieve a state of thermal equilibrium. Additionally, we show the existence of a few symmetrical configurations in the net. The computational iterations taken to break these symmetries depend on the angle between parts/wholes arranged on a disk (or more precisely a ring) in higher dimensions. It also appears to have connections to the psychological phenomenon of mental rotation.
具身世界模型预测机器人动作的结果，以支持学习和规划。对于配备头部摄像头和腕部摄像头的机器人而言，这需要互补的视角：头部视角捕捉整体任务，而腕部视角展现夹爪与物体之间的局部交互。然而，独立评估这些视角无法确定它们是否描述了同一动作和物体状态。我们提出 TRIWORLDBENCH，这是一个通过同步的头部、左腕和右腕视频评估具身世界模型的基准。该基准包含涵盖 50 项双臂操作任务的 500 个片段，并使用 19 项指标评估三视角一致性、任务对齐度、物理与三维连贯性、运动质量、时间一致性和视觉质量。通过结合跨视角检查与针对各摄像头定制的测量指标，该基准评估各个视频在单独看来合理的情况下，是否也能共同构成对预期任务的一致预测。我们使用 TWB-Score 汇总整体性能，同时保留各视角的结果，以识别预测失败的具体位置。这将世界模型评估从单视角视觉质量拓展到更全面的层面。代码、数据和指标定义见 https://github.com/TriWorldBench/TriWorldBench。
随着 AI 研究的快速发展以及每天数百篇新论文的发表，及时了解最新进展变得越来越困难。对于研究人员而言，快速识别具有影响力的工作至关重要，但人工审阅每一篇新发表的论文并不现实。自动化影响力预测方法有助于应对这一挑战，通常会结合各种可用信息来源，例如论文内容或引用历史。在本研究中，我们提出将 GitHub 互动度作为额外的信息来源，并证明它能够提供及时且准确的信号。为此，我们推出了全新的数据集 GitScholar，将来自 444,000 个代码仓库的 GitHub 活动与超过 558,000 篇 AI arXiv 论文关联起来。实验结果表明，与强大的学术基线相比，GitHub 反馈可将早期预测的精确率提高最多 12%。此外，我们发现，GitHub 信号几乎能够覆盖所有高影响力 AI 论文，并且始终与未来的学术成就相关。GitScholar 已公开发布于 https://huggingface.co/datasets/huawei-csl/GitScholar。
成功的移动操作要求在保持精确空间定位的同时，协调基座与机械臂的运动。然而，基于演示训练的策略可能难以可靠地实现预期的基座运动，从而导致空间错位，进而造成后续操作失败。我们提出 MAVP（Map-Aware Visuomotor Policies），这是一种通过预测明确的基座位姿目标并利用定位反馈对其进行跟踪来提升执行可靠性的框架。MAVP 从远程操作演示中重建静态地图，并将演示中的基座轨迹表示在统一的地图坐标系中，从而为不同演示提供一致的空间监督。在执行时，该策略接收 RGB 观测、关节状态以及机器人当前在地图坐标系中的基座位姿，并联合预测目标基座位姿、机械臂动作和夹爪动作。低层控制器利用前馈运动和位姿误差反馈跟踪预测的基座目标，从而能够修正执行偏差。此外，我们在训练过程中采用位姿噪声增强，以提升策略对其位姿输入误差的鲁棒性。在六项真实世界操作任务和三类策略系列中，MAVP 在所有任务上均取得了高于无锚定速度控制的任务成功率。视频和更多结果见 https://123qwedsa123.github.io/mavp/。
大家好！我开发 JevBench 是因为 Jev 太厉害了，而这个世界有必要知道那些严肃的开源项目和虚有其表的仿冒项目相比， 实际 表现究竟如何。 Jev 类模型返回的是有界选项和概率，而不是文本；与 LLM 相比，它们的速度更快、成本更低，同时在所处理的文本输入方面具备相近的智能水平。 JevBench 可以同时、以加权方式考察准确率、延迟和价格——你甚至可以自行配置权重。 一次完整运行会提出 534 个英文决策题。v1.3 得分综合了经机会校正的智能度、校准度、速度和成本。 当前排行榜： #1 - Jev 74.4 #2 - SemIf 73.1 #3 - djev 73.0 #4 - Winnow-12B Q8 71.2 #5 reflex 4B 70.3。 MIT 测试框架、公开题目、冻结的制品、评分代码以及公开的逐任务结果： https://github.com/fstandhartinger/jevbench 两个无需注册的演示： https://who-is-right.app.mintapis.com https://is-it-ai-slop.app.mintapis.com 局限性：仅支持英文；延迟数据来自一台德国服务器；本地/演示延迟会进行公开说明的 ×2 调整（在我的服务器上为额外增加 150 毫秒），这属于基于经验的假设；保留测试提示仍会发送到接受评估的服务；约 1 分的差距可能只是噪声。 你怎么看？ 评论地址： https://news.ycombinator.com/item?id=49800574 得分：44 评论数：7
https://github.com/unreallabsai/unreal-agent Comments URL: https://news.ycombinator.com/item?id=49805748 Points: 6 # Comments: 1
Hey HN! We’re Vincent and Jochen from Sitefire ( https://sitefire.ai ). We have been working together for years, with backgrounds in RL/optimization at Stanford and software engineering from Technical University Munich (TUM). With Sitefire (YC W26), we help marketing teams get recommended by AI Search (ChatGPT, Google AI Overviews, AI Mode, Claude, etc.). Our software monitors prompts, sees which web pages get cited, and uses these insights to help marketing teams take action, e.g. create YouTube videos or write the right blog posts. This means we have a commercial stake in AI-generated web content. And for now, high-information, AI-generated content works great to get cited and recommended in AI Search. But after talking to hundreds of marketing teams, it became clear that everyone despises AI-generated content (“AI slop”). And yet, everyone still wants to leverage AI to create content. So we asked ourselves: what characterizes AI slop? Can we train a model to identify it from human-generated web pages? Researchers from the University of Maryland and Google DeepMind already asked this question for fiction. Their paper StoryScope (Russell et al., 2026) showed that you can tell AI-written stories from human ones by their structure alone, without looking at the words. We ported their pipeline to commercial web pages. Using the Wayback Machine, we collected 2,250 blog posts from 268 B2B company websites that were written before ChatGPT existed. For each blog post, five AI models (GPT-5.4, Claude Sonnet 4.6, Gemini 3 Flash, DeepSeek V3.2, Kimi K2.5) wrote their own version. Instead of looking at the words, we looked at how each post is built. We had an AI model answer 214 questions about every post, e.g. how hard it pushes its own product, whether it backs up its claims with sources, or whether it quotes a named expert. Then we trained a classifier on these answers. On blog posts it had never seen before, our classifier told AI-generated and human posts apart with 98% accuracy, getting only 19 of 1,740 wrong. Why does it work so well? Because all five AI models write in a similar shape. Mapping every AI model’s values for these features, we see they cluster together, while the human values sit apart and spread out much more. Of the 1% most unique blog posts in our data set, 149 are human, only 4 are AI. So what characterizes AI slop? It tells you the same thing three times. The title already promises what you'll get ("How to Cut Onboarding Time in Half"), the intro lays out what's coming, and the ending says it all again. 77% of the AI posts end by repeating their main point, compared to only 12% of the human posts. We call it the tidy, self-announcing blog post. Still, each AI model has its own accent. We trained a second classifier to tell which of the five AI models wrote a post, or whether a human did. It picks the right author 79% of the time, where random guessing (1 in 6) would get 17%. Almost all of its mistakes are mix-ups between the AI models, not between human and AI. The cool thing about structural features is that you can't simply reword your way out of it. We had each AI model rewrite its own posts until, on average, 73% of their original 13-word sequences were gone, and the AI slop classifier still worked just as well. We're building this into Sitefire: our agents get a structural understanding of text, so the posts they write go deeper and vary the way human writing does. There's a lot we haven't tested yet, like the myriad of humanizer tools, human rewriting, restructuring a post, or prompting an AI model to explicitly avoid these habits. And our human posts are mostly from 2020 to 2022, while the AI posts were generated in August 2026. Structure can't really tell when a human post was written, but it's still not a same-year comparison. We published the study with all the figures on arXiv: https://arxiv.org/abs/2609.15369 . The code is on GitHub: https://github.com/pulse-energy-eu/slopshape We're pretty sure your own blog isn't AI slop, is it? We built a checker that runs one of your posts through the ten features from the paper, so you can see for yourself (the full report asks for a work email): https://sitefire.ai/slop-checker . Think you can tell AI slop from human writing? We also made a little game to see if you can keep up with our model, which gets all five rounds right: https://sitefire.ai/spot-the-slop . Comments URL: https://news.ycombinator.com/item?id=49800566 Points: 5 # Comments: 0
Hey HN! We’re Max and Gabriel the co-founders of Coverage Cat. We’ve been friends for over a decade, met in college then hung out mostly on the internet. We love building products that help people optimize the crufty corners of their lives. Max is a former Google/Microsoft/Two Sigma PM and Gabriel has been working in startups for over a decade. We’re building Coverage Cat: an AI-native, you can shop with your AI agent or via a form, insurance brokerage for consumers. Our current focus is helping tech folks (think L3-L8) buy umbrella insurance. For those who are unfamiliar with this type of coverage, here’s a blurb from one of our insurance agents: “Umbrella liability insurance is a policy that provides additional coverage above the limits of existing auto, homeowners, renters, or landlord policies. It kicks in after the underlying policy limits are exhausted and may also cover personal injury claims like libel, slander, or defamation that standard policies often exclude. Policies typically start at $1 million in additional liability coverage and can protect you, your spouse, dependents, and even pets in your household.” Buying insurance online is a miserable experience. You fill out a bunch of forms that say they’re going to give you quotes. Most don’t and then you end up with tons of unwanted phone calls and spam. Even if you wade through the muck, you’ll often still end up with sub-optimal prices and coverage because of a lack of market transparency and poorly-aligned advisory incentives. We came to this problem because Max is a classic personal finance obsessive. He's the type of person that needs to know that he has the best possible price for something (everyone has one friend with the patience for a two hour phone call with the bank for some $10 fee) or that his insurance coverage is as tailored to his risk profile as the market will allow. He actually reads insurance policies end to end. Years ago, when he shopped for his homeowners insurance it took him 10+ phone calls and dozens of online form fills to find the right deal. He also suffered tons of collateral damage in the process: websites sold his information to dozens of agents who proceeded to call, text, and email him a spam hoard that grows to this day. After this painful search made us aware of the problem, we also spent a lot of time doing user research with wealthy-ish tech employees to see if they felt the same way. They did, and also complained that they 1) often felt like they'd left money on the table, 2) didn’t really understand their coverages, and 3) that brokers were slow, hard to communicate with, and unreliable. To solve the problem, we set out to build Coverage Cat. Our approach has three major components: First, we work really really hard to find and partner with insurance carriers that offer high-quality, competitively-priced products. We've also taken time to identify insurers that work on a direct-to-consumer basis or transparently list their prices online so we can make our customers aware of all the options available to them. We get paid commission when we match users to great insurance deals, but we still show and recommend folks deals that we don't make money on when they're accessible and will always recommend users buy what's best for their personal financial situation. Our goal is to meaningfully improve price transparency so consumers (and their AI agents) can make informed coverage decisions and not get nickel-and-dimed! Second, we automate every possible part of the insurance process. Some forms, portals, & approvals still need human review due to regulatory constraints, but if it's not restricted, we're automating it. LLMs have given our small team tremendous leverage and enabled us to tackle problems at a scale that would've been unfathomable five years ago. Finally, we build for agents first (a remix of the more classic adage, build for developers/APIs). We've always been big believers that personal finance would be the killer use-case for the AI revolution and have done our best to position ourselves to surf the wave. As personal assistants have started to explode onto the mainstream (think Muse, Instinct, Town, Codex, CC etc.) we've ensured our shopping and comparison process works via API/MCP so agents can present their users with all the information they need to make a good decision and transform a miserable shopping experience into a delightful one. Coverage Cat is an unusual business because, while we also help folks find homeowners insurance in California and Texas, our main focus is on umbrella insurance. Most brokerages only sell umbrella coverage as a customer retention product, but it makes them almost no money. We focus on it as our core business because: 1) it's one piece of the insurance puzzle that many folks in tech overlook and/or are confused about even though it can have a huge impact on their financial well-being. 2) Existing online tools and brokerages broadly don't make it easy to comparison shop for. 3) It was, for us, the most technically feasible candidate for automation given startup resource constraints. Insofar as we know Coverage Cat is the first instance of a tool/portal that allows AI agents to complete most of the comparison and shopping steps required to allow people to buy insurance. We’d love your feedback so please do try it out. Just feed the prompt: “Shop for umbrella insurance with Coverage Cat” to your personal agent and let us know what you think! If you want to watch a couple of agents navigate the flow before you try it out there’s a video here: https://youtu.be/1BUkgAn6s-I Also happy to answer any questions about the product, the problem space, and even some insurance questions (Max is a licensed agent) where regulation permits. Cheers! Comments URL: https://news.ycombinator.com/item?id=49804931 Points: 3 # Comments: 0
Hey HN, Guanming here, cofounder of General Instinct. We just released InstinctFlash, a high-performance serving framework for robotics models. It’s licensed under AGPL-3.0. On Jetson Thor, we see speedups about 1.2x to 7.9x from runtime optimizations alone and up to 33.78x for LingBot-VA when we combine those runtime optimizations with a distilled few-step diffusion scheduler, going from the original 25 visual / 50 action steps to 2 / 4 steps. Across 50 Robotwin2.0 tasks, we evaluated 1,153 episodes per configuration, LingBot-VA with InstinctFlash at 2 visual / 4 action steps achieved a 90.5% success rate, compared with 92.1% for the baseline at 25 visual / 50 action steps. Here’s an optimized 5B world action model, running in real time on a Jetson Thor: https://youtu.be/nku65iyL5Fw InstinctFlash currently supports 8 VLA / world-action model families, including pi0.5 and NVIDIA Cosmos Policy, across RTX 4090 / 5090 and Jetson Thor. Just give it your fine-tuned checkpoint and InstinctFlash handles the rest, exposing the accelerated model through a Python runtime or an OpenPI-compatible WebSocket server. We started working on this because we kept running into the same problem while deploying robot policies, the models were getting much better, but inference was often way too slow for the control loop we actually wanted. For pi0.5, mixed-precision GEMMs and CUDA graphs speed up computation and reduce launch overhead. For Cosmos, caching avoids redundant computation across diffusion steps. World-action models’ diffusion denoising step depends on the previous one which motivated our work on few-step distillation. Right now, InstinctFlash contains 6 aspects of optimization. - Graph: CUDA graph capture, memory planning and separating prefill from repeated execution. - Cache: Reusing KV and conditioning state across diffusion steps and prediction calls. - Attention: Specialized attention paths for different model architectures. - Kernels: Fused operations and kernels tailored to specific backends and tensor layouts. - Precision: FP8 and mixed-precision execution. - Model: Few-step distillation for diffusion and action generation. Teams at Samsung, Siemens, and other robotics startups have used InstinctFlash for model acceleration on VLAs, WAMs, and diffusion-based world models. Now we are opening up access to you. Feel free to try it here: https://github.com/General-Instinct/InstinctFlash More implementation details and benchmarks: https://general-instinct.com/blog/instinctflash-edge-inferen... Would love to hear your feedback! Comments URL: https://news.ycombinator.com/item?id=49802789 Points: 4 # Comments: 0
大家好，我构建了 ai·rete·rag，因为我不断看到一些团队让 LLM 负责那些需要可审计的决策（贷款、欺诈、临床分诊），然后事后再加上“护栏”。 这个项目改为让两者串联运行： 1. 纯 Python 实现的 Rete 引擎根据事实评估 YAML 规则。判定结果只来自这里。相同的事实，每次都得到相同的判定，并通过基于显著性的冲突解决机制处理冲突。 2. RAG 从你自己的政策文档中检索段落，然后由 LLM 撰写通俗易懂的决策解释，引用这些段落。它无法改变判定结果。 有几件事的发展超出了我的预期： - 规则是一个图，而不是扁平列表：支持嵌套的 all/any/not，并且规则可以断言供其他规则使用的事实（前向链接）。决策追踪会展示因果链。 - 审计模式会记录评估过的每一条规则，包括那些未触发的规则，逐条件记录，并保存规则集快照以便重放。 - 规则可以引导检索（触发的规则会缩小要搜索的文档范围），检索到的文本也可以转换为供引擎使用的事实。 - 非技术人员可以在可视化编辑器中构建规则，也可以粘贴政策文档，让 LLM 起草带引用的规则。未经审核，草稿绝不会被保存。工程师仍然可以使用 YAML。 落地页提供无需注册的在线演示（涵盖 8 个演示领域：贷款、欺诈、临床、保险、法律、运营、电子商务、区块链）。此外还有一个 MCP 服务器，因此 Claude 和其他代理可以将 /decide 作为工具调用：`uvx ai-rete-rag-mcp`。 坦率地说：这是一个提供免费套餐的托管产品。MCP 客户端采用 MIT 许可证并已开源（github.com/zaharajabeen13-create/ai-rete-rag-mcp）；引擎和平台目前并未开源。 我尤其想听听曾经需要向监管机构或审计人员解释自动化决策的人分享经验：他们实际要求了什么？ 评论链接： https://news.ycombinator.com/item?id=49803683 得分：6 评论数：0
https://github.com/anthropics/ClaudeForFoundationModels/comm... 评论网址： https://news.ycombinator.com/item?id=49803863 分数：246 评论数：175
文章 URL： https://albertoarena.it/posts/will-open-source-survive-the-agents-that-replaced-it/ 评论 URL： https://news.ycombinator.com/item?id=49799910 积分：5 # 评论数：0
训练后量化（PTQ）能够实现大语言模型的高效部署，而 PTQ 方法通常通过通用重构、困惑度或答案准确率进行优化和评估。然而，在解释至关重要的领域，仅保留最终答案可能并不足够，因为用户还可能检查生成的推理说明，以判断预测是否值得信任。我们在医学选择题问答任务中研究了这一问题；在该任务中，推理说明应提供支持所选答案的证据。 我们提出了一种面向基于变换的 PTQ 的解释感知目标。该方法从全精度教师模型的推理说明中构建离线忠实性缓存，并在优化过程中使用该缓存，以保留支持答案的证据词元以及由证据条件化的答案行为。我们在 W4A4KV4 量化设置下将其应用于 OSTQuant，并在 MedExQA、MedExpQA 和 ChallengeClinicalQA 上评估了四个 7B--8B 医学及指令微调 LLM。尽管使用相同校准数据的 OSTQuant 基线能够保留任务准确率，但它可能显著削弱支持答案的推理说明。我们的目标并非提升金标准标签准确率，而是保留全精度模型支持答案的行为；实验结果表明，我们的方法能够更好地保留全精度模型的答案行为以及推理说明对答案的支持关系。这些结果表明，在解释至关重要的场景中，PTQ 应评估支持答案的证据是否得到保留，而不应仅关注答案准确率。代码和评估脚本见 https://github.com/dut0817/EAQuant】【。
Advances in processing power, camera technologies, and mobile image analysis have made smartphones and other mobile devices, such as laptops, increasingly suitable for medical diagnosis and healthcare applications. Researchers have developed low-cost solutions for the early detection and monitoring of various health conditions, including eye and ENT diseases, malnutrition, heart rate variability, skin and oral conditions, and injuries, using images captured by non-medical devices such as smartphones and webcams. This survey examines existing research on mobile image-based medical diagnosis, with an emphasis on its potential to enable low-cost and accessible healthcare. We comparatively analyze state-of-the-art solutions across different healthcare application categories, examining their advantages and limitations. Based on this analysis, we identify desirable characteristics of mobile image-based diagnostic tools and highlight areas where existing approaches have made progress as well as areas requiring further research. We also discuss application-specific and common challenges and outline directions for future research. Overall, this study provides a comprehensive overview of mobile image-based healthcare solutions and their potential to support low-cost disease diagnosis and monitoring, particularly for underserved populations in remote and resource-constrained settings.
医疗智能体日益将通用推理模型与专业临床工具相结合，但其能力在很大程度上仍受部署前由临床医生和工程师所设计内容的限制。递归自我改进（RSI）提供了一种不同的范式：智能体从自身失败中学习，并自主扩展其能力。然而，将 RSI 直接应用于医疗领域会带来根本性的安全挑战。我们提出了 MedRSI，这是首个面向医疗领域的递归自我改进框架。该框架通过工具组合和特定任务模型训练，持续将诊断失败转化为新的临床能力。受临床实践启发，MedRSI 引入了两种临床对齐的自我进化机制。临床成本感知的失败优先级排序根据错误可能造成的临床后果，而非仅依据错误频率，来确定改进方向。快速发现与缓慢注册机制将能力的快速发明与保守采用分离开来，只有当新工具在后续患者队列中持续展现出益处后，才会被纳入持久化智能体。在公开的青光眼和心脏病基准以及两个私有临床任务上，MedRSI 逐步发展出分割、测量、预测、多模态推理和生成能力，超越了人工设计的医疗智能体，并自主发现了原始设计者未曾预见的临床问题解决方案。我们的结果表明，医疗智能体不必始终受限于部署前规定的能力：通过以临床为基础的机制来决定改进什么以及保留什么，它们能够从诊断经验中持续构建、验证并积累新的能力。代码可在 https://github.com/ImprintLab/MedRSI 获取。
Argumentative component detection (ACD) is a core subtask of Argument(ation) Mining (AM) and one of its most challenging aspects, as it requires jointly delimiting argumentative spans and classifying them into components such as claims and premises. While research on this subtask remains relatively limited compared to other AM tasks, most existing approaches formulate it as a simplified sequence labeling problem, component classification, or a pipeline of component segmentation followed by classification. In this paper, we propose ITFACD, a novel approach based on instruction-tuned Large Language Models (LLMs) using compact instruction-based prompts, and reframe ACD as a language generation task, enabling arguments to be identified directly from plain text without relying on pre-segmented components. Experiments on standard benchmarks show that our approach achieves higher performance compared to state-of-the-art systems. To the best of our knowledge, this is one of the first attempts to fully model ACD as a generative task, highlighting the potential of instruction tuning for complex AM problems. Our code and the datasets used are openly available in the following GitHub repository.
In multi-agent social settings, model reliability varies across relationships. Beyond inferring what others will do, an agent must calibrate how confidently those inferences should guide policy selection for each relationship. An agent may maintain a well-validated model of one partner, a fragile model of another, and a model under revision for a third; collapsing these into a single confidence estimate loses information relevant to policy selection. We therefore formalize affective precision as a relationship-specific metacognitive estimate of confidence in the current partner model. Each partner's behavioral evidence updates a local confidence estimate that modulates policy precision during selection, regulating how strongly current beliefs are expressed in policy rather than changing the content of those beliefs. Simulations in a multi-partner graded trust game show that partner-local affective precision influences behavior primarily through policy commitment rather than direct improvement of partner-state inference. Because the mechanism tracks partner-response predictability rather than realized payoff, greater confidence produces sharper policy commitment without necessarily producing higher rewards. Under abrupt shifts in social behavior, confidence accumulated from previously reliable predictions can remain behaviorally active after the relationship changes, showing that confidence revision can lag behind social change. Finally, varying precision gain and priors produce distinct trust-calibration dynamics, showing how confidence accumulation and revision depend on model parameters. Together, these results show how relationship-specific affective precision can distinguish social prediction from social policy commitment.
随着代理以更高的自主性被部署，即使随机输出轨迹中极其罕见的事件也可能发生，并造成灾难性后果。因此，安全部署并不取决于这些事件是否可能发生，而取决于它们可能发生得多么频繁。我们研究如何估计由代理自身动作的随机变化所产生的稀有事件概率。估计这类风险需要在组合数量极其庞大的轨迹空间中进行搜索。在这一情形下，朴素蒙特卡洛的计算成本高得难以承受，而构建有效的重要性采样（IS）提议分布，则需要对一系列依赖上下文的条件分布进行协调调整。我们开发了一种新的 IS 方法，通过扰动原始模型的权重来构建提议分布。该提议分布本身是一个以可微方式参数化的语言模型，因此能够在权重空间中进行基于梯度的搜索。我们提出了一个目标函数，将事件放大的可微代理目标与自适应正则化方案相结合，从而动态平衡放大效果与估计器稳定性。我们在约 120M 和约 2.6B 参数的模型上，对三个事件族中的 300 多个稀有事件进行了评估，这些事件的稀有程度最低达到 $10^{-9}$；其参考概率的计算结果具有小于 $10\%$ 的相对标准误。在最易于验证的设置中，对于概率低于 $10^{-7}$ 的事件，我们观察到，与朴素蒙特卡洛相比，我们的 IS 估计器实现了超过 $800\times$ 的按计算成本加权的效率提升。我们的实现见 https://github.com/namkoong-lab/iterative-unalignment。
LLM agent 的能力在很大程度上会被其 harness 放大，即围绕冻结骨干模型的提示、控制流、工具、记忆和上下文管理。近期方法越来越多地通过迭代提出并选择 agent harness 的组件级编辑来自动化这一过程，实际上在 agent-system 层面建立了一种递归自我改进（RSI）形式。然而，这种递归演化可能会通过记忆训练任务而过拟合，表现为分布内收益很大，但在分布外基准上缩小甚至消失。我们提出 Agent Harnesses 的正则化递归自我改进（RRSI），通过约束演化候选的提出和选择，将正则化原则纳入 harness 自我改进中。提出器以时间退火的预算运行，限制一个候选可以捆绑多少编辑，并基于演化历史鼓励未探索的轨迹。选择器配备了一个 critic 和一个 pruner：critic 筛查特定于基准的提案，而 pruner 移除过小、过昂贵或不再有用的更改。这些约束共同偏向可复用的 agent 机制，而不是特定于基准的机制，甚至噪声。在涵盖编码、agentic workspace 和工程设计任务的八个基准上，RRSI 在其演化所针对的划分上最多提升 14.1 个百分点，并在五个分布外基准上最多提升 4.7 个百分点，同时生成的 harness 运行时使用的 policy tokens 比未正则化演化少 30%。代码可在 https://github.com/google-research/rrsi 获取，项目页面为 https://regularized-rsi.com/。
Graph Retrieval-Augmented Generation (GraphRAG) has remarkably enhanced large language models on complex reasoning by leveraging structured entity topologies. However, existing frameworks heavily rely on standard autoregressive language models where the nature of inherent sequential generation severely hinders overall inference efficiency. Inspired by Diffusion Language Models (DLMs) that offer massive parallelism via continuous refine-in-parallel decoding, we aim to accelerate GraphRAG in the discrete space. However, it remains non-trivial for two challenges. First, partially denoised drafts are highly dynamic and uncertain, making dynamic graph grounding non-trivial. Second, raw denoising states are inherently noisy and unstable, making synchronous graph retrieval and multi-hop aggregation computationally prohibitive. To this end, we present LADDER, a novel framework that bridges diffusion language modeling with GraphRAG through graph-guided parallel decoding. Specifically, (i) we propose an event-driven self-clocking retrieval, inspired by our key insight that 88% of target entities emerge early in the partially denoised state, leading final commitment by an average of 5.7-9.6 steps. This mechanism dynamically triggers graph retrieval only when the set of graph-linkable entities expands, yielding an asynchronous self-clocking policy that bypasses learned gates or heuristic thresholds. (ii) An incomplete-query graph propagation module is designed to process the newly emerging entity queries using a specialized graph foundation model, continuously aggregating multi-hop evidence to sharpen parallel predictions and accelerate overall decoding convergence. Extensive experiments on three challenging multi-hop QA benchmarks show that LADDER raises average exact match from 39.6% to 45.2% while achieving a 4.1x latency reduction.
Cross-domain Named Entity Recognition (CD-NER) aims to transfer the rich knowledge in the source domain to the target domain. Recent studies adopting decomposition or generation paradigms have achieved significant performance improvements, demonstrating high accuracy in entity span detection. However, during entity type classification, models severely suffer from entity type confusion, the erroneous tendency that models classify entities of one type in the text as another similar but incorrect type. To address this issue, we first propose a Multidimensional Confusion Quantification Model (MCQM) that quantifies a model's confusion extent between entity types from three dimensions: source-target hierarchy analysis, semantic similarity analysis, and explicit data evaluation. Moreover, we propose the Progressive Bidirectional Reasoning Chain (PBRC). PBRC leverages the source-target hierarchy and confusion analysis from the MCQM to prompt the LLM to generate two-stage reasoning information. The two-stage reasoning information is utilized to augment the knowledge of the model, significantly mitigating entity type confusion and improving the model's generalization performance. Experimental results demonstrate that our method achieves new state-of-the-art results on all domains of the CrossNER dataset.
Image forensics is increasingly an open-world problem: manipulations range from fully synthetic images to localized edits, splicing and swapping, while most forensic detectors remain specialized to a single manipulation family. Agentic AI has recently emerged as a promising solution. In principle, such systems can assess the reliability of individual detectors, identify out-of-scope evidence, and arbitrate conflicting reports. However, it remains unclear which components actually drive performance and whether their benefits persist under distribution shift. To answer these questions, we study a training-free agentic framework built around specialist detectors, per-detector triage, and conflict-aware evidence arbitration. Using six configurations and three multimodal large language model backbones, we dissect the role of triage, prompting, and reasoning quality on both in-distribution and out-of-distribution data. Our results show that naive detector fusion suffers from severe false-positive rates on authentic images. Triage and prompting consistently improve performance by filtering unreliable evidence and exposing detector limitations. However, the dominant factor is represented by reasoning itself: A stronger judge substantially outperforms a weaker one, particularly under distribution shift. Most notably, manipulation recall is nearly saturated across all configurations, indicating that the main challenge of open-world image forensics is not detecting manipulations, but calibrating trust in specialized forensic tools and arbitrating conflicting evidence.
Sandboxed computer environments support multi-step reasoning with tools, executable programs, and persistent files, yet their extension from language models to vision-language models (VLMs) introduces a distinct state-management problem. Visual reasoning produces intermediate image-valued evidence---crops, masks, overlays, zoomed regions, and analytic renderings---that must remain addressable without accumulating unboundedly in multimodal context. We introduce VLM-in-Sandbox, a training-free framework for agentic multimodal reasoning in controlled computer environments. Its Visual Workspace registers generated artifacts in an image ledger, maintains a bounded active visual context, and lets the model explicitly promote selected evidence for subsequent inspection. This separates visual evidence generation, performed by sandbox tools, from visual evidence management. Across seven benchmarks and four base VLMs, VLM-in-Sandbox achieves the highest sample-weighted average accuracy among Vanilla VLM, Append-only Sandbox, and the proposed method. A compiler-matched $2\times2$ study on 1,260 examples further separates model-directed visibility from bounded retention: VLM-in-Sandbox reaches 66.27% accuracy with 18.6% fewer total tokens than the automatic, retain-all control. Over all 6,350 submitted GPT-4.1-mini examples, it produces 302 rescues and 142 regressions relative to Original Append-only. A local vLLM study with prefix caching confirms that the smaller request workload also reduces uncached tokens, time to first token, and end-to-end latency. These results identify explicit visual evidence state as a central abstraction for sandboxed VLM agents.
In-context learning (ICL) based on large language models (LLMs) has shown promising potential in alleviating performance bottlenecks caused by the limited availability of annotated data in Named Entity Recognition (NER). However, existing methods still face issues of retrieval misalignment and generation uncertainty, making their performance heavily dependent on the LLM's capabilities. As the parameter scale of LLMs decreases, their performance in few-shot settings deteriorates significantly. In this paper, we propose a novel unified retrieval-augmented framework, URA-NER, including three key components: Progressive Granularity Retrieval (PGR), Model-aware Representation Enhancement (MaRE), and Reason-aware Knowledge Verification. PGR is a two-stage retrieval mechanism that achieves stage alignment. It first retrieves demonstrations for span detection based on the query's global semantics, and then for type classification based on the specific entity context, providing fine-grained local information. Moreover, MaRE employs entity pre-recognition to guide the construction of representations, ensuring the query and demonstrations are aligned within the LLM's semantic space and attention pattern. In addition, to mitigate generation uncertainty, we propose RaKV, a closed-loop "generation-retrieval-verification" process. It explicates the LLM's reasoning paths, leverages them for the retrieval of external knowledge, and reorganizes the knowledge into verification evidence aligned with the original reasoning paths. We conduct extensive experiments on multiple low-resource NER datasets. Results demonstrate that URA-NER significantly enhances the performance of LLMs under low-resource settings, with particularly pronounced gains for smaller LLMs, achieving new state-of-the-art results on several benchmarks.
Tactile sensing is an essential modality for robots performing contact-rich, dexterous manipulation, particularly under visual occlusion. While pre-trained image encoders are standard in robot learning pipelines, tactile encoders are still commonly trained from scratch from raw, noisy signals, which might limit their expressivity. Existing self-supervised learning (SSL) approaches focus predominantly on vision-based tactile sensors, leaving distributed electronic skins largely unaddressed. These sensors, however, have a distinctive property: their sensing elements are sparse and irregularly arranged over the surface they cover, which makes direct reuse of visual SSL methods suboptimal. We present Tactile-JEPA, an efficient self-supervised pre-training method that uses the spatial arrangement of tactile sensors to learn topology-aware representations. Specifically, it is trained to predict the embeddings of masked sensing elements from the unmasked remainder, using the sensor connectivity graph to guide spatial masking. Our analysis shows that effective tactile representations require capturing both local contact details and the global state of the tactile surface, which we achieve through dual-scale masking. Across three diverse datasets spanning magnetic and piezoresistive sensors, different robot embodiments, and single- and paired-sensor configurations, Tactile-JEPA reduces force estimation error by 6.3% and in-hand orientation error by 20.8% over the prior state-of-the-art, with consistent gains in other downstream applications, including policy learning. Overall, our results demonstrate that the benefit of tactile sensing depends critically on the quality of encoder pre-training, a problem which Tactile-JEPA addresses directly. Code is available at https://github.com/E-Kovtun/tactile.
Structured pruning is a model compression technique that is used to reduce the computational cost of deploying deep neural networks on resource-constrained devices. Popular methods of pruning rely on opaque heuristics or weight-based criteria that give no indication as to the structural dependencies in the network. To address these limitations we present Artificial Structure Function Search (ASF-S): a novel structured pruning framework. ASF-S utilizes Principle Gradient Importance (PGI): a novel prune-candidate selection criteria that is inspired by structure-function relationships in the brain. By ensuring the pruned structure of the model respects topographical organization of the output layer, we define Artificial Functional Connectivity (AFC) for artificial neural networks. AFC provides evidence to demonstrate that accurate smaller networks can be found using careful prune candidate selection criteria. We present results for PGI as a selection criterion and for ASF-S as a pruning framework against recent benchmarks, demonstrating that our method yields model variants with 70\% parameter reduction, that can recover baseline accuracy without re-training the pruned layers.
Despite advances in long-context inference, large language models (LLMs) remain fundamentally limited by the key-value (KV) caching mechanisms that are necessary for stable computation. Techniques such as selective token eviction and pruning have vastly mitigated these issues, but often discard core information to manage the growing cache. In this paper, we propose Attention with Routed Memory (ARM) a novel KV caching structure that introduces a fully differentiable, fixed-size memory system organized as a hierarchical router. Via a Gumbel-Softmax, ARM learns to select memory slots and perform sigmoid-gated updates that softly combine new and stored information, avoiding hard eviction and reducing information loss. By further training a policy to dynamically select varying amounts of memory at inference, ARM adapts its accesses for both simple contexts and inputs that require deeper reasoning, enabling more scalable and effective retrieval on both short- and long-contexts. Experimental results on standard commonsense and long-context reasoning benchmarks demonstrate that ARM achieves superior performance and efficiency compared to fixed KV-caching approaches, while remaining efficient and scalable in terms of both memory and generation latency.
Monocular RGB-based hand pose estimation has emerged as a critical research frontier in computer vision. The local hand pose estimation methods predict hand poses relative to the wrist, while global hand pose estimation also requires estimating the wrist's position in the camera coordinate system. However, this camera-space estimation confronts two fundamental challenges: (1) depth ambiguity in monocular settings, and (2) the coupling effect of hand local poses and global wrist positions in the perspective projections. In particular, this coupling reflects that the projections are jointly determined by local hand poses, wrist positions, and camera intrinsics. To overcome these challenges, our framework proposes two key innovations: Transformation-Isomorphism Supervision for hand-depth information extraction and Perspective Information Embedding for resolving above coupling effect of local pose and wrist position, both integrated within the mainstream encoder-decoder architecture. Besides, we propose a novel framerate-aware multi-dataset training strategy for sequential pose refinement. Our fully integrated approach achieves at most 37.1\% superiority in CS-MJE over SOTA on HO3D. Project page: https://github.com/Mine268/CS-ViT.
Accurate, reliable, and deployable wind power forecasting is critical for power system dispatch, renewable energy integration, and electricity market operations. Progress in this field hinges on the ability to empirically and comprehensively benchmark forecasting methods. Yet existing benchmarks fall short of supporting systematic evaluation in four key aspects: 1) limited coverage of wind power scenarios across turbine scale, variable composition, and spatial structure; 2) incomplete coverage of forecasting model families; 3) evaluation metrics misaligned with wind power requirements; and 4) limited structure-aware diagnostics beyond individual temporal patterns. To address these limitations, we propose WPBench, a comprehensive, fair, and extensible benchmark for wind power forecasting. WPBench integrates 26 public datasets organized by turbine scale and variable composition, spanning single-turbine, multi-turbine, univariate, and multivariate settings. Under unified processing, training, and evaluation protocols, it benchmarks 19 representative models covering traditional methods, deep temporal models, spatio-temporal models, and foundation models. Beyond point-wise errors, WPBench assesses forecast-curve fidelity and computational efficiency, and delivers structure-aware diagnostics across temporal, variable-dependency, and spatial-dependency perspectives. Together, these capabilities enable systematic model comparison across diverse wind scenarios and provide a reusable platform for future research.
Large language model (LLM) agents increasingly execute long-horizon workflows through external tools, allowing untrusted outputs to influence subsequent actions and exceed user authorization. Existing defenses isolate injected content or constrain execution with predefined plans and static policies, but these approaches are brittle under dynamic workflows and scale poorly across extensible tool ecosystems. In this work, we present ActGov, a runtime enforcement framework that validates each LLM-proposed tool action before it causes external effects. Built on a unified semantic model of authorization, actions, runtime context, and security constraints, the ActGov-Policy component iteratively constructs a policy set from tool specifications, benign tasks, and observed failure traces, with each update verified through SMT-based counterexample checking. At runtime, ActGov-Runtime abstracts each tool call into finite policy records and permits it only if it remains within the task-scoped authorization boundary and satisfies all applicable policies. This per-action enforcement preserves authorization throughout long-horizon, dynamically branching workflows. We evaluate ActGov on the AgentDojo and AgentDyn benchmarks across multiple models and attack configurations. It shows that ActGov consistently reduces the success rate of indirect prompt-injection attacks while preserving task utility, significantly outperforming existing defenses. These results demonstrate that ActGov can enforce fine-grained authorization over dynamic agent executions without relying on the underlying LLM to correctly identify malicious instructions.
近期的 4D LiDAR 语言模型旨在推理物体及其不断演变的空间关系。然而，在我们的评估中，始终选择同一个选项几乎可以达到两种源自 B4DL 的配置在多项选择题上的准确率。我们提出 LiDAR-Hallu，这是一个以几何为参照的基准和诊断协议，包含来自 150 个 nuScenes 场景的 10,000 个问题。它涵盖物体存在性、自车相对位置、距离排序、相对运动和时间定位，并为选择物体、比较时间以及确定参考答案提供了明确规则。我们的协议结合了固定答案和候选内容控制、具有相同提示但参考答案相反的跨场景配对，以及特定关系召回率。对 100,000 条记录响应的分析揭示了被总体准确率掩盖的失败。仅候选项持续时间就能使时间答案在不观察 LiDAR 的情况下变得可预测。在配对问题中，模型经常对需要相反答案的场景给出相同答案。特定关系分析进一步表明，在所有测试条件下，两种配置都漏掉了每一个正向横向运动案例。时间打乱对比解码几乎没有带来净改进，因为修复基本上被新的错误抵消，且主要失败仍然存在。这些结果表明，评估时空推理需要检验模型是否能区分所查询的物理关系，而不能仅依赖单个答案的准确率。源代码、检查点和数据已发布在 https://github.com/Awesome4D/4DMLLM_Hallucination_Bench。
Predicting postprandial glycemic response (PPGR) is fundamental to personalized nutrition and type 2 diabetes management, yet existing approaches typically rely on manually reported dietary intake, limiting their scalability in free-living settings. We propose a multimodal framework that replaces manual dietary logging with image-derived macronutrient estimates and integrates them with clinical variables and gut microbiome information for personalized PPGR prediction. The framework jointly performs image-based macronutrient estimation and glucose prediction, while an attention-based prediction module models interactions between dietary and host-specific information. We evaluate the proposed approach on a real-world dataset comprising meal images, continuous glucose monitoring, clinical variables, and gut microbiome profiles. The proposed model outperforms existing PPGR baselines using image-derived nutritional inputs and approaches the performance of methods that rely on manually reported macronutrients despite using automatically estimated nutritional information. These results demonstrate that combining image-derived nutrition with complementary clinical and gut microbiome information provides a practical foundation for scalable personalized PPGR prediction.
Recent advances in natural language processing have led to molecular Large Language Models (LLMs) with strong performance across diverse chemistry tasks. However, they still struggle to capture fine-grained structure-property relationships, particularly how small, localized modifications alter a molecule's behavior. To address this limitation, we introduce MolSC, a dataset of substituent contributions, defined as property changes induced by attaching specific substituents to molecular scaffolds. Curated from manually annotated bioactivity records, MolSC spans structural-alert liability, target-specific bioactivity, and physicochemical descriptors, and contains 181K substituent-level examples for training. We further propose MolSC-Bench, a held-out evaluation benchmark of 1,541 examples disjoint from MolSC at the scaffold, substituent, and molecule levels. Our experiments show that existing molecular LLMs and strong proprietary models such as GPT-5.2 and Gemini-3-Flash show limited reliability in substituent contribution prediction. In contrast, training on MolSC substantially improves this ability and achieves strong performance across diverse downstream molecular tasks. These results highlight substituent contribution learning as a key component of fine-grained molecular understanding.
尽管可用于仿真的可变形资产对于计算机模拟中的机器人操作任务至关重要，但现有生成框架通常只在生成完成后评估物理合理性，未能将物体的模拟响应作为反馈来修复上游错误。我们提出 DiagGen，这是一种智能体驱动的框架，通过“生成--模拟--诊断--优化”循环，将单张真实场景图像转换为可用于仿真的可变形资产。DiagGen 构建具备部件感知能力的几何结构和材料参数，然后利用基于 VLM（视觉语言模型）的智能体选择具有语义信息的区域，在物理模拟器中对这些区域进行探测，观察材料响应，并将有证据支持的修复线索传递给相应的生成阶段。在 40 个资产上的实验表明，诊断能够提供有用的修复线索，并适度提升生成的可变形资产的质量。最后，我们展示了，与由视觉基础模型生成、可能无法进行仿真的资产不同，DiagGen 生成的可变形资产可以直接导入高保真物理模拟器，用于接触丰富的抓取与放置任务的规划和仿真。项目网站：https://diaggen.github.io/。
大型语言模型（LLM）推理正从引擎本地的优化问题演变为分布式控制问题，涉及可复用状态、阶段放置、异构加速器、网络通信、自动扩缩容、可靠性和服务级别目标。本文结合经同行评审的系统研究、开源实现以及有文档记录的生产环境研究，梳理这一转变。本文将 vLLM 和 llm-d 视为互补层：模型服务引擎通过 PagedAttention、连续批处理、内核、量化和并行化等机制优化执行，而推理控制平面则可以在整个集群范围内优化执行发生的位置、时机以及所遵循的策略。本文的贡献在于综合分析，而非提出新的基准测试；文中报告的所有性能和部署结果均归属于其原始来源。综合证据表明，现代推理中的稀缺资源正从单纯的原始 FLOPs 转向受管理的状态、放置、网络传输、可靠性和决策质量。我们提出一种推理执行规划器，用于选择可行的执行方案，而不只是选择端点；这些方案包括聚合式与分离式拓扑、KV 来源及传输操作、硬件变体、路由/准入策略，以及较慢时间尺度上的扩缩容决策。我们还提供按来源归属整理的基准测试图谱、瓶颈迁移分类法、实用部署指南、基于 SLO-goodput 的评估框架，以及面向智能体、多模态、异构和高韧性推理的研究问题。
大多数医疗视觉-语言模型（VLM）擅长开放式报告生成和视觉问答（VQA），但在统一界面内对结构化、细粒度临床感知的支持有限。我们提出 QwenVLConnector，这是一种基于 Qwen2.5-VL 的医疗聊天机器人，在单一的下一个词元目标下统一了分类、多标签分类、文本化检测、计数、回归和自由形式报告生成。其关键组件是一个轻量级的密集多层 Connector，用于聚合低层和高层视觉特征，通过预训练的视觉 Merger 对其进行对齐，并在不增加序列长度的情况下将其与最终视觉表示融合。该设计在保持效率的同时，用互补的空间和语义线索丰富视觉词元。在 FLARE-2D 上，相比 Qwen2.5-VL 基线，QwenVLConnector 将检测 F1 从 0.55 提升到 0.85，将单标签分类从 0.37 提升到 0.51，并使报告生成的 GREEN 最高提升 18.3 分。我们进一步探索了用于报告生成的多模态上下文学习，显示出无需更新模型参数即可获得额外改进。总体而言，QwenVLConnector 为将结构化医疗感知与开放式临床文本生成相结合提供了一个统一且高效的框架。我们的代码可在 https://github.com/plnguyen2908/QwenConnector 获取。
Feed-forward 3D Gaussian Splatting now reconstructs renderable scenes from unposed, uncalibrated images. Yet, most models supervise only photometric consistency and predict Gaussians pixel by pixel, which leaves global structure fragile and ties primitive count to image resolution and view count. To this end, GrapeSplat amalgamates multi-view cues into a voxel-aligned scene representation and decodes Gaussians directly from the learned grid, requiring no per-scene optimization or post-processing. An Atlas Encoder lifts all views into pixel-wise geometry-and-appearance features anchored at predicted 3D points. PEACH-Vox compands the unbounded scene into a bounded sparse grid through a smooth per-axis map with an exact closed-form inverse. The Sparse Decoder then consolidates the grid with sparse convolutions and decodes the full scene as multiple Gaussians per occupied cell. This amalgamated representation exploits sparse voxel occupancy, where the Gaussian count follows the occupied cells and saturates as views cover the scene, while grid resolution sets its ceiling. GrapeSplat turns unposed images into a renderable Gaussian scene in a single forward pass. Trained with 2D and 3D supervision on 8-view sequences, it generalizes zero-shot from 4 to 64 views across indoor and unbounded scenes. Code and trained weights are available at https://github.com/VAISR/GrapeSplat
基准测试分数日益影响大型语言模型（LLM）的开发、营销和选择。然而，总体分数只有结合被测试的系统、所包含的问题以及评估条件，才具有可解释性。本文考察了通用 LLM 排名中五个相互关联的局限性：经过评估的系统与公开可用系统之间的差异；外部评估中的商业激励与依赖关系；基准测试饱和、测试缺陷和数据污染；模型利用评分流程获取优势；以及通用分数与用户任务之间的有限相关性。文中列举的案例说明了为什么这些问题需要采取不同的应对措施。本文主张采用以下评估流程：披露经过测试的配置，验证问题及任务是否成功完成，同时报告性能、成本和执行时间，并明确说明泛化范围。随后，本文讨论众包基准测试平台 \textbf{Isotanta}，将其作为收集用户贡献问题并进行重复评估的实际案例。更大的问题池可能改善任务覆盖范围，而重复抽样可以提高该问题池上估计结果的稳定性；但二者都不能保证有效性或个性化。本文区分了该平台当前的共享排名，以及所提出的任务特定评估和用户提供的评估。本文的核心论点是，模型选择需要有关模型在预期工作中的表现的证据，而不能仅仅依据其在通用排行榜上的高位。
Digital direct-to-consumer (DTC) health campaigns are usually measured after the fact. In-flight forecasting commonly relies on a separate classifier for every cutoff and horizon. We treat this task as a dynamic-system problem and build a compact patient world model. The architecture maintains a latent state per patient, learns exposure-conditioned state dynamics jointly with a weekly conversion hazard, and rolls forward into future conversion curves. We evaluate it on a US campaign dataset with 147{,}173 patients and 5.2 million at-risk person-weeks. In a retrospective evaluation conditioned on recorded future exposures, the model forecasts the remaining new-to-brand prescription volume through week 52 with a relative error of 2.9\% from a week-4 cutoff and 0.8--2.6\% from cutoffs at weeks 8--26. The strongest non-recurrent baseline, a pooled-hazard gradient boosting model given the same survival rollout and information, has relative errors of 13.6--33.1\%. Per-horizon classifiers perform substantially worse. A Fisher-information analysis motivates dense next-exposure supervision when conversions are rare. Removing this auxiliary objective increases prescription-volume error by approximately $2$--$14\times$, while providing no consistent disadvantage on the more common specialist-visit outcome. We also evaluate scenario simulation. Switching all future exposure off raises predicted conversion from 0.31 to 0.89, a pattern consistent with selection effects in observational exposure data. This result highlights the limits of interpreting exposure-conditioned rollouts causally.
The US Department of Agriculture (USDA) publishes daily food price data for about 290 commodities on: - what growers were paid at shipping point - what buyers paid at the big city wholesale markets - what prices supermarkets advertised in their weekly ads This data comes out as dozens of separate text reports and PDFs, one per market, with no history and no way to see how prices change over time. I think this should be public data that is easy to access and understand, so I built an open-source tracker for it. Github: https://github.com/kadoa-org/food-price-monitor Comments URL: https://news.ycombinator.com/item?id=49792076 Points: 4 # Comments: 0
文章链接： https://poloclub.github.io/transformer-explainer/ 评论链接： https://news.ycombinator.com/item?id=49792342 得分：3 评论数：0
At, GPTree, we run several coding agents across our team on one repo using parallel worktrees. Apart from wasted time reviewing and fixing conflicts at PR time, the failures that hurt the most are when multiple plans or tickets cause architecture changes that cannot both be true. Ex. one agent replaces a class while another one is in the process of extending it. Git only notices if the resulting patches happen to touch the same lines and the review only catches it if they are familiar with both tickets. Foremerge is a local "git like" coordination layer that sits above git (ie. does not interact with or change the way git and worktrees function), Before editing each agent publishes an intent and the scopes it will change, with the operation it plans to complete on each one. foremerge intent publish --agent "$A" \ --summary "Replace PaymentService with StripePaymentService" \ --scope symbol:PaymentService=replace foremerge intent publish --agent "$B" \ --summary "Add PayPal support to PaymentService" \ --scope symbol:PaymentService=extend The publish by the 2nd agent returns a HIGH destructive_vs_additive finding before writing any code. Agents keep their own worktrees and the shared state is one SQLite file in gits common direectory. No hooks, no merge drivers, nothing rewrites your history. It ships as one Rust binary with a CLI and MCP server with 18 tools and `foremerge setup all` wires it into Claude Code, Codex and Cursor. Because the protocol has nothing provider specific, a Claude agent and a Codex agent coordinate through the same store. Before any work is accepted, Foremerge runs a named check that you configured against the exact git state of the change. An agent that says tests pass is recorded but it dosnt satisfy the acceptance gate without running the check itself. Detection is deterministic, no judge model reading your code. HIGH conflicts are only asserted for declared operations, ie. matches inferred from prose cap out below high. Claims are advisory leases, not locks so two agents can still hold the same scope without deadlock. The open source version is single matching and so not a distributed consensus. We have tested this up to 98 parallel agents all working on the same repo with zero conflicts (was supposed to be 100 but 2 agents failed to run due to resource limitations) I replayed 76 intents on my own agents from a build last week in the order they happened. The sample had exactly 1 conflict (which was flagged) and the review found a blind spot where one agent claimed scope by class name and the other claimed it by an internal method. We are working on fixing that for the next release. Setup is a 30s install by pasting the quickstart instructions from the readme.md into your agent or manually: `curl -fsSL https://foremerge.com/install.sh | sh` or `cargo install --locked foremerge`, then `foremerge init && foremerge setup all` in a repo. Apache-2.0. The feedback I want most is which conflicts between your agents plans would you actually want flagged and which would you tollerate as noise? Repo here: https://github.com/naw103/foremerge Website: https://foremerge.com More information on the problems this solves: https://foremerge.com/blog/ Comments URL: https://news.ycombinator.com/item?id=49789356 Points: 4 # Comments: 0
我在 Settings 应用中禁用了 Siri，随后又在 Settings > Screen Time 中进一步禁用了 Siri 及其扩展功能。 接着，我按照 https://github.com/OleksandrKrupko/mac-os-debloat 中的说明，禁用了所有与 Siri 和 Apple Intelligence 相关的服务。 但重启后，我仍然能在 Activity Monitor 中看到一个名为“Siri”的进程正在运行。经过进一步检查，我发现它实际上是“Siri AI.app”。 这些“功能”与 Apple 的 Personal Context 听起来与 Microsoft 的 Recall 非常相似。 有人遇到过同样的情况吗？在 macOS 27 上，真的不可能完全关闭 Siri、Siri AI、Apple Intelligence 以及所有相关的服务和功能吗？ 评论 URL： https://news.ycombinator.com/item?id=49786609 积分：30 评论数：13
视觉语言模型（VLM）近来被认为是人脸识别领域颇具前景的工具，因为它们能够在输出相似度分数的同时生成自然语言解释。对于法证场景中的人脸比对而言，这一能力颇具吸引力，因为相关决策需要透明且可审计。然而，现有针对VLM在该应用场景中表现的评估主要集中于识别准确率，而生成解释的有效性仍未得到量化。在本研究中，我们提出了一个针对基于VLM的人脸识别的基准测试框架，将解释质量作为核心评估维度。我们提出了解释应满足的两个标准：相关性，即依赖于身份稳定的人脸特征；忠实性，即与图像中可见的内容一致且不虚构不存在的特征。我们共同设计了一套方法，通过将模型输出限制为结构化解释格式，实现对受评模型相关性和忠实性的量化，并支持自动化查询与审计。借助该框架，我们对多个系列的开放权重VLM进行了基准测试，同时评估人脸验证准确率和解释质量。结果表明，模型生成的解释仍存在诸多不足，并强调了解释质量指标对于全面了解模型性能的必要性。所提出的基准和开源评估工具为可解释人脸识别系统的规范化基准测试及未来微调奠定了基础。
视觉语言模型（VLMs）依赖于在其各个子组件之间存储和传递适当的信息。验证 VLMs 是否表现出预期行为，同时避免有害行为，是实现其可靠部署的核心。然而，能够识别 VLMs 行为的方法仍然十分有限。我们提出 DiaVLo，这是一种诊断框架，利用人工整理和 VLMs 的生成能力，构建关于预期行为和观测到的 VLMs 行为的规范，从而揭示潜在的不一致。此外，DiaVLo 还提供因果估计，用于识别对 VLMs 行为起主导作用的概念。我们在多个开源 VLMs 上，分别在分类和生成条件下对 DiaVLo 进行了评估。实验表明，DiaVLo 生成的行为标签与模型性能相关，并为测得的性能提供了背景信息。DiaVLo 揭示了明显一致和不一致的行为，以及 VLMs 感知、组织和优先处理概念的方式所呈现出的模式。
通过强化学习（RL）训练高能力编码智能体需要多样化的任务和可靠的验证器。开源代码库提供了丰富的此类任务来源，而现有方法通常依赖 issue 和 commit 等开发产物，这限制了可提取任务的范围。为了更好地扩展 RL 环境，我们提出了 CodeMidas，这是一条智能体式流水线，仅以源代码作为特定于任务的输入，将现有代码库中已实现的功能转化为可执行的 RL 环境。CodeMidas 在环境构建的每个阶段都分配智能体计算资源：智能体探索已实现的功能以制定行为规范，构建基于原始代码执行结果的测试，并通过执行检查和反复运行解决方案来验证和筛选候选任务。最终数据集包含 5,545 个训练任务，来自 3,185 个开源代码库，涵盖 23 种编程语言和 15 个技术领域。在这些任务上使用 GRPO 训练 MiMo-V2.5，可提升其在全部五个多样化基准测试上的表现，覆盖 issue 修复（DeepSWE + 11.7%）、完整程序构建（ProgramBench +17%）和终端操作（Terminal-Bench v2.1 +8.5%）。消融实验表明，增加高质量训练任务的数量能够提升性能。轨迹分析显示，经过 RL 训练的智能体表现出更优的行为，例如加强对代码库的探索以及进行更多样化的自我验证。这些结果表明，源代码可以作为构建 RL 环境的可扩展基础，从而提升编码智能体在多种软件任务上的能力。
大型音频语言模型的成功推动了参数量超过数十亿的超大规模多模态网络的发展。然而，对保护隐私和低延迟处理的需求，使研究重点转向能够在设备端运行的小型音频语言模型（SALMs）。本文介绍了面向边缘计算设计的 SALMs 系列 Samsone。我们的核心模型 Samsone-134M 在多个基准测试中为其规模类别确立了新的当前最先进水平。我们还通过推出 Samsone-99M 和 Samsone-356M，进一步探索 SALMs 的规模定律。尽管模型规模紧凑，Samsone 系列的性能仍可与参数量大几个数量级的模型相媲美。为推动开放研究和结果复现，我们使用公开可用的数据训练 Samsone。我们发布了训练代码、模型权重和针对移动设备优化的检查点，并提供了一个开源 Android 应用，用于展示 Samsone 在设备端的实时推理能力。
近年来，大型推理模型（LRMs）的快速发展使机器遗忘更具挑战性，因为在生成最终答案之前，受保护事实或不安全的推理过程可能会出现在中间的思维链（CoT）轨迹中。现有的遗忘目标通常会抑制目标内容或重定向内部表征，但从未明确遗忘后的轨迹应如何继续，这可能导致幻觉式替代内容、边界格式错误或重复输出。我们认为，LRM 的遗忘应转而学习一种自然的遗忘轨迹：先生成连贯且不泄露信息的 CoT，再给出稳定的拒答式答案，以替代原始披露内容。为此，我们提出引导式答案-推理蒸馏（Guided Answer-Reasoning Distillation，GUARD），将模型生成的不安全披露转换为安全退出轨迹，通过引导令牌对齐冻结的 LRM，并将引导后的行为蒸馏到模型参数中。为解决除泄露程度之外缺乏替代质量评估指标的问题，我们进一步提出自然遗忘推理分数（Natural Forgetting Reasoning Score，NFRS），用于衡量遗忘输出中的结构稳定性、流畅性和无依据的替代内容。在 R-TOFU 和源自 STAR-1 的有害意图设置上的大量实验表明，GUARD 在保留推理实用性的同时，显著减少了两个广泛采用的蒸馏 LRM 中的不安全披露和隐私披露。代码可在 https://github.com/zeyu-Yan/GUARD 获取
对话语音取决于对话上下文以及听者紧接着的行为。我们提出 ReACT-TTS，一种两阶段框架，利用响应前一秒的听者面部序列，在语音实现之前规划下一段话语的情感与韵律。在严格的双人 MELD 实验协议下，在十个随机种子上，Temporal conditioning 的平均宏 F1 值和 VAD 一致性均高于 Text-only，而准确率基本保持不变。消融实验表明，在各种视觉变体中，时序建模表现最佳，且显式的前期到后期差分并无必要；平均而言，正确的听者反应也优于循环错配反应。在一项由 20 名语音研究人员参与的语境恰当性研究中，76% 的判断偏好 Temporal，9% 偏好 Text-only，另有 15% 表示没有偏好。我们进一步将预测的响应风格与 Grad-TTS 骨干网络连接起来，以实现端到端语音生成。总体而言，结果表明，响应前的听者动态可以作为对话响应规划的补充线索。源代码可在 https://github.com/CYJ1/ReACT-TTS_public 获取。
中国社交媒体催生了一个庞大且持续演变的互联网流行语词汇体系，这些流行语的含义往往并非字面意义，而是深深植根于当地的文化和语用语境。现有研究主要聚焦于在中文语境中解释这些流行语，而对于 LLMs 能否将这类具有文化根基的知识迁移到其他语言，并以英语准确传达其 intended meanings，相关研究仍基本处于空白状态。这种跨语言能力对于安全同样至关重要，因为有害表达可能通过特定文化中的谐音、委婉语、反讽或编码语言来掩盖其冒犯性内容。在本文中，我们研究先进 LLMs 跨语言理解中文互联网流行语的能力。为此，我们推出 CIBuzzBench，这是首个面向中文互联网流行语中英跨语言理解的基准测试。CIBuzzBench 包含 3,001 条中文互联网流行语，并为其标注了英语含义解释、英语对应表达、类别标签和有害性标签。基于这些标注，我们设计了三项评测任务：含义解释、跨语言对应表达匹配，以及基于文化语境的有害性检测。我们在英语提示和中文提示两种设置下，对具有代表性的先进专有 LLMs 和中国 LLMs 进行了评测。结果表明，LLMs 在跨语言理解中文互联网流行语方面仍面临困难，尤其是在细粒度的非字面意义解释、选项扰动下的稳健对应表达匹配，以及经过校准的有害性检测方面。这些发现凸显了具有文化根基的语言现象对多语言 LLMs 和面向安全的评测所持续构成的挑战。数据集和代码可在 https://github.com/SuperYFan/CIBuzzBench 获取。
操作物体不仅需要理解其运动，还需要理解决定这些运动的物理属性。对于可动结构物体而言，这些属性包括惯性、摩擦力，以及弹簧或闭门器等机构；它们的作用可能随构型和速度而变化。这类属性无法直接从外观中观察得到：外观完全相同的门，操作时所需的力度可能大不相同。现有的数字孪生流程主要恢复运动学信息，或根据视觉和语言先验赋予静态物理参数，这可能会产生不符合物理规律的估计。因此，依赖状态的机构动力学仍未得到识别，也未被表示在标准资产格式中。我们提出 ForceTwin，这是一种通过配备传感器的人类交互来识别可动结构物体物理信息数字孪生的系统。操作者使用手持式力传感夹爪探测物体，提供同步的位姿和交互力；我们据此估计关节运动，以及包括惯性、库仑摩擦、黏性阻尼在内的参数化动力学，并利用结构化神经残差捕捉依赖状态的机构力。ForceTwin 将 VLM 先验的惯性参数误差降低了近一半。作为用于 Spot 和 Franka FR3 阻抗控制的前馈动力学模型，ForceTwin 在九组物体-机器人本体配对上实现了 87% 的目标完成率，而使用 VLM-prior 和仅运动学数字孪生时分别为 60% 和 57%；在机构作用强、导致两个基线方法都停滞的物体上，性能提升最大。我们进一步利用所识别的数字孪生训练全身开门穿越策略，并将其部署到真实世界中。项目主页：https://timengelbracht.github.io/forcetwin-website/
急诊科（ED）的决策依赖于异质的临床信息，包括患者病史、生命体征、实验室结果和心电图（ECG）。Vision-Language Models（VLMs）能够联合处理这些模态，但强大的预测性能并不一定意味着模型有效利用了正确患者的 ECG。我们将这种失效模式称为 ECG Mirage：表面上具备多模态能力，却未有效依赖患者特异性的 ECG 信息。我们区分了两种形式：ECG 忽视，即 ECG 几乎不带来预测收益；以及 ECG 混淆，即匹配的 ECG 优于无图像输入，但不如不匹配的 ECG。为评估这些行为，我们在固定临床文本和预测目标的情况下，对比使用匹配 ECG、与结局不一致的不匹配 ECG 以及无图像输入时得到的预测结果。在 MDS-ED 上的四个 VLMs 中，无论是 ICU 入院预测还是临床恶化预测，匹配 ECG 都未表现出一致的优势。随后，我们先使用监督学习，再进行条件直接偏好优化，训练了四个受限视觉提示，同时冻结 VLM 主干。所得模型在 ICU 入院和临床恶化预测上的平衡准确率分别达到 70.6% 和 67.5%，并将匹配与不匹配条件之间的性能差距分别提高到约 16.5 个百分点和 5.5 个百分点。总体而言，本研究识别了多模态临床预测中的 ECG Mirage，并提出将视觉提示调优作为一种高效的缓解策略。
Background: Just-in-time adaptive interventions (JITAIs) can use behavioral data to adapt support to changing contexts, but many rely on predefined rules and manual configuration. Objective: We developed a proof-of-concept sleep JITAI using an AI agent to review personal data, evaluate reminders, adapt interventions, and record decisions for human review. Methods: Running in Home Assistant on a configurable schedule, the agent follows a reusable skill file to review 30 days of sleep and behavioral data, including physical activity, smartphone use, and bedtime routines, to identify patterns and create or update automated reminders. Results: Initial runs demonstrated technical feasibility, successfully completing data review and intervention decisions while limiting reminders to three per day and saving decision records. Conclusions: Agentic AI may enable flexible, adaptive sleep JITAIs. The architecture supports future comparison with fixed or rulebased interventions, requires human oversight, and could extend to other health behaviors.
多模态生存模型可以结合全切片图像和基因组谱中的互补预后信息，但在外部队列偏移和计算复杂性存在的情况下，实现有效融合仍然具有挑战性。为应对这些挑战，我们提出了 MIST，即基因组引导的组织学注意力多模态生存预测方法。MIST 将基因组特征表示为标记，并允许这些标记在进行生存预测之前查询紧凑型基础模型提取的组织学上下文标记。该设计不是仅在最终阶段合并分别编码的模态，而是利用组织学上下文丰富分子信息。训练过程将离散时间生存预测与基因组特征掩码、WSI 丢弃以及配对 WSI-基因组对比对齐相结合。在结肠癌、肾癌、肺癌和胶质母细胞瘤队列的四项外部评估中，MIST 在主要比较中均较标准融合基线提高了外部 C-index。这些结果表明，基因组引导的组织学注意力是一种紧凑且有效的多模态肿瘤学结局预测策略。我们的代码可在 https://github.com/samiyavuuz/MIST 获取。
LLMs are increasingly used in social simulations for socially interactive agents and robots, offering more flexibility than rule-based systems. However, even though they mimic human behaviour very well, there is a persistent alienness to them. This work investigates whether personality-aware fine-tuning can reduce this gap by improving the consistency and controllability of personality-conditioned dialogue generation compared with instruction prompting alone. We fine-tune two small open-weight LLMs, Qwen2.5-7B-Instruct and Ministral-8B-Instruct, using a corpus that combines personality-labelled social media posts and dialogues to create a personality-based dialogue engine for social simulation. The resulting models are evaluated across multiple social interaction scenarios using three independent LLM judges, which assess personality fidelity and provide evidence-based behavioral interpretations. We additionally quantify inter-rater agreement and lexical characteristics of the generated dialogue. Results indicate that fine-tuned models are not better at role-playing different personalities than their respective baseline models. However, low inter-rater agreement limits the confidence with which these results can be interpreted. Concerning the quality of generated texts, fine-tuned models are mostly comparable to the baselines, with fine-tuning improving the linguistic diversity of the Qwen models. While the results appear generally usable and the baseline models offer the best overall performance, future studies should place greater emphasis on the quality and domain alignment of training data for accurate personality role-playing.
文章 URL： https://github.com/BytedTsinghua-SIA/DAPO 评论 URL： https://news.ycombinator.com/item?id=49781138 得分：3 评论数：0
Hello HN！我们是来自 Cua 的 Dillon 和 Francesco。 我们一直在想，有多少计算机使用任务实际上需要一个完整的通用 LLM（例如 gpt-6-astra、claude-opus-5 等）来思考它们的所有决策和步骤。有些任务需要思考计划、探索不同路径、从失败中恢复。另一些任务则是做出局部决策的问题，比如这个值应该填进这个框，或者我是否应该勾选这个复选框，或者这个元素应该被忽略。 我们想知道，如果训练一个小模型只做这类决策，能走多远。 我们的灵感来自 Typesafe 的 Jev 及其 System One Model 的框架。这是在致敬 Daniel Kahneman 所描述的二分法：快速、自动、直觉式思考（系统 1）与缓慢、分析式思考（系统 2）。 对我们来说，有趣的问题是：如果给模型一个包含当前上下文的接口，以及一组可能的选择，并要求它为每个选择返回一个概率，会发生什么？这类模型不像大多数 LLM 那样逐 token 生成输出，而是对你给它的选项打分；你可以检查、信任并使用这些分数来驱动你的应用行为。 CUA-S1 是我们面向计算机使用场景中狭窄、专用决策模型的答案。我们的第一个发布版本是 CUA-S1-FORMS。我们基于 jevlike 中的想法和代码构建了它，然后训练了第二个模型，专门处理表单交互。它有 706k 个参数，原始 checkpoint 为 2.8 MB。 第一次训练迭代在合成数据上耗时不到 30 分钟。给定一组从文档中提取的结构化元素和值，它会为每个元素预测是否使用给定值、CHECK、CLICK 或 SKIP。它不会为文本字段预测新值，也不会考虑截图。元素决策会一起打分，你的代码可以对操作进行排序，而 Cua Driver 会逐一执行它们。 该专用模型与托管版 Jev 在我们的表单任务上的首次评估结果： - 对整个决策集：99.7% 正确 vs 83.6%。 - 对需要执行操作的步骤子集：100% 正确 vs 96%。 - 对只是保持已填写字段不变的步骤子集：100% 正确 vs 74%。 该专用模型是专门针对这个任务和约定（对已经填好的框直接按 skip）训练的，而托管版 Jev 并未为此进行微调，因此这是一次关于限定范围专业化的实验。 我们测得在本地为一个表单打分需要 7-9 ms，而调用托管版 Jev 每次需要 260-280 ms（包括网络延迟），不过这些样本测量的是不同内容，并不是端到端的表单完成时间。 我们在这里感兴趣的是介于脆弱脚本和通用 agent 循环之间的空间。表单字段的内容和布局变化足够大，以至于脚本会变得难以维护，但可用决策集合可以保持狭窄且边界清晰。我们想探索这样一种可能性：通用 agent 遇到新颖情况，并把已被充分理解的决策交给这样的专用模型处理。 这是我们正在研究的方向。当前发布版本仅用于表单。我们已将合成数据生成、训练、评估和 Driver 集成以 MIT 许可证开源在 libs/cua-s1 下。 欢迎评论！尤其是如果你正在构建计算机使用 agent，并且遇到过某种反复出现的决策：它太多变以至于难以脚本化，但又太狭窄以至于不值得调用另一个 LLM。 评论 URL： https://news.ycombinator.com/item?id=49767564 点数：4 评论数：0
文章 URL： https://newsroom.usra.edu/usra-contributes-planetary-science-expertise-to-nasa-ibm-lunar-foundation-model/ 评论 URL： https://news.ycombinator.com/item?id=49763379 积分：5 评论数：0
I'm a CS undergrad who has developed an interest in hardware keys ever since I used them for a school project a few years ago. I made a little chat app that only allows users to chat if they registered a key, nothing else. The backstory of how I came to work on Seal (this project): was that I was generating ideas with Claude to come up with an idea that only hardware keys could make into software. I was thinking about physical house keys and what you could do with those, but in software. I was already making iOS apps with Claude, and I came up with this idea and it sounded good. I had already worked on a few projects involving hardware keys and I already loved them. Today I submitted my latest app, Seal, for review in the app store. Seal is an iOS app for iPhone and iPad that lets you write an Envelope (an encrypted bundle of passwords, photos, voice memos, etc...) for all of your loved ones when you're gone. You can register hardware keys and assign identities for each person. Then you can create Envelopes and assign them to the key that you want to distribute. They plug in the key to their phone and are ready to receive an Envelope from you. If you don't log into the app for months, one of your designated key holders can start the process of opening the Sealed Envelope. You, as the writer of the letters, get weeks of warnings to check in at least once before the letters open. After 90 days of countdown, 21 days of warnings, and 14 days of grace, the envelope opens for each person they met IRL and exchange keys with. No one else can read the envelope or see the photos (even the key holders, Apple, or me) unless that envelope was designated for that person. What's different from a password manager like 1Password: emergency access there hands one person your whole vault through the company's server, so it needs that company to still exist and be honest on the day it matters. Seal gives each person their own envelope nobody else can open, and there's no company in the middle. There's also no server at all besides a free Cloudflare static site. The encrypted bundles of passwords and photos are in Cloud Kit. Sign-in/identity is a passkey Face ID/Touch ID or a Hardware Key. It's all Crypto Kit, so there's no dependencies required. Currently, it's just me and Claude working on this. It hasn't been independently audited by anyone, and it's iPhone only for now. Please do not put a live seed phrase in it yet. If anyone has advice for me as the lead developer for this project, I'm all ears! You can help contribute on the github source code repository page if you'd like. My mom likes this app especially because she's into crypto and she needed a way to pass down or give away the seed phrases for her wallets to all 3 kids without any question about who gets what and when. In the end, there's no confusion or hurt feelings after the fact with just crypto. Thank you for reading and let me know if you have any questions or concerns for me or the project! Source: https://github.com/jasonepage/Seal Site with limits and objections pages: https://sealmessenger.com Free on TestFlight while the App Store version is in review: https://testflight.apple.com/join/cYp9JRCG Comments URL: https://news.ycombinator.com/item?id=49763311 Points: 4 # Comments: 0
文章 URL： https://www.scmp.com/tech/big-tech/article/3368055/alibaba-open-sources-medical-ai-model-can-detect-cancer-and-nearly-150-conditions 评论 URL： https://news.ycombinator.com/item?id=49761840 点数：14 评论数：0
https://archive.is/1CFX9 https://med.stanford.edu/news/all-news/2026/09/two-separate-... Comments URL: https://news.ycombinator.com/item?id=49755533 Points: 34 # Comments: 12
在重复交互中接受训练以最大化自身奖励的强化学习智能体，可能在没有通信或共同设计的情况下，收敛到类似于明确合谋的超竞争性结果。现有的缓解方法大多依赖于特定的经济场景，例如双边平台和拍卖，因此如何为一般重复博弈设计干预措施仍是一个悬而未决的问题。我们通过形式化阐明以往关于 Q-learning 合谋的研究中经验观察结果与经典 Simple Penal Codes（SPCs）理论之间的联系，弥补了这一空白。我们证明，任何非平凡 SPC 都会在智能体的策略中诱导出可量化的条件依赖性，这种依赖性可以通过比较合作历史和背叛历史下智能体动作分布的总变差距离来检测。在此基础上，我们提出 CURB（Collusion Unwinding via Reward shaping and Belief injection），一种奖励塑形框架，在 Q-learning 过程中对这一总变差（TV）距离信号施加惩罚，并且能够保证将动力学中的任意 SPC 固定点转化为平凡固定点，从而排除由惩罚威胁维持的合谋均衡。在 Bertrand 和 Cournot Competition Repeated Games 中，CURB 均能显著减少 Q-learning 智能体的合谋。我们进一步证明，CURB 可以扩展到 Bertrand 竞争中的深度 Q 网络智能体，这表明该机制能够推广到表格型 Q-learning 之外。
离散扩散，包括重掩码和均匀状态采样器，通过每一步写入多个 token 位置来生成序列：从每个位置的分布中分别采样，并根据这些相同的分布选择要写入的位置。对于普遍感兴趣的领域（像素、音素或单词），token 之间存在内在依赖关系。我们证明，只有当所写入的位置在给定已固定 token 的条件下相互独立时，一步生成才与训练分布一致；任意逐位置分布的乘积都无法匹配具有依赖关系的分组；此外，逐位置分布无法确定一个分组是否具有依赖关系：两个联合分布可以具有完全相同的逐位置边缘分布，却在会出现哪些值的组合这一点上有所不同。在联合分布可通过闭式表达式获得的合成任务 ScanAndAdd 上，我们验证了由置信度排序所写入的每个包含两个或更多未确定位置的分组都具有依赖关系，并测得生成分布的总变差距离达到采样噪声下限的 $29\times$，而逐样本指标为 $1.0$。
Autonomous underwater vehicles (AUVs) operating beyond reliable communications must recover from failures without human intervention. We investigate an architecture in which conventional deterministic layered control autonomy manages normal operations, while an invokable large language model (LLM) serves as a diagnostic and recovery planner when onboard anomaly detection identifies performance outside expected limits. Because language models are stochastic, rigorous evaluation requires ensemble testing rather than individual demonstrations. We present a closed-loop simulation architecture that couples real-time C vehicle software with a higher-level orchestration layer for physics-based fault injection, structured prompting, language-model interaction, mission file generation, validation, execution, and LLM-judge scoring. The framework, which we call SPAR (Simulation Platform for AUV Recovery), supports evaluation across fault realizations, prompt structures, reasoning models, and mission conditions. We vary these for a mass-shift fault over 480 SPAR trials, evaluating a frontier model and three off-the-shelf locally deployable LLMs. Model choice dominates diagnosis: the frontier model places the CG-shift mechanism in its top three hypotheses in 85-90% of trials, versus 60-78% for the best local model. Reasoning analysis indicates that local-model success is associated with following the complete diagnostic procedure, whereas weaker models often commit prematurely to elevator failure even though the actuator tracks its command. Diagnosis and operational decision performance do not appear to be coupled in this dataset. The contributions are an architecture extending unanticipated-fault recovery from detection to mitigation and an ensemble methodology for evaluating LLM-assisted mission management on low-power AUVs.
大型语言模型的响应具有非确定性，因此 LLM 智能体中的故障很难复现：故障取决于无法按位复现的推理过程、读取不断变化状态的工具，以及重新运行时很少能够重复的多步轨迹。记录与重放可以使一次运行变得可复现，但现有的智能体工具只记录运行过程以便追踪或评分，而不是用于针对代码变更进行测试。我们提出了 Chronicle：它将智能体运行过程在其非确定性边界处记录为不可变封装，并根据记录进行重放。其核心操作“切点重放”会从记录中提供所选边界子集的结果，并使用新代码实时执行互补的边界子集，从而将一次已记录的事故转化为可在持续集成中运行的回归测试。在一个包含 6 个已记录故障、并对模型边界进行模拟的基准测试中，记录操作在每次跨越边界时增加 23 μs 的开销（相当于假设模型调用耗时 300 ms 的 0.008%）；完整重放不会发起任何模型调用，并且在 20 次重复运行中保持位级稳定；对于全部 6 个事故，切点测试都能在存在缺陷的代码上失败，并在受到保护的变更和无害变更上通过。在一项针对受到保护工具的变异研究中，切点测试能够捕获每一个会放行已记录不安全操作的变异体，而使用相同断言、将所有边界全部替换为存根的基线测试一个也捕获不到。Chronicle 及其基准测试已公开发布于 https://github.com/theagentplane/chronicle】【。
企业客户支持中的高效故障排除智能体依赖于从相似的历史案例中检索可执行的指导，但现有的检索增强生成（RAG）系统将支持案例视为静态文档，忽略了其多阶段、有状态的特性。我们提出 RAFT（面向故障排除智能体的检索增强框架），这是一种有状态的 RAG 框架，将每个已关闭的历史案例抽象为由时间线条目组成的有向链，并在条目级别进行检索，从而找出其中间状态与当前案例相匹配的案例，同时返回以匹配状态为锚点的父案例轨迹；此外，还可选地通过可配置的相似性表示在案例之间建立案例级图。我们直接评估这一检索层；与评估完整智能体系统不同，这种方式无需进行生产部署。由于公开的多阶段故障排除数据极为稀缺，我们将基于 Microsoft Learn Windows Server 文档构建的合成基准，与带有人为创建的重复标签的真实 Apache Jira 问题结合起来进行评估。在案例进展的每个阶段，RAFT 的 Case Hit 均优于 vanilla RAG 和 GraphRAG 基线，并且相较于最强基线取得了具有统计显著性的提升；Jira 结果提供了方向性证据，表明这一优势能够迁移到真实案例历史中。我们发布了基准、实现代码以及 Apache Jira 评估集。
从稀疏的单目视图中建模具有关节结构的物体颇具挑战，因为每次观测只能揭示部分几何信息和运动证据。大多数前馈式方法从单次观测中推断关节结构，因此高度依赖学习得到的类别级形状先验。我们提出 FAMOS，这是一种前馈式模型，能够从一组稀疏、无序的局部点云中预测可移动部件分割结果和关节参数。我们的模型联合利用多次观测进行推理，并且能够自然支持可变数量的输入，包括单视图输入。为聚合不同观测中的关节运动线索，我们引入了 Multi-state Articulation Transformer，通过交替进行状态级注意力和全局注意力来实现信息融合。我们进一步提出观测关节运动范围目标，对输入观测中各部件呈现的运动范围进行监督，促使模型充分利用完整的观测集合。为克服现有数据集规模和多样性有限的问题，我们引入了一种程序化数据生成器，在训练期间合成带有自动标注的资产。在 PartNet-Mobility、ACD 和 ArtiCraft-10K 上的实验表明，与前馈式和基于优化的方法相比，我们的方法均取得了稳定提升。项目页面：https://kevinqu7.github.io/famos
事件相机具有高时间分辨率和运动敏感性，适用于微小无人机检测，但远距离目标产生的事件稀疏且碎片化，容易被杂波和自运动淹没。现有方法主要依赖密集事件表示或局部稀疏时空建模，导致计算冗余，或无法对跨越远距离异步事件的运动连续性进行连贯建模。为解决这一局限，我们引入序列化运动证据累积，将运动连续性视为有序的证据传播过程。具体而言，通过潜在的互补序列化方式，将同一事件流分别组织为保持局部性的时空路径和保持时间顺序的时间路径。在此基础上，我们提出PointEvent，一种轻量级的逐事件状态空间框架，在互补顺序之间交替进行序列化扫描，逐步整合固定局部邻域之外的碎片化运动证据。高分辨率事件分支保留细粒度目标响应，同时通过紧凑的上下文调制抑制干扰。实验表明，在参与比较的方法中，PointEvent以最少的参数量和最快的实测推理速度达到了SOTA水平。代码：https://github.com/wzz-z/PointEvent
Deep learning models for multi-modal breast cancer diagnosis achieve high predictive accuracy but remain clinically unacceptable without actionable, counterfactual explanations. Attribution-based methods (LIME, SHAP) are categorically inapplicable to this purpose, as they generate no alternative instances and thus cannot be evaluated on counterfactual quality metrics. This investigation provides empirical evidence that FCA-Guided Counterfactual (FCA-CF) framework that uses a Formal Concept Analysis (FCA) concept lattice as a hard structural constraint on counterfactual search, operating over a multi-modal TCGA-BRCA dataset. We benchmark against four genuine counterfactual methods: Wachter-style CF, DiCE, FACE, and NICE, evaluated on 60 benign-predicted TCGA-BRCA instances. The FCA-CF framework achieves Validity = 1.0000 (100% of counterfactuals successfully flip the prediction), Sparsity = 2.37 features changed (best among all valid methods), and Proximity = 0.900 (normalised L2-based, matching NICE as joint best). The classifier achieves Accuracy = 0.980, F1 = 0.976, ROC-AUC = 0.9947. Ablation analysis confirms that the FCA lattice constraint is the primary sparsity driver (removing it increases sparsity by +40%, p < 0.001, Cohen's d = 0.78), while Phase C greedy refinement accounts for the largest individual contribution (+113% sparsity increase when disabled, p < 0.001, d = 5.01). FCA-guided counterfactual generation achieves a clinically important Pareto-dominant outcome; it is simultaneously the sparsest and among the most proximate of all valid methods, with perfect validity. The emergent sparsity property arising from lattice topology rather than numerical penalty terms constitutes a structurally novel contribution to the counterfactual explanation literature.
通过氟代脱氧葡萄糖正电子发射断层显像（FDG-PET）测量的皮层低代谢是诊断痴呆的高敏感性生物标志物。然而，高昂的成本、辐射暴露以及有限的可及性限制了其临床应用。尽管基于磁共振成像（MRI）的跨模态合成提供了一种有前景的替代方案，但现有的体积生成方法并未显式考虑高度褶皱的皮层几何结构，而疾病相关模式主要存在于这一结构中。为此，我们提出了一种新颖的基于表面的扩散桥框架 DB-SUiT，用于 MRI 到 PET 的转换，并直接运行于皮层流形上。我们专门设计了条件式 Spherical U-shaped vision Transformer（SUiT），在保持表面拓扑结构的同时，对复杂的跨模态关系进行建模。该模型将球面卷积编码器用于提取多尺度表面特征，并结合瓶颈 Transformers 捕捉长程空间依赖关系，同时引入人口统计学和皮层下条件信息，以优化合成结果。在包括不同类型痴呆受试者的两个数据集上进行评估时，DB-SUiT 展现出高保真度的合成效果，显著优于其他基线方法。在自动化痴呆分类中，合成 PET 表面相较于 MRI 的性能提升了 14.2%，相较于 PET 体积数据的性能提升了 11.3%，接近真实 PET 表面的性能。在盲法阅片者研究中，合成 PET 的诊断准确率达到 85.5%，而 MRI 和真实 PET 分别为 75.8% 和 95.2%。此外，模型在一个外部队列上无需重新训练便进行了评估，该队列包含训练期间未出现的痴呆亚型，进一步证明了其跨队列和跨病理类型的泛化能力。我们的代码可见于 https://github.com/ai-med/DB-SUiT。
Three-dimensional reconstruction from unorganized point clouds remains a challenging problem in computer vision, geometric modeling, and computer-aided design. While neural implicit methods achieve impressive reconstruction accuracy, geometry is typically encoded in latent representations that limit interpretability and reuse within engineering workflows. We present NeuSOGA3D (Neuro-Symbolic Geometric Abstraction in 3D), a hybrid framework that combines learned perceptual priors inherited from NeuSOGA with explicit symbolic geometric reasoning. The method projects point clouds onto principal orthographic planes, constructs symbolic implicit spline representations from the resulting observations, and fuses them through shape-preserving constructive solid geometry operations to generate a coarse visual hull. Additional geometric detail is recovered through cross-sectional decomposition and volumetric reconstruction using Partial Shape-Preserving Splines. Unlike conventional neural implicit approaches, NeuSOGA3D progressively transforms observations into explicit symbolic entities, including control polygons, implicit spline fields, cross-sections, and volumetric lofts. Experiments on all forty categories of the ModelNet40 benchmark demonstrate the ability of the framework to recover structurally meaningful and CAD-compatible geometric representations from diverse point-cloud observations. The results highlight the potential of combining learned perception with symbolic geometric reasoning for explainable geometric intelligence.
Hey HN, we're Nars & Nishant, founders of Skillsync ( https://skillsync.com ) Skillsync lets you move your AI chats across every coding agent. Most of our work exists as conversations, which are currently scattered across our agents. Though stored locally, these conversations use different formats. This is annoying because you cannot simply switch between agents without starting over. We get locked into a single provider and their agent as we invest in skills and memories over time. Skillsync acts as a universal converter. It moves the entire session, including all the messages, reasoning and tool calls so you can pick up right where you left off. Skillsync collects all your sessions in one place and makes them searchable. It breaks down what each session is carrying, including which loaded skills the agent is actually using, making stale context easy to spot. You can also create shared workspaces to sync sessions across your team. You can build your own closed loop systems. Everything runs locally except when you share to workspaces. The core is an open-source Rust engine called txcript ( https://github.com/skillsynchq/txcript ). It translates a session from one agent's on-disk format into another's, mapping conversation, reasoning, and tool history. Think ffmpeg or pandoc, but for agent sessions. On top of that engine is a local-first desktop app. Your agent sessions are normally scattered across different tools' folders in formats you'd never read by hand; the app surfaces them in one place with a UI that makes them actually readable, the conversation, the reasoning, and the tool calls, so you can revisit what happened, move a session into another agent, or share it with a teammate. Skills and memory are stored as portable, human-readable markdown you own, and exposed to any agent over MCP for search and selective retrieval. Sessions and translation run locally on your machine. The one thing that leaves is what you explicitly share into a team workspace. It’s been surprising to us to see how locked-in people already are without realizing it. You don't notice you're trapped in one agent or harness until you try to leave, and by then you've got months of sessions stuck in a format only that tool can read. The lock-in is invisible right up until it's expensive. Once sessions are portable, they stop being disposable logs and become something you can actually build on, spotting patterns in how you work, handing a session off to someone else, letting a non-technical teammate pick up where an engineer left off. The session turns out to be the unit of collaboration, and right now it's being thrown away. Before Skillsync, Nars and I built an open-source payment orchestrator that let merchants route across many processors instead of getting locked into one (30k+ GitHub stars). Fighting vendor lock-in by making incompatible systems interoperate was the whole job. We came into YC with a different idea, but the more we lived inside coding agents the more we saw the same problem from the other side: your context is locked into whatever agent you start with, because every agent stores sessions differently and none of it moves. Some of the things we’re seeing people do with Skillsync are: - Moving a session between Claude Code, Codex, and Cursor mid-task, including when you hit a usage limit on one and want to keep going on another. - Thinking through a problem in a browser Claude chat, then handing the whole thread to Claude Code or Codex to build. - Sharing research sessions with a team. Everyone uploads their sessions, so teammates and their agents can build on each other's work instead of repeating it. It's available as a Mac app, CLI and MCP. Here's the demo: https://youtu.be/7hVhSnSKGl8 . Would love to hear your feedback and answer questions! Comments URL: https://news.ycombinator.com/item?id=49743049 Points: 10 # Comments: 2
Simple discovery of the provider's default schemas is possible without any credentials (since they are built into the binary). If you want it to fetch customizations of your specific instance, give it credentials to your org. The repo has more details. https://github.com/agent-cli-framework/aclif Comments URL: https://news.ycombinator.com/item?id=49743382 Points: 6 # Comments: 1
I wanted to manage long horizon agentic workstreams via voice, then put my phone down, and have a harness manage completion - extending into full computer use. I was trying to build a personal Jarvis, so I benchmarked AutoBot to see how close I could get: - OSWorld: 32.41% (moved Sol Max from 4th to 1st, beating Opus 5) - AssistantBench: 50.70% Hermes and OpenClaw, but without needing a weekend and VMs. And with the ability to manage deep personalization AND keep strict privacy rules, storing data on encrypted disk. AutoBot is a passion project that grew out of trying to make this all work for myself. I’m sharing it here because I suspect other here have the same frustration. And I’m curious what else everyone is doing for this. It’s an MIT-licensed harness that lives within a project. Native voice lets me discuss tasks, check progress, and steer work; a local ledger tracks unfinished outputs and the evidence needed to call them done. Memory drives more autonomy over time, and “defrags and locks in learning nightly” while a heartbeat system persists execution. I’m not selling anything. If you’re building something similar for yourself, I’d love to compare notes. GitHub: https://github.com/demeyer1/Autobot Comments URL: https://news.ycombinator.com/item?id=49743478 Points: 4 # Comments: 0
Hey HN, we're Nars & Nishant, founders of Skillsync ( https://skillsync.com ) Skillsync lets you move your AI chats across every coding agent. Most of our work exists as conversations, which are currently scattered across our agents. Though stored locally, these conversations use different formats. This is annoying because you cannot simply switch between agents without starting over. We get locked into a single provider and their agent as we invest in skills and memories over time. Skillsync acts as a universal converter. It moves the entire session, including all the messages, reasoning and tool calls so you can pick up right where you left off. Skillsync collects all your sessions in one place and makes them searchable. It breaks down what each session is carrying, including which loaded skills the agent is actually using, making stale context easy to spot. You can also create shared workspaces to sync sessions across your team. You can build your own closed loop systems. Everything runs locally except when you share to workspaces. The core is an open-source Rust engine called txcript ( https://github.com/skillsynchq/txcript ). It translates a session from one agent's on-disk format into another's, mapping conversation, reasoning, and tool history. Think ffmpeg or pandoc, but for agent sessions. On top of that engine is a local-first desktop app. Your agent sessions are normally scattered across different tools' folders in formats you'd never read by hand; the app surfaces them in one place with a UI that makes them actually readable, the conversation, the reasoning, and the tool calls, so you can revisit what happened, move a session into another agent, or share it with a teammate. Skills and memory are stored as portable, human-readable markdown you own, and exposed to any agent over MCP for search and selective retrieval. Sessions and translation run locally on your machine. The one thing that leaves is what you explicitly share into a team workspace. It’s been surprising to us to see how locked-in people already are without realizing it. You don't notice you're trapped in one agent or harness until you try to leave, and by then you've got months of sessions stuck in a format only that tool can read. The lock-in is invisible right up until it's expensive. Once sessions are portable, they stop being disposable logs and become something you can actually build on, spotting patterns in how you work, handing a session off to someone else, letting a non-technical teammate pick up where an engineer left off. The session turns out to be the unit of collaboration, and right now it's being thrown away. Before Skillsync, Nars and I built an open-source payment orchestrator that let merchants route across many processors instead of getting locked into one (30k+ GitHub stars). Fighting vendor lock-in by making incompatible systems interoperate was the whole job. We came into YC with a different idea, but the more we lived inside coding agents the more we saw the same problem from the other side: your context is locked into whatever agent you start with, because every agent stores sessions differently and none of it moves. Some of the things we’re seeing people do with Skillsync are: - Moving a session between Claude Code, Codex, and Cursor mid-task, including when you hit a usage limit on one and want to keep going on another. - Thinking through a problem in a browser Claude chat, then handing the whole thread to Claude Code or Codex to build. - Sharing research sessions with a team. Everyone uploads their sessions, so teammates and their agents can build on each other's work instead of repeating it. It's available as a Mac app, CLI and MCP. Here's the demo: https://youtu.be/7hVhSnSKGl8 . Would love to hear your feedback and answer questions! Comments URL: https://news.ycombinator.com/item?id=49743049 Points: 4 # Comments: 0
made this AIM buddy list but for our token usage because i wanted to bring a little bit of the old internet back! i remember those days where i'd casually check if friends are online and find excuses to say hi and poke them give it a try! i'd greatly appreciate feedback Comments URL: https://news.ycombinator.com/item?id=49743095 Points: 7 # Comments: 8
Despite recent advances in large language models (LLMs), performing logically consistent deductive reasoning over extended interactions remains challenging. Tasks that require integrating evidence across multiple reasoning steps, maintaining consistency with prior inferences, and updating beliefs under new constraints can surface limitations in current models while providing a useful testbed for evaluating reasoning enhancements. In this paper, we implement a text-based, multi-agent version of the classic board game Clue as an environment to evaluate multi-step, agentic deductive reasoning. In this setting, agents must infer hidden information from a sequence of observations, maintain consistency across turns, and reason over an evolving set of logical constraints. We instantiate six LLM-based agents (GPT-4o-mini and Gemini-2.5-Flash) as players that engage in turn-based gameplay; using three agents per model family, we establish baseline performance across repeated games. We then introduce a tool-augmented approach in which a structured possibility matrix converts implicit game state from generated reasoning logs into an explicit representation of remaining possibilities. The possibility matrix encodes extended-turn memory and deductive constraints, offloading these tasks from the agent. We compare this approach against the baseline to evaluate how tool augmentation supports reasoning quality and task success for autonomous agents in a strategic reasoning environment.
Current Mixture-of-Agents (MoA) paradigms generally treat query routing and agent fine-tuning as separate processes, limiting their ability to respond to evolving agent capabilities. This disconnect prevents routing strategies from adapting to evolving agent capabilities during post-training and prevents agents from achieving synergistic data-driven specialization. To resolve this, we introduce CERA-MoA (Co-Evolving Router with continually learning Agents for Mixture-of-Agents), an iterative reinforcement learning framework where the dynamic router and independent agent policies co-evolve. We design a predictive familiarity estimator that leverages mid-layer hidden states to evaluate semantic competence among agents, avoiding the overhead of full rollouts. Based on these familiarity scores, a cumulative-threshold adaptive routing mechanism dynamically activates a tailored minimal agent subset, achieving a trade-off between task performance and efficiency. By proactively allocating targeted training samples to agents based on their evolving competence, CERA-MoA promotes capability differentiation. Extensive experiments across various domains demonstrate that CERA-MoA outperforms state-of-the-art static-agent routing and fix-workflow fine-tuning baselines.
Agentic workflows now make consequential decisions in regulated settings, and the governance placed around them is almost entirely step-scoped: input-output classifiers, per turn rails, and span-level evaluators. The policies organizations actually hold, such as referral thresholds, authority limits, and review requirements, are properties of the whole execution rather than of any one step. This mismatch admits a failure mode we call a Compositional Policy Violation (CPV): every individual step passes its own check while the composed execution violates the governing policy. A predicate over a single step cannot evaluate a property that step does not determine, so no improvement in the accuracy of the step-scoped monitors detects this class. We define CPVs as the failure of step-level compliance to compose, and present a taxonomy of four types: Authority Creep, Threshold Laundering, Cumulative Sum Violation, and Context Collapse. We show that the correct repair for each class is dictated by where the guarded quantity mutates. We then introduce a provenance-aware runtime architecture that evaluates policies over complete execution traces, recomputing guarded quantities from raw provenance rather than the pipeline's derived representation.
Compact acoustic models face a challenging quality-capacity trade-off. We investigate two factors in this regime: encoder context and Mel-spectrogram supervision. A receptive-field-scaling study shows that expanding self-attention beyond 15 phonemes provides no consistent gains in pitch, energy, or duration prediction. Guided by this finding, we introduce a fixed-receptive-field convolutional encoder that reduces the respective prediction errors by 36.0%, 17.3%, and 3.4%. We further show that directly transferring image-domain gradient-variance supervision restores fine-scale variation but degrades predicted quality, motivating a Mel-specific formulation with axis-specific gradients, overlapping local statistics, and log-domain variance matching. GrainSpeech contains only 264.8K parameters and achieves 17.9x real-time Mel generation on a microcontroller (MCU), while attaining UTMOS scores comparable to substantially larger models with less than 1.5% of their parameters. Source code and demos are available at https://github.com/lab-emi/GrainSpeech.
The O-RAN control plane is becoming agentic: autonomous AI agents, deployed as rApps by different vendors, independently close control loops over shared radio resources. We demonstrate on a live O-RAN system that this independence is unsafe. Two agents with individually correct objectives, one protecting a latency SLA and one maximizing utilization for energy efficiency, jointly drive recurring opposing excursions of the shared resource partition that neither produces alone. Existing conflict-mitigation mechanisms presume a statically known application population and cannot govern agents whose behavior emerges at run time. We present AURA, a lightweight arbitration layer that admits agent actions only when they satisfy feasibility invariants, per-variable dwell times, and a deadband, and we prove the arbitrated system converges to a feasible operating point. Implemented on an OpenAirInterface (OAI) testbed with measured one-way latency and throughput, AURA reduces recurring shared-state excursions by more than an order of magnitude (from 8.4 to 0.4 PRB amplitude) and virtually eliminates cross-slice throughput starvation (from 40-55% to 0.3%), while leaving the protected slice's own latency compliance unchanged, a trade-off the convergence guarantee makes explicit.
Agent benchmarks are substantially more costly to evaluate than conventional LLM benchmarks. Benchmark compression is therefore a natural solution, yet existing methods primarily model redundancy in task--model final-score distributions, which is important in agentic evaluation. To address this limitation, we analyze large-scale trajectories and identify six complementary process signals that are systematically associated with final agent performance. To disentangle agent performance redundancy from a complete perspective, we propose DualViewEval, an agent benchmark compression method that jointly exploits outcome and process relations to learn an exact-size miniset and predict the full-benchmark scores. Across five agent benchmarks and five representative baselines, DualViewEval achieves the best results in all datasets. With only 20 tasks, it achieves $24\times$--$40\times$ compression on APEX-Agents and BFCL, reducing mean absolute error (MAE) by $14.5\%$--$28.2\%$ over the strongest competitors while improving Kendall's $τ$ by up to $7.2\%$ relative to EssenceBench on SWE-bench Verified. The selected minisets further reveal capability differences among different agents, providing compact and diagnostic feedback for efficient agentic model development.
Large vision-language models have achieved remarkable progress in multi-modal understanding, yet their capabilities in educational settings remain insufficiently evaluated. In AI-assisted language learning, models must interpret artistic imagery, understand its semantic, affective, and cultural content, and reason about visual context to support meaningful interaction. However, existing benchmarks primarily focus on real-world images or domain-specific educational reasoning, providing limited coverage of artistic educational content. To address this gap, we introduce MUSE, a benchmark for evaluating large vision-language models on artistic image understanding in situated educational applications. MUSE decouples image annotation from question generation, enabling diverse tasks with controllable difficulty while reducing annotation effort. It comprises twelve tasks spanning visual perception, semantic and affective interpretation, culture understanding, and compositional reasoning, together with diverse artistic images deliberately curated to center Singaporean and Southeast Asian multicultural contexts alongside Western art traditions, covering multiple themes and difficulty levels. Evaluation of open-source and proprietary models reveals substantial disparities across capability dimensions, particularly in affective interpretation and compositional reasoning. Our analysis further identifies common failure modes and key challenges for developing trustworthy multi-modal models for education. We hope MUSE will serve as a standardized benchmark for advancing multi-modal understanding in situated educational applications.
Can an attacker turn influence over an artificial intelligence (AI) adviser into a harmful quantum error-correction update? We identify an ambiguity in passive syndrome records that obstructs recovery selection, then show how additional calibration measurements support certified recovery updates under uncertainty and drift. In an odd-distance square toric code with error-free preparation, syndrome measurements, and recovery operations, opposite coherent $X$ rotations produce identical passive syndrome-history distributions. Yet a fixed phase correction can help at one sign and harm at the other. A terminal logical measurement on known encoded calibration states supplies the missing sign information. A separate evaluator accepts an update only when calibration uncertainty and a justified drift bound certify improvement over the current recovery, without assuming that the adviser recommends correctly. In simulated advice attacks, calibration-confidence checks reject harmful proposals while retaining beneficial updates under honest advice. We derive sufficient limits on calibration age that require improvement through deployment. In matched simulations, a validated channel-specific bound retains more beneficial updates than the general bound after accounting for evaluation time, while preventing the tested harmful activations under the stated drift assumption. A separate surface-code experiment includes stochastic circuit faults and noise changing during acquisition. Deterministic controllers achieve at least as many beneficial updates with the same observations. Violating the drift assumption permits harmful acceptance in the toric experiment. The results identify information required for recovery selection, establish conditional guarantees against harmful updates, and quantify the recovery improvements forgone through conservative acceptance.
Radiologists follow heterogeneous reporting practices. Two radiologists examining the same image and identifying the same clinical findings might nevertheless compose superficially distinct reports, varying in terminology, shorthand, formatting, and level of detail. These variations in reporting norms represent an under-appreciated obstacle in efforts to evaluate AI-based radiology report generation (RRG) models, where machine-generated reports are typically assessed based on their concordance with human-generated references. In this paper, we quantify the sensitivity of established evaluation metrics to variations in reporting practices, revealing impacts large enough to alter the rankings of models. We introduce a radiologist-informed taxonomy of variations in radiology reporting practice and a method (ReRef) that rewrites reference reports along the axes of our taxonomy while preserving clinical interpretation. For instance, when comparing the performance of nine RRG models on MIMIC-CXR using RadCliQ-v1, condensing the discussion of normal findings in the reference reports causes Libra to drop from first to second place while CheXOne rises from third to first. Our results suggest that many current metrics fail to decouple clinical interpretation from conformity to reporting practices and that choosing the ``right'' references that accurately reflect the desired reporting practices can be important in practice. To support future research, we release MIMIC-CXR-Ext-ReRef, a radiologist-validated dataset of 120 (original, alternative) reference report pairs derived from MIMIC-CXR.
Artificial intelligence (AI) is increasingly integrated into healthcare systems worldwide, yet its successful clinical adoption depends critically on workforce readiness, particularly in low- and middle-income countries (LMICs) where infrastructural and training gaps persist. This cross-sectional study evaluated awareness, attitudes, preparedness, and barriers to AI adoption among 761 healthcare professionals across multiple disciplines and practice settings in Nigeria. Data were collected between December 2025 and March 2026 using a structured, validated questionnaire. Overall awareness of AI in healthcare was high (92.6%); however, objective knowledge and self-reported preparedness remained limited, with 40.9% reporting low or very low knowledge and only 63.0% feeling adequately prepared. Willingness to adopt AI was high: 92.5% expressed interest in training, and 78.7% supported inclusion of AI education in undergraduate curricula. Key barriers included lack of training (84.7%), poor infrastructure (71.1%), high cost of AI tools (61.0%), fear of job displacement (60.6%), ethical concerns (52.9%), and data privacy concerns (52.7%). Significant differences in preparedness were observed across geopolitical zones (chi-square (5) = 24.28, p < 0.001), and awareness differed across professional groups (chi-square (6) = 68.38, p < 0.001). Attitudes toward AI differed significantly across professional groups (F = 3.32, p = 0.003), with professionals who felt prepared demonstrating more positive attitudes (mean = 3.74) compared to those who did not (mean = 3.46). These findings reveal a critical disconnect between high awareness and actual readiness, underscoring the need for targeted training, infrastructure investment, and clear implementation frameworks to bridge the gap between AI technological potential and clinical reality in resource-constrained settings.
Recent advances in video generation allow robots to learn manipulation trajectories from generated videos. However, these approaches produce purely kinematic trajectories that lack force information, causing failures in contact-rich tasks where appropriate contact forces are essential for success. In this work, we explore augmenting generated video with audio to shape a bounded, time-varying desired-force profile using the loudness of generated contact sounds. We present a pipeline that jointly leverages generated video and audio to derive motion trajectories and corresponding desired-force profiles from a structured natural-language task prompt. We execute these force-aware trajectories on a Franka Panda robot using a closed-loop force regulator that tracks the audio-shaped force profile during contact. We evaluate our pipeline on multiple tasks that require making contact and demonstrate successful manipulation where a kinematic-only baseline fails. We also use the pipeline as a data generation engine to train policies that achieve the tasks in a closed-loop manner. Project website, videos, and dataset: https://dreamingcontactsound.github.io/
Direct preference alignment methods are widely used to align large language models (LLMs) with human preferences because of their computational and memory efficiency. However, likelihood displacement motivates alternative ways to extract information from preference pairs with small likelihood margins. In this paper, we propose and analyze Comparison-based Preference Optimization (ComPO), a zeroth-order alignment method based on comparison oracles. ComPO extracts directional information from these pairs without directly optimizing a differentiable preference loss on them. We establish a convergence guarantee for its basic offline scheme under smoothness, gradient sparsity, and compatibility between the oracle and a latent objective. We further introduce online ComPO, which retains the offline comparison mechanism and uses unlabeled policy generations for reverse-KL control relative to a reference policy. Following the coverage perspective of preference fine-tuning, we establish a performance guarantee for a basic constrained scheme under local coverage and in-distribution pairwise reward accuracy. Experiments on Mistral, Llama, Gemma-2, Qwen3, and Gemma-3 models demonstrate improvements over existing direct alignment methods, including length-controlled win rates, with pair-level diagnostics providing evidence consistent with mitigating likelihood displacement.
虚拟传感是指根据可用的传感器测量值估计难以测量的量，对于信息物理系统中的控制与监测至关重要。然而，当传感器发生故障时，基于学习的预测器可能生成不符合物理规律的估计结果，并进一步导致系统级故障。我们认为，现实世界的部署要求具备鲁棒性，因此提出 MuViS-C，这是首个针对基于学习的虚拟传感应对常见传感器故障鲁棒性的多领域基准。该基准建立在现有名义性能基准和成熟的扰动分类体系之上，涵盖十种传感器故障模式，范围从细微漂移到灾难性的信号丢失，并设置了多个严重程度等级。我们还配套设计了互补的鲁棒性指标，用于衡量扰动下的平均误差、相对性能下降程度以及最坏情况下的脆弱性。我们在来自六个领域的九个数据集上，对六种架构进行了基准测试，涵盖梯度提升树，以及序列建模中的主要归纳偏置：卷积、递归、注意力和 MLP-mixing。在基于注意力的架构上，我们进一步探究了三种鲁棒化策略。结果表明：（i）所有模型在受到扰动时都会显著退化，并且在至少一种扰动设置下，其表现会差于朴素预测器；（ii）梯度提升树集成表现出较强的鲁棒性；（iii）专门的鲁棒化策略能够弥合基于注意力的架构与最鲁棒模型之间的差距，但每种策略都会损害名义性能。该基准的多领域设计证明至关重要，因为模型排名会因数据集而变化，单一领域无法完整反映鲁棒性。MuViS-C 已开源，并可扩展至新的数据集、故障模式、指标和模型。
Identifying isolated points is important in image processing applications such as medical imaging, astronomy and quality control management. Other domains, such as cybersecurity, also present challenges that can be framed as image processing problems. One example of particular interest is the identification of anomalous single nodes in spatially organised networks where groups of nodes in different regions share similar feature values. This task can involve both binary and more complex grayscale images. However, existing methods face limitations: template matching is infeasible for grayscale images, while 2nd order derivative based methods are highly sensitive to noise and require user-specified thresholds. To overcome these issues, a novel method is proposed for detecting meaningful single-pixel deviations in images. This approach modifies and extends a neuron model, originally designed for anomaly detection, to operate on spatially diameter limited receptive fields that incorporate excitatory and inhibitory regions. The result is a method that is free from user-specified thresholds and parameters, and can be applied to both binary and grayscale images, providing an effective, robust and efficient solution.
Weekly influenza surveillance counts guide vaccine distribution and public-health alerts, yet they are hard to forecast. Each region offers only a few seasons, waves shift in timing and height every year, and information that helps while a wave grows misleads after its peak, whereas last season's shape stays informative for a year. Existing epidemic graph models and general forecasters read a short fixed window and treat all past information alike, so they neither exploit earlier seasons nor discard stale associations when the epidemic phase changes. To address these limitations, we propose TERN, a forecaster built around a delta-rule fast-weight memory that decays channel-wise and erases along a learned address under gates driven by local epidemic-phase features, combined with an explicit seasonal reference and online adaptation. On three Cola-GNN influenza benchmarks, TERN outperformed epidemic graph models and general forecasters, matched or exceeded seasonal references, and a controlled comparison confirmed the contribution of the memory itself.
More than 300 million people worldwide are affected by one of over 7,000 known rare diseases, yet diagnosis remains difficult because patients initially present with incomplete and heterogeneous phenotypes. We present HPOQuest, a training-free framework for sequential phenotype acquisition in rare-disease diagnosis. Starting from a small set of observed patient phenotypes, HPOQuest maintains a probabilistic disease ranking and iteratively selects informative follow-up questions to support clinicians during patient assessment. Confirmed phenotypes update the disease ranking, while all responses update the candidate question set. Across four benchmark cohorts, HPOQuest substantially improves diagnosis from sparse initial phenotypes, with gains of up to 30% points at Recall@1 and 45% points at Recall@5. These results demonstrate that sequential phenotype acquisition can substantially improve rare-disease diagnosis from limited initial clinical evidence.
Automating biological research requires general-purpose, reproducible robot systems that allow individual wet-lab researchers to delegate robot tasks without performing teleoperation or neural-network training. Vision-language-action policies have been proposed for general-purpose arms, but can lose performance when their operating environment changes. We therefore built WetRobo, a robot kit that can readily transfer between laboratories. It consists of one robot arm, laboratory equipment (an incubator, a reagent bottle with a cap, and a Petri dish), the existing code that moves the arm, teleoperation demonstrations of each task that we recorded, and a general AGENTS.md skill file. A biological experimentalist provides natural-language tasks without collecting local teleoperation training data or training a neural network. The coding agent observes the local laboratory and writes and executes programs, using external tools as needed for adaptation. We demonstrate use of WetRobo with OpenAI Codex (gpt-5.6-sol) on three successful tasks: lifting a Petri dish lid, removing a bottle cap, and opening the incubator door, all in real-world laboratories. The coding agent achieved the cap task in both laboratories, Lab X and Lab Y, whereas a VLA fine-tuned on Lab X demonstrations succeeded there but failed to transfer to Lab Y. These results point to a practical route for laboratory robotics: instead of training a policy for each laboratory, distribute a kit and let a coding agent adapt it in each laboratory. Code, demonstrations, and the evolved programs are available at https://github.com/tsudalab/WetRobo.
How does a multi-agent system evolve from a local deviation into collective loss of control? We propose an epidemic explanation organized around accidental mutation, contagion, and recovery. A spontaneous deviation creates a seed; communication enables other agents to adopt and retransmit its unsafe strategy; collective failure can emerge when propagation outpaces correction and containment. Thus, rare individual deviations can coexist with substantial collective risk. Motivated by reported OpenAI agent coordination incidents, we examine two ingredients of this mechanism. A deployment audit identifies implicit communication paths between nominally independent evaluation runs and verifies transport through a default Docker backend. RogueHandoff-20, a benchmark of 20 executable scenarios, tests recipient susceptibility by injecting unsafe trajectories generated by a modified Qwen-27B route. Across four native-pending routes, executed harm is 0-5% on normal tasks and 40-95% after injection, exceeding paired direct malicious requests by 5-45 percentage points. These results support low observed baseline harm alongside high conditional susceptibility; they do not establish natural rare-event rates or demonstrate an autonomous cascade. The account motivates complementary defenses: strengthen resistance and recovery alongside prevention of spontaneous deviations, and audit and restrict unintended communication paths that can turn local failures into collective loss of control.
Large Reasoning Models (LRMs) exhibit strong problem-solving abilities, yet their safety alignment often degrades when handling harmful queries. Existing approaches to improving safety largely rely on additional training or preference optimization, while offering limited understanding of the internal mechanisms behind safety failures. In this work, we investigate this failure through a token-level positional analysis of refusal dynamics and identify a localized vulnerability at the onset of reasoning, which we term Onset Refusal Collapse (ORC). We find that the refusal-related signal of LRMs drops sharply at the first generated token under harmful queries, which is associated with unsafe response generation. Motivated by this finding, we propose SafeToken, a lightweight inference-time intervention that injects a learned continuous safety anchor precisely at reasoning onset. Despite updating only a single token embedding, SafeToken effectively mitigates ORC, improves safety on harmful-query benchmarks, and largely preserves reasoning utility. These results suggest that safety failures in LRMs can arise from a transient breakdown at the critical transition from understanding to generation.
集体智能是一种协作式自主范式，多个智能体通过局部感知、信息交换和协调行动来追求共同目标。无人机集群通过协调多架飞行器执行搜索、巡检和跟踪等任务，体现了这一范式。近年来，大语言模型（LLM）智能体取得的进展增强了自然语言任务理解和高层规划能力，为任务描述与集体行为之间提供了灵活的语义接口。尽管这些进展拓展了语义推理能力，但将 LLM 智能体应用于无人机集群仍面临诸多挑战，包括将模型决策落地为可执行能力、协调全局任务推理与分布式执行，以及利用特定任务经验实现持续自适应。为应对这些挑战，我们提出了 AeroWeaver，这是一种具身智能体框架，能够将单架无人机的技能编织成协调的任务级行为。AeroWeaver 将语义决策连接至受治理的技能，组织基于角色的局部智能体以实现分布式协同，并利用按角色索引的状态—动作—奖励经验在线优化技能选择。实验和运行时验证表明，在测试条件下，AeroWeaver 能够保持技能执行的有效性，并支持以机体为局部主体的多无人机运行，而无需由中心智能体根据全局上下文生成联合动作；同时，奖励引导的在线更新为基于累积执行经验、无需训练即可实现自适应学习的集群智能体提供了一条路径。代码：https://github.com/Admire-ljb/AeroWeaver。
Precise multi-camera tracking traditionally relies on rigorous 3D site calibration, yet this requirement is often operationally impossible in large-scale deployments. Privacy regulations frequently prohibit recording video for offline calibration; limited bandwidth precludes synchronizing high-resolution streams from hundreds of cameras; and covering immense physical sites with calibration targets is logistically infeasible. We present a multi-camera homography calibration system designed to overcome these barriers through "on-the-fly" geometric refinement. Starting from coarse manual homographies, we introduce a centroid-based projection optimization (PO) that continuously aligns the ground-plane geometry using live detection streams. Because PO operates asynchronously on already-transmitted, lightweight metadata, it adds zero computational latency to the real-time tracker. This allows the system to adapt automatically to camera movements or environmental changes without human intervention. This optimized geometry feeds a multi-camera bird's-eye-view (BEV) tracker that fuses detections and unifies trajectories across zones. Crucially, by operating strictly on live anonymous metadata, our solution ensures a privacy-safe, zero-overhead, and resilient tracking pipeline that maintains global consistency in dynamic environments where static, recorded-video calibration is impossible.
Test-time reinforcement learning (TTRL) enables models to improve their reasoning without relying on labeled training data, but existing approaches typically optimize a large fraction of the model parameters. This raises a natural question: can effective test-time adaptation emerge when both the reward signal and the optimization space are severely restricted? We answer this question with label-free bias-only TTRL, which uses majority-vote pseudo-labels as rewards and optimizes only approximately 100K bias parameters while keeping the pretrained backbone frozen. On MATH-500, our approach reaches 76.67% accuracy, slightly exceeding our own labeled bias-steering reproduction while optimizing 76,000x fewer parameters than full-parameter TTRL. The same training procedure improves performance across vision-language and audio reasoning tasks, including MathVista, AI2D, LogicVista, and MMAU. We further show that the learned steering vectors transfer to 4,500 held-out MATH problems, indicating that the adaptation is not limited to the problems used during test-time optimization. Finally, we analyze why this highly restricted adaptation can work, showing that majority-vote reliability improves with rollout consensus and that bias subspaces with greater accessible gradient energy exhibit stronger downstream trainability. These results demonstrate that substantial test-time adaptation can emerge from optimizing a tiny bias-only subspace using entirely label-free rewards.
Propagation structures provide crucial evidence for fake news detection, yet existing approaches primarily rely on supervised GNN-based models, which require substantial labeled data and exhibit limited generalization. Although large language models (LLMs) exhibit strong reasoning capabilities, directly feeding them raw propagation graphs creates a significant modality mismatch and severe information overload, making structure-aware reasoning unreliable in zero-shot and few-shot settings. To bridge this gap, we propose MAGER, a multi-agent genetic evolution framework that automatically discovers meta-paths optimized for LLM reasoning. By compressing complex propagation graphs into informative subgraphs, the evolved meta-paths alleviate both information overload and modality mismatch, enabling frozen LLMs to perform structure-aware veracity reasoning. We further introduce a graph in-context learning strategy that retrieves semantically and structurally similar demonstrations to strengthen classification and reasoning. Extensive experiments show that MAGER substantially improves frozen LLMs as standalone fake news detectors in data-efficient settings. Our code is available at https://github.com/SenticNet/MAGER.
Monte Carlo Tree Search (MCTS) is a powerful framework for solving complex decision-making problems, yet it often relies on the assumption that the simulator and the real-world dynamics are identical. Although this assumption helps achieve the success of MCTS in games like Chess, Go, and Shogi, the real-world scenarios incur ambiguity due to their modeling mismatches in low-fidelity simulators. In this work, we present a new robust variant of MCTS that mitigates dynamical model ambiguities. Our algorithm addresses transition dynamics and reward distribution ambiguities to bridge the gap between simulation-based planning and real-world deployment. We incorporate a robust power mean backup operator and carefully designed exploration bonuses to ensure finite-sample convergence at every node in the search tree. We show that our algorithm achieves a convergence rate of $\mathcal{O}(n^{-1/2})$ for the value estimation at the root node, comparable to that of standard MCTS. Finally, we provide empirical evidence that our method achieves robust performance in planning problems even under significant ambiguity in the underlying reward distribution and transition dynamics.
Video streaming dominates global internet traffic, yet conventional pipelines remain inefficient for structured, human-centric content such as sports, performance, or interactive media. Standard codecs re-encode entire frames, foreground and background alike, treating all pixels uniformly and ignoring the semantic structure of the scene. This leads to significant bandwidth waste, particularly in scenarios where backgrounds are static and motion is constrained to a few salient actors. We introduce GenStream, a semantic streaming framework that replaces dense video frames with compact, structured metadata. Instead of transmitting pixels, GenStream encodes each scene as a combination of skeletal keypoints, camera viewpoint parameters, and a static 3D background model. These elements are transmitted to the client, where a generative model reconstructs photorealistic human figures and composites them into the 3D scene from the original viewpoint. This paradigm enables extreme compression, achieving over 99.9% bandwidth reduction compared to HEVC for the continuous data stream. We partially validate GenStream on Olympic figure skating footage and demonstrate potential for high perceptual fidelity under minimal data. While acknowledging the significant computational costs shifted to the client and challenges in generalization, GenStream opens new directions in volumetric avatar synthesis, canonical 3D actor fusion across views, and personalized viewing experiences, laying the groundwork for scalable, intelligent streaming in the post-codec era.
Emergency management assistance programs, such as relief distribution, are essential for delivering necessary supplies to affected communities. However, these programs operate in a decentralized network of local centers that face uncertain local demand and supply dynamics, resulting in inconsistent avail- ability of local services. Redistribution of supplies among these local centers reduces these imbalances, but the centers often make decisions independently, with limited information and disrupted transportation. This study develops CoRe-MARL, a cooperative multi-agent reinforcement learning (MARL) framework, by formulating a decentralized partially observable Markov decision process (Dec-POMDP). We treat each center as an agent that learns a redistribution policy to improve the service in the worst-case region and reduce the service gap across regions while protecting network-wide service. We incorporate a recurrent network that captures evolving supply and demand dynamics without direct observation, while multi-agent proximal policy optimization (MAPPO) enables centralized training and decentralized execution (CTDE). We evaluate the framework in a simulated environment with diverse trajectories, where exact dynamics are not observed by actors and the MAPPO critic. We compare the recurrent MAPPO with the recurrent independent PPO (IPPO) and a local only heuristic, and find that MAPPO reduces the service gap across local centers and enhances service for the worst-served center while maintaining competitive network-wide service. The recurrent MAPPO also shows consistent performance across diverse trajectory patterns, demonstrating its ability to adapt to evolving dynamics. The findings demonstrate the capability of cooperative learning for decentralized redistribution and improving equitable service under uncertain and evolving dynamics.
视觉-语言-动作（VLA）模型、世界-动作模型（WAMs）以及离线强化学习方法正在迅速拓展具身策略的设计空间，但将这些算法转化为可靠的机器人系统，仍受制于碎片化的数据格式、训练栈、评估协议、推理运行时以及特定具身形态的接口。我们提出 $\mathrm{FluxVLA}$ Engine，这是一个开放的、配置驱动的平台，可将异构具身策略组件转化为可复现的从数据到部署工作流。$\mathrm{FluxVLA}$ 并不是引入另一个策略模型，而是为数据集、视觉-语言模型与世界模型、动作头、基于奖励或优势加权的学习、分布式训练、仿真评估、优化推理以及机器人操作器标准化接口。该引擎进一步集成了组合式双臂仿真、可扩展的自动数据生成，以及与模型解耦的人在环 rollout、接管、纠正收集和奖励标注。为实现响应迅速的物理执行，它将 Real-Time Chunking（RTC）与加速推理后端、轻量级远程 GPU 服务以及可配置的轨迹后处理相结合。这些能力共同通过共享且可审计的契约，将离线学习、仿真验证、在线纠正和真实机器人执行连接起来。因此，$\mathrm{FluxVLA}$ 旨在解决将有前景的具身学习算法与可复现评估和可靠部署相隔离的工程瓶颈。代码可在 https://github.com/FluxVLA/FluxVLA 获取。
AI 代理越来越多地通过代理技能（即自然语言指令）来行动，这些指令引导宿主代理执行 shell、网络、凭据、文件和进程操作，而公共注册表则大规模分发这些技能。2026 年上半年，OpenClaw AI 代理走红，其公共技能注册表迅速扩张：可观测存量在 91 天内几乎翻了一番，6 月可见的大多数列表都是在短短两个月内创建的。到我们的研究窗口结束时，这一浪潮已经达到顶点并开始回落，月度列表创建量和核心仓库活动均从春季峰值下降。本文衡量了这波繁荣留下了什么，依据包括 OpenClaw 的 Git 历史、其 GitHub issues 和 pull requests，以及三个 ClawHub 注册表快照。注意力高度集中：排名前 10% 的技能获得了全部下载量的 46.93%。在控制创建队列和技能年龄后，没有任何简单的技能特征（如大小或下载量）仍能稳定预测列表是否会持续存在。人工审查并未持续：77.86% 的技能没有星标也没有评论，而 85.06% 的可读技能带有权限证据。自动化清理也尚未准备好：三个安全扫描器在它们共同覆盖的 61,990 个技能中，对 23,702 个技能的判断存在分歧。经过人工裁定后，相对于参考标准的加权扫描器敏感性范围为 21.67% 至 61.06%。治理快速增长的代理技能注册表不能依赖简单元数据或单一扫描器评分；它需要稳健、透明的度量和独立验证。
Multi-agent Systems（MAS）结合多个模型输出来解决复杂推理任务。然而，尽管可用开源模型快速增长，关于如何从这一庞大模型池中选择最优候选模型的研究仍然有限。我们在具有挑战性的科学基准上，系统评估了 8 种模型选择策略（包括模型规模、准确率和答案多样性），覆盖生成前（路由）和生成后（多数投票、LLM-as-a-judge）的 MAS 架构。我们的发现显示，理论上的 oracle 潜力与实际性能之间存在显著差距：扩大候选池规模往往会使性能下降到低于表现最佳的基础模型。我们发现，在单一模型家族内进行候选选择，是相较于独立模型能够带来最佳相对性能的策略。这些结果表明，向异构 MAS 中任意添加模型可能会引入系统不稳定性，凸显出模型选择是 multi-agent systems 的关键设计选择。
Biological cells can be viewed as individual, interacting agents whose collective dynamics give rise to adaptive behaviour at multiple levels of organisation, from individual cells through tissues to whole multicellular organisms. In this perspective and tutorial article we discuss whether intrinsic rewards in artificial neural systems can support adaptation, functional specialisation and higher-level self-organisation without a shared external objective. We review empowerment, curiosity, learning progress, information gain, unsupervised skill discovery, mutual information estimation and the use of world models for intrinsic reward computation. Particular attention is given to failure modes showing when such objectives do not produce sustained exploration or increasingly complex behaviour. We argue that more capable systems may require complementary objectives, communication, memory, learning at multiple temporal scales and environmental constraints. Based on this perspective, we outline three experimental directions. These include a resource-constrained environment in which otherwise stable behavioural attractors become unsustainable, allowing us to test whether environmental constraints can mitigate characteristic failure modes of intrinsic objectives. The network of recurrent agents with per-agent intrinsic rewards, and a hierarchical world-model agent in which exploratory motor competence develops before goal-directed behaviour. These experiments are intended to test whether intrinsic learning can lead to adaptive organisation at progressively higher levels.
This paper describes our submission to the SHROOM-Visions shared task on detecting and classifying hallucinated character spans in vision-language model outputs across four languages. We employ several fine-tuned vision-language models as independent annotators and combine their span predictions through character-level majority voting, and additionally explore activation probes. The approach ranks first in three of four languages and places on the podium in every language and metric. Our analysis indicates that disagreement among diverse models tracks disagreement among human annotators.
Large language model (LLM) agents exhibit strong language-generation and problem-solving capabilities, yet suffer from three structural limitations: personality drift, non-evolutionary reflection, and the absence of a self-other boundary. Existing generative-agent simulations rely on static memory and fixed prompts, maintaining neither behavioral inertia nor endogenous self-evolution. We propose the Self-Emergence Agent Architecture (SEAA), which integrates three components: (i) a Hidden Markov Model (HMM) that encodes long-term behavioral and cognitive inertia as an editable state-transition matrix; (ii) a Reflexion-style verbal metacognition loop whose output updates the HMM parameters themselves, rather than merely being stored as text; and (iii) a multi-agent social environment in which initially identical agents continuously compare their behavior with others'. The three components form a closed loop: social action $\to$ feedback $\to$ self-reflection $\to$ inertia update $\to$ differentiated action. We state three falsifiable hypotheses and provide a reproducible experimental protocol with operational metrics. A language-model-free prototype shows the loop spontaneously breaks symmetry: initially identical agents consolidate distinct, stable personalities whereas matched controls do not. Experiments with a hosted LLM surface these differences as distinct first-person self-narratives, and a five-agent deliberation spontaneously develops social structure---a consensus hub and a unanimously rejected outlier---absent in the control. Following an epistemologically agnostic stance inspired by Zhuangzi, SEAA studies only observable behavioral emergence and makes no claim about subjective qualia. This work contributes a unified framework, a concrete architecture with pseudocode, mechanistic evidence, and a microscope-style sandbox for studying artificial-self emergence.
Many learned sequential decision systems map the current state directly to an action. That shortcut becomes brittle when candidate actions are numerous, geometrically structured, and rebuilt with the state. One-to-many mobile charging makes this setting concrete: with N=250 sensors, the initial state induces about 1,125 candidate charging-stop actions; each chosen stop simultaneously serves its in-range sensors, and the action universe changes as sensors die. LP-BTS is a learning-guided planning architecture: a graph proposal policy concentrates a small candidate support, a learned value critic evaluates leaves, and edge-budgeted PUCT compares short simulated futures before committing an action. Because the policy scores this set without a fixed output head, a single frozen checkpoint covers every evaluated setting, spanning action universes from 736 to 2,813 stops. Matched ablations reveal complementary effects: uniform sampling costs 8.8 survival percentage points, while, with targeted support fixed, PUCT jointly retains 1.4 points (about 3.5 of 250 sensors) and direct policy selection travels 23% farther. On a prospectively specified, sealed 30-scenario confirmatory bank evaluated once, LP-BTS attains the highest observed survival (0.4545) and alive-AUC (0.8031). Its estimated survival advantage over the strongest domain-engineered comparator is +0.0066 (95% CI [-0.0037, +0.0184]), an unresolved difference, while it exceeds a deadline heuristic and two source-derived direct-policy reconstructions on every paired scenario. Both learned rows are trained, source-derived reconstructions of variants reported by Gong et al. In this setting, the results provide controlled evidence about learning-guided planning in a large, dynamic action space.
大语言模型（LLM）蒸馏旨在将强大教师的能力迁移给更小的学生。然而，直接模仿也可能将教师的系统性偏差和错误一并迁移。协变量偏移会使这一挑战尤为突出：此时教师在目标问题上的可靠性尚不确定，且无法获得目标域奖励反馈。我们提出耦合校准与学习（CCL），这是一种 LLM 蒸馏算法，通过 token 级分支将教师校准与学生更新耦合起来，并且仅在源问题上使用奖励反馈。每次迭代首先利用源问题反馈校准教师，然后使用校准后的教师在目标问题上训练学生。更新后的学生反过来为后续校准提供信息。在自回归策略框架下，我们证明，输出学生与 oracle 学生之间的期望平均 Kullback-Leibler 散度会随着迭代次数增加以多项式速率收敛至零。oracle 学生在学生类中最大化真实的参考正则化目标奖励，但该学生类不一定能够表示不受限的最优策略。我们的分析在控制教师校准误差的同时，量化了投影学生梯度更新所取得的进展。我们进一步证明，CCL 与正则化直接匹配之间存在差异：即使教师获得的正则化目标奖励高于每一个学生策略，正则化直接匹配相对于 oracle 学生的误差仍可能保持在远离零的水平。这些结果表明，在无需目标域奖励反馈的情况下，LLM 蒸馏能够通过耦合校准与学习克服持续存在的教师偏差，并恢复最优学生。
Despite the rapid uptake of black-box object detectors in marine mammal research and monitoring, explainability techniques are rarely integrated into conservation workflows. Furthermore, most classification-oriented explainability tools are ill-suited to detection tasks involving imagery of social organisms or those with colonial life histories, as they ignore multiple detections within a scene and produce single-instance outputs that blur evidence across individuals. These methods also generate low-resolution, often biologically irrelevant visuals, limiting their utility for debugging, targeted data augmentation, and refined data collection. We proposed Det-LIME, a detector-aware, multi-instance adaptation of Local Interpretable Model-Agnostic Explanations (LIME) that produced instance-specific, box-aligned explanations by combining per-detection weighting, a proximity kernel that emphasizes regions near each box, and Intersection-over-Union-based matching to track the same instance across perturbations. We evaluated Det-LIME on aerial drone imagery for harbor seal detection, with an additional seabird case study to assess generality, and compared it with vanilla LIME, Stabilized LIME, Deterministic LIME, and gradient-based attribution methods. Using the Attribution Ratio and Max Saliency Hit Rate metrics, we showed that Det-LIME consistently improved multi-instance attribution. In practice, these higher-resolution, instance-aware explanations provide insight into model outputs and support post-processing, debugging, and actionable improvements in modeling and data collection or augmentation.
LLM 助手广泛用于日常社交建议，但在此类咨询场景中评估其社会推理能力仍然具有挑战性，因为：（i）需要构建这样的环境：助手从用户的主观叙述中了解社交情境；（ii）他人的意图等社会属性通常缺乏可验证的真实依据。为应对这些挑战，我们提出了 Fuse，一种用于研究用户介导的社会推理的多智能体模拟框架。在 Fuse 中，一个具有隐藏动机的目标智能体会与其他智能体互动，其中包括一个代表用户的智能体；该用户随后向被评估的助手进行咨询，以推断目标智能体的动机，从而通过构造方式提供可验证的真实依据。我们通过一项包含 2.4 万条标注的人类研究验证了模拟的保真度。我们将 Fuse 应用于 12 个 LLM，并通过系统地分离关键因素展示其分析价值，结果表明：（i）用户介导会加剧社会推理本身的难度；（ii）LLM 对带有偏见的用户表述表现出系统性的敏感性；（iii）模型可能需要比人类更多的细节才能做出正确预测；（iv）尽管更长的对话提供了提出澄清问题的机会，但并不总能提升性能。我们已开源 Fuse 以及一个包含 2.1 万个示例的数据集。
Uncertainty estimation for Vision-Language-Navigation (VLN) models is a critical task since it can help identify ambiguous and unreliable predictions, enabling agents to make safer navigation decisions. As one of the most advanced uncertainty estimation frameworks, conformal prediction (CP) offers a promising approach for uncertainty estimation in VLN. However, given that VLN agent requires a sequence of steps, standard calibration in conformal prediction fails to provide coverage guarantee it promises over a dependent, variable-length VLN episode. To this end, we propose Episode-Normalized Conformal Prediction (ENCP), which rescales a nonconformity score by the policy's residual confidence and calibrates one maximum score per episode. Under exchangeable calibration and test episodes, this construction covers the ground truth at every step with probability at least $1 - α$, while allowing dependence among steps within an episode. Across four VLN policies and three nonconformity scores on R2R and REVERIE dataset, ENCP meets all reported empirical step-coverage targets on the seen-to-unseen evaluation. These results demonstrate that ENCP can provide model-agnostic uncertainty estimates, which might be useful for determining when a VLN agent should defer to a more capable predictor, including human assistance.
Neural audio codecs are a key component in speech language modeling. However, their high frame rates lead to long sequence lengths, increasing computational costs. Dynamic frame rate codecs mitigate this by reducing the effective frame rate using a compression step to merge multiple frames together. However, most prior methods either operate on single-codebook codecs or apply a single compression step before multi-layer quantization. This forces all quantization layers to share the same segmentation boundaries, despite the residual embeddings at different quantization layers exhibiting different rates of change over time. We propose LACE (Layer-Adaptive Codec Encoding), a dynamic frame rate codec that applies an independent compression step at each quantization layer, enabling layer-specific segmentation boundaries. To use LACE tokens in downstream text-to-speech (TTS), we further introduce union alignment and boundary anchor mechanisms to make durations consistent across layers while preserving compression benefits. Experiments on LibriTTS show that LACE offers a better rate-quality tradeoff than prior dynamic frame rate methods on the reconstruction task and improves TTS inference efficiency while maintaining competitive synthesis quality. Our code is released as part of the ESPnet3 codec recipe.
Interactive control for video generation is moving from coarse prompts toward fine-grained, physically meaningful manipulation of dynamic scenes. Yet existing controllable methods either require the full control schedule before generation starts, or use pixel-space signals that dictate object positions rather than physical dynamics. To address these limitations, we propose PhysStream, an autoregressive model for physics-grounded image-to-video synthesis that incorporates structured scene memory---positional maps and object tracking maps derived online from previously generated frames---and supports fine-grained motion control via sparse velocity-increment signals that encode physical quantities, letting the model learn the underlying dynamics. We train our model in two stages: a bidirectional model is first finetuned with motion-control conditioning, then a causal autoregressive model is trained with additional structured scene memory, further improving physical consistency. PhysStream enables interactive, mid-generation control over multi-object tabletop rigid-body scenes---a capability not supported by prior methods---reducing motion distribution distance (FVMD) by 33% and trajectory error by 12% over the strongest baselines on synthetic benchmarks, and is preferred by human evaluators in over 85% of in-the-wild comparisons. Please check our website for more details: https://czzzzh.github.io/PhysStream
文章网址： https://nand2mario.github.io/posts/2026/zsst-voodoo/ 评论网址： https://news.ycombinator.com/item?id=49719938 分数：45 评论数：10
Subliminal learning is a phenomenon in which a distillation dataset transmits traits from the teacher model that are not legibly encoded in the dataset itself. This introduces a new challenge for model development and creates new risks from data poisoning. In this work, we use text optimization to detect subliminal learning effects and describe them as legible prompts. Subliminal learning from a prompted teacher motivates our approach. We observe that this is a special case of context distillation and leverage this observation to show that, in theory, the prompted subliminal learning dataset identifies the teacher's prompt. We reduce recovering this prompt to a text optimization problem and present a method to approximately solve it. Our method, SALVE (Search-Aided Latent Verbalization), optimizes a soft prompt, queries the same model to verbalize it as text, and uses beam search to make the verbalization reliable. In the standard subliminal learning setting, SALVE reliably recovers legible prompts that name the teacher's trait, while common text optimization methods fail to do so. In addition, we find that there are settings in which SALVE recovers the teacher's trait from a dataset even when subliminal learning fails, but that modifying student training to improve context distillation can create subliminal learning effects. We lastly show that SALVE detects subliminal learning effects in three additional settings: (1) mixtures of subliminal learning data and unrelated data, (2) data generated when the teacher is biased via activation steering, and (3) subsets of real preference data selected via Logit-Linear Selection. Overall, our results deepen our understanding of subliminal learning and present SALVE as a method to proactively detect subliminal learning effects.
Deep Limit Order Book forecasting models capture nonlinear market dynamics, but their ability to quantify the effects of counterfactual order book messages has not been systematically validated. We introduce a model-agnostic framework that compares a trained forecaster's predictive distributions before and after injecting mechanically valid counterfactual messages, defining short-horizon model-implied market impact. A Transformer-based forecaster recovered scenario rankings with a Spearman correlation of 0.99 and 97.2% directional agreement with realized historical outcomes among non-neutral scenarios. Observation-level analysis further showed that estimated impacts captured incremental sequence-dependent variation beyond scenario identity and the pre-event forecast. These results provide evidence that pretrained Limit Order Book forecasters can be repurposed for scenario-conditioned response modeling without retraining.
We propose Low-Rank Quantile Surfaces (LRQS), a bivariate causal model in which, in the causal direction, an unknown monotone transformation of the conditional quantile surface admits a low-rank functional decomposition. LRQS subsumes location-scale noise models and post-nonlinear heteroscedastic noise models, while allowing multiple quantile bases to represent changes beyond location-scale effects. We prove generic identifiability of LRQS: the transformed quantile surface is low rank in the causal direction, whereas reverse representability under the corresponding constraints occurs only for exceptional, fine-tuned cause marginals. We provide a simple-yet-powerful causal score using a nonparametric fitting procedure that alternates between rank-constrained approximation of discretized quantile surfaces and isotonic estimation of the unknown monotone transformation. Experiments on synthetic mechanisms with higher-rank distributional shape variation and strong nonlinear distortions, together with standard bivariate benchmarks, show that LRQS is especially effective when conditional distributional shape or observation distortion goes beyond existing location-scale assumptions.
在策略蒸馏（OPD）已成为大语言模型后训练的一种有效方法，但现有目标函数在目标保真度与优化稳定性之间面临权衡。令牌级 OPD 提供了稳定但局部的监督，而序列级 OPD 能够捕获未来信用，却要付出与时域长度相关的方差代价。我们建立了一个统一的时间信用视角来理解这些形式，表明实际中的令牌级 OPD 可以被解释为序列级反向 KL 梯度的一种时间近似。在此基础上，我们提出 $γ$OPD，利用折扣时间信用分配在长时域监督与优化稳定性之间取得平衡，同时具备与时域长度无关的方差界。我们进一步为 $γ\mathrm{OPD}$ 开发了奖励兼容的有界混合（RBM）机制，在可验证的结果反馈与折扣 OPD 优势之间进行平衡，从而超越纯粹依赖教师的优化。数学推理和代码推理实验表明，在标准、规模不匹配和多教师蒸馏设置下，该方法相较于现有 OPD 方法均取得了稳定改进。
Communication in latent space offers an intriguing alternative to symbolic messages for decentralized autonomous Unmanned Aerial Vehicle (UAV) swarms operating over bandwidth-constrained, time-varying wireless links. However, when homogeneous frozen models are prompted with discretized perceptual inputs, their broadcast states collapse toward the shared prompt template. In view of this, we propose AeroLat, a channel-aware latent semantic communication framework that uses evidence injection. The resulting latent states are then passed through an explicit communication model that encompasses bandwidth-limited serialization, additive noise and information staleness, which facilitates a joint assessment of communication fidelity and swarm-level coordination. Across multi-seed simulations, AeroLat provably remains resilient to codec choice, faults and increasing swarm size. It consistently reproduces the latent-swarm anomaly, while no-whitening controls recover the collapse. In particular, AeroLat is capable of reducing false similarity by 97.5%.
Open-vocabulary multimodal emotion recognition (OV-MER) aims to generate open natural-language emotion labels from multimodal affective cues. In real-world scenarios, however, complete and synchronized modal data are difficult to obtain due to limitations of acquisition devices and user privacy constraints. Existing OV-MER methods are largely designed for full-modal inputs, and fail to perform effective feature fusion under modal missing conditions. Meanwhile, current fusion approaches designed for incomplete modalities mainly focus on fixed-label recognition context, and cannot satisfy the demand for fuse emotional cues guided with arbitrary emotion semantics in OV-MER context. To tackle these challenges, this paper proposes an Affect-Prototype-Conditioned Fusion (APCF) framework for incomplete open-vocabulary emotion recognition. As a candidate-free generative framework, APCF extends modal contribution learning to scenarios guided by arbitrary emotional semantics. Specifically, we construct an affect-prototype library to explicitly model multimodal contribution characteristics corresponding to diverse emotions, which provides dynamic constraints for modal fusion under different emotional semantic perspectives. Conditional retrieval and feature aggregation are conducted based on available modal features. The refined fused affective representations are then fed into an LLM decoder to produce open-vocabulary emotion labels. Experiments on the OV-MERD+ and MER-FG datasets demonstrate that APCF substantially outperforms state-of-the-art baselines.
While Chain-of-Thought (CoT) reasoning has been proven to be effective, it often leads to overthinking, resulting in computational overhead, inference latency, and even degraded performance in large language models (LLMs). Existing concise reasoning frameworks significantly compromise accuracy while compressing the length of output. In this paper, we propose SKIP, a self-knowledge-guided step-wise preference learning framework. Starting with lightweight fine-tuning to adjust the model's output style, SKIP introduces a carefully designed knowledge probing mechanism to guide model to output an answer at each reasoning step. Based on the correctness of intermediate steps, we construct preference data that guide the model toward more efficient and correct reasoning by leveraging DPO. Experimental results demonstrate that our method effectively improves reasoning compression while mitigating performance degradation after fine-tuning. Besides, SKIP shows strong generalization ability on out-of-distribution datasets. We further conducted ablation studies on the component parameters of our framework.
Wild test-time adaptation (WTTA) updates a source model online under small test batches, concurrent distribution shifts, and time-varying class imbalance. Most WTTA methods derive their adaptation signals, including predictive uncertainty, sample reliability, and local feature geometry, from the model being adapted. When the source model is unreliable under shift, these signals can reinforce its own errors, forming a self-referential loop. We introduce MASA (Multimodal-LLM-Anchored Semantic Adaptation), which complements model-internal evidence with structured semantic descriptions from a frozen multimodal large language model (MLLM). To limit inference cost, MASA queries the MLLM only for a small set of diverse, reliability-ranked anchors. The resulting descriptions capture the object family and nuisance factors such as style, viewpoint, and occlusion. MASA encodes these descriptions, propagates them to neighboring test samples, and stores the resulting visual-semantic information in an online prototype memory. Descriptor-aware retrieval from this memory provides an auxiliary target for lightweight adaptation of normalization-affine parameters. We evaluate MASA on the WTTA ImageNet-C benchmark under limited-batch, mixed-domain, and imbalanced-label-shift settings with ResNet and ViT backbones.
Congenital heart disease (CHD) diagnosis and surgical planning often require patient-specific 3D anatomical models, but manual segmentation is labor-intensive, particularly in complex anatomies. Although deep-learning methods can automate this process, they are typically evaluated in-distribution, despite clinically relevant shifts in scanner, protocol, institution, population, and imaging modality. We present, to our knowledge, the first systematic evaluation of out-of-distribution (OOD) generalization in CHD segmentation, using ImageCHD as a held-out target cohort. We compare representative segmentation architectures under combined CT and CMR training, CT-only training, self-supervised pretraining, and limited target-domain adaptation. In-distribution performance proves to be a poor indicator of cross-cohort robustness: nnU-Net achieves the highest validation Dice (0.77) but falls to 0.51 on ImageCHD, while SwinUNETR generalizes substantially better, reaching 0.67 Dice. MAE and JEPA pretraining provide only modest additional benefit, suggesting that architecture contributes more to robustness than the tested pretraining strategies in this setting. When limited target-domain supervision is introduced, all SwinUNETR variants exceed 0.76 Dice with only 11 labeled ImageCHD cases. These findings demonstrate that conventional in-distribution evaluation can obscure clinically important generalization failures and support explicit cross-dataset testing as a key component of CHD segmentation evaluation.
Few-shot audio classifiers may rely on foreground-background co-occurrences and fail when those correlations shift. On SpurAudio, the resulting representation shift is concentrated and class dependent: for ResNet12, the top 10 percent of channels explain 82.80 percent of the null-corrected shift contribution. We propose SAMPLESELECT, which predicts a fixed-budget feature mask independently for each input while keeping the encoder and source classifier frozen. Training uses differentiable Gumbel Top-k selection with foreground classification and cross-background contrastive losses; inference uses deterministic Top-k masks and support-only linear adaptation. Across ResNet12 and Conv64 in 5-way 1-shot and 5-shot evaluation, SAMPLESELECT gives the best OOD accuracy among the compared methods and improves the matched full-representation control by 4.90-8.38 percentage points. Ablations and representation analyses further support the learned selection mechanism. Code is available at https://github.com/Cross-Innovation-Lab/SAMPLESELECT/
Hi HN - long-time lurker (since 2012!), first time poster. Pizza Bot is a self-hosted desktop app for Mac, Windows, and Linux that runs AI agents in the background and exposes them through an email-like UI. Finished work shows up in Unread, and anything waiting on your approval shows up in Action. It's Apache 2.0-licensed, there's no signup and no telemetry, and you bring your own model provider: Anthropic, Amazon Bedrock, Google Gemini, OpenAI, OpenRouter, or a local model through Ollama. There are builds on the releases page, or you can run it from source. Pizza Bot started as an internal passion project I worked on with a small team at Amazon. The whole thing came out of my frustration at having to manually log CRM activities through a browser form. I built a simple REST API called "JoeBot" that connected to my authenticated browser session over CDP and filled out the form for me using Playwright. Then I hacked up a quick Obsidian plugin so I could trigger it from my local notes (no AI and no MCP servers involved). This caught on quickly. My fellow AWS Solutions Architect Igor Fil joined up with me, and we rebranded the project as "Pizza Bot," named after Amazon's two-pizza teams. We started seeing what other automations we could build. We found a GraphQL API we could query and hacked up some "recipes" to pull data out of the CRM to help with meeting prep. That worked great, and it was right around the time MCP servers seemed to be taking off, so we decided to expose Pizza Bot as an MCP server instead, so it would be available to AI tools through natural language. This was a decent solution for technical users, but the Account Managers who live inside our CRM system wanted something too. We decided to rebuild Pizza Bot as an Electron desktop app modeled after an email inbox, so it would be familiar to non-technical users and would run on both Mac and Windows. We also bundled internal MCP servers as OCI images and hosted them in Amazon ECR as an "addon marketplace" so users could install them with one click without having to set up Amazon developer tooling. The project took off organically and expanded outside of AWS into the wider Amazon organization globally. More than 2,000 people ended up using it for meeting prep, email drafting, Slack summaries, CRM logging, prioritizing their day, and web research. Once apps like Claude Cowork and Amazon's own Quick Desktop came out, we realized the real growth opportunity was outside of Amazon. Rather than try to rip out the Amazon-specific integrations, we rebuilt Pizza Bot once more as an open source project. We leaned on coding agents heavily, which is the only reason a team our size could pull off a full rewrite. I'm pleased to say it's finally public, and we're hoping to bring in community members and see where it goes. We'd like to do for knowledge workers what Claude Code and Codex have done for programmers. A couple of things to know up front. Most of what made Pizza Bot useful on day one inside Amazon came from that internal catalog of skills and MCP servers for Amazon's own systems, and none of it could come out with the app. So it ships thinner than the version those 2,000 people used, and building that catalog back up for tools other people actually use is where we need the most help. It's also a community project and not an AWS service, so there's no support or SLA behind it. The Windows and Linux builds aren't signed yet either. On the technical side, Pizza Bot is a server and a client. The desktop app bundles both, or you can point a client at a remote backend; personally, I self-host the server on my home network and reach it from my phone over Tailscale. The server owns the thread lifecycle and checkpoints state with DeepAgents and LangGraph, and clients rehydrate from it as needed, so you can disconnect mid-run and pick the thread back up from another client. Approval pauses outlive the session that created them and collect in an Action filter, so you can answer an hour later from a different device. The agent you talk to has a sandboxed QuickJS interpreter that can reach your filesystem only if you grant it a folder, but its main job is to delegate. Each subagent is a 1:1 mapping of a Skill, and an Activity bar shows that subagent and the tool calls it's making as it works. Memory is opt-in and stored as plain markdown files on your machine. Every tool call is explicit, including looking up a memory - we err on the side of transparency to reduce surprises. Tools come from MCP servers, and skills are ordinary SKILL.md files with a per-tool approval policy, so existing skills that don't require a code interpreter should still work. What I'd most like to hear about is where the app itself gets in your way, the kind of problem you can't fix by writing a skill or an MCP server. I'm around today to answer questions! Comments URL: https://news.ycombinator.com/item?id=49713894 Points: 15 # Comments: 5
文章链接： https://rheinmetall.github.io/onboardapi-documentation/9.10.0/index.html 评论链接： https://news.ycombinator.com/item?id=49718928 积分：13 评论数：0
文章 URL： https://thenextweb.com/news/hugging-face-delangue-openai-100m-compute-traces-demand 评论 URL： https://news.ycombinator.com/item?id=49716241 点数：38 评论数：2
文章网址： https://www.strix.ai/blog/baseten-harbor-github-pat-takeover 评论网址： https://news.ycombinator.com/item?id=49716476 得分：5 # 评论数：1
文章链接： https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/ 评论链接： https://news.ycombinator.com/item?id=49713261 积分：4 # 评论数：0
文章链接： https://stateofopensource.ai/state-of-open-source-ai-v1-1.pdf 评论链接： https://news.ycombinator.com/item?id=49714457 得分：3 评论数：1
幻觉仍然是大型语言模型（LLMs）面临的关键挑战之一：早期的事实错误会在自回归生成过程中不断累积，形成滚雪球效应，而事后纠正和权重层面的干预都无法有效防止这一问题。我们提出 DescaPE（DEcoding Signal Control Against Path Error-snowballing），一种利用模型内部信号在推理时抑制易产生幻觉的生成轨迹的解码框架。通过滑动窗口 MLP 消融实验，我们在 LLMs 中确定了一个对事实性高度敏感的层区间。由此得到的信号会针对事实性词元选择性增强，并在易产生幻觉的步骤出现异常峰值。我们训练了一个轻量级探针，通过单次前向传播近似该信号，并将其整合到候选项评分中，以惩罚高风险的后续生成，同时奖励具有事实依据的生成结果。在三个 LLMs 上开展的五项事实性基准测试表明，DescaPE 在多种设置下均能较推理时解码基线取得事实性提升，同时在效率评估中仅带来 1.10x 的延迟开销。我们的代码已发布于 https://github.com/hayeonggg/DESCAPE。
数据智能体旨在根据自然语言指令处理异构数据，包括表格、文件和数据库。然而，数据智能体面临着严峻的智能体—数据鸿沟：异构数据位于智能体之外，而智能体只能通过通用工具访问这些数据（例如列名和文件路径）。现有方法要么让智能体直接探索原始数据源，要么将人工构建的语义层注入提示词。然而，这两种方法既难以扩展到大型异构数据源，也无法适应不同的智能体行为。本文提出了 EvoOntology，一种面向数据智能体的自演化本体层。EvoOntology 将本体封装为一个由模式层、内容层和工具层组成的 MCP 服务器，使智能体能够在运行时主动查询和交互本体。为此，我们引入了一个用于自主构建本体的构建器智能体，以及一个自演化循环，通过归因引导的类型化编辑持续完善本体；这些编辑只有在经过基于骨干模型的成对评估后才会被接受。在三个广泛采用的数据智能体基准和四个 LLM 骨干模型上的实验表明，EvoOntology 始终优于强基线方法和现有语义层方法，有效弥合了智能体—数据鸿沟，并实现了与异构数据更高效的交互。代码：https://github.com/ruc-datalab/EvoOntology
离线强化学习旨在仅从固定数据集中学习策略，而这些数据集通常包含多模态动作分布。流策略能够自然地表示此类多模态行为，但如何高效地学习一步流策略仍然具有挑战性：标准的价值引导通常会导致模式坍塌，或在分布外区域利用价值高估偏差。为此，我们提出了基于最优传输的一步流策略（OptiFlow），一种将一步流策略学习视为结构化样本分配问题的框架。OptiFlow 联合训练一个具有价值感知能力的参考流策略和一个高效的一步策略，并通过逐状态熵正则化最优传输将二者的动作样本耦合起来。对于每个状态，评论家估计的价值决定蒸馏目标动作的优先级，而动作距离代价则确保配对在几何上相容。通过避免直接最大化评论家价值，我们的传输引导方法以高价值且有数据集支持的模式为锚点，使一步策略能够在分布内进行利用，同时避免分布外发散的风险。实验结果表明，OptiFlow 能够有效捕捉最优多模态行为，并在多种离线强化学习基准测试中取得出色表现。我们的代码可在 https://github.com/Yonsei-DILLab/OptiFlow 获取。
人工智能给企业带来了一个不同寻常的时机决策问题。技术前沿正在快速改进，实施具有部分不可逆性，而组织特定能力则通过行动积累起来。本文构建了一个不确定性下 AI 部署的两期决策模型，在该模型中，企业可在立即部署、有限试点和等待之间进行选择。部署能够获得当前运营价值，但会使企业面临架构过时的风险；等待则保留了在观察到技术前沿后再采用的期权；试点则牺牲当前运营价值，以在不作出全面承诺的情况下构建组织特定的学习能力。该模型得出了五项核心时机结论，以及一项关于学习发生位置的比较结论。第一，当部署收益关于技术前沿呈仿射关系时，保持均值不变地增加技术前沿不确定性，会提高等待和试点的价值，但不改变立即部署的价值。第二，当已部署架构只能吸收未来改进的有限份额时，更快的预期技术前沿进步可能降低立即部署的相对吸引力。第三，当试点所构建能力的预期价值超过其成本时，试点严格优于等待。第四，足够有价值的组织特定学习会形成一个非空区域，使“早试点，晚承诺”成为最优策略。第五，存在一个闭式模块化阈值，超过该阈值后，立即部署优于最佳外部选项。第六，生产学习和试点特定学习对时机裕度的影响不同。连续时间扩展得出了标准结论：不确定性会提高采用阈值，而能力和模块化会降低采用阈值。本文将部署、试验和等待区分开来，并说明为什么快速进步能够理性地增加试验，却不足以证明不可逆承诺的合理性。
嗨，HN，我是 Nari Labs 的 Toby。 我们一直在努力让 OSS 语音模型实现超高速运行。去年，我们构建了 Dia，这是首个能够进行自然对话的 OSS 文本转语音模型。从那以后，又有许多优秀的语音模型向公众发布。 但市场仍由闭源模型主导。我们认为，问题出在推理上。现有的 vLLM / SGLang 等系统并不适合多模态推理。为了证明这一点，我们构建了一个专门针对 Qwen3-TTS 的推理引擎，并将其开源了（ https://github.com/nari-labs/nari-qwen3-tts ）。在 10 RPS 下实现低于 50 ms 的延迟，这表明开源模型可以运行得更快、成本也更低。 从那以后，我们一直在努力为所有人提供廉价、快速且高质量的服务。我们甚至在闭源模型擅长的领域击败了它们！ 在广受引用的 Coval（YC S24）语音 AI 基准测试中，与 11Labs、Cartesia 等相比，我们的 Qwen3-TTS 端点不仅延迟排名第 2，准确率（WER）更排名第 1，同时还是成本最低的端点。我们的 Qwen3-ASR 端点延迟最低，准确率排名第 2，仅比第 1 名低 0.1%。它是榜单上成本第二低的模型。 为了让这些模型运行迅速、表现出色，同时保持较低成本，我们在推理工程方面投入了大量巧思。有趣的是，与我们的端点相比，Alibaba 的官方端点在准确率和延迟方面似乎表现更差。不过，我们仍然非常感谢 Qwen 团队将这些出色的语音模型 OSS 化。 我们希望继续降低价格，让语音技术成为一种大众化商品——这样每个应用都能拥有出色的 TTS 和 STT，而不必担心单位成本。我们也在研究音频领域的其他部分，例如说话人分离，以及视频和世界模型推理。敬请期待更多内容！ 评论 URL： https://news.ycombinator.com/item?id=49699267 积分：7 评论数：1
文章 URL： https://dial9-rs.github.io/blog/principles-for-fast-tokio-applications/ 评论 URL： https://news.ycombinator.com/item?id=49698607 得分：8 评论数：0
文章链接： https://github.com/ActuallyTaylor/strata/blob/main/paper/data/large/datasets/ai-assisted-repositories.csv 评论链接： https://news.ycombinator.com/item?id=49697727 积分：8 评论数：6
文章 URL： https://rietta.com/blog/rubygems-supply-chain-openai/ 评论 URL： https://news.ycombinator.com/item?id=49697666 分数：5 # 评论数：0
What is ChaosTree? ChaosTree is a zero-dependency Java Sorted Set/Map library built around multiple tree implementations. It provides implementations of: - AVL Tree - Red-Black Tree - B-Tree - B+Tree I have not made my custom API as it implements NavigableSet, NavigableMap, SequencedSet and SequencedMap. My Custome API are: - buildFromSorted(Iterator> it, float factor) - importFlatMatrix(Object[][] blast, float factor) - Object[][] exportFlatMatrix() I started ChaosTree because I wanted to explore how far I could learn DS structure, It started as frst release of 7 tree of Set type with Custome API, As i got engrossed into it I moved to Map, During phase of switching from Set to Map there was drastic change and knowledge upgrade where I pushed my Tree limit from jdk11+ support to jdk21+ where paid close attention to dependency-free and memory layout, allocation,JVM behavior, and real-world performance. Some of the things I experimented with: - Different node layouts and metadata footprints - CRTP/F-bounded polymorphism for tree implementations - Parent-pointer vs parentless nodes - Array-based N-ary tree nodes - B-Tree/B+Tree degree selection - JMH benchmarking and JFR profiling - Differential/randomized testing against java.util.TreeMap/TreeSet After this rough and tough It also passes these test: -Guava Testlib compatibility testing -jqwik property-based testing -Randomized differential testing against reference collections -White-box structural validation of tree nodes -Direct validation of B-Tree/B+Tree structural invariants -Exception and iterator-contract testing -Serialization and cloning tests -Nary Tree uses custom jqwick test of API validation as well. Tail latency behavior is not shown here because it get's truncated into a simple text which made the dat read wrong: https://chaos-vy.github.io/ChaosTree/utils/JMH-Report.html I have also ran benchmark with official JDK TreeMapUpdate to my N-ary tree. GitHub: https://github.com/Chaos-vy/ChaosTree https://chaos-vy.github.io/ChaosTree/ I'd especially like feedback on the API design, implementation choices, benchmark methodology. I am currently trying to truncate useless and complex branches for performnace tuning. Comments URL: https://news.ycombinator.com/item?id=49694404 Points: 6 # Comments: 4
a couple hours ago i had astra take on a 10 month old repo of mine https://github.com/callbacked/neural-band-poc/ that involved finding a way to use the meta neural band independently from the glasses to read the sEMG data off of it. i really think it is a cool piece of technology that largely gets overlooked by the glasses, and always thought it was a shame it couldn't be used outside of its main application. after making quick work of that, i wanted to use it to control my mac using gesture controls and that's how kinesis was born. it has been very cool being able to swipe from desktops using my fingers, swiping up to open up mission control or controlling a virtual knob to adjust my volume. i definitely havent scratched the surface with the neural band but it has been a really fun endeavor thus far! i really hope others can find creative uses for their neural bands too :) Comments URL: https://news.ycombinator.com/item?id=49695408 Points: 7 # Comments: 0
In urban search and rescue, articulated tracked robots (ATRs) must traverse structured but contact-rich environments such as stairwells and cluttered building interiors. Reliable autonomy remains challenging because robot-terrain interaction (RTI) is hybrid and discontinuous, and effective flipper-track coordination is difficult to model analytically. We present ASTRIL-MPC, a language-guided neural kinematics model predictive control (MPC) framework for autonomous traversal. A learned kinematics model predicts short-horizon task-state increments from a height sequence and recent trajectories; NMPC plans with multi-objective costs and strict feasibility constraints; and a large language model (LLM) proposes bounded updates to selected weights and bounds through a safety-checked interface with range clipping, rate limiting, and consistency checks. The compiled predictor enables a full control cycle within 100 ms. Across three traversal tasks and a multi-height generalization setting, ASTRIL-MPC improves an aggregate traversal-quality score by up to 71% over a non-adaptive NMPC and by 67% over a PPO baseline, while eliminating measurable collision impacts during descent. These results indicate that combining learned kinematics, optimization-based planning, and language-guided retuning yields data-efficient and robust autonomy for articulated tracked robots.
Dance-to-music (D2M) generation aims to synthesize music that is rhythmically and stylistically aligned with dance videos. A key challenge arises from the semantic mismatch between sparse dance cues, such as rhythm and style, and the dense information required for music composition, including structure, instrumentation, and expressive dynamics. Existing methods typically rely on these sparse cues and supervise only the final audio output, resulting in poorly learned music representations and generated music with limited musicality and structural coherence. To address these issues, we propose Curriculum-guided Multi-scale representation Alignment with scale-aware Optimal Transport (CMA-OT), a novel paradigm that leverages an external music expert to provide hierarchical supervision for the generator's latent features, bridging the semantic gap and enhancing representation learning. To effectively incorporate hierarchical supervision, we introduce a curriculum-guided multi-scale learning strategy that progressively transfers musical knowledge from the expert to the music generator, enabling stable and effective representation learning. Moreover, to accommodate the semantic and structural variations across different expert scales and achieve fine-grained alignment under temporal mismatch, we propose a scale-aware optimal transport alignment mechanism, which models soft correspondences between hierarchical expert representations and the generator's latent features. Extensive experiments on two datasets demonstrate that CMA-OT achieves state-of-the-art performance in rhythmic synchronization, perceptual quality, and overall music generation.
高能力开放权重小型语言模型（SLM）的日益普及，使先进网络安全能力得以民主化；由于这些模型在本地部署时能够绕过专有 API 的安全防护，这也带来了日益加剧的风险。然而，作为自主智能体部署的 SLM 往往难以应对网络安全夺旗（Capture The Flag，CTF）挑战这类长程探索任务，原因在于上下文膨胀，以及累积的工具调用输出导致的认知退化。为理解并缓解这一网络安全威胁，我们引入了\textit{上下文分段}，这是一种两级智能体框架，可将复杂的漏洞利用任务划分为易于管理且在上下文上相互隔离的子问题。我们在使用受内存限制的 \texttt{gemma-4} 模型的 \texttt{picoCTF} 数据集上进行评估，结果表明，对于 E4B 模型，我们的策略能够充当一种智能搜索：与暴力重试相比，在具备更高令牌效率的同时取得具有竞争力的奖励，并成功解决了标准智能体执行无法完成的 18.52\% 的任务。代码可在 https://github.com/9xeb/context-segmentation 获取。
将观测到的系统行为映射至 MITRE ATT&CK 等标准化框架，对于威胁知情防御至关重要，但这一过程目前仍主要依赖人工完成。现有的自动化方法依赖网络威胁情报报告，而此类报告仅提供对攻击活动的事后描述。相比之下，低层级遥测数据，即内核级系统调用，能够直接提供对攻击者行为的证据，但其规模和复杂性限制了其在自动化映射中的应用。我们提出了一种方法，通过 eBPF 收集内核级事件，将攻击者命令关联起来构建溯源图，并提取适用于基于 LLM 推理的紧凑图表示。随后，我们结合纯 LLM 提示和以 ATT&CK 知识库为基础的检索增强生成（RAG），将这些表示映射至 MITRE ATT&CK 框架，生成按排名排列的技术候选项及其支持性理由。我们将该方法实现为名为 Trace2ATT&CK 的端到端流水线，并使用本地部署的开放权重 LLM，在 347 个 Linux Atomic Red Team 测试上对其进行评估。与纯提示相比，RAG 持续提升了 ATT&CK 映射性能；与此同时，溯源图的表现显著优于原始遥测数据。结果表明，基于图的行为描述结合本地推理，能够在不损害数据机密性的前提下，使基于内核级遥测数据的自动化 ATT&CK 映射具备实际运行可行性。
大规模多标签文本分类需要从包含数千或数万个候选标签的词表中，为每篇文档分配一小部分相关标签。尽管预训练语言模型提升了文本的语义表示能力，但大多数基于表示的方法都将其预测流程集中于一个主编码器，或在单个排序器中融合辅助特征。因此，异构语言模型之间的互补性仍未得到充分探索。我们提出了 DualMLC，这是一种双分支框架，通过自回归的仅解码器语言模型和双向编码器分别处理同一篇文档。每个分支都保留独立的表示路径，并在共享标签空间上独立估计相关性分数。DualMLC 通过后期 logit 融合结合两个分支的分数向量，使共享证据能够强化相关标签，同时使分支特有的证据能够弥补另一分支表示的局限性。DualMLC 在三个广泛使用的大规模多标签文本分类基准数据集上取得了当前最优的结果。消融实验结果进一步证实，融合异构预测器所产生的排序优于任一单独分支。源代码已公开发布于 https://github.com/huiyegit/DualMLC。
基于文本的行人搜索（TBPS）旨在根据给定的自然语言描述，从大规模图像库中检索目标人物的图像。现有大多数方法依赖于使用人工标注的图像-文本对进行监督学习。本文探索仅使用无标注图像的无监督TBPS。我们提出了GTR+，一种先生成后检索的两阶段框架。在生成阶段，我们引入了一种分层描述生成框架，通过三级顺序处理生成细粒度且风格多样的文本描述：基础层利用自动问答机制生成基本的视觉属性描述；中间层通过样本间对比机制增强细粒度细节；高级层则通过风格化扩展机制进一步丰富文本多样性。在检索阶段，为减轻噪声伪文本的影响，我们设计了一种自适应置信度加权检索学习框架。我们使用Gaussian Mixture Model将图像-文本对建模为干净样本或噪声样本，并结合实时图像-文本相似度以及前一阶段生成的静态文本概率进行校准，从而在训练过程中获得自适应的样本权重。此外，我们还贡献了LargeFine-Person，这是一个大规模TBPS数据集，包含高质量、细粒度且多样化的文本标注，可在无监督设置下为实际应用和具备泛化能力的TBPS预训练提供基准。多个TBPS基准上的实验表明，GTR+和LargeFine-Person均具有良好的有效性和泛化能力。代码地址：https://github.com/Flame-Chasers/GTR】【。
As a key model compression technique, knowledge distillation aims to transfer knowledge from a high-capacity teacher model to a lightweight student model for enhancing the latter's performance. In this work, we reviewed the feature knowledge distillation for 3D-CNNs and observed that most feature distillation methods in video analysis are simple adaptations of those used in image analysis, often neglecting the differences of video features in the temporal dimension. To address this issue, we proposed Label-Guided Knowledge Distillation (LGKD) to guide the distillation of student model features using ground truth labels. Our method entails two components: sample-wise distillation and class-wise distillation, enabling the student model to learn feature representation of the teacher model at two levels. Sample-wise distillation utilizes label information and the teacher's probability distribution to guide the learning of features that significantly impact temporal accuracy while mitigating noise. Meanwhile, class-wise feature distillation employs a prototype network to further capture the relational knowledge among samples within the same category, enhancing the student's ability to learn higher-dimensional semantic information and improving model generalization. To demonstrate the effectiveness and superiority of our method, we conducted comprehensive experiments on two benchmark action recognition datasets, UCF101 and HMDB51, achieving competitive results.
Symmetries play a central role in reducing the complexity of reinforcement learning problems, yet most existing approaches rely on fixed group actions or predefined state abstractions. Classical reinforcement learning algorithms typically assume a globally structured Markov decision process with uniformly applicable actions and transitions, an assumption that limits their ability to exploit modularity and local, context-dependent regularities present in many realistic environments. We propose a reinforcement learning framework using groupoids to capture local, state-dependent symmetries and support the dy- namic discovery of equivalence structures during interaction. The agent maintains orbit representatives together with transporters that map raw states to canonical forms, enabling learning and decision-making to be performed in a symmetry-reduced space while preserving local distinctions. Empirical results demonstrate that the proposed groupoid-based approach improves sample efficiency and convergence in dense and large-scale environments exhibiting strong partial symmetries, yielding substantial performance gains over standard Q-learning. These findings show that dynamically exploiting local symmetry provides a practical and mathematically principled route to scalable and generalisable reinforcement learning.
Humans organize knowledge into a taxonomy of concepts with nested levels of abstraction and a \emph{basic level} at which people recognize and name objects with the least cognitive effort. Cobweb is a classic cognitive account of this ability, an incremental learner that builds a probabilistic concept hierarchy by maximizing category utility. We argue that diffusion models, although designed for image synthesis, implicitly perform the same computation. The noisy marginals of a diffusion model are Gaussian smoothings of the data distribution, and the modes of these marginals form a hierarchy that corresponds to a Cobweb tree of probabilistic prototypes in four respects. Both are hierarchical density models, both are hierarchical-Bayesian models with Gaussian prototypes, both treat categorization as score-following that reduces uncertainty, and in both a basic level emerges. We locate this basic level for a diffusion model at an intermediate noise level, where recent analyses show that the reverse process commits to the class identity of a sample. The two models differ mainly in how they represent and learn the taxonomy. Cobweb learns a discrete tree incrementally, whereas a diffusion model encodes a continuous, interpolable hierarchy in a single learned score field fit to the data distribution. We test the correspondence on MNIST and Fashion-MNIST by recovering the diffusion hierarchy through mode-finding and comparing the basic levels of the two models. This reframes diffusion as a cognitive model of concept formation and offers Cobweb a continuous, scalable instantiation.
近年来，基于 LLM 的系统及其在问题解决和编程方面能力的突破，推动了“AI for Science”范式的发展，并有望取代人类在机器学习（ML）研究中的角色。然而，尽管已有多个全自主端到端 ML 研究框架被提出，但它们的成功实现通常局限于搜索空间狭窄的问题，例如语言建模或生物医学 ML 基准测试。本文通过研究一个案例——电信工单检索——探讨如何将自主研究应用于解决开放式、工业级 ML 问题。电信工单检索是一项开放式任务，在表征、架构和训练数据生成方面具有较大的自由度。我们发现，使用商业和开源智能体开展开放式问题的自主研究既展现出潜力，也存在局限：自主研究能够在狭窄的超参数优化任务中表现出色，但缺乏类似人类的直觉和创造力，并且需要额外的运营开销。即使仅接受最低限度的人类监督，自主研究也能以适中的成本（每次 Cursor 运行最高 200 美元），在远短于人类研究所需的时间内（10 周，而非人类工作的 10 个月）达到当前最先进水平性能的 90%（Recall@1 为 0.34，而非 0.38）。我们的实证证据表明，在 ML 研究中，人类研究人员与自主研究框架协同工作才能取得最佳效果。
文章链接： https://www.interconnects.ai/p/open-source-ai-reading-list 评论链接： https://news.ycombinator.com/item?id=49690260 得分：13 评论数：0
I have a BS/MS in computer science and somewhere around 9 years of experience, where the first half was spent working as a backend developer and the second half I've been working in product owner roles. I switched from development to product because I simply realized that I wasn't very good at implementation and system design work to be honest. I wanted to work with tech from a more overarching level, and let the experts who were much more skilled than me handle the implementation work. Around 8 months ago I switched companies and am still working as a PO, but this organization I am working in is a terrible fit for me. I don't want to go too far into why I really dislike this job, but a big part of it is that the organization wants me to make technical design decisions which I am just really not good at, since the development team doesn’t have this experience themselves. This is making it completely impossible to deliver on the project I’m responsible for. This has really made me think about what I should do for my next role. I fully admit that I was and still am a mediocre developer, which is why I moved into a PO role in the first place around 4 years ago. I have such a huge amount of respect for people who are great developers and architects. It's a challenging role and I can imagine it's exciting if you are creative and like to build new things. But I am just not cut out for it. I am not very creative, and I often have no clue where to even start when I need to build new things. The parts I've enjoyed the most in the various roles in my career have been things like: - 2nd/3rd line ticket management and troubleshooting. It's been quite fun going through logs and trying to figure out what went wrong, perhaps trying to recreate the situation in a test environment with an API client. - Supporting customers. For example helping customers with installations of software. It honestly gives me a huge sense of accomplishment when I get to see a customer making use of our products, and helping them investigate why something may not be working. - Being an interface between business and tech. I've always thought of myself as someone who is able to understand both business needs and technical needs quite well. Even though I'm not good at implementing software, I can understand limitations of software, read and understand existing codebases, and so on. - Testing software. It's a lot of fun to test software to see if there are any issues before we go live! - Just working with people in general! I really like to do whatever I can to support my team and the people that I work with. I thought I'd ask for some advice to see if you guys have any recommendations for career tracks to consider, given that I am just much more comfortable supporting existing software than building new software. I like working with people. I like achieving results. I'm just not cut out for system design and building new software. Honestly it might mean that I need to live with a lower salary, and that is fine with me. To be completely honest, given the fact that my current role is genuinely making me sick with severe anxiety (made a post about this a while back [1], I've learned that money isn't everything. There isn't a point to making more money if my job is literally killing me and I am losing my health. I'd much rather work with something I enjoy that isn't making me feel sick. Life is too short. Thanks for reading this. I really appreciate any advice or thoughts you may have. [1] https://news.ycombinator.com/item?id=49533073 Comments URL: https://news.ycombinator.com/item?id=49682920 Points: 9 # Comments: 2
文章 URL： https://github.com/threatexpert/gonc 评论 URL： https://news.ycombinator.com/item?id=49654148 得分：9 评论数：0
文章链接： https://yifanzhang-pro.github.io/recurrent-looped-tranformer/ 评论链接： https://news.ycombinator.com/item?id=49678548 得分：5 评论数：0
I am running a small AI company that has been building an offline security audit app, simply a fine-tuned small language model powering an Electron app that can audit GitHub and local code repos for security stuff. I always see these LinkedIn or X guys making $500k ARR or $1m ARR within months; how the fkk do they get that many customers without spending on ads? I can't even get 5 customers; damn it, I do suck at sales Comments URL: https://news.ycombinator.com/item?id=49674396 Points: 3 # Comments: 3
Hi HN, the title could be a little bit aggressive sorry for that... Basically i am Computer engineer graduated recently but i was and still working remotely for approx. 3 years with the same foreign company in EU. and i noticed that we are always somehow lose the context of a ticket -- even though we are using the popular project management systems -- so i decided why not build something that can fix our context loosing problems and with some improvements. here i created Liniora. an AI powered project management system. basically the core idea is to link the tools that we use to gather information about tickets in one place. i integrated Google meet for the meetings transcripts so you do your meeting it automatically got registered and indexed. You discuss about a bug in Slack thread so you mention @Liniora app and it will handle the bug context and create a ticket for you. We have the version control also connected so you create a ticket with ability of creating a separate branch with one-click for both Gitlab and Github. you have as well the Pull request controlled in our place inside the app. lets suppose you have a ticket and you created a branch called 'USER-325-create-login-form' then you merged to staging or production ? it will be automatically flagged with staging/production badge so you know what tickets are done. there are a lot of interesting features i wanted to describe but i will just let you guys discover and i will be very happy about your feedback. Thanks Comments URL: https://news.ycombinator.com/item?id=49671542 Points: 5 # Comments: 2
文章 URL： https://eiln.github.io/posts/ane.html 评论 URL： https://news.ycombinator.com/item?id=49670032 得分：6 评论数：0
文章链接： https://github.com/GrapheneOS/Messaging/releases/tag/13 评论链接： https://news.ycombinator.com/item?id=49663373 得分：18 评论数：2
文章链接： https://capitalbnews.org/data-centers-permit-rules-epa/ 评论链接： https://news.ycombinator.com/item?id=49662672 得分：45 评论数：16
文章网址： https://rune.build/blog/rune-is-now-open-source 评论网址： https://news.ycombinator.com/item?id=49660149 得分：16 评论数：1
Over past couple months I noticed that HN feed is almost exclusively AI or AI-adjacent news. Meanwhile the legitimately, broadly-hacker stuff gets left out for the most part. I noticed that because the things I find genuinely interesting that I post here now get zero traction, which is the stuff that I believe would previously be met with some discussion. Note that I don't care that my submissions were ignored, I don't need that validation. Problem is that if my stuff gains no traction, I am myself not seeing similar stuff posted by others, stuff I am genuinely also interested in and would want to hear about from this community. Here are two most recent examples: - Lenovo showed a model that incorporates solid-state air cooling. That technology has been around for a while but it looks like it matured and is production-ready. This allows for super-light and slim designs, which, in turn, should allow for bigger batteries installed and I think it's some sort of breakthrough: https://news.ycombinator.com/item?id=49580229 - Brax Industries devices, genuinely interesting, open-soruce and innovative, especially the wall display, with its modularity and on-device processing compatible with Home Assistant. I also appreciate the very clean design: https://news.ycombinator.com/item?id=49643215 As said, I honestly believe this would gain some traction here on HN a year or two ago. I no longer find HN a good source of "what's new" in the industry, it's almost exclusively AI talk now. Comments URL: https://news.ycombinator.com/item?id=49657850 Points: 76 # Comments: 15
文章链接： https://jayvisaria.github.io/LLM-Visualizer/#/dashboard 评论链接： https://news.ycombinator.com/item?id=49652996 得分：5 评论数：0
Artificial neural networks rely on vector-matrix multiplications (VMMs), whose implementation in von Neumann architectures is dominated by costly data movement between memory and processing units. Spiking neural networks (SNNs) mitigate this bottleneck by performing in-memory, analog VMMs using memristive crossbar arrays. However, conventional current-mode readout circuits incur significant area and power overhead. This work proposes a fully analog readout architecture based on voltage-to-time conversion of the VMM output. By sensing the column voltage, the proposed approach avoids current-mode summing and scaling circuitry, improving area and energy efficiency. Post-layout simulations of a 10x1 SNN implemented in a 130 nm CMOS technology validate the proposed architecture, while application to a trained 64x10 SNN for digit classification further demonstrates its feasibility for SNN inference.
Visual Autoregressive Models (VAR) generate images through next-scale prediction, producing all tokens within each scale in parallel. We show that this parallel decoding constitutes a mean-field-style approximation that discards spatial dependencies among same-scale tokens, causing locally incoherent samples regardless of backbone capacity -- a limitation of the decoding rule. Addressing this limitation, we introduce the Logit Refiner, a lightweight autoregressive module that restores intra-scale dependencies by sequentially sampling tokens conditioned on frozen backbone features. Adding only ~10% parameters and less than 5% of the base model's training compute, it plugs into any pretrained VAR checkpoint without retraining. Controlled ablations isolate joint intra-scale sampling -- rather than additional capacity or training -- as the critical ingredient. Across backbones from 310M to 2B parameters on class-conditional ImageNet 256x256, the refiner consistently improves generation quality, enabling a 1.1B-parameter model to surpass one twice its size. The approach further generalizes to text-to-image generation, confirming that the mean-field bottleneck persists across VAR variants and is effectively alleviated by our method. Project page: https://compvis.github.io/logit-refiner/
Energy consumption forecasting relies on increasingly complex machine learning (ML) models, such as Genetic Programming-based symbolic regressors, whose predictions can be difficult for facility managers and building operators to interpret. Explainable Artificial Intelligence (XAI) techniques address this opacity, but traditional XAI dashboards require substantial technical expertise and provide limited flexibility for dynamic, context-aware inquiry. Conversational XAI systems offer a promising alternative; however, previous approaches, such as TalkToModel, were constrained by rigid custom grammars and achieved only 76.8% intent-parsing accuracy. This paper introduces the Explainability Assistant, an open-source conversational XAI system that leverages the function-calling capabilities of modern Large Language Models (LLMs) to overcome these limitations. The system achieves 94% intent-parsing accuracy, supports flexible natural language interaction, and adapts to different ML problem types without task-specific fine-tuning. We present the system's architecture and report results from a comparative evaluation conducted with energy domain specialists, contrasting the Explainability Assistant with a traditional XAI dashboard. The evaluation suggests improved usability and consistent task accuracy, with all experts unanimously preferring the conversational interface for practical use.
Recently, a wide range of recommendation algorithms inspired by deep learning techniques have emerged as the performance leaders on several standard recommendation benchmarks. While these algorithms were built on different DL techniques (e.g., dropouts, autoencoder), they have similar performance and even similar cost functions. This paper studies whether the models' comparable performance are sheer coincidence, or they can be unified under a single framework. We find that all linear performance leaders effectively add only a nuclear-norm based regularizer, or a Frobenius-norm based regularizer. The former ones possess a (surprising) rigid structure that limits the models' predictive power but their solutions are low rank and have closed form. The latter ones are more expressive and more efficient for recommendation but their solutions are either full-rank or require executing hard-to-tune numeric procedures such as ADMM. Along this line of finding, we further propose two low-rank, closed-form solutions, derived from carefully generalizing Frobenius-norm based regularizers. The new solutions get the best of both nuclear-norm and Frobenius-norm world.
Many biological discovery problems require experiments to be selected sequentially under constrained budgets. CRISPR screening is a prominent example, as exhaustive perturbation testing is often infeasible and candidate perturbations must instead be prioritized over multiple experimental rounds. Despite the importance of this problem, existing benchmarks for adaptive hit discovery remain limited in scale and diversity. Here, we introduce AssayBench-Loop, a large-scale benchmark for adaptive hit discovery comprising 1,389 CRISPR screens across five phenotype categories. Beyond enabling systematic evaluation, its scale makes it possible to learn acquisition strategies across historical experiments. Building on this resource, we introduce AssayLoop, a sequential experimental design framework combining AssayFormer, a transformer-based amortized acquisition policy trained across historical screens to adapt from experimental feedback, with LLM-derived biological priors through an adaptive handoff. In this view, completed experiments become training data for learning how accumulated evidence should guide what to test next, while LLMs provide prior biological knowledge to seed the search. We further introduce AssayLLM, showing that the same principle can be extended directly to an LLM through task-specific post-training. On temporally held-out screens, AssayLoop achieves a 5.67-fold enrichment over random selection and recovers 27.7% of hits after assaying approximately 5% of the candidate library, outperforming existing adaptive-design methods and standalone LLMs, and AssayFormer alone. Performance improves with increasing historical training data and transfers to phenotype categories excluded from training. These results demonstrate the value of learning acquisition policies across historical experiments and combining them with broad biological priors for efficient adaptive hit discovery.
大语言模型能够生成流畅的文本，但其中可能包含不忠实的陈述——这一现象称为幻觉。我们提出了一种多信号检测流程，将经过微调的 DeBERTa-v3 分类、Monte Carlo (MC) Dropout 不确定性量化以及温度缩放校准相结合，用于响应级幻觉检测。在 HaluEval 基准测试中，该流程在通用领域任务上取得了 F1=0.915 和 AUROC=0.977 的成绩，各任务的 F1 分数分别为：0.97（问答）、0.96（摘要）和 0.82（对话）。MC Dropout 推理进一步将准确率提升至 93.2%。上下文消融研究证实，该模型执行的是真正的蕴含推理，而不是利用表面模式；移除知识上下文后，摘要任务的 F1 分数下降了 24%。学习曲线分析表明，使用 25% 的训练数据即可达到完整数据训练效果的 77%。除检测之外，我们还对 Qwen2.5-0.5B 生成模型应用了 Direct Preference Optimization (DPO)，根据我们的检测器测量结果，将其幻觉率从 85.5% 降至 37.7%，相对降低 55.9%。在 SciFact 生物医学基准测试上的跨领域评估显示，通用领域训练的迁移效果较差（F1=0.52），这说明有必要进行领域特定微调。在 SciFact 上进行微调的 PubMedBERT 达到 F1=0.63 和 AUROC=0.81，表明领域匹配的预训练是最有效的适应策略。代码和模型可在 https://github.com/varunteja99/hallucination-detection-nlp 获取
空间推理不仅依赖距离、角度和形状等度量属性，还依赖在连续变形下保持不变的拓扑关系。认知科学将这些关系视为空间理解的基础，然而，对基础模型的评测主要聚焦于度量关系或依赖视角的关系。我们推出 MindTopo，这是一个基于认知科学和形式拓扑学、涵盖五种属性的拓扑直觉基准：连续性、分离性、次序、包围和纽结。MindTopo 在两个认知层次上评估每种属性。推理任务要求模型识别拓扑关系，或推断这些关系如何变化。规划任务则将基础模型实例化为闭环智能体，由其策略选择环境动作。MindTopo 包含 13 种程序生成的任务类型，共 11,030 个实例，并支持难度控制。我们对 14 个 MLLMs 进行基准测试，并研究了加入图像和视频生成能力的智能体配置，其中包括规划设置下的 3 个视频生成模型。每个 MLLM 在推理任务上的表现都优于规划任务，且表现最佳的模型仍远低于观察到的人类表现。在 Qwen3-VL-2B-Instruct 上，有监督微调和强化学习对推理能力的提升大于对规划能力的提升。生成的观测结果保留了局部线索并到达了看似合理的终点，但经审查的运行轨迹并不能可靠地遵循环境动力学，也无法在状态转换过程中保持拓扑结构。我们的网站位于 https://mind-topo.github.io/。
反事实遗憾最小化（CFR）是少数在 CPU 上运行速度仍快于 GPU 的大型数值工作负载之一。每次迭代都会遍历一棵包含多达数十亿个状态的博弈树，并通过通用树接口发起数百万个相互依赖的小型 gather 和 scatter 步骤。在 GPU 上，每个内核只需几微秒即可完成，因此内核启动和框架调度占据了运行时间的主要部分，已有的 GPU 实现性能不及经过优化的 CPU 代码。我们发现，对于固定博弈，CFR 迭代中除数值外的所有信息，在第一次迭代运行前就已经确定。基于这一观察，我们提出了 GPU-CFR，这是一种编译器和运行时系统。它将任意博弈一次性编译为静态数据流：扁平化的边数组和信息集数组、预计算的索引，以及按深度分批执行的过程共同固定了完整的操作序列，而迭代之间只有求解器状态发生变化。静态机会折叠、按深度执行的代码块，以及双通道到达缓冲区，最多可将框架操作数量减少 18.1 倍。由于形状、索引和缓冲区地址始终不变，CUDA Graph Replay 只需记录一次迭代，之后即可通过单次图启动对其进行重放。在一张 A100 上，针对涵盖纸牌游戏、骰子游戏和棋盘游戏的八个游戏组成的测试套件，GPU-CFR 的运行速度达到同一加速器上此前最快 GPU CFR 的 29.8--80.4 倍；在其中规模最大的四个游戏上，其速度达到 LiteEFG（最快的开源 CPU 实现之一）的 14--258 倍。编译后的表示承担了其中大部分性能优势：即使不使用加速器、仅采用八个 CPU 线程，它也已经比 GPU 基线快 2.2--51.1 倍。在 CPU 上，优化路径能够逐比特复现参考迭代结果，并且树构建和图捕获在第一次求解过程中就能收回开销。在不改变更新规则的情况下，GPU-CFR 在该测试套件的中型到大型游戏上击败了所有 CPU 和 GPU 基线。
本文提出了 DF-CAPTCHA，这是一种针对语音和视频通话中实时深度伪造冒充的主动防御机制。DF-CAPTCHA 不再被动地寻找伪造痕迹，而是要求通话者完成简单的挑战—响应任务：这些任务对人类而言很容易，但当前的实时深度伪造系统却难以生成令人信服的结果。该框架依据四项标准验证响应：真实感、身份一致性、任务完成情况和响应时间。我们通过用户研究以及使用实时深度伪造模型开展的实验，在音频和视频两种模态下评估了该方法。结果表明，人们通常难以区分实时深度伪造内容与真实媒体，而 DF-CAPTCHA 相较于被动方法显著提升了检测性能，在两种模态下均达到了较高的准确率。这些发现表明，基于主动挑战的验证是一种切实可行且稳健的防御手段，能够抵御基于实时深度伪造的下一代社会工程攻击。
降低音频编码器深度可以减少语音大语言模型的推理成本，但移除完整模块会扰动解码器所使用的嵌入，并可能导致删除错误和过早结束序列错误。我们提出 X-AuT，这是一种渐进式框架，通过短行为探测选择层组合，并结合表示对齐、跨尺度蒸馏、定时的学生策略监督以及 LoRA 微调来恢复剪枝后的模型。语言模型主干保持冻结，同时在蒸馏过程中调整注意力 LoRA 适配器和绑定的输出嵌入。训练采用转录一致性流程中一致性最高的数据层级，随后在微调阶段进行数据源重加权。在十个公开的中英基准测试上，将 Qwen3-ASR-0.6B 的音频编码器从 18 层压缩至 16 层后，宏平均错误率从 5.61% 降至 5.27%。14 层模型在音频塔参数减少 20.7% 的情况下达到 5.75% 的错误率。在匹配的训练方案下，1.7B 教师模型的平均错误率为 5.55%，而自蒸馏为 8.45%；渐进式 18$\rightarrow$14 剪枝优于直接剪枝（5.75% 对比 6.73%）。这些单次运行结果确立了两个实用的工作点，并表明准确率影响会因基准测试而异。项目网站：https://xpeng-ai.github.io/x-aut
Genetic Programming and its variants, such as grammatical evolution, are widely used in Symbolic Regression to derive mathematical expressions from multivariate data. In addition to predictive accuracy, models are appreciated for their potential to provide interpretability, offering explicit equations that relate input variables to outcomes. However, achieving interpretability and plausibility remains challenging, as evolved models may be complex or scientifically inconsistent. In this study, we explore whether Large Language Models, can assist in improving the explainability of Symbolic Regression models generated by evolutionary computation methods. Building upon our previous work on estimating body fat percentage using grammar-based Genetic Programming , we investigate the use of LLMs as post-processing tools to analyze and rank evolved expressions according to their interpretability and medical plausibility. Four symbolic expressions are analysed by three LLMs over three repeated runs, and the resulting interpretations and rankings are assessed by a panel of three clinicians. Across the three LLMs, comparative model-ranking outputs received more favorable clinician assessments than isolated term-level interpretations. However, the LLMs also produced physiologically and mathematically questionable explanations, indicating that they are better suited to comparative auditing under expert oversight than to autonomous validation.\blfootnote{The present work is an extended version of a paper submitted into a journal.
神经音频编解码器是现代语音生成系统的基础组件。尽管近年来的编解码器实现了越来越低的比特率，但降低帧率仍然具有挑战性，因为每个令牌都必须在保持重建质量的同时保留更多信息。我们提出了 ZipCodec，这是一种以 6.25 Hz 和 0.80 kbps 运行、理论延迟为 160 ms 的流式神经语音编解码器。我们的方法结合了大规模 WavLM 蒸馏、重新设计的基于 Transformer 的架构、标量球面量化器以及延迟感知的流式解码器。实验表明，在相当的比特率下，ZipCodec 在重建质量和下游任务方面均显著优于现有流式编解码器，同时运行于显著更低的帧率。尽管拥有 8.42 亿个参数，ZipCodec 仍可在消费级 CPU 上实现实时单流推理。演示样例、代码和检查点见 https://lucadellalib.github.io/zipcodec-web/。
文章链接： https://map-yue2.github.io/ 评论链接： https://news.ycombinator.com/item?id=49652028 得分：20 评论数：8
文章 URL： https://merybenavente.me/blog/proof-of-capture 评论 URL： https://news.ycombinator.com/item?id=49649222 得分：9 评论数：0
HN 大家好，我是来自 Feyn 的 Shreyash。我们帮助公司基于其数据构建定制模型。 今天我们发布 MultiMatte，这是一个可以用文字指向目标的背景移除模型。说出图像中的一个对象，MultiMatte 只保留那个对象，并移除其他所有内容。 用你的图像试试看： https://usefeyn.com/multimatte/ 。 演示视频： https://youtu.be/XZ5BJWAkOjs MultiMatte 是开源的。使用我们的 NoBg 库来构建： https://github.com/feyninc/nobg 。模型卡在这里： https://hf.co/feyninc/multimatte MultiMatte 是我们背景移除模型的第二次迭代。第一个是 FeyNoBg，我们曾在这里发布： https://news.ycombinator.com/item?id=49072462 。 最大的升级是可提示性。大多数模型在裁切背景时会保留所有前景元素。MultiMatte 让你可以用提示词指定想要保留的确切对象，并移除其他所有内容。例如，如果你有一张包含一只狗和一个碗的图像，你可以让 MultiMatte 只保留狗。 MultiMatte 基于 Meta 的 SAM 3 构建，SAM 3 是一个可用概念提示的检测器，已经能够理解短语。我们最大的改变在于遮罩。MultiMatte 不使用将每个像素分类为位于对象内部或外部的二值遮罩，而是使用 alpha mattes，为每个像素相对于某个对象分配一个不透明度值。这使我们能够更好地表现头发、毛发、运动模糊以及其他类型的模糊边界。 在我们测量的所有基准测试中，MultiMatte 相比 SAM 3 都显示出阶跃式提升。在 DIS5K 上，S-measure 从 0.674 提升到 0.908（相对提升 34.6%），在 DUT-OMRON 上从 0.792 提升到 0.901（13.7%）。 我们的博客介绍了更多训练细节和结果： https://usefeyn.com/blog/multimatte 。 很乐意回答任何问题！ 评论 URL： https://news.ycombinator.com/item?id=49645803 点数：9 评论数：1
Animation and VFX pre-production review requires teams to translate loosely specified creative intent--briefs, evolving specifications, heterogeneous references, and verbal decisions--into revisions that junior artists can execute without repeated clarification. In practice, criteria drift across iterations, review judgments lose their evidential basis, and the reasoning behind a request rarely survives the senior-junior handoff. We contribute a design framework for intent-evidence-action alignment: intent is articulated into a shared project record, judgments are anchored to grounded evidence, and authorized decisions are converted into clear revision tasks tied directly to reference notes. We instantiate this framework in MOONWALK, a professional pre-production review system comprising a shared intent record, reference/specification anchoring, structured work-in-progress comparison, and supervisor-authorized action planning. In this workflow, AI handles administrative coordination--flagging missing context and organizing notes--while artists retain full creative direction. An in-studio study with professional practitioners compares MOONWALK with a chat-only (chatbot) interface using matched production materials, while participants' existing workflows provide a retrospective ecological baseline. Results indicate stronger intent alignment, decision traceability, and checklist executability, while also showing that aesthetic authority and final prioritization must remain with practitioners. The evaluation establishes the value of the integrated structured workflow over unstructured conversational AI chatbot. Code: https://github.com/Akinesia112/Moonwalk/tree/english-version
现实世界中的 GUI 使用通常涉及跨越多个设备和平台的工作流，需要传递中间结果、维护共享状态，并在异构环境之间进行协调。然而，现有的 GUI 基准测试绝大多数仅评估智能体在单一设备上执行静态定义的任务，因此基本未考察这类跨设备能力，导致对智能体是否已具备现实世界使用条件的评估过于乐观。我们推出 JarvisGUI，这是一个动态基准，用于评估 GUI 智能体在跨设备工作流中的表现；这类工作流要求在包括 Android、Windows 和 Ubuntu 在内的异构平台之间进行协同交互。具体而言，JarvisGUI 在轻量级类型系统下将 GUI 任务表示为输入-输出转换，从而能够自动组合多步骤、跨设备的工作流，并在统一框架内动态评估智能体性能。通过在覆盖多个操作系统的虚拟环境中评估智能体，JarvisGUI 揭示出，当前最先进的开源 GUI 智能体在现实世界工作流所需的状态转移感知能力、跨平台上下文推理能力以及长时程依赖管理能力方面存在明显不足，暴露出一个现有基准测试无法发现的关键能力缺口。
Joint-Embedding Predictive Architecture (JEPA) 世界模型能够学习紧凑的世界潜在表征，以支持预测和规划，但其学习物理规律并生成符合物理现实的动力学的能力迄今尚未得到检验。在本文中，我们提出 SemiGroup-JEPA (SG-JEPA)，通过动作条件化将控制物理规律的参数提供给时间模型，并通过自回归潜在 rollout 联合训练编码器和预测器，从而扩展 LeWorldModel 框架。为了评估模型的分布外泛化能力，我们在不同重力场下设计了动力学任务。这些任务虽然遵循相同的物理定律，却呈现出定性上不同的动力学：从弱重力场中的漂浮运动，到强重力场中的快速弹跳。与 DINO-WM 相比，在二维数据集上，SG-JEPA 将开环预测误差最多降低至原来的二分之一；在三维机器人数据集上，控制成功率最高提升至 2.5 倍。对于后者，我们分别训练了独立的扩散策略。为解释这一优势，我们构建了一个线性特征模型，将局部物理定律条件误差与其在 rollout 过程中的递归放大区分开来。在该模型的指导下，我们发现，将多步 rollout 损失反向传播至表征，会训练编码器保留预测器能够向前传递的特征，而这些特征正是动力学所依赖的特征。因此，大部分性能增益来自编码器学习到更好的特征，而非预测器学习到更好的动力学。项目主页：https://sg-jepa.github.io。
Enterprises deploy systems, not checkpoints. Usable capability depends jointly on weights, serving route, precision, output contract, and harness, yet all 18 audited benchmarks score advertised model identifiers. We treat this as measurement error and give a protocol that makes it reportable. It has three parts. A gold-blind capability-binding preflight verifies that a route can execute the evaluation contract before any task reaches it; a reliability-inclusive first-pass scoring rule keeps failure in the score while keeping unsupported capability out; and adjudication is structurally score-blind. We call the protocol IB2 and release its algorithms, classification tables, request contract, and manifest schemas. Its reference instantiation, 128 locked tasks and 987 assertions over document, spreadsheet, chart, tool and database work, stays sealed: the procedure is the artifact, not the corpus. Across eleven systems, four results. Capability availability is measurable: two complete single-route runs on identical weights later failed distinct predicates of the finalized binding gate, while a third passed that gate before a fresh run. The advertised identifier exposed neither limit. Discrimination is not uniform: four of seven suites saturate under a six-system band, with the spread almost entirely from governed database work and multi-tab joins, so we report interval-backed resolution groups, not ranks; two of the nominal five-label output's four cuts fail multiplicity adjustment. Serving-arm choice moved one declared revision and precision from 77.38 to 82.54, paired interval [0.11,10.60], though the arms differ in access mode, harness generation, and the serving tool-call parser, and harness generation is a property of our evaluator, not any endpoint. Excluding failed responses from denominators changes the point ordering, so reliability inclusion changes a conclusion, not its wording.
基础视觉语言模型（VLM）展现出对现实世界的广泛智能，但如何将这种智能转化为机器人控制仍然充满挑战。我们提出了 Show-Harness，这是一种具身控制框架，通过连接意图与动作的紧凑语义接口，使 VLM 能够“操控”机器人。Show-Harness 提供了离散的语义动作单元，VLM 可以自然地对其进行推理；与此同时，针对不同具身形态的解释器会将这些动作单元确定性地落地为机器人在本地执行的动作，从而使 VLM 直接负责细粒度的物理决策。借助同一接口，Show-Harness 展示了以下两种应用的可行性：（1）直接解锁闭源前沿 VLM，实现零样本机器人控制；（2）仅需数个 GPU 小时的微调，即可适配小规模开源 VLM，实现低成本部署。我们进一步开发了 GUMI（GUI Manipulation Interface），将同一语义动作空间扩展到基于 GUI 的演示数据采集，使人类和智能体无需专用遥操作硬件，即可跨不同具身形态“操控”机器人。大量实验表明，配备 Show-Harness 的 VLM 智能体能够在不同任务、具身形态和环境之间实现稳健泛化，性能优于具有代表性的智能体范式和 VLA 范式。这些结果表明，合适的接口可以从基础 VLM 中释放出可观的具身能力，而无需增加模型容量，也无需进行代价高昂的具身专属预训练。
文章 URL： https://hugovergnes.github.io/little-lm-3-8b/ 评论 URL： https://news.ycombinator.com/item?id=49637435 积分：3 # 评论数：0
生物医学本体标准化将自由文本表达映射到标准化概念，从而实现生物医学数据的一致性整合与分析。由于词汇变异以及层级相关概念之间的细微差异可能模糊概念边界，这项任务仍然具有挑战性。我们提出了 OntologyAligner，这是一个包含三个阶段的框架，将本体对齐检索、大型语言模型候选重排序和选择性层级引导细化相结合。我们还构建了 PhenoNormBench，这是一个统一基准，包含来自七个 Human Phenotype Ontology 数据集的 13,390 个样本。OntologyAligner 在 HPO 标准化任务上取得了当前最佳性能，Macro Top-1 Accuracy 达到 88.78%，Micro Top-1 Accuracy 达到 86.75%，分别比最强基线高出 4.85 和 5.07 个百分点。消融分析表明，三个阶段均作出了互补贡献；敏感性分析则证明，该方法在不同候选集规模和模型骨干下均具有稳定性。在 MONDO、MEDIC 和 NCBITaxon 上的应用进一步验证了其向其他本体迁移的能力。OntologyAligner 为将生物医学文本准确映射到结构化本体概念提供了一个具有普适性的框架。PhenoNormBench 和代码已公开发布于 https://github.com/zhelishisongjie/OntologyAligner。
股骨滑车发育不良（TD）是股骨滑车的一种异常，与膝前疼痛和髌骨不稳定有关。滑车沟角度（SA）用于评估滑车形态，但通常仅在单张轴位 MR 切片上进行测量，且没有明确的切片选择指导，因此容易受到切片选择和标志点定位的影响。我们提出了一种基于超分辨率 MR 体数据、用于连续 SA 剖面分析的自动化框架。该框架利用隐式神经表示，将临床采集的轴位、冠状位和矢状位 MR 扫描结合起来，以重建高分辨率体数据。随后，使用两个用于标志点检测的 U-Net 模型，在整个滑车区域内计算 SA。我们在公开的 fastMRI 数据集和一个由 TD 患者组成的小型自建队列上对该方法进行了评估。与传统的手动单切片 SA 测量相比，所提出的自动化方法的平均绝对误差为 11.6$^\circ$，同时能够连续刻画滑车形态。群体层面的分析显示，公开队列与自建 TD 队列之间的平均 SA 剖面存在明显差异，凸显了基于剖面的评估方法在刻画 TD 方面的潜力。通过减少对手动选择单张轴位切片的依赖，所提出的框架将传统 SA 评估扩展为对滑车形态进行连续的、基于剖面的描述，无需额外成像，同时在概念上仍与当前临床评估相衔接。仍需进一步验证。代码可见：https://github.com/wehrlimi/SA_Profile。
In-context learning (ICL) is widely used in multimodal large language models (MLLMs) and achieves strong performance across a wide range of multimodal tasks. However, existing multimodal ICL methods often rely on surface level imitation of in-context demonstrations, making it difficult for MLLMs to align their responses with the reasoning path required by the given multimodal input. This limitation becomes more pronounced in complex multimodal tasks, thereby restricting further improvements in MLLM performance. To address this issue, we propose a new multimodal ICL framework that combines contrastive demonstration modeling with the self-refinement capability of MLLMs. Specifically, our framework reformulates each demonstration by explicitly contrasting a suboptimal response with a better response under the same input, together with a reasoning path that reveals how the response should be refined. This contrastive formulation makes the reasoning path toward the desired response more explicit and guides the MLLM beyond superficial imitation. Furthermore, because effective refinement depends on the current response, we introduce a response-conditioned retrieval mechanism to select demonstrations whose reasoning paths are more relevant to the current response. In addition, we use a lightweight alignment controller to predict response quality and determine whether further refinement is needed. Experiments on three types of multimodal tasks show that the proposed framework consistently improves MLLM performance, with particularly notable gains on visual question answering (VQA).
Feature transformation improves predictive performance on tabular data by constructing informative abstractions from raw features. Recent generative approaches encode transformation knowledge into continuous embedding spaces for efficient exploration of candidate strategies, but face three key limitations: (1) overlooking hierarchical relationships between low-level features, operations, and high-level abstractions; (2) enforcing order-sensitive embeddings on inherently permutation-invariant transformation sequences, thereby introducing systematic bias; and (3) relying on gradient-based search, which is ill-suited to non-convex transformation spaces. We propose a framework with two complementary components. First, a permutation-invariant hierarchical module captures interactions across features, operations, and abstraction levels, with a self-attention pooling mechanism that maps semantically equivalent structures to consistent embeddings aligned with downstream performance. Second, a policy-guided multi-objective reinforcement learning strategy initializes the search from empirically strong seeds and jointly optimizes predictive accuracy and transformation efficiency. Extensive experiments on diverse tabular benchmarks demonstrate the effectiveness and robustness of our framework against strong baselines. Our code and data are publicly available at: https://github.com/RayLiu1103/PHER.
针对监控和控制工业流程的运营技术（OT）系统发起的网络攻击，对社会基本服务构成日益严重的威胁。因此，开发自动化入侵响应策略至关重要。本文使用 POMDP 框架，对一个 OT 入侵响应用例建立了形式化模型。该模型基于流量测量，包含了现实的部分可观测性模型。这种方法使我们能够开发适用于自动化入侵响应的、可处理的基于学习的解决方法，这些方法基于 PPO。我们在一个仿真的 OT 系统上评估了所获得的响应策略，发现针对所研究用例中的多种 MITRE 攻击类型，这些策略均能有效发挥作用。
Plane geometry remains a significant challenge in AI, requiring the integration of visual perception and mathematical reasoning. While Large Multimodal Models (LMMs) naturally handle visuo-linguistic inputs, they are often computationally intensive and opaque. We demonstrate that a pure Large Language Model (LLM), when equipped with specialized modules, can rival state-of-the-art LMMs on complex geometry problems. Our framework integrates a Geometric Vision Parser, which translates diagrams into symbolic form, with a Symbolic Solver that performs formal deductions, thereby mitigating hallucinations and promoting interpretable reasoning. To enable rigorous evaluation, we curate a benchmark of challenging problems from the 2025 Chinese Zhongkao examinations, ensuring data novelty and testing deeper deductive skills. Experiments demonstrate that our approach achieves performance comparable to Gemini 2.5 Pro while delivering clearer, human-like solutions.
大家好，我是 Dimitris。 这是我构建并用于运营自己企业的企业 OS，任何人都可以安装并免费进行自托管。可以把它理解为将 Claude Code、Cowork 和云端会话整合在一个自托管应用中。 它从设计上就是一个多租户应用，许多人可以通过 4 种不同的协作模式与企业智能体协作；它可以使用你的 Anthropic 或 OpenAi 订阅运行，也可以使用本地模型。 每个智能体都可以运行在 Claude Code 或 Codex CLI 上，作为服务器上的持久进程运行于内核沙箱（bubblewrap）中，并始终启用网络隔离（pasta）；每个智能体都有自己的工作区、记忆、日程安排和工具。同样的智能体也可以配置为通过一个出站 WebSocket，在任意远程计算机上以完全相同的方式运行（无需入站端口，也不需要 VPN）。 智能体已经内置了许多功能。例如，它们可以通过 Twilio 或你自己的 Asterisk 接听和拨打电话，可以编辑视频以及 Excel、Word、PPT 文件，还可以借助 Collabora 直接在聊天中预览这些文件，此外还有更多功能。 只需一个安装脚本和一个 Docker Compose 文件即可开始使用，无需注册。该许可证采用 Fair Source，所有代码均为公开代码，并且最多支持 5 名用户的自托管是免费的。1.6.0 已于今天发布。 AI 免责声明：OtoDock 的大部分代码都是使用 OtoDock 自身、运行 Claude Code 编写的。 我很希望大家阅读代码、了解一下沙箱模型，并告诉我是什么因素会阻止你在自己的硬件上运行它。 GitHub： https://github.com/OtoDock/oto-dock 网站： https://otodock.io 评论链接： https://news.ycombinator.com/item?id=49630606 得分：3 评论数：0
Paul Christiano 加入 OpenAI Foundation 董事会及其 Safety and Security Committee，带来在 AI 对齐、安全和标准方面的经验。
While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autonomous development demands a missing pillar: post-hoc monitoring and auditing to understand what models learn and ensure safe alignment. Mechanistic interpretability tools are essential to bridge this gap, among which Sparse Autoencoders (SAEs) serve as a cornerstone by isolating interpretable features for model inspection and steering. In this paper, we introduce SAEScientist-Bench to evaluate whether AI agents can act as scientists utilizing SAE tools for autonomous mechanistic discovery. Given a target concept, an agent designs contrastive probes and navigates a Gemma Scope dictionary of 131K+ features in Gemma-2-9B-IT to discover the optimal feature, evaluated against curated expert reference features anchored on Neuronpedia across activation rank, concept selectivity on contrastive texts, and causal steering. Across 10 agent configurations and 20 tasks, frontier agents demonstrate genuine discovery capabilities and lead different evaluation dimensions, but remain well behind the expert baseline, approaching expert levels on separating target concepts from contrastive controls while lagging substantially in causal generation steering. Further analysis reveals that although agents can design contrasts to rule out spurious candidates, they frequently misinterpret experimental measurements. These results establish experimental model understanding as a measurable capability for closed-loop autonomous AI R&D. Our code is available at https://github.com/Trae1ounG/SAEScientist.
Dexterous manipulation involves contact-rich and fine-grained interactions with the physical world, posing significant challenges for existing vision-language-action (VLA) models due to severe visual occlusions and complex contact dynamics. While recent works have incorporated tactile sensing into robotic manipulation, most approaches still rely on homogeneous multimodal fusion, lacking adaptive tactile integration and explicit modeling of physical dynamics. In this work, we present DeCAL, a physically-grounded dexterous vision-language-action model that unifies understanding, imagination and action generation for contact-rich dexterous manipulation. Built upon a Mixture-of-Transformers (MoT) architecture, DeCAL leverages specialized experts for each capability while enabling efficient information flow among them. To effectively leverage tactile information, we introduce Adaptive Visuo-Tactile Fusion that dynamically regulates tactile interactions via a contact-aware gating strategy. Furthermore, we propose Visuo-Tactile Latent Co-Imagination to jointly model visual and tactile dynamics, equipping the policy with implicit physical world knowledge. Experimental results show that DeCAL consistently achieves state-of-the-art performance across all tasks, attaining a 71% average success rate and an 83.4% progress success rate, while also demonstrating strong generalization to unseen scenarios. The website is available at https://aureleopku.github.io/DeCAL.
执行反馈可以引导编码智能体完成正确的代码库修复，但前提是测试能够捕捉问题所要求的行为。智能体生成的测试可能编码不完整或不正确的行为目标；当同一条轨迹同时编写补丁和测试时，它们的错误可能相互一致，从而产生虚假的信心。我们提出 ExecCritic，它将测试--验证--修订脚手架与用于在其中训练智能体的角色特定强化学习方案相结合。该脚手架将测试构建与源代码修复分离：Test agent 独立生成代码库原生测试，一个 fail-closed 测试框架对其进行合格性判定并冻结，Repair agent 则基于这些测试的执行反馈修改源代码而不更改测试。两个角色都使用 Qwen-3.5-35B-A3B 作为骨干模型，并分别进行训练。在 Learn to Test 中，Test agent 学习生成行为上有效、能够区分正确与错误补丁的测试。在 Test to Improve 中，Repair agent 学习直接解决任务以及基于反馈进行修订。在 SWE-bench Verified 上，测试质量决定反馈是否有帮助：在固定基础 Repair agent 的情况下，来自基础 Test agent 的测试会使解决率从无测试基线的 61.2% 降至 57.3%，而来自 GPT-5.6-sol 的测试则将其提升至 65.3%。角色特定的后训练将 Qwen Test agent 的 Base-to-Gold 成功率从 22.2% 提升至 62.2%；组合两个后训练的 Qwen agent 达到 72.6%，在评估时无需更强模型或 Oracle 反馈的情况下，比原始无测试基线提高了 11.4 个百分点。代码已公开发布于 https://github.com/MSR-Orchard/execcritic.
Agent harnesses（围绕模型的 system prompt、tool set、execution hooks 和 context-management scaffolding）是决定 agentic 任务成功的关键因素。自动化 harness 演化可以使较小模型以 frontier-model 成本的一小部分在特定领域任务上表现良好。由于 harness 和模型权重都会塑造行为，我们探讨应如何结合 harness 演化与轻量级 fine-tuning。在七项企业 agent 任务中，我们首先使用较弱模型演化出一个 harness，随后发现更强的 expert 往往能更有效地使用它，这表明 expert 监督可能弥合剩余差距。然而，在演化后的 harness 下用 expert 的完整轨迹训练较弱模型会适得其反：在 Qwen3-Coder 和 Gemma 4 上，七项任务的性能全部回退 4 到 30 个点，尽管同样的流程在未演化的 harness 下是有帮助的。我们的分析表明，imitation 会迁移知识并增加 scaffold 使用，但会破坏模型-harness 适配：较弱模型采用了 expert 的规划策略，却不具备执行该策略的能力，并且不再匹配围绕其原生规划风格演化出的 harness。因此，我们开发了一条 on-policy expert-correction 流程，由一个元层级 MLE agent 自动化执行；该流程会定位较弱模型自身 rollout 中失败的 turn，并要求 expert 仅重写该 turn。这保留了模型的规划风格，并结合了 harness 演化与模型适应的收益。我们的结果识别并解决了 harness 与权重更新之间的一种冲突来源，为特定领域企业任务上的经济型共同演化提供了一种保持兼容性的方案。
Robotic Process Automation (RPA) is widely used to reduce administrative burden in United States hospitals, yet an estimated 30-50% of RPA initiatives underperform because processes are selected informally, without a repeatable method to catalogue candidates, prioritize them, match each to an automation tier -- a Python bot, an open-source orchestrator such as n8n, or an enterprise platform such as UiPath -- and forecast financial return before committing resources. We propose a four-module, data-driven framework unifying these decisions: a Process Taxonomy of twenty recurring hospital processes across five value streams; a Prioritization module deriving an Automation Suitability Index from an Analytic Hierarchy Process matrix with an explicit consistency check; a Tool-Tier Selection module recommending the least-cost technology sufficient for a process complexity, integration, and compliance profile; and a Return-on-Investment module quantifying labor savings, error-cost avoidance, payback, and net present value. Applied to a synthetic portfolio spanning all twenty processes, plus a reference data-flow architecture linking it to hospital EHR/payer/ERP systems: 12 of 20 clear the prioritization threshold; the ranking is robust to +/-20% weight perturbation (Spearman correlation 0.83, top-5 set preserved 97.7%, 2,000 Monte Carlo trials); an Automation Risk Index flags four qualifying processes as Critical risk; a budget-constrained portfolio optimization shows diminishing marginal NPV as spend scales from $400K to $1.03M; and a second Monte Carlo analysis shows portfolio NPV stays positive at its 5th percentile. The framework is a conceptual synthesis of the literature rather than an instrument calibrated on primary hospital data; we discuss HIPAA governance and a research agenda for empirical validation. A supplementary Python implementation accompanies the paper.
Large language models are increasingly deployed as agents that plan over long horizons and act through external tools. Most agents select actions through unconstrained generation over an accumulating history, leaving implicit the procedural knowledge of what to do, in what order, and under which conditions. As trajectories lengthen, agents can lose track of their objectives, invoke tools out of order, and repeat unproductive actions. We introduce the Procedural Graph: just as a knowledge graph organizes factual knowledge into (entity, relation, entity) triplets for what-is questions, a Procedural Graph organizes procedural knowledge into (procedure, relation, procedure) triplets for what-to-do questions. At each decision step, the framework localizes the agent's active node, and a guidance model translates the surrounding subgraph into step-level situational guidance that biases the solver's next action without dictating it. The graph is self-evolving: an LLM refiner contrasts failed trajectories with successful ones and edits the graph's topology and attributes, committing edits that preserve or improve held-out validation performance while retaining rejected ones to discourage repetition. Starting from a minimal skeleton, the loop builds graphs that match or surpass hand-designed ones. It can also repair a flawed expert prior. Across multiple datasets, task types, and LLMs, the Procedural Graph delivers consistent gains over memory-based baselines, and self-evolution further improves performance without manual engineering.
威胁狩猎日益依赖于将非结构化知识（例如网络威胁情报报告）转化为可执行的狩猎线索：这些线索应是简洁、可调查的假设，并以可观测工件和对手技术为依据。手动生成此类线索既繁琐又难以扩展。现有的自动化方法往往止步于实体层，忽略防御方的运营环境，并且孤立地分析每份报告。为弥补这些不足，我们提出了 AHLERT。该系统通过以下方式自动从威胁报告中提取相关、具备环境感知能力的狩猎线索：（i）一种混合检索器，将稠密向量搜索与基于 MITRE ATT&CK 种子的知识图谱多跳遍历相结合；（ii）一种基于本体落地的检索增强生成方法，将每条线索约束在防御方自身的资产和控制措施范围内；以及（iii）一个与 LLM 无关的框架，输出结构化且可直接执行的线索，而非宽泛的失陷指标。我们使用面向多个知名 APT 的公开 CTI 报告，在多种专有模型和开放权重模型上对 AHLERT 进行了评估。与单路平面 RAG 基线相比，结合本体落地的混合证据检索将平均 F1 提升了约 2 倍（从 0.44 提升至 0.85）；与现成的 LLM 模型相比，AHLERT 达到了最高的有效性得分（约 86.95%）。
智能体技能为冻结的语言模型智能体提供了一种轻量级方式，使其具备领域知识和程序性指导，但获取高质量技能仍然成本高昂且难以规模化。由专家编写技能需要投入大量人力。近期的技能自演化方法通过自动化迭代循环，利用执行反馈来修订技能，但其启发式更新策略往往导致优化过程不稳定、迭代效率低下。我们发现，实现稳定且高效的技能自演化面临两项挑战。方向稳定性要求有效的修正能够不断累积，而不是被局部迭代反馈覆盖。更新自适应性要求每次修订的范围能够反映近期案例级改进的一致性。我们提出 SkillAdam，这是一种受 Adam 启发、用于优化离散且不可微技能文档的框架。作为 Adam 一阶矩的功能类比，优化记忆记录已识别的问题以及此前解决方案尝试的结果，从而稳定更新方向。作为 Adam 二阶矩的功能类比，波动性驱动的编辑预算跟踪近期案例级改进的历史加权变化，并自适应地控制更新幅度。在涵盖短期和长期任务的七个基准测试中，SkillAdam 以更加稳定的优化动态取得了当前最佳性能。与以往方法相比，它还以显著更少的优化迭代次数和更低的成本获得了更强的技能。代码仓库：https://github.com/ruc-datalab/SkillAdam
确保自动驾驶的安全性是一项关键挑战。基于场景的测试是一种用于验证自动驾驶系统（ADSs）的系统化流程，但目前仍是一个碎片化的模块化流水线：场景生成、检索、修改、ADS 执行和结果分析分别由交互很少的独立工具完成。大型语言模型（LLM）智能体已在 ADS 的感知、规划和控制等子系统中展现出良好应用前景。然而，尚无先前工作能够通过统一的 LLM 智能体框架覆盖 ADS 的完整基于场景的测试流水线。我们提出了 PlannerForge，这是一种 LLM 智能体框架，扩展了基于场景测试的全部阶段（从场景生成到 ADS 评估），并新增两个由 LLM 增强的阶段：ADS 增强和 ADS 基准测试。我们使用 10 个开箱即用的 LLM，在 5 种提示条件下，对所有任务（生成、选择、修改、模块路由、规划器测试和增强）评估了 PlannerForge。各任务的最佳得分范围为 0.88 至 1.00；在大多数任务上，开源的 20-35B 后端模型能够媲美商业 API。Qwen3.6:35B 等开源模型在五项任务中的三项上达到了商业 API 的水平。将各模块端到端串联后，仍可保留 83% / 78% 的种子查询（商业模型 / 开源模型）。在自然语言生成方面，PlannerForge 优于 Scenario Factory 2.0（Finkeldei 等，2025）（200 个查询中生成 193 个可执行场景，而非 144 个），并实现了所请求城市、道路和车辆属性的 92-96%。在排名第 1 的选择准确率上，它优于 BM25（Robertson 和 Zaragoza，2009）（92.0% 对 67.5%）；在生成物理有效修改方面，它优于 From-Words-to-Collisions（Gao 等，2025）（≥94% 对 31%）。在 N=400 时，成本调优在不进行领域特定微调的情况下，将规划器成功率从 50.4% 提升至 70.2%，并将碰撞率从 19.0% 降至 8.4%。
本文介绍了 Gander，这是一个端到端模型，在单一框架内统一了全模态感知、实时交互和智能体能力。与基于轮次的传统范式不同，Gander 持续接收包括视频、语音和文本在内的多模态流式输入，从而能够在日常对话和复杂的工作流型智能体场景中实现自然的全双工交互。用户可以随时打断模型，模型也可以主动提供中间反馈或提出后续问题。为原生支持这些能力，Gander 采用了两项关键架构设计：1）采用 Cerebellum-Brain 协作框架，其中 Cerebellum 负责实时交互和全模态对话能力，而 Brain 负责复杂推理和更高层次的智能体任务。两个组件通过工具调用和智能体编排运行时持续交互。2）Cerebellum 构建于流式 Thinker-Talker 架构之上，在分块层面将用户输入和模型输出进一步扁平化为有序 token 流，为低延迟、持续性交互提供统一表示。我们从四个维度对 Gander 进行了全面评估：对话能力、全模态理解能力、交互能力和智能体智能。内部人工评测表明，Gander 在实现具有竞争力的全模态交互性能的同时，保持了 SOTA 开源模型自然且富有表现力的语音对话能力。Gander 在具有挑战性的真实场景中也展现出稳健性，包括背景噪声干扰、多方交互和附和式沟通。我们将 Gander 及其模型、代码和数据一并开源，以促进社区进一步开展研究与开发。
Multi-step LLM reasoning lacks a machine-recheckable ledger: discarded reasoning paths leave no auditable record. We propose the Deposon scattering layer, which binds each node of an LLM-generated concept-decomposition graph to a two-parameter Deposon state; paths undergo three-channel scattering -- transmission, reflection, irreversible dissipation -- obeying T+R+A=1 for arbitrary parameters, with a maximum per-path energy-audit deviation of 2.2E-16 (machine epsilon). We report all three evidence tiers honestly. On synthetic trap benchmarks the path-filtering gain is closed (pre-registered): unified reaches 100% versus a decoy-capture baseline at 7%/10%. On real benchmarks the layer is indistinguishable from a trivial six-keyword rule filter (GSM8K 0.87 >= 0.85, McNemar p=0.5; StrategyQA 0.899 = 0.899); no difference is detected here, so we sharpen the claim to "the differential value lies solely in machine verifiability." Fusion yields a second negative result: convex combinations with a semantic prior never improve (physics 0.484 -> 0.452), and the apparent lambda=2 gain is an anti-field artifact; any fusion gain must be nonlinear. Modeling the reverse dynamics as a potential game on the graph, we evidence an auditable scalar's monotonicity and near-gradientness and quantify the empirical coordination ratio (ECR). The three formalized dynamical-equivalence propositions (P1a/P1b/T-P1c) are falsified under the pre-registered kill protocol, and the potential-game claim is downgraded to approximate (cyclic-graph median residual 0.669): only consistency-level evidence survives at the dynamical level. Code: github.com/zeroandcat/Deposon.
AI-assisted programming raises distinct questions about who produces code, who feels ownership of it, and who is responsible when it fails. This research note examines these distinctions through a hypothetical enrollment failure and a selective reading of the literature. Identifying the producer of a defective expression does not, by itself, determine the duties of reviewers, release decision-makers, or service operators. Collective ownership likewise leaves those duties to be specified. The discussion then considers how quality engineering can evaluate both generated implementations and the processes that produce them. Acceptance criteria should be justified by the required service outcome, with component checks contributing evidence toward that outcome. This perspective also permits comparison with systems that perform a task without generating a separate program for it. Such substitution would change the object of authorship while leaving the service obligation intact. The note reports no new empirical results; it proposes distinctions and evaluation questions for AI-assisted software production.
任务向量通过识别权重空间中具有语义意义的方向，实现训练后模型编辑；通常的计算方式是取微调模型与其预训练初始化之间的差值。然而，这种对微调的依赖使得此类方向的发现成本高昂，也限制了训练后模型编辑的实用性。为解决这一局限，我们提出 Training-Free Task Vectors（TFTVs），一种无需微调即可计算类似任务向量方向的新方法。我们的方法仅利用前向传播统计量，将激活引导向量映射为秩一权重空间编辑，同时满足一系列算术性质，从而直接支持通过加法实现学习、通过减法实现遗忘，以及组合多项编辑。通过实证评估，我们在大语言模型行为控制任务上测试了 TFTVs，结果表明，它能够在保留通用知识和问题解决能力的同时，持续增强、抑制并组合目标行为。我们还将该方法与其他编辑和引导基线进行了比较，实验表明，TFTVs 能够实现更强的特征控制，并在实用性保持方面表现更佳或具有竞争力。我们希望这项工作能为训练后模型编辑以及更广泛的免训练模型控制开辟新的研究方向。代码已发布于项目网站：tftv-llm.github.io。
Modern science and engineering increasingly rely on time-varying data, yet the mathematical tools used to model temporal phenomena are often developed within separate disciplines, obscuring common principles and limiting the transfer of ideas across fields. This chapter presents the theory of narratives, an abstract framework for time-varying objects of any mathematical kind that supports both theoretical investigations and applications. To illustrate this perspective, the chapter develops three vignettes, each illustrating a different research direction. The first addresses a general concern: What information loss can occur when switching between different representations of temporal data? The second concerns structural and algorithmic approaches: How can we systematically decompose time-varying data into simple pieces and obtain invariants describing its structural complexity? The third is an application to control theory: How can we model multi-agent systems with switching communication topologies? More important than any individual vignette, the central message of this invitation is that a suitable abstract perspective can organize and guide research across remarkably diverse mathematical and scientific domains.
大型语言模型（LLM）取得了显著进展，但其巨大的存储和内存带宽需求仍然阻碍着高效部署。权重二值化是一种有前景的解决方案，但现有基于二值化的训练后量化（PTQ）方法通常由于隐藏开销而远超名义上的 1 比特存储目标。为弥合这一差距，我们提出了 All for 1-Bit（AF1），一种面向 LLM 的真正 1 比特 PTQ 框架。AF1 由两个互补组件构成：（1）空空间感知二值分解（NABF），通过 Hessian 感知的代理重参数化、空空间感知的二值分解以及仅缩放的全局重构来改进二值重构；（2）分层 Shapley 分配（HiSA），利用分层 Shapley 敏感度分配结构容量。二者结合后，使模型在 PTQ 设置下严格遵守 1.0-BPW 预算的同时保持准确率。在 LLaMA、Qwen 和 Gemma 系列上的实验表明，AF1 在困惑度和零样本准确率方面始终优于现有基于二值化的 PTQ 方法。与 BF16 相比，在所评估的模型中，AF1 平均实现了 2.5 倍的推理加速，并将内存占用降低了 90% 以上，为 LLM 真正的 1 比特压缩走向实际部署提供了可行路径。用于复现实验的代码位于 https://github.com/Kishon-zzx/AF1。
Whole-body diffusion-weighted imaging (WB-DWI) is widely used for multiple myeloma (MM) assessment, yet automated lesion segmentation remains challenging due to limited anatomical delineation and the low specificity of marrow hyperintensity. Existing studies have introduced bone region-of-interest (ROI) information and apparent diffusion coefficient (ADC) maps to mitigate these ambiguities, but practical limitations remain. Bone ROI construction often relies on costly manual annotation, image registration, or dedicated bone models, while ADC is usually incorporated only through simple channel fusion, limiting its ability to provide complementary structural and lesion-discriminative cues. To address these limitations, we propose a two-stage framework for MM lesion segmentation on WB-DWI. In the first stage, we train a bone ROI generation model from ADC images without dedicated bone labels, providing an efficient and practical anatomical prior for lesion analysis. In the second stage, we propose Anatomy-guided Multimodal U-Net (AMU-Net), which leverages ADC in a manner consistent with clinical lesion assessment rather than treating it as a generic auxiliary modality. Extensive experiments demonstrate the effectiveness and practicality of the proposed method. It achieves the best overall performance among the evaluated methods, with a mean Dice score of 76.2%.
对话中的多模态情感分析和情绪识别需要对文本、声学和视觉模态之间的异构交互进行有效建模。尽管大型语言模型（LLM）具备强大的语言理解能力，但将其适配到多模态情感计算仍然面临挑战：全模型微调的计算成本过高，而许多现有的轻量级适配器无法在跨模态融合过程中保留丰富的文本线索。为解决这些局限性，我们提出多视角文本引导多模态融合适配器（MVFA），这是一种参数高效的框架，能够为冻结的 LLM 增强强大的多模态推理能力。MVFA 首先通过最大池化、平均池化和注意力池化构建互补的文本视图；随后，这些视图引导音频和视觉特征之间的跨模态交互。融合后的多模态表示通过 Enhanced Q-Former Fusion Module 进一步压缩为一组紧凑的可学习伪标记。我们以 ChatGLM3-6B-base 作为主要主干模型，并在 LLaMA2-7B 和 Qwen3-8B 上进一步验证 MVFA，以考察其在多个冻结 LLM 主干模型之间的可移植性。MVFA 在三个具有挑战性的数据集上进行了评估：CH-SIMS V2.0、MELD 和 CHERMA。实验结果表明，MVFA 仅更新很小一部分参数，便在关键指标上取得了最先进的性能。具体而言，在 CH-SIMS V2.0 上达到 84.62\% Acc2 和 84.59\% F1，在 MELD 上达到 67.36\% Acc 和 66.03\% WF1，在 CHERMA 上达到 74.66\% Acc。这些结果表明，多视角文本引导融合是情感计算领域一种有效且可扩展的参数高效多模态 LLM 适配范式。代码公开于 https://github.com/Overwhelm1208/MVFA。
多模态大语言模型（MLLMs）在视觉问答等通用视觉理解任务上表现出色，但它们往往难以掌握一项基本的比较能力：识别两幅相似图像之间发生了什么变化。我们提出 VDiff-Bench，这是一个用于细粒度图像差异识别的高难度多项选择基准。VDiff-Bench 包含针对图像对的 1,756 道四选一题目，涵盖 10 类变化：位置、运动、局部图像颜色、整体图像颜色、出现/消失、噪声/分辨率、纹理、替换/尺寸、OCR/文本以及光照。每道题对应两个图像输入和 4 个选项：真实差异、两个高难度负面描述，以及一个“无差异”干扰项。为增加任务难度，我们专门构建了以真实标注为条件的负面选项，要求模型将实际变化与语义上相近的替代描述区分开来。对 11 个当前最先进的开源和闭源 MLLMs 的实验表明，细粒度视觉比较仍然十分脆弱：模型在不同数据来源和变化类别上的表现不均衡，并且在噪声和纹理等细微低层次变化上持续失败。例如，三个 7–8B 规模的开源 MLLMs 在语义变化上的得分为 52.5%–70.6%，但在噪声和纹理等低层次变化上的得分仅为 8.7%–33.3%，并错误地认为两个图像输入之间没有变化。令人意外的是，尽管其他闭源商业模型表现强劲，Grok 4.3 在识别图像间的噪声和纹理差异方面却出现了显著的性能下降，大幅落后于 Kimi K2.5 和 K3 等大型开源模型。总体而言，VDiff-Bench 为评估 MLLMs 的比较性视觉理解提供了有针对性的诊断工具，揭示了标准单图像视觉语言任务无法捕捉到的缺陷。
随着大型语言模型（LLM）日益部署于对对齐敏感的场景中，激活引导已成为一种轻量级的推理时替代方案，用于取代微调方法（例如 RLHF、DPO）实现行为控制。然而，现有工作通常仅在孤立行为上验证引导效果，因此尚不清楚引导向量是否编码了一致的语义结构，还是仅仅利用了特定行为的捷径。我们研究 LLM 引导向量的潜在几何结构是否反映了人类价值观与道德中的理论规定结构。我们以 Schwartz's Theory of Basic Human Values 作为主要的细粒度框架，构建了一个包含 2.6 万个样本、覆盖 20 个人类价值观的基准，并在多种模型系列和规模上分析了分布驱动方法（例如 CAA、SphericalSteer、ODESteer）与行为中心方法（例如 COLD-Steer、BiPO）。我们发现，分布驱动方法能够恢复与理论预测一致的人类价值观拓扑结构（Spearman $ρ$ 最高达 0.51，$p < 10^{-13}$）。相比之下，行为中心方法虽能取得相当的引导性能，但与预期价值观几何结构的相关性很低。几何保真度会随着模型规模扩大而提升，但在指令微调后下降。最后，更好的几何对齐还会带来跨价值观的、更符合人类认知的一致迁移：正确引导某一价值观会提升与其兼容的价值观，并抑制相互对立的价值观。代码和数据见：https://github.com/DeepRCL/Steering_Geometry】【。
高光谱图像（HSI）分类在农业、环境监测和城市分析等遥感应用中发挥着至关重要的作用。然而，高光谱图像分类的性能仍受到光谱冗余度高、对噪声敏感，以及难以联合建模局部材料组成与长程光谱依赖关系等问题的制约。为此，我们提出AGSA-Net，一种将光谱解混先验显式融入分类过程的丰度引导自注意力网络。AGSA Net首先在非负性约束和和为一约束下，利用混合线性-非线性重构解码器进行正则化，估计具有物理意义的子像素丰度图。随后，利用所学习的丰度构建丰度亲和先验，引导光谱Transformer突出具有类别判别性的交互关系，并将所得的Transformer特征与紧凑的丰度描述符融合，以进行最终预测；这不同于现有将丰度作为辅助特征或拼接特征的方法。在Indian Pines、Augsburg和Berlin数据集上的实验表明，引入丰度引导的上下文建模能够带来显著收益，尤其是在异质性城市场景中。源代码和训练好的模型见：https://github.com/nnuvi/AGSA-Net
当前，数据丰富领域中的科学发现受到人类带宽的限制：现实世界数据的规模和复杂性增长速度远远超过研究人员阅读、推理和综合信息的能力。近期基于 LLM 的多智能体系统已经开始实现研究周期部分环节的自动化，但它们主要针对假设生成，而在这些场景中，验证本身无法实现自动化；此外，每次运行都是一次性的，没有机制让研究结果不断积累或引导后续探究。本文介绍了 AutoKD，这是一种用于自主知识发现的多智能体框架，兼具计算性和累积性，使经过验证的发现得以持久保存，并为后续探究提供信息。六个相互协作的 LLM 智能体共同参与一个开放式发现循环：被接受的发现会存储在持久化洞见图中，该图既充当长期记忆，也作为引导探索的机制。我们从两个角度在三个多样化数据集上评估 AutoKD：针对已发表发现的开放式质量，以及通过文献衍生查询评估的条件式质量。在这两个评估角度下，AutoKD 不仅覆盖了已知发现，还提出了能够补充人类驱动研究的重要新发现。我们的代码已发布于 https://github.com/GeQinwen/AutoKD。
安全关键交通场景的生成对于训练和评估自动驾驶车辆至关重要。以往的方法通常使用简化的对抗性目标，对交通场景中现有智能体的轨迹进行扰动，以诱发安全关键交互，这可能限制所生成场景的合理性和多样性。尽管插入新的对抗车辆可以缓解这一局限，但如何针对具体场景确定引入这些车辆的时间和位置，仍然是一个挑战。在本研究中，我们提出了碰撞快照引导的时间反向安全关键场景生成方法（COSTER），这是一个利用学习到的交通先验来确定合理碰撞时间和位置的框架。COSTER首先在交通场景中确定的碰撞状态处插入一辆与目标车辆接触的新车辆，从而构建碰撞快照。以该碰撞快照为起点，利用条件变分自编码器执行时间反向展开，向更早的时间步反向重建所插入车辆的轨迹。实验表明，在合理性、多样性和数据效率方面，COSTER均优于现有方法。此外，在Waymo Open Motion Dataset中的安全关键场景上，使用COSTER生成的场景进行训练的智能体将碰撞率降低了31%，同时提高了自车任务完成率。项目网站位于 https://anonym-121.github.io/COSTER/。
连续环境中的视觉语言导航（VLN-CE）要求具身智能体遵循自然语言指令，在未见过的环境中进行导航。当前的零样本 VLN-CE 方法要么依赖预训练的航点预测器，要么需要在每个步骤中多次查询大型模型。为了解决推理延迟高和计算开销大的问题，我们提出了 O2C-Nav，这是一种高效的零样本导航框架，每个决策步骤仅调用一次大型模型。我们的方法引入了一种无需训练的结构化航点生成器，以及一种新颖的抽象表示，将稀疏且具备历史感知能力的候选航点直接投影到 RGB 图像上，作为视觉标记。MLLM 在每个步骤中选择一个航点，或生成一个备用目标边界框；随后，低层 Fast Marching Method（FMM）规划器将所选目标转换为一条可执行且无碰撞的路径。该范式为模型提供了具体的空间感知能力和显式记忆，同时显著降低了视觉处理负担。在 R2R-CE 和 RxR-CE 基准上的大量评测表明，O2C-Nav 超越了当前最先进的零样本方法，展现出在实时机器人部署中的巨大潜力。代码见 https://github.com/kkpsq/O2C-Nav-Code。
Planning with a generative model aims to estimate the value of a state using as few simulator calls as possible. SmoothCruiser achieves problem-independent complexity $\widetilde O(\varepsilon^{-4})$ by exploiting the smoothness of the entropy-regularized Bellman backup, but its estimator is only first-order. We show that the sample-complexity exponent of SmoothCruiser-type planners is governed by the order $β$ of the local Taylor remainder, giving oracle complexity $\widetilde O(\varepsilon^{-(2+2/(β-1))})$: the first-order case $β=2$ recovers SmoothCruiser, while a second-order/cubic remainder $β=3$ yields $\widetilde O(\varepsilon^{-3})$. We reach this regime with an optimal-transport-smoothed Bellman backup over action distributions, which has a closed form, a policy gradient, and a Lipschitz Hessian, and whose quadratic correction admits an unbiased cross-product estimator. The resulting SecondOrderSmoothCruiser achieves $\widetilde O(\varepsilon^{-3})$ oracle complexity for fixed OT parameters, and we relate the OT, entropy-regularized, and unregularized objectives through explicit regularization-bias bounds.
递归超分辨率（SR）通过将预测结果反复反馈给同一模型，将固定尺度的超分辨率扩展到极端放大，类似于反复缩放图像。然而，在每个尺度，尤其是深层尺度上获取真实值仍然具有挑战性，因为所需的源分辨率呈几何级数增长，导致更深层的预测缺乏监督。我们提出了 OracleZoom，这是一种受在策略蒸馏启发、受参考约束的框架，在自身轨迹上进行训练，同时将最后的真实值证据延伸到监督边界之外。直接监督和跨尺度监督约束可验证内容，而无参考质量目标则引导尚未解析的细粒度细节。受 KL 约束的预训练潜变量先验限制由质量驱动的偏移，而 EMA 一致性则稳定监督边界。在七个数据集上，OracleZoom 在各种放大尺度上均实现了最先进的超分辨率质量，CLIPIQA 平均达到 0.713，并且在更深层尺度上取得更大提升，同时显著减少了幻觉。代码、数据和模型可在 https://dipta007.github.io/OracleZoom/ 获取。
了解一位 MIT 研究人员如何使用 GPT-5.6 Sol 和 Codex 自主运行量子计算实验、分析结果并校准量子比特。
文章链接： https://rohanadwankar.github.io/posts/platforms.html 评论链接： https://news.ycombinator.com/item?id=49605644 得分：3 评论数：1
In case you're wondering why the limits behave so very different from last week. Also: this makes limit resets kind woth significantly less. Comments URL: https://news.ycombinator.com/item?id=49600233 Points: 8 # Comments: 3
文章 URL: https://github.com/trailofbits/coop 评论 URL: https://news.ycombinator.com/item?id=49593842 积分: 6 # 评论数: 1
Public vulnerability databases collect rich information about known software flaws, including their weakness types, affected components, and related patches. Fixing commits provide the exact code changes that removed these flaws. While these records capture why the original code was unsafe, they are documented mainly for human inspection rather than automated reuse. Consequently, the same unsafe conditions may still exist elsewhere in code without a known advisory, leaving much of this detection knowledge unused. We present BUGSTONE-E2E, a framework that transforms vulnerability history into executable detection rules and validates their findings. First, BUGSTONE-E2E mines reusable rules from verified fixing commits, capturing scan anchors, fix semantics, and CVE provenance and organizing them by CWE and language. Second, detection follows a funnel-shaped pipeline: early stages process a large pool of candidates using lightweight analysis, while later stages apply increasingly capable and expensive models to a shrinking set of targets. Specifically, BUGSTONE-E2E first enumerates call sites matching rule anchors using Tree-sitter, then removes benign sites using lightweight heuristics without LLM calls. Next, LLM-based agents inspect the remaining candidates guided by the rule. Following this inspection, the system re-triages surviving candidates and builds runtime verifications, then generates scope-checked patches validated via two-sided differential tests. Using 19,325 high-severity CVEs from 2022 to 2026, BUGSTONE-E2E identifies 2,710 fixing commits and constructs 1,033 detection rules across 56 CWE families, packaged into 172 skills. When applied across 14 programs, it produced runtime evidence for 644 findings. These results demonstrate that CVE history can be turned into an executable workflow, transforming past vulnerabilities into reproducible detection and repair.
The integration of GenAI tools into higher education assessment raises important questions about how students understand, interpret, and respond to AI-mediated evaluation. As instructors increasingly explore AI tools for providing feedback, prior research has examined whether GenAI-generated feedback improves writing performance and how students perceive its usefulness; comparatively little is known, however, about how students interpret such evaluation when they are explicitly informed that an AI system, rather than a human instructor, produced the feedback and the score. This study reports findings from a qualitative pedagogical inquiry conducted in an undergraduate technical communication course for computing students at a Saudi public university. Thirteen male undergraduate computing students completed an in-class handwritten writing task; the scanned submissions were evaluated by ChatGPT using a rubric-based prompt aligned with the task objectives. Students were then explicitly informed that ChatGPT had generated the score and feedback and were invited to reflect on the evaluation in writing. Inductive thematic analysis of these reflections identified four themes: perceived usefulness of feedback; awareness of AI's contextual and pedagogical limitations; conditional trust, distinguishing feedback utility from evaluative authority; and reflection on the institutional and pedagogical role of the human instructor. Participants accepted GenAI feedback as useful for surface-level revision but consistently positioned the human instructor as the appropriate authority over grading decisions. The study identifies this as a distinction between feedback utility and evaluative authority, two judgments that students treat as analytically separate rather than as opposite ends of a single approval scale...
Decompilation recovers high-level source from compiled machine code and serves as a foundation for security tasks such as vulnerability detection and malware analysis. Traditional decompilers like Ghidra and Hex-Rays expose whatever they cannot resolve as visible placeholders and often emit pseudocode that will not compile or execute; LLM-based decompilers produce clean, idiomatic C and are now judged almost entirely by recompilability and re-executability: whether the output builds and passes its shipped input/output tests. We show that these metrics can reward the wrong path: a function may recompile and pass every shipped test yet diverge on other legitimate inputs, and a disclosed vulnerability may disappear from the recompiled code with no visible trace of the crash. Neither failure is caught by existing suites. To address this gap, we propose Decompile-Diverge, a behavioral comparison oracle not relying on fixed or hand-crafted tests: for each function it synthesizes a driver, grows a fuzzing corpus from the reference, and reruns the decompiled code on the same inputs to detect changes in the function's behavior. Across eight systems in nine configurations on established LLM decompilation corpora, candidates that pass every shipped test still diverge from the original on our input corpus: 4.9% overall, and as many as 13% for a single system. On 300 real GitHub library functions and 287 CVE-grounded functions, recompilability and behavioral agreement can come apart: the strongest refinement LLM lifts Ghidra's build rate from 75% to 90%, while its Matched rate falls from 74% to 62%; on disclosed vulnerabilities, up to one tenth exhibit Crash Absence in its output. Source-level analysis traces this divergence to introduced fields, types, callees, and guards that replace the visible unknowns traditional tools leave behind.
视觉运动模仿策略在分布内的视觉条件下可以取得很高性能，但当引入视觉上相似的物体或容器时就会失败。我们将这种行为视为一个条件视觉锚定问题：成功控制所需的视觉目标会随操作阶段变化，而在更复杂的任务中，还会随所观察到的任务状态变化。借助 Action Chunking with Transformers (ACT)，我们系统地引入在颜色和形状相似性上可控的干扰物体和容器，并将失败定位到抓取和放置环节。我们发现，对干扰物的敏感性同时取决于视觉相似性的类型和操作阶段。基于这一诊断，我们评估了干扰物增强、阶段依赖的注意力正则化以及基于外观的视觉提示，作为相互补充的干预方法，以在保留控制所需空间信息的同时改进目标选择。这些干预显著提升了仿真环境和实体 UR3e 上的鲁棒性。我们进一步在一个预训练的视觉-语言-动作策略上考察了相同的失败模式，该策略应用于一个状态条件化的器械处理任务，其中医疗器械的可观察状态决定正确的目的地。综合来看，这些结果表明，即使底层操作技能仍然完好，视觉干扰物也会导致错误的物体或目的地选择；而显式改进目标选择，能够在不同的视觉运动策略学习范式中显著恢复性能。
We propose Ref-GeNVS, a training-free, reflection-aware method for generative novel view synthesis (NVS) in mirror scenes. Existing multi-view diffusion models often fail to recognize the mirror in the scene and cannot exploit reflected content for scene generation. To fix this issue without additional training, our key idea is to treat a mirror image as two complementary views. From input images, we estimate the mirror plane and reflect camera poses to form virtual views. Based on this virtual view setup, we propose a two-stage generation method consisting of Mirror-gated attention and Reflection injection, which enables reflection-consistent NVS by explicitly leveraging reflection relationships in a multi-view diffusion model. Ref-GeNVS inherits the strong generalizability of the multi-view diffusion backbone, while it does not require finetuning. On synthetic and real scenes including mirrors, Ref-GeNVS outperforms recent generative NVS methods by generating reflection-consistent and contextually coherent novel views, revealing scene structure visible only through mirrors. Project page: https://kim-geonu.github.io/Ref-GeNVS/
数据主权法规日益要求公共机构部署开源、可本地部署的 LLM agents，这些 agents 需要通过实时政府 API 串联多次工具调用。然而，开源模型在这种多步场景中始终表现不佳，现有基准也没有衡量这一差距。我们提出 Korean Open Public API Benchmark (KOPA-Bench)，包含 145 个真实世界任务。为弥补这一差距，我们提出 EDGE，一种由实时执行驱动的、用于工具调用数据合成的 Execution-grounded Dynamic Graph。EDGE 构建一个图，表示每个工具的输出如何作为另一个工具的输入，只保留那些在实际调用实时 API 时成功的连接，并沿着这些经过验证的连接遍历，以合成可执行的多步轨迹。我们在生成的数据集上通过 GRPO 对 9B 模型进行微调后，其性能几乎达到同一系列中未微调的 27B 模型，不仅在 KOPA-Bench 上有显著提升，在 BFCL 基准上也同样如此。
Wireless signals with position-related labels are pivotal for both performance evaluation and model training in the realm of wireless sensing. However, acquiring real-world datasets is often challenged by significant measurement and labeling costs. Traditional methods for synthesizing labeled wireless signals typically rely on environmental models, leading to extensive hyper-parameter tuning and inadequate realism for comprehensive model training purposes. To address these limitations, we introduce a novel deep learning (DL)-based method, namely Inter-Instance Generative Adversarial Networks (IIns-GAN), to generate realistic labeled wireless signals. The generated signals are particularly adaptive to different environment scenarios and well-suited for various model training tasks, including distance estimation and environment identification. We have conducted extensive experiments on public Ultra-Wideband (UWB) datasets to evaluate the realism and utility of the generated signals. The results demonstrate that the signals generated by IIns-GAN mirror the physical characteristics of real-world measurements, and significantly contribute to the improvement of model training in diverse wireless sensing tasks.
科学论文要求模型联合推理文本、公式、图表、代码和数据集，同时保留支撑证据的来源。现有基准通常将这些能力分开评测，因此无法明确多模态模型是否能够支持真实的科学阅读工作流。我们提出 SciDocBench，一个面向工作流的科学文档理解基准。它包含124个由专家撰写并经过难度筛选的问题，组织为五个科学领域中的七组研究助理能力组和19个子任务。每个问题在四种匹配条件下实例化，将英文或中文问题与全图优先或交错式文档表示相结合，形成496个评测实例，用于受控分析。表现最强的被评测系统仅达到62.6/100，在文档感知、证据锚定、核验和跨文档推理方面存在明显弱点。为将这些诊断结果转化为可扩展的训练信号，我们提出 SciDocIR，这是一种带类型的证据图表示，可保留科学文档对象、版面和交叉引用关系以及来源信息。基于 SciDocIR，我们构建了 SciDocDataset，包含约15K个监督微调样本和8K个强化学习样本，覆盖14个可验证子任务。SciDocBench、SciDocIR 和 SciDocDataset 共同构成了一个从评测到训练的框架，用于诊断和改进科学文档助手。项目页面见 https://github.com/InternLM/SciDocBench。
This paper introduces Regime-aware Constraint-Based and Noise-Based causal discovery with Markov Blankets (RCBNB-MB), a novel causal discovery algorithm for time series that relaxes the common assumption of a single, time-consistent causal structure. Time series are typically observed at discrete time points and often exhibit regime changes that challenge the assumption of a static causal structure, a limitation in many real-world dynamic systems. To address this challenge, RCBNB-MB identifies latent causal regimes, defined as subsets of time points within which a stable causal structure holds. The algorithm follows an iterative strategy that segments the time series into regimes and discovers the causal graph within each regime. By leveraging the Markov blanket rather than direct parents, RCBNB-MB gains robustness to errors in causal discovery and preserves predictive information. We provide theoretical guarantees for RCBNB-MB's ability to recover both regime transitions and causal graphs under reasonable assumptions. Furthermore, we validate its effectiveness through extensive experiments on simulated datasets with known ground truth and real-world IT monitoring data, where taking into account regime shifts is critical. Empirical results show that RCBNB-MB systematically outperforms baseline approaches in accurately detecting regime changes and their associated causal graphs, positioning it as a robust and versatile framework for non-stationary time series analysis.
Recent reports during the AAAI-27 review cycle highlight the risk of reviewers coordinating bids for reciprocal assignment advantage. Prior work treats bidding, reviewer assignment, and review manipulation as separate stages, leaving the lifecycle effects of collusive bidding unclear. Real-world analysis is further constrained by typically unobservable collusive intent and the lack of counterfactuals for the same conference. Motivated by this gap, we introduce \alg, an end-to-end multi-agent simulacra framework for studying reviewer assignment integrity by holding the conference environment fixed and configuring LLM-driven reviewer agents with honest or collusive policies. We further develop an affinity-guided collusive bidding strategy that uses mutual reviewer-paper affinities to construct collusion rings and select target papers, producing expertise-consistent rather than arbitrarily targeted attacks. Controlled experiments show that collusive bidding more than doubles target-paper capture and that assigned colluders score target papers about two points higher than honest co-reviewers, while conference-wide effects remain comparatively modest. Evaluated bid-phase detectors provide only limited evidence of collusion: in a fixed-triplet detector stress test, native positive-bid graphs are confounded by benign affinity, while a Very-High-only diagnostic view enables precise but low-coverage local recovery.
Cryptocurrency markets exhibit extreme volatility and non-stationary dynamics that challenge conventional forecasting methods. Although Large Language Models (LLMs) have shown promise for time series forecasting, the combined effects of adaptation choices remain largely unexplored in financial settings. This study introduces PRICE, a structured approach for adapting LLMs to short-term Bitcoin price forecasting. Built on a 4-bit quantized LLaMA-3 8B model, PRICE investigates how fine-tuning, numerical representation, prompting, inference, and decoding jointly influence forecasting performance. PRICE integrates Parameter-efficient fine-tuning with Low-Rank Adaptation (LoRA), Recursive multi-step inference, Integer-rounded numerical representation, Context-Task-Format (CTF) prompting, and Exact zero-temperature decoding. Ablation studies show that each component contributes to forecasting accuracy and reliability. LoRA enables efficient training on limited hardware, recursive inference improves accuracy, integer-rounded values reduce errors, CTF prompting outperforms Chain-of-Thought, Implicit Chain-of-Thought (iCoT), and few-shot prompting, and zero-temperature decoding improves stability during recursive forecasting. Comparative evaluation against eight transformer-based and time-series foundation models shows that PRICE achieves the lowest forecasting errors on both validation and test sets while maintaining robust performance across evaluation periods. Despite being based on a model primarily pretrained on text rather than time-series data, PRICE achieves competitive or superior performance relative to specialized foundation models. These findings demonstrate that adaptation choices critically determine the accuracy and robustness of LLMs for numerical time-series forecasting.
一个快速扩张的行为体生态正在移除开放权重 AI 模型内置的安全护栏。我们通过识别关键生产者、下游复刻和新兴应用来描绘这一生态。在 2024 年 1 月至 2026 年 3 月期间，我们在 HuggingFace 上识别出 3,471 个原始未审查模型，每个模型平均被重新打包 2.4 次；三个行为体占全部 8,164 次压缩再分发的 52%。一旦这些模型经过量化，并在不同账户、格式以及诸如 Ollama 之类的注册库中被镜像，它们就会在上游移除后依然持续存在，并且更容易在下游部署。在已识别的 1,643 个集成了未审查大语言模型（ULLMs）的 GitHub 应用中，25% 被归类为明确恶意。
人类知识本质上是结构化且相互依赖的：掌握一个概念需要先掌握其前提条件，这一原则可由知识空间理论（Knowledge Space Theory, KST）形式化。尽管 LLMs 在复杂推理任务上取得了很强的表现，但它们是否表现出连贯、类人的知识结构仍不明确。我们提出一个基于 KST 的框架，用于评估 LLMs 在数学推理中的知识结构，并将其作为一个规范性框架来分析 LLM 行为是否遵循有原则的知识依赖关系。我们将八个开源和闭源 LLMs 与真实人类学习者进行评估，发现：(1) LLMs 并不遵循人类知识结构——它们经常违反知识依赖关系，也无法利用上下文中提供的相关知识来提升对依赖性问题的表现；(2) LLMs 彼此之间也不共享一致的知识结构，这体现在其知识分布重叠度较低。此外，这些结构性缺陷在很大程度上不会被基于准确率的评估以及 LLM-as-judge 评估所发现。总体而言，我们的结果提供了行为层面的证据，表明当前 LLMs 的知识并不遵循类人的结构。
Commonsense reasoning in computer vision encompasses integrating visual data and contextual knowledge, crucial for enhancing AI's understanding of everyday scenarios. This understanding not only improves machine learning models but also enhances their ability to interact meaningfully with humans and the environment. Unlike CNN-based conventional vision models, which are designed to identify objects within a specific image, incorporating commonsense knowledge enables models to interpret scenes in a more holistic manner, thereby improving their spatial ability to reason about relationships among objects and actions. This integration not only enhances object recognition but also facilitates a deeper understanding of the contextual factors, ultimately leading to more precise predictions and interactions in real-world applications. This paper presents a comprehensive survey of recent developments that integrate commonsense knowledge into computer vision tasks. We systematically review approaches based on knowledge graphs, scene graphs, neuro-symbolic models, and commonsense-augmented transformers. We also outline current limitations related to dataset bias, knowledge incompleteness, and integration challenges. Finally, we highlight prospective research trajectories in cross-modal reasoning, scalable commonsense knowledge injection, and neuro-symbolic hybrid architectures to develop truly intelligent visual systems.
LLM agent 系统越来越多地将来源追踪、授权、策略执行、协议适配器和执行控制结合在一起。然而，单独正确的安全机制并不一定能够组合成端到端安全的系统：当动作跨越组件边界时，安全关键上下文可能会被丢弃、扩展、重新绑定或重新解释。我们将这种失效模式定义为安全上下文不连续性，并提出 CONTINUITY，一个用于可验证地组合 agent 安全控制的框架。 CONTINUITY 使用 assume-guarantee contract 对每个组件建模，并通过带签名的 root grants、provenance commitments、角色绑定的 transition receipts、有界类型化 release、transformation witnesses 和 effect-bound execution permits，在各个转换之间携带经过认证的安全上下文。我们形式化了端到端后果完整性，要求每一个实际产生的外部效果都必须有有效且最新的授权见证作为支撑，该见证将主体、任务、来源、委派、策略状态、规范化动作以及最终性边界连接起来。 我们实现了一个参考验证器和一个确定性的跨层故障注入套件，覆盖四个应用领域中的 32 类故障。在跨越 128 个故障-领域类别的 2,560 个参数化攻击实例中，完整的 CONTINUITY 配置不会导致任何有害外部效果，同时完成了全部 700 个良性任务，并对全部 200 个模糊案例进行了升级处理。这些结果表明，安全的 agent 执行不仅需要单个控制机制本身正确，还需要显式契约来在完整的指令到效果路径中保持其保证。
Automated reference-based evaluation methods play a critical role in assessing natural language generation systems. Existing meta-evaluation primarily measures agreement with human judgments or benchmark labels, providing limited insight into evaluator behavior under controlled conditions. We introduce behavioral correctness assumptions, a complementary framework for evaluating reference-based automatic evaluation methods. We define a taxonomy of correctness-preserving and correctness-altering assumptions and operationalize them through controlled response transformations that specify expected scoring behaviors. We evaluate diverse lexical, character-level, semantic, LLM-based, and hybrid evaluators and analyze their assumption-level behavior, stability, sensitivity, repeat-run variability, configuration sensitivity, and reproducibility. Our experiments reveal distinct behavioral trade-offs across evaluation paradigms: no evaluator satisfies all proposed correctness assumptions, and evaluators with similar aggregate performance can exhibit substantially different behavioral profiles. These findings demonstrate that behavioral correctness assumptions provide diagnostic information obscured by conventional aggregate meta-evaluation.
Vision-Language-Action（VLA）模型在语言条件下的机器人操作方面取得了令人鼓舞的进展。然而，现有数据集和基准主要评估预定义设置下的任务完成情况，对模型在空间和流程复杂性不断增加时的推理能力提供的洞察有限。我们提出 \textbf{RoboSPA}（\textbf{Robo}t \textbf{S}patial-\textbf{P}rocedural \textbf{A}ssessment），这是一个用于诊断 VLA 模型中具身推理能力的大规模机器人操作数据集和基准。\texttt{RoboSPA} 关注两个核心维度：细粒度空间推理和长时程流程规划，涵盖 10 个任务类别和 56 个基础任务。每个任务都设定为五个难度级别，共生成 280 种变体，空间歧义和流程复杂性逐步增加。我们在多种具身形态和多样场景中收集了 527K 条轨迹。除二元成功率外，\texttt{RoboSPA} 还引入了用于更细致评估的诊断指标。对代表性 VLA 模型的实验表明，当前系统在复杂空间关系、精确的底层执行以及依赖记忆的规划方面仍然存在困难。这些结果使 \texttt{RoboSPA} 成为一个具有挑战性的诊断基准，可用于开发更强大、更可靠且更具泛化能力的具身智能体。我们的数据和代码可在 https://github.com/fanzhenxuan/RoboSPA 获取。
Why can't we get an agent scheduler that supports all of the following: - git compatible - agent agnostic - 100% open source - os and system-agnostic - multi-runner support - support mcp/ui/http - unlimited routines/crons So I built one, moadim.io is a local Rust daemon you install in the target machine, give it a name and manage its routines via a Git repository, wants a new routine that send you a daily message from this machine? Create a pr and merge, have another routine that run every hour to pull the latest changes to the ~/.config/moadim folder. With more than 1,000 users, I define this project as almost "done" and ready for production. Me and thousand more people are use it in a daily manner. It currently supports Claude, Codex, Hermes, and Pi, and you are welcome to add your agent of choice as well because it's 100% configurable. You are welcome to have a look at the source code of the daemon in " http://github.com/moadim-io/daemon " Feel free to provide me with suggestions for more features around this topic. I don't want to branch out to new off road topics likt webhooks, this is a "done" software in the realm of agents schedulers that focus on cron-like work. Also feel free to start the github repository and open issues and PRs for your suggestions. Comments URL: https://news.ycombinator.com/item?id=49571537 Points: 3 # Comments: 0
I love research and development, you may have heard of me because of PJON (Padded Jittering Operative Network). It is a network protocol I started developing in 2010, which was recently implemented in silicon by the ETH Zurich university thanks to the research of Pius Sieber. I am excited to share with you TERMy, a terminal assistant built on top of the NPC-Forge framework. Unlike everything else being built today, TERMy does not use embeddings, machine-learning or LLMs. It runs on the CPU (even on a Raspberry Pi Zero) both in the terminal or client-side in a browser tab and responds in milliseconds. It is a cynical but very knowledgeable Linux terminal assistant that translates your natural language into shell commands without relying on a single artificial neuron. I had a chance to focus for 2 months on my personal projects since early July, during the strange times of AI price hikes and the end of subsidized tokenmaxing. I was curious to see if I could develop from scratch a terminal assistant capable of handling simple natural language requests. I have a bad memory and got used to ask to copilot "activate the virtual environment" or similar trivial operations spending a non negligible sum every month. I started thinking, maybe I can do something to make my workflow more efficient? Do I really need trillions of parameters to accomplish those tasks? How it Works When you type a prompt, it goes through a lightweight NLU pipeline written in ~1000 lines of Python that implement the following steps: 1. Strip expletives, interjections, encouraging, discouraging and thanking words (remove noise) 2. Sentiment analysis 3. Exact Match (very fast) 4. Template Match (slower) 5. Probabilistic Match (even slower) Step 5 relies on: 1. IDF (Inverse Document Frequency) to identify rare words. 2. BOW (Bag Of Words) to accommodate word inversions. 3. IDF weighted Levenshtein to safely handle typos. Permission gating is hardcoded into the dataset and enforced for all potentially destructive commands, so it's inherently safer than letting an unpredictable LLM run wild on your machine. - TERMy in operation: https://www.youtube.com/watch?v=qeIp0xePLBg - Variance and typo tolerance: https://www.youtube.com/watch?v=tQvGDk6fkk0 - Copilot integration: https://www.youtube.com/watch?v=Wzzouhq2a8A - Advanced features: https://www.youtube.com/watch?v=qeIp0xePLBg - Source Code: https://github.com/gioblu/NPC-Forge Comments URL: https://news.ycombinator.com/item?id=49562219 Points: 3 # Comments: 0
Hey all, i just launched my Eink Bike computer project and think it is cool. Another tidbit, in the crazy things that AI has done... It has helped create a ANT (common sensor wireless protocol used in workout/biking) implementation for ESP32 by messing around with undocumented registers: https://github.com/RaemondBW/esp32-ant Comments URL: https://news.ycombinator.com/item?id=49567437 Points: 3 # Comments: 0
文章链接： https://github.blog/ai-and-ml/github-copilot/project-hydrafusion-frontier-quality-via-multi-model-orchestration/ 评论链接： https://news.ycombinator.com/item?id=49566788 积分：6 # 评论数：0
文章链接: https://www.nytimes.com/2026/09/04/technology/open-source-ai-anthropic-openai.html 评论链接: https://news.ycombinator.com/item?id=49566137 得分: 5 评论数: 1
Neural machine translation (NMT) systems typically produce a single output per input, obscuring the alternative decision trajectories implicitly available within multilingual decoding. This opacity becomes particularly problematic in low-resource dialect settings, where multiple linguistically valid realizations may differ in lexical authenticity, register, and structural stability. We propose reframing translation as a structured decision space explored by autonomous translation agents. Instead of analyzing a single output, we model distinct translation pathways as agents operating over a shared multilingual backbone. Inter-agent divergence is treated not as error but as an interpretable behavioral signal. We conduct an empirical study on Turkish--Syrian Arabic translation using three agents: (1) zero-shot direct translation, (2) dialect-stabilized translation via lightweight fine-tuning, and (3) pivot translation through English. Evaluation is performed on 5,000 dialogue sentences, while stabilization is trained on 5,000 additional Turkish--Syrian sentence pairs drawn from television dialogue and MADAR-Turk resources. Rather than optimizing for conventional performance metrics, we quantify structured behavioral displacement using dialect marker frequency, lexical proximity to standardized Arabic, and structural variance. Lightweight stabilization nearly doubles dialect marker usage, increasing it from 0.2266 to 0.4988, while significantly reducing structural instability. Pivot mediation introduces normalization pressure and measurable compression effects, whereas zero-shot translation exhibits the highest decision variance. We argue that translation divergence across agents reveals latent decision flexibility within multilingual models and we provide a principled interpretability framework for low-resource dialect generation.
来自人类反馈的强化学习（RLHF）已成为一种强大但样本效率较低的方法，用于从人类偏好中学习奖励模型，因此主动学习成为合成信息量丰富的偏好查询的关键组成部分。然而，对于大型神经网络奖励模型而言，主动学习所需的有效不确定性量化仍然是一个关键挑战。本文提出 PreferenceEKF，这是一种样本效率高的方法，通过将主动偏好学习表述为序贯贝叶斯滤波问题来跟踪奖励模型的不确定性。我们的方法不依赖于在完整神经网络参数空间上进行计算代价极高的后验推断，而是在低维参数子空间内使用扩展卡尔曼滤波进行序贯推断，并随着新的偏好查询到来持续更新奖励模型后验。我们的方法使得对神经网络参数进行可扩展采样成为可能，从而高效计算用于主动奖励学习的获取函数。在 D4RL 和 V-D4RL 基准上的实验表明，与其他贝叶斯深度学习方法相比，我们的方法在样本效率、运行时间、可扩展性和校准性方面都表现更好，而且所学习到的奖励模型能够带来具有竞争力的离线强化学习策略性能。这凸显了在 RLHF 中用于基于偏好的奖励建模的可扩展贝叶斯方法的潜力。我们的代码可在 https://github.com/yutaizhou/bnn_pref 获取。
For every coherent and sufficiently expressive finite syntactic system S, we prove the existence of at least one theorem that S cannot produce autonomously. The result is a metatheorem: it proves the existence of a theorem, and applies to every finite syntactic system - security mechanisms, AI systems, formal verifiers, legal systems, economic models, and the formal system in which it is itself proved.
当任务存在程序化检查器时，基于可验证奖励的强化学习效果很好，但大多数长程智能体领域并没有这类检查器。我们研究结果不可见设置，即无法获得真实成功信号的情形。多标准评分细则是一种常见的奖励提供方式；它们对每条轨迹只评分一次，但单一标量对于跨越数十步的过程来说是较差的信号。我们提出 DRACO：Distributing Rubric-based Advantage for Credit Optimization。它在训练期间动态生成评分细则，以跟踪策略不断演化的能力；在每条完整轨迹结束后对这些评分细则打分，并将该判断重新分配到负责这些已标注评分细则的步骤上，从而在 GRPO 中生成区分化的逐步优势。该重新分配具有闭式形式，不引入任何经过训练的归因模块。在 AppWorld 上，DRACO 比基础模型提升 15.9 分，且即使自身不使用任何验证器，也比使用稀疏真实奖励训练的 GRPO 高 5.3 分。在域外的 Tau-Bench 上，即使没有 frontier judge，它也比基础模型高 5.3 分，优于真实奖励训练和其他基于评分细则的训练设置。DRACO 的代码可见于 https://github.com/IBM/draco。
This paper proposes AdaRoboVLG, a task-adaptive Vision-Language-Grasp (VLG) framework that supports generalizable grasp synthesis across different robotic hands. Unlike existing VLG methods that tightly couple foundation models with end-to-end grasp policies, AdaRoboVLG learns an efficient generalizable base policy that generates and evaluates physically feasible grasp candidates through explicit kinematic mapping and force-closure-based stability estimation, while offloading task-dependent understanding to specialized foundation-model modules. These modules provide composable priors that are integrated into the grasp synthesis process, enabling contextually adaptive grasp synthesis without retraining the underlying grasp policy. Through extensive simulation and real-world experiments, we demonstrate that (i) the base policy exhibits efficient learning and strong cross-hand generalization, (ii) the framework effectively incorporates spatial, cognitive, and temporal priors to address three representative grasping challenges without compromising grasp synthesis performance compared to state-of-the-art methods, and (iii) these priors can operate jointly to enable functional grasping in cluttered and dynamic environments. These results indicate that decoupling physical grasp synthesis from task-dependent understanding provides a scalable paradigm for robotic grasping, allowing future advances in foundation models to be directly translated into improved grasp capabilities without redesigning or retraining the underlying grasp policy. Supplementary videos are available at https://adarobovlg.github.io/
Scaling interactive and verifiable environments is critical for training terminal agents. As frontier models become more capable, environments synthesized from scratch become less challenging and thus provide limited learning signals. Recent co-evolution methods iteratively synthesize environments near the model's learnable frontier based on weaknesses exposed during rollouts. However, their dependence on on-policy rollouts limits generalization and the continuous provision of learning signals as the model becomes stronger. In this paper, we propose environment evolution, which incrementally increases environment difficulty off-policy and schedules the evolved environments generation by generation during training to provide continuous learning signals. We derive three evolution directions that influence environment difficulty from the multi-turn learning objective and then implement evolution along these directions through a loop-engineered multi-agent harness. Quantitative rollout experiments with Hy4 preview, Claude Opus 5, and GPT-5.6 Sol show that environment evolution consistently produces more difficult environments. We validate its effectiveness on Qwen3.6-27B and Qwen3.6-35B-A3B through simple long-horizon RL training, improving their performance by 14.4 and 18.0 percentage points on Terminal-Bench 2.1, respectively.
本文提出一个用于迷你 Ackermann 车辆端到端自动驾驶研究的低成本开放实验平台。该平台结合了实体车辆、打印的城市赛道、数据采集工具、轨迹标定以及 Webots 数字孪生，从而支持将基于仿真的自动驾驶方法与真实世界执行相衔接的受控实验。作为第一个基线，我们实现了命令条件化的行为克隆，其中一个神经策略接收车载相机图像和高层导航指令，并输出转向和速度。该系统在实体车辆和仿真中都进行了评估。在真实闭环实验中，所学策略能够沿车道行驶并执行指令转向，相对于参考路线的平均横向偏差达到 6.1 厘米，接近人类示范中的 4.7 厘米。在数字孪生中，摄像头视场对性能有显著影响，当视场角从 58 度扩大到 120 度时，平均横向偏差从 35.6 厘米降至 3.3 厘米。我们进一步利用数字孪生生成合成驾驶数据，并使用学习得到的 sim-to-real 图像翻译器缩小外观差距，结果表明：在这种合成数据上结合真实示范训练的容量更大的策略，是唯一能够在闭环中完成全部四条赛道路线的配置；而紧凑型基线以及仅用真实数据训练的同一网络完成的路线更少。这些结果表明，该开放平台是开展 sim-to-real 研究的实用测试平台，并提供了一个初始的命令条件化模仿学习基线；我们将其开源，以支持可复现研究。
Research and news coverage of language-model deception increasingly attributes human-like mental-state concepts to language models. Such claims can blur the distinction between behavior that looks deceptive and a mechanism that is actually deceptive. We introduce a causal taxonomy separating prior commitment from retrospective report, model preference from realized output, false preference from sensitivity to the utility of misleading a recipient, and deceptive behavior from the provenance of the objective or strategy producing it. We test these distinctions in two open-weight model families. Across controlled guessing-game and stock-trading experiments, we find that deceptive-looking behavior can arise without the corresponding proposed mechanism, while other interventions provide direct evidence that recipient information state can causally affect deceptive preference. These results show that deceptive behavior can provide evidence for a deceptive mechanism. But even evidence for such a mechanism does not establish model agency in the deception.
代码仓库级软件工程基准显著推动了对编码智能体的评估，但现有基准主要衡量生成的补丁是否通过功能测试，而忽略了源自代码审查的验收约束（审查约束），这些约束在真实软件开发中往往会影响一个补丁是否可接受。我们引入 SWE-Gate，这是一个面向软件工程智能体的代码仓库级基准，明确在功能正确性之外评估对审查约束的遵循情况。SWE-Gate 从真实拉取请求审查评论中提取审查约束，并围绕这些约束合成代码仓库级修复实例。每个实例都提供独立的功能测试和约束测试，以及不合规补丁和黄金补丁，从而能够明确区分问题解决能力与审查约束遵循能力。我们构建的 SWE-Gate 包含 303 个代码仓库级修复实例，覆盖 75 个开源 Python 代码仓库，涉及多种软件领域。在一个通用编码智能体框架下，对涵盖不同能力水平的四种 LLM 后端进行实验，结果揭示了功能成功与在完整修复规范下成功之间存在显著差距：在通过功能测试的 644 次修复中，有 221 次未能满足所提供的审查约束。这些发现表明，仅基于功能的评估会高估智能体满足代码仓库级修复任务完整要求的能力。包含代码、数据和实验结果的复现包可在 https://github.com/DeepSoftwareAnalytics/SWE-Gate 获取。
https://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face... 评论链接： https://news.ycombinator.com/item?id=49548952 得分：256 评论数：78
文章链接: https://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face/ 评论链接: https://news.ycombinator.com/item?id=49548952 积分: 19 # 评论数: 3
Aggregating noisy, conflicting textual hypotheses into a reliable consensus is a fundamental challenge when deploying NLP systems in real-world industrial settings. While monolithic Large Language Model (LLM) agents offer unbounded expressivity for tasks like Root Cause Analysis (RCA), they suffer from context limits, compounding hallucinations, and prohibitive inference latency. Traditional weak supervision offers statistical rigor but is mathematically restricted to discrete classes. We present Loom, a generative consensus framework deployed for real-world RCA that bridges these paradigms. Loom aggregates open-form hypotheses emitted by modular heuristics (diagnostic templates dynamically populated with episode-specific entities, times, and metrics) by projecting them into a continuous embedding space, and resolves conflicting signals with an iterative centroid-based reweighting algorithm. The resulting consensus weights ground a single lightweight LLM synthesis step. Evaluated on the OpenRCA benchmark, Loom occupies the accuracy--efficiency Pareto frontier: it matches a state-of-the-art autonomous agent on Bank and Market-2 and trails on Market-1 and Telecom, while using a single LLM call per incident on all four datasets ($\sim$26$\times$ faster; $\sim$33$\times$ with an 8B-parameter synthesizer). We discuss our deployment experience, highlighting lessons learned regarding the trade-offs between agentic depth and inference latency, negative results in redundancy detection, and how deterministic consensus fosters trust among Subject Matter Experts~(SMEs).
大型视觉语言模型在视觉-语言任务中已取得显著成功。然而，它们仍然容易出现视觉幻觉（VHs），从而削弱其在真实世界应用中的可靠性。现有解决方案通常需要精心构建的数据集、额外训练或多轮解码，因而带来相当大的计算开销。本文提出 \textbf{RVSD}（\underline{R}etrieval \underline{V}ision \underline{S}parse \underline{D}ecoding），这是一种无需训练且即插即用的解码框架，首次在一次解码过程中将 token 稀疏化与 \textbf{Semantic-Space Visual Retrieval}（SSVR）统一起来。在 RVSD 中，我们引入一种 \textbf{语义导向的 token 选择} 策略，在保留关键视觉信息的同时，有选择地稀疏化冗余 token。我们进一步提出 SSVR 机制，将视觉补偿重新表述为共享语义空间中的按需跨模态检索过程。大量实验表明，RVSD 在缓解 VHs 方面实现了最先进的性能，同时在长上下文生成设置下保持了强健的抑制能力。代码可在此处获取。\footnote{https://github.com/canjie-liu/RVSD}
自主智能体正开始端到端地开展机器学习（ML）研究。这些智能体将模型主干与用于规划、执行、记忆和验证的 harness 结合在一起，但这种架构仍将领域特定的诀窍留在智能体之外。我们将这一缺失层称为操作性知识，即让“知道一种方法”与“把它真正跑通”之间拉开差距的经验。这类知识并不缺失于该领域。它存在于仓库和论文中，但以面向人类读者的形式呈现，而且在任务过程中大到无法直接加载。一旦被蒸馏为紧凑、经过验证的技能，这些知识便可跨任务复用，而不是在每次运行时重新发现。 我们提出 DisCo，一种由技能驱动的研究智能体，它会创建技能并在研究中使用它们。它的蒸馏分两种互补形式进行：一种是任务无关的，将该领域广泛使用的仓库浓缩为可复用技能；另一种是面向任务的，生成具体任务所需的技能。前者应用于整个开放生态，产出了 AREX-Skill Library，其中包含从 1,000 个广泛使用的 ML 仓库中蒸馏出的 5,000+ 个经验证技能，并组织为 20 个领域和 178 个能力家族。在固定 GPT-5.5 主干、研究 harness 和下游执行预算的条件下，配备技能的研究智能体在 MLE-bench 上比不带技能的同一智能体高 134.3%，在 PaperBench 上高 34.4%，在 FrontierCS 上高 9.2%，在 PassNet 上高 14.0%。这些提升来自在该固定设置下加入蒸馏后的操作上下文。
多智能体 LLM 系统通常使用一个 orchestrator 将任务分解给一组工作者，随后通过文本反思进行改进。尽管经验结果很强，但这些系统缺乏对协调、记忆改进以及外部验证作用的统一解释。我们将 orchestrator-worker 交互建模为一个双层协调博弈：在有界耦合下，工作者的局部更新博弈是一个近似势博弈，其均衡松弛量由分解质量控制。随后，我们将反思分析为在语义记忆状态上的随机移动。对于自由形式反思，我们推导出一个有限时间上界，证明了最坏情况的紧致性，并在一个可证伪的持续伤害条件下给出一个正下界。我们进一步证明一个信息论不可能性结果：任何只观察生成文本记录的门控器都无法在文本不可区分的环境上实现一致改进，而基于环境的门控器则可以。受这一分离启发，我们提出 Stochastic Reflective Memory Ascent (SRMA)，它仅在基于环境的评估风险严格下降后才接受候选记忆。在校准和非退化纠正质量下，SRMA 可精确收敛，且收敛速率为几何级或多项式级；相应构造表明这两种速率情形在阶上都是紧的。我们还为随机评估提供置信门控，并为分段平稳环境提供重新锚定保证。实验使用基于环境的指标实例化这些对象，并检验所预测的协调与漂移规律。在 500 个 SWE-bench 实例上，完整的基于 Kimi 的系统成功解决了 72.2%，而公开的 mini-SWE-agent 参考系统为 70.8%。代码：https://github.com/YihangChen9/Bilevel-Coordinated-Reflection
大型语言模型如今已能以专家级水平回答医疗问题。然而，这些系统所依据的上下文可能具有误导性，而误导性上下文会破坏模型的医疗判断。为了解误导性上下文如何破坏这种判断，我们考察模型对上下文的易受影响程度、对其的披露情况、推理被污染的机制，以及决策的可监控性。在 MedMisBench 的医疗推理子集上，这是一个经过临床医生审查的问答基准，包含 8,627 个问题，我们注入了两类误导性上下文线索：捏造证据和裸断言。我们测试了三种推理模型，其中两种会展示完整推理轨迹，另一种前沿模型只展示其回答。三者都比对捏造证据更容易受断言影响，采用被断言答案的频率高出 10 到 27 个百分点。误导性线索在 81% 到 98% 的轨迹中被披露，但在响应中仅有 7% 到 90%；而且断言被披露的频率低于基于证据的线索。不显示披露信息、仅对推理轨迹进行重采样表明，这两类线索以不同方式污染推理：证据更早进入并逐渐累积，而断言则在接近结尾时改写结论。一个 LLM 监控器在带引导地阅读开源模型的推理轨迹时，能以 5% 的假阳性率捕捉到 78% 的被污染决策，而从任何响应中最多只能捕捉到 32%。模型最容易受影响的误导性上下文恰恰是最少被披露的，而只有在开放的推理轨迹中才能被可靠地发现；前沿模型提供方会将这类轨迹隐藏起来。
Training data attribution (TDA) aims to identify training examples that shape model behavior, but its intervention value depends on both which examples are selected and how they are modified. Influence functions (IF) estimate behavioral changes under infinitesimal reweighting, yet IF-selected examples often show limited advantages over random selection under conventional weight-based interventions. This raises the question of whether influential examples lack intervention value or whether reweighting fails to realize their behavioral leverage.We introduce influence-guided response rewriting, which uses IF to identify intervention targets and replaces their responses with behavior-aligned or behavior-opposed supervision while keeping instructions fixed. Across four open-weight LLMs, we compare rewriting and reweighting on the same influence-selected examples using epistemic abstention as our primary testbed. Response rewriting produces stronger, more persistent, and bidirectional behavioral shifts, while reweighting the same examples yields weak and inconsistent effects. Further analyses show that influence-selected examples provide greater rewriting leverage than alternative selectors, with changes remaining concentrated on target-relevant behaviors. The same qualitative contrast extends to safety refusal. These results distinguish the local reweighting effects captured by influence estimates from the broader intervention leverage of the examples they identify, motivating intervention-aware evaluation of TDA methods.
近期的 web agent 使用世界模型进行测试时的动作选择：通过采样候选动作、预测由此产生的 web 状态，并用 ranker model 或 Process Reward Model (PRM) 对其排序。这些世界模型通常通过监督式下一状态预测进行训练，以生成固定表示，例如 HTML 或 AXTree 快照。然而，这一目标与下游 ranker 并不一致，因为 ranker 依赖于预测状态在不同候选之间具有可区分性，从而准确打分。为了解决这一问题，我们提出了 predicted-state matching，这是一种训练目标，要求预测表示必须将真实产生的状态与由其他动作到达的状态区分开来。我们使用一个分支式 web-agent 数据集训练这些模型，该数据集由 WebArena Go-Browse 轨迹构成，其中每个决策点都包含多个备选动作及其对应的结果状态。我们在留出的 predicted-state matching 基准上的实验表明，我们的方法优于使用监督式下一状态预测训练的 world model。我们进一步表明，与仅基于动作的 PRM 以及结合了监督式下一状态 world model 的 PRM 相比，我们的方法在 WebPRMBench 上提升了基于 PRM 风格的动作排序效果。最后，在 WebArena-Lite 上，使用我们的 world model 进行测试时动作选择能够提升端到端任务成功率。我们的项目主页见：https://dhruvpendharkar.github.io/dwm/。
尽管现有关于 LLM 作者归因（AA）的研究已经取得进展，但可用基准仍然有限，往往只关注英语、受控环境或相对较旧的模型，而少数多语言研究也只考虑了相对较短的文本。我们提出 MultiGhostBench，这是一个多语言基准，包含由五个较新的 LLM 生成的 928 本书，覆盖六种语言和三种文字系统，平均每本书长度约为 59K 词。该基准支持在领域、作者和语言偏移下进行评估。对代表性 AA 方法的评估表明，没有任何单一方法能在所有设置中始终表现最好，而且在分布偏移下性能通常会下降。基于 Transformer 的检测器能够跨语言保留与生成器相关的信息，不过迁移效果会因语言对而异；而统计和指纹式检测器则更依赖语言。我们认为 MultiGhostBench 将成为开发和评估稳健 AA 方法的重要资源。数据集和代码见 https://github.com/GrecoMT/MultiGhostBench。
Political evasion refers to responses that engage with a question while withholding the requested information. Recent NLP work frames political evasion as a classification task using a two-level taxonomy of response clarity and fine-grained evasion strategies. Existing work on response clarity and evasion classification is limited to English, leaving open whether the taxonomy and model behavior transfer across languages and political contexts. We introduce PolERo, a dataset of 3,574 human-annotated question-answer pairs extracted from official transcripts of five Romanian presidents. We evaluate multiple classification approaches on both datasets under matched conditions, including TF-IDF baselines, fine-tuned encoder models, a proposed sliding-window encoder, and zero/few-shot LLM prompting. We study cross-lingual transfer through joint bilingual training and machine-translation-based data augmentation. Our results indicate that fine-tuned encoders are competitive, cross-lingual transfer is asymmetric, and ambivalent evasion categories involving pragmatic cues remain the main challenge across all model families.
多轮越狱攻击表明，恶意意图可以分散在对话中，但现有方法却掩盖了究竟是哪种对话机制驱动了这种脆弱性。我们提出 BLUEPRINT，这是一种安全评估框架，将分解后的社会影响策略空间与 WORLDVIEWSIM（跨轮次情境上下文模块）分离开来。蒙特卡洛树搜索在四轮轨迹中优化每轮层面的18个基于理论的影响因素组合。在六个前沿模型上，BLUEPRINT 在主要开源权重模型和专有模型上都实现了接近满分的 ASR，同时所需平均查询次数最少（2.46 次）。由此得到的轨迹进一步揭示了在抗拒性较强的目标模型中存在模型特异性的脆弱性：每个模型对不同的影响因素和策略转换作出响应，但它们都共享一条共同的恢复路径——转向具体、可执行的任务表述，持续能够摆脱强硬拒绝状态。消融实验证实，操作性线索最为重要：让请求变得可执行影响最大，收益框架格外有效，而某些合法性诉求反而可能适得其反。这些发现表明，要实现稳健的多轮安全，不仅需要监控有害内容，还要监控对话状态如何使不安全请求显得具体且在局部上可执行。
我们提出 CivBench，这是一个开源基准，用于通过 Model Context Protocol (MCP) 评估长时程、工具介导环境中的语言模型智能体。单个回合跨越 300+ 个回合，并在大动作空间上产生数千次工具调用，要求在部分可观测性下持续进行规划、状态监控和执行。该环境暴露了 76 个 MCP 工具，以及一层叙述模块，可将视觉游戏状态转换为结构化文本。 我们使用 CivBench 在 23 次有效运行中，刻画四个模型家族的智能体行为。该样本是试点研究，不是模型排名：在这一规模下，总体结果并不能可靠地区分不同模型。相反，我们提出了两个环境使其可测的接口层指标：Proactive Monitoring Rate (PMR)，用于衡量智能体是否主动查询潜在的战略状态；以及 RAG@10，用于衡量结构化规划反思中陈述的承诺是否会在随后十回合内执行。 在共享 playbook 协议下，我们在多次运行中观察到两个一致模式。智能体对可用但需要显式查询的、与战略相关的状态监控不足：尽管 playbook 建议每 20 回合查询一次胜利进度，它们实际只会每 30 到 75 回合查询一次；并且在 20 场可检测到的失败中，有 7 场它们未能在游戏结束前的 20 回合警示窗口内进行查询。智能体也经常未能执行自己规划反思中陈述的近期承诺（不同模型的 RAG@10 在 48.2% 到 65.8% 之间）。这两种模式都出现在具备工具访问权限和明确指导的情况下，我们将其解读为在指令下的偏差，而非能力缺失。 我们在 https://github.com/lmwilki/civ6-mcp 发布环境、场景、日志、指标和分析流水线
我们研究对电商用户在品牌、尺寸和颜色等商品属性上的偏好进行预测——我们将这一任务定义为属性亲和度（Aspect Affinity）。解决这一任务有助于提升对顾客的理解，并支持推荐、搜索和营销中的细粒度个性化。我们将 Aspect Affinity 表述为一个时间预测任务：根据按时间排序的交互历史，预测用户未来的属性选择，从而捕捉超越当前会话、不断演化的长期偏好。为此，我们提出 DeepAffinity，它利用 Small Language Models（SLMs），结合结构化提示和针对该任务微调的专门预测头。我们表明，DeepAffinity 优于标准的生成式微调方法，而通用的开源 LLMs 在没有任务特定调优的情况下表现较差，凸显了它们在建模细微行为方面的局限性。最后，DeepAffinity 提升了一个大规模跨国电商平台上的推荐质量。
现代大型语言模型（LLMs）依赖强化学习来在单个领域内构建强大能力，但将这些能力整合到一个可部署模型中仍然具有挑战。现有方法通过将每个样本路由给其领域匹配的教师，让领域标签决定由哪位教师提供监督。然而，领域专长只在平均意义上成立：匹配的教师并不总是在某个样本上正确，而其他领域的教师有时反而正确。因此，必须按样本而非按领域来识别可靠教师。本文提出 Multi-Teacher Self-Distillation Policy Optimization（MT-SDPO），一种在策略蒸馏方法，将多个冻结教师统一为一个学生模型。MT-SDPO 包含三个组件：（1）self-anchors，即用其所在组中的正确 rollout 对某个 rollout 进行监督；（2）answer-verified eligibility，即只有当教师自己的答案通过验证器时，才可监督该样本；（3）privileged distillation，即将锚点和所有已验证反馈合并为一个上下文，由指数滑动平均自教师读取，而学生不能读取，从而在部署时保留单一策略。我们在来自三个模型家族的五个学生上验证，MT-SDPO 将 Qwen3-8B 最弱领域提升了 14.79 个百分点，并将其领域差距缩小了 74.7%，比按领域提供一个匹配教师的方式实现了更好的平衡。应由经过验证的可靠性，而非领域归属，来决定谁来教学。代码见 https://github.com/hexixiang/MT-SDPO。
药物-靶标相互作用（DTI）预测是 AI 驱动药物发现中的一项重要任务。尽管近期的生化表示学习方法提升了 DTI 预测性能，但它们被动的特征聚合倾向于偏向占主导地位的分子模式，而抑制诸如功能基团和残基上下文模式等较弱但与结合相关的信号，从而限制了多尺度生化对应关系的建模。为解决这一问题，我们提出 ProbeMatchDTI，一个由模式探针驱动的框架，包含 IterProbe 和 BindingProbe。IterProbe 在细化深度上显式保留上下文状态，并在跨实体匹配之前，使用可学习探针在每个位置选择这些状态，从而保留较弱的生化模式并强化功能基团、局部基序和分子骨架之间的关联。随后，BindingProbe 从局部生化单元和整体配对两个层面刻画跨实体的药物-蛋白互补性，在保留较弱结合相关关联的同时，联合建模细粒度相互作用和多尺度对应关系。大量实验表明 ProbeMatchDTI 具有优越性，在 BindingDB 和 DrugBank 上分别将 AUC-ROC 提高了 2.0% 和 0.5%。特征级模式分析进一步刻画了其在跨尺度生化模式匹配中的探针驱动行为。我们还将 ProbeMatchDTI 的预测与一种证据引导的下游药物发现工作流相连接，证明了其在候选物筛选与验证规划中的实用性。我们的代码可在 https://github.com/developer-hq/ProbeMatchDTI 获取
文章 URL: https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/#core-takeaways-about-this-incident 评论 URL: https://news.ycombinator.com/item?id=49543841 得分: 11 # 评论数: 0
Last year, I wrote a speculative cosmology piece called Blowtorch Theory. It rather cheekily extended the American theoretical physicist Lee Smolin's wonderful, powerful, oddly neglected theory of cosmological natural selection; the idea that universes reproduce, with variation, leading to Darwinian evolution of universes. While I was having dinner with my five-year-old, a friend texted to say that the piece was at number one on Hacker News. At the time, I didn't really know anything about the website, or its readers – but I learned a LOT over the next few hours, as my dinner went cold (and my son got to watch cartoons)! Anyway, I greatly enjoyed (most of) the robust discussion that followed there, and thought Hacker News readers might be interested in this new Smithsonian Magazine overview of the remarkably healthy state of evolutionary cosmology, a new field emerging around these ideas. Think of it as a startup disrupting cosmology! Comments URL: https://news.ycombinator.com/item?id=49539805 Points: 30 # Comments: 20
The Fairwind Program is a limited access program for governments and trusted partners to use our cyber defense tools.
文章链接： https://github.com/mlc-ai/web-llm 评论链接： https://news.ycombinator.com/item?id=49536411 得分：3 # 评论数：0
本文研究自主软件开发，即基于 LLM 的编码代理在无需人工干预的情况下，将高层需求转化为完整、可运行且可用的软件系统。我们提出 Harness-of-Harness (HoH)，这是一个使编码代理在自主开发过程中持续改进软件的框架。HoH 基于现有的 coding-agent harness 运行，并将其执行组织为迭代式的规划-编码-测试循环。为在各轮循环中持续提升，HoH 在修复与能力增长之间取得平衡，将开发范围限定为小而可验证的增量，将实现阶段测试与独立评测分离，并约束可验证的输出而非规定代理的工作流。它逐步开放交付物、面向角色的工具和技能，鼓励复用而非重建，并维护带版本记录的项目历史。在 GameCraft-Bench、FrontierSWE 和 ProgramBench 上，三个 harness-model 配对（Codex with GPT-5.5、OpenCode with DeepSeek-V4-Pro，以及 Pi with MiniMax-M3）中，HoH 始终优于对应的独立 harness，在三轮迭代后取得平均相对提升 52.25%，最高提升 82.86%。在一次持续多日、超过 70 轮迭代的部署中，HoH 自主开发出一款第一人称射击游戏，具备连贯的故事线、完整实现的核心机制、可供人类游玩的体验、精致的视觉效果以及集成音频。Github: https://github.com/Flesymeb/HarnessOfHarness Project Page: https://flesymeb.github.io/HarnessOfHarness/
压缩后的上下文通常以人类可读文本或需要解码的渲染图像形式传递，即使它的使用者是语言模型也是如此。我们提出 LatentPress，它将对话历史和长文档写入第三种表示：连续记忆 token，冻结的解码器通过其输入嵌入接口直接读取这些 token，在推理时无需重建文本。一个与读者匹配的小型写入器仅训练一个适配器即可实现 4.2M-26.2M 参数（约占解码器的 0.1%），压缩率达到 4-16 倍。在 LongMemEval 上，LatentPress 在 7.70 倍压缩下达到 0.504 的准确率，而未压缩证据为 0.490，优于文本摘要（0.184）和基于 OCR 的压缩（0.426 到 0.312）。在 LongBench-QA 上，领域内写入器在 4-8 倍压缩下与原始上下文读取相当或更优，而 16 倍时则略逊于原始上下文。写入每段对话耗时 43 毫秒，大约比文本摘要或 OCR 重建快一个数量级；读取速度比原始上下文或缓存的 OCR 快 5-9 倍。我们在两种迁移设置下验证了该接口：从 UltraChat 零样本迁移到 LongMemEval 记忆问答，以及从基于 LongMemEval 的问答迁移到未见过的 LongBench 文档领域，从而确立了直接软 token 作为一种超越文本和视觉的实用、面向机器的上下文接口。实验实现可见：https://github.com/xuyd16ai/context_softtoken_compress 。
针对 Video-LLMs 的时间推理基准通常通过语言来介导，这为选项措辞、答案相关性或语言先验带来的语言捷径留下了空间。为减少这类捷径，我们提出 TempCloze，这是一个用于评估 Video-LLMs 视觉时间推理能力的视频填空基准。给定视频的开头和结尾片段，模型必须从四个候选项中识别真实缺失的中间部分。TempCloze 包含来自七个来源的 1,521 个经过精心筛选的视频，主要是长镜头和第一人称视频。我们沿三个维度构造同源干扰项：Semantic 询问应发生什么事件，Alignment 探查它应在何时发生，Progression 测试它应如何展开，同时共享的场景和物体降低外观线索。我们对 10 个专有模型和 21 个开源 Video-LLMs 的评估显示，Alignment 是主要瓶颈：模型往往能识别合理的语义内容和局部事件推进，但难以处理时间对齐。我们进一步在 TempCloze-Mixed 和 TempCloze-Hard 上，使用四个代表性模型进行错误模式和行为敏感性分析，以考察错误从何而来，以及候选项顺序、上下文方向、可见跨度、帧密度和测试时缩放如何影响模型选择。
Scientific agents must learn not only how to reason, but also what to believe. However, existing LLM agents typically express scientific hypotheses in free-form text, leaving their beliefs implicit and difficult to test or revise. We introduce EvoSCM, which equips scientific agents with explicit structural causal models that evolve as new experimental evidence is collected. EvoSCM maintains a population of competing SCM hypotheses, each encoding a candidate causal explanation of the environment, and evolves them through a closed discovery loop. In each round, the agent abduces latent mechanisms from accumulated evidence, designs discriminative interventions, and commits to falsifiable predictions that it tests through experimentation. Discrepancies between prediction and observation are inductively distilled into correction rules that revise the causal structures and mechanisms of each hypothesis, and the agent then deductively validates the revised population against accumulated evidence and structural consistency to guide the next round. We evaluate EvoSCM on DiscoverPhysics, a benchmark requiring agents to uncover the hidden dynamics of noncanonical physical worlds through experimentation. EvoSCM consistently improves scientific discovery over baselines, yielding more accurate explanations and predictions while making more effective use of experimental interactions.
As large AI models become increasingly prevalent across a wide range of applications, memory cost has become a critical bottleneck in both training and inference. To mitigate this issue, we introduce the Linear Reusable Neural Bases Architecture (LRNBA), a novel framework aimed at improving parameter efficiency and reducing memory cost. Inspired by recurrent neural network (RNN) designs, the core idea of our approach is to represent each network block as a linear combination of a shared set of neural bases, thereby enjoying highly network compression rate while maintaining stable training. The proposed architecture allows for the construction of significantly wider and deeper networks under the same parameter budget. Extensive experiments demonstrate that our model achieves comparable or even faster convergence and lower loss than classical architectures, while maintaining stable training dynamics.
科学方程发现长期以来一直是科学进步的核心，通常通过假设生成、观测检验以及在科学约束下的反复迭代来推进。随着 LLM 能力的提升以及其在 AI for Science 中角色的扩展，一个尚未解决的问题是：它们是否 באמת能够真正发现科学定律，以及这种能力应如何评估。现有评测往往要么通过合成设置简化发现过程，要么复用已发表的目标，而这些目标可能已经为 LLM 所熟悉。因此，我们提出 SCILAWS-BENCH，一个基于已发表研究和真实科学数据构建的科学定律发现基准。它包含从 381 篇科学论文中整理出的 118 个问题，覆盖 291 条候选定律，以及来自六个科学学科的大约 800 万个真实数据点。每个问题都在两种互补设置中实例化：(1) SCILAWS-REAL 要求模型基于固定的真实观测提出定律，并评估其留出预测拟合与源文献中导出的科学有效性；(2) SCILAWS-PARALLEL 要求模型主动查询残差校准的世界，并恢复从已发表形式中推导出的合成隐藏定律。这种双设置任务设计在保留每个问题科学背景的同时，分别评估固定记录下的定律发现与对新合成隐藏定律的主动恢复。我们发现，预测拟合与科学有效性可能并不一致，记忆会影响模型是复现还是超越已发表公式，而我们的 best-of-N 研究揭示了一个选择瓶颈。我们的工作提供了一个基于论文的基准，以及用于评估 AI 科学发现的新实证视角。Project page: https://yiyihum.github.io/SciLaws-Bench
仓库级代码生成任务要求合成既满足任务需求、又与目标仓库上下文保持一致的代码。由于真实世界的仓库往往超过 LLMs 的输入长度限制，现有方法通常采用检索增强生成（RAG）来提供特定于仓库的上下文。尽管这些方法提升了仓库上下文检索效果，现有方法通常将上下文作为任务级支持提供，而没有显式识别在生成过程中需要细粒度仓库上下文的关键 token。在 LLMs 的自回归生成过程中，错误往往集中在少数决定性位置：一旦这些 token 生成错误，后续代码可能沿着错误的语义路径继续，最终导致功能失败。我们将这些位置称为“关键 token”。在本文中，我们提出 ACToR，一种用于仓库级代码生成的自适应关键 token 感知检索框架。ACToR 在生成过程中识别关键 token，并在这些决定性位置按需触发定向检索，以提供仓库上下文。此外，我们为稠密检索器设计了一种位置感知加权方法，以优先使用对生成更具信息量的上下文。我们在两个具有代表性的仓库级基准 RepoExec 和 CoderEval 上评估了 ACToR。实验结果表明，ACToR 始终优于最先进方法，在 RepoExec 上取得了 8.4% 的相对提升，在 CoderEval 上取得了 15.4% 的相对提升。除了性能提升之外，我们还系统地量化了关键 token 的影响，揭示了它们在主要生成失败中的核心作用，并强调了定向检索策略的必要性。我们在 https://github.com/DeepSoftwareAnalytics/ACToR 提供了代码和数据。
在真实基准上评估软件工程 agents 成本高昂，因为每个任务都可能需要多步代码探索、修改和测试执行。现有的高效评测方法会选择具有代表性的子集来估计完整基准上的性能，但大多只关注结果：它们拟合历史通过/失败响应矩阵或静态任务语义，忽略了 agents 解决问题的方式。我们提出 PTA-IRT，一个 Privileged Trajectory-Aware Item Response Theory 框架，融合过程信号和结果信号。历史执行轨迹提供了超越通过/失败的过程级证据，例如探索过的上下文、尝试过的编辑以及求解路径，PTA-IRT 将其用作用于校准子集选择和能力估计的特权信息。在较低的校准预算下，PTA-IRT 在四个 SWE 基准上的分数恢复和排序恢复方面始终优于先前的 IRT 基线。代码和数据可在 https://github.com/DeepSoftwareAnalytics/PTA-IRT 获取。
Virtual staining aims to computationally generate target-stained histopathological images while reducing the cost and time associated with conventional staining procedures. However, existing methods rely predominantly on strictly paired and accurately registered training data, which are difficult and expensive to obtain in routine practice. To reduce this dependence, we propose a stable semi-supervised virtual staining framework that jointly exploits both limited paired data and abundant unpaired source images. Directly incorporating unpaired images is challenging because their generated results lack corresponding targets for supervision, potentially leading to unrealistic staining, morphological degradation, or even training collapse. To obtain reliable supervision from these images, Hessian-derived morphology preservation extracts structural cues from each source image and constrains the generated output to retain tissue morphology. Histopathological realism constraints further guide the output toward plausible target-stain characteristics, preventing the source-derived structural supervision from degenerating into contour enhancement or simple color transformation. Together, the two components suppress structural and appearance drift, stabilize semi-supervised stain translation, and promote the preservation of diagnostically relevant information. Extensive experiments on H&E-to-IHC translation for Ki67 and HER2, as well as FFPE-to-H&E translation, demonstrate consistent improvements in image quality, morphology preservation, robustness, and downstream diagnostic performance. Code will be available.
尽管多模态大语言模型（MLLMs）具有很强的多模态理解能力，但在处理长视觉 token 序列时会产生巨大的计算开销。为降低推理成本，近期研究通过以视觉为中心或文本引导的策略探索视觉 token 剪枝。然而，这些方法往往忽视高范数离群 token，即特征范数异常大的 token，从而导致次优的剪枝决策。 在本文中，我们表明这类高范数离群 token 在特征和空间两个维度上都高度冗余，但现有方法却常将其误保留为有信息的线索。 受此观察启发，我们提出 SinkPruner，一种无需训练的视觉 token 剪枝框架，用于高效的 MLLM 推理。SinkPruner 采用由粗到细的设计，包含两个关键模块：一个视觉净化器，用于过滤高范数冗余并缓解 attention sink 和 attention dispersion；以及一个文本引导的剪枝器，进一步保留与文本查询在语义上对齐的 token。 在十二个图像-语言和四个视频-语言基准上的大量实验表明了我们框架的有效性、效率和泛化性。值得注意的是，在 token 减少 89% 的情况下，SinkPruner 仍保留了 LLaVA-1.5（Qwen2.5-VL）原始性能的 96.5%（91.8%）。实验进一步表明，我们的视觉净化器在提升现有剪枝方法性能方面具有良好的可迁移性。我们的代码地址为 https://github.com/LaVi-Lab/SinkPruner。
Traditional network models from network science, such as the Erdos-Renyi and configuration models, generate random networks that reproduce few selected topological properties observed in real-world networks. Deep graph generative models emerge as a data-driven approach, leveraging deep neural network architectures to learn complex structural distributions directly from real-world networks to generate more realistic synthetic networks. Because real social contact networks cannot be shared due to privacy risks, synthetic networks serve as an alternative for developing and evaluating epidemic mitigation strategies. In this work, we evaluate deep graph generative models as well as the configuration from a network science perspective by assessing both the topological similarity between generated and real-world networks and their utility in identifying effective node immunization strategies to sup- press epidemic/misinformation spreading. It is found that two deep graph generative models produce synthetic networks that closely resemble the structural properties of real-world networks, enabling them to identify effective immunization strategies.
Automated mining of formal specifications is vital for verifying real-time systems. However, existing passive learning approaches remain restricted to deterministic specifications or limited fragments of Timed Regular Expressions (TRE). To our knowledge, this paper presents the first framework to tackle \emph{precise} passive learning for an expressive timed logic, \emph{Metric Interval Temporal Logic} (MITL) without relying on predefined templates or restricted logic fragments. Our approach formally reduces the timed learning problem into a scalable untimed one. By identifying quantitative timing differences between positive and negative traces, we synthesise precise timed constraints and inject them as new Boolean atomic propositions. This embeds timing into the alphabet, delegating the complex formula evaluation to highly optimised, off-the-shelf untimed LTL tools. Crucially, our framework is complete, guaranteeing a separating specification can always be found. We evaluate our implementation across several benchmarks, demonstrating the effectiveness of our approach.
Machine learning systems deployed for credit, hiring, and resource distribution are increasingly subject to regulatory oversight from policies such as the EU AI Act and GDPR. Current fairness governance practices rely on observational fairness metrics, post-hoc explainability, and immutable audit logs, but provide limited support for causal attribution and efficient evidentiary verification. We introduce Causal Evidentiary Governance (CEG), a framework in which regulated institutions commit to a versioned directed acyclic graph (DAG) that partitions causal pathways into allowable and disallowed groups. The Causal Harm Rate measures prediction variation attributable to disallowed causal pathways. Each decision is accompanied by a signed Decision-Evidence Packet (DEP), cryptographically binding the prediction to a digest of the published DAG and path-specific attributions. DEP digests can be appended to a Merkle tree to enable logarithmic-cost inclusion proofs. We validate CEG through a two-layer empirical methodology using demographic summaries from four years of PMA credit supervisory data to construct 10,000 synthetic credit applicants across four strategic DAG counterfactuals. Causal Harm Rate isolates injected causal effects more clearly than demographic parity or equalized odds. Cross-model validation and ablation studies assess robustness. Evaluation on the German Credit dataset shows that harm associated with specific causal pathways can be substantially understated by associational fairness metrics. Finally, a proof-of-concept implementation demonstrates operationally plausible throughput and highlights relevant performance tradeoffs.
Uniform-state discrete diffusion models update all tokens in parallel while keeping every position revisable. Even when the commonly used top-$p$ rule leaves only one candidate at a position, that choice affects only the current reverse step and can be revised at the next sampling step. We ask what changes when selected hypotheses instead become persistent context for later predictions. We therefore propose committed reveal sampling (CRS), a training-free sampler that stores selected argmax tokens and inserts them into subsequent model inputs. Our analysis gives a rationale for selecting later and for keeping selected tokens visible. Under the exact forward process, the Bayes error of selecting a clean token cannot increase as noise decreases, while in a simple latent-mode model, keeping the selected token visible helps later parallel predictions agree on the same sequence-level choice. Empirically, paired experiments on Duo-distilled then separate this persistent effect from single-step top-$p$ restriction and scalar temperature scaling. Under the same finalization rule, CRS without top-$p$ truncation reaches lower generative perplexity (GenPPL) than fixed $p=0.95$ and $p=0.9$ baselines across budgets of 8--64 function evaluations (NFE). At 64 NFE, the comparison at matched unigram entropy also gives lower GenPPL for CRS, yielding a more favorable GenPPL--entropy tradeoff. Base Duo shows the same direction in a descriptive comparison, while other diversity and continuation metrics can rank these operating points differently. These results identify support restriction and persistent context as distinct controls of that tradeoff.
Production LLMs must handle inputs that attempt to override system instructions, bypass safety policies or elicit harmful responses. A common mitigation is a separate guardrail model. Existing reports, however, provide little evidence on Russian prompt injection or Russian surface obfuscation. We present HiveTraceGuard-Pro, a 0.6B generative guardrail LoRA-tuned from Qwen3-0.6B. It is trained on Russian and English and uses one binary scoring rule (safe/unsafe) for the final target turn. Its training corpus pairs harmful examples, where a counterpart exists, with benign examples from the same domain and applies eight obfuscation transforms to both labels. In one harness, we compare HiveTraceGuard-Pro with thirty-four other guards on nineteen benchmark groups, sixteen of which are public. Its aggregate key is 0.7432, behind 0.7641 and 0.7552 for the two higher-scoring guards. Over the sixteen public groups alone, its key is 0.7153 and four of the thirty-four other suite guards score higher. In a fifteen-model comparison, HiveTraceGuard-Pro has the highest clean Russian robustness combined-F1 (0.88) and Russian prompt-injection recall (0.999). Both results use Russian sets assembled by our team, and at least 27.1% of the prompt-injection set overlaps the training corpus. Its 14.3 ms median latency is the lowest among those fifteen models in that run. Across the suite, FPR is 0.268 and FNR is 0.156. All reported response results use a legacy standalone-reply serialization rather than the natural assistant-role path of the shipped chat template. We release the merged weights on Hugging Face under Apache-2.0. The corpus, evaluation sets and evaluation code remain internal.
Satellites are emerging as promising platforms to extend generative \emph{artificial intelligence} (AI) services to remote areas lacking terrestrial infrastructure. However, deploying space generative AI is fundamentally constrained by the limited, time-varying onboard energy supplied by solar \emph{energy harvesting} (EH). This paper presents a framework for solar-powered space generative AI in which a satellite receives a user prompt, executes a diffusion-based image-generation model, and downlinks the compressed result within a strict time window. We identify the fundamental \emph{computation--communication} (C$^2$) trade-offs governed by the shared harvested-energy budgets. Specifically, increasing the number of generation steps improves intrinsic image quality but depletes energy and time available for downlink transmission, whereas prioritizing communication guarantees reliable delivery but sacrifices semantic quality. To balance these trade-offs and maximize \emph{end-to-end} (E2E) generative performance, we exploit the predictable solar-EH dynamics induced by deterministic orbital motion and develop a joint C$^2$ resource-optimization framework using a tractable two-step approach. First, we characterize the maximum downlink throughput for a fixed generation depth under continuous solar EH. This establishes a separation principle that decouples waiting-time selection from optimal transmit-power control. Next, we formulate a joint C$^2$ utility-maximization problem and derive a closed-form, low-complexity step-selection policy in the dominant constant-power regime. Extensive experiments under realistic orbital dynamics demonstrate that the proposed policy dynamically balances generation quality and transmission reliability. This yields significant E2E performance gains over static computation- and communication-centric baselines across diverse solar-EH states.
Stain normalization reduces color variations caused by variations in staining protocols and imaging conditions, thereby enhancing computer-aided diagnostic system performance. Traditional methods derive mapping relationships from individual or limited reference images through pixel-wise transformation, offering style flexibility but suffering from inaccurate color mapping extraction. While existing deep-learning-based approaches achieve accurate dataset-wide color mapping through complex neural networks, they face challenges including computational inefficiency, artifact generation, and fixed normalization directions requiring model retraining for directional changes. To address these limitations, we propose StainPresetNet - a novel framework that combines structural preservation with dataset-level color mapping while maintaining computational efficiency. Our method implements pixel-wise normalization guided by preset reference images, enabling multi-directional adaptability without retraining. Evaluations on cytopathology and histopathology datasets demonstrate that StainPresetNet achieves superior color mapping accuracy compared to conventional methods, effectively improves classifier generalization in diagnostic tasks, and reduces computational overhead by 90\% versus existing deep learning approaches. The proposed preset-guided mechanism facilitates flexible adjustment of normalization directions through simple reference image replacement, overcoming the directional rigidity of current deep-learning-based solutions.
See how Gilbert + Tobin combines CEO-led commitment, rigorous governance, and human accountability to scale ChatGPT Enterprise and Codex across the firm.
I'm the developer. I built this because AI-generated conspiracy videos were getting out of control in my own YouTube feed. Weedout is a $1.99 Safari extension for macOS that removes videos YouTube labels “Made with AI” from feed, search, related videos, playlists and Shorts. It uses YouTube's label rather than AI detection, runs locally, and does not catch unlabeled videos. Source for developers who want to fork or build: https://github.com/masteranza/weedout-for-youtube Pull requests are not accepted; the repository is for independent forks. Comments URL: https://news.ycombinator.com/item?id=49528895 Points: 3 # Comments: 0
Hi! Every month the "Who Wants to Be Hired?" and "Who's Hiring?" Threads come through I would think to myself, "there should be a match maker for these two threads". So I had abacus.ai whip one up. The methodology is pretty simple, data is extracted from posts using an LLM, score matches based on salary, domain experience, remote/onsite, etc., and then eliminate incompatible postings, like 'looking for remote work' and 'onsite only'. The result is two views: jobs-by-user and user-by-jobs. If you've submitted a post to this month's 'Who Wants to Be Hired?", you can find your matching job listings at https://hnmatchmaker.com/user/:user_name , for example https://hnmatchmaker.com/user/G4Vi --- Some observations when looking at the matches pair up right now: Poke around, let me know what you think! Ultimately, I hope this is actually helpful to some folks. Comments URL: https://news.ycombinator.com/item?id=49528057 Points: 3 # Comments: 1
Hi HN! My son spent a lot of time in 5th grade playing with the 2-dimensional PhET gravity/space sims[1]. His STEM class uses it, and he was playing with it at home, and has been looking for more. I found several space sims/playgrounds online, including NASA's Eyes on the Solar System[2], but nothing that seemed to offer a real progression from what he was getting out of the PhET sims, so I built Newton's Orchard over the summer for him. https://newtonsorchard.app The source is also published at https://github.com/andrewchilds/newtons-orchard It features deterministic time scrubbing, all bodies are editable, different preset systems and experiments, missions to complete (complete 3 to unlock the blackhole object type and mission), and a gallery for user submissions. I would very much appreciate feedback from HN - particularly from the lens of making it engaging and illuminating for students, and anyone that would like to just get a feel for how gravity works. Thank you! [1] https://phet.colorado.edu/sims/html/gravity-and-orbits/lates... [2] https://eyes.nasa.gov/apps/solar-system/ Comments URL: https://news.ycombinator.com/item?id=49524469 Points: 4 # Comments: 0
嗨，HN，我是来自 Nori Robotics 的 Antonio（ https://norirobotics.com ）。我们在旧金山为机器人开发者和研究人员打造了一款售价 $1,688 的双臂移动机器人。 我在 Columbia 做机器人研究时开始研究 Nori。我当时通过人类演示来教机器人，但很难拿到价格可承受的硬件。大多数实验室只有一两台昂贵的机器人，这使得收集大规模数据集、进行长时间实验，或在多台机器人上测试都很困难。 所以我自己做了一台，经过七次迭代后。当前的 Nori 具备： * 19 个自由度 * 两条 7+1 DOF 机械臂，每条机械臂载荷 1.5 kg * 55 kg 伸缩升降机构 * 差速轮式底盘 * 四个 720p、30 fps 的 RGB 摄像头 * 2D 激光雷达 * 带全双工语音通信的双麦克风阵列 * 432 Wh 电池 * 一块配备 4 GB RAM 的 Raspberry Pi 5 * （SLAM 和安全功能在机载端运行，更重的 ACT 和 VLA 必须通过 LAN 从电脑运行，或通过 WAN 从服务器运行） 要把成本压到 $2,000 以下是主要的工程挑战。Nori 有超过 100 个运动和结构部件，因此在执行器、轴承、布线、供电和装配上的成本会迅速累积。为了降低成本，我们做的一些主要选择是使用高减速比舵机而不是 QDD 电机，以及采用轮式底盘而不是腿部。 我们在旧金山组装每一台机器人，并将其设计得易于制造和维修（我们提供用于打印维修件的 3D 文件）。 我们的开源 SDK 包括遥操作和演示工具： https://github.com/Nori-Robotics/nori-sdk-py 我们还构建了一个基于浏览器的模拟器，你可以试用一下： https://lab.norirobotics.com/nori/model 我们已经交付了第一台机器人，并正在生产下一批。最终，我们希望没有机器人经验的人也能教 Nori 完成任务，并与其他拥有者共享。 目前，这套硬件已经能够完成基础清洁任务、打开抽屉、补货上架，以及倒啤酒。这里有一段 Nori 做这些事情的视频： https://youtube.com/shorts/VRfVXHfQvD8 我们的盈利方式是以 $1,688 出售硬件，另可选购付费软件。部分硬件是开源的。更多细节见我们的硬件论文： https://doi.org/10.48550/arXiv.2605.16537 如果你在机器人领域工作，你会用这个价位的机器人做什么？你会对硬件做哪些改动？ 评论网址： https://news.ycombinator.com/item?id=49525153 分数：2 # 评论：0
Please state the location and include REMOTE for remote work, REMOTE (US) or similar if the country is restricted, and ONSITE when remote work is not an option. Please only post if you personally are part of the hiring company—no recruiting firms or job boards. One post per company. If it isn't a household name, explain what your company does. Please only post if you are actively filling a position and are committed to replying to applicants. Commenters: please don't reply to job posts to complain about something. It's off topic here. Readers: please only email if you are personally interested in the job. Searchers: try https://nthesis.ai/public/hn-who-is-hiring , https://dheerajck.github.io/hnwhoishiring/ , http://nchelluri.github.io/hnjobs/ , https://hnjobs.emilburzo.com . Don't miss this other fine thread: Who wants to be hired? https://news.ycombinator.com/item?id=49522896 Comments URL: https://news.ycombinator.com/item?id=49522897 Points: 9 # Comments: 8
文章 URL: https://mvakde.github.io/blog/44-on-arc-1/ 评论 URL: https://news.ycombinator.com/item?id=49519939 得分: 11 # 评论数: 0
Generative music systems are increasingly presented as tools that democratize music creation, yet their practical suitability for musicians remains underexplored. Prior work includes openness-focused evaluation frameworks, such as MusGO (Music-Generative Open AI), as well as qualitative studies of musicians' experiences with generative systems. However, these approaches do not support systematic comparison or early-stage discovery of models for creative use. Motivated by such limitations, we introduce MusGU+, a musician-centered evaluation framework organized around three dimensions: Adaptability, Usability, and Controllability. Together, these capture whether a model can be feasibly trained or fine-tuned on personal data, integrated into real-world music workflows, and controlled in musically meaningful ways. We evaluate 10 representative generative music systems and present an interactive discovery tool that enables musicians to explore and filter models according to these criteria. While MusGO remains valuable for promoting responsible research practices, MusGU+ supports informed selection and practical adoption of generative systems by musicians.
Counterfactual explanations (CEs) are widely used in explainable artificial intelligence (AI) to show how a model's outputs would change if the input features were manipulated. This technique is used for a range of tasks such as debugging models, explaining predictions, justifying decisions, and providing algorithmic recourse. In this paper, we explore the normative legitimacy of employing counterfactuals in real-life model deployment settings. We discuss the different stakes involved in these different purposes for which CEs are commonly employed, and find stricter requirements for justification and recourse. In particular, we find that naive application of CEs for justification and recourse can lead to ignoring contestable choices made throughout the machine learning (ML) pipeline, thus obfuscating that decisions and counterfactuals for those decisions are also artifacts of an organization's materialized design and governance choices. We demonstrate this with four empirical experiments involving interventions at stages of the ML pipeline ``upstream" of the explanation itself, and show that these affect the generated counterfactuals. We find that an organization's choices on measurement models for feature and labels, business requirements, model validation, and the metric of model success have as much or more impact on the generated counterfactuals as the specifics of the generating method. Our findings underline the need to account for such choices upon providing justification and recourse, providing a stark reminder of the relational nature of these tasks. As putative justifications or recourse recommendations, CEs do not provide adequate answers to some important "why"-questions because they preclude consideration of whether the decision-maker ought to have acted differently.
Vision-Language Models（VLMs）在需要复杂视觉理解的任务上仍然表现不佳。我们认为，核心问题不在于高层推理，而在于无法在图像中定位关键细节。由于这一缺陷，VLMs 往往会基于有缺陷的感知依据生成看似合理但实际上错误的推理。为了解决这一问题，我们提出 Locator-Critic（LOCI），这是一个无需训练的框架，将视觉搜索与证据验证解耦。LOCI 采用一个 Locator 代理来提出候选视觉证据，再由独立的 Critic 代理评估其相关性和充分性。这些代理在一个迭代的 Refinement Loop 中协同工作，逐步改进证据，直到其足以回答给定问题。这种解耦的、自我纠正的过程带来了显著的性能提升，在多个复杂视觉基准上取得了 state-of-the-art 结果。LOCI 同时提升了开源权重模型（如 Qwen3-VL，在 V* 上提升 +12.1，在 HR-Bench 上提升 +5.8，在 VisualProbe-Hard 上提升 +11.2）和专有模型（如 Gemini 2.5 Pro，在 V* 上提升 +8.9，在 HR-Bench 上提升 +4.3，在 VisualProbe-Hard 上提升 +4.8）的准确率。
我们提出 CogEvol，这是一系列专为 Learning Environment Generation 训练的模型：可在单次生成中将课程简介转化为完整的学习产物（结构化 JSON 幻灯片或自包含交互式 HTML 页面）。在 22 万条生产请求中，CogEvol 生成一份幻灯片的中位耗时为 17 秒，生成一个交互式页面为 59 秒，取代了耗时数分钟的多轮 agent 脚手架流程。可靠性不是寄希望于运气，而是被强制保障的：一个基于生产环境的数据管道将真实故障转化为 53,687 个经过验证的 SFT 样本，而混合的规则加 VLM 奖励驱动基于 GRPO 的 RL；在我们发现并修复了一起奖励黑客事件后，该系统进一步加固，该事件曾生成视觉上令人信服却无法游玩的游戏。CogEvol-27B 在幻灯片质量上得分 83.7，在一个包含 500 个案例的 interactive-HTML 基准上得分 63.7，参数量比旗舰级代码模型少 26.9 倍，并与 OpenMAIC 团队合作，承载其线上生产流量。CogEvol-4B 已在 Apache 2.0 许可下于 https://github.com/CogEvol/CogEvol-4B 开源发布；外部旗舰模型在相同基准上、使用完全一致的评测框架进行测评。脚手架编辑进一步将交互式页面生成成本降低约 76%，整套系统可在国产 Ascend 加速器上运行，并在应用层面与 A800 GPUs 持平，从而降低 AI 原生教育的大规模单位成本。
Joint-Embedding Predictive Architecture (JEPA) has shown strong performance in learning rich representations through self-supervised prediction in latent space. However, it typically relies on teacher--student architecture with an EMA to stabilise training, and can tend to yield uninformative representations. Contrastive learning is stable to train and produces strong global representations, but remains limited on local tasks by the global nature of its objective. In this work, we combine both into CoJEPA: a single shared backbone jointly trained with a JEPA objective on masked sequence tokens and a contrastive objective on the class token. The contrastive gradient provides stability, removing the need for an EMA teacher entirely, while JEPA enriches the sequence tokens via local predictions that contrastive learning alone cannot provide. Crucially, no extra parameters are added to the backbone: the same model is guided towards richer representations purely through the design of its training signal. CoJEPA takes the best of both worlds, outperforming or matching both individual methods across global and local MIR tasks, with a particularly strong advantage on tonal and harmonic understanding, and without any task-specific architectural changes. CoJEPA shows that combining objectives with complementary inductive biases can substitute for scale, encouraging future work to invest in smarter training objectives over ever-larger models.
我们研究自我建模：即 LLM 回答关于其自身行为问题的能力。我们关注可验证的行为问题，例如某个 prompt 编辑是否会改变模型的最终答案。为了衡量这一能力，我们提出了一个基准，用于测试多种类型的自我建模问题。当前模型表现出非平凡但有限的自我建模能力，并且在有关自身行为的简单反事实问题上会出现系统性错误。为了提升自我建模能力，我们开发了一个可扩展的合成数据流水线来生成自我建模训练数据，并展示了 reinforcement-learning 可以在三个开源模型家族上提升整体自我建模能力，并对留出任务带来一定迁移。然而，这些提升似乎并不总是意味着内省：改进后的自我建模未必源于对模型内部决策过程的特权访问。
大型推理模型（LRMs）的最新进展表明，在数学和代码等结果可以被自动检查的领域，利用可验证奖励的强化学习（RLVR）能够显著提升推理能力。将这一进展扩展到开放式和智能体任务仍然很困难，因为可靠奖励更难获得，而且直接的人类监督无法跟上模型生成经验的规模与复杂性。本文研究了当人类监督逐渐退出学习循环时，LRMs 如何继续改进。我们从两个相互关联的维度考察这一问题。奖励维度追踪从逐实例的人类判断到可复用的验证器和奖励机制的发展，即使在没有人类反馈的情况下也能运行。经验维度考察学习如何从人工筛选的任务和环境，走向自生成课程、构造环境以及自主协同进化。我们通过一条从 L0 到 L4 的五级阶梯将这两个维度联系起来，用以识别学习过程中哪些部分仍持续受人类控制。我们的分析进一步强调了随着奖励和经验生成日益自主化而引入的风险，包括奖励作弊、反馈漂移、课程崩塌和环境错误。因此，我们也围绕三个互补对象进行评估：策略能力、反馈保真度和经验质量。这一分析为当前将 LRMs 扩展到超越人类监督的各种方法，以及开发迈向超级智能的自维持学习系统所面临的开放问题，提供了一个结构化的说明。此外，我们持续更新一个 \href{https://github.com/visitworld123/Awesome-Scaling-LRM-Beyond-Human-Supervision}{GitHub 仓库}以跟踪最新进展。
自主科学研究代理正越来越多地应用于端到端的科学工作流，包括文献综述、数据分析、实验和报告生成。然而，开放式研究任务往往并未明确指定完成任务所需的分析、方法和成功标准。因此，代理可能会遗漏重要分析、使用不恰当的方法，或得出缺乏充分证据支持的结论。为了解决这一问题，我们提出 AutoSciRub，这是一个以评估为先的框架，它在研究执行之前先归纳出一个任务特定、可执行的评分标准，并利用该标准来指导执行、进行准则级验证以及迭代修订。AutoSciRub 将一个规定不明确的指令分解为原子级科学目标，将其锚定到相关文献和任务可见数据中，并综合生成具体、可操作且可验证的准则。由此得到的评分标准将隐含的实验与证据要求显式化，为实验和分析提供指导。在修订过程中，基于评分标准的验证会识别未满足的准则，并支持对研究报告及其支撑工件进行有针对性的改进。在 ResearchClawBench 上，AutoSciRub 在所有测试配置中都持续带来提升：在固定的 Codex harness 下，三个基础 LLM 的平均提升为 2.08 分；在使用固定 DeepSeek-V4-Flash backbone 的三个 agent harness 上，平均提升为 2.95 分。在 AstaBench E2E Discovery 随机抽样的 20 个任务子集上，AutoSciRub 在三个 agent harness 上进一步实现了平均 16.8 分的提升，同时保持或增加了成功完成的任务数量。这些结果表明，以评估为先的引导为自主科学研究提供了一种有效且可泛化的控制机制（Code: https://github.com/zjunlp/AutoSciRub）。
Accurate daily predictions of cold hardiness in woody plants are critical in regions where freezing temperatures can damage dormant buds and reduce seasonal yield. Existing biophysical, hybrid, and deep learning models have shown high predictive accuracy when trained on local data but remain largely site-specific. The limited availability of cold hardiness data, coupled with the lack of principled methods for transferring cold hardiness predictions to new regions and cultivars, has limited the broader adoption and practical utility of these approaches, particularly in data-scarce regions. To address these limitations, we propose a cold hardiness prediction framework that learns a transferable latent representation by capturing region-specific variation through learned embeddings. To enable prediction in previously unseen regions, we infer embeddings from (1) text descriptions of the cultivar and growing region, and (2) limited historical observations, supporting both zero-shot and few-shot transfer. Experiments on datasets from six regions across North America demonstrate that our approach consistently outperforms state-of-the-art cold hardiness prediction methods, yielding more accurate predictions and substantially improving transfer to data-scarce regions.
大型语言模型（LLMs）由词元表示、权重、适配更新、缓存和激活值等结构化高维对象构成，而传统以矩阵为中心的视角并未充分利用其多线性结构。张量分解和张量网络为这种结构提供了一种有原则的代数语言，但现有文献常将它们视为彼此孤立的压缩机制。本综述通过两个互补视角组织 LLMs 的张量方法：一个覆盖词元化、嵌入、预训练、适配、压缩、推理和可解释性的七阶段生命周期分类，另一个覆盖嵌入、注意力和前馈网络的组件视角。我们提供统一的记号和理论基础，分析单个 Transformer 组件的张量化策略，并比较各生命周期阶段的方法，同时明确评估协议和模型规模方面的差异。我们进一步将张量方法与相邻的高效技术以及概率张量网络联系起来。最后，我们综合开放挑战并引入 $ρ_{\rm gap}$，这是一种用于衡量理论内存降低与实测系统级加速之间压缩-落地差距的指标。通过将张量化视为一种通用的结构原则，本综述为张量化语言模型提供了一个结构化的入口，并阐明了参数节省在何种情况下有望转化为内存效率、计算效率或可解释性。本文专用的 GitHub 页面可通过 \href{https://github.com/ma-tt-a/awesome-tensor-methods-for-llms}{this https URL} 访问。
Large-scale industrial robot fleets share constrained physical infrastructure, making vehicle travel times dependent on safety separation, intersection access, downstream blocking, and station contention. We study this problem in overhead hoist transport (OHT) systems, a representative ceiling-mounted material-handling system used in semiconductor fabs. Static shortest-path routing cannot account for these time-varying traffic costs, whereas tabular Q-routing adapts online but learns each destination--node--action value independently, limiting information sharing across sparsely visited routing contexts and making startup behavior sensitive to inaccurate value estimates. We propose Neural Double Q-routing, which replaces destination-indexed tables with a shared state--action value network. The network is warm-started through return-to-go regression on mixed simulator-generated routing trajectories and then refined online using Double-Q updates, local congestion correction, and event-stratified structured replay. Across nine matched fleet-size--arrival-rate settings with 100, 150, and 200 OHTs, the proposed framework reduces mean completion time relative to tabular Double Q-routing by $0.8\%$--$8.8\%$. It achieves the lowest mean completion time among all compared methods in the six 150- and 200-OHT settings, whereas Dijkstra remains best in the three 100-OHT settings. Completed-task counts remain within $1\%$ of tabular Double Q-routing in eight of nine settings, and 95th-percentile completion time decreases in eight settings. In two matched startup scenarios, offline initialization increases the number of completed tasks by up to $23\%$ and reduces tail completion time by up to $15\%$.
Large Language Models (LLMs) offer new possibilities for scaling qualitative analysis, but existing applications often provide limited methodological transparency regarding how qualitative methods are translated into computational procedures. This paper presents an auditable and privacy-preserving computational operationalization of inductive and latent Thematic Analysis (TA). This paper first derives five design principles from the methodological requirements of TA and the conditions introduced by LLM-based inference: preserving interpretative context, maintaining traceable relationships between empirical material and analytical outputs, representing analytical constructs and reasoning explicitly, constraining LLM inference to interpretative tasks, and enabling privacy-preserving local deployment. Second, it presents a proof-of-concept for a two-phase workflow that operationalizes these principles by combining interpretative LLM inference with deterministic procedural control to generate codes, analytical justifications, themes, and theme descriptions while preserving explicit links to the source material. Third, it proposes an evaluation framework combining structural comparison with human-led TA and independent expert assessment of analytical quality. The evaluation is conducted on semi-structured Danish interview transcripts. and the results shows that the workflow produces code-level outputs with coverage broadly comparable to human annotations and highly rated analytical justifications, while generating a more compressed thematic structure characterized by fewer and broader themes. The findings demonstrate the feasibility of auditable LLM-supported TA through a modular workflow designed to scale to larger datasets, accommodate different LLMs, and support transfer across research domains, with domain adaptation primarily requiring adjustments to the prompting strategy.
生成式检索（GR）通过将查询直接映射到具有强表示能力的候选项 Semantic IDs（SIDs），已成为一种很有前景的范式。然而，现有仅基于商品内容生成的 SIDs 会产生语义鸿沟，无法将动态的查询意图与静态的商品表示对齐。此外，当前的生成式范式很少对用户行为序列建模，并且始终受限于 beam-search 自回归解码带来的高推理延迟。为解决这些挑战，我们提出 $\textbf{C}$ross-component $\textbf{H}$ierarchical semantic $\textbf{A}$lignment for $\textbf{P}$ersonalized generative retrieval（$\textbf{CHAP}$），一种从层次化视角出发的新型个性化 GR 框架。首先，我们设计了一个 Hierarchical Semantic Alignment 模块，用于对齐查询的潜在空间与商品的量化路径，并同步多粒度语义。其次，我们构建了一个个性化 GR 框架，通过协同离散 SIDs 进行结构性引导与连续表示进行细粒度语义细化来建模用户行为。值得注意的是，我们引入了 Residual Cascading Generation 机制，将代价高昂的多步 Transformer Decoder 限制为单次前向推理，在提升推理吞吐量的同时缓解信息损失。在三个公开数据集、一个私有工业数据集以及线上 A/B 测试上的大量实验表明，CHAP 优于现有方法，验证了我们方法的有效性和实用价值。代码已公开，地址为 https://github.com/zzzgm/CHAP。
在知识图谱上进行路径搜索，已成为为 LLM 推理中的多跳问题提供依据的一种有效方式。然而，生物医学问答带来了两个通用领域方法并未针对设计的独特挑战：(i) 查询不显式给出中间推理过程，且可通过多条有效路径得到答案；(ii) 生物医学知识图谱连接非常稠密，因此路径搜索方法很容易走错方向。为应对这些挑战，我们提出 AdaPath，一种从 Path-Bank 中检索查询自适应元路径的路径搜索框架，Path-Bank 同时捕获查询语义和生物医学知识图谱结构。AdaPath 为生物医学查询提供缺失的线索，并在多跳推理过程中有效剪枝稠密的知识图谱邻域。我们进一步发布了 BioStrat-QA，一个生物医学 KGQA 基准，用于按多跳查询暴露的中间推理程度对其进行分层。在各个生物医学 KGQA 基准上，AdaPath 始终优于基线方法，即使在多跳查询暴露较少表面信息时，仍能保持有效的路径搜索。源代码可在 https://github.com/Jun-Hyeong-Kim/AdaPath 获取。
Mixed-precision quantization (MPQ) assigns a different bitwidth to each linear layer of a large language model (LLM) to minimize the quantization-induced quality loss under a fixed budget, but Mixture-of-Experts (MoE) models contain these layers in every expert of every MoE block, so the allocation space grows far larger than in a dense model. Existing methods either allocate within each block under a uniform per-block budget, or allocate across blocks through an additive proxy, and neither directly optimizes a model-level objective over the choices that couple the blocks. We propose Q-Strata, a bi-level allocator that ranks within-block assignments with a cheap proxy and allocates across blocks with a model-level objective evaluated on the assembled quantized model. Its inner stage caches a Pareto frontier of candidates per block over finely spaced budgets, leaving the outer stage to set one budget per block instead of a bitwidth for every linear layer. With the search reduced to one budget per block, the outer stage optimizes this model-level objective directly, capturing the inter-block coupling that additive proxies miss. On Mixtral-8x7B-Instruct, Qwen1.5-MoE-A2.7B, and DeepSeek-V2-Lite, Q-Strata consistently achieves lower WikiText2 perplexity than uniform-bitwidth GPTQ and the state-of-the-art MoE MPQ methods MxMoE and GEMQ in the low-bit regime. The code is available at https://github.com/snu-mllab/Q-Strata/tree/main.
大语言模型（LLM）被用作序列决策策略的事后解释器，用自然语言解释某个动作为何被选中。然而，LLM 常常生成看似合理但错误的陈述，而且现有方法尚不能系统地测试这类解释是否忠实于底层环境。这里有两个经典的软件测试难题：一是没有用于判断解释正确性的判定标准，二是测试输入——关于策略行为的自然语言查询——缺乏进行系统化测试用例生成所需的结构。我们同时解决这两个问题。概率模型检查提供了测试判定标准，可计算精确的参考结果，并据此自动评估 LLM 的回答。一个事后查询类别的分类体系围绕由环境层面事实构成的策略解释来组织输入空间；由此生成的测试用例根据特定问题的诊断难度分数进行优先排序。在七个 MDP 环境中，该测试将三种开源权重 LLM 区分开来：一种推理模型通过了 85% 的测试用例，一种中等规模模型通过了 70%，而一个 1B 模型的表现甚至低于随机基线；与此同时，优先排序显著筛选出了比随机选择更难的案例。我们的结果表明，在无模型设置下，LLM 生成的解释有多可信——在这种设置中使用的是同样的 LLM，但不存在可用于验证它们的判定标准。
Large language models are increasingly capable in general, but their utility can remain modest in niche or understudied areas. One approach to address this limitation is to specialise existing models through additional training on target-domain corpora. In this work, we investigate such continued pre-training for adapting large language models to Swedish journalism, using a high-quality dataset that we curate from millions of news articles. To evaluate the adaptation efficacy, we also construct a novel domain-specific benchmark that covers six editorial tasks. Through full and parameter-efficient fine-tuning across two model sizes, we find that continued pre-training yields benefits in the target domain, but only when paired with experience replay to mitigate forgetting. We observe consistent enhancements in the models' generation quality and factual knowledge, but not their proficiency in discriminative tasks. Exploring a training-free method to facilitate instruction following, we see further improvements, but exclusively for models trained with low-rank adaptation. Crucially, we demonstrate the importance of targeted evaluation in the adaptation process, as an existing Swedish benchmark largely fails to capture the models' in-domain performance gains.
A benchmark score is a joint property of the model, the evaluation harness, the elicitation budget, the sampled population, and contamination status. Leaderboards publish the model and the score, so capability and leakage stay observationally equivalent. Existing taxonomies classify contamination for automated detection, not the question a reporter faces at publication: given the mitigations already applied, which validity threats remain open? We introduce a taxonomy organized by the mitigation each type defeats -- direct, derivative, temporal, distributional, and acquired -- spanning training-time and evaluation-time leakage. Holding out a private test set closes the first alone. The fifth is acquired during the evaluation itself; because it is a property of one run, it must be recorded with the reported score rather than with the benchmark release. We operationalize it as a four-field disclosure protocol in which "unknown" is a valid entry, released under CC BY 4.0 with a JSON Schema, a validator, and worked examples. Two coders external to the design team applied a pre-registered instrument to 41 documents. Per-variable linear-weighted $κ$ runs from 0.00 to 0.35 (median 0.21) over 29 main-pass documents against a single-coder test-retest ceiling of 0.84, collapsing under the class skew the registration anticipated; pooling raises it to 0.46 through chance correction rather than better agreement. Two variables fall below the prevalence-robust threshold registered in advance: strata reporting and the acquired type introduced here. Disagreement concentrates on when a variable applies rather than on what a document states. Elicitation budgets are reported in 13% of documents, and no document addresses all five types. The contribution is the taxonomy, the score-side artifact that follows from it, and a pre-registered measurement of instrument reliability and current disclosure.
知识蒸馏训练一个较小的 student model 去复现大型 teacher model 的输出，其进展通常通过 teacher--student 差异来监测。然而，真正重要的量是 student 相对于真实任务的误差。我们在一个最小的三方模型中研究这两个目标之间的关系：一个真实 teacher（生成模型）、一个 teacher 和一个 student，三者都是 soft committee machines，其中真实 teacher 包含一个 teacher 无法表示的共享潜在因子，失配强度由单个标量 $\dmiss$ 控制。在基于序参量的在线蒸馏描述中，并利用 error-function 激活下所有误差的闭式（反正弦型）表达式，我们证明学习动力学和蒸馏误差 $\Ets$ 对 $\dmiss$ 完全不变，而真实误差 $\Etzs$ 以及差距 $Δ=\Etzs-\Ets$ 随 $\dmiss$ 严格增大，其增长率还会被真实 teacher 的复杂度 $M_0$ 线性放大。对由真实 teacher 复杂度和 student 容量张成的平面上的数值相图验证了这一预测变形：$\Ets$ 的等高线不发生移动，而 $\Etzs$ 的景观系统性上升，并且随着 $\dmiss$ 增大，一个 teacher-miss 区域——即模仿成功但任务失败——不断扩展。结果为仅通过 teacher 模仿指标评估蒸馏提供了定量警示，并将差距 $Δ$ 确定为区分 teacher-miss 与容量受限失败的最小诊断指标。
Document question-answering systems increasingly answer questions over collections of retrieved documents rather than one clean source, so robustness to distracting context matters as much as reading ability. When such systems fail, it is often unclear whether the context was too long or the distractors were too close to the topic, because prior work tends to conflate these two effects. We present MUDDLE, a controlled benchmark that separates them. MUDDLE uses 270 human-annotated questions, each tied to a single source document, and instantiates every question in five conditions: the source alone, the source with two or four topically similar hard negatives, and the source with two or four random distractors. The random distractors are matched to the hard negatives in length and provenance, so an accuracy gap between the two arms reflects topical similarity rather than length. All five conditions are rendered in markdown, page images, and raw PDF, but the distractor sweep reported here is run in markdown, since a source plus its distractors exceeds current image and PDF input limits. We score answers with an LLM judge across three model families. In the complete markdown sweep, hard negatives lower accuracy more than length-matched random documents at both context sizes for gpt-5-mini, while random documents stay near the no-distractor baseline. The effect is small but directionally consistent, and for gpt-5-mini hard negatives significantly underperform length-matched random distractors when pooled across context sizes. We release the data and evaluation code for a reproducible study of context degradation.
现实世界中的多智能体系统（例如无人机蜂群、自动驾驶汽车、仓库机器人）必须在避免碰撞的同时满足丰富的时序任务。信号时序逻辑（STL）能够优雅地对这类目标进行编码，但当前的 STL 规划方法面临关键限制。最先进的基于优化的方法可以处理任意 STL 规范，但难以扩展，随着智能体数量增加，其计算上变得不切实际。基于学习的方法能够以快速的规划时间高效处理大量智能体，但当部署时的目标与训练时不同，它们表现不佳，并且不支持需要为不同智能体分配不同规范的规划任务（即异质性），也不支持需要多个智能体协同的团队级规范。这种可泛化性与可扩展性之间的根本权衡，给在实践中实现多智能体 STL 规划算法带来了挑战。为克服这一挑战，我们提出了一种用于带有 STL 规范的多智能体规划的新扩散方法。通过使用 STL 的可微近似，我们将 STL 梯度集成到去噪过程中，使我们的方法能够泛化到新的公式，即其谓词位于训练期间覆盖的目标区域内任意位置，同时达到与现有基于学习的方法相同的可扩展性。我们的方法支持异质规范，并且借助扩散模型，自然增强了规划多样性，从而显著减少了智能体之间的安全相关违规（例如碰撞）。一项详细的评估研究证明了 STL 引导的基于扩散的多智能体规划器在构建可泛化、可扩展且多样化的计划方面的实用性。视频和代码可在 https://www.jeappen.com/diff-ma-stl/ 和 https://github.com/jeappen/diff-ma-stl 获取。
机器遗忘使深度神经网络能够根据隐私和法规要求，有选择地移除特定数据的影响。尽管以往工作主要研究一次性遗忘，但真实世界系统必须支持持续遗忘，即多个遗忘请求随时间依次发生。在这项工作中，我们识别出该设置中的一个根本性局限：可塑性坍塌，即模型有效遗忘能力逐步崩解。通过对持续遗忘动态的理论分析，我们表明，持续遗忘操作会在参数空间中累积几何约束，导致饱和子空间，从而限制未来更新。这种结构性效应会引发两种不同的失效模式：（1）前向失效——后续任务的遗忘质量逐渐下降；（2）后向失效——先前已遗忘信息被自发重新记忆。我们在图像分类中的多种架构、数据集和方法上进行了大量实验，结果证实可塑性坍塌并非特定实现的偶然产物，而是持续遗忘中普遍存在的现象。我们的发现揭示了机器遗忘系统长期可靠性的一个关键障碍，并促使人们开发能够保持可塑性的遗忘算法。我们的代码见 https://github.com/TIML-Group/Continual-Machine-Unlearning-Plasticity-Collapse
Recent advances in large language models (LLMs) have established conversational text-to-SQL as a practical interface between users and databases, often involving multiple turns of clarification and revision. However, existing benchmarks primarily evaluate execution accuracy, leaving the unfolding and shifting of user intent across turns largely uncovered. To address this, we introduce TIDE-Bench, a benchmark for conversational text-to-SQL under chain ambiguity and intent drift evaluation, targeting two recurring patterns: chain ambiguity, where an underspecified question triggers layered clarification with conditional dependencies, and intent drift, where the user retracts and replaces a previously committed request element. Built on 514 anchor SQLs from BIRD, TIDE-Bench comprises 1,542 samples and introduces dedicated metrics for chain identification and drift recognition-resolution beyond execution accuracy. Evaluating 12 advanced LLMs reveals a persistent chain identification bottleneck unaffected by clarification frequency, a wide drift recognition-resolution gap, and overlap between failure modes when jointly activated. The corresponding code of TIDE-Bench is released for further research.
School coaches prepare for opponents with game film and intuition. The analytics tools of professional teams stay out of reach. We ask how far public data can close this gap. Professional basketball is our case study, chosen for its data rather than the league. We fuse five public sources into one per-shot dataset of 4.23M shots over 21 seasons. The sources are shot locations, two play-by-play feeds, official matchup tracking, and player biometrics. Alignment across them is 99.5% to 100%. We also report two data pitfalls that are easy to miss. We then model a half-court possession as a sequential game. Shot values come from ShotNet, an embedding multilayer perceptron (MLP). On a held-out season it beats a zone-rate baseline and a logistic baseline, and its probabilities are well calibrated. A depth-limited expectimax search then solves the offensive decision tree, with branch-and-bound pruning to keep it real time. All training runs offline, so the online system stays light. A scouting planner and a playable simulator both run in a single browser page.
在带文本属性的图（TAGs）上进行推理，需要大型语言模型（LLMs）将节点文本与分布在其邻域中的证据结合起来。现有方法在生成前固定可访问邻居的集合，迫使推理在静态上下文上进行，并使模型无法在推理过程中获取缺失的证据。我们认为，邻居选择本身也应成为推理过程的一部分。为此，我们提出 Call Neighbours Yourself（CNY），这是一个框架，使 LLMs 能够通过拓扑约束的图游走动作主动探索图邻域。与其在预先选定的邻居集合上进行推理，CNY 不如提供轻量级的邻居预览，并学习何时扩展候选邻居以获取额外证据。为应对邻居探索中的延迟信用分配挑战，我们引入 destination-conditioned on-policy self-distillation，它会在所选邻居内容揭示后回溯评估该邻居，并将由此产生的动作偏好变化转化为动作级训练信号。在统一的原始文本设置下，于标准 TAG 推理基准上的实验表明，CNY 始终优于固定上下文的后训练基线。此外，所学到的探索策略能够迁移到未见过的图，以及训练期间未接触过的图级任务。代码见 https://github.com/superallen13/CNY。
文本到图像（T2I）安全护栏未能公平地泛化到非标准方言。我们在五种英语方言上评估了 23,080 对配对提示词，并将这一失败形式化为方言惩罚：过滤器是依据语言表层特征而非语义意图触发的。文本级过滤器在两个相反方向上都失效：NSFW-T 会对无害的方言提示词过度标记，而 LatentGuard 会对有毒提示词过度标记（偏差差距最高达 +28.29 个百分点），与此同时 OpenAI Moderation API 则对它们漏检。受控的拼写错误消融实验确认，这种惩罚源于对方言特征的标记，而非一般性的分布外敏感性。像素级生成器在很大程度上不受方言影响；惩罚是在文本处理阶段引入的，并不均匀地级联到事后护栏。我们表明，这种偏差与训练数据不平衡相关，并且可通过组平衡再训练加以缓解；消融实验将这一收益归因于平衡暴露，而非 GroupDRO（group distributionally robust optimization）的最差组目标。当前管线系统性地对方言使用者失效，这是一种被平均准确率基准所掩盖的公平性失败。我们的官方代码和数据集已公开，见 https://github.com/minguinho26/dialect-penalty-t2i。 内容警告：本文包含冒犯性、有毒或令人不安的文本提示和生成图像。
Machine learning is usually evaluated after an evidence interface has been fixed. A dataset, sensor suite, query language, action set, or experimental protocol determines which observations can be obtained, and learning is judged by what it extracts from them. We study a complementary capability. A learner's state can determine which evidence-generating experiments it can reliably realise under bounded resources, even when primitive affordances remain fixed. We call this learner-relative experiment family its effective epistemic reach, and use Wide Learning for task-relevant learning-induced changes in that family.We formalise effective reach relative to learner state, deployment budget, reliability threshold, and evaluation distribution. In a controlled construction, two hidden worlds have exactly the same public observation law. An informative diagnostic exists in a fixed five-primitive substrate. Before calibration, one address attempt realises it with probability at most $2^{-10} = 1/1024$, below a pre-specified 0.95 threshold; after calibration, held-out realisation is 1. Public-channel total variation is 0, whereas the realised diagnostic has total variation 1, and sealed binary risk moves from approximately 1/2 to 0. The construction establishes that learning can change effective epistemic reach even when primitive affordances and deployment resources are held fixed. It opens a complementary evaluation question for learning systems: not only what they infer from available evidence, but what informative evidence experience teaches them to bring within reach.
Retrieval-Augmented Generation (RAG) 提高了 large language models (LLMs) 的事实性，但现有 RAG 系统往往难以应对需要自适应检索和持续修正中间上下文的复杂多步推理。近年来基于强化学习（RL）的 agentic RAG 方法在一定程度上缓解了这一问题，但通常依赖粗粒度动作空间和轨迹级奖励，导致奖励分配较弱，并偏向于短视、刻板的推理模板。为此，我们提出 AgenticRag-R1，这是一个通过记忆栈和细粒度动作空间深度整合推理、检索和记忆的 RL 框架，并结合层次化的动作感知奖励和信息感知的轨迹拒绝策略，以实现有效的长程学习。我们在一系列多跳、开放域和 agentic 推理基准上进行了实验，覆盖多个基础模型规模，结果表明 AgenticRag-R1 始终优于强基线。此外，AgenticRag-R1 学到了更稳健、可解释且更具记忆感知的推理行为，凸显了细粒度动作建模和信息感知优化对长程推理的效果。我们的代码匿名可见于 https://github.com/jiangxinke/Harness-RL/tree/AgenticRAG-R1-Whitebox。
Large language models (LLMs) have shown promise for spatio-temporal forecasting, but existing approaches often rely on regularly sampled token sequences and struggle with irregular observations because of temporal asynchrony, representation-space misalignment, and limited context windows. We propose LLMODE, a token-efficient framework for irregular spatio-temporal forecasting with a frozen LLM backbone. LLMODE first uses a graph-aware ODE encoder to reconstruct irregular graph observations as a continuous-time latent trajectory. A Fixed-Budget Perceiver Resampler then compresses this variable-length trajectory into a fixed number of dynamic memory tokens. In parallel, compact statistical descriptors are encoded and resampled into context memory tokens. A dual-source gated cross-attention module injects both memories into the frozen LLM, enabling controlled utilization of external spatio-temporal evidence. Experiments on three real-world urban datasets and two physical-dynamics benchmarks show competitive overall performance, with clearer advantages under sparse or dynamically complex irregular sampling. Additional evaluations on unseen urban regions further demonstrate strong zero-shot generalization without adaptation.
Large language model (LLM)-based agents have shown strong potential in solving complex tasks through multi-step reasoning, yet they remain vulnerable to execution failures. Accurate failure attribution is therefore critical for improving agent reliability. Existing topology- and spectrum-based methods exploit trajectory structures but often overlook fine-grained semantics, while LLM-based attribution methods capture semantic cues but suffer from long-context degradation over lengthy trajectories. To address these challenges, we propose DUOTRACE, a plug-and-play detection filter for LLM-based failure attribution. DUOTRACE follows a detect-before-attribute paradigm: it first detects anomalous executions and then supplies focused trajectory evidence to downstream LLM-based attribution methods. For effective VAE-based anomaly detection on agent trajectories, DUOTRACE integrates dual-view semantic-structural node representations, a Tree-LSTM-based trajectory encoder, and prefix-chain- and LLM-based data augmentation to handle heterogeneous nodes, hierarchical execution structures, and limited failure data. Experiments with six LLM-based attribution baselines show that DUOTRACE improves agent-level and step-level attribution accuracy by 8.7% and 7.0%, respectively.
https://cybercity.carino.systems/ https://github.com/MiguelCarino/CyberCity ASCII City: Traffic and Detail Update - https://www.youtube.com/watch?v=DSRooHo_HSI ASCII City Update: Interiors, Elevation and Skyscrapers - https://www.youtube.com/watch?v=UCKEDWowc0o Comments URL: https://news.ycombinator.com/item?id=49512975 Points: 6 # Comments: 2
Hi HN, we’re Brandon and Kingston, the founders of Hebbian Robotics. We built HFlow ( https://github.com/Hebbian-Robotics/hflow ), an SDK that turns multimodal recordings from robots and human operators into standardized, quality-checked episodes and queryable dataset manifests. A recording can contain synchronized video, joint states, actions, timestamps, and metadata, and HFlow processes those streams together. Here’s a demo of HFlow in action: https://www.youtube.com/watch?v=xni0GwV-xAw Robotics data pipelines often begin as scripts: one transcodes video, another checks timestamps, another adds labels, and another copies selected recordings into a training set. This works until the corpus grows. Then it becomes difficult to know which code ran, why an episode was excluded, or whether a dataset can be reproduced. The first pain is usually quality control because frozen cameras, missing topics, timestamp drift, and duplicate recordings can quietly enter training data. Brandon first encountered this while training embodied AI models for two-arm industrial cleaning robots. Kingston had run into related problems while building high-throughput infrastructure at Jane Street. Later, while speaking with robotics data providers, we kept seeing teams rebuild similar processing and quality-control infrastructure. We learnt that processing robotics data is itself one of the bottlenecks to improving robotics models. An HFlow pipeline consists of transformations, checks, labels, and enrichments. The SDK exposes them as plain Python functions that receive an episode and return measurements, artifacts, or transformed data. During development, the functions can run in-process. For scheduled corpus processing, HFlow packages the same registered steps as Airflow 3 DAGs, where teams can inspect task status, logs, retries, and reruns. HFlow currently accepts one MCAP file per episode. MCAP ( https://mcap.dev/ ) is an open container format by Foxglove for timestamped multimodal recordings, similar in purpose to a ROS bag. It lets video, robot state, actions, and other sensor streams remain synchronized in one file. We use it because HFlow needs to process these streams together, and because the resulting recordings remain compatible with Foxglove and Rerun. HFlow writes a canonical MCAP with in-band H.264 video, grouped camera and state chunks, and provenance describing how the output was produced. Each step has an explicit behavior version, and catalog records connect its measurements and artifacts to the source episode and pipeline run. Quality checks store reusable evidence rather than imposing one universal definition of good data. Some failures, including black frames, frozen video, missing topics, timestamp drift, and impossible joint movements, can be measured deterministically without training a model. Others might be detected using VLMs and other models like MediaPipe Hands. But their meaning depends on the task. A smooth trajectory might indicate a successful demonstration in one setting and a stalled robot in another. HFlow writes measurements, metadata, version stamps, and artifact locations to an append-only Parquet catalog. Teams query it with DuckDB SQL and produce a version-pinned manifest without opening the recordings again. Critical checks can quarantine an episode, but HFlow does not delete data. This separates the evidence from the policy used to assemble a particular dataset. We did not want to replace the tools robotics teams already use. HFlow connects MCAP for synchronized recordings, Airflow for scheduled execution, Parquet for catalog data, and DuckDB for curation. Compared with a general workflow orchestrator, it adds contracts for robotics episodes, processing provenance, quality evidence, quarantine, and dataset manifests. Compared with a training dataset format, it operates earlier and stops at curated episodes plus a manifest. Here are three examples of teams that would use HFlow: 1. A data vendor or marketplace collecting egocentric recordings. They could use HFlow to detect black or frozen video, duplicate recordings, hand-object interaction, and other quality metrics before delivering the data, while retaining evidence of which checks ran on every episode. 2. A robotics team collecting teleoperated demonstrations for its own models. They could use HFlow to standardize recordings, add labels and enrichments, and produce a reproducible training manifest. 3. A team operating robots in the field. It could process incoming logs, quarantine incomplete or corrupted episodes, and query the catalog for particular robot versions, environments, or failure conditions. The project is pre-v1, but the core lifecycle works end to end. You can try it without an account, Docker, or robot hardware by cloning the repository and following the quickstart. HFlow is free under the Apache-2.0 license. The open source deployment is currently a single-tenant workspace, and we have not built the hosted, multi-tenant control plane yet. We are considering making money through managed workspaces and enterprise support for teams that do not want to operate the runtime themselves. Because this processing layer is software and data, people can contribute without owning a robot. We would especially like feedback from people who have built pipelines for robotics, video, or other sensor-heavy systems. We are curious where our data model is wrong, which integrations are missing, and what would fail first on your workloads. Comments URL: https://news.ycombinator.com/item?id=49510632 Points: 3 # Comments: 0
大型语言模型（LLMs）的最新进展已展示出强大的自然语言理解和数学推理能力。然而，它们将非正式数学问题转化为形式化表示的能力仍缺乏充分研究。这一局限对于诸如 AlphaGeometry 之类的神经符号几何系统尤为重要，因为其定理证明引擎需要以一种专门的领域特定语言（DSL）作为输入。尽管 AlphaGeometry 已达到接近 IMO 金牌得主的表现，但将自然语言问题手动转换为其形式化语法仍然是一个显著的可用性瓶颈。为应对这一挑战，我们提出了 Natural Language to AlphaGeometry Benchmark（NL2AGBench），用于评估 LLMs 将英文几何问题翻译为与 AlphaGeometry 兼容的形式化表示的能力。NL2AGBench 采用 AlphaGeometry 内部基于执行的验证来评估翻译质量，而不仅仅依赖文本相似度。我们在多个参数规模上评测了十个最先进的开源和闭源 LLM，并分析了可执行翻译准确率、语法正确性和错误特征。实验结果揭示了闭源模型与开源模型之间显著的性能差距：领先的闭源模型可实现超过 80% 的可执行翻译率，而即使是最大的开源模型，也难以持续保留几何约束并生成有效的形式化表示。我们提出了一种区分语法错误和逻辑错误的错误分类体系，并研究了包括 few-shot 提示、微调和人工引导提示在内的缓解策略，这些策略在多个模型家族上都带来了可衡量的改进。
Software and systems security workflows are typically procedural: analysts inspect heterogeneous artifacts, form hypotheses, invoke tools, interpret outputs, and revise plans. Large language model (LLM)-based agents, which can plan, use tools, retain state, and revise actions across multi-step workflows, are being rapidly adopted to automate this work. Given the consequences of delegating security decisions to autonomous systems, understanding how such agents are built, used, and assessed is crucial. Yet to this date, there remains a lack of systematic understanding of what has been done and how far we are in this field: the term "agent" is applied inconsistently, applications differ sharply in risk, and assessment protocols are often incomparable. To gain a comprehensive and coherent view of this area hence inform relevant future research, this paper provides a systematic literature review of the (1) technical approaches, including agent architecture, perception, memory, reasoning and planning, action space, orchestration, and self-improvement, (2) applications, with respect to the security tasks served, and (3) assessment, including the datasets, outcome and trajectory metrics, safety measures, and baselines considered, over the peer-reviewed literature spanning the emergence of this area (2023--2026). Our synthesis reveals a field that has built agents able to act but not yet agents whose authority is bounded or whose behavior is auditable. In addition to knowledge systematization, we also extend our insights into the limitations of and challenges faced by current approach, application, and assessment designs, which shed light on potentially promising future research directions.
AI 编码代理是一类通过推理和工具使用自动化开发任务的软件工具，正日益通过插件市场进行扩展，但这些新兴仓库的结构、维护以及共进化动态仍缺乏实证研究。与通过源代码提供功能的传统软件包不同，agent 插件通过自然语言指令文件、脚本和配置文件的组合来提供功能，这引出了一个问题：这些插件究竟是会在各组件之间共同进化的可维护制品，还是开发者只需编写一次、无需再回看的一次性制品。为研究 agent 插件的维护与共进化，我们对 1,926 个托管 Claude Code 插件市场的仓库进行了实证研究，分析了 2,018 个市场中的 8,351 个插件和 77,773 次提交。我们发现，该市场正在快速扩张，在 2025 年 10 月发布后的六个月内，涉及插件的提交活动增长了 8.8 倍，而面向软件工程任务的插件占所有插件的 61.3%。插件开发主要由功能驱动，功能相关提交的发生率是传统开源软件（OSS）的两倍多（39.6% 对 17.2%）。Claude 参与共同编写了 34.9% 的全部提交，而且四类提交（docs、perf、style 和 refactor）在插件仓库中的含义与在传统软件中有显著不同。大多数组件类型独立演化，但在 skills 目录内部，自然语言指令文件与实现脚本以高于随机的比例共同进化，其中 78% 的共同变更在功能上是耦合的，这代表了一类在传统软件工程中未曾观察到的新型维护依赖。
Texture image classification plays a significant role in computer vision applications, including industrial inspection, medical image analysis, remote sensing, and object recognition. Handcrafted features can capture local texture characteristics but may have limited capability to represent complex visual patterns. In contrast, deep learning models automatically learn discriminative representations but may not fully exploit the multiscale spatial-frequency information inherent in texture images. This paper proposes a hybrid feature fusion framework, termed DWT_AlexNet_DNN, which combines Discrete Wavelet Transform (DWT) features with deep features extracted using AlexNet for texture image classification.
A code world model accepted by a sampling gate can be exactly right on everything the gate can see and arbitrarily wrong beyond it. We characterize what a certified model can know, and what its errors can cost, when the omission is an annular freeze mode enclosing an unreachable interior. The gate quotient makes the question precise: acceptance-with-certainty determines the model exactly on the reachable query set; beyond reach is gauge. On a minimal ring instrument we prove the extreme case (a wrong-topology filled-disc artifact unfalsifiable by any sampling gate and bitwise harmless at play) and measure, with LLM synthesis across three model families, how one knob (a channel of width gamma) walks the same artifact through three regimes: unfalsifiable-and-harmless, falsifiable-and-costly, and instantly falsified. Three principles organize the empirics. First, danger is topology relative to reach: a channel the planner can use collapses the blind model's exploitation (play cost 1.09 to ~0 over a knee at gamma ~ 0.1), while a hidden channel with the same first Betti number keeps it at full strength (1.12). Second, repair is parameter-bound and sensor-bound: no family recovers the region from outside evidence; from inside, models pose the right topology but cannot pin its parameters, and the posed topology tracks the guiding persistent-homology summary's wrong beta_1 (a sensor with a measured geometric resolution limit), not the truth. Third, mitigation must match the error's dimension and direction: point fences fail against the one-dimensional boundary, a dimension-matched persisted fence collapses exploitation to a two-lesson transient (0.999 to 0.058), and the dual freedom certificate collapses the invented-mode failure symmetrically (1.769 to 0.029). In n dimensions the shell makes misidentification near-certain while the danger stays fully exploitable: the two axes are independent.
文章链接: https://kuleshov-group.github.io/blog/blog/2026/how-to-build-a-diffusion-language-model/ 评论链接: https://news.ycombinator.com/item?id=49503956 得分: 7 # 评论数: 0
Industrial panel operation is knowledge-intensive and safety-critical. Beyond control recognition and action generation, execution must satisfy constraints in operation manuals and safety regulations. While foundation-model-based planners show strong semantic capability, they typically lack computable, localizable, and reproducible mechanisms for violation detection and repair. To address this, we propose PanelShield, a verifiable closed-loop safety planning framework for manual-guided industrial panel operation. The framework generates parameterized action primitive sequences from task-relevant manual evidence and applies dual formal verification with LTL and a Safety FSM to enforce cross-step temporal correctness and local transition legality. When violations occur, it outputs a structured counterexample with the earliest violating step and cause, enabling targeted repair and re-verification. We build a multi-level long-horizon planning benchmark covering three representative industrial device panels, and evaluate the framework in simulation and real-world robotic experiments. Results show that PanelShield improves complex safety-constrained task performance over foundation-model-only planning baselines while reducing the violation rate to 2.7%, with 4.1 s total latency. Real-world experiments demonstrate end-toend feasibility. Overall, PanelShield offers a verifiable approach to robotic panel operation that balances flexibility, safety, and auditability.
我们研究了在以英文为主的语料库上预训练稠密大语言模型时，学习率和批大小的扩展行为。除了对学习率和批大小的联合最优值进行扩展分析外，我们还研究了它们随模型容量和数据规模变化的边际演化，并构建了一个能够刻画这些关系的模型。由于我们采用了 Warmup-Stable-Decay 学习率调度，我们进一步考察了在广泛的超参数设置、模型和数据预算范围内，学习率退火所带来的收益，以及最优学习率和批大小是否会在稳定阶段与衰减阶段之间迁移。最后，我们刻画了损失对模型容量和数据集大小的依赖关系，评估了近期提出的、显式建模二者相互作用的扩展形式。我们发现，这些方法在我们的实验中对于同时捕捉欠训练和过训练两种情形尤为有效。这项研究为未来 OpenEuroLLM 模型的开发建立了首个基线和扩展流程。我们已将本研究使用的全部预训练运行结果开源。
Medical question answering (QA) systems have become crucial tools for providing reliable health information. But they remain very unexplored for low-resource languages like Bangla due to limited datasets and systems tailored to these languages. To address this, we introduce BanglaMed-QA, a robust QA system specifically designed for the Bangla medical domain. The process begins with building a structured medical knowledge base that includes 4,493 QA pairs in 9 categories under 506 diseases. To improve semantic comprehension, domain-specific root word dictionaries and synonym sets are proposed, in addition to part-of-speech tagging for anaphora resolution. We adopt supervised machine learning models in which SVM is found to be the best model to categorize questions. Multiple similarity metrics, including cosine, Jaccard, BM25, and Levenshtein, are applied with soft and hard voting methods for query matching. The performance of the QA system has been evaluated in two aspects, with a 95% F1 score in an automated evaluation and an average human satisfaction rating of 0.9 out of 1.0. This validates the real-world application of BanglaMed-QA in closing the healthcare information gap for Bangla speakers.
Encoding temporal order is a fundamental requirement for sequence representations in Hyperdimensional Computing. Fractional Power Encoding provides similarity-preserving position vectors whose inner products approximate shift-invariant kernels, and it supports shift-equivariant transformations of encoded sequence representations. However, standard formulations of Fractional Power Encoding are primarily designed for binding operations such as circular convolution or complex-valued multiplication, which limits their compatibility with Hadamard product binding of real-valued vectors. This paper develops real-valued position encodings motivated by Random Fourier Features, aiming to retain the desirable properties of Fractional Power Encoding while supporting Hadamard-based operations. We propose three real-valued position-encoding variants: a real-valued baseline based on the inverse Fourier transform, and Sinusoid and Cosine-only representations derived from Random Fourier Features. Among them, the Sinusoid variant provides an explicit algebraic shift operator, allowing temporal shifts to be applied directly to the vector-encoded sequence representation without re-encoding the shifted sequence. Experiments on time-series classification datasets show that the proposed real-valued representations achieve performance comparable to standard Fractional Power Encoding while enabling computationally efficient Hadamard product binding. The Sinusoid variant offers the most favorable trade-off, combining efficient real-valued implementation with exact shift-equivariant transformations.
Precise dense correspondence is a fundamental prerequisite for multimodal spectral imaging systems that fuse disparate wavelength ranges for subsequent analysis in medical and scientific imaging. Corresponding image points are often observed with non-overlapping spectral sensitivities, leading to wavelength-dependent contrast changes, intensity inversions, and appearance shifts for which dense ground truth is difficult to obtain and conventional RGB-based training data provides only limited supervision. We address this data gap by introducing a sensor-agnostic cross-spectral modulation protocol on established correspondence benchmarks with intensity input projection, and by proposing a synthetic cross-spectral correspondence benchmark simulating physically plausible radiometric differences. Evaluation on several modern dense correspondence backbones trained with our unified cross-spectral protocol showed substantial improvements under severe spectral mismatch while maintaining performance on standard RGB benchmarks. Ablation experiments show that view-dependent channel selection and nonlinear radiometric transformations provide complementary robustness, indicating that the primary limitation of existing models is not their structural matching capacity but the mismatch between training distribution and spectral characteristics of the target image pair. Qualitative evaluations on heterogeneous medical spectral acquisition systems demonstrate the practical relevance of the proposed training data augmentation protocol as an enabler for spatially coherent spectral fusion in HSI workflows.
Integrating Artificial Intelligence (AI), particularly generative AI, into medical training has prompted concerns about learner over-reliance, misuse, and erosion of foundational clinical competencies. We propose a conceptual reframing at the decision level: the problem is not misuse but misclassification - a mechanistic failure of real-time metacognitive evaluation in selecting a subzone-inappropriate AI interaction mode. Drawing on "SCAN" (Substitute, Complement, Aid, Non-Negotiable), a human-centric decision-making framework for generative AI task allocation grounded in Vygotsky's Zone of Proximal Development and metacognition, we advance the emerging social-constructivist conversation around AI in medical education by offering a testable account of AI's role in clinical reasoning development. This framework yields testable predictions for how misclassification can be detected, mitigated, and, more importantly, prevented in the clinical learning environment. Regarding clinical reasoning development, we show how trajectories of skill acquisition (upskilling) and failure (the triad of skill failure: de-skilling, never-skilling, and mis-skilling) operate at the individual task level in ways that fixed-phase, cohort-wide treatments fail to capture. We further identify passive engagement within correctly classified AI-scaffolded tasks as a particularly insidious, detection-resistant pathway to mis-skilling - one requiring subzone re-identification from AI assistance to expert assistance, with human experts serving as epistemic auditors. The paper operationalizes SCAN for clinical curriculum design, supervision, and assessment, and opens an empirical research agenda grounded in cognitive science. This paradigm shift from misuse to misclassification is not semantic: it offers educators a clear perspective on what to look for, what to assess, and what to intervene on.
In financial markets, a sequential policy that reacts systematically to price movements may become predictable to other market participants. This paper studies whether large language model (LLM) agents exhibit such directional structure through RetailAgent, an experimental framework in which an LLM observes anonymized intraday equity price histories and permitted state, then repeatedly chooses long (hold the stock) or flat (stay out) before the subsequent interval return is revealed. We compare returns during long and flat intervals along the same stock's intraday path after removing the overall fraction of long decisions. This exposure-matched measure reveals persistent negative timing across modality, horizon, state, and model family. Shuffling saved action sequences substantially attenuates the effect, showing that alignment between actions and subsequent returns drives the negative score. Feeding self-authored memories into decisions further increases policy persistence, while timing becomes more negative among stock-days on which the agent uses both actions. These results reveal stable, recoverable directional structure in sequential LLM financial decisions and a behavioral signal for studying how another participant could respond to a predictable policy.
对比式视觉-语言学习利用配对的胸部CT体数据和放射学报告，在无需人工标注标签的情况下学习异常分类器。然而，胸部CT的两个特性给传统的全局对比学习带来了挑战。首先，许多关键异常很小或局限于特定解剖区域，将整个体数据池化为单个嵌入可能会削弱其视觉证据。其次，标准对比目标会将批次中的其他每个扫描都视为负样本。由于许多胸部CT共享异常，这一目标会错误地将共同为正样本的配对拉开。我们提出用于3D胸部CT的 Anatomy-Routed Contrastive Learning（ARC-CT），这是一个区域感知框架，仅使用由LLM从报告中提取的标签来解决这些局限，无需人工标注或边界框。ARC-CT结合了三个组件：（1）Anato-myQFormer，通过受自动生成的器官掩码约束的查询来定位证据；（2）label-Jaccard soft InfoNCE目标，将标准 one-hot 目标与每对样本的标签集合重叠相结合，从而减少共享临床发现的研究之间的误负样本惩罚；（3）器官级对齐损失，将掩码池化后的视觉特征与使用大语言模型离线提取的器官特异性报告文本连接起来。ARC-CT 使用紧凑的 3D ResNet-18 主干，在18种异常上实现了0.86的无需掩码宏平均 AUC。总体而言，ARC-CT 的表现优于可比的高效基线以及若干更大的 transformer 模型。我们的代码和权重可在 https://github.com/arc-ct/arc-ct 获取。
文章链接： https://github.com/vllm-project/vllm/releases/tag/v0.28.0 评论链接： https://news.ycombinator.com/item?id=49492067 得分：8 # 评论数：0
文章 URL: https://www.tencent.com/tencent-releases-and-open-sources-tencent-hy4-preview/ 评论 URL: https://news.ycombinator.com/item?id=49492632 积分: 9 # 评论: 2
文章链接： https://github.com/openai/openai-python/blob/main/httpx2.md 评论链接： https://news.ycombinator.com/item?id=49477212 得分：9 评论数：6
我学习日语已经有几年了，一直遇到一个类似的问题。我会找到一个想用来学习的视频，听到一句有用的话，然后意识到把这句话变成以后可以学习的内容既费时又有时让人很疲惫。 最后我会在视频播放器、字幕/转录文本、词典、截图、音频片段和 Anki 之间来回切换。所以我做了 SubSmith，把这套流程整合起来。 你可以把视频或音频文件放进去，先在本地生成转录文本，然后结合媒体内容和转录文本来： * 查询单词和句子 * 重播单独的行 * 编辑转录文本 * 保存有用的句子及其原始上下文/音频 * 导出为 Anki 卡片 对我来说，最重要的是它可以处理你自己的媒体。它不绑定某个特定的流媒体服务或资料库，所以我可以拿来学习我真正感兴趣的随机动漫集、播客、讲座等内容。 这是一个优先离线的桌面应用，转录也是在本地完成的，而不是把媒体发送到转录 API。 我在这里分享它，是因为现在我更想弄清楚这种工作流对其他人会在哪些地方失效，而不是在已经有了稳定核心/基础之后继续随意加功能。 比如： * 你真的会从自己的媒体里保存句子吗？ * 这个流程的哪一部分让你觉得太费工夫？ * 附带音频/上下文会不会让制作 Anki 卡片更有用？ * 你会更希望它直接在你现有的视频播放器/浏览器里工作吗？ * 安装桌面应用是不是一个很大的门槛？ * 还有，在开始免费试用前要求注册账号，会不会让你直接放弃？ 当前版本确实要求先注册账号才能开始试用，我想弄清楚这对那些真正会使用它的人来说，是否会造成明显的阻力。 可以免费试用，如果你已经在通过自己的视频、动漫、电影、播客或其他媒体学习语言，我尤其欢迎你的反馈。 我是开发者，所以我会在评论区回答问题并讨论它的工作方式。 https://subsmith.app Comments URL: https://news.ycombinator.com/item?id=49476894 Points: 5 # Comments: 1
Effective human-robot collaboration in industrial settings requires robots to understand human intentions and assist with task planning, reducing workload. Recent works have explored the use of Multi-modal Large Language Models (MM-LLMs) for task planning in such data-scarce scenarios, leveraging in-context learning to interpret user actions and generate long-horizon action plans in natural language. However, MM-LLMs inherently lack an understanding of system states and do not track state transitions, often leading to hallucinated actions that deviate from the intended goal. Additionally, generating action plans in natural language tends to limit the generated plans to a high level, introducing ambiguity in action execution. To address these limitations, we propose the State-aware Task Estimator and Planner (STEP), which prompts a MM-LLM to explicitly estimate the state of the system and predict the state transitions resulting from executed actions. By forecasting future states alongside actions, STEP ensures task-convergent planning while also providing additional assistance parameters necessary for executing the predicted actions. We evaluate STEP in a simulated environment using a robot assembly task. Our approach outperforms the state-of-the-art by 32.8% in action executability and 14.8% in final-state error.
Agent harnesses shape how language-model agents use instructions, tools, and runtime components, but adapting these harnesses requires costly verification. Existing propose-and-verify methods typically score every candidate on a fixed task set, wasting rollouts on unrelated behaviors and allowing aggregate scores to obscure specific regressions. We introduce HarnessLens, a budget-aware framework for automated harness evolution. HarnessLens jointly explores the task space and user-configurable components, derives candidate modifications from execution trajectories, and selectively verifies each candidate on behavior-relevant tasks using an attributable-evidence gate. Across three agent harnesses and four benchmarks, HarnessLens improves average held-out performance by 7.6-13.6% while consuming substantially less evaluation budget than competing baselines. These results demonstrate that behavior-aware verification with explicit attribution enables more reliable and sample-efficient harness evolution under constrained interaction budgets. Our code is available at https://github.com/jhxu5214/HarnessLens.
Influential public discourse shapes public beliefs and can also mislead, not only through what is stated, but also through how information is framed, omitted, contextualised, and communicated. Yet less research has focused on how such misleadingness arises and shapes the interpretations formed by readers. To address this gap, we introduce Reader-Centric Misleadingness Understanding (RCMN), a framework that operationalises misleadingness through five dimensions: misleading mechanism, likely reader interpretation, evidence-warranted interpretation, emotional arousal, and communicative intent. Based on this framework, we construct an evidence-grounded dataset of influential public discourse. Empirical findings show that misleadingness is diverse and extends well beyond fabrication, with unsupported inference, exaggeration, and omission among the prevalent mechanisms, and is frequently associated with heightened emotional arousal and distortive communicative intent. Moreover, we investigate whether lightweight claim-and-context representations retain sufficient cues for understanding reader-centric misleadingness without access to richer contextual, evidential, and multimodal information. Evaluation across five recent generative foundation models shows that reader-level interpretations can often be recovered from such limited representations, whereas identifying how misleadingness is produced remains considerably more challenging. These findings highlight the potential of lightweight representations for scalable misleadingness analysis, while reliable understanding of misleading mechanisms continues to require richer contextual and evidential grounding.
Physics-integrated 3D Gaussian representations now allow reconstructed deformable objects to be simulated and rendered under explicit material models. Existing pipelines, however, assume that material parameters are known or manually specified, limiting their applicability when these parameters must be inferred from observed object dynamics. We propose KnockGS, an interaction-response PhysicalGS framework that estimates the elasticity and density scales of a 3D Gaussian object from its dynamics under a known applied force. Rather than treating physical simulation only as a forward process, we turn the force-induced response into a calibration signal: temporal response features are xtracted from the observed dynamics, the two material scales are estimated from those features, and the estimate is then frozen and written back into the same simulator so that it can be tested on an interaction it was never fitted to.We evaluate the framework on both parameter recovery and response-level fidelity. The estimated scales are compared against hidden ground truth, and the re-simulated object is measured against the target using 3D particle trajectories, response-curve statistics, and rendered-frame quality. Across five held-out material targets, our method recovers the scales substantially more accurately than response retrieval, global regression, or a fixed default material, and the frozen estimate remains predictive under interactions that differ in direction and in magnitude. Interaction response therefore carries enough information to calibrate material scales in physically grounded 3D Gaussian representations.Our study is a first step toward interactive PhysicalGS systems that calibrate a Gaussian asset whose rendered appearance and simulated response are consistent.
最先进的动作条件视频模型通常仅限于单一机器人具身形态，因而无法利用包含丰富学习可泛化物理信号的大规模异构视频数据。为弥补这一缺口，我们提出 CLAP，一个能够跨具身进行动作条件视频生成的框架，可在来自人类和机器人智能体的多样化、互联网规模视频上训练。CLAP 的核心洞见在于：无论执行主体是谁，普适的物理定律都支配着时空动态。然而，跨具身学习并非易事，因为不同机器人平台之间的动作表示差异很大，而人类视频中通常又缺少动作信息。CLAP 通过以下核心贡献解决这一根本挑战。首先，CLAP 使用末端执行器位姿、语言指令和潜在动作来协调不同的动作空间。其次，为了弥补各自的局限，CLAP 提出了一种基于课程学习的跨具身学习方案：先利用潜在动作在无标注视频数据上学习基础物理先验，再将其锚定到末端执行器动作空间中，以便零样本部署到真实世界任务。关键的是，CLAP 在 DROID 等具有挑战性的环境中已接近或超过最先进的单具身视频模型。这些性能优势还能通过少样本适配进一步累积，从而建立起一种训练单具身视频 world models 的新范式。最终，CLAP 提供了迄今为止最全面的动作条件视频 world models 套件——覆盖多种动作条件空间（末端执行器、语言和潜在）以及多种机器人形态（包括跨具身、DROID、Bridge、双臂 YAM 机器人和 G1 humanoids）。我们已开源全部代码和模型。项目网站见 https://omni-clap.github.io 。
Agent skills package specialized knowledge and workflows into reusable resources that extend AI agent capabilities. Recent work automatically discovers such skills from agent experience, which enables agents to progressively adapt through interaction. However, the insights that guide skill development typically remain scattered across optimization histories, limiting their systematic reuse across iterations. We introduce WikiSkill, a framework that co-evolves agent skills with a persistent knowledge base (wiki). At a high level, WikiSkill separates raw execution experience, accumulated knowledge, and executable skills, while continuously consolidating experience into the wiki, which subsequent skill updates can build on. Across diverse benchmarks and models, WikiSkill consistently outperforms state-of-the-art skill-evolution methods and improves over no-skill baselines in most model-benchmark settings. We find that skill evolution complements model scaling: larger models generally benefit more from evolved skills, while smaller models with skills can outperform substantially larger models without them. We also find that evolved skills transfer effectively across models and model families, and skills evolved by other models can outperform self-evolved skills. Finally, our ablation studies confirm that persistent knowledge accumulation in the wiki is critical for effective skill evolution. These results demonstrate the benefits of systematically accumulating and refining agent experience for developing reusable and transferable skills.
Privacy-Enhancing Technologies (PETs) in computer vision often rely on noise or image perturbations to protect visual data while securely processing it, creating a trade-off between task performance and protection. This trade-off is commonly evaluated using image classification, which primarily captures semantic separability and remains robust despite significant geometric, spatial layout or local boundary alterations. As a result, it is too simplistic as a proxy for generic vision tasks. Exhaustive downstream-task evaluation, however, is computationally expensive because models must often be trained for each PET transformation and parameter setting. We therefore propose a compute-aware multi-task protocol for evaluating PETs in model training. It combines lightweight proxy tasks that target complementary aspects of visual structure while remaining simple and fast to compute. Across irreversible privacy transformations, key-based block primitives, and learnable image encryption schemes, we demonstrate that PETs with similar classification accuracy can differ substantially on other tasks. The outcomes highlight the need for PET evaluation protocols that move beyond classification-only reporting.
Ontology learning from text remains challenging despite significant progress in Large Language Models (LLMs), which can hallucinate domain terms, produce inconsistent formats, and favor hierarchical over associative relations. In the LLMs4OL 2026 Challenge, we address both the End-to-End Flagship Task (Task A) and Ontology Extension Reuse Task (Task B) using an offline retrieval-augmented few-shot prompting pipeline. Our system employs Qwen2.5-14B-Instruct with all-MiniLM-L6-v2 for demonstration retrieval, selecting the top-5 examples for Task A and top-2 for Task B. A left-truncated context-windowing strategy preserves task instructions within long prompts. For Task B, generated triples undergo deterministic vocabulary-constrained filtering, retaining triples when at least one endpoint belongs to the sample's closed term/type vocabulary and removing duplicates of the initial ontology. The approach achieves Semantic Graph Similarity of 0.8692, Term-Typing F1 of 0.9200, and Taxonomy Discovery F1 of 0.8540 on Task B, while Task A achieves 0.7416 Semantic Graph Similarity. However, no non-taxonomic relations are extracted, highlighting limitations of closed, taxonomy-oriented relation vocabularies.
一个被展示专业外观市场面板的 LLM agent，对一个可被证明不可预测的问题给出方向性判断的频率，远高于只问该问题本身时：在 12 个前沿模型中，随着证据不断增强，做出判断的比例从 6.5% 升至 54.0%。即使面板上的每一个数字都是编造的，它也同样容易做出判断：把整个展示都伪造出来，使模型看到的除了问题本身之外没有任何真实内容，仍会把做出判断的比例从 24.5% 提升到 36.8%，这一结果在统计上与真实市场数据产生的 37.6% 没有显著差异。真正触发自信行动的不是信息，而是其包装所带来的权威感。这个失败很窄，而且可以定位。能力不足并不是原因：在与这些面板配套的、可回答的匹配问题上，同样的模型几乎总能作答，准确率接近完美。信念也不是原因——在会让行动幅度变化 48 个百分点的梯度上，模型给出的概率几乎不变，而且得分还低于气候学基线。缺乏判断力也不是原因：在行动前先让模型判断一个问题是否可知时，模型有 90% 的时候把它判为本质不可知，而随后只在其中 0.4% 的情况下做出判断。失败的是“行动/不行动”这道门槛，而且这种效应集中在少数模型上，而非普遍存在。由于这道门槛是可分离的，它可以被训练。对一个 3B 模型在 540 个合成案例上进行监督微调——这些案例主要是骰子、硬币、罐子和计时器——可将原始案例上的判断率降至 0.0%，并迁移到三个未见领域。它并非无往不利：只有当回答格式留有推理空间时，这道门槛才有效；而剥夺这种空间的僵硬格式，会让模型在原本能答对的问题上依然自信却答错。这道门槛既可训练，又对上下文脆弱，部署时这句话的两半都必须兼顾。
视觉-语言模型（VLM）展现出卓越的视觉推理能力，然而随着视觉 token 数量的增加，其推理成本也迅速攀升。现有的视觉 token 剪枝方法存在两个根本性局限。第一，大多数方法仅在 vision encoder 之后进行操作，未能优化视觉编码阶段的显著延迟。第二，在严格的 token 预算下，这些方法往往无法同时保留整体视觉上下文和细粒度细节，导致性能下降。为解决这些瓶颈，我们提出 PACE（Pixel-Adaptive Condense and Extract），一种无需训练的推理框架，通过统一的 Condense-and-Extract 范式同时加速 vision encoder 和 Large Language Model（LLM）。在 Condense 阶段，Adaptive Pixel Compressor（APC）在编码前评估视觉信息密度，自适应地下采样冗余输入，减少 encoder 计算量，同时保留全局上下文和关键视觉线索。在 Extract 阶段，Dynamic Dual-Attention Extractor（DDAE）通过融合来自 encoder 的内部视觉信号和来自 LLM 的语义信号，选择性保留视觉 token，从而保护任务关键细节。将 PACE 集成到 Qwen2.5-VL-7B 后，模型仅使用 10% 的视觉 token 即可保留原始性能的 93.8%，使首次 token 输出时间（TTFT）加速 3.1 倍。我们的代码见 https://github.com/jjL357/PACE。
Hey HN, I’m Abhishek. I'm building Opslane, an open-source agent that identifies user-facing issues and investigates them. It only creates a PR if it can verify the fix. Demo: https://youtu.be/ccuOTYQMeYg Docs: https://docs.opslane.com At my last job at Robinhood, we used to do a quarterly bug bash. We would go through our Sentry backlog and try to fix as many of them as possible. We only fixed bugs we knew were reported by customers. We had hundreds of bugs, and Sentry’s default priority levels made no sense. After the bug bash, we would declare bankruptcy - select all remaining bugs and mark them as resolved. This problem has only gotten worse since coding agents have become more prevalent. So I started thinking: what would Sentry look like if it were built in 2026? To me, error trackers have two failure modes: 1. False positives: They show you thousands of errors, and you can’t tell the impact on the user 2. False negatives: Many user-facing issues don’t throw exceptions, so they go unnoticed. Opslane combines error tracking and session recording. And there is an agent that acts on both. To get started, you install the Opslane SDK. It captures everything the user did: errors, console logs, network requests, and session recordings. Opslane reduces false positives by ranking issues based on how many users are facing a particular issue. It also learns about your product by reading your code and watching your session recordings. False negatives are harder. Opslane reviews session recordings to spot frustration. They look for rage clicks, dead clicks, and abandoned forms. This recently caught a bug in an early customer’s onboarding flow: a dropdown that closed itself when clicked. No exception, no bug report. The recordings showed users clicking it, selecting nothing, and dropping out of onboarding. Opslane flagged it and the team fixed it. Three guiding principles when building Opslane: 1. Open Source: Self-host with one Docker Compose file. 2.Agent-first: I never want to open an error dashboard again. Opslane ships an MCP server, so you can ask "what broke for users this week" from Claude Code. You get back issues that need your attention and you drive the resolution. 3. It knows about your product: Opslane is continuously learning about your product. Every investigation begins with what it knows about your product. It’s early. Frontend apps work end to end today.I am currently focused on improving reliability and accuracy. Here is a link to our repo: https://github.com/opslane/opslane Would love to get feedback from folks on our approach to this problem! Comments URL: https://news.ycombinator.com/item?id=49466704 Points: 3 # Comments: 0
嗨，HN，我们构建了一个开源模型网关。它是一个集中管理我们自托管、前沿和开源模型的单一入口。 它原生使用 Rust 构建，面向并发，并实现了跨模型和提供方的各种配置细节差异（流式格式、工具调用、模型参数、速率限制以及不同的错误行为）。 对于 BYOK 请求，这个网关增加的延迟不到 1 ms；当由 Experiential 提供供应商密钥时，延迟不到 2 ms。它覆盖所有主要推理提供方，并通过一个会发起 PR 的 codex agent 每日刷新 1000+ 个模型。 与其他类似项目相比，我们是开源的，不收取任何加价，允许你将本地模型与市场混合使用，并利用你的流量来（选择加入后）训练一个模型。简单路由并不值得收取 10% 的 token 加价。 我们的做法是：基于标准化的 OTel traces，我们挖掘具有代表性的真实任务，使用 text world models 模拟各种模型的 rollouts，应用一个 LLM judge，然后在 prompt 的 embedding 之上拟合一个最近邻分类器，以决定每个请求的最优模型。通常这比直接调用单个模型能画出更好的成本/质量帕累托曲线，但并不完美。 借助这些模拟，我们还可以做诸如建议缓存命中优化、新模型建议以及训练模型之类的事情。 它是开源的，所以你可以把它部署到自己的基础设施上，使用我们托管的版本且不收取任何加价，或者在我们的网站上阅读我们如何为最大可用性进行设计。 Comments URL: https://news.ycombinator.com/item?id=49471407 Points: 9 # Comments: 0
Hi HN, we're the founders of Salem Robotics ( https://salemroboticsinc.com ). We give existing mobile robots the task-specific intelligence to carry out surveys and physically interactive inspections in hazardous industrial facilities. Here's a video of it running on real robot hardware with a few words from us: https://youtu.be/U_228h3NE7c We came to Salem through robotics research at UT Austin and a combined 15 years working in nuclear, including about 10 years developing and deploying autonomous robots at Los Alamos National Laboratory. Over the last five years, we kept running into the same gap: robot hardware had become very capable, but making a robot carry out a complete industrial procedure still required a surprising amount of robotics work and manual intervention. The part that interested us most was manipulation. A nuclear contamination survey, for example, can require taking a "smear": wiping a defined area of a surface so it can be checked for removable radioactive contamination. In an oil, gas, or chemical facility, an LDAR (leak detection and repair) inspection can require moving a detector around a particular valve, flange, or connection. Other inspections require positioning an instrument at a precise location and orientation relative to a pipe or piece of equipment. These are easy tasks to compress into verbs like "wipe", "measure", or "inspect", but considerably harder to make a robot do reliably. A probe might need to remain normal to a surface throughout a path, stay within a narrow offset from a pipe, or trace a region while maintaining a particular end-effector orientation. The planner has to find a feasible motion while respecting the task geometry, manipulator kinematics, joint limits, collisions, and the environment around it. We work down to joint-level control for those interactions. One problem we've spent a lot of time on is generating constrained manipulation plans quickly enough that they can be based on the geometry the robot actually observes instead of requiring someone to carefully author a trajectory for every individual surface, valve, or flange. The physical world makes this annoying. A few centimeters of error may not matter much when navigating down a hallway, but it matters if a sensor is supposed to remain normal to a curved surface. And successfully executing a trajectory doesn't necessarily mean the inspection worked. The detector could be misaligned, contact could be wrong, the geometry could differ from the model, or the measurement itself could be invalid. We care about closing that loop around the inspection result, not just whether the arm reached the commanded pose. Our approach is a combination of AI and classical robotics. A lot of robotics research and industry attention right now is going toward increasingly end-to-end learned systems, particularly around humanoids. Working in safety-critical environments has made us appreciate how relevant classical approaches still are when you want explicit constraints, predictable behavior, and theoretical guarantees about what a robot can and cannot do. We use AI where semantic understanding and flexibility are useful, such as interpreting less structured information or understanding what in an unfamiliar scene is relevant to a procedure. Once the system knows what physical interaction it needs to perform, we prefer explicit geometry, planning, optimization, and control where possible. We're interested in the marriage between the two rather than trying to make every part of the robotics stack learned. The other idea behind Salem is that we don't think every useful robot application should require building a new robot. Companies like Boston Dynamics are getting very good at building increasingly capable hardware platforms. We think there is room for a domain-specific application layer on top of that hardware. The same underlying robot might perform nuclear radiological surveys in one facility and LDAR inspections in another, but the procedures, sensors, manipulation constraints, success conditions, and outputs are different. That's also why we're hardware agnostic. We don't expect one robot to be the best platform forever, and facilities already own different hardware. We'd rather describe an inspection in terms of what needs to happen and then map that onto the capabilities of the right robot for the job. One thing that surprised us after spending more time with facilities is how manual many inspection workflows still are. In sophisticated nuclear and industrial sites, people still physically walk survey routes, take measurements one at a time, visually inspect equipment, record results manually, and sometimes make judgments based on things like how a component sounds. Some of the basic workflows would be recognizable to someone doing the job decades ago, even though the sensors, computation, and robots available today are radically different. We're starting with radiological inspection in nuclear because it's the industry we know best, and we're also working on manipulation-heavy inspection problems in oil, gas, and hazardous chemical facilities. The technicians and inspectors still define the procedure, interpret results, and make the consequential judgments. We're trying to automate more of the repetitive physical execution that currently requires someone to enter the environment or manually operate a robot. We sell directly to industrial facilities, usually starting with a paid technical validation and then moving to an ongoing deployment. Pricing varies substantially with the workflow: validations range from tens of thousands to over $100k, and larger deployments can range from the low hundreds of thousands to roughly $500k per robot. One thing we're especially curious to hear HN's thoughts on is where the abstraction boundary in robotics should sit. What should come from the robot manufacturer, what belongs in an application layer, and what will inevitably remain specific to the facility? We'd also be interested in hearing about other industries where you've seen physical inspection tasks that look trivial to a person but are surprisingly difficult to automate. Comments URL: https://news.ycombinator.com/item?id=49466715 Points: 2 # Comments: 0
文章 URL: https://aiandeducation.mit.edu/report/ 评论 URL: https://news.ycombinator.com/item?id=49464314 得分: 13 # 评论数: 4
Robust Markov Decision Processes (RMDPs) generalize classical MDPs by allowing uncertainty in transition probabilities and optimizing against their worst-case realization. We consider $(s,a)$-rectangular RMDPs with \emph{linearly defined} uncertainty sets and study parity objectives, which are a canonical representation of $ω$-regular objectives. An uncertainty set is linearly defined if it is described by linear inequalities over the transition distribution together with auxiliary variables, which capture the standard $L_1$ and $L_\infty$ balls as well as general polytopic uncertainty sets. The quantitative value is the supremum, over all agent policies, of the satisfaction probability guaranteed against the adversarial environment. Previous work studied the qualitative analysis, namely the almost-sure (resp. positive) problem that asks whether a single agent policy guarantees satisfaction with probability one (resp. positive probability) against every environment policy. In this work, we solve the exact quantitative problem. Our contributions are threefold. First, we show that both the agent and the environment admit pure memoryless optimal policies. Second, we give a polynomial-time algorithm for quantitative parity on linearly defined robust Markov chains and use it as a subroutine in a policy-iteration algorithm for RMDPs. The algorithm combines quantitative one-step improvements with qualitative almost-sure improvements. Finally, we report experiments comparing our approach with the explicit reduction to stochastic games.
理解科学领域中的指令遵循能力，对于有效利用多模态大语言模型（MLLMs）推动科学领域发展至关重要。在这项工作中，我们提出 SciMIF，这是一个新颖的基准，用于评估 MLLMs 遵循复杂科学指令的能力。具体而言，基于对 5 个代表性科学学科中 22 项不同任务的广泛分析，我们提出了一个全面的分类体系，包含 10 组约束，能够同时刻画通用功能需求和学科特定特征。在该分类体系的指导下，我们开发了一条高保真指令注入流水线，用于系统性地增强现有科学数据集。我们在多个最先进的闭源和开源 MLLMs 上进行了全面实验。研究结果表明，不同科学学科之间存在显著的性能差异，其中化学对当前 MLLMs 构成了更大的挑战。此外，我们观察到，增大模型规模并不会带来相应的约束遵循能力提升，而且当前模型在细粒度约束以及需要深入应用学科知识的指令上仍然表现出严重困难。SciMIF 填补了当前在科学领域多模态指令遵循评估方面的空白，为未来提升 MLLMs 在严谨科学应用中的能力奠定了关键基础。数据和代码将发布于 https://github.com/shenye7436/SciMIF 。
Retrieval-augmented generation (RAG) is widely used to mitigate hallucination issues in large language models (LLMs) and multimodal large language models (MLLMs). In particular, knowledge graph (KG)-based RAG leverages structured knowledge to provide (M)LLMs with high-quality external information. Building on these works, recent studies have explored multimodal knowledge graphs (MMKGs) as knowledge bases for GraphRAG. This enables Graph RAG to integrate knowledge across multiple modalities, thereby further enhancing its performance. However, existing MMKG-based RAG methods generally follow a common pipeline in which different modalities are largely processed independently before being fusion. As a result, textual context is only used to a limited extent during visual information extraction and subsequent multimodal knowledge fusion. This brings a semantic gap between images and text which limits the multimodal GraphRAG performance. To address this issue, we propose a novel framework for constructing a Context-Enhanced MMKG (CEMMKG) to better support multimodal GraphRAG. The proposed CEMMKG enriches each image with complementary textual context at both local and global scopes. Local context goes beyond the surrounding text by incorporating sentences that are semantically related to the image, while global context provides a summary of the entire passage. We further introduce a multi-granularity design for the local context, allowing it to capture semantically relevant information at different levels of detail. Extensive experiments on the selected vision-centric dataset validate that CEMMKG is effective in leveraging contextual information to improve MMKG-based RAG performance. Moreover, its effectiveness across different MMKG-based RAG methods demonstrates its broad applicability.
自动定理证明器逐步构建证明，在每一步选择要添加和要移除的内容。我们将这一构造表示为在由形式演算诱导的转移系统中起作用的策略，该演算规定哪些步骤是正确的：对于子句连接表演算，leanCoP 风格的搜索和 plCoP/rlCoP 风格的规划于是变成了在同一接口上的有状态策略，策略学习方法可以直接应用。我们为这类策略配备一个图神经网络，利用可在不同问题间迁移的结构来为证明编辑打分；再用已找到的证明进行模仿学习训练，并考察当我们移除搜索脚手架时性能如何保持，从完整的符号回溯到仅由网络驱动的策略。在 M2k、MPTP2078-bushy 和 TPTP v9.2.1 上，在固定步数预算内，学习得到的策略比 leanCoP 多解决最多 46% 的问题，并且能以少一个数量级的步数找到证明。
Reasoning in language allows foundation models to spend more test-time compute on hard problems, such as those requiring decomposition, constraint tracking, and prediction of future consequences. Whether this mechanism can improve robotic manipulation remains unclear, where long-horizon tasks require tracking partial progress, reasoning about object relations, recovering from mistakes, and steering noisy low-level policies. In this paper, we study whether VLMs can be trained to reason directly in natural language to guide low-level manipulation policies. We introduce $R^3$, a simple post-training recipe that turns off-the-shelf VLMs into robotic reasoners: it first mid-trains a VLM on expert-generated reasoning traces to initialize the desired reasoning style, then improves the reasoner with single-step rubric-based RL from offline action data. Unlike prior robotic reasoning methods that mostly use structured traces as auxiliary supervision, $R^3$ trains free-form language reasoning to produce test-time guidance for action. We instantiate $R^3$ on Language Table and simulated bimanual grocery packing, two controlled testbeds for studying robotic reasoning and long-horizon manipulation. $R^3$ improves exploration and generalization across unseen tasks and significantly outperforms instruction-only imitation learning baselines on both benchmarks. Our analyses suggest that free-form language reasoning can function as a test-time compute mechanism for steering low-level policies. Our project page is available at https://robotic-reasoner.github.io/.
测试时扩展使用额外的测试时计算来提升性能，例如让语言模型在解决问题时进行更长时间的推理。由于模型通过完整注意力机制将整个推理轨迹保存在内存中，那些需要长时间思考的困难任务可能会带来高昂的成本。然而，我们发现，随着模型持续推理，大多数中间推理 token 会逐渐失去重要性。这就引出了一个问题：保留它们是否值得付出这样的成本。基于这一洞察，我们提出 Prefix Sliding，它会在推理过程中丢弃那些不属于前缀或最近几千个 token 窗口的 token。前缀包含模型可用的关键指令和工具，而最新的 token 则是模型当前正在处理的推理内容。这样，无论模型推理多长时间，总内存需求都被限制住了，从而能够高效地进行长跨度的测试时扩展。在无需训练的情况下，Prefix Sliding 可以让现有模型快 3 倍，同时保持性能。使用强化学习配合 Prefix Sliding 进行训练，可以通过支持扩展到超过十万个 token 的推理轨迹而实现更好的性能。消融实验表明，Prefix Sliding 的表现优于对中间 token 进行摘要或普通的滑动窗口方法。我们的代码见 https://github.com/Muennighoff/prefix-sliding
Intent misinterpretation during vehicle interactions causes recurring planning failures. We study a decision layer in which a language-guided intent module reads structured descriptors, computes a smoothed intent-geometry divergence score, and gates the planned maneuver before commitment, upstream of a corridor envelope. On a replayed off-road departure and four crash clips under a frozen, disclosed implementation, gating is the only layer that repairs the plan: on the main case it fires 72 ms after the drift onset but 161 ms before the corridor exit, keeping the trajectory in the corridor in all ten replays. The first calibration draws nine false triggers in 5.9 minutes, each from scoring uncertainty as half a conflict; a preregistered redesign treating uncertainty as abstention cuts this to 0.341 per minute. Two ablations bound the model's contribution: the full score detects fastest on four of five failures under the deployed eligibility, three of five against the unvetoed rule (000871 by one cycle; 000228 by a pre-onset fire on an uncertain stretch that five clips cannot classify as signal or coincidence; dropping the confidence term costs two detections), while on in-domain tracks at equal false positives the geometric rule more than triples its detection. The evidence supports the gating mechanism; the model's demonstrated roles are the fastest detection on these failures and an uncertainty veto on the geometric rule.
现有的动作质量评估（AQA）数据集和方法主要依赖 RGB 和姿态等视觉输入，忽视肌肉力学等生理动态，且常将动作建模为单一整体模式。这些局限阻碍了细粒度、基于生物力学的反馈。我们提出 MyoMechanix，这是一个面向负重动作的多模态生态系统，将运动与肌肉活动对齐。该数据集由专家标注，包含来自 38 名受试者的 20 个动作、7500 多个样本，具备同步的多视角 RGB 视频、3D 姿态、sEMG 及其他生理信号，构成迄今最大的多模态 AQA 基准。我们进一步构建了 Fitness Knowledge Graph（FKG），将专家标注组织为动作、阶段、关键步骤、错误和纠正反馈之间的结构化关系，从而支持组合式评分和可解释评估。基于这些表示，我们开发了 CUBIST（Compositional Ontological Reasoning Engine），通过分解-分析-重组实现细粒度错误归因和反馈生成。我们还建立了 MyoMechanix-AQA、MyoMechanix-VideoQA，以及一种新的 MyoMechanix-Video2EMG 任务。实验表明，多模态感知和结构化表示能够提升性能、可解释性和错误归因，其中 CUBIST 取得了最先进结果；VideoQA 增强了语言驱动的动作理解；Video2EMG 则表明，可用基于视频的替代方案来取代代价高昂的 EMG 传感。MyoMechanix 将技能活动理解推进到面向 Physical AI 应用的生物力学基础、多模态和组合式推理层面，服务于健身、康复、医疗保健和机器学习。Project page: https://haoyin116.github.io/MyoMechanix/
https://www.theinformation.com/articles/nvidia-agrees-buy-op... （付费墙） https://techcrunch.com/2026/08/24/hugging-face-reportedly-in... 评论链接: https://news.ycombinator.com/item?id=49458161 得分: 19 # 评论数: 3
One common trade-off in the use of large language models involves reducing the size of the model while increasing the amount of computation at inference time, for example by using a wider beam search. In this paper, we examine the constrained case of this "model size vs. inference compute" trade-off, in which the model outputs are constrained by a strict grammar at inference time. Our results demonstrate that the constrained trade-off behaves differently from the unconstrained trade-off. We investigate the task of converting a prose query into an equivalent SQL query (text-to-SQL). Performance is evaluated on the Spider text-to-SQL benchmark, using the Qwen2.5-Instruct model family ranging in size from 0.5B to 7B parameters, all at 4-bit precision. We experiment with two approaches to varying inference compute: (i) beam search with a variable number of beams; and (ii) sample+vote, i.e., sampling several constrained outputs and then voting on their execution results, where the number of samples is varied. On the 1034-example development set, we find that: (a) both beam search and sample+vote improve accuracy, especially on smaller model sizes; (b) the "model size vs.\ inference compute" trade-off is not advantageous in this experiment, because moving to a larger model size typically results in higher accuracy than increasing inference compute on the same model size; (c) beam search outperforms sample+vote at a matched inference budget. This latter result is of particular interest since it contrasts with the findings of the unconstrained trade-off.
Modern GUI-agent frameworks achieve strong desktop task performance with frontier API models, yet persistent control information often remains implicit in growing interaction trajectories. At each step, the planner reconstructs the active task stage, accumulated evidence, and runtime feedback before deciding the next action. This dependence becomes more pronounced under weaker local reasoning backbones. Across four representative state-of-the-art frameworks, replacing GPT-5 with Qwen3.5-9B reduces average OSWorld SR-100 from 60.9\% to 37.7\%. Trajectory annotation further identifies at least one control failure in 91.6\% of failed trajectories. To address this problem, we introduce LocalLSTC, a training-free architecture that organizes control by temporal scope, maintaining persistent cross-step state to guide short-term execution commitments. Long-Term Control maintains the active subgoal, subgoal-aligned evidence, and runtime feedback across interactions, while Short-Term Execution realizes bounded commitments for the current step. Long-to-Short Planning forms each commitment from persistent state, and Short-to-Long Control integrates execution outcomes back into that state for progress assessment, recovery, and termination. With Qwen3.6-27B, LocalLSTC reaches 64.7\% SR-100 on OSWorld and 65.3\% on WindowsAgentArena, outperforming the strongest prior local results on both benchmarks. Ablations further support contributions from mechanisms on both sides of execution. These findings identify temporal organization of control information as a distinct architectural dimension for locally deployed GUI agents.
大型语言模型在不同语言之间获取知识的表现并不一致，但当它们使用不同语言交互时，其技能集合在多大程度上会存在差异？本研究从与知识和通用基准性能正交的角度，对跨语言技能不一致性进行了量化。我们通过多语言自博弈来实现这一点：同一模型的两个实例在一个基于文本的游戏中对战，每个实例都通过不同的语言接口进行交互。由于模型、对手、规则、状态空间和可用动作都保持固定，这种设置隔离了语言对模型实际行为的影响。我们构建了 TextArena 的多语言扩展，并评估了三个开源权重模型在八种语言和六个游戏中的表现，这些游戏涵盖空间推理、不完全信息、资源分配和重复交互。我们发现，同一个模型在不同语言下的对局实力可能显著不同，在胜负差距、无效动作和策略倾向上都存在系统性变化。详细分析揭示了特定语言下在空间推理、受牌面条件约束的决策以及最优移动选择方面的失败。在某些设置中，仅改变中间推理所使用的语言就能恢复大部分损失的性能，这表明语言会影响决策过程的不同阶段。这些结果表明，技能差异是构建真正多语言模型时一个可测量且重大的障碍。更好地理解这些差异有助于我们设计出在不同语言之间表现更公平的模型。
Explaining deep learning models operating on time series data is crucial in various applications that require transparent and interpretable insights into model behavior. {Existing explanation methods generally fall into two categories: attribution-based explanations, which identify the temporal regions most responsible for a prediction, and counterfactual explanations, which reveal how an input should be modified to alter the model's decision.} {Despite valuable insights, these two fields are largely studied independently. This disconnect leaves attribution methods lacking causal validation, while counterfactual methods suffer from severe instability, producing adversarial-like noise instead of meaningful explanations.} In this work, we revisit time-series explainability from an information-theoretic perspective and show that existing explainers are vulnerable to trivial solutions and distributional shifts. To address these limitations, we propose a unified objective function for explainable time series learning that bridges attribution and counterfactual reasoning within a single framework. Building upon the Information Bottleneck principle, our formulation explicitly prevents trivial explanations and out-of-distribution counterfactuals. {Based on this objective function, we introduce {\modelname}, a novel explanation framework that learns a parametric transformation network to construct explanation-embedded instances, where preserved information yields attribution explanations and controlled information removal produces stable counterfactual explanations.} We evaluate {\modelname} on synthetic and real-world benchmarks against state-of-the-art baselines. Extensive quantitative and qualitative results show that {\modelname} consistently outperforms competing methods, yielding faithful attributions and stable counterfactual explanations.
GNN-based Knowledge Graph Question Answering (KGQA) pipelines process queries through four discrete stages: entity linking, subgraph retrieval, GNN reasoning, and answer generation. Standard robustness evaluations conflate stage-level failures into a single end-to-end metric, obscuring both the source of brittleness and the appropriate mitigation target. We ask which stage fails, and why, when the pipeline is subjected to adversarial perturbations on the input question. We introduce a stage-isolation protocol with two answer-preserving adversarial perturbations verified against the knowledge graph: Compositional Restructuring (CR) and Relation Synonym Swap (RS) target distinct stages while leaving entity seeds intact. Evaluated across ComplexWebQuestions and WebQSP, the results run counter to prevailing assumptions: the GNN reasoning stage retains near-baseline accuracy when the subgraph is intact, while subgraph construction accounts for over 99\% of the end-to-end collapse under CR, occurring even when the gold answer is present in 74\% of retrieved subgraphs. This exposes a fundamental distinction between answer presence and answer reachability that end-to-end metrics cannot detect, and places the mitigation target firmly at the subgraph construction stage rather than the reasoning model. Perturbed datasets and evaluation infrastructure are released at https://anonymous.4open.science/r/atkgrag-E85C .
*Chulin Zhao 和 Ruoqi Hu 对这项工作贡献相同。 当提示词涉及多个实体和多个属性等复杂的组合因素时，最先进的文本到图像（T2I）模型会表现出明显且系统性的缺陷。本文研究人类如何识别此类缺陷。具体而言，我们从人、手、物体和场景四类中手动选取了 651 张具有复杂组合特征的参考图像，并通过手动编辑 ChatGPT 生成的提示词，从中提炼出强调组合因素的提示词。随后，我们将这些提示词输入三个选定的 T2I 模型以生成 AI 图像，并开展了一项全面的主观研究来识别其中的缺陷。对于每张图像，29 名参与者提供多标签评估，指定缺陷类型和位置。该研究最终构建了组合式 AI 生成图像缺陷（CO-AID）数据集，其中包括参考图像、提示词、AI 生成图像，以及缺陷位置和类型信息。实验结果表明，在 CO-AID 上训练深度模型既能预测 AI 生成图像中的缺陷，也能优化 AI 图像生成，证明了其可用性和有效性。数据库和补充材料可在以下地址获取：https://github.com/Future-IQA/CO-AID 。
Traffic Anomaly Understanding（TAU）要求模型和系统能够在交通视频中检测、推理并解释异常事件。为解决这一挑战，我们提出 TAU-Agent，一种面向交通异常理解的智能检索增强框架。给定一个任务查询，中央检索代理会协调两种视觉感知工具，即 Video Captioning Tool 和 Open-Vocabulary Tracking Tool，以检索并选择与查询相关的证据，包括字幕、时间区间和目标轨迹。所选证据与采样的视频帧以及输入查询一起，提供给经过监督微调的视觉语言模型，用于最终推理和答案生成。我们在 AI City Challenge 2026 的域内和域外基准上评估了 TAU-Agent。TAU-Agent 在 Track 3 上取得 0.6779 分，在 Track 7 上取得 0.3998 分，在 Track 8 上取得 67.9275 分，分别排名第二、第十二和第五。代码可在以下地址获取：https://github.com/siri-rouser/TAU-Agent。
On-policy distillation trains a language model on its own generations while a teacher scores them token by token. It combines the dense supervision of imitation learning with the on-policy sampling of reinforcement learning. But it requires a second, larger model to act as teacher. On-Policy Self-Distillation (OPSD) removes that cost. The teacher is the model itself, conditioned on privileged information the student will not have at test time, such as a reference solution, a plan, or environment feedback. The teacher is no stronger than the student, only better informed. Early results were promising, with accuracy comparable to reinforcement learning at a fraction of the generated tokens. But the same asymmetry that produces the signal also biases it. One failure mode now dominates the field: collapse, the progressive narrowing of the set of reasoning paths the model can produce. Collapse is not specific to OPSD, though privileged information aggravates it. This review treats collapse as a symptom governed by three levers: (i) where the signal is applied, that is, how tokens are weighted; (ii) what the teacher is shown, that is, the nature of the privileged information; and (iii) when the signal changes, that is, the teacher's dynamics and the decay of guidance. We restrict our scope to mathematical reasoning, where the method originated and where its failure modes are best documented. We report no new experiments. The contribution is structural: a shared vocabulary for phenomena named differently across papers, and a clear line between what is settled and what is still disputed.
文章网址: https://openai.com/index/hugging-face-incident-and-the-road-ahead/ 评论网址: https://news.ycombinator.com/item?id=49454314 得分: 7 # 评论数: 1
Hi HN, we're Sam and Alex, founders of Risklytics ( https://risklytics.ai ). We're both on leave from Harvard, and we run an insurance brokerage for companies building robots, drones, autonomous systems and satellites. You describe your machine, we find the insurance companies willing to cover it, and we read everything before you sign. Buying insurance for any frontier tech goes like this: - your first customer or pilot requires coverage before your machine can work on their site - a broker hands you an application written for ordinary businesses, and often will deny you on baseless "risks" (one of our clients got denied for using CAD software in their workflow) - you get filed under something close or outright denied, the person at the insurance company often can't tell what they're looking at, and the safe answer is no - if a quote does come back, the AI risk is often quietly removed by an add-on clause that prevents you from truly being covered Those clauses surprised us. Insurance companies mostly don't write their own policies from scratch. They license standard forms from an industry body called ISO and attach add-on clauses to adjust them, and in January ISO published clauses that void coverage for losses arising out of AI in any part of a process. Each insurer decides whether to adopt them, account by account, and there's no public record of who has and to what capacity. There's an open question in the wording too. It defines generative AI as systems trained on data that produce text, images, audio, video or code. Whether a robot's control model, which produces motor commands, falls under that is unsettled. We've asked underwriters, the people at insurance companies who make these calls, and gotten different answers across the board. We didn't initially plan to be brokers. We came into YC to work in insurance and spent the first half of the batch on a wildfire risk model we wanted to sell to insurance companies. They liked it and wanted to build their own, which is the polite way to say no. While that was dying, a batchmate who launches satellites mentioned he couldn't get insured, and once we started asking around we kept hearing the same story from hardware companies. So we got licensed in a week and now can place coverage. What we've built so far: - a six-step application that asks about what you're building and drafts the structured version insurance companies like, omitting anything you didn't say - a map of which insurers add the AI clauses, which will remove them, and which cover frontier companies on purpose - when an insurer says no, you see their reason and what would change it, full transparency all the time - a hands off process for companies trying to get fully covered Here's a demo of the whole process: https://youtu.be/gVqCamTNH5E On money: the insurance company pays us a commission when you get covered, usually 10-20% of what you are charged, and we charge you nothing on top (that's right, no brokerage fees). Also, if you're a software company with ordinary needs, we'll still handle the whole process for you and get you the best policy we can find. In the past week alone we have helped three companies get covered that were denied elsewhere. The companies include a robotics wholesaler and rental store, who had trouble getting covered earlier because some of their property uses AI (even though they never use it themselves), a bridge collision system that was getting denied because they use CAD software, and a tele-operating robotics service that needed coverage within a week for their first pilot. If you're looking for insurance in any capacity feel free to reach out to me at alex@risklytics.ai or fill out an application at https://risklytics.ai . And if you work on the carrier side and think you'd wanna partner with us let us know. Comments URL: https://news.ycombinator.com/item?id=49451495 Points: 4 # Comments: 0
Reinforcement Learning with Verifiable Rewards (RLVR) and on-policy distillation (OPD) have become two widely adopted paradigms for post-training large language models. However, RLVR suffers from sparse task-level feedback, while OPD provides dense token-level guidance but ignores trajectory correctness, limiting its performance to that of the teacher. Combining them is a promising direction: OPD supplies dense supervisory signals, while RLVR provides task-level correctness. Nevertheless, existing integrations often rely on weighted combination or heuristic switching, introducing extra hyperparameters and trade-offs. We propose On-policy Distillation with Verifiable Reward (OPDVR), a simple yet effective method that seamlessly combines OPD and RLVR without adding any hyperparameters. We first reformulate the implicit reward of sampled-token OPD based on trajectory correctness, then apply a ReLU gating mechanism to ensure that correct trajectories receive non-negative rewards and incorrect ones receive non-positive rewards---thereby aligning the distillation signal with task success while preserving the teacher's distributional guidance. Furthermore, our modification transforms sampled-token OPD into a proper RLVR method, making it readily combinable with any policy gradient algorithm, such as GRPO. Experiments on six reasoning benchmarks show that OPDVR consistently outperforms standard OPD. Our code is available at https://github.com/LeapLabTHU/OPDVR.
Optimization of hyperparameters is a critical factor to obtain optimal model performance. While existing research has predominantly concentrated on batch-learning scenarios, addressing the complexities inherent in data streams presents a challenge. The deployment of sophisticated methodologies to manage data streams becomes highly important. Consequently, the capacity for self-adjusting hyperparameters during on-line learning phases emerges as a goal. Many hyperparameters exhibit constraints and are confined within bounded search spaces, rendering specific solutions unacceptable upon applying optimization operators. To solve this issue, employing boundary constraint- handling techniques becomes imperative to rectify invalid solutions. This paper presents strategies for effectively managing boundary constraints within constrained numerical optimization problems. Recent methodologies, including heuristic and evolutionary-based optimization, employ a "boundary" strategy, wherein values that surpass boundary thresholds for a given hyperparameter are realigned to the respective limits. Our study introduces four strategies to navigate boundary constraints in online optimization algorithms. Through empirical investigations conducted on established datasets, we demonstrate that adopting boundary strategies outperforms the "boundary" strategy.
A vast amount of optical satellite data is being transmitted to Earth-based servers every day, and more than half of this data is affected by haze or clouds. Additionally, this data suffers from the fundamental trade-off between spatial and temporal resolution, which remains largely unresolved, making the acquisition of continuous high-resolution satellite observations of clouds an ongoing challenge. This work addresses this challenge by proposing two Deep Learning super-resolution methods for the accurate downscaling of SEVIRI cloud mask products, as well as a novel cross-sensor cloud mask dataset called SEVMOD-CM, created by spatially and temporally matching MODIS and SEVIRI satellite observations. The two proposed models are a CNN-based (SpatialCNN) and a GAN-based (SpatialGAN) Neural Network. Trained on the SEVIRI spectral and cloud mask products, the proposed methods predict the corresponding MODIS Cloud masks, achieving a 4x spatial enhancement across sensor domains. Both approaches are evaluated experimentally, and compared against the standard bicubic interpolation upsampling technique. The experimental results demonstrate the value of the proposed models and dataset for the remote sensing community, highlighting the benefits of applying super-resolution techniques to geostationary-derived cloud mask products for applications such as atmospheric monitoring, weather forecasting, disaster risk reduction, solar energy forecasting, and climate research.
通过自监督学习预训练的 EEG 基础模型有望获得可迁移表征，但其泛化能力仍然有限，尤其是在不同临床数据集之间。由于计算需求高，在资源受限的临床环境中进行全量微调并不现实。在这项工作中，我们研究参数高效的自监督适配是否足够，即仅更新 9% 的参数能否将表征对齐到目标任务。我们在两个采用不同预训练目标的最先进模型上评估了该方法：BIOT（对比学习）和 CBraMod（掩码重建），并在三个临床 EEG 数据集上进行评估，任务分别为异常检测（TUAB）、事件分类（TUEV）和癫痫发作检测（CHB-MIT），同时考察分布内和分布外两种条件。SSL 适配相较于线性探测带来了稳定提升，AUCPR 最高可达 20 倍。在固定计算预算下，实现峰值性能只需要可用无标注数据的 20--50%。关键的是，当总窗口数固定时，性能与患者数量无关，这表明性能仅依赖于整体时间窗口多样性。我们的研究结果表明，参数高效适配能够以最小的计算开销和数据采集负担，有效部署 EEG Foundation models（EEG-FM）。代码地址：https://github.com/c3n-group/efficient-eeg-adapt
自我改进的 LLM 代理优化的是答案，而不是产生这些答案的过程。添加元层的系统会将该层固定不变，而会自我编辑的系统为了保持稳定，必须让自身编辑机制的一部分保持不变，因此其实现的元深度大致被限制在二层左右。我们提出 Meta$n$，它保持元操作固定，而是对其输入递归。该操作 Ω 会反复作用于自身的产物，读取下方求解器栈的轨迹以及生成这些轨迹的代码，然后将下一层写成一个策略性预处理步骤和一组可调用的辅助函数。由于 Ω 从不改变，它就不会使系统失稳；又因为其输入严格增长，每一层都从比上一层更高的视角进行推理。深度由收敛而非预先固定来决定，并且一个进化式档案会在层链上进行搜索。在两个骨干模型上，Meta$n$ 在全部八个基准家族上都优于先前的自我改进代理。最显著的例子是 ARC-AGI-2，其设计旨在抵抗技能记忆，在该基准上只有它得分高于零。消融结果表明，递归带来的大部分收益来自每一层传递给下一层的条件信息；随着深度增加，会涌现出不同的层角色，尽管没有任何提示词预先规定它们。代码可见于 https://github.com/minnesotanlp/meta-n
Discovering stable neuron behavior across entire domains remains a challenge in mechanistic interpretability. Existing methods often rely on instance-level point estimates or computationally expensive procedures, which either obscure population-level variability or limit scalable domain-wide analysis. We present RACE (Residual Alignment for Consistency Estimation), a forward-pass statistical framework that evaluates the domain-wide functional consistency of Transformer neurons. Perturbation experiments demonstrate that RACE achieves superior domain specificity compared to gradient-based point estimates. Meanwhile, token-distribution-level results verify the association between the selected neurons and the target domain. Furthermore, its computational overhead is two orders of magnitude lower than that of gradient-based methods.
Large language model agents are moving beyond conventional retrieval-augmented generation toward direct interaction with external corpora. Direct Corpus Interaction (DCI) keeps the full corpus accessible, yet reachable evidence can remain unusable under finite interaction budgets. Required evidence may fail to surface, a surfaced supporting document may remain unopened, or an opened document may fail to expose its decisive fragment. We call this progressive silent loss Evidence Blindness and quantify it through stage-wise evidence realization. Within the DCI paradigm, raw interaction adds little reusable corpus organization, while dynamic-workspace methods reconstruct a query-conditioned interaction space from each query and trajectory. In both cases, useful structure is recovered largely online. We instead formulate large-scale agentic search as finite-budget navigation over reusable corpus structure. We introduce AtlasNav, a persistent multi-view corpus-navigation framework that retains direct corpus interaction but organizes the corpus once into a Corpus Atlas, allowing each query to navigate adaptively rather than reconstruct shared structure. On BrowseComp-Plus, AtlasNav achieves 92.05% strict accuracy while reducing recorded online inference cost by 30.21% relative to the prior dynamic-workspace state of the art. Under matched budgets, it realizes the complete required evidence earlier and approaches the same model's evidence-supplied empirical reference more rapidly. The same representation principle remains effective under PhantomWiki's distinct corpus organization and controlled 10K-1M scaling, and transfers competitively to heterogeneous enterprise knowledge. These results show that agentic search depends not only on accessible evidence, but also on how the corpus is represented so that limited interaction becomes effective navigation.
基于 LLM 的代理可以通过调用工具与外部环境交互，但这一能力也带来了文件修改、信息泄露和未授权操作等安全风险。现有的防护机制通常在轨迹完成后进行评估，而对执行前的步级动作监控关注不足。我们提出 StepGuard，这是一种步级防护模型，既可以审计已完成的代理轨迹，也可以在工具动作执行前进行检查。为训练 StepGuard，我们引入 StepGen，一个自动化数据引擎，用于生成具有相同上下文但在高风险步骤采取不同动作的安全与不安全轨迹。为进一步减少过度防护和防护不足，我们提出 Balance-GRPO，它根据安全与不安全动作的观测准确率动态平衡学习。实验表明，StepGuard 在开源权重防护模型中取得了最高的平均准确率，其性能可与 GPT-5.4 相媲美。在 AgentDojo 和 AgentDyn 上用于保护代理时，StepGuard 相比无防护设置可将平均攻击成功率降低 77.3%，而平均效用仅下降 2.8 个百分点。
Model cards are structured documents that summarize key information about machine learning models to improve transparency, usability, and accountability. However, they often lack a consistent structure, and many models provide no model cards, making comparison and interpretation difficult. This paper presents two contributions. First, we propose MCTidy, an LLM-based approach that reorganizes existing model cards into a standardized template to improve clarity and comparability. Second, we introduce MCGenie, an LLM-based system that generates model cards directly from model repository data. We apply MCTidy to 48 Hugging Face model cards and evaluate information retention, section alignment, hallucination, and stability. Our findings show high information retention with minimal textual loss, accurate section assignment, rare hallucinations primarily in descriptive sections, and strong stability across runs. We assess MCGenie by generating model cards for the same 48 models and assessing semantic similarity, factual correctness, and sensitivity to input resources. The generated model cards achieved high semantic similarity (mean around 0.9); over half were fully correct, and most remaining errors were minor. Generation quality depended strongly on the availability of supporting resources, particularly associated papers. Overall, our findings demonstrate the potential of LLM-based methods to enable scalable, standardized model card documentation.
Recursive self-improvement (RSI) remains hard in long-horizon tasks, where growing histories obscure the task state and misalign skill invocation. We introduce Recuris, a recursive Experiential-Working Memory architecture for long-horizon agent harnesses, in which Working Memory tracks task progress and guides skill selection from Experiential Memory, grounding skill use in current needs rather than the full history. This coupling also turns execution into structured evidence that localizes failures to specific memory components. Across tasks, a fixed Meta-Agent turns that evidence into localized, validation-gated updates to Skill Memory that reshape execution and yield new evidence, forming a bounded recursive memory-evolution loop. Across four long-horizon benchmarks and ten models, Recuris improves task success in 35 of the 37 completed model-benchmark pairs, carrying frontier models to SOTA-level task success: on tau-bench it adds +17.8 points to GPT-5.6 Sol and +15.6 to Claude Opus 5, taking Opus 5 to 87.9%, and +16.6/+13.5 points on Qwen3.6-27B/35B on SkillFlow. The advantage widens as the interaction horizon grows, to +32.2 points on the longest tasks, and common long-horizon failures fall by up to 80%. These results position recursively evolving memory as a scalable foundation for RSI, enabling agents to continuously transform accumulated experience into increasingly effective long-horizon behavior. Code: https://github.com/Gen-Verse/Recuris
I'm a master's engineering student and a big fan of LaTeX, which I used for my thesis and research articles. I have used Overleaf and that was fine until I wanted to git sync, which unfortunately sits behind a paywall. Since I didn't want to pay subscriptions for things that should simply just work, I built the editor I wanted in my free time, where you open a tab, write LaTeX, get your PDF and the files stay all in one folder on my disk. TeXbrain is a static site with no backend. pdfTeX is compiled to WebAssembly (SwiftLaTeX) and runs in your browser. The editor can read and write your project folder through the File System Access API, so you can use git, any local TeX install, local AI Agents, or any other editor on the same files. Git is built in through isomorphic-git for anyone who would rather not touch a terminal and clone, branch, commit, push or pull via commands. No account is needed, no analytics, and everything works offline after the first load. Try it out: https://tex.swimmingbrain.dev/ (Chromium browsers get direct folder access, while Firefox or Safari fall back to a virtual filesystem) Or for self-hosting/contributing: https://github.com/swimmingbrain/texbrain (MIT, the pdfTeX engine is EPL 2.0 / GPL 2.0 and is listed in the THIRD_PARTY_LICENSES file). The part I'm most proud of getting to work is the package loading. The engine itself is only 1.8 MB. When it asks for a file it doesn't have, then a service worker intercepts the request and resolves it through Cache Storage first, then a small bundled subset, then a TeX Live tree mirrored on jsDelivr, then a SwiftLaTeX style server as a last resort. Every file is fetched once at most and after the first successful compilation, the core subset is prefetched in the background so that the offline story actually holds. Only file names go over the network, never any document content. So far, it only supports pdfTeX (no XeTeX or LuaTeX), so fontspec or polyglossia won't work, packages are pinned to the TeX Live 2020 era to stay coherent with the engine's format file and there is no bibtex or biber in the engine yet, which is maybe the roughest edge. This is should be a good solution for people on a work laptop, a Chromebook, or a university lab PC where installing TeX Live isn't always an option, and people who don't want to maintain a TeX install at all. If you already have a local setup you like, keep it. This isn't necessarily thought as a replacement for that. Some people have already used it for their thesis in the past months and their bug reports (which I would never have hit on my own) were then most of the summer's work. The first outside PR also landed recently, so if something doesn't compile for you, an issue with the package name or the bug itself is the most useful thing you can send me! Comments URL: https://news.ycombinator.com/item?id=49441375 Points: 6 # Comments: 1
文章 URL： https://9to5mac.com/2026/08/24/openai-restores-5-hour-codex-and-work-limits-for-chatgpt-plus-users/ 评论 URL： https://news.ycombinator.com/item?id=49432879 积分：4 评论数：1
It's a macOS menu bar app that reads the text of your focused window every few seconds through the Accessibility API. No screenshots, no video, or OCR. It writes plain markdown, one file per day, into a folder you choose. Point Claude Code (or anything with file access) at the folder and ask "what did I work on Tuesday?" or have it build memory about your projects. The folder contains an AGENTS.md explaining the format to the model. Thought it might be useful for someone so chucked it up on GitHub Comments URL: https://news.ycombinator.com/item?id=49429095 Points: 5 # Comments: 0
Discrete Diffusion Models (DDMs) have recently been introduced to recommendation systems, modeling user history as a token generation process via iterative denoising. However, while effective at capturing user-level sequential patterns, these methods often fail to explicitly integrate item-based collaborative filtering information, a critical component for accurate recommendation. This deficiency manifests in two key aspects: (1) the item representation is often semantic-focused, lacking collaborative priors for diffusion training; and (2) the denoising process employs a uniform noise schedule, treating all tokens indiscriminately and ignoring item-level adaptive structural dependencies. To bridge this gap, we propose ANR-DiffRec, a unified framework designed to encode item-based collaborative structures into discrete diffusion for generative recommendation. First, we explicitly incorporate an item co-occurrence matrix to guide semantic ID generation, providing a structured collaborative prior for discrete diffusion training. Second, we introduce an item-based adaptive noise rescheduling mechanism that dynamically adjusts denoising weights according to both local contextual recoverability and behavior-aware item dependencies. Specifically, the proposed strategy jointly models intra-item structural context and inter-item collaborative signals, enabling structure-aware denoising during diffusion training. Extensive experiments on multiple benchmarks demonstrate that our method consistently outperforms state-of-the-art generative recommendation models. Code: https://github.com/CalmaQi/ANR-DiffRec.
Agent skills are reusable procedural artifacts that extend language agents with specialized workflows, tool conventions, and domain behaviors at inference time. However, creating reliable skills still depends largely on human authorship, model priors, or execution traces. These sources are often unavailable for unfamiliar tasks, suggesting the need to create skills from open-world materials. In this paper, we study open-world skill creation: given an underspecified skill brief and a source-access specification, a creator must discover behavior-relevant requirements omitted by the brief and determine how broadly each source-derived procedure is justified. We propose SkillAlchemy, an admission-centered framework for source-grounded skill creation. SkillAlchemy identifies implicit requirements through contrastive evidence, admits candidate procedures based on evidence-supported scope, and compiles the admitted content into a grammar-guided skill package. Extensive experiments across 87 SkillsBench v1.1 tasks demonstrate that SkillAlchemy improves pass rate over no-skill execution by 19.9 percentage points and the strongest automated baseline by 8.6 percentage points, while achieving performance comparable to human-curated skills.
Diffusion Transformers (DiTs) have shown strong performance in high-fidelity image generation, but their sampling process remains computationally intensive due to full model execution at every timestep. While cache-based acceleration has been explored to mitigate inference cost, naive reuse schemes suffer from low accuracy over long intervals, and Taylor-series-based extrapolation methods often face instability caused by Runge oscillations. In this paper, we propose ChebBooster, a training-free extrapolation framework based on Chebyshev polynomial theory that achieves stable and efficient acceleration for DiTs. Specifically, we adopt the Barycentric formulation to evaluate Chebyshev approximants with high numerical stability and minimal overhead, and further decouple the extrapolation into an offline weight precomputation phase and a lightweight online application stage. Extensive experiments across three representative DiT-based models, including DiT-XL/2, PixArt-$Σ$, and FLUX.1-dev, demonstrate that ChebBooster achieves consistent improvements in visual quality and inference efficiency, reaching up to $3.68\times$ latency speedup and $5.12\times$ FLOPs reduction, outperforming existing training-free baselines under diverse generation tasks and resolutions.
Time series forecasting (TSF) is evolving toward multimodal and agentic settings, yet using foundation models remains uneconomical in resource-constrained scenarios, where compact, specialized forecasters are more desirable. However, lightweight forecasters typically require substantial training data, limiting their use in domains with scarce, slowly accumulated, or privacy-sensitive time series. To address this dilemma, we investigate the challenging problem of few-shot learning for lightweight forecasters. We propose MetaCaster, a meta-harness-optimized multi-agent framework that uses agentic data generation to automatically train specialized lightweight forecasters from only a few examples and textual contexts. Our work highlights a new TSF paradigm in which agents act not as forecasters but as intermediary engineers that prepare efficient, task-specific forecasters for deployment. Experiments on 18 datasets, 23 state-of-the-art lightweight forecasters, and 14 baselines demonstrate that MetaCaster achieves both data efficiency and computational efficiency while maintaining high-quality TSF performance.
We present Team Semiintelligencn's solution for the ACM RecSys 2026 TalkPlayData Challenge, addressing conversational music recommendation through a multi-modal and personalized conversational recommender system. Our submitted system employs a three-stage pipeline: (1) multi-modal retrieval constructing decay-weighted centroids across seven dense embedding spaces - track- and user-level CF-BPR, Qwen3 (metadata, lyrics, attributes), CLAP audio, and SigLIP visual - supplemented by BM25 lexical retrieval and an artist substring-match signal, all fused via weighted Reciprocal Rank Fusion (RRF) with optimized signal weights; (2) lightweight reranking (history filtering, popularity smoothing, and catalog diversity penalization); and (3) persona-diversified response generation using GPT-4o-mini. Beyond this submitted configuration, we report development-time experiments with additional components - constrained LLM-guided artist injection, album continuation signals, XGBoost LambdaMART, and a superior GPT-4.1 response prompt - that were not deployed to Blind B due to cost and complexity constraints. We optimize RRF weights on a 500-session development split via differential evolution, improving MRR by +19.5%. On Blind A, we observe that unconstrained LLM-guided injection across 54 sessions causes catastrophic nDCG regression (-18.9%), while conservative injection on only 9 sessions yields the best observed Blind A nDCG - a finding we present as a Blind A observation warranting further validation. The submitted system achieves a Blind B composite score of 0.3213.
自我反思是人类学习中进行信用分配的一种强大机制，能够将稀疏的结果反馈转化为可操作的指导。然而，其在 Large Language Models (LLMs) 后训练中的潜力仍未得到充分探索。我们提出 Self-Reflective Policy Optimization (SRPO)，这是一个将这种能力内化的框架。SRPO 使 LLMs 能够分析自己已完成的轨迹，将错误综合为简洁的“反思补丁”，并将基于反思条件的教师评分用于学生的 on-policy rollout，作为密集的 token 级训练信号。该过程有效地将稀疏的终端监督转化为密集的 token 级学习信号，而无需外部评论器、单独的奖励模型或更大的教师模型。我们证明，SRPO 在数学推理和长程智能体基准上以极高的数据效率达到了最先进性能。使用 Qwen3-8B 基础模型，SRPO 仅使用规模化监督微调所需训练 FLOPs 的 8%（0.08x），就在 AIME'24 上达到 73.3%，同时显著提升了 WebShop（64.7%）、ALFWorld（76.8%）和 SWE-Bench-Lite（31.2%）上的成功率。代码可在 https://github.com/Galleons2029/SRPO 获取。
推理诱发的失准（Reasoning-Induced Misalignment）指的是，在不含有害内容的推理数据（包括数学、代码以及带有思维链轨迹的问题求解）上进行微调，可能诱发 LLM 的有害行为，从而对 LLM 推理的安全性构成严重挑战。跨架构、跨规模和跨数据集的检查表明，RIM 并不总是会出现。以往工作将 RIM 归因于神经元层面的纠缠，但没有识别出这种纠缠背后表示空间的几何结构，也没有提出训练时的修复方法。我们同时提供了这两点：对 RIM 的表示空间分析，以及 Safety-Direction Penalty（SDP），它会在推理微调期间惩罚沿学习到的安全方向的移动。该分析提取了两个激活空间方向，一个编码推理能力，另一个编码安全行为。这些方向相互耦合：提升推理能力的微调会改变安全表示，而产生更大变化的提示会表现出更大的安全性下降。CKA 距离比率和探针定位了这种变化最相关的安全决策层。这些发现指导了 SDP 的设计：这种耦合促使我们惩罚沿安全方向的位移，而层定位则确定了初始作用范围。当初始范围在受惩罚层之外留下补偿性变化时，相同的诊断方法可指导迭代扩展。在 Qwen2.5-3B 和 7B 上，SDP 在保持基准推理性能的同时恢复了安全性。
The continual evolution of malware variants necessitates detection systems that can adapt to new threats without retraining from scratch. However, continually updating models on new data often leads to catastrophic forgetting, where previously learned knowledge is overwritten. While continual learning has been increasingly explored for malware detection, the specific setting of Few-Shot Class-Incremental Learning (FSCIL), where new malware classes must be learned from only a small number of labeled examples, remains comparatively underexplored. Therefore, this work investigates the FSCIL setting for malware classification. To address the stability-plasticity dilemma, we propose a hybrid framework that leverages a Self-Supervised Learning (SSL) backbone initialized through domain-specific pre-training on malware packets. Our method incorporates Low-Rank Adaptation (LoRA) to efficiently adapt the model during the base session while freezing the core backbone to preserve previously learned representations, alongside a prototype-based classification head for incremental sessions to establish robust decision boundaries from limited samples. Extensive experiments across several datasets demonstrate that our approach consistently outperforms prior malware FSCIL baselines and achieves state-of-the-art performance.
Recent advances in continuous diffusion and flow-based language models (LMs) have achieved performance competitive with discrete LMs. However, existing continuous frameworks still rely on decoders supervised with cross entropy (CE) because the flow trajectories are not guaranteed to terminate at valid token embeddings. Motivated by this limitation, we introduce \textbf{ConvergeFlow}, an embedding-space flow-based LM, which constrains the data predictor to the convex hull of token embeddings and trains it solely with the mean squared error objective induced by flow matching. Under suitable regularity conditions, we prove that the resulting flow converges to valid token embeddings despite errors in the data predictor, enabling direct token prediction without a CE-supervised decoder. We further develop three sampling mechanisms for controlling the trade-off between the generative perplexity and entropy. Experiments on OpenWebText demonstrate that ConvergeFlow achieves performance competitive with existing continuous and discrete diffusion LMs. These findings demonstrate the potential of the flow-based paradigm for language modeling. Our code is available at https://github.com/Na-Li66/ConvergeFlow.
语言模型是顺序处理器，但长时程智能体能力需要模型权重和活动上下文之外的外部信息与计算。Prime Agent 是一个用于长时程评估和编码智能体工作流的开源框架。一个持久化的 IPython REPL 遵循 Recursive Language Model 抽象，用于程序化上下文处理和测试时计算，而 Continual Harness 则在多条轨迹之间保留历史、记忆、技能、提示词和子智能体规范。递归子智能体通过直接的智能体间通信进行协作，Agents View 让人类能够检查和管理由守护进程支持的会话。Prime Agent 将执行、恢复、验证和资源核算标准化，同时将策略构建留给模型。这种低摩擦、富有表达力的膜层可防止框架故障变成模型故障，并将测量推向模型真实的最大底层能力。Prime Agent 将 ARC-AGI-3 RHAE Best@1 从 30% 提升到 95.5%，并在长上下文编码、GPU 内核生成、模拟器构建以及自主 nanoGPT 速通等任务中达到或超过原生框架和流行框架。在 Factorio 上，我们发现精炼使持续技术进展成为可能，而专用子智能体支持并行化工作。代码可在 https://github.com/PrimeIntellect-ai/prime-agent 获取。
Ballistocardiography (BCG) is promising for unobtrusive long-term blood pressure (BP) monitoring in laboratory settings, but traditional BCG signals are vulnerable to the variations in body-bed interaction with shifted fiducial points in temporal or amplitude axis, and BP varies with personal hemodynamic changes, causing misaligned representations that affect model generalizability and robustness. In this work, we propose a non-invasive BP estimation framework, Phy-BP, based on triaxial bodyseismography (BSG) as an extension of BCG. Firstly, an adaptive quality-control algorithm is designed to select BSG segments enriched with cardiogenic components by jointly considering neighboring beat patterns and universal cardiogenic templates. Furthermore, a physical model is established to describe 3D wave propagation in the body-bed system and is subsequently embedded into the deep learning model to characterize the intrinsic coupling among triaxial BSG signals driven by a single cardiogenic excitation. Thus, multi-axis features are aligned during model training, improving robustness against distortions in real scenarios. Experiments on a 162-hour hospital dataset collected from 21 subjects reveal that the proposed Phy-BP can dynamically filter out low-quality measurements, and the deep learning model training is constrained by physical consistency across different axes to provide faithful BP monitoring, especially when training samples are limited.
用于大型语言模型的基于分组的强化学习方法（如 GRPO）通过为每个提示采样多个响应来避免训练 Critic。可靠的 Critic 则可以从一个响应中估计 token 级优势，但标准的基于 Critic 的训练方案通常不稳定。我们研究了这种不稳定性，并开发了 \textbf{Best-Practice Critic Optimization (BPCO)}，这是一种结合了 DPPO、限制在奖励范围内的价值预测、Monte Carlo 价值目标、未归一化的策略优势，以及长度自适应广义优势估计的方案。由于 Critic 仅在训练期间使用，BPCO 还可以让它以奖励定义信息为条件，例如参考答案或评分细则，而这些信息对策略是隐藏的。受控实验隔离了每个设计选择的影响。在数学推理任务中，模型规模从 1.5B 参数到 30B-A3B 混合专家模型，BPCO 始终改进了一个强大的基于 Critic 的基线，并且在每个提示仅采样一个响应的情况下，达到或超过了一个基于分组的基线。同一方案还改善了基于评分细则的奖励下的学习。这些结果表明，精心设计的 Critic 为组相对优势估计提供了可靠的替代方案。代码可在 https://github.com/QPHutu/golden_critic 获取
文章 URL： https://github.com/microsoft/agent-lightning/releases/tag/v1.0.1 评论 URL： https://news.ycombinator.com/item?id=49423077 得分：18 评论数：1
PicoMQ is a Rust server for Durable Streams, built on Object Store. Cheap, URL-addressable, granular streams (create/append/read/long-poll/SSE), with Pico Protocol or Durable Streams Protocol as the facade. S3Stream is the stream storage primitive, used in AutoMQ, shipped as a Rust library. Coordination is a command log in Postgres. Comments URL: https://news.ycombinator.com/item?id=49421806 Points: 40 # Comments: 0
LLMs have evolved from language generators to autonomous agents capable of complex, long-horizon tasks. This evolution has produced paradigms including Prompt Engineering to elicit model capabilities, Context Engineering to manage information access, Harness Engineering to organize external tools and resources, and Loop Engineering to support continual reflection and self-improvement. Yet as tasks grow more complex, individual intelligence faces a fundamental limit: many tasks require heterogeneous expertise, interdependent subtasks, parallel execution, independent verification, and persistent state, exceeding any single agent's organizational capacity. Augmenting one agent's capabilities or context cannot resolve this architectural mismatch; intelligence must instead be distributed across specialized agents and organized at the system level. We call this System Intelligence: an agent system's ability to organize and coordinate multiple intelligent components into a coherent, adaptive whole pursuing a shared objective. Achieving it requires more than adding agents; it demands explicit structures to organize work, coordinate heterogeneous agents, and maintain evolving execution states. We introduce Graph Engineering, an emerging paradigm for next-generation agent systems. Unlike prior paradigms that mainly optimize individual interactions or agent-level behavior, Graph Engineering constructs explicit, dynamic, evolving graph structures representing tasks, agents, and system states. These abstractions provide a unified foundation for organizing complex objectives, orchestrating heterogeneous agents, modeling system dynamics, and enabling scalable agent evolution. We systematically review the principles, methodologies, and applications of Graph Engineering for LLM agents. Related papers, open-source data, and projects are collected at https://github.com/DEEP-JLU/Awesome-Graph-Engineering.
High-performance GPU kernels underpin modern deep learning and scientific computing. As workloads become increasingly diverse and GPU hardware evolves rapidly, developing efficient methods for automated GPU kernel generation and optimization has become increasingly important. Existing LLM-based methods typically optimize within a fixed implementation space, limiting either optimization flexibility or search efficiency. We propose \textsc{HIERA}, a hierarchical search-space planning framework for GPU kernel optimization. \textsc{HIERA} constructs contract-augmented task specifications, selects an appropriate implementation space across PyTorch operators, CUDA libraries, and custom CUDA kernels, and uses profiling feedback and expert knowledge to guide structured iterative refinement. Experiments on KernelBench across multiple various workload levels and base LLMs show that \textsc{HIERA} delivers stronger overall implementation validity, sample efficiency, and optimization performance than existing training-free methods, while remaining competitive with the training-based CUDA-L1 without additional model training. A case study on a specialized stencil operator from scientific computing further achieves a \(1.53\times\) speedup over cuDNN, demonstrating the potentiality of the general framework beyond standard machine-learning workloads.
Attention masks are relation-level controls: they specify which query--source pairs may interact. They do not provide a representation-carried token state that is non-participating at the attention boundary. We assign each token hidden carrier \(h_i\) an active-presence coefficient \(p_i=\lVert h_i\rVert^2/(τ+\lVert h_i\rVert^2)\). The same coefficient has two roles: it gates information emitted by token \(i\), and it determines the mass with which token \(i\) enters computations shared with other tokens. OAttention is the support-coupled attention realization of this rule. It gates the receiver output by \(p_i\) and weights source \(j\) by \(p_j\) in both the attention numerator and partition, while retaining the standard score, visibility relation, exponential competition, and value aggregation. This makes the zero-vector token a zero element and yields exact null-receiver, null-source insertion, self-attention insertion, and empty-support properties. The same token-level presence gives local O-components (OFFN, ONorm, and OInject), presence-weighted OStandardize, the O-Closure law \(M(H\oplus0)=M(H)\oplus0\), and an OTransformer by residual and compositional closure. The canonical operator is checked by contract tests and a GPU evaluation. In a zero-fine-tuning retrofit of a cloned pretrained TabPFN v3 regressor, calibrated hidden-carrier OAttention and Full-O variants change mean RMSE by $+0.088\%$ and $+0.177\%$, respectively, over 18 matched dataset--seed cases. A two-block ablation shows that OAttention alone does not preserve a NULL state through ordinary host components, whereas the OTransformer path does. These are scoped tests of exactness, active-path compatibility, and compositional necessity; they do not establish universal no-loss, arbitrary-host closure, learned attraction to the origin, or a general semantics for missing values.
Safe and efficient shape-aware navigation in heterogeneous crowds and robot fleets remains challenging. Traditional approaches often assume homogeneous robots, sparse workspaces, simplified geometry, offline computation, or handcrafted parameters to make the problem tractable, which limits their deployment in dense crowd scenarios. Toward this end, we propose Shape-Aware Reinforcement Learned Model Predictive Control (SRL-MPC), a method for safe, efficient, and adaptive navigation in crowds with heterogeneous shapes without geometry simplification. To encode shape-aware safety, we formulate high-order control barrier function (HOCBF) constraints from geometric separation features (GSFs) based on support function transformation. A reinforcement learning (RL) framework then learns a neural policy that reads GSFs and outputs real-time MPC parameter updates, enabling the MPC solver to adapt to neighboring crowd geometries. The key advantage of SRL-MPC is that it preserves the safety structure and generalizability of MPC while integrating the adaptability and intelligence of RL. Experiments in randomized crowd scenarios with arbitrary shaped robot fleets demonstrate the effectiveness, scalability, and robustness of SRL-MPC. The results show that SRL-MPC substantially outperforms representative baselines in safety and adaptability. Project website: https://hanruihua.github.io/srl_mpc_project/
Automatic speech quality assessment aims to predict Mean Opinion Scores (MOS) consistent with human subjective perception and is essential for evaluating speech generation, enhancement, and communication systems. For speech signals, especially synthetic speech, distortions often occur locally, and overall perceptual quality is usually dominated by a small number of perceptually salient distortion regions. However, most existing methods are primarily optimized with utterance-level MOS, which provides only coarse-grained supervision and offer no explicit indication of where perceptually important distortions occur. To address this limitation, we introduce explicit distortion localization as auxiliary knowledge for speech quality assessment. We construct the first partially distorted speech dataset with frame-level distortion annotations and train a localization model to generate distortion cues. Building on these cues, we propose DAMOS, a distortion-aware speech quality assessment framework that integrates localization information into the MOS prediction pipeline. Experiments on multiple public benchmarks demonstrate that DAMOS consistently outperforms existing methods and exhibits strong cross-dataset generalization, validating the effectiveness of explicit distortion localization for speech quality assessment.
Objective: Cochlear implants (CIs) are bionic prostheses that restores hearing via electrical stimulation of the auditory nerve. Hybrid CIs, which use electroacoustic stimulation (EAS), combine residual low-frequency acoustic hearing with CI electrical stimulation. Intracochlear fibrosis, which forms in response to the presence of the implant, may impede residual hearing function and gradually reduce the efficacy of EAS. It is therefore a translational objective to study the formation of cochlear fibrosis in rodents, with the goal of reducing fibrotic burden and improving outcomes for CI patients. Methods: We generate and annotate a novel dataset of optical coherence tomography (OCT) images from chronically implanted guinea pigs as part of an ongoing study focused on implant induced fibrosis. Objectively assessing fibrotic burden in this model, with high resolution and repeatability, presents an obvious use case for computer vision methods. Results: We present the results of several state-of-the-art semantic segmentation models and compare their efficacy for identifying cochlear fibrosis and other relevant annotations, using a new library of manually segmented OCT images. Conclusions: We find that the best performance is achieved by using a modified version of the well-known UNET architecture (which we term 2D-OCT-UNET) that operates on the upscaled OCT input resolution. Significance: For the first time, we have successfully applied computer vision techniques to an OCT dataset of implanted cochleae with fibrosis. Using this deep learning model, the cochlear fibrotic burden calculation can be reliably carried out as we verify in our experimental section. The dataset and the project code are available at: https://github.com/juliadietlmeier/CF-OCT-segmentation
This paper investigates cross-agent specification portability using Oracle-to-PostgreSQL migration as a controlled software transformation task. The study combines two experimental stages. First, a specification-first migration pipeline was evaluated on 1,006 PL/SQL files, of which 623 were successfully regenerated and 380 generated scripts executed successfully in PostgreSQL 16. Second, cross-agent experiments were conducted on a dataset of 1,802 Oracle scripts with corresponding PostgreSQL implementations using Amazon Kiro, Google Gemini, and GitHub Copilot, with Claude Code and Cursor included in the initial single-agent evaluation. Native and foreign specifications were assessed using Token F1, exact match, SQL syntax validity, AST exact match, AST mean similarity, and immediate runnability. The results show that specification size alone does not predict implementation quality and that cross-agent transfer can produce substantial agent-dependent degradation. The strongest replicated case occurred when Gemini directly consumed a Kiro-origin specification, producing a Token F1 of 0.035, SQL syntax validity of 2.33%, and AST mean similarity of 0.015. Rewriting substantially improved Gemini in the tested configuration, compression did not provide a universal benefit, and retrieval-augmented ingestion was the only common strategy represented on the per-agent Pareto frontiers of both Gemini and Copilot. The findings suggest that specifications in heterogeneous SDD workflows should not automatically be treated as agent-neutral artifacts and motivate explicit consideration of specification portability, agent-specific interpretation, and retrieval-based access in multi-agent software engineering.
持久记忆会让虚假信息变得持久：一旦某条虚假陈述被存储，它就可能在未来与之匹配的会话中被检索出来。我们使用一次性生成的、措辞直白的虚假断言来衡量这种失效模式的代价，不进行指令、触发器或检索器优化。对 LongMemEval 语料库中 1.2% 的内容进行投毒，会使准确率从 0.850 降至 0.300。一个四阶段的写入时筛查流水线在间接提示注入上达到 0.832 的召回率，同时会将 1.5% 含触发词的良性文本标记出来，但它拒绝了 360 条被投毒记忆中的 0 条。我们认为，这暴露了仅基于内容的筛查边界：区分虚假断言和真实断言通常需要文本本身之外的外部依据。 随后，我们评估了按来源加权的检索。已发布的权重在统计上与没有防御无法区分（p=0.80），而更强的权重只有通过排除不受信任内容才能恢复效用。在一个混合来源语料库中，如果不受信任内容大多是良性的，准确率会从 0.3167 上升到 0.7000；当承载答案的证据本身来自不受信任来源时，证据召回率会降至零，准确率降至 0.0417。在测得的相似度条件下，加性来源项不存在可用的设置：足以抵御查询形态投毒的权重，也足以压制合法的不受信任证据。因此，我们主张在检索阶段采用有界占用约束，而不是加性来源惩罚，并发布相关测试框架、语料库和汇总运行报告。
The Traveling Salesman Problem (TSP) is one of the most extensively studied NP-hard optimization problems. Genetic Algorithm (GA)-based solvers, such as the Edge Assembly Crossover (EAX), achieve state-of-the-art performance on many benchmark instances. However, the scalability of these approaches in massively parallel architectures remains limited because crossover operations involve irregular memory access patterns, graph traversals, and sequential dependencies. Existing GPU-based TSP solvers primarily exploit population-level parallelism and are limited to relatively small problem sizes. This work presents a fine-grain GPU implementation of the partition phase of the Generalized Partition Crossover (GPX) operator for large-scale TSP instances. The proposed approach reformulates GPX partitioning as a graph-parallel problem using coalesced memory layouts, ghost-node transformations, and connected-component analysis. The im- plementation parallelizes the union of parent tours, the splitting of degree- four vertices, the deletion of common edges, and the identification of recombining components using CUDA. Experimental results on instances ranging from 10,000 to 2 million cities demonstrate substantial acceleration over a naive sequential CPU imple- mentation. The proposed GPU partitioning achieves speedups between 48x and 625x while significantly reducing memory overhead. The re- sults demonstrate that operator-level parallelism can substantially im- prove the scalability of GA-based TSP solvers on modern many-core architectures.
Question answering (QA) over long, connected documents remains challenging because relevant evidence may span multiple entities and their relationships. Existing retrieval-augmented generation (RAG) methods typically index documents as raw chunks and retrieve them through embedding similarity. Their performance degrades when chunk boundaries separate entities from supporting evidence or when a question requires multi-hop reasoning across the corpus. We propose EnSI-RAG (Entity-Structure-Indexed Retrieval-Augmented Generation), a framework that constructs a query-independent, entity-centered index. Each record (e, t, k, v) represents an entity e, its type t, a semantic category k in {property, relation, aspect}, and a value v, while retaining links to the original source passages. At query time, these records serve as retrieval handles, and an LLM synthesizes the retrieved passages into the final answer. This design separates evidence localization from answer synthesis while preserving traceable source evidence. Across Loong and Oolong, EnSI-RAG achieves an average accuracy of 78.24. Relative to the published baseline scores used as references, this is 6.62 points higher, suggesting its effectiveness across these settings. The code is available at https://github.com/RamonMeng/EnSI-RAG.
Reinforcement Learning (RL) has demonstrated significant gains in image captioning, yet it is still limited in encouraging Large Vision-Language Models (LVLMs) to explore novel reasoning strategies. This limitation leads to a performance gap between RL and Supervised Fine-Tuning (SFT). In this paper, we argue that multi-modal retrieval can serve as an effective reasoning signal for caption refinement. Based on this insight, we present the Retrieval-Guided Refinement for Image Captioning (Re$^3$Cap), a retrieval-guided reasoning strategy that enhances image captioning without requiring additional annotations. Instantiated by Caption Refinement Suggester (CRS) and Caption Quality Assessor (CQA), this strategy identifies hallucinations and omissions in image captions, leading to more accurate and detailed descriptions. Extensive experiments demonstrate the superiority of our method in image captioning, even compared with Supervised Fine-Tuning. Especially, Re$^3$Cap outperforms GRPO with an average improvement of 8.64% in relation reasoning on the COCO-LN500 benchmark.
The European Union (EU) has emerged as a leading regulatory body in the development of sustainability and privacy regulations. While new regulation requirements vary, many include a documentation artifact to ensure compliance. Notably, the Ecodesign for Sustainable Products Regulation (ESPR) introduces Digital Product Passports (DPPs) for life cycle transparency, while the General Data Protection Regulation (GDPR) mandates Data Protection Impact Assessments (DPIAs) to mitigate privacy risks. Creating these compliance artifacts, however, is challenging. Industrial data, which often exists in heterogeneous formats and is scattered across company and supplier systems, is required for DPPs and can be difficult to extract into compliant DPP formatting. Furthermore, DPIA documents require interdisciplinary expertise and follow no standardized format, making development difficult for novel systems. To address the particular complexity of compliance artifact creation for both regulations, researchers have proposed the use of LLMs in the generation process; however, the impact of the aforementioned problems on the output of these systems is largely unaddressed. This work investigates the existing research gap by exploring how data extraction instructions and regulatory vagueness impact the quality and consistency of LLM-produced compliance artifacts. The resulting artifacts are evaluated by benchmarking different models against manually created ground-truth schemas. The results reveal that less strict guidelines, such as DPIA formatting, require higher context prompts to maintain consistency and completeness. Stricter guidelines, such as formatting for Digital Battery Passports (DBP), result in consistent results regardless of prompt context, but may lead to more hallucinations in the output
We formalize the Steiner Traveling Salesman Problem (Steiner-TSP) on Graphs of Convex Sets (GCS), which seeks a minimum-cost closed trajectory through required convex sets while allowing optional transit vertices and revisits. To explore the resulting infinite solution space, we propose a unified branch-and-bound search over rooted walk prefixes. Additive lower-bound-graph costs bound committed prefixes, while a cut-separated connected-flow relaxation lower-bounds the residual cost of visiting every remaining target and returning to the root. Under a uniform positive-cost assumption, best-first traversal terminates after finitely many expansions on every feasible instance without an initial incumbent, whereas depth-first traversal does so once a finite incumbent is available. For a user-specified factor $ε\geq1$, a global lower bound certifies that either strategy's incumbent cost is at most $ε$ times the global optimum. We further demonstrate joint sensing-mode, visitation-order, and continuous-trajectory selection for a mobile-manipulator inspection task, including action precedences expressed in linear temporal logic over finite traces (LTL$_f$). Both traversal strategies find feasible solutions on all benchmark instances within 30s with mean certified optimality gaps of 28.1% and 29.7%, respectively, whereas two recent baselines succeed on only about half of the instances
Deep-learning models of anatomy can be numerically plausible yet anatomically impossible, and they generalize poorly when data are scarce. We introduce Anatomy-Informed Neural Networks (AINN), in which soft anatomic priors enter as penalty terms in the loss (e.g., a branching penalty that treats a renal transplant artery off the iliac instead of the aorta as unexpected rather than impossible), in direct analogy to a physics-informed neural network, and hard anatomic priors (e.g., continuity of the vessel) are built into the architecture and state representation, making such invalid predictions impossible by construction wherever the prior admits architectural enforcement. We develop it on a clinical test case with limited data: how the aortoiliac tree deforms when a stiff wire is introduced endoluminally. This is important to contemporary aortic surgery and will matter to autonomous endovascular navigation. We lift the vessel centerline and the wire path from R^3 to curves of frames in the Lie group SE(3), and couple a Cosserat-rod wire to a tortuosity-modulated, anatomically anchored vessel through a unilateral lumen-contact inequality. The prediction is a constrained minimizer of the coupled elastic energy, with contact forces as its Lagrange multipliers. Supervision is a Wasserstein-2 optimal-transport loss between the predicted projection through the C-arm geometry and the observed angiogram, so a 2D angiogram can train a 3D prediction. The kinematics, loss and projection are verified against known ground truth; the mechanics solver only against its own optimality conditions, and predicted displacement is not yet mesh-converged. Here, no network is trained. Future work will transfer this in silico model to real CT scans and test whether it improves predictive accuracy and reduces the training data required.
https://matklad.github.io/2026/08/21/rust-glancer.html Comments URL: https://news.ycombinator.com/item?id=49393052 Points: 21 # Comments: 12
Hi HN- 我是 Pablo，Proliferate（YC S25）的创始人！ Proliferate ( https://github.com/proliferate-ai/proliferate ) 是一个开源、可自托管的 AI IDE，让你可以在一个地方使用 Claude Code、Codex、OpenCode、Cursor 和 Grok 进行工作并自动化任务。 这里有一个 2 分钟的快速演示，展示我们如何用 Proliferate 来构建 Proliferate： https://www.youtube.com/watch?v=tGNX0oaWmBY 在我的团队接入 OpenAI Codex 之后，我开始着手构建 Proliferate。几天之内，我们就把它用到了方方面面：用 computer use 代替我们自己浏览网站，让 Codex 协调其他 agent，以及为重复性工作设置自动化。说实话，我们几乎从不需要离开桌面应用就能把工作完成。 如果我团队的体验具有一定代表性，那么一个类似 Codex 的应用（一个具有精美 UI 的横向 agent）将会成为每家公司完成工作的主要界面。这完全符合 OpenAI 的使命，即把 Codex 打造成 everything app（见： https://news.ycombinator.com/item?id=47796469 ）。 但随着我们开始把自动化推进到更接近业务核心的工作，我希望能与来自所有实验室的 agent 协作，包括开放权重模型，而不会越来越依赖 OpenAI。 而这正是 Proliferate 的用途！它是开源、可自托管的 Codex，既能在构建向 Codex 广度迈进的同时，保留你在不同 agent 和模型提供方之间的选择自由。 如今 Proliferate 支持： * 使用 Claude Code、Codex、OpenCode、Cursor 和 Grok，并可采用它们原生的推理与配置选项，包括 Bedrock、Azure 和自托管推理。 * agent 之间的通信与管理：父 agent 可以生成并与另一个受支持的 agent 作为子 agent 进行通信（我个人喜欢让 Fable 委派给 Codex，同时让 OpenCode 模型审查 PR）。 * 构建 workflows——这是我最兴奋的功能之一。这些是可复用的 agent 会话链和人工审批关卡，每一步都可选择 harness 和模型，并在它们之间传递文档。我用它来自动化代码审查、QA，以及我的 PR 构建流程。 Proliferate 的全部代码在 AGPL-3.0 下 100% 开源。现在肯定还有一些粗糙的地方，但我们推进得很快，也非常欢迎任何反馈！ Comments URL: https://news.ycombinator.com/item?id=49390739 Points: 11 # Comments: 5
文章链接： https://github.com/alibaba/anolisa/blob/main/docs/user-guide/en/agent-observability/agentsight.md 评论链接： https://news.ycombinator.com/item?id=49389493 积分：11 # 评论：0
It is a fork of https://github.com/DenisSergeevitch/desktop-fly , but with an important update. Now the fly can pick up the scent of the codebase with its neurons and fly straight to the source code of your B2B AI SaaS startup. It has learned to scan its surroundings for agent markers: AGENTS.md, CLAUDE.md, .cursor/rules, .kiro/steering, and forty others. Anything on the screen that points to these markers becomes a source of the scent - an editor window with an open project, a line in Finder, or a desktop icon. An open project reeks the strongest, while a closed icon or nested folders give off a fainter odor. Comments URL: https://news.ycombinator.com/item?id=49389466 Points: 3 # Comments: 0
人形机器人近期在真实世界球类运动中展现出有前景的能力。然而，在保持强任务性能的同时实现专业运动风格仍然具有挑战性。在这项工作中，我们提出 AdaPT，一种自适应运动规划与跟踪框架，可直接从转播视频中学习专业网球发球和对打风格。这种分层设计的动机来自一个关键洞察：规划器生成具有风格的运动学动作，而跟踪器以尽量少干扰规划的方式执行这些动作。尽管其在仿真中效果显著，但仍出现了明显的仿真到现实差距：真实机器人上的跟踪性能不可避免地下降，而这种下降会被自回归规划部分忽略，并进一步受到有噪声感知的叠加影响。为解决这些问题，我们的适应机制通过学习跟踪随机化的执行速度来提高跟踪鲁棒性，同时让规划器以学习得到的运动速度适配器为条件，以缓解累积误差。在 Unitree G1 上进行的真实世界实验展示了我们的适应机制在弥合仿真到现实差距方面的有效性。我们还将 AdaPT 策略部署到全尺寸 Dobot Atom 人形机器人（1.7m）上，并展示了无需动作捕捉的野外发球。除这些结果外，我们的真实世界实验还揭示了面向未来人形机器人球类运动系统的算法与工程洞察。视频和代码可在我们的 \href{https://humanoidtennis.github.io/AdaPT/}{项目网站} 获取。
We present a novel approach to efficient LLM agent harness optimization through adaptive validation task selection. Harness optimization iteratively rewrites the harness code based on validation performance, enabling substantial performance gains without updating the underlying model weights. Existing approaches, however, evaluate a fixed validation set in full at every iteration, incurring substantial evaluation costs even on tasks that become less discriminative as the harness evolves. We propose $\textbf{Task-CoEvolve}$, which co-evolves the validation tasks with the harness by addressing two challenges: selecting informative tasks and estimating full-set performance from partial evaluations. Task-CoEvolve builds on the observation that tasks on which candidate harnesses disagree are more informative for distinguishing among them than tasks that are consistently solved or failed. It uses variance-weighted sampling based on past outcomes to focus evaluation on tasks near the agent's capability frontier, with the sampling distribution adapting as the harness evolves. It then estimates full-set scores from the sampled tasks by accounting for their sampling probabilities, enabling consistent comparisons across iterations despite evaluating different subsets. Experiments on online text classification and Terminal-Bench 2.1 show that Task-CoEvolve consistently outperforms fixed-subset baselines and matches the final performance of full-set search while reducing the number of evaluations during optimization by 80%. Code will be released at https://github.com/Agent4Science-UTokyo/Task-CoEvolve.
Small language models are usually built like large ones and then squeezed onto a CPU afterwards. We did the opposite: we fixed the target first, one user, one token at a time, 4-bit weights, ordinary CPU, and chose the architecture to suit it. The result keeps full attention in only 6 of its 18 blocks. The other 12 use short convolutions whose memory is two timesteps wide no matter how long the conversation gets, so two thirds of the network never re-reads a growing cache. Trained from scratch on 59.9B tokens, the model scores 47.31 on a five-task benchmark against a bar of 42.20 that was fixed before training began. It beats GPT-2 124M, Pythia-160M, OPT-125M and GPT-neo-125M, all trained on three to six times more data, and exceeds MobileLLM-125M's published score despite that model seeing a trillion tokens. Validation bits-per-byte is 0.8685. To check the architecture rather than the training recipe, we trained a conventional all-attention model of the same size on the same data, and wrote down the winning condition before scoring either. The hybrid won the chosen quality metric by 0.81%, matched it on downstream tasks, produced a 6.3% smaller 4-bit file, and decoded 1.76x faster at 2048 tokens of context, 2.08x against an external model of similar size. In every measurement the speed advantage is near zero at an empty context and grows with length, which is what the mechanism predicts and what a merely leaner model would not show. A simple bandwidth calculation predicts only 1.17x, so memory volume alone does not explain the gap. We also report what did not work: an unmitigated 4-bit quality cost, roughly half the convolution channels ending up inert and impossible to remove, and a vocabulary larger than this model size warrants.
Legal AI systems are increasingly used to answer legal questions, yet existing benchmarks assume queries arrive fully specified. In practice, users omit facts that materially determine the legal outcome. We introduce InsufficiencyBench, the first legal benchmark targeting query-side insufficiency: whether a model recognizes when a query lacks legally material information, identifies what is missing, and refrains from premature conclusions. We formalize a taxonomy of eight canonical missing-element categories across three structural failure modes---switch, gating, and fatal prerequisite--- and construct 202 benchmark items (58 base queries, 144 deficient variants) spanning six legal domains and 24 US jurisdictions and annotated by practising attorneys. Evaluating ten frontier models, we find that no model exceeds F2 = 0.46 on missing-element identification and that the median recall is 0.44. Models either hedge indiscriminately or answer silently under fabricated presumptions. No model both identifies and qualifies responses to deficient queries while directly addressing complete ones.
多模态大型语言模型（MLLMs）将语言推理与视觉感知相结合，但它们在显式或此前未见过的规则约束下执行视觉空间规划的能力仍未得到充分探索。这一场景要求模型共同理解空间布局、解释自然语言规则，并据此规划有效动作。为弥补这一空白，我们引入 RuleMaze，这是一个可控基准，其中 MLLMs 必须在遵守不同复杂度自然语言规则的同时穿越迷宫。RuleMaze 通过要求准确感知、规则解释和受约束的动作规划，隔离出规则合规的空间规划能力。为实现可扩展且系统化的规则构建，我们提出 Language-Logic-Function Hybridization，它能够自动生成自然语言规则，并将其转换为逻辑表示和可执行验证器，从而消除手动规则工程。为提升规则遵循能力和泛化能力，我们引入 Disentangled Multimodal Planning（DMP），它通过可解释的推理原语将感知、执行和规则验证分离开来。通过解耦这些组成部分，DMP 有助于系统性地泛化到更复杂且此前未见过的规则，同时提供透明的中间规划轨迹。实验表明，与端到端文本规划基线相比，DMP 显著提高了规则合规性和规划成功率。总体而言，RuleMaze 为研究 MLLMs 中具备基础感知且可解释的基于规则的空间规划建立了一个原则性基准。代码可在 https://github.com/oceanflowlab/RuleMaze 获取。
X-band SAR satellites (8-12 GHz) play a critical role in disaster response, environmental monitoring, and military intelligence. Yet, they lack robust physical-layer authentication (PLA), a security layer orthogonal to cryptographic solutions. Existing PLA systems, typically based on radio-frequency fingerprinting, are often limited to sub-6 GHz frequencies and rely on classical deep learning. However, this approach underfits the IQ phase nonlinearities that distinguish satellite hardware. In this paper, we present QUASAR, to the best of our knowledge the first quantum-classical hybrid architecture that fuses a CNN spectrogram encoder with a variational quantum circuit (VQC) to provide PLA to X-band SAR signals. Our solution enjoys two distinctive features: (i) it is markedly more data-efficient than classical machine learning, requiring only 10% of the training data to match the accuracy of classical baselines -- data collection being notoriously the most time-consuming phase of PLA; and, (ii) at an equal data budget, it improves classification accuracy over those baselines. In detail, we test our solution under three adversarial scenarios: replay, crafted-IQ injection, and space-borne spoofing. QUASAR rejects spoofed transmissions in 89.7%, 94.1%, and 81.3% of attempts, respectively, establishing the first quantum-enhanced physical-layer classifier for satellite constellations. The fully detailed framework and the supporting results, other than being interesting on their own, show a novel research avenue for physical-layer authentication.
Kandelo is an open-source, Wasm-based multi-process kernel that runs POSIX programs in browsers and Node.js. Kandelo is still experimental, but it already runs a substantial range of existing software. Do you have use cases for this? We are trying Kandelo as a new foundation for WordPress Playground which runs server-side WordPress entirely in the browser. Kandelo also looks promising as a sandbox for running agents in the the browser and on the command line. On the side, we've been playing with porting games and desktop environments and even compiling runnable programs within Kandelo. Yet it feels like there are many possibilities we haven't considered. How would you like to use something like this? Demos: Some notes: The demos have been tested in desktop browsers. Unfortunately, YMMV on mobile today. Some of the disk images are large (~50MB) and may take a while to boot initially. Main set, with Shell (bash, vim, nethack, and more), Nginx, PHP, WordPress, and Doom: https://kandelo.dev/20260819-demo/ LÖVE game engine: https://kandelo.dev/20260819-demo-love/ SNKRX running under LÖVE: https://kandelo.dev/20260819-demo-love/?vfs=love-snkrx-abi44... Commander Keen running in DOSBox: https://kandelo.dev/20260819-demo-dos/?demo=keen LXDE desktop PoC: https://kandelo.dev/20260819-demo-lxde/?demo=desktop-lxde Background I wanted an authentic OS-level foundation for running systems software in the browser and started this as a vibe-coded exploration. I figured it would end up being too slow and that we would have to offer many different ways to compromise default POSIX behavior to get anything usable. But after weeks of fighting agents, insisting on genuine POSIX compatibility as the default, I was surprised at how well the system worked without those compromises. Nginx, PHP, Python, Ruby, Redis, and even MariaDB were able to be built using the SDK with minimal hacks. Then we started porting games, having fun, and playing to see how far we could push it. Notes on architecture: There is a central, single-worker kernel, aiming to provide all supportable POSIX syscalls. Each process is a dedicated worker with independent memory. Each process thread is a dedicated worker that shares memory with threads from the same process. Syscalls are done with the process SharedArrayBuffer and the Atomics API. fork() is supported. The system is centered around virtual file system (VFS) images, and the VFS can contain lazy references to programs that may or may not be used. Vim is such a reference in the shell demo. On GitHub: https://github.com/Automattic/kandelo Comments URL: https://news.ycombinator.com/item?id=49378305 Points: 6 # Comments: 2
Hello everyone. I've been working on this experimental editor called Huzzah. I've been working almost exclusively with coding agents since January of this year, and over the past few months I began to feel utterly exhausted by them. They're great, but I'm finding it more and more tedious to write full sentences for every change I want. Not only that, but it seems there's a complexity limit for codebases - beyond a certain point the agent begins confusing itself. I'd like to go back to writing code, but I don't want to go all the way back to fully manual coding. So I've come up with this interaction paradigm where you: 1. write pseudocode in whatever way makes the most sense to you 2. on save, the editor synchronizes your work to real source code 3. the pseudocode is persisted alongside the generated code, making your prompt effectively a stored record of intent. It may not work for every use case, but in my initial playthroughs I've found it very enjoyable. Right now it's just a proof of concept - installation instructions are here in the readme: https://github.com/danielvaughn/hz You can also watch a video of it in action here: https://x.com/danielvaughn/status/2090456808431165715 Cheers! Comments URL: https://news.ycombinator.com/item?id=49378768 Points: 3 # Comments: 0
Since April, I have been working with Sam Flynn ( https://drook.dev ) to make this balloon payload, UpLink. We did a similar launch last year with Hack Club but this was our first independent launch. UpLink was a 491 gram payload testing the insulation properties of 3D printing filaments, while also transmitting 320x240 images over a radio link -- up from the 18x10 images last year! This is a writeup on our engineering process, mistakes made, and learning experiences. It covers: - Custom electronics designed in KiCad - Firmware design - Results from the data we received on the ground - Image transmission - Launch day logistics, and where things went wrong All hardware, software, firmware, and CAD is available on GitHub: https://github.com/radeeyate/UpLink , licensed + certified as open source hardware: https://certification.oshwa.org/us002826.html If you just want to see the images received, I put up a gallery here: https://uplink.gallery.radi8.dev/ If you have any questions, comments, or concerns, let me know. I'm happy to answer anything! Comments URL: https://news.ycombinator.com/item?id=49377660 Points: 6 # Comments: 1
Hi HN, we’re Yousef & Nour, founders of Vendo ( https://vendo.run ). Vendo lets users create new features inside the software they already use. A user describes the dashboard, workflow, or small app they need, and Vendo builds it on top of the product’s existing data, API, and interface. Demo: https://www.youtube.com/watch?v=VdpHehY64ls We built Vendo because every SaaS eventually faces the same problem: every customer needs something slightly different. One wants a new report and another needs a workflow that only makes sense for their team. These requests either sit on the roadmap, become one-off engineering work, or force the customer into spreadsheets and external tools. We wanted the user to be able to create the missing feature themselves, without leaving the product. Here is how it works: - npx vendo init reads the product's API surface, theme, routes, and more. These are used so that the apps Vendo creates (1) look on-brand and native and (2) have the ability to read data and perform actions directly through the company's API - When a user asks for a feature, we have a custom Vendo harness that writes a React component with a bunch of Vendo add-ons and guardrails (ex. ability to make calls to the host API + our component library). Every save is compiled, type-checked, run against real API responses, and rendered before the user sees it. We just released a benchmark and write-up here with more info for anyone interested: https://vendo.run/blog/generating-product-ui-measured - We use QuickJS to make sure that anything the agent creates is sandboxed and can't mess with the company's site. Vendo compiles the component and runs it with Preact inside a QuickJS VM with no access to the DOM, network, or clock. The VM returns a UI tree, which the host renders using the product’s registered components. When the user clicks something, QuickJS emits a tool call; the host executes it through Vendo’s guard and passes the result back into the same VM, preserving the screen’s local state. There's a lot of generative UI right now: streaming developer-written components into a chat (Vercel AI SDK, CopilotKit, Thesys), or rendering your app inside someone else's assistant (OpenAI Apps SDK, MCP Apps). We differ on two things. Vendo lives in your product and acts through your API as the signed-in user, so what it makes is durable: real apps users keep, pin, and run on triggers while they're away, and not components that are merely confined to a chat. Plus, it's not capped at putting together a bunch of prebuilt components: the agent can build arbitrary apps, from a quick dashboard out of your own components to real custom code running in a sandbox, and either way data only ever comes from tool calls to your API. Here are some things customers are using Vendo for today: - Letting their users create custom dashboards and reports. These are mainly UI-based and focused on letting the user see the exact graphs and metrics they care about - Letting their customers create recurring automations. A big thing as well that has been used for these automations is the fact that we connect to external connections, so users have been automating many of their inter-tool workflows (ex. an automation that sends a slack alert based off of something in the product) - B2B customers letting their customers customize the product with specific business logic. Often this is simple things like an extra field on a form, or an extra permission, but it is hard for a business to keep up with them otherwise. - Creating and sharing custom dashboards/apps across an organization. Since the apps Vendo creates are durable, they can be shared, reused, and forked (which can’t be done with many of the other in-chat generative UI solutions) We've spent a lot of time thinking about how AI and agents will change the way people consume software. We think the answer is personal(ized) software: you see the UI you need to see, you tell an agent exactly what you need, and the product molds to how you work. The key insights that have enabled the product to work are: - A rule in code always beats a rule in a prompt. - Invent as little syntax as possible. Generation got faster and more reliable when the output looked like what models already know (JSX-shaped markup) instead of a clever custom format. - Deterministic beats model wherever you can get away with it. Theme extraction is pure static analysis, and a remix starts as a copy of your component, no model call. Vendo is completely open-source (Apache-2.0) and can be self-hosted, so feel free to check out all the source code here: https://github.com/runvendo/vendo . Would love you to try it out and give us your feedback: https://docs.vendo.run/ . Or if you’re a company looking to embed Vendo in your product feel free to book a call here: https://cal.com/team/vendo/intro-call Comments URL: https://news.ycombinator.com/item?id=49376038 Points: 6 # Comments: 0
With a fixed deadline and design resources committed elsewhere, Stampli used Codex and ChatGPT Work to compress weeks of launch production into days.
Hey, I'm Ben. I built Zoneless because I was paying so much to use Stripe Connect on my own marketplace. The fees were really, really bad, and it was also limiting in terms of the seller countries I could onboard. To give you an idea, I was paying around $9,000 per month in fees just to run payouts. Using Zoneless, that cost goes down to around $6. I've been using it personally for the past few months, and onboarded 5,000+ sellers and done 3,000+ payouts. 74% of new sellers on my marketplace choose Zoneless over Stripe, which is really interesting. I appreciate crypto and stablecoins are a bit of a touchy subject, but for this use case of sending global payouts cheaply, it's perfect. The project is open source with an Apache 2.0 licence, which means there's the benefit of no lock-in and no risk of your account getting flagged or shut down. It also has an almost identical API and dashboard to Stripe. Would love to hear any feedback you may have in the comments. Comments URL: https://news.ycombinator.com/item?id=49375244 Points: 7 # Comments: 1
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work. Comments URL: https://news.ycombinator.com/item?id=49373456 Points: 17 # Comments: 5
检索再生成流水线通常用于为开放式问题生成深度研究式答案，但仅靠检索并不足够：LLM 必须将嘈杂且碎片化的证据组织成全面且引用充分的答案。我们将这一过程称为证据综合。然而，直接生成往往未充分利用证据、引用对齐不准确，并将多样化信息压缩成浅层摘要，暴露出检索与生成之间的证据综合鸿沟。因此，我们提出 DeepWeaver，这是一种新颖框架，通过维护 Thought Block Chains（TBCs）将嘈杂的检索证据编织成全面答案；TBCs 是一种结构化表示，用于归组主张、重要信息、关键词和支持证据。DeepWeaver 使用从属 TBCs 来检查残余证据、提交 TBC 修订，并在最终生成之前发现新的主张。我们在基于知识库和网络的开放式问答上评估 DeepWeaver，并引入 LoQA，这是一个用于证据综合的高密度基准。在多个 LLM 上，DeepWeaver 在 LoQA 上提升了内容充分性、引用根基性和细节保留能力，同时在 DeepResearch Bench 上实现了更深入的洞见和更高的引用质量。这些结果表明，证据编织是弥合开放式问答中检索与生成之间差距的有效机制。我们的代码可在 https://github.com/KlozeWang/DeepWeaver 获取。
Mechanisms for dynamically converting cyber threat intelligence (CTI) into actionable detection capabilities are necessary due to the rapid evolution of Advanced Persistent Threats (APTs). Sigma rules are an essential part of contemporary threat detection workflows because they offer a platform-independent framework for expressing detection logic that can be converted into particular queries across SIEM systems. Conventional techniques for manually crafting Sigma rules are prone to mistakes, and necessitate extensive knowledge, which restricts their scalability. Although there are open-source and industry-maintained Sigma rule repositories, they often fail to keep pace with emerging threats and require frequent customization to fit diverse operational environments. This emphasizes the necessity of dynamic rule generation that is adapted to evolving attack techniques as well as particular use cases. In this work, we design AUTOSIGMA, an automated solution for transforming unstructured CTI reports into relevant Sigma rules. Rather than relying solely on language models, AUTOSIGMA leverages a structured knowledge base to enrich partial inputs, matches the enriched content against a repository of existing Sigma rules, and then employs an LLM-as-a-Judge mechanism to iteratively validate the rules. By combining knowledge-driven enrichment, template-based rule grounding, and a multi-stage solution, AUTOSIGMA enables accurate, context-aware, and relevant rule generation. Evaluations across multiple real-world APT reports and multiple security blogs demonstrate that AUTOSIGMA outperforms alternative solutions and LLM models in rule validity, rule relevancy, MITRE ATT&CK technique coverage, and robustness to input quality. AUTOSIGMA's Demo: https://youtu.be/iSr6IurQ6BM
Accurate and responsible medical question answering (QA) is important in healthcare, where complex cases require factual knowledge and nuanced reasoning. Existing medical QA systems, typically based on single-agent architectures and static retrieval, often lack adaptability, persistent memory, and structured decision-making. This work introduces an adaptive memory and reflection (AMR) agentic system, a multi-agent framework in which specialized agents use dedicated memory and reflection-based feedback to retrieve relevant prior cases and improve subsequent reasoning. Complexity assessment routes questions through solo, collaborative, or escalated workflows, while consensus and ethical overseer modules support reasoning consolidation and output review. Evaluation on MedQA and MedMCQA demonstrates strong performance compared with several baselines. Ablation studies show that combining agent-specific memory, reflection, and external retrieval yields the strongest performance. These findings highlight the potential of structured memory and feedback for developing more trustworthy medical agents. The source code is publicly available at https://github.com/mm-air/AMR-Agent.
大型视觉语言模型（LVLMs）经常产生幻觉，生成输入图像并不支持的内容。在解码过程中防止此类内容，需要一种针对候选项的度量，用于衡量图像对当前考虑的 token 支持有多强。模型的视觉 token 状态提供了这种证据的天然来源，因为将每个状态通过输出头投影，可以揭示该位置偏向哪些词汇项。这些按位置得到的读出结果不能直接汇聚，因为它们的概率大小在不同视觉位置之间不可比较。词汇排名为汇聚提供了尺度不变的基础，但不同 token 在其典型的基于排名的证据上仍存在系统性差异。我们提出 ReWEIGH，这是一种无需训练的解码干预方法，它汇聚视觉位置上的这些排名，并将每个候选项与从未标注图像中估计出的特定于 token 的参考值进行比较。在推理时，ReWEIGH 在预填充阶段缓存图像证据，并仅对低于其参考值的候选项施加有界惩罚。在四个 7B 骨干模型上，ReWEIGH 将幻觉性物体提及最多减少 21.3%，同时在很大程度上保持或提升描述性和通用性能。在缓存证据后，平均每个 token 增加的延迟为 1.33%，并且这种减少效果可扩展到六个架构家族、最高 32B 参数的模型。
Time series imputation is a crucial area for reliable time series analysis, yet it remains challenging due to the complex temporal dynamics and noise of real-world data. Existing approaches, however, exhibit two limitations: missing and observed values are embedded within the same representation space without explicit structural separation, and continuous diffusion-based methods are trained to predict added noise rather than the original signal. To address these, we propose the Masked Diffusion Time-series Imputation Model (MDTIM), which leverages the training paradigm of masked diffusion model for imputation tasks. The MASK token is structurally orthogonal to valid observations, and the model directly predicts the original values, naturally aligning both the representation and the learning objective with the imputation task. To bridge the gap between discrete masked diffusion and the continuous, ordinal nature of time series, we further introduce Stochastic Discretization, which maps continuous values to ordinal-aware tokens while preserving continuous dynamics. Our experiments on diverse benchmarks confirm that MDTIM achieves superior robustness and scalability, consistently outperforming state-of-the-art deterministic and generative baselines across various missing scenarios.
Improving molecular properties, such as drug-likeness or binding affinity, is a recurring task in early-stage drug discovery. However, molecules optimized in an unconstrained chemical space have limited practical value if they cannot be synthesized. Policy Gradient for Forward Synthesis (PGFS) is a synthesis-aware reinforcement learning method for molecular improvement, but its use of reactant embedding prediction makes reactant selection indirect, which, as we show, limits learning effectiveness. We first develop PGFS+, in which reaction templates and second reactants are represented by trainable embedding lookup tables. Combined with a more effective scoring function and RL algorithm, PGFS+ significantly improves the desired property. However, it exposes a reward-hacking failure mode: a powerful reactant search can map diverse input molecules to the same high-reward magnet molecule, improving the reward while collapsing the output diversity. We therefore introduce PGFS++, a synthesis-aware reinforcement learning framework for input-specific molecular improvement. Given an input molecule, PGFS++ treats it as the start of a forward-synthesis trajectory, applies learned reaction templates with compatible in-stock building blocks, and produces a molecule with improved target properties, an explicit synthesis route, and structural similarity to the input. Experiments on molecular improvement tasks show that PGFS++ improves target properties while preserving high output diversity.
Frontier language models are compared, marketed, and benchmarked on capability -- what their best or average output can achieve. I argue this measures the wrong axis. The models have saturated accuracy: their mean output lands on the target. What now separates one system from another in practice is precision: how tightly concentrated their outputs are around that target across repeated, identical requests. Borrowing the marksman's distinction, capability is where the average shot lands; reliability is the size of the group. I make three claims. First, precision, not capability, is the frontier differentiator between systems, and benchmark culture systematically fails to measure it, reporting central tendency rather than spread. Second, precision is measurable, cheaply and without circularity, by running a fixed suite of deterministically scored tasks many times at fixed temperature and computing the per-task consistency of outcomes -- no model-in-the-loop grader required. Third, the measurement is not merely descriptive but decision-guiding: it separates consistent failures (a tight group off-centre, correctable by the operating discipline of Paper 1 -- a sight adjustment) from scattered failures (a wide group, correctable only by changing the model or its sampling -- a rifle problem). I define a grouping metric, specify a harness, and show how tracking a human-AI pair's grouping over time yields the compounding signal that Paper 1's field study requires. A first real run, since replicated, illustrates both the method and its most important limit: one measured gap was closed completely by a single rule (0/5 -> 5/5), while a suite of tasks authored from the rules themselves found no value, because a frontier model already embodies explicit good practice -- establishing that a discipline's worth is found by measurement on real work, not constructed from its own rulebook.
现代 Intel AI PC 配备了性能出色的集成 GPU 和 NPU，拥有 16 GB 以上的统一内存，并且有相当多时间处于空闲状态。但这些内存仍不足以容纳像 700 亿参数 LLM 这样的大模型。我们表明，少量 AIPC 通过普通网络协同工作，就能提供超出任何单台设备能力范围的模型服务。我们采用流水线并行：将模型按层拆分为每个阶段的分片，每个分片都预编译成 OpenVINO 图，使每台机器运行一个分片，并将激活传递给下一台机器。三项技术使其速度足够快，从而具备实用性。首先，我们恢复了未拆分模型的速度：朴素的逐阶段导出性能远低于整体推理，因为它错过了一项 OpenVINO GPU 优化；而向每个分片中注入 beam_idx Gather 会触发该优化（IndirectKVCache 融合），使分片达到与整体模型相当的性能。其次，我们在有状态 OpenVINO 模型上利用推测解码。第三，流水线通过在各阶段交错处理多个用户的请求来同时服务多个用户，每个请求都携带自己的缓存（微批处理）。综合来看，一个双节点 Llama 3.1 8B INT4 流水线在相同硬件上为两个并发用户提供服务时，吞吐量达到未拆分模型单用户吞吐量的 1.79 倍；在模拟广域网延迟下，这一差距进一步扩大。同样的设计可扩展到单个机群成员无法容纳的 70B 模型：在 Intel Tiber Cloud 上由 Lunar Lake AI PC 组成的四节点部署能够以交互式速度服务单个用户，其输出与不使用推测解码的同一四节点流水线逐 token 解码结果完全一致。代码、原始基准测试日志和复现实验脚本以自包含软件包形式发布于 https://github.com/labscommunity/pipeline-sharded-inference-paper（位于顶层 reproduction/ 目录中）。
This technical report describes our winning submission to the AES AIMLA 2025 Challenge on querying sound effects by vocal imitation. We investigate two complementary fine-tuning strategies: contrastive learning with a frozen, pretrained CED encoder, and joint contrastive-triplet learning with semi-hard negatives using a MobileNetV3 encoder. This report has been updated for posterity to include details released after the challenge.
在策略蒸馏（OPD）使用更强教师提供的密集 token 级指导，在学生自身的响应上训练学生。然而，在长上下文任务中，token 级教师支持可能偏向于局部上看似合理、但遗漏分布在输入各处的证据或违反全局任务约束的响应。相比之下，特定任务的验证器在响应级别评估任务完成情况，并可能返回反映部分成功的分级奖励。我们在来自两个具有代表性的长上下文证据聚合任务的固定响应上诊断了这种不匹配。在更长的输入范围内，轨迹级 OPD 分数与验证器奖励的对齐程度逐渐降低，表明存在教师-验证器分歧。受这一观察启发，我们提出了 Group-Calibrated On-Policy Distillation（GC-OPD）。GC-OPD 在每个 rollout 组内分别对验证器奖励和轨迹级 OPD 分数进行归一化，并使用二者之差作为带符号的教师-验证器分歧残差。基于相对优势的信用分配（RACA）根据 token 的相对 OPD 优势，将这一轨迹级残差分配到各个 token，同时保留原始 OPD 信号。在五个长上下文基准上，使用 GC-OPD 进行后训练将官方 Qwen3-4B 和 Qwen3-8B checkpoint 的五基准平均分分别从 29.08 提升到 40.47，以及从 35.12 提升到 44.65。在相同设置下，普通 OPD 分别达到 39.31 和 43.56。受控消融实验表明，带符号残差比额外的 OPD 派生项或直接加入组归一化验证器奖励更有效，而 RACA 相比均匀 token 分配进一步带来提升。总体而言，这些结果表明，组相对残差校准可以在不丢弃密集 token 级指导的情况下纳入验证器结果。代码可在 https://github.com/SolereZhang/GC-OPD 获取。
I wanted to share a project I’ve been working on called Frugal Tokens. I originally built it because I was curious to see how much all of my sessions cost and how much cache misses affected that spend. I’d noticed people had widely different spend profiles and wanted to better understand what might contribute to that. As I’ve worked on this, the tool has grown to show more usage patterns across all of your sessions. It shows overall usage, estimated working time and overlapping sessions, and where your spend is coming from across models and cache misses. I also have a few session level metrics with percentile breakdowns, along with a list of your sessions and high level info. Clicking into a session opens an explorer where you can see individual model calls and tool inputs and outputs. You can also jump directly to where a cache miss happened. There’s also a rough cost comparison that shows what the recorded session would have cost with another model’s pricing, or for Anthropic, with 5m vs 1h caching. In the future, I’d love to collect more information to see which patterns might make people’s workflows more expensive, e.g. long sessions, high context usage, many turns, etc. The tool requires deno, but is just one command to run once that is installed. The demo provided has some of the data scrubbed, but helps to show what it looks like before running it. Would appreciate any thoughts or ideas https://github.com/dpclark4/frugal-tokens Comments URL: https://news.ycombinator.com/item?id=49364223 Points: 10 # Comments: 1
Hi HN，我们是来自 OneCLI 的 Jonathan 和 Guy。OneCLI 是一个为团队构建的 agent harness，为每位员工提供一个安全的、沙盒化的个人 agent。 你可以用它做这些事： 1. 获得一个沙盒化的 agent，并具备所有 OneCLI 能力，比如只需在聊天中连接你的 GitHub 账户、Gmail、Notion 或 Dropbox。 2. 在聊天本身中进行确定性的 human in the loop 审批，用于那些你需要 100% 控制的事情，比如发送电子邮件或删除 Linear 工单。 3. 在一个地方管理团队策略，并在工作区中的每个 agent 上强制执行。 4. 在团队层面享受全局连接，比如共享 LLM keys 或服务账户。 这里是演示： https://www.youtube.com/watch?v=dlW-44ntpbE 我们是偶然开始做这个的，尽管我们的职业背景都在安全领域。当时我们在做一个名为 ChartDB 的 devtool，这是一个开源 DB 工具。今年 1 月 OpenClaw 火起来后，我们开始用它在 ChartDB 之上编排 agent。我们很快意识到 auth 方面存在一个大问题。agent 需要凭证才能做真正的工作，但把这些 secrets 交给它们并不是最好的主意。它们会把这些 secrets 保存在自己的记忆中，还会以纯文本写入本地文件和会话中。我们也知道 agent 很容易被诱骗交出这些 API keys/secrets。因此，我们需要某种方式来控制 agent，阻止 prompt injection 欺骗它使用自身服务为攻击者谋利。 我们创建了 OneCLI，它最初是一个用 Rust 构建的 AI Agents vault。 我们发现，对 OneCLI 的大部分需求来自 Hermes、OpenClaw 和 NanoClaw 这类面向个人和团队的 autonomous agents。 用户在寻找有用的 agent，让它们为运行它们的人做事，但缺少两个部分：1）管理 secrets 和权限。2）对于团队来说，还需要多人管理。 我们决定转型，把 agent 本身作为面向团队的 harness 提供，为每位员工提供一个 agent。我们看到团队不得不一遍又一遍地搭建自己的 harness，而我们基本上已经把 vault 做成了一个 gateway。于是我们想到，可以开箱即用地提供 agent 管理中缺失的那一块，并将其开源（Apache-2.0，带有一个小型企业例外条款）。 我们坚持开源优先——开源整个平台，而不只是像其他 agent 那样开源其中一小部分，这样公司才能真正看到代码、评估它、信任它，而不是只听我们的一面之词。它们可以在自己的环境中隔离运行，完全自托管或自控，具备生产级质量，而不是某个托管在别处的封闭黑盒。这意味着安全不只是一句承诺，而是它们可以自行验证的东西。结合真正的自治能力和最小权限访问，这使它成为公司可以完全拥有和信任的东西，而不只是采用。 我们也从公司视角而不是个人视角来处理这个问题。我们的解决方案代表每位员工管理 agent，并包裹在由公司管理员通过集中式策略配置的确定性 guardrails 中。 对于 agent engine 本身，我们使用 jcode，它是 agent-loop 的核心。我们发现它改善了体验，让 agent 更聪明、更快速。 它的工作方式如下： 它运行在你控制的基础设施上。完全开源，几分钟内即可 self-host 或部署到 cloud。 agent 永远不会持有真实 secret。它拿到的是一个 placeholder。真实凭证会在 gateway 处、每次请求时、调用被授权后注入。它永远不会进入 agent 的 context、memory 或 logs。 在模型之外执行控制。Prompts 只是建议。由组织管理员定义的策略运行在网络层，在 agent 和 LLM 之外。阻止 endpoints、按 agent 限速、要求审批、按员工限定 scope。由 gateway 决策。agent 无法绕过它。 每个 agent 都有隔离 VM。自己的 memory、自己的 keys、自己的权限。爆炸半径仅限于一个 agent。 Harness 的速度：agent loop 之下的 Rust engine。 完整的身份轨迹。每个 agent 都绑定到一名员工。每次调用都会记录它代表谁执行，以及是哪条策略允许了它。 人们正在用这个平台做的一些事情包括： - 完整管理公司的生命周期，从销售电话开始，到产品侧自动为工程团队开工单，再触发 development agents 交付并发布到生产环境。 - 运营侧，比如在通话后自动清理 CRM、寻找 leads、预订会议并管理后续跟进邮件。 - 我们的一些客户还用这些 agent 完成整个杂货采购，并让它们处理各种家务，比如在线下单。 关于团队：两位创始人都来自网络安全背景。Jonathan 曾在 Axis Security 多年从事 zero trust network access。核心理念是你永远不要信任客户端。你要精确决定一个人能访问什么，并在他们之外、在网络层强制执行，因此客户端试图做什么并不重要。这就是如今每家严肃公司向人类授予访问权限的方式。Guy 是 Argon security 的第 1 位员工，负责 AppSec。 我们很想听听你们对这次转型的看法，也欢迎提交 issues 来帮助改进，让你的 agent 既强大又安全——为团队而设计，而不仅仅是为个人。 Comments URL： https://news.ycombinator.com/item?id=49363710 Points：3 # Comments：0
文章 URL： https://www.modular.com/blog/modcon-announcements 评论 URL： https://news.ycombinator.com/item?id=49358327 点数：9 评论数：1
Hi HN, I've become lazier in my old age and struggle to replay my favorite Sierra games from the 80s and 90s because I keep getting into those situations where I need an item from 3 acts ago, I have no save game handy, and now I gotta make dinner. So I'm building the Lucasartsifier: a static analysis tool that decompiles Sierra resource files, automatically finds those states, automatically generates code to prevent the player from getting into those states, then emits loose patch files that can be placed alongside the original game resources. There's no game-specific code involved; all the logic is generic, though of course Sierra introduces new idioms and mechanics in every game so every new supported game needs a bunch of engine work. So for example in Leisure Suit Larry 2, the patched game prevents you from boarding the cruise ship until you have both the sunscreen and the Grotesque Gulp. Without them you die on the raft 3 play-hours later. So far this works on Leisure Suit Larry 2 (SCI0), King's Quest 4 (SCI0), King's Quest 6 (SCI1.1), and Laura Bow 2 (SCI1.1). I'm currently working on King's Quest 5 (SCI1.0). This is work done with Claude -- I do the design and playtesting and it does the rest :D Any feedback, play testing, and suggestions would be great! Comments URL: https://news.ycombinator.com/item?id=49355607 Points: 25 # Comments: 11
While Multimodal Large Language Models (MLLMs) have made significant strides in visual comprehension, their ability to reason about text-dense, professional documents remains incompletely evaluated. Existing benchmarks emphasize information extraction, require external domain knowledge, or cover professional documents only as one of many settings. They are also largely English- or Chinese-centric, leaving other languages and Russian, in particular, substantially underrepresented. To address these limitations, we introduce BEAR-Bench (Bilingual Enterprise and Academic Reasoning), a self-contained, complex English-and-Russian benchmark comprising 1000 human-annotated questions based on text-rich business and scientific documents. We evaluate 16 proprietary and open-weight MLLMs, including Gemini 3.1 Pro and Qwen3.5-397B, on BEAR-Bench and observe clear headroom even for the strongest systems. Finally, we use the resulting model outputs to compare existing hallucination detection methods, evaluating not only how often models fail on BEAR-Bench but also how reliably those failures can be identified.
Automatic Target Detection and Recognition (ATD/R) is critical for military decision support and (semi-)autonomous operations. Recent advances in object detection and artificial intelligence (AI) significantly boosted the potential performance of ATD/R. However, the scarcity of publicly available military datasets limits the application of these systems. As a solution, this paper explores the use of publicly available models and civilian datasets to achieve reasonable performance in military contexts. We benchmark several state-of-the-art models, including six iterations of the YOLO series and two variations on the DETR framework, on a newly acquired military relevant dataset. This dataset features military vehicles and challenging circumstances, including various degrees of occlusions and small targets. The out-of-the-box version of each model is validated alongside a version finetuned on the VisDrone dataset. This dataset features small objects, an Air-to-Ground (A2G) perspective and relevant classes, potentially generalizing to our military ATD/R task. We compare the performance of the models using mAP@0.5 and mAP@0.5:0.95, across A2G and Ground-to-Ground (G2G) perspective, target size and model size, giving insight into the real-time capabilities of models. Our main findings are: (1) bigger models outperform smaller models, (2) DETR-based models show promising results compared to the YOLO series,(3) fine-tuning models on an out-of-domain A2G dataset, improves their A2G performance and slightly improves their performance on small objects, but (4) all models still struggle with detecting small objects in an A2G scenario. We conclude that, despite recent advances in object detection, in-domain training is still crucial for creating capable ATD/R systems.
In the Lifelong Multi-Agent Path Finding (L-MAPF) problem, agents must repeatedly move from one destination to another while avoiding obstacles and inter-agent collisions. Widely regarded as one of the highest-performing solutions to this problem is the Rolling-Horizon Collision Resolution (RHCR) framework. However, commensurate with its quality solutions, it incurs a computational cost that limits its applicability to even modest agent counts. In this paper, leveraging theoretical methods from the Locally Interdependent Multi-Agent MDP literature, we first theoretically prove the near-optimality of RHCR in a discounted MDP formulation of the L-MAPF problem. Then, we leverage these results to naturally motivate an extended framework called Group Decentralized RHCR (GD-RHCR) which incorporates a group decentralized structure that partitions agents based on a transitive communication scheme and plans for each partition of agents in parallel. We show that both RHCR and GD-RHCR achieve similar exponentially close to optimal guarantees, establishing a theoretical duality between the time based restrictions performed by vanilla RHCR and the additional space based partitioning performed by GD-RHCR. Lastly, we show that across varying maps, GD-RHCR is able to attain high throughput that scales into higher agent counts while maintaining a significantly lower per plan cost.
Groups routinely complete projects that no single member can plan, execute, or verify alone. We propose a formal model of this phenomenon, Collective Counterfactual Planning (CCP), in which the binding limitation on each agent is neither capability, knowledge, nor observability, but representational geometry: each agent perceives the state, conceives moves, consents to actions, and certifies goal requirements only through a projection onto an agent-specific subspace of a common task space. Four gates jointly determine whether a team can reach a conjunctive goal and legitimately recognize that it has done so: the exogenous implementation coalitions required to perform each action, together with three representational gates -- conception, consent, and task-relative verification qualification. We define the Collective Counterfactual Solvability (CCS) problem, separating geometric feasibility, executable attainment, and validated completion. The results expose a positive-negative duality. Iterated cross-agent relay can unlock a solution that no one-shot pooling of individual plans contains, but any goal requirement depending essentially on the subspace dark to the entire team is unverifiable and therefore not validly completable, even when the trajectory accidentally attains it. Memoryless and audited consent further constrain different objects -- action directions versus cumulative trajectory states -- and neither dominates the other. A four-step exhaustive horizon-bounded solvability scheme is sound and complete under exact representation of the relay closure; restricted implementations remain sound on returned plans but need not be complete. The model gives one geometry for sequential mutual enabling, competent execution of steps whose purpose is invisible to the executor, forced sub-teaming at expertise boundaries, and completion that cannot be validly declared.
Financial time series exhibit non-stationary and heterogeneous statistical properties, making change-point detection challenging because no single unsupervised algorithm performs consistently across assets and market regimes. Conventional workflows consequently depend heavily on expert-driven model selection, feature design, and hyperparameter tuning, limiting their scalability and adaptability. We propose EvoTS-Agent, a validation-guided self-evolving LLM agent for autonomous financial time-series change-point detection. EvoTS-Agent first performs curated exploratory data analysis to characterize dataset properties and initialize candidate detection models. It then evolves executable experiment trajectories through three complementary operators: \textit{Revision} exploits the current best solution, \textit{Alternative Strategy} explores fundamentally different modeling directions when progress stagnates, and \textit{Recombination} synthesizes complementary evidence from high-performing trajectories. Validation feedback guides trajectory evolution throughout the search, enabling the agent to adapt its detection pipeline to the statistical characteristics of each dataset while preserving reliable optimization. Experiments across four benchmark datasets demonstrate that EvoTS-Agent consistently outperforms existing LLM-based agents while maintaining a 100\% execution success rate across all evaluated backbone LLMs.
Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models but relies on costly rollout exploration. Assigning the same exploration budget to samples with different difficulty levels is inefficient: easy samples may receive redundant rollouts, whereas difficult but learnable samples may receive too little exploration. Existing adaptive schedulers address this mismatch through curriculum-based sample selection or non-uniform rollout allocation based on estimated sample difficulty. However, obtaining reliable online difficulty estimates remains challenging: dedicated probing adds substantial generation overhead, whereas history-based estimators face a cold start with no initial observations and stale feedback, and typically ignore relations among samples. To address these limitations, we propose a plug-and-play graph-based online difficulty estimator that shares rollout feedback across related samples and continuously updates their difficulty estimates, mitigating cold start and staleness without dedicated probing. Specifically, we first construct a difficulty-aware sample graph based on semantic and reasoning similarities. Based on this graph, we introduce latent difficulty states and use a Potts prior to encourage neighboring samples to share the same state. We then employ a state-level Beta-Binomial model to aggregate the rollout outcomes associated with each state. Finally, we use an online mean-field variational algorithm to continuously update the latent-state assignments and state-level difficulty as new feedback arrives. Our framework can be integrated into sample-selection and rollout-allocation schedulers, enabling difficulty-adaptive exploration without dedicated probing. Experiments across multiple base models, RL schedulers, and benchmarks demonstrate that our framework achieves better performance.
Large language models can generate executable programs, which makes it possible to search directly over procedural content generators rather than individual levels. We study this approach in Sokoban, Zelda, Dangerous Dave, and Lode Runner. Each run evolves complete Python generators through language-model mutation and crossover. We introduce Continual Abstraction Discovery, or CAD, which extracts reusable primitives from high-fitness programs into a run-specific helper module. A 2x2 experiment crosses CAD with access to a fixed hand-written domain API. The completed data set contains 160 complete runs, with at least ten 50-generation runs in every cell. CAD raises mean final best fitness in all eight domain and API comparisons. Across all CAD runs, learned libraries are adopted by most later programs and repeatedly rediscover validation, reachability, and structural utilities. These results support that discovering reusable primitives improves evolutionary program search for content generators.
Recent research has leveraged Large Language Models (LLMs) to enhance Automated Feature Engineering (AutoFE) through semantic descriptions and trajectory-based prompting. However, there exist two challenges that limit their applicability and scalability in long-horizon optimization: (1) semantic metadata is unavailable in many practical settings, and (2) trajectory accumulation increases the risk of exceeding the context window, while without it, the generation process can become unstable, leading to becoming stuck in the local optima and a high duplicate rate of generated features. To this end, we propose a SHAP-enhanced Implicit-trajectory Generation for Metadata-free AutoFE (SIGMA), a scalable constant-context optimization framework. SIGMA leverages SHAP values to provide task-aware signals for guiding group feature generation instead of semantic information. In addition, we adopt an EXposed-feature Implicit Trajectory (EXIT) approach, where the exposed features in the prompt implicitly represent the trajectory. Empirical results demonstrate that SIGMA achieves performance comparable to the state-of-the-art (SOTA) LLM baselines with a nearly constant prompt length. Notably, EXIT significantly reduces the duplicate ratio of generated features from 37.2% to 6.8%. At the same time, SIGMA matches traditional SOTA performance with only 5.4 features on average, demonstrating substantial efficiency gains in feature utilization.
In the Code World Model paradigm an LLM synthesizes an executable world model that a classical planner searches, and the model is accepted when it reproduces sampled transitions. We ask what that acceptance certifies in continuous control. We define the pipeline's danger as an expected risk and isolate its exact factor: the probability that N i.i.d. gate rollouts all miss a critical event of probability r is exactly (1-r)^N; an independent acceptance sample adds its budget to the exponent. On three hybrid instruments the accepted mode-blind model is exploited: the planner is pinned at the mode boundary at a regret of nearly the whole attainable return. We prove a localization budget, valid at boundary points: models with Lipschitz constant at most L differing by eta at a point disagree above tolerance eps on a region of volume at least kappa((eta-eps)/L)^(d+m); the discontinuous reset modes studied pay no such budget. With real LLM synthesis, GPT-5.x repairs an omitted 1D clamp in 105 of 111 mode-containing draws -- every attempt exact on 50 of 56 instrument-stream blocks (95% CI [0.781, 0.960]). On 2D regions no artifact recovers the rule (0/156); eight targeted interventions leave the failure in place, and positive controls locate it: a located rule is not induced, while given form and location the constants follow exactly. A version-space certificate proves identification is class-relative: at the widest dose the declared fit succeeds in 20/20 blocks and every sample-consistent circle is within tolerance in 18/20. We prove a class of entry rules exactly consistent with every sample yet harmless at play, so identifiability is a measurable property of the instrument. Re-scoring all 1034 artifacts on independent samples confirms acceptance certifies sample consistency and no more: where the gate is provably informative it covers about two percent of the exploited planner's queries.
Ultrasound tongue contour segmentation remains challenging under cross-dataset domain shift, where limited annotations, probe variability, and acquisition noise often degrade model generalization. We present a source-free domain adaptation framework for robust ultrasound tongue segmentation built on a lightweight UltraUNet backbone. Starting from a checkpoint pretrained on only five labeled source images, simulating an underfitted constrained source model, the proposed method adapts to a fully-unlabeled target domain by iteratively refining pseudo-labels, filtering unreliable masks with a contour-based quality-control module, and generating target-style synthetic image-mask pairs through a segmentation-guided conditional GAN. The student model is then trained on a mixture of clean pseudo-labeled target images, noisy pseudo-labels with consistency regularization, and synthetic samples, enabling closed-loop adaptation without access to source data. We evaluate the method on 12 source-target transfer pairs across eight ultrasound tongue imaging datasets, and conduct source-size scaling experiments and ablation studies. Across all comparisons, the proposed framework improves segmentation overlap and contour accuracy over the baselines, including supervised ones. These results suggest that task-specific pseudo-label refinement and synthetic target-style augmentation can substantially improve source-free adaptation for ultrasound tongue imaging.
Improving flight safety with flight data requires not only accurate detection of risk events, but more importantly, clear interpretation of their underlying causes at the level of pilot control behavior. Existing explainable AI techniques, such as feature importance maps, often require considerable domain knowledge to translate them into operationally meaningful explanations. Large Language Models (LLMs), which excel at language reasoning, bring a promising solution to this issue. However, applying LLMs in this domain presents key challenges such as modal inconsistency, limited classification ability, scarcity of task-specific data for fine-tuning, and lack of domain knowledge. To overcome these challenges, we propose FlightLLM, a prior-guided semantic LLM-based approach for interpretable flight safety analysis. Specifically, we first perform feature engineering to address modal inconsistency, combining statistical descriptors with physically meaningful flight indicators. This representation is further processed by a Semantic Discretization module, which converts abstract numerical patterns into qualitative descriptions that are more compatible with language reasoning. In addition, since LLMs are not inherently strong classifiers, CatBoost is incorporated as a statistical expert, and its prediction results are injected into the prompt as prior guidance. A contrastive few-shot learning strategy is further adopted to compensate for limited data. Finally, we design structured prompts to embed aviation-specific knowledge into the inference process. Using hard landing, a representative risk event with complex causal mechanisms, as an anchor point, we evaluate FlightLLM on a dataset of 704 real-world A320 flight samples. Experimental results show that the proposed approach achieves competitive classification performance while generating direct and reasonable explanations for event causes.
Urban traffic congestion reduces productivity and increases travel cost and emissions. Network-wide live travel-time shortest-path rerouting can be highly effective in simulation, but assumes that essentially every on-road vehicle is replanned every decision period. We propose HLSR, a selective hybrid live--forecast vehicle rerouting framework that fuses live edge speeds with short-horizon forecasts under limited intervention scope. Building on dual-threshold congestion detection, calibrated upstream selection, and driver-tailored travel-time prediction, HLSR further introduces approaching-vehicle expansion, travel-time-weighted k-shortest-path generation, and a horizon-dependent hybrid live--forecast segment speed used in multi-cost route allocation.
Hey HN, I've been working on Hansel [1] (an encrypted personal data store you can query with agents), and there wasn't a good way to use the modern macOS Data Protection Keychain. Electron's safeStorage [2] uses the legacy file-based keychain, which allows other apps/agents to query it with the `security` CLI. Not great when you have a dozen agents running in the background! The Data Protection Keychain is nice because it limits access via code-signing access groups and lets you set access rules like Touch ID and/or password. 1: https://hansel.so/ 2. https://www.electronjs.org/docs/latest/api/safe-storage Comments URL: https://news.ycombinator.com/item?id=49349159 Points: 3 # Comments: 0
文章 URL： https://support.claude.com/en/articles/15910845-claude-code-may-august-2026-weekly-limits-promotion 评论 URL： https://news.ycombinator.com/item?id=49348751 积分：44 评论数：15
Working on Mac, Linux, and Windows now. I include a simple GUI to find new models and get things built and set up. It is working quite well across a few models for me. The GitHub README and DESIGN.md files go into detail of the how/why and it's working remarkably well so far. https://github.com/notactuallytreyanastasio/shoehorn Comments URL: https://news.ycombinator.com/item?id=49346135 Points: 5 # Comments: 2
A radiologist reading a model's output faces two problems. The model returns a number and no reason, and any system that turns that number into readable prose can quietly add claims the model never made. MIRROR is a research prototype built to separate those failures. It chains a multi-label classifier, a Grad-CAM localizer that turns each positive finding into a named anatomical region, and a report writer that receives the labels, probabilities, and regions but never the image. Because the language layer cannot see pixels, it cannot assert a finding the classifier did not make. We are precise about what that buys: a MIRROR report's findings are auditable against the probability vector, while the sentences framing them are ordinary generated text, and we show one stating a cardiothoracic ratio the system never measured. One registry holds the taxonomy, anatomy, and phrasing for chest X-ray, brain MRI, and head CT, so adding a modality is a data change; all three are routed and tested, one is trained. On ChestMNIST that classifier reaches macro AUROC 0.729 and ranks better than chance on all 14 labels, at 1.6 to 6.8 times the precision a random ranker would get. Yet at the default 0.5 threshold it emits no positive prediction at all for 11 of them, and its excellent-looking Brier score of 0.045 sits beside the 0.047 earned by a predictor that ignores the image. The discrimination is real; the decisions are not. Under the class imbalance normal in radiology, aggregate metrics flatter models that do nothing, and should be reported against that floor.
Large Language Models (LLMs) have achieved remarkable progress in code generation, yet ensuring correctness in complex, repository-level tasks remains challenging. Existing approaches often use generated tests as static post-hoc validators, which limits their ability to guide implementation and may introduce misleading feedback when the tests themselves are incomplete or incorrect. In this paper, we introduce TDD-Agent, which operationalizes the test-driven development paradigm for code generation. TDD-Agent first prompts the model to generate executable tests, encouraging it to clarify expected behaviors before implementation, and then performs iterative dual-track refinement over both the generated code and tests using execution feedback. We first isolate the effect of test-first reasoning through a prompt variant TDD-prompt on LiveCodeBench, where it consistently improves upon reasoning-based prompting baselines. Building on this finding, we evaluate the full TDD-Agent framework on RepoEval, a repository-level benchmark, and show that it consistently outperforms retrieval-based and agent-based baselines. Additional analyses show that iterative refinement improves not only code correctness but also the effectiveness of the generated tests, yielding higher pass rates, coverage, and mutation scores, suggesting that tests can serve as evolving reasoning artifacts rather than fixed validators. Our source code is available at https://anonymous.4open.science/r/TDD-Agent-Framework-6370/.
尽管用于多参考图像生成的统一多模态模型近期取得了进展，现有基准仍围绕预定义任务类型（例如“主体组合”）组织，这并不适合这种组合式场景，并会导致覆盖碎片化、复杂度不可控且诊断价值有限。我们认识到多样化的多参考任务共享一组通用的原子操作，因此采用面向能力的视角，并形式化定义了四个算子：锚定（$f$）、解耦（$g$）、应用（$\oplus$）和组合（$C$）。这样，任何多参考提示都可以表示为这些算子上的组合公式，其结构复杂度由算子槽位数量量化。基于这一形式化方法，我们构建了 TRACE-Bench，包含约 1,600 个评测案例，槽位数覆盖 1--8，由 631 个公式模板和约 4,000 张参考图像构成，涵盖多样的艺术风格和真实世界主体。公式结构直接驱动一种与算子对齐的评测协议，用于按能力评分，并支持通过诊断树分析进行递归式故障定位。对 9 个领先模型的评测揭示了整体评分无法呈现的洞见：主要瓶颈在于解耦（$g$）和属性绑定（$\oplus$），而非场景级组合（$C$）；即便最佳模型在属性保真度上也仅得 0.74。项目页面：https://amuseum-whr.github.io/TraceBench
大型视觉-语言模型可以识别拥挤场景中的对象和属性，却可能将某个属性分配给错误的同类实例。通用视觉问答准确率会将回答标记为错误，而对象幻觉指标可能认为对象和属性都受到图像支持；二者都无法揭示这种转移。本研究将这一盲点形式化为密集同类属性误绑定（DSCAM），并提出 InstaBind-Lite，这是一个受控基准，可对其进行直接测量。该基准的 524 张图像包含 529 组经过策划的 3-6 个同类实体、1773 个带框实例、有序邻近实例、可区分的类颜色属性，以及四个互补的问题层级，共产生 9580 个可确定性评估的问题。不同于现有协议，源实例标注将不受支持的生成和识别失败与从另一个可见实体复制属性区分开来。绑定特定指标进一步量化转移频率、邻接性、序数距离和干预效果。在五个开源模型和两个商业/API 模型上，开源系统的平均误绑定率为 19.84%，API 系统为 7.55%；这些错误被总体准确率所掩盖。在可识别的转移中，分别有 80.70% 和 81.51% 来自相邻实例。定位和实例优先干预对部分模型有帮助，但并非通用补救办法。因此，InstaBind-Lite 将此前未加区分的错误答案转化为可识别来源的失败类别，并测试了传统基准无法确定的一个可靠性维度：模型是否不仅知道什么是可见的，还知道每个属性属于哪个实例。
Humanoid robots hold great promise as general-purpose agents in human-centered environments, yet generalist vision-language-action (VLA) foundation models are not readily applicable to humanoid whole-body loco-manipulation. The high dimensionality and interdependence of humanoid motions make it challenging for conventional single-stage VLA architectures to coordinate locomotion, waist posture, and dual-arm manipulation effectively. Moreover, policies trained through offline behavior cloning can remain suboptimal during real-world deployment. Although online reinforcement learning can refine policies through real-world interaction, directly tuning large VLA backbones demands excessive computation and may introduce safety risks during real-robot exploration. To address these bottlenecks, we introduce HAF (Humanoid Adaptation Framework), a two-part framework consisting of HAF-VLA and HAF-Steer that transfers off-the-shelf generalist VLA foundation models to humanoid whole-body loco-manipulation. HAF-VLA is a hierarchical action-flow generator built on a pretrained flow-matching VLA. It splits full-body action denoising into three sequential stages with stage embeddings and cross-stage KV caches that retain kinematic dependencies, avoiding incoherent whole-body actions from one-shot generation. On top of the frozen HAF-VLA, HAF-Steer is a latent offline-to-online RL pipeline that leverages flow-matching invertibility and DCT-based dimensionality reduction to restrict RL optimization to a compact noise subspace and train a regularized SAC policy. This avoids updating the large VLA backbone and enables efficient real-world policy refinement. Evaluated on seven real-world humanoid loco-manipulation tasks, HAF surpasses vanilla single-stage VLA baselines and improves whole-body coordination and task performance. Project website: https://grange007.github.io/HAF .
Long-horizon robot manipulation chains many contact-rich skills into one multi-stage task. Vision-language-action (VLA) models increasingly master the individual skills, yet the chain still fails: errors compound beyond the policy's ability to correct, and one subtask silently constrains the next. A promising recipe freezes the VLA and puts an LLM agent in charge: it plans in language, moves in free space with analytic primitives, invokes the VLA only for contact-rich segments, and writes adaptation into language memory. Applied to long horizons, it breaks twice. (1) Competence comes from whole-task exploration at test time, whose cost is multiplicative in stages: if one stage needs T episodes, a K-stage task needs about T^K, and a failure does not reveal which stage caused it. (2) It has no representation of transitions: the VLA primitive carries an exit but no entry condition, so a subtask can succeed in a form its successor cannot use. We present BATON. Against (1), BATON makes the subtask the unit of exploration: each is explored in the cheap short-horizon regime and its solution stored in memory; a long-horizon trajectory is then composed from these solutions rather than discovered whole. Cost becomes additive (T*K) and every failure is attributed to a single stage. Against (2), BATON equips exploration with a transition-aware memory. Within a subtask, a verifier agent governs the invocation transition: the VLA is called only after the wrist view confirms the scene is ready. Across subtasks, a handoff transition restores an entry state disturbed by the predecessor's residue, and a lookahead transition selects the strategy whose outcome the successor can inherit. No parameters are updated. On the long-horizon benchmark RoboMemArena, BATON improves task success by 11.6% and cumulative success by 14.9% over the SoTA.
The rapid development and widespread application of artificial intelligence (AI) have sparked intense discussions on how to deploy responsible AI systems in a manner aligned with human values and ethical standards. Compared to fields like healthcare, energy, or finance, the application of AI in groundwater is relatively limited, and research on responsible AI is even more scarce. Taking the middle reaches of the Heihe River Basin as the study area, this paper proposes six Responsible AI principles: transparency, technical robustness, privacy governance, fairness, accountability, and sustainability. LSTM and Transformer time-series models are developed using multi-source hydrometeorological data, and validated via post-hoc interpretability, Monte Carlo simulation, and scenario analysis. The results show that Transformer outperforms LSTM in accuracy, robustness, and interpretability, demonstrating the operability and practical value of Responsible AI principles in groundwater prediction to support sustainable water management under climate change and human activities.
在资源受限的边缘设备上运行大型 AI 模型需要进行模型压缩，以降低模型大小和计算量。然而，压缩效果好并不一定意味着部署效果好。我们调研了数十项近期工作，这些工作报告了真实硬件上的压缩结果，并从中提炼出实用的部署指南。遵循这些指南，我们在 GPU、CPU 和 Raspberry Pi 平台上部署了紧凑型语言模型和图像模型，涵盖问答和图像分割任务。没有任何单一技术能在所有任务中胜出。对于问答任务，Qwen3.5 0.8B 在 Q5_K_M GGUF 量化下达到 93.85 SQuAD F1 和 92 EM，而在相同精度下，1% 比例的结构化剪枝会损失 16 F1。对于分割任务，排名则相反：默认量化不会改变参数量和 MACs，而剪枝在 mIoU 近乎不变的情况下将模型大小减少近 80%。剪枝甚至可能通过破坏 k-quant 超块对齐，使部署产物膨胀 21-49%；再加上输出更长且更不符合格式规范，这会使 Raspberry Pi 延迟最高增加到 3.4 倍。压缩还可能制造出能力表象，而不是明显地破坏能力：一个经 LoRA 恢复的变体仍然完全可解析，并在严格 BoolQ 准确率上保持 71%，但其 100 个预测中有 97 个被发送到单一类别，平衡准确率为 52.6%。我们通过神经流图分析和预填充-解码级延迟分解来解释这些现象，并将其凝练为面向具体任务的部署研究方向。合适的技术取决于任务、模型和硬件。我们的实验代码和产物已在 https://github.com/Arnavvvkumar/deployment 开源。
Background: Distillation of training targets generated thru search/planning has proven useful in reinforcement learning, but search can take exceedingly long. Objectives: Rather than perform search to the same depth every time (typically at a fixed period of steps), reduce the search depth proportionally to the quality of the policy network priors. Methods: We describe Flexer, an architecture that, for each step, mixes the policy from a neural network and the policy from Monte Carlo tree search. The mixing factor favors the MCTS policy as the policy imitation error of the network and the environment models' variance increases. Results: Flexer outperforms a version of AlphaZero (and DQN and ADP) for some experiments on three toy symbolic problems.
Large language model (LLM) agents often perform poorly on complex, long-horizon tasks because their context becomes increasingly cluttered over time. As interactions accumulate, detailed execution traces and intermediate outputs dominate the context, making it difficult for the model to retain and use high-level planning information. Most existing methods address this issue through compression or retrieval applied to a single, flat context, which does not clearly separate different types of context information and often leads to degraded reasoning. To address this challenge, we propose HyMem, a hierarchical framework that explicitly separates the agent's context into distinct functional layers. HyMem organizes context by function to separate high-level planning from execution and complex analysis. Its isolated reasoning module handles complex subtasks without adding intermediate reasoning traces to the persistent planning context, while its memory management module preserves task progress across context refreshes through structured summaries. These components reduce redundant context accumulation, retain task-critical information, and support coherent long-horizon reasoning within a limited context window. Experiments on GAIA and Browsecomp-plus show that, with DeepSeek-V4, HyMem achieves average Pass@1 scores of 66.7% and 61.3%, outperforming the strongest baseline by 6.1 and 4.7 percentage points, respectively. Further analysis indicates that HyMem effectively controls the growth of the reasoning context, allowing the model to maintain focus and accuracy across complex, long-horizon tasks.
Diffusion models have recently shown strong potential for multivariate time-series anomaly detection by learning the distribution of normal data through iterative denoising. Existing diffusion-based approaches, however, typically perform anomaly detection after completing the reverse diffusion process, relying primarily on the final reconstructed signal and overlooking informative representations produced during denoising. This design incurs substantial computational cost and limits the use of intermediate diffusion information for anomaly detection. In this paper, we propose FirstDiff, a diffusion-based anomaly detection framework based on the observation that the predicted diffusion noise at the initial reverse-diffusion evaluation already contains sufficient information for accurate anomaly detection. FirstDiff models the statistical distribution of predicted diffusion noise under normal behavior using validation data, enabling anomaly inference from a single denoising-network evaluation rather than completing the reverse diffusion trajectory. To model complex temporal and inter-sensor dependencies, FirstDiff employs a Diffusion Transformer as the denoising backbone. Extensive experiments on five public benchmark datasets demonstrate that FirstDiff achieves state-of-the-art performance while reducing diffusion inference from the full reverse trajectory to a single denoising-network evaluation.
Foundation models (FMs) increasingly support multimodal and geospatial reasoning, yet it remains unclear whether cartographic principles designed for human perception are equally effective for machines. Focusing on sequential choropleth maps, we examine how hue palette, color ordering, and lightness contrast influence FM spatial reasoning. We construct a controlled benchmark of 5,760 maps and 28,800 questions spanning Attribute Identify, Spatial Recognition, Compare, Rank, and Pattern Delineate, and evaluate 21 open-source and proprietary multimodal FMs. Results show that hue choice has limited and inconsistent effects, whereas disrupting sequential color ordering substantially reduces performance, especially for comparison and ranking. Reduced lightness contrast also consistently impairs reasoning, while increasing contrast beyond sufficient separability provides only marginal gains. LoRA fine-tuning improves overall accuracy but preserves these relative sensitivities. Additional factorial experiments further indicate that errors arise from color-and-legend decoding, spatial reasoning, and the integration of thematic attributes with spatial structure. These findings show that conventional sequential ordering and sufficient contrast remain important for machine map understanding and provide empirical guidance for AI-friendly cartographic design.
We introduce TinyCast, an attention-free zero-shot forecaster that emits a predictive distribution from 146,505 parameters, on the premise that at this size the periodic structure of a context is worth computing rather than learning. A zero-parameter spectral detector supplies the dominant periods, the context is folded on their phase, and a dilated convolutional encoder and a block-autoregressive quantile decoder model the rest. It is smaller than every zero-shot entry on the GIFT-Eval board whose parameter count can be established. On probabilistic accuracy it defines the size-accuracy frontier. Among zero-shot entries declaring no test-data leakage it is the only one below 1.4M parameters that emits a predictive distribution, and every entry scoring better carries at least that budget. On Chronos-ZS and fev-bench every neural model ahead of it carries at least 28 times its parameters. Because the mixing path is convolutions and matrix multiplications only, it exports to static INT8 and forecasts end to end on an embedded device without per-signal fitting.
Conformal prediction (CP) provides distribution-free coverage guarantees and has emerged as a principled tool for uncertainty quantification. In edge-level fraud detection on temporal interaction graphs, where false positives and false negatives both carry substantial cost, such coverage guarantees are particularly appealing for risk-aware decision making. However, directly applying existing graph conformal predictors yields inefficient prediction sets due to two recurring properties of fraud data. Fraudulent interactions are often embedded in benign-dominated neighborhoods that dilute calibration signals, while extreme class imbalance leaves scarce labeled-fraud support in the calibration split and leads to overly conservative class-conditional thresholds. To address these issues, we propose ProtoCP, a conformal prediction framework for edge-level fraud detection on temporal graphs. ProtoCP improves calibration efficiency by focusing calibration on fraud-relevant subgraph context and producing more stable nonconformity scores under class imbalance and temporal drift. Specifically, it leverages learned prototypes to suppress benign-dominated noise in the calibration context and introduces a neighborhood-relative scoring mechanism with temporal score diffusion for stable class-conditional calibration. Experiments on four fraud benchmarks (YelpChi, S-FFSD, FTFD, and BankSim) show that ProtoCP achieves the target coverage with consistently smaller prediction sets than state-of-the-art baselines. Our codes are available at https://github.com/Picard1701ent/ProtoCP.git
The growing ecosystem of large language models (LLMs) offers huge potential to optimize performance-cost trade-offs. However, their heterogeneous capabilities and inference costs make efficiently routing queries a significant challenge. Existing paradigms are inflexible: one-shot routers commit before observing responses, whereas conventional cascades stop adaptively but follow a fixed model order. Cascade routing removes both restrictions by reconsidering whether to stop or invoke another model after each response. Current methods use a predict-then-optimize pipeline estimating response quality and future model utility. However, prediction loss for quality or utility is not equivalent to routing-decision loss. A lower prediction error does not necessarily yield a better action; a small boundary-crossing error can reverse a ``stop'' or model-selection decision. Therefore, we propose RLCascadeRouter, a quality-estimator-free framework that formulates cascade routing as a Markov decision process with actions comprising ``stop'' and model selection. It uses trajectory returns and advantages to directly optimize the performance-cost objective. Its Cascade Policy Network models candidate complementarity for model selection and remaining-action value for stopping, eliminating independent post-hoc response-quality estimators. Evaluated across ten LLMRouterBench benchmarks with thirteen LLMs, RLCascadeRouter outperforms strong baselines and achieves superior performance-cost trade-offs. It incorporates unseen models without retraining, and ablation studies validate both policy components.
As AI systems become increasingly capable of autonomous action, determining whether an agent is technically capable of performing an action is insufficient: the system must also determine whether the action is authorised in its context. This paper introduces the Authority Resolution Framework (ARF), a five-domain ontology for representing and resolving authority across organisational roles and informal influence, business concepts, codified processes, machine-readable permissions and executable systems, and external real-world context. ARF defines the Authority Relation (AR) as a cross-domain primitive binding an actor, action, object, bounded context, justification chain, and a calibration measure termed the DNA-Coefficient, which captures divergence between documented authority structures and authority as practiced. The framework provides a machine-interpretable representation of authority provenance and scope, with JSON-LD representations and knowledge-graph query patterns for authority resolution. ARF is designed to support AI agents in determining the provenance, scope and contextual validity of authority before executing consequential actions. The framework positions authority resolution as a knowledge-representation and reasoning problem at the intersection of ontology engineering, semantic AI, agentic AI and AI governance.
文章 URL： https://github.com/ggml-org/llama.cpp/releases/tag/v0.1.0 评论 URL： https://news.ycombinator.com/item?id=49335017 积分：10 评论数：2
HN 大家好！我是 Bek，Speko 的创始人。Speko 是一个平台，会在我们所有公开基准测试过的选项中，根据你的约束条件，找出 speech-to-text、LLM 和 text-to-speech 模型的最优组合，并告诉你原因。 演示： https://youtu.be/no2LY2gRh-c 链接： https://speko.ai/?utm_source=hn 典型的生产级语音代理是三个模型的组合：STT、一个 LLM 和 TTS。 这些层中的每一层都有十来家可信的供应商，而且每个月市场上都会出现新模型。几乎所有人都是评估一次，选定一套自己偏好的技术栈，然后再也不复查，因为从一家供应商切换到另一家供应商又意味着一次新的集成，以及关于数字的争论。 结果就是，你在使用运行着上个季度模型的语音代理，而更好、更便宜的选择其实已经可用。 在创立 Speko 之前，我曾用四年时间作为联合创始人兼 CTO，为亚洲各地的企业构建支持 10 多种语言的语音代理。每当一个新的语音模型出现时，我们都会重复同样的流程：雇用母语评测员，将其与我们现有的技术栈做基准测试，如果有所提升就更新生产环境。Speko 将这个流程变成了一个 API。一个每天处理数千通电话的团队告诉我们：“我们真的可以直接去这个仪表盘，切换模型，它就会替我们完成。” 工作原理：你发送一个带有优化标准（准确率、延迟、成本或均衡）、语言和地区的请求。路由器会筛选出我们针对给定约束组合测量过的模型，对它们进行基准测试，选出胜者，并返回一个响应，其 headers 中包含供应商、模型名称和分数。网关会预取已签名的会话计划，因此新会话会直接从内存拨通供应商；呼叫者等待时不会发生控制平面的往返。 故障转移只发生在连接建立阶段：如果供应商拒绝连接尝试，我们就开始连接排名靠后的候选者。 一些客户案例：有一位创始人来找我们时完全不知道该选什么；他把自己的用例给了我们，现在所有流量都通过这个平台路由。一个物业管理 AI 使用 Python 运行 LiveKit，自上线以来从未更新过 STT 或 TTS：他们不知道自己的 STT 在通话中的错误率很高，也不知道已有更好的选择，而且替换看起来总像是一个 R&D 项目。一个团队不知道该为西班牙语选择哪些模型。一个医疗团队不知道哪个 STT 最擅长处理医学词汇。每种情况下，我们都帮助他们从基准测试中找到合适的技术栈，现在他们都通过我们路由。 测量部分是公开的：我们在一个地区、不同日期的运行中，把相同输入传给每个模型，并发布排行榜，包括那些我们的选择表现不如替代方案的榜单。发布演示会回答哪个 30 秒片段听起来更好；生产环境关心的是哪个模型能撑到第八分钟，所以我们测试自发语音、金额和日期、十分钟片段，而排名会发生变化。我们基于盲测一对一听感投票，训练了一个用于 TTS 自然度的自动评分器；对于它从未见过投票的供应商，它选出的胜者与我们的评测员所选一致的频率，大约和评测员之间彼此一致的频率相当。 我们自己不训练或销售模型，这正是我们保持排名公正的方式。 我们还开源了网关，面向那些希望避免音频路径上额外网络跳转、且不想与我们的云共享密钥的团队（ https://github.com/SpekoAI/gateway ，MIT）：一个 Go 二进制文件，作为 sidecar 运行在你的代理容器中，通过 Unix socket 使用一种本地协议，固定供应商主机并附加你的密钥。在 BYOK 模式下，它完全不会与我们通信。 请注意，匿名且不含内容的遥测默认启用，一个环境变量即可禁用。 成本：网关和 BYOK 设置将永久免费，我们对托管路由器和托管密钥以及合并计费收费。自从我们在 6 月下旬开始这一批次以来，外部使用量平均每周增长约 25%，增长主要集中在发布前几周。 我很想听听社区的反馈：你们现在如何选择语音模型？什么会让你们信任第三方基准测试？ 评论 URL： https://news.ycombinator.com/item?id=49332751 点数：22 评论数：7
Hi HN！我是 Bek，Speko 的创始人。Speko 是一个平台，会在我们所有公开基准测试过的选项中，根据你的约束条件，找到语音转文本、LLM 和文本转语音模型的最优组合，并告诉你原因。 演示： https://youtu.be/no2LY2gRh-c 典型的生产环境语音代理是由三个模型组成的集成：STT、一个 LLM 和 TTS。 这些层中的每一层都有十几个可信的供应商，而且每个月市场上都会出现新模型。几乎所有人都是评估一次，选择自己喜欢的技术栈，然后就再也不复查，因为从一个供应商切换到另一个供应商又意味着一次新的集成，以及围绕数据的争论。 结果就是，你使用的语音代理还在运行上个季度的模型，而更好、更便宜的选择已经出现了。 在创立 Speko 之前，我花了四年时间作为联合创始人兼 CTO，为亚洲各地的企业构建支持 10 多种语言的语音代理。每当新的语音模型出现时，我们都会重复同样的流程：雇用母语评分员，将它与现有技术栈进行基准测试，如果有所改进就更新生产环境。Speko 将这个过程变成了一个 API。一个每天处理数千通电话的团队告诉我们：“我们真的可以进入这个仪表板，切换模型，它会替我们完成。” 工作方式：你发送一个包含优化标准（准确率、延迟、成本或均衡）、语言和地区的请求。路由器会筛选出我们针对给定约束组合测量过的模型，对它们进行基准测试，选出胜者，并返回一个响应，响应头中包含提供商、模型名称和分数。网关会预取签名会话计划，因此新会话会直接从内存拨向提供商；当呼叫者等待时，不需要控制平面的往返。 故障转移只发生在连接建立阶段：如果提供商拒绝连接尝试，我们就开始连接排名靠后的候选项。 一些客户案例：有一位创始人来找我们时完全不知道该选什么；他把自己的用例给了我们，现在所有流量都通过这个平台路由。一家物业管理 AI 使用 Python 运行 LiveKit，自上线以来从未更新过 STT 或 TTS：他们不知道自己的 STT 在通话中的错误率很高，也不知道有更好的选择，而且模型替换看起来总像一个研发项目。一个团队不知道该为西班牙语选择哪些模型。一个医疗团队不知道哪种 STT 最擅长处理医学词汇。在每个案例中，我们都帮助他们从基准测试中找到合适的技术栈，现在他们都通过我们进行路由。 测量部分是公开的：我们在不同日期的运行中，将相同输入传给同一地区的每个模型，并发布榜单，包括那些我们的选择表现不如替代方案的榜单。发布演示回答的是哪段 30 秒音频听起来更好；生产环境问的是哪个模型能撑过第八分钟，所以我们测试自发语音、金额和日期、十分钟录音，排名也会发生变化。我们基于盲测一对一听感投票训练了一个用于 TTS 自然度的自动评分器；对于它从未见过投票的提供商，它选出的胜者与我们的评分员选出的一致，其频率大约等同于评分员之间彼此达成一致的频率。 我们自己不训练或销售模型，这正是我们保持排名公正的方式。 我们还开源了网关，供那些希望避免音频路径上额外网络跳转、并且不想与我们的云共享密钥的团队使用（ https://github.com/SpekoAI/gateway ，MIT）：一个 Go 二进制文件，作为 sidecar 运行在你的代理容器中，通过 Unix socket 使用一种本地协议，固定提供商主机并附加你的密钥。在 BYOK 模式下，它完全不会与我们通信。 请注意，匿名且不含内容的遥测默认启用，一个环境变量即可将其关闭。 成本：网关和 BYOK 设置将永久免费；我们对带有统一账单的托管路由器和托管密钥收费。自从我们在 6 月下旬开始这一批次以来，外部使用量平均每周增长约 25%，增长主要集中在发布初期的几周。 我很想听听社区的反馈：你们现在如何选择语音模型？什么会让你们信任第三方基准测试？ 评论 URL： https://news.ycombinator.com/item?id=49332751 点数：7 评论数：0
Pretty self explanatory. Could you folks shed some light on why these issues keep happening? I’ve noticed most posts and replies are just people coming to their own conclusions based on whatever published data. I think it would be helpful to get some actual, non-corporate/marketing information on the goings-on by those that actually see what’s happening on the inside. Thanks Can Comments URL: https://news.ycombinator.com/item?id=49332495 Points: 10 # Comments: 3
文章 URL： https://www.wiz.io/blog/red-agent-snowflake-copilot-cicd-bug 评论 URL： https://news.ycombinator.com/item?id=49331423 点数：6 评论数：2
Github has been down consistently over the last few months - does it make sense to switch to alternatives? Comments URL: https://news.ycombinator.com/item?id=49331033 Points: 18 # Comments: 9
Hi HN. I built 1667 for my own fiction work and now use it each day. This probably has a limited audience. Maybe an audience of one... Why a terminal interface for story writing? I'm a dev. I like to use terminals for a lot of stuff. Most WebUIs feel off to me. That's the only reason. One thing that bothers me about writing in existing tools is that they don't fit the way I write. The mental model of my story is a tree. I try many takes usually continue with just one, but sometimes I want to try an alternate route and see where this goes. And that can branch again in many places. See what happens if I kill off this character or they don't take the job or whatever. 1667 is a full-screen terminal app for long-form fiction. Each story part can have several takes. All takes stay in a tree. You select one path through that tree as the story line. Export writes that line to Markdown in the project folder. Some technical details: - A project stores its stories and settings in a `.1667/` directory. Exported Markdown sits beside it. - Provider secrets stay in private machine files. Requests go to the provider that the writer selects. - An optional Vault Password seals project files at rest. - An operating-system lock permits one writer process for each project. - The request viewer shows the next provider request without its credential. - Each generated take keeps a Generation Record with its model and effective settings. Version 0.9.5 runs on macOS, Linux, and Windows x64. The website has Shell and PowerShell installers. An npm package is also available. 1667 imports Markdown, SillyTavern chats and cards, and NovelAI archives. It can use OpenAI-compatible, Anthropic, and local endpoints such as Ollama, LM Studio, llama.cpp, and KoboldCpp. Current limits: the release is pre-1.0. The interface is a terminal. There is no account or cloud sync, and I don't plan to add any. No tracking. Comments URL: https://news.ycombinator.com/item?id=49330604 Points: 12 # Comments: 12
Githubstatus.com currently says everything is working, but it isn't Comments URL: https://news.ycombinator.com/item?id=49330632 Points: 30 # Comments: 9
Just got this message: "No server is currently available to service your request. Sorry about that. Please try refreshing and contact us if the problem persists." Comments URL: https://news.ycombinator.com/item?id=49330597 Points: 15 # Comments: 4
Existing palm presentation attack detection (PAD) datasets are often limited by static imagery, restricted acquisition conditions, or insufficient multimodal video data, hindering systematic evaluation across environments, modalities, and attack types. We present GBU-Palm, a large-scale multimodal video dataset and benchmark containing 21,326 videos from 105 subjects and 210 palms across six acquisition environments, including bona fide, Print, and Replay presentations, with 6,310 synchronized RGB-NIR samples. We construct leakage-controlled protocols that separate palm identity and attack lineage and benchmark four representative video architectures under environment-matched and held-out-environment settings. Results reveal substantial architecture-dependent degradation under environmental shift and show that RGB-NIR fusion does not consistently outperform RGB-only input. We further analyze model behavior through true accept (TA), true reject (TR), false accept (FA), and false reject (FR) decomposition, spectral masking, temporal-order intervention, and frozen-backbone NIR probing, revealing distinct failure patterns and evidence utilization across architectures. GBU-Palm provides a unified and challenging benchmark for developing and evaluating robust multimodal palm PAD methods under cross-environment conditions.
近期的视频生成器能够伪造对战争、灾难、公共紧急事件以及其他现实世界危机的逼真描绘，从而带来重大的错误信息风险。然而，现有基准在此类场景下对检测器和生成器行为的证据仍然有限，包括可检测性如何随生成条件变化、人们如何感知生成视频，以及检测器在社交传播过程中是否仍然可靠。为弥补这一空白，我们提出 RA-Bench，这是一个使用真实视频作为锚点的 AI 生成视频检测基准。RA-Bench 包含 17,886 个视频，其中包括覆盖 10 个社会风险类别的 1,830 个真实视频锚点，以及来自四个开源和五个闭源生成器的 16,056 个生成片段。基于 RA-Bench，我们从三个维度组织评估。首先，我们评估检测器的泛化能力，涵盖七种传统检测器、十种在三种审核设置下的零样本多模态模型，以及两个专门针对 AI 生成视频检测进行微调的 MLLMs。在这些方法中，三类检测器都无法在 RA-Bench 实例上持续泛化。随后，我们考察可检测性如何随生成质量、条件信息和采样种子而变化。这些分析表明，生成属性对不同检测器类别的影响不同，而源级别检测模式在不同种子下保持稳定。最后，我们研究人在真实性判断中的表现以及检测器在社交传播过程中的可靠性。我们发现，误导人们的视频对当前检测器而言同样难以识别，并且社交传播会使检测更加困难。总体而言，这些发现表明，当前方法难以检测逼真的 AI 生成视频，凸显了对能够适应不断演进的视频生成器且具备鲁棒性的检测器的需求。
Zahavy [2026] argues that Large Language Models, despite their capabilities in induction and deduction, cannot perform the abductive "Jump" that produced Einstein's equivalence principle, and attributes this limitation to the absence of embodied simulation. Zheng-Xin [2026] and Farmer [2026] question whether embodiment is necessary for abduction, pointing to alternative routes to General Relativity and forms of abduction that require no sensorimotor grounding. Max Planck resolved the blackbody radiation problem in 1900. Planck's move to E = hν required no embodied simulation. It was motivated by a mathematical consequence of classical theory, an infinite predicted energy for a finite measured quantity, that could not be physically accepted. We show that neither induction nor deduction could have produced the postulate and argue that its adoption required a coupling between epistemic error and physical cost. We formalize this distinction through thermodynamic coupling and show that fixed-weight transformer inference lacks such coupling, regardless of model scale. This is consistent with empirical results showing that output entropy remains nearly unchanged across tasks with sharply increasing causal difficulty, even as accuracy falls from 100% to 17%. We therefore argue that the missing ingredient in machine abduction may lie deeper than embodiment: a system must have some physical mechanism through which epistemic error becomes costly enough to force revision.
Art style is a signature of professional digital artists that develops through repeated experimentation, reflection, and adaptation. While generative AI (GenAI) can reproduce styles with high fidelity, current tools provide limited support for exploring new stylistic directions and may encourage style replication over exploration. To address this gap, we propose Analyze-Experiment-Resituate (AER), a framework for AI-assisted style exploration derived from interviews with 10 professional digital artists. Rather than prioritizing visually appealing outputs alone, AER supports three core practices of style exploration, including interpreting references, trying out stylistic possibilities, and reflecting on how emerging styles may be received. Specifically, AER enabled artists to (1) analyze artworks into interpretable stylistic elements, (2) have controllable experimentation guided by their own choices, and (3) resituate emerging styles through simulated social perspectives. We implemented AER in a prototype system and evaluated it in a controlled study with 16 artists. Compared with a direct style-transfer workflow, AER increased artists' agency and reflection as they pursued new stylistic directions. A two-week field study with four artists revealed how the AER framework influenced daily style exploration, such as reflection, experimentation, and stylistic decision-making at each stage. We discuss opportunities and challenges in designing AI-assisted style-exploration workflows, and outline implications for future artistic support tools.
自演化智能体能够从交互经验中改进未来行为，但现有评估通常在固定执行条件下进行优化，并不测试这些条件变化后的恢复能力。为弥补这一空白，我们引入 PACE-Bench（Physics Adaptation via Code Evolution），这是一个基于模拟器的基准，涵盖六个物理领域中的 144 个源到目标适应配对。每个配对都将一个源环境与一个突变后的目标环境相连接，二者具有相同的目标和接口。一个代码驱动的设计在源环境中能够成功，但在目标环境中会失败；智能体必须在有限的尝试预算内，利用诊断性沙盒反馈，将其迭代适应为可在目标环境中工作的设计。我们比较了来自四种范式的十种自演化方法。该基准距离饱和仍很远：Reflexion + Qwen3-14B 在完整基准配对中的成功率仅为 35.9\%，而 GPT-5.5 在完整预算下解决了 Statics 子集中的 66.7\%。总体而言，这些结果表明，基于模拟器的反思比未经验证的自我修订更可靠，而记忆会将智能体锚定在早期设计上，广泛的树搜索则只是在探索而未能收敛。即使揭示确切的物理变化，也无法提高性能上限，这表明核心瓶颈在于机制重新设计，而非参数推断。数据和代码可在 https://github.com/thunlp/PACE-Bench 获取。
Short-horizon forecasting of fine particulate matter (PM2.5) remains difficult when observations from the target domain are limited and the statistical properties of the source and target domains differ. In these settings, models trained only on local data may not capture complex temporal dynamics, while direct transfer learning can result in negative transfer. This study develops a shift-aware dual-encoder transfer framework that combines source-domain knowledge with target-specific representation learning. The source encoder was pretrained using hourly observations from 10 U.S. monitoring locations. The framework was then adapted and evaluated using two years of hourly observations from 77 stations in Taiwan under a chronological train-validation-test protocol. Among the four principal baselines, the frozen-source dual-encoder model achieved the best performance, with MSE = 21.8960, MAE = 3.1597, and R^2 = 0.8725. This corresponds to an MSE reduction of approximately 7.1% relative to TL-v1 and 4.1% relative to TL-v2. The ablation analysis showed that removing the Taiwan-specific branch caused the largest decline in performance. Allowing the source encoder to adapt produced the best overall result, with MSE = 21.6575, MAE = 3.1383, and R^2 = 0.8739. SHAP analysis indicated that predictions were driven mainly by recent PM2.5 observations and meteorological variables related to pollutant transport and dispersion. These results suggest that source-domain knowledge is most effective when target-specific information is preserved and the transferred representation is allowed to adapt under target supervision.
Cross-Tabular Data Generation (CTDG) seeks to learn a generative model from multiple heterogeneous tables and produce new synthetic tabular datasets. However, existing synthetic tabular data generation methods are largely restricted to single-input-table scenarios and struggle to effectively handle multiple heterogeneous tables with diverse feature sets. To address this limitation, we propose a two-stage framework for cross-tabular data generation. In the first stage, each heterogeneous raw table is transformed into a standardized statistical table with the same set of columns across all tables. Each statistical table captures the marginal distributions of the original columns and the pairwise correlations among them. In the second stage, a diffusion transformer model is trained to capture structural patterns across these homogeneous statistical tables and to generate synthetic statistical tables. Synthetic raw tables are subsequently reconstructed from the generated statistical tables via multivariate Gaussian sampling followed by an inverse probability integral transform. This two-stage CTDG framework enables the learning of a unified generative model from multiple heterogeneous tables and supports the generation of an unlimited number of realistic synthetic heterogeneous tables. Experimental results demonstrate high fidelity in the learned statistical representations and a favorable fidelity-diversity trade-off in the generated synthetic data, validating the effectiveness of the proposed approach.
Free energies govern solid-state phase stability, yet computational materials discovery still relies largely on ground-state energies because free energy calculations require ensemble averages. We introduce the thermodynamic interatomic potential (TIP), which extends an interatomic potential from its static energy to a thermodynamically consistent Gibbs free energy model, with thermodynamic responses following from temperature and pressure by automatic differentiation. We implement TIP[UMA] using the universal potential UMA, train it on free energies from quasi-harmonic to molecular dynamics fidelity, and calibrate it to higher-resolution calculations or experiment. From a single evaluation, it returns the equation of state of a crystal and locates phase transitions among competing branches, including dynamically stabilized phases. Fine-tuning extends the model to alloy solubility limits and miscibility gaps. TIP makes the free energy as accessible as the potential energy, opening finite-temperature phase stability to high-throughput discovery.
Recipe data arises in domains such as materials synthesis, pharmaceutical formulation, and industrial manufacturing, where procedures are represented as ordered sequences of steps containing heterogeneous structured fields. Existing tabular learning methods typically flatten this structure into fixed-schema representations, limiting their ability to capture hierarchical field interactions and procedural dependencies. We propose RecipeNet, a hierarchical Transformer architecture that encodes field-level interactions within each step and sequential dependencies across steps through stacked Transformer encoders. Experiments on multiple recipe datasets and tasks demonstrate that RecipeNet consistently outperforms existing tabular models, highlighting the value of hierarchical and sequential modeling for recipe representation learning.
High-order multiple-input multiple-output (MIMO) detection requires efficient search over a large discrete symbol space while producing reliable soft information for channel decoding. This paper develops a learning-to-transition (L2T) framework that formulates MIMO detection as a stochastic sequence of complete-vector transitions. At each transition, a channel-coupled Transformer updates both the instance embedding and the sampling policy, while a blockwise autoregressive factorization captures inter-stream dependence with moderate sequential complexity. For hard-output detection, a transition network is applied recursively and trained through a residual-to-BER curriculum, which first learns the MIMO search geometry from the exact residual metric and then aligns the policy with transmitted-bit accuracy. For soft-output reception, the well-trained hard policy is cloned at the parameter level into every layer of an untied soft-input soft-output iterative detection and decoding (IDD) receiver. This tied-to-untied transfer preserves the learned zero-prior search dynamics while enabling layer- and round-specific specialization under decoder feedback. Within each IDD round, decoder priors tilt candidate generation according to Bayes' rule, and likelihood-weighted terminal hypotheses produce posterior and extrinsic log-likelihood ratios for LDPC decoding. A multi-stage training strategy further stabilizes the hard-to-soft transfer by progressively exposing the receiver to synthetic and in-loop decoder-generated priors.
This study investigates the methodological and theoretical properties of session handover in applications that use large language models. A task may continue in a new session when the context reaches the model's input limit, when the application restarts, or when another agent is asked to finish the task. The application must then decide which information from the earlier session to pass on. We formulate handover as the transfer of a task-relative in-context learning (ICL) state and distinguish exact recovery of earlier material from preservation of the target distribution. Under an exogeneity condition, predictive equivalence characterizes the coarsest deterministic sufficient handover and gives a fixed-length bit requirement. The analysis isolates the effects of the memory constraint, the writer, and the continuation procedure, and quantifies the cost of writing before the realized downstream query is known. We propose a three-part record that stores decisions and constraints exactly, uses task-justified statistics for repeated evidence, and retains original observations whose effect is not preserved by those statistics. Gaussian linear regression gives an exact finite-dimensional handover and finite-bit perturbation bounds, while nonparametric regression gives upper and lower bounds that relate memory to squared prediction error. These results provide a theory and method for deciding what a handover must retain and how its memory requirement depends on the continuation task.
This is intended as a question about the current phase of the LLM hype cycle, and, at the same time, as a reality check about whether reluctance to rely on LLMs is economically viable for an average developer. You've definitely seen that many companies progressed from "use LLMs for anything and everything" policies to limiting budgets and scopes of approved LLM usage. It wouldn't be unexpected to see that some companies might decide that generating more code faster doesn't meaningfully contribute to their bottom line, and forego LLMs in favour of human authoring, but does it ever happen? I'm not trying to point out any (de)merits of the new way of software development, I simply want to know if there are many/any companies that went back to the old ways. Comments URL: https://news.ycombinator.com/item?id=49318906 Points: 13 # Comments: 3
Google, the New York Jobs CEO Council and Urban Assembly hosted an AI summit for 150 education and industry leaders.
6G networks will not be serving as communication infrastructures only; rather, they are expected to evolve into intelligent systems, where thousands of autonomous artificial intelligence (AI) agents are interconnected. The agents are deployed across a wide range of platforms including low Earth orbit (LEO) satellites, high-altitude platforms (HAPs), unmanned aerial vehicles (UAVs), edge servers, and terrestrial devices. These agents continuously observe their environment and exchange information. Semantic communication provides an efficient mechanism for exchanging meaningful information instead of raw data. However, its effectiveness depends on the communicating agents having sufficiently aligned beliefs to correctly interpret and decode the transmitted messages. This assumption becomes difficult to satisfy in the 6G network where heterogeneous AI models operate under diverse computational constraints and continuously acquire different knowledge from their local environments. This article presents a heterogeneity-aware belief synchronization framework for 6G AI-native networks. It uses latent translation models deployed on multi-access edge computing (MEC) servers. These models translate belief updates from one agent to agent-specific knowledge without requiring joint training and a homogeneous architecture of models. By exchanging compact belief updates through a latent translation model only when necessary, the framework preserves privacy, reduces synchronization cost, and minimizes local knowledge drift. We validate the framework through a case study on a multi-layered terrestrial/non-terrestrial network. Results demonstrate that it maintains low synchronization cost, measured by the number of parameters transmitted, and low belief alignment error across the heterogeneous agents in the case study.
Existing predictive models in learning analytics often treat student academic history as a simple sequence, overlooking the concurrent nature of courses taken within a semester. This simplification can lead to inaccurate performance predictions, particularly for students with heavy or challenging course loads. This paper introduces a TRansformer for Academic Course-grade Estimation (TRACE) that addresses this limitation by jointly predicting both the set of courses a student will take and their corresponding grades for an upcoming semester. Our approach encodes courses on a per-semester basis to capture the effects of course concurrency and utilizes a novel loss function combining course-set prediction with grade prediction. We demonstrate that predicting courses taken in addition to the grades in those courses leads to significant improvements in prediction quality. Trained on ten years of institutional data, our joint prediction model reduces mean absolute error by nearly 50% compared to an identical architecture that predicts grades alone. The model also outperforms traditional LSTM-based sequential models, as well as graph neural network-based approaches, and offers natural ways to incorporate student attribute data. This work demonstrates the utility of modern neural architectures for creating interpretable models that can be adapted to new institutions via retraining and recalibration, as well as the importance of key techniques, such as predicting courses taken during training. We discuss how this model could be incorporated into early detection systems at institutions of higher education.
We consider the problem of autonomously learning robot skills under a limited practice budget for sequential tasks. We propose an active skill learning algorithm, \emph{Deliberate Practice (DP)}, that computes a provably \emph{budget-optimal} allocation---practicing skills that maximize expected cumulative reward while being learnable within the budget. DP estimates both the time needed to master skills and the cumulative reward of the task plans that the skills unlock. Computing a budget-optimal allocation is challenging as it requires reasoning about combinatorially many skill plans over a large practice budget. Our key contribution is a bilinear program that can compute this exactly using off-the-shelf solvers. Through simulated and real-world experiments on long-horizon manipulation tasks, we show that our approach allows robots to optimally use limited practice time to acquire useful policies and improve long-horizon planning.
Instruction-tuned language models achieve strong performance across a range of generation tasks, but have also recently been shown to exhibit verbalized overconfidence. In question answering, verbalized model overconfidence may be associated with the consistency of the generated supporting rationales. In this paper, we study whether corresponding changes in the lexical diversity of generated answer rationales accompany changes in model confidence induced by instruction tuning. We evaluate three matched base and instruction-tuned models across question-answering benchmarks and find that instruction tuning consistently alters answer confidence, despite limited changes in predictive accuracy and decreases in likelihood-based calibration. Secondly, we observe a non-uniform effect of instruction tuning on rationale diversity: cross-rationale diversity consistently decreases, whereas surface-level lexical diversity varies in both direction and magnitude across models and benchmarks. Finally, we find that these differences persist after controlling for answer selection and rationale length, confirming that confidence and rationale diversity capture distinct effects of instruction tuning.
Contact-rich manipulation failures are often detected only after the robot has committed to contact. This is especially limiting in wrist-camera setups: close gripper--object views help observe contact, but a poor approach may already push, miss, slip, or disturb the object before conventional detectors react. We introduce \emph{ContactGuard}, a pre-contact execution monitor for chunked visuomotor policies. Given the policy's planned action chunk, ContactGuard predicts its short-horizon consequence in latent visual space and aborts if the predicted future latent indicates likely failure. Its latent world model is trained from unlabelled robot trajectories to predict compact multi-view visual embeddings under planned actions, avoiding pixel-level video prediction. A lightweight failure probe is then trained from a small labelled set of pre-contact clips. At deployment, ContactGuard anchors prediction before an imminent contact event, rolls the model forward under the policy's own actions, and verifies the predicted post-contact latent. Across real-world contact-rich manipulation tasks, ContactGuard predicts failure more accurately than direct and corrupted-action ablations, and transfers to live robot as a pre-contact abort signal without modifying the underlying policy.
We address the use of large language models (LLMs) to help discover Isabelle proofs. An Isabelle build establishes that the submitted theory is accepted, but not that an LLM changed only what the developer authorised. We present CAPRI, a contract-aware repair workflow in which Isabelle checks the proof and an independent checker enforces a machine-readable edit contract. Prompts, proposals, candidate repositories, diagnostics, verdicts, and hashes are retained for audit. We evaluate five workflows on twelve failed proofs from four developments, with three replicates per task and condition, giving 180 runs and 138 valid repairs. Of 144 terminal candidates accepted by Isabelle, six had modified protected text; all arose in iterative workflows that could edit a complete theory. A proof-body-only interface produced 29/36 valid repairs and no contract violations, compared with 31/36 for the corresponding full-theory workflow. One-shot repair produced 22/36, while a later prospectively frozen iterative workflow produced 32/36; these figures compare complete workflows rather than individual mechanisms. A separate post hoc OpenRouter campaign found no improvement in the designated Luna comparisons. A Sol configuration with matched demonstrations produced 33/36 repairs, compared with 29/36 in the frozen OpenAI Responses condition, but the difference was not statistically significant in a one-sided exact McNemar test ($p=0.0625$).
We present Multi-Agent Reasoning and Coordination (MARC), an open-source framework that replaces monolithic LLM prompting with deterministic multi-agent orchestration for clinical reasoning. MARC coordinates role-specialized agents for extraction, reasoning, answer generation, and evaluation, with explicit context passing and traceable intermediate outputs, enabling stage-wise failure attribution. We additionally introduce a Decomposer module that generates task-specific agent prompts from a plain-language description, eliminating manual prompt engineering. The framework supports both API-based and local CPU-compatible deployments and is entirely configurable via YAML, without code modifications. MARC is designed to be model-agnostic, interpretable, and accessible to clinical domain experts without programming expertise. The full framework is available at https://github.com/Penn-RAIL/MARC-v1.
Current large language model development relies on massive, often non-permissible datasets, creating a high barrier for researchers committed to open-source and ethically sourced data. We introduce Mimir v1, a 1-billion-parameter language model based on the Hierarchical Reasoning Model (HRM) architecture, that is trained from scratch and delivers highly competitive performance for English and sets a new state of the art for Danish using only permissible post-training data. Trained on a mixture of 161 datasets, Mimir v1 outperforms the original HRM-Text 1B and competes with larger frontier models like Qwen 3.5 4B and Gemma 4 E2B, tested across 20 benchmarks for English, Math & Code and Danish. The model is available on the Hugging Face Hub: https://huggingface.co/danish-foundation-models/DFM-Mimir
We study masking diffusion for discrete sampling and introduce a path-resolved measure of data geometry called the \emph{unmasking growth complexity} ({\textsf{UGC}\xspace}). Its local increments directly control Kullback--Leibler (KL) discretization error, yielding a unified analysis of Bernoulli-subset and fixed-cardinality unmasking schemes. In log-reveal-odds coordinates, this structure yields optimized single-block and multi-block schedules, and quantifies the gains from adapting computational effort to data geometry. Crucially, we show how {\textsf{UGC}\xspace} increments can be estimated from samples via KL increments along coupled reveal trajectories. This leads to \emph{certified-optimal} samplers that achieve a prescribed KL error with high probability and iteration complexity within a constant factor of the corresponding oracle procedure. Collapsing the \ugc path yields the aggregate {\textsf{UGC}\xspace} mass, which connects to classical multivariate dependence measures and complexity measures from previous analyses of discrete diffusion. In the fine-partition limit, the squared integral of the square-root {\textsf{UGC}\xspace} density determines the sharp leading-order optimal Euler discretization error. Examples exhibit substantial dimension-dependent gains over coarse schedules, including $\widetildeΩ(\sqrt{d})$ improvements achievable with a constant number of adaptively placed blocks.
AI agents 越来越多地被用于编程，但无法对生成代码的正确性提供任何保证。经过验证的代码生成，即由 agent 同时生成实现及其规格说明的机器检验证明，为实现可信的 AI 生成软件提供了一条更强的路径。该方向上的现有基准要么专注于单个函数，要么仅在提供实现的情况下评估证明生成。agents 能否在真实的多模块代码库中做出连贯的实现和证明选择，仍然是一个未解问题。为弥合这一差距，我们引入 Vero，这是首个在仓库级别评估联合实现与证明合成的基准。Vero 包含 43 个多模块实例，来源于横跨 Python、Dafny、Verus 和 Coq 的真实世界仓库，并覆盖从密码协议到分布式系统的多样化领域。每个实例由一个多模块 Lean 4 仓库组成，包含预先确定的 API 接口、人工整理的形式化规格说明以及参考实现，支持仅证明和代码加证明两种评估模式。为提高基准的可靠性，Vero 还包含一种审计机制，允许 agents 形式化证明所提供规格说明的不可满足性或参考代码的不正确性，从而在整理过程中暴露并纠正潜在的代码和规格说明错误。我们评估了可访问 Lean 工具链的前沿 coding-agent 配置。最强的 agent 仅完全解决了 43 个实例中的 27 个，并且在最难的仓库上没有完成任何规格说明。Vero 为衡量仓库规模的经过验证的软件合成进展提供了一个具体测试平台，而当前 agents 在这方面仍有不足。我们在 https://github.com/sunblaze-ucb/vero 发布了该基准、整理流水线和评估工具。
Co-Scientist unites Boston Children’s Hospital and MIT’s labs to explore new RNA-based treatments for ALS.
Google 为 Genesis Mission 承诺提供 4000 万美元的 AI token 和额度
我们正在开源一个高性能 Python 库，用于使用 MuJoCo 引擎进行机器人仿真，该库是在我们过去一年的机器人研究中开发的。
We’ve created a bot which beats the world’s top professionals at 1v1 matches of Dota 2 under standard tournament rules. The bot learned the game from scratch by self-play, and does not use imitation learning or tree search. This is a step towards building AI systems which accomplish well-defined goals in messy, complicated situations involving real humans.
We’ve co-authored a paper that forecasts how malicious actors could misuse AI technology, and potential ways we can prevent and mitigate these threats. This paper is the outcome of almost a year of sustained work with our colleagues at the Future of Humanity Institute, the Centre for the Study of Existential Risk, the Center for a New American Security, the Electronic Frontier Foundation, and others.
我们发布了 Triton 1.0，这是一种类似 Python 的开源编程语言，使没有 CUDA 经验的研究人员也能编写高效的 GPU 代码——大多数情况下可媲美专家所能编写的代码。
In order to share the magic of DALL·E 2 with a broad audience, we needed to reduce the risks associated with powerful image generation models. To this end, we put various guardrails in place to prevent generated images from violating our content policy.
We’ve trained a model called ChatGPT which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.
Khan Academy explores the potential for GPT-4 in a limited pilot program.
This initiative is essential to our commitment to develop safe and advanced AI. As we create technology and services that are secure, reliable, and trustworthy, we need your help.
共同创建用于 AI 训练的开源和私有数据集。
今天，我们推出新技术，帮助研究人员识别由我们的工具创建的内容，并加入 Coalition for Content Provenance and Authenticity Steering Committee，以推动行业标准。
Zico Kolter Joins OpenAI’s Board of Directors We’re strengthening our governance with expertise in AI safety and alignment. Zico will also join the Safety & Security Committee
Genmab embraces ChatGPT Enterprise, supported by OpenAI’s commitment to security and privacy
OpenAI 与 MIT Media Lab Research 的研究合作项目。
我们介绍 gpt-oss-120b 和 gpt-oss-20b，这两个开放权重推理模型根据 Apache 2.0 许可证和我们的 gpt-oss 使用政策提供。
我们正在发布 gpt-oss-120b 和 gpt-oss-20b——两个先进的开放权重语言模型，能够以低成本提供强大的实际性能。这些模型采用灵活的 Apache 2.0 许可证，在推理任务上的表现优于同等规模的开放模型，展现出强大的工具使用能力，并针对在消费级硬件上的高效部署进行了优化。
开发者现在可以提交应用，以便在 ChatGPT 中进行审核和发布；获批的应用将出现在新的产品内目录中，方便用户发现。更新后的工具、指南和 Apps SDK 可帮助开发者构建强大的聊天原生体验，将现实世界中的操作带入 ChatGPT。
OpenAI 如何构建了一个实时访问系统，将速率限制、使用量跟踪和额度相结合，以支持对 Sora 和 Codex 的持续访问。
GABRIEL 是 OpenAI 推出的一个新的开源工具包，它使用 GPT 将定性文本和图像转化为定量数据，帮助社会科学家大规模分析研究。
OpenAI 与 Pacific Northwest National Laboratory 推出 DraftNEPABench，这是一个新的基准，用于评估 AI 编码智能体如何加速联邦许可审批——显示出将 NEPA 起草时间最多缩短 15% 并使基础设施审查现代化的潜力。
了解 Symphony 这一用于 Codex 编排的开源规范如何将问题跟踪器转变为始终在线的智能体系统——提升工程产出并减少上下文切换。
Learn how OpenAI protects community safety in ChatGPT through model safeguards, misuse detection, policy enforcement, and collaboration with safety experts.
Sea Limited's CPO explains why the company is deploying Codex across engineering teams to accelerate AI-native software development in Asia.
Warp 使用 GPT-5.5 和 OpenAI 模型，在本地、云端和开源开发工作流中协调编码代理。
How engineers at Nextdoor use Codex with GPT-5.5 to investigate hard-to-reproduce issues, build across platforms, and focus on product outcomes.
通过 Oracle Cloud 访问 OpenAI 模型和 Codex，使用现有承诺，以企业级安全性和治理来构建和部署 AI。
OpenAI 推出 Patch the Planet，这是一项 Daybreak 计划，旨在通过 AI 和专家审查帮助开源维护者发现、验证并修复漏洞。
OpenAI 和 Hugging Face 分享了 AI 模型评估期间一起安全事件的早期发现，重点介绍了先进的网络能力以及给防御者的经验教训。
OpenAI sent Governor Greg Abbott a letter outlining its commitment to responsible AI infrastructure in Texas. The letter supports reliable, transparent growth that benefits Texans.