arXiv:2610.13020v1 公告类型：新论文 摘要：强化学习（RL）已成为增强 LLM 推理能力的主流范式。然而，长时域自回归轨迹采样带来了一项根本性局限：局部策略失配会沿轨迹累积，造成严重的状态分布偏移。现有 LLM RL 方法（如 GRPO 和 GSPO）在词元或序列层面控制策略失配，却忽视了这种由轨迹引起的偏移，导致训练不稳定，并随着时域延长而最终崩溃。为弥合状态分布差距，我们提出 Trajectory-Aware Policy Optimization（TAPO），这一框架具有坚实的理论基础，从离策略策略梯度出发，在理论上推导出状态分布校正项，并将其纳入 RL 目标函数。TAPO 进一步采用几何平均估计器来降低方差，同时结合长度感知的动态裁剪，防止整条轨迹上的更新偏离过大。大量实验表明，TAPO 在八项竞赛级基准测试中显著优于现有 LLM RL 算法，相比 GRPO，在数学推理和算法编程任务上分别取得最高 48% 和 12% 的提升，同时大幅增强了训练稳定性。
arXiv:2610.13015v1 公告类型：新提交 摘要：强化学习已成为增强 LLM 推理能力的基础范式。然而，主流方法近端策略优化（PPO）依赖于一种几何退化的裁剪代理目标，无法自适应地调整对策略偏移的阻力，导致过早裁剪和探索能力下降。后续变体通过零散的启发式修改缓解了这些问题。我们提出广义近端策略优化（GPPO），这是一个原理清晰的统一框架，用于从散度视角理解和设计 LLM 强化学习算法。我们建立了具有性能单调改进保证的广义 $f$-散度信赖域理论，并表明不同的 $f$-散度会诱导不同的几何结构和策略更新动力学。我们进一步引入了新颖的估计器拓扑结构，相较于传统的单样本蒙特卡洛估计，能够提供更忠实的散度估计。在这一框架下，PPO 及 DAPO、DPPO 和 RIPO 等近期变体均可视为特例，由此建立了具有理论依据的 {散度 $\times$ 估计器} 二维设计空间。在八个竞赛级数学和编程基准上的广泛实验表明，这两个维度都至关重要。表现最佳的 GPPO 实例在数学和编程任务上分别比标准 GRPO 提升最高 50% 和 14%，达到当前最先进的性能水平。
arXiv:2610.13013v1 Announce Type: new Abstract: Softmax-based multimodal emotion recognition methods often assign high confidence even when sensor data are corrupted or contradictory, for example when facial expressions suggest positive valence while vocal prosody indicates negative arousal. We present an Evidential Multimodal Transformer with Dempster-Shafer Fusion (EMT-DS) that addresses this problem through formal uncertainty quantification. EMT-DS models predictions as Subjective Opinions (Dirichlet distributions), separating epistemic uncertainty (lack of evidence) from aleatoric uncertainty (sensor noise). Modality-specific transformer encoders produce belief mass distributions that are fused with Dempster's Rule of Combination, which normalizes the joint belief mass by a conflict coefficient 1/(1-c), so that contradictory evidence is explicitly quantified as conflict and the fused prediction becomes diffuse rather than being forced into an unjustified consensus. Under subject-independent and speaker-independent evaluation on four benchmarks, EMT-DS achieves 84.3% accuracy on IEMOCAP versus 81.7% for the best baseline (p < 0.01), 82.4% on DEAP under leave-one-subject-out evaluation versus 78.1%, and 96.8% on the highly imbalanced KEMDy20, and it improves accuracy on CMU-MOSEI by 2.4 points over the strongest evidential baseline. In controlled ablation, it reduces Expected Calibration Error from 0.153 (softmax transformer) to 0.042. It maintains 71.2% accuracy under severe Gaussian noise on audio features (sigma = 1.0), where attention baselines drop to 45.3%, and degrades gracefully when entire modalities are absent. The fusion step has O(MK^2) complexity, compared with O(M^2 d^2) for cross-modal attention. The learned conflict metric also flags likely errors: reviewing 20% of samples captures 80% of misclassifications, which supports selective prediction in clinical and safety-critical systems.
arXiv:2610.12990v1 Announce Type: new Abstract: We study how offline data can help online learning in stochastic linear bandits. We focus on partially covered offline data, where only $k$ of the $d$ feature directions are poorly covered. Existing analyses of algorithms based on phased elimination and optimism give regret bounds of order $\widetilde O(\sqrt{dkT})$ in this regime. We show that this dependence arises from loose control of the reward prediction error, and obtain tighter bounds using concentration of Lipschitz functions of Gaussian vectors. Based on this analysis, we develop a phased elimination algorithm with regret $\widetilde O(k\sqrt T)$. For the unit Euclidean ball, we prove a matching regret lower bound under partial coverage, as well as matching upper and lower bounds under uniform offline coverage. We also study pure exploration with offline data, providing algorithms and matching upper and lower bounds on simple regret for fixed-design algorithms and also for adaptive algorithms under partial and uniform coverage of offline data.
arXiv:2610.12976v1 公告类型：新论文 摘要：准确的日降水预报对天气敏感型决策至关重要。尽管近期天气基础模型（WFMs）取得了显著进展，但它们使用再分析数据进行训练，而这类数据存在随区域和降水强度而变化的偏差。由于基于观测目标重新训练 WFMs 的成本高昂，我们研究了 Dynamics-Precipitation Decoupling，这是一种即插即用框架，利用单独训练的生成器，将冻结天气模型输出的非降水预报映射为与观测一致的日降水。其关键挑战在于捕捉异质性降雨分布，并使日累计降水与随时空演变的大气信号相匹配。为应对这些挑战，我们以大气状态为条件，利用卫星与雨量计融合降水估计训练扩散生成器。多尺度循环编码器概括大气演变过程，交叉注意力则在去噪过程中对齐大气特征与降水特征。针对 Pangu-Weather、GraphCast 和 HRES 等多种骨干模型开展的全球及区域预报广泛评估表明，该方法提高了日降水预报准确性和事件检测能力，并能更好地表征局地累计降水分布。全球集合预报评估显示，在所有报告的预报时效下，该方法的 Brier 评分均低于 ENS 和从头训练的 GenCast，且在大多数比较中具有更低的 CRPS。值得注意的是，一项利用中国 2,288 个站点开展的强降雨个例研究，通过与地面观测对比，进一步证实了该方法在日累计降水分布表征和事件检测方面的改进。
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:2610.12904v1 Announce Type: new Abstract: Tabular foundation models (TFMs) based on prior-data fitted networks typically use sampled hard query labels, leaving avoidable sampling variability in pretraining. A key observation is that intermediate generator states can reveal label uncertainty beyond a sampled outcome. We propose Generator-state-Informed Supervision for TFMs (GIST), which uses these states to average query losses over conditional label distributions. GIST jointly designs task generation and query supervision, using entropy-calibrated encodings to obtain tractable, input-consistent targets on supported label-dependent branches. We show that this averaging preserves the population objective and yields no larger gradient covariance at fixed parameters within each generating distribution. Under matched pretraining schedules, GIST improves mean rank across three architectures on both TabArena and TALENT without changing architecture or inference. Across these benchmarks and architectures, GIST wins 62.1% of paired dataset-level comparisons against the corresponding hard-label controls.
arXiv:2610.12885v1 Announce Type: new Abstract: Federated Direct Preference Optimization (FedDPO) enables privacy-preserving alignment of language models across distributed agents without centralizing sensitive preference data. However, existing FedDPO analysis assumes all clients converge to a single shared policy, which is fundamentally at odds with the heterogeneous preference distributions present in practice. We propose Personalized FedDPO (pFedDPO), in which each client maintains a small private LoRA adapter atop a shared global base model. In pFedDPO, the base model is synchronized via FedAvg each round, while LoRA adapters remain local and capture client-specific preference patterns. We prove that this decomposition replaces the gradient diversity $\kappa^2$ in the FedDPO bound with a residual heterogeneity $\kappa^2_{\mathrm{eff}}$, which is no larger than $\kappa^2$ when adapters absorb client-specific directions, at the cost of an additive adapter-bias term that vanishes as adapters converge. Experiments on preference datasets show that pFedDPO consistently attains a lower gradient norm than vanilla FedDPO while remaining competitive in training loss and per-client reward accuracy.
arXiv:2610.12884v1 Announce Type: new Abstract: Accurate estimation of reservoir properties from seismic data remains challenging when well-log labels are sparse but three-dimensional seismic observations are abundant. We present CoGameReg, a cooperative-game-based semi-supervised regression framework that uses contribution-weighted predictions of heterogeneous regressors to construct shared pseudo-labels. This design makes each regressor's influence explicit in the learning targets generated for unlabeled seismic samples under sparse well control. The Tri-Game Shapley mechanism combines a rank-MSE performance game with a pairwise consistency game to estimate nonnegative regressor weights from labeled training data. These weights remain fixed during collaborative training and are reused for final prediction fusion. Candidate pseudo-labels are screened using a common local k-nearest-neighbor error-reduction criterion before retraining. We evaluate CoGameReg on 10 public regression datasets, a synthetic three-dimensional P-wave velocity prediction task, and a field water-saturation application. The benchmark advantage is most pronounced at the 2.5% labeled-data ratio and decreases as supervision increases. CoGameReg obtains the best blind-well metrics in the synthetic experiment. In the field application, predictions clipped to 0-100% yield a pooled coefficient of determination of 0.5942, a root mean square error of 5.7441 percentage points, and a Pearson correlation of 0.8265 across three blind wells. CoGameReg has the highest Pearson correlation among the four field methods. The results support the use of CoGameReg to integrate sparse well labels with dense seismic attributes, with performance varying across metrics and well locations.
arXiv:2610.12883v1 公告类型：新发布 摘要：非自回归多步预测（NAF）使时间序列基础模型能够通过单次前向传播预测整个未来时域。尽管 RWKV-7 具有强大的序列建模能力，但在我们的 NAF 实验中，它在长预测时域下出现了性能退化。在这一设置中，未来的数据片段被表示为缺失输入，因此循环状态需要在没有新观测值的情况下沿预测时域演化。我们对 RWKV-7 广义 delta 规则的分析揭示了一种机制，我们称之为占位符空转衰减（PID）：当占位符驱动的写入无法提供充分补偿时，反复的状态转移可能会削弱历史信息。逐层分析表明，大多数层的状态注入范数有所降低，而历史区间与预测区间的转移矩阵范数仍然相近，这与 PID 机制一致。基于这一分析，我们提出了 HART，这是一种混合注意力与 RWKV 的时间序列基础模型，交替使用注意力模块和 RWKV 模块，将对历史词元的直接访问与循环时序建模相结合。在 GIFT-Eval 和 TIME-Benchmark 上的受控实验表明，HART 在中长预测时域下优于纯 RWKV，同时在短预测时域下保持相当的性能，并且其综合性能优于规模相近的纯 Transformer。HART 仅有 3600 万个参数，在 GIFT-Eval 上实现了与 Toto-2.0-1B 相当的长时域预测 MASE（0.752 对 0.754），而参数量约为后者的 1/28。
arXiv:2610.12865v1 公告类型：新提交 摘要：现有针对随机梯度 Langevin 蒙特卡洛（SG-LMC）非对数凹采样的收敛性保证通常假设梯度全局 Lipschitz 连续且噪声方差一致有界。即使在基本的采样问题中，这两个假设也可能不成立，放宽它们能更好地反映大型语言模型（LLM）的贝叶斯微调等实际应用。我们通过广义光滑性以及所提出的具有有限 $p$ 阶矩（$1<p\leq 2$）的状态依赖重尾预言机，放宽了这两个假设。该预言机允许在 $p<2$ 时方差无穷大，并涵盖 Blum-Gladyshev（BG-0）条件。我们分别及联合研究这两种放宽情形，并提出三种方法。在梯度全局 Lipschitz 连续的条件下，鲁棒 LMC（R-LMC）通过几何中位数中心化与残差裁剪来处理所提出的预言机。在使用精确梯度时，自适应 LMC（A-LMC）在广义光滑性条件下控制离散化误差。鲁棒自适应 LMC（RA-LMC）通过鲁棒估计和噪声感知步长，同时应对这两种放宽情形。对于这三种方法，我们均在不假设耗散性或函数不等式的情况下，建立了以 Fisher 信息（FI）衡量的非渐近收敛性保证。合成实验展示了 FI 收敛性以及各项机制的益处。针对 LLM 和视觉语言模型的贝叶斯微调实验支持了我们的预言机模型，并表明 RA-LMC 在准确率和不确定性量化方面达到或超越了当前最先进的基线方法。
arXiv:2610.12848v1 公告类型：新投稿 摘要：在最大熵强化学习和行为约束强化学习（RL）中，最优策略由 Boltzmann 分布给出。从该分布中生成样本可能颇具挑战性，因为它通常复杂且具有多模态特征。尽管通过流匹配训练的流策略为表示此类分布提供了强大的框架，但传统上它们依赖训练数据，而在线 RL 中通常无法获得这些数据。为解决这一局限，我们提出不动点流匹配（FFM），一种用于在 RL 中训练和微调流策略的无数据方法。FFM 利用目标流恒等式，将演员更新表述为不动点迭代，并采用流匹配式回归目标，其中评论家对动作的梯度作为回归目标值。我们证明，最优策略是该迭代的一个不动点。通过广泛的实验，我们表明，FFM 能够在离线到在线的设置下有效微调预训练流策略，并能在在线 RL 中从零开始高效、准确地训练流策略。这些结果确立了不动点流匹配作为一种具有理论依据的方法的地位，使其无需训练数据即可学习表达能力强的基于流的策略。
arXiv:2610.12847v1 公告类型：新论文 摘要：模拟器将环境概括为用户可以读取和修改的状态。生成模型无需显式指定即可学习世界的表示，但并未提供编辑该世界状态的内置方式。我们通过 $\textit{可编辑性}$ 来研究生成模型在何种条件下能够充当模拟器：改变模型的潜在表示，是否能产生与编辑模拟器状态后所得到的相同的下一步输出。简而言之，我们的目标是在生成模型的潜在状态与模拟器状态之间找到一种映射，其中模型仅基于观测进行训练，而映射则在之后学习。我们测试了两类映射，两者均从与模拟器状态配对的观测中学习。一类方法训练探针，从模型中解码模拟器状态，然后沿探针所指示的方向进行引导。另一类方法借助逆向映射直接学习这种映射，即使在引导失败的情况下也可能成功。我们在两个截然不同的环境中研究这两类方法：离散棋盘游戏 Othello，以及我们自行构建的 Rayworld——一个通过局部渲染来观测运动圆盘的连续世界，更接近物理场景和视频。我们发现，哪种映射能够成功高度依赖于环境，尽管在所有变体中都可以解码出状态。在 Othello 中，随着游戏被简化，可编辑性平滑下降，即使细微的规则变化也会产生重大影响。在 Rayworld 中，探针无法沿模拟器的连续状态进行引导，但逆向映射仍能根据该状态设定模型。因此，模型可以由一个它可能从未学会的模拟器来驱动。
arXiv:2610.12830v1 公告类型：新提交 摘要：推理后训练利用离线示范和策略生成的采样轨迹，但传统的先监督微调（SFT）、后强化学习（RL）的方法将两者的目标分离，限制了对已收集回答的持续复用。我们提出 \textbf{Cross-Behavior Policy Optimization}（CBPO），通过基于固定参考的目标，使这些数据源能够兼容同一个回归损失。由于该目标不涉及生成策略，正向和负向离线补全、参考回答，以及当前和历史采样轨迹可以共享统一的带奖励标签的记录格式和回放缓冲区。当新回答进入缓冲区时，优化过程仍保持相同的回归目标，无需协调独立的 SFT 和 RL 目标，也不需要存储行为策略概率、进行重要性加权或学习评价器。通过限制可用数据源，可以得到离线、同策略和异策略变体，而 CBPO-Unified 则将它们结合起来。在两种模型规模和四个数学推理基准上，CBPO 在不同基准和数据设置下均取得领先的 \texttt{avg@32} 和 \texttt{pass@32} 表现。在 Qwen3-1.7B 上，相较于 SFT DAPO，CBPO-Unified 在各基准上平均的 \texttt{avg@32} 和 \texttt{pass@32} 分别提高了 2.33 和 4.44 个百分点。在回放消融实验中，相较于不保留历史轨迹的配置，保留历史采样轨迹使这两项指标分别提高了 1.87 和 3.72 个百分点。
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:2610.12815v1 公告类型：新论文 摘要：RDMA 可以降低分布式机器学习系统中的通信开销，但要在生产环境中实现这些收益，仅仅替换 TCP 还不够。大小可变的消息、并发连接、突发性的多对一流量以及服务故障带来了仅靠硬件卸载无法解决的挑战。本文介绍了我们在 ByteDance 基于参数服务器的推理和训练系统中部署 ByteExpress 的经验。ByteExpress 是构建于硬件 RDMA 之上的软件通信层。我们总结了此次部署的三点经验：高效的 RDMA 需要跨应用层和传输层进行缓冲区管理；可扩展的通信需要协调管理连接、接收容量和尚未完成传输的数据；服务可用性需要传输层之上的恢复机制。ByteExpress 结合了支持 RDMA 的内存分配器、零拷贝传输、QP 多路复用、基于接收信用的流量控制、软件发送窗口以及 RPC 辅助恢复机制。在生产环境的推理任务中，与内核 TCP 相比，ByteExpress 将模型参数拉取的平均延迟和 P99 延迟分别降低了 35.7% 和 43.7%，并将 CPU 总使用量降低了 10.69%。在训练任务中，与 TCP 相比，ByteExpress 将模型参数推送的 P99 延迟最多降低了 74.0%，且运行期间未观察到 PFC 暂停。
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:2610.12794v1 公告类型：新论文 摘要：大语言模型（LLM）引导的进化优化已成功改进提示词、程序，乃至整个 AI 系统。大多数优化方法从搜索前沿选择一个候选方案，然后仅依据其失败情况，通过单次 LLM 调用对其进行变异。这种对搜索前沿的聚焦可能使生产环境 AI 系统的优化陷入停滞，因为孤立的失败很难揭示复杂的隐性交互。我们提出 CoCoEvolve（Cortex Code Evolve），这是一种围绕我们所称的“元反思”构建的进化优化器：智能体式变异器（Cortex Code）选择性地从整个进化历史中检索证据，并探查实际运行的系统，然后验证并提出变异方案。我们将 CoCoEvolve 应用于三类生产环境 AI 系统的优化：数据智能体、LLM 驱动的数据流水线和数据工程工作流。在 DABStep 上，CoCoEvolve 位居公开排行榜榜首，在留出的高难度测试集上达到 92.8% 的准确率，而在相同条件下，Meta-Harness 和 GEPA 的准确率分别为 78.4% 和 30.4%。在六项 LLM 数据流水线工作负载中，它均优于所有接受评估的基线。在 data-eng-bench 上，CoCoEvolve 将测试集求解率从 59.1% 提升至 77.4%，超过了最佳基线的 73.6%。它在 HotpotQA 上的准确率也比 GEPA 高出 11 个百分点，同时在成熟的数学和系统基准测试上仍可与专用优化器保持竞争力。同等条件下的消融实验确认，CoCoEvolve 的每个组件都对这些提升有所贡献，体现了元反思在优化生产环境 AI 系统方面的价值。
arXiv:2610.12776v1 公告类型：新论文 摘要：自动微分（AD）是现代机器学习的基石。由于所有 AD 框架都隐式地固定了欧几里得空间的一组基来定义梯度和雅可比矩阵，因此它们原生处理的数学对象都是由一组固定坐标构成的数组。这使得旋转、体积形式和曲率等对象难以表达，而这些对象会出现在例如人工智能应用于科学的场景中。为特定问题定制的变通方案缺乏可复用性。我们证明：（i）单个与基无关的导数算子可以表示多种几何微分运算；（ii）该算子可以完全用现有的 AD 原语来表达。这种可复用的变换可以与 JAX 的批处理、编译和硬件加速功能组合使用。几何结构以代数数据的形式引入。我们通过实验展示了该变换在不同流形算子之间的复用、在时空逆问题处理流程中的组合，以及在不同旋转子子群上的优化中的应用。
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:2610.12766v1 公告类型：新稿 摘要：钙钛矿太阳能电池的机器学习模型通常采用随机数据划分进行评估，而我们发现，大量内部重复数据会使这种评估方式产生系统性的信息泄漏。由于经过整理的器件中，近三分之二与其他器件具有完全相同的配方，随机划分主要奖励的是对历史数据的简单记忆，而非检验真正的物理泛化能力。为解决这一问题，我们梳理了 Perovskite Database 中数据的机构来源，并建立了一套严格的分组划分层级，逐步将重复器件、发表文献和机构相互隔离。对十三种模型的基准评测在所有划分方式下均取得了新的最先进性能，并揭示了一种关键的表现格局转变：树集成模型虽然能通过识别熟悉的配方在随机划分中占据优势，但面对真正未见过的机构和未来发现时，其排名却大幅下滑，而具备组成感知能力的神经网络架构则表现出色。此外，特征重要性分析表明，器件结构和制备方法对最终效率的决定作用，与吸收层化学组成同样显著。这一标准化评估方案使研究界能够区分数据记忆与真正的物理发现。
arXiv:2610.12754v1 Announce Type: new Abstract: We propose NV-Tesseract-AD, a curriculum learning-enhanced conditional diffusion framework for multivariate time-series anomaly detection. Our central hypothesis is that reconstruction difficulty can be explicitly structured along two complementary axes, masking severity and diffusion noise level, and that training from easier to harder tasks along both axes simultaneously stabilizes optimization and improves residual discriminability. We integrate a Transformer-based diffusion backbone with a three-stage dual curriculum over masking ratios and diffusion timesteps, retaining anomalous samples throughout training and relying on structured corruption to drive self-supervised representation learning. We evaluate NV-Tesseract-AD under two regimes: in-distribution training on the TSB-UAD-M corpus and strict zero-shot transfer to ten EasyTSAD benchmark families without per-dataset tuning. The dual-axis curriculum produces masking-invariant reconstruction, confirmed by a near-zero loss gap between complementary masking strategies within the first training stage, and yields an approximately 3.7-fold improvement in F1 and a 56 percent reduction in MSE relative to a non-curriculum baseline. Zero-shot performance reaches F1 = 0.984 on NAB and F1 = 0.948 on NormA. Weaker results on heterogeneous families are traced to unsupervised threshold calibration rather than reconstruction quality, identifying calibration as the dominant remaining detection bottleneck. Accelerated inference via DPM-Solver++ reduces sampling cost by a factor of 3.5 to 5.7 at the dataset level but couples to thresholding behavior, establishing that the sampler and threshold must be validated jointly. Together, these findings demonstrate that aligning diffusion denoising difficulty with curriculum learning principles provides a principled and practical approach for robust anomaly detection in complex multivariate time series.
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:2610.12741v1 Announce Type: new Abstract: We interpret Transformer networks as learned approximations to a hybrid probabilistic-inference procedure that combines max-product Loopy Belief Propagation (LBP) with Expectation--Maximization (EM). In this view, a Transformer performs inference in partially observed Markov Random Fields (MRFs) whose structure and parameters are not given a priori but are inferred from data. We show that self-attention implements a differentiable approximation to max-product LBP on an input-conditioned, dynamically structured MRF, with attention weights encoding learned structural affinities. We further establish an EM correspondence at the level of depth: each layer updates latent representations in an E-step-like forward pass, while the transition $\theta^{(l)} \to \theta^{(l+1)}$ across layers acts as an implicit M-step. Building on this interpretation, we introduce \emph{meta-models}, a design paradigm in which inputs generate instance-specific parameters, unifying mechanisms such as self-attention, mixture-of-experts, and gated units within a single framework. Empirical results support the proposed perspective.
arXiv:2610.12738v1 Announce Type: new Abstract: Forecasting rare, high-cost events in multivariate time-series systems is hard because their precursors are seldom observed, so a purely data-driven model has little to learn from. The engineers who run these systems, however, hold knowledge that precedes such events, information absent from the scarce labels. We call this a semantic prior and inject it into a self-supervised event forecaster through two switchable mechanisms, a covariate mechanism and an attention-bias mechanism, each the identity at its null, so one architecture serves every prior type with no per-dataset architecture search. A prior helps when it supplies information the observed window does not already contain: in a short context an informative prior beats no prior in-domain (pooled h-AUROC over the no-prior baseline: a periodicity clock +14%, p<0.001; a conservation residual +47% at a 0.25 label fraction), and a working prior transfers out-of-domain (+53%), substituting for feature transfer, while priors the window already recovers stay inert and the attention-bias mechanism shows no measurable gain. KITE is a switchable, attributable route to contextual conditioning, and a substrate for scaling toward cross-domain, pretrained time-series foundation world models that carry their governing structure explicitly.
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:2610.12728v1 公告类型：新论文 摘要：参数高效微调已有数百种方法，通常以参数数量和基准测试得分进行比较。我们将它们作为映射来比较。预训练检查点是权重空间中的一个基点，而权重空间方法则是从一个小参数空间出发、以该点为起点的光滑映射。这些映射构成一个切片范畴，其中冻结模型是始对象，全量微调是终对象。映射的三个不变量承担了大部分分析工作。其在基点附近的像，通过一阶像、维数、秩的最小割界，以及该点是光滑点还是锥顶，决定表达能力。其纤维与优化器的度量共同决定训练动力学。基变换移动基点，并涵盖分裂初始化、量化、重启与合并。例如，我们证明：零初始化的 LoRA，以及在所述秩与奇偶性条件下采用默认初始设置的 Householder 反射适配，均从锥顶出发；任何精确初始化都无法使 LoRA 达到秩超过自身秩两倍的更新；在精确算术下，严格正交的方法无论合并多少次，都不会改变权重的奇异值；在不使用 epsilon、权重衰减或裁剪且共用同一调度方案的 Adam 下，从标准零初始化出发的 LoRA+，其轨迹与将缩放系数乘以学习率之比的普通 LoRA 完全一致；采用共享超参数的 Adam，在 LoRA 满秩因子的连续对称群中仅保持带符号置换对称性；激活函数决定缩放适配器何时可以折叠进前面的权重。我们依据从公式中提取的七个坐标，对来自 138 篇论文的 382 种方法进行了分类，其中包含 403 个模拟箭头和 485 个障碍。每个编号结果均给出完整证明，对重新推导的他人成果也均注明出处。
arXiv:2610.12723v1 公告类型：新论文 摘要：通过稀疏、含噪传感器预测物理场的神经代理模型，会读取附近传感器构成的空间窗口以及过去快照组成的历史序列。实践者通过比较拟合后的模型来选择窗口大小和历史长度。然而，这类比较无法区分模型自身的不足与信息缺失所造成的误差。我们以相同观测数据对应的贝叶斯最优误差为基准，衡量窗口与历史之间的权衡。精确的高斯分析表明，对具有局部动力学的场进行含噪或稀疏观测时，最优预测仅通过空间相关性利用远处传感器的信息。在最优解已知的受迫线性场上，仅读取窗口的训练后模型，其误差与最优误差的差距不超过 1%。随后，我们在黏性 Burgers 流和混沌 Kuramoto-Sivashinsky（KS）方程上训练此类模型，并在任何测试评估之前确定所有用于得出结论的比较方案。在除 KS 之外的所有场中，随着预测提前量（即到预测目标的步数）增大，八个快照相较于单个快照所消除的误差比例变小。在 KS 中，当窗口不超过一个相关长度时，这一比例反而变大。这种随预测提前量变化的反转现象在不同传感器密度、三种模型架构以及二维区域中均持续存在。随着阻尼使 Lyapunov 指数降低，这种反转减弱。阻尼也会消除 KS 的线性不稳定性。因此，我们将这种反转与不稳定性联系起来，但并未区分混沌与线性放大的作用。这些结果为代理模型设计提供了两条规则：以场的相关长度为单位确定窗口大小；根据模型将要服务的预测提前量，在窗口与历史之间分配固定的输入预算，因为随着预测提前量增大，最优分配会向更宽的窗口倾斜。在 Burgers 流中，按预测提前量为一步时选定的分配方案，在预测提前量为八步时产生的误差达到该提前量下最优误差的 13.5 倍。
arXiv:2610.12711v1 Announce Type: new Abstract: Federated Large Language Models (FedLLMs) are promising to deliver privacy-preserving, collaborative intelligence. The decentralized training process is vulnerable to Byzantine attacks that corrupt semantic consistency and factual resilience through malicious model updates. This paper presents a new resilience strategy that prevents semantically corrupted knowledge from propagating during federated aggregation. A defense problem is formulated to jointly maximize global accuracy and suppress manipulated model updates, enabling trustworthy knowledge aggregation without accessing private local data. To strengthen the resilience of FedLLMs, we develop a novel Hypergraph Message Passing Graph Autoencoder (HMP-GAE) framework, where low-rank model updates are represented as nodes in a hypergraph and high-order semantic relationships among multiple agents are exploited through iterative message passing. By leveraging historical checkpoints and cross-agent relational reasoning, HMP-GAE identifies and suppresses semantically inconsistent updates while preserving knowledge alignment among agents throughout federated training. HMP-GAE is implemented using Qwen-2.5 and Llama-3.2 as pretrained FedLLM backbones and evaluated on AG News and Yahoo! Answers datasets. Compared to existing defenses, HMP-GAE improves global accuracy by $1.21$\%, reduces final-model classification semantic entropy by $44.08$\%, and lowers perplexity by $13.34$\%. HMP-GAE outperforms the GAE-based defense by improving the accuracy $1.39$\%, reducing the semantic entropy by $54.5$\%, and lowering perplexity by $35.5$\%.
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:2610.12694v1 公告类型：新论文 摘要：分子机器学习使用结构不相交划分来评估分布外性能，其构建规则在最初制定时便被固定在当时分子库的规模上：对成对距离设定阈值，再将完整的连通分量分配到各折中。在根据质谱预测结构的任务中，该领域已普遍采用基于短视的最大公共边子图（MCES）距离、阈值为 10 的单链接聚类。这一惯例是在 29,000 个分子的规模上确立的，但如今公共分子库已包含 227,690 个分子。每对分子的 MCES 计算都是 NP-hard 问题，而分子对的数量又呈平方增长，因此这一准则从未在如此规模上得到评估。我们提出了该准则的优化版本，使其能够以计算上可行的方式应用于公共分子库：通过可容许界和并查集跳过策略，将 $2.6 \times 10^{10}$ 个分子对的计算缩减为仅 $1.15 \times 10^{7}$ 个整数规划问题，计算量减少至原来的 $1/2{,}259$，同时可证明所得划分完全无损。我们发现，在这一规模下，图发生了渗流。单个连通分量包含分子库中 $95.5\%$ 的分子，使任何测试集的占比上限仅为 $4.5\%$。剩余部分不仅规模更小，还存在系统性差异：留出化合物的分子量有 88.1\% 的概率大于训练化合物，因为固定的绝对差异预算对较大分子意味着更严格的相对容差。因此，在大型分子库基准测试中直接使用这一划分规则，会将结构新颖性与向更大、更复杂化学结构的外推混为一谈。
arXiv:2610.12680v1 公告类型：新论文 摘要：背景：新生儿脑电图的不连续性是脑功能障碍的重要标志，但难以通过目视进行量化。爆发间期（IBIs）是不连续性的关键特征，但其定义不易直接用于人工智能检测。目的：引入仅由振幅和持续时间定义的低电压间期（LVIs），作为不依赖上下文的 IBIs 替代指标，开发 LVIs 的人工智能检测方法，并评估其作为脑损伤标志物的临床效用。方法：我们开发了用于检测 LVIs 的 ConvNeXt 模型，并将其性能与现有方法进行比较。随后，我们在四种应用场景中检验了所检测 LVIs 的临床相关性：1）在开放基准数据集上对脑电图背景活动异常进行分级；2）对临床缺氧缺血性脑病（HIE）进行分级，并与正常脑电图数据集进行比较；3）预测癫痫发作；4）分析其与癫痫发作负荷的关联。结果：为开展训练和验证，我们审阅了来自 123 名新生儿、按单通道累计 654 小时的脑电图，并标注了 51,162 个 LVIs。在留出验证集上，ConvNeXt 的曲线下面积（AUC）为 0.996，Matthews 相关系数为 0.926。模型与阅图者之间的一致性高于阅图者之间的一致性（$\Delta \kappa = 0.025$，95% 置信区间：0.007 至 0.045，$p<0.001$），达到与专家水平相当的表现。在包含 444 名新生儿的合并数据集中，LVI 时间占比与脑电图分级呈强相关（Spearman $\rho=0.801$），与 HIE 分级呈中等相关（$\rho=0.384$），与癫痫发作负荷呈较弱相关（$\rho=0.244$），并具有较强的癫痫发作预测能力（AUC $=0.872$）。结论：人工智能对 LVIs 的检测达到与专家水平相当的表现，而定量特征未达到这一水平。自动化 LVI 量化为新生儿脑病提供了一种稳健的临床生物标志物，与脑电图背景活动、HIE 严重程度及癫痫发作风险相关。
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:2610.12668v1 公告类型：新论文 摘要：语言模型智能体越来越多地在长交互流中运行，需要通过持续学习在线适应分布偏移。然而，现有智能体运行框架通常要么仅通过前向信息流更新状态，要么依赖额外的回放数据或验证数据来优化运行框架。为此，我们提出 ReHarness，这是一种面向智能体持续学习的循环式运行框架，在严格在线的流式协议下，以反向修订补充前向读写。我们探究：运行框架状态中的哪些信息本可以通过不同的方式记录，从而为处理流中的当前实例提供更有效的参考。具体而言，在每个实例处理完成后，我们利用产生的轨迹作为反馈，修订为该实例检索出的反思内容，从而使反馈沿文本形式的运行框架状态反向传播。在三个持续学习基准和三种语言模型骨干上，ReHarness 均持续优于强大的原始上下文基线、基于记忆和反思的方法以及提示词优化方法。进一步分析表明，ReHarness 能够生成有来源依据且有意义的修订，在分布偏移下纠正过度泛化或过时的经验，并提升持续学习过程中的稳定性和可塑性。
arXiv:2610.12658v1 Announce Type: new Abstract: We develop a data-driven sensor-placement framework and apply it to aerosol monitoring in the United Kingdom (UK). A trained convolutional Gaussian neural process is embedded within the DeepSensor optimisation framework. Genetic optimisation with lazy evaluation and caching reduces mean computational time by 60\% (10$\pm$3 versus 25$\pm$7 minutes) while producing sequential placement performance comparable to the greedy solver. Using the CAMS reference field to construct a retrospective oracle benchmark, Joint NLL improves by about 22\% for Joint-MI-driven genetic placement after five locations for aerosol stations are identified. Seasonal and wind-regime analyses show that agreement between model-based criteria and oracle benchmarks varies with meteorological transport conditions. These results show a computationally efficient approach for data-informed expansion of environmental observing systems.
arXiv:2610.12657v1 公告类型：新提交 摘要：高效自回归生成日益受到键值（KV）缓存内存占用的限制，该占用随序列长度、批量大小和模型深度线性增长。现有方法，如 Multi-Query Attention（MQA）、Grouped-Query Attention（GQA）、Multi-Head Latent Attention（MLA）和 Tensor Product Attention（TPA），通过不同的权衡降低这一成本，但往往会牺牲表达能力、需要额外的位置编码机制，或施加限制性的低秩结构。我们提出 Kronecker Product Attention（KPA），将键和值的激活表示为若干 Kronecker 积之和，每个积由紧凑且依赖于词元的矩阵因子构成。KPA 是 TPA 的严格推广：秩一向量外积分解是 Kronecker 划分的一种退化情形，而非退化矩阵因子则能表示更丰富的块结构模式。我们进一步表明，与 Kronecker 表示相匹配的结构化旋转嵌入算子，仅旋转紧凑因子即可保留 RoPE 的相对位置性质，从而无需像 MLA 那样使用解耦的 RoPE 分支。实验结果表明，在相同训练条件下，KPA 在四种模型规模和九项下游基准测试中的表现达到或超过 TPA、MHA、MQA、GQA 和 MLA，同时实现高达 10 倍的 KV 缓存压缩。Kronecker 注意力头划分是一个可调超参数，使实践者能够在设计阶段权衡压缩率与准确率，而无需更改架构。
arXiv:2610.12651v1 公告类型：新论文 摘要：执行多步骤工作流的 LLM 智能体会以两种根本不同的方式失败：一种是可恢复的失误，可通过进一步尝试修正；另一种是知识缺口，无法通过重试解决。专家反馈可以弥补后者，但专家的注意力资源稀缺。因此，关键挑战不在于是否寻求反馈，而在于在哪些环节反馈能够改变结果。现有方法要么避开专家反馈，任由智能体强化自身错误；要么在任务层面分配反馈，却未能识别会对结果产生实质影响的决策。我们提出 SynAct，这是一种主动学习框架，将专家注意力视为关键制约因素，并在单个决策层面分配反馈。SynAct 根据重复执行轨迹中所选动作与最终结果之间的互信息对决策排序，在价值最高的决策处向专家征询意见，并将所得修正提炼为可迁移的规则，使智能体在部署时无需访问专家即可复用这些规则。我们从理论上刻画了这一信号在有限次重复采样下能够识别的内容：在 k 次执行轨迹、每个动作至少有 m 次观测的条件下，结果恒定的决策的互信息为零，而能够区分结果的决策的互信息至少为 H2(m/k) 比特，因此，对于区间 (0,H2(m/k)) 内的任意阈值 t，均可实现精确分离。这也解释了为什么在所有执行轨迹均失败时，基于成功与否的采集策略可能失去辨别能力：即使失败的运行产生了不同结果，成功与否这一变量仍保持恒定。在涵盖三个模型家族的四个模型和四项智能体基准测试中，SynAct 在全部 16 种评估设置下均提高了留出集准确率，平均提升 11.7 个百分点。同一信号还可迁移至推理时路由：在 16 种设置中的 15 种中，SynAct 优于预算相同的级联路由器、基于不确定性的路由器和专家评判路由器；同时，一次有针对性的修正优于额外进行七次无针对性的重试。
arXiv:2610.12638v1 Announce Type: new Abstract: Day-ahead electricity forecasts can help building operators schedule heating, ventilation, and other services around anticipated demand. Choosing a forecasting model also requires balancing accuracy with its forecast-time inputs, local training, and computational requirements. We benchmark statistical, machine-learning, recurrent neural-network, and pretrained TimesFM models for a university building. We use chronological validation and evaluate all configurations on the same 57 test days, comparing accuracy across workdays, weekends, and holidays alongside local fitting and forecasting time. Weather covariates use retrospective observations. TimesFM 3.0 with covariates achieves the lowest overall errors, reducing mean absolute error by approximately 43\% relative to weekly seasonal-naive forecasting. XGBoost remains competitive, although paired daily comparisons do not establish a clear mean-absolute-error advantage for TimesFM. TimesFM also performs strongly using only past electricity consumption. Adding covariates improves its holiday forecasts but slightly worsens workday and weekend forecasts, leaving overall mean absolute error almost unchanged. TimesFM requires no building-specific weight training, while XGBoost offers rapid local fitting and inference. The evaluated LSTM takes longer to fit without outperforming either approach. An exploratory historical evaluation reveals seasonal variation, while a separate weather-source comparison produces only small changes in the selected TimesFM configuration's accuracy. These results support pretrained TimesFM as a strong initial forecasting baseline for this building, with XGBoost as a competitive locally fitted alternative. Covariates are particularly useful when improved holiday forecasts justify the additional input preparation.