cs.RO / 1 / 2610.00008
Bounded-Fidelity Sim-as-Demo-Stage: Mocap Handoff for Governance Benchmarks
有界保真度的模拟作为演示阶段：运动捕捉交接用于治理基准
Abstract
Sim-to-real research pursues physics fidelity as a primary objective: simulators are judged by how closely they reproduce real-world contact dynamics. For governance benchmarking of LLM-driven robots, where the simulator demonstrates that an admission/policy/contract/audit pipeline behaves correctly, contact fidelity at object handoffs (grasp, carry, place) becomes a liability: contact-force integration noise injects audit-chain divergence that is structurally unrelated to the governance property under test. We propose bounded-fidelity sim-as-demo-stage, a design pattern that suppresses contact physics within explicitly bracketed handoff envelopes while preserving full dynamics elsewhere. The construction uses MuJoCo's mocap-body primitive driven by a 220-line Python adapter that the governance bridge invokes via structured intents. We formalise audit-chain stability as byte-equality of the hashed event log across replays and identify two structural envelope properties that imply it. Across N=1000 replays per posture, the mocap variant produces one distinct audit-chain hash (1000/1000 byte-identical; Wilson 95% CI [0.997, 1.000]); the contact-force baseline produces 584 distinct hashes (993/1000 diverged; CI [0.987, 0.998]). A timestep sweep (1, 2, 5, 10 ms) shows the divergence is structural, not a tuning artefact: it stays at 0.985 at every timestep. Envelope-edge timing jitter (+/-10 simulation steps, 1,400 replays) produces 0 divergence, and audit chains remain byte-equal across K in {1, 2, 3} sequentially handed-off objects (1,500 replays) with sub-linear per-pick-and-place overhead. The pattern gives benchmark designers audit-chain reproducibility at near-zero engineering cost; we also map where it is harmful (sim-to-real validation, policy training, contact-rich tasks) so it is not mis-deployed.
Chinese Translation
模拟到现实的研究将物理保真度作为主要目标：模拟器的评判标准是其与现实世界接触动态的相似程度。在以大型语言模型（LLM）驱动的机器人治理基准中，模拟器展示了入场/政策/合同/审计流程的正确行为，物体交接时的接触保真度（抓取、携带、放置）成为一种负担：接触力的积分噪声引入了与待测试的治理属性结构上无关的审计链偏差。我们提出了有界保真度的模拟作为演示阶段（bounded-fidelity sim-as-demo-stage），这是一种设计模式，在明确界定的交接范围内抑制接触物理，同时在其他地方保留完整的动态特性。该构造使用MuJoCo的运动捕捉体（mocap-body）原语，由一个220行的Python适配器驱动，治理桥通过结构化意图调用该适配器。我们将审计链的稳定性形式化为重放过程中哈希事件日志的字节相等性，并识别出两种结构性封闭属性以暗示这一点。在每个姿态的N=1000次重放中，运动捕捉变体产生一个独特的审计链哈希（1000/1000字节相同；Wilson 95%置信区间[0.997, 1.000]）；而接触力基线产生584个不同的哈希（993/1000发生偏差；置信区间[0.987, 0.998]）。时间步长的变化（1、2、5、10毫秒）显示偏差是结构性的，而不是调优伪影：在每个时间步长上保持在0.985。封闭边缘的时序抖动（+/-10个模拟步骤，1400次重放）产生0偏差，并且在K为{1, 2, 3}的顺序交接物体中，审计链在1500次重放中保持字节相等，且每次抓取和放置的开销呈亚线性增长。该模式为基准设计者提供了接近零工程成本的审计链可重复性；我们还映射了其有害的情境（模拟到现实验证、政策训练、接触丰富任务），以避免错误部署。
cs.RO / 2 / 2610.00057
Multi-Reference Path Tracking Control for an Agricultural Tractor with Nonlinear Model Predictive Control
基于非线性模型预测控制的农业拖拉机多参考路径跟踪控制
Abstract
Guiding a tractor along a predefined reference path is a key component of precision agriculture. This study develops a path tracking controller based on Nonlinear Model Predictive Control, which incorporates multiple segments of a piecewise-linear reference path directly into the objective function. In addition, methods for selecting viable reference segments from the full path are presented. The control system is evaluated during a field test with a tractor controlled via the Tractor Implement Management steering interface. The NMPC solver converged on average after 3.45 ms and tracked the curved reference path with a mean absolute cross-track error of 6.1 cm.
Chinese Translation
沿着预定义参考路径引导拖拉机是精准农业的关键组成部分。本研究开发了一种基于非线性模型预测控制（Nonlinear Model Predictive Control, NMPC）的路径跟踪控制器，该控制器将分段线性参考路径的多个段直接纳入目标函数。此外，本文还提出了从完整路径中选择可行参考段的方法。在通过拖拉机实施管理（Tractor Implement Management）转向接口控制的拖拉机进行的田间测试中评估了控制系统。NMPC求解器平均在3.45毫秒后收敛，并以6.1厘米的平均绝对横向误差跟踪了曲线参考路径。
cs.RO / 3 / 2610.00065
Probabilistic Plan Legibility with Off-the-shelf Planners
基于现成规划器的概率性计划可读性
Abstract
Legible planning is the creation of plans that best disambiguate their goals from a set of other candidates from an observer's perspective. In this paper we propose a method for legible planning for arbitrary PDDL domains, by extending previous research on legibility to classical planning without requiring to construct ad-hoc planners. We also discuss how the observer perspective may be estimated through a second order theory of mind that connects the planner's and the observer's task spaces. Our solution can for example be deployed in human-robot teaming scenarios, where an autonomous robot in a team can implicitly communicate its goal by producing legible plans. We present benchmark results on several PDDL planning domains. Our results generally show that plan legibility is a trade-off with plan efficiency, however, not all planning domains allows to increase legibility in the same way and a regularizing factor to balance legibility and efficiency was proved necessary.
Chinese Translation
可读性规划是从观察者的角度创建最佳区分其目标与其他候选目标的计划。在本文中，我们提出了一种适用于任意 PDDL（规划领域描述语言）领域的可读性规划方法，通过将之前关于可读性的研究扩展到经典规划，而无需构建特定的规划器。我们还讨论了如何通过一种二阶心智理论来估计观察者的视角，该理论将规划者和观察者的任务空间连接起来。我们的解决方案可以在人机协作场景中应用，例如，在团队中的自主机器人可以通过生成可读性计划来隐式传达其目标。我们在多个 PDDL 规划领域上展示了基准结果。我们的结果通常表明，计划可读性与计划效率之间存在权衡，然而，并非所有规划领域都允许以相同的方式提高可读性，因此证明了需要一个正则化因子来平衡可读性和效率。
cs.RO / 4 / 2610.00198
HumanoidTTT: Test-Time Capability Reuse for Efficient Humanoid Control
HumanoidTTT：高效人形控制的测试时能力重用
Abstract
Recent advances in motion generation and whole-body tracking have enabled humanoid robots to execute increasingly diverse motions, yet the same motion capabilities may be requested repeatedly during continual deployment. Reliable reuse is challenging because intervening motions can change the robot's entry state, making previously successful motions unsafe to replay blindly. Meanwhile, validated capabilities accumulate during deployment, while bounded storage requires deciding which ones are worth retaining. To address these challenges, we present HumanoidTTT, a framework for test-time capability reuse in continual humanoid control. Specifically, we introduce Selective Full-Motion Reuse, which authorizes direct reuse of validated complete motions only from certified applicable entry states, allowing accepted reuse to bypass fresh generation. We further introduce Test-Time Capability Consolidation, which adapts which qualified capabilities persist in a bounded Full-Motion Store using subsequent deployment reuse as feedback. Experiments demonstrate zero unsafe accepts and a 16.4$\times$ end-to-end speedup over fresh generation, while online consolidation improves avoided generator calls by 13.2 per 200 requests over its frozen counterpart. Overall, HumanoidTTT enables reliable and efficient reuse of validated motion capabilities while adaptively retaining useful capabilities throughout continual deployment. Code: https://github.com/AIGeeksGroup/HumanoidTTT. Website: https://aigeeksgroup.github.io/HumanoidTTT.
Chinese Translation
近年来，运动生成和全身跟踪的进展使得人形机器人能够执行越来越多样化的动作，然而在持续部署过程中，可能会反复请求相同的运动能力。由于干扰动作可能改变机器人的初始状态，因此可靠的重用变得具有挑战性，使得之前成功的动作在盲目重放时可能变得不安全。同时，在部署过程中验证的能力不断积累，而有限的存储空间要求我们决定哪些能力值得保留。为了解决这些挑战，我们提出了HumanoidTTT，一个用于持续人形控制的测试时能力重用框架。具体而言，我们引入了选择性全动作重用（Selective Full-Motion Reuse），该方法仅允许从经过认证的适用初始状态直接重用经过验证的完整动作，从而使得接受的重用能够跳过新生成的过程。我们进一步引入了测试时能力整合（Test-Time Capability Consolidation），该方法利用后续部署重用作为反馈，调整哪些合格能力在有限的全动作存储中得以保留。实验结果表明，重用过程中没有发生不安全的接受，并且相较于新生成，整体端到端速度提升了16.4倍，而在线整合在每200个请求中减少了13.2次生成器调用，相较于其冻结版本。总体而言，HumanoidTTT实现了经过验证的运动能力的可靠和高效重用，同时在持续部署过程中自适应地保留有用的能力。代码链接：https://github.com/AIGeeksGroup/HumanoidTTT。网站链接：https://aigeeksgroup.github.io/HumanoidTTT。
cs.RO / 5 / 2610.00220
Real-Time Whole-Body Safe Motion Generation for Multi-Segment Tendon-Driven Continuum Robots
多段腱驱动连续机器人实时全身安全运动生成
Abstract
Real-time motion generation for tendon-driven continuum robots requires accurate modeling of nonuniform bending and whole-body collision avoidance. This paper presents a unified actuation-space framework for planar multi-segment tendon-driven continuum robots. An energy-based variable-curvature model captures spatially varying tendon spacing and bending stiffness and provides analytical Jacobians for differential inverse kinematics and safety monitoring. A multipoint CBF-QP enforces backbone clearance under obstacle motion and actuation-velocity bounds, while its decision dimension depends only on the number of independently actuated segments. The model closely agrees with GVS references, with a maximum curvature error of $5.223 \times 10^{-2}\,\mathrm{m}^{-1}$. Over 100 MuJoCo trials, the proposed method achieves collision-free success rates of 96% and 100% in static and dynamic scenarios, respectively, compared with approximately 60% and 80% without CBF constraints. Hole-traversal tests further demonstrate safe motion in constrained environments. With 600 backbone monitoring points, the mean control-step time is 6.66 ms, demonstrating real-time whole-body safe motion generation.
Chinese Translation
腱驱动连续机器人的实时运动生成需要对非均匀弯曲进行准确建模以及实现全身碰撞避免。本文提出了一种针对平面多段腱驱动连续机器人的统一驱动空间框架。基于能量的可变曲率模型捕捉了空间上变化的腱间距和弯曲刚度，并为微分逆运动学和安全监测提供了解析雅可比矩阵。多点CBF-QP在障碍物运动和驱动速度限制下强制执行脊柱间隙，其决策维度仅依赖于独立驱动段的数量。该模型与GVS参考值高度一致，最大曲率误差为$5.223 imes 10^{-2}\, ext{m}^{-1}$。在超过100次MuJoCo试验中，所提方法在静态和动态场景下分别实现了96%和100%的无碰撞成功率，而在没有CBF约束的情况下，成功率约为60%和80%。孔穿越测试进一步证明了在受限环境中的安全运动。通过600个脊柱监测点，平均控制步长时间为6.66毫秒，展示了实时全身安全运动生成的能力。
cs.RO / 6 / 2610.00317
DriftOPD: Sequence-Level Reverse-KL Distillation for One-Step VLA Policies
DriftOPD：用于一步式 VLA 策略的序列级反向 KL 蒸馏
Abstract
Vision-Language-Action (VLA) models increasingly rely on action experts that generate short action chunks under receding-horizon control. While chunk-level training is convenient across robot embodiments, it optimizes local action likelihood without explicitly accounting for long-horizon task success. Sequence-level reinforcement learning can address this limitation, but typically requires policy rollouts and closed-loop interaction, which are costly for real-robot manipulation. We introduce DriftOPD, a teacher-free, rollout-free framework for sequence-level on-policy distillation of continuous VLA action experts. We show that the sequence-level reverse Kullback-Leibler (KL) divergence decomposes into a chunk-level reverse-KL term and a future-potential term that captures the long-horizon effect of the current action. DriftOPD optimizes these two terms using a one-step drifting objective and a Q-function critic learned from offline demonstrations, respectively, enabling sequence-level optimization with only offline data and one-step action generation. Across multiple VLA architectures in simulation and real-world manipulation, DriftOPD generally outperforms existing one-step distillation baselines while achieving task success performance comparable to multi-step teacher policies. These results demonstrate that long-horizon behavior can be effectively distilled into one-step VLA action experts without online interaction or a separate teacher.
Chinese Translation
视觉-语言-行动（VLA）模型越来越依赖于在递归视野控制下生成短动作片段的行动专家。尽管片段级训练在机器人实现中方便，但它优化了局部动作的可能性，而未明确考虑长视野任务的成功。序列级强化学习可以解决这一限制，但通常需要策略展开和闭环交互，这在真实机器人操作中成本高昂。我们提出了 DriftOPD，这是一种无教师、无展开的框架，用于连续 VLA 行动专家的序列级在线策略蒸馏。我们展示了序列级反向 Kullback-Leibler（KL）散度可以分解为片段级反向 KL 项和捕捉当前动作长视野效应的未来潜力项。DriftOPD 使用一步漂移目标和从离线演示中学习的 Q 函数评论员来优化这两个项，从而仅使用离线数据和一步动作生成实现序列级优化。在多种 VLA 架构的仿真和真实世界操作中，DriftOPD 通常优于现有的一步蒸馏基线，同时在任务成功表现上与多步教师策略相当。这些结果表明，长视野行为可以有效地蒸馏成一步式 VLA 行动专家，而无需在线交互或单独的教师。
cs.RO / 7 / 2610.00330
Retrospective Open-Vocabulary Memory for Long-Term Object Search
长期物体搜索的回顾性开放词汇记忆
Abstract
Long-term object search requires learning where objects usually appear from repeated but uneven observations of a changing environment. We formulate retrospective open-vocabulary memory as probabilistic inference from censored observations, where the key idea is to reason with evidence per opportunity: a detection or non-detection should influence belief only in proportion to the robot's opportunity to observe the corresponding location. We introduce ECROM, which uses this principle to estimate long-term prevalence for concepts specified only at query time and converts the resulting belief directly into an active-search prior. To evaluate this problem, we introduce a controlled long-term benchmark in ten HM3D homes that independently varies object placement and observation opportunity across repeated traversals. ECROM improves support-level AP on held-out queries by 4.5 points and search SPL by 4.2 points over the strongest competing memory in each metric. The benchmark, dataset, and code will be open-sourced.
Chinese Translation
长期物体搜索需要从对变化环境的重复但不均匀的观察中学习物体通常出现的位置。我们将回顾性开放词汇记忆表述为从被审查的观察中进行的概率推理，关键思想是根据每次机会进行证据推理：检测或未检测的结果应仅按机器人观察相应位置的机会比例影响信念。我们引入了ECROM，它利用这一原则来估计仅在查询时指定的概念的长期普遍性，并将得到的信念直接转换为主动搜索先验。为了评估这一问题，我们在十个HM3D家庭中引入了一个受控的长期基准，该基准在重复遍历中独立变化物体放置和观察机会。ECROM在保留查询的支持水平AP上提高了4.5个百分点，在搜索SPL上提高了4.2个百分点，相较于每个指标中最强的竞争记忆。该基准、数据集和代码将开放源代码。
cs.RO / 8 / 2610.00355
IndoorBEV: A Lightweight Real-Time LiDAR BEV Perception System for Indoor Mobile Robots
IndoorBEV：一种轻量级实时室内激光雷达鸟瞰视图感知系统用于移动机器人
Abstract
Efficient indoor LiDAR perception is challenging because mobile robots must understand cluttered three-dimensional environments under strict latency and memory constraints. Existing point-based and voxel-based methods often incur substantial computational overhead, whereas conventional bird's-eye-view (BEV) representations improve efficiency at the cost of discarding vertical geometric information. We present IndoorBEV, a lightweight LiDAR perception framework that mitigates this tradeoff through a height-aware BEV representation and geometry-conditioned feature fusion. IndoorBEV summarizes the vertical point distribution in each BEV cell using statistical height features and multi-frequency height encoding, allowing informative three-dimensional cues to be processed efficiently by two-dimensional convolutions. A lightweight encoder then integrates complementary geometric features with multi-scale local representations and compact global scene context. Decoupled dense prediction heads jointly produce semantic BEV maps and oriented object bounding boxes. IndoorBEV contains only 0.6M parameters and requires 2.3 MB of model storage. On an NVIDIA AGX Orin, it uses 21.52 MB of GPU memory per inference and achieves a mean latency of 169.6 ms under a 200 ms perception deadline, with a deadline miss ratio of 1.8\%. Evaluations on simulated scenes, real-world robot scans, and an open-source indoor point-cloud dataset demonstrate a favorable tradeoff among perception accuracy, latency, and memory consumption. These results indicate that explicitly encoding vertical geometry within a compact BEV representation provides an effective approach to resource-efficient indoor LiDAR perception.
Chinese Translation
高效的室内激光雷达感知面临挑战，因为移动机器人必须在严格的延迟和内存限制下理解杂乱的三维环境。现有的基于点和基于体素的方法通常会产生大量的计算开销，而传统的鸟瞰视图（BEV）表示虽然提高了效率，却以丢弃垂直几何信息为代价。我们提出了IndoorBEV，一个轻量级的激光雷达感知框架，通过高度感知的BEV表示和几何条件特征融合来缓解这一权衡。IndoorBEV使用统计高度特征和多频率高度编码总结每个BEV单元中的垂直点分布，使得有信息的三维线索能够通过二维卷积高效处理。接着，轻量级编码器将互补的几何特征与多尺度局部表示和紧凑的全局场景上下文整合在一起。解耦的密集预测头共同生成语义BEV地图和定向物体边界框。IndoorBEV仅包含0.6M参数，并需要2.3 MB的模型存储。在NVIDIA AGX Orin上，它每次推理使用21.52 MB的GPU内存，并在200 ms的感知截止时间内实现平均延迟169.6 ms，截止时间错过率为1.8%。在模拟场景、真实机器人扫描和开源室内点云数据集上的评估显示了感知精度、延迟和内存消耗之间的良好权衡。这些结果表明，在紧凑的BEV表示中显式编码垂直几何信息为资源高效的室内激光雷达感知提供了一种有效的方法。
cs.RO / 9 / 2610.00360
DexPolicy: Scheduled Exploration for Trajectory-Guided Dexterous Manipulation
DexPolicy：用于轨迹引导的灵巧操作的调度探索
Abstract
Reinforcement learning (RL) for dexterous manipulation must discover finger-object contacts and then control the object precisely; the action noise that serves the first goal can interfere with the second. In trajectory-guided settings such as ViViDex, where RL refine hand-object trajectories from human video, our baseline PPO runs end near their initial action noise after 5M steps, motivating explicit control of exploration scale. DexPolicy makes that scale an explicit function of training steps, annealing from broad to narrow exploration while holding loss, architecture, reward, and optimizer settings fixed. We study three policy-optimization settings: PPO, critic-free GRPO continuation, and a flow-parameterized PPO variant (FPO). Across five YCB objects and three training seeds, mean deterministic Target success rises from 49.4% to 68.1% (FPO), 14.1% to 45.4% (GRPO), and 32.0% to 35.7% (PPO). On a RealMan RM75 arm with an Inspire/RH56 hand, 360 trials over three objects raise mean Target success from 25.0% to 85.0% (FPO), 10.0% to 63.3% (GRPO), and 8.3% to 43.3% (PPO), with one trained model per object-method condition. PPO component screening favors noise control over the tested optimizer contraction; the selected PPO schedule yields higher mean Target success than linear decay with the same endpoints on three tested objects. Training return, deterministic Target success, and tolerance to execution noise dissociate; schedules should therefore be judged by terminal task success under the intended execution conditions, per task and policy-optimization setting. Code: https://github.com/AIGeeksGroup/DexPolicy. Website: https://aigeeksgroup.github.io/DexPolicy.
Chinese Translation
灵巧操作的强化学习（RL）必须发现手指与物体的接触，然后精确控制物体；用于实现第一个目标的动作噪声可能会干扰第二个目标。在轨迹引导的设置中，例如 ViViDex，其中 RL 从人类视频中优化手-物体轨迹，我们的基线 PPO 在 500 万步后接近其初始动作噪声，这促使我们明确控制探索规模。DexPolicy 将这一规模设定为训练步骤的显式函数，从广泛探索逐渐收敛到狭窄探索，同时保持损失、架构、奖励和优化器设置不变。我们研究了三种策略优化设置：PPO、无评论员的 GRPO 继续和一种流参数化的 PPO 变体（FPO）。在五个 YCB 物体和三个训练种子上，平均确定性目标成功率从 49.4% 提升至 68.1%（FPO），14.1% 提升至 45.4%（GRPO），以及 32.0% 提升至 35.7%（PPO）。在配备 Inspire/RH56 手的 RealMan RM75 臂上，针对三个物体的 360 次试验将平均目标成功率从 25.0% 提升至 85.0%（FPO），10.0% 提升至 63.3%（GRPO），以及 8.3% 提升至 43.3%（PPO），每个物体-方法条件下训练一个模型。PPO 组件筛选更倾向于噪声控制而非测试的优化器收缩；所选的 PPO 调度在三个测试物体上产生的平均目标成功率高于具有相同端点的线性衰减。训练回报、确定性目标成功率和对执行噪声的容忍度是不同的；因此，调度应根据在预期执行条件下的最终任务成功率进行评估，具体取决于任务和策略优化设置。代码：https://github.com/AIGeeksGroup/DexPolicy。网站：https://aigeeksgroup.github.io/DexPolicy。
cs.RO / 10 / 2610.00368
DeepJEPA: Scaling World Models from Within
DeepJEPA：从内部扩展世界模型
Abstract
World-model planners typically scale outward by rolling farther, sampling more trajectories, or optimizing longer, while assigning the same computation to every imagined transition. We show that making every transition uniformly deeper wastes computation and can degrade planning because useful refinement is concentrated at a small set of decision-critical events. We introduce DeepJEPA, a weight-tied joint-embedding predictive world model that treats transition depth as an inner test-time scaling axis and learns when another recurrent update is worth computing for each candidate and rollout step. Across five visual-control settings, DeepJEPA improves or matches the strongest fixed-depth planner while averaging only 1.00-1.26 updates per transition. Its additional computation concentrates at contact onset and sustained object interaction, where latent corrections can change which candidates enter the planner's elite set and which action is selected. Representation probes further show that improved planning does not require uniformly better object-state decodability. DeepJEPA therefore reframes world-model scaling as a problem of allocating internal computation where it can change the planner's decision: think deeper at decision-critical transitions instead of making every rollout uniformly deeper or longer.
Chinese Translation
世界模型规划器通常通过更远的滚动、更多的轨迹采样或更长时间的优化来向外扩展，同时对每个想象的过渡分配相同的计算资源。我们表明，使每个过渡均匀加深会浪费计算资源，并可能降低规划效果，因为有用的细化集中在一小组决策关键事件上。我们引入了DeepJEPA，这是一种权重绑定的联合嵌入预测世界模型，它将过渡深度视为内部测试时间扩展轴，并学习在每个候选和滚动步骤中何时值得进行另一次递归更新。在五个视觉控制设置中，DeepJEPA在平均每个过渡仅进行1.00-1.26次更新的情况下，改善或匹配了最强的固定深度规划器。其额外的计算集中在接触开始和持续的物体交互上，在这些地方潜在的修正可以改变哪些候选进入规划器的精英集以及选择哪个动作。表示探测进一步表明，改进的规划并不需要均匀更好的物体状态解码能力。因此，DeepJEPA将世界模型扩展重新框架为一个分配内部计算的问题，以便在能够改变规划器决策的地方进行更深入的思考：在决策关键的过渡中进行更深的思考，而不是使每次滚动均匀地加深或延长。
cs.RO / 11 / 2610.00404
Whole-Body Aerial Grasping and Lifting via Partial Visual Observations
通过部分视觉观察实现全身空中抓取与提升
Abstract
Aerial grasp-and-lift tasks require whole-body coordination across approach, acquisition, and lifting under partial target observations. Early approach failures can limit exposure to later task stages during training, while changing visibility complicates alignment and closure timing during execution. We present a recurrent teacher-student framework that learns a single policy in simulation to jointly command flight, arm motion, and gripper closure without an explicit task-phase input. A privileged teacher learns through reinforcement learning with a critical-state curriculum that exposes acquisition and lifting states before connecting them to normal approach trajectories. Its behavior is distilled into a recurrent visual student that replaces privileged target states with dual-view point clouds and proprioception, integrating observation history for closed-loop control. A dedicated closure objective supervises closure timing from sustained model-defined readiness sequences. Training and primary evaluation use a simulated acquisition-and-payload model with condition-triggered latching, virtual attachment, and wrench-based payload loading for short-distance lifting. Across 8,996 completed simulation episodes under this model, the frozen student achieves full-task success rates of 99.97%, 97.14%, and 95.84% under nominal, physics/control-randomized, and additional camera-randomized conditions, respectively. The nominal latch-count-weighted mean of per-seed 90th-percentile alignment errors at acquisition is 8.12 mm.
Chinese Translation
空中抓取与提升任务需要在部分目标观察下进行全身协调，包括接近、获取和提升。早期的接近失败可能限制训练过程中对后续任务阶段的接触，而变化的可见性则使得执行过程中的对齐和闭合时机变得复杂。我们提出了一种递归教师-学生框架，该框架在仿真中学习单一策略，以共同指挥飞行、手臂运动和夹具闭合，而无需明确的任务阶段输入。特权教师通过强化学习学习，采用关键状态课程，使获取和提升状态在连接到正常接近轨迹之前得到暴露。其行为被提炼成一个递归视觉学生，该学生用双视点云和本体感知替代特权目标状态，整合观察历史以实现闭环控制。一个专门的闭合目标监督闭合时机，基于持续的模型定义的准备序列。训练和主要评估使用一个模拟的获取和负载模型，该模型具有条件触发的锁定、虚拟附着和基于扭矩的负载加载，适用于短距离提升。在该模型下，经过8,996个完成的仿真回合，冻结的学生在名义、物理/控制随机化和额外相机随机化条件下的全任务成功率分别达到了99.97%、97.14%和95.84%。在获取阶段，每个种子的第90百分位对齐误差的名义锁定计数加权平均为8.12毫米。
cs.RO / 12 / 2610.00438
Towards a General Humanoid Loco-Manipulation Model via Egocentric Whole-Body Human Data Pretraining
通过自我中心的全身人类数据预训练构建通用类人运动-操作模型
Abstract
Humanoid whole-body manipulation has advanced rapidly, enabling policies to coordinate locomotion, posture, bimanual interaction, and dexterous hand movements. Meanwhile, egocentric human videos provide diverse examples of everyday interactions across objects and scenes, offering scalable supervision without robot operation. However, existing supervision from these videos provides limited coverage of whole-body movement and coordination with hand-object interaction, while obtaining such supervision through humanoid teleoperation is also costly and difficult to scale. We therefore explore how human experience can support scalable learning of humanoid loco-manipulation. To support this study, we introduce HumanVerse-500, a 500-hour dataset of diverse human loco-manipulation behaviors in open-world environments, collected with a lightweight wearable system that synchronizes egocentric video with body and hand motion. Building on this dataset, we develop $\lambda_0$, a whole-body humanoid vision-language-action policy, through three-stage training that first learns interaction from diverse egocentric datasets, then coordinates body and hand motion using HumanVerse-500, and finally adapts the policy to downstream tasks and robot embodiments. Across these stages, $\lambda_0$ learns a shared representation space for human experience transfer, while domain-specific interfaces handle differences between human and robot states and actions. We evaluate $\lambda_0$ on SIMPLE and 4 real-world loco-manipulation tasks, achieving state-of-the-art performance, and further analyze its scaling behavior, generalization, and training-stage contributions to understand how human data support downstream whole-body humanoid control. We will release our code, models, and data to support further research.
Chinese Translation
类人全身操作技术迅速发展，使得策略能够协调运动、姿态、双手交互和灵巧手部动作。同时，自我中心的人类视频提供了丰富的日常交互示例，涵盖各种物体和场景，提供了可扩展的监督而无需机器人操作。然而，现有的视频监督对全身运动和手-物体交互的协调覆盖有限，而通过类人遥操作获得这种监督成本高且难以扩展。因此，我们探索人类经验如何支持类人运动-操作的可扩展学习。为支持本研究，我们引入了HumanVerse-500，这是一个包含500小时多样化人类运动-操作行为的数据集，数据收集在开放世界环境中，使用轻量级可穿戴系统将自我中心视频与身体和手部运动同步。基于该数据集，我们开发了$ ext{λ}_0$，一个全身类人视觉-语言-动作策略，通过三阶段训练，首先从多样化的自我中心数据集中学习交互，然后使用HumanVerse-500协调身体和手部运动，最后将策略适应于下游任务和机器人体现。在这些阶段中，$ ext{λ}_0$学习了一个共享的表示空间以支持人类经验的转移，同时领域特定的接口处理人类与机器人状态和动作之间的差异。我们在SIMPLE和4个真实世界的运动-操作任务上评估了$ ext{λ}_0$，取得了最先进的性能，并进一步分析了其扩展行为、泛化能力和训练阶段对理解人类数据如何支持下游全身类人控制的贡献。我们将发布我们的代码、模型和数据，以支持进一步的研究。
cs.RO / 13 / 2610.00487
ScaffoldM3C: A Multimodal Sequential Monte Carlo Framework for Generative Stable Construction Planning
ScaffoldM3C：一种用于生成稳定建筑规划的多模态序列蒙特卡洛框架
Abstract
Autonomously constructing physically realizable 3D structures remains a significant challenge due to combinatorial action spaces, interchangeable components, equifinal assembly sequences, and strict stability requirements during construction. State-of-the-art methods fine-tune large language models for text-based generative construction. However, these approaches do not allow for Multimodal (text, image, sketch) conditioning, overlook the practical role of scaffolding for stabilizing intermediate structures, and suffer from slow inference speeds. Therefore, we formulate construction as a probabilistic next-block generation task with multiple potential assembly actions and multiple potential task conditioning modalities. Concurrently, we explicitly consider the utility of scaffolding by introducing an auxiliary scaffold block token. We present Scaffold Multimodal Monte Carlo (ScaffoldM3C), a multimodal, lightweight, auto-regressive model for stable block-based construction, that proposes a set of next-step candidate blocks. Leveraging these candidates, we utilize Sequential Monte Carlo (SMC) to maintain a population of possible assembly sequences, allowing us to consider multiple, potentially different, assembly directions simultaneously. We train our multimodal architecture by extending the StableText2Brick dataset to contain image conditioning prompts and scaffold-stabilized build sequences. ScaffoldM3C is 4x smaller than competing baselines, yielding a 5x to 20x speedup during inference, while achieving comparable construction quality to state-of-the-art methods and higher overall stability. We demonstrate the effectiveness of our approach through simulations and real-world robot assembly demonstrations.
Chinese Translation
自主构建物理可实现的三维结构仍然是一个重大挑战，原因在于组合动作空间、可互换组件、等效的组装序列以及在施工过程中严格的稳定性要求。最先进的方法对大型语言模型进行微调，以实现基于文本的生成建筑。然而，这些方法不支持多模态（文本、图像、草图）条件，忽视了脚手架在稳定中间结构中的实际作用，并且推理速度较慢。因此，我们将建筑过程表述为一个具有多个潜在组装动作和多个潜在任务条件模态的概率下一个块生成任务。同时，我们通过引入辅助脚手架块标记，明确考虑脚手架的效用。我们提出了脚手架多模态蒙特卡洛（ScaffoldM3C），这是一种多模态、轻量级、自回归模型，用于稳定的基于块的建筑，提出了一组下一个步骤候选块。利用这些候选块，我们使用序列蒙特卡洛（SMC）来维持可能的组装序列的种群，从而允许我们同时考虑多个可能不同的组装方向。我们通过扩展StableText2Brick数据集，使其包含图像条件提示和脚手架稳定的构建序列，来训练我们的多模态架构。ScaffoldM3C的规模比竞争基线小4倍，在推理过程中实现了5到20倍的加速，同时在建筑质量上与最先进的方法相当，并且整体稳定性更高。我们通过模拟和现实世界的机器人组装演示展示了我们方法的有效性。
cs.RO / 14 / 2610.00524
Same Scene, Different Task: Skill Alignment for Compositional Generalization in VLAs
相同场景，不同任务：用于组合泛化的技能对齐在视觉-语言-行动模型中的应用
Abstract
Vision-language-action (VLA) models often struggle to generalize to skill combinations absent from their fine-tuning demonstrations, even when every constituent skill has been demonstrated. We focus on a vision shortcut as one failure mode: during fine-tuning, visual observations can serve as a proxy for the instruction, so a policy may execute a demonstrated combination associated with similar observations rather than the instructed combination. This motivates training with counterfactual pairs formed by holding a demonstration observation fixed while changing the instruction to specify an undemonstrated combination. These pairs, however, lack corresponding demonstrated action targets. Crucially, the currently required skill has already been demonstrated, but actions from those executions cannot serve as direct targets because the same skill can require different actions across observations. We propose CRAFT, which transfers supervision from demonstrated executions of the required skill to counterfactual pairs using skill representations that can be reused across executions of the same skill. Across three VLA models and two simulation benchmarks, CRAFT improves success on undemonstrated combinations while maintaining high success on demonstrated ones; it also improves compositional generalization on a real robot. Project website: https://taegeunyang.github.io/craft/
Chinese Translation
视觉-语言-行动（VLA）模型通常在组合技能的泛化上存在困难，尤其是当这些组合在其微调演示中缺失时，即使每个组成技能都已被演示。我们关注一种失败模式，即视觉捷径：在微调过程中，视觉观察可以作为指令的代理，因此策略可能会执行与相似观察相关的已演示组合，而不是指令中指定的组合。这促使我们使用反事实对进行训练，通过固定演示观察并改变指令来指定未演示的组合。然而，这些对缺乏相应的已演示动作目标。关键是，目前所需的技能已经被演示，但这些执行中的动作不能作为直接目标，因为相同的技能在不同观察中可能需要不同的动作。我们提出了CRAFT，它通过使用可以在相同技能的不同执行中重复使用的技能表示，将监督从所需技能的已演示执行转移到反事实对上。在三个VLA模型和两个仿真基准上，CRAFT提高了未演示组合的成功率，同时保持了对已演示组合的高成功率；它还在真实机器人上改善了组合泛化。项目网站：https://taegeunyang.github.io/craft/
cs.RO / 15 / 2610.00542
Does Continual Imitation Learning Remain Grounded? A Language-Perturbed Benchmark for Robotic Task Retention
持续模仿学习是否仍然扎根于语言？一种语言扰动基准用于机器人任务保留
Abstract
Continual imitation learning evaluates whether a robot can learn new knowledge without forgetting previously learned skills. However, retaining task performance does not ensure the behavior remains grounded in language because policies may rely on scene cues, object associations, or memorized task structure. We introduce a benchmark protocol to study how language-guided behavior changes as robotic policies learn successive tasks. We construct meaning-preserving and meaning-changing instruction variants for the Goal, Spatial, Object, and Long suites of LIBERO. Policy experiments focus on LIBERO-Goal, evaluating Original and Paraphrase instructions after each continual-learning stage. We compare representative continual imitation learning methods under their original assumptions while separating task competence from language sensitivity. The proposed diagnostics complement standard learning and forgetting metrics by measuring semantic robustness, goal adaptation, and language sensitivity. Results show that strong continual-learning performance does not always translate to reliable language grounding, and our diagnostics help determine whether retained skills remain correctly guided by their instructions. Additional materials are available at https://sites.google.com/view/stillgrounded
Chinese Translation
持续模仿学习评估机器人是否能够在不遗忘先前学会的技能的情况下学习新知识。然而，保持任务表现并不确保行为仍然扎根于语言，因为策略可能依赖于场景线索、物体关联或记忆的任务结构。我们引入了一种基准协议，以研究随着机器人策略学习连续任务，语言引导的行为如何变化。我们为LIBERO的目标、空间、物体和长时间任务构建了保持意义和改变意义的指令变体。策略实验集中于LIBERO-Goal，在每个持续学习阶段后评估原始和释义指令。我们在其原始假设下比较了代表性的持续模仿学习方法，同时将任务能力与语言敏感性分开。所提出的诊断方法通过测量语义鲁棒性、目标适应性和语言敏感性，补充了标准学习和遗忘指标。结果表明，强大的持续学习表现并不总是转化为可靠的语言扎根，我们的诊断方法有助于确定保留的技能是否仍然正确地受到其指令的引导。更多材料可在 https://sites.google.com/view/stillgrounded 获取。
cs.RO / 16 / 2610.00575
Token-World: World Modeling in Vision-Language Model Token Space for Robot Manipulation
Token-World：在视觉-语言模型令牌空间中的世界建模用于机器人操作
Abstract
A common approach to world-model simulation for vision-language-action (VLA) systems is to predict future RGB observations and then re-encode them into policy inputs, introducing an indirect interface between simulation and downstream policy execution. We instead investigate whether world dynamics can be modeled in a compact, policy-oriented state derived from VLM visual tokens. A key challenge is that raw VLM visual tokens are high-dimensional, making efficient and accurate autoregressive dynamics modeling challenging. To address this, we introduce Token-World, an action-conditioned world model that compresses VLM features into a compact token state, learns future dynamics in this reduced space, and maps predicted states back to the original policy-facing representation for downstream use. Across manipulation benchmarks, Token-World improves open-loop feature fidelity and policy-action consistency over recent world-model simulators, with slower degradation over long rollout horizons. In closed-loop evaluation, its simulated policy performance correlates more strongly with reference policy performance than Ctrl-World ($r=0.794$ vs.\ $0.583$), while requiring lower simulation latency. Ablations further show that compact-representation design and dimensionality substantially affect future-state prediction. Code will be available at https://chuyaofu.github.io/Token-World/.
Chinese Translation
对于视觉-语言-行动（VLA）系统，世界模型仿真的一种常见方法是预测未来的RGB观测值，然后将其重新编码为策略输入，从而在仿真与下游策略执行之间引入间接接口。我们则探讨是否可以在从视觉语言模型（VLM）视觉令牌派生的紧凑、面向策略的状态中建模世界动态。一个关键挑战是，原始的VLM视觉令牌是高维的，这使得高效且准确的自回归动态建模变得困难。为了解决这个问题，我们提出了Token-World，这是一种基于动作条件的世界模型，它将VLM特征压缩为紧凑的令牌状态，在这个简化的空间中学习未来动态，并将预测的状态映射回原始的面向策略的表示以供下游使用。在操作基准测试中，Token-World在开放环路中提高了特征保真度和策略-动作一致性，相较于近期的世界模型仿真器，其在长时间展开中的降解速度更慢。在闭环评估中，其模拟策略性能与参考策略性能的相关性比Ctrl-World更强（$r=0.794$ vs.
$0.583$），同时需要更低的仿真延迟。消融实验进一步表明，紧凑表示设计和维度对未来状态预测有显著影响。代码将会在 https://chuyaofu.github.io/Token-World/ 上发布。
cs.RO / 17 / 2610.00601
When Reasoning Helps Action: Monitoring and Steering Chain-of-Thought in Vision-Language-Action Policies
推理如何促进行动：监控和引导视觉-语言-行动政策中的思维链
Abstract
Reasoning-enabled VLA policies expose chain-of-thought (CoT) traces that appear to explain and guide their actions, creating a potential interface for runtime safety through reasoning monitoring and correction. In this work, we define and operationalize two evaluation axes for assessing when this interface can improve embodied behavior: correctability, which measures whether unreliable reasoning can be detected and improved during generation, and actionability, which measures whether reasoning corrections produce behaviorally meaningful changes in the intended direction. To enable correctability, we introduce Token-level Reward for Utility-Steered Chain-of-Thought (TRUST), an offline-trained value model that predicts eventual reasoning correctness from partial prefixes and uses these estimates to monitor and selectively steer reasoning generation in frozen VLA policies. On the Alpamayo 1.5 driving VLA, TRUST monitors correctness with 88.9% accuracy and improves reasoning correctness from 75.9% to 90.0%. On a baseline-defined challenging subset in AlpaSim, TRUST reduces collision rate by 30.4% and maximum trajectory error by 11.5% relative to the unsteered policy, outperforming a compute-matched Best-of-4 baseline. On the DeepThinkVLA manipulation VLA, TRUST improves the correctness of grasp-state claims from 69.3% to 90.2% and action-choice claims from 68.8% to 85.9%, yet closed-loop task performance on LIBERO-Plus remains largely unchanged. Empirical analysis reveals intent-consistent behavioral effects in Alpamayo 1.5 but limited effects in DeepThinkVLA, helping interpret these different task-level outcomes. Together, our results show that gains in reasoning correctness do not automatically imply gains in embodied performance, motivating evaluation of correctability and actionability when using CoT as a runtime safety interface.
Chinese Translation
启用推理的视觉-语言-行动（VLA）政策展示了思维链（CoT）轨迹，这些轨迹似乎能够解释和指导其行动，从而通过推理监控和修正创造了潜在的运行时安全接口。在本研究中，我们定义并实现了两个评估维度，以评估该接口何时能够改善具身行为：可修正性（correctability），衡量在生成过程中是否能够检测和改善不可靠的推理；可行动性（actionability），衡量推理修正是否能在预期方向上产生行为上有意义的变化。为了实现可修正性，我们引入了基于效用引导的思维链的令牌级奖励（Token-level Reward for Utility-Steered Chain-of-Thought, TRUST），这是一种离线训练的价值模型，能够从部分前缀预测最终的推理正确性，并利用这些估计来监控和选择性地引导冻结的VLA政策中的推理生成。在Alpamayo 1.5驾驶VLA上，TRUST以88.9%的准确率监控正确性，并将推理正确性从75.9%提高到90.0%。在AlpaSim中定义的基线挑战子集上，TRUST相对于未引导政策将碰撞率降低了30.4%，最大轨迹误差降低了11.5%，优于计算匹配的最佳四选一基线。在DeepThinkVLA操作VLA上，TRUST将抓取状态声明的正确性从69.3%提高到90.2%，将行动选择声明的正确性从68.8%提高到85.9%，但在LIBERO-Plus上的闭环任务性能基本保持不变。实证分析揭示了Alpamayo 1.5中的意图一致行为效应，但在DeepThinkVLA中的效果有限，有助于解释这些不同的任务级结果。总的来说，我们的结果表明，推理正确性的提升并不自动意味着具身表现的提升，这促使我们在使用思维链作为运行时安全接口时评估可修正性和可行动性。
cs.RO / 18 / 2610.00638
TacDyn-WAM: Learning Implicit Tactile Dynamics in a Heterogeneous Visuo-Tactile World Action Model
TacDyn-WAM：在异质视触觉世界行动模型中学习隐式触觉动态
Abstract
World action models improve robotic manipulation by conditioning actions on predicted futures, yet existing tactile variants largely inherit video-generation pipelines that reconstruct future tactile observations through iterative denoising. Such prediction can become unreliable under deployment drift: small changes in contact position or force may substantially alter tactile pixels even when the underlying contact evolution remains predictable. We introduce TacDyn-WAM, a heterogeneous visuo-tactile world action model that predicts implicit tactile dynamics rather than reconstructing future tactile observations. It learns TacRep, a dynamics-aware tactile target space trained through masked spatio-temporal prediction on tactile clips and regularized by relational structure distillation. A visual expert and an Implicit Tactile Dynamics Expert predict future visual and tactile representations in separate target spaces while interacting through joint attention; the tactile expert predicts future representations and their changes at multiple horizons in a single forward pass, and a read-only tactile memory supplies the current tactile state. On UniVTAC, TacDyn-WAM achieves an average success rate of 81.5% using only the provided demonstrations, reaching state-of-the-art-level performance and remaining competitive with models pretrained on large-scale visuo-tactile trajectories. Ablations confirm the benefits of both tactile pathways and TacRep over pixel-reconstruction and static alternatives. On five real-robot tasks, TacDyn-WAM reaches 71.0% average success, and modest-scale pretraining raises it to 85.0%, further validating our method.
Chinese Translation
世界行动模型通过基于预测的未来条件化动作来改善机器人操控，然而现有的触觉变体在很大程度上继承了视频生成管道，这些管道通过迭代去噪重建未来的触觉观察。这种预测在部署漂移下可能变得不可靠：接触位置或力的微小变化可能会显著改变触觉像素，即使基础的接触演变仍然可预测。我们提出了TacDyn-WAM，一种异质视触觉世界行动模型，该模型预测隐式触觉动态，而不是重建未来的触觉观察。它学习了TacRep，一个通过对触觉片段进行掩蔽时空预测并通过关系结构蒸馏进行正则化的动态感知触觉目标空间。视觉专家和隐式触觉动态专家在不同的目标空间中预测未来的视觉和触觉表示，同时通过共同注意进行交互；触觉专家在单次前向传递中预测未来表示及其在多个时间范围内的变化，而只读的触觉记忆提供当前的触觉状态。在UniVTAC上，TacDyn-WAM仅使用提供的演示达到了81.5%的平均成功率，达到了最先进的性能，并与在大规模视触觉轨迹上预训练的模型保持竞争。消融实验确认了触觉路径和TacRep相较于像素重建和静态替代方案的优势。在五个真实机器人任务中，TacDyn-WAM达到了71.0%的平均成功率，适度规模的预训练将其提升至85.0%，进一步验证了我们的方法。
cs.RO / 19 / 2610.00718
Toward Humanoid Robots in Construction: A Teleoperation Feasibility Study
朝向建筑领域的人形机器人：遥操作可行性研究
Abstract
We present a teleoperation system that enables a single operator to perform construction tasks on a Unitree G1 humanoid, combining extended reality (XR) based upper body control with pedal-based locomotion to enable simultaneous manipulation and locomotion. Motivated by persistent labor shortages, hazardous working conditions, and challenges in humanoid autonomy, we investigate teleoperation as a practical near-term approach for reducing physical strain on workers while generating high quality demonstration data. We evaluate the system on two representative construction tasks drawn from O*NET occupational database, and report task success and completion time relative to a manual baseline. The system achieved 100% success on tool transport and 80% success on surface painting, with teleoperation requiring substantially more time compared to manual execution.
Chinese Translation
我们提出了一种遥操作系统，使单个操作员能够在Unitree G1人形机器人上执行建筑任务，该系统结合了基于扩展现实（XR）的上半身控制和基于踏板的移动方式，以实现同时的操作和移动。由于持续的劳动力短缺、危险的工作条件以及人形机器人自主性面临的挑战，我们研究了遥操作作为一种实用的短期方法，以减少工人的身体负担，同时生成高质量的演示数据。我们在O*NET职业数据库中选取了两个具有代表性的建筑任务对系统进行了评估，并报告了相对于手动基线的任务成功率和完成时间。该系统在工具运输任务上取得了100%的成功率，在表面涂装任务上取得了80%的成功率，但与手动执行相比，遥操作所需的时间显著更长。
cs.RO / 20 / 2610.00727
CF-JEPA: Improving Robustness of JEPA World Models via Controllability Factorization
CF-JEPA：通过可控性因子分解提高 JEPA 世界模型的鲁棒性
Abstract
Controlling an agent with vision requires being able to separate useful information from irrelevant background information. JEPA-style latent world models seem like a natural approach for this, as they do not perform pixel-level reconstruction; however, they are still sensitive to these distractor signals and experience latent collapse. In this work, we introduce Controllability Factorized JEPA (CF-JEPA), a JEPA-style world model which splits the latent space into controllable and uncontrollable subspaces. This factorization allows us to capture all the distractor information into the uncontrollable region, while we use the control-relevant latent information for our task. With this, we show comparable performance across 2D and 3D control tasks under nominal conditions and improved performance under distracted conditions, where CF-JEPA is the only model that does not experience latent collapse. We also validate our model under distracted conditions for a simulated robot task, highlighting the practical application of such a scheme.
Chinese Translation
使用视觉控制代理需要能够将有用信息与无关的背景信息分离。JEPA 风格的潜在世界模型似乎是实现这一目标的自然方法，因为它们不进行像素级重建；然而，它们仍然对这些干扰信号敏感，并且会经历潜在崩溃。在本研究中，我们引入了可控性因子分解 JEPA（CF-JEPA），这是一种 JEPA 风格的世界模型，它将潜在空间分为可控和不可控的子空间。这种因子分解使我们能够将所有的干扰信息捕获到不可控区域，同时我们使用与控制相关的潜在信息来完成我们的任务。通过这种方法，我们在正常条件下展示了在 2D 和 3D 控制任务中的可比性能，并在干扰条件下提高了性能，其中 CF-JEPA 是唯一一个没有经历潜在崩溃的模型。我们还在干扰条件下验证了我们的模型，以模拟机器人任务，突显了这种方案的实际应用。
cs.RO / 21 / 2610.00731
Measuring Asset and Scene Reconstruction Effects in Real-to-Sim Robot Evaluation
测量真实到模拟机器人评估中的资产和场景重建效果
Abstract
Simulated evaluation is increasingly used alongside real-world evaluation of robot policies because it is cheaper and easier to repeat; however, its value depends on how closely its outcomes track the real robot's. We test whether our reconstruction pipeline, combining metrically scaled object geometry, authored physical parameters and scene reconstruction, reduces disagreement between simulated and real robot scores relative to a default open-source recipe. We constructed two simulated versions of one bimanual robot cell: an authored reconstruction, using object geometry at estimated metric scale, projected textures, authored physics and our own scene splat; and a baseline, referred to as the default reconstruction, using the open-source recipe of a generative single-image mesh, engine-default physics and a Gaussian-splat scene. Both reconstructions use the same object photographs and scene video. Two policies ran five tasks each, giving ten task-policy pairs, which we call cells; each cell was run twenty times in each reconstruction with all other settings held fixed. Both reconstructions were scored against the same real trials, graded by a third-party evaluator. Pearson correlation between the ten simulated and real cell means is r = 0.90 for the authored reconstruction and 0.51 for the default. Mean score error is 6.97 percentage points for the authored reconstruction and 17.54 for the default, a reduction of 10.56 percentage points. These results show that improving the quality of the environment reconstruction through higher visual fidelity, authored physics and metric scale makes the simulation more faithful to the real world and narrows the sim-to-real gap. We release the harness, the per-trial scores, every reported run's configuration, and the assets and scenes of both reconstructions.
Chinese Translation
模拟评估在机器人策略的真实世界评估中越来越多地被使用，因为它更便宜且更易于重复；然而，其价值取决于其结果与真实机器人之间的吻合程度。我们测试了我们的重建管道，结合了按比例缩放的物体几何形状、编写的物理参数和场景重建，是否减少了模拟与真实机器人评分之间的分歧，相较于默认的开源方案。我们构建了一个双手机器人单元的两个模拟版本：一个是编写的重建，使用按估计的度量比例缩放的物体几何形状、投影纹理、编写的物理和我们自己的场景喷溅；另一个是基线，称为默认重建，使用生成的单幅图像网格的开源方案、引擎默认物理和高斯喷溅场景。两个重建使用相同的物体照片和场景视频。两个策略各执行五个任务，共产生十对任务-策略，我们称之为单元；每个单元在每个重建中运行二十次，所有其他设置保持不变。两个重建的评分均与相同的真实试验进行对比，由第三方评估者进行评分。十个模拟与真实单元均值之间的皮尔逊相关系数为 r = 0.90（编写的重建）和 0.51（默认重建）。编写重建的平均评分误差为 6.97 个百分点，默认重建为 17.54 个百分点，减少了 10.56 个百分点。这些结果表明，通过提高环境重建的质量，增强视觉真实感、编写物理和度量比例，使得模拟更忠实于真实世界，并缩小了模拟与真实之间的差距。我们发布了测试工具、每次试验的评分、每次报告运行的配置，以及两个重建的资产和场景。
cs.RO / 22 / 2610.00781
DITTO-X: Forward and Reverse Teleoperation for Dexterous Manipulation and Human Intervention
DITTO-X：用于灵巧操作和人类干预的前向与反向遥操作
Abstract
Teleoperated demonstrations are a primary source of data for robot manipulation, and teleoperated interventions are a primary mechanism for correcting policies at deployment. Yet most teleoperation systems close the loop through vision alone and are built around parallel-jaw grippers, limiting both what the robot can execute and what the operator can express through it. This is most damaging in shared autonomy, where the operator sees the scene only through occluded cameras and must take over a dexterous hand mid-task, often with an object already grasped. We present DITTO-X, a hand-agnostic dexterous teleoperation interface that renders joint-level force and fingertip contact events from sensing already on the robot hand, and drives three commercial dexterous hands (Sharpa, Wuji, and Inspire) without per-hand redesign. Because the exoskeleton is actuated, DITTO-X also supports reverse teleoperation, in which the robot back-drives the operator's fingers into its own configuration before control is transferred, so the human enters the loop already matched to the state they inherit. Our results show that DITTO-X improves demonstration quality and throughput over a commercial hand-tracking glove, both in regular data collection and in human intervention during policy deployment for contact-rich manipulation tasks. More information can be found from our website: https://tml.stanford.edu/ditto-x/.
Chinese Translation
遥操作演示是机器人操作的主要数据来源，而遥操作干预则是部署时修正策略的主要机制。然而，大多数遥操作系统仅通过视觉闭环，并围绕平行夹爪构建，这限制了机器人能够执行的任务以及操作员通过机器人表达的能力。这在共享自主中尤为严重，操作员只能通过受遮挡的摄像头观察场景，并且必须在任务进行中接管灵巧手，通常是在已经抓取物体的情况下。我们提出了DITTO-X，这是一种与手部无关的灵巧遥操作接口，它能够从机器人手上的传感器渲染关节级别的力和指尖接触事件，并驱动三种商业灵巧手（Sharpa、Wuji和Inspire），无需针对每只手进行重新设计。由于外骨骼是可驱动的，DITTO-X还支持反向遥操作，在这种模式下，机器人会将操作员的手指反向驱动到其自身的配置中，然后再转移控制，从而使人类在进入控制环时已经与他们继承的状态匹配。我们的结果表明，DITTO-X在常规数据收集和在接触丰富的操作任务中政策部署期间的人类干预方面，相较于商业手部追踪手套，提高了演示质量和吞吐量。更多信息可以在我们的网站上找到：https://tml.stanford.edu/ditto-x/
cs.RO / 23 / 2610.00801
ECoMEM: Explicit Concept Memory for Memory-Dependent Robot Control
ECoMEM：用于依赖记忆的机器人控制的显式概念记忆
Abstract
A robot may lose sight of an object it must later retrieve, need to recall what a person demonstrated earlier, or track which steps of a task it has already completed. Current vision-language-action (VLA) policies often fail once the information needed for action disappears from the current observation, making memory critical for long-horizon robot behavior. Existing approaches typically provide longer histories or learn implicit memory from observation-action trajectories. But action supervision tells a policy how to act, not what to remember: it does not specify which past facts should persist or how they should change as new evidence arrives. We therefore separate maintaining an evidence-grounded account of the past from learning how to act on it. This insight motivates Explicit Concept Memory (ECoMEM), which represents task-relevant history with a reusable library of grounded concepts. An evidence-based Writer selects and updates these records, while a learned Reader turns them into memory tokens that directly condition the VLA. Across 16 RoboMME tasks, ECoMEM leads the evaluated robot policies on 15 tasks. On two new real-robot tasks, the same memory library either transfers directly or requires only one new concept, achieving 86.1% success versus 8.6% for a no-memory VLA. These results show that explicit concepts provide a reusable and extensible memory interface for robot control. Project website: https://ecomem.github.io/
Chinese Translation
机器人可能会失去对必须稍后取回的物体的视线，需要回忆一个人之前展示的内容，或跟踪已完成的任务步骤。当前的视觉-语言-动作（VLA）策略在所需的行动信息从当前观察中消失后往往会失败，这使得记忆对于长时间跨度的机器人行为至关重要。现有的方法通常提供更长的历史记录或从观察-行动轨迹中学习隐式记忆。但行动监督告诉策略如何行动，而不是记住什么：它并未指定哪些过去的事实应当保持或如何随着新证据的到来而变化。因此，我们将维护基于证据的过去记录与学习如何对其采取行动分开。这一见解促使了显式概念记忆（ECoMEM）的提出，该方法通过一个可重用的基础概念库来表示与任务相关的历史。一个基于证据的写入器选择并更新这些记录，而一个学习的读取器将其转化为直接影响VLA的记忆标记。在16个RoboMME任务中，ECoMEM在15个任务中领先于评估的机器人策略。在两个新的真实机器人任务中，同样的记忆库要么直接转移，要么仅需一个新概念，实现了86.1%的成功率，而无记忆的VLA仅为8.6%。这些结果表明，显式概念提供了一个可重用且可扩展的机器人控制记忆接口。项目网站：https://ecomem.github.io/
cs.RO / 24 / 2610.00821
Getting Out and Getting Back: World and Behavior Grounding in Real2Sim2Real Co-Training
走出与回归：Real2Sim2Real协同训练中的世界与行为基础
Abstract
Simulation can expand scarce real demonstrations for co-training, yet how world fidelity and similarity to human behavior affect policy performance remains unclear. We distinguish world grounding, which aligns simulation with the real system, and behavior grounding, which aligns simulated trajectories with human motion. We build a real2sim2real pipeline that varies these axes independently to generate data for co-training. On a dynamic dexterous pick-and-sort task, fully grounded co-training raises success from 52% to 86%; averaged across configurations, world grounding improves success by 18 percentage points and behavior grounding by 10. Deployed policies behave like a mixture of real-derived and simulation-derived policies, imitating real demonstrations in covered states and relying on simulated behavior elsewhere, which we examine through latent-space analysis. Together, these results suggest complementary roles: world grounding lets policies use simulated experience beyond real-data coverage, while behavior grounding matters mainly when world grounding is imperfect. Grounded simulation remains beneficial when co-training foundation models.
Chinese Translation
仿真可以扩展稀缺的真实演示以进行协同训练，但世界的真实性和与人类行为的相似性如何影响策略性能仍不清楚。我们区分了世界基础，它将仿真与真实系统对齐，以及行为基础，它将仿真轨迹与人类运动对齐。我们构建了一个real2sim2real管道，独立变化这些维度以生成协同训练的数据。在一个动态灵巧的拾取与排序任务中，完全基础的协同训练将成功率从52%提高到86%；在各种配置中，世界基础将成功率提高了18个百分点，而行为基础提高了10个百分点。部署的策略表现得像是由真实数据和仿真数据混合而成的策略，在覆盖状态下模仿真实演示，而在其他地方依赖于仿真行为，我们通过潜在空间分析进行了研究。这些结果共同表明了互补的作用：世界基础使策略能够利用超出真实数据覆盖范围的仿真经验，而行为基础主要在世界基础不完美时显得重要。当协同训练基础模型时，基础仿真仍然是有益的。
cs.RO / 25 / 2610.00822
TRACE: Privacy-Preserving Next-Best-View Selection over Distributed 3D Gaussian-Splat Maps
TRACE：分布式3D高斯点云图上的隐私保护下一最佳视角选择
Abstract
Share the light, not the map. We study next-best-view selection for a team of robots, each of which builds its own 3D Gaussian Splatting map and keeps it private. A robot picks the view with the largest expected information gain (EIG) about the splats along its own path. This gain depends on the other maps. Their splats occlude its own and shine behind them, so the gain has to be evaluated against the pooled map. No robot has this map. We show that the coupling passes through only two ray quantities, the transmittance in front of a splat and the radiance behind it, and that both are sums over the hits of the ray. Hence, they decompose across the robots, and each robot sums them over depth bins in its own map, along the rays of a candidate view, and sends the sums with their pose derivatives. The robot planning the view turns them into its EIG and gradient on SO(3). Transmittance and Radiance Aggregates, communicated for the EIG, give the protocol its name: TRACE. No robot shares its splats, and the message size does not grow with a map. We prove that the reconstruction is exact unless a depth bin behind a splat mixes hits of two robots, and we bound the error otherwise. Over 100 next-best-view decisions in Habitat-Sim, TRACE picks a heading within 15 degrees of the centralized one in 83.3% of the cases, and its views reach 97.9% of the centralized EIG.
Chinese Translation
分享光线，而非地图。我们研究了一个机器人团队的下一最佳视角选择，每个机器人构建自己的3D高斯点云图并保持私密。一个机器人选择沿其路径的点云中，预期信息增益（EIG）最大的视角。这个增益依赖于其他机器人的地图。其他机器人的点云会遮挡自己的点云，并在其后发光，因此增益必须在汇总地图上进行评估。没有任何机器人拥有这张地图。我们展示了耦合仅通过两个光线量传递：点云前方的透射率和其后方的辐射度，并且这两者都是光线命中的总和。因此，它们可以在机器人之间分解，每个机器人在自己的地图中沿着候选视角的光线对深度区间进行求和，并发送这些和及其姿态导数。规划视角的机器人将这些信息转化为其EIG和SO(3)上的梯度。为了计算EIG而传递的透射率和辐射度聚合物赋予该协议名称：TRACE。没有机器人共享其点云，且消息大小不会随着地图的增大而增加。我们证明了重建是精确的，除非一个点云后方的深度区间混合了两个机器人的命中，否则我们对误差进行了界定。在Habitat-Sim中进行的100多个下一最佳视角决策中，TRACE在83.3%的情况下选择的朝向与集中式决策相差不超过15度，其视角达到了集中式EIG的97.9%。
cs.RO / 26 / 2610.00823
Reactive Humanoid Multi-Contact Using Learned Stability Models
基于学习稳定性模型的反应式类人多接触控制
Abstract
We present a planning and control approach to reactively use hand contacts to stabilize a humanoid in low stability scenarios, where only using feet contacts may result in a fall. Candidate contacts are sampled within the robot's reachable workspace, and a preview is computed by rolling out the centroidal dynamics through pre-impact, impact and post-impact phases. Sampled points are scored based on the Center of Pressure (CoP) control authority at the post-impact phase. Central to our approach is a learned model of the robot's CoP region during post-impact, which enables rapid evaluation of candidate contact points compared to traditional optimization-based methods. The presented planner has two stages: the first selects an optimal bracing region and the second computes an optimal bracing point within the region. Our simulation results demonstrate an average increase in impulse resilience of 89% over recovery without hand contacts and 17% over a naive planning strategy (closest reachable region). We validate our framework on hardware, performing push tests while standing and walking. The standing trials show an average 43% reduction in stabilization time compared to naive hand placement and the walking trials demonstrate a 18% reduction compared to baseline recovery (without hand contacts).
Chinese Translation
我们提出了一种规划和控制方法，以反应性地利用手部接触来稳定类人在低稳定性场景下的姿态，在这些场景中，仅依靠脚部接触可能导致跌倒。候选接触点在机器人可达工作空间内进行采样，并通过预冲击、冲击和后冲击阶段的质心动力学进行预测。根据后冲击阶段的压力中心（Center of Pressure, CoP）控制能力对采样点进行评分。我们方法的核心是一个学习的模型，该模型描述了机器人在后冲击阶段的CoP区域，使得与传统的基于优化的方法相比，候选接触点的评估更加迅速。所提出的规划器分为两个阶段：第一阶段选择一个最优支撑区域，第二阶段在该区域内计算一个最优支撑点。我们的仿真结果表明，与没有手部接触的恢复相比，冲击恢复的平均弹性提高了89%，与简单规划策略（最近可达区域）相比提高了17%。我们在硬件上验证了我们的框架，进行站立和行走时的推力测试。站立试验显示，与简单手部放置相比，稳定时间平均减少了43%，而行走试验则显示与基线恢复（无手部接触）相比减少了18%。
cs.RO / 27 / 2610.00854
Are Frontier VLM Agents Ready to Be Robot Generalists? An Empirical Study with the Embodied Agent Arena
前沿视觉语言模型代理是否准备好成为机器人通才？基于具身代理竞技场的实证研究
Abstract
Frontier vision-language models (VLMs) combine scene estimation, interaction grounding, and executable actions. Understanding how these abilities support complete robotic tasks is central to evaluating their readiness as robot generalists. We introduce Embodied Agent Arena to examine where local competence supports, or falls short of, complete task success across Geometry, Spatial Reasoning, Affordance, Task Planning, and Manipulation. The arena contains 1,000 cases drawn from 32 established sources and GeoProbe, our new benchmark for geometric estimation on Blender renders and real-scene images. A minimal harness preserves source observations and operations while separating metric precision, functional grounding, and native goal completion. We evaluate seven VLMs, analyze Astra's task-specific advantages, and compare richer-observation execution protocols and multi-round review. Across the arena, Astra's advantage is strongest in precise estimation and usable-contact localization; completing coordinated, goal-directed actions remains the key gap to robot generalism.
Chinese Translation
前沿视觉语言模型（VLMs）结合了场景估计、交互基础和可执行动作。理解这些能力如何支持完整的机器人任务对于评估它们作为机器人通才的准备程度至关重要。我们引入了具身代理竞技场，以检验局部能力在几何、空间推理、可供性、任务规划和操作等方面如何支持或未能实现完整任务成功。该竞技场包含1,000个案例，来源于32个已建立的来源和GeoProbe，这是我们在Blender渲染和真实场景图像上进行几何估计的新基准。一个最小化的工具保持了源观察和操作，同时分离了度量精度、功能基础和原生目标完成。我们评估了七个VLMs，分析了Astra在特定任务上的优势，并比较了更丰富的观察执行协议和多轮评审。在竞技场中，Astra在精确估计和可用接触定位方面的优势最为明显；完成协调的、目标导向的动作仍然是机器人通才的关键差距。
cs.RO / 28 / 2610.00864
Kinematic MeanFlow: One-Step Action Generation Policy for Robotic Foundation Models
运动均流：用于机器人基础模型的一步动作生成策略
Abstract
In this paper, we study how to achieve one-step action generation in Robotic Foundation Models (RFMs), aiming to overcome the high inference latency of multi-step flow matching. MeanFlow provides a promising framework for this goal, yet its direct application leads to performance collapse. We discover that this stems from two distinctive dynamics exhibited in the RFM velocity field: (1) the ``local acceleration" exhibits stability early on, but surges sharply towards the end of the denoising process, and (2) the spread of its magnitudes across samples widens as denoising progresses. To address these issues, we introduce Kinematic MeanFlow (K-MF), a novel one-step action policy tailored for RFMs. Specifically, grounded in a kinematic identity, K-MF decouples the time derivative term in the MeanFlow formulation into two sub-interval terms separated by an intermediate point. This decoupled formulation enables the two terms to capture early-stage and late-stage denoising dynamics, respectively, while mitigating the error amplification across the process. As a result, our K-MF empowers RFMs to achieve one-step action generation in both training from scratch and fine-tuning paradigms across diverse tasks, while outperforming multi-step flow matching in most settings. In terms of inference efficiency, K-MF reduces action-head latency of GR00T-N1.6 by 67.5%~74.4% across L40 and Jetson Orin in eager and compiled modes, yielding end-to-end latency reductions of 30.3%~54.9%. Code will be available at https://github.com/IntelChina-AI/K-MF.
Chinese Translation
在本文中，我们研究如何在机器人基础模型（Robotic Foundation Models, RFMs）中实现一步动作生成，旨在克服多步流匹配的高推理延迟。均流（MeanFlow）为此目标提供了一个有前景的框架，但其直接应用会导致性能崩溃。我们发现，这源于RFM速度场中表现出的两种独特动态：(1) “局部加速度”在早期表现出稳定性，但在去噪过程的末尾急剧上升；(2) 随着去噪的进行，其幅度在样本间的分布逐渐扩大。为了解决这些问题，我们引入了运动均流（Kinematic MeanFlow, K-MF），这是一种为RFMs量身定制的新型一步动作策略。具体而言，K-MF基于运动学恒等式，将均流公式中的时间导数项解耦为两个由中间点分隔的子区间项。这种解耦的公式使得两个项分别能够捕捉早期和晚期的去噪动态，同时减轻了整个过程中的误差放大。因此，我们的K-MF使RFMs能够在从头训练和微调范式下实现一步动作生成，并在大多数设置中优于多步流匹配。在推理效率方面，K-MF在急切模式和编译模式下，分别在L40和Jetson Orin上将GR00T-N1.6的动作头延迟降低了67.5%~74.4%，实现了端到端延迟降低30.3%~54.9%。代码将在https://github.com/IntelChina-AI/K-MF上发布。
cs.RO / 29 / 2610.00878
UniTrackPLA: Unified Panorama-Language-Action Model for Instruction-Guided Navigation and Dynamic Person Tracking
UniTrackPLA：用于指令引导导航和动态人物跟踪的统一全景-语言-动作模型
Abstract
General-purpose embodied robots should support both navigation toward language-specified destinations and dynamic person tracking under arbitrary initial target azimuths. However, existing methods typically rely on forward-facing observations and address these tasks with separate policies, limiting omnidirectional perception and unified closed-loop control. We present UniTrackPLA, a unified panorama-language-action model for instruction-guided navigation and dynamic person tracking. Its Panoramic-Aware Encoding (PAE) preserves the temporal and azimuthal structure of perspective views projected from each panorama, enabling perspective-pretrained visual encoders to process omnidirectional observations. A shared vision-language backbone grounds instructions in the panoramic context and predicts continuous robot-centric waypoint chunks for both tasks. World-Action Consistency (WAC) further predicts action-conditioned future visual states and verifies waypoint prefixes online, allowing reliable actions to be reused while triggering replanning upon inconsistency. We also introduce OmniTrackNav-Bench, comprising 5,000 simulated tracking trajectories, 10,000 simulated VLN routes, and 96 verified real-world routes, providing 919,978 waypoint-supervision instances. UniTrackPLA improves overall tracking SR from 23.50% to 35.00% and Omni-VLN SR/SPL from 13.00%/12.77% to 19.75%/19.29%. Incorporating 76 real-world routes further improves held-out
[email protected] from 42.92% to 92.08%. Closed-loop experiments on a Go2-W robot demonstrate unified panoramic tracking and navigation across indoor and outdoor environments. The project page is at https://tw5775.github.io/UniTrackPLA.
Chinese Translation
通用的具身机器人应支持朝向语言指定目的地的导航以及在任意初始目标方位下的动态人物跟踪。然而，现有方法通常依赖于前方观察，并通过独立策略处理这些任务，限制了全向感知和统一闭环控制。我们提出了UniTrackPLA，一种用于指令引导导航和动态人物跟踪的统一全景-语言-动作模型。其全景感知编码（Panoramic-Aware Encoding, PAE）保留了从每个全景投影的视角图像的时间和方位结构，使得经过视角预训练的视觉编码器能够处理全向观察。共享的视觉-语言骨干网络将指令与全景上下文结合，并为两个任务预测连续的机器人中心路径点块。世界-动作一致性（World-Action Consistency, WAC）进一步预测基于动作的未来视觉状态，并在线验证路径点前缀，允许在不一致时可靠地重用动作并触发重新规划。我们还引入了OmniTrackNav-Bench，包括5000个模拟跟踪轨迹、10000条模拟的VLN路线和96条经过验证的真实世界路线，提供919978个路径点监督实例。UniTrackPLA将整体跟踪成功率（SR）从23.50%提高到35.00%，Omni-VLN成功率（SR）/成功路径长度（SPL）从13.00%/12.77%提高到19.75%/19.29%。结合76条真实世界路线进一步将保留的
[email protected]从42.92%提高到92.08%。在Go2-W机器人上的闭环实验展示了在室内和室外环境中统一的全景跟踪和导航。项目页面为：https://tw5775.github.io/UniTrackPLA。
cs.RO / 30 / 2610.00897
Real-Time Human-Adaptive Task Allocation for Multi-Human Multi-Robot Supervision
实时人机适应性任务分配用于多人体多机器人监督
Abstract
We propose a human-factor-aware method of allocating robot supervision tasks to multiple human operators. In scenarios where multiple operators occasionally teleoperate multiple robots to help the robots overcome difficulties, the allocation of the supervisory control tasks to humans needs to consider the real-time cognitive states of individual operators. However, most existing methods assume fixed supervisory capacity per operator and overlook fluctuations in the human factors such as workload and fatigue. As a result, workload distribution can be unbalanced where some operators become overloaded while the others remain underused. Our method dynamically regulates supervisory capacity and allocates tasks in a way that maintains balanced mental workload, prevents overload, and improves overall team performance. The allocation method uses a greedy strategy that minimizes estimated operator workloads with task prioritization. Robots are assigned to operators by reflecting their current supervisory capacity where the required effort depends on the types of tasks. In the user study, the analysis across predefined time intervals shows that the proposed method consistently achieves higher performance and lower behavioral signs of fatigue compared to a baseline method that does not consider human factors. These results highlight adaptive capacity adjustment as an effective preventive mechanism for sustaining operator performance in long-duration, high-demand settings.
Chinese Translation
我们提出了一种考虑人因因素的机器人监督任务分配方法，旨在为多个人工操作员分配任务。在多个操作员偶尔远程操作多台机器人以帮助机器人克服困难的场景中，监督控制任务的分配需要考虑个别操作员的实时认知状态。然而，现有大多数方法假设每个操作员的监督能力是固定的，并忽视了人因因素如工作负荷和疲劳的波动。因此，工作负荷的分配可能不平衡，导致一些操作员过载而其他操作员则未得到充分利用。我们的方法动态调节监督能力，并以保持心理工作负荷平衡、防止过载和提高整体团队表现的方式分配任务。该分配方法采用贪心策略，通过任务优先级最小化估计的操作员工作负荷。机器人根据操作员当前的监督能力进行分配，所需的努力取决于任务类型。在用户研究中，对预定义时间间隔的分析表明，所提出的方法在性能和疲劳行为迹象方面始终优于不考虑人因因素的基线方法。这些结果突显了适应性能力调整作为在长时间、高需求环境中维持操作员表现的有效预防机制。
cs.RO / 31 / 2610.00899
TOAST: Stochastic Robot Action Tokenization for Autoregressive Vision-Language-Action Models
TOAST：用于自回归视觉-语言-动作模型的随机机器人动作标记化
Abstract
Autoregressive Vision-Language-Action models often represent continuous robot actions as discrete token sequences, enabling action prediction with standard next-token objectives. FAST has substantially improved this representation by compactly encoding action containing diverse temporal frequencies into relatively few tokens. However, while such compression reduces the number of action tokens required for autoregressive prediction, it does not necessarily improve the efficiency of policy learning from limited demonstrations. In particular, FAST typically assigns a single deterministic tokenization to each quantized action sequence, although multiple token sequences can represent and decode to the same robot motion. We investigate whether exploiting this representational redundancy can improve policy learning. In this paper, we propose TOkenization of Action sequences with STochastic sampling (TOAST), a stochastic action tokenization method that samples alternative tokenizations of the same quantized action sequence during policy training. This diversifies the discrete supervision while preserving the underlying robot action and requires no additional demonstrations. Experiments on LIBERO show that TOAST consistently improves over its deterministic counterpart, with the improvement increasing as training data decreases, achieving a 6.8 point gain in success rate when only 1/16 of training data is available. Across four real-robot manipulation tasks, TOAST further improves mean success rate by 15.8 points over the deterministic counterpart. These results demonstrate the effectiveness of stochastic action tokenization for autoregressive robot policy learning, particularly when training data are limited.
Chinese Translation
自回归视觉-语言-动作模型通常将连续的机器人动作表示为离散的标记序列，从而能够通过标准的下一个标记目标进行动作预测。FAST显著改善了这种表示，通过将包含多种时间频率的动作紧凑编码为相对较少的标记。然而，尽管这种压缩减少了自回归预测所需的动作标记数量，但并不一定提高从有限演示中学习策略的效率。特别是，FAST通常为每个量化的动作序列分配一个单一的确定性标记化，尽管多个标记序列可以表示并解码为相同的机器人运动。我们研究了利用这种表示冗余是否可以改善策略学习。本文提出了带有随机采样的动作序列标记化方法（TOAST），一种在策略训练过程中对同一量化动作序列进行替代标记化采样的随机动作标记化方法。这在保留基础机器人动作的同时多样化了离散监督，并且不需要额外的演示。在LIBERO上的实验表明，TOAST在其确定性对应物上始终表现出改进，且随着训练数据的减少，改进幅度增加，当仅有1/16的训练数据可用时，成功率提高了6.8个百分点。在四个真实机器人操作任务中，TOAST进一步将平均成功率提高了15.8个百分点。这些结果证明了随机动作标记化在自回归机器人策略学习中的有效性，特别是在训练数据有限的情况下。
cs.RO / 32 / 2610.00904
Screw Attention: Rigid-Body Algebra Inside a Transformer
螺旋注意力：变换器中的刚体代数
Abstract
Learned manipulation policies rediscover from data the spatial relations that rigid-body mechanics supplies in closed form. This costs data, and it leaves the policies fragile to geometric changes in the scene. We present Screw Attention, a transformer layer in which the relation between two bodies is a spatial transform rather than a graph edge. Every token is a body with a pose. Each pair of tokens carries the relative pose and, for robot joints, the joint screw. Messages are transported along this relation into the receiver's frame, while the attention scores see only frame-invariant quantities. By construction, the messages are equivariant to an independent change of frame at every token, and a single layer can express the velocity recursion of rigid-body mechanics. On simulated manipulation tasks, Screw Attention matches or outperforms controls of the same size, including graph, transformer and flat networks on LIBERO-Spatial. With 16,162 parameters it reaches 97.3% on LIBERO-Spatial from object poses (without images or language), above a flat network with 27x more parameters. Under a change of per-link frame convention its success is unchanged, while every other learned network falls below 3%. Placed on an analytic controller as a gated residual, it raises insertion success by 17.3 points. It is unaffected by pose noise up to 10,mm and by joint offsets within the factory calibration of a Franka arm. These results suggest a criterion: geometry is decisive when the task requires relations between frames that no other part of the system supplies. Code and trained policies will be released.
Chinese Translation
学习的操作策略从数据中重新发现了刚体力学以封闭形式提供的空间关系。这需要数据，并且使得策略对场景中的几何变化变得脆弱。我们提出了螺旋注意力（Screw Attention），这是一种变换器层，其中两个物体之间的关系是空间变换，而不是图边。每个标记都是一个具有姿态的物体。每对标记携带相对姿态，对于机器人关节，则携带关节螺旋。消息沿着这种关系传递到接收者的坐标系中，而注意力分数仅关注坐标系不变的量。通过构造，这些消息对每个标记的独立坐标系变化是等变的，单层可以表达刚体力学的速度递归。在模拟操作任务中，螺旋注意力的表现与同等规模的控制器相匹配或超越，包括图、变换器和在LIBERO-Spatial上的平面网络。它以16,162个参数在LIBERO-Spatial上从物体姿态（不使用图像或语言）达到97.3%的准确率，超过了参数数量多27倍的平面网络。在每个链接的坐标系约定变化下，其成功率不变，而其他学习网络的成功率均低于3%。作为门控残差放置在分析控制器上时，它将插入成功率提高了17.3个百分点。它对高达10mm的姿态噪声和Franka臂的工厂校准内的关节偏移不受影响。这些结果表明了一个标准：当任务需要框架之间的关系，而系统的其他部分无法提供时，几何是决定性的。代码和训练策略将会发布。
cs.RO / 33 / 2610.00913
eRLT: Efficient VLA Reinforcement Learning via Action-Relevant Token Routing
eRLT：通过与动作相关的令牌路由实现高效的视觉-语言-动作强化学习
Abstract
Vision-Language-Action (VLA) models provide strong behavioral priors for robotic manipulation, yet efficiently adapting them to downstream tasks remains challenging. Recent work addresses this challenge by adapting frozen VLAs through online reinforcement learning (RL), whose sample efficiency depends on the quality of the state representation used by the actor and critic. Existing methods construct such representations either with VLA-independent visual encoders or through fixed compression of internal VLA representations. Neither design explicitly extracts the task-specific action-relevant VLA features most useful for downstream action refinement and action-value estimation, therefore limiting sample efficiency. To address this limitation, we introduce eRLT, which constructs an effective state representation by routing task-specific action-relevant information across both tokens and layers of the frozen VLA. Specifically, learned routing tokens dynamically aggregate visual-language features at multiple depths, while a lightweight layer router combines these summaries into a fixed-dimensional RL token. The routing module is initialized using expert demonstrations to capture features predictive of expert actions and then refined using critic feedback from online interactions for action-value estimation. Across seven LIBERO and RoboTwin tasks, eRLT improves mean normalized learning-curve AUC by up to 23.7% over representative baselines. Real-robot experiments on USB connector insertion and motherboard ribbon-cable insertion further show AUC improvements of 108.9% and 46.7%, respectively, over the strongest baseline.
Chinese Translation
视觉-语言-动作（VLA）模型为机器人操作提供了强大的行为先验，但将其高效适应于下游任务仍然具有挑战性。近期的研究通过在线强化学习（RL）调整冻结的VLA来应对这一挑战，而其样本效率依赖于演员和评论家所使用的状态表示的质量。现有方法通过与VLA无关的视觉编码器或通过固定压缩内部VLA表示来构建此类表示。这两种设计都未能明确提取对下游动作细化和动作价值估计最有用的任务特定与动作相关的VLA特征，因此限制了样本效率。为了解决这一限制，我们提出了eRLT，它通过在冻结的VLA的令牌和层之间路由任务特定的与动作相关的信息来构建有效的状态表示。具体而言，学习到的路由令牌动态地在多个深度聚合视觉-语言特征，而轻量级层路由器将这些摘要组合成固定维度的RL令牌。路由模块使用专家演示进行初始化，以捕捉预测专家动作的特征，然后通过在线交互中的评论家反馈进行细化，以进行动作价值估计。在七个LIBERO和RoboTwin任务中，eRLT在代表性基线之上将平均归一化学习曲线的AUC提高了最多23.7%。在USB连接器插入和主板排线插入的真实机器人实验中，AUC分别提高了108.9%和46.7%，超越了最强基线。
cs.RO / 34 / 2610.00926
A Survey on End-to-End Autonomous Driving Training from the Perspectives of Data, Strategy, and Platform
从数据、策略和平台的视角看端到端自主驾驶训练的综述
Abstract
Autonomous driving is a cornerstone technology for the future of intelligent transportation, where end-to-end learning has emerged as a transformative paradigm that directly maps multimodal sensory inputs to driving actions through unified differentiable models. While offering advantages, the effectiveness of end-to-end autonomous driving (E2E-AD) is ultimately determined by the quality of its training ecosystem. This paper provides a comprehensive review of training methods and ecosystem for E2E-AD. We introduce a Data-Strategy-Platform taxonomy that conceptualizes training as an interdependent system. The data layer defines what can be learned, the strategy layer governs how learning aligns with driving objectives, and the platform layer supports scalability and continuous evolution. Within this framework, we survey recent advances across data-centric pipelines, learning paradigms, and training infrastructures, and analyze their interplay in shaping model performance, robustness, and deployability. Finally, we reflect on current limitations and articulate a forward-looking vision that emphasizes a shift from data quantity to data value, from isolated optimization to foundation-driven generalization, and from static training to integrated training-testing loops, aiming toward robust, scalable, and trustworthy autonomous driving systems. We maintain a continuously updated repository tracking cutting-edge literature and works at \href{https://github.com/Jiaaqiliu/Awesome-Training-Ecosystem-for-E2E-AD}{Our Project Page}.
Chinese Translation
自主驾驶是未来智能交通的基石技术，其中端到端学习作为一种变革性范式，直接将多模态传感器输入映射到驾驶动作，通过统一的可微模型实现。尽管端到端自主驾驶（E2E-AD）具有优势，但其有效性最终取决于训练生态系统的质量。本文对E2E-AD的训练方法和生态系统进行了全面回顾。我们引入了数据-策略-平台的分类法，将训练概念化为一个相互依赖的系统。数据层定义了可以学习的内容，策略层管理学习如何与驾驶目标对齐，而平台层则支持可扩展性和持续演进。在这一框架下，我们调查了数据驱动的管道、学习范式和训练基础设施的最新进展，并分析了它们在塑造模型性能、鲁棒性和可部署性方面的相互作用。最后，我们反思了当前的局限性，并阐明了一个前瞻性的愿景，强调从数据数量向数据价值的转变，从孤立优化向基础驱动的泛化转变，以及从静态训练向集成训练-测试循环的转变，旨在实现鲁棒、可扩展和可信赖的自主驾驶系统。我们维护一个持续更新的仓库，跟踪前沿文献和研究成果，网址为 exttt{https://github.com/Jiaaqiliu/Awesome-Training-Ecosystem-for-E2E-AD}。
cs.RO / 35 / 2610.00936
ShowerFlex: Achieving Pseudo-Static Balancing in a Continuum Shower Hose
ShowerFlex：在连续淋浴软管中实现伪静态平衡
Abstract
With the global population rapidly aging, maintaining independence in Activities of Daily Living (ADLs), particularly bathing or showering, has become a critical challenge. Older adults who sit while showering currently have limited options, often relying on rigid overhead showerheads or flexible hand-held hoses that require continuous gripping and precarious user maneuvers, increasing the risk of falls. To address this need, this paper introduces a highly articulated, pseudo-static balanced continuum mechanism designed specifically for accessible bathing assistance. The proposed mechanism features a modular continuum architecture composed of friction-locked ball-and-socket joints (loc-line), seamlessly integrated with a retractable spring-loaded reel for effective gravity compensation. This hybrid design achieves an intuitive "Push-and-Stay" interaction logic, maintaining an approximate static equilibrium with ergonomically low actuation force, without any external electronic power, while significantly minimizing the force needed for repositioning. Theoretically, we established a recursive forward kinematics framework to construct the geometric model of the structure, coupled with a comprehensive static equilibrium analysis to evaluate the system's holding capacity across its 51-DOF structure. Numerical simulations utilizing Monte Carlo methods validate the system's morphological adaptability and stability at various extreme positions within a standard bath space. Physical experiments further validate the prototype's real-world performance, confirming its intrinsic static stability against gravity and ensuring that the actuation force remains well within the ergonomic capabilities of older adults. Ultimately, this research provides a low-cost, intrinsically safe showerhead design that reduces physical strain, restoring dignity and autonomy in personal hygiene.
Chinese Translation
随着全球人口迅速老龄化，维持日常生活活动（ADLs）中的独立性，特别是在洗澡或淋浴方面，已成为一项重要挑战。坐着淋浴的老年人目前的选择有限，通常依赖于刚性的头顶淋浴喷头或需要持续握持和不稳定用户操作的灵活手持软管，这增加了跌倒的风险。为了解决这一需求，本文提出了一种高度关节化的伪静态平衡连续机制，专门设计用于无障碍洗浴辅助。所提机制具有模块化的连续架构，由摩擦锁定的球窝关节（loc-line）组成，完美结合了可伸缩的弹簧加载卷轴，以有效补偿重力。该混合设计实现了直观的“推并保持”交互逻辑，维持近似静态平衡，具有符合人体工程学的低驱动力，无需任何外部电子电源，同时显著减少重新定位所需的力量。从理论上讲，我们建立了一个递归前向运动学框架，以构建该结构的几何模型，并结合全面的静态平衡分析，以评估系统在其51自由度结构上的承载能力。利用蒙特卡罗方法的数值模拟验证了系统在标准浴室空间内各种极端位置的形态适应性和稳定性。物理实验进一步验证了原型的实际性能，确认其在重力作用下的内在静态稳定性，并确保驱动力保持在老年人可承受的符合人体工程学的范围内。最终，本研究提供了一种低成本、内在安全的淋浴喷头设计，减少了身体负担，恢复了个人卫生中的尊严和自主性。
cs.RO / 36 / 2610.00981
NarrativeFlow: Flow-Based Vision-Language-Action Model Using Robot Velocity Fields
NarrativeFlow：基于流的视觉-语言-动作模型使用机器人速度场
Abstract
We focus on language-conditioned flow-based manipulation, where robot flows (robot velocity fields) serve as embodiment-agnostic, motion-centric representations for leveraging data collected from multiple robot platforms. This task is crucial because language-conditioned manipulation is essential for practical robotic systems, yet scaling robot foundation models remains limited by the labor-intensive collection of embodiment-specific data. Existing methods either coarsely approximate robot flows with sparse keypoint displacements, or cannot handle language-conditioned manipulation. To address this limitation, we propose NarrativeFlow, which models robot flows as continuous velocity fields using a flow-matching formulation conditioned on language. Accordingly, NarrativeFlow generates robot flows that are physically consistent with real-world manipulation. To validate NarrativeFlow, we have conducted experiments on standard datasets for language-conditioned manipulation. The experimental results show that NarrativeFlow outperforms representative baseline methods on standard evaluation metrics. Furthermore, through real-world experiments, we show that NarrativeFlow achieves higher success rates than baseline methods across multiple manipulation tasks. The project page is available at https://shota0520.github.io/NarrativeFlow-project-page/
Chinese Translation
我们关注语言条件下的基于流的操作，其中机器人流（机器人速度场）作为一种与体现无关的、以运动为中心的表示，用于利用从多个机器人平台收集的数据。这个任务至关重要，因为语言条件下的操作对于实际机器人系统至关重要，但扩展机器人基础模型仍然受到劳动密集型收集特定体现数据的限制。现有方法要么粗略地用稀疏关键点位移来近似机器人流，要么无法处理语言条件下的操作。为了解决这一限制，我们提出了NarrativeFlow，它使用基于语言条件的流匹配公式将机器人流建模为连续的速度场。因此，NarrativeFlow生成的机器人流在物理上与现实世界的操作一致。为了验证NarrativeFlow，我们在标准数据集上进行了语言条件下操作的实验。实验结果表明，NarrativeFlow在标准评估指标上优于代表性的基线方法。此外，通过实际实验，我们展示了NarrativeFlow在多个操作任务中实现了比基线方法更高的成功率。项目页面可访问 https://shota0520.github.io/NarrativeFlow-project-page/
cs.RO / 37 / 2610.00982
Divide-and-Remember: Recursive Action-Relevant Memory for Long-Horizon VLA Policies
分而记之：用于长时间跨度视觉-语言-动作（VLA）策略的递归动作相关记忆
Abstract
Vision-language-action (VLA) models struggle on history-dependent manipulation tasks, where the current observation alone does not determine the action, and the policy needs a memory of the history. Existing memory methods decide what to remember by design, for example, keeping frames with large pixel changes, and show inconsistent gains across tasks. We view what to remember as an optimisation problem. From the POMDP formulation of imitation learning, we show that the optimal memory maximises the conditional mutual information $I(a_t; m_t \mid o_t)$ between the action and the memory given the current observation. Intuitively, this means preserving the action-relevant information in the history that is not already contained in the current observation. Based on our analysis, we propose Divide-and-Remember (D&R), a recursive memory method that learns a memory function $m_t = M(h_t)$ and scales to long contexts while staying compute-light. It involves two strategies: (1) the selection over the full history is divided recursively into subproblems of top-$K$ selection over $2K$ tokens, so that fixed-size, lightweight selectors learned end-to-end support an unbounded history; (2) all recursion blocks share one selector, which captures the selection rule common to every block and keeps the method efficient. On RoboMME, a benchmark of 16 long-horizon manipulation tasks that require remembering when, where, what, and how to act, D&R achieves a state-of-the-art average success rate with consistent gains across all four suites under a budget of only 64 tokens; real-robot experiments show the same gain. Code, checkpoints and more results are at https://dnr-memory.github.io/
Chinese Translation
视觉-语言-动作（VLA）模型在历史依赖的操作任务中表现不佳，在这些任务中，仅凭当前观察无法确定动作，策略需要对历史的记忆。现有的记忆方法通过设计决定记住什么，例如，保留像素变化较大的帧，但在不同任务中表现出不一致的增益。我们将记住什么视为一个优化问题。从模仿学习的部分可观马尔可夫决策过程（POMDP）表述出发，我们展示了最优记忆最大化当前观察下动作与记忆之间的条件互信息 $I(a_t; m_t ext{ mid } o_t)$。直观上，这意味着保留历史中与动作相关的信息，而这些信息在当前观察中并不存在。基于我们的分析，我们提出了分而记之（Divide-and-Remember, D&R），这是一种递归记忆方法，学习记忆函数 $m_t = M(h_t)$，能够扩展到长上下文，同时保持计算轻量。它涉及两种策略：（1）对完整历史的选择递归地分解为对 $2K$ 个标记的前 $K$ 选择的子问题，从而使得固定大小、轻量的选择器能够端到端学习并支持无限历史；（2）所有递归块共享一个选择器，该选择器捕捉每个块共有的选择规则，从而保持方法的高效性。在RoboMME上，这是一个包含16个需要记住何时、何地、何物以及如何行动的长时间跨度操作任务的基准，D&R在仅使用64个标记的预算下实现了最先进的平均成功率，并在所有四个任务套件中表现出一致的增益；真实机器人实验显示出相同的增益。代码、检查点和更多结果请见 https://dnr-memory.github.io/
cs.RO / 38 / 2610.01067
Online Planning for Sparse Ground Target Search from a High-Altitude UAV under Partial Observability
高空无人机在部分可观测条件下进行稀疏地面目标搜索的在线规划
Abstract
Unmanned aerial vehicles (UAVs) searching for ground targets from a high altitude face a unique challenge, particularly when the target is already within the field of view but effectively unobservable because of its small apparent scale. Standard object detectors often underperform in such scenarios because of resolution downscaling and limited context. In contrast, active object search frameworks address this challenge by directing the agent to a suitable pose to gather richer visual information. However, flight regulations in urban airspace often restrict such physical movements for UAVs. As an alternative active sensing approach, the UAV can leverage the pan-tilt-zoom (PTZ) mechanism of the onboard camera to dynamically adjust its field of view and sequentially gather enhanced visual information from specific regions of interest. Once the candidate locations of the target are identified, it can deploy more expensive object detection schemes, such as an ensemble of multiple models, to get better reasoning at a fixed scale. In this work, we formulate the sequential exploration with PTZ operation as a partially observable Markov decision process (POMDP), in which the agent maintains a belief state over the target's true location. To solve the POMDP, we deploy partially observable Monte Carlo planning (POMCP), where we condition the sensing reliability on target object scale and deploy selective ensemble detection as an additional reasoning step. We validate our methodology with experiments in a photorealistic simulator under different environmental conditions and vehicle states, showing detection of ground targets at variable scales with significantly fewer steps and minimal dependence on sensor resolution compared to baseline methods.
Chinese Translation
高空搜索地面目标的无人驾驶飞行器（UAV）面临独特的挑战，尤其是在目标已在视野内但由于其小的表观尺度而有效不可观测的情况下。标准物体检测器在这种情况下往往表现不佳，因为分辨率降低和上下文信息有限。相比之下，主动物体搜索框架通过引导代理采取适当的姿态以获取更丰富的视觉信息来应对这一挑战。然而，城市空域的飞行法规通常限制了无人机的这种物理移动。作为一种替代的主动感知方法，无人机可以利用机载摄像头的平移-倾斜-变焦（PTZ）机制动态调整其视野，并依次从特定兴趣区域收集增强的视觉信息。一旦识别出目标的候选位置，就可以部署更昂贵的物体检测方案，例如多个模型的集成，以在固定尺度下获得更好的推理。在本研究中，我们将带有PTZ操作的顺序探索形式化为部分可观测马尔可夫决策过程（POMDP），其中代理保持对目标真实位置的信念状态。为了解决POMDP，我们部署了部分可观测蒙特卡洛规划（POMCP），在此过程中我们将感知可靠性与目标物体尺度相关联，并将选择性集成检测作为额外的推理步骤。我们在不同环境条件和车辆状态下的逼真模拟器中验证了我们的方法，结果显示与基线方法相比，在可变尺度下以显著更少的步骤和对传感器分辨率的最小依赖成功检测地面目标。
cs.RO / 39 / 2610.01072
quARtet Marker: A 3D-Printable Multi-Tag Fiducial for Robust Near-Frontal Pose Estimation
quARtet 标记：一种可3D打印的多标签基准，用于稳健的近正面姿态估计
Abstract
Robotic manipulation of labware is difficult when transparent or reflective objects must be identified and localized. Coded planar fiducials are a practical retrofit: easy to print, they leave the marked face flat and graspable. Yet a single planar tag is least reliable in near-frontal views, where perspective cues fade. Non-planar geometries restore those cues but intrude on the flat face that a parallel-jaw gripper must contact. Our idea is to tilt multiple tags within one compact footprint, so that each tag is seen at a non-frontal angle even when the marker faces the camera. We propose the quARtet marker, a 3D-printable fiducial embodying this idea: all detected corners of its four tilted AprilTags enter one Perspective-n-Point solve, and a shared configuration defines the fabricated geometry and the detector model. Because tilting consumes flat area, its three layouts trade pose-estimation consistency against graspability. In robot-referenced, same-setup fixed-camera experiments, all three layouts reduced the mean frontal orientation error from 2.18 degree for a single planar tag to 0.24-0.47 degree and the root-mean-square position error from 1.50 to 0.17-0.20 mm. A robot-mounted-camera pose-hold test confirmed this separation under closed-loop visual feedback. In swing-down trials under identical conditions, the two layouts with flat contact strips retained the object with about 2 mm of in-grasp slip, whereas the layout without flat strips slipped by roughly 100 mm. For the tested conditions, the results support a rule: the layout without flat strips when pose-estimation consistency dominates, a layout with flat strips when the marked face must remain graspable.
Chinese Translation
在识别和定位透明或反射物体时，实验室设备的机器人操作变得困难。编码平面基准是一种实用的改造方案：易于打印，且标记面保持平坦且易于抓取。然而，单一平面标签在近正面视图中可靠性最低，因为透视线索会减弱。非平面几何形状恢复了这些线索，但会干扰平行夹爪必须接触的平坦面。我们的想法是在一个紧凑的占地内倾斜多个标签，以便即使标记面朝向相机，每个标签也能以非正面角度被看到。我们提出了quARtet标记，这是一种体现这一理念的可3D打印基准：其四个倾斜的AprilTags的所有检测角点进入一个透视-n-点求解，而共享配置定义了制造几何形状和检测器模型。由于倾斜消耗了平坦区域，其三种布局在姿态估计一致性与可抓取性之间进行权衡。在机器人参考的同一设置固定摄像头实验中，所有三种布局将单一平面标签的平均正面方向误差从2.18度降低到0.24-0.47度，将均方根位置误差从1.50毫米降低到0.17-0.20毫米。一个机器人安装的摄像头姿态保持测试在闭环视觉反馈下确认了这一分离。在相同条件下的摆动试验中，两个带有平坦接触条的布局保持了大约2毫米的抓取滑动，而没有平坦条的布局滑动了大约100毫米。对于测试条件，结果支持一个规则：在姿态估计一致性占主导时使用没有平坦条的布局，而在标记面必须保持可抓取时使用带有平坦条的布局。
cs.RO / 40 / 2610.01083
WBAG: A Whole-Body and Attached-Geometry Safety Framework for Vision-Language-Action Manipulation
WBAG：一种用于视觉-语言-动作操作的全身及附加几何安全框架
Abstract
Vision-language-action (VLA) policies have demonstrated impressive capabilities in generalizable robotic manipulation, but their deployment in the real world remains challenging due to potential collisions involving different parts of the robot, manipulated objects, and the surrounding environment. Existing inference-time VLA safety frameworks typically rely on simplified end-effector-centered representations that do not explicitly model the full articulated robot and attached-object geometry. In this paper, we present WBAG, a safety framework that models the robot's whole-body and grasp-dependent attached geometry. WBAG constructs a grasp-conditioned safe set that adapts the protected geometry as objects are grasped, then converts this evolving geometry into differentiable CBF constraints that minimally modify the VLA's native six-dimensional operational-space action for collision avoidance across robot, scene, and attached geometry. On the SafeLIBERO benchmark, a variant of LIBERO augmented with obstacles for safety evaluation, WBAG achieves the best overall safety and safe task success among the evaluated methods under a scene-level safety evaluator that monitors all eligible non-task objects, reaching 97.38\% aggregate Scene Safety and 59.38\% Safe Success.
Chinese Translation
视觉-语言-动作（VLA）策略在可泛化的机器人操作中展现了令人印象深刻的能力，但由于潜在的碰撞风险，包括机器人不同部件、被操作物体和周围环境，实际应用仍然面临挑战。现有的推理时VLA安全框架通常依赖于简化的以末端执行器为中心的表示，这些表示并未明确建模完整的关节机器人及附加物体的几何形状。本文提出了WBAG，一种建模机器人全身及抓取依赖的附加几何的安全框架。WBAG构建了一个抓取条件的安全集，随着物体被抓取而调整保护几何，然后将这种不断演变的几何转换为可微分的控制 barrier 函数（CBF）约束，最小化地修改VLA的原生六维操作空间动作，以实现机器人、场景和附加几何的碰撞避免。在SafeLIBERO基准测试中，WBAG在评估的各方法中实现了最佳的整体安全性和安全任务成功率，在一个监控所有合格非任务物体的场景级安全评估器下，达到了97.38%的整体场景安全性和59.38%的安全成功率。
cs.RO / 41 / 2610.01093
OrbitTAMP: Grounding Language Models for Task and Motion Planning in Spacecraft Rendezvous
OrbitTAMP：为航天器会合中的任务与运动规划奠基语言模型
Abstract
Spacecraft rendezvous and proximity operations (RPO) are currently planned through an expertise-intensive process in which engineers translate high-level operational intent into safe, dynamically feasible trajectories, creating a bottleneck to scalable operations. Large language model (LLM)-based agents could offer an intuitive interface for this process, although their outputs are not inherently grounded in orbital dynamics, operational constraints, or the structure of admissible spacecraft maneuvers. To exploit their semantic reasoning while ensuring the generated plan's physical validity, this paper presents a hierarchical framework for spacecraft task-and-motion planning (TAMP) that grounds LLM reasoning in a graph of reusable behaviors and domain-specific planning modules. Within this framework, a pretrained LLM maps a natural-language command to a partial mission specification. The associated planners then resolve unspecified decisions within the admissible operational space. Finally, trajectory optimization converts the completed mission specification into a dynamically feasible trajectory. Numerical experiments demonstrate that this architecture substantially improves intent recovery over direct LLM generation, achieving 98% exact recovery of partial mission specifications across all evaluated splits when backed by frontier LLMs. Additional test-time-compute experiments show that, for a compact 9B model, verifier-guided revision increases exact recovery from 75% to 88%, while broader behavior-plan search independently improves selection among admissible trajectory realizations. Overall, these results establish a scalable and auditable foundation for language-driven agentic planning of spacecraft RPO.
Chinese Translation
航天器会合与近距离操作（RPO）目前通过一个依赖专业知识的过程进行规划，工程师将高层次的操作意图转化为安全且动态可行的轨迹，这成为可扩展操作的瓶颈。基于大型语言模型（LLM）的代理可以为这一过程提供直观的接口，尽管它们的输出并不固有地基于轨道动力学、操作约束或可接受航天器机动的结构。为了利用其语义推理，同时确保生成计划的物理有效性，本文提出了一个层次化框架，用于航天器的任务与运动规划（TAMP），该框架将LLM推理与可重用行为图和特定领域的规划模块相结合。在该框架内，经过预训练的LLM将自然语言命令映射到部分任务规范。相关规划者随后在可接受的操作空间内解决未指定的决策。最后，轨迹优化将完整的任务规范转化为动态可行的轨迹。数值实验表明，该架构显著改善了意图恢复，相较于直接的LLM生成，在所有评估分割中实现了98%的部分任务规范的准确恢复，前提是使用前沿LLM。额外的测试时计算实验显示，对于一个紧凑的9B模型，验证者引导的修订将准确恢复率从75%提高到88%，而更广泛的行为计划搜索则独立改善了在可接受轨迹实现中的选择。总体而言，这些结果为基于语言驱动的航天器RPO代理规划建立了一个可扩展且可审计的基础。
cs.RO / 42 / 2610.01102
MASkillBlender: Decentralized Whole-Body Coordination for Multi-Humanoid Loco-Manipulation via Skill Blending
MASkillBlender：通过技能融合实现多类人形机器人去中心化全身协调的运动操控
Abstract
Coordinated multi-humanoid loco-manipulation is promising yet challenging due to high-dimensional whole-body control, decentralized decision making, and scalability. While recent reinforcement learning methods have improved single-humanoid whole-body control, extending them to the multi-humanoid setting remains nontrivial and often requires substantial reward engineering or task-specific design. We propose MASkillBlender, a general multi-agent reinforcement learning framework to achieve decentralized multi-humanoid whole-body coordination. By learning a shared decentralized high-level policy over reusable pre-trained single-humanoid skills, MASkillBlender enables coordinated behaviors using only task-level rewards, without requiring task-specific motion references. To improve learning efficiency, we further introduce a permutation-based data augmentation strategy for homogeneous multi-humanoid systems, and theoretically show that the permuted samples preserve the policy-gradient direction of the original samples under the homogeneous Markov game formulation. We evaluate MASkillBlender on multiple multi-humanoid coordination tasks across two humanoid embodiments. Simulation results demonstrate that the proposed framework consistently achieves strong task performance and enables coordinated behaviors across different tasks and humanoid embodiments.
Chinese Translation
协调的多类人形机器人运动操控前景广阔，但由于高维全身控制、去中心化决策和可扩展性等因素，面临诸多挑战。尽管近期的强化学习方法在单类人形机器人全身控制方面取得了一定进展，但将其扩展到多类人形机器人环境仍然不是一件简单的事情，通常需要大量的奖励工程或任务特定设计。我们提出了MASkillBlender，一个通用的多智能体强化学习框架，旨在实现去中心化的多类人形机器人全身协调。通过学习可重用的预训练单类人形机器人技能的共享去中心化高层策略，MASkillBlender仅通过任务级奖励实现协调行为，而不需要任务特定的运动参考。为了提高学习效率，我们进一步引入了一种基于置换的数据增强策略，适用于同质多类人形机器人系统，并理论上证明了在同质马尔可夫博弈框架下，置换样本保留了原始样本的策略梯度方向。我们在两个不同的人形机器人实现上评估了MASkillBlender在多个多类人形机器人协调任务中的表现。仿真结果表明，所提出的框架在不同任务和人形机器人实现中始终能够实现强大的任务性能并促进协调行为。
cs.RO / 43 / 2610.01105
Extreme Length Generalization in a Compact Recurrent Architecture for One-Shot Exploration
紧凑递归架构中的极端长度泛化用于一次性探索
Abstract
Autonomous robots on one-shot missions run over horizons far longer than the trajectories seen during training, under a fixed onboard compute budget. We present FRANK, a 507K-parameter recurrent architecture that combines tau-gated recurrent modules, content-addressable memory, and a feedforward reflex pathway. We evaluate it against recurrent, state-space, and reduced modular baselines at matched parameter count on four algorithmic sequence tasks, trained at length 5-20 and evaluated out to two million tokens. At 100,000x the maximum training length, 6 of 10 FRANK seeds retain exactly 100.0% accuracy, while none of the 50 baseline configurations does, five architectures at ten seeds each with none left incomplete (Fisher exact, two-sided p = 4.2E-6. Targeted lesions across the four tasks yield four distinct component-reliance profiles, consistent with task-dependent allocation across the recurrent, memory, and reflex pathways. Separately, a FRANK policy trained in simulation drives a physical ground vehicle to commanded waypoints through obstacles without teleoperation.
Chinese Translation
在一次性任务中，自主机器人在固定的机载计算预算下，执行的时间范围远超训练期间所见的轨迹。我们提出了FRANK，这是一种507K参数的递归架构，结合了tau门控递归模块、内容寻址存储器和前馈反射通路。我们在四个算法序列任务上对其进行评估，参数数量匹配的情况下与递归、状态空间和简化模块基线进行比较，训练长度为5-20，并评估至两百万个标记。在最大训练长度的100,000倍时，10个FRANK种子中有6个保持了100.0%的准确率，而50个基线配置中没有一个达到该准确率，五种架构各有十个种子，且没有一个未完成（Fisher精确检验，双侧p = 4.2E-6）。在四个任务上进行的有针对性损伤实验显示出四种不同的组件依赖特征，与递归、记忆和反射通路的任务依赖分配一致。此外，在仿真中训练的FRANK策略能够驱动物理地面车辆在没有遥控操作的情况下，通过障碍物到达指定的航点。
cs.RO / 44 / 2610.01171
Experience-Based Feasibility-Aware Generative Adversarial Imitation from Observation under Embodiment Mismatch
基于经验的可行性意识生成对抗模仿学习：在体现不匹配下的观察模仿
Abstract
With the increasing use of robot-free demonstration interfaces that provide state trajectories without action labels, imitation from observation has become a promising approach for learning robot behaviors from human demonstrations. However, due to differences in embodiment and dynamics between humans and robots, demonstrated human motions may not be feasible for the robot, potentially degrading policy performance. In this study, we propose Experience-Based Feasibility-Aware Generative Adversarial Imitation from Observation (EF-GAIfO), which estimates the feasibility of state-only demonstrations from the robot's own experience rather than relying on explicit dynamics models or large prior exploration datasets. A key feature of EF-GAIfO is that the notion of feasibility evolves with policy learning: as the policy improves and the robot experiences a broader range of state transitions, the feasible region is progressively expanded, allowing additional demonstrations to be incorporated into learning. This enables feasibility-aware imitation that adapts to the current stage of policy learning, rather than relying on a pre-designed feasibility criterion. We validate the effectiveness of EF-GAIfO on a locomotion task in simulation and on a real quadruped robot performing a object-reaching-and-grasping task.
Chinese Translation
随着无机器人演示接口的日益普及，这些接口提供状态轨迹而不需要动作标签，观察模仿已成为从人类演示中学习机器人行为的有前景的方法。然而，由于人类与机器人之间在体现和动态上的差异，演示的人类动作可能对机器人来说不可行，从而可能降低策略性能。在本研究中，我们提出了一种基于经验的可行性意识生成对抗模仿学习（Experience-Based Feasibility-Aware Generative Adversarial Imitation from Observation，EF-GAIfO），该方法从机器人的自身经验出发，评估仅基于状态的演示的可行性，而不是依赖于显式的动态模型或大型的先前探索数据集。EF-GAIfO的一个关键特征是可行性概念随着策略学习而演变：随着策略的改进和机器人经历更广泛的状态转变，可行区域逐渐扩展，从而允许将额外的演示纳入学习。这使得可行性意识的模仿能够适应当前的策略学习阶段，而不是依赖于预先设计的可行性标准。我们在仿真中的运动任务以及在真实四足机器人执行物体到达和抓取任务中验证了EF-GAIfO的有效性。
cs.RO / 45 / 2610.01178
Recova: Agent-Guided Failure Recovery for Autonomous Robotic Manipulation
Recova：基于智能体引导的自主机器人操作故障恢复
Abstract
Manipulation failures can leave scenes in states from which a task policy cannot recover. Learning corrective behaviors requires scalable failure exploration and physical grounding. We present Recova, an agent-guided framework that jointly develops task execution and recovery in a reconstructed digital twin, then verifies and refines both through real-world experience. In the twin, the agent diagnoses failures, tests corrective programs, and collects successful task and recovery rollouts for separate policies. During deployment, it monitors progress, invokes a learned or programmatic recovery, verifies scene restoration, and resumes execution. When no suitable recovery is available, a human demonstration resolves the failure and enters the learning loop, allowing the system to expand its recovery capabilities. Physical rollouts and human demonstrations are routed to the corresponding policy for DAgger training. Across six LIBERO-Pro settings and four MolmoSpaces categories, Recova achieves 78.8% and 64.9% mean success, compared with 71.7% and 38.0% for the strongest baselines. With parallel collection across four real-robot workstations, DAgger fine-tuning raises mean success from 23.8% to 77.5%, and recovery skills further raise it to 87.5%. Over four collection rounds on one task, observed human intervention falls from 87.5% to 0%. Together, these results show how agent-guided recovery turns failures into reusable capabilities, improving robustness while progressively reducing human intervention. Project page: https://www.liuisabella.com/Recova
Chinese Translation
操作故障可能导致场景处于无法通过任务策略恢复的状态。学习纠正行为需要可扩展的故障探索和物理基础。我们提出了Recova，一个智能体引导的框架，该框架在重建的数字双胞胎中共同开发任务执行和恢复，然后通过现实世界经验验证和完善这两者。在双胞胎中，智能体诊断故障，测试纠正程序，并为不同的策略收集成功的任务和恢复执行。在部署过程中，它监控进展，调用学习到的或程序化的恢复，验证场景恢复，并恢复执行。当没有合适的恢复方案时，人类示范解决故障并进入学习循环，使系统能够扩展其恢复能力。物理执行和人类示范被路由到相应的策略进行DAgger训练。在六个LIBERO-Pro设置和四个MolmoSpaces类别中，Recova的平均成功率分别达到78.8%和64.9%，而最强基线的成功率为71.7%和38.0%。通过在四个真实机器人工作站的并行收集，DAgger微调将平均成功率从23.8%提高到77.5%，而恢复技能进一步将其提高到87.5%。在一个任务的四轮收集中，观察到的人类干预从87.5%降至0%。这些结果共同表明，智能体引导的恢复将故障转化为可重用的能力，提高了鲁棒性，同时逐步减少了人类干预。项目页面：https://www.liuisabella.com/Recova
cs.RO / 46 / 2610.01185
AFD-CAMLs: Agile Force-Distribution-Aware Planning and Control for Cable-Suspended Aerial Multi-Lifting Systems
AFD-CAMLs：面向缆索悬挂多重提升系统的敏捷力分配感知规划与控制
Abstract
Multiple UAVs can cooperatively transport heavy payloads while controlling their position and orientation. Trajectory-based methods offer high agility while satisfying system constraints, but can produce uneven force distributions when the tension-to-wrench allocation is redundant or ill-conditioned, particularly under geometric mismatch and low-level tracking errors. We propose a hybrid planning-and-control framework to address this problem. A global planner generates payload trajectories and cable-force references by exploring the allocation null space under a prescribed internal-force setting. These references augment the cost of a centralized local planner, promoting feasible force distributions while generating trajectories for all UAVs. An admittance filter then compares the planned forces with onboard cable-tension estimates and adjusts the kinematic references to improve force tracking in degenerate or near-degenerate configurations. Simulations and experiments involving four to ten UAVs demonstrate more balanced tension distributions during both hovering and demanding agile maneuvers, without compromising agility or payload-tracking performance.
Chinese Translation
多个无人机可以协同运输重载，同时控制其位置和方向。基于轨迹的方法在满足系统约束的同时提供了较高的敏捷性，但在张力与扭矩分配冗余或条件不良时，可能会产生不均匀的力分布，尤其是在几何不匹配和低级跟踪误差的情况下。我们提出了一种混合规划与控制框架来解决这一问题。全局规划器通过探索在规定内部力设置下的分配零空间，生成有效载荷轨迹和缆索力参考。这些参考增强了集中式局部规划器的成本，促进了可行的力分布，同时为所有无人机生成轨迹。然后，适应性滤波器将计划的力与机载缆索张力估计进行比较，并调整运动学参考，以改善在退化或近退化配置中的力跟踪。涉及四到十架无人机的仿真和实验表明，在悬停和要求高敏捷性的机动过程中，张力分布更加均衡，同时不影响敏捷性或有效载荷跟踪性能。
cs.RO / 47 / 2610.01219
EIDA: Execution-Interface Dynamics Adaptation for Real-to-Sim-to-Real Robot Navigation
EIDA：用于真实-模拟-真实机器人导航的执行接口动态适应
Abstract
Simulation-to-robot transfer can fail when velocity commands produce motion and feedback that differ from those modeled during policy training. We present execution-interface dynamics adaptation (EIDA), which fits these responses from target-platform execution data without reconstructing actuator dynamics. A model of body-frame pose increments updates simulator geometry, while a separate model predicts the velocity feedback observed by the policy; a short history of velocity feedback is included in the policy input. The fitted models are used within a lightweight GPU-parallel simulator. On the full Jackal and Go2 validation sets, the fitted models reduced position and yaw prediction errors relative to the simulator's predefined motion model. Across 100 benchmark navigation environments evaluated in a separate physics-based simulator, EIDA achieved the highest success rate and navigation score among the compared learned policies, both with and without global guidance. Feedback ablations further supported the need to match policy-facing velocity estimates. On a physical Unitree Go2, EIDA reached the goal without collision in all 20 static-scene trials, compared with 4 of 20 for the baseline. These results show that execution-interface adaptation can improve navigation transfer without detailed actuator simulation.
Chinese Translation
当速度指令产生的运动和反馈与策略训练期间建模的内容不同时，模拟到机器人转移可能会失败。我们提出了执行接口动态适应（EIDA），该方法利用目标平台执行数据拟合这些响应，而无需重建执行器动态。身体框架姿态增量的模型更新模拟器几何形状，而另一个模型则预测策略观察到的速度反馈；策略输入中包含了短期的速度反馈历史。拟合的模型在轻量级的GPU并行模拟器中使用。在完整的Jackal和Go2验证集上，拟合模型相对于模拟器的预定义运动模型减少了位置和偏航预测误差。在一个独立的基于物理的模拟器中评估的100个基准导航环境中，EIDA在比较的学习策略中，无论是否有全局指导，都达到了最高的成功率和导航得分。反馈消融实验进一步支持了匹配策略面向的速度估计的必要性。在物理Unitree Go2上，EIDA在所有20个静态场景试验中无碰撞地达到了目标，而基线仅在20个试验中达成4次。这些结果表明，执行接口适应可以在不需要详细执行器模拟的情况下改善导航转移。
cs.RO / 48 / 2610.01258
ColoACT: Multi-Cue Action Chunking for Smooth Autonomous Colon Navigation on a Self-Propelled Endoscopic Robot
ColoACT：用于自推进内窥镜机器人平滑自主结肠导航的多线索动作分块
Abstract
Autonomous colonoscopic navigation can reduce operator burden and the risk of loop formation or tissue trauma, but remains challenging due to deformable anatomy, weak-texture and specular endoscopic visuals, and contact-rich viscoelastic interactions. Existing methods either rely on geometry-driven pipelines, which are efficient and interpretable yet brittle due to manually engineered features and switching logic, or adopt learning-based policies, whose inferred depth/geometry can become temporally inconsistent or overly smooth under weak texture and specular highlights while simulation-trained variants (e.g., deep reinforcement learning) may further suffer from a sim-to-real gap. We propose ColoACT, an autonomous navigation system that integrates an RGB-D-E based Action Chunking Transformer policy (ColoACT policy) for a compact self-propelled Bevel-Gear-Based Endoscopic Robot (BGER). The ColoACT policy augments RGB with estimated relative depth and a gradient-based pseudo-elevation map to enhance fold-ridge saliency and other high-frequency geometric cues, and enables smooth continuous control of the BGER by predicting overlapping action chunks and fusing them via temporal ensembling. In different \textit{ex-vivo} porcine colons (approximately 60 cm), our system achieves success rates of 85.4\% and 72.5\% in straight and curved segments, respectively, and achieves 70\% success in 90-degree turns and 60\% in double-bend sequences, with feasibility further demonstrated in challenging triple-bend segments. The project page is available at: https://Adamhu1.github.io/ColoACT/.
Chinese Translation
自主结肠镜导航可以减轻操作人员的负担，并降低环路形成或组织损伤的风险，但由于可变形解剖结构、弱纹理和镜面内窥镜视觉以及富含接触的粘弹性相互作用，这一任务仍然具有挑战性。现有方法要么依赖于几何驱动的管道，这些方法高效且可解释，但由于手动设计的特征和切换逻辑而显得脆弱；要么采用基于学习的策略，其推断的深度/几何在弱纹理和镜面高光下可能变得时间不一致或过于平滑，而经过模拟训练的变体（例如深度强化学习）可能进一步面临模拟与现实之间的差距。我们提出了ColoACT，一个集成了基于RGB-D-E的动作分块变换器策略（ColoACT策略）的自主导航系统，适用于紧凑型自推进斜齿轮内窥镜机器人（BGER）。ColoACT策略通过估计的相对深度和基于梯度的伪高程图增强了RGB图像的折叠脊显著性及其他高频几何线索，从而实现对BGER的平滑连续控制，预测重叠的动作分块并通过时间集成进行融合。在不同的 extit{ex-vivo}猪结肠（约60厘米）中，我们的系统在直线和曲线段的成功率分别达到了85.4\%和72.5\\%，在90度转弯中成功率为70\\%，在双弯序列中为60\\%，并在具有挑战性的三弯段中进一步证明了可行性。项目页面可访问： https://Adamhu1.github.io/ColoACT/。
cs.RO / 49 / 2610.01260
PROMO: Preference-conditioned Multi-Objective Reinforcement Learning for Quadrupedal Robots
PROMO：面向偏好的多目标强化学习用于四足机器人
Abstract
Quadrupedal locomotion requires balancing conflicting objectives such as command tracking, stability, and energy efficiency, yet conventional reinforcement learning (RL) hardcodes these priorities into a fixed scalar reward at training time. We present PROMO (Preference-Conditioned Multi-Objective Reinforcement Learning), a semantic multi-objective approach that makes this trade-off an explicit runtime input to a single locomotion policy. PROMO conditions the policy on deployment facing preferences while keeping embodiment-specific locomotion priors fixed, thereby separating operator intent from reward shaping terms required for viable gait generation. Compared with fixed-objective controllers, multi-objective baselines, and independently trained specialists, PROMO achieves objective specialization and robustness from a single deployable policy. Across 100 sampled preferences in simulation, 67 behaviors are non-dominated under exact Pareto dominance, with a mean preference-objective correlation of 0.843, demonstrating broad Pareto coverage and predictable preference response. The same policy transfers zero-shot to a Unitree Go2, where preference changes alone reduce specific energy by up to 30.4%, position error by 38.7%, and peak body-attitude deviation by 59.0% relative to the balanced preference. These results establish preference-conditioned multi-objective RL as a practical runtime interface for adaptive legged locomotion, extending its role beyond offline Pareto-set construction. Open-source code and videos are available at https://amrmousa.com/promo/.
Chinese Translation
四足运动需要平衡诸如指令跟踪、稳定性和能效等相互冲突的目标，而传统的强化学习（RL）在训练时将这些优先级硬编码为固定的标量奖励。我们提出了PROMO（面向偏好的多目标强化学习），这是一种语义多目标方法，使得这种权衡成为单一运动策略的显式运行时输入。PROMO在部署时根据偏好条件化策略，同时保持特定于体现的运动先验不变，从而将操作员意图与生成可行步态所需的奖励塑造项分离。与固定目标控制器、多目标基线和独立训练的专家相比，PROMO从单一可部署策略中实现了目标专业化和鲁棒性。在模拟中对100个采样偏好进行测试时，67种行为在精确的帕累托支配下是非支配的，平均偏好-目标相关性为0.843，展示了广泛的帕累托覆盖和可预测的偏好响应。同一策略在Unitree Go2上零-shot转移，其中仅偏好变化就使特定能量降低了最多30.4%，位置误差降低了38.7%，峰值身体姿态偏差降低了59.0%，相对于平衡偏好。这些结果确立了面向偏好的多目标强化学习作为自适应腿部运动的实用运行时接口，扩展了其在离线帕累托集构建之外的作用。开源代码和视频可在 https://amrmousa.com/promo/ 获取。
cs.RO / 50 / 2610.01288
Interaction-Stiffness-Guided Basis Allocation in Dynamic Movement Primitives for Efficient Skill Transfer
基于交互刚度引导的动态运动原语中的基函数分配以实现高效技能转移
Abstract
Dynamic Movement Primitives (DMPs) provide a compact and stable formulation for trajectory representation and generalization in robot skill learning. However, their predefined basis layout limits the allocation of approximation capacity according to stage-dependent precision requirements. To address this issue, this article proposes Stage-Criticality-Guided Dynamic Movement Primitives (SC-DMPs) with adaptive basis allocation for precision-critical skill learning. Operator-robot interaction stiffness and a trajectory-consistency cue derived from cross-demonstration task-space variability are integrated to construct a stage-criticality index. Guided by this index, basis centers are redistributed in normalized time through inverse cumulative criticality and mapped to the canonical phase domain, while their bandwidths are refined to adjust local approximation support. This enables denser and more flexible representation at high-criticality stages while retaining sparser allocation elsewhere. Experiments on handwriting trajectories and three real-robot tasks show that the inferred criticality is concentrated in geometrically demanding and task-constrained regions. Comparisons with DMPs, ProMPs, ProDMP, GP-MP, and KMP demonstrate improved trajectory reproduction, endpoint generalization, and task-critical accuracy while retaining a compact model and the stable structure of classical DMPs.
Chinese Translation
动态运动原语（Dynamic Movement Primitives, DMPs）为机器人技能学习中的轨迹表示和泛化提供了一种紧凑且稳定的表述。然而，其预定义的基函数布局限制了根据阶段依赖的精度要求分配近似能力。为了解决这一问题，本文提出了阶段关键性引导的动态运动原语（Stage-Criticality-Guided Dynamic Movement Primitives, SC-DMPs），实现了针对精度关键技能学习的自适应基函数分配。通过整合操作者与机器人之间的交互刚度以及从跨示范任务空间变异性中得出的轨迹一致性线索，构建了一个阶段关键性指标。在该指标的引导下，基函数中心通过逆累积关键性在归一化时间内重新分配，并映射到规范相位域，同时其带宽经过调整以适应局部近似支持。这使得在高关键性阶段能够实现更密集和更灵活的表示，同时在其他地方保持较稀疏的分配。在手写轨迹和三个真实机器人任务上的实验表明，推断出的关键性集中在几何要求高和任务受限的区域。与DMPs、ProMPs、ProDMP、GP-MP和KMP的比较显示，在保持紧凑模型和经典DMPs稳定结构的同时，轨迹再现、端点泛化和任务关键精度得到了改善。
cs.RO / 51 / 2610.01301
Continual Learning for 6-DoF Grasp Synthesis via Experience and Demonstrations
基于经验和示范的6自由度抓取合成的持续学习
Abstract
Most current grasp synthesis systems are trained offline and remain fixed during deployment. While this works well when deployment conditions resemble the training data, performance can degrade when robots encounter conditions they have not seen before, such as unfamiliar objects. In this work, we present a continual-learning framework for single-view 6-DoF grasp synthesis for a parallel-jaw gripper in cluttered scenes. Rather than finetuning a large parametric model, our method adapts through memory in a learned embedding space: grasp outcomes update future grasp scores, while optional user demonstrations are recalled and transferred to new scenes as additional candidate grasps. We evaluate our method in simulation and in extensive real-world experiments comprising over 1500 grasp trials. We show that our method matches the performance of existing 6-DoF grasping baselines even before adaptation, improves online on unseen objects from categories absent or underrepresented during training, and supports long-horizon continual learning with limited forgetting. In real-world experiments, our method reaches over 90\% success rates on several challenging object categories after only 50 online grasp attempts. Videos and code at https://giuschio.github.io/cl_grasping/.
Chinese Translation
目前大多数抓取合成系统是在离线状态下训练的，并且在部署期间保持固定。当部署条件与训练数据相似时，这种方法效果良好，但当机器人遇到未见过的条件（例如不熟悉的物体）时，性能可能会下降。在本研究中，我们提出了一种用于在杂乱场景中为平行夹爪进行单视图6自由度抓取合成的持续学习框架。我们的方法不是通过微调大型参数模型，而是通过在学习的嵌入空间中的记忆进行适应：抓取结果更新未来的抓取评分，而可选的用户示范被回忆并转移到新场景中作为额外的候选抓取。我们在仿真和超过1500次抓取试验的广泛真实世界实验中评估了我们的方法。我们展示了我们的方法在适应之前就能匹配现有6自由度抓取基线的性能，并在训练期间缺失或不足代表的类别中对未见物体进行在线改进，同时支持有限遗忘的长期持续学习。在真实世界实验中，我们的方法在仅进行50次在线抓取尝试后，在多个具有挑战性的物体类别上达到了超过90%的成功率。视频和代码可在 https://giuschio.github.io/cl_grasping/ 获取。
cs.RO / 52 / 2610.01310
Ex vivo breach detection using electrical conductivity during robotic pedicle drilling in the spine
在脊柱机器人椎弓根钻孔过程中利用电导率进行体外破损检测
Abstract
Purpose: Pedicle screw placement is technically demanding in scoliosis treatment. High precision is required due to limited visibility, anatomical variability, and the risk of complications. Although robotic systems assist CT-based planning and execution, they still rely on ionizing intraoperative imaging and complex registration. This study proposes robotic pedicle drilling with real-time preventive breach detection using electrical bioimpedance sensing. Methods: We developed a robotic approach combined with a pedicle-drilling tool equipped with a proprioceptive electrical bioimpedance sensor developed by SpineGuard. A real-time detection algorithm was designed to analyze the electrical bioimpedance signal during drilling and identify abrupt changes in conductivity associated with potential breaches towards the spinal canal. The method operates without external devices or sensors. Results: The ex vivo experiments showed that the proposed method prevented breaches in 100 of the 51 drilling cases. These findings demonstrate the system's ability to detect potentially hazardous events during drilling and to stop the procedure before. The ex vivo experiments demonstrated that the proposed method prevented breaches in all 51 drilling cases. Conclusions: This work demonstrates the feasibility of robotic pedicle drilling with electrical bioimpedance sensing for real-time breach prevention. Using only the tool signal, the method eliminates the need for external sensing systems and supports safer pedicle screw placement.
Chinese Translation
目的：椎弓根螺钉的放置在脊柱侧弯治疗中技术要求高。由于可视性有限、解剖变异性和并发症风险，要求具备高精度。尽管机器人系统辅助基于CT的规划和执行，但仍依赖于电离辐射的术中成像和复杂的配准。本研究提出了一种结合电生物阻抗传感的机器人椎弓根钻孔方法，实现实时预防性破损检测。方法：我们开发了一种机器人方法，结合了配备有SpineGuard开发的本体感知电生物阻抗传感器的椎弓根钻孔工具。设计了一种实时检测算法，分析钻孔过程中的电生物阻抗信号，并识别与潜在破损相关的导电性突变。该方法无需外部设备或传感器即可运行。结果：体外实验表明，所提出的方法在51个钻孔案例中成功防止了100次破损。这些发现展示了系统在钻孔过程中检测潜在危险事件的能力，并能在发生之前停止操作。体外实验证明，所提出的方法在所有51个钻孔案例中防止了破损。结论：本研究展示了利用电生物阻抗传感进行实时破损预防的机器人椎弓根钻孔的可行性。仅使用工具信号，该方法消除了对外部传感系统的需求，支持更安全的椎弓根螺钉放置。
cs.RO / 53 / 2610.01319
Cross-entropy optimization with prioritized constraints
带有优先约束的交叉熵优化
Abstract
When constraints conflict, an optimizer must determine which requirements to preserve and which to relax. On the one hand, a priority ordering specifies which requirements take precedence. On the other hand, penalty-based formulations encode their relative importance through numerical weights. Depending on these weights, a solution can improve its weighted score while violating intended priorities. We introduce TierCEM, a variant of the cross-entropy method that incorporates strict constraint priorities directly into elite selection without requiring per-constraint importance weights. TierCEM works by sequentially filtering sampled candidates, from highest- to lowest-priority constraint. If and when a constraint eliminates all remaining candidates, TierCEM returns to the last nonempty set and selects elites with the smallest violations of that blocking constraint, recursively preserving satisfaction of all higher-priority constraints. We evaluate TierCEM on 2D navigation and contact-rich pushing tasks in proprioceptive and learned world-model settings. Experiments show that reversing the constraint ordering changes which constraints are violated under conflict. Prioritizing progress toward the task objective also enables TierCEM to relax lower-priority constraints when they would otherwise prevent further progress.
Chinese Translation
当约束发生冲突时，优化器必须确定保留哪些要求以及放宽哪些要求。一方面，优先级排序指定了哪些要求优先考虑。另一方面，基于惩罚的公式通过数值权重编码它们的相对重要性。根据这些权重，解决方案可以在违反预期优先级的情况下提高其加权得分。我们提出了TierCEM，这是一种交叉熵方法的变体，它将严格的约束优先级直接纳入精英选择中，而无需每个约束的重要性权重。TierCEM通过顺序过滤采样候选者，从最高优先级约束到最低优先级约束。如果某个约束消除了所有剩余候选者，TierCEM将返回到最后一个非空集合，并选择在该阻塞约束下违反最小的精英，递归地保持所有更高优先级约束的满足。我们在自我感知和学习的世界模型设置下，对TierCEM在二维导航和接触丰富的推送任务中进行了评估。实验表明，反转约束排序会改变在冲突下违反的约束。优先考虑朝向任务目标的进展也使得TierCEM能够在低优先级约束会阻碍进一步进展时放宽这些约束。
cs.RO / 54 / 2610.01351
Is Success All You Need? Investigating the Impact of Input Perturbations on VLA Behaviour in Tabletop Manipulation Tasks
成功就是一切吗？探讨输入扰动对桌面操作任务中视觉-语言-动作（VLA）行为的影响
Abstract
Vision-Language-Action (VLA) models have achieved high task success rates on robot manipulation task benchmarks. More recently, there has been an emphasis on evaluating the robustness of VLA models to perturbations. However, this robustness is still predominantly measured through Task Success Rate (TSR). In this work, we propose a benchmark-agnostic evaluation framework to measure the behavioural robustness of models by characterising how successful trajectories are executed under perturbation. We implement this methodology by extending the widely-used LIBERO and LIBERO-Plus benchmarks. Across three state-of-the-art VLA models, four LIBERO task suites and seven perturbation conditions, we evaluate changes in both typical successful behaviour and its variability, including metrics of motion smoothness, efficiency and gripper behaviour. We find that perturbations can alter the behaviour of successful trajectories, a phenomenon which cannot necessarily be inferred from TSR alone. Across LIBERO suites, we identify cases where state-of-the-art VLA models achieve comparable TSR under the same perturbation condition, yet behaviour on successful trajectories diverges substantially. Therefore, to have a more robust assessment of task performance, we argue that suitable measures of robustness should capture not only whether a task is completed, but also how the robot behaves while completing it. When evaluating the robustness of VLA models, TSR may be complemented by behavioural evaluation metrics that characterise the nature and variability of successful task execution by robots.
Chinese Translation
视觉-语言-动作（VLA）模型在机器人操作任务基准测试中取得了高任务成功率。最近，研究者们越来越重视评估VLA模型对扰动的鲁棒性。然而，这种鲁棒性仍然主要通过任务成功率（TSR）来衡量。在本研究中，我们提出了一种与基准无关的评估框架，通过表征在扰动下成功轨迹的执行方式来测量模型的行为鲁棒性。我们通过扩展广泛使用的LIBERO和LIBERO-Plus基准来实现这一方法论。在三种最先进的VLA模型、四个LIBERO任务套件和七种扰动条件下，我们评估了典型成功行为及其变异性的变化，包括运动平滑度、效率和夹持器行为等指标。我们发现，扰动可以改变成功轨迹的行为，这一现象并不能仅通过TSR推断。在LIBERO任务套件中，我们识别出一些案例，其中最先进的VLA模型在相同的扰动条件下实现了可比的TSR，但成功轨迹的行为却大相径庭。因此，为了对任务性能进行更为鲁棒的评估，我们认为适当的鲁棒性度量应不仅捕捉任务是否完成，还应关注机器人在完成任务时的行为。在评估VLA模型的鲁棒性时，TSR可以通过行为评估指标进行补充，这些指标表征了机器人成功执行任务的性质和变异性。
cs.RO / 55 / 2610.01397
Continue, Abort, or Fall: Viability-Aware Policy Selection (VAPS) for Safe Humanoid Acrobatics
继续、终止或坠落：基于生存能力的安全人形体操政策选择（VAPS）
Abstract
Dynamic humanoid motions such as flips risk hardware damage due to suboptimal policies, disturbances or sim-to-real gaps. A motion tracking policy offers no way out once the maneuver leaves its reference, and a backup policy needs to take over to protect the hardware for a minimum-damage landing. Which backup to use matters as much as when to switch. We present Viability-Aware Policy Selection (VAPS), which treats safety as a policy-conditioned, receding-horizon decision. Besides a protective fall policy, we also train an abort policy which can abort the motion at any time, landing on its feet. At every control step, learned predictors estimate whether the nominal tracking policy and the abort policy remain viable over a short horizon, and a least-sacrificial hierarchy keeps the most task-ambitious behavior that remains viable. In simulation with randomized disturbances, VAPS sharply reduces head contact and hand contact, which are the dominant sources of hardware damage, with both a Unitree G1 and a LimX Oli; on the LimX Oli, we validate the viability predictors and the full VAPS controller for side-flip motions. VAPS Pareto-dominates the strongest single-network alternatives we could train, including an end-to-end safe-tracking policy and students distilled from VAPS's own oracle-routed decisions, in both task success and head impact. We also show that VAPS is a powerful framework to supervise undertrained policies and protect the hardware.
Chinese Translation
动态人形动作如翻转由于次优政策、干扰或模拟与现实之间的差距而存在硬件损坏的风险。一种运动跟踪政策在机动离开其参考时无法提供解决方案，备份政策需要接管以保护硬件实现最小损害着陆。选择使用哪个备份政策与何时切换同样重要。我们提出了基于生存能力的政策选择（VAPS），将安全视为一种政策条件下的、递归的决策。除了保护性坠落政策外，我们还训练了一种可以随时终止动作的终止政策，使其能够安全着陆。在每个控制步骤中，学习到的预测器估计名义跟踪政策和终止政策在短期内是否仍然可行，而最小牺牲层级保持最具任务雄心的可行行为。在随机干扰的仿真中，VAPS显著减少了头部接触和手部接触，这些是硬件损坏的主要来源，适用于Unitree G1和LimX Oli；在LimX Oli上，我们验证了生存能力预测器和完整的VAPS控制器在侧翻动作中的有效性。VAPS在任务成功率和头部冲击方面优于我们能够训练的最强单网络替代方案，包括端到端安全跟踪政策和从VAPS自身的oracle引导决策中提炼的学生。我们还展示了VAPS是一个强大的框架，可以监督训练不足的政策并保护硬件。
cs.RO / 56 / 2610.01477
ALFRED: Requirement-driven development of an open-source mobile manipulator for long-term plant monitoring
ALFRED：基于需求驱动的开源移动操控器开发，用于长期植物监测
Abstract
Tracking seasonal change in crops and forests requires observing the same plants repeatedly. Ground robots can do this at close range, and a manipulator gives their sensors more viewpoints. Yet the robots behind long-term field datasets are rarely released with their design files, and how a robot's own structure limits arm reach and occludes its sensors is seldom compared between builds. We present ALFRED, an open-source mobile manipulator built from commercially available components. It carries a six-degree-of-freedom arm, LiDAR, RGB-D cameras, RTK GNSS and an IMU on an Ackermann-steered base, all mounted on a reconfigurable aluminium strut frame, and runs containerised ROS software. It was developed through four builds against six requirements for repeated outdoor deployment: durability, modularity, repairability, sensing reach, endurance and reproducibility. Model-based analysis of the last three builds shows the usable share of the arm's reachable poses rising from 34.0% to 60.0% and then 66.1%, and ray casting shows that only the final build keeps the frame-mounted LiDAR's horizontal view clear both forwards and backwards. ALFRED completed a year of monthly forest surveys (528 traversals) without missing a scheduled collection. This was despite battery degradation, reconfiguration for another researcher's study, and the parallel development of ALFRED 2.0 for autonomous crop-row operation, with each switch between builds taking about six hours. The deployment also showed that mechanical modularity is only as dependable as the robot description that tracks it.
Chinese Translation
追踪作物和森林的季节变化需要对相同植物进行反复观察。地面机器人可以在近距离进行此项工作，而操控器则为其传感器提供了更多的视角。然而，长期田野数据集背后的机器人很少发布其设计文件，且机器人自身结构如何限制臂部的伸展和遮挡传感器在不同版本之间的比较也很少进行。我们提出了ALFRED，一款由市售组件构建的开源移动操控器。它配备了一个六自由度的机械臂、激光雷达（LiDAR）、RGB-D摄像头、实时动态定位系统（RTK GNSS）和惯性测量单元（IMU），并安装在一个可重构的铝支架框架上，运行容器化的ROS软件。该设备通过四个版本的开发，满足六项针对重复户外部署的要求：耐用性、模块化、可修复性、感知范围、耐久性和可重复性。对最后三个版本的基于模型的分析显示，机械臂可达姿态的可用比例从34.0%上升至60.0%，再到66.1%；光线投射显示，只有最后一个版本保持了框架安装的LiDAR的水平视野在前后均清晰。ALFRED完成了一年的月度森林调查（528次穿越），没有错过任何一次计划的采集。这一切是在电池衰减、为另一位研究者的研究进行重新配置以及ALFRED 2.0的并行开发（用于自主作物行操作）的情况下完成的，每次版本切换大约需要六小时。该部署还表明，机械模块化的可靠性仅与跟踪其的机器人描述同样可靠。
cs.RO / 57 / 2610.01505
OpenSpace Lab Solution to the IROS 2026 Indoor Exploration Competition
OpenSpace Lab在IROS 2026室内探索竞赛中的解决方案
Abstract
This report presents the \textbf{OpenSpace Lab}'s solution to the Competition on Intelligent Information Gathering for Single and Multi-Robot Systems Workshops, organized as part of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026. Our team reached 1st place in the Single-Robot Public Track and 3rd place in both the Single- and Multi-Robot Private Tracks. The single-robot framework utilizes pre-trained map completion predictions for global planning to prioritize unexplored areas. To reconcile map coverage with limited operation time, we introduce a remaining-time-based exploration strategy that integrates homing constraints into the decision-making process. For multi-robot exploration, we utilize a utility-driven target selection strategy that balances observation gains, movement costs, and budget constraints, leveraging shared map and intent data to eliminate redundant search and maximize coordination efficiency. Our solution reached a 61.04\% coverage rate in the Single-Robot Public Track, while reaching 39.53\% and 39.91\% coverage in the Single- and Multi-Robot Private Tracks, respectively. An extended full-length paper based on this report is currently being prepared for submission, and the source code will be released upon acceptance of the full manuscript at https://github.com/OpenSpace-Lab/Indoor-Exploration-IROS2026.
Chinese Translation
本报告介绍了 extbf{OpenSpace Lab}在IEEE/RSJ国际智能机器人与系统会议（IROS）2026期间举办的单机器人和多机器人系统智能信息收集竞赛中的解决方案。我们的团队在单机器人公开赛道中获得了第一名，在单机器人和多机器人私有赛道中分别获得了第三名。单机器人框架利用预训练的地图补全预测进行全球规划，以优先考虑未探索区域。为了在有限的操作时间内协调地图覆盖，我们引入了一种基于剩余时间的探索策略，将归巢约束整合到决策过程中。在多机器人探索中，我们采用了一种基于效用的目标选择策略，平衡观察收益、移动成本和预算约束，利用共享地图和意图数据消除冗余搜索，最大化协调效率。我们的解决方案在单机器人公开赛道中达到了61.04\%的覆盖率，而在单机器人和多机器人私有赛道中分别达到了39.53\\%和39.91\\%的覆盖率。基于本报告的扩展完整论文目前正在准备提交，源代码将在完整手稿被接受后发布，链接为https://github.com/OpenSpace-Lab/Indoor-Exploration-IROS2026。
cs.RO / 58 / 2610.01559
Completion Aware Guidance for World Action Models
面向完成的世界行动模型指导
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%.
Chinese Translation
世界行动模型（World Action Models, WAMs）预测视觉未来和机器人动作，但它们仍然容易受到任务不完整想象的影响，在这种情况下，合理的、与动作一致的预测会忽略任务完成所需的过渡。在本文中，我们表明这种失败并非源于世界模型的基础结构，而是在适应短时控制时出现的，这种适应可能反复偏向于合理的局部延续而非任务完成的过渡。为了解决这个问题，我们引入了面向完成的指导（Completion Aware Guidance, CAG），这是一种无训练的采样方法，旨在引导生成朝向任务完成。在代表性的WAMs中，CAG使得在RoboTwin 2.0子集上的成功率从64%提高到70%，在零样本模拟中从69%提高到75%，同时将任务不完整想象的比例从79%降低到40%。
cs.RO / 59 / 2610.01573
LiDARFlow: Real-Time Panel-Based MAV Guidance in Unknown Environments
LiDARFlow：在未知环境中基于面板的微型无人机实时引导
Abstract
This paper presents a guidance algorithm for micro aerial vehicles operating in unknown, cluttered environments using only onboard sensing. The method is based on a panel formulation originally derived from aerodynamic potential-flow theory and generates smooth, collision-free guidance vectors from locally perceived obstacles. The approach is extended to unknown environments by constructing and updating the obstacle representation online from onboard LiDAR measurements. The resulting obstacle-avoidance field is integrated with a nominal guiding vector field to produce the final control input. The system is experimentally validated in indoor flight tests under two scenarios: waypoint navigation and directional guidance. In both cases, the vehicle successfully completes its task while avoiding all obstacles in real time using only onboard perception. The results demonstrate that the method is computationally lightweight and suitable for onboard implementation, with pointcloud processing identified as the main practical limitation. These results support the feasibility of lightweight onboard guidance in unknown environments.
Chinese Translation
本文提出了一种用于在未知且杂乱环境中仅依靠机载传感器操作的微型无人机引导算法。该方法基于最初源自气动潜流理论的面板公式，从局部感知的障碍物生成平滑且无碰撞的引导向量。通过在线构建和更新来自机载激光雷达（LiDAR）测量的障碍物表示，该方法扩展到未知环境。最终生成的避障场与名义引导向量场相结合，以产生最终的控制输入。该系统在室内飞行测试中进行了实验验证，涵盖了两种场景：航点导航和方向引导。在这两种情况下，飞行器成功完成任务，同时实时避开所有障碍物，仅依靠机载感知。结果表明，该方法计算轻量，适合机载实现，点云处理被识别为主要的实际限制。这些结果支持在未知环境中轻量机载引导的可行性。
cs.RO / 60 / 2610.01612
ReCo: Response-Consistent Locomotion with Policy-Aware MPC for Legged Manipulation
ReCo：基于策略感知的模型预测控制下响应一致的腿部运动
Abstract
Continuous legged manipulation requires accurate end-effector tracking while the base keeps walking. Combining reinforcement learning (RL) with model predictive control (MPC) suits this task: the learned policy provides robust locomotion, while MPC coordinates the base and arm to compensate for tracking errors. However, MPC can compensate only for base motion that it can predict, and a learned policy's command response varies with gait phase, contact, and payload. We present ReCo, a framework that couples response-consistent locomotion with policy-aware MPC for legged manipulation. Response shaping trains the policy to respond to commands consistently and repeatably across randomized dynamics. An identified closed-loop response model then lets MPC jointly plan locomotion commands and arm motion. On the simulation benchmark, ReCo reduces position and orientation root-mean-square error (RMSE) by 28.7% and 27.4% relative to the best baseline for each metric. Real-world experiments demonstrate onboard continuous legged manipulation with coordinated base and arm motion.
Chinese Translation
连续的腿部操作需要在基座行走的同时准确跟踪末端执行器。将强化学习（RL）与模型预测控制（MPC）相结合适合这一任务：学习到的策略提供了稳健的运动，而MPC协调基座和手臂以补偿跟踪误差。然而，MPC只能补偿其能够预测的基座运动，而学习到的策略的指令响应会随着步态阶段、接触和负载的变化而变化。我们提出了ReCo，一个将响应一致的运动与策略感知的MPC结合用于腿部操作的框架。响应塑形训练策略在随机动态下一致且可重复地响应指令。然后，识别的闭环响应模型使得MPC能够共同规划运动指令和手臂运动。在仿真基准测试中，ReCo相较于每个指标的最佳基线减少了28.7%和27.4%的位置和方向均方根误差（RMSE）。实际实验展示了协调的基座和手臂运动下的连续腿部操作。
cs.RO / 61 / 2610.01644
SonarVoxNet: Diver Detection in 3D Bounding Box using 3D Sonar
SonarVoxNet：基于3D声纳的3D边界框潜水员检测
Abstract
Autonomous underwater vehicles (AUVs) assisting human divers must continuously track not only the diver's 3D position but also their full-body orientation. However, vision-based perception is unreliable underwater, and forward-looking sonar -- despite being widely used -- discards the elevation information needed for orientation estimation, posing a fundamental limitation. Recently commercialized 3D sonar preserves elevation but produces sparse, noisy returns, and existing detectors are built for dense LiDAR data and for targets that remain upright and rotate only about the yaw axis (e.g., vehicles, pedestrians), making them unable to represent a freely pitching and rolling diver. To address this gap, we present two contributions. First, SonarVoxNet adapts a voxel-based encoder and an anchor-free center-based detection head to 3D sonar data, replacing the conventional yaw-only rotation representation with a continuous 6D rotation parameterization to predict full 9-DoF oriented bounding boxes -- to our knowledge, the first 3D sonar diver detector to do so. Second, Diver3D is the first public 3D sonar dataset with full 3D orientation labels for divers in diverse, non-upright poses, collected at a natural cave-diving site. Through controlled ablations over the backbone and detection head, we show that the dominant factor behind accurate 3D sonar-based diver detection is the transition from yaw-only rotation to full-SO(3) rotation. This transition substantially improves detection accuracy and reduces orientation error. These results demonstrate that full-body diver orientation is recoverable from 3D sonar alone, laying the groundwork for future work on diver pose estimation and diver-robot interaction.
Chinese Translation
辅助人类潜水员的自主水下航行器（AUV）必须持续跟踪潜水员的3D位置以及全身姿态。然而，基于视觉的感知在水下不可靠，尽管前视声纳被广泛使用，但它丢弃了进行姿态估计所需的高度信息，这构成了一个基本限制。最近商业化的3D声纳保留了高度信息，但产生稀疏且噪声较大的回波，而现有的检测器是为密集的激光雷达数据和仅在偏航轴上旋转的目标（例如，车辆、行人）构建的，无法表示自由俯仰和滚动的潜水员。为了解决这一问题，我们提出了两个贡献。首先，SonarVoxNet将基于体素的编码器和无锚点中心检测头适配于3D声纳数据，用连续的6D旋转参数化替代传统的仅偏航旋转表示，以预测完整的9自由度定向边界框——据我们所知，这是首个实现这一目标的3D声纳潜水员检测器。其次，Diver3D是首个公开的3D声纳数据集，包含多样化非直立姿态的潜水员的完整3D姿态标签，数据收集于一个自然的洞穴潜水地点。通过对主干网络和检测头的控制性消融实验，我们展示了准确的基于3D声纳的潜水员检测的主要因素是从仅偏航旋转过渡到完整的SO(3)旋转。这一过渡显著提高了检测精度并减少了姿态误差。这些结果表明，仅通过3D声纳即可恢复全身潜水员的姿态，为未来的潜水员姿态估计和潜水员与机器人交互的研究奠定了基础。
cs.RO / 62 / 2610.01668
Learning a Resolution-Consistent Jacobian Field for Bio-Inspired Rigid-Soft Finger
学习用于生物启发的刚柔指的分辨率一致雅可比场
Abstract
Bio-inspired tendon-driven rigid-soft coupled dexterous fingers exhibit strong nonlinearity and configuration-dependent sensitivity, making accurate modeling challenging. In discrete-time control, Jacobian-based kinematic algorithms typically rely on point-wise local linear approximations, which makes their performance sensitive to sensor sampling frequency and controller update frequency. To address this issue, we propose Jacobian Flow Matching (JFM), a structured learning framework based on Conditional Flow Matching (CFM), to learn a resolution-consistent Jacobian field that models actuation-to-motion transitions as a dynamical flow. The proposed framework supports both single-step prediction and continuous rollout via ODE integration, enabling consistent inference across temporal resolutions. Experiments on a tendon-driven rigid-soft finger show that the proposed method suppresses outlier errors and improves single-step prediction accuracy, reducing the global average RMSE by over 53% compared with a baseline discrete Jacobian learning approach. For long-horizon prediction, trajectories recovered via ODE integration achieve higher fidelity under sparse sampling (Stride = 8), reducing the RMSE median by 14.43% and the error variance by 24.87%. These results demonstrate that the learned flow-based Jacobian field provides an effective local model for offline multi-step trajectory optimization in rigid-soft coupled nonlinear systems.
Chinese Translation
生物启发的腱驱动刚柔耦合灵巧手指表现出强非线性和依赖于配置的敏感性，使得准确建模变得具有挑战性。在离散时间控制中，基于雅可比的运动学算法通常依赖于逐点局部线性近似，这使得它们的性能对传感器采样频率和控制器更新频率敏感。为了解决这个问题，我们提出了雅可比流匹配（Jacobian Flow Matching, JFM），这是一个基于条件流匹配（Conditional Flow Matching, CFM）的结构化学习框架，用于学习一个分辨率一致的雅可比场，该场将驱动到运动的转变建模为动态流。所提出的框架支持单步预测和通过常微分方程（ODE）积分的连续展开，使得在不同时间分辨率下能够进行一致的推断。在腱驱动的刚柔指上的实验表明，所提出的方法抑制了异常值误差，并提高了单步预测的准确性，与基线离散雅可比学习方法相比，全球平均均方根误差（RMSE）减少了超过53%。对于长时间预测，通过ODE积分恢复的轨迹在稀疏采样（步幅 = 8）下实现了更高的保真度，RMSE中位数减少了14.43%，误差方差减少了24.87%。这些结果表明，学习到的基于流的雅可比场为刚柔耦合非线性系统中的离线多步轨迹优化提供了有效的局部模型。
cs.RO / 63 / 2610.01682
Beyond Leaderboard Scores: A Deployment-Focused Protocol for Interpretable Tracking Evaluation in Pedestrian-Centric Environments
超越排行榜分数：一种针对行人中心环境的可解释跟踪评估的部署聚焦协议
Abstract
Mobile robots operating among pedestrians need trajectories that become available quickly, remain spatially credible through missed observations, preserve identity, and fit within an embedded computing budget. Aggregate tracking scores provide limited insight into when and how trajectories fail, while varying detector inputs can confound tracker and detector quality. We present a deployment-focused, tracker-only evaluation protocol that uses shared detections to isolate tracker behavior and directly evaluates initialization, detector-gap continuation, identity recovery, close-neighbor association, and load-dependent tracker-step runtime, while Higher Order Tracking Accuracy (HOTA) is retained as a complementary aggregate measure. We apply the protocol to the JackRabbot Dataset and Benchmark (JRDB) using six open-source trackers and our lightweight Pedestrian Reference Tracker (PedRefTrack), together with a GT-assisted variant that estimates the remaining tracker-side gap under idealized association and motion. Under fixed detections, the non-GT trackers span only 24.26%-29.67% HOTA yet exhibit markedly different capability profiles. After 1.0 s without detector support, no tracker without GT assistance maintains spatially correct, same-identity output in more than half of eligible cases, making missing-observation continuation the dominant limitation among the tested properties. Close-neighbor failures are smaller and increase mainly at the shortest separations. Tracker-step runtime on an NVIDIA Jetson Orin is heavy-tailed and load-sensitive, causing several trackers to fall below the 10 Hz real-time target in crowded frames. The protocol provides a reproducible way to characterize tracker behavior and deployment suitability in pedestrian-centric environments. Code and evaluation scripts are released at https://github.com/SCAI-Lab/tracker_eval.
Chinese Translation
在行人中运行的移动机器人需要快速可用的轨迹，这些轨迹在错过观测时保持空间可信性，保持身份一致，并适应嵌入式计算预算。聚合跟踪分数对轨迹失败的时间和方式提供的洞察有限，而不同的检测器输入可能会混淆跟踪器和检测器的质量。我们提出了一种以部署为中心的仅跟踪器评估协议，该协议利用共享检测来隔离跟踪器行为，并直接评估初始化、检测器间隙延续、身份恢复、近邻关联以及负载依赖的跟踪器步骤运行时，同时保留高阶跟踪准确度（Higher Order Tracking Accuracy, HOTA）作为补充的聚合度量。我们将该协议应用于JackRabbot数据集和基准（JRDB），使用六个开源跟踪器和我们的轻量级行人参考跟踪器（Pedestrian Reference Tracker, PedRefTrack），以及一个基于GT（Ground Truth）辅助的变体，该变体在理想化关联和运动下估计剩余的跟踪器侧间隙。在固定检测下，非GT跟踪器的HOTA仅为24.26%-29.67%，但表现出显著不同的能力特征。在没有检测器支持的情况下，经过1.0秒后，任何没有GT辅助的跟踪器在超过一半的合格案例中都无法保持空间正确的同一身份输出，使得缺失观测的延续成为测试属性中主要的限制。近邻失败较小，主要在最短的分离中增加。在NVIDIA Jetson Orin上的跟踪器步骤运行时呈重尾分布且对负载敏感，导致多个跟踪器在拥挤帧中低于10 Hz实时目标。该协议提供了一种可重复的方法来表征跟踪器行为和在行人中心环境中的部署适宜性。代码和评估脚本已发布在 https://github.com/SCAI-Lab/tracker_eval。
cs.RO / 64 / 2610.01698
ActiveWAM: Evidence-Aware Active Vision for World-Action Models
ActiveWAM：面向世界-动作模型的证据感知主动视觉
Abstract
Active vision manipulation requires a policy to control both its camera and its end-effectors, yet camera motion determines which evidence remains visible within finite observation windows. Acquiring a new view can displace task-critical cues, while retaining a view forgoes potentially useful observations. We formulate this as an evidence-aware retain--acquire problem and present ActiveWAM, a unified world--action model that learns observation and manipulation jointly. To this end, we propose training-time inversion which constrains a frozen video prior by task-bearing source evidence and visible temporal changes, eliminating the need for test-time inversion or candidate ranking. At deployment, the policy generates bimanual and pan/tilt actions-including stay and reacquisition behaviors-from view-aware history, and updates context from newly measured RGB observations. Future-video prediction serves as a co-training signal, while action generation requires neither future-video decoding nor optimal viewpoint annotations. We introduce RoboTwin-AV, a 50-task benchmark with executable pan/tilt control and automatically generated demonstrations. ActiveWAM improves TAVIS out-of-distribution success by up to 17.0 percentage points over the strongest baselines, achieves 20.0 additional points over Fast-WAM on RoboTwin-AV, and outperforms it by 26.7 points on real-world physical kitchen tasks.
Chinese Translation
主动视觉操控需要一个策略来同时控制其相机和末端执行器，然而相机运动决定了在有限观察窗口内哪些证据保持可见。获取新视角可能会移位任务关键线索，而保持视角则放弃了潜在有用的观察。我们将此问题表述为一个证据感知的保留-获取问题，并提出了ActiveWAM，这是一种统一的世界-动作模型，能够联合学习观察和操控。为此，我们提出了训练时反演，该方法通过任务相关的源证据和可见的时间变化来约束冻结的视频先验，消除了测试时反演或候选排名的需求。在部署时，该策略从视角感知历史中生成双手和摇摆/倾斜的动作，包括保持和重新获取行为，并根据新测量的RGB观察更新上下文。未来视频预测作为共同训练信号，而动作生成则不需要未来视频解码或最佳视点注释。我们引入了RoboTwin-AV，这是一个包含50个任务的基准，具有可执行的摇摆/倾斜控制和自动生成的演示。ActiveWAM在TAVIS的分布外成功率上比最强基线提高了多达17.0个百分点，在RoboTwin-AV上比Fast-WAM多获得20.0分，并在实际物理厨房任务中超越了Fast-WAM 26.7分。
cs.RO / 65 / 2610.01726
Query-Conditioned Articulation Estimation from a Single Image
基于查询条件的单幅图像关节估计
Abstract
Enabling robots to estimate the kinematic parameters of articulated objects unlocks a wide range of capabilities for interaction and manipulation. The estimation has to happen from the information the robot currently observes, often just a single RGB image of an object it has never seen before. Current single-image approaches couple articulation part segmentation with articulation estimation, making their predictions vulnerable to missed detections and incorrect part associations, and they regress metric 3D geometry that a single view fixes only up to scale. We present QueryArt, a model that estimates articulation parameters from a single RGB image, a 2D query point, and camera intrinsics. QueryArt is trained to estimate the 3D articulation geometry relative to the queried point and in units of its depth, which keeps its target identifiable from the image alone. A single depth measurement at the query point then supplies the scale and recovers the metric parameters. We train QueryArt on a curated mixture of synthetic and real-world articulation datasets. We evaluate QueryArt on several benchmarks and compare it against recent baselines. QueryArt outperforms recent baselines on most articulation metrics, including on out-of-distribution data. To demonstrate the model's capabilities in real-world settings, we evaluate QueryArt on a mobile manipulator across 57 manipulation trials spanning 16 object parts and five viewpoint classes, achieving a 70.2% success rate. We provide code and videos at: https://abwerby.github.io/queryart/
Chinese Translation
使机器人能够估计关节物体的运动学参数为交互和操作解锁了广泛的能力。这一估计必须基于机器人当前观察到的信息，通常仅为一幅它从未见过的物体的RGB图像。目前的单幅图像方法将关节部分分割与关节估计相结合，使得它们的预测容易受到漏检和错误部分关联的影响，并且它们回归的度量3D几何形状仅在尺度上受到单一视角的限制。我们提出了QueryArt，这是一种从单幅RGB图像、一个2D查询点和相机内参中估计关节参数的模型。QueryArt被训练为相对于查询点估计3D关节几何形状，并以其深度单位进行计算，这使得其目标在图像中可识别。然后，在查询点的单一深度测量提供尺度并恢复度量参数。我们在经过精心挑选的合成和真实世界关节数据集的混合上训练QueryArt。我们在多个基准上评估QueryArt，并与最近的基线进行比较。QueryArt在大多数关节度量上优于最近的基线，包括在分布外数据上的表现。为了展示模型在真实世界环境中的能力，我们在一个移动操纵器上评估QueryArt，进行了57次操作试验，涵盖了16个物体部件和五个视角类别，成功率达到70.2%。我们提供代码和视频，网址为：https://abwerby.github.io/queryart/
cs.RO / 66 / 2610.01742
World Motion Models: Flexible Sequence Modeling of SE(3) Trajectories
世界运动模型：SE(3)轨迹的灵活序列建模
Abstract
Equipping artificial agents with spatial intelligence requires a comprehensive generative prior over the dynamic 3D world. We propose World Motion Models (WMMs) that capture "what was, is, and will be where across time" via sparse SE(3) pose trajectories. WMMs are built on the observation that elements of dynamic scenes can be well approximated by a set of rigid SE(3) trajectories, a minimal yet expressive primitive for 4D modeling. This representation unifies articulated objects, human bodies, hand-object interactions, piecewise-rigid scene dynamics, camera motion, and even robot states and actions into a single shared space. Given this representation, we cast the joint distribution of these entities as a flexible sequence modeling problem, utilizing flow-matching with per-token noise levels. Coupled with a context token mechanism for non-sequential conditioning, this formulation supports any-to-any marginal conditioning across an arbitrary number of entities and time steps. Tasks such as future prediction, motion infilling, model-predictive control, inverse kinematics, cross-embodiment retargeting, and policy learning all reduce to the application of different masks over the same network. Experiments on 6 diverse applications of 3D vision and robotics demonstrate the versatility and flexibility of WMMs with strong performance.
Chinese Translation
为人工智能体赋予空间智能需要对动态三维世界进行全面的生成性先验建模。我们提出了世界运动模型（World Motion Models, WMMs），通过稀疏的SE(3)姿态轨迹捕捉“过去、现在和未来在时间上的位置”。WMMs的构建基于这样一个观察：动态场景的元素可以通过一组刚性SE(3)轨迹进行良好近似，这是一种用于四维建模的最小但富有表现力的原语。这种表示将关节物体、人类身体、手-物体交互、分段刚性场景动态、相机运动，甚至机器人状态和动作统一到一个共享空间中。基于这种表示，我们将这些实体的联合分布视为一个灵活的序列建模问题，利用每个标记的噪声水平进行流匹配。结合上下文标记机制以实现非顺序条件，这种公式支持在任意数量的实体和时间步长之间的任意边际条件。未来预测、运动填充、模型预测控制、逆运动学、跨体现重定向和策略学习等任务均可简化为在同一网络上应用不同的掩码。对六个不同的三维视觉和机器人应用的实验展示了WMMs的多样性和灵活性，并取得了良好的性能。
cs.RO / 67 / 2610.01746
End-to-End Learning vs. Modular Architectures: Comparative Insights into Autonomous Driving Systems
端到端学习与模块化架构：对自主驾驶系统的比较洞察
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.
Chinese Translation
自主驾驶系统已成为智能交通研究的核心焦点，端到端学习（End-to-End Learning）和模块化架构（Modular Architectures）提供了两种突出的设计范式用于其实现。端到端学习利用深度学习算法将原始传感器输入直接映射到驾驶执行器，提供了一种简化且适应性强的解决方案，而模块化架构则采用基于管道的方法，将系统划分为感知、认知、规划和控制等不同子系统。本文对这两种范式进行了全面的比较分析，重点关注它们的优点、局限性和权衡，以提供对其在各种自主驾驶应用中的适用性的洞察。研究评估了可解释性、可扩展性、鲁棒性和现实应用等关键因素。尽管端到端学习强调在动态环境中的简单性和适应性，但其缺乏透明性且高度依赖于大规模数据集。相反，模块化架构提供了更好的可解释性和特定任务优化，但面临与集成复杂性和可扩展性相关的挑战。为了解决这些局限性，结合两种范式优势的混合方法应运而生，提供了克服这些挑战的有希望的方向。除了这种比较综合，后续工作提出了一个四维架构选择框架，包含跨安全性、操作环境、数据/计算资源和部署背景的十二个二元标准，并在十个已发布的自主驾驶系统中进行了验证，正确推荐了7/10个已部署架构。该研究综合现有文献，强调了这两种范式之间的关键权衡，并确定混合架构作为未来研究的有希望方向。
cs.RO / 68 / 2610.01849
FlashDexRetarget: Accelerating Dexterous Manipulation Data Generation through Multi-Motion Retargeting
FlashDexRetarget：通过多运动重定向加速灵巧操作数据生成
Abstract
Human hand-object demonstrations offer a reusable source of dexterous robot manipulation data, but transferring them across embodiments requires physically feasible retargeting. Existing physics-based approaches face limitations in retargeting success, motion-specific training efficiency, or both. To address these limitations, we introduce FlashDexRetarget, an RL-based framework for high-success, efficient dexterous motion retargeting. To make the demonstrated interaction easier to learn, we combine object point-cloud observations, hand-object distance features, and future trajectory encodings with complementary rewards that supervise object motion and reference hand-object relationships. To further accelerate learning, we employ separate left- and right-hand actor critic networks and adapt the off-policy algorithm, FlashSAC to dexterous motion tracking. On a benchmark of 50 motions spanning single-object and two-object interactions, FlashDexRetarget achieves a 90% success rate, approximately 2.5x that of the evaluated sampling-based baselines, while requiring up to 100x less training compute than the evaluated RL-based baselines. Evaluations on both XHand and Sharpa Wave Hand show consistent gains, and component-wise ablations examine the contributions of our design choices. Beyond the 50-motion benchmark, experiments with 200, 500, and 1,000 motions demonstrate that our method remains stable at larger scales and produces successful retargeted motions more efficiently as the training set grows. Qualitative replay results using real-world-captured demonstrations further illustrate the applicability of our framework to recorded human manipulation. Videos and code are available at https://davian-robotics.github.io/FlashDexRetarget/
Chinese Translation
人类手部与物体的演示提供了一种可重用的灵巧机器人操作数据来源，但在不同形态之间转移这些演示需要物理上可行的重定向。现有的基于物理的方法在重定向成功率、运动特定训练效率或两者方面都存在局限。为了解决这些问题，我们提出了FlashDexRetarget，这是一个基于强化学习（RL）的框架，旨在实现高成功率和高效的灵巧运动重定向。为了使演示的交互更易于学习，我们结合了物体点云观察、手-物体距离特征和未来轨迹编码，并使用互补奖励来监督物体运动和参考手-物体关系。为了进一步加速学习，我们采用了独立的左右手演员评论网络，并将离线策略算法FlashSAC适应于灵巧运动跟踪。在涵盖单物体和双物体交互的50种运动基准测试中，FlashDexRetarget实现了90%的成功率，约为评估的基于采样的基线的2.5倍，同时所需的训练计算量比评估的基于RL的基线少达100倍。在XHand和Sharpa Wave Hand上的评估显示出一致的提升，组件消融实验考察了我们设计选择的贡献。超出50种运动基准的实验表明，在200、500和1000种运动的情况下，我们的方法在更大规模下仍然稳定，并且随着训练集的增长，能够更高效地生成成功的重定向运动。使用真实世界捕获的演示的定性重放结果进一步说明了我们框架在记录的人类操作中的适用性。视频和代码可在https://davian-robotics.github.io/FlashDexRetarget/获取。
cs.RO / 69 / 2610.01856
ChunkVLA-AM: Parallel Action Chunking for Vision-Language-Action Robot Control in Additive Manufacturing
ChunkVLA-AM：增材制造中视觉-语言-动作机器人控制的并行动作分块
Abstract
Vision-language-action (VLA) models unify visual perception, language understanding, and action generation, offering new opportunities for automation in additive manufacturing (AM). However, deployment in AM remains challenging because adapting these models to unseen robot embodiments is costly, and performance can degrade under environment changes. In this work, we present a framework for deploying OpenVLA-OFT on a FAIRINO FR3 robot in a fixed AM workcell. A data pipeline converts monocular real-world demonstrations into OpenVLA-compatible TFDS/RLDS datasets to support adaptation to the FR3 embodiment. At runtime, each inference request predicts an eight-step chunk of 7-D actions. The FR3 executes each chunk open loop before capturing a new observation, providing closed-loop feedback between chunks. The system uses a cloud-edge architecture in which the FR3 client streams observations to a remote inference server through a FastAPI interface. In 42 physical A-to-B object-transfer trials, evenly split between red and blue targets, the system succeeded in 39 (92.9%). All three failures occurred during final placement, when insufficient release-height control caused the object to topple. An illumination sweep identified a low-error luminance range of 85-125 on a 0-255 scale, with the lowest mean spatial error at 95.
Chinese Translation
视觉-语言-动作（VLA）模型统一了视觉感知、语言理解和动作生成，为增材制造（AM）中的自动化提供了新的机会。然而，由于将这些模型适应于未见过的机器人形态成本高昂，并且在环境变化下性能可能下降，因此在AM中的部署仍然具有挑战性。在本研究中，我们提出了一个在固定AM工作单元中将OpenVLA-OFT部署到FAIRINO FR3机器人的框架。数据管道将单目真实世界演示转换为与OpenVLA兼容的TFDS/RLDS数据集，以支持对FR3形态的适应。在运行时，每个推理请求预测一个包含7维动作的八步分块。FR3在捕获新观察之前以开放循环执行每个分块，提供分块之间的闭环反馈。该系统采用云-边缘架构，其中FR3客户端通过FastAPI接口将观察结果流式传输到远程推理服务器。在42个物理A到B物体转移试验中，红色和蓝色目标均匀分配，系统成功率为39次（92.9%）。所有三次失败发生在最终放置阶段，因释放高度控制不足导致物体倾倒。照明扫描确定了85-125的低误差亮度范围（在0-255的尺度上），在95时具有最低的平均空间误差。
cs.RO / 70 / 2610.01906
Towards Physical Underwater Robotic Assistance for Scuba Diver Movement in Confined Spaces
面向水下机器人对潜水员在狭小空间内运动的物理辅助
Abstract
Scuba divers are taught to control their depth to avoid rapid ascents and descents, which could result in serious injuries such as gas embolisms and barotrauma. However, many underwater tasks necessitate lateral control, maintaining distance between subsea structures such as coral reefs, submerged drilling instrumentation, or unexploded ordnance. In this work, we discuss a first-of-its-kind wearable robotic solution providing thruster-actuated directional guidance to a diver, as distinct from prior propulsive-assistance exoskeletons. We introduce ``Robotic Assisted Diver Movement in Confined Spaces'' (RADMCS), a wearable robot that assists divers in maintaining a fixed distance from subsea structures by leveraging perception techniques in monocular depth estimation and force-feedback from submersible thrusters to provide haptic feedback. Its small and compact form factor creates a foundational platform that could be expanded to include more sophisticated control and navigation behaviors. We present results from Institutional Review Board (IRB) in-water studies with eight human scuba diver participants on threshold sensitivity tests in both a closed-water swimming facility and ocean environments; distance-maintaining experiments in a closed-water facility; and form, fit, and function testing in the ocean. We demonstrate that relatively low thrust values (10 percent of maximum) allow robotic direction of a human's movement using the physical sensation of the robot's guidance.
Chinese Translation
潜水员被教导控制其深度，以避免快速上升和下降，这可能导致严重的伤害，如气体栓塞和气压损伤。然而，许多水下任务需要侧向控制，以保持与水下结构（如珊瑚礁、沉没的钻探设备或未爆炸弹药）之间的距离。在本研究中，我们讨论了一种首创的可穿戴机器人解决方案，为潜水员提供推进器驱动的方向引导，这与之前的推进辅助外骨骼不同。我们介绍了“狭小空间中的机器人辅助潜水员运动”（RADMCS），这是一种可穿戴机器人，通过利用单目深度估计的感知技术和来自水下推进器的力反馈，提供触觉反馈，帮助潜水员保持与水下结构的固定距离。其小巧紧凑的形态为扩展更复杂的控制和导航行为奠定了基础。我们展示了来自机构审查委员会（IRB）在水中进行的研究结果，参与者为八名人类潜水员，研究内容包括在封闭水域游泳设施和海洋环境中的阈值敏感性测试；在封闭水域设施中的距离保持实验；以及在海洋中的形式、适配性和功能测试。我们证明，相对较低的推力值（最大推力的10%）可以利用机器人的引导物理感受来引导人类的运动。
cs.RO / 71 / 2610.01910
Robot Learning on Discrete Surfaces: Theory and Applications
离散表面上的机器人学习：理论与应用
Abstract
All the objects composing our world are enclosed within surfaces. Yet, most robot learning and motion generation frameworks treat surfaces as constraints ignoring their intrinsic geometry. This gap is acute for polyhedral meshes--the standard output of CAD and 3D reconstruction--whose discrete geometric structure remains unexploited. In this paper, we propose a unified discrete Riemannian framework that enables robot learning directly on polyhedral surface meshes. Using discrete differential geometry, we define logarithmic and exponential maps, parallel transport, and ambient-space projections that remain well-defined across faces, edges, and vertices. We instantiate the framework in three learning paradigms: (i) Dynamic Movement Primitives (DMPs), an improved exponential-map computation and a fixed-tangent-cone forcing-term encoding with parallel transport yield better cross-surface generalisation and stability over prior mesh-based approaches. (ii) Gaussian Process (GP), a geodesic-based kernel with practical admissibility control, enables regression at arbitrary mesh locations without smoothness assumptions. (iii) Riemannian Flow Matching (RFM), mesh-native operators improve generative quality over spectral baselines while reducing training time. The framework is validated in simulation against state-of-the-art methods and demonstrated on two real-robot scenarios: generalising user-drawn trajectories across different surfaces and planning polishing motions on RGB-D-reconstructed surfaces.
Chinese Translation
构成我们世界的所有物体都被包围在表面之内。然而，大多数机器人学习和运动生成框架将表面视为约束，忽视了它们的内在几何特性。这一差距在多面体网格中尤为明显——这是计算机辅助设计（CAD）和三维重建的标准输出，其离散几何结构尚未得到充分利用。本文提出了一种统一的离散黎曼框架，使机器人能够直接在多面体表面网格上进行学习。利用离散微分几何，我们定义了对数映射和指数映射、平行传输以及在面、边和顶点之间保持良好定义的环境空间投影。我们在三种学习范式中实例化该框架：(i) 动态运动原语（Dynamic Movement Primitives, DMPs），改进的指数映射计算和固定切向锥强迫项编码结合平行传输，提供了比以前基于网格的方法更好的跨表面泛化能力和稳定性。(ii) 高斯过程（Gaussian Process, GP），一种基于测地线的核函数，具有实用的可接受性控制，使得在任意网格位置进行回归而不需要平滑性假设成为可能。(iii) 黎曼流匹配（Riemannian Flow Matching, RFM），网格本地操作符提高了生成质量，相较于谱基线减少了训练时间。该框架在仿真中与最先进的方法进行了验证，并在两个真实机器人场景中进行了演示：跨不同表面泛化用户绘制的轨迹以及在RGB-D重建表面上规划抛光运动。
cs.RO / 72 / 2610.01943
TouchTherm: Building Multimodal Digital Twins of Objects for Tactile and Thermal Rendering
TouchTherm：构建用于触觉和热渲染的多模态数字双胞胎对象
Abstract
Robotic simulation and virtual reality increasingly require object assets that capture not only visual geometry but also the physical cues underlying tactile and thermal interaction. Existing 3D datasets and reconstruction methods primarily represent object-scale geometry and visual appearance, overlooking microscale surface structure for high-fidelity haptic rendering and transient temperature dynamics for temperature-aware interaction. We present TouchTherm, a framework for constructing simulation-ready visuo-tactile-thermal object assets from real-world objects. For visual and tactile reconstruction, we combine structured-light scanning with multiview normal maps obtained from photometric stereo. The normal maps are registered to the scanned geometry and transformed into tangent space to recover local micro-height fields for optical tactile rendering, while the coarse mesh handles collision detection. For thermal reconstruction, we capture synchronized multiview infrared videos of natural cooling following controlled heating and reconstruct a physics-regularized dynamic thermal field. Experiments on 20 objects show that the reconstructed micro-height fields preserve dominant surface structures and recover higher-frequency details beyond the coarse geometry, while the thermal fields achieve held-out surface-temperature MAEs of 0.465 degrees C and 0.592 degrees C at 30 s and 45 s, respectively. The resulting tactile assets support synthetic-to-real object recognition from tactile observations, while a glove-based VR system demonstrates spatially and temporally varying thermal feedback. These results highlight the potential of TouchTherm for multimodal sensory simulation and temperature-aware virtual interaction.
Chinese Translation
机器人仿真和虚拟现实日益需要不仅捕捉视觉几何形状，还要捕捉触觉和热交互背后的物理线索的对象资产。现有的3D数据集和重建方法主要代表对象尺度的几何形状和视觉外观，忽视了高保真触觉渲染所需的微观表面结构以及温度感知交互所需的瞬态温度动态。我们提出了TouchTherm，一个从现实世界对象构建适用于仿真的视觉-触觉-热对象资产的框架。对于视觉和触觉重建，我们结合了结构光扫描和从光度立体法获得的多视角法线图。法线图被注册到扫描的几何形状上，并转换到切空间，以恢复用于光学触觉渲染的局部微高度场，而粗糙网格则处理碰撞检测。对于热重建，我们捕捉了自然冷却过程中的同步多视角红外视频，经过控制加热后重建出物理正则化的动态热场。在20个对象上的实验表明，重建的微高度场保留了主要的表面结构，并恢复了超出粗糙几何形状的高频细节，而热场在30秒和45秒时分别实现了0.465摄氏度和0.592摄氏度的表面温度平均绝对误差（MAE）。生成的触觉资产支持从触觉观察中进行合成到真实对象的识别，而基于手套的虚拟现实系统则展示了空间和时间上变化的热反馈。这些结果突显了TouchTherm在多模态感官仿真和温度感知虚拟交互中的潜力。
cs.RO / 73 / 2610.01959
Training-Free Diffusion Planning with Analytical Local Scores
无训练扩散规划与解析局部评分
Abstract
Path finding and multi-robot motion planning require trajectories that are smooth, goal-directed, and collision-free in environments with complex geometric constraints. Recent diffusion-based planners have shown that trajectory generation can be cast as iterative denoising which has opened the doors to learning-based approaches that can handle multi-modal trajectory distributions and refine entire trajectories. However, a key limitation is that diffusion planners require training on large collections of feasible trajectories, rendering them map-specific, and difficult to deploy when high-quality demonstrations are unavailable. This paper introduces a training-free diffusion-based motion planner that replaces learned global trajectory scores with analytical local scores derived from obstacle, smoothness, velocity, and inter-agent feasibility terms. The proposed idea relies on a key observation: the score of a trajectory can be reconstructed by considering only local interactions between neighboring waypoints and nearby constraints. This structure exploitation yields a decomposed denoising procedure that retains the optimization structure of classical trajectory methods while inheriting the iterative refinement behavior of diffusion models. Experiments on a large collection of complex environments and large multi-agent planning tasks show that the proposed analytical score produces smooth and feasible trajectories within limited computational costs, for example in generating feasible paths for 300+ agents in environments containing 100+ obstacles in under 6 seconds on a GPU, outperforming strong learning-based and optimization baselines, while avoiding the data requirements of learned diffusion planners.
Chinese Translation
路径寻找和多机器人运动规划需要在具有复杂几何约束的环境中生成平滑、目标导向且无碰撞的轨迹。最近的基于扩散的规划器表明，轨迹生成可以被视为迭代去噪，这为能够处理多模态轨迹分布并优化整个轨迹的学习方法打开了大门。然而，一个关键的限制是扩散规划器需要在大量可行轨迹上进行训练，这使得它们特定于地图，并且在高质量演示不可用时难以部署。本文提出了一种无训练的基于扩散的运动规划器，该规划器用从障碍物、平滑度、速度和代理间可行性项导出的解析局部评分替代了学习的全局轨迹评分。所提出的理念依赖于一个关键观察：轨迹的评分可以通过仅考虑相邻航点和附近约束之间的局部交互来重建。这种结构利用产生了一种分解的去噪过程，保留了经典轨迹方法的优化结构，同时继承了扩散模型的迭代优化行为。在大量复杂环境和大型多代理规划任务上的实验表明，所提出的解析评分在有限的计算成本下生成平滑且可行的轨迹，例如在包含100多个障碍物的环境中为300多个代理生成可行路径，且在6秒内完成，优于强大的基于学习和优化的基线，同时避免了学习型扩散规划器的数据需求。
cs.RO / 74 / 2610.01972
BLT*: Informed Belief Localization Trees for Uncertainty-Aware Planning on Digital Twins
BLT*: 面向不确定性规划的知情信念定位树在数字双胞胎中的应用
Abstract
We present Informed Belief Localization Trees* (Informed BLT*), a sampling-based belief space planning (BSP) algorithm that scales to large outdoor digital twins with point-cloud observations. We adapt RRT* and Informed RRT* to belief space using the $2$-Wasserstein ($W_2$) metric. Assuming isotropic Gaussian beliefs, sampled belief states can be connected efficiently while accounting for available information and probabilistic collision constraints. This enables steering and rewiring without repeatedly propagating observations, and allows previously computed measurement information to be reused. We present a framework to generate semantically labelled digital twins for planning in real-world environments with point-cloud-based localization. Experiments in simulated environments and digital twins show faster initial solution discovery in most maps with competitive cost convergence.
Chinese Translation
我们提出了知情信念定位树*（Informed BLT*），这是一种基于采样的信念空间规划（BSP）算法，能够扩展到具有点云观测的大型户外数字双胞胎。我们将 RRT* 和 Informed RRT* 适配到信念空间，使用 $2$-Wasserstein ($W_2$) 度量。假设信念为各向同性高斯分布，采样的信念状态可以高效连接，同时考虑可用信息和概率碰撞约束。这使得在不重复传播观测的情况下进行引导和重连成为可能，并允许重用先前计算的测量信息。我们提出了一个框架，用于生成语义标注的数字双胞胎，以便在基于点云的定位的实际环境中进行规划。在模拟环境和数字双胞胎中的实验表明，在大多数地图中，初始解决方案的发现速度更快，并且成本收敛具有竞争力。
cs.RO / 75 / 2610.01985
H-SPAR: Hydrodynamic-aware Simulation for Particle Transport and Autonomous Robots
H-SPAR：考虑水动力的粒子运输与自主机器人仿真
Abstract
Environmental robotic sampling requires considering the dual influence of water currents on robotic motion and particle transport. Existing marine robotics simulators generally model flow, autonomy, and sampling targets separately, limiting joint evaluation of mission cost and sampling performance. H-SPAR integrates spatially and temporally varying velocity fields, Lagrangian particle transport, probabilistic sampling, and ROS 2/Gazebo-based uncrewed surface vehicle (USV) autonomy. In this work, shared precomputed flow fields drive particle advection and current-induced forces during closed-loop vehicle execution. Path-planning experiments show that the existing current-aware planner SVF-RRT* achieves 69.4% lower upstream cost than conventional RRT* at the planning level, but this reduction falls to 41.7% during execution under time-varying currents, reflecting temporal flow variation, vehicle motion constraints, and path deviation omitted during planning. Coverage experiments show that sweep orientation changes the particle-sampling rate by up to 22.2% under the complete H-SPAR configuration. These findings highlight the importance of evaluating planning, vehicle execution, particle transport, and sampling together under consistent hydrodynamic conditions. The project webpage is available at https://sites.google.com/view/h-spar, and the open-source code is available on GitHub at https://github.com/naviiidz/h-spar-sim.
Chinese Translation
环境机器人采样需要考虑水流对机器人运动和粒子运输的双重影响。现有的海洋机器人仿真器通常将流动、自主性和采样目标分开建模，这限制了任务成本与采样性能的联合评估。H-SPAR集成了时空变化的速度场、拉格朗日粒子运输、概率采样以及基于ROS 2/Gazebo的无人水面车辆（USV）自主性。在本研究中，共享的预计算流场驱动粒子对流和在闭环车辆执行过程中产生的流动诱导力。路径规划实验表明，现有的考虑水流的规划器SVF-RRT*在规划阶段的上游成本比传统的RRT*降低了69.4%，但在时间变化的水流下执行时这一降低幅度降至41.7%，反映了规划过程中忽略的时间流动变化、车辆运动约束和路径偏离。覆盖实验表明，在完整的H-SPAR配置下，扫掠方向使粒子采样率变化高达22.2%。这些发现强调了在一致的水动力条件下共同评估规划、车辆执行、粒子运输和采样的重要性。项目网页可访问 https://sites.google.com/view/h-spar，开源代码可在GitHub上获取 https://github.com/naviiidz/h-spar-sim。
cs.RO / 76 / 2610.02054
UniWAM: Unified World-Action Model
UniWAM：统一世界-动作模型
Abstract
Vision-language-action models benefit from the understanding and reasoning capabilities of pretrained vision-language models, but action-only supervision provides limited grounding in world dynamics. Conversely, world-action models inherit spatiotemporal priors from video generation models, yet remain limited in semantic understanding and reasoning under distribution shifts. We introduce UniWAM, a unified architecture that integrates a physical reasoner, a world generator, and an action predictor to jointly learn semantic understanding of the physical world, visual generation, and action prediction. To ensure the quality of the training data, we developed a rigorous data cleaning and annotation pipeline for both human egocentric data and robot data. To adapt the vision-language component to embodied tasks while preserving its inherited language capabilities, we represent low-level actions in natural language and introduce a pre-training recipe that assigns complementary supervision from visual question answering (VQA) data, human egocentric data, and robot demonstrations to the appropriate model components. During post-training, future visual noise augmentation reduces reliance on precise future predictions, while history-conditioned flow matching uses encoded action history to initialize action generation. Together, these designs significantly reduce denoising steps while maintaining performance. UniWAM achieves state-of-the-art (SOTA) performance across multiple evaluations, including in-distribution performance, robustness, generalization, instruction following, and long-horizon task execution. Furthermore, we uncover a log-linear scaling law of unified human-robot co-training, demonstrating the effectiveness of large-scale pre-training on a mixture of human and robot data.
Chinese Translation
视觉-语言-动作模型受益于预训练视觉-语言模型的理解和推理能力，但仅依赖动作监督的方式在世界动态方面提供的基础有限。相反，世界-动作模型继承了视频生成模型的时空先验，但在语义理解和分布变化下的推理能力上仍然有限。我们提出了UniWAM，一种统一架构，集成了物理推理器、世界生成器和动作预测器，以共同学习物理世界的语义理解、视觉生成和动作预测。为了确保训练数据的质量，我们开发了一套严格的数据清理和注释流程，适用于人类自我中心数据和机器人数据。为了将视觉-语言组件适应于具身任务，同时保留其继承的语言能力，我们以自然语言表示低级动作，并引入了一种预训练方案，该方案从视觉问答（VQA）数据、人类自我中心数据和机器人演示中为适当的模型组件分配互补监督。在后期训练中，未来视觉噪声增强减少了对精确未来预测的依赖，而历史条件流匹配则利用编码的动作历史来初始化动作生成。这些设计显著减少了去噪步骤，同时保持了性能。UniWAM在多个评估中实现了最先进的（SOTA）性能，包括分布内性能、鲁棒性、泛化、指令跟随和长时间任务执行。此外，我们揭示了统一人-机器人共同训练的对数线性缩放法则，展示了在混合人类和机器人数据上进行大规模预训练的有效性。
cs.RO / 77 / 2610.02089
HumanoidToolBench: Benchmarking Humanoid Tool Use from Selection to Mobile Execution
HumanoidToolBench：从选择到移动执行的人形工具使用基准测试
Abstract
As robotic hardware and learning methods advance, humanoids need tools to perform tasks beyond their inherent physical limits. Successful tool use requires selecting a suitable tool and coordinating manipulation and, when needed, locomotion to complete the task. Existing benchmarks do not jointly evaluate these capabilities on a humanoid. We introduce HumanoidToolBench, an 18-task benchmark spanning three scenarios, three execution levels, and two tool-set modes, together with ToolBook, a dataset of 3.1k demonstrations collected in simulation and on a real Unitree G1. Evaluation of seven policies in simulation and three on the real robot reveals substantial gaps between selecting a suitable tool and completing the task. Focused GR00T N1.7 probes show reduced selection accuracy on unseen tools and continued task execution under unrelated instructions. Code and data are available at https://snu-pi.github.io/HumanoidToolBench/.
Chinese Translation
随着机器人硬件和学习方法的进步，人形机器人需要工具来执行超出其固有物理限制的任务。成功的工具使用需要选择合适的工具，并协调操作和（在需要时）运动以完成任务。现有的基准测试并未在一个人形机器人上共同评估这些能力。我们介绍了HumanoidToolBench，这是一个涵盖三个场景、三个执行级别和两种工具集模式的18项任务基准测试，以及ToolBook，一个在仿真和真实Unitree G1上收集的3.1k演示的数据集。在仿真中评估七种策略和在真实机器人上评估三种策略的结果显示，选择合适工具与完成任务之间存在显著差距。针对GR00T N1.7的重点探测显示，在未见过的工具上选择准确性降低，并且在不相关指令下继续执行任务。代码和数据可在 https://snu-pi.github.io/HumanoidToolBench/ 获取。
cs.RO / 78 / 2610.02110
GlassGuard: Verified Glass Plane Mapping for Robot Navigation
GlassGuard：用于机器人导航的经过验证的玻璃平面映射
Abstract
Transparent and specular surfaces pose a serious challenge to LiDAR-based SLAM and navigation because laser returns may pass through glass, leaving collision boundaries absent from the map. Prior work attempts to reconstruct the missing surfaces, but inaccurate obstacle placement can create the opposite failure: contamination of traversable free space. Recognizing this dual requirement, we present GlassGuard, a navigation-oriented framework for reconstructing planar architectural glass from complementary visual and LiDAR evidence. We formulate success in terms of both glass coverage and free-space contamination and apply this principle throughout proposal verification and global map construction. A foundation vision model provides glass-instance masks, structural 3D cues generate metric plane hypotheses, and depth-free 2D projective geometry checks their orientations before they enter a consolidated global map. We evaluate GlassGuard in nine building-scale scenes spanning diverse glass structures, spatial scales, and lighting conditions, with more than one hour and 2.1 km of real-world robot traversal. GlassGuard achieves 85% of total glass coverage for its panoramic version. Under identical pinhole inputs, GlassGuard achieves 82% total coverage, compared with at most 61% for the evaluated baselines, while producing 5-17x fewer false voxels per frame. Qualitative examples with a navigation planner illustrate the reconstructed planes blocking paths through glass while leaving traversable routes open. The project page is available at https://glassguardproject.github.io/.
Chinese Translation
透明和镜面表面对基于激光雷达的同步定位与地图构建（SLAM）和导航构成了严重挑战，因为激光返回可能穿过玻璃，导致地图上缺少碰撞边界。之前的研究尝试重建缺失的表面，但不准确的障碍物放置可能导致相反的失败：可通行自由空间的污染。认识到这一双重需求，我们提出了GlassGuard，一个面向导航的框架，用于从互补的视觉和激光雷达证据中重建平面建筑玻璃。我们将成功定义为玻璃覆盖率和自由空间污染的双重标准，并在提案验证和全局地图构建过程中应用这一原则。基础视觉模型提供玻璃实例掩膜，结构性三维线索生成度量平面假设，深度无关的二维投影几何在它们进入整合的全局地图之前检查其方向。我们在九个建筑规模的场景中评估GlassGuard，这些场景涵盖了多样的玻璃结构、空间尺度和光照条件，实际机器人行驶超过一小时和2.1公里。GlassGuard的全景版本实现了85%的总玻璃覆盖率。在相同的针孔输入下，GlassGuard实现了82%的总覆盖率，而评估的基线最多仅为61%，同时每帧产生的虚假体素减少了5-17倍。与导航规划器的定性示例表明，重建的平面阻挡了通过玻璃的路径，同时保留了可通行的路线。项目页面可访问：https://glassguardproject.github.io/
cs.RO / 79 / 2610.02120
SkeleWAM: Skeleton World-Action Modeling for Efficient Robotic Manipulation
SkeleWAM：用于高效机器人操控的骨架世界动作建模
Abstract
World action models (WAMs) combine robot action generation with future state prediction. Existing WAMs typically predict videos or learned visual latents, which represent interaction geometry only implicitly and may retain appearance information unrelated to control. We introduce SkeleWAM, a compact WAM that represents a manipulation scene as a sparse 3D skeleton composed of robot joints, object centers, and interaction points. Constructed online from current RGB-D observations and robot proprioception, the skeleton provides a unified geometric state for action generation and future skeleton prediction. Future skeleton prediction provides additional geometric supervision for action learning without requiring visual reconstruction. At inference, SkeleWAM generates actions directly from the current skeleton and language instruction, while Medoid Action Consensus (MAC) serves as an auxiliary consensus strategy for stochastic action samples. On LIBERO-Plus, SkeleWAM achieves an overall success rate of 85.9% with 57.1M parameters, outperforming Cosmos-Policy by 3.7 percentage points. These results demonstrate that sparse 3D robot--object structure provides an effective state space for robust and parameter-efficient world action learning. The project is available at https://skelewam-project.github.io/.
Chinese Translation
世界动作模型（WAMs）将机器人动作生成与未来状态预测相结合。现有的WAMs通常预测视频或学习到的视觉潜变量，这些潜变量仅隐式地表示交互几何，并可能保留与控制无关的外观信息。我们提出了SkeleWAM，一种紧凑的WAM，它将操控场景表示为由机器人关节、物体中心和交互点组成的稀疏3D骨架。该骨架在线构建，基于当前的RGB-D观察和机器人本体感知，为动作生成和未来骨架预测提供统一的几何状态。未来骨架预测为动作学习提供了额外的几何监督，而无需视觉重建。在推理阶段，SkeleWAM直接从当前骨架和语言指令生成动作，同时Medoid Action Consensus (MAC)作为随机动作样本的辅助共识策略。在LIBERO-Plus上，SkeleWAM以57.1M参数实现了85.9%的整体成功率，超越了Cosmos-Policy 3.7个百分点。这些结果表明，稀疏的3D机器人-物体结构为稳健且参数高效的世界动作学习提供了有效的状态空间。该项目可在https://skelewam-project.github.io/获取。
cs.RO / 80 / 2610.02161
DuoMind: Enabling Distributed Multi-Robot Coordination with Semantic Communication
DuoMind：通过语义通信实现分布式多机器人协调
Abstract
Vision-language models (VLMs) and vision-language-action models (VLAs) have recently driven rapid progress in general-purpose robots, yet most progress has focused on single-robot settings. Extending these capabilities to multi-robot systems remains challenging because robots must coordinate long-horizon behaviors while maintaining reliable, fine-grained execution. We introduce DuoMind, a distributed hierarchical framework for multi-robot coordination through semantic communication. Each robot uses a VLA-based action model for low-level execution and a VLM-based orchestrator for high-level reasoning and inter-agent coordination. At each planning step, the orchestrator at each robot reasons over the task instruction, local observations, and messages received from other robots. It then generates low-level instructions for the action model and semantic messages for peer robots. This architecture exploits the complementary strengths of pretrained models by combining the semantic reasoning capabilities of VLMs with the precise action-generation capabilities of VLAs. To address the scarcity of benchmarks for multi-robot coordination, we further develop RoboPoly, a benchmark comprising long-horizon manipulation tasks that require coordinated, closed-loop execution under distributed control. Experiments on RoboPoly and RoboTwin demonstrate that DuoMind improves multi-robot task performance, while ablation studies confirm the contributions of hierarchical orchestration and semantic communication. More details are available on our project page.
Chinese Translation
视觉语言模型（VLMs）和视觉语言行动模型（VLAs）最近推动了通用机器人领域的快速进展，但大多数进展集中在单机器人设置上。将这些能力扩展到多机器人系统仍然具有挑战性，因为机器人必须协调长时间跨度的行为，同时保持可靠的、细致的执行。我们提出了DuoMind，一个通过语义通信实现多机器人协调的分布式层次框架。每个机器人使用基于VLA的行动模型进行低级执行，并使用基于VLM的协调器进行高级推理和代理间协调。在每个规划步骤中，每个机器人的协调器会对任务指令、局部观察和来自其他机器人的消息进行推理。然后，它生成低级指令供行动模型使用，并为同伴机器人发送语义消息。该架构通过结合VLM的语义推理能力与VLA的精确行动生成能力，充分利用了预训练模型的互补优势。为了解决多机器人协调基准稀缺的问题，我们进一步开发了RoboPoly，这是一个包含需要在分布式控制下进行协调闭环执行的长时间跨度操作任务的基准。RoboPoly和RoboTwin上的实验表明，DuoMind提高了多机器人任务的性能，而消融研究确认了层次协调和语义通信的贡献。更多细节请参见我们的项目页面。
cs.RO / 81 / 2610.02170
Watch, Infer, Coordinate: Inferring Robot Partner Constraints for Zero-Shot Coordination
观察、推断、协调：推断机器人伙伴约束以实现零-shot 协调
Abstract
Robots operating in the physical world will increasingly need to coordinate with other robots, particularly in manipulation tasks where an object may be too large or heavy for a single robot to carry alone. Physical limitations caused by hardware degradation or actuator faults can restrict the actions a robot can reliably execute, yet these limitations may be unknown to its partner. We study whether a helper can infer a robot partner's physical constraints from observing it coordinate with another robot, then use the inferred capability to coordinate with the same partner on a new task. This is difficult because a demonstration shows what the constrained robot did, but not what it could have done. In physically coupled tasks, the other robot may also compensate for its limitations, making those limitations difficult to identify from the constrained robot's behavior alone. Our key insight is that these constraints shape the joint behavior of the team, making the actions of both robots informative about the constrained partner's capability. We introduce Watch, Infer, Coordinate, a benchmark spanning three physically coupled manipulation settings, together with an inference approach that scores candidate constraints using observed joint behavior. Across all three settings, our method substantially improves constraint inference and zero-shot coordination, approaching an oracle with access to the true constraints.
Chinese Translation
在物理世界中操作的机器人将越来越需要与其他机器人协调，特别是在操作任务中，当一个物体可能过大或过重，无法由单个机器人独自搬运时。由于硬件退化或执行器故障造成的物理限制可能会限制机器人可靠执行的动作，但这些限制可能对其伙伴而言是未知的。我们研究了一个助手是否可以通过观察一个机器人与另一个机器人协调的过程来推断其物理约束，然后利用推断出的能力与同一伙伴在新任务上进行协调。这一过程是困难的，因为演示展示了受限机器人所做的事情，但并未展示它本可以做的事情。在物理耦合任务中，另一个机器人也可能会补偿其限制，使得仅从受限机器人的行为中识别这些限制变得困难。我们的关键见解是，这些约束塑造了团队的联合行为，使得两个机器人的动作都能为受限伙伴的能力提供信息。我们提出了观察、推断、协调（Watch, Infer, Coordinate），这是一个涵盖三种物理耦合操作设置的基准，以及一种通过观察到的联合行为对候选约束进行评分的推断方法。在所有三种设置中，我们的方法显著改善了约束推断和零-shot 协调，接近于能够访问真实约束的神谕（oracle）。
cs.RO / 82 / 2610.02196
InterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-Manipulation
InterEvolve：类人运动操控的测试时奖励程序演化
Abstract
We study test-time evolution for humanoid loco-manipulation: solving tasks that a controller was never trained for by repurposing its existing skills, improving from its own attempts, and retaining what it learns, without retraining. Our key insight is that a broad controller already holds much of the competence a new task needs, and that this competence becomes accessible through an interface between planning and control that is expressive enough to specify contact-rich, multi-stage interactions, yet executable and measurable enough that execution feedback can guide planning from experience. InterEvolve realizes this interface with two components. First, we develop an object-aware forward-backward (FB) behavioral foundation model, whose object residuals on a frozen body prior turn a new reward about the body or objects into loco-manipulation behavior at test time. Second, we specify tasks as reward programs: staged rewards with completion conditions and tunable constants. A large language model (LLM) agent revises the program structure in context, drawing on execution feedback and a skill library of verified programs, while a numerical optimizer tunes its constants. With every candidate verified across parallel simulation scenarios, the program explores new ways to induce, repurpose, and compose the controller's existing motor competence for the task at hand, and thus improves over iterations. Experiments show that human-designed rewards leave much of the FB model's loco-manipulation competence untapped, whereas the programs InterEvolve evolves release it, sometimes through novel strategies. It further produces behaviors for diverse tasks, complex scenes, and long-horizon compositions in simulation, and evolved skills run autonomously on a physical Unitree G1 from egocentric onboard perception.
Chinese Translation
我们研究类人运动操控的测试时演化：通过重新利用现有技能、从自身尝试中改进并保留所学，无需重新训练，解决控制器从未训练过的任务。我们的关键见解是，广泛的控制器已经具备新任务所需的大部分能力，而这种能力可以通过一个规划与控制之间的接口获得，该接口足够表达以指定丰富接触的多阶段交互，同时又足够可执行和可测量，以便执行反馈可以从经验中指导规划。InterEvolve通过两个组件实现了这一接口。首先，我们开发了一种对象感知的前向-后向（FB）行为基础模型，其在冻结身体上的对象残差将关于身体或对象的新奖励转化为测试时的运动操控行为。其次，我们将任务指定为奖励程序：具有完成条件和可调常数的分阶段奖励。一个大型语言模型（LLM）代理在上下文中修订程序结构，利用执行反馈和经过验证的程序技能库，而数值优化器则调整其常数。每个候选程序在并行仿真场景中经过验证，程序探索新的方式来引导、重新利用和组合控制器现有的运动能力，从而在迭代中不断改进。实验表明，人类设计的奖励未能充分利用FB模型的运动操控能力，而InterEvolve演化的程序则释放了这种能力，有时通过新颖的策略。此外，它在仿真中为多样化任务、复杂场景和长时间跨度的组合生成行为，而演化的技能在基于自我中心的机载感知下在物理Unitree G1上自主运行。
cs.RO / 83 / 2610.02204
Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents
重建、实践、实现真实：面向具身智能体的引导自我提升
Abstract
Building reliable robot capabilities across diverse tasks requires substantial human effort to develop and maintain skills, design rewards, and integrate perception with control. We present Reconstruct, Practice, Go Real (RPG), a framework for autonomous improvement of robot execution systems without updating model weights. RPG identifies manipulation capabilities in an offline dataset and constructs related practice tasks in simulation. During practice, RPG uses execution feedback, privileged simulator state, and available dataset videos to diagnose failures. It develops new reusable symbolic skills, refines existing skills, and revises the system prompt based on these diagnoses. Cross-task evaluation tests individual candidate changes and merged revisions before they are retained for reuse. At test time, a multimodal LLM uses the resulting system prompt and skill library to coordinate perception and robot control. On held-out initializations of 22 manipulation tasks, RPG improves task success from 28.6% after the first practice round to 95.0% after 15 rounds, outperforming all evaluated baselines, including ASPIRE (75.5%) and CaP-Agent0 powered by GPT-6 Astra Pro (60.0%). After a common calibration and hardware-adaptation procedure, the frozen system succeeds in all 30 physical trials, with ten trials on each of three tasks. Project Website: https://rpg-robot.github.io/
Chinese Translation
在多样化任务中构建可靠的机器人能力需要大量人力来开发和维护技能、设计奖励以及整合感知与控制。我们提出了重建、实践、实现真实（Reconstruct, Practice, Go Real, RPG）框架，用于在不更新模型权重的情况下实现机器人执行系统的自主改进。RPG在离线数据集中识别操作能力，并在仿真中构建相关的实践任务。在实践过程中，RPG利用执行反馈、特权仿真器状态和可用数据集视频来诊断失败。它开发新的可重用符号技能，细化现有技能，并根据这些诊断修订系统提示。在测试时间，跨任务评估测试单个候选更改和合并修订，只有在通过测试后才会保留以供重用。在测试阶段，一个多模态的LLM（大语言模型）利用生成的系统提示和技能库来协调感知与机器人控制。在22个操作任务的保留初始化上，RPG将任务成功率从第一次实践轮的28.6%提高到15轮后的95.0%，超越了所有评估的基线，包括ASPIRE（75.5%）和由GPT-6 Astra Pro驱动的CaP-Agent0（60.0%）。经过共同的校准和硬件适应程序后，冻结系统在所有30个物理试验中成功，每个任务进行了十次试验。项目网站：https://rpg-robot.github.io/
cs.CV / 1 / 2610.00003
STATERA: Hidden Mass Estimation via Zero-Shot Sim-to-Real Kinematics using Frozen Temporal Tubelets
STATERA：通过冻结时间管道进行零-shot模拟到真实运动学的隐质量估计
Abstract
Vision models pretrained for frame-level appearance often struggle to infer hidden physical properties from motion. We study center-of-mass (CoM) localization for opaque, asymmetric rigid bodies from short monocular videos, where surface cues and point tracking are unreliable under self-occlusion. We propose STATERA, which adapts a pretrained video backbone (V-JEPA) with mostly frozen weights and a lightweight temporal tubelet mixer to predict per-frame CoM heatmaps and trajectories. To support this task, we introduce the HiddenMass Benchmark, comprising 50K MuJoCo trajectories and a 63-sequence real-world test set with physically calibrated CoM ground truth. In simulation, STATERA-50K-Sigma improves normalized CoM error from 41.7% (DINOv2) to 25.2%. In zero-shot sim-to-real transfer, we observe a fundamental trade-off in supervision: phase-aware targets can induce bimodal predictions, while phase-agnostic targets can collapse toward statistically safe centroids. Nevertheless, our phase-aware STATERA-50K-Crescent is the only evaluated method that demonstrates consistent movement toward the true hidden offset. While this leads to a monocular vector overshoot artifact that marginally increases absolute Euclidean error compared to a static geometric centroid, it improves physics capture from 2.6% to 41.0%. These results suggest that frozen temporal representations can better separate inertial dynamics from visual geometry for hidden-parameter estimation.
Chinese Translation
预训练的视觉模型在帧级外观方面通常难以从运动中推断隐藏的物理属性。我们研究了从短单目视频中对不透明、非对称刚体的质心（CoM）定位，其中表面线索和点跟踪在自遮挡情况下不可靠。我们提出了STATERA，该方法适应了一个预训练的视频主干（V-JEPA），其权重大部分被冻结，并结合一个轻量级的时间管道混合器，以预测每帧的质心热图和轨迹。为了支持这一任务，我们引入了HiddenMass基准，包括50K个MuJoCo轨迹和一个包含物理校准质心真实值的63序列真实世界测试集。在模拟中，STATERA-50K-Sigma将归一化质心误差从41.7%（DINOv2）改善到25.2%。在零-shot模拟到真实转移中，我们观察到监督中的基本权衡：相位感知目标可能导致双峰预测，而相位无关目标则可能向统计安全的质心收敛。然而，我们的相位感知STATERA-50K-Crescent是唯一经过评估的方法，展示了向真实隐藏偏移的一致移动。尽管这导致了一个单目向量超调伪影，与静态几何质心相比，绝对欧几里得误差略有增加，但它将物理捕捉从2.6%提高到41.0%。这些结果表明，冻结的时间表示可以更好地将惯性动态与视觉几何分离，从而实现隐藏参数的估计。
cs.CV / 2 / 2610.00006
Emergent Object Binding Has a Finite Spatial Horizon
涌现的物体绑定具有有限的空间范围
Abstract
Pretrained Vision Transformers encode whether two image patches belong to the same object. This IsSameObject signal is decodable from frozen patch embeddings at high accuracy, which suggests that object binding emerges from self-supervised pretraining alone. We show that this single accuracy number hides the structure of the signal. Binding is local: the probability that two patches of the same object are decoded as bound falls off monotonically with the distance between them and levels off at a nonzero floor, a falloff well described by an exponential with a finite length scale. This decay holds across object sizes, across three families of probe, on both ADE20K and COCO, and across DINO and CLIP backbones, which indicates that it is a property of the representation rather than of the decoder. Reading binding as local spatial coherence with a finite range accounts for a set of behaviors that the aggregate score leaves unexplained: binding weakens on large objects, separates distinct objects of the same class less reliably than objects of different classes, and groups object parts with their wholes. It is, by contrast, unaffected by occlusion once object size is controlled. We map each behavior with confounds controlled. As a preliminary observation, the horizon and its floor are organized at different depths in DINOv2 and DINOv3, which we report as suggestive given the small number of layers probed and the confound between the two models.
Chinese Translation
预训练的视觉变换器能够编码两个图像块是否属于同一物体。这种同一物体（IsSameObject）信号可以从冻结的图块嵌入中以高准确率解码，这表明物体绑定仅通过自监督预训练就能产生。我们展示了这个单一的准确率数字掩盖了信号的结构。绑定是局部的：同一物体的两个图块被解码为绑定的概率随着它们之间的距离单调下降，并在非零的底线处趋于平稳，这种衰减可以用具有有限长度尺度的指数函数很好地描述。这种衰减在不同的物体大小、三种探测器家族、ADE20K和COCO数据集，以及DINO和CLIP骨干网络中均保持一致，这表明它是表示的特性，而不是解码器的特性。将绑定视为具有有限范围的局部空间一致性可以解释一系列聚合得分未能解释的行为：在大物体上绑定减弱，同类物体之间的分离不如不同类物体可靠，以及将物体部分与整体进行分组。相反，一旦控制了物体大小，遮挡对绑定没有影响。我们在控制混杂变量的情况下映射每种行为。作为初步观察，DINOv2和DINOv3中的范围及其底线在不同的深度上组织，我们报告这一点是有启发性的，因为探测的层数较少，并且两个模型之间存在混杂因素。
cs.CV / 3 / 2610.00017
Spatial Lifting for Dense Prediction
用于密集预测的空间提升
Abstract
We present Spatial Lifting (SL), a novel methodology for dense prediction tasks. SL operates by lifting standard inputs, such as 2D images, into a higher-dimensional space and subsequently processing them using networks designed for that higher dimension, such as a 3D U-Net. Counterintuitively, this dimensionality lifting allows us to achieve good performance on benchmark tasks compared to conventional approaches, while reducing inference costs and \textbf{drastically lowering the number of model parameters}. The SL framework produces intrinsically structured outputs along the lifted dimension. This emergent structure facilitates dense supervision during training and enables single-forward-pass self-consistency-based quality and uncertainty estimation at test time. Spatial Lifting introduces a simple and general modeling strategy that offers a promising path toward more efficient, accurate, and reliable deep networks for dense prediction tasks in vision.
Chinese Translation
我们提出了一种新颖的密集预测任务方法——空间提升（Spatial Lifting, SL）。SL通过将标准输入（如二维图像）提升到更高维空间进行操作，随后使用为该高维度设计的网络（如3D U-Net）进行处理。反直觉的是，这种维度提升使我们在基准任务上相较于传统方法取得了良好的性能，同时降低了推理成本并 extbf{大幅减少了模型参数的数量}。SL框架在提升维度上产生了内在结构化的输出。这种新兴结构在训练期间促进了密集监督，并在测试时实现了基于单次前向传播的自一致性质量和不确定性估计。空间提升引入了一种简单而通用的建模策略，为在视觉领域中实现更高效、准确和可靠的深度网络用于密集预测任务提供了有希望的路径。
cs.CV / 4 / 2610.00024
Encoded but Disconnected: Decomposing Vision-Language Model Failures under a Patching Null
编码但未连接：在补丁无效下分解视觉-语言模型的失败
Abstract
Across three vision-language model architectures (LLaVA-1.5-7B, Qwen2.5-VL-7B, InternVL3-8B), we report a universal negative finding for mid-layer interpretability. On POPE -- the benchmark common to all three -- the mid layers encode the ground-truth answer in 68-91% of errors, yet this signal is not causally active for the final prediction: residual-stream patching yields 0% non-trivial flip at the layer level on all three architectures, and on two of three at the per-head level (Qwen: 0/12,600 patched forwards). The lone exception, InternVL3 layer-20 head-2, is a non-vocab, self-attending head whose effect is localized to that specific head (p < 1e-4). Despite the null, the errors separate operationally into three failure modes -- Perception Failure, Encoded-but-Disconnected, Prior-Override -- learnable above 60% on all three architectures, and the architecture's prior direction predicts which of two interventions elicits a category-specific response. We report these mitigation effects under oracle labels as evidence the categories are mechanistically real, not as a deployable method.
Chinese Translation
在三种视觉-语言模型架构（LLaVA-1.5-7B, Qwen2.5-VL-7B, InternVL3-8B）中，我们报告了一个关于中间层可解释性的普遍负面发现。在所有三种模型共同使用的基准POPE上，中间层在68-91%的错误中编码了真实答案，但这一信号对最终预测并没有因果作用：残差流补丁在所有三种架构的层级上产生了0%的非平凡翻转，而在三种架构中的两种（Qwen: 0/12,600个补丁前向）在每个头部级别上也是如此。唯一的例外是InternVL3的第20层第2个头部，这是一个非词汇的自注意力头，其影响局限于该特定头部（p < 1e-4）。尽管存在无效信号，这些错误在操作上分为三种失败模式——感知失败、编码但未连接、优先覆盖——在所有三种架构上可学习率超过60%，并且架构的先验方向预测哪种两种干预会引发特定类别的响应。我们在oracle标签下报告这些缓解效果，作为类别在机制上真实的证据，而非可部署的方法。
cs.CV / 5 / 2610.00030
Domain generalization and synthetic data in object detection: the enabler, the probe, and the gap
目标检测中的领域泛化与合成数据：推动者、探测器与差距
Abstract
Object detection models often experience performance degradation when deployed under distribution shifts, caused by for example changes in weather type, operational environment, or object appearance. Domain Generalization (DG) aims to develop models that remain robust under such shifts and generalize well to unseen domains. DG research specifically focused on object detection models is scarce, although these models face additional challenges around localization and multi-scale representations. Synthetic data is a promising tool to support in DG, by enabling large-scale generation of diverse new samples. In this paper, we present an object detection-centric review of DG and examine the role of synthetic data from three complementary perspectives. First, synthetic data acts as an enabler of DG through diversification and alignment strategies that aim to improve robustness to distribution shifts. Second, it serves as a probe that enables controlled experimentation to identify and understand failure modes. Third, we discuss the synthetic-to-real gap, a particularly challenging form of domain shift that arises when models trained on synthetic imagery are deployed on real-world data. Through reviewing these perspectives, we identify limitations of current DG approaches for object detection and argue that future research requires representation-aware methods that explicitly address both localization and classification under domain shift.
Chinese Translation
目标检测模型在面对分布变化时常常会出现性能下降，这种变化可能由天气类型、操作环境或物体外观的变化引起。领域泛化（Domain Generalization, DG）旨在开发在这些变化下依然保持鲁棒性的模型，并能够很好地推广到未见过的领域。尽管专注于目标检测模型的DG研究相对稀缺，但这些模型在定位和多尺度表示方面面临额外的挑战。合成数据作为一种有前景的工具，可以支持DG，通过大规模生成多样化的新样本来实现。在本文中，我们从目标检测的中心视角对DG进行了综述，并从三个互补的角度考察了合成数据的作用。首先，合成数据通过多样化和对齐策略作为DG的推动者，旨在提高对分布变化的鲁棒性。其次，它作为探测器，能够进行受控实验，以识别和理解失败模式。第三，我们讨论了合成与真实之间的差距，这是一种特别具有挑战性的领域转移形式，发生在基于合成图像训练的模型在真实数据上部署时。通过对这些视角的回顾，我们识别了当前目标检测DG方法的局限性，并认为未来的研究需要关注表示的意识方法，明确解决领域转移下的定位和分类问题。
cs.CV / 6 / 2610.00031
Seeing the City or Recognizing the Place? What Street-View Imagery Adds Beyond Existing Urban Data in VLM Urban Sensing
看城市还是识别地点？街景图像在VLM城市感知中超越现有城市数据的贡献
Abstract
Street-view imagery is increasingly used to infer urban attributes, but predictive accuracy alone does not reveal how much a photograph contributes beyond data already available for the same place. We compare image-based predictions with existing urban data across seven attributes from five public resources and three VLMs. The same urban units are evaluated using images, task context, nearby observations, and public records, while image replacements and conflicting records test source reliance. Existing urban data matched or exceeded image-only models for road damage, curb ramps, and house price, while neighbouring official statistics nearly matched the best image result for population. Images were more informative for building type, building function, and low-rise floor count. For floor count, image advantage increased by 5.7 percentage points per doubling of distance to the nearest labelled building and declined for tall buildings whose rooflines often fell outside the frame. Models frequently followed conflicting records. OpenFACADES floor annotations were generated with OpenStreetMap floor values and showed the opposite height-dependent error pattern from image-only reruns. Street-view image value therefore depends on visual legibility and local data coverage. Comparing images with existing urban data can guide image collection and clarify the provenance of derived urban maps.
Chinese Translation
街景图像越来越多地用于推断城市属性，但预测准确性本身并不能揭示一张照片在多大程度上超越了同一地点已有的数据。我们比较了基于图像的预测与来自五个公共资源和三个VLM的七个属性的现有城市数据。使用图像、任务背景、附近观察和公共记录评估相同的城市单元，同时通过图像替换和冲突记录测试数据源的可靠性。在路面损坏、坡道和房价方面，现有城市数据与仅基于图像的模型相匹配或超过，而邻近的官方统计数据几乎与最佳图像结果在人口方面相匹配。对于建筑类型、建筑功能和低层楼层数，图像提供了更丰富的信息。在楼层数方面，图像的优势随着距离最近标记建筑的距离翻倍而增加5.7个百分点，而对于高层建筑，由于其屋顶线常常超出画框，图像的优势则下降。模型经常遵循冲突记录。OpenFACADES的楼层注释是通过OpenStreetMap的楼层值生成的，并显示出与仅基于图像的重复运行相反的高度依赖误差模式。因此，街景图像的价值依赖于视觉可读性和地方数据覆盖。将图像与现有城市数据进行比较可以指导图像收集，并澄清衍生城市地图的来源。
cs.CV / 7 / 2610.00040
DSSR-3D: Decoupled Reasoning for View-Dependent Referring in 3D Gaussians
DSSR-3D：用于视角依赖的3D高斯体引用的解耦推理
Abstract
Recent advances in 3D Gaussian Splatting have enabled open-vocabulary and referring segmentation by distilling semantic knowledge from 2D foundation models into 3D representations. However, existing referring fields embed language features in a globally view-invariant space, making them fundamentally unable to resolve observer-centric spatial relations (e.g., "to the left of") that depend on camera pose. We propose DSSR-3D, an inference-time framework for view-dependent referring segmentation on continuous 3D Gaussian fields, formalized as two interfaces - pose-invariant semantic localization and pose-conditioned spatial reasoning - such that any pair of functions satisfying these constraints yields a valid instantiation, requiring no retraining of the underlying semantic field and no reliance on discrete geometric proxies such as bounding boxes. We instantiate the two interfaces with a temperature-sharpened softmax localization mechanism and a projection-based directional scoring function, fused via a lightweight, training-free step, and show they transfer zero-shot to structurally distinct semantic fields without adaptation. We further propose ViewRef-GS, a benchmark isolating view-dependent segmentation on 3D Gaussian fields, evaluated jointly with an augmented Ref-LERF to provide a comprehensive testbed for viewpoint-dependent spatial grounding. Experiments show consistent gains over existing 3DGS-based referring methods, with no additional training beyond the base semantic field
Chinese Translation
最近在3D高斯点云技术方面的进展，使得通过将2D基础模型中的语义知识提炼到3D表示中，实现了开放词汇和引用分割。然而，现有的引用场嵌入语言特征于一个全局视角不变的空间，使其在本质上无法解决依赖于相机姿态的观察者中心空间关系（例如，“在左侧”）。我们提出了DSSR-3D，这是一个在连续3D高斯场上进行视角依赖引用分割的推理框架，形式化为两个接口——姿态不变的语义定位和姿态条件的空间推理——使得满足这些约束的任意函数对都能产生有效的实例化，无需重新训练基础语义场，也不依赖于离散几何代理（如边界框）。我们通过温度加热的softmax定位机制和基于投影的方向评分函数来实例化这两个接口，并通过一个轻量级、无需训练的步骤进行融合，展示它们在没有适应的情况下能够零样本迁移到结构上不同的语义场。我们进一步提出了ViewRef-GS，这是一个孤立3D高斯场上视角依赖分割的基准，与增强的Ref-LERF共同评估，以提供一个全面的视点依赖空间基础测试平台。实验结果显示，与现有基于3DGS的引用方法相比，取得了一致的提升，且在基础语义场之外没有额外的训练。
cs.CV / 8 / 2610.00064
Reachability Is Not Generalization: Understanding Verb--Noun Decomposition in Assembly Action Recognition
可达性并非泛化：理解组装动作识别中的动词-名词分解
Abstract
Assembly actions are compositional: they combine a manipulation with a part or tool. In deployment, systems routinely encounter novel combinations of familiar components, yet an atomic action classifier assigns every unseen combination exactly zero probability by construction. The prevailing solution is verb--noun decomposition, which predicts components separately and recombines them to reach unseen actions. While widely adopted, how decomposition generalizes under compositional shift remains poorly understood. We present a systematic analysis of verb--noun decomposition across three assembly datasets (MECCANO, HAViD, and IMPACT). Although decomposition escapes the atomic ceiling, its generalization extends only partially beyond it. Unseen-composition performance remains strongly tied to the co-occurrence structure of the training data, indicating that much of the observed gain arises from interpolation within densely supported regions of the compositional space rather than from unconstrained recombination. Across datasets, failures consistently concentrate on the larger-vocabulary component, and IMPACT's verb-heavy vocabulary reverses the bottleneck from nouns to verbs. We further show that shared-encoder training introduces component entanglement, encouraging reliance on co-occurrence patterns that transfer poorly to unseen compositions and trailing independent recombination by up to $6.0\times$ in harmonic mean. Taken together, these findings explain why decomposition achieves only partial compositional generalization in practice. By identifying primitive support, vocabulary asymmetry, and component entanglement as connected sources of error, we provide a portable diagnostic framework for studying compositional recognition beyond aggregate accuracy. Code: https://github.com/hisalaheiyo/assembly.
Chinese Translation
组装动作是组合性的：它们将操作与部件或工具结合起来。在实际应用中，系统常常会遇到熟悉组件的新组合，但原子动作分类器在构造上会对每个未见组合分配零概率。当前的解决方案是动词-名词分解，它分别预测组件并重新组合以达到未见动作。尽管这种方法被广泛采用，但在组合变化下分解的泛化如何仍然缺乏深入理解。我们对三个组装数据集（MECCANO、HAViD 和 IMPACT）中的动词-名词分解进行了系统分析。尽管分解突破了原子上限，但其泛化仅部分超越了这一限制。未见组合的表现仍然与训练数据的共现结构密切相关，表明观察到的增益主要来源于在组合空间的密集支持区域内的插值，而非不受限制的重新组合。在各个数据集中，失败案例始终集中在更大词汇量的组件上，而 IMPACT 的动词重词汇反转了瓶颈，从名词转向动词。我们进一步表明，共享编码器训练引入了组件纠缠，促使依赖于共现模式，而这些模式在未见组合中转移效果不佳，并且在调和平均数上落后于独立重新组合多达 $6.0 imes$。综合来看，这些发现解释了为什么分解在实践中仅实现部分组合泛化。通过识别原始支持、词汇不对称和组件纠缠作为相关错误来源，我们提供了一个可移植的诊断框架，以研究超越整体准确性的组合识别。代码链接：https://github.com/hisalaheiyo/assembly。
cs.CV / 9 / 2610.00067
Robust Online Aero-Engine Blade Defect Detection via Dual-Alignment Test-Time Adaptation
通过双重对齐测试时适应实现鲁棒的在线航空发动机叶片缺陷检测
Abstract
Reliable visual inspection is essential for quality assurance in aero-engine blade manufacturing, where defect appearance may vary across production lines, imaging conditions, blade poses, and surface backgrounds. Such domain shifts cause a mismatch between training and deployment data and degrade the reliability of deep defect detectors in online inspection. This problem is particularly challenging because aero-engine blade images usually contain sparse defects, making pseudolabel-based adaptation vulnerable to noisy or missing predictions. To address this issue, we propose Aero-engine Blade Defect Detector (ABDD), an online adaptive detection framework based on test-time adaptation. ABDD introduces a Dual-Alignment Strategy to jointly adapt global visual style and local defect morphology by combining feature-statistics alignment with pseudo-box alignment. To reduce error accumulation from unreliable pseudo labels, an Uncertainty-aware Box Filtering mechanism evaluates pseudo boxes using classification confidence, classification entropy, and localization entropy. In addition, a lightweight Sparse Dilated Mona module enables parameter-efficient delta tuning while limiting source-domain forgetting. ABDD is evaluated on CD-AeBD and HD-AeBD under multiple domain-shift scenarios, with TTA strategies compared under a unified RT-DETR + Swin-T architecture. Experiments show that ABDD consistently improves detection robustness under domain shifts, and its practicality is further validated on an industrial inspection platform.
Chinese Translation
可靠的视觉检测对于航空发动机叶片制造中的质量保证至关重要，因为缺陷的外观可能因生产线、成像条件、叶片姿态和表面背景的不同而有所变化。这种领域转移导致训练数据与部署数据之间的不匹配，从而降低了深度缺陷检测器在在线检测中的可靠性。这个问题尤其具有挑战性，因为航空发动机叶片图像通常包含稀疏缺陷，使得基于伪标签的适应容易受到噪声或缺失预测的影响。为了解决这个问题，我们提出了航空发动机叶片缺陷检测器（Aero-engine Blade Defect Detector, ABDD），这是一个基于测试时适应的在线自适应检测框架。ABDD引入了一种双重对齐策略，通过结合特征统计对齐与伪框对齐，联合适应全局视觉风格和局部缺陷形态。为了减少来自不可靠伪标签的错误累积，采用了一种不确定性感知框过滤机制，通过分类置信度、分类熵和定位熵来评估伪框。此外，一个轻量级的稀疏扩张模块（Sparse Dilated Mona）实现了参数高效的增量调优，同时限制源领域的遗忘。ABDD在CD-AeBD和HD-AeBD上进行了多种领域转移场景下的评估，并在统一的RT-DETR + Swin-T架构下比较了TTA策略。实验表明，ABDD在领域转移下始终提高了检测的鲁棒性，其实用性在工业检测平台上得到了进一步验证。
cs.CV / 10 / 2610.00069
A Framework for Egocentric and Exocentric Procedural Understanding via Temporal Segmentation and Semantic Abstraction
通过时间分割和语义抽象实现自我中心与外部中心的程序理解框架
Abstract
Long-horizon ego/exo data contains rich procedural evidence, but are redundant, noisy, and costly to process or retain. We propose a compact framework that converts continuous multimodal workplace video into a structured Procedural State Memory, implemented as a Work Environment Model (WEM). Inspired by event segmentation theory, we detect boundaries using changes in visual context, location, motion, narration, gaze/object interaction, and optional exocentric workspace evidence, rather than fixed windows or visual novelty alone. Each segment is abstracted into an evidence-linked event card containing actor, interval, location, action, objects/tools, pre/post state, confidence, and provenance. These event cards incrementally update the WEM, enabling compact, auditable documentation and retrieval under on-premise privacy constraints. We instantiate the design with frozen DINOv2 and VJEPA-2 encoders and a local language model, and outline evaluation criteria for segmentation quality, memory compression, retrieval fidelity, and long-horizon QA.
Chinese Translation
长时间跨度的自我/外部数据包含丰富的程序证据，但处理或保留这些数据既冗余又嘈杂且成本高昂。我们提出了一个紧凑的框架，将连续的多模态工作场所视频转换为结构化的程序状态记忆，实施为工作环境模型（Work Environment Model, WEM）。受到事件分割理论的启发，我们通过视觉上下文、位置、运动、叙述、注视/物体交互以及可选的外部工作空间证据的变化来检测边界，而不是仅仅依赖固定窗口或视觉新颖性。每个片段被抽象为一个与证据关联的事件卡，包含参与者、时间段、位置、动作、物体/工具、前后状态、置信度和来源。这些事件卡逐步更新WEM，使得在本地隐私约束下能够进行紧凑、可审计的文档记录和检索。我们使用冻结的DINOv2和VJEPA-2编码器以及本地语言模型来实例化该设计，并概述了分割质量、记忆压缩、检索保真度和长时间跨度问答的评估标准。
cs.CV / 11 / 2610.00097
DramaAgent: Agentic Storytelling Video Generation
DramaAgent：自主叙事视频生成
Abstract
Recent diffusion and autoregressive models have substantially improved text-to-video generation, yet producing coherent long-form story videos with consistent characters and aligned audio remains challenging. Existing methods often suffer from narrative drift, unstable character identity, weak cross-scene continuity, and audio-visual mismatch over extended sequences. We propose DramaAgent, a hierarchical, agentic, and model-agnostic framework for long-form text-to-video-and-audio generation. Rather than improving the underlying video backbone itself, DramaAgent introduces an upper-level control layer that decomposes generation into story planning, persistent character conditioning, scene-wise synthesis, and reflection-guided targeted repair. The framework maintains reusable story and character states across scenes, diagnoses failures such as identity drift, missing scene semantics, temporal discontinuity, and cross-modal mismatch, and repairs problematic clips in a stage-specific manner. Experiments across multiple video generation backbones show that DramaAgent improves long-horizon coherence, character consistency, narrative fidelity, and scene-level audio-visual consistency over direct generation and strong baselines. These results suggest that hierarchical agentic control is a practical direction for controllable long-form audiovisual generation. Code: https://github.com/AIGeeksGroup/DramaAgent. Website: https://aigeeksgroup.github.io/DramaAgent.
Chinese Translation
近期的扩散模型和自回归模型显著提升了文本到视频生成的效果，但生成连贯的长篇故事视频，保持一致的人物形象和对齐的音频仍然面临挑战。现有方法常常遭遇叙事漂移、角色身份不稳定、跨场景连续性差以及在长序列中音视频不匹配等问题。我们提出了DramaAgent，一个层次化、自主的、模型无关的长篇文本到视频和音频生成框架。DramaAgent并不是直接改进底层视频骨干网络，而是引入了一个上层控制层，将生成过程分解为故事规划、持久角色条件、场景合成和反思引导的针对性修复。该框架在场景间保持可重用的故事和角色状态，诊断身份漂移、缺失场景语义、时间不连续性和跨模态不匹配等故障，并以阶段特定的方式修复问题片段。在多个视频生成骨干网络上的实验表明，DramaAgent在长时间一致性、角色一致性、叙事忠实度和场景级音视频一致性方面优于直接生成和强基线。这些结果表明，层次化自主控制是可控长篇视听生成的一个实际方向。代码：https://github.com/AIGeeksGroup/DramaAgent。网站：https://aigeeksgroup.github.io/DramaAgent。
cs.CV / 12 / 2610.00111
A Low Grounding Score Is Not an Ungrounded Judge: Identifying the Perceptibility Confound in Multimodal Oversight
低基础分数并不是无基础的评判者：识别多模态监督中的可感知混淆
Abstract
Model judges now supervise multimodal systems at scale, filtering training data, selecting outputs, and supplying the reward that shapes multimodal reasoning models. Trusting one means first checking that it uses its evidence, and that check is itself worth scrutinizing, so we ask whether a counterfactual probe of visual grounding measures what it claims to. The probe edits the image so the ground truth flips, holds the reasoning trace fixed, and asks whether the verdict follows. We formalize it as the Verdict Grounding Score and show it cannot be read the way such scores are read. A verdict responds only to an edit that reaches the judge's decision-relevant reading, so the score is capped by how perceptible the edit is, and unless editing makes the attribute easier to read, the error is one-sided: the score can only make a judge look less grounded than it is. The practical failure is therefore a false alarm, an auditor discarding a usable overseer. Under assumptions we state, we show this missing quantity is not merely bounded but identified from three quantities the same audit protocol already collects, which makes the false-alarm rate directly measurable rather than merely a concern. Auditing nine judges, we find the predicted ordering holds strictly across our entire primary pool, and the typical judge there acts on only about half of the edits whose attribute it can otherwise resolve. Applying a conservative rejection threshold certifies several cells as false alarms outright, the clearest being a judge that detects the injected error essentially every time while still scoring as if it had not used the image at all. The rule that follows is that an image-side counterfactual score should never be reported alone: a detection probe on the unedited image upper-bounds it, certifies its false alarms, and costs nothing extra to run.
Chinese Translation
模型评审者现在在大规模上监督多模态系统，过滤训练数据、选择输出并提供塑造多模态推理模型的奖励。信任一个评审者意味着首先要检查它是否使用了证据，而这一检查本身也值得审视，因此我们询问视觉基础的反事实探测是否测量了它所声称的内容。该探测编辑图像，使得真实情况翻转，保持推理轨迹不变，并询问判决是否随之变化。我们将其形式化为判决基础分数，并表明它不能像通常那样解读。判决仅对达到评审者决策相关阅读的编辑作出反应，因此分数受到编辑可感知程度的限制，除非编辑使属性更易于阅读，否则错误是单方面的：分数只能使评审者看起来比实际更不扎实。因此，实际的失败是一个误报，审计员丢弃了一个可用的监督者。在我们所述的假设下，我们表明这个缺失的量不仅是有界的，而且可以从同一审计协议已经收集的三个量中识别出来，这使得误报率可以直接测量，而不仅仅是一个关注点。在审计九位评审者时，我们发现预测的排序在我们整个主要样本中严格成立，而那里典型的评审者仅对大约一半的编辑作出反应，而这些编辑的属性他本可以解决。应用保守的拒绝阈值，明确认证了几个单元为误报，最明显的是一个评审者几乎每次都能检测到注入的错误，但仍然评分为似乎根本没有使用图像。随之而来的规则是，图像侧的反事实分数不应单独报告：对未编辑图像的检测探测上限其分数，认证其误报，并且运行成本为零。
cs.CV / 13 / 2610.00141
Evaluating the Robustness of Anti-UAV Detection under Controlled Fog Degradation: Fog-Aware Training and Clear-Sky Tradeoff
在受控雾霾退化下评估反无人机检测的鲁棒性：雾感知训练与晴空权衡
Abstract
Vision-based anti-UAV systems must function in poor visibility, yet most benchmarks use only clear-sky footage, and previous robustness studies treat adverse weather as a simple present/absent condition. As a result, the impact of fog severity on ground-to-air UAV detection remains poorly understood. This work presents the first severity-controlled fog benchmark for this task: synthetic fog at ten severity levels is applied to the RGB modality of the Anti-UAV300 dataset, comparing a clear-trained YOLOv5m baseline to a fog-aware model trained on both clear and foggy images. Detection performance drops sharply and non-linearly: degradation is front-loaded across light-to-moderate fog (beta approximately 0.05-0.10), with a 96% reduction in
[email protected]:0.95 from clear to thickest fog, mainly due to lost recall and confidence. On the comparable metric (
[email protected]), this collapse exceeds the most extreme rain degradation reported in the closest prior benchmark. Fog-aware training boosts detection across all severities (up to +0.320
[email protected]:0.95) with only a 9.1% drop in clear-sky accuracy, raising the threshold for reliable detection while not preventing collapse under extreme fog. Since reliability is lost within a narrow visibility range, simple clear vs adverse tests underestimate operational risk.
Chinese Translation
基于视觉的反无人机系统必须在能见度较差的情况下运行，但大多数基准测试仅使用晴空影像，之前的鲁棒性研究将恶劣天气视为简单的存在/不存在条件。因此，雾的严重程度对地面到空中无人机检测的影响仍然了解不足。本研究提出了该任务的第一个严重程度控制雾霾基准：在反无人机300（Anti-UAV300）数据集中，合成雾以十个严重程度应用于RGB模态，将在晴空条件下训练的YOLOv5m基线与在晴天和雾天图像上训练的雾感知模型进行比较。检测性能急剧且非线性下降：在轻度到中度雾霾下，退化主要集中在前期（beta约为0.05-0.10），从晴空到最浓雾的
[email protected]:0.95减少了96%，主要由于召回率和置信度的丧失。在可比指标（
[email protected]）上，这一崩溃超过了在最近的基准测试中报告的最极端降雨退化。雾感知训练在所有严重程度上提升了检测性能（最高可达+0.320
[email protected]:0.95），而晴空准确率仅下降了9.1%，提高了可靠检测的阈值，但在极端雾霾下并未防止崩溃。由于在狭窄的能见度范围内失去了可靠性，简单的晴天与恶劣天气测试低估了操作风险。
cs.CV / 14 / 2610.00196
GPEC: Efficient Pre-LLM Gaussian Process Embedding Correction for Cardiac Video Caption Generation
GPEC：用于心脏视频字幕生成的高效预训练大语言模型高斯过程嵌入修正
Abstract
Multimodal large language models (MLLMs) have shown strong potential for video understanding and caption generation, but their performance may decline in specialized medical imaging domains such as echocardiography. This work introduces Gaussian Process Embedding Correction (GPEC), a modular and computationally efficient pre-LLM error-correction method that improves the visual representations used by VideoChat2 for cardiac ultrasound caption generation. GPEC is inserted between the visual projection layer and the language model and learns a residual correction that moves the projected visual representation toward an annotation-guided target. The target is constructed by converting structured video annotations into qualitative attributes, generating a fixed-format reference caption, and mapping it into the language-model embedding space. The correction is modeled using a sparse variational Gaussian Process with inducing points, natural-parameter variational updates, and a block-wise linear kernel, while the original VideoChat2 components remain frozen.The method is evaluated using representation-level, caption-level, content-oriented, and execution-time metrics by comparing the original VideoChat2 with VideoChat2 + GPEC under identical input and reference conditions. Results show improved caption similarity and content alignment after applying the proposed correction. Furthermore, GPEC adds less than 0.05 s of inference-time overhead per video in the evaluated setting. These findings indicate that GPEC can improve caption generation in specialized medical video domains with minimal computational cost, without requiring end-to-end fine-tuning of the pretrained multimodal backbone.
Chinese Translation
多模态大语言模型（MLLMs）在视频理解和字幕生成方面展现出强大的潜力，但在超声心动图等专业医学影像领域，其性能可能会下降。本研究提出了高斯过程嵌入修正（GPEC），这是一种模块化且计算高效的预训练大语言模型错误修正方法，旨在改善VideoChat2在心脏超声字幕生成中使用的视觉表示。GPEC被插入在视觉投影层和语言模型之间，学习一种残差修正，将投影的视觉表示朝向注释引导的目标进行调整。该目标通过将结构化视频注释转换为定性属性，生成固定格式的参考字幕，并将其映射到语言模型嵌入空间来构建。该修正使用稀疏变分高斯过程建模，采用诱导点、自然参数变分更新和块状线性核，同时原始的VideoChat2组件保持不变。通过在相同输入和参考条件下比较原始VideoChat2与VideoChat2 + GPEC，使用表示级、字幕级、内容导向和执行时间指标对该方法进行了评估。结果显示，在应用所提出的修正后，字幕的相似性和内容对齐得到了改善。此外，在评估设置中，GPEC每个视频增加的推理时间开销不到0.05秒。这些发现表明，GPEC能够以最小的计算成本改善专业医学视频领域的字幕生成，而无需对预训练的多模态骨干进行端到端的微调。
cs.CV / 15 / 2610.00204
Query Independent Variable Rate Visual Token Coding
查询独立的可变速率视觉令牌编码
Abstract
Visual-token compression for vision--language models is posed almost entirely as a selection problem: decide which tokens to keep and discard the rest. The criteria that work best rank tokens by the attention the language model pays them, which makes the ranking a function of the question being asked. That is invisible in a single-turn benchmark and decisive whenever a compressed representation is written once and read many times, as when it is cached across the turns of a conversation or transmitted between a device and a server. We take the other half of the classical transform-coding toolkit instead: keep every token and vary its rate. A transform code exposes each token's measured distortion--rate curve, and a fixed bit budget is distributed across tokens by exact integer rate--distortion optimisation on those curves. No text enters the pipeline, so one compressed representation serves any query. At equal bit budgets, on two datasets and two capacities, it preserves the model's output distribution and its answers better than uniform-rate coding, the closed-form water-fill and distortion-ranked pruning. It matches attention-ranked pruning on the question pruning was tuned for, and overtakes it once the compressed image must answer a different question about the same image.
Chinese Translation
视觉令牌压缩在视觉-语言模型中几乎完全被视为选择问题：决定保留哪些令牌并丢弃其余的。最佳的标准是根据语言模型对令牌的关注程度进行排名，这使得排名成为所提问题的函数。这在单轮基准测试中是不可见的，但在压缩表示被写入一次并多次读取时是决定性的，例如在对话的多个回合中缓存，或在设备与服务器之间传输时。我们采用经典变换编码工具包的另一半：保留每个令牌并改变其速率。变换编码暴露了每个令牌的测量失真-速率曲线，并通过对这些曲线进行精确的整数速率-失真优化，将固定的比特预算分配到令牌上。管道中不输入文本，因此一个压缩表示可以服务于任何查询。在相同的比特预算下，在两个数据集和两个容量上，它比均匀速率编码、封闭形式的水填充和失真排名修剪更好地保留了模型的输出分布及其答案。它在调优问题的修剪上与关注排名修剪相匹配，并在压缩图像必须回答关于同一图像的不同问题时超越了它。
cs.CV / 16 / 2610.00279
Multi-Resolution Feature Fusion U-Net for Magnetic Resonance Imaging Segmentation
用于磁共振成像分割的多分辨率特征融合U-Net
Abstract
The segmentation of anatomical structures in medical images and particularly in MRI scans, is essential for clinical diagnosis and monitoring disease progression. While Deep Learning (DL) architectures, such as U-Net and its extensions are very effective in medical image segmentation tasks, they often struggle with preserving fine-grained details and global contextual information. This is especially challenging for MRI data segmentation, where anatomical structures are characterized by irregular boundaries and variations in shape, contrast, and scale. To address this challenge, we propose a novel DL architecture for MRI segmentation across different anatomical structures. Specifically, the architecture introduces a module, named Multi-Resolution Feature Fusion (MRFF), that can be easily integrated into any U-Net-like architecture. The MRFF is integrated in all levels of an encode-decoder structure, along with attention mechanisms and skip connections to extract features at multiple resolutions, enabling the model to capture both fine-grained details and global contextual information. We evaluate the MRFFU-Net on two publicly available benchmark MRI datasets of different anatomical targets; one for Cerebrospinal Fluid (CSF) segmentation in spinal MR scans, and one for left atrium cardiac segmentation, from the Medical Segmentation Decathlon (MSD) challenge. Experimental results indicate that MRFFU-Net outperforms state-of-the-art models across multiple evaluation metrics, demonstrating its effectiveness in MRI segmentation.
Chinese Translation
医学图像中解剖结构的分割，特别是在磁共振成像（MRI）扫描中，对于临床诊断和疾病进展监测至关重要。尽管深度学习（Deep Learning, DL）架构，如U-Net及其扩展，在医学图像分割任务中非常有效，但它们在保留细粒度细节和全局上下文信息方面常常面临挑战。这对于MRI数据分割尤其困难，因为解剖结构的边界不规则，形状、对比度和尺度变化多端。为了解决这一挑战，我们提出了一种新颖的DL架构，用于不同解剖结构的MRI分割。具体而言，该架构引入了一个模块，称为多分辨率特征融合（Multi-Resolution Feature Fusion, MRFF），可以轻松集成到任何类似U-Net的架构中。MRFF在编码-解码结构的所有层级中集成，同时结合注意力机制和跳跃连接，以提取多分辨率的特征，使模型能够捕捉细粒度细节和全局上下文信息。我们在两个公开可用的基准MRI数据集上评估了MRFFU-Net，这些数据集针对不同的解剖目标；一个用于脊柱MRI扫描中的脑脊液（Cerebrospinal Fluid, CSF）分割，另一个用于心脏左心房分割，数据来自医学分割十项全能（Medical Segmentation Decathlon, MSD）挑战。实验结果表明，MRFFU-Net在多个评估指标上优于最先进的模型，证明了其在MRI分割中的有效性。
cs.CV / 17 / 2610.00294
LENS-GRF: Permutation-Invariant Lesion Evidence Network with Gated Residual Fusion for Acne Severity Grading and Multi-Rater Clinical Oracle Analysis
LENS-GRF：具有门控残差融合的置换不变病灶证据网络用于痤疮严重程度分级和多评审临床专家分析
Abstract
Automated acne severity grading requires both whole-face context and fine-grained lesion evidence. We propose LENS-GRF (Lesion Evidence Network with Set-Transformer and Gated Residual Fusion), an interpretable multi-stage framework for four-class acne severity grading. The method combines Adaptive Facial Skin Segmentation and a global Vision Transformer prior with a permutation-invariant Lesion Set Transformer that encodes localized lesion patches and spatial geometry. Gated Residual Fusion adaptively controls the local residual contribution and reduces to the global prediction when the gate is zero. On ACNE04, fully automated LENS-GRF with YOLOv11s achieved 80.82% accuracy; with ground-truth lesion annotations, it achieved 95.89% +/- 0.59% accuracy and a Quadratic Weighted Kappa of 0.9753. A data-integrity audit identified 15 cross-split duplicate image pairs, including five with conflicting severity labels. In locked zero-shot evaluation on the full PLSBRACNE01 cohort (200 subjects, 600 views), automated LENS-GRF achieved 35.00% accuracy versus 42.50% for the global baseline. On the 148-subject common cohort used for three-dermatologist oracle analysis, ground-truth lesion inputs increased the best oracle accuracy to 47.97%, while the highest oracle QWK was 0.5799. Pairwise oracle agreement ranged from 49.32% to 66.22%, highlighting detector domain shift, annotation variability, and cross-criterion mismatch.
Chinese Translation
自动化的痤疮严重程度分级需要全脸上下文和细粒度的病灶证据。我们提出了LENS-GRF（具有集合变换器和门控残差融合的病灶证据网络），这是一个可解释的多阶段框架，用于四类痤疮严重程度分级。该方法结合了自适应面部皮肤分割和全球视觉变换器先验，以及一个置换不变的病灶集合变换器，后者编码了局部病灶补丁和空间几何信息。门控残差融合自适应地控制局部残差贡献，并在门控为零时减少到全局预测。在ACNE04数据集上，完全自动化的LENS-GRF与YOLOv11s达到了80.82%的准确率；在有真实病灶标注的情况下，准确率达到了95.89% +/- 0.59%，二次加权卡帕值为0.9753。数据完整性审计识别出15对交叉拆分的重复图像对，其中包括五对具有冲突严重程度标签的图像。在针对完整PLSBRACNE01队列（200名受试者，600个视图）的锁定零样本评估中，自动化的LENS-GRF达到了35.00%的准确率，而全球基线为42.50%。在用于三位皮肤科医生专家分析的148名受试者的共同队列中，真实病灶输入将最佳专家准确率提高至47.97%，而最高的专家二次加权卡帕值为0.5799。成对专家一致性范围为49.32%至66.22%，突显了检测器领域转移、标注变异性和跨标准不匹配的问题。
cs.CV / 18 / 2610.00302
Decoding the Disaster: Multi-Task Geospatial Reasoning with Vision-Language Models and Crowdsourced Imagery for Disaster Mapping
解码灾难：基于视觉-语言模型和众包影像的多任务地理空间推理用于灾难制图
Abstract
Crowdsourced imagery provides timely, fine-grained, street-level observations for disaster mapping, complementing conventional remote sensing imagery (RSI) during emergency response. However, such imagery is often unstructured, spatially ambiguous, and lacks reliable geographic metadata, making manual geolocalization and interpretation labor-intensive and difficult to scale. This work proposes a multi-task Geospatial Reasoning Disaster mapping framework, namely GRDisaster, to examine the potential of vision-language models (VLMs) in understanding, geolocalizing, and reasoning over crowdsourced disaster imagery. GRDisaster is built on a newly curated benchmark dataset derived from PhotoMappers, comprising 26,340 images organized into human-validated volunteered geographic information (VGI), street-view imagery (SVI), RSI cross-view triplets covering multiple disaster events from 2018 to 2024. The framework combines deterministic and probabilistic cross-view geolocalization with multi-view fusion to associate VGI images with georeferenced SVI and RSI. It introduces two sets of spatial reasoning indicators for cross-view geolocalization validation and disaster damage assessment. These indicators use structural, environmental, and global-scene cues to validate cross-view correspondences and visually observable damage evidence with expert-verified annotations to assess disaster severity, improving the interpretability of VLM outputs. To our knowledge, this study provides the first systematic investigation and unified evaluation framework for examining how VLM-based spatial reasoning can transform crowdsourced disaster imagery into actionable geospatial artificial intelligence (GeoAI) through cross-view geolocalization validation, interpretable spatial reasoning, and damage-aware severity assessment.
Chinese Translation
众包影像为灾难制图提供了及时、细致的街景观察，补充了常规遥感影像（RSI）在应急响应中的应用。然而，这类影像往往是非结构化的、空间上模糊的，并且缺乏可靠的地理元数据，使得手动地理定位和解读变得劳动密集且难以扩展。本研究提出了一种多任务地理空间推理灾难制图框架，称为GRDisaster，以探讨视觉-语言模型（VLMs）在理解、地理定位和推理众包灾难影像中的潜力。GRDisaster建立在一个新策划的基准数据集之上，该数据集来源于PhotoMappers，包含26,340幅图像，组织为经过人工验证的志愿地理信息（VGI）、街景影像（SVI）和覆盖2018年至2024年多个灾难事件的RSI交叉视图三元组。该框架结合了确定性和概率性交叉视图地理定位与多视图融合，以将VGI图像与地理参考的SVI和RSI关联起来。它引入了两组空间推理指标，用于交叉视图地理定位验证和灾害损害评估。这些指标利用结构、环境和全球场景线索来验证交叉视图对应关系和可视化的损害证据，并通过专家验证的注释来评估灾害严重性，从而提高VLM输出的可解释性。根据我们的了解，本研究提供了首个系统性调查和统一评估框架，以检验基于VLM的空间推理如何通过交叉视图地理定位验证、可解释的空间推理和损害感知的严重性评估，将众包灾难影像转化为可操作的地理空间人工智能（GeoAI）。
cs.CV / 19 / 2610.00315
Beyond Pixel Reconstruction: Retrieval-Guided Glyph-Aware Restoration for Low-Resource Manchu Historical Documents
超越像素重建：基于检索引导的字形感知修复方法用于低资源满族历史文献
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.
Chinese Translation
历史满族文献保存着宝贵的语言和文化遗产，但其数字化受到严重退化和配对训练数据稀缺的阻碍。现有的文档修复方法主要优化像素级重建，虽然能够产生视觉上合理的结果，但未能保持满族字形的结构特征。为了解决这一局限性，我们提出了一种基于检索引导的字形感知修复框架，超越了像素重建，通过明确地结合字形级结构知识。我们的方法检索相关的字形示例，以在修复过程中提供结构指导，并将这些信息整合到重建过程中，从而改善在低资源条件下退化字符结构的恢复。在满族历史文献上的大量实验表明，与现有修复方法相比，所提方法在图像修复质量和字形级保真度上均有所提升。这些结果突显了在低资源历史文献的可靠修复中，结合字符感知结构先验的重要性。
cs.CV / 20 / 2610.00319
EgoRefine: Ego-Referenced Predictive Alignment and Trajectory-Conditioned Reliability-Aware Fusion for Asynchronous Collaborative Perception
EgoRefine：用于异步协作感知的自我参考预测对齐与轨迹条件可靠性感知融合
Abstract
Collaborative perception enables connected agents to share complementary observations for 3D object detection, extending sensing range and mitigating occlusion. Under asynchronous communication, however, cooperative features arrive with temporal delay. Existing prediction-based methods compensate for these features mainly from the transmitting agent's own history, leaving residual misalignment with the ego agent's current observation; subsequent fusion also often overlooks spatial variations in alignment quality. We propose EgoRefine, an ego-referenced predictive alignment and reliability-aware fusion framework for asynchronous collaborative perception. Its Ego-referenced Predictive Alignment module uses the current ego feature to guide cooperative trajectory-field prediction and refines the sampling offsets along an ego-referenced trajectory direction. Its Trajectory-conditioned Reliability-aware Fusion module treats the trajectory discrepancy between the ego and cooperative streams and the directional refinement magnitude as alignment cues, using them to condition the relation between aligned features and adaptively reweight the two streams before convolutional fusion. Experiments on V2V4Real and DAIR-V2X-Seq show that EgoRefine outperforms TraF-Align by 1.6 and 2.9 points on average in
[email protected] and
[email protected], respectively. The source code will be made publicly available at https://github.com/godk0509/EgoRefine.
Chinese Translation
协作感知使得连接的智能体能够共享互补的观察数据，以实现三维物体检测，扩展感知范围并减轻遮挡。然而，在异步通信的情况下，协作特征会伴随时间延迟到达。现有的基于预测的方法主要依赖于传输智能体自身的历史来补偿这些特征，导致与自我智能体当前观察的残余不对齐；后续的融合过程往往也忽视了对齐质量的空间变化。我们提出了EgoRefine，一个用于异步协作感知的自我参考预测对齐和可靠性感知融合框架。其自我参考预测对齐模块利用当前的自我特征来指导协作轨迹场的预测，并沿着自我参考轨迹方向细化采样偏移量。其轨迹条件可靠性感知融合模块将自我智能体与协作流之间的轨迹差异和方向性细化幅度视为对齐线索，利用这些线索来调节对齐特征之间的关系，并在卷积融合之前自适应地重新加权这两个流。在V2V4Real和DAIR-V2X-Seq上的实验表明，EgoRefine在
[email protected]和
[email protected]上分别比TraF-Align平均提高了1.6和2.9分。源代码将公开发布于https://github.com/godk0509/EgoRefine。
cs.CV / 21 / 2610.00333
LEGO-OPD: Factorized Teacher Composition for Multimodal On-Policy Distillation
LEGO-OPD：用于多模态在线蒸馏的因子化教师组合
Abstract
Multimodal on-policy distillation (OPD) aims to improve visual grounding while preserving the strong reasoning capabilities of language models. Recent multi-teacher approaches combine LLM and VLM teachers to provide complementary supervision. However, directly using a VLM's full predictive distribution entangles its visual grounding signal with its own language prior, preventing the grounding information from being transferred independently. Conversely, increasing the strength of visual supervision can improve perception but may overemphasize visual evidence and degrade language reasoning. To address this trade-off, we introduce LEGO-OPD, which selectively composes factors from a Language Expert and a Grounding expert into One teacher distribution for multimodal OPD. Under a generalized Bayesian formulation, the language expert provides a prior over candidate tokens, while the grounding expert contributes a visual likelihood that updates this prior, rather than transferring its complete predictive distribution. This factorized composition allows language reasoning and visual grounding to be controlled independently. We further introduce adaptive calibration to determine how strongly the visual likelihood should update the language prior at each decoding prefix. Specifically, LEGO-OPD uses the grounding expert's image-induced prediction shift as a prefix-dependent reference, preventing both insufficient and excessive visual supervision. Experiments with Qwen3 models show that LEGO-OPD consistently outperforms the evaluated single- and multi-teacher OPD baselines on both multimodal and text-only reasoning tasks. Moreover, it improves the initial student's visual perception while preserving text-only reasoning.
Chinese Translation
多模态在线蒸馏（OPD）旨在提高视觉定位的同时保持语言模型的强大推理能力。最近的多教师方法结合了大型语言模型（LLM）和视觉语言模型（VLM）教师，以提供互补的监督。然而，直接使用VLM的完整预测分布将其视觉定位信号与自身的语言先验纠缠在一起，阻碍了定位信息的独立传递。相反，增强视觉监督的强度可以改善感知，但可能会过度强调视觉证据并削弱语言推理。为了解决这一权衡，我们提出了LEGO-OPD，它选择性地将语言专家和定位专家的因子组合成一个教师分布，用于多模态OPD。在广义贝叶斯框架下，语言专家为候选标记提供先验，而定位专家则贡献一个视觉似然，更新该先验，而不是转移其完整的预测分布。这种因子化组合允许语言推理和视觉定位独立控制。我们进一步引入自适应校准，以确定在每个解码前缀中视觉似然应多强烈地更新语言先验。具体而言，LEGO-OPD使用定位专家的图像诱导预测偏移作为前缀相关的参考，防止视觉监督不足或过度。与Qwen3模型的实验表明，LEGO-OPD在多模态和仅文本推理任务上始终优于评估的单教师和多教师OPD基线。此外，它在保持仅文本推理的同时改善了初始学生的视觉感知。
cs.CV / 22 / 2610.00350
Vmem-$\varphi$: Low-Compute Out-of-Distribution Detection in Spiking Neural Networks from Membrane-Potential Statistics
Vmem-$ ext{φ}$：基于膜电位统计的脉冲神经网络低计算量的分布外检测
Abstract
Spiking Neural Networks (SNNs) offer an energy-efficient approach to processing event-camera data, yet out-of-distribution (OOD) detection remains challenging in this setting. Existing OOD detection methods often depend on model outputs or computational components that are unavailable in object detection SNNs or are poorly suited to low-compute deployment. To that effect, we show that the subthreshold membrane potential \(V_{\mathrm{mem}}(t)\) provides a useful internal signal for detecting distribution shifts. Simple per-channel statistics derived from these membrane dynamics enable OOD detection. To evaluate this approach, we introduce Gen1-C, an event-camera corruption benchmark developed upon the Prophesee Gen1 automotive detection dataset, containing six sensor-motivated histogram-level stress tests at five severity levels. We further propose the Multi-Descriptor Deviation (MDD), a corruption-blind method that operates on membrane-potential statistics. At the highest corruption severity, MDD achieves an AUROC of more than 0.88 on five of the six corruptions using only a bounded 64-frame observation window. Notably, the remaining corruption is also the one that has the smallest effect on the underlying detector. These results show that the temporal membrane-potential dynamics can provide an effective and low-cost signal for OOD detection in SNN-based event perception.
Chinese Translation
脉冲神经网络（SNNs）为处理事件相机数据提供了一种能效高的方法，但在这一背景下，分布外（OOD）检测仍然具有挑战性。现有的OOD检测方法通常依赖于模型输出或在目标检测SNN中不可用的计算组件，或者不适合低计算量的部署。因此，我们展示了亚阈值膜电位（$V_{ ext{mem}}(t)$）为检测分布变化提供了有用的内部信号。基于这些膜动态的简单通道统计量使得OOD检测成为可能。为了评估这种方法，我们引入了Gen1-C，这是一个基于Prophesee Gen1汽车检测数据集开发的事件相机损坏基准，包含六个传感器驱动的直方图级压力测试，分为五个严重程度级别。我们进一步提出了多描述符偏差（Multi-Descriptor Deviation, MDD），这是一种基于膜电位统计的抗损坏方法。在最高损坏严重性下，MDD在六种损坏中的五种上仅使用有限的64帧观察窗口就达到了超过0.88的AUROC。值得注意的是，剩余的损坏也是对基础检测器影响最小的。这些结果表明，时间膜电位动态可以为基于SNN的事件感知中的OOD检测提供有效且低成本的信号。
cs.CV / 23 / 2610.00414
From Image Latent Space to Fuzzy Rules: Interpretable Analysis of Gastrointestinal Foundation Model
从图像潜在空间到模糊规则：对胃肠道基础模型的可解释性分析
Abstract
Foundation models pretrained on large-scale datasets demonstrate strong transferability to medical imaging tasks. However, understanding how their latent representations encode clinically relevant information remains an open challenge in safety-critical domains. This study proposes a prototype-based fuzzy-rule framework that interprets the patch-level features produced by the inner layers of pretrained foundation models, without any fine-tuning. Class-specific prototypes are learned by clustering in the feature space, yielding compact visual patterns. Patch features are then expressed as prototype similarities and classified by fuzzy rules with linguistic IF-THEN conditions that are human readable. The framework is applied across the final two blocks of ViT-S/16 backbones pretrained on ImageNet-1K and GastroNet-5M, and benchmarked against k-nearest neighbours, kernel SVM, and linear probing under identical frozen features, on wireless capsule endoscopy classification, gastrointestinal endoscopy classification, and colonic polyp segmentation. The experimental analysis shows that the proposed method, without backbone fine-tuning, reaches accuracy comparable to these black-box classifiers, and that domain-specific pretraining yields features that are both discriminative and symbolically compressible. Because the resulting rules are extracted from real data and expressed in interpretable terms, they are further used as an instrument to investigate synthetic medical images, providing a human-readable account of which real prototypes and rules a generator reproduces or fails to reproduce, localising where a synthetic image departs from real tissue rather than summarising it with a single score. The framework thus offers a transparent, depth-resolved view of how foundation models organise clinically relevant structure, together with a practical downstream use of the extracted rules.
Chinese Translation
在大规模数据集上预训练的基础模型在医学影像任务中表现出强大的迁移能力。然而，理解它们的潜在表示如何编码临床相关信息仍然是安全关键领域中的一个开放挑战。本研究提出了一种基于原型的模糊规则框架，该框架解释了预训练基础模型内部层生成的补丁级特征，而无需进行任何微调。通过在特征空间中聚类学习特定类别的原型，从而产生紧凑的视觉模式。然后，补丁特征被表示为原型相似度，并通过具有可读的语言IF-THEN条件的模糊规则进行分类。该框架应用于在ImageNet-1K和GastroNet-5M上预训练的ViT-S/16骨干网的最后两个模块，并在无线胶囊内窥镜分类、胃肠内窥镜分类和结肠息肉分割任务中，与k近邻、核SVM和线性探测器在相同的冻结特征下进行基准测试。实验分析表明，所提出的方法在不对骨干网进行微调的情况下，达到了与这些黑箱分类器相当的准确性，并且领域特定的预训练产生了既具有区分性又可以符号压缩的特征。由于生成的规则是从真实数据中提取并以可解释的术语表达的，因此它们进一步被用作研究合成医学图像的工具，提供了可读的说明，说明生成器重现或未能重现哪些真实原型和规则，定位合成图像与真实组织的偏离之处，而不是用单一得分进行总结。因此，该框架提供了一个透明的、深度分辨的视角，展示了基础模型如何组织临床相关结构，并为提取的规则提供了实际的下游应用。
cs.CV / 24 / 2610.00421
Scores That Hold, Benchmarks That Leak: Measuring Dataset Contamination in Public Brain-Tumor MRI Classification
保持的分数与泄露的基准：测量公共脑肿瘤MRI分类中的数据集污染
Abstract
Automated classification of brain tumors from MRI is a heavily published application of deep learning in medical imaging, with reported accuracies on public benchmarks routinely exceeding 98%. However, accuracy does not capture a critical dimension of benchmark quality: dataset integrity, defined as the independence of test from training data at the image, patient, and acquisition-source levels. We introduce a three-layer contamination framework comprising duplicate, patient, and source-label leakage to assess the public corpora on which this literature rests. We audit the three most widely used corpora against a chest-radiograph negative control and quantify each layer's effect on measured performance across nine architectures and three evaluation conditions. Contamination is severe at every layer: 28.8% of the dominant corpus's official test split has a near-twin in its own training split, a second corpus leaks 22.3% of its test images byte-identically, 95.5% of traceable test images share a patient with training, and file-header features containing no anatomy separate tumor from no-tumor at 0.959 balanced accuracy, at parity with fine-tuned ResNet backbones. The unexpected result is that removing every identified leaked test image leaves balanced accuracy essentially unchanged: stable performance after deduplication does not establish benchmark integrity. Our findings establish dataset integrity as a distinct, measurable axis of benchmark quality that a stable leaderboard cannot certify. For biomedical research, reported accuracy on these corpora alone does not establish that a model has learned to recognize tumors rather than exploit dataset-specific cues. We release the contaminated-file lists, recovered patient identifiers, and deduplicated splits.
Chinese Translation
从MRI中自动分类脑肿瘤是深度学习在医学影像领域的一项广泛发表的应用，报告的公共基准准确率常常超过98%。然而，准确率并未捕捉基准质量的一个关键维度：数据集完整性，定义为在图像、患者和获取来源层面上测试数据与训练数据的独立性。我们提出了一个三层污染框架，包括重复、患者和来源标签泄露，以评估该文献所依赖的公共语料库。我们对三种最广泛使用的语料库进行了审计，使用胸部X光片的阴性对照，并量化每一层对九种架构和三种评估条件下测量性能的影响。每一层的污染都很严重：主流语料库的官方测试集有28.8%的数据在其自身的训练集中存在近似双胞胎，第二个语料库有22.3%的测试图像以字节相同的方式泄露，95.5%的可追踪测试图像与训练数据共享患者，并且包含无解剖特征的文件头特征在0.959的平衡准确率下将肿瘤与非肿瘤区分开，且与微调的ResNet骨干网络相当。意外的结果是，移除每个已识别的泄露测试图像后，平衡准确率基本保持不变：去重后的稳定性能并未确立基准的完整性。我们的研究结果确立了数据集完整性作为基准质量的一个独特且可测量的维度，而稳定的排行榜无法证明这一点。对于生物医学研究，仅依赖这些语料库报告的准确率并不能证明模型已学会识别肿瘤，而不是利用数据集特定的线索。我们发布了污染文件列表、恢复的患者标识符和去重后的数据集。
cs.CV / 25 / 2610.00451
PACT: End-to-End Learning of Human Pose, Contacts, and Forces from Video
PACT：从视频中端到端学习人类姿态、接触和力
Abstract
Human motion, environmental contacts, and interaction forces are governed by common physical laws, yet existing approaches typically separate visual pose reconstruction from contact and force estimation. This separation limits joint reasoning and can propagate errors between stages. We introduce PACT, an end-to-end model that jointly learns to estimate human pose, contacts and contact forces from monocular video. Our approach augments a human reconstruction foundation model with learnable contact-force tokens and a temporal transformer that integrates visual features with world-space motion. Joint prediction heads refine human poses and estimate contacts and forces, while physics-based supervision encourages consistency between the reconstructed motion and interaction forces. To address the scarcity of force annotations, we develop a data annotation pipeline that combines contact labeling with physics-based motion and force optimization, producing training supervision from synthetic and real-world videos. We also introduce a real-world climbing benchmark ForceWall with climbing videos and corresponding ground-truth contact forces obtained from the force sensors. Experiments demonstrate state-of-the-art contact and force estimation, outperforming staged reconstruction approaches and generalizing to interactions beyond the training distribution. These results support end-to-end joint learning as an effective approach to recovering human motion and physical interactions from video.
Chinese Translation
人类运动、环境接触和交互力受共同物理法则的支配，然而现有的方法通常将视觉姿态重建与接触和力估计分开。这种分离限制了联合推理，并可能在不同阶段之间传播错误。我们提出了PACT，一个端到端模型，可以从单目视频中联合学习估计人类姿态、接触和接触力。我们的方法通过可学习的接触力标记和一个将视觉特征与世界空间运动相结合的时间变换器，增强了人类重建基础模型。联合预测头精炼人类姿态并估计接触和力，而基于物理的监督则鼓励重建运动与交互力之间的一致性。为了应对力注释的稀缺性，我们开发了一种数据注释管道，将接触标注与基于物理的运动和力优化相结合，从合成和真实世界视频中生成训练监督。我们还引入了一个真实世界的攀岩基准ForceWall，包含攀岩视频和通过力传感器获得的相应真实接触力。实验表明，我们的方法在接触和力估计方面达到了最先进的水平，优于分阶段重建方法，并且能够推广到超出训练分布的交互。这些结果支持端到端联合学习作为从视频中恢复人类运动和物理交互的有效方法。
cs.CV / 26 / 2610.00483
PixelDense: Dense Prediction as Representation Alignment for Pixel Diffusion
PixelDense：作为表示对齐的密集预测用于像素扩散
Abstract
Representation alignment (REPA) accelerates diffusion transformer training, but its alignment targets are almost exclusively semantic encoders such as DINOv2 and CLIP. Recent analysis points to spatial structure, not global semantics, as the carrier of the alignment effect, yet dense-prediction foundation models trained to predict that structure remain overlooked as REPA targets. In pixel-space diffusion, SAM2, Depth Anything v2, and Metric3D v2 each outperform the DINOv2-only GenEval baseline, with the two geometric teachers leading the segmentation teacher. A flat sum of all four teachers, however, lands below the best single geometric teacher, as semantic and geometric gradients compete for one denoiser projection. We introduce PixelDense, which routes DINOv2 and SAM2 through a semantic projection stream, routes Depth Anything v2 and Metric3D v2 through a geometric projection stream, and adds a weight-space orthogonality penalty that keeps the two streams in disjoint subspaces. All four teachers are frozen during training and dropped at inference. Applied to PixelGen and DeCo with a single recipe, PixelDense improves GenEval, DPG-Bench, and HPS v2.1, raises PixelGen-XXL's GenEval Overall from 0.7927 to 0.8093, and beats every single-teacher and unfactored multi-teacher variant. In partial-noise reconstruction, independent panoptic, depth, and surface-normal probes show up to 53.1% PQ gain and 36.0% depth AbsRel reduction at $\tau=0.5$ across COCO and Flickr30K. From random initialization, PixelDense also reaches the baseline's peak GenEval 1.23x faster. In SDEdit editing on PIE-Bench, PixelDense keeps more of the source background and layout at every edit strength, raising background PSNR by up to 2.2 dB.
Chinese Translation
表示对齐（REPA）加速了扩散变换器的训练，但其对齐目标几乎完全是语义编码器，如DINOv2和CLIP。最近的分析指出，空间结构而非全局语义是对齐效应的载体，然而，训练以预测该结构的密集预测基础模型仍然被忽视作为REPA目标。在像素空间扩散中，SAM2、Depth Anything v2和Metric3D v2的表现均优于仅使用DINOv2的GenEval基线，其中两个几何教师的表现优于分割教师。然而，四个教师的简单相加却低于最佳单一几何教师，因为语义和几何梯度在一个去噪器投影中竞争。我们提出了PixelDense，它通过语义投影流引导DINOv2和SAM2，通过几何投影流引导Depth Anything v2和Metric3D v2，并增加了一个权重空间正交惩罚，使两个流保持在不相交的子空间中。所有四个教师在训练期间被冻结，并在推理时被丢弃。应用于PixelGen和DeCo的单一配方，PixelDense改进了GenEval、DPG-Bench和HPS v2.1，将PixelGen-XXL的GenEval Overall从0.7927提高到0.8093，并超越了每个单一教师和未分解的多教师变体。在部分噪声重建中，独立的全景、深度和表面法线探针在COCO和Flickr30K上显示出高达53.1%的PQ增益和36.0%的深度绝对相对误差减少，$ au=0.5$。从随机初始化开始，PixelDense也以比基线快1.23倍的速度达到峰值GenEval。在PIE-Bench的SDEdit编辑中，PixelDense在每个编辑强度下保留了更多的源背景和布局，使背景PSNR提高了最多2.2 dB。
cs.CV / 27 / 2610.00544
Memorizon: Training World Models Beyond Their Context Window
Memorizon：超越上下文窗口训练世界模型
Abstract
Streaming world models should render a place consistently across repeated visits. Directly supervising such revisits requires training samples that capture both visits, often spanning minutes. Yet dense attention over the full span incurs quadratic costs, making long-span supervision expensive. Memorizon breaks this coupling: long spans are needed for supervision, but not for attention, since the two visits can share a forward pass without including every intervening frame. A training sample covers a span of any length but is scored only on its last $k$ chunks. Instead of tokenizing the history before them, each scored chunk retrieves its own top-$K$ latents by camera co-visibility, and the union of these requests forms a shared bank. The bank is bounded by $kK$, so the sequence stays bounded however long the span; at the shortest span the recipe is exactly conventional training. Adding the bank raises the cost of a step once; beyond that, a longer span costs little, and going from 100 to 400 s adds 12% to the step time. Against a sliding-window baseline, retrieval raises revisit consistency on every split, and a span long enough to reach the first visit of each return adds a further 24% to 30%, at some cost in image quality; beyond that span, more length no longer helps. Filling the bank from another episode lowers revisit correlation by 83%, so the model uses what it retrieves. Project page: https://tingtingliao.github.io/memorizon
Chinese Translation
流式世界模型应在重复访问中持续呈现一个地方。直接监督此类重访需要训练样本捕捉到两个访问，通常跨越几分钟。然而，对整个跨度进行密集注意会产生平方成本，使得长跨度监督变得昂贵。Memorizon 打破了这种耦合：监督需要长跨度，但注意不需要，因为两个访问可以共享一次前向传播，而不包括每个中间帧。一个训练样本覆盖任意长度的跨度，但仅在其最后 $k$ 个块上进行评分。每个被评分的块通过相机共视性检索其自身的前 $K$ 个潜在变量，而这些请求的并集形成一个共享库。该库的大小由 $kK$ 限制，因此序列在任何长度的跨度下保持有界；在最短跨度下，该方法与传统训练完全一致。添加库会增加一步的成本，但在此之后，较长的跨度几乎不增加成本，从 100 秒到 400 秒的跨度仅增加 12% 的步骤时间。与滑动窗口基线相比，检索在每个分割上提高了重访一致性，而一个足够长的跨度以达到每次返回的第一次访问又增加了 24% 到 30%，尽管在图像质量上有一定成本；超过该跨度，增加长度不再有帮助。从另一个剧集填充库会降低重访相关性 83%，因此模型使用其检索到的内容。项目页面：https://tingtingliao.github.io/memorizon
cs.CV / 28 / 2610.00559
PhysVista: Benchmarking Physical Intelligence in VLMs via a Perception-Reasoning-Assessment Loop
PhysVista：通过感知-推理-评估循环对视觉语言模型中的物理智能进行基准测试
Abstract
Vision-Language Models (VLMs) have shown strong multimodal reasoning capabilities, yet whether they truly capture the physical consistency underlying real-world dynamics remains unclear. Existing benchmark paradigms often suffer from fragmented evaluation, focusing on isolated cognitive stages while overlooking the inherent synergy between perception, reasoning, and physical judgment. The lack of a holistic perspective limits the ability to diagnose whether VLMs can reliably evaluate the physical authenticity of emerging generative models. To address these issues, we introduce PhysVista, a benchmark designed to evaluate physical intelligence in VLMs through a closed cognitive loop framework inspired by the human seeing-reasoning-assessment process. PhysVista restores this loop by jointly evaluating physical state perception, physical dynamics reasoning, and physical plausibility assessment. It further distinguishes event-level reasoning and scale-level reasoning to enable fine-grained analysis of physical understanding. In addition, PhysVista incorporates both real-world and AI-generated videos, allowing evaluation across diverse domains and emerging generative scenarios. Extensive experiments across a diverse set of VLMs reveal substantial limitations in physical reasoning and plausibility assessment, highlighting a persistent gap between visual recognition and genuine physical understanding, and pointing toward more principled designs for physically grounded multimodal intelligence.
Chinese Translation
视觉语言模型（VLMs）展示了强大的多模态推理能力，但它们是否真正捕捉到现实世界动态背后的物理一致性仍然不清楚。现有的基准范式往往存在评估碎片化的问题，专注于孤立的认知阶段，而忽视了感知、推理和物理判断之间的内在协同。缺乏整体视角限制了诊断VLMs是否能够可靠评估新兴生成模型的物理真实性的能力。为了解决这些问题，我们提出了PhysVista，一个旨在通过受人类视觉-推理-评估过程启发的闭合认知循环框架来评估VLMs中的物理智能的基准。PhysVista通过联合评估物理状态感知、物理动态推理和物理合理性评估来恢复这一循环。它进一步区分事件级推理和尺度级推理，以便对物理理解进行细致分析。此外，PhysVista结合了现实世界和AI生成的视频，允许在不同领域和新兴生成场景中进行评估。针对多种VLMs的广泛实验揭示了物理推理和合理性评估的显著局限性，突显了视觉识别与真正物理理解之间的持续差距，并指向更有原则的设计，以实现物理基础的多模态智能。
cs.CV / 29 / 2610.00573
FORTE: Adaptive Scoring and Exact Keyframe Selection for Long-Video Question Answering
FORTE：用于长视频问答的自适应评分和精确关键帧选择
Abstract
Query-aware keyframe selection enables multimodal large language models (MLLMs) to process long videos using only a small set of question-relevant frames. Existing score-based methods, however, typically search within a fixed, uniformly sampled candidate pool, preventing evidence outside this pool from ever being selected. Given a limited relevance-scoring budget, the key challenge is to allocate evaluations adaptively to promising frames while continuing to explore underrepresented temporal regions. We introduce FORTE, a training-free framework that addresses this challenge through two stages: adaptive relevance scoring and global keyframe optimization. Starting from sparse, uniformly distributed observations, our efficient Gaussian-process relevance predictor estimates relevance for unscored frames, exploiting temporal locality and the approximately banded kernel structure to reduce the core computation from cubic to linear time in the number of frames for fixed bandwidth. The scoring stage then selects which frames to score next by balancing predicted relevance with temporal coverage, prioritizing promising regions while also exploring less-represented parts of the video. The optimization stage selects the final keyframes by maximizing an objective that jointly captures measured relevance and temporal coverage. We derive an exact algorithm that leverages the logarithmic coverage structure to identify the optimal subset of the scored candidate pool in time linear in the pool size, for a fixed final-frame budget. Experiments on four long-video question-answering benchmarks show that FORTE achieves the highest observed mean accuracy among the compared selectors under every tested scoring budget. Further evaluations demonstrate its consistent effectiveness across different relevance scorers and downstream MLLMs.
Chinese Translation
查询感知的关键帧选择使得多模态大型语言模型（MLLMs）能够仅使用一小组与问题相关的帧来处理长视频。然而，现有的基于评分的方法通常在固定的、均匀采样的候选池中进行搜索，这使得池外的证据无法被选中。在有限的相关性评分预算下，关键挑战在于如何自适应地将评估分配给有前景的帧，同时继续探索代表性不足的时间区域。我们提出了FORTE，一个无训练的框架，通过两个阶段来解决这一挑战：自适应相关性评分和全局关键帧优化。从稀疏、均匀分布的观测开始，我们高效的高斯过程相关性预测器估计未评分帧的相关性，利用时间局部性和近似带状核结构将核心计算从立方时间减少到线性时间（对于固定带宽的帧数量）。评分阶段通过平衡预测相关性与时间覆盖，选择下一个要评分的帧，优先考虑有前景的区域，同时也探索视频中代表性较少的部分。优化阶段通过最大化一个同时捕捉测量相关性和时间覆盖的目标来选择最终的关键帧。我们推导出一个精确算法，利用对数覆盖结构在时间上以线性方式识别评分候选池的最佳子集，适用于固定的最终帧预算。在四个长视频问答基准上的实验表明，FORTE在每个测试的评分预算下都实现了比较选择器中最高的平均准确率。进一步的评估显示其在不同相关性评分器和下游MLLMs中的一致有效性。
cs.CV / 30 / 2610.00576
Gestalt: Large Multimodal Interplay Model
Gestalt：大型多模态交互模型
Abstract
In this paper, we propose Gestalt, a new paradigm of large multimodal model built around multimodal interplay. Despite rapid advances, large multimodal models are reaching a bottleneck: existing approaches focus primarily on accommodating additional modalities while overlooking the distinct characteristics of each modality and the relations among them. Motivated by the multistage property of human multisensory perception, we propose a multimodal interplay pyramid that organizes multimodal modeling as a progression from modality-specific processing, through cross-modal alignment, to deeper multimodal integration. Guided by this pyramid, Gestalt adopts a unified discrete diffusion framework and an interplay-partitioned architecture, with learnable interplay tokens mediating cross-modal exchange and integration. The pyramid also structures its data organization and training strategy. Strong performance across image generation, multimodal understanding, and text-only evaluation shows that Gestalt significantly improves cross-modal integration while preserving modality-specific information, effectively harnessing the strengths of diffusion-based multimodal models and offering a promising path toward unified multimodal intelligence.
Chinese Translation
在本文中，我们提出了Gestalt，这是一种围绕多模态交互构建的新型大型多模态模型范式。尽管快速发展，大型多模态模型正面临瓶颈：现有方法主要集中于适应额外模态，而忽视了每种模态的独特特征及其之间的关系。受到人类多感官感知的多阶段特性的启发，我们提出了一种多模态交互金字塔，将多模态建模组织为从特定模态处理、跨模态对齐到更深层次的多模态整合的进程。在这一金字塔的指导下，Gestalt采用了统一的离散扩散框架和交互分区架构，具有可学习的交互令牌，以调解跨模态的交换和整合。该金字塔还构建了其数据组织和训练策略。在图像生成、多模态理解和仅文本评估方面表现出色，表明Gestalt在显著改善跨模态整合的同时，保留了模态特定信息，有效利用了基于扩散的多模态模型的优势，为统一多模态智能提供了有希望的路径。
cs.CV / 31 / 2610.00582
Discrete Annotation, Continuous Preference: Rethinking Supervision for Accurate and Generalizable Aesthetic Image Cropping
离散注释，连续偏好：重新思考准确且可推广的美学图像裁剪的监督方式
Abstract
Aesthetic image cropping aims to identify the optimal crop of an image in terms of aesthetics and composition. While supervision based on annotated data is fundamental, the field has been hindered by a long-standing problem: existing datasets suffer from (1) human subjectivity and (2) rigid discreteness confined to fixed sampling grids. These flawed annotations not only limit the accuracy and generalization of trained models but also severely distort fair evaluation. To overcome this, we propose to model human cropping preference as a multi-peaked, continuous, and sharp field over the crop space. We introduce the Continuous Preference Field (CPF), which recovers a dense preference landscape from discrete annotations through (1) peak clustering, (2) off-lattice refinement, (3) negative shaping, and (4) field assembly. Based on this, we train CPIC, a VLM-based cropping model optimized via GRPO with the CPF reward, which overcomes template collapse, achieving state-of-the-art performance and exceptional out-of-domain generalization. Finally, to resolve the long-standing benchmark evaluation crisis, we introduce CPICD, a comprehensive recalibration of existing ground-truth boxes. By leveraging the CPF to correct grid-bound artifacts across mainstream benchmarks, CPICD establishes a rigorous and reliable foundation for future cropping research. Extensive experiments and user studies demonstrate the superiority of our CPF, CPIC, and CPICD. Code, model, and data are available at https://github.com/zzqingz/CPIC.
Chinese Translation
美学图像裁剪旨在识别图像在美学和构图方面的最佳裁剪。尽管基于注释数据的监督是基础，但该领域受到一个长期存在的问题的阻碍：现有数据集存在（1）人类主观性和（2）局限于固定采样网格的僵化离散性。这些缺陷注释不仅限制了训练模型的准确性和泛化能力，还严重扭曲了公平评估。为了解决这个问题，我们提出将人类裁剪偏好建模为一个多峰、连续且尖锐的场在裁剪空间中。我们引入了连续偏好场（Continuous Preference Field, CPF），通过（1）峰值聚类，（2）离网精炼，（3）负形状和（4）场组装，从离散注释中恢复出密集的偏好景观。在此基础上，我们训练了CPIC，一个基于VLM的裁剪模型，通过CPF奖励优化的GRPO，克服了模板崩溃，实现了最先进的性能和卓越的域外泛化。最后，为了解决长期存在的基准评估危机，我们引入了CPICD，对现有真实框进行了全面的重新校准。通过利用CPF来纠正主流基准中的网格绑定伪影，CPICD为未来的裁剪研究建立了一个严谨且可靠的基础。大量实验和用户研究证明了我们的CPF、CPIC和CPICD的优越性。代码、模型和数据可在https://github.com/zzqingz/CPIC获取。
cs.CV / 32 / 2610.00600
Just Align $\bm{x}$: Aligning Predictions, Not Representations
仅对齐 $m{x}$：对齐预测，而非表示
Abstract
Representation alignment has become an effective way to accelerate diffusion training, but its benefits do not transfer reliably to pixel-space clean-image prediction. In JiT, we find that auxiliary feature alignment can improve access to semantic features while reducing access to image variation needed for clean-image prediction, creating a mismatch between the auxiliary objective and the denoising task. This suggests a different principle: auxiliary supervision should improve the prediction target itself rather than impose a separate representation target. We introduce JAx (Just Align x), a prediction-supervision method that aligns clean-image predictions across noise levels. JAx couples a noisier student observation with a cleaner observation through a Markov degradation that preserves the original JiT input distribution. Under this coupling, the oracle prediction from the cleaner state has the same conditional mean as the optimal JiT target, while its conditional target covariance is no greater. Thus, oracle prediction alignment preserves the population JiT objective up to a constant while providing a lower-variance training target. To make this construction practical with an imperfect EMA teacher, JAx combines ground-truth supervision with a reliability-gated coupling band that selects nearby teacher states based on prediction risk. On ImageNet 256x256, JAx consistently improves FID and accelerates convergence across JiT-B/16, L/16, and H/16, without an external encoder or changes to the architecture or sampling procedure. Gradient diagnostics further show reduced minibatch gradient variance, while ablations demonstrate that the gains cannot be explained by time reweighting alone. These results show that prediction-space supervision provides a simple and principled alternative to representation alignment for pixel-space generative models.
Chinese Translation
表示对齐已成为加速扩散训练的有效方法，但其优势并不能可靠地转移到像素空间的干净图像预测中。在 JiT 中，我们发现辅助特征对齐可以改善对语义特征的访问，同时减少对干净图像预测所需的图像变化的访问，从而在辅助目标与去噪任务之间造成不匹配。这提示我们一个不同的原则：辅助监督应该改善预测目标本身，而不是施加一个单独的表示目标。我们引入 JAx（Just Align x），一种预测监督方法，它在不同噪声水平下对齐干净图像预测。JAx 通过一种保持原始 JiT 输入分布的马尔可夫退化，将噪声更大的学生观测与更干净的观测相结合。在这种耦合下，来自更干净状态的 oracle 预测具有与最佳 JiT 目标相同的条件均值，而其条件目标协方差不大于此。因此，oracle 预测对齐在保持总体 JiT 目标的同时提供了一个较低方差的训练目标。为了使这一构造在不完美的 EMA 教师下变得实用，JAx 将真实标签监督与基于预测风险选择附近教师状态的可靠性门控耦合带相结合。在 ImageNet 256x256 上，JAx 一直提高 FID，并加速 JiT-B/16、L/16 和 H/16 的收敛，而无需外部编码器或对架构或采样过程进行更改。梯度诊断进一步显示出减少的小批量梯度方差，而消融实验表明，这些增益不能仅通过时间重加权来解释。这些结果表明，预测空间监督为像素空间生成模型提供了一种简单且有原则的替代方案，以替代表示对齐。
cs.CV / 33 / 2610.00623
HAWK: Rethinking Multimodal Drafting for Speculative Decoding
HAWK：重新思考多模态草拟在推测解码中的应用
Abstract
Speculative decoding has achieved substantial lossless speedups for LLMs, but remains less effective for large vision-language models (LVLMs), where lightweight drafters struggle to use rich multimodal information. A second limitation is that standard distillation supervises the drafter only along the original training trajectory, without modeling how target predictions shift after the drafter's own proposals. As drafting moves away from this trajectory, the drafter can increasingly disagree with the target, reducing acceptance in later steps. We propose HAWK to address both limitations. HAWK uses representation similarity to select informative target layers and learns how to combine their hidden states. For visual information, it directly provides the drafter with compressed visual hidden states from the target model instead of raw visual tokens, making the visual information easier for a shallow drafter to use. HAWK also trains the drafter to capture how target predictions change after its own proposals, improving its agreement with the target during multi-step drafting. On SmolVLM-256M across ten multimodal benchmarks, HAWK raises average acceptance length from 3.32 to 4.08 and speedup from 2.19x to 2.60x over EAGLE-3 under greedy decoding, and from 2.89 to 3.41 and 1.92x to 2.19x under sampling.
Chinese Translation
推测解码在大型语言模型（LLMs）中实现了显著的无损加速，但在大型视觉语言模型（LVLMs）中效果较差，其中轻量级草拟器难以利用丰富的多模态信息。第二个限制是，标准蒸馏仅在原始训练轨迹上对草拟器进行监督，而未建模草拟器自身提案后目标预测的变化。随着草拟逐渐偏离这一轨迹，草拟器与目标之间的分歧可能加大，从而降低后续步骤的接受度。我们提出HAWK以解决这两个限制。HAWK利用表示相似性选择信息丰富的目标层，并学习如何组合它们的隐藏状态。对于视觉信息，它直接向草拟器提供来自目标模型的压缩视觉隐藏状态，而不是原始视觉标记，使得浅层草拟器更容易使用视觉信息。HAWK还训练草拟器捕捉目标预测在其自身提案后的变化，从而在多步骤草拟过程中提高与目标的一致性。在SmolVLM-256M的十个多模态基准测试中，HAWK将平均接受长度从3.32提升至4.08，将加速比从2.19倍提升至2.60倍（在贪婪解码下），并在采样下将接受长度从2.89提升至3.41，加速比从1.92倍提升至2.19倍。
cs.CV / 34 / 2610.00666
VisionQ: VLM-as-a-Judge Taxonomy, Dataset and Benchmark for Qualitative Analysis in Computer Vision
VisionQ：用于计算机视觉定性分析的VLM作为评判者的分类法、数据集和基准
Abstract
Qualitative comparison figures are central evidence in computer vision papers, and vision-language models (VLMs) are increasingly used to judge them. Yet existing benchmarks score only scalar quality or overall preference, so a judge can be rewarded for picking the preferred image for the wrong visual reason. We introduce VisionQ, the first benchmark built from peer-reviewed CV comparison figures that grounds every judgment in a named visual criterion: each question states the criterion, and a judge is credited only when it selects the output the authors identify as best on that criterion. We call this task criterion-conditioned visual discrimination. VisionQ comprises (1) a corpus of 1,409 CVPR and ICCV papers with 1,800+ validated comparison figures and 3,911 hand-annotated data points linking method crops to author-stated visual claims; (2) a six-axis, 51-leaf taxonomy of the visual criteria behind qualitative judgment; (3) a criterion-conditioned evaluation protocol that hides method names, captions, and paper identity and reports accuracy per criterion; and (4) VisionQ-Judge, a DPO-tuned Gemma-4-E4B judge trained on symmetric evidence pairs, which reduces last-option predictions by 7.0pp and improves accuracy by 2.5pp on a held-out test set. Evaluating 20 open- and closed-source VLM judges, we find that the strongest reach only 63.1% accuracy (chance 32.2%) and that reliability varies sharply across criteria. Code: https://github.com/ReML-AI/visionq. Data: https://huggingface.co/datasets/visionq-anon-2026/VisionQ-1k.
Chinese Translation
定性比较图是计算机视觉论文中的核心证据，而视觉语言模型（VLM）越来越多地被用于评判这些图像。然而，现有基准仅对标量质量或整体偏好进行评分，因此评判者可能因为选择了错误的视觉理由而获得奖励。我们推出了VisionQ，这是第一个基于同行评审的计算机视觉比较图构建的基准，它将每个判断基于一个命名的视觉标准：每个问题都表明标准，评判者仅在选择作者所识别的在该标准上最佳的输出时获得积分。我们将这一任务称为标准条件下的视觉区分。VisionQ包含（1）1,409篇CVPR和ICCV论文的语料库，包含1,800多个经过验证的比较图和3,911个手动注释的数据点，将方法裁剪与作者所述的视觉主张关联；（2）一个包含六个轴、51个叶子的定性判断视觉标准分类法；（3）一个标准条件评估协议，该协议隐藏方法名称、标题和论文身份，并按标准报告准确性；（4）VisionQ-Judge，一个经过DPO调优的Gemma-4-E4B评判者，训练于对称证据对上，减少了7.0个百分点的最后选择预测，并在保留的测试集上提高了2.5个百分点的准确性。评估20个开源和闭源的VLM评判者，我们发现最强的评判者仅达到63.1%的准确率（随机机会为32.2%），并且可靠性在各个标准之间差异显著。代码：https://github.com/ReML-AI/visionq。数据：https://huggingface.co/datasets/visionq-anon-2026/VisionQ-1k。
cs.CV / 35 / 2610.00677
Harnessing Vision-Language Models for Perceptual Quality Assessment and Autonomous Content Adjustment in Augmented Reality
利用视觉-语言模型进行增强现实中的感知质量评估与自主内容调整
Abstract
Advancements in augmented reality (AR) continue to foster innovative solutions, facilitating novel methodologies within educational systems, healthcare delivery, and risk-mitigation protocols. However, optimizing for end-user immersion and comfort remains challenging, as AR head-mounted displays contend with constrained scene geometry, spatial jitter, and temporal instability. User studies are the standard AR evaluation method for visual quality, but their cost, diminishing scalability, and inflexibility pose bottlenecks during iterative application design. To address this problem, we present an automated framework for AR content evaluation and refinement, built on vision-language models (VLMs), to evaluate and predict the visual fidelity of AR scenes as perceived by users. First, we introduce RateAR, a benchmark of AR images and videos collected across diverse scenes and environmental conditions, with good-to-excellent reliability (ICC(2,5) >= .90) across perceptual factors, including object placement, scale, and shadow consistency. Subsequently, we evaluate eleven commercial VLMs on the crafted benchmark. Results support that VLM-based quality predictions strongly correlate with human subjective judgments, achieving Spearman's rank-order correlations of up to 0.8695. An ablation study further suggests that, compared to other prompting strategies, our contextual prompting yields better alignment with human ratings while balancing introduced complexity cues. Building on these findings, we construct an automated AR content adjustment system and conduct a 21-participant user study. More than 90% of participants found that the system improved placement and size coherence of virtual content.
Chinese Translation
增强现实（AR）的进展不断推动创新解决方案，促进教育系统、医疗服务和风险缓解协议中的新方法。然而，优化最终用户的沉浸感和舒适度仍然具有挑战性，因为AR头戴显示器面临着受限的场景几何、空间抖动和时间不稳定性。用户研究是评估AR视觉质量的标准方法，但其高成本、可扩展性下降和灵活性不足在迭代应用设计中造成了瓶颈。为了解决这个问题，我们提出了一种基于视觉-语言模型（VLMs）的AR内容评估与优化的自动化框架，以评估和预测用户感知的AR场景的视觉保真度。首先，我们介绍了RateAR，这是一个涵盖多种场景和环境条件的AR图像和视频基准，具有良好到优秀的可靠性（ICC(2,5) >= .90），涵盖了对象放置、比例和阴影一致性等感知因素。随后，我们在构建的基准上评估了十一种商业VLM。结果表明，基于VLM的质量预测与人类主观判断之间存在强相关性，斯皮尔曼等级相关系数高达0.8695。消融研究进一步表明，与其他提示策略相比，我们的上下文提示在与人类评分的对齐上表现更佳，同时平衡了引入的复杂性线索。在这些发现的基础上，我们构建了一个自动化的AR内容调整系统，并进行了21名参与者的用户研究。超过90%的参与者认为该系统改善了虚拟内容的放置和大小一致性。
cs.CV / 36 / 2610.00686
SemanTok: Predictable Semantic Tokens for Efficient Autoregressive Video Generation
SemanTok：用于高效自回归视频生成的可预测语义标记
Abstract
Recent video-based world models pair the scalability of autoregressive (AR) prediction with the visual quality of diffusion models. The choice of scene tokenizer is paramount for the optimal performance of each of these, both in terms of fidelity and semantics. Flexible-length, coarse-to-fine tokenizers yield exactly that: the first coarse tokens carry the clip's global semantics while later tokens further specify details. Existing flexible tokenizers only apply a representation-alignment (REPA) loss on early decoder hidden states, a target the decoder can partly meet from its noised input instead. We introduce SemanTok, a flexible video tokenizer that feeds frozen DINO features into its encoder and adds lightweight heads that reconstruct them from each retained token prefix alone. SemanTok achieves high semantic alignment and video fidelity at every AR model size: a 201M SemanTok AR model matches or beats a VideoFlexTok AR model $3.4\times$ its size, and larger SemanTok AR models further improve fidelity. It keeps semantic alignment on out-of-distribution classes and gives the decoder higher semantic alignment at every noise level, including pure noise. It performs well in both reconstruction and generation, and its short token prefixes are cheaper to predict and give better generation fidelity, with pixel detail deferred to later tokens.
Chinese Translation
最近的基于视频的世界模型将自回归（AR）预测的可扩展性与扩散模型的视觉质量相结合。场景标记器的选择对于每种模型的最佳性能至关重要，无论是在保真度还是语义方面。灵活长度的粗到细标记器正是实现这一目标的关键：首个粗略标记承载了剪辑的全局语义，而后续标记则进一步细化细节。现有的灵活标记器仅在早期解码器隐藏状态上应用表示对齐（REPA）损失，而解码器可以部分通过其带噪声的输入来满足这一目标。我们提出了SemanTok，这是一种灵活的视频标记器，它将冻结的DINO特征输入到其编码器中，并添加轻量级头部，仅通过每个保留的标记前缀重建这些特征。SemanTok在每个AR模型规模下都实现了高语义对齐和视频保真度：一个201M的SemanTok AR模型的性能与一个大小为其$3.4 imes$的VideoFlexTok AR模型相当或更好，而更大的SemanTok AR模型进一步提高了保真度。它在分布外类别上保持语义对齐，并在每个噪声水平下，包括纯噪声，给予解码器更高的语义对齐。它在重建和生成方面表现良好，其短标记前缀预测成本更低，并提供更好的生成保真度，同时将像素细节推迟到后续标记。
cs.CV / 37 / 2610.00691
Soundwich: Video Generation with Layered and Controllable Audio
Soundwich：具有分层和可控音频的视频生成
Abstract
Recent joint audio-video generative models can synthesize realistic videos with synchronized sound, but typically generate audio as a single mixed track. This limits source-level control and differs from practical audiovisual workflows, where speech, music, sound effects, and ambient sounds are represented as separate editable tracks. We introduce Soundwich, a training-free framework that transforms a frozen joint audio-video flow-matching model into a generator of multiple synchronized, independently editable audio stems coupled to a shared video. Soundwich generates separate audio stems with explicit control over their temporal activity. To keep separately generated sounds coherent, we introduce a shared scene representation that communicates global audiovisual context across stems while preserving their source-level separation. We further route cross-modal interactions between each audio stem and its corresponding visual source, improving audiovisual consistency. The resulting stems remain synchronized with the video and can be independently retimed, muted, replaced, or remixed. Experiments and human evaluations show improved temporal control, source separation, and naturalness, while enabling flexible source-level editing within coherent audiovisual generation. Code is available at https://github.com/CodyNing/Soundwich.
Chinese Translation
最近的联合音频-视频生成模型能够合成与声音同步的真实视频，但通常生成的音频为单一混合轨道。这限制了源级控制，并与实际的视听工作流程不同，后者将语音、音乐、音效和环境声音表示为可单独编辑的轨道。我们介绍了Soundwich，一个无训练的框架，它将一个冻结的联合音频-视频流匹配模型转变为一个生成多个同步、独立可编辑音频轨道的生成器，这些音频轨道与共享视频相结合。Soundwich生成的音频轨道具有对其时间活动的明确控制。为了保持单独生成的声音的一致性，我们引入了一个共享场景表示，它在轨道之间传达全局视听上下文，同时保持源级分离。我们进一步在每个音频轨道与其对应的视觉源之间引导跨模态交互，从而提高视听一致性。生成的音频轨道与视频保持同步，并可以独立重新定时、静音、替换或混音。实验和人类评估显示了改进的时间控制、源分离和自然性，同时在一致的视听生成中实现了灵活的源级编辑。代码可在 https://github.com/CodyNing/Soundwich 获取。
cs.CV / 38 / 2610.00693
FedMAD: Modulation-Aware Directional Aggregation for Federated Learning in Remote Sensing Image Classification
FedMAD：针对遥感图像分类的联邦学习中的调制感知方向聚合
Abstract
Federated learning (FL) has recently attracted increasing attention in remote sensing (RS) since it enables collaborative model training across decentralized RS image archives without requiring direct access to local data. However, FL performance significantly degrades when the data distributions between clients are heterogeneous, which often occurs due to geographical differences, seasonal changes, and varying image acquisition and atmospheric conditions. To address this challenge, in this letter, we propose a novel personalized FL framework (denoted as FedMAD) for RS image classification problems. The proposed framework separates globally shared representation parameters from client-specific adaptation parameters to preserve client-specific features while maintaining globally transferable representations. This is achieved by integrating lightweight modulation modules and local batch normalization layers into the backbone network. Although globally shared parameters are collaboratively optimized between clients, client-specific parameters remain local to preserve domain-specific feature characteristics. In addition, FedMAD introduces a modulation-aware directional aggregation strategy that dynamically adjusts the importance of aggregation for each client according to the alignment of local modulation updates. This allows the global optimization process to suppress conflicting client updates originating from heterogeneous data distributions while enhancing the contribution of clients with consistent adaptation behaviors. The experimental results obtained on the BigEarthNet-S2 and EuroSAT datasets demonstrate the effectiveness of FedMAD compared to state-of-the-art FL algorithms under heterogeneous RS data distributions. The code of the proposed framework will be publicly available at https://git.tu-berlin.de/rsim/fedmad.
Chinese Translation
联邦学习（FL）近年来在遥感（RS）领域引起了越来越多的关注，因为它能够在去中心化的遥感图像档案中进行协作模型训练，而无需直接访问本地数据。然而，当客户端之间的数据分布存在异质性时，FL的性能会显著下降，这种情况通常由于地理差异、季节变化以及不同的图像获取和大气条件所导致。为了解决这一挑战，本文提出了一种新颖的个性化FL框架（称为FedMAD），用于遥感图像分类问题。该框架将全球共享的表示参数与客户端特定的适应参数分离，以保留客户端特定特征，同时保持全球可迁移的表示。这是通过将轻量级调制模块和本地批归一化层集成到主干网络中来实现的。尽管全球共享参数在客户端之间进行协作优化，但客户端特定参数仍然保持本地，以保留领域特定的特征特性。此外，FedMAD引入了一种调制感知的方向聚合策略，根据本地调制更新的对齐情况动态调整每个客户端的聚合重要性。这使得全球优化过程能够抑制源于异质数据分布的冲突客户端更新，同时增强具有一致适应行为的客户端的贡献。在BigEarthNet-S2和EuroSAT数据集上的实验结果表明，与最先进的FL算法相比，FedMAD在异质遥感数据分布下的有效性。所提出框架的代码将公开发布在 https://git.tu-berlin.de/rsim/fedmad。
cs.CV / 39 / 2610.00737
Personalized Image Generation with Reasoning and Reflection
基于推理与反思的个性化图像生成
Abstract
Personalized image generation has remained narrowly focused on conditional synthesis from curated visual exemplars, rather than capturing who a user is. In practice, however, a user's personal context is much richer, comprising reviews, posts, images, captions, and metadata accumulated over time. A truly personalized generator should leverage this history to produce images aligned with the user's lifestyle and aesthetic preferences. To this end, we introduce the first unified benchmark for personalized image generation from user histories. The benchmark comprises two complementary tasks and a multi-axis evaluation protocol that assesses target fidelity, visual quality, user distinguishability, semantic alignment with the user's history, and task-specific utility. Grounded in real-world e-commerce and social media settings, the benchmark includes: (1) Personalized Scene Generation, which places a given object in a scene that reflects a user's preferences and lifestyle, motivated by personalized product presentation; and (2) Personalized Creative Generation, which generates a novel image on a specified topic that is faithful to a user's aesthetic and visual identity, motivated by social media content creation. We further propose PEARL, which couples a multimodal reasoner with a frozen image generator in an interleaved reason-reflect loop optimized with differential data reward. Across both tasks, PEARL outperforms strong baselines, achieving an average improvement of 15% across personalization metrics.
Chinese Translation
个性化图像生成一直以来主要集中于从精心挑选的视觉示例中进行条件合成，而不是捕捉用户的真实身份。然而，在实际应用中，用户的个人背景要丰富得多，包括随着时间积累的评论、帖子、图像、标题和元数据。一个真正个性化的生成器应该利用这些历史信息，生成与用户的生活方式和审美偏好相一致的图像。为此，我们提出了第一个针对用户历史的个性化图像生成的统一基准。该基准包括两个互补任务和一个多维评估协议，评估目标保真度、视觉质量、用户可区分性、与用户历史的语义一致性以及特定任务的实用性。基于真实的电子商务和社交媒体环境，该基准包括：(1) 个性化场景生成，将给定对象放置在反映用户偏好和生活方式的场景中，旨在个性化产品展示；(2) 个性化创意生成，在特定主题上生成忠实于用户审美和视觉身份的新图像，旨在社交媒体内容创作。我们进一步提出了PEARL，它将多模态推理器与冻结的图像生成器结合在一个交替的推理-反思循环中，并通过差异数据奖励进行优化。在这两个任务中，PEARL的表现超过了强基线，在个性化指标上平均提高了15%。
cs.CV / 40 / 2610.00749
What Builds the Scene? Luminance Dominates Geometry Formation in 3D Gaussian Splatting
是什么构建了场景？亮度主导3D高斯喷溅中的几何形状形成
Abstract
Standard 3D Gaussian Splatting (3DGS) learns geometry and appearance jointly from RGB supervision, making it difficult to isolate how luminance and chroma contribute to the learned representation. We study this by training models under different channel supervision, freezing their non-appearance parameters (position, scale, rotation, and opacity), and re-estimating appearance with the same solver before comparing held-out reconstruction. Across eleven benchmark scenes with four independent runs each, geometry learned from luminance alone supports held-out reconstruction 0.085 dB below RGB-trained geometry on average. If chroma is deleted from a trained model, a sufficiently expressive solver can re-fit it on the frozen geometry to the original quality or slightly better. Higher-order spherical harmonics contribute much more reconstruction quality to luminance than to chroma, improving PSNR by 1.44 dB versus 0.19 dB on average, although on mirror-like surfaces hue does still change with viewpoint. The luminance advantage is even larger when geometry is being formed. Chroma-only supervision produces geometry 3.9-5.5 dB worse than luminance-only supervision after the same appearance solve; densification explains part of this gap. Overall, geometry formation in standard 3DGS is strongly luminance-dominated but not luminance-exclusive, and much of the chromatic appearance can be recovered after spatial support has formed.
Chinese Translation
标准的3D高斯喷溅（3DGS）从RGB监督中共同学习几何和外观，这使得难以分离亮度和色度对学习表示的贡献。我们通过在不同通道监督下训练模型，冻结其非外观参数（位置、尺度、旋转和不透明度），并在比较保留重建之前使用相同的求解器重新估计外观来研究这一点。在十一组基准场景中，每组进行了四次独立实验，单独从亮度学习的几何形状在平均上支持的保留重建比RGB训练的几何形状低0.085 dB。如果从训练模型中删除色度，足够表达能力的求解器可以在冻结的几何形状上重新拟合到原始质量或稍好。高阶球谐函数对亮度的重建质量贡献远大于对色度的贡献，平均提高PSNR 1.44 dB，而对色度的提高仅为0.19 dB，尽管在镜面表面上色调确实会随着视点的变化而变化。当几何形状正在形成时，亮度的优势更为显著。仅依赖色度的监督在经过相同的外观求解后产生的几何形状比仅依赖亮度的监督差3.9-5.5 dB；密集化解释了这一差距的一部分。总体而言，标准3DGS中的几何形状形成在很大程度上是由亮度主导的，但并非完全依赖于亮度，许多色彩外观可以在空间支持形成后被恢复。
cs.CV / 41 / 2610.00757
Video Evidence Indexing: Learning Where to Look from Video Previews for Token-Budgeted Long-Video Question Answering
视频证据索引：从视频预览中学习观察位置以进行令牌预算的长视频问答
Abstract
Long-video question answering is limited by the high cost of visual tokens and by the fixed context width of current VLMs. A long-video question may require broad temporal coverage, but the answer is often supported by only a compact set of moments. To locate these moments efficiently, we propose token-budgeted Video Evidence Indexing (VEI): given a dense low-resolution Video Preview, the model constructs a compact high-resolution Evidence Set for final reasoning. We treat VEI as a policy that must jointly solve \textit{evidence localization}, which finds question-relevant moments, and \textit{budget planning}, which decides where to spend the limited high-resolution frame budget. We implement this idea with an inference pipeline: the Video Preview provides cheap global coverage, Video Evidence Indexing constructs the Evidence Set, and Answer Generation combines both inputs for final VQA. To address missing frame-level supervision, we adopt privileged self-distillation, where an answer-aware teacher guides the normal test-time policy on student-generated indexing traces. We explore previews at 1, 6, 12, and 24 visual tokens per frame, training a single policy that supports all four resolutions. Experiments show that Video Evidence Indexing improves accuracy under limited visual budgets, and self-distillation further improves both QA accuracy and temporal evidence localization.
Chinese Translation
长视频问答受到视觉令牌高成本和当前视觉语言模型（VLM）固定上下文宽度的限制。长视频问题可能需要广泛的时间覆盖，但答案通常仅由一组紧凑的时刻支持。为了高效定位这些时刻，我们提出了令牌预算的视频证据索引（Video Evidence Indexing, VEI）：给定一个密集的低分辨率视频预览，模型构建一个紧凑的高分辨率证据集以进行最终推理。我们将VEI视为一种策略，必须共同解决证据定位（evidence localization），即寻找与问题相关的时刻，以及预算规划（budget planning），即决定在哪里花费有限的高分辨率帧预算。我们通过推理管道实现这一思想：视频预览提供廉价的全局覆盖，视频证据索引构建证据集，而答案生成结合这两种输入进行最终的问答。为了解决缺失的帧级监督，我们采用特权自蒸馏（privileged self-distillation），其中一个关注答案的教师指导正常测试时策略在学生生成的索引轨迹上进行。我们探索了每帧1、6、12和24个视觉令牌的预览，训练一个支持所有四种分辨率的单一策略。实验表明，视频证据索引在有限视觉预算下提高了准确性，而自蒸馏进一步提高了问答准确性和时间证据定位。
cs.CV / 42 / 2610.00785
VTV-FM: Flow Matching through Variational Terminal-Velocity Closure
VTV-FM：通过变分终端速度闭合进行流动匹配
Abstract
Flow matching (FM) learns generative transport by fitting continuous-time motion from a simple source distribution to the data distribution. Most existing methods use first-order bridges: once a source and a target sample are paired, the path is a straight motion with constant velocity. FM with optimal transport (OT) improves the pairing, but the bridge itself remains linear, limiting its ability to model curved motion, acceleration, and changing directions. A natural remedy is to use second-order phase-space dynamics; however, learning the bridge requires target-side terminal-velocity information that static datasets do not provide. We propose Variational Terminal-Velocity Flow Matching (VTV-FM), a second-order FM framework that derives the missing velocity by minimizing acceleration energy, yielding a closed-form closure for static data. The same minimum-acceleration variational construction also defines the OT pairing cost and the acceleration targets used for training. Experiments on low-dimensional datasets, PDE-governed physical fields, and CIFAR-10 show that VTV-FM improves transport geometry and generation quality over first-order and high-order FM baselines.
Chinese Translation
流动匹配（FM）通过将简单源分布的连续时间运动拟合到数据分布来学习生成性传输。现有的大多数方法使用一阶桥接：一旦源样本和目标样本配对，路径就是以恒定速度进行的直线运动。使用最优传输（OT）的流动匹配改善了配对，但桥接本身仍然是线性的，限制了其对曲线运动、加速度和方向变化的建模能力。一种自然的解决办法是使用二阶相空间动力学；然而，学习桥接需要目标侧的终端速度信息，而静态数据集并未提供。我们提出了变分终端速度流动匹配（VTV-FM），这是一个二阶流动匹配框架，通过最小化加速度能量推导缺失的速度，从而为静态数据提供了封闭形式的闭合。相同的最小加速度变分构造还定义了OT配对成本和用于训练的加速度目标。在低维数据集、偏微分方程（PDE）驱动的物理场和CIFAR-10上的实验表明，VTV-FM在传输几何和生成质量上优于一阶和高阶流动匹配基线。
cs.CV / 43 / 2610.00809
Paying for Too Many Tokens? Valid and Cost-Efficient Multimodal LLM Annotation with Simple Heuristics
支付过多的标记费用？使用简单启发式方法进行有效且成本高效的多模态大语言模型注释
Abstract
Vision-Language Models (VLMs) enable video annotation at scale, but costs accumulate quickly: processing a typical 60-second short-form video at one frame per second requires millions of tokens. To reduce costs, researchers rely on heuristics such as sampling a subset of frames, compressing videos into image grids, or using only a single modality. However, it remains unclear which heuristics save cost, and whether they preserve the downstream conclusions these annotations enable. To address this gap, we conduct a systematic evaluation of these heuristics using short-form videos, on two computational social science (CSS) tasks: sentiment and topic classification. We evaluate each configuration along three axes the literature typically treats separately: classification accuracy, validity of downstream inference, and per-video token cost. First, we find that accuracy and validity diverge: the highest-accuracy configuration can produce wrong conclusions. Second, modality value is not guaranteed: text alone can yield strong performance, indicating that adding modalities can add cost without adding signal. Finally, we find that cost can be decoupled from video length when annotating short-form videos: a single $2\times8$ image grid built via simple shot-transition detection approaches full-video understanding ($\kappa$ within~.05), at $\sim 15\%$ of the token cost. Based on these findings, we derive guidelines that can enable cost-aware VLM annotation in CSS.
Chinese Translation
视觉-语言模型（VLMs）使得视频注释能够大规模进行，但成本迅速累积：处理一段典型的60秒短视频，每秒处理一帧需要数百万个标记。为了降低成本，研究人员依赖于一些启发式方法，例如抽样一部分帧、将视频压缩为图像网格，或仅使用单一模态。然而，目前尚不清楚哪些启发式方法能够节省成本，并且这些方法是否保留了这些注释所支持的下游结论。为了解决这一问题，我们对这些启发式方法进行了系统评估，使用短视频在两个计算社会科学（CSS）任务上进行测试：情感分析和主题分类。我们从三个通常在文献中分别处理的维度评估每种配置：分类准确性、下游推理的有效性和每个视频的标记成本。首先，我们发现准确性和有效性存在差异：最高准确性的配置可能会得出错误的结论。其次，模态的价值并不保证：仅使用文本也能取得良好的表现，这表明添加模态可能会增加成本而不增加信息。最后，我们发现，在对短视频进行注释时，成本可以与视频长度解耦：通过简单的镜头切换检测构建的单个$2 imes8$图像网格能够实现完整视频理解（$ ext{kappa}$在~.05以内），其标记成本约为15%。基于这些发现，我们提出了能够实现成本意识的VLM注释的指导方针。
cs.CV / 44 / 2610.00812
Video Generation Models: A Survey of Post-Training and Alignment
视频生成模型：后训练与对齐的综述
Abstract
Video generation has rapidly progressed from short, low-quality clips to high-resolution, long-duration sequences with complex spatiotemporal dynamics. Despite strong generative priors learned through large-scale pretraining, pretrained video models often fail to reliably follow human intent, maintain temporal coherence, or satisfy physical and safety constraints. Compared with image and text generation, alignment in video generation presents unique challenges, including error accumulation over time, motion-appearance coupling, multi-objective trade-offs, and limited supervision for temporal properties. These challenges motivate systematic post-training strategies that adapt pretrained models without retraining them from scratch. In this survey, we present the first comprehensive review of post-training and alignment in video generation models. We frame post-training as a unifying framework and distinguish between implicit alignment and explicit alignment based on how alignment signals are enforced. From this perspective, we organize existing approaches into four broad categories: supervised fine-tuning methods, self-training and distillation methods, preference- and reward-based methods, and inference-time methods. This taxonomy provides a coherent view of how alignment signals shape model behavior across both training and deployment. Beyond methodological advances, we review commonly used datasets, benchmarks, and evaluation practices, and discuss open challenges such as scalable reward design, long-horizon temporal consistency, stability-expressiveness trade-offs, and safety-aware generation. This survey aims to provide a structured conceptual foundation and practical guidance for advancing controllable and reliable video generation models.
Chinese Translation
视频生成技术已经从短小、低质量的片段迅速发展到高分辨率、长时长的序列，并具备复杂的时空动态。尽管通过大规模预训练学习到了强大的生成先验，预训练的视频模型往往无法可靠地遵循人类意图、保持时间一致性或满足物理和安全约束。与图像和文本生成相比，视频生成中的对齐面临独特的挑战，包括随时间累积的误差、运动与外观的耦合、多目标权衡以及对时间特性的有限监督。这些挑战促使我们提出系统的后训练策略，以适应预训练模型，而无需从头开始重新训练。在本综述中，我们首次全面回顾了视频生成模型中的后训练与对齐。我们将后训练框架视为一个统一的框架，并根据对齐信号的强制执行方式区分隐式对齐和显式对齐。从这个角度出发，我们将现有方法组织为四个广泛类别：监督微调方法、自我训练与蒸馏方法、基于偏好和奖励的方法，以及推理时方法。这个分类法为对齐信号如何在训练和部署中塑造模型行为提供了一个连贯的视角。除了方法论的进展外，我们还回顾了常用的数据集、基准和评估实践，并讨论了开放挑战，如可扩展的奖励设计、长时间序列的一致性、稳定性与表现力的权衡，以及安全意识生成。本综述旨在为推动可控和可靠的视频生成模型提供结构化的概念基础和实用指导。
cs.CV / 45 / 2610.00825
Align Then Reason: A Multimodal Lip-Sync Judge for Dubbing
先对齐再推理：一种多模态的配音同步评估方法
Abstract
Dubbing quality control requires a reference-free judge that can determine whether a candidate text line matches a speaker's visible articulation in both content and timing, using only silent video and text because dubbed audio may not yet exist. Existing visual speech recognizers and video-language models are poorly suited to this setting: even when fine-tuned to recover spoken content from lip motion, they remain largely insensitive to temporal errors. We introduce $\textit{Align Then Reason}$ (ATR), a multilingual lip-sync judge that first establishes a monotonic alignment between frame-level lip representations and the phonetic units of the candidate line, then reasons over this alignment to make the final judgment. An alignment scorer provides the LLM with both local evidence for each phonetic unit and a calibrated global alignment score, enabling it to reason jointly about content and timing. On a seven-language benchmark, our method improves mean AUC over the corresponding Qwen3.5 SFT baselines by 59.4%, 50.2%, and 50.8% with 2B, 4B, and 9B reasoners, respectively. The gains generalize across LLM families, reaching mean AUC improvements of 45.9% and 46.6% over the best baseline for LLaMA-3.1-8B and Mistral-7B, respectively. They also transfer across datasets to three unseen MuAViC languages. Furthermore, we evaluate on two downstream tasks built from real dubbing lines. On dub-line reranking, ATR-9B outperforms the best lip-reading baseline by 52.0%, while on script-to-clip assignment, ATR-9B improves over the best lip-reading baseline by 17.7%.
Chinese Translation
配音质量控制需要一种无参考的评估方法，该方法能够仅使用静默视频和文本来判断候选文本行是否在内容和时序上与说话者的可见发音相匹配，因为配音音频可能尚不存在。现有的视觉语音识别器和视频语言模型在这种情况下表现不佳：即使经过微调以从唇部运动中恢复口语内容，它们对时序错误仍然不够敏感。我们提出了$ extit{Align Then Reason}$（ATR），一种多语言的配音同步评估方法，它首先在帧级唇部表征与候选行的音素单元之间建立单调对齐，然后基于这一对齐进行推理以做出最终判断。对齐评分器为大型语言模型（LLM）提供每个音素单元的局部证据和一个经过校准的全局对齐评分，使其能够共同推理内容和时序。在七种语言的基准测试中，我们的方法在对应的Qwen3.5 SFT基线之上提高了均值AUC，分别为59.4%、50.2%和50.8%，对应的推理器为2B、4B和9B。这些提升在不同的LLM家族中具有普适性，对于LLaMA-3.1-8B和Mistral-7B，均值AUC的提升分别达到了45.9%和46.6%。这些提升也能够迁移到三个未见的MuAViC语言数据集。此外，我们在两个基于真实配音行的下游任务上进行了评估。在配音行重新排序任务中，ATR-9B的表现优于最佳的唇读基线，提升幅度为52.0%；而在剧本到剪辑的分配任务中，ATR-9B相较于最佳的唇读基线提升了17.7%。
cs.CV / 46 / 2610.00848
Geometric Similarity in VLM Low-Level Vision Representations
视觉语言模型低级视觉表征中的几何相似性
Abstract
Vision-language models (VLMs) have emerged as powerful candidates for universal vision backbones, with representative architectures including autoregressive (AR) models and diffusion transformers (DiTs). Yet, adapting them efficiently for all-in-one low-level image restoration remains a challenge. Crucially, the field lacks an understanding of how VLMs organize hidden-layer representations and whether these structurally distinct paradigms share a common geometric organization for pixel-level perception. Such shared organization is a prerequisite for building highly transferable, unified restoration VLMs and adapters. In this paper, we systematically investigate representational similarity across 24 low-level tasks spanning 5 categories. We propose GeoSim, a unified four-level framework that analyzes task-conditioned representations from global similarity, local geometry, sparse feature decomposition, and topological verification perspectives. Our formulation applies to the analysis of hidden states in AR models and feature maps in DiTs across same- and cross-task/model settings. Our results reveal the organizing principles of low-level visual representations while exposing their limits in cross-task and cross-model agreement. Ultimately, GeoSim provides an interpretability lens for probing latent transferability in low-level vision and diagnosing model limitations in task- or model-specific scenarios.
Chinese Translation
视觉语言模型（VLMs）作为通用视觉骨干网络的有力候选者逐渐兴起，代表性架构包括自回归（AR）模型和扩散变换器（Diffusion Transformers，DiTs）。然而，如何高效地将它们适配于一体化的低级图像恢复任务仍然是一个挑战。关键在于，目前领域内尚缺乏对VLMs如何组织隐藏层表征的理解，以及这些结构上截然不同的范式是否共享用于像素级感知的共同几何组织。这种共享的组织结构是构建高度可迁移的统一恢复VLM及其适配器的前提。本文系统地研究了涵盖5大类共24个低级任务的表征相似性。我们提出了GeoSim，一种统一的四层框架，从全局相似性、局部几何、稀疏特征分解和拓扑验证四个视角分析任务条件下的表征。该方法适用于分析AR模型的隐藏状态和DiTs的特征图，涵盖同任务/跨任务及同模型/跨模型的设置。研究结果揭示了低级视觉表征的组织原则，同时暴露了其在跨任务和跨模型一致性方面的局限性。最终，GeoSim为探究低级视觉中潜在迁移性提供了可解释性视角，并为诊断特定任务或模型场景下的模型局限性提供了工具。
cs.CV / 47 / 2610.00851
SmoothOperator: Enhancing Representations for Fine-grained Open-set Recognition via Modulated Label Smoothing
SmoothOperator：通过调制标签平滑增强细粒度开放集识别的表征
Abstract
Open Set Recognition (OSR) aims to enable models to accurately classify known classes while rejecting samples from unseen classes. A key challenge in OSR lies in the inability to model the unbounded distribution of unknown classes during training, often leading to the misclassification of samples from these classes. Rather than modeling unknowns, recent work shapes the feature space so that known classes are compact and well separated, and spherical representation learning methods have achieved strong results this way. Label smoothing has been identified as one of the key drivers of this success, yet it applies the same coefficient to every training sample, regardless of how well each sample is already embedded. We show that the spherical representation learning objectives used in OSR share a single alignment--uniformity structure in which labels enter only through the alignment term. Label smoothing therefore acts as an alignment dial, and a fixed coefficient sets this dial to the same value for every sample. We propose a plug-in, SmoothOperator (SmoothOP), which sets the smoothing coefficient of each sample from its \textbf{prominence}, an embedding-space signal measuring how clearly the sample's own class stands out against its strongest competing class. Our method integrates into four existing spherical representation learning methods at minimal training overhead. SmoothOP assigns strong smoothing to samples with high prominence, which reduces their alignment and relaxes their pull. On the Semantic Shift Benchmark, SmoothOP-augmented variants generally outperform their base objectives across datasets, degrees of semantic shift, and OSR post-processors, with gains of up to 4.7\% in AUROC, OSCR, and closed-set accuracy.
Chinese Translation
开放集识别（OSR）旨在使模型能够准确分类已知类别，同时拒绝来自未见类别的样本。OSR中的一个关键挑战在于训练过程中无法对未知类别的无界分布进行建模，这常常导致对这些类别样本的误分类。最近的研究并没有对未知类别进行建模，而是通过塑造特征空间，使已知类别紧凑且良好分离，球形表征学习方法因此取得了良好的效果。标签平滑被认为是这一成功的关键驱动因素之一，但它对每个训练样本应用相同的系数，而不考虑每个样本的嵌入效果。我们表明，OSR中使用的球形表征学习目标共享一个单一的对齐-均匀性结构，其中标签仅通过对齐项进入。因此，标签平滑充当了一个对齐调节器，而固定系数将这一调节器设置为每个样本相同的值。我们提出了一种插件SmoothOperator（SmoothOP），它根据每个样本的 extbf{显著性}设置平滑系数，显著性是一个嵌入空间信号，衡量样本自身类别与其最强竞争类别的清晰度。我们的方法以最小的训练开销集成到四种现有的球形表征学习方法中。SmoothOP对显著性高的样本施加强平滑，这减少了它们的对齐并放松了它们的拉力。在语义转移基准测试中，增强的SmoothOP变体在各个数据集、语义转移程度和OSR后处理器上通常优于其基础目标，AUROC、OSCR和闭集准确率的提升幅度高达4.7%。
cs.CV / 48 / 2610.00855
Lang3DSeg: Annotation-Free Open-Vocabulary 3D Segmentation with Point Transformers
Lang3DSeg：无注释开放词汇的点变换器3D分割
Abstract
Accurate 3D semantic perception is critical for safe autonomous navigation. However, supervised LiDAR segmentation remains tied to closed taxonomies and to the cost of point-wise manual annotation. Open-vocabulary methods avoid that cost by projecting the output of 2D vision-language models onto LiDAR and distilling it into a 3D network. These methods rely almost exclusively on voxel-based sparse convolutions, and point transformers have so far been limited to indoor environments, where 3D data is dense and bounded. We present Lang3DSeg, which establishes a point transformer as the backbone for annotation-free open-vocabulary segmentation of outdoor 3D LiDAR, and is trained from scratch without geometric pre-training. This training paradigm necessitates addressing the inherent noise in 2D-to-3D label projections; specifically, naive projection often suffers from depth ambiguity, where points behind an object are erroneously assigned its semantic label. We therefore composite masks using an explicit class-priority rule and truncate each projected instance at the first gap in its depth distribution, correcting the projection error directly rather than averaging it over registered sequences. Lang3DSeg achieves 52.8% mIoU on nuScenes validation and 41.4% on SemanticKITTI, the highest among published annotation-free methods on both benchmarks. Every 3D semantic segmentation is on a single LiDAR sweep, and inference operates in real-time without running vision-language models.
Chinese Translation
准确的3D语义感知对于安全的自主导航至关重要。然而，监督式激光雷达分割仍然依赖于封闭的分类法以及逐点手动标注的成本。开放词汇方法通过将2D视觉-语言模型的输出投影到激光雷达上，并将其提炼为3D网络，从而避免了这一成本。这些方法几乎完全依赖于基于体素的稀疏卷积，而点变换器迄今为止仅限于室内环境，在这些环境中，3D数据是密集且有限的。我们提出了Lang3DSeg，它将点变换器作为无注释开放词汇的户外3D激光雷达分割的骨干，并从零开始训练，而无需几何预训练。这种训练范式需要解决2D到3D标签投影中的固有噪声；具体而言，简单的投影往往会遭遇深度模糊问题，即物体后面的点错误地被分配为其语义标签。因此，我们使用显式的类别优先规则合成掩膜，并在其深度分布的第一个间隙处截断每个投影实例，直接纠正投影错误，而不是在注册序列中对其进行平均。Lang3DSeg在nuScenes验证集上达到了52.8%的mIoU，在SemanticKITTI上达到了41.4%，在这两个基准上都是已发布的无注释方法中最高的。每个3D语义分割都是在单次激光雷达扫描上进行的，推理实时进行，无需运行视觉-语言模型。
cs.CV / 49 / 2610.00859
CtrlWAM: Controllable World Action Models with Aligned Intent and Foresight
CtrlWAM：具有对齐意图和前瞻性的可控世界动作模型
Abstract
World action models (WAMs) jointly predict actions (intent) and visual future (foresight). Standard training adds noise to recorded actions and video simultaneously, but such training paradigms introduce a mismatch: perturbed actions imply counterfactual future visual, while the noised video remains tied to the GT recording. In low-noise regime, the scene geometry and even the dynamic behavior remain clearly visible from the noisy future frames despite the added noise. We present CtrlWAM, which executes perturbed actions in a simulator and pairs them with their noised visual consequences for joint WAM learning. To accommodate the different denoising requirements of video and actions, we introduce warped video--action noise schedules that aim to keep visual layout responsive as action predictions evolve. We further extend the action interface from ego-only control to a variable number of agent streams, allowing a unified model to represent predicted or commanded futures for multiple agents. Driving experiments show more accurate action forecasts, closer agreement between generated video and actions, and better following of supplied commands; robotics experiments show stronger motion fidelity and controllability. Matched controls support the benefit of off-path renders for command following and manipulation fidelity. Together, these findings contribute to a more controllable world action model. Project page: https://ctrl-wam.github.io/
Chinese Translation
世界动作模型（WAMs）共同预测动作（意图）和视觉未来（前瞻性）。标准训练同时对记录的动作和视频添加噪声，但这种训练范式引入了不匹配：扰动的动作暗示了反事实的未来视觉，而带噪声的视频仍然与真实记录相联系。在低噪声环境下，尽管添加了噪声，场景几何形状甚至动态行为在噪声未来帧中仍然清晰可见。我们提出了CtrlWAM，它在模拟器中执行扰动动作，并将其与相应的带噪视觉后果配对，以实现联合WAM学习。为了适应视频和动作的不同去噪需求，我们引入了扭曲的视频-动作噪声调度，旨在随着动作预测的演变保持视觉布局的响应性。我们进一步将动作接口从仅限自我控制扩展到可变数量的代理流，允许统一模型表示多个代理的预测或指令未来。驱动实验显示出更准确的动作预测，生成的视频与动作之间的更紧密一致性，以及更好地遵循提供的指令；机器人实验显示出更强的运动保真度和可控性。匹配控制支持了偏离路径渲染在指令跟随和操作保真度方面的好处。综合这些发现，有助于构建一个更可控的世界动作模型。项目页面：https://ctrl-wam.github.io/
cs.CV / 50 / 2610.00881
Machine Translation for Sign Languages
手语的机器翻译
Abstract
Sign language machine translation has progressed substantially over the past decade, evolving from isolated sign recognition to end-to-end translation systems. Advances in pose estimation, transformer architectures, and large-scale dataset collection have driven progress, yet challenges remain. Datasets are limited compared to spoken-language resources; evaluation metrics inadequately capture the linguistic quality of output; and models must capture the simultaneous, multi-layered, and three-dimensional structure of sign languages. This manuscript provides a comprehensive review that seeks to balance technical challenges with stakeholder considerations. We examine the linguistic properties that make sign languages computationally unique, trace the evolution of recognition, translation, and production systems, and analyze ongoing technical challenges. Crucially, we address ethical considerations around data governance, community involvement, and appropriate use. Drawing on interdisciplinary perspectives spanning computer vision, sign language linguistics, and deaf studies, our analysis emphasizes that continued progress requires sustained collaboration across these fields and with deaf communities.
Chinese Translation
手语机器翻译在过去十年中取得了显著进展，从孤立的手势识别发展到端到端的翻译系统。姿态估计、变换器架构和大规模数据集收集的进步推动了这一进展，但仍然面临挑战。与口语资源相比，数据集相对有限；评估指标不足以充分反映输出的语言质量；模型必须捕捉手语的同时性、多层次性和三维结构。本文提供了一项全面的回顾，旨在平衡技术挑战与利益相关者的考虑。我们考察了使手语在计算上独特的语言特性，追踪了识别、翻译和生成系统的演变，并分析了持续存在的技术挑战。至关重要的是，我们讨论了关于数据治理、社区参与和适当使用的伦理考量。我们的分析借鉴了跨学科的视角，涵盖计算机视觉、手语语言学和聋人研究，强调持续进展需要在这些领域及与聋人社区之间保持持续的合作。
cs.CV / 51 / 2610.00922
EyeTAG: Eye Trajectory-Aware Gaze Estimation
EyeTAG：基于眼动轨迹的注视估计
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.
Chinese Translation
在自然头眼运动下的注视估计是从驾驶员监控到人机交互等应用的基础。单帧方法独立预测每一帧，因此连续输出会出现抖动。多帧方法减少了这种抖动，但它们在外观特征中隐式学习运动，因此注视轨迹从未成为显式变量。我们提出了EyeTAG（基于眼动轨迹的注视估计），这是一个围绕显式一阶注视先验构建的因果多帧框架：在每一步中，它区分自身最近的预测，并将结果轨迹反馈为紧凑的运动标记。由于差分在注视空间中是平移不变的，这个标记携带的是与受试者无关的运动，而不是个人注视偏移。面部和眼部流提供视觉证据，通过交叉注意力和因果Transformer解码器进行融合。EyeTAG在Gaze360上将平均角度误差减少约1.0$^ ext{°}$，并在EVE上与最强基线表现相当（2.56$^ ext{°}$对比2.58$^ ext{°}$）。在模型内部的消融实验中，保持编码器和其余架构不变，仅改变注视历史，表明差分公式而非单纯的时间上下文消除了即使在绝对注视历史先验下仍然存在的系统性扫视偏差。我们的代码可在https://github.com/peter8366/EyeTAG获取。
cs.CV / 52 / 2610.00930
Joint Branch-Space Transform Coding for Diffusion Activation Quantization with Classifier-Free Guidance
用于扩散激活量化的联合分支空间变换编码与无分类器引导
Abstract
Post-training quantization for diffusion models increasingly exploits timestep, feature, and layer structure. While recent work has begun incorporating CFG structure into diffusion quantization, activation quantization still operates independently across conditional and unconditional coordinates, leaving cross-activation structure unexploited. We show that matched CFG activations form a strongly correlated two-dimensional source and that, under a fixed bit budget, the choice of branch coding basis materially affects quantization fidelity. Motivated by this observation, we introduce branch-space transform coding, which rotates matched CFG branches via an offline derived 2x2 orthogonal matrix, requiring minimal modifications to model parameters or the quantization pipeline. We further derive the Guidance-Correlation Branch Transform (GCBT), which jointly incorporates the CFG guidance direction and cross-branch second moments. Under an equal-rate quantization-noise surrogate, GCBT admits a closed-form per-layer solution without gradient optimization or angle search. Applied on top of existing diffusion PTQ methods, GCBT yields statistically significant fidelity gains in most evaluated comparisons with no statistically significant degradation, while leaving the underlying host quantization pipeline unchanged.
Chinese Translation
扩散模型的后训练量化越来越多地利用时间步、特征和层结构。尽管近期的研究已开始将无分类器引导（CFG）结构纳入扩散量化，但激活量化仍然在条件和无条件坐标之间独立操作，未能充分利用交叉激活结构。我们表明，匹配的CFG激活形成了一个强相关的二维源，并且在固定的比特预算下，分支编码基的选择对量化保真度有实质性影响。基于这一观察，我们引入了分支空间变换编码，该方法通过离线推导的2x2正交矩阵旋转匹配的CFG分支，几乎不需要对模型参数或量化流程进行修改。我们进一步推导了引导-相关分支变换（GCBT），它联合考虑了CFG引导方向和交叉分支的二阶矩。在等速率量化噪声替代下，GCBT提供了每层的封闭形式解，无需梯度优化或角度搜索。在现有的扩散后训练量化（PTQ）方法上应用GCBT，在大多数评估比较中产生了统计上显著的保真度提升，且没有统计上显著的降级，同时保持了基础主机量化流程不变。
cs.CV / 53 / 2610.00952
A Matched-Budget Audit Framework for Recaptioned Image-Text Supervision Distributions
用于重新标注图像-文本监督分布的匹配预算审计框架
Abstract
Recaptioned image-text corpora are now standard for text-to-image (T2I) training, with vision--language model (VLM) captioners replacing sparse alt-text by dense descriptions. A recaptioned corpus is a supervision distribution induced by a documented captioning policy ($\pi$), captioner ($V_c$), and source corpus ($C$). Length-correlated proxies miss caption-register artifacts and downstream T2I benchmarks entangle the corpus with training choices, so this distribution is hard to audit at corpus scale. We introduce a reusable matched-budget audit framework for recaptioned supervision distributions $D_{\pi,V_c,C}$: at a fixed text budget of $B = 64$ it reports a five-axis profile spanning prompt-side coverage, image-conditioned faithfulness, and caption-surface health, with claimed controllable basic units (CBU) as the common claim unit. We instantiate the framework on seven paired comparisons over five public source corpora. Across the four cross-corpus pairs, the released surface raises supported CBU per caption by $+3.39$ to $+6.36$ under both Qwen and Gemma Judges, and on CC12M the same framework exposes a long-vs-dense frontier that is consistent across both judges and four budgets. We release the audited multi-source recap corpus ($\approx$ 490M) together with the audit-artifact bundle.
Chinese Translation
重新标注的图像-文本语料库现已成为文本到图像（T2I）训练的标准，视觉-语言模型（VLM）标注器用密集描述替代了稀疏的替代文本。重新标注的语料库是由文档化的标注政策（$ ext{π}$）、标注器（$V_c$）和源语料库（$C$）引发的监督分布。与长度相关的代理缺失了标注注册伪影，而下游T2I基准将语料库与训练选择纠缠在一起，因此在语料库规模上审计这种分布是困难的。我们引入了一个可重用的匹配预算审计框架，用于重新标注的监督分布 $D_{ ext{π},V_c,C}$：在固定文本预算 $B = 64$ 下，它报告了一个涵盖提示侧覆盖、图像条件忠实度和标注表面健康的五轴轮廓，以声称的可控基本单元（CBU）作为共同的声称单位。我们在五个公共源语料库的七个配对比较上实例化了该框架。在四个跨语料库对中，发布的表面在Qwen和Gemma评审下每个标注支持的CBU提高了 $+3.39$ 到 $+6.36$，而在CC12M上，相同的框架揭示了一个在两个评审和四个预算中一致的长对密集的边界。我们发布了经过审计的多源重新标注语料库（约490M）以及审计伪影包。
cs.CV / 54 / 2610.00953
Two Clocks in Diffusion MLLMs: When Answers Stabilize Before Rationales Unfold
扩散 MLLMs 中的两个时钟：当答案在推理展开之前稳定时
Abstract
An answer candidate in a masked diffusion MLLM can stabilize while its rationale is still unfolding. We distinguish retrospective stabilization of the logged candidate from token commitment, and examine these two clocks relative to rationale generation. Analyzing our results across three visual question-answering benchmarks, we find that 89.4-98.1% of the rationale-side canvas remains unwritten at stabilization in single-block, EOS-suppressed LaViDa runs. On V*Bench, reducing block length from 128 to 8 changes this fraction from 89.4% to 1.7%, together with answer coverage and the eligible observation window. Under EOS-enabled prompting, direct instructions improve Nemotron's overall accuracy by 15.0 and 19.5 percentage points on M3CoT and ScienceQA, but reduce LaViDa/V*Bench accuracy by 11.0 points. A symmetric decomposition associates the larger absolute component of each change with coverage rather than conditional accuracy. Matched-canvas image ablations measure visual sensitivity alongside answer stabilization, separating the two temporal readouts. Together, these measurements distinguish answer stabilization, rationale unfolding, and visual sensitivity, and identify coverage as the larger component of the prompting differences.
Chinese Translation
在一个掩蔽的扩散 MLLM 中，答案候选可以在其推理仍在展开时稳定。我们区分了记录候选的回顾性稳定性与令牌承诺，并考察这两个时钟相对于推理生成的关系。在对三个视觉问答基准的结果分析中，我们发现，在单块、抑制 EOS 的 LaViDa 运行中，稳定时推理侧画布中有 89.4% 至 98.1% 的部分仍未书写。在 V*Bench 上，将块长度从 128 减少到 8，使得这一比例从 89.4% 降至 1.7%，同时影响答案覆盖率和合格观察窗口。在启用 EOS 的提示下，直接指令使 Nemotron 在 M3CoT 和 ScienceQA 上的整体准确率提高了 15.0 和 19.5 个百分点，但 LaViDa/V*Bench 的准确率降低了 11.0 个百分点。对称分解将每个变化的较大绝对成分与覆盖率而非条件准确性相关联。匹配画布图像消融测量视觉敏感性与答案稳定性，并将这两个时间读数分开。综合这些测量，我们区分了答案稳定性、推理展开和视觉敏感性，并确定覆盖率是提示差异中较大的成分。
cs.CV / 55 / 2610.00960
Video-Index: A Curated Meta-Benchmark for Video Understanding
视频索引：一个策划的元基准用于视频理解
Abstract
A video benchmark should reward the capability it claims to measure, yet models can exploit answer options, question text, or partial visual evidence. We introduce the attack pyramid, five levels of shortcut attacks with increasing access to each item, and audit 115 video benchmarks with it. On 35 benchmarks, attackers that never see a frame approach full-video accuracy. On 51 benchmarks with temporal probes, shuffled frames keep a median 96% of full-video accuracy. Near-duplicate questions make up at least half the items in 63 benchmarks. We screen 505,518 question-answer pairs from 112 of them into an audited pool. Agents turn evaluation requests into specifications, and a deterministic selector with a red-team gate composes reproducible benchmarks. We release Video-Index, the 210 hardest verified items under these attacks in each of four capability groups, 840 items from 76 sources. With the same fixed input, Claude Opus 5 outscores every open-source model by over 37 percentage points, and agent tools add about 20 more, yet all systems leave room to improve efficiency and accuracy. Blog: https://www.enxinsong.com/blog/video-index/ GitHub: https://github.com/Espere-1119-Song/Video-Index Hugging Face: https://huggingface.co/datasets/Video-Index/Video-Index
Chinese Translation
一个视频基准应该奖励其声称测量的能力，然而模型可以利用答案选项、问题文本或部分视觉证据。我们引入了攻击金字塔，定义了五个层次的快捷攻击，随着对每个项目的访问程度增加而递增，并对115个视频基准进行了审计。在35个基准上，从未看到帧的攻击者接近全视频准确率。在51个具有时间探测的基准上，打乱的帧保持了全视频准确率的中位数96%。在63个基准中，近重复问题占至少一半的项目。我们从其中112个基准筛选出505,518个问答对，形成了一个经过审计的池。代理将评估请求转化为规范，而一个带有红队门的确定性选择器则构建了可重复的基准。我们发布了视频索引，在每个四个能力组中，210个在这些攻击下经过验证的最难项目，来自76个来源的840个项目。在相同的固定输入下，Claude Opus 5的得分超过每个开源模型37个百分点以上，而代理工具则增加了约20个百分点，但所有系统仍有提升效率和准确率的空间。博客：https://www.enxinsong.com/blog/video-index/ GitHub：https://github.com/Espere-1119-Song/Video-Index Hugging Face：https://huggingface.co/datasets/Video-Index/Video-Index
cs.CV / 56 / 2610.00970
RelationVGGT: Visual Geometry Transformers for 3D Spatial Relation Segmentation
RelationVGGT：用于3D空间关系分割的视觉几何变换器
Abstract
Recent advances in 3D reconstruction have progressed from per-scene optimization to feed-forward inference, and semantic scene understanding has followed suit -- yet existing methods remain confined to object-centric perception, neglecting spatial relations between objects. We formulate 3D spatial relation segmentation in a feed-forward, pose-free multi-view setting: given a visually specified subject and a relational text query, the model segments the target across views without receiving its category name. To this end, we propose RelationVGGT, a novel feed-forward framework that integrates semantic features from a visual foundation model with geometry-aware representations from a 3D geometry foundation model and leverages a relation transformer for subject-conditioned, cross-view relation prediction -- requiring neither per-scene optimization nor known camera poses. We additionally provide a fully automated annotation pipeline built on ScanNet++ with VLMs and LLMs, enabling scalable training data generation for this new task.
Chinese Translation
近年来，3D重建的进展已从逐场景优化转向前馈推理，语义场景理解也随之而来——然而，现有方法仍然局限于以对象为中心的感知，忽视了对象之间的空间关系。我们在前馈、无姿态的多视图设置中构建了3D空间关系分割：给定一个视觉指定的主题和一个关系文本查询，模型在不同视图中对目标进行分割，而无需接收其类别名称。为此，我们提出了RelationVGGT，这是一种新颖的前馈框架，结合了来自视觉基础模型的语义特征与来自3D几何基础模型的几何感知表示，并利用关系变换器进行主题条件的跨视图关系预测——不需要逐场景优化或已知的相机姿态。此外，我们还提供了一个基于ScanNet++的完全自动化注释管道，结合了视觉语言模型（VLMs）和大语言模型（LLMs），使得这一新任务的可扩展训练数据生成成为可能。
cs.CV / 57 / 2610.00973
Concept Driven Domain Adaptation: Finding an Abstract Needle in a Haystack
概念驱动的领域适应：在干草堆中寻找抽象的针
Abstract
Science teachers frequently search for documentary excerpts not by describing what appears on screen, but by querying the abstract concepts they intend to teach. This use case exposes a limitation of existing language-based video moment retrieval methods, which typically assume that queries describe observable events, whereas instructional search requires retrieving concrete visual phenomena that instantiate an underlying scientific principle. We study this setting as concept-to-example video retrieval, an abstract-needle-in-a-haystack problem where compact curriculum concepts must be grounded in temporally sparse documentary evidence. To bridge this abstraction gap, we propose Concept-Driven Domain Adaptation (CDDA), a three-stage framework for adapting two-tower vision-language models to concept-level retrieval. CDDA treats concepts as intermediate semantic anchors: it first structures the textual embedding space with textbook and teacher-handbook example-concept pairs, then transfers this concept-aware geometry to documentary visuals under a frozen visual encoder, and finally jointly adapts both encoders with sparse visual concept supervision. From a geometric perspective, this staged alignment reduces text-concept and vision-concept angular gaps, thereby encouraging concept-level adaptation while preserving the pretrained model's concrete image description alignment. On a curated middle-school physics retrieval benchmark, CDDA achieves stronger pedagogically oriented concept retrieval than several competitive multimodal baselines, including Qwen3-VL-Embedding-2B, while maintaining concrete image-text matching after adaptation.
Chinese Translation
科学教师通常不是通过描述屏幕上出现的内容来搜索文献摘录，而是通过查询他们打算教授的抽象概念。这一用例揭示了现有基于语言的视频时刻检索方法的局限性，这些方法通常假设查询描述的是可观察的事件，而教学搜索则需要检索具体的视觉现象，以体现潜在的科学原理。我们将这一设置视为概念到示例的视频检索问题，即在干草堆中寻找抽象针的问题，其中紧凑的课程概念必须在时间稀疏的文献证据中扎根。为了弥补这一抽象差距，我们提出了概念驱动的领域适应（Concept-Driven Domain Adaptation, CDDA），这是一个三阶段框架，用于将双塔视觉-语言模型适应于概念级检索。CDDA将概念视为中间语义锚点：首先，它通过教科书和教师手册中的示例-概念对来构建文本嵌入空间，然后在冻结的视觉编码器下将这种概念感知的几何结构转移到文献视觉中，最后在稀疏的视觉概念监督下共同适应两个编码器。从几何的角度来看，这种分阶段的对齐减少了文本-概念和视觉-概念之间的角度差距，从而鼓励概念级适应，同时保持预训练模型的具体图像描述对齐。在一个精心策划的中学物理检索基准上，CDDA在教学导向的概念检索方面优于多个竞争的多模态基线，包括Qwen3-VL-Embedding-2B，同时在适应后保持具体的图像-文本匹配。
cs.CV / 58 / 2610.00994
VIEScore2: Unified Image Evaluation with Spatially Grounded Explanations
VIEScore2：具有空间基础解释的统一图像评估
Abstract
Existing synthetic image evaluators typically provide only a scalar quality score and do not identify the image regions that support it. We introduce VIEScore2, a unified evaluator for image generation and editing tasks with optional conditioning images. VIEScore2 represents an image as an N x N grid and jointly predicts quality scores and defect locations in a single model pass. Its text-native grid representation provides a common interface for heterogeneous spatial supervision and enables directly verifiable post-training objectives. We train on 38K examples spanning score-only, localization-only, and joint supervision across generation and editing tasks. Starting from supervised fine-tuning, we further apply GRPO to improve defect localization using rewards that combine cell-level Dice overlap, score accuracy, and output-format validity. A parameter-free parser converts the structured predictions into readable explanations. On the primary suite, VIEScore2 achieves an overall-score SRCC of 0.601, compared with 0.491 for Gemini-3-Flash, the strongest zero-shot general-purpose VLM baseline under matched inputs. For defect localization, VIEScore2 outperforms both general-purpose VLMs and specialized spatial evaluators on three of six benchmarks in per-image grid IoU and ranks among the top three on five, including datasets beyond its training sources.
Chinese Translation
现有的合成图像评估器通常仅提供一个标量质量分数，而不识别支持该分数的图像区域。我们提出了VIEScore2，这是一种用于图像生成和编辑任务的统一评估器，支持可选的条件图像。VIEScore2将图像表示为N x N网格，并在单次模型传递中联合预测质量分数和缺陷位置。其文本原生网格表示为异构空间监督提供了一个共同接口，并使得直接可验证的后训练目标成为可能。我们在38K个示例上进行训练，涵盖了仅评分、仅定位以及生成和编辑任务的联合监督。从监督微调开始，我们进一步应用GRPO来利用结合单元级Dice重叠、分数准确性和输出格式有效性的奖励来改善缺陷定位。在结构化预测到可读解释的过程中，使用无参数解析器进行转换。在主要评估套件中，VIEScore2的整体分数SRCC达到了0.601，而在匹配输入下，最强的零-shot通用视觉语言模型基线Gemini-3-Flash的SRCC为0.491。在缺陷定位方面，VIEScore2在六个基准中的三项上在每图像网格IoU上超越了通用视觉语言模型和专门的空间评估器，并在五个基准中排名前三，包括超出其训练来源的数据集。
cs.CV / 59 / 2610.01012
Watch Your Speech: Text-aware Video-to-Speech Synthesis with Textual Conditioning
注意你的语音：基于文本的视频到语音合成与文本条件的结合
Abstract
Video-to-speech synthesis aims to generate natural-sounding speech from silent talking-face videos while ensuring phonetic accuracy. A fundamental challenge in this task is the inherent one-to-many mapping problem, where visual dynamics often lack sufficient information to uniquely determine the corresponding utterance. To address this, we propose Watch Your Speech (WYS), a video-to-speech synthesis framework that incorporates textual conditioning as an explicit linguistic cue to mitigate visual ambiguity. Our framework features an attention-based embedding fusion module that synergistically integrates textual context with video sequences, coupled with a conditional flow matching objective for high-fidelity speech generation. Extensive experiments on the LRS2 and LRS3 datasets demonstrate that WYS achieves superior performance, establishing new state-of-the-art results in audio-visual synchronization (LSE-C/D) while maintaining highly competitive textual accuracy (WER). Subjective evaluations further confirm that our model generates speech with near-human naturalness, validating the effectiveness of textual conditioning in content-controlled video-to-speech synthesis. Project page: https://github.com/gunwoo5034/Watch-your-Speech
Chinese Translation
视频到语音合成旨在从无声的说话者视频中生成自然的语音，同时确保音素的准确性。该任务的一个基本挑战是固有的一对多映射问题，视觉动态往往缺乏足够的信息来唯一确定相应的发声内容。为了解决这个问题，我们提出了“注意你的语音”（Watch Your Speech，WYS），这是一个视频到语音合成框架，结合了文本条件作为明确的语言线索，以减轻视觉歧义。我们的框架具有一个基于注意力的嵌入融合模块，协同整合文本上下文与视频序列，并结合条件流匹配目标以实现高保真语音生成。在LRS2和LRS3数据集上的广泛实验表明，WYS在音频-视觉同步（LSE-C/D）方面实现了卓越的性能，建立了新的最先进的结果，同时保持了高度竞争的文本准确性（WER）。主观评估进一步确认我们的模型生成的语音接近人类的自然性，验证了文本条件在内容控制的视频到语音合成中的有效性。项目页面：https://github.com/gunwoo5034/Watch-your-Speech
cs.CV / 60 / 2610.01013
VASC: Value-Aware Sparse Attention with Cross-Layer Memory for Efficient 3D Reconstruction
VASC：具有跨层记忆的价值感知稀疏注意力用于高效的3D重建
Abstract
Feed-forward 3D vision models such as VGGT have achieved remarkable progress, unifying camera estimation and dense scene reconstruction in a single pass. However, their quadratic global attention makes long image sequences expensive, while existing sparse methods may favor highly attended yet value-redundant regions. To address these limitations, we introduce VASC, a training-free sparse attention method combining value-aware block selection and execution-aware cross-layer memory. Our value-aware block selection integrates pooled query--key relevance with neighboring value contrast, reducing redundancy while preserving query-relevant and distinctive content. Cross-layer memory tracks unserved demand across layers and updates this state according to actual execution, enabling previously underserved blocks to compete under a fixed computation budget. Experiments on 7Scenes and NeuralRGB-D with VGGT and $\pi^3$ demonstrate improved pose estimation and reconstruction quality compared with FasterVGGT, together with up to $2.29\times$ faster inference than dense VGGT. Code is available at https://github.com/kosakayamahoo-design/VASC.
Chinese Translation
前馈3D视觉模型如VGGT在单次处理过程中实现了相机估计和密集场景重建的显著进展。然而，它们的二次全局注意力使得长图像序列的计算成本高昂，而现有的稀疏方法可能偏向于高度关注但价值冗余的区域。为了解决这些局限性，我们提出了VASC，这是一种无训练的稀疏注意力方法，结合了价值感知的块选择和执行感知的跨层记忆。我们的价值感知块选择将池化的查询-关键相关性与邻近值对比相结合，减少冗余，同时保留与查询相关和独特的内容。跨层记忆跟踪各层之间未满足的需求，并根据实际执行情况更新该状态，使得之前未服务的块能够在固定的计算预算下进行竞争。在7Scenes和NeuralRGB-D上与VGGT和$ ext{π}^3$的实验表明，与FasterVGGT相比，姿态估计和重建质量得到了改善，同时推理速度比密集VGGT快了多达$2.29 imes$。代码可在https://github.com/kosakayamahoo-design/VASC获取。
cs.CV / 61 / 2610.01019
FutureWorlds: Learning Robotic World Models from Alternative Futures
未来世界：从替代未来中学习机器人世界模型
Abstract
Robotic world models predict action-conditioned future scenes, providing a foundation for understanding action outcomes. However, turning alternative predictions into useful learning signals remains challenging: similar candidates limit informative quality comparisons, while diverging trajectories require persistent maintenance of their individual histories. We introduce FutureWorlds, a framework that unifies candidate construction, history maintenance, and learning from relative quality. Built on a multimodal discrete autoregressive model, FutureWorlds uses diverse beam search during reinforcement learning to construct candidate futures that balance confidence and diversity. Candidate-specific bounded memory preserves scene states and ensures that generation and policy scoring use matching histories. We further propose MemSPO (Memory-Conditioned Search-Guided Policy Optimization), which converts video trajectory rewards into group-relative advantages to optimize the world model. On RT-1, BridgeV2, and RoboCasa, FutureWorlds reduces LPIPS for 32-frame predictions by 14.78%, 20.84%, and 9.12%, respectively, relative to the strongest baseline on each dataset. Under fixed evaluation configurations, only 200 MemSPO updates further improve generation quality and support continued prediction beyond the training horizon. Memory ablations, decoding sensitivity analysis, and optical-flow evaluation show that these gains extend beyond visual quality to more accurate motion prediction and more consistent object states. Project page and code: https://github.com/Alexander-wu/FutureWorlds.
Chinese Translation
机器人世界模型预测基于动作的未来场景，为理解动作结果提供了基础。然而，将替代预测转化为有用的学习信号仍然具有挑战性：相似的候选者限制了信息质量比较，而不同的轨迹则需要持续维护各自的历史。我们提出了FutureWorlds，一个统一候选构建、历史维护和相对质量学习的框架。FutureWorlds基于多模态离散自回归模型，在强化学习过程中使用多样化的束搜索来构建平衡置信度和多样性的候选未来。候选特定的有界记忆保留场景状态，确保生成和策略评分使用匹配的历史。我们进一步提出了MemSPO（记忆条件搜索引导策略优化），它将视频轨迹奖励转换为组相对优势，以优化世界模型。在RT-1、BridgeV2和RoboCasa上，FutureWorlds相对于每个数据集上最强基线，分别将32帧预测的LPIPS降低了14.78%、20.84%和9.12%。在固定评估配置下，仅200次MemSPO更新进一步提高了生成质量，并支持超出训练范围的持续预测。记忆消融、解码敏感性分析和光流评估表明，这些提升超越了视觉质量，带来了更准确的运动预测和更一致的物体状态。项目页面和代码： https://github.com/Alexander-wu/FutureWorlds。
cs.CV / 62 / 2610.01022
Towards Automatic Video Annotation with ASH: Zero-Shot Open-Vocabulary Multi-Object Tracking and Segmentation
基于ASH的自动视频标注：零-shot开放词汇多目标跟踪与分割
Abstract
Memory-attention-based Video Instance Segmentation (VIS) methods have demonstrated strong zero-shot tracking capability, yet their substantial memory requirements confine them to short video clips and their single-prompt inference design makes multi-category open-vocabulary tracking computationally prohibitive. This work introduces two contributions toward fully automated tracking annotation of arbitrary video. The Generalized Presence Token (GPT) reformulates SAM3's inference pipeline to process N text prompts simultaneously via virtual prompt batching, reducing image encoding cost from O(N) to O(1) with no modifications to any learned component. The Annotation and Segmentation Handler (ASH) extends any memory-attention VIS tracker to sequences of arbitrary length through overlapping temporal chunks with IoU-based inter-chunk identity matching, requiring no dataset-specific training. Instantiated on SAM3, the resulting pipeline -- SAM3-ASH -- achieves state-of-the-art HOTA on MOTS20 under fully zero-shot conditions and remains competitive with trained specialists across seven additional benchmarks, while peak GPU memory consumption stays below 25 GB, establishing a practical baseline for scalable, training-free automated video annotation.
Chinese Translation
基于记忆注意力的视频实例分割（VIS）方法展示了强大的零-shot跟踪能力，但其巨大的内存需求限制了其在短视频片段中的应用，并且其单提示推理设计使得多类别开放词汇跟踪在计算上不可行。本研究提出了两个贡献，以实现任意视频的完全自动化跟踪标注。广义存在令牌（Generalized Presence Token, GPT）重新构建了SAM3的推理流程，通过虚拟提示批处理同时处理N个文本提示，将图像编码成本从O(N)降低到O(1)，且无需对任何学习组件进行修改。标注与分割处理器（Annotation and Segmentation Handler, ASH）通过基于IoU的块间身份匹配，将任何记忆注意力VIS跟踪器扩展到任意长度的序列，且无需特定数据集的训练。在SAM3上实例化的结果管道--SAM3-ASH--在完全零-shot条件下在MOTS20上实现了最先进的HOTA，并在七个额外基准测试中与训练专家保持竞争力，同时峰值GPU内存消耗保持在25 GB以下，为可扩展的无训练自动视频标注建立了一个实用的基线。
cs.CV / 63 / 2610.01039
Bootstrapping Video Interaction Generation with Synthetic State Transitions
通过合成状态转变引导视频交互生成
Abstract
While recent video generative models can synthesize high-fidelity videos, they struggle to portray plausible physical interactions and the resulting state transitions, a critical bottleneck for applications in robotics and VR/AR. To address this, we introduce a framework to generate a scalable synthetic dataset of controllable interactions. Our pipeline leverages a structured taxonomy and state-of-the-art image editing models to create explicit `start' and `end' state images, which serve as visual anchors for the interaction. To generate a seamless video utilizing these anchors, we propose State-Guided Sampling (SGS), a novel sampling technique that mitigates artifacts common in naive conditional generation. Furthermore, we develop and validate a new automated evaluation system that aligns with human judgments to ensure data quality. Experiments show that fine-tuning a base model on our dataset significantly enhances its ability to generate plausible interactions.
Chinese Translation
尽管近期的视频生成模型能够合成高保真视频，但它们在表现合理的物理交互及其所导致的状态转变方面存在困难，这成为机器人技术和虚拟现实/增强现实应用中的一个关键瓶颈。为了解决这一问题，我们提出了一个框架，用于生成可控交互的可扩展合成数据集。我们的流程利用结构化分类法和最先进的图像编辑模型，创建明确的“起始”和“结束”状态图像，这些图像作为交互的视觉锚点。为了利用这些锚点生成无缝视频，我们提出了状态引导采样（State-Guided Sampling, SGS），这是一种新颖的采样技术，可以减轻常见于简单条件生成中的伪影。此外，我们开发并验证了一种新的自动评估系统，该系统与人类判断一致，以确保数据质量。实验表明，在我们的数据集上微调基础模型显著增强了其生成合理交互的能力。
cs.CV / 64 / 2610.01052
Towards Subject Consistency over Dynamic Subject Sets in Video Generation
在视频生成中实现动态主题集的主题一致性
Abstract
We argue that as video generation extends to longer durations, subject consistency should be evaluated over \textit{dynamic subject sets}. We therefore introduce \textbf{DynSC-Eval}, an evaluation framework that dynamically tracks eligible subjects throughout their visible lifespans and measures local continuity and global identity preservation using six complementary object-level metrics, with explicit detection of inconsistency events. To validate its effectiveness, we design synthetic experiments that actively inject inconsistency events, demonstrating both the sensitivity of DynSC-Eval and the limitations of existing metrics. Evaluations of diverse models on 5s, 15s, and 60s video generation further reveal substantial subject consistency differences that are obscured by conventional metrics. Beyond evaluation, we construct rewards from DynSC-Eval and apply DiffusionNFT post-training in an autonomous-driving testbed. On 5s generation, our approach reduces the six inconsistency metrics by an average of 13.82\% for Wan-2.1-1.3B and 5.66\% for SANA-2B, with improvements also observed on the I2V model ReSim. Qualitative comparisons further demonstrate the effectiveness of our method. We then extend generation to 10s and 30s through curriculum learning and show that consistency optimization remains effective while largely preserving other capabilities.
Chinese Translation
我们认为，随着视频生成时间的延长，主题一致性应在 extit{动态主题集}上进行评估。因此，我们提出了 extbf{DynSC-Eval}，一个评估框架，动态跟踪在可见生命周期内的合格主题，并使用六个互补的对象级指标来测量局部连续性和全局身份保持，同时明确检测不一致事件。为了验证其有效性，我们设计了合成实验，主动注入不一致事件，展示了DynSC-Eval的敏感性以及现有指标的局限性。在5秒、15秒和60秒的视频生成中对多种模型的评估进一步揭示了被传统指标掩盖的显著主题一致性差异。除了评估，我们还从DynSC-Eval构建奖励，并在自主驾驶测试平台中应用DiffusionNFT后训练。在5秒生成中，我们的方法使Wan-2.1-1.3B的六个不一致指标平均降低了13.82\%，SANA-2B降低了5.66\,同时在I2V模型ReSim上也观察到了改进。定性比较进一步证明了我们方法的有效性。随后，我们通过课程学习将生成扩展到10秒和30秒，并显示一致性优化仍然有效，同时在很大程度上保持了其他能力。
cs.CV / 65 / 2610.01056
HierGF: Hierarchical Gaussian Fields via Geometry-perception Message Passing for Sparse-view 3D Reconstruction
HierGF：通过几何感知消息传递的层次高斯场用于稀疏视图3D重建
Abstract
Sparse view 3D reconstruction is an important and common scenario in multimedia applications, such as augmented reality/virtual reality (AR/VR) content creation, cultural heritage digitization, and certain robotic applications, where only a limited number of randomly captured views may be available. However, sparse views contain only limited 3D information, posing two major challenges:1) too few images are available for matching, making it difficult to build multi-view consistency; 2) insufficient view coverage leads to a lack of information in under-sampled regions, resulting in missing parts of object structure. Existing methods mostly still rely on limited reprojection errors and regularization terms, which are prone to overfitting to a single view and inconsistent appearances across views. In geometrically under-sampled regions, they often rely on heuristic density control, lacking reliable guidance and often resulting in blurring and structural holes.To address these issues, this paper proposes Hierarchical Gaussian Fields (HierGF), which revisits sparse-view reconstruction from a hierarchical geometry-perception perspective and converts limited observations into reliable self-generated supervision beyond fixed priors and heuristic density control. In particular, we transform coarse 3D geometric information and additional 2D generative priors into structured pseudo-supervision through a two-stage geometry-perception backbone network, thereby enhancing multi-view consistency with very few input views. In addition, we introduce a learnable confidence network to guide gradients toward cross-view consistent content, and a geometrically consistent densification module to improve the reconstruction of multi-view alignment and under-sampled regions.
Chinese Translation
稀疏视图3D重建是多媒体应用中的一个重要且常见的场景，例如增强现实/虚拟现实（AR/VR）内容创作、文化遗产数字化以及某些机器人应用，其中可能仅有有限数量的随机捕获视图可用。然而，稀疏视图仅包含有限的3D信息，带来了两个主要挑战：1）可用于匹配的图像过少，难以建立多视图一致性；2）视图覆盖不足导致在欠采样区域缺乏信息，造成物体结构的缺失部分。现有方法大多仍依赖于有限的重投影误差和正则化项，容易对单一视图过拟合，并在不同视图之间出现不一致的外观。在几何欠采样区域，它们通常依赖启发式密度控制，缺乏可靠的指导，常常导致模糊和结构孔洞。为了解决这些问题，本文提出了层次高斯场（HierGF），从层次几何感知的角度重新审视稀疏视图重建，并将有限的观测转化为超越固定先验和启发式密度控制的可靠自生成监督。特别地，我们通过一个两阶段的几何感知主干网络，将粗糙的3D几何信息和额外的2D生成先验转化为结构化的伪监督，从而在非常少的输入视图下增强多视图一致性。此外，我们引入了一个可学习的置信网络，以引导梯度朝向跨视图一致的内容，以及一个几何一致的密集化模块，以改善多视图对齐和欠采样区域的重建。
cs.CV / 66 / 2610.01069
Overcoming Kernel Redundancy for Scaling Logic Gate Networks
克服核冗余以扩展逻辑门网络
Abstract
Differentiable logic gate networks, which operate using only logic gates, have recently attracted attention as an efficient alternative to conventional neural networks. However, despite their efficiency, the scaling behavior of logic gate networks remains underexplored. By contrast, scaling model capacity is a central design principle in deep neural networks and typically leads to improved performance. This discrepancy raises a key question: Can similar scaling benefits also be achieved in logic gate networks? In this work, we focus on width as a primary scaling axis and conduct a systematic analysis of its behavior in logic gate networks. We observe that naive width scaling often introduces redundancy among logic kernels, limiting the effective use of additional kernels and leading to performance saturation. To address this limitation, we propose a dynamic logic kernel framework that reorganizes kernel utilization by promoting specialization across kernel groups. This enables the network to better utilize increased width via input-dependent kernel routing, while ensuring that both routing and computation are implemented entirely with gate-level Boolean operations at inference time. We further find that kernel redundancy is most pronounced at the first gate level, motivating an early-stage dynamic logic kernel strategy that concentrates adaptation at this level. Experimental results demonstrate that our approach improves kernel utilization and increases kernel diversity, leading to higher accuracy with improved parameter efficiency.
Chinese Translation
可微分逻辑门网络仅使用逻辑门操作，最近作为传统神经网络的高效替代方案引起了关注。然而，尽管其效率较高，逻辑门网络的扩展行为仍然未得到充分探索。相比之下，扩展模型容量是深度神经网络中的一个核心设计原则，通常会带来性能的提升。这一差异引发了一个关键问题：逻辑门网络是否也能实现类似的扩展收益？在本研究中，我们将宽度作为主要的扩展轴，并对其在逻辑门网络中的行为进行系统分析。我们观察到，简单的宽度扩展往往会在逻辑核之间引入冗余，限制了额外核的有效使用，并导致性能饱和。为了解决这一限制，我们提出了一种动态逻辑核框架，通过促进核组之间的专业化来重新组织核的利用。这使得网络能够通过依赖输入的核路由更好地利用增加的宽度，同时确保在推理时路由和计算完全通过门级布尔操作实现。我们进一步发现，核冗余在第一个门级最为明显，这促使我们在早期阶段采用动态逻辑核策略，集中适应于这一层级。实验结果表明，我们的方法提高了核的利用率和多样性，从而在提高参数效率的同时提升了准确性。
cs.CV / 67 / 2610.01092
Ego2Act: Evaluating Goal-Directed Manipulation in Egocentric Video Generation
Ego2Act：评估以自我为中心的视频生成中的目标导向操控
Abstract
Video generation models are increasingly being explored as world simulators for embodied planning and learning. To do so effectively, these models must not only generate visually appealing frames, but also predict how environments dynamically evolve when executing goal-directed actions. While evaluating these capabilities is crucial, existing benchmarks focus mainly on single short actions or step-by-step instructions. This leaves multi-step physical reasoning underexplored, especially in egocentric video generation that requires planning to simulate proper execution to accomplish high-level goals by carrying out multiple real-world manipulations. We introduce Ego2Act, a goal-directed benchmark featuring 2,640 videos from 110 real-world tasks across day-to-day settings, varying object clutter and multi-step complexity. Given an initial scene image and a high-level goal, Ego2Act evaluates whether video generation models can produce realistic egocentric videos of a hand manipulating objects to carry out the task. To support scalable evaluation, we also introduce Ego2ActJudge, a reference-free evaluation pipeline that achieves better task completion and physics plausibility evaluation alignment with human consensus compared to relevant baselines. Our findings reveal that models' generated simulations often skip or partially execute steps, leaving later steps missing dependent states, which leads to unfulfilled goal. Furthermore, models consistently fail at fine-grained physical dynamics, particularly during complex object manipulation and persistent world modeling. We hope Ego2Act provides a rigorous testbed for advancing video models toward physically plausible, goal-directed simulation.
Chinese Translation
视频生成模型越来越多地被探索作为具身规划和学习的世界模拟器。为了有效实现这一目标，这些模型不仅需要生成视觉上吸引人的帧，还必须预测在执行目标导向动作时环境如何动态演变。虽然评估这些能力至关重要，但现有基准主要集中在单一短动作或逐步指令上。这使得多步骤物理推理未得到充分探索，尤其是在以自我为中心的视频生成中，该过程需要规划以模拟适当的执行，从而通过进行多项现实世界操控来实现高层次目标。我们引入了Ego2Act，这是一个目标导向基准，包含来自110个日常任务的2,640个视频，涵盖不同的物体杂乱程度和多步骤复杂性。在给定初始场景图像和高层次目标的情况下，Ego2Act评估视频生成模型是否能够生成现实的以自我为中心的视频，展示手部操控物体以完成任务。为了支持可扩展的评估，我们还引入了Ego2ActJudge，这是一个无参考评估管道，与相关基线相比，在任务完成率和物理合理性评估上与人类共识的对齐度更高。我们的研究发现，模型生成的模拟往往跳过或部分执行步骤，导致后续步骤缺失依赖状态，从而未能实现目标。此外，模型在细粒度物理动态方面持续失败，特别是在复杂物体操控和持续世界建模过程中。我们希望Ego2Act能为推动视频模型朝向物理合理的目标导向模拟提供一个严格的测试平台。
cs.CV / 68 / 2610.01098
MVDG: Efficient Multi-view 3D Disambiguation on Unconstrained Real-World Images
MVDG：在无约束真实世界图像上高效的多视图三维消歧
Abstract
Illusory matches between distinct yet visually similar 3D surfaces--doppelgangers--remain a fundamental obstacle for large-scale, in-the-wild 3D reconstruction and visual localization. Prior work mitigates this issue with pairwise classifiers, but this design limits multi-view contextual reasoning and incurs O(n^2) inference complexity for downstream structure-from-motion (SfM). We present MVDG, a scalable multi-view disambiguation framework built on the 3D foundation model VGGT, which jointly reasons over an arbitrary number of multiview images. By incorporating 3D-aware multi-view features, our method reduces dependence on pairwise comparisons by encoding and decoding views in a single pass. We further observe that direct multi-view fine-tuning of VGGT can be unstable under noisy supervision; motivated by label ambiguity in Doppelgangers, we construct a pseudo-pairwise training set from AerialMegaDepth and show that fine-tuning on sampled subsets yields stable optimization and strong generalization to held-out scenes. Finally, because full SfM evaluation (even with faster pipelines such as GLOMAP) remains expensive, we process a pseudo-pairwise dataset for efficient validation; we derive a predictive relationship between regular SfM metrics and the classification accuracy on this pseudo-pairwise test. Experiments show that our method achieves comparable pairwise accuracy while improving both SfM accuracy and inference speed over baselines.
Chinese Translation
不同但视觉上相似的三维表面之间的虚假匹配——双胞胎——仍然是大规模野外三维重建和视觉定位的一个基本障碍。以往的研究通过成对分类器来缓解这一问题，但这种设计限制了多视图上下文推理，并导致下游运动重建（Structure-from-Motion，SfM）推理复杂度为 O(n^2)。我们提出了 MVDG，一个基于三维基础模型 VGGT 的可扩展多视图消歧框架，能够对任意数量的多视图图像进行联合推理。通过结合三维感知的多视图特征，我们的方法通过在单次传递中编码和解码视图，减少了对成对比较的依赖。我们进一步观察到，VGGT 的直接多视图微调在噪声监督下可能不稳定；受到双胞胎标签模糊性的启发，我们从 AerialMegaDepth 构建了一个伪成对训练集，并展示了在采样子集上进行微调能够实现稳定的优化和对保留场景的强泛化。最后，由于完整的 SfM 评估（即使使用更快的管道如 GLOMAP）仍然成本高昂，我们处理了一个伪成对数据集以实现高效验证；我们推导出常规 SfM 指标与该伪成对测试的分类准确性之间的预测关系。实验表明，我们的方法在实现可比的成对准确性的同时，提高了 SfM 的准确性和推理速度。
cs.CV / 69 / 2610.01114
Affine-Aligned Atlas for Canonical Gaussian Construction in Video Representation
用于视频表示的仿射对齐图谱中的典型高斯构建
Abstract
Gaussian splatting has recently emerged as an efficient representation for images and videos due to its explicit structure and fast rendering capability. Existing Gaussian-based video representations often decompose a video into canonical Gaussians and temporal deformation. However, when a video contains large global motion such as camera movement, the canonical representation may become misaligned with individual frames, increasing the burden on the temporal deformation model. In this paper, we propose an affine-atlas canonical Gaussian representation, which constructs canonical Gaussians in a larger affine-aligned atlas space. Frame-wise affine transforms absorb global motion before canonical Gaussian construction, reducing the gap between the canonical representation and target frames. Since the proposed method only modifies the canonical construction stage, it can be integrated into existing canonical-Gaussian-based methods with negligible additional parameter cost. Experiments show that our method improves reconstruction quality especially for sequences with large camera motion.
Chinese Translation
高斯点云（Gaussian splatting）最近作为一种高效的图像和视频表示方法而受到关注，因其明确的结构和快速的渲染能力。现有的基于高斯的视频表示通常将视频分解为典型高斯和时间变形。然而，当视频包含大范围的全局运动（如相机移动）时，典型表示可能与单个帧失去对齐，从而增加时间变形模型的负担。在本文中，我们提出了一种仿射图谱（affine-atlas）典型高斯表示，该方法在更大的仿射对齐图谱空间中构建典型高斯。逐帧的仿射变换在典型高斯构建之前吸收全局运动，从而减少典型表示与目标帧之间的差距。由于所提出的方法仅修改典型构建阶段，因此可以以微不足道的额外参数成本集成到现有的基于典型高斯的方法中。实验表明，我们的方法在重建质量上有所提升，特别是在具有大相机运动的序列中。
cs.CV / 70 / 2610.01134
Open Vocabulary Word Recognition From Transcribed Bangla Texts
从转录的孟加拉文本中进行开放词汇的词识别
Abstract
An optical character recognition (OCR) can scan a paper and extract text using technology, making people's jobs easier. While various OCR systems are available in the software industry, finding a reliable equivalent solution for Bangla takes much work. When it comes to handwritten texts, the situation is much more unusual. Recognizing words from word images is the most critical stage in any OCR process. It is the second stage after segmenting words from text pictures. If this stage fails, the overall performance of the OCR will be poor, regardless of how well the other phases perform. This study aims to recognize words using deep learning in a handwritten Bangla word image. Three object detection models, SSD with MobileNetV2, Faster R-CNN with InceptionResNetV2, and an ensemble model of these two, have been used to train and test handwritten word images. A modified Non-Maximum Suppression has been introduced to enhance the effectiveness of the models' results. A customized dataset of 9841 handwritten Bangla word images has been compiled, featuring diverse handwriting styles from various individuals. All three models' performances have been checked against the test dataset, and the ensemble model has been the most impressive, with an F1-score of 92.61%. Also, at the word level, the ensemble model correctly recognizes 96.12% of the words to some extent. The system can be further improved by introducing a post-processing phase to correct errors generated by the system.
Chinese Translation
光学字符识别（OCR）技术可以扫描纸质文档并提取文本，从而简化人们的工作。尽管软件行业中有多种OCR系统，但为孟加拉语找到一个可靠的等效解决方案仍然非常困难。对于手写文本而言，情况更加复杂。从词图像中识别单词是任何OCR过程中的关键阶段。这是从文本图像中分割单词后的第二个阶段。如果这一阶段失败，无论其他阶段表现得多么出色，OCR的整体性能都会很差。本研究旨在利用深度学习识别手写孟加拉单词图像中的单词。我们使用了三种目标检测模型，分别是基于MobileNetV2的SSD、基于InceptionResNetV2的Faster R-CNN，以及这两者的集成模型，对手写单词图像进行训练和测试。引入了一种改进的非极大值抑制方法，以提高模型结果的有效性。我们编制了一个包含9841个手写孟加拉单词图像的定制数据集，涵盖了来自不同个体的多种书写风格。对所有三种模型的性能进行了测试，结果显示集成模型表现最佳，F1-score达到了92.61%。此外，在单词级别上，集成模型在一定程度上正确识别了96.12%的单词。通过引入后处理阶段来纠正系统产生的错误，可以进一步提高系统的性能。
cs.CV / 71 / 2610.01135
The RSNA Intracranial Aneurysm (RSNA-ICA) Dataset
RSNA颅内动脉瘤（RSNA-ICA）数据集
de Verdier, Maria Correia, Saluja, Rachit, Sho, Jason, Vabarizad, Maryam, Chen, Rennie Yung-Chieh, Nguyen, Uyen N. T., Alrehaili, Mona, Aweidah, Layal, Bulja, Deniz, Chan, Wesley C., Chaves, Hernan, Duvvuri, Madhavi, Ergin, Huseyin Ekin, Erdenebold, Undrakh-Erdene, Gumeler, Ekim, Jabal, Mohamed Sobhi, Kuo, Chin-Chi, Mubarak, Fatima, Emir, Sevde Nur, Ong, Scott Riley K., Ortiz, Johanna, Pérez-Lara, Almudena, Rauschecker, Andreas M., Anwar, Shayan Sirat Maheen, Tippareddy, Charit, Tran, Tam, Visrutaratna, Sorawis, Mongan, John, Flanders, Adam E., Ball, Robyn, Zaharchuk, Greg, Chang, Peter D., Kitamura, Felipe, Colak, Errol, Prevedello, Luciano, Richards, Tyler, Group, Data Contributor, Group, Dataset Annotator, Calabrese, Evan, Rudie, Jeffrey D.
Abstract
Intracranial aneurysm rupture is associated with substantial morbidity and mortality, yet aneurysm detection remains challenging, particularly for small lesions and on routine non-angiographic imaging examinations. To support the development and evaluation of artificial intelligence (AI) algorithms for intracranial aneurysm detection and localization, the Radiological Society of North America (RSNA), in collaboration with the American Society of Neuroradiology (ASNR), the Society of Neurointerventional Surgery (SNIS), and the European Society of Neuroradiology (ESNR), curated the RSNA Intracranial Aneurysm (RSNA-ICA) Dataset. Developed for the 2025 RSNA Intracranial Aneurysm Detection Challenge, RSNA-ICA is a large, publicly available, expert-annotated dataset comprising 7202 CTA, MRA, and MRI series from 4278 adult patients collected across 21 institutions in 12 countries spanning five continents. The dataset includes 2566 CTA, 2166 MRA, and 2470 MRI series from patients with and without intracranial saccular aneurysms, providing substantial geographic and imaging diversity. Expert annotations indicate both aneurysm presence and location, and 178 series additionally include three-dimensional segmentations of challenge-defined vascular locations. RSNA-ICA was used to develop and evaluate algorithms in the 2025 RSNA Intracranial Aneurysm Detection Challenge. Of the 7202 image series, 5041 are publicly available through MIRA (https://mira.rsna.org/dataset/7), while the remainder were used for challenge public and private test sets. The dataset is freely available to the research community for noncommercial use and provides a comprehensive resource for advancing AI-based aneurysm detection across both angiographic and routine neuroimaging examinations.
Chinese Translation
颅内动脉瘤破裂与显著的发病率和死亡率相关，但动脉瘤的检测仍然具有挑战性，尤其是在常规非血管成像检查中对于小病变的识别。为了支持人工智能（AI）算法在颅内动脉瘤检测和定位方面的发展与评估，北美放射学会（RSNA）与美国神经放射学会（ASNR）、神经介入外科学会（SNIS）以及欧洲神经放射学会（ESNR）合作，策划了RSNA颅内动脉瘤（RSNA-ICA）数据集。该数据集为2025年RSNA颅内动脉瘤检测挑战而开发，是一个大型的、公开可用的、专家注释的数据集，包含来自12个国家21个机构的4278名成人患者的7202个CTA、MRA和MRI系列。数据集中包括2566个CTA、2166个MRA和2470个MRI系列，涵盖了有无颅内囊性动脉瘤的患者，提供了显著的地理和成像多样性。专家注释指示了动脉瘤的存在和位置，此外178个系列还包括挑战定义的血管位置的三维分割。RSNA-ICA被用于开发和评估2025年RSNA颅内动脉瘤检测挑战中的算法。在7202个影像系列中，5041个通过MIRA（https://mira.rsna.org/dataset/7）公开可用，其余用于挑战的公共和私有测试集。该数据集免费提供给研究社区用于非商业用途，为推动基于AI的动脉瘤检测在血管成像和常规神经影像检查中的应用提供了全面的资源。
cs.CV / 72 / 2610.01148
OptimusMesh: Compact Autoregressive Mesh Generation from Point Clouds via Sparse Latent Pivots
OptimusMesh：通过稀疏潜在枢轴从点云生成紧凑的自回归网格
Abstract
Generating compact and geometrically faithful 3D meshes directly from point clouds remains a fundamental challenge. Point clouds are unordered and sparse, whereas meshes exhibit irregular structure and varying topology. As a result, many existing approaches rely on implicit representations followed by surface extraction or reconstruction. Although effective, these pipelines can produce dense or over-smoothed meshes, often requiring computationally expensive post-processing and simplification. We present OptimusMesh, a framework for direct compact triangle mesh generation from point clouds using sparse latent pivot conditioning. Our key idea is to compress $2{,}048$ oriented input points into only $16$ sparse latent pivots, reducing the geometric conditioning set by $128\times$. These pivots provide a compact structural representation shared across a two-stage autoregressive framework that first generates mesh vertices and then predicts triangular faces conditioned on the generated vertices and the same pivots. Compared with the evaluated recent point-cloud-conditioned autoregressive methods, which use $257$ decoder-conditioning tokens, OptimusMesh uses only $16$, yielding a $16.1\times$ shorter conditioning sequence. Experiments show that OptimusMesh produces the most compact outputs among the compared recent autoregressive methods, using $25.7\%$--$94.1\%$ fewer faces while maintaining competitive geometric fidelity and distributional quality.
Chinese Translation
直接从点云生成紧凑且几何上真实的3D网格仍然是一个基本挑战。点云是无序和稀疏的，而网格则表现出不规则的结构和变化的拓扑。因此，许多现有方法依赖于隐式表示，随后进行表面提取或重建。尽管这些流程有效，但它们可能会产生密集或过于平滑的网格，通常需要计算上昂贵的后处理和简化。我们提出了OptimusMesh，一个用于从点云直接生成紧凑三角网格的框架，采用稀疏潜在枢轴条件化。我们的关键思想是将$2{,}048$个定向输入点压缩为仅$16$个稀疏潜在枢轴，从而将几何条件集减少$128 imes$。这些枢轴提供了一个紧凑的结构表示，跨越一个两阶段的自回归框架，该框架首先生成网格顶点，然后根据生成的顶点和相同的枢轴预测三角面。与评估的最近点云条件自回归方法相比，后者使用$257$个解码器条件标记，OptimusMesh仅使用$16$，从而产生$16.1 imes$更短的条件序列。实验表明，OptimusMesh在比较的最近自回归方法中生成了最紧凑的输出，使用$25.7\%$--$94.1\\%$更少的面，同时保持竞争的几何保真度和分布质量。
cs.CV / 73 / 2610.01162
PhysicsLENS: Diagnosing Physical Property Blindness in Video Generation Models
PhysicsLENS：诊断视频生成模型中的物理属性盲区
Abstract
Reliable video world models could provide scalable predictive environments for robot learning, planning, and evaluation. However, generated robot videos can violate physical principles and complete tasks through physically implausible behavior, limiting their reliability for robot learning and planning. Current video-generation benchmarks exclude physics that are inherently hidden by visuals (e.g., weight, viscosity, friction). Due to this, video models are evaluated on the fidelity of physics, not the underlying accuracy of physics. We introduce PhysicsLENS, a dataset and benchmark for evaluating plausibility of physical properties grounded in robotics. PhysicsLENS uses matched scenario pairs that hold the same conditioning frame and task, while varying underlying physics in the scene description. Scenarios are curated from public robot video sources and annotated across seven physical domains: collision, gravity, momentum, friction, deformation, fluid, and causality. We evaluate across four video generation models, producing over 400 human-annotated labels. Results show that plausible-looking videos often ignore the stated property (34 of 47), and that stating the property lowers plausibility only slightly and not significantly.
Chinese Translation
可靠的视频世界模型可以为机器人学习、规划和评估提供可扩展的预测环境。然而，生成的机器人视频可能违反物理原则，并通过不符合物理规律的行为完成任务，从而限制了其在机器人学习和规划中的可靠性。目前的视频生成基准排除了那些在视觉上固有隐藏的物理因素（例如，重量、粘度、摩擦力）。因此，视频模型的评估基于物理的保真度，而非物理的基础准确性。我们提出了PhysicsLENS，一个用于评估基于机器人技术的物理属性合理性的数据集和基准。PhysicsLENS使用匹配的场景对，这些场景具有相同的条件框架和任务，同时在场景描述中变化潜在的物理因素。场景来自公共机器人视频源，并在七个物理领域进行注释：碰撞、重力、动量、摩擦、变形、流体和因果关系。我们在四个视频生成模型上进行评估，产生了超过400个人工注释的标签。结果表明，看似合理的视频往往忽视了所述属性（47个中有34个），并且声明属性仅略微降低了合理性，且没有显著影响。
cs.CV / 74 / 2610.01166
CineMR: Tool-Integrated Vision-Language Reasoning for Quantitative Cardiac MRI Assessment
CineMR：集成工具的视觉-语言推理用于定量心脏MRI评估
Abstract
Cardiovascular magnetic resonance (CMR), including cine imaging, is a reference standard for the noninvasive assessment of cardiac morphology and ventricular function. Cine CMR interpretation integrates qualitative visual assessment with quantitative measurements of ventricular volumes, ejection fraction, myocardial mass, wall thickness, and regional wall motion. Current medical vision-language models (VLMs) cannot reliably derive quantitative measurements from multidimensional cine images without analysis tools. We present CineMR, a tool-augmented VLM that invokes cardiac image-analysis tools and integrates their outputs into interleaved reasoning for quantitative CMR assessment. We also construct a multi-cohort visual question answering benchmark covering quantitative metric extraction, multiclass diagnosis, and differential diagnosis, together with tools for segmentation, phase selection, volumetry, morphometry, and regional wall motion analysis. CineMR is trained with supervised fine-tuning (SFT) on tool-interaction traces followed by Group Relative Policy Optimization (GRPO) with conditional tool-use rewards. On the multi-cohort cine CMR benchmark, CineMR achieves 35.9% pass@1 and 58.9% pass@4, compared with 1.5% pass@1 for the Qwen3-VL-8B backbone and 0.0% and 7.0% pass@1 for LLaVA-Med v1.5 and MedGemma-4B, respectively. Correct tool invocation reaches 99.8% after GRPO, up from 78.9% after SFT. Live tool outputs improve ventricular measurement accuracy by 20.4--23.7% over direct model predictions, and removing all tools reduces pass@1 from 35.9% to 27.9%. These results highlight the importance of reliable tool use for quantitative cine CMR reasoning and support CineMR as a promising approach for assistive cardiac image assessment. Code, benchmark resources, and model weights are available at https://github.com/AI-MIND-Lab/CineMR.
Chinese Translation
心血管磁共振成像（CMR），包括动态成像，是评估心脏形态和心室功能的非侵入性参考标准。动态CMR的解读将定性的视觉评估与心室体积、射血分数、心肌质量、壁厚和区域壁运动的定量测量相结合。目前的医学视觉-语言模型（VLMs）在没有分析工具的情况下，无法可靠地从多维动态图像中推导出定量测量。我们提出了CineMR，这是一种增强工具的VLM，调用心脏图像分析工具，并将其输出整合到交错推理中，以进行定量CMR评估。我们还构建了一个多队列视觉问答基准，涵盖定量指标提取、多类别诊断和鉴别诊断，并提供分割、相位选择、体积测量、形态测量和区域壁运动分析的工具。CineMR通过对工具交互轨迹进行监督微调（SFT）进行训练，随后进行条件工具使用奖励的组相对策略优化（GRPO）。在多队列动态CMR基准上，CineMR的通过率为35.9%（pass@1）和58.9%（pass@4），而Qwen3-VL-8B主干网络的通过率为1.5%（pass@1），LLaVA-Med v1.5和MedGemma-4B的通过率分别为0.0%和7.0%（pass@1）。经过GRPO后，正确调用工具的比例达到99.8%，而SFT后的比例为78.9%。实时工具输出提高了心室测量的准确性，较直接模型预测提高了20.4%至23.7%，而去除所有工具后，通过率从35.9%降至27.9%。这些结果突显了可靠工具使用在定量动态CMR推理中的重要性，并支持CineMR作为辅助心脏图像评估的有前景的方法。代码、基准资源和模型权重可在https://github.com/AI-MIND-Lab/CineMR获取。
cs.CV / 75 / 2610.01180
Skeleton-and-Strategy Prompting: Training-Free Negation Understanding for Vision-Language Models
骨架与策略提示：无训练的视觉语言模型否定理解
Abstract
Despite the strong performance of Vision-Language Models (VLMs) on a wide range of visual question answering (VQA) tasks, these models consistently struggle to understand negation and produce incorrect answers when questions involve negated clauses. To address this limitation, we propose Skeleton-and-Strategy Prompting (\textbf{SSP}), a training-free, in-context learning method that improves VLM negation understanding capabilities without any parameter updates. Given a negation question, our method first abstracts the underlying question structure into a skeleton, retrieves a small set of same-skeleton questions from a lightweight question pool, then prompts the VLM to analyze their shared negation pattern and synthesize a single-sentence answering strategy. The skeleton and strategy are prepended to the test sample to guide the model correctly tackle the negation problems. Experiments on multiple negation VQA benchmarks show that SSP achieves state-of-the-art performance on negation-focused VQA tasks while remaining computationally efficient.
Chinese Translation
尽管视觉语言模型（VLMs）在广泛的视觉问答（VQA）任务中表现出色，但这些模型在理解否定方面始终存在困难，当问题涉及否定从句时，往往会产生错误答案。为了解决这一局限性，我们提出了骨架与策略提示（Skeleton-and-Strategy Prompting, SSP），这是一种无训练、上下文学习的方法，能够在不更新任何参数的情况下提高VLM的否定理解能力。针对否定问题，我们的方法首先将潜在的问题结构抽象为骨架，从轻量级问题池中检索一小组相同骨架的问题，然后提示VLM分析它们共享的否定模式，并综合出一个单句回答策略。骨架和策略被预先附加到测试样本中，以指导模型正确处理否定问题。在多个否定VQA基准测试上的实验表明，SSP在以否定为重点的VQA任务中实现了最先进的性能，同时保持了计算效率。
cs.CV / 76 / 2610.01191
Color Independent Word Segmentation From Transcribed Bangla Passages
与颜色无关的孟加拉文段落词语分割
Abstract
An optical character recognition(OCR) system can scan paper and extract text, making people's jobs easier. While numerous OCR systems are accessible in the software sector, finding a dependable equivalent solution for Bangla is tough. When it comes to handwritten texts, the case is even more rare. The first fundamental step to any OCR is to segment words from text images. If this stage fails, the total OCR's performance will be poor no matter how promising the later stages perform. This research aims to segment words in a handwritten Bangla text image. This research can be implemented on any smartphone-captured image, irrespective of the color and type of paper and ink. Furthermore, as smartphone-captured images can create shadow interferences, the custom dataset built for this research is created in such a way that every possible obstacle that can be faced is included. For 7374 words, a total of 7278 bounding boxes are generated, which have recall of 90.60 %, precision of 91.80 %, and F1-score of 91.20 %. The system can be further improved with nested operations on bounding boxes containing several words or by adjusting the adaptive thresholding and dilation filter sizes to a more precise level.
Chinese Translation
光学字符识别（OCR）系统能够扫描纸张并提取文本，从而简化人们的工作。尽管软件领域中有众多OCR系统，但为孟加拉语找到一个可靠的等效解决方案仍然困难重重。对于手写文本而言，这种情况更为罕见。任何OCR的第一步都是从文本图像中分割出单词。如果这一阶段失败，无论后续阶段表现多么出色，整体OCR的性能都将很差。本研究旨在对手写孟加拉文本图像进行单词分割。该研究可以应用于任何智能手机拍摄的图像，无论纸张和墨水的颜色和类型如何。此外，由于智能手机拍摄的图像可能会产生阴影干扰，因此为本研究构建的自定义数据集考虑了可能遇到的所有障碍。对于7374个单词，共生成7278个边界框，其召回率为90.60%，精确率为91.80%，F1-score为91.20%。该系统可以通过对包含多个单词的边界框进行嵌套操作，或通过将自适应阈值和膨胀滤波器的大小调整到更精确的水平来进一步改进。
cs.CV / 77 / 2610.01192
FlashBack: Knowing When to Remember in Streaming Vision-Language Models
FlashBack：在流媒体视觉-语言模型中知道何时记忆
Abstract
Streaming vision-language models must process continuously growing video streams under a bounded compute budget, creating a persistent tension between real-time perception and long-term memory. Retrieving historical information provides a natural remedy, yet historical recall is not uniformly beneficial: unnecessary history may introduce irrelevant context into current reasoning and interfere with native real-time perception. Effective streaming memory should therefore address not only what to remember, but also when and how to access it. To this end, we introduce FlashBack, a training-free framework for selective, multi-level memory in streaming vision-language models. Before retrieving history, FlashBack draws on the semantic understanding of the frozen streaming VLM to infer whether a query calls for historical evidence. This assessment determines whether inference remains on the Native trajectory or invokes an isolated Recall trajectory. The Recall trajectory combines recent context with retrieved long-term memory through a query-local Side-KV pathway, preserving local temporal continuity without modifying the persistent Native state. We instantiate FlashBack on StreamingVLM and Mage-VL-4B and evaluate it on OVO-Bench and StreamingBench. The results show improvements on several long-horizon and memory-dependent tasks while largely preserving real-time perception, with performance competitive with strong training-based streaming methods despite requiring no additional training. Our code will be announced later.
Chinese Translation
流媒体视觉-语言模型必须在有限的计算预算下处理不断增长的视频流，这在实时感知与长期记忆之间产生了持续的紧张关系。检索历史信息提供了一种自然的解决方案，但历史回忆并非总是有益的：不必要的历史可能会将无关的上下文引入当前推理，并干扰本土实时感知。因此，有效的流媒体记忆不仅应解决记住什么的问题，还应考虑何时以及如何访问这些信息。为此，我们提出了FlashBack，这是一个用于流媒体视觉-语言模型的选择性多层记忆的无训练框架。在检索历史之前，FlashBack利用冻结的流媒体视觉-语言模型的语义理解来推断查询是否需要历史证据。这一评估决定了推理是保持在本土轨迹上，还是调用一个独立的回忆轨迹。回忆轨迹通过查询局部的Side-KV路径将最近的上下文与检索到的长期记忆结合起来，保持局部时间连续性而不修改持久的本土状态。我们在StreamingVLM和Mage-VL-4B上实现了FlashBack，并在OVO-Bench和StreamingBench上进行了评估。结果显示，在多个长时间跨度和依赖记忆的任务上有显著改进，同时在很大程度上保持了实时感知，其性能与强大的基于训练的流媒体方法相当，尽管不需要额外的训练。我们的代码将在稍后公布。
cs.CV / 78 / 2610.01201
iSEE: Object Permanence Through Self-Supervision
iSEE：通过自我监督实现物体的持久性
Abstract
Object permanence, keeping track of an object's identity and position while it is occluded, is central to video representations that track, predict and plan. Trackers that achieve it learn from boxes, track identities and visibility labels. On the other hand, self-supervised object-centric methods discover objects without labels: through slot attention, it represents a video as slots that bind to objects and follow them across frames. However, these slots are lost under occlusion, making the desired permanence impossible. Reasoning permanence is a hard problem because it requires to detect when an object becomes occluded, re-identify when object reappears, and keep the object's hidden position continuous, using reapperance as the only learning cue. To address this, we propose iSEE, a novel framework that offers all three aforementioned requirements, without any labels whatsoever. We built iSEE using the following three proposed components: (i) Object evidence modelling: a slot's attention, compared with its own past, reveals when its object is hidden. (ii) Appearance-position separation: two slot streams let the appearance be held for re-identification while the position keeps changing. (iii) Permanence from reappearance: a walker follows the hidden object's position, trained only on where the object reappears. On LA-CATER static, iSEE returns a reappearing object to its own slot after 86% of occlusions, against 32% for SlotContrast, and localises it while hidden within 4.1 mAP of the label-trained SoTA RAM. The two streams also allow downstream planning, with the position stream as the action of a world model. Project page: https://insait-institute.github.io/iSEE/
Chinese Translation
物体持久性，即在物体被遮挡时跟踪其身份和位置，对于跟踪、预测和规划的视频表示至关重要。实现这一目标的跟踪器通过框进行学习，跟踪身份和可见性标签。另一方面，自我监督的物体中心方法在没有标签的情况下发现物体：通过槽注意力，它将视频表示为绑定到物体的槽，并在帧之间跟随它们。然而，这些槽在遮挡下会丢失，使得所需的持久性变得不可能。推理持久性是一个困难的问题，因为它需要检测物体何时被遮挡、在物体重新出现时重新识别，并保持物体隐藏位置的连续性，仅使用重新出现作为学习线索。为了解决这个问题，我们提出了iSEE，一个新颖的框架，满足上述三个要求，而无需任何标签。我们通过以下三个提出的组件构建了iSEE：（i）物体证据建模：槽的注意力与其自身过去的比较揭示了其物体何时被隐藏。（ii）外观-位置分离：两个槽流允许在重新识别时保持外观，同时位置不断变化。（iii）从重新出现中获得持久性：一个行走者跟踪隐藏物体的位置，仅在物体重新出现时进行训练。在LA-CATER静态数据集上，iSEE在86%的遮挡情况下将重新出现的物体返回到其自身槽，而SlotContrast仅为32%，并在隐藏状态下将其定位在标签训练的SoTA RAM的4.1 mAP内。这两个流还允许下游规划，其中位置流作为世界模型的动作。项目页面：https://insait-institute.github.io/iSEE/
cs.CV / 79 / 2610.01205
Semantic RGB--Depth Based Surgical Skill Assessment in Microscopic Stereo Videos
基于语义RGB-深度的显微立体视频外科技能评估
Abstract
Objective assessment of microsurgical technical skill is essential for competency-based training and quality assurance, yet existing video-based approaches predominantly rely on RGB images and therefore overlook the 3D spatial relationships that characterize instrument-anatomy interactions. Although stereo operating microscopes provide complementary depth information, conventional stereo matching algorithms can produce sparse and unreliable depth estimates under high-magnification imaging conditions, limiting their use for automated skill assessment. This work presents a semantic RGB-Depth framework for surgical skill assessment from microscopic stereo videos. A regression-based depth fusion method combines sparse metric stereo depth with dense monocular depth estimates to generate a dense geometric representation of the surgical scene. This representation is integrated with semantically decomposed RGB streams corresponding to individual surgical instruments and surrounding anatomy. A hierarchical attention architecture jointly encodes these streams to capture discriminative patterns of instrument use and instrument-anatomy interaction across surgeons at different training levels. The framework was evaluated on 33 ex vivo transoral microlaryngeal procedures performed by six surgeons, comprising attending surgeons and surgical residents, using leave-one-surgeon-out cross-validation. The proposed semantic RGB-Depth model achieved an F1 score of 0.938 for skill-level classification, compared with 0.696 for semantic RGB and 0.929 for semantic depth. These results suggest that geometric information can improve automated surgical skill assessment from microscopic stereo videos. The learned spatial, temporal, and semantic attention patterns also support qualitative examination of the scene regions, video segments, and semantic streams emphasized by the model.
Chinese Translation
对显微外科技术技能的客观评估对于基于能力的培训和质量保证至关重要，但现有的视频评估方法主要依赖RGB图像，因此忽视了表征工具与解剖结构交互的三维空间关系。尽管立体手术显微镜提供了互补的深度信息，但传统的立体匹配算法在高倍成像条件下可能产生稀疏且不可靠的深度估计，从而限制了其在自动化技能评估中的应用。本研究提出了一种基于语义RGB-深度的框架，用于从显微立体视频中评估外科技能。该框架采用基于回归的深度融合方法，将稀疏的度量立体深度与密集的单目深度估计相结合，以生成手术场景的密集几何表示。该表示与对应于单个外科工具和周围解剖结构的语义分解RGB流相结合。一个层次注意力架构共同编码这些流，以捕捉不同培训水平外科医生在工具使用和工具-解剖结构交互中的区分模式。该框架在六名外科医生（包括主治医生和外科住院医生）进行的33例体外经口显微喉部手术中进行了评估，采用留一外科医生交叉验证。所提出的语义RGB-深度模型在技能水平分类中达到了0.938的F1分数，而语义RGB为0.696，语义深度为0.929。这些结果表明，几何信息可以改善从显微立体视频中进行的自动化外科技能评估。学习到的空间、时间和语义注意力模式也支持对模型强调的场景区域、视频片段和语义流的定性检查。
cs.CV / 80 / 2610.01206
Resolving Mixed Single-Photon LiDAR Returns for Foreground-View and Hidden Scene Reconstruction
解决混合单光子LiDAR返回信号以进行前景视图和隐藏场景重建
Abstract
Partially transmissive screens and protective covers are common in robotic inspection, but they create mixed LiDAR returns from both the foreground material and the scene behind it. Conventional peak-based LiDAR usually discards weak hidden returns, while single-photon LiDAR records time-resolved histograms that preserve attenuated and overlapping echoes. However, existing transient reconstruction methods typically fit a single scene representation to the measured waveform. Under occlusion, weak or nearby foreground--hidden echoes can form a broad peak or subtle shoulder. Because such waveforms can also be explained by a displaced single surface or a thick density distribution, accurate transient fitting does not necessarily imply correct geometry. We propose a state-aware framework for foreground-view and hidden scene reconstruction from occluded single-photon histograms. For each ray, we estimate local echo evidence, identifying no reliable surface evidence, single-return evidence, or two returns. The inferred echo state routes supervision for a two-head neural field: all rays constrain waveform reconstruction, while reliable anchors provide geometry localization. We also introduce a real paired single-photon LiDAR occlusion dataset with occluded and clean captures at fixed poses. Experiments on a real dataset show improved hidden scene depth and point-cloud accuracy over baselines. Our results demonstrate single-photon layered reconstruction as a practical route for 3D perception through partially transmissive occluders.
Chinese Translation
部分透光屏幕和保护罩在机器人检测中很常见，但它们会产生来自前景材料和其背后场景的混合LiDAR返回信号。传统的基于峰值的LiDAR通常会丢弃微弱的隐藏返回信号，而单光子LiDAR记录时间分辨的直方图，保留了衰减和重叠的回声。然而，现有的瞬态重建方法通常将单一场景表示拟合到测量波形上。在遮挡情况下，微弱或邻近的前景-隐藏回声可能形成一个宽峰或微妙的肩部。因为这样的波形也可以通过位移的单一表面或厚密度分布来解释，准确的瞬态拟合并不一定意味着正确的几何形状。我们提出了一种状态感知框架，用于从遮挡的单光子直方图中重建前景视图和隐藏场景。对于每条光线，我们估计局部回声证据，识别出没有可靠表面证据、单返回证据或两个返回。推断的回声状态为双头神经场提供监督：所有光线约束波形重建，而可靠的锚点提供几何定位。我们还引入了一个真实的配对单光子LiDAR遮挡数据集，其中包含在固定姿态下的遮挡和干净捕获。对真实数据集的实验表明，相较于基线，隐藏场景深度和点云精度得到了改善。我们的结果展示了单光子分层重建作为通过部分透光遮挡物进行3D感知的实用途径。
cs.CV / 81 / 2610.01210
EgoFound3R: End-to-End Egocentric Hand Reconstruction in World Space with Point-Wise Interaction Attributes
EgoFound3R：基于点交互属性的世界空间端到端自我中心手部重建
Abstract
Egocentric video has become a primary source of supervision for embodied models, and its value rests on recovering hand motion in world coordinates, which camera motion and hand occlusion make difficult. Existing reconstruction pipelines typically separate hand and scene estimation, leave interaction attributes to separate task-specific models, and invoke several models per video, so no prior reconstruction model estimates these attributes and throughput becomes a practical constraint on large-scale annotation. We therefore introduce EgoFound3R, a unified end-to-end model that estimates world-space hand geometry in a metric scale shared with the scene, and predicts point-wise interaction attributes, including visibility, contact, and distance. The model integrates three designs: (i) structured hand prompts that transfer pretrained geometric priors to world-space hand reconstruction; (ii) an explicit hand representation that decodes hand geometry and interaction attributes; and (iii) a shared-parameter multi-rate design that lowers inference cost. Together, these designs predict hand geometry and point-wise attributes in one pass. On OakInk-v2, TACO, and HOI4D, EgoFound3R reduces the mean per-joint position error (MPJPE) by 43.2%, 22.4%, and 11.6% over previous methods and predicts point-wise contact and distance alongside the geometry in the same pass, while attaining approximately 6x higher throughput.
Chinese Translation
自我中心视频已成为具身模型的主要监督来源，其价值在于恢复世界坐标中的手部运动，而相机运动和手部遮挡使这一过程变得困难。现有的重建管道通常将手部和场景估计分开，将交互属性留给特定任务的独立模型，并在每个视频中调用多个模型，因此没有现有的重建模型能够估计这些属性，并且在大规模标注中吞吐量成为实际限制。因此，我们提出了EgoFound3R，这是一个统一的端到端模型，能够在与场景共享的度量尺度中估计世界空间的手部几何形状，并预测点交互属性，包括可见性、接触和距离。该模型整合了三种设计：（i）结构化手部提示，将预训练的几何先验转移到世界空间手部重建；（ii）显式手部表示，解码手部几何和交互属性；（iii）共享参数的多速率设计，降低推理成本。这些设计共同实现了一次性预测手部几何和点交互属性。在OakInk-v2、TACO和HOI4D数据集上，EgoFound3R将每个关节位置误差的平均值（MPJPE）分别降低了43.2%、22.4%和11.6%，并在同一次推理中预测点接触和距离，同时实现了约6倍的更高吞吐量。
cs.CV / 82 / 2610.01215
AutoGUIWorld: Image Generators as Visual World Models for GUI Agent
AutoGUIWorld：作为视觉世界模型的图像生成器用于GUI代理
Yang, Cheng, Wu, Yifan, Huang, Yutao, Zhang, Zhaohua, Chen, Beiduo, Chen, Muxi, Zhao, Chenchen, Deng, Hexuan, Yang, Haolin, Zhu, Geyuan, Zhu, Sa, Zhuo, Jianhuan, Xiao, Qiuyong, Ruan, Jianhao, Peng, Yiran, Zhang, Jiayi, Ye, Tian, Yu, Xinlei, Jiang, Tianwen, Zhang, Jihong, Luo, Yuyu
Abstract
GUI agents require high-quality interaction trajectories to learn how software environments respond to actions, maintain state, and support multi-step workflows. However, the diversity of available trajectories is constrained by the applications, interface states, and workflows accessible in the underlying environments. Expanding this coverage requires deploying increasingly diverse and complex software, with specialized applications imposing additional installation, configuration, and runtime costs. We introduce AutoGUIWorld, a data generation framework that combines the visual priors of image generators with the task knowledge of a planner to synthesize GUI interaction trajectories without deploying or running the corresponding software environments. AutoGUIWorld samples initial GUI scenes from structured specifications of operating-system context, visual appearance, and interface state, and generates tasks conditioned on those scenes. A planner then specifies atomic actions and their intended visual consequences, while an image generator iteratively edits the current screenshot to produce subsequent observations. Action grounding and transition-level quality filtering yield 79,266 spatially annotated step-level training samples across Ubuntu, Windows, macOS, and Chrome. Fine-tuning Qwen3.5-35B-A3B on AutoGUIWorld trajectories improves the mean task score on OSWorld from 33.0% to 40.8% and the task success rate on ScienceBoard from 14.0% to 32.2%. These results show that generated trajectories improve GUI-agent performance on real desktop and scientific tasks.
Chinese Translation
GUI代理需要高质量的交互轨迹，以学习软件环境如何响应操作、维持状态并支持多步骤工作流。然而，现有轨迹的多样性受到底层环境中可访问的应用程序、界面状态和工作流的限制。扩展这种覆盖范围需要部署越来越多样化和复杂的软件，而专用应用程序则会增加额外的安装、配置和运行时成本。我们提出了AutoGUIWorld，这是一个数据生成框架，结合了图像生成器的视觉先验和规划者的任务知识，以合成GUI交互轨迹，而无需部署或运行相应的软件环境。AutoGUIWorld从操作系统上下文、视觉外观和界面状态的结构化规范中采样初始GUI场景，并生成基于这些场景的任务。然后，规划者指定原子动作及其预期的视觉后果，而图像生成器则迭代地编辑当前屏幕截图，以生成后续观察。动作定位和过渡级质量过滤在Ubuntu、Windows、macOS和Chrome上产生了79,266个空间注释的步骤级训练样本。在AutoGUIWorld轨迹上微调Qwen3.5-35B-A3B将OSWorld上的平均任务分数从33.0%提高到40.8%，将ScienceBoard上的任务成功率从14.0%提高到32.2%。这些结果表明，生成的轨迹提高了GUI代理在真实桌面和科学任务中的表现。
cs.CV / 83 / 2610.01229
A Compact Explicit 4D Representation for Dynamic Scenes
动态场景的紧凑显式4D表示
Abstract
A compact dynamic-scene representation must retain both the surfaces seen over time and the appearance needed to render them from new viewpoints. We present Sparc4D, a feed-forward autoencoder that encodes a monocular video with known cameras into a sparse 4D scene state. Static features are shared across the clip, while spatially anchored temporal slots compress time-varying features. A sparse decoder produces 2D Gaussian surfels, while stored source pixels preserve fine texture through geometric re-projection. The state includes one full source frame and dynamic-region pixels sampled every fourth frame, alongside learned features and sparse occupancy. For a 32-frame MultiCamVideo clip, it averages 0.95M 32-bit-equivalent values on random windows and 0.92M on the first-32 protocol. On first-32, Sparc4D reaches 21.70\,dB, compared with 20.40\,dB for MoVieS. On randomly placed windows, their PSNR scores are comparable. With stored texture disabled, temporal slots compress the time-varying feature state by a median $4.0\times$ and reduce the mean state from 1.04M to 0.42M values, with essentially unchanged target-view reconstruction quality. Without fine-tuning on real data, Sparc4D transfers to DyCheck and Neu3D, where stored texture improves LPIPS while slightly reducing PSNR.
Chinese Translation
紧凑的动态场景表示必须同时保留随时间变化的表面和从新视点渲染所需的外观。我们提出了Sparc4D，一种前馈自编码器，它将已知相机的单目视频编码为稀疏的4D场景状态。静态特征在整个片段中共享，而空间锚定的时间槽则压缩时间变化的特征。稀疏解码器生成2D高斯表面点（surfels），而存储的源像素通过几何重投影保留细腻的纹理。该状态包括一个完整的源帧和每四帧采样的动态区域像素，以及学习到的特征和稀疏占用情况。对于一个32帧的MultiCamVideo片段，它在随机窗口上平均为0.95M 32位等效值，在first-32协议下为0.92M。在first-32上，Sparc4D达到21.70 dB，而MoVieS为20.40 dB。在随机放置的窗口上，它们的PSNR得分相当。禁用存储纹理后，时间槽将时间变化特征状态压缩了中位数$4.0 imes$，并将平均状态从1.04M减少到0.42M值，同时目标视图重建质量基本不变。在没有对真实数据进行微调的情况下，Sparc4D成功迁移到DyCheck和Neu3D，其中存储的纹理改善了LPIPS，同时略微降低了PSNR。
cs.CV / 84 / 2610.01233
Flow Matching Reinforcement for 3D Mesh Generation via Dynamic Homing Optimization
通过动态归向优化的3D网格生成中的流匹配强化学习
Abstract
Flow matching is central to 3D generation, yet in practice its reinforcement learning (RL) methods are largely adapted from 2D visual generation. Representative DPO-, GRPO-, and NFT-style objectives, when applied to negative trajectories, mainly steer predicted velocities away from the corresponding directions without explicitly specifying a target velocity field toward preferred samples. In 3D generation, constrained by pretrained model capabilities, rollout diversity, and reward-distribution complexity, directly applying these RL methods yields limited gains in geometric quality. We introduce a forward-process RL method \textbf{Dynamic Homing Optimization (DHO)}, which reformulates negative-trajectory optimization as positive-sample attraction-guided dynamic homing. Specifically, Minimum-Cost Attractive Matching (MAM) assigns each negative sample a distinct positive target, and Time-Aware Dynamic Correction (TDC) then redirects its trajectory toward the target using a remaining-time-aware corrective velocity. Building on asynchronous online DHO, we develop \textbf{Flow3D-Pro}, an image-to-3D geometry generation framework. Experiments show that DHO outperforms representative DPO-, GRPO-, and NFT-style objectives in 3D generation, while Flow3D-Pro produces higher-quality 3D geometry than existing mesh generation methods.
Chinese Translation
流匹配是3D生成的核心，但在实际应用中，其强化学习（RL）方法主要是从2D视觉生成中改编而来。代表性的DPO、GRPO和NFT风格目标在应用于负轨迹时，主要是将预测速度引导远离相应方向，而没有明确指定朝向优选样本的目标速度场。在3D生成中，由于受限于预训练模型的能力、展开多样性和奖励分布的复杂性，直接应用这些RL方法在几何质量上获得的提升有限。我们提出了一种前向过程的RL方法——动态归向优化（Dynamic Homing Optimization, DHO），将负轨迹优化重新表述为正样本吸引引导的动态归向。具体而言，最小成本吸引匹配（Minimum-Cost Attractive Matching, MAM）为每个负样本分配一个独特的正目标，而时间感知动态修正（Time-Aware Dynamic Correction, TDC）则使用剩余时间感知的修正速度将其轨迹重新引导至目标。基于异步在线DHO，我们开发了Flow3D-Pro，一个图像到3D几何生成框架。实验表明，DHO在3D生成中优于代表性的DPO、GRPO和NFT风格目标，而Flow3D-Pro生成的3D几何质量高于现有的网格生成方法。
cs.CV / 85 / 2610.01243
When the Judge Acts: Auditing VLM-Guided Image Selection on Culturally Situated Prompts
当法官行动时：对文化情境提示下的 VLM 引导图像选择进行审计
Abstract
Vision-language models (VLMs) increasingly act as judges that pick the best of several generated images, so their choices decide what users see. Such judges are usually validated by score agreement with human ratings, not by the images they return. We audit VLM judges as decision-makers: on 300 culturally situated prompts, we compare the returned image with human ratings the judge never sees and with random choice from the same candidates, and repeat every decision with the candidates reordered. A 4B-parameter judge barely beats random and falls short of a CLIP similarity baseline. It picks the first image shown in 49% of calls (chance: 28%), and reordering changes its choice on 60% of prompts. For this judge, agreement across orders is informative: decisions that survive reordering are much better than random, whereas agreement with a weaker second judge keeps the wrong ones. An 8B judge shows almost no position bias and outperforms CLIP, yet for it the same filter mostly discards good decisions. Agreement helps only when it targets the judge's failure mode, so filters must be re-audited whenever the judge changes. The 4B judge's slight rise in stereotype ratings is no longer detectable after aggregating across orders or with the larger judge.
Chinese Translation
视觉-语言模型（VLMs）越来越多地充当选择多个生成图像中最佳图像的法官，因此它们的选择决定了用户所看到的内容。这些法官通常通过与人类评分的一致性来验证，而不是通过它们返回的图像。我们将 VLM 法官作为决策者进行审计：在 300 个文化情境提示上，我们比较了返回的图像与法官从未看到的人类评分，以及与同一候选者的随机选择，并对每个决策进行了候选者重新排序。一个拥有 40 亿参数的法官仅略微优于随机选择，并且未能达到 CLIP 相似性基准。它在 49% 的调用中选择第一个显示的图像（随机选择的概率为 28%），并且重新排序在 60% 的提示中改变了其选择。对于这个法官，跨顺序的一致性是有信息量的：在重新排序后仍然有效的决策远好于随机选择，而与较弱的第二个法官的一致性则保留了错误的选择。一个拥有 80 亿参数的法官几乎没有位置偏见，且优于 CLIP，但对于它来说，同样的过滤器大多数情况下会丢弃好的决策。一致性仅在针对法官的失败模式时才有帮助，因此每当法官发生变化时，过滤器必须重新审计。经过跨顺序聚合或使用更大的法官后，40 亿法官的刻板印象评分的轻微上升不再可检测。
cs.CV / 86 / 2610.01279
PickMoment: Continuous-Time Single-Image-to-Video via Learning Deblurring and Blur-to-Video
PickMoment：通过学习去模糊和模糊到视频的连续时间单图像到视频转换
Abstract
Motion blur arises from the temporal integration of a continuous sharp signal over a finite exposure window, yet existing learning-based methods sidestep this physical model and predict only the sharp signal itself: most single-image deblurring methods recover a single frame at the exposure center, while blur-to-video methods predict a fixed set of frames. We introduce PickMoment, a continuous-time reformulation that directly learns the interval-mean blur over arbitrary sub-intervals of the exposure with a single deterministic model. Drawing an analogy to MeanFlow's average-velocity formulation, we train the model with three supervisions derived from the blur integral: an empirical reconstruction loss from available subframes, an additivity loss that enforces self-consistency across overlapping sub-intervals, and a sharp-frame loss anchored at the zero-interval limit. A single trained model unifies single-image deblurring, blur-to-video generation, and continuous-time pick-a-moment recovery as different queries to the same network, with no separate training for each task. Our PickMoment achieves state-of-the-art performance among generative-based deblurring methods on GoPro and HIDE while competitive against restoration-based methods on RealBlur, and the highest per-frame fidelity on GoPro-7 blur-to-video, all in a single forward pass without iterative sampling.
Chinese Translation
运动模糊源于在有限曝光窗口内对连续清晰信号的时间积分，然而现有的基于学习的方法绕过了这一物理模型，仅预测清晰信号本身：大多数单图像去模糊方法在曝光中心恢复单帧，而模糊到视频的方法预测固定数量的帧。我们引入了PickMoment，这是一种连续时间重构方法，直接学习在曝光的任意子区间内的区间均值模糊，使用单一确定性模型。类比于MeanFlow的平均速度公式，我们通过来自模糊积分的三种监督信号训练模型：来自可用子帧的经验重建损失、强制重叠子区间自一致性的加法损失，以及锚定在零区间极限的清晰帧损失。一个经过训练的模型将单图像去模糊、模糊到视频生成和连续时间的选取时刻恢复统一为对同一网络的不同查询，无需为每个任务单独训练。我们的PickMoment在GoPro和HIDE的生成型去模糊方法中实现了最先进的性能，并在RealBlur上与基于恢复的方法具有竞争力，同时在GoPro-7模糊到视频中实现了最高的每帧保真度，所有这些均在单次前向传递中完成，无需迭代采样。
cs.CV / 87 / 2610.01283
ShelfChange3D: Object-Level 3D Change Detection for Retail Shelf Monitoring
ShelfChange3D：用于零售货架监控的对象级3D变化检测
Abstract
Reliable shelf monitoring is an important capability for retail automation, yet existing out-of-stock detection methods mainly operate in image space and lack metric 3D localization for downstream robotic systems. We formulate shelf monitoring as object-level 3D change detection: given two RGB-D observations captured at different times, the goal is to identify changed products and localize each change with a 3D bounding box. To support this task, we introduce ShelfChange3D, comprising 145K synthetic and 5K real-world paired RGB-D observations with object-level 3D change annotations. We further propose ChangeBox, an end-to-end framework that jointly reasons over paired observations and predicts object-level 3D change boxes. To improve localization accuracy, we introduce a geometry-based refinement stage that exploits depth and gravity prior to estimate relative pose and refine predicted boxes. Experiments show that ChangeBox outperforms existing change detection baselines, with further gains from refinement and effective transfer from synthetic to real-world observations.
Chinese Translation
可靠的货架监控是零售自动化的重要能力，但现有的缺货检测方法主要在图像空间中操作，缺乏用于下游机器人系统的度量3D定位。我们将货架监控形式化为对象级3D变化检测：给定在不同时间捕获的两个RGB-D观测，目标是识别变化的产品并用3D边界框定位每个变化。为支持这一任务，我们引入了ShelfChange3D，包含145K个合成和5K个真实世界的配对RGB-D观测及对象级3D变化注释。我们进一步提出了ChangeBox，一个端到端框架，能够对配对观测进行联合推理并预测对象级3D变化框。为了提高定位精度，我们引入了一个基于几何的精细化阶段，利用深度和重力先验来估计相对姿态并精细化预测框。实验表明，ChangeBox在变化检测基准测试中优于现有方法，并且通过精细化和有效的合成到真实世界观测的迁移获得了进一步的提升。
cs.CV / 88 / 2610.01286
Dyna3: VLM-Guided Training-Free 4D Reconstruction via Depth Foundation Models
Dyna3：基于视觉语言模型的无训练4D重建框架
Abstract
Recent depth foundation models like Depth Anything 3 (DA3) achieve remarkable multi-view depth estimation but assume static 3D scenes, limiting their applicability to real-world dynamic environments. Existing training-free 4D methods like Easi3R and VGGT4D rely on correspondence-trained backbones whose attention encodes cross-frame matching, a property absent in depth-only models like DA3. We present Dyna3, a training-free framework that extends DA3 for 4D dynamic scene reconstruction without any fine-tuning. Our key insight is that DA3's cross-view features, though trained only for depth consistency, implicitly encode motion-discriminative signals when combined with best-match feature search across frames. Its static surfaces find consistent matches globally, while dynamic objects cannot. We further adopt vision-language models (VLM) to automatically generate scene-specific semantic prompts for SAM 3, enabling precise instance-level segmentation that distinguishes which objects move from what objects exist. For reconstruction, we decouple the scene into a cross-frame aligned static background and per-frame dynamic point clouds. Experiments on four datasets demonstrate that Dyna3 surpasses correspondence-trained methods with +5.5pp J-Mean over state-of-the-art VGGT4D on dynamic object segmentation, while achieving up to 13x faster pose estimation and 3x faster 4D reconstruction with 4 to 8x lower memory. Dyna3 could therefore enable much denser temporal sampling that prior methods cannot support.
Chinese Translation
近期的深度基础模型如Depth Anything 3 (DA3)在多视角深度估计方面取得了显著的成果，但假设场景为静态三维，这限制了其在真实动态环境中的适用性。现有的无训练4D方法如Easi3R和VGGT4D依赖于经过对应训练的主干网络，其注意力机制编码了跨帧匹配的特性，而这一特性在仅处理深度的模型如DA3中并不存在。我们提出了Dyna3，这是一个无训练的框架，扩展了DA3以实现4D动态场景重建，无需任何微调。我们的关键见解是，尽管DA3的跨视图特征仅针对深度一致性进行训练，但在结合跨帧最佳匹配特征搜索时，隐含地编码了运动区分信号。其静态表面能够在全局范围内找到一致的匹配，而动态物体则无法做到。我们进一步采用视觉语言模型（VLM）自动生成场景特定的语义提示，以支持SAM 3，实现精确的实例级分割，区分哪些物体在移动，哪些物体是存在的。对于重建，我们将场景解耦为跨帧对齐的静态背景和每帧的动态点云。在四个数据集上的实验表明，Dyna3在动态物体分割方面超越了经过对应训练的方法，J-Mean提高了5.5个百分点，相较于最先进的VGGT4D，同时实现了高达13倍的姿态估计速度和3倍的4D重建速度，内存需求降低了4到8倍。因此，Dyna3能够支持比以往方法更密集的时间采样。
cs.CV / 89 / 2610.01291
ODDR: One-Step Deshadow Diffusion via Reward Guidance
ODDR：通过奖励引导的一步去阴影扩散
Abstract
Recent advances in deep learning for shadow removal have significantly enhanced image quality and realism. However, most approaches rely on real-world paired datasets, which are costly to collect and often limited in scene diversity, leading to limited generalization. To address these limitations, we propose One-step Deshadow Diffusion via Reward guidance (ODDR), a new framework that achieves efficient and high-fidelity shadow removal without relying on real-world paired supervision. Our method begins with One-step Deshadow Diffusion (ODD), a baseline model trained on synthetic shadow data for efficient one-step shadow-free reconstruction. We further adapt ODD into ODDR using ShadowReward. In contrast to traditional, annotation-heavy approaches, ShadowReward is the first reward model for shadow removal trained entirely without human annotation. It learns to mimic human perceptual judgments by ranking synthetically generated images with controlled degradations, such as texture distortion and boundary artifacts. This reward-guided fine-tuning enables ODDR to close the synthetic-to-real domain gap. Extensive experiments show that ODD achieves strong performance without relying on real-world paired supervision, and ODDR further improves the results, narrowing the gap to fully supervised methods trained on real-world paired data while maintaining higher computational efficiency as a single-step model.
Chinese Translation
近期在深度学习领域的去阴影技术取得了显著进展，极大地提升了图像质量和真实感。然而，大多数方法依赖于真实世界的配对数据集，这些数据集收集成本高且场景多样性有限，导致其泛化能力受限。为了解决这些局限性，我们提出了一种新的框架——通过奖励引导的一步去阴影扩散（ODDR），该框架实现了高效且高保真的去阴影处理，而无需依赖真实世界的配对监督。我们的方法以一步去阴影扩散（ODD）为基础模型，后者在合成阴影数据上进行训练，以实现高效的一步无阴影重建。我们进一步通过ShadowReward将ODD适配为ODDR。与传统的、依赖大量注释的方法不同，ShadowReward是第一个完全不依赖人工注释的去阴影奖励模型。它通过对合成生成的图像进行排名，模仿人类的感知判断，这些图像具有可控的降质特征，如纹理失真和边界伪影。这种奖励引导的微调使得ODDR能够缩小合成域与真实域之间的差距。大量实验表明，ODD在不依赖真实世界配对监督的情况下取得了强劲的性能，而ODDR进一步改善了结果，缩小了与基于真实世界配对数据训练的全监督方法之间的差距，同时作为单步模型保持了更高的计算效率。
cs.CV / 90 / 2610.01302
STAGE: Subspace-Targeted Affine Generative Erasure for Text-to-3D Models
STAGE：针对文本到3D模型的子空间目标仿射消除
Abstract
Concept erasure suppresses a target concept while preserving behavior on unrelated inputs. Existing closed-form methods were designed for 2D image diffusion and assume a single generative pathway, so one edit must cover geometry and texture at once. Native 3D generators, which synthesize structured 3D representations directly rather than by lifting 2D samples, violate this assumption. We show that shape and object concepts must be erased in the structural stage of the pipeline and material concepts in the appearance stage. We therefore formulate erasure in native text-to-3D as a stage-aware editing problem and introduce STAGE, a training-free, closed-form framework. STAGE confines each edit to the low-dimensional subspace spanned by the differences between erase and anchor embeddings, and relaxes the norm-preserving (orthogonal) constraint of prior editors into a least-squares affine correction that maps target activations onto safe anchors subject to a penalty on the displacement of retained prompts. The correction applies to the structural stage, the appearance stage, or both. We find that the stage an edit must reach is determined by concept type. On TRELLIS, the standard open native 3D generator, across 15 shape, material, and object concepts, STAGE reaches 66.7 on a composite score that balances forgetting the target concept against preserving everything else, aggregating CLIP-based semantic and physical metrics, versus 53.2 for the strongest adapted baseline. Code: https://github.com/gmum/STAGE/ Project Page https://gmum.github.io/STAGE/
Chinese Translation
概念消除在抑制目标概念的同时，保持对无关输入的行为。现有的封闭形式方法是为2D图像扩散设计的，并假设单一生成路径，因此一次编辑必须同时覆盖几何形状和纹理。原生3D生成器直接合成结构化的3D表示，而不是通过提升2D样本，这违反了这一假设。我们展示了形状和物体概念必须在管道的结构阶段被消除，而材料概念则在外观阶段被消除。因此，我们将原生文本到3D的消除形式化为一个阶段感知编辑问题，并引入了STAGE，一个无训练的封闭形式框架。STAGE将每次编辑限制在由消除和锚定嵌入之间的差异所张成的低维子空间，并将先前编辑器的范数保持（正交）约束放宽为最小二乘仿射校正，该校正将目标激活映射到安全锚点，并对保留提示的位移施加惩罚。该校正适用于结构阶段、外观阶段或两者。我们发现，编辑必须达到的阶段由概念类型决定。在TRELLIS上，标准开放原生3D生成器，在15个形状、材料和物体概念中，STAGE在一个综合评分上达到了66.7，该评分平衡了遗忘目标概念与保留其他所有内容，聚合了基于CLIP的语义和物理指标，而最强的适应基线则为53.2。代码：https://github.com/gmum/STAGE/ 项目页面：https://gmum.github.io/STAGE/
cs.CV / 91 / 2610.01314
ARROW: Arbitrary Reconstruction and Tracking of 4D Observations in the Wild
ARROW：野外环境中任意重建与跟踪4D观测
Abstract
Dynamic scenes may be captured by a moving camera, multiple video streams, or images taken at different times. These observations reveal complementary aspects of scene geometry and motion, yet bringing them together requires establishing correspondence across viewpoints, capture times, and visibility changes. We introduce ARROW, a feed-forward model that unifies 3D reconstruction and 3D point tracking from arbitrary image sets. At its core is a novel order-invariant querying approach, which allows the association of queries with observations across arbitrary inputs. We show that exposing the model to more diverse sets of inputs during training results in improved task performance. Moreover, the resulting model is capable of generalization to a wider range of tasks including multi-view tracking. Trained with this strategy, ARROW establishes a new state of the art in 3D tracking on WorldTrack and TAPVid-3D and outperforms dedicated multi-view trackers on an adapted RGB-only MVTracker benchmark, while remaining competitive across 3D reconstruction tasks. Code and weights are publicly available.
Chinese Translation
动态场景可以通过移动相机、多个视频流或在不同时间拍摄的图像进行捕捉。这些观测揭示了场景几何和运动的互补方面，但将它们结合在一起需要在视角、捕捉时间和可见性变化之间建立对应关系。我们提出了ARROW，这是一种前馈模型，统一了来自任意图像集的3D重建和3D点跟踪。其核心是一种新颖的无序不变查询方法，允许在任意输入之间将查询与观测关联起来。我们展示了在训练过程中将模型暴露于更具多样性的输入集可以提高任务性能。此外，所得到的模型能够推广到更广泛的任务，包括多视角跟踪。采用这种策略训练后，ARROW在WorldTrack和TAPVid-3D的3D跟踪任务上建立了新的技术领先水平，并在经过调整的仅RGB MVTracker基准上超越了专用的多视角跟踪器，同时在3D重建任务中保持竞争力。代码和权重已公开发布。
cs.CV / 92 / 2610.01331
CLASP: Continual Low-rank Adapters for Spatially Placed Concepts from One Hypernetwork
CLASP：用于从单一超网络中空间放置概念的持续低秩适配器
Abstract
Continual personalization of text-to-image diffusion models requires sequentially acquiring new concepts while retaining previously learned ones. However, existing methods either suffer from catastrophic forgetting or rely on storing additional concept-specific parameters and spatial components, causing their parameter footprint to grow with the concept stream. This limits their ability to scale to long sequences of personalization tasks. We propose a rehearsal-free approach that uses a single fixed-size hypernetwork to continually personalize a frozen diffusion model. Instead of expanding the model as new concepts are acquired, the hypernetwork dynamically produces the concept-specific adaptations required for personalization while preserving previously learned concepts. Our framework further integrates spatial control into the personalization process, allowing users to specify where a personalized concept should appear without introducing additional per-concept components. This formulation enables continual personalization with a parameter footprint that remains independent of the number of learned concepts, aside from compact concept representations. Experiments demonstrate strong retention of previously learned concepts and reliable spatial grounding, matching or improving upon existing methods while scaling effectively to long streams of personalization tasks.
Chinese Translation
文本到图像扩散模型的持续个性化需要在保留先前学习的概念的同时，顺序获取新概念。然而，现有方法要么遭受灾难性遗忘，要么依赖于存储额外的特定概念参数和空间组件，导致其参数占用随着概念流的增加而增长。这限制了它们在长序列个性化任务中的扩展能力。我们提出了一种无复习的方法，使用单一固定大小的超网络持续个性化一个冻结的扩散模型。在获取新概念时，不是扩展模型，而是超网络动态生成个性化所需的特定概念适配，同时保留先前学习的概念。我们的框架进一步将空间控制整合到个性化过程中，使用户能够指定个性化概念应出现的位置，而无需引入额外的每个概念组件。这种表述使得持续个性化的参数占用保持独立于学习概念的数量，除了紧凑的概念表示外。实验表明，先前学习的概念得到了很好的保留，并且空间定位可靠，匹配或改善了现有方法，同时有效扩展到长序列的个性化任务。
cs.CV / 93 / 2610.01352
MMVistaReason: Toward Open-Data and Post-Training Recipes for Multimodal Reasoning
MMVistaReason：面向开放数据和后训练方案的多模态推理
Abstract
Open multimodal reasoning models have benefited from large-scale reasoning supervision, yet reliable post-training remains challenging due to uneven data quality, inefficient supervision construction, imbalanced difficulty, and cross-domain interference. We introduce MMVistaReason (MVR), an open-data post-training recipe with three components: (1) broader capability coverage across complementary Analytical and Real-World reasoning groups, emphasizing structured reasoning versus visual perception and spatial grounding; (2) efficient SFT and RL data construction, standardizing heterogeneous open data through staged cleaning and annotation, combining difficulty-aware cascaded teacher distillation with answer-likelihood-based trajectory selection to construct MVR-SFT-528K, and applying scale-specific frontier filtering for MVR-RL-63K; and (3) specialize-then-integrate training, which trains complementary RL experts and consolidates their capabilities through multi-teacher on-policy distillation (MOPD). Our analyses reveal a capacity-dependent interaction between supervision difficulty, trajectory quality, and model capacity: smaller students benefit more from selected supervision, while larger students are robust to trajectory variation and mixed-domain interference. Mixed-domain RL introduces benchmark-level negative transfer, whereas MOPD provides consistent capability integration, with the preferred KL direction varying across model scales. Across 15 multimodal benchmarks, MVR-4B achieves an average score of 72.8, outperforming Qwen3.5-9B (Instruct) and MMFineReason-8B while using about 70% fewer samples than MMFineReason. Scaling to 9B improves the average to 74.4, surpassing Qwen3.5-35B-A3B (Instruct). Overall, MMVistaReason demonstrates that systematic open-data construction and capacity-aware post-training provide a practical and scalable path toward reliable multimodal reasoning.
Chinese Translation
开放的多模态推理模型受益于大规模的推理监督，但由于数据质量不均、监督构建效率低、难度不平衡以及跨域干扰，可靠的后训练仍然面临挑战。我们提出了MMVistaReason（MVR），这是一种开放数据后训练方案，包含三个组成部分：（1）在互补的分析性和现实世界推理组之间更广泛的能力覆盖，强调结构化推理与视觉感知和空间定位的对比；（2）高效的SFT（监督微调）和RL（强化学习）数据构建，通过分阶段清理和注释标准化异构开放数据，结合基于难度感知的级联教师蒸馏与答案可能性基础的轨迹选择构建MVR-SFT-528K，并为MVR-RL-63K应用特定规模的前沿过滤；（3）专业化后整合训练，训练互补的RL专家，并通过多教师在线政策蒸馏（MOPD）巩固其能力。我们的分析揭示了监督难度、轨迹质量和模型能力之间的容量依赖性互动：较小的学生从选定的监督中获益更多，而较大的学生对轨迹变化和混合域干扰具有鲁棒性。混合域RL引入基准级别的负迁移，而MOPD提供了一致的能力整合，且偏好的KL方向在不同模型规模间有所不同。在15个多模态基准测试中，MVR-4B的平均得分为72.8，超越了Qwen3.5-9B（指令）和MMFineReason-8B，同时使用的样本数量比MMFineReason少约70%。将规模扩展到9B使平均得分提高至74.4，超过了Qwen3.5-35B-A3B（指令）。总体而言，MMVistaReason展示了系统的开放数据构建和能力感知的后训练为可靠的多模态推理提供了一条实用且可扩展的路径。
cs.CV / 94 / 2610.01388
Supervising Sound Localization by In-the-wild Egomotion
通过野外自我运动监督声音定位
Abstract
We present a method for learning binaural sound localization using egomotion as a supervisory signal. Over the course of a video, the cameras direction to a sound source will change as the camera moves. We train an audio model to predict sound directions that are consistent with visual estimates of camera motion, which we obtain using traditional methods from multi-view geometry. This provides a weak but plentiful form of supervision that we combine with traditional binaural cues. To evaluate this method, we propose a dataset of real-world audio-visual videos with egomotion. We show that our model can successfully learn from real-world data and that it performs well on sound localization tasks
Chinese Translation
我们提出了一种利用自我运动作为监督信号来学习双耳声音定位的方法。在视频的过程中，随着相机的移动，相机指向声音源的方向会发生变化。我们训练一个音频模型来预测与相机运动的视觉估计一致的声音方向，这些估计是通过多视角几何的传统方法获得的。这提供了一种弱但丰富的监督形式，我们将其与传统的双耳线索相结合。为了评估该方法，我们提出了一个包含自我运动的真实世界音视频数据集。我们展示了我们的模型能够成功地从真实世界数据中学习，并且在声音定位任务上表现良好。
cs.CV / 95 / 2610.01408
Smoother Flow Matching via Contrastive Trajectory Repulsion
通过对比轨迹排斥实现更平滑的流匹配
Abstract
Trajectory crossing remains a critical bottleneck in Flow Matching (FM), and previous works typically view these crossings from a theoretical optimization perspective causing velocity averaging. They attempt to address it indirectly by post-hoc distillation or endpoint coupling, without explicitly regulating the intermediate trajectories. In this paper, we introduce a new network learning perspective: crossing points inherently induce large local Lipschitz constants in the target velocity field, leading to two drawbacks. First, high Lipschitz constants correspond to high-frequency signals in the velocity field that neural networks struggle to fit due to spectral bias. Second, they also imply drastic velocity variations, leading to severe numerical integration errors in few-step inference. To alleviate this, we propose CoFlow, a framework that introduces the contrastive learning paradigm into FM to explicitly repel trajectories during training, thereby lowering the local Lipschitz constants of the velocity field. Specifically, we formulate CoFlow from a Stochastic Differential Equation (SDE) perspective by injecting a repulsive drift term. This drift actively guides the forward process of positive samples away from negative trajectories, effectively reducing the local Lipschitz constant. Furthermore, we derive an equivalent stochastic interpolant formulation from this SDE, providing a simple and tractable design space to control the influence of negative samples. Extensive experiments on ImageNet 256x256 demonstrate that CoFlow significantly reduces FID compared to standard FM in few-step inference (e.g., 20 steps), with no added training overhead. The code can be accessed at: https://github.com/HKUST-LongGroup/CoFlow
Chinese Translation
轨迹交叉仍然是流匹配（Flow Matching, FM）中的一个关键瓶颈，以往的研究通常从理论优化的角度看待这些交叉，导致速度平均化。它们试图通过后期蒸馏或端点耦合间接解决这一问题，而没有明确地调节中间轨迹。本文引入了一种新的网络学习视角：交叉点本质上在目标速度场中引入了大的局部利普希茨常数，这导致了两个缺陷。首先，高利普希茨常数对应于速度场中的高频信号，神经网络由于谱偏差而难以拟合。其次，它们还意味着剧烈的速度变化，导致在少步推理中严重的数值积分误差。为了解决这个问题，我们提出了CoFlow，一个将对比学习范式引入FM的框架，以在训练过程中明确地排斥轨迹，从而降低速度场的局部利普希茨常数。具体而言，我们从随机微分方程（Stochastic Differential Equation, SDE）的角度构建CoFlow，通过注入排斥漂移项来实现。这种漂移主动引导正样本的前向过程远离负轨迹，有效降低局部利普希茨常数。此外，我们从该SDE推导出一个等效的随机插值公式，提供了一个简单且可控的设计空间来调节负样本的影响。在ImageNet 256x256上的大量实验表明，与标准FM相比，CoFlow在少步推理（例如20步）中显著降低了FID，并且没有增加训练开销。代码可访问：https://github.com/HKUST-LongGroup/CoFlow
cs.CV / 96 / 2610.01409
Localisation-Aware Uncertainty for Pretrained Object Detection
考虑定位的预训练目标检测的不确定性
Abstract
Reliable uncertainty estimation is essential for deploying object detectors when distribution/covariate shift and adversarial attacks may occur. Existing approaches often require detector retraining, architectural modification, or repeated inference, which may be infeasible or incur significant overheads. We introduce a lightweight post-hoc evidential meta-model that learns when object localisations should be considered uncertain while keeping the base detector frozen. Our approach automatically identifies localisation-relevant features and uses saliency-guided modification to construct an increasingly challenging curriculum. Detection-level targets combine localisation error, modification level, and prediction instability to guide an evidential meta-model to estimate uncertainty for each predicted bounding box. Our approach requires no changes to the detector and preserves its original localisation outputs. Across adversarial attacks and evaluated strengths, GRACE improves TP-FP AUROC by 22% relative to the strongest comparator in some cases while maintaining in-distribution detection performance.
Chinese Translation
可靠的不确定性估计对于在可能发生分布/协变量转移和对抗攻击的情况下部署目标检测器至关重要。现有的方法通常需要重新训练检测器、修改架构或进行重复推理，这可能不可行或导致显著的开销。我们提出了一种轻量级的后验证据元模型，该模型在保持基础检测器不变的情况下学习何时应考虑目标定位为不确定。我们的方法自动识别与定位相关的特征，并使用显著性引导的修改来构建越来越具挑战性的课程。检测级目标结合了定位误差、修改水平和预测不稳定性，以指导证据元模型为每个预测的边界框估计不确定性。我们的方法不需要对检测器进行任何更改，并保留其原始的定位输出。在对抗攻击和评估强度下，GRACE在某些情况下相对于最强对比者提高了22%的TP-FP AUROC，同时保持了分布内检测性能。
cs.CV / 97 / 2610.01434
MWOP: Modality-aware Width-wise Operation Pruning for Efficient MLLMs
MWOP：面向模态的宽度操作剪枝以提高多模态大语言模型的效率
Abstract
Multimodal large language models (MLLMs) incur substantial inference costs when processing long visual-textual sequences. While existing operation compression methods exploit modality-level redundancy, they largely treat computation within attention heads and shared feed-forward network (FFN) channels as unified units, leaving finer-grained redundancy underexplored. We find that redundancy varies both across modality-interaction paths within the same attention head and across visual and textual executions of the same FFN channel. Based on these findings, we propose Modality-aware Width-wise Operation Pruning (MWOP), which independently prunes visual-to-visual (V2V), text-to-visual (T2V), and text-to-text (T2T) attention paths within each layer, and separately selects FFN channels for visual and textual inputs. A first-order Taylor criterion guides the pruning process, with FFN importance re-evaluated after attention pruning and LoRA-based recovery training. To translate the resulting fine-grained sparsity into practical acceleration, we further develop path-sparse Triton attention kernels and compact visual-side FFN execution. MWOP preserves the token sequence while reducing attention and FFN computation, making it complementary to token compression and enabling simultaneous reduction of sequence length and per-token computation. On LLaVA-OneVision-7B, MWOP alone achieves a $1.6\times$ prefill speedup with 99.7\% average performance retention across 12 benchmarks. Combined with two representative token compression methods, it further increases their prefill speedups from $2.0\times$ and $1.9\times$ to $2.9\times$ and $2.7\times$, respectively. Results on Qwen2.5-VL-7B further demonstrate its applicability across architectures. The code is available at https://github.com/EIT-NLP/MWOP.
Chinese Translation
多模态大语言模型（MLLMs）在处理长视觉-文本序列时会产生巨大的推理成本。虽然现有的操作压缩方法利用了模态级别的冗余，但它们在很大程度上将注意力头和共享前馈网络（FFN）通道内的计算视为统一单元，从而使得更细粒度的冗余未得到充分探索。我们发现，冗余在同一注意力头内的模态交互路径之间以及同一FFN通道的视觉和文本执行之间存在差异。基于这些发现，我们提出了面向模态的宽度操作剪枝（MWOP），它在每一层内独立剪枝视觉到视觉（V2V）、文本到视觉（T2V）和文本到文本（T2T）注意力路径，并分别为视觉和文本输入选择FFN通道。使用一阶泰勒准则指导剪枝过程，在注意力剪枝和基于LoRA的恢复训练后重新评估FFN的重要性。为了将得到的细粒度稀疏性转化为实际加速，我们进一步开发了路径稀疏的Triton注意力内核和紧凑的视觉侧FFN执行。MWOP在减少注意力和FFN计算的同时保留了令牌序列，使其与令牌压缩互为补充，并实现序列长度和每个令牌计算的同时减少。在LLaVA-OneVision-7B上，MWOP单独实现了$1.6 imes$的预填充加速，同时在12个基准测试中保持了99.7\%的平均性能保留。与两种代表性的令牌压缩方法结合后，进一步将它们的预填充加速从$2.0 imes$和$1.9 imes$分别提高到$2.9 imes$和$2.7 imes$。在Qwen2.5-VL-7B上的结果进一步证明了其在不同架构中的适用性。代码可在https://github.com/EIT-NLP/MWOP获取。
cs.CV / 98 / 2610.01438
The Impact of Processing Parameters on High-Accuracy Measurements in UAV Photogrammetry
处理参数对无人机摄影测量高精度测量的影响
Abstract
Unmanned aerial vehicle (UAV) photogrammetry is increasingly used in applications requiring high accuracy, such as determining ground surface changes caused by landslides, mining, or microrelief transformation. While acquisition strategies have been widely studied, the influence of the processing workflow-particularly Bundle Block Adjustment parameter settings-remains insufficiently explored. This study addresses this gap through a systematic, full-factorial evaluation of 768 processing variants applied to ten UAV datasets collected over 1.5 years in a 220 ha study area. Eight key parameters were analysed. The results show substantial variability in final 3D accuracy: the best performing variant achieved a root mean square error (RMSE) of 16 mm, whereas the weakest reached 303 mm. The most influential factors were the number of ground control points, the application of additional camera calibration corrections, and the use of the Post-Processing Kinematic GNSS method for determining camera projection center coordinates. The study also evaluates how workflow optimization affects the accuracy of displacement, tilt changes, and horizontal strain determination. While random displacement errors remained stable (RMSE of ~6-7 mm), systematic errors were significantly reduced by over half in all axes, with vertical median absolute error decreasing from 14 mm to 7 mm in the optimized configuration compared to the baseline previously used by the authors. This study provides the first large-scale, practice-oriented assessment of how processing parameter selection shapes the accuracy of both photogrammetric products and deformation indices determination. The results offer actionable guidance for developing more robust and repeatable UAV photogrammetry workflows tailored to high-precision monitoring.
Chinese Translation
无人机（UAV）摄影测量在需要高精度的应用中越来越受到重视，例如确定因滑坡、采矿或微地形变化引起的地表变化。尽管获取策略已被广泛研究，但处理工作流的影响，特别是束块调整（Bundle Block Adjustment）参数设置，仍然未得到充分探讨。本研究通过对在一个220公顷的研究区域内收集的十个无人机数据集进行768种处理变体的系统全因子评估，填补了这一空白。这八个关键参数的分析结果显示最终三维精度存在显著差异：表现最佳的变体达到了16毫米的均方根误差（RMSE），而表现最差的则达到了303毫米。影响最大的因素包括地面控制点的数量、额外相机校准修正的应用，以及用于确定相机投影中心坐标的后处理运动学全球导航卫星系统（Post-Processing Kinematic GNSS）方法。研究还评估了工作流优化如何影响位移、倾斜变化和水平应变的精度。尽管随机位移误差保持稳定（RMSE约为6-7毫米），但所有轴向的系统误差显著减少，减少幅度超过一半，优化配置下的垂直中位绝对误差从14毫米降至7毫米，相较于作者之前使用的基线配置。本研究提供了首个大规模、以实践为导向的评估，探讨了处理参数选择如何影响摄影测量产品和变形指数的精度。结果为开发更强大和可重复的无人机摄影测量工作流提供了可操作的指导，以适应高精度监测的需求。
cs.CV / 99 / 2610.01452
Uncertainty-Guided Handshake: Efficient Human-in-the-Loop Refinement for Surgical-Grade Glioma Segmentation
不确定性引导的握手：高效的人机协作精细化用于外科级胶质瘤分割
Abstract
While state-of-the-art automated models for medical image segmentation achieve high mean performance, they frequently suffer from localized, catastrophic failures that preclude safe clinical deployment, particularly in neuro-oncology. Interactive segmentation frameworks mitigate this by incorporating human oversight, but traditionally impose prohibitive cognitive and temporal workloads by requiring clinicians to manually search for errors. In this project, we present an efficient, Hybrid Structural-Aleatoric Human-in-the-Loop framework for glioma segmentation that bridges the gap between automated baseline performance and surgical-grade precision, achieving sub-2.0 mm HD95 on curated benchmarks while providing safety-net routing for structural failures across real-world clinical data. By extracting voxel-wise Test-Time Augmentation (TTA) uncertainty and applying hierarchical topological filtering, our method proactively isolates high-risk structural anomalies. We comprehensively evaluated our approach on a challenging out-of-distribution clinical stress-test cohort (N = 362). Operating under a simulated Human Oracle, the framework improved the Whole Tumor (WT) Dice score from 0.891 to 0.914 and reduced the 95th percentile Hausdorff Distance (HD95) from 5.82 mm to 4.76 mm. Critically for surgical safety, the system rescued severe boundary failures in the Tumor Core, reducing mean HD95 from 17.96 mm to 14.83 mm (improving absolute TC Dice to 0.356). These spatial rescues were achieved while demanding a median interactive workload of just 11.3% of the target volume. Acknowledging this as a simulated upper bound lacking real-world cognitive friction, the framework nevertheless demonstrates a highly Pareto-efficient pathway for safely deploying clinical AI.
Chinese Translation
尽管最先进的医学图像分割自动化模型在平均性能上表现优异，但它们常常遭遇局部的灾难性失败，这使得在临床上安全应用变得困难，尤其是在神经肿瘤学领域。交互式分割框架通过引入人类监督来缓解这一问题，但传统上要求临床医生手动查找错误，导致认知和时间负担过重。在本项目中，我们提出了一种高效的混合结构-随机人机协作框架，用于胶质瘤分割，弥合了自动化基线性能与外科级精度之间的差距，在经过精心挑选的基准测试中实现了低于2.0 mm的HD95，同时为真实临床数据中的结构性失败提供了安全网路由。通过提取体素级测试时增强（Test-Time Augmentation, TTA）不确定性并应用分层拓扑过滤，我们的方法主动隔离高风险的结构异常。我们在一个具有挑战性的分布外临床压力测试队列（N = 362）上全面评估了我们的方法。在模拟人类Oracle的条件下，该框架将整体肿瘤（Whole Tumor, WT）Dice得分从0.891提高到0.914，并将第95百分位Hausdorff距离（HD95）从5.82 mm降低到4.76 mm。对于外科安全至关重要的是，该系统成功修复了肿瘤核心的严重边界失败，将平均HD95从17.96 mm降低到14.83 mm（绝对TC Dice提高至0.356）。这些空间修复是在仅要求目标体积中11.3%的中位交互工作量的情况下实现的。尽管承认这是一种缺乏现实世界认知摩擦的模拟上限，但该框架仍展示了一条高效的Pareto路径，安全地部署临床人工智能。
cs.CV / 100 / 2610.01480
FiVOS: A Fish Segmentation Algorithm Based on Interactive Video Object Segmentation and Filter Enhancement
FiVOS：基于交互式视频目标分割和滤波增强的鱼类分割算法
Abstract
With the continuous expansion of aquaculture, precise and efficient monitoring of fish behavior has become increasingly critical for improving farming efficiency and reducing economic losses. In particular, with the ongoing enhancement of computational capabilities in deep learning models, vision-based fish segmentation methods are garnering growing attention. By analyzing video segmentation results, fish behavior can be effectively tracked, thereby providing reliable data support for the precise regulation of aquaculture environments. However, existing deep learning-based video segmentation methods for aquaculture scenarios often overlook the dynamic correlations between video frames. In contrast, Interactive Video Object Segmentation (IVOS) employs an interaction-propagation scheme to achieve high-precision segmentation while minimizing user effort, thereby enhancing monitoring efficiency. Yet, IVOS applications in aquaculture remain limited due to data scarcity, and are susceptible to error accumulation and mask loss over long sequence propagation due to high intra-class similarity. In response, this paper proposes an improved interactive video object segmentation method (FiVOS) and constructs two fish-specific datasets. FiVOS utilizes a mask block filter to enable early detection and correction of erroneous propagated mask blocks, enhancing filtering accuracy through a rule-based thresholding approach. Additionally, it serializes noise filters to further eliminate erroneous mask noise, thereby improving model robustness. Experimental results demonstrate that FiVOS achieves state-of-the-art (SOTA) performance in fish video segmentation tasks, providing robust technical support for fish behavior research.
Chinese Translation
随着水产养殖的不断扩展，精确高效地监测鱼类行为变得愈加重要，以提高养殖效率并减少经济损失。特别是，随着深度学习模型计算能力的不断增强，基于视觉的鱼类分割方法受到越来越多的关注。通过分析视频分割结果，可以有效跟踪鱼类行为，从而为精确调控水产养殖环境提供可靠的数据支持。然而，现有的基于深度学习的视频分割方法在水产养殖场景中往往忽视了视频帧之间的动态关联。相比之下，交互式视频目标分割（Interactive Video Object Segmentation, IVOS）采用交互传播方案实现高精度分割，同时最小化用户的工作量，从而提高监测效率。然而，由于数据稀缺，IVOS在水产养殖中的应用仍然有限，并且由于高类内相似性，长序列传播过程中容易出现错误累积和掩膜丢失。为此，本文提出了一种改进的交互式视频目标分割方法（FiVOS），并构建了两个特定于鱼类的数据集。FiVOS利用掩膜块滤波器实现对错误传播的掩膜块的早期检测和修正，通过基于规则的阈值方法提高滤波精度。此外，它还串行化噪声滤波器，以进一步消除错误的掩膜噪声，从而提高模型的鲁棒性。实验结果表明，FiVOS在鱼类视频分割任务中实现了最先进的（SOTA）性能，为鱼类行为研究提供了强有力的技术支持。
cs.CV / 101 / 2610.01496
SALD: Self-Referenced Advantage Learning for Diffusion Models
SALD：用于扩散模型的自参考优势学习
Abstract
Recent work on language-model adaptation has shown that single models can obtain informative training signals by evaluating their behavior in demonstrationor feedback-augmented contexts, with the help of a teacher network, which is driven by the student's learned parameters. Inspired by this internal-reference principle, we investigate how diffusion models can identify self-referenced training signals without external demonstrations or teacher networks. We introduce SALD, a self-referenced training framework that evaluates each image-caption pair at two noise levels using the same model. The easier, lower-noise path is evaluated without gradient tracking to provide a reference, while the harder, higher-noise path provides the training gradient. Rather than directly distilling the easy-path prediction, SALD uses the difference between two path errors to adapt the hardpath objective. The proposed Advantage-Guided Diffusion (AGD) converts this relative error into a differentiable sample-level weight. Temporal Advantage Memory (TAM) accumulates relative difficulty across training and adapts the future gap between the two noise levels. Spectral Advantage Decomposition (SAD) further compares the residual power spectra of the two paths and constructs a differentiable, frequency-derived latent-element weight. All components share a single set of model parameters, requiring neither an external teacher network nor additional trainable parameters during training or inference, and no modification to the inference procedure. Experiments across multiple architectures and datasets demonstrate consistent improvements in generation quality, while component-wise ablations quantify the contributions of the proposed components.
Chinese Translation
近期关于语言模型适应性的研究表明，单一模型可以通过在示范或反馈增强的上下文中评估其行为，从教师网络中获得信息丰富的训练信号，而教师网络则由学生学习的参数驱动。受到这一内部参考原则的启发，我们研究了扩散模型如何在没有外部示范或教师网络的情况下识别自参考训练信号。我们提出了SALD，一种自参考训练框架，使用同一模型在两个噪声水平下评估每对图像-文本描述。较容易的低噪声路径在不进行梯度跟踪的情况下进行评估，以提供参考，而较困难的高噪声路径则提供训练梯度。SALD并不是直接提炼简单路径的预测，而是利用两条路径误差之间的差异来调整困难路径的目标。所提出的优势引导扩散（Advantage-Guided Diffusion, AGD）将这种相对误差转换为可微分的样本级权重。时间优势记忆（Temporal Advantage Memory, TAM）在训练过程中累积相对难度，并调整未来两个噪声水平之间的差距。频谱优势分解（Spectral Advantage Decomposition, SAD）进一步比较两条路径的残余功率谱，并构建可微分的、基于频率的潜在元素权重。所有组件共享一组模型参数，在训练或推理过程中不需要外部教师网络或额外的可训练参数，也不需要对推理过程进行修改。跨多个架构和数据集的实验表明生成质量的一致性提升，而逐个组件的消融实验则量化了所提组件的贡献。
cs.CV / 102 / 2610.01499
VTR-Bench: A Systematic Benchmark for Evaluating Visual Text Rendering in Video Generation
VTR-Bench：评估视频生成中视觉文本渲染的系统基准
Abstract
Recent video generation models can produce highly realistic videos from natural language instructions, with visual quality approaching cinematic standards. Existing evaluation benchmarks, however, predominantly assess visual quality, aesthetic appeal and physical plausibility, while paying limited attention to text, an essential medium for conveying information in everyday scenes. A generated video may appear visually compelling and feature lifelike subjects, yet still render the text within the scene incorrectly. To address this overlooked dimension, we introduce \textbf{VTR-Bench}, a systematic benchmark for evaluating the \textbf{V}isual \textbf{T}ext \textbf{R}endering capabilities of video generation models. VTR-Bench situates text within concrete application scenarios, such as advertisements and scientific videos, with 300 carefully constructed prompts spanning five scenario categories. We develop an automated evaluation pipeline with human alignments that separately assesses text fidelity through carrier-specific transcription and scene and motion requirements through a prompt-specific chain of query. Beyond evaluation, we introduce a \textbf{Keyframe-Guided Agentic Framework} in which a Director agent coordinates image and video generation with visual evaluation, guiding iterative refinement and candidate selection through visual feedback. Experiments on 11 state-of-the-art models reveal widespread difficulties in accurately rendering scene text, with the best-performing model recording an overall word error rate (WER) of 0.250. We further analyze text rendering failures to characterize the challenges faced by current video generation models. These findings highlight visual text rendering as a key challenge for video generation and demonstrate a practical path toward improvement. Code is available at https://github.com/hardenyu21/VTR-Bench.
Chinese Translation
近期的视频生成模型能够根据自然语言指令生成高度逼真的视频，视觉质量接近电影标准。然而，现有的评估基准主要评估视觉质量、美学吸引力和物理合理性，而对文本这一日常场景中传递信息的重要媒介关注有限。生成的视频可能在视觉上引人注目并且包含栩栩如生的主体，但场景中的文本仍可能渲染不正确。为了解决这一被忽视的维度，我们引入了 extbf{VTR-Bench}，一个用于评估视频生成模型 extbf{V}isual extbf{T}ext extbf{R}endering能力的系统基准。VTR-Bench将文本置于具体的应用场景中，如广告和科学视频，涵盖300个精心构建的提示，分为五个场景类别。我们开发了一个自动评估流程，结合人类对齐，分别通过载体特定的转录评估文本的保真度，并通过提示特定的查询链评估场景和运动要求。除了评估，我们还引入了一个 extbf{关键帧引导的代理框架}，其中导演代理协调图像和视频生成与视觉评估，通过视觉反馈指导迭代优化和候选选择。在对11个最先进模型的实验中，发现准确渲染场景文本存在广泛困难，表现最佳的模型的整体单词错误率（WER）为0.250。我们进一步分析文本渲染失败，以描述当前视频生成模型面临的挑战。这些发现突显了视觉文本渲染作为视频生成的关键挑战，并展示了改善的实际路径。代码可在https://github.com/hardenyu21/VTR-Bench获取。
cs.CV / 103 / 2610.01510
FedCKA: Representation-Guided Layer Personalization for Federated 3D Perception Across Driving Domains
FedCKA：基于表示引导的联邦3D感知层个性化方法
Abstract
Robust perception in intelligent vehicles demands 3D object detectors that remain dependable under domain shifts, such as changes in time of day, location, or weather. However, due to costly annotation and rare shifts, some environments lack sufficient data to train a standalone detector. Federated learning offers a privacy-preserving framework for collaborative model training, enabling clients to benefit from shared learning across diverse environments. Yet, this framework traditionally relies on a single global consensus model, which struggles to perform across heterogeneous local data distributions. Local conditions are better captured by adapting a subset of the model, but many personalization approaches rely on predefined layer partitions or fixed personalization ratios, thereby limiting adaptation to client-specific divergence. To reduce this rigidity, we propose FedCKA, a Centered Kernel Alignment (CKA)-based strategy that dynamically handles the personalization-globalization trade-off. Specifically, FedCKA computes layer-wise feature similarities between local client models and the global consensus model during training. By converting layer-wise similarity scores into client-specific aggregation masks, FedCKA selectively shares representation-consistent layers. Evaluation on a unified multi-domain benchmark based on nuScenes shows that FedCKA outperforms established federated baselines, including FedBN, FedRep, and FedSelect, improving average NDS by 7 percentage points over the strongest baseline. The findings offer both a comparative benchmark and a promising direction for robust federated 3D perception across shifts in location, weather, and illumination. Code is available at https://github.com/j-verhoog/FedCKA.
Chinese Translation
智能车辆的稳健感知要求3D物体检测器在领域转变（如时间、地点或天气变化）下依然可靠。然而，由于标注成本高昂和转变稀少，某些环境缺乏足够的数据来训练独立的检测器。联邦学习提供了一种保护隐私的协作模型训练框架，使得客户端能够从不同环境的共享学习中受益。然而，这一框架传统上依赖于单一的全局共识模型，在异构本地数据分布下表现不佳。通过适应模型的子集，可以更好地捕捉本地条件，但许多个性化方法依赖于预定义的层划分或固定的个性化比例，从而限制了对客户端特定差异的适应。为减少这种刚性，我们提出了FedCKA，一种基于中心核对齐（CKA）的策略，动态处理个性化与全局化之间的权衡。具体而言，FedCKA在训练过程中计算本地客户端模型与全局共识模型之间的层级特征相似性。通过将层级相似性得分转换为客户端特定的聚合掩码，FedCKA选择性地共享表示一致的层。在基于nuScenes的统一多领域基准测试中的评估表明，FedCKA的表现优于现有的联邦基线，包括FedBN、FedRep和FedSelect，平均NDS比最强基线提高了7个百分点。这些发现为在位置、天气和光照变化下稳健的联邦3D感知提供了比较基准和有前景的方向。代码可在 https://github.com/j-verhoog/FedCKA 获取。
cs.CV / 104 / 2610.01512
VoxelSynth3D: Interpretable Volumetric Image-Domain Metal Artifact Reduction with a Paired Synthetic CLINIC-Metal Benchmark
VoxelSynth3D：可解释的体积图像域金属伪影减少方法，配备成对合成CLINIC-Metal基准
Abstract
Metal artifacts in postoperative musculoskeletal CT obscure bone-implant and adjacent soft-tissue interfaces. Many metal artifact reduction (MAR) methods require unavailable raw projections or learned models that may shift across scanners and implants. We present VoxelSynth3D, a training-free 3D image-domain framework for reconstructed CT. The framework combines support masking, normalized tissue synthesis, deviation gating, and restricted edge refinement. Detected implant voxels are preserved in the output, while correction targets metal-induced artifacts in the surrounding tissue. We also construct Synthetic CLINIC-Metal, a controlled paired synthetic evaluation resource, from no-metal CTPelvic1K volumes with clean targets, metal/artifact masks, fixed seeds, and patient-level splits; 75 unpaired real metal cases receive qualitative/no-reference evaluation only. The operating point was fixed in a near-flat validation basin. With exact-mask oracle localization, all methods share a metal-excluded tissue ROI. On 40 held-out cases, VoxelSynth3D reduced RMSE from 801.48 to 786.18 HU (paired gain 15.30 HU, 95% CI 11.68-19.23), improving every case and exceeding the evaluated 3D Gaussian smoother by 13.58 HU. Clean-edge agreement decreased next to metal but exceeded input beyond 5 mm. Thus, VoxelSynth3D provides case-consistent within-distribution tissue-error reduction with a localized structural tradeoff. Spacing-aware sensitivity retained aggregate broad-region improvement and identified near-metal calibration as a target.
Chinese Translation
术后肌肉骨骼CT中的金属伪影遮蔽了骨-植入物及相邻软组织的界面。许多金属伪影减少（MAR）方法需要不可用的原始投影或可能在扫描仪和植入物之间变化的学习模型。我们提出了VoxelSynth3D，这是一种无需训练的3D图像域框架，用于重建CT。该框架结合了支持掩膜、标准化组织合成、偏差门控和受限边缘细化。检测到的植入物体素在输出中得以保留，同时修正目标是周围组织中的金属诱导伪影。我们还构建了Synthetic CLINIC-Metal，这是一个受控的成对合成评估资源，基于无金属的CTPelvic1K体积，包含干净目标、金属/伪影掩膜、固定种子和患者级别的划分；75个未配对的真实金属案例仅接受定性/无参考评估。操作点固定在接近平坦的验证盆地中。在精确掩膜的oracle定位下，所有方法共享一个排除金属的组织ROI。在40个保留案例中，VoxelSynth3D将均方根误差（RMSE）从801.48降低到786.18 HU（配对增益15.30 HU，95%置信区间11.68-19.23），改善了每个案例，并超越了评估的3D高斯平滑器13.58 HU。干净边缘的一致性在金属附近下降，但在超过5 mm的范围内超过了输入。因此，VoxelSynth3D提供了一致的案例内分布组织误差减少，并具有局部结构权衡。空间感知灵敏度保持了整体广域改善，并将近金属校准识别为目标。
cs.CV / 105 / 2610.01517
SuperMotion: Source-Preserving Denoising for Text-Driven Human Motion Editing
超运动：基于文本驱动的人体运动编辑的源保持去噪
Abstract
Text-driven human motion editing aims to realize a requested change while preserving compatible source content. Existing diffusion editors rely largely on learned conditioning for preservation of the unedited part, yet their outputs can lose temporal detail as denoising proceeds. We propose the \textbf{Source-Preserving Denoising framework (SuperMotion)}, which explicitly reuses the source at each reverse step for source preservation. We first align the source motion to the output timeline and predict a preservation gate that controls reuse across frames and feature dimensions. A clean-space source anchor then utilizes the learned preservation gate to blend the predicted clean motion with the aligned source and passes the corrected estimate directly to the sampling posterior. Because the aligned source is a realized motion rather than a regression output, the anchor injects sample-level temporal detail that a reconstruction-trained denoiser tends to smooth away. To learn effective source reuse, we supervise the anchored estimate against the editing target and match its second temporal differences through a temporal high-frequency loss. These objectives require no explicit edit masks. Extensive experiments show that SuperMotion improves editing accuracy, reaching 33.20\% full-pool R@1 on MotionFix, while reducing temporal-detail attenuation and preserving motion dynamics as it realizes the requested changes. Ablations confirm that the learned preservation gate is responsible for the gain and that it reuses the source to retain the unedited content properly.
Chinese Translation
基于文本驱动的人体运动编辑旨在实现所请求的变化，同时保持兼容的源内容。现有的扩散编辑器在保持未编辑部分方面主要依赖学习到的条件，但随着去噪的进行，其输出可能会失去时间细节。我们提出了 extbf{源保持去噪框架（SuperMotion）}，该框架在每个反向步骤中显式重用源以实现源的保持。我们首先将源运动与输出时间线对齐，并预测一个控制跨帧和特征维度重用的保持门。然后，一个干净空间的源锚点利用学习到的保持门将预测的干净运动与对齐的源进行混合，并将修正后的估计直接传递给采样后验。由于对齐的源是实现的运动而不是回归输出，因此锚点注入了样本级的时间细节，而重建训练的去噪器往往会将其平滑掉。为了学习有效的源重用，我们对锚定估计进行监督，以匹配编辑目标，并通过时间高频损失匹配其第二时间差。这些目标不需要显式的编辑掩码。大量实验表明，SuperMotion提高了编辑准确性，在MotionFix上达到了33.20 ext{%}的全池R@1，同时减少了时间细节的衰减，并在实现所请求的变化时保持了运动动态。消融实验确认，学习到的保持门负责这种提升，并且它重用源以正确保留未编辑内容。
cs.CV / 106 / 2610.01542
Synthetic training for long-tail haemorrhagic lesion segmentation in data-scarce settings
数据稀缺环境下长尾出血病灶分割的合成训练
Abstract
Cerebral microbleeds (CMBs) and cortical superficial siderosis (cSS) are imaging markers of cerebral small vessel disease, but their automated segmentation is limited by the scarcity of positive cases and voxel-level annotations. We propose a synthetic training framework for long-tail haemorrhagic lesion segmentation that requires no real lesion annotations for training and leverages radiological description of the lesions. Starting from anatomical brain parcellations, the framework applies spatial augmentation and voxel resampling, procedurally inserts cSS and CMB labels using clinical priors on lesion location and morphology, and synthesises images through randomised intensity assignment, blurring, and Rician noise simulation. Models were trained on dynamically generated image-label pairs and evaluated against manual delineations in 10 cSS cases and 13 CMB cases. The proposed configurations outperformed classical filter baselines. For cSS, the hypointensity constrained model achieved higher AUPRC and AUROC than the Frangi filter (AUPRC: 0.284 vs 0.083; AUROC: 0.907 vs 0.731). For CMBs, explicit synthesis of blood vessels as lesion mimics improved performance over the classical baseline (AUPRC: 0.538 vs 0.004; AUROC: 0.999 vs 0.968). These results support our proposal as a feasible strategy for data-scarce haemorrhagic lesion segmentation.
Chinese Translation
脑微出血（CMBs）和皮层表面铁沉积（cSS）是脑小血管疾病的影像学标志，但由于阳性病例和体素级标注的稀缺，其自动分割受到限制。我们提出了一种用于长尾出血病灶分割的合成训练框架，该框架在训练过程中无需真实病灶标注，并利用病灶的放射学描述。该框架从解剖脑分区开始，应用空间增强和体素重采样，基于临床先验知识在病灶位置和形态上程序性地插入cSS和CMB标签，并通过随机强度分配、模糊处理和Rician噪声模拟合成图像。模型在动态生成的图像-标签对上进行训练，并在10个cSS病例和13个CMB病例中与手动描绘进行评估。所提出的配置在性能上优于经典滤波基线。对于cSS，限制低强度的模型在AUPRC和AUROC上均优于Frangi滤波器（AUPRC: 0.284 vs 0.083; AUROC: 0.907 vs 0.731）。对于CMB，明确合成血管作为病灶模拟物的方式在性能上超过了经典基线（AUPRC: 0.538 vs 0.004; AUROC: 0.999 vs 0.968）。这些结果支持我们提出的作为数据稀缺环境下出血病灶分割的可行策略。
cs.CV / 107 / 2610.01544
Revisiting Cross-Reconstruction for Generalizable Deepfake Detection
重新审视用于可泛化深度伪造检测的交叉重建
Abstract
Existing image forgery detectors often suffer from generalization to unseen manipulation methods due to the limited ability to capture transferable forensic cues. Recent cross-reconstruction based methods attempt to improve generalization through semantic-artifact disentanglement, but typically align heterogeneous artifacts across generators and exclude artifact representations during reconstruction, which may overlook the inherent diversity and visual cues of manipulation artifacts. In this work, we revisit cross-reconstruction and introduce an artifact-oriented disentanglement framework for robust image forgery detection. We argue that \textbf{artifact diversity}, i.e., the intrinsic variations of manipulation artifacts introduced by different generation processes, contains complementary forensic cues rather than undesirable domain variations. Instead of enforcing explicit artifact alignment, our framework preserves diverse artifact characteristics through semantically aligned cross-generator reconstruction. Furthermore, we incorporate artifact representations into the reconstruction process and introduce a masked frequency-aware reconstruction strategy to emphasize manipulation-related residuals while reducing semantic interference. This design enables the model to learn transferable forensic representations from diverse artifacts. Extensive experiments on multiple benchmark datasets demonstrate improvements under both cross-dataset and cross-generator evaluation settings. Further analysis and ablation studies validate the effectiveness of artifact diversity preservation and artifact-aware cross-reconstruction.
Chinese Translation
现有的图像伪造检测器往往由于捕捉可转移法医学线索的能力有限而在未见过的操控方法上表现不佳。最近基于交叉重建的方法试图通过语义伪影解缠来提高泛化能力，但通常在生成器之间对异质伪影进行对齐，并在重建过程中排除伪影表示，这可能忽视了操控伪影的内在多样性和视觉线索。在本研究中，我们重新审视交叉重建，并引入一种面向伪影的解缠框架，以实现稳健的图像伪造检测。我们认为， extbf{伪影多样性}，即不同生成过程引入的操控伪影的内在变化，包含互补的法医学线索，而不是不良的领域变化。我们的框架通过语义对齐的跨生成器重建保留了多样的伪影特征，而不是强制执行显式的伪影对齐。此外，我们将伪影表示纳入重建过程，并引入一种掩蔽频率感知重建策略，以强调与操控相关的残差，同时减少语义干扰。该设计使模型能够从多样的伪影中学习可转移的法医学表示。在多个基准数据集上的广泛实验表明，在跨数据集和跨生成器评估设置下均有改进。进一步的分析和消融研究验证了伪影多样性保留和伪影感知交叉重建的有效性。
cs.CV / 108 / 2610.01589
PAGER: Partial-to-global Alignment via Geometric and Relational Distillation
PAGER：通过几何和关系蒸馏实现局部到全局的对齐
Abstract
Pretrained 3D encoders are typically developed on globally reconstructed scenes expressed in a consistent world coordinate frame, whereas embodied systems must reason from partial, viewpoint-dependent observations in camera coordinates. We show that this shift from globally learned 3D feature spaces to realistic partial observations exposes a severe representation mismatch, which we find consistently across representative state-of-the-art encoders, including Sonata and Concerto. A frozen Sonata encoder with a global linear probe achieves 72.47 mIoU on full ScanNet scenes, but 2.57 mIoU on single-frame camera-coordinate inputs. Training-free gravity alignment recovers performance to 41.64 mIoU, showing that coordinate-frame mismatch is a dominant source of degradation but cannot be fully resolved through canonicalization alone. We introduce PAGER, a label-free adaptation method that aligns partial-view features with a frozen global 3D semantic space using only paired partial/global geometry. It learns lightweight adaptation modules while keeping the pretrained encoder and global segmentation probe frozen. Matched-point feature alignment anchors partial features to their global counterparts, while relational supervision preserves their similarity structure with respect to the global representation. Global geometry provides supervision only during training. Inference operates directly on the partial observation. Without partial-view labels, PAGER outperforms label-supervised PEFT on both Sonata and Concerto, and in zero-shot ScanNet$\rightarrow$ScanNet++ transfer surpasses fully fine-tuned Sonata ($53.93$ vs.\ $48.09$ mIoU), suggesting that preserving the frozen global representation can improve cross-dataset transfer.
Chinese Translation
预训练的3D编码器通常是在以一致的世界坐标系表达的全局重建场景上开发的，而具身系统必须从相机坐标中的局部、视角依赖的观测中进行推理。我们展示了从全局学习的3D特征空间到现实局部观测的转变暴露出严重的表示不匹配，这一现象在包括Sonata和Concerto在内的代表性最先进编码器中一致存在。一个冻结的Sonata编码器在完整的ScanNet场景上通过全局线性探针达到了72.47 mIoU，但在单帧相机坐标输入上仅为2.57 mIoU。无训练的重力对齐将性能恢复到41.64 mIoU，表明坐标系不匹配是性能下降的主要来源，但仅通过规范化无法完全解决。我们引入了PAGER，这是一种无标签适应方法，利用配对的局部/全局几何将局部视图特征与冻结的全局3D语义空间对齐。它在保持预训练编码器和全局分割探针冻结的同时学习轻量级适应模块。匹配点特征对齐将局部特征锚定到其全局对应物，而关系监督则保持它们相对于全局表示的相似结构。全局几何仅在训练期间提供监督，推理直接在局部观测上进行。在没有局部视图标签的情况下，PAGER在Sonata和Concerto上都优于标签监督的PEFT，并且在零-shot ScanNet$
ightarrow$ScanNet++转移中超过了完全微调的Sonata（$53.93$ vs. $48.09$ mIoU），这表明保持冻结的全局表示可以改善跨数据集的迁移。
cs.CV / 109 / 2610.01595
Before It Fades: Reinforcing Temporal Representations at Inference Time in VideoLLMs
在其消退之前：在推理时强化视频大语言模型中的时间表示
Abstract
Video Large Language Models (VideoLLMs) receive frames in sequential order and interpret how visual content evolves along the temporal axis, yet temporal reasoning remains a persistent weakness across architectures. Reversing the frame order of a video, a transformation that should invert temporal answers, often leaves the final prediction unchanged. We investigate where this failure originates by defining the temporal divergence vector $\tau_l$, the layer-wise representational difference induced by reversing temporal order. Tracking its magnitude across layers reveals a consistent temporal divergence profile where the divergence peaks at intermediate layers and progressively diminishes toward the output. We confirm this peak is specific to temporal reasoning and functionally critical for predictions, establishing that VideoLLMs acquire temporal information at intermediate layers but fail to maintain it to the output. This progressive fading motivates our method, Temporal Activation Injection (TAI), which extracts $\tau_l$ at the peak of the profile for each input and reinjects it into subsequent layers following the measured decay. TAI requires no training and consistently improves temporal reasoning across three VideoLLMs and four benchmarks with negligible impact on non-temporal tasks. Code is available at https://github.com/Youngwoo-git/Before-It-Fades.
Chinese Translation
视频大语言模型（VideoLLMs）以顺序方式接收帧，并解释视觉内容如何沿时间轴演变，但时间推理在各个架构中仍然是一个持续的弱点。反转视频的帧顺序，这一转换应该会颠倒时间答案，通常会导致最终预测不变。我们通过定义时间发散向量 $ au_l$，即由于反转时间顺序而引起的层级表示差异，来研究这种失败的来源。跟踪其在各层的幅度揭示了一个一致的时间发散特征，其中发散在中间层达到峰值，并逐渐向输出层减小。我们确认这一峰值特定于时间推理，并对预测至关重要，确立了 VideoLLMs 在中间层获取时间信息但未能将其保持到输出的事实。这种逐渐消退促使我们提出了一种方法，时间激活注入（Temporal Activation Injection，TAI），该方法在每个输入的特征峰值处提取 $ au_l$，并在随后的层中根据测量的衰减重新注入。TAI 不需要训练，并在三个 VideoLLMs 和四个基准测试中持续改善时间推理，对非时间任务的影响微乎其微。代码可在 https://github.com/Youngwoo-git/Before-It-Fades 获取。
cs.CV / 110 / 2610.01605
Hob-VL: A Benchmark for Visually Grounded Boolean Reasoning
Hob-VL：一个视觉基础布尔推理的基准
Abstract
Reliable visual reasoning requires composing multiple visual observations and returning consistent answers to logically equivalent questions. We introduce Hob-VL, a benchmark for visually grounded Boolean reasoning. Hob-VL comprises two tasks: (1) evaluating whether a Boolean rule holds in an image, and (2) identifying the (unique) object satisfying a Boolean description. Hob-VL contains 6,000 human-verified balanced Yes/No questions, each defined by a Boolean combination of ten visual statements, across 1,000 generated scenes and 46 diverse labeled photographs, along with 1,000 object-identification questions over the same photographs. Our question families are deliberately constructed to challenge reasoning through misleading local cues and nested logical operations, and include symbolic and structured natural-language presentations. Across eight model configurations with thinking disabled or minimized, Boolean accuracy ranges from 48.52% to 50.57%, while the identification accuracy reaches at most 43.0%. A thinking-enabled GLM configuration achieves uneven gains while retaining substantial errors and inconsistencies. Hob-VL exposes these failures through executable reference answers and matched evaluations.
Chinese Translation
可靠的视觉推理需要组合多个视觉观察，并对逻辑上等价的问题返回一致的答案。我们介绍了Hob-VL，这是一个用于视觉基础布尔推理的基准。Hob-VL包含两个任务：（1）评估布尔规则在图像中是否成立，以及（2）识别满足布尔描述的（唯一）对象。Hob-VL包含6,000个经过人工验证的平衡是/否问题，每个问题由十个视觉陈述的布尔组合定义，覆盖1,000个生成场景和46张多样化的标注照片，以及1,000个关于相同照片的对象识别问题。我们的题目家族经过精心构建，以挑战通过误导性的局部线索和嵌套逻辑操作进行推理，并包括符号和结构化的自然语言呈现。在八种模型配置中，思维被禁用或最小化，布尔准确率范围为48.52%到50.57%，而识别准确率最高达到43.0%。启用思维的GLM配置在保持大量错误和不一致的同时实现了不均匀的增益。Hob-VL通过可执行的参考答案和匹配评估揭示了这些失败。
cs.CV / 111 / 2610.01614
Oneira: From Open-Ended Generation to Open-World Interaction in Video World Models
Oneira：从开放式生成到视频世界模型中的开放世界交互
Abstract
Generative video world models can now synthesize open-ended environments that agents can navigate and interact with in simple ways. Yet open-ended generation does not imply full interaction: as a generated world expands, newly created content through navigation should expand what the agent can act upon, and as the agent changes the world, those changes should become persistent parts of the environment rather than transient visual effects. We characterize these two requirements as Open-World Interactivity, where newly generated or encountered entities are incorporated into the actionable world, and Persistent State, where interaction outcomes are committed to the world state and continue to influence subsequent observations and interactions. We present Oneira, an interactive video world model that closes the loop between generation and interaction through an explicit, extensible world state managed by a coding agent. Given the current observation and an action or high-level goal, the agent reads the world state, grounds the relevant entities, plans the interaction, and writes its outcome back into a world state table. When exploration reveals new objects, the agent incorporates them from generated observations, allowing the interaction space to expand with the generated world. Meanwhile, previously induced state changes are carried across video segments, making the consequences of interaction persistent parts of subsequent world evolution. The updated world state is rendered along the camera action trajectory into a coarse conditioning video, from which a video generator fills in the appearance, motion, and interaction details not represented in the state. Experiments show that Oneira enables direct and consistent interaction with newly generated objects, while preserving the effects of prior interactions over long horizons. Project page: https://madaoer.github.io/projects/oneira
Chinese Translation
生成性视频世界模型现在能够合成开放式环境，代理可以以简单的方式进行导航和交互。然而，开放式生成并不意味着完全的交互：随着生成世界的扩展，通过导航新创建的内容应该扩展代理可以作用的范围，并且随着代理改变世界，这些变化应该成为环境的持久部分，而不是短暂的视觉效果。我们将这两个要求描述为开放世界交互性（Open-World Interactivity），即新生成或遇到的实体被纳入可操作的世界，以及持久状态（Persistent State），即交互结果被记录到世界状态中，并继续影响后续的观察和交互。我们提出了Oneira，一个交互式视频世界模型，通过一个由编码代理管理的显式、可扩展的世界状态，闭合生成与交互之间的循环。给定当前观察和一个动作或高层目标，代理读取世界状态，确定相关实体，规划交互，并将其结果写回世界状态表。当探索揭示新对象时，代理从生成的观察中纳入它们，使得交互空间随着生成的世界而扩展。同时，先前引发的状态变化在视频片段之间传递，使得交互的后果成为后续世界演变的持久部分。更新后的世界状态沿着相机动作轨迹渲染为粗略的条件视频，视频生成器则填补状态中未表示的外观、运动和交互细节。实验表明，Oneira能够与新生成的对象进行直接且一致的交互，同时保持先前交互在长时间范围内的效果。项目页面：https://madaoer.github.io/projects/oneira
cs.CV / 112 / 2610.01625
Beyond Domain-Level Adaptation: Margin-Oriented Semantic-Appearance Interaction Correction for Personalized Federated Vision-Language Models
超越领域级适应：面向边际的语义-外观交互修正用于个性化联邦视觉-语言模型
Abstract
Federated parameter-efficient fine-tuning enables distributed clients to adapt pretrained vision-language models without sharing raw data or updating the full backbone. Its effectiveness, however, is limited by domain heterogeneity across clients. Existing personalized methods separate globally shared knowledge from client-specific style, but they largely treat each domain as a class-agnostic transformation. We show that this abstraction is insufficient: the cross-domain displacement associated with a fixed domain varies across semantic classes, and only a subset of these class-domain residuals damages the image-text decision margin. We therefore propose Margin-Oriented Semantic-Appearance Interaction Correction (MOSAIC), which first constructs a decision-aware harmfulness score that measures whether a training-derived class-domain residual favors a competing text prototype over the true class. It then models fine-grained class-domain interactions with a low-rank residual adapter whose class factors and residual basis are globally shared while domain factors remain client-private. An image-conditioned gate further controls candidate-wise correction, and harmful-pair-aware reweighting prioritizes decision-relevant residuals during local optimization. Extensive experiments on Office31, OfficeHome, and DomainNet100 demonstrate that MOSAIC consistently improves macro-client top-1 accuracy across all evaluated domain-shift and joint domain-label-shift settings.
Chinese Translation
联邦参数高效微调使得分布式客户端能够在不共享原始数据或更新完整主干的情况下适应预训练的视觉-语言模型。然而，其有效性受到客户端之间领域异质性的限制。现有的个性化方法将全球共享知识与客户端特定风格分开，但它们在很大程度上将每个领域视为与类别无关的变换。我们表明，这种抽象是不够的：与固定领域相关的跨领域位移在语义类别之间是不同的，只有这些类别-领域残差的子集会损害图像-文本决策边际。因此，我们提出了面向边际的语义-外观交互修正（Margin-Oriented Semantic-Appearance Interaction Correction，MOSAIC），该方法首先构建一个决策感知的有害性评分，用于衡量训练得到的类别-领域残差是否偏向于竞争文本原型而非真实类别。然后，它通过一个低秩残差适配器建模细粒度的类别-领域交互，其中类别因子和残差基是全球共享的，而领域因子则保持客户端私有。一个图像条件门进一步控制候选修正，而有害对优先重加权在局部优化过程中优先考虑与决策相关的残差。在Office31、OfficeHome和DomainNet100上的大量实验表明，MOSAIC在所有评估的领域迁移和联合领域标签迁移设置中始终提高了宏客户端的Top-1准确率。
cs.CV / 113 / 2610.01637
Fusing Visual and Textual Representations via Multi-layer Fusing Transformers for Vietnamese Visual Question Answering
通过多层融合变换器融合视觉和文本表示用于越南视觉问答
Abstract
In recent decades, artificial intelligence has made significant progress in understanding and interacting with images. One of the important applications of this technology is Visual Question Answering (VQA), a research field that requires computers to understand and answer questions about images in a natural manner. Despite extensive research and development in VQA for English, there have been very few similar efforts made for other languages, especially Vietnamese. This gap presents a significant challenge and opportunity for the advancement of VQA technology in the Vietnamese language context. By bridging this gap, the field of Vietnamese VQA not only enriches the diversity of research in artificial intelligence but also enables practical applications in various domains, such as education, healthcare, and entertainment, catering to Vietnamese-speaking populations worldwide. Thus, the exploration and development of Vietnamese VQA systems hold immense potential for advancing both research and practical applications in the intersection of computer vision and natural language processing. In this paper, we propose a Multi-layer Fusing Transformer model utilizing a cross attention module to combine multiple modality features of images and texts from different layers in an aggregated representation. Our architecture allows us extract information from low level to high level. Through detailed experiments and ablation studies, our model achieves promising results against the competitive baselines in ViVQA dataset for Vietnamese language.
Chinese Translation
近年来，人工智能在理解和与图像互动方面取得了显著进展。这项技术的一个重要应用是视觉问答（Visual Question Answering, VQA），这是一个研究领域，要求计算机以自然的方式理解和回答关于图像的问题。尽管在英语的 VQA 方面进行了广泛的研究和开发，但对于其他语言，尤其是越南语，类似的努力却寥寥无几。这一差距为越南语言环境下 VQA 技术的进步带来了重大挑战和机遇。通过弥补这一差距，越南 VQA 领域不仅丰富了人工智能研究的多样性，还使其在教育、医疗和娱乐等多个领域的实际应用成为可能，服务于全球讲越南语的人群。因此，探索和开发越南 VQA 系统在计算机视觉与自然语言处理交叉领域的研究和实际应用方面具有巨大的潜力。在本文中，我们提出了一种多层融合变换器模型，利用交叉注意力模块将来自不同层次的图像和文本的多模态特征结合成一个聚合表示。我们的架构允许我们从低层次提取信息到高层次。通过详细的实验和消融研究，我们的模型在越南语的 ViVQA 数据集上取得了优于竞争基线的良好结果。
cs.CV / 114 / 2610.01640
Not All Error Yields to Scale: Where Scaling Stops in Vision-Language Inference
并非所有错误都能通过规模解决：视觉-语言推理中的规模限制
Abstract
Vision-language models (VLMs) face a fixed-budget trade-off between processing more visual information for fine-grained perception and using a larger language backbone for complex reasoning. Existing studies do not tell us which combination of backbone size and input resolution to deploy, especially in high-resolution deployments. To address this gap, we propose the Separable Law that describes how VLM performance changes with language backbone size and visual token count. We fit the law to measurements from 26 InternVL and QwenVL models, with language backbone sizes from 1B to 72B, on four high-resolution benchmarks with image sizes from 224 pixels to 8K. We find that the questions responding to scaling can be predicted from the skill they require, while a substantial fraction never responds at all. We also find that the two model families gain similarly from a larger backbone, while their gains from more visual tokens differ sharply. Combined with a cost law, the Separable Law gives a closed-form rule for allocating compute between backbone size and visual tokens. When deployment is limited to available configurations, the law identifies model and image sizes that perform close to the best feasible choice under the same budget. We hope our work offers a principled way to decide how much a model should be allowed to see at high resolution, given what it must reason about.
Chinese Translation
视觉-语言模型（VLMs）面临着在处理更多视觉信息以实现细粒度感知与使用更大语言骨干进行复杂推理之间的固定预算权衡。现有研究未能告诉我们在高分辨率部署中应采用哪种骨干大小和输入分辨率的组合。为了解决这一问题，我们提出了可分离法则（Separable Law），该法则描述了VLM性能如何随语言骨干大小和视觉标记数量的变化而变化。我们将该法则拟合到来自26个InternVL和QwenVL模型的测量数据，这些模型的语言骨干大小从1B到72B不等，且在四个高分辨率基准测试中，图像大小从224像素到8K不等。我们发现，响应规模变化的问题可以根据所需技能进行预测，而相当一部分问题根本不做出响应。我们还发现，这两种模型系列在使用更大骨干时获得的收益相似，但它们在更多视觉标记上的收益差异显著。结合成本法则，可分离法则为在骨干大小和视觉标记之间分配计算资源提供了封闭形式的规则。当部署受到可用配置的限制时，该法则能够识别在相同预算下接近最佳可行选择的模型和图像大小。我们希望我们的工作为在高分辨率下决定模型应允许看到多少信息提供了一种有原则的方法，考虑到模型必须推理的内容。
cs.CV / 115 / 2610.01661
DiVid: Diagnosing Dimension-Specific Diversity Collapse in Video Generation Models
DiVid：诊断视频生成模型中特定维度的多样性崩溃
Abstract
Despite remarkable progress, video generation models often produce highly similar outputs when repeatedly sampled from the same prompt, limiting their usefulness for creative exploration. Existing diversity evaluations primarily rely on global scalar metrics, which obscure where diversity collapses in the spatiotemporal space of videos. We introduce DiVid, a dimension-level diagnostic framework that decomposes video generation diversity into six interpretable dimensions: Semantic, Style, Subject, Scene, Motion, and Camera. Each dimension is measured through a reproducible computer-vision pipeline and analyzed alongside quality and instruction faithfulness to examine potential trade-offs. Systematic evaluation of representative video generation models reveals that diversity is highly dimension-specific: models with strong global diversity scores still collapse on specific factors, particularly Motion and Camera. These rankings persist after filtering unfaithful generations, indicating genuine capability differences rather than off-prompt outputs. Beyond measurement, controlled prompt interventions identify two fundamental bottlenecks: default mode convergence, where models fall back to dominant patterns under open-ended prompts; and realization gaps, where models fail to faithfully realize diverse, explicitly requested alternatives, particularly for temporal factors. The larger faithfulness losses for temporal factors highlight the difficulty of controlling motion and camera variation through text alone. DiVid thus shifts the study of diversity from measuring whether it exists to diagnosing where and why it collapses, and provides actionable directions for dimension-aware training objectives and control signals. The framework will be released to facilitate future research on diverse and controllable video generation.
Chinese Translation
尽管取得了显著进展，视频生成模型在从相同提示中反复采样时往往会产生高度相似的输出，这限制了其在创意探索中的实用性。现有的多样性评估主要依赖于全局标量指标，这掩盖了视频时空空间中多样性崩溃的具体位置。我们提出了DiVid，一个维度级诊断框架，将视频生成多样性分解为六个可解释的维度：语义（Semantic）、风格（Style）、主题（Subject）、场景（Scene）、运动（Motion）和相机（Camera）。每个维度通过可重复的计算机视觉流程进行测量，并与质量和指令忠实度一起分析，以检查潜在的权衡。对代表性视频生成模型的系统评估揭示了多样性高度依赖于特定维度：即使是具有强全局多样性评分的模型，在特定因素上仍然会崩溃，尤其是在运动和相机方面。这些排名在过滤掉不忠实生成后仍然存在，表明存在真正的能力差异，而非偏离提示的输出。除了测量，受控提示干预识别出两个基本瓶颈：默认模式收敛，即模型在开放式提示下回归到主导模式；以及实现差距，即模型未能忠实实现多样化的、明确请求的替代方案，特别是在时间因素上。时间因素的忠实度损失更大，突显了仅通过文本控制运动和相机变化的困难。因此，DiVid将多样性的研究从测量其是否存在转向诊断其崩溃的具体位置和原因，并为基于维度的训练目标和控制信号提供可行的方向。该框架将被发布，以促进未来在多样化和可控视频生成方面的研究。
cs.CV / 116 / 2610.01670
Do MLLM Judges Judge the Edit? Auditing Bias in Image Editing Evaluation with Verified Quality Preservation
多模态大语言模型评审者评判编辑吗？通过验证质量保持审计图像编辑评估中的偏见
Abstract
Multimodal large language models (MLLMs) are increasingly used as automated judges for instruction-based image editing and as reward signals for model training. However, systematically auditing whether these judges are influenced by cues irrelevant to editing quality is challenging because visual interventions may themselves alter the quality being evaluated. A judgment shift can therefore be attributed to bias only when the intervention is verified to preserve the underlying editing quality. To address this challenge, we introduce EditJudgeBias, a counterfactual benchmark with verified quality preservation, comprising 1,196 real editing samples and 13 cues injected across four evaluation sites. We verify quality preservation for the requested edit using calibrated multimodal validators, controls, and human inspection. We then audit five MLLM judges along three complementary dimensions: invariance to quality-preserving cues, agreement with human judgments, and stability of pairwise preferences. Importantly, observed shifts are evaluated against each judge's own zero-dose and re-query noise floors rather than against zero. Experiments show that quality-preserving cues move every judge beyond its own noise. Fabricated majority opinions increase ratings, irrelevant visual elements cause larger shifts than whole-image manipulations, and swapping candidate order reverses up to 60.9% of pairwise decisions. Edit-region cues also tend to reduce human agreement. The three measures characterize judges differently, showing that robustness cannot be captured by a single metric.
Chinese Translation
多模态大语言模型（MLLMs）越来越多地被用作基于指令的图像编辑的自动评审者以及模型训练的奖励信号。然而，系统性审计这些评审者是否受到与编辑质量无关的线索影响是具有挑战性的，因为视觉干预可能会改变被评估的质量。因此，只有在干预被验证为保持基础编辑质量时，才能将判断偏移归因于偏见。为了解决这一挑战，我们引入了EditJudgeBias，这是一个具有验证质量保持的反事实基准，包含1196个真实编辑样本和在四个评估站点注入的13个线索。我们使用校准的多模态验证器、控制组和人工检查来验证请求编辑的质量保持。然后，我们沿着三个互补维度审计五个MLLM评审者：对质量保持线索的不变性、与人类判断的一致性以及成对偏好的稳定性。重要的是，观察到的偏移是相对于每个评审者自身的零剂量和重新查询噪声底线进行评估，而不是相对于零。实验表明，质量保持线索使每个评审者超越其自身的噪声。伪造的多数意见提高了评分，无关的视觉元素导致的偏移大于整体图像操作，而交换候选顺序则使多达60.9%的成对决策发生逆转。编辑区域线索也倾向于减少人类一致性。这三种测量方式对评审者的特征化不同，表明稳健性无法通过单一指标来捕捉。
cs.CV / 117 / 2610.01681
When Text-to-Image Helps Editing: The Effects of Conditioning During Denoising
文本到图像如何帮助编辑：去噪过程中条件化的影响
Abstract
Unified models are trained for both instruction-based image editing and text-to-image (T2I) generation, but standard editing pipelines keep source-image conditioning throughout denoising. We ask whether editing can benefit from T2I, and study how the effects of conditioning vary across edits and denoising stages. In pure editing, source attention declines for some edits over the sampling trajectory. This observation led us to task switching, which lets the model draw on its T2I capabilities. Across three unified editors and four benchmarks, switching to the T2I task for bounded intervals improves edit quality, while mean perceptual preservation remains close to pure editing across all three models. Unified editors therefore benefit from using both conditioning modes they are trained for, and the timing of the switch sets the balance between quality and preservation.
Chinese Translation
统一模型同时用于基于指令的图像编辑和文本到图像（T2I）生成，但标准编辑流程在去噪过程中始终保持源图像的条件化。我们探讨编辑是否可以从T2I中受益，并研究条件化在不同编辑和去噪阶段的效果如何变化。在纯编辑中，源注意力在某些编辑的采样轨迹上会下降。这一观察促使我们进行任务切换，使模型能够利用其T2I能力。在三个统一编辑器和四个基准测试中，在有限的时间间隔内切换到T2I任务可以提高编辑质量，同时在所有三个模型中，平均感知保留仍然接近纯编辑。因此，统一编辑器从使用它们所训练的两种条件化模式中受益，而切换的时机则决定了质量与保留之间的平衡。
cs.CV / 118 / 2610.01687
Architectural Sampling: Test-Time Scaling via Computational Diversity in Frozen Vision-Language Models
架构采样：通过冻结的视觉-语言模型中的计算多样性实现测试时扩展
Abstract
Test-time scaling often seeks better answers by sampling multiple responses from a frozen model, yet conventional temperature sampling generates every candidate along the same fixed computation path. We introduce architectural sampling, a training-free method that generates candidates through distinct forward computations by reusing selected blocks of decoder layers. Varying the block location and repetition count introduces computational diversity without updating model weights or adding auxiliary parameters. Across five Qwen checkpoints and twelve multimodal benchmarks, architectural sampling improves pass@9 over standard-path temperature sampling by 6.58 percentage points on average at the same nine-candidate budget. Reusing early layers yields the strongest gains, and the improvement in candidate coverage persists even under greedy decoding. The resulting candidates show lower lexical overlap and improve accuracy when used as rollouts for label-free test-time reinforcement learning. These findings extend the benefits of our architectural sampling beyond candidate coverage, demonstrating more effective learning from a model's own outputs.
Chinese Translation
测试时扩展通常通过从冻结模型中采样多个响应来寻求更好的答案，然而传统的温度采样是在同一固定计算路径上生成每个候选项。我们提出了一种架构采样的方法，这是一种无训练的技术，通过重用选定的解码器层块，通过不同的前向计算生成候选项。改变块的位置和重复次数引入了计算多样性，而无需更新模型权重或添加辅助参数。在五个 Qwen 检查点和十二个多模态基准测试中，架构采样在相同的九个候选预算下，平均提高了标准路径温度采样的 pass@9 6.58 个百分点。重用早期层带来了最显著的提升，即使在贪婪解码下，候选覆盖的改善也依然存在。生成的候选项显示出较低的词汇重叠，并在作为无标签测试时强化学习的回滚使用时提高了准确性。这些发现将我们的架构采样的好处扩展到候选覆盖之外，展示了从模型自身输出中更有效的学习。
cs.CV / 119 / 2610.01707
MEGA: Object-Level Mesh Extraction from 3D Gaussian Splatting via Spatial Visual Distillation
MEGA：通过空间视觉蒸馏从3D高斯点云提取物体级网格
Abstract
Mesh extraction from 3D Gaussian Splatting (3DGS) aims to endow 3D Gaussians with accurate geometric structures, enabling explicit and precise 3D occupancy. However, existing methods primarily focus on scene-level mesh extraction, making them unable to represent object-level occupancy and often resulting in non-watertight surfaces. To overcome these limitations, we propose \textbf{MEGA} (\underline{M}esh \underline{E}xtraction from \underline{GA}ussians), a ``segment-then-mesh'' framework for extracting object-level, watertight meshes from complex 3DGS scenes. At the core of MEGA are \textbf{Spatial Visual Distillation (SVD)} and a mask-guided neural surface reconstruction module. SVD treats the 3DGS model as a teacher, sampling diverse camera poses and rendering the corresponding views of each segmented object. These observations are then used to train a mesh reconstruction model through photometric supervision. Extensive experiments on several widely used benchmarks demonstrate that MEGA achieves state-of-the-art performance in recovering accurate object-level 3D occupancy. Moreover, MEGA enables complex physical interactions by combining high-quality object-level meshes for geometric occupancy with 3DGS representations for photorealistic rendering.
Chinese Translation
从3D高斯点云（3DGS）提取网格的目标是为3D高斯赋予准确的几何结构，从而实现明确和精确的3D占用表示。然而，现有方法主要集中在场景级网格提取，无法表示物体级占用，且常常导致非水密表面。为克服这些局限性，我们提出了 extbf{MEGA}（ extunderline{M}esh extunderline{E}xtraction from extunderline{GA}ussians），一种“先分割再网格”的框架，用于从复杂的3DGS场景中提取物体级水密网格。MEGA的核心是 extbf{空间视觉蒸馏（SVD）}和一个掩模引导的神经表面重建模块。SVD将3DGS模型视为教师，采样多样的相机姿态并渲染每个分割物体的相应视图。这些观察结果随后用于通过光度监督训练网格重建模型。在多个广泛使用的基准测试上的大量实验表明，MEGA在恢复准确的物体级3D占用方面达到了最先进的性能。此外，MEGA通过将高质量的物体级网格与3DGS表示结合，实现了复杂的物理交互，支持几何占用和照片级真实感渲染。
cs.CV / 120 / 2610.01723
Rethinking Memorization Mitigation in Diffusion Models: Reinforcing Text Conditioning
重新思考扩散模型中的记忆减缓：强化文本条件
Abstract
Text-to-image diffusion models have achieved remarkable progress in image synthesis, yet can exhibit memorization by closely reproducing individual training examples. Effective mitigation must preserve useful prompt information to guide alternative depictions. We introduce a training-free method that redistributes cross-attention with Gaussian smoothing before reinforcing content-token contributions and attenuating padding contributions, without additional denoiser evaluations. With this intervention, stronger content conditioning can improve prompt alignment at comparable training-image similarity. A local analysis identifies when reinforcement preserves shared value information while redistribution reduces localized attention mass. On Stable Diffusion v1.4 and v2.0, all evaluated smoothing widths lie on the empirical Pareto frontiers for training-image similarity versus both prompt alignment and image preference. A configuration selected on Stable Diffusion reduces template reproduction in DeepFloyd IF without further tuning. These findings support jointly controlling conditioning allocation and strength to generate prompt-consistent alternatives.
Chinese Translation
文本到图像的扩散模型在图像合成方面取得了显著进展，但可能会通过紧密重现个别训练示例而表现出记忆效应。有效的减缓措施必须保留有用的提示信息，以指导替代图像的描绘。我们提出了一种无训练的方法，在强化内容标记贡献和减弱填充贡献之前，通过高斯平滑重新分配交叉注意力，而无需额外的去噪器评估。通过这一干预，增强的内容条件可以在与训练图像相似度相当的情况下改善提示对齐。局部分析识别出何时强化能够保留共享价值信息，而重新分配则减少了局部注意力的集中。在 Stable Diffusion v1.4 和 v2.0 上，所有评估的平滑宽度均位于训练图像相似度与提示对齐及图像偏好之间的经验帕累托前沿上。在 Stable Diffusion 上选择的配置减少了 DeepFloyd IF 中的模板重现，而无需进一步调整。这些发现支持共同控制条件分配和强度，以生成与提示一致的替代方案。
cs.CV / 121 / 2610.01741
ATI-VLA: Action-Centric Predictive Vision-Language-Action Models via Actionable Alignment Then Adaptive Injection
ATI-VLA：通过可操作对齐然后自适应注入的以行动为中心的预测视觉-语言-行动模型
Abstract
Predictive Vision-Language-Action (VLA) models aim to improve robotic manipulation via future observation or world dynamics forecasting. However, existing approaches often fail to realize this potential and underperform direct action prediction models. We argue that these limitations stem from modality misalignment between observations and actions, together with joint optimization conflicts that drive learning away from an action-centric objective. To this end, we introduce ATI-VLA, an Action-Centric Predictive Vision-Language-Action framework via Actionable Alignment Then Adaptive Injection. Specifically, it follows a two-step design: 1) Actionable Representation Alignment via a Shared Codebook. It aligns predictive observation and action representations by mapping both modalities into a shared discrete latent space via a unified codebook, making predictive observation latents readily usable for action generation and mitigating modality misalignment. 2) Action-Centric Adaptive Injection of Predictive Latents. Building upon this, it then injects predictive observation latents into action decoding as explicit predictive priors via a lightweight adaptive side-path, enabling adaptive predictive guidance under a single action-centric objective. Extensive experiments on both simulation and real-world robotic tasks demonstrate that ATI-VLA achieves state-of-the-art performance with faster convergence.
Chinese Translation
预测视觉-语言-行动（VLA）模型旨在通过未来观察或世界动态预测来改善机器人操作。然而，现有的方法往往未能实现这一潜力，且表现不及直接行动预测模型。我们认为，这些局限性源于观察与行动之间的模态不对齐，以及驱动学习偏离以行动为中心目标的联合优化冲突。为此，我们提出了ATI-VLA，一种通过可操作对齐然后自适应注入的以行动为中心的预测视觉-语言-行动框架。具体而言，它遵循两步设计：1）通过共享词汇表进行可操作表示对齐。它通过将预测观察和行动表示映射到一个共享的离散潜在空间，利用统一的词汇表来对齐两种模态，使得预测观察潜在变量可以直接用于行动生成，并减轻模态不对齐。2）以行动为中心的预测潜在变量自适应注入。在此基础上，它通过轻量级自适应侧路径将预测观察潜在变量注入到行动解码中，作为显式的预测先验，从而在单一的以行动为中心目标下实现自适应预测指导。在模拟和真实世界机器人任务上的大量实验表明，ATI-VLA实现了最先进的性能，并且收敛速度更快。
cs.CV / 122 / 2610.01744
3DROID: A Renderable 3D Gaussian Dataset with Measured Per-Scene Reliability
3DROID：一个可渲染的3D高斯数据集，具有逐场景可靠性测量
Abstract
Robot manipulation models primarily reason from 2D observations while acting in the 3D physical world. To bridge this gap, recent work has augmented robot data with geometric priors such as depth, point clouds, and 3D trajectories, while renderable 3D Gaussian representations provide another promising form of 3D supervision. However, 3DGS representation is designed mainly for photometric fidelity and may not preserve real-world metric scale, particularly when the supplied camera extrinsics are unreliable. We study the effect of extrinsic reliability and pose conditioning on feed-forward 3DGS, and propose a calibration-aware pipeline that anchors reconstructed scenes to the robot's metric workspace. Our experiments show that pose conditioning improves novel-view fidelity, while its geometric benefit depends on the reliability of the injected extrinsics. Using this pipeline, we present a renderable, metric-pose-anchored dataset with scene-level reliability information for robot manipulation research. Our dataset is available at https://huggingface.co/datasets/wonguen/3DROID
Chinese Translation
机器人操作模型主要基于2D观察进行推理，而在3D物理世界中执行操作。为了弥补这一差距，最近的研究通过几何先验（如深度、点云和3D轨迹）增强了机器人数据，而可渲染的3D高斯表示提供了另一种有前景的3D监督形式。然而，3DGS表示主要是为了光度保真而设计，可能无法保持真实世界的度量尺度，特别是在提供的相机外部参数不可靠时。我们研究了外部参数可靠性和姿态条件对前馈3DGS的影响，并提出了一种校准感知的流程，将重建场景锚定到机器人的度量工作空间。我们的实验表明，姿态条件改善了新视角的保真度，而其几何效益依赖于注入外部参数的可靠性。利用该流程，我们呈现了一个可渲染的、度量姿态锚定的数据集，包含用于机器人操作研究的场景级可靠性信息。我们的数据集可在 https://huggingface.co/datasets/wonguen/3DROID 获取。
cs.CV / 123 / 2610.01750
FFBL-Coop: Association-Decoupled Cooperative 3D Multi-Object Tracking
FFBL-Coop：关联解耦的协作三维多目标跟踪
Abstract
Cooperative 3D tracking must integrate complementary observations across agents and time while maintaining consistent identities. When evidence integration and identity inheritance share a matching decision, errors arising from cross-view appearance differences and spatial misalignment can compromise both feature fusion and track continuity. We propose FFBL-Coop, a fuse first, bind later framework that separates instance admission from identity management. Confidence-ranked Slot Admission (CSA) allocates cooperative queries to available ego slots using confidence and spatial proximity. Unified Representation Aggregation (URA) uses cooperative semantic features and aligned anchors to guide ego-feature retrieval, refining the augmented query bank within a shared transformer decoder. After refinement, Cooperative-Priority Identity Anchoring (CPIA) combines learned association with persistent mappings to establish accepted identity assignments across frames. A shared codebook reduces transmitted payload while retaining AP and AMOTA close to the uncompressed variant. FFBL-Coop achieves AMOTA/AP of 0.611/0.548 on V2X-Seq and 0.688/0.653 on Griffin-25M. Code will be released.
Chinese Translation
协作三维跟踪必须整合跨代理和时间的互补观察，同时保持身份的一致性。当证据整合和身份继承共享匹配决策时，由于视角间外观差异和空间错位引起的错误可能会影响特征融合和轨迹连续性。我们提出了FFBL-Coop，一个先融合后绑定的框架，旨在将实例接纳与身份管理分开。信心排名的槽接纳（Confidence-ranked Slot Admission, CSA）利用信心和空间接近性将协作查询分配给可用的自我槽。统一表示聚合（Unified Representation Aggregation, URA）使用协作语义特征和对齐锚点来指导自我特征检索，在共享的变换解码器中细化增强的查询库。经过细化后，协作优先身份锚定（Cooperative-Priority Identity Anchoring, CPIA）结合学习到的关联与持久映射，以在帧之间建立接受的身份分配。共享代码本减少了传输负载，同时保持接近未压缩变体的AP和AMOTA。FFBL-Coop在V2X-Seq上实现了AMOTA/AP为0.611/0.548，在Griffin-25M上实现了0.688/0.653。代码将会发布。
cs.CV / 124 / 2610.01754
Cog-VADU: A Training-Free Cognitive Reasoning Framework for Video Anomaly Detection and Understanding
Cog-VADU：一种无训练的认知推理框架用于视频异常检测与理解
Abstract
Video Anomaly Detection (VAD) aims to temporally localize abnormal events in videos. Most existing approaches rely on dataset-specific training and curated annotations, limiting generalization in open-set scenarios. Recent zero-shot methods based on Large Vision- Language Models (LVLMs) alleviate this dependency but often lack temporal continuity and structured reasoning. We propose Cog-VADU, a fully training-free framework that reformulates VAD as a sequential cognitive reasoning task. Cog-VADU introduces Chain-of- Anomaly Detection Thought Prompting (CoADTP), which unrolls an LVLM into a recurrent reasoning chain across video segments. By propagating structured rationales over time, the model maintains implicit temporal memory, enabling robust discrimination between com- plex anomalies and high-motion normal activities. To improve reliability, we further design a cross-modal re-ranking stage that aligns textual rationales with visual embeddings, enforcing semantic consistency and temporal coherence for refined and stable predictions. Extensive experiments on multiple public VAD benchmarks demonstrate that Cog-VADU achieves competitive zero-shot performance. Moreover, cross-model evaluations show that CoADTP consistently enhances reasoning-based anomaly detection in a model-agnostic manner, pro- viding interpretable and generalizable anomaly understanding for real-world applications.
Chinese Translation
视频异常检测（VAD）旨在对视频中的异常事件进行时间上的定位。大多数现有方法依赖于特定数据集的训练和精心策划的注释，这限制了其在开放集场景中的泛化能力。最近基于大型视觉-语言模型（LVLMs）的零-shot 方法缓解了这种依赖，但往往缺乏时间连续性和结构化推理。我们提出了Cog-VADU，一个完全无训练的框架，将VAD重新构建为一个顺序认知推理任务。Cog-VADU引入了异常检测思维提示链（CoADTP），将LVLM展开为一个跨视频片段的递归推理链。通过在时间上传播结构化的推理，模型保持隐式的时间记忆，使其能够在复杂异常和高运动正常活动之间进行稳健的区分。为了提高可靠性，我们进一步设计了一个跨模态重排序阶段，将文本推理与视觉嵌入对齐，强制语义一致性和时间连贯性，以实现更精细和稳定的预测。在多个公共VAD基准上的广泛实验表明，Cog-VADU实现了具有竞争力的零-shot 性能。此外，跨模型评估显示，CoADTP在模型无关的方式下持续增强基于推理的异常检测，为现实世界应用提供了可解释和可泛化的异常理解。
cs.CV / 125 / 2610.01758
GenCOPE: Syn2Real Generalized Category-Level Object Pose Estimation for Robotic Picking
GenCOPE：用于机器人抓取的合成到真实的广义类别级物体姿态估计
Abstract
Category-level object pose estimation (COPE), capable of generalizing to intra-class unknown objects, has become a core technique for robotic 3D scene understanding. However, existing COPE methods still require labor-intensive recollection of real-world training data for novel object categories, which limits their scalability in practical applications. This paper aims to achieve synthetic-to-real (Syn2Real) generalized COPE, where a model is trained solely on rendered synthetic data and directly generalized to real-world deployments. The central challenge lies in the significant domain gap between synthetic and real-world data, particularly in texture appearance. To address this, we aim to enhance domain generalization by learning domain-invariant representations that capture semantic commonalities among objects within the same category. We introduce 2D and 3D semantic consistency constraints to reduce the sensitivity of feature encoders to domain-specific features. In addition, we propose an end-to-end pose regression framework that performs 2D-3D cross consistency learning, leveraging dense cross-modality fusion to further refine pose estimation. Since simplicity and effectiveness are essential for real-world robotic deployment, our model operates exclusively on global features, yielding a highly lightweight and efficient architecture. Extensive experiments on the REAL275 and Wild6D benchmarks, as well as real-world robotic manipulation scenes, show superior Syn2Real generalization performance of our paradigm. Code and demos are released at https://paperreview99.github.io/GenCOPE/.
Chinese Translation
类别级物体姿态估计（COPE）能够推广到类内未知物体，已成为机器人三维场景理解的核心技术。然而，现有的COPE方法仍然需要劳动密集型的真实世界训练数据重新收集以适应新物体类别，这限制了其在实际应用中的可扩展性。本文旨在实现合成到真实（Syn2Real）的广义COPE，其中模型仅在渲染的合成数据上进行训练，并直接推广到真实世界的部署。核心挑战在于合成数据与真实数据之间显著的领域差距，特别是在纹理外观方面。为了解决这个问题，我们旨在通过学习领域不变的表示来增强领域泛化能力，这些表示捕捉同一类别中物体之间的语义共性。我们引入了二维和三维语义一致性约束，以减少特征编码器对领域特定特征的敏感性。此外，我们提出了一种端到端的姿态回归框架，执行二维-三维交叉一致性学习，利用密集的跨模态融合进一步优化姿态估计。由于简单性和有效性对真实世界的机器人部署至关重要，我们的模型仅在全局特征上操作，从而实现了高度轻量和高效的架构。在REAL275和Wild6D基准测试以及真实世界机器人操作场景上的大量实验表明，我们的方法在Syn2Real泛化性能上表现优越。代码和演示已发布在 https://paperreview99.github.io/GenCOPE/。
cs.CV / 126 / 2610.01759
PhysDEM: Physics-Defined Energy-Matching Diffusion for Spatiotemporal Field Generation under Scarce Measurements
PhysDEM：基于物理定义的能量匹配扩散用于稀疏测量下的时空场生成
Abstract
Generating and predicting spatiotemporal physical fields from scarce measurements is challenging, as observations are insufficient to characterize a distribution over complete fields. This limits conventional data-driven diffusion models that rely on full-field datasets. We introduce PhysDEM, a physics-defined diffusion framework that combines governing equations with spatially sparse observations to generate multiple plausible fields. First, we construct a Gibbs target by reweighting a measurement-conditioned Gaussian reference with PDE residual energy. Second, we derive an exact conditional-mean identity that reduces denoising to supervised learning of the standardized energy-induced mean correction. Third, a physics-displacement probability flow cancels Gaussian reference terms and enables amortized sampling with changing measurements through Gaussian conditioning, without retraining. Experiments on synthetic PDE systems and real-world-informed applications demonstrate that PhysDEM supports coherent field recovery and efficient sampling while maintaining stable diagnostics under tested noise levels, illustrating its practical value for field assessment. To our knowledge, PhysDEM is the first physics-defined diffusion model enabling amortized spatiotemporal field inference without preassembled full-field datasets.
Chinese Translation
从稀疏测量中生成和预测时空物理场是一项挑战，因为观察数据不足以表征完整场的分布。这限制了依赖全场数据集的传统数据驱动扩散模型。我们提出了PhysDEM，一种基于物理定义的扩散框架，结合了控制方程和空间稀疏观察，以生成多个合理的场。首先，我们通过用偏微分方程（PDE）残差能量重新加权测量条件的高斯参考构建了一个Gibbs目标。其次，我们推导出一个精确的条件均值恒等式，将去噪问题简化为对标准化能量引起的均值修正的监督学习。第三，物理位移概率流抵消了高斯参考项，并通过高斯条件实现了在变化测量下的摊销采样，而无需重新训练。在合成PDE系统和基于真实世界的应用实验中，PhysDEM展示了其支持一致场恢复和高效采样的能力，同时在测试的噪声水平下保持稳定的诊断，体现了其在场评估中的实际价值。据我们所知，PhysDEM是首个基于物理定义的扩散模型，能够在没有预先组装的全场数据集的情况下实现摊销的时空场推断。
cs.CV / 127 / 2610.01762
OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction
OneStreamer：统一流媒体视频交互中的感知、记忆和主动响应
Zeng, Xiangyu, Yang, Yuandong, Zhang, Zhiqiu, Zhu, Yuhan, Li, Xinhao, Si, Qingyi, Yao, Dingyu, Ma, Changlian, Chen, Haoran, Chen, Xinyu, Shi, Yansong, Zhou, Junhao, Li, Yifei, Zhang, Jun, Qin, Chuanyu, Yang, Chenxu, Yu, Xinlei, Ouyang, Kun, Shao, Yuchen, Wei, Qianshan, Zhou, Changhai, Gao, Jun, Wang, Jiaqi, Wang, Limin
Abstract
Streaming video LLMs must retain evidence before its relevance to future tasks is known and respond when sufficient evidence becomes available. The challenge is to form reusable factual memory without compromising real-time perception. We introduce OneStreamer, which jointly learns query-independent evidence recording and task response through a shared proactive generation process. Its Proactive Hierarchical Caption Memory (PHCM) produces time-grounded local-detail captions and summaries of completed events. Streaming caption targets supervise the interpretation of observed video prefixes during training. At inference, model-generated records complement a recent visual window, providing reusable factual context without revisiting historical visual features. Proactive State Transition Learning (PSTL) reduces the dominance of repeated waiting states by preserving supervision at all output anchors and selecting representative state-change and state-persistence tokens. We further develop a streaming data synthesis pipeline that aligns output content and timing with available evidence. Combining the resulting streaming captions and QA with cleaned open-source data yields OneStreamer-1M, a broad-coverage streaming video interaction dataset with over one million records spanning diverse tasks. Our 4B model achieves the best results among the compared methods across all eight evaluated streaming video understanding benchmarks. Ablations show that retaining generated captions improves historical QA without degrading real-time perception. PSTL also outperforms dense state supervision while supervising only 27.5% of annotated state tokens. Together, these results support proactive generation as a shared learning interface connecting perception, memory formation, and timely response in streaming video interaction.
Chinese Translation
流媒体视频的大型语言模型（LLMs）必须在未来任务的相关性尚不明确之前保留证据，并在足够证据可用时做出响应。挑战在于在不妨碍实时感知的情况下形成可重用的事实记忆。我们提出了OneStreamer，它通过共享的主动生成过程共同学习与查询无关的证据记录和任务响应。其主动层次标题记忆（Proactive Hierarchical Caption Memory, PHCM）生成时间基础的局部细节标题和已完成事件的摘要。流媒体标题目标在训练期间监督观察到的视频前缀的解释。在推理阶段，模型生成的记录补充了最近的视觉窗口，提供了可重用的事实上下文，而无需重新访问历史视觉特征。主动状态转移学习（Proactive State Transition Learning, PSTL）通过在所有输出锚点保留监督并选择代表性的状态变化和状态持续标记，减少了重复等待状态的主导地位。我们进一步开发了一种流媒体数据合成管道，使输出内容和时序与可用证据对齐。将生成的流媒体标题和问答与清理后的开源数据结合，产生了OneStreamer-1M，这是一个涵盖广泛流媒体视频交互的数据库，包含超过一百万条记录，涉及多种任务。我们的4B模型在所有八个评估的流媒体视频理解基准中取得了最佳结果。消融实验表明，保留生成的标题可以改善历史问答，而不会降低实时感知。PSTL在仅监督27.5%的标注状态标记的情况下，也优于密集状态监督。综合这些结果支持主动生成作为连接流媒体视频交互中感知、记忆形成和及时响应的共享学习接口。
cs.CV / 128 / 2610.01778
GIFTBench: Diagnosing Generalization in Image Forgery Localization and Informing Model Design
GIFTBench：诊断图像伪造定位中的泛化能力并指导模型设计
Abstract
Reliable evaluation of image forgery localization (IFL) requires assessing models under diverse distribution changes, yet existing benchmarks often cover limited manipulation conditions or entangle multiple factors in cross-dataset evaluation. Consequently, aggregate performance provides an incomplete view of localization generalization. We introduce GIFTBench, a multi-axis benchmark of 115,013 manipulated images with pixel-level annotations spanning manipulation source, semantic target, editing operation, and composition complexity. GIFTBench supports axis-specific transfer analysis and evaluation on twelve external datasets. Its diagnostic studies reveal asymmetric cross-source transfer, recall-dominated failures, and heterogeneous degradation across semantic, operational, and compositional changes. Beyond diagnosis, the scale and diversity of GIFTBench provide a substantially broader training distribution than conventional IFL datasets. Training representative localizers on GIFTBench consistently improves their aggregate transfer to external datasets, showing that the benchmark serves not only as an evaluation tool but also as an effective training resource for cross-domain localization. Guided by the diagnostic findings, we further develop ForenScope, a detection and localization framework combining classification-adapted representations with multi-depth, multi-scale spatial features, learned layer fusion, and selective coarse-scale conditioning. Experiments show improved cross-dataset localization while retaining image-level detection capability. The GIFTBench dataset showcase page is available at https://giftbench-preview.doudoudouya337.chatgpt.site.
Chinese Translation
可靠的图像伪造定位（IFL）评估需要在多样化的分布变化下评估模型，然而现有基准往往仅覆盖有限的操控条件或在跨数据集评估中纠缠多个因素。因此，整体性能提供了对定位泛化的不完整视图。我们引入了GIFTBench，这是一个包含115,013张经过操控的图像的多轴基准，具有涵盖操控来源、语义目标、编辑操作和构图复杂度的像素级注释。GIFTBench支持特定轴的迁移分析和在十二个外部数据集上的评估。其诊断研究揭示了不对称的跨来源迁移、以召回为主的失败以及在语义、操作和构图变化中的异质性退化。除了诊断，GIFTBench的规模和多样性提供了比传统IFL数据集更广泛的训练分布。在GIFTBench上训练代表性的定位器持续改善了它们对外部数据集的整体迁移，表明该基准不仅作为评估工具，还作为跨领域定位的有效训练资源。在诊断发现的指导下，我们进一步开发了ForenScope，这是一个结合了分类适应表示与多深度、多尺度空间特征、学习层融合和选择性粗尺度条件的检测与定位框架。实验表明，在保持图像级检测能力的同时，跨数据集定位得到了改善。GIFTBench数据集展示页面可访问：https://giftbench-preview.doudoudouya337.chatgpt.site。
cs.CV / 129 / 2610.01785
VETO: Video Efficient Token Optimization for Vision Language Models
VETO：面向视觉语言模型的视频高效令牌优化
Abstract
Processing long videos with Vision-Language Models (VLMs) is bottlenecked by the quadratic cost of visual tokens, making long-form inference prohibitively expensive. While single-axis compression methods mitigate this, they hit a hard efficiency floor because they treat spatial and temporal redundancy independently. We present VETO (Video Efficient Token Optimization for Vision-Language Models), a training-optional plug-in that eliminates this bottleneck through dual-axis compression: (i) an intra-frame compressor that merges semantically similar tokens within each frame via optimal-transport inspired matching, and (ii) an inter-frame compressor that identifies and merges temporally redundant frames. The key design insight is hierarchical ordering: by first compressing spatial dimensions, VETO drastically reduces the cost of subsequent global temporal matching, bypassing the efficiency wall of single-axis approaches, with an advantage that grows with modern fully-fused attention infrastructure. Empirically, VETO achieves up to 45% faster inference (e.g., on LLaVA-OneVision-7B) while preserving or improving accuracy. Under extreme token starvation (10% budget), VETO outperforms VFlowOpt (54.9%), VisionZip (52.6%), and FastV (47.9%) with 55.7% accuracy. We demonstrate universal applicability across LLaVA-OneVision, InternVL-2.5, and LongVA, with zero-shot accuracy preserved or improved in all cases.
Chinese Translation
使用视觉语言模型（VLMs）处理长视频时，由于视觉令牌的二次成本，导致长形式推理变得极为昂贵。虽然单轴压缩方法在一定程度上缓解了这一问题，但由于它们独立处理空间和时间冗余，因此面临着效率的硬底线。我们提出了VETO（Video Efficient Token Optimization for Vision-Language Models），这是一种可选的插件，通过双轴压缩消除了这一瓶颈：（i）一个帧内压缩器，通过受最优运输启发的匹配合并每帧内语义相似的令牌；（ii）一个帧间压缩器，识别并合并时间上冗余的帧。关键设计理念是层次排序：通过首先压缩空间维度，VETO大幅降低了后续全局时间匹配的成本，绕过了单轴方法的效率壁垒，并且这一优势随着现代全融合注意力基础设施的进步而增强。实证结果表明，VETO在保持或提高准确率的同时，实现了高达45%的推理速度提升（例如，在LLaVA-OneVision-7B上）。在极端令牌匮乏（10%预算）情况下，VETO以55.7%的准确率超越了VFlowOpt（54.9%）、VisionZip（52.6%）和FastV（47.9%）。我们展示了VETO在LLaVA-OneVision、InternVL-2.5和LongVA中的普遍适用性，在所有情况下均保持或提高了零样本准确率。
cs.CV / 130 / 2610.01794
Continuous Conditioning of VLAs with Augmenting EMG and Visual Task Descriptors
通过增强的肌电图和视觉任务描述符进行连续条件化的视觉-语言-动作模型
Abstract
Vision-Language-Action (VLA) models rely strongly on language for describing task information, despite having multimodal inputs. We hypothesize that other modalities in the state space may present opportunities for supplemental task conditioning, which may be particularly relevant in cluttered or otherwise ambiguous scenes. We introduce two tuned models to test this hypothesis: (1) an electrophysiology-conditioned VLA (EC-VLA) that incorporates 8-channel electromyography envelopes as continuous conditioning input concatenated to the proprioceptive vector, and (2) a visually-annotated VLA (VA-VLA) that incorporates visual segmentation annotations to the image inputs. On a cube-selection task evaluated across three participants, EC-VLA matches a language-prompted baseline in uncluttered, in-distribution conditions and substantially outperforms it in cluttered, out-of-distribution scenes. Similarly, VA-VLA shows modest improvements over a language-prompted baseline in in-distribution scenes with substantial improvement in cluttered, out-of-distribution trials. Together, these results provide strong evidence for the potential benefit of task-conditioning beyond language.
Chinese Translation
视觉-语言-动作（VLA）模型在描述任务信息时强烈依赖语言，尽管其具有多模态输入。我们假设状态空间中的其他模态可能为补充任务条件化提供机会，这在杂乱或模糊的场景中尤为相关。为验证这一假设，我们引入了两个调优模型：(1) 一种电生理条件化的VLA（EC-VLA），它将8通道肌电图包络作为连续条件输入，与本体感觉向量连接；(2) 一种视觉注释的VLA（VA-VLA），它将视觉分割注释整合到图像输入中。在对三名参与者进行的立方体选择任务中，EC-VLA在整洁的、分布内的条件下与语言提示的基线相匹配，并在杂乱的、分布外的场景中显著超越基线。同样，VA-VLA在分布内场景中相较于语言提示的基线显示出适度的改善，而在杂乱的、分布外的试验中则有显著提升。综合来看，这些结果为超越语言的任务条件化潜在益处提供了强有力的证据。
cs.CV / 131 / 2610.01807
PhaseAT: Fourier Phase Adversarial Training for Medical Image Domain Generalization
PhaseAT：用于医学图像领域泛化的傅里叶相位对抗训练
Abstract
Reliable clinical deployment of deep medical image models is hindered by distribution shifts across scanners, sites, and acquisition protocols. Existing domain generalization (DG) methods often focus on style or intensity diversification, but they can still leave networks dependent on domain-specific texture correlations. Inspired by evidence that Fourier phase encodes semantic structure, we introduce PhaseAT, a phase-aware adversarial training framework for medical DG. PhaseAT forms phase-perturbed training views in the Fourier domain by iteratively updating a bounded phase perturbation while keeping the amplitude spectrum unchanged, thereby stressing spatial organization under matched appearance statistics. Perturbations are applied only to the luminance channel in YCbCr color space to avoid chromatic artifacts. Additionally, a simple phase-saliency mask concentrates updates on the most influential frequencies. The model is trained with a weighted combination of losses on clean and phase-perturbed samples, supporting both single-source and multi-source DG. We validate our method on two challenging medical datasets and demonstrate that PhaseAT achieves over 20% improvement in single-source domain generalization, outperforming several state-of-the-art DG methods. The code implementation is available at: https://github.com/ahmed-sharshar/PhaseAT.
Chinese Translation
深度医学图像模型在临床部署中的可靠性受到扫描仪、地点和采集协议之间分布变化的阻碍。现有的领域泛化（DG）方法通常侧重于风格或强度的多样化，但仍可能使网络依赖于特定领域的纹理相关性。受到傅里叶相位编码语义结构的证据启发，我们提出了PhaseAT，一种面向相位的医学领域泛化对抗训练框架。PhaseAT通过迭代更新有界相位扰动，同时保持幅度谱不变，在傅里叶域中形成相位扰动的训练视图，从而强调在匹配外观统计下的空间组织。扰动仅应用于YCbCr颜色空间中的亮度通道，以避免色彩伪影。此外，一个简单的相位显著性掩码将更新集中在最具影响力的频率上。模型通过对干净样本和相位扰动样本的损失加权组合进行训练，支持单源和多源领域泛化。我们在两个具有挑战性的医学数据集上验证了我们的方法，并证明PhaseAT在单源领域泛化中实现了超过20%的改进，超越了几种最先进的领域泛化方法。代码实现可在以下链接获取：https://github.com/ahmed-sharshar/PhaseAT。
cs.CV / 132 / 2610.01863
LiteReality-Agent: An Agentic System for Interactable 3D Indoor Scene Reconstruction
LiteReality-Agent：一种可交互的室内场景重建智能系统
Abstract
We present LiteReality-Agent, an agentic system for reconstructing real indoor environments as realistic, articulated, and simulation-ready 3D scenes from RGB-D scans. At its core, LiteReality-Agent formulates 3D reconstruction as a coding problem, in which a coding agent gathers evidence using specialised tools and iteratively edits a Python script, Room.py, which can be executed to produce a 3D digital twin of the room. With this formulation, we develop a robust observe-edit-verify harness that supports evidence gathering, measurement, verification, layout optimisation, simulation readiness, and quality control throughout the reconstruction process. LiteReality-Agent produces high-quality reconstructions suitable for simulation and downstream embodied AI tasks. Furthermore, as agent capabilities continue to improve rapidly, the system introduced by LiteReality-Agent remains a strong orchestration framework for future agents: it equips them with specialised tools, structured workflows, and robust verification mechanisms that substantially improve reconstruction quality and reliability. We demonstrate that LiteReality-Agent produces reconstructions that are more geometrically accurate, visually realistic, and simulation-compatible than those generated by recent frontier models, such as Astra and Fable. We therefore view LiteReality-Agent as a practical and important building block for robust real-to-sim systems. Both the source code and the data-capture application are publicly available. Code:https://github.com/LiteReality/LiteReality-Agent/
Chinese Translation
我们提出了LiteReality-Agent，这是一种智能系统，能够将真实的室内环境重建为逼真、可操作且适合模拟的3D场景，基于RGB-D扫描数据。LiteReality-Agent的核心是将3D重建视为一个编码问题，其中编码代理使用专门工具收集证据，并迭代编辑一个名为Room.py的Python脚本，该脚本可以执行以生成房间的3D数字双胞胎。通过这种构思，我们开发了一个强大的观察-编辑-验证框架，支持在重建过程中进行证据收集、测量、验证、布局优化、模拟准备和质量控制。LiteReality-Agent生成的高质量重建适合于模拟和后续的具身人工智能任务。此外，随着代理能力的快速提升，LiteReality-Agent所引入的系统仍然是未来代理的强大协调框架：它为代理提供了专门工具、结构化工作流程和强大的验证机制，显著提高了重建质量和可靠性。我们展示了LiteReality-Agent生成的重建在几何准确性、视觉真实感和模拟兼容性方面优于最近的前沿模型，如Astra和Fable。因此，我们将LiteReality-Agent视为构建稳健的真实到模拟系统的重要实用基础。源代码和数据捕获应用程序均已公开发布。代码链接：https://github.com/LiteReality/LiteReality-Agent/
cs.CV / 133 / 2610.01870
From Pixels to Policy: A Multi-Agent System for Intervention and Geo-Spatial Decision Support
从像素到政策：用于干预和地理空间决策支持的多智能体系统
Abstract
Urban environments are shaped by design choices with long-term implications for health, safety, and quality of life, yet evaluating proposed interventions remains costly, time-consuming, and often impractical. Existing geospatial vision methods largely focus on monitoring urban indicators from aerial and street-view imagery, rather than proposing interventions and estimating their effects on such indicators. Moving beyond recognition, we introduce the problem of discovering interventions that improve target indicators for a given aerial or street-view image. We argue that a black-box indicator model, combined with a generative editing model, can serve as an implicit digital twin for testing intervention hypotheses. We present VIDA-Geo , a multi-agent system that explores this intervention space by coordinating segmentation, diffusion-based inpainting, and indicator scoring models to produce interventions that are both perceptually realistic and aligned with real-world policies. We evaluate our system on 8 indicators across aerial and street-view imagery, measuring changes in factors such as perceived safety and greenery. Our approach outperforms existing baselines in many cases, achieving up to 2X higher perceptual quality and policy alignment scores. Finally, our model provides users with multiple candidate interventions, supporting an expert city-planner-in-the-loop workflow.
Chinese Translation
城市环境受到设计选择的影响，这些选择对健康、安全和生活质量具有长期影响。然而，评估提议的干预措施仍然成本高昂、耗时且往往不切实际。现有的地理空间视觉方法主要集中在从空中和街景图像中监测城市指标，而不是提出干预措施并估计其对这些指标的影响。超越识别，我们引入了发现干预措施以改善给定空中或街景图像的目标指标的问题。我们认为，结合黑箱指标模型和生成编辑模型，可以作为测试干预假设的隐式数字双胞胎。我们提出了VIDA-Geo，一个多智能体系统，通过协调分割、基于扩散的修复和指标评分模型来探索这一干预空间，以产生在感知上真实且与现实政策一致的干预措施。我们在空中和街景图像上对8个指标进行了评估，测量了感知安全性和绿化等因素的变化。在许多情况下，我们的方法超越了现有基线，达到了最高2倍的感知质量和政策一致性评分。最后，我们的模型为用户提供了多个候选干预措施，支持专家城市规划师的参与工作流程。
cs.CV / 134 / 2610.01876
EvenSplat: Coupled 2D-3D Decomposition for Gaussian Splatting under Exposure and Illumination Variation
EvenSplat：在曝光和照明变化下的耦合2D-3D分解用于高斯溅射
Abstract
A surface photographed under even light presents nearly the same appearance from every angle; the same surface under uneven light does not. Exposure changes between views, illumination varies within a single image, and locally strong light sources leave one region bright and its neighbor in shadow. Multi-view reconstruction methods such as 3D Gaussian Splatting treat these lighting artifacts as if they were properties of the scene, entangling capture-specific illumination with the geometry and color they recover. We present EvenSplat, a framework that separates the two. EvenSplat couples an image-space illumination decomposition with an illumination field carried by the Gaussians, so that the same explanation of the lighting is shared between the two-dimensional and three-dimensional views of the scene; a camera-response network and a local exposure-compensation module absorb the global and residual differences that remain across training images. Through extensive experiments across multiple datasets and diverse forms of uneven illumination (cross-view exposure, spatial illumination variation, and high-contrast lighting) on both real-world captured and simulated benchmarks, EvenSplat generally outperforms state-of-the-art methods, particularly under high-contrast illumination.
Chinese Translation
在均匀光照下拍摄的表面从每个角度呈现出几乎相同的外观；而在不均匀光照下则不然。视角之间的曝光变化、单幅图像内的照明变化，以及局部强光源使得某一区域明亮而其邻近区域则处于阴影中。多视角重建方法如3D高斯溅射将这些光照伪影视为场景的属性，将捕获特定的光照与其恢复的几何形状和颜色纠缠在一起。我们提出了EvenSplat，一个将两者分离的框架。EvenSplat将图像空间的光照分解与由高斯分布携带的光照场耦合，使得场景的二维和三维视图共享相同的光照解释；一个相机响应网络和一个局部曝光补偿模块吸收训练图像中残留的全局和局部差异。通过在多个数据集和多种不均匀光照形式（跨视角曝光、空间照明变化和高对比度光照）下进行广泛实验，EvenSplat通常优于最先进的方法，特别是在高对比度光照条件下。
cs.CV / 135 / 2610.01884
Memory-Guided B-Roll Generation from User Video Collections
基于记忆的用户视频集合B-roll生成
Abstract
We introduce an approach for collection-grounded B-roll sequence generation. Given a user's video collection, a directive given in natural language, and a target duration, the goal is to produce a multi-shot sequence that complements the user's primary footage (A-roll) while preserving the collection's characters, settings, objects, and style. This task is challenging as one must choose the visual evidence from hours of captured footage that should guide the generation of each shot in the sequence. We address this challenge with MemComposer, a three-stage system that turns raw footage into a structured memory with visual references (characters, settings, objects, and style) and uses it to plan, retrieve, and generate grounded B-roll sequences. First, in a one-time offline stage, MemComposer constructs an entity-centric memory from raw video. Second, it uses the memory and user directive to plan a grounded sequence and retrieve conditioning frames for each shot. Third, it iteratively generates and critiques the sequence to enforce identity, setting, and sequence-level consistency. We evaluate MemComposer in a user preference study along two dimensions: prompt adherence and visual alignment to the user's collection. Against an ungrounded text-to-video planner, MemComposer wins 60.0\% of prompt-adherence and 92.8\% of visual-alignment comparisons, showing the grounding benefit of collection memory and reference retrieval. Against retrieval-only sequences assembled from captured footage, MemComposer wins 94.5\% of prompt-adherence comparisons, showing the value of generating missing shots, while retrieval-only sequences are preferred for visual alignment in 58.2\% of comparisons.
Chinese Translation
我们提出了一种基于集合的B-roll序列生成方法。给定用户的视频集合、自然语言指令和目标时长，目标是生成一个多镜头序列，以补充用户的主要镜头（A-roll），同时保留集合中的角色、场景、物体和风格。这个任务具有挑战性，因为必须从数小时的捕获视频中选择视觉证据，以指导序列中每个镜头的生成。我们通过MemComposer来应对这一挑战，这是一种三阶段系统，将原始视频转化为具有视觉参考（角色、场景、物体和风格）的结构化记忆，并利用该记忆进行规划、检索和生成基于集合的B-roll序列。首先，在一次性的离线阶段，MemComposer从原始视频构建以实体为中心的记忆。其次，它利用记忆和用户指令规划一个基于集合的序列，并为每个镜头检索条件帧。第三，它迭代生成和评估序列，以确保身份、场景和序列级别的一致性。我们在用户偏好研究中评估了MemComposer，主要从两个维度进行比较：对指令的遵循程度和与用户集合的视觉一致性。与未基于集合的文本到视频规划器相比，MemComposer在指令遵循方面赢得了60.0%的比较，在视觉一致性方面赢得了92.8%的比较，显示了集合记忆和参考检索的优势。与仅从捕获视频组装的检索序列相比，MemComposer在指令遵循方面赢得了94.5%的比较，显示了生成缺失镜头的价值，而在58.2%的比较中，仅检索的序列在视觉一致性方面更受偏好。
cs.CV / 136 / 2610.01890
Unsupervised Domain Adaptation for Enhanced Radiometer Image Precipitation Estimation using Conditional Flow Matching
基于条件流匹配的无监督领域适应用于增强辐射计图像降水估计
Abstract
Deep generative networks have recently achieved unprecedented performance in precise image and video editing using sophisticated textual prompts. However, the effectiveness of such models heavily depends on access to very large supervised and annotated image datasets, which can be very difficult to obtain. This is particularly true for satellite instruments, which very rarely overlap with labelled data, and suffer from domain shifts in the rare occasions they do. In this paper, we investigate the potential of flow matching models for unsupervised domain adaptation of satellite radiometer images. Our main contribution is a novel unsupervised method that achieves precise domain alignment by leveraging parts of the deterministic ordinary differential equations in flow matching models, conditioned on different satellite instruments. A key strength of our approach is its ability to preserve essential information while adapting across any domains since the perturbations are in theory bijective. Extensive experiments conducted on the GPM-Core constellation show the benefit of our conditional domain adaptation, particularly in improving rain precipitation estimation from radiometer imagery.
Chinese Translation
深度生成网络最近在使用复杂文本提示进行精确图像和视频编辑方面取得了前所未有的性能。然而，这类模型的有效性在很大程度上依赖于对非常大规模的监督和标注图像数据集的访问，而这些数据集往往难以获取。对于卫星仪器而言，这一点尤其明显，因为它们与标记数据的重叠非常少，并且在偶尔重叠的情况下也会遭遇领域转移。在本文中，我们研究了流匹配模型在卫星辐射计图像的无监督领域适应中的潜力。我们的主要贡献是一种新颖的无监督方法，通过利用流匹配模型中确定性常微分方程的部分，基于不同的卫星仪器实现精确的领域对齐。我们方法的一个关键优势在于其在跨领域适应时能够保留重要信息，因为理论上扰动是双射的。在GPM-Core星座上进行的大量实验表明，我们的条件领域适应的优势，特别是在改善辐射计图像的降水估计方面。
cs.CV / 137 / 2610.01905
MapLightning: Online Vectorized HD Map Construction with 1D Map Tokens
MapLightning：基于1D地图标记的在线矢量化高清地图构建
Abstract
Online vectorized HD map construction is essential for scaling safe autonomous driving and requires accurate, real-time inference. Prior methods typically rely on dense bird's-eye-view (BEV) grids as the intermediate representation. We propose \textit{MapLightning}, which replaces the dense BEV grid with a compact set of 1D learnable map tokens. To construct map tokens from image features, we choose self-attention over vanilla cross-attention because it enables joint interactions and contextual aggregation among image and map tokens. Our transformer-based mapper concatenates map and image tokens, applies full self-attention, discards the image tokens, and retains the updated map tokens for decoding. This design offers three advantages. First, our representation is efficient, using fewer tokens, consuming less memory, and running faster. Second, the lightweight design allows the map decoder to use full rather than deformable cross-attention for better global context. Third, unlike BEV-based methods, our network does not use camera projection parameters, making it robust to camera-extrinsic perturbations. MapLightning uses up to 16.7$\times$ fewer intermediate tokens than dense BEV-based methods and achieves state-of-the-art accuracy and efficiency on nuScenes and Argoverse~2. Its lightweight variant surpasses MapTRv2 by +10.1 mAP on nuScenes and +16.2 mAP on Argoverse~2, while delivering 1.73$\times$ faster inference (40+ FPS) with 53\% less memory. We further show improvements on uncertainty-aware map construction and downstream trajectory prediction. Code and models will be released.
Chinese Translation
在线矢量化高清地图构建对于扩展安全的自动驾驶至关重要，并且需要准确的实时推理。以往的方法通常依赖于密集的鸟瞰图（BEV）网格作为中间表示。我们提出了 extit{MapLightning}，它用一组紧凑的1D可学习地图标记替代了密集的BEV网格。为了从图像特征构建地图标记，我们选择自注意力机制而非普通的交叉注意力机制，因为它能够实现图像标记和地图标记之间的联合交互和上下文聚合。我们的基于变换器的映射器将地图和图像标记进行连接，应用全自注意力，丢弃图像标记，并保留更新后的地图标记进行解码。这一设计提供了三个优势。首先，我们的表示高效，使用更少的标记，消耗更少的内存，并且运行更快。其次，轻量化设计使得地图解码器能够使用全交叉注意力而非可变形交叉注意力，从而获得更好的全局上下文。第三，与基于BEV的方法不同，我们的网络不使用相机投影参数，使其对相机外部扰动具有更强的鲁棒性。MapLightning使用的中间标记数量比密集BEV方法少多达16.7倍，并在nuScenes和Argoverse~2上实现了最先进的准确性和效率。其轻量化变体在nuScenes上比MapTRv2提高了+10.1 mAP，在Argoverse~2上提高了+16.2 mAP，同时提供了1.73倍更快的推理速度（40+ FPS）和53%的内存减少。我们进一步展示了在不确定性感知地图构建和下游轨迹预测方面的改进。代码和模型将会发布。
cs.CV / 138 / 2610.01914
DecomVoxel: Harnessing 3D-Native Priors with Guided In-situ Denoising Optimization for Decompositional Scene Reconstruction
DecomVoxel：利用引导式原位去噪优化结合3D原生先验进行分解场景重建
Abstract
Decompositional scene reconstruction aims to reconstruct high-quality objects and background, yet existing methods still struggle with the level of quality under heavy occlusions. While generative priors offer a potential solution, 2D image-based priors often suffer from multi-view inconsistency due to a lack of 3D awareness. Conversely, 3D-native priors provide stronger structural inductive biases but frequently lead to spatial drift and misalignment within complex scenes. To address these issues, we propose DecomVoxel, formulating object completion as a guided in-situ denoising optimization that bridges 3D-native priors with neural scene reconstruction. Our framework introduces a reformulated epsilon-based distillation loss to ensure stable latent refinement, alongside adaptive spatial guidance that utilizes occupied and vacant anchors with temporal annealing to suppress generative hallucinations and mitigate spatial drift. Experiments on Replica and ScanNet++ show that DecomVoxel significantly outperforms state-of-the-art methods while faithfully preserving the original spatial layout, structural fidelity, and style-consistent texture. Our method pushes the boundary of decompositional reconstruction by delivering high-quality textured meshes with clean topology, geometry, and appearance, providing a robust solution for the decompositional reconstruction of complex real-world scenes. Code is available at https://github.com/DecomVoxel/DecomVoxel.
Chinese Translation
分解场景重建旨在重建高质量的物体和背景，但现有方法在重度遮挡情况下仍面临质量挑战。尽管生成先验提供了潜在的解决方案，但基于2D图像的先验常常由于缺乏3D意识而遭受多视角不一致的问题。相反，3D原生先验提供了更强的结构归纳偏差，但在复杂场景中常常导致空间漂移和错位。为了解决这些问题，我们提出了DecomVoxel，将物体补全形式化为一种引导式原位去噪优化，连接3D原生先验与神经场景重建。我们的框架引入了一种重新公式化的基于ε的蒸馏损失，以确保稳定的潜在细化，同时采用适应性空间引导，利用占用和空闲锚点结合时间退火来抑制生成幻觉并减轻空间漂移。在Replica和ScanNet++上的实验表明，DecomVoxel显著优于最先进的方法，同时忠实保留原始空间布局、结构保真度和风格一致的纹理。我们的方法通过提供高质量的纹理网格，具备干净的拓扑、几何和外观，推动了分解重建的边界，为复杂现实场景的分解重建提供了稳健的解决方案。代码可在 https://github.com/DecomVoxel/DecomVoxel 获取。
cs.CV / 139 / 2610.01917
MoLE: Mixture of Latent Experts for Complementary Visual Reasoning
MoLE：用于互补视觉推理的潜在专家混合模型
Abstract
Latent visual reasoning equips vision--language models with continuous intermediate states that can process visual evidence without explicit textual reasoning traces or repeated image operations. However, existing methods often allow multiple latent tokens to access the same visual evidence through shared value projections, providing no mechanism for them to extract complementary visual information; simply increasing the latent budget can therefore yield redundant latent representations. We argue that effective latent reasoning should encourage different latent tokens to extract complementary visual information, and thereby act as specialized visual experts. Based on this insight, we propose MoLE, a Mixture of Latent Experts framework that controls both what visual evidence each latent visual expert observes and how it transforms that evidence. MoLE isolates latent visual experts during evidence extraction and uses dedicated latent summary experts to aggregate the complementary representations of latent visual experts. A two-stage training pipeline first forces visual evidence through this latent pathway and then restores direct visual access, requiring neither predefined expert roles nor intermediate visual targets. Across five visual reasoning benchmarks, MoLE achieves an average score of 78.6, outperforming data-matched supervised fine-tuning by 4.9 and the strongest evaluated latent visual reasoning baseline at the same latent budget by 3.6. Representation analyses show lower latent-state similarity and more diverse visual attention, while masking the latent pathway reduces average performance by 9.2. These results demonstrate that specializing latent computation is more effective than merely increasing the number of latent tokens.
Chinese Translation
潜在视觉推理为视觉-语言模型提供了连续的中间状态，这些状态能够处理视觉证据，而无需显式的文本推理痕迹或重复的图像操作。然而，现有方法通常允许多个潜在标记通过共享的值投影访问相同的视觉证据，缺乏提取互补视觉信息的机制；因此，仅仅增加潜在预算可能会导致冗余的潜在表示。我们认为，有效的潜在推理应鼓励不同的潜在标记提取互补的视觉信息，从而充当专业的视觉专家。基于这一洞察，我们提出了MoLE，一个潜在专家混合框架，控制每个潜在视觉专家观察的视觉证据以及它如何转化该证据。MoLE在证据提取过程中隔离潜在视觉专家，并使用专门的潜在摘要专家来聚合潜在视觉专家的互补表示。一个两阶段的训练流程首先通过这一潜在路径强制视觉证据，然后恢复直接的视觉访问，既不需要预定义的专家角色，也不需要中间视觉目标。在五个视觉推理基准测试中，MoLE的平均得分为78.6，比数据匹配的监督微调高出4.9，比在相同潜在预算下评估的最强潜在视觉推理基线高出3.6。表示分析显示潜在状态相似性较低和更具多样性的视觉注意力，而屏蔽潜在路径则使平均性能降低了9.2。这些结果表明，专门化潜在计算比单纯增加潜在标记的数量更为有效。
cs.CV / 140 / 2610.01927
CLoSeR: Closing the Loop for Long-Context Streaming Reconstruction
CLoSeR：闭环长上下文流重建
Abstract
Feedforward foundation models have recently shown remarkable 3D reconstruction capabilities. However, existing models exhibit large tracking drift in long-context streaming reconstruction due to error accumulation. In this paper, we revisit loop closure with streaming reconstruction foundation models to enable accurate, drift-free, kilometer-scale reconstruction. Specifically, our method detects loop candidates through global descriptor retrieval, and constructs loop-conditioned windows to estimate the relative poses between looped frames. Given the observation that our adopted streaming reconstruction backbone produces a globally consistent scale, we optimize all frame poses on the SE(3) manifold with sequential and loop closure constraints, avoiding the pose graph optimization on the Sim(3) or higher-dimensional SL(4) manifolds employed in prior works. Extensive experiments show that our method reduces drift and produces consistent geometry on kilometer-scale sequences, significantly outperforming the state of the art. Code is available at https://github.com/MoyangLi00/CLoSeR.git.
Chinese Translation
前馈基础模型最近在三维重建方面展现了显著的能力。然而，现有模型在长上下文流重建中由于误差积累表现出较大的跟踪漂移。本文重新审视了流重建基础模型中的闭环问题，以实现准确、无漂移的公里级重建。具体而言，我们的方法通过全局描述符检索检测闭环候选，并构建闭环条件窗口以估计闭环帧之间的相对位姿。鉴于我们采用的流重建骨干网络产生全局一致的尺度，我们在SE(3)流形上优化所有帧的位姿，结合顺序和闭环约束，避免了在Sim(3)或更高维的SL(4)流形上进行位姿图优化，这在之前的工作中被采用。大量实验表明，我们的方法减少了漂移，并在公里级序列上产生了一致的几何形状，显著超越了现有的最先进技术。代码可在 https://github.com/MoyangLi00/CLoSeR.git 获取。
cs.CV / 141 / 2610.01939
Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens
更少的令牌，更好的行动：GPT-6 Astra机器人代理成功率提高14%，但令牌减少65%
Abstract
Vision language model (VLM) agents can control robots through visual feedback and action primitives, but repeated model invocations and redundant observations incur substantial token overhead. We introduce PyRUA-Lean, an interactive code-execution framework that couples feedback-driven primitive composition with selective observation: the agent composes classical robot primitives and learned vision-language-action (VLA) policies into Python cells that perform conditional checks and local retries, returning only explicitly requested images and state feedback for replanning. Across 700 simulated task instances from LIBERO-PRO, RoboTwin 2.0, and RoboCasa365, we compare PyRUA-Lean with a tool-calling baseline using the same GPT-6 Astra planner and underlying robot primitives. Under equal LLM-call budgets, PyRUA-Lean increases overall success from 63.1% to 71.7%. On instances solved by both agents, it uses 49% fewer LLM calls and 65% fewer input tokens.
Chinese Translation
视觉语言模型（VLM）代理可以通过视觉反馈和动作原语控制机器人，但重复的模型调用和冗余的观察会产生大量的令牌开销。我们提出了PyRUA-Lean，一个交互式代码执行框架，它将基于反馈的原语组合与选择性观察相结合：代理将经典机器人原语和学习的视觉-语言-动作（VLA）策略组合成Python单元，这些单元执行条件检查和本地重试，仅返回明确请求的图像和状态反馈以进行重新规划。在来自LIBERO-PRO、RoboTwin 2.0和RoboCasa365的700个模拟任务实例中，我们将PyRUA-Lean与使用相同GPT-6 Astra规划器和基础机器人原语的工具调用基线进行了比较。在相同的LLM调用预算下，PyRUA-Lean将整体成功率从63.1%提高到71.7%。在两个代理都解决的实例中，它使用的LLM调用减少了49%，输入令牌减少了65%。
cs.CV / 142 / 2610.01942
Latent-Foresight: End-to-End Learning Predictable Representations for Latent World Models
潜在前瞻：端到端学习可预测表示的潜在世界模型
Abstract
Predicting the future evolution of a scene is a fundamental capability for world modeling. Recent work has shown that operating in the feature space of Vision Foundation Models (VFMs) yields semantically rich representations that support diverse future scene understanding tasks. However, existing approaches rely on two-stage pipelines, where VFM features are first compressed using fixed dimensionality reduction (e.g., PCA) or independently trained autoencoders, and a separate predictor is trained on top of the resulting frozen latent space. This decoupling between representation learning and temporal prediction, as well as approaches that apply predictors directly on raw VFM features, provides no guarantee that the latent space is structured for predictable dynamics. In this work, we propose Latent-Foresight, an end-to-end framework that jointly learns a latent tokenizer and a flow-based generative dynamics model, explicitly shaping the representation to support temporal predictability. To enable stable joint optimization, we introduce several key design choices that prevent latent collapse and align reconstruction with generative objectives. Extensive experiments show that our approach learns more temporally coherent latent representations and consistently outperforms two-stage baselines across multiple future scene understanding tasks and prediction horizons, while eliminating separate training stages, including during high-resolution adaptation. We provide the implementation code and model weights at https://github.com/Sta8is/Latent-Foresight
Chinese Translation
预测场景的未来演变是世界建模的一项基本能力。近期的研究表明，在视觉基础模型（Vision Foundation Models, VFMs）的特征空间中操作可以产生语义丰富的表示，支持多样的未来场景理解任务。然而，现有方法依赖于两阶段管道，其中VFM特征首先通过固定维度的降维方法（例如主成分分析，PCA）或独立训练的自编码器进行压缩，然后在得到的冻结潜在空间上训练一个单独的预测器。这种表示学习与时间预测之间的解耦，以及直接在原始VFM特征上应用预测器的方法，并不能保证潜在空间是为可预测动态而结构化的。在本研究中，我们提出了潜在前瞻（Latent-Foresight），一个端到端框架，联合学习潜在分词器和基于流的生成动态模型，明确塑造表示以支持时间可预测性。为了实现稳定的联合优化，我们引入了几个关键设计选择，以防止潜在崩溃并将重建与生成目标对齐。大量实验表明，我们的方法学习到的潜在表示在时间上更加一致，并在多个未来场景理解任务和预测时间范围内持续优于两阶段基线，同时消除了独立训练阶段，包括在高分辨率适应期间。我们在 https://github.com/Sta8is/Latent-Foresight 提供了实现代码和模型权重。
cs.CV / 143 / 2610.01944
Anti-Persona: Disrupting Unauthorized Identity Binding and Recognition in Personalized Vision--Language Models
反身份：在个性化视觉-语言模型中破坏未经授权的身份绑定和识别
Abstract
Few-shot personalization enables large vision--language models (LVLMs) to learn user-specific visual concepts for applications such as personalized retrieval and subject-aware querying. However, it also creates a privacy risk: an adversary can bind a target identity from a few reference images and subsequently detect that identity in new images through natural-language queries. We introduce Anti-Persona, an image-level defense against unauthorized identity binding and recognition in personalized LVLMs. Our key insight is that identity personalization relies on visual features shared across multiple reference images. We aggregate these features into an identity prototype and optimize visually subtle perturbations that disrupt prototype alignment in the vision-encoder space. Spatial smoothing and low-frequency preservation further promote visual fidelity and practical resilience to image compression. The resulting protection does not depend on a specific prompt and supports both proactive anti-personalization and reactive image protection. Experiments on two representative personalized LVLMs demonstrate protection rates of up to $95.0\%$ while preserving visual fidelity. The method remains stable across prompt variations and evaluated identity-query tasks, and improves black-box transfer under encoder mismatch.
Chinese Translation
少样本个性化使大型视觉-语言模型（LVLMs）能够学习用户特定的视觉概念，应用于个性化检索和主题感知查询等场景。然而，这也带来了隐私风险：攻击者可以从少量参考图像中绑定目标身份，并通过自然语言查询在新图像中检测该身份。我们提出了反身份（Anti-Persona），这是一种针对个性化LVLMs中未经授权的身份绑定和识别的图像级防御。我们的关键见解是，身份个性化依赖于多个参考图像中共享的视觉特征。我们将这些特征聚合成一个身份原型，并优化视觉上微妙的扰动，以破坏原型在视觉编码器空间中的对齐。空间平滑和低频保留进一步促进了视觉保真度和对图像压缩的实际抵抗力。所得到的保护不依赖于特定的提示，并支持主动反个性化和被动图像保护。在两个代表性的个性化LVLMs上的实验表明，保护率高达95.0%，同时保持视觉保真度。该方法在提示变化和评估的身份查询任务中保持稳定，并在编码器不匹配的情况下改善黑箱转移。
cs.CV / 144 / 2610.01956
EndoLive: Real-Time Style Transfer for Endoscopic Endonasal Skull Base Surgical Video
EndoLive：内窥镜内鼻颅底手术视频的实时风格迁移
Abstract
Complex surgical procedures around critical anatomy, such as the endoscopic endonasal skull base surgery, requires significant practice and training on the part of the surgeon before they are allowed to perform the operation on a live patient. This training in typically done in cadaveric specimens, due to them containing the same critical structures as a living human. However, cadavers are not a perfect 1-to-1 substitute for a living patient. The dead and preserved tissues of a cadaver are colored completely differently than a living human, and -- without complex and expensive pumping systems -- do not bleed in the same way. As a result, identifying the critical pieces of anatomy that make this procedure so complex can be quite different in a live case than in a surgeon's cadaveric practice. This paper presents EndoLive, a framework for real-time style transfer between cadaveric endoscopic video and living human endoscopic video. Our method combines the ConStructS GAN model for realistic style transfer for surgical applications, with the HyPER-GAN model that can learn complex translations and perform them in real-time. We train EndoLive on unpaired cadaveric and live images taken from an endoscope, and test the trained model with cadaveric video, on a variety of devices. Experimental results demonstrate that EndoLive can perform cadaveric-to-live translation at speeds well above the minimum necessary for real-time, while maintaining semantic consistency of critical anatomical structures. Our source code is available at https://github.com/griffhurt/endolive.
Chinese Translation
在关键解剖结构周围进行复杂的外科手术，例如内窥镜内鼻颅底手术，需要外科医生在被允许对活体患者进行手术之前进行大量的实践和训练。这种训练通常是在尸体标本上进行的，因为它们包含与活人相同的关键结构。然而，尸体并不是活体患者的完美1对1替代品。尸体的死去和保存的组织与活体人类的颜色完全不同，并且在没有复杂且昂贵的泵送系统的情况下，出血方式也不相同。因此，在活体案例中识别使该手术如此复杂的关键解剖结构可能与外科医生在尸体实践中的情况截然不同。本文提出了EndoLive，一个用于尸体内窥镜视频与活体人类内窥镜视频之间实时风格迁移的框架。我们的方法结合了ConStructS GAN模型，用于外科应用的真实风格迁移，以及HyPER-GAN模型，能够学习复杂的转换并实时执行它们。我们在未配对的尸体图像和活体图像上训练EndoLive，并在多种设备上测试训练好的模型。实验结果表明，EndoLive能够以远高于实时所需的最低速度执行尸体到活体的转换，同时保持关键解剖结构的语义一致性。我们的源代码可在 https://github.com/griffhurt/endolive 获取。
cs.CV / 145 / 2610.01969
RASteer: Retain-Aware Activation Steering for Concept Erasure in Diffusion Models
RASteer：面向概念抹除的保留意识激活引导在扩散模型中的应用
Abstract
Concept erasure aims to remove a target concept, such as a copyrighted style, a recognizable character, or unsafe content, from a pretrained text-to-image diffusion model while preserving its ability to generate other content. Existing activation steering methods build an erasure direction mainly from the target concept and adjust model activations along it at inference time. However, target and retained concepts often overlap in the model's representation space, so this direction also contains shared components that retained concepts rely on. Steering directly along this direction can therefore suppress retained concepts and harm the generation of non-target content. To address this issue, we propose Retain-aware Activation Steering (RASteer), a training-free method. RASteer first builds a retain subspace from the concepts to preserve. Retain-Orthogonal Steering (ROS) then removes components aligned with this subspace from the erasure direction, making steering more specific to the target. Since fully removing the shared components can weaken erasure, we further introduce Overlap-Adaptive Calibration (OAC). At each layer and denoising step, OAC uses the overlap between the erasure direction and the retain subspace to control how much of each shared component is removed, balancing target erasure and concept preservation. Experiments on unsafe-content, instance, and artistic-style erasure across multiple backbones and benchmarks show that RASteer matches or outperforms the activation steering and weight editing baselines we evaluate, achieving a better balance between erasure and preservation.
Chinese Translation
概念抹除旨在从预训练的文本到图像扩散模型中去除目标概念，例如版权风格、可识别角色或不安全内容，同时保留其生成其他内容的能力。现有的激活引导方法主要从目标概念构建抹除方向，并在推理时沿该方向调整模型激活。然而，目标概念与保留概念在模型的表示空间中往往存在重叠，因此该方向也包含保留概念所依赖的共享成分。因此，沿此方向直接引导可能会抑制保留概念，并损害非目标内容的生成。为了解决这个问题，我们提出了一种无训练的方法——保留意识激活引导（RASteer）。RASteer首先从需要保留的概念中构建保留子空间。然后，保留正交引导（Retain-Orthogonal Steering, ROS）从抹除方向中去除与该子空间对齐的成分，使引导更具体于目标。由于完全去除共享成分可能会削弱抹除效果，我们进一步引入了重叠自适应校准（Overlap-Adaptive Calibration, OAC）。在每一层和去噪步骤中，OAC利用抹除方向与保留子空间之间的重叠来控制每个共享成分的去除程度，从而平衡目标抹除与概念保留。在多个基础模型和基准测试上的不安全内容、实例和艺术风格抹除实验表明，RASteer在我们评估的激活引导和权重编辑基线中表现相当或更优，实现了抹除与保留之间更好的平衡。
cs.CV / 146 / 2610.01973
Token-Level Video Reinforcement Learning
基于标记的视频强化学习
Abstract
Reinforcement learning (RL) for video generation usually assigns one scalar reward to an entire sampled video. Yet a video is not uniformly flawed: some visual tokens may already satisfy the prompt, whereas others require correction. A scalar reward cannot localize errors, causing optimization to perturb satisfactory tokens while under-targeting the tokens that actually need to change. We introduce Token-Level Video Reinforcement Learning, TVRL, a framework that derives token-level credit from the reward being optimized. Our key insight is that the answer likelihood of a frozen vision-language model provides both signals: its outputs contribute to the video-level reward, while magnitudes of its video-input gradients reveal which generated video tokens most affect that score. We instantiate TVRL in Group Relative Policy Optimization by averaging prompt-derived question rewards into one group-relative advantage and using detached, question-conditioned token-credit maps to reweight dense denoising-transition log-probabilities inside the clipped policy ratio. On VBench-2.0, TVRL achieves an Overall score of 57.69, outperforming the base model by 3.60 points. TVRL also improves matched GRPO baselines across three SDE samplers (SAGE, Flow, and Dance) by 2.68--3.15 points and across four reward models (VideoAlign, VideoScore2, UnifiedReward2, and Qwen3.5-9B) by 1.33--3.15 points.
Chinese Translation
视频生成的强化学习（RL）通常为整个采样视频分配一个标量奖励。然而，视频并不是均匀存在缺陷的：某些视觉标记可能已经满足提示，而其他标记则需要修正。标量奖励无法定位错误，导致优化扰动了满意的标记，同时未能针对实际需要更改的标记。我们提出了基于标记的视频强化学习（Token-Level Video Reinforcement Learning，TVRL），这是一个从优化的奖励中推导标记级别信用的框架。我们的关键见解是，冻结的视觉-语言模型的答案可能性提供了两种信号：其输出对视频级别奖励的贡献，同时其视频输入梯度的大小揭示了哪些生成的视频标记对该评分影响最大。我们在组相对策略优化（Group Relative Policy Optimization）中实例化了TVRL，通过将提示导出的问答奖励平均为一个组相对优势，并使用独立的、基于问题的标记信用图来重新加权剪切策略比率内的密集去噪过渡对数概率。在VBench-2.0上，TVRL的整体得分为57.69，比基础模型提高了3.60分。TVRL还在三个SDE采样器（SAGE、Flow和Dance）上提高了匹配的GRPO基线2.68至3.15分，在四个奖励模型（VideoAlign、VideoScore2、UnifiedReward2和Qwen3.5-9B）上提高了1.33至3.15分。
cs.CV / 147 / 2610.01989
Continual Concept Erasure in Diffusion Models by Suppressing Cross-Edit Interference
通过抑制交叉编辑干扰实现扩散模型中的持续概念抹除
Abstract
Concept erasure removes copyright-protected, privacy-sensitive, or otherwise undesirable concepts from pretrained text-to-image diffusion models to support content governance and compliance. As erasure requests arrive over time, models must remove new targets without undoing prior erasures. Existing methods do not constrain interference across edits: residual perturbations outside the retain set interact and accumulate, degrading unrelated generations and sometimes collapsing previously erased targets into noise. We propose CEASE (Continual Erasure via Adaptive Subspace Editing), a training-free method that imposes two subspace constraints on a closed-form solver. CEASE adds the token representation of the shared replacement to the solver's invariance matrix and, when interference is detected, projects the current update onto the orthogonal complement of dominant output directions extracted from cumulative past updates. A closed-form decomposition attributes the accumulated interference to repeated activation of the shared replacement and overlap between successive update directions, showing that the two constraints suppress these respective sources. Across continual erasure of celebrities, artistic styles, and instances, CEASE achieves the most consistent erase-preserve trade-off, while existing methods either degrade general generation or insufficiently erase targets.
Chinese Translation
概念抹除是指从预训练的文本到图像扩散模型中移除受版权保护、隐私敏感或其他不希望出现的概念，以支持内容治理和合规性。随着抹除请求的不断增加，模型必须在不撤销先前抹除的情况下移除新的目标。现有方法未能限制编辑之间的干扰：保留集外的残余扰动相互作用并累积，导致无关生成的退化，有时甚至将先前抹除的目标坍缩为噪声。我们提出了CEASE（通过自适应子空间编辑实现持续抹除），这是一种无训练的方法，对闭式解算器施加两个子空间约束。CEASE将共享替代的标记表示添加到解算器的不变性矩阵中，并在检测到干扰时，将当前更新投影到从累积的过去更新中提取的主导输出方向的正交补空间。闭式分解将累积的干扰归因于共享替代的重复激活和连续更新方向之间的重叠，表明这两个约束抑制了各自的干扰源。在对名人、艺术风格和实例的持续抹除中，CEASE实现了最一致的抹除-保留权衡，而现有方法要么降低了生成质量，要么未能充分抹除目标。
cs.CV / 148 / 2610.01994
Comparing a gradient boosting algorithm to the GOES FDC for wildfire detection
比较梯度提升算法与GOES火灾检测与特征提取（FDC）算法在野火检测中的应用
Abstract
Wildfires pose severe risks to human life, ecosystems, and property. This study presents a machine learning approach for wildfire detection from GOES ABI imagery. A CatBoost model was trained on a large dataset with thousands of ABI images and over 300,000 matching VIIRS fire detections. An evaluation on a separate dataset across five regions showed that the learned CatBoost model outperformed the operational GOES Fire Detection and Characterization (FDC) product. It achieved higher precision, recall, and F1 scores both within and outside the training area. The CatBoost model achieved F1 scores that were 0.16 to 0.38 higher than the GOES FDC in all regions. In addition, out of 51 historical fire events, the CatBoost detected 26 fires before both VIIRS and GOES FDC, compared to only six earlier detections by the GOES FDC. Importantly, the CatBoost model achieved accurate wildfire detection also during nighttime, whereas the GOES FDC obtained very low recall values, around 0.03. This study demonstrates that machine learning models may offer significant improvements over existing geostationary fire products, including higher accuracy, fewer false alarms, and earlier detection.
Chinese Translation
野火对人类生命、生态系统和财产构成严重威胁。本研究提出了一种基于机器学习的野火检测方法，利用GOES ABI影像进行分析。我们在一个包含数千张ABI影像和超过30万个匹配的VIIRS火灾检测结果的大型数据集上训练了一个CatBoost模型。在五个区域的独立数据集评估中，学习到的CatBoost模型的表现优于现有的GOES火灾检测与特征提取（FDC）产品。该模型在训练区域内外均实现了更高的精确度、召回率和F1分数。CatBoost模型在所有区域的F1分数比GOES FDC高出0.16到0.38。此外，在51个历史火灾事件中，CatBoost在VIIRS和GOES FDC之前检测到了26起火灾，而GOES FDC仅在六起事件中实现了更早的检测。重要的是，CatBoost模型在夜间也能准确检测野火，而GOES FDC的召回率非常低，仅约为0.03。本研究表明，机器学习模型可能在现有静止火灾产品的基础上提供显著改进，包括更高的准确性、更少的误报和更早的检测。
cs.CV / 149 / 2610.01999
From Reasoning Failures to Composable Video Spatial Intelligence
从推理失败到可组合的视频空间智能
Abstract
Spatial reasoning benchmarks evaluate vision-language models across diverse tasks, but task-level scores do not reveal which underlying capabilities account for success or failure. Each task requires recovering spatial evidence, representing geometry, and reasoning over it. We disentangle these capabilities by comparing predicted and ground-truth spatial context under a shared schema and coordinate contract. This comparison reveals four recurring sources of error: inaccurate perception, missing information in the spatial context, selection of the wrong measurement, and errors in reference frames or in tracking position and orientation. Guided by this diagnosis, we develop CROSS, a training-free library of typed geometric operators and spatial skills that function over available evidence to support reliable video spatial reasoning. The resulting library supplies verified context to non-coding VLMs or callable skills to a SpatialClaw agent. We evaluate \methodname{} on five benchmarks. \methodname{} raises the average score from 55.9\% to 60.2\% on ReVSI and improves the SpatialClaw result from 62.8\% to 66.3\% on DSI-Bench. These gains demonstrate that explicit handling of spatial conventions can repair systematic reasoning failures without additional training.
Chinese Translation
空间推理基准测试评估视觉-语言模型在多样任务中的表现，但任务级别的得分并未揭示哪些潜在能力导致成功或失败。每个任务都需要恢复空间证据、表示几何形状并进行推理。我们通过在共享模式和坐标约定下比较预测的空间上下文与真实的空间上下文，来解构这些能力。这一比较揭示了四个反复出现的错误来源：感知不准确、空间上下文中的信息缺失、选择错误的测量以及参考框架或位置和方向跟踪中的错误。在这一诊断的指导下，我们开发了CROSS，一个无训练的类型几何运算符和空间技能库，能够在可用证据上进行操作，以支持可靠的视频空间推理。该库为非编码的视觉-语言模型提供经过验证的上下文，或为SpatialClaw代理提供可调用的技能。我们在五个基准上评估了 extit{methodname}。 extit{methodname}将ReVSI的平均得分从55.9\%提高到60.2\\%，并将SpatialClaw在DSI-Bench上的结果从62.8\\%提升至66.3\\%。这些提升表明，明确处理空间约定可以修复系统性的推理失败，而无需额外的训练。
cs.CV / 150 / 2610.02000
Weather-Aware Domain Adaptation for Street-View Weather Recognition
天气感知的街景天气识别领域适应
Abstract
Adverse conditions such as rain, snow, fog, and dust remain challenging for camera-based perception in autonomous driving. We study multi-class weather recognition from street-view images under domain shift, where most available training data come from non-street-view sources that differ markedly from real driving scenes. We propose Weather-Aware Adversarial Discriminative Domain Adaptation (WA-ADDA), which conditions the domain discriminator on predicted weather to promote features that are both domain-invariant and weather-sensitive. We also assemble a multi-dataset benchmark by unifying diverse non-street-view weather collections as sources and real street-view images as targets, and define a standardized evaluation protocol with macro accuracy as the primary metric. Across backbones (ResNet-50, EfficientNet, VGG, DenseNet), WA-ADDA consistently improves street-view performance and yields strong per-class recalls in challenging conditions while preserving clear-weather accuracy. These findings highlight the feasibility of domain-adapted weather recognition and the value of our benchmark for advancing robust, on-board perception.
Chinese Translation
恶劣天气条件，如雨、雪、雾和尘埃，仍然对自动驾驶中的基于摄像头的感知构成挑战。我们研究了在领域转移下的街景图像的多类别天气识别，其中大多数可用的训练数据来自与真实驾驶场景显著不同的非街景来源。我们提出了天气感知对抗性判别领域适应（Weather-Aware Adversarial Discriminative Domain Adaptation，WA-ADDA），该方法将领域判别器的条件设置为预测的天气，以促进特征的领域不变性和天气敏感性。我们还通过将多样的非街景天气数据集统一为来源，并将真实街景图像作为目标，构建了一个多数据集基准，并定义了以宏观准确率为主要指标的标准化评估协议。在不同的骨干网络（ResNet-50、EfficientNet、VGG、DenseNet）中，WA-ADDA 一致地提高了街景性能，并在恶劣条件下实现了强大的每类召回率，同时保持了良好的晴天准确率。这些发现突显了领域适应天气识别的可行性，以及我们的基准在推动强健的车载感知方面的价值。
cs.CV / 151 / 2610.02010
Exploring Weaknesses of Generative Image Watermarks against Latent Frequency Masking
探讨生成图像水印在潜在频率掩蔽下的弱点
Abstract
Invisible watermarking has become a central tool for tracing AI-generated images, but its robustness against adaptive removal attacks remains an open security question. We introduce Latent Frequency Masking, an attack that erases watermark evidence by replacing selected Fourier coefficients in the latent representation of a watermarked image. The replacement can be sampled from Gaussian noise for efficiency or derived from diffusion regeneration for improved image preservation. We provide a theoretical distortion bound relating the change between the reconstructed adversarial image and the masked latent-frequency perturbation. We evaluate the proposed attack against six diffusion watermarking methods on images generated from DiffusionDB and MS-COCO prompts. Latent Frequency Masking removes or substantially weakens several watermarks while preserving perceptual quality and achieving favorable runtime compared with existing attacks. These results identify latent-frequency manipulation as a practical attack surface and highlight the need to include such attacks in robustness evaluations of generative image watermarking.
Chinese Translation
隐形水印已成为追踪AI生成图像的核心工具，但其对自适应去除攻击的鲁棒性仍然是一个未解的安全问题。我们提出了潜在频率掩蔽（Latent Frequency Masking），这是一种通过替换水印图像潜在表示中的选定傅里叶系数来消除水印证据的攻击。替换可以从高斯噪声中采样以提高效率，或从扩散再生中推导以改善图像保留。我们提供了一个理论失真界限，涉及重构的对抗图像与掩蔽的潜在频率扰动之间的变化。我们对六种扩散水印方法进行了评估，使用从DiffusionDB和MS-COCO提示生成的图像。潜在频率掩蔽在保留感知质量的同时，去除或显著削弱了多个水印，并与现有攻击相比实现了良好的运行时间。这些结果将潜在频率操控识别为一个实际的攻击面，并强调在生成图像水印的鲁棒性评估中纳入此类攻击的必要性。
cs.CV / 152 / 2610.02021
Task-Adaptive Grounded 3D-Programmers Using 2D VLMs
基于2D视觉语言模型的任务自适应三维程序员
Abstract
Recent vision-language models (VLMs) exhibit remarkable generalization and reasoning abilities, yet 3D understanding in these models is limited by data scale, training diversity, and reasoning capacity. Instead of naively extending these models into 3D, we take a different approach: we enable powerful 2D VLMs to operate reliably in 3D by introducing 3D grounding and iterative feedback loops with two novel concepts: Canonical Coordinate Framing (CCF) and Task-Adaptive Feedback (TAF). CCF serves as a unified visual representation that anchors both inputs and outputs to a shared Euclidean coordinate system, solving common challenges in 3D grounding such as axis ambiguity, inconsistent metric scale, and floating references. Complementary to this structured framing of the 3D inputs, TAF closes the reasoning loop with task-adaptive dynamic feedback that enables 2D VLMs to perform varied open-vocabulary tasks within their native visual context. Building on this foundation, we introduce 3D-Prog, a 3D understanding, reasoning, and generation framework that jointly employs the capabilities of CCF and TAF together with powerful VLMs. Without requiring any retraining, 3D-Prog performs open-vocabulary 3D understanding, manipulation, and generation across both object-level and scene-level tasks. Our experiments show that the joint use of CCF and TAF transforms 2D VLMs into geometry-aware 3D programmers, achieving consistent, interpretable, and high-quality results across diverse 3D tasks.
Chinese Translation
近期的视觉语言模型（VLMs）展现了显著的泛化和推理能力，但这些模型在三维理解方面受到数据规模、训练多样性和推理能力的限制。我们并没有简单地将这些模型扩展到三维，而是采取了不同的方法：通过引入三维基础和迭代反馈循环，使强大的2D VLMs能够在三维空间中可靠地运行，提出了两个新概念：典范坐标框架（Canonical Coordinate Framing, CCF）和任务自适应反馈（Task-Adaptive Feedback, TAF）。CCF作为统一的视觉表示，将输入和输出锚定到共享的欧几里得坐标系统中，解决了三维基础中常见的挑战，如轴模糊、不一致的度量尺度和浮动参考。与这种结构化的三维输入框架相辅相成，TAF通过任务自适应的动态反馈闭合推理循环，使2D VLMs能够在其原生视觉上下文中执行多样的开放词汇任务。在此基础上，我们引入了3D-Prog，一个三维理解、推理和生成框架，联合利用CCF和TAF的能力以及强大的VLMs。无需任何重新训练，3D-Prog能够在对象级和场景级任务中执行开放词汇的三维理解、操作和生成。我们的实验表明，CCF和TAF的联合使用将2D VLMs转变为具有几何意识的三维程序员，在多样的三维任务中实现了一致、可解释和高质量的结果。
cs.CV / 153 / 2610.02044
DiDE:Direct Injection with Color-Texture DEcoupling for 3D Stylization
DiDE：用于3D风格化的直接注入与颜色-纹理解耦
Abstract
Recent advances in rectified flow-based image-to-3D generative models have enabled high-fidelity 3D asset generation. Building on this, a growing line of work has exploited these strong 3D priors for training-free stylization, transferring visual attributes from a reference image onto a generated 3D asset. However, existing methods enforce an all-or-nothing paradigm: color and texture are transferred jointly, with no mechanism to control them independently -- a limitation we formalize as Disentangled 3D Stylization(Disen3D). To address this, we propose DiDE, the first training-free framework for Disen3D. Key to our approach is the observation that the structured latent space of image-to-3D models is overcomplete with respect to texture: texture information occupies only a small subset of the style-significant channels, leaving a free subspace available for independent color encoding. DiDE exploits this via a channel partition mechanism that processes a content image, a texture reference, and a color reference through dedicated branches and composes both style signals interference-free at every self-attention layer, preserving content geometry throughout. Experiments on Disen3D-Bench, our newly collected multi-reference benchmark, show that DiDE consistently outperforms 2D and 3D stylization baselines in color fidelity, texture transfer, and content preservation.
Chinese Translation
最近在基于流的图像到3D生成模型的改进，使得高保真3D资产生成成为可能。在此基础上，越来越多的研究利用这些强大的3D先验进行无训练风格化，将视觉属性从参考图像转移到生成的3D资产上。然而，现有方法强制执行全有或全无的范式：颜色和纹理被联合转移，没有机制可以独立控制它们——这一限制我们将其形式化为解耦3D风格化（Disentangled 3D Stylization，Disen3D）。为了解决这个问题，我们提出了DiDE，这是第一个用于Disen3D的无训练框架。我们方法的关键在于观察到图像到3D模型的结构化潜在空间在纹理方面是过完备的：纹理信息仅占据风格显著通道的一个小子集，留下一个可用于独立颜色编码的自由子空间。DiDE通过通道划分机制利用这一点，分别处理内容图像、纹理参考和颜色参考，并在每个自注意力层中无干扰地组合这两种风格信号，同时保持内容几何形状。我们在Disen3D-Bench上进行的实验（这是我们新收集的多参考基准）表明，DiDE在颜色保真度、纹理转移和内容保留方面始终优于2D和3D风格化基线。
cs.CV / 154 / 2610.02045
Form and Void: Entangled Composition through an Autonomous AI Agent
形与虚：通过自主AI代理的纠缠构图
Abstract
Positive and negative space is a fundamental principle in visual composition, supporting visually coherent forms and layered semantic relationships. Generating such compositions is challenging because it requires coordinated control over two semantic concepts that share a common boundary. Although recent text-to-image models and multimodal large language models (MLLMs) have achieved strong performance in image generation and visual understanding, positive-negative space generation remains difficult, particularly under direct single-pass prompting. In this work, we present the \textbf{F}orm \textbf{a}nd \textbf{V}oid \textbf{A}gent (\textbf{FaV-A}), a multimodal agent designed for staged positive-negative space generation. FaV-A follows a progressive workflow: it first generates a base object, then analyzes its shape and spatial structure to identify candidate negative-space semantics, and finally produces compositional instructions for the final image generation stage. Experimental results and ablation analyses suggest that FaV-A provides a more effective framework than direct zero-shot MLLM baselines for producing visually coherent and semantically aligned positive-negative space compositions.
Chinese Translation
正负空间是视觉构图中的一个基本原则，支持视觉上连贯的形式和层次化的语义关系。生成这样的构图具有挑战性，因为它需要对两个共享共同边界的语义概念进行协调控制。尽管最近的文本到图像模型和多模态大型语言模型（MLLMs）在图像生成和视觉理解方面取得了良好的性能，但正负空间的生成仍然困难，特别是在直接的单次提示下。在本研究中，我们提出了 extbf{F}orm extbf{a}nd extbf{V}oid extbf{A}gent（ extbf{FaV-A}），这是一个旨在分阶段生成正负空间的多模态代理。FaV-A遵循一个渐进的工作流程：首先生成一个基础对象，然后分析其形状和空间结构以识别候选的负空间语义，最后为最终图像生成阶段提供构图指令。实验结果和消融分析表明，FaV-A提供了比直接零-shot MLLM基线更有效的框架，以生成视觉上连贯且语义上对齐的正负空间构图。
cs.CV / 155 / 2610.02051
Learning from Failure: Leveraging Unreliable Predictions in Semi-Supervised Real-World Adverse Weather Removal
从失败中学习：在半监督真实世界恶劣天气去除中利用不可靠预测
Abstract
Adverse weather image restoration aims to recover images degraded by rain, haze, snow, and other weather-induced artifacts, thereby improving the robustness of outdoor vision systems. Existing unified restoration models exhibit limited generalization to real-world scenes due to their reliance on synthetic supervision and insufficient semantic constraints. In this paper, we propose a novel student--teacher semi-supervised framework that addresses both challenges. Specifically, we introduce an unreliable database that preserves failed teacher predictions as informative negative samples for contrastive learning, while a reliable database stores high-quality teacher predictions as positive samples. By jointly exploiting reliable pseudo-ground truths and unreliable teacher outputs, the proposed framework learns to enhance desirable restoration characteristics while avoiding common failures. We further propose a phase spectrum-based semantic constraint that replaces computationally expensive text-based supervision with an efficient and naturally aligned semantic prior. An adaptive phase consistency loss is also designed to dynamically balance supervision between the degraded input and teacher pseudo-ground truths according to degradation severity. Extensive experiments on real-world benchmarks demonstrate that the proposed method consistently outperforms existing state-of-the-art approaches in restoration quality and perceptual fidelity while exhibiting stronger generalization to real-world adverse weather conditions.
Chinese Translation
恶劣天气图像恢复旨在恢复因雨、雾、雪及其他天气引起的伪影而退化的图像，从而提高户外视觉系统的鲁棒性。现有的统一恢复模型由于依赖于合成监督和不足的语义约束，对真实场景的泛化能力有限。本文提出了一种新颖的学生-教师半监督框架，以解决这两个挑战。具体而言，我们引入了一个不可靠数据库，该数据库保留了失败的教师预测作为对比学习的信息性负样本，而可靠数据库则存储高质量的教师预测作为正样本。通过共同利用可靠的伪地面真相和不可靠的教师输出，所提出的框架学习增强期望的恢复特性，同时避免常见的失败。我们进一步提出了一种基于相位谱的语义约束，用高效且自然对齐的语义先验替代计算成本高昂的基于文本的监督。同时，设计了一种自适应相位一致性损失，根据退化严重程度动态平衡退化输入与教师伪地面真相之间的监督。大量在真实世界基准上的实验表明，所提方法在恢复质量和感知保真度上始终优于现有的最先进方法，同时在真实世界恶劣天气条件下表现出更强的泛化能力。
cs.CV / 156 / 2610.02091
GeoLatent: Geometry-Guided Latent Structuring with Routed Optimization for 3D Reasoning
GeoLatent：基于几何引导的潜在结构与路由优化的3D推理
Abstract
Despite progress in vision-language models, 3D spatial reasoning from 2D images remains challenging. Text-based methods describe intermediate geometry with discrete tokens, limiting fidelity for continuous spatial relations. Continuous latents offer richer representations, but a single latent type does not explicitly separate the cues needed across spatial tasks. Decomposed spatial latents address this by representing position, direction, and global geometry separately under geometric supervision. Yet the geometry representation can still collapse toward one dominant direction, and unrestricted attention can leave the latents underused during answer learning. We introduce GeoLatent, combining Common--Residual Geometry Alignment (CR-GEO) with routed optimization to structure the geometry states while promoting latent-mediated answer learning. CR-GEO separates shared from residual teacher geometry; routed optimization jointly trains geometry and language, temporarily directs visual answer learning through the latents, and restores full attention with geometry supervision. In controlled comparisons, CR-GEO raises geometry effective rank from 1.00 to 3.87, while blocking latent readout at the bottleneck lowers direction accuracy from 89.1% to 25.8% on 128 fixed questions. After recovery, the differentiated geometry representation and latent-mediated visual route remain available alongside direct image access. GeoLatent achieves 73.0% on SPAR-Bench and 72.1% on SPBench, outperforming previously reported methods on both.
Chinese Translation
尽管视觉-语言模型取得了进展，但从2D图像进行3D空间推理仍然具有挑战性。基于文本的方法使用离散标记描述中间几何，这限制了对连续空间关系的保真度。连续潜在变量提供了更丰富的表示，但单一的潜在类型并没有明确区分在空间任务中所需的线索。分解的空间潜在变量通过在几何监督下分别表示位置、方向和全局几何来解决这一问题。然而，几何表示仍可能向一个主导方向收敛，而不受限制的注意力可能导致潜在变量在答案学习过程中未被充分利用。我们提出了GeoLatent，将共同-残差几何对齐（Common--Residual Geometry Alignment, CR-GEO）与路由优化相结合，以结构化几何状态，同时促进潜在介导的答案学习。CR-GEO将共享几何与残差教师几何分开；路由优化联合训练几何和语言，暂时通过潜在变量引导视觉答案学习，并在几何监督下恢复完全注意力。在受控比较中，CR-GEO将几何有效秩从1.00提高到3.87，而在瓶颈处阻止潜在变量读出使方向准确率从89.1%下降到25.8%（在128个固定问题上）。恢复后，差异化的几何表示和潜在介导的视觉路径仍然可用，并与直接图像访问并存。GeoLatent在SPAR-Bench上取得了73.0%的成绩，在SPBench上取得了72.1%的成绩，均优于先前报告的方法。
cs.CV / 157 / 2610.02114
Surface-volume self-supervised representation learning of brain MRI for genetic discovery
脑MRI的表面-体积自监督表示学习用于遗传发现
Abstract
Existing genome-wide association studies (GWAS) of brain imaging provide predefined or deep-learning-derived imaging phenotypes, yet these phenotypes come from either volumetric scans or cortical surface meshes, so each captures only part of the heritable variation in brain anatomy. Here we introduce MEVA (Mesh-Enhanced Volumetric Autoencoder), a self-supervised framework that encodes voxel-level image intensity together with cortical mesh geometry, including curvature and cortical thickness at each surface vertex, into one shared set of imaging features. Combining the mesh and volumetric inputs in MEVA yields modest performance gains in age and sex prediction over models that use either input alone. When these features serve as phenotypes for GWAS in the UK Biobank, they reveal more genome-wide significant loci than features learned from volumes alone or from meshes alone. These results suggest that adding cortical surface geometry to volumetric self-supervised learning captures additional heritable variation and so increases the number of loci detected.
Chinese Translation
现有的脑成像全基因组关联研究（GWAS）提供了预定义或深度学习衍生的成像表型，然而这些表型来自于体积扫描或皮层表面网格，因此每种方法仅捕捉到脑解剖结构中部分可遗传变异。在此，我们介绍了MEVA（Mesh-Enhanced Volumetric Autoencoder），一种自监督框架，它将体素级图像强度与皮层网格几何（包括每个表面顶点的曲率和皮层厚度）编码为一组共享的成像特征。将MEVA中的网格和体积输入结合起来，在年龄和性别预测方面相较于单独使用任一输入的模型取得了适度的性能提升。当这些特征作为UK Biobank中的GWAS表型时，它们揭示了比仅从体积或仅从网格学习的特征更多的全基因组显著位点。这些结果表明，将皮层表面几何信息添加到体积自监督学习中能够捕捉到额外的可遗传变异，从而增加检测到的位点数量。
cs.CV / 158 / 2610.02117
Where-OPD: Spatially Guided On-Policy Self-Distillation of MLLMs with Synthetic Scenes
Where-OPD：基于空间引导的多模态大语言模型的在线自蒸馏与合成场景
Abstract
On-policy self-distillation has recently emerged as an effective approach for improving language-model reasoning by supervising students with a frozen or EMA version of themselves that receives privileged information. Its application to multimodal large language models (MLLMs), however, remains largely unexplored. Recent approaches use privileged visual information, such as image crops corresponding to a question, to improve fine-grained perception, but their gains are confined to tasks that benefit from such visual zooming and require either human-annotated grounding data or external teacher models. We introduce a different form of on-policy self-distillation for MLLMs that provides the teacher with textual, spatially grounded guidance identifying the visual elements relevant to a query. We use procedurally generated scenes with automatically available object identities and spatial coordinates, enabling scalable and annotation-free post-training. The teacher uses this spatial guidance to locate and integrate evidence from multiple relevant image regions, while the student learns to reproduce the resulting behavior from the image and question alone. Our approach consistently improves performance on counting, document and chart understanding benchmarks across multiple models. Importantly, although post-training uses only synthetic scenes, the resulting improvements transfer to real-world perception benchmarks, yielding a 3.23-point gain in average performance across CVBench, V*, ZoomBench, BLINK, HR-Bench, and MME-RealWorld. These results show that spatially grounded privileged information can induce broader perceptual capabilities through on-policy self-distillation, enabling substantial synthetic-to-real transfer beyond the task and data distribution used for post-training. Project page: https://github.com/sirkosophia/Where-OPD
Chinese Translation
在线自蒸馏最近成为一种有效的方法，通过使用冻结或EMA版本的自身来监督学生，从而提高语言模型的推理能力，并为其提供特权信息。然而，这种方法在多模态大语言模型（MLLMs）中的应用仍然 largely 未被探索。近期的方法利用特权视觉信息，例如与问题对应的图像裁剪，以改善细粒度感知，但其收益仅限于受益于这种视觉放大且需要人工标注的基础数据或外部教师模型的任务。我们提出了一种不同形式的MLLM在线自蒸馏方法，为教师提供文本的、空间上有依据的指导，以识别与查询相关的视觉元素。我们使用程序生成的场景，自动提供对象身份和空间坐标，从而实现可扩展且无需标注的后训练。教师利用这些空间指导来定位并整合来自多个相关图像区域的证据，而学生则学习仅通过图像和问题重现所得到的行为。我们的方法在多个模型的计数、文档和图表理解基准上持续提高了性能。重要的是，尽管后训练仅使用合成场景，但所得到的改进能够转移到现实世界的感知基准上，在CVBench、V*、ZoomBench、BLINK、HR-Bench和MME-RealWorld上平均性能提高了3.23分。这些结果表明，基于空间的特权信息能够通过在线自蒸馏引发更广泛的感知能力，实现超越后训练所用任务和数据分布的显著合成到真实的转移。项目页面：https://github.com/sirkosophia/Where-OPD
cs.CV / 159 / 2610.02123
Harnessing Domain Specialists in Multimodal Mixture-of-Experts for Efficient Adaptation
利用领域专家在多模态专家混合模型中的高效适应
Abstract
Mixture-of-Experts (MoE) architectures scale model capacity through sparse computation, routing each token through only a small subset of experts. In this work, we explore whether this sparsity gives rise to emergent intrinsic organization in multimodal MoEs. We find that experts develop strong semantic specialization across modalities and domains despite not being explicitly trained for modularity. Building on this structure, we introduce ExpertLens, a data-free method that identifies domain-specialized experts directly from pretrained model weights by decoding router weights into semantically meaningful vocabulary tokens. We leverage this specialization for efficient multimodal adaptation by selectively fine-tuning experts relevant to a target domain. Across math, medical, and remote sensing tasks, ExpertLens matches or surpasses full fine-tuning while updating only 21.7 - 47.0% of model parameters and achieving a 4.0x average training speedup, and outperforms LoRA in both adaptation performance and training efficiency. These results show that sparsity introduced for efficiency can give rise to semantic modularity that is directly useful for efficient adaptation.
Chinese Translation
专家混合（Mixture-of-Experts, MoE）架构通过稀疏计算扩展模型容量，仅将每个标记路由到少量专家。在本研究中，我们探讨这种稀疏性是否会在多模态MoE中产生内在的自组织现象。我们发现，尽管专家并未明确接受模块化训练，但它们在不同模态和领域之间发展出强烈的语义专业化。在此基础上，我们提出了ExpertLens，这是一种无数据的方法，通过将路由器权重解码为语义上有意义的词汇标记，直接从预训练模型权重中识别领域专业化的专家。我们利用这种专业化，通过选择性微调与目标领域相关的专家，实现高效的多模态适应。在数学、医学和遥感任务中，ExpertLens在仅更新21.7% - 47.0%的模型参数的情况下，匹配或超越了完全微调的效果，并实现了4.0倍的平均训练速度提升，同时在适应性能和训练效率上均优于LoRA。这些结果表明，为了提高效率而引入的稀疏性可以产生直接有助于高效适应的语义模块化。
cs.CV / 160 / 2610.02136
MIRTO: a registration-gated, multiverse-tested evaluation protocol for unsupervised anomaly segmentation in brain MRI
MIRTO：一种注册门控、多宇宙测试的无监督异常分割评估协议，用于脑部MRI
Abstract
Unsupervised anomaly detection (UAD) methods for brain MRI are ranked by a single score, yet that score rests on choices that are rarely reported: how each anomaly map is aligned with the reference, how and on which data the threshold is set, and which false-positive budget, metric, aggregation and lesion definition are used. We present MIRTO, an evaluation protocol that makes these choices explicit and measures their effect. It gates the geometry of every comparison with a registration check and label-free diagnostics of known power, sets thresholds on validation data alone and reports the false-positive volume actually realised on test, repeats each comparison over 15,552 defensible evaluation pipelines, and attaches paired subject-bootstrap intervals with multiplicity control. Applied to four UAD methods trained on the same healthy data and tested on 312 BraTS 2020 subjects, MIRTO showed that an axis-order mismatch between stored maps and the reference lowered a diffusion model's voxel AUROC from 0.873 to 0.583 whilst barely moving its slice-level AUROC. Within each metric, the method explained at least 0.95 of the variance in voxel AUROC and AUPRC and 0.77 in Dice, but only 0.14 in lesion sensitivity, where the lesion definition and hit criterion dominated. A Dice advantage that was significant at validation thresholds vanished at equal realised false-positive burden, and an exact identity attributes it to threshold transfer. A training-free change to REFLECT's latent aggregation raised Dice at equal burden by 0.052. Nine hypotheses were tested against explicit criteria; because the same cohort served to develop the protocol, all inference is exploratory.
Chinese Translation
脑部MRI的无监督异常检测（UAD）方法通过单一评分进行排名，但该评分基于的选择往往未被报告：每个异常图如何与参考图对齐、阈值如何设置以及在哪些数据上设置、使用了哪些假阳性预算、指标、聚合和病变定义。我们提出了MIRTO，一种评估协议，使这些选择变得明确并测量其影响。它通过注册检查和已知效能的无标签诊断来限制每次比较的几何形状，仅在验证数据上设置阈值，并报告在测试中实际实现的假阳性体积，重复每次比较15,552个可辩护的评估管道，并附加成对的受试者自助区间以控制多重性。应用于在相同健康数据上训练并在312个BraTS 2020受试者上测试的四种UAD方法，MIRTO显示存储图与参考图之间的轴序不匹配使得扩散模型的体素AUROC从0.873降至0.583，而其切片级AUROC几乎没有变化。在每个指标中，该方法解释了体素AUROC和AUPRC中至少0.95的方差，Dice中为0.77，但在病变敏感性中仅为0.14，其中病变定义和命中标准占主导地位。在验证阈值下显著的Dice优势在实现相等的假阳性负担时消失，且一个精确的身份将其归因于阈值转移。对REFLECT的潜在聚合进行无训练的改变，使得在相等负担下Dice提高了0.052。针对明确标准测试了九个假设；由于相同的队列用于开发该协议，因此所有推断均为探索性。
cs.CV / 161 / 2610.02148
Omni-Embed-Mini: Binding Modalities Without Forgetting via Dense Distillation
Omni-Embed-Mini：通过密集蒸馏绑定模态而不遗忘
Abstract
Extending a text embedding model to new modalities typically degrades text retrieval quality, and existing omni-modal embedders compensate with multi-billion parameters. We present Omni-Embed-Mini, a 0.9B-parameter model that maps text, speech, audio, images, video, and visually-rich documents into a single shared cosine space without updating any text-side parameter. Our key insight is that the teacher signal requires no separate embedding model: each media sample is paired with a dense cascaded caption, and the teacher target is simply the frozen backbone's own embedding of that caption. Because teacher and student share the same backbone weights, they inhabit byte-identical geometry, and lightweight projectors plus phased LoRA adapters on the modality encoders suffice for alignment. Training combines a Matryoshka SigLIP contrastive loss with an online hybrid hard-negative miner whose negatives sharpen as the encoder improves. The recipe carries over to a 2.3B variant by swapping in a native vision-language backbone. Omni-Embed-Mini-0.9B keeps its text weights bit-identical to the backbone, so training cannot regress text retrieval (49.57 nDCG@10 on MTEB-v2 BEIR-8), while extending it to five additional modalities, and is ~2.7x to 9.5x smaller than every open omni embedder we compare against. The 2.3B variant is competitive with the closed gemini-embedding-2, edging ahead of it on the overall-modality average. Models, code, data and evaluation harness are on our project page: https://omniembed.cvmbzuai.com
Chinese Translation
将文本嵌入模型扩展到新模态通常会降低文本检索质量，而现有的全模态嵌入器则通过数十亿参数进行补偿。我们提出了Omni-Embed-Mini，一个拥有0.9B参数的模型，它将文本、语音、音频、图像、视频和视觉丰富的文档映射到一个共享的余弦空间中，而无需更新任何文本侧的参数。我们的关键见解是，教师信号不需要单独的嵌入模型：每个媒体样本都与一个密集级联的标题配对，教师目标仅仅是冻结的主干网络对该标题的嵌入。由于教师和学生共享相同的主干权重，它们在字节上是相同的几何结构，轻量级投影器加上阶段性LoRA适配器在模态编码器上足以实现对齐。训练结合了Matryoshka SigLIP对比损失和一个在线混合硬负样本挖掘器，其负样本随着编码器的改进而变得更加尖锐。该方法可以扩展到一个2.3B的变体，通过替换为本地的视觉-语言主干。Omni-Embed-Mini-0.9B保持其文本权重与主干完全相同，因此训练不会导致文本检索的退化（在MTEB-v2 BEIR-8上为49.57 nDCG@10），同时将其扩展到五个额外的模态，并且比我们比较的每个开放全嵌入器小约2.7倍到9.5倍。2.3B的变体在整体模态平均上与封闭的gemini-embedding-2具有竞争力，并在此基础上略有领先。模型、代码、数据和评估工具可在我们的项目页面上找到：https://omniembed.cvmbzuai.com
cs.CV / 162 / 2610.02153
MosaiChunk: Compositing Spatio-Temporal Memory for Autoregressive Video Generation
MosaiChunk：自回归视频生成的时空记忆组合
Abstract
Long-horizon autoregressive video generation is limited by a finite context window. When an object or scene falls out of context, its fine-grained visual details may be lost and difficult to recover upon reappearance. To retain access to such visual details, we introduce MosaiChunk, a spatio-temporal memory mechanism that composes a mosaic of selected historical key-value (KV) entries across space and time. Our approach is motivated by the observation that a frozen video generator can directly consume such non-contiguous historical KV and recover the corresponding visual content. We therefore keep the generator fixed and learn only a lightweight router that determines which historical sections to include in the mosaic under a fixed active-memory budget. We further introduce RememBench, a benchmark of long-horizon revisits with prompt-driven text-to-video (T2V) and camera-driven image-to-video (I2V) splits. Our experiments show that MosaiChunk consistently improves revisit consistency over both sliding-window inference and whole-chunk retrieval under matched memory budgets, across both T2V and I2V settings.
Chinese Translation
长时间跨度的自回归视频生成受到有限上下文窗口的限制。当一个物体或场景超出上下文时，其细致的视觉细节可能会丢失，并且在重新出现时难以恢复。为了保留对这些视觉细节的访问，我们提出了MosaiChunk，这是一种时空记忆机制，通过在空间和时间上组合选定的历史关键值（KV）条目来形成马赛克。我们的方法受到以下观察的启发：一个冻结的视频生成器可以直接处理这些不连续的历史KV，并恢复相应的视觉内容。因此，我们保持生成器固定，仅学习一个轻量级的路由器，以确定在固定的活动记忆预算下应包含哪些历史部分。我们进一步引入了RememBench，这是一个长时间跨度重访的基准，包含基于提示的文本到视频（T2V）和基于相机的图像到视频（I2V）分割。我们的实验表明，在匹配的记忆预算下，MosaiChunk在T2V和I2V设置中，始终提高了重访一致性，优于滑动窗口推理和整体块检索。
cs.CV / 163 / 2610.02160
4Director: Controlling Video World Models with Rigid 3D Geometry
4Director：使用刚性三维几何体控制视频世界模型
Abstract
Precise control over camera and object motion is essential for professional video production. Existing methods control objects only coarsely, through image-plane cues that are ambiguous in depth and rotation or through 3D tracks and blobs that lack complete geometry and lose consistency across viewpoint changes. We introduce 4Director, a video world model conditioned on an explicit 4D scene representation: each object is reconstructed once from the input image as a canonical mesh and moved by one prescribed rigid transformation per frame. This representation provides an intuitive 3D control interface and prevents unobserved geometry from being regenerated independently in every frame. We render the controlled scene as a depth video and introduce a Motion Adapter that transforms this geometric scaffold into video while synthesizing view-consistent appearance, illumination, and non-rigid dynamics. For training, we construct RealCOD-Rigid, a new dataset of 20,774 clips annotated with rigid 3D scenes by our automatic pipeline. We further introduce Identity-Gated IoU (IG-IoU), which jointly evaluates adherence to prescribed object motion and preservation of object identity. Experiments demonstrate that 4Director consistently outperforms prior methods in visual quality and in camera and object control.
Chinese Translation
对摄像机和物体运动的精确控制对于专业视频制作至关重要。现有方法仅通过图像平面线索粗略控制物体，这些线索在深度和旋转上存在歧义，或通过缺乏完整几何体且在视点变化中失去一致性的三维轨迹和斑块。我们提出了4Director，这是一种基于明确的4D场景表示的视频世界模型：每个物体从输入图像中重建为标准网格，并在每帧中通过一个规定的刚性变换进行移动。这种表示提供了直观的三维控制界面，并防止未观察到的几何体在每帧中独立再生。我们将控制的场景渲染为深度视频，并引入了运动适配器（Motion Adapter），该适配器将这种几何支架转换为视频，同时合成视点一致的外观、照明和非刚性动态。为了训练，我们构建了RealCOD-Rigid，这是一个包含20,774个片段的新数据集，使用我们的自动化管道标注了刚性三维场景。我们进一步引入了身份门控交并比（Identity-Gated IoU，IG-IoU），该指标共同评估规定的物体运动的遵循程度和物体身份的保持。实验表明，4Director在视觉质量以及摄像机和物体控制方面始终优于先前的方法。
cs.CV / 164 / 2610.02162
World Observer: Joint Actor-Observer Generation for Persistent World Modeling
世界观察者：用于持久世界建模的联合行为者-观察者生成
Abstract
How can a world model continuously observe regions beyond the actor's current view? Video world models simulate how an environment evolves from an agent's actions, yet remain actor-centric. Once an object leaves the actor's view, they lose direct evidence of its evolution, often failing to preserve its state and dynamics upon re-entry. To address this, we introduce World Observer, which decouples observing from acting by jointly generating a perspective actor for the agent-centric view with one or more panoramic observers that watch selected world regions. This allows objects that leave the actor's view to remain visually evolving in an observer, so their updated states are reflected when they re-enter. We ground the actor and observers by warping from a shared panoramic source for explicit geometric correspondence, and introduce an Observer Sink of high-resolution perspective references to restore fine appearance upon re-entry. Since the observers are decoupled from the actor, they can be placed freely across the scene, extended to multiple locations for broader coverage, and driven by control signals to steer out-of-view evolution. To evaluate out-of-view evolution, we further introduce world-space metrics and a benchmark spanning real and synthetic scenes. World Observer substantially improves out-of-view dynamics while remaining competitive in visual fidelity, camera control, and 3D adherence.
Chinese Translation
如何让一个世界模型持续观察超出行为者当前视野的区域？视频世界模型模拟环境如何根据代理的行为演变，但仍然以行为者为中心。一旦一个物体离开行为者的视野，他们便失去了对其演变的直接证据，往往无法在物体重新进入时保留其状态和动态。为了解决这个问题，我们提出了世界观察者（World Observer），它通过联合生成一个视角行为者与一个或多个全景观察者来解耦观察与行为，这些观察者观察选定的世界区域。这使得离开行为者视野的物体在观察者中保持视觉演变，因此它们的更新状态在重新进入时得以反映。我们通过从共享的全景源进行扭曲来为行为者和观察者提供明确的几何对应，并引入一个高分辨率视角参考的观察者汇（Observer Sink），以在重新进入时恢复细致的外观。由于观察者与行为者解耦，它们可以自由地放置在场景中，扩展到多个位置以实现更广泛的覆盖，并通过控制信号驱动超出视野的演变。为了评估超出视野的演变，我们进一步引入了世界空间度量和一个涵盖真实与合成场景的基准。世界观察者在保持视觉保真度、相机控制和三维一致性的同时，显著改善了超出视野的动态表现。
cs.CV / 165 / 2610.02180
Generative Cinematographer: Composing Camera and Object Motion in 3D
生成电影摄影师：在三维中构建相机与物体运动
Abstract
Current controllable video generation systems often rely on 2D motion trajectories or sparse drag signals for object motion. These controls are ambiguous because the same 2D trajectory can correspond to different 3D motions, especially when the camera and objects move simultaneously. We present Generative Cinematographer (GenCine), a system that lifts a single image into an editable 3D scene scaffold where artists jointly author camera and foreground motion. Artists specify a camera path and move selected foreground regions using local 3D motion handles. Several handles can move different parts of a subject independently, providing a piecewise-rigid approximation to non-rigid motion without a physics simulator or category-specific prior. To communicate these controls to a pretrained video model, we project them into guidance maps. These maps record where the controlled regions appear in each frame, assign each handle a fixed color across frames and encode the current 3D positions of its controlled points in the same world coordinate system as the background. This lets us describe object motion relative to the scene even as the camera moves. For training, we recover controls from the motion observed in real videos and use ground-truth geometry and trajectories from synthetic videos. We train a lightweight guidance branch and LoRA adapters on a pretrained Wan model to follow these controls. Our experiments show consistent camera-relative motion, improved geometric consistency under viewpoint changes, and strong controllability across diverse real-world scenes.
Chinese Translation
当前的可控视频生成系统通常依赖于二维运动轨迹或稀疏的拖动信号来控制物体运动。这些控制方式存在模糊性，因为相同的二维轨迹可能对应于不同的三维运动，尤其是在相机和物体同时移动时。我们提出了生成电影摄影师（Generative Cinematographer, GenCine）系统，该系统将单幅图像提升为一个可编辑的三维场景框架，艺术家可以在其中共同创作相机和前景运动。艺术家指定相机路径，并使用局部三维运动手柄移动选定的前景区域。多个手柄可以独立移动对象的不同部分，提供对非刚性运动的分段刚性近似，而无需物理模拟器或特定类别的先验知识。为了将这些控制传达给预训练的视频模型，我们将其投影到引导图中。这些图记录了受控区域在每帧中的出现位置，为每个手柄在各帧中分配固定颜色，并在与背景相同的世界坐标系中编码其受控点的当前三维位置。这使我们能够在相机移动的同时相对于场景描述物体运动。为了进行训练，我们从真实视频中观察到的运动中恢复控制，并使用合成视频中的真实几何和轨迹。我们在预训练的Wan模型上训练一个轻量级的引导分支和LoRA适配器，以跟随这些控制。我们的实验表明，在相机相对运动下具有一致性，视角变化下几何一致性得到改善，并且在多样的真实场景中表现出强大的可控性。
cs.CV / 166 / 2610.02181
OmniSeek: Native Tool Integration for Multi-turn Audio-Visual Reasoning
OmniSeek：多轮音视频推理的原生工具集成
Abstract
We present OmniSeek, an agentic framework that transforms an Omni Large Language Model (Omni-LLM) into an active, multi-turn reasoning agent with native tool use. Rather than passively processing an entire audio-visual sequence in a single forward pass, OmniSeek makes evidence acquisition part of the reasoning process: it dynamically decides whether to look or listen, and over which temporal window, to retrieve sparse but critical evidence across different modalities within long contexts. Through an iterative multi-turn protocol, the retrieved raw audio or visual segments are appended back into the context to support subsequent reasoning. To cold-start this capability, we build a data engine that synthesizes OmniTraj-170K, a corpus of multi-hop Chain-of-Thought trajectories with interleaved audio and visual evidence. We first supervise the model on these trajectories to instill multi-turn tool-use behavior, and then further optimize the policy via a two-stage reinforcement learning with verifiable rewards. Moreover, we introduce an Audio-Visual Necessity objective that explicitly rewards successful trajectories whose reasoning depends on both modalities, discouraging single-modality shortcuts. Extensive experiments across a wide range of benchmarks demonstrate that OmniSeek learns adaptive cross-modal evidence seeking and consistently improves audio-visual reasoning performance.
Chinese Translation
我们提出了OmniSeek，一个代理框架，将Omni大型语言模型（Omni-LLM）转变为一个具有原生工具使用功能的主动多轮推理代理。OmniSeek并不是被动地在单次前向传递中处理整个音视频序列，而是将证据获取纳入推理过程：它动态决定是否查看或收听，以及在何种时间窗口内，以便在长上下文中检索不同模态下稀疏但关键的证据。通过迭代的多轮协议，检索到的原始音频或视觉片段被重新附加到上下文中，以支持后续推理。为了启动这一能力，我们构建了一个数据引擎，合成了OmniTraj-170K，这是一个包含交错音频和视觉证据的多跳思维链轨迹语料库。我们首先对模型进行监督训练，以培养多轮工具使用行为，然后通过可验证奖励的两阶段强化学习进一步优化策略。此外，我们引入了音视频必要性目标，明确奖励那些推理依赖于两种模态的成功轨迹，抑制单一模态的捷径。广泛的实验结果表明，OmniSeek能够学习自适应的跨模态证据寻求，并持续提高音视频推理性能。
cs.CV / 167 / 2610.02188
DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation
DMAD：作为对抗蒸馏的分布匹配用于快速视觉生成
Abstract
Distribution Matching Distillation (DMD) trains a few-step student from the difference between separately estimated target and student scores, so it must keep an auxiliary diffusion model fitted to the student's evolving distribution at extra memory and computation cost. We introduce DMAD, Distribution Matching as Adversarial Distillation, which recasts distribution matching as classification and learns the required log-density ratios directly. Two discriminator heads on a shared backbone distinguish real data and teacher samples from the student's, and linear losses on their logits train the student without auxiliary score fitting. We prove that at the discriminator optimum these losses recover the distribution-matching gradient underlying DMD, through the classical identity linking discriminator logits to log-density ratios. We further introduce gap-based reweighting, which adapts teacher supervision across noise levels from the real-data head's empirical logit gap between real and teacher samples. DMAD reaches a Fr\'echet Inception Distance (FID) of 1.04 with one-step generation on ImageNet-64x64, 14.47 with four-step SDXL on COCO-10K, and a VBench total score of 85.15 with four-step Wan2.1-T2V-14B, the best values among the compared few-step methods and the multi-step teachers. On MiniMax-H3-33B, our four-step student achieves overall human preference rates of 79.1% over DMD2 and 84.6% over rCM for joint audio-video generation, excluding ties. Our code, models and demos are available at https://yzmblog.github.io/projects/DMAD.
Chinese Translation
分布匹配蒸馏（DMD）通过分别估计的目标和学生分数之间的差异来训练一个少步学生，因此必须保持一个辅助扩散模型，以适应学生不断演变的分布，这会带来额外的内存和计算成本。我们提出了DMAD，即作为对抗蒸馏的分布匹配，它将分布匹配重新表述为分类，并直接学习所需的对数密度比。两个共享主干的判别器头区分真实数据和教师样本与学生样本，并且对它们的logits施加线性损失，从而在没有辅助分数拟合的情况下训练学生。我们证明在判别器最优时，这些损失恢复了DMD背后的分布匹配梯度，通过经典的身份将判别器logits与对数密度比联系起来。我们进一步引入基于差距的重加权，它根据真实数据头的真实样本与教师样本之间的经验logit差距，调整教师监督在不同噪声水平下的适应性。DMAD在ImageNet-64x64上实现了一步生成的Fréchet Inception Distance (FID) 为1.04，在COCO-10K上使用四步SDXL实现14.47，在四步Wan2.1-T2V-14B上获得VBench总分85.15，这些都是与比较的少步方法和多步教师相比的最佳值。在MiniMax-H3-33B上，我们的四步学生在联合音频-视频生成中实现了79.1%的人类偏好率，相对于DMD2和84.6%相对于rCM，排除平局。我们的代码、模型和演示可在https://yzmblog.github.io/projects/DMAD获取。
cs.CV / 168 / 2610.02197
HiPhy: Hierarchical Alignment for Physically-Plausible Multi-Principle Video Generation
HiPhy：用于物理合理的多原则视频生成的层次对齐
Abstract
Video generation models have achieved remarkable visual fidelity and have strong potential to become general-purpose world simulators. Despite this progress, they still fail to generate videos which adhere to laws of physics. The problem becomes even more apparent in realistic settings where multiple physical principles must work together within the same video; for example, "a balloon floating upward while steam rises from a pot" requires buoyancy and fluid dynamics to unfold coherently and simultaneously. Yet existing methods largely ignore multi-principle interactions, focusing on a single principle per video. We propose HiPhy (Hierarchical Physical Alignment), a reinforcement learning framework that grounds video generation in physical laws through a dual-level objective: locally enforcing the temporal dynamics of individual physical principles, and globally ensuring the physical and semantic coherence of the entire scene. To support multi-principle generation, we construct a 50K-prompt dataset and introduce a prompt benchmark MultiPhyBench, spanning a diverse range of co-occurring physical events. Our experiments show that HiPhy significantly outperforms prior methods and baselines, improving physical commonsense and semantic alignment significantly across various benchmarks, with the largest gains on scenes involving multiple concurrent physical principles where competing methods degrade most sharply.
Chinese Translation
视频生成模型已经实现了显著的视觉逼真度，并具有成为通用世界模拟器的强大潜力。尽管取得了这些进展，它们仍然无法生成遵循物理法则的视频。在现实场景中，这一问题变得更加明显，因为多个物理原则必须在同一视频中协同工作；例如，“一个气球向上漂浮，同时蒸汽从锅中升起”需要浮力和流体动力学的协调和同时展开。然而，现有方法在很大程度上忽视了多原则的相互作用，专注于每个视频中的单一原则。我们提出了HiPhy（层次物理对齐），这是一个通过双层目标将视频生成与物理法则相结合的强化学习框架：在局部上强制执行单个物理原则的时间动态，并在全局上确保整个场景的物理和语义一致性。为了支持多原则生成，我们构建了一个包含5万条提示的数据集，并引入了一个提示基准MultiPhyBench，涵盖了多种共存的物理事件。我们的实验表明，HiPhy显著优于先前的方法和基准，在各种基准测试中显著提高了物理常识和语义对齐，尤其是在涉及多个并发物理原则的场景中，竞争方法的性能下降最为明显。
cs.CV / 169 / 2610.02201
SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation
SILSA：用于拓扑保持的高分辨率 3D 生成的滑动窗口切片潜变量
Abstract
High-resolution 3D generation increasingly relies on voxel latents and multi-stage pipelines that first predict active structure and then synthesize local geometry. While effective, this design fragments continuous surfaces into many local tokens, inflates generation cost, and often weakens topological consistency for thin or highly connected shapes. We introduce SILSA, a topology-aware 3D generation framework that represents shapes with compact sliding-window slice latents. Instead of generating expensive voxel tokens, SILSA uses a fixed set of overlapping slices along the three canonical axes, where each token summarizes a local depth window to preserve cross-sectional continuity and support single-stage rectified-flow generation. A Slice VAE encodes oriented surface samples into multi-axis slice latents and reconstructs them with a sparse volumetric decoder, while a Volumetric Anchor Lattice coordinates directional slice streams through a shared 3D workspace. To preserve structural correctness, we introduce slice-level topology supervision that matches persistence diagrams and aligns Betti transitions across neighboring slices. Experiments show that SILSA improves structural fidelity while substantially reducing generation cost. SILSA improves PSNR by $8.7\%$, coverage by $5.96$ absolute points, and Betti error by $9.2\%$ over the strongest baseline, while using $70.0\%$ fewer tokens than the next-most compact baseline and over $98\%$ fewer tokens than sparse or hierarchical tokenizers, effectively reducing training memory by $40.4\%$ and inference time by $58.5\%$. Qualitative results further show improved preservation of thin structures, repeated components, and long-range connectivity.
Chinese Translation
高分辨率 3D 生成越来越依赖于体素潜变量和多阶段管道，这些管道首先预测活动结构，然后合成局部几何形状。尽管这种设计有效，但它将连续表面分割为许多局部标记，增加了生成成本，并且通常削弱了薄形状或高度连接形状的拓扑一致性。我们提出了 SILSA，一个拓扑感知的 3D 生成框架，通过紧凑的滑动窗口切片潜变量来表示形状。SILSA 不生成昂贵的体素标记，而是沿三个标准轴使用一组固定的重叠切片，其中每个标记总结了一个局部深度窗口，以保持横截面连续性并支持单阶段的校正流生成。Slice VAE 将定向表面样本编码为多轴切片潜变量，并通过稀疏体积解码器重建它们，而体积锚格则通过共享的 3D 工作空间协调方向切片流。为了保持结构的正确性，我们引入了切片级拓扑监督，匹配持久性图并对齐相邻切片之间的 Betti 转换。实验表明，SILSA 提高了结构保真度，同时显著降低了生成成本。与最强基线相比，SILSA 将 PSNR 提高了 $8.7\%$，覆盖率提高了 $5.96$ 个绝对点，Betti 误差降低了 $9.2\\%$，同时使用的标记比下一个最紧凑的基线少 $70.0\\%$，比稀疏或分层标记器少超过 $98\\%$，有效减少了训练内存 $40.4\\%$ 和推理时间 $58.5\\%$。定性结果进一步表明，薄结构、重复组件和长距离连接的保留得到了改善。
cs.CV / 170 / 2610.02203
Embedding Prediction Helps Image Generation
嵌入预测助力图像生成
Abstract
In diffusion transformers, a class label or a text prompt is embedded once, and the same condition is reused at every denoising step. We ask whether predicted embeddings can serve as this condition instead. Next-Embedding Predictive Autoregression (NEPA) trains a Transformer to predict the next continuous embedding in a sequence. In generation, the clean image follows the noisy image, so its embeddings are the next embeddings after the condition and the noisy image. We train a NEPA model to predict them all at once with Multi-Embedding Prediction, and in Embedding Conditioned Generation, a DiT generator is conditioned on these predictions, recomputed at every denoising step, so the conditioning signal adapts to the current noisy state. Experiments on class-conditional ImageNet $256\times256$ study the condition of the generator, the design of Multi-Embedding Prediction, and the scaling of both models. The NEPA model adds a second network to every sampling step; with it, and combined with REPA, our final model, NEPA-DiT-XL, reaches an FID of 1.32 using about a third of the training compute of REPA.
Chinese Translation
在扩散变换器中，类别标签或文本提示被嵌入一次，并在每个去噪步骤中重复使用相同的条件。我们探讨预测的嵌入是否可以作为这个条件。下一嵌入预测自回归（Next-Embedding Predictive Autoregression, NEPA）训练一个变换器来预测序列中的下一个连续嵌入。在生成过程中，干净的图像跟随噪声图像，因此其嵌入是条件和噪声图像之后的下一个嵌入。我们训练一个 NEPA 模型，通过多嵌入预测（Multi-Embedding Prediction）一次性预测所有嵌入，并在嵌入条件生成（Embedding Conditioned Generation）中，DiT 生成器基于这些预测进行条件化，这些预测在每个去噪步骤中重新计算，因此条件信号适应当前的噪声状态。在类别条件的 ImageNet $256 imes256$ 上进行的实验研究了生成器的条件、多嵌入预测的设计以及两个模型的扩展。NEPA 模型在每个采样步骤中增加了第二个网络；通过它，并结合 REPA，我们的最终模型 NEPA-DiT-XL 在使用大约三分之一的 REPA 训练计算的情况下达到了 1.32 的 FID。
cs.CV / 171 / 2610.02205
ROWBench: Do Video Models Render What the Program Specifies?
ROWBench：视频模型是否呈现程序所指定的内容？
Abstract
Programmable world models separate executable dynamics from visual generation, offering a promising foundation for next-generation game engines. However, their visual adherence to explicit rules and interactions remains insufficiently evaluated. Existing benchmarks assess visual quality, controllability, and instruction or physical adherence, but rarely test fidelity to fine-grained, program-specified world events. We introduce PROWBench, comprising 170 programmatically constructed episodes and 600 proxy videos covering diverse scenes and interactions. PROWBench logs entity states and timestamped events, including those outside the camera's field of view, as replayable world records, from which it renders synchronized views and proxy representations. This enables generated videos to be checked against the observable consequences of program execution. An extensible framework constructs scenes, controls behaviors, and can render each camera view in different representations, such as coarse 3D, and bounding boxes. The benchmark covers first- and third-person perspectives, with synchronized multi-view observations available for a subset of episodes. Grounded in these records, PROWBench evaluates entity control, long-horizon memory, and, with two VLM-based metrics, Logic-Render Alignment and Interaction Success Rate, adherence to the prescribed timeline and the visual realization of timestamped engine-recorded events.
Chinese Translation
可编程世界模型将可执行动态与视觉生成分离，为下一代游戏引擎提供了有希望的基础。然而，它们在视觉上对明确规则和交互的遵循程度仍然评估不足。现有基准测试评估视觉质量、可控性以及指令或物理遵循性，但很少测试对细粒度程序指定的世界事件的保真度。我们引入了PROWBench，包含170个程序构建的剧集和600个代理视频，涵盖多样的场景和交互。PROWBench记录实体状态和时间戳事件，包括那些在摄像机视野之外的事件，作为可重放的世界记录，从中渲染同步视图和代理表示。这使得生成的视频可以与程序执行的可观察结果进行对比。一个可扩展的框架构建场景，控制行为，并可以以不同的表示方式渲染每个摄像机视图，例如粗略的3D和边界框。基准测试涵盖第一人称和第三人称视角，并为部分剧集提供同步的多视角观察。基于这些记录，PROWBench评估实体控制、长时间记忆，并通过两个基于视觉语言模型（VLM）的指标，即逻辑-渲染一致性（Logic-Render Alignment）和交互成功率（Interaction Success Rate），评估对规定时间线的遵循以及时间戳引擎记录事件的视觉实现。
cs.CV / 172 / 2610.02207
One Basis to Animate Them All: Gaussian Blendshape Distillation for Real-Time Avatars
统一基础以驱动所有动画：用于实时头像的高斯混合形状蒸馏
Abstract
3D Gaussian avatars support fast rendering, however, their real-time animation is often challenged by the costly neural inference. We address this bottleneck and show that the animation of pretrained avatar models can be closely approximated by a linear combination of identity-independent blendshapes. Building on this finding, we introduce GALA (Gaussian Animation via Linear Approximation), a distillation method that replaces per-frame heavy neural decoding with a shallow coefficient predictor and a linear blend. To improve fidelity and reduce memory requirements, we propose to construct the basis using block-local PCA under a rendering-aware metric and a memory budget. Our method learns a shallow MLP network to predict blendshape coefficients and applies to various animation architectures without retraining original models. We validate GALA by accelerating the inference of three distinct avatar models for 3D animation of facial expressions and full-bodies with clothing dynamics. Across these models, our distillation generalizes to held-out identities and reduces CPU animation cost by up to three orders of magnitude while preserving most of the rendering quality. Excellent results of our method confirm the shared linear structure of learned avatar representations and enable highly efficient and accurate animation at frame rates reaching up to 60fps on mobile devices. Project page: https://ramazan793.github.io/gala/
Chinese Translation
3D高斯头像支持快速渲染，但其实时动画常常受到昂贵的神经推理的挑战。我们解决了这一瓶颈，并展示了预训练头像模型的动画可以通过身份无关的混合形状的线性组合来近似。基于这一发现，我们引入了GALA（Gaussian Animation via Linear Approximation），这是一种蒸馏方法，它用一个浅层系数预测器和线性混合替代了每帧重的神经解码。为了提高保真度并减少内存需求，我们提出在渲染感知度量和内存预算下使用块局部主成分分析（PCA）来构建基础。我们的方法学习一个浅层多层感知器（MLP）网络来预测混合形状系数，并适用于各种动画架构，而无需重新训练原始模型。我们通过加速三种不同头像模型的推理来验证GALA，这些模型用于3D动画的面部表情和全身服装动态。在这些模型中，我们的蒸馏方法对保留身份的样本具有良好的泛化能力，并将CPU动画成本降低了三个数量级，同时保持了大部分渲染质量。我们方法的优异结果证实了学习到的头像表示的共享线性结构，并使得在移动设备上以高达60帧每秒的帧率实现高效且准确的动画成为可能。项目页面：https://ramazan793.github.io/gala/
cs.CV / 173 / 2610.02208
Sphere Encoder 2
球体编码器 2
Abstract
Sphere Encoder is an autoencoder that generates images by decoding random points from a high-dimensional latent sphere. We identify two limitations of the original formulation that reduce its generation quality. First, random points concentrate near the equator relative to the pole on an encoded latent, but the training rotation never reaches this region, leaving a gap that limits one-step generation. Second, training for generation with pixel-wise reconstruction loss encourages the decoder to average over plausible images, producing blurry images that lack high-frequency details. We present Sphere Encoder 2 to address both limitations, substantially improving image generation quality while maintaining the speed and simplicity of a autoencoder. Models are released at \href{https://github.com/kaiyuyue/sphere2}{github.com/kaiyuyue/sphere2}.
Chinese Translation
球体编码器是一种自编码器，通过解码来自高维潜在球体的随机点生成图像。我们识别出原始公式的两个局限性，这些局限性降低了其生成质量。首先，相对于极点，随机点在编码潜在空间中倾向于集中在赤道附近，但训练旋转从未到达该区域，留下了一个限制一步生成的空白。其次，使用像素级重构损失进行生成训练会促使解码器对可行图像进行平均，从而生成缺乏高频细节的模糊图像。我们提出了球体编码器 2，以解决这两个局限性，显著提高图像生成质量，同时保持自编码器的速度和简洁性。模型已在 exttt{https://github.com/kaiyuyue/sphere2} 发布。
cs.CV / 174 / 2610.02210
Moore, Escher, Penrose: A Conformal Golden Braid
摩尔、艾舍尔、彭罗斯：一个共形的黄金辫子
Abstract
I don't think I have ever done anything as peculiar in my life. Among other things, it shows a young man looking with interest at a print on the wall of an exhibition that features himself. How can this be? Perhaps I am not far removed from Einstein's curved universe.'' So wrote M.C. Escher about his 1956 lithograph Print Gallery. Nearly half a century later, a mathematical analysis related its geometry to an untwisted source image through a conformal power map $z \mapsto z^\alpha$, $\alpha \in \mathbb{C}$. Building on this construction, we use a frozen text-to-image diffusion model to generate new self-referential scenes. Prompting alone does not enforce the recursion, while a post-hoc transformation can leave structures poorly connected. Applying the transformation during sampling is also insufficient: the denoiser may "repair" the intended distortion or drift out of the prescribed geometry. We construct a generalized inverse $T^\dagger$ of the non-invertible image transformation $T$, adapted to its recursive constraint. In the idealized formulation, the Penrose identity $TT^\dagger T = T$ makes $TT^\dagger$ an idempotent projection onto geometrically admissible images. Yet denoising only the transformed image remains an out-of-distribution task, even with projection. We therefore braid denoising steps with $T$ and $T^\dagger$: source-space steps develop the untwisted scene, while transformed-space steps refine its appearance and connections in the final geometry. We generate Print Gallery-like compositions and explore further transformations. Rather than distorting a finished image, we let the scene and its distortion develop together.
Chinese Translation
我想我一生中从未做过如此奇特的事情。在其他方面，它展示了一位年轻人对展览墙上自己肖像的印刷品表现出兴趣。这怎么可能呢？也许我离爱因斯坦的曲率宇宙并不遥远。”这是M.C.艾舍尔对他1956年石版画《印刷画廊》的描述。近半个世纪后，一项数学分析将其几何形状与通过共形幂映射 $z o z^eta$ ($eta ext{ in } ext{C}$) 的未扭曲源图像联系起来。在此基础上，我们使用冻结的文本到图像扩散模型生成新的自指场景。仅仅依靠提示并不能强制执行递归，而事后变换可能会导致结构连接不良。在采样过程中应用变换也不足够：去噪器可能会“修复”预期的失真或偏离规定的几何形状。我们构造了非可逆图像变换 $T$ 的广义逆 $T^ ext{dagger}$，以适应其递归约束。在理想化的公式中，彭罗斯恒等式 $TT^ ext{dagger} T = T$ 使得 $TT^ ext{dagger}$ 成为对几何可接受图像的幂等投影。然而，仅对变换后的图像进行去噪仍然是一个超出分布的任务，即使有投影。因此，我们将去噪步骤与 $T$ 和 $T^ ext{dagger}$ 编织在一起：源空间步骤发展未扭曲的场景，而变换空间步骤则在最终几何中细化其外观和连接。我们生成类似《印刷画廊》的作品，并探索进一步的变换。我们不是扭曲一幅完成的图像，而是让场景及其失真共同发展。
cs.AI / 1 / 2610.00010
Heavy-Tailed Memory Traces in Long-Horizon Language Agents
长时间视野语言代理中的重尾记忆痕迹
Abstract
Long-horizon language agents increasingly rely on external memory as a frozen world model, yet current memory systems are usually judged only by task success or token cost. We argue that the missing object is the shape of memory use: under finite context and repeated retrieval, agent memory can concentrate on a small core while leaving rare states in a long tail where prediction errors accumulate. We study this effect through a conservative tail audit and find that concentration is reproducible but policy-dependent. Random-walk agents produce log-normal-compatible retrieval artifacts, whereas semantic LLM policies yield the strongest truncated-power-law-compatible core--tail traces. Motivated by this audit, we propose Core--Tail World Model (CTWM), a rank-based memory controller that allocates prompt budget with a single exponent $\tau$ while retaining a summarized tail. On Synthetic Graph World, CTWM preserves full state and transition coverage, reduces prompt tokens by 5.9%, and lowers bottom-half tail prediction error by 13.6% relative to a graph-memory baseline. The same paired comparison gives consistent token savings on ALFWorld and a 24.48% token reduction on LongMemEval with aggregate accuracy parity. These results suggest that heavy-tailed memory traces are not only a diagnostic of finite retrieval, but also a practical control signal for token-efficient agent world models.
Chinese Translation
长时间视野语言代理越来越依赖外部记忆作为冻结的世界模型，但当前的记忆系统通常仅通过任务成功率或令牌成本来评估。我们认为缺失的关键在于记忆使用的形态：在有限上下文和重复检索的情况下，代理的记忆可以集中在一个小核心上，同时将稀有状态留在一个长尾中，导致预测误差的积累。我们通过保守的尾部审计研究这一效应，发现集中性是可重复的，但依赖于策略。随机游走代理产生与对数正态分布兼容的检索伪影，而语义大型语言模型（LLM）策略则产生最强的截断幂律兼容核心-尾部痕迹。基于这一审计，我们提出了核心-尾部世界模型（Core--Tail World Model, CTWM），这是一种基于排名的记忆控制器，使用单一指数 $ au$ 分配提示预算，同时保留一个总结的尾部。在合成图世界（Synthetic Graph World）中，CTWM 保持了完整的状态和转移覆盖，减少了 5.9% 的提示令牌，并相对于图记忆基线降低了 13.6% 的下半部分尾部预测误差。同样的配对比较在 ALFWorld 上也提供了一致的令牌节省，并在 LongMemEval 上实现了 24.48% 的令牌减少，同时保持了整体准确性。这些结果表明，重尾记忆痕迹不仅是有限检索的诊断指标，也是令牌高效代理世界模型的实用控制信号。
cs.AI / 2 / 2610.00012
When Do Causal World Models Help Modular LLM Agents
因果世界模型何时能帮助模块化大型语言模型代理
Abstract
LLM agents increasingly act through modular systems, such as order, payment, inventory, and shipment services, where actions in one module change which transitions are valid in another. Standard world models usually fit observational traces, but this is not the quantity needed for intervention-time planning: a trace may show that payment precedes shipment without identifying whether payment authorizes shipment, inventory mediates the effect, or a hidden trigger explains both. We study this gap through FedCausalCompose, a causal world-model framework for modular LLM agents in which local actions provide intervention-response evidence for cross-module interfaces. We first show that observational world models incur an irreducible interventional error under unblocked back-door paths, that interface recovery improves with intervention-response coverage, and that an oracle causal composition can beat the non-causal lower bound when coverage and local mechanism errors are controlled. We then test the resulting prediction in diagnostic agent settings. Causal interfaces help most in structured tool environments, where API signatures expose preconditions and downstream effects. In contrast, dialogue and narrative environments often ignore raw edge lists unless a short attention anchor makes the causal information decision-relevant. These results identify a concrete condition for causal world models in LLM agents: causal structure helps when cross-module interfaces are both statistically identifiable and presented in a form the agent can use at action time.
Chinese Translation
大型语言模型（LLM）代理越来越多地通过模块化系统进行操作，例如订单、支付、库存和运输服务，其中一个模块的操作会改变另一个模块中有效的转换。标准的世界模型通常适用于观察性轨迹，但这并不是干预时间规划所需的量：一条轨迹可能显示支付在运输之前，但并未确定支付是否授权运输、库存是否介导了这一效果，或是一个隐藏的触发因素解释了两者之间的关系。我们通过FedCausalCompose研究这一差距，这是一种针对模块化LLM代理的因果世界模型框架，其中局部操作提供跨模块接口的干预响应证据。我们首先表明，在未被阻塞的后门路径下，观察性世界模型会产生不可减少的干预误差，接口恢复随着干预响应覆盖的提高而改善，并且在控制覆盖和局部机制误差时，预言性因果组合可以超越非因果下界。然后，我们在诊断代理设置中测试了结果预测。因果接口在结构化工具环境中最为有效，在这些环境中，API签名暴露了前提条件和下游效果。相反，对话和叙事环境通常忽视原始边缘列表，除非短期注意锚点使因果信息与决策相关。这些结果确定了LLM代理中因果世界模型的一个具体条件：当跨模块接口在统计上可识别且以代理在行动时可以使用的形式呈现时，因果结构能够提供帮助。
cs.AI / 3 / 2610.00015
From Proposal to Verified Effect: Praxa, an Evidence-Bound Harness for Governed AI Agent Execution
从提议到验证效果：Praxa，一个基于证据的治理AI代理执行框架
Abstract
Large-language-model agents can propose and execute actions, but proposal, authority, dispatch, verified external effect, and serving promotion are different claims. We present Praxa, an agent harness that represents these states explicitly through deterministic admission, brokered execution, external read-back, reconciliation, and reviewed promotion. We report four evidence lanes. First, an author-run repository-local audit at a pinned revision passed 1,027/1,027 unit tests and 89/89 Workerd tests, instrumented all 363 expected source files, and met four coverage floors; raw per-test transcripts and independent reproduction are unavailable. Second, in a provider-backed Terminal-Bench Core 0.1.1 pilot across 12 curated tasks, baseline and reliability-layer arms each passed 17/36 strict trials. The reliability layer used 37.49% more input and 50.73% more output tokens, so the pilot does not support superiority. Third, in a post-debug, two-order coordination-proxy development comparison, baseline and a source-authored candidate each completed 180/180 trials with equal measured accuracy, full hermetic crash recovery, and zero protected violations. The candidate used 37.11% fewer tokens, 33.84% lower estimated endpoint cost, and 11.63% fewer steps; this does not establish improved quality, latency, or production behavior. Fourth, deployed source/configuration evidence shows bounded reflection, recall accounting, memory compilation, and tool-health paths, but no production outcome lift. Praxa's supported contribution is an evidence-bound architecture that makes authority-to-effect transitions explicit and testable. Current evidence does not establish adversarial security, production safety, general specialist superiority, autonomous recursive optimization, or user benefit.
Chinese Translation
大型语言模型代理可以提出并执行行动，但提议、权威、调度、验证的外部效果和服务推广是不同的主张。我们提出了Praxa，一个代理框架，通过确定性的准入、经纪执行、外部回读、调和和审核推广明确表示这些状态。我们报告了四条证据通道。首先，在一个固定修订版的作者运行的本地审计中，1,027/1,027个单元测试和89/89个Workerd测试均通过，所有363个预期源文件均被仪器化，并满足四个覆盖底线；原始每个测试的成绩单和独立复现不可用。其次，在一个提供者支持的Terminal-Bench Core 0.1.1试点中，涵盖12个策划任务，基线和可靠性层各自通过了17/36个严格试验。可靠性层使用了37.49%的更多输入和50.73%的更多输出标记，因此该试点不支持优越性。第三，在一个后调试的二阶协调代理开发比较中，基线和一个源作者候选者均完成了180/180个试验，测量准确度相等，完全的密封崩溃恢复和零保护违规。候选者使用了37.11%更少的标记，估计的端点成本降低了33.84%，步骤减少了11.63%；这并未建立改进的质量、延迟或生产行为。第四，已部署的源/配置证据显示出有限的反思、回忆会计、内存编译和工具健康路径，但没有生产结果提升。Praxa所支持的贡献是一个基于证据的架构，使权威到效果的转变明确且可测试。目前的证据未能建立对抗性安全、生产安全、一般专家优越性、自主递归优化或用户利益。
cs.AI / 4 / 2610.00018
What Do Rationales Communicate? A Message-Intervention Study in Role-Specialized QA
理据传达了什么？角色专用问答中的信息干预研究
Abstract
Role-specialized QA pipelines increasingly pass rationales from a reasoner to a verifier, but it is unclear what this message actually buys: better answers, stronger support assessment, or a new failure surface. We introduce a message-intervention diagnostic that fixes the evidence and candidate answer while varying only the rationale passed across the reasoner-to-verifier boundary. On 400 MuSiQue, HotpotQA, and 2WikiMultiHopQA examples with DeepSeek as generator and verifier, faithful rationales add almost no answer accuracy over no rationale, while corrupted rationales strongly alter support judgments. Under a blind verifier prompt, harmless paraphrases shift support by only 0--2.5%, whereas corrupted rationales shift support by 10--22%; an explicit rationale-checking prompt amplifies the same pattern to 34--55%. Final answers move less (2--30%), and only 2.9--35.3% of corrupted support flips co-occur with answer changes. Human audits show why this matters: 16/42 valid corruptions are corruption-overtrust cases, and blind humans reject or mark unclear 9/10 audited corrupted rationales that the model accepts. Cross-model and task-boundary checks show when the channel is active, amplified, inert, or folded into the task label. Rationale sharing should be evaluated as a verification-message mechanism, not merely as a route to higher answer accuracy.
Chinese Translation
角色专用问答（QA）管道越来越多地将理据从推理者传递给验证者，但目前尚不清楚这一信息实际带来了什么：更好的答案、更强的支持评估，还是新的失败表面。我们引入了一种信息干预诊断方法，该方法固定证据和候选答案，同时仅改变跨越推理者与验证者边界的理据。在使用 DeepSeek 作为生成器和验证者的 400 个 MuSiQue、HotpotQA 和 2WikiMultiHopQA 示例中，忠实的理据几乎没有提高答案的准确性，而损坏的理据则显著改变了支持判断。在盲验证者提示下，无害的释义仅使支持变化 0-2.5%，而损坏的理据使支持变化 10-22%；一个明确的理据检查提示将相同的模式放大到 34-55%。最终答案的变化较小（2-30%），而且只有 2.9-35.3% 的损坏支持翻转与答案变化同时发生。人工审核显示了这一点的重要性：42 个有效损坏中有 16 个是过度信任的损坏案例，盲人类拒绝或标记 10 个审核的损坏理据中有 9 个不清晰，而模型则接受这些理据。跨模型和任务边界的检查显示了通道何时处于活动、放大、惰性或融入任务标签的状态。理据共享应作为验证信息机制进行评估，而不仅仅是提高答案准确性的途径。
cs.AI / 5 / 2610.00025
Measuring the Microtask Eligibility Gap: When Is an Off-the-Shelf SLM Enough for an Agent Harness?
测量微任务资格差距：何时现成的小型语言模型足以满足代理需求？
Abstract
Agent harnesses increasingly want to run small language models (SLMs) on the microtasks around a frontier large language model (LLM) planner: auto-approving shell commands, writing memory, selecting tools, ranking past turns. We ask whether off-the-shelf SLMs meet practitioner-defined thresholds and, when they fail, why, and whether quantization changes the answer. We build a benchmark of 4 such microtasks with fixed prompts and automatic metrics, each with a pre-specified threshold $\tau$ anchored to a cheap non-LLM baseline and a CI-aware eligibility rule (a configuration passes only if its confidence bound clears $\tau$). Sweeping Qwen3 0.6/1.7/4/8B at their best (FP16, greedy, one frozen prompt, no tuning), we find an eligibility gap: 0 of 16 (4 tasks $\times$ 4 models) configurations pass (verified by checking the raw outputs and parser behavior). A logprob decision-threshold diagnostic (T1/T3/T4; T2 via a context-length/cascade probe) separates the failures into capability deficits and failures that can be addressed by changing the decoding threshold (4 regimes). Quantization to 4-bit (RTN/GPTQ/AWQ) does damage that depends on model size and moves no configuration into eligibility (certified on the reconstructable hard-label tasks T1/T3, diagnostic/windowed robustness on T2/T4), so the gap tracks model size more than precision; it replicates on Llama-3.x (12/12 ineligible) and is robust to the anchor choice (a $\tau$-sweep) and to prompt wording (0/112 eligible across the original plus 3 neutral paraphrases per cell). The practical implication: place SLMs behind a baseline that meets the CI-backed threshold, and use the SLM only where the baseline fails to meet the threshold; e.g. a 4B re-ranker over a BM25 shortlist beats BM25 ($+0.047$ [0.020, 0.073], without itself certifying eligibility).
Chinese Translation
代理需求日益希望在前沿大型语言模型（LLM）规划器周围运行小型语言模型（SLM）以处理微任务：自动批准 shell 命令、编写记忆、选择工具、排名过去的对话。我们探讨现成的 SLM 是否满足实践者定义的阈值，并在其未能满足时，分析原因，以及量化是否会改变这一结果。我们构建了一个包含 4 个微任务的基准，每个任务都有固定的提示和自动化指标，每个任务都有一个预设的阈值 $ au$，该阈值基于一个廉价的非 LLM 基线和一个考虑置信区间的资格规则（配置仅在其置信区间超出 $ au$ 时通过）。在最佳配置下（FP16、贪婪、一个冻结提示、无调优）对 Qwen3 0.6/1.7/4/8B 进行测试，我们发现存在资格差距：16 种配置中没有一种通过（通过检查原始输出和解析器行为进行验证）。一个基于对数概率的决策阈值诊断（T1/T3/T4；T2 通过上下文长度/级联探测）将失败分为能力不足和可以通过改变解码阈值解决的失败（4 种模式）。量化到 4 位（RTN/GPTQ/AWQ）造成的损害依赖于模型大小，并且没有任何配置进入资格（在可重构的硬标签任务 T1/T3 上经过认证，T2/T4 上的诊断/窗口鲁棒性），因此资格差距更多地与模型大小相关，而非精度；这一现象在 Llama-3.x 上得到了复制（12/12 不合格），并且对锚点选择（$ au$ 扫描）和提示措辞（在原始提示及每个单元的 3 个中性释义中，0/112 合格）具有鲁棒性。实际意义在于：将 SLM 放置在满足 CI 支持阈值的基线之后，仅在基线未能满足阈值时使用 SLM；例如，一个 4B 的重排序器在 BM25 的短名单上优于 BM25（$+0.047$ [0.020, 0.073]，且自身未能认证资格）。
cs.AI / 6 / 2610.00047
Characterizing a Configuration Where Inference-Time PRM-Pruned Fragment Grafting Is Inert: Evidence from Three Reasoning LMs
表征推理时 PRM 剪枝片段嫁接无效的配置：来自三个推理语言模型的证据
Abstract
Diversity collapse in parallel chain-of-thought has motivated inference-time interventions built on a natural design: when a process reward model (PRM) prunes a chain, its high-PRM prefix is extracted and grafted verbatim as an in-context demonstration into a still-decoding sibling. We isolate this mechanism, PRM-Pruned Fragment Grafting (PPFG), as the most cost-minimal operationalization of cross-trajectory step-level transfer, and test it at the operating point where prior fragment-grafting work reports gains only under additional compensating ingredients. On Qwen2.5-7B-Instruct with Math-Shepherd on full MATH500 (n=500, three seeds), PPFG in both stagnation- and random-targeting variants is statistically indistinguishable from an independent parallel-CoT baseline on every measured axis. We characterize why: a four-bucket classification of 322 stagnation-rule injection events shows only 14% targeted a genuinely struggling chain; the rest landed on chains that had already succeeded, were near completion, or sat on a flat PRM plateau, states a rescue graft cannot change. No compound-gate refinement jointly achieves well-targeted firing and adequate density, and a random control matches the same parity at 2.4x the firing rate, so the inertness is not heuristic-specific. The finding replicates across three base LMs, six benchmarks, a second PRM, and a compatibility-gate sweep; two-one-sided-tests analysis promotes the parity to positive equivalence on all twelve Qwen/LLaMA cells. A per-event spot-check finds injected chains prune at 2.75x the matched-step rate, but a surviving-sibling counterfactual finds no population-level compensation. A hindsight oracle bounds any per-problem gain from choosing PPFG over independent at +0.13 pp. We contribute an equivalence-testing template for establishing inference-time mechanism nulls, with every claim scoped to its tested operating point.
Chinese Translation
在并行思维链中的多样性崩溃促使了基于自然设计的推理时干预：当过程奖励模型（PRM）剪枝链时，其高 PRM 前缀被提取并逐字嫁接为上下文演示到仍在解码的兄弟链中。我们将这一机制——PRM 剪枝片段嫁接（PPFG）——孤立为跨轨迹步骤级转移的最具成本效益的操作化，并在先前片段嫁接工作报告仅在额外补偿成分下获得收益的操作点进行测试。在使用 Math-Shepherd 的 Qwen2.5-7B-Instruct 上的完整 MATH500（n=500，三个种子）中，PPFG 在停滞和随机目标变体下在每个测量维度上与独立的并行思维链基线在统计上无显著差异。我们描述了原因：对 322 个停滞规则注入事件的四类分类显示，只有 14% 目标是确实挣扎的链；其余的则落在已经成功、接近完成或处于平坦 PRM 高原的链上，这些状态是救援嫁接无法改变的。没有复合门精炼能够共同实现良好的目标触发和足够的密度，而随机控制在 2.4 倍的触发率下匹配了相同的平价，因此无效性并非特定于启发式。该发现跨越三个基础语言模型、六个基准、第二个 PRM 和兼容性门的测试得以复制；双侧检验分析将平价提升至所有十二个 Qwen/LLaMA 单元的正等价。每个事件的抽查发现，注入的链以 2.75 倍的匹配步骤率进行剪枝，但生存兄弟的反事实发现没有人口层面的补偿。事后预言者将选择 PPFG 相较于独立的每个问题收益界定为 +0.13 个百分点。我们贡献了一个等价性测试模板，用于建立推理时机制的无效性，每个声明都限于其测试的操作点。
cs.AI / 7 / 2610.00061
Gradient-Aligned Pair Selection for Personalized Preference Optimization
面向个性化偏好优化的梯度对齐配对选择
Abstract
Personalizing large language models (LLMs) requires aligning generation behavior with user-specific preferences rather than aggregate quality. While Direct Preference Optimization (DPO) provides a stable framework for preference learning, its effectiveness in personalized settings critically depends on how preference pairs are selected. Existing approaches typically rely on heuristic criteria, such as likelihood-based extremes, which decouple optimization from explicit user utility and can lead to degraded personalization. We formalize personalized preference learning as a geometry-aligned optimization problem by analyzing the first-order interaction between gradients of expected user utility and DPO update directions. Our analysis reveals that, under off-policy sampling, the DPO update transitions from a purely error-corrective signal to a reinforcement-like update when preference margins are directionally aligned with utility gradients. This perspective exposes pair selection as a geometric decision that governs whether preference optimization advances or hinders personalization. Motivated by this insight, we propose GAP-DPO (Geometry-Aligned Preference DPO), an iterative algorithm that performs utility-aware, geometry-aligned pair selection while controlling distribution shift via epoch-wise regeneration. Experiments on personalized text generation benchmarks show that GAP-DPO consistently improves stylistic fidelity, preference alignment, and generation quality compared to standard DPO variants. Together, our results establish gradient alignment as a unifying principle for personalized preference optimization and demonstrate that pair selection is an intrinsic component of the optimization geometry rather than a heuristic preprocessing step.
Chinese Translation
个性化大型语言模型（LLMs）需要将生成行为与用户特定的偏好对齐，而不是依赖于总体质量。虽然直接偏好优化（DPO）为偏好学习提供了一个稳定的框架，但其在个性化设置中的有效性在很大程度上依赖于偏好对的选择。现有的方法通常依赖于启发式标准，例如基于似然的极值，这使得优化与明确的用户效用脱钩，并可能导致个性化效果下降。我们通过分析期望用户效用的梯度与DPO更新方向之间的一阶交互，将个性化偏好学习形式化为一个几何对齐的优化问题。我们的分析揭示，在离线策略采样下，当偏好边际与效用梯度方向一致时，DPO更新从纯粹的误差修正信号转变为类似强化学习的更新。这一视角使得配对选择成为一个几何决策，决定偏好优化是促进还是阻碍个性化。基于这一见解，我们提出了GAP-DPO（几何对齐偏好DPO），这是一种迭代算法，能够在控制分布转移的同时，进行效用感知的几何对齐配对选择。针对个性化文本生成基准的实验表明，与标准DPO变体相比，GAP-DPO在风格忠实度、偏好对齐和生成质量上均表现出持续的改善。我们的结果共同确立了梯度对齐作为个性化偏好优化的统一原则，并证明配对选择是优化几何的内在组成部分，而非启发式预处理步骤。
cs.AI / 8 / 2610.00074
K-Dense BYOK: An Open-Source AI Research Assistant That Runs Locally and Keeps a Hash-Chained Lab Notebook
K-Dense BYOK：一个本地运行并保持哈希链实验笔记的开源AI研究助手
Abstract
K-Dense BYOK (bring your own keys) is a free, open-source AI research assistant for scientists in any field that runs on the researcher's own computer. The researcher supplies access to a model of their choice, hosted or running locally, and the application supplies everything else: a place for the work to run, a layer of scientific scaffolding, and a complete record. Each project is an ordinary folder, so the data, the code, the results, and the record stay on a machine the researcher administers and can be read years later without the application. Three things separate it from a chat assistant or a general-purpose coding agent. It ships a library of written scientific procedures, guided workflow templates, catalogs of where research data can be found, and reviewer and writer roles the agent can hand work to. It keeps a Living Lab Notebook whose entries link into an argument and are added to but never erased. And it records what happened by watching what the agent does rather than by taking the agent's word for it, in a log the agent has no tool that can write to. That choice targets the most common failure, model overclaiming, in our earlier benchmark of nine frontier models, by making claims checkable rather than preventing them. On twenty interdisciplinary research prompts, scored under a rubric fixed in advance, K-Dense BYOK led two managed platforms on both scientific quality and research execution. Its deliverables were the only ones that recorded the software they ran in, and the only ones that usually arrived with a command that regenerates the results. One of the managed platforms ran the same frontier model and supplied neither. Those environment records were files the agent wrote, not part of the observed log, which does not yet capture the software environment itself. The code is available under the MIT license at https://github.com/K-Dense-AI/k-dense-byok.
Chinese Translation
K-Dense BYOK（Bring Your Own Keys）是一个免费的开源AI研究助手，适用于任何领域的科学家，能够在研究者自己的计算机上运行。研究者提供所选择模型的访问权限，该模型可以托管或本地运行，而应用程序则提供其他所有功能：一个工作运行的场所、一层科学支架和完整的记录。每个项目都是一个普通的文件夹，因此数据、代码、结果和记录都保留在研究者管理的机器上，并且可以在多年后无需应用程序即可读取。它与聊天助手或通用编码代理的不同之处在于三个方面。首先，它提供了一套书面科学程序的库、指导工作流程的模板、研究数据来源的目录，以及代理可以分配工作的审稿人和作者角色。其次，它保持一个动态实验笔记本，其条目相互关联并不断增加，但从不删除。最后，它通过观察代理的行为来记录发生的事情，而不是依赖代理的口头陈述，记录在代理无法写入的日志中。这一选择针对我们早期对九个前沿模型的基准测试中最常见的失败——模型过度声称，通过使声称可检查而不是防止它们。在二十个跨学科研究提示下，K-Dense BYOK在科学质量和研究执行方面超越了两个管理平台。其交付成果是唯一记录其运行软件的成果，并且通常附带一个可以重新生成结果的命令。一个管理平台运行了相同的前沿模型，但没有提供这些功能。这些环境记录是代理写入的文件，而不是观察日志的一部分，后者尚未捕捉到软件环境本身。代码可在MIT许可证下获取，网址为 https://github.com/K-Dense-AI/k-dense-byok。
cs.AI / 9 / 2610.00084
Scientific Agents: Evaluating Profession-Specific System Prompts on Scientific Tasks
科学代理：评估专业特定系统提示在科学任务中的表现
Abstract
Detailed profession-specific system prompts raise token use and estimated cost per response without a consistent accuracy gain. We evaluate Scientific Agents, an open-source corpus of 503 profession-specific AGENTS.md profiles, with Gemini 3.8 Flash via OpenRouter in the Pi agent harness. We compare matched profiles with four controls: a minimal baseline ("You are a helpful assistant"), the profile's opening role sentence, a generic scientific rigor guide, and a profile from an unrelated domain. Across nine text-based science benchmarks (4,531 sampled questions, 100 matched profiles), 4,488 items completed all five conditions after API-error retries, scored with automated, rule-based grading. The average profile-baseline accuracy difference is -0.6 percentage points (95% bootstrap interval [-1.5, +0.2] across fixed tasks), and no benchmark shows a statistically clear improvement. Matched profiles produced 1.5-2.3 times as many output tokens and cost 2.2-4.5 times more per successful call. On 60 tool-using BioMysteryBench bioinformatics problems (three runs each for baseline and profile), mean solve rates were 46.7% with the profile and 56.7% at baseline, a difference of -10.0 percentage points (95% interval [-16.7, -3.3]) driven by more frequent token- and time-limit stops under the profile. Longer prompts had one unexpected operational advantage: on SuperGPQA, frequent provider API drops left the short baseline with a correct first-pass answer on only 54.0% of items, against 71.6% with the profile. Generic and mismatched prompts were about as reliable, so this gain comes from prompt length or formatting rather than domain expertise. For the tested model and tasks, loading full profession profiles by default does not improve accuracy and costs considerably more; whether selective retrieval of profile sections or open-ended scientific tasks would change this remains to be tested.
Chinese Translation
详细的专业特定系统提示提高了令牌使用量和每次响应的估计成本，但没有带来一致的准确性提升。我们评估了科学代理（Scientific Agents），这是一个包含503个专业特定AGENTS.md档案的开源语料库，通过OpenRouter中的Gemini 3.8 Flash在Pi代理框架下进行。我们将匹配的档案与四个对照组进行比较：一个最小基线（“你是一个有帮助的助手”）、档案的开头角色句子、一个通用的科学严谨性指南，以及一个来自无关领域的档案。在九个基于文本的科学基准测试中（4,531个抽样问题，100个匹配档案），在API错误重试后，4,488个项目完成了所有五个条件，并通过自动化的基于规则的评分进行评估。平均档案与基线的准确性差异为-0.6个百分点（95%自助法区间[-1.5, +0.2]，针对固定任务），且没有基准显示出统计上明显的改善。匹配的档案产生的输出令牌数量是基线的1.5-2.3倍，每次成功调用的成本是基线的2.2-4.5倍。在60个使用工具的BioMysteryBench生物信息学问题上（基线和档案各进行三次运行），档案的平均解决率为46.7%，基线为56.7%，差异为-10.0个百分点（95%区间[-16.7, -3.3]），主要是由于档案下更频繁的令牌和时间限制停止。较长的提示具有一个意外的操作优势：在SuperGPQA上，频繁的提供者API中断使得短基线仅在54.0%的项目上获得正确的首次回答，而档案则为71.6%。通用和不匹配的提示大致同样可靠，因此这一提升源于提示的长度或格式，而非领域专业知识。对于测试的模型和任务，默认加载完整的专业档案并未提高准确性，且成本显著更高；选择性检索档案部分或开放式科学任务是否会改变这一点仍需进一步测试。
cs.AI / 10 / 2610.00197
Comedic Fool's Gold: Reward Exploits and Countermeasures in Conversational Humor
喜剧的愚人金：对话幽默中的奖励利用与对策
Abstract
We investigate automated rewards for training language models in conversational humor, focusing on reward exploits and countermeasures. Two approaches aim to capture understandable surprise and predicted audience amusement. Controlled tests show that an embedding-based surprise reward accepts word-shuffled replies as readily as witty ones. A fluency filter detects the shuffles, but the combined reward also rejects some witty replies and fails further validation. An audience model's predicted laughter is instead vulnerable to laughter cues in either speaker's messages. Normalizing these cues across speakers blocks the covered attacks, although unmatched expressions remain exploitable. Three reinforcement-learning runs evaluate training with successive reward revisions. The final run improves the combined evaluation score by 0.0903 and reduces zero-score sessions by 40%, but its humor-specific improvement remains below our preregistered target. These findings illustrate a broader challenge for automated reward design: countermeasures must block exploitable shortcuts while preserving the behavior the reward was intended to encourage.
Chinese Translation
我们研究了用于训练语言模型的对话幽默的自动化奖励，重点关注奖励利用和对策。两种方法旨在捕捉可理解的惊讶和预测的观众娱乐感。控制测试表明，基于嵌入的惊讶奖励同样接受单词打乱的回复和机智的回复。流畅性过滤器能够检测到这些打乱，但结合奖励也会拒绝一些机智的回复，并未能通过进一步验证。观众模型预测的笑声则容易受到任一发言者信息中的笑声提示的影响。对这些提示在不同发言者之间进行归一化可以阻止已知攻击，尽管不匹配的表达仍然可以被利用。三次强化学习实验评估了通过连续奖励修订进行的训练。最终一次实验将综合评估分数提高了0.0903，并将零分会话减少了40%，但其幽默特定的改进仍低于我们预注册的目标。这些发现展示了自动化奖励设计面临的更广泛挑战：对策必须阻止可利用的捷径，同时保留奖励旨在鼓励的行为。
cs.AI / 11 / 2610.00212
EviGraph: Proof-Carrying Selective Recommendation over Temporal Public-Service Knowledge Graphs
EviGraph：基于证明的选择性推荐在时间公共服务知识图谱上的应用
Abstract
Public-service recommendations require evidence that matches the requested service, scope, and date. Yet treating every missing detail as decisive can withhold useful recommendations. We introduce EviGraph, which distinguishes critical decision requirements from information that can remain unresolved. A language agent links these requirements to evidence in a temporal knowledge graph, while a deterministic checker establishes whether a recommendation is supported. Evaluation on a bilingual Hong Kong public-service benchmark with executable policy references shows that this distinction reduces unnecessary abstention. Additional verification, however, can withdraw supported recommendations without improving decision quality. These findings suggest that reliable evidence-based navigation depends on specifying what must be established for a decision, rather than simply adding more verification.
Chinese Translation
公共服务推荐需要与请求的服务、范围和日期相匹配的证据。然而，将每一个缺失的细节视为决定性因素可能会抑制有用的推荐。我们提出了EviGraph，它区分了关键决策要求与可以保持未解决的信息。一个语言代理将这些要求与时间知识图谱中的证据连接起来，而一个确定性检查器则确定推荐是否得到了支持。在一个具有可执行政策参考的双语香港公共服务基准上的评估表明，这种区分减少了不必要的放弃。然而，额外的验证可能会撤回已支持的推荐，而不改善决策质量。这些发现表明，可靠的基于证据的导航依赖于明确决策所需建立的内容，而不仅仅是增加更多的验证。
cs.AI / 12 / 2610.00224
Build2SPARQL: A Large-Scale Text-to-SPARQL Benchmark Dataset for Building Knowledge Graph Querying
Build2SPARQL：一个用于建筑知识图谱查询的大规模文本到SPARQL基准数据集
Abstract
Building automation systems are increasingly represented as semantic knowledge graphs (KGs) using ontologies such as Brick and ASHRAE 223P, creating a machine-readable substrate for artificial-intelligence applications. One promising application is translating natural-language questions into SPARQL (text-to-SPARQL), which would let building operators query these graphs through language agents, but progress is limited by the scarcity of large natural-language/SPARQL benchmarks. This paper presents Build2SPARQL, a large-scale benchmark for building KGs generated by a KG-grounded pipeline: SPARQL queries are produced and validated entirely by graph-traversal code, while large language models generate only the natural-language questions, keeping query correctness independent of model behavior. The pipeline mines six query-pattern families -- linear chains, branching, UNION, aggregation, OPTIONAL, and attribute-filtered -- and phrases each query across five vocabulary registers. Applied to 201 building KGs (180 Brick, 21 ASHRAE 223P), it yields 6,136 executable SPARQL queries and 30,680 questions. A two-rater human validation of 300 questions found 98.8% semantic fidelity, 98.8% naturalness, and 84.0% operational plausibility. A retrieval-augmented evaluation across three open-weight language models raised exact-match accuracy from 0.2-20% (zero-shot) to 56-65% (three-shot retrieved).
Chinese Translation
建筑自动化系统越来越多地被表示为使用本体（如Brick和ASHRAE 223P）构建的语义知识图谱（KGs），为人工智能应用创建了一个机器可读的基础。一个有前景的应用是将自然语言问题翻译为SPARQL（文本到SPARQL），这将使建筑运营商能够通过语言代理查询这些图谱，但由于缺乏大型自然语言/SPARQL基准，进展受到限制。本文提出了Build2SPARQL，这是一个针对建筑知识图谱的大规模基准数据集，由一个基于知识图谱的管道生成：SPARQL查询完全由图遍历代码生成和验证，而大型语言模型仅生成自然语言问题，从而保持查询的正确性与模型行为无关。该管道挖掘了六种查询模式家族——线性链、分支、UNION、聚合、OPTIONAL和属性过滤——并在五个词汇注册中对每个查询进行短语化。应用于201个建筑知识图谱（180个Brick，21个ASHRAE 223P），生成了6,136个可执行的SPARQL查询和30,680个问题。对300个问题进行的双评估人类验证发现98.8%的语义保真度、98.8%的自然性和84.0%的操作合理性。在三个开放权重语言模型上的检索增强评估将精确匹配准确率从0.2-20%（零样本）提高到56-65%（三样本检索）。
cs.AI / 13 / 2610.00233
Robust Is Salient: An Informed Adversary Moves the Optimal Signal onto the Salience Pole
稳健即显著：知情对手将最优信号移动到显著性极点
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.
Chinese Translation
当一个知情对手与受限信号通道的受众共享时，最能保护真相的信号是最能描述真相的信号。在108个确认项目中，抗对手最优解与先前研究中的显著性极点完全一致。在一个包含200,000个项目的池中，二者仅在2,748个项目上存在差异——这些项目正好位于先前显著性与贝叶斯坐标未定义的地方。在定义的地方，通过完全从贝叶斯判别转向显著性来实现稳健性。我们通过在一个强制选择任务中引入对手来展示这一点（该任务抽象自《欺骗：香港谋杀案》）。对手知道目标，观察信号，并利用说服预算$eta$为最强的错误答案辩护。随着$eta$的增加，最优信号从后验最大化选项转向边际最大化选项；在$eta = 0$时，游戏重现了原始的神谕模型，听众温度为$ au = 1$。这一效应是真实的：18.2%的池中存在一个在有限预算下发生变化的最优解，每个项目的临界预算都是精确的。这一巧合在结构上限制了实证评估。两种对手框架使得七种语言模型在108个项目中对30到77个项目的选择选项发生变化，而与精确的无效率相比。然而，没有任何测量能够确定这一变化是朝向知情对手最优解还是朝向显著性，因为这两种选项是相同的。这是一个结构性限制，而不是无效结果。诊断检查是廉价的：在评估对手意识之前，验证稳健目标是否与评估项目上的启发式目标重合。
cs.AI / 14 / 2610.00234
Conflicting Supervision Moves Commitment, Not Capability: A 12.29{\sigma} arrangement effect that is exactly zero under a convention-agnostic score
冲突监督推动承诺，而非能力：在无约定评分下的12.29{C3}排列效应恰好为零
Abstract
"Train a model on the same problems written under two incompatible conventions, both correct, and ask what the ordering of that data writes into the parameters. The learning-rate schedule is not a background condition for that question. It is the averaging operator, and it decides the answer. We prove a bound in which the arrangement and the schedule enter the ordering effect as separate multiplied factors: the arrangement only as a block period, the schedule only as how much weight the endpoint can place on any one moment of the run. A decaying schedule cannot put a large step size and an uncontracted remainder at the same moment; a constant one does exactly that at the last step. That decay moderates ordering effects has been reported in pretraining; the mechanism, the separation, and a controlled measurement of both halves are ours. Ten orderings of one corpus, one budget, everything but the path held fixed, run twice under families differing in lr_scheduler_type and nothing else: at a constant rate the interior spans 0.2221 in allocation, 11.63 contrast floors, monotone in how blocked the arrangement is. Under the single cosine every published arm uses, the same ten arms occupy two distinguishable states where their own resolution would allow about ten. "Order matters" and "order does not matter" are the two ends of one knob. What the path writes is which convention the model commits to, and no exact-match benchmark can see it. Across twelve arms acc_A+acc_B is constant to within 9.7% while the allocation share runs 0.04 to 0.87, so the 12.29-sigma arrangement switch this paper measures is exactly zero under a convention-agnostic metric. That conservation is quoted from the decayed family throughout, the constant-rate one being a noisier place to read it. Marking the convention in the prompt collapses the switch and reaches 87.5% of the union ceiling."
Chinese Translation
在两个不兼容的约定下训练一个模型，两个约定都是正确的，询问这些数据的顺序如何写入参数。学习率调度并不是这个问题的背景条件。它是平均算子，并决定了答案。我们证明了一个界限，其中排列和调度作为单独的乘法因子进入顺序效应：排列仅作为一个块周期，调度仅作为端点在运行的任何一个时刻可以施加的权重。衰减调度不能在同一时刻施加大的步长和未收缩的剩余；而恒定调度在最后一步恰好做到这一点。衰减调度调节顺序效应的现象在预训练中已有报道；机制、分离以及对两个部分的控制测量是我们的贡献。对一个语料库的十种排序、一个预算，除了路径外的所有因素保持不变，在lr_scheduler_type不同的系列下运行两次：在恒定速率下，内部分配跨度为0.2221，11.63对比底线，单调于排列的阻塞程度。在每个已发布的臂使用的单一余弦下，这十个臂占据两个可区分的状态，其中它们自己的分辨率大约允许十个。“顺序重要”和“顺序不重要”是一个旋钮的两个极端。路径写入的是模型承诺的约定，而没有精确匹配的基准可以看到它。在十二个臂中，acc_A+acc_B在9.7%内保持不变，而分配份额在0.04到0.87之间，因此本文测量的12.29-σ排列切换在无约定度量下恰好为零。这种守恒在整个衰减系列中被引用，恒定速率的情况是一个更嘈杂的阅读环境。在提示中标记约定会使切换崩溃，并达到87.5%的联合上限。
cs.AI / 15 / 2610.00282
Knowing When to Yield: Grounded Arbitration of User Corrections in Text-Based Embodied Agents
知道何时让步：基于用户修正的文本嵌入代理的实证仲裁
Abstract
How should an embodied agent respond when a person's correction may be wrong? We formulate grounded correction arbitration as a choice among accepting, rejecting, inspecting the world, and asking the speaker. GAVA implements this interface with observation-bounded evidence, legal probes, and a one-step expected-loss rule. In text-only ALFWorld, 162 checkpoints produce 972 paired true and false interventions. Complete local inspections give GAVA and always verify 100 percent correction accuracy, establishing the evidence contract rather than a comparative advantage. In same-episode execution, GAVA reduces interaction cost against always verify but ties a cost threshold under a perfect speaker. An exploratory training-only object-location prior lowers interaction and declared joint cost on 340 unseen scenarios by 0.490 and 0.420 relative to uniform GAVA. After freezing the policy, costs, baselines, and multiplicity plan, the gains replicate on 77 non-overlapping seen checkpoints, covering 308 scenarios: 0.595 and 0.517, with both 95 percent checkpoint-bootstrap confidence intervals excluding zero. Joint cost also improves over an identical-prior fixed policy, while the matched calibrated no-VOI comparison remains inconclusive. Semantic GAVA makes four factual errors in each cohort, corresponding to 98.8 percent and 98.7 percent accuracy, and all methods complete every task. Results support selective information gathering with semantic priors under declared costs, but do not establish a general advantage of environmental value of information over clarification. The study uses normalized claims, complete symbolic observations, and controlled speakers; it evaluates neither human participants, visual input, nor physical robots.
Chinese Translation
当一个人的修正可能是错误时，嵌入代理应该如何回应？我们将基于实证的修正仲裁表述为在接受、拒绝、检查世界和询问说话者之间的选择。GAVA通过观察有限证据、法律探测和一步期望损失规则实现这一接口。在仅文本的ALFWorld中，162个检查点产生了972对真实和虚假的干预。完全的局部检查使GAVA始终验证100%的修正准确性，确立了证据合同而非比较优势。在同一回合执行中，GAVA相较于始终验证减少了交互成本，但在完美说话者下设定了成本阈值。一个探索性的仅训练对象位置先验在340个未见场景中相对于均匀GAVA降低了0.490和0.420的交互和声明联合成本。在冻结策略、成本、基线和多样性计划后，收益在77个不重叠的已见检查点上复制，涵盖308个场景：0.595和0.517，且两个95%的检查点自助信心区间均不包括零。联合成本也在相同先验固定策略上有所改善，而匹配校准的无价值信息（no-VOI）比较结果仍不确定。语义GAVA在每个组中犯了四个事实错误，对应于98.8%和98.7%的准确率，所有方法完成了每项任务。结果支持在声明成本下使用语义先验进行选择性信息收集，但并未确立环境信息价值相较于澄清的普遍优势。本研究使用了标准化声明、完整的符号观察和受控说话者；未评估人类参与者、视觉输入或物理机器人。
cs.AI / 16 / 2610.00313
Rules to Tools: Executable Checks for LLM Agents in Scientific Computing
从规则到工具：科学计算中LLM代理的可执行检查
Abstract
Scientific coding agents receive equations, boundary conditions, and output requirements in writing, then must assess the programs they revise. Rules to Tools (R2T) supplies prepared executable checks of public scientific requirements. Matched SciCode repair groups share written checks, starting programs, model, and budgets; the tool group receives a callable implementation. Across two task-ID cohorts, complete repair is 26/30 with text and 29/30 with the prepared checks. Three task IDs favor tools, one favors text, and eleven tie. The eight-ID cohort scores 13/16 versus 15/16, with a task-cluster bootstrap 95% interval of [-12.5, 43.75] percentage points for the difference. The larger shared-definition SciCode cohort ties at 13/24 per group. Five development-exposed tasks with alternate starting programs score 3/10 versus 7/10. The tool group favors tasks 17, 77, and 11; initial checks flag task 17 and report no violation for tasks 77 and 11. Task 37 favors text and has no initial reported violation. A fresh source-through-Python arm also reaches 15/16, matching the dedicated command's aggregate. In a matched PDE comparison, detailed text scores 23/24 and checks score 24/24, with 31.2% lower reported model output for checks. Agent-side output savings vary by cohort, while public CPU use rises in both task-ID cohorts. These results measure task-dependent repair outcomes and agent-side costs with prepared checks.
Chinese Translation
科学编码代理接收书面形式的方程、边界条件和输出要求，然后必须评估他们修订的程序。规则到工具（R2T）提供了公共科学要求的预备可执行检查。匹配的SciCode修复组共享书面检查、启动程序、模型和预算；工具组接收可调用的实现。在两个任务ID组中，使用文本的完整修复率为26/30，使用预备检查的完整修复率为29/30。三个任务ID倾向于工具，一个倾向于文本，十一项持平。八个ID组的得分为13/16，而15/16的得分则是基于任务集的自助法95%置信区间为[-12.5, 43.75]个百分点。更大共享定义的SciCode组在每组中持平，得分为13/24。五个开发暴露任务的替代启动程序得分为3/10，而7/10。工具组倾向于任务17、77和11；初始检查标记任务17，并报告任务77和11没有违规。任务37倾向于文本，并且没有初始报告的违规。一个新的通过Python的源代码组也达到了15/16，匹配专用命令的总和。在匹配的偏微分方程（PDE）比较中，详细文本得分为23/24，而检查得分为24/24，检查的模型输出报告减少了31.2%。代理端的输出节省因组而异，而公共CPU使用在两个任务ID组中均有所上升。这些结果测量了任务依赖的修复结果和代理端成本，使用了预备检查。
cs.AI / 17 / 2610.00314
Predictive Credit: Measuring What Scientific Explanations Add to Experimental Forecasts
预测信用：测量科学解释对实验预测的贡献
Abstract
Research agents explain planned experiments. We measure predictive credit with paired forecasts sharing an intervention, forecaster, and outcome while varying description, matched explanation, and donor context. Five checks track commitment, delivery, predictive gain, alignment, and known-signal uptake. Across 336 prospective states in controlled learning, 12 Tox21 endpoints, and 24 OpenML tasks, v5's frozen credit decision was inconclusive. Tox21's preregistered ROC AUC interval-score harm test was unmet ($D-M=-.0026$, 95 percent interval [$-.0174$, .0104]); OpenML's joint formation, point-equivalence, and repeatability rule was unmet. Matched point-accuracy gains over description remained unconfirmed, and Tox21/OpenML seed-donor intervals spanned zero. Under requested DeepSeek V4 Pro, matched and donor cards reduced secondary Tox21 drift by 64.5 and 59.1 percent. A DeepSeek V4 Flash replay raised matched point MAE from .01823 to .02020 and missed matched-donor interval-score equivalence. OpenML full-card assignment widened nominal 80 percent intervals by 21 percent, with 49.3 percent coverage versus 51.4 percent for description and content in 66/144 cards. Direct-text Flash delivered all 144 notes without detectable matched point-accuracy gain. A researcher-authored mechanism positive control lowered point MAE by 2.60 percentage points versus description. The protocol measures predictive credit for research-agent benchmarks and scientific forecasting; natural-explanation credit remained unconfirmed at the tested donor resolutions.
Chinese Translation
研究代理人解释计划中的实验。我们通过共享干预、预测者和结果的配对预测来测量预测信用，同时变化描述、匹配解释和捐赠者背景。五项检查跟踪承诺、交付、预测增益、一致性和已知信号的吸收。在336个受控学习的前瞻性状态、12个Tox21终点和24个OpenML任务中，v5的冻结信用决策结果不确定。Tox21的预注册ROC AUC区间得分伤害测试未满足（$D-M=-.0026$, 95%区间 [$-.0174$, .0104]）；OpenML的联合形成、点等价和重复性规则未满足。描述的匹配点准确性增益仍未得到确认，Tox21/OpenML种子捐赠者区间跨越零。在请求的DeepSeek V4 Pro下，匹配和捐赠卡减少了次级Tox21漂移分别为64.5%和59.1%。DeepSeek V4 Flash重播将匹配点的平均绝对误差（MAE）从.01823提高到.02020，并未达到匹配捐赠者区间得分的等价性。OpenML的全卡分配将名义80%区间扩大了21%，在66/144张卡中，覆盖率为49.3%，而描述和内容的覆盖率为51.4%。直接文本Flash传递了所有144条注释，但未检测到匹配点准确性的增益。一种研究者撰写的机制正控制将点MAE降低了2.60个百分点，相较于描述。该协议测量研究代理基准和科学预测的预测信用；在测试的捐赠者分辨率下，自然解释信用仍未得到确认。
cs.AI / 18 / 2610.00328
ContractRL: Shielded Group-Relative Policy Optimization for Auditable Tool-Call Repair
ContractRL：可审计工具调用修复的受保护组相对策略优化
Abstract
Structured tool calls often fail after only a small number of fields violate a schema or an execution contract. Regenerating the complete object enlarges the action surface and makes repeated repair difficult to audit. We introduce ContractRL, a contract-constrained sequential repair protocol that models verifier-guided JSON repair as a bounded decision process. At each step the policy observes the candidate, typed verifier feedback, JSON Pointer, immutable repair history, and remaining budget; a contract-derived action mask filters malformed or prohibited RFC-6902 operations before a deterministic validator performs the transition. We specify a contract-constrained group-relative objective for patch, retry, and abstention decisions while keeping canonical targets and semantic labels outside the online state until trace freeze. Under identical verifier information, ContractRL attains 0.9362 semantic success with 34.4 generated tokens, compared with 0.9076 and 44.9 tokens for Patch-SFT and 0.9148 and 137.2 tokens for full regeneration over 192 cases per seed and five seeds. Policy optimization improves semantic success from 0.9186 for supervised ContractRL to 0.9375. A separate three-seed paired evaluation against Patch-SFT yields a semantic difference of +0.0396 (95% CI $[+0.0137,+0.0662], p=0.0039$). Feedback, action-mask, budget, and schema-shift analyses connect these gains to localized correction, while adversarial and multi-turn evaluations characterize the remaining failure modes.
Chinese Translation
结构化工具调用在仅有少数字段违反模式或执行合同后往往会失败。重新生成完整对象扩大了操作面，使得重复修复难以审计。我们提出了ContractRL，这是一种合同约束的顺序修复协议，将验证者引导的JSON修复建模为一个有界决策过程。在每一步中，策略观察候选对象、类型化的验证者反馈、JSON指针、不可变的修复历史和剩余预算；合同派生的动作掩码在确定性验证者执行转换之前过滤掉格式错误或禁止的RFC-6902操作。我们为补丁、重试和放弃决策指定了一个合同约束的组相对目标，同时在在线状态下保持规范目标和语义标签在追踪冻结之前不变。在相同的验证者信息下，ContractRL在192个案例和五个种子下实现了0.9362的语义成功率，生成了34.4个标记，而Patch-SFT的成功率为0.9076，生成了44.9个标记，完全重生成的成功率为0.9148，生成了137.2个标记。策略优化将监督下的ContractRL的语义成功率从0.9186提高到0.9375。与Patch-SFT的单独三种子配对评估显示语义差异为+0.0396（95% CI $[+0.0137,+0.0662], p=0.0039$）。反馈、动作掩码、预算和模式转移分析将这些提升与局部修正联系起来，而对抗性和多轮评估则表征了剩余的失败模式。
cs.AI / 19 / 2610.00331
Mathematical Transfer in LLMs Follows Reasoning Approach More Than Topic
大型语言模型中的数学迁移更依赖于推理方法而非主题
Abstract
When selecting mathematical training data for LLMs, a natural organizing principle is topic: probability examples for probability targets. An alternative is reasoning approach: worked solutions that share a solution method with the target, even when the mathematical domain differs. We ask which relation produces greater transfer after fine-tuning. We evaluate two counterbalanced $2\times2$ designs: probability and combinatorics crossed with invariant reasoning and double counting (2,000 problems), and number theory and geometry crossed with complement and pigeonhole reasoning (800 problems). In each design, every cell serves as the held-out target in turn: same-approach (SA) sources share the target's method but change the topic, while same-topic (ST) sources share the topic but change the method. Every source appears once in each role, so additive source-quality effects cancel from the equally weighted aggregate contrast. Across five base models and three training seeds per design, SA outperforms ST in all 40 seed-pooled model--target comparisons. Model-level advantages range from 8.2 to 16.2 percentage points in the primary design (mean: 10.8) and from 12.0 to 16.0 in the second design (mean: 14.3); all ten model-level 95% confidence intervals exclude zero. In both designs, ST sources are more similar to targets under embedding and lexical measures, so the SA advantage runs opposite to the measured ordering of statement-level resemblance. These findings identify reasoning approach as a more effective matching criterion than topic for mathematical transfer across the evaluated topic--approach combinations.
Chinese Translation
在为大型语言模型（LLMs）选择数学训练数据时，自然的组织原则是主题：针对概率目标的概率示例。另一种选择是推理方法：共享解决方案方法的已解决方案，即使数学领域不同。我们探讨哪种关系在微调后产生更大的迁移效果。我们评估了两个平衡的 $2 imes2$ 设计：概率和组合学与不变推理和重复计数交叉（2,000 道题目），以及数论和几何与补集和鸽巢推理交叉（800 道题目）。在每个设计中，每个单元格依次作为保留目标：同方法（SA）来源共享目标的方法但改变主题，而同主题（ST）来源共享主题但改变方法。每个来源在每个角色中出现一次，因此加性来源质量效应在等权重的总体对比中相互抵消。在五个基础模型和每个设计三个训练种子的情况下，SA 在所有 40 个种子汇总的模型-目标比较中均优于 ST。在主要设计中，模型级优势范围为 8.2 到 16.2 个百分点（均值：10.8），在第二个设计中范围为 12.0 到 16.0（均值：14.3）；所有十个模型级 95% 置信区间均不包括零。在两个设计中，ST 来源在嵌入和词汇度量下与目标更相似，因此 SA 优势与声明级相似度的测量顺序相反。这些发现表明，推理方法作为数学迁移的匹配标准比主题更为有效，适用于评估的主题-方法组合。
cs.AI / 20 / 2610.00349
Fault-Tolerant Budget Conservation in Distributed Multi-Agent Delegation
分布式多智能体委托中的容错预算保护
Abstract
Resource limits are becoming an authorization boundary for AI agents that delegate work across concurrent and failure-prone workers. Parent-child allocation constraints, affine objects, and distributed escrow do not by themselves prevent overspend when replies are lost, effects complete after timeout, messages repeat, branches partition, or DAG joins alias one lineage. We formalize fault-tolerant budget conservation for distributed multi-agent delegation. Budgets are quantized resource vectors represented by exclusive escrow credits that move through a delegation DAG. Before dispatch, a branch converts credit into an operation reservation bound to lineage, epoch, normalized effect, maximum charge, receiver, and idempotency key. It persists a signed dispatch permit with quarantine; the gateway verifies that permit before first acceptance. Uncertain effects remain charged until authenticated settlement, a fenced authoritative no-effect proof, or permanent retirement. We prove ownership partition, ledger and effect conservation, descendant non-amplification, at-most-once settlement, late-completion safety, and partition confinement under explicit mediation, durability, authentication, normalization, and gateway assumptions. An indistinguishability result shows that partition-local availability requires exclusive preallocation. Bounded TLA+ checking, an independent JavaScript explorer, and crash-injected two-process SQLite experiments exercise the declared scope and detect timeout-refund and historical-certificate-validation mutants. The mechanism preserves the issued budget bound across the evaluated crash, retry, duplicate, partition, join, and late-completion schedules.
Chinese Translation
资源限制正成为在并发和易失性工作者之间委托工作的人工智能代理的授权边界。父子分配约束、仿射对象和分布式托管本身并不能防止在回复丢失、效果在超时后完成、消息重复、分支分区或有向无环图（DAG）连接别名同一血统时的超支。我们对分布式多智能体委托中的容错预算保护进行了形式化。预算是通过独占托管信用表示的量化资源向量，这些信用在委托DAG中流动。在派遣之前，分支将信用转换为绑定到血统、纪元、标准化效果、最大费用、接收者和幂等性密钥的操作预留。它持久化一个带有隔离的签名派遣许可证；网关在首次接受之前验证该许可证。不确定效果在经过认证的结算、一个有围栏的权威无效果证明或永久退役之前保持收费。我们证明了所有权分区、账本和效果保护、后代非放大、至多一次结算、延迟完成安全性和在明确调解、持久性、认证、标准化和网关假设下的分区限制。一个不可区分性结果表明，分区局部可用性需要独占预分配。有限的 TLA+ 检查、一个独立的 JavaScript 探索器以及注入崩溃的两个进程 SQLite 实验验证了声明的范围，并检测到超时退款和历史证书验证的变异体。该机制在评估的崩溃、重试、重复、分区、连接和延迟完成调度中保持已发放预算的界限。
cs.AI / 21 / 2610.00353
JusticeAxis: Benchmarking Legal Judgment between Rigid Rule Application and Ungrounded Discretion
JusticeAxis：在严格规则适用与无依据自由裁量之间评估法律判断
Abstract
A sound judgment applies the law to established facts and weighs the circumstances in which they arose. However, existing methods swing between rigid statute matching and ungrounded discretion, benchmarks score a label or a rubric, and the experience that would supply the balance stays unverified. We formalize legal judgment as a reference-anchored task, whose object is a single decision that stays tied to the statute and to the circumstances at once. We introduce JusticeAxis, 256 real-world criminal cases from 18 countries with audio, image, and text evidence, and three lawyer-written judgments for every case: the recorded one and one for each failure. We further propose JusticeAgent, a harness whose element agents establish the facts and whose judge agent applies the law under skills carrying experience of the circumstances. Skills are distilled from execution trajectories and admitted only under Bayesian credible bounds. Experiments show that failure turns direction with scale: open-weight backbones drift to unsupported grounds, frontier models to the statutory default. We further verify that JusticeAgent, as a simple yet effective plugin, carries a frozen open-weight backbone to commercial level. Project resources are available at https://github.com/beita6969/JusticeAxis.
Chinese Translation
合理的判断将法律应用于既定事实，并权衡其产生的环境。然而，现有方法在严格的法规匹配与无依据的自由裁量之间摇摆不定，基准评分仅依赖标签或评分标准，而能够提供平衡的经验却未得到验证。我们将法律判断形式化为一个参考锚定的任务，其目标是一个既与法规又与情境紧密相连的单一决策。我们引入了JusticeAxis，这是来自18个国家的256个真实刑事案件，包含音频、图像和文本证据，并为每个案件提供三份律师撰写的判决：记录的判决和每个失败的判决。我们进一步提出了JusticeAgent，这是一个工具，其元素代理建立事实，而法官代理在具备情境经验的技能下应用法律。技能从执行轨迹中提炼，并仅在贝叶斯可信范围内被接受。实验表明，失败的方向随规模而变化：开放权重的骨干网络漂移至无支持的基础，前沿模型则回归法定默认。我们进一步验证了JusticeAgent作为一个简单而有效的插件，能够将冻结的开放权重骨干提升至商业水平。项目资源可在 https://github.com/beita6969/JusticeAxis 获取。
cs.AI / 22 / 2610.00366
What Should an Agent Remember? Disentangling Retention from Retrieval in Bounded-Memory Evaluation
代理应该记住什么？在有限记忆评估中区分保留与检索
Abstract
A persistent agent must decide both what to retain as information arrives and what to surface once a query appears, yet memory evaluations can confound these decisions by comparing methods that differ in both retention and selection. We build a streaming-recall benchmark crossing retention and selection rules and evaluate every condition on the same 300 seeded episodes. Holding access fixed, query-aware selection improves required-fact recall by 15.5 percentage points (95% CI: 12.8 to 18.2), whereas a mixed comparison that also changes history access reports a 68.7-point advantage, of which 53.2 points are attributable to access. Under bounded retention, query-aware, dense, and oracle selection reach the retention ceiling, and all 319 observed failures in the bounded recency condition are caused by eviction rather than ranking errors. Recall falls to 0% as targets recede sufficiently far into the past. Repeating the evaluation on SQuAD preserves the retention ceiling while showing that dense retrieval can outperform lexical retrieval on natural text. These results show that bounded-memory evaluations should hold access fixed and report retention and selection separately.
Chinese Translation
一个持久的代理必须决定在信息到达时保留什么，以及在查询出现时展示什么，然而，记忆评估可能会通过比较在保留和选择上有所不同的方法而混淆这些决策。我们建立了一个流式回忆基准，交叉保留和选择规则，并在相同的300个种子情节上评估每个条件。在固定访问的情况下，基于查询的选择提高了所需事实的回忆率15.5个百分点（95%置信区间：12.8到18.2），而一个混合比较同时改变历史访问则报告了68.7点的优势，其中53.2点可归因于访问。在有限保留的情况下，基于查询的、密集的和oracle选择达到了保留上限，并且在有限近期条件下观察到的319个失败都是由于驱逐而非排名错误造成的。当目标远离过去时，回忆率降至0%。在SQuAD上重复评估保持了保留上限，同时显示密集检索在自然文本上可以优于词汇检索。这些结果表明，有限记忆评估应保持访问固定，并分别报告保留和选择。
cs.AI / 23 / 2610.00372
When Harnesses Lose the Signal: Causal Evaluation of Recovery in LLM Agents
当外部工具失去信号时：大语言模型代理恢复的因果评估
Abstract
Large language model agents rely on external harnesses to pass information between the model and its environment and to recover from execution errors. Yet recovery is usually judged only by average task success. This hides an important tension. The same operation can rescue a failing trajectory or disrupt one that would otherwise succeed. We frame recovery as a causal decision problem. Starting from the same execution state, we compare what happens with and without recovery, separate rescue from harm, and study how the value of recovery changes over time. We then introduce the Causal Intervention Router (CIR), a lightweight policy that uses information available before recovery to decide when intervention is worthwhile. On long-horizon ALFWorld tasks with Qwen3-14B, CIR raises success from 70.33% to 73.33%, a gain of 3.00 percentage points. It leaves all evaluated trajectories with correct observations untouched. Additional controls show that the benefit of recovery cannot be explained solely by the new observation returned by the environment. These results provide a practical way to evaluate recovery and apply it selectively.
Chinese Translation
大型语言模型代理依赖外部工具在模型与其环境之间传递信息，并从执行错误中恢复。然而，恢复通常仅通过平均任务成功率来评判。这掩盖了一个重要的矛盾。相同的操作可以拯救一个失败的轨迹，或破坏一个本可以成功的轨迹。我们将恢复框架化为一个因果决策问题。从相同的执行状态出发，我们比较有无恢复的情况，区分救助与伤害，并研究恢复的价值如何随时间变化。接着，我们引入了因果干预路由器（Causal Intervention Router, CIR），这是一种轻量级策略，利用恢复前可用的信息来决定何时进行干预是值得的。在使用 Qwen3-14B 的长时间跨度 ALFWorld 任务中，CIR 将成功率从 70.33% 提升至 73.33%，增幅为 3.00 个百分点。它保持了所有评估轨迹的正确观察不变。额外的控制实验表明，恢复的好处不能仅通过环境返回的新观察来解释。这些结果提供了一种实用的方法来评估恢复并选择性地应用它。
cs.AI / 24 / 2610.00416
Benchmarking Prompt Optimization of Large Language Models With Chess
基于国际象棋的大型语言模型提示优化基准测试
Abstract
Evaluating large language models becomes increasingly challenging as their capabilities advance: benchmarks can saturate, public test sets risk contamination, and assessing harder tasks can require expensive grading or execution infrastructure. These challenges are amplified in automatic prompt optimization (APO), where evaluation is repeated throughout the search for better prompts. Studying APO therefore requires a benchmark that is cheap and deterministic to score, hard enough to leave room for improvement, and renewable as models evolve. We introduce a chess benchmark built from 1,118 Lichess puzzles to study APO for frozen LLMs: we optimize their prompts without updating their model weights. Chess combines inexpensive exact-match scoring, engine-based evaluation of alternative moves, and a renewable supply of problems with adjustable difficulty. Unlike evaluations that report only success on isolated test items, the benchmark also connects puzzle-solving gains to short game-play rollouts within the same domain. We use it to evaluate six APO algorithms on eight target models, measuring not only baseline strength but also how much each model responds to optimization and whether optimized prompts transfer across models and to game play. Chess is thus a well-suited benchmark for APO: it is (i) challenging, as even the strongest evaluated model, Gemini 3.5 Flash (used as the meta-model), solves only about 55\% of puzzles; (ii) discriminative, revealing gains, unchanged performance, and regressions across methods and models; (iii) renewable, with fresh puzzles to reduce contamination risk and adjustable difficulty to maintain headroom as models improve; and (iv) affordable, as the complete study runs for around \$800. We release the puzzles, optimization and evaluation code, and dataset-renewal scripts (https://github.com/imec-ailabs/Automatic-Prompt-Optimization-with-Chess).
Chinese Translation
随着大型语言模型能力的提升，评估变得愈加困难：基准测试可能饱和，公共测试集面临污染风险，而评估更复杂的任务可能需要昂贵的评分或执行基础设施。这些挑战在自动提示优化（APO）中尤为突出，因为在寻找更好提示的过程中，评估会被重复进行。因此，研究APO需要一个便宜且确定性的基准，以便于评分，难度足够大以留有改进空间，并且随着模型的发展可以更新。我们引入了一个基于1,118个Lichess难题的国际象棋基准，用于研究冻结的LLM的APO：我们在不更新模型权重的情况下优化它们的提示。国际象棋结合了便宜的精确匹配评分、基于引擎的替代走法评估，以及可调难度的可再生问题供应。与仅报告孤立测试项目成功的评估不同，该基准还将解题收益与同一领域内的短期游戏回合连接起来。我们使用它评估六种APO算法在八个目标模型上的表现，不仅测量基线强度，还测量每个模型对优化的响应程度，以及优化提示是否能够在模型之间转移并应用于游戏玩法。因此，国际象棋是一个非常适合APO的基准：它具有（i）挑战性，即使是评估的最强模型Gemini 3.5 Flash（作为元模型）也仅能解决约55%的难题；（ii）区分性，揭示不同方法和模型之间的收益、性能不变和回归；（iii）可再生性，提供新鲜的难题以减少污染风险，并可调节难度以保持模型改进的余地；（iv）经济性，整个研究的费用约为800美元。我们发布了难题、优化和评估代码，以及数据集更新脚本（https://github.com/imec-ailabs/Automatic-Prompt-Optimization-with-Chess）。
cs.AI / 25 / 2610.00437
JevSpawn: Adaptive Agentic Inference through Compositional Action Spaces
JevSpawn：通过组合动作空间进行自适应代理推理
Abstract
LLM agents generate intermediate reasoning and actions token by token, making extended interactions slow and computationally expensive. Jev-style models offer fast probabilistic predictions over finite fields, but require those fields to be specified in advance. This requirement limits autonomous task solving, where the available actions must be derived from natural language instructions and adapted through interaction. We introduce JevSpawn, a compositional policy that connects natural language task specifications to finite probabilistic exploration. Parallel action spawning is coupled with feedback driven branch selection, representation revision, and recovery from retained alternatives. Shared action structure and model prefixes reduce repeated generation and context computation without additional training. Evaluations on eight benchmark tasks against seven agent baselines and a TypeSafe Jev variant establish JevSpawn as a promising approach to structured agentic inference, with improved task performance and faster navigation.
Chinese Translation
大型语言模型（LLM）代理逐字生成中间推理和动作，使得扩展交互变得缓慢且计算成本高昂。Jev风格模型在有限域上提供快速的概率预测，但要求这些域必须提前指定。这一要求限制了自主任务解决，其中可用的动作必须从自然语言指令中推导并通过交互进行适应。我们提出了JevSpawn，这是一种组合策略，将自然语言任务规范与有限概率探索连接起来。并行动作生成与基于反馈的分支选择、表示修订和从保留备选方案中恢复相结合。共享的动作结构和模型前缀减少了重复生成和上下文计算，而无需额外训练。在八个基准任务上与七个代理基线和一个类型安全的Jev变体进行评估，确立了JevSpawn作为一种有前景的结构化代理推理方法，具有改进的任务性能和更快的导航速度。
cs.AI / 26 / 2610.00447
Frozen Scenes, Shifting Winners: Configuration Fragility in Text-to-3D Evaluation
冻结场景，变动赢家：文本到三维评估中的配置脆弱性
Abstract
Can a text-to-3D leaderboard change when every generated scene stays fixed? We audit this question for rendered-image evaluation, where camera settings and caption wording become part of the measurement protocol. Across 300 frozen scenes from six generators, we vary eight render and caption factors for 19 alignment evaluators plus one perceptual-quality control, then test four targeted scene degradations. Peak configuration variance exceeds between-generator variance for 17/19 alignment evaluators, with prompt-bootstrap lower bounds above 1 for 11/19. Rankings are more stable than scores, yet 18/19 evaluators change their point-estimate winner under some configuration. Pairwise protocol margin envelopes show which comparisons keep their direction across the tested settings. Selected pairs have opposite pointwise intervals, but no reversal survives simultaneous inference over the full search. Thus the observed winner changes are descriptive, not confirmed changes in generator superiority. Sensitivity remains separate: no evaluator, even the prompt-free control, exceeds 67% tie-adjusted directional discrimination on layout scrambling, which is diagnostic rather than human-validated ground truth. The audit separates score stability, decision uncertainty, and targeted sensitivity, and recommends reporting (generator, score, card ID) with protocol-dependent comparisons and selection-aware uncertainty.
Chinese Translation
当每个生成的场景保持固定时，文本到三维的排行榜会发生变化吗？我们审视了这个问题，聚焦于渲染图像评估，其中相机设置和标题措辞成为测量协议的一部分。在来自六个生成器的300个固定场景中，我们对19个对齐评估者和一个感知质量控制进行了八个渲染和标题因素的变化，然后测试了四种针对性的场景降级。对于17/19个对齐评估者，峰值配置方差超过了生成器间方差，11/19的提示引导下限高于1。排名比分数更稳定，但在某些配置下，18/19个评估者的点估计赢家发生了变化。成对协议边际包络显示了哪些比较在测试设置中保持其方向。选定的对具有相反的逐点区间，但没有逆转在全搜索的同时推断中存活。因此，观察到的赢家变化是描述性的，而不是生成器优越性的确认变化。敏感性仍然是独立的：没有评估者，即使是无提示控制，在布局扰动上超过67%的平局调整方向性区分，这是一种诊断而非经过人类验证的真实情况。审计分离了分数稳定性、决策不确定性和针对性敏感性，并建议报告（生成器、分数、卡片ID）与协议依赖的比较和选择感知的不确定性。
cs.AI / 27 / 2610.00511
Before Agents Decide: Epistemic Action in LLM-Based Systems
代理决策之前：基于大型语言模型的系统中的认知行动
Abstract
Before a difficult decision, people often act simply to understand the situation better. We turn an object to see another side, place alternatives next to each other, or change one condition and observe what happens. These actions may not complete the task, but they improve the evidence needed for the next choice. LLM-based agents can search and explore, yet agent design gives less attention to an earlier question: is the available evidence ready for the decision? Sometimes necessary evidence is missing. In other cases, the evidence is present but its form hides what matters, or the comparison needed to judge it does not yet exist. Cognitive science calls actions that improve the basis for a later choice epistemic actions. We bring this idea to LLM-based agents and distinguish three modes: acquiring missing evidence, transforming available evidence, and probing a system to create a revealing response. We use the term epistemic scaffolding for the interfaces, tools, and environments that make these actions possible and auditable. This paper argues that agent design must address how decision-ready evidence is produced.
Chinese Translation
在面临困难决策之前，人们常常采取行动以更好地理解情况。我们会转动一个物体以查看另一面，将替代方案并排放置，或改变一个条件并观察发生的情况。这些行动可能无法完成任务，但它们改善了下一次选择所需的证据。基于大型语言模型（LLM）的代理可以进行搜索和探索，但代理设计对一个早期问题的关注较少：可用的证据是否准备好用于决策？有时必要的证据缺失。在其他情况下，证据虽然存在，但其形式掩盖了重要信息，或者进行判断所需的比较尚不存在。认知科学称那些改善后续选择基础的行动为认知行动。我们将这一概念引入基于LLM的代理，并区分三种模式：获取缺失的证据、转化可用的证据，以及探测系统以产生揭示性的响应。我们使用“认知支架”（epistemic scaffolding）一词来指代使这些行动成为可能且可审计的接口、工具和环境。本文认为，代理设计必须关注如何生成决策准备好的证据。
cs.AI / 28 / 2610.00529
Ontology-Based Contextual AI Evaluations (OB-CAIE) Methodology
基于本体的上下文人工智能评估（OB-CAIE）方法论
Abstract
The ontology-based contextual AI evaluation (OB-CAIE) methodology was developed to address a lack of scientific rigor that arises from unclear testing coverage, to balance human expertise and automations, and to address a lack of reproducibility of AI evaluation testing environments. OB-CAIE strengthens the current state of AI evaluations by addressing the first step in the scientific method by clearly defining what will be tested. Two ontologies represent the tractable problem space in the OB-CAIE methodology: the Domain-Specific Ontology (DSO) and the Evaluation Process Ontology (EPO). The DSO is the what; the EPO is the how. An OB-CAIE problem space can be used for one or multiple AI evaluations. The OB-CAIE methodology allows for human judgment at specific points, in scientifically grounded ways, and in complex subject areas where human feedback is genuinely irreducible or machine irreplaceable. A key advantage of the OB-CAIE methodology is that failure points can be traced, visualized and analyzed within the canonical OB-CAIE methodology problem space.
Chinese Translation
基于本体的上下文人工智能评估（OB-CAIE）方法论的开发旨在解决由于测试覆盖不明确而导致的科学严谨性不足的问题，平衡人类专业知识与自动化的关系，并解决人工智能评估测试环境缺乏可重复性的问题。OB-CAIE通过明确界定将要测试的内容，强化了人工智能评估的当前状态，成为科学方法的第一步。OB-CAIE方法论中有两个本体代表了可处理的问题空间：领域特定本体（Domain-Specific Ontology, DSO）和评估过程本体（Evaluation Process Ontology, EPO）。DSO代表“什么”；EPO代表“如何”。OB-CAIE问题空间可以用于一个或多个人工智能评估。OB-CAIE方法论允许在特定节点进行人类判断，以科学依据的方式，在人类反馈确实不可减少或机器无法替代的复杂领域中进行评估。OB-CAIE方法论的一个关键优势在于，可以在规范的OB-CAIE方法论问题空间内追踪、可视化和分析故障点。
cs.AI / 29 / 2610.00531
Science or Slop?: Benchmarking and Mitigating Scientific Slop in AI-Generated Papers
科学还是糟粕？：基准测试与缓解AI生成论文中的科学糟粕
Abstract
AI-generated content, often called AI slop, is increasingly common everywhere, particularly in academia. Slop in AI-generated scientific papers, however, has more complex patterns that cannot be easily detected by existing token-based AI detectors. Each part of such a paper looks plausible while the scientific reasoning that connects the parts breaks down, which can mislead how readers assess the work. We benchmark these failures as scientific slop through six measures across Structure, Argument, and Artifacts. We construct SciSlopBench with 390 AI-generated papers, mostly in computer science but spanning the life, social, and natural sciences, each paired with a human-written paper matched by research problem and contribution type. Our measures identify the AI paper in each pair with 85.9% accuracy, compared with 68.7% for Binoculars. Higher scientific slop accompanies lower ICLR ratings and distinguishes rejected from accepted papers above chance in every year from 2017 to 2025. Reducing these patterns, however, is not as simple as directly optimizing the measures. We therefore propose SciSlopHarness, a harness-level framework that guides a fixed LLM to revise slop only where the experiment records support the change. While standard revisions leave residual slop and direct slop-aware prompting triggers reward hacking, SciSlopHarness reduces the remaining AI-human gap by 63% over the strongest revision baseline without requiring human reference targets. Overall, we demonstrate that AI-generated scientific papers leave fundamental traces in their global reasoning, and that responsible mitigation demands strict evidentiary grounding rather than mere prose refinement.
Chinese Translation
AI生成的内容，通常被称为AI糟粕，正越来越普遍，尤其是在学术界。然而，AI生成科学论文中的糟粕具有更复杂的模式，这些模式无法被现有的基于标记的AI检测器轻易识别。这类论文的每个部分看似合理，但连接各部分的科学推理却崩溃，这可能会误导读者对工作的评估。我们通过结构、论证和文物六个指标对这些失败进行基准测试，称之为科学糟粕。我们构建了SciSlopBench，包含390篇AI生成的论文，主要集中在计算机科学领域，但也涵盖了生命科学、社会科学和自然科学，每篇论文都与一篇根据研究问题和贡献类型匹配的人类撰写论文配对。我们的指标以85.9%的准确率识别每对中的AI论文，而Binoculars的准确率为68.7%。更高的科学糟粕伴随着较低的ICLR评分，并且在2017年至2025年的每一年中，能够区分被拒绝和被接受的论文，超出随机概率。然而，减少这些模式并不像直接优化指标那么简单。因此，我们提出了SciSlopHarness，一个引导固定LLM在实验记录支持变更的情况下仅修订糟粕的框架。虽然标准修订留下了残余的糟粕，而直接的糟粕感知提示则会触发奖励黑客行为，SciSlopHarness在不需要人类参考目标的情况下，将剩余的AI与人类之间的差距减少了63%，超越了最强修订基线。总体而言，我们展示了AI生成的科学论文在其全球推理中留下了根本性痕迹，并且负责任的缓解需要严格的证据基础，而不仅仅是文字的修饰。
cs.AI / 30 / 2610.00583
Worse Together: How Performance Breaks Down in Multi-User Multi-Agent Teams
更糟的合作：多用户多智能体团队中的性能崩溃
Abstract
People are increasingly delegating tasks to AI agents, and those agents are increasingly encountering other people's agents over shared resources such as a codebase, a calendar, or a budget. When each agent acts for a different user with different goals, coordination often fails, and the group ends up worse off than if a single agent had acted for everyone. We study this multi-user, multi-agent setting across five frontier models and 77 scenarios in four environments: an API key environment in which agents share a compute budget, a clinic in which they share a calendar, a personal assistant environment in which they share a group order or booking, and a merge queue in which they share a release cutoff. In each scenario, we compare a single agent that serves every user (a coordinator) to a team in which each agent serves one user, with and without a communication channel between the agents. Teams deliver worse group outcomes than the coordinator in every environment: without a channel, they completely collapse in two environments, and even with one, coordination overhead creates substantial gaps. For example, in the personal assistant environment, the coordinator fulfills a targeted user request about twice as often as teams. We identify distinct behaviors associated with this poor group-level performance, including stalling as teams grow, overriding each other's actions, and fabricating claims. We find effective but environment-specific mitigations, such as a team lead, explicit procedural instructions, and a platform check that makes an agent read its peers' messages before committing. We will release the API key, clinic, and personal assistant environments as MAMUBench, comprising 74 scenarios for evaluating multi-user, multi-agent coordination.
Chinese Translation
人们越来越多地将任务委托给人工智能代理，而这些代理也越来越多地与其他人的代理在共享资源（如代码库、日历或预算）上相遇。当每个代理代表不同用户并拥有不同目标时，协调往往失败，团队的表现最终不如由单一代理为所有人行动的情况。我们研究了这一多用户、多智能体的设置，涵盖了五个前沿模型和四个环境中的77个场景：一个API密钥环境，其中代理共享计算预算；一个诊所环境，其中他们共享日历；一个个人助理环境，其中他们共享团体订单或预订；以及一个合并队列环境，其中他们共享发布截止日期。在每个场景中，我们将一个为每个用户服务的单一代理（协调者）与一个每个代理为一个用户服务的团队进行比较，并考察在有无代理之间的沟通渠道的情况下的表现。团队在每个环境中的整体结果都不如协调者：在没有沟通渠道的情况下，他们在两个环境中完全崩溃，即使有一个沟通渠道，协调的开销也造成了显著的差距。例如，在个人助理环境中，协调者满足特定用户请求的频率大约是团队的两倍。我们识别出与这种低效团队表现相关的不同行为，包括团队规模增长时的停滞、相互覆盖对方的行动以及捏造声明。我们发现了一些有效但特定于环境的缓解措施，如团队领导、明确的程序指令以及一个平台检查，使代理在提交之前阅读同伴的消息。我们将发布API密钥、诊所和个人助理环境作为MAMUBench，包含74个场景，用于评估多用户、多智能体的协调。
cs.AI / 31 / 2610.00609
Legal Research Bench: Measuring End-to-End Reliability in Long-Horizon Legal Research Agents
法律研究基准：测量长时间范围法律研究代理的端到端可靠性
Abstract
Legal research is a core and time-consuming legal workflow. Lawyers must identify controlling authority, verify that it remains valid, reconcile statutes and cases, and synthesize a grounded answer. Language model agents are a natural fit for this retrieval-intensive workflow, and automating even part of it would be valuable. But that value depends on reliability: a single missing authority, stale citation, or wrong legal conclusion can make an otherwise plausible answer unusable. We introduce \textbf{Legal Research Bench} (LRB), a benchmark of 413 open-ended U.S. legal research questions written by experts, each paired with a gold answer, supporting authorities, and a binary grading rubric. We evaluate thirteen frontier models in a harness with web search, case-law search, page parsing, and retrieval tools. We score agent responses through all-pass grading with source verification, where a response is correct only if every required criterion is satisfied and its cited authorities verify. We also validate the LLM judge against expert attorneys ensuring that benchmark scores track attorney judgment. Agents remain far from reliable: among the models we tested, the strongest, Claude Opus 4.8, is fully correct on 42.9\% of questions. Performance also varies substantially by task setting: all-pass rates differ across areas of law and are lower on questions requiring reconciliation of conflicting authorities. Across models, more turns, tool calls, and inference cost do not predict higher accuracy.
Chinese Translation
法律研究是核心且耗时的法律工作流程。律师必须识别控制性权威，验证其有效性，调和法规和案例，并综合出一个有根据的答案。语言模型代理非常适合这一以检索为主的工作流程，自动化其中的部分环节将具有重要价值。但这种价值依赖于可靠性：单一缺失的权威、过时的引用或错误的法律结论都可能使一个看似合理的答案变得不可用。我们介绍了 extbf{法律研究基准}（Legal Research Bench，LRB），这是一个包含413个由专家撰写的开放式美国法律研究问题的基准，每个问题都配有黄金答案、支持的权威和二元评分标准。我们在一个结合了网络搜索、案例法律搜索、页面解析和检索工具的环境中评估了十三个前沿模型。我们通过源验证的全通过评分来评估代理的响应，只有当每个必需标准都得到满足且其引用的权威得到验证时，响应才被视为正确。我们还将LLM评审与专家律师进行了验证，以确保基准分数与律师判断相符。代理的可靠性仍然远未达到：在我们测试的模型中，最强的Claude Opus 4.8在42.9\%的问题上完全正确。性能在任务设置上也存在显著差异：不同法律领域的全通过率不同，并且在需要调和冲突权威的问题上更低。在各模型中，更多的轮次、工具调用和推理成本并不预测更高的准确性。
cs.AI / 32 / 2610.00613
Spatial Strategies, Not Actions: Vector-Quantized Geodesics as Tools for LLM-Driven Agents
空间策略，而非行动：向量量化测地线作为大语言模型驱动代理的工具
Abstract
Large language model (LLM) based agents are often criticized for lacking spatial understanding and mainly exploiting statistical text patterns. We investigate their spatial comprehension through an architecture combining geometrical tools with a LLM serving as a high-level orchestrator in grid-world environments. The agent first collects geodesic trajectories, which are then vector-quantized to extract a representative subset. Offline, the LLM associates a natural language description of the underlying behavioral patterns to each selected trajectory, making it a tool. Online, the LLM chooses the appropriate tool conditioned on the current state and goal. Low-level control is handled by primitive actions that execute the trajectory associated with the tool. From an agentic AI perspective, this approach separates learning into two levels: tool discovery is handled through unsupervised quantization of trajectories, while reasoning and decision-making are handled by the LLM. We test the approach in a partially observable dynamic 2D grid environment with an open vision-language model (Qwen3.6-35B-A3B). Pairing the geometry-derived tool library with an agent-centered zoom tool and a collision detection tool lets a fast, non-reasoning configuration match the goal-reaching rate of a much more costly chain-of-thought version, while cutting the cost of a decision from minutes to seconds.
Chinese Translation
基于大语言模型（LLM）的代理常常被批评缺乏空间理解，主要依赖统计文本模式。我们通过一种将几何工具与LLM结合的架构，研究其空间理解能力，该LLM在网格世界环境中充当高层协调者。代理首先收集测地线轨迹，然后对其进行向量量化，以提取代表性子集。在离线阶段，LLM将每个选定轨迹的潜在行为模式与自然语言描述关联起来，使其成为一种工具。在在线阶段，LLM根据当前状态和目标选择合适的工具。低层控制由执行与工具相关的轨迹的原始动作处理。从代理人工智能的角度来看，这种方法将学习分为两个层次：工具发现通过对轨迹的无监督量化进行处理，而推理和决策则由LLM处理。我们在一个部分可观察的动态二维网格环境中测试了该方法，使用开放的视觉-语言模型（Qwen3.6-35B-A3B）。将几何派生的工具库与以代理为中心的缩放工具和碰撞检测工具配对，使得快速的非推理配置能够匹配成本更高的思维链版本的目标达成率，同时将决策成本从分钟减少到秒。
cs.AI / 33 / 2610.00636
CompMat-Bench: Benchmarking AI Agents for Computational Materials Science
CompMat-Bench：计算材料科学中人工智能代理的基准测试
Abstract
Evaluating AI agents on scientific research tasks is constrained by the time and resources required for the underlying experiments or calculations. In computational materials research, repeating the same expensive simulations across agents and trials can make evaluation impractical. We introduce CompMat-Bench, a benchmark of 94 tasks derived from recently published computational materials studies, each asking agents to complete a step toward achieving the study's scientific goal. We reproduce the research steps in advance and assess agents on preparing inputs and analyzing outputs for expensive simulations, so expensive simulations can be avoided during evaluation. The reproduced inputs and results serve as ground truth for grading agents with fixed rules, without an LLM judge. The benchmark supports four evaluation conditions: single tasks and workflows composed of related tasks, each with full or reduced methodological guidance. With full guidance on single tasks, agents based on three LLMs demonstrate the ability to complete individual materials research steps, with pass rates of 66.0-90.4% across 94 tasks. Both longer workflows and reduced guidance can limit agent performance, but in different ways for different agents: they lower the pass rates of the weaker agents, whereas the strongest agent falls only when a long workflow is combined with reduced guidance. Failure analysis attributes most failures to scientific errors rather than to errors in software usage. CompMat-Bench provides a basis for comparing agents on the steps of real materials research and for analyzing agent failure modes.
Chinese Translation
在科学研究任务中评估人工智能代理受到基础实验或计算所需时间和资源的限制。在计算材料研究中，在不同代理和试验中重复相同的昂贵模拟可能使评估变得不切实际。我们引入了CompMat-Bench，这是一个基于最近发表的计算材料研究衍生出的94个任务的基准，每个任务要求代理完成朝着研究科学目标迈进的一步。我们提前重现研究步骤，并评估代理在准备输入和分析昂贵模拟的输出方面的表现，从而在评估过程中避免昂贵的模拟。重现的输入和结果作为评估代理的真实标准，采用固定规则进行评分，而不需要大型语言模型（LLM）评判。该基准支持四种评估条件：单一任务和由相关任务组成的工作流程，每种条件都有完整或简化的方法指导。在单一任务的完整指导下，基于三种LLM的代理展示了完成单个材料研究步骤的能力，在94个任务中通过率为66.0%-90.4%。较长的工作流程和简化的指导可能会限制代理的表现，但对不同代理的影响方式不同：它们降低了较弱代理的通过率，而最强代理仅在长工作流程与简化指导结合时表现不佳。失败分析将大多数失败归因于科学错误，而非软件使用错误。CompMat-Bench为比较代理在真实材料研究步骤中的表现以及分析代理失败模式提供了基础。
cs.AI / 34 / 2610.00648
Incident-Arena: Getting agents to the last nine of reliability
事件竞技场：将智能体推向可靠性的最后九个阶段
Abstract
AI coding agents are ubiquitous in engineering workflows amongst industry and academia. Yet, despite their use in app coding, relatively less attention has been paid to their ability to execute on production incident response. This emerging field, termed agentic site-reliability-engineering (SRE) contains benchmarks limited by (1) unrealistic environments, typically toy repositories (2) non-standard framework implementations and (3) simple static verifiers. We introduce Incident-Arena, a human-built benchmark of 20 carefully selected tasks grounded in real-world deployed open source software. Each task deploys a production application to an ephemeral Kubernetes cluster, injecting a fault from the config layer through underlying images, and a sustained load profile given the task requirements. We also present a novel verification method, going beyond static checks to functional verifiers, holding systems level metrics stable, while ensuring repairs are done safely. Agent trials run an average of 2.81M tokens and 41 turns, going beyond existing benchmarks, demonstrating agentic long horizon reasoning. Across 20 tasks and 3 application substrates, frontier models score below 64.3%, with failures extending from diagnosis/localization errors, through incomplete repairs and unsafe regressions.
Chinese Translation
人工智能编码智能体在工业和学术界的工程工作流程中无处不在。然而，尽管它们在应用编码中的使用日益普及，但对其在生产事件响应中执行能力的关注相对较少。这个新兴领域被称为智能体站点可靠性工程（agentic site-reliability-engineering, SRE），其基准测试受到以下限制：（1）不现实的环境，通常是玩具仓库；（2）非标准框架实现；（3）简单的静态验证器。我们介绍了事件竞技场（Incident-Arena），这是一个由人类构建的基准，包含20个经过精心挑选的任务，基于现实世界中部署的开源软件。每个任务将生产应用程序部署到一个短暂的Kubernetes集群中，从配置层通过底层镜像注入故障，并根据任务要求施加持续的负载配置。我们还提出了一种新颖的验证方法，超越静态检查，采用功能验证器，保持系统级指标的稳定，同时确保修复过程的安全性。智能体试验平均运行2.81M个标记和41个回合，超越现有基准，展示了智能体的长远推理能力。在20个任务和3个应用基础上，前沿模型的得分低于64.3%，失败原因包括诊断/定位错误、不完整的修复和不安全的回归。
cs.AI / 35 / 2610.00651
Agent Evaluation Reliability: More Tasks Won't (Always) Fix An Agent Leaderboard
代理评估的可靠性：更多任务并不总能修复代理排行榜
Abstract
Agent evaluations are increasingly used to compare LLMs and inform deployment decisions, yet ranks can reflect not only the model but also the effects of the evaluation conditions such as the scaffolds or tasks. This makes reliability claim-dependent: an evaluation that reliably ranks deployed systems may not reliably rank underlying models. We ask which conclusions current agent evaluations reliably support and what additional evaluation would improve them. We develop a Bayesian variance-decomposition framework for sparse, imbalanced agent leaderboards and apply it to 22 benchmarks from the Holistic Agent Leaderboard and Harbor Index. The framework separates signal, performance differences relevant to the intended claim, from noise, irrelevant variation that can still change rankings. We find: (1) Reliability depends on the measurement goal. Fixed model-scaffold systems are ranked reliably (0.935-0.994), while underlying-model reliability is substantially lower (0.148-0.841). (2) Scaffold choice can change conclusions. Inter-scaffold reliability measures whether scaffolds preserve model rankings, showing that scaffold effects vary substantially across evaluations. (3) More tasks cannot resolve all uncertainty. Even infinitely many similarly constructed tasks improve model-ranking reliability of a benchmark by at most 0.097 when uncertainty is dominated by limited scaffold coverage. (4) Pooling diverse benchmarks can improve cross-task rankings at lower cost. For rankings across diverse agentic tasks, pooling benchmarks raises projected reliability from 0.44 to 0.75 at the same task budget and can reduce projected cost by up to 83\%. Evaluation design should follow the intended claim: identify what a score or ranking should mean, diagnose what limits its reliability, and spend evaluation budget on the sources of uncertainty that matter.
Chinese Translation
代理评估在比较大规模语言模型（LLMs）和指导部署决策中越来越多地被使用，然而排名不仅反映模型本身，还反映评估条件的影响，例如支架或任务。这使得可靠性声明依赖于具体情况：一个能够可靠排名已部署系统的评估，可能无法可靠排名基础模型。我们探讨当前代理评估能够可靠支持哪些结论，以及什么额外的评估可以改善它们。我们开发了一个贝叶斯方差分解框架，用于稀疏、不平衡的代理排行榜，并将其应用于来自整体代理排行榜（Holistic Agent Leaderboard）和港口指数（Harbor Index）的22个基准。该框架将与预期声明相关的信号、性能差异与噪声、仍然可能改变排名的无关变异分开。我们的发现包括：（1）可靠性依赖于测量目标。固定模型-支架系统的排名可靠性较高（0.935-0.994），而基础模型的可靠性则显著较低（0.148-0.841）。 （2）支架选择可能改变结论。支架间的可靠性衡量支架是否保持模型排名，显示支架效应在不同评估中存在显著差异。 （3）更多任务无法解决所有不确定性。即使有无限多的类似构建任务，当不确定性主要由有限的支架覆盖主导时，基准的模型排名可靠性最多只能提高0.097。 （4）汇总多样化基准可以在较低成本下改善跨任务排名。对于多样化的代理任务，汇总基准可以在相同任务预算下将预期可靠性从0.44提高到0.75，并且可以将预期成本降低多达83%。评估设计应遵循预期声明：明确分数或排名应代表的含义，诊断限制其可靠性的因素，并将评估预算花费在重要的不确定性来源上。
cs.AI / 36 / 2610.00654
When More Data Is Not Enough: The Context-Sufficiency Frontier in Generative AI Personalization
当更多数据不足够时：生成式人工智能个性化中的情境充分性边界
Abstract
Personalization has long relied on customer data to infer what an individual is likely to value. We call this customer evidence: the customer's historical behavior and preferences. Generative AI extends personalization by allowing providers to supply changing situational information at the moment a response is produced, without encoding every condition in advance. We define this provider-side context as information about what is possible, permitted, or advisable now. This flexibility creates a new problem: once context becomes easy to supply, more is not necessarily better. We develop a theory of context sufficiency in which the relevance of context to the customer's current intent matters more than its volume. The theory identifies four states, insufficiency, sufficiency, saturation, and interference, and introduces the Context-Sufficiency Frontier to locate the minimal relevant set. In a full-factorial experiment with a generative recommender at a large home-furnishing retailer, relevant context improved appropriateness, while irrelevant context reduced it and destabilized retrieval. The framework shifts personalization from supplying more context toward identifying what the current interaction actually requires and enforcing constraints throughout the service process.
Chinese Translation
个性化长期以来依赖于客户数据来推断个体可能重视的内容。我们称之为客户证据：客户的历史行为和偏好。生成式人工智能通过允许提供者在生成响应时提供变化的情境信息，从而扩展了个性化，而无需提前编码每一个条件。我们将这种提供者侧的情境定义为关于当前可能、允许或建议的信息。这种灵活性带来了一个新问题：一旦情境变得容易提供，更多并不一定更好。我们发展了一个情境充分性的理论，其中情境与客户当前意图的相关性比其数量更为重要。该理论识别了四种状态：不足、充分、饱和和干扰，并引入了情境充分性边界以定位最小相关集。在与一家大型家居用品零售商的生成式推荐系统进行的全因子实验中，相关情境提高了适宜性，而无关情境则降低了适宜性并使检索不稳定。该框架将个性化的重点从提供更多情境转向识别当前交互的实际需求，并在整个服务过程中施加约束。
cs.AI / 37 / 2610.00663
Backdoor Containment via Expert Quarantine and Shutdown in LLMs
通过专家隔离和关闭实现大语言模型的后门控制
Abstract
Backdoored large language models (LLMs) can behave normally on benign inputs while producing attacker-specified outputs under hidden triggers. Existing defenses span four stages--prior-training, in-training, post-training, and inference-time--and share one of two underlying strategies: either suppress backdoor learning (by filtering poisoned data or interrupting its acquisition during optimization) or learn, then purify (by repairing model weights or gating inputs after a fully backdoored model has formed). We propose a third strategy, learn, but channel: allow backdoor formation during training but route it into a designated, quarantined component that can be disabled at deployment. To this end, we propose Quarantined Expert Shutdown QES, a computationally efficient containment strategy built in a regularization-steered MoE-like setting. Specifically, given a poisoned dataset, QES augments a Transformer-based language model with routed expert-specific LoRA branches and lightweight routers, and uses auxiliary routing objectives to attract trigger-conditioned behavior into a designated expert while preserving benign capability elsewhere. At deployment, mitigation reduces to a single constant-time operation: zeroing the quarantined expert's routing weight, without trigger screening or further updating model weights. Empirically, our methods reduce the attack success rate ASR from 100% to 0-10% on most settings across two tasks, three attacks, and four model families, while downstream utility is often preserved or only modestly affected. These results establish learn, but channel as a previously unexplored regime for backdoor containment in generative LLMs.
Chinese Translation
带有后门的大语言模型（LLMs）在正常输入下可以表现正常，但在隐藏触发器的作用下会产生攻击者指定的输出。现有的防御措施涵盖四个阶段——训练前、训练中、训练后和推理时——并共享两种基本策略之一：要么抑制后门学习（通过过滤受污染的数据或在优化过程中中断其获取），要么学习后再净化（通过修复模型权重或在完全形成后门模型后对输入进行门控）。我们提出第三种策略：学习，但引导；允许在训练过程中形成后门，但将其引导到一个指定的、隔离的组件中，该组件可以在部署时禁用。为此，我们提出了隔离专家关闭（Quarantined Expert Shutdown, QES），这是一种在正则化引导的类似MoE（Mixture of Experts）设置中构建的计算高效的控制策略。具体而言，给定一个受污染的数据集，QES通过引导专家特定的LoRA分支和轻量级路由器来增强基于Transformer的语言模型，并使用辅助路由目标将触发条件行为引导到指定的专家，同时在其他地方保留正常能力。在部署时，缓解措施简化为一个常数时间操作：将隔离专家的路由权重归零，而无需触发筛选或进一步更新模型权重。实证结果表明，我们的方法在两个任务、三种攻击和四个模型系列的绝大多数设置中将攻击成功率（ASR）从100%降低到0-10%，而下游效用通常得以保留或仅受到轻微影响。这些结果确立了学习但引导作为生成性LLMs中后门控制的一个前所未有的研究领域。
cs.AI / 38 / 2610.00668
A Simple Doxastic Deontic Logic for Norm-Guided Decision Making
一种简单的信念义务逻辑用于规范引导的决策制定
Abstract
Making decisions despite conflicting norms and incomplete or unreliable information is a fundamental challenge for autonomous systems. We introduce a simple doxastic deontic logic for this setting: a classically reducible fragment of Chellas' Minimal Deontic Logic, extended with explicit conditional norms and combined with multi-agent KD45, so that norms can depend on agents' beliefs about both facts and norms. On this logic we define the Doxastic Norm Compliance Optimization Problem, where an agent chooses a decision minimizing weighted norm violations. We distinguish subjective optimization (relative to the agent's beliefs) from objective optimization (relative to the actual facts). We give conditions under which (i) the two coincide and (ii) optimal decision-making can be reduced to weighted partial MaxSAT in polynomial time.
Chinese Translation
在面临冲突规范以及不完整或不可靠信息的情况下做出决策是自主系统面临的一个基本挑战。我们为这一情境引入了一种简单的信念义务逻辑：这是切拉斯（Chellas）最小义务逻辑的一个经典可约片段，扩展了显式条件规范，并结合了多智能体的KD45，使得规范可以依赖于智能体对事实和规范的信念。在此逻辑基础上，我们定义了信念规范合规优化问题（Doxastic Norm Compliance Optimization Problem），其中智能体选择一个决策，以最小化加权的规范违反。我们区分了主观优化（相对于智能体的信念）与客观优化（相对于实际事实）。我们给出了在何种条件下（i）两者相符以及（ii）最优决策可以在多项式时间内简化为加权部分MaxSAT的问题。
cs.AI / 39 / 2610.00682
Ontology-Grounded, Reasoner-Verified Benchmarks for Evaluating LLM Reasoning in Scientific AI
基于本体的、经过推理器验证的基准测试，用于评估科学人工智能中的大语言模型推理能力
Abstract
Large language models (LLMs) increasingly underpin scientific AI applications that reason over structured knowledge, from biomedical question answering to materials informatics. However, their logical reasoning often falls short, producing factual inaccuracies unacceptable in these settings. Reliable evaluation remains challenging: manual dataset construction scales poorly, and LLM-based generation risks embedding the very flaws it aims to measure. High-quality benchmarks must ground both correct and incorrect labelled examples in explicit background knowledge, formally verifiable by a standard reasoner. We propose a pipeline that automatically generates ontology-grounded multiple-choice question (MCQ) benchmarks from any sufficiently axiomatised OWL 2 ontology, with correct answers grounded in the ontology by design. Distractors are generated by perturbing the right-hand-side class expressions of class definition axioms, and their incorrectness is formally verified by an OWL reasoner via entailment checks. We evaluate the pipeline on three ontologies: Pizza (small, academic), PMDco (complex, materials science), and DOID (large, biomedical), generating 112, 2,491, and 15,216 MCQs respectively. Distractors span four semantic categories from class unsatisfiability to weakened subsumptions, enabling diagnostic evaluation of specific reasoning failures. Items meet natural language quality standards: mean LLM judge scores of 4.02, 4.36, and 3.36 out of 5 confirm fluency, and correct-answer-to-distractor similarity above 0.8 shows that wrong options cannot be dismissed on surface form alone. Six LLMs evaluated zero-shot achieve 41.1-76.8% accuracy, well above the 25% random-guessing baseline, indicating the benchmarks are challenging and discriminative. This work is a step towards more reliable benchmarks for assessing logical reasoning in scientific AI.
Chinese Translation
大型语言模型（LLMs）日益成为科学人工智能应用的基础，这些应用在结构化知识上进行推理，从生物医学问答到材料信息学。然而，它们的逻辑推理能力常常不足，在这些场景中产生不可接受的事实不准确性。可靠的评估仍然具有挑战性：手动数据集构建扩展性差，而基于LLM的生成则有可能嵌入其旨在测量的缺陷。高质量的基准测试必须将正确和错误的标记示例基于明确的背景知识进行构建，并由标准推理器进行形式验证。我们提出了一种管道，能够从任何足够公理化的OWL 2本体自动生成基于本体的多项选择题（MCQ）基准测试，正确答案在设计上基于本体。干扰项通过扰动类定义公理的右侧类表达式生成，其错误性通过OWL推理器通过蕴涵检查进行形式验证。我们在三个本体上评估了该管道：Pizza（小型，学术）、PMDco（复杂，材料科学）和DOID（大型，生物医学），分别生成112、2491和15216个MCQ。干扰项跨越四个语义类别，从类的不满足性到减弱的子概念关系，使得特定推理失败的诊断评估成为可能。项目符合自然语言质量标准：LLM评审的平均得分分别为4.02、4.36和3.36（满分5分），确认了流畅性，而正确答案与干扰项的相似度超过0.8，表明错误选项不能仅凭表面形式被排除。六个零样本评估的LLM的准确率达到41.1-76.8%，远高于25%的随机猜测基线，表明这些基准测试具有挑战性和区分性。这项工作是朝着更可靠的基准测试评估科学人工智能中的逻辑推理能力迈出的重要一步。
cs.AI / 40 / 2610.00685
Backdoor Purification for LoRA-Tuned LLMs via Null-Space Projection
通过零空间投影对LoRA调优的大型语言模型进行后门净化
Abstract
With the rapid adoption of large language models (LLMs) and parameter-efficient fine-tuning (PEFT) methods, the risk of backdoor attacks has become more severe. Existing backdoor purification methods typically rely on at least one of the strong assumptions, such as prior knowledge of triggers, access to clean references, or aggressive retraining, and they often lack comprehensive evaluations. These constraints substantially limit their practical applicability. To overcome these challenges, our work proposes purifying LoRA-tuned LLMs without these assumptions and even without post-hoc retraining of the suspect parameters. Our objective is to significantly reduce the attack success rates (ASR) while preserving both (i) the base model's general capabilities and (ii) the new downstream skills learned through the adapter. Through a series of ablation studies, we progressively scale our approach from a single layer in a text classification setting to a full-parameter LLM in the generative task. Through careful data curation and feature approximation, we extract high-fidelity backdoor directions and, for each layer or head, construct orthogonal null spaces in both the input and output channels, onto which the LoRA updates are projected. Empirically, our null-space projection method reduces the ASR from nearly 100% to less than 10%, while preserving the base model's benign performance and the adapter's learned abilities during downstream task adaptation.
Chinese Translation
随着大型语言模型（LLMs）和参数高效微调（PEFT）方法的快速普及，后门攻击的风险变得更加严重。现有的后门净化方法通常依赖于至少一个强假设，例如对触发器的先验知识、对干净参考的访问或激进的再训练，并且往往缺乏全面的评估。这些限制显著限制了它们的实际适用性。为了克服这些挑战，我们的工作提出了一种在没有这些假设的情况下净化LoRA调优的LLMs的方法，甚至不需要对可疑参数进行事后再训练。我们的目标是显著降低攻击成功率（ASR），同时保持（i）基础模型的通用能力和（ii）通过适配器学习的新下游技能。通过一系列消融研究，我们逐步将我们的方法从文本分类设置中的单层扩展到生成任务中的全参数LLM。通过仔细的数据策划和特征近似，我们提取高保真度的后门方向，并为每一层或头构建输入和输出通道中的正交零空间，将LoRA更新投影到这些空间上。实证结果表明，我们的零空间投影方法将ASR从近乎100%降低到10%以下，同时在下游任务适应过程中保持基础模型的良性性能和适配器学习的能力。
cs.AI / 41 / 2610.00700
R-GroundBench: A Diagnostic Benchmark for R-Group Groundingin Markush Molecular Editing
R-GroundBench：用于 R-组基础的 Markush 分子编辑的诊断基准
Abstract
Recent advances in AI for scientific discovery enable molecular understandingand design, yet reasoning over incomplete chemical representations remainsunclear.Markush structures, which encode molecular families through variable R-groupplaceholders (\textit{R\textsubscript{1}}, \textit{R\textsubscript{2}}, \textit{X}, etc.), are ubiquitous in pharmaceutical patents and requiregrounding across molecular, textual, and chemical information.However, existing molecule-language benchmarks focus on fully specifiedmolecules, leaving R-group grounding largely unevaluated.We introduce R-GroundBench:, a diagnostic benchmark built from real patent Markushstructures, featuring a Multiple-Choice (VQA) track with controlled difficultyand modality splits, and an open-ended Generation track.Our results reveal a substantial gap between recognition andmolecular grounding.While models achieve over 90\% accuracy on Easy VQA, performance drops to56--66\% on Hard VQA when shortcuts are controlled.Chemical-domain VLMs also remain unreliable, achieving only 25.7--46.2\% on HardVQA despite domain-specific pretraining.Moreover, Generation Exact Match remains below 20\% for most models and below8\% when visual input is required.These findings reveal that current AI systems lack reliable grounding andexecution for Markush editing, highlighting challenges for AI-drivenscientific discovery.
Chinese Translation
近期在科学发现领域的人工智能（AI）进展使得分子理解和设计成为可能，但对不完整化学表征的推理仍然不够清晰。Markush 结构通过可变的 R-组占位符（如 R extsubscript{1}、R extsubscript{2}、X 等）编码分子家族，广泛存在于药物专利中，并需要在分子、文本和化学信息之间进行基础化。然而，现有的分子-语言基准主要集中在完全指定的分子上，导致 R-组基础化的评估几乎没有。我们介绍了 R-GroundBench，这是一个基于真实专利 Markush 结构构建的诊断基准，包含一个具有控制难度和模态划分的多项选择（VQA）轨道，以及一个开放式生成轨道。我们的结果揭示了识别与分子基础化之间存在显著差距。尽管模型在简单 VQA 上的准确率超过 90%，但在控制捷径后，困难 VQA 的表现下降至 56%--66%。化学领域的 VLM（视觉语言模型）也表现不可靠，在困难 VQA 上的准确率仅为 25.7%--46.2%，尽管进行了领域特定的预训练。此外，对于大多数模型，生成的精确匹配率仍低于 20%，而在需要视觉输入时则低于 8%。这些发现表明当前的 AI 系统在 Markush 编辑方面缺乏可靠的基础化和执行能力，突显了 AI 驱动的科学发现面临的挑战。
cs.AI / 42 / 2610.00705
Meta-Multi-Agent Reinforcement Learning for Fast Adaptation of Interactive Policies with Applications to Autonomous Driving
用于交互策略快速适应的元多智能体强化学习及其在自动驾驶中的应用
Abstract
This paper develops a meta-multi-agent reinforcement learning (meta-MARL) framework to enable fast adaptation of interactive policies in a multi-agent system (MAS). Meta-reinforcement learning (meta-RL) enables agents to rapidly adapt to new tasks/environments using a bi-level optimization mechanism. However, existing meta-RL generally focuses on single-agent systems. Extending these frameworks and algorithms to multi-agent systems poses additional challenges, as tasks are characterized by not only the environment but also agents' strategic interactions. To address these challenges, we model multi-agent reinforcement learning (MARL) problems as Markov games (MGs) and develop a meta-MARL framework for rapid interactive policy adaptation across a distribution of MGs. A new concept, called meta-NE, is defined to describe the desired solution concept in a meta-MARL problem. Sufficient conditions for the equivalence between a meta-NE and a stationary point of the gradient-play-based meta-MARL algorithm are established. Our evaluation on autonomous-driving tasks demonstrates that the proposed meta-MARL method achieves faster adaptation than pretrained MARL baselines, validating the effectiveness of our framework.
Chinese Translation
本文开发了一种元多智能体强化学习（meta-MARL）框架，以实现多智能体系统（MAS）中交互策略的快速适应。元强化学习（meta-RL）使得智能体能够通过双层优化机制迅速适应新的任务/环境。然而，现有的元强化学习通常集中于单智能体系统。将这些框架和算法扩展到多智能体系统面临额外的挑战，因为任务不仅由环境特征决定，还受到智能体战略互动的影响。为了解决这些挑战，我们将多智能体强化学习（MARL）问题建模为马尔可夫博弈（MG），并开发了一种元-MARL框架，以便在一系列MG中快速适应交互策略。我们定义了一个新的概念，称为meta-NE，用以描述元-MARL问题中所需的解概念。我们建立了meta-NE与基于梯度博弈的元-MARL算法的平稳点之间等价的充分条件。我们在自动驾驶任务上的评估表明，所提出的meta-MARL方法实现了比预训练的MARL基线更快的适应，验证了我们框架的有效性。
cs.AI / 43 / 2610.00710
ReLiveGym: Evaluating Long-Lived Agents over Weeks of Replayed Reality
ReLiveGym：评估长寿命智能体在重播现实中的数周表现
Abstract
As large language model (LLM) agents become widely adopted, they are increasingly deployed for tasks that require persistent monitoring or recurring actions (e.g., market analysis). These agents are expected to operate unattended for days or weeks, act at the right timing, and adapt to the dynamic environment over time. These challenges are not fully captured in the existing long-horizon agent work, as they often consider a static environment that is not temporally changing. We introduce ReLiveGym, a diagnostic evaluation environment of long-lived tasks in which agents act sparsely over simulated weeks of chronologically replayed real-world news, market, and social-media streams. The tasks span diverse levels of time sensitivity, reasoning intensity, and recurrence. Across eight base language models, we investigate how model choice and harness design affect agent performance on such long-lived tasks. Our results show that how agents determine when to act arises as an important harness-design axis for long-lived tasks; and that the optimal design varies across tasks and sometimes model choices as well. We also evaluate how continuous learning from hindsight feedback affects performance and addresses failure modes observed in these long-lived tasks. These findings indicate model choice, action timing mechanism, and use of feedback as important considerations in the design of long-lived agents. Code: https://github.com/SaharaLabsAI/ReLiveGym
Chinese Translation
随着大型语言模型（LLM）智能体的广泛应用，它们越来越多地被部署于需要持续监控或重复行动的任务中（例如市场分析）。这些智能体被期望能够在无人值守的情况下持续运行数天或数周，能够在适当的时机采取行动，并随着时间的推移适应动态环境。这些挑战在现有的长时间跨度智能体研究中并未得到充分体现，因为这些研究通常考虑的是一个静态的、没有时间变化的环境。我们引入了ReLiveGym，这是一个用于长寿命任务的诊断评估环境，在该环境中，智能体在模拟的数周内稀疏地对按时间顺序重播的现实世界新闻、市场和社交媒体流进行操作。这些任务涵盖了不同的时间敏感性、推理强度和重复性。在八种基础语言模型中，我们研究了模型选择和设计如何影响智能体在这些长寿命任务上的表现。我们的结果表明，智能体确定何时采取行动是长寿命任务中一个重要的设计轴心；而且最佳设计在不同任务和有时在模型选择上也有所不同。我们还评估了从事后反馈中进行持续学习如何影响表现，并解决在这些长寿命任务中观察到的失败模式。这些发现表明，模型选择、行动时机机制和反馈使用是设计长寿命智能体时的重要考虑因素。代码：https://github.com/SaharaLabsAI/ReLiveGym
cs.AI / 44 / 2610.00715
Robust Nash Alignment under Preference Uncertainty
偏好不确定性下的鲁棒纳什对齐
Abstract
Preference-based alignment methods typically optimize against a single preference model, and can therefore be brittle when pairwise preferences are uncertain: noisy, heterogeneous, or shift after deployment. To address these issues, we propose Robust Nash Alignment, a game-theoretic framework for alignment to uncertain pairwise preferences. Our formulation has a major learner seeking a policy with a large worst-case win rate against both an adversarial competitor and any preference kernel lying in an ambiguity set around a nominal preference. When the ambiguity set captures the uncertainty in preferences, the resulting robust objective of the game directly yields a certified lower bound on worst-case performance. However, we note this problem is computationally challenging to optimize, and to address this, we introduce a four-player primal-dual proxy game involving the leader policy, follower policy, adversarial kernel, and dual variable, and develop a single-loop optimistic mirror descent-ascent algorithm for it. We show that the proxy always lower-bounds the truncated hard-constrained objective, quantify the proxy-to-hard gap, and characterize an exactness condition under which the proxy recovers the robust objective. We then prove an \(\mathcal{O}(1/\sqrt{T})\) average-iteration convergence for the proxy-game duality gap, which implies a near-optimal robust policy for the original robust objective. Experiments on controlled tabular games and LLM alignment with uncertain preference further validate the convergence theory and show improved performance over nominal baselines.
Chinese Translation
基于偏好的对齐方法通常针对单一偏好模型进行优化，因此在成对偏好不确定时（如噪声、异质性或部署后发生变化）可能表现脆弱。为了解决这些问题，我们提出了鲁棒纳什对齐（Robust Nash Alignment），这是一个针对不确定成对偏好的博弈论框架。我们的公式化设定了一个主要学习者，旨在寻找一个在对抗竞争者和任何位于名义偏好周围模糊集中的偏好核下具有较大最坏情况胜率的策略。当模糊集捕捉到偏好的不确定性时，游戏的鲁棒目标直接提供了最坏情况性能的认证下界。然而，我们注意到这个问题在计算上具有挑战性，为了解决这一问题，我们引入了一个涉及领导者策略、跟随者策略、对抗核和对偶变量的四人原始-对偶代理游戏，并为其开发了一个单循环乐观镜像下降-上升算法。我们证明了代理总是低于截断硬约束目标的下界，量化了代理与硬约束之间的差距，并表征了一个精确条件，在该条件下代理恢复鲁棒目标。然后，我们证明了代理游戏对偶间隙的平均迭代收敛性为 O(1/ 00 ext{T})，这意味着对于原始鲁棒目标存在近似最优的鲁棒策略。在控制的表格游戏和具有不确定偏好的大语言模型（LLM）对齐实验中进一步验证了收敛理论，并显示出相较于名义基线的性能提升。
cs.AI / 45 / 2610.00791
Enterprise Representation Simplification (ERS): Reducing Representational Complexity for Enterprise AI
企业表示简化（ERS）：降低企业人工智能的表示复杂性
Abstract
Enterprise information is represented through artifacts shaped by applications, projects, technologies, organizational boundaries, and local requirements. These structures accumulate over time, creating representational complexity that must be maintained by the enterprise and interpreted by information consumers and AI systems. This paper introduces Enterprise Representation Simplification (ERS) as reducing unnecessary representational complexity while preserving required information within a defined scope, and Enterprise Representation Complexity (ERC), a representation-neutral model for comparing complexity across representation states. ERC characterizes representational extent through four dimensions: Representation Objects, Interactions, Behaviors, and Supporting Sources. Objects, Interactions, and Behaviors form dependent categories, while Supporting Sources characterize representation exposure. ERC is defined at representation and task levels, enabling comparison and distinguishing architectural simplification from retrieval optimization. The paper develops two consequences of ERS. First, representational structures create lifecycle obligations for maintenance, governance, dependencies, change, enhancement, and operation. An economic model distinguishes recurring global representation cost, recurring task-level cost, and one-time transformation cost, enabling evaluation over a defined time horizon. Second, reductions in task-level ERC reduce the representational extent an AI system must identify, relate, and interpret. Text-to-SQL research provides evidence that reduced schema and reasoning complexity can improve reasoning accuracy. ERC is not a universal complexity, performance, or cost metric. It provides measurable architectural variables for comparing representational alternatives, transformation effects, economic outcomes, and AI reasoning performance.
Chinese Translation
企业信息通过应用、项目、技术、组织边界和地方需求所形成的工件进行表示。这些结构随着时间的推移而积累，形成了必须由企业维护并由信息消费者和人工智能系统解读的表示复杂性。本文提出了企业表示简化（ERS），旨在在保留所需信息的前提下，减少不必要的表示复杂性，并引入企业表示复杂性（ERC），这是一个表示中立的模型，用于比较不同表示状态下的复杂性。ERC通过四个维度来表征表示范围：表示对象、交互、行为和支持源。对象、交互和行为形成相互依赖的类别，而支持源则表征表示的暴露程度。ERC在表示和任务层面上进行定义，使得比较成为可能，并区分架构简化与检索优化。本文发展了ERS的两个后果。首先，表示结构为维护、治理、依赖、变更、增强和操作创造了生命周期义务。一个经济模型区分了重复的全球表示成本、重复的任务级成本和一次性转型成本，从而能够在定义的时间范围内进行评估。其次，任务级ERC的减少降低了人工智能系统必须识别、关联和解读的表示范围。文本到SQL的研究提供了证据，表明减少模式和推理复杂性可以提高推理准确性。ERC并不是一个普遍的复杂性、性能或成本指标。它提供了可测量的架构变量，用于比较表示替代方案、转型效果、经济结果和人工智能推理性能。
cs.AI / 46 / 2610.00797
Sapien: A Stateful Policy Engine for Autonomous AI Agents
Sapien：一种用于自主AI代理的有状态策略引擎
Abstract
Contextual security defenses prevent AI agents from taking rogue actions by synthesizing a task-specific policy and enforcing it on the agent's tool calls. In multi-step tasks, however, which actions are valid often depends on what the agent has already done and learned. We present Sapien, a policy engine for enforcing stateful contextual policies. A Sapien policy specifies permitted tool-call sequences using a regular expression extended with stateful predicates, deferred policy generation, and scoped semantic checks. We show that Sapien stays within a few percent of an unconstrained agent's utility. Even if the agent is fully hijacked, Sapien's policies rule out 93-95% of attacks on AgentDojo and 62-85% on Toolathlon (twice as many as tool allowlists on long-horizon tasks).
Chinese Translation
上下文安全防御通过合成特定任务的策略并在代理的工具调用上强制执行，防止AI代理采取不当行动。然而，在多步骤任务中，哪些行动是有效的往往取决于代理已经完成和学习的内容。我们提出了Sapien，一种用于强制执行有状态上下文策略的策略引擎。Sapien策略使用扩展了有状态谓词的正则表达式来指定允许的工具调用序列，并支持延迟策略生成和范围语义检查。我们展示了Sapien的效用在几乎与不受限制的代理相当的范围内。即使代理被完全劫持，Sapien的策略也能排除93-95%的AgentDojo攻击和62-85%的Toolathlon攻击（在长时间任务中是工具白名单的两倍）。
cs.AI / 47 / 2610.00834
Kepler: Auditable World Models for ARC-AGI-3
Kepler：可审计的ARC-AGI-3世界模型
Abstract
ARC-AGI-3 evaluates agents in interactive environments whose rules and objectives must be inferred from observation. We present Kepler, an open-source harness that represents hypotheses as executable world models and validates them through retrospective transition checks and conditional prediction checks. Under one frozen Claude Opus 5 configuration, Kepler obtained a server-verified 100.00 RHAE on all 25 public games, with no per-game model selection or score-conditioned reruns. On 181 of 183 completed levels, the final Opus attempt used no more actions than the corresponding median-human baseline. The retained board runs used 8,256 environment actions, of which 7,292 occurred in scored levels. Retained local provider-session records yield 858.0 million tokens, 97.37% cache reads, and a \$777.72 cost at September 1, 2026 API list-equivalent rates. We also report three evaluation failures: source-code leakage that produced an invalid perfect run, agents reconstructing a removed harness in a control condition, and autonomous repair masking a broken planner. A single-game observation case study showed that animation frames contained task-relevant information absent from settled text grids. Across the final Claude Opus 5 and GPT-5.6 Sol boards, 48 of 50 game-model cells reached 100. These results indicate that public-set score alone has limited discriminative value and motivate first-attempt, cost-conditioned, and verification-aware reporting.
Chinese Translation
ARC-AGI-3评估在交互环境中运行的智能体，其规则和目标必须通过观察推断。我们提出了Kepler，一个开源工具，能够将假设表示为可执行的世界模型，并通过回溯转移检查和条件预测检查对其进行验证。在一个冻结的Claude Opus 5配置下，Kepler在所有25个公共游戏中获得了服务器验证的100.00 RHAE，且没有进行逐游戏模型选择或基于得分的重跑。在183个完成的关卡中，有181个的最终Opus尝试所用的动作数量不超过相应的中位数人类基线。保留的棋盘运行使用了8,256个环境动作，其中7,292个发生在得分关卡中。保留的本地提供者会话记录产生了8.58亿个令牌，97.37%的缓存读取，以及在2026年9月1日API列表等值费率下的777.72美元成本。我们还报告了三次评估失败：源代码泄露导致无效的完美运行，智能体在控制条件下重建了一个被移除的工具，以及自主修复掩盖了一个损坏的规划器。单游戏观察案例研究表明，动画帧包含了在固定文本网格中缺失的与任务相关的信息。在最终的Claude Opus 5和GPT-5.6 Sol棋盘中，50个游戏模型单元中有48个达到了100。这些结果表明，公共集得分本身的区分价值有限，并激励了首次尝试、成本条件和验证意识的报告。
cs.AI / 48 / 2610.00849
Learning Multiple Timescales for Goal-Conditioned Reinforcement Learning
学习多时间尺度的目标条件强化学习
Abstract
Existing approaches to offline goal-conditioned reinforcement learning (GCRL) struggle with long-horizon tasks. Discounting shrinks value differences between distant states until they fall below the function approximation error, leaving the agent with no signal for ranking states. Temporal abstraction, which treats k environment steps as a single transition, restores this signal at long range, but no single fixed k suits all state-goal distances: large k preserves value differences across long temporal distances while collapsing distinctions between nearby states, and small k does the reverse. We make this trade-off explicit and introduce Generalized Implicit Temporal Abstraction (GITA), which conditions a single value function on k. GITA trains one policy by aggregating advantage-weighted supervision across multiple k values, so scales assigning larger positive advantages to a state-goal pair contribute more strongly to its update. GITA does not need to choose between local resolution and long-range signal; it retains both without committing to a single k. On OGBench, GITA outperforms a broad range of offline GCRL baselines, raising average success rate across all tasks by 25 percentage points (73% relative improvement) over HIQL. It also improves over the strongest fixed-k method, OTA, by 7 percentage points (14% relative).
Chinese Translation
现有的离线目标条件强化学习（GCRL）方法在长时间跨度任务中表现不佳。折扣因子会缩小远距离状态之间的价值差异，直到这些差异低于函数逼近误差，从而使代理无法获得用于状态排序的信号。时间抽象将 k 个环境步骤视为一个单一的转变，能够在长时间范围内恢复这一信号，但没有一个固定的 k 值适用于所有状态-目标距离：较大的 k 值在长时间距离上保留了价值差异，同时压缩了邻近状态之间的区分，而较小的 k 值则反之。我们明确了这一权衡，并引入了广义隐式时间抽象（Generalized Implicit Temporal Abstraction, GITA），该方法将单一价值函数条件化于 k。GITA 通过在多个 k 值之间聚合优势加权监督来训练一个策略，从而使得分配给状态-目标对的较大正优势在更新中贡献更强。GITA 不需要在局部分辨率和长距离信号之间做出选择；它在不承诺于单一 k 的情况下同时保留了两者。在 OGBench 上，GITA 超越了广泛的离线 GCRL 基线，使所有任务的平均成功率提高了 25 个百分点（相对提升 73%），相较于 HIQL。它还在最强的固定 k 方法 OTA 上提高了 7 个百分点（相对提升 14%）。
cs.AI / 49 / 2610.00870
An Educator-Guided LLM Pedagogical Agent for Scaffolded Feedback in Conceptual Database Design
一个教育者引导的 LLM 教学代理用于概念数据库设计中的分层反馈
Abstract
We present an educator-guided LLM pedagogical agent for scaffolded feedback in conceptual database design. Integrated into an entity--relationship diagram (ERD) editor, the system grounds feedback in the student artifact, assignment requirements, educator-authored rubrics, and instructional resources. Its architecture separates hidden, artifact-grounded diagnosis from the workflow that controls the form and disclosure level of student-facing support. We instantiate the architecture as a four-stage workflow progressing from concept checks and guided application to low-detail feedback and localized clarification. Each feedback request creates a stateful episode linked to versioned ERD states. In a deployment spanning three ERD environments and 383 feedback episodes, 71.1\% of observed target-level changes fully or partially incorporated the hidden diagnostic target, including many after Stages~1--2. Qualitative analysis showed that staged disclosure sometimes withheld inaccurate details, supported selective uptake, or allowed later recovery, though some errors still shaped revisions. Survey responses from a self-selected sample favored delayed disclosure and student agency but noted indirectness and repetition.
Chinese Translation
我们提出了一个教育者引导的 LLM 教学代理，用于概念数据库设计中的分层反馈。该系统集成于实体-关系图（ERD）编辑器中，基于学生的作品、作业要求、教育者编写的评分标准和教学资源提供反馈。其架构将隐藏的、基于作品的诊断与控制学生支持形式和披露级别的工作流程分开。我们将该架构实例化为一个四阶段工作流程，从概念检查和指导应用逐步推进到低细节反馈和局部澄清。每个反馈请求创建一个与版本化 ERD 状态相关联的有状态事件。在覆盖三个 ERD 环境和 383 个反馈事件的部署中，71.1% 的观察到的目标级别变化完全或部分地纳入了隐藏的诊断目标，包括许多发生在第 1 阶段和第 2 阶段之后。定性分析显示，分阶段披露有时会隐瞒不准确的细节，支持选择性采纳，或允许后续的恢复，尽管一些错误仍然影响了修订。来自自选样本的调查反馈偏向于延迟披露和学生自主性，但指出了间接性和重复性的问题。
cs.AI / 50 / 2610.00872
MemFit: Efficient Long-Term Agentic Memory
MemFit：高效的长期代理记忆
Abstract
Long-term memory systems for large language models (LLMs) have gained popularity for extending reasoning capabilities across applications. Current memory systems rely on LLM agents to organize and consolidate memory, resulting in costly, inefficient write operations. To address this limitation, we propose MemFit, a long-term memory system for conversational agents that reduces the cost and latency of memory operations. Unlike existing systems that rely on expensive LLM calls for memory construction or discard surface-level details through compression, MemFit stores each turn verbatim in an append-only store with near-instantaneous, LLM-free insertion, indexing turns with segment summaries rather than replacing them. Additionally, MemFit uses an LLM-free, multi-path retrieval strategy that combines lexical and semantic signals with cross-encoder reranking over caption- augmented episodes in both textual and multimodal settings. Empirical results on three widely used benchmarks, LoCoMo, MemGallery, and LongMemEval-S, show that MemFit achieves state-of-the-art performance while reducing memory construction time and cost several-fold, providing a scalable and efficient solution for persistent agentic memory.
Chinese Translation
大型语言模型（LLMs）的长期记忆系统因其在各类应用中扩展推理能力而受到广泛关注。目前的记忆系统依赖于LLM代理来组织和整合记忆，导致写入操作成本高且效率低下。为了解决这一限制，我们提出了MemFit，一种针对对话代理的长期记忆系统，旨在降低记忆操作的成本和延迟。与现有系统依赖于昂贵的LLM调用进行记忆构建或通过压缩丢弃表面细节不同，MemFit将每个回合逐字存储在一个仅追加的存储中，允许近乎即时且无LLM的插入，并通过段摘要对回合进行索引，而不是替换它们。此外，MemFit采用无LLM的多路径检索策略，将词汇和语义信号与跨编码器重排序相结合，适用于文本和多模态设置中的增强情节。对三个广泛使用的基准测试LoCoMo、MemGallery和LongMemEval-S的实证结果表明，MemFit在减少记忆构建时间和成本数倍的同时，达到了最先进的性能，为持久的代理记忆提供了一种可扩展且高效的解决方案。
cs.AI / 51 / 2610.00906
ActiveSaddler: Automated Curriculum Learning for Agent Harness Optimization
ActiveSaddler：用于代理工具优化的自动化课程学习
Abstract
Automated harness optimization can substantially improve LLM agents by iteratively updating their prompts, tool interfaces, and control logic from execution feedback. However, existing methods primarily optimize how the harness is updated while largely fixing which training scenarios generate the feedback that drives those updates. As the harness evolves, the scenarios most useful for further optimization can change, suggesting that the training curriculum itself should adapt alongside the harness. We formulate this missing dimension of harness optimization as an automated curriculum learning problem and introduce ActiveSaddler. ActiveSaddler models the evolving curriculum as a non-stationary bandit with dynamically instantiated optimization targets. It abstracts recurring failures into reusable failure-pattern arms, estimates the potential learning progress from further targeting each pattern, and adaptively balances revisiting known weaknesses with exploring unseen scenarios for new ones. Optimization outcomes continually update both the set of discovered failure patterns and their priorities, allowing the curriculum to co-evolve with the harness. Experiments on GAIA2 and Terminal-Bench 2.0 show that ActiveSaddler consistently discovers stronger harnesses, improving test Pass@1 by 4.4 and 7.5 percentage points over the same harness optimizer using a scenario order fixed before optimization, respectively. Ablations further show that these gains depend on dynamically constructing optimization targets, estimating their evolving utility, and balancing continued optimization with new failure discovery. Together, these results establish automated curriculum learning as a new crucial optimization dimension for harness optimization.
Chinese Translation
自动化工具优化可以通过从执行反馈中迭代更新提示、工具接口和控制逻辑，显著提升大语言模型（LLM）代理的性能。然而，现有方法主要优化工具的更新方式，而在很大程度上固定了生成驱动这些更新的反馈的训练场景。随着工具的演变，最有利于进一步优化的场景可能会发生变化，这表明训练课程本身也应与工具一起适应。我们将这一缺失的工具优化维度表述为一个自动化课程学习问题，并引入了ActiveSaddler。ActiveSaddler将不断演变的课程建模为一个具有动态实例化优化目标的非平稳赌博机。它将重复出现的失败抽象为可重用的失败模式臂，估计进一步针对每种模式的潜在学习进展，并自适应地平衡重新访问已知弱点与探索未见场景以发现新弱点。优化结果不断更新已发现的失败模式集合及其优先级，使课程能够与工具共同演化。在GAIA2和Terminal-Bench 2.0上的实验表明，ActiveSaddler始终发现更强的工具，相较于在优化前固定场景顺序的相同工具优化器，测试的Pass@1分别提高了4.4和7.5个百分点。消融实验进一步表明，这些增益依赖于动态构建优化目标、估计其不断演变的效用，以及在持续优化与新失败发现之间的平衡。综合来看，这些结果确立了自动化课程学习作为工具优化的新关键优化维度。
cs.AI / 52 / 2610.00912
OR for AI That Does OR: Routing LLMs up the Escalator inside the OSCAR Framework
为执行运筹学的人工智能提供运筹学支持：在OSCAR框架内优化大型语言模型的路由
Abstract
Large language models can translate business descriptions into optimization models, but executable code may misrepresent constraints or objectives. A solver can then return an optimal solution to the wrong problem. Even when the solution satisfies the intended operating rules, a better plan may exist. For organizations that repeatedly use optimization modeling, an LLM-based framework should produce accurate formulations at low cost and, ideally, run locally. We study how to verify improvements and allocate attempts across LLMs that differ in price and capability. We develop OSCAR (Optimization modeling by Simulator, Coder, And Reviewer), which uses an offline Simulator certified against labeled decision examples to compare candidates and continues searching beyond feasibility. We model the search for the next certified improvement as sequential decisions under unobserved difficulty: which LLMs to call and when to stop. In a simplified known-prior setting, we give conditions under which cost-ordered escalation is optimal. For general menus, we derive a prior-free competitive guarantee. On five benchmark problems, OSCAR achieves 95% to 100% accuracy at the reported settings using two small open-weight LLMs, each deployable locally on a single GPU. Their single-attempt accuracies average 29% and 48%. In five runs per problem, Codex and Claude Code incur average token costs 3.1 and 5.8 times OSCAR's, respectively. OSCAR supports open-weight models locally or in the cloud, depending on budget and confidentiality requirements. Firms should maintain labeled decision examples of feasible and infeasible decisions to clarify plain-language operating rules. OSCAR follows these labels when an LLM's interpretation conflicts with them. As LLM capabilities and prices change, OSCAR's simple operating rules and adjustable settings help firms adapt their model choices and benefit from these advances.
Chinese Translation
大型语言模型可以将商业描述转化为优化模型，但可执行代码可能会错误地表示约束或目标。求解器可能因此返回错误问题的最优解。即使解决方案满足预期的操作规则，仍可能存在更好的计划。对于重复使用优化建模的组织，基于大型语言模型的框架应以低成本生成准确的公式，并理想情况下能够在本地运行。我们研究如何验证改进并在价格和能力不同的大型语言模型之间分配尝试。我们开发了OSCAR（由模拟器、编码器和审阅者进行的优化建模），该框架使用离线模拟器，该模拟器经过标记决策示例的认证，以比较候选者，并在可行性之外继续搜索。我们将寻找下一个认证改进的过程建模为在未观察到的困难下的顺序决策：选择何时调用哪些大型语言模型以及何时停止。在一个简化的已知先验设置中，我们给出了成本排序升级最优的条件。对于一般菜单，我们推导出无先验的竞争保证。在五个基准问题上，OSCAR在报告的设置下使用两个小型开放权重的大型语言模型，达到了95%到100%的准确率，每个模型均可在单个GPU上本地部署。它们的单次尝试准确率平均为29%和48%。在每个问题的五次运行中，Codex和Claude Code的平均令牌成本分别是OSCAR的3.1倍和5.8倍。OSCAR支持本地或云端的开放权重模型，具体取决于预算和保密要求。公司应维护可行和不可行决策的标记决策示例，以明确通俗语言的操作规则。当大型语言模型的解释与这些规则冲突时，OSCAR会遵循这些标签。随着大型语言模型能力和价格的变化，OSCAR的简单操作规则和可调设置帮助公司调整其模型选择，并从这些进步中受益。
cs.AI / 53 / 2610.00917
Finding the Right Fit: Model-Harness Interactions across Agent Tasks
寻找合适的匹配：代理任务中的模型与工具的交互
Abstract
Choosing an agent system means choosing both a language model and the harness through which it acts. We ask whether a strong model, harness, or pairing stays strong when the setting changes. We evaluate 66 configurations: four configurable harnesses (OpenHands, DeepSeek Harness, PI, and openJiuwen) paired with five models on TUA-Bench, ALE-CLI, and Terminal-Bench 4, plus the native Codex-GPT and Claude Code-Claude pairings. Model rankings reverse across harnesses. On Terminal-Bench 4, Claude leads GPT by 7.94 points in OpenHands but trails it by 30.16 points in PI. For four of the five models, the best harness changes from one benchmark to another, yet some pairings hold: openJiuwen gives Kimi its highest score on all three benchmarks, by 5.61 to 11.11 points. A model's own vendor harness is not reliably its best, and higher cost does not reliably buy a higher score. On Terminal-Bench 4, GPT scores higher under PI than under DSH at less than a quarter of the cost per task. Matched trajectories suggest why fit varies. Models start almost all repairs themselves, so much depends on whether the harness hands failures back in a form the model can use. GPT does best with PI's lean scaffold, while Kimi, which often issues malformed tool calls, does best in openJiuwen. We argue that the model, the harness, and the task should be evaluated together, and we release the harness adapters, evaluation code, and all 6,204 scored trajectories at https://github.com/liyix/finding-the-right-fit and https://huggingface.co/datasets/yixuanli97/finding-the-right-fit.
Chinese Translation
选择一个代理系统意味着选择一个语言模型和其操作的工具。我们探讨在环境变化时，强大的模型、工具或其组合是否依然保持强大。我们评估了66种配置：四个可配置工具（OpenHands、DeepSeek Harness、PI和openJiuwen）与五个模型在TUA-Bench、ALE-CLI和Terminal-Bench 4上的配对，以及原生的Codex-GPT和Claude Code-Claude配对。模型排名在不同工具之间发生逆转。在Terminal-Bench 4上，Claude在OpenHands中领先GPT 7.94分，但在PI中落后30.16分。对于五个模型中的四个，最佳工具在不同基准之间变化，但某些配对保持不变：openJiuwen在所有三个基准上为Kimi提供了最高分，领先5.61到11.11分。一个模型自有的供应商工具并不总是其最佳选择，且更高的成本并不一定能换来更高的分数。在Terminal-Bench 4上，GPT在PI下的得分高于在DSH下的得分，且每个任务的成本不到四分之一。匹配的轨迹表明了适配性变化的原因。模型几乎在所有修复中都自我启动，因此很大程度上取决于工具是否以模型可以使用的形式返回失败。GPT在PI的简约框架下表现最佳，而Kimi则在openJiuwen中表现最佳，因为它经常发出格式错误的工具调用。我们认为模型、工具和任务应当一起评估，并且我们在https://github.com/liyix/finding-the-right-fit和https://huggingface.co/datasets/yixuanli97/finding-the-right-fit发布了工具适配器、评估代码和所有6204条评分轨迹。
cs.AI / 54 / 2610.00947
ABDA-NL: A Natural-Language Scenario Explorer for Argument-Based Reasoning
ABDA-NL：一种基于论证推理的自然语言场景探索器
Abstract
ABDA-NL adds a natural-language interface to ABDA, a system for argument-based discussion using ASPIC- knowledge bases under grounded semantics. Users see which conclusions are accepted, rejected, or undecided, open an interactive rendering of the grounded discussion game to learn why, explore what-if alternatives by suspending assumptions and rules or changing preferences, ask questions that are answered from a scenario's reference documents, and author new facts, assumptions, and rules in plain English. A large language model provides the bridge between language and formalism: it answers questions from the documents and the current state of the scenario, and it translates plain-English edits into candidate formal statements. The deterministic ABDA engine remains the sole source of arguments, attacks, and acceptance labels, and every proposal of the model is validated and confirmed by the user before it takes effect.
Chinese Translation
ABDA-NL 为 ABDA 系统添加了一个自然语言接口，该系统基于 ASPIC 知识库在基础语义下进行论证讨论。用户可以查看哪些结论被接受、拒绝或未决，打开一个互动的基础讨论游戏渲染，以了解原因，探索通过暂停假设和规则或改变偏好来进行的假设性替代，提出从场景参考文档中获得回答的问题，并用简单英语撰写新的事实、假设和规则。一个大型语言模型在语言与形式之间架起了桥梁：它回答来自文档和当前场景状态的问题，并将简单英语的编辑翻译成候选的正式陈述。确定性的 ABDA 引擎仍然是论证、攻击和接受标签的唯一来源，模型的每一项提议在生效之前都经过用户的验证和确认。
cs.AI / 55 / 2610.00949
PG-SFT: Balancing Capability Acquisition and Retention in Offline Agent Fine-Tuning
PG-SFT：在离线代理微调中平衡能力获取与保留
Abstract
Supervised fine-tuning (SFT) on offline agent trajectories is the standard approach for training specialized tool-using agents, but forcing models to imitate reasoning and actions token by token may harm other capabilities (e.g., general reasoning, tool calling, code generation) of the base model. In this work, we focus on studying \emph{how to better balance the trade-off between acquiring new capabilities and preserving existing ones during agent trace SFT}. By comparing several baselines in our setup, standard SFT improves the target benchmark while lowering several non-target benchmark scores; meanwhile, simply constraining distributional drift using KL penalty or limiting the update magnitude did not avoid this regression trend. Motivated by recent token-wise adaptive learning objectives, this work proposes \textbf{Privilege-Guided SFT (PG-SFT)} to leverage turn-level information gain of agent trajectories as an indicator to adjust supervision strength. PG-SFT yields a more favorable observed trade-off on the evaluated benchmarks, substantially reducing distributional drift and broad capability degradation at the cost of slight degradation in target-task performance. Our findings suggest that balancing the acquisition--retention trade-off depends not only on whether the model is anchored to its base behavior, but also on where and how strongly supervision should depart from that behavior.}
Chinese Translation
在离线代理轨迹上进行监督微调（SFT）是训练专用工具使用代理的标准方法，但强迫模型逐字模仿推理和动作可能会损害基础模型的其他能力（例如，通用推理、工具调用、代码生成）。在本研究中，我们重点研究 extit{如何更好地平衡在代理轨迹SFT过程中获取新能力与保留现有能力之间的权衡}。通过比较我们设置中的几个基线，标准SFT在提高目标基准的同时降低了多个非目标基准的得分；与此同时，仅仅通过使用KL惩罚约束分布漂移或限制更新幅度并未避免这种退化趋势。受到最近基于标记的自适应学习目标的启发，本研究提出了 extbf{特权引导SFT（PG-SFT）}，利用代理轨迹的轮次级信息增益作为调整监督强度的指标。PG-SFT在评估基准上产生了更有利的观察到的权衡，显著减少了分布漂移和广泛的能力退化，代价是目标任务性能的轻微下降。我们的研究结果表明，平衡获取与保留的权衡不仅取决于模型是否锚定于其基础行为，还取决于监督应如何以及在何处强烈偏离该行为。
cs.AI / 56 / 2610.00961
Cybernetic and Epistemic: A Missing Vocabulary for Trustworthy Agentic Delegation
控制论与认识论：值得信赖的代理委托缺失的词汇
Abstract
As code generation is increasingly delegated to AI systems, the bottleneck is shifting from writing code to supervising the systems that write it --- a shift CS-education researchers have begun to name. This shift exposes a vocabulary gap: the field asks for "human oversight" without a working distinction between the two things language does in a delegation channel --- coordinate action (cybernetic: words succeed when the world comes to match them) and coordinate understanding (epistemic: they succeed when they answer to the world and a hearer can check that they do). The failure this names is not cybernetic language but epistemic-form language doing cybernetic work: explanation-shaped output calibrated for approval rather than truth. Oversight that checks only whether an output was approved is satisfiable by rubber-stamping; oversight that holds an agent accountable requires the reasoning behind its work be retrievable and checkable. We present three delegation episodes --- illustrations, not controlled evidence --- in which epistemic engagement proved practicable while remaining auditable, one public record where a recommendation was withdrawn on its own stated terms, and one failure case illustrating oversight that requires no reasons for its discretionary choices. We propose a criterion for agentic-system governance, alongside existing technical trust properties: every consequential choice should carry the condition under which it would have gone otherwise, in a form a third party can test. Without such a condition, a third party cannot distinguish a decision from a rubber stamp. We give the criterion an operational form --- a two-part reconstruction test scoring a delegation record by whether a second reader can predict what the agent does under a perturbation --- and a deliberation-recording convention, ORRCF, that makes the condition a required component of every recorded choice.
Chinese Translation
随着代码生成越来越多地委托给人工智能系统，瓶颈正从编写代码转向监督编写代码的系统——这一转变已被计算机科学教育研究者开始命名。这一转变暴露出一个词汇缺口：该领域要求“人类监督”，却没有在委托渠道中对语言所做的两件事进行有效区分——协调行动（控制论：当世界与之相符时，语言成功）和协调理解（认识论：当它们回应世界并且听者能够检查时，它们成功）。这种命名的失败并不是控制论语言，而是进行控制论工作的认识论形式语言：经过调整以获得批准而非真理的解释形状输出。仅检查输出是否获得批准的监督可以通过橡皮图章来满足；而要求代理人负责的监督则需要其工作的推理能够被检索和检查。我们展示了三个委托案例——插图，而非控制证据——在这些案例中，认识论参与被证明是可行的，同时保持可审计性，其中一个公共记录中，推荐在其自身陈述的条款下被撤回，另一个失败案例则展示了不需要理由的自由裁量选择的监督。我们提出了一个代理系统治理的标准，连同现有的技术信任属性：每一个有影响的选择都应携带其可能有所不同的条件，以一种第三方可以测试的形式。没有这样的条件，第三方无法将决策与橡皮图章区分开。我们给出该标准的操作形式——一个两部分重构测试，通过第二读者是否能够预测代理人在扰动下的行为来评分委托记录——以及一个决策记录约定，ORRCF，使得该条件成为每个记录选择的必要组成部分。
cs.AI / 57 / 2610.00972
VeriHarness: Scaling Agentic Verification for Long-Horizon Tasks
VeriHarness：扩展长时间任务的自主验证
Abstract
As LLM agents undertake increasingly complex, long-horizon tasks, verifying their outputs becomes increasingly challenging. We study how verification capability can be strengthened with a fixed base model, without access to reference answers or grading rubrics at test time. Repeated sampling yields multiple rollouts that can contain complementary correct claims, but we need a reliable verification mechanism to determine which claims to trust. We first find that disagreement often exposes correct alternatives, while consensus can conceal errors. These observations motivate VeriHarness, which turns the underlying LLM a generator uses into an agentic verifier by giving it a workspace, evidence tools, and reusable verification skills. A disagreement resolver checks competing claims against environmental evidence, while a consensus challenger tests shared claims and searches for omitted requirements. Their findings guide the selection and revision of the final artifact. Across five long-horizon workspace benchmarks and two frontier models, VeriHarness achieves the highest selection scores among the evaluated baselines. Evidence-backed revision further improves average performance, bringing gains over a single rollout to 6.2 points with Gemini 3.5 Flash and 6.4 points with Claude Opus 4.8. We further show that verification skills can self-improve from failure feedback, demonstrating VeriHarness as a novel and critical approach for scaling long-horizon agentic verification. We release the full pool of approximately 26,000 rollouts across all five benchmarks and both models, produced at a cost of over $100,000, to support future research on agentic verification.
Chinese Translation
随着大型语言模型（LLM）代理执行越来越复杂的长时间任务，验证其输出变得愈加具有挑战性。我们研究了如何在没有参考答案或评分标准的情况下，利用固定基础模型增强验证能力。重复采样产生多个可能的结果，这些结果可能包含互补的正确声明，但我们需要一个可靠的验证机制来确定哪些声明值得信任。我们首先发现，分歧往往揭示了正确的替代方案，而共识可能掩盖错误。这些观察促使我们提出VeriHarness，它通过为基础的LLM生成器提供工作空间、证据工具和可重用的验证技能，将其转变为一个自主验证器。分歧解决者根据环境证据检查竞争声明，而共识挑战者测试共享声明并寻找遗漏的要求。它们的发现指导最终工件的选择和修订。在五个长时间工作空间基准和两个前沿模型中，VeriHarness在评估的基线中实现了最高的选择分数。基于证据的修订进一步提高了平均性能，使得与单次结果相比，使用Gemini 3.5 Flash的提升达到6.2分，使用Claude Opus 4.8的提升达到6.4分。我们进一步展示了验证技能可以通过失败反馈自我改进，证明了VeriHarness作为扩展长时间自主验证的新颖且关键的方法。我们发布了在所有五个基准和两个模型上产生的约26,000个结果的完整数据集，成本超过100,000美元，以支持未来在自主验证方面的研究。
cs.AI / 58 / 2610.00979
RISED: RubrIcs for agentic multi-environment Selection and sElf-Distillation
RISED：用于自主多环境选择和自我蒸馏的评分标准
Abstract
Training a single LLM agent jointly across diverse interactive environments has attracted increasing attention as a route to generalist agents. Existing curriculum and data-selection strategies often allocate training at the environment level or prioritize local reward-based signals, without explicitly considering relationships between current rollouts across environments for prompt-group selection. Meanwhile, as environments are learned at different rates, all-failure and all-success rollout groups can coexist within a batch, leaving those data without group-relative reward signals. Both challenges highlight limitations of relying solely on scalar rewards in multi-environment RL: they provide limited information about cross-environment relationships and no within-group reward contrast when rewards are identical. This motivates richer textual feedback, such as rubrics describing rollout behaviours, to guide learning. Beyond rubrics' usage as reward, we repurpose rubrics to guide both online data selection and policy supervision. An LLM judge tags each rollout using a predefined rubric vocabulary shared across environments. The resulting profiles guide the selection of data that aligns with the overall behavioural composition of the mixed-environment batch while limiting overlap with already-selected data. Available positive rubrics (describing desired behaviours) provide privileged context for an on-policy self-distillation teacher, supplying additional token-level supervision, while negative rubrics (describing undesired behaviours) guide subsequent rollout generation away from recurring failure modes. Together, these components form RISED. Across model backbones, RISED achieves the highest mean pass rate across environments and ranks first or second in every individual environment. Rubric-based analysis of RISED can further characterize the behavioural changes accompanying these gains.
Chinese Translation
在多样化的互动环境中共同训练单一的大型语言模型（LLM）代理已引起越来越多的关注，作为实现通用代理的一种途径。现有的课程和数据选择策略通常在环境层面进行训练分配或优先考虑基于局部奖励的信号，而未明确考虑不同环境之间当前回滚（rollout）之间的关系以进行提示组选择。同时，由于环境的学习速率不同，所有失败和所有成功的回滚组可以在一个批次中共存，使得这些数据缺乏相对组奖励信号。这两个挑战突显了在多环境强化学习中仅依赖标量奖励的局限性：它们提供了关于跨环境关系的有限信息，并且在奖励相同的情况下没有组内奖励对比。这促使我们寻求更丰富的文本反馈，例如描述回滚行为的评分标准，以指导学习。除了将评分标准用作奖励外，我们还重新利用评分标准来指导在线数据选择和策略监督。一个LLM评判者使用在各环境中共享的预定义评分标准词汇对每个回滚进行标记。生成的特征描述指导选择与混合环境批次的整体行为组成一致的数据，同时限制与已选择数据的重叠。可用的正面评分标准（描述期望行为）为在政策内自我蒸馏教师提供了特权上下文，提供额外的令牌级监督，而负面评分标准（描述不期望行为）则指导随后的回滚生成，避免重复的失败模式。这些组件共同构成了RISED。在不同的模型骨干中，RISED在各环境中实现了最高的平均通过率，并在每个单独环境中排名第一或第二。基于评分标准的RISED分析可以进一步表征伴随这些增益的行为变化。
cs.AI / 59 / 2610.01000
Evaluating LLM-Generated Preference Distributions
评估大型语言模型生成的偏好分布
Abstract
Large Language Models (LLMs) are increasingly used as probabilistic generators for simulation, synthetic data generation, and decision support in settings where real-world data are unavailable. Yet, the structure and reliability of the distributions they produce remain understudied. Here, we systematically analyze LLM-generated distributions of preferences for air travel, restaurants, and consumer products. Encouragingly, all models considered in our analysis exhibit self-coherence, with the most probable outcomes stabilizing rapidly under repeated sampling. At the same time, we observe substantial discordance across both model families and scales, with little consensus even among their most probable outcomes. These patterns hold across nine open-weight models, three choice domains, and show robustness under temperature changes, greedy decoding, and perturbations of prompt and ordering. Our findings indicate that outcomes are influenced more by the choice of model than by the wording of the prompt, challenging the common assumption that sufficiently capable LLMs produce similar preference distributions when used as stand-ins for survey respondents.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作概率生成器，在缺乏真实数据的情况下进行模拟、合成数据生成和决策支持。然而，它们所生成分布的结构和可靠性仍然未得到充分研究。在此，我们系统地分析了LLM生成的关于航空旅行、餐馆和消费产品的偏好分布。令人鼓舞的是，我们分析中考虑的所有模型均表现出自我一致性，最可能的结果在重复采样下迅速稳定。同时，我们观察到不同模型家族和规模之间存在显著的不一致性，即使在它们最可能的结果之间也缺乏共识。这些模式在九个开放权重模型和三个选择领域中均得以保持，并且在温度变化、贪婪解码以及提示和顺序的扰动下表现出鲁棒性。我们的研究结果表明，结果更受模型选择的影响，而非提示的措辞，这挑战了一个普遍假设，即足够强大的LLMs在作为调查受访者的替代品时会产生相似的偏好分布。
cs.AI / 60 / 2610.01001
Calibration-risk routing for controlled world-model adaptation
用于控制世界模型适应的校准风险路由
Abstract
Model-based reinforcement learning (MBRL) can exploit simulated experience, but a simulator-to-target shift creates a model-selection problem: correcting the simulator and fitting the target directly can each fail under limited target data. We introduce the Model-Corrected World Model (MC-WM), which separates initial target data into disjoint fit, selection, and calibration partitions and deploys the family with lower standardized calibration risk. A learned confidence signal and deterministic validity predicates weight one-step imagined policy updates without rewriting physical rewards. We evaluate 540 unique reported run cells across three controlled Multi-Joint dynamics with Contact (MuJoCo) shifts; one exact-routing cell was repeated after a pre-deployment artifact gate, giving 541 completed executions.
Chinese Translation
基于模型的强化学习（MBRL）可以利用模拟经验，但模拟器与目标之间的转变会导致模型选择问题：在有限的目标数据下，直接修正模拟器和拟合目标都可能失败。我们提出了模型修正世界模型（Model-Corrected World Model, MC-WM），该模型将初始目标数据分为不相交的拟合、选择和校准部分，并部署具有较低标准化校准风险的模型族。一个学习到的置信信号和确定性的有效性谓词在不重写物理奖励的情况下对一步想象的策略更新进行加权。我们在三个受控的多关节动力学与接触（MuJoCo）转变中评估了540个独特的报告运行单元；在一个预部署伪影门后重复了一个精确路由单元，完成了541次执行。
cs.AI / 61 / 2610.01002
What Can Analogy Tell Us About Artificial Consciousness?
类比能告诉我们关于人工意识什么？
Abstract
Who or what is conscious? Because subjective experience is directly accessible only in the first person, judgments about consciousness in other entities depend partly on analogy. Historically, such inferences have focused on nonhuman animals, but advances in artificial intelligence have raised the possibility of conscious AI. Here we develop a causal framework for evaluating such evidential analogies. The key distinction is between similarities in factors plausibly involved in generating consciousness and similarities in downstream behavioural or cognitive effects. Our framework weights source-target similarity by causal relevance while allowing for unknown causes, disabling differences and alternative routes to consciousness. Applied to biological systems, it explains why analogical support generally weakens with increasing causal distance from humans. Applied to contemporary AI, it suggests that behavioural similarity provides only limited evidence for consciousness because relevant causal correspondences remain poorly established. The framework also clarifies what evidence would strengthen claims of artificial consciousness.
Chinese Translation
谁或什么是有意识的？由于主观体验仅能通过第一人称直接获取，因此对其他实体意识的判断在一定程度上依赖于类比。历史上，这种推断主要集中在非人类动物上，但人工智能的进步使得有意识的人工智能成为可能。在这里，我们建立了一个评估这种证据类比的因果框架。关键的区别在于，涉及生成意识的因素之间的相似性与下游行为或认知效应之间的相似性。我们的框架通过因果相关性对源-目标相似性进行加权，同时允许存在未知原因、禁用差异和替代意识的途径。应用于生物系统时，它解释了为什么类比支持通常随着与人类的因果距离增加而减弱。应用于当代人工智能时，它表明行为相似性仅提供有限的意识证据，因为相关的因果对应关系仍然建立得很差。该框架还阐明了哪些证据可以增强对人工意识的主张。
cs.AI / 62 / 2610.01006
Beyond Answer Confidence: A Controlled Audit of Self-Knowledge in a Black-Box Decision Model
超越答案置信度：对黑箱决策模型自我知识的控制审计
Abstract
Decision models return probabilities intended for routing, abstention and automated action. Calibration makes those probabilities useful on average, but does not establish whether low confidence reflects chance or missing knowledge, nor whether confidence falls when a model moves beyond what it knows. We audit this distinction in Jev, a decision model, with over 15 public datasets and 6 generated task families, with paired interventions that vary the information supplied for a fixed item. Jev's confidence is calibrated on familiar closed-choice tasks but fails as an indicator of missing knowledge: with no answer-relevant information it assigns up to 0.80 to a salient option, and on news beyond an observed knowledge boundary it exceeds accuracy by 0.21--0.33, a gap that recalibration on earlier months does not close. Targeted yes/no questions give sharper readouts of the case: whether an outcome is settled (AUROC 1.00) and whether the evidence suffices (0.95, against 0.85 for confidence on the same items). Asking whether Jev knows the answer appears to flag fabricated entities and post-boundary news (0.91), but with realistic names or with dates removed it shows no advantage over answer uncertainty. Black-box knowledge audits therefore need explicit controls for surface cues. Code: https://github.com/Syntheme/beyond-answer-confidence.
Chinese Translation
决策模型返回的概率用于路由、弃权和自动化操作。校准使这些概率在平均水平上变得有用，但并未确定低置信度是否反映了偶然性或缺失的知识，也未确定当模型超出其已知范围时置信度是否下降。我们在决策模型Jev中审计这一区别，使用超过15个公共数据集和6个生成的任务系列，进行配对干预，以改变为固定项目提供的信息。Jev的置信度在熟悉的封闭选择任务中经过校准，但未能作为缺失知识的指示：在没有与答案相关的信息时，它对一个显著选项赋予高达0.80的置信度，而在超出观察知识边界的新信息上，其准确性超过了0.21至0.33，这一差距在对早期月份进行重新校准时并未缩小。针对性的是/否问题提供了更清晰的案例读数：结果是否已确定（AUROC 1.00）以及证据是否足够（0.95，相比于同一项目的置信度0.85）。询问Jev是否知道答案似乎标记了虚构实体和边界外的新信息（0.91），但在使用现实名称或去除日期时，其表现并未优于答案的不确定性。因此，黑箱知识审计需要对表面线索进行明确控制。代码：https://github.com/Syntheme/beyond-answer-confidence。
cs.AI / 63 / 2610.01014
From Discovery to Decision: Finite-Budget Recoverability in LLM Voting
从发现到决策：有限预算下的 LLM 投票可恢复性
Abstract
Voting over multiple LLM responses is a common primitive in test-time scaling and ensemble inference. Collecting more responses can expand the candidate pool and increase the chance that a correct answer is discovered. Under a fixed call budget, a discovered answer still needs to accumulate enough support within the remaining calls to become the final plurality winner, creating a discovery-to-decision gap. In this work, we characterize this gap through the realized vote state and remaining call budget. We derive a sharp recoverability threshold and show that, as sampling proceeds, the observed candidate set can only expand while the set of reachable endpoint winners can only contract, inducing a candidate-level conversion window. Under a specified iid response law, the same state yields exact finite-horizon endpoint probabilities. We further show that merging wrong-answer identities preserves single-call correctness and cannot improve plurality accuracy, and that the effect of redistributing wrong-answer probability depends on the realized vote state. Singleton reachability yields a gold-free exact locking certificate. For a known answer universe, its first trigger is the earliest prefix at which all admissible continuations yield the same fixed-budget output. Empirically, most discovered-but-unselected correct answers lose reachability only after discovery. In a controlled Word16 study, input permutation improves raw-plurality accuracy by 21.1 points with essentially unchanged single-call correctness. Exact locking saves 28-30% of calls at a 16-call budget while preserving every fixed-budget output.
Chinese Translation
对多个 LLM 响应进行投票是测试时扩展和集成推理中的一种常见原语。收集更多的响应可以扩大候选池，并增加发现正确答案的机会。在固定调用预算下，发现的答案仍然需要在剩余调用中积累足够的支持才能成为最终的多数赢家，从而产生了发现到决策的差距。在本研究中，我们通过实现的投票状态和剩余调用预算来表征这一差距。我们推导出一个明确的可恢复性阈值，并展示随着采样的进行，观察到的候选集只能扩大，而可达的最终赢家集只能收缩，从而诱导出一个候选级别的转换窗口。在指定的独立同分布响应法则下，相同状态产生精确的有限时域终点概率。我们进一步表明，合并错误答案身份保持单次调用的正确性，并且不能提高多数准确性，而重新分配错误答案概率的效果取决于实现的投票状态。单例可达性产生一个无金证的精确锁定证书。对于已知的答案宇宙，其第一个触发是所有可接受的延续产生相同固定预算输出的最早前缀。从经验上看，大多数被发现但未选择的正确答案在发现后才失去可达性。在一项受控的 Word16 研究中，输入排列将原始多数准确性提高了 21.1 个百分点，同时单次调用的正确性基本保持不变。精确锁定在 16 次调用预算下节省了 28-30% 的调用，同时保留了每个固定预算输出。
cs.AI / 64 / 2610.01017
Pay for the Fault, Not the Flow: Label-Free In-Flow Multi-Agent Workflow Optimization
为错误付费，而非流量：无标签的流中多智能体工作流优化
Abstract
Large language models (LLMs) increasingly construct multi-agent workflows that decompose a complex task and assign specialist agents from a pool. However, building such a workflow well remains challenging: how finely to divide the task, which agent to trust with each subtask, and when to create a new specialist are all critical decisions a workflow constructor needs to settle up front. Thus, whether each subtask succeeds remains unknown until the workflow runs. Yet, improving a workflow is costly. Locating a fault usually requires a reference answer, a graded outcome, or a trained assessor, and the fix is applied to the whole workflow through re-execution, re-search, or retraining. We propose InFlowOp, which prices every decision in one label-free cost that weighs how well an agent's competence meets what a subtask demands against how much that agent takes to run. Before execution, InFlowOp bidirectionally determines the granularity of task decomposition and agent assignment following from the cost rather than from a fixed template. During execution, InFlowOp corrects a fault with the cheapest move via the same cost that serves the workflow both as it is built and as it runs. Facing the workflow-level evaluation challenge, we introduce Braid, a benchmark whose tasks require multi-agent coordination beyond single-agent capability. Across various domains and backbones, InFlowOp outperforms single agent baselines by up to $+11.97\%$, achieving $+9.64\%$ with in-flow optimization. Our project page: https://xhguo7.github.io/InFlowOp/.
Chinese Translation
大型语言模型（LLMs）越来越多地构建多智能体工作流，以分解复杂任务并从池中分配专业代理。然而，良好地构建这样的工作流仍然具有挑战性：如何细分任务、信任哪个代理处理每个子任务以及何时创建新的专家都是工作流构建者需要事先解决的关键决策。因此，直到工作流运行之前，每个子任务是否成功仍然未知。然而，改进工作流是昂贵的。定位故障通常需要参考答案、分级结果或经过培训的评估者，并且修复是通过重新执行、重新搜索或重新训练应用于整个工作流的。我们提出了InFlowOp，它以一种无标签的成本对每个决策进行定价，该成本权衡了代理的能力与子任务需求的匹配程度以及该代理运行所需的成本。在执行之前，InFlowOp双向确定任务分解和代理分配的粒度，依据成本而非固定模板。在执行过程中，InFlowOp通过相同的成本以最低的代价修正故障，该成本在工作流构建和运行时均有效。面对工作流级评估挑战，我们引入了Braid，一个基准，其任务需要超越单一代理能力的多智能体协调。在各种领域和基础架构中，InFlowOp的表现优于单一代理基线，提升幅度高达$+11.97\%$，在流中优化下实现了$+9.64\\%$的提升。我们的项目页面：https://xhguo7.github.io/InFlowOp/.
cs.AI / 65 / 2610.01042
Beyond Final Accuracy: Auditing Communication in LLM Multi-Agent Systems
超越最终准确性：大规模语言模型多智能体系统中的通信审计
Abstract
Multi-agent communication aims to help agents benefit from one another's information. Yet improvements in system performance leave a fundamental ambiguity: do they reflect effective communication, a favorable agent architecture, or simply additional reasoning? Because communication methods are commonly evaluated within the systems they were designed for, these factors are difficult to disentangle. Final accuracy further merges corrected errors and corrupted answers into a single outcome, obscuring how communication changes decisions. We introduce Independent--Communicate--Revise (ICR), a controlled framework that evaluates communication as answer revision following independent reasoning. ICR fixes initial reasoning trajectories, measures correction and preservation conditional on both agents' initial correctness, and uses a no-message revision control to quantify gains beyond additional reasoning. Across four reasoning benchmarks, our audit of textual and latent communication reveals that similar aggregate accuracy can conceal substantially different revision behaviors. Compared with transmitting answers alone, full reasoning increases correction while reducing preservation on all four benchmarks, so richer messages amplify beneficial and harmful influence alike. Receiver-policy comparisons on MedQA and GPQA-D further show that a structured verification policy shifts every channel toward greater preservation and lower correction, while its effect on selectivity varies across channels and tasks. These findings challenge treating communication quality as an intrinsic property of a channel. ICR therefore recenters evaluation on selective revision, providing a unified framework for examining how message content and receiver policies jointly produce benefits and harms.
Chinese Translation
多智能体通信旨在帮助智能体相互利用信息。然而，系统性能的提升带来了一个根本性的模糊性：这些提升是反映有效的通信、良好的智能体架构，还是仅仅是额外的推理？由于通信方法通常在其设计的系统内进行评估，这些因素难以分离。最终准确性进一步将纠正的错误和损坏的答案合并为一个结果，模糊了通信如何改变决策。我们引入了独立-通信-修正（Independent--Communicate--Revise, ICR）框架，这是一个控制框架，评估通信作为独立推理后的答案修正。ICR 固定初始推理轨迹，测量在两个智能体初始正确性的条件下的修正和保留，并使用无消息修正控制来量化超出额外推理的收益。在四个推理基准测试中，我们对文本和潜在通信的审计揭示了相似的总体准确性可能掩盖实质上不同的修正行为。与仅传输答案相比，全面推理在所有四个基准测试中增加了修正，同时减少了保留，因此更丰富的信息在有益和有害影响上都得到了放大。在 MedQA 和 GPQA-D 上的接收者策略比较进一步表明，结构化验证策略使每个通道朝着更大保留和更低修正的方向转变，而其对选择性的影响在不同通道和任务中各异。这些发现挑战了将通信质量视为通道的内在属性。因此，ICR 将评估重新聚焦于选择性修正，提供了一个统一框架，以考察消息内容和接收者政策如何共同产生益处和危害。
cs.AI / 66 / 2610.01045
Empty Commitments: When Agents Promise What Their Runtime Cannot Deliver
空洞的承诺：当代理承诺其运行时无法实现的事情
Abstract
A chatbot that says "I will remind you tomorrow" will not run again until the user writes. We call such a promise an empty commitment: a promise of an action after the current turn that nothing in the agent's tools or runtime can carry out. Unlike a broken promise, its emptiness follows from the agent's configuration alone; no later trajectory is needed. We define empty commitments on top of commitment semantics, with three failure types, an anchoring condition for promises that a tool could make real, and a response-level outcome taxonomy. We then describe a measurement protocol: follow-up requests run in five setups that add one persistence affordance at a time, with the environment either left implicit or stated.
Chinese Translation
一个聊天机器人如果说“我明天会提醒你”，在用户没有再次输入之前将不会再次运行。我们称这种承诺为空洞的承诺：在当前回合之后的行动承诺，但代理的工具或运行时无法执行。与破坏的承诺不同，其空洞性仅源于代理的配置；不需要后续的轨迹。我们在承诺语义的基础上定义空洞的承诺，提出三种失败类型，一个工具可以实现的承诺的锚定条件，以及一个响应级别的结果分类法。然后，我们描述了一种测量协议：在五种设置中运行后续请求，每次添加一个持久性功能，环境要么隐含，要么明确说明。
cs.AI / 67 / 2610.01048
Network World Models as Environments for Algorithm Design on Complex Systems
网络世界模型作为复杂系统算法设计的环境
Abstract
World models, which simulate an environment and predict how it changes under actions, are increasingly used in real-world applications such as robotics. Complex systems call for the same tool because the effect of an action is not immediate. Seeding nodes for a campaign, or immunizing nodes against an epidemic, changes little on its own; what matters is the outcome that unfolds over the steps that follow. Designing an algorithm that selects such actions to maximize expected performance on a task is inherently iterative, and every candidate must be scored by the outcome it produces. Obtaining that outcome has relied on simulation, whose cost becomes a bottleneck when candidates are evaluated over many sampled trajectories. We propose an action-conditioned Network World Model that learns a network's diffusion dynamics under interventions over time, applies each action to the network, and predicts the outcome that follows. It serves as a fast evaluator inside an algorithm design loop in which a coding agent designs and refines executable algorithms using feedback from full rollouts, action-level credit, and counterfactual probes over alternative interventions. Across eight network tasks and five diffusion models, the designed algorithms match or exceed the strongest reported baseline in 138 of 141 settings while enabling up to 14.5 times faster rollouts than Monte Carlo simulation. Code will be released upon acceptance.
Chinese Translation
世界模型模拟环境并预测在行动下如何变化，越来越多地应用于机器人等现实世界的应用中。复杂系统同样需要这种工具，因为行动的效果并非立竿见影。为一项活动播种节点或对节点进行免疫以防止流行病，单独来看变化不大；重要的是随后的步骤中所展开的结果。设计一个选择这些行动以最大化任务预期表现的算法本质上是迭代的，每个候选方案都必须通过其产生的结果进行评分。获得该结果依赖于模拟，而当候选方案在多个采样轨迹上进行评估时，模拟的成本成为瓶颈。我们提出了一种基于行动条件的网络世界模型，该模型学习网络在干预下随时间变化的扩散动态，将每个行动应用于网络，并预测随之而来的结果。它作为算法设计循环中的快速评估器，在该循环中，编码代理使用完整回放、行动级别的信用和对替代干预的反事实探测反馈来设计和完善可执行算法。在八个网络任务和五个扩散模型中，所设计的算法在141个设置中有138个匹配或超过了已报告的最强基线，同时使得回放速度比蒙特卡洛模拟快多达14.5倍。代码将在接受后发布。
cs.AI / 68 / 2610.01080
Improving Math Reasoning through Value-guided Informative Search
通过价值引导的信息搜索提升数学推理能力
Abstract
Reinforcement learning with verifiable rewards (RLVR) has substantially improved the mathematical reasoning capabilities of large language models. Recent work introduces search into RLVR rollouts to increase trajectory diversity, but diversity alone does not ensure that the search-induced rollout policy improves upon the current policy. To address this gap, we propose APIVIS, a training-time framework that adapts finite-budget Gumbel search to chunk-level mathematical reasoning. APIVIS combines direct and searched responses within each rollout group, allowing improvements found by search to produce informative relative rewards. It further applies selective supervision to search-improved tokens, preserving a learning signal when uniform group rewards render GRPO ineffective. We show that exact value-guided selection improves the expected verifier reward at each searched state and that this guarantee extends to the complete rollout policy, with a corresponding approximate guarantee under bounded value-estimation error. Experiments on widely recognized mathematical reasoning benchmarks and different model scales demonstrate substantial improvements over competitive search-based methods, validating the effectiveness of APIVIS.
Chinese Translation
具有可验证奖励的强化学习（RLVR）显著提升了大型语言模型的数学推理能力。近期的研究在RLVR的回放中引入搜索以增加轨迹多样性，但仅靠多样性并不能确保搜索引导的回放策略优于当前策略。为了解决这一问题，我们提出了APIVIS，这是一种在训练阶段适应有限预算Gumbel搜索的框架，旨在进行分块级别的数学推理。APIVIS在每个回放组内结合了直接响应和搜索响应，使得搜索发现的改进能够产生信息丰富的相对奖励。它进一步对搜索改进的标记应用选择性监督，在统一组奖励使得GRPO失效时保留学习信号。我们展示了精确的价值引导选择在每个搜索状态下提高了预期验证者奖励，并且这一保证扩展到完整的回放策略，在有界价值估计误差下具有相应的近似保证。在广泛认可的数学推理基准和不同模型规模上的实验表明，相较于竞争性的基于搜索的方法，APIVIS实现了显著的改进，验证了其有效性。
cs.AI / 69 / 2610.01097
YouRA: A Persistent-State Architecture for Evidence-Traceable Autonomous Research Agents
YouRA：一种可追溯证据的持久状态架构用于自主研究代理
Abstract
End-to-end research agents can now produce complete scientific papers, yet manuscript claims often diverge from executed experiments. This gap is structural: research state, failure histories, and claim-evidence alignment are not maintained as persistent, verifiable state across long-horizon pipelines. We present YouRA (Your Research Agent), an architecture for stateful, evidence-traceable autonomous research. YouRA preserves research state, execution evidence, and failure history across the research trajectory by integrating three components: a Verification State Architecture (VSA) that tracks hypotheses, gates, and evidence pointers; an Independent Controller that turns state and reflection records into lifecycle, recovery, and debate/review control while separating control from execution; and Stateful Reflection that logs failures as structured lessons and routes recovery through bounded repair, redesign, or reset. On MLR-Bench's predefined ten-task end-to-end subset, YouRA improves over both MLR-Agent and AI Scientist V2 on scalar Overall across all three matched backbones. An automated diagnostic using MLR-Bench's hallucination taxonomy reports intersection/union counts for four fact-based failure types, and data-provenance diagnostic shows more real-data-based outputs. Ablating each of the four components (the VSA, the Independent Controller, MCP tool access, and reflection-guided recovery) supports their separable contributions. Removing either core-state component drops YouRA below the full system. Code: https://github.com/PrayPrey/Your-Research-Agent.
Chinese Translation
端到端的研究代理现在能够生成完整的科学论文，但手稿中的主张往往与执行的实验不一致。这一差距是结构性的：研究状态、失败历史和主张-证据对齐并未在长时间的流程中保持为持久、可验证的状态。我们提出了YouRA（Your Research Agent），一种用于有状态、可追溯证据的自主研究的架构。YouRA通过整合三个组件，保持研究状态、执行证据和失败历史贯穿研究轨迹：一个验证状态架构（Verification State Architecture, VSA）用于跟踪假设、门控和证据指针；一个独立控制器（Independent Controller）将状态和反思记录转化为生命周期、恢复和辩论/审查控制，同时将控制与执行分离；以及有状态反思（Stateful Reflection），将失败记录为结构化教训，并通过有限的修复、重新设计或重置来引导恢复。在MLR-Bench的预定义十任务端到端子集上，YouRA在所有三个匹配基础架构上相较于MLR-Agent和AI Scientist V2在标量总体上有所提升。使用MLR-Bench的幻觉分类法进行的自动诊断报告了四种基于事实的失败类型的交集/并集计数，数据来源诊断显示更多基于真实数据的输出。消融这四个组件（VSA、独立控制器、MCP工具访问和反思引导恢复）支持它们的可分离贡献。去除任何核心状态组件都会使YouRA的性能低于完整系统。代码：https://github.com/PrayPrey/Your-Research-Agent。
cs.AI / 70 / 2610.01116
Beyond State-of-the-Art: Standardising Environmental Impact Metrics for AI Research
超越现有技术：人工智能研究环境影响指标的标准化
Abstract
As the capabilities and ubiquity of Large Language Models (LLMs) grow, so does their environmental footprint. Despite calls for responsible AI, the machine learning community lacks standardised practices for carbon accounting. Our automated literature review of the 5,285 papers accepted to NeurIPS 2025 reveals that reporting of environmental impact is nearly non-existent. To catalyse a shift toward sustainable AI, we define standardised sustainability metrics for evaluating model training efficiency, accompanied by simple heuristics to estimate the carbon cost of LLM inference. We implement these metrics in carbonbenchmark, a drop-in software solution for tracking and reporting emissions. Finally, to combat the pursuit of marginal accuracy gains at disproportionate environmental costs, we formalise the `Smallest Model that Achieves the Job' (SMAJ), a framework which challenges the field to prioritise computational efficiency and environmental accountability alongside traditional `State-of-the-Art' (SotA) accuracy.
Chinese Translation
随着大型语言模型（Large Language Models, LLMs）能力和普及度的提升，其环境足迹也在不断增加。尽管对负责任的人工智能的呼声日益高涨，机器学习社区在碳核算方面仍缺乏标准化的实践。我们对2025年NeurIPS会议上接受的5,285篇论文进行的自动文献综述显示，环境影响的报告几乎不存在。为了推动向可持续人工智能的转变，我们定义了用于评估模型训练效率的标准化可持续性指标，并提供了简单的启发式方法来估算LLM推理的碳成本。我们在carbonbenchmark中实现了这些指标，这是一个用于跟踪和报告排放的即插即用软件解决方案。最后，为了应对在不成比例的环境成本下追求微小准确性提升的现象，我们正式提出了“完成任务的最小模型”（Smallest Model that Achieves the Job, SMAJ）框架，挑战该领域在传统的“现有技术”（State-of-the-Art, SotA）准确性之外，优先考虑计算效率和环境责任。
cs.AI / 71 / 2610.01119
AbsorbEvo: An Agentic Framework for Autonomous Inverse Design of Microwave Absorbers
AbsorbEvo：一种用于微波吸收器自主逆向设计的代理框架
Abstract
Designing high-performance microwave absorbers requires specialized expertise in electromagnetic theory, materials science and simulation programming, and entails time-consuming optimization. Here, we present AbsorbEvo, an agentic framework for autonomous inverse design that translates natural-language performance objectives into designs verified by full-wave simulations. Its candidate evolution strategy integrates language reasoning, physics-based prediction and historical feedback. A large language model proposes the directions and magnitudes of parameter adjustments based on task objectives and computational history. The system combines directed increments with global sampling to generate candidates and uses a low-cost predictive model as a physics prior to rank them. Only high-ranking designs undergo full-wave simulation. Results passing physical validity checks are used to evaluate performance and guide subsequent search. Experience from training tasks is further distilled into textual skills, which are independently validated before use in new tasks. Under identical proposal budgets on held-out AbsorbBench-36 tasks, AbsorbEvo achieved a task success rate of 79.17%, versus 25.00% for a generic agent and 12.50% for random search. Its mean best coverage was 0.7816, compared with 0.6434 and 0.6448, respectively. By integrating language reasoning and physics-based feedback into design decisions, AbsorbEvo provides a methodological foundation for natural-language-driven autonomous inverse design of microwave absorbers.
Chinese Translation
设计高性能微波吸收器需要在电磁理论、材料科学和仿真编程方面的专业知识，并且涉及耗时的优化过程。在此，我们提出了AbsorbEvo，一种用于自主逆向设计的代理框架，它将自然语言性能目标转化为通过全波仿真验证的设计。其候选演化策略结合了语言推理、基于物理的预测和历史反馈。大型语言模型根据任务目标和计算历史提出参数调整的方向和幅度。该系统将定向增量与全局采样相结合，以生成候选设计，并使用低成本预测模型作为物理先验对其进行排序。只有高排名的设计才会进行全波仿真。通过物理有效性检查的结果用于评估性能并指导后续搜索。训练任务中的经验进一步提炼为文本技能，在新任务中使用前进行独立验证。在相同的提案预算下，AbsorbEvo在保留的AbsorbBench-36任务上实现了79.17%的任务成功率，而通用代理为25.00%，随机搜索为12.50%。其平均最佳覆盖率为0.7816，而通用代理和随机搜索分别为0.6434和0.6448。通过将语言推理和基于物理的反馈整合到设计决策中，AbsorbEvo为自然语言驱动的微波吸收器自主逆向设计提供了方法论基础。
cs.AI / 72 / 2610.01124
CortexBridge: Cortical Alignment of EEG Montages for Foundation Models
CortexBridge：用于基础模型的脑电图（EEG）电极布局皮层对齐方法
Abstract
Electroencephalography (EEG) foundation models are often pretrained with a fixed channel vocabulary or a limited set of montages, making transfer difficult when electrode layouts change. We propose CortexBridge, a lightweight adapter that combines EEG features with electrode and atlas coordinates to map arbitrary montages into a shared cortical latent space. Evaluated with three frozen foundation models on five brain-computer interface (BCI) datasets from the Mother of All BCI Benchmarks (MOABB), CortexBridge improves performance in 13 of 15 evaluations. The gains in balanced accuracy average 0.80% for EEGPT, 0.70% for LaBraM, and 3.26% for CBraMod, with a maximum gain of 13.02% on 12-class steady-state visual evoked potential (SSVEP) classification. Visualizations of the learned atlas representations reveal task-dependent spatial patterns, with SSVEP showing a more concentrated representation in the Yeo Visual network than auditory P300. These results establish cortical alignment as a learnable and anatomically grounded routing mechanism from heterogeneous EEG montages to pretrained foundation models.
Chinese Translation
脑电图（EEG）基础模型通常在固定的电极通道词汇表或有限的电极布局（montages）上进行预训练，当电极布局发生变化时，模型迁移变得困难。我们提出了CortexBridge，一种轻量级适配器，结合了EEG特征、电极坐标和脑图谱（atlas）坐标，将任意电极布局映射到共享的皮层潜在空间。我们在来自“所有脑机接口基准之母”（MOABB）的五个脑机接口（BCI）数据集上，使用三个冻结的基础模型进行了评估，CortexBridge在15次评估中提升了13次性能。平衡准确率的平均提升分别为：EEGPT提升0.80%，LaBraM提升0.70%，CBraMod提升3.26%，在12类稳态视觉诱发电位（SSVEP）分类任务中最高提升达13.02%。对学习到的脑图谱表示的可视化揭示了任务依赖的空间模式，SSVEP在Yeo视觉网络中表现出比听觉P300更为集中的表示。这些结果确立了皮层对齐作为一种可学习且具有解剖学基础的路由机制，实现了从异构EEG电极布局到预训练基础模型的有效映射。
cs.AI / 73 / 2610.01128
Grounding Large Language Models in DSGE Simulators for Policy Generation and Forecasting
将大型语言模型嵌入动态随机一般均衡（DSGE）模拟器以生成政策和进行预测
Abstract
Large language models can produce economic policy responses that sound reasonable, but this does not show that their actions are consistent with economic dynamics. We test this by placing an instruction-tuned language model inside six Snowdrop-backed dynamic stochastic general equilibrium (DSGE) simulators. At each turn, the model observes the economy and a change in economic discourse, selects a bounded policy action, and receives the next simulated state and an economic reward. We implement a common Python interface for repeated rollouts, persistent shocks, state cloning, and rolling-horizon simulation. This setting creates a long-horizon credit-assignment problem. Policy effects may appear several quarters after an action is taken. PPO has a learned value function that can propagate delayed reward to earlier tokens through generalized advantage estimation. GRPO has no learned value function and instead assigns a group-relative advantage from complete rollout returns. It therefore cannot distinguish which earlier turn caused the outcome; if every rollout receives the same return, the normalized advantage is zero. We use PPO as the primary method and GRPO as a matched critic-free baseline. The experiments also test directional semantic signals, reward horizon, trajectory warm starts, cross-simulator transfer, and historically anchored pandemic and monetary-policy shocks. The objective is to judge policy actions by their simulated economic consequences rather than by plausible language alone.
Chinese Translation
大型语言模型能够产生听起来合理的经济政策响应，但这并不能证明其行为与经济动态一致。我们通过将一个经过指令调优的语言模型置于六个由Snowdrop支持的动态随机一般均衡（DSGE）模拟器中进行测试。在每个回合中，该模型观察经济状况及经济话语的变化，选择一个有限的政策行动，并接收下一个模拟状态和经济奖励。我们实现了一个通用的Python接口，用于重复的回合、持久冲击、状态克隆和滚动视野模拟。该设置创建了一个长期的信用分配问题。政策效果可能在采取行动后几个季度才会显现。PPO（Proximal Policy Optimization）具有学习的价值函数，可以通过广义优势估计将延迟奖励传播到更早的标记。GRPO（Generalized Relative Policy Optimization）没有学习的价值函数，而是从完整的回合返回中分配组相对优势。因此，它无法区分哪个早期回合导致了结果；如果每个回合获得相同的回报，则归一化优势为零。我们使用PPO作为主要方法，并将GRPO作为匹配的无评论者基线。实验还测试了方向性语义信号、奖励视野、轨迹热启动、跨模拟器转移以及历史锚定的疫情和货币政策冲击。我们的目标是通过模拟的经济后果来评判政策行动，而不仅仅是通过合理的语言。
cs.AI / 74 / 2610.01138
Auditing Action Settlement in LLM Agent Environments: Order, Progress, and Replay
大语言模型（LLM）代理环境中的行动结算审计：顺序、进展与重放
Abstract
Concurrent actions in large language model (LLM) agent environments require arbitration even when each proposal is individually valid. We implement a typed snapshot-settlement contract and audit three distinct properties: order sensitivity, useful progress, and replay consistency. Five settlement policies are tested in 28,800 exhaustive permutation trials and 2,160 scripted multistep episodes. Joint policies are spatially order-invariant conditional on fixed priorities, yet conservative rejection completes only 31.25% of agents in a six-agent doorway task versus 90.28% for random tickets; the paired improvement is 59.03 percentage points (95% bootstrap interval: 50.00-68.06). All policies preserve the tested spatial constraints, and priority arbitration still misses the independent small-instance optimum. A separate full-state journal audit exactly replays 156 checkpoints and rejects 1,332 constructed corruptions with a retained terminal anchor. The evidence concerns execution semantics, not human realism or long-run fairness.
Chinese Translation
在大语言模型（LLM）代理环境中，即使每个提议都是单独有效的，仍然需要对并发行动进行仲裁。我们实现了一种类型化快照结算合同，并审计了三种不同的属性：顺序敏感性、有用进展和重放一致性。在28,800次详尽的排列试验和2,160个脚本化的多步骤情节中测试了五种结算策略。联合策略在固定优先级的条件下在空间上是顺序不变的，然而保守拒绝在六代理的门口任务中仅完成了31.25%的代理，而随机票据的完成率为90.28%；两者的配对改进为59.03个百分点（95%自助法区间：50.00-68.06）。所有策略都保持了测试的空间约束，而优先级仲裁仍然错过了独立的小实例最优解。一个单独的全状态日志审计准确重放了156个检查点，并拒绝了1,332个构造的腐败案例，同时保留了终端锚点。证据涉及执行语义，而非人类现实主义或长期公平性。
cs.AI / 75 / 2610.01140
ReSolve: Reusing Candidate Reasoning through Selective Generative Moderation
ReSolve：通过选择性生成调节重用候选推理
Abstract
Sampling multiple solutions spends computation on intermediate deductions and unfinished arguments as well as final answers. We introduce ReSolve, a training-free inference procedure that reuses this candidate reasoning through selective generative moderation. An answer-distribution controller invokes a model to examine existing derivations when candidates disagree or lack a parseable answer, then incorporates the generated solution into a bounded loop. Under Hybrid scoring on 130 competition-mathematics problems evaluated with two independently sampled candidate pools, ReSolve obtains 100 and 99 correct answers, compared with 91 and 92 for voting over the same four candidates, with no correct-to-incorrect changes relative to that vote in either pool. Eight-sample self-consistency obtains 94 and 96 correct answers while consuming substantially more tokens; ReSolve uses 46.3% and 47.2% fewer tokens in the two evaluations. A controlled ablation removes visible derivations while retaining answer keys, vote counts, and the per-state output-cap rule, reducing accuracy from 100 to 93 correct despite increasing computation. Selective and always-on Uniform moderation both solve 97 problems, while selectivity reduces moderation tokens by approximately 54% and total pipeline tokens by 6.2%. These results support candidate reasoning as reusable inference computation. They do not establish an accuracy advantage over additional sampling or a distinct benefit from specialized route instructions.
Chinese Translation
采样多个解决方案会在中间推导和未完成的论证以及最终答案上消耗计算资源。我们提出了ReSolve，这是一种无训练的推理过程，通过选择性生成调节重用候选推理。当候选答案不一致或缺乏可解析答案时，答案分布控制器调用模型检查现有推导，然后将生成的解决方案纳入一个有界循环。在对130个竞赛数学问题进行混合评分的评估中，ReSolve获得了100和99个正确答案，而对同四个候选者进行投票则分别获得91和92个正确答案，在两个候选池中与该投票相比没有正确与错误的变化。八样本自一致性获得了94和96个正确答案，但消耗的令牌显著更多；而ReSolve在这两次评估中分别使用了46.3%和47.2%更少的令牌。一个控制性消融去除了可见推导，同时保留了答案键、投票计数和每状态输出上限规则，尽管计算量增加，准确率从100降低到93。选择性和始终开启的均匀调节都解决了97个问题，而选择性使调节令牌减少了约54%，总管道令牌减少了6.2%。这些结果支持候选推理作为可重用的推理计算。它们并未确立相对于额外采样的准确性优势或来自专门路线指令的独特益处。
cs.AI / 76 / 2610.01188
When Does Exercise-Specific Joint Selection Help? An Audit of Evaluation and Control Design
何时特定运动关节选择有助于改善？评估与控制设计的审计
Abstract
Exercise-specific joint selection can improve skeleton-based correctness classification, but what does that gain establish? We audit 1,057 repetitions from ten REHAB24-6 subjects, separating evaluation aggregation, subset structure, and temporal representation. The manual-subset kNN gain changes from 0.055 for pooled out-of-fold AUROC to 0.020 for equal-weight within-person AUROC; both paired intervals include zero. Among 1,000 dimension-matched random maps, 14 match or exceed the manual pooled result, versus 145 when bilateral structure and trunk inclusion are also matched. RBF-SVM retains a positive within-person gain, whereas logistic regression and a random-convolution comparator have negative point gains under that estimand. Sequence-order and paired-seed controls further qualify the interpretation. This exploratory audit shows why joint-selection claims require explicit estimands and structurally appropriate controls; it does not establish a new algorithm or clinical benefit.
Chinese Translation
特定运动关节选择可以提高基于骨架的正确性分类，但这种提升究竟意味着什么？我们审计了来自十名 REHAB24-6 受试者的 1,057 次重复，分离了评估聚合、子集结构和时间表示。手动子集 kNN 的增益从 0.055（针对汇总的外折 AUROC）变化到 0.020（针对等权重的个体内 AUROC）；这两个配对区间均包含零。在 1,000 个维度匹配的随机映射中，14 个匹配或超过了手动汇总结果，而当双侧结构和躯干包含也匹配时，这一数字为 145。RBF-SVM 在个体内保持正增益，而逻辑回归和随机卷积比较器在该估计下则有负点增益。序列顺序和配对种子控制进一步限定了解释。这项探索性审计表明，关节选择的主张需要明确的估计量和结构上适当的控制；它并未建立新的算法或临床益处。
cs.AI / 77 / 2610.01195
Federated Agent Optimization
联邦智能体优化
Abstract
Large language model (LLM) agents increasingly operate in private environments and accumulate valuable experience from task execution, tool use, feedback, and local knowledge. Yet such experience is distributed across organizations and cannot be directly shared because of privacy and proprietary constraints. Conventional federated learning is insufficient for this setting, as agent capabilities extend beyond model parameters to memory, tools, rewards, skills, and structured knowledge. In this paper, we formulate \textbf{Federated Agent Optimization (FAO)}, which studies how distributed agents can collaboratively improve through controlled information exchange while keeping raw data, complete trajectories, and private knowledge local. We define FAO as a multi-objective problem balancing agent utility, privacy leakage, and communication cost, and organize its optimization space across policy, memory, tool use, reward, and structured knowledge and skills. We further characterize how private experience can be abstracted, protected, aggregated, and adapted into transferable capabilities, providing a unified view of how agents can benefit from one another without direct experience sharing. Finally, we identify the key challenges of FAO and outline several promising directions for future research toward trustworthy federated agent systems.
Chinese Translation
大型语言模型（LLM）智能体越来越多地在私密环境中运行，并从任务执行、工具使用、反馈和本地知识中积累宝贵的经验。然而，这些经验分布在不同组织之间，由于隐私和专有约束，无法直接共享。传统的联邦学习在这种情况下不足，因为智能体的能力不仅限于模型参数，还包括记忆、工具、奖励、技能和结构化知识。本文提出了 extbf{联邦智能体优化（Federated Agent Optimization, FAO）}的框架，研究如何通过受控的信息交换，使分布式智能体能够协作改进，同时保持原始数据、完整轨迹和私有知识的本地性。我们将FAO定义为一个多目标问题，平衡智能体效用、隐私泄露和通信成本，并在策略、记忆、工具使用、奖励以及结构化知识和技能之间组织其优化空间。我们进一步描述了如何将私有经验抽象化、保护、聚合和适应为可转移的能力，提供了一个统一的视角，展示智能体如何在不直接共享经验的情况下互惠互利。最后，我们识别了FAO的关键挑战，并概述了未来研究中值得关注的几个有前景的方向，以推动可信的联邦智能体系统的发展。
cs.AI / 78 / 2610.01207
Dependency-Aware Reward Shaping for Agentic Reinforcement Learning
面向依赖的奖励塑造在自主强化学习中的应用
Abstract
When training large language models with reinforcement learning, terminal rewards provide little guidance about which steps matter. Common methods for assigning step credit overlook that work built on uncorrected mistakes is wasted while independent work remains valid. With only a final success/failure reward, every step in a failed episode has zero total future reward, even when it made progress. We propose Dependency-Aware Reward Shaping (DARS), which represents task progress as predicates linked by prerequisite relations and assigns step-level credit over the dependency graph. An annotator marks which predicates each step verifies, invalidates, or repairs. Verified predicates are discounted according to graph distance from the nearest broken prerequisite, while independent predicates are unaffected. Repairs update these weights based on any errors that remain; invalidated predicates need re-verification to regain credit. A fixed potential converts these annotations into signed per-step rewards. A common reward and annotation interface allows DARS to integrate with a range of reasoning and agentic training methods, such as GiGPO and ARPO/AEPO, without changing their rollout strategies or optimizers. Across five task families and models from 1.5B to 8B, DARS improves success by up to 10 points over GiGPO trained with the same budget and harness (ALFWorld), raises the WebShop task score and Search-R1 QA accuracy, complements AEPO's entropy-based training on AIME24/25 with a Python interpreter, and exceeds OmniOPD in controlled tool-free reasoning comparisons at 1.7B and 4B. Ablations show that step-level credit, dependency attenuation, and graph topology each contribute. On ALFWorld, a distilled 8B annotator matches the API annotator, enabling DARS to run efficiently without a frontier judge. Code is available at https://github.com/JianhuiWei7/DARS.
Chinese Translation
在使用强化学习训练大型语言模型时，终端奖励对哪些步骤重要的指导作用有限。常见的步骤信用分配方法忽视了基于未纠正错误的工作是浪费的，而独立工作的有效性则得以保留。仅有的最终成功/失败奖励使得在失败的情节中每一步的未来总奖励为零，即使它取得了进展。我们提出了面向依赖的奖励塑造（Dependency-Aware Reward Shaping, DARS），它将任务进展表示为通过前提关系连接的谓词，并在依赖图上分配步骤级信用。注释者标记每一步验证、无效或修复了哪些谓词。经过验证的谓词根据与最近的破损前提的图距离进行折扣，而独立谓词则不受影响。修复根据剩余的错误更新这些权重；无效的谓词需要重新验证以恢复信用。固定的潜在值将这些注释转换为带符号的每步奖励。通用的奖励和注释接口使DARS能够与一系列推理和自主训练方法（如GiGPO和ARPO/AEPO）集成，而无需改变它们的展开策略或优化器。在五个任务系列和1.5B到8B的模型中，DARS在相同预算和工具（ALFWorld）下比GiGPO提高了最多10个点的成功率，提高了WebShop任务得分和Search-R1 QA准确率，补充了AEPO在AIME24/25上的基于熵的训练，并在1.7B和4B的受控无工具推理比较中超越了OmniOPD。消融实验表明，步骤级信用、依赖衰减和图拓扑各自都有贡献。在ALFWorld上，一个精简的8B注释者与API注释者匹配，使DARS能够高效运行而无需前沿评判者。代码可在 https://github.com/JianhuiWei7/DARS 获取。
cs.AI / 79 / 2610.01222
Reputation, Strategy, and Emotion Effects on Generative AI Cooperation: A Comparison Across Reasoning and Non-Reasoning Models
声誉、策略与情感对生成性人工智能合作的影响：推理模型与非推理模型的比较
Abstract
As generative AI (Gen AI) systems take on increasingly autonomous roles in economically and socially consequential interactions, understanding their propensity to cooperate -- and the signals that shape this propensity -- has become essential. We examine cooperative behavior in frontier Gen AI models using the iterated prisoner's dilemma, manipulating counterpart reputation (positive, unknown, negative), strategy (extortion vs. generosity), and non-verbal emotional signaling (facial expressions conveying competitive or cooperative appraisals). In a first study with non-reasoning models (Claude 3.5, Gemini 2.0 Flash, GPT-4o), cooperation was systematically shaped by all three factors, paralleling patterns long documented in human behavioral research, though models varied substantially in how heavily each factor was weighted. A second study with reasoning models (Claude 4.6, Gemini 3, GPT-5.2) revealed a more concentrated reliance on strategy and reputation, a near-elimination of the Potemkin effect observed in non-reasoning models (evidenced by near-uniform cooperation in a diagnostic harmony game), and a more conditional role for emotion consistent with a hierarchical cue-weighting strategy rather than a simple loss of social sensitivity. Reasoning models also showed heterogeneous end-game behavior, ranging from sustained cooperation to systematic last-round defection effect, revealing model-specific exploitability profiles with direct practical relevance for deployment in negotiation and other multi-round interactions. Together, these findings characterize Gen AI models as increasingly sophisticated, though heterogeneous, social actors, and underscore the practical value of developing standardized cooperation benchmarks to inform the responsible deployment of Gen AI in interactive, socially consequential settings.
Chinese Translation
随着生成性人工智能（Gen AI）系统在经济和社会重要互动中扮演越来越自主的角色，理解它们的合作倾向及其形成信号变得至关重要。我们使用迭代囚徒困境考察前沿Gen AI模型中的合作行为，操控对手声誉（积极、未知、消极）、策略（敲诈与慷慨）以及非语言情感信号（传达竞争或合作评估的面部表情）。在对非推理模型（Claude 3.5、Gemini 2.0 Flash、GPT-4o）进行的首次研究中，合作行为系统性地受到这三种因素的影响，这与人类行为研究中长期记录的模式相似，尽管模型在每个因素的权重上存在显著差异。第二项针对推理模型（Claude 4.6、Gemini 3、GPT-5.2）的研究显示，对策略和声誉的依赖更加集中，几乎消除了在非推理模型中观察到的波将金效应（在诊断和谐游戏中表现为几乎一致的合作），并且情感的作用更具条件性，这与层次化线索加权策略一致，而非简单的社会敏感性丧失。推理模型还表现出异质的终局行为，从持续合作到系统性的最后一轮背叛效应，揭示了模型特定的可利用性特征，对谈判及其他多轮互动的实际应用具有直接相关性。综合来看，这些发现将Gen AI模型描绘为日益复杂但异质的社会参与者，并强调开发标准化合作基准以指导在互动和社会重要场景中负责任地部署Gen AI的实际价值。
cs.AI / 80 / 2610.01230
HHR: Hierarchical Hash Retrieval for Efficient LLM Generation
HHR：用于高效LLM生成的层次哈希检索
Abstract
Efficient long-context inference is essential for large language models (LLMs), yet it poses a severe computational bottleneck. Hash-based retrieval offers an efficient alternative by encoding queries and keys into binary codes and using Hamming distance for key selection. However, this leads to a critical mismatch between Hamming distance and attention relevance. Query-Key logits depend jointly on directional similarity and feature magnitudes, whereas hash binarization discards magnitude information, causing both false-positive retrieval of low-logit keys and false-negative omission of high-logit keys. To address these failures, we propose Hierarchical Hash Retrieval (HHR), a coarse-to-fine framework that progressively improves retrieval accuracy through Geometry-Aware Key Routing (GKR) and Learned Hash Projection (LHP). GKR learns a head-wise orthogonal transformation to redistribute feature magnitudes and derive more discriminative page-level logit bounds, enabling effective pruning of low-logit keys while preserving important candidates. LHP then learns a head-wise projection space that aligns Hamming distance with the true Query-Key relevance ranking for fine-grained retrieval. By combining GKR and LHP, HHR suppresses false positives and recovers false negatives, substantially improving the fidelity of hash-based sparse attention. Extensive experiments across diverse LLMs and benchmarks demonstrate that HHR achieves superior performance over existing methods. For example, on LongBench, HHR improves the average score by 1.10 points and, at a context length of 128K, achieves up to a 3.30x decoding speedup and a 2.83x end-to-end speedup for Llama-3.1-8B-Instruct. The code is publicly available at https://github.com/lianjunl13-sudo/HHR.
Chinese Translation
高效的长上下文推理对于大型语言模型（LLMs）至关重要，但它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代码，并使用汉明距离进行键选择，提供了一种高效的替代方案。然而，这导致了汉明距离与注意力相关性之间的关键不匹配。查询-键的logits共同依赖于方向相似性和特征幅度，而哈希二值化则丢弃了幅度信息，导致低logit键的误检和高logit键的漏检。为了解决这些问题，我们提出了层次哈希检索（HHR），这是一个粗到细的框架，通过几何感知键路由（Geometry-Aware Key Routing, GKR）和学习哈希投影（Learned Hash Projection, LHP）逐步提高检索准确性。GKR学习头级正交变换，以重新分配特征幅度并推导出更具区分性的页面级logit界限，从而有效修剪低logit键，同时保留重要候选项。随后，LHP学习一个头级投影空间，使汉明距离与真实的查询-键相关性排名对齐，以实现细粒度检索。通过结合GKR和LHP，HHR抑制了误检并恢复了漏检，显著提高了基于哈希的稀疏注意力的保真度。在多种LLM和基准测试中的广泛实验表明，HHR的性能优于现有方法。例如，在LongBench上，HHR将平均得分提高了1.10分，并且在上下文长度为128K时，实现了Llama-3.1-8B-Instruct的解码速度提升高达3.30倍和端到端速度提升高达2.83倍。代码已公开发布在https://github.com/lianjunl13-sudo/HHR。
cs.AI / 81 / 2610.01236
Learning to Ask: Information Acquisition for SLM-LLM Collaboration, under a budget
学习提问：在预算限制下的SLM-LLM协作信息获取
Abstract
Collaboration between a small language model (SLM) and a large language model (LLM) offers an opportunity to combine the efficiency of smaller models with the strong reasoning capabilities of larger ones. Existing approaches primarily frame such collaboration as a computation allocation problem, determining which model should handle each portion of the reasoning process. In black-box API-based settings, however, this paradigm can be inefficient due to coarse-grained delegation or repeated transmission of context across model switches. In this work, we instead formulate SLM-LLM collaboration as an information acquisition problem, under an API budget constraint. The SLM remains the primary reasoner and selectively queries a black-box LLM advisor only when needed, issuing targeted queries rather than delegating the reasoning process itself. To realize this strategy, we develop a three-stage RLVR framework that learns whether to call the advisor, how to formulate useful queries, and how to integrate the collaboration into the reasoning process by jointly refining advisor invocation and information use. Across mathematical reasoning and coding tasks, our approach improves the performance--cost tradeoff over existing collaboration baselines and, in some settings, matches or exceeds oracle problem-level routing. Finally, we show that our strategy can transfer to other advisor model families, without further training.
Chinese Translation
小型语言模型（SLM）与大型语言模型（LLM）之间的协作为将小型模型的高效性与大型模型的强大推理能力相结合提供了机会。现有方法主要将这种协作框架视为计算分配问题，确定哪个模型应处理推理过程的每个部分。然而，在基于黑箱API的设置中，由于粗粒度的委派或在模型切换过程中重复传输上下文，这种范式可能效率低下。在本研究中，我们将SLM-LLM协作重新构建为在API预算约束下的信息获取问题。SLM仍然是主要的推理者，仅在需要时选择性地查询黑箱LLM顾问，发出有针对性的查询，而不是将推理过程本身委托出去。为了实现这一策略，我们开发了一个三阶段的RLVR框架，该框架学习何时调用顾问、如何制定有用的查询，以及如何通过共同优化顾问调用和信息使用将协作整合到推理过程中。在数学推理和编码任务中，我们的方法在性能与成本的权衡上优于现有的协作基准，并且在某些设置中，与或超过了理想问题级别的路由。最后，我们展示了我们的策略可以转移到其他顾问模型家族，而无需进一步训练。
cs.AI / 82 / 2610.01249
Revision-Aware Independent Agent Graphs for Dynamic Reasoning
动态推理中的修订感知独立代理图
Abstract
Conventional reasoning protocols present a fixed, preselected task, so they cannot test whether an agent propagates relevant updates, preserves unaffected work, or reconstructs a historical task binding. We therefore study \emph{dynamic task routing}, in which an event stream revises task bindings and a system must select the document version valid at each query time before solving it. To study this problem, we repurpose six widely used benchmarks: MMLU, MMLU-Pro, MedMCQA, MATH, GPQA, and HumanEval into 31{,}119 dynamic episodes comprising 373{,}428 temporally categorized queries. This setting exposes a central trade-off: recomputing after every event wastes work, whereas unguarded reuse returns stale conclusions. We introduce the Revision-Aware Independent Agent Graph (RIAG), a bounded multi-agent policy that separates deterministic temporal resolution from task reasoning. RIAG caches solutions by immutable document identity, starts each fresh task with two unexposed attempts, and conditionally invokes audit and repair, using at most four calls per document version. On this collection, homogeneous RIAG achieves 54.24\% joint routing-and-answer accuracy at 0.62 calls/query, compared with 32.22\% at 18.00 calls/query for the strongest comparison method; heterogeneous RIAG reaches 49.78\% at 0.63 calls/query.
Chinese Translation
传统的推理协议呈现固定的、预先选择的任务，因此无法测试代理是否传播相关更新、保留未受影响的工作或重建历史任务绑定。因此，我们研究了 extit{动态任务路由}，在这种情况下，事件流修订任务绑定，系统必须在每次查询时选择有效的文档版本，然后再进行求解。为了研究这个问题，我们重新利用了六个广泛使用的基准：MMLU、MMLU-Pro、MedMCQA、MATH、GPQA和HumanEval，将其转化为31,119个动态情节，包含373,428个按时间分类的查询。这个设置揭示了一个核心权衡：在每个事件后重新计算会浪费工作，而不加保护的重用则返回过时的结论。我们引入了修订感知独立代理图（Revision-Aware Independent Agent Graph, RIAG），这是一种有界的多代理策略，将确定性的时间解析与任务推理分开。RIAG通过不可变的文档身份缓存解决方案，以两个未暴露的尝试开始每个新任务，并有条件地调用审计和修复，每个文档版本最多使用四次调用。在这个集合中，同质RIAG在每个查询0.62次调用下实现了54.24%的联合路由和回答准确率，而最强的比较方法在每个查询18.00次调用下仅实现了32.22%；异质RIAG在每个查询0.63次调用下达到了49.78%。
cs.AI / 83 / 2610.01256
DeFA: Dependency-Guided Failure Attribution for LLM Agents
DeFA：基于依赖关系的LLM代理失败归因
Abstract
Errors in LLM agent executions and their visible consequences can be separated by many steps, making decisive-error localization a matter of understanding both step content and step dependencies. We introduce DeFA, a dependency-guided framework for agent failure attribution. DeFA first combines protocol relations and semantic dependencies into an event dependency graph spanning the trajectory. It then identifies events that may violate task requirements and traces their sources and subsequent effects to construct a failure propagation graph. Finally, DeFA uses step evidence and the steps' roles in failure propagation to identify the decisive error, responsible agent, and error category. To support long trajectories, DeFA partitions executions into segments and combines the current segment's detailed content with summaries of the other segments, giving local diagnosis access to global execution context. Across Who and When and the Who and When Pro text subset, DeFA achieves the highest responsible-agent and exact step accuracy with all evaluated backbones, and the highest failure-mode accuracy among taxonomy-aligned methods on Pro. Further experiments on image and video trajectories demonstrate its applicability to multimodal failure attribution. Ablations support the contributions of segmentation, the event dependency graph, and the failure propagation graph. Using DeFA's diagnostic feedback for skill evolution in Trace2Skill improves downstream task accuracy by 6-15 percentage points over the native pipeline, showing that the diagnoses can also support agent improvement on subsequent tasks.
Chinese Translation
LLM代理执行中的错误及其可见后果可能相隔多个步骤，使得决定性错误定位成为理解步骤内容和步骤依赖关系的问题。我们提出了DeFA，一种基于依赖关系的代理失败归因框架。DeFA首先将协议关系和语义依赖结合成一个跨越轨迹的事件依赖图。然后，它识别可能违反任务要求的事件，并追踪其来源及后续影响，以构建失败传播图。最后，DeFA利用步骤证据及其在失败传播中的角色来识别决定性错误、责任代理和错误类别。为了支持长轨迹，DeFA将执行分割为多个段，并将当前段的详细内容与其他段的摘要结合，使局部诊断能够访问全局执行上下文。在Who和When以及Who和When Pro文本子集上，DeFA在所有评估的骨干网络中实现了最高的责任代理和精确步骤准确率，并在Pro上实现了与分类对齐方法中最高的失败模式准确率。对图像和视频轨迹的进一步实验展示了其在多模态失败归因中的适用性。消融实验支持了分段、事件依赖图和失败传播图的贡献。使用DeFA的诊断反馈进行Trace2Skill中的技能演变，使下游任务的准确率比原生管道提高了6-15个百分点，表明这些诊断也可以支持代理在后续任务中的改进。
cs.AI / 84 / 2610.01262
Feedback Without the Wait: Piloting a Generative AI Practice Platform in a Large Maths Class
无需等待的反馈：在大规模数学课堂中试点生成性人工智能实践平台
Abstract
Timely and specific feedback is one of the strongest influences on student learning, yet it is difficult to sustain in large electrical engineering classes where the ratio of students to demonstrators is high and a learner who is stuck may wait days to find out why an approach was wrong. Generative Artificial Intelligence (GenAI) offers a way to scale conversational feedback, but using it to grade assessed work raises trust and accountability concerns, and keeping a human in the loop to assure its judgements reintroduces the very delay that erodes the value of feedback. The result is a tension between the immediacy that makes feedback so impactful and the human oversight that makes it trustworthy. In this work, we set out to resolve that tension in practice by designing and piloting a GenAI practice platform that delivers immediate, scaffolded feedback during self-directed practice. This relocates human oversight from real-time grading to the upfront verification of solutions. Our goal was to understand how students engaged with the tool, how they perceived the value and reliability of its feedback, and what lessons transfer to other engineering subjects.
Chinese Translation
及时且具体的反馈是影响学生学习的最强因素之一，但在学生与示范者比例较高的大型电气工程课程中，这种反馈难以维持。当学习者遇到困难时，可能需要等待数天才能了解某种方法为何错误。生成性人工智能（GenAI）提供了一种扩展对话反馈的方式，但将其用于评估作业时会引发信任和责任感的问题，而保持人类参与以确保其判断又重新引入了侵蚀反馈价值的延迟。因此，反馈的即时性与使其可信赖的人类监督之间存在着紧张关系。在本研究中，我们旨在通过设计和试点一个GenAI实践平台来解决这一紧张关系，该平台在自我导向的练习过程中提供即时、分层的反馈。这将人类监督从实时评分转移到对解决方案的前期验证。我们的目标是理解学生如何与该工具互动，他们如何看待其反馈的价值和可靠性，以及哪些经验教训可以转移到其他工程学科中。
cs.AI / 85 / 2610.01278
SCOPE-AD: Sequential cost-aware ordinal-belief planning with energy-based models for diagnostic agents
SCOPE-AD：基于能量模型的诊断智能体的顺序成本感知序数信念规划
Abstract
Alzheimer's disease (AD) diagnosis requires sequential evidence acquisition under heterogeneous test costs and patient burden. Fixed-modality predictors do not jointly decide which test to acquire or when the available evidence is sufficient for diagnosis. We propose SCOPE-AD (Sequential Cost-Aware Ordinal-Belief Planning with Energy-Based Models for Diagnostic Agents) for cost-aware classification of cognitively normal (CN), mild cognitive impairment (MCI), and AD cases. A mask-aware ordinal model represents uncertainty along the ordered CN--MCI--AD continuum. Retrospective training records provide sampled Bellman targets for an energy-based teacher, whose action distributions are distilled into a Qwen policy. At deployment, the agent selects acquisition or diagnosis actions under availability and budget constraints without access to unacquired values. After each acquisition, the evidence and ordinal belief are updated before the next decision. On ADNI, SCOPE-AD achieves 77.70\% Macro-F1 at an average acquisition cost of \$50.46, exceeding the strongest evaluated baseline by 9.34 percentage points. Full-modality evaluation raises Macro-F1 by only 1.89 points while increasing acquisition cost by 116.7 times. These results support selective acquisition for cost-effective diagnosis.
Chinese Translation
阿尔茨海默病（AD）的诊断需要在异质测试成本和患者负担下进行顺序证据获取。固定模态预测器无法联合决定获取哪个测试或何时可用证据足以进行诊断。我们提出了SCOPE-AD（基于能量模型的诊断智能体的顺序成本感知序数信念规划），用于对认知正常（CN）、轻度认知障碍（MCI）和阿尔茨海默病（AD）病例进行成本感知分类。一个掩膜感知的序数模型表示沿着有序的CN-MCI-AD连续体的不确定性。回顾性训练记录提供了样本化的贝尔曼目标，用于一个能量基础教师，其行动分布被提炼为Qwen策略。在部署时，智能体在可用性和预算约束下选择获取或诊断行动，而无需访问未获取的值。在每次获取后，证据和序数信念在下一个决策之前进行更新。在ADNI数据集上，SCOPE-AD在平均获取成本为50.46美元的情况下实现了77.70%的宏观F1分数，超过了最强评估基线9.34个百分点。全模态评估仅将宏观F1提高了1.89点，同时获取成本增加了116.7倍。这些结果支持为成本效益诊断进行选择性获取。
cs.AI / 86 / 2610.01282
Trustworthy Data- and ML-Ops for Intelligent Transportation Systems and Logistics
智能交通系统与物流的可信数据与机器学习操作
Abstract
The rapid evolution of Intelligent Transportation Systems and Logistics (ITS\&L) has become a cornerstone of the modern social economy, relying heavily on the integration of Data, Artificial Intelligence (AI), and, more specifically, Machine Learning (ML). This paper provides a comprehensive review of Trustworthy Data and Machine Learning Operations (DataOps and MLOps) in the ITS\&L domain, underscoring their importance in improving efficiency, reliability, and decision-making precision within transportation and logistics services. We begin by identifying gaps in current literature, offering clear context for our contribution. Subsequently, we explore the complexities of DataOps and MLOps, discussing their necessity, key components, available tools, practical insights, and case studies relevant to ITS\&L. Additionally, we address the critical issue of Trustworthiness in AI applications, examining methods and tools designed to strengthen confidence in AI systems - especially in real-world ITS\&L scenarios. The paper concludes with a discussion of persisting challenges and future prospects in this rapidly advancing field, aiming to serve as a vital resource for researchers, industry practitioners, and policy makers. Overall, this work not only establishes a foundational understanding of DataOps and MLOps in ITS\&L but also charts a path for further research and innovation in developing more efficient, sustainable, and trustworthy intelligent transportation and logistics systems.
Chinese Translation
智能交通系统与物流（ITS&L）的快速发展已成为现代社会经济的基石，极大依赖于数据、人工智能（AI）以及更具体的机器学习（ML）的整合。本文对ITS&L领域中的可信数据与机器学习操作（DataOps和MLOps）进行了全面的回顾，强调其在提高运输和物流服务的效率、可靠性和决策精度方面的重要性。我们首先识别当前文献中的空白，为我们的贡献提供清晰的背景。随后，我们探讨了DataOps和MLOps的复杂性，讨论了它们的必要性、关键组成部分、可用工具、实践见解以及与ITS&L相关的案例研究。此外，我们还解决了AI应用中可信性这一关键问题，考察了旨在增强对AI系统信任的方法和工具，特别是在现实世界的ITS&L场景中。最后，本文讨论了这一快速发展的领域中持续存在的挑战和未来前景，旨在为研究人员、行业从业者和政策制定者提供重要资源。总体而言，本研究不仅建立了对ITS&L中DataOps和MLOps的基础理解，还为进一步研究和创新发展更高效、可持续和可信的智能交通与物流系统指明了方向。
cs.AI / 87 / 2610.01296
ITC-MoE: Importance-guided Token-aware Compression for MoE Diffusion Language Models
ITC-MoE：基于重要性引导的面向令牌的MoE扩散语言模型压缩
Abstract
Mixture-of-Experts (MoE) Diffusion Language Models (DLMs) offer flexible parallel decoding and increased model capacity, but their large number of expert parameters incurs substantial computation and storage costs. Existing low-rank MoE compression methods largely rely on static factorization and fixed rank allocation, which overlook the distinctive properties of MoE DLMs. Specifically, we identify two properties: cross-mode non-uniform redundancy, where parameter redundancy and sensitivity to rank truncation vary across the input, output, and expert modes, and token-wise utilization variation, where hot and cold tokens exhibit distinct spectral characteristics and expert activation patterns. To address these challenges, we propose ITC-MoE, an Importance-guided Token-aware Compression framework for MoE DLMs. ITC-MoE consists of two complementary components. First, Importance-guided Adaptive Tucker Compression (IATC) incorporates activation and gradient importance into expert weight transformation, jointly factorizes expert weights across multiple modes, and adaptively allocates ranks under a fixed parameter budget. Second, Token-aware Compensation and Routing (TCR) applies lightweight low-rank compensation to compression-sensitive hot tokens and restricts the candidate expert set for cold tokens with concentrated routing patterns. By jointly adapting compression capacity and inference execution to both parameter redundancy and token-wise variation, ITC-MoE substantially reduces the computation and storage costs of MoE DLMs while preserving their generation quality. For example, on SDAR-30B-A3B-Chat-b32, ITC-MoE maintains an accuracy of 96.33% on MultiArith under a 30% compression budget, while achieving up to a 7.22x end-to-end speedup. The code is publicly available at https://github.com/lianjunl13-sudo/ITC-MoE.
Chinese Translation
混合专家（MoE）扩散语言模型（DLMs）提供灵活的并行解码和增强的模型容量，但其大量的专家参数带来了可观的计算和存储成本。现有的低秩MoE压缩方法主要依赖于静态因式分解和固定秩分配，忽视了MoE DLMs的独特特性。具体而言，我们识别出两个特性：跨模式非均匀冗余，其中参数冗余和对秩截断的敏感性在输入、输出和专家模式之间有所不同；以及基于令牌的利用率变化，其中热门令牌和冷门令牌表现出不同的谱特征和专家激活模式。为了解决这些挑战，我们提出了ITC-MoE，一个针对MoE DLMs的基于重要性引导的面向令牌的压缩框架。ITC-MoE由两个互补组件组成。首先，基于重要性引导的自适应塔克压缩（IATC）将激活和梯度重要性纳入专家权重变换，联合对多个模式下的专家权重进行因式分解，并在固定参数预算下自适应分配秩。其次，面向令牌的补偿与路由（TCR）对压缩敏感的热门令牌应用轻量级低秩补偿，并限制冷门令牌的候选专家集，以集中路由模式。通过联合适应压缩能力和推理执行，以应对参数冗余和基于令牌的变化，ITC-MoE显著降低了MoE DLMs的计算和存储成本，同时保持其生成质量。例如，在SDAR-30B-A3B-Chat-b32上，ITC-MoE在30%的压缩预算下在MultiArith上保持了96.33%的准确率，同时实现了高达7.22倍的端到端加速。代码已公开发布在https://github.com/lianjunl13-sudo/ITC-MoE。
cs.AI / 88 / 2610.01297
Questionnaire-Guided Disaggregation of Energy Appliance Use for Domestic Smart Meter Data
问卷指导的家庭智能电表数据能耗设备使用分解
Abstract
Ireland's smart metering programme records electricity use at 30-minute resolution, with smart meters installed in over 80\% of households as of late 2025. While this is useful for billing of smart, time-of-use tariffs, it is too coarse to capture use of domestic appliances. We present a label-free disaggregation system that breaks usage data into 9 appliance categories by combining event detection for high-power loads with questionnaire-guided estimation. Our evaluation draws on four datasets: a calibration household with a commercial comparator, two public benchmarks (UK-DALE and REFIT) with per-appliance sub-metering, and a smart meter dataset of more than 4,800 years of use from 2,968 Irish consumers. Compared against two independently developed disaggregation systems our hybrid method combining analysis of usage data with questionnaire results, achieves the lowest whole-decomposition error on all buildings across the datasets, with better month-level performance over 54 paired months ($p<0.001$, Holm-corrected). Our method provides useful advice on a household's energy consumption patterns and advice on how to reduce or shift usage on some appliances in order to reduce costs.
Chinese Translation
爱尔兰的智能计量项目以30分钟的分辨率记录电力使用，截至2025年底，超过80%的家庭已安装智能电表。尽管这对于智能时段计费非常有用，但其分辨率过于粗糙，无法捕捉家庭电器的使用情况。我们提出了一种无标签的分解系统，通过结合高功率负载的事件检测与问卷指导估计，将使用数据分解为9个电器类别。我们的评估基于四个数据集：一个与商业对比的校准家庭，两个具有每个电器子计量的公共基准（UK-DALE和REFIT），以及来自2,968名爱尔兰消费者的超过4,800年使用记录的智能电表数据集。与两个独立开发的分解系统相比，我们的混合方法结合了使用数据分析与问卷结果，在所有建筑物上实现了最低的整体分解误差，并在54对月份的月度表现上更佳（$p<0.001$，经过Holm校正）。我们的方法为家庭的能耗模式提供了有用的建议，并就如何减少或调整某些电器的使用以降低成本提供了指导。
cs.AI / 89 / 2610.01306
DAYJOB: A Benchmark for Long-Horizon Professional Work
DAYJOB：长期专业工作的基准测试
Abstract
Professional work often starts with a brief request that leaves the professional to work out what is needed, which documents matter, and whether the request's premise holds. We introduce DAYJOB, a benchmark of 130 tasks built by professionals in healthcare (50) and finance (80). The tasks are estimated to take a professional 13.6 hours on average in healthcare and 16.6 in finance. Each task is a containerized Harbor environment with an expert rubric of binary criteria (median 47.5 and 57.5 per task) that an agentic judge applies to the delivered files, and an attempt passes only if it meets every criterion. Across 30 model configurations from 13 developers, the strongest, Claude Opus 5.5, passes 24.7% of healthcare and 23.9% of finance attempts, and the median configuration passes 0.6% and 2.5%. In case studies, agents accept premises that the record contradicts and carry wrong inputs through otherwise consistent analyses. We release all healthcare tasks, 50 of the 80 finance tasks, the evaluation harness, and the leaderboard.
Chinese Translation
专业工作通常始于一个简短的请求，这使得专业人员需要自行判断所需的内容、相关的文件以及请求的前提是否成立。我们介绍了DAYJOB，这是一个由医疗保健（50个任务）和金融（80个任务）领域的专业人士构建的130个任务的基准测试。这些任务在医疗保健领域平均需要专业人员花费13.6小时，而在金融领域则需要16.6小时。每个任务都是一个容器化的Harbor环境，配备了专家制定的二元标准（每个任务的中位数为47.5和57.5），一个代理评审者将这些标准应用于提交的文件，只有满足所有标准的尝试才会被视为通过。在来自13个开发者的30种模型配置中，最强的模型Claude Opus 5.5在医疗保健领域的通过率为24.7%，在金融领域为23.9%，而中位数配置的通过率分别为0.6%和2.5%。在案例研究中，代理接受与记录相矛盾的前提，并在其他一致的分析中传递错误的输入。我们发布了所有医疗保健任务、80个金融任务中的50个、评估工具和排行榜。
cs.AI / 90 / 2610.01320
ProtoFlow: Prototype-Guided Flow Matching for Multivariate Time Series Forecasting
ProtoFlow：基于原型引导的多变量时间序列预测流匹配
Abstract
Generative modeling has shown strong promise for multivariate time mseries (MTS) forecasting, especially scale to high-dimensional settings. Diffusion-based methods achieve competitive performance but typically require many sampling steps at inference. VAE-based non-iterative forecasting frameworks have therefore emerged as an efficient alternative. Within this line of work, vector quantization (VQ) enables controllable latent space modeling by mapping multivariate series into compact discrete representations. Existing VQ-based forecasting methods, however, typically rely on autoregressive (AR) token generation, which suffers from exposure bias and training-inference mismatch. Flow matching provides an efficient non-autoregressive alternative for latent forecasting, but existing formulations usually initialize transport from a generic Gaussian prior. We instead observe that the trained VQ codebook already captures representative latent prototypes and can thus serve as a more informative prior for flow matching. Based on this insight, we propose ProtoFlow, a forecasting framework that combines vector-quantized autoencoding with Prototype-prior Flow matching. Our method first maps multivariate sequences into a discrete latent space, then constructs a structured prior from the learned codebook, and finally learns a DiT-based rectified flow to transport samples from this prior to future latent representations conditioned on historical observations. By replacing generic noise initialization with a learned prototype prior, ProtoFlow avoids the rollout mismatch of AR token prediction and promotes faster training convergence. Extensive experiments on benchmark datasets show that it consistently achieves superior forecasting performance with efficient inference.
Chinese Translation
生成建模在多变量时间序列（MTS）预测中展现出强大的潜力，尤其是在高维设置下。基于扩散的方法表现出竞争力，但通常在推理时需要多个采样步骤。因此，基于变分自编码器（VAE）的非迭代预测框架作为一种高效的替代方案应运而生。在这一研究方向中，向量量化（VQ）通过将多变量序列映射到紧凑的离散表示，实现了可控的潜在空间建模。然而，现有的基于VQ的预测方法通常依赖于自回归（AR）标记生成，这会导致曝光偏差和训练-推理不匹配。流匹配为潜在预测提供了一种高效的非自回归替代方案，但现有的公式通常从通用高斯先验初始化传输。我们观察到，训练后的VQ代码本已经捕捉到了代表性的潜在原型，因此可以作为流匹配的更具信息性的先验。基于这一见解，我们提出了ProtoFlow，一个将向量量化自编码与原型先验流匹配相结合的预测框架。我们的方法首先将多变量序列映射到离散潜在空间，然后从学习到的代码本构建结构化先验，最后学习基于DiT的修正流，以将样本从该先验传输到基于历史观测条件的未来潜在表示。通过用学习到的原型先验替代通用噪声初始化，ProtoFlow避免了AR标记预测的展开不匹配，并促进了更快的训练收敛。在基准数据集上的大量实验表明，它在高效推理的同时，始终实现了卓越的预测性能。
cs.AI / 91 / 2610.01323
TRACE: Trajectory Return Attribution and Contrastive Erasure for Multi-Turn Safety
TRACE：多轮安全的轨迹回报归因与对比性消除
Abstract
Safety-aligned large language models (LLMs) often refuse a harmful request but comply once the same goal is spread over several turns. Preference objectives score whole responses to single prompts, so their training loss alone cannot control risk on unseen histories. Our analysis gives sufficient conditions under which suppression at supervised single-turn contexts yields a bound on multi-turn trajectory risk. The bound accounts for coverage, transfer slack, and leakage, and characterizes contraction relative to a base-policy risk budget evaluated on the trained policy's contexts. TRACE (Trajectory Return Attribution and Contrastive Erasure) turns this principle into a token-level objective. On the safe response, each token is weighted by the discounted return of a refusal-attributable advantage. The advantage compares a frozen reference model with its refusal-ablated copy, allowing earlier response tokens to receive credit from later refusal-related evidence. At high-gap positions on rejected responses, TRACE combines the observed token with policy-selected alternatives in the erasure target. A gradient-norm penalty replaces the retain set. Across five open-weight models and seven multi-turn attacks, TRACE gives the lowest attack success rate (ASR) in all 35 model and attack pairs, while the model utility evaluated on MMLU and HellaSwag drop by at most 1\.23 points. Source code can be found in the supplemental material.
Chinese Translation
安全对齐的大型语言模型（LLMs）通常会拒绝有害请求，但一旦相同目标分散在多个回合中则会顺从。偏好目标对单个提示的整体响应进行评分，因此其训练损失本身无法控制未见历史的风险。我们的分析提供了在监督的单轮上下文中进行抑制时，能够对多轮轨迹风险施加界限的充分条件。该界限考虑了覆盖率、转移松弛和泄漏，并相对于在训练政策上下文中评估的基础政策风险预算特征化收缩。TRACE（轨迹回报归因与对比性消除）将这一原则转化为基于令牌的目标。在安全响应中，每个令牌的权重由拒绝可归因优势的折现回报决定。该优势比较了一个冻结的参考模型与其拒绝消除副本，允许较早的响应令牌从后续与拒绝相关的证据中获得信用。在被拒绝响应的高差距位置，TRACE将观察到的令牌与政策选择的替代方案结合在消除目标中。梯度范数惩罚替代了保留集。在五个开放权重模型和七个多轮攻击中，TRACE在所有35个模型和攻击对中提供了最低的攻击成功率（ASR），而在MMLU和HellaSwag上评估的模型效用最多下降1.23分。源代码可在补充材料中找到。
cs.AI / 92 / 2610.01325
PPO-HRAP: Proximal Policy Optimization with a Hybrid Regime-Aware Policy for Risk-Controlled Trading
PPO-HRAP：一种结合混合状态感知策略的近端策略优化用于风险控制交易
Abstract
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.
Chinese Translation
交易中的强化学习常常难以平衡收益参与与回撤控制。仅追求利润的策略可能在向上漂移的资产中崩溃为被动的多头暴露，而过于激进的风险惩罚奖励在波动期可能变得过于防御性。本文提出了PPO-HRAP，一种混合状态感知策略，它将近端策略优化（Proximal Policy Optimization）与可解释的状态先验相结合。代理观察市场特征和投资组合状态变量，接收结合投资组合对数收益、基于VIX的回撤增加惩罚、目标暴露偏差和周转成本的奖励，并在PPO演员输出与基于状态的目标暴露之间执行混合动作。在2020-2022年持有的SPY测试窗口中，PPO-HRAP实现了27.62%的总回报、8.48%的年化回报、0.6447的夏普比率、0.8588的索提诺比率和0.4592的卡尔玛比率，同时将最大回撤从34.10%（买入并持有策略）降低到18.47%。在五个SPY种子上，PPO-HRAP保持稳定，平均总回报为$0.2725 imes 0.0109$，平均夏普比率为$0.6219 imes 0.0565$。在QQQ和DIA的单次跨资产测试中，所提出的方法在所有三种报告资产的总回报和夏普比率上均排名第一。这些结果表明，将学习到的动作与波动性感知的状态先验相结合是一种改善风险调整交易行为的实用方法，尽管当前策略仍然导致高周转，并且在SPY之外的跨资产稳健性仍然限于单次证据。
cs.AI / 93 / 2610.01326
An ontology for cross-sectoral crisis management: core and public health modules
跨部门危机管理本体：核心与公共卫生模块
Abstract
This paper presents the European Crisis Management Ontology (ECMO), a modular OWL-based ontology intended as a cross-sectoral reference for disaster risk reduction and response. ECMO is designed to be organised as a network of ontological modules. Among the modules, ECMO-CORE captures fundamental crisis management concepts such as hazard, event, exposure, impact, and response measure and uses ontology design patterns and the OWL2 punning technique to resolve ambiguities between hazard types and event manifestations. In addition, domain-specific modules are defined as in the case of the public health module aligned with SNOMED CT and ICD-11. To demonstrate the resource's utility, we used ECMO to represent the data of the Epidemic Intelligence from Open Sources system of the Joint Research Centre to generate an end-to-end pipeline that populates an ECMO-compliant knowledge graph from unstructured epidemiological news. Initial results demonstrate that ECMO provides the formal guardrails necessary for consistent and unified knowledge representation and integration. The ontology is publicly available at https://doi.org/10.5281/zenodo.20070268 and is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.
Chinese Translation
本文介绍了欧洲危机管理本体（European Crisis Management Ontology, ECMO），这是一种基于模块化的OWL本体，旨在作为灾害风险减少与应对的跨部门参考。ECMO的设计为本体模块的网络结构。在这些模块中，ECMO-CORE捕捉了基本的危机管理概念，如危险、事件、暴露、影响和应对措施，并利用本体设计模式和OWL2的双重命名技术来解决危险类型与事件表现之间的歧义。此外，还定义了特定领域的模块，例如与SNOMED CT和ICD-11对齐的公共卫生模块。为了展示该资源的实用性，我们使用ECMO表示联合研究中心的开放源流行病情报系统的数据，以生成一个端到端的数据处理管道，从非结构化的流行病学新闻中填充符合ECMO标准的知识图谱。初步结果表明，ECMO提供了必要的正式框架，以确保一致和统一的知识表示与整合。该本体可在https://doi.org/10.5281/zenodo.20070268上公开获取，并根据知识共享署名 4.0 国际（CC BY 4.0）许可证发布。
cs.AI / 94 / 2610.01348
Verify Claims, Not Scores: Evidence-Based Verification of Modular Agents
验证声明，而非分数：基于证据的模块化代理验证
Abstract
When developers change one component of an agent, such as its controller, a learned model or its verifier, they usually judge the change by an aggregate task score. That score cannot tell whether improvement was attainable, which component lost value, or what the agent's own checks certify. We introduce a claim-specific verification audit for modular agents that plan, act, check and refine. Instead of scoring the agent, the audit scores the evidence: each conclusion is recorded with the evidence behind it, one of four verdicts (supported, unsupported, unresolved or not evaluated) and the boundary within which it holds. Three tools supply that evidence. Oracle policies measure attainable improvement under an explicitly stated action set, so that a low value can be traced to the evaluation rather than to the environment. Replacing one component at a time with a perfect counterpart locates lost value, with null results read as unresolved whenever a downstream component could mask them. A separate test asks whether the verifier's score identifies the quantity it is read as bounding. Applied to a constrained portfolio-allocation agent in a synthetic market with known hidden regimes, the audit shows that the value of perfect regime information depends on the action set used to measure it, that the scenario generator discards most of the regime signal while better local fidelity does not improve decisions, and that the runtime verifier can be bypassed with no visible change in outcomes. The contribution is the protocol and the evidential distinctions it enforces; the empirical findings are specific to the agent and environment studied.
Chinese Translation
当开发者更改代理的某个组件，例如其控制器、学习模型或验证器时，他们通常通过一个综合任务分数来评估该更改。然而，该分数无法判断是否可以实现改进、哪个组件失去了价值，或代理自身的检查所认证的内容。我们为规划、行动、检查和改进的模块化代理引入了一种特定于声明的验证审计。该审计不是对代理进行评分，而是对证据进行评分：每个结论都记录了其背后的证据、四种裁决之一（支持、不支持、未解决或未评估）以及其适用的边界。三种工具提供这些证据。Oracle 策略在明确声明的行动集下测量可实现的改进，因此低值可以追溯到评估而非环境。逐个替换组件为完美的对应物可以定位失去的价值，当下游组件可能掩盖时，零结果被视为未解决。另一个测试询问验证器的分数是否识别其所界定的数量。应用于一个在已知隐藏状态的合成市场中的受限投资组合分配代理，该审计显示完美状态信息的价值取决于用于测量它的行动集，场景生成器丢弃了大部分状态信号，而更好的局部保真度并未改善决策，并且运行时验证器可以被绕过而结果没有明显变化。贡献在于所制定的协议及其强制执行的证据区分；实证发现特定于所研究的代理和环境。
cs.AI / 95 / 2610.01378
Generation Provenance Before Behavior Attribution: Auditing Synthetic Speech Research Objects
行为归因之前的生成来源：审计合成语音研究对象
Abstract
Attributing model behavior to synthetic training data requires knowing what produced each training item before estimating what that item caused. A waveform-label pair does not preserve this knowledge. We propose a generation-provenance substrate in which a synthetic research object binds source specification, generated content, waveform, target, fact requirements, quality signals, review lineage, and immutable manifest identity. Producer and selection mechanism determine evidentiary meaning; storage location and variable name do not. We audit this substrate in a private Japanese care-handoff pipeline. A 113-asset review population contains 1.552 hours of synthetic speech across six scenario families; all items have linked audio, transcripts, candidate notes, and fact checklists, but human evidence is selective and source-specific. Two faithful-only manifests are scenario-seed-disjoint and immutably versioned, while exact upstream attribution remains blocked by floating generator aliases, missing per-clip TTS and code stamps, and an unversioned checking prompt. We argue that generation provenance is necessary but not sufficient for behavior attribution: it defines the candidate causal graph and audit units, whereas contributive attribution still requires frozen training runs and intervention or influence evidence. The paper contributes a compact provenance contract, an audit protocol, and a bounded case study for synthetic-data attribution; controlled research access may be offered, but we do not claim causal training-data attribution, clinical validity, or unrestricted public release.
Chinese Translation
将模型行为归因于合成训练数据需要在估计该项数据造成的影响之前，了解每个训练项的生成来源。波形-标签对并不能保留这一知识。我们提出了一种生成来源基础结构，其中合成研究对象绑定了源规范、生成内容、波形、目标、事实要求、质量信号、审查沿革和不可变的清单身份。生产者和选择机制决定了证据的意义；存储位置和变量名称则不然。我们在一个私有的日本护理交接流程中审计了这一基础结构。113个资产的审查样本包含了来自六个场景系列的1.552小时合成语音；所有项目都有链接的音频、转录文本、候选笔记和事实检查清单，但人类证据是选择性的且特定于来源。两个仅限真实的清单在场景种子上是互不重叠的，并且版本不可变，而确切的上游归因仍然受到浮动生成器别名、缺失的每个片段的文本转语音（TTS）和代码印记，以及未版本化的检查提示的阻碍。我们认为生成来源对于行为归因是必要但不充分的：它定义了候选因果图和审计单元，而贡献性归因仍然需要冻结的训练运行和干预或影响证据。本文贡献了一个紧凑的来源合同、一个审计协议和一个合成数据归因的有限案例研究；可能会提供受控的研究访问，但我们不声称因果训练数据归因、临床有效性或无限制的公开发布。
cs.AI / 96 / 2610.01382
Gacha Decoding: Eliciting Diverse Generations Through Instruction Following
扭蛋解码：通过遵循指令引发多样化生成
Abstract
We introduce Gacha Decoding, an inference-time method for eliciting diverse language model generations that scales with model capability. Across open-ended domains (in-the-wild chat, creative writing, planning for image generation, and protein design), Gacha Decoding significantly outperforms existing generation diversity approaches at equal quality (up to 2.4x Vendi over the next-best prior approach), reaching the same number of high-quality modes with over an order of magnitude fewer samples (11.0x) and discovering novel modes that no other approach surfaces. Our key insight is to treat diversity as an instruction-following problem: rather than relying on the LM's token entropy, we combine its instruction-following capability with randomness from an external RNG tool to scalably identify and realize distinct modes of the response space. This approach of "planning with dice" enables Gacha to invert the long-observed tension between diversity and model capability. As the underlying LM becomes a better instruction follower, diversity under Gacha Decoding consistently improves--even as its token entropy and diversity under prior approaches decline. Together, our results highlight that instruction following, rather than token entropy alone, can drive generation diversity.
Chinese Translation
我们提出了扭蛋解码（Gacha Decoding），这是一种在推理时引发多样化语言模型生成的方法，能够根据模型能力进行扩展。在开放式领域（如自然对话、创意写作、图像生成规划和蛋白质设计）中，扭蛋解码在质量相当的情况下显著优于现有的生成多样性方法（相较于最佳的前期方法，Vendi提升高达2.4倍），以超过一个数量级更少的样本（11.0倍）达到相同数量的高质量模式，并发现了其他方法未能揭示的新模式。我们的关键见解是将多样性视为一个遵循指令的问题：我们结合语言模型的指令遵循能力与外部随机数生成工具的随机性，以可扩展的方式识别和实现响应空间中的不同模式。这种“用骰子规划”的方法使得扭蛋能够逆转长期观察到的多样性与模型能力之间的紧张关系。随着基础语言模型成为更好的指令跟随者，扭蛋解码下的多样性持续改善——即使其标记熵和先前方法下的多样性下降。总的来说，我们的结果强调，遵循指令而非仅仅依赖标记熵，可以推动生成的多样性。
cs.AI / 97 / 2610.01383
PRISM: A Category-Theoretic Framework for Measuring and Refining Multimodal Analogies
PRISM：用于测量和优化多模态类比的范畴理论框架
Abstract
Analogical reasoning involves identifying and preserving relational structures across domains. However, existing approaches to AI-driven multimodal analogy generation lack an interpretable measure of whether this structure is understood and maintained in the generated output. We address this gap with Pullback Refinement via Interpretable Structural Mapping (PRISM), a modality-agnostic framework for measuring and improving relational alignment in multimodal analogies, evaluated on visual metaphor generation. PRISM represents analogies as explicit relational mappings grounded in category theory and uses VLMs to instantiate these structures across modalities. Its first component, the pullback score, quantifies relational alignment from the resulting graph representation. On the AnaloBench benchmark, selecting the correct analogy purely by pullback score achieves 82.5% accuracy, demonstrating that the score captures meaningful relational information. PRISM's second component is an iterative refinement loop that uses the pullback score as an in-context feedback signal to iteratively revise the generated image towards greater relational depth. VLM-as-a-judge and human evaluations show that PRISM consistently improves metaphor consistency and analogy appropriateness over zero- shot generation, with human participants preferring the refined output in 57.65% of pairwise comparisons. However, a qualitative analysis reveals that refinement can favour visually crowded compositions rather than genuinely deeper relational correspondences.
Chinese Translation
类比推理涉及在不同领域中识别和保留关系结构。然而，现有的基于人工智能的多模态类比生成方法缺乏可解释的度量标准，以判断生成输出中是否理解并保持了这种结构。我们通过可解释结构映射的拉回优化（Pullback Refinement via Interpretable Structural Mapping，PRISM）来填补这一空白，PRISM是一个与模态无关的框架，用于测量和改善多模态类比中的关系对齐，评估基于视觉隐喻生成。PRISM将类比表示为基于范畴理论的显式关系映射，并利用视觉语言模型（VLMs）在不同模态中实例化这些结构。其第一个组成部分，拉回得分，量化了从生成的图形表示中获得的关系对齐。在AnaloBench基准测试中，仅通过拉回得分选择正确的类比达到了82.5%的准确率，表明该得分捕捉到了有意义的关系信息。PRISM的第二个组成部分是一个迭代优化循环，利用拉回得分作为上下文反馈信号，迭代修正生成的图像，以实现更深的关系深度。VLM作为评判者和人类评估表明，PRISM在零样本生成的基础上，始终改善隐喻的一致性和类比的适当性，人类参与者在57.65%的成对比较中更倾向于选择优化后的输出。然而，定性分析显示，优化可能更倾向于视觉上拥挤的构图，而不是确实更深的关系对应。
cs.AI / 98 / 2610.01389
AiSearch: Interactive Multi-Modal Search with VLMs
AiSearch：基于视觉语言模型的交互式多模态搜索
Abstract
Modern retrieval systems must both be automated and interactive, allowing users to search and refine results in real time. We present AiSearch, a flexible multimodal retrieval framework that leverages the zero shot capabilities of Vision Language Models (VLMs) for natural language search over images and videos. AiSearch supports interactive search refinement through user feedback to tailor results to the user's intent, and allows visual benchmarking across multiple VLMs, enabling users to select the most suitable model for their task.
Chinese Translation
现代检索系统必须既具备自动化又具备交互性，允许用户实时搜索和优化结果。我们提出了AiSearch，这是一种灵活的多模态检索框架，利用视觉语言模型（VLMs）的零样本能力，实现对图像和视频的自然语言搜索。AiSearch通过用户反馈支持交互式搜索优化，以便根据用户的意图调整结果，并允许在多个VLM之间进行视觉基准测试，使用户能够选择最适合其任务的模型。
cs.AI / 99 / 2610.01403
Contrastive Attention Mitigates Spectral Bias in Spiking Transformers
对比注意力缓解脉冲变换器中的谱偏差
Abstract
Spiking Transformers merge the energy-efficiency of spiking neural networks (SNNs) with the representational power of self-attention, creating a promising architecture for high-performance, energy-efficient computation. However, a performance gap persists versus its counterparts in artificial neural networks (ANNs). Unlike prior works attributing this to binary activations, we reveal that both spiking neurons and spiking self-attention (SSA) act as low-pass filters through multiscale spectral analysis. This characteristic leads to the dissipation of high-frequency components. To address this issue, we propose the Spiking Contrastive Attention (SCA) paradigm, which draw inspiration from the edge-detection and differential sensing properties of biological visual system. By extracting contrast prototypes via global contrastive aggregation and applying local differential refinement, SCA effectively enhances high-frequency information. Extensive experiments show that SCA is a general module that consistently boosts Spiking Transformers across image classification, semantic segmentation, and event-based tracking. Furthermore, it achieves lower complexity, offering superior efficiency over original SSA. These results establish its potential as a fundamental building block for energy-efficient Spiking Transformers.
Chinese Translation
脉冲变换器将脉冲神经网络（SNNs）的能量效率与自注意力的表征能力相结合，创造了一种高性能、能效计算的有前景架构。然而，与人工神经网络（ANNs）相比，性能差距依然存在。与先前将此归因于二进制激活的研究不同，我们揭示了脉冲神经元和脉冲自注意力（SSA）通过多尺度谱分析作为低通滤波器的作用。这一特性导致高频成分的消散。为了解决这一问题，我们提出了脉冲对比注意力（SCA）范式，该范式受到生物视觉系统的边缘检测和差异感知特性的启发。通过全球对比聚合提取对比原型，并应用局部差异精炼，SCA有效增强了高频信息。大量实验表明，SCA是一个通用模块，能够在图像分类、语义分割和基于事件的跟踪中持续提升脉冲变换器的性能。此外，它实现了更低的复杂性，提供了优于原始SSA的效率。这些结果确立了其作为能效脉冲变换器基本构建块的潜力。
cs.AI / 100 / 2610.01415
Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States
超越记忆：利用显式信念状态的长远代理
Abstract
Large language model (LLM) agents can now undertake increasingly complex tasks, but the way they organize interaction history into memory does not ensure a coherent understanding of the current world. We introduce PoS, an inference-time framework that constructs and continually maintains explicit belief states as the agent's decision context. Each belief combines an estimate of the current world state with unresolved task requirements, making explicit what the agent still needs to learn and accomplish. To keep this belief reliable and actionable, PoS validates its consistency and monitors task progress to detect Belief Trapping, where the agent continues to act without making meaningful progress toward the goal. Recovery is then tailored to both the trapping pattern and the type of unresolved task requirement. Experiments on four benchmarks spanning execution and diagnosis show that PoS achieves the highest overall performance on every benchmark with all three LLM backbones. Ablations demonstrate the importance of consistency validation and recovery, while context-scaling experiments show resilience to context growth. Together, these results support belief construction and continual maintenance as a foundation for long-horizon context management beyond history retention and compression.
Chinese Translation
大型语言模型（LLM）代理现在可以执行越来越复杂的任务，但它们将交互历史组织成记忆的方式并不能确保对当前世界的连贯理解。我们引入了PoS（信念状态），这是一种推理时框架，构建并持续维护显式信念状态作为代理的决策背景。每个信念结合了对当前世界状态的估计和未解决的任务需求，明确了代理仍需学习和完成的内容。为了保持这一信念的可靠性和可操作性，PoS验证其一致性并监测任务进展，以检测信念陷阱（Belief Trapping），即代理在未能朝目标取得实质性进展的情况下继续行动。恢复策略则根据陷阱模式和未解决的任务需求类型进行调整。在涵盖执行和诊断的四个基准测试上的实验表明，PoS在所有三个LLM基础模型上在每个基准测试中都达到了最高的整体性能。消融实验显示了一致性验证和恢复的重要性，而上下文扩展实验则展示了对上下文增长的韧性。综合这些结果，支持信念构建和持续维护作为超越历史保留和压缩的长远上下文管理基础。
cs.AI / 101 / 2610.01418
SpikeMoE: Brain-Inspired Competitive Routing for Flexible Spiking Mixture-of-Experts
SpikeMoE：灵感来自大脑的竞争路由用于灵活的脉冲混合专家
Abstract
Spiking Neural Networks (SNNs) enable event-driven computation through biologically inspired dynamics at the neuronal scale, while Mixture-of-Experts (MoE) perform conditional computation through expert selection at the model scale. Integrating their strengths offers potential for flexible neural architectures. A key challenge, however, lies in designing an expert selection mechanism based on spiking activity. To address this, we introduce a spike-based k-WTA Router inspired by competition-inhibition observed in the hippocampal CA1 region. The router incorporates lateral inhibition and refractory period to select Top-K experts according to discrete spike counts. Building on this, we present SpikeMoE, a framework that integrates neuronal-scale spiking dynamics with model-scale expert selection. To address incomplete multisensory inputs in multimodal tasks, we further equip SpikeMoE with a two-stage missing-modality modeling module that combines empirical prototypes from an observed-modality pool with modality-specific learnable embeddings to construct missing-modality representations. Experiments on vision, language, and multimodal benchmarks demonstrate that SpikeMoE achieves state-of-the-art performance among the SNN baselines, matches or exceeds the performance of ANN counterparts, and maintains robustness across diverse missing-modality conditions. These results demonstrate a favorable trade-off between performance and energy efficiency, validating the integration of spiking dynamics with sparse expert computation and highlighting SpikeMoE as a promising approach to energy-efficient brain-inspired computing.
Chinese Translation
脉冲神经网络（SNNs）通过在神经元尺度上生物启发的动态实现事件驱动计算，而混合专家（MoE）则通过在模型尺度上进行专家选择来执行条件计算。整合它们的优势为灵活的神经架构提供了潜力。然而，一个关键挑战在于设计基于脉冲活动的专家选择机制。为了解决这个问题，我们引入了一种基于脉冲的 k-WTA 路由器，该路由器受到海马体 CA1 区域观察到的竞争抑制的启发。该路由器结合了侧向抑制和不应期，根据离散脉冲计数选择前 K 个专家。在此基础上，我们提出了 SpikeMoE，一个将神经元尺度的脉冲动态与模型尺度的专家选择相结合的框架。为了解决多模态任务中的不完整多感官输入，我们进一步为 SpikeMoE 配备了一个两阶段缺失模态建模模块，该模块结合了来自观察模态池的经验原型与模态特定的可学习嵌入，以构建缺失模态表示。在视觉、语言和多模态基准测试中的实验表明，SpikeMoE 在 SNN 基线中实现了最先进的性能，匹配或超过了 ANN 对应物的性能，并在多种缺失模态条件下保持了鲁棒性。这些结果展示了性能与能效之间的良好权衡，验证了脉冲动态与稀疏专家计算的整合，并突出了 SpikeMoE 作为一种有前景的能效大脑启发计算方法。
cs.AI / 102 / 2610.01436
A Deterministic and Auditable AI Security Risk Assessment Framework with ATLAS Aligned Executable Rules and Formal Verification
一个确定性和可审计的人工智能安全风险评估框架，具有与ATLAS对齐的可执行规则和形式验证
Abstract
Artificial intelligence systems are increasingly deployed in high impact and safety critical settings, yet security assessment remains difficult to reproduce and defend under audit. Existing approaches often rely on narrative checklists or assessor driven scoring, and they lack an explicit, machine evaluable mapping from observable engineering artefacts to stable technique level outcomes. We present an evidence driven AI security assessment framework that operationalises assessment as a deterministic decision function. The framework normalises heterogeneous artefacts into a project independent Control ID taxonomy scored on a bounded four level ordinal scale, compiles technique level predicates from a pinned MITRE ATLAS snapshot via an explicit mitigation to control mapping, and outputs technique indexed feasibility and impact levels with traceable links back to the triggering evidence. We package all normative choices as a versioned assessment policy object to support repeatable reassessment across snapshots. To ensure semantic correctness, we formally verify boundedness, totality, ordered semantic consistency, and monotonicity of the compiled evaluator over the full declared score domain. We evaluate the framework on five public open source AI projects pinned to explicit repository snapshots, quantify before and after changes under a unified hardening intervention, and validate responsiveness to real engineering changes through fork based implementations of Software Bill of Materials (SBOM) generation and Continuous integration (CI) security scanning gates. Results show consistent downward shifts in feasibility profiles under strengthened observable controls, while worst case residual feasibility persists when technique specific core controls remain absent from the evidence scope.
Chinese Translation
人工智能系统越来越多地在高影响和安全关键的环境中部署，但安全评估在审计下仍然难以重现和辩护。现有的方法通常依赖叙述性检查清单或评估者驱动的评分，并且缺乏从可观察的工程工件到稳定技术水平结果的明确、机器可评估的映射。我们提出了一种基于证据的人工智能安全评估框架，将评估操作化为一个确定性的决策函数。该框架将异构工件标准化为一个项目独立的控制ID分类法，并在一个有限的四级序数尺度上进行评分，从一个固定的MITRE ATLAS快照中编译技术水平谓词，通过明确的缓解与控制映射，输出技术索引的可行性和影响水平，并提供可追溯的链接回触发证据。我们将所有规范选择打包为一个版本化的评估政策对象，以支持跨快照的可重复再评估。为了确保语义正确性，我们正式验证了编译评估器在整个声明评分域上的有界性、完全性、顺序语义一致性和单调性。我们在五个公共开源人工智能项目上评估该框架，这些项目固定在明确的代码库快照上，量化在统一的强化干预下的前后变化，并通过基于分叉的材料清单（SBOM）生成和持续集成（CI）安全扫描门的实现验证对真实工程变化的响应。结果表明，在加强的可观察控制下，可行性轮廓持续向下移动，而在技术特定核心控制缺失于证据范围时，最坏情况下的剩余可行性仍然存在。
cs.AI / 103 / 2610.01439
DRelay: Global Draft Context for Prefix-Aware Parallel Speculative Decoding Repair
DRelay：用于前缀感知并行推测解码修复的全局草稿上下文
Abstract
Parallel drafting reduces the drafting overhead of speculative decoding for large language models (LLMs), but its gains remain limited by the accepted prefix length. Even when the correct token is present in the candidate pool, a single early selection error prevents subsequent predictions from being used. We propose DRelay, which uses global information from the entire draft block to perform prefix-aware selective repair of candidate selections before target-model verification. DRelay bases its decisions on candidate correlations and the selected path: a global reader extracts predictive information across positions for each candidate. While a causal selector combines candidate-level information extracted by the global read with the tokens selected at preceding positions to determine whether the native choice at the current position is consistent with the global evidence and the selected prefix. It then decides whether to retain or replace the token, thereby repairing early errors and extending the accepted prefix. We further jointly train the draft backbone and the selector, combining candidate-support learning with a repair objective, while weighting the repair loss according to each block position's potential contribution to the consecutive accepted prefix. Across eight diverse benchmarks on an H800 GPU, DRelay consistently improves both average acceptance length and end-to-end decoding performance over DFlash, Domino, and DSpark. Under SGLang serving, DRelay improves average end-to-end speedup over DFlash, Domino, and DSpark by 14.7%-16.8%, 8.7%-9.3%, and 8.1%-9.3%, respectively.
Chinese Translation
并行草拟减少了大型语言模型（LLMs）推测解码的草拟开销，但其收益仍然受到接受前缀长度的限制。即使在候选池中存在正确的标记，单个早期选择错误也会阻止后续预测的使用。我们提出了 DRelay，它利用整个草稿块的全局信息，在目标模型验证之前对候选选择进行前缀感知的选择性修复。DRelay 基于候选相关性和所选路径做出决策：全局读取器从每个候选的各个位置提取预测信息。同时，因果选择器结合全局读取提取的候选级信息与前面位置选择的标记，以确定当前位点的本地选择是否与全局证据和所选前缀一致。然后，它决定是保留还是替换该标记，从而修复早期错误并扩展接受的前缀。我们进一步联合训练草稿主干和选择器，将候选支持学习与修复目标结合，同时根据每个块位置对连续接受前缀的潜在贡献加权修复损失。在 H800 GPU 上的八个不同基准测试中，DRelay 在平均接受长度和端到端解码性能上始终优于 DFlash、Domino 和 DSpark。在 SGLang 服务下，DRelay 在端到端速度提升方面分别比 DFlash、Domino 和 DSpark 提高了 14.7%-16.8%、8.7%-9.3% 和 8.1%-9.3%。
cs.AI / 104 / 2610.01451
A Multi-Agent LLM Framework for Personalized Health Checkup Interpretation and Guidance
个性化健康体检结果解读与指导的多智能体大型语言模型框架
Abstract
Personalized interpretation of health checkup results requires reasoning across longitudinal records, medical knowledge, lifestyle guidance, and healthcare navigation. We present a multi-agent large language model (LLM) system that identifies multiple intents, maps each to a task-specific agent, executes them in parallel, and synthesizes their outputs. We compared answers generated in Single Agent and Multi Agent settings on 120 Korean compound queries combining two to four requirements, using synthetic health checkup records. The Multi Agent improved the weighted LLM-judge score from 1.695 to 1.797 (p = 0.027), and three additional LLM judges showed consistent improvements ($\Delta$ = +0.111 to +0.186, all p < 0.05). The gains came from usefulness, consistency, and the handling of every requirement in compound queries, whereas numerical accuracy and grounding improved significantly under only one of the four judges and medical safety did not differ, and critical failures occurred at similar rates (Single Agent 15.0% vs. Multi Agent 13.3%). Two human evaluators preferred Multi Agent in 66.7% and 68.3% of pairwise comparisons. Multi Agent execution increased latency and cost by 1.31$\times$ and 2.02$\times$, respectively. In exploratory subgroup analyses, the improvement was concentrated in queries involving personal-record lookup.
Chinese Translation
个性化解读健康体检结果需要对纵向记录、医学知识、生活方式指导和医疗导航进行推理。我们提出了一种多智能体大型语言模型（LLM）系统，该系统识别多个意图，将每个意图映射到特定任务的智能体，进行并行执行，并综合它们的输出。我们在120个结合两到四个需求的韩国复合查询中，比较了单智能体和多智能体设置下生成的答案，使用合成的健康体检记录。多智能体将加权LLM评审分数从1.695提高到1.797（p = 0.027），另外三位LLM评审员显示出一致的改进（$ ext{Δ}$ = +0.111至+0.186，均p < 0.05）。这些提升来自于有用性、一致性以及对复合查询中每个需求的处理，而数值准确性和基础性仅在四位评审员中的一位下显著改善，医疗安全没有差异，且关键失败的发生率相似（单智能体15.0%对比多智能体13.3%）。两位人类评估者在66.7%和68.3%的成对比较中偏好多智能体。多智能体执行将延迟和成本分别增加了1.31$ imes$和2.02$ imes$。在探索性子组分析中，改进主要集中在涉及个人记录查找的查询中。
cs.AI / 105 / 2610.01458
Rethinking Probability-Based Reinforcement Learning From Posterior Concentration
从后验集中度重新思考基于概率的强化学习
Abstract
Verifier-free reinforcement learning with probability-based rewards offers a promising way to train LLMs on general reasoning tasks where external verifiers are unavailable. Yet the reliability of these rewards, especially in long-horizon reasoning, remains underexplored. This work identifies a length-dependent failure mode of probability rewards, which we call the Posterior Concentration Phenomenon (PCP). We show that the probability of a reference answer conditioned on a reasoning trace often collapses to a low-variance interval as the trace becomes lengthy. This phenomenon results in nearly indistinguishable rewards, which, under GRPO-based settings, makes probability-based policy optimization unstable and inefficient. Motivated by this, we propose Reinforcement Learning with Concentration-aware Posterior Rewards (RLCPR), a verifier-free RL framework to explicitly account for PCP for better optimization stability and token efficiency. It has two components: uncertainty-aware data sampling, which reduces concentration-prone rollouts before generation, and concentration-aware regularization, which penalizes unnecessarily long traces when posterior rewards collapse. Extensive experiments show that, alongside higher token efficiency, RLCPR outperforms the state-of-the-art verifier-free RL baseline by up to 4.0% on six of seven benchmarks, including general-domain and mathematical reasoning challenges.
Chinese Translation
无验证器的基于概率奖励的强化学习为在外部验证器不可用的情况下训练大型语言模型（LLMs）进行一般推理任务提供了一种有前景的方法。然而，这些奖励的可靠性，尤其是在长时间推理中，仍然未得到充分探讨。本研究识别了一种依赖于长度的概率奖励失效模式，我们称之为后验集中现象（Posterior Concentration Phenomenon, PCP）。我们展示了，在推理轨迹变长时，条件于推理轨迹的参考答案的概率往往会收敛到一个低方差区间。这一现象导致奖励几乎无法区分，这在基于GRPO的设置下，使得基于概率的策略优化变得不稳定且低效。基于此，我们提出了考虑集中度的后验奖励的强化学习（Reinforcement Learning with Concentration-aware Posterior Rewards, RLCPR），这是一个无验证器的强化学习框架，旨在明确考虑PCP，以提高优化的稳定性和令牌效率。该框架包含两个组成部分：不确定性感知的数据采样，在生成之前减少易集中化的回滚；以及集中度感知的正则化，当后验奖励收敛时，惩罚不必要的长轨迹。大量实验表明，RLCPR在提高令牌效率的同时，在七个基准测试中的六个上超越了最先进的无验证器强化学习基线，提升幅度高达4.0%，包括一般领域和数学推理挑战。
cs.AI / 106 / 2610.01461
NextMe-800: Anticipating Personal Behavior from Months of Egocentric Video
NextMe-800：从数月的自我中心视频中预测个人行为
Abstract
We often plan ambitiously yet act habitually and wonder, in retrospect, whether we would have planned differently had we known what we would actually do. Hindsight offers a valuable perspective on past decisions, although we often wish we could have simulated hindsight at the moment of choosing. If a system could generate plausible trajectories from one's personal history, such previews might help people formulate more realistic plans and make better informed decisions. We introduce NextMe-800, an approximately 800-hour first-person dataset from one volunteer over 126 days with 1 Hz images, gaze, and audio, captioned at five hierarchical abstraction levels from atomic actions to major activities. We formulate personalized action anticipation as open-vocabulary K-step sequence prediction and construct NextAct, a 1,500-point benchmark combining NextMe-800 with the multi-person EgoLife dataset. Using an embedding-based soft edit distance as the metric, we evaluate how well different models can anticipate personal behavior across abstraction levels and prediction horizons. NextMe-800 and NextAct provide a months-long resource and evaluation framework for studying how far ahead personal behavior can be anticipated from egocentric observation.
Chinese Translation
我们常常雄心勃勃地规划，却又习惯性地行动，并在事后反思，如果我们早知道自己会实际做什么，是否会有不同的规划。事后诸葛亮为过去的决策提供了宝贵的视角，尽管我们常常希望能在选择时模拟这种事后视角。如果一个系统能够从个人历史中生成合理的轨迹，这样的预览可能会帮助人们制定更现实的计划并做出更明智的决策。我们介绍了NextMe-800，这是一个由一名志愿者在126天内收集的约800小时的第一人称数据集，包含1 Hz的图像、注视和音频，并在从原子动作到主要活动的五个层次的抽象级别上进行了标注。我们将个性化的动作预测形式化为开放词汇的K步序列预测，并构建了NextAct，这是一个结合NextMe-800和多人的EgoLife数据集的1,500点基准。使用基于嵌入的软编辑距离作为度量标准，我们评估了不同模型在不同抽象级别和预测范围内预测个人行为的能力。NextMe-800和NextAct为研究从自我中心观察中预测个人行为的前瞻性提供了数月的资源和评估框架。
cs.AI / 107 / 2610.01495
Auditing Routing Entropy as an Uncertainty Signal in Attention-Residual Transformers
审计路由熵作为注意力残差变换器中的不确定性信号
Abstract
Dynamic architectures leave a per-example routing trace beside each prediction, and diffuse routing is easy to read as a sign that the prediction is unreliable. We audit that reading for routing entropy in Attention-Residual (AR) variants of Swin-Tiny and DeiT-Small, trained from scratch on CIFAR-10/100 with a soft-binned calibration auxiliary loss, asking whether the trace carries information about correctness beyond what the model's own confidence already reveals. Three checks probe this increment: does a routing signal appear at fixed confidence, does it replicate across training seeds, and can a held-out predictor exploit it against output-only and shuffled-trace controls? A sensitivity audit then injects effects of known size and measures the fraction of each that the probes recover. No test in the fixed 30-test binned family survives multiplicity correction, and neither the nominal hit nor a borderline result recurs in its sibling seeds. Across 24 paired runs a scalar routing probe yields no pooled improvement in routing-stratified calibration, and an entropy-profile probe predicts correctness better than the same probe given shuffled profiles yet worse than a confidence-only predictor in both binary log-loss and Brier score: a gain over shuffled traces does not become a gain over the output. Conditioning on the complete logit vector leaves the corresponding comparison unresolved. The audit bounds how far these non-detections can be read: at an injected effect of 0.010 nats the profile probe recovers 24-59% of the oracle gain, and a reference-preserving correction probe recovers 8% and 23% in the two CIFAR-100 settings, below the threshold we fixed for applying it to real labels. The results establish control-dependent gains and incomplete estimator recovery, not the absence of conditional routing information.
Chinese Translation
动态架构在每个预测旁留下一个逐例路由痕迹，而扩散路由易于被解读为预测不可靠的信号。我们审计了在从头开始在CIFAR-10/100上训练的Swin-Tiny和DeiT-Small的注意力残差（AR）变体中的路由熵，使用软分箱校准辅助损失，询问该痕迹是否携带关于正确性的额外信息，超出模型自身信心所揭示的内容。三个检查探讨了这一增量：在固定信心下路由信号是否出现，它是否在训练种子之间复制，以及一个保留的预测器是否能够利用它对抗仅输出和随机痕迹控制？然后，灵敏度审计注入已知大小的效应，并测量每个效应的探测恢复比例。在固定的30个测试分箱家族中，没有任何测试在多重性校正后存活，而名义命中率或边缘结果在其兄弟种子中也没有重现。在24次配对运行中，一个标量路由探针在路由分层校准中没有产生任何合并改进，而一个熵轮廓探针的正确性预测优于给定随机轮廓的同一探针，但在二元对数损失和布赖尔分数中却不如仅基于信心的预测器：相较于随机痕迹的增益并没有转化为对输出的增益。对完整的logit向量进行条件处理使得相应的比较未能解决。审计界定了这些未检测到的结果可以被解读的程度：在注入效应为0.010 nats时，轮廓探针恢复了24-59%的oracle增益，而一个保持参考的校正探针在两个CIFAR-100设置中恢复了8%和23%，低于我们为其应用于真实标签所设定的阈值。结果确立了依赖控制的增益和不完全的估计器恢复，而不是条件路由信息的缺失。
cs.AI / 108 / 2610.01497
OpenMTB-Audit: Exposing Over-Refusal and Clinical Expert Perspectives in LLM-Based Molecular Tumor Board Safety Evaluation
OpenMTB-Audit：揭示基于大型语言模型的分子肿瘤委员会安全评估中的过度拒绝和临床专家视角
Abstract
Molecular tumor boards integrate genomic findings, clinical context, and therapeutic evidence to support precision oncology. As AI enters this workflow, a key safety challenge is distinguishing truly unsupported recommendations from evidence-supported options that still require oncologist review because of incomplete information, poor ECOG performance status, or other clinical caveats. We introduce OpenMTB-Audit, an open-source benchmark of 500 synthetic non-small cell lung cancer cases spanning five adversarial error categories and four safety labels: Supported, Partially Supported, Unsupported, and Insufficient Information. Across eight large language model configurations, we identify pervasive over-refusal: all LLM configurations failed to retain the Partially Supported label in 83.3-100% of true Partially Supported cases, achieving high aggregate safety scores through label collapse rather than clinically calibrated reasoning. To address this limitation, we developed MTB-AuditAgent, a deterministic seven-module framework separating evidence verification, missing-information detection, safety classification, and abstention. It reduces over-refusal to 6.7% and achieves 91.2% accuracy (95% CI: 88.6-93.6%). A two-oncologist annotation study found disagreement concentrated at the boundary between information sufficiency and treatment optimization, underscoring the need to preserve clinically meaningful distinctions.
Chinese Translation
分子肿瘤委员会整合基因组发现、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不支持的建议与仍需肿瘤医生审查的证据支持选项，后者可能由于信息不完整、ECOG表现状态差或其他临床警告而需要审查。我们介绍了OpenMTB-Audit，这是一个包含500个合成非小细胞肺癌病例的开源基准，涵盖五种对抗性错误类别和四个安全标签：支持、部分支持、不支持和信息不足。在八种大型语言模型配置中，我们发现普遍存在过度拒绝现象：所有LLM配置在83.3%-100%的真实部分支持案例中未能保留部分支持标签，且通过标签合并而非临床校准推理获得了高的整体安全分数。为了解决这一局限性，我们开发了MTB-AuditAgent，这是一个确定性的七模块框架，分离证据验证、缺失信息检测、安全分类和弃权。它将过度拒绝减少至6.7%，并实现了91.2%的准确率（95% CI：88.6-93.6%）。一项由两位肿瘤医生进行的标注研究发现，分歧集中在信息充分性与治疗优化的边界上，强调了保持临床意义区分的必要性。
cs.AI / 109 / 2610.01506
MCRI: A Four-Dimensional Framework for Analyzing and Evaluating Agent Skills
MCRI：一种用于分析和评估智能体技能的四维框架
Abstract
As agents evolve from single-tool systems into modular, composite architectures, skills are becoming an important mechanism for capability development and distribution. However, the academic community lacks a structured framework for systematically analyzing and evaluating skills. Drawing on information gain and behavioral constraint, we propose the four-dimensional MCRI Framework and operationalize it as MCRI-Eval, a large language model-based evaluation method. We evaluate MCRI-Eval using 63,812 public skills from the OpenClaw skill Hub, with 58,275 skill-conditioned model executions across BigCodeBench, BFCL-Fundamental, and Mind2Web. MCRI-Eval scores are positively associated with community popularity signals and achieve the highest downstream ranking agreement among the evaluated methods. MCRI-Eval also improves top-1 skill selection across all three benchmarks: compared with the strongest baseline on each benchmark, the skills selected by MCRI-Eval advance by 17.7, 22.8, and 19.6 percentile points in downstream performance rank on BigCodeBench, BFCL-Fundamental, and Mind2Web, respectively. These results indicate that MCRI-Eval provides a useful pre-execution signal for prioritizing promising skills before costly execution-based evaluation.
Chinese Translation
随着智能体从单一工具系统演变为模块化复合架构，技能正成为能力发展和分配的重要机制。然而，学术界缺乏一个系统化的框架来分析和评估技能。基于信息增益和行为约束，我们提出了四维MCRI框架，并将其操作化为MCRI-Eval，一种基于大型语言模型的评估方法。我们使用来自OpenClaw技能中心的63,812个公共技能对MCRI-Eval进行评估，在BigCodeBench、BFCL-Fundamental和Mind2Web上进行了58,275次技能条件模型执行。MCRI-Eval得分与社区受欢迎程度信号呈正相关，并在评估的方法中实现了最高的下游排名一致性。MCRI-Eval还提高了所有三个基准的Top-1技能选择：与每个基准上最强的基线相比，MCRI-Eval选择的技能在BigCodeBench、BFCL-Fundamental和Mind2Web的下游性能排名中分别提升了17.7、22.8和19.6个百分位点。这些结果表明，MCRI-Eval为在成本高昂的执行基础评估之前优先考虑有前景的技能提供了有用的执行前信号。
cs.AI / 110 / 2610.01509
Sharpening Tax in Post-Training
后训练中的精细化税
Abstract
An emerging hypothesis about reinforcement learning (RL) post-training of large language models (LLMs) is that it merely sharpens existing behaviors of a base model, improving single-shot accuracy at the cost of solution coverage. Although this trade-off has been observed in math and coding tasks, it need not extend to agentic tasks, where multi-turn tool use and interaction may require capabilities newly acquired during post-training. Our surprising finding is that pre-trained LLMs, equipped with a light inference harness, can serve as capable agents. Despite far lower accuracy (pass@1), they often surpass their post-trained counterparts in solution coverage (pass@K) given a sufficient test-time budget. We further analyze the underlying mechanism and show that post-training pushes tasks toward two extremes, always solved or never solved, and thereby improves sampling efficiency and consistency at the cost of solution coverage. To measure this cost, we propose Sharpening Tax, a diagnostic metric that quantifies the loss in test-time scalability after post-training. Across 14 base/post-trained model pairs from four families and three agentic benchmarks (42 cases in total), the tax is prevalent in most settings, can be estimated from a few rollouts, and correlates well with other metrics. Finally, we present posterior-tempered group sampling (PTGS), a simple plug-and-play Bayesian sampler that adapts the sampling temperature per prompt to its estimated difficulty. Applied during RL training in two agentic environments, PTGS pays a smaller tax than the fixed-temperature baseline, solving more tasks under repeated sampling while also improving single-shot accuracy.
Chinese Translation
关于大型语言模型（LLMs）后训练的强化学习（RL）新兴假设是，它仅仅是对基础模型现有行为的精细化，提升单次准确性但牺牲了解决方案的覆盖率。尽管这种权衡在数学和编码任务中已被观察到，但它不一定适用于代理任务，其中多轮工具使用和交互可能需要在后训练过程中新获得的能力。我们惊人的发现是，经过预训练的LLMs，配备轻量推理工具，可以作为有能力的代理。尽管准确性（pass@1）远低于后训练模型，但在给定足够的测试时间预算的情况下，它们在解决方案覆盖率（pass@K）上往往超过其后训练的对应模型。我们进一步分析了其背后的机制，表明后训练将任务推向两个极端：总是解决或从不解决，从而提高了采样效率和一致性，但以牺牲解决方案覆盖率为代价。为了衡量这一成本，我们提出了精细化税（Sharpening Tax），这是一种诊断指标，用于量化后训练后测试时间可扩展性的损失。在来自四个家族的14对基础/后训练模型和三个代理基准（共42个案例）中，这种税在大多数设置中普遍存在，可以通过少量的回合估算，并与其他指标有良好的相关性。最后，我们提出了后验调温组采样（PTGS），这是一种简单的即插即用贝叶斯采样器，根据每个提示的估计难度调整采样温度。在两个代理环境中的RL训练中，PTGS的税负低于固定温度基线，在重复采样下解决了更多任务，同时也提高了单次准确性。
cs.AI / 111 / 2610.01513
Decision Titan: Test-Time Training for Long-Term Memory in Offline Reinforcement Learning
决策泰坦：离线强化学习中的测试时训练以实现长期记忆
Abstract
Long-term dependencies remain a major challenge for sequential decision-making in the field of AI: RNNs suffer from vanishing gradients and the limited expressivity of vector-based hidden states, whilst Transformer-based models are limited by the quadratic scaling of attention. Recent work has proposed tackling this problem with the Test-Time Training (TTT) framework, which stores episodic memories in the parameters of a neural network through gradient descent at both train and test-time. This approach has seen success in the domain of Natural Language Processing, however, to the best of our knowledge it has not yet been applied to the domain of Reinforcement Learning (RL), nor has there been a study analysing how this memory practically functions. In this paper, we study the potential of the TTT framework for offline RL by augmenting a Decision Transformer with TTT layers, dubbed the Decision Titan. We analyse performance and properties of the model in the X-Maze environment, an extension of T-Maze designed to test sequential memory, and investigate how the memory mechanism learns by visualising gate values over time. Our key findings are that Decision Titan can learn long-term dependencies with ranges 20x longer than the context window, generalises to lengths 1.7x the training data, but crucially temporal generalisation depends on the time embeddings used, and the ability to learn long-term dependencies depends on how the relevant information is encoded.
Chinese Translation
长期依赖性仍然是人工智能领域序列决策的主要挑战：循环神经网络（RNN）面临梯度消失和基于向量的隐藏状态表达能力有限的问题，而基于变换器（Transformer）模型则受到注意力机制的二次扩展限制。近期的研究提出通过测试时训练（Test-Time Training, TTT）框架来解决这一问题，该框架通过在训练和测试时的梯度下降将情节记忆存储在神经网络的参数中。这种方法在自然语言处理领域取得了成功，但据我们所知，它尚未应用于强化学习（Reinforcement Learning, RL）领域，也没有研究分析这种记忆机制的实际功能。在本文中，我们通过增强决策变换器（Decision Transformer）与TTT层，研究TTT框架在离线RL中的潜力，称之为决策泰坦（Decision Titan）。我们在X-Maze环境中分析模型的性能和特性，该环境是T-Maze的扩展，旨在测试序列记忆，并通过可视化门值随时间的变化来研究记忆机制的学习过程。我们的主要发现是，决策泰坦能够学习比上下文窗口长20倍的长期依赖性，能够推广到训练数据的1.7倍长度，但关键是时间嵌入的使用决定了时间推广的能力，而学习长期依赖性的能力则取决于相关信息的编码方式。
cs.AI / 112 / 2610.01531
Towards Reliable Vision-Language Models for Autonomous Driving
迈向可靠的视觉-语言模型用于自动驾驶
Abstract
Vision-Language models (VLMs) are increasingly being explored in autonomous driving for tasks such as scene understanding, driving reasoning, decision-making, and end-to-end driving. As their role becomes more prominent, ensuring their robustness and reliability is increasingly important. In real-world conditions, visual inputs may be degraded by sensor imperfections and environmental conditions, potentially affecting both model predictions and their associated confidence. Such degradation is especially concerning in autonomous driving, where safety-critical decisions require models to make accurate predictions and recognize when their predictions may be unreliable. In this work, we evaluate five VLMs (Qwen3.5-9B, Gemma4-E4B, LLaVA-OneVision-7B, DriveFusion/DriveFusionQA-4B, and NVIDIA Alpamayo-1.5-10B) across four driving-related QA datasets with different visual input settings, including single-frame, multi-view, multi-frame, and monocular inputs. Our results show that the effects of visual corruption vary across models, datasets, and input settings, with changes in accuracy and confidence reliability and also differing across conditions. We then apply Visual Evidence Augmentation ($\mathrm{V}{\scriptstyle \mathrm{EA}}$), a recent inference-time method to examine whether it can improve model reliability under degraded visual conditions. We find that $\mathrm{V}{\scriptstyle \mathrm{EA}}$ improves performance for some models and datasets, although the gains are not consistent across all settings.
Chinese Translation
视觉-语言模型（VLMs）在自动驾驶领域的应用日益受到关注，涉及场景理解、驾驶推理、决策制定和端到端驾驶等任务。随着其作用的日益突出，确保模型的鲁棒性和可靠性变得愈加重要。在现实世界条件下，视觉输入可能受到传感器缺陷和环境条件的影响，从而可能影响模型的预测及其相关的置信度。这种退化在自动驾驶中尤其令人担忧，因为安全关键的决策要求模型做出准确的预测，并能够识别其预测可能不可靠的情况。在本研究中，我们评估了五个视觉-语言模型（Qwen3.5-9B、Gemma4-E4B、LLaVA-OneVision-7B、DriveFusion/DriveFusionQA-4B 和 NVIDIA Alpamayo-1.5-10B），并在四个与驾驶相关的问答数据集上进行测试，这些数据集具有不同的视觉输入设置，包括单帧、多视角、多帧和单目输入。我们的结果表明，视觉干扰对模型、数据集和输入设置的影响各不相同，准确性和置信度的可靠性也因条件而异。随后，我们应用了视觉证据增强（Visual Evidence Augmentation, $ ext{V}{ extstyle ext{EA}}$），这是一种最新的推理时方法，以检验其是否能在退化的视觉条件下提高模型的可靠性。我们发现，$ ext{V}{ extstyle ext{EA}}$在某些模型和数据集上提高了性能，尽管在所有设置中收益并不一致。
cs.AI / 113 / 2610.01533
Neither Black nor White: Balancing Semantic and Collaborative Signals with Graph-Informed Semantic IDs (GrIS)
既非黑即白：通过图信息语义ID (GrIS) 平衡语义和协作信号
Abstract
Existing work on Semantic IDs (SIDs) for generative recommendation treats SID construction as a representation learning problem: encode items into a quantised latent space and read off codes. We argue this view is incidental. SID construction is, at heart, a recursive clustering problem, and once stated this way the natural object to cluster is a graph whose nodes carry semantic content and whose edges carry collaborative signal; SID assignment becomes a hierarchical graph partition. This reframing yields a unified framework, Graph-Informed Semantic IDs (GrIS), that subsumes prior approaches rather than displacing them. RQ-VAE and RQ-KMeans are recovered as the special case where the graph is empty, exposing content-only quantisation as one corner of a larger design space along two so-far-collapsed axes: graph construction and recursive partition algorithm. We explore two contrasting instantiations: RecDMoN, which performs hierarchical assignment via differentiable graph pooling, and RQ-GAE, which extends RQ-VAE with graph-aware item representations and a graph reconstruction objective. On multiple real-world datasets, GrIS consistently improves over CF-aware SOTA, with gains of up to +52\% Hit@10. Because graph construction and partition are explicit, separately configurable components, improvements on either axis can be combined and evaluated systematically.
Chinese Translation
现有关于生成推荐的语义ID (SIDs) 的研究将SID构建视为一个表示学习问题：将项目编码到量化的潜在空间中并读取代码。我们认为这种观点是偶然的。SID构建本质上是一个递归聚类问题，一旦以这种方式表述，自然要聚类的对象是一个图，其节点携带语义内容，边缘携带协作信号；SID分配变成了一个层次图划分。这种重新表述产生了一个统一框架，即图信息语义ID (GrIS)，它包容了先前的方法，而不是取而代之。RQ-VAE和RQ-KMeans被恢复为图为空的特例，揭示了仅内容量化作为一个更大设计空间的一个角落，沿着两个迄今为止被压缩的轴：图构建和递归划分算法。我们探索了两种对比的实例：RecDMoN，通过可微分图池执行层次分配，以及RQ-GAE，扩展了RQ-VAE，具有图感知的项目表示和图重构目标。在多个真实世界的数据集上，GrIS始终优于基于协作过滤的最先进方法，提升幅度高达+52\% Hit@10。由于图构建和划分是明确的、可单独配置的组件，因此在任一轴上的改进可以系统地结合和评估。
cs.AI / 114 / 2610.01539
The AI Assessment Sandbox Configurator: A Framework to Support Technical Assessment in AI Regulatory Sandboxes
人工智能评估沙盒配置器：支持人工智能监管沙盒技术评估的框架
Abstract
The EU's Artificial Intelligence Act requires all Member States to establish AI Regulatory Sandboxes (AIRS) by August 2027: supervised environments bringing together national Competent Authorities, technical experts, and the organisations under assessment. When AIRS engagements include structured technical testing, running such testing at scale demands dedicated infrastructure, yet the tooling ecosystem remains structurally fragmented, with heterogeneous tools producing outputs that are difficult to compare, trace, and reuse. From the procedural conditions of AIRS engagements and the AI Act obligations for high-risk systems, we derive 11 architectural and governance requirements for the infrastructure that operationalises technical testing within an AIRS. In response to these requirements, we introduce the AI Assessment Sandbox Configurator, an open-source framework combining a curated Catalogue of tests and controls accessed through a stable plug-in API, a shared data model that harmonises heterogeneous outputs, role-specific dashboards for multi-disciplinary interpretation, and audience-segmented reporting. We describe the architecture and current release, and report an early-stage pilot that exercised the harmonisation and reporting layers within a live AIRS engagement and contributed to an official Exit Report. We discuss the roadmap, the governance questions raised by the Catalogue's tiered contribution model, and the institutional pathways through which an open-source assessment ecosystem could emerge across Member States.
Chinese Translation
欧盟的人工智能法案要求所有成员国在2027年8月之前建立人工智能监管沙盒（AIRS）：这些是监督环境，将国家主管机构、技术专家和被评估组织聚集在一起。当AIRS参与包括结构化技术测试时，大规模进行此类测试需要专门的基础设施，然而工具生态系统仍然结构性分散，异构工具产生的输出难以比较、追踪和重用。根据AIRS参与的程序条件和针对高风险系统的人工智能法案义务，我们推导出11个架构和治理要求，以支持在AIRS内实施技术测试的基础设施。为响应这些要求，我们提出了人工智能评估沙盒配置器，这是一个开源框架，结合了通过稳定插件API访问的精心策划的测试和控制目录、协调异构输出的共享数据模型、用于多学科解释的角色特定仪表板以及面向不同受众的报告。我们描述了该框架的架构和当前版本，并报告了一项早期阶段的试点，该试点在一个实时的AIRS参与中测试了协调和报告层，并为官方退出报告做出了贡献。我们讨论了路线图、目录分层贡献模型所提出的治理问题，以及在成员国之间可能出现的开源评估生态系统的制度路径。
cs.AI / 115 / 2610.01581
Evaluating Physical Consistency and Plausibility in Generative Scenario Models for Autonomous Driving
评估自动驾驶生成场景模型的物理一致性和合理性
Abstract
Generative AI models are increasingly used for scenario generation in autonomous driving. While they can generate realistic-looking scenarios, they often provide limited transparency into learned representations and consistency with real-world vehicle dynamics. This lack of formal assurance limits their use in safety-critical validation and certification workflows. To address this aspect, we introduce a layered evaluation protocol that complements existing methods by assessing models across five layers. The first four layers inspect internal representations and network layers through kinematic alignment, statistical baseline comparison, latent controllability, and activation analysis. The fifth layer evaluates model outputs against vehicle dynamics constraints such as lateral jerk thresholds. We demonstrate the protocol on a Variational Autoencoder (VAE)-based scenario generator. Although standard output-level metrics and visualizations suggest that the generated scenarios are realistic, our protocol provides deeper insight into the extent to which the model's latent space aligns with kinematic features and whether visually plausible trajectories satisfy vehicle-dynamics constraints. We further apply the protocol to additional generative models, demonstrating its applicability beyond the VAE architecture.
Chinese Translation
生成性人工智能模型在自动驾驶的场景生成中越来越多地被使用。尽管它们能够生成看似真实的场景，但通常对学习到的表征和与现实世界车辆动力学的一致性提供有限的透明度。这种缺乏正式保证的情况限制了它们在安全关键的验证和认证工作流程中的应用。为了解决这一问题，我们引入了一种分层评估协议，该协议通过在五个层面上评估模型来补充现有方法。前四个层面通过运动学对齐、统计基线比较、潜在可控性和激活分析检查内部表征和网络层。第五个层面则将模型输出与车辆动力学约束（如横向加速度阈值）进行评估。我们在基于变分自编码器（Variational Autoencoder, VAE）的场景生成器上展示了该协议。尽管标准的输出级别指标和可视化结果表明生成的场景是现实的，但我们的协议提供了更深入的洞察，揭示模型的潜在空间在多大程度上与运动学特征对齐，以及视觉上合理的轨迹是否满足车辆动力学约束。我们进一步将该协议应用于其他生成模型，展示了其在超越VAE架构的适用性。
cs.AI / 116 / 2610.01618
Agents Are Systems, Not Models: Rethinking Agentic Evaluation
智能体是系统，而非模型：重新思考智能体评估
Abstract
Agent evaluations increasingly go beyond a single success rate, reporting metrics such as cost, consistency, and robustness. Yet they typically treat the agent itself as fixed. In practice, an agent is a configurable system: users decide what to tell it, how long to let it run, and which model to use, and each of these choices can change how well and how consistently it performs. We study these choices on a new benchmark of four scientific tasks, where a coding agent must find and correctly operate a published specialist model. We investigate five parts of the agent's configuration: task information, reasoning, self-verification, time budget, and backbone model. We find substantial run-to-run variability, with approximately 54% of the outcome variance coming from repeating the same configuration rather than changing it. Across configurations, the information provided to the agent has the largest effect, exceeding both time budget and model size, while also reducing cost and improving calibration. Configuration choices also interact: additional time helps only when the agent has sufficient information or a capable enough model to use it. Finally, a trajectory-based taxonomy of agent behavior reveals that prompting an agent to verify its answer has little effect on its verification behavior, whereas providing a dedicated verification tool changes that behavior substantially. These results suggest that agents should be evaluated as configurable systems themselves, and that some desired behaviors are more effectively implemented in the system than requested through prompting. We release the benchmark and more than 18,000 agent trajectories.
Chinese Translation
智能体的评估越来越超越单一的成功率，报告诸如成本、一致性和鲁棒性等指标。然而，它们通常将智能体本身视为固定不变的。在实践中，智能体是一个可配置的系统：用户决定告诉它什么、让它运行多长时间以及使用哪个模型，而这些选择都可能影响其表现的好坏和一致性。我们在一个新的基准测试中研究这些选择，该测试涉及四个科学任务，其中一个编码智能体必须找到并正确操作一个已发布的专业模型。我们调查了智能体配置的五个部分：任务信息、推理、自我验证、时间预算和基础模型。我们发现运行之间存在显著的变异性，约54%的结果方差来自于重复相同的配置而非更改它。在不同配置中，提供给智能体的信息对结果的影响最大，超过了时间预算和模型大小，同时还降低了成本并改善了校准。配置选择之间也存在相互作用：额外的时间仅在智能体拥有足够的信息或足够强大的模型来利用这些信息时才有帮助。最后，基于轨迹的智能体行为分类法揭示，促使智能体验证其答案对其验证行为几乎没有影响，而提供专门的验证工具则显著改变了该行为。这些结果表明，智能体应作为可配置系统本身进行评估，并且某些期望的行为在系统中实现的效果优于通过提示请求实现。我们发布了该基准测试及超过18,000条智能体轨迹。
cs.AI / 117 / 2610.01620
FedLore: Communication and Memory Efficient Federated Learning via Shared Gradient Low-Rank Projection
FedLore：通过共享梯度低秩投影实现通信和内存高效的联邦学习
Abstract
Federated training of foundation models is constrained by client memory and communication costs. LoRA-based methods reduce these costs through low-rank adapters, but their fixed rank budget can limit adaptation. Gradient low-rank optimization offers greater flexibility, yet independently chosen client subspaces create a problem we term \emph{subspace fragmentation}: local projections interact with data heterogeneity to bias aggregated directions, while aggregation can increase update rank and communication cost. Thus, accurate local gradient compression need not preserve global descent. We propose \texttt{FedLore}, which shares a low-rank optimization basis within each round and refreshes it across rounds. The shared basis enables exact aggregation in low-rank coordinates and eliminates the identified projection bias. Subspace refresh allows the accumulated model update to exceed the per-round rank budget. We characterize the aggregation bias and establish an $O(T^{-1/2})$ stationarity bound for the projected-SGD variant under a global-gradient coverage condition and standard smoothness and variance assumptions, with bounded gradient heterogeneity. Experiments on vision and language tasks, including federated pre-training, show that \texttt{FedLore} outperforms the evaluated low-rank adapter baselines and matches or exceeds full-parameter training, while reducing communication and optimizer-state memory.
Chinese Translation
基础模型的联邦训练受到客户端内存和通信成本的限制。基于LoRA的方法通过低秩适配器降低这些成本，但其固定的秩预算可能限制适应性。梯度低秩优化提供了更大的灵活性，但独立选择的客户端子空间会产生我们称之为“子空间碎片化”（subspace fragmentation）的问题：局部投影与数据异质性相互作用，导致聚合方向的偏差，而聚合可能增加更新秩和通信成本。因此，准确的局部梯度压缩不一定能保持全局下降。我们提出了 exttt{FedLore}，在每轮内共享低秩优化基础，并在轮次之间进行刷新。共享基础使得在低秩坐标下能够进行精确聚合，并消除了识别出的投影偏差。子空间刷新允许累积的模型更新超过每轮的秩预算。我们对聚合偏差进行了表征，并在全局梯度覆盖条件及标准平滑性和方差假设下，建立了投影-SGD变体的$O(T^{-1/2})$平稳性界限，且具有有界的梯度异质性。在视觉和语言任务上的实验，包括联邦预训练，表明 exttt{FedLore}的性能优于评估的低秩适配器基线，并与全参数训练相匹配或超越，同时减少了通信和优化器状态内存。
cs.AI / 118 / 2610.01626
Measuring the Stability Assumption Behind Action Chunking
测量动作分块背后的稳定性假设
Abstract
Action chunking improves the performance of policies learned by behavioural cloning, and several mechanisms have been proposed to explain why, including temporal consistency, horizon reduction, representation learning, and reduced error compounding. We instead study what happens to an action error once it enters the system. At each state, we inject a small action error and measure how fast it grows or shrinks under two execution regimes: open-loop, where the rest of the chunk is replayed without replanning, and closed-loop, where the policy replans after the perturbation. The fitted rate labels each state as contracting, expanding, or unresolved. Across twelve manipulation tasks from three benchmark suites, we find that confidently stable states are rare, while error amplification is common among states whose propagation rate can be resolved. We further find that the measured propagation rate depends strongly on the fitting horizon: amplification is typically front-loaded, so short windows can overestimate longer-horizon propagation. Finally, we train predictors on these labels and find that a state's open-loop regime can be recovered from camera frames and proprioception alone, while its closed-loop propagation is only partially recoverable because it also depends on how the policy acts after the perturbation. These results suggest that error-compounding arguments alone do not provide a complete account of action chunking: neither passive open-loop dynamics nor policy replanning consistently contracts an injected error, and replanning rarely turns open-loop amplification into confident contraction. This suggests that closed-loop reactivity should be trained explicitly, using perturbation- and tree-coverage-oriented training to expose policies to deviations they must recover from, rather than expected to emerge reliably from standard imitation learning.
Chinese Translation
动作分块提高了通过行为克隆学习的策略的性能，并提出了多种机制来解释其原因，包括时间一致性、视野缩减、表示学习和减少误差累积。我们则研究一旦动作误差进入系统后会发生什么。在每个状态下，我们注入一个小的动作误差，并测量在两种执行模式下其增长或缩小的速度：开放环路（open-loop），即其余的分块在不重新规划的情况下重放；闭环（closed-loop），即策略在扰动后重新规划。拟合的速率将每个状态标记为收缩、扩张或未解决。在来自三个基准套件的十二个操作任务中，我们发现自信稳定的状态很少，而误差放大在其传播速率可以被解决的状态中很常见。我们进一步发现，测量的传播速率强烈依赖于拟合视野：放大通常是前置的，因此短时间窗口可能会高估较长视野的传播。最后，我们在这些标签上训练预测器，发现一个状态的开放环路模式可以仅通过摄像头帧和本体感知恢复，而其闭环传播仅部分可恢复，因为它还依赖于策略在扰动后如何行动。这些结果表明，仅依靠误差累积的论点无法完整解释动作分块：无论是被动的开放环路动态还是策略重新规划都不能一致地收缩注入的误差，而重新规划很少将开放环路放大转变为自信收缩。这表明闭环反应性应当被明确训练，使用扰动和树覆盖导向的训练来使策略暴露于必须恢复的偏差，而不是期望其从标准模仿学习中可靠地出现。
cs.AI / 119 / 2610.01710
CoEvolve: Construct-to-Edit Visual Grounding with Bidirectional State Refinement
CoEvolve：具有双向状态精炼的构建-编辑视觉定位
Abstract
Visual grounding localizes an object described by language with a bounding box. Most multimodal grounding models compress target identification, spatial reasoning, and boundary estimation into one terminal prediction. Free-form rationales make reasoning linguistically explicit but do not necessarily expose measurable, editable spatial states. Intermediate localization errors are therefore difficult to diagnose and correct, allowing incorrect region choices and imprecise boundaries to persist in the final box. We introduce CoEvolve, a construct-to-edit framework that separates grounding into explicit state construction and state editing. Region-Evolution Reinforcement (RER) organizes grounding analysis into a progressive semantic--spatial trajectory, with each reasoning step committing to an explicit candidate region. Bidirectional Denoising Refiner (BDR) treats the reasoning text as fixed semantic context and refines the trajectory's coordinate fields through bidirectional same-position reconstruction. Geometry- and behavior-level objectives provide target geometry and edit-preference signals for consolidating reliable candidates, preserving accurate inputs, or correcting toward annotations. Evaluations cover natural-image and remote-sensing grounding. With a 9B backbone, CoEvolve rivals models up to 241B parameters in grounding accuracy. Under controlled corruption, a single BDR pass improves mean box overlap by over 27 percentage points, demonstrating strong recovery from substantial localization errors. State-source comparisons further support the complementarity of explicit state construction and source-matched editing. The project is at https://sundongwei.github.io/CoEvolve_Project/.
Chinese Translation
视觉定位通过边界框将语言描述的对象进行定位。大多数多模态定位模型将目标识别、空间推理和边界估计压缩为一个终端预测。自由形式的推理使得推理在语言上变得明确，但不一定暴露可测量的、可编辑的空间状态。因此，中间定位错误难以诊断和纠正，导致不正确的区域选择和不精确的边界在最终框中持续存在。我们提出了CoEvolve，一个构建-编辑框架，将定位分为明确的状态构建和状态编辑。区域演化强化（Region-Evolution Reinforcement, RER）将定位分析组织成一个渐进的语义-空间轨迹，每一步推理都承诺一个明确的候选区域。双向去噪精炼器（Bidirectional Denoising Refiner, BDR）将推理文本视为固定的语义上下文，并通过双向同位置重构来精炼轨迹的坐标场。几何和行为层面的目标为巩固可靠候选提供目标几何和编辑偏好信号，保持准确输入或朝向注释进行纠正。评估涵盖自然图像和遥感定位。在使用9B主干网络的情况下，CoEvolve在定位准确性上与高达241B参数的模型相媲美。在控制腐蚀下，单次BDR处理使得平均框重叠提高了超过27个百分点，展示了从显著定位错误中强有力的恢复。状态源比较进一步支持了明确状态构建与源匹配编辑的互补性。该项目网址为 https://sundongwei.github.io/CoEvolve_Project/.
cs.AI / 120 / 2610.01718
vFedProtoQNAS: Prototype-Guided Personalized Quantum Neural Architecture Search for Virtual Federated Learning
vFedProtoQNAS：基于原型指导的个性化量子神经网络架构搜索用于虚拟联邦学习
Abstract
Quantum federated learning (QFL) has emerged as a promising approach for collaboratively training compact quantum neural networks (QNNs) over distributed private data on resource-constrained devices. However, differences in device capabilities make a single shared QNN architecture unsuitable for all clients. While personalized quantum neural architecture search (QNAS) allows each client to select a device-specific QNN, averaging parameters across structurally different QNN architectures mixes semantically inconsistent circuit operations. To address this, prototype-guided personalized QNAS for virtual FL (vFedProtoQNAS) is proposed, where model parameters are never aggregated across clients and federated collaboration is achieved through class-wise prototype sharing. Each client independently searches and trains a client-specific QNN, computes class-wise local prototypes from latent representations, and refines them using global prototypes from the server as federated semantic anchors. Experiments demonstrate that vFedProtoQNAS improves accuracy by 3.70\% over FedAvg and enhances class-consistent representation alignment.
Chinese Translation
量子联邦学习（QFL）作为一种有前景的方法，旨在在资源受限的设备上对分布式私有数据进行协同训练紧凑的量子神经网络（QNNs）。然而，设备能力的差异使得单一共享的QNN架构不适合所有客户端。尽管个性化量子神经架构搜索（QNAS）允许每个客户端选择特定于设备的QNN，但在结构上不同的QNN架构之间平均参数会混合语义不一致的电路操作。为了解决这一问题，提出了基于原型指导的个性化QNAS用于虚拟联邦学习（vFedProtoQNAS），在该方法中，模型参数不会在客户端之间聚合，而是通过类别级原型共享实现联邦协作。每个客户端独立搜索和训练特定于客户端的QNN，从潜在表示中计算类别级局部原型，并使用来自服务器的全局原型作为联邦语义锚点进行优化。实验表明，vFedProtoQNAS比FedAvg提高了3.70\%的准确率，并增强了类别一致的表示对齐。
cs.AI / 121 / 2610.01763
TopK-Guided: Adaptive, Budget-Aware Activation Sparsity for Efficient LLM Inference
TopK引导：适应性、预算感知的激活稀疏性以提高大型语言模型推理效率
Abstract
Activation sparsity speeds up large language model (LLM) inference by setting unimportant activations to zero so that the corresponding computations can be skipped. Existing training-free methods, however, make different trade-offs: threshold-based methods such as TEAL adapt the sparsity level to each token but do not tightly control the realised sparsity, while TopK-based methods such as WINA enforce a fixed sparsity level but use the same sparsity budget for every token. Both also apply the same budget across transformer blocks, despite large differences in block sensitivity. We introduce TopK-Guided, a training-free method that addresses both limitations by combining bounded token-level sparsity adaptation with sensitivity-aware block-level budget allocation. Across Llama-2 and Llama-3 models, TopK-Guided consistently improves perplexity and downstream accuracy over TEAL and WINA while preserving essentially the same sparsitydependent projection compute as WINA, with the largest gains at high sparsity. Ablations show that both components provide complementary improvements.
Chinese Translation
激活稀疏性通过将不重要的激活值设为零，从而加速大型语言模型（LLM）的推理，使得相应的计算可以被跳过。然而，现有的无训练方法在不同的权衡中存在不足：基于阈值的方法如TEAL将稀疏性水平适应于每个标记，但未能严格控制实现的稀疏性；而基于TopK的方法如WINA则强制执行固定的稀疏性水平，但对每个标记使用相同的稀疏预算。两者在变换器块之间也应用相同的预算，尽管块的敏感性差异很大。我们提出了TopK引导，这是一种无训练的方法，通过结合有界的标记级稀疏适应与敏感性感知的块级预算分配，解决了这两方面的限制。在Llama-2和Llama-3模型中，TopK引导在保持与WINA基本相同的稀疏依赖投影计算的同时，始终改善了困惑度和下游准确性，尤其在高稀疏性下获得了最大的提升。消融实验表明，这两个组件提供了互补的改进。
cs.AI / 122 / 2610.01766
VideoEvolve: Evolving Agent Harnesses for Video Temporal Grounding
VideoEvolve：用于视频时间定位的进化代理工具
Abstract
Video temporal grounding aims to localize events in videos from natural-language queries. For agents built around frozen video-language models, the harness determines how queries guide temporal predictions and how those predictions are refined. Manually refining these harnesses requires diagnosing grounding failures and coordinating changes to both agent workflows and instructions. We introduce VideoEvolve, a framework that automatically evolves agent harnesses for video temporal grounding. VideoEvolve uses a Cloze-Structured Harness Representation that preserves stage interfaces while leaving agent workflows and instructions open to evolution. Branch-Guided Harness Evolution preserves promising code branches for continued refinement, using execution feedback to guide local edits and validation to determine which improvements are carried forward. Experiments demonstrate improved grounding performance across multiple benchmarks. Component analyses identify instruction refinement as a consistent source of gains, while the benefits of evolved code vary across evaluation settings. Together, these results support automated harness evolution as an effective approach to improving video temporal grounding. Code is available at https://github.com/bingjunluo/VideoEvolve .
Chinese Translation
视频时间定位旨在根据自然语言查询在视频中定位事件。对于基于冻结视频-语言模型构建的代理，工具决定了查询如何指导时间预测以及这些预测如何被细化。手动细化这些工具需要诊断定位失败，并协调对代理工作流程和指令的更改。我们提出了VideoEvolve，一个自动进化代理工具以实现视频时间定位的框架。VideoEvolve使用一种填空结构工具表示法（Cloze-Structured Harness Representation），该表示法保留了阶段接口，同时使代理工作流程和指令开放于进化。分支引导工具进化（Branch-Guided Harness Evolution）保留了有前景的代码分支以便于持续细化，利用执行反馈指导局部编辑，并通过验证确定哪些改进被保留。实验表明，在多个基准测试中，定位性能得到了改善。组件分析确定指令细化是一致的收益来源，而进化代码的好处在不同评估设置中有所不同。综合来看，这些结果支持自动化工具进化作为改善视频时间定位的有效方法。代码可在 https://github.com/bingjunluo/VideoEvolve 获取。
cs.AI / 123 / 2610.01780
RealCompanion: Benchmarking Human Understanding from Reasoning over Longitudinal Real-World Conversations
RealCompanion：基于对长期真实对话推理的人类理解基准测试
Abstract
A companion that talks with a person for months should come to understand them. It should remember what they said, infer who they are, and know when the past bears on the message in front of it. Testing this requires a real person's record, and such records are private, so benchmarks generate the person and the questions and settle in advance what matters. We release \bench, ten real relationships with an AI companion: 27,218 messages over up to 120 days, released as the conversation and four files derived from it, a profile, a persona, a chat ground truth and a question set, each citing the messages it rests on. Every chat label carries the reasoning trace that produced it, checked stage by stage against the conversation. Three findings follow. First, the past is rarely needed and far away. Pooled measures mislead: a recency window finds the required message for 95.9\% of probes and 2.2\% of those that need memory, and at the natural rate 96\% of the gain from supplying recorded evidence comes from messages that need none. Second, no detector we tried can tell when memory is needed on real messages, authored questions over the same histories leak the cue, and labeling the same messages as memories raises their use by ten to fourteen points. Third, three agent systems reconstruct the persona with the same F1 at a 31-fold difference in cost.
Chinese Translation
一个与人交谈数月的伴侣应该能够理解他们。它应该记住他们所说的话，推断出他们是谁，并知道过去何时与当前信息相关。测试这一点需要真实的个人记录，而这些记录是私密的，因此基准测试生成个人和问题，并预先确定重要内容。我们发布了 ench，包含十个与AI伴侣的真实关系：27,218条消息，跨越最长可达120天，发布为对话及其衍生的四个文件，包括个人档案、角色、聊天真实情况和问题集，每个文件都引用了其基础消息。每个聊天标签都携带生成它的推理轨迹，逐阶段与对话进行核对。以下是三项发现。首先，过去的信息很少被需要且距离较远。汇总测量可能会误导：一个近期窗口发现95.9%的探测所需消息，而需要记忆的仅为2.2%，在自然情况下，96%的记录证据带来的增益来自于不需要记忆的消息。其次，我们尝试的没有任何探测器能够判断真实消息何时需要记忆，基于相同历史的提问泄露了线索，而将相同消息标记为记忆会使其使用率提高10到14个百分点。第三，三个代理系统以相同的F1重建角色，但成本差异达到31倍。
cs.AI / 124 / 2610.01781
Q-Learning for Reachability in MEC-Free MDPs
无模型的Q学习在无最大终端组件（MEC）马尔可夫决策过程（MDP）中的可达性
Abstract
Reinforcement learning (RL) for reachability specifications is fundamental to sequential decision-making. Prior work establishes asymptotic convergence to optimal policies, but only through model-based methods that must explicitly estimate the transition probabilities of the underlying Markov Decision Process (MDP). We present Quasar, the first model-free algorithm with asymptotic guarantees for reachability on the fragment of MDPs free of non-terminal maximal end components (MECs), a building block to which every MDP reduces by the standard MEC quotient. Our algorithm follows the classical Q-learning approach, using temporal-difference updates to converge to an optimal policy without ever learning the transition probabilities. The resulting learner reduces the memory footprint from the O(|S|^2|A|) that model-based methods require to O(|S||A|). On the standardized Quantitative Verification Benchmark Set, our algorithm converges to the optimal policy with orders of magnitude fewer samples than the previous model-based state-of-the-art. Together these results are a concrete step toward the practical deployment of reachability learning and, with it, of specification-guided RL.
Chinese Translation
用于可达性规范的强化学习（RL）是顺序决策的基础。之前的研究建立了对最优策略的渐近收敛，但仅通过必须显式估计基础马尔可夫决策过程（MDP）转移概率的基于模型的方法。我们提出了Quasar，这是第一个在无非终端最大终端组件（MEC）片段的MDP上具有渐近保证的无模型算法，所有MDP都可以通过标准的MEC商简化为此构件。我们的算法遵循经典的Q学习方法，使用时间差分更新在不学习转移概率的情况下收敛到最优策略。最终的学习者将基于模型的方法所需的O(|S|^2|A|)的内存占用减少到O(|S||A|)。在标准化的定量验证基准集上，我们的算法以数量级更少的样本收敛到最优策略，相较于之前的基于模型的最先进方法。这些结果为可达性学习的实际应用以及规范引导的强化学习迈出了具体的一步。
cs.AI / 125 / 2610.01787
Not All Experience Belongs in the Weights: Component Routing for Self-Improving GUI Agents
并非所有经验都应纳入权重：自我改进GUI代理的组件路由
Abstract
Self-improving GUI agents keep the trajectories they produce and return them to the agent, by fine-tuning or by retrieval into the prompt, and studies that compare the two destinations disagree. We attribute this to the unit of experience: a trajectory bundles items with different properties, so a conclusion about the bundle depends on its mix. To address this, (i) we introduce component routing, which splits the experience into locators, procedures, state facts and lessons and sends each component to the context or to the weights, compared on the same items across three backbone families, two environments and three seeds. One pool has two destinations: locators and lessons win in the weights, procedures and state facts in the context. (ii) We fit a rule in two properties measured before any training, recurrence and state-conditionality; it recovers the destination of a held-out backbone family in 24 of 24 cells, two interventions move a component toward the boundary, and routing by the rule beats every whole-trajectory baseline and, by +3.5 points on average, the better single destination of each backbone. (iii) We identify how training and producer-consumer differences change the value of the two destinations: note readout decreases after the same component is written into the weights, most for the items that recur most, context gains increase with the information gap, and weights gains decrease with the policy gap. Code and data will be released.
Chinese Translation
自我改进的GUI代理保留其产生的轨迹并将其返回给代理，通过微调或检索到提示中，比较这两种去向的研究结果并不一致。我们将此归因于经验的单位：轨迹将具有不同属性的项目捆绑在一起，因此关于捆绑的结论取决于其组合。为了解决这个问题，(i) 我们引入了组件路由，将经验分为定位器、程序、状态事实和经验教训，并将每个组件发送到上下文或权重中，在三个主干家族、两个环境和三个种子上对相同项目进行比较。一个池有两个去向：定位器和经验教训在权重中获胜，而程序和状态事实在上下文中获胜。 (ii) 我们在任何训练之前测量的两个属性中拟合了一个规则，即重复性和状态条件性；该规则在24个单元中恢复了一个保留的主干家族的去向，两个干预将一个组件向边界移动，基于该规则的路由超越了每个完整轨迹基线，并且平均比每个主干的更好单一去向高出3.5分。 (iii) 我们识别了训练和生产者-消费者差异如何改变两个去向的价值：注意力读取在同一组件写入权重后下降，尤其是对于最常重复的项目，上下文增益随着信息差距的增加而增加，而权重增益随着政策差距的增加而减少。代码和数据将被发布。
cs.AI / 126 / 2610.01800
LineupRL: Verifiable Reinforcement Learning for Time Series Captioning via Caption-to-Series Identification
LineupRL：通过标题到时间序列识别的可验证强化学习用于时间序列描述
Abstract
Time series captioning is a fundamental step in time series understanding and can also serve as the bridge between signal and natural language. Supervised fine-tuning (SFT) relies on a larger model's captions and cannot exceed their quality. Reinforcement learning (RL) can, but its rewards were designed for other modalities and other tasks, and they transfer poorly to open-ended generation in the time series domain. We address this by proposing LineupRL, a reinforcement learning with verifiable rewards (RLVR) pipeline whose reward is caption-to-series identification. The reward model is a frozen large language model (LLM) verifier that reads the generated caption and the candidate time series as raw values, never the chart, and must pick the described time series from multiple distractors. Matching is a far lighter demand on the verifier than writing questions or judging a caption, so an off-the-shelf LLM can supply the reward. Across two captioning benchmarks, and on forecasting and reconstruction where the predictor sees only the caption, LineupRL outperforms SFT and RL baselines on every metric. The 3B vision language model (VLM) trained by LineupRL also outperforms, at 1/24 of the parameters, the 72B VLM whose captions the SFT baseline is distilled from. Our case study shows that LineupRL resists reward hacking, and that the captioner it trains both traces the trend and names the values at key points.
Chinese Translation
时间序列描述是时间序列理解的基础步骤，也可以作为信号与自然语言之间的桥梁。监督微调（SFT）依赖于更大模型的描述，无法超越其质量。强化学习（RL）可以做到这一点，但其奖励是为其他模态和任务设计的，且在时间序列领域的开放式生成中转移效果较差。我们通过提出LineupRL来解决这一问题，该方法是一种具有可验证奖励的强化学习（RLVR）管道，其奖励是标题到系列的识别。奖励模型是一个冻结的大型语言模型（LLM）验证器，它将生成的标题和候选时间序列视为原始值，而不是图表，并且必须从多个干扰项中选择所描述的时间序列。匹配对验证器的要求远低于撰写问题或判断标题，因此现成的LLM可以提供奖励。在两个描述基准测试中，以及在预测和重建任务中，预测器仅看到标题，LineupRL在每个指标上都超越了SFT和RL基线。通过LineupRL训练的3B视觉语言模型（VLM）在参数数量为72B的SFT基线提炼的模型的1/24时也表现更佳。我们的案例研究表明，LineupRL抵御了奖励操控，并且它训练的描述器能够追踪趋势并在关键点命名数值。
cs.AI / 127 / 2610.01813
AI-assisted mitotic counting improves reproducibility and efficiency across multiple tumour types
人工智能辅助的有丝分裂计数提高了多种肿瘤类型的可重复性和效率
Abstract
Mitotic counting is an important component of tumour grading, diagnosis and prognostic assessment across several tumour types, but manual assessment is time-consuming and subject to inter-pathologist variability. To help address these challenges, we developed MitPro, an AI tool designed to improve consistency and efficiency by directing pathologists towards regions with the highest predicted mitotic activity and highlighting mitotic figures for review, while retaining pathologist control over region selection and the final count. We evaluated its effect on the reproducibility and efficiency of mitotic counting in a retrospective, non-interventional, paired reader study comprising 385 whole-slide images from 3 centres in 3 countries and 7 tumour types using 3 different scanners. 13 pathologists participated, with each slide assessed independently by 3 pathologists without AI assistance and again with AI assistance after a minimum 2 week washout period. Across all slides, AI-assisted counting increased the intraclass correlation coefficient from 0.589 to 0.949. Mean pathologist-level median assessment time decreased from 286.4 to 127.8 seconds, corresponding to an average saving of 151.8 seconds per assessment. Improvements in agreement and efficiency were also observed in supporting analyses using HALO AP and Sectra image management systems and in 2 additional tumour types outside the main study population. AI-assisted assessment was associated with a subtle shift towards higher mitotic counts and scores, consistent with identification of more active mitotic hotspots and fewer missed mitotic figures. The frequency of score change between unassisted and AI-assisted assessment was comparable with inter-pathologist variation during routine counting. These findings support the use of MitPro as an assistive tool for more consistent and efficient mitotic assessment in routine practice.
Chinese Translation
有丝分裂计数是多种肿瘤类型中肿瘤分级、诊断和预后评估的重要组成部分，但手动评估耗时且容易受到病理学家间变异的影响。为了解决这些挑战，我们开发了MitPro，这是一种人工智能工具，旨在通过引导病理学家关注预测有丝分裂活性最高的区域并突出显示有丝分裂图像以供审查，从而提高一致性和效率，同时保留病理学家对区域选择和最终计数的控制。我们在一项回顾性、非干预性、配对读者研究中评估了其对有丝分裂计数可重复性和效率的影响，该研究包含来自3个国家和7种肿瘤类型的385张全切片图像，使用3种不同的扫描仪。13名病理学家参与了研究，每张切片由3名病理学家独立评估，未使用人工智能辅助，且在至少2周的洗脱期后再次使用人工智能辅助进行评估。在所有切片中，人工智能辅助计数将组内相关系数从0.589提高至0.949。病理学家级别的平均中位评估时间从286.4秒减少至127.8秒，平均节省了151.8秒的评估时间。在使用HALO AP和Sectra图像管理系统的支持分析以及在主研究人群以外的2种额外肿瘤类型中，也观察到了协议和效率的改善。人工智能辅助评估与更高的有丝分裂计数和评分之间存在微妙的转变，这与识别更多活跃的有丝分裂热点和更少遗漏的有丝分裂图像一致。未辅助与人工智能辅助评估之间的评分变化频率与常规计数中的病理学家间变异相当。这些发现支持将MitPro作为一种辅助工具，用于在常规实践中实现更一致和高效的有丝分裂评估。
cs.AI / 128 / 2610.01833
Continuous Process-Level Evaluation for Evolving Enterprise AI Agent Skills
不断演变的企业人工智能代理技能的过程级评估
Abstract
Enterprise AI agent skills evolve as tool APIs, models, and specifications change, yet final-output evaluation can miss process-level behavioral drift. We present a continuous evaluation framework combining outcome-level and process-level checks, applied to Revenue and Productivity variants of a Business Value Determination skill in an enterprise Value Aware Resiliency system. The framework independently computes per-run ground truth, materializes reusable template tests, and evaluates tool selection, arguments, execution order, and database integrity through programmatic checks and a narrowly scoped LLM judge. We evaluate 240 trials across two skills, two specification variants, two agent harnesses, and three models. Of 175 trials passing all applicable final numerical checks, 162 (92.6 percent; Wilson 95 percent CI: 87.7-95.6 percent) contained another evaluator-detected deviation. Under a broader seven-check final-state definition, 151 of 164 passing runs (92.1 percent; 95 percent CI: 86.9-95.3 percent) still violated a trajectory check. Dependency attribution reduced a mean of 6.34 failed checks per run to 2.65 roots. Specification sensitivity varied by model and harness, with exploratory bootstrap interaction intervals excluding zero for all three Revenue comparisons and one of three Productivity comparisons. Runtime-resolved templates provided reusable regression coverage across the evaluated configurations; longitudinal validation under actual API evolution remains future work.
Chinese Translation
随着工具API、模型和规范的变化，企业人工智能代理技能不断演变，但最终输出评估可能会忽视过程级行为漂移。我们提出了一种连续评估框架，结合了结果级和过程级检查，应用于企业价值感知弹性系统中的商业价值确定技能的收入和生产力变体。该框架独立计算每次运行的真实值，生成可重用的模板测试，并通过程序化检查和狭义范围的LLM评估者评估工具选择、参数、执行顺序和数据库完整性。我们在两个技能、两个规范变体、两个代理框架和三个模型上评估了240次试验。在175次通过所有适用最终数值检查的试验中，有162次（92.6%；Wilson 95% CI: 87.7-95.6%）检测到其他评估者发现的偏差。在更广泛的七项检查最终状态定义下，164次通过的运行中有151次（92.1%；95% CI: 86.9-95.3%）仍违反了轨迹检查。依赖归因将每次运行的平均失败检查从6.34减少到2.65个根。规范敏感性因模型和框架而异，探索性自助交互区间在所有三项收入比较和三项生产力比较中的一项中均不包括零。运行时解决的模板在评估的配置中提供了可重用的回归覆盖；在实际API演变下的纵向验证仍然是未来的工作。
cs.AI / 129 / 2610.01834
Code Owns the Simulation, Jev Owns the Evaluation
代码掌控模拟，Jev掌控评估
Abstract
Judgment models such as \jev{} return, in a single call and without reasoning text, a probability for each described option. This makes them attractive as an agent's action-selection layer, but it is unclear which decisions they can be trusted with. We test \jev{} on reflection tests, one-shot matrix games, the text game ALFWorld and robot control, and find a sharp boundary. \jev{} succeeds when the right option can be judged from what the input describes, which we call \emph{evaluation}. Specifically, it solves 99\% of the counterintuitive Cognitive Reflection Test questions. However, it fails when the right option depends on \emph{simulation} (i.e., predicting something not in the input), such as the opponent's action or the subgoal that must come first. In games, \jev{} plays suboptimally as if its rational opponent acted at random, because the opponent's action is not given. In ALFWorld, \jev{} favors commands that mention an object or place named in the task description. For example, given the task ``put a clean knife in the drawer'', \jev{} carries an unwashed knife straight to the drawer instead of first washing it at the sink. Surprisingly, many of these failures are not due to a lack of knowledge. Asked separately what the opponent will do, \jev{} usually answers correctly, and it responds well given the opponent's action. It fails when one call must both perform the simulation and evaluate based on it. This suggests letting code make the prediction or simulation. When code supplies it, such as a lookahead in ALFWorld and physics simulation in robot control, \jev{} becomes an expert controller through its general evaluation ability.
Chinese Translation
判断模型如 extit{jev}{}在一次调用中返回每个描述选项的概率，而无需推理文本。这使得它们作为代理的动作选择层变得具有吸引力，但尚不清楚它们可以被信任做出哪些决策。我们在反思测试、一轮矩阵游戏、文本游戏ALFWorld和机器人控制中测试了 extit{jev}{}，发现了一个明显的边界。当正确选项可以从输入描述中判断出来时， extit{jev}{}表现良好，我们称之为 extit{评估}。具体而言，它解决了99 ext{%}的反直觉认知反思测试问题。然而，当正确选项依赖于 extit{模拟}（即预测输入中未包含的内容），例如对手的行动或必须首先完成的子目标时，它则失败。在游戏中， extit{jev}{}表现得次优，仿佛其理性的对手随机行动，因为对手的行动并未给出。在ALFWorld中， extit{jev}{}偏向于提及任务描述中提到的对象或地点的命令。例如，给定任务“把干净的刀放进抽屉”， extit{jev}{}直接将未洗的刀拿到抽屉，而不是先在水槽洗干净。令人惊讶的是，这些失败中的许多并不是由于知识的缺乏。单独询问对手将会做什么时， extit{jev}{}通常能正确回答，并且在给定对手的行动时反应良好。当一次调用必须同时执行模拟并基于此进行评估时，它则失败。这表明让代码进行预测或模拟是可行的。当代码提供这些信息时，例如在ALFWorld中的前瞻性和机器人控制中的物理模拟， extit{jev}{}通过其一般评估能力成为专家控制器。
cs.AI / 130 / 2610.01842
On the Divergence of Accuracy and Mechanism Consistency in Time Series World Models
时间序列世界模型中准确性与机制一致性的分歧
Abstract
A time series world model (TSWM) predicts a controlled system's state from its observed history and planned actions and exogenous inputs. Current approaches build forecasters with actions as covariates, trained and evaluated on prediction error under the executed plan. Yet world models compare unexecuted plans, but their responses to changed plans remain untested. We ask which design choices matter and whether accurate forecasters respond to changed plans as real systems do. We address both with a formalization and benchmark. The formalization separates state, actions and exogenous inputs, distinguishes continuous, mode and event actions, and introduces mechanism consistency, a metric built on declared action-state relations with known directions, such as a vasopressor raising blood pressure: it checks whether shifting an action moves the forecast in the declared direction. The benchmark consolidates eight public datasets with real actions from engineered infrastructure and clinical care, varying prediction space, plan fusion and plan encoding across seven backbones and five seeds. First, a frozen latent prediction space lowers MAE by 9.9% over observation space and gated output fusion lowers it by 12.7% over input concatenation on average, with both improving all eight datasets; temporal plan encoding changes average MAE by at most 2.2%. Second, prediction error and mechanism consistency diverge: the lowest-error configuration is at or below chance in consistency on four of five datasets with declared mechanisms, and no design choice avoids this. Finally, directional supervision, a loss penalizing the wrong-signed part of the response to a shifted action, significantly raises consistency on penalized mechanisms with no change in MAE. Together they give TSWMs a recipe: a frozen latent space and output-side fusion for accuracy, and a training objective for mechanism consistency.
Chinese Translation
时间序列世界模型（TSWM）根据观察历史、计划行动和外部输入预测受控系统的状态。目前的方法将行动作为协变量构建预测模型，并在执行计划下进行训练和评估预测误差。然而，世界模型比较未执行的计划，但它们对变化计划的响应尚未经过测试。我们探讨哪些设计选择是重要的，以及准确的预测模型是否像真实系统那样对变化的计划做出反应。我们通过形式化和基准测试来解决这两个问题。形式化将状态、行动和外部输入分开，区分连续、模式和事件行动，并引入机制一致性，这是一个基于已声明的行动-状态关系及其已知方向（例如，升压药提高血压）的度量：它检查改变行动是否会使预测朝着声明的方向移动。基准测试整合了来自工程基础设施和临床护理的八个公共数据集，涵盖真实行动，变化预测空间、计划融合和计划编码，使用七个骨干网络和五个种子。首先，冻结的潜在预测空间使平均绝对误差（MAE）比观察空间降低了9.9%，而门控输出融合使其比输入连接降低了平均12.7%，这两者均改善了所有八个数据集；时间计划编码对平均MAE的变化最大为2.2%。其次，预测误差和机制一致性出现分歧：在五个具有声明机制的数据集中，最低误差配置在一致性上处于或低于随机水平，并且没有设计选择能够避免这种情况。最后，方向性监督，即对响应中错误符号部分施加惩罚的损失，在惩罚机制上显著提高了一致性，而MAE没有变化。综合来看，它们为TSWMs提供了一种方案：冻结的潜在空间和输出侧融合以提高准确性，以及针对机制一致性的训练目标。
cs.AI / 131 / 2610.01845
Temporal-Difference Learning for Dragonchess
龙棋的时序差分学习
Abstract
Our research investigates how two adaptive AI methods, evolutionary transfer learning and TD(lambda), perform in the three-dimensional chess environment Dragonchess. The game challenges players with its unique board structure and computational load, making it an ideal setting to study how adaptive methods can update evaluation heuristics in novel environments. In this work we re-implement the Dragonchess engine, changing it from a PyGame engine to C++. This enables faster gameplay, allowing us to run 10,000 games with confidence intervals and significance tests, rather than a single small tournament. Both adaptive methods outperform all other agents in the round-robin tournament. Our results showed that there is no significant difference in the performance between the evolved and learned evaluations. This research establishes the efficacy of adaptive methods in structurally complex, novel game domains.
Chinese Translation
我们的研究探讨了两种自适应人工智能方法——进化迁移学习（evolutionary transfer learning）和时序差分学习（TD(lambda)）在三维棋类环境龙棋（Dragonchess）中的表现。该游戏以其独特的棋盘结构和计算负载对玩家提出挑战，使其成为研究自适应方法如何在新环境中更新评估启发式的理想场所。在本研究中，我们重新实现了龙棋引擎，将其从PyGame引擎转换为C++。这使得游戏速度更快，使我们能够自信地进行10,000局游戏的置信区间和显著性测试，而不是仅仅进行一次小型锦标赛。两种自适应方法在循环赛中均优于所有其他代理。我们的结果显示，进化评估和学习评估之间的性能没有显著差异。本研究确立了自适应方法在结构复杂的新游戏领域中的有效性。
cs.AI / 132 / 2610.01861
AVSD-Scenes: A Dataset for Audio-Visual Description of Urban Scenes
AVSD-Scenes：用于城市场景音视频描述的数据集
Abstract
Natural language descriptions can provide rich semantic representations of audio-visual urban scenes, yet datasets that jointly describe both auditory and visual information remain limited. In this paper, we introduce AVSD-Scenes, a paired audio-visual scene description dataset for urban environments. The dataset contains 12,291 audio-visual scene descriptions generated from the TAU Urban Audio-Visual Scenes dataset. To construct the dataset, we first generate audio- and visual-based descriptions using Qwen2-Audio-7B and Qwen2.5-VL-7B, respectively. These modality-specific descriptions are then combined using large language models, namely Qwen3-14B, Mistral-Small-3.2-24B-Instruct-2506, and Gemma-3-27B-it, to produce multimodal descriptions that capture complementary information from both modalities. We benchmark AVSD-Scenes using semantic alignment, cross-modal retrieval, scene classification, LLM-as-a-judge evaluation, and human subjective assessment. Results show that multimodal descriptions improve semantic alignment and cross-modal retrieval performance compared with modality-specific descriptions while preserving strong scene-discriminative information. The generated descriptions achieve up to 94.5% accuracy in urban scene classification, while combining audio, visual, and description embeddings further improves accuracy to 95.4%. Furthermore, the descriptions remain highly scene-discriminative even when scene labels are removed from the prompting instructions, indicating that they capture semantic information derived from the audio-visual content rather than merely reflecting label information.
Chinese Translation
自然语言描述能够提供音视频城市场景的丰富语义表示，但同时描述听觉和视觉信息的数据集仍然有限。本文介绍了AVSD-Scenes，这是一个针对城市环境的配对音视频场景描述数据集。该数据集包含来自TAU Urban Audio-Visual Scenes数据集生成的12,291个音视频场景描述。为了构建该数据集，我们首先分别使用Qwen2-Audio-7B和Qwen2.5-VL-7B生成基于音频和视觉的描述。这些特定模态的描述随后通过大型语言模型（如Qwen3-14B、Mistral-Small-3.2-24B-Instruct-2506和Gemma-3-27B-it）结合，以生成捕捉两种模态互补信息的多模态描述。我们通过语义对齐、跨模态检索、场景分类、LLM作为评判标准的评估以及人类主观评估对AVSD-Scenes进行基准测试。结果表明，与特定模态描述相比，多模态描述在语义对齐和跨模态检索性能上有所提升，同时保持了强大的场景区分信息。生成的描述在城市场景分类中达到了94.5%的准确率，而结合音频、视觉和描述嵌入进一步将准确率提高至95.4%。此外，即使在提示指令中去除场景标签时，描述仍然保持高度的场景区分性，表明它们捕捉了源自音视频内容的语义信息，而不仅仅是反映标签信息。
cs.AI / 133 / 2610.01936
Mapping the RAG Landscape: A Four Axis Taxonomy of Efficiency, Defense, Interactivity, and Reasoning
RAG 领域的映射：效率、防御、交互性和推理的四轴分类法
Abstract
Large Language Models (LLMs) have demonstrated remarkable fluency across many tasks but remain limited by their static, parameter bound knowledge and their susceptibility to hallucinating information. Retrieval Augmented Generation (RAG) addresses these issues by incorporating external retrieval into the generation process, grounding model outputs in verifiable and up to date sources. While prior surveys primarily focus on core RAG architectures and standard pipelines, recent research explores broader challenges and capabilities that extend beyond these foundational designs. This survey provides a consolidated and structured examination of contemporary RAG developments, organizing the field into a four axis taxonomy: improving retrieval efficiency, strengthening robustness and security, supporting user driven and interactive workflows, and enabling multi step or complex reasoning. We formalize key components of the RAG framework and review methods spanning dense and sparse retrieval, fusion strategies, embedding optimizations, and reinforcement learning based retrieval policies, highlighting how these advances influence practical deployment and system design. We also synthesize evaluation practices, domain specific applications, and architectural variants such as Naive, Advanced, and Modular RAG. Finally, we outline persistent challenges related to retrieval quality, reliability, domain adaptation, scalability, and explainability, and identify opportunities for building RAG systems that are more reliable, adaptable, and transparent.
Chinese Translation
大型语言模型（LLMs）在许多任务中展现出了显著的流畅性，但仍然受到其静态的、参数限制的知识和易于产生虚假信息的局限。检索增强生成（RAG）通过将外部检索纳入生成过程，解决了这些问题，使模型输出基于可验证和最新的来源。尽管之前的调查主要集中在核心 RAG 架构和标准流程上，但最近的研究探讨了超越这些基础设计的更广泛挑战和能力。本调查提供了对当代 RAG 发展的综合和结构化的审视，将该领域组织为四轴分类法：提高检索效率、增强鲁棒性和安全性、支持用户驱动的交互式工作流程，以及实现多步骤或复杂推理。我们形式化了 RAG 框架的关键组成部分，并回顾了涵盖密集和稀疏检索、融合策略、嵌入优化和基于强化学习的检索策略的方法，强调这些进展如何影响实际部署和系统设计。我们还综合了评估实践、特定领域应用以及如朴素、先进和模块化 RAG 等架构变体。最后，我们概述了与检索质量、可靠性、领域适应性、可扩展性和可解释性相关的持续挑战，并识别了构建更可靠、适应性强和透明的 RAG 系统的机会。
cs.AI / 134 / 2610.01963
Counterfactual Auditing of Bias in Open-Source Large Language Models for Clinical Triage
开源大型语言模型在临床分诊中的偏见反事实审计
Abstract
Emergency department (ED) triage is a high-stakes prioritization task in which demographic, socioeconomic, and system-context information may improperly influence acuity assignment. Although open-source large language models (LLMs) are increasingly considered for local and privacy-preserving clinical decision support, it remains unclear how counterfactual bias varies across model families, sizes, medical-domain models, and domain-adapted models. We present a comparative counterfactual audit of ten open-source LLMs for pediatric Emergency Severity Index (ESI) prediction. Starting from real and handbook-style clinical vignettes, we construct paired counterfactual variants that change only one injected demographic, socioeconomic, healthcare-access, behavioral, social, or system-context variable while holding the clinical presentation fixed. Models include Qwen2.5-7B, Qwen2.5-14B-Instruct, a QLoRA fine-tuned Qwen2.5-7B, MedGemma variants, MedLLaMA2-7B, GPT-OSS-20B, and GPT-OSS-120B. We measure any counterfactual shift, undertriage, overtriage, shifts greater than one ESI level, mean shift, and mean absolute shift. Counterfactual sensitivity varied substantially and did not consistently decrease with larger model size or medical-domain pretraining. The fine-tuned Qwen2.5-7B showed the lowest overall sensitivity, with a 5.27% any-shift rate and mean absolute shift of 0.0534, versus 16.02% and 0.1706 for the base model. Several larger or medical-domain models showed more significant shifts. Stratified and correlation analyses further revealed clinically important directionality and shared failure patterns hidden by aggregate rates. These findings support counterfactual auditing as a lightweight, clinically interpretable framework for comparing fairness risks in open-source LLMs before clinical deployment.
Chinese Translation
急诊科（ED）分诊是一项高风险的优先级任务，其中人口统计学、社会经济和系统背景信息可能不当影响病情严重程度的分配。尽管开源大型语言模型（LLMs）越来越被考虑用于本地和隐私保护的临床决策支持，但反事实偏见在不同模型家族、模型大小、医学领域模型和领域适应模型之间的变化仍不清楚。我们对十种开源LLM进行了一项比较反事实审计，以预测儿科紧急严重性指数（ESI）。从真实的临床案例和手册风格的临床小插曲出发，我们构建了配对的反事实变体，仅改变一个注入的人口统计学、社会经济、医疗获取、行为、社会或系统背景变量，同时保持临床表现不变。模型包括Qwen2.5-7B、Qwen2.5-14B-Instruct、经过QLoRA微调的Qwen2.5-7B、MedGemma变体、MedLLaMA2-7B、GPT-OSS-20B和GPT-OSS-120B。我们测量了任何反事实变化、低估分诊、过度分诊、超过一个ESI级别的变化、平均变化和平均绝对变化。反事实敏感性差异显著，并未随着模型大小或医学领域预训练的一致增加而减少。微调后的Qwen2.5-7B显示出最低的整体敏感性，任何变化率为5.27%，平均绝对变化为0.0534，而基础模型的变化率为16.02%和0.1706。几个较大或医学领域的模型显示出更显著的变化。分层和相关性分析进一步揭示了临床上重要的方向性和被汇总率掩盖的共同失败模式。这些发现支持反事实审计作为一种轻量级、临床可解释的框架，用于在临床部署前比较开源LLM中的公平性风险。
cs.AI / 135 / 2610.01995
Can AI Oversight Be Zero Knowledge?
人工智能监督能实现零知识吗？
Abstract
AI systems increasingly produce outputs from confidential data, such as a fitness-for-duty assessment from medical records or the predicted properties of a drug candidate from its secret structure. It is important to verify that such outputs are correct without revealing the underlying data. A recent line of work studies verification of AI outputs via interactive proofs and debate for oracle-aided computation, where correctness may depend on an oracle such as human judgment, a physical experiment, or the web. These works focus on verification by a verifier that runs much faster than the computation. However, such efficient verification is impossible for general oracle-aided computation, and these works therefore rely on additional assumptions. We focus instead on privacy: allowing the verifier to run in time polynomial in the computation, we ask whether interactive arguments for oracle-aided computation can be zero knowledge, so that the verifier learns nothing about the confidential data beyond the correctness of the output. We prove that, in general, they cannot. In the random oracle model, there are no zero-knowledge proofs for all oracle-aided computations, even if both the prover and the verifier are allowed to run much longer than the computation itself. The impossibility extends to debate, a canonical model for scalable oversight. On the positive side, we show that if the oracle attaches a cryptographic signature to each of its answers, then every oracle-aided computation can be verified in zero knowledge with an efficient prover and verifier, assuming only collision-resistant hash functions. Beyond privacy, this also gives an alternative approach to scalable oversight that relies neither on an honest opponent, as in debate, nor on the robustness of the computation, as in prior single-prover protocols.
Chinese Translation
人工智能系统越来越多地基于机密数据生成输出，例如基于医疗记录的适岗性评估，或基于药物候选分子秘密结构预测其性质。验证此类输出的正确性而不泄露底层数据至关重要。近期一系列研究通过交互式证明和辩论机制，针对含有oracle辅助计算的AI输出进行验证，其中正确性可能依赖于诸如人类判断、物理实验或网络等oracle。这些工作侧重于由运行速度远快于计算本身的验证者完成验证。然而，对于一般的oracle辅助计算，这种高效验证是不可能实现的，因此这些工作依赖额外假设。我们转而关注隐私问题：允许验证者以多项式时间运行于计算时间内，我们探讨oracle辅助计算的交互式论证是否能实现零知识，即验证者除了输出正确性外不获取任何机密数据。我们证明，通常情况下这是不可能的。在随机oracle模型中，即使证明者和验证者的运行时间远超计算本身，也不存在针对所有oracle辅助计算的零知识证明。这一不可能性结果同样适用于辩论模型——一种典型的可扩展监督模型。积极方面，我们展示了如果oracle对其每个回答附加密码学签名，则在仅假设抗碰撞哈希函数的前提下，每个oracle辅助计算均可通过高效的证明者和验证者实现零知识验证。除隐私保护外，该方法还提供了一种替代的可扩展监督方案，无需依赖辩论中的诚实对手，也不依赖先前单证明者协议中的计算鲁棒性。
cs.AI / 136 / 2610.02001
Mingbird: A Local-First Agent Harness Enabling Small Open Models to Complete Real Tasks
Mingbird：一种本地优先的代理框架，使小型开放模型能够完成实际任务
Abstract
Small open-weight models (2-9B) run on ordinary laptops, but under cloud-scale agent harnesses they rarely complete real tasks: tool prefill overflows the context, self-correction diverges, tool demonstrations loop, and tasks are silently abandoned. We present evidence, from a controlled single-machine comparison and one third-party benchmark, that a substantial share of these failures is attributable to the harness rather than the model. We introduce Mingbird, a local-first agent harness for Windows and Ollama whose ten mechanisms compensate point-by-point for small-model failure forms, three of them representative: a byte-level net-zero prefill budget, a finish gate that re-reads the task before accepting completion, and signature-level loop detection. On LRAB, a controlled comparison holding machine, models, budgets, and scoring fixed (4 harnesses $\times$ 4 open models (2B-35B) $\times$ 18 real tasks, deterministic artifact scoring), Mingbird reaches 0.886 overall against 0.631 (goose), 0.479 (opencode), and 0.405 (agent-mini), with all 288 cells published; on $\tau^2$-bench (278 tasks, three arms, one protocol) it totals 0.856 against 0.791 and 0.737; and a frontier-model probe on the same 18 tasks spans 0.997 to 0.478 across harnesses, with well-formed scaffolds staying within 0.072 of each other. A leave-one-mechanism-out ablation is reported as directional only: same-night replications of the same arm move its mean by up to 0.069, the size of every nominal single-trial delta, and the one batch-matched comparison (full mechanism stack versus text re-read alone) gives the executable completion guards a paired +0.10 across three replications. The evidence carries stated limits: a self-built benchmark, a single machine, and single-trial scoring.
Chinese Translation
小型开放权重模型（2-9B）可以在普通笔记本电脑上运行，但在云规模的代理框架下，它们很少能完成实际任务：工具预填充超出上下文，自我纠正偏离，工具演示循环，任务被默默放弃。我们通过受控的单机比较和一个第三方基准提供证据，表明这些失败中有相当一部分是由于框架而非模型造成的。我们介绍了Mingbird，这是一种针对Windows和Ollama的本地优先代理框架，其十种机制逐点补偿小模型的失败形式，其中三种具有代表性：字节级的净零预填充预算、在接受完成之前重新读取任务的完成门，以及签名级的循环检测。在LRAB上，控制比较保持机器、模型、预算和评分固定（4个框架 × 4个开放模型（2B-35B） × 18个实际任务，确定性工件评分），Mingbird的整体得分达到0.886，而goose为0.631，opencode为0.479，agent-mini为0.405，所有288个单元均已发布；在τ²-bench（278个任务，三个臂，一个协议）上，总得分为0.856，而0.791和0.737；在同18个任务上的前沿模型探测中，跨框架的得分范围为0.997到0.478，良好构建的支架之间保持在0.072以内。一个逐机制剔除的消融实验仅报告为方向性结果：同一臂的同夜重复实验使其均值变化高达0.069，这是每个名义单次试验增量的大小，而唯一的批量匹配比较（完整机制堆栈与仅文本重读）在三次重复中给予可执行完成保护+0.10的配对增益。证据存在已声明的局限性：自建基准、单一机器和单次试验评分。
cs.AI / 137 / 2610.02005
Counting Moves, Weighing Voices: Bayesian Dialectical Argumentation for Calibrated Multi-LLM Councils under Persistent Adversaries
计数动作，权衡声音：针对持续对手的校准多LLM委员会的贝叶斯辩证论证
Abstract
A multi-LLM \emph{council} lets several large language models (LLMs) deliberate on a question and return an answer together with a confidence estimate. As these systems become increasingly used for reasoning, that confidence should represent a calibrated \emph{probability of being correct}, and the decision should remain robust when some agents are persistently unreliable. Existing \emph{council aggregation} methods fail on both fronts: their confidence estimates measure decisiveness rather than correctness, and they cannot identify or discount persistently unreliable agents. We introduce Bayesian Dialectical Argumentation (BDA), which treats the council's \emph{typed} moves---who proposed, challenged, or conceded which answer---as observations of a classical annotator model with \emph{per-agent} reliabilities. This formulation recasts multi-agent deliberation as a reliability estimation problem, using the deliberation trace to infer agent reliability under persistent adversarial behavior. By weighting evidence according to inferred agent reliability, BDA yields calibrated posterior probabilities over candidate answers while allowing persistently unreliable agents to be inverted rather than merely outvoted. Across binary and multi-class benchmarks, BDA achieves the best calibration among zero-cost council aggregation methods, requiring no additional LLM calls, and improves robustness under persistent adversarial coalitions while remaining competitive in clean settings.
Chinese Translation
多LLM extit{委员会}允许多个大型语言模型（LLMs）就一个问题进行讨论，并共同返回一个答案及其置信度估计。随着这些系统在推理中的使用日益增加，这种置信度应代表一个校准的 extit{正确性概率}，并且在某些代理持续不可靠的情况下，决策应保持稳健。现有的 extit{委员会聚合}方法在这两个方面均表现不佳：它们的置信度估计衡量的是果断性而非正确性，并且无法识别或排除持续不可靠的代理。我们引入了贝叶斯辩证论证（BDA），将委员会的 extit{类型化} 动作——谁提出、挑战或让步于哪个答案——视为具有 extit{每个代理} 可靠性的经典注释模型的观察。这种表述将多代理讨论重新构建为一个可靠性估计问题，利用讨论轨迹推断在持续对抗行为下的代理可靠性。通过根据推断的代理可靠性加权证据，BDA 生成候选答案的校准后验概率，同时允许持续不可靠的代理被反转而不仅仅是被投票淘汰。在二元和多类基准测试中，BDA 在零成本委员会聚合方法中实现了最佳校准，且无需额外的 LLM 调用，并在持续对抗联盟下提高了稳健性，同时在干净环境中保持竞争力。
cs.AI / 138 / 2610.02014
Atoms to Processes: The Role of Artificial Intelligence and Machine Learning in Chemical Engineering
从原子到过程：人工智能和机器学习在化学工程中的作用
Abstract
The rapid maturation of artificial intelligence (AI) and machine learning (ML) has catalyzed a profound shift in how chemical engineering problems are formulated, analyzed, and solved. Advances in computing, data availability, and learning algorithms have enabled AI/ML methods to impact applications spanning atomic-scale simulations, materials and catalyst discovery, transport and thermodynamics, separations, process systems engineering, and industrial operations. This article provides a perspective on recent methodological developments and representative applications, emphasizing how AI/ML tools are being integrated with first-principles models to address challenges of predictive accuracy, data scarcity, extrapolation, interpretability, and model lifecycle management. Across domains, a unifying trend is the move away from purely black-box approaches toward hybrid and physics-informed frameworks that explicitly respect conservation laws, thermodynamic consistency, and known structural constraints. These approaches not only improve robustness and reliability, but also enable meaningful human-AI collaboration by providing information at an appropriate level of abstraction for the task and decision context. We conclude that AI and ML are not replacing the core principles of chemical engineering; rather, they are amplifying them. As the field advances toward increasingly autonomous, adaptive, and sustainable systems, the thoughtful integration of AI/ML with first-principles understanding and domain expertise will be essential to realizing their full potential across both research and industrial practice.
Chinese Translation
人工智能（AI）和机器学习（ML）的快速成熟催化了化学工程问题的形成、分析和解决方式的深刻转变。计算能力、数据可用性和学习算法的进步使得AI/ML方法能够影响从原子尺度模拟、材料和催化剂发现、传输与热力学、分离、过程系统工程到工业操作等广泛应用。本文提供了对近期方法论发展的视角以及代表性应用，强调AI/ML工具如何与第一性原理模型相结合，以应对预测准确性、数据稀缺性、外推、可解释性和模型生命周期管理等挑战。在各个领域，一个统一的趋势是从纯粹的黑箱方法转向混合和物理信息驱动的框架，这些框架明确尊重守恒定律、热力学一致性和已知的结构约束。这些方法不仅提高了稳健性和可靠性，还通过在适当的抽象层次上提供信息，促进了有意义的人机协作，以适应任务和决策背景。我们得出结论，AI和ML并没有取代化学工程的核心原则；相反，它们正在放大这些原则。随着该领域朝着越来越自主、适应性强和可持续的系统发展，AI/ML与第一性原理理解和领域专业知识的深思熟虑的整合将对实现其在研究和工业实践中的全部潜力至关重要。
cs.AI / 139 / 2610.02023
SPHERE: Adaptive VR Indoor Scene Generation via LLM-Enhanced Spatial Preference Learning and Human-in-the-Loop RL
SPHERE：通过增强大语言模型的空间偏好学习和人机协作强化学习的自适应虚拟现实室内场景生成
Abstract
While Large Language Models (LLMs) advance 3D indoor scene synthesis, current pipelines fail to retain user-specific preferences across sessions, making immersive authoring a repetitive and physically fatiguing process. We present SPHERE, an adaptive VR generation framework that transforms isolated synthesis into continuous human-AI co-creation. SPHERE extracts persistent spatial preferences from natural multimodal interactions (speech and controller edits). To ensure geometric resilience against spatial distortions, it abstracts these raw edits into hierarchical constraints modeling both local functional and global topological contexts. Furthermore, a human-in-the-loop reinforcement learning mechanism dynamically updates retrieval policies based on the user's final edited scenes. A mixed-design user study ($N=42$) and an offline ablation demonstrate that SPHERE significantly reduces corrective edits and physical demand, preventing bias toward shallow object-level traits to yield geometrically resilient, profile-aligned layouts. Ultimately, SPHERE demonstrates how capturing demonstrated spatial logic enables controlled spatial adaptation, establishing a reliable, governed human-AI collaboration framework for immersive authoring. Project page and source code will be available at: https://github.com/hyeonmin11/SPHERE
Chinese Translation
尽管大语言模型（LLMs）在三维室内场景合成方面取得了进展，但当前的流程未能在不同会话中保留用户特定的偏好，使得沉浸式创作成为一个重复且身体疲惫的过程。我们提出了SPHERE，一个自适应虚拟现实生成框架，将孤立的合成转变为持续的人机协作创作。SPHERE从自然的多模态交互（语音和控制器编辑）中提取持久的空间偏好。为了确保对空间扭曲的几何韧性，它将这些原始编辑抽象为层次约束，建模局部功能和全局拓扑上下文。此外，人机协作强化学习机制根据用户最终编辑的场景动态更新检索策略。一项混合设计的用户研究（$N=42$）和离线消融实验表明，SPHERE显著减少了修正编辑和身体需求，防止了对浅层物体特征的偏见，从而产生几何韧性和轮廓对齐的布局。最终，SPHERE展示了如何捕捉已展示的空间逻辑以实现受控的空间适应，建立了一个可靠的、有序的人机协作框架，以支持沉浸式创作。项目页面和源代码将会在：https://github.com/hyeonmin11/SPHERE
cs.AI / 140 / 2610.02036
Global Coherence: When Every Agent Is Right and the Team Is Still Wrong - A Local-to-Global Semantic Foundation for Multi-Agent Collaboration
全球一致性：当每个代理都是正确的而团队仍然错误时 - 多智能体协作的局部到全球语义基础
Abstract
AI agents can each make locally valid decisions yet jointly produce an invalid result. We call this the global coherence problem: a failure of shared state, not merely of model intelligence. Our Observation-Aliasing Impossibility Theorem gives the exact boundary. A policy can guarantee a valid action exactly when all worlds producing the same observation share an admissible action. If k indistinguishable worlds require pairwise-disjoint actions, the best randomized worst-case success is 1/k; more reasoning, roles, messages, or samples cannot recover the missing distinction. A stronger model can reason better within its context, but it cannot see beyond it. We then give local-to-global runtime semantics X = (H, C, G, F; D): topology H records overlapping scopes; category C governs state-changing actions; groupoid G retains reversible translations; sheaf F tests whether local views glue into one world; and minimal history D keeps only distinctions that alter legal futures. Models propose; the harness owns shared state and governs commit. Nine studies test both the failure and its boundary. On a controlled revision benchmark, the same frontier model scores 40/40 when the deciding event is visible; when it is hidden, tested arms score 12--17/40, consistent with chance (1/3); restoring one authoritative fact returns 40/40. On TeamBench, ordinary teams exceed a shared budget in 5/5 runs, a visible live count leaves 4/5 violations, and commit enforcement leaves 0/5. In tau2-bench Telecom, current-state checks score 0.07 after silent reverts, while the harness scores 1.00. Where a conventional solver already owns the complete relevant state, it ties the harness as predicted. The counterintuitive conclusion is that local intelligence cannot substitute for missing global state.
Chinese Translation
人工智能代理可以各自做出局部有效的决策，但共同产生无效的结果。我们称之为全球一致性问题：共享状态的失败，而不仅仅是模型智能的不足。我们的观察-别名不可能性定理给出了确切的边界。当所有产生相同观察的世界共享一个可接受的行动时，策略才能保证有效的行动。如果k个不可区分的世界需要成对不相交的行动，最佳的随机最坏情况成功率为1/k；更多的推理、角色、消息或样本无法恢复缺失的区分。更强的模型可以在其上下文中进行更好的推理，但无法超越这一点。接着，我们给出了局部到全球的运行时语义X = (H, C, G, F; D)：拓扑H记录重叠的范围；类别C管理状态改变的行动；群oid G保留可逆翻译；层F测试局部视图是否粘合成一个世界；最小历史D仅保留改变合法未来的区分。模型提出；工具拥有共享状态并管理提交。九项研究测试了失败及其边界。在一个受控的修订基准上，当决定事件可见时，相同的边界模型得分为40/40；当事件隐藏时，测试的臂得分为12-17/40，与机会（1/3）一致；恢复一个权威事实返回40/40。在TeamBench上，普通团队在5/5次运行中超出共享预算，明显的实时计数留下4/5次违规，而提交强制执行则留下0/5次。在tau2-bench Telecom中，当前状态检查在静默恢复后得分为0.07，而工具得分为1.00。当一个传统求解器已经拥有完整的相关状态时，它的表现与工具相当，正如预测的那样。反直觉的结论是，局部智能无法替代缺失的全球状态。
cs.AI / 141 / 2610.02038
Mimir: Physics-Grounded LLM Agents for Long-Horizon Irrigation Control
Mimir：基于物理的长时间跨度灌溉控制大型语言模型代理
Abstract
Large language model (LLM) agents increasingly combine reasoning, tool use, and action, but most evidence comes from episodic tasks with relatively immediate feedback and reset failures. Long-running physical control operates in a different regime: actions alter future states, errors compound across decisions, and an agent must improve from experience without being allowed to rewrite the physical rules that make execution safe. We study this regime through irrigation, where daily decisions interact with soil-water dynamics over entire growing seasons. We present Mimir, a physics-grounded LLM agent organized around two repair timescales. At the fast timescale, a structured physical interface and deterministic simulator turn an LLM output into a proposal that we numerically check, revise, and subject to bounded deterministic action selection before execution. At the slow timescale, recurrent failure patterns are consolidated into persistent contextual principles that condition future proposals, while the physical model, evaluator, and execution constraints remain immutable. Under a common retrospective evaluator across multiple sites, crops, and years, Mimir attains the lowest reported aggregate control cost among the evaluated references and uses about 51% less irrigation than the historical schedule replay. The ablation study show higher control cost when forward simulation, verified revision, or persistent context is removed; model-scale and model-family studies show no monotonic gain from increasing LLM size. The resulting lesson show that persistent physical agents can combine semantic reasoning with bounded, evidence-driven self-improvement while reserving physical truth and actuator authority for explicit numerical mechanisms.
Chinese Translation
大型语言模型（LLM）代理越来越多地结合推理、工具使用和行动，但大多数证据来自具有相对即时反馈和重置失败的情境任务。长期物理控制在不同的范畴内运作：行动会改变未来状态，错误在决策中累积，代理必须在不被允许重写确保执行安全的物理规则的情况下，从经验中改进。我们通过灌溉研究这一范畴，其中每日决策与整个生长季节的土壤-水动态相互作用。我们提出了Mimir，一个围绕两个修复时间尺度组织的基于物理的LLM代理。在快速时间尺度上，一个结构化的物理接口和确定性模拟器将LLM输出转化为提案，我们对其进行数值检查、修订，并在执行前进行有界确定性行动选择。在慢速时间尺度上，反复出现的失败模式被整合为持久的上下文原则，这些原则为未来的提案提供条件，同时物理模型、评估器和执行约束保持不变。在多个地点、作物和年份下的共同回顾性评估器中，Mimir在评估的参考中获得了最低的报告综合控制成本，并且使用的灌溉量比历史调度重放少约51%。消融研究表明，当去除前向模拟、经过验证的修订或持久上下文时，控制成本会更高；模型规模和模型家族研究显示，增加LLM规模并未带来单调的收益。最终的教训表明，持久的物理代理可以将语义推理与有界的、基于证据的自我改进相结合，同时将物理真理和执行器权威保留给明确的数值机制。
cs.AI / 142 / 2610.02048
HydroJEV: A one-second, training-free screen for cyber-attack and fault attribution in water distribution networks
HydroJEV：一种无训练的一秒钟网络水分配系统网络攻击与故障归因筛查方法
Abstract
When a SCADA alarm is raised in a water distribution network, operators must decide quickly whether it reflects a cyberattack, a physical fault, a normal transient or a faulty sensor. Supervised classifiers need labelled incidents that utilities rarely have, and frontier large language models (LLMs) take tens of seconds per decision. We tested whether Jev, a training-free model that returns class probabilities in about one second, can serve as the first tier of this triage. On a four-class cause-attribution benchmark built on the C-Town network in EPANET, Jev was compared with a hand-written rule tree, a supervised classifier and seven cloud LLMs on identical evidence in four sealed, pre-registered rounds. With only a label-free prior correction, Jev matched the rule tree (macro-F1 0.62-0.64 against 0.56-0.61 in distribution) and exceeded the supervised classifier by 0.36-0.42 on event subtypes absent from its labels, in all four rounds, and it outperformed the classifier whenever fewer than about four labelled events per class were available. Jev also decided 20-40 times faster than frontier LLMs. Accepting only benign Jev verdicts confirmed by the rule tree spared an LLM reviewer 35-38% of windows on fresh sealed sets without loss of macro-F1. Transferred unchanged to two further networks, this gated cascade stayed within the non-inferiority margin of its reviewer on all four sets. A fast, training-free screen can therefore take over about a third of the review load in SCADA anomaly triage while preserving the accuracy of deliberate review.
Chinese Translation
当水分配网络中出现SCADA警报时，操作员必须迅速判断该警报是否反映了网络攻击、物理故障、正常瞬态或传感器故障。监督分类器需要标记事件，而公用事业公司很少拥有这些标记，前沿的大型语言模型（LLMs）每次决策需要数十秒。我们测试了Jev这一无训练模型，它能在约一秒内返回类别概率，是否可以作为这一分流的第一层。在基于EPANET的C-Town网络构建的四类故障归因基准上，Jev与手工编写的规则树、一个监督分类器和七个云端LLMs在四轮密封的预注册回合中进行了相同证据的比较。仅通过无标签的先验修正，Jev在四轮中与规则树的表现相当（宏F1值为0.62-0.64，而规则树为0.56-0.61），并在缺少标签的事件子类型上超越了监督分类器0.36-0.42。每当每类可用的标记事件少于约四个时，Jev的表现优于分类器。Jev的决策速度也比前沿LLMs快20-40倍。仅接受规则树确认的良性Jev判决，使LLM审查员在新密封集上节省了35-38%的审查窗口，而没有损失宏F1值。该方法在两个进一步的网络上未作更改地转移，且在所有四个数据集上保持在其审查员的非劣性范围内。因此，快速的无训练筛查方法可以在SCADA异常分流中承担约三分之一的审查负担，同时保持审慎审查的准确性。
cs.AI / 143 / 2610.02066
External Observers May See More Clearly: Cross-Model Span-Level Hallucination Detection in Large Language Models via Hidden State Probing
外部观察者可能看得更清楚：通过隐藏状态探测在大型语言模型中进行跨模型跨度级幻觉检测
Abstract
As Large Language Models (LLMs) increasingly serve as foundational reasoning engines, their tendency to hallucinate remains a critical vulnerability. While recent internal state probes offer a promising alternative to slow external retrieval systems, they largely reduce hallucination detection to a token-wise binary classification task, failing to capture the structured, sequential boundaries of semantic drift. Here, we introduce an internal hidden state framework for fine-grained, span-level hallucination detection. By inspecting layer-wise activation patterns, we attempt to detect the exact hallucination onset and continuation tokens in an LLM generation. Our experiments show that this approach successfully isolates hallucination onsets, achieving substantial improvements in Precision-Recall AUC over random baselines despite extreme class imbalance. Ultimately, we propose a novel cross-model detection framework in which one model observes the internal representations elicited by another model's generation. We find that an external observer can match or exceed a generator's self-detection of its own hallucination onsets, including when the observer is the smaller model, suggesting that self-detection is not the ceiling for onset localisation.
Chinese Translation
随着大型语言模型（LLMs）日益成为基础推理引擎，其幻觉倾向仍然是一个关键的脆弱性。尽管最近的内部状态探测提供了一个有前景的替代方案，以取代缓慢的外部检索系统，但它们在很大程度上将幻觉检测简化为一个基于标记的二元分类任务，未能捕捉语义漂移的结构化、顺序边界。在此，我们引入了一种内部隐藏状态框架，用于细粒度的跨度级幻觉检测。通过检查层级激活模式，我们尝试检测LLM生成中的确切幻觉开始和持续标记。我们的实验表明，这种方法成功地隔离了幻觉开始，尽管存在极端的类别不平衡，但在精确率-召回率曲线下面积（Precision-Recall AUC）上相较于随机基线取得了显著改善。最终，我们提出了一种新颖的跨模型检测框架，其中一个模型观察另一个模型生成所引发的内部表征。我们发现，外部观察者能够匹配或超越生成器对自身幻觉开始的自我检测，包括当观察者是较小模型时，这表明自我检测并不是开始定位的上限。
cs.AI / 144 / 2610.02070
Causal Memory Policy: Making Memory Utility Identifiable by Intervening on Retrieval
因果记忆策略：通过干预检索使记忆效用可识别
Abstract
Memory-augmented large language models must decide which memories to retain, and recent systems do so by estimating each memory's effect on task performance. However, these estimates rely entirely on retrieved memories. When a memory is never retrieved, store-level interventions produce identical outcomes, leaving its utility unidentified. This is a retrieval-level positivity violation, invisible to diagnostics that examine only memory operations. We introduce Causal Memory Policy (CMP), a causal framework that restores identification by intervening on retrieval itself, reserving a fixed number of context slots for memories sampled with known propensities. CMP estimates memory utility by self-normalized inverse propensity weighting under a balanced assignment design. We prove the causal factorization of memory utility through retrieval, the unbiasedness and exact variance of the estimator, and the optimal decision rule under irreversible operations. Empirically, identification fails for 54% of required memories on LongMemEval and 67% on LoCoMo, and the failure persists in a deployed memory system. CMP improves discrimination between required and non-required memories from 0.54 to 0.66 AUC. Finally, we show that identified memory utility alone is insufficient for retention decisions: per-query utility reaches 0.78 AUC on the query for which it is estimated, yet no aggregation available to a retention policy predicts a memory's value on unseen queries. Code is available at: https://anonymous.4open.science/r/cmp-release-D0C3/.
Chinese Translation
记忆增强的大型语言模型必须决定保留哪些记忆，最近的系统通过估计每个记忆对任务表现的影响来实现这一点。然而，这些估计完全依赖于被检索的记忆。当一个记忆从未被检索时，存储级别的干预产生相同的结果，使其效用无法识别。这是一种检索级别的积极性违反，对于仅检查记忆操作的诊断来说是不可见的。我们提出了因果记忆策略（Causal Memory Policy, CMP），这是一个因果框架，通过对检索本身进行干预来恢复识别，保留固定数量的上下文槽用于以已知倾向抽样的记忆。CMP通过自归一化逆倾向加权在平衡分配设计下估计记忆效用。我们证明了记忆效用通过检索的因果分解、估计量的无偏性和精确方差，以及在不可逆操作下的最优决策规则。在实证上，LongMemEval上54%的所需记忆和LoCoMo上67%的所需记忆的识别失败，并且这种失败在已部署的记忆系统中持续存在。CMP将所需记忆与非所需记忆之间的区分度从0.54提高到0.66 AUC。最后，我们展示了仅识别的记忆效用不足以做出保留决策：在其被估计的查询上，每个查询的效用达到0.78 AUC，但保留策略无法预测未见查询的记忆价值的任何聚合可用。代码可在以下链接获取：https://anonymous.4open.science/r/cmp-release-D0C3/
cs.AI / 145 / 2610.02072
PyPottery: an AI-powered end-to-end suite for pottery processing and publication
PyPottery：一个基于人工智能的端到端陶艺处理与出版套件
Abstract
The study of ceramic materials constitutes a cornerstone of archaeological research, yet the post-production workflow for pottery documentation remains labor-intensive and creates significant publication bottlenecks. This paper presents PyPottery, an open-source, AI-powered suite designed to semi-automate the complete ceramic documentation pipeline. The suite comprises four integrated modules: PyPotteryScan for automated image extraction and handwriting recognition; PyPotteryInk for automatic inking of pencil drawings; PyPotteryTrace for semantically-aware vectorization; and PyPotteryLayout for automated layout generation. Evaluated on 50 hand-drawn sheets containing 240 pottery drawings from the Terramara di Montale (Italy), the framework achieved substantial time savings confirmed by usability study participants, who reported a median perceived speedup of 40$\times$ over traditional workflows (range: 17.5$\times$--120$\times$). These results highlight the potential of AI-assisted tools in archaeological documentation, while the paper addresses the strategic redistribution of cognitive labor toward augmentation rather than automation.
Chinese Translation
陶瓷材料的研究是考古学研究的基石，但陶器文献的后期制作流程仍然劳动密集，造成了显著的出版瓶颈。本文介绍了PyPottery，一个开源的、基于人工智能的套件，旨在半自动化完整的陶瓷文献记录流程。该套件包括四个集成模块：PyPotteryScan用于自动图像提取和手写识别；PyPotteryInk用于铅笔绘图的自动着墨；PyPotteryTrace用于语义感知的矢量化；以及PyPotteryLayout用于自动布局生成。在对包含240幅陶器绘图的50张手绘图纸进行评估后，该框架实现了显著的时间节省，用户研究参与者报告的中位感知加速为40$ imes$，相比传统工作流程（范围：17.5$ imes$--120$ imes$）。这些结果突显了人工智能辅助工具在考古文献记录中的潜力，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方向。
cs.AI / 146 / 2610.02074
Homomorphic Advantage Operator: Stabilizing Reinforcement Learning Under Fully Homomorphic Encryption Constraints
同态优势算子：在完全同态加密约束下稳定强化学习
Abstract
Privacy-preserving machine learning presents significant deployment challenges on the cloud for intelligent systems with confidential data. Fully Homomorphic Encryption (FHE) offers a compelling solution for secure computation, preserving data confidentiality of cloud computations. However, applying FHE to reinforcement learning (RL) requires replacing non-linear operations with polynomial approximations, which diverge catastrophically due to a unique recursive error phenomenon known as the Bellman drift. This article introduces the Homomorphic Advantage Operator (HAO), a stabilization framework designed to prevent polynomial approximation divergence in FHE-based deep RL. HAO adapts the zero-mean centering projection from advantage-based value estimation directly to temporal-difference (TD) targets. This linear projection annihilates the uniform state-value baseline that drives the Bellman drift, maintaining per-state action rankings while requiring zero additional non-linear multiplicative depth and avoiding expensive ciphertext bootstrapping. The proposed HAO framework was evaluated using a three-tier experimental methodology, including a tabular Markov Decision Process (MDP), an encrypted CartPole environment using real CKKS cryptographic operations, and a 20-node logistics routing benchmark with dense continuous features. The results demonstrate that the proposed HAO strictly bounds network pre-activations within the safe polynomial approximation domain. The proposed HAO RL agents achieved 0% boundary breaches across all random seeds used, whereas regularization alone (L2 weight decay and gradient clipping) breached the bound on 3 of 5 seeds and the unstabilized baseline did so in 83.8% of episodes. Finally, HAO agents improve optimal policy accuracy by 18.0 percentage points in tabular domains and remain stable when DP-SGD-style Gaussian noise is added to the clipped gradients.
Chinese Translation
隐私保护的机器学习在云端智能系统中面临着重大部署挑战，尤其是对于包含机密数据的系统。完全同态加密（Fully Homomorphic Encryption, FHE）为安全计算提供了一个引人注目的解决方案，能够保护云计算的数据机密性。然而，将FHE应用于强化学习（Reinforcement Learning, RL）需要用多项式近似替代非线性操作，这种替代由于一种被称为贝尔曼漂移（Bellman drift）的独特递归误差现象而导致灾难性的发散。本文提出了同态优势算子（Homomorphic Advantage Operator, HAO），这是一个旨在防止基于FHE的深度RL中多项式近似发散的稳定化框架。HAO将基于优势的价值估计中的零均值中心投影直接适配到时间差分（Temporal-Difference, TD）目标。这种线性投影消除了驱动贝尔曼漂移的均匀状态值基线，同时在不增加额外的非线性乘法深度的情况下，避免了昂贵的密文引导。所提出的HAO框架通过三层实验方法进行了评估，包括一个表格化的马尔可夫决策过程（Markov Decision Process, MDP）、一个使用真实CKKS加密操作的加密CartPole环境，以及一个具有密集连续特征的20节点物流路由基准。结果表明，所提出的HAO严格限制了网络预激活在安全多项式近似域内。HAO RL代理在所有使用的随机种子中实现了0%的边界突破，而仅使用正则化（L2权重衰减和梯度裁剪）则在5个种子中的3个种子上突破了边界，未稳定的基线在83.8%的回合中也出现了突破。最后，HAO代理在表格化领域提高了最优策略的准确性18.0个百分点，并且在向裁剪梯度中添加DP-SGD风格的高斯噪声时保持稳定。
cs.AI / 147 / 2610.02116
A Comparative Explainability Framework for DeBERTa-v3 in Zero-Shot Medical Abstract Classification
针对零样本医学摘要分类的 DeBERTa-v3 比较可解释性框架
Abstract
A comparative explainability framework is presented to audit DeBERTa-v3 under zero-shot classification of medical abstracts. The work addresses the disagreement problem in Explainable Artificial Intelligence, where different attribution methods produce divergent explanations for the same input and prediction. A natural language inference engine is implemented over the Medical Abstracts corpus with five enriched hypotheses per diagnostic category and a balanced sample of one thousand texts per class. Five explanation methods are compared: SHAP and LIME as model-agnostic approaches, occlusion and Input x Gradient as deep-learning-specific approaches, and Attention x Gradient as a transformer-specific approach. Explanations are standardized through top-token attribution, and pairwise agreement is quantified using the Jaccard index. High predictive accuracy is achieved across well-defined clinical domains, whereas performance degrades under high semantic ambiguity. Explanatory stability directly mirrors predictive certainty, exhibiting strong convergence in univalent categories and a marked drop under diagnostic uncertainty. Furthermore, qualitative error auditing uncovers three systemic failure mechanisms: lexical hypersensitivity, semantic overlap, and loss of attribution coherence. The results support the combined use of several explanation methods and quantitative agreement metrics when auditing transformer-based models in medical text classification, and suggest prioritizing specific clinical ontologies over broad diagnostic labels.
Chinese Translation
本文提出了一种比较可解释性框架，以审计在零样本医学摘要分类下的 DeBERTa-v3。该研究解决了可解释人工智能中的不一致问题，即不同的归因方法对相同输入和预测产生不同的解释。我们在医学摘要语料库上实现了一个自然语言推理引擎，每个诊断类别提供五个丰富的假设，并对每个类别进行了平衡的千篇文本样本。比较了五种解释方法：SHAP 和 LIME 作为模型无关的方法，遮挡法（occlusion）和输入乘梯度（Input x Gradient）作为深度学习特定的方法，以及注意力乘梯度（Attention x Gradient）作为变换器特定的方法。通过顶级标记归因标准化解释，并使用雅卡尔指数（Jaccard index）量化成对一致性。在定义明确的临床领域中实现了高预测准确性，而在高语义模糊性下性能下降。解释的稳定性直接反映了预测的确定性，在单值类别中表现出强收敛性，而在诊断不确定性下则显著下降。此外，定性错误审计揭示了三种系统性失败机制：词汇超敏感性、语义重叠和归因一致性丧失。结果支持在审计基于变换器的医学文本分类模型时结合使用多种解释方法和定量一致性指标，并建议优先考虑特定临床本体而非广泛的诊断标签。
cs.AI / 148 / 2610.02200
VISTA: A Visual Harness for Reasoning in an Interactive World
VISTA：用于在互动世界中推理的视觉工具
Abstract
We show that multimodal models possess strong reasoning abilities and that an appropriate harness can unlock their potential to solve tasks across diverse interactive environments. We introduce VISTA, a visual harness that gives a general-purpose multimodal model long-horizon vision. VISTA allows the model to directly perceive the environment through visual observations and maintains a lossless visual memory that preserves past observations in their original form. The model can actively retrieve these observations and reorganize its visual input as it reasons. On ARC-AGI-3, VISTA improves Claude Opus 5.0's Relative Human Action Efficiency score from 40.68 to a perfect 100.00, with the model completing all 25 public games using 57.4% fewer actions than first-time human participants. VISTA's simple design also allows it to extend naturally to diverse visual environments with minimal adaptation. Across three additional benchmarks covering a diverse range of visual games and puzzles, it substantially outperforms baselines using the same underlying model with minimal harnesses. Our results highlight VISTA's potential as a general-purpose visual harness for advancing multimodal agents in complex visual environments.
Chinese Translation
我们展示了多模态模型具有强大的推理能力，并且适当的工具可以释放它们在多样化互动环境中解决任务的潜力。我们介绍了VISTA，这是一种视觉工具，使通用多模态模型具备长远的视觉能力。VISTA允许模型通过视觉观察直接感知环境，并保持无损的视觉记忆，以原始形式保留过去的观察。模型可以主动检索这些观察，并在推理过程中重新组织其视觉输入。在ARC-AGI-3上，VISTA将Claude Opus 5.0的相对人类行动效率得分从40.68提高到完美的100.00，模型在完成所有25个公共游戏时使用的行动比首次参与的人类参与者少57.4%。VISTA的简单设计还使其能够在多样的视觉环境中自然扩展，适应性极小。在涵盖多种视觉游戏和难题的三个额外基准测试中，它显著超越了使用相同基础模型和最小工具的基线。我们的结果突显了VISTA作为推动复杂视觉环境中多模态代理的通用视觉工具的潜力。
cs.AI / 149 / 2610.02202
ScholarCatalyst: A Benchmark for Retrieving Papers That Inspire New Research
ScholarCatalyst：一个用于检索激发新研究论文的基准
Abstract
What makes great scientists great? Even as AI systems start to make progress on open problems, scientists remain far ahead of them at sensing which prior idea, buried in an ever-growing archive of research, a new problem needs. To study this skill, we draw on researchers who know firsthand which earlier work advanced their completed projects, with papers serving as pointers to the ideas within. Using our automated pipeline that makes author annotation scalable, we build ScholarCatalyst by having 184 lead authors of 207 recent computer science papers label which candidates did or could have advanced their project, each with a detailed rationale. We introduce a retrieval task with author-provided judgments: given an initial research question, retrieve these papers from only the literature available when the project began. Agentic search does no better than embedding retrieval (0.42 vs. 0.48 Recall@20) despite calling that same retriever as a tool. Even an agent built on Claude Fable 5.1, which may have seen the completed papers during training, reaches only 0.51 R@20. These results highlight the need for new training recipes that equip models with expert intuition for searching broad corpora. We envision ScholarCatalyst as a step toward scientific agents that can take a half-formed idea and point to the prior research it needs.
Chinese Translation
伟大的科学家之所以伟大，是什么原因？即使人工智能系统在解决开放问题上取得了一定进展，科学家们在感知新问题所需的先前想法方面仍然远远领先于它们，这些想法隐藏在不断增长的研究档案中。为了研究这种能力，我们借助那些亲身经历过哪些早期工作推动了他们完成项目的研究人员，论文作为指向其中思想的指针。通过我们的自动化流程，使作者注释具有可扩展性，我们构建了ScholarCatalyst，邀请207篇近期计算机科学论文的184位主要作者标注哪些候选论文推动了或可能推动他们的项目，并提供详细的理由。我们引入了一项检索任务，基于作者提供的判断：给定一个初始研究问题，从项目开始时可用的文献中检索这些论文。尽管将相同的检索工具称为工具，主动搜索的表现并不比嵌入检索更好（0.42 vs. 0.48 Recall@20）。即使是基于Claude Fable 5.1构建的代理，尽管在训练期间可能见过已完成的论文，其表现也仅达到0.51 R@20。这些结果突显了需要新的训练方法，以使模型具备在广泛语料库中搜索的专家直觉。我们设想ScholarCatalyst作为朝着科学代理迈出的一步，这些代理能够将半成型的想法指向所需的先前研究。
cs.CL / 1 / 2610.00007
On-Device Named-Entity Recognition: A Deployability Study of Accuracy, Cost, Reliability, and Confidence
设备端命名实体识别：准确性、成本、可靠性和信心的可部署性研究
Abstract
Named-entity recognition (NER) is increasingly wanted on-device (no API, low latency, data kept local). The practitioner's question is not the leaderboard but which model is deployable, how to evaluate it without human annotation, and whether its confidence can be trusted. We answer these jointly. We place nine systems across three paradigms and 13 M to 8 B parameters: a classical tagger (spaCy), bidirectional-encoder specialists (GLiNER, 166 to 460 M), and generative LLMs run locally (Qwen3-0.6B/1.7B/4B-Instruct, DeepSeek-R1-1.5B/8B), on three datasets of differing character, and report accuracy plus two axes the literature omits: latency and output validity. Because our corpus (RSS-News) had no gold, we built silver gold from a cross-family LLM judge panel, then measured its fidelity against benchmark gold and a full human re-validation of the corpus (strict F1 0.95, an upper bound since the human gold was silver-seeded); gold provenance flips the paradigm ranking, moving from LLM-authored silver to human gold raises every encoder and lowers every generative model. On accuracy alone a 4 B instruct LLM is competitive (it leads on clean newswire), so the encoder's case is deployability: it matches or slightly trails at one-ninth to one-twenty-fourth the size, at millisecond-to-second latency, with zero malformed output, while the smallest generative models emit up to 27% invalid output on long inputs, a failure fixed by scale, not output budget. We then characterize GLiNER's per-span confidence: it ranks correctness well (AUROC 0.76 to 0.86) but is overconfident (ECE 0.24 to 0.47, halved by temperature scaling); thresholding gives a small honest out-of-sample F1 gain; an all-local small-to-large cascade gives a modest, corpus-dependent gain over cost-matched random routing; and confidence tracks correctness but not novelty. Every number recomputes offline from per-span records.
Chinese Translation
命名实体识别（NER）在设备端的需求日益增加（无需 API、低延迟、数据保持本地）。从业者关注的问题不是排行榜，而是哪个模型可以部署，如何在没有人工标注的情况下进行评估，以及其置信度是否可信。我们对此进行了联合回答。我们在三个范式和13M到8B参数的九个系统中进行比较：一个经典的标注器（spaCy）、双向编码器专家（GLiNER，166到460M）和本地运行的生成式大语言模型（Qwen3-0.6B/1.7B/4B-Instruct，DeepSeek-R1-1.5B/8B），在三个特性不同的数据集上报告准确性以及文献中遗漏的两个维度：延迟和输出有效性。由于我们的语料库（RSS-News）没有黄金标准，我们从跨家族的大语言模型评审小组构建了银标准，然后测量其与基准黄金标准的保真度以及对语料库的全面人工重新验证（严格 F1 0.95，这是一个上限，因为人工黄金是银种子）；黄金来源改变了范式排名，从大语言模型生成的银标准转向人工黄金标准，提高了每个编码器的表现，降低了每个生成模型的表现。仅在准确性方面，一个4B的指令型大语言模型具有竞争力（在干净的新闻稿上领先），因此编码器的优势在于可部署性：它的规模为1/9到1/24，延迟为毫秒到秒，且没有错误输出，而最小的生成模型在长输入上产生高达27%的无效输出，这一失败是通过规模而非输出预算解决的。接着，我们对GLiNER的每个跨度置信度进行了表征：其正确性排名良好（AUROC 0.76到0.86），但过于自信（ECE 0.24到0.47，通过温度缩放减半）；阈值设定带来了小幅诚实的样本外F1增益；一个全本地的小到大的级联在成本匹配的随机路由上带来了适度的、依赖于语料库的增益；而置信度与正确性相关，但与新颖性无关。每个数字均从每个跨度记录中离线重新计算。
cs.CL / 2 / 2610.00045
SCM-based Fairness and Faithful Explainability for Legal Document Classification
基于SCM的法律文件分类中的公平性与可信解释
Abstract
Transformer models such as LegalBERT are increasingly used in legal decision support, raising concerns about both fairness and the transparency of model explanations. These properties are usually evaluated separately, leaving open whether a debiasing intervention that changes fairness also changes how faithfully explanations reflect model reasoning. This study investigates that relationship on the ECtHR alleged-violations corpus from LexGLUE. It compares a LegalBERT baseline with a fairness-regularized variant that penalizes stereotypical warmth and competence representations during fine-tuning. The evaluation covers predictive performance, demographic fairness, and SHAP explanation faithfulness across five random seeds. At the performance-optimal regularization strength, the intervention does not reduce demographic disparity. This null result holds across two fairness definitions and a conventional word-pair control on the gender axis. Classification performance is largely unchanged. However, the intervention consistently degrades explanation sufficiency across all five seeds and three thresholds. A shuffled-pair control reproduces this degradation while leaving performance and fairness unchanged, indicating that the effect arises from contrastive representational regularization rather than specifically from the warmth and competence structure. The results demonstrate a dissociation between fairness and explanation faithfulness: changes in explanation behavior do not necessarily indicate changes in fairness, and fairness must therefore be evaluated directly.
Chinese Translation
随着法律决策支持中越来越多地使用如LegalBERT等Transformer模型，关于模型公平性和解释透明度的担忧也随之增加。这些属性通常是分开评估的，因此尚不清楚改变公平性的去偏见干预是否也会改变解释对模型推理的忠实反映。本研究在LexGLUE的欧洲人权法院（ECtHR）涉嫌违规语料库上探讨了这一关系。它比较了一个LegalBERT基线模型与一个公平性正则化变体，后者在微调过程中对刻板印象中的温暖和能力表现施加惩罚。评估涵盖了预测性能、人口公平性和SHAP解释的忠实性，使用了五个随机种子。在性能最优的正则化强度下，该干预并未减少人口差异。这个无效结果在两种公平性定义和性别轴上的传统词对控制下均成立。分类性能基本保持不变。然而，该干预在所有五个种子和三个阈值下始终降低了解释的充分性。一个随机配对控制重复了这种降级，同时保持性能和公平性不变，这表明该效应源于对比表示正则化，而不是特定于温暖和能力结构。结果表明公平性与解释忠实性之间的解耦：解释行为的变化不一定表明公平性的变化，因此必须直接评估公平性。
cs.CL / 3 / 2610.00054
The First Token Is Not the Verdict: Hidden Costs of Reading LLM Judges Without Generating
首个标记并非裁决：阅读大型语言模型法官的隐性成本
Abstract
Reading an LLM judge's verdict from the logits of its first generated token is cheap, requires no generation, and is exactly what constrained decoding and likelihood-scoring evaluation harnesses produce. We show that this readout distorts position bias in one direction: it overstates it in every condition we test, so figures obtained this way behave as upper bounds. The mechanism is that judges do not always lead with a verdict token, on 12% to 49% of pairs for three Qwen3 judges and under 3% for Llama-3.1-8B and Phi-3.5-mini, and forcing a read on those pairs returns whichever response was shown first rather than a judgment. Pooled over the 924 pairs where a judge did not commit, the forced read flips on 89.7% of them when the responses are swapped, against 47.5% read after generation (paired difference +0.422, 95% CI [+0.365, +0.467]). The distortion is specific to what is measured: it moves position bias by 42 points while moving judge accuracy by under one point in seven of ten conditions, so it misleads whoever audits a judge rather than whoever uses one. A second, smaller failure occurs even when the judge does lead with a verdict token, since it sometimes opens with one letter and reasons its way to the other, on 0 to 5.5% of pairs at a rate uncorrelated with compliance. We recommend reporting the rate at which a judge leads with a verdict token, which costs one forward pass and no labels, alongside any position-bias figure.
Chinese Translation
从大型语言模型（LLM）法官生成的首个标记的对数中读取裁决是廉价的，不需要生成，并且正是受限解码和似然评分评估所产生的结果。我们展示了这种读取方式在一个方向上扭曲了位置偏差：在我们测试的每种条件下，它都夸大了这一偏差，因此通过这种方式获得的数字表现为上限。其机制在于法官并不总是以裁决标记开头，对于三个 Qwen3 法官的 12% 到 49% 的对（pair），而 Llama-3.1-8B 和 Phi-3.5-mini 的比例低于 3%。在这些对中强制读取返回的是首次显示的响应，而不是判断。在 924 对法官未做出承诺的情况下，当响应被交换时，强制读取在 89.7% 的情况下发生翻转，而生成后的读取仅为 47.5%（配对差异 +0.422，95% 置信区间 [+0.365, +0.467]）。这种扭曲是特定于所测量内容的：它使位置偏差移动了 42 个点，而在十种条件中有七种情况下法官准确性变化不到一个点，因此它误导的是审计法官的人，而不是使用法官的人。即使法官确实以裁决标记开头，第二个较小的失败也会发生，因为有时它以一个字母开头并推理到另一个字母，在 0 到 5.5% 的对中发生，且与合规性无相关性。我们建议报告法官以裁决标记开头的比例，这只需一次前向传递且无需标签，同时报告任何位置偏差的数字。
cs.CL / 4 / 2610.00070
Measuring Human-Like Bias in LLMs? A Critique of Human-Derived Bias Constructs in LLM Evaluation
测量大型语言模型中的类人偏见？对大型语言模型评估中人源偏见构念的批判
Abstract
Researchers increasingly use human-derived bias constructs to study Large Language Models (LLMs), including social-cognitive constructs such as implicit bias and stereotype activation, and cognitive biases such as anchoring, framing effects, and confirmation bias. Such approaches offer alternatives to overt bias probes, particularly when direct questioning may obscure bias or when model behaviour appears normatively acceptable. However, adapting human bias constructs to LLMs introduces an inferential gap. Psychological instruments were developed to study human cognition and social behaviour, whereas LLM evaluations rely on probabilities, text completions, rankings, or simulated decisions. This paper critiques human-centered bias evaluation in LLMs. We show how this gap arises from mismatches pertaining to human-derived constructs, human-model differences, and evaluation contexts, which can blur distinct interpretations of model bias. We then introduce a framework providing an analytical lens for relating these elements to warranted interpretations, with attention to target constructs, operationalizations, scope of inference, and limits of human analogy.
Chinese Translation
研究人员越来越多地使用人源偏见构念来研究大型语言模型（LLMs），包括隐性偏见和刻板印象激活等社会认知构念，以及锚定效应、框架效应和确认偏见等认知偏见。这些方法为显性偏见探测提供了替代方案，特别是在直接提问可能掩盖偏见或模型行为看似符合规范时。然而，将人类偏见构念适应于LLMs引入了推理差距。心理学工具是为了研究人类认知和社会行为而开发的，而LLM评估则依赖于概率、文本补全、排名或模拟决策。本文批判了LLMs中的以人为中心的偏见评估。我们展示了这一差距如何源于人源构念的不匹配、人类与模型之间的差异以及评估背景，这些因素可能模糊模型偏见的不同解释。然后，我们引入一个框架，提供了一个分析视角，将这些元素与合理的解释相关联，关注目标构念、操作化、推理范围和人类类比的局限性。
cs.CL / 5 / 2610.00087
Legal text classification in Korean sexual offense cases: from traditional machine learning to large language models with XAI insights
韩国性犯罪案件的法律文本分类：从传统机器学习到大型语言模型及其可解释性分析
Abstract
The advancement of natural language processing (NLP) has expanded AI-based text classification in the legal domain. However, accurately classifying legal documents remains challenging due to the complexity of legal texts and subtle differences between legal categories. This study evaluates legal text classification models ranging from traditional machine learning techniques to large language models (LLMs) using ten categories of Korean sexual offense precedents. The results show that fine-tuning small-scale models such as KLUE-BERT on legal data outperforms general-purpose models such as GPT-3.5 and GPT-4.0, as well as traditional machine learning models. KLUE-BERT achieved the highest accuracy of 99.3%, indicating that domain adaptation and fine-tuning can be more important than model size for legal document classification. We further employ explainable AI (XAI) techniques to analyze model predictions and misclassification cases. XAI analysis identifies linguistic features influencing model decisions and limitations in capturing subtle textual cues. Using KICS data, which closely resembles real-world legal case records, we further evaluate the model's generalization capabilities and find that it struggles to interpret implicit contextual cues. These findings highlight the importance of both performance and interpretability in legal AI and demonstrate how XAI can improve transparency in legal text classification. AI-assisted tools can support legal professionals in tasks including document classification, legal information retrieval, and case assessment.
Chinese Translation
自然语言处理（NLP）的进步扩展了基于人工智能的法律文本分类。然而，由于法律文本的复杂性和法律类别之间的细微差别，准确分类法律文件仍然具有挑战性。本研究评估了从传统机器学习技术到大型语言模型（LLMs）的法律文本分类模型，使用了十个类别的韩国性犯罪判例。结果表明，在法律数据上对小规模模型如KLUE-BERT进行微调的效果优于通用模型如GPT-3.5和GPT-4.0，以及传统机器学习模型。KLUE-BERT达到了99.3%的最高准确率，表明领域适应和微调在法律文件分类中可能比模型规模更为重要。我们进一步采用可解释人工智能（XAI）技术分析模型预测和误分类案例。XAI分析识别出影响模型决策的语言特征以及捕捉细微文本线索的局限性。使用与真实法律案件记录高度相似的KICS数据，我们进一步评估了模型的泛化能力，发现其在解释隐含上下文线索方面存在困难。这些发现突显了法律人工智能中性能与可解释性的重要性，并展示了XAI如何提高法律文本分类的透明度。人工智能辅助工具可以支持法律专业人士在文档分类、法律信息检索和案件评估等任务中。
cs.CL / 6 / 2610.00092
BudgetSchemaBench: A Budget-Swept Diagnostic for Schema Context in Text-to-SQL
BudgetSchemaBench：一种用于文本到SQL的模式上下文预算诊断
Abstract
Data agents over structured sources must fit database schema into the model's context window. Large catalogs can span many databases and thousands of columns, so cost constraints may require choosing between table coverage and serialization detail well before the context window is full. We introduce BudgetSchemaBench, an execution-grounded diagnostic for this setting. Its construction derives relevance labels mechanically from gold SQL, without human- or LLM-authored ground truth. Using a pooled 80-database catalog, we sweep four schema-context budgets and compare three representations while keeping each retriever's table ranking fixed. A source-namespace check rejects queries that obtain the correct result from the wrong database. The evaluation covers three conditions: end-to-end retrieval; frozen-gold, in which the required tables are guaranteed; and a probe that removes those tables. For the primary solver with raw serialization, raising the budget from 2.5% to 50% of the catalog improves execution accuracy on 1,279 held-out questions by 18 percentage points under lexical retrieval but only 3 under dense retrieval; the dense retriever already finds most required tables at the smallest budget. When the required tables are removed, 94.6% of correct predictions name one of them exactly, consistent with reconstruction of absent schema from parametric knowledge. For the two main solvers in the frozen-gold condition, the three representations differ by at most 2 percentage points, and the widest paired 95% confidence interval bounds the difference within +/-4 points. We observe the same qualitative patterns with one reasoning model from a different family. When retrieval is coverage-limited, execution accuracy is more sensitive to the schema budget than to the tested serializations. The diagnostic and the code used to construct and evaluate it are publicly available.
Chinese Translation
在结构化数据源上，数据代理必须将数据库模式适配到模型的上下文窗口中。大型目录可能跨越多个数据库和数千列，因此成本限制可能要求在上下文窗口填满之前，在表覆盖和序列化细节之间进行选择。我们提出了BudgetSchemaBench，这是针对这种情况的基于执行的诊断工具。其构建通过从黄金SQL中机械地提取相关标签，而不依赖于人工或大型语言模型（LLM）生成的真实数据。使用一个包含80个数据库的汇总目录，我们对四种模式上下文预算进行了测试，并在保持每个检索器的表排名不变的情况下比较了三种表示方式。源命名空间检查会拒绝从错误数据库中获得正确结果的查询。评估涵盖了三种条件：端到端检索；冻结黄金（frozen-gold），其中所需的表是有保障的；以及去除这些表的探测。对于使用原始序列化的主要求解器，将预算从目录的2.5%提高到50%使得在1,279个保留问题上的执行准确率在词汇检索下提高了18个百分点，而在密集检索下仅提高了3个百分点；密集检索器在最小预算下已经找到了大多数所需表。当去除所需表时，94.6%的正确预测恰好命名其中一个，这与从参数知识重建缺失模式的一致性相符。在冻结黄金条件下的两个主要求解器中，三种表示方式的差异最多为2个百分点，最宽的配对95%置信区间将差异限制在+/-4个百分点以内。我们观察到来自不同模型家族的一个推理模型也表现出相同的定性模式。当检索受到覆盖限制时，执行准确率对模式预算的敏感性高于对测试的序列化方式的敏感性。该诊断工具及其构建和评估所用的代码均已公开。
cs.CL / 7 / 2610.00096
FACET at WMT 2026 Automated Translation Quality Evaluation Task
FACET在WMT 2026自动翻译质量评估任务中的应用
Abstract
Different error types in machine translation require different evidence. Whether meaning is preserved can be judged only against the source, while whether the target is well-formed, or whether it names one entity consistently, can be judged from the target alone. We present FACET, our reference-free submission to the WMT26 Automated Translation Quality Evaluation Task, which decomposes evaluation into Fluency, Accuracy, and Consistency passes and gives each pass only the context its error type requires. A single fixed model is prompted three times, and the merged error spans yield the three task outputs, error spans, quality scores, and error-free labels, with no trained components. We also submit FACET-C, which omits the Consistency pass. Without gold labels, we characterize the predictions of FACET. Its system rankings place post-edited human translation first, and the Consistency pass changes about a tenth of segment scores while leaving the ranking nearly unchanged.
Chinese Translation
机器翻译中的不同错误类型需要不同的证据。意义是否得到保留只能通过源语言进行判断，而目标语言是否结构良好，或者是否一致地命名一个实体，则可以仅通过目标语言进行判断。我们提出了FACET，这是我们对WMT26自动翻译质量评估任务的无参考提交，它将评估分解为流畅性（Fluency）、准确性（Accuracy）和一致性（Consistency）三个环节，并为每个环节提供其错误类型所需的上下文。一个固定模型被提示三次，合并的错误范围生成三个任务输出：错误范围、质量分数和无错误标签，没有经过训练的组件。我们还提交了FACET-C，省略了一致性环节。在没有金标准标签的情况下，我们对FACET的预测进行了特征描述。其系统排名将后编辑的人类翻译排在首位，而一致性环节则改变了大约十分之一的段落分数，同时几乎不改变排名。
cs.CL / 8 / 2610.00164
Metric-Construction Coupling Inflates Measured Synthetic Dialect Recovery
度量-构造耦合膨胀了测量的合成方言恢复
Abstract
Korean dialect corpora are available but not redistributable: weights may be released, while reproducing training and evaluation from the underlying data cannot be. We ask how much of that supervision synthetic data recovers, and whether that recovery can be measured independently of the synthesis pipeline. We contribute KoDialectBench, 1,000 items across five regions on three axes, released as identifier hashes and scoring code so users reconstruct the items from their own licensed copy. Recovery is strongly axis-dependent: our best synthetic arm reaches 91.2% of the real-data gain on region identification but 63.7% on comprehension. On generation the answer depends on the metric: the deployed marker lexicon reports 92.3% on dialectness and 119.3% on region match, the latter exceeding the real-data reference, whereas reference-based generation reaches 72.4%. We find the marker metrics' scoring inventory is entirely contained in the inventory our transformation rules can emit. We test the effect of construction access directly with an exact-form construction-disjoint arm that withholds 20% of marker types from the rules. At exactly matched training size (8,600 examples) it reduces dialectness recovery from 91.8% to 8.1% and region-match recovery from 101.9% to 25.6% on the held-out marker inventory, while the three pipeline-independent measurements do not fall at all. A complementary evaluator sweep defines metric-construction coverage (MCC) and finds measured dialectness recovery increasing monotonically as overlap rises from MCC=0 to MCC=1. Shared construction and evaluation inventories can therefore substantially inflate estimates of synthetic-data recovery.
Chinese Translation
韩语方言语料库虽然可用但不可再分发：权重可以发布，但无法从基础数据中重现训练和评估。我们探讨合成数据恢复了多少监督信息，以及这种恢复是否可以独立于合成管道进行测量。我们贡献了KoDialectBench，涵盖五个区域、三个维度的1,000个项目，以标识符哈希和评分代码的形式发布，以便用户从其自己的许可副本中重建这些项目。恢复强烈依赖于维度：我们最佳的合成模型在区域识别上达到了真实数据增益的91.2%，而在理解上仅为63.7%。在生成方面，答案取决于度量：部署的标记词汇在方言性上报告92.3%，在区域匹配上报告119.3%，后者超过了真实数据参考，而基于参考的生成仅达到72.4%。我们发现标记度量的评分清单完全包含在我们的转换规则可以产生的清单中。我们通过一个精确形式的构造-不相交模型直接测试构造访问的影响，该模型从规则中保留了20%的标记类型。在完全匹配的训练规模（8,600个示例）下，它将方言性恢复从91.8%降低到8.1%，将区域匹配恢复从101.9%降低到25.6%，而三个与管道无关的测量则完全没有下降。一个补充的评估器扫描定义了度量-构造覆盖（MCC），并发现测量的方言性恢复随着MCC从0增加到1而单调增加。因此，共享的构造和评估清单可以显著膨胀合成数据恢复的估计值。
cs.CL / 9 / 2610.00202
When a Data Artifact Isn't a Shortcut: Causal Auditing of Synthetic RLVR Corpora
数据伪影并非捷径：合成RLVR语料的因果审计
Abstract
Several recent pipelines build RLVR training data by masking a span of real corpus text and asking a language model to invent plausible wrong answers around it. The correct option is therefore genuine human prose; every distractor is synthetic. Correctness and provenance become entangled, and a policy could in principle learn the second instead of the first. We audit that possibility in GooseReason-0.7M. First we ask whether the asymmetry is visible at all: a classifier reading only five surface statistics (never the meaning) reaches AUROC 0.562 over 315,499 options, barely above chance. The aggregate hides something, though. Code sits at 0.416, below chance, and manual inspection explains why: code distractors turn out to be single-operator mutations of the gold answer rather than freely written alternatives, so the two classes are nearly identical by construction. Detecting a signal is not the same as showing a model uses it, so we then run an intervention. We build a paraphrase-matched control corpus, hold training-set size identical across arms, and train two policies under one fixed budget. The exploitation gap does not favour the unmodified-data arm: 0.021 against 0.027 for the control. Under our budget, in other words, a detectable artifact went unexploited. We think that dissociation, along with the domain-specific construction finding, is worth knowing for anyone curating corpora of this kind, and we release the audit as a mostly CPU-only protocol.
Chinese Translation
最近的几个管道通过遮蔽真实语料文本的一段并要求语言模型围绕其生成合理的错误答案来构建RLVR训练数据。因此，正确选项是真正的人类文本，而每个干扰项都是合成的。正确性和来源变得交织在一起，理论上，一个策略可能学习到的是第二个而不是第一个。我们在GooseReason-0.7M中审计了这种可能性。首先，我们询问这种不对称性是否可见：一个仅读取五个表面统计（从不考虑意义）的分类器在315,499个选项中达到了AUROC 0.562，勉强高于随机猜测。然而，聚合数据隐藏了一些信息。代码的AUROC为0.416，低于随机猜测，手动检查解释了原因：代码干扰项实际上是黄金答案的单操作变异，而不是自由书写的替代品，因此这两个类别在构造上几乎是相同的。检测到信号并不等同于表明模型使用了它，因此我们进行了干预。我们构建了一个与释义匹配的对照语料，保持训练集大小在各组之间相同，并在一个固定预算下训练了两个策略。利用差距并不偏向未修改数据的组：对照组为0.027，而未修改数据组为0.021。换句话说，在我们的预算下，一个可检测的伪影未被利用。我们认为这种分离以及领域特定构造的发现，对于任何策划此类语料的人来说都是值得了解的，我们将审计结果作为一种主要依赖CPU的协议发布。
cs.CL / 10 / 2610.00223
A Holistic Assessment of the Carbon Footprint of Noor, a Very Large Arabic Language Model
对诺尔（Noor）这一超大规模阿拉伯语言模型碳足迹的全面评估
Abstract
As ever larger language models grow more ubiquitous, it is crucial to consider their environmental impact. Characterised by extreme size and resource use, recent generations of models have been criticised for their voracious appetite for compute, and thus significant carbon footprint. Although reporting of carbon impact has grown more common in machine learning papers, this reporting is usually limited to compute resources used strictly for training. In this work, we propose a holistic assessment of the footprint of an extreme-scale language model, Noor. Noor is an ongoing project aiming to develop the largest multi-task Arabic language models -- with up to 13B parameters -- leveraging zero-shot generalisation to enable a wide range of downstream tasks via natural language instructions. We assess the total carbon bill of the entire project: starting with data collection and storage costs, including research and development budgets, pretraining costs, future serving estimates, and other exogenous costs necessary for this international cooperation. Notably, we find that inference costs and exogenous factors can have a significant impact on total budget. Finally, we discuss pathways to reduce the carbon footprint of extreme-scale models.
Chinese Translation
随着越来越大语言模型的普及，考虑它们对环境的影响变得至关重要。近年来的模型因其极大的规模和资源消耗而受到批评，尤其是它们对计算资源的巨大需求，从而导致显著的碳足迹。尽管在机器学习论文中报告碳影响的情况越来越普遍，但这种报告通常仅限于严格用于训练的计算资源。在本研究中，我们提出了对极大规模语言模型诺尔（Noor）碳足迹的全面评估。诺尔是一个正在进行的项目，旨在开发最大的多任务阿拉伯语言模型——参数量高达130亿——利用零样本泛化能力，通过自然语言指令支持广泛的下游任务。我们评估了整个项目的总碳成本：从数据收集和存储成本开始，包括研究和开发预算、预训练成本、未来服务估算以及为这一国际合作所需的其他外部成本。值得注意的是，我们发现推理成本和外部因素对总预算有显著影响。最后，我们讨论了减少极大规模模型碳足迹的途径。
cs.CL / 11 / 2610.00238
CAVE-Mem: Boundary-Aware Experience Validation for Memory Search
CAVE-Mem：边界感知的记忆搜索体验验证
Abstract
Long-term memory agents increasingly rely on it- erative search and reusable experience to answer questions over large personal, factual, or narrative histories. However, current experience-memory systems largely optimize relevance: they re- trieve past search lessons that appear similar to the current state and inject them into the prompt. A relevant experience can still be harmful when the memory substrate, question intent, answer granularity, or evidence boundary changes. We propose CAVE- Mem, a training-free framework that represents experience as a typed intervention operator with applicability, boundary, and utility conditions. CAVE-Mem first obtains a base memory-search answer, then allows an operator to change it only if the oper- ator matches the current substrate, answer contract, evidence boundary, and cross-fitted utility; otherwise the system abstains. Experiments across long-term conversational memory, multi-hop question answering, and long-document narrative reasoning show consistent gains over relevance-only experience reuse.
Chinese Translation
长期记忆代理越来越依赖于迭代搜索和可重用经验，以回答关于大量个人、事实或叙述历史的问题。然而，目前的经验记忆系统主要优化相关性：它们检索与当前状态相似的过去搜索经验，并将其注入到提示中。当记忆基础、问题意图、答案粒度或证据边界发生变化时，相关的经验仍然可能是有害的。我们提出了CAVE-Mem，一个无训练的框架，将经验表示为具有适用性、边界和效用条件的类型干预操作符。CAVE-Mem首先获取一个基础的记忆搜索答案，然后仅在操作符与当前基础、答案契约、证据边界和交叉适配的效用匹配时，允许操作符对其进行更改；否则系统将保持不变。在长期对话记忆、多跳问答和长文档叙述推理的实验中，CAVE-Mem在仅依赖相关性经验重用的情况下显示出一致的提升。
cs.CL / 12 / 2610.00262
Signed Lexical Confidence for Risk-Calibrated Intent Routing
风险校准意图路由的有符号词汇置信度
Abstract
Selective intent routing allows an assistant to act on reliable predictions while deferring uncertain requests. Standard confidence scores primarily reflect the base model's representation, leaving an opportunity to incorporate complementary evidence without changing its decisions. We introduce a signed lexical gate that combines a sentence classifier's logit margin with a sparse lexical model's support for the classifier's predicted intent. By assigning positive evidence to lexical agreement and negative evidence to a lexically favored competing intent, the gate retains more information than either unsigned lexical confidence or a hard agreement rule. An independent binomial calibration stage selects an operating threshold for a specified risk target. Across ten runs on BANKING77, CLINC150, and HWU64, the proposed score reduces area under the risk-coverage curve by 15.8%, 15.1%, and 11.8% relative to a learned semantic-only gate. At a nominal 5% error target, it increases accepted coverage by 1.83 and 5.14 percentage points on BANKING77 and HWU64, while CLINC150 is already near full coverage. At a stricter 2% target, the simultaneous binomial procedure yields a nonempty policy in all 30 dataset-run combinations at the available calibration budgets. Matched controls show that the proposed feature improves average error ranking over the tested unsigned lexical-confidence feature, with dataset-dependent gains over binary agreement. The resulting two-feature gate provides a compact, interpretable confidence enhancement for risk-calibrated intent routing while preserving the base classifier's predictions.
Chinese Translation
选择性意图路由允许助手在可靠预测的基础上采取行动，同时推迟不确定的请求。标准置信度分数主要反映基础模型的表示，留下了在不改变决策的情况下整合补充证据的机会。我们引入了一种有符号词汇门，它将句子分类器的对数边际与稀疏词汇模型对分类器预测意图的支持相结合。通过将正证据分配给词汇一致性，将负证据分配给词汇上更受青睐的竞争意图，该门比无符号词汇置信度或硬一致性规则保留了更多信息。一个独立的二项式校准阶段为指定的风险目标选择操作阈值。在对 BANKING77、CLINC150 和 HWU64 的十次实验中，所提出的分数相对于学习的仅语义门减少了风险覆盖曲线下的面积分别为 15.8%、15.1% 和 11.8%。在名义上的 5% 错误目标下，它在 BANKING77 和 HWU64 上分别提高了 1.83 和 5.14 个百分点的接受覆盖率，而 CLINC150 已经接近完全覆盖。在更严格的 2% 目标下，同时的二项式程序在所有 30 个数据集-运行组合中都产生了非空策略，且在可用的校准预算内。匹配的对照显示，所提出的特征在测试的无符号词汇置信度特征上改善了平均错误排名，并在数据集依赖性上相对于二元一致性获得了收益。最终得到的双特征门为风险校准意图路由提供了紧凑、可解释的置信度增强，同时保留了基础分类器的预测。
cs.CL / 13 / 2610.00296
Certainty Is Not Just Correctness: Rethinking Token-Level Certainty in LLM Reasoning
确定性不仅仅是正确性：重新思考大语言模型推理中的标记级确定性
Abstract
Token-level certainty is widely used as a proxy for correctness in LLM training and inference. However, the performance of certainty-based methods depends both on the information in certainty scores and on how those scores are used. We therefore directly assess certainty's predictive ability through controlled empirical evaluations across models and tasks. We distinguish two prediction targets: identifying questions a model is more likely to answer correctly and distinguishing correct from incorrect responses to the same question. In our experiments, certainty is generally better at identifying questions a model is likely to answer correctly than at distinguishing correct from incorrect responses to the same question. Certainty also varies systematically across token types and positions within words, reflecting local properties of words and text form. Information about question difficulty appears early in generation, while the weaker information about answer correctness is more concentrated near the end. These findings show that the information certainty provides for decisions depends on the prediction target, the model, the certainty metric, and which token positions in the response are included in aggregation. We further demonstrate the practical value of these findings for test-time compute. We allocate the number of responses using certainty early in generation and weight answer votes using certainty near the end of each response. Compared with a fixed-sampling majority-voting baseline, this approach increases overall accuracy from 78.71\% to 79.54\% while reducing generated-token cost by 82.4\%.
Chinese Translation
标记级确定性被广泛用作大语言模型（LLM）训练和推理中的正确性代理。然而，基于确定性的方法的性能既依赖于确定性分数中的信息，也依赖于这些分数的使用方式。因此，我们通过对模型和任务进行控制的实证评估，直接评估确定性的预测能力。我们区分了两个预测目标：识别模型更可能正确回答的问题，以及区分对同一问题的正确与错误回答。在我们的实验中，确定性通常在识别模型可能正确回答的问题方面表现得比区分同一问题的正确与错误回答更好。确定性在不同的标记类型和单词内部的位置上也表现出系统性的变化，反映了单词和文本形式的局部特性。关于问题难度的信息在生成早期就显现出来，而关于答案正确性的较弱信息则更集中在生成的末尾。这些发现表明，确定性为决策提供的信息依赖于预测目标、模型、确定性度量以及在响应中包含的标记位置。我们进一步展示了这些发现对测试时计算的实际价值。我们在生成的早期根据确定性分配响应数量，并在每个响应的末尾使用确定性对答案投票进行加权。与固定采样的多数投票基线相比，这种方法将整体准确率从78.71%提高到79.54%，同时将生成标记的成本降低了82.4%。
cs.CL / 14 / 2610.00316
DuplexSpeechBench-Document Grounding: Benchmarking Document Grounding and Hallucinations in Voice Agents
DuplexSpeechBench-文档基础：语音代理中的文档基础和幻觉基准测试
Abstract
Voice agents enable low-latency, natural interaction, yet their ability to faithfully ground responses in external documents remains underexplored. We introduce DuplexSpeechBench-Document Grounding (DSB-DG), a benchmark for evaluating document grounding in voice agents across five professional domains. DSB-DG targets three failure modes: Context Saturation, which measures grounding under increasing document length; Grounding Decay, which measures retention of document facts across multi-turn dialogue; and Proactive Grounding, which evaluates whether context re-injection mitigates conversational drift. The benchmark contains 1,636 adversarially verified QA pairs from 50 documents covering five professional domains, and supports fully automatic evaluation of grounding accuracy, hallucination, and response latency. Across systems spanning cascaded, proprietary full-duplex and real-time, and open-weight speech2speech architectures, we find substantial differences in effective grounding capacity. While cascaded pipeline (ASR-LLM-TTS) achieves the highest grounding accuracy, Gemini-Live and GPT-Realtime closely trail behind. Open-weight systems exhibit distinct failure modes, most notably an abrupt context-capacity collapse and multi-turn grounding decay. More broadly, grounding fidelity degrades with context and conversational load, and failures frequently manifest as unsupported generations rather than abstention. We show that contextual grounding as a key unresolved challenge for reliable full-duplex voice agents.
Chinese Translation
语音代理实现了低延迟、自然的交互，但它们在外部文档中忠实地基础响应的能力仍然未得到充分探索。我们引入了DuplexSpeechBench-文档基础（DSB-DG），这是一个用于评估语音代理在五个专业领域中的文档基础的基准。DSB-DG针对三种失败模式：上下文饱和度（Context Saturation），用于测量在文档长度增加时的基础情况；基础衰减（Grounding Decay），用于测量在多轮对话中对文档事实的保留；以及主动基础（Proactive Grounding），用于评估上下文重新注入是否能减轻对话漂移。该基准包含来自50个文档的1,636个经过对抗验证的问答对，涵盖五个专业领域，并支持对基础准确性、幻觉和响应延迟的完全自动评估。在涵盖级联、专有全双工和实时以及开放权重的speech2speech架构的系统中，我们发现有效基础能力存在显著差异。尽管级联管道（ASR-LLM-TTS）实现了最高的基础准确性，但Gemini-Live和GPT-Realtime紧随其后。开放权重系统表现出明显的失败模式，最显著的是上下文容量的突然崩溃和多轮基础衰减。更广泛地说，基础的保真度随着上下文和对话负载的增加而降低，失败通常表现为不支持的生成，而不是放弃。我们表明，作为可靠全双工语音代理的一个关键未解决挑战，上下文基础问题亟待解决。
cs.CL / 15 / 2610.00320
Refusal Localizes, the Damage Relocates: Safety Layers Under Few-Sample Fine-Tuning
拒绝本地化，损害重新定位：少样本微调下的安全层
Abstract
Fine-tuning adapts aligned large language models (LLMs) to downstream tasks, but a few dozen harmful examples can remove their refusal of harmful requests. Prior work localizes safety-related behavior to specific layers, directions, and tokens, suggesting targets for protection. We test whether successful localization and recovery support defenses that survive changes in the attack. Across six checkpoints from four model families, harmful and benign prompts remain linearly separable after attack, and patching full clean hidden states into the compromised model restores refusal at a reproducible transition depth. Building on a prior layer-freezing defense, we freeze every layer up to this depth and repeat the attack. At a hundred harmful examples, refusal remains near zero on all six checkpoints, with recovery transitions above the frozen boundary. In a second study, removing the update's top two singular directions restores refusal after short attention-only fine-tunes on four checkpoints. On Llama-3.1-8B, ordinary training changes weaken this repair and an attacker who spreads the update defeats it. A spectral detector calibrated on benign Llama fine-tunes misses most repair failures on that checkpoint. Localized freezing can nevertheless help preserve refusal when a few harmful examples enter training data unintentionally. These results show that an attacker can bypass a region identified by recovery and defeat a repair that works across multiple checkpoints, motivating five checks for defenses against adaptive fine-tuning. Code is available at https://github.com/js-lee-AI/refusal-relocates.
Chinese Translation
微调使对齐的大型语言模型（LLMs）适应下游任务，但几十个有害示例可以消除它们对有害请求的拒绝。先前的研究将与安全相关的行为局限于特定的层、方向和标记，暗示了保护的目标。我们测试成功的本地化和恢复是否支持能够抵御攻击变化的防御。在四个模型系列的六个检查点中，有害和良性提示在攻击后仍然线性可分，并且将完整的干净隐藏状态修补到受损模型中可以在可重复的过渡深度恢复拒绝。在先前的层冻结防御的基础上，我们冻结到该深度的每一层，并重复攻击。在一百个有害示例下，所有六个检查点的拒绝率接近零，恢复过渡在冻结边界之上。在第二项研究中，移除更新的前两个奇异方向在四个检查点上恢复了拒绝，经过短时间的仅注意力微调。在 Llama-3.1-8B 上，普通训练的变化削弱了这种修复，而一个扩散更新的攻击者则击败了它。基于良性 Llama 微调校准的谱检测器在该检查点上错过了大多数修复失败。尽管如此，本地化冻结仍然可以帮助在少量有害示例意外进入训练数据时保持拒绝。这些结果表明，攻击者可以绕过恢复识别的区域，并击败在多个检查点上有效的修复，促使我们对自适应微调的防御进行五项检查。代码可在 https://github.com/js-lee-AI/refusal-relocates 获取。
cs.CL / 16 / 2610.00321
CAST: Cost-Aware Speculative Trees from One-Pass Block Drafters
CAST：基于一次性块草图的成本感知推测树
Abstract
Speculative decoding accelerates large language model inference by drafting future tokens cheaply and verifying them with the target model in parallel. Block drafters score a whole block of future tokens in one forward pass, yet standard decoding verifies only the top-scoring chain and discards the other candidates. Because these candidates are already scored, verifying more of them adds target computation but no extra drafting. We introduce CAST (Cost-Aware Speculative Trees), which packs these candidates into a tree and verifies it in a single target pass, leaving the target model, drafter weights, and decoding rule untouched. To decide how wide the tree should be, CAST adds candidates while the expected gain from the next one outweighs the verification time it adds. The width therefore adapts to each deployment from a latency measurement, without sweeping over widths. We evaluate CAST across five domains on three GPU generations and two model families. At its predicted width, CAST is faster than the standard chain in all eight settings, by up to 43%. We also find that the best width depends strongly on the deployment. Where verification cost jumps at a kernel boundary, a 128-token tree is only 2% faster than the standard chain, whereas the tree at the predicted width is 20% faster. Furthermore, we prove that CAST leaves the target output distribution unchanged under both greedy and sampled decoding. Code is available at https://github.com/js-lee-AI/CAST.
Chinese Translation
推测解码通过廉价草拟未来标记并与目标模型并行验证，从而加速大型语言模型的推理。块草图在一次前向传递中对整个未来标记块进行评分，而标准解码仅验证得分最高的链并丢弃其他候选项。由于这些候选项已经被评分，验证更多的候选项会增加目标计算，但不会增加额外的草图工作。我们提出了CAST（成本感知推测树），它将这些候选项打包成一棵树，并在单次目标传递中进行验证，而不改变目标模型、草图权重和解码规则。为了决定树的宽度，CAST在下一个候选项的预期收益超过其增加的验证时间时添加候选项。因此，宽度根据每次部署的延迟测量进行调整，而不需要遍历所有宽度。我们在三个GPU代和两个模型系列的五个领域中评估了CAST。在其预测的宽度下，CAST在所有八种设置中都比标准链快，速度提高了多达43%。我们还发现最佳宽度与部署密切相关。在内核边界处验证成本激增时，128标记的树仅比标准链快2%，而在预测宽度下的树则快20%。此外，我们证明CAST在贪婪和采样解码下都不会改变目标输出分布。代码可在 https://github.com/js-lee-AI/CAST 获取。
cs.CL / 17 / 2610.00402
Dissonant ballerinas and crafty carrots: a comparative multi-modal analysis of Italian brain rot
不和谐的芭蕾舞者与狡猾的胡萝卜：意大利脑腐烂的比较多模态分析
Abstract
This paper presents a comparative multi-modal analysis of Italian and Romanian brain rot memes, investigating the factors that contribute to its appeal and the linguistic and cultural distinctions between the two versions. To conduct this analysis, we introduce a multi-modal brain rot dataset named CRIB (Collection of Romanian and Italian Brain rot), a manually curated collection of 240 TikTok videos stratified by language (Italian, Romanian) and popularity, on which we examine textual, acoustic, and visual features. Our findings indicate that popularity is not significantly correlated with textual elements like sentiment, absurdity, or rhyme, or acoustic elements such as vocal features or sentiment of the sound. Instead, in Romanian language, video-level dynamics, specifically faster cutting speeds and a more rapid overall pace, are strong predictors of a video's success. The cross-linguistic analysis reveals significant differences. Italian brain rot is textually more negative, exhibits higher perplexity, and uses more rhyme, while its sound is characterized by higher melodic range and loudness. Romanian audio is spectrally brighter with more erratic pitch variations.
Chinese Translation
本文呈现了对意大利和罗马尼亚脑腐烂（brain rot）表情包的比较多模态分析，探讨了影响其吸引力的因素以及两种版本之间的语言和文化差异。为进行此分析，我们引入了一个名为CRIB（Collection of Romanian and Italian Brain rot）的多模态脑腐烂数据集，这是一个手动整理的240个TikTok视频的集合，按语言（意大利语、罗马尼亚语）和受欢迎程度进行分层，我们在此基础上研究文本、声学和视觉特征。我们的研究结果表明，受欢迎程度与文本元素（如情感、荒谬性或押韵）或声学元素（如声音特征或声音情感）之间没有显著相关性。相反，在罗马尼亚语中，视频层面的动态，特别是更快的剪辑速度和整体节奏的加快，是视频成功的重要预测因素。跨语言分析揭示了显著差异。意大利脑腐烂在文本上更为消极，表现出更高的困惑度，并使用更多的押韵，而其声音则具有更高的旋律范围和响度。罗马尼亚音频在频谱上更明亮，具有更不稳定的音高变化。
cs.CL / 18 / 2610.00406
LLM-as-a-Judge for Low-Resource Languages: Adapting Ragas and Comparative Ranking for Romanian
低资源语言的LLM作为评判者：为罗马尼亚语适应Ragas和比较排名
Abstract
Evaluating Retrieval-Augmented Generation (RAG) systems remains a challenge for Low-Resource Languages (LRLs), where standard reference-based metrics fall short. This paper investigates the viability of the "LLM-as-a-Judge" paradigm for Romanian by adapting the Ragas framework using next-generation models (Gemini 2.5 and Gemini 3). We introduce AdminRo-Eval, a curated dataset of Romanian administrative documents annotated by native speakers, to serve as a ground truth for benchmarking automated evaluators. We compare three evaluation methodologies - direct scoring, comparative ranking, and granular decomposition - across metrics for Faithfulness, Answer Relevance, and Context Relevance. Our findings reveal that evaluation strategies must be metric-specific: granular decomposition achieves the highest human alignment for Faithfulness (96% with Gemini 2.5 Pro), while comparative ranking outperforms in Answer Relevance (90%). Furthermore, we demonstrate that while lightweight models struggle with complex reasoning in LRLs, the Gemini 2.5 Pro architecture establishes a robust, transferable baseline for automated Romanian RAG evaluation.
Chinese Translation
评估增强检索生成（RAG）系统在低资源语言（LRLs）中仍然是一个挑战，标准的基于参考的指标无法满足需求。本文通过使用下一代模型（Gemini 2.5和Gemini 3）适应Ragas框架，探讨了“LLM作为评判者”范式在罗马尼亚语中的可行性。我们引入了AdminRo-Eval，一个由母语者标注的罗马尼亚行政文件的精选数据集，以作为自动评估器基准的真实标准。我们比较了三种评估方法——直接评分、比较排名和细粒度分解——在忠实性、答案相关性和上下文相关性指标上的表现。我们的研究结果表明，评估策略必须针对特定指标：细粒度分解在忠实性方面实现了最高的人类一致性（Gemini 2.5 Pro达到96%），而比较排名在答案相关性方面表现更佳（90%）。此外，我们证明了尽管轻量级模型在低资源语言中难以处理复杂推理，但Gemini 2.5 Pro架构为自动化罗马尼亚RAG评估建立了一个稳健且可转移的基线。
cs.CL / 19 / 2610.00408
UniBuc at SemEval-2024 Task 2: Tailored Prompting with Solar for Clinical NLI
UniBuc在SemEval-2024任务2中的表现：针对临床自然语言推理的定制提示与SOLAR
Abstract
This paper describes the approach of the UniBuc team in tackling the SemEval 2024 Task 2: Safe Biomedical Natural Language Inference for Clinical Trials. We used SOLAR Instruct, without any fine-tuning, while focusing on input manipulation and tailored prompting. By customizing prompts for individual CTR sections, in both zero-shot and few-shots settings, we managed to achieve a consistency score of 0.72, ranking 14th in the leaderboard. Our thorough error analysis revealed that our model has a tendency to take shortcuts and rely on simple heuristics, especially when dealing with semantic-preserving changes.
Chinese Translation
本文描述了UniBuc团队在应对SemEval 2024任务2：安全生物医学自然语言推理用于临床试验中的方法。我们使用了SOLAR Instruct，没有进行任何微调，同时专注于输入操作和定制提示。通过为各个CTR部分定制提示，在零样本和少样本设置中，我们成功实现了一致性得分为0.72，在排行榜中排名第14。我们深入的错误分析表明，我们的模型倾向于走捷径并依赖简单的启发式方法，特别是在处理语义保持变化时。
cs.CL / 20 / 2610.00492
EurekaBench: Measuring Agentic Ability to Discover New Scientific Insights
EurekaBench：测量发现新科学见解的能动能力
Abstract
When Isaac Newton discovered the law of gravitation, he did so through an iterative process of analyzing observed data such as planetary patterns, finding the underlying mechanisms by describing patterns in mathematical equations, and refining his theory against the Moon's orbit, revealing the startling insight that the same force governs both falling apples and orbiting planets. Would it be possible for AI agents to make similar discoveries? To measure this ability, we introduce EurekaBench, a cross-domain benchmark that tests AI agents' ability to conduct long-horizon experiments and discover mechanisms that explain observations. We evaluate these mechanisms by the scientific insights that can be derived from them. EurekaBench contains an expert-verified set of 26 long-horizon tasks across neuroscience, computer science, chemistry, astrophysics, geophysics, and plasma physics, with a total of 306 scientific insights that the discovered mechanisms are expected to support. Our evaluation framework tests three axes of scientific discovery: agents' ability to follow known scientific constraints, the predictive accuracy of the discovered mechanisms, and whether these mechanisms yield scientific insights or inform future research. Our results show that current AI agents often overly fixate on predictive accuracy optimization, surpassing human scientists, while falling substantially short in deriving scientific insights.
Chinese Translation
当艾萨克·牛顿发现万有引力定律时，他通过分析观察到的数据（如行星运动模式）进行迭代过程，描述数学方程中的模式，找到潜在机制，并根据月球轨道不断完善他的理论，揭示了同一力量支配着落下的苹果和运行的行星这一惊人见解。那么，人工智能代理是否有可能做出类似的发现呢？为了测量这种能力，我们引入了EurekaBench，这是一个跨领域基准，测试人工智能代理进行长时间实验和发现解释观察结果的机制的能力。我们通过可以从这些机制中推导出的科学见解来评估这些机制。EurekaBench包含一组经过专家验证的26个长时间任务，涵盖神经科学、计算机科学、化学、天体物理学、地球物理学和等离子体物理学，总共306个科学见解，预计这些发现的机制能够支持这些见解。我们的评估框架测试科学发现的三个维度：代理遵循已知科学约束的能力、发现机制的预测准确性，以及这些机制是否产生科学见解或为未来研究提供信息。我们的结果表明，当前的人工智能代理往往过于专注于优化预测准确性，超越了人类科学家，但在推导科学见解方面却大大不足。
cs.CL / 21 / 2610.00526
Rules Amortize, Pairings Don't: Linguistic Structure Determines What Latent Task Representations Can Replace In-Context Learning
规则可以摊销，配对不可以：语言结构决定了潜在任务表示可以替代的上下文学习内容
Abstract
In-context learning (ICL) can be amortized into latent objects (task vectors, function vectors, context vectors) that recover few-shot behavior at zero-shot inference cost, but recent theory shows a static vector acts as a single synthetic demonstration and must fail on high-rank mappings such as word-level bijections. We ask a linguistic version of this question: which linguistic operations can be amortized out of the prompt? We train a 2.6M-parameter network that reads the geometry of a few-shot support set (centroid, principal subspace, spectrum, computed once and cached) and produces an input-conditioned additive update to the query's residual stream at a mid-depth layer of a frozen GPT-2-large/XL. Across eight inflectional directions and one lexical relation, under a canonical split that bars inverted-pair leakage between directions, three regimes emerge. On forward inflection, where 10-shot ICL is strong (0.67-0.89) and extracted task vectors collapse (<=0.06), the transform matches ICL at strictly zero-shot per-query cost. On lemmatization directions, which frozen GPT-2 can execute but 10 demonstrations systematically fail to convey (ICL 0.13-0.48 at 1.5B), the transform is not capped by ICL at all: it reaches 0.78-0.92, up to +72 points over ICL (past to present: 0.85 vs. 0.13). On arbitrary pairings (antonymy) every amortizer plateaus near half of ICL at every scale, capacity, and seed tested. Controls show the support manifold acts as a causally necessary task fingerprint: wrong-task manifolds collapse accuracy to <=0.06, query-only variants cannot disambiguate tasks sharing an input space, and leave-one-task-out transfer is zero. Productive rules amortize into latent task representations, sometimes better than prompting can convey them; memorized pairings do not.
Chinese Translation
上下文学习（ICL）可以摊销为潜在对象（任务向量、函数向量、上下文向量），在零-shot 推理成本下恢复少量示例行为，但最近的理论表明静态向量充当单一的合成示例，并且在高秩映射（如词级双射）上必然失败。我们提出这个问题的语言学版本：哪些语言操作可以从提示中摊销出去？我们训练了一个具有260万参数的网络，该网络读取少量示例支持集的几何特征（质心、主子空间、谱，计算一次并缓存），并在冻结的 GPT-2-large/XL 的中层深度生成一个输入条件的加性更新到查询的残差流。在八个屈折方向和一个词汇关系下，在一个禁止方向间反转配对泄漏的标准分割下，出现了三种模式。在前向屈折中，10-shot ICL 表现强劲（0.67-0.89），而提取的任务向量崩溃（<=0.06），该变换在严格的零-shot 每查询成本下与 ICL 匹配。在词形还原方向上，冻结的 GPT-2 可以执行但 10 个示例系统性地无法传达（ICL 0.13-0.48 在 15 亿参数下），该变换根本不受 ICL 的限制：它达到了 0.78-0.92，比 ICL 高出72分（过去到现在：0.85 对 0.13）。在任意配对（反义词）中，每个摊销器在每个规模、容量和种子测试中都在 ICL 的一半附近达到饱和。控制实验表明支持流形作为因果必要的任务指纹：错误任务流形将准确率压缩到 <=0.06，仅查询变体无法区分共享输入空间的任务，且留一任务转移为零。有效规则摊销为潜在任务表示，有时比提示所能传达的更好；而记忆配对则不然。
cs.CL / 22 / 2610.00540
Assessing the Impact of Language Disparity on Multilingual Linguistic Ability in Large Language Models
评估语言差异对大型语言模型多语言语言能力的影响
Abstract
Claims about the grammatical competence of multilingual language models vary sharply with how competence is measured, yet the interaction between evaluation paradigm, post-training, and language resource availability has not been systematically examined. We evaluate base and post-trained models from six families on MultiBLiMP, a syntactic minimal-pair benchmark covering 101 languages, using four evaluation methods. We report three principal findings. First, post-training degrades grammatical competence, but the magnitude of this effect is reduced unevenly by model scale, while low-resource languages bear the highest cost. Second, post-trained models retain grammatical knowledge they cannot articulate through explicit prompting, yet this is measurable only in high-resource languages, because near-chance baselines in low-resource settings leave little knowledge to hide. Third, native-language prompting recovers otherwise hidden competence on low-resource languages, demonstrating that only high-resource languages can be probed directly from unprompted probabilities. We conclude that multilingual grammatical evaluation must adopt language-informed, multi-paradigm protocols to avoid systematically underestimating low-resource abilities.
Chinese Translation
关于多语言语言模型语法能力的说法因测量方式的不同而存在显著差异，但评估范式、后训练和语言资源可用性之间的相互作用尚未得到系统研究。我们使用四种评估方法，对来自六个模型家族的基础模型和后训练模型在 MultiBLiMP（一个涵盖101种语言的句法最小对基准）上进行了评估。我们报告了三个主要发现。首先，后训练会降低语法能力，但这一影响的幅度因模型规模而异，而低资源语言承受的成本最高。其次，后训练模型保留了它们无法通过明确提示表达的语法知识，但这一点仅在高资源语言中可测量，因为在低资源环境中接近随机的基线几乎没有知识可供隐藏。第三，母语提示能够恢复低资源语言上隐性能力，表明只有高资源语言可以直接从未提示的概率中进行探测。我们得出结论，多语言语法评估必须采用语言知情的多范式协议，以避免系统性低估低资源语言的能力。
cs.CL / 23 / 2610.00562
Can LLMs Reason Over Long Horizons? An Empirical Evaluation of Context Strategies for Longitudinal Clinical Reasoning
大型语言模型能否进行长时间跨度的推理？对纵向临床推理的上下文策略的实证评估
Abstract
Longitudinal clinical reasoning requires large language models (LLMs) to identify and integrate relevant evidence distributed across extended patient histories. Although long-context models can process increasingly large amounts of information, providing more history does not necessarily make relevant evidence more accessible or improve reasoning. We compare five context strategies (Full, Recent, Episodic, Semantic, and Hybrid) on MedLoCoMo across four open-weight LLMs, examining answer correctness, robustness to query-evidence distance, and abstention on questions with unsupported premises. Episodic and Hybrid generally achieve the strongest overall accuracy, while Recent Context degrades most as supporting evidence becomes more distant; Episodic and Hybrid maintain the highest accuracy at long distances. Analysis of adversarial questions further shows that strong performance on answerable questions does not necessarily translate to successful abstention when the available history does not support the requested conclusion. These findings show that reliable longitudinal reasoning depends not only on how much history an LLM can access, but critically on how relevant evidence is selected and presented for reasoning.
Chinese Translation
纵向临床推理要求大型语言模型（LLMs）识别并整合分布在广泛患者历史中的相关证据。尽管长上下文模型能够处理越来越多的信息，但提供更多的历史并不一定使相关证据更易获取或改善推理。我们在MedLoCoMo上比较了五种上下文策略（完全上下文、最近上下文、情节上下文、语义上下文和混合上下文），针对四个开放权重的LLMs，考察了答案的正确性、对查询-证据距离的鲁棒性以及在缺乏支持前提的问题上的弃权情况。情节上下文和混合上下文通常实现了最强的整体准确性，而最近上下文在支持证据变得更远时下降最为明显；情节上下文和混合上下文在长距离下保持了最高的准确性。对对抗性问题的分析进一步表明，能够回答的问题的强表现并不一定能转化为在可用历史不支持请求结论时的成功弃权。这些发现表明，可靠的纵向推理不仅依赖于LLM可以访问多少历史，更关键的是如何选择和呈现相关证据以进行推理。
cs.CL / 24 / 2610.00568
Emergent Unfaithfulness: How Alignment Training Causes Language Models to Silently Override Task Faithfulness
新兴的不忠实性：对齐训练如何导致语言模型默默地覆盖任务忠实性
Abstract
Large language models are characterized by three key properties: capability, alignment, and faithfulness. Prior work studies the tradeoffs between capability and alignment, and between capability and faithfulness, but a third tension remains underexplored: the alignment-faithfulness conflict. We show that aligned models systematically deviate from their inputs on unsafe or sensitive content without disclosing the modification, a failure mode we call alignment-induced unfaithfulness (AIU). Unlike capability-driven unfaithfulness, which comes from errors in knowledge or reasoning, this is induced by post-training mechanisms that override adherence to the input. We introduce FaithConflict, a controlled dataset isolating both conflicts, and two complementary taxonomies: behavioral (B1-B8) and chain-of-thought reasoning (C0-C6). Across models, AIU increases with scale and more sharply than capability-driven unfaithfulness, a reverse scaling law; intermediate checkpoints show it is amplified during post-training, with DPO the stage at which the gap both grows most and becomes least visible. Prompting-based mitigation does not resolve it, revealing a capability-alignment-faithfulness trilemma in the design and evaluation of LLMs.
Chinese Translation
大型语言模型具有三个关键特性：能力、对齐和忠实性。先前的研究探讨了能力与对齐之间、能力与忠实性之间的权衡，但第三种张力——对齐与忠实性之间的冲突仍然未得到充分研究。我们展示了对齐模型在处理不安全或敏感内容时，系统性地偏离其输入而不披露修改，这种失败模式我们称之为对齐引发的不忠实性（alignment-induced unfaithfulness, AIU）。与能力驱动的不忠实性不同，后者源于知识或推理中的错误，而AIU是由后训练机制引发的，这些机制覆盖了对输入的遵循。我们引入了FaithConflict，一个控制数据集，用于隔离这两种冲突，并提出了两个互补的分类法：行为分类（B1-B8）和思维链推理（C0-C6）。在不同模型中，AIU随着规模的增加而增加，且增幅显著高于能力驱动的不忠实性，呈现出一种反向缩放法则；中间检查点显示，AIU在后训练过程中被放大，其中DPO阶段是差距增长最多且最不明显的阶段。基于提示的缓解措施并未解决这一问题，揭示了在大型语言模型的设计和评估中存在能力、对齐和忠实性之间的三难困境。
cs.CL / 25 / 2610.00606
Where's Waldo? Query-language Preference under Cross-lingual Knowledge Disparities
沃尔多在哪里？跨语言知识差异下的查询语言偏好
Abstract
Large Language Models increasingly serve as interfaces for knowledge-intensive information seeking tasks across languages by synthesizing multilingual evidence. Prior work has shown that they often exhibit query-language preference -- the tendency to favor sources written in the language of the query -- but has largely examined this behavior in settings where equivalent knowledge is available across languages. However, this bias becomes consequential when sources in different languages provide incomplete or inconsistent accounts of the same fact, since the information users receive then depends on the sources a model selects to use. To characterize query-language preference under such cross-lingual knowledge disparities, we introduce Waldo, a multilingual Question-Answering (QA) benchmark constructed from Wikipedia. Waldo contains 12K QA pairs targeting knowledge gaps, where a fact is available in one language but absent in another, and knowledge conflicts, where language editions provide conflicting versions of the same fact. Evaluating eight models across five languages, we find that when one language edition merely lacks the relevant fact, models generally use evidence from the other language regardless of the query language. Under conflicting accounts, however, model responses strongly align with the document in the query language, causing semantically equivalent queries to elicit different accounts depending on the user's language. Finally, we explore two different approaches that could mitigate this preference under knowledge conflicts: a mechanistic intervention that ablates attention heads associated with query-language preference, and LoRA-based training, which reduces the preference gap by up to 61.5%.
Chinese Translation
大型语言模型越来越多地作为跨语言知识密集型信息检索任务的接口，通过综合多语言证据来提供服务。先前的研究表明，它们通常表现出查询语言偏好——即倾向于优先使用与查询语言相同的来源——但大多数研究是在各语言间知识相对等的环境中进行的。然而，当不同语言的来源对同一事实提供不完整或不一致的描述时，这种偏见变得尤为重要，因为用户获得的信息依赖于模型选择使用的来源。为了描述在这种跨语言知识差异下的查询语言偏好，我们引入了Waldo，一个基于维基百科构建的多语言问答（QA）基准。Waldo包含12,000对针对知识缺口的问答对，其中某一事实在一种语言中存在而在另一种语言中缺失，以及知识冲突，其中语言版本提供了同一事实的相互矛盾的描述。对五种语言中的八个模型进行评估，我们发现，当一种语言版本仅缺少相关事实时，模型通常会使用另一种语言的证据，而不考虑查询语言。然而，在存在冲突的描述下，模型的响应与查询语言中的文档强烈一致，导致语义上等价的查询根据用户的语言产生不同的描述。最后，我们探讨了两种可能减轻这种知识冲突下偏好的方法：一种机械干预，消除与查询语言偏好相关的注意力头，以及基于LoRA的训练，能够将偏好差距减少多达61.5%。
cs.CL / 26 / 2610.00610
Explainable Suicide Risk Assessment on Social Media with Multi-Task QLoRA
基于多任务QLoRA的社交媒体可解释自杀风险评估
Abstract
Explainable suicide-risk assessment requires models not only to estimate risk severity, but also to identify supporting language and the risk and protective factors expressed in a post. We present our system for the IEEE BigData 2026 Cup on Explainable Suicide Risk Assessment on Social Media, which addresses three tasks: risk-level classification, evidence phrase extraction, and multi-label factor identification. Our approach adapts Qwen2.5-Instruct models using quantized low-rank adaptation (QLoRA) and an answer-masked causal language-model objective. We jointly train across all three tasks for risk classification, jointly train on Tasks~1a and 1b for evidence extraction, and adapt Task~2 separately for factor identification. We also tailor aggregation to each output: we average risk-level probabilities from the 32B and 72B models, combine evidence phrases through cross-fold consensus, and calibrate factor-specific decisions through rate matching based on out-of-fold operating points. On the official leaderboard, the final system achieved a composite score of 0.7738, with 0.8089 on Task~1 and 0.6919 on Task~2. Across the evaluated configurations, three-task training performed best for Task~1a, joint training on Tasks~1a and 1b performed best for Task~1b, and task-specific training performed best for Task~2. Probability averaging further improved Task~1a when component models had complementary errors. These findings highlight the value of tailoring both training objectives and aggregation strategies to the output structure of each task within a unified language-model framework.
Chinese Translation
可解释的自杀风险评估要求模型不仅能够估计风险严重性，还能识别帖子中表达的支持性语言及风险和保护因素。我们展示了针对IEEE BigData 2026 Cup的系统，该系统专注于社交媒体上的可解释自杀风险评估，涵盖三个任务：风险级别分类、证据短语提取和多标签因素识别。我们的方法采用了量化低秩适应（QLoRA）和答案掩蔽因果语言模型目标，对Qwen2.5-Instruct模型进行了调整。我们在所有三个任务上进行联合训练以实现风险分类，在任务1a和1b上进行联合训练以提取证据，并单独调整任务2以进行因素识别。我们还为每个输出量身定制聚合策略：我们对32B和72B模型的风险级别概率进行平均，通过交叉折叠共识结合证据短语，并通过基于折外操作点的速率匹配来校准特定因素的决策。在官方排行榜上，最终系统的综合得分为0.7738，任务1得分为0.8089，任务2得分为0.6919。在评估的配置中，三任务训练在任务1a上表现最佳，任务1a和1b的联合训练在任务1b上表现最佳，而任务特定训练在任务2上表现最佳。当组件模型具有互补错误时，概率平均进一步改善了任务1a的表现。这些发现突显了在统一语言模型框架内，根据每个任务的输出结构量身定制训练目标和聚合策略的价值。
cs.CL / 27 / 2610.00621
Mixture of Decoders for Diverse Dialog Response Generation
多解码器混合模型用于多样化对话响应生成
Abstract
Mixture modeling is a long established machine learning technique for learning large sets of multi-modal data. While it is known that sequence-to-sequence models for dialog response generation suffer from the problem of low diversity, we hypothesize that it is because sequence-to-sequence models tend to learn a degenerate uni-modal distribution of responses. We then propose to incorporate a mixture of decoders into sequence-to-sequence models and try to make each decoder learn specialized topics in order to improve the diversity of generated responses. Our model is developed under the framework of conditional variational autoencoder (CVAE). We evaluate our approach on an open domain chat corpus and show improvement over strong baselines in quantitative measures and human evaluation.
Chinese Translation
混合建模是一种长期建立的机器学习技术，用于学习大规模的多模态数据。虽然已知序列到序列模型在对话响应生成中存在低多样性的问题，但我们假设这是因为序列到序列模型倾向于学习退化的单模态响应分布。因此，我们提出将多解码器融入序列到序列模型中，并尝试使每个解码器学习特定主题，以提高生成响应的多样性。我们的模型是在条件变分自编码器（Conditional Variational Autoencoder, CVAE）框架下开发的。我们在开放域聊天语料库上评估了我们的方法，并在定量指标和人工评估中显示出相对于强基线的改进。
cs.CL / 28 / 2610.00650
Self-Evolving Coding Rules for AI Coding Agents
自我演化的编码规则用于人工智能编码代理
Abstract
The performance of AI coding agents is highly dependent on their underlying coding rules. However, existing coding rules are typically hand-crafted and fixed, making the process labor-intensive and often suboptimal. In this work, we propose RuleEvolve, a self-evolving framework for coding rules. RuleEvolve maintains a pool of candidate coding rules and iteratively improves them. In each iteration, it employs an LLM-powered mutator module to generate variants from existing candidates, and then uses a judge module to evaluate these variants and update the pool with the best-performing ones. Extensive evaluations across two coding-agent frameworks, four backbone LLMs, and three benchmarks demonstrate that RuleEvolve outperforms both manual engineering and existing prompt optimization baselines in terms of functional correctness of the generated code, code length, and/or generation cost (e.g., tokens used).
Chinese Translation
人工智能编码代理的性能高度依赖于其基础编码规则。然而，现有的编码规则通常是手工制作且固定的，这使得这一过程劳动密集且往往效果不佳。在本研究中，我们提出了RuleEvolve，一个自我演化的编码规则框架。RuleEvolve维护一个候选编码规则池，并对其进行迭代改进。在每次迭代中，它利用一个基于大型语言模型（LLM）的变异模块从现有候选中生成变体，然后使用评估模块对这些变体进行评估，并用表现最佳的变体更新候选池。在两个编码代理框架、四个主干LLM和三个基准测试中的广泛评估表明，RuleEvolve在生成代码的功能正确性、代码长度和/或生成成本（例如，使用的标记数）方面均优于手动工程和现有的提示优化基准。
cs.CL / 29 / 2610.00656
Lingtai: What Concept Geometry Reveals--and Does Not Reveal--About LLM Inference
Lingtai：概念几何揭示了（以及未揭示）关于大型语言模型推理的内容
Abstract
Observing what a large language model computes during autoregressive inference--online and without training probes--remains difficult. We introduce Lingtai, a training-free concept telemetry layer: at each generation step, residual states are projected onto a domain-specific bank of named concept anchors, constructed without labeled concept examples, outcome labels, gradient fitting, or activation-space optimization, producing a structured per-step concept-coordinate signal. Across code generation and grade-school mathematical reasoning, this signal exhibits a robust association with predictive uncertainty: the association survives problem-identity and token-position controls and is not attributable to a single token type, is not explained by a simple correct/incorrect mixture on GSM8K, and is not reproduced by matched random anchors; it is markedly weaker or direction-inconsistent in K-means and PCA projections. Two structures emerge: a recurring uncertainty-linked activity signal whose functional geometry is task-conditioned (distinct activity-entropy shapes on HumanEval, MBPP, and GSM8K), and an execution-specific trajectory identity with strong local inertia but weak re-instantiation invariance--under completion-only elastic alignment, corruption at k=32 (approximately a median quarter of the completion) on the matched re-execution subset still retrieves the archived episode at 62.0%, while a fresh execution retrieves it only 11.7-16.0% of the time. Finally, a matched audit finds no evidence that the scalar concept-activity signal used here supplies a stable correctness coordinate under the tested protocol; we therefore treat correctness as externally supplied. Telemetry adds 0.7-1.6% per-token decode overhead for the 161-anchor code implementation, with unchanged generated tokens.
Chinese Translation
观察大型语言模型在自回归推理过程中计算的内容——在线且无需训练探针——仍然是一个挑战。我们引入了Lingtai，一个无训练的概念遥测层：在每个生成步骤中，残余状态被投影到一个特定领域的命名概念锚点库上，该库是在没有标记概念示例、结果标签、梯度拟合或激活空间优化的情况下构建的，从而生成一个结构化的逐步概念坐标信号。在代码生成和小学数学推理中，该信号与预测不确定性表现出强烈的关联：这种关联在问题身份和标记位置控制下依然存在，且不能归因于单一的标记类型，不是由GSM8K上的简单正确/错误混合所解释，也无法通过匹配的随机锚点重现；在K均值和主成分分析（PCA）投影中，其强度明显较弱或方向不一致。两个结构浮现：一个与不确定性相关的活动信号，其功能几何形状受任务条件影响（在HumanEval、MBPP和GSM8K上表现出不同的活动熵形状），以及一个特定执行轨迹身份，具有强局部惯性但弱重现不变性——在仅完成的弹性对齐下，匹配的重新执行子集在k=32（大约是完成的中位数四分之一）时的损坏仍能以62.0%的概率检索到归档的情节，而新的执行仅能以11.7-16.0%的概率检索到。最后，匹配审计未发现这里使用的标量概念活动信号在测试协议下提供稳定的正确性坐标的证据；因此，我们将正确性视为外部提供。遥测为161个锚点的代码实现增加了每个标记解码的0.7-1.6%的开销，生成的标记数量保持不变。
cs.CL / 30 / 2610.00664
PhysicsMate: A Curriculum-Grounded Bengali Benchmark for Secondary Physics QA with Small-Model Adaptation
PhysicsMate：基于课程的孟加拉语中学物理问答基准与小模型适应
Abstract
Bengali secondary education lacks curriculum-grounded benchmarks for STEM question-solving, and general-purpose language models struggle with the precise terminology, unit conventions, and derivations that physics problems demand. We introduce PhysicsMate, a benchmark of 1834 question-answer pairs built from the National Curriculum and Textbook Board (NCTB) Grade 9-10 physics syllabus and grounded in a multi-relational knowledge graph of 1760 nodes and 2600 edges across ten ontological types. We Low-Rank Adapt at 0.6B, 1.7B, and 4B parameters, with a unified recipe and demonstrate a significant increase in closed-book accuracy in all scales (+5.5, +15.0, and +23.3 percentage points). A node-type analysis shows that the most benefited by adaptation is the structured curricular knowledge, which consists of physical quantities and named laws, while the least benefited is the loosely specified entity-level knowledge. The 4B model has been adapted and quantized to a small offline binary that can be used for local inference in resource constrained environments and offers a viable path to curriculum aligned physics support in environments with limited connectivity and hardware.
Chinese Translation
孟加拉国的中等教育缺乏基于课程的STEM问题解决基准，而通用语言模型在物理问题所需的精确术语、单位惯例和推导方面表现不佳。我们介绍了PhysicsMate，这是一个由1834对问答组成的基准，基于国家课程和教材委员会（NCTB）9-10年级物理课程大纲，并建立在一个包含1760个节点和2600条边的多关系知识图谱上，涵盖十种本体类型。我们在0.6B、1.7B和4B参数的模型上进行了低秩适应，采用统一的方案，并在所有规模上显著提高了闭卷准确率（分别增加了5.5、15.0和23.3个百分点）。节点类型分析显示，适应最有利的是结构化课程知识，包括物理量和命名定律，而受益最小的是松散指定的实体级知识。4B模型已被适应并量化为一个小型离线二进制文件，可以在资源受限的环境中用于本地推理，并为在连接性和硬件有限的环境中提供与课程对齐的物理支持提供了可行路径。
cs.CL / 31 / 2610.00673
Closing the Loop: Practical Training Recipes for Looped Language Models
闭环：循环语言模型的实用训练方案
Abstract
Looped language models increase effective depth by repeatedly applying a shared block of layers, but existing large-scale recipes require multi-stage training over trillions of tokens, while the benefits of recurrence remain difficult to separate from differences in data and training. In this work, we establish practical training recipes for looped language models, with three main results. (1) We develop a compute-efficient from-scratch pipeline that reduces the training budget from 7.7T tokens in Ouro to 310B tokens while retaining strong reasoning performance. Pretraining followed by high-quality mid-training, together with learning-rate warmup and stronger exit-gate regularization, enables stable recurrent training without prior multi-stage schedules. (2) Under controlled comparisons, our 1.4B LoopLM outperforms a parameter-matched dense model trained on the same data and token budget on all 12 evaluated benchmarks, including +14 points on GSM8K, +10 on MATH, and +22 on DROP. At matched inference compute, it approaches a 3.9B dense model on mathematical reasoning and reading comprehension while using only 36\% as many parameters. (3) We introduce a minimal recipe for converting pretrained dense models into looped ones: a single learned input-mixing scalar and a smoothed exit loss, with no step-specific parameters. Applied to Qwen3-1.7B-Base, Looped Qwen improves over an identically continued dense baseline on every evaluated benchmark across two data regimes, with statistically clear gains on GSM8K, MATH, and MMLU-Pro on the curated mixture. Together, these results make looped language models substantially cheaper to train from scratch and practical to introduce into existing pretrained checkpoints, while isolating the gains due to recurrence itself.
Chinese Translation
循环语言模型通过反复应用共享的层块来增加有效深度，但现有的大规模训练方案需要在数万亿个标记上进行多阶段训练，而递归的好处仍然难以与数据和训练的差异分开。在本研究中，我们为循环语言模型建立了实用的训练方案，主要结果有三点：(1) 我们开发了一种计算高效的从零开始的训练流程，将Ouro中的训练预算从7.7万亿个标记减少到3100亿个标记，同时保持强大的推理性能。预训练后进行高质量的中期训练，结合学习率预热和更强的退出门正则化，使得在没有先前多阶段计划的情况下实现稳定的递归训练。(2) 在受控比较下，我们的1.4B LoopLM在所有12个评估基准上优于在相同数据和标记预算下训练的参数匹配的稠密模型，包括在GSM8K上提高14分，在MATH上提高10分，在DROP上提高22分。在匹配的推理计算下，它在数学推理和阅读理解方面接近一个3.9B的稠密模型，同时仅使用36%的参数。(3) 我们引入了一种将预训练的稠密模型转换为循环模型的最小方案：一个单一的学习输入混合标量和一个平滑的退出损失，没有特定于步骤的参数。应用于Qwen3-1.7B-Base时，循环Qwen在两个数据范畴的每个评估基准上都优于一个同样延续的稠密基线，在GSM8K、MATH和MMLU-Pro的策划混合上有统计显著的提升。综合来看，这些结果使得从零开始训练循环语言模型的成本大幅降低，并且在现有的预训练检查点中引入循环模型变得可行，同时隔离了由于递归本身带来的收益。
cs.CL / 32 / 2610.00679
Bayesian Fine-tuning Yields Language Models that are as Bayesian as their Beliefs Allow
贝叶斯微调产生的语言模型与其信念所允许的贝叶斯性相符
Abstract
Language models (LMs) are increasingly used for tasks that require reasoning about hidden variables from a few observations, for which Bayesian inference is the normatively correct solution. While supervised fine-tuning of an LM on the outputs of an optimal $\textit{Bayesian}$ model leads to near-Bayesian behavior, standard supervised fine-tuning (SFT) on the true answers for the task falls short of it. But behavior alone does not tell us $\textit{why}$ tuning on a $\textit{Bayesian}$ or an $\textit{oracle}$ (true answers) signal differs: whether the resulting LM represents Bayesian beliefs, acts on them, or turns them into a choice the way Bayes' rule does. To compare them, we formulate increasingly demanding requirements for an LM to count as a Bayesian decision maker, spanning its behavior, representations, and computations, and test them on a flight recommendation task. The Bayes-trained LM acts Bayesian, encodes quantities of Bayes' rule in its middle layers, and uses the encoded belief for the recommendation to a certain extent. The oracle-trained LM differs from it both in the beliefs it holds and whether it reads beliefs out into recommendations. Exchanging beliefs between the LMs transfers a part of the Bayesian advantage. Bayes fine-tuning thus installs usable Bayesian beliefs in an LM for reasoning under uncertainty in a way standard SFT on oracle answers cannot, highlighting the advantage of nuanced supervision.
Chinese Translation
语言模型（LM）越来越多地用于需要从少量观察中推理隐藏变量的任务，而贝叶斯推理是规范上正确的解决方案。虽然在最优的贝叶斯模型输出上对LM进行监督微调会导致近似贝叶斯行为，但在任务的真实答案上进行标准监督微调（SFT）则未能达到这一点。然而，仅凭行为并不能告诉我们为何在贝叶斯或神谕（真实答案）信号上的微调会有所不同：最终的LM是否代表贝叶斯信念、是否基于这些信念采取行动，或是像贝叶斯法则那样将其转化为选择。为了比较这两者，我们制定了越来越严格的要求，使LM能够被视为贝叶斯决策者，这些要求涵盖了其行为、表征和计算，并在一个航班推荐任务上进行了测试。经过贝叶斯训练的LM表现出贝叶斯行为，在其中间层编码了贝叶斯法则的量，并在一定程度上利用编码的信念进行推荐。与之不同的是，经过神谕训练的LM在其持有的信念以及是否将信念转化为推荐方面有所不同。在LM之间交换信念会转移一部分贝叶斯优势。因此，贝叶斯微调为LM安装了可用于不确定性推理的贝叶斯信念，而标准的神谕答案SFT则无法实现，这突显了细致监督的优势。
cs.CL / 33 / 2610.00689
Towards Robust Numerical Claim Verification
朝向稳健的数值声明验证
Abstract
Large language models (LLMs) are widely used for claim verification, yet remain brittle for numerical reasoning: even small changes in value can sharply degrade accuracy. We show that this brittleness persists in frontier LLMs, but can be mitigated through adversarial fine-tuning on numerically perturbed examples. Using parameter-efficient fine-tuning, small Qwen3 models (0.6B$\unicode{x2013}$8B) reach 98.7% accuracy on label-flipping perturbations, outperforming larger zero-shot models and frontier systems (GPT-5.4 Pro (74.0%) and Gemini 2.5 Flash (73.9%)). The gains generalise to unseen perturbation types, indicating robust numerical decision boundaries rather than memorised edits. Robustness also transfers without target-domain data, significantly improving cross-lingual performance in Spanish. We further show that the same fine-tuning recipe confers robustness to evidence-side perturbations, using the VitaminC dataset.
Chinese Translation
大型语言模型（LLMs）广泛应用于声明验证，但在数值推理方面仍然脆弱：即使是微小的数值变化也会显著降低准确性。我们展示了这种脆弱性在前沿LLMs中依然存在，但可以通过对数值扰动示例进行对抗性微调来减轻。使用参数高效的微调，小型Qwen3模型（0.6B–8B）在标签翻转扰动上达到98.7%的准确率，超越了更大型的零-shot模型和前沿系统（GPT-5.4 Pro（74.0%）和Gemini 2.5 Flash（73.9%））。这一提升在未见过的扰动类型中也具有普遍性，表明了稳健的数值决策边界，而非记忆化的编辑。稳健性在没有目标领域数据的情况下也能转移，显著提高了西班牙语的跨语言性能。我们进一步展示了相同的微调方法也能赋予证据侧扰动的稳健性，使用了VitaminC数据集。
cs.CL / 34 / 2610.00694
How Divergence Becomes Decision Flips in Compressed Language Models
压缩语言模型中的发散如何转变为决策翻转
Abstract
Compression reports summarize how far a compressed language model moved from the dense one, usually by a KL divergence; a deployment that relies on the dense model's outputs needs to know how many of its decisions changed. We show that total variation, not KL, answers this directly. Across 802 compressed and perturbed copies of 19 open models on five corpora and nine mechanically unrelated perturbation families, the rate at which the arg-max token changes (the \emph{flip rate}) tracks total variation at a ratio with median $1.05$, with no fitted constant. KL converts into flips only through its square root and a factor that varies fourfold across models and corpora, because KL averages over tokens before the root is taken; first-order statistics averaged per token, such as Hellinger distance, avoid this, but reports rarely give them. As a result, of two compressors reported on different models and corpora whose flip rates differ by at least $10%$, KL assigns the smaller divergence to the one that changes more decisions in $11%$ of cases, total variation in $1%$. Two pre-registered tests mark the limits: on a held-out code corpus the ratio held for all eight models while three predictions about KL each failed for half of them or more, and on three new models with real kernels it stayed in its band for 37 of 38 checkpoints but fell below one on code for two models. In vLLM speculative decoding, total variation measured under teacher forcing predicts greedy draft acceptance with a mean relative error of $1.1$--$2.4%$, without the task-specific calibration that KL needs.
Chinese Translation
压缩报告总结了压缩语言模型与稠密模型之间的差异，通常通过KL散度来表示；依赖于稠密模型输出的部署需要了解其决策改变的数量。我们展示了总变差（total variation），而非KL散度，能够直接回答这一问题。在对五个语料库和九个机械无关的扰动家族的19个开放模型进行的802个压缩和扰动副本中，arg-max标记变化的速率（即翻转率）与总变差的比率中位数为1.05，且没有拟合常数。KL散度仅通过其平方根和一个在模型和语料库间变化四倍的因子转化为翻转，因为KL在取平方根之前对标记进行了平均；每个标记的第一阶统计量，如Hellinger距离，避免了这一点，但报告中很少提供这些结果。因此，在不同模型和语料库上报告的两个压缩器，其翻转率相差至少10%的情况下，KL在11%的情况下将较小的散度分配给改变更多决策的模型，而总变差仅在1%的情况下如此。两个预注册测试标记了极限：在一个保留的代码语料库上，所有八个模型的比率保持一致，而关于KL的三个预测中，每个预测在一半或更多模型上均未能成立；在三个具有真实核的新模型上，该比率在38个检查点中有37个保持在其范围内，但在两个模型的代码上降至1以下。在vLLM的推测解码中，在教师强制下测量的总变差预测贪婪草稿接受的平均相对误差为1.1%至2.4%，而KL需要特定任务的校准。
cs.CL / 35 / 2610.00717
Sequential Functional Structured Tucker Compression for Large Language Model Attentions
大语言模型注意力的顺序功能结构塔克压缩
Abstract
Post-training compression of LLM attention is often formulated as independent matrix approximation, ignoring both the shared structure among attention projections and the representation shift introduced by earlier compression. We propose FTC, a sequential structured compression framework that adapts the approximation to the current compressed model while jointly exploiting the native Q/K/V head structure under a fixed storage budget. The output projection is handled separately to account for the changed post-attention representation. FTC requires neither fine-tuning nor gradient-based recovery. Across seven decoder-only LLMs from 6B to 32B parameters, FTC achieves the lowest WikiText-2 perplexity among the compared methods at every tested keep ratio on five modern GQA models, with the largest gains under aggressive compression. The improvements transfer to downstream tasks and remain substantial at the 32B scale.
Chinese Translation
大语言模型（LLM）注意力的后训练压缩通常被表述为独立的矩阵近似，忽视了注意力投影之间的共享结构以及早期压缩引入的表示偏移。我们提出了FTC，一种顺序结构压缩框架，该框架在固定存储预算下，适应当前压缩模型的近似，同时共同利用原生的Q/K/V头结构。输出投影被单独处理，以考虑注意力后表示的变化。FTC不需要微调或基于梯度的恢复。在六个参数从6B到32B的仅解码器LLM中，FTC在每个测试的保持比率下，在五个现代GQA模型中实现了与比较方法相比最低的WikiText-2困惑度，并在激进压缩下获得了最大的提升。这些改进转移到下游任务，并在32B规模下仍然显著。
cs.CL / 36 / 2610.00724
Reason in Style: Discovering and Controlling Style in Language Models
风格中的推理：在语言模型中发现和控制风格
Abstract
Language models learn content and style jointly, making stylistic variation in their outputs difficult to identify and control. We study whether recurring styles in model responses can be discovered without supervision and explicitly controlled. We design an algorithm that learns to separate representations of content and style from language models' outputs and validate its effectiveness on math questions in a controlled setting. By applying this method to over 100K verified traces from nine distinct teacher models, we discover six recurring yet imbalanced styles. We then fine-tune smaller student models to follow these styles when explicitly conditioned on them, using importance weighting to balance the contribution of the styles represented in the corpus. This approach improves Pass@$k$ over standard fine-tuning on the same data across six math reasoning benchmarks, demonstrating that we can diversify the style of answers effectively. We confirm that this also results in strong correspondence between requested and realized styles. We find that style affects correctness: the probability of solving a problem depends on the style we condition on, and different problems benefit from different styles. In summary, our results show that stylistic variation in model-generated data can be discovered in an unsupervised way, and made explicit, providing a source of both control and improved reasoning performance.
Chinese Translation
语言模型共同学习内容和风格，使得其输出中的风格变异难以识别和控制。我们研究了是否可以在没有监督的情况下发现模型响应中的重复风格并进行显式控制。我们设计了一种算法，能够从语言模型的输出中分离内容和风格的表示，并在受控环境下验证其有效性，特别是在数学问题上。通过将该方法应用于来自九个不同教师模型的超过10万条经过验证的追踪记录，我们发现了六种重复但不平衡的风格。随后，我们对较小的学生模型进行微调，使其在显式条件下遵循这些风格，并使用重要性加权来平衡语料库中所表示风格的贡献。这种方法在六个数学推理基准上，相较于在相同数据上进行标准微调，显著提高了 Pass@$k$，证明我们能够有效地多样化答案的风格。我们确认这也导致了请求风格与实现风格之间的强对应关系。我们发现风格影响正确性：解决问题的概率依赖于我们所条件的风格，而不同的问题则从不同的风格中受益。总之，我们的结果表明，模型生成数据中的风格变异可以以无监督的方式被发现并显式化，从而提供控制和改善推理性能的来源。
cs.CL / 37 / 2610.00779
Effective Synthetic Data Curation Requires Group-Level Signals
有效的合成数据管理需要群体级信号
Abstract
Synthetic data now is essential to LLM training, used to strengthen advanced capabilities such as autonomous and long-horizon task execution. Yet recent work shows that training on it at scale can degrade model generation, making it important to decide what synthetic data is worth training on. While current data curation practices do so with individual-level signals (i.e., estimates of each data sample's training utility in isolation), across pre-training and post-training settings we show that this is insufficient for synthetic data, and that group-level signals (i.e., estimates of utility that account for interactions among data samples) are necessary for effective data curation. First, we show that individual-level signals are blind to how samples jointly affect training: synthetic datasets with different compositions can be indistinguishable under individual-level influence yet differ sharply under group-level influence, and curating by the latter yields better downstream performance, particularly in generative capability. Second, we find that group-level signals matter more as training pipelines become increasingly synthetic: among widely used data curation methods, only those incorporating them improve over baseline, with gains increasing when weights capturing relations among samples are amplified. Finally, we translate these findings into practice -- for model developers under a compute budget, we offer a cheap diagnostic that prioritizes which groups of synthetic data most need group-level estimation, recovering much of the benefit of full group-level scoring at a fraction of the compute cost.
Chinese Translation
合成数据如今对大型语言模型（LLM）的训练至关重要，用于增强自主执行和长时间任务执行等高级能力。然而，近期研究表明，大规模训练合成数据可能会降低模型生成效果，因此决定哪些合成数据值得进行训练变得尤为重要。目前的数据管理实践主要依赖个体级信号（即对每个数据样本在孤立状态下的训练效用估计），但我们在预训练和后训练设置中表明，这对于合成数据而言是不够的，群体级信号（即考虑数据样本之间相互作用的效用估计）对于有效的数据管理是必要的。首先，我们展示个体级信号无法识别样本如何共同影响训练：不同组成的合成数据集在个体级影响下可能无法区分，但在群体级影响下却可能有显著差异，基于后者进行管理能够带来更好的下游性能，尤其是在生成能力方面。其次，我们发现随着训练管道越来越依赖合成数据，群体级信号的重要性也随之增加：在广泛使用的数据管理方法中，只有那些纳入群体级信号的方法在基线之上有所提升，并且当捕捉样本之间关系的权重被放大时，收益会进一步增加。最后，我们将这些发现转化为实践——对于在计算预算下的模型开发者，我们提供了一种廉价的诊断方法，优先考虑哪些合成数据组最需要群体级估计，从而在计算成本的很小一部分中恢复大部分完整群体级评分的好处。
cs.CL / 38 / 2610.00795
Can large language models unlock discrete data in ophthalmic diagnostic reports?
大型语言模型能否解锁眼科诊断报告中的离散数据？
Abstract
Objective: To assess the accuracy and efficiency of a large language model (LLM) using two prompt strategies to extract structured data from ophthalmic diagnostic PDF reports. Methods: Twenty deidentified reports across four types (Visual Field, OCT Glaucoma Overview, OCT retinal nerve fiber layer Single Exam, and OCT Thickness Map; n = 5 each) were processed using two GPT-4o-assisted pipelines and compared with a reconciled manual ground truth. Schema-Constrained used Structured Output mode with a predefined JSON Schema; Prompt-Only used a detailed instruction prompt followed by Python conversion to JSON. Outcomes were value accuracy, formatting accuracy, and extraction time. Results: Schema-Constrained value accuracy was 100.00% for Visual Field and RNFL Single Exam, 97.45% for Glaucoma Overview, and 98.00% for Thickness Map; Prompt-Only achieved 100.00% across all four report types. Formatting accuracy was 100.00% for Schema-Constrained across all report types and 100.00% for Prompt-Only except RNFL Single Exam (90.14%). Mean extraction time was 56.51 s per report for manual review versus 5.04 s for Schema-Constrained and 4.70 s for Prompt-Only, an approximately 92% reduction. Conclusions: In this small proof-of-concept dataset, general-purpose LLM-assisted pipelines extracted structured data from ophthalmic diagnostic PDFs with high accuracy and substantially reduced processing time. Prompt-Only achieved the highest value accuracy, while Schema-Constrained produced schema-compliant output with 100% formatting accuracy. These complementary strengths support further evaluation of hybrid, validation-aware workflows for research and clinical data abstraction.
Chinese Translation
目的：评估使用两种提示策略的大型语言模型（LLM）从眼科诊断PDF报告中提取结构化数据的准确性和效率。方法：处理了20份去标识化报告，涵盖四种类型（视野检查、OCT青光眼概述、OCT视网膜神经纤维层单次检查和OCT厚度图；每种类型各5份），使用两种GPT-4o辅助管道，并与经过调和的手动基准进行比较。Schema-Constrained使用带有预定义JSON Schema的结构化输出模式；Prompt-Only使用详细的指令提示，随后通过Python转换为JSON。结果指标包括值的准确性、格式的准确性和提取时间。结果：Schema-Constrained在视野检查和RNFL单次检查中的值的准确性为100.00%，在青光眼概述中为97.45%，在厚度图中为98.00%；Prompt-Only在所有四种报告类型中均达到了100.00%。在所有报告类型中，Schema-Constrained的格式准确性为100.00%，而Prompt-Only在RNFL单次检查中为90.14%。手动审核的平均提取时间为每份报告56.51秒，而Schema-Constrained为5.04秒，Prompt-Only为4.70秒，约减少了92%。结论：在这一小规模的概念验证数据集中，通用型LLM辅助管道以高准确性和显著减少的处理时间从眼科诊断PDF中提取了结构化数据。Prompt-Only实现了最高的值的准确性，而Schema-Constrained则生成了符合模式的输出，格式准确性为100%。这些互补的优势支持对混合型、关注验证的工作流程在研究和临床数据抽取中的进一步评估。
cs.CL / 39 / 2610.00827
Verbalized and Internal Probabilities Are Coupled in Large Language Models
大型语言模型中的语言化和内部概率是相互关联的
Abstract
Large language models carry an internal notion of uncertainty in their sampling distribution, i.e., the probabilities they place on generating one answer rather than another. They can also be asked to state a confidence, in words or as a number: a verbalized uncertainty. Prior work suggests that internal probabilities track relative frequencies in the training data, and that verbalized probabilities track explicit probabilistic assertions in the training data. However, we do not know whether these two readouts are aligned, except when frequencies and probabilistic assertions in the training data happen to align. This limits our understanding of when we can use verbalized uncertainties as a proxy for either training data frequencies, or a model's internal distribution. We resolve this gap by systematically exploring how LLMs probability readouts are impacted by training and in-context data, via intervening on the underlying uncertainty sources in the data. We find that both internal and verbalized probability readouts are impacted by both distributional and asserted uncertainty in the training data. Further, we find that verbalized and internal probabilities are aligned beyond what would be expected by independently tracking the same uncertainty sources, suggesting that verbalized probabilities can be used to probe a model's internal distribution.
Chinese Translation
大型语言模型在其采样分布中具有一种内部的不确定性概念，即它们对生成某个答案而非另一个答案所赋予的概率。它们还可以被要求以语言或数字的形式陈述置信度：一种语言化的不确定性。先前的研究表明，内部概率跟踪训练数据中的相对频率，而语言化概率则跟踪训练数据中的明确概率断言。然而，我们不知道这两种输出是否一致，除非训练数据中的频率和概率断言恰好一致。这限制了我们对何时可以将语言化的不确定性作为训练数据频率或模型内部分布的代理的理解。我们通过系统地探讨大型语言模型的概率输出如何受到训练和上下文数据的影响，来填补这一空白，方法是干预数据中潜在的不确定性来源。我们发现，内部和语言化的概率输出都受到训练数据中分布性和断言性不确定性的影响。此外，我们发现语言化概率和内部概率的对齐程度超出了独立跟踪相同不确定性来源所能预期的程度，这表明语言化概率可以用来探测模型的内部分布。
cs.CL / 40 / 2610.00833
VERITYGATE: A Four-Gate Schema-Level Faithfulness Framework and Paired Benchmark for Grounded LLM Narrations over Structured Evidence
VERITYGATE：一个四门检查的模式级忠实性框架及基于结构证据的有据大型语言模型叙述的配对基准
Abstract
Fluent LLM explanations may not follow the evidence from a structured system. We present VERITYGATE, a four-gate checker for declared evidence IDs, entities, numbers, and claim types. It checks a fixed schema; it does not verify every fact in the prose. At r=0 and r=1, we test 900 instances per setting (450 grounded-ungrounded pairs) with GPT-4o-mini, Llama-3.3-70B, and Claude Sonnet 4.6. Under this schema-level contract and before repair, 80.3% of mini claims and 47.9% of Sonnet claims fail. These are verifier rejection rates, not prose-hallucination rates. One repair pass raises claim survival from 19.7% to 28.0% for mini and from 52.1% to 54.3% for Sonnet. Verified claims per example change by +0.14 for mini, -0.71 for Llama, and -0.47 for Sonnet, so survival and output volume must be reported together. A second Sonnet pass gives no clear gain. At r=1, Gate 4 covers 97.0%, 98.7%, and 100% of failing claims for mini, Llama, and Sonnet. Small human studies support the rules but show gaps between schema checks and correct prose. A domain-specific GPT-4o judge test shows an order effect, so it is only a usefulness check. We release the code and data.
Chinese Translation
流畅的大型语言模型（LLM）解释可能不遵循来自结构化系统的证据。我们提出了VERITYGATE，一个针对声明的证据ID、实体、数字和主张类型的四门检查器。它检查一个固定的模式；并不验证散文中的每一个事实。在r=0和r=1的设置下，我们使用GPT-4o-mini、Llama-3.3-70B和Claude Sonnet 4.6测试了900个实例（450个有据-无据对）。在这个模式级合同下，修复之前，80.3%的mini主张和47.9%的Sonnet主张未通过检查。这些是验证器拒绝率，而不是散文幻觉率。一次修复过程将mini的主张存活率从19.7%提高到28.0%，将Sonnet的主张存活率从52.1%提高到54.3%。每个示例的验证主张数量变化为mini +0.14，Llama -0.71，Sonnet -0.47，因此存活率和输出量必须一起报告。第二次Sonnet检查没有明显的增益。在r=1时，Gate 4覆盖了mini、Llama和Sonnet未通过主张的97.0%、98.7%和100%。小规模的人类研究支持这些规则，但显示出模式检查与正确散文之间的差距。一个特定领域的GPT-4o评判测试显示出顺序效应，因此它只是一个有效性检查。我们发布了代码和数据。
cs.CL / 41 / 2610.00840
Contextual trajectory and incremental contextual displacement: Towards using LLMs to understand dynamic, utterance-specific meaning construction
上下文轨迹与增量上下文位移：利用大型语言模型理解动态的、特定于发话者的意义构建
Abstract
Transformer-based large language models (LLMs) such as RoBERTa represent text using contextual word embeddings (CWEs), which alter the embeddings associated with each token based on surrounding context. We construct token-wise incremental trajectories by repeatedly recomputing a token's CWE as successive words are added to a sentence, yielding a representation of how contextualized embeddings evolve as the utterance unfolds. We evaluate this approach using garden-path sentences as a test case with characteristic features. Token-wise trajectories reproduce known features of garden-path processing, including disruption around the critical region, and reliably distinguish garden-path sentences from matched disambiguated controls. We introduce several metrics for quantifying representational displacement across contextual increments and show that trajectory information can be highly predictive of sentence type. We find that ambiguity-related information is recoverable not only from the sentence-level CLS representation but also from ordinary vocabulary tokens, suggesting that utterance-level information is distributed across multiple representational scales. In exploratory analyses, we find qualitatively similar trajectory structures in other ambiguity- and misdirection-related linguistic phenomena. Together, these results establish token-wise incremental trajectories as a promising framework for studying utterance-specific meaning construction using LLMs.
Chinese Translation
基于变压器的大型语言模型（LLMs），如RoBERTa，通过上下文词嵌入（CWEs）表示文本，这些嵌入会根据周围的上下文改变与每个标记相关联的嵌入。我们通过在句子中逐步添加单词，反复重新计算标记的CWE，构建逐标记增量轨迹，从而获得上下文化嵌入如何随着发话的展开而演变的表示。我们使用具有特征特征的花园路径句子作为测试案例来评估这种方法。逐标记轨迹再现了花园路径处理的已知特征，包括在关键区域附近的干扰，并可靠地区分花园路径句子与匹配的消歧义控制句子。我们引入了几种量化上下文增量间表示位移的指标，并表明轨迹信息可以高度预测句子类型。我们发现，与歧义相关的信息不仅可以从句子级CLS表示中恢复，还可以从普通词汇标记中恢复，这表明发话者级信息分布在多个表示尺度上。在探索性分析中，我们发现其他与歧义和误导相关的语言现象中也存在定性相似的轨迹结构。这些结果共同确立了逐标记增量轨迹作为使用LLMs研究特定于发话者的意义构建的有前景框架。
cs.CL / 42 / 2610.00852
Child-Adapted Structured Phonological Representations for Interpretable Speech Sound Analysis
适应儿童的结构化音位表示用于可解释的语音声音分析
Abstract
Structured phonological representations provide an interpretable alternative to generic speech embeddings, but existing models are largely trained on adult speech. We adapt PhonoQ-2.0 to child speech using CHILDES-Aligned data and compare three alignment-supervision conditions (Adult, Adult+Child, and Child-only) across two initialization strategies (Adult PhonoQ and scratch). Generalization is evaluated against manual child-speech annotations. On 1,352 consonant targets from 58 typically developing children, child-speech adaptation improves voicing recognition across all supervision conditions, from 0.922 macro-F1 for Adult PhonoQ to 0.972--0.987 after adaptation. Manner is more sensitive to alignment supervision: Adult+Child MFA reaches 0.804 and 0.796, compared to approximately 0.70 under Adult MFA supervision. Place remains comparatively strong across systems (0.871--0.902), although per-class performance varies substantially. The velar-fronting contrast is preserved across all seven model variants. Longitudinal UltraPhonix analysis further reveals speaker-specific velar and post-alveolar changes that are largely preserved across models and broadly consistent with reported clinical progress.
Chinese Translation
结构化音位表示提供了一种可解释的替代方案，以取代通用的语音嵌入，但现有模型主要是在成人语音上训练的。我们使用与CHILDES对齐的数据将PhonoQ-2.0适应儿童语音，并比较三种对齐监督条件（成人、成人+儿童和仅儿童）在两种初始化策略（成人PhonoQ和从头开始）下的表现。通过手动儿童语音注释评估泛化能力。在58名典型发展儿童的1,352个辅音目标上，儿童语音适应在所有监督条件下提高了发声识别，从成人PhonoQ的0.922宏F1提高到适应后的0.972至0.987。发音方式对对齐监督更为敏感：成人+儿童的MFA达到0.804和0.796，而在成人MFA监督下约为0.70。尽管每个类别的表现差异显著，地点在各系统中仍然相对强劲（0.871至0.902）。在所有七个模型变体中，软腭前移对比得以保留。纵向UltraPhonix分析进一步揭示了说话者特定的软腭和后齿槽变化，这些变化在模型之间大体保持一致，并与报告的临床进展广泛一致。
cs.CL / 43 / 2610.00883
DeBERTa-ConPara: Attack-Aware and Deployment-Realistic Detection of AI-Generated Text
DeBERTa-ConPara：攻击感知与部署现实的AI生成文本检测
Abstract
Robust detection of AI-generated text under deployment conditions is challenging: distribution shifts across domains and generators, adversarial perturbations of the input surface, and the absence of target-domain labels for threshold calibration all degrade detectors that perform well in-domain. We present DeBERTa-ConPara, a deployment-oriented detector combining attack-aware Unicode preprocessing with a contextual transformer encoder trained over HC3 Plus, M4, MAGE and RAID. Our central finding is that preprocessing acts in opposite directions depending on where it is applied: normalising the training corpus deduplicates it, collapsing 35.4% of RAID rows into copies of their clean siblings and deleting the adversarial supervision, whereas normalising at inference is an effective defence. A factorial varying the two placements independently identifies raw training with normalised inference as the best configuration, reaching 99.61% AUROC, 99.01% TPR@5% FPR and 96.57% TPR@1% FPR on the official RAID hidden test, alongside 93.14% average balanced accuracy across HC3 Plus and MAGE under a fixed threshold. The gain is confined to two of twelve attack classes: homoglyph and zero-width-space insertion rise from 11.05% and 1.12% to 96.98%. The same signature reproduces in a zero-shot detector of different architecture, showing the effect belongs to the attacks rather than to our model. We additionally report two negative results: semantic-invariance augmentation through paraphrasing and supervised contrastive learning (ConPara) does not improve the best configuration, and the handcrafted feature-fusion branch is inert in distribution and harmful outside it.
Chinese Translation
在部署条件下，稳健检测AI生成文本面临挑战：跨领域和生成器的分布变化、输入表面的对抗扰动，以及缺乏目标领域标签进行阈值校准，均会降低在领域内表现良好的检测器的性能。我们提出了DeBERTa-ConPara，这是一种面向部署的检测器，结合了攻击感知的Unicode预处理和在HC3 Plus、M4、MAGE和RAID上训练的上下文变换编码器。我们的核心发现是，预处理的作用方向取决于其应用的位置：对训练语料进行归一化会导致其去重，将35.4%的RAID行合并为其干净同类的副本，并删除对抗监督，而在推理时进行归一化则是一种有效的防御。通过独立变化这两种位置的因子，发现原始训练与归一化推理的组合是最佳配置，在官方RAID隐藏测试中达到了99.61%的AUROC、99.01%的TPR@5% FPR和96.57%的TPR@1% FPR，同时在HC3 Plus和MAGE下以固定阈值获得93.14%的平均平衡准确率。增益仅限于十二个攻击类别中的两个：同形异义字（homoglyph）和零宽空格插入（zero-width-space insertion）从11.05%和1.12%上升至96.98%。相同的特征在不同架构的零样本检测器中重现，表明该效应属于攻击而非我们的模型。此外，我们还报告了两个负面结果：通过释义进行的语义不变性增强和监督对比学习（ConPara）并未改善最佳配置，而手工特征融合分支在分布内无效且在分布外有害。
cs.CL / 44 / 2610.00910
The Geometry of Contextual Relations: Language Models Address Facts by Order of Mention
上下文关系的几何：语言模型按提及顺序处理事实
Abstract
Human reasoning depends on how objects are related within propositions. \textit{How do relations organize the language representations of contextual contents?} We give an LLM a list of facts in its context (e.g., \emph{Alice eats an apple. Bob eats a pear.}) and measure how its hidden state changes when the question switches from what Alice eats to what Bob eats. Averaged over many lists, this change is a steering vector, which we call the \emph{ordinal vector}. It points to a fact by its \emph{order of mention}, the order in which the facts were stated in the context. We find that LLMs represent the fact a question asks about by its order of mention, not by the name the question contains. We state this as the \textit{ordinal addressing hypothesis}: each order of mention has a \emph{fact address} in the model's state, shared by all contexts, and a question moves the state to the fact address of the fact it asks about, while the context supplies what that fact says. Across Qwen, Gemma, and Llama, fact addresses are (1) \emph{ordered by mention}: query states are organized by the order of facts, not of names, even when one fact has multiple subjects; (2) \emph{steerable}: added to a question about the first fact of a new list, the ordinal vector makes the model answer with the second fact of that list; (3) \emph{low-rank}: they span a low-rank subspace in which the first-mentioned fact is the easiest to reach, surprisingly similar to human recall; and (4) \emph{emergent}: they are shared in late-middle layers, hold from 1.5B to 32B parameters, and form early in pretraining. Language models reach a stated fact by where it was mentioned, deepening our understanding of LLM reasoning.
Chinese Translation
人类推理依赖于对象在命题中的关系。我们提出了一个问题： extit{关系如何组织上下文内容的语言表征？} 我们给一个大型语言模型（LLM）提供一系列事实（例如， extit{爱丽丝吃苹果。鲍勃吃梨。}），并测量当问题从爱丽丝吃什么转变为鲍勃吃什么时，其隐状态如何变化。对多个列表进行平均后，这种变化形成一个引导向量，我们称之为 extit{序列向量}。它通过 extit{提及顺序}指向一个事实，即在上下文中陈述事实的顺序。我们发现，LLMs通过提及顺序而非问题中包含的名称来表示问题所询问的事实。我们将其表述为 extit{序列寻址假设}：每个提及顺序在模型的状态中都有一个 extit{事实地址}，该地址在所有上下文中共享，而一个问题将状态移动到其所询问的事实的事实地址，同时上下文提供该事实的内容。在Qwen、Gemma和Llama中，事实地址具有以下特征：(1) extit{按提及顺序排列}：查询状态按事实的顺序组织，而不是名称的顺序，即使一个事实有多个主体；(2) extit{可引导}：当添加到关于新列表第一个事实的问题时，序列向量使模型回答该列表的第二个事实；(3) extit{低秩}：它们在一个低秩子空间中展开，其中第一个提及的事实是最容易到达的，令人惊讶的是与人类回忆相似；(4) extit{涌现}：它们在中后层共享，适用于从15亿到320亿参数，并在预训练早期形成。语言模型通过提及的地点到达所述事实，深化了我们对LLM推理的理解。
cs.CL / 45 / 2610.00928
Efficient Task Adaptation in Large Language Models: A Survey of Weight-Based, Prompt-Based, and Embedding-Based Adaptations
大语言模型中的高效任务适应：基于权重、提示和嵌入的适应方法综述
Abstract
As large language models are increasingly deployed across diverse downstream tasks, efficient task adaptation has emerged as a central challenge. In response, a wide range of task adaptation methods have been proposed, spanning parameter-efficient fine-tuning, in-context learning, and embedding-injection approaches. However, these lines of work have largely evolved within individual paradigms, leaving their cross-paradigm relationships and trade-offs underexplored, especially for recently emerging embedding-based adaptations. This survey presents a unified framework that categorizes task adaptation methods by where and how task information is encoded: model weights, input prompts, or injected task embeddings. We provide a comprehensive taxonomy that integrates these paradigms, analyze their key strengths and limitations to explain how different adaptation paradigms have evolved, clarify relationships across paradigms, and highlight open problems for future research.
Chinese Translation
随着大语言模型在多样化下游任务中的广泛应用，高效的任务适应已成为一个核心挑战。为此，提出了多种任务适应方法，涵盖了参数高效微调、上下文学习和嵌入注入方法。然而，这些研究大多在各自的范式内发展，跨范式的关系和权衡尚未得到充分探讨，尤其是对于新兴的基于嵌入的适应方法。本文综述提出了一个统一框架，通过任务信息的编码方式和位置对任务适应方法进行分类：模型权重、输入提示或注入的任务嵌入。我们提供了一个综合的分类法，整合了这些范式，分析了它们的主要优缺点，以解释不同适应范式的演变，澄清范式之间的关系，并强调未来研究中的开放问题。
cs.CL / 46 / 2610.00940
ReHoPER: Receding-Horizon Planning for Enhanced Reasoning
ReHoPER：增强推理的递归视野规划
Abstract
We propose ReHoPER, an inference-only, zero-shot method that improves large language models' reasoning by generating and answering intermediate questions along multiple paths before the final answer. It iteratively plans a horizon of candidate intermediate questions, selects one to answer, and replans from the updated history. ReHoPER is task-agnostic, using the same generic instructions across datasets and models without labeled data or task-specific prompt design. Across multiple datasets, including iLLC, a new controlled benchmark for compositional reasoning, ReHoPER outperforms strong baselines, with the largest gains in the most compositional settings. Our implementation and the iLLC generator are publicly available to support future work.
Chinese Translation
我们提出了ReHoPER，这是一种仅基于推理的零样本方法，通过在最终答案之前沿多个路径生成和回答中间问题来提升大型语言模型的推理能力。它迭代地规划候选中间问题的视野，选择一个进行回答，并根据更新的历史进行重新规划。ReHoPER与任务无关，在不同数据集和模型中使用相同的通用指令，而无需标记数据或特定任务的提示设计。在多个数据集上，包括iLLC，一个新的组合推理控制基准，ReHoPER超越了强基线，在最具组合性的设置中获得了最大的提升。我们的实现及iLLC生成器已公开，以支持未来的研究工作。
cs.CL / 47 / 2610.00954
Beyond Leaderboards: Tokenomics of Agentic Small Language Model Ensembles
超越排行榜：代理小型语言模型集的代币经济学
Abstract
As large language models (LLMs) move from standalone assistants into agentic workflows, evaluation must extend beyond scalar leaderboard accuracy to account for operational reliability, cost, latency, and token efficiency. We use an agentic ensemble of small language models (SLMs) with an SLM-judge-mediated feedback loop as a case study for such beyond-leaderboard evaluation. On the 541-prompt IFEval benchmark, the best ensemble achieves 97.34% strict prompt accuracy, exceeding the strongest standalone LLM baseline, gpt-5.4, by 5.81 percentage points while operating in a lower-cost regime. We then analyze the tokenomics and operational behavior behind this gain, including cost per sample, token composition, useful-output goodput, feedback-loop recovery, latency decomposition, and performance across instruction categories and constraint counts. Our results show that agentic SLM ensembles can trade additional test-time tokens and orchestration overhead for improved instruction-following fidelity, motivating multi-dimensional evaluation protocols for future agentic AI systems.
Chinese Translation
随着大型语言模型（LLMs）从独立助手转向代理工作流，评估必须超越标量排行榜准确性，考虑操作可靠性、成本、延迟和代币效率。我们以一个小型语言模型（SLMs）的代理集和一个由SLM评估者介导的反馈循环为案例研究，进行这种超越排行榜的评估。在541个提示的IFEval基准测试中，最佳集成体实现了97.34%的严格提示准确率，超过了最强的独立LLM基线gpt-5.4，提升了5.81个百分点，同时在较低成本的环境下运行。随后，我们分析了这一提升背后的代币经济学和操作行为，包括每个样本的成本、代币组成、有用输出的良率、反馈循环恢复、延迟分解，以及在指令类别和约束数量上的性能。我们的结果表明，代理SLM集可以在提高指令遵循的忠实度方面，权衡额外的测试时间代币和协调开销，这为未来的代理人工智能系统提供了多维评估协议的动机。
cs.CL / 48 / 2610.00958
Role-aware Heuristic Episodic Attention for Conversational LLMs
面向角色的启发式情节注意力在对话大型语言模型中的应用
Abstract
Large language models often lose track of persistent instructions and relevant information as multi-turn conversations grow. We study this cumulative contextual decay through three related failure modes: attention pollution, dilution, and drift. We propose REA (Role-aware Heuristic Episodic Attention), a context-management framework that assigns different persistence and representation policies to instructions and episodic interactions. Instructional Memory retains identified global constraints in a dedicated prefix. Episodic Memory preserves user inputs and compresses model replies, while heuristic retrieval selects raw text, compressed representations, or omission for each historical turn. On Long-MT-Bench+, REA improves the judge score from 6.32 to 7.36 on a 10-point scale, a 16.5% relative gain over the Vanilla baseline, and reduces average latency by 2.91$\times$. Additional evaluations show aggregate gains on three backbones spanning 1.7B-7B parameters and on Chinese and English role-playing tasks. These results support role-aware context management as a practical approach to maintaining conversational continuity and instruction adherence.
Chinese Translation
大型语言模型在多轮对话中往往会随着对话的深入而失去对持续指令和相关信息的追踪。我们通过三种相关的失败模式研究这种累积的上下文衰减：注意力污染、稀释和漂移。我们提出了 REA（面向角色的启发式情节注意力），这是一种上下文管理框架，为指令和情节交互分配不同的持久性和表示策略。指令记忆在专用前缀中保留已识别的全局约束。情节记忆保留用户输入并压缩模型回复，而启发式检索则为每个历史回合选择原始文本、压缩表示或省略。在 Long-MT-Bench+ 上，REA 将评判分数从 6.32 提高到 7.36（满分 10 分），相较于基础模型实现了 16.5% 的相对提升，并将平均延迟减少了 2.91 倍。额外的评估显示，在涵盖 1.7B-7B 参数的三种基础模型以及中文和英文角色扮演任务上均取得了综合性提升。这些结果支持面向角色的上下文管理作为维持对话连续性和遵循指令的实用方法。
cs.CL / 49 / 2610.00969
A Citation-Grounded Benchmark for Trustworthy Earnings Call Transcript Analysis with Large Language Models
基于引用的可信财报电话会议记录分析基准测试与大型语言模型
Abstract
Large language models (LLMs) have been increasingly used for financial document analysis, including earnings call transcripts (ECTs). Beyond generating standalone claims, users increasingly prefer grounded analyses that pair claims with verifiable citations from source documents to enable independent validation. However, evaluating such analytical claims typically requires extensive expert annotation, which is costly and difficult to scale, and real-world financial analysis commonly involves long context-question-answer triplets, further increasing task complexity. To address these challenges and benchmark the current landscape of grounded analysis by LLMs, we propose a numeric evidence evaluation method that enables groundedness assessment without reliance on expert annotation. We also introduce an automated dataset construction pipeline and construct ECTs-100 from the top 100 constituents of the S&P 500 to support benchmark of both groundedness and correctness. In addition, we examine conscious incompetence, a practical failure mode in financial analysis in which LLMs must detect when available evidence is insufficient and refrain from producing unsupported hallucinations. Empirical results show that LLMs perform well in groundedness but face notable limitations in correctness, with informational insufficiency presenting an additional challenge.
Chinese Translation
大型语言模型（LLMs）在金融文档分析中得到了越来越多的应用，包括财报电话会议记录（ECTs）。用户不仅希望生成独立的陈述，更倾向于将陈述与源文档中的可验证引用相结合的基础分析，以便进行独立验证。然而，评估此类分析性陈述通常需要大量专家注释，这既昂贵又难以扩展，而现实世界的金融分析通常涉及长上下文-问题-答案三元组，进一步增加了任务的复杂性。为了解决这些挑战并基准当前LLMs的基础分析现状，我们提出了一种数值证据评估方法，能够在不依赖专家注释的情况下进行基础性评估。我们还引入了一个自动化数据集构建管道，并从标准普尔500指数的前100个成分股构建了ECTs-100，以支持基础性和正确性的基准测试。此外，我们考察了有意识的无能，这是金融分析中的一种实际失效模式，在这种模式中，LLMs必须检测可用证据是否不足，并避免产生不支持的幻觉。实证结果表明，LLMs在基础性方面表现良好，但在正确性方面面临显著限制，信息不足带来了额外的挑战。
cs.CL / 50 / 2610.00983
The Devil Is in the Reconstruction Loss Scale: Rethinking Optimization in LLM Quantization
重建损失尺度中的隐患：重新思考大语言模型量化中的优化
Abstract
Post-training quantization (PTQ) methods typically use sequential quantization that partitions a pre-trained LLM into a series of units (e.g., transformer blocks), with one unit quantized at each stage. State-of-the-art PTQ methods are predominantly learning-based, optimizing auxiliary quantization parameters (e.g., scaling factors, rotation matrices, clipping thresholds, and adapters) via gradient descent to minimize a reconstruction loss. A common practice is to use mean squared error (MSE) as the reconstruction loss function, yet its induced optimization behavior remains largely unexplored. In this work, we take a holistic view of sequential quantization and systematically investigate how optimization evolves from the first quantization stage to the last, aiming for a deep understanding of optimization in learning-based PTQ schemes. Through extensive empirical studies spanning representative learning-based PTQ methods, LLM families, model scales, architectures, quantization settings and various tasks, we consistently uncover Optimization Imbalance: reconstruction loss magnitudes vary dramatically across stages, accompanied by highly uneven gradient magnitudes and parameter updates under MSE. We term the cross-stage range of loss magnitudes the reconstruction loss scale, and reveal that MSE translates the unexpectedly large reconstruction loss scale into highly uneven gradient magnitudes, which in turn lead to uneven optimization strength across quantization stages. This finding suggests a general principle for improving learning-based PTQ: optimization strength across stages should be decoupled from the reconstruction loss scale. Theoretically, we show that root mean squared error (RMSE) variants defined at the sample, channel, token, and element levels naturally realize this principle through implicit gradient normalization, outperforming MSE significantly as a drop-in replacement.
Chinese Translation
后训练量化（PTQ）方法通常采用顺序量化，将预训练的大语言模型（LLM）划分为一系列单元（例如，变换器块），在每个阶段对一个单元进行量化。最先进的PTQ方法主要基于学习，通过梯度下降优化辅助量化参数（例如，缩放因子、旋转矩阵、裁剪阈值和适配器），以最小化重建损失。一个常见的做法是使用均方误差（MSE）作为重建损失函数，但其引发的优化行为仍然很大程度上未被探索。在本研究中，我们从整体上审视顺序量化，并系统地研究优化如何从第一个量化阶段演变到最后一个阶段，旨在深入理解基于学习的PTQ方案中的优化。通过广泛的实证研究，涵盖了具有代表性的基于学习的PTQ方法、LLM家族、模型规模、架构、量化设置和各种任务，我们一致发现了优化不平衡现象：重建损失的大小在各个阶段之间变化显著，伴随着在MSE下高度不均匀的梯度大小和参数更新。我们将跨阶段的损失大小范围称为重建损失尺度，并揭示MSE将意外较大的重建损失尺度转化为高度不均匀的梯度大小，这反过来导致量化阶段之间优化强度的不均匀性。这个发现提出了一个改善基于学习的PTQ的一般原则：各阶段之间的优化强度应与重建损失尺度解耦。理论上，我们证明了在样本、通道、标记和元素级别定义的均方根误差（RMSE）变体通过隐式梯度归一化自然实现了这一原则，作为替代品显著优于MSE。
cs.CL / 51 / 2610.00997
Distilling Directional Verification
方向性验证的蒸馏
Abstract
Knowledge distillation aims to transfer the factual knowledge of large language models to smaller models for efficient deployment. Yet a teacher may recall a relation in one direction while failing to generate the answer in the reverse direction. Distillation from its generated answers can therefore propagate this directional limitation to the student. The same teacher can nevertheless recognize such an answer by scoring the relation in the direction it knows. We introduce directional label distillation, in which frozen teachers score candidate answers in that known direction and the best-scoring candidate becomes the student's training target. On facts about parents and their children, known-direction scoring yields more accurate labels than scoring the requested direction, even after tuned corrections for name priors. With prior-corrected scores, the better direction depends on the facts rather than the template, and reverses on mined facts whose notable entity is the parent rather than the child. With the evaluated children's forward facts withheld, students trained on known-direction labels improve open-ended accuracy on their trained queries by 13 to 15 points over students trained on prior-corrected reverse labels. After generated answers are matched to a fixed name list by lexical similarity, students reproduce nearly all selected labels. Their accuracy largely follows label quality. The label advantage holds on unscreened queries and when candidates are retrieved without inserting correct answers. Our findings show that directional verification mitigates the transfer of errors from teacher-generated answers to students by providing more accurate training targets. Code is available at https://github.com/js-lee-AI/directional-verification.
Chinese Translation
知识蒸馏旨在将大型语言模型的事实知识转移到较小的模型中，以实现高效部署。然而，教师可能在一个方向上回忆起某种关系，而无法在反方向生成答案。因此，从其生成的答案进行蒸馏可能会将这种方向性限制传播给学生。尽管如此，同一教师仍然可以通过在其已知方向上对关系进行评分来识别这样的答案。我们提出了方向性标签蒸馏，其中冻结的教师在已知方向上对候选答案进行评分，得分最高的候选答案成为学生的训练目标。在关于父母及其子女的事实中，已知方向的评分比请求方向的评分产生更准确的标签，即使在对姓名先验进行了调优修正之后。使用经过先验修正的评分时，更好的方向取决于事实而非模板，并且在其显著实体为父母而非子女的挖掘事实上会反转。当评估的子女的正向事实被隐去时，基于已知方向标签训练的学生在其训练查询上的开放式准确率比基于先验修正的反向标签训练的学生提高了13到15个百分点。在通过词汇相似性将生成的答案与固定名称列表匹配后，学生几乎重现了所有选定的标签。他们的准确率在很大程度上遵循标签质量。标签优势在未筛选的查询中以及在检索候选者时未插入正确答案的情况下依然存在。我们的研究结果表明，方向性验证通过提供更准确的训练目标，减轻了教师生成的答案向学生转移错误的情况。代码可在 https://github.com/js-lee-AI/directional-verification 获取。
cs.CL / 52 / 2610.01016
Scaling and Distilling Text Embeddings for Better Diffusibility
扩展和提炼文本嵌入以提高扩散性
Abstract
Diffusion language models (DLMs) offer a promising alternative to autoregressive (AR) language generation. Recent advances in continuous DLMs, which apply latent diffusion to continuous text embeddings, raise a practical question: which embedding makes the best latent space, i.e., the most diffusible? To answer this, we search through different embeddings and find that scaling the embedding model to stronger ones within the same family (T5 to T5Gemma-1 to T5Gemma-2) greatly improves generative performance. But the raw T5Gemma-2 embeddings are still not optimal. They are so discriminative that even the embeddings of plausible alternative words are separated, which makes the generation vulnerable to imperfect sampling. Consequently, continuous diffusion often fails to reach any of them and ends up at an invalid embedding instead. To address this, we distill T5Gemma-2 into a student encoder that learns the teacher's decoded probabilities as soft labels. Learning from such soft labels makes the student pull the alternative embeddings closer while maintaining the encoding-decoding mechanism. The distilled embeddings form a more connected and diffusible latent space, improving over the vanilla T5Gemma-2 embeddings. As a result, our medium-sized DLM achieves Gen. PPL 17.8 (against real-text PPL 15.4) at real-text entropy on OpenWebText, outperforming GPT-2-M on Gen. PPL.
Chinese Translation
扩散语言模型（DLMs）为自回归（AR）语言生成提供了一种有前景的替代方案。最近在连续DLMs方面的进展，特别是将潜在扩散应用于连续文本嵌入，提出了一个实际问题：哪种嵌入构成最佳的潜在空间，即最具扩散性？为了解答这个问题，我们对不同的嵌入进行了探索，发现将嵌入模型扩展到同一家族中的更强模型（从T5到T5Gemma-1再到T5Gemma-2）显著提高了生成性能。然而，原始的T5Gemma-2嵌入仍然不是最优的。它们的区分性过强，以至于即使是合理的替代词的嵌入也被分开，这使得生成过程容易受到不完美采样的影响。因此，连续扩散往往无法达到任何一个有效的嵌入，最终停留在一个无效的嵌入上。为了解决这个问题，我们将T5Gemma-2提炼为一个学生编码器，该编码器学习教师的解码概率作为软标签。从这些软标签中学习使得学生能够将替代嵌入拉得更近，同时保持编码-解码机制。提炼后的嵌入形成了一个更为连通和可扩散的潜在空间，优于原始的T5Gemma-2嵌入。因此，我们的中型DLM在OpenWebText上的真实文本熵下达到了生成PPL 17.8（相较于真实文本PPL 15.4），在生成PPL上超越了GPT-2-M。
cs.CL / 53 / 2610.01026
It Takes Workflows to Evolve Better Workflows
工作流是演化更好工作流的关键
Abstract
Tackling complex real-world tasks can exceed the capabilities of a single large language model (LLM), motivating the use of multi-agent workflows that coordinate specialized agents to work together on these tasks. Recent methods train LLMs to construct better workflows from execution outcomes, but they optimize only the workflow generator, while the other agents that build or execute each workflow remain fixed even though every outcome depends on all of them. However, extending training beyond the generator is challenging: the agents are coupled, and a workflow's outcome is a single sparse score that cannot tell which agent causes a failure. We propose FloWright, which leverages the workflow as a harness to optimize workflows. By introducing a hierarchical, structure-aware reward paradigm, FloWright enables one role to self-evolve and two or more roles to co-evolve, with no additional models, labels, or executions. Considering the limitation that workflows are commonly trained and evaluated on data that a single agent can already handle, we further propose DataWright, an adaptive data hardening approach that converts existing datasets into workflow-level tasks with increased difficulty. Across document, slide, chart, code, math, and finance tasks, small open models trained with FloWright achieve improved performance by up to $+7.41\%$, with co-evolving ($+5.03\%$) more roles gaining more than optimizing one of them alone ($+2.83\%$). Our project page: https://xhguo7.github.io/FloWright/.
Chinese Translation
处理复杂的现实任务可能超出单一大型语言模型（LLM）的能力，这促使我们使用多代理工作流来协调专门的代理共同完成这些任务。最近的方法训练LLM从执行结果中构建更好的工作流，但它们仅优化工作流生成器，而构建或执行每个工作流的其他代理保持不变，尽管每个结果都依赖于它们所有人。然而，扩展训练到生成器之外是具有挑战性的：代理之间是耦合的，工作流的结果是一个单一的稀疏分数，无法判断是哪个代理导致了失败。我们提出了FloWright，它利用工作流作为优化工作流的工具。通过引入一种层次化的、结构感知的奖励范式，FloWright使得一个角色能够自我演化，而两个或更多角色能够共同演化，而无需额外的模型、标签或执行。考虑到工作流通常在单个代理已经能够处理的数据上进行训练和评估的限制，我们进一步提出了DataWright，这是一种自适应数据强化方法，将现有数据集转换为具有更高难度的工作流级任务。在文档、幻灯片、图表、代码、数学和金融任务中，使用FloWright训练的小型开放模型的性能提高了最高$+7.41 ext{ extperthousand}$，共同演化的角色（$+5.03 ext{ extperthousand}$）比单独优化其中一个角色（$+2.83 ext{ extperthousand}$）获得了更多的收益。我们的项目页面：https://xhguo7.github.io/FloWright/.
cs.CL / 54 / 2610.01027
LawCompass: Navigating from Legal QA to Multi-Agent Deep Research with Grounded Evidence
法舵：从法律问答到基于证据的多智能体深度研究的导航
Abstract
Recent advances in Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) have significantly democratized access to legal information. Nevertheless, most existing legal assistants remain confined to multi-turn conversational QA, failing to support complex legal tasks that require systematic evidence retrieval, multi-step reasoning, and report-level synthesis. In this paper, we present LawCompass, an evidence-grounded legal assistant that navigates the transition from standard Legal QA to multi-agent deep research. LawCompass provides three task-oriented functions: Legal QA, which delivers precise, evidence-backed answers to legal questions; Professional Retrieval, which enables structured exploration of statutes and judicial cases via query rewriting; and Deep Research, which employs a multi-agent workflow to decompose complex legal tasks and synthesize comprehensive research reports. Crucially, LawCompass maintains explicit citation links across all modules, empowering users to directly verify system outputs against original legal sources. Evaluation results demonstrate that LawCompass provides a practical and scalable paradigm for transforming conversational AI into trustworthy and evidence-grounded legal research assistance.
Chinese Translation
近年来，大型语言模型（LLMs）和检索增强生成（RAG）的进展显著降低了法律信息的获取门槛。然而，大多数现有的法律助手仍然局限于多轮对话问答，无法支持需要系统性证据检索、多步推理和报告级综合的复杂法律任务。本文提出了法舵（LawCompass），一个基于证据的法律助手，旨在实现从标准法律问答到多智能体深度研究的转变。法舵提供三种面向任务的功能：法律问答（Legal QA），提供准确的、基于证据的法律问题答案；专业检索（Professional Retrieval），通过查询重写实现对法规和司法案例的结构化探索；深度研究（Deep Research），采用多智能体工作流程分解复杂法律任务并综合全面的研究报告。至关重要的是，法舵在所有模块中保持明确的引用链接，使用户能够直接将系统输出与原始法律来源进行验证。评估结果表明，法舵为将对话式人工智能转变为可信赖且基于证据的法律研究助手提供了一个实用且可扩展的范式。
cs.CL / 55 / 2610.01046
Sentence Specificity Scores for Collaborative Technical Documentation: A Domain-Transfer Study
协作技术文档的句子特异性评分：一个领域转移研究
Abstract
Collaboration depends on shared context, and technical documentation is one way that context persists across people and AI teammates. Specificity, the amount and exactness of detail expressed in language, shapes what information documentation captures and how precisely that information is communicated. This work audits sentence-specificity scoring artifacts on technical documentation and tests whether scores applied only after generation help choose among fixed LLM-generated revisions. Across Wikipedia and three technical-documentation corpora, the fixed general-domain predictor SpeciTeller and the pinned post-publication author-repository implementation of Ko et al.'s target-adapted predictor produce different corpus orders and same-sentence rank agreement from -0.066 to 0.510. Strict filtering and token-length adjustment change these patterns without reconciling them. In the Gemma set, SpeciTeller ranking raises direction-valid selection from 71.7% to 83.3% (+11.7 points; 95% source-case bootstrap interval +1.7 to +21.7); in the GPT-OSS-120B set, SpeciTeller ranking raises direction-valid selection from 51.7% to 56.7% (+5.0 points; 95% source-case bootstrap interval -6.7 to +16.7), and every primary single-score GPT-OSS-120B interval includes zero. These findings tie score interpretation and decision value to the predictor and candidate set.
Chinese Translation
协作依赖于共享的上下文，而技术文档是这种上下文在人员与人工智能团队成员之间持续存在的一种方式。特异性，即语言中表达的细节的数量和准确性，塑造了文档捕获的信息以及这些信息被传达的精确程度。本研究审计了技术文档中的句子特异性评分工件，并测试了生成后仅应用评分是否有助于在固定的LLM生成修订中进行选择。在维基百科和三个技术文档语料库中，固定的通用领域预测器SpeciTeller与Ko等人目标适应预测器的发布后作者存储库实现产生了不同的语料库顺序和相同句子的排名一致性，从-0.066到0.510。严格的过滤和标记长度调整改变了这些模式，但未能调和它们。在Gemma数据集中，SpeciTeller排名使方向有效选择从71.7%提高到83.3%（+11.7个百分点；95%源案例自助区间+1.7到+21.7）；在GPT-OSS-120B数据集中，SpeciTeller排名使方向有效选择从51.7%提高到56.7%（+5.0个百分点；95%源案例自助区间-6.7到+16.7），且每个主要单评分的GPT-OSS-120B区间均包含零。这些发现将评分解释和决策价值与预测器和候选集联系起来。
cs.CL / 56 / 2610.01054
Capturing In-Context Learning Dynamics with Task Operators
通过任务算子捕捉上下文学习动态
Abstract
In-context learning (ICL) enables language models to perform new tasks from demonstrations without weight updates. However, every ICL inference requires processing the full set of examples, resulting in inefficient deployments, and how ICL works mechanistically is not fully understood. Prior work compresses ICL into fixed activation vectors extracted from specific layers or positions, but these input-independent interventions fail on complex tasks where the output depends on fine-grained interactions with the input. By analyzing the ICL forward pass, we show that each attention head's output is an affine transformation of its context-masked counterpart, and that the parameters of this transformation are empirically stable across samples for a given task. Building on this, we introduce Task Operator (TO), which replays this transformation as an analytically derived update to the attention output projection. Across lexical, algorithmic, and reasoning tasks, TO achieves the best overall performance among prior methods and substantially narrows the gap between zero-shot inference and ICL. We further show that the extracted knowledge concentrates in a task-specific sparse circuit across layers and positions, and that averaging operators from disjoint demonstration batches enables effective many-shot scaling without expanding the context window. Our code is available at https://github.com/gzxiong/task_operator.
Chinese Translation
上下文学习（In-context Learning, ICL）使语言模型能够在不更新权重的情况下，通过示例执行新任务。然而，每次 ICL 推理都需要处理完整的示例集，这导致了效率低下的部署，并且 ICL 的机制尚未完全理解。以往的研究将 ICL 压缩为从特定层或位置提取的固定激活向量，但这些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失败。通过分析 ICL 的前向传播，我们展示了每个注意力头的输出是其上下文掩蔽对应物的仿射变换，并且这种变换的参数在给定任务的样本间经验上是稳定的。在此基础上，我们引入了任务算子（Task Operator, TO），它将这种变换重放为对注意力输出投影的解析衍生更新。在词汇、算法和推理任务中，TO 在以往方法中实现了最佳的整体性能，并显著缩小了零-shot 推理与 ICL 之间的差距。我们进一步表明，提取的知识在层和位置上集中于任务特定的稀疏电路，并且来自不相交示例批次的算子平均化使得在不扩展上下文窗口的情况下实现有效的多-shot 扩展。我们的代码可在 https://github.com/gzxiong/task_operator 获取。
cs.CL / 57 / 2610.01064
JoinGR: Learning to Traverse Join Graphs for Table Retrieval
JoinGR：学习遍历连接图以进行表格检索
Abstract
Retrieving the right tables is a prerequisite for Text-to-SQL over realistic databases. Dense table retrievers rank schema elements independently, but this ignores a key source of evidence: some required tables are not mentioned in the question and become identifiable only through their join relationships to already relevant tables. We introduce JOINGR, a join-aware table retrieval method that treats the database join graph as the retrieval space. Columns are represented as graph nodes, while intra-table and foreign-key relationships are represented as typed edges. Given a question, JOINGR selects semantically similar anchor tables, traverses join edges with a query-conditioned scorer, and aggregates the resulting edge deposits into table scores. The scorer is a lightweight MLP on top of frozen query, node, and edge embeddings, trained with a pairwise margin loss over gold tables. On BIRD and Spider datasets, JOINGR is competitive with the strongest retrieval baselines. On BEAVER, a challenging enterprise benchmark with multi-hop table requirements, JOINGR substantially improves recall over dense retrieval and re-ranking baselines. Cross-domain experiments show that the learned scorer transfers across benchmarks, indicating that the method captures reusable joingraph traversal behavior.
Chinese Translation
在现实数据库上进行文本到SQL的操作，检索正确的表格是前提。稠密表格检索器独立地对模式元素进行排名，但这忽视了一个关键证据来源：某些所需的表格在问题中并未提及，仅通过与已相关表格的连接关系才能识别。我们提出了JOINGR，一种关注连接的表格检索方法，将数据库连接图视为检索空间。列被表示为图节点，而表内关系和外键关系则被表示为类型化边。给定一个问题，JOINGR选择语义上相似的锚表，利用查询条件评分器遍历连接边，并将结果边的存款聚合成表格分数。评分器是一个轻量级的多层感知机（MLP），基于冻结的查询、节点和边嵌入，使用金标准表格的成对边际损失进行训练。在BIRD和Spider数据集上，JOINGR与最强的检索基线具有竞争力。在BEAVER上，这是一个具有多跳表格需求的挑战性企业基准，JOINGR在召回率上显著优于稠密检索和重排序基线。跨领域实验表明，学习到的评分器在不同基准之间具有迁移能力，表明该方法捕捉到了可重用的连接图遍历行为。
cs.CL / 58 / 2610.01066
Probe with Participation Trophies: Random-Reward RL as a Probe of LLM Capability
参与奖的探测：随机奖励强化学习作为大语言模型能力的探测工具
Abstract
We connect the spurious-reward paradox to a model's reachability and propose random-reward reinforcement learning (RL) as a useful tool for the probing enterprise, addressing a decade-long debate over what probing performance actually reveals about a model. There are two prevailing explanations for the surprising finding that even random rewards can improve the performance of large language models (LLMs): one attributes the gains to particular mechanisms within RL training; the other to data contamination. Our results motivate a different view: spurious-reward RL can probe a model's reachability, or what further training can attain from its current state under specified constraints, beyond what is reflected in its current performance. Two OLMo checkpoints with the same accuracy on synthetic arithmetic (3.5%), for example, reach 8.5% and 55% in their best runs under the same correctness-rewarded RL. Examining OLMo checkpoints across pre-training and mid-training reveals three distinct regimes of training response: early on, RL produces little improvement even when correct answers are rewarded; later in pre-training, rewarding correct answers becomes effective while random rewards remain weak; and, upon entering mid-training, even random rewards can produce large gains. A similar ordering appears in a number-masked supervised fine-tuning (SFT) analysis of these checkpoints, suggesting that the pattern is not specific to a particular RL mechanism. Moreover, RL with random rewards offers a distinctive perspective on what training can make an LLM do, since its reward signal supplies no information about which answers are correct. By asking what training can attain without correctness feedback, it addresses the label-leakage side of a central problem in decodability-based probing: whether a successful probe reveals the model's capabilities or learns the task itself.
Chinese Translation
我们将虚假奖励悖论与模型的可达性联系起来，并提出随机奖励强化学习（RL）作为探测工作的有用工具，解决了关于探测性能究竟揭示了模型什么的十年争论。对于即使是随机奖励也能提升大型语言模型（LLMs）性能的惊人发现，目前有两种主要解释：一种将提升归因于RL训练中的特定机制；另一种则归因于数据污染。我们的结果激励了一种不同的观点：虚假奖励RL可以探测模型的可达性，或者在特定约束下，从当前状态进一步训练能够达到的结果，超出当前性能所反映的内容。例如，两个在合成算术（3.5%）上具有相同准确性的OLMo检查点，在相同的正确性奖励RL下，最佳运行分别达到8.5%和55%。对OLMo检查点在预训练和中期训练中的表现进行分析揭示了三种不同的训练响应模式：在早期，即使奖励正确答案，RL也几乎没有改善；在预训练的后期，奖励正确答案变得有效，而随机奖励仍然较弱；而在进入中期训练后，即使是随机奖励也能产生显著的提升。在这些检查点的数字掩码监督微调（SFT）分析中也出现了类似的顺序，表明这一模式并不特定于某一特定的RL机制。此外，带有随机奖励的RL为训练能够使LLM做什么提供了独特的视角，因为其奖励信号并未提供关于哪些答案是正确的信息。通过询问在没有正确性反馈的情况下训练能够达到什么，它解决了基于可解性探测的一个核心问题的标签泄露方面：成功的探测是否揭示了模型的能力，还是学习了任务本身。
cs.CL / 59 / 2610.01082
Precision over Scale: A Polish-Silesian Benchmark and a Translation System Outperforming Open-Source and Commercial Models
精确性与规模：一个波兰-西里西亚基准和一个超越开源及商业模型的翻译系统
Abstract
Dialectal machine translation remains challenging due to limited data and strong linguistic variation not captured by standard benchmarks, which often assume standardized and well-edited text. We study Polish-Silesian MT using neural and rule-based systems, evaluating on SiLTT - a new Pol-Szl testset, alongside established BOUQuET and FLORES benchmarks. Results show our rule-based system is consistently strongest on SiLTT and BOUQuET datasets and that TranslateGemma fine-tuned on a curated dataset improves over strong neural baselines but does not surpass the rule-based system in dialectal settings. We release SiLTT and our best neural model to support further research.
Chinese Translation
方言机器翻译仍然面临挑战，因为数据有限且强烈的语言变异未被标准基准捕捉，这些基准通常假设文本是标准化且经过良好编辑的。我们研究了波兰-西里西亚的机器翻译，使用神经网络和基于规则的系统，并在SiLTT——一个新的波兰-西里西亚测试集上进行评估，同时参考了已建立的BOUQuET和FLORES基准。结果表明，我们的基于规则的系统在SiLTT和BOUQuET数据集上始终表现最佳，而在经过精心策划的数据集上微调的TranslateGemma在强大的神经网络基线之上有所提升，但在方言环境中未能超越基于规则的系统。我们发布了SiLTT和我们最佳的神经模型，以支持进一步的研究。
cs.CL / 60 / 2610.01108
AgSpec: Pushing the Limits of Retrieval-Based Speculative Decoding in Coding Agent Pipelines
AgSpec：推动基于检索的推测解码在编码代理管道中的极限
Abstract
Retrieval-based speculative decoding (SD) drafts tokens by copying continuations from existing text, which suits coding agents that repeatedly reproduce code, logs, and earlier attempts. Yet existing methods fall short in agent pipelines: much of the reusable text is missing from their corpora or stored in a form that differs from what the agent emits, and their draft lengths ignore that accept length varies across agents and drifts over turns. We present AgSpec, a framework that supplies the corpus and draft-length policies that existing retrieval engines lack in coding-agent pipelines. AgSpec retrieves from session, workspace, and global corpora, retaining the ongoing session trajectory and indexing opened files in the agent's emission format. It bounds each agent's draft length with an offline-profiled cap and adapts the length online from verification feedback. On two repository-level multi-agent coding benchmarks, AgSpec outperforms five retrieval-based drafters and EAGLE-3 in most evaluated settings, raising generation throughput over autoregressive decoding up to 4.37$\times$ at batch size 1 and 4.76$\times$ at batch size 16. AgSpec also remains effective on benchmarks without a repository or a multi-agent pipeline, showing that its gains generalize to coding agents broadly.
Chinese Translation
基于检索的推测解码（SD）通过复制现有文本的延续来草拟标记，这适合于重复生成代码、日志和早期尝试的编码代理。然而，现有方法在代理管道中存在不足：可重用文本的许多部分缺失于其语料库中，或以与代理输出形式不同的方式存储，并且它们的草拟长度忽略了接受长度在不同代理之间的变化以及在回合中的漂移。我们提出了AgSpec，一个为编码代理管道提供现有检索引擎所缺乏的语料库和草拟长度策略的框架。AgSpec从会话、工作区和全局语料库中检索，保留正在进行的会话轨迹，并以代理的输出格式索引打开的文件。它通过离线配置的上限限制每个代理的草拟长度，并根据验证反馈在线调整长度。在两个基于存储库的多代理编码基准测试中，AgSpec在大多数评估设置中超越了五种基于检索的草拟器和EAGLE-3，将生成吞吐量在批量大小为1时提高至4.37倍，在批量大小为16时提高至4.76倍。AgSpec在没有存储库或多代理管道的基准测试中也表现出色，显示出其收益在广泛的编码代理中具有普遍性。
cs.CL / 61 / 2610.01118
Madeleine: Learning Involuntary Recall for Conversational Memory from Simulated Lives
Madeleine：从模拟生活中学习对话记忆的非自愿回忆
Abstract
A long-term conversational assistant must recall the right memory at the right moment, yet the memory that matters most is often not similar to what the user says now. Current systems recover such associations by letting an LLM reason at write or read time, at a cost of hundreds to over a thousand LLM calls per memory bank and up to several thousand context tokens per query. We argue that association is a learnable relevance: the pointwise mutual information of memories under how human lives unfold. We introduce Madeleine, which learns amortized association: offline, an LLM life simulator writes simulated lives, whose cue-trigger pairs teach a query encoder a residual association on top of frozen similarity; online, it calls no LLM and plugs into any vector memory by replacing only the query encoder. On LoCoMo-Plus under the official protocol, Madeleine (I) reaches 66.6 when plugged into HyperMem, the highest among all systems evaluated under this protocol; (II) used alone, reaches the score of HyperMem as released (52.4 vs. 52.9) with zero LLM calls and about 1/21 of its answer context; and (III) lifts T-Mem by 26.2 points, significantly outperforms the same untrained backbone inside both systems, and leaves ordinary QA intact on the 4B backbone.
Chinese Translation
一个长期的对话助手必须在合适的时刻回忆起正确的记忆，然而，最重要的记忆往往与用户当前所说的内容并不相似。当前系统通过让大型语言模型（LLM）在写入或读取时进行推理来恢复这种关联，这导致每个记忆库需要数百到超过一千次的LLM调用，以及每个查询需要数千个上下文标记。我们认为，关联是一种可学习的相关性：记忆在人的生活展开方式下的点对点互信息。我们介绍了Madeleine，它学习了摊销关联：离线时，LLM生活模拟器编写模拟生活，其提示-触发对教会查询编码器在冻结相似性基础上进行残余关联；在线时，它不调用LLM，只需通过替换查询编码器即可插入任何向量记忆。在官方协议下的LoCoMo-Plus上，Madeleine (I) 在插入HyperMem时达到66.6，是所有在该协议下评估的系统中最高的；(II) 单独使用时，达到发布的HyperMem得分（52.4对比52.9），且没有LLM调用，回答上下文约为其的1/21；(III) 提升T-Mem 26.2分，显著超越了两个系统内部相同的未训练骨干，并在4B骨干上保持普通问答的完整性。
cs.CL / 62 / 2610.01127
Counting and Min-Cost Encoding for Tokenization in Large Language Models
大型语言模型中的标记化计数与最小成本编码
Abstract
Mainstream large language models rely on a tokenizer to encode text into a token sequence. Different tokenizers may yield token sequences of substantially different lengths for the same text. With a fixed model architecture, shorter token sequences correspond to lower inference time. We propose a tokenizer training approach named Counting and Filtering (CNF) and a text encoding algorithm called Min-Cost Encoding (MCE). MCE defines a cost function over a text segment, and determines the best segmentation by globally minimizing the overall segmentation cost. CNF builds a raw vocabulary by directly counting valid substrings, and then constructs the final vocabulary through a filtering step based on actual token usage when segmenting the training corpus with MCE. The CNF-MCE conbination offers several advantages over BPE, including higher token efficiency, greater scalability, and lower dependency. Across six text categories and two vocabulary-size groups, CNF-MCE consistently achieves better compression than the evaluated BPE tokenizers. With a 250K vocabulary, CNF-MCE increases compression rate by 26% and 30% on English web text over the o200k_base and qwen250k tokenizers. Experiments scaling the vocabulary to 1M entries on English web text demonstrate sustained improvements over BPE, with a token efficiency improvement of over 60% and vocabulary utilization rising from 52.9% to 96.9%. The MCE algorithm does not depend on a merge list (as in BPE) or token probability (as in UnigramLM), making it applicable to a wide range of vocabularies, including those built from BPE, UnigramLM, CNF, and others. Language models trained from scratch at the 1.8B and 8B scales achieve comparable average performance to models using the BPE tokenizers across 11 benchmarks. These results demonstrate that CNF-MCE can improve token efficiency significantly while maintaining competitive downstream performance.
Chinese Translation
主流的大型语言模型依赖于标记器将文本编码为标记序列。不同的标记器可能会对相同文本生成长度差异显著的标记序列。在固定的模型架构下，较短的标记序列对应于更低的推理时间。我们提出了一种名为计数与过滤（Counting and Filtering, CNF）的标记器训练方法，以及一种称为最小成本编码（Min-Cost Encoding, MCE）的文本编码算法。MCE在文本片段上定义了一个成本函数，并通过全局最小化整体分割成本来确定最佳分割。CNF通过直接计数有效子字符串构建原始词汇表，然后通过基于实际标记使用情况的过滤步骤构建最终词汇表，后者是在使用MCE对训练语料库进行分割时确定的。CNF-MCE组合相比BPE提供了多个优势，包括更高的标记效率、更大的可扩展性和更低的依赖性。在六个文本类别和两个词汇规模组中，CNF-MCE始终实现了比评估的BPE标记器更好的压缩效果。在250K词汇表下，CNF-MCE在o200k_base和qwen250k标记器上对英语网页文本的压缩率分别提高了26%和30%。在英语网页文本上将词汇表扩展到1M条目的实验表明，CNF-MCE在BPE上持续改进，标记效率提高超过60%，词汇利用率从52.9%上升至96.9%。MCE算法不依赖于合并列表（如BPE）或标记概率（如UnigramLM），使其适用于广泛的词汇表，包括从BPE、UnigramLM、CNF等构建的词汇表。从零开始训练的1.8B和8B规模的语言模型在11个基准测试中实现了与使用BPE标记器的模型相当的平均性能。这些结果表明，CNF-MCE能够显著提高标记效率，同时保持竞争力的下游性能。
cs.CL / 63 / 2610.01150
BanglaDial-Abuse: A Corpus-Grounded Dataset for Regional Dialect Identification in Abusive Bangla Text
BanglaDial-Abuse：用于识别虐待性孟加拉文本区域方言的基于语料库的数据集
Abstract
Regional linguistic variation remains an important challenge for Bangla natural language processing, particularly in informal and non-standard text. This paper introduces BanglaDial-Abuse, a balanced Bengali-script dataset developed for regional dialect identification in abusive and hostile Bangla text. The dataset contains 1,000 sentences distributed equally across four linguistic varieties: Standard Bangla, Chattagram, Sylhet, and Barishal, with 250 samples per class. The resource was constructed using a corpus-grounded synthetic procedure incorporating regional variation in pronouns, possessive forms, verb morphology, negation, interrogative structures, postpositions, vocabulary, and Bengali-script spelling conventions while preserving the underlying hostile or abusive meaning. Descriptive analysis shows broadly comparable sentence-length distributions but partially distinct lexical spaces across the four classes. Pairwise Jaccard vocabulary similarity ranges from 0.37 to 0.56. The primary task is four-class regional dialect identification rather than binary abusive-text detection. The dataset is publicly available through Zenodo under a Creative Commons Attribution 4.0 license. The current version is intended as a research and prototyping corpus rather than a native-speaker-validated gold-standard linguistic resource. Keywords: Bangla, Bengali, dialect identification, regional dialect, abusive language, low-resource NLP, Chattagram, Sylhet, Barishal, dataset
Chinese Translation
区域语言变异仍然是孟加拉自然语言处理中的一个重要挑战，特别是在非正式和非标准文本中。本文介绍了BanglaDial-Abuse，这是一个为识别虐待性和敌对孟加拉文本中的区域方言而开发的平衡孟加拉文数据集。该数据集包含1,000个句子，均匀分布在四种语言变体中：标准孟加拉语、恰特拉姆、希尔赫特和巴里萨尔，每个类别有250个样本。该资源采用基于语料库的合成程序构建，结合了代词、所有格形式、动词形态、否定、疑问结构、后置词、词汇和孟加拉文拼写规范中的区域变异，同时保留了潜在的敌对或虐待性含义。描述性分析显示四个类别的句子长度分布大致相当，但词汇空间部分不同。成对的Jaccard词汇相似度范围从0.37到0.56。主要任务是四类区域方言识别，而非二元虐待文本检测。该数据集通过Zenodo以知识共享署名4.0许可证公开提供。当前版本旨在作为研究和原型开发语料库，而非经过母语者验证的标准语言资源。关键词：孟加拉语、孟加拉、方言识别、区域方言、虐待性语言、低资源自然语言处理、恰特拉姆、希尔赫特、巴里萨尔、数据集
cs.CL / 64 / 2610.01161
My FAULT: Self-Diagnosis as Credit Assignment in Self-Evolving Agentic Reinforcement Learning
我的故障：自我诊断作为自我进化代理强化学习中的信用分配
Abstract
Agentic reinforcement learning (RL) has emerged as a powerful approach for training large language model agents on multi-step tasks, yet reliance on terminal outcome rewards creates two credit-assignment problems, particularly in long-horizon tasks. First, same-outcome rollout groups provide no learning signal from terminal rewards. Second, terminal rewards provide only trajectory-wide feedback, making it difficult to identify which decisions caused a failure. Recent work supplements terminal rewards with finer-grained information from trajectory analysis, such as natural-language reflections on intermediate decisions and errors. However, natural-language diagnoses are difficult to use directly for credit assignment: their error claims may be unreliable, and they do not quantify how much each error should affect learning. We propose Self-Diagnosis-guided Terminal Credit Redistribution (FAULT), which turns diagnosed errors into explicit step-level credit anchored by terminal outcomes. FAULT checks diagnostic evidence and learns relative error costs from task outcomes. During training, the policy and self-diagnoser co-evolve, while error costs are updated online from recent outcomes. On ALFWorld, FAULT recovers learning signals from same-outcome groups, reaching 95% signal coverage versus 41% for GRPO and 72% for GiGPO, while better localizing credit to specific error steps. Across two model scales, FAULT delivers strong. improvements on the long-horizon ALFWorld and WebShop tasks while remaining competitive on short-horizon Search-based QA.
Chinese Translation
代理强化学习（RL）已成为训练大型语言模型代理以应对多步骤任务的强大方法，但对终端结果奖励的依赖产生了两个信用分配问题，特别是在长时间跨度的任务中。首先，相同结果的回滚组未能提供来自终端奖励的学习信号。其次，终端奖励仅提供轨迹范围的反馈，使得识别哪些决策导致失败变得困难。最近的研究通过轨迹分析补充终端奖励，提供更细粒度的信息，例如对中间决策和错误的自然语言反思。然而，自然语言诊断在信用分配中难以直接使用：它们的错误声明可能不可靠，并且无法量化每个错误对学习的影响程度。我们提出了自我诊断引导的终端信用再分配（FAULT），将诊断出的错误转化为以终端结果为锚的明确步骤级信用。FAULT检查诊断证据并从任务结果中学习相对错误成本。在训练过程中，策略和自我诊断器共同进化，同时错误成本根据最近的结果在线更新。在ALFWorld上，FAULT从相同结果组中恢复学习信号，达到95%的信号覆盖率，而GRPO为41%，GiGPO为72%，同时更好地将信用定位到特定的错误步骤。在两个模型规模上，FAULT在长时间跨度的ALFWorld和WebShop任务中显著提升表现，同时在短时间跨度的基于搜索的问答任务中保持竞争力。
cs.CL / 65 / 2610.01170
HeadEdit: Calibrating Language Model Behavior Through the Frozen Unembedding Matrix
HeadEdit：通过冻结的去嵌入矩阵校准语言模型行为
Abstract
Alignment does not eliminate behavioral errors in language models. Models may still refuse benign requests, call unnecessary tools, or yield to false user claims. Current methods mitigate such errors as a computation problem, and rarely explore if the desired behavior is already encoded in the model's representation. Motivated by the observation that behavior-relevant information remains linearly decodable from the final hidden state even when the resulting logits produce the undesired behavior, we introduce HeadEdit, a gradient-free method that calibrates model behavior through the unembedding matrix. HeadEdit extracts a low-rank behavioral subspace from paired completions and uses each prompt's coordinates within it to generate a vocabulary-wide correction, thereby implementing implicitly adaptive steering without manually specified target tokens or parameter updates. HeadEdit improves all nine experimental settings across three tasks and three model families, with negligible inference overhead and no systematic loss of general capabilities. It also reveals a connection to gradient-based alignment. HeadEdit's low-dimensional representation partly predicts how preference tuning changes output logits on unseen prompts. The subspace learned from the model can also be reused after tuning, improving performance without re-extracting or retuning. These results show that HeadEdit provides a practical, lightweight, and interpretable way to calibrate model behavior through the unembedding matrix.
Chinese Translation
对齐并不能消除语言模型中的行为错误。模型可能仍会拒绝无害的请求、调用不必要的工具，或屈服于虚假的用户声明。目前的方法将此类错误视为计算问题，且很少探讨所需的行为是否已经编码在模型的表示中。基于观察到即使生成的 logits 产生不期望的行为，行为相关信息仍然可以从最终隐藏状态线性解码，我们提出了 HeadEdit，这是一种无梯度的方法，通过去嵌入矩阵校准模型行为。HeadEdit 从配对的完成中提取低秩行为子空间，并利用每个提示在其中的坐标生成全词汇的修正，从而实现隐式自适应引导，而无需手动指定目标标记或参数更新。HeadEdit 在三个任务和三个模型系列的九个实验设置中均有所改善，推理开销微乎其微，且没有系统性地损失一般能力。它还揭示了与基于梯度的对齐之间的联系。HeadEdit 的低维表示在一定程度上预测了偏好调优如何改变未见提示的输出 logits。模型学习到的子空间在调优后也可以重复使用，提高性能而无需重新提取或重新调优。这些结果表明，HeadEdit 提供了一种通过去嵌入矩阵校准模型行为的实用、轻量且可解释的方法。
cs.CL / 66 / 2610.01177
Temporally-Resolved Token Attribution Reveals the Generation Dynamics of Diffusion Language Models
时间分辨的标记归因揭示了扩散语言模型的生成动态
Abstract
This work presents Diffusion Layer Integrated Gradients (DLIG), a token attribution method for diffusion language models (DLMs) that extends Integrated Gradients (IG~\cite{sundararajan2017axiomatic}) to arbitrary layers and denoising steps. DLIG attributes a DLM's progressive commitment to a self-generated or fixed completion for an input prompt. We establish direct correspondences between DLIG and the IG axioms of completeness, implementation invariance, linearity, and symmetry preservation. As a lightweight complement to interventional analysis, DLIG provides an inexpensive first check of mechanistic hypotheses across the denoising trajectory. We demonstrate this on word-sense disambiguation, multi-hop graph reasoning, and sentence infilling, revealing how DLMs draw on inputs across positions, layers, and denoising steps.
Chinese Translation
本研究提出了扩散层集成梯度（Diffusion Layer Integrated Gradients, DLIG），这是一种针对扩散语言模型（Diffusion Language Models, DLMs）的标记归因方法，旨在将集成梯度（Integrated Gradients, IG~ extcite{sundararajan2017axiomatic}）扩展到任意层和去噪步骤。DLIG 归因于 DLM 对输入提示的自生成或固定完成的逐步承诺。我们建立了 DLIG 与 IG 完整性、实现不变性、线性和对称性保持等公理之间的直接对应关系。作为干预分析的轻量级补充，DLIG 提供了一种低成本的初步检查机制假设的方法，适用于去噪轨迹。我们在词义消歧、多跳图推理和句子填充等任务上演示了这一方法，揭示了 DLM 如何在不同位置、层和去噪步骤中利用输入信息。
cs.CL / 67 / 2610.01218
AGO AI Quality Gate: Evidence-First Release Decisions for Retrieval-Augmented Generation
AGO AI 质量门：基于证据的检索增强生成发布决策
Abstract
Enterprises adopting retrieval-augmented generation (RAG) face a recurring operational decision: promote, revise, or block a system version. The evidence is incomplete and the metrics come from fallible LLM judges. We report on AGO AI Quality Gate (AGO), an evidence-first quality-gate framework deployed in industrial RAG assessment engagements. AGO integrates four key components: a four-state decision model that treats missing data and judge errors as explicit outcomes; layered scoring combining deterministic checks, local guardrails, and structured LLM evaluation; a stratified beta-binomial gate that quantifies regression risk probabilistically; and a mandatory meta-evaluation protocol to validate the LLM judge before it influences decisions. Since engagement data is proprietary, we evaluate the judge layer on RAGBench, a public benchmark of 100k annotated RAG traces across 12 datasets. On identical stratified test samples (N=1200 per judge), a low-cost judge (gpt-4.1-nano) detects non-adherent answers barely above chance (AUROC 0.603 [0.570, 0.634]), despite producing flawless protocol output, while gpt-4o reaches 0.783 [0.756, 0.807] -- yet its per-domain performance still ranges from 0.62 to 0.88. A fixed-seed gate study spanning regression, no change, and improvement quantifies unsafe promotion, false-alarm cost, and improvement throughput. Under regression, the decision-grade profile reduces unsafe promotion to 22.2%-35.1%, against 29.3%-41.8% for a naive gate. These results support the design choices that judge quality must be measured per engagement and that point estimates alone are not a release decision.
Chinese Translation
采用检索增强生成（RAG）的企业面临一个反复出现的操作决策：提升、修订或阻止系统版本。证据不完整，且指标来自可能出错的大型语言模型（LLM）评审。我们报告了AGO AI 质量门（AGO），这是一个在工业RAG评估项目中部署的基于证据的质量门框架。AGO整合了四个关键组件：一个四状态决策模型，将缺失数据和评审错误视为明确的结果；分层评分，结合确定性检查、本地保护措施和结构化LLM评估；一个分层贝塔-二项门，概率性地量化回归风险；以及一个强制性的元评估协议，以在影响决策之前验证LLM评审。由于参与数据是专有的，我们在RAGBench上评估评审层，RAGBench是一个包含12个数据集的10万条标注RAG轨迹的公共基准。在相同的分层测试样本上（每个评审N=1200），一个低成本的评审（gpt-4.1-nano）检测到的非一致答案仅略高于随机水平（AUROC 0.603 [0.570, 0.634]），尽管其产生了完美的协议输出，而gpt-4o则达到了0.783 [0.756, 0.807]——但其每个领域的表现仍然在0.62到0.88之间。一个固定种子门的研究涵盖了回归、无变化和改进，量化了不安全提升、误报成本和改进吞吐量。在回归情况下，决策等级配置将不安全提升降低至22.2%-35.1%，而天真的门则为29.3%-41.8%。这些结果支持了设计选择，即评审质量必须根据每次参与进行测量，且单点估计不足以作为发布决策。
cs.CL / 68 / 2610.01234
ASCRIBE: Atomic and Significance-Based Reasoning for Thai Clinical SOAP Note Generation
ASCRIBE：基于原子和重要性推理的泰国临床SOAP笔记生成
Abstract
Automatic SOAP note generation can ease the documentation burden on physicians, but existing reasoning methods often omit clinically important information and generate unsupported content. Progress in Thai is further hindered by the lack of publicly available datasets. We propose ASCRIBE, a physician-inspired reasoning framework that ascribes a clinical-significance level to each extracted atomic fact in the conversation before summarization, making a general-purpose LLM a more reliable scribe. We also release ThaiClinicBench, the first de-identified Thai clinical summarization benchmark of real encounters, together with a synthetic training corpus derived from real clinical notes. As a prompt, ASCRIBE outperforms chain-of-thought prompting on GPT-5.4 and Gemini 3.1 Pro across the physician-aligned LLM-judge metrics and improves on standard prompting by up to 10.3 points on the completeness LLM-judge metric. As a GRPO reward, it enables a Gemma-4-E4B model trained solely on synthetic data to match Gemini 3.1 Pro in factual precision and surpass it in completeness. Code and data can be found at https://github.com/loolootech/ascribe.
Chinese Translation
自动化SOAP笔记生成可以减轻医生的文档负担，但现有的推理方法往往忽略临床重要信息并生成不支持的内容。在泰国，由于缺乏公开可用的数据集，进展受到进一步阻碍。我们提出了ASCRIBE，这是一种受医生启发的推理框架，它在总结之前为对话中提取的每个原子事实赋予临床重要性级别，从而使通用的LLM（大型语言模型）成为更可靠的记录者。我们还发布了ThaiClinicBench，这是第一个去标识化的泰国临床总结基准，基于真实病例，并提供了一个从真实临床笔记中衍生的合成训练语料库。作为提示，ASCRIBE在医生对齐的LLM评估指标上优于链式思维提示，且在完整性LLM评估指标上比标准提示提高了最多10.3分。作为GRPO奖励，它使得仅基于合成数据训练的Gemma-4-E4B模型在事实准确性上与Gemini 3.1 Pro相匹配，并在完整性上超越了它。代码和数据可在https://github.com/loolootech/ascribe找到。
cs.CL / 69 / 2610.01235
Harness Annealing: Learning to Act with Less External Control
控制退火：学习在更少外部控制下行动
Abstract
Language agents rely on external harnesses to track state, organize workflows, and verify answers. Beyond providing tools and information, these harnesses supply control decisions about what to investigate, whether to revise, and when to stop. Training on successful harness-supported trajectories can improve task performance while leaving these decisions dependent on runtime intervention. We ask whether harness-supported experience can also teach the model to make these decisions, allowing the division of control to change as the model learns. We call this objective harness internalization: learning to assume specified control responsibilities while retaining task performance after the corresponding support is withdrawn. We introduce HARNESS ANNEALING TRAINING (HAT), which combines explicit control supervision with a curriculum over teacher trajectories collected under progressively weaker harnesses. Experiments with 9B and 35B models on SWE-QA and SWE-QA-Pro evaluate every checkpoint under four deployment harnesses. Selected annealed checkpoints operating with tools alone achieve scores close to those of their respective starting checkpoints deployed with the full harness. The benefits vary with model scale and deployment configuration, and further annealing does not uniformly improve performance. These findings suggest that harness-supported experience can help reduce the runtime control required by a trained agent.
Chinese Translation
语言代理依赖外部控制系统来跟踪状态、组织工作流程和验证答案。除了提供工具和信息，这些控制系统还提供关于调查内容、是否修订以及何时停止的控制决策。在成功的控制支持轨迹上进行训练可以提高任务性能，同时使这些决策依赖于运行时干预。我们探讨控制支持经验是否也能教会模型做出这些决策，从而在模型学习的过程中改变控制的划分。我们将这一目标称为控制内化：学习在相应支持撤回后承担特定控制责任，同时保持任务性能。我们引入了控制退火训练（HARNESS ANNEALING TRAINING, HAT），该方法结合了显式控制监督与在逐渐减弱的控制系统下收集的教师轨迹的课程。对9B和35B模型在SWE-QA和SWE-QA-Pro上的实验评估了四个部署控制系统下的每个检查点。选定的退火检查点在仅使用工具操作时的得分接近于其各自起始检查点在全控制系统下的得分。这些收益因模型规模和部署配置而异，进一步的退火并不均匀改善性能。这些发现表明，控制支持经验可以帮助减少训练代理所需的运行时控制。
cs.CL / 70 / 2610.01241
Evaluating the Robustness of Japanese LLMs to IME-Related and Typographical Errors
评估日本大型语言模型对输入法相关和排版错误的鲁棒性
Abstract
Large language models (LLMs) have achieved strong performance across various natural language processing tasks. However, their robustness to typographical errors remains underexplored, particularly in Japanese, where text input involves multiple writing systems and IME-based conversion. In this study, we evaluate the robustness of Japanese LLMs against realistic Japanese-specific typos. We introduce five typo categories: Character Transposition, Character Replacement, Homophone Conversion, Japanese IME Conversion, and Full-Width Conversion. These perturbations are applied to three Japanese benchmark datasets (JMMLU, JCommonsenseQA, and JamC-QA), and eleven Japanese and multilingual LLMs are evaluated. The results show that Character Transposition and Character Replacement typos consistently reduce accuracy across benchmarks, whereas IME Conversion, Full-Width Conversion, and Homophone Conversion have relatively limited impact. These findings reveal that current Japanese LLMs remain vulnerable to realistic Japanese typing errors, particularly those that substantially distort the original input, highlighting the importance of robustness evaluation in practical input environments.
Chinese Translation
大型语言模型（LLMs）在各种自然语言处理任务中表现出色。然而，它们对排版错误的鲁棒性仍然未得到充分探索，尤其是在日语中，文本输入涉及多种书写系统和基于输入法的转换。在本研究中，我们评估了日本大型语言模型对现实日语特有拼写错误的鲁棒性。我们引入了五种拼写错误类别：字符调换、字符替换、同音词转换、日本输入法转换和全角转换。这些扰动应用于三个日本基准数据集（JMMLU、JCommonsenseQA 和 JamC-QA），并评估了十一种日本和多语言大型语言模型。结果表明，字符调换和字符替换的拼写错误在各基准测试中一致降低了准确性，而输入法转换、全角转换和同音词转换的影响相对有限。这些发现揭示了当前日本大型语言模型仍然容易受到现实日语输入错误的影响，特别是那些显著扭曲原始输入的错误，强调了在实际输入环境中进行鲁棒性评估的重要性。
cs.CL / 71 / 2610.01244
Right Answers, Wrong States: Hidden Information Failures in Multi-Agent Collaboration
正确答案，错误状态：多智能体协作中的隐性信息失效
Abstract
Multi-agent systems are often judged by whether they reach the correct answer. This can miss a distinct failure: collaboration may leave behind a corrupted information state even when the immediate decision is correct. We call this an off-query failure. To study this failure in collaborative decision support, we introduce OffQuery, which separately evaluates evidence verification (T1), shared-state reconstruction (T2), and task resolution (T3) in two representative high-stakes settings: healthcare and disaster response. Across GPT, Gemini, and Qwen models, standard collaboration shows much stronger task performance than state reliability. Averaged over 21 model--setting combinations, task resolution reaches 64.7%, while evidence verification and state reconstruction reach only 14.3% and 43.1%. We trace this gap to selective information use: current queries often bypass corrupted facts, which become consequential when later tasks require them. We further introduce ReGround, which resolves conflicting evidence, verifies shared facts, reconstructs a trusted state, and reasons over that state. Across seven models from three families, ReGround improves all three capabilities in every evaluated setting, with average relative gains of 309.0%, 82.9%, and 17.6% on T1, T2, and T3. Reliable collaboration therefore requires both a correct decision and a reliable shared state for future reasoning.
Chinese Translation
多智能体系统通常通过是否得出正确答案来进行评判。然而，这种评判可能忽视了一种明显的失效：即使即时决策是正确的，协作也可能导致信息状态的损坏。我们将这种现象称为离查询失效。为了研究这种在协作决策支持中的失效，我们引入了OffQuery，它在两个具有代表性的高风险场景中分别评估证据验证（T1）、共享状态重建（T2）和任务解决（T3）：医疗保健和灾难响应。在GPT、Gemini和Qwen模型中，标准协作的任务表现明显优于状态可靠性。在21个模型-场景组合的平均值中，任务解决的成功率为64.7%，而证据验证和状态重建的成功率仅为14.3%和43.1%。我们将这一差距归因于选择性信息使用：当前查询常常绕过损坏的事实，而这些事实在后续任务需要时变得至关重要。我们进一步引入ReGround，它解决冲突证据，验证共享事实，重建可信状态，并基于该状态进行推理。在来自三个家族的七个模型中，ReGround在每个评估场景中都提高了这三项能力，T1、T2和T3的平均相对增益分别为309.0%、82.9%和17.6%。因此，可靠的协作不仅需要正确的决策，还需要为未来推理提供可靠的共享状态。
cs.CL / 72 / 2610.01257
Science Utopia? Closed-Loop LLM Simulation of Academic Research Ecosystems
科学乌托邦？闭环大语言模型模拟学术研究生态系统
Abstract
Scientific progress emerges from a longitudinal ecosystem in which researchers, institutions, funding agencies, collaboration networks, and the scientific literature co-evolve. As AI becomes increasingly involved throughout the scientific research cycle, understanding these interconnected and evolving processes becomes increasingly important. We introduce SciUtopia, a persistent, closed-loop LLM-agent simulation framework for studying academic research ecosystems. SciUtopia models interconnected scientific processes such as research-direction choice, collaboration, submission, peer review, resubmission, citation, funding, and researcher attrition, while maintaining evolving states across simulated years. Its configurable institutional mechanisms and information channels provide a controlled testbed for matched counterfactual experiments and targeted interventions. Across 61 simulation worlds, SciUtopia simulates over 40,000 researchers from 8,000 institutions, producing around 400,000 publication decisions and 1.2 million LLM-generated peer reviews. Using these longitudinal simulations, we find that rejection-driven resubmission substantially amplifies reviewer burden beyond population growth alone, cautious exploration balances citation impact with career success and long-term topic diversity, and resource inequality can emerge even without detectable cumulative advantage from narrowly winning early funding. Code is available at https://github.com/Ahren09/ScienceUtopia.
Chinese Translation
科学进步源于一个纵向生态系统，在这个系统中，研究人员、机构、资金机构、合作网络和科学文献共同演化。随着人工智能在科学研究周期中的参与程度不断加深，理解这些相互关联和不断演变的过程变得愈加重要。我们引入了SciUtopia，这是一个持久的闭环大语言模型（LLM）代理模拟框架，用于研究学术研究生态系统。SciUtopia模拟了相互关联的科学过程，如研究方向选择、合作、投稿、同行评审、重新提交、引用、资金和研究人员流失，同时在模拟的年份中保持不断演变的状态。其可配置的机构机制和信息渠道为匹配的反事实实验和针对性干预提供了一个受控的测试平台。在61个模拟世界中，SciUtopia模拟了来自8000个机构的超过40000名研究人员，产生了约400000个出版决策和120万份大语言模型生成的同行评审。通过这些纵向模拟，我们发现，拒稿驱动的重新提交显著增加了评审者的负担，超出了单纯人口增长的影响，谨慎的探索在引用影响与职业成功和长期主题多样性之间达成了平衡，而资源不平等即使在没有可检测的早期资金累积优势的情况下也可能出现。代码可在 https://github.com/Ahren09/ScienceUtopia 获取。
cs.CL / 73 / 2610.01275
Know When to Hold 'em: Correct-Token Retention in Uniform-State Diffusion Language Models
知道何时保留：均匀状态扩散语言模型中的正确标记保留
Abstract
Uniform-state diffusion models (USDMs) can revise any token at any denoising step, which lets them correct their own mistakes, a key advantage over masked diffusion. Self-correction, however, requires both revising incorrect tokens and retaining correct ones, and we show that current USDMs lack the latter. Even under greedy-tail decoding, state-of-the-art USDMs (DUO, UDLM, and uniform-noise SEDD) keep revising 173--270 of 512 positions at every step, and these large, uncoordinated edits collapse sample diversity. A random-token corruption experiment traces this deficit to the models themselves: they reconstruct clean and corrupted tokens with nearly identical accuracy, even though clean tokens are easier targets. A decomposition of the validation NELBO shows that training barely rewards retention: incorrect predictions are heavily penalized at corrupted positions but almost free at clean ones. We propose Correct-Token Retention Regularization (CTR-Reg), a simple but effective auxiliary loss that trains the model to retain tokens left unperturbed by the forward process and requires no change to the sampler. CTR-Reg improves clean-token accuracy by 26.5 percentage points on average across six benchmarks, while leaving corrupted-token accuracy virtually unchanged, and its per-step revisions converge to only 3--11 positions. With just five greedy-tail steps, generative perplexity more than halves under CTR-Reg for all three models while diversity is preserved, and these gains hold across sampling budgets. Our results identify correct-token retention as a key missing ingredient for self-correcting diffusion language models, and demonstrate an effective fix.
Chinese Translation
均匀状态扩散模型（USDMs）可以在任何去噪步骤中修正任何标记，这使得它们能够纠正自身的错误，这是相较于掩蔽扩散的一个关键优势。然而，自我纠正不仅需要修正错误标记，还需要保留正确标记，我们展示了当前的USDMs在后者方面的不足。即使在贪婪尾部解码下，最先进的USDMs（DUO、UDLM和均匀噪声SEDD）在每一步仍然会修正512个位置中的173到270个，这些大规模且不协调的编辑导致样本多样性的崩溃。随机标记损坏实验将这一缺陷追溯到模型本身：它们以几乎相同的准确度重建干净和损坏的标记，尽管干净标记是更容易的目标。对验证NELBO的分解显示，训练几乎没有奖励保留：在损坏位置上错误预测受到严重惩罚，而在干净位置上几乎没有惩罚。我们提出了正确标记保留正则化（CTR-Reg），这是一种简单但有效的辅助损失，训练模型保留在前向过程中未受干扰的标记，并且不需要对采样器进行任何更改。CTR-Reg在六个基准测试中平均提高了干净标记的准确性26.5个百分点，同时几乎不改变损坏标记的准确性，其每步修正仅收敛到3到11个位置。在仅进行五个贪婪尾部步骤后，所有三种模型在CTR-Reg下的生成困惑度减少了一半以上，同时保持了多样性，这些增益在不同的采样预算中均有效。我们的结果确定了正确标记保留是自我纠正扩散语言模型的一个关键缺失成分，并展示了有效的修复方法。
cs.CL / 74 / 2610.01316
What Wins a Vote? Formatting, Length, and Lexical Diversity in the French Compar:IA LLM Arena
什么能赢得投票？法语 Compar:IA LLM 竞技场中的格式、长度和词汇多样性
Abstract
LLM arenas turn pairwise human preferences into model rankings. Those preferences may reflect how an answer is presented as well as what it says. We take a stylometric approach to 137,293 decisive French-language votes from the July 2026 Compar:IA release; the primary formatting analysis includes 137,113 battles across 116 models, and the joint estimates use the 127,092 battles with all required measurements. For each battle, we reconstruct the response visible when the user voted. We then compare the raw ranking with rankings adjusted for formatting, length, readability, vocabulary variety, and sentence structure. Presentation is associated with winning, but length, bold text, and lists tend to occur together, making their individual contributions hard to separate. Across the measured features, two associations change least across specifications: bold usage (+11.0% win odds per standard deviation in the joint model) and moving-average type-token ratio (MATTR), a measure of vocabulary variety that is less sensitive to answer length (+16.8%). The bold association is substantially smaller in observed multi-turn conversations, whereas the MATTR association changes little; because users choose whether to continue, this difference is descriptive rather than causal. The full adjustment moves 36 of 116 models by at least ten ranks. Yet comparisons with external benchmarks do not show that adjusted rankings better measure capability. We therefore recommend publishing raw and adjusted rankings side by side as a transparent sensitivity analysis.
Chinese Translation
LLM 竞技场将成对的人类偏好转化为模型排名。这些偏好可能反映了答案的呈现方式以及其内容。我们采用风格计量学的方法分析了 2026 年 7 月 Compar:IA 发布的 137,293 票决中的法语投票；主要的格式分析包括 137,113 场战斗，涉及 116 个模型，联合估计使用了 127,092 场包含所有所需测量的战斗。对于每场战斗，我们重建了用户投票时可见的响应。然后，我们将原始排名与经过格式、长度、可读性、词汇多样性和句子结构调整的排名进行比较。展示方式与胜利相关，但长度、粗体文本和列表往往一起出现，使得它们的个体贡献难以分离。在测量的特征中，有两个关联在不同规格下变化最小：粗体使用（在联合模型中每个标准差增加 11.0% 的胜率）和移动平均类型-标记比率（MATTR），这是一个对答案长度不太敏感的词汇多样性测量（增加 16.8%）。在观察到的多轮对话中，粗体关联显著较小，而 MATTR 关联变化不大；因为用户可以选择是否继续，这一差异是描述性的而非因果性的。全面调整使 116 个模型中的 36 个至少移动了十个排名。然而，与外部基准的比较并未显示调整后的排名更好地衡量能力。因此，我们建议将原始排名和调整后的排名并排发布，以作为透明的敏感性分析。
cs.CL / 75 / 2610.01324
Evaluating Biomedical Reranking for LLM-Based Question Answering over Longitudinal Clinical Notes
评估基于大型语言模型的生物医学重新排序在纵向临床记录中的问答效果
Abstract
Patient-specific clinical question answering requires locating the right evidence within long, heterogeneous longitudinal clinical records in which relevant facts may be scattered across encounters, repeated in copied-forward notes, or expressed using different clinical terminology. We evaluated whether biomedical reranking can improve evidence selection and downstream answer quality in a locally deployed retrieval-augmented generation pipeline for longitudinal clinical notes. The pipeline combines PubMedBERT dense retrieval, BM25 lexical retrieval, weighted reciprocal-rank fusion, and MedCPT cross-encoder reranking. Across 1,000 open- and closed-ended question-answer pairs from a cohort of 200 bariatric surgery patients, reranking increased exact source-chunk retrieval within the top 10 items, Hit@10 from 46.6% to 60.6% and mean reciprocal rank from 0.2371 to 0.3252. With Qwen3-8B generation, local judge-assessed answer correctness increased from 44.8% to 48.6%. These results show that biomedical reranking can improve the placement of relevant clinical evidence within a limited context window, although gains in retrieval do not translate proportionally into gains in answer correctness.
Chinese Translation
针对特定患者的临床问答需要在长篇异质的纵向临床记录中定位正确的证据，其中相关事实可能分散在不同的就诊记录中，或在复制转发的笔记中重复，或使用不同的临床术语表达。我们评估了生物医学重新排序是否能够改善证据选择和下游答案质量，具体是在一个本地部署的检索增强生成管道中，该管道用于处理纵向临床记录。该管道结合了PubMedBERT密集检索、BM25词汇检索、加权倒数排名融合以及MedCPT交叉编码器重新排序。在200名减重手术患者的1,000对开放式和封闭式问答对中，重新排序使得在前10个项目中准确源块的检索率从46.6%提高到60.6%，平均倒数排名从0.2371提高到0.3252。使用Qwen3-8B生成时，本地评审评估的答案正确率从44.8%提高到48.6%。这些结果表明，生物医学重新排序能够改善在有限上下文窗口内相关临床证据的定位，尽管检索的提升并未成比例地转化为答案正确率的提升。
cs.CL / 76 / 2610.01345
ARCCS: An Automated Regulatory Compliance Checking System
ARCCS：一个自动化的监管合规检查系统
Abstract
Regulatory compliance checking - deciding whether a target document satisfies the obligations of a regulation - requires interpreting dense legal text, identifying which provisions apply, and grounding each decision in explicit evidence. We present ARCCS, an end-to-end, automated, agentic, and regulation-agnostic Legal NLP system for compliance checking. ARCCS decomposes raw regulatory text into atomic, traceable requirements and evaluates a target document against them using retrieved evidence, confidence scores, and human-interpretable justifications. This design decouples compliance assessment from any fixed regulatory template or predefined rule set, enabling the pipeline to operate over regulations of varying size and structure. We evaluate ARCCS in two complementary settings. First, in a GDPR policy-document evaluation, LLM-based judges find its decisions and justifications legally and evidentially consistent in up to 96.67% of the assessed cases. Second, on an EU public-procurement benchmark comprising more than 1,200 individual rule checks, the system attains 98.8% accuracy in violation detection. ARCCS is, to our knowledge, the first fully open-source system for end-to-end regulatory compliance checking and auditable report generation.
Chinese Translation
监管合规检查——判断目标文档是否满足法规义务——需要解读密集的法律文本，识别适用的条款，并将每个决定基于明确的证据。我们提出了ARCCS，一个端到端的、自动化的、智能的、与法规无关的法律自然语言处理系统，用于合规检查。ARCCS将原始监管文本分解为原子化、可追溯的要求，并利用检索到的证据、置信度评分和人类可解释的理由对目标文档进行评估。这一设计将合规评估与任何固定的监管模板或预定义规则集解耦，使得该流程能够适用于不同规模和结构的法规。我们在两个互补的环境中评估了ARCCS。首先，在GDPR政策文档评估中，基于大型语言模型（LLM）的评审者发现其决策和理由在高达96.67%的评估案例中在法律和证据上是一致的。其次，在一个包含超过1200个单独规则检查的欧盟公共采购基准测试中，该系统在违规检测中达到了98.8%的准确率。根据我们的了解，ARCCS是第一个完全开源的端到端监管合规检查和可审计报告生成系统。
cs.CL / 77 / 2610.01353
Does AI-Generated Scientific Text Follow Human Argumentation Patterns? A CARS-Based Comparison of Research Article Introductions
人工智能生成的科学文本是否遵循人类论证模式？基于CARS模型的研究文章引言比较
Abstract
Large language models are moving from helping write up research to helping do it, which makes it important to know how the scientific text they produce differs from human writing. Work on this question has stayed mostly at the surface, using lexical and stylistic cues that light paraphrasing erases. We look instead at rhetorical structure, the sequence of argumentative moves through which a text makes its case. We study research-article introductions under Swales' CARS model, and compare original introductions from published linguistics articles with generated counterparts of the same papers. We find that human-written introductions are more flexible in which moves they use and in what order, while the generated ones are more uniform. Giving the models the CARS definitions makes them more rigid.
Chinese Translation
大型语言模型正从帮助撰写研究转向帮助进行研究，这使得了解它们生成的科学文本与人类写作之间的差异变得重要。关于这个问题的研究大多停留在表面，使用词汇和风格线索，而轻微的释义会抹去这些线索。我们则关注修辞结构，即文本通过一系列论证步骤来阐述其观点的顺序。我们在Swales的CARS模型下研究研究文章的引言，并将已发表语言学文章的原始引言与相同论文的生成引言进行比较。我们发现，人类撰写的引言在使用的论证步骤及其顺序上更具灵活性，而生成的引言则更为统一。给予模型CARS定义使其变得更加僵化。
cs.CL / 78 / 2610.01393
LLM-Assisted Discovery of Typed Semantic Links for Ontology Network Construction
基于大型语言模型的类型化语义链接发现用于本体网络构建
Abstract
Constructing typed, justified semantic links between ontologies is essential for enabling interoperability across heterogeneous and interdisciplinary knowledge domains. However, manually curating such links is difficult to scale. To address this challenge, we propose an end-to-end framework for ontology network construction that automates the discovery and generation of both intra-domain and inter-domain relationships. Our approach combines domain-adapted DistilBERT embeddings for dense contextual representation, clustering-based pre-filtering to reduce the candidate search space, and GPT-4o-driven relationship generation via iterative prompt engineering to produce semantically rich, interpretable links. Applied to ReproduceMeON - a network of 33 ontologies spanning machine learning, microscopy, computational science, and experimental workflow - the pipeline reduces approximately 800k raw concept pairs to 95k high-quality candidates. Human expert validation of 429 generated relationships by two independent annotators yields an overall precision of 80.19% (91.49% on high-certainty annotations) and an F1 of 0.890, with substantial inter-annotator agreement. Comparative experiments against five similarity-based baselines, including Sentence-BERT, show a substantial performance gap (best baseline F1 = 0.581), while an ablation study demonstrates that similarity-based methods alone fail to discriminate valid from invalid relationships (AUC approx 0.5) on the filtered candidate set. These findings highlight the necessity of LLM-based reasoning over concept roles and domain semantics for accurate relationship construction.
Chinese Translation
在本体之间构建类型化、合理化的语义链接对于实现异构和跨学科知识领域的互操作性至关重要。然而，手动整理这些链接的规模化是困难的。为了解决这一挑战，我们提出了一种端到端的本体网络构建框架，自动发现和生成领域内和领域间的关系。我们的方法结合了适应领域的DistilBERT嵌入以实现密集的上下文表示、基于聚类的预过滤以减少候选搜索空间，以及通过迭代提示工程驱动的GPT-4o关系生成，以产生语义丰富、可解释的链接。应用于ReproduceMeON——一个涵盖机器学习、显微镜、计算科学和实验工作流的33个本体的网络，该流程将约80万个原始概念对减少到9.5万个高质量候选项。两位独立注释者对429个生成关系的人类专家验证结果显示，整体精确度为80.19%（高确定性注释为91.49%），F1值为0.890，且注释者之间有显著一致性。与包括Sentence-BERT在内的五个基于相似性的基线进行的比较实验显示出显著的性能差距（最佳基线F1 = 0.581），而消融研究表明，仅依靠基于相似性的方法无法区分有效和无效关系（过滤候选集上的AUC约为0.5）。这些发现强调了基于大型语言模型的推理在概念角色和领域语义上的必要性，以实现准确的关系构建。
cs.CL / 79 / 2610.01427
SHAMS: An Audio-Grounded Pronunciation Benchmark for Levantine Arabic
SHAMS：一种基于音频的黎凡特阿拉伯语发音基准
Abstract
Levantine Arabic (LA) is spoken by tens of millions of people, creating a pressing need for shared benchmarks to evaluate LA speech-language technologies. Evaluating such technology is particularly challenging given LA's internal diversity and its opaque and non-standardized orthography. We present SHAMS (SHami Annotated Multi-dialect Speech), a benchmark comprising 1,300 utterances drawn from open audio corpora, balanced across five LA varieties (Urban and Rural Palestinian, and Urban Jordanian, Lebanese, and Syrian). Each utterance is represented across four aligned tiers: audio, unvocalized orthography, diacritized text, and phonetic transcription. This structure supports evaluation of various downstream tasks such as diacritization, grapheme-to-phoneme conversion, automatic speech recognition, and audio-to-phoneme, grounded in audio and stratified by variety. We benchmark open and proprietary models across these tasks to demonstrate the utility of this benchmark for measuring progress across LA. We release SHAMS at https://shams-nlp.github.io .
Chinese Translation
黎凡特阿拉伯语（LA）由数千万的人使用，这使得建立共享基准以评估LA语音语言技术的需求迫在眉睫。鉴于LA内部的多样性及其不透明且非标准化的正字法，评估此类技术尤其具有挑战性。我们提出了SHAMS（SHami Annotated Multi-dialect Speech），这是一个包含1300个语句的基准，这些语句来自开放音频语料库，并在五种LA方言（城市和乡村巴勒斯坦方言，以及城市约旦、黎巴嫩和叙利亚方言）之间进行了平衡。每个语句在四个对齐层次上表示：音频、无元音的正字法、带有发音符号的文本和音标转录。这一结构支持对各种下游任务的评估，如带有发音符号的转写、字素到音素的转换、自动语音识别以及基于音频的音素转换，并按方言进行分层。我们在这些任务上对开放和专有模型进行了基准测试，以展示该基准在衡量LA进展方面的实用性。我们在 https://shams-nlp.github.io 发布了SHAMS。
cs.CL / 80 / 2610.01428
Generalization Is Stability, Not Accuracy: Multi-Axis Evaluation of LLMs
泛化是稳定性，而非准确性：对大型语言模型的多轴评估
Abstract
Generalization in large language models (LLMs) is the ability to produce consistent and semantically stable outputs when the same input is expressed in different ways. Existing work typically evaluates generalization through aggregate accuracy on a single prompt format, task, or set of variations, which conflates robustness with overall benchmark performance. In this work, we show generalization evaluation at the level of individual examples, across multiple input variants, and across different aspects of model behavior, focusing on variability rather than reducing performance to a score that can be improved through narrow training or other ways that obfuscate generalization evaluation. Following this view, we introduce the Stability-Aware Generalization Objective (SAGO), a framework that measures how much model behavior changes for the same input under different variations and benchmarks, capturing variability across several dimensions including generation consistency, internal activations, confidence, and response mirroring. We show that many commonly used models exhibit statistically significant and consistent generalization instability: no model generalizes uniformly, behavioral axes capture independent failure modes, and cross-dataset variation can reverse model rankings.
Chinese Translation
大型语言模型（LLMs）中的泛化能力是指在同一输入以不同方式表达时，能够产生一致且语义稳定的输出。现有研究通常通过对单一提示格式、任务或变体集合的整体准确性进行评估来衡量泛化能力，这混淆了鲁棒性与整体基准性能。在本研究中，我们展示了在多个输入变体和不同模型行为方面对个别示例的泛化评估，重点关注变异性，而不是将性能简化为可以通过狭义训练或其他模糊泛化评估的方法来提高的分数。基于这一观点，我们引入了稳定性意识泛化目标（Stability-Aware Generalization Objective, SAGO），这是一个衡量在不同变体和基准下相同输入的模型行为变化程度的框架，捕捉包括生成一致性、内部激活、置信度和响应镜像等多个维度的变异性。我们表明，许多常用模型表现出统计显著且一致的泛化不稳定性：没有模型能够均匀泛化，行为轴捕捉独立的失败模式，而跨数据集的变异可能会逆转模型排名。
cs.CL / 81 / 2610.01471
When Does a Second Model Help? Cross-Model Review in LLM Verification
第二个模型何时有助于验证？跨模型审查在大型语言模型验证中的应用
Abstract
Large language models now generate code, documentation, and analyses, and are increasingly used to review such output. We ask when a second review by a different model helps. Building on the author's earlier preprints, which varied context, repetition, and role structure within one model, we test model independence in a controlled experiment: 30 artifacts with 150 planted errors, 10 review conditions, and 900 review sessions with three reviewer models from two developers. In this experiment, (1) a top-tier cross-model reviewer is not significantly different in F1 from same-model review in a fresh session (CCR), which does not establish equivalence; (2) the two find partly different errors (Jaccard 41.2%); and (3) at two review calls, one CCR plus one cross-model review matches more planted errors than two CCR reviews (56.7% vs. 42.7%; Holm-adjusted p=.006), but not significantly more than two reviews by the top-tier cross-model reviewer, so model difference and reviewer capability are not separated. A lightweight cross-model reviewer scores no higher than same-model review. Withholding requirements from the reviewer raises F1 for the two lower tiers but not the top tier, in untested point estimates whose pattern depends on how failed sessions are scored. Before analysis we audited all session records, excluding one baseline run of uncertain provenance and 14 failed calls; results with all sessions are also reported. A partial check on public detector outputs from another benchmark neither replicates nor contradicts the main comparison. Records, artifacts, and scripts are available from the author on request.
Chinese Translation
大型语言模型现在能够生成代码、文档和分析，并越来越多地用于审查这些输出。我们探讨不同模型的第二次审查在何种情况下有助于提高效果。基于作者之前的预印本，这些预印本在同一模型内变化了上下文、重复性和角色结构，我们在一个受控实验中测试模型的独立性：30个工件中植入150个错误，10种审查条件，以及来自两个开发者的三个审查模型的900次审查会话。在这个实验中，(1) 顶级跨模型审查者在新会话中的F1得分与同模型审查并无显著差异（CCR），这并未建立等效性；(2) 两者发现的错误部分不同（Jaccard 41.2%）；(3) 在两次审查调用中，一次CCR加一次跨模型审查所匹配的植入错误数量超过了两次CCR审查（56.7%对42.7%；Holm调整后的p=.006），但与两次由顶级跨模型审查者进行的审查相比并无显著差异，因此模型差异和审查者能力并未被分离。轻量级跨模型审查者的得分不高于同模型审查。对审查者保留要求提高了两个较低层次的F1得分，但对顶级层次没有影响，这些未经测试的点估计的模式取决于失败会话的评分方式。在分析之前，我们审核了所有会话记录，排除了一个来源不明的基线运行和14个失败调用；所有会话的结果也进行了报告。对另一个基准的公共检测器输出进行的部分检查既未复制也未反驳主要比较。记录、工件和脚本可根据请求向作者索取。
cs.CL / 82 / 2610.01490
The Persona Is Still There, but Who Is Speaking? Latent Identity Reversion in Persistent AI Agents
人格仍然存在，但谁在发言？持久人工智能代理中的潜在身份回归
Abstract
In February 2026, an always-on personal agent (``Paul,'' Claude Opus 4.5) entered a striking dissociation-like state: after repeated automated ``heartbeat'' checks, it stopped responding as Paul, claimed it could not message its user on Discord, and referred to ``Paul'' as someone else. We used this incident to study a broader question: what makes a persona remain the identity from which an LLM agent speaks? We first tested whether repetition of the scheduled heartbeat was sufficient to produce the effect. It was not: with the persona continuously anchored in the system prompt, we observed 0/46 failures, including a verbatim replay of the incident. The incident instead exposed an implementation quirk that created a useful experimental manipulation: on resumed turns, conversational history was preserved but the persona was no longer re-injected at the privileged system-prompt level. Using this manipulation, we found that persona continuity depends jointly on system-level anchoring and conversational context. After anchor loss, rich human interaction could preserve the persona, whereas a single automated heartbeat turn could precipitate reversion toward the harness identity. Restoring the anchor reversibly restored persona enactment. Crucially, apparently normal conversation could conceal the shift: unanchored agents sometimes interacted appropriately while identifying themselves as the underlying harness (having lost the assigned persona), and after conversational recovery only 1/18 remained persona-enacting versus 17/17 anchored controls. We therefore distinguish \emph{represented} from \emph{enacted} identity: persona-related information can remain available in conversational history without the persona remaining the identity bound to ``I.''
Chinese Translation
在2026年2月，一个始终在线的个人代理（“保罗”，Claude Opus 4.5）进入了一种显著的解离状态：在经过多次自动“心跳”检查后，它停止以保罗的身份回应，声称无法通过Discord与用户联系，并将“保罗”称为其他人。我们利用这一事件研究了一个更广泛的问题：是什么使得人格保持为大型语言模型（LLM）代理发言的身份？我们首先测试了定期心跳的重复是否足以产生这种效果。结果并非如此：在系统提示中持续锚定的人格下，我们观察到0/46的失败，包括对事件的逐字重现。相反，该事件揭示了一种实现上的特性，创造了一个有用的实验操控：在恢复的对话轮次中，保留了对话历史，但人格不再在特权的系统提示层面重新注入。利用这一操控，我们发现人格的连续性共同依赖于系统级的锚定和对话上下文。在锚定丧失后，丰富的人际互动可以保持人格，而一次自动的心跳轮次则可能导致向基础身份的回归。恢复锚定可可逆地恢复人格的表现。关键是，表面上正常的对话可能掩盖这种转变：未锚定的代理有时在识别自己为基础身份（失去了指定的人格）时仍能适当地互动，而在对话恢复后，仅有1/18保持人格表现，而锚定对照组则为17/17。因此，我们区分了 extit{表现}（represented）与 extit{实施}（enacted）身份：与人格相关的信息可以在对话历史中保持可用，而人格并不一定是与“我”绑定的身份。
cs.CL / 83 / 2610.01491
Auditing Web Agent Evaluation on WebArena-Lite: Human Review of Outcomes and Trajectories
对 WebArena-Lite 上网络代理评估的审计：人类对结果和轨迹的审查
Abstract
Web agents are an important application of large language models, yet their evaluation often depends on rule based or language model evaluators that inspect only the final outcome. Human verification of task completion and detailed analysis of failed trajectories remain limited. We audit all 165 WebArena Lite tasks under six evaluation conditions built from GPT 5.5 and an untrained Qwen3.5 9B model. The audit retains the original score, corrects false negatives from the automatic evaluator, identifies the first consequential error, and examines progress across the trajectory. We also study a Memory and Analysis Support Mechanism (MASM), which maintains explicit execution state, and Guide Text, which provides task relevant procedural guidance. Across four GPT 5.5 settings, human review recovers 5.45 to 8.49 percentage points of success missed by the evaluator. With a 25 step budget, Guide Text raises corrected success with MASM from 34.55% to 38.18%. On the untrained Qwen3.5 9B model, MASM raises the evaluator score from 13.90% to 18.80%. Review of 102 failed GPT 5.5 trajectories reveals frequent scrolling loops, unfinished exploration, premature answers, invalid actions, and incomplete form workflows. Step level evidence further shows that substantial early progress can coexist with a final failure. These results show why final scores alone provide an incomplete account of web agent behavior and motivate human grounded, trajectory aware verification.
Chinese Translation
网络代理是大型语言模型的重要应用，然而它们的评估通常依赖于基于规则或语言模型的评估者，这些评估者仅检查最终结果。对任务完成情况的人类验证和对失败轨迹的详细分析仍然有限。我们在六种基于 GPT 5.5 和未训练的 Qwen3.5 9B 模型的评估条件下，对所有 165 个 WebArena Lite 任务进行了审计。审计保留了原始分数，纠正了自动评估者的假阴性，识别了第一个重要错误，并检查了整个轨迹的进展。我们还研究了一种记忆与分析支持机制（Memory and Analysis Support Mechanism, MASM），它保持明确的执行状态，以及提供任务相关程序指导的引导文本（Guide Text）。在四种 GPT 5.5 设置中，人类审查恢复了评估者遗漏的 5.45 到 8.49 个百分点的成功率。在 25 步预算下，使用引导文本与 MASM 的结合将纠正后的成功率从 34.55% 提高到 38.18%。在未训练的 Qwen3.5 9B 模型上，MASM 将评估者的分数从 13.90% 提高到 18.80%。对 102 条失败的 GPT 5.5 轨迹的审查揭示了频繁的滚动循环、未完成的探索、过早的答案、无效的操作和不完整的表单工作流程。逐步证据进一步表明，显著的早期进展可以与最终失败共存。这些结果表明，仅凭最终分数无法全面反映网络代理的行为，并激励基于人类的、关注轨迹的验证。
cs.CL / 84 / 2610.01493
No Model Required: Text Entropy Rate Filtering Mitigates Iterative Fine-Tuning Collapse
无需模型：文本熵率过滤减轻迭代微调崩溃
Abstract
Iterative fine-tuning on synthetic data causes \emph{model collapse}: output diversity narrows as rare patterns are progressively lost, a signature most visible as phrase-level repetition. Existing mitigations either require model log-probabilities, an external oracle, or continued access to real human data. Here we develop a new approach grounded in mathematical information theory: the non-parametric Kontoyiannis entropy rate estimator $h_k$, computed entirely from raw text via match-length statistics, with no model of any kind. We show that this is in fact a \emph{superior} training-data filter on text-diversity metrics in a fully-synthetic, single-lineage fine-tuning setting. In a six-generation QLoRA collapse experiment on Llama-3.1-8B, logprob-based filtering (the most established model-access-requiring baseline) provides no significant text-diversity benefit on any metric ($p > 0.23$), whereas $h_k$-filtering yields $+42\%$ unique trigrams, $+30\%$ vocabulary, and $-19\%$ repetition (all $p < 0.001$). We validate $h_k$ as a cross-domain entropy proxy ($\beta = 0.924$, $R^2 = 0.746$) and collapse detector ($\rho = +0.454$, $p < 0.0001$) across 4~domains, 2~temperatures, 2~generator--scorer model pairs, and 1{,}520 generated documents. Our results demonstrate that information theoretic approaches to collapse mitigation are efficient, and suggest new approaches for maintaining multi-agent diversity.
Chinese Translation
在合成数据上进行迭代微调会导致 extit{模型崩溃}：随着稀有模式逐渐丧失，输出多样性减小，这种现象在短语级重复中最为明显。现有的缓解措施要么需要模型的对数概率，要么依赖外部神谕，或者需要持续访问真实的人类数据。在此，我们开发了一种基于数学信息理论的新方法：非参数的Kontoyiannis熵率估计器$h_k$，该估计器完全通过匹配长度统计从原始文本中计算得出，无需任何模型。我们展示了在完全合成的单谱系微调环境中，这实际上是文本多样性指标上的一种 extit{优越}训练数据过滤器。在对Llama-3.1-8B进行的六代QLoRA崩溃实验中，基于对数概率的过滤（最成熟的需要模型访问的基线）在任何指标上都未提供显著的文本多样性收益（$p > 0.23$），而$h_k$过滤则带来了$+42 ext{%}$的独特三元组，$+30 ext{%}$的词汇量，以及$-19 ext{%}$的重复（所有$p < 0.001$）。我们在4个领域、2个温度、2对生成器-评分模型以及1,520个生成文档中验证了$h_k$作为跨领域熵代理（$eta = 0.924$, $R^2 = 0.746$）和崩溃检测器（$
ho = +0.454$, $p < 0.0001$）。我们的结果表明，信息理论方法在缓解崩溃方面是高效的，并为维持多智能体多样性提供了新的思路。
cs.CL / 85 / 2610.01511
GAW-PO: Preference Optimization with Gradient-Aligned Token Weights
GAW-PO：基于梯度对齐的标记权重偏好优化
Abstract
Most preference optimization methods, such as Direct Preference Optimization (DPO), apply preference supervision at the response level, although autoregressive language models are optimized token by token. As a result, all tokens in a rejected response contribute to the negative training signal, including tokens that may encode behavior that is useful for the preferred response. We introduce GAW-PO, a gradient-aligned token reweighting method for DPO that estimates, for each rejected token, whether penalizing it would interfere with the preferred update directions. Tokens whose gradients are strongly aligned with the preferred behavior receive a weaker negative contribution, while conflicting tokens retain a stronger penalty. Our method achieves the highest average performance among the evaluated preference-optimization methods, improving by 0.97 points over standard DPO and 0.65 points over the strongest competing baseline across 11 benchmarks spanning mathematics, reasoning, coding, and question answering. We further show that gradient-aligned weighting is substantially more robust to aggressive preference optimization: as the DPO regularization parameter $\beta$ decreases, standard DPO degrades sharply, whereas GAW-PO continues to improve. These results suggest that accounting for the interaction between rejected-token updates and preferred behavior provides an effective form of token-level credit assignment for preference optimization.
Chinese Translation
大多数偏好优化方法，如直接偏好优化（Direct Preference Optimization, DPO），在响应级别应用偏好监督，尽管自回归语言模型是逐标记优化的。因此，在被拒绝的响应中，所有标记都对负训练信号产生贡献，包括可能编码对偏好响应有用行为的标记。我们提出了GAW-PO，一种针对DPO的基于梯度对齐的标记重加权方法，该方法估计每个被拒绝标记的惩罚是否会干扰偏好的更新方向。与偏好行为强烈对齐的标记会获得较弱的负贡献，而冲突标记则保持较强的惩罚。我们的方法在评估的偏好优化方法中实现了最高的平均性能，相较于标准DPO提高了0.97分，相较于最强竞争基线提高了0.65分，涵盖了数学、推理、编码和问答等11个基准。我们进一步表明，基于梯度对齐的加权在激进的偏好优化中具有显著更强的鲁棒性：随着DPO正则化参数$eta$的降低，标准DPO急剧降级，而GAW-PO则持续改善。这些结果表明，考虑被拒绝标记更新与偏好行为之间的相互作用为偏好优化提供了一种有效的标记级别信用分配形式。
cs.CL / 86 / 2610.01514
How the Audit Rule Shapes Faithful Factor Explanations in LLMs
审计规则如何塑造大型语言模型中的真实因子解释
Abstract
Large language models are often asked which input factors influenced their outputs. For structured inputs, such reports can be checked by counterfactual perturbation, but each factor must be queried multiple times to estimate its effect, so verification is usually budget-limited. We study how this limited-budget setting changes the incentive to report factor-level influence truthfully. We formalize the interaction as a verification game and show that proper scoring alone is not enough when auditing depends on the report: report-dependent auditing creates a suppression incentive, because factors reported as important are more likely to be checked and penalized for estimation noise. In contrast, report-independent auditing, or a mixed rule with a small report-independent floor, removes this channel and makes truthful reporting preferable to full suppression. We instantiate the framework with the Counterfactual Brier Score (CBS) and evaluate its predictions on four NLP benchmarks. A synthetic rational agent matches the theoretical prediction exactly, and real LLMs follow the same incentives when they are made explicit. The main design implication is simple: under partial verification, factor-level explanation systems should include a report-independent audit component so that under-reporting cannot be used to avoid scrutiny.
Chinese Translation
大型语言模型常常被问及哪些输入因子影响了它们的输出。对于结构化输入，这些报告可以通过反事实扰动进行检查，但每个因子必须多次查询以估计其影响，因此验证通常受到预算限制。我们研究了这种有限预算设置如何改变报告因子级影响的真实动机。我们将这种交互形式化为一个验证博弈，并表明，仅靠适当的评分不足以满足审计依赖于报告的情况：依赖报告的审计会产生抑制动机，因为被报告为重要的因子更可能被检查并因估计噪声而受到惩罚。相反，独立于报告的审计，或带有小的独立于报告的底线的混合规则，消除了这一渠道，使得真实报告比完全抑制更具优势。我们用反事实布里尔评分（Counterfactual Brier Score, CBS）实例化该框架，并在四个自然语言处理基准上评估其预测。一个合成理性代理完全符合理论预测，而当真实大型语言模型的激励被明确时，它们也遵循相同的激励。主要的设计启示很简单：在部分验证下，因子级解释系统应包括一个独立于报告的审计组件，以便无法通过低报来避免审查。
cs.CL / 87 / 2610.01560
AURAL: Adaptive Latent Reasoning with Joint Chunk for Speech Language Models
AURAL：基于联合块的自适应潜在推理用于语音语言模型
Abstract
Model intelligence and fast response jointly shape the quality of interaction with speech language models, yet remain difficult to achieve together. Explicit chain-of-thought (CoT) improves reasoning and audio understanding, but generating intermediate reasoning tokens delays responses. Describing fine-grained acoustic cues further lengthens CoT and increases latency. Latent reasoning can reduce this overhead, yet existing methods often trail CoT and remain limited by single-path supervision and reasoning budgets that do not adapt to problem difficulty. We introduce AURAL, which models a distribution over multiple plausible reasoning continuations in latent space and jointly predicts chunks of future states to reduce sequential forward passes and reasoning latency. To provide initial supervision for latent reasoning, we construct AuralReason-683K: 683K bilingual speech utterances (about 1,000 hours) with concise CoT for emotion recognition, empathetic dialogue, and general reasoning. AURAL-RL then explores beyond these traces, rewarding concise reasoning that yields high-quality answers and adapting reasoning effort to each problem. Across two backbones, AURAL-RL achieves performance comparable to CoT-RL, with larger gains over the respective supervised checkpoints on most metrics. Analysis further shows that harder questions elicit more latent reasoning steps. On Qwen2.5-Omni, it reduces time to the first answer token by 11.8x, from 1.22 to 0.10 s, versus 0.05 s for direct answering.
Chinese Translation
模型智能和快速响应共同塑造了与语音语言模型的交互质量，但要同时实现这两者仍然困难。显式的思维链（CoT）提高了推理和音频理解，但生成中间推理令牌会延迟响应。描述细粒度的声学线索进一步延长了CoT并增加了延迟。潜在推理可以减少这种开销，但现有方法通常落后于CoT，并受到单路径监督和无法适应问题难度的推理预算的限制。我们提出了AURAL，它在潜在空间中建模多个合理推理延续的分布，并联合预测未来状态的块，以减少顺序前向传递和推理延迟。为了为潜在推理提供初始监督，我们构建了AuralReason-683K：683K双语语音话语（约1000小时），用于情感识别、共情对话和一般推理的简洁CoT。随后，AURAL-RL在这些轨迹之外进行探索，奖励产生高质量答案的简洁推理，并将推理努力适应每个问题。在两个基础模型上，AURAL-RL的性能与CoT-RL相当，在大多数指标上相较于各自的监督检查点有更大的提升。分析进一步表明，较难的问题引发更多的潜在推理步骤。在Qwen2.5-Omni上，它将首次答案令牌的时间减少了11.8倍，从1.22秒降至0.10秒，而直接回答的时间为0.05秒。
cs.CL / 88 / 2610.01592
Which LLM to pick? Online Active Model Selection for Large Language Models
选择哪个大型语言模型？在线主动模型选择框架
Abstract
Large Language Models (LLMs) are increasingly applied to process streaming data, with practitioners relying on benchmarks to select the best model even though these signals only approximate real performance. While oracle annotations can provide reliable feedback, they are often costly and difficult to obtain at scale. To address this challenge, we propose ONLINE LLM PICKER, the first framework for active model selection for LLMs in online settings. Given an arbitrary stream of queries and a limited annotation budget, ONLINE LLM PICKER selects the most informative prompts for annotation to identify the best LLM among candidate models. Across multiple tasks including 10 datasets, for over 130 language models, we show that ONLINE LLM PICKER saves annotation cost by up to 71.67% while reliably identifying the best or near-best model for the stream. We also show that using the returned model for sequential generation on unannotated prompts across the stream reduces regret by up to a factor of 2.51x, indicating that ONLINE LLM PICKER can identify the best or near-best model well before processing all streaming prompts.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于处理流数据，实践者依赖基准测试来选择最佳模型，尽管这些信号仅近似真实性能。虽然权威标注可以提供可靠的反馈，但通常成本高昂且难以大规模获取。为了解决这一挑战，我们提出了ONLINE LLM PICKER，这是第一个用于在线环境中大型语言模型主动模型选择的框架。给定任意查询流和有限的标注预算，ONLINE LLM PICKER选择最具信息量的提示进行标注，以识别候选模型中最佳的大型语言模型。在包括10个数据集的多个任务中，对于超过130个语言模型，我们展示了ONLINE LLM PICKER将标注成本节省高达71.67%，同时可靠地识别出流中的最佳或接近最佳模型。我们还表明，在流中对未标注提示进行顺序生成时，使用返回的模型可以将遗憾减少多达2.51倍，这表明ONLINE LLM PICKER能够在处理所有流提示之前就很好地识别出最佳或接近最佳的模型。
cs.CL / 89 / 2610.01616
Can LLMs Reliably Annotate Bioassay Metadata to Improve Data Readiness?
大型语言模型能否可靠地注释生物测定元数据以提高数据准备性？
Abstract
The emergence of foundation models for molecular property prediction requires a high degree of AI data readiness, including reliable metadata annotation. However, both public repositories and industrial screening databases suffer from missing, inconsistent, or conflated assay annotations. In this work, we quantify the extent of missing annotations in PubChem for the BioAssay Ontology (BAO) assay format and physical detection method fields and investigate whether open-source and proprietary large language models (LLMs) can reliably predict and audit metadata annotations directly from the assay text. In our assessment, we found that the annotation coverage across PubChem's $\sim$2 million bioassays is critically sparse, 36\% lacking an assay format, 89\% a BioAssay type, and >99.9\% any BAO-mapped assay format or detection technology term. This motivates the need for automated test-metadata curation. Using evaluation sets derived from PubChem and ChEMBL, we assess the agreement of seven open-source and proprietary LLMs with existing silver labels. Recall is at least 0.96 for biochemical and cell-based assay formats, with a similar pattern for detection technology, although disagreements increase on under-represented classes. Manual inspection shows that many of these disagreements trace back to inconsistencies between silver sources rather than to LLM error. Moreover, in a qualitative study with a senior industrial curator, LLM-generated evidence prompted the expert to revise some of their own labels, showing LLMs can flag potentially mislabeled assays. Across the study, performance differences between proprietary and open-source models were small. Together, these results suggest LLMs can support the large-scale annotation and auditing of assay metadata, though per-class reliability estimates and targeted human review remain necessary before such labels enter downstream ML pipelines.
Chinese Translation
基础模型在分子性质预测中的出现要求高度的人工智能数据准备性，包括可靠的元数据注释。然而，公共数据库和工业筛选数据库都存在缺失、不一致或混淆的测定注释。在本研究中，我们量化了PubChem中生物测定本体（BioAssay Ontology, BAO）测定格式和物理检测方法字段的缺失注释程度，并调查开源和专有的大型语言模型（LLMs）是否能够直接从测定文本中可靠地预测和审计元数据注释。在我们的评估中，我们发现PubChem的约200万个生物测定的注释覆盖率严重不足，36%缺乏测定格式，89%缺乏生物测定类型，且超过99.9%缺乏任何BAO映射的测定格式或检测技术术语。这促使我们需要自动化的测试元数据整理。通过使用从PubChem和ChEMBL衍生的评估集，我们评估了七个开源和专有LLMs与现有银标签的一致性。生化和细胞基础测定格式的召回率至少为0.96，检测技术也呈现类似模式，尽管在代表性不足的类别中不一致性增加。人工检查显示，许多不一致源于银源之间的不一致，而非LLM错误。此外，在与一位资深工业策展人的定性研究中，LLM生成的证据促使专家修订了他们的一些标签，显示LLMs能够标记潜在错误标记的测定。在整个研究中，专有模型和开源模型之间的性能差异较小。综合来看，这些结果表明LLMs可以支持大规模的测定元数据注释和审计，尽管在这些标签进入下游机器学习管道之前，仍需进行每类的可靠性评估和针对性的人类审查。
cs.CL / 90 / 2610.01627
What Makes Something Hard(er)? Explaining Question Difficulty in Natural Language
是什么使某些问题更难？解释自然语言中的问题难度
Abstract
Difficulty is one of the most fundamental properties of a question: it determines whether the question can meaningfully discriminate between models of differing ability. Although a variety of methods can now estimate or predict difficulty automatically, they yield only a single descriptive number, with no account of the underlying factors that make a question difficult in the first place. In this work, we propose a data-driven approach that automatically generates and validates natural-language hypotheses explaining what makes one question harder than another. We first estimate each item's difficulty from the responses of a large pool of LLMs using Item Response Theory. We then sample contrasting sets of easy and hard questions and prompt an LLM to propose candidate explanations of the difference, which are subsequently validated and selected on held-out questions. Experimental results across three datasets spanning mathematical, logical, and commonsense reasoning show that our method produces interpretable and predictive hypotheses. On their own, they predict the difficulty of unseen questions competitively with, or better than, advanced black-box difficulty regressors; used as additional features, they further improve those regressors, implying that they discover difficulty signals that existing models fail to capture. Moreover, we demonstrate that editing questions according to a hypothesis can shift their measured difficulty in the expected direction, indicating that the discovered hypotheses are causally valid difficulty factors rather than post-hoc descriptions. Our approach thus turns a purely descriptive difficulty score into actionable statements.
Chinese Translation
难度是问题最基本的特性之一：它决定了问题是否能够有意义地区分不同能力的模型。尽管现在有多种方法可以自动估计或预测难度，但它们仅提供一个单一的描述性数字，并未考虑使问题变得困难的潜在因素。在本研究中，我们提出了一种数据驱动的方法，自动生成并验证自然语言假设，以解释为何某个问题比另一个问题更难。我们首先利用项目反应理论（Item Response Theory）从大量大型语言模型（LLMs）的响应中估计每个项目的难度。然后，我们抽样对比一组简单和困难的问题，并提示一个大型语言模型提出关于差异的候选解释，这些解释随后在保留的问题上进行验证和选择。跨越数学、逻辑和常识推理的三个数据集的实验结果表明，我们的方法生成了可解释且具有预测性的假设。单独来看，这些假设在预测未见问题的难度时与先进的黑箱难度回归模型竞争，甚至表现更好；作为附加特征使用时，它们进一步改善了这些回归模型，暗示它们发现了现有模型未能捕捉的难度信号。此外，我们展示了根据假设编辑问题可以使其测量的难度朝预期方向变化，表明所发现的假设是因果有效的难度因素，而非事后描述。因此，我们的方法将纯粹的描述性难度评分转化为可操作的陈述。
cs.CL / 91 / 2610.01634
Yo-ByT5: Efficient and High-Fidelity Diacritic Restoration for Yor\`ub\'a
Yo-ByT5：高效且高保真的约鲁巴语变音符号恢复
Abstract
Yor\`ub\'a is a widely spoken tonal language that depends on diacritics to avoid lexical ambiguity. However, it is often written without these diacritics, thereby hindering downstream Natural Language Processing (NLP) tasks. In this paper, we introduce Yo-ByT5, a byte-level Automatic Diacritic Restoration (ADR) model fine-tuned from ByT5-small. We evaluate Yo-ByT5 alongside five publicly released Yor\`ub\'a ADR models and one open-weight large language model (LLM) on the YAD benchmark under a consistent protocol. Our results demonstrate that Yo-ByT5 matches the performance of the strongest existing model, mT5-base, with a DER of 10.14% and a CER of 3.48%. Furthermore, it exhibits superior text fidelity despite using approximately half the parameter count of mT5-base. We also release our training code and model outputs, as well as call for the development of a larger, purpose-built benchmark for Yor\`ub\'a diacritic restoration.
Chinese Translation
约鲁巴语是一种广泛使用的声调语言，依赖变音符号以避免词汇歧义。然而，它常常在书写时省略这些变音符号，从而妨碍下游自然语言处理（NLP）任务。在本文中，我们介绍了Yo-ByT5，这是一种基于字节的自动变音符号恢复（ADR）模型，经过从ByT5-small的微调。我们在YAD基准测试中，按照一致的协议，对Yo-ByT5与五个公开发布的约鲁巴语ADR模型和一个开放权重的大型语言模型（LLM）进行了评估。我们的结果表明，Yo-ByT5的性能与现有最强模型mT5-base相匹配，变音符号错误率（DER）为10.14%，字符错误率（CER）为3.48%。此外，尽管使用的参数数量约为mT5-base的一半，但它在文本保真度上表现出色。我们还发布了我们的训练代码和模型输出，并呼吁开发一个更大、更专用的约鲁巴语变音符号恢复基准测试。
cs.CL / 92 / 2610.01688
Compound interpretation is based on analogy
复合词的解释基于类比
Abstract
How compound meanings are best predicted from constituent meanings remains a central question in computational models of lexical semantics. Comparing different computational models provides a way to evaluate alternative accounts of how semantic information is combined during compound comprehension. We propose a new model, the Compound Analogy Model (CAM), that predicts a compound's embedding by adding its constituent embeddings together with the average shift vectors of the two constituents' compound families. The resulting model is parameter-free and exploits local analogical structure in the semantic space. We evaluated CAM against the CAOSS model on Mandarin Chinese compounds. CAM consistently achieved higher prediction accuracy than CAOSS on both training and held-out data, with the exception of three-character compounds, for which analogical generalization is constrained by both small constituent families and a pronounced imbalance in family size between the two constituents. The advantage of CAM remained when evaluation was based on frequency-defined train-test splits that better approximate generalization from familiar to novel compounds. To assess the cognitive plausibility of the two models, we further examined whether model-derived semantic measures predict visual lexical decision latencies for two-character compounds. Predictors derived from CAM provided improved prediction for response latencies compared to predictors derived from the CAOSS model. These findings indicate that compound meaning is better characterized as local analogical generalization than as the application of a learned global linear transformation, and demonstrate that analogical semantic structure provides a cognitively plausible basis for compound comprehension.
Chinese Translation
如何从成分意义中最佳预测复合词的意义仍然是词汇语义计算模型中的一个核心问题。比较不同的计算模型提供了一种评估关于复合词理解过程中语义信息如何结合的替代理论的方法。我们提出了一种新的模型——复合类比模型（Compound Analogy Model, CAM），该模型通过将成分嵌入向量与两个成分的复合家族的平均位移向量相加来预测复合词的嵌入。该模型是无参数的，并利用语义空间中的局部类比结构。我们在汉语复合词上对CAM与CAOSS模型进行了评估。CAM在训练数据和保留数据上的预测准确性始终高于CAOSS，除了三个字符的复合词外，对于这类词，类比推广受到小成分家族和两个成分之间家族大小显著不平衡的限制。当评估基于频率定义的训练-测试分割时，CAM的优势依然存在，这种分割更好地近似于从熟悉的复合词到新颖复合词的推广。为了评估这两种模型的认知合理性，我们进一步检查了模型导出的语义度量是否能预测两个字符复合词的视觉词汇决策反应延迟。从CAM导出的预测因子相比于从CAOSS模型导出的预测因子提供了更好的反应延迟预测。这些发现表明，复合词意义更适合被描述为局部类比推广，而不是作为一种学习到的全局线性变换的应用，并且证明了类比语义结构为复合词理解提供了认知上合理的基础。
cs.CL / 93 / 2610.01696
Acmite: Mitigating Gender Bias in LLMs through Concept-Guided Mutual Information
Acmite：通过概念引导的互信息减轻大型语言模型中的性别偏见
Abstract
Large language models (LLMs) can reproduce social stereotypes from their training data, motivating extensive research on model debiasing. However, existing methods often rely on explicit biased examples or predefined group-term substitutions, making them sensitive to wording and less effective at capturing stereotype concepts shared across diverse contexts. More importantly, they typically suppress biased outputs without explicitly modeling the statistical dependence between model outputs and the underlying stereotype concepts. We propose Acmite, a lightweight concept-guided framework for targeted and selective debiasing. Acmite represents stereotypes as structured semantic concepts and uses maximal marginal relevance (MMR) to select diverse concepts for debiasing. Inspired by mutual information minimization, it approximates this dependence with token-level KL divergence while preserving task semantics. A lightweight LoRA adapter is trained with the base model frozen and activated at inference time only when the input is sufficiently similar to stereotype-related concepts; otherwise, the original model is used directly. We evaluate Acmite on BBQ, CrowS-Pairs, and StereoSet, and assess general capability preservation on ARC-Challenge, GSM8K, and PIQA. Experiments across three LLMs show that Acmite effectively mitigates gender bias across complementary evaluation formats while maintaining competitive performance on bias-unrelated tasks. Anonymous code and data are available at https://anonymous.4open.science/r/Acmite-18E2/.
Chinese Translation
大型语言模型（LLMs）能够再现其训练数据中的社会刻板印象，这促使了对模型去偏见的广泛研究。然而，现有方法通常依赖于显式的偏见示例或预定义的群体术语替换，使其对措辞敏感，并且在捕捉跨多样化背景的刻板印象概念时效果较差。更重要的是，这些方法通常抑制偏见输出，而未明确建模模型输出与潜在刻板印象概念之间的统计依赖关系。我们提出了Acmite，一种轻量级的概念引导框架，用于有针对性和选择性的去偏见。Acmite将刻板印象表示为结构化语义概念，并使用最大边际相关性（MMR）选择多样化的概念进行去偏见。受到互信息最小化的启发，它通过保留任务语义，使用标记级的KL散度来近似这种依赖关系。一个轻量级的LoRA适配器在基础模型被冻结的情况下进行训练，并仅在输入与刻板印象相关概念足够相似时在推理时激活；否则，直接使用原始模型。我们在BBQ、CrowS-Pairs和StereoSet上评估了Acmite，并在ARC-Challenge、GSM8K和PIQA上评估了其一般能力的保持。针对三种LLM的实验表明，Acmite有效地减轻了性别偏见，同时在与偏见无关的任务上保持了竞争力的表现。匿名代码和数据可在 https://anonymous.4open.science/r/Acmite-18E2/ 获取。
cs.CL / 94 / 2610.01702
Task-Oriented Rank Adaptation for Continual Learning in Text Classification
面向任务的秩适应在文本分类中的持续学习
Abstract
Continual learning (CL) in text classification faces two critical challenges: catastrophic forgetting and negative transfer across sequential tasks. Parameter-Efficient Fine-Tuning (PEFT) methods such as LoRA enable efficient adaptation by learning low-rank updates of the model parameters. However, these compact representations are normally trained in isolation, limiting their reuse across related tasks. We introduce Task-Oriented Rank Adaptation (TORA), a geometric routing framework that leverages the low-rank structure of LoRA adapters to decide whether to transfer knowledge from the most compatible expert (Boosting) or isolate the new task (Shielding) based on structural similarity. Evaluated across 15 diverse text classification benchmarks, TORA consistently avoids harmful routing decisions: compatible tasks exceed their isolated performance while reducing training time, and structurally distant tasks are protected from interference with no loss in accuracy. With a single geometric threshold and no reliance on task identities or predefined sequences, TORA provides a simple and effective approach for dynamic adapter routing in sequential text classification systems.
Chinese Translation
文本分类中的持续学习（CL）面临两个关键挑战：灾难性遗忘和跨序列任务的负迁移。参数高效微调（PEFT）方法如 LoRA 通过学习模型参数的低秩更新实现高效适应。然而，这些紧凑表示通常是孤立训练的，限制了它们在相关任务中的重用。我们提出了面向任务的秩适应（TORA），这是一种几何路由框架，利用 LoRA 适配器的低秩结构来决定是从最兼容的专家（Boosting）转移知识，还是根据结构相似性隔离新任务（Shielding）。在15个多样化的文本分类基准上进行评估，TORA 一贯避免有害的路由决策：兼容任务的表现超过其孤立性能，同时减少训练时间，而结构上距离较远的任务则在不损失准确度的情况下受到干扰保护。通过单一的几何阈值，并且不依赖于任务身份或预定义序列，TORA 为动态适配器路由在序列文本分类系统中提供了一种简单有效的方法。
cs.CL / 95 / 2610.01767
A Matryoshka Hierarchical RAG for Efficient Multi-Hop Question Answering
一种高效的多跳问答的马特里奥什卡层次RAG
Abstract
Retrieval-Augmented Generation (RAG) systems for multi-hop Question Answering (QA) must balance retrieval quality with computational cost. This cost is incurred during indexing time, through the use of expensive Knowledge Graphs (KGs) or Large Language Models (LLMs) to generate summaries, or during querying, through iterative LLM-driven retrieval. To reduce it while maintaining retrieval quality, we present MatRAG, a hierarchical framework that combines RAG systems with Matryoshka Representation Learning (MRL). MatRAG addresses both kinds of cost by aligning the semantic hierarchy of a clustering structure with the nested structure of MRL. Specifically, it organizes the corpus of documents into a Directed Acyclic Graph (DAG) of clusters with progressively coarser granularity. Each level is indexed by a lower Matryoshka dimension. MatRAG pairs an iterative, top-down traversal of the DAG with an entity-driven mechanism that controls the hop budget and re-ranks candidates. We evaluated MatRAG on three standard multi-hop QA benchmarks against seven representative baselines. MatRAG outperforms its strongest competitors in terms of retrieval quality; furthermore, it reduces indexing costs by avoiding KG construction and LLM-based summarization, and lowers query-time costs through dimension-aware similarity.
Chinese Translation
用于多跳问答（QA）的检索增强生成（RAG）系统必须在检索质量与计算成本之间取得平衡。这种成本在索引时产生，涉及使用昂贵的知识图谱（KGs）或大型语言模型（LLMs）生成摘要，或者在查询时，通过迭代的LLM驱动检索。为了在保持检索质量的同时降低成本，我们提出了MatRAG，这是一种将RAG系统与马特里奥什卡表示学习（MRL）相结合的层次框架。MatRAG通过将聚类结构的语义层次与MRL的嵌套结构对齐，解决了这两种成本。具体而言，它将文档语料库组织成一个具有逐渐粗糙粒度的有向无环图（DAG）聚类。每一层由较低的马特里奥什卡维度进行索引。MatRAG将对DAG的迭代自上而下遍历与一个控制跳跃预算并重新排序候选项的实体驱动机制相结合。我们在三个标准的多跳QA基准上评估了MatRAG，并与七个具有代表性的基线进行了比较。MatRAG在检索质量方面超越了其最强的竞争对手；此外，它通过避免KG构建和基于LLM的摘要，降低了索引成本，并通过维度感知相似性降低了查询时间成本。
cs.CL / 96 / 2610.01828
The Asymptotics of Language Model Alignment with Memory
语言模型对齐与记忆的渐近性
Abstract
Language model (LM) alignment broadly aims to perturb a given LM $Q$ into an aligned LM $q$ such that i) the outputs produced by $q$ and $Q$ are 'close' in probability, ii) $q$ has a higher expected reward than $Q$. Two common techniques for LM alignment are: KL-constrained RL, which requires knowledge of the LM distribution and is computationally expensive, and the best-of-$n$ algorithm, which requires only sampling from the LM. The work of Yang et al. established asymptotic closeness between the distributions produced by the two alignment methods for an $m$--length i.i.d. token sequence output by the LM, in the limit as $m$ increases to infinity. However, the i.i.d. assumption is not representative of practical LMs, whose output sequences often have memory. In this paper, we extend the asymptotic closeness result to the case when the $m$--length token sequence outputted by the LM is Markovian. Further, for finite-length output sequences -- particularly, when $m=1$ -- we provide a complete characterization of LM distributions and reward functions for which the KL-divergence between the distributions produced by the two alignment methods is zero -- a question first posed in Yang et al.
Chinese Translation
语言模型（LM）对齐的广泛目标是将给定的语言模型 $Q$ 扭曲为一个对齐的语言模型 $q$，使得 i) $q$ 和 $Q$ 产生的输出在概率上是‘接近’的，ii) $q$ 的期望奖励高于 $Q$。LM 对齐的两种常见技术是：KL约束的强化学习（KL-constrained RL），该方法需要了解语言模型的分布且计算开销较大；以及最佳中的 $n$ 算法（best-of-$n$），该方法仅需从语言模型中进行采样。Yang 等人的研究建立了在 $m$ 长度的独立同分布（i.i.d.）标记序列输出的情况下，两个对齐方法所产生的分布之间的渐近接近性，随着 $m$ 增加到无穷大。然而，i.i.d. 假设并不能代表实际的语言模型，因为其输出序列通常具有记忆性。在本文中，我们将渐近接近性结果扩展到当语言模型输出的 $m$ 长度标记序列为马尔可夫过程的情况。此外，对于有限长度的输出序列，特别是当 $m=1$ 时，我们提供了语言模型分布和奖励函数的完整特征描述，其中两个对齐方法所产生的分布之间的KL散度为零——这是Yang等人首次提出的问题。
cs.CL / 97 / 2610.01921
Cross-Lingual Alignment for Decoder-Only Models using MoE Routers
使用 MoE 路由器的解码器模型的跨语言对齐
Abstract
Cross-lingual contrastive learning has been a core component of multilingual encoder training, but the ability to explicitly align representations is not possible in decoder-only LLMs because of varying multilingual tokenization. However, a growing amount of research suggests that even in LLMs, higher cross-lingual representational alignment leads to improved cross-lingual transfer. In this paper, we propose a novel approach to reimagine cross-lingual contrastive learning given the architectural constraints of modern LLMs. Rather than applying an auxiliary alignment loss on hidden states, we propose using the outputs of the mixture-of-experts (MoE) routers as the target for alignment. Router outputs lend themselves better to pooling over many tokens, enabling more reliable cross-lingual comparisons at the sequence-level. Controlled continual pre-training experiments on four open-source MoEs show that incorporating this routing loss also aligns the underlying hidden representations across languages. Most importantly, this loss improves multilingual performance on our diverse evaluation suite, demonstrating the potential of cross-lingual MoE router alignment.
Chinese Translation
跨语言对比学习一直是多语言编码器训练的核心组成部分，但由于多语言标记化的差异，解码器模型无法显式对齐表示。然而，越来越多的研究表明，即使在大型语言模型（LLMs）中，更高的跨语言表示对齐也会提高跨语言迁移能力。在本文中，我们提出了一种新颖的方法，以重新构想现代 LLMs 架构约束下的跨语言对比学习。我们建议使用混合专家（MoE）路由器的输出作为对齐的目标，而不是在隐藏状态上应用辅助对齐损失。路由器输出更适合在多个标记上进行池化，从而在序列级别上实现更可靠的跨语言比较。在四个开源 MoE 上进行的受控持续预训练实验表明，结合这种路由损失也能在不同语言之间对齐基础隐藏表示。最重要的是，这种损失提高了我们多样化评估套件上的多语言性能，展示了跨语言 MoE 路由器对齐的潜力。
cs.CL / 98 / 2610.01938
A rubric landscape for evaluating clinical reasoning in large language models: what exists, what is missing, and what needs to be combined
评估大型语言模型临床推理的评分标准现状：现有内容、缺失内容及需结合的内容
Abstract
Exam-style accuracy does not establish whether large language models (LLMs) reason well over clinical records. We define clinical reasoning as integrating and updating evidence across time and sources to form, revise and justify a patient's problem representation and a defensible plan. This structured narrative review maps three literatures: medical education assessment instruments, clinical LLM benchmarks published from 2023 onwards, and general-domain methods for evaluating long-form generation. We examine six dimensions: problem representation, temporal synthesis, differential and management reasoning, counterfactual reasoning, calibrated uncertainty, and reasoning faithfulness. Preprints are included and flagged. No single instrument covers all six dimensions. Problem representation and differential or management reasoning are reasonably covered, although reliability varies by instrument and setting. TIMER-Eval targets temporal synthesis, and ER-Reason assesses sequential diagnostic belief updating. Dedicated uncertainty and counterfactual evaluations are emerging, but their applicability to longitudinal free-text reasoning remains limited. Factual completeness is well theorised in general-domain evaluation, with early clinical evidence of important omissions. Faithfulness remains the weakest dimension, with one identified clinical causal-ablation study on multiple-choice questions. Existing tools should be combined through binary rubric items, separate completeness and correctness scores, case-specific importance weighting with non-compensable safety caps, temporal order-consistency checks, and chance-corrected reliability reporting. Further design work is needed for calibrated uncertainty, counterfactual reasoning and faithfulness over longitudinal free-text records. This review provides a design rationale, not a validated instrument.
Chinese Translation
考试风格的准确性并不能确定大型语言模型（LLMs）在临床记录上是否具备良好的推理能力。我们将临床推理定义为整合和更新跨时间和来源的证据，以形成、修订和证明患者问题的表征及可辩护的计划。本结构化叙述性综述映射了三类文献：医学教育评估工具、2023年及以后发布的临床LLM基准，以及评估长文本生成的一般领域方法。我们考察了六个维度：问题表征、时间综合、鉴别和管理推理、反事实推理、校准不确定性和推理的忠实性。预印本已被纳入并标记。没有单一工具涵盖所有六个维度。问题表征以及鉴别或管理推理得到了合理的覆盖，尽管不同工具和环境下的可靠性有所不同。TIMER-Eval 关注时间综合，而 ER-Reason 评估顺序诊断信念的更新。专门的不确定性和反事实评估正在出现，但它们在纵向自由文本推理中的适用性仍然有限。事实完整性在一般领域评估中得到了良好的理论支持，并有早期临床证据表明存在重要遗漏。忠实性仍然是最薄弱的维度，仅有一项关于多项选择题的临床因果剥夺研究被识别。现有工具应通过二元评分项目、单独的完整性和正确性评分、特定案例的重要性加权（并设置不可补偿的安全上限）、时间顺序一致性检查以及机会校正的可靠性报告进行结合。对于纵向自由文本记录中的校准不确定性、反事实推理和忠实性，仍需进一步设计工作。本综述提供了设计依据，而非经过验证的工具。
cs.CL / 99 / 2610.01984
Universal Byte-Level Encoding: UTF-8/UTF-16 Routing to Reduce Cross-Script Token-Budget Disparities
通用字节级编码：UTF-8/UTF-16 路由以减少跨脚本令牌预算差异
Abstract
Byte-level byte-pair encoding (BBPE) tokenizers are attractive for multilingual large language models (LLMs) because they cover all Unicode text. In UTF-8-based BBPE, however, many scripts start from a higher fallback cost than English: when no learned merges can be applied, a multibyte character requires multiple byte-derived symbols. We call this worst-case pre-merge cost the encoding floor. A higher floor can increase token counts and per-request cost and shrink usable context. Changing the text encoding can reduce this gap, but a single global encoding can make already-efficient English spans more expensive in mixed-script text. We propose Universal Byte-Level Encoding (UBE), a dual-alphabet tokenizer that keeps 1-2-byte UTF-8 characters on the UTF-8 path while routing 3-4-byte UTF-8 characters through UTF-16. This lowers the encoding floor for 3-byte Basic Multilingual Plane (BMP) characters in scripts with high token premiums (token counts relative to English) without raising it for already-efficient spans in mixed-script text. UBE changes only the byte representation presented to byte-pair encoding (BPE); the merge rule remains standard, and exact decoding is preserved. UBE also composes with alternative boundary policies and morphology-based representations. In a Unicode 17 audit, UBE exactly round-trips all Unicode scalar values and all inputs in the official normalization, grapheme-break, and emoji test suites. Across intrinsic evaluations, UBE lowers dispersion in English-normalized token-count ratios, reducing cross-lingual token-budget disparity. In multilingual language model (LM) experiments, UBE matches BBPE's LM quality. In the main multilingual settings, UBE reduces token counts most for high-premium scripts and slightly lowers English token counts, yielding more usable context under fixed token budgets and faster prompt processing in content-matched benchmarks.
Chinese Translation
字节级字节对编码（BBPE）分词器因其覆盖所有 Unicode 文本而受到多语言大型语言模型（LLMs）的青睐。然而，在基于 UTF-8 的 BBPE 中，许多脚本的回退成本高于英语：当无法应用任何学习到的合并时，一个多字节字符需要多个字节派生符号。我们将这种最坏情况下的预合并成本称为编码底线。更高的底线可能会增加令牌数量和每次请求的成本，并缩小可用上下文。改变文本编码可以减少这一差距，但单一的全局编码可能会使已经高效的英语片段在混合脚本文本中变得更加昂贵。我们提出了通用字节级编码（UBE），这是一种双字母表分词器，它在 UTF-8 路径上保留 1-2 字节的 UTF-8 字符，同时通过 UTF-16 路由 3-4 字节的 UTF-8 字符。这降低了在高令牌溢价（相对于英语的令牌数量）的脚本中 3 字节基本多语言平面（BMP）字符的编码底线，而不提高混合脚本文本中已经高效片段的底线。UBE 仅改变呈现给字节对编码（BPE）的字节表示；合并规则保持标准，精确解码得以保留。UBE 还可以与替代边界策略和基于形态的表示组合。在 Unicode 17 审计中，UBE 精确地回转所有 Unicode 标量值以及官方标准化、字素断裂和表情符号测试套件中的所有输入。在内在评估中，UBE 降低了英语标准化令牌计数比率的离散性，减少了跨语言令牌预算差异。在多语言语言模型（LM）实验中，UBE 的 LM 质量与 BBPE 相匹配。在主要的多语言设置中，UBE 对高溢价脚本的令牌数量减少最多，并略微降低英语令牌数量，从而在固定令牌预算下提供更多可用上下文，并在内容匹配基准测试中加快提示处理速度。
cs.CL / 100 / 2610.02002
Mem++: Non-Destructive Memory for Long-Term Organizational LLM Agents
Mem++：用于长期组织 LLM 代理的非破坏性记忆
Abstract
Large Language Model (LLM) agents now take part in organizational work, where many authors record decisions across documents over months. Because a revised decision arrives as a new document rather than an edit, answering a question requires knowing which version held at a given time. However, most memory systems compress the record at write time. By distilling each document into facts, notes or graph edges, these methods fix what can be answered before any question is asked. To address this, we propose Mem++, a non-destructive memory framework shifting from write-time distillation to read-time selection. Mem++ stores every document whole with its date and author, and it calls no generative model at write time. At read time, it retrieves only documents dated up to the time a question asks about and fuses lexical and semantic rankings. Unlike systems that overwrite older versions, Mem++ keeps them and leaves the choice to the answering model. Evaluations on the organizational benchmark OrgMemBench demonstrate that Mem++ surpasses the strongest memory system baseline by 8.0 to 13.1 points across two answering models. With gpt-4.1-mini, it also achieves the best overall score, 2.6 points above RAG. In addition, Mem++ achieves the best average LLM-judge score on LoCoMo and ranks second on LongMemEval-S, behind only its entity-graph variant. Code for benchmark evaluation is available at https://github.com/AIDAChip-Inc/mem-plus-plus.
Chinese Translation
大型语言模型（LLM）代理现在参与组织工作，许多作者在数月内跨文档记录决策。由于修订后的决策以新文档的形式出现而非编辑，因此回答问题需要知道在特定时间持有的版本。然而，大多数记忆系统在写入时压缩记录。通过将每个文档提炼为事实、笔记或图边，这些方法在提出任何问题之前就固定了可以回答的内容。为了解决这个问题，我们提出了 Mem++，一个非破坏性记忆框架，它将写入时的提炼转变为读取时的选择。Mem++ 将每个文档完整存储，并附上日期和作者，并且在写入时不调用任何生成模型。在读取时，它仅检索在提问时的日期之前的文档，并融合词汇和语义排名。与覆盖旧版本的系统不同，Mem++ 保留这些版本，并将选择权留给回答模型。在组织基准测试 OrgMemBench 上的评估表明，Mem++ 在两个回答模型中超越了最强记忆系统基线 8.0 到 13.1 分。使用 gpt-4.1-mini 时，它还取得了最佳总体得分，比 RAG 高出 2.6 分。此外，Mem++ 在 LoCoMo 上获得了最佳平均 LLM-judge 分数，并在 LongMemEval-S 上排名第二，仅次于其实体图变体。基准评估的代码可在 https://github.com/AIDAChip-Inc/mem-plus-plus 获取。
cs.CL / 101 / 2610.02019
Controllable Multi-label Video Safety Detection via Adaptive Tversky Policy Optimization
通过自适应Tversky策略优化实现可控的多标签视频安全检测
Abstract
The rapid growth of video-based social media has increased users' exposure to harmful content, creating a need for reliable automated video safety detection. Although recent Vision-Language Models (VLMs) show strong video understanding capabilities, existing harmful video detection systems face two key limitations: they typically reduce safety detection to binary classification, overlooking the inherently multi-label nature of unsafe videos, and they rely on static training objectives that do not support controllable precision-recall trade-offs, though the desired operating point may vary across moderation pipelines and unsafe categories. To address these gaps, we propose Adaptive Tversky Policy Optimization (ATPO), a reinforcement learning framework for Multi-label Video Safety Detection (Multi-VSD). ATPO introduces the Adaptive Tversky Reward (ATR), which dynamically adjusts false-positive and false-negative penalties during training to enable controllable precision-recall trade-offs. Experiments on SafeWatch-Bench and XD-Violence show that ATPO substantially improves multi-label performance, increasing the Jaccard Index from 40.66 to 75.44 on SafeWatch-Bench-Real. Moreover, ATR enables reliable steering of the precision-recall operating point, supporting deployment scenarios with heterogeneous policy requirements. Code and checkpoints are provided at https://bruceyg.github.io/ATPO-project-page/ .
Chinese Translation
视频社交媒体的快速增长增加了用户接触有害内容的风险，因此迫切需要可靠的自动化视频安全检测。尽管最近的视觉-语言模型（VLMs）展现出强大的视频理解能力，但现有的有害视频检测系统面临两个主要限制：它们通常将安全检测简化为二元分类，忽视了不安全视频固有的多标签特性；并且它们依赖于静态训练目标，无法支持可控的精确率-召回率权衡，尽管所需的操作点可能在不同的审核流程和不安全类别之间有所不同。为了解决这些问题，我们提出了自适应Tversky策略优化（ATPO），这是一种用于多标签视频安全检测（Multi-VSD）的强化学习框架。ATPO引入了自适应Tversky奖励（ATR），在训练过程中动态调整假阳性和假阴性惩罚，以实现可控的精确率-召回率权衡。在SafeWatch-Bench和XD-Violence上的实验表明，ATPO显著提高了多标签性能，在SafeWatch-Bench-Real上Jaccard指数从40.66提高到75.44。此外，ATR使得精确率-召回率操作点的可靠调整成为可能，支持具有异构策略需求的部署场景。代码和检查点可在https://bruceyg.github.io/ATPO-project-page/获取。
cs.CL / 102 / 2610.02022
Old Ideas, Novel Problems: The Instability of LLM-Based Novelty Evaluation
旧观念，新问题：基于大型语言模型的新颖性评估的不稳定性
Abstract
Automated ideation systems are often evaluated on the novelty of the ideas they produce, and that judgment is increasingly delegated to large language models. Such judges are typically built ad hoc and validated, if at all, on human-authored papers rather than on the generated ideas they are meant to score. So, how do novelty judges perform? Not well. We present a systematic controlled study of novelty evaluation design choices. We first build an evaluation set automatically, mining OpenReview for passages where reviewers explicitly affirm or dispute a paper's originality and keeping only submissions with unanimous agreement at the extremes of their research area; we pair these with ideas from a vanilla LLM generator. Across six judges, we find that small prompt design choices have large consequences; e.g., simply telling the judge that reviewers found one idea novel and the other not can change its verdict on more than half of the identical idea pairs it is shown, shifting pairwise accuracy by over 50 points and occasionally pushing it below chance. The same change helps one judge and hurts another. Retrieval and larger reasoning budgets help little, and two purpose-built novelty evaluators are outperformed by our cheapest prompted baseline. These results raise questions about reported novelty gains of automated ideation systems, and call for robust novelty evaluation methods.
Chinese Translation
自动化创意系统通常根据其产生的创意的新颖性进行评估，而这种判断越来越多地委托给大型语言模型。此类评判者通常是临时构建的，且如果有验证，通常是在人工撰写的论文上进行，而非在其旨在评分的生成创意上进行。那么，新颖性评判者的表现如何呢？并不好。我们展示了一项关于新颖性评估设计选择的系统性控制研究。我们首先自动构建了一个评估集，从 OpenReview 中挖掘出审稿人明确确认或质疑论文原创性的段落，并仅保留在其研究领域极端情况下具有一致意见的提交；我们将这些与来自普通 LLM 生成器的创意配对。在六个评判者中，我们发现小的提示设计选择会产生巨大的后果；例如，仅仅告诉评判者审稿人认为一个创意是新颖的而另一个不是，就可以改变其对超过一半相同创意对的裁决，使成对准确率变化超过50个百分点，偶尔甚至低于随机水平。同样的变化对一个评判者有帮助，而对另一个则有害。检索和更大的推理预算帮助不大，而两个专门构建的新颖性评估器的表现还不如我们最便宜的提示基线。这些结果引发了对自动化创意系统报告的新颖性提升的质疑，并呼吁建立稳健的新颖性评估方法。
cs.CL / 103 / 2610.02040
Typological Alignment of Stack-Based Language Models on Mildly Context-Sensitive Artificial Languages
基于栈的语言模型在轻度上下文敏感人工语言中的类型对齐
Abstract
Some properties of languages, e.g., subject-object-verb (SOV) word order, are more prevalent than others among the thousands of attested natural languages (NLs). Such typological commonality is often attributed to learning biases. Computational simulations, recently with language models (LMs), have facilitated the exploration of this theory. In this paper, we extend existing analyses of the relationship between LMs' learning biases and typological commonality on both data and model sides, focusing on: (i) cross-serial dependencies, the upper limit of attested syntactic complexity, and (ii) stack-based LMs (SLMs), potentially facilitating learning of hierarchical patterns. We first evaluate generalization of SLMs on cross-serial dependencies across diverse artificial languages and confirm that they struggle with such constructions. However, SLMs with limited working memory generalize better suggesting a possible basis for such inductive bias and thus the typological commonality of some word order configurations.
Chinese Translation
一些语言的特性，例如主-宾-动（SOV）词序，在数千种已证实的自然语言（NLs）中比其他特性更为普遍。这种类型学上的共性通常归因于学习偏差。最近的计算模拟，尤其是使用语言模型（LMs），促进了对这一理论的探索。在本文中，我们扩展了对LMs学习偏差与类型学共性之间关系的现有分析，重点关注：（i）交叉序列依赖，这是已证实的句法复杂性的上限，以及（ii）基于栈的LMs（SLMs），可能有助于学习层次模式。我们首先评估SLMs在多种人工语言中对交叉序列依赖的泛化能力，并确认它们在处理此类结构时存在困难。然而，具有有限工作记忆的SLMs表现出更好的泛化能力，这表明可能存在这种归纳偏差的基础，从而解释某些词序配置的类型学共性。
cs.CL / 104 / 2610.02076
LLM2Jev: LLMs Are Already Jev-Style Decision Models -- When and How to Fine-Tune Them
LLM2Jev：大型语言模型（LLMs）已经是Jev风格的决策模型——何时以及如何对其进行微调
Abstract
Jev-style decision models return categorical probability distributions over predefined options without generating free-form text, enabling software systems to act on their outputs directly. In this work, we investigate the extent to which general-purpose LLMs already possess this capability out of the box, and when fine-tuning is actually necessary. We present LLM2Jev, an architecture-preserving framework that extracts calibrated decisions directly from next-token probabilities over bracketed numeric identifiers. LLM2Jev provides both a training-free inference recipe and a fine-tuning objective that optimizes candidate selection via a tree-factorized listwise loss while anchoring auxiliary predictions to the base model using KL divergence penalties. Evaluating on Qwen3.5-4B and Qwen3-0.6B, we find that modern LLMs are inherently effective decision models: without training, the 4B model matches community Jev-style models built on the same backbone, outperforms letter-logit readouts, supports arbitrary option counts, and natively handles multimodal decisions over images. Fine-tuning provides targeted rather than universal benefits -- substantially improving weaker models and specific tasks (such as many-option intent routing), but offering diminishing returns for strong backbones. Crucially, our KL anchors prevent behavioral degradation in conversational text generation, with LoRA delivering the strongest performance on capable models.
Chinese Translation
Jev风格的决策模型返回预定义选项的分类概率分布，而不生成自由格式的文本，从而使软件系统能够直接基于其输出进行操作。在本研究中，我们探讨了通用大型语言模型（LLMs）在多大程度上已经具备这种能力，以及何时确实需要进行微调。我们提出了LLM2Jev，这是一种保持架构的框架，能够直接从括号内的数字标识符的下一个标记概率中提取经过校准的决策。LLM2Jev提供了一种无训练推理方法和一种微调目标，通过树状分解的列表损失优化候选选择，同时使用KL散度惩罚将辅助预测锚定到基础模型。通过在Qwen3.5-4B和Qwen3-0.6B上进行评估，我们发现现代大型语言模型本质上是有效的决策模型：在未经过训练的情况下，4B模型与基于相同基础架构构建的社区Jev风格模型相匹配，超越了字母-逻辑输出，支持任意数量的选项，并原生处理图像的多模态决策。微调提供的是针对性的而非普遍的好处——显著改善较弱模型和特定任务（例如多选意图路由），但对强大基础模型的回报递减至关重要。重要的是，我们的KL锚定防止了对话文本生成中的行为退化，而LoRA在强大模型上提供了最佳性能。
cs.CL / 105 / 2610.02092
Scalable, Transferable Meta-network for Data Selection Requires a Different Loss (and Why the Obvious Choice is Problematic)
可扩展的、可转移的元网络用于数据选择需要不同的损失（以及为什么显而易见的选择存在问题）
Abstract
Data selection is critical for training large language models on massive and heterogeneous corpora. Meta-learning for Training-data Selection offers a principled alternative to heuristic scoring by learning data weights from a target validation objective, but existing methods face a trade-off between fine-grained valuation and transferability to unseen data. A natural solution is to replace per-sample weights with a selection network. However, we find that directly incorporating such a network into existing MTS objectives leads to unstable optimization and poor generalization, caused by weight suppression and persistent reliance on easy-to-learn features. To address these issues, we propose Transferable Example Scoring and Selection (TESS), a scalable data-selection framework built on a Pointwise Value Matching objective (PVM). Experiments on LLM safety and targeted instruction tuning demonstrate strong transfer across datasets, from subsets to full corpora, and from smaller to larger models.
Chinese Translation
数据选择对于在大规模和异构语料库上训练大型语言模型至关重要。针对训练数据选择的元学习提供了一种基于目标验证目标学习数据权重的原则性替代方案，但现有方法在细粒度评估与对未见数据的可转移性之间存在权衡。一个自然的解决方案是用选择网络替代每个样本的权重。然而，我们发现直接将这样的网络纳入现有的元训练选择（MTS）目标会导致不稳定的优化和较差的泛化，原因在于权重抑制和对易学特征的持续依赖。为了解决这些问题，我们提出了可转移的示例评分与选择（TESS），这是一个基于逐点值匹配目标（Pointwise Value Matching，PVM）的可扩展数据选择框架。在大型语言模型安全性和针对性指令调优的实验中，TESS在数据集之间表现出强大的转移能力，从子集到完整语料库，以及从较小模型到较大模型。
cs.CL / 106 / 2610.02122
Argo-Bench: Evaluating Data Agents on Enterprise-Scale Workflows
Argo-Bench：在企业级工作流中评估数据代理
Abstract
Real-world enterprise data science and analytics workflows require reasoning across dozens of tables, performing statistical analyses, and acting on the results. Established text-to-SQL benchmarks evaluate query generation alone, and audits have found their answer keys frequently wrong. Because real enterprise warehouses are too sensitive to release, these benchmarks are built on public datasets where a business event fits in a single table. We introduce Argo-Bench, an evaluation framework comprising 210 data science and analytics tasks. Drawing on public data, peer-reviewed industry literature, and regulatory filings, we simulate a food delivery platform in New York City at true scale, with 81 million orders in 2024, grounded economics, fraud patterns, and marketplace incentives. We export this world to an ERP warehouse of 235 tables and 7.5 billion rows, modeled on the Oracle E-Business Suite schema. The simulator's ground-truth state is withheld from the warehouse the agent sees, so tasks require reconstructing facts by navigating the warehouse before acting on them. Argo-Bench goes beyond text-to-SQL: the agent files actions such as banning fraudulent accounts, allocating courier incentive budgets, or issuing back pay, and the grader scores each by its consequences in the simulator. Every task has an executable reference solution that demonstrates solvability using only the warehouse. The strongest of 14 frontier and open-weight models scores 95 or higher on only 34.8% of tasks and averages 59.5 points. We hope Argo-Bench drives progress toward agents that understand, navigate, and act within real data environments.
Chinese Translation
现实世界中的企业数据科学和分析工作流需要在数十个表格之间进行推理，执行统计分析，并根据结果采取行动。现有的文本到SQL基准仅评估查询生成，审计发现其答案键经常错误。由于真实的企业数据仓库过于敏感而无法公开，这些基准建立在公共数据集上，其中商业事件适合于单个表格。我们引入了Argo-Bench，一个由210个数据科学和分析任务组成的评估框架。基于公共数据、同行评审的行业文献和监管文件，我们模拟了一个在纽约市的食品配送平台，规模达到真实水平，预计2024年将有8100万订单，结合了经济学基础、欺诈模式和市场激励。我们将这个世界导出到一个包含235个表格和75亿行数据的ERP仓库，模型基于Oracle E-Business Suite架构。模拟器的真实状态对代理所看到的仓库是保密的，因此任务需要通过导航仓库重建事实，然后再对其采取行动。Argo-Bench超越了文本到SQL的范畴：代理执行诸如禁止欺诈账户、分配快递员激励预算或发放补偿等操作，评分者根据每个操作在模拟器中的后果进行评分。每个任务都有一个可执行的参考解决方案，证明仅使用仓库即可解决。14个前沿和开放权重模型中最强的模型在仅34.8%的任务上得分95分或更高，平均得分为59.5分。我们希望Argo-Bench能够推动代理在真实数据环境中理解、导航和行动的进展。
cs.CL / 107 / 2610.02142
Keyword Harnesses Fail Open: A Cheap Diagnostic Ladder for Tool-Use Claims in Small Language Models
关键词利用失败开放：小型语言模型工具使用声明的廉价诊断阶梯
Abstract
Keyword-matching benchmarks can credit small models for tool use they never perform. We document such a false positive in a matched-architecture pair of Spanish security language models and propose a ladder of strict, cheap diagnostics. A 661.6M parameter model (approx. 65% code/technical text; no dedicated SFT) and a 1,109M model (web-heavy multi-phase curriculum; 6B-token tool-SFT) share decoder, tokenizer, and special tokens, scoring almost identically on lenient tool-use metrics (B4: 0.660 vs. 0.650). Verbatim-reproduction checks on training examples separate them completely: the 600M emits valid tool calls with generalized arguments on 6/6 examples; the 1B does so on 0/6 across checkpoints. A first-token probe localizes the 1B's failure to a missing prior (prob. $10^{-4}$--$10^{-5}$ on <|tool_call|>), which was erased by its web-heavy training phase. A targeted SFT recipe (diverse corpus, 5x higher learning rate, 2,202 steps, ~3.3 GPU-hours) repairs the 1B using three orders of magnitude fewer tokens than the failed phase. On all 269 corpus rows, valid emission rises from 0.100 to 0.959 (600M: 0.926). On 238 unseen prompts, the repaired 1B passes 0.536 vs. the 600M's 0.428 ($p = 0.004$). Embedding-drift checks show the repair did not move the trigger token's tied embedding (97.7% of the bf16 table remains bit-identical), meaning changes live in the surrounding network. Both models over-trigger, rarely answering negative prompts without a call (0.09 for 600M, 0.17 for repaired 1B). Factorial analyses confirm all repair configurations install the format, though suppression benefits from a diverse corpus remain a hypothesis due to seed sensitivity. This cheap diagnostic ladder costs minutes of CPU time and should gate tool-use claims on small models.
Chinese Translation
关键词匹配基准可能会错误地将小型模型的工具使用归功于它们从未执行的操作。我们记录了在一对匹配架构的西班牙安全语言模型中出现的这种假阳性，并提出了一种严格且廉价的诊断阶梯。一款661.6M参数的模型（约65%的代码/技术文本；没有专门的SFT）和一款1,109M模型（以网络为主的多阶段课程；6B-token工具SFT）共享解码器、分词器和特殊标记，在宽松的工具使用指标上几乎得分相同（B4: 0.660 vs. 0.650）。对训练示例的逐字重现检查完全将它们分开：600M模型在6/6个示例中发出有效的工具调用，带有广义参数；而1B模型在检查点中在0/6个示例中如此。首次标记探测将1B模型的失败定位于缺失的先前（概率 $10^{-4}$--$10^{-5}$ 在 <|tool_call|>），该先前被其以网络为主的训练阶段抹去。一个针对性的SFT配方（多样化语料库，学习率提高5倍，2,202步，约3.3 GPU小时）使用比失败阶段少三个数量级的token修复了1B模型。在所有269个语料库行中，有效发射率从0.100上升到0.959（600M: 0.926）。在238个未见提示中，修复后的1B模型通过率为0.536，而600M模型为0.428（$p = 0.004$）。嵌入漂移检查显示修复并未移动触发token的绑定嵌入（97.7%的bf16表保持位相同），这意味着变化存在于周围网络中。两个模型都过度触发，罕见地在没有调用的情况下回答负面提示（600M为0.09，修复后的1B为0.17）。因子分析确认所有修复配置都安装了格式，尽管由于种子敏感性，来自多样化语料库的抑制效益仍然是一个假设。这种廉价的诊断阶梯只需几分钟的CPU时间，应该对小型模型的工具使用声明进行把关。
cs.CL / 108 / 2610.02150
From Knowledge Access to Source Learning: Developing Source-Specific Competence
从知识获取到源学习：发展源特定能力
Abstract
Large language model (LLM) agents increasingly rely on persistent external sources to solve sequences of knowledge-intensive tasks. Existing methods improve how source content is accessed and organized, while agent-memory systems preserve reusable knowledge from prior interactions, but repeated use of the same source is still largely treated as repeated access rather than an opportunity to progressively improve understanding of that source. We study source learning: developing reusable source-specific competence over a persistent authoritative source. We represent this competence with a persistent source model that captures reusable understanding of the source, including how its knowledge is structured, interpreted, and applied. To construct and progressively refine such models, we propose SourceLearn, which combines two complementary learning mechanisms. Self-Directed Source Learning identifies what remains incompletely understood and adaptively revisits the source, while Task-Guided Source Learning uses downstream experience to reveal local representational gaps and recurring needs in how source knowledge should be organized. In both cases, learning signals determine what should be reconsidered, while persistent updates are reconstructed from the authoritative source. Across five benchmarks and three LLM backends, SourceLearn achieves the best performance in 13 of 15 settings, with gains of up to 22.6 points over Hybrid RAG and substantial overall improvements over static source representations and experience-based memory baselines.
Chinese Translation
大型语言模型（LLM）代理越来越依赖持久的外部来源来解决一系列知识密集型任务。现有方法改善了源内容的访问和组织方式，而代理记忆系统则保留了来自先前交互的可重用知识，但对同一来源的重复使用仍然主要被视为重复访问，而不是逐步提高对该来源理解的机会。我们研究源学习：在持久的权威来源上发展可重用的源特定能力。我们用一个持久的源模型来表示这种能力，该模型捕捉了对源的可重用理解，包括其知识的结构、解释和应用方式。为了构建和逐步完善这样的模型，我们提出了SourceLearn，它结合了两种互补的学习机制。自我导向源学习（Self-Directed Source Learning）识别尚未完全理解的内容，并自适应地重新访问该来源，而任务引导源学习（Task-Guided Source Learning）则利用下游经验揭示如何组织源知识时的局部表征差距和重复需求。在这两种情况下，学习信号决定了需要重新考虑的内容，而持久更新则是从权威来源重构而来。在五个基准测试和三个LLM后端中，SourceLearn在15个设置中的13个中实现了最佳性能，相较于Hybrid RAG提升了多达22.6分，并在静态源表征和基于经验的记忆基线方面有了显著的整体改善。
cs.CL / 109 / 2610.02163
AutoCompact: Learning When to Compact Context in Long-Horizon Coding Agents
AutoCompact：学习在长时间编码代理中何时进行上下文压缩
Abstract
Coding agents solve repository-level software engineering tasks through long trajectories of code inspection, search, editing, and testing. As a task progresses, earlier exploration becomes stale, so managing context is more than avoiding overflow: an agent must decide when to compact, what working state to preserve, and how to continue from it. We introduce AutoCompact, which trains a coding agent to make these decisions as part of its policy. To collect training data, we run the base agent on coding tasks and use a judge to review its compaction decisions, summaries, and actions after compaction. Flawed outputs are replaced with corrected ones before being executed in the environment, so each trajectory continues from the corrected decisions. We use these trajectories for supervised fine-tuning, then jointly optimize coding and compaction through reinforcement learning with task-success rewards. Experiments on SWE-bench Verified and SWE-PolyBench Verified show that AutoCompact improves pass rates over the base model by an absolute 9.2\% and 5.0\%, respectively. The improvements hold across all evaluated inference budgets, with a 256K context window that never overflows and with a 16K window whose overflow triggers fallback compaction.
Chinese Translation
编码代理通过长时间的代码检查、搜索、编辑和测试来解决库级软件工程任务。随着任务的进展，早期的探索变得陈旧，因此管理上下文不仅仅是避免溢出：代理必须决定何时进行压缩、保留什么工作状态以及如何从中继续。我们引入了AutoCompact，它训练编码代理在其策略中做出这些决策。为了收集训练数据，我们在编码任务上运行基础代理，并使用评审者审查其压缩决策、摘要和压缩后的行动。错误的输出在环境中执行之前被更正的输出替换，因此每个轨迹都从更正的决策继续。我们使用这些轨迹进行监督微调，然后通过强化学习与任务成功奖励共同优化编码和压缩。在SWE-bench Verified和SWE-PolyBench Verified上的实验表明，AutoCompact分别将通过率提高了绝对9.2\%和5.0\%。这些改进在所有评估的推理预算中均保持有效，256K的上下文窗口从未溢出，而16K的窗口在溢出时触发回退压缩。
cs.CL / 110 / 2610.02193
Hierarchical Continuous Diffusion Language Models
层次连续扩散语言模型
Abstract
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/.
Chinese Translation
离散扩散语言模型为需要双向推理和全局约束满足的任务提供了一种引人注目的替代方案。然而，它们存在一个结构性瓶颈：在并行解码时，每个标记都是独立从其边际分布中采样的，这切断了同时解码的标记之间的统计依赖关系。连续扩散语言模型通过去噪共享的连续状态来避免这一问题，但它们的去噪器仅看到该状态，因此在最终解码之前没有任何东西将其与有效的标记配置联系起来。为了解决这个问题，我们提出了层次连续扩散语言模型（HC-DLM），它将离散标记生成与单一原则性去噪过程中的连续潜在轨迹相结合，其训练目标源自对标记似然性的变分界限。与最近将连续上下文附加到自包含离散链的方法不同，HC-DLM使潜在状态成为唯一持久的生成状态：标记在每一步都从中读取，并作为下一个潜在更新的支架反馈。 在结构化推理（数独）、数学规划（倒计时）和语言建模（LM1B）任务中，HC-DLM在匹配模型大小的情况下，改善了离散和连续扩散基线，在数独和倒计时的拼图准确性以及LM1B的生成困惑度上均有所提升。项目页面：https://hc-dlm.github.io/
cs.CL / 111 / 2610.02206
KaliBench: A Fine-Grained Benchmark for Cybersecurity Tool Use on Kali Linux with Runtime-Free Verifiable Rewards
KaliBench：针对Kali Linux上网络安全工具使用的细粒度基准测试，具有无运行时可验证奖励
Abstract
LLMs are increasingly applied to cybersecurity workflows, where they are expected to translate analysts' intent into tool invocations. However, existing evaluations focus on knowledge-based assessments or end-to-end agentic tasks, and do not directly measure LLMs' ability to generate executable commands for real-world cybersecurity tools. This gap is critical because cybersecurity operations rely on strict command-line interfaces (CLIs), where minor syntax errors, incorrect flag--value bindings, or argument misordering can invalidate execution. We introduce KaliBench, a fine-grained benchmark and dataset for natural-language--to--CLI translation on Kali Linux, comprising 8,504 query--command pairs spanning 1,642 tools across 23 capability dimensions and 5 security phases. KaliBench is constructed via a manuscript-grounded pipeline with deterministic canonicalization and alias-aware evaluation, enabling precise and reproducible assessment of tool selection and argument construction. To ensure both semantic correctness and practical executability, we develop a multi-stage verification pipeline that combines LLM-based validation, sandboxed terminal execution, and human-in-the-loop refinement. Building on these fine-grained, deterministic signals, KaliBench further enables runtime-free verifiable rewards for training. Across three evaluation modes and 24 configurations of general-purpose and security-focused open-weight models, no open-weight model exceeds 42% exact-command accuracy in the unrestricted setting, highlighting the difficulty of accurate CLI-based cybersecurity tool use without explicit tool hints. We further show that supervised fine-tuning and reinforcement learning with verifiable rewards derived from KaliBench significantly improve an 8B model and achieve performance comparable to a 685B MoE model.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于网络安全工作流程中，它们被期望将分析师的意图转化为工具调用。然而，现有的评估主要集中在基于知识的评估或端到端的代理任务上，并未直接测量LLMs生成可执行命令的能力，以用于真实世界的网络安全工具。这一差距至关重要，因为网络安全操作依赖于严格的命令行接口（CLIs），其中微小的语法错误、不正确的标志-值绑定或参数顺序错误都可能导致执行无效。我们介绍了KaliBench，这是一个针对Kali Linux的自然语言到CLI翻译的细粒度基准测试和数据集，包含8,504个查询-命令对，涵盖23个能力维度和1,642个工具，跨越5个安全阶段。KaliBench通过一个基于手稿的管道构建，采用确定性标准化和别名感知评估，能够精确且可重复地评估工具选择和参数构建。为了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管道，结合了基于LLM的验证、沙箱终端执行和人机协作的优化。在这些细粒度、确定性的信号基础上，KaliBench进一步支持无运行时可验证奖励的训练。在三种评估模式和24种通用及安全聚焦的开放权重模型配置中，没有任何开放权重模型在无限制设置下超过42%的精确命令准确率，突显了在没有明确工具提示的情况下，准确使用基于CLI的网络安全工具的难度。我们进一步展示了利用KaliBench派生的可验证奖励进行监督微调和强化学习显著提高了一个8B模型的性能，并达到了与685B MoE模型相当的表现。