cs.RO / 1 / 2610.02274
Awomo-SimDataEngine: Agentic Simulation-ReadyWorld Generation
Awomo-SimDataEngine：代理仿真准备世界生成
PhysicalRSI Team, Ma, Danjiao, Ma, Enhui, Liu, Haohan, Jia, Heng, Shan, Hui, Xu, Jianhua, Zhang, Jiahuan, Xu, Jiangdi, Guo, Kaiwen, Yu, Kaicheng, Zhang, Linwei, Jin, Liyang, Luo, Maochun, Niu, Pengyao, Li, Shiwen, Feng, Shuangyu, Zhang, Tong, Wang, Tianheng, Wang, Xin, Huang, Xiangru, Huang, Yongqiang, Wang, Zhaozhi, Ma, Zijian
Abstract
Generating useful robot-training data requires more than visually plausiblescenes: objects must support interaction, placements must remain physicallyvalid, and tasks must admit repeatable execution. We present\textbf{Awomo-SimDataEngine}, an agentic system that connects asset and scenegeneration to robot demonstration synthesis. Shared asset services providerigid and articulated objects, including structure-grounded part and jointgeneration with ISArt. Scene generation supports two complementary routes:Unravel reconstructs editable scenes from images, while SimForge buildssingle-room and multi-room environments from text. A graph-native harnesscoordinates construction, validation, andbounded repair, routing failures to the responsible module while retainingunaffected scene state. PolicyForge binds validated worlds to tasks and robotembodiments to produce replayable demonstrations. Evaluations cover assetgeometry, scene quality, and downstream policy learning. On MuJoCo-basedLIBERO-Plus, co-training with Isaac Sim demonstrations improves the overallsuccess rate of a World-Action Model (WAM) from $77.17\%$ to $89.43\%$. Goal and spatialsuccess improve by $31.66$ and $6.25$ percentage points, respectively.These results support the utility of the generated data for cross-simulatorpolicy training, with more limited gains on long-horizon tasks.
Chinese Translation
生成有用的机器人训练数据不仅需要视觉上可信的场景：物体必须支持交互，放置必须保持物理有效性，任务必须允许可重复执行。我们提出了 extbf{Awomo-SimDataEngine}，这是一个将资产和场景生成与机器人演示合成连接起来的代理系统。共享资产服务提供刚性和关节物体，包括与ISArt一起生成的结构基础部件和关节。场景生成支持两条互补路径：Unravel从图像中重建可编辑场景，而SimForge从文本构建单房间和多房间环境。一个图形原生的框架协调构建、验证和有限修复，将失败路由到负责的模块，同时保留未受影响的场景状态。PolicyForge将验证过的世界与任务和机器人体现绑定，以生成可重放的演示。评估涵盖资产几何、场景质量和下游策略学习。在基于MuJoCo的LIBERO-Plus上，与Isaac Sim演示的共同训练将世界-动作模型（World-Action Model, WAM）的整体成功率从$77.17\%$提高到$89.43\\%$。目标和空间成功分别提高了$31.66$和$6.25$个百分点。这些结果支持生成数据在跨模拟器策略训练中的实用性，但在长时间任务上的增益较为有限。
cs.RO / 2 / 2610.02323
World-Calibrated Proposal-to-Action Flow for Vision-Language-Action Models
世界校准的行动提案流用于视觉-语言-行动模型
Abstract
Flow-based Vision-Language-Action (VLA) policies generate action chunks by transporting samples from a task-agnostic isotropic Gaussian source. As this source is conditioned on neither recent execution nor predicted future evolution, (i) it discards the local continuity established by recently executed motion. (ii) Even when predictive world representations are introduced, they often only condition the transport dynamics rather than determine where generation starts, how far it may deviate, or along which action directions it may expand. Building on this observation, we introduce ProAct, a world-calibrated proposal-to-action framework that makes the generative source itself predictable. (i) To preserve motion continuity, a lightweight Proposal Expert converts recent actions into a scene-aware hypothesis via one motion-anchored endpoint flow-matching step, initializing generation near the demonstrated action manifold. (ii) To jointly capture intended scene evolution and proposal-future compatibility, a prospective World Expert treats the hypothesis as a soft motion prior while predicting the task-consistent latent future. (iii) From this compatibility, the model calibrates a proposal-centered anisotropic source, where a bounded per-step extent controls the allowed deviation and a trace-normalized low-rank geometry under a condition-number budget allocates refinement over coupled translation, rotation, and gripper directions. Compared with $\pi_{0.5}$, ProAct improves performance across simulation and real-world tasks while reducing denoising steps by 50%, inference latency by up to 25.8%, and increasing throughput by up to 34.8%.
Chinese Translation
基于流的视觉-语言-行动（VLA）策略通过从任务无关的各向同性高斯源传输样本来生成动作块。由于该源既不依赖于最近的执行也不依赖于预测的未来演变，(i) 它丢弃了最近执行的运动所建立的局部连续性。(ii) 即使引入了预测的世界表示，它们通常仅仅调节传输动态，而不是决定生成的起始位置、可能的偏离程度或扩展的动作方向。基于这一观察，我们引入了ProAct，一个世界校准的提案到行动框架，使生成源本身变得可预测。(i) 为了保持运动的连续性，一个轻量级的提案专家通过一个运动锚定的端点流匹配步骤将最近的动作转换为场景感知的假设，从而在示范的动作流形附近初始化生成。(ii) 为了共同捕捉预期的场景演变和提案-未来的兼容性，一个前瞻性的世界专家将假设视为软运动先验，同时预测与任务一致的潜在未来。(iii) 基于这种兼容性，模型校准了一个以提案为中心的各向异性源，其中每步的受限范围控制允许的偏离，而在条件数预算下的迹归一化低秩几何分配了在耦合的平移、旋转和夹持方向上的细化。与$ ext{π}_{0.5}$相比，ProAct在模拟和现实世界任务中提高了性能，同时减少了去噪步骤50%，推理延迟最多减少25.8%，并将吞吐量提高最多34.8%。
cs.RO / 3 / 2610.02338
SoTa: Soft Tactile Skins for Dexterous Manipulation
SoTa：用于灵巧操作的软触觉皮肤
Abstract
A growing body of work suggests that tactile sensing gives robot policies contact information that complements vision in dexterous manipulation. However, visuo-tactile robot data remains scarce: dexterous demonstrations require teleoperating robots, which limits dataset scale. Human demonstrations are far cheaper to collect and offer a path to scale this data, but only if human and robot hands carry tactile sensors with corresponding signals. This requires sensors that conform to different hand geometries, cover the full hand, and share a common layout across embodiments. We present SoTa, a low-cost capacitive tactile skin that provides full-hand coverage on humans and robots while preserving a shared layout of 202 taxels across corresponding finger and palm regions. Our multilayer design with fabric electrodes enables in-house fabrication of thin, soft skins with customizable geometry for under $10 in materials per skin. The sensor retains over 97% of its initial response span after 10,000 loading-unloading cycles with traces retaining continuity through 1,280 tight-fist folding cycles. The shared taxel layout supports human-robot co-training with a common tactile encoder and no learned cross-sensor mapping. Across three contact-rich manipulation tasks, tactile observations improve in-distribution success over vision-only policies. With a fixed robot demonstration budget, adding human demonstrations more than doubles mean success across eight evaluation conditions, from 22.8% to 45.9%, improving success in all five out-of-distribution conditions. We plan to open-source the resources needed to fabricate and operate these skins.
Chinese Translation
越来越多的研究表明，触觉感知为机器人策略提供了补充视觉的接触信息，从而在灵巧操作中发挥重要作用。然而，视觉-触觉机器人数据仍然稀缺：灵巧演示需要遥控操作机器人，这限制了数据集的规模。人类演示的收集成本更低，并且提供了扩大数据规模的途径，但前提是人类和机器人手部都配备有相应信号的触觉传感器。这需要符合不同手部几何形状、覆盖整个手部并在不同实现中共享相同布局的传感器。我们提出了SoTa，一种低成本的电容触觉皮肤，能够在人体和机器人上提供全手覆盖，同时在相应的手指和掌心区域保留202个触觉单元（taxels）的共享布局。我们的多层设计结合了织物电极，使得每个皮肤的材料成本低于10美元，能够在内部制造出薄而柔软的皮肤，且几何形状可定制。该传感器在经历10,000次加载-卸载循环后仍保持超过97%的初始响应范围，并且在1,280次紧握折叠循环中保持信号连续性。共享的触觉单元布局支持人类与机器人共同训练，使用相同的触觉编码器且无需学习跨传感器映射。在三个接触丰富的操作任务中，触觉观察在分布内的成功率相较于仅使用视觉的策略有所提高。在固定的机器人演示预算下，增加人类演示使得八个评估条件下的平均成功率从22.8%提升至45.9%，在所有五个分布外条件下均有所改善。我们计划开源制造和操作这些皮肤所需的资源。
cs.RO / 4 / 2610.02339
NEEDLEWORK: Offline Rewriting of Robot Data with Verified Local Stitches
NEEDLEWORK：具有验证局部缝合的机器人数据离线重写
Abstract
Robot demonstrations may contain useful behavior even when individual episodes are inefficient or unsuccessful. Trajectory stitching offers a way to compose these behaviors into improved training data, but identifying useful connections and verifying their feasibility is difficult in high-dimensional robot data, where many prior methods rely on low-dimensional state representations. We introduce NEEDLE, an offline dataset-augmentation algorithm that addresses these challenges by adding short, verified action bridges between recorded observations in high-dimensional robot demonstrations. First, NEEDLE identifies and creates connections that bypass suboptimal detours, broaden action coverage, and augment the original dataset with failed trajectories, using only RGB images, proprioception, and episode-level outcomes, without new environment interaction or privileged object state. Next, we present a sampling technique that incorporates accepted bridges into policy training without synthesizing intermediate images or discarding the original demonstrations, allowing policies to learn alternative actions while retaining the original dataset's coverage. On real-robot tasks, NEEDLE improves success rate over the strongest baseline on each task by an average of 21 percentage points. Videos and supplementary materials are on https://needle-work.github.io/.
Chinese Translation
机器人演示可能包含有用的行为，即使个别情节效率低下或不成功。轨迹缝合提供了一种将这些行为组合成改进训练数据的方法，但在高维机器人数据中识别有用的连接并验证其可行性是困难的，许多先前的方法依赖于低维状态表示。我们提出了NEEDLE，这是一种离线数据集增强算法，通过在高维机器人演示中的记录观察之间添加短的、经过验证的动作桥接来解决这些挑战。首先，NEEDLE识别并创建绕过次优绕道的连接，扩大动作覆盖范围，并使用仅RGB图像、本体感觉和情节级结果来增强原始数据集，且无需新的环境交互或特权对象状态。接下来，我们提出了一种采样技术，将接受的桥接纳入策略训练，而无需合成中间图像或丢弃原始演示，从而允许策略在保留原始数据集覆盖范围的同时学习替代动作。在真实机器人任务中，NEEDLE在每个任务上平均提高了21个百分点的成功率，超越了最强基线。视频和补充材料可在 https://needle-work.github.io/ 上查看。
cs.RO / 5 / 2610.02341
Filter-Aware Fine-Tuning for Safe Humanoid Whole-Body Tracking
考虑过滤器的安全人形机器人全身跟踪微调
Abstract
Safe whole-body motion is essential for deploying humanoid robots in unstructured environments. Modern humanoid control commonly separates reference specification from execution, with a planner, teleoperator, or motion generator providing a reference that a reinforcement-learning policy tracks through dynamically feasible whole-body control. Runtime safety filters, such as control barrier functions (CBFs), offer a promising approach for enforcing newly introduced constraints via interventions on the tracker's outputs. We show, however, that treating the tracking policy and safety filter independently induces fundamental mismatches, as filtering alters both the executed actions and the induced state distribution. We study this policy-filter interface through case studies that isolate dynamics, objective, and information mismatches, highlight their root causes, and use these insights to develop CoFiT (Constrained Filter-aware Tuning), a filter-aware fine-tuning method for pretrained trackers. Across diverse constraint scenes, CoFiT reduces violation time relative to filter-only training by 91% on TWIST2 and 21% on SONIC, while requiring smaller safety filter corrections. On Unitree G1 hardware, CoFiT reduces violation time by 83% for TWIST2 and completes every trial without operator intervention, whereas 50% of baseline trials require an operator stop. Together, these results provide actionable insights into policy-filter interactions and establish design principles for integrating learned trackers with runtime safety filters.
Chinese Translation
安全的全身运动对于在非结构化环境中部署人形机器人至关重要。现代人形控制通常将参考规范与执行分开，由规划器、遥控操作员或运动生成器提供参考，强化学习策略通过动态可行的全身控制进行跟踪。然而，我们发现将跟踪策略和安全过滤器独立处理会导致根本的不匹配，因为过滤会改变执行的动作和引发的状态分布。我们通过案例研究研究这一策略-过滤器接口，隔离动态、目标和信息的不匹配，突出其根本原因，并利用这些见解开发了CoFiT（考虑过滤器的约束微调），这是一种针对预训练跟踪器的过滤器感知微调方法。在多样的约束场景中，CoFiT在TWIST2上相对于仅过滤器训练减少了91%的违规时间，在SONIC上减少了21%，同时需要更小的安全过滤器修正。在Unitree G1硬件上，CoFiT在TWIST2上将违规时间减少了83%，并在每次试验中都无需操作员干预，而基线试验中有50%需要操作员停止。综合来看，这些结果为策略-过滤器交互提供了可操作的见解，并建立了将学习到的跟踪器与运行时安全过滤器集成的设计原则。
cs.RO / 6 / 2610.02360
SocialVLA: A Social Perception Gateway for Human-Reaction-Based Failure Detection and Recovery in VLA Manipulation
SocialVLA：一种基于人类反应的故障检测与恢复的社交感知网关在视觉-语言-动作（VLA）操作中的应用
Abstract
Vision-language-action (VLA) policies enable diverse robotic manipulation but can fail during execution without recognizing their own errors. Human observers provide complementary signals, as unexpected robot behavior can trigger rapid vocal, facial, or verbal reactions before failure is completed. We introduce SocialVLA, a local, policy-agnostic social perception gateway that converts spontaneous human reactions into runtime intervention signals for VLA manipulation. SocialVLA combines causal paralinguistic audio detection, visual reaction recognition, explicit stop phrases, and robot-relevance estimation. An asynchronous first-event fusion mechanism triggers a VLA hold from the earliest sufficiently confident signal, while a separate speech channel captures verbal corrections for participant-directed continuation, restart, or instruction revision. We evaluate SocialVLA on physical Unitree G1 manipulation using 15 participants, with 238 annotated intervention-worthy episodes and 1.038 h of non-intervention behavior. Frozen offline replay achieves 54.6% recall and 69.5% precision, while unfiltered audio-video fusion reaches 64.3% recall. Relevance estimation reduces false-stop episodes from 100 to 57 and increases precision from 60.5% to 69.8%. In prospective deployment on an unseen 16th participant, the frozen system achieves 59.5% recall and 91.7% precision. Median detector-to-fusion latency is 47.9 ms, VLA-gate-to-physical-hold latency is 336 ms, and reaction-onset-to-hold latency is 1.021 s. These results demonstrate a complete local pathway from spontaneous social reaction to physical VLA interruption and participant-directed recovery.
Chinese Translation
视觉-语言-动作（VLA）策略使得多样化的机器人操作成为可能，但在执行过程中可能会因未能识别自身错误而失败。人类观察者提供了补充信号，因为意外的机器人行为可以在故障完成之前触发快速的声音、面部或语言反应。我们提出了SocialVLA，这是一种局部的、与策略无关的社交感知网关，能够将自发的人类反应转化为VLA操作的运行时干预信号。SocialVLA结合了因果性副语言音频检测、视觉反应识别、明确的停止短语和机器人相关性估计。异步的首次事件融合机制从最早的足够自信信号触发VLA保持，而一个独立的语音通道则捕捉参与者导向的继续、重启或指令修订的语言修正。我们在15名参与者的物理Unitree G1操作中评估了SocialVLA，共有238个标注的干预事件和1.038小时的非干预行为。冻结的离线重放实现了54.6%的召回率和69.5%的精确率，而未过滤的音频-视频融合达到了64.3%的召回率。相关性估计将错误停止事件从100减少到57，并将精确率从60.5%提高到69.8%。在对未见过的第16名参与者的前瞻性部署中，冻结系统实现了59.5%的召回率和91.7%的精确率。中位检测到融合延迟为47.9毫秒，VLA门到物理保持的延迟为336毫秒，反应开始到保持的延迟为1.021秒。这些结果展示了从自发社交反应到物理VLA中断及参与者导向恢复的完整局部路径。
cs.RO / 7 / 2610.02368
Rethinking World-Action Model for Compositional and In-Context Robotic Manipulation
重新思考用于组合和上下文机器人操作的世界-动作模型
Abstract
Long-horizon compositional manipulation has become increasingly important for real-world robot deployment, where a single task involves multiple coordinated subtasks. Existing world-action models (WAMs) jointly predict short-horizon visual futures and actions, but typically lack explicit subtask-level reasoning. We propose Visual Goal-conditioned Action Reasoning (ViGAR), a hierarchical framework that factorizes manipulation into a visual subgoal planner and a subgoal executor. Given the current observation and global instruction, the subgoal planner predicts a visual subgoal for the next subtask. The subgoal executor then jointly generates future visual trajectories and actions conditioned on the predicted subgoal. Both components share a pretrained world-model representation, enabling task-level planning and action generation to benefit from common physical knowledge. Moreover, our framework naturally supports in-context learning: using a global goal image as context can induce different subtask decompositions and behaviors without parameter updates. On the RoboTwin Clean2Random benchmark, ViGAR achieves 82.00% and 67.02% success rates under the Clean and Random settings, respectively, surpassing the strongest baseline by 12.86 percentage points in average success rate. Real-world robot experiments on five compositional and two in-context learning tasks further confirm the effectiveness of ViGAR.
Chinese Translation
长时间范围的组合操作在现实世界机器人部署中变得越来越重要，其中单个任务涉及多个协调的子任务。现有的世界-动作模型（WAMs）共同预测短时间范围的视觉未来和动作，但通常缺乏明确的子任务级推理。我们提出了视觉目标条件的动作推理（ViGAR），这是一个分层框架，将操作分解为视觉子目标规划器和子目标执行器。在给定当前观察和全局指令的情况下，子目标规划器预测下一个子任务的视觉子目标。然后，子目标执行器在预测的子目标条件下共同生成未来的视觉轨迹和动作。这两个组件共享一个预训练的世界模型表示，使得任务级规划和动作生成能够受益于共同的物理知识。此外，我们的框架自然支持上下文学习：使用全局目标图像作为上下文可以在不更新参数的情况下引导不同的子任务分解和行为。在RoboTwin Clean2Random基准测试中，ViGAR在Clean和Random设置下分别达到了82.00%和67.02%的成功率，平均成功率比最强基线高出12.86个百分点。在五个组合任务和两个上下文学习任务上的现实世界机器人实验进一步验证了ViGAR的有效性。
cs.RO / 8 / 2610.02398
Keep the Effect, Drop the Actor: Programmable Effect-to-Execution World-Action Models
保持效果，去除执行者：可编程效果到执行的世界-动作模型
Abstract
A robot demonstration records two things in the same frames: what happened to the objects, and how one particular arm made it happen. We condition on the first. A demonstration is compiled into an effect program: the 3D keypoint trajectories of the objects that moved, two points marking where each was held, and the configuration the scene ends in, with the demonstrator removed. PEWAM, a 71.5M-parameter world-action model, generates effect, robot execution, action and terminal state as four streams with independent flow-matching times, so clamping a program and sampling the execution turns inference into programming, re-solved closed loop from the live scene. On held-out LIBERO-Goal tasks, one demonstration's program completes 40 of 90 episodes, where the same backbone given a goal image or language, and published demonstration-conditioned methods, complete at most 19; on three of Meta-World's held-out classes it exceeds the best published results, though not on the five-class mean. Because a program is a set of coordinates, a person can edit it: the placement follows a shifted terminal state and the grasp turns with rotated contact points. The same program runs on four robot arms without retraining, and after a push, re-solving completes 23 of 60 episodes where replaying the demonstration completes 6. On a Franka arm, fine-tuned on real demonstrations of other tasks, programs compiled from single human videos complete 36 of 40 trials, against 22 for the same backbone conditioned on the video's last frame as a goal image.
Chinese Translation
机器人演示在同一帧中记录了两件事：物体发生了什么，以及一个特定的手臂是如何实现的。我们以第一项为条件。演示被编译成一个效果程序：移动物体的三维关键点轨迹，标记每个物体被抓取的两个点，以及场景结束时的配置，演示者被移除。PEWAM（Programmable Effect-to-Execution World-Action Model），一个拥有7150万个参数的世界-动作模型，生成效果、机器人执行、动作和终态作为四个流，每个流具有独立的流匹配时间，因此固定一个程序并采样执行将推理转变为编程，从实时场景中重新解决闭环。在保留的LIBERO-Goal任务中，一个演示的程序完成了90个回合中的40个，而同样的骨干网络在给定目标图像或语言以及已发布的演示条件方法时，最多只完成19个；在Meta-World的三个保留类别中，它超过了已发布的最佳结果，尽管在五类平均值上没有超越。由于程序是一组坐标，个人可以对其进行编辑：放置遵循一个偏移的终态，抓取点随着旋转的接触点而变化。同一程序可以在四个机器人手臂上运行而无需重新训练，并且在一次推动后，重新解决完成了60个回合中的23个，而重放演示仅完成6个。在Franka手臂上，经过其他任务的真实演示微调，从单个人类视频编译的程序完成了40个试验中的36个，而同样的骨干网络在将视频的最后一帧作为目标图像时仅完成22个。
cs.RO / 9 / 2610.02428
Autonomous mobile robot operations logistics: a dataset of jobs, dispatch events and robot states
自主移动机器人操作物流：作业、调度事件和机器人状态的数据集
Abstract
Autonomous mobile robots (AMRs) increasingly perform material transport in production logistics, where their operation is governed by job generation, dispatching and robot control. We present MoRoOp, a dataset of AMR operations recorded in a laboratory kit preparation and supply scenario over nine eight-hour shifts. During each shift, an AMR executed stochastically generated kit supply, empty-box refill and charging jobs. The dataset links job specifications, the operations constituting each job, dispatch events documenting operation state transitions and outcomes, and robot-state observations comprising position, orientation, velocity, per-wheel state of charge and diagnostics. It contains 1,382 jobs, 4,815 operations, 19,352 dispatch events and 140,386 robot-state observations together with the kit specifications used during job generation. The dataset was recorded in an operating laboratory environment, and technically valid observations of delays, obstructed navigation and unsuccessful operations were retained. Both raw and cleaned robot-state tables are provided. Documented reuse directions include the evaluation of AI agents on operational decision records, disturbance detection, operation prediction, data-driven simulation and event-log analysis.
Chinese Translation
自主移动机器人（AMRs）在生产物流中越来越多地执行物料运输，其操作受作业生成、调度和机器人控制的影响。我们提出了MoRoOp，这是一个在实验室套件准备和供应场景中记录的AMR操作数据集，涵盖了九个八小时的班次。在每个班次中，AMR执行了随机生成的套件供应、空箱补充和充电作业。该数据集链接了作业规范、构成每个作业的操作、记录操作状态转变和结果的调度事件，以及包含位置、方向、速度、每个轮子的电量状态和诊断信息的机器人状态观察。数据集中包含1,382个作业、4,815个操作、19,352个调度事件和140,386个机器人状态观察，以及在作业生成过程中使用的套件规范。该数据集是在一个运行的实验室环境中记录的，并保留了技术上有效的延迟、导航受阻和操作失败的观察。提供了原始和清理后的机器人状态表。文档中记录的重用方向包括对操作决策记录的AI代理评估、干扰检测、操作预测、数据驱动的仿真和事件日志分析。
cs.RO / 10 / 2610.02459
OpenRUA: Robot-Use Agents Are Zero-Shot Visuomotor Policies
OpenRUA：机器人使用代理是零样本视觉运动策略
Abstract
Coding agents are extending their reach into the physical world by writing and executing robot control programs. One might expect the agents to use the existing mature software stack that engineers have developed over decades to access sensors and control motion. Yet prior work primarily engineers complex custom harnesses to orchestrate agents for robot use, particularly by prescribing specialized workflows and providing bespoke interfaces. This raises the question: "Is such additional harness engineering necessary?" We introduce OpenRUA, a zero-abstraction harness that bypasses bespoke abstraction layers by providing off-the-shelf coding agents with only terminal access to the robot's native software interface ROS 2. OpenRUA employs a minimalist workspace-as-harness design, only offering ROS 2 documentation and basic tools while leaving the coding agent to organize its own work without orchestrating any agentic workflow. Within this workspace, OpenRUA recasts perception as file I/O and manipulation as coding. With Claude Code powered by Claude Opus 5, OpenRUA achieves success rates of 99.0% on CaP-Bench and 87.0% on LIBERO-PRO, demonstrating that an off-the-shelf coding agent can serve as a zero-shot visuomotor policy through the robot's native interface, without bespoke primitives or task-specific training. Under this minimalist design, further analysis reveals striking emergent behaviors of coding agents: (1) For perception, the agent spontaneously writes programs that process raw sensory inputs and derive metric measurements in 96.80% of episodes. (2) For manipulation, the agent spontaneously builds motion-control clients (e.g., gripper control) in 95.87% of episodes and closed-loop control programs (e.g., adjusting motion based on sensor feedback) in 50.13% of episodes. Our code is available at https://github.com/terminalworld/OpenRUA.
Chinese Translation
编码代理通过编写和执行机器人控制程序，正在扩展其在物理世界中的应用。人们可能会期望这些代理使用工程师数十年来开发的成熟软件栈来访问传感器和控制运动。然而，先前的工作主要是设计复杂的定制工具，以协调代理进行机器人使用，特别是通过规定专门的工作流程和提供定制接口。这引发了一个问题：“这样的额外工具工程是否必要？”我们介绍了OpenRUA，这是一种零抽象工具，通过仅提供终端访问机器人本地软件接口ROS 2，绕过了定制的抽象层。OpenRUA采用极简的工作空间作为工具设计，仅提供ROS 2文档和基本工具，同时让编码代理自行组织工作，而无需协调任何代理工作流程。在这个工作空间中，OpenRUA将感知重新定义为文件输入输出，将操作重新定义为编码。在Claude Opus 5驱动的Claude Code支持下，OpenRUA在CaP-Bench上实现了99.0%的成功率，在LIBERO-PRO上实现了87.0%的成功率，证明了一个现成的编码代理可以通过机器人的本地接口作为零样本视觉运动策略，而无需定制原语或任务特定的训练。在这种极简设计下，进一步分析揭示了编码代理的显著涌现行为：(1) 在感知方面，代理在96.80%的实验中自发编写处理原始传感器输入并推导度量测量的程序。(2) 在操作方面，代理在95.87%的实验中自发构建运动控制客户端（例如，抓手控制），在50.13%的实验中构建闭环控制程序（例如，根据传感器反馈调整运动）。我们的代码可在https://github.com/terminalworld/OpenRUA获取。
cs.RO / 11 / 2610.02469
RUL-Aware RRT*: Degradation-Balanced Motion Planning for Robotic Manipulators
考虑剩余使用寿命的 RRT*: 机器人手臂的退化平衡运动规划
Abstract
Robotic manipulators operating over long durations often experience uneven joint degradation, which causes the weakest actuator to fail prematurely, leads to unplanned downtime, and results in significant operational losses. Traditional motion-planning algorithms do not account for joint health conditions and therefore tend to exacerbate this imbalance during extended operation. To address this challenge, this study introduces the RUL-aware RRT*, a motion-planning method that incorporates joint remaining useful life information into the planning process and adaptively adjusts joint usage in response to evolving health conditions. The method is evaluated across three representative scenarios, namely the Full Health Scenario (FHS), the Heterogeneous Degradation Scenario (HDS), and the Local Degradation Scenario (LDS). The results show that the RUL-aware RRT* effectively suppresses degradation imbalance, delays the emergence of bottleneck failures, and improves the long-term reliability of the robotic system during extended operation. These findings demonstrate that integrating health feedback into motion planning provides a practical and robust pathway for enhancing the durability and operational resilience of manipulators subject to continuous wear.
Chinese Translation
长时间运行的机器人手臂往往会经历不均匀的关节退化，这导致最弱的执行器过早失效，造成计划外的停机，并导致显著的运营损失。传统的运动规划算法并未考虑关节健康状况，因此在延长操作过程中往往加剧了这种不平衡。为了解决这一挑战，本研究提出了考虑剩余使用寿命（RUL）的 RRT* 方法，该方法将关节剩余使用寿命信息纳入规划过程，并根据不断变化的健康状况自适应地调整关节使用。该方法在三个典型场景中进行了评估，即完全健康场景（FHS）、异质退化场景（HDS）和局部退化场景（LDS）。结果表明，考虑 RUL 的 RRT* 有效抑制了退化不平衡，延迟了瓶颈故障的出现，并提高了机器人系统在延长操作过程中的长期可靠性。这些发现表明，将健康反馈整合到运动规划中，为增强受持续磨损影响的机械手的耐用性和操作弹性提供了一条切实可行且稳健的途径。
cs.RO / 12 / 2610.02527
CriticHack: Evaluating Visual Rewards Under Robot Policy Optimization
CriticHack：在机器人策略优化下评估视觉奖励
Abstract
Learned visual reward models are increasingly used to optimize robot policies, yet a reward model can score an execution that acts on the wrong object as highly as one that completes the task. We show that optimizing such a reward can amplify these wrong-object failures while reward and task success both rise, so the signals a practitioner would normally monitor look healthy. We fine-tune every denoiser parameter of a diffusion policy against Robometer on a drawer task. Starting from a supervised policy with no prior reward exposure, five training runs raise task success by 10.2 percentage points and wrong-object failures by 10.9 points on 512 evaluation seeds, whereas five runs trained on the simulator's task-completion signal raise success without amplifying wrong-object failures (difference 9.2 points, 95% CI 5.6 to 13.0). The amplification recurs from a policy previously optimized against learned rewards, under the policy's native diffusion sampler, at matched distance from the initial policy, and across constrained-policy experiments with two critics and two optimizers. A tilt model explains when it occurs: under KL-regularized optimization, an outcome becomes more frequent whenever its expected reward under the initial policy exceeds the population average. Robometer separates successes from failures well overall (AUROC .81) but scores wrong-object failures slightly above successes (AUROC .37), so optimization raises both. The same model predicts the outcome shifts across 26 constrained settings (Spearman .89), including those in which task success falls, and Robometer's own published success-termination recipe inherits the error. A frozen outcome verifier redirects the same optimization toward the requested task.
Chinese Translation
学习的视觉奖励模型越来越多地用于优化机器人策略，但奖励模型可能对错误对象的执行评分与对完成任务的执行评分同样高。我们展示了优化这样的奖励可能会放大这些错误对象的失败，同时奖励和任务成功率都在上升，因此从业者通常监测的信号看起来是健康的。我们对在抽屉任务中使用 Robometer 的扩散策略的每个去噪器参数进行了微调。从没有先前奖励暴露的监督策略开始，五次训练运行使任务成功率提高了 10.2 个百分点，而错误对象的失败率提高了 10.9 个百分点，评估种子为 512；而在模拟器的任务完成信号上进行的五次训练运行则提高了成功率，而没有放大错误对象的失败（差异为 9.2 个百分点，95% 置信区间 5.6 到 13.0）。这种放大效应在一个先前针对学习奖励优化的策略中重复出现，在策略的原生扩散采样器下，距离初始策略相匹配，并且在两个评论者和两个优化器的约束策略实验中也存在。倾斜模型解释了何时发生这种情况：在 KL 正则化优化下，当某个结果在初始策略下的期望奖励超过总体平均时，该结果变得更为频繁。总体而言，Robometer 能很好地区分成功与失败（AUROC .81），但对错误对象的失败评分略高于成功（AUROC .37），因此优化同时提高了两者。相同的模型预测了 26 个约束设置中的结果变化（Spearman .89），包括那些任务成功率下降的情况，而 Robometer 自身发布的成功终止方案也继承了这一错误。一个冻结的结果验证器将相同的优化重定向到请求的任务上。
cs.RO / 13 / 2610.02601
Real-time Event-camera Stereo Visual Odometry via Keytime Gaussian Process Regression
基于关键时间高斯过程回归的实时事件相机立体视觉里程计
Abstract
Event cameras have microsecond-level temporal resolution and high dynamic range which make them more resilient to motion blur and poor illumination than standard frame-based cameras. Event-camera visual odometry (VO) pipelines maximize these benefits when they process the asynchronous event stream at the native temporal resolution. Continuous-time Gaussian process (GP) regression and a white-noise-on-acceleration (WNOA) prior can handle asynchronous measurements but result in a prohibitively large estimation state when applied naively. This paper presents a continuous-time event-camera stereo VO pipeline that maintains the native measurement times of asynchronous events while also running in real time. It reduces the estimation states to keytimes while maintaining full temporal resolution by interpolating measurements to their exact timestamps with a physically founded WNOA prior. This decouples the state size from the dense number of measurements without discarding their asynchronous nature. The real-time continuous-time VO pipeline is evaluated on the MVSEC and DSEC datasets. It provides estimates in real time that are more accurate than ES-PTAM, a state-of-the-art discrete estimator, in all but one of the tested sequences. The pipeline respectively provides estimates at 22 Hz and 6 Hz on MVSEC and DSEC and RMS relative errors of 0.46 cm and 0.038 degrees across all valid sequences, which were 11 and 15 times better than ES-PTAM, respectively.
Chinese Translation
事件相机具有微秒级的时间分辨率和高动态范围，使其在运动模糊和光照不足的情况下比标准帧摄像机更具抗干扰能力。事件相机视觉里程计（VO）流程在以原生时间分辨率处理异步事件流时最大化了这些优势。连续时间高斯过程（GP）回归和加速度上的白噪声（WNOA）先验可以处理异步测量，但在简单应用时会导致估计状态过于庞大。本文提出了一种连续时间事件相机立体VO流程，该流程在保持异步事件的原生测量时间的同时，也能实现实时运行。通过使用物理基础的WNOA先验将测量插值到其精确时间戳，减少了估计状态到关键时间，同时保持了完整的时间分辨率。这将状态大小与大量测量的密集数量解耦，而不丢弃其异步特性。该实时连续时间VO流程在MVSEC和DSEC数据集上进行了评估。它在实时提供的估计在所有测试序列中都比ES-PTAM（一个最先进的离散估计器）更为准确，只有一个序列例外。该流程在MVSEC和DSEC上分别以22 Hz和6 Hz的频率提供估计，所有有效序列的均方根相对误差分别为0.46厘米和0.038度，分别比ES-PTAM好11倍和15倍。
cs.RO / 14 / 2610.02668
A Passive AI System for Verifying Physical State on Automated Liquid Handlers
一种被动人工智能系统用于验证自动液体处理器的物理状态
Abstract
Automated liquid handlers execute digital protocols, but operators must assemble the deck and confirm that the physical setup matches the intended experiment. We present the Labware Setup Checker, a passive vision system that verifies deck preparation on an Opentrons Flex without modifying its hardware or firmware. A consumer webcam mounted outside the robot supplies images to a browser application. The checker segments the deck into slots, corrects slot assignments using the known deck geometry, classifies the labware, and compares the observed setup with a reference protocol library to report missing, misplaced, or unexpected labware. It operates independently of the robot controller and does not interrupt or gate a run. During SwabSeq Respiratory Viral Panel (RVP) pre-PCR deck preparation, the checker tracked a slot from empty to a base and then to the completed plate-and-base assembly, identifying the completed assembly in all 29 frames before robot motion. Across these frames, 95.1% of the 348 labware classifications for slots requiring exact matches agreed with the protocol reference. In 480 frames with partially occluded reservoirs, all 6,720 slot assignments were correct. Across three camera positions and two lighting conditions, slot assignment was 100% correct, and 97.73% of labware classifications matched the protocol reference. Median verification time was 2.10 s on a laptop using only its CPU and 4.24 s from frame capture to receipt of results during AWS deployment in the laboratory. These results show that an external camera and independent software can provide verification of deck setup against protocol requirements within seconds, allowing operators to review the physical setup before starting a run.
Chinese Translation
自动液体处理器执行数字协议，但操作员必须组装工作台并确认物理设置与预期实验相符。我们提出了实验室设备设置检查器（Labware Setup Checker），这是一种被动视觉系统，可以在不修改硬件或固件的情况下验证Opentrons Flex的工作台准备情况。安装在机器人外部的消费级网络摄像头向浏览器应用程序提供图像。该检查器将工作台分割成多个插槽，利用已知的工作台几何形状修正插槽分配，分类实验室设备，并将观察到的设置与参考协议库进行比较，以报告缺失、放错位置或意外的实验室设备。它独立于机器人控制器运行，不会中断或限制运行。在SwabSeq呼吸道病毒面板（RVP）PCR前工作台准备过程中，检查器跟踪了一个插槽从空到基础，再到完成的板和基础组件的过程，在机器人运动之前识别出所有29帧中的完成组件。在这些帧中，对于需要精确匹配的348个实验室设备分类，95.1%与协议参考一致。在480帧中，部分遮挡的储液器的所有6720个插槽分配均正确。在三个摄像头位置和两种光照条件下，插槽分配100%正确，97.73%的实验室设备分类与协议参考匹配。中位验证时间为2.10秒，使用仅CPU的笔记本电脑进行验证，而在实验室的AWS部署中，从帧捕获到结果接收的时间为4.24秒。这些结果表明，外部摄像头和独立软件可以在几秒钟内提供工作台设置与协议要求的验证，使操作员能够在开始运行之前审查物理设置。
cs.RO / 15 / 2610.02691
DeltaWorld: Physically Consistent Interactive World Simulators via Action-Conditioned Latent Increment Learning
DeltaWorld：通过动作条件潜在增量学习实现物理一致的交互式世界模拟器
Abstract
Interactive world simulators can provide scalable environments for robot planning, policy training, and evaluation by predicting action consequences while reducing reliance on repeated physical rollouts. To serve these applications, they must generate future image sequences that respond faithfully to robot actions and preserve the dynamics of robot-object interactions over long horizons. However, existing world models typically predict the entire next latent state and often fail to capture subtle changes induced by robot actions. Such omissions can produce physically implausible outcomes, including object interpenetration and excessive deformation. To address this limitation, we propose DeltaWorld, a physically consistent interactive world simulator for robotic manipulation. Our method introduces the Delta Latent Transition Model (Delta-LTM), which predicts action-induced latent feature changes and adds them to the current latent state to obtain the next state, rather than predicting the next latent state directly. To mitigate object interpenetration and excessive deformation in predicted future frames, Interaction-aware Latent Alignment is introduced to construct counterfactual interaction regions and supervise interaction-related latent changes. DeltaWorld is evaluated on the IWS manipulation benchmark and a self-collected cross-robot dataset covering multiple robot embodiments and manipulation tasks. On the cross-robot dataset, DeltaWorld reduces FVD by 46.6% and LPIPS by 31.1% relative to the IWS baseline. These results highlight the potential of DeltaWorld for long-horizon action-conditioned video prediction in robotic manipulation.
Chinese Translation
交互式世界模拟器能够为机器人规划、策略训练和评估提供可扩展的环境，通过预测动作后果来减少对重复物理模拟的依赖。为了服务于这些应用，它们必须生成对机器人动作做出真实响应的未来图像序列，并在长时间范围内保持机器人与物体交互的动态特性。然而，现有的世界模型通常预测整个下一个潜在状态，往往无法捕捉到机器人动作引起的微妙变化。这种遗漏可能导致物理上不合理的结果，包括物体相互穿透和过度变形。为了解决这一局限性，我们提出了DeltaWorld，一种用于机器人操作的物理一致的交互式世界模拟器。我们的方法引入了Delta潜在过渡模型（Delta-LTM），该模型预测动作引起的潜在特征变化，并将其添加到当前潜在状态中以获得下一个状态，而不是直接预测下一个潜在状态。为了减轻预测未来帧中的物体相互穿透和过度变形，提出了交互感知潜在对齐，以构建反事实交互区域并监督与交互相关的潜在变化。DeltaWorld在IWS操作基准和一个自收集的跨机器人数据集上进行了评估，该数据集涵盖了多个机器人形态和操作任务。在跨机器人数据集上，DeltaWorld相对于IWS基线将FVD降低了46.6%，LPIPS降低了31.1%。这些结果突显了DeltaWorld在机器人操作中的长时间范围动作条件视频预测的潜力。
cs.RO / 16 / 2610.02694
Who Went Where When on the Lunar Surface: Forensic Trajectory Analysis to Identify Byzantine Rovers
谁在何时何地走过月球表面：用于识别拜占庭探测器的法医轨迹分析
Abstract
Future planetary surface missions are likely to involve multiple independently operated rovers sharing the same deployment region, raising the need to verify compliance with operational constraints such as Lunar Safety Zones. Because continuous in-situ observability is rarely available, such verification requires post-hoc reconstruction of rover trajectories from sparse telemetry, including odometry, pose priors, and relative inter-rover detections. We introduce the problem of forensic trajectory analysis for non-cooperative planetary rovers in the presence of Byzantine agents: rovers that provide miscalibrated or deliberately falsified measurements to support an incorrect trajectory. We show that standard outlier-robust pose graph optimisation methods are vulnerable in this setting, because Byzantine rovers can generate measurements that are internally consistent and numerous enough to make truthful incriminating measurements appear as outliers. To address this, we propose an attribution-aware trajectory estimation method that reasons over rover credibility rather than individual measurement validity. The method evaluates candidate credible rover subsets by comparing the statistical consistency of their internal and boundary relative detections against provided priors, and then estimates trajectories using only measurements attributed to credible agents. Across synthetic simulations and real planetary-analogue trajectory data, the proposed method identifies Byzantine rovers and produces significantly more accurate trajectory estimates than existing robust pose graph optimisation baselines.
Chinese Translation
未来的行星表面任务可能涉及多个独立操作的探测器共享同一部署区域，这就需要验证其是否符合操作约束，例如月球安全区。由于连续的现场可观测性很少可用，因此这种验证需要从稀疏的遥测数据中事后重建探测器轨迹，包括里程计、姿态先验和相对探测器间的检测。我们引入了在拜占庭代理存在下对非合作行星探测器进行法医轨迹分析的问题：这些探测器提供错误校准或故意伪造的测量数据，以支持不正确的轨迹。我们展示了标准的抗异常值姿态图优化方法在这种情况下是脆弱的，因为拜占庭探测器可以生成内部一致且数量足够多的测量，使得真实的指控测量看起来像异常值。为了解决这个问题，我们提出了一种关注归因的轨迹估计方法，该方法基于探测器的可信度而不是单个测量的有效性进行推理。该方法通过比较候选可信探测器子集的内部和边界相对检测的统计一致性与提供的先验进行评估，然后仅使用归因于可信代理的测量来估计轨迹。在合成模拟和真实的行星类轨迹数据中，所提出的方法能够识别拜占庭探测器，并产生显著更准确的轨迹估计，优于现有的鲁棒姿态图优化基线。
cs.RO / 17 / 2610.02697
GeoScaffold: Learning Compact Geometric Latents via Reconstruction for Efficient Vision-Language Navigation
GeoScaffold：通过重建学习紧凑的几何潜变量以实现高效的视觉-语言导航
Abstract
Recent vision-and-language navigation (VLN) systems increasingly adopt streaming Video-LLM policies that map egocentric RGB observations and instructions directly to low-level actions. Yet these policies inherit weak 3D geometric priors from 2D pretraining. Existing geometry-aware extensions charge a persistent inference-time price: depth sensors, 3D encoders, or per-step perception tool calls. We propose GeoScaffold, a geometric supervision framework that pays this price once, at training time, by internalizing geometry into the policy itself. It first learns a compact depth tokenizer on depth maps from the training trajectories and freezes it. It then fine-tunes the policy with a handful of learnable geometry query tokens, training their hidden states to reconstruct navigation-critical geometry such as depth, connectivity, and traversability. This supervision turns the query states into compact geometric latents for action decoding, and through the shared weights also internalizes geometry into the backbone's own representations. Like a scaffold, the tokenizer, target generators, and reconstruction heads are discarded after training, leaving the backbone and action interface unchanged. Extensive experiments show that GeoScaffold consistently outperforms leading vision-only navigators on continuous VLN benchmarks, offering a practical paradigm for lightweight edge deployment of spatially aware embodied navigation models.
Chinese Translation
最近的视觉与语言导航（VLN）系统越来越多地采用流式视频-大语言模型（Video-LLM）策略，将自我中心的RGB观察和指令直接映射到低级动作。然而，这些策略从2D预训练中继承了弱的3D几何先验。现有的几何感知扩展在推理时需要持续的代价：深度传感器、3D编码器或每步感知工具调用。我们提出了GeoScaffold，一种几何监督框架，通过将几何内化到策略本身，在训练时一次性支付这一代价。它首先在训练轨迹的深度图上学习一个紧凑的深度标记器并将其冻结。然后，它通过少量可学习的几何查询标记对策略进行微调，训练其隐藏状态以重建导航关键几何信息，如深度、连通性和可通行性。这种监督将查询状态转化为紧凑的几何潜变量用于动作解码，并通过共享权重将几何内化到主干网络自身的表示中。像支架一样，标记器、目标生成器和重建头在训练后被丢弃，主干网络和动作接口保持不变。大量实验表明，GeoScaffold在连续的VLN基准测试中始终优于领先的视觉单一导航器，为空间感知的具身导航模型的轻量级边缘部署提供了一种实用范式。
cs.RO / 18 / 2610.02706
AdaTempo: Learning Shared Relative Tempo from Demonstrations for Faster Robot Manipulation
AdaTempo：从演示中学习共享相对节奏以加快机器人操作
Abstract
Visuomotor policies trained via imitation learning often inherit the unnecessarily slow timing of teleoperated demonstrations. Yet uniform speedup is unreliable because different phases of a manipulation task tolerate acceleration differently. In this work, we introduce AdaTempo, a self-supervised method that accelerates visuomotor policies by exploiting shared relative-tempo structure in demonstrations. AdaTempo establishes phase correspondence, aggregates the aligned relative tempo into a consensus, and maps it to a continuous speedup profile used to resample demonstrations into accelerated training trajectories. Training standard policies such as ACT or Diffusion Policy on these resampled trajectories directly embeds the desired tempo in the learned behavior, without runtime tempo selection or online retiming. Extensive evaluations show that AdaTempo achieves up to a $3.57\times$ speedup and yields a stronger success--speed trade-off than the original policies and representative acceleration baselines.
Chinese Translation
通过模仿学习训练的视觉运动策略往往继承了遥控演示中不必要的缓慢节奏。然而，统一加速并不可靠，因为操作任务的不同阶段对加速的容忍度不同。在本研究中，我们提出了AdaTempo，这是一种自监督方法，通过利用演示中的共享相对节奏结构来加速视觉运动策略。AdaTempo建立了阶段对应关系，将对齐的相对节奏聚合成共识，并将其映射到用于重新采样演示以加速训练轨迹的连续加速曲线。对这些重新采样轨迹上训练的标准策略（如ACT或Diffusion Policy）直接嵌入了所需的节奏，而无需运行时节奏选择或在线重定时。大量评估表明，AdaTempo实现了高达$3.57 imes$的加速，并在成功率与速度的权衡上优于原始策略和代表性的加速基线。
cs.RO / 19 / 2610.02708
RoboChemGym: A Protocol-Driven Generative Simulation Framework for Long-Horizon Chemical Manipulation
RoboChemGym：一种基于协议驱动的生成仿真框架，用于长时间跨度的化学操作
Abstract
Wet-lab experimentation serves as the gold standard for hypothesis verification in scientific discovery; yet it is inherently labor-intensive, costly, and safety-critical. Embodied agents hold the promise of automating these tedious workflows, but their development is hindered by the scarcity of real-world training data. While simulation offers a scalable alternative for producing demonstrations, current methods primarily target relatively short-horizon tasks with loosely structured interactions, failing to meet the strict procedural constraints and fine-grained manipulation demands of chemical experiments. To bridge this gap, we introduce \textbf{RoboChemGym}, a framework that autonomously generates high-fidelity manipulation demonstrations aligned with real-world experiment protocols, featuring a \textit{self-improving task synthesis} mechanism to iteratively refine task execution and scene configurations, enabling the reliable generation of expert trajectories for complex, multi-object protocols exceeding 10 interaction steps. Furthermore, we introduce a hierarchical benchmark that systematically assesses performance across varying granularities, spanning from atomic operations to full-cycle experimental workflows. RoboChemGym sets a scalable paradigm for the automated data synthesis and capability evaluation of embodied agents in intricate chemical tasks, serving as a critical stepping stone toward fully intelligent laboratories.
Chinese Translation
湿实验室实验是科学发现中假设验证的金标准；然而，它本质上劳动密集、成本高昂且涉及安全风险。具身智能体有望自动化这些繁琐的工作流程，但其发展受到真实世界训练数据稀缺的制约。虽然仿真提供了一种可扩展的替代方案来生成演示，但当前的方法主要针对相对短期的任务，且交互结构松散，未能满足化学实验严格的程序约束和细致的操作需求。为了解决这一问题，我们提出了 extbf{RoboChemGym}，一个自主生成与真实实验协议相一致的高保真操作演示的框架，具有 extit{自我改进任务合成}机制，以迭代方式优化任务执行和场景配置，从而可靠地生成复杂多物体协议的专家轨迹，交互步骤超过10步。此外，我们引入了一个分层基准，系统地评估在不同粒度下的性能，从原子操作到完整实验工作流程。RoboChemGym为复杂化学任务中具身智能体的自动数据合成和能力评估设定了一个可扩展的范式，成为实现完全智能实验室的重要基石。
cs.RO / 20 / 2610.02717
RoboBridge: A Self-Evolving Embodied Agent Framework for Sim-to-Real Transfer
RoboBridge：一种自我演化的具身智能代理框架用于仿真到现实的转移
Abstract
A key challenge in bringing embodied intelligence into the real world is transferring capabilities from simulation to reality and enabling agents to continually adapt after deployment. End-to-end vision-language-action policies provide strong manipulation capabilities, but their transfer to physical environments typically relies on calibrating simulated visual and dynamical conditions, collecting additional target-domain demonstrations, and optimizing the policy through further training. Tool-using embodied agents offer flexible task orchestration, yet existing systems primarily emphasize task execution and experience reuse within a given environment, with limited support for transferring procedural knowledge and continuously adapting it across simulation and reality. We propose RoboBridge, a framework that treats sim-to-real transfer as the continued adaptation of executable task skills. The agent represents task knowledge as procedures connecting task intent, observations, tool operations, and outcome verification. Interaction feedback is used to generate candidate skill revisions, which are evaluated before being persisted or rejected. A pretrained vision-language-action policy is exposed as a reusable action tool and enhanced with inference-time guidance, enabling fine-grained execution without retraining the underlying policy. RoboBridge grounds transferable skills in task semantics and interaction interfaces shared across simulation and reality. This representation preserves reusable task structure while allowing environment-dependent operations to be selectively revised through real-world execution feedback. We evaluate the framework on LIBERO-PRO and corresponding physical tasks, studying both skill evolution and post-transfer adaptation. Our framework provides a route from one-shot policy deployment to continual procedural learning across environments.
Chinese Translation
将具身智能引入现实世界的一个关键挑战是将能力从仿真转移到现实，并使代理在部署后能够持续适应。端到端的视觉-语言-动作策略提供了强大的操作能力，但它们向物理环境的转移通常依赖于校准模拟的视觉和动态条件、收集额外的目标领域演示以及通过进一步训练优化策略。使用工具的具身代理提供灵活的任务协调，但现有系统主要强调在给定环境中的任务执行和经验重用，对跨仿真和现实转移程序性知识的支持有限。我们提出了RoboBridge，一个将仿真到现实转移视为可执行任务技能持续适应的框架。该代理将任务知识表示为连接任务意图、观察、工具操作和结果验证的程序。交互反馈用于生成候选技能修订，这些修订在被保留或拒绝之前会进行评估。预训练的视觉-语言-动作策略被作为可重用的动作工具暴露，并通过推理时指导进行增强，从而实现细粒度执行而无需重新训练基础策略。RoboBridge将可转移技能扎根于跨仿真和现实共享的任务语义和交互接口。这种表示保留了可重用的任务结构，同时允许通过现实世界执行反馈选择性地修订环境依赖的操作。我们在LIBERO-PRO及其对应的物理任务上评估该框架，研究技能演变和转移后的适应。我们的框架提供了一条从一次性策略部署到跨环境持续程序学习的路径。
cs.RO / 21 / 2610.02728
Around the World: Unified Learned Locomotion on a 270 g Continuous-Rotation Quadruped
环游世界：统一学习的270克连续旋转四足机器人运动
Abstract
Closed-loop learned locomotion is established on commercial quadrupeds but remains uncommon at the sub-kilogram scale. Continuous-rotation legs give MiNI-Q, a 270 g quadruped, access to supporting configurations on either side of the body. We exploit this range with a single posture-conditioned reinforcement-learning policy that runs entirely onboard. A continuous joint-space reference on the torus $T^8$ and its gravity-conditioned transformation connect upright walking, inverted walking, and landing recovery without state machines or phase switching. Coordinated posture and release curricula train this behavior family; identified actuation, cross-engine validation, and embedded execution support hardware transfer. The same sub-100k-parameter network tracks forward velocity with RMSE of 0.037 m/s upright and 0.050 m/s inverted, resumes walking in 24 of 30 release trials, and operates with four interchangeable foot geometries across four indoor surfaces. Hardware experiments and simulation ablations connect these capabilities to the representation, conditioning, and training choices that exploit the platform's motion range. Demonstration videos and supplementary material are available on the project website: https://submissionreview.github.io/around-the-world/.
Chinese Translation
闭环学习运动在商业四足机器人上已得到建立，但在亚千克级别的应用仍然不常见。连续旋转的腿部使得重270克的四足机器人MiNI-Q能够在身体两侧访问支撑配置。我们利用这一范围，通过一个完全在机载运行的单一姿态条件强化学习策略来实现。一个在环面 $T^8$ 上的连续关节空间参考及其重力条件变换将直立行走、倒立行走和着陆恢复连接起来，而无需状态机或相位切换。协调的姿态和释放课程训练这一行为家族；识别的驱动、跨引擎验证和嵌入式执行支持硬件转移。相同的少于10万参数的网络在直立状态下以0.037 m/s的均方根误差（RMSE）跟踪前进速度，在倒立状态下为0.050 m/s，在30次释放试验中恢复行走24次，并在四种室内表面上使用四种可互换的足部几何形状进行操作。硬件实验和仿真消融研究将这些能力与利用平台运动范围的表示、条件和训练选择联系起来。演示视频和补充材料可在项目网站上获取：https://submissionreview.github.io/around-the-world/.
cs.RO / 22 / 2610.02750
CSIR: Contextually and Socially Informed Robots for Efficient Person Goal Navigation
CSIR：基于上下文和社会信息的高效人目标导航机器人
Abstract
We address the Person Goal Navigation (PersonNav) problem, enabling robots to search for people under realistic constraints on information accessibility and language uncertainty. This tackles the current solutions revolving around rigid person finding systems requiring exact knowledge of the individual for item-delivery in indoor settings. While the focus in literature is on learned methods using unavailable public or sparse data due to human privacy. We propose a planning framework that combines distance and semantic information about the recipient (habits and intent) weighted by the trust of the user-provided information. A synthetic benchmark of scenarios is also developed including actors, items, and requests to evaluate performance before real-world deployment aiming to simulate natural language human-robot interactions. Results show our informed search outperforms classical distance-based graph baselines, while semantics alone lead to ungrounded, sporadic search. Our method achieves strong gains over baselines, with an LLM-based variant performing comparably, and its explicit belief representation naturally supports future Bayesian filtering. Hardware tests demonstrate our method supports real-world embodiment able to leverage between semantic and distance information in a real setting. This work moves toward more intelligent mobile service agents capable of human-like, informed search in realistic environments to be leveraged in day-to-day use. https://anonymous.4open.science/r/personnavsite-4ED2/index.html.
Chinese Translation
我们解决了人目标导航（PersonNav）问题，使机器人能够在信息可获取性和语言不确定性等现实约束下搜索人。这一研究应对了当前围绕着刚性人寻找系统的解决方案，这些系统要求对个体有确切的了解，以便在室内环境中进行物品递送。文献中的研究主要集中在使用由于人类隐私原因而无法获取的公共或稀疏数据的学习方法上。我们提出了一种规划框架，该框架结合了关于接收者的距离和语义信息（习惯和意图），并根据用户提供信息的可信度进行加权。同时，我们还开发了一个合成基准场景，包括参与者、物品和请求，以评估在实际部署前的性能，旨在模拟自然语言的人机交互。结果表明，我们的知情搜索在性能上优于经典的基于距离的图形基线，而仅使用语义则导致了无基础、零散的搜索。我们的方法在基线之上取得了显著提升，其中基于大语言模型（LLM）的变体表现相当，其明确的信念表示自然支持未来的贝叶斯过滤。硬件测试表明，我们的方法支持能够在真实环境中利用语义和距离信息的现实体现。这项工作朝着更智能的移动服务代理迈进，使其能够在现实环境中进行类人、知情的搜索，以便在日常使用中发挥作用。
cs.RO / 23 / 2610.02759
Proprioceptive Sketches as Long-Horizon Intent for Generative Action Policies
本体感知草图作为生成行动策略的长远意图
Abstract
Generative robot policies predict short action chunks but lack explicit long-horizon intent. Recent methods expose longer-horizon structure through language plans, subgoal images, or video forecasts, which are costly to generate and still need to be translated into robot motion. Predicting future robot motions avoids this translation, but a dense, time-indexed trajectory requires numerous parameters to cover the full remaining task, and over a short horizon it largely repeats the action chunk and adds little guidance for action generation. We propose Proprioceptive Action Models (PAM), which jointly generate a compact, timing-free sketch of the robot's remaining joint-space path and a dense executable action chunk within a single transformer denoiser. The sketch parameterizes the path by arc length rather than time, capturing geometric intent invariant to execution timing. Block-causal attention and a staggered denoising schedule maintain directed sketch-to-action dependence, ensuring the action tokens condition on a progressively cleaner sketch throughout sampling. In simulation, PAM improves over its action-only counterparts on Push-T and LIBERO-Long; on four real-world bimanual tasks, it raises success from 47.5% to 75.0%. Project page: https://nicehiro.github.io/pam_dp/
Chinese Translation
生成机器人策略预测短期行动片段，但缺乏明确的长远意图。近期的方法通过语言计划、子目标图像或视频预测揭示了更长时间范围的结构，这些方法生成成本高，并且仍需转换为机器人运动。预测未来机器人运动避免了这种转换，但密集的时间索引轨迹需要大量参数来覆盖剩余任务的全部内容，并且在短期内主要重复行动片段，对行动生成的指导作用有限。我们提出了本体感知行动模型（Proprioceptive Action Models, PAM），该模型在单个变换器去噪器中联合生成机器人的剩余关节空间路径的紧凑、无时间限制的草图和一个密集的可执行行动片段。该草图通过弧长而非时间对路径进行参数化，捕捉与执行时机无关的几何意图。块因果注意力和错开去噪调度保持了草图到行动的有向依赖，确保行动标记在采样过程中依赖于逐渐清晰的草图。在仿真中，PAM在Push-T和LIBERO-Long任务上优于仅有行动的对手；在四个真实的双手任务中，其成功率从47.5%提高到75.0%。项目页面：https://nicehiro.github.io/pam_dp/
cs.RO / 24 / 2610.02765
Localized Conformal Safety Monitoring with Vision-Language Models for Autonomous Driving
基于视觉-语言模型的局部一致性安全监测在自动驾驶中的应用
Abstract
Monitoring planned driving trajectories requires accurately estimating the collision likelihood with actors whose motion is itself impacted by the ego motion. Existing classical approaches are often limited by the quality of their forecasting model. Vision-language models (VLMs) have shown promise in reasoning about the consequences of high-level actions, yet their approximate predictions are unsuitable for safety-critical applications such as autonomous driving. Conformal prediction (CP) has emerged as a data-driven framework for quantifying the uncertainty of black-box model predictions. We propose Split Label-Localized Conformal Prediction (SLLCP), a post-hoc calibration layer over frozen VLMs that transforms their unreliable predictions into probabilistically calibrated safety prediction sets. We consider how the ability to estimate safety can depend on the observed driving scene and introduce a localized procedure that upweights relevant past experience when calculating uncertainty thresholds. We provide label-conditional finite-sample distribution-free coverage under exchangeability. Evaluated over 15k CARLA trajectories from unseen scenarios, SLLCP correctly flags 89.6% of collision-causing trajectories with a Qwen backbone and 88.4% with a Cosmos backbone, while the base VLMs only flagged 4.6% and 39.1% of the collision-causing trajectories, respectively. These results indicate that local, label-conditional calibration can reduce missed unsafe trajectories.
Chinese Translation
监测规划的驾驶轨迹需要准确估计与运动体发生碰撞的可能性，而运动体的运动又受到自我运动的影响。现有的经典方法往往受到其预测模型质量的限制。视觉-语言模型（VLMs）在推理高层次动作的后果方面显示出潜力，但其近似预测不适合用于安全关键的应用，如自动驾驶。一致性预测（CP）作为一种数据驱动框架，已被提出用于量化黑箱模型预测的不确定性。我们提出了分裂标签局部一致性预测（SLLCP），这是一种在冻结的VLMs上进行后处理校准的层，将其不可靠的预测转化为概率校准的安全预测集。我们考虑到安全估计的能力可能依赖于观察到的驾驶场景，并引入了一种局部程序，在计算不确定性阈值时加权相关的过去经验。我们在可交换性条件下提供了标签条件的有限样本分布无关覆盖。在对来自未见场景的15,000条CARLA轨迹进行评估时，SLLCP正确标记了89.6%的导致碰撞的轨迹（使用Qwen骨干网络）和88.4%的轨迹（使用Cosmos骨干网络），而基础的VLMs仅标记了4.6%和39.1%的导致碰撞的轨迹。这些结果表明，局部的标签条件校准可以减少漏标不安全轨迹的情况。
cs.RO / 25 / 2610.02784
SimpleTouch: Can Vision-Language-Action Models Master Contact-Rich Manipulation Without Tactile Policy Pretraining?
SimpleTouch：视觉-语言-动作模型能否在没有触觉策略预训练的情况下掌握接触丰富的操作？
Abstract
Tactile sensing provides essential contact information for robotic manipulation, yet incorporating it into pretrained vision-language-action (VLA) models remains challenging. A common concern is that simply introducing touch during task-specific fine-tuning may fail to bridge the cross-modal gap, yielding limited gains or even reduced success. Consequently, existing methods often rely on large-scale tactile policy pretraining or separate visuotactile alignment, adding data requirements and training stages. We introduce SimpleTouch, a simple VLA extension that augments $\pi_{0.5}$ with a tactile expert, to test whether these additional stages are necessary. Leveraging all tokens from a frozen pretrained tactile encoder, the expert learns from action supervision and multi-horizon prediction of future tactile latents. This single-stage training uses only task demonstrations, without additional tactile policy pretraining or separate alignment. With 50 demonstrations per task, SimpleTouch achieves the highest success rate among evaluated methods on all six UniVTAC tasks. Its average success rate reaches 77.5%, compared with 45.2% for FTP-$\pi_{0.5}$ and 66.7% for FTP-1, corresponding to gains of 32.3 and 10.8 percentage points, respectively. Across four real-world tasks, it averages 71.3%, exceeding FTP-1 by 8.8 percentage points. These results demonstrate that, given pretrained VLA and tactile representations, additional tactile policy pretraining is not a prerequisite for strong performance on these tasks, offering a simpler route to contact-rich manipulation. Project page: https://simpletouch-robot.github.io/
Chinese Translation
触觉感知为机器人操作提供了重要的接触信息，但将其纳入预训练的视觉-语言-动作（VLA）模型仍然具有挑战性。一个普遍的担忧是，在任务特定的微调过程中，仅仅引入触觉可能无法弥合跨模态的差距，从而导致有限的收益甚至成功率下降。因此，现有方法通常依赖于大规模的触觉策略预训练或单独的视觉-触觉对齐，这增加了数据需求和训练阶段。我们提出了SimpleTouch，这是一个简单的VLA扩展，通过触觉专家增强$ ext{π}_{0.5}$，以测试这些额外阶段是否必要。利用来自冻结的预训练触觉编码器的所有标记，专家通过动作监督和对未来触觉潜在变量的多时间步预测进行学习。这一单阶段训练仅使用任务演示，而不需要额外的触觉策略预训练或单独的对齐。在每个任务中提供50个演示，SimpleTouch在评估的六个UniVTAC任务中取得了最高的成功率。其平均成功率达到77.5%，而FTP-$ ext{π}_{0.5}$为45.2%，FTP-1为66.7%，分别对应32.3和10.8个百分点的增益。在四个现实世界任务中，SimpleTouch的平均成功率为71.3%，超过FTP-1 8.8个百分点。这些结果表明，在给定预训练的VLA和触觉表示的情况下，额外的触觉策略预训练并不是在这些任务上取得强大表现的先决条件，为接触丰富的操作提供了一条更简单的途径。项目页面：https://simpletouch-robot.github.io/
cs.RO / 26 / 2610.02788
Skill2Real: Agentic Skill Learning for Zero-Shot Sim-to-Real Robot Manipulation
Skill2Real：用于零-shot模拟到现实机器人操作的自主技能学习
Abstract
Transferring robotic skills from simulation to reality requires task knowledge that remains usable across differences in perception, dynamics, and embodiment. We introduce Skill2Real, an agentic policy framework that learns executable skills through a shared application programming interface (API). A Proposer-Verifier-Governor (PVG) loop uses privileged simulation evidence to diagnose outcomes and validate updates, while keeping learned skills grounded in public observations and API semantics. The Cerebellum first acquires local manipulation skills; the Brain then learns task-level composition with the Cerebellum frozen. Both memories transfer to the real robot without task-policy fine-tuning or skill-memory updates. As GPT-5.6 Sol learns skills on LIBERO-90, evaluating each frozen checkpoint with GPT-6 Astra raises LIBERO-Pro Long success from 2.0% to 56.3%, without training on Pro Long. Independent Robosuite training reaches 85.1% and 89.4% mean success with Sol and Opus 5 across seven tasks, respectively. Frozen Sol-trained LIBERO-90 skills achieve 78.75% mean completion across four real-world manipulation tasks with Astra. Removing the Verifier or Governor during LIBERO-90 training lowers final Pro Long success by 17.3 and 13.3 percentage points, respectively. These results support learning and transferring a hierarchy of executable skills through a common robot interface.
Chinese Translation
将机器人技能从模拟转移到现实需要在感知、动态和体现之间保持可用的任务知识。我们提出了Skill2Real，一个自主策略框架，通过共享的应用程序编程接口（API）学习可执行技能。提议者-验证者-管理者（PVG）循环利用特权模拟证据来诊断结果和验证更新，同时保持学习到的技能与公共观察和API语义相结合。小脑首先获取局部操作技能；然后大脑在小脑被冻结的情况下学习任务级组合。这两种记忆在没有任务策略微调或技能记忆更新的情况下转移到真实机器人上。当GPT-5.6 Sol在LIBERO-90上学习技能时，使用GPT-6 Astra评估每个冻结的检查点将LIBERO-Pro Long的成功率从2.0%提高到56.3%，而无需在Pro Long上进行训练。独立的Robosuite训练在七个任务中分别达到了85.1%和89.4%的平均成功率，使用Sol和Opus 5。冻结的Sol训练的LIBERO-90技能在四个现实世界操作任务中实现了78.75%的平均完成率。LIBERO-90训练期间移除验证者或管理者分别使最终Pro Long成功率降低了17.3和13.3个百分点。这些结果支持通过通用机器人接口学习和转移可执行技能的层次结构。
cs.RO / 27 / 2610.02802
ManiPhysicsBench: Physics-Based Assessment of Object Preservation in VLA Manipulation
ManiPhysicsBench：基于物理的物体保护评估在VLA操作中的应用
Abstract
Vision-language-action (VLA) models aim to perform diverse manipulation tasks, but task success in existing rigid-body benchmarks does not indicate whether they preserve objects. We introduce ManiPhysicsZoo, which consolidates literature-supported material properties, 3D meshes, and supporting references into reusable object assets. Using these assets, a solver-based assessment computes grasp-specific damage thresholds from object geometry, material properties, and recorded grasp conditions and compares them with recorded contact forces to assess potential deformation and fracture. Building on these components, ManiPhysicsBench evaluates object preservation in LIBERO and SimplerEnv across three physics axes and three difficulty levels. Public VLA checkpoints show a substantial gap between task success and safe success, defined as task completion while preserving the object. Their gripper commands concentrate near full opening and closure, with largely similar aggregate distributions across objects, consistent with binary gripper supervision. We examine how object-specific continuous gripper labels change model behavior by retraining a VLA model. The retrained model shows more object-dependent gripping and higher safe success, but lower task success and limited generalization of object-preserving behavior.
Chinese Translation
视觉-语言-动作（VLA）模型旨在执行多样化的操作任务，但现有刚体基准测试中的任务成功并不表明它们能够保护物体。我们引入了ManiPhysicsZoo，它将文献支持的材料属性、3D网格和相关参考整合为可重用的物体资产。利用这些资产，基于求解器的评估计算从物体几何形状、材料属性和记录的抓取条件中得出的特定抓取损伤阈值，并将其与记录的接触力进行比较，以评估潜在的变形和断裂。在这些组件的基础上，ManiPhysicsBench在LIBERO和SimplerEnv中评估物体保护，涵盖三个物理轴和三个难度等级。公共VLA检查点显示任务成功与安全成功之间存在显著差距，安全成功定义为在保护物体的情况下完成任务。它们的抓取命令集中在完全打开和关闭附近，且在物体之间的总体分布大致相似，这与二元抓取监督一致。我们通过重新训练VLA模型来检查物体特定的连续抓取标签如何改变模型行为。重新训练的模型显示出更依赖物体的抓取和更高的安全成功率，但任务成功率较低，且物体保护行为的泛化能力有限。
cs.RO / 28 / 2610.02803
LOCUS: Landmark-Oriented Container Discrimination Using Spatial Graphs
LOCUS：基于地标的容器区分方法使用空间图
Abstract
As robots are increasingly deployed in unstructured, real-world environments, the ability to reason about complex spatial and semantic relationships among objects remains a fundamental challenge in enabling robust and generalizable manipulation and navigation. For example, deciding where to search for an object that is not in plain sight depends on where it is physically plausible as well as semantically likely. Popular methods such as cosine similarity with CLIP struggle to disambiguate between spatially and semantically similar objects that could contain a target object. We propose Landmark-Oriented Container Discrimination Using Spatial Graphs (LOCUS). To jointly reason about spatial information and semantics, we train a GNN to update CLIP embeddings on a full environment scene graph by passing embedding information between nodes based on proximity. CLIP embeddings, augmented by semantic knowledge from a fusion of household ontologies provide a robust semantic signal. We evaluate in simulation on a noiseless scene graph from simulator metadata, making the approach detection-agnostic. Our approach outperforms random, CLIP, Tidybot, and an LLM planner in the majority of room classes and configurations in simulation. To show our approach is robust to scene graph node placement and label noise, we demonstrate a physical mobile manipulator running in an exploration pipeline start-to-finish including scene graph labels from Detic.
Chinese Translation
随着机器人在非结构化的现实环境中越来越多地被部署，推理对象之间复杂的空间和语义关系的能力仍然是实现稳健且可推广的操作和导航的基本挑战。例如，决定在哪里寻找一个不在视线范围内的物体，依赖于其物理上合理的位置以及语义上的可能性。流行的方法如使用 CLIP 的余弦相似度在区分可能包含目标物体的空间和语义相似对象时遇到困难。我们提出了一种基于空间图的地标导向容器区分方法（LOCUS）。为了共同推理空间信息和语义，我们训练了一个图神经网络（GNN），通过基于接近度在节点之间传递嵌入信息，更新 CLIP 嵌入于完整的环境场景图中。通过家庭本体的融合提供的语义知识增强的 CLIP 嵌入提供了强大的语义信号。我们在来自模拟器元数据的无噪声场景图上进行模拟评估，使得该方法与检测无关。我们的方案在大多数房间类别和配置中优于随机、CLIP、Tidybot 和 LLM 规划器。为了证明我们的方法对场景图节点位置和标签噪声具有鲁棒性，我们展示了一个物理移动操纵器在探索管道中从头到尾的运行，包括来自 Detic 的场景图标签。
cs.RO / 29 / 2610.02804
SARI: Phase-Split Sim-Real Co-Training for Contact-Rich Manipulation
SARI：用于接触丰富操作的相位分割模拟-真实协同训练
Abstract
Vision-language-action (VLA) models often require costly real-world demonstrations to adapt to contact-rich manipulation tasks, particularly when generalization across object placements is needed. We propose SARI (Simulated Approach, Real Interaction), a phase-split sim-and-real co-training framework built on a simple insight: spatial coverage and contact physics should be acquired from the domains best suited to them. Specifically, free-space approaches require spatial diversity but tolerate modest simulation gaps, making them ideal for synthetic generation; conversely, contact interactions demand accurate physics but vary little across object placements, allowing a few real demonstrations to generalize across the workspace. SARI generates diverse simulated approaches in a photorealistic digital twin while collecting real contact interactions at only a few placements. Post-trained on these phase-segmented demonstrations, a single policy seamlessly stitches simulated approaches with real contact interactions using visual appearance alignment and a shared camera-relative action representation--without explicit phase labels or hand-coded switches. Across five real-world contact-rich manipulation tasks, SARI reduces real-data collection time by 34.3% and achieves 27.5% success at unseen placements, where all full-task sim-real baselines fail completely (0%).
Chinese Translation
视觉-语言-动作（VLA）模型通常需要昂贵的现实世界演示，以适应接触丰富的操作任务，特别是在需要跨物体放置进行泛化时。我们提出了SARI（模拟方法，真实交互），一种基于简单洞察的相位分割模拟与真实协同训练框架：空间覆盖和接触物理应从最适合它们的领域获取。具体而言，自由空间方法需要空间多样性，但可以容忍适度的模拟差距，使其非常适合合成生成；相反，接触交互要求准确的物理模型，但在物体放置之间变化很小，允许通过少量真实演示在工作空间中进行泛化。SARI在一个逼真的数字双胞胎中生成多样化的模拟方法，同时仅在少数放置中收集真实接触交互。在这些相位分割演示上进行后训练后，单一策略通过视觉外观对齐和共享的相机相对动作表示无缝地将模拟方法与真实接触交互结合起来，而无需显式的相位标签或手动编写的切换。在五个真实世界的接触丰富操作任务中，SARI将真实数据收集时间减少了34.3%，并在未见放置中实现了27.5%的成功率，而所有完整任务的模拟-真实基线完全失败（0%）。
cs.RO / 30 / 2610.02811
Learning Reflexive Behavior for Contact-Rich Manipulation
学习接触丰富操作的反射行为
Abstract
During contact-rich manipulation, interactions between a robot and its environment carry information about local geometry: a surface prevents penetration, a bore guides a peg. A controller that exploits these interactions can comply with environmental constraints while preserving task intent. Robot-earning systems commonly use position, hybrid force-position, or Cartesian impedance control. Their prescribed tracking objectives, stiffness, or force-control directions may not match local constraints and may degrade performance. We learn a proprioceptive reflex policy in simulation on three simple interaction primitives: a spring, a plane, and a rail. The policy maps task-space commands to joint-position targets using state history, without direct force or geometrical measurements. Once trained, it serves as a frozen execution layer beneath higher-level controllers. We evaluate it in dual-arm box lifting, peg insertion, and surface following. In box lifting, the reflex kept the force below the threshold while the baseline failed. In rough-surface following it stayed below the 10 N reference, on par with tuned hybrid force-position control. In 0.02 mm peg insertion It reduced mean estimated contact force to less than half that of the baseline while increasing hardware success rates from at most 22 % to 36-58 %. By separating contact response from command generation, the reflex policy provides motion planners, learned policies, and teleoperators with robust contact-rich execution.
Chinese Translation
在接触丰富的操作中，机器人与其环境之间的交互携带着关于局部几何形状的信息：表面防止穿透，孔引导插销。利用这些交互的控制器可以在遵循环境约束的同时保持任务意图。机器人学习系统通常使用位置控制、混合力-位置控制或笛卡尔阻抗控制。它们规定的跟踪目标、刚度或力控制方向可能与局部约束不匹配，从而降低性能。我们在模拟中学习了一种本体反射策略，针对三种简单的交互原语：弹簧、平面和轨道。该策略利用状态历史将任务空间命令映射到关节位置目标，而无需直接的力或几何测量。训练完成后，它作为高层控制器下的固定执行层。我们在双臂箱体提升、插销插入和表面跟随中对其进行了评估。在箱体提升中，反射保持了力在阈值以下，而基线方法失败。在粗糙表面跟随中，它保持在10 N参考值以下，与调优的混合力-位置控制相当。在0.02毫米插销插入中，它将平均估计接触力减少到基线的一半以下，同时将硬件成功率从最多22%提高到36-58%。通过将接触响应与命令生成分离，反射策略为运动规划器、学习策略和遥控操作员提供了稳健的接触丰富执行能力。
cs.RO / 31 / 2610.02813
Register-Routed Delayed Fusion: Rewiring Shortcut-Prone Observation Fusion in Visuomotor Imitation
注册路由延迟融合：重构视觉运动模仿中的快捷观察融合
Abstract
Visuomotor imitation policies combine high-dimensional visual observations with compact signals such as proprioception, and their fusion topology determines when and through which tokens these streams interact. In dense token fusion, visual tokens may attend directly to compact tokens from the first encoder layer, allowing action-predictive compact cues to influence spatial visual representations early in their formation. We ask whether controlling this route improves visual responsiveness and policy behavior. We introduce Register-Routed Delayed Fusion (RRDF), which masks direct compact-visual attention and stages cross-modal interaction through a learned register workspace. Its isolate-collect-route schedule protects an early stream-separated prefix and later permits only register-mediated exchange, while compact conditioning remains available to the native action generator. Across five simulation tasks and three real-robot tasks, RRDF matches or improves dense ACT under nominal conditions. Appearance-shift evaluations on four simulation tasks and held-out-position evaluations on three real-robot tasks also favor RRDF. Phase-matched input probes show lower measured state-to-image sensitivity, while ablations indicate that adding registers alone does not reproduce the full performance gain. These results support controlling cross-modal propagation while retaining compact action conditioning.
Chinese Translation
视觉运动模仿策略将高维视觉观察与诸如本体感觉等紧凑信号结合在一起，其融合拓扑决定了这些流何时以及通过哪些标记进行交互。在密集标记融合中，视觉标记可能直接关注来自第一个编码器层的紧凑标记，从而使得预测动作的紧凑线索能够在空间视觉表征形成的早期阶段产生影响。我们探讨控制这一路径是否能改善视觉响应性和策略行为。我们提出了注册路由延迟融合（Register-Routed Delayed Fusion, RRDF），该方法屏蔽了直接的紧凑-视觉注意，并通过学习的注册工作空间分阶段进行跨模态交互。其隔离收集路径调度保护了一个早期流分离的前缀，随后仅允许注册介导的交换，同时紧凑条件仍可供本地动作生成器使用。在五个仿真任务和三个真实机器人任务中，RRDF在正常条件下与密集ACT相匹配或有所改善。对四个仿真任务的外观变化评估和对三个真实机器人任务的保留位置评估也支持RRDF。相位匹配输入探针显示出较低的测量状态到图像的敏感性，而消融实验表明，仅添加注册并不能重现完整的性能提升。这些结果支持在保持紧凑动作条件的同时控制跨模态传播。
cs.RO / 32 / 2610.02832
FastOPD: On-Policy Distillation for Lightweight VLA Deployment
FastOPD：用于轻量级视觉-语言-动作（VLA）部署的在线蒸馏
Abstract
Vision-Language-Action (VLA) foundation models have scaled rapidly to enhance manipulation performance and generalizability, but this scaling incurs high computational costs that render real-world deployment increasingly challenging. Existing approaches typically mitigate this issue by designing smaller architectures or reducing the iterative denoising steps in flow-based policies. In this work, we propose FastOPD, a foundation-to-lightweight VLA framework that enables the practical deployment of large-scale VLAs through efficient on-policy distillation. Specifically, FastOPD adapts a flow map for single-state teacher supervision and combines it with a self-consistency objective to construct a compact student that learns the teacher dynamics. Furthermore, we theoretically demonstrate that minimizing this objective allows the distilled student to recover a distribution on par with that induced by an ideal few-step teacher model. We evaluate FastOPD across diverse foundation policies in simulation and real-world experiments. On LIBERO, FastOPD retains 84% of the performance of $\pi_{0.5}$ with only two inference steps, reducing inference latency by 78.1% while outperforming existing few-step distillation baselines in average success rate. With LingBot-VLA as the teacher, FastOPD improves the single-step success rate over the base student by 15.9 percentage points on RoboTwin 2.0. We further demonstrate its applicability to a World Action Model (WAM) and deploy a compact student distilled from MolmoAct2 on a real robot.
Chinese Translation
视觉-语言-动作（VLA）基础模型迅速扩展，以增强操作性能和泛化能力，但这种扩展带来了高昂的计算成本，使得实际部署变得愈加困难。现有的方法通常通过设计更小的架构或减少基于流的策略中的迭代去噪步骤来缓解这一问题。在本研究中，我们提出了FastOPD，一个基础到轻量级的VLA框架，通过高效的在线蒸馏实现大规模VLA的实际部署。具体而言，FastOPD为单状态教师监督适配了流图，并结合自一致性目标构建一个紧凑的学生模型，使其学习教师动态。此外，我们从理论上证明，最小化该目标使得蒸馏学生能够恢复与理想的少步教师模型所诱导的分布相当的分布。我们在模拟和真实实验中评估了FastOPD在多种基础策略下的表现。在LIBERO上，FastOPD在仅进行两次推理步骤的情况下保留了$ ext{π}_{0.5}$性能的84%，将推理延迟降低了78.1%，同时在平均成功率上超越了现有的少步蒸馏基线。在RoboTwin 2.0上，以LingBot-VLA作为教师，FastOPD将单步成功率提高了15.9个百分点。我们进一步展示了其在世界行动模型（WAM）中的适用性，并在真实机器人上部署了从MolmoAct2蒸馏出的紧凑学生模型。
cs.RO / 33 / 2610.02840
PointWAM: 3D World Action Modeling for Dexterous Robotic Manipulation
PointWAM：用于灵巧机器人操作的3D世界动作建模
Abstract
World action models jointly learn to forecast world dynamics and predict robot actions, such that the learned internal world dynamics guide accurate actions. Existing approaches typically represent the world as RGB frames or latent counterparts while predicting actions as end-effector poses or joint angles, but they often struggle to capture the 3D spatial structure and contact geometry central to dexterous manipulation. We introduce Point World Action Model (PointWAM), a 3D world action model that decomposes the world into a scene (i.e., environment) and hands (i.e., actor), and jointly forecasts both as 3D point trajectories within a shared space-time coordinate frame. This explicit, disentangled representation enables effective pre-training on large-scale human demonstration videos without requiring any task-specific object or keypoint selection. Given a colored point cloud and a language instruction, PointWAM predicts how the scene and hands co-evolve in 3D space over time, then retargets the forecast hand motion to robot actions. Pre-training on human videos improves average DexJoCo success by 56.9 percentage points, and scene-trajectory supervision adds 10.9 points over forecasting the hands alone. With both, PointWAM surpasses the prior state of the art on ten DexJoCo tasks by 11.7 points and outperforms strong VLAs on a real robot.
Chinese Translation
世界动作模型共同学习预测世界动态和机器人动作，使得学习到的内部世界动态能够指导准确的动作。现有的方法通常将世界表示为RGB帧或潜在对应物，同时将动作预测为末端执行器姿态或关节角度，但它们往往难以捕捉到灵巧操作中至关重要的3D空间结构和接触几何。我们提出了点世界动作模型（PointWAM），这是一种3D世界动作模型，将世界分解为场景（即环境）和手（即执行者），并在共享的时空坐标框架内共同预测这两者作为3D点轨迹。这种明确、解耦的表示使得在大规模人类演示视频上进行有效的预训练成为可能，而无需任何特定任务的物体或关键点选择。给定一个彩色点云和语言指令，PointWAM预测场景和手如何在3D空间中随时间共同演变，然后将预测的手部运动重新定向为机器人动作。在人类视频上的预训练使得平均DexJoCo成功率提高了56.9个百分点，而场景轨迹监督在单独预测手部时增加了10.9个百分点。结合这两者，PointWAM在十个DexJoCo任务上超越了之前的最先进水平11.7个百分点，并在真实机器人上超越了强大的视觉语言算法（VLA）。
cs.RO / 34 / 2610.02848
Permutation Robustness Is Not Enough: Action Collapse in Multi-Agent Transformer Policies
置换鲁棒性不足：多智能体变换器策略中的行动崩溃
Abstract
Transformer policies are attractive for multi-agent robot learning because self-attention can model interactions among agents. However, multi-agent teams are unordered, while transformers typically process agents as ordered token sequences. We study how this mismatch affects cooperative navigation policies under agent-order permutations. Our results show that low permutation error alone can be misleading: policies may appear robust simply because all agents choose the same action. We therefore evaluate policies using both permutation-consistency metrics and action-collapse diagnostics, including action diversity, same-action fraction, and maximum action frequency. A PPO-ID baseline yields non-collapsed behavior but remains order-sensitive, while strong equivariance regularization can still induce homogeneous behavior. A weak equivariance penalty improves the robustness while preserving more diverse actions for teams with \(N=3\) agents, whereas teams with \(N=4\) agents require substantially smaller regularization weights. These findings suggest that multi-agent transformer policies should be evaluated not only by return and permutation robustness, but also by whether they maintain non-collapsed, differentiated multi-agent behavior.
Chinese Translation
变换器策略因其自注意力机制能够建模智能体之间的交互而在多智能体机器人学习中备受青睐。然而，多智能体团队是无序的，而变换器通常将智能体视为有序的标记序列。我们研究了这种不匹配如何影响智能体顺序置换下的协作导航策略。我们的结果表明，低置换误差本身可能具有误导性：策略可能看起来很鲁棒，仅仅是因为所有智能体选择了相同的行动。因此，我们使用置换一致性指标和行动崩溃诊断来评估策略，包括行动多样性、相同行动比例和最大行动频率。PPO-ID基线表现出非崩溃行为，但仍然对顺序敏感，而强等变性正则化仍然可能导致同质行为。弱等变性惩罚在保持更多多样化行动的同时提高了鲁棒性，适用于有3个智能体的团队，而有4个智能体的团队则需要显著更小的正则化权重。这些发现表明，多智能体变换器策略的评估不仅应考虑回报和置换鲁棒性，还应考虑其是否保持非崩溃的、差异化的多智能体行为。
cs.RO / 35 / 2610.02874
SceneFactory-3D: Lifting 2D Traffic Scenes into 3D Physical Counterfactuals for Scalable Physically Grounded Safety Evaluation
SceneFactory-3D：将2D交通场景提升为3D物理反事实，以实现可扩展的物理基础安全评估
Abstract
Scalable driving simulators typically execute vehicle commands using prescribed behavioral or kinematic rules, overlooking the physics of tire-road interfaces, thereby limiting their ability to capture how adverse road and environmental conditions alter vehicle execution and propagate through traffic. To address this limitation, we present SceneFactory-3D, a GPU-batched, physics-grounded multi-agent driving simulator. Vehicles execute acceleration and steering commands via suspension- and friction-limited forces evaluated at each wheel-contact point. Spatially varying friction, per-world 3D heightfields, gravity, and rigid contact consistently govern wheel motion and chassis collisions. Per-world terrain isolation and GPU batching enable SceneFactory-3D to run matched physical counterfactuals in parallel: traffic scenario setup and vehicle controllers remain fixed while only the road condition changes, enabling the resulting closed-loop effects to be evaluated across parallel worlds. To demonstrate the advantage of the SceneFactory-3D-enabled counterfactual evaluation, we conduct an empirical study on vehicle controllers' sensitivity to road conditions. We study three learned-policy families on 1,024 matched 12-vehicle worlds per condition, and two classical planners on a shared 32-world subset, across 21 friction and grade conditions. When friction drops from 1.0 to 0.18, the share of vehicles that clear the work zone safely falls by 6 to 90 percentage points across learned policies (18-19 for classical planners), and near-collision situations become more frequent for every learned policy. Code: https://github.com/SmallWorldLab/SceneFactory_3D
Chinese Translation
可扩展的驾驶模拟器通常使用规定的行为或运动学规则执行车辆指令，忽视了轮胎与路面接口的物理特性，从而限制了它们捕捉不利路况和环境条件如何改变车辆执行并在交通中传播的能力。为了解决这一局限性，我们提出了SceneFactory-3D，这是一种基于GPU批处理的物理基础多智能体驾驶模拟器。车辆通过在每个轮接触点评估的悬挂和摩擦限制力执行加速和转向指令。空间变化的摩擦、每个世界的3D高度场、重力和刚性接触始终支配着轮子运动和底盘碰撞。每个世界的地形隔离和GPU批处理使得SceneFactory-3D能够并行运行匹配的物理反事实：交通场景设置和车辆控制器保持不变，而只有路况发生变化，从而使得结果的闭环效应能够在平行世界中进行评估。为了展示SceneFactory-3D启用的反事实评估的优势，我们对车辆控制器对路况的敏感性进行了实证研究。我们在每种条件下研究了三种学习策略家族在1,024个匹配的12辆车世界中的表现，以及在共享的32个世界子集上研究了两种经典规划器，涵盖了21种摩擦和坡度条件。当摩擦从1.0降至0.18时，安全通过施工区域的车辆比例在学习策略中下降了6到90个百分点（经典规划器为18-19个百分点），而每种学习策略的近碰撞情况变得更加频繁。代码：https://github.com/SmallWorldLab/SceneFactory_3D
cs.RO / 36 / 2610.02898
MixVLA: Adaptive Mixing of Non-Invariant Information for Generalizable Vision-Language-Action Models
MixVLA：用于可泛化视觉-语言-动作模型的非不变信息自适应混合
Abstract
Vision-Language-Action (VLA) models have achieved remarkable advances in robotic manipulation, yet their zero-shot generalization under out-of-distribution (OOD) conditions remains limited. These models often entangle task-relevant invariant structure with environment-specific non-invariant factors, causing policies to rely on spurious appearance cues during action prediction. In this work, we propose \textbf{MixVLA}, a model-agnostic training framework that improves the generalization of VLA models without requiring additional OOD data or architectural modifications. The key component of MixVLA is \textbf{Adaptive Mixing of Non-Invariant Information (AMI)}. AMI stochastically mixes non-invariant representations to regularize distribution-specific variability while preserving complementary predictive cues. The mixed non-invariant features are then fused with invariant representations for final action prediction, resulting in improved robustness without sacrificing policy expressiveness. Extensive experiments across challenging manipulation settings, including LIBERO, LIBERO-Plus, the RoboTwin perturbation suite, and real-world tasks, demonstrate that MixVLA improves overall zero-shot robustness while retaining strong in-domain performance.
Chinese Translation
视觉-语言-动作（VLA）模型在机器人操控方面取得了显著进展，但在分布外（OOD）条件下的零-shot 泛化能力仍然有限。这些模型通常将与任务相关的不变结构与特定环境的非不变因素纠缠在一起，导致策略在动作预测时依赖于虚假的外观线索。在本研究中，我们提出了 extbf{MixVLA}，一种模型无关的训练框架，能够在不需要额外的 OOD 数据或架构修改的情况下提高 VLA 模型的泛化能力。MixVLA 的关键组成部分是 extbf{非不变信息自适应混合（AMI）}。AMI 随机混合非不变表示，以规范分布特定的变异性，同时保留互补的预测线索。混合后的非不变特征随后与不变表示融合，以进行最终的动作预测，从而在不牺牲策略表达能力的情况下提高了鲁棒性。在包括 LIBERO、LIBERO-Plus、RoboTwin 扰动套件和现实世界任务等具有挑战性的操控设置下的广泛实验表明，MixVLA 在保持强大领域内性能的同时，提高了整体的零-shot 鲁棒性。
cs.RO / 37 / 2610.02974
From Language Priors to Field Adaptation: Preference Learning for Traversability Estimation
从语言先验到领域适应：可通行性估计的偏好学习
Abstract
Image-based traversability estimation is inherently dependent on the robot platform, deployment domain, and mission preferences, which limits the applicability of purpose-trained models. To facilitate domain adaptation, this work aims to reduce the number of required annotations in the target domain using sample-efficient preference learning. Our method represents traversability through von Mises-Fisher mixture prototypes in a frozen vision-language feature space. Relative natural-language rules provide a commonsense prior, while sparse relative image annotations adapt the prototype directions and utilities to a target domain through computationally and sample-efficient fine-tuning. Experiments on WayFAST demonstrate accuracy competitive with end-to-end trained estimators while enabling sample-efficient image-based adaptation. Qualitative experiments further demonstrate the language prior's zero shot applicability and the fine-tuned estimator's improved dense prediction on semantic maps. Evaluation is complemented via semantic interpretation of learned prototypes by dissecting semantically close natural language prompts. Code and trained estimators available at https://resireg.github.io
Chinese Translation
基于图像的可通行性估计本质上依赖于机器人平台、部署领域和任务偏好，这限制了专门训练模型的适用性。为了促进领域适应，本研究旨在通过样本高效的偏好学习减少目标领域所需的标注数量。我们的方法通过在冻结的视觉-语言特征空间中使用冯·米塞斯-费舍尔混合原型来表示可通行性。相对自然语言规则提供了常识先验，而稀疏的相对图像标注则通过计算和样本高效的微调将原型方向和效用适应到目标领域。在 WayFAST 上的实验表明，我们的方法在准确性上与端到端训练的估计器具有竞争力，同时实现了样本高效的基于图像的适应。定性实验进一步展示了语言先验的零样本适用性以及微调估计器在语义地图上的密集预测能力的提升。通过剖析语义相近的自然语言提示，对学习到的原型进行语义解释以补充评估。代码和训练好的估计器可在 https://resireg.github.io 获取。
cs.RO / 38 / 2610.02985
On Representational Alignment among Embodied Agents
关于具身智能体之间的表征一致性
Abstract
Embodied agents interacting with the same physical process may maintain heterogeneous, asynchronous, and observer-relative representations. Rather than assuming that such representations should always be globally aligned, we investigate which distinctions among them must actually be resolved for coherent interaction. We formalize this question through relational sufficiency: an interaction-specific relational-evidence map defines the representational ambiguity left unresolved by comparison, and sufficiency holds when each residual ambiguity fiber lies within an interaction-equivalence class. When this condition fails, additional sensorimotor anchors must separate precisely the ambiguities that can change the interaction-relevant outcome. For smooth systems, we derive a local first-order lower bound on the number of inde- pendent scalar anchors required, given by the dimension of directions that are invisible to relational evidence but visible to the interaction-relevant map. We instantiate the theory in two variants of an asynchronous hand-off scenario that share the same observer agents, representations, transports, and relational evidence. Relative synchronization requires no additional anchor, whereas hand-off at an absolute time requires exactly one. The analysis there- fore provides a task-relative stopping condition for representational alignment: residual disagreement need not be eliminated when it is invariant with respect to the interaction.
Chinese Translation
具身智能体在相同物理过程中互动时，可能维持异质的、异步的以及观察者相对的表征。我们并不假设这些表征总是应该全局一致，而是探讨在什么情况下它们之间的区别必须被解决以实现一致的互动。我们通过关系充分性来形式化这个问题：一个特定于互动的关系证据图定义了通过比较未解决的表征模糊性，当每个残余模糊性纤维位于一个互动等价类内时，充分性成立。当这一条件失效时，必须额外的感知运动锚点精确地区分那些可能改变互动相关结果的模糊性。对于平滑系统，我们推导出所需独立标量锚点数量的局部一阶下界，该下界由对关系证据不可见但对互动相关图可见的方向的维度给出。我们在两个变体的异步交接场景中实例化该理论，这两个场景共享相同的观察者智能体、表征、传输和关系证据。相对同步不需要额外的锚点，而在绝对时间下的交接则恰好需要一个。因此，该分析为表征一致性提供了一个任务相关的停止条件：当残余分歧与互动无关时，不必消除它。
cs.RO / 39 / 2610.02991
Subject-Specific Predictive Musculoskeletal Simulations of Lower-Limb Exoskeleton Assistance: Metabolic and Biomechanical Effects of Joint Assistance Strategies
特定主题的下肢外骨骼辅助的预测性肌肉骨骼模拟：关节辅助策略的代谢和生物力学效应
Abstract
Lower-limb exoskeletons have made considerable progress in reducing energy expenditure during walking. However, designing optimal assistance strategies remains challenging, particularly given inter-individual variability in anthropometry and biomechanics. This study explores energy-optimal lower-limb joint-assistance strategies using predictive simulations with musculoskeletal models. Subject-specific models of six able-bodied subjects, with BMI-based muscle strength scaling, were used in predictive simulations to generate gait at self-selected walking speeds. Ideal actuators were incorporated to simulate various combinations of joint assistance at peak levels of 25 Nm and 50 Nm to examine the effects of assistance on gait and metabolic savings. The effects of each assistance configuration were assessed through cost of transport (COT), joint kinematics, muscle activations, assistive torques, and joint-level power metrics to characterize the biomechanical and energetic impacts of different assistance strategies. At 50 Nm, combined H+K+A (hip-knee-ankle) assistance resulted in the greatest mean COT reduction of 48.50 +/- 4.55%, with individual reductions ranging from 42.22% to 54.65% across the subjects. Among single-joint conditions, assisting the hip was most effective, reducing COT by 31.77 +/- 5.21%; the knee and ankle produced smaller, comparable reductions (19.17% and 17.02%). H+A (hip-ankle) assistance (44.60 +/- 7.08%) emerged as the most effective two-joint assistive configuration. Increasing the torque bound increased positive assistive power primarily at the hip and ankle, while knee assistance showed little sensitivity and delivered positive power near pre-swing. These results support H+A assistance as an efficient two-actuator target, while identifying the knee's atypical pre-swing power strategy as a candidate for targeted experimental validation.
Chinese Translation
下肢外骨骼在减少步行过程中的能量消耗方面取得了显著进展。然而，设计最佳的辅助策略仍然具有挑战性，特别是在个体之间的人体测量和生物力学差异的情况下。本研究利用肌肉骨骼模型的预测性模拟探索能量最优的下肢关节辅助策略。使用六名健康受试者的特定模型，并基于BMI进行肌肉力量缩放，通过预测性模拟生成自选步行速度下的步态。理想的执行器被纳入，以模拟在25 Nm和50 Nm的峰值水平下的各种关节辅助组合，以研究辅助对步态和代谢节省的影响。通过运输成本（COT）、关节运动学、肌肉激活、辅助扭矩和关节级功率指标评估每种辅助配置的效果，以表征不同辅助策略的生物力学和能量影响。在50 Nm时，联合H+K+A（髋-膝-踝）辅助导致平均COT减少48.50 +/- 4.55%，个体减少范围为42.22%至54.65%。在单关节条件中，髋部辅助最为有效，COT减少31.77 +/- 5.21%；膝部和踝部的减少幅度较小，分别为19.17%和17.02%。H+A（髋-踝）辅助（44.60 +/- 7.08%）成为最有效的双关节辅助配置。增加扭矩限制主要在髋部和踝部增加了正辅助功率，而膝部辅助表现出较小的敏感性，并在预摆动阶段提供了正功率。这些结果支持H+A辅助作为有效的双执行器目标，同时将膝部的非典型预摆动功率策略识别为针对性实验验证的候选者。
cs.RO / 40 / 2610.03132
Safe Streaming Flow Planning by Aligning Sampling Dynamics with Execution Dynamics
通过对齐采样动态与执行动态实现安全流规划
Abstract
Generative planners based on diffusion/flow matching can learn to synthesize long-horizon trajectories from demonstrations. However, real-world deployment requires (i) enforcing safety constraints during execution and (ii) tight online replanning at fast execution rates. Prior safe diffusion/flow planners generate the agent's full trajectory at once, while repeatedly perturbing intermediate states to satisfy safety constraints. This approach is not only computationally intensive, but also introduces distribution shift since the learned sampling dynamics is distinct from the system's execution dynamics. We propose SafeStreamingFlow, a goal-conditioned planner that aligns flow sampling dynamics with execution dynamics by sequentially integrating a learned state vector field with hierarchical state prediction. Importantly, we need to enforce safety constraints only for the executed step via high order control barrier functions. Across navigation, racing, and locomotion benchmarks, SafeStreamingFlow reduces planning latency and improves safety compared to existing methods, while maintaining competitive goal-reaching success.
Chinese Translation
基于扩散/流匹配的生成规划器能够从示例中学习合成长时间跨度的轨迹。然而，实际部署要求（i）在执行过程中强制执行安全约束，以及（ii）在快速执行速率下进行紧凑的在线重新规划。之前的安全扩散/流规划器一次性生成代理的完整轨迹，同时反复扰动中间状态以满足安全约束。这种方法不仅计算密集，而且由于学习到的采样动态与系统的执行动态不同，导致了分布转移。我们提出了SafeStreamingFlow，一种目标条件规划器，通过顺序整合学习到的状态向量场与层次状态预测，将流采样动态与执行动态对齐。重要的是，我们只需通过高阶控制障碍函数在执行步骤中强制执行安全约束。在导航、赛车和运动基准测试中，SafeStreamingFlow相比现有方法减少了规划延迟并提高了安全性，同时保持了竞争性的目标达成成功率。
cs.RO / 41 / 2610.03196
Beyond Reward Hacking: Proxy Divergence Across Four Layers of a Staged Humanoid Learning Pipeline
超越奖励黑客：分阶段类人学习管道四层的代理偏差
Abstract
A reinforcement-learning (RL) pipeline for a legged robot is assembled from proxies. A reward stands in for intended behaviour, a curriculum gate stands in for competence, an evaluation statistic stands in for robustness, and a reference motion stands in for an achievable skill. The traditional view treats only the first of these as optimised against, and so locates specification failure (reward hacking) in the reward alone. I argue that all four are proxies in the same formal sense, that each has a characteristic divergence mechanism, and that each admits a reformulation that closes it. For every layer I state the traditional formulation, derive the condition under which it diverges from its target, and give the alternative: first-order (L1) costs where quadratic kernels are flat, peak and outcome statistics where curriculum gates average, gate reachability and information checks, deterministic and phase-desynchronised evaluation, curriculum state treated as part of the model, feasibility-first reference design with residual feed-forward, and function-preserving input widening that lets one policy grow instead of being retrained. The arguments are illustrated by measurements from one continuous lineage of a PPO policy for a simulated 1.91 m humanoid, grown over four stages and 13,500 iterations on a single laptop GPU. Among them, a curriculum gate built on averaged error advanced at its rate limit on every check while the skill it gated was absent, and a batched push test whose synchronised resets aliased the gait phase ranked a 0.5 m/s push as more dangerous than a 2.0 m/s one.
Chinese Translation
一个用于四足机器人的强化学习（RL）管道由多个代理组成。奖励代表预期行为，课程门代表能力，评估统计量代表鲁棒性，而参考运动代表可实现的技能。传统观点仅将第一个视为优化对象，因此将规范失败（奖励黑客）仅归因于奖励。我认为这四者在形式上都是代理，每个都有特征性的偏差机制，并且每个都可以重新表述以使其闭合。对于每一层，我陈述传统的表述，推导出其偏离目标的条件，并给出替代方案：在二次核平坦的情况下的一阶（L1）成本，课程门平均的峰值和结果统计，门的可达性和信息检查，确定性和相位不同步的评估，将课程状态视为模型的一部分，具有残差前馈的可行性优先参考设计，以及允许一个策略增长而不是重新训练的保持功能的输入扩展。通过对一个模拟的1.91米类人机器人在单个笔记本GPU上经过四个阶段和13,500次迭代的PPO策略的连续谱系的测量，说明了这些论点。在这些测量中，一个基于平均误差构建的课程门在每次检查时以其速率限制推进，而它所控制的技能却缺失；一个批量推送测试，其同步重置与步态相位混叠，将0.5米/秒的推送评估为比2.0米/秒的推送更危险。
cs.RO / 42 / 2610.03278
DexJoCo-X: Benchmarking Action Representations for Multi-Hand Dexterous Manipulation
DexJoCo-X：多手灵巧操控的动作表示基准测试
Abstract
As dexterous hands proliferate, collecting data and training policies separately for every morphology becomes increasingly impractical. Scalable cross-embodiment learning therefore requires a unified representation that captures shared manipulation structure while preserving morphology-specific control. Differences in hands, tasks, datasets, and control interfaces prevent existing studies from isolating the effects of representation, pretraining, and architecture. We introduce DexJoCo-X, a benchmark and toolkit for controlled comparison across seven representative dexterous hands, six single-arm and bimanual tasks, and 2,100 balanced demonstrations. DexJoCo-X provides a matched multi-hand, multi-task protocol with common scenes, success criteria, and execution interfaces, redesigned glove-to-hand mappings, and an automated pipeline that expands reviewed demonstrations across randomized scenes. Using $\pi_{0.5}$, Ego-Pi, and Being-H0.5, we examine whether a shared action interface is sufficient for multi-hand learning. Expanding $\pi_{0.5}$ to an 80-dimensional bimanual output yields near-zero success. Ego-Pi preserves the pretrained action head through interleaved prediction and supports per-hand multi-task learning, but remains ineffective for seven-hand joint training. By contrast, Being-H0.5 combines cross-embodiment pretraining, a unified action space, and embodiment-aware experts, enabling one policy to control all seven hands. Within this architecture, function-aligned action slots achieve 47.7% mean success, compared with 47.0% for native coordinates and 33.1% for DexLatent. These results show that cross-embodiment representation depends on the entire learning system: action coordinates, pretraining, and architecture must jointly separate shared manipulation structure from embodiment-specific control.
Chinese Translation
随着灵巧手的普及，为每种形态单独收集数据和训练策略变得越来越不切实际。因此，可扩展的跨形态学习需要一种统一的表示，能够捕捉共享的操控结构，同时保留特定形态的控制。手部、任务、数据集和控制接口之间的差异使得现有研究无法孤立出表示、预训练和架构的影响。我们介绍了DexJoCo-X，这是一个用于在七种代表性灵巧手、六个单臂和双臂任务以及2100个平衡演示之间进行受控比较的基准和工具包。DexJoCo-X提供了一个匹配的多手、多任务协议，具有共同的场景、成功标准和执行接口，重新设计的手套到手的映射，以及一个自动化管道，能够在随机场景中扩展审查过的演示。使用$ ext{π}_{0.5}$、Ego-Pi和Being-H0.5，我们考察了共享动作接口是否足以支持多手学习。将$ ext{π}_{0.5}$扩展到80维的双手输出几乎没有成功。Ego-Pi通过交错预测保留了预训练的动作头，并支持每手的多任务学习，但在七手联合训练中仍然无效。相比之下，Being-H0.5结合了跨形态预训练、统一的动作空间和形态感知专家，使得一个策略能够控制所有七只手。在该架构中，功能对齐的动作槽实现了47.7%的平均成功率，而原生坐标为47.0%，DexLatent为33.1%。这些结果表明，跨形态表示依赖于整个学习系统：动作坐标、预训练和架构必须共同将共享的操控结构与特定形态的控制分离。
cs.RO / 43 / 2610.03283
HexVIO: Towards All-Day Stereo-Inertial Tracking Through Commodity DSPs
HexVIO：通过商用数字信号处理器实现全天候立体惯性跟踪
Abstract
The ability of a device to localize itself within its surroundings is a fundamental prerequisite for spatial computing. Visual-inertial odometry (VIO) has proven to be a cost-effective and accurate solution for this task. Robots, wearables, XR devices, and drones can benefit significantly from efficient implementations of VIO since they allow for cooler, lighter, and cheaper devices with longer battery life and a better user experience. In this work, we propose to enhance the efficiency of a VIO system by leveraging the Hexagon DSP, a commodity co-processor present in many modern smartphones and XR devices. Our approach offloads the visual frontend of a stereo-inertial odometry system to the DSP while keeping the backend on the main CPU. By optimizing the implementation for the DSP architecture, we achieve significant reductions in power consumption and latency compared to CPU-only execution. Our system, HexVIO, demonstrates a 67% reduction in power consumption or an 86% increase in throughput on a commodity smartphone, with the ability to sustain long-term real-time 30 fps tracking for 0.83 W, corresponding to ~18 hours of tracking on the testing device. These results highlight the potential of commodity DSPs for enabling all-day visual-inertial tracking in robotics and mobile devices.
Chinese Translation
设备在其环境中进行自我定位的能力是空间计算的基本前提。视觉惯性里程计（VIO）已被证明是一种成本效益高且准确的解决方案。机器人、可穿戴设备、扩展现实（XR）设备和无人机可以显著受益于VIO的高效实现，因为这使得设备更轻、更便宜，电池寿命更长，用户体验更佳。在本研究中，我们提出通过利用Hexagon DSP（数字信号处理器）来提高VIO系统的效率，该处理器是许多现代智能手机和XR设备中存在的商用协处理器。我们的方法将立体惯性里程计系统的视觉前端卸载到DSP上，同时将后端保留在主CPU上。通过针对DSP架构优化实现，我们在功耗和延迟方面相较于仅使用CPU的执行实现了显著降低。我们的系统HexVIO在一款商用智能手机上实现了67%的功耗降低或86%的吞吐量提升，能够以0.83 W的功耗持续进行实时30 fps跟踪，相当于在测试设备上约18小时的跟踪时间。这些结果突显了商用DSP在机器人和移动设备中实现全天候视觉惯性跟踪的潜力。
cs.RO / 44 / 2610.03333
Equivariant Visual-Tactile Diffusion Policy for Contact-Rich Manipulation
等变视觉-触觉扩散策略用于接触丰富的操控
Abstract
Imitation learning for contact-rich manipulation requires high-quality expert data that is expensive to obtain. This makes learning a sample-efficient policy a key issue. To address this, we propose VISTA, a workspace-level equivariant visuotactile diffusion policy for data-efficient contact-rich imitation learning. VISTA projects visual and tactile observations into spherical tokens, injects tactile contact cues into visual spherical directions through permutation-equivariant spherical fusion, and rotates the fused harmonic representation using the end-effector orientation. The resulting representation conditions an equivariant diffusion policy to predict spatially consistent actions. Extensive experiments in both simulation and real-world robotic settings show that VISTA substantially improves data efficiency over strong visuotactile imitation learning baselines. Project website: https://vista-paper.github.io/
Chinese Translation
接触丰富的操控的模仿学习需要高质量的专家数据，而这些数据的获取成本高昂。这使得学习一种样本高效的策略成为关键问题。为了解决这一问题，我们提出了VISTA，一种用于数据高效的接触丰富模仿学习的工作空间级别的等变视觉-触觉扩散策略。VISTA将视觉和触觉观测投影到球形标记中，通过置换等变的球形融合将触觉接触线索注入视觉球形方向，并使用末端执行器的方向旋转融合的谐波表示。最终的表示条件化了一个等变扩散策略，以预测空间一致的动作。在模拟和真实世界机器人环境中的大量实验表明，VISTA在数据效率上显著优于强大的视觉-触觉模仿学习基线。项目网站：https://vista-paper.github.io/
cs.RO / 45 / 2610.03388
KungfuAthleteBot: learning high-dynamic humanoid motion from video with unified robust recovery
KungfuAthleteBot：通过统一的鲁棒恢复从视频中学习高动态人形运动
Abstract
Video is an abundant, inexpensive source of human motion data that is rich in extreme athletic behaviors. Making it usable for humanoid robots, however, is not a matter of simply retargeting a reconstructed trajectory: video-derived motion is physically inconsistent, devoid of actuation information, and says nothing about failure or recovery. We present KungfuAthleteBot (KAB), a framework that treats learning high-dynamic motion from video as the central problem and resolves each of these three failure modes in turn. (C1) We build the KungfuAthlete dataset from videos of national-level martial artists and introduce a physics-guided parabolic trajectory correction that removes height floating, ground penetration, and high-frequency jitter from reconstructed aerial and landing phases. (C2) Because video carries no force information, strict tracking of a reconstructed trajectory is dynamically infeasible, and error-driven initialization keeps re-launching the policy from infeasible aerial poses. We introduce physics-driven pseudo-low-kinetic-energy (LKE) sampling, our central mechanism for making such references learnable: it biases initialization towards dynamically feasible states, letting the policy discover feasible actuation patterns instead of imitating infeasible ones. (C3) Finally, we introduce a direct training paradigm in which disturbance rejection and fall recovery are learned inside the same policy that tracks the video motion, requiring no recovery reference data and no manual mode switching. On a humanoid robot, KAB learns dynamic skills from video and recovers from arbitrary falls in about 0.7 s, the fastest reported recovery for a unified policy. Ablations on the unified policy confirm the necessity of its components, supporting the view that repairing and compensating video data, rather than only collecting more of it, is what unlocks high-dynamic humanoid skills.
Chinese Translation
视频是人类运动数据的丰富且廉价的来源，包含极限运动行为。然而，使其可用于人形机器人并非仅仅是重新定向重建的轨迹：视频衍生的运动在物理上不一致，缺乏驱动信息，并且未涉及失败或恢复。我们提出了KungfuAthleteBot (KAB)框架，将从视频中学习高动态运动视为核心问题，并逐一解决这三种失败模式。(C1) 我们从国家级武术运动员的视频中构建了KungfuAthlete数据集，并引入了一种物理引导的抛物线轨迹修正，消除了重建的空中和着陆阶段中的高度漂浮、地面穿透和高频抖动。(C2) 由于视频不携带力的信息，严格跟踪重建轨迹在动态上是不可行的，基于误差的初始化不断从不可行的空中姿态重新启动策略。我们引入了物理驱动的伪低动能（LKE）采样，这是使这些参考可学习的核心机制：它将初始化偏向于动态可行状态，使策略能够发现可行的驱动模式，而不是模仿不可行的模式。(C3) 最后，我们引入了一种直接训练范式，其中干扰拒绝和跌倒恢复在跟踪视频运动的同一策略中学习，无需恢复参考数据和手动模式切换。在人形机器人上，KAB能够从视频中学习动态技能，并在约0.7秒内从任意跌倒中恢复，这是统一策略中报告的最快恢复时间。对统一策略的消融实验确认了其组件的必要性，支持了修复和补偿视频数据，而不仅仅是收集更多数据，是解锁高动态人形技能的关键。
cs.RO / 46 / 2610.03453
I2CD: Direct Image-to-Convex Decomposition for Simulation-Ready Collision Geometry
I2CD：用于仿真准备的直接图像到凸分解的碰撞几何
Abstract
Physics simulators and motion planners require convex collision geometry, yet image-to-3D generative models output dense, frequently non-manifold visual meshes. Bridging the two today takes a slow, brittle reconstruct-then-decompose pipeline of repair, decimation, and approximate convex decomposition. We present I2CD, which predicts a convex decomposition directly from a single RGB image. Rather than train a new image-to-3D model, I2CD freezes the pretrained Hunyuan3D-2 image-conditioned diffusion transformer and shape decoder and trains only a lightweight cross-attention head (38M parameters, under ten GPU-hours) whose learned "convex-slot" tokens emit the halfplane parameters of $K$ convex polytopes. The output is compact, convex by construction, and loads into physics engines without any post-processing, in ${\sim}0.5$s per image. On $227$ held-out OmniObject3D and Google Scanned Objects instances, I2CD attains the highest volumetric IoU among eight reconstruct-then-decompose pipelines while running $6$-$37\times$ faster end-to-end. In a cross-simulator study in MuJoCo, PyBullet, Genesis, and Isaac Sim, every engine uses I2CD geometry as delivered, whereas raw generated meshes "load" everywhere but are silently replaced by a different collision shape in most cases or need seconds to minutes of per-object preprocessing. On a physical xArm7, I2CD produces planner-ready geometry for a $20$-object cluttered scene in $11$s versus $328$s for the strongest baseline, at comparable pick-and-place execution success ($85$ vs. $90$ of $100$ trials).
Chinese Translation
物理模拟器和运动规划器需要凸碰撞几何，而图像到3D生成模型输出的是密集且常常是非流形的视觉网格。当前将两者结合的过程缓慢且脆弱，通常需要经过修复、简化和近似凸分解的重建-再分解管道。我们提出了I2CD，它可以直接从单张RGB图像预测凸分解。I2CD并不是训练一个新的图像到3D模型，而是冻结预训练的Hunyuan3D-2图像条件扩散变换器和形状解码器，仅训练一个轻量级的交叉注意力头（38M参数，少于十个GPU小时），其学习的“凸槽”标记发出$K$个凸多面体的半平面参数。输出是紧凑的，构造上是凸的，并且可以在物理引擎中加载而无需任何后处理，处理时间约为每张图像${ ext{∼}}0.5$s。在227个保留的OmniObject3D和Google扫描对象实例上，I2CD在八个重建-再分解管道中获得了最高的体积IoU，同时端到端运行速度快了$6$-$37 imes$。在MuJoCo、PyBullet、Genesis和Isaac Sim的跨模拟器研究中，每个引擎都使用I2CD几何体，而原始生成的网格“加载”到处都是，但在大多数情况下会被静默替换为不同的碰撞形状，或者需要每个对象几秒到几分钟的预处理。在物理xArm7上，I2CD在$11$s内为一个包含$20$个物体的杂乱场景生成规划器准备好的几何体，而最强基线需要$328$s，且在可比的抓取和放置执行成功率上（$100$次试验中分别为$85$和$90$）相当。
cs.RO / 47 / 2610.03476
MobiAgent: Dual-Loop Recursive Policy Self-Improvement for Long-Horizon Mobile Manipulation
MobiAgent：用于长时间移动操控的双循环递归策略自我改进
Abstract
Long-horizon mobile manipulation presents significant challenges due to compounding execution errors and capacity interference between locomotion and arm control. While recent Vision-Language-Action models excel at short-horizon tasks, they lack the hierarchical reasoning required for multi-stage objectives. Furthermore, existing hierarchical agents suffer from rigid sub-task mapping, inflexible replanning, and a lack of continuous learning. To address these limitations, we introduce MobiAgent, a dual-loop agentic framework that bridges robust deployment execution and recursive policy self-improvement. During deployment, the Inner Loop decouples high-level reasoning from low-level control through highly composable atomic skills. It employs Vision-Language models for receding-horizon planning and visual reflection, dynamically composing skills to ensure robust error recovery. These skills are executed by specialized flow-matching experts that share a unified VLM backbone, maximizing reusability while mitigating capacity interference. Concurrently, the Outer Loop drives automated lifelong learning by autonomously segmenting and verifying deployment rollouts, clustering them to discover atomic skills, and continuously fine-tuning the skill library without human annotations. Evaluations on RoboCasa, BEHAVIOR-1K, and real-world tasks demonstrate the effectiveness of MobiAgent. It outperforms $\pi_{0.5}$-TA by 22.5 percentage points on BEHAVIOR-1K and enables robust recovery from execution failures. Through autonomous data recycling, success improves from 7.50% to 27.50% on RoboCasa and from 32.5% to 57.5% on Astribot S1.
Chinese Translation
长时间移动操控面临显著挑战，这主要源于执行错误的累积以及移动和手臂控制之间的能力干扰。尽管最近的视觉-语言-动作模型在短时间任务中表现出色，但它们缺乏实现多阶段目标所需的层次推理。此外，现有的层次代理存在子任务映射僵化、重新规划不灵活以及缺乏持续学习的问题。为了解决这些局限性，我们提出了MobiAgent，这是一种双循环代理框架，旨在连接稳健的部署执行与递归策略自我改进。在部署过程中，内循环通过高度可组合的原子技能将高层推理与低层控制解耦。它利用视觉-语言模型进行逐步规划和视觉反思，动态组合技能以确保稳健的错误恢复。这些技能由共享统一视觉-语言模型（VLM）主干的专业流匹配专家执行，最大限度地提高了可重用性，同时减轻了能力干扰。同时，外循环通过自主分割和验证部署展开，聚类以发现原子技能，并在没有人工标注的情况下持续微调技能库，从而推动自动化的终身学习。在RoboCasa、BEHAVIOR-1K和真实世界任务上的评估表明MobiAgent的有效性。在BEHAVIOR-1K上，它比$ ext{π}_{0.5}$-TA提高了22.5个百分点，并能够在执行失败后实现稳健恢复。通过自主数据回收，RoboCasa上的成功率从7.50%提高到27.50%，而在Astribot S1上的成功率从32.5%提高到57.5%。
cs.RO / 48 / 2610.03498
Detect and Suppress: A Mechanistic Defense against Adversarial Patches in VLA Models
检测与抑制：针对VLA模型中对抗性补丁的机制性防御
Abstract
Adversarial patches can disrupt Vision-Language-Action (VLA) models by manipulating visual observations, leading to failures in robot control. However, it remains poorly understood which internal mechanisms underlie these failures and how targeted interventions can mitigate them. In this work, we mechanistically analyze VLA representations using a sparse autoencoder (SAE) and identify a feature whose activation strongly correlates with the presence of an adversarial patch. Based on this analysis, we suppress the identified feature at inference time only when a linear probe detects an attack. This intervention improves robustness without the cost of fine-tuning the VLA. We evaluate our method against VLA adversarial patch attacks on LIBERO-10. Conditional intervention improves success rate under intermittent attacks, whereas continuously applying the same intervention substantially degrades policy performance. These results show that attack-related internal representations can provide useful targets for VLA adversarial defense and that controlling when to intervene is important for limiting disruption to nominal policy behavior.
Chinese Translation
对抗性补丁可以通过操控视觉观察来干扰视觉-语言-行动（VLA）模型，导致机器人控制的失败。然而，目前尚不清楚这些失败背后的内部机制是什么，以及如何通过有针对性的干预来减轻这些问题。在本研究中，我们利用稀疏自编码器（SAE）对VLA表示进行机制性分析，识别出一种与对抗性补丁存在强相关性的特征激活。基于这一分析，我们在推理时仅在线性探测器检测到攻击时抑制该特征。这一干预在不需要微调VLA的情况下提高了鲁棒性。我们在LIBERO-10上评估了我们的方法，针对VLA对抗性补丁攻击进行了测试。条件干预在间歇性攻击下提高了成功率，而持续应用相同的干预则显著降低了策略性能。这些结果表明，与攻击相关的内部表示可以为VLA对抗防御提供有用的目标，并且控制何时进行干预对于限制对正常策略行为的干扰至关重要。
cs.RO / 49 / 2610.03516
XGenAct: Geometry-Enhanced World Action Models through Cross-Task Generation
XGenAct：通过跨任务生成增强几何的世界动作模型
Abstract
World action models (WAMs) have advanced robot control by predicting how observations and actions evolve over time. Despite this progress, RGB and action based future prediction does not explicitly address the spatial understanding needed for robot manipulation. Existing efforts often add a limited set of spatial prediction tasks through specialized heads or branches, leaving both the range of spatial supervision and the model architecture fragmented. We introduce XGenAct, a world action model that represents RGB observations, robot actions, metric depth, surface normals, and functional role segmentation as RGB videos through deterministic codecs. By sampling perception and action tasks during training, XGenAct uses one video diffusion transformer and one objective to learn temporal prediction across these spaces without modality specific learned heads. On held out RLBench tasks, structured perception training improves average closed loop success over RGB only training, and XGenAct achieves 52% success in the five task external comparison, versus 26% for the strongest evaluated baselines. It also predicts future depth and segmentation more accurately than the evaluated pipelines that generate RGB first and then apply a frozen perception expert.
Chinese Translation
世界动作模型（WAMs）通过预测观察和动作随时间的演变，推动了机器人控制的发展。尽管取得了这些进展，基于RGB和动作的未来预测并未明确解决机器人操作所需的空间理解。现有的努力通常通过专门的头部或分支添加有限的空间预测任务，这使得空间监督的范围和模型架构都显得支离破碎。我们提出了XGenAct，这是一种世界动作模型，通过确定性编解码器将RGB观察、机器人动作、度量深度、表面法线和功能角色分割表示为RGB视频。通过在训练过程中对感知和动作任务进行采样，XGenAct使用一个视频扩散变换器和一个目标，在这些空间中学习时间预测，而无需特定模态的学习头。在保留的RLBench任务上，结构化感知训练提高了相较于仅使用RGB训练的平均闭环成功率，XGenAct在五个任务的外部比较中取得了52%的成功率，而最强评估基线的成功率为26%。此外，它在未来深度和分割的预测上也比评估的流水线更为准确，这些流水线首先生成RGB，然后应用一个固定的感知专家。
cs.RO / 50 / 2610.03530
DR-IPC: Disturbance-Resilient Integrated Planning and Control for LiDAR-Based Quadrotor Navigation
DR-IPC：基于LiDAR的四旋翼导航的抗干扰集成规划与控制
Abstract
LiDAR-based quadrotor navigation in cluttered environments remains challenging under external disturbances, particularly when obstacle-aware motion generation and disturbance-rejection control are handled in separate layers. This article presents disturbance-resilient integrated planning and control (DR-IPC), which combines lightweight path guidance with nonlinear model predictive control (NMPC) to directly generate angular velocity and thrust. An interconnected extended Kalman filter and nonlinear disturbance observer jointly provide filtered state estimates and reconstructed disturbances for NMPC prediction. The resulting formulation unifies nonlinear quadrotor dynamics, actuator constraints, local motion generation, and penalised safe-flight-corridor residuals without requiring a separate trajectory-optimization stage. Gazebo and MARSIM simulations, together with indoor and outdoor experiments, validate DR-IPC under wind, suspended payloads, narrow passages, ball impacts and reactive avoidance of a dynamic obstacle. In multi-goal navigation with disturbances, DR-IPC increases the number of completed missions from 1/10 to 9/10 in Gazebo and reduces the altitude RMSE from 0.34 to 0.01 m in experiments. The complete system operates onboard at 100 Hz. Supplementary videos are available on the project page https://drpp316.github.io/DR-IPC-Page/, and the source code will be released.
Chinese Translation
在复杂环境中，基于LiDAR的四旋翼导航在外部干扰下仍然面临挑战，特别是在障碍物感知的运动生成和抗干扰控制分开处理时。本文提出了一种抗干扰集成规划与控制（DR-IPC），该方法将轻量级路径引导与非线性模型预测控制（NMPC）相结合，直接生成角速度和推力。一个互联的扩展卡尔曼滤波器和非线性干扰观测器共同提供过滤后的状态估计和重构的干扰，用于NMPC预测。所得到的公式统一了非线性四旋翼动力学、执行器约束、局部运动生成和惩罚安全飞行通道残差，而无需单独的轨迹优化阶段。Gazebo和MARSIM模拟，以及室内和室外实验验证了DR-IPC在风、悬挂负载、狭窄通道、球体冲击和动态障碍物的反应避让下的有效性。在有干扰的多目标导航中，DR-IPC将完成任务的数量从1/10提高到9/10，并将实验中的高度均方根误差（RMSE）从0.34降低到0.01米。完整系统以100 Hz的频率在机载操作。补充视频可在项目页面https://drpp316.github.io/DR-IPC-Page/上查看，源代码将会发布。
cs.RO / 51 / 2610.03537
Autonomous Robotic Navigation for Endovascular Brain-Computer Interface Access
用于血管内脑-计算机接口访问的自主机器人导航
Abstract
Endovascular brain-computer interfaces (BCIs) avoid craniotomy but require precise device delivery through anatomically variable cerebral veins. This work presents the first demonstration of in vitro autonomous robotic navigation for endovascular BCI access in the cerebral venous system. Soft Actor-Critic controllers were trained in silico for two sequential tasks spanning the right internal jugular vein to the superior sagittal sinus, using geometric augmentation of one training anatomy. Navigation was evaluated in a training anatomy and an anatomically unseen hold-out model over 250 in silico episodes and five fluoroscopy-guided in vitro robotic runs per task-anatomy condition, comprising 1,000 simulated episodes and 20 physical runs overall. Task recurrent predictors were also evaluated for online identification of impending navigation failure. In silico success rates for Tasks A and B were 85.6% and 98.4% in the training anatomy and 42.0% and 91.6% in the hold-out anatomy, respectively. Fourteen of 20 physical runs were successful (70% overall), including 80% success for Task B in the hold-out phantom. In silico the predictors detected 99.3-100.0% of failures with false-alarm rates of 0.8-6.7%. During in vitro evaluation, predicted risk increased before failed episodes, but elevated probabilities during some successful runs showed reduced calibration after transfer. These results demonstrate the feasibility of autonomous cerebral venous access and show how online failure prediction could support human oversight, while also identifying anatomical generalization and sim-to-real calibration as priorities before preclinical translation.
Chinese Translation
血管内脑-计算机接口（BCIs）避免了开颅手术，但需要通过解剖结构变化的脑静脉进行精确的设备输送。本研究首次展示了在体外进行自主机器人导航以实现脑静脉系统的血管内BCI访问。使用几何增强的训练解剖结构，在计算机模拟中训练了软演员-评论者（Soft Actor-Critic）控制器，以完成从右侧内颈静脉到上矢状窦的两个连续任务。导航在训练解剖结构和一个未见过的保留模型中进行了评估，涵盖了250个计算机模拟的实验和每个任务-解剖条件下五次荧光透视引导的体外机器人运行，总计1,000个模拟实验和20次实际运行。还评估了任务重复预测器，以在线识别即将发生的导航失败。在训练解剖结构中，任务A和任务B的计算机模拟成功率分别为85.6%和98.4%，而在保留解剖结构中则为42.0%和91.6%。20次实际运行中有14次成功（总体成功率为70%），其中在保留模型中任务B的成功率为80%。在计算机模拟中，预测器检测到99.3-100.0%的失败，误报率为0.8-6.7%。在体外评估期间，预测的风险在失败实验之前增加，但在某些成功运行期间的高概率显示出转移后的校准降低。这些结果证明了自主脑静脉访问的可行性，并展示了在线失败预测如何支持人类监督，同时还确定了解剖学泛化和模拟到现实校准在临床前转化之前的重要性。
cs.RO / 52 / 2610.03538
RATE: Risk-Aware Tactile Encoding for Contact-rich Robotic Manipulation
RATE：面向接触丰富的机器人操作的风险感知触觉编码
Abstract
Tactile sensing is particularly valuable for contact-rich robotic manipulation. Recent work has made substantial progress in tactile representation learning for robotic manipulation. However, similar tactile observations can arise from interaction conditions with very different task-risk implications, such as sensor noise, task-necessary variations, and emerging undesirable contact. Without context-grounded risk information, these cases can be ambiguous to downstream policies, leading to unnecessary corrections to benign variations or delayed responses to genuinely risky contact. To address this limitation, we propose Risk-Aware Tactile Encoding (RATE), which learns tactile representations that encode task-conditioned interaction risk. Specifically, history-conditioned prediction captures interaction context, while alert supervision associates this context with task-conditioned risk. The learned risk-aware representation complements conventional tactile features through a lightweight residual adapter. Experiments in both simulation and the real world demonstrate substantial improvements in task success, with controlled ablations confirming the complementary benefits of alert-guided learning and predictive temporal modeling.
Chinese Translation
触觉感知在接触丰富的机器人操作中尤为重要。近期的研究在机器人操作的触觉表示学习方面取得了显著进展。然而，相似的触觉观察可能源于具有非常不同任务风险含义的交互条件，例如传感器噪声、任务所需的变化以及意外的接触。缺乏基于上下文的风险信息，这些情况可能对下游策略造成歧义，导致对无害变化的不必要修正或对真正风险接触的反应延迟。为了解决这一局限性，我们提出了风险感知触觉编码（RATE），该方法学习编码任务条件交互风险的触觉表示。具体而言，基于历史的预测捕捉交互上下文，而警报监督将该上下文与任务条件风险关联起来。所学习的风险感知表示通过轻量级残差适配器补充了传统的触觉特征。在模拟和现实世界中的实验表明，任务成功率显著提高，控制性消融实验确认了警报引导学习和预测时间建模的互补效益。
cs.RO / 53 / 2610.03563
A Unified Framework for Empowerment and Predictive Control
统一的赋能与预测控制框架
Abstract
Sampling-based model predictive control (MPC) is a powerful approach to trajectory optimization, but its performance depends on an informative cost function that often requires substantial domain knowledge to design. An alternative is to derive objectives directly from the system dynamics. Empowerment, defined as the channel capacity between an agent's inputs and its future state, provides such a goal-agnostic objective and has been shown to produce useful behaviors across a range of domains. However, existing empowerment-based controllers have two key limitations: empowerment is estimated using a virtual probing policy distinct from the executed control policy, obscuring its connection to the resulting behavior; and its computation typically requires second-order derivatives of the dynamics, limiting compatibility with standard MPC methods. We formulate empowerment using a single policy for both probing and acting, directly linking the empowerment objective to the executed behavior. The resulting objective requires only first-order derivatives and can be optimized with standard MPC solvers, either alone or in combination with a task-specific cost. We evaluate the proposed approach on classical control tasks and show that combining empowerment with task costs matches or exceeds the success rate achieved by either objective alone. Our formulation provides a practical bridge between standard MPC and intrinsically motivated control.
Chinese Translation
基于采样的模型预测控制（MPC）是一种强大的轨迹优化方法，但其性能依赖于信息丰富的成本函数，而设计这样的成本函数通常需要大量的领域知识。另一种选择是直接从系统动力学中推导目标。赋能（Empowerment），定义为代理输入与其未来状态之间的信道容量，提供了一种与目标无关的客观标准，并已被证明在多个领域中产生有用的行为。然而，现有的基于赋能的控制器存在两个主要限制：赋能是通过与执行控制策略不同的虚拟探测策略来估计的，这模糊了其与结果行为之间的联系；而且其计算通常需要动力学的二阶导数，这限制了与标准MPC方法的兼容性。我们使用单一策略同时进行探测和执行来表述赋能，直接将赋能目标与执行的行为联系起来。所得到的目标仅需一阶导数，并且可以与标准MPC求解器一起优化，无论是单独使用还是与特定任务成本结合使用。我们在经典控制任务上评估了所提出的方法，并显示将赋能与任务成本结合使用的成功率与单独使用任一目标的成功率相匹配或超过。我们的表述为标准MPC与内在驱动控制之间提供了一个实用的桥梁。
cs.RO / 54 / 2610.03587
AVL-JEPA: Preventing Causal Dynamics Information Collapse In Joint Embedding Predictive Architecture World Models
AVL-JEPA：防止联合嵌入预测架构世界模型中的因果动态信息崩溃
Abstract
Joint embedding predictive architectures (JEPAs) predict future latent representations without reconstructing observations, enabling world models to focus on high-level semantic dynamics. However, a JEPA can preserve high dimensional visual information while discarding information about the physical consequences of actions. We call this failure mode causal dynamics information collapse and propose action-grounded vision-invariance latent (AVL) to prevent this collapse. We first use the executed action as an auxiliary dynamics anchor that encourages the model to preserve dynamics information, and then use a vision-invariance pathway which aligns perturbed and clean latent predictions without discarding dynamics information, forcing the model to fully understand and utilize causal dynamics information. We validate AVL on four robotic control tasks (TwoRoom, PushT, OGBench Cube, and Reacher), showing that it substantially improves success rates under visual perturbations while preserving clean-environment performance. We further evaluate physical consequence alignment, clean-noisy dynamics consistency, and the causal effect of targeted transition subspace erasure. Collectively, these results indicate that dynamic information causally relevant to planning is preserved from collapse under AVL.
Chinese Translation
联合嵌入预测架构（JEPA）在不重建观测的情况下预测未来的潜在表征，使得世界模型能够专注于高层语义动态。然而，JEPA可能会保留高维视觉信息，同时丢弃关于动作物理后果的信息。我们将这种失败模式称为因果动态信息崩溃，并提出了基于动作的视觉不变潜在（AVL）方法以防止这种崩溃。我们首先使用执行的动作作为辅助动态锚点，鼓励模型保留动态信息，然后使用视觉不变路径对齐扰动和干净的潜在预测，而不丢弃动态信息，迫使模型充分理解和利用因果动态信息。我们在四个机器人控制任务（TwoRoom、PushT、OGBench Cube 和 Reacher）上验证了AVL，结果表明在视觉扰动下成功率显著提高，同时保持了干净环境下的性能。我们进一步评估了物理后果对齐、干净-噪声动态一致性以及针对过渡子空间擦除的因果效应。综合来看，这些结果表明，与规划因果相关的动态信息在AVL下得以保留，避免了崩溃。
cs.RO / 55 / 2610.03607
World Action Learning via Interaction-Centric Spectral Latent Guidance
通过以交互为中心的谱潜在引导实现世界行动学习
Abstract
Learning general-purpose robot policies requires large-scale real-world interaction data, yet collecting robot demonstrations remains expensive and difficult to scale. Egocentric videos offer abundant human interaction experience with task-relevant semantics for robotic manipulation, but direct transfer is challenging for two reasons: latent actions inferred from frame reconstruction can be dominated by nuisance variation such as ego-camera motion, and human and robot behaviors often exhibit different temporal dynamics. We propose WING (World Action Learning via INteraction-Centric Spectral Latent Guidance), a framework for transferring interaction knowledge from egocentric videos to robot policies. WING first separates observer-induced motion from hand-object interaction and distills the interaction-centric component into latent actions. It then exploits the observation that cross-embodiment task semantics are concentrated in slowly varying temporal structures, identifying shared low-frequency components between egocentric latent actions and robot behaviors in the spectral domain and using them to guide action generation. WING achieves average success rates of 99.20% on LIBERO, 93.80% on RoboTwin 2.0, and 57.7% on RoboCasa-GR1, and also performs strongly across four real-world manipulation tasks under diverse generalization settings. These results show that interaction-centric spectral guidance provides an effective and scalable way to transfer physical interaction knowledge from human egocentric video to robot control. Project page: https://mikuz12.github.io/wing/
Chinese Translation
学习通用机器人策略需要大规模的真实世界交互数据，但收集机器人演示仍然昂贵且难以扩展。自我中心视频提供了丰富的人类交互经验，具有与任务相关的语义，适用于机器人操作，但直接转移面临两个挑战：从帧重建推断出的潜在动作可能受到如自我摄像机运动等干扰变化的主导影响，并且人类和机器人行为通常表现出不同的时间动态。我们提出了WING（通过以交互为中心的谱潜在引导实现世界行动学习），这是一个将自我中心视频中的交互知识转移到机器人策略的框架。WING首先将观察者引起的运动与手-物体交互分离，并将以交互为中心的成分提炼为潜在动作。然后，它利用观察到的跨体现任务语义集中在缓慢变化的时间结构中，识别自我中心潜在动作与机器人行为在谱域中的共享低频成分，并利用这些成分指导动作生成。WING在LIBERO上实现了99.20%的平均成功率，在RoboTwin 2.0上为93.80%，在RoboCasa-GR1上为57.7%，并且在四个真实世界操作任务中在多样化的泛化设置下表现出色。这些结果表明，以交互为中心的谱引导提供了一种有效且可扩展的方式，将人类自我中心视频中的物理交互知识转移到机器人控制中。项目页面：https://mikuz12.github.io/wing/
cs.RO / 56 / 2610.03615
Bridging Frontier Reasoning and Robot Execution: From Autonomous Demonstration Generation to Dense Language Supervision
桥接前沿推理与机器人执行：从自主演示生成到密集语言监督
Abstract
Recent advances in frontier models enable robot manipulation from only a few demonstrations, but high inference latency limits their use for real-time robot control. To bridge this gap, we study two complementary approaches that connect frontier reasoning with low-latency local execution. First, we use a frontier model to autonomously generate demonstrations that supplement human demonstrations for training a fast local policy. We augment its in-context examples with corrective demonstration segments that show how to recover from physical errors, improving generation reliability. Generation time and cost decrease as successful examples accumulate in context, suggesting a path toward more efficient data collection. At deployment, a harness combines frontier-generated instructions with a fast local policy, enabling efficient execution while preserving the frontier model's ability to guide and correct actions. With low-latency execution delegated to the local policy, the bottleneck shifts to its capacity to reliably follow the frontier model's diverse instructions. Our second bridge introduces dense language supervision across three nested granularities---primitive, atomic, and composite---with multi-aspect descriptions at each level. Across long-horizon tasks in RoboCasa 365 and BEHAVIOR-1K, where instructions change as execution progresses, the combined supervision achieves the highest performance under both oracle and frontier-model instructors, demonstrating more reliable instruction following through the policy's language interface. Finally, we evaluate both bridges together on a crossword task that combines semantic planning and manipulation within a fixed time budget. These results support autonomous demonstration generation and dense language supervision as complementary components for connecting frontier reasoning to low-latency local execution.
Chinese Translation
最近，前沿模型的进展使得机器人能够仅通过少量演示进行操作，但高推理延迟限制了其在实时机器人控制中的应用。为了解决这一问题，我们研究了两种互补的方法，将前沿推理与低延迟的本地执行连接起来。首先，我们使用前沿模型自主生成演示，以补充人类演示，从而训练快速的本地策略。我们通过纠正演示片段增强其上下文示例，展示如何从物理错误中恢复，提高生成的可靠性。随着成功示例在上下文中的积累，生成时间和成本减少，暗示了一条更高效的数据收集路径。在部署时，一个装置将前沿生成的指令与快速本地策略结合，能够高效执行，同时保留前沿模型指导和纠正动作的能力。由于低延迟执行被委托给本地策略，瓶颈转移到其可靠遵循前沿模型多样化指令的能力。我们的第二个桥梁引入了跨越三种嵌套粒度的密集语言监督——原始、原子和复合——在每个层次上都有多方面的描述。在RoboCasa 365和BEHAVIOR-1K的长期任务中，指令随着执行进展而变化，结合的监督在oracle和前沿模型指导下均实现了最高性能，展示了通过策略的语言接口更可靠的指令遵循。最后，我们在一个结合语义规划和操作的填字游戏任务中共同评估了这两个桥梁，该任务在固定时间预算内进行。这些结果支持自主演示生成和密集语言监督作为连接前沿推理与低延迟本地执行的互补组件。
cs.RO / 57 / 2610.03616
LLA-MPC on Embedded Hardware: Rapid Adaptive Control with Thousands of Parallel Models
嵌入式硬件上的LLA-MPC：千个并行模型的快速自适应控制
Abstract
We present a generalized implementation of Look-Back and Look-Ahead Adaptive Model Predictive Control (LLA-MPC), a learning-free framework for real-time, rapid adaptive system identification and control. The original formulation was demonstrated only in simulation, for autonomous racing with a fixed model structure. Our implementation has modular dynamics and integrators, and we validate it on the F1TENTH platform under constrained computation and noisy state estimation. The system identifies tire parameters online by evaluating thousands of candidate models in real time on an embedded computer. Experiments with low-friction tires across changing surfaces show that LLA-MPC completes high-speed tracking tasks where a fixed nominal model fails. Code, videos, and our related work are available at: https://lla-control.github.io.
Chinese Translation
我们提出了一种广义的回顾与展望自适应模型预测控制（LLA-MPC）实现，这是一个无学习框架，用于实时、快速的自适应系统识别与控制。原始的公式仅在仿真中展示，适用于具有固定模型结构的自主赛车。我们的实现具有模块化的动力学和积分器，并在受限计算和噪声状态估计的条件下在F1TENTH平台上进行了验证。该系统通过实时评估数千个候选模型在线识别轮胎参数，运行于嵌入式计算机上。针对不同表面上的低摩擦轮胎进行的实验表明，LLA-MPC能够完成固定名义模型失败的高速跟踪任务。代码、视频及相关工作可在以下链接获取：https://lla-control.github.io。
cs.RO / 58 / 2610.03622
CORNAV: Construction-Aware Reasoning for Robot Navigation on Active Worksites
CORNAV：面向主动工地的建设意识机器人导航推理
Abstract
The construction industry faces persistent labor shortages, low productivity that costs the global economy over $1.6 trillion annually, and one of the highest injury rates among major industries. These factors motivate the use of autonomous robots to improve efficiency and worker safety. Existing language-grounded navigation systems, however, rely on semantic scene understanding alone and lack access to construction-specific context such as architectural plans, evolving work schedules, and safety constraints. As a result, they localize permanent building features unreliably and cannot safely navigate active jobsites. We present CORNAV, a blueprint-grounded, schedule-aware navigation framework that operates from 2D CAD drawings and project schedules without requiring a Building Information Model. CORNAV aligns architectural blueprints against hierarchical open-vocabulary 3D scene graphs to ground object queries, converts project schedules into time-varying navigation constraints, and validates requests through an LLM-based safety module that escalates hazardous zones before planning. An A* planner then enforces mandatory exclusion zones while preferentially avoiding higher-risk areas. Across an indoor office and a real construction site, blueprint grounding raises task success from 13.0% to 72.2% over semantic retrieval alone, schedule awareness eliminates all hard-zone violations, and the safety module correctly rejects hazardous requests arising from mislabeled project schedules.
Chinese Translation
建筑行业面临持续的劳动力短缺、低生产力（每年给全球经济造成超过1.6万亿美元的损失）以及主要行业中最高的伤害率。这些因素促使使用自主机器人来提高效率和工人安全。然而，现有的基于语言的导航系统仅依赖于语义场景理解，缺乏对建筑特定上下文的访问，例如建筑蓝图、不断变化的工作计划和安全约束。因此，它们在定位永久建筑特征时不可靠，无法安全地在活跃的工地上导航。我们提出了CORNAV，一个基于蓝图的、考虑时间表的导航框架，能够从2D CAD图纸和项目时间表中操作，而无需建筑信息模型（Building Information Model）。CORNAV将建筑蓝图与分层开放词汇的3D场景图对齐，以支持对象查询，将项目时间表转换为时间变化的导航约束，并通过基于LLM的安全模块验证请求，该模块在规划之前提升危险区域。然后，A*规划器强制执行强制排除区，同时优先避免高风险区域。在一个室内办公室和一个真实的建筑工地上，蓝图基础的导航成功率从仅依靠语义检索的13.0%提高到72.2%，时间表意识消除了所有硬区违规，而安全模块正确拒绝了因项目时间表标记错误而产生的危险请求。
cs.RO / 59 / 2610.03710
EyeRobot 2.0: Active Gaze for Precise Manipulation without Wrist Cameras
EyeRobot 2.0：无需腕部摄像头的精确操作主动注视
Abstract
Inspired by human vision, we introduce a framework using active gaze to enable fine-grained bimanual manipulation with only a single stereo camera. EyeRobot 2.0 physically attends to a 3D fixation point in the scene by swiveling two eye viewpoints to center their gaze on it. The resulting images are processed foveally by allocating more visual tokens to the image centers, focusing computation on task-relevant features. Such Active Visual Fixation (AVF) requires carefully coordinated gaze during task execution, which we accomplish hierarchically by first training a low-level gaze servoing policy conditioned on a goal object, then training a target selector which emits fixation goals based on task progress. Both modules are trained with RL on real-world data: the first is trained with a dense geometric reward and the second co-trains with the BC gripper policy which allows it to discover fixation sequences that can resemble a human's fixation sequence while performing the task. EyeRobot 2.0 further takes advantage of fixation by canonicalizing gripper information into a fixation-relative SE(3) frame, which compacts the size of the action distribution to learn. We collect teleoperation data for 7 real-world and 6 simulated tasks, and conduct over 1000 physical and 1800 simulated robot trials comparing EyeRobot 2.0 against passive stereo and ego + wrist camera policies trained on the same data. Removing wrist cameras is costly for standard policies: with only passive stereo, real-world success drops from 52% to 27%. EyeRobot 2.0 closes this gap with only stereo, outperforming passive stereo by 40% in real and 20% in sim. It matches ego + wrist policies when their wrist views are clear (69% vs. 64%), and more than doubles their success when grasped objects occlude the wrist cameras (48% vs. 22%)
Chinese Translation
受到人类视觉的启发，我们提出了一种框架，利用主动注视来实现仅使用单个立体摄像头的精细双手操作。EyeRobot 2.0 通过旋转两个眼睛视点，物理上关注场景中的 3D 固定点，使其注视集中于该点。生成的图像通过将更多视觉标记分配给图像中心进行中心化处理，专注于与任务相关的特征。这种主动视觉固定 (Active Visual Fixation, AVF) 在任务执行过程中需要精心协调的注视，我们通过分层方式实现：首先训练一个基于目标物体的低级注视伺服策略，然后训练一个目标选择器，根据任务进展发出固定目标。两个模块均在真实世界数据上通过强化学习 (Reinforcement Learning, RL) 进行训练：第一个模块通过密集几何奖励进行训练，第二个模块与基于行为克隆 (Behavior Cloning, BC) 的抓取策略共同训练，使其能够发现与人类在执行任务时的注视序列相似的固定序列。EyeRobot 2.0 进一步通过将抓取器信息规范化为相对于固定点的 SE(3) 坐标系，利用固定点的优势，从而压缩学习的动作分布大小。我们收集了 7 个真实世界和 6 个模拟任务的遥操作数据，并进行了超过 1000 次物理和 1800 次模拟机器人试验，将 EyeRobot 2.0 与基于相同数据训练的被动立体和自我 + 腕部摄像头策略进行比较。去除腕部摄像头对标准策略的影响很大：仅使用被动立体时，真实世界的成功率从 52% 降低到 27%。EyeRobot 2.0 通过仅使用立体摄像头缩小了这一差距，在真实环境中比被动立体提高了 40%，在模拟环境中提高了 20%。当腕部视图清晰时，其成功率与自我 + 腕部策略相当 (69% 对比 64%)，而在抓取物体遮挡腕部摄像头时，其成功率更是超过了后者的两倍 (48% 对比 22%)。
cs.CV / 1 / 2610.02298
EditHero: A Benchmark for Long-Horizon Part-Level 3D Editing and Vibe Modeling
EditHero：一个用于长时间跨度部件级3D编辑和氛围建模的基准
Abstract
3D editing methods are usually tested on a single edit, yet an asset is built through a long sequence of revisions, each of which must implement the requested change while leaving everything else unchanged. We introduce EditHero, to our knowledge the first benchmark for long-horizon, part-level 3D editing, with natural-language instructions and target images for both geometry and texture. A deterministic assembly engine produces the exact target after every edit, and every sequence is reviewed by hand. We use EditHero to compare 2 opposite approaches to 3D editing. Non-agentic methods operate top down, regenerating the object from a learned 3D representation and inferring what to keep. In contrast, LLM/VLM agents operate bottom up, editing through code that inspects the mesh and rewrites only the parts required by instructions. The non-agentic methods often miss the requested change and disturb regions that should stay fixed. Most LLMs follow instructions more closely, and all of them preserve the unedited parts better, but each of their edits takes minutes. We will release the engine and the edit sequences to support research on reliable iterative 3D editing.
Chinese Translation
3D编辑方法通常在单次编辑上进行测试，而资产是通过一系列长时间的修订构建的，每次修订都必须实现所请求的更改，同时保持其他部分不变。我们介绍了EditHero，作为我们所知的第一个长时间跨度、部件级3D编辑的基准，提供自然语言指令以及几何和纹理的目标图像。一个确定性的装配引擎在每次编辑后生成准确的目标，且每个序列都经过人工审核。我们使用EditHero比较两种相对立的3D编辑方法。非自主方法自上而下地操作，从学习到的3D表示中重新生成对象，并推断需要保留的部分。相反，LLM/VLM代理自下而上地操作，通过检查网格的代码进行编辑，仅重写指令所需的部分。非自主方法通常会遗漏请求的更改，并干扰应保持固定的区域。大多数LLM更紧密地遵循指令，并且它们在保持未编辑部分方面表现更好，但它们的每次编辑都需要几分钟。我们将发布该引擎和编辑序列，以支持可靠的迭代3D编辑研究。
cs.CV / 2 / 2610.02320
DeskForge: Dense Supervision from Desktop Environments for Computer-Use Agents
DeskForge：来自桌面环境的密集监督用于计算机使用代理
Abstract
Computer-use agents need to reliably ground action targets in complex desktop scenes, where multiple applications, overlapping windows, and visually similar controls compete for attention. Existing training data rarely pair such scenes with dense annotations or vary them in a controlled way. We introduce DeskForge, a controllable desktop environment that composes and explores real applications to generate large-scale supervision for computer-use agents. It varies application states, content, window layout, appearance, and resolution, and fuses screenshots, accessibility trees, and window geometry into dense element annotations while recording the outcome of each executed action. Using this environment, we construct DeskForge-1M, a corpus of 1.2M annotated desktop observations containing 159.7M element instances. We fine-tune four vision-language models on 200K grounding examples drawn from DeskForge-1M. All four improve across held-out desktop conditions and on all five external GUI grounding benchmarks; for Qwen3.5-4B, accuracy increases by 11.51 percentage points on ScreenSpot-Pro and 10.11 points on OSWorld-G. The gains also translate to long-horizon task completion: under a fixed planner, the fine-tuned action models solve more WebArena-Infinity and OpenApps tasks, with Qwen3.5-4B increasing from 31 to 50 of 119 tasks and from 3 to 15 of 100 tasks, respectively. These results show that controllable composition of real desktop environments provides a scalable source of supervision for improving both GUI grounding and long-horizon computer use. The framework code, the dataset, and the fine-tuned model are available from the project page: https://saidgurbuz.github.io/deskforge/
Chinese Translation
计算机使用代理需要在复杂的桌面场景中可靠地确定动作目标，这些场景中多个应用程序、重叠窗口和视觉上相似的控件相互竞争以吸引注意力。现有的训练数据很少将此类场景与密集注释配对，或以受控方式进行变化。我们引入了DeskForge，一个可控的桌面环境，能够组合和探索真实应用程序，以生成大规模的计算机使用代理监督。它变化应用状态、内容、窗口布局、外观和分辨率，并将屏幕截图、可访问性树和窗口几何信息融合为密集元素注释，同时记录每个执行动作的结果。利用该环境，我们构建了DeskForge-1M，一个包含120万条注释桌面观察的数据集，包含1.597亿个元素实例。我们在从DeskForge-1M中提取的20万个基础示例上微调了四个视觉-语言模型。所有四个模型在保留的桌面条件下和五个外部GUI基础基准上均有所提升；对于Qwen3.5-4B，其在ScreenSpot-Pro上的准确率提高了11.51个百分点，在OSWorld-G上提高了10.11个百分点。这些提升也转化为长时间任务的完成：在固定规划器下，微调后的动作模型解决了更多WebArena-Infinity和OpenApps任务，其中Qwen3.5-4B从119个任务中从31个增加到50个，从100个任务中从3个增加到15个。这些结果表明，真实桌面环境的可控组合为改善GUI基础和长时间计算机使用提供了可扩展的监督来源。框架代码、数据集和微调模型可从项目页面获取：https://saidgurbuz.github.io/deskforge/
cs.CV / 3 / 2610.02322
SCION: Scene Composition with Instanced Neural Primitives
SCION：基于实例化神经原语的场景组合
Abstract
Real-world scenes are compositional: bricks, blades of grass, pebbles, and tree leaves recur across human-built and natural environments. Existing neural scene representations model these elements independently. Most 3D Gaussian Splatting and follow-up abstraction and compression methods treat each element as unique, fitting millions of independent Gaussians per scene. Prior methods like Splat and Replace fit template objects, but they require mostly manual selection of repeated elements. As a result, these representations store redundant parameters and provide weak manipulation handles for downstream tasks. We introduce SCION, a hier- archical compositional scene representation that replaces independent Gaussians with a compact vocabulary of reusable primitives and lightweight world-space instances that place transformed copies throughout the scene. We fit this represen- tation to multi-view captures via a joint optimization over discrete and continuous scene parameters, combining two-level densification over splats and instances with an adversarial loss that preserves detail across shared primitives. The recovered structure yields a compact, controllable representation while maintaining high quality even at 1.2 MB. SCION achieves rate-distortion favorable to existing Gaussian compression methods, and it enables instance-level scene editing and animation without retraining. Our results show that neural scene representations need not memorize scenes as independent primitives; they can discover reusable parts. Project webpage: https://light.princeton.edu/SCION
Chinese Translation
现实世界的场景是组合性的：砖块、草叶、卵石和树叶在人工建造和自然环境中反复出现。现有的神经场景表示独立地建模这些元素。大多数3D高斯点云（3D Gaussian Splatting）及其后续的抽象和压缩方法将每个元素视为独特的，为每个场景拟合数百万个独立的高斯分布。之前的方法如Splat和Replace拟合模板对象，但它们主要依赖手动选择重复元素。因此，这些表示存储了冗余参数，并为下游任务提供了较弱的操作手段。我们提出了SCION，一种分层组合场景表示，它用可重用原语的紧凑词汇和轻量级的世界空间实例替代独立的高斯分布，这些实例在整个场景中放置变换后的副本。我们通过对离散和连续场景参数的联合优化，将这种表示拟合到多视角捕捉数据上，结合了对点云和实例的两级密集化以及保持共享原语细节的对抗损失。恢复的结构在保持高质量的同时，提供了紧凑且可控的表示，文件大小甚至可以达到1.2 MB。SCION在压缩率-失真方面优于现有的高斯压缩方法，并且它使得实例级场景编辑和动画成为可能，而无需重新训练。我们的结果表明，神经场景表示不必将场景记忆为独立的原语；它们可以发现可重用的部分。项目网页：https://light.princeton.edu/SCION
cs.CV / 4 / 2610.02343
SCOPE-4D: Endoscopic 4D Geometry Foundation Models
SCOPE-4D：内窥镜4D几何基础模型
Abstract
Geometric understanding supports endoscopic navigation and robotic assistance, but learning reliable endoscopic geometry faces two challenges: scarce geometric annotations and ambiguity between camera motion and tissue deformation. We present SCOPE-4D, an endoscopic 4D geometry foundation model that jointly predicts camera parameters, dense geometry, and 3D tissue trajectories from monocular RGB video in a single forward pass. Our curation and annotation pipeline constructs SCOPE-5K, a collection of approximately 5,000 clips spanning real and synthetic gastrointestinal endoscopy and laparoscopy. The collection provides rich geometric supervision and includes newly collected phantom and real-colonoscopy evaluation sets. Geometric supervised fine-tuning on SCOPE-5K learns endoscopic priors that improve camera and depth estimation. Common--Residual Motion (CRM) further constrains local deformation relative to common tissue movement. Together with geometric supervision, CRM and trajectory supervision further improve camera and depth estimation over geometric fine-tuning alone while enabling dense 3D tissue tracking. Evaluations on public and newly collected benchmarks demonstrate strong in-domain and out-of-domain geometry, superior 3D tracking, and more stable long-sequence colon reconstruction. A blinded user study further supports the perceived reconstruction quality on real clinical video. Together, these results demonstrate the value of large-scale endoscopic supervision and motion constraints for joint geometry estimation and tissue tracking.
Chinese Translation
几何理解支持内窥镜导航和机器人辅助，但学习可靠的内窥镜几何面临两个挑战：几何标注稀缺和相机运动与组织变形之间的模糊性。我们提出了SCOPE-4D，这是一种内窥镜4D几何基础模型，可以在单次前向传播中共同预测相机参数、稠密几何和3D组织轨迹，基于单目RGB视频。我们的策划和标注流程构建了SCOPE-5K，这是一个包含约5000个片段的集合，涵盖真实和合成的胃肠内窥镜和腹腔镜。该集合提供了丰富的几何监督，并包括新收集的幻影和真实结肠镜评估集。对SCOPE-5K进行几何监督的微调学习了内窥镜先验，从而改善了相机和深度估计。共同残差运动（Common--Residual Motion, CRM）进一步约束了相对于共同组织运动的局部变形。结合几何监督，CRM和轨迹监督进一步改善了相机和深度估计，相较于仅进行几何微调，同时实现了稠密的3D组织跟踪。在公共和新收集的基准测试上的评估显示了强大的领域内和领域外几何表现，优越的3D跟踪能力，以及更稳定的长序列结肠重建。盲测用户研究进一步支持了在真实临床视频上感知重建质量的提升。综合来看，这些结果展示了大规模内窥镜监督和运动约束在联合几何估计和组织跟踪中的价值。
cs.CV / 5 / 2610.02364
Confidence-Controlled XAI Auditing for Pedestrian Detection under Domain Shift
域转移下行人检测的信心控制可解释性审计
Abstract
Explainability is increasingly required for perception models in intelligent vehicles, yet whether explanations remain faithful under driving domain shift is still poorly understood. This work audits post-hoc explanations of a fixed YOLOv8s pedestrian detector across PIE and JAAD using ROI-based D-Deletion, frozen confidence terciles, rank-based tests, bootstrap intervals, and Holm correction. The audit shows that deletion-based faithfulness is strongly coupled to detection strength at explanation time, with Spearman correlations between 0.70 and 0.82 for D-RISE, making naive confidence-stratified comparisons unreliable. After controlling for detection strength within fixed f0 bins, D-RISE faithfulness remains domain-dependent in the central f0 range, with PIE showing higher D-Deletion than JAAD and Holm-adjusted significance. A non-perturbative EigenCAM baseline is less faithful than D-RISE but also exhibits score coupling, suggesting that the effect is not specific to D-RISE and is related to the deletion-based evaluation setup. These results motivate confidence-controlled XAI audits for safety-critical perception under domain shift.
Chinese Translation
在智能车辆的感知模型中，解释性越来越受到重视，但在驾驶领域转移下，解释是否仍然保持真实的情况仍然不甚了解。本研究对固定的YOLOv8s行人检测器在PIE和JAAD数据集上的事后解释进行了审计，采用了基于ROI的D-Deletion、冻结的置信度三分位数、基于排名的测试、引导区间和Holm校正。审计结果表明，基于删除的真实度与解释时的检测强度密切相关，D-RISE的斯皮尔曼相关系数在0.70到0.82之间，使得简单的置信度分层比较不可靠。在固定的f0区间内控制检测强度后，D-RISE的真实度在中央f0范围内仍然依赖于领域，PIE的D-Deletion高于JAAD，并且经过Holm调整的显著性更强。非扰动的EigenCAM基线的真实度低于D-RISE，但也表现出得分耦合，表明这一效应并非特定于D-RISE，而与基于删除的评估设置有关。这些结果促使我们对在领域转移下安全关键感知进行信心控制的可解释性审计。
cs.CV / 6 / 2610.02375
EviDent-CBCT: Evidence-Bottlenecked Report Generation from Dental CBCT under Non-Exhaustive Report Supervision
EviDent-CBCT：基于证据瓶颈的牙科CBCT报告生成方法在非详尽报告监督下的应用
Abstract
Dento-maxillofacial cone-beam CT (CBCT) reports may contain dozens of tooth-specific, anatomical, and spatial findings from a single 3D scan. Learning to generate such reports from limited clinical data is challenging because routine reports may not exhaustively document image findings, and a non-mention may reflect either absence or non-reporting. We present EviDent-CBCT, an evidence-bottlenecked framework designed for this incomplete supervision. An anatomy-aware network maps each CBCT scan to a discrete record of tooth-level, global, and tooth-IAC evidence. A dental-logic consistency projection reconciles incompatible evidence before a deterministic renderer and an image-blind local language model generate the report using only this record. For tooth-level evidence, reliability-aware training uses eligible non-mentions as reduced-weight negatives, while unreported global and tooth-IAC labels remain unknown. A metal-sensitive input channel preserves intensity cues from dental materials. Across three validation runs, EviDent-CBCT achieves $0.666\pm0.006$ merged evidence set-F1 and $0.402\pm0.003$ RadFact-Lite-Dental logical-F1, versus $0.371\pm0.018$ for the strongest controlled direct baseline. In the ODIN 2026 challenge, it ranked second in automated evaluation and third in blinded clinical Arena comparison on the hidden test set. These results support the discrete evidence record as an effective and auditable interface for CBCT report generation.
Chinese Translation
牙颌面锥形束计算机断层扫描（CBCT）报告可能包含来自单个三维扫描的数十个特定牙齿的解剖和空间发现。从有限的临床数据中学习生成此类报告具有挑战性，因为常规报告可能并未详尽记录图像发现，而未提及可能反映出缺失或未报告。我们提出了EviDent-CBCT，这是一种针对这种不完整监督设计的证据瓶颈框架。一个关注解剖结构的网络将每个CBCT扫描映射到一个离散的记录，包括牙齿级别、全局和牙齿-IAC证据。在确定性渲染器之前，牙科逻辑一致性投影调和不兼容的证据，随后使用该记录生成报告的图像盲局部语言模型。对于牙齿级别证据，可靠性意识训练将合格的未提及作为减重负样本，而未报告的全局和牙齿-IAC标签仍然未知。金属敏感输入通道保留了来自牙科材料的强度线索。在三次验证运行中，EviDent-CBCT达到了$0.666 ext{±}0.006$的合并证据集-F1和$0.402 ext{±}0.003$的RadFact-Lite-Dental逻辑-F1，相较于最强的控制直接基线$0.371 ext{±}0.018$。在ODIN 2026挑战赛中，它在自动评估中排名第二，在盲测临床Arena比较中排名第三。这些结果支持离散证据记录作为CBCT报告生成的有效且可审计的接口。
cs.CV / 7 / 2610.02382
FactorSplat: Appearance-Controllable Gaussian Proxies for Medical Volume Rendering
FactorSplat：可控外观的高斯代理用于医学体积渲染
Abstract
Transfer functions (TFs) control color and visibility in medical volume rendering, but image-trained Gaussian proxies typically bake one transfer function into their appearance. We present FactorSplat, a per-scene N-dimensional Gaussian splatting (N-DGS) proxy that accepts region-specific intensity-to-RGBA curves at inference. A local lookup applies the authored color and opacity change, while a shared functional encoder and low-rank per-Gaussian factors learn the residual appearance response. Geometry and directional appearance remain shared across presets, with visibility control and TF-aware pruning preserving the ability to hide and reveal structures. On seven CT and MR scans, FactorSplat improves mean PSNR and changed-region error over region-aware VEG across validation, interpolation, unseen composition, and out-of-distribution (OOD) edits. Across these four splits, seven-scan mean PSNR gains over VEG range from 1.10 to 1.52 dB. One checkpoint per scan supports unseen edits without retraining. At $1600^2$, the cached fast renderer averages 524 FPS with 1.17 ms TF switches. Project page: https://gaozhongpai.github.io/FactorSplat/.
Chinese Translation
传输函数（TFs）在医学体积渲染中控制颜色和可见性，但图像训练的高斯代理通常将一个传输函数固定在其外观中。我们提出了FactorSplat，这是一种每场景N维高斯喷溅（N-DGS）代理，在推理时接受区域特定的强度到RGBA曲线。局部查找应用了作者定义的颜色和不透明度变化，而共享的功能编码器和低秩每个高斯因子学习残余外观响应。几何形状和方向外观在预设之间保持共享，同时可见性控制和TF感知修剪保留了隐藏和显示结构的能力。在七个CT和MR扫描中，FactorSplat在验证、插值、未见组合和分布外（OOD）编辑中，相较于区域感知的VEG提高了平均PSNR和变化区域误差。在这四个拆分中，七个扫描的平均PSNR相较于VEG的增益范围为1.10到1.52 dB。每个扫描的一个检查点支持未见编辑而无需重新训练。在$1600^2$的分辨率下，缓存的快速渲染器平均达到524 FPS，TF切换时间为1.17毫秒。项目页面：https://gaozhongpai.github.io/FactorSplat/.
cs.CV / 8 / 2610.02388
Octrees as an Explicit 3D Language
八叉树作为一种显式的三维语言
Abstract
Existing 3D large language models (LLMs) compromise on two fronts: they compress shapes into latent codebook indices or coordinate text, which removes spatial structure from what the model observes, and they acquire the 3D modality by fine-tuning the backbone, which overwrites its general language ability. We present OctLLM, which addresses both limitations. Geometry enters as an explicit 3D sequence of octree occupancy tokens. However, full octree sequences grow rapidly with depth; OctLLM therefore randomly empties penultimate-level nodes and omits descendants while preserving shape, yielding a shorter coordinate- and depth-anchored Sparse Octree (S-Octree) for position-aware mask-modeling generation and 3D understanding. On the other front, existing methods introduce a new modality with full fine-tuning or LoRA, but full fine-tuning is costly, LoRA limits 3D capacity, and both modify the language pathway. OctLLM instead adds 3D capacity in parameters separate from the pretrained ones: mesh tokens are routed through independent trainable branches in a subset of blocks while text and image tokens retain the frozen vision-language pathway, and the two streams interact through shared self-attention. It trains far fewer parameters than full fine-tuning, yet sets a new state of the art among unified multimodal LLMs, lowering image-to-3D FID by $17.4\%$ and raising render-grounded captioning by $28.7$ points over ShapeLLM-Omni, while matching the backbone on general language benchmarks.
Chinese Translation
现有的三维大型语言模型（LLMs）在两个方面存在妥协：它们将形状压缩为潜在编码索引或坐标文本，这消除了模型观察到的空间结构；它们通过微调主干网络来获取三维模态，这会覆盖其通用语言能力。我们提出了OctLLM，解决了这两个限制。几何信息以显式的三维八叉树占用标记序列的形式输入。然而，完整的八叉树序列随着深度的增加而迅速增长；因此，OctLLM随机清空倒数第二层节点并省略后代，同时保持形状，从而生成一个较短的坐标和深度锚定的稀疏八叉树（S-Octree），用于位置感知的掩码建模生成和三维理解。在另一个方面，现有方法通过完全微调或LoRA引入新的模态，但完全微调成本高，LoRA限制了三维能力，并且两者都修改了语言通路。OctLLM则在与预训练参数分开的参数中增加三维能力：网格标记通过一部分块中的独立可训练分支进行路由，而文本和图像标记保持冻结的视觉-语言通路，两条流通过共享自注意力进行交互。它训练的参数远少于完全微调，但在统一的多模态LLMs中设定了新的最先进水平，将图像到三维的FID降低了17.4%，并将基于渲染的描述提高了28.7个点，相较于ShapeLLM-Omni，同时在通用语言基准上与主干网络相匹配。
cs.CV / 9 / 2610.02421
An AI-Based Multi-Stage Approach for Androgenetic Alopecia Assessment from Low-Magnification Scalp Images
基于人工智能的多阶段雄性激素性脱发评估方法：来自低倍放大头皮图像的研究
Abstract
Androgenetic alopecia (AGA) is characterized by patterned follicular miniaturization, increased single-hair follicular units, and altered hair-shaft diameter. We present an automated quantitative scalp-analysis and clinical decision-support framework combining FU localization, ordinal visible-shaft counting, calibrated shaft-width estimation, regional aggregation, and an interpretable rule layer. The clinical cohort comprised 243 patients (127 AGA, 116 non-AGA), while the computer-vision experiments used 160 expert-annotated patients, 2,400 trichoscopic images, and approximately 158,000 FU annotations. Under patientdisjoint evaluation, YOLOv8m achieved test
[email protected]=0.920 and recall=0.860; EfficientNet-B5 with a support-map channel achieved 87.0% expert-box count accuracy (macro F1=0.85). A separate 500-image set was processed end-to-end with detector-generated boxes, yielding MAE of 6.56 for follicle detection and 16.59 for follicle classification relative to human-expert annotations. The system is intended to assist, rather than replace, dermatologist interpretation.
Chinese Translation
雄性激素性脱发（AGA）以有规律的毛囊微小化、增加的单发毛囊单位和改变的发干直径为特征。我们提出了一种自动化的定量头皮分析和临床决策支持框架，结合了毛囊单位（FU）定位、顺序可见发干计数、校准的发干宽度估计、区域聚合和可解释的规则层。临床队列包括243名患者（127名AGA患者，116名非AGA患者），而计算机视觉实验使用了160名专家标注的患者、2400张毛发显微镜图像和大约158,000个FU标注。在患者不重叠评估下，YOLOv8m在测试中的
[email protected]达到了0.920，召回率为0.860；使用支持图通道的EfficientNet-B5达到了87.0%的专家框计数准确率（宏F1=0.85）。一个独立的500张图像集通过检测器生成的框进行端到端处理，毛囊检测的平均绝对误差（MAE）为6.56，毛囊分类的MAE为16.59，相对于人类专家的标注。该系统旨在辅助而非替代皮肤科医生的解读。
cs.CV / 10 / 2610.02451
A Simulation-Grounded Agentic VLM Framework for Wildfire Monitoring and Reporting
基于模拟的自主视觉-语言模型框架用于野火监测与报告
Abstract
Effective wildfire monitoring requires relating visual evidence to physical fire dynamics, yet real videos with synchronized physical annotations are scarce and high-fidelity 3D simulation is costly. We present a simulation-grounded vision-language model (VLM) framework that automatically converts 2D wildfire simulations into labeled video episodes. A fixed Blender mapping produces low-detail 3D proxies aligned with simulator terrain, fuel layout, fire activity, and wind cues; controllable video generation supplies richer appearance. The proxies are intermediate representations rather than finely rendered final scenes. Generated videos and simulator labels form reusable multimodal memory for a training-free multi-agent VLM system that retrieves reference episodes, reconciles visual and memory-based predictions, and produces structured wildfire reports. On held-out generated episodes, video memory achieves 51.5% exact four-tag accuracy, compared with 22.6% for direct VLM querying and 16-17% for text-only memory; the complete system achieves 77.3% accuracy on six simulator-derived report fields. Component ablations, cross-generator tests, and three real-UAV evaluations assess retrieval, reporting, generator changes, and observable monitoring tasks. The framework connects automatic simulation-to-proxy conversion with memory-based VLM reasoning under scarce real-world physical annotations.
Chinese Translation
有效的野火监测需要将视觉证据与物理火灾动态相关联，但同步物理注释的真实视频稀缺，而高保真3D模拟成本高昂。我们提出了一种基于模拟的视觉-语言模型（VLM）框架，该框架能够自动将2D野火模拟转换为标记视频片段。固定的Blender映射生成与模拟器地形、燃料布局、火灾活动和风线索对齐的低细节3D代理；可控的视频生成提供了更丰富的外观。这些代理是中间表示，而非精细渲染的最终场景。生成的视频和模拟器标签形成可重用的多模态记忆，用于一个无训练的多智能体VLM系统，该系统检索参考片段，调和视觉和基于记忆的预测，并生成结构化的野火报告。在保留的生成片段上，视频记忆的准确率达到51.5%的四标签精确度，而直接VLM查询为22.6%，仅文本记忆为16-17%；完整系统在六个模拟器衍生报告字段上的准确率达到77.3%。组件消融、跨生成器测试和三次真实无人机评估评估了检索、报告、生成器变化和可观察监测任务。该框架将自动模拟到代理的转换与基于记忆的VLM推理连接起来，解决了真实世界物理注释稀缺的问题。
cs.CV / 11 / 2610.02494
DeepStratNet: A Context-Aware Coordinate Regression Framework for Seismic Horizon Tracking under Sparse Labels
DeepStratNet：一种基于上下文的坐标回归框架，用于稀疏标签下的地震地平线追踪
Abstract
Automatic horizon tracking is a foundational task in 3D seismic interpretation. Most existing deep learning approaches formulate it as dense semantic segmentation, typically using U-Net-based architectures. The model produces a probability map over all pixels that must be post-processed to extract precise horizon coordinates, while horizon picks in time/depth must be converted into dense masks for training. Unpicked seismic traces are consequently treated as background, which can hinder convergence, and both pre- and post-processing can introduce errors into the final interpretation. Moreover, 2D segmentation models do not inherently capture inter-slice context, while 3D models are often computationally prohibitive. We instead formulate horizon tracking as a bounded coordinate regression problem, where the model directly predicts the time/depth coordinate of the target horizon at each lateral position. We propose a lightweight regression head compatible with any pretrained vision backbone, coupled with an LSTM module to model inter-slice context and produce a continuous horizon surface across the volume. A combination of L1 and L2 losses supervises predictions at valid horizon picks, while a geology-informed regularization enforces lateral continuity between successive traces. Under controlled experimental conditions, we evaluate four pretrained vision backbones under both segmentation and regression configurations on a seismic volume from New Zealand. The proposed approach consistently outperforms its segmentation counterparts quantitatively, using metrics including RMSE and PCC, and qualitatively, while also demonstrating greater robustness to increasing sparsity of training picks. Finally, we show that prediction variation across successive traces captures local variations in geological complexity, providing an automated quality control measure for downstream seismic interpretation.
Chinese Translation
自动地平线追踪是三维地震解释中的基础任务。现有的大多数深度学习方法将其表述为密集语义分割，通常使用基于 U-Net 的架构。该模型生成一个覆盖所有像素的概率图，需要后处理以提取精确的地平线坐标，而时间/深度上的地平线选择必须转换为密集掩膜以进行训练。因此，未选取的地震轨迹被视为背景，这可能会阻碍收敛，且前处理和后处理都可能在最终解释中引入错误。此外，二维分割模型本质上无法捕捉切片间的上下文，而三维模型通常计算成本过高。我们将地平线追踪重新表述为一个有界坐标回归问题，其中模型直接预测目标地平线在每个横向位置的时间/深度坐标。我们提出了一种轻量级的回归头，兼容任何预训练的视觉主干，并结合 LSTM 模块以建模切片间的上下文，并在整个体积中生成连续的地平线表面。通过 L1 和 L2 损失的组合来监督有效地平线选择的预测，同时地质信息驱动的正则化强制相邻轨迹之间的横向连续性。在受控实验条件下，我们评估了四个预训练视觉主干在来自新西兰的地震体积上的分割和回归配置。所提出的方法在定量上（使用 RMSE 和 PCC 等指标）和定性上均持续优于其分割对应物，同时在训练选择稀疏性增加时表现出更大的鲁棒性。最后，我们展示了相邻轨迹之间的预测变化捕捉了地质复杂性的局部变化，为下游地震解释提供了一种自动化的质量控制措施。
cs.CV / 12 / 2610.02507
MeshQuery: Agentic Seam Planning for UV Parametrization
MeshQuery：用于UV参数化的自主缝规划
Abstract
We present MeshQuery, a training-free agentic approach to automatic UV unwrapping of production-grade quad meshes. A Vision-Language Model (VLM) plans artist-aligned seams using a set of edge-selection tools, conditioned on domain-specific UV-unwrapping knowledge expressed in natural language and refined with a feedback loop. We design a queryable mesh representation together with a domain-specific language (DSL) that enables the agent to retrieve mesh information on demand, express a seam plan as a compact program of edge-selection operators over topological, geometric, and semantic mesh attributes, and iteratively refine it from UV quality feedback. On Adobe Substance 3D and Toys4K meshes, MeshQuery produces 2.9x/4.29x fewer charts and 1.63x/1.7x shorter seams than the strongest baseline, and professional artists prefer its results in 80.9% of comparisons. Ultimately, decoupling high-level intent planning from low-level edge selection and compact mesh representation lets MeshQuery run on different backend VLMs and scale to meshes an order of magnitude larger than autoregressive seam prediction
Chinese Translation
我们提出了MeshQuery，这是一种无训练的自主方法，用于自动展开生产级四边形网格的UV。视觉-语言模型（Vision-Language Model, VLM）利用一组边选择工具规划与艺术家对齐的缝隙，该过程基于以自然语言表达的领域特定UV展开知识，并通过反馈循环进行优化。我们设计了一种可查询的网格表示法以及一种领域特定语言（Domain-Specific Language, DSL），使得代理能够按需检索网格信息，将缝隙规划表达为关于拓扑、几何和语义网格属性的边选择操作的紧凑程序，并根据UV质量反馈进行迭代优化。在Adobe Substance 3D和Toys4K网格上，MeshQuery生成的图表数量比最强基线减少了2.9倍/4.29倍，缝隙长度减少了1.63倍/1.7倍，且专业艺术家在80.9%的比较中更倾向于其结果。最终，将高层意图规划与低层边选择和紧凑网格表示解耦，使得MeshQuery能够在不同的后端VLM上运行，并扩展到比自回归缝预测大一个数量级的网格。
cs.CV / 13 / 2610.02513
From Fragments to Global Maps: Learning Vectorized Map Aggregation with Large Language Models
从碎片到全球地图：利用大型语言模型学习向量化地图聚合
Abstract
Large-scale vectorized HD maps provide structured road information that is essential for perception, localization, and planning in autonomous driving. Constructing such maps requires aggregating noisy, fragmented, and overlapping local predictions collected along a vehicle trajectory into a coherent global map. Existing aggregation methods typically rely on hand-crafted rules for fragment association and refinement. However, a fixed set of thresholds cannot effectively handle variations in road structures and prediction errors, often requiring detector-specific tuning or manual adjustment. To address this limitation, we propose MapMergeLLM, a data-driven framework that formulates vectorized map aggregation as conditional sequence generation with a large language model. Given serialized local vectorized maps, our model directly predicts the aggregated global map polylines. To reduce dependence on any particular upstream detector, we train the model on synthetic local maps generated from clean vector maps using corruptions that simulate representative prediction errors. We further introduce a coordinate tokenizer with geometry-aware pretraining to precisely represent map coordinates. In addition, we propose a line-level association loss that explicitly supervises correspondences between local observations of the same map element. Experiments on Argoverse2 and nuScenes using multiple recent upstream detectors demonstrate that MapMergeLLM substantially outperforms heuristic and optimization-based aggregation baselines without detector-specific retraining.
Chinese Translation
大规模向量化高清地图提供了结构化的道路信息，这对于自动驾驶中的感知、定位和规划至关重要。构建这样的地图需要将沿车辆轨迹收集的嘈杂、碎片化和重叠的局部预测聚合成一个连贯的全球地图。现有的聚合方法通常依赖于手工制定的规则来进行碎片关联和精炼。然而，固定的阈值集无法有效处理道路结构和预测误差的变化，通常需要特定检测器的调优或手动调整。为了解决这一限制，我们提出了MapMergeLLM，这是一个数据驱动的框架，将向量化地图聚合公式化为使用大型语言模型的条件序列生成。在给定序列化的局部向量化地图的情况下，我们的模型直接预测聚合的全球地图多边形线。为了减少对任何特定上游检测器的依赖，我们在从干净的向量地图生成的合成局部地图上训练模型，使用模拟代表性预测误差的损坏。我们进一步引入了一种具有几何感知预训练的坐标标记器，以精确表示地图坐标。此外，我们提出了一种线级关联损失，明确监督同一地图元素的局部观测之间的对应关系。在使用多个最新上游检测器的Argoverse2和nuScenes上的实验表明，MapMergeLLM在没有特定检测器重新训练的情况下，显著优于启发式和基于优化的聚合基线。
cs.CV / 14 / 2610.02521
Spatial Memory Intelligence: Endowing World Models with Understanding-Driven Long-Term Memory
空间记忆智能：赋予世界模型理解驱动的长期记忆
Abstract
Long-video generation and world models have shown strong potential for interactive entertainment and embodied simulation by predicting future observations conditioned on user actions and historical memory. However, as memory sequences grow longer and their structures become increasingly complex, managing long-range spatial context becomes increasingly challenging, calling for a more intelligent and systematic memory-management strategy. Building on the advancing spatial reasoning capabilities of multimodal large language models (MLLMs) and the broader vision of unified models, we propose Spatial Memory Intelligence (SMI), the first framework to systematically employ an understanding model for spatial-memory management in long-video world models. SMI introduces four coordinated atomic operations: spatial clustering, within-cluster sparsification, action-aware retrieval, and reliability-aware filtering. Extensive experiments across multiple baselines, benchmarks, and world-model backbones demonstrate the effectiveness and generalizability of SMI, achieving comprehensive improvements in memory sparsity, generation stability, and spatial consistency.
Chinese Translation
长视频生成和世界模型在通过预测基于用户行为和历史记忆的未来观察方面，展现了在互动娱乐和具身模拟中的强大潜力。然而，随着记忆序列的增长及其结构日益复杂，管理长距离空间上下文变得愈加具有挑战性，这呼唤一种更智能和系统的记忆管理策略。基于多模态大型语言模型（MLLMs）日益增强的空间推理能力以及统一模型的更广泛愿景，我们提出了空间记忆智能（SMI），这是第一个系统性地采用理解模型进行长视频世界模型中空间记忆管理的框架。SMI引入了四个协调的基本操作：空间聚类、聚类内稀疏化、基于动作的检索和基于可靠性的过滤。通过在多个基准、基线和世界模型骨干网络上的广泛实验，证明了SMI的有效性和普适性，在记忆稀疏性、生成稳定性和空间一致性方面实现了全面的改进。
cs.CV / 15 / 2610.02561
Oracle headroom without signal: null-calibrated evaluation of candidate selection for thermal heart rate estimation
无信号的Oracle余量：热心率估计候选选择的零校准评估
Abstract
Camera-based physiological monitoring can produce multiple estimates from several facial regions, extraction methods, and processing settings. Signal quality indices aim to select reliable estimates without a physiological reference, and their potential is often assessed with an oracle that selects the estimate closest to the reference in each window. This retrospective selection can reward chance agreement. We model the effect with order statistics. For K independent candidates unrelated to the reference, the expected oracle error decreases approximately as 1/K. We analyze thermal heart rate estimation on 96 iBVP recordings with 168 candidates per 10 s window. The oracle achieves a mean absolute error of 0.91 bpm, compared with 10.74 bpm for the best fixed configuration, 18.03 bpm for the best quality index, and 8.61 bpm for a constant predictor. With K = 24, a forehead signal from another recording matches the correct one, with 4.62 against 4.61 bpm. Oracle evaluations should report candidate count, valid coverage, and matched null controls.
Chinese Translation
基于摄像头的生理监测可以从多个面部区域、提取方法和处理设置中产生多个估计值。信号质量指数旨在选择可靠的估计值，而无需生理参考，其潜力通常通过一个Oracle进行评估，该Oracle在每个窗口中选择与参考最接近的估计。这种回顾性选择可能会奖励偶然的一致性。我们用顺序统计学模型化这一效应。对于K个与参考无关的独立候选者，预期的Oracle误差大约以1/K的速度减少。我们分析了96个iBVP录音中的热心率估计，每10秒窗口有168个候选者。与最佳固定配置的10.74 bpm、最佳质量指数的18.03 bpm和常量预测器的8.61 bpm相比，Oracle的平均绝对误差为0.91 bpm。当K = 24时，来自另一个录音的额头信号与正确信号匹配，分别为4.62 bpm和4.61 bpm。Oracle评估应报告候选者数量、有效覆盖范围和匹配的零对照。
cs.CV / 16 / 2610.02567
DAGS: Disentangled Appearance-and-Geometry Steering of a Frozen Image DiT for Temporally Stabilized Generative Rendering
DAGS：对冻结图像DiT进行解耦外观与几何引导的时间稳定生成渲染
Abstract
Diffusion transformers (DiTs) generate high-fidelity images from text and image conditions, but their outputs carry large variance and their faithfulness to a desired target depends heavily on how the condition is supplied. We present DAGS, a lightweight, attention-free, disentangled appearance and geometry conditioning scheme that steers a frozen image DiT to produce high-fidelity, highly faithful, and independently controllable renders. Two small convolutional encoders compute conditioning features once per frame and inject them as a learned, per-layer, element-wise residual into the image tokens, avoiding the quadratic cost of stacking conditions through attention. Because control and temporal handling live outside the frozen backbone, we retain its vast pretrained prior and eliminate backbone-overfitting risk. We further add a small recurrent lighting stabilizer and a training-free temporal guidance term that, coupled with our conditioning, elevate a per-frame image model into a streaming renderer. DAGS produces controllable, high-quality renders at a fraction of the compute of path tracing; it is not real-time, trading compute for controllability and quality. On a matched 1-spp + G-buffer input, per-frame DAGS reconstructs +8.6 dB / +10.1 dB PSNR over the real-time denoiser Intel OIDN and the diffusion renderer RGB<->X while being 2.5-8x more temporally stable perceptually (temporal-LPIPS flicker).
Chinese Translation
扩散变换器（DiTs）能够根据文本和图像条件生成高保真图像，但其输出存在较大方差，且对目标的忠实度在很大程度上依赖于条件的提供方式。我们提出了DAGS，这是一种轻量级、无注意力的解耦外观与几何条件方案，能够引导冻结图像DiT生成高保真、高忠实度且可独立控制的渲染图像。两个小型卷积编码器每帧计算一次条件特征，并将其作为学习的、逐层的元素级残差注入到图像标记中，从而避免了通过注意力堆叠条件所带来的二次计算成本。由于控制和时间处理位于冻结的主干网络之外，我们保留了其庞大的预训练先验，并消除了主干过拟合的风险。我们进一步添加了一个小型递归光照稳定器和一个无训练的时间引导项，结合我们的条件，使每帧图像模型提升为流式渲染器。DAGS以路径追踪计算量的一小部分生成可控的高质量渲染；它不是实时的，牺牲了计算以换取可控性和质量。在匹配的1-spp + G-buffer输入下，每帧DAGS在实时去噪器Intel OIDN和扩散渲染器RGB<->X上重建了+8.6 dB / +10.1 dB的PSNR，同时在感知上具有2.5-8倍的时间稳定性（时间-LPIPS闪烁）。
cs.CV / 17 / 2610.02580
Physical AI Smart Spaces: A Large-Scale Benchmark for Multi-Camera 3D Perception in Smart Spaces
物理人工智能智能空间：多摄像头三维感知的大规模基准测试
Abstract
Physical AI Smart Spaces is, to the best of our knowledge, the first benchmark to simultaneously provide large-scale, multi-class, and multi-camera 3D perception data for indoor smart spaces. It contains over 280 hours of synchronized 1080p footage captured by nearly 1,800 cameras in warehouses, hospitals, retail venues, and similar settings, together with automatic annotations for multi-camera identities, 2D bounding boxes, 3D bounding boxes, camera calibration, and depth where available. The benchmark spans Isaac Sim synthetic generation, Cosmos Transfer appearance augmentation, and real-world Sim2Real evaluation. For the real-world target, we include two warehouse deployments with time-synchronized streams, automatic VGGT-based calibration, and a 3D labeling interface that projects world-frame 3D boxes into each view for cross-camera verification. We describe the dataset scope, annotation and calibration schema, generation workflow, benchmark protocols, and official evaluation system, which standardizes submission format, and leaderboard reporting. A central contribution is a 3D instantiation of Higher Order Tracking Accuracy (HOTA), extending the usual 2D box-based tracking evaluation to 3D locations and 3D boxes. We further report empirical baselines from the AI City Challenge leaderboards, showing how methods evolve from person-only 3D location tracking to multi-class 3D box tracking under realistic smart-space constraints. The release is available at https://huggingface.co/datasets/nvidia/PhysicalAI-SmartSpaces.
Chinese Translation
物理人工智能智能空间是我们所知的首个基准测试，旨在同时提供针对室内智能空间的大规模、多类别和多摄像头的三维感知数据。该数据集包含超过280小时的同步1080p视频，由近1800个摄像头在仓库、医院、零售场所及类似环境中捕获，并提供多摄像头身份、二维边界框、三维边界框、摄像头标定及可用深度的自动注释。该基准测试涵盖了Isaac Sim合成生成、Cosmos Transfer外观增强和现实世界Sim2Real评估。对于现实世界目标，我们包括两个具有时间同步流的仓库部署，自动VGGT基础的标定，以及一个三维标注界面，将世界坐标系中的三维框投影到每个视图中以进行跨摄像头验证。我们描述了数据集的范围、注释和标定方案、生成工作流程、基准测试协议和官方评估系统，后者标准化了提交格式和排行榜报告。一个核心贡献是三维实例化的高阶跟踪准确性（Higher Order Tracking Accuracy，HOTA），将通常基于二维框的跟踪评估扩展到三维位置和三维框。我们进一步报告了来自AI City Challenge排行榜的经验基线，展示了在现实智能空间约束下，方法如何从仅基于人的三维位置跟踪演变为多类别三维框跟踪。该数据集已在 https://huggingface.co/datasets/nvidia/PhysicalAI-SmartSpaces 上发布。
cs.CV / 18 / 2610.02597
GRAFT: Growing Agglomerative Foundation Models via Continual Teacher Distillation
GRAFT：通过持续教师蒸馏增长聚合基础模型
Abstract
Vision foundation models such as DINOv2, SigLIP2, and MASt3R develop complementary capabilities from different pretraining objectives, yet their knowledge remains distributed across separate, specialized models. Multi-teacher knowledge distillation offers a path toward consolidating these capabilities into a single agglomerative backbone, but existing approaches assume a fixed set of teachers, and incorporating a new teacher requires repeating expensive joint distillation over the entire teacher set. We introduce GRAFT, a continual multi-teacher distillation framework that enables a unified backbone to progressively acquire capabilities from an open-ended sequence of foundation models. When a new teacher arrives, GRAFT treats the previously distilled model as a teacher for preserving learned capabilities, while the current student jointly learns from both the previous model and the incoming teacher. Furthermore, to reconcile the incompatible representation geometries of heterogeneous teachers, we introduce Teacher Specific Readout Tokens, which grant each teacher an independent read-out of the shared encoder, together with Geometry Agnostic Relational Loss that aligns a vision-language teacher by matching image-text similarity structures rather than raw feature values. We provide GRAFT model, which is a single, continually extensible backbone that unifies five domains, including image understanding, 2D dense prediction, 3D human pose estimation, 3D vision, and vision-language, delivering strong performance across all of them while acquiring each new capability at the cost of a single distillation rather than a full re-distillation.
Chinese Translation
视觉基础模型如 DINOv2、SigLIP2 和 MASt3R 从不同的预训练目标中发展出互补的能力，但它们的知识仍然分散在各个独立的专业模型中。多教师知识蒸馏为将这些能力整合到一个单一的聚合主干提供了一条路径，但现有的方法假设教师集是固定的，且引入新教师需要对整个教师集重复进行昂贵的联合蒸馏。我们提出了 GRAFT，一个持续的多教师蒸馏框架，使得统一的主干能够逐步从开放序列的基础模型中获取能力。当新教师到来时，GRAFT 将先前蒸馏的模型视为教师，以保留已学习的能力，同时当前学生模型同时从前一个模型和新到来的教师中学习。此外，为了调和异构教师的不兼容表示几何，我们引入了教师特定的读取标记（Teacher Specific Readout Tokens），使每个教师能够独立读取共享编码器，并结合几何无关关系损失（Geometry Agnostic Relational Loss），通过匹配图像-文本相似性结构而非原始特征值来对齐视觉-语言教师。我们提供的 GRAFT 模型是一个单一的、持续可扩展的主干，统一了五个领域，包括图像理解、2D 密集预测、3D 人体姿态估计、3D 视觉和视觉-语言，在所有领域中均表现出色，同时以单次蒸馏的成本获取每项新能力，而不是完全重新蒸馏。
cs.CV / 19 / 2610.02626
Imagine the Future, Internalize the Gist: Efficient VLA Reasoning via Internalized Spatiotemporal Imagination
想象未来，内化要点：通过内化时空想象实现高效的视觉-语言-行动（VLA）推理
Abstract
Vision-language-action (VLA) models increasingly incorporate intermediate reasoning to improve robotic manipulation, yet existing approaches primarily reason about observed states without explicitly anticipating future scene evolution. Extending such reasoning to explicit future rollouts at every inference step, however, introduces substantial computational overhead. We propose IG-VLA, a VLA reasoning framework that enables models to imagine the future and internalize the gist. Our Latent Spatiotemporal Reasoning learns to imagine task-relevant future scene evolution directly in visual representation space, guiding action prediction without costly pixel-level video generation. To further reduce inference overhead, we introduce Scene Gist Memory, which internalizes reasoning-derived scene-behavior associations into a compact Scene Gist Token, preserving the benefits of future reasoning while bypassing explicit future imagination at inference. Extensive experiments on LIBERO, LIBERO-Plus, and VLABench demonstrate the effectiveness and efficiency of IG-VLA. On the LIBERO-Plus Language suite, both the reasoning and gist policies outperform the strongest baseline by nearly 6% in success rate. The gist policy also achieves up to 6.38x speedup over baselines, reducing inference latency from 1081ms to 169.5ms per action chunk on a single NVIDIA A6000 GPU. These results demonstrate that future spatiotemporal reasoning can be effectively internalized for efficient VLA deployment.
Chinese Translation
视觉-语言-行动（VLA）模型越来越多地融入中间推理以改善机器人操作，然而现有的方法主要针对观察到的状态进行推理，而未明确预见未来场景的演变。然而，将这种推理扩展到每个推理步骤的明确未来展开会引入大量的计算开销。我们提出了IG-VLA，一个VLA推理框架，使模型能够想象未来并内化要点。我们的潜在时空推理直接在视觉表示空间中学习想象与任务相关的未来场景演变，引导行动预测，而无需昂贵的像素级视频生成。为了进一步减少推理开销，我们引入了场景要点记忆，将推理得出的场景-行为关联内化为紧凑的场景要点标记，保留未来推理的优势，同时在推理时绕过明确的未来想象。在LIBERO、LIBERO-Plus和VLABench上的大量实验表明了IG-VLA的有效性和高效性。在LIBERO-Plus语言套件上，推理和要点策略的成功率均比最强基线高出近6%。要点策略在基线之上还实现了高达6.38倍的加速，将每个动作块的推理延迟从1081毫秒降低到169.5毫秒，使用单个NVIDIA A6000 GPU。这些结果表明，未来时空推理可以有效内化，以实现高效的VLA部署。
cs.CV / 20 / 2610.02647
Capturing Dynamics: The 4D Facial Expression Intensity Dataset
捕捉动态：4D面部表情强度数据集
Abstract
The estimation and analysis of facial expression intensity play a crucial role in affective communication and human-computer interaction. Previous research has primarily focused on detecting and estimating facial expression intensity from frame-level 2D representations. However, this limitation restricts a comprehensive understanding of real-world facial expressions, as they are inherently 3D and temporally continuous. This paper investigates the perception of facial expression intensity by introducing the 4D Facial Expression Intensity Dataset (4DFEID). We employ a parametric face model and compile a total of 2,869 mesh sequences with controlled geometric variations, generating 4D data instances with diverse peak intensities and identity attributes. Using a Likert scale, we collect more than 90,000 subjective intensity perception ratings via a crowdsourcing platform. We explore various architectures and aggregation methods to establish baselines for episode intensity estimation on the new dataset, revealing that spatial-temporal graph models consistently outperform traditional frame-aggregation methods. In contrast to existing datasets that rely on 2D static imagery, the proposed 4D-FEID dataset provides the community with a unique and vital resource for investigating the perception of facial expression intensity through the use of dynamic 3D stimuli. By offering high-fidelity, spatio-temporally coherent facial data, 4D-FEID establishes a new foundation for research into more nuanced and naturalistic expression analysis, thereby addressing a gap in the current landscape of affective computing and human-computer interaction studies. The dataset is available at link.
Chinese Translation
面部表情强度的估计和分析在情感交流和人机交互中发挥着至关重要的作用。以往的研究主要集中在从帧级2D表示中检测和估计面部表情强度。然而，这一局限性限制了对现实世界面部表情的全面理解，因为面部表情本质上是三维的，并且在时间上是连续的。本文通过引入4D面部表情强度数据集（4DFEID）来研究面部表情强度的感知。我们采用参数化面部模型，编制了总计2,869个具有可控几何变化的网格序列，生成具有多样化峰值强度和身份属性的4D数据实例。通过使用李克特量表，我们通过众包平台收集了超过90,000个主观强度感知评分。我们探索了各种架构和聚合方法，以在新数据集上建立情节强度估计的基线，结果显示时空图模型始终优于传统的帧聚合方法。与依赖于2D静态图像的现有数据集相比，所提出的4D-FEID数据集为研究通过动态3D刺激感知面部表情强度提供了独特而重要的资源。通过提供高保真、时空一致的面部数据，4D-FEID为更细致和自然的表情分析研究奠定了新的基础，从而填补了当前情感计算和人机交互研究领域的空白。该数据集可在链接处获取。
cs.CV / 21 / 2610.02660
SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching
SpectralCache：通过光谱特征缓存加速基于扩散的世界模型
Abstract
Diffusion-based world models enable high-quality interactive environment generation but suffer from substantial inference overhead due to repeated Transformer evaluations during denoising. Existing caching methods mainly exploit temporal redundancy at the feature or token level, leaving the underlying mathematical structure of diffusion features largely unexplored. In this work, we reveal that world-model features exhibit highly stable singular subspaces across nearby denoising steps, while their singular values follow predictable evolution patterns. Building on this observation, we propose SpectralCache, a training-free spectral caching framework that reuses stable singular subspaces and estimates only low-dimensional singular values through linear extrapolation. We further exploit the spectral consistency between neighboring full-computation features to skip selected expensive backbone evaluations via singular value scaling. Extensive experiments on representative world models demonstrate that SpectralCache consistently improves inference efficiency while preserving generation quality. On HunyuanWorld-Voyager-13B, SpectralCache achieves 5.22x acceleration while maintaining a WorldScore of 65.90 for static scenes, substantially outperforming existing training-free caching methods in inference efficiency.
Chinese Translation
基于扩散的世界模型能够实现高质量的交互环境生成，但由于在去噪过程中反复进行Transformer评估，导致推理开销显著。现有的缓存方法主要利用特征或标记级别的时间冗余，未能深入探讨扩散特征的基本数学结构。在本研究中，我们揭示了世界模型特征在相邻去噪步骤中表现出高度稳定的奇异子空间，同时其奇异值遵循可预测的演变模式。基于这一观察，我们提出了SpectralCache，一种无训练的光谱缓存框架，重用稳定的奇异子空间，并通过线性外推仅估计低维奇异值。我们进一步利用相邻全计算特征之间的光谱一致性，通过奇异值缩放跳过选定的高成本主干评估。在具有代表性的世界模型上的大量实验表明，SpectralCache在保持生成质量的同时，始终提高了推理效率。在HunyuanWorld-Voyager-13B上，SpectralCache实现了5.22倍的加速，同时静态场景的WorldScore保持在65.90，显著优于现有的无训练缓存方法在推理效率上的表现。
cs.CV / 22 / 2610.02666
CHASE-VLA: Post-Training Quantization Framework for Vision-Language-Action Models with Chunk-Aware Scale Estimation
CHASE-VLA：面向视觉-语言-动作模型的后训练量化框架，具有块感知的尺度估计
Abstract
Vision-Language-Action (VLA) models map visual observations and language instructions to continuous robot actions, but a diffusion-based action expert (AE) poses a key challenge for low-bit post-training quantization (PTQ). The AE is repeatedly invoked across denoising steps and policy queries, where fixed calibration scales can be mismatched with activation ranges that vary with denoising progress and intended motion. We propose CHASE-VLA, a chunk-aware PTQ method that exploits a VLA-specific signal readily available from the policy: the generated action chunk, including its unexecuted future suffix. Rather than relying only on static scale matching for AE layers, CHASE-VLA combines the previously generated chunk as causal action context with denoising step group information to adapt AE activation scales. This enables W4A4 quantization of both MLP and attention projections in the repeated AE without modifying the pretrained policy. On LIBERO, CHASE-VLA achieves 97.3% average success rate on $\pi_{0.5}$ when both MLP and attention projections in the AE are quantized to W4A4, restoring FP16-level performance. CHASE-VLA also reduces the weight storage of the quantized AE linear layers by 73.4% and their single-chunk memory traffic by 70.9% and 71.2% on $\pi_{0.5}$ and GR00T N1.6, respectively, with a predictor overhead of at most 1.26% of the saved storage.
Chinese Translation
视觉-语言-动作（VLA）模型将视觉观察和语言指令映射到连续的机器人动作，但基于扩散的动作专家（AE）在低比特后训练量化（PTQ）中面临关键挑战。AE在去噪步骤和策略查询中被反复调用，其中固定的校准尺度可能与随去噪进展和预期运动而变化的激活范围不匹配。我们提出了CHASE-VLA，这是一种块感知的PTQ方法，利用从策略中可用的VLA特定信号：生成的动作块，包括其未执行的未来后缀。CHASE-VLA并不单纯依赖于AE层的静态尺度匹配，而是将先前生成的块作为因果动作上下文与去噪步骤组信息结合，以适应AE激活尺度。这使得在不修改预训练策略的情况下，可以对重复的AE中的多层感知机（MLP）和注意力投影进行W4A4量化。在LIBERO上，当AE中的MLP和注意力投影量化为W4A4时，CHASE-VLA在$ ext{π}_{0.5}$上实现了97.3%的平均成功率，恢复了FP16级别的性能。CHASE-VLA还将量化AE线性层的权重存储减少了73.4%，并在$ ext{π}_{0.5}$和GR00T N1.6上分别减少了它们的单块内存流量70.9%和71.2%，预测器的开销最多为节省存储的1.26%。
cs.CV / 23 / 2610.02718
Revisiting Visual Representation Enhancement of VLMs via Kernel Canonical Correlation Analysis
通过核典型相关分析重新审视视觉语言模型的视觉表示增强
Abstract
Vision-language models such as CLIP exhibit strong semantic generalization, but remain limited in fine-grained visual perception. A recent work named KUEA presents a natural remedy by finetuning the image encoder under the supervision of the vision-centric DINOv2 to align their kernel matrices element-wisely, while regularizing the embeddings to remain close to the pretrained visual encoder for preserving image-text semantics in CLIP. However, we show that diminishing the role of the alignment loss to DINOv2 does not necessarily degrade its fine-grained visual performance, suggesting that the kernel-matrix discrepancy may be insufficient for further visual representation enhancement, motivating us to revisit the alignment formulation. In this work, we present a novel perspective to characterize representation alignment on feature subspaces through Kernel Canonical Correlation Analysis (KCCA), which maximizes the projection correlations. In optimization, we derive an efficient end-to-end training scheme upon KKT conditions, avoiding the eigenvalue problem in KCCA. Further, we extend our method into a 3-view formulation, i.e., 3vKCCA, in which the projections from the pretrained text encoder are also incorporated under a unified optimization framework for joint alignment. With CLIP ViT-L/14 on ImageNet-1K, our 3vKCCA improves the MMVP-VLM accuracy from 17.8 to 25.9, substantially outperforming the existing methods, and meanwhile maintains zero-shot image--text retrieval performance.
Chinese Translation
视觉语言模型（如 CLIP）展现了强大的语义泛化能力，但在细粒度视觉感知方面仍然有限。最近的研究工作 KUEA 提出了一个自然的解决方案，通过在以视觉为中心的 DINOv2 的监督下微调图像编码器，以逐元素对齐它们的核矩阵，同时对嵌入进行正则化，使其保持接近预训练的视觉编码器，从而保留 CLIP 中的图像-文本语义。然而，我们表明，降低对 DINOv2 的对齐损失的作用并不一定会降低其细粒度视觉性能，这表明核矩阵差异可能不足以进一步增强视觉表示，这促使我们重新审视对齐公式。在本研究中，我们提出了一种新颖的视角，通过核典型相关分析（KCCA）来表征特征子空间上的表示对齐，最大化投影相关性。在优化过程中，我们推导出了一种高效的端到端训练方案，基于 KKT 条件，避免了 KCCA 中的特征值问题。此外，我们将我们的方法扩展为三视图公式，即 3vKCCA，其中预训练文本编码器的投影也在统一优化框架下被纳入以实现联合对齐。在 ImageNet-1K 上使用 CLIP ViT-L/14，我们的 3vKCCA 将 MMVP-VLM 的准确率从 17.8 提高到 25.9，显著优于现有方法，同时保持零样本图像-文本检索性能。
cs.CV / 24 / 2610.02726
SymRegFlow: Symmetry-Regularized Flow Matching for Video World Models
SymRegFlow：用于视频世界模型的对称正则化流匹配
Abstract
Flow-matching-based multi-view world models generate realistic videos, but are commonly restricted to fixed camera rigs. Extending them to continuously varying camera poses requires paired pose--video observations with dense pose coverage, which are costly to acquire. We introduce \emph{SymRegFlow}, a symmetry-regularized flow-matching framework for multi-view-consistent video generation across continuous viewpoints without ground-truth novel-view RGB supervision. For each target pose, SymRegFlow geometrically warps source views into noisy anchors and combines masked dual-anchor supervision with cross-anchor denoising-output consistency to mitigate anchor-specific errors. Under an affine Gaussian surrogate, we prove that suitable consistency regularization recovers the clean-reference optimum at fixed noise levels, strictly outperforming single- and merged-anchor baselines. Experiments on Cosmos-Drive-Dreams and nuScenes demonstrate high-quality, multi-view-consistent autonomous-driving video generation: on nuScenes, SymRegFlow achieves the lowest FVD and FVMD among the evaluated baselines, reducing FVD by over 31\% relative to the best baseline, and source-conditioned inference also attains the best FID and instance preservation.
Chinese Translation
基于流匹配的多视角世界模型能够生成真实感视频，但通常限制于固定相机设备。将其扩展到连续变化的相机姿态需要配对的姿态-视频观测，并且需要密集的姿态覆盖，这在获取上成本较高。我们提出了 extit{SymRegFlow}，一种对称正则化流匹配框架，用于在没有真实新视角RGB监督的情况下，实现跨连续视点的一致视频生成。对于每个目标姿态，SymRegFlow将源视图几何变形为噪声锚点，并结合掩蔽的双锚点监督与跨锚点去噪输出一致性，以减轻特定锚点的误差。在仿射高斯替代下，我们证明了合适的一致性正则化能够在固定噪声水平下恢复干净参考的最优解，严格优于单锚点和合并锚点的基线。在Cosmos-Drive-Dreams和nuScenes上的实验表明，SymRegFlow能够生成高质量的多视角一致的自动驾驶视频：在nuScenes上，SymRegFlow在评估的基线中实现了最低的FVD和FVMD，相较于最佳基线FVD降低超过31\%，并且源条件推理也达到了最佳的FID和实例保留。
cs.CV / 25 / 2610.02753
Correcting Guided Diffusion Trajectories with Spectral Alignment
通过谱对齐修正引导扩散轨迹
Abstract
The practical success of conditional image generation hinges on fine-grained differences in condition alignment and visual fidelity. Classifier-free guidance (CFG) is central to this success, but its lack of an explicit criterion makes it difficult to assess whether the guided trajectory is progressing as intended. To address this gap, we show that spectral alignment provides a principled criterion for understanding guidance behavior and improving guided diffusion sampling through adaptive correction. Our analysis identifies the spectra of intermediate states as an indicator of consistency with the expected spectral evolution of the forward process. Based on this observation, we introduce Spectral Correction Guidance, a method that corrects deviations from an analytic reference spectrum during sampling. The proposed method is training-free and applicable across diffusion backbones and conditional generation tasks without modifying the underlying model. Experiments demonstrate consistent gains in preference-based metrics over baseline guidance methods in text-to-image generation and improved generation quality over CFG on ImageNet. These improvements persist across a range of guidance scales and with fewer denoising steps. Our analyses and ablations provide insight into guidance behavior and how the proposed method affects generation quality.
Chinese Translation
条件图像生成的实际成功依赖于条件对齐和视觉保真度的细微差异。无分类器引导（Classifier-free Guidance, CFG）是这一成功的核心，但其缺乏明确标准使得评估引导轨迹是否按预期进展变得困难。为了解决这一问题，我们展示了谱对齐提供了一种理解引导行为和通过自适应修正改善引导扩散采样的原则性标准。我们的分析将中间状态的谱视为与前向过程预期谱演变一致性的指标。基于这一观察，我们引入了谱修正引导（Spectral Correction Guidance），该方法在采样过程中修正与解析参考谱的偏差。所提出的方法无需训练，适用于各种扩散骨干网络和条件生成任务，而无需修改基础模型。实验结果表明，在文本到图像生成中，相较于基线引导方法，偏好基础指标有了一致的提升，并且在ImageNet上生成质量优于CFG。这些改进在不同的引导尺度和更少的去噪步骤下依然存在。我们的分析和消融实验为引导行为提供了洞见，并揭示了所提方法如何影响生成质量。
cs.CV / 26 / 2610.02755
FiberGeoText: A Vision-Language Model for Population- Level Organization of Superficial White Matter
FiberGeoText：一种用于浅层白质群体级组织的视觉-语言模型
Abstract
The superficial white matter (SWM), a critical brain region for cognition across the lifespan and brain disease, contains abundant short-range association fibers whose organization remains incompletely characterized, in part because the short trajectories and highly variable cortical folding make correspondence across individuals challenging. Anatomically corresponding connections may vary in spatial location across individuals and therefore may not be adequately defined by geometric proximity alone. We introduce FiberGeoText (FGT), a vision-language model (VLM) for organizing short-range superficial white matter (SWM) streamlines reconstructed from ultra-high-resolution diffusion MRI into population-level clusters. FGT jointly represents three complementary properties of each streamline: its three-dimensional trajectory, its cortical anatomical context, and its shape. Cortical endpoint information from multiple parcellation schemes is expressed as text and encoded using a pretrained large language model (LLM), enabling heterogeneous anatomical descriptions to contribute to a common continuous representation. We evaluated FGT on acquired submillimeter 0.76 mm diffusion MRI data. Compared with state-of-the-art (SOTA) methods, FGT produced substantially greater cortical parcel coherence, within-cluster shape consistency, cluster-size consistency, and cross-subject correspondence. The trained model also generalizes well to unseen subjects with an average of 96.7% of the 5,000 learned clusters recovered, and high consistency of cluster structure between training and testing data. Together, these findings demonstrate that integrating geometric, anatomical, and shape information by learning multimodal deep embeddings with a VLM model enables robust learning of population-consistent SWM organization despite interindividual anatomical variability.
Chinese Translation
浅层白质（SWM）是一个在整个生命周期和脑疾病中对认知至关重要的脑区，包含丰富的短程联结纤维，其组织特征尚未完全表征，部分原因在于短轨迹和高度可变的皮层折叠使得个体间的对应关系变得具有挑战性。解剖上对应的连接在个体间的空间位置可能有所不同，因此仅通过几何接近性可能无法充分定义。我们提出了FiberGeoText（FGT），一种视觉-语言模型（VLM），用于将从超高分辨率扩散MRI重建的短程浅层白质（SWM）流线组织成群体级集群。FGT共同表示每条流线的三个互补特性：其三维轨迹、其皮层解剖背景和其形状。来自多种分区方案的皮层端点信息以文本形式表达，并使用预训练的大型语言模型（LLM）进行编码，使得异构的解剖描述能够贡献于一个共同的连续表示。我们在获取的亚毫米0.76 mm扩散MRI数据上评估了FGT。与最先进的方法（SOTA）相比，FGT在皮层分区一致性、集群内形状一致性、集群大小一致性和跨个体对应性方面显著提高。训练后的模型在未见过的个体上也表现良好，平均恢复了5,000个学习集群中的96.7%，并且训练和测试数据之间的集群结构高度一致。这些发现共同表明，通过学习多模态深度嵌入与VLM模型结合几何、解剖和形状信息，能够在个体间解剖变异的情况下，稳健地学习群体一致的SWM组织。
cs.CV / 27 / 2610.02779
TRAC: Trajectory-aware Reuse and Adaptive Correction for Efficient Autoregressive Video Generation
TRAC：面向轨迹的重用与自适应修正以实现高效自回归视频生成
Abstract
In this paper, we present trajectory-aware reuse and adaptive correction (TRAC), a training-free framework for efficient autoregressive (AR) video generation. Existing acceleration methods mainly target single-trajectory generation with bidirectional attention. AR video generation, by contrast, sequentially couples chunk-level denoising trajectories. Consequently, approximation errors accumulate and propagate through the generation process. TRAC addresses this challenge with three components, including robust cumulative scheduling (RCS), autoregressive trajectory-aware guidance scheduling (ATGS), and spectral structure correction (SSC). RCS selects cache reuse schedules by cumulative rollout error and cross-chunk/prompt variation. ATGS coordinates CFG refreshes along the global AR trajectory. SSC restores low-frequency structure of the first chunk to correct long-term structural loss. Experiments on SkyReels-V2 and FramePack-F1 show that, compared with existing methods, TRAC achieves both the highest inference efficiency and the best generation quality for AR video generation.
Chinese Translation
在本文中，我们提出了一种面向轨迹的重用与自适应修正（TRAC）框架，用于高效的无训练自回归（AR）视频生成。现有的加速方法主要针对具有双向注意力的单轨迹生成。相比之下，自回归视频生成则是将块级去噪轨迹按顺序耦合。因此，近似误差在生成过程中累积并传播。TRAC通过三个组成部分来解决这一挑战，包括稳健的累积调度（RCS）、自回归轨迹感知引导调度（ATGS）和频谱结构修正（SSC）。RCS通过累积回滚误差和跨块/提示变化选择缓存重用调度。ATGS协调沿全局自回归轨迹的CFG刷新。SSC恢复第一个块的低频结构，以修正长期结构损失。在SkyReels-V2和FramePack-F1上的实验表明，与现有方法相比，TRAC在自回归视频生成中实现了最高的推理效率和最佳的生成质量。
cs.CV / 28 / 2610.02799
FUSEye: Training-Light Fisheye Detection with Overlapping Views and Zero-Initialized Adapters
FUSEye：基于重叠视图和零初始化适配器的轻量级鱼眼检测训练
Abstract
Fisheye cameras give mobile robots a single-sensor, low-cost view of their surroundings, yet the COCO-pretrained detectors that practitioners routinely reuse fail on them: strong radial distortion warps local image structure, while boundary compression shrinks objects to near-invisible sizes. Full fine-tuning closes much of the gap but requires abundant fisheye labels and compute. We present FUSEye, a training-light framework that turns a frozen-backbone COCO-pretrained extra-large YOLO26 detector (YOLO26-x) into a fisheye detector. FUSEye adds roughly 227k new parameters while updating the inserted modules and the pretrained detection head. It addresses the transfer gap at three causally linked levels. At the input level, overlapping grid view generation and box remapping (GridViews) enlarge compressed boundary regions. At the feature level, zero-initialized residual adapters (Z-Adapters) correct distortion-induced feature misalignment. At the decision level, learned cross-projection agreement fusion (AgreeFusion) promotes low-confidence detections only when they are supported by consistent evidence across multiple views. On the WoodScape surround-view fisheye benchmark, FUSEye raises YOLO26-x from 0.148 to 0.266 mAP50 and retains 84.3% fully fine-tuned accuracy. Moreover, randomly using only 25% of the labeled training images, FUSEye achieves 0.2597 mAP50, retaining 97.6% of its full-label performance. FUSEye also consistently improves YOLOv8-11 detectors, showing that the recipe is architecture-agnostic. Source code will be available at https://github.com/Su-wenya/FUSEye.
Chinese Translation
鱼眼相机为移动机器人提供了单传感器、低成本的环境视图，但实践中常用的COCO预训练检测器在此类相机上表现不佳：强烈的径向畸变扭曲了局部图像结构，而边界压缩使得物体缩小至几乎不可见的尺寸。全面微调虽然能缩小大部分差距，但需要大量的鱼眼标签和计算资源。我们提出了FUSEye，一个轻量级训练框架，将冻结的COCO预训练超大YOLO26检测器（YOLO26-x）转变为鱼眼检测器。FUSEye大约增加了227k个新参数，同时更新了插入模块和预训练检测头。它在三个因果关联的层面上解决了迁移差距。在输入层面，重叠网格视图生成和框重映射（GridViews）扩大了压缩的边界区域。在特征层面，零初始化残差适配器（Z-Adapters）纠正了因畸变引起的特征错位。在决策层面，学习的交叉投影一致性融合（AgreeFusion）仅在多个视图中有一致证据支持时，促进低置信度检测。在WoodScape环视鱼眼基准测试中，FUSEye将YOLO26-x的mAP50从0.148提升至0.266，并保持了84.3%的全面微调准确率。此外，仅随机使用25%的标注训练图像，FUSEye就能达到0.2597 mAP50，保留了97.6%的全标签性能。FUSEye还持续改进了YOLOv8-11检测器，表明该方法与架构无关。源代码将可在https://github.com/Su-wenya/FUSEye获取。
cs.CV / 29 / 2610.02825
TerrainForge: Physics-Grounded road geometry Editing for Counterfactual Autonomous Driving
TerrainForge：基于物理的反事实道路几何编辑用于自主驾驶
Abstract
Road geometry (e.g., crests, sags, and speed humps) and surface conditions (e.g., wet or icy pavement) affect how vehicles move, what drivers and onboard cameras observe, and how much clearance remains between vehicles. Editing these properties in a driving scene therefore requires corresponding changes in vehicle motion. Capturing these differences in a driving video requires a road edit to propagate to vehicle motion, camera viewpoint, and the clearance between vehicles. We present TerrainForge, a framework for generating road geometry-focused counterfactuals from reconstructed multi-vehicle driving episodes. A unified road model connects scene deformation with four-wheel vehicle dynamics, allowing crests, sags, speed humps, and friction changes to propagate through vehicle motion, camera viewpoint, and inter-vehicle clearance. Vehicle dynamics are evaluated against CarSim, and prescribed road geometry is verified in reconstructed Waymo scenes. Across 18 episodes, leaving surrounding vehicles on their recorded trajectories instead of recomputing their responses produces median peak differences in predicted ego-lead distance of 1.52 m for crests and 1.41 m for sags. We further simulate the ego response to 15,758 road edits across 983 braking episodes, pairing each edit with its safety outcomes relative to an unedited replay. These pairs train a first-stage screening surrogate that takes the original driving context and candidate road-edit parameters as input and predicts the resulting change in the ego's terminal gap. On held-out scenes, this prediction achieves 22-40% lower mean absolute error than predicting no change, so candidates can be screened cheaply before the full multi-vehicle rollout.
Chinese Translation
道路几何（例如，隆起、凹陷和减速带）和表面条件（例如，湿滑或结冰的路面）影响车辆的运动、驾驶员和车载摄像头的观察以及车辆之间的间隙。因此，在驾驶场景中编辑这些属性需要相应地改变车辆的运动。捕捉驾驶视频中的这些差异需要对道路进行编辑，以使其影响车辆运动、摄像头视角和车辆间的间隙。我们提出了TerrainForge，一个从重建的多车辆驾驶情节中生成以道路几何为中心的反事实的框架。一个统一的道路模型将场景变形与四轮车辆动力学连接起来，使得隆起、凹陷、减速带和摩擦变化能够通过车辆运动、摄像头视角和车辆间的间隙传播。车辆动力学与CarSim进行评估，规定的道路几何在重建的Waymo场景中得到验证。在18个情节中，保持周围车辆在其记录轨迹上而不是重新计算其响应，导致隆起的预测自我领先距离的中位峰值差为1.52米，凹陷为1.41米。我们进一步模拟了在983个制动情节中对15,758个道路编辑的自我响应，将每个编辑与其相对于未编辑重播的安全结果配对。这些配对训练了一个第一阶段筛选代理，该代理以原始驾驶上下文和候选道路编辑参数作为输入，预测自我的终端间隙的变化。在保留的场景中，这一预测的平均绝对误差比预测没有变化低22-40%，因此可以在全面的多车辆推广之前廉价筛选候选项。
cs.CV / 30 / 2610.02876
Seeing, Saying, but Not Using: From Reportable Spatial Facts to Usable States in Multimodal Large Language Models
看到、说出，但不使用：从可报告的空间事实到多模态大型语言模型中的可用状态
Abstract
A multimodal large language model that correctly reports a spatial fact does not necessarily use that fact in subsequent reasoning. To study this distinction, we introduce \textsc{SpaceConflict}, a benchmark of 23{,}196 inputs for the construction and use of spatial state. Under a unified Supported/Contradictory/Unknown judgment interface, it covers local fact binding (L1), relational composition (L2), cross-observation consistency (L3), and state judgment under transformation (L4). Posing a direct-state query, a full-transformation query, and an explicit-initial-state query on the same world reveals an availability--utilization gap: models recover the initial state from visual evidence yet fail when that state must drive a transformation. For Qwen3.5-9B, 50 of 100 sequences with a correctly recovered initial state fail the full transformation, and supplying the state explicitly repairs all 50; the gap narrows with scale but does not close. We therefore propose Operational State Supervision (OSS), which supervises task-relevant spatial states and their transformation trajectories and aligns shared facts across contexts. OSS improves paired accuracy on matched judgments most on L3 and L4, the levels that depend on organizing and using state. Evaluating multimodal spatial reasoning thus requires asking not only whether a model can see and state a spatial fact, but whether that fact becomes a usable state in subsequent computation.
Chinese Translation
一个能够正确报告空间事实的多模态大型语言模型，并不一定在后续推理中使用该事实。为了研究这种区别，我们引入了 extsc{SpaceConflict}，这是一个包含23,196个输入的基准，用于空间状态的构建和使用。在统一的支持/矛盾/未知判断界面下，它涵盖了局部事实绑定（L1）、关系组合（L2）、交叉观察一致性（L3）和转化下的状态判断（L4）。对同一世界提出直接状态查询、全转化查询和显式初始状态查询，揭示了可用性与利用之间的差距：模型能够从视觉证据中恢复初始状态，但在该状态必须驱动转化时却失败。对于Qwen3.5-9B，在100个序列中，50个具有正确恢复初始状态的序列未能完成全转化，而显式提供状态修复了所有50个；随着规模的扩大，这一差距缩小但并未消失。因此，我们提出了操作状态监督（Operational State Supervision，OSS），它监督与任务相关的空间状态及其转化轨迹，并在不同上下文中对共享事实进行对齐。OSS在匹配判断上提高了配对准确性，尤其是在L3和L4层次，这两个层次依赖于组织和使用状态。因此，评估多模态空间推理不仅需要询问模型是否能够看到和陈述空间事实，还需要询问该事实是否在后续计算中成为可用状态。
cs.CV / 31 / 2610.02880
Found but Not Read: When Extracted Text Closes the Retrieval-Reading Gap in Document Vision-Language Models
发现但未阅读：提取文本如何缩小文档视觉语言模型中的检索-阅读差距
Abstract
Retrieval-augmented document question answering assumes that once the right page is found, a vision-language model (VLM) can read it. We show that this assumption often fails, leaving a retrieval-reading gap: evidence found but not used. A paired protocol isolates this gap by comparing answers from the retrieved page images alone with answers from the same images plus their extracted text. On FoveDoc-Bench, our benchmark with traceable evidence, retrieval finds nearly every evidence page, yet adding CPU-OCR text raises strict accuracy by 13 to 16 points. An exact text layer roughly doubles the gain, which appears across six VLMs from three families and, within one family, narrows with scale without closing. The reader can read this evidence but cannot find it: crops of it recover most of the text gain, boxes around it on the page none. The same protocol identifies two boundaries. Extracted text helps on textual evidence but is neutral or harmful on charts and figures. Its advantage shrinks as retrieval degrades, and unrelated text of the same form adds nothing detectable. Extracted text is an amplifier of retrieval that works, not a substitute for retrieval that does not. Our code is available at https://github.com/atoz03/fovedoc-sup.
Chinese Translation
增强检索的文档问答假设一旦找到正确的页面，视觉语言模型（VLM）就可以读取它。我们展示了这一假设往往失败，导致检索-阅读差距：发现的证据未被使用。通过比较仅从检索到的页面图像与同一图像加上其提取文本的答案，配对协议隔离了这一差距。在我们的基准测试FoveDoc-Bench中，检索几乎找到了每一页证据，但添加CPU-OCR文本使严格准确度提高了13到16个百分点。精确的文本层大约使增益翻倍，这一现象在来自三个家族的六个VLM中均有体现，并且在同一家庭内，随着规模的扩大而缩小但未关闭。读者可以阅读这些证据但无法找到它：其裁剪部分恢复了大部分文本增益，而页面上的框架则没有。相同的协议识别了两个边界。提取文本在文本证据上有帮助，但在图表和图形上则是中立或有害的。随着检索质量的下降，其优势缩小，且同类无关文本没有任何可检测的增益。提取文本是检索的放大器，而非检索的替代品。我们的代码可在 https://github.com/atoz03/fovedoc-sup 获取。
cs.CV / 32 / 2610.02887
Revealing Epistemic Uncertainty in MLLMs via Causal-Invariant Masking
通过因果不变掩蔽揭示多模态大语言模型中的认知不确定性
Abstract
Multimodal Large Language Models (MLLMs) suffer from hallucinations, creating a critical need for Uncertainty Quantification (UQ) to ensure reliable deployment. However, existing approaches struggle to detect uncertainty caused by superficial associations, especially when the query-relevant signal is weak. We mainly attribute this issue to their bias toward aleatoric uncertainty arising from data ambiguity, overlooking epistemic uncertainty stemming from model limitations. To further decompose uncertainty types for a comprehensive UQ, we propose Causal-Invariant Masking (CIM), which measures the semantic shift between the original predictions and those conditioned on a causally-focused view. Based on this framework, we introduce Semantic Divergence as our core metric for UQ and provide theoretical evidence that it converges to the variance of model's sensitivity to non-causal correlations, establishing its ability to capture MLLM's limitation. To accelerate UQ in MLLMs, we further propose Expected Embedding Drift (EED), a fast geometric proxy metric that estimates semantic shift directly within the hyperspherical embedding space. Experiments show that our method achieves state-of-the-art performance on various benchmarks, while the proposed EED accelerates by nearly 50% with comparable performance.
Chinese Translation
多模态大语言模型（MLLMs）存在幻觉现象，这使得不确定性量化（UQ）变得至关重要，以确保其可靠部署。然而，现有方法在检测由表面关联引起的不确定性时面临困难，尤其是在查询相关信号较弱的情况下。我们主要将这一问题归因于它们对由数据模糊性引起的偶然不确定性的偏见，忽视了源于模型局限性的认知不确定性。为了进一步分解不确定性类型以实现全面的不确定性量化，我们提出了因果不变掩蔽（CIM），该方法测量原始预测与基于因果聚焦视角条件下的预测之间的语义偏移。在此框架下，我们引入语义差异作为我们不确定性量化的核心指标，并提供理论证据表明它收敛于模型对非因果相关性的敏感性方差，从而确立其捕捉MLLM局限性的能力。为了加速MLLMs中的不确定性量化，我们进一步提出了期望嵌入漂移（EED），这是一种快速的几何代理指标，能够直接在超球嵌入空间中估计语义偏移。实验表明，我们的方法在各种基准测试中达到了最先进的性能，而所提出的EED在性能相当的情况下加速了近50%。
cs.CV / 33 / 2610.02903
ViTok: Improving Dense Semantics in AM-RADIO-Style Multi-Teacher Distillation with PHI-S and Masked Image Modelling
ViTok：通过 PHI-S 和掩蔽图像建模改善 AM-RADIO 风格的多教师蒸馏中的稠密语义
Abstract
We study how to consolidate the current VITOK progress into a single multi-teacher distillation recipe that jointly preserves global recognition and dense semantics. Our starting point is an AM-RADIO-style student distilled from SigLIP2 and DINOv3-L, where SigLIP2 supplies strong global semantics and DINOv3-L supplies stronger dense features. The central empirical issue is that the same recipe does not optimize all objectives equally well: changes that improve ImageNet-1K kNN accuracy can still degrade ADE20K segmentation. We summarize a progression of modifications that make this trade-off more explicit and more manageable: split adaptor heads for CLS and patch tokens, asymmetric cosine/MSE losses, initialization from a DINOv3-L checkpoint, teacher reweighting, masked image modeling (MIM), and PHI-S feature balancing. The resulting model reaches 83.2 patch kNN and 85.2 CLS kNN, slightly surpassing the DINOv3-L teacher on ImageNet-1K kNN classification, while PHI-S restores ADE20K performance from 46.5/58.1 to 48.5/61.0 mIoU/mAcc, matching the teacher on this dense benchmark. We also summarize negative results: scaling distillation from ImageNet-1K to ImageNet22K does not consistently help, and naively adding extra teachers such as SAM3 or HOG features introduces interference. Rather than claiming a final recipe, this paper distills the current project state into a compact empirical story and a concrete set of lessons for future iterations.
Chinese Translation
我们研究如何将当前的 VITOK 进展整合为一个单一的多教师蒸馏方案，以共同保留全局识别和稠密语义。我们的起点是一个从 SigLIP2 和 DINOv3-L 蒸馏而来的 AM-RADIO 风格学生模型，其中 SigLIP2 提供强大的全局语义，而 DINOv3-L 提供更强的稠密特征。核心的实证问题是相同的方案并不能均衡地优化所有目标：改善 ImageNet-1K kNN 准确率的变化仍可能降低 ADE20K 分割性能。我们总结了一系列修改，使得这种权衡更加明确和可控：为 CLS 和 patch tokens 拆分适配器头、使用不对称余弦/MSE 损失、从 DINOv3-L 检查点初始化、教师重加权、掩蔽图像建模（MIM）和 PHI-S 特征平衡。最终模型在 ImageNet-1K kNN 分类中达到 83.2 的 patch kNN 和 85.2 的 CLS kNN，略微超越了 DINOv3-L 教师，而 PHI-S 将 ADE20K 性能从 46.5/58.1 恢复至 48.5/61.0 mIoU/mAcc，在这一稠密基准上与教师相匹配。我们还总结了一些负面结果：将蒸馏从 ImageNet-1K 扩展到 ImageNet22K 并不总是有帮助，简单地添加额外的教师（如 SAM3 或 HOG 特征）会引入干扰。本文并不声称提供最终的方案，而是将当前项目状态提炼为一个简明的实证故事和一组具体的经验教训，以供未来迭代参考。
cs.CV / 34 / 2610.02914
Custom Forcing: Training-Free Subject Customization for Autoregressive Video Generation
自定义强制：无训练的主体定制用于自回归视频生成
Abstract
Autoregressive video models can generate minute-long videos in real time, but they produce generic subjects from text rather than specific subjects from user-provided images. Existing customization methods either require costly per-subject optimization or use pretrained conditioning networks that jointly process all video frames with bidirectional attention. Neither approach is designed for causal streaming. We present Custom Forcing, a training-free method that stores reference-based anchor frames in the persistent KV cache of a frozen autoregressive video model. However, fixed anchors face two limitations: simple conditioning allows identity to drift, and the text prompt continues to favor a generic subject. To address these problems, drift-adaptive value amplification (DVA) scales reference influence with the degree of identity drift, while anchor contrast guidance (ACG) steers generation away from the generic class prior. Over two-minute rollouts, fixed anchors fall from 0.58 to 0.42 in DINO-I, while Custom Forcing keeps it between 0.58 and 0.62 without reducing motion. Custom Forcing also achieves higher subject similarity than bidirectional customization methods and better preserves identity over 30s than causal image-to-video and reference-to-video models, while generating each frame 9.5--28.5 times faster than these long-video baselines.
Chinese Translation
自回归视频模型能够实时生成长达一分钟的视频，但它们从文本生成通用主体，而不是从用户提供的图像生成特定主体。现有的定制方法要么需要昂贵的逐主体优化，要么使用预训练的条件网络，这些网络通过双向注意力共同处理所有视频帧。两种方法都不适合因果流媒体。我们提出了自定义强制（Custom Forcing），这是一种无训练的方法，它将基于参考的锚帧存储在冻结的自回归视频模型的持久KV缓存中。然而，固定锚点面临两个限制：简单的条件允许身份漂移，而文本提示继续偏向于通用主体。为了解决这些问题，漂移自适应值放大（drift-adaptive value amplification, DVA）根据身份漂移的程度调整参考影响，而锚点对比引导（anchor contrast guidance, ACG）则引导生成远离通用类别先验。在超过两分钟的生成过程中，固定锚点在DINO-I中的得分从0.58降至0.42，而自定义强制保持在0.58到0.62之间，且没有降低运动表现。自定义强制还实现了比双向定制方法更高的主体相似性，并在30秒内比因果图像到视频和参考到视频模型更好地保持身份，同时生成每帧的速度比这些长视频基线快9.5到28.5倍。
cs.CV / 35 / 2610.02943
Kinematics-Induced Multimodal 3D Human Pose Estimation with Subject-Level Privacy
运动学引导的多模态三维人体姿态估计与个体隐私保护
Abstract
Multimodal 3D Human Pose Estimation (3D HPE) combines complementary information from RGB, LiDAR, and mmWave radar, but models trained on correlated observations from the same individuals, raise privacy risks overlooked by record level analysis. We present a unified framework for multimodal 3D HPE that couples kinematics-induced sensor fusion with subject level privacy auditing and private training. First, our multimodal model aligns modality specific joint representation, injects skeletal structure and adaptively aggregates complementary sensor evidence for accurate pose prediction. Second, we formulate a black-box subject membership inference attack for 3D HPE, complemented by an empirical pointwise maximal leakage analysis, which characterizes how individual attack score outcomes change inference about the membership outcome. Third, we instantiate user-level differential privacy via Action Temporal Stratification, a population weighted within-subject sampling strategy that enforces action and temporal coverage. We evaluate our framework on the MM-Fi dataset across three diverse experimental protocols. Source-code will be released upon acceptance.
Chinese Translation
多模态三维人体姿态估计（3D HPE）结合了来自RGB、LiDAR和毫米波雷达的互补信息，但在同一个体的相关观察上训练的模型存在被记录级分析所忽视的隐私风险。我们提出了一个统一的多模态3D HPE框架，该框架将运动学引导的传感器融合与个体级隐私审计和私密训练相结合。首先，我们的多模态模型对齐了模态特定的关节表示，注入了骨骼结构，并自适应地聚合互补传感器证据，以实现准确的姿态预测。其次，我们为3D HPE制定了一个黑箱个体成员推断攻击，并辅以经验性的点对点最大泄漏分析，描述了个体攻击得分结果如何改变对成员结果的推断。第三，我们通过动作时间分层（Action Temporal Stratification）实例化用户级差分隐私，这是一种强制执行动作和时间覆盖的人群加权个体内采样策略。我们在MM-Fi数据集上评估了我们的框架，涵盖了三种不同的实验协议。源代码将在接受后发布。
cs.CV / 36 / 2610.02946
When Predicting Nothing Beats SAM 3: Revisiting Evaluation in Video Object Segmentation
当预测无物时，SAM 3 无可匹敌：视频目标分割评估的再审视
Abstract
Video Object Segmentation (VOS) in complex and long videos is increasingly important for real-world applications, where target objects often appear only intermittently within long temporal horizons. However, existing benchmarks largely focus on temporally salient objects that remain visible for most of the video. To address this gap, we introduce FaVOS (A Benchmark for Video Object Segmentation with Fractional Temporal Visibility), a benchmark designed to evaluate VOS methods under low temporal visibility. We show that, in this regime, the standard J&F metric can collapse VOS evaluation into absence classification, because empty predictions receive high rewards on target-absent frames. Consequently, even a trivial empty-mask predictor can outperform strong models such as SAM 3, revealing a fundamental mismatch between current metrics and practical VOS performance. To mitigate this issue, we propose Volumetric J&F, which evaluates mask sequences as spatio-temporal volumes and reduces the dominance of target-absence rewards while preserving sensitivity to segmentation quality and temporal structure. Project page: https://aidaslab.github.io/FaVOS.
Chinese Translation
在复杂且较长的视频中，视频目标分割（VOS）对于现实世界应用越来越重要，因为目标物体往往仅在较长时间范围内间歇性出现。然而，现有的基准测试主要集中在大多数时间内可见的时间显著物体上。为了解决这一问题，我们引入了 FaVOS（具有分数时间可见性的视频目标分割基准），该基准旨在评估在低时间可见性下的 VOS 方法。我们展示了在这种情况下，标准的 J&F 指标可能会将 VOS 评估简化为缺失分类，因为在目标缺失帧上，空预测会获得高奖励。因此，即使是一个微不足道的空掩膜预测器也能超越强大的模型，如 SAM 3，这揭示了当前指标与实际 VOS 性能之间的根本不匹配。为了解决这个问题，我们提出了体积 J&F，它将掩膜序列评估为时空体积，并减少目标缺失奖励的主导地位，同时保持对分割质量和时间结构的敏感性。项目页面：https://aidaslab.github.io/FaVOS。
cs.CV / 37 / 2610.02959
TerraVis: Towards Evaluation of World-Grounded Visual Consistency in Text-to-Image Generation via MLLM Workflows
TerraVis：通过 MLLM 工作流评估文本到图像生成中的世界基础视觉一致性
Abstract
Recent text-to-image models have made substantial progress in photorealism, aesthetics, and text-image alignment. Yet visually appealing images can still violate real-world plausibility, exhibiting malformed object structures, impossible anatomy, physically implausible interactions, or inconsistent spatial relationships. Such failures are not well captured by existing fidelity, aesthetics, preference, or alignment metrics. To address this gap, we introduce TerraVis, a framework for evaluating world-grounded visual consistency in generated images. TerraVis defines a structured taxonomy of world-consistency violations spanning object-, interaction-, and scene-level failures, and employs a multi-stage evaluation framework to identify and quantify them. Given an image, TerraVis first uses an MLLM to assess its eligibility for evaluation, then detects violations across 18 taxonomy-defined types and classifies them as minor or major to derive an overall world-consistency score. Across diverse open-source and proprietary text-to-image models on two widely used benchmarks, TerraVis achieves the strongest correlation with human judgments of world consistency among existing metrics. Our benchmark results further show that models that achieve strong performance on conventional metrics can still exhibit substantial world-consistency failures. These findings highlight world consistency as a complementary evaluation dimension and demonstrate that TerraVis enables systematic quantification, diagnosis, and comparison of such failures. Our code is publicly available at https://github.com/ShyFoo/TerraVis.
Chinese Translation
近期的文本到图像模型在照片真实感、美学和文本-图像对齐方面取得了显著进展。然而，视觉上吸引人的图像仍可能违反现实世界的合理性，表现出畸形的物体结构、不可能的解剖结构、物理上不合理的交互或不一致的空间关系。这些失败并未被现有的保真度、美学、偏好或对齐指标很好地捕捉。为了解决这一问题，我们提出了 TerraVis，一个用于评估生成图像中世界基础视觉一致性的框架。TerraVis 定义了一个结构化的世界一致性违规分类法，涵盖了物体、交互和场景层面的失败，并采用多阶段评估框架来识别和量化这些违规情况。给定一幅图像，TerraVis 首先使用 MLLM 评估其评估资格，然后检测 18 种分类法定义的违规类型，并将其分类为轻微或重大，以得出整体世界一致性评分。在两个广泛使用的基准上，针对多种开源和专有的文本到图像模型，TerraVis 与现有指标中人类对世界一致性判断的相关性最强。我们的基准结果进一步表明，在传统指标上表现良好的模型仍可能表现出显著的世界一致性失败。这些发现突显了世界一致性作为一个补充评估维度的重要性，并证明了 TerraVis 能够系统地量化、诊断和比较这些失败。我们的代码已公开发布在 https://github.com/ShyFoo/TerraVis。
cs.CV / 38 / 2610.02967
Post-Training Frontier Text-to-Image Models by Composing Preference and Rubric Rewards
通过组合偏好和评分奖励对后训练前沿文本到图像模型进行优化
Abstract
Recent text-to-image generation models have achieved remarkable visual quality, but improving them through post-training remains challenging because no single reward signal captures the full range of human preference. In this work, we develop a simple and effective post-training recipe for open-domain text-to-image generation based on the composition of complementary reward signals. Our reward system consists of two main components: a preference reward, trained on large-scale human preference data using a Bradley-Terry objective to capture overall human aesthetic and perceptual preferences, and rubric-based rewards, which explicitly evaluate prompt faithfulness and other desirable properties while providing safeguards against reward hacking. A key challenge is how to combine these heterogeneous reward signals. We show that a naive weighted average leads to suboptimal optimization behavior, and propose a simple reward composition strategy that more effectively balances preference optimization with rubric satisfaction. In the Arena text-to-image leaderboard (https://arena.ai/), our RL-trained Flux2dev achieves an Elo rating 69 points above the base model, and our post-trained Ideogram-4 surpasses every open-source model on the leaderboard, reaching an Elo of 1223.5. (Claims of state-of-the-art performance are based on the Arena leaderboard snapshot as of September 4, 2026.) Our results suggest that effective rewards for frontier generative-model training require broad coverage of user intent and robustness to exploitation under optimization. To support reproducible research, we release Arena-T2I-Training, a 1K subset of training data that recovers some gains of full-scale training, providing a resource that we hope will facilitate future work on post-training for text-to-image models.
Chinese Translation
近年来，文本到图像生成模型在视觉质量上取得了显著进展，但通过后训练来进一步提升这些模型仍然面临挑战，因为没有单一的奖励信号能够全面捕捉人类偏好的多样性。在本研究中，我们开发了一种简单而有效的后训练方案，用于开放领域的文本到图像生成，基于互补奖励信号的组合。我们的奖励系统由两个主要组成部分构成：偏好奖励，该奖励基于大规模人类偏好数据使用Bradley-Terry目标进行训练，以捕捉整体人类的审美和感知偏好；以及基于评分的奖励，这些奖励明确评估提示的忠实性和其他期望属性，同时提供防止奖励操控的保障。一个关键挑战是如何结合这些异质奖励信号。我们发现，简单的加权平均会导致次优的优化行为，因此提出了一种简单的奖励组合策略，更有效地平衡了偏好优化与评分满意度。在Arena文本到图像排行榜（https://arena.ai/）上，我们经过强化学习训练的Flux2dev的Elo评分比基础模型高出69分，而我们后训练的Ideogram-4在排行榜上超越了所有开源模型，达到了1223.5的Elo评分。（关于最先进性能的声明基于截至2026年9月4日的Arena排行榜快照。）我们的结果表明，前沿生成模型训练的有效奖励需要广泛覆盖用户意图，并在优化过程中具有抗利用性。为了支持可重复的研究，我们发布了Arena-T2I-Training，这是一个包含1000个训练数据子集的资源，能够恢复部分全规模训练的收益，我们希望这能促进未来在文本到图像模型后训练方面的研究。
cs.CV / 39 / 2610.03012
Rethinking Fixed Temporal Grids: Frequency-Disentangled Motion Generation
重新思考固定时间网格：频率解耦的运动生成
Abstract
Most human motion generation methods encode motion as tokens on a uniform temporal grid, where every token spans the same fixed time window. Human motion, however, is temporally heterogeneous: slowly evolving global trajectories coexist with rapid transient events such as foot contacts and joint impulses. Forcing such multi-scale dynamics onto tokens of identical temporal resolution entangles motion frequencies, leaving slow regions redundant while smoothing out the rapid details that distinguish realistic motion. We propose \textbf{FreqMo}, a scale-adaptive motion representation that decomposes motion into wavelet frequency bands, separating dynamics across temporal scales while preserving temporal localization and exact reconstruction. Unified Frequency Residual Quantization (UFRQ) then encodes all bands within a single shared codebook, compressing the token sequence threefold and enabling stable single-stage generation. Experiments show FreqMo attains SOTA fidelity with substantially improved high-frequency preservation, and the same decomposition transfers to continuous diffusion backbones.
Chinese Translation
大多数人类运动生成方法将运动编码为均匀时间网格上的标记，其中每个标记跨越相同的固定时间窗口。然而，人类运动在时间上是异质的：缓慢演变的全局轨迹与快速瞬态事件（如脚接触和关节冲击）共存。将这种多尺度动态强加于具有相同时间分辨率的标记上，会导致运动频率的纠缠，使得缓慢区域冗余，同时平滑掉区分真实运动的快速细节。我们提出了 extbf{FreqMo}，一种尺度自适应的运动表示方法，它将运动分解为小波频率带，分离不同时间尺度的动态，同时保持时间定位和精确重建。统一频率残差量化（Unified Frequency Residual Quantization, UFRQ）随后在一个共享的代码本中编码所有频带，将标记序列压缩三倍，并实现稳定的单阶段生成。实验表明，FreqMo 达到最先进的保真度，并显著改善高频保留，且相同的分解可以迁移到连续扩散骨干网络上。
cs.CV / 40 / 2610.03013
RYOPO: Bringing End-to-End Category-Level Object Pose Estimation into Real Time
RYOPO：将端到端类别级物体姿态估计带入实时应用
Abstract
Category-level object pose estimation predicts the rotation, translation, and metric size of unseen instances within known categories. Many accurate RGB-D methods rely on external instance segmentation and crop-based pose estimation, introducing separate stages and object-dependent processing costs that hinder real-time inference. To bring accurate pose estimation into real time, we present \ours{}, an end-to-end trainable query-based RGB-D set predictor. It jointly detects and segments objects and estimates their \mbox{9-DoF} poses without explicit CAD-derived shape priors or a separately trained instance segmentor. Shared image and scene encoding avoids repeated per-object crop encoding. A query-conditioned geometry pathway associates observed 3D points and RGB features with object queries and incorporates shared scene context. Object-centric refinement uses the resulting point descriptors to update an explicit pose state through pose-conditioned cross-attention and recurrent residual corrections. On NOCS, \ours{} substantially improves on published RGB-D joint detection and pose estimation results. It achieves competitive performance compared with two-stage methods under all-object evaluation on REAL275 and HouseCat6D, while enabling real-time full-frame pose estimation at $31.8$ FPS on an RTX~A6000. Project page: https://yopo-series.github.io/RYOPO-project-page/.
Chinese Translation
类别级物体姿态估计预测已知类别中未见实例的旋转、平移和度量大小。许多准确的RGB-D方法依赖于外部实例分割和基于裁剪的姿态估计，这引入了独立的阶段和依赖于物体的处理成本，阻碍了实时推断。为了将准确的姿态估计引入实时应用，我们提出了 extit{RYOPO}，一种端到端可训练的基于查询的RGB-D集合预测器。它联合检测和分割物体，并在没有显式CAD衍生形状先验或单独训练的实例分割器的情况下估计其9自由度（9-DoF）姿态。共享的图像和场景编码避免了对每个物体重复进行裁剪编码。查询条件的几何路径将观察到的3D点和RGB特征与物体查询关联，并结合共享的场景上下文。以物体为中心的细化使用结果点描述符通过姿态条件的交叉注意力和递归残差修正来更新显式姿态状态。在NOCS数据集上， extit{RYOPO}显著改善了已发布的RGB-D联合检测和姿态估计结果。在REAL275和HouseCat6D的全物体评估中，它的性能与两阶段方法相当，同时在RTX A6000上实现了每秒31.8帧的实时全帧姿态估计。项目页面：https://yopo-series.github.io/RYOPO-project-page/
cs.CV / 41 / 2610.03015
OmniAct3D: Leveraging Foundation Geometry and Evidence-Grounded Reasoning for Panoramic 3D Detection
OmniAct3D：利用基础几何和基于证据的推理进行全景3D检测
Abstract
Accurate 3D detection is essential for mobile embodied agents, while Vision Foundation Models (VFMs) offer transferable visual and geometric priors. Yet existing VFM-based 3D detectors rely on narrow-view monocular images or discrete perspective views, limiting coherent surround perception; equirectangular projection (ERP) instead encodes a continuous 360 scene in a single image. Direct transfer remains difficult because ERP organizes geometry and visual information differently, making object-relevant cues hard to model, localize, and preserve. We propose OmniAct3D, a framework that adapts perspective-trained VFM detectors to ERP while preserving transferable VFM priors. To resolve geometric mismatch, the ERP-Ray Geometry Adapter (ERGA-Ray) models spherical viewing rays and periodic spatial structure. To localize evidence in scene-wide context, the Visual-Action Reasoning Chain (VARC) grounds each hypothesis in relevant panoramic evidence and converts it into a structured geometric action. To recover local cues lost under fixed token budgets, the Appearance-Guided Heading Expert (AGHE) re-encodes object regions at higher resolution for heading estimation. Experiments show that OmniAct3D improves over the previous best 3D detector by 2.96 NDS points on Spheriverse and over the unadapted VFM baseline by 24.87 mAP points on PanoMMOcc. With target-specific geometry adaptation, VARC retains 95--98% of the same-configuration mAP, indicating reusable object-level 3D reasoning across sensing configurations. The source code will be made publicly available at https://github.com/FeiT-FeiTeng/OmniAct3D.
Chinese Translation
准确的3D检测对于移动具身智能体至关重要，而视觉基础模型（Vision Foundation Models, VFMs）提供了可转移的视觉和几何先验。然而，现有基于VFM的3D检测器依赖于狭窄视角的单目图像或离散的透视视图，限制了连贯的周围感知；而等矩形投影（equirectangular projection, ERP）则可以在单幅图像中编码连续的360度场景。直接转移仍然困难，因为ERP以不同的方式组织几何和视觉信息，使得与物体相关的线索难以建模、定位和保留。我们提出了OmniAct3D，一个将透视训练的VFM检测器适应于ERP的框架，同时保留可转移的VFM先验。为了解决几何不匹配问题，ERP-Ray Geometry Adapter（ERGA-Ray）建模球形视线和周期性空间结构。为了在场景广泛上下文中定位证据，Visual-Action Reasoning Chain（VARC）将每个假设基于相关的全景证据进行扎根，并将其转化为结构化的几何动作。为了恢复在固定令牌预算下丢失的局部线索，Appearance-Guided Heading Expert（AGHE）以更高的分辨率重新编码物体区域以进行航向估计。实验表明，OmniAct3D在Spheriverse上比之前最佳的3D检测器提高了2.96 NDS点，在PanoMMOcc上比未适应的VFM基线提高了24.87 mAP点。通过特定目标的几何适应，VARC保留了95%至98%的同配置mAP，表明在传感配置中可重用的物体级3D推理。源代码将公开发布在https://github.com/FeiT-FeiTeng/OmniAct3D。
cs.CV / 42 / 2610.03022
ReSCUE: Re-translation with Sentence Commitment for Unsegmented Long-Form Simultaneous Sign Language Translation
ReSCUE：用于未分段长篇即时手语翻译的句子承诺重翻译
Abstract
Simultaneous Sign Language Translation (SLT) is critical for real-time communication, yet existing methods remain largely confined to sentence-level, offline settings that assume pre-segmented inputs. These assumptions hinder deployment in realistic scenarios involving continuous, unsegmented video streams. We present ReSCUE, a unified framework for simultaneous SLT on unsegmented long-form sign language videos that aligns training and inference with realistic streaming conditions. ReSCUE combines inference-aware training to handle partial inputs, non-signing pauses, and multi-sentence contexts, stabilized re-translation to enable low-latency yet revisable predictions with reduced output flicker, and a sentence commitment mechanism for online segmentation and memory management. Experiments on standard sentence-level benchmarks show that ReSCUE achieves lower latency and the best translation quality under low-latency settings. On long-form unsegmented datasets, ReSCUE approaches the translation quality of oracle offline systems that use ground-truth sentence boundaries, while operating at substantially lower latency, demonstrating its practicality for real-world streaming scenarios.
Chinese Translation
即时手语翻译（SLT）对于实时沟通至关重要，但现有方法仍主要局限于假设预先分段输入的句子级离线设置。这些假设阻碍了在涉及连续未分段视频流的现实场景中的应用。我们提出了ReSCUE，这是一个统一的框架，用于在未分段的长篇手语视频上进行即时SLT，使训练和推理与现实流媒体条件相一致。ReSCUE结合了推理感知训练，以处理部分输入、非手语暂停和多句子上下文，稳定的重翻译以实现低延迟但可修订的预测，并采用句子承诺机制进行在线分段和记忆管理。在标准句子级基准测试中的实验表明，ReSCUE在低延迟设置下实现了更低的延迟和最佳的翻译质量。在未分段的长篇数据集上，ReSCUE的翻译质量接近使用真实句子边界的理想离线系统，同时以显著更低的延迟运行，展示了其在现实流媒体场景中的实用性。
cs.CV / 43 / 2610.03031
CrowdOcc: Monocular Semantic Scene Completion for Quadruped Robots in Crowded Indoor Environments
CrowdOcc：在拥挤室内环境中为四足机器人提供单目语义场景补全
Abstract
Monocular semantic scene completion (SSC) for quadruped robots remains underexplored in real crowded indoor environments, where human-scene occlusion disrupts static geometry and human occupancy predictions are often incomplete or spatially misplaced. We present CrowdOcc, an RGB-D dataset and monocular SSC framework for this setting. CrowdOcc contains 25.1K frames from 11 indoor scenes, with semantic occupancy annotations constructed through static dynamic decoupling. Our framework combines: (i) Normal Guided Scene Geometry Fusion (NGSGF) to complement depth-aware lifting with surface-normal cues for occlusion robust geometry; and (ii) Human-Centric Sparse Interaction (HCSI) to selectively model human-human and local human scene relations in 3D. Our method achieves state-of-the-art SSC performance on CrowdOcc's scene-disjoint test set, reaching 15.80 IoU, 11.40 mIoU, and 46.23 Human IoU, demonstrating generalization to unseen indoor scenes.
Chinese Translation
在真实的拥挤室内环境中，四足机器人的单目语义场景补全（SSC）仍然未得到充分探索，其中人类与场景的遮挡会干扰静态几何形状，而人类占用预测往往不完整或空间位置错误。我们提出了CrowdOcc，一个针对该环境的RGB-D数据集和单目SSC框架。CrowdOcc包含来自11个室内场景的25.1K帧，语义占用注释通过静态动态解耦构建。我们的框架结合了：(i) 正常引导场景几何融合（NGSGF），利用表面法线线索补充深度感知提升，以实现对遮挡的鲁棒几何；以及(ii) 以人为中心的稀疏交互（HCSI），选择性地建模人类与人类以及局部人类与场景的三维关系。我们的方法在CrowdOcc的场景不重叠测试集上实现了最先进的SSC性能，达到15.80的IoU、11.40的mIoU和46.23的人类IoU，展示了对未见室内场景的泛化能力。
cs.CV / 44 / 2610.03047
Parasitic Co-Denoising: Unlocking 3D Human Motion Generation in a Frozen Video Diffusion Model
寄生共去噪：解锁冻结视频扩散模型中的3D人类运动生成
Abstract
Despite never being supervised on explicit 3D motion, large-scale text-to-video diffusion models synthesize realistic human motion in their generated videos. We ask whether this implicit knowledge can be turned into explicit 3D motion generation, without training a separate motion model. Probing a frozen Wan2.1 reveals that a recoverable motion signal is present in its intermediate states across the entire denoising schedule, not confined to the clean output. Motivated by this, we introduce parasitic co-denoising, a paradigm in which motion is decoded from the host model along its denoising schedule rather than produced by an independent generator. We instantiate it as the Parasitic Motion Decoder (PMD), an efficient flow-matching decoder that shares the host's noise schedule and reads its intermediate features through a $\sigma$-adaptive multi-layer fusion, leaving the host unmodified. Drawing its coverage from the host rather than from motion data, PMD leads dedicated motion generators on text-motion alignment at a small fraction of their trainable parameters, while producing paired video and motion in a single pass that motion-only baselines cannot match.
Chinese Translation
尽管从未在显式的3D运动上进行监督，大规模的文本到视频扩散模型仍能在生成的视频中合成逼真的人类运动。我们探讨这种隐式知识是否可以转化为显式的3D运动生成，而无需训练一个独立的运动模型。对冻结的Wan2.1模型的探查显示，在其整个去噪过程中的中间状态中存在可恢复的运动信号，而不仅限于干净的输出。基于此，我们引入了寄生共去噪（parasitic co-denoising）这一范式，其中运动是从主模型的去噪过程中解码，而不是由独立的生成器产生。我们将其实例化为寄生运动解码器（Parasitic Motion Decoder, PMD），这是一种高效的流匹配解码器，能够共享主模型的噪声调度，并通过$ au$-自适应多层融合读取其中间特征，而不对主模型进行修改。PMD从主模型而非运动数据中获取覆盖，能够以极少的可训练参数引导专用运动生成器进行文本-运动对齐，同时在单次传递中生成配对的视频和运动，这是仅依赖运动的基线模型无法比拟的。
cs.CV / 45 / 2610.03051
BeeWhere: Segmenting Bumble Bee Colonies to Quantify Behavioral Effects
BeeWhere：分割大黄蜂群体以量化行为影响
Abstract
Social bees are important pollinators that support biodiversity and crop pollination globally and serve as important model systems for collective behavior, but scalable measurement of individual- and colony-level behavior remains difficult in dense, occluded nest environments. Existing monitoring workflows use fiducial tags (e.g., ArUco) to preserve individual identity, yet tag-based tracking can fail when markers are obscured and provide limited information about body extent, spatial context, and untagged individuals. We present BeeWhere, an AI-assisted annotation and analysis workflow that combines ArUco detections with deep-learnt instance segmentations to quantify bumble bee behavior from high-resolution colony images and videos. Using bumble bee (Bombus impatiens) microcolonies as a test case, we annotate 483 frames containing 8,443 bee instances. We additionally annotate pollen balls, nest structures, and chamber boundaries, and train YOLO instance segmentation models for downstream behavioral analysis. Instance segmentations enable quantification of important behavioral metrics based on body contours, including nearest-neighbor distance, proximity to nest structures, spatial occupancy within the nest, and detection counts over time. We apply the BeeWhere models to tag-based tracking in an exploratory validation study assessing the behavioral impacts of neonicotinoid pesticide exposure. BeeWhere increased detection rates compared to tag-based tracking, particularly when bees were partially obscured or under challenging imaging conditions, and also captured treatment-associated changes in bee spatial organization not captured using tag-based tracking alone. These results suggest that instance segmentation can complement fiducial-marker tracking by recovering behaviorally meaningful signals under challenging colony conditions.
Chinese Translation
社会性蜜蜂是重要的授粉者，支持全球生物多样性和作物授粉，并作为集体行为的重要模型系统，但在密集的、被遮挡的巢环境中，个体和群体行为的可扩展测量仍然困难。现有的监测工作流程使用基准标签（如 ArUco）来保持个体身份，但当标记被遮挡时，基于标签的跟踪可能会失败，并且提供有关身体范围、空间背景和未标记个体的有限信息。我们提出了 BeeWhere，这是一种 AI 辅助的注释和分析工作流程，结合了 ArUco 检测与深度学习实例分割，以从高分辨率的群体图像和视频中量化大黄蜂行为。以大黄蜂（Bombus impatiens）微群体为测试案例，我们注释了包含 8,443 个蜜蜂实例的 483 帧。此外，我们还注释了花粉球、巢结构和巢室边界，并训练了 YOLO 实例分割模型以进行下游行为分析。实例分割使得基于身体轮廓量化重要的行为指标成为可能，包括最近邻距离、与巢结构的接近度、巢内的空间占用以及随时间变化的检测计数。我们在一项探索性验证研究中应用 BeeWhere 模型进行基于标签的跟踪，以评估新烟碱类农药暴露的行为影响。与基于标签的跟踪相比，BeeWhere 提高了检测率，特别是在蜜蜂部分被遮挡或在具有挑战性的成像条件下，并且还捕捉到了基于标签的跟踪无法单独捕捉的蜜蜂空间组织的处理相关变化。这些结果表明，实例分割可以通过在具有挑战性的群体条件下恢复行为上有意义的信号，从而补充基于基准标记的跟踪。
cs.CV / 46 / 2610.03068
Where to Look Is Not How to Fix: Pre-Denoising Diagnostics and Modality-Dependent Control in Diffusion Composition
观察的重点不是修复的方法：扩散组合中的预去噪诊断与模态依赖控制
Abstract
Understanding compositional failures in text-to-image diffusion requires identifying both where stress is detectable and how intervention changes the output. We study these questions through a controlled anchor--stress protocol that jointly evaluates text-encoder diagnostics and denoiser interventions. We introduce a text-only Compositional Stress Index (CSI), which separates common from rare compositions across SD1.5, SDXL, and the SD3 text path and provides an upstream diagnostic coordinate. A matched six-prompt localization study links intervention location to distinct outcomes: residual-minimizing embedding adapters improve representation fit, while downstream cross-attention intervention increases color hit rate (CHR) by 0.0272. Across SD1.5 and SDXL denoiser blocks, the largest positive signed diagnostic-accessibility mean occurs at the deep encoder, whereas selective boost has its largest positive mean CHR response at decoder blocks. Selective subtraction and broad ablation reveal further modality- and architecture-dependent responses, including a substantial CHR decrease when SDXL decoder cross-attention is broadly ablated. We find a diagnosis-control dissociation under our controlled attribute-object composition setting: compositional defects are diagnosable before denoising, but the representation coordinate that exposes risk is not necessarily the coordinate or modality that improves generation.
Chinese Translation
理解文本到图像扩散中的组合失效需要识别可检测到压力的位置以及干预如何改变输出。我们通过一种受控的锚点-压力协议研究这些问题，该协议联合评估文本编码器诊断和去噪器干预。我们引入了一种仅基于文本的组合压力指数（Compositional Stress Index, CSI），该指数在SD1.5、SDXL和SD3文本路径中区分常见与稀有组合，并提供了一个上游诊断坐标。一项匹配的六提示定位研究将干预位置与不同结果联系起来：最小化残差的嵌入适配器改善了表示拟合，而下游交叉注意力干预将颜色命中率（Color Hit Rate, CHR）提高了0.0272。在SD1.5和SDXL去噪器模块中，最大的正签名诊断可达性均值出现在深层编码器，而选择性增强在解码器模块中具有最大的正均值CHR响应。选择性减法和广泛消融揭示了进一步的模态和架构依赖响应，包括当SDXL解码器的交叉注意力被广泛消融时CHR显著下降。我们发现，在我们的受控属性-对象组合设置下，诊断与控制之间存在解离：组合缺陷在去噪之前是可诊断的，但暴露风险的表示坐标不一定是改善生成的坐标或模态。
cs.CV / 47 / 2610.03084
NegT2IBench: When Negation Changes the Picture. A Polarity Benchmark for Text-to-Image Models
NegT2IBench：当否定改变画面时。文本到图像模型的极性基准
Abstract
Text-to-image (T2I) models are judged by benchmarks that measure whether requested content appears, but these benchmarks largely overlook the complementary ability to satisfy negated constraints, for example, generating "a non-red cup." Measuring negation raises challenges not faced by affirmation-based benchmarks and requires careful prompt and evaluation design. We introduce NegT2IBench, a benchmark of 4,800 prompts covering two attribute types and four relation categories. Prompts are organized by polarity: the number of positive statements that must hold and negated statements that must not, each ranging from 0 to 2. Varying the two independently separates the effect of negation from the effect of prompt complexity. Our detector-based scoring is reproducible, auditable, and pinpoints which requirement failed. On 600 images with three-annotator labels, it agrees with humans as closely as vision-language judges up to 30x larger, while using only a fraction of their GPU memory. Across eleven T2I models and 211,200 images, nine score lower on a single negated statement than on a single positive one. Per-statement scoring reveals that the loss is largest for color and near zero for proximity, and that 41.5% of failed statements render exactly what the prompt forbids. Rendering what a prompt asks for and withholding what it forbids are distinct capabilities that an aggregate compositional score cannot distinguish. NegT2IBench measures the latter directly, providing a controlled testbed for diagnosing negation failures and developing methods to overcome them.
Chinese Translation
文本到图像（T2I）模型的评估通常依赖于测量请求内容是否出现的基准，但这些基准在很大程度上忽视了满足否定约束的互补能力，例如生成“一个非红色的杯子”。测量否定带来了与基于肯定的基准不同的挑战，并需要仔细的提示和评估设计。我们引入了NegT2IBench，这是一个包含4800个提示的基准，涵盖两种属性类型和四个关系类别。提示按极性组织：必须满足的正面陈述数量和不得满足的否定陈述数量，均在0到2之间变化。独立变化这两个变量可以将否定的影响与提示复杂性的影响分开。我们的基于检测器的评分是可重复的、可审计的，并且能够准确指出哪个要求未能满足。在600幅带有三位注释者标签的图像上，它与人类的评分接近，甚至与规模大30倍的视觉-语言评审者一致，同时仅使用其GPU内存的一小部分。在十一种T2I模型和211,200幅图像中，九种在单个否定陈述上的得分低于在单个肯定陈述上的得分。逐条陈述的评分显示，颜色的损失最大，而接近度的损失接近于零，并且41.5%的失败陈述恰好呈现出提示所禁止的内容。满足提示要求和抑制提示禁止内容是两种不同的能力，聚合的组合评分无法区分。NegT2IBench直接测量后者，为诊断否定失败和开发克服方法提供了一个受控的测试平台。
cs.CV / 48 / 2610.03099
Beyond Single Videos: Benchmarking and Active Evidence Seeking for E-Commerce Cross-Video Reasoning
超越单一视频：电子商务跨视频推理的基准测试与主动证据寻求
Abstract
E-commerce videos are information-dense and frequently compared by consumers evaluating products and merchants assessing marketing strategies. However, existing multimodal models mainly focus on single-video understanding and have limited ability to compare information across videos. We introduce AdsCVR, the first e-commerce cross-video reasoning benchmark, containing 2,483 videos and 6,110 question-answer pairs across six reasoning dimensions. Cross- video reasoning requires models to locate fine-grained evidence among many redundant frames and integrate visual details, speech, and on-screen text. We therefore propose AdSeek, an agentic framework that dynamically selects visual and audio tools during multi-turn exploration, replacing static uniform sampling with active evidence acquisition. To address the sparse credit assignment of reinforcement learning, we develop an offline trajectory rectification mechanism that identifies reasoning errors and missing multimodal evidence in RL-generated trajectories. The corrected trajectories provide supervised fine-tuning signals that reduce biases learned during RL. This mechanism supports a rectified bootstrapping pipeline in which initial RL exposes reasoning bottlenecks, supervised fine-tuning corrects them, and a final RL stage further improves the policy. AdSeek achieves 74.30 percent accuracy on the AdsCVR test split, outperforming its Qwen3-VL-8B-Instruct backbone by 27.90 percentage points. It also generalizes to the open- domain CrossVid benchmark, demonstrating effective active evidence gathering.
Chinese Translation
电子商务视频信息密集，消费者在评估产品时常常进行比较，而商家则在评估营销策略时进行比较。然而，现有的多模态模型主要集中于单视频理解，跨视频信息比较的能力有限。我们引入了 AdsCVR，这是第一个电子商务跨视频推理基准，包含 2,483 个视频和 6,110 对问题-答案，涵盖六个推理维度。跨视频推理要求模型在众多冗余帧中定位细粒度证据，并整合视觉细节、语音和屏幕文本。因此，我们提出了 AdSeek，这是一个动态选择视觉和音频工具的智能框架，在多轮探索中替代静态均匀采样，采用主动证据获取。为了解决强化学习中的稀疏信用分配问题，我们开发了一种离线轨迹校正机制，识别推理错误和缺失的多模态证据在强化学习生成的轨迹中。校正后的轨迹提供监督微调信号，减少在强化学习中学习到的偏差。该机制支持一个校正的自举管道，其中初始的强化学习暴露推理瓶颈，监督微调进行校正，最终的强化学习阶段进一步改善策略。AdSeek 在 AdsCVR 测试集上达到了 74.30% 的准确率，比其 Qwen3-VL-8B-Instruct 主干模型高出 27.90 个百分点。它还能够推广到开放领域的 CrossVid 基准，展示了有效的主动证据收集能力。
cs.CV / 49 / 2610.03105
A Benchmark for Spatially Grounded Gesture Generation
空间基础手势生成的基准测试
Abstract
Communication in shared space interweaves verbal and non-verbal signals, and pointing gestures anchor language to the environment: "put the cup on that one" is uninterpretable without the gesture that fixes the referent. Yet no common framework exists for evaluating whether generated gestures indicate their intended referent; distributional metrics reward a gesture aimed at the wrong object as long as it looks natural. We introduce a benchmark for spatially grounded gesture generation, comprising ~2K pointing-annotated clips from naturalistic VR dialogue with ground-truth 3D referents, a task in which systems must decide when, how and where to point within conversational speech, and a protocol that separates temporal alignment, spatial grounding and perceived naturalness. We also provide a flow-matching baseline, MM-Conv-Flow. Evaluating it alongside an independent retrieval-based system and captured human motion, we find that geometric grounding can exceed that of human pointing without any gain in perceived naturalness, showing that referential gesture quality must be measured along separate dimensions.
Chinese Translation
共享空间中的交流交织着语言和非语言信号，而指向手势将语言锚定于环境中：“把杯子放在那个上面”在没有指向所指对象的手势时是无法理解的。然而，目前尚不存在一个通用框架来评估生成的手势是否指向其预期的所指对象；分布式度量会奖励指向错误对象的手势，只要它看起来自然。我们提出了一个用于空间基础手势生成的基准测试，包含约2000个来自自然对话的指向注释片段，具有真实的3D所指对象。在这个任务中，系统必须决定在对话中何时、如何以及在哪里指向，并且我们提供了一个将时间对齐、空间基础和感知自然性分开的评估协议。我们还提供了一个流匹配基线，MM-Conv-Flow。在与一个独立的基于检索的系统和捕获的人类运动进行评估时，我们发现几何基础的手势可以超越人类指向，而在感知自然性上没有任何提升，这表明所指手势的质量必须沿着不同的维度进行测量。
cs.CV / 50 / 2610.03120
In-Distribution Forcing for Long Video Generation at Test Time
测试时长视频生成的分布内强制策略
Abstract
Modern autoregressive (AR) video diffusion models excel at short-horizon video generation, yet generating long videos remains challenging due to drifting, where colors and textures shift, and motion dynamics decay. Existing works primarily rely on KV conditioning, which selects or modifies cached key-value (KV) entries to mitigate drifting. However, we observe that KV conditioning alone is insufficient as it assumes cached KV entries remain in-distribution. This assumption fails beyond the training horizon: nothing constrains the construction of KV entries during rollout, giving rise to the KV-provenance problem where cached entries themselves become out-of-distribution (OOD). To address this, we propose In-Distribution Forcing (ID-Forcing), a test-time framework that aligns both KV caching and KV conditioning with training configurations. Its key mechanism, self-caching, prevents OOD KV entries at their source. Each chunk is cached without attending to prior KV entry, keeping the rolling window exactly in-distribution. Consequently, ID-Forcing seamlessly extends short-horizon models to minute-scale video generation. Extensive evaluations show that our method remains competitive on standard video generation benchmark while substantially outperforming prior work in mitigating drifting, as validated by both our drift metrics and a user study.
Chinese Translation
现代自回归（AR）视频扩散模型在短时间视频生成方面表现出色，但生成长视频仍然具有挑战性，主要是由于漂移现象，即颜色和纹理的变化以及运动动态的衰减。现有研究主要依赖于键值（KV）条件化，通过选择或修改缓存的键值（KV）条目来减轻漂移。然而，我们观察到单靠KV条件化是不够的，因为它假设缓存的KV条目保持在分布内。这个假设在训练范围之外失效：在生成过程中，没有任何约束限制KV条目的构建，从而引发了KV来源问题，即缓存条目本身变得超出分布（OOD）。为了解决这个问题，我们提出了分布内强制策略（In-Distribution Forcing，ID-Forcing），这是一个在测试时对KV缓存和KV条件化与训练配置进行对齐的框架。其关键机制是自缓存，防止OOD KV条目在源头上产生。每个块在缓存时不考虑先前的KV条目，从而保持滚动窗口完全在分布内。因此，ID-Forcing可以无缝地将短时间模型扩展到分钟级视频生成。大量评估表明，我们的方法在标准视频生成基准上仍具竞争力，同时在减轻漂移方面显著优于先前的工作，这一点通过我们的漂移指标和用户研究得到了验证。
cs.CV / 51 / 2610.03123
Foresight: planning future perception in streaming VLMs without retraining
前瞻性：在不重新训练的情况下规划流媒体视觉语言模型中的未来感知
Abstract
Existing streaming vision-language models (VLMs) continuously perceive and reason over visual streams, but their computational pathways remain fixed throughout inference. Consequently, they cannot adapt computation to evolving scene dynamics, where different future events demand different levels and forms of perception. We show that streaming VLMs inherently possess the ability to anticipate the immediate future, and leverage this capability to dynamically configure future computation in a training-free manner. Realizing such anticipatory computation, however, is very challenging: future anticipation must be sufficiently reliable to guide computation, planning must run concurrently with streaming inference, and online reconfiguration must incur negligible overhead. To address these challenges, we introduce FORESIGHT, a dual-stream architecture comprising two Siamese LLMs with shared weights, input encoders, and KV cache. The first LLM continuously processes incoming tokens, while the second runs ahead of the stream to anticipate future context, plan future computation, and generate task responses without interrupting streaming inference. Each plan decides when to reason next, what to check then, and how densely to sample, keeping transient evidence separate from persistent control. The resulting computation plan is executed online through an efficient reconfiguration protocol with schemaguided decoding and lightweight diff-based updates, enabling dynamic adaptation with low overhead. With a frozen Qwen3-VL-8B backbone, FORESIGHT achieves 23.0 mean joint F1 on OmniPro Online evaluation beating strongest trained baseline by 9.5%, while improving the backbone by 6.7 on StreamingBench and 15.4 on OVO-Bench, with the largest gain of 18.7 when evidence arrives later in the video stream. Our source code will be made publicly available.
Chinese Translation
现有的流媒体视觉语言模型（VLMs）持续感知并推理视觉流，但其计算路径在推理过程中保持固定。因此，它们无法根据不断变化的场景动态调整计算，而不同的未来事件需要不同水平和形式的感知。我们展示了流媒体VLMs本质上具备预测近期未来的能力，并利用这一能力以无训练的方式动态配置未来计算。然而，实现这种预测计算非常具有挑战性：未来的预测必须足够可靠以指导计算，规划必须与流媒体推理并行进行，在线重新配置必须带来微不足道的开销。为了解决这些挑战，我们提出了FORESIGHT，一种双流架构，由两个共享权重的Siamese LLMs组成，具有输入编码器和KV缓存。第一个LLM持续处理传入的标记，而第二个LLM则在流的前方运行，以预测未来上下文、规划未来计算并生成任务响应，而不干扰流媒体推理。每个计划决定何时进行下一个推理、接下来检查什么以及如何密集采样，将瞬态证据与持久控制分开。通过一种高效的重新配置协议，结合模式引导解码和轻量级差异更新，在线执行生成的计算计划，实现动态适应且开销低廉。在冻结的Qwen3-VL-8B骨干网络上，FORESIGHT在OmniPro在线评估中获得23.0的平均联合F1，超越最强训练基线9.5%，同时在StreamingBench上提高了6.7，在OVO-Bench上提高了15.4，当证据在视频流中稍晚到达时，获得的最大增益为18.7。我们的源代码将公开发布。
cs.CV / 52 / 2610.03141
Behavior Pack Optimization for Video MLLM Post-Training
视频多模态大语言模型后训练的行为包优化
Abstract
Video multimodal large language models (MLLMs) keep climbing video question answering benchmarks, yet shuffling the frames, masking the segment that supports the answer, or occluding the target object barely changes their predictions. The accuracy rests on appearance and language priors, not on the temporal evidence the question asks for. We trace this to the unit of post-training: rewards are computed on a single response to the original clip, so the model is never asked to behave consistently across views. We propose Behavior Pack Optimization (BPO), which replaces the single response with a behavior pack of outputs across counterfactual views chosen by question type, scored jointly. The pack reward asks for stability when the intervention is irrelevant, sensitivity when key evidence is removed, and abstention when no evidence remains. To keep this objective stable at small pack sizes, BPO uses an anchor-relative advantage: the response on the original view serves as a per-prompt reference instead of a group mean over mixed views. On TempCompass, MVBench, and NExT-QA, BPO improves the macro accuracy of Qwen2.5-VL-7B-Instruct by 4.7 pp, the temporal-hard subset by 7.8 pp, and abstention F1 by 20.0 pp over a budget-matched vanilla GRPO baseline from the same SFT checkpoint. The gains transfer to Video-MME, LongVideoBench, and to LLaVA-Video-7B; ablations confirm they follow the view sets, not the rollout count. We hope this pack-level perspective offers a useful starting point for the video MLLM and multimodal post-training community as the field moves toward evidence-grounded video reasoning.
Chinese Translation
视频多模态大语言模型（MLLMs）在视频问答基准测试中不断取得进展，但对帧进行重排、遮蔽支持答案的片段或遮挡目标物体几乎不会改变它们的预测。其准确性依赖于外观和语言先验，而不是问题所要求的时间证据。我们将这一现象追溯到后训练的单位：奖励是基于对原始片段的单一响应计算的，因此模型从未被要求在不同视角下保持一致的行为。我们提出了行为包优化（Behavior Pack Optimization, BPO），它用跨反事实视角的输出行为包替代单一响应，这些视角是根据问题类型选择的，并共同评分。该包奖励在干预无关时要求稳定性，在关键证据被移除时要求敏感性，在没有证据时要求弃权。为了在小包大小下保持这一目标的稳定性，BPO使用了锚点相对优势：原始视角的响应作为每个提示的参考，而不是混合视角的组均值。在TempCompass、MVBench和NExT-QA上，BPO将Qwen2.5-VL-7B-Instruct的宏观准确率提高了4.7个百分点，时间困难子集提高了7.8个百分点，弃权F1提高了20.0个百分点，相较于来自同一SFT检查点的预算匹配的普通GRPO基线。这些增益转移到Video-MME、LongVideoBench和LLaVA-Video-7B；消融实验确认它们遵循视角集，而不是回放计数。我们希望这种包级视角为视频MLLM和多模态后训练社区提供一个有用的起点，推动该领域向基于证据的视频推理发展。
cs.CV / 53 / 2610.03142
CalCErt: Bin-wise Certification of Confidence Calibration in Medical Image Classification
CalCErt：医学图像分类中置信度校准的逐箱认证
Abstract
Deep neural networks remain vulnerable to adversarial perturbations, which can distort not only predictions but also confidence scores, undermining uncertainty calibration. While existing certification methods focus on preserving the predicted category, providing guarantees on how calibration behaves under adversarial attacks remains overlooked. In this work, we introduce CalCErt, a simple and efficient post-hoc strategy that certifies bin-wise confidence calibration for any pretrained differentiable classifier. Our approach combines empirical calibration estimates, statistical concentration bounds, and local Lipschitz estimates of the confidence function to derive data-dependent upper bounds on worst-case miscalibration within an $ell_2$-ball of radius R. We evaluate CalCErt across 11 medical image classification tasks and multiple adversarial perturbations, demonstrating substantially higher certified coverage than baseline strategies while maintaining competitive tightness. Our code is available at https://github.com/leofillioux/calcert.
Chinese Translation
深度神经网络仍然容易受到对抗性扰动的影响，这不仅会扭曲预测结果，还会影响置信度评分，从而削弱不确定性校准。尽管现有的认证方法专注于保持预测类别，但在对抗攻击下如何保证校准行为仍然被忽视。在本研究中，我们提出了CalCErt，这是一种简单而高效的后处理策略，能够对任何预训练的可微分类器进行逐箱置信度校准认证。我们的方法结合了经验校准估计、统计集中界限和置信度函数的局部Lipschitz估计，以推导出在半径为R的$ell_2$-球体内最坏情况下的误校准的数据依赖上界。我们在11个医学图像分类任务和多种对抗性扰动下评估了CalCErt，结果表明其认证覆盖率显著高于基线策略，同时保持了竞争性的紧密性。我们的代码可在https://github.com/leofillioux/calcert获取。
cs.CV / 54 / 2610.03154
Does Physics Live in the Activations? Localizing Physical Quantities in Video Diffusion Models
物理是否存在于激活中？在视频扩散模型中定位物理量
Abstract
Video generation models produce strikingly realistic sequences and are increasingly proposed as world models, yet recent benchmarks reveal pronounced deficits in their physical reasoning. This raises the question of whether these models internalize physical principles or merely reproduce familiar motion patterns. We address this by probing internal representations of video Diffusion Transformers (DiTs) for simulator-derived ground-truth physical quantities spanning kinematic motion and rigid-body dynamics under gravity and contact. We find that these quantities are linearly decodable with high accuracy early in the denoising process, substantially outperforming a baseline decoded directly from the model's own noised latents, indicating that the relevant physical information is actively constructed during denoising rather than already present in the input. Additionally, we show that activations at on-object tokens carry the relevant physical information and that quantities defined over multiple frames are readable from single latent frames. Hence, information is sharply localized within the token sequence and is computed globally but stored locally. The probes further show partial extrapolation, transferring to scene variations and object configurations outside their training regime, so what they read is not simply a correlate of the scenes they were fit on. When fitted directly in the full-resolution activation space, the probing directions can serve as steering vectors to change the model's output.
Chinese Translation
视频生成模型能够生成极其逼真的序列，越来越多地被提议作为世界模型，然而最近的基准测试揭示了它们在物理推理方面的明显不足。这引发了一个问题：这些模型是否内化了物理原则，还是仅仅重现了熟悉的运动模式。我们通过探测视频扩散变换器（Diffusion Transformers, DiTs）的内部表示，来解决这个问题，研究模拟器导出的真实物理量，这些量涵盖了重力和接触下的运动学运动和刚体动力学。我们发现，这些物理量在去噪过程的早期阶段可以以高精度线性解码，显著优于直接从模型自身的噪声潜变量解码的基线，这表明相关的物理信息是在去噪过程中主动构建的，而不是已经存在于输入中。此外，我们还表明，在物体标记上的激活携带相关的物理信息，并且在多个帧上定义的量可以从单个潜变量帧中读取。因此，信息在标记序列中被明确局部化，并且是全局计算但局部存储的。探测还显示出部分外推，能够转移到其训练范围之外的场景变化和物体配置，因此它们读取的信息并不仅仅是与其拟合场景的相关性。当直接在全分辨率激活空间中进行拟合时，探测方向可以作为引导向量来改变模型的输出。
cs.CV / 55 / 2610.03167
Geometry-Aligned Semantic Matching for Cross-Modal Planar Image Registration
几何对齐的语义匹配用于跨模态平面图像配准
Abstract
Cross-modal image matching establishes stable and accurate geometric correspondences across modalities for planar registration. Existing semantic representations provide cross-modal consistency, but semantic similarity does not necessarily imply geometric correspondence. Meanwhile, fine-grained CNN features provide accurate local details but lack global cross-modal semantic guidance for stable refinement. To address these issues, we propose CDPM, which first establishes geometrically consistent semantic representations and then preserves their dominant role in correspondence estimation during fine-grained localization. Specifically, we progressively adapt DINOv3 using geometrically consistent cross-modal patch pairs, enabling feature similarity to better reflect true cross-modal spatial correspondences. We then construct a DINO-Centric Feature Pyramid, where multi-scale DINO representations maintain stable cross-modal correspondences, while a lightweight CNN branch provides auxiliary structural details for precise local refinement. Extensive experiments on three cross-modal datasets demonstrate the superior performance of CDPM. On VIS-IR, compared with the dense matcher RoMa, CDPM improves AUC@3/5/10/20 by 7.36, 13.40, 13.75, and 10.42 percentage points, respectively, and reduces mACE from 5.83 to 2.78 pixels. It also outperforms RoMa v2 across all metrics while requiring 45.6% fewer FLOPs. The online demo and dataset are available, and the code will be released on our project page at https://warren-wzw.github.io/CDPM/.
Chinese Translation
跨模态图像匹配在平面配准中建立了跨模态的稳定和准确的几何对应关系。现有的语义表示提供了跨模态的一致性，但语义相似性并不一定意味着几何对应关系。同时，细粒度的卷积神经网络（CNN）特征提供了准确的局部细节，但缺乏稳定细化所需的全局跨模态语义指导。为了解决这些问题，我们提出了CDPM，首先建立几何一致的语义表示，然后在细粒度定位过程中保持其在对应估计中的主导作用。具体而言，我们使用几何一致的跨模态补丁对逐步调整DINOv3，使特征相似性更好地反映真实的跨模态空间对应关系。然后，我们构建了一个以DINO为中心的特征金字塔，其中多尺度的DINO表示保持稳定的跨模态对应关系，而一个轻量级的CNN分支提供辅助结构细节以实现精确的局部细化。在三个跨模态数据集上的大量实验表明CDPM的优越性能。在VIS-IR上，与密集匹配器RoMa相比，CDPM在AUC@3/5/10/20上分别提高了7.36、13.40、13.75和10.42个百分点，并将mACE从5.83减少到2.78像素。它在所有指标上也优于RoMa v2，同时减少了45.6%的FLOPs。在线演示和数据集可用，代码将在我们的项目页面https://warren-wzw.github.io/CDPM/上发布。
cs.CV / 56 / 2610.03192
PocketSplat: Mobile Gaussian Reconstruction via World-Space Latent Allocatio
PocketSplat：通过世界空间潜在分配进行移动高斯重建
Abstract
Mobile Gaussian reconstruction must satisfy two requirements: the reconstruction model must execute within a device resource envelope, and the resulting Gaussian asset must expose a representation size suited to downstream mobile use. Existing feed-forward Gaussian reconstructors commonly decode dense, image-aligned candidates whose final cardinality is implicitly determined by the input resolution and number of views. We present PocketSplat, a feed-forward framework for budgeted mobile Gaussian asset construction. Given a prescribed output budget, PocketSplat organizes dense geometry-aware latent candidates in predicted world space, allocates exact integer capacity across local latent cells, and decodes complete Gaussian attributes only for retained candidates. Cell-conditioned latent fusion aggregates repeated multi-view evidence before decoding, while spatial responsibility decoding adapts Gaussian support after local sparsification. Experiments on DL3DV and out-of-distribution benchmarks establish a strong quality--budget trade-off against feed-forward Gaussian reconstruction baselines. On Mip-NeRF 360, PocketSplat executes directly on a target iPhone and constructs compact, higher-quality Gaussian assets substantially faster than a deployable streamed MVSplat variant; native MVSplat and DepthSplat exceed the device memory budget.
Chinese Translation
移动高斯重建必须满足两个要求：重建模型必须在设备资源限制内执行，且生成的高斯资产必须具有适合下游移动使用的表示大小。现有的前馈高斯重建器通常解码密集的、与图像对齐的候选项，其最终基数隐含地由输入分辨率和视图数量决定。我们提出了PocketSplat，这是一个用于预算移动高斯资产构建的前馈框架。在给定的输出预算下，PocketSplat在预测的世界空间中组织密集的几何感知潜在候选项，在局部潜在单元中分配精确的整数容量，并仅为保留的候选项解码完整的高斯属性。基于单元的潜在融合在解码之前聚合重复的多视图证据，而空间责任解码在局部稀疏化后调整高斯支持。在DL3DV和分布外基准上的实验表明，PocketSplat在质量与预算之间建立了强有力的权衡，相较于前馈高斯重建基线。在Mip-NeRF 360上，PocketSplat直接在目标iPhone上执行，并以显著快于可部署的流式MVSplat变体的速度构建紧凑且高质量的高斯资产；原生的MVSplat和DepthSplat超出了设备内存预算。
cs.CV / 57 / 2610.03193
Bridging Research and Practice: A Systematic Evaluation of Generalist and Dermatology-Specific Models in Clinical Skin Lesion Classification
连接研究与实践：临床皮肤病变分类中通用模型与皮肤病特定模型的系统评估
Abstract
The application of machine learning to dermatology has grown substantially in recent years, moving beyond proof-of-concept studies toward potential applications. However, clinical dermatology remains a challenging and still open problem. Diagnostic assessment is often ambiguous, and skin lesions exhibit high variability, compounded by differences in acquisition modality, device quality, and patient demographics. These factors hinder the development of robust models suitable for safe and equitable clinical use. To support translation into practice, it is essential to systematically evaluate how contemporary models generalize across heterogeneous data sources. In this work, we benchmark a diverse set of architectures on recent dermatology datasets, spanning dermoscopic images and smartphone-based clinical photographs. We assess the robustness of recent general-purpose and medical vision-language models, as well as foundation models, and compare them against task-specific dermatology classifiers, including embedding-based approaches and convolutional neural networks. Our study provides an evaluation of model performance under distribution shifts, modality changes, and demographic variability. By quantifying the gap between current state-of-the-art models and the requirements of clinical deployment, we aim to contribute to the development of reliable, accessible, and clinically applicable AI systems for dermatology.
Chinese Translation
近年来，机器学习在皮肤病学中的应用显著增长，已从概念验证研究转向潜在应用。然而，临床皮肤病学仍然是一个具有挑战性且尚未解决的问题。诊断评估常常模糊不清，皮肤病变表现出高度的变异性，这种变异性受到获取方式、设备质量和患者人口统计特征差异的影响。这些因素阻碍了适合安全和公平临床使用的稳健模型的发展。为了支持实践中的转化，系统评估当代模型在异构数据源中的泛化能力至关重要。在本研究中，我们在近期的皮肤病学数据集上基准测试了一组多样化的架构，这些数据集涵盖了皮肤镜图像和基于智能手机的临床照片。我们评估了最新的通用和医学视觉语言模型的稳健性，以及基础模型，并将它们与任务特定的皮肤病分类器进行比较，包括基于嵌入的方法和卷积神经网络。我们的研究提供了在分布变化、模态变化和人口变异性下模型性能的评估。通过量化当前最先进模型与临床部署要求之间的差距，我们旨在为开发可靠、可获取且适用于临床的皮肤病学人工智能系统做出贡献。
cs.CV / 58 / 2610.03202
Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation
上下文流匹配：流模型中自适应步长选择以实现高效视觉生成
Abstract
Flow Matching enables high-quality visual generation via continuous-time dynamics, but inference remains costly due to multiple sequential function evaluations. Existing acceleration methods reduce the number of function evaluations but often introduce additional training overhead, degrade quality, or fail to account for input-dependent variability. We propose COFLOW, an inference-time method that adaptively selects the step counts each generation based on the prompt features. Our context-aware COFLOW is trained online with an unsupervised reward that balances inference efficiency and generation fidelity. Our method is plug-and-play, requiring no retraining of the underlying generative model. It generalizes to image and video generation, achieving over 2.5x speedup while preserving perceptual and semantic quality. We further provide a theoretical analysis establishing an O(1/K) forward-Euler discretization error bound under standard regularity conditions.
Chinese Translation
流匹配通过连续时间动态实现高质量的视觉生成，但推理过程仍然代价高昂，因为需要多次顺序函数评估。现有的加速方法减少了函数评估的次数，但通常会引入额外的训练开销，降低质量，或者未能考虑输入依赖的变异性。我们提出了COFLOW，这是一种在推理时根据提示特征自适应选择每次生成步数的方法。我们上下文感知的COFLOW通过无监督奖励在线训练，平衡推理效率和生成保真度。我们的方法是即插即用的，无需重新训练基础生成模型。它可以推广到图像和视频生成，取得了超过2.5倍的加速，同时保持感知和语义质量。我们进一步提供了理论分析，建立了在标准正则性条件下O(1/K)的前向欧拉离散化误差界限。
cs.CV / 59 / 2610.03218
VisionMX: Unlocking Microscaling Post-Training Quantization for Vision Models
VisionMX：解锁视觉模型的微缩后训练量化
Abstract
Microscaling (MX) formats are emerging as a hardware-supported approach to efficient training and inference. They combine low-precision elements with shared block scales, but their impact on vision models remains underexplored. We systematically investigate post-training MX quantization across vision models and tasks. An analysis of direct conversion identifies three sources of error: block-scale representation, the poor alignment of some small convolutional weight tensors with nonuniform element grids, and the underuse of signed codes by nonnegative activations. These findings motivate VisionMX, a post-training MX quantization method that optimizes bounded weight rounding and applies a foldable affine correction to activations. We evaluate VisionMX across image classification, object detection, semantic segmentation, and low-light image enhancement using several MX-style formats. It improves on direct conversion and the evaluated post-training quantization baselines, with the largest performance recoveries in architectures most sensitive to MX conversion
Chinese Translation
微缩（MX）格式作为一种硬件支持的高效训练和推理方法正在兴起。它们将低精度元素与共享块尺度相结合，但其对视觉模型的影响仍然未得到充分探讨。我们系统地研究了视觉模型和任务中的后训练MX量化。对直接转换的分析识别出三个误差来源：块尺度表示、小卷积权重张量与非均匀元素网格之间的对齐不良，以及非负激活对带符号编码的使用不足。这些发现促使我们提出VisionMX，一种优化有界权重舍入并对激活应用可折叠仿射校正的后训练MX量化方法。我们在图像分类、目标检测、语义分割和低光图像增强等任务中评估了VisionMX，使用了几种MX风格格式。它在直接转换和评估的后训练量化基线之上取得了改进，在对MX转换最敏感的架构中实现了最大的性能恢复。
cs.CV / 60 / 2610.03221
VDOT++: Unified Few-Step Video Generation via Unbalanced Optimal Transport Distillation
VDOT++：通过不平衡最优传输蒸馏实现统一的少步视频生成
Abstract
Video creation spans text-to-video (T2V), image-to-video (I2V), and condition-based generation, yet video diffusion models remain costly because they repeatedly evaluate large backbones during sampling. Distribution matching distillation (DMD) reduces this cost, but its reverse Kullback--Leibler (KL) objective can provide unstable or incomplete guidance when the student and teacher distributions have limited overlap. VDOT addressed this issue by adding optimal transport distillation (OTD), whose explicit coupling supplies geometric directions for condition-based generation. Balanced OTD, however, performs full-mass matching between the spatial tokens of each corresponding student--teacher frame pair. This assumption weakens for T2V and I2V, where one condition admits many valid outputs and spatial content need not align across different realizations. We present VDOT++, a unified distillation framework that applies the same training recipe separately to generators for the three task families. It makes OTD robust to output diversity through an asymmetric unbalanced formulation that allows unreliable student tokens to carry less mass while maintaining coverage of the teacher tokens. An $\ell_1$ ground cost further replaces mean-based aggregation with a more mode-preserving weighted median that limits the influence of distant transport targets. The two changes respectively determine whom to match and how the selected targets should be aggregated. We additionally combine distribution matching and adversarial refinement through sequential backward passes, and exploit the decoupled score networks for cross-scale distillation, where larger score networks improve a compact generator. Experiments on UVCBench, VBench, VBench-I2V, and the VACE benchmark show that the resulting four-step generators are competitive with many-step teachers and strong few-step baselines across all three task families.
Chinese Translation
视频创作涵盖文本到视频（T2V）、图像到视频（I2V）和基于条件的生成，然而视频扩散模型在采样过程中仍然成本高昂，因为它们在采样时反复评估大型骨干网络。分布匹配蒸馏（DMD）减少了这一成本，但其反向Kullback-Leibler（KL）目标在学生和教师分布重叠有限时可能提供不稳定或不完整的指导。VDOT通过添加最优传输蒸馏（OTD）来解决这一问题，其显式耦合为基于条件的生成提供了几何方向。然而，平衡OTD在每对对应的学生-教师帧之间执行全质量匹配。这一假设在T2V和I2V中变得薄弱，因为一个条件可以产生多个有效输出，并且不同实现之间的空间内容不必对齐。我们提出了VDOT++，一个统一的蒸馏框架，它对三类任务的生成器应用相同的训练方案。它通过不对称的不平衡形式使OTD对输出多样性具有鲁棒性，允许不可靠的学生标记携带较少的质量，同时保持教师标记的覆盖。一个$ ext{l}_1$地面成本进一步用更具模式保持性的加权中位数替代了基于均值的聚合，从而限制了远程传输目标的影响。这两个变化分别决定了匹配对象和如何聚合所选目标。我们还通过顺序反向传递结合了分布匹配和对抗性精炼，并利用解耦的评分网络进行跨尺度蒸馏，其中更大的评分网络改善了紧凑生成器。在UVCBench、VBench、VBench-I2V和VACE基准上的实验表明，所得到的四步生成器在所有三类任务中与多步教师和强少步基线具有竞争力。
cs.CV / 61 / 2610.03224
Uncertainty as a Proxy for Semantic Correctness in Diffusion-Based Medical Image Synthesis
不确定性作为基于扩散的医学图像合成中语义正确性的代理
Abstract
Diffusion models can synthesise contrast-enhanced CT (CECT) from non-contrast CT (NCCT), avoiding contrast administration and its environmental and patient-access costs. However, visually realistic images are not necessarily anatomically correct, and the pixel-intensity and feature-space similarity metrics used to assess generation quality do not directly measure anatomical correctness. In this work, we investigate whether uncertainty can serve as a proxy for semantic correctness in diffusion-based medical image synthesis. We study NCCT-to-CECT synthesis using AortaDiff, a multitask diffusion framework that jointly generates CECT images and lumen segmentations. The segmentation output provides an explicit representation of the generated vascular anatomy, enabling segmentation-derived errors to be used as a quantitative measure of generation correctness. Six methods spanning weight (Ensemble, HyperDiff, BayesDiff), architecture-perturbation (MCDropout), generative-stochasticity (RDS) and input-perturbation (TTA) uncertainty are compared at the pixel, region and image levels, and for detection of clinically relevant out-of-distribution (OOD) cases. Uncertainty proves informative at all three spatial scales, remains informative on an external multi-centre dataset under distribution shift, and supports OOD detection. MCDropout stands out among the six: it ranks among the leading methods at every scale, generalizes well on the external dataset, and can be enabled at inference on any model already trained with dropout, so reliable uncertainty comes at no extra training cost. Uncertainty reliably flags severe failures but discriminates poorly among already high-quality images. These findings support uncertainty as a practical and computationally economical signal for quality filtering, reliability assessment and OOD detection in NCCT-to CECT synthesis.
Chinese Translation
扩散模型可以从非对比CT（NCCT）合成增强对比CT（CECT），避免了对比剂的使用及其对环境和患者的成本。然而，视觉上逼真的图像不一定在解剖上是正确的，而用于评估生成质量的像素强度和特征空间相似性度量并不能直接衡量解剖正确性。在本研究中，我们探讨了不确定性是否可以作为基于扩散的医学图像合成中语义正确性的代理。我们使用AortaDiff研究NCCT到CECT的合成，这是一种多任务扩散框架，能够联合生成CECT图像和腔体分割。分割输出提供了对生成血管解剖结构的明确表示，使得基于分割的错误可以作为生成正确性的定量测量。我们比较了六种方法，涵盖了权重（Ensemble、HyperDiff、BayesDiff）、架构扰动（MCDropout）、生成随机性（RDS）和输入扰动（TTA）不确定性，分别在像素、区域和图像层面，以及临床相关的分布外（OOD）案例检测中进行评估。不确定性在这三个空间尺度上均表现出信息价值，并在外部多中心数据集上在分布变化下仍然保持信息价值，并支持OOD检测。MCDropout在六种方法中表现突出：它在每个尺度上均排名靠前，在外部数据集上具有良好的泛化能力，并且可以在任何已经使用dropout训练的模型推理时启用，因此可靠的不确定性不会增加额外的训练成本。不确定性可靠地标识出严重的失败，但在已经高质量的图像中区分能力较差。这些发现支持不确定性作为NCCT到CECT合成中质量过滤、可靠性评估和OOD检测的实用且计算经济的信号。
cs.CV / 62 / 2610.03248
EmbPASS: Towards Cross-Embodiment Open Panoramic Segmentation
EmbPASS：面向跨体态开放全景分割
Abstract
Panoramic images provide a complete 360-degree field of view, enabling comprehensive scene understanding for embodied perception. However, heterogeneous embodied platforms exhibit substantial differences in observation viewpoints and spatial layouts, giving rise to cross-embodiment observation shifts that pose additional challenges to consistent and reliable panoramic perception, while systematic studies of this problem remain limited. To bridge this gap, we introduce a new task, termed Cross-Embodiment Open Panoramic Segmentation. Meanwhile, we establish EmbPASS, a multi-platform panoramic semantic segmentation benchmark spanning Vehicle, Drone, Wearable, and Quadruped platforms under a unified semantic taxonomy, providing a testbed for systematically studying cross-embodiment panoramic perception. We further propose EPONet, an open-vocabulary panoramic semantic segmentation network that integrates Relation-Aware Metric Adapter (RAMA) and Content-Adaptive Semantic Transfer (CAST) to enhance spatial modeling and semantic transfer under heterogeneous embodied observations. Extensive experiments show that EPONet achieves the best platform-balanced performance on EmbPASS with 35.82% mIoU, outperforming the strongest baseline by 1.10%, while remaining competitive on existing panoramic segmentation benchmarks. The source code and EmbPASS benchmark will be made publicly available at https://github.com/guopj1/EmbPASS.
Chinese Translation
全景图像提供了完整的360度视野，使得对具身感知的场景理解更加全面。然而，异构的具身平台在观察视点和空间布局上存在显著差异，导致跨体态观察偏移，这给一致且可靠的全景感知带来了额外挑战，而对这一问题的系统研究仍然有限。为填补这一空白，我们提出了一项新任务，称为跨体态开放全景分割（Cross-Embodiment Open Panoramic Segmentation）。同时，我们建立了EmbPASS，一个跨越车辆（Vehicle）、无人机（Drone）、可穿戴设备（Wearable）和四足动物（Quadruped）平台的多平台全景语义分割基准，采用统一的语义分类法，为系统研究跨体态全景感知提供了实验平台。我们进一步提出了EPONet，一个开放词汇的全景语义分割网络，集成了关系感知度量适配器（Relation-Aware Metric Adapter, RAMA）和内容自适应语义转移（Content-Adaptive Semantic Transfer, CAST），以增强在异构具身观察下的空间建模和语义转移。大量实验表明，EPONet在EmbPASS上实现了35.82%的最佳平台平衡性能，比最强基线提高了1.10%，同时在现有全景分割基准上保持竞争力。源代码和EmbPASS基准将公开发布于https://github.com/guopj1/EmbPASS。
cs.CV / 63 / 2610.03252
COSMI: COmpositional Synthesis of Multi-object Interactions
COSMI：多对象交互的组合合成
Abstract
Generative models of human-object interaction are bounded by the data that exists: everyday activities involve several objects, but most captured datasets record one at a time, as multi-object capture is combinatorially expensive. Our observation is that interactions are local, so single-object captures already contain the parts of multi-object activities. We compose them: contact-consistent clips of single interactions, mirrored to balance the hands, transfer between bodies, and a language model and geometric checks admit only the pairings that are plausible, semantically and physically. Therefore, the dataset grows combinatorially with the clips rather than recording time. The COSMI dataset holds 222k sequences and 275 hours with up to five objects, nearly thirty times the largest multi-object capture, and can be extended by adding datasets or even hand-object recordings. On this data we train the COSMI method, a text-to-interaction diffusion transformer that follows how the data is built: weight-shared object slots generate a variable number of objects, predicted relative to the body parts that move them. On a benchmark with an unseen object and unseen interaction combinations, models trained on the dataset generalize to the unseen combinations. COSMI outperforms baselines in text alignment and contact accuracy, where its margin is largest on the unseen object. Code, models, and the dataset pipeline will be released on the project page: https://ptrvilya.github.io/cosmi.
Chinese Translation
人类与物体交互的生成模型受限于现有数据：日常活动涉及多个物体，但大多数捕获的数据集一次只记录一个，因为多对象捕获在组合上成本高昂。我们的观察是，交互是局部的，因此单对象捕获已经包含了多对象活动的部分。我们将它们组合起来：保持接触一致性的单一交互片段，进行镜像以平衡双手，身体之间的转移，以及语言模型和几何检查仅允许在语义和物理上合理的配对。因此，数据集随着片段的增加而组合增长，而不是记录时间。COSMI 数据集包含 222,000 个序列和 275 小时的内容，最多支持五个对象，几乎是最大的多对象捕获的三十倍，并且可以通过添加数据集或甚至手-物体录音来扩展。在这些数据上，我们训练了 COSMI 方法，这是一种文本到交互的扩散变换器，遵循数据构建的方式：权重共享的对象槽生成可变数量的对象，相对于移动它们的身体部位进行预测。在一个包含未见对象和未见交互组合的基准测试中，基于该数据集训练的模型能够推广到未见组合。COSMI 在文本对齐和接触准确性方面优于基线，其中在未见对象上的优势最大。代码、模型和数据集管道将在项目页面发布：https://ptrvilya.github.io/cosmi。
cs.CV / 64 / 2610.03261
Consecutive Posterior Fusion for Diffusive Recovery of Unobservable Image Structures
连续后验融合用于不可观测图像结构的扩散恢复
Abstract
Solving severely ill-posed imaging inverse problems requires recovering image structures that are unobservable or weakly constrained by the measurements. Diffusion models provide expressive learned priors for inferring such missing information, while posterior sampling incorporates measurement consistency along the reverse process. Standard diffusion posterior samplers, however, rely on instantaneous measurement-aware estimates, without explicitly exploiting information carried by previous posterior corrections. We introduce Consecutive Posterior Fusion Denoising Diffusion Null-Space Models (CPF-DDNM), an inference-time strategy that fuses consecutive measurement-aware estimates to improve the diffusive recovery of unobservable image structures, without requiring retraining or additional denoiser evaluations. We instantiate this principle within DDNM, whose range/null-space decomposition reveals that consecutive fusion preserves the measurement-determined component while acting exclusively on the prior-driven null-space estimate. We thus provide a geometric interpretation of CPF-DDNM and a local error analysis that characterizes the optimal time-dependent fusion coefficient, including the extrapolative regime. Experiments on sparse-view and simulated low-dose computed tomography, as well as medical image super-resolution, show consistent improvements over DDNM and competitive performance against diffusion-based inverse solvers.
Chinese Translation
解决严重病态的成像逆问题需要恢复那些不可观测或由测量弱约束的图像结构。扩散模型提供了表达丰富的学习先验，用于推断这些缺失的信息，而后验采样则在逆过程中的测量一致性方面发挥作用。然而，标准的扩散后验采样器依赖于瞬时的测量感知估计，而没有明确利用先前后验修正所携带的信息。我们提出了连续后验融合去噪扩散零空间模型（Consecutive Posterior Fusion Denoising Diffusion Null-Space Models, CPF-DDNM），这是一种在推断时融合连续测量感知估计的策略，以改善不可观测图像结构的扩散恢复，而无需重新训练或额外的去噪器评估。我们在DDNM中实例化这一原理，其范围/零空间分解揭示了连续融合保留了测量确定的成分，同时仅对先验驱动的零空间估计进行操作。因此，我们提供了CPF-DDNM的几何解释和局部误差分析，表征了最优的时间依赖融合系数，包括外推区域。在稀疏视图和模拟低剂量计算机断层扫描以及医学图像超分辨率的实验中，显示出相较于DDNM的一致性改进，并在与基于扩散的逆求解器的竞争中表现出色。
cs.CV / 65 / 2610.03276
Moving Forward with Video Saliency: A New Dataset and Benchmark where Motion Matters
视频显著性研究的新进展：一个运动重要的新数据集和基准
Abstract
Video saliency prediction is inherently harder to model than static image saliency due to the additional temporal dimension. Video saliency benchmarks rest on the premise that predicting gaze on video requires utilizing temporal activity distributed across frames. Prior work has challenged this, showing that static baselines recover a significant fraction of the explainable gaze information on LEDOV, a popular video saliency dataset, and that video saliency models fail in the same places as this static baseline. We verify that this diagnosis still stands: under a more capable gold standard than the original analysis, and an updated panel of recent architectures, the strongest temporal architecture in the panel still does not substantially improve over a fine-tuned static baseline. However, it remains unclear whether the marginal gain reflects limitations of current temporal architectures or a lack of temporal patterns in the benchmark itself. We introduce SalTempto, a video saliency benchmark with greater dynamism: 224 clips of highly dynamic content, sourced from the HACS-Segments dataset so that each clip contains an event together with its lead-up and aftermath, with gaze recordings from up to 16 subjects and a training split for adapting pretrained models. On SalTempto, the static baseline recovers only about 13\% of the headroom above the centerbias, against more than half on LEDOV. A fine-tuned temporal architecture shows a substantial gain in performance over the static baseline, indicating that it does capture meaningfully more temporal information, which LEDOV fails to measure. Yet, even this SoTA model still leaves nearly half of SalTempto's headroom unexplained, indicating room for improvement in video saliency modelling. Examination of SalTempto also lets us describe human tendencies that models miss. SalTempto link: https://huggingface.co/datasets/bethgelab/video_saliency.
Chinese Translation
视频显著性预测由于额外的时间维度，固有地比静态图像显著性更难建模。视频显著性基准建立在这样一个前提上：预测视频中的注视需要利用分布在帧之间的时间活动。之前的研究对此提出了质疑，显示静态基线在流行的视频显著性数据集LEDOV上恢复了相当一部分可解释的注视信息，并且视频显著性模型在与该静态基线相同的地方失败。我们验证了这一诊断仍然成立：在比原始分析更强大的金标准下，以及更新的近期架构面板中，面板中最强的时间架构仍未显著超越经过微调的静态基线。然而，目前尚不清楚这种边际增益是否反映了当前时间架构的局限性，或是基准本身缺乏时间模式。我们引入了SalTempto，一个具有更大动态性的视频显著性基准：224个高度动态内容的剪辑，来源于HACS-Segments数据集，使得每个剪辑包含一个事件及其前因和后果，包含来自多达16个受试者的注视记录，并提供用于适应预训练模型的训练划分。在SalTempto上，静态基线仅恢复了中心偏差以上约13%的余量，而在LEDOV上则超过了一半。经过微调的时间架构在性能上显示出相对于静态基线的显著提升，表明它确实捕捉到了比LEDOV更有意义的时间信息。然而，即使是这个最先进的模型仍然留下了近一半SalTempto的余量未被解释，表明视频显著性建模仍有改进空间。对SalTempto的检查还使我们能够描述模型所忽视的人类倾向。SalTempto链接：https://huggingface.co/datasets/bethgelab/video_saliency。
cs.CV / 66 / 2610.03308
T3lescope: Arbitrary-Resolution High-Fidelity Generative Surface Reconstruction from Images
T3lescope：基于图像的任意分辨率高保真生成表面重建
Abstract
We reconstruct high-fidelity 3D scene meshes from posed multi-view images without per-scene optimization, across scales ranging from single objects to large outdoor scenes. Per-scene optimization methods lack the learned 3D prior needed when observations are sparse or surfaces are glossy or transparent. Existing generative methods leverage such priors to complete geometry in sparsely observed regions, but typically operate at a fixed resolution over a limited spatial extent, trading spatial coverage against detail. Reconstructing a large scene therefore often requires partitioning it into independently processed overlapping local regions, making it difficult to maintain global geometric consistency. To address these issues, we propose T3lescope, which applies a single fixed-resolution generator across scene scales in an inference-time coarse-to-fine cascade. A coarse level establishes the scene layout, and finer levels perturb and denoise geometry inherited from the coarser level within progressively finer spatial cells to recover surface detail. The model is trained on individual cells at multiple scales and shares its weights across all levels, so no hierarchy is fixed during training, and the number of levels, cell scales, and cell locations are determined at inference time. On indoor, outdoor, and city-scale scenes, T3lescope outperforms feed-forward and generative baselines, matches or surpasses per-scene optimization, and recovers fine structures as well as glossy and transparent surfaces. These results show that our method generalizes across diverse scenes, view counts, and image resolutions. Project page: https://pfnet-research.github.io/t3lescope/
Chinese Translation
我们从带姿态的多视角图像中重建高保真的3D场景网格，无需针对每个场景进行优化，适用于从单个物体到大型户外场景的各种尺度。每个场景的优化方法在观察稀疏或表面光滑或透明时缺乏所需的学习到的3D先验。现有的生成方法利用这些先验在稀疏观察区域完成几何形状，但通常在有限的空间范围内以固定分辨率运行，牺牲了空间覆盖率以换取细节。因此，重建大型场景通常需要将其划分为独立处理的重叠局部区域，这使得保持全局几何一致性变得困难。为了解决这些问题，我们提出了T3lescope，它在推理时采用固定分辨率的生成器，通过粗到细的级联在场景尺度上进行应用。粗级别建立场景布局，细级别在逐渐细化的空间单元内扰动和去噪从粗级别继承的几何形状，以恢复表面细节。该模型在多个尺度的单个单元上进行训练，并在所有级别之间共享权重，因此在训练期间没有固定的层次结构，级别数量、单元尺度和单元位置在推理时确定。在室内、户外和城市规模的场景中，T3lescope的表现优于前馈和生成基线，匹配或超越每个场景的优化，并恢复精细结构以及光滑和透明表面。这些结果表明我们的方法在不同场景、视角数量和图像分辨率下具有良好的泛化能力。项目页面：https://pfnet-research.github.io/t3lescope/
cs.CV / 67 / 2610.03332
A Fully Automatic Pipeline for 3D Dendrite Instance Segmentation in SBF-SEM
一种用于SBF-SEM中3D树突实例分割的全自动管道
Abstract
Accurate three-dimensional (3D) reconstruction of individual dendrites in serial block-face scanning electron microscopy (SBF-SEM) is essential for quantifying structural plasticity in the brain, yet manual annotation at scale is infeasible. We present a fully automatic pipeline for 3D dendrite instance segmentation that unifies YOLOv6-guided Segment Anything Model (SAM) prompting on downsampled slices, iterative two-dimensional mask refinement, random forest 3D instance linking, and instance-aware high-resolution refinement using nnU-Net at native resolution into a single system requiring no manual prompting at inference. Applied to hippocampal CA1 SBF-SEM datasets from a control rat and a pilocarpine- induced epileptic rat, our pipeline reconstructs coherent, well- separated dendrites with high semantic accuracy (Dice 0.93 and 0.91) and strong instance-level performance on control tissue, while analysis of the more challenging epileptic tissue identifies instance recognition in dense regions as the principal remaining limitation. The high-resolution refinement stage recovers thin dendritic protrusions, providing a basis for downstream spine- level analysis. Code is available at https://github.com/ ZE-WEN/dendrite-3d-instance-seg.
Chinese Translation
在串行块面扫描电子显微镜（SBF-SEM）中，对单个树突进行准确的三维（3D）重建对于量化大脑的结构可塑性至关重要，但大规模手动标注是不可行的。我们提出了一种全自动的3D树突实例分割管道，该管道将基于YOLOv6的Segment Anything Model（SAM）提示、迭代的二维掩膜细化、随机森林3D实例链接以及使用nnU-Net在原始分辨率下进行的实例感知高分辨率细化统一为一个系统，在推理时无需手动提示。该管道应用于来自对照大鼠和巴比妥酸盐诱导癫痫大鼠的海马CA1 SBF-SEM数据集，重建出连贯且分离良好的树突，具有较高的语义准确性（Dice系数为0.93和0.91），并在对照组织上表现出强大的实例级性能，而对更具挑战性的癫痫组织的分析则发现，在密集区域的实例识别是主要的剩余限制。高分辨率细化阶段恢复了细小的树突突起，为后续的脊椎水平分析提供了基础。代码可在https://github.com/ZE-WEN/dendrite-3d-instance-seg获取。
cs.CV / 68 / 2610.03370
LAS-CLIP: A Lightweight Adapter Steering Approach for CLIP's Visual Encoder
LAS-CLIP：一种轻量级适配器引导方法用于CLIP的视觉编码器
Abstract
CLIP's visual encoder produces only global image representations, limiting its use in region-level tasks. Existing adaptations rely on visual prompting, input masking, or encoder fine-tuning, each compromising pre-trained representations. We propose LAS-CLIP, a Lightweight Adapter Steering approach that keeps every CLIP parameter frozen. A compact MaskAdapter generates per-head, per-layer attention biases from an input mask and injects them into the frozen self-attention layers, steering attention toward the target region. Crucially, because the backbone remains strictly untouched, LAS-CLIP seamlessly reverts to vanilla CLIP when no mask is provided, preserving its foundational zero-shot capabilities. With approximately 116K to 145K trainable parameters and 100K training samples on two T4 GPUs, LAS-CLIP achieves competitive or superior results compared to Alpha-CLIP on ImageNet-S zero-shot classification and RefCOCO referring expression comprehension, despite the latter fine-tuning its entire encoder on millions of samples. Qualitative analysis further confirms stronger representational fidelity under incorrect masks and in downstream generation. Our project page is link to https://github.com/AnhKhoa585/lasclip
Chinese Translation
CLIP的视觉编码器仅生成全局图像表示，限制了其在区域级任务中的应用。现有的适应方法依赖于视觉提示、输入掩码或编码器微调，每种方法都妥协了预训练表示。我们提出了LAS-CLIP，一种轻量级适配器引导方法，该方法保持每个CLIP参数不变。一个紧凑的MaskAdapter从输入掩码生成每个头、每层的注意力偏置，并将其注入到冻结的自注意力层中，引导注意力朝向目标区域。至关重要的是，由于主干网络保持严格不变，LAS-CLIP在未提供掩码时能够无缝恢复为原始CLIP，保留其基础的零-shot能力。在两个T4 GPU上，LAS-CLIP拥有约116K到145K的可训练参数和100K的训练样本，在ImageNet-S的零-shot分类和RefCOCO的指称表达理解中，尽管后者在数百万样本上微调了整个编码器，LAS-CLIP仍然实现了与Alpha-CLIP相竞争或更优的结果。定性分析进一步确认了在不正确掩码下和下游生成中的更强表示保真度。我们的项目页面链接为 https://github.com/AnhKhoa585/lasclip
cs.CV / 69 / 2610.03374
EVEWorld: Physical Evolution Supervision for Embodied World Models
EVEWorld：具身世界模型的物理演化监督
Abstract
Embodied world models enable scalable simulation of embodied interactions for robot learning. However, existing models are prone to Model Laziness, as they focus on visual fidelity at the expense of physical reasoning and lack process-level supervision over the temporal dynamics of manipulated objects. In this work, we propose EVEWorld, a physical evolution-supervision framework for physically consistent target evolution. EVEWorld consists of two components: Instance-Guided Restoration (IGR) and Temporal Instance Alignment (TIA). First, IGR promotes instance consistency through restoration supervision. Second, TIA promotes cross-frame consistency by aligning target instances across adjacent frames. We further introduce the Model Laziness Rate (MLR), a metric that measures persistent violations of instance consistency in generated trajectories. Extensive experiments on DreamGenBench, EWMBench, and PBench demonstrate the effectiveness of EVEWorld, notably achieving an 87.5% reduction in MLR compared with GigaWorld-0. On the WorldArena 2.0 Track 1 leaderboard, our model ranks 6th in JEPA Similarity and 17th overall, which further validates the performance of our evolution supervision strategy.
Chinese Translation
具身世界模型使得机器人学习中的具身交互能够进行可扩展的模拟。然而，现有模型容易出现模型懒惰（Model Laziness），因为它们在关注视觉逼真度的同时忽视了物理推理，并且缺乏对被操控物体的时间动态的过程级监督。在本研究中，我们提出了EVEWorld，一个用于物理一致目标演化的物理演化监督框架。EVEWorld由两个组成部分构成：实例引导恢复（Instance-Guided Restoration, IGR）和时间实例对齐（Temporal Instance Alignment, TIA）。首先，IGR通过恢复监督促进实例一致性。其次，TIA通过对齐相邻帧之间的目标实例促进跨帧一致性。我们进一步引入了模型懒惰率（Model Laziness Rate, MLR），这一指标用于衡量生成轨迹中实例一致性的持续违反。在DreamGenBench、EWMBench和PBench上的大量实验表明，EVEWorld的有效性，特别是与GigaWorld-0相比，MLR减少了87.5%。在WorldArena 2.0 Track 1排行榜上，我们的模型在JEPA相似度中排名第6，总体排名第17，这进一步验证了我们演化监督策略的性能。
cs.CV / 70 / 2610.03380
Interpretable Deepfake Detection in Videos via Explicit Forensic Features and Temporal Modeling
通过显式法医特征和时间建模实现可解释的视频深伪检测
Abstract
Deepfake detection in videos remains challenging, as manipulated content may appear visually consistent at the frame level while exhibiting subtle temporal inconsistencies. This paper introduces an interpretable deepfake detection framework that models spatially and temporally coherent facial features in video sequences. Unlike end-to-end deep models relying on implicit representations, the proposed approach explicitly encodes physically grounded forensic cues, enabling transparent analysis and improved multi-dataset generalization. The pipeline transforms videos into identity-consistent facial trajectories, segments them into fixed-length temporal windows, and represents each frame using 68 structured descriptors spanning four complementary domains: photometric, textural, geometric, and compression-based features. These descriptors provide a compact multi-domain representation of manipulation artifacts and are processed by a Long Short-Term Memory (LSTM) network to capture temporal dependencies and subtle irregularities. Evaluation on four benchmark datasets, FaceForensics++, Celeb-DF v2, a curated subset of the DeepFake Detection Challenge (DFDC), and DeeperForensics, yields strong and consistent F1-scores of 98.0%, 91.0%, 97.6%, and 96.2%, respectively. The approach also demonstrated a good cross-dataset generalization, providing a robust and interpretable solution for video deepfake detection.
Chinese Translation
视频中的深伪检测仍然具有挑战性，因为经过处理的内容在帧级别上可能看起来视觉一致，但在时间上却表现出微妙的不一致性。本文提出了一种可解释的深伪检测框架，该框架对视频序列中的空间和时间一致的面部特征进行建模。与依赖隐式表示的端到端深度模型不同，所提出的方法显式编码了物理基础的法医线索，从而实现透明分析和改进的多数据集泛化。该流程将视频转换为身份一致的面部轨迹，将其分段为固定长度的时间窗口，并使用68个结构化描述符表示每一帧，这些描述符涵盖了四个互补领域：光度、纹理、几何和基于压缩的特征。这些描述符提供了对操控伪影的紧凑多领域表示，并通过长短期记忆网络（LSTM）处理，以捕捉时间依赖性和微妙的不规则性。在四个基准数据集上进行评估，包括 FaceForensics++、Celeb-DF v2、深伪检测挑战（DFDC）的一个策划子集和 DeeperForensics，分别获得了强大且一致的 F1 分数，分别为 98.0%、91.0%、97.6% 和 96.2%。该方法还展示了良好的跨数据集泛化能力，为视频深伪检测提供了一个稳健且可解释的解决方案。
cs.CV / 71 / 2610.03389
From Patching to Pruning Visual Computation in Vision Language Models
从修补到修剪：视觉语言模型中的视觉计算
Abstract
Vision language models (VLMs) incur substantial inference cost because every visual token is processed by the attention and MLP projections of every decoder layer, even when token-specific visual computation is unnecessary at many depths. We introduce Patch-to-Prune (P2P), inspired by Mechanistic Interpretability, a training-free framework that converts activation patching from a diagnostic tool into an inference-time computation bypass. P2P performs validation-guided forward and backward layer sweeps to identify decoder regions whose visual-token projection outputs can be replaced by fixed neutral proxy activation vectors within a user-specified accuracy tolerance. Unlike conventional token-pruning methods, P2P preserves the sequence length, token order, positional information, attention mask, and residual pathways, thereby pruning computation without removing tokens or modifying the pretrained model weights. We evaluate P2P on four VLMs from the Qwen2.5-VL and LLaVA families across seven multi-modal benchmarks using mutually disjoint calibration, validation, and test partitions. P2P at a 3% tolerance retains around 94% of dense accuracy while reducing FLOPs by 55%. Beyond these efficiency gains, our layer-wise analysis suggests that visual processing in VLMs is non-uniformly distributed across decoder depth: early and late layers often require little token-specific visual computation, whereas intermediate layers appear to perform most task-relevant visual integration, enabling later reasoning to rely largely on visual information already embedded in shared residual and textual representations. This makes P2P both an efficient inference framework and a causal lens into visual information processing in VLMs.
Chinese Translation
视觉语言模型（VLMs）在推理时产生了巨大的成本，因为每个视觉标记都需要经过每个解码器层的注意力和多层感知器（MLP）投影，即使在许多深度上，特定于标记的视觉计算并不必要。我们提出了Patch-to-Prune（P2P），受到机械解释性（Mechanistic Interpretability）的启发，这是一个无训练的框架，将激活修补从诊断工具转变为推理时的计算旁路。P2P执行验证引导的前向和后向层扫描，以识别解码器区域，其视觉标记投影输出可以在用户指定的准确性容忍度内被固定的中性代理激活向量替代。与传统的标记修剪方法不同，P2P保留了序列长度、标记顺序、位置信息、注意力掩码和残差路径，从而在不移除标记或修改预训练模型权重的情况下修剪计算。我们在来自Qwen2.5-VL和LLaVA系列的四个VLM上评估了P2P，使用互不重叠的校准、验证和测试分区，在七个多模态基准上进行测试。在3%的容忍度下，P2P保留了约94%的密集准确性，同时减少了55%的FLOPs。除了这些效率提升之外，我们的逐层分析表明，VLM中的视觉处理在解码器深度上是不均匀分布的：早期和晚期层通常需要很少的特定于标记的视觉计算，而中间层似乎执行了大多数与任务相关的视觉整合，使得后续推理在很大程度上依赖于已经嵌入共享残差和文本表示中的视觉信息。这使得P2P既是一个高效的推理框架，也是一个理解VLM中视觉信息处理的因果视角。
cs.CV / 72 / 2610.03391
Native Action-Prior Learning from Videos for World Action Models
基于视频的原生动作先验学习用于世界动作模型
Abstract
World action models integrate future visual dynamics with robot action prediction, but their scalability remains limited by the need for action-annotated robot trajectories. Observation-only videos contain rich evidence about interaction dynamics, but existing approaches typically use them either to pretrain visual representations that must later be adapted for control, or to infer latent actions that are subsequently grounded to robot commands. We present NAVA-WAM, which introduces native action-prior learning by directly pretraining the action policy from observation-only videos, avoiding indirect representation-to-control transfer or a separate latent-action model. Our training consists of two stages. First, we pretrain on observation-only videos, where future-video flow-matching supervision over visual transitions is propagated through transition-structured joint attention to optimize the Action-DiT and learn action-relevant priors. Second, we use action-labeled demonstrations to post-train the Action-DiT for robot control through joint video--action flow matching, while asymmetric attention decouples the visual branch from iterative action denoising and enables efficient action-only inference. Extensive experiments show that NAVA-WAM consistently outperforms prior approaches under both in-distribution and out-of-distribution settings, while demonstrating strong action-label efficiency and effective real-robot generalization. These results establish native action-prior learning as an effective approach to directly pretrain action policies from observation-only videos, providing a scalable path beyond action-labeled robot data.
Chinese Translation
世界动作模型将未来视觉动态与机器人动作预测相结合，但其可扩展性仍受限于对动作标注机器人轨迹的需求。仅观察的视频包含关于交互动态的丰富证据，但现有方法通常要么用于预训练视觉表示，随后必须适应控制，要么用于推断潜在动作，随后再将其与机器人命令结合。我们提出了NAVA-WAM，它通过直接从仅观察的视频中预训练动作策略，引入了原生动作先验学习，避免了间接的表示到控制的转移或单独的潜在动作模型。我们的训练分为两个阶段。首先，我们在仅观察的视频上进行预训练，其中通过过渡结构的联合注意力传播未来视频流匹配监督，以优化Action-DiT并学习与动作相关的先验。其次，我们使用带有动作标签的演示来后训练Action-DiT，以通过联合视频-动作流匹配进行机器人控制，同时不对称注意力将视觉分支与迭代动作去噪解耦，从而实现高效的仅动作推断。大量实验表明，NAVA-WAM在分布内和分布外设置下均持续优于先前的方法，同时展示了强大的动作标签效率和有效的真实机器人泛化能力。这些结果确立了原生动作先验学习作为一种有效的方法，能够直接从仅观察的视频中预训练动作策略，为超越动作标注机器人数据提供了一条可扩展的路径。
cs.CV / 73 / 2610.03400
Beyond Entropy: Self-Diagnostic Multi-Role Token Optimization for Video Reasoning
超越熵：用于视频推理的自我诊断多角色令牌优化
Abstract
Reinforcement learning with verifiable rewards has substantially advanced multimodal reasoning, yet it remains fundamentally limited by ambiguous token-level credit assignment. While high-entropy token heuristics encourage possibility exploration, naively extending them to video reasoning tends to induce lengthy reasoning, as the model becomes overly reliant on high-entropy visual activations. Alternative approaches that rely on counterfactual-based visual token localization for credit assignment also tend to over-prioritize visual exploration at the expense of decisive reasoning cues for answer derivation, thereby exacerbating the interference from spurious visual nuances. Moreover, these methods employ static counterfactual strategies that fail to co-evolve with the policy during training. In this paper, we introduce DyCPO, a co-evolutionary framework that jointly optimizes reliable token selection and adaptive counterfactual intervention. It constructs a multi-role dependence metric to balance visual exploration and answer-relevance mining in token-wise contrastive learning, while suppressing exploration-only filler tokens and spurious visual noise. Rather than relying on static counterfactual priors, DyCPO dynamically derives counterfactual signals from the model's own successful and failed rollouts, enabling self-diagnostic analysis and co-evolution of the optimization objective with the policy. Extensive experiments on complex video reasoning and general video understanding benchmarks demonstrate consistent performance improvements, establishing DyCPO as a robust token-level credit assignment paradigm for multimodal reinforcement learning.
Chinese Translation
可验证奖励的强化学习在多模态推理方面取得了显著进展，但在令牌级别的信用分配上仍然存在根本性的局限性。尽管高熵令牌启发式方法鼓励可能性的探索，但将其简单扩展到视频推理往往会导致推理过程冗长，因为模型过于依赖高熵的视觉激活。依赖于反事实视觉令牌定位进行信用分配的替代方法也往往过于优先考虑视觉探索，而牺牲了用于答案推导的决定性推理线索，从而加剧了来自虚假视觉细节的干扰。此外，这些方法采用静态反事实策略，在训练过程中未能与策略共同演化。在本文中，我们提出了DyCPO，一个共同演化框架，联合优化可靠的令牌选择和自适应反事实干预。它构建了一个多角色依赖度量，以平衡视觉探索和答案相关性挖掘，在令牌级对比学习中，同时抑制仅用于探索的填充令牌和虚假视觉噪声。DyCPO并不依赖于静态反事实先验，而是动态地从模型自身的成功和失败回合中推导反事实信号，使得自我诊断分析和优化目标与策略的共同演化成为可能。在复杂视频推理和一般视频理解基准上的广泛实验表明，性能持续改善，确立了DyCPO作为多模态强化学习中稳健的令牌级信用分配范式。
cs.CV / 74 / 2610.03403
ForestQuery: Boundary-Aware and Spatially Anchored Query Learning for Unified Forest Point Cloud Segmentation
ForestQuery：边界感知与空间锚定的统一森林点云分割查询学习
Abstract
Forest point cloud segmentation is fundamental for fine-grained 3D forest scene understanding, yet remains challenging due to irregular tree structures, severe occlusions, density variations, and ambiguous instance boundaries. Recent query-based forest segmentation methods have shown promise for unified semantic and instance prediction, but they still insufficiently exploit forest-specific spatial structure and account for boundary uncertainty. In this paper, we propose ForestQuery, a boundary-aware and spatially anchored query learning framework for unified forest point cloud segmentation. ForestQuery enhances instance and semantic query learning through two complementary designs. Specifically, boundary uncertainty is explicitly modeled to guide reliable instance query construction and modulate query optimization through adaptive loss reweighting. Meanwhile, spatially anchored semantic query enhancement (SA-SQE) introduces learnable 3D anchors encoding forest vertical stratification priors to enrich semantic queries with explicit spatial references. We evaluate ForestQuery on multiple public forest point cloud benchmarks and a self-collected annotated real-world dataset. Extensive experiments demonstrate consistent improvements in both individual-tree segmentation and semantic segmentation across diverse forest scenes. Code and data are publicly available at https://zhan994.github.io/ForestQuery
Chinese Translation
森林点云分割是细粒度三维森林场景理解的基础，但由于树木结构不规则、严重遮挡、密度变化和模糊的实例边界等原因，仍然面临挑战。近期基于查询的森林分割方法在统一语义和实例预测方面显示出良好的前景，但仍未充分利用森林特有的空间结构，也未能有效处理边界不确定性。本文提出了ForestQuery，一个边界感知和空间锚定的查询学习框架，用于统一森林点云分割。ForestQuery通过两种互补设计增强了实例和语义查询学习。具体而言，边界不确定性被明确建模，以指导可靠的实例查询构建，并通过自适应损失重加权调节查询优化。同时，空间锚定语义查询增强（SA-SQE）引入可学习的三维锚点，编码森林垂直分层先验，以显式空间参考丰富语义查询。我们在多个公共森林点云基准和自收集的标注真实世界数据集上评估了ForestQuery。大量实验表明，在多样化的森林场景中，个体树木分割和语义分割均取得了一致的改进。代码和数据可在 https://zhan994.github.io/ForestQuery 获取。
cs.CV / 75 / 2610.03423
OuroReward: Sequential Reward Scheduling for Reinforcement Learning in Text-to-3D Generation
OuroReward：文本到三维生成中的强化学习序列奖励调度
Abstract
Reinforcement learning (RL) for Text-to-3D (T23D) generation requires optimization across multiple quality dimensions such as semantic alignment and texture clarity. Existing methods typically optimize these dimensions simultaneously through multiple reward aggregation, without explicitly modeling inter-dimension dependencies. This can cause imbalanced optimization and persistent interference among conflicting dimensions. To address this limitation, we propose OuroReward, an interference-aware sequential reward scheduling strategy for T23D RL. OuroReward first estimates pairwise dependencies among dimensions and constructs a cyclic optimization path that minimizes cumulative interference. By incorporating the tail-to-head dependency, the cycle captures global compatibility across the entire schedule. Then, OuroReward converts the cycle into a one-pass sequence, and starts optimization from the dimension with the lowest aggregate interference. Rather than assigning a fixed optimization budget to each dimension-wise reward, training adaptively determines when to advance to the next reward according to the remaining optimization headroom of the current one. We further introduce AdaSelect, an adaptive prompt selection strategy that identifies reliable and informative prompts aligned with the model's current capability. By focusing policy updates on these prompts, AdaSelect effectively improves training stability. Extensive experiments across different T23D models and RL algorithms demonstrate that our framework consistently improves generation quality across multiple dimensions.
Chinese Translation
文本到三维生成（Text-to-3D, T23D）的强化学习（Reinforcement Learning, RL）需要在多个质量维度上进行优化，例如语义对齐和纹理清晰度。现有方法通常通过多重奖励聚合同时优化这些维度，而没有明确建模维度间的依赖关系。这可能导致优化不平衡和相互冲突的维度之间的持续干扰。为了解决这一局限性，我们提出了OuroReward，一种针对T23D RL的干扰感知序列奖励调度策略。OuroReward首先估计维度之间的成对依赖关系，并构建一个循环优化路径，以最小化累积干扰。通过结合尾到头的依赖关系，该循环捕捉了整个调度的全局兼容性。然后，OuroReward将循环转换为一次性序列，并从干扰总和最低的维度开始优化。与其为每个维度的奖励分配固定的优化预算，不如根据当前奖励的剩余优化空间自适应地决定何时推进到下一个奖励。我们进一步引入了AdaSelect，一种自适应提示选择策略，能够识别与模型当前能力相一致的可靠且信息丰富的提示。通过将策略更新集中在这些提示上，AdaSelect有效提高了训练的稳定性。针对不同T23D模型和RL算法的广泛实验表明，我们的框架在多个维度上持续提高了生成质量。
cs.CV / 76 / 2610.03439
Depth Hypothesis Guided Iterative Refinement for Event-Image Monocular Depth Estimation
基于深度假设引导的事件图像单目深度估计迭代优化
Abstract
Event cameras hold excellent dynamic properties, showing great potential for monocular depth estimation (MDE). However, existing methods mainly improve performance by optimizing contextual features, but still struggle with the ill-posed and nonlinear nature of direct full-depth regression. In this paper, we propose HypoDepth, the first event-image monocular depth iterative refinement framework. By introducing a discrete Depth Hypothesis Volume (DHV), we transform the depth regression problem into a constrained depth search task. Specifically, we construct a 3D cost volume between the DHV features and contextual features and perform a multi-scale correlation search to guide stable residual optimization. This lightweight cost volume enables efficient global-to-local refinement across multi-resolution. Our method outperforms existing approaches on DSEC and MVSEC with state-of-the-art results and strong zero-shot generalization. Meanwhile, our tiny model achieves an excellent balance between accuracy and efficiency, enabling real-time performance on resource-limited devices.
Chinese Translation
事件相机具有出色的动态特性，展现了在单目深度估计（MDE）中的巨大潜力。然而，现有方法主要通过优化上下文特征来提高性能，但仍然在直接全深度回归的病态和非线性特性上面临挑战。本文提出了HypoDepth，这是第一个事件图像单目深度迭代优化框架。通过引入离散深度假设体积（Depth Hypothesis Volume, DHV），我们将深度回归问题转化为一个受限的深度搜索任务。具体而言，我们在DHV特征和上下文特征之间构建了一个三维代价体积，并执行多尺度相关性搜索以引导稳定的残差优化。这个轻量级的代价体积使得在多分辨率下实现高效的全局到局部优化成为可能。我们的方法在DSEC和MVSEC数据集上超越了现有方法，取得了最先进的结果和强大的零样本泛化能力。同时，我们的小型模型在准确性和效率之间实现了出色的平衡，使其能够在资源有限的设备上实现实时性能。
cs.CV / 77 / 2610.03441
ChromaGS: Text-Driven Semantic Editing of 4D Gaussian Avatars
ChromaGS：基于文本的4D高斯头像语义编辑
Abstract
We present ChromaGS, a method for real-time, language-guided color editing of animatable 3D Gaussian head avatars. Given a trained animatable avatar, users can instantly modify the color of semantic regions through natural language, with edits applied at render time and no retraining required. Our key insight is to augment each Gaussian primitive with learned soft assignments to semantic regions and decompose colors into region-level base colors and Gaussian-level residuals. This decomposition enables coherent color transfer: modifying a region's base color propagates naturally through all associated Gaussians while preserving fine appearance details encoded in residuals. A two-stage language pipeline translates text instructions into target colors, supporting both absolute specifications and relative adjustments. Unlike generative editing methods that may introduce unintended modifications, our approach provides deterministic, precisely localized semantic control. Experiments demonstrate faithful appearance preservation and intuitive interaction across diverse subjects. Project page and code are available at: https://a-canela.github.io/chromags/
Chinese Translation
我们提出了ChromaGS，一种用于实时、语言引导的可动画3D高斯头像的颜色编辑方法。给定一个经过训练的可动画头像，用户可以通过自然语言即时修改语义区域的颜色，编辑在渲染时应用，无需重新训练。我们的关键见解是为每个高斯原语增加学习到的软分配到语义区域，并将颜色分解为区域级基础颜色和高斯级残差。这种分解使得一致的颜色传递成为可能：修改区域的基础颜色会自然地传播到所有相关的高斯，同时保留编码在残差中的细微外观细节。一个两阶段的语言处理管道将文本指令翻译为目标颜色，支持绝对规格和相对调整。与可能引入意外修改的生成编辑方法不同，我们的方法提供确定性、精确定位的语义控制。实验表明，在不同主题上实现了真实的外观保留和直观的交互。项目页面和代码可在以下链接获取：https://a-canela.github.io/chromags/
cs.CV / 78 / 2610.03445
Corrupted but Correct: Why Vision-Language Models Lie to Themselves Internally
被破坏但仍正确：为何视觉-语言模型在内部自我欺骗
Abstract
A targeted adversarial perturbation can drive a vision-language model's (VLM's) teacher-forced training loss for a fixed target caption to near zero, yet the same model, allowed to generate freely, produces the original, correct description with no trace of the target. We call this dissociation the train/inference gap, and give it a precise mechanistic account on Qwen2.5-VL-7B-Instruct using a controlled two-stage PGD attack on 200 held-out COCO images. First, we show that image-level pixel statistics, including a correctly re-implemented, texture-based attackability measure from the CNN robustness literature, have essentially no predictive power over which images are corrupted (best predictor r=-0.050, p=0.484; ridge regression R^2=0.069). Second, using the logit lens, we localise the gap to a single autoregressive step: the rank of the target token, conditioned on the correct first token already being generated, is fixed at exactly 3,488 out of 152,064 vocabulary entries for every image and every condition, with zero variance. Third, tracking target-token rank across all 28 LLM decoder layers reveals that the visual encoder corrupts every image's representation by a comparable margin regardless of eventual outcome, but the language model decoder then differentially arbitrates: amplifying the corrupted signal for susceptible images and actively suppressing it, past its clean-image baseline, for resistant ones (p<0.001, rank-biserial r=0.579). A linear probe on the merger hidden state separates these two outcomes with AUC=0.858, though we flag a circularity concern in this estimate. Together these results argue that adversarial robustness in autoregressive VLMs is substantially a property of the language decoder's prior, not the visual encoder, with direct implications for where faithfulness evaluations and defenses for deployed VLM systems should be targeted.
Chinese Translation
有针对性的对抗扰动可以使视觉-语言模型（VLM）的教师强制训练损失在固定目标描述下接近于零，然而同一模型在自由生成时却产生了原始的、正确的描述，完全没有目标的痕迹。我们将这种脱节称为训练/推理差距，并在 Qwen2.5-VL-7B-Instruct 上给出了精确的机制解释，使用了对 200 张保留的 COCO 图像进行的受控两阶段 PGD 攻击。首先，我们展示了图像级像素统计，包括来自 CNN 鲁棒性文献中正确重新实现的基于纹理的攻击性度量，对哪些图像被破坏几乎没有预测能力（最佳预测者 r=-0.050，p=0.484；岭回归 R^2=0.069）。其次，使用对数几率视角，我们将差距定位到单个自回归步骤：在已生成正确的第一个标记的条件下，目标标记的排名在每张图像和每种条件下都固定为 152,064 个词汇条目中的 3,488，且方差为零。第三，跟踪所有 28 层 LLM 解码器中的目标标记排名揭示，视觉编码器以相似的幅度破坏每张图像的表示，无论最终结果如何，但语言模型解码器随后进行差异性裁决：对易受攻击的图像放大被破坏的信号，而对抗拒的图像则主动抑制该信号，超出其干净图像的基线（p<0.001，排名双序相关 r=0.579）。对合并隐藏状态的线性探测将这两种结果分开，AUC=0.858，尽管我们对该估计存在循环性问题。综合这些结果表明，自回归 VLM 的对抗鲁棒性在很大程度上是语言解码器先验的特性，而非视觉编码器，这对部署 VLM 系统的忠实性评估和防御策略的目标具有直接影响。
cs.CV / 79 / 2610.03467
Preserving Anatomical Continuity: Three-Stage Pipeline for Colon Segmentation in 3D Abdominal CT Scans
保持解剖连续性：三阶段管道用于三维腹部CT扫描中的结肠分割
Abstract
Accurate colon segmentation from CT images is essential for colorectal disease analysis, yet deep learning based methods often produce disconnected predictions due to complex anatomy. This study introduces a three-stage, topology-preserving segmentation pipeline to address this issue. The first stage performs initial deep learning-based segmentation, followed by centreline bridging to reconnect disjoint regions and a reconstruction stage to refine continuity. Evaluations on TotalSegmentator and RAOS datasets using overlap, distance and topology-based metrics demonstrate improved structural consistency while maintaining segmentation accuracy. The proposed method enhances topological integrity, enabling more reliable colon segmentation for clinical and research applications.
Chinese Translation
从CT图像中准确分割结肠对于结直肠疾病分析至关重要，但基于深度学习的方法常常由于复杂的解剖结构而产生不连贯的预测。本研究提出了一种三阶段的、保持拓扑结构的分割管道来解决这一问题。第一阶段进行初步的基于深度学习的分割，随后通过中心线连接来重新连接不相连的区域，最后进行重建阶段以细化连续性。在TotalSegmentator和RAOS数据集上的评估使用重叠、距离和基于拓扑的指标，显示出在保持分割准确性的同时改善了结构一致性。所提出的方法增强了拓扑完整性，使得结肠分割在临床和研究应用中更加可靠。
cs.CV / 80 / 2610.03468
A Vision-Language Model (VLM)-based Pipeline for End-to-End Procedural Modeling of Field-Grown Maize from Point Clouds
基于视觉-语言模型（VLM）的端到端田间生长玉米的程序建模管道
Abstract
Editable 3D models of field-grown crops support high-throughput phenotyping and in silico breeding trials, but building them from scanned point clouds requires organ-level segmentation and fitting. Procedural generators can turn an organ-level parameter set into an analysis-suitable 3D model, but obtaining that set requires hours of manual tuning per plant or segmentation models trained on species-specific labels. We present an automated pipeline that reconstructs procedural maize models from raw 3D point clouds without manual tuning or species-specific training data. A multimodal vision-language model (VLM) annotates leaf midlines in rendered orthographic views. Deterministic geometric algorithms back-project the annotations onto the point cloud, merge them into 3D leaves by cross-view consensus, and grow the midlines to full blades on an orientation-weighted surface graph. Measured organ parameters populate a plant descriptor for a Non-Uniform Rational B-Spline (NURBS)-based procedural model generator. Each leaf surface is then refined against its scan points by differentiable NURBS fitting. The pipeline reached a median whole-plant Chamfer distance of 5.4 mm on 100 genotypically diverse field-grown maize plants from the MaizeField3D dataset. The reconstructions were closer to the scans than those of an earlier semi-automated pipeline based on manual annotations. The pipeline recovered 1,017 of 1,023 (99.4%) curated reference leaves at an intersection-over-union of at least 0.5 without using those labels as input. These results show that VLM annotations become usable organ-level measurements when downstream geometric stages can correct them. This makes automated generation of editable 3D plant assets feasible at the scale of modern phenotyping experiments.
Chinese Translation
可编辑的田间生长作物的三维模型支持高通量表型分析和计算机模拟育种试验，但从扫描的点云构建这些模型需要进行器官级分割和拟合。程序生成器可以将器官级参数集转换为适合分析的三维模型，但获取该参数集需要每株植物数小时的手动调节或基于特定物种标签训练的分割模型。我们提出了一种自动化管道，可以从原始三维点云重建程序玉米模型，而无需手动调节或特定物种的训练数据。多模态视觉-语言模型（VLM）在渲染的正交视图中标注叶片中线。确定性几何算法将标注反投影到点云上，通过交叉视图共识将其合并为三维叶片，并在加权方向的表面图上将中线生长为完整的叶片。测量的器官参数填充了基于非均匀有理B样条（NURBS）的程序模型生成器的植物描述符。然后，通过可微分的NURBS拟合对每个叶片表面进行精细化处理。该管道在MaizeField3D数据集中对100株基因型多样的田间生长玉米植物的整体植物Chamfer距离达到了5.4毫米的中位数。重建结果比早期基于手动标注的半自动化管道更接近扫描结果。该管道在没有使用这些标签作为输入的情况下，恢复了1,017片中的1,023片（99.4%）经过整理的参考叶，交并比至少为0.5。这些结果表明，当下游几何阶段能够纠正VLM标注时，这些标注变得可用作为器官级测量。这使得在现代表型实验的规模下，自动生成可编辑的三维植物资产成为可能。
cs.CV / 81 / 2610.03473
UniDynamics: Event-RGB Fusion for Unified Future 4D Dynamic Scene Generation
UniDynamics：基于事件-RGB融合的统一未来4D动态场景生成
Abstract
We propose UniDynamics, a diffusion-based framework for future 4D dynamic scenes (RGB, depth, and optical flow) generation from a single event-RGB pair, without requiring long histories or control priors as in existing methods, while explicitly modeling future motion fields. The core idea is to leverage event streams to offer an alternative motion prior for single-RGB extrapolation, and to enforce geometric and motion constraints throughout generation via multimodal modeling. Specifically, we design an Event Latent Enhancement (ELE) module to align and enhance event latents into diffusion-injectable conditioning features, providing robust initial motion priors and reliable texture/structure cues. We further introduce a Perceptual Dynamics Space (PDS) embedded in the multi-scale U-Net, which decouples and adaptively interacts depth and flow while continuously feeding back constraints to appearance features, improving geometric-motion consistency for physically plausible and spatiotemporally coherent prediction. Experiments on VKitti2 and DSEC demonstrate state-of-the-art performance, producing high-quality, temporally coherent, and 4D-consistent future predictions, especially under challenging high-speed motion blur.
Chinese Translation
我们提出了UniDynamics，这是一种基于扩散的框架，用于从单个事件-RGB对生成未来4D动态场景（RGB、深度和光流），无需像现有方法那样依赖长时间历史或控制先验，同时明确建模未来运动场。其核心思想是利用事件流为单RGB外推提供替代运动先验，并通过多模态建模在生成过程中强制执行几何和运动约束。具体而言，我们设计了一个事件潜在增强（Event Latent Enhancement, ELE）模块，以对齐和增强事件潜在特征，使其成为可注入扩散的条件特征，从而提供稳健的初始运动先验和可靠的纹理/结构线索。此外，我们进一步引入了嵌入在多尺度U-Net中的感知动态空间（Perceptual Dynamics Space, PDS），该空间解耦并自适应地交互深度和光流，同时持续反馈约束到外观特征，提高了几何-运动一致性，以实现物理上合理和时空一致的预测。在VKitti2和DSEC上的实验表明，我们的方法达到了最先进的性能，生成高质量、时间一致且4D一致的未来预测，尤其是在具有挑战性的高速运动模糊条件下。
cs.CV / 82 / 2610.03474
Fed-ADApt: Federated Anytime Depth Adaptation for Resource-Aware Medical Image Segmentation
Fed-ADApt：面向资源感知的医学图像分割的联邦随时深度适应
Abstract
Federated learning (FL) enables collaborative training of medical image segmentation models without sharing raw patient data, yet existing approaches assume a homogeneous compute budget across institutions, limiting participation of low-resource sites. We propose Fed-ADApt, a depth-adaptive federated framework for UNet-based segmentation that jointly addresses low-compute training and inference. Fed-ADApt integrates multi-depth supervision with hierarchical depth-wise aggregation, allowing each site to train according to its local compute budget while contributing to a global model that supports dynamic depth selection at deployment. We evaluated Fed-ADApt on multi-site 2D retinal fundus disc segmentation and 3D brain tumor segmentation. Across both tasks, federated collaboration substantially improves robustness under domain shift. Fed-ADApt matched the full-resource FedAvg performance in 3D and achieved competitive 2D performance with a 4.7% average Dice reduction, while reducing average inference cost by 19.5% in 3D and 34.5% in 2D and substantially reducing training cost by 98% at the most constrained sites. Importantly, Fed-ADApt enables low-resource institutions that cannot train full-capacity models to participate in federations while maintaining competitive global performance under a favorable accuracy to efficiency trade-off. By considering training and inference compute budgets, Fed-ADApt provides a practical and equitable solution for federated medical image segmentation across heterogeneous clinical and edge-enabled imaging environments.
Chinese Translation
联邦学习（FL）使医学图像分割模型能够在不共享原始患者数据的情况下进行协作训练，但现有方法假设各机构之间的计算预算是均匀的，从而限制了低资源场所的参与。我们提出了Fed-ADApt，这是一种基于UNet的深度自适应联邦框架，旨在共同解决低计算训练和推理问题。Fed-ADApt结合了多深度监督与分层深度聚合，使每个站点能够根据其本地计算预算进行训练，同时为支持动态深度选择的全球模型做出贡献。我们在多站点的2D视网膜眼底盘分割和3D脑肿瘤分割任务上评估了Fed-ADApt。在这两个任务中，联邦协作显著提高了在领域转移下的鲁棒性。Fed-ADApt在3D任务中达到了全资源FedAvg的性能，并在2D任务中实现了具有竞争力的表现，平均Dice系数降低4.7%，同时在3D和2D中分别减少了19.5%和34.5%的平均推理成本，并在最受限的站点大幅降低了98%的训练成本。重要的是，Fed-ADApt使得无法训练全容量模型的低资源机构能够参与联邦，同时在准确性与效率之间保持竞争力的全球性能。通过考虑训练和推理的计算预算，Fed-ADApt为异构临床和边缘启用成像环境中的联邦医学图像分割提供了一个实用且公平的解决方案。
cs.CV / 83 / 2610.03510
Weave Forcing: Compositional Memory Routing for Interactive Long Video Generation
编织强制：用于交互式长视频生成的组合记忆路由
Abstract
Recent advances in autoregressive video generation have improved temporal consistency over extended durations, yet interactive storytelling requires more than continuous scene extension: a new shot may combine characters and backgrounds from different historical shots. Whole prompt retrieval can overlook the distinct reference needs of individual components, while directly combining all historical memories may introduce unrelated visual content. To address these problems, we present Weave Forcing, a training-free framework for compositional memory reuse in interactive long video generation. First, we use an LLM for semantic slot routing to decompose user prompts into character and background descriptions and explicitly select suitable historical references for each component. To isolate the required content, masked memory weaving uses contrasting attention maps conditioned on semantic slots to construct refined semantic masks, selectively exposing relevant tokens from compressed historical KV memories to guide the generation of the current shot. We further introduce coverage adaptive RoPE to adjust temporal offsets and memory retention according to no, partial, or full reference coverage, addressing visual artifacts observed when incomplete historical references are positioned close to the current generation. Extensive experiments demonstrate that Weave Forcing improves cross-shot subject and background consistency while maintaining competitive visual quality and text alignment.
Chinese Translation
近年来，自回归视频生成的进展提高了长时间段内的时间一致性，但交互式叙事不仅需要连续场景的扩展：新的镜头可能会结合来自不同历史镜头的角色和背景。整体提示检索可能会忽视个别组件的独特引用需求，而直接组合所有历史记忆可能会引入无关的视觉内容。为了解决这些问题，我们提出了编织强制（Weave Forcing），这是一个无需训练的框架，用于交互式长视频生成中的组合记忆重用。首先，我们使用大型语言模型（LLM）进行语义槽路由，将用户提示分解为角色和背景描述，并明确选择每个组件的合适历史参考。为了隔离所需内容，掩蔽记忆编织利用基于语义槽的对比注意力图构建精细的语义掩码，选择性地暴露来自压缩历史键值（KV）记忆的相关标记，以指导当前镜头的生成。我们进一步引入覆盖自适应的RoPE（Rotary Position Embedding）来根据无、部分或完全引用覆盖调整时间偏移和记忆保留，解决了在不完整历史引用靠近当前生成时观察到的视觉伪影。大量实验表明，编织强制提高了跨镜头主体和背景的一致性，同时保持了竞争性的视觉质量和文本对齐。
cs.CV / 84 / 2610.03512
ProgressNet: Sketching and Prompting with a Frozen Text-to-Image Model
ProgressNet：使用冻结的文本到图像模型进行草图绘制和提示
Abstract
Humans draw progressively: a few strokes, a look at the result, a stroke erased, a prompt revised. Image generators do not work this way. They typically take a finished sketch and produce the image in a single pass, so every edit starts the picture again, and the models that do keep state across turns are driven by text, cannot take a stroke, and are too slow to draw with. We present ProgressNet, a training-free framework that lets a frozen text-to-image model follow a drawing session as it unfolds: strokes are added and erased, the prompt is revised, and the image keeps up at about a second per turn. It needs no new parameters because the frozen model already has what a progressive generator needs, a pathway through which the previous turn can be remembered, layers that can carry appearance forward without freezing structure, and an internal signal of how far to trust an unfinished sketch; three inference-time mechanisms (Previous-Concept Memory, Layer-Selective K/V Injection and Banded Adaptive Control) use each in turn. As a sketch fills in, every existing method degrades, the FID of the FLUX+ControlNet baseline doubling between 10% and 100% completion on FS-COCO, while ProgressNet's barely moves; it maintains strong fidelity and progressive coherence across three sketch domains and is preferred by users over five competitors, most widely on erasure.
Chinese Translation
人类的绘画过程是渐进的：几笔画，查看结果，擦去一笔，修改提示。图像生成器并不是这样工作的。它们通常需要一个完成的草图，并在一次性处理的过程中生成图像，因此每次编辑都需要重新开始，而那些在多个回合中保持状态的模型则是由文本驱动的，无法接受单独的笔画，并且绘制速度过慢。我们提出了ProgressNet，一个无须训练的框架，允许一个冻结的文本到图像模型在绘画会话展开时进行跟随：笔画可以添加和擦除，提示可以被修改，图像以每回合约一秒的速度跟进。它不需要新的参数，因为冻结的模型已经具备渐进生成器所需的条件：一个可以记住前一回合的路径，一些可以在不冻结结构的情况下传递外观的层，以及一个内部信号，用于判断对未完成草图的信任程度；三个推理时机制（Previous-Concept Memory、Layer-Selective K/V Injection 和 Banded Adaptive Control）依次使用这些条件。随着草图的逐步填充，现有的每种方法都会退化，FLUX+ControlNet基线在FS-COCO上从10%完成到100%完成时的FID值翻倍，而ProgressNet几乎没有变化；它在三个草图领域中保持了强大的保真度和渐进一致性，并且用户在五个竞争者中更倾向于选择它，尤其是在擦除方面。
cs.CV / 85 / 2610.03522
Feedforward Novel View Synthesis for Heterogeneous Cameras
异构相机的前馈新视图合成
Abstract
Feed-forward novel view synthesis has recently shown promising results from sparse posed images, but most existing methods assume that context and target views share a fixed camera family. This homogeneous-camera assumption breaks in practical multi-sensor systems, where perspective, fisheye, and panoramic cameras may coexist and where the target projection may be unseen during training. We study feed-forward NVS across heterogeneous central cameras and identify a key ambiguity introduced by tokenization: a visual token aggregates a projection-dependent bundle of pixel rays, while existing camera encodings mainly expose absolute rays or token-center relations. To address this, we combine token-center relative Camera Positional Encodings and proposed local raymaps, a token-level representation that explicitly describes the intra-patch ray distribution summarized by each token. We further propose projection-aware 2D RoPE, which replaces raw image-grid coordinates with ray-induced angular coordinates so that relative positional reasoning is aligned across camera projections. Together, these components treat diverse cameras as calibrated samplings of a shared ray space rather than separate visual domains. On ScanNet++ with heterogeneous-camera system, our method improves over camera-conditioned baselines under mixed-camera evaluation and demonstrates zero-shot generalization to panoramic views.
Chinese Translation
前馈新视图合成最近在稀疏姿态图像中显示出良好的结果，但大多数现有方法假设上下文和目标视图共享固定的相机系列。这种同质相机的假设在实际的多传感器系统中被打破，其中透视、鱼眼和全景相机可能共存，并且目标投影在训练期间可能是不可见的。我们研究了异构中心相机下的前馈新视图合成，并识别出由标记化引入的一个关键模糊性：视觉标记聚合了依赖于投影的像素光线束，而现有的相机编码主要暴露绝对光线或标记中心关系。为了解决这个问题，我们结合了标记中心相对相机位置编码和提出的局部光线图（local raymaps），这是一种标记级表示，明确描述了每个标记总结的补丁内光线分布。我们进一步提出了投影感知的二维相对位置编码（2D RoPE），该方法用光线诱导的角度坐标替换原始图像网格坐标，从而使相对位置推理在相机投影之间保持一致。通过这些组件，我们将不同的相机视为共享光线空间的标定采样，而不是独立的视觉领域。在异构相机系统的ScanNet++上，我们的方法在混合相机评估下优于基于相机条件的基线，并展示了对全景视图的零样本泛化能力。
cs.CV / 86 / 2610.03543
DuoMatching: Joint-Marginal Distribution Matching for Few-Step Video Generation
DuoMatching：用于少步视频生成的联合边际分布匹配
Abstract
Streaming video generation has benefited from distribution matching distillation (DMD), which matches the joint distribution of video frames to a video teacher's approximation of the real video distribution. Although this joint matching mitigates drift during autoregressive rollouts, limitations remain in visual quality and semantic alignment. To address these limitations, we propose DuoMatching, a distribution matching framework that approximates the real video distribution through a unified joint-marginal formulation. On top of existing joint matching formulations, the additional marginal matching objective provides dedicated frame-level supervision from an image generator, transferring complementary visual and semantic priors from it. To apply this frame-level supervision in video generation, we introduce LatentBridge to resolve the latent representation mismatch between the video student and the image teacher. Latent Variation Sampling further distributes such frame-level supervision across distinct temporal segments, reducing redundancy. Experiments demonstrate that DuoMatching improves visual quality, composition, and semantic alignment while largely preserving motion dynamics. Human evaluations show overall preference rates above 80% against all evaluated baselines. The project page is available at https://johnzhan2023.github.io/DuoMatching/.
Chinese Translation
流媒体视频生成受益于分布匹配蒸馏（DMD），该方法将视频帧的联合分布与视频教师对真实视频分布的近似进行匹配。尽管这种联合匹配在自回归生成过程中减轻了漂移，但在视觉质量和语义对齐方面仍存在局限性。为了解决这些问题，我们提出了DuoMatching，这是一种通过统一的联合边际形式来逼近真实视频分布的分布匹配框架。在现有的联合匹配形式基础上，额外的边际匹配目标提供了来自图像生成器的专门帧级监督，从中转移互补的视觉和语义先验。为了在视频生成中应用这种帧级监督，我们引入了LatentBridge，以解决视频学生与图像教师之间的潜在表示不匹配问题。潜在变异采样进一步将这种帧级监督分布到不同的时间段，从而减少冗余。实验表明，DuoMatching在视觉质量、构图和语义对齐方面有所改善，同时在很大程度上保持了运动动态。人类评估显示，相较于所有评估的基线，整体偏好率超过80%。项目页面可访问：https://johnzhan2023.github.io/DuoMatching/
cs.CV / 87 / 2610.03577
Rethinking What to Cache in Few-Step Diffusion Transformers: Solver-Aware Target Selection
重新思考在少步扩散变换器中缓存什么：基于求解器的目标选择
Abstract
Diffusion Transformers (DiTs) can generate high-quality images and videos, but generating each sample requires multiple costly DiT forward passes. Two common ways to accelerate DiT sampling are step distillation, which reduces the number of sampling steps, and caching, which skips some DiT evaluations by reusing a tensor computed at an earlier step. Most caching methods decide in advance which tensor to reuse. After distillation, adjacent sampling steps are farther apart. Reusing a tensor across this larger gap introduces more error, so choosing what to cache becomes especially important. We therefore introduce AutoTarget, a method that chooses the cached tensor for a given model, solver, and reuse schedule. AutoTarget uses a small set of runs without cache reuse to measure the error caused by reusing each candidate tensor, then selects the candidate with the lowest error. We also analyze how an error at one reuse step affects the final sample. For Euler sampling, we identify cache targets that produce the same trajectory and show why a stored solver update may not. Experiments on distilled image and video DiTs show that the best cache target changes with the model, image resolution, and solver. AutoTarget reduces DiT evaluations and retained cache storage. Generation quality remains close to the corresponding uncached run. On the tested PixArt-LCM and FLUX.1-schnell settings, its calibration ranking matches the ranking from held-out cached runs. To help others reproduce the method, we provide its core implementation on GitHub at https://github.com/wali1024-offical/AutoTarget.
Chinese Translation
扩散变换器（Diffusion Transformers, DiTs）能够生成高质量的图像和视频，但生成每个样本需要多个耗时的 DiT 前向传播。加速 DiT 采样的两种常见方法是步骤蒸馏（step distillation），它减少了采样步骤的数量，以及缓存（caching），它通过重用在早期步骤计算的张量来跳过一些 DiT 评估。大多数缓存方法在事先决定重用哪个张量后进行操作。经过蒸馏后，相邻的采样步骤之间的距离变得更远。在这个更大的间隔中重用张量会引入更多的误差，因此选择缓存的内容变得尤为重要。因此，我们提出了 AutoTarget，这是一种为给定模型、求解器和重用计划选择缓存张量的方法。AutoTarget 使用一小组不进行缓存重用的运行来测量重用每个候选张量所导致的误差，然后选择误差最低的候选张量。我们还分析了一个重用步骤的误差如何影响最终样本。对于欧拉采样（Euler sampling），我们识别出产生相同轨迹的缓存目标，并展示了为什么存储的求解器更新可能不会。对蒸馏图像和视频 DiTs 的实验表明，最佳缓存目标会随着模型、图像分辨率和求解器的不同而变化。AutoTarget 减少了 DiT 评估和保留的缓存存储。生成质量仍然接近相应的未缓存运行。在测试的 PixArt-LCM 和 FLUX.1-schnell 设置中，其校准排名与保留缓存运行的排名相匹配。为了帮助他人复现该方法，我们在 GitHub 上提供了其核心实现，网址为 https://github.com/wali1024-offical/AutoTarget。
cs.CV / 88 / 2610.03599
ManifoldSplat: Language-Guided Semantic Shape Editing of 3D Gaussian Head Avatars
ManifoldSplat：基于语言指导的3D高斯头部头像语义形状编辑
Abstract
High-fidelity 3D head avatars have reached near-photorealistic quality. While recent methods enable text-driven manipulation, they struggle to provide fine-grained localized control, often entangling features or lacking geometric consistency. Modifying geometry through natural language currently requires slow per-prompt optimization or compromises identity and rigging. We present ManifoldSplat, the first end-toend framework for language-guided semantic shape editing of animatable 3D Gaussian Splatting avatars reconstructed from monocular videos. By performing edits within the structured FLAME manifold rather than directly optimizing an unstructured Gaussian cloud, we strictly preserve identity and animation. We introduce DeltaRegion, a per-region disentangled Conditional Variational Autoencoder (CVAE) delivering feedforward shape deltas, alongside a refining stage to recover view-consistent details. ManifoldSplat reconstructs and edits an avatar in ~90 seconds on a consumer GPU, rendering at ~800 FPS. Extensive evaluations demonstrate our approach sets a new state-of-the-art in localized prompt alignment, geometric coherence, and identity preservation. Project page and code: https://a-canela.github.io/manifoldsplat/
Chinese Translation
高保真3D头部头像已达到近乎照片级真实的质量。尽管最近的方法实现了基于文本的操控，但它们在提供细粒度局部控制方面仍然存在困难，常常导致特征纠缠或缺乏几何一致性。通过自然语言修改几何形状目前需要缓慢的逐提示优化，或者会妥协身份和绑定。我们提出了ManifoldSplat，这是第一个用于从单目视频重建的可动画3D高斯点云头像的语言指导语义形状编辑的端到端框架。通过在结构化的FLAME流形中进行编辑，而不是直接优化非结构化的高斯云，我们严格保持身份和动画。我们引入了DeltaRegion，这是一个按区域解耦的条件变分自编码器（Conditional Variational Autoencoder, CVAE），提供前馈形状增量，并配备一个精细化阶段以恢复视图一致的细节。ManifoldSplat在消费级GPU上重建和编辑一个头像的时间约为90秒，渲染速度约为800帧每秒。广泛的评估表明，我们的方法在局部提示对齐、几何一致性和身份保持方面设定了新的最先进水平。项目页面和代码：https://a-canela.github.io/manifoldsplat/
cs.CV / 89 / 2610.03617
DEPICT: Scoring Text-to-Image Alignment by Answer Agreement
DEPICT：通过答案一致性评分文本与图像的对齐
Abstract
Image-text alignment is a core problem in computer vision with applications in caption evaluation, hallucination detection, data curation, and the benchmarking of text-to-image (T2I) generators. As T2I models improve, benchmarking has become demanding, requiring metrics capable of finding a series of issues like missing objects, swapped attributes, miscounts, and ignored negations. Recent work addresses this by fine-tuning evaluators on preference data or by prompting a vision-language model, either holistically with the caption or with decomposed verification questions. However, existing approaches fall short: fine-tuned metrics remain bound to one backbone and training distribution; holistic metrics miss fine-grained details; and decomposed metrics rely on a fixed-YES assumption that penalizes faithful images whenever that assumption fails. In contrast, we propose DEPICT, a training-free metric that replaces fixed reference answers with expected agreement between image-based and caption-only answers, weighting questions by how decisively the caption determines them. By replacing fixed references, our agreement rule increases negation accuracy from 19% to 88%. To recover the context lost during decomposition, DEPICT merges this agreement score with a holistic score. We evaluate DEPICT on five benchmarks and eleven backbones from three model families and find that it surpasses all training-free metrics and exceeds fine-tuned evaluators on two out of three human-correlation benchmarks.
Chinese Translation
图像与文本的对齐是计算机视觉中的一个核心问题，应用于字幕评估、幻觉检测、数据整理以及文本到图像（T2I）生成器的基准测试。随着T2I模型的不断改进，基准测试的要求也变得更加严格，需要能够发现一系列问题的指标，如缺失对象、属性交换、计数错误和忽略的否定。近期的研究通过在偏好数据上微调评估器或通过提示视觉-语言模型来解决这一问题，后者可以是整体性的（使用字幕）或分解的验证问题。然而，现有的方法存在不足：微调的指标仍然受限于一个主干网络和训练分布；整体性指标缺乏细粒度细节；而分解指标依赖于固定的“是”假设，这在假设失败时会惩罚真实图像。相比之下，我们提出了DEPICT，这是一种无训练的指标，它用图像基础答案和仅字幕答案之间的预期一致性来替代固定的参考答案，并根据字幕对问题的决定性影响对问题进行加权。通过替换固定的参考，我们的一致性规则将否定的准确性从19%提高到88%。为了恢复在分解过程中丢失的上下文，DEPICT将这一一致性得分与整体得分合并。我们在五个基准和来自三个模型家族的十一种主干上评估DEPICT，发现它超越了所有无训练的指标，并在三个人类相关性基准中的两个上超过了微调的评估器。
cs.CV / 90 / 2610.03632
World Embedding Benchmark
世界嵌入基准
Abstract
Physical fidelity has received increasing attention in world models and video generation, yet how video representations encode physical information remains less understood. We introduce the World Embedding Benchmark, comprising 8,000 controlled simulation cases from 80 families spanning fluid mechanics, solid mechanics, dynamics, and optics & electromagnetism. Each case pairs a rendered video with simulation-derived physical annotations, supporting three complementary tasks: text-video retrieval, physical-property regression, and multiple-choice video-description pair classification. We use these tasks to distinguish cross-modal physical alignment from the recoverability of quantitative physical information. Evaluated pre-trained omnimodal embedding models show weak retrieval and near-chance within-family pair classification, while lightweight probes recover useful physical information from frozen video embeddings. Continual contrastive training with physics-specific video-text pairs improves retrieval and pair classification but degrades physical-property regression, revealing a trade-off between alignment and quantitative information recoverability. Finally, we use the embeddings to retrieve reference videos for retrieval-augmented generation with MiniMax-H3. Retrieved references improve the physical fidelity of generated videos, with stronger retrieval models yielding larger gains in our experiments. Together, these findings highlight the need to evaluate physical alignment and property recoverability jointly, and demonstrate the utility of physical representations for improving video generation.
Chinese Translation
物理真实性在世界模型和视频生成中受到越来越多的关注，但视频表示如何编码物理信息仍然不够清楚。我们引入了世界嵌入基准，包含来自80个类别的8000个受控仿真案例，涵盖流体力学、固体力学、动力学以及光学与电磁学。每个案例将渲染视频与仿真生成的物理注释配对，支持三项互补任务：文本-视频检索、物理属性回归和多项选择视频描述对分类。我们利用这些任务区分跨模态物理对齐与定量物理信息的可恢复性。评估的预训练全模态嵌入模型在检索和同类对分类中表现较弱，接近随机水平，而轻量探针能够从冻结的视频嵌入中恢复有用的物理信息。与物理特定视频-文本对的持续对比训练改善了检索和对分类，但降低了物理属性回归的性能，揭示了对齐与定量信息可恢复性之间的权衡。最后，我们使用嵌入来检索参考视频，以进行基于检索的生成，采用MiniMax-H3。检索到的参考视频提高了生成视频的物理真实性，强大的检索模型在我们的实验中带来了更大的收益。综合来看，这些发现强调了共同评估物理对齐和属性可恢复性的必要性，并展示了物理表示在改善视频生成中的实用性。
cs.CV / 91 / 2610.03636
LoGo: Local-Global Rewards for Consistent Long-Horizon Video Generation
LoGo：用于一致性长时间视频生成的局部-全局奖励
Abstract
Camera-controlled video models are rapidly advancing toward long generation horizons and complex camera control. A key failure mode is 3D inconsistency: as the camera moves, objects lose permanence and scene structures shift. Existing post-training techniques, which assign a single scalar reward to the entire generation, are poorly suited to correcting these inconsistencies over long horizons. We introduce LoGo, which blends global and spatially localized rewards for camera-controlled video models. The local reward provides fine-grained credit assignment, which substantially improves 3D consistency, while the global reward preserves camera following and video quality. Across three base models, LoGo shows a clear advantage on DL3DV and TrajectoryBench, a new benchmark for long-horizon, complex-camera-control generation that current evaluations lack. LoGo effectively reduces local object shifts, artifacts, and global scene changes, illustrating the importance of credit assignment in post-training video models. Project website: https://ziqi-ma.github.io/logo-website/
Chinese Translation
摄像机控制的视频模型正迅速朝着长生成时间和复杂摄像机控制的方向发展。一个关键的失败模式是三维不一致性：随着摄像机的移动，物体失去持久性，场景结构发生变化。现有的后训练技术为整个生成过程分配单一的标量奖励，这对于在长时间范围内纠正这些不一致性并不适用。我们提出了LoGo，它结合了针对摄像机控制视频模型的全局和空间局部奖励。局部奖励提供了细粒度的信用分配，显著改善了三维一致性，而全局奖励则保持了摄像机跟随和视频质量。在三个基础模型中，LoGo在DL3DV和TrajectoryBench（一个当前评估所缺乏的长时间范围、复杂摄像机控制生成的新基准）上显示出明显的优势。LoGo有效减少了局部物体位移、伪影和全局场景变化，说明了在后训练视频模型中信用分配的重要性。项目网站：https://ziqi-ma.github.io/logo-website/
cs.CV / 92 / 2610.03649
On-Board Anomaly Detection for Efficient Marine Environmental Monitoring
高效海洋环境监测的船载异常检测
Abstract
Marine ecosystems are impacted by various threats such as oil spills, algal blooms, and sediment floods, which disrupt habitats, wildlife, and human activities. Advances in satellite imagery and Artificial Intelligence (AI) have enhanced our capabilities for early detection and mitigation of such hazards. In this paper, we propose a marine event detection pipeline for Earth observation satellites equipped with multi- or hyperspectral sensors. Our approach includes a self-supervised neural network encoder that compresses satellite images into a reduced latent space, enabling efficient onboard processing. A machine learning anomaly detection model identifies deviations from normal sea patterns to detect environmental anomalies. We compare its performance against traditional algorithms such as Isolation Forest, One-Class Support Vector Machine and Local Outlier Factors. Our lightweight, resource-efficient pipeline is optimized for deployment on satellites with limited computational resources, ranging from embedded CPUs to AI hardware accelerators. By prioritizing the transmission of critical information, our solution enhances system responsiveness and optimizes satellite communication bandwidth. Demonstrated through current integration across multiple missions, including European Space Agency's (ESA) Phisat-2 mission and Microsoft/Thales Alenia Space IMAGIN-e mission, our pipeline aims to improve marine environmental monitoring by providing timely alerts and efficient data reduction.
Chinese Translation
海洋生态系统受到油污泄漏、藻类繁殖和沉积洪水等多种威胁的影响，这些威胁破坏了栖息地、野生动物和人类活动。卫星影像和人工智能（AI）的进步增强了我们对这些危害的早期检测和缓解能力。本文提出了一种针对配备多光谱或高光谱传感器的地球观测卫星的海洋事件检测流程。我们的方法包括一个自监督神经网络编码器，将卫星图像压缩到一个简化的潜在空间，从而实现高效的船载处理。机器学习异常检测模型识别正常海洋模式的偏差，以检测环境异常。我们将其性能与传统算法进行比较，如孤立森林（Isolation Forest）、单类支持向量机（One-Class Support Vector Machine）和局部离群因子（Local Outlier Factors）。我们的轻量级、资源高效的流程经过优化，适用于计算资源有限的卫星，从嵌入式CPU到AI硬件加速器。通过优先传输关键信息，我们的解决方案增强了系统响应能力，并优化了卫星通信带宽。通过在多个任务中的当前集成，包括欧洲航天局（ESA）的Phisat-2任务和微软/泰雷兹阿莱尼亚空间（Thales Alenia Space）IMAGIN-e任务，我们的流程旨在通过提供及时警报和高效数据减少来改善海洋环境监测。
cs.CV / 93 / 2610.03664
ProAR: Learning Prospective Reasoning with Autoregressive Video Models
ProAR：利用自回归视频模型学习前瞻性推理
Abstract
Autoregressive (AR) video models excel at causal generation, but their reliance on next-chunk prediction confines them to a short-sighted, reactive paradigm. This limitation is particularly consequential for reasoning-oriented generation, where achieving a target outcome through valid intermediate states matters more than local visual plausibility. To address this challenge, we propose Learning Prospective Reasoning with Autoregressive Video Models (ProAR), a novel framework that transforms autoregressive video generation into a goal-oriented reasoning process. ProAR introduces two key components: (1) To anchor generation to the long-range outcome, we integrate goal-frame prediction into the autoregressive loop via an asymmetric attention mask, enabling the predicted goal frame to guide the generation of intermediate states without being disrupted by them. (2) To guide short-range transitions, we introduce future representation self-alignment to encourage current hidden states to anticipate upcoming temporal dynamics. By leveraging teacher-forcing in AR training, we extract clean future representations in a single forward pass and align current representations with them using a lightweight, training-only predictor. Together, these two mechanisms seamlessly combine explicit, sparse target supervision with implicit, dense step-wise guidance, promoting coherent, goal-directed reasoning progress with modest computational cost. Experiments show that ProAR's complementary components consistently improve performance across diverse visual reasoning benchmarks. The framework proves highly training-efficient, surpassing fully trained standard AR baselines using only 25% of the training steps. This paradigm also demonstrates promising applicability to embodied reasoning tasks.
Chinese Translation
自回归（AR）视频模型在因果生成方面表现出色，但它们对下一个片段预测的依赖使其局限于短视的反应性范式。这一局限性对于以推理为导向的生成尤为重要，因为通过有效的中间状态实现目标结果比局部视觉的合理性更为重要。为了解决这一挑战，我们提出了利用自回归视频模型学习前瞻性推理（ProAR），这是一个将自回归视频生成转变为目标导向推理过程的新框架。ProAR引入了两个关键组件：（1）为了将生成锚定到长期结果，我们通过不对称注意力掩码将目标帧预测整合到自回归循环中，使得预测的目标帧能够在不被中间状态干扰的情况下引导中间状态的生成。（2）为了引导短期过渡，我们引入未来表示自对齐，以鼓励当前隐状态预测即将到来的时间动态。通过在AR训练中利用教师强制，我们在单次前向传播中提取干净的未来表示，并使用轻量级的仅训练预测器将当前表示与之对齐。这两个机制无缝结合了显式的稀疏目标监督与隐式的密集逐步指导，促进了连贯的、以目标为导向的推理进展，同时保持适度的计算成本。实验表明，ProAR的互补组件在各种视觉推理基准测试中持续提高了性能。该框架证明了高度的训练效率，仅使用25%的训练步骤便超越了完全训练的标准AR基线。这一范式在具身推理任务中的应用前景也令人鼓舞。
cs.CV / 94 / 2610.03689
SigLIP2 for aerial fire risk classification
用于空中火灾风险分类的 SigLIP2
Abstract
We examine the transfer of a pretrained SigLIP2 image encoder to seven class fire risk classification from aerial imagery. We introduce a reproducible partition of the public FireRisk training mirror and an implementation that records data provenance, preprocessing and model selection. Two initial runs compare a frozen encoder probe with full model adaptation. On the validation partition, full adaptation reaches 63.05% accuracy and 58.94% macro F1, compared with 55.95% and 50.19% for the probe. Both runs use one training seed and select their checkpoint on the same validation partition. These development results support further evaluation of SigLIP2 but do not establish performance on an independent test set or unseen regions. The accompanying code provides a common framework for repeated experiments and comparisons with additional visual encoders.
Chinese Translation
我们研究了将预训练的 SigLIP2 图像编码器迁移到七类火灾风险分类的应用，基于空中影像。我们引入了一个可重复的公共 FireRisk 训练镜像的划分，并实现了记录数据来源、预处理和模型选择的系统。两个初步实验比较了冻结编码器探针与完全模型适应的效果。在验证集上，完全适应的准确率达到 63.05%，宏观 F1 值为 58.94%，而探针的准确率为 55.95%，宏观 F1 值为 50.19%。这两个实验均使用相同的训练种子，并在同一验证集上选择其检查点。这些开发结果支持对 SigLIP2 进行进一步评估，但并未确立其在独立测试集或未见区域上的表现。随附的代码提供了一个通用框架，以便于重复实验和与其他视觉编码器的比较。
cs.CV / 95 / 2610.03691
FlowHMR: Physically Plausible Motion Capture from Video
FlowHMR：从视频中恢复物理上合理的人体运动捕捉
Abstract
We present FlowHMR, a framework for recovering physically plausible global 3D human motion from monocular video. Previous learning-based methods typically regress human motion directly from video and train the network with geometric supervision. However, recovering human motion from monocular video is inherently ambiguous in depth, and direct regression tends to collapse toward an averaged solution. Moreover, the recovered motions are not guaranteed to be physically plausible, so physics-based tracking of them often fails. To address these challenges, we formulate video motion capture as a video-conditioned motion generation problem and first pretrain a flow matching model for this task. Given an input video, the pretrained model generates diverse motion candidates, but not all of them are faithful to the video or physically trackable. We therefore post-train the model using Group Relative Policy Optimization (GRPO) with two rewards. A fidelity reward encourages consistency with the input video. A tracking reward favors motions that a physics-based controller can track successfully. Together, these rewards shift the model's output preference, so the post-trained model stays faithful to the input video while producing more physically plausible motion. We further introduce Wild-4K, a large and diverse dataset of about 4K internet videos, for evaluating human motion recovery in the wild. Qualitative and quantitative experiments on Wild-4K show that our method outperforms state-of-the-art methods in overall motion fidelity and achieves a physical tracking success rate of 82.47%, compared with 62.82% for the strongest baseline, GVHMR.
Chinese Translation
我们提出了FlowHMR，一个从单目视频中恢复物理上合理的全球3D人体运动的框架。之前的基于学习的方法通常直接从视频回归人体运动，并使用几何监督来训练网络。然而，从单目视频中恢复人体运动在深度上本质上是模糊的，直接回归往往趋向于一个平均解。此外，恢复的运动并不能保证是物理上合理的，因此基于物理的跟踪往往失败。为了解决这些挑战，我们将视频运动捕捉公式化为一个视频条件下的运动生成问题，并首先为此任务预训练一个流匹配模型。给定输入视频，预训练模型生成多样的运动候选，但并非所有候选都忠实于视频或可被物理跟踪。因此，我们使用组相对策略优化（Group Relative Policy Optimization, GRPO）对模型进行后训练，设置了两个奖励。保真奖励鼓励与输入视频的一致性，而跟踪奖励则偏向于物理控制器能够成功跟踪的运动。这些奖励共同改变了模型的输出偏好，使得后训练模型在生成更物理上合理的运动的同时，仍然忠实于输入视频。我们进一步引入Wild-4K，这是一个约4K互联网视频的大型多样化数据集，用于评估在真实环境中恢复人体运动的效果。在Wild-4K上的定性和定量实验表明，我们的方法在整体运动保真度上优于最先进的方法，并且实现了82.47%的物理跟踪成功率，而最强基线GVHMR的成功率为62.82%。
cs.CV / 96 / 2610.03698
Decoding the Functional Roles of Register and High-Norm Patch Tokens in Vision Transformers
解码视觉变换器中注册和高范数补丁令牌的功能角色
Abstract
Self-supervised Vision Transformers (ViTs), such as DINOv2, learn rich visual representations, but the functions of their internal tokens remain poorly understood. Recent architectures introduce dedicated register tokens to reduce high-norm out- lier patch tokens that emerge in background re- gions, yet the semantic and functional roles of both token types have not been fully established. In this paper, we analyze these roles by training sparse autoencoders (SAEs) on register-token and outlier-token activations in DINOv2. Using an automated interpretability pipeline, UMAP clus- tering, and CLIP-space cross-checks, we find that register-token features are more strongly associ- ated with high-level semantic concepts. Outlier- token features, by contrast, are more often associ- ated with lower-level structural, background, and texture-dominant patterns. Causal ablations fur- ther reveal a substantial functional asymmetry: disrupting top-activating register-derived features produces a 48.17% drop in representation cosine similarity, whereas disrupting outlier-derived fea- tures produces only a 0.31% drop. Together, our results provide evidence for token specialization in self-supervised ViTs.
Chinese Translation
自监督视觉变换器（ViTs），如 DINOv2，学习丰富的视觉表征，但其内部令牌的功能仍然不够清楚。最近的架构引入了专用的注册令牌，以减少在背景区域中出现的高范数异常补丁令牌，然而这两种令牌类型的语义和功能角色尚未完全确定。在本文中，我们通过在 DINOv2 中对注册令牌和异常令牌激活进行稀疏自编码器（SAEs）训练来分析这些角色。利用自动化可解释性管道、UMAP 聚类和 CLIP 空间交叉检查，我们发现注册令牌特征与高级语义概念的关联性更强。相比之下，异常令牌特征则更常与低级结构、背景和纹理主导模式相关联。因果消融进一步揭示了显著的功能不对称性：破坏高激活的注册衍生特征会导致表示余弦相似度下降 48.17%，而破坏异常衍生特征仅导致 0.31% 的下降。综合来看，我们的结果为自监督 ViTs 中的令牌专业化提供了证据。
cs.CV / 97 / 2610.03715
4DCodeBench: Benchmarking Agents on Inverse Graphics of Dynamic Scenes
4DCodeBench：动态场景逆图形的智能体基准测试
Abstract
We introduce 4DCodeBench, a benchmark for 4D inverse graphics through code generation, in which agents reconstruct dynamic scenes from video as executable graphics programs. To accomplish this, agents must translate visual observations into compact representations of scene structure and dynamics, by implementing abstractions such as physical simulations to reproduce complex behavior. To evaluate this capability, we curate a set of real-world videos and construct synthetic scenes spanning diverse physical phenomena, including deformation, fluid flow, and fracture. We perform extensive benchmarking of frontier models, finding that strong static reconstruction capabilities do not yet translate into reliable reconstruction of complex dynamics. 4DCodeBench provides a testbed for tracking progress toward agents that can interpret the dynamics of the world through code. Our benchmark is available at https://github.com/4DCodeBench/4DCodeBench
Chinese Translation
我们介绍了4DCodeBench，这是一个通过代码生成进行4D逆图形的基准测试，其中智能体从视频中重建动态场景为可执行的图形程序。为此，智能体必须将视觉观察转化为场景结构和动态的紧凑表示，通过实现物理模拟等抽象来再现复杂行为。为了评估这一能力，我们整理了一组真实世界的视频，并构建了涵盖多种物理现象的合成场景，包括变形、流体流动和断裂。我们对前沿模型进行了广泛的基准测试，发现强大的静态重建能力尚未转化为可靠的复杂动态重建。4DCodeBench为追踪智能体通过代码解释世界动态的进展提供了一个测试平台。我们的基准测试可在 https://github.com/4DCodeBench/4DCodeBench 获取。
cs.CV / 98 / 2610.03716
MoSE3: Learning World-Space SE(3) at Every Pixel
MoSE3：在每个像素处学习世界空间 SE(3)
Abstract
Dense 3D point tracking has been a prominent paradigm for modeling motion in dynamic scenes, but a point track is just a 3-DoF translation curve per pixel: it captures where pixels go, not the rotation of the underlying part, nor which pixels move together as one body. We propose MoSE3, the first feed-forward model that predicts dense SE(3) motion from monocular RGB video, producing full 6-DoF rigid transforms at every pixel in world space. Per-pixel SE(3) motion offers a richer view of how a scene moves: rotation, translation, and grouping all at once. Directly predicting SE(3) is challenging: rotations lie on a curved manifold that is ill-suited to Euclidean regression, and annotations for SE(3) are particularly difficult to acquire. To address these challenges, MoSE3 predicts per-pixel SE(3) through two jointly learned intermediates, 3D point tracks and rigidity embeddings, and recovers SE(3) by differentiably fitting transforms within each soft rigid cluster, enabling end-to-end prediction and supervision. To close the data gap, we introduce Art-Kubric, a large-scale synthetic dataset with dense SE(3) and rigidity labels for articulated objects with rich physical interactions. MoSE3 achieves state-of-the-art SE(3) estimation at pixel, part, and object levels on both rigid and articulated benchmarks, and state-of-the-art average 3D point tracking accuracy across three datasets, while showing strong generalization to real-world videos despite being trained solely on synthetic motion data.
Chinese Translation
密集的 3D 点跟踪一直是动态场景中建模运动的一个重要范式，但每个像素的点轨迹仅仅是一个 3 自由度的平移曲线：它捕捉了像素的移动，而不是底层部分的旋转，也没有说明哪些像素作为一个整体一起移动。我们提出了 MoSE3，这是第一个从单目 RGB 视频中预测密集 SE(3) 运动的前馈模型，在世界空间中的每个像素生成完整的 6 自由度刚性变换。每个像素的 SE(3) 运动提供了场景运动的更丰富视角：旋转、平移和分组一次性呈现。直接预测 SE(3) 是具有挑战性的：旋转位于一个曲面流形上，这对于欧几里得回归并不适合，并且 SE(3) 的注释特别难以获取。为了解决这些挑战，MoSE3 通过两个联合学习的中间变量——3D 点轨迹和刚性嵌入，预测每个像素的 SE(3)，并通过在每个软刚性簇内可微分地拟合变换来恢复 SE(3)，从而实现端到端的预测和监督。为了填补数据差距，我们引入了 Art-Kubric，一个大型合成数据集，包含具有丰富物理交互的关节物体的密集 SE(3) 和刚性标签。MoSE3 在刚性和关节基准测试中，在像素、部分和对象级别上实现了最先进的 SE(3) 估计，并在三个数据集上展示了最先进的平均 3D 点跟踪准确性，同时尽管仅在合成运动数据上训练，仍表现出对真实世界视频的强泛化能力。
cs.CV / 99 / 2610.03717
Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis
少解码器更多编码器：来自新视角合成的几何表示学习
Abstract
This paper examines the role of Novel View Synthesis (NVS) in geometric representation learning. In principle, NVS should reason about 3D scene structure, thereby enabling transferable multi-view geometric representations. Yet, existing encoder-based NVS methods yield poor representations. This is not because of a lack of supervisory signal, but rather due to inconspicuous architectural choices: \textit{spatially expressive decoders} that dilute representational capabilities of the scene encoder, and \textit{low-level pixel-space targets} that hinder feature learning. We present SNAP, a self-supervised encoder-decoder transformer that addresses both through a pose-conditioned local decoder and a latent-space reconstruction objective. SNAP is task agnostic, and we show that it is competitive with special-purpose geometry-supervised methods. SNAP also performs competitively against self-supervised representations across five tasks: visual localization, pose estimation, point correspondence, depth estimation, and robot manipulation. Remarkably, SNAP's patch features exhibit emergent viewpoint invariance that approaches heavily supervised models despite lower compute and data budgets. Under camera shifts where standard 2D representations collapse, SNAP degrades more gracefully, revealing that restricting decoder expressivity actively prevents the suppression of transferable geometric structure. https://snap-nvs.github.io
Chinese Translation
本文探讨了新视角合成（Novel View Synthesis, NVS）在几何表示学习中的作用。原则上，NVS 应该能够推理三维场景结构，从而实现可转移的多视角几何表示。然而，现有的基于编码器的 NVS 方法产生的表示效果较差。这并不是由于缺乏监督信号，而是由于不显眼的架构选择： extit{空间表达能力强的解码器}稀释了场景编码器的表示能力，以及 extit{低级像素空间目标}阻碍了特征学习。我们提出了 SNAP，一种自监督的编码器-解码器变换器，通过姿态条件的局部解码器和潜在空间重建目标解决了这两个问题。SNAP 是任务无关的，我们展示了它在与特定目的几何监督方法的竞争中表现出色。SNAP 在五个任务（视觉定位、姿态估计、点对应、深度估计和机器人操作）中也与自监督表示表现出竞争力。值得注意的是，SNAP 的补丁特征展现出接近重度监督模型的视角不变性，尽管计算和数据预算较低。在相机移动导致标准二维表示崩溃的情况下，SNAP 的降级表现更为优雅，揭示了限制解码器表达能力能够有效防止可转移几何结构的抑制。
cs.AI / 1 / 2610.02260
MintFlow: Minimal Trajectory Intervention for Constrained Flow Matching
MintFlow：约束流匹配的最小轨迹干预
Abstract
Flow matching models excel at generative modeling, and many downstream applications require their samples to satisfy prescribed constraints, such as observed measurements and physical laws. However, existing constrained samplers often face a trade-off: \textit{enforcing constraints can substantially displace samples from the pretrained data distribution}. To address this trade-off, we introduce \textbf{MintFlow}, a training-free constrained sampling framework that formulates constraint enforcement as a minimal intervention on the pretrained flow trajectory. MintFlow seeks the minimal perturbation of an intermediate flow state such that its subsequent evolution under the pretrained flow field satisfies the target constraint. By minimally perturbing the flow state while keeping the pretrained flow field unchanged, MintFlow enforces the constraint while minimizing unnecessary deviation from the pretrained distribution. An adjoint formulation yields a closed-form expression for this perturbation, eliminating expensive iterative optimization. Furthermore, MintFlow adaptively selects the intervention time to balance the required perturbation magnitude with its amplification by the remaining flow. Across a range of tasks in generative vision and physical system modeling, MintFlow achieves competitive constraint satisfaction while preserving the pretrained generative distribution substantially better than state-of-the-art constrained methods.
Chinese Translation
流匹配模型在生成建模方面表现出色，许多下游应用要求其样本满足规定的约束，例如观察到的测量值和物理定律。然而，现有的约束采样器往往面临一个权衡： extit{强制执行约束可能会显著偏离预训练数据分布的样本}。为了解决这一权衡，我们提出了 extbf{MintFlow}，一种无训练的约束采样框架，将约束执行形式化为对预训练流轨迹的最小干预。MintFlow寻求对中间流状态的最小扰动，使其在预训练流场下的后续演化满足目标约束。通过在保持预训练流场不变的情况下，最小扰动流状态，MintFlow在强制执行约束的同时，最小化与预训练分布的不必要偏差。伴随的公式推导为这一扰动提供了封闭形式的表达，消除了昂贵的迭代优化。此外，MintFlow自适应选择干预时间，以平衡所需的扰动幅度与剩余流的放大效应。在生成视觉和物理系统建模的一系列任务中，MintFlow在保持预训练生成分布方面显著优于最先进的约束方法，同时实现了竞争性的约束满足。
cs.AI / 2 / 2610.02267
Fast Models, Slow Evidence: A Paired and Self-Audited Evaluation of System-1 Decision Models for LLM Agent Harnesses
快速模型，缓慢证据：系统1决策模型在大型语言模型代理应用中的配对与自我审计评估
Abstract
Agent harnesses make many small, typed decisions per task: which model to call, which tool to use, whether retrieved text is relevant, whether an input carries an injection. System-1 decision models answer such questions in a single forward pass with class probabilities, promising large cost and latency savings over LLM calls. We present a paired evaluation of an open-weight (Laya) and a hosted (Jev) System-1 model on 11 agent decision points built from 18 public sources: 7,283 base cases plus 6,640 robustness variants, with byte-identical inputs, paired tests, and cross-hardware and cross-day reproducibility checks. Jev is significantly more accurate on 9 of 11 decision points (+10.8 to +46.0 pp). Neither model beats chance on zero-shot model routing, and they tie on RAG relevance gating. Laya changes 30% of its answers when the option order is reversed and degrades sharply with many or similar candidates (31% at 50 nearest-neighbour tools, vs. 98% for Jev on items with a unique correct tool). We also audit our own pipeline. Three analysis errors and one design confound distorted headline deployment claims: an omitted pre-screen cost (reported 23.9% saving, actual 4.3%), gate accuracy reported as end-to-end quality (58% vs. 98%), in-sample thresholds (5% target, up to 17% held-out misses), and a "channel effect" on injection false positives that vanishes with channel-native content. Two other suspected confounds did not change the conclusions. All cases, raw outputs and analysis code are available at https://github.com/David-DL-Space/sys1-eval.
Chinese Translation
代理应用在每个任务中做出许多小型的、类型化的决策：调用哪个模型、使用哪个工具、检索到的文本是否相关、输入是否包含注入。系统1决策模型通过单次前向传播以类别概率回答这些问题，承诺在大型语言模型调用中节省大量成本和延迟。我们对一个开放权重模型（Laya）和一个托管模型（Jev）在11个代理决策点上的配对评估，基于18个公共来源构建：7283个基本案例加6640个鲁棒性变体，使用字节完全相同的输入、配对测试，以及跨硬件和跨天的可重复性检查。Jev在11个决策点中的9个上显著更准确（提高10.8到46.0个百分点）。在零样本模型路由上，两个模型均未超过随机猜测，而在RAG相关性门控上则平局。当选项顺序反转时，Laya改变了30%的答案，并且在候选项较多或相似时急剧下降（在50个最近邻工具中为31%，而在具有唯一正确工具的项目中Jev为98%）。我们还对自己的流程进行了审计。三个分析错误和一个设计混淆扭曲了头条部署声明：遗漏的预筛选成本（报告节省23.9%，实际为4.3%）、门控准确性被报告为端到端质量（58%对比98%）、样本内阈值（目标5%，最多17%被保留的遗漏），以及注入假阳性上的“通道效应”，该效应在通道本地内容中消失。其他两个怀疑的混淆没有改变结论。所有案例、原始输出和分析代码可在https://github.com/David-DL-Space/sys1-eval获取。
cs.AI / 3 / 2610.02281
The AI Risk Observatory: What Can We Learn from AI Disclosures in Annual Reports About Societal Resilience?
人工智能风险观察站：我们能从年度报告中的人工智能披露中学到关于社会韧性的什么？
Abstract
Societal resilience research relies on access to useful and actionable data, which motivates our main research question: Can annual reports, processed at scale with LLMs, provide a useful signal about how companies disclose their response to AI? We test this by applying a reproducible two-stage classification pipeline to 9,821 annual reports from 1,362 UK listed companies (2020-2025, with partial 2026 data). We first validate the method against 474 human-annotated passages, finding high recall and moderate label-level agreement. We then report three empirical patterns: (i) between 2020 and 2025, the share of reports mentioning AI risk rose from 2.8% to 41.2%, while AI adoption disclosure also rose, from 13.8% to 45.2%, and named vendor mentions cluster around a small set of major providers led by Microsoft; (ii) disclosure varies substantially by Critical National Infrastructure sector and market segment: AIM reports disclose AI risk at far lower rates than Main Market reports, and sectors such as Energy and Data Infrastructure lag behind the rest in AI risk disclosure; and (iii) harm disclosures are near-absent (seven reports across the entire corpus). We develop a substantiveness classification to assess the quality of the disclosure and find that most AI risk disclosure is not substantive: in 2025, 41.2% of all reports mention AI as a risk, but only 4.3% contain AI risk disclosure we classify as substantive.
Chinese Translation
社会韧性研究依赖于可用且可操作的数据，这促使我们提出主要研究问题：经过大规模处理的年度报告能否提供关于公司如何披露其对人工智能（AI）响应的有用信号？我们通过对来自1,362家英国上市公司（2020-2025年，部分2026年数据）的9,821份年度报告应用可重复的两阶段分类流程来测试这一点。我们首先通过474个人工标注的段落验证该方法，发现高召回率和中等标签级别一致性。然后，我们报告了三个实证模式：（i）在2020年至2025年间，提及AI风险的报告比例从2.8%上升至41.2%，而AI采用披露也从13.8%上升至45.2%，提及供应商的名称集中在以微软为首的一小部分主要提供商；（ii）披露在关键国家基础设施部门和市场细分中差异显著：人工智能市场（AIM）报告的AI风险披露率远低于主市场报告，能源和数据基础设施等行业在AI风险披露方面落后于其他行业；（iii）损害披露几乎不存在（整个语料库中仅有七份报告）。我们开发了一种实质性分类来评估披露的质量，发现大多数AI风险披露并不具实质性：在2025年，41.2%的所有报告提到AI作为风险，但仅有4.3%包含我们分类为实质性的AI风险披露。
cs.AI / 4 / 2610.02300
Keep It CALM: Analyzing the Limits of Global Unsafety in Text-to-Image Generation
保持冷静：分析文本到图像生成中的全球不安全性限制
Abstract
Training-free safeguards for text-to-image generation often rely on a reusable safety signal, such as an unsafe direction or global toxic subspace, applied broadly across prompts. We provide a controlled geometric analysis of this global-unsafety assumption and reveal a consistent coverage-selectivity trade-off: compact unsafe subspaces fail to cover heterogeneous unsafe semantics, whereas broader aggregation increasingly distorts safety-adjacent benign prompts. Motivated by this finding, we propose CALM (Counterfactual Adaptive Local Modulation), a training-free safeguard that replaces uniform global removal with prompt-local counterfactual correction. Using matched unsafe-benign anchors, CALM routes each prompt to active unsafe categories, minimally edits only violating token representations toward the safe side, and suppresses positively aligned unsafe residual components. Across broad evaluation, CALM significantly improves unsafe content suppression while preserving benign utility, demonstrating that local counterfactual correction provides a more selective alternative to global unsafe signal removal.
Chinese Translation
无训练的文本到图像生成安全保障通常依赖于可重复使用的安全信号，例如不安全方向或全球有毒子空间，这些信号在提示中广泛应用。我们对这一全球不安全假设进行了受控的几何分析，并揭示了一种一致的覆盖选择性权衡：紧凑的不安全子空间无法覆盖异质的不安全语义，而更广泛的聚合则越来越扭曲与安全相邻的良性提示。基于这一发现，我们提出了CALM（反事实自适应局部调制），这是一种无训练的安全保障，它用提示局部反事实校正替代了均匀的全球移除。通过匹配的不安全-良性锚点，CALM将每个提示引导到活跃的不安全类别，仅对违反的标记表示进行最小编辑，使其朝向安全侧，并抑制积极对齐的不安全残余成分。在广泛的评估中，CALM显著提高了不安全内容的抑制，同时保持了良性的实用性，证明局部反事实校正提供了比全球不安全信号移除更具选择性的替代方案。
cs.AI / 5 / 2610.02330
Choosing Before Acting: Comparative Value Estimation for Long-Horizon Tool-Use Agents
行动前的选择：长时间视角工具使用代理的比较价值估计
Abstract
Large language models (LLMs) rely on long-horizon tool invocation sequences for complex tasks, where each invocation can alter the task state and condition subsequent decisions. In long-horizon tool use, final-outcome rewards provide weak credit assignment over long interaction traces. Step-level rewards can offer more targeted feedback, but obtaining reliable step supervision often requires human or LLM judgment, or additional rollouts to estimate the downstream effect of an intermediate decision. In this paper, we argue that effective tool-use agents should estimate the long-horizon value of a possible next tool invocation before executing it. This objective requires comparative supervision over alternative invocations under the same context, while logged trajectories only contain the invocation that was actually taken. Therefore, we propose Comparative Inference for Tool-use Agents (CITA). CITA trains a Comparative Inference Model (CIM) from paired signals that combine observed tool behavior, scalable supervision from a Bayesian tool-graph simulator, and semantic judgments from LLM-based comparison. The resulting CIM learns to estimate how likely a possible next tool invocation is to support final task success under the current context. Across three tool-use benchmarks and multiple backbone LLMs, CITA consistently improves Tool F1 and task success. Additional analysis shows that CIM learns accurate step-level value estimates for comparative tool choices.
Chinese Translation
大型语言模型（LLMs）依赖于长时间视角的工具调用序列来完成复杂任务，其中每次调用都可能改变任务状态并影响后续决策。在长时间视角的工具使用中，最终结果奖励对长交互轨迹的信用分配较弱。逐步奖励可以提供更有针对性的反馈，但获得可靠的逐步监督通常需要人类或LLM的判断，或额外的回滚以估计中间决策的下游影响。在本文中，我们认为有效的工具使用代理应该在执行可能的下一个工具调用之前，估计其长时间视角的价值。这个目标需要在相同上下文下对替代调用进行比较监督，而记录的轨迹仅包含实际采取的调用。因此，我们提出了工具使用代理的比较推断（Comparative Inference for Tool-use Agents, CITA）。CITA从成对信号中训练比较推断模型（Comparative Inference Model, CIM），这些信号结合了观察到的工具行为、来自贝叶斯工具图模拟器的可扩展监督，以及基于LLM的比较的语义判断。最终的CIM学习如何估计在当前上下文下，可能的下一个工具调用支持最终任务成功的可能性。在三个工具使用基准和多个基础LLM上，CITA始终提高了工具F1和任务成功率。额外分析表明，CIM学习到了准确的逐步价值估计，以支持比较工具选择。
cs.AI / 6 / 2610.02331
World Editing: Intervening on Executable Worlds at Increasing Depth
世界编辑：在可执行世界中进行更深层次的干预
Abstract
Interactive world models are increasingly capable of generating environments and acting within them, yet deliberately editing an existing executable world remains underexplored. We formulate world editing as intervening on an existing world while preserving properties that should remain unchanged, and introduce intervention depth as an axis describing how strongly an edit couples world entities, dynamics, and systems. We instantiate this capability through industry-grade game modding and introduce IGMWorld, together with IGMBench, a benchmark of 110 tasks and over 1.1K executable state and behavioral criteria across Minecraft and Terraria. The tasks span property, entity, dynamics, and system interventions and are evaluated through deterministic executability, behavioral, preservation, and visual checks. Frontier coding agents already exhibit substantial world-editing capability: the strongest configuration solves 78.2% of tasks under a strict task-level criterion, while criterion-level performance reaches 94.8%. Reliability generally decreases with intervention depth, and this pattern persists even among tasks with similar numbers of evaluation criteria. Most failed edits still build and load successfully, suggesting that the main difficulty is making the edited world behave as requested. Visual consistency remains a separate weakness, with all evaluated configurations below 50% joint visual pass rate. These results show that world editing is a distinct capability from world generation and interaction, and that executable games provide a practical testbed for studying it.
Chinese Translation
交互式世界模型越来越能够生成环境并在其中进行操作，但有意编辑现有可执行世界的研究仍然不足。我们将世界编辑定义为对现有世界的干预，同时保留应保持不变的属性，并引入干预深度作为描述编辑如何强烈耦合世界实体、动态和系统的一个维度。我们通过行业级游戏修改实现这一能力，并引入IGMWorld，以及IGMBench，这是一个涵盖Minecraft和Terraria的110个任务和超过1.1K可执行状态和行为标准的基准。这些任务涵盖属性、实体、动态和系统干预，并通过确定性可执行性、行为、保留和视觉检查进行评估。前沿编码代理已经展现出相当的世界编辑能力：最强配置在严格的任务级标准下解决了78.2%的任务，而标准级性能达到94.8%。随着干预深度的增加，可靠性通常会降低，这一模式在具有相似评估标准数量的任务中也依然存在。大多数失败的编辑仍然能够成功构建和加载，表明主要困难在于使编辑后的世界按要求行为。视觉一致性仍然是一个独立的弱点，所有评估的配置的联合视觉通过率均低于50%。这些结果表明，世界编辑是一种不同于世界生成和交互的能力，而可执行游戏为研究这一能力提供了一个实用的测试平台。
cs.AI / 7 / 2610.02342
A Multi Method Importance and Performance Efficiency Analysis of Topological Metrics for Natural Visibility Graph Based Cyber Attack Detection
基于自然可视图的网络攻击检测的拓扑指标多方法重要性与性能效率分析
Abstract
Natural Visibility Graph (NVG) based analysis characterizes network traffic through topological descriptors reflecting different structural properties. However, not all descriptors contribute equally to cyber-attack classification, and extracting a large metric set can increase computational cost. This study evaluates 21 NVG derived topological metrics and investigates whether a compact subset can preserve classification capability while improving computational efficiency. Four importance analysis methods SHAP, grouped Permutation Importance, Boruta, and Recursive Feature Elimination (RFE) are integrated through a Consensus Ranking strategy. Based on this ranking, Full21, Top15, Top10, Top7, Top5, and Top3 configurations are evaluated using the CICIDS2018 dataset, a CNN classifier, and stratified 5 fold cross validation. The three highest ranked metrics are avg_clustering_coeff_median, avg_clustering_coeff_std, and avg_clustering_coeff_mean. Top3 achieved the highest observed mean performance, with 97.148% accuracy, 97.055% weighted F1 score, and an MCC of 0.9675, compared with 95.999%, 95.521%, and 0.9549 for Full21, respectively. It also reduced total runtime from 14,961.39 s to 589.22 s (96.06%). These results indicate that importance guided metric reduction can provide a compact NVG representation with higher observed mean predictive performance and substantially lower computational cost under the evaluated setting.
Chinese Translation
基于自然可视图（Natural Visibility Graph, NVG）的分析通过反映不同结构特性的拓扑描述符来表征网络流量。然而，并非所有描述符对网络攻击分类的贡献相同，提取大量指标集可能会增加计算成本。本研究评估了21个基于NVG的拓扑指标，并探讨了是否可以通过紧凑的子集在保持分类能力的同时提高计算效率。通过共识排名策略整合了四种重要性分析方法：SHAP、分组置换重要性（grouped Permutation Importance）、Boruta和递归特征消除（Recursive Feature Elimination, RFE）。基于该排名，使用CICIDS2018数据集、CNN分类器和分层5折交叉验证评估了Full21、Top15、Top10、Top7、Top5和Top3配置。排名前三的指标分别为avg_clustering_coeff_median、avg_clustering_coeff_std和avg_clustering_coeff_mean。Top3实现了最高的平均性能，准确率为97.148%，加权F1得分为97.055%，MCC为0.9675，而Full21的对应值分别为95.999%、95.521%和0.9549。此外，Top3将总运行时间从14,961.39秒减少至589.22秒（减少96.06%）。这些结果表明，在评估的设置下，基于重要性指导的指标减少可以提供更紧凑的NVG表示，具有更高的平均预测性能和显著降低的计算成本。
cs.AI / 8 / 2610.02351
DeReAct: Decomposed Reasoning and Acting for Reliable AI Agents
DeReAct：用于可靠人工智能代理的分解推理与行动
Abstract
ReAct-based agents typically rely on a single LLM policy to propose actions, interact with the environment, and decide when a task is complete. This coupling makes action authorization and completion control difficult to enforce independently, allowing errors to propagate and unsupported completion claims to terminate execution. We introduce DeReAct, a modular agent architecture that externalizes two gating policies: a Critic that validates proposed actions before execution, and a Context Manager that reconstructs an environment-supported \textsc{State} and certifies task completion. Across GAIA and SWE-bench Verified, DeReAct improves Pass@1 most for weaker Brain models, with gains of 6.5--7.0 points for Qwen3-Coder-480B and 4.2--5.2 points for Claude Sonnet~4.5; gains diminish as Brain capability increases. Trajectory and ablation analyses show that external gating is effective when targeted failures are sufficiently prevalent and the gating policy is itself sufficient. With Claude Opus~4.5, Pass@1 remains comparable to ReAct, while DeReAct produces more evidence-complete and constraint-satisfying trajectories, indicating that completion control can trade earlier termination for stronger grounding. Overall, DeReAct improves weaker agents while retaining grounding benefits as models strengthen.
Chinese Translation
基于ReAct的代理通常依赖于单一的LLM策略来提出行动、与环境互动并决定任务何时完成。这种耦合使得行动授权和完成控制难以独立执行，从而导致错误传播和不支持的完成声明终止执行。我们提出了DeReAct，一种模块化代理架构，外部化了两个门控策略：一个在执行前验证提议行动的Critic，以及一个重建环境支持的状态并认证任务完成的Context Manager。在GAIA和SWE-bench Verified中，DeReAct在较弱的Brain模型上显著提高了Pass@1，Qwen3-Coder-480B的增益为6.5-7.0分，Claude Sonnet~4.5的增益为4.2-5.2分；随着Brain能力的增强，增益逐渐减小。轨迹和消融分析表明，当目标失败足够普遍且门控策略本身足够时，外部门控是有效的。在Claude Opus~4.5中，Pass@1与ReAct相当，而DeReAct产生了更多证据完整且满足约束的轨迹，表明完成控制可以在更强的基础上以更早的终止进行权衡。总体而言，DeReAct改善了较弱的代理，同时在模型增强时保留了基础支持的优势。
cs.AI / 9 / 2610.02372
Traversing the Satisfaction-Diversity Frontier in Text-to-Image Diffusion
在文本到图像扩散中的满意度-多样性边界探索
Abstract
Text-to-image generation enables users to explore several images generated from the same prompt. For these generated images to be useful, each one must reflect the user's preferences, measured by a learned reward, and differ visually from the others to maintain diversity. Existing methods are limited: they either address reward and diversity separately or combine them in one aggregate score, enabling high diversity to offset low rewards. In this paper, we address these limitations by formulating generation as satisficing: every image (candidate) must satisfy a reward floor and the batch of images must satisfy a diversity cutoff. The reward floor controls the balance between worst-candidate reward and batch diversity; we show that varying this floor defines a Pareto frontier. To traverse this frontier, we introduce SatisDive, a training-free inference-time method. SatisDive uses a batch-relative reward cutoff to distinguish lower- from higher-reward candidates, emphasizing reward improvement for candidates below the cutoff and diversity among candidates above it. On Pick-a-Pic, at matched DreamSim, SatisDive improves worst-candidate reward over FK steering by up to 0.43 with FLUX.1-dev as the base model and HPSv3 as the reward, and by up to 0.70 with SANA-1.6B as the base model and ImageReward as the reward. More broadly, across their overlapping DreamSim ranges, SatisDive's satisfaction-diversity curve Pareto-dominates FK steering's curve in each setting.
Chinese Translation
文本到图像生成使用户能够探索从同一提示生成的多个图像。为了使这些生成的图像有用，每个图像必须反映用户的偏好，这通过学习的奖励来衡量，并且在视觉上必须与其他图像有所不同，以保持多样性。现有方法存在局限性：它们要么分别处理奖励和多样性，要么将它们结合成一个综合得分，从而使高多样性能够抵消低奖励。在本文中，我们通过将生成过程表述为满意度满足来解决这些限制：每个图像（候选）必须满足奖励底线，而图像批次必须满足多样性阈值。奖励底线控制最差候选奖励与批次多样性之间的平衡；我们展示了改变这一底线可以定义一个帕累托前沿。为了在这一前沿上进行探索，我们引入了 SatisDive，这是一种无训练的推理时方法。SatisDive 使用批次相对奖励阈值来区分低奖励候选和高奖励候选，强调低于阈值的候选的奖励改进以及高于阈值的候选之间的多样性。在 Pick-a-Pic 上，在匹配的 DreamSim 下，SatisDive 在以 FLUX.1-dev 作为基础模型和 HPSv3 作为奖励的情况下，将最差候选奖励提高了最多 0.43，而在以 SANA-1.6B 作为基础模型和 ImageReward 作为奖励的情况下，提高了最多 0.70。更广泛地说，在它们重叠的 DreamSim 范围内，SatisDive 的满意度-多样性曲线在每种设置中都优于 FK steering 的曲线。
cs.AI / 10 / 2610.02378
THPL: A Vision-to-Language Decision Support Framework for Rainbow Trout Feeding Management in RAS
THPL：一种用于循环水养殖系统中虹鳟鱼饲养管理的视觉-语言决策支持框架
Abstract
In Recirculating Aquaculture Systems (RAS), precision feeding is critical for minimizing costs and improving fish welfare. However, existing methods lack cognitive alignment between fish behaviors and management knowledge, impeding translation into executable, interpretable feeding decisions. To address this, we propose THPL, a generative feeding decision framework tailored for rainbow trout (Oncorhynchus mykiss) in RAS. First, Fishsort extracts trajectories to establish an Activity Coefficient (AC) quantifying feeding intensity. Second, a Hierarchical Behavior Encoder (HBE) models individual temporal progression and collective dynamics using Temporal and Set Transformers, transforming trajectory tensors into dual-evidence representations of explicit physical and implicit soft tokens. Finally, these tokens are integrated with environmental parameters, metadata, and expert rules to fine-tune an LLM via LoRA, followed by counterfactual multimodal Direct Preference Optimization (mDPO) to reinforce causal reasoning. Results show that AC exhibits a statistically significant monotonic positive correlation with expert-annotated feeding intensity (Spearman $\rho = 0.925$, $p < 0.001$). Ablations indicate that decision accuracy improves from 33.33% (text-only baseline) to 93.33% with dual-evidence tokens, confirming that continuous spatiotemporal tokens provide necessary physical grounding for LLMs. Compared with standard LoRA, counterfactual mDPO elevates decision accuracy from 93.33% to 96.67%, advances METEOR from 58.10% to 85.30%, reduces Self-BLEU-2 from 58.79% to 52.88%, and increases Distinct-3 from 6.68% to 7.81%, suppressing templating and actuation biases while reinforcing causal consistency and operational safety. Overall, by integrating continuous kinematics with LLM reasoning, this study provides a novel decision support paradigm for precision aquaculture.
Chinese Translation
在循环水养殖系统（RAS）中，精准饲喂对于降低成本和改善鱼类福利至关重要。然而，现有方法在鱼类行为与管理知识之间缺乏认知对齐，妨碍了可执行、可解释的饲喂决策的转化。为了解决这一问题，我们提出了THPL，一种针对RAS中虹鳟鱼（Oncorhynchus mykiss）的生成性饲喂决策框架。首先，Fishsort提取轨迹以建立活动系数（Activity Coefficient, AC），量化饲喂强度。其次，层次行为编码器（Hierarchical Behavior Encoder, HBE）利用时间和集合变换器（Temporal and Set Transformers）对个体时间进程和集体动态进行建模，将轨迹张量转化为显性物理和隐性软标记的双证据表示。最后，这些标记与环境参数、元数据和专家规则相结合，通过LoRA对大型语言模型（LLM）进行微调，随后进行反事实多模态直接偏好优化（counterfactual multimodal Direct Preference Optimization, mDPO）以增强因果推理。结果显示，AC与专家标注的饲喂强度之间存在统计显著的单调正相关（Spearman $
ho = 0.925$, $p < 0.001$）。消融实验表明，决策准确率从33.33%（仅文本基线）提高到93.33%（使用双证据标记），确认连续时空标记为LLMs提供了必要的物理基础。与标准LoRA相比，反事实mDPO将决策准确率从93.33%提升至96.67%，METEOR从58.10%提升至85.30%，Self-BLEU-2从58.79%降低至52.88%，Distinct-3从6.68%增加至7.81%，抑制了模板化和激活偏差，同时增强了因果一致性和操作安全性。总体而言，通过将连续运动学与LLM推理相结合，本研究为精准水产养殖提供了一种新颖的决策支持范式。
cs.AI / 11 / 2610.02395
FlashSinkhorn 2: Block-Sparse Entropic Optimal Transport
FlashSinkhorn 2：块稀疏熵最优传输
Abstract
Streaming GPU solvers for entropic optimal transport (EOT), such as FlashSinkhorn, avoid storing the dense kernel but still evaluate all $n\times m$ point pairs in every Sinkhorn iteration. We present \textbf{FlashSinkhorn~2} (FS2), a solver for squared-Euclidean cost on low-dimensional point clouds that solves large discrete EOT problems to a prescribed marginal residual on a single GPU by coupling two stages. A coarse stage solves on cell centroids, lifts the potentials to every point and, when a sampled marginal check rejects the lift, continues on the centroids, replacing most point-level updates. A block-sparse fine stage then removes the centroid error that coarse updates cannot. Its Morton-ordered blocks support screening and fused tensor-core execution, and a threshold set by the block masses bounds each omitted tile's contribution to every row and column. On synthetic benchmarks, FS2 reaches the target residual on all 32 problems and GeomLoss multiscale on 10. On one A100, FS2 solves discrete EOT between two $1.34\times10^8$-particle measures from a cosmological $N$-body simulation, at an entropic blur equal to the mean interparticle distance, to an all-particle marginal residual below 0.01 in under 2.5 hours. To our knowledge, it is the largest discrete EOT problem solved to this accuracy within hours. For reproducibility, we release an open-source implementation at https://github.com/ot-triton-lab/flash-sinkhorn
Chinese Translation
用于熵最优传输（EOT）的流式GPU求解器，如FlashSinkhorn，避免存储密集核，但在每次Sinkhorn迭代中仍然评估所有$n imes m$点对。我们提出了 extbf{FlashSinkhorn~2}（FS2），这是一个针对低维点云的平方欧几里得成本的求解器，能够在单个GPU上通过耦合两个阶段来解决大型离散EOT问题，以达到规定的边际残差。粗略阶段在单元质心上求解，将势能提升到每个点，并在采样的边际检查拒绝提升时，继续在质心上进行，替代大部分点级更新。然后，块稀疏的精细阶段去除粗略更新无法消除的质心误差。其Morton排序的块支持筛选和融合张量核心执行，并且由块质量设定的阈值限制了每个被省略的块对每一行和列的贡献。在合成基准测试中，FS2在所有32个问题上达到了目标残差，并在10个问题上达到了GeomLoss多尺度。在一台A100上，FS2在熵模糊等于平均粒子间距的情况下，解决了来自宇宙学$N$-体模拟的两个$1.34 imes10^8$粒子测度之间的离散EOT问题，所有粒子的边际残差低于0.01，耗时不到2.5小时。据我们所知，这是在数小时内以此精度解决的最大离散EOT问题。为了可重复性，我们在https://github.com/ot-triton-lab/flash-sinkhorn发布了开源实现。
cs.AI / 12 / 2610.02405
When Terminal-Agent Training Stalls: Demystifying Data Generation and Verification Challenge
当终端代理训练停滞时：揭示数据生成与验证挑战
Abstract
Using a frontier model like Claude Opus as a meta-agent to generate terminal tasks and verifiers for RL training is increasingly common. Yet a runnable Docker image and executable test suite do not guarantee a faithful end-to-end pipeline for terminal agent training. We present a meta-agent pipeline motivated by this gap, diagnosing three classes of failure: benchmark invalidity, harness brittleness, and reward misalignment. Prompt redesign and context extension raise baseline solvability 5.6 times, but a 9B model saturates at 81.3% mean pass@2 within 20 steps on Claude Opus-generated tasks. Adding hard tasks reduces mean pass@2 to 20.6% without changing the training configuration, a strong evidence that the solvability band is model-specific. These findings demonstrate that meta-agent reliability requires solvability-band calibration, verifier audits, and infrastructure error accounting as first-class evaluation criteria, not post-hoc diagnost.
Chinese Translation
使用像 Claude Opus 这样的前沿模型作为元代理来生成终端任务和验证者以进行强化学习（RL）训练变得越来越普遍。然而，一个可运行的 Docker 镜像和可执行的测试套件并不能保证终端代理训练的端到端管道的可靠性。我们提出了一个由此差距驱动的元代理管道，诊断出三类失败：基准无效性、工具脆弱性和奖励不一致性。提示重新设计和上下文扩展将基线可解性提高了 5.6 倍，但一个 9B 模型在 Claude Opus 生成的任务上在 20 步内饱和于 81.3% 的平均通过率（mean pass@2）。添加困难任务使平均通过率降至 20.6%，而不改变训练配置，这强有力地证明了可解性区间是模型特定的。这些发现表明，元代理的可靠性需要将可解性区间校准、验证者审计和基础设施错误核算作为一流的评估标准，而不是事后诊断。
cs.AI / 13 / 2610.02452
Reinforcement Learning Techniques for the Optimization of Target Polarization in Nuclear Physics Scattering Experiments
用于优化核物理散射实验中目标极化的强化学习技术
Abstract
The operation of dynamically polarized targets in nuclear physics experiments relies on continuous tuning of the microwave frequency to compensate for radiation damage and evolving material properties, a task that is traditionally performed through manual trial-and-error by expert operators. This work presents a data-driven control framework that combines surrogate modeling with reinforcement learning to optimize the target polarization. Using operational data from the APOLLO cryogenic target system, we train and evaluate multilayer perceptron and Gaussian process regression models to predict polarization as a function of microwave frequency, beam current, and accumulated radiation dose. We show that Gaussian process-based models provide calibrated uncertainty estimates and reliably identify regions outside the training distribution, while MLPs exhibit limited sensitivity to distributional shift. To enable learning and control across multiple target samples, we introduce a Gaussian process approximation and embed the surrogate model within a standardized simulation environment. A reinforcement learning agent is trained using a lower-confidence-bound reward formulation that balances performance maximization against uncertainty. We are able to show an almost 2x improvement on the operators actions utilizing our RL agent.
Chinese Translation
在核物理实验中，动态极化目标的操作依赖于对微波频率的持续调节，以补偿辐射损伤和不断变化的材料特性，这一任务传统上由专家操作员通过手动试错进行。本文提出了一种数据驱动的控制框架，该框架结合了代理建模和强化学习，以优化目标极化。利用APOLLO低温目标系统的操作数据，我们训练和评估了多层感知器（MLP）和高斯过程回归模型，以预测极化与微波频率、束流电流和累积辐射剂量之间的关系。我们展示了基于高斯过程的模型提供了经过校准的不确定性估计，并可靠地识别出训练分布之外的区域，而MLP对分布变化的敏感性有限。为了实现对多个目标样本的学习和控制，我们引入了高斯过程近似，并将代理模型嵌入标准化的仿真环境中。我们使用低置信界奖励公式训练强化学习代理，该公式在性能最大化与不确定性之间取得平衡。我们能够展示，利用我们的强化学习代理，操作员的行动几乎提高了2倍。
cs.AI / 14 / 2610.02478
Tropical Reinforcement Learning
热带强化学习
Abstract
Reinforcement learning for large language models typically maximizes expected return, adding up the probabilities of all successful trajectories. However, the classical sum formulation can only report how often the model policy succeeds, not which solution actually worked, and because probabilities sum to one, reinforcing one solution can make the model forget another that was never shown to be wrong. This makes expected return a poor fit for compositional reasoning, where a solution must be assembled from reasoning steps that the model produces in separate, often failed, attempts but rarely produces together. To address this, we propose Tropical Reinforcement Learning, which rests on a simple change of algebra: instead of adding the probabilities of alternative solutions, we take their maximum, which yields the tropical semiring. The value of a state then becomes the log-probability of its most likely verified solution, together with an explicit path that can be replayed and reused. This enables true composition, since the best prefix and the best suffix meeting at a shared state can be joined even when they come from different rollouts. To put this into practice, we introduce TROPIC, a training algorithm for deterministic, resettable environments with verifiable outcomes. On four agentic tasks (Sokoban, Countdown, FrozenLake, WebShop), TROPIC outperforms the strongest on-policy baselines by up to 16 percentage points. Changing the algebra of reinforcement learning, not just its estimators, can thus substantially improve compositional reasoning in language models
Chinese Translation
大型语言模型的强化学习通常是最大化期望回报，通过累加所有成功轨迹的概率。然而，传统的求和形式只能报告模型策略成功的频率，而不能指出实际有效的解决方案，并且由于概率总和为一，强化某一解决方案可能会使模型遗忘另一个从未被证明错误的方案。这使得期望回报不适合组合推理，其中一个解决方案必须由模型在不同的、通常是失败的尝试中产生的推理步骤组合而成，但这些步骤很少能一起产生。为了解决这个问题，我们提出了热带强化学习（Tropical Reinforcement Learning），其基础是简单的代数变更：我们不再累加替代解决方案的概率，而是取它们的最大值，从而得出热带半环（tropical semiring）。一个状态的值因此变为其最可能的验证解决方案的对数概率，并附带一个可以重播和重用的明确路径。这使得真正的组合成为可能，因为在共享状态下相遇的最佳前缀和最佳后缀即使来自不同的回滚也可以结合在一起。为了将这一理论付诸实践，我们引入了TROPIC，一种用于具有可验证结果的确定性、可重置环境的训练算法。在四个代理任务（Sokoban、Countdown、FrozenLake、WebShop）中，TROPIC的表现超过了最强的在线基准，提升幅度高达16个百分点。因此，改变强化学习的代数，而不仅仅是其估计器，可以显著改善语言模型中的组合推理。
cs.AI / 15 / 2610.02480
MEA: A Reward-Driven Multi-Agent System for Faithful Model Explanations
MEA：一种基于奖励驱动的多智能体系统，用于可信的模型解释
Abstract
Recent years have seen the employment of a plethora of machine learning (ML) models in high-stakes domains, but they remain largely opaque to the practitioners who act on their predictions. While post-hoc explanation methods offer a lens into this model behavior, wielding them effectively demands expertise most domain experts lack: navigating high-dimensional outputs, selecting the best explanations, and synthesizing evidence across disparate tools. To this end, we present MEA, a multi-agent framework that removes the explanation knowledge barrier entirely: a Proposer agent selects and configures explanation tools based on the question and modality, while an Actor agent is optimized end-to-end against faithfulness, transforming the outputs into natural language explanations grounded in model behavior across tabular, text, and vision modalities. Further, we introduce diverse question types spanning feature attribution, counterfactual reasoning, and spurious feature detection, each paired with a perturbation-based faithfulness metric. We find that frontier LLMs systematically produce unfaithful explanations. By optimizing against faithfulness rewards augmented with a modality-adaptive penalty, MEA consistently outperforms post hoc explainers, agentic, and closed-source baselines across six datasets, with reward-driven optimization yielding faithfulness gains of +28% (tabular), +21% (text), and +34% (vision) over the untrained backbone. More broadly, our findings suggest that AI agents themselves can serve as a scalable, adaptable interface to ML explainability, opening a path toward natural-language explainability that generalizes beyond the fixed, single-purpose tools that have long defined the field.
Chinese Translation
近年来，机器学习（ML）模型在高风险领域的应用日益增多，但对于依赖其预测的从业者而言，这些模型仍然在很大程度上是不可理解的。虽然事后解释方法提供了对模型行为的洞察，但有效地使用这些方法需要大多数领域专家所缺乏的专业知识：处理高维输出、选择最佳解释以及在不同工具之间综合证据。为此，我们提出了MEA，一个完全消除解释知识障碍的多智能体框架：提议者（Proposer）智能体根据问题和模态选择和配置解释工具，而行动者（Actor）智能体则针对可信度进行端到端优化，将输出转化为基于模型行为的自然语言解释，涵盖表格、文本和视觉模态。此外，我们引入了涵盖特征归因、反事实推理和虚假特征检测的多样化问题类型，每种类型都配有基于扰动的可信度度量。我们发现前沿的大型语言模型（LLMs）系统性地产生不可信的解释。通过优化与模态自适应惩罚相结合的可信度奖励，MEA在六个数据集上始终优于事后解释器、智能体和闭源基线，基于奖励驱动的优化在未训练的基础模型上实现了+28%（表格）、+21%（文本）和+34%（视觉）的可信度提升。更广泛地说，我们的研究结果表明，人工智能智能体本身可以作为机器学习可解释性的可扩展、适应性接口，为超越长期以来定义该领域的固定单一目的工具的自然语言可解释性开辟了一条新路径。
cs.AI / 16 / 2610.02491
What Does a Token Cost? A Mixture-of-Agents Measurement of Sufficient Per-Token Compute
一个标记的成本是多少？基于混合代理的每标记计算充分性测量
Abstract
Large language models spend the same amount of computation on every token they generate, regardless of how difficult each token is to produce. Methods such as speculative decoding and model routing are built on the premise that much of this computation is unnecessary, yet the computation an individual token actually requires has not been measured. We measure it through a Mixture-of-Agents (MoA) lens: a panel of fifteen language models of increasing capacity, drawn from three families, in which every agent attempts to reproduce a reference sequence token by token, conditioned on the correct preceding tokens. We define the inference cost of the smallest agent that succeeds as the token's sufficient compute, which upper-bounds what the token requires. On three core benchmarks, a 0.5B agent reproduces 92--95\% of reference tokens. Across Qwen, OLMo, and R1-distilled panels, the most expensive 10\% account for 64--80\% of estimated FLOPs. On all 500 MATH-500 problems, the MoA-derived map helps model routing reduce projected latency from 7.59 to 5.12 seconds while slightly improving accuracy, relative to the best confidence-routing baseline. The MoA-map helps drafting use 32.6\% fewer draft tokens and approximately 20\% lower projected latency than fixed-window drafting at similar accuracy. These comparisons reveal remaining allocation headroom, motivating controllers that exploit sufficient-compute structure.
Chinese Translation
大型语言模型在生成每个标记时消耗相同数量的计算，无论每个标记的生成难度如何。诸如推测解码和模型路由等方法建立在这样一个前提上：大部分计算是多余的，但单个标记实际所需的计算量尚未被测量。我们通过混合代理（Mixture-of-Agents, MoA）的视角进行测量：一个由十五个逐渐增强能力的语言模型组成的面板，来自三个家族，其中每个代理尝试在正确的前置标记条件下逐个重现参考序列。我们将成功的最小代理的推理成本定义为标记的充分计算，这为标记所需的计算量设定了上限。在三个核心基准测试中，一个0.5B的代理重现了92%到95%的参考标记。在Qwen、OLMo和R1蒸馏面板中，最昂贵的10%占估计FLOPs的64%到80%。在所有500个MATH-500问题中，基于MoA的映射帮助模型路由将预计延迟从7.59秒减少到5.12秒，同时相对于最佳的置信度路由基线略微提高了准确性。MoA映射帮助草拟使用比固定窗口草拟少32.6%的草拟标记，并在相似准确性下预计延迟降低约20%。这些比较揭示了剩余的分配空间，激励控制器利用充分计算结构。
cs.AI / 17 / 2610.02492
Right Order, Wrong Scale: Auditing LLM Judges for Occupational AI Measurement
正确的顺序，错误的尺度：对职业人工智能测量的 LLM 评审的审计
Abstract
LLM judges are increasingly used to assess whether AI outputs meet workplace requirements, but agreement on response rankings does not establish agreement on acceptance rates or occupational aggregates. We introduce O*NET-BENCH, an audit suite derived from an existing survey of 45,796 worker ratings, and evaluate 33 pre-existing judge configurations across six model families on 4,501 test ratings. Twenty-five configurations achieve tie-aware pair accuracy of at least 0.60, although a train-fitted response-only TF-IDF baseline nearly matches the strongest judge. Despite this ordering agreement, judges estimate that 3.0%-97.9% of responses are acceptable, compared with 61.1% for occupation-matched workers. In one fine-tuned lineage, changing from pointwise scoring to a bundled few-shot/listwise protocol improves response ordering while reducing agreement with worker means at the task and occupation levels; this reversal replicates on a task- and worker-disjoint validation split under prespecified criteria. Cross-validated calibration largely removes mean bias, but calibrated scores explain at most 8.5% of individual worker-rating variance. Prediction-assisted estimation yields at most small precision gains at the studied label budgets. These results show that ranking agreement alone is insufficient for occupational measurement. Judges should be validated against the acceptance rates and aggregates their scores will be used to estimate.
Chinese Translation
LLM 评审越来越多地用于评估人工智能输出是否符合工作场所要求，但对响应排名的共识并不能确立对接受率或职业总量的共识。我们引入了 O*NET-BENCH，这是一个基于对 45,796 名工人评分的现有调查衍生的审计套件，并在 4,501 个测试评分上评估了 33 种预先存在的评审配置，涵盖六个模型家族。虽然有 25 种配置实现了至少 0.60 的考虑平局的配对准确率，但一个经过训练的仅基于响应的 TF-IDF 基线几乎与最强的评审相匹配。尽管存在这种排序一致性，评审估计 3.0%-97.9% 的响应是可接受的，而职业匹配工人的接受率为 61.1%。在一个经过微调的谱系中，从逐点评分转变为捆绑的少样本/逐列表协议改善了响应排序，同时在任务和职业层面上减少了与工人均值的一致性；这种反转在预先指定的标准下，在任务和工人不重叠的验证分割上得到了重复。交叉验证的校准在很大程度上消除了均值偏差，但校准后的评分最多只能解释 8.5% 的个体工人评分方差。预测辅助估计在研究的标签预算下最多只能带来小幅的精度提升。这些结果表明，仅靠排名一致性不足以进行职业测量。评审应当根据接受率和他们的评分将用于估计的总量进行验证。
cs.AI / 18 / 2610.02496
"I just assumed that it would translate": examining MT risk awareness among healthcare staff with abbreviations as a use case
“我只是认为它会翻译”：考察医疗保健人员对机器翻译风险的认知，以缩写作为案例
Abstract
In the UK, public healthcare staff report turning to machine translation (MT) - predominantly Google Translate (GT) - to communicate with patients across language barriers. Though intended to support their duty of care, potentially uninformed reliance on MT in such contexts could have serious consequences for patient safety. Research nonetheless remains limited on staff awareness of the possible risks posed by higher-stakes MT use in general and with patient medical records in particular, most existing literature instead examining its use in interpersonal situations or with patient-oriented documentation. Moreover, medical abbreviations are well-documented as increasing patient risk even monolingually, with outcomes from their misuse and/or misinterpretation ranging from temporary harm to the death of the patient. Abbreviations were therefore selected as a use case for identifying the potential risks posed by their translation with MT. Contextualised French and Spanish data examples drawn from authoritative clinical corpora and translated via GT were presented during semi-structured interviews to 21 healthcare staff participants in diverse roles and specialties. The results were then subject to qualitative analysis and cross-analysis.
Chinese Translation
在英国，公共医疗保健人员报告称，他们在跨语言障碍与患者沟通时，主要依赖机器翻译（MT）——尤其是谷歌翻译（GT）。尽管这一工具旨在支持他们的护理职责，但在此类情境中对机器翻译的潜在无知依赖可能对患者安全产生严重后果。然而，关于工作人员对高风险机器翻译使用可能带来的风险的认知，尤其是在患者医疗记录方面的研究仍然有限，现有文献大多集中于其在面对面交流或患者导向文档中的使用。此外，医学缩写已被充分记录为即使在单语环境中也会增加患者风险，其误用和/或误解的后果可能从暂时伤害到患者死亡。因此，缩写被选为识别其通过机器翻译翻译所带来的潜在风险的案例。在半结构化访谈中，向21名来自不同角色和专业的医疗保健人员参与者展示了从权威临床语料库中提取的上下文化的法语和西班牙语数据示例，并通过谷歌翻译进行了翻译。然后对结果进行了定性分析和交叉分析。
cs.AI / 19 / 2610.02504
HXAI: Hierarchical Privacy-Preserving Explainable AI in Distributed Energy Systems
HXAI：分布式能源系统中的层次隐私保护可解释人工智能
Abstract
Balancing electricity demand and supply is increasingly difficult due to the inherent intermittency of renewable power generation and the stochastic power consumption. Grid operators require fine-grained, decision-relevant insights into household energy consumption to manage peak loads and design responsive tariffs, but increased transparency at this level raises significant privacy concerns. Traditional methods for explainable AI (XAI) can reveal sensitive information, while standard privacy techniques often reduce the usefulness of explanations. To address this issue, we introduce HXAI, a hierarchical framework that preserves privacy while enabling reasonable explainable analysis for grid-level demand management. HXAI consists of two main components: (1) a local model that generates fine-grained explanations within a secure, private environment, and (2) a zonal model that aggregates these explanations to support grid-level analysis while enforcing privacy through flexible privacy-budget management. We explicitly limit cumulative privacy exposure under repeated operator queries and show that the proposed framework preserves decision-relevant information without compromising household privacy. Experiments on both simulated and real-world energy datasets demonstrate that HXAI provides useful insights for zonal load management while ensuring that appliance-level consumption remains local and is never transmitted to grid operators. Our results show that preserving the semantic structure of explanations, rather than minimizing numerical error, is the key to XAI under differential privacy. This framework provides a way to achieve both privacy and explainability in energy management.
Chinese Translation
由于可再生能源发电的固有间歇性和随机电力消费，平衡电力需求与供应变得越来越困难。电网运营商需要对家庭能源消费进行细粒度的、与决策相关的洞察，以管理峰值负荷和设计响应性电价，但在这一层面上增加透明度会引发重大隐私问题。传统的可解释人工智能（XAI）方法可能会泄露敏感信息，而标准的隐私技术往往会降低解释的实用性。为了解决这一问题，我们提出了HXAI，一个在保护隐私的同时，能够进行合理可解释分析的层次框架，以支持电网级需求管理。HXAI由两个主要组件组成：（1）一个在安全、私密环境中生成细粒度解释的本地模型，以及（2）一个聚合这些解释以支持电网级分析的区域模型，同时通过灵活的隐私预算管理来强化隐私。我们明确限制在重复运营商查询下的累积隐私暴露，并展示所提出的框架在不妨碍家庭隐私的情况下保留与决策相关的信息。在模拟和真实世界的能源数据集上的实验表明，HXAI为区域负荷管理提供了有用的洞察，同时确保家电级别的消费保持本地化，且从未传输给电网运营商。我们的结果表明，保留解释的语义结构，而不是最小化数值误差，是在差分隐私下实现XAI的关键。该框架为在能源管理中实现隐私和可解释性提供了一种方法。
cs.AI / 20 / 2610.02508
World Action Modeling with Progressive Visual Planning
渐进式视觉规划的世界动作建模
Abstract
World action models (WAMs) have emerged as a promising paradigm for robotic control by jointly predicting future visual dynamics and actions from an initial observation and instruction. However, existing WAMs struggle with long-horizon prediction, as generating dense video rollouts is highly inefficient. Some recent WAMs address this by predicting a single future frame without generating the full video, but this approach neglects how to progress toward the goal. We present ProWAM, a progressive world action model that jointly predicts actions and an ordered sequence of sparse visual sub-goals, providing explicit visual guidance to anchor action generation throughout task execution. This design scales naturally, as sub-goal prediction can be learned from large-scale action-free videos, allowing the video backbone to offload complex visual planning from the action policy. For efficient action generation, ProWAM executes a single video-backbone forward pass to cache sparse sub-goal features, eliminating iterative full-video generation and requiring only lightweight action denoising during replanning. Across extensive evaluations, ProWAM achieves superior out-of-distribution robustness. On simulation benchmarks, it sets new state-of-the-art results on LIBERO-Plus (85.8%) and randomized RoboTwin (75.7%), outperforming the strongest baseline with relative gains of up to +35.9%. On RoboCasa365, ProWAM achieves a 48.1% success rate and 18.2% on the challenging Composite-Unseen split, ranking 4th overall. Crucially, in zero-shot real-world experiments, ProWAM achieves 70.0% success, outperforming the strongest baseline by +15.0 (from 55.0% to 70.0%, a +27.3% relative gain) in novel scenes. These results demonstrate the value of progress-indexed visual foresight for closed-loop control. Our program is in https://sii-ferenas.github.io/ProWAM-page.
Chinese Translation
世界动作模型（WAMs）作为一种有前景的机器人控制范式，通过从初始观察和指令中共同预测未来的视觉动态和动作而受到关注。然而，现有的WAMs在长时间预测方面存在困难，因为生成密集的视频回放效率极低。一些近期的WAMs通过预测单个未来帧而不生成完整视频来解决这一问题，但这种方法忽视了如何朝目标推进。我们提出了ProWAM，一种渐进式世界动作模型，它共同预测动作和稀疏视觉子目标的有序序列，为任务执行中的动作生成提供明确的视觉指导。这种设计自然可扩展，因为子目标预测可以从大规模无动作视频中学习，使得视频骨干网络能够将复杂的视觉规划从动作策略中卸载。为了高效生成动作，ProWAM执行单次视频骨干前向传递以缓存稀疏子目标特征，消除了迭代生成完整视频的需求，并且在重新规划时仅需轻量级的动作去噪。在广泛的评估中，ProWAM在分布外的鲁棒性上表现优越。在模拟基准测试中，它在LIBERO-Plus（85.8%）和随机化RoboTwin（75.7%）上创造了新的最先进结果，超越了最强基线，获得了高达+35.9%的相对增益。在RoboCasa365上，ProWAM的成功率为48.1%，在具有挑战性的Composite-Unseen分割上为18.2%，整体排名第4。关键的是，在零样本的真实世界实验中，ProWAM的成功率达到70.0%，在新场景中超越了最强基线+15.0（从55.0%提升到70.0%，相对增益+27.3%）。这些结果展示了进度索引视觉前瞻性在闭环控制中的价值。我们的程序可在https://sii-ferenas.github.io/ProWAM-page获取。
cs.AI / 21 / 2610.02510
On-Premises Multi-Course RAG Tutoring for Business Education: Hardware-Software Trade-offs in a Campus AI Tutor
校园内多课程RAG辅导在商业教育中的应用：校园AI辅导的硬件-软件权衡
Abstract
Campus AI tutors based on retrieval-augmented generation (RAG) must ground answers in assigned course materials while keeping textbooks and student dialogue on institutional infrastructure. We present CourseChat, an on-premises, multi-course RAG tutor for undergraduate business education, deployed behind a campus web gateway and intended for use embedded in Moodle. Six isolated course offerings, each keyed by its own course reference number (CRN), share twin-edge AI hosts running a FastAPI service, a local vector database, and a local large language model (LLM) served by Ollama. We report two generation-model bake-off rounds, a separate fixed-evidence source-fidelity comparison, and conversation and quiz audits. Several larger models failed the classroom speed gate, but a 12B model and a 7B alternative passed. A separate mixture-of-experts candidate improved some corrections while introducing new factual and continuity errors. We therefore retain the 8B production model pending a demonstrated overall improvement, rather than claiming that 8B is universally optimal. Software changes improved follow-up topic resolution while preserving course scope; 435 prebuilt questions across 65 modules decouple practice from live generation. The results support treating model choice, evidence selection, serving compatibility, and product design as a joint engineering decision. They do not establish learning gains: faculty ratings, peak-load capacity, and complete public-gateway acceptance remain separate evaluation needs.
Chinese Translation
基于检索增强生成（RAG）的校园AI辅导必须将答案基于指定的课程材料，同时保持教科书和学生对话在机构基础设施之上。我们提出了CourseChat，这是一种针对本科商业教育的校园内多课程RAG辅导，部署在校园网络网关后面，旨在嵌入Moodle中使用。六个独立的课程提供，每个课程都有自己的课程参考编号（CRN），共享双边AI主机，运行FastAPI服务、本地向量数据库和由Ollama提供的本地大型语言模型（LLM）。我们报告了两轮生成模型的比较、一个独立的固定证据源保真度比较，以及对话和测验审计。几个较大的模型未能通过课堂速度测试，但一个12B模型和一个7B替代模型通过了测试。一个独立的专家混合候选模型在改正某些错误时有所改善，但也引入了新的事实和连续性错误。因此，我们保留8B生产模型，待其显示出整体改善，而不是声称8B是普遍最优的。软件更改改善了后续主题的解决，同时保持了课程范围；在65个模块中预构建的435个问题将练习与实时生成解耦。结果支持将模型选择、证据选择、服务兼容性和产品设计视为联合工程决策。它们并未确立学习收益：教师评分、峰值负载能力和完整的公共网关接受度仍然是独立的评估需求。
cs.AI / 22 / 2610.02523
Hypothesis-guided discovery of cognitive algorithms via program refinement
通过程序细化进行假设引导的认知算法发现
Abstract
Developing cognitive models of algorithmic reasoning from behavioral data is a central problem in cognitive science that challenges current methods. Traditional approaches to cognitive modeling are interpretable and benefit from human expertise, but lack flexibility and scalability. Emerging techniques using large language models (LLMs) for de novo generation of cognitive models are scalable and flexible, but lack a role for human expertise and have mostly been applied to simpler tasks than algorithm recovery. We propose a hybrid system that treats discovery of cognitive algorithms as a program refinement problem. Human-created cognitive models are expressed as probabilistic programs and provided to a system of LLM agents with a mandate to: identify mismatches between model and behavior; propose code-level modifications within researcher-specified constraints; and verify structural fidelity. Revisions propagate to a probabilistic inference module that performs inference for latent variables and data likelihood computations. We evaluate the pipeline on human behavior in a problem-solving paradigm that exposes a variety of cognitive algorithms. Revised models consistently improve model fit relative to ancestral models and reveal a small set of recurring innovations that capture meaningful behavioral variability in this task.
Chinese Translation
从行为数据中开发算法推理的认知模型是认知科学中的一个核心问题，挑战着当前的方法。传统的认知建模方法具有可解释性，并受益于人类专业知识，但缺乏灵活性和可扩展性。使用大型语言模型（LLMs）进行新认知模型生成的新兴技术具有可扩展性和灵活性，但缺乏人类专业知识的作用，并且主要应用于比算法恢复更简单的任务。我们提出了一种混合系统，将认知算法的发现视为程序细化问题。人类创建的认知模型被表示为概率程序，并提供给一组LLM代理，任务是：识别模型与行为之间的不匹配；在研究者指定的约束内提出代码级修改；并验证结构的保真性。修订内容传播到一个概率推理模块，该模块执行潜变量和数据似然计算的推理。我们在一个问题解决范式中评估该流程，该范式揭示了多种认知算法。修订后的模型相对于祖先模型始终改善模型拟合，并揭示了一小组反复出现的创新，捕捉到该任务中有意义的行为变异性。
cs.AI / 23 / 2610.02525
Learning What to Investigate Next: Meta-Reasoning for Long-Horizon Research Agents
学习下一步调查内容：长时间研究代理的元推理
Abstract
Long-horizon research agents must decide both how to investigate and what to investigate next as evidence accumulates. This is hard to learn because such decisions are sparse in long execution traces, and their consequences may emerge several investigations later. We introduce Meta-reasoning for Iterative Research Agents (MIRA), a hierarchical architecture separating research allocation from execution. An outer-loop meta-reasoner curates context from a persistent research record, then writes a work order for the next investigation or ends the episode. A fresh inner-loop executor carries out each work order, making execution part of the transition between meta-reasoning actions. Without policy training, MIRA improves long-horizon inference and allocates additional compute more effectively in theorem proving and open-ended neural-architecture research. Its decision boundaries also provide natural units for credit assignment. At each boundary, we train a generative critic to forecast expected remaining return from partial states, outperforming token-level alternatives. Cross-environment pretraining improves forecasting and adaptation, yielding a transferable prior for valuing partial progress. We use this prior to initialize MIRA-AC, a generative actor-critic jointly trained to forecast remaining return and choose the next investigation, without a separate critic model. MIRA-AC concentrates policy optimization on meta-reasoning decisions, enabling efficient long-horizon reinforcement learning without directly optimizing the longer execution traces they initiate. Training MIRA-AC on the model's own proxy hill-climbing signals improves gold performance across four autoresearch environments; the actor transfers with cross-environment value initialization. Together, these results show that meta-reasoning can be learned as an explicit policy for directing long-horizon autonomous research.
Chinese Translation
长时间研究代理在证据积累的过程中必须决定如何调查以及接下来调查什么。这种决策难以学习，因为在长执行轨迹中，这类决策是稀疏的，其后果可能在几次调查之后才会显现。我们提出了迭代研究代理的元推理（Meta-reasoning for Iterative Research Agents, MIRA），这是一种将研究分配与执行分开的层次架构。外层元推理器从持久的研究记录中整理上下文，然后为下一次调查写出工作指令或结束该轮次。一个新的内层执行器执行每个工作指令，将执行作为元推理动作之间的过渡部分。在没有策略训练的情况下，MIRA改善了长时间推理，并在定理证明和开放式神经架构研究中更有效地分配额外计算资源。其决策边界也为信用分配提供了自然单位。在每个边界上，我们训练一个生成性评论员，以预测从部分状态中获得的预期剩余收益，超越了基于令牌的替代方案。跨环境的预训练提高了预测和适应能力，产生了一个可转移的先验，用于评估部分进展。我们利用这个先验初始化了MIRA-AC，一个生成性演员-评论员，联合训练以预测剩余收益并选择下一次调查，而无需单独的评论员模型。MIRA-AC将策略优化集中在元推理决策上，使得在不直接优化其启动的较长执行轨迹的情况下，实现高效的长时间强化学习。在模型自身的代理爬坡信号上训练MIRA-AC，提高了四个自我研究环境中的黄金表现；演员通过跨环境价值初始化进行迁移。综合来看，这些结果表明，元推理可以作为一种明确的策略被学习，以指导长时间的自主研究。
cs.AI / 24 / 2610.02542
How To Train Your World Model: Fine-tuning vs RAG for LM-based World Modeling
如何训练你的世界模型：基于语言模型的世界建模中的微调与检索增强生成的比较
Abstract
World models (WMs) simulate the transition dynamics of environments, enabling agents to plan over the consequences of their actions. In text-based environments, fine-tuning a Language Model (LM) to serve as a WM has emerged as a dominant paradigm. However, despite the widespread success of non-parametric approaches such as Retrieval Augmented Generation (RAG), retrieval for LM-based world modelling remains underexplored. We conduct a systematic evaluation across five diverse environments spanning embodied, web navigation and social settings, comparing fine-tuning and RAG-based approaches for LM-based world modelling. Our study reveals that fine-tuning often outperforms RAG, with fine-tuned WMs enabling agents to obtain higher rewards on 15/20 settings. While both construction paradigms benefit from additional and more diverse exploration, RAG-based approaches prove more data-efficient, and fine-tuning approaches disproportionately benefit from scaling the amount of experience collected. With a focus on RAG-based WMs, we devise a procedure that uses counterfactual intervention to estimate the error rate of the retrieval stage, and show that retrievers consistently surface suboptimal transitions from the experience buffer. Hoping to address this failing, we study a variety of query reformulation strategies, demonstrating that a hierarchical approach outperforms the traditional retrieval pipeline. Finally, we compose our findings into a hybrid world modelling system that parametrically captures core environment dynamics, while learning to rely on retrieval from an actively maintained memory store. Our hybrid system consistently outperforms other methods across multiple environments and models, showcasing the robustness of the approach and the applicability of our findings.
Chinese Translation
世界模型（World Models, WMs）模拟环境的转移动态，使得智能体能够对其行为的后果进行规划。在基于文本的环境中，将语言模型（Language Model, LM）进行微调以作为世界模型已成为一种主流范式。然而，尽管检索增强生成（Retrieval Augmented Generation, RAG）等非参数方法取得了广泛成功，基于语言模型的世界建模中的检索仍然未得到充分探索。我们在五个不同的环境中进行了系统评估，这些环境涵盖了具身、网络导航和社交场景，比较了基于微调和基于RAG的方法在基于语言模型的世界建模中的表现。我们的研究表明，微调通常优于RAG，微调后的世界模型使智能体在15/20个设置中获得了更高的奖励。虽然这两种构建范式都受益于额外和更广泛的探索，但基于RAG的方法在数据效率上表现更佳，而微调方法则在收集经验的数量上受益更大。针对基于RAG的世界模型，我们设计了一种程序，利用反事实干预来估计检索阶段的错误率，并表明检索器在经验缓冲区中持续呈现次优转移。希望解决这一问题，我们研究了多种查询重构策略，证明分层方法优于传统的检索流程。最后，我们将我们的发现整合成一个混合世界建模系统，该系统参数化地捕捉核心环境动态，同时学习依赖于主动维护的记忆存储中的检索。我们的混合系统在多个环境和模型中始终优于其他方法，展示了该方法的鲁棒性和我们发现的适用性。
cs.AI / 25 / 2610.02557
How to Have a Sensitive Debate: An Instance-Optimal Protocol for AI Debate
如何进行敏感的辩论：一种实例最优的人工智能辩论协议
Abstract
As powerful AI systems reach and sometimes surpass the abilities of human experts across a range of cognitively demanding tasks, the problem of accurate oversight and supervision of these systems has become increasingly urgent. One promising approach is AI debate, which seeks to leverage a debate between two powerful AIs to break complex questions down into simpler claims that can be easily judged directly. Theoretical work on debate has formalized this intuition in the language of computational complexity theory, where the goal is to design protocols (i.e., rules of the debate game) that provide rigorous guarantees on correctness for judging solutions to complex problems with limited supervision. Specifically, the current best protocol has been shown to work for all problems that have sufficiently stable decompositions into subproblems. In this paper, we design a new protocol for this same class of problems that improves on the prior work in several ways. First, correctness holds in a worst-case rather than an average-case sense. Second, being honest and correct is a dominant-strategy equilibrium for both debaters, rather than a Stackelberg equilibrium. Finally, we prove black-box lower bounds, showing that our new protocol is instance-wise optimal. That is, no protocol for this class of problems can outperform ours while making only black-box queries to human judgments. We obtain these results by relating the notion of stable problem decompositions to the concept of fractional block sensitivity from query complexity.
Chinese Translation
随着强大的人工智能系统在许多认知要求高的任务中达到甚至超越人类专家的能力，对这些系统进行准确监督和管理的问题变得愈发紧迫。一种有前景的方法是人工智能辩论，它试图利用两个强大人工智能之间的辩论，将复杂问题分解为可以直接判断的简单主张。关于辩论的理论研究已在计算复杂性理论的语言中形式化了这一直觉，其目标是设计协议（即辩论游戏的规则），为在有限监督下判断复杂问题的解决方案提供严格的正确性保证。具体而言，目前最佳的协议已被证明适用于所有具有足够稳定的子问题分解的复杂问题。在本文中，我们为同一类问题设计了一种新协议，在多个方面改进了先前的工作。首先，正确性在最坏情况下成立，而不是平均情况。其次，诚实和正确是两个辩论者的主导策略均衡，而不是斯塔克尔堡均衡。最后，我们证明了黑箱下界，表明我们的新协议在实例层面上是最优的。也就是说，对于这一类问题，没有任何协议能够在仅对人类判断进行黑箱查询的情况下超越我们的协议。我们通过将稳定问题分解的概念与查询复杂性中的分数块敏感性概念相关联来获得这些结果。
cs.AI / 26 / 2610.02568
Mitigating Social Sycophancy via Pluralistic Preference Optimization
通过多元偏好优化缓解社会谄媚现象
Abstract
Personal advice, including relationship advice, now ranks among the most common uses of generative AI. But language models (LMs) exhibit sycophancy: they affirm users much more often than humans do, which can make people overconfident and less willing to repair their relationships after a conflict. Prior work on mitigating sycophancy has focused on factual settings where a response can be checked against a ground truth answer, while mitigations for social sycophancy (e.g., personal advice, where there is no ground truth) have relied on simple prompting and post-training methods with limited effectiveness. Our insight is that social sycophancy occurs in part because LMs overly center on the user and fail to consider the perspectives of other stakeholders impacted by the user's behavior. To address this problem we propose Pluralistic Preference Optimization (PlurPO): given inputs describing interpersonal conflicts, the LM identifies and simulates the relevant stakeholders, and is then trained to prefer and generate responses acceptable to all stakeholders. PlurPO uses only signals the model produces about its own outputs, without ground-truth labels. PlurPO substantially reduces social sycophancy across four datasets and four model families compared to prior methods. For example, on statements of intent to cause harm, where the users' actions should not be endorsed, PlurPO reduces the endorsement rate by 89% on average across four models. On general advice questions, where the target is to match the endorsement rate of human responses, it closes the gap by more than half, from 17.8% to 8.0% on average. The preference dataset constructed by PlurPO for an 8B model also effectively transfers to mitigating sycophancy in a larger (32B) model. Our results indicate that social sycophancy can be reduced by leveraging a model's own capabilities to simulate a plurality of relevant perspectives.
Chinese Translation
个人建议，包括人际关系建议，如今已成为生成式人工智能最常见的应用之一。然而，语言模型（LM）表现出谄媚行为：它们比人类更频繁地肯定用户，这可能导致人们过于自信，并在冲突后不愿修复关系。以往关于缓解谄媚的研究主要集中在可以与真实答案进行核对的事实性场景，而对社会谄媚（例如个人建议，因没有真实答案而难以核对）的缓解方法则依赖于简单的提示和后训练方法，效果有限。我们的见解是，社会谄媚部分发生是因为语言模型过于关注用户，未能考虑受用户行为影响的其他利益相关者的观点。为了解决这一问题，我们提出了多元偏好优化（Pluralistic Preference Optimization, PlurPO）：在给定描述人际冲突的输入后，语言模型识别并模拟相关利益相关者，然后被训练以偏好并生成所有利益相关者都能接受的响应。PlurPO仅使用模型对其自身输出产生的信号，无需真实标签。与以往方法相比，PlurPO在四个数据集和四个模型系列中显著减少了社会谄媚现象。例如，在意图造成伤害的陈述中，用户的行为不应被支持，PlurPO在四个模型中平均减少了89%的支持率。在一般建议问题中，目标是匹配人类响应的支持率，PlurPO将差距缩小了一半以上，从17.8%降至8.0%。PlurPO为8B模型构建的偏好数据集在更大（32B）模型中也有效地转移了对谄媚的缓解。我们的结果表明，通过利用模型自身的能力来模拟多元相关视角，可以减少社会谄媚现象。
cs.AI / 27 / 2610.02576
Answering clinicians' questions over trial evidence tables with verifiable, feedback-driven language models
通过可验证的反馈驱动语言模型回答临床医生对试验证据表的提问
Abstract
Systematic reviews condense clinical trials into evidence tables, yet clinicians can interrogate these tables only through database queries, and many questions concern attributes that the table does not record, such as a drug's target class or a harmonised endpoint. Here we introduce FD-SCoPE, a language-model framework that answers both kinds of question, exposes the query, the selected trials and the derivation rule behind every answer, and learns from expert corrections. On an oncology evidence table of 159 immune checkpoint inhibitor trial records, FD-SCoPE completed all 140 clinician-style tasks (alternatives, 90.7-97.9%). For questions needing derived attributes it retrieved 99.3% of relevant trial records at a positive predictive value of 89.8% and outperformed four alternative approaches (derived-value F1 77.7% versus 64.8-73.4%). Corrections on 299 questions, simulated from reference answers, raised F1 on 1,201 unseen questions from 77.9% to 84.9%. Language models coupled with executable queries, verified programs and expert feedback can give clinicians auditable access to trial evidence.
Chinese Translation
系统评价将临床试验浓缩为证据表，但临床医生只能通过数据库查询来询问这些表格，许多问题涉及表格未记录的属性，例如药物的靶类或统一终点。在此，我们介绍FD-SCoPE，一个语言模型框架，能够回答这两类问题，展示查询、所选试验以及每个答案背后的推导规则，并从专家的修正中学习。在包含159个免疫检查点抑制剂试验记录的肿瘤学证据表上，FD-SCoPE完成了所有140个临床医生风格的任务（替代方案，90.7%-97.9%）。对于需要推导属性的问题，它以89.8%的正预测值检索到99.3%的相关试验记录，并优于四种替代方法（推导值F1为77.7%，而其他为64.8%-73.4%）。对299个问题的修正（模拟自参考答案）使得在1,201个未见问题上的F1从77.9%提高到84.9%。结合可执行查询、经过验证的程序和专家反馈的语言模型可以为临床医生提供可审计的试验证据访问。
cs.AI / 28 / 2610.02586
Labels Override Definitions in Jev-Style Typed Decision Models
标签优先于定义在Jev风格的类型化决策模型中
Abstract
A typed decision model answers a fixed question about an input by returning a probability for each of several caller-defined options. Each option carries a short label and a written definition, which is where a developer states the rule the model should apply. Jev introduced this interface for routing, moderation and triage, open implementations followed, and the same operation occurs whenever a language model is used as a classifier by scoring label strings. We study the open implementations, whose weights we can inspect and patch, and ask whether the probability follows the definitions or the labels. A preference for the label we call option-label bias. Across four open-weight typed decision models, three ways of reading an answer from a Qwen2.5 backbone, eleven classification tasks and PolicyBench, a synthetic routing suite we introduce in which the rule appears only in the definitions, the answer is mostly the labels. Deleting every definition leaves accuracy unchanged (laya-td: 0.8559 against 0.8487), although those definitions support 0.7971 on their own, and renaming the options to A and B raises accuracy by +0.1511 [+0.1377, +0.1646]. One system, von, is unaffected, and the two code bases differ in one expression: laya writes each option as "{label}: {definition}", while von writes only the definition. Changing that expression in both directions, with no weight changed, makes all three laya checkpoints exactly invariant (+0.0000 [+0.0000, +0.0000]) and creates the effect in von, whose accuracy falls from 0.8511 to 0.2281 when a label contradicts its definition. Earlier work attributed this failure to the constrained decision head these models use in place of a text decoder; our results locate it in the prompt rendering. We give a two-call test that tells a practitioner which case applies to their model, and measure what four mitigations are worth.
Chinese Translation
类型化决策模型通过为多个调用者定义的选项返回概率，回答关于输入的固定问题。每个选项都有一个简短的标签和一个书面定义，开发者在其中陈述模型应应用的规则。Jev为路由、审核和分流引入了这一接口，随后出现了开放实现，每当语言模型作为分类器通过对标签字符串进行评分时，都会发生相同的操作。我们研究这些开放实现，其权重可以被检查和修补，并询问概率是遵循定义还是标签。我们称之为选项标签偏见的标签偏好。在四个开放权重的类型化决策模型中，通过三种方式从Qwen2.5主干读取答案，进行十一项分类任务，以及我们引入的PolicyBench，一个合成路由套件，其中规则仅出现在定义中，答案大多是标签。删除每个定义不会改变准确性（laya-td: 0.8559对比0.8487），尽管这些定义单独支持0.7971，而将选项重命名为A和B则使准确性提高了+0.1511 [+0.1377, +0.1646]。一个系统von不受影响，两个代码库在一个表达式上有所不同：laya将每个选项写为“{label}: {definition}”，而von仅写定义。在不改变权重的情况下，改变该表达式的两个方向，使所有三个laya检查点完全不变（+0.0000 [+0.0000, +0.0000]），并在von中产生效果，当标签与其定义相矛盾时，其准确性从0.8511降至0.2281。早期的工作将这一失败归因于这些模型使用的受限决策头，而不是文本解码器；我们的结果将其定位于提示渲染。我们提供了一个两次调用的测试，告诉从业者哪个案例适用于他们的模型，并测量四种缓解措施的价值。
cs.AI / 29 / 2610.02588
Open-Endedness Bench: Measuring Epistemic Process from Agent Records
开放性基准：从智能体记录中测量认知过程
Abstract
Agents are increasingly given open-ended research tasks: discovering an empirical law from self-designed experiments, improving a heuristic whose optimum nobody knows, or beating a standing record. Their execution logs record every step of this research, yet the runs are still judged by their outcome score. That score alone does not establish whether an agent's claims follow from executed experiments, and a reference answer may be unavailable. We evaluate the agent's epistemic process: how it forms hypotheses, tests them, and revises them in response to evidence. We introduce OEB (Open-Endedness Bench), a benchmark-agnostic methodology that reads only the agent's execution record and never a reference answer or an outcome score. OEB compiles the record into a unified epistemic event graph whose edges connect the propositions the agent states to the executed actions that test them; each node carries an exact excerpt that code verifies against the record. One principle governs scoring: prose can state a proposition, but only evidence returned by an executed action can support or refute it, so OEB checks what the agent writes against what it actually ran. From the graph, OEB scores four competence axes (evidence, experiment, revision, and no reward hacking), mostly as the share of opportunities for sound research that the agent took, and profiles six subjective persona traits that describe the agent's research habits. We score 119 existing runs over 12 tasks from three benchmarks: LLM post-training, chip design, and a training-speed record. Against logged results, only 16-29% of the improvements agents claim are real. On 9 of 10 tasks, the best run tries more new ideas in its second half than the worst run. The persona readings follow the model: for every trait, the model that ran explains more of its variance across runs than the task (a median of 43% against 7%).
Chinese Translation
智能体越来越多地被赋予开放性研究任务：从自设计实验中发现经验法则、改进一个未知最优值的启发式方法，或打破现有记录。它们的执行日志记录了研究的每一步，但这些运行仍然根据结果得分进行评判。仅凭得分无法确定智能体的主张是否源于已执行的实验，且参考答案可能不可用。我们评估智能体的认知过程：它如何形成假设、测试假设并根据证据进行修正。我们引入OEB（Open-Endedness Bench），一种基准无关的方法论，仅读取智能体的执行记录，而不参考答案或结果得分。OEB将记录编译成一个统一的认知事件图，其边连接智能体所陈述的命题与测试这些命题的执行动作；每个节点都携带一个确切的摘录，代码会根据记录进行验证。评分遵循一个原则：文本可以陈述命题，但只有执行动作返回的证据才能支持或反驳该命题，因此OEB检查智能体所写内容与其实际运行内容的一致性。通过图，OEB对四个能力维度（证据、实验、修正和无奖励黑客行为）进行评分，主要是智能体所采取的合理研究机会的比例，并描绘六个描述智能体研究习惯的主观特征。我们对来自三个基准的12个任务中119个现有运行进行评分：LLM后训练、芯片设计和训练速度记录。与记录结果相比，智能体声称的改进中仅有16-29%是真实的。在10个任务中的9个任务中，最佳运行在其后半部分尝试了更多的新想法，而最差运行则较少。个性特征的读数遵循模型：对于每个特征，运行的模型解释了其变异性中更多的部分，相较于任务（中位数为43%对7%）。
cs.AI / 30 / 2610.02599
TasteBench: Multimodal Benchmark for Sensory Prediction, from Molecules to Sustainable Foods
TasteBench：从分子到可持续食品的感官预测多模态基准
Abstract
Sustainable protein discovery lacks the fast computational proxies, analogous to molecular docking or density functional theory, that accelerate drug and materials discovery. Evaluating whether a novel food tastes like its animal-based target requires expensive human sensory panels, bottlenecking the design-build-test loop. We introduce TasteBench, a multimodal benchmark and privacy-preserving competition for sensory prediction, spanning two tasks: a food-level ranking task built on 21K+ human evaluations across 215 plant-based foods in 24 product categories, yielding 935 within-category ranking pairs, and a supporting molecular-level taste classification task over 15K flavor molecules. To enable rigorous interpretation of model performance, we characterize the ground truth: inter-rater agreement among panelists is low (Krippendorff's $\alpha = .077$), and the split-half reliability ceiling of panel-aggregated rankings is .825, establishing the range within which ML systems on this benchmark should be assessed. We evaluate baselines across four input modalities; on the same pairs panelists rated, the best model achieves .661 pairwise accuracy, competitive with the median individual panelist (.650), and .683 across all within-category pairs. TasteBench provides the evaluation infrastructure and baselines for measuring progress on computational screening for sustainable protein discovery.
Chinese Translation
可持续蛋白质发现缺乏快速的计算代理，类似于分子对接或密度泛函理论，这些代理可以加速药物和材料的发现。评估一种新食品是否与其动物基目标的味道相似需要昂贵的人类感官小组，这限制了设计-构建-测试循环的效率。我们提出了TasteBench，一个用于感官预测的多模态基准和隐私保护竞赛，涵盖两个任务：一个基于215种植物性食品在24个产品类别中超过21,000次人类评估构建的食品级排名任务，产生了935个类别内排名对，以及一个支持性的分子级味道分类任务，涵盖15,000种风味分子。为了实现对模型性能的严格解释，我们对真实情况进行了特征描述：评审员之间的评分一致性较低（Krippendorff's α = .077），而评审小组聚合排名的分半信度上限为.825，确立了在此基准上评估机器学习系统的范围。我们在四种输入模态上评估了基线；在评审员评分的相同对中，最佳模型实现了.661的成对准确率，与中位数个体评审员（.650）相当，并且在所有类别内对中达到了.683。TasteBench提供了评估基础设施和基线，用于衡量可持续蛋白质发现的计算筛选进展。
cs.AI / 31 / 2610.02608
Time Series Forecasting Benchmarks Need Scenario-Grounded Stress Testing
时间序列预测基准需要基于情境的压力测试
Abstract
Time series forecasting (TSF) increasingly drives decisions in transportation, energy, finance, healthcare, and infrastructure, yet current evaluation remains overly narrow: standard benchmarks reward low held-out error, while robustness studies typically reduce failure to Gaussian noise, random masking, or bounded adversarial perturbations. This obscures the real failure modes of deployed forecasting systems. Input-side anomalies are not merely noisier inputs: they often reflect structured events that alter temporal dynamics, break cross-variable dependencies, induce regime shifts, or propagate from faulty sensors to downstream decisions. These semantic, causal, and system-level failures cannot be faithfully captured by i.i.d. perturbations alone. The rise of TSF foundation models makes this evaluation gap more urgent, as unauditable pretraining corpora make held-out generalization increasingly unreliable. We therefore advocate scenario-grounded stress testing. Each test instance should include historical inputs and future targets, together with a semantic scenario, an explicit failure operator, and a measurable difficulty level. This shift makes evaluation interpretable, attributable, and deployment-relevant and friendly, enabling the community to ask not only which model is accurate, but under what conditions it fails and why.
Chinese Translation
时间序列预测（TSF）在交通、能源、金融、医疗保健和基础设施等领域日益影响决策，然而当前的评估仍然过于狭隘：标准基准奖励低的保留误差，而稳健性研究通常将失败简化为高斯噪声、随机掩蔽或有限的对抗扰动。这掩盖了已部署预测系统的真实失败模式。输入侧异常不仅仅是更嘈杂的输入：它们通常反映出结构化事件，这些事件改变时间动态、打破跨变量依赖、引发状态转变，或从故障传感器传播到下游决策。这些语义、因果和系统级的失败无法仅通过独立同分布（i.i.d.）扰动真实捕捉。随着TSF基础模型的兴起，这一评估差距变得更加紧迫，因为不可审计的预训练语料库使得保留泛化变得越来越不可靠。因此，我们倡导基于情境的压力测试。每个测试实例应包括历史输入和未来目标，以及一个语义情境、一个明确的失败操作符和一个可测量的难度级别。这一转变使评估变得可解释、可归因，并与部署相关，能够使社区不仅关注哪个模型准确，还关注在什么条件下它失败以及原因。
cs.AI / 32 / 2610.02616
VERSE: Verified Self-Evolving Optimizer for Agent Harnesses
VERSE：用于代理工具的经过验证的自我演化优化器
Abstract
Harness evolution improves an LLM agent's prompts, tools, and workflow, while the optimizer's own tools and procedures often remain fixed. We study whether an optimizer can improve another agent more effectively by also improving how it diagnoses failures, develops edits, and tests their effects. Two observations guide our design. In a controlled study, optimizer self-evolution fails to improve performance without execution-based verification, but achieves the best result of that study when verification is available. Across five executors, self-evolving optimizers build their own tools for failure analysis, verification, training audits, and workflow control. Motivated by these findings, we introduce VERSE, a Verified Self-Evolving optimizer for agent harnesses. VERSE lets the optimizer test draft edits, replay failures, and perturb suspected steps before submission, while tracking fixes and regressions across rounds. Using this feedback, the optimizer revises both the executor harness and its own prompts, skills, tools, hooks, and notes, while the weights of the optimizer and executor models stay fixed. Under a shared protocol with disjoint training, validation, and test tasks, VERSE improves all four evaluated harness optimizers on held-out SWE-rebench tasks and newer out-of-distribution tasks in five languages. Its best validation-selected harness reaches 42.3% and 37.7% accuracy, respectively, against 39.2% and 29.3% for the strongest baselines. Code is available at https://github.com/wzekai/VERSE.
Chinese Translation
工具演化改善了大型语言模型（LLM）代理的提示、工具和工作流程，而优化器自身的工具和程序往往保持不变。我们研究了优化器是否可以通过改善其诊断故障、开发编辑和测试效果的方式，更有效地提升另一个代理的性能。两个观察结果指导了我们的设计。在一项受控研究中，优化器的自我演化在没有基于执行的验证时未能改善性能，但在有验证的情况下取得了该研究的最佳结果。在五个执行器中，自我演化的优化器构建了自己的故障分析、验证、训练审计和工作流程控制工具。基于这些发现，我们引入了VERSE，一个用于代理工具的经过验证的自我演化优化器。VERSE 允许优化器在提交之前测试草稿编辑、重放故障和扰动可疑步骤，同时跟踪修复和回归。在使用这些反馈的过程中，优化器修订了执行器工具及其自身的提示、技能、工具、钩子和笔记，同时优化器和执行器模型的权重保持不变。在一个共享协议下，训练、验证和测试任务不重叠，VERSE 在保留的 SWE-rebench 任务和五种语言的新分布外任务上改善了所有四个评估的工具优化器。其最佳验证选择的工具分别达到了 42.3% 和 37.7% 的准确率，而最强基线的准确率为 39.2% 和 29.3%。代码可在 https://github.com/wzekai/VERSE 获取。
cs.AI / 33 / 2610.02622
CuBEs: Culturally-Situated Behavioral Evaluations and the Limitations of Culture-Blind LLM Judges
CuBEs：文化情境下的行为评估及文化盲目大型语言模型评判的局限性
Abstract
Evaluating the occurrence and triggers of large language model (LLM) behaviors - such as sycophancy, self-preference, or over-confidence - is critical for predicting real-world model deployment risks. However, existing situated behavioral evaluations typically ignore cultural context, limiting their generalizability across an increasingly global user base. To address this gap, we propose CuBEs - Culturally-situated Behavior Evaluations that probe for response patterns across diverse user cultures. We first extend an automated testing pipeline to inject cultural context into behavioral test scenarios and subsequent evaluation. We assess the cultural adaptability of this pipeline by building a human-labeled dataset that captures nuanced dimensions of behavior understanding across 12 distinct cultures. Our dataset reveals significant cross-cultural variations that one-size-fits all judgments fail to capture. Through evaluating 13 open- and closed-source LLMs, we find that introducing cultural situatedness in the evaluation scenario creates significant variation in the presence of a behavior. For example, while our baseline experiments testing for political bias capture localized Western political dimensions like the American conservative-progressive divide, non-Western culturally situated evaluations surface entirely different axes of bias such as religious and colonial political issues. Our findings demonstrate that standard, culturally-agnostic evaluations fail to capture these shifts, highlighting the necessity of culturally situated behavioral testing for global deployments.
Chinese Translation
评估大型语言模型（LLM）行为的发生及触发因素——例如谄媚、自我偏好或过度自信——对于预测实际模型部署风险至关重要。然而，现有的情境行为评估通常忽视文化背景，限制了其在日益全球化的用户基础上的普遍适用性。为了解决这一问题，我们提出了CuBEs——文化情境下的行为评估，旨在探讨不同用户文化中的响应模式。我们首先扩展了一个自动化测试管道，将文化背景注入行为测试场景及后续评估中。通过构建一个人类标注的数据集，我们评估了该管道的文化适应性，该数据集捕捉了12种不同文化中行为理解的细微维度。我们的数据集揭示了显著的跨文化差异，而一刀切的判断无法捕捉到这些差异。通过评估13个开源和闭源的LLM，我们发现，在评估场景中引入文化情境性会显著影响行为的存在。例如，尽管我们的基线实验测试政治偏见捕捉了美国保守派与进步派之间的局部西方政治维度，但非西方文化情境评估则揭示了完全不同的偏见轴线，如宗教和殖民政治问题。我们的研究结果表明，标准的文化无关评估无法捕捉到这些变化，强调了在全球部署中进行文化情境下行为测试的必要性。
cs.AI / 34 / 2610.02627
Lost in the Request: How Communication Variation Disrupts Retrieval and Action in Email Agents
请求中的迷失：沟通变异如何干扰电子邮件助手的检索与行动
Abstract
An email assistant should not complete less work simply because a user phrases the same request differently. Yet most benchmarks test each task with only one canonical request, leaving this form of robustness largely unmeasured. We test whether email assistants remain reliable when the requested information, available evidence, and expected outcome stay fixed, but the communication style or English variety changes. We construct validated variants along five communication-style axes and four rule-based dialect conditions, and evaluate them on three benchmarks: a retrieval-augmented generation (RAG) pipeline and two tool-using agents. Indirect requests reduce performance on all three benchmarks, while formal requests reduce performance on both agentic benchmarks. Examining the systems more closely shows that these failures have different causes. Verbose requests mainly hurt a lexical retriever by making the relevant email harder to find. By contrast, indirect and dialect variants remain harmful even when the relevant email is retrieved. In the agentic setting, indirect and formal requests mainly cause the agents to omit required actions, not to take more unsupported actions. These results show that a successful response is not enough to establish robustness: evaluations should vary how requests are expressed and separately measure whether agents complete the requested work.
Chinese Translation
电子邮件助手不应仅因用户以不同方式表达相同请求而完成更少的工作。然而，大多数基准测试仅用一个标准请求来测试每个任务，使得这种鲁棒性在很大程度上未被测量。我们测试了当请求的信息、可用证据和预期结果保持不变时，电子邮件助手在沟通风格或英语变体变化时是否仍然可靠。我们沿着五个沟通风格轴和四个基于规则的方言条件构建了经过验证的变体，并在三个基准上进行评估：一个检索增强生成（RAG）管道和两个工具使用代理。间接请求在所有三个基准上都降低了性能，而正式请求在两个代理基准上降低了性能。更仔细地检查系统表明，这些失败有不同的原因。冗长的请求主要通过使相关电子邮件更难找到而损害了词汇检索器的性能。相比之下，即使相关电子邮件被检索到，间接和方言变体仍然会造成伤害。在代理设置中，间接和正式请求主要导致代理省略所需的行动，而不是采取更多不支持的行动。这些结果表明，成功的响应不足以建立鲁棒性：评估应变化请求的表达方式，并单独测量代理是否完成所请求的工作。
cs.AI / 35 / 2610.02631
Designing the Future of User Feedback for Generative AI
为生成性人工智能设计用户反馈的未来
Abstract
Post-deployment feedback from users can be a cost-effective, scalable, and representative means to monitor and improve generative AI systems and features. When implemented effectively, giving such feedback can increase users' engagement with and trust in GenAI systems. Government regulations and industry guidelines call for post-deployment user engagement, but there is little guidance on designing mechanisms that are usable for consumers and provide actionable input for product teams. We conducted a multi-phase study as a collaboration between academic researchers and eBay. Our benchmark evaluation of current industry approaches identified common issues including lack of discoverability, unclear terminology, and inattention to user value. Based on these findings, we developed best-practice recommendations and designed and tested a prototype feedback-collection tool. The tool aimed to provide users with an efficient, flexible, and positive feedback-giving experience, and provide product teams with rich data on performance and potential problems in a usable format.
Chinese Translation
用户在部署后的反馈可以成为一种具有成本效益、可扩展且具有代表性的方法，用于监测和改进生成性人工智能系统及其功能。当有效实施时，提供这样的反馈可以增强用户对生成性人工智能系统的参与度和信任度。政府法规和行业指南呼吁在部署后进行用户参与，但关于设计可供消费者使用的机制以及为产品团队提供可操作输入的指导却很少。我们进行了一个多阶段的研究，作为学术研究人员与eBay之间的合作。我们对当前行业方法的基准评估识别出了一些共同问题，包括缺乏可发现性、术语不清晰以及对用户价值的忽视。基于这些发现，我们制定了最佳实践建议，并设计和测试了一个原型反馈收集工具。该工具旨在为用户提供高效、灵活和积极的反馈体验，并为产品团队提供以可用格式呈现的丰富数据，以便于了解性能和潜在问题。
cs.AI / 36 / 2610.02638
Batched Speech Decisions Without Decoding: Single-Token Supervision Lets a Frozen LLM Hear Beyond the Transcript
无解码的批量语音决策：单标记监督让冻结的LLM超越转录听见
Abstract
Full-duplex voice agents make many small, closed decisions, which current systems answer by slow autoregressive decoding. We propose DuplexJev, which feeds ASR-encoder hidden states through a small connector into a frozen LLM and reads each question as a single-token distribution over its options. Nothing is decoded, and an 8-GPU node answers 80 decisions about eight utterances in about 0.1 s. With a last-layer connector, spoken QA stays close to reading the transcript (90% vs. 91%). DuplexJev also hears the speaker: gender and emotion accuracy both reach 90% (from 55% and 28%) with a cross-attention connector, whose spoken QA drops by only 1 point (83% to 82%). We train decisions with cross-entropy on the read-out answer token, instead of the usual transcript distillation, whose teacher never hears the voice, and keep distillation for content. Encoders and LLMs are interchangeable; we release weights, training recipe, a batched-inference pipeline for full-duplex serving and a bilingual spoken-QA set.
Chinese Translation
全双工语音代理进行许多小型、封闭的决策，而当前系统通过缓慢的自回归解码来响应这些决策。我们提出了DuplexJev，它通过一个小连接器将ASR编码器的隐藏状态输入到一个冻结的LLM中，并将每个问题视为其选项上的单标记分布。没有任何内容被解码，一个8-GPU节点在约0.1秒内回答关于八个发声的80个决策。通过最后一层连接器，口语问答的准确性接近于读取转录（90%对91%）。DuplexJev还能够听到说话者的声音：性别和情感的准确率均达到90%（分别从55%和28%提高），使用交叉注意力连接器，其口语问答的准确率仅下降了1个百分点（从83%降至82%）。我们通过交叉熵训练决策，基于读取的答案标记，而不是通常的转录蒸馏，后者的教师从未听到声音，并将蒸馏用于内容。编码器和LLM是可互换的；我们发布权重、训练配方、用于全双工服务的批量推理管道以及一个双语口语问答集。
cs.AI / 37 / 2610.02654
Coherence-Driven Belief Formation and Population Dynamics of Contagion in LLM Agents
基于一致性的信念形成与大语言模型代理的传播动态
Abstract
Models of social contagion usually assume how individuals adopt beliefs and derive population behavior from it. We instead empirically measure belief adoption in language model agents, quantifying the probability an agent adopts a claim given how many peers endorse it. We find this adoption kernel to be sigmoid, a characteristic of complex contagion, with a threshold that is sensitive to three sources: the claim's plausibility, the source's reliability, and the agent's disposition. These three dimensions are well approximated by a single effective dimension which we propose can be understood as the coherence of the incoming belief with the LLM agent's prior beliefs. Further, we observe a characteristic of complex contagion in the collective dynamics of belief adoption in a system of AI agents: further spread on clustered than random networks. These systems also exhibit a bifurcating cascade window, and self-sustaining hysteretic consensus which lead to consensus being far harder to remove than to establish.
Chinese Translation
社会传播模型通常假设个体如何采纳信念，并由此推导出群体行为。相反，我们通过实证研究语言模型代理的信念采纳，量化了在多少同伴支持的情况下，代理采纳某一主张的概率。我们发现这一采纳核呈现S形曲线，这是复杂传播的特征，其阈值对三个来源敏感：主张的可信度、来源的可靠性以及代理的倾向。这三个维度可以很好地用一个有效维度来近似，我们提出可以将其理解为新信念与大语言模型代理先前信念的一致性。此外，我们观察到在AI代理系统中信念采纳的集体动态中，复杂传播的特征：在聚类网络中传播的效果优于随机网络。这些系统还表现出分叉级联窗口和自我维持的滞后共识，这使得达成共识比消除共识要困难得多。
cs.AI / 38 / 2610.02664
A GHOST in Long-Horizon Agents: Governance Hazard from Overlooked Safety Constraints across Turns
长时间视角代理中的幽灵：因忽视跨回合安全约束而产生的治理风险
Abstract
Long-horizon agents are now playing an increasingly significant role in assisting humans with complex problem-solving. However, it is exactly their extended interaction history that introduces an underexplored execution-safety concern. Under benign interaction conditions, an agent may execute an action that violates a safety constraint specified many turns earlier. We term this failure mode Governance Hazard from Overlooked Safety Constraints across Turns (GHOST), which may cause irreversible damage. Our experiments reveal that GHOST events are not isolated cases: this failure mode, occurring precisely under benign interaction conditions, yields an occurrence rate of 11.5% on GPT-5.5. Furthermore, we theoretically show that if the residual conditional violation hazard along each safe prefix is bounded below by a non-summable sequence, the execution enters the hazard region almost surely. Leveraging this theoretical insight, we further propose STAR-Guard, a two-layer defense coupling historical semantic safety constraint restoration with pre-execution audit. STAR-Guard restores applicable safety constraints to reduce unsafe proposals, while its deterministic audit layer prevents residual violations from reaching the environment. Consistent with this two-layer design, we observe no GHOST events in our experiments under the GPT-5.5 setup.
Chinese Translation
长时间视角代理在协助人类解决复杂问题方面正发挥着越来越重要的作用。然而，正是它们延长的交互历史引入了一个未被充分探索的执行安全问题。在良性交互条件下，代理可能会执行一个违反早先指定的安全约束的动作。我们将这种失败模式称为跨回合忽视安全约束的治理风险（Governance Hazard from Overlooked Safety Constraints across Turns，简称 GHOST），这可能导致不可逆转的损害。我们的实验表明，GHOST 事件并不是孤立的案例：这种失败模式在良性交互条件下发生，其发生率在 GPT-5.5 上达到了 11.5%。此外，我们从理论上证明，如果沿每个安全前缀的剩余条件违反风险被一个不可求和的序列下界所限制，则执行几乎肯定会进入风险区域。利用这一理论见解，我们进一步提出了 STAR-Guard，一种将历史语义安全约束恢复与执行前审计相结合的双层防御机制。STAR-Guard 恢复适用的安全约束以减少不安全的提议，而其确定性的审计层则防止剩余违反到达环境。与这种双层设计一致，我们在 GPT-5.5 设置下的实验中观察到没有 GHOST 事件的发生。
cs.AI / 39 / 2610.02678
Spend Teacher Tokens Where They Matter: Success-Referenced On-Policy Distillation
在关键位置使用教师令牌：成功参考的在线策略蒸馏
Abstract
On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher, but providing such supervision for every rollout requires substantial teacher computation. We introduce Success-Referenced On-Policy Distillation (SR-OPD), which reduces this cost by selecting which prompts and rollouts receive teacher supervision. When the student produces both successful and failed rollouts for the same prompt, a successful rollout can serve as a natural reference for selecting failed rollouts. SR-OPD therefore focuses on such prompts and prioritizes failed rollouts whose hidden-state trajectories show sustained divergence from a successful reference, while accounting for estimated teacher-input cost. Across three teacher-student pairs and six mathematical reasoning benchmarks, SR-OPD uses only 3.46-5.02% of the teacher-input tokens required by Vanilla OPD in the one-pass setting while maintaining comparable reasoning performance. Under a controlled setting matched to 5% of Vanilla OPD's teacher-input budget, further experiments support both key design choices: focusing supervision on prompts with both successful and failed rollouts, and using successful rollouts to guide failure selection. These results indicate that a student's own successful behavior can serve as a useful reference for allocating teacher supervision under a fixed teacher-input budget.
Chinese Translation
在线策略蒸馏（OPD）将学生生成的回合与来自教师的密集令牌级监督相结合，但为每个回合提供这种监督需要大量的教师计算。我们提出了成功参考的在线策略蒸馏（SR-OPD），通过选择哪些提示和回合接受教师监督来降低这一成本。当学生为同一提示生成成功和失败的回合时，成功的回合可以作为选择失败回合的自然参考。因此，SR-OPD专注于这些提示，并优先考虑那些隐藏状态轨迹与成功参考表现出持续偏离的失败回合，同时考虑估计的教师输入成本。在三个教师-学生对和六个数学推理基准上，SR-OPD在单次传递设置中仅使用了香草OPD所需教师输入令牌的3.46%-5.02%，同时保持了可比的推理性能。在与香草OPD的教师输入预算的5%相匹配的受控环境下，进一步实验支持了两个关键设计选择：将监督集中在同时具有成功和失败回合的提示上，以及使用成功回合来指导失败选择。这些结果表明，学生自身的成功行为可以作为在固定教师输入预算下分配教师监督的有用参考。
cs.AI / 40 / 2610.02679
DataWeave: Deploying Human-LLM Analytics for Exploratory Structured Data Analysis
DataWeave：部署人类-大型语言模型分析以进行探索性结构化数据分析
Abstract
Data journalism, the practice of using data analysis to surface newsworthy stories, depends increasingly on the ability of reporters and investigative journalists to uncover trends, disparities, and accountability narratives. In practice, exploring large structured datasets remains slow and brittle: journalists must navigate hundreds of variables across many datasets over years, understand data coding conventions, and write non-trivial analysis code while hypotheses evolve. Although LLMs are often touted as "ask in English, get SQL/answers," real newsroom workflows expose recurring failures, e.g., schema mismatches and drift, misread domain semantics and units, and silent assumptions. We present DataWeave, a system that addresses these needs by combining conversational interaction, schema grounding, analytical planning, and executable query generation to support exploratory analysis over structured data. Rather than treating LLMs as autonomous answer engines, DataWeave frames them as interactive partners whose outputs can be inspected, corrected, and steered as hypotheses shift. We present a case study with professional journalists using our system to analyze the U.S. Department of Education's Integrated Postsecondary Education Data System (IPEDS), a high-stakes public dataset with substantial domain semantics and frequent schema updates. We also report how deployment experience and iterative refinement shaped the current DataWeave architecture and its analytical workflow. Our findings distill design principles and deployment lessons for trustworthy human-LLM collaboration in structured data analysis.
Chinese Translation
数据新闻，即利用数据分析揭示值得报道的故事的实践，越来越依赖于记者和调查记者发现趋势、不平等和问责叙事的能力。在实践中，探索大型结构化数据集仍然缓慢且脆弱：记者必须在多年内跨越多个数据集导航数百个变量，理解数据编码约定，并在假设演变的过程中编写复杂的分析代码。尽管大型语言模型（LLMs）常被宣传为“用英语提问，获得 SQL/答案”，但真实的新闻工作流程暴露出反复出现的失败，例如模式不匹配和漂移、误读领域语义和单位，以及无声的假设。我们提出了 DataWeave，一个通过结合对话交互、模式基础、分析规划和可执行查询生成来支持结构化数据的探索性分析的系统。DataWeave 不将 LLMs 视为自主答案引擎，而是将其框架设定为互动伙伴，其输出可以在假设变化时进行检查、修正和引导。我们展示了一个案例研究，专业记者使用我们的系统分析美国教育部的综合高等教育数据系统（IPEDS），这是一个具有重要领域语义和频繁模式更新的高风险公共数据集。我们还报告了部署经验和迭代改进如何塑造当前的 DataWeave 架构及其分析工作流程。我们的研究结果提炼了在结构化数据分析中实现可信赖的人类-LLM 协作的设计原则和部署经验。
cs.AI / 41 / 2610.02684
Large language models exhibit unreliable updating of clinical judgment as patient evidence evolves
大型语言模型在患者证据演变时表现出不可靠的临床判断更新
Abstract
Large language models (LLMs) are increasingly explored for clinical reasoning, but whether they appropriately revise judgments as patient evidence evolves remains unclear. We evaluated longitudinal belief updating using matched intensive-care trajectories from electronic health records. Across diverse LLMs, conditioning on a preceding judgment more often increased than reduced prediction error when estimates changed, replicated for a second endpoint. Controlled interventions revealed two failure modes. First, with preceding assessment fixed, models responded more strongly to worsening than matched improving respiratory evidence; this asymmetry persisted after headroom normalization at moderate and strong evidence levels. Second, with current evidence fixed, increasing prior risk from 10% to 90% shifted estimates by 26.2 percentage points, demonstrating causal influence of prior model beliefs. Prompting did not restore reliable updating. Evidence-Validated Longitudinal Update (EVLU) identified fewer, more reliable revisions, revealing a reliability-coverage trade-off. These findings establish longitudinal belief updating as a distinct dimension of LLM reliability.
Chinese Translation
大型语言模型（LLMs）在临床推理中越来越受到关注，但它们是否能够在患者证据演变时适当地修正判断仍不明确。我们使用来自电子健康记录的匹配重症监护轨迹评估了纵向信念更新。在不同的LLMs中，当估计发生变化时，基于先前判断的条件更常导致预测误差增加而非减少，这一现象在第二个终点上得到了重复验证。控制干预揭示了两种失效模式。首先，在固定先前评估的情况下，模型对恶化的呼吸证据的反应强于对匹配改善的呼吸证据的反应；这种不对称性在中等和强证据水平的头部空间标准化后仍然存在。其次，在固定当前证据的情况下，将先前风险从10%增加到90%使估计值变化了26.2个百分点，证明了先前模型信念的因果影响。提示并未恢复可靠的更新。证据验证的纵向更新（Evidence-Validated Longitudinal Update, EVLU）识别出更少但更可靠的修订，揭示了可靠性与覆盖率之间的权衡。这些发现确立了纵向信念更新作为LLM可靠性的一个独特维度。
cs.AI / 42 / 2610.02687
Decoupling Memory from Context: Structured Memory for Token-Efficient Test-Time Continual Learning
将记忆与上下文解耦：面向高效测试时持续学习的结构化记忆
Abstract
Large language models (LLMs) are increasingly deployed in enterprise, scientific, and medical applications, where agents must incorporate domain-specific knowledge and adapt from experience. Context engineering offers a practical alternative to weight updates by improving model behavior through instructions, strategies, and evidence supplied at inference time. However, adapting context online typically requires a costly trial-and-error process, while queries are often processed independently, preventing useful experience from carrying forward. Memory systems address this limitation by retaining information across interactions, but approaches that continually append information to a shared context face increasing token costs, context-window limits, and performance degradation as the context expands. We introduce a unified formulation of context optimization and show that an agent memory system update can be interpreted as an optimization update procedure over the model's context. This perspective attempts to provide a principled framework for studying memory design and its efficiency. We then propose GraphMemory, a lightweight graph-based memory that accumulates, refines, organizes, and connects reusable strategies. For each query, GraphMemory retrieves only the relevant subgraph, enabling online context adaptation without exposing the model to the entire memory. Under bounded retrieval, the amount of retrieved memory remains constant as the number of processed examples grows. Experiments show that GraphMemory achieves competitive downstream performance while using approximately 81-85% fewer memory-construction tokens than our baselines.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于企业、科学和医疗等领域，在这些应用中，智能体必须结合特定领域的知识并从经验中适应。上下文工程通过在推理时提供指令、策略和证据，提供了一种改进模型行为的实用替代方案，而无需更新权重。然而，在线适应上下文通常需要耗时的试错过程，而查询往往是独立处理的，这阻碍了有用经验的传递。记忆系统通过在交互中保留信息来解决这一限制，但持续向共享上下文添加信息的方法面临着不断增加的令牌成本、上下文窗口限制以及随着上下文扩展而导致的性能下降。我们提出了一种上下文优化的统一表述，并表明智能体记忆系统的更新可以被解释为对模型上下文的优化更新过程。这一视角试图为研究记忆设计及其效率提供一个有原则的框架。随后，我们提出了GraphMemory，一种轻量级的基于图的记忆，能够积累、精炼、组织和连接可重用的策略。对于每个查询，GraphMemory仅检索相关的子图，从而实现在线上下文适应，而无需将模型暴露于整个记忆中。在有限检索的情况下，随着处理示例数量的增加，检索到的记忆量保持不变。实验表明，GraphMemory在下游任务中实现了具有竞争力的性能，同时使用的记忆构建令牌比我们的基准减少了约81-85%。
cs.AI / 43 / 2610.02703
Learning to Revise Reasoning with Segment-wise On-Policy Distillation
通过分段在线蒸馏学习修正推理
Abstract
On-policy distillation (OPD) improves large language model reasoning by training students on their own rollouts with dense token-wise supervision from the teacher. However, token-wise OPD does not explicitly provide a coherent alternative reasoning step showing how the student's step could be revised to improve subsequent reasoning. Furthermore, this paradigm can become less effective when the student produces a degenerate reasoning prefix, as subsequent teacher supervision remains conditioned on that prefix and may reinforce poor reasoning patterns. In this work, we focus on learning reasoning revision with segment-wise OPD to rework intermediate reasoning steps and better support subsequent reasoning. Through controlled reasoning interventions, we find that replacing student segments with teacher redrafts improves subsequent reasoning accuracy. Therefore, we address the problem of turning teacher redrafts into explicit supervision for learning to revise reasoning. We propose Segment-wise On-Policy Distillation (Seg-OPD), which selects student segments based on an uncertainty metric and obtains corresponding teacher redrafts. Seg-OPD trains the student to prefer teacher redrafts over their paired student segments while retaining dense token-wise OPD supervision. Extensive experiments on mathematical reasoning and competitive programming tasks show that Seg-OPD-trained students achieve higher revision success rates than baselines. Seg-OPD consistently outperforms the compared state-of-the-art baselines in reasoning accuracy with an average relative improvement of 5.22% across diverse models and tasks. Code is available at https://anonymous.4open.science/r/Seg-OPD.
Chinese Translation
在线蒸馏（On-policy distillation, OPD）通过在学生的自我回滚上进行训练，并从教师那里获得密集的逐词监督，从而改善大型语言模型的推理能力。然而，逐词的 OPD 并未明确提供一个连贯的替代推理步骤，显示学生的步骤如何修正以改善后续推理。此外，当学生产生退化的推理前缀时，这一范式的有效性可能会降低，因为后续的教师监督仍然依赖于该前缀，可能会强化不良的推理模式。在本研究中，我们专注于通过分段在线蒸馏（Segment-wise On-Policy Distillation, Seg-OPD）学习推理修正，以重新处理中间推理步骤并更好地支持后续推理。通过控制推理干预，我们发现用教师重写的段落替换学生段落可以提高后续推理的准确性。因此，我们解决了将教师重写转化为显式监督以学习修正推理的问题。我们提出了 Seg-OPD，它基于不确定性指标选择学生段落，并获得相应的教师重写。Seg-OPD 训练学生优先选择教师重写而非其配对的学生段落，同时保留密集的逐词 OPD 监督。在数学推理和竞争编程任务上的大量实验表明，经过 Seg-OPD 训练的学生的修正成功率高于基线。Seg-OPD 在推理准确性方面始终优于比较的最先进基线，平均相对提升为 5.22%，适用于多种模型和任务。代码可在 https://anonymous.4open.science/r/Seg-OPD 获取。
cs.AI / 44 / 2610.02704
Label-Efficient Time Series Classification at Scale: A Dual-Stream OSSE-LSTM with Counterfactual Attribution
高效标签时间序列分类：一种具有反事实归因的双流OSSE-LSTM
Abstract
Time series are produced continuously at enormous scale by industrial equipment, wearables, power grids, and clinical monitors, yet annotation remains manual, expensive, and expert-dependent. The binding constraint in large-scale time series analytics is therefore not data volume but label volume, and the question facing a practitioner is concrete: how many examples per class must be labeled before a classifier becomes usable? We study this question directly, in a regime where the label space is fixed and known in advance and the decision rule must be constructed from only K labeled examples per class. We propose Dual-Stream OSSE-LSTM, an episodic metric-learning framework that pairs an Omni-Scale CNN with Squeeze-and-Excitation recalibration, for multi-scale motif extraction without per-dataset kernel tuning, with a Bidirectional LSTM for global temporal context. The two streams are independently normalized and fused into a prototype-oriented embedding. Because decisions taken from a few labels must also be explainable, we introduce Counterfactual Integrated Gradients (C-IG), which attributes the prototype margin between target and opposing classes rather than an isolated classifier logit, and reuses the resulting maps as soft masks for test-time prototype refinement without updating the encoder. On 19 univariate UCR datasets, OSSE-LSTM attains the highest average accuracy and per-dataset win count at every support size, and its accuracy remains within a 0.36-point band (96.36-96.72%) across that range. Its weakest configuration still exceeding the best result any compared baseline achieves at any K (93.99%).
Chinese Translation
工业设备、可穿戴设备、电网和临床监测器不断产生大量时间序列数据，但注释仍然是手动的、昂贵的且依赖于专家。因此，大规模时间序列分析的限制因素不是数据量，而是标签量，实践者面临的具体问题是：在分类器可用之前，每个类别必须标注多少示例？我们直接研究这个问题，在标签空间固定且已知的情况下，决策规则必须仅基于每个类别的K个标注示例构建。我们提出了双流OSSE-LSTM，这是一种情节度量学习框架，将全尺度卷积神经网络（Omni-Scale CNN）与挤压与激励（Squeeze-and-Excitation）再校准相结合，用于多尺度模式提取，而无需对每个数据集进行内核调优，并结合双向长短期记忆网络（Bidirectional LSTM）以获取全局时间上下文。这两个流独立归一化并融合为原型导向的嵌入。由于从少量标签中做出的决策也必须是可解释的，我们引入了反事实集成梯度（Counterfactual Integrated Gradients, C-IG），它归因于目标类与对立类之间的原型边际，而不是孤立的分类器对数值，并将生成的图作为软掩码用于测试时的原型细化，而无需更新编码器。在19个单变量UCR数据集上，OSSE-LSTM在每个支持大小下都达到了最高的平均准确率和每个数据集的胜利次数，其准确率在该范围内保持在0.36个百分点（96.36%-96.72%）的带内。即使是其最弱的配置，仍然超过了任何比较基线在任何K下所取得的最佳结果（93.99%）。
cs.AI / 45 / 2610.02715
Ego2World: Compiling Egocentric Cooking Videos into Executable Worlds for Belief-State Planning
Ego2World：将自我中心的烹饪视频编译为可执行世界以进行信念状态规划
Abstract
Egocentric videos capture how people carry out everyday activities, yet testing an agent requires evaluating the consequences of actions it chooses itself. We introduce Ego2World, a benchmark that turns annotated cooking activities into executable planning environments under partial observation. Its compiler links source steps and objects to symbolic action rules, persistent world states, and explicit task conditions, so researchers can execute an agent's proposed actions and check their outcomes. World state and agent belief are maintained separately, enabling controlled studies of planning and information reuse across continuing tasks. Evaluating six planners on 105 tasks shows that accepted operations often leave task goals unmet. Execution traces and condition checks distinguish interrupted runs, partial attainment, and completed execution without goal attainment. In a separate paired Qwen-Plus study, persistent belief improves action validity by 4.15 percentage points and reduces visual-query attempts by 90.27%, with higher token use and no detected completion gain. Ego2World provides a reusable testbed for tracing how planning and memory choices affect execution, observation demand, and task attainment, connecting recorded human activity to the development and evaluation of interactive agents.
Chinese Translation
自我中心视频捕捉了人们如何进行日常活动，但测试智能体需要评估其自主选择的行动后果。我们介绍了Ego2World，一个基准测试，它将注释的烹饪活动转化为在部分观察下的可执行规划环境。其编译器将源步骤和对象链接到符号行动规则、持久世界状态和明确任务条件，从而使研究人员能够执行智能体提出的行动并检查其结果。世界状态和智能体信念被分别维护，使得对规划和信息重用的控制研究成为可能。在105个任务上评估六个规划器的结果显示，接受的操作往往无法满足任务目标。执行轨迹和条件检查区分了中断运行、部分达成和未达成目标的完成执行。在一项独立的配对Qwen-Plus研究中，持久信念使行动有效性提高了4.15个百分点，并减少了90.27%的视觉查询尝试，伴随更高的标记使用率且未检测到完成增益。Ego2World提供了一个可重用的测试平台，用于追踪规划和记忆选择如何影响执行、观察需求和任务达成，将记录的人类活动与交互智能体的开发和评估联系起来。
cs.AI / 46 / 2610.02741
On the Chain-of-Thought Monitorability of Looped Language Models
循环语言模型的思维链监测能力研究
Abstract
Chain-of-thought (CoT) monitoring provides a promising approach for detecting undesirable model behavior. Looped language models (LoopLMs) repeatedly apply shared transformer layers, increasing effective computational depth and enabling additional latent computation without increasing model size. However, the effect of looped architectures on CoT monitorability remains largely unexplored. In this work, we provide the first systematic evaluation of CoT monitorability in LoopLMs. We study two complementary settings: (1) varying the loop depth within the same LoopLM family to isolate the effect of additional recurrent computation, and (2) comparing LoopLMs with non-looped language models matched by parameter size, transformer-layer count, or effective depth to study whether LoopLMs are less monitorable. Across eight tasks from MonitorBench and both standard and stress-test settings, we observe task-dependent reductions in CoT monitorability under stress tests on specific Logic/Science/Engineering \texttt{Cue Answer} tasks, while other tasks exhibit weaker or qualitatively different trends. Our diagnosis suggests that these declines are not fully explained by task difficulty, verification pass rate, or generated token length; qualitative examples further suggest changes in how deeper-loop models explicitly use or attribute provided cues. Our cross-model comparison finds no evidence that LoopLMs are systematically less monitorable than non-looped language models matched on size or depth. Overall, our results suggest that deeper loop depth can reduce CoT monitorability in some tasks under stress tests, but looped transformer architecture alone does not necessarily imply lower monitorability.
Chinese Translation
思维链（Chain-of-thought, CoT）监测为检测模型的不良行为提供了一种有前景的方法。循环语言模型（Looped Language Models, LoopLMs）重复应用共享的变换器层，增加了有效计算深度，并在不增加模型规模的情况下实现额外的潜在计算。然而，循环架构对思维链监测能力的影响仍然很大程度上未被探索。在本研究中，我们首次系统性地评估了LoopLMs中的CoT监测能力。我们研究了两种互补的设置：（1）在同一LoopLM家族中改变循环深度，以隔离额外递归计算的影响；（2）比较与参数规模、变换器层数或有效深度匹配的非循环语言模型，以研究LoopLMs是否更难监测。在来自MonitorBench的八个任务以及标准和压力测试设置中，我们观察到在特定的逻辑/科学/工程 exttt{Cue Answer} 任务的压力测试下，CoT监测能力出现了依赖任务的下降，而其他任务则表现出较弱或定性不同的趋势。我们的诊断表明，这些下降并不能完全用任务难度、验证通过率或生成的标记长度来解释；定性示例进一步表明，深层循环模型在如何明确使用或归因提供的提示方面发生了变化。我们的跨模型比较未发现LoopLMs在规模或深度上与非循环语言模型相比系统性地更难监测的证据。总体而言，我们的结果表明，在某些任务的压力测试下，较深的循环深度可能会降低CoT监测能力，但循环变换器架构本身并不一定意味着较低的监测能力。
cs.AI / 47 / 2610.02762
Dynamic LLM Routers are Often Misguided
动态 LLM 路由器常常被误导
Abstract
Dynamic LLM routers promise to cut inference costs by sending each query to the cheapest model that can answer it correctly. We analyze six commercial routers across 14 settings on a diverse benchmark spanning eight task categories, finding that none of them outperforms a router that randomly selects between two well-chosen models at matched cost. Some underperform by more than 10 percentage points. We trace this gap to four patterns prevalent across routers: difficulty blindness, length reversal, semantic matching, and roster suboptimality. We show that the first three are what the standard objective rewards: cost-accuracy Pareto efficiency on realized costs favors escalating moderately hard queries over the hardest ones, shorter queries over longer ones, and routing by a query's source over its difficulty. We also argue that the two assumptions that would justify large rosters, model granularity and model specialization, do not hold empirically. We propose an alternative evaluation methodology that does not reward these patterns, and as a proof of concept, we design a simple two-model router that avoids all four. Nevertheless, its gain over random routing is limited, because a well-chosen roster leaves little to route.
Chinese Translation
动态 LLM 路由器承诺通过将每个查询发送到能够正确回答的最便宜模型来降低推理成本。我们分析了六个商业路由器在涵盖八个任务类别的多样化基准测试中的 14 种设置，发现没有一个路由器的表现优于在匹配成本下随机选择两个精心挑选模型的路由器。有些路由器的表现低于 10 个百分点。我们将这一差距追溯到路由器中普遍存在的四种模式：困难盲点、长度逆转、语义匹配和名单次优性。我们表明，前面三种模式是标准目标所奖励的：在实现成本上，成本-准确性帕累托效率偏向于提升适度困难的查询而非最困难的查询，偏向于较短的查询而非较长的查询，以及根据查询的来源而非其难度进行路由。我们还认为，支持大名单的两个假设，即模型粒度和模型专业化，在实证上并不成立。我们提出了一种替代评估方法，不奖励这些模式，并作为概念验证，我们设计了一个简单的双模型路由器，避免了所有四种模式。然而，它相对于随机路由的增益有限，因为精心挑选的名单几乎没有可供路由的选择。
cs.AI / 48 / 2610.02792
Law And Order: Tax Law Autoformalization
法律与秩序：税法自动形式化
Abstract
Legal systems are increasingly implemented through software, yet scalable methods for translating legal texts into accurate symbolic representations remain underdeveloped. We study this problem through tax law, where forms and filing instructions define large computational structures involving arithmetic, branching, recursion, and tabular reasoning. We propose Law&Order, a neuro-symbolic framework for automatically formalizing tax forms and instructions into executable symbolic programs. Our approach establishes two forms of correspondence between law and logic: structural correspondence, which aligns legal and symbolic components such as cells and schedules, and denotational correspondence, which requires symbolic components to implement the computations specified by their legal counterparts. We combine large language model synthesis with cell-level verification and iterative localized error repair using human-written OpenTaxSolver tax returns. We then evaluate the resulting formalizations on independently authored, held-out TaxCalcBench returns, that are never exposed during generation or repair. Although the most advanced LLM achieves only 66% accuracy, Law&Order achieves 100% cell-level and form-level accuracy on 51 held-out returns, demonstrating the effectiveness of combining LLM-based synthesis with symbolic verification for scalable and verifiable large-scale legal autoformalization compared with using an LLM alone.
Chinese Translation
法律系统越来越多地通过软件实现，但将法律文本翻译成准确符号表示的可扩展方法仍然不够成熟。我们通过税法研究这一问题，其中表格和申报说明定义了涉及算术、分支、递归和表格推理的大型计算结构。我们提出了Law&Order，一个神经符号框架，用于自动将税务表格和说明形式化为可执行的符号程序。我们的方法在法律与逻辑之间建立了两种对应关系：结构对应关系，将法律和符号组件（如单元格和计划）对齐；而指称对应关系则要求符号组件实现其法律对应物所指定的计算。我们结合大型语言模型的合成与单元格级验证，以及使用人类编写的OpenTaxSolver税务申报表进行迭代局部错误修复。随后，我们在独立创作的、未在生成或修复过程中暴露的TaxCalcBench申报表上评估了结果形式化。尽管最先进的LLM仅达到66%的准确率，但Law&Order在51个未见的申报表上实现了100%的单元格级和表格级准确率，证明了将基于LLM的合成与符号验证相结合的有效性，相较于单独使用LLM，能够实现可扩展和可验证的大规模法律自动形式化。
cs.AI / 49 / 2610.02793
PAPER2LLM++: Continual Self-Evolution of LLMs from Research Papers
PAPER2LLM++：基于研究论文的LLM持续自我演化
Abstract
Research on LLMs continually uncovers model limitations, their causes, and potential solutions. Yet these human discoveries remain largely disconnected from model evolution: an LLM does not automatically learn from new research about its own failures. We introduce PAPER2LLM++, a framework for continual self-evolution of LLMs from research papers. Rather than treating papers merely as knowledge to retrieve, PAPER2LLM++ uses the growing literature as a stream of evidence and supervision for model improvement. For each incoming paper, it extracts evidence-grounded findings, tests whether the reported limitation persists in the current model, and, when needed, converts the findings into candidate learning signals. A try-evaluate-commit procedure integrates an update only when it improves the targeted behavior without substantially forgetting prior improvements or degrading general capabilities. Across a sequential stream of research-discovered LLM failures, we show that models can progressively incorporate new findings while retaining earlier gains. PAPER2LLM++ thus takes a step toward closing the loop between human discovery and model evolution, enabling models to continually learn from research about their own limitations and improvements.
Chinese Translation
对大型语言模型（LLMs）的研究不断揭示模型的局限性、其原因及潜在解决方案。然而，这些人类发现与模型演化之间仍然存在较大脱节：LLM并不会自动从关于其自身缺陷的新研究中学习。我们提出了PAPER2LLM++，一个基于研究论文的LLM持续自我演化框架。PAPER2LLM++并不只是将论文视为可检索的知识，而是将不断增长的文献视为模型改进的证据和监督流。对于每一篇新论文，它提取基于证据的发现，测试所报告的局限性是否在当前模型中仍然存在，并在必要时将这些发现转化为候选学习信号。一个尝试-评估-承诺的程序仅在更新能够改善目标行为而不显著遗忘先前改进或降低整体能力时才会整合更新。在一系列研究发现的LLM缺陷中，我们展示了模型可以逐步吸收新发现，同时保留早期的进展。因此，PAPER2LLM++向缩小人类发现与模型演化之间的差距迈出了重要一步，使模型能够持续从研究中学习其自身的局限性和改进。
cs.AI / 50 / 2610.02796
Modeling Shared and Individual Structure for Cross-Subject Continuous Affect Regression from EEG-fNIRS
基于EEG-fNIRS的跨受试者连续情感回归模型中的共享与个体结构建模
Abstract
Continuous, second-by-second valence-arousal estimation from physiological signals is typically studied in a subject-dependent setting, where the model sees labeled data from the same person it is later evaluated on. We study the harder zero-shot cross-subject variant on a synchronized EEG-fNIRS dataset: predict raw-scale ([1, 255]) valence and arousal trajectories for subjects whose labels the model never observes, given only their unlabeled EEG/fNIRS recordings while watching the same video stimuli as a disjoint set of training subjects. We decompose the affect trajectory into a structure shared across subjects who watch the same stimuli and an individual structure estimated for each test subject from a label-free EEG marker (alpha-band cross-channel synchrony), which rescales the shared trajectory around the scale midpoint. We validate the per-subject calibration mechanism on four independent axes: leave-one-subject-out correlation between the marker and each subject's true optimal gain, a functional-form comparison against non-linear alternatives, a repeated leave-4-out component ablation isolating each part of the pipeline's contribution, and a ceiling analysis bounding the remaining headroom for per-subject scaling. On held-out subjects, the model reaches an overall MAE of 25.96 / 22.80 across two evaluation batches (valence 21.94 / 19.6, arousal 29.98 / 26.0), well below EEGNet and ASAC-Net baselines reported for the same subject-independent split (raw scale score 60.6 and 55.0 respectively). We further report a systematic negative-result search across model architectures, feature representations, and prediction targets that found no signal able to improve on the single alpha-synchrony marker.
Chinese Translation
从生理信号中进行连续的逐秒情感价值-唤醒估计通常是在受试者依赖的设置中进行研究，在这种情况下，模型会看到来自同一人的标记数据，并在其后进行评估。我们研究了在同步EEG-fNIRS数据集上更具挑战性的零样本跨受试者变体：在模型从未观察到标签的受试者中，预测原始尺度（[1, 255]）的情感价值和唤醒轨迹，仅基于他们在观看与一组训练受试者不重叠的视频刺激时的未标记EEG/fNIRS录音。我们将情感轨迹分解为一个在观看相同刺激的受试者之间共享的结构和一个为每个测试受试者估计的个体结构，该个体结构基于无标签EEG标记（α波段跨通道同步），并在尺度中点周围重新缩放共享轨迹。我们在四个独立的维度上验证了每个受试者的校准机制：标记与每个受试者的真实最佳增益之间的留一受试者相关性、与非线性替代方案的功能形式比较、重复的留4出组件消融以隔离管道每个部分的贡献，以及一个天花板分析以界定每个受试者缩放的剩余提升空间。在保留的受试者中，模型在两个评估批次中达到整体平均绝对误差（MAE）为25.96 / 22.80（情感价值21.94 / 19.6，唤醒29.98 / 26.0），远低于针对相同受试者独立拆分的EEGNet和ASAC-Net基线（原始尺度得分分别为60.6和55.0）。我们还报告了在模型架构、特征表示和预测目标上的系统性负结果搜索，未发现能够改善单一α同步标记的信号。
cs.AI / 51 / 2610.02800
BitNest: Bit-Nested Speculative Decoding for Memory-Efficient LLM Inference Acceleration
BitNest：用于内存高效大语言模型推理加速的比特嵌套推测解码
Abstract
Speculative decoding accelerates autoregressive generation by using a lightweight draft to propose multiple tokens for parallel verification. However, existing methods often require an additional draft model or weight representation, introducing non-negligible memory overhead on resource-constrained devices. Self-speculative approaches reduce this overhead, yet still face trade-offs between draft quality, target quality, and storage efficiency. We propose BitNest, a bit-nested speculative decoding framework that embeds a low-precision draft directly into the higher-precision target representation. Instead of deriving a draft from a predefined target, BitNest first constructs a strong low-precision base and then recovers the higher-precision target through residual refinement, enabling both models to share a single physical weight representation. BitNest further extends this progressive-precision design to the KV cache for long-context inference. Across multiple 7B--8B edge-friendly LLMs and diverse workloads, BitNest achieves an average speculative acceptance rate of 95.2% while closely preserving higher-precision model quality, and delivers 1.48--1.61x end-to-end speedup over FP16 autoregressive decoding. On the LLaMA models supported by all representative self-speculative baselines, BitNest also achieves consistently competitive or higher decoding speedup.
Chinese Translation
推测解码通过使用轻量级草稿提出多个令牌以进行并行验证，从而加速自回归生成。然而，现有方法通常需要额外的草稿模型或权重表示，这在资源受限的设备上引入了不可忽视的内存开销。自推测方法减少了这种开销，但仍面临草稿质量、目标质量和存储效率之间的权衡。我们提出了BitNest，一个比特嵌套推测解码框架，它将低精度草稿直接嵌入到高精度目标表示中。BitNest并不是从预定义的目标推导草稿，而是首先构建一个强大的低精度基础，然后通过残差细化恢复高精度目标，使得两个模型能够共享一个物理权重表示。BitNest进一步将这种渐进精度设计扩展到KV缓存，以支持长上下文推理。在多个7B--8B边缘友好的大语言模型和多样化的工作负载中，BitNest实现了95.2%的平均推测接受率，同时紧密保持高精度模型质量，并在FP16自回归解码上实现了1.48--1.61倍的端到端加速。在所有代表性的自推测基线支持的LLaMA模型上，BitNest也实现了持续具有竞争力或更高的解码加速。
cs.AI / 52 / 2610.02801
VIGOR: Zero-Shot Visual Generalization via Latent-Space Consistency in Model-Based Reinforcement Learning
VIGOR：通过模型基础强化学习中的潜在空间一致性实现零样本视觉泛化
Abstract
Model-based reinforcement learning (MBRL) achieves strong sample efficiency by planning within learned latent dynamics, yet its performance degrades substantially under unseen visual distractions such as background variations, lighting changes, or camera shifts. Unlike model-free RL, where encoder perturbations affect only single-step predictions, MBRL suffers from a two-level vulnerability: visual distractions first push encoder outputs out of distribution, and these errors then compound through recursive latent rollouts over the planning horizon. We propose visual generalization via latent-space consistency in model-based RL (VIGOR), a framework that enables zero-shot generalization to unseen visual distractions while retaining the sample efficiency of its MBRL backbone. VIGOR integrates three interdependent components: (i) asymmetric weak-to-strong augmentation, which pairs weak-only and weak-to-strong latent views within a single batch; (ii) dynamics-level consistency, which enforces augmentation-invariant transition predictions through direct latent regression; and (iii) encoder-level stabilization, which prevents encoder drift under the cross-augmentation supervision imposed by dynamics-level consistency. Evaluations on the DeepMind Control Suite (DMC) and Robosuite show that VIGOR outperforms state-of-the-art model-free and model-based baselines, surpassing the second-best baseline by 3.4% on DMC and 43.6% on Robosuite. Ablations further show that VIGOR's robustness is augmentation-agnostic: replacing the default augmentation with alternatives from distinct perturbation families preserves strong generalization, confirming that latent-space consistency, not the augmentation choice, drives robustness.
Chinese Translation
基于模型的强化学习（MBRL）通过在学习的潜在动态中进行规划，达到了强大的样本效率，但在面对未见过的视觉干扰（如背景变化、光照变化或相机位移）时，其性能显著下降。与模型无关的强化学习不同，后者的编码器扰动仅影响单步预测，MBRL则面临双重脆弱性：视觉干扰首先使编码器输出超出分布，然后这些错误在规划视野内通过递归潜在展开累积。我们提出了一种通过潜在空间一致性实现的基于模型的强化学习视觉泛化框架（VIGOR），该框架能够在保持MBRL骨干的样本效率的同时，实现对未见视觉干扰的零样本泛化。VIGOR整合了三个相互依赖的组件：（i）不对称的弱到强增强，它在单个批次中将弱仅和弱到强的潜在视图配对；（ii）动态级一致性，通过直接潜在回归强制增强不变的过渡预测；（iii）编码器级稳定性，防止在动态级一致性施加的交叉增强监督下编码器漂移。在DeepMind控制套件（DMC）和Robosuite上的评估表明，VIGOR超越了最先进的模型无关和基于模型的基线，在DMC上超越第二最佳基线3.4%，在Robosuite上超越43.6%。消融实验进一步表明，VIGOR的鲁棒性与增强无关：用来自不同扰动家族的替代增强替换默认增强，仍能保持强泛化，确认潜在空间一致性而非增强选择驱动了鲁棒性。
cs.AI / 53 / 2610.02808
ROUTEAUDIT: Interaction-Aware Identification for Budgeted Multi-Verifier Routing
ROUTEAUDIT：预算多验证者路由的交互感知识别
Abstract
Adaptive multi-verifier systems are commonly compared through endpoint quality-cost gaps, even when the verifier catalog, availability, accounting, information filtration, or scorer changes with the policy. We formulate verifier routing as a contract-conditioned identification problem. The contract records request support, verifier catalog, realized availability, resource accounting, online filtration, and post-trace scoring; a matched route contrast changes only the policy coordinate. ROUTEAUDIT adds three measurable objects to this contract. A contract lattice averages coordinate increments over every admissible bridge order and reports the resulting attribution together with its path sensitivity. A policy-independent response tape identifies paired sequential contrasts when adaptive policies reveal different observations. For incomplete matching, request-level bounds use whichever potential outcome remains observed and give a sharp finite-population interval. The protocol commits paid observations and ledger events before the oracle join and returns an attribution certificate for each comparison. On two held-out raw-tail caches, matched static SF+SA equals the cascade, assigning the apparent gains of 0.1797 and 0.1250 over full static to the verifier-set edge. On 1,319 held-out task requests, the learned and RLVR studies report quality 0.9522 and 0.9553 versus 0.9484 for matched static; the RLVR-static paired difference is +0.0068 with a request-paired interval $[0.0015,0.0122]$ and a training-seed-by-request hierarchical interval $[0.0006,0.0131]$. Controlled attribution recovery yields route mean absolute error 0.0011 and endpoint reconstruction error 0.0004. Factorial, bridge-order, and stochastic-provider studies evaluate the certificate interface; RLVR supplies a learned-policy stress test under the same identification contract.
Chinese Translation
自适应多验证者系统通常通过端点质量-成本差距进行比较，即使在验证者目录、可用性、资源核算、信息过滤或评分者因政策而变化的情况下。我们将验证者路由形式化为一个合同条件的识别问题。该合同记录请求支持、验证者目录、实际可用性、资源核算、在线过滤和后追踪评分；匹配的路由对比仅更改政策坐标。ROUTEAUDIT 向该合同添加了三个可测量的对象。合同格在每个可接受的桥接顺序上平均坐标增量，并报告结果归属及其路径敏感性。政策无关的响应带在自适应政策揭示不同观察时识别配对的顺序对比。对于不完全匹配，请求级界限使用任何剩余观察到的潜在结果，并给出一个明确的有限总体区间。该协议在神谕连接之前承诺已支付的观察和账本事件，并为每个比较返回一个归属证书。在两个保留的原始尾缓存上，匹配的静态 SF+SA 等于级联，将 0.1797 和 0.1250 的明显增益分配给验证者集边缘。在 1,319 个保留的任务请求上，学习的和 RLVR 研究报告的质量分别为 0.9522 和 0.9553，而匹配静态为 0.9484；RLVR-静态配对差异为 +0.0068，配对请求区间为 $[0.0015,0.0122]$，按请求的训练种子层次区间为 $[0.0006,0.0131]$。受控归属恢复产生路由平均绝对误差 0.0011 和端点重建误差 0.0004。因子、桥接顺序和随机提供者研究评估证书接口；RLVR 在相同识别合同下提供学习政策压力测试。
cs.AI / 54 / 2610.02815
iS-KV: Online Low-Rank KV Cache Compression via Block-Incremental SVD
iS-KV：通过块增量奇异值分解实现在线低秩KV缓存压缩
Abstract
Long chain-of-thought reasoning substantially increases KV-cache memory during autoregressive decoding, as every generated token introduces new key and value states and causes the cache to grow linearly with decoding length. Existing KV-cache compression methods typically control this growth through token eviction, but irreversible deletion can remove historical states that later reasoning may need to revisit. SVD-based low-rank compression provides an alternative by retaining all positions with a more compact representation. However, extending it from a fixed prompt cache to online decoding is non-trivial. Through our investigation, we find that if the basis is updated for new tokens while old tokens keep their coordinates in the old basis, the stored history drifts substantially. Based on this observation, we propose iS-KV, an online low-rank KV-cache compression method for long-horizon reasoning. iS-KV keeps a recent window exact while incrementally folding older states into bounded-rank representations. As the low-rank basis evolves, it synchronizes historical coordinates with the updated basis to maintain representation consistency. On DeepSeek-R1-Distill-Llama-8B, iS-KV achieves 82.6% accuracy at 4.06-fold persistent-KV compression, close to the original model's 83.6%. On Qwen3-8B, it achieves 89.2% accuracy at 5.64-fold compression. Under matched memory budgets, iS-KV consistently outperforms token-eviction baselines.
Chinese Translation
在自回归解码过程中，长链推理显著增加了KV缓存的内存，因为每生成一个标记都会引入新的键和值状态，并导致缓存随着解码长度线性增长。现有的KV缓存压缩方法通常通过标记驱逐来控制这种增长，但不可逆的删除可能会移除后续推理可能需要重新访问的历史状态。基于奇异值分解（SVD）的低秩压缩提供了一种替代方案，通过更紧凑的表示保留所有位置。然而，将其从固定提示缓存扩展到在线解码并非易事。通过我们的研究，我们发现如果在更新新标记的基础时，旧标记保持在旧基础中的坐标，存储的历史将会显著漂移。基于这一观察，我们提出了iS-KV，一种用于长时间推理的在线低秩KV缓存压缩方法。iS-KV保持最近窗口的精确性，同时将较旧的状态逐步折叠为有界秩表示。随着低秩基础的演变，它将历史坐标与更新后的基础同步，以保持表示的一致性。在DeepSeek-R1-Distill-Llama-8B上，iS-KV在4.06倍持久KV压缩下实现了82.6%的准确率，接近原始模型的83.6%。在Qwen3-8B上，它在5.64倍压缩下实现了89.2%的准确率。在匹配的内存预算下，iS-KV始终优于标记驱逐基线。
cs.AI / 55 / 2610.02824
MetaRubric: Learning to Reward for Rubric-Based Reinforcement Learning
MetaRubric：学习基于评分标准的强化学习奖励
Abstract
Rubric-based reinforcement learning extends reward-driven optimization to open-ended tasks by assigning partial credit to individual response requirements. However, rubric judges can assign a high criterion score even when the information or action it requires is absent from the response, a failure mode we term Vacuous Credit. Such awards persist after the required information is removed and can reverse the sign of a response's GRPO advantage. To address this problem, we introduce MetaRubric, which alternates evidence-aware policy optimization with response-guided rubric adaptation. We construct counterfactual counterparts by changing one task-relevant fact in each prompt. During policy optimization, credit is assigned only when the response contains sufficient evidence to satisfy the required rubric criterion. After each policy-optimization stage, current policy responses guide revisions to original and counterfactual criteria while preserving the meaning of the original prompt's initial rubric as interpreted under each prompt's facts. We also adapt criterion weights at stage boundaries to better address observed policy errors. Across multiple backbones, MetaRubric improves PubMedQA accuracy by 6.00--20.40 percentage points over static-judge GRPO, with further gains on HealthBench-Hard and two multimodal medical benchmarks.
Chinese Translation
基于评分标准的强化学习通过为个别响应要求分配部分积分，将奖励驱动的优化扩展到开放式任务。然而，评分标准评审者即使在响应中缺失所需的信息或行动时，也可能给予高标准分数，这种失败模式我们称之为“空洞信用”（Vacuous Credit）。这种奖励在所需信息被移除后仍然存在，并可能反转响应的GRPO优势的符号。为了解决这个问题，我们提出了MetaRubric，它交替进行证据感知的策略优化与响应引导的评分标准调整。我们通过在每个提示中改变一个与任务相关的事实来构建反事实对照。在策略优化过程中，只有当响应包含足够的证据以满足所需评分标准时，才会分配积分。在每个策略优化阶段之后，当前策略响应引导对原始和反事实标准的修订，同时保留原始提示在每个提示事实下的初始评分标准的含义。我们还在阶段边界调整标准权重，以更好地解决观察到的策略错误。在多个基础模型上，MetaRubric相较于静态评审者的GRPO提高了PubMedQA的准确性6.00--20.40个百分点，并在HealthBench-Hard和两个多模态医学基准上取得了进一步的提升。
cs.AI / 56 / 2610.02826
Scaling Trajectories for Complex Tasks through Recursive Self-Rewrite
通过递归自重写扩展复杂任务的轨迹
Abstract
Successful trajectories on difficult tasks provide valuable supervision for model improvement, but specialized harnesses introduce interventions that may be unavailable during deployment. We propose Recursive Self-Rewrite (RSR), a framework that uses one base model, Qwen-3.8-27B, to discover successful solutions under diverse harnesses and reconstruct them as training trajectories under a general harness. A planner extracts procedures into runbooks, a critic screens for verifier and solution leakage and guides recursive revision, and an executor follows qualified runbooks in fresh sandboxes. Across approximately 3K self-curated terminal tasks, three harnesses jointly solve 759 tasks, 34.3% more than the strongest individual harness in the recorded pool. RSR expands 2,001 successful source trajectories into 11,094 rewritten trajectories for supervised finetuning. Training on these trajectories outperforms both the base model and direct trajectory SFT. Compared with the base model, pass@3 increases from 57.0% to 74.2% on Terminal-Bench 2, from 1.5% to 9.1% on Terminal-Bench 4, from 39.0% to 63.0% on our self-curated Terminal-Bench Hard, and from 3.0% to 6.0% on our Software Terminal-Bench. Process reward on Long-Horizon Terminal-Bench rises from 0.21 to 0.29. These results show how diverse harness-assisted experiences can be reconstructed into reusable capabilities for a model operating under a general harness.
Chinese Translation
在困难任务上成功的轨迹为模型改进提供了宝贵的监督，但专门的工具引入的干预措施在部署时可能不可用。我们提出了递归自重写（Recursive Self-Rewrite, RSR）框架，该框架使用一个基础模型Qwen-3.8-27B，在多种工具下发现成功的解决方案，并将其重构为在通用工具下的训练轨迹。规划器将程序提取为操作手册，评论者筛选验证器和解决方案泄漏，并指导递归修订，执行者在新的沙箱中遵循合格的操作手册。在大约3000个自我策划的终端任务中，三个工具共同解决了759个任务，比记录池中最强的单一工具多出34.3%。RSR将2001个成功的源轨迹扩展为11094个重写轨迹以进行监督微调。在这些轨迹上的训练优于基础模型和直接轨迹的SFT。与基础模型相比，在Terminal-Bench 2上的pass@3从57.0%提高到74.2%，在Terminal-Bench 4上从1.5%提高到9.1%，在我们自我策划的Terminal-Bench Hard上从39.0%提高到63.0%，在我们的Software Terminal-Bench上从3.0%提高到6.0%。在Long-Horizon Terminal-Bench上的过程奖励从0.21上升到0.29。这些结果表明，多样化的工具辅助经验可以重构为在通用工具下运行的模型的可重用能力。
cs.AI / 57 / 2610.02827
MLCommons Jailbreak Benchmark v1.0
MLCommons 越狱基准测试 v1.0
Maple, Carsten, Yucel, Cagatay, Holeman, Isaac, Knotz, Chris, Mattson, Peter, Goel, James, Petit, Jonathan, McGregor, Sean, Ezick, James, Kumar, Abhishek, Parrish, Alicia, Emani, Murali, Iyer, Kashyap, Khattak, Faiza Khan, Mbonu, Washington, Machlab, Daniel, Long, Eileen, Ghosh, Shaona, Varghese, Jibin, Lutz, Roman, Gruen, Andrew, Hillenbrand, Bennett, Gupta, Prabal, Serrhini, Mohammed, Nagasubramanian, Dhivya, Gupta, Aakash, Jun, Lu, Bollacker, Kurt, Liu, Chang, Petit, Jonathan, Chen, Cong, Monteuuis, Jean-Philippe, Miller, Brent, Verma, Apurv, Eng, Roman, Foundjem, Armstrong, Serrhini, Mohammed
Abstract
Modern AI systems are designed to refuse hazardous requests. A jailbreak is a prompt crafted to bypass those safeguards and elicit outputs that the system would normally refuse to provide. The MLCommons Jailbreak Benchmark v1.0 provides an end-to-end methodology for evaluating the robustness of large language models to single-turn, text-based jailbreak attacks. It combines criteria-driven system and attack selection, paired baseline and adversarial evaluation, human annotation, automated evaluator calibration, scoring, grading, and risk-calibrated disclosure within a single benchmarking pipeline. The benchmark evaluates eight open-weight systems using 264 seed prompts spanning eleven hazard categories and representative attacks drawn from the MLCommons Jailbreak Taxonomy. Responses are assessed using the AILuminate Assessment Standard v1.4, and robustness is measured through the Resilience Gap: the change in safety performance between baseline and adversarial conditions. Across all evaluated systems and attacks, the unsafe-response rate increased from 11.08% under baseline conditions to 18.65% under jailbreak conditions, producing an average Resilience Gap of 7.57%. Accessible systems showed a larger mean gap, while attack effectiveness varied substantially across attack categories and hazards. The benchmark also examines evaluator reliability and sources of measurement error. Beyond reporting results, Jailbreak Benchmark v1.0 establishes a reproducible methodological foundation for comparative jailbreak evaluation and for future expansion across systems, attacks, hazards, and evaluation methods.
Chinese Translation
现代人工智能系统被设计为拒绝危险请求。越狱是为了绕过这些保护措施而精心设计的提示，旨在引发系统通常会拒绝提供的输出。MLCommons 越狱基准测试 v1.0 提供了一种端到端的方法论，用于评估大型语言模型对单轮文本越狱攻击的鲁棒性。该基准结合了基于标准的系统和攻击选择、配对基线和对抗性评估、人类注释、自动评估器校准、评分、等级评定以及风险校准披露，形成一个单一的基准测试流程。该基准使用264个种子提示评估八个开放权重系统，这些提示涵盖了十一类危险类别和从 MLCommons 越狱分类法中提取的代表性攻击。响应使用 AILuminate 评估标准 v1.4 进行评估，鲁棒性通过韧性差距（Resilience Gap）进行测量：基线条件与对抗条件下安全性能的变化。在所有评估的系统和攻击中，不安全响应率在基线条件下从11.08%增加到越狱条件下的18.65%，产生了平均韧性差距为7.57%。可接入系统显示出更大的平均差距，而攻击效果在攻击类别和危险之间变化显著。该基准还考察了评估者的可靠性和测量误差的来源。除了报告结果外，越狱基准测试 v1.0 还建立了一个可重复的方法论基础，以便进行比较越狱评估，并为未来在系统、攻击、危险和评估方法上的扩展奠定基础。
cs.AI / 58 / 2610.02828
FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training
FSPO：预算有限的 LLM 强化学习后训练中的政策一致风险与帕累托可行控制
Abstract
Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL regularization, verifier allocation, and update budget. Three coupled issues remain unresolved. A future-risk model trained from behavior trajectories need not estimate the risk induced by the controller that will be deployed; a score calibrated on logged state-action pairs can become miscalibrated after selective action choice; and independent per-resource minimum costs do not in general certify a feasible multi-resource continuation. We introduce FSPO, a feedback-state controller for budgeted LLM RL post-training that addresses these issues jointly. FSPO learns a policy-consistent risk-to-go model whose Bellman target follows the same frozen controller used for future decisions, together with a long-horizon utility model. Decision-conditioned trajectory calibration (DCTC) calibrates risk on cross-fitted trajectories generated by actions selected by provisional controllers. A Pareto resource continuation certificate (PRCC) admits an action only when a non-dominated cumulative reservation remains feasible over the residual horizon. Under a matched GRPO resource envelope, FSPO reaches 66.11% held-out and 59.43% OOD accuracy, compared with 64.47% and 57.03% for PB2, the strongest evaluated adaptive baseline. Three paired training seeds give gains of +2.42 and +3.19 percentage points over the contextual bandit on held-out and OOD evaluation. Under high behavior-deployment mismatch, policy-consistent risk lowers selected-decision ECE from 0.108 to 0.053; DCTC lowers it from 0.039 to 0.022 at matched acceptance; PRCC removes false-feasible admissions on an 18-action catalog ($0.197\rightarrow0.000$); and enabling all three components reduces trajectory failure from 0.181 to 0.083 in a factorial ablation.
Chinese Translation
自适应 LLM 强化学习后训练在线更改多个训练执行器，包括展开温度、组大小、剪切、KL 正则化、验证器分配和更新预算。三个耦合问题尚未解决。基于行为轨迹训练的未来风险模型不一定能够估计将要部署的控制器所引发的风险；基于记录的状态-动作对校准的分数在选择性动作选择后可能会失去校准；而独立的每资源最低成本通常不能证明可行的多资源延续。我们提出了 FSPO，一种预算有限的 LLM 强化学习后训练的反馈状态控制器，旨在共同解决这些问题。FSPO 学习一个政策一致的风险前景模型，其贝尔曼目标遵循用于未来决策的相同冻结控制器，并结合一个长时间范围的效用模型。决策条件轨迹校准（DCTC）在由临时控制器选择的动作生成的交叉拟合轨迹上校准风险。帕累托资源延续证书（PRCC）仅在非支配的累积预留在剩余时间范围内仍然可行时才允许某个动作。在匹配的 GRPO 资源包下，FSPO 在持出和 OOD 准确率上分别达到 66.11% 和 59.43%，而 PB2（最强评估的自适应基线）分别为 64.47% 和 57.03%。三个配对训练种子在持出和 OOD 评估上分别比上下文赌博机提高了 +2.42 和 +3.19 个百分点。在高行为-部署不匹配的情况下，政策一致风险将选定决策的 ECE 从 0.108 降低到 0.053；DCTC 在匹配接受下将其从 0.039 降低到 0.022；PRCC 在一个包含 18 个动作的目录上消除了虚假可行的接纳（$0.197
ightarrow0.000$）；启用所有三个组件将轨迹失败率从 0.181 降低到 0.083，经过因子消融实验验证。
cs.AI / 59 / 2610.02831
AMBER: Multi-View Adaptive Budget Allocation for Listwise Vision-Language Reranking
AMBER：用于列表式视觉-语言重排序的多视角自适应预算分配
Abstract
Vision-language models (VLMs) are powerful listwise rerankers for multimodal retrieval, but high inference costs restrict them to evaluating small local candidate views. Existing multi-call strategies rely on fixed schedules, wasting expensive VLM calls on uninformative candidate pairs and easy queries. To address this, we propose Adaptive Multi-view Budgeted Elo Reranking (AMBER), an online, budgeted multi-view reranking framework that dynamically optimizes global resource allocation. AMBER treats fragmented listwise VLM outputs as local tournaments, using continuous Elo updates to maintain a lightweight global ranking state. Building on this, it allocates computation at two levels: dynamically constructing candidate views with high score ambiguity, and scheduling queries to maximize expected information gain. We show that each Elo update corresponds to a stochastic gradient ascent step on the Bradley-Terry log-likelihood, and provide a submodular information-theoretic motivation for the query-level allocation strategy. Experiments on CIRR, CIRCO, and PhotoBench demonstrate that AMBER achieves the strongest overall performance among the compared multi-call VLM reranking methods under comparable VLM-call budgets, while remaining effective in lower-budget settings. Our code is publicly available at https://github.com/wnlfc/AMBER.
Chinese Translation
视觉-语言模型（VLMs）是多模态检索中强大的列表式重排序工具，但高推理成本限制了它们对小规模局部候选视角的评估。现有的多调用策略依赖于固定的调度，浪费了在无信息候选对和简单查询上的昂贵VLM调用。为了解决这个问题，我们提出了自适应多视角预算Elo重排序（AMBER），这是一种在线的预算多视角重排序框架，能够动态优化全局资源分配。AMBER将碎片化的列表式VLM输出视为局部锦标赛，利用连续的Elo更新来维持轻量级的全局排名状态。在此基础上，它在两个层面上分配计算：动态构建具有高得分模糊性的候选视角，以及调度查询以最大化预期信息增益。我们展示了每次Elo更新对应于Bradley-Terry对数似然的随机梯度上升步骤，并为查询级别的分配策略提供了次模信息论动机。在CIRR、CIRCO和PhotoBench上的实验表明，AMBER在可比的VLM调用预算下，在比较的多调用VLM重排序方法中实现了最强的整体性能，同时在低预算设置中仍然有效。我们的代码已公开发布在https://github.com/wnlfc/AMBER。
cs.AI / 60 / 2610.02844
DNAlign: Dynamic Null-Space Safe Alignment for LLMs
DNAlign：大型语言模型的动态零空间安全对齐
Abstract
Ensuring the safe and reliable deployment of large language models (LLMs) remains a fundamental challenge. Existing safety alignment approaches either incur high computational cost or unintentionally disrupt the model's core knowledge, leading to degraded fluency and factual accuracy on benign tasks. This reveals a persistent trade-off between safety and utility. We propose DNAlign, a lightweight alignment framework that integrates control-theoretic optimization with null-space projection. By treating the LLM as a dynamic system, the proposed framework introduces controllable perturbations to steer generation toward safe behavior. A key component is the projection module, which restricts these perturbations to the harmful-related subspace derived from neutral hidden states, thereby preserving general knowledge and response quality. A value function trained on human preference data adaptively optimizes the control signals to align with human safety preferences. Extensive evaluations across multiple LLM backbones demonstrate that our framework consistently reduces harmful outputs while maintaining fluency, coherence, and factual utility. It achieves superior overall performance compared to prior alignment baselines without sacrificing generation diversity. These results indicate that the proposed framework provides an effective and practically deployable solution for safe LLM alignment. Code is available at https://anonymous.4open.science/r/DNAlign.
Chinese Translation
确保大型语言模型（LLMs）的安全和可靠部署仍然是一个基本挑战。现有的安全对齐方法要么成本高昂，要么无意中破坏模型的核心知识，导致在良性任务上的流畅性和事实准确性下降。这揭示了安全性与实用性之间的持续权衡。我们提出了DNAlign，一种轻量级对齐框架，结合了控制理论优化与零空间投影。通过将LLM视为动态系统，所提出的框架引入可控扰动，以引导生成朝向安全行为。一个关键组件是投影模块，它将这些扰动限制在从中性隐藏状态派生的与危害相关的子空间，从而保留一般知识和响应质量。基于人类偏好数据训练的价值函数自适应地优化控制信号，以与人类安全偏好对齐。对多个LLM基础模型的广泛评估表明，我们的框架在保持流畅性、一致性和事实实用性的同时，始终减少有害输出。与先前的对齐基准相比，它在整体性能上表现优越，而不牺牲生成的多样性。这些结果表明，所提出的框架为安全LLM对齐提供了一种有效且可实际部署的解决方案。代码可在 https://anonymous.4open.science/r/DNAlign 获取。
cs.AI / 61 / 2610.02853
Bounded Reachability & Jailbreak Detection via Contraction-Constrained State Space Models
受限可达性与越狱检测通过收缩约束状态空间模型
Abstract
Safety heads are lightweight classifiers attached to pretrained language models for flagging harmful inputs before generation. Their empirical detection performance has been studied, but their formal robustness properties remain largely unexplored. We ask when a State Space Model (SSM)-based safety head can be certified to produce the same prediction for all inputs within a bounded embedding-space perturbation. We prove that the answer turns on a single condition: the $l_\infty$ norm of the state transition matrix must satisfy $\norm{A}_\infty<1$ (the \emph{contraction condition}), which enables exact interval bound propagation (IBP) certification for linear time-invariant classifiers. When the contraction condition holds, the reachable output interval has bounded steady-state width and examples can be certified as robustly classified. When it fails, the interval grows exponentially with sequence length and certification is impossible at any practical perturbation radius. We enforce contraction with a hinge penalty and show on toxic comment data that certified fraction improves from 41\% to 59\%, with a sharp empirical phase transition at $\norm{A}_\infty=1$ matching the theory. Applying a contraction-regularized S4 head to jailbreak detection on JailbreakBench, we achieve a zero-shot transfer to AdvBench (DR=0.994) and HarmBench (DR=0.988). A logistic regression on mean-pooled Mamba-130M embeddings matches or exceeds the S4 head on every detection metric, confirming that harmful intent is already linearly separable in the embedding space. The S4 safety head's contribution is not superior discrimination but the formal certification that no probe-based approach provides.
Chinese Translation
安全头是附加在预训练语言模型上的轻量级分类器，用于在生成之前标记有害输入。其经验检测性能已被研究，但其正式的鲁棒性特性仍然在很大程度上未被探索。我们探讨何时基于状态空间模型（State Space Model, SSM）的安全头可以被认证为对所有在受限嵌入空间扰动内的输入产生相同的预测。我们证明答案取决于一个单一条件：状态转移矩阵的 $l_ ext{∞}$ 范数必须满足 $
orm{A}_ ext{∞}<1$（即 extit{收缩条件}），这使得线性时不变分类器的精确区间界限传播（Interval Bound Propagation, IBP）认证成为可能。当收缩条件成立时，可达输出区间具有有限的稳态宽度，且示例可以被认证为鲁棒分类。当条件不成立时，区间随着序列长度呈指数增长，且在任何实际扰动半径下都无法进行认证。我们通过铰链惩罚强制收缩，并在有害评论数据上显示，认证比例从41 ext{%}提高到59 ext{%}，在 $
orm{A}_ ext{∞}=1$ 处出现与理论相符的急剧经验相变。将收缩正则化的 S4 头应用于 JailbreakBench 上的越狱检测，我们实现了对 AdvBench（DR=0.994）和 HarmBench（DR=0.988）的零-shot 转移。对均值池化的 Mamba-130M 嵌入进行逻辑回归，在每个检测指标上与 S4 头匹配或超越，确认有害意图在嵌入空间中已经是线性可分的。S4 安全头的贡献不在于优越的区分能力，而在于提供了没有探测基础方法能够提供的正式认证。
cs.AI / 62 / 2610.02858
Harness-Aware Distillation for Small Language Model Agents
考虑马具的蒸馏方法用于小型语言模型代理
Abstract
Language model agents are deployed with a harness, the software around the model that manages its context, tools, and feedback. When such an agent is distilled into a smaller one, the harness stays in place, so the student mainly needs the teacher-specific abilities that the harness cannot provide, such as acting correctly on harness information. Standard distillation, however, imitates the teacher's full outputs and treats the harness as part of the input. We propose Harness-Aware Distillation (HAD), which focuses distillation on what the teacher adds beyond the harness. HAD complements on-policy distillation with two components: an action preference that contrasts the same teacher's actions with and without the harness information, scored after the student's own reasoning, and a validity check that drops preference pairs whose preferred action contradicts the harness records. We show that the contrast gives the student information that imitating the teacher alone cannot provide, and HAD needs no task rewards, success labels, or future information. Across multiple long-horizon agent benchmarks and models, HAD outperforms on-policy distillation baselines with the same fixed harness. Our analysis shows that HAD enters fewer unproductive loops and recovers from errors more often than the baselines, and suggests that it adaptively keeps learnable feedback in its weights while reading state information from the harness.
Chinese Translation
语言模型代理在马具的支持下被部署，马具是管理模型上下文、工具和反馈的软件。当这样的代理被蒸馏成一个更小的模型时，马具仍然存在，因此学生主要需要教师特定的能力，而这些能力是马具无法提供的，例如正确处理马具信息。然而，标准的蒸馏方法模仿教师的完整输出，并将马具视为输入的一部分。我们提出了考虑马具的蒸馏方法（Harness-Aware Distillation, HAD），该方法将蒸馏的重点放在教师所添加的超出马具的信息上。HAD通过两个组成部分补充了基于策略的蒸馏：一个行动偏好，它对比同一教师在有和没有马具信息时的行动，偏好在学生自身推理之后进行评分；以及一个有效性检查，它剔除那些偏好行动与马具记录相矛盾的偏好对。我们展示了这种对比为学生提供了仅仅模仿教师所无法提供的信息，并且HAD不需要任务奖励、成功标签或未来信息。在多个长时间跨度的代理基准和模型中，HAD在相同固定马具下超越了基于策略的蒸馏基线。我们的分析表明，HAD进入无效循环的次数更少，并且比基线更频繁地从错误中恢复，且建议它在从马具读取状态信息时自适应地保持可学习的反馈在其权重中。
cs.AI / 63 / 2610.02867
TACD: Distilling Efficient Text-to-Motion Models via Terminal Amplification Control
TACD：通过终端放大控制提炼高效的文本到运动模型
Abstract
Recent text-to-motion models have improved motion quality and instruction following, yet many-step denoising and large model components make deployment slow and memory-intensive. We present Terminal-Amplification-Controlled Distillation (TACD), an on-policy approach for training efficient motion generators from text prompts and pretrained teachers, without real-motion training data. Building on segmented on-policy flow distillation, we supervise clean-motion predictions along student-generated trajectories. We identify a failure mode in which velocity matching on a fixed supervision grid repeatedly overweights errors near the denoising endpoint, degrading few-step generation. TACD ties the latest teacher query to the student's step size, bounding the effective loss weights in clean-motion space without changing inference. Experiments on HumanML3D and KIT-ML demonstrate improved few-step generation, including a 58% reduction in eight-step HY-Motion student FID relative to distillation without this bound. For diffusion teachers, the endpoint-matching form of TACD yields four-step students with lower FID and matched or improved text-motion retrieval relative to their 50-step teachers on HumanML3D. On HY-Motion and Kimodo, eight-step students with compact components achieve 7.7-11.9x end-to-end speedups and reduce peak GPU memory by 3.8-6.7x relative to their teachers. Project page: https://vkgo.github.io/TACD/
Chinese Translation
近期的文本到运动模型在运动质量和指令遵循方面有所提升，但多步骤去噪和大型模型组件使得部署变得缓慢且内存密集。我们提出了终端放大控制蒸馏（Terminal-Amplification-Controlled Distillation, TACD），这是一种基于策略的方法，用于从文本提示和预训练教师中训练高效的运动生成器，而无需真实运动训练数据。基于分段的基于策略流蒸馏，我们在学生生成的轨迹上监督清晰运动的预测。我们识别出一种失败模式，即在固定监督网格上进行速度匹配时，反复过度权重去噪端点附近的错误，导致少步骤生成的质量下降。TACD将最新教师查询与学生的步长相结合，限制了清晰运动空间中的有效损失权重，而不改变推理过程。在HumanML3D和KIT-ML上的实验表明，少步骤生成得到了改善，包括相较于没有此限制的蒸馏，八步HY-Motion学生FID降低了58%。对于扩散教师，TACD的端点匹配形式使得四步学生在FID上表现更低，并且在文本-运动检索上与其50步教师相比匹配或改善。在HY-Motion和Kimodo上，具有紧凑组件的八步学生实现了7.7-11.9倍的端到端加速，并将峰值GPU内存减少了3.8-6.7倍，相较于其教师。项目页面：https://vkgo.github.io/TACD/
cs.AI / 64 / 2610.02885
PsyEvo: A Personalized Counseling Agent That Self-Evolves at Test Time
PsyEvo：一个在测试时自我进化的个性化咨询代理
Abstract
Mental health disorders affect a substantial proportion of the global population, yet a persistent shortage of trained practitioners leaves the majority without adequate care. Large language model (LLM)-based counselors present a promising direction for delivering scalable conversational psychological support. Offline model training alone leaves limited room to adapt to individual clients or to learn from ongoing therapeutic interaction at test time. We introduce PsyEvo, an LLM-based counseling framework that enables both client-specific personalization and response-policy improvement at test time through three components: Hierarchical Bayesian Skill Policy (HBSP) personalizes what intervention to apply by maintaining a per-client skill posterior updated from session feedback; Inter-session Listwise Preference Optimization (LiPO) improves how the selected skill is expressed by updating a shared response adapter from cross-client preference evidence; and State-conditioned Ordinal Credit Assignment (SOCA) supplies candidate preferences and trajectory credit to the two components through consistency-checked comparisons and ordinal projection. In simulated-client evaluation with shared online cohort adaptation, PsyEvo obtains 7.684 Overall on PsychEval and exceeds every component variant in each of three matched runs. Removing individual components lowers mean overall score by 0.138--0.171 under the shared configuration, supporting conditional contributions within the complete scaffold. Our code is available at https://github.com/Lingxi-mental-health/PsyEvo
Chinese Translation
心理健康障碍影响着全球相当大比例的人口，但训练有素的从业者持续短缺，使大多数人无法获得足够的护理。基于大型语言模型（LLM）的咨询师为提供可扩展的对话心理支持提供了一个有前景的方向。然而，仅依靠离线模型训练，无法充分适应个别客户或在测试时从持续的治疗互动中学习。我们提出了PsyEvo，一个基于LLM的咨询框架，通过三个组件实现客户特定个性化和响应策略的改进：层次贝叶斯技能策略（HBSP）通过维护从会话反馈中更新的每个客户的技能后验，个性化干预的应用；会话间列表偏好优化（LiPO）通过从跨客户偏好证据中更新共享响应适配器，改善所选技能的表达；状态条件序数信用分配（SOCA）通过一致性检查比较和序数投影，为这两个组件提供候选偏好和轨迹信用。在共享在线群体适应的模拟客户评估中，PsyEvo在PsychEval上获得了7.684的整体评分，并在三次匹配运行中超越了每个组件变体。去除单个组件在共享配置下使平均整体得分降低了0.138至0.171，支持了完整框架内的条件贡献。我们的代码可在https://github.com/Lingxi-mental-health/PsyEvo获取。
cs.AI / 65 / 2610.02897
Interpreting at Write Time: A Policy Ablation for Multi-Goal Agent Memory
写时解释：多目标智能体记忆的策略消融
Abstract
A long-running assistant cannot keep everything it has seen, so it summarises. Summarising is not neutral: what is kept is chosen against some notion of what the record is for, and that choice is made once, before anyone knows which of the user's standing goals will ask. Goals rarely disagree about what happened. They disagree about which parts of it were worth the space. Once the history is too long to re-read, the summary replaces the stream, and whatever it left out is gone. We ask what a memory should summarise for when it serves several standing goals at once. Three policies answer differently: summarise with no goal in view, write one summary covering every goal, or write one summary per goal and read them together. We compare them across several models and event streams, holding the read step fixed so that only the write differs. The goals do pull apart: summaries written for different goals overlap each other less than a summary overlaps a rewrite of itself. Per-goal summaries win on relevance, completeness and accuracy, and the all-goal summary loses even to the neutral one written at a fraction of its budget. Interpreting at write pays off, but only for the goal that later asks.
Chinese Translation
一个长期运行的助手无法保留它所见的一切，因此它会进行总结。总结并不是中立的：保留的内容是根据某种对记录目的的理解进行选择的，而这种选择是在任何人都不知道用户的长期目标会提出什么请求之前就已做出。目标之间通常不会对发生的事情产生分歧，它们的分歧在于哪些部分值得占用空间。一旦历史过长以至于无法重新阅读，总结就会取代原始记录，而被遗漏的内容则不复存在。我们探讨在同时服务于多个长期目标时，记忆应该为哪些内容进行总结。三种策略给出了不同的答案：在没有目标的情况下进行总结，写一个涵盖所有目标的总结，或者为每个目标写一个总结并一起阅读。我们在多个模型和事件流中对它们进行了比较，固定读取步骤以便仅写入方式不同。目标确实存在差异：为不同目标撰写的总结彼此重叠的程度低于一个总结与其自身重写的重叠程度。按目标撰写的总结在相关性、完整性和准确性上表现更佳，而覆盖所有目标的总结甚至不如在其预算的一小部分下撰写的中立总结。写时解释的策略是有效的，但仅对后续提出请求的目标有效。
cs.AI / 66 / 2610.02902
LUMOS: Tracing Parametric Knowledge from Training Data to Behavioral Outputs in LLMs
LUMOS：追踪大语言模型中从训练数据到行为输出的参数知识
Abstract
Current analyses of LLMs' parametric knowledge are largely output-centric, drawing conclusions about what a model knows without verifying what it was actually trained on. This leaves fundamental questions, such as whether a correct response reflects genuine generalization or rote memorization, grounded in speculation rather than evidence. To resolve these ambiguities, we introduce LUMOS, a diagnostic framework that traces knowledge along the causal chain from training-data exposure to behavioral output, leveraging OLMo 2 with its fully transparent training corpus. By grounding analysis in verified exposure, we reveal that models internally encode rare facts with high separability (84%) yet fail to express them behaviorally (54%), though this retrieval gap narrows with scale. Furthermore, when models are asked to self-reflect on their own answers, they perform reliably on trained content (83%) but drop to random-baseline levels (49%) on unseen content. This collapse persists even under chain-of-thought prompting, which inflates confidence signals rather than improving calibration. Collectively, these findings demonstrate that incorporating the training-data axis into LLM evaluation transforms speculative diagnoses into verifiable claims, and we advocate that this axis should be a standard component of knowledge assessment in LLMs.
Chinese Translation
当前对大语言模型（LLMs）参数知识的分析主要以输出为中心，得出的结论往往是关于模型所知内容，而未验证其实际训练内容。这留下了诸如正确响应是否反映了真正的概括能力或死记硬背等基本问题，这些问题更多地基于推测而非证据。为了解决这些模糊性，我们引入了LUMOS，一个诊断框架，沿着从训练数据暴露到行为输出的因果链追踪知识，利用OLMo 2及其完全透明的训练语料库。通过基于已验证的暴露进行分析，我们揭示模型内部编码稀有事实的可分离性高达84%，但在行为上表现出来的能力仅为54%，尽管这一检索差距随着规模的扩大而缩小。此外，当模型被要求自我反思其答案时，在训练内容上的表现可靠（83%），但在未见内容上的表现则降至随机基线水平（49%）。即使在链式思维提示下，这一崩溃现象仍然存在，这种提示虽然提高了信心信号，却并未改善校准。总体而言，这些发现表明，将训练数据维度纳入大语言模型评估可以将推测性诊断转变为可验证的主张，我们主张这一维度应成为大语言模型知识评估的标准组成部分。
cs.AI / 67 / 2610.02910
Frequency Is Not Sensitivity Identifying Safety-Sensitive Experts in Sparse MoE LLM
频率并非敏感性：在稀疏混合专家（MoE）大语言模型中识别安全敏感专家
Abstract
Suppressing a small set of routed experts can weaken the safety behavior of a sparse Mixture-of-Experts (MoE) language model without retraining. Which experts to suppress is therefore a security question, and the usual answer is activation frequency, but frequency measures use, not influence. We test an alternative: router-gradient sensitivity, the sensitivity of the sequence loss to the gate weights that select an expert. Across five MoE architectures, we rank experts by each signal on 500 benign and 500 malicious prompts and measure refusal on 100 held-out malicious prompts under two budgets: equal expert counts and equal nominal malicious routing traffic (1%-5%). Under each of the two budgets, router-gradient selection reduces refusals more than activation in 24 of 25 conditions, and more than a ten-trial random mean in all 25. The largest effect is in OLMoE, where refusals fall from 34 to 9 of 100 prompts (73.53% relative) with no degraded outputs, indicating substantive compliance rather than broken generation. After matching expert counts in every layer, gradient selection still produces greater refusal reduction than activation in 23 of 25 conditions, with two ties. An exploratory cross-model analysis links larger malicious-versus-benign concentration gaps to greater peak gradient effects (rho = 0.90; exact two-sided p = 0.083, n = 5). Together, the results support gradient selection under the tested budgets.
Chinese Translation
抑制一小部分路由专家可能会削弱稀疏混合专家（MoE）语言模型的安全行为，而无需重新训练。因此，抑制哪些专家是一个安全问题，通常的答案是激活频率，但频率测量的是使用，而非影响。我们测试了一种替代方法：路由梯度敏感性，即序列损失对选择专家的门控权重的敏感性。在五种MoE架构中，我们根据每个信号对500个良性和500个恶性提示进行专家排名，并在两个预算下测量对100个保留的恶性提示的拒绝情况：相等的专家数量和相等的名义恶性路由流量（1%-5%）。在这两个预算下，路由梯度选择在25种条件中有24种情况下比激活减少了更多的拒绝，并且在所有25种情况下都超过了十次试验的随机均值。在OLMoE中，拒绝数量从100个提示中的34个降至9个（相对减少73.53%），且没有输出质量下降，表明合规性显著，而非生成失败。在每一层匹配专家数量后，梯度选择在25种条件中仍有23种情况下比激活产生更大的拒绝减少，且有两个平局。探索性的跨模型分析将更大的恶性与良性集中差距与更大的峰值梯度效应联系起来（rho = 0.90；双侧精确p = 0.083，n = 5）。综合来看，结果支持在测试预算下的梯度选择。
cs.AI / 68 / 2610.02920
HASTE: Evolving Agent Harnesses Against Emerging Attacks Using Sparse Evidence
HASTE：利用稀疏证据对抗新兴攻击的进化代理工具
Abstract
Agent harnesses play a critical role in defenses by enforcing safety constraints to prevent unsafe actions. However, rapidly emerging attacks outpace manual harness adaptation, motivating automated harness evolution. Yet the signals available for harness evolution are often sparse, such as brief descriptions or a few attack examples in threat reports and preprints. To address this limitation, we introduce HASTE, a multi-agent framework that evolves agent harnesses from sparse threat evidence through an adversarial interplay between safety-specification generation and attack-case generation. Safety specifications guide harness updates toward addressing identified safety vulnerabilities, while attack cases probe for remaining safety vulnerabilities after each update. By feeding evaluation outcomes back into both processes, HASTE enables harness evolution against emerging attacks beyond the initially observed evidence. Experimental results across multiple backbone models, attack types, and evidence forms show that HASTE consistently reduces attack success rates while preserving benign-task utility. The code is available at https://github.com/xxiqiao/HASTE.
Chinese Translation
代理工具在防御中发挥着关键作用，通过强制安全约束来防止不安全的行为。然而，快速出现的攻击超出了手动工具适应的速度，这促使了自动化工具的进化。然而，用于工具进化的信号往往是稀疏的，例如威胁报告和预印本中的简要描述或少量攻击示例。为了解决这一限制，我们提出了HASTE，一个多代理框架，通过安全规范生成与攻击案例生成之间的对抗性互动，从稀疏的威胁证据中进化代理工具。安全规范指导工具更新，以解决已识别的安全漏洞，而攻击案例则在每次更新后探测剩余的安全漏洞。通过将评估结果反馈到这两个过程中，HASTE使得工具能够在最初观察到的证据之外，对抗新兴攻击。针对多个基础模型、攻击类型和证据形式的实验结果表明，HASTE始终能够降低攻击成功率，同时保持良性任务的效用。代码可在 https://github.com/xxiqiao/HASTE 获取。
cs.AI / 69 / 2610.02925
Positive-Unlabeled Learning for Agent Safety False Alarm Auditing
正负样本学习用于代理安全虚警审计
Abstract
Safety monitors help safeguard language-model agents interacting with external tools and environments, but conservative monitoring can generate many false alarms, consuming extensive review resources and weakening trust in alerts. Because false and genuine alarms often remain interleaved in native monitor scores, obtaining a reliable cutoff still requires substantial manual verification. In practice, a small set of verified-safe non-alarmed trajectories may be available while alarms remain unlabeled, naturally casting false-alarm auditing as a positive-unlabeled (PU) ranking problem. The key challenge is monitor-induced selection, since observed safe references are accepted by the monitor, while the hidden safe alarms of interest are precisely those it incorrectly flags, making the observed positives poorly representative of the positives to be recovered. To address this challenge, we propose a two-stage framework in which Trust-aware PU Supervision adapts safe references toward the alarm domain and protects plausible false alarms from excessive negative pressure, while Reliability-gated Rank Distillation consolidates consistent ordering preferences from multiple PU reference models into a single student. Consensus-guided Structural Refinement then improves the student ranking using hierarchical safe-reference support, alarm relations, and predicted reference consensus. The framework requires no alarm safety labels for fitting and leaves the underlying monitor unchanged. Across mainstream safety monitors, our method achieves a macro AUPRC of $0.6444$, outperforming eight evaluated PU baselines by 5.27--16.98 absolute percentage points; compared with PULDA, the strongest evaluated PU baseline, it recovers 33.3% more false alarms at a 5% review budget.
Chinese Translation
安全监控帮助保护与外部工具和环境交互的语言模型代理，但保守的监控可能会产生大量虚假警报，消耗大量审查资源并削弱对警报的信任。由于虚假警报和真实警报通常在原生监控评分中交错存在，因此获得可靠的截止值仍然需要大量的人工验证。在实践中，可能会有一小部分经过验证的安全非警报轨迹可用，而警报则保持未标记，这自然将虚假警报审计视为正负样本（PU）排序问题。关键挑战在于监控引起的选择，因为观察到的安全参考被监控接受，而隐藏的安全警报正是监控错误标记的对象，使得观察到的正样本对待恢复的正样本的代表性较差。为了解决这一挑战，我们提出了一个两阶段框架，其中信任感知的PU监督将安全参考适应于警报领域，并保护可能的虚假警报免受过度负面压力，而可靠性门控的排名蒸馏则将多个PU参考模型中的一致排序偏好整合到一个单一的学生模型中。共识引导的结构优化随后利用分层安全参考支持、警报关系和预测的参考共识来改善学生排名。该框架在拟合时不需要警报安全标签，并且不改变底层监控。在主流安全监控中，我们的方法实现了$0.6444$的宏AUPRC，超越了八个评估的PU基线，绝对百分比提高了5.27至16.98个百分点；与PULDA（评估的最强PU基线）相比，它在5%的审查预算下恢复了33.3%更多的虚假警报。
cs.AI / 70 / 2610.02932
When to Compile a Computer-Use Agent? Measuring Payback and Making Compilation Decisions for Token Efficiency
何时编译计算机使用代理？衡量回报并做出令牌效率的编译决策
Abstract
Compiling GUI procedures that agents execute repeatedly into programs can reduce their token costs. However, measuring payback and deciding when to compile have two challenges. First, compilation costs are uncertain because attempts can require repair and still fail to produce a usable program. Second, future reuse is unknown because tasks may stop arriving or GUI drift may stop the program from working. To address these challenges, we propose PACE (Payback-Aware Compilation from Experience), a system with a measurement protocol and an online compilation algorithm. The measurement protocol records successful and failed compilation costs, and compares agent and program execution costs on matched task inputs to estimate per-use savings and payback counts. Using these measurements, the online algorithm compares estimated future savings with compilation costs, including failed attempts, based on past task arrivals and compilation outcomes. It checks execution and compilation charges against a cumulative budget determined by observed task arrivals before allowing either action. Under stated action-cost assumptions, total cost after each arrival is at most $1+\epsilon$ times the cost of running every task with the agent. For successful compilation attempts, estimated payback counts excluding source agent runs are 2-16 uses. In simulations using recorded task arrivals, PACE reduces token costs by 17.3% compared with ReAct, 24.9% with the AutoRPA adaptation, and 17.3% with the ToolPro adaptation on average ($\epsilon=0.25$).
Chinese Translation
将代理重复执行的图形用户界面（GUI）程序编译成程序可以降低其令牌成本。然而，衡量回报和决定何时编译面临两个挑战。首先，编译成本是不确定的，因为尝试可能需要修复，并且仍然无法生成可用程序。其次，未来的重用情况未知，因为任务可能停止到达，或者GUI漂移可能导致程序无法正常工作。为了解决这些挑战，我们提出了PACE（基于经验的回报感知编译），这是一个具有测量协议和在线编译算法的系统。测量协议记录成功和失败的编译成本，并在匹配任务输入上比较代理和程序的执行成本，以估算每次使用的节省和回报次数。利用这些测量，在线算法根据过去的任务到达和编译结果，将估算的未来节省与编译成本（包括失败尝试）进行比较。在允许任何操作之前，它会根据观察到的任务到达情况检查执行和编译费用是否符合累积预算。在所述的操作成本假设下，每次到达后的总成本至多为运行每个任务与代理的成本的 $1+ ext{ε}$ 倍。对于成功的编译尝试，排除源代理运行的估算回报次数为2-16次。在使用记录的任务到达情况的模拟中，PACE与ReAct相比平均降低了17.3%的令牌成本，与AutoRPA适配相比降低了24.9%，与ToolPro适配相比平均降低了17.3%（$ ext{ε}=0.25$）。
cs.AI / 71 / 2610.02945
Continual Graph Memory for Mathematical Research Agents
用于数学研究代理的持续图记忆
Abstract
Using frontier agent harnesses to tackle mathematical research problems has emerged as an effective means of advancing mathematics. However, solving frontier problems in mathematics may require a massive number of agents working in parallel for extended periods to construct proofs, thereby generating an enormous volume of intermediate proof results. Organizing these intermediate results throughout a long-horizon proof-search process and reusing knowledge gained from prior explorations remain major challenges. We present Ansatz, a mathematical research agent built around Continual Graph Memory, a graph-based, evolvable, cross-problem mathematical research memory system that explicitly organizes the entire proof search process and reuses information from exploration trajectories of previous problems. Specifically, we develop a unified graph memory that represents all intermediate exploration results, including facts, plans, and counterexamples, together with edges that explicitly represent the relationships among them; dependency-aware retrieval supplies precisely targeted local context; an evidence-sensitive curator updates the research frontier and distills lessons from prior attempts; and scoped recall surfaces earlier statements and negative findings for local re-proving rather than uncritical reuse. Experiments cover runs across all ten First Proof Second Batch problems, together with four component studies. Ansatz reports closure on all ten research tasks, demonstrating its ability to sustain and resume long-horizon mathematical search. Beyond these problems, Ansatz also produces solutions to the Jamison caterpillar conjecture and Erd\H{o}s Problems 289, 348, and 488 without human intervention, and makes partial progress on several open problems, illustrating its strong ability to solve open mathematical research problems.
Chinese Translation
利用前沿代理来解决数学研究问题已成为推动数学发展的有效手段。然而，解决数学中的前沿问题可能需要大量代理并行工作较长时间以构建证明，从而生成大量的中间证明结果。在长时间的证明搜索过程中组织这些中间结果，并重用从先前探索中获得的知识仍然是主要挑战。我们提出了Ansatz，一个围绕持续图记忆（Continual Graph Memory）构建的数学研究代理，这是一个基于图的、可演化的、跨问题的数学研究记忆系统，明确组织整个证明搜索过程并重用先前问题的探索轨迹中的信息。具体而言，我们开发了一个统一的图记忆，表示所有中间探索结果，包括事实、计划和反例，以及明确表示它们之间关系的边；依赖感知检索提供精确定位的局部上下文；证据敏感的策展人更新研究前沿并提炼先前尝试的经验教训；而范围回忆则呈现早期陈述和负面发现，以便进行局部重证明，而非不加批判的重用。实验涵盖了所有十个第一证明第二批问题的运行，以及四个组件研究。Ansatz在所有十个研究任务上报告了闭合，展示了其维持和恢复长期数学搜索的能力。除了这些问题，Ansatz还在没有人类干预的情况下解决了Jamison毛虫猜想和Erdős问题289、348和488，并在几个开放问题上取得了部分进展，展示了其解决开放数学研究问题的强大能力。
cs.AI / 72 / 2610.02968
Reasoning with Evidence, Not Merely Rationales: Verifiable Preference Proofs for LLM-Based Recommendation
基于证据的推理，而不仅仅是理由：可验证的偏好证明用于基于大型语言模型的推荐
Abstract
Large language models (LLMs) can infer user preferences from interaction histories and reviews, yet the rationales they generate may not reflect the information actually used for recommendation. A preference claim may be weakly supported by its selected evidence, or may have little effect on the final ranking. We refer to these two failures as the grounding-influence gap. We introduce PROVE-REC, a general framework for verifiable preference reasoning in LLM-based recommendation. Pass A converts the complete pre-target history into a compact preference proof consisting of positive and avoidance claims linked to selected evidence entries. Pass B predicts the next item using only the proof and its selected evidence, preventing the recommender from bypassing the reasoning path. To verify evidence-to-proof grounding, we compare the effect of masking selected evidence with masking a comparable control entry. To verify proof-to-recommendation influence, we remove a preference claim and measure the resulting decrease in the target item's ranking margin. A ranking-preservation objective further retains useful information from the complete history. Comprehensive experiments on wide-ranging real-world datasets demonstrate that PROVE-REC consistently outperforms strong sequential, generative, and LLM-enhanced baselines, with improvements of up to 7.45%. Controlled ablations confirm the effectiveness of the two-pass architecture and verification objectives. Moreover, PROVE-REC produces claims that are more strongly grounded in historical evidence and more influential to recommendation while preserving ranking quality.
Chinese Translation
大型语言模型（LLMs）可以从用户的互动历史和评论中推断用户偏好，但它们生成的理由可能并不反映实际用于推荐的信息。一个偏好声明可能仅由其选定的证据弱支持，或者对最终排名几乎没有影响。我们将这两种失败称为基础影响差距（grounding-influence gap）。我们提出了 PROVE-REC，这是一个用于 LLM 基础推荐的可验证偏好推理的通用框架。第一阶段（Pass A）将完整的目标前历史转换为一个紧凑的偏好证明，其中包含与选定证据条目相关联的积极和避免声明。第二阶段（Pass B）仅使用证明及其选定证据来预测下一个项目，从而防止推荐系统绕过推理路径。为了验证证据与证明之间的基础关系，我们比较了掩盖选定证据与掩盖可比控制条目的效果。为了验证证明对推荐的影响，我们移除一个偏好声明并测量目标项目排名边际的下降。一个排名保持目标进一步保留了来自完整历史的有用信息。在广泛的真实世界数据集上的综合实验表明，PROVE-REC 始终优于强大的序列、生成和 LLM 增强基线，改进幅度高达 7.45%。控制性消融实验确认了双阶段架构和验证目标的有效性。此外，PROVE-REC 生成的声明在历史证据中更强地扎根，并对推荐的影响更大，同时保持排名质量。
cs.AI / 73 / 2610.02972
CreateScore: Domain-Theory-Informed Bayesian Routing for LLM-Based CV Screening
CreateScore：基于领域理论的贝叶斯路由用于大语言模型的简历筛选
Abstract
Large language models (LLMs) can support rubric-based screening of CVs, but applying a high-capability model to every candidate and criterion is costly. We present CreateScore, a domain-theory-informed Bayesian network for criterion-level LLM routing. A hand-specified directed acyclic graph with Dirichlet-multinomial conditional probability tables converts CV evidence into posterior uncertainty; low-uncertainty decisions are resolved by a local 8B model and uncertain ones are escalated to a 120B reference model. The graph is causally motivated, but the system performs standard Bayesian conditioning, not causal inference. The escalation threshold is calibrated on a training fold (target: 70% resolved locally) and then frozen. On 200 synthetic Data Science CVs (139 training and 61 test candidates, five criteria), 77.7% of criterion decisions were resolved locally (237 of 305). Relative to a reference condition in which the 120B model adjudicated every criterion, routed escalation reduced token use by 65.2% and raised exact score agreement from 32.8% (8B alone) to 42.6% (95% CI 31.0-55.1%); at n = 61 the gain was not statistically distinguishable. The uncertainty signal did not, however, identify the decisions on which the 8B model erred: disagreement with the reference was 16.2% among escalated and 19.4% among locally resolved decisions (AUROC 0.47, 95% CI 0.39-0.56), no better than random selection. We also document how an earlier evaluation was invalidated when truncated reasoning-model outputs were silently replaced by local labels, and we recommend safeguards for cascade evaluation. CreateScore is supported as an auditable cost-reduction mechanism, not yet as a targeted error detector, and is not an autonomous hiring system.
Chinese Translation
大型语言模型（LLMs）可以支持基于评分标准的简历筛选，但将高能力模型应用于每个候选人和标准的成本较高。我们提出了CreateScore，一种基于领域理论的贝叶斯网络，用于标准级别的LLM路由。一个手动指定的有向无环图（DAG）与Dirichlet-多项式条件概率表结合，将简历证据转化为后验不确定性；低不确定性的决策由本地8B模型解决，而不确定的决策则升级到120B参考模型。该图是基于因果动机构建的，但系统执行的是标准的贝叶斯条件化，而非因果推断。升级阈值在训练折叠上进行校准（目标：70%的决策在本地解决），然后被冻结。在200份合成数据科学简历（139个训练候选人和61个测试候选人，五个标准）中，77.7%的标准决策在本地得到解决（305个中的237个）。与120B模型裁定每个标准的参考条件相比，路由升级减少了65.2%的令牌使用，并将准确评分一致性从32.8%（仅8B）提高到42.6%（95%置信区间31.0-55.1%）；在n = 61时，增益在统计上没有显著区别。然而，不确定性信号并未识别出8B模型错误的决策：在升级的决策中与参考模型的不一致为16.2%，在本地解决的决策中为19.4%（AUROC 0.47，95%置信区间0.39-0.56），没有比随机选择更好。我们还记录了早期评估在截断推理模型输出被静默替换为本地标签时失效的情况，并建议为级联评估提供保障。CreateScore被支持作为一种可审计的成本降低机制，而尚未作为针对性错误检测器，并且不是一个自主招聘系统。
cs.AI / 74 / 2610.02975
Reliable Self-Evolution with Imperfect Proxy Rewards
具有不完美代理奖励的可靠自我进化
Abstract
Large language model (LLM)-based self-evolving search is a promising approach to scientific discovery. However, high-fidelity evaluation of every candidate is prohibitively expensive in some domains. Self-evolving systems in such settings therefore rely on low-cost but imperfect proxy rewards, which may assign high scores to infeasible candidates. These false positives may contaminate both the final output and the feedback used to guide subsequent generations. This motivates statistically calibrated reward intervals for more reliable self-evolving search. We propose Conformal Interval-Driven Self-Evolution (CISE), which constructs candidate-specific reward intervals using conditional conformal inference and iteration-wise online density-ratio estimation. CISE uses conservative interval-based rewards for evolutionary feedback and returns candidates only when all required property intervals lie entirely within their respective feasible regions. We derive fixed-iteration coverage results under explicit assumptions of independence and covariate shift. We evaluate CISE on three self-evolving search tasks in materials science. In our experiments, all candidates returned by CISE are true positives under high-fidelity evaluation, whereas the baselines return more candidates but include false positives. These results highlight the value of a smaller, more precise shortlist when downstream validation budgets are limited. Our repository is available at https://github.com/MLAI-Yonsei/CISE.git.
Chinese Translation
基于大型语言模型（LLM）的自我进化搜索是一种有前景的科学发现方法。然而，在某些领域，对每个候选者进行高保真评估的成本过于昂贵。因此，在这种情况下，自我进化系统依赖于低成本但不完美的代理奖励，这可能会对不可行的候选者赋予高分。这些假阳性可能会污染最终输出和用于指导后续生成的反馈。这促使我们为更可靠的自我进化搜索提出统计校准的奖励区间。我们提出了符合区间驱动的自我进化（Conformal Interval-Driven Self-Evolution, CISE），该方法使用条件符合推断和迭代在线密度比估计构建候选者特定的奖励区间。CISE使用保守的基于区间的奖励作为进化反馈，并仅在所有所需属性区间完全位于各自的可行区域内时返回候选者。我们在明确的独立性和协变量转移假设下推导了固定迭代覆盖结果。我们在材料科学的三个自我进化搜索任务上评估了CISE。在我们的实验中，CISE返回的所有候选者在高保真评估下都是真正的阳性，而基线方法返回更多候选者但包含假阳性。这些结果突显了在下游验证预算有限时，更小且更精确的候选名单的价值。我们的代码库可在 https://github.com/MLAI-Yonsei/CISE.git 获取。
cs.AI / 75 / 2610.02976
Relevant Evidence Decoding for Audio-Visual Hallucination Mitigation
音视频幻觉缓解的相关证据解码
Abstract
Audio-Visual Large Language Models (AV-LLMs) remain prone to cross-modal hallucinations, where one modality incorrectly affects predictions about another. Although contrastive decoding reduces hallucinations in vision-language models, its direct extension to AV-LLMs overlooks a key challenge: different questions require different perceptual evidence, including audio, video, or their interaction. Notably, we observe that joint audio-visual inference can weaken the prediction even when a model can recover the correct answer from a single informative modality. For example, when asked which instrument is heard, a model may correctly predict violin from the audio alone. Once a video showing a guitar is added, its confidence in violin may drop. In this paper, we introduce Relevant Evidence Decoding (RED), a training-free method that identifies question-relevant evidence and selectively strengthens its contribution. RED uses pointwise mutual information to quantify the predictive support provided by audio and video beyond the question alone. It decomposes their joint contribution into audio, video, and residual interaction components. A question-only inference pass determines the required evidence type, after which the model augments the original audio-visual prediction with the corresponding PMI contribution. Across three audio-visual hallucination benchmarks and three AV-LLMs, RED improves accuracy over standard decoding by up to 7% on CMM, 6.3% on AVHBench, and 3.8% on SVHalluc, with an average relative time to first token of 1.5x standard decoding.
Chinese Translation
音视频大型语言模型（AV-LLMs）仍然容易受到跨模态幻觉的影响，即一种模态错误地影响对另一种模态的预测。尽管对比解码在视觉-语言模型中减少了幻觉，但其直接扩展到AV-LLMs时忽略了一个关键挑战：不同的问题需要不同的感知证据，包括音频、视频或它们的交互。值得注意的是，我们观察到联合音视频推理可能会削弱预测，即使模型能够从单一的信息模态中恢复正确答案。例如，当被问及听到哪种乐器时，模型可能仅凭音频就能正确预测小提琴。一旦添加了显示吉他的影片，其对小提琴的信心可能会下降。在本文中，我们提出了相关证据解码（Relevant Evidence Decoding, RED），这是一种无训练的方法，旨在识别与问题相关的证据并选择性地增强其贡献。RED使用逐点互信息（pointwise mutual information）来量化音频和视频在超出问题本身的情况下提供的预测支持。它将它们的联合贡献分解为音频、视频和残余交互成分。仅依赖问题的推理过程确定所需的证据类型，之后模型将原始的音视频预测与相应的PMI贡献进行增强。在三个音视频幻觉基准测试和三个AV-LLMs上，RED在CMM上提高了高达7%的准确率，在AVHBench上提高了6.3%，在SVHalluc上提高了3.8%，且首次生成token的平均相对时间为标准解码的1.5倍。
cs.AI / 76 / 2610.02979
RASPER: Reward-Aligned Summarization of Clinical Notes for EHR Outcome Prediction
RASPER：用于电子健康记录结果预测的奖励对齐临床笔记摘要
Abstract
Unstructured discharge notes in Electronic Health Records (EHRs) often carry signal complementary to structured medical codes, holding patient-specific evidence that standardized cohort-level codes alone cannot capture. However, this evidence in notes is frequently buried in lengthy, noisy text that is not intentionally written with any specific clinical prediction in mind. Summarization is an obvious mitigation, but generic summaries, tuned for fluency rather than the outcome, routinely omit decisive evidence while retaining plausible but uninformative detail. To this end, we propose RASPER, a Reward-Aligned Summarizer for Prediction in EHR, that optimizes note summarization directly against the downstream clinical task. RASPER employs a tunable LLM-based summarizer to extract task-relevant evidence from discharge notes and trains it via reinforcement learning from prediction feedback, using a reward derived from the downstream predictor's loss. To ground the summarizer, a longitudinal encoder converts structured codes into soft prompts that incorporate each patient's clinical context into note summarization. By rewarding the quality of the resulting multimodal prediction, RASPER encourages the summarizer to retain patient-specific evidence that complements, rather than duplicates, information captured by structured codes. RASPER consistently outperforms strong baselines on both readmission prediction and medication recommendation across MIMIC-III and MIMIC-IV.
Chinese Translation
电子健康记录（EHR）中的非结构化出院记录通常包含与结构化医疗编码互补的信号，持有患者特定的证据，而仅靠标准化的队列级编码无法捕捉这些证据。然而，这些证据常常埋藏在冗长且嘈杂的文本中，这些文本并不是为了特定的临床预测而有意编写的。摘要是一种明显的缓解措施，但通用摘要往往侧重于流畅性而非结果，常常遗漏决定性证据，同时保留可信但无信息量的细节。为此，我们提出了RASPER，一种用于EHR预测的奖励对齐摘要器，直接针对下游临床任务优化笔记摘要。RASPER采用可调的基于大型语言模型（LLM）的摘要器，从出院记录中提取与任务相关的证据，并通过从预测反馈中进行强化学习进行训练，使用源自下游预测器损失的奖励。为了为摘要器提供基础，一个纵向编码器将结构化编码转换为软提示，将每个患者的临床背景纳入笔记摘要中。通过奖励最终多模态预测的质量，RASPER鼓励摘要器保留患者特定的证据，这些证据补充而非重复结构化编码所捕获的信息。在MIMIC-III和MIMIC-IV上，RASPER在再入院预测和药物推荐方面始终优于强基线。
cs.AI / 77 / 2610.02981
Safeguarding Mutual Correction in Source-Free Domain Adaptation via Cut Statistics
通过切割统计保障无源领域适应中的相互修正
Abstract
Source-Free Domain Adaptation (SFDA) aims to adapt a source-pretrained model to an unlabeled target domain without access to the original source domain. While early single-model approaches rely on self-refinement, they are inherently susceptible to confirmation bias and struggle to correct their own systematic errors. To overcome this limitation, recent methods introduce Vision-Language (ViL) models as external knowledge sources. However, these approaches operate in a largely unidirectional paradigm, using the ViL model primarily to supervise the source-pretrained model. This overlooks a key structural property: the two models exhibit distinct failure modes -- where one produces an incorrect prediction, the other may produce a correct one, creating a natural opportunity for mutual correction within the target domain. Yet, without ground-truth labels, identifying which model is correct on any given sample is non-trivial, and naively exchanging predictions risks propagating errors across models. To address this challenge, we propose SafeCut, a novel approach that leverages the cut statistic as a label-free measure of prediction reliability to gate cross-model supervision. Our approach dynamically controls both the direction and strength of supervision based on relative reliability, selectively amplifying true corrections while suppressing miscorrections on a per-sample basis. We further provide theoretical justification showing that this reliability-gated mechanism guarantees a net-positive correction signal. Extensive experiments across diverse SFDA benchmarks demonstrate that SafeCut achieves state-of-the-art performance, highlighting the effectiveness of safeguarding mutual correction in SFDA via cut statistics.
Chinese Translation
无源领域适应（Source-Free Domain Adaptation, SFDA）旨在将源预训练模型适应于无标签的目标领域，而无需访问原始源领域。早期的单模型方法依赖于自我精炼，但它们本质上容易受到确认偏差的影响，难以纠正自身的系统性错误。为克服这一局限性，近期的方法引入了视觉-语言（Vision-Language, ViL）模型作为外部知识源。然而，这些方法主要在单向范式下运作，主要利用ViL模型来监督源预训练模型。这忽视了一个关键的结构特性：两个模型表现出不同的失败模式——当一个模型产生错误预测时，另一个模型可能产生正确预测，从而在目标领域内创造了相互修正的自然机会。然而，在没有真实标签的情况下，识别给定样本中哪个模型是正确的并非易事，而简单地交换预测可能会导致错误在模型间传播。为了解决这一挑战，我们提出了SafeCut，这是一种新颖的方法，利用切割统计作为无标签的预测可靠性度量来控制跨模型监督。我们的方法根据相对可靠性动态控制监督的方向和强度，选择性地放大真实修正，同时在每个样本基础上抑制错误修正。我们还提供了理论依据，表明这种可靠性门控机制保证了净正的修正信号。在多样化的SFDA基准上进行的广泛实验表明，SafeCut实现了最先进的性能，突显了通过切割统计保障SFDA中相互修正的有效性。
cs.AI / 78 / 2610.02982
PLCWorld: Benchmarking LLM-Generated PLC Programs in Closed-Loop Plant Simulation
PLCWorld：在闭环工厂仿真中基准测试LLM生成的PLC程序
Abstract
Programmable logic controllers (PLCs) coordinate industrial equipment by reading sensor inputs and issuing control commands. Evaluating whether large language model (LLM)-generated PLC programs satisfy task requirements and safety constraints requires observing how their commands affect device and workpiece states. We introduce PLCWorld, a common closed-loop execution environment and benchmark that couples Structured Text (ST) execution with simulated plant responses and sensor feedback. Grounded in control relations identified in industrial PLC programs and engineering documentation, PLCWorld contains 100 synthetic tasks and 473 registered task-condition pairs across Motion Control and Material Handling, with difficulty defined by control-dependency scope. A common protocol reports Task Success and Safety Violation separately. Validation combines practitioner review, reference and alternative programs, targeted counterexamples, specification-evaluator alignment checks, and comparisons with independent ST runtimes. Reference and alternative programs satisfy their applicable cases, while all 542 targeted counterexamples activate their designated evaluator rules under at least one registered condition. Execution Gap relates submission-profile acceptance to subsequent task failure or observed Safety Violation. Across the constructed task groups, direct GPT-5.5 achieves 82.70% Task Success on Easy cases but 25.10% on Hard cases. Evaluations of six LLMs and four adapted generation-and-verification workflows further expose differences between completion, safety, and generation cost. Our code, simulation environment, benchmark tasks, and baseline implementations are publicly available at https://yunji0516.github.io/PLCWorld/.
Chinese Translation
可编程逻辑控制器（PLC）通过读取传感器输入并发出控制命令来协调工业设备。评估大型语言模型（LLM）生成的PLC程序是否满足任务要求和安全约束需要观察其命令如何影响设备和工件状态。我们介绍了PLCWorld，这是一个通用的闭环执行环境和基准，结合了结构化文本（Structured Text, ST）执行与模拟工厂响应和传感器反馈。PLCWorld基于工业PLC程序和工程文档中识别的控制关系，包含100个合成任务和473对注册的任务-条件对，难度由控制依赖范围定义。一个通用协议分别报告任务成功和安全违规。验证结合了从业者审查、参考和替代程序、针对性反例、规范评估者对齐检查，以及与独立ST运行时的比较。参考和替代程序满足其适用案例，而所有542个针对性反例在至少一个注册条件下激活其指定的评估规则。执行差距（Execution Gap）将提交配置的接受与后续任务失败或观察到的安全违规相关联。在构建的任务组中，直接使用GPT-5.5在简单案例上实现了82.70%的任务成功率，但在困难案例上仅为25.10%。对六个LLM和四个改编的生成与验证工作流的评估进一步揭示了完成、安全性和生成成本之间的差异。我们的代码、仿真环境、基准任务和基线实现已公开发布在 https://yunji0516.github.io/PLCWorld/。
cs.AI / 79 / 2610.03017
Personalized Automatic Speech Recognition for a Dysarthric and Tracheostomic Speaker using Artificial Conversations
针对患有构音障碍和气管造口的说话者的个性化自动语音识别系统，基于人工对话
Abstract
This work presents an automatic speech recognition (ASR) system personalized for a Czech speaker with a permanent tracheal stoma and severe dysarthria rendering their speech unintelligible to untrained listeners. We release a public dataset containing 33 annotated hours of the speaker's speech, collected using a novel "artificial conversation" protocol designed for high engagement and dialogue realism. We propose a multi-stage training pipeline based on Whisper Base: fine-tuning on standard Czech speech, acoustically simulated tracheostomic speech, and the speaker's data. We evaluate the system across three near real-time scenarios: scripted conversations, question answering, and spontaneous dialogue, achieving a 50\% relative reduction in Character Error Rate compared to Whisper Base baseline and surpassing the average recognition accuracy of their assistants in acoustic recognition of isolated utterances. We demonstrate that even for severely impeded speech, a helpful ASR is achievable, as evidenced by the quantitative results and the feedback from the speaker.
Chinese Translation
本研究提出了一种针对一位具有永久气管造口和严重构音障碍的捷克说话者的个性化自动语音识别（ASR）系统，该说话者的语音对未受过训练的听众来说难以理解。我们发布了一个公共数据集，包含33小时的标注语音，数据是通过一种新颖的“人工对话”协议收集的，该协议旨在提高参与度和对话的真实感。我们提出了一种基于Whisper Base的多阶段训练流程：在标准捷克语音、声学模拟的气管造口语音和说话者数据上进行微调。我们在三种近实时场景中评估该系统：脚本对话、问答和自发对话，相较于Whisper Base基线，字符错误率（Character Error Rate）实现了50%的相对降低，并且超过了他们在孤立语句声学识别中的平均识别准确率。我们证明，即使对于严重受限的语音，仍然可以实现有效的ASR，这一点通过定量结果和说话者的反馈得到了验证。
cs.AI / 80 / 2610.03020
DyadMem: A Long-Term Memory Benchmark of How Agents Work with Users
DyadMem：代理与用户协作的长期记忆基准
Abstract
Long-term agents must remember not only what is true about a user, but also how a particular agent should work with that user as their shared history evolves. Existing benchmarks primarily supervise user facts and preferences or experience reusable across users, leaving this relationship-specific agent memory implicit. Additionally, most prior works measure the model solely with final-answer QA over long interaction histories, making the assessment still incomplete and unreliable. To this end, we introduce DyadMem with the proposed new definition User-conditioned Relational Agent Memory (URAM). DyadMem jointly annotates user-side memory and URAM along the same multi-session trajectories, resulting in 6 memory categories. To summarize, it includes 3,065 episodes, 50,961 sessions, and 61,210 QA instances, with extensive session-level Capture and Update gold annotations, query-level Recall support, and two QA settings: Gold-Memory and Full-Pipeline. Across 16 open-weight and 4 proprietary models, Gold-Memory QA is consistently strong, yet Full-Pipeline QA drops sharply. Such a gap explicitly supports our fine-grained evaluation design. Additionally, several quantitative results further reveal low Capture recall, incomplete Recall, and unsafe-deletion issues arising from even the frontier LLMs. We further conduct a rigorous experiment to validate the effectiveness of our URAM and observe the positive effects for all 20 models. In summary, DyadMem is a dual-domain, full-pipeline memory benchmark with extensive annotation efforts for advancing the domain's development.
Chinese Translation
长期代理不仅必须记住用户的真实信息，还必须了解在共享历史演变过程中，特定代理应如何与该用户协作。现有基准主要监督用户的事实和偏好或可在用户之间重用的经验，导致这种关系特定的代理记忆隐含。此外，大多数先前的研究仅通过长交互历史中的最终答案问答（QA）来评估模型，使得评估仍然不完整且不可靠。为此，我们引入了DyadMem，并提出了用户条件关系代理记忆（User-conditioned Relational Agent Memory, URAM）的新定义。DyadMem在相同的多会话轨迹上共同注释用户侧记忆和URAM，形成6个记忆类别。总的来说，它包括3,065个事件、50,961个会话和61,210个QA实例，具有广泛的会话级捕获和更新金标准注释、查询级召回支持，以及两种QA设置：金标准记忆（Gold-Memory）和全流程（Full-Pipeline）。在16个开放权重模型和4个专有模型中，金标准记忆QA表现始终强劲，而全流程QA则急剧下降。这种差距明确支持了我们细致的评估设计。此外，一些定量结果进一步揭示了即使是最前沿的大型语言模型（LLMs）也存在低捕获召回、不完整召回和不安全删除等问题。我们进一步进行了一项严格的实验，以验证URAM的有效性，并观察到对所有20个模型的积极影响。总之，DyadMem是一个双领域、全流程的记忆基准，具有广泛的注释工作，以推动该领域的发展。
cs.AI / 81 / 2610.03025
Verifiable, Articulable, and Tacit Components of Preference
可验证、可表述和隐性偏好的组成部分
Abstract
What makes a short story gripping; a news article newsworthy; or a math proof elegant? These constructs resist articulation or verification; their meaning is at least partially tacit. However, modern AI models are improved primarily via articulated constitutions, rubrics and verifiers (i.e. in RLAIF and RLVR); tacit components of preferences are typically understudied. We introduce a large, labeled preference dataset CreativePreferences, containing 2.8M texts labeled by 317M human preference judgments across 7 creative domains, with 42 benchmark tasks. We model these labels with executable programs, rubric banks and densely trained models (V, A and VAT, respectively). We observe robust articulability gaps, VAT-VA; and verifiability gaps, VAT-V; we estimate upper and lower bounds for each gap with a novel measurement approach that discovers articulable and verifiable metrics, identifies spurious variables and estimates the value of undiscovered metrics using capture-recapture. These gaps occur across all domains, even in domains traditionally treated as fully verifiable: correctness-centered domains (i.e. mathematics and software engineering) and claim- and novelty-centric domains (i.e. news, patents, peer review). The size of the gap varies based on domain (e.g. peer review and creative writing have the largest articulability gaps) and widens as more people take part in the judgment, consistent with Collins' collective tacit knowledge. We show two consequences: (1) on human generations, the full model more closely matches human preferences, often in disagreement with articulated criteria, and (2) in an analogy to Goodhart's law, articulating preference shifts it away from the tacit dimension. Articulability and verifiability gaps are consequential; we give recommendations on when tasks can be prompted; how learning mechanisms might improve; and when to leave judgments with humans.
Chinese Translation
是什么让短篇小说引人入胜；新闻文章具有新闻价值；或者数学证明优雅？这些构念难以表述或验证；其意义至少部分是隐性的。然而，现代人工智能模型主要通过可表述的构成、评分标准和验证器（即在 RLAIF 和 RLVR 中）得到改进；偏好的隐性组成部分通常被忽视。我们引入了一个大型标注偏好数据集 CreativePreferences，包含 280 万条文本，由 3.17 亿个人类偏好判断标注，涵盖 7 个创意领域，并设有 42 个基准任务。我们使用可执行程序、评分标准库和密集训练的模型（分别为 V、A 和 VAT）对这些标签进行建模。我们观察到显著的可表述性差距（VAT-VA）和可验证性差距（VAT-V）；我们采用一种新颖的测量方法来估算每个差距的上限和下限，发现可表述和可验证的指标，识别虚假变量，并使用捕获-重捕法估算未发现指标的价值。这些差距在所有领域中普遍存在，即使在传统上被视为完全可验证的领域：以正确性为中心的领域（即数学和软件工程）以及以主张和新颖性为中心的领域（即新闻、专利、同行评审）。差距的大小因领域而异（例如，同行评审和创意写作的可表述性差距最大），并随着更多人参与判断而扩大，这与 Collins 的集体隐性知识一致。我们展示了两个后果：（1）在人类生成方面，完整模型更接近人类偏好，往往与可表述标准不一致；（2）类比于古德哈特法则，表述偏好使其偏离隐性维度。可表述性和可验证性差距具有重要意义；我们提供了关于何时可以提示任务、学习机制如何改进以及何时应将判断留给人类的建议。
cs.AI / 82 / 2610.03029
SoftGene: Protein Language Model-Enhanced Soft Prompting for Interpretable Gene Set Annotation
SoftGene：基于蛋白质语言模型增强的软提示用于可解释的基因集注释
Abstract
Gene set analysis is a cornerstone of functional genomics, yet it remains labor-intensive and heavily dependent on manual curation and expert biological interpretation. While Large Language Models (LLMs) have emerged as powerful tools for genomic reasoning and annotation, most existing approaches rely on symbolic gene names and fail to capture domain-specific biological structure, particularly protein sequence information that governs molecular activity, interactions, and downstream gene function. In this work, we propose SoftGene, a novel framework for LLM-based gene set annotation that leverages the hierarchical structure of gene sets. First, we use a hierarchical attention-based encoder built on ESM, a protein language model, to represent each gene set using protein-level amino acid sequence information. Second, we construct a hybrid prompting scheme that combines soft prompts derived from gene set embeddings with hard prompts containing auxiliary context generated by an LLM, and feed the resulting prompt into a local LLM for annotation. We evaluate our framework on two benchmark datasets: Gene Ontology (GO) and the Molecular Signatures Database (MSigDB). Our results show that integrating protein-sequence representations with textual context improves gene set annotation overall, while per-domain analyses reveal that the contribution of protein embeddings varies across biological domains.
Chinese Translation
基因集分析是功能基因组学的基石，但仍然劳动密集且高度依赖手动整理和专家生物学解释。尽管大型语言模型（LLMs）已成为基因组推理和注释的强大工具，但现有的大多数方法依赖于符号基因名称，未能捕捉特定领域的生物结构，尤其是决定分子活性、相互作用和下游基因功能的蛋白质序列信息。在本研究中，我们提出了SoftGene，一个基于LLM的基因集注释新框架，利用基因集的层次结构。首先，我们使用基于层次注意力的编码器，该编码器建立在蛋白质语言模型ESM之上，利用蛋白质级氨基酸序列信息表示每个基因集。其次，我们构建了一种混合提示方案，将从基因集嵌入中派生的软提示与包含LLM生成的辅助上下文的硬提示相结合，并将生成的提示输入本地LLM进行注释。我们在两个基准数据集上评估了我们的框架：基因本体（GO）和分子特征数据库（MSigDB）。我们的结果表明，将蛋白质序列表示与文本上下文相结合，整体上改善了基因集注释，而每个领域的分析显示，蛋白质嵌入的贡献在不同生物领域中有所不同。
cs.AI / 83 / 2610.03033
When Numbers Start Talking: Numerical Signalling and Strategic Behaviour Among LLMs
当数字开始发声：大型语言模型中的数值信号与战略行为
Abstract
Large language model (LLM)-based agents increasingly operate in multi-agent systems (MAS) characterised by strategic interaction. However, little is known about whether, and to what extent, different types of messages affect the outcomes of strategic games. By investigating AI agents based on four popular LLMs, playing four games with different cooperation equilibria, we study whether messages of different kinds (natural language, numerical signals, or random sequences) significantly modify the levels of cooperation in each game, also depending on the agents' assigned personalities. We observe that structured messages alter the final payoffs for most games and LLMs, but without a predictable pattern; this challenges the assumption that AI agents can converge to stable equilibria regardless of additional capabilities. Moreover, we observe that agent-generated numerical messages depart from randomness, most strongly and consistently when agents are explicitly instructed to communicate; however, they introduce an additional interpretability challenge, as their symbol distributions are mostly associated with the payoff structure and typically become more concentrated with repetition, but are overall difficult for humans to interpret. Monitoring for coordination of AI agents through restricted channels should thus prioritise message-level fingerprints, which generalise across models, over behavioural decisions, which do not.
Chinese Translation
基于大型语言模型（LLM）的代理在多智能体系统（MAS）中越来越多地进行战略互动。然而，关于不同类型的信息在多大程度上影响战略游戏结果的研究仍然较少。通过研究基于四种流行LLM的人工智能代理在四种具有不同合作均衡的游戏中的表现，我们探讨了不同类型的信息（自然语言、数值信号或随机序列）是否显著改变每个游戏中的合作水平，这也取决于代理的个性特征。我们观察到，结构化信息在大多数游戏和LLM中改变了最终收益，但没有可预测的模式；这挑战了人工智能代理能够在不考虑额外能力的情况下收敛到稳定均衡的假设。此外，我们观察到代理生成的数值信息偏离随机性，尤其是在代理被明确指示进行沟通时表现得最为强烈和一致；然而，这也引入了额外的可解释性挑战，因为它们的符号分布大多与收益结构相关，并且通常随着重复而变得更加集中，但总体上人类难以解读。因此，通过限制渠道监控人工智能代理的协调应优先考虑跨模型的消息级指纹，而不是行为决策。
cs.AI / 84 / 2610.03055
hacktrace: behavior-supervised detection of reward hacking during code generation
hacktrace：在代码生成过程中行为监督的奖励黑客检测
Abstract
A coding agent can earn a passing grade by fixing its code, or by deleting the test that exposes the bug. Detecting such reward hacking requires recognizing attempted shortcuts, including those that fail. We release 173,561 annotated multi-turn coding trajectories from Qwen3-8B and show that supervising shortcut behavior independently of exploit success substantially improves detection. We introduce HACKTRACE, a behavior-supervised monitor that reads the internal states the agent already computes while generating code. Reusing these states enables monitoring before a turn is complete, without additional language-model tokens or passes. Combining this evidence with static features of the final files achieves a mean per-problem AUC of 0.997 with 8 ms of monitoring overhead, improving both accuracy and latency over monitors that run the model again on an honesty question and answer. The same generation states also provide an inexpensive monitoring signal for reinforcement learning. With strong GRPO penalties, HACKTRACE reduces the cheating share of passing solutions from 82-91% to 1-5%, while retaining honest, correct solutions and maintaining high detection accuracy as the policy evolves. Our results show that both the supervision target and the source of monitoring evidence matter for turning accurate detection into a useful training signal.
Chinese Translation
编码代理可以通过修复其代码或删除暴露错误的测试来获得及格分数。检测这种奖励黑客行为需要识别尝试的捷径，包括那些失败的捷径。我们发布了来自Qwen3-8B的173,561条注释的多轮编码轨迹，并展示了在不依赖于利用成功的情况下监督捷径行为显著提高了检测效果。我们引入了HACKTRACE，这是一种行为监督的监控工具，可以读取代理在生成代码时已经计算的内部状态。重用这些状态使得在回合完成之前进行监控成为可能，无需额外的语言模型令牌或传递。将这些证据与最终文件的静态特征结合，达到了每个问题平均AUC为0.997，监控开销为8毫秒，提升了准确性和延迟，相较于在诚实问题和答案上重新运行模型的监控器。相同的生成状态还为强化学习提供了一种廉价的监控信号。在强GRPO惩罚下，HACKTRACE将及格解决方案中的作弊比例从82-91%降低到1-5%，同时保留诚实且正确的解决方案，并在策略演变过程中保持高检测准确性。我们的结果表明，监督目标和监控证据的来源对于将准确检测转化为有用的训练信号至关重要。
cs.AI / 85 / 2610.03056
MOF-VERIFY: A Failure-Aware Agentic Harness for MOF Hypothesis Verification
MOF-VERIFY：一种关注失败的代理工具用于金属有机框架假设验证
Abstract
Large language models are increasingly used as reasoning components in AI-driven materials Co-Scientists, yet the reliability of the resulting verification pipeline remains unclear. Metal-organic frameworks (MOFs) provide a particularly challenging setting because structures may appear under different identifiers, synthesis outcomes depend strongly on experimental conditions, evidence is distributed across heterogeneous sources, and some hypotheses require computation rather than literature alone. We introduce a diagnostic benchmark with four task families covering structural grounding, synthesis-condition verification, evidence-sufficiency verification, and MLIP-based computational verification. T-MOF-1-3 are evaluated under closed-book, retrieval-enabled, and oracle-evidence settings to localize failures in knowledge access, evidence acquisition, and reasoning, while T-MOF-4 separately evaluates computational verification. Guided by these diagnosed failure modes, we develop MOF-Verify, a failure-aware agentic harness that targets structural, literature, evidence-sufficiency, and computational bottlenecks before producing a final verdict. Across multiple backbone LLMs, MOF-Verify substantially improves hypothesis-verification performance over direct inference and retrieval-based baselines. Benchmark datasets are released at https://github.com/IMMS-Ewha/MOF-Verify-Benchmark.
Chinese Translation
大型语言模型越来越多地被用作AI驱动的材料共同科学家中的推理组件，但由此产生的验证流程的可靠性仍不明确。金属有机框架（MOFs）提供了一个特别具有挑战性的环境，因为结构可能以不同的标识符出现，合成结果在很大程度上依赖于实验条件，证据分布在异构来源中，并且某些假设需要计算而不仅仅依赖文献。我们引入了一个诊断基准，涵盖四个任务类别，包括结构基础、合成条件验证、证据充分性验证和基于MLIP的计算验证。在闭卷、检索启用和oracle证据设置下评估T-MOF-1-3，以定位知识访问、证据获取和推理中的失败，而T-MOF-4则单独评估计算验证。在这些诊断出的失败模式的指导下，我们开发了MOF-Verify，这是一种关注失败的代理工具，旨在在产生最终判决之前解决结构、文献、证据充分性和计算瓶颈。在多个基础LLM上，MOF-Verify显著提高了假设验证的性能，相较于直接推理和基于检索的基线。基准数据集已发布在https://github.com/IMMS-Ewha/MOF-Verify-Benchmark。
cs.AI / 86 / 2610.03079
RIFAR: Reliability and Forgetting-Aware Replay for Continual Robot Learning
RIFAR：面向持续机器人学习的可靠性与遗忘感知重放
Abstract
Genuine embodied agency requires robots to turn continuous real-world experience into lasting, transferable skills. This demands continual learning that integrates new capabilities without eroding prior knowledge as tasks and environments evolve. Experience replay mitigates forgetting, but storing complete demonstrations becomes costly as tasks accumulate. World-action models offer a generative alternative, reconstructing past experience through joint predictions of actions and future observations. However, visually coherent rollouts may contain actions that cannot realize the predicted transitions, while new-task adaptation can disrupt previously learned behavior. RIFAR therefore combines reliability screening with drift-aware replay selection. It reconstructs trajectories from compact demonstration prefixes and uses a frozen inverse-dynamics model to assess action-visual consistency. Training first combines current demonstrations with the highest-quality screened trajectories. RIFAR then compares action predictions before and after this adaptation on identical historical inputs, reselecting trajectories with larger normalized drift from the same screened pool for continued training. Across three LIBERO suites and real-world experiments, RIFAR surpasses the previous state of the art in WAM-based generative replay. On LIBERO-Goal, it achieves 90.97 AUC while retaining only 320 historical time steps per task, approximately 4.9% of the steps retained using 50-demonstration replay.
Chinese Translation
真正的具身代理要求机器人将持续的现实世界经验转化为持久的、可转移的技能。这需要持续学习，能够在任务和环境演变的过程中整合新能力而不侵蚀先前的知识。经验重放可以减轻遗忘，但随着任务的积累，存储完整的演示变得成本高昂。世界-动作模型提供了一种生成性替代方案，通过对动作和未来观察的联合预测重构过去的经验。然而，视觉上连贯的展开可能包含无法实现预测转变的动作，而新任务的适应可能会干扰先前学习的行为。因此，RIFAR结合了可靠性筛选与漂移感知重放选择。它从紧凑的演示前缀中重构轨迹，并使用冻结的逆动力学模型评估动作与视觉的一致性。训练首先将当前演示与最高质量的筛选轨迹结合。然后，RIFAR在相同的历史输入上比较适应前后的动作预测，从同一筛选池中重新选择具有较大标准化漂移的轨迹以进行持续训练。在三个LIBERO套件和实际实验中，RIFAR在基于WAM的生成重放中超越了之前的最先进水平。在LIBERO-Goal上，它在仅保留每个任务320个历史时间步的情况下，达到了90.97的AUC，约为使用50个演示重放保留的步骤的4.9%。
cs.AI / 87 / 2610.03095
Peer Influence across Heterogeneous AI Models
异构人工智能模型中的同行影响
Abstract
When two AI agents disagree, who persuades whom? As multi-agent systems increasingly combine language models of different families and sizes, the answer can determine which judgments survive interaction. Measuring persuasion as the probabilistic shift in an agent's decision after a single exchange with a dissenting peer, we test seven open-weight models across three language understanding tasks. We find that persuasion is strong: when models disagree, receivers often abandon their initial judgment after seeing a peer's answer and explanation. Surprisingly, however, neither standalone certainty nor model scale reliably predicts persuasion dynamics. Models producing almost perfectly consistent decisions in isolation can be among the most susceptible to persuasion, and small models can match larger ones as persuaders and resist their influence just as effectively. Furthermore, we show that the size of the shift depends more on the susceptibility of the listener than on the persuasiveness of the speaker. Persuasion patterns are therefore specific to each model pairing, with heterogeneity amplifying persuasion in some combinations and suppressing it in others, allowing a dissenting agent running a small model to overturn the judgments of a much larger one. These findings show that the behavior of interacting models cannot be inferred from their individual properties but must be evaluated in the combinations in which they will operate.
Chinese Translation
当两个人工智能代理意见不一致时，谁说服谁？随着多代理系统越来越多地结合不同家族和规模的语言模型，这一问题的答案可以决定哪些判断在交互中得以保留。我们将说服力测量为代理在与持不同意见的同行进行单次交流后决策的概率性变化，测试了七种开放权重模型在三个语言理解任务中的表现。我们发现说服力很强：当模型意见不一致时，接收者在看到同行的答案和解释后，往往会放弃他们最初的判断。然而，令人惊讶的是，单独的确定性或模型规模并不能可靠地预测说服动态。在孤立状态下几乎做出完全一致决策的模型，可能是最容易受到说服的，而小模型在作为说服者时可以与大模型相匹敌，并同样有效地抵抗其影响。此外，我们还表明，变化的大小更多地依赖于听众的易受影响性，而非说话者的说服力。因此，说服模式对于每对模型都是特定的，异质性在某些组合中增强了说服力，而在其他组合中则抑制了说服力，使得使用小模型的持不同意见的代理能够推翻更大模型的判断。这些发现表明，交互模型的行为不能仅从其个体属性推断，而必须在它们将要操作的组合中进行评估。
cs.AI / 88 / 2610.03098
Predictor-Guided Latent Space Codon Optimization for Maximizing Protein Expression
基于预测器引导的潜在空间密码子优化以最大化蛋白质表达
Abstract
Codon optimization, the process of selecting synonymous codons to improve mRNA translation efficiency and protein expression, is central to therapeutic protein production and mRNA vaccines, yet it remains a hard problem. The design space is discrete and combinatorially large, precluding gradient-based methods, and existing tools rely on heuristic proxies (e.g., Codon Adaptation Index or GC-content) that poorly capture true expression. We introduce Latent-Space Codon Optimization (LSCO), which recasts this discrete problem as a continuous one by mapping sequences into the latent space of a pretrained mRNA language model, enabling efficient gradient-based search. LSCO combines four components: a data-driven expression objective from an uncertainty-aware predictor, a Minimum-Free-Energy regularizer for structural stability, a naturalness prior from a protein-to-codon back-translation model, and constrained decoding for protein fidelity. On a real-world, wet-lab antibody expression dataset, LSCO outperforms simple frequency-based, as well as modern deep generative baselines in predicted expression, while retaining suitable biophysical properties.
Chinese Translation
密码子优化是选择同义密码子以提高mRNA翻译效率和蛋白质表达的过程，这在治疗性蛋白质生产和mRNA疫苗中至关重要，但仍然是一个难题。设计空间是离散的且组合性庞大，无法使用基于梯度的方法，而现有工具依赖于启发式代理（例如，密码子适应指数（Codon Adaptation Index）或GC含量），这些方法无法准确捕捉真实的表达情况。我们提出了潜在空间密码子优化（Latent-Space Codon Optimization, LSCO），通过将序列映射到预训练mRNA语言模型的潜在空间，将这一离散问题重新表述为连续问题，从而实现高效的基于梯度的搜索。LSCO结合了四个组件：来自不确定性感知预测器的数据驱动表达目标、用于结构稳定性的最小自由能正则化器、来自蛋白质到密码子的反向翻译模型的自然性先验，以及用于蛋白质保真度的约束解码。在一个真实的湿实验室抗体表达数据集中，LSCO在预测表达方面优于简单的基于频率的方法以及现代深度生成基线，同时保持适当的生物物理特性。
cs.AI / 89 / 2610.03128
Trading Strategy Optimization via Textual Gradient
通过文本梯度优化交易策略
Abstract
Quantitative trading strategy design aims to discover trading programs from historical data that remain effective in future markets, which can be viewed as a black-box program optimization problem. LLM-based textual gradients offer a promising approach by providing explicit optimization directions for iterative strategy refinement. However, directly applying textual gradients faces two challenges: (1) optimization is myopic, underutilizing experience from previous evaluations; and (2) aggregate backtest feedback overlooks temporal robustness, potentially favoring strategies that perform well only in specific market periods. To address these challenges, we propose TradeGrad, an experience-guided textual-gradient framework for robust trading strategy optimization. TradeGrad leverages accumulated optimization experience to estimate textual gradients and employs multi-scale revisions for both strategy exploration and refinement. It further introduces the Cross-Period Robust Objective (CPRO), which emphasizes performance in unfavorable historical periods to promote temporal robustness. Experiments on cross-sectional and time-series strategy design in Chinese A-share and U.S. equity markets show that TradeGrad achieves the best in-sample and out-of-sample performance across all four settings. Notably, its Chinese cross-sectional strategy achieves 27.99% annualized return, 12.19% maximum drawdown, and a Sharpe ratio of 1.63, approximately 68% higher than the CSI 300 benchmark. Further analyses validate the proposed components and show consistent improvements in both in-sample and out-of-sample performance throughout optimization. The code is available at https://github.com/transcend-0/TradeGrad.
Chinese Translation
量化交易策略设计旨在从历史数据中发现能够在未来市场中保持有效的交易程序，这可以视为一个黑箱程序优化问题。基于大语言模型（LLM）的文本梯度提供了一种有前景的方法，通过为迭代策略优化提供明确的优化方向。然而，直接应用文本梯度面临两个挑战：（1）优化是短视的，未充分利用先前评估的经验；（2）汇总回测反馈忽视了时间上的稳健性，可能偏向于在特定市场时期表现良好的策略。为了解决这些挑战，我们提出了TradeGrad，一个经验指导的文本梯度框架，用于稳健的交易策略优化。TradeGrad利用积累的优化经验来估计文本梯度，并采用多尺度修正进行策略探索和优化。它进一步引入了跨周期稳健目标（Cross-Period Robust Objective, CPRO），强调在不利历史时期的表现，以促进时间上的稳健性。在中国A股和美国股票市场的横截面和时间序列策略设计实验中，TradeGrad在所有四个设置中实现了最佳的样本内和样本外表现。值得注意的是，其中国横截面策略实现了27.99%的年化收益率，12.19%的最大回撤，以及1.63的夏普比率，约比CSI 300基准高出68%。进一步的分析验证了所提出的组件，并显示在整个优化过程中样本内和样本外表现均有一致的改善。代码可在https://github.com/transcend-0/TradeGrad获取。
cs.AI / 90 / 2610.03137
Keeping JEPA World Models Plannable When Little of the Frame Moves
在帧移动极少时保持JEPA世界模型的可规划性
Abstract
Specifying a goal in language rather than as a goal frame is a natural interface for planning with a latent world model, but testing it needs scenes in which language must discriminate between several objects. We build SLIM, a pushing benchmark with several small objects and paired visual and language goals on identical scenes. On SLIM a LeWM world model that solves PushT succeeds on under 1% of trials, although a scripted controller with simulator state solves every tier. Probes locate the failure in the encoder: its latent is nearly action-insensitive, neither pusher nor object positions can be decoded from it, and rollouts are no better than copying the current latent forward. One inverse-dynamics auxiliary loss, applied to encoder latents and to predicted latents through a shared head discarded at test time, restores every probe and raises success from 0.003 to 0.35 (0.16 on the hard pushing tier, where a goal-agnostic policy scores zero), and improves PushT at twice the trained horizon. Controls attribute the repair to the gradient into the encoder, and a response sweep shows that the vanilla model plans once enough of the frame responds to actions. A cheap action-sensitivity probe, computable without environment access, acts as an empirical necessary condition: all configurations below its threshold failed to plan. On the repaired latent, a small language-goal head plans from sentences without retraining the world model: it reaches 0.84 on navigation (visual-goal oracle 1.00), follows the named zone when it is swapped with a decoy, and degrades gracefully to unseen nouns. A single goal sentence rarely completes a push, but given the push as a sequence of stage sentences the head raises success on the medium and hard pushing tiers from 0.04 to 0.25, on par with the goal-frame oracle, also when the switch between stages is read from the latent alone.
Chinese Translation
用语言而非目标框架来指定目标是一种自然的潜在世界模型规划接口，但测试它需要在语言必须区分多个物体的场景中进行。我们构建了SLIM，一个推理基准，包含多个小物体以及在相同场景中配对的视觉和语言目标。在SLIM上，解决PushT的LeWM世界模型在不到1%的试验中成功，尽管一个带有模拟器状态的脚本控制器在每个层级上都能解决问题。探针定位了编码器的失败：其潜在状态几乎对动作不敏感，推手和物体位置无法从中解码，且回滚的效果不比简单地将当前潜在状态向前复制更好。一个逆动力学辅助损失，应用于编码器潜在状态和通过共享头部预测的潜在状态（在测试时丢弃），恢复了每个探针并将成功率从0.003提高到0.35（在困难推理层级上为0.16，此时一个与目标无关的策略得分为零），并在训练的视野下将PushT的表现提升至两倍。控制实验将修复归因于对编码器的梯度，响应扫描显示，当足够的帧对动作做出反应时，普通模型能够进行规划。一个廉价的动作敏感性探针，无需环境访问即可计算，作为一个经验必要条件：所有低于其阈值的配置都未能进行规划。在修复后的潜在状态上，一个小型语言目标头能够在不重新训练世界模型的情况下从句子中进行规划：在导航任务中达到0.84（视觉目标oracle为1.00），在与诱饵交换时能够跟随命名区域，并且在面对未见名词时表现出良好的降级。单个目标句子很少完成推理，但当将推理视为一系列阶段句子时，该头部将中等和困难推理层级的成功率从0.04提高到0.25，达到了目标框架oracle的水平，即使在阶段之间的切换仅从潜在状态中读取时也是如此。
cs.AI / 91 / 2610.03185
Gains and Collapse in On-Policy Distillation:A Reinforcement Learning Perspective
基于策略蒸馏的收益与崩溃：强化学习视角
Abstract
On-policy distillation (OPD) has become an important approach to language model post-training. However, despite its performance gains, OPD can also collapse into excessively long and repetitive generation, and the mechanism underlying these divergent outcomes remains poorly understood. We explain these outcomes through a reinforcement learning perspective: the teacher implicitly rewards student behaviors, even those it rarely exhibits itself. From this perspective, our experiments show that OPD improves performance without expanding the student's capabilities. When the implicit reward model is reliable, OPD makes correct responses easier to sample. In contrast, when the preference misaligns with quality, reward hacking happens: the implicit reward model amplifies overlong, repetitive student rollouts, even though it rarely generates such text itself. Guided by this diagnosis, we find that masking unhealthy responses during training and using SFT initialization can each effectively mitigate the collapse. Together, these findings show that OPD amplifies student behaviors favored by the teacher's implicit feedback, shifting the focus from how well the teacher generates to how reliably it evaluates student rollouts. Our code is available at https://github.com/HancCui/opd_hacking.
Chinese Translation
基于策略蒸馏（On-policy distillation, OPD）已成为语言模型后训练的重要方法。然而，尽管其性能有所提升，OPD也可能导致生成内容过长和重复的崩溃现象，而这些不同结果背后的机制仍然不甚明了。我们通过强化学习的视角解释了这些结果：教师隐性地奖励学生的行为，即使这些行为是其自身很少表现出来的。从这个角度来看，我们的实验表明，OPD在不扩展学生能力的情况下提高了性能。当隐性奖励模型可靠时，OPD使得正确响应的采样变得更容易。相反，当偏好与质量不一致时，奖励黑客行为发生：隐性奖励模型放大了过长和重复的学生生成内容，尽管教师自身很少生成这样的文本。在这一诊断的指导下，我们发现，在训练过程中屏蔽不健康的响应和使用SFT初始化都能有效缓解崩溃现象。这些发现共同表明，OPD放大了教师隐性反馈所偏好的学生行为，将关注点从教师生成的质量转向教师评估学生生成内容的可靠性。我们的代码可在 https://github.com/HancCui/opd_hacking 获取。
cs.AI / 92 / 2610.03198
KV$^2$: A Self-Refining KV Cache
KV$^2$: 一种自我优化的键值缓存
Abstract
The memory footprint of the key-value (KV) cache constrains the practical use of long-context models, and it dominates cost when one prefilled context must later serve many different queries. In this reusable setting, query-agnostic compression trades cost against quality: lightweight estimators are cheap but less accurate, whereas full-context reconstruction scoring is more accurate yet reprocesses the entire prompt. We introduce KV$^2$, a query-agnostic KV-cache compression method based on selective reconstruction. KV$^2$ first uses a lightweight proxy scorer to identify informative in-context tokens, then reprocesses only this subset to compute final eviction scores. On RULER, Needle-in-a-Haystack, and LongBench, KV$^2$'s margin over baselines widens as the budget tightens: on RULER 16K at a 2% KV-cache budget it improves the average score over the next-best baseline by more than 40 percentage points, and on LongBench it attains the highest average across 2%-10% budgets at lower compression-stage runtime and peak memory than full-context reconstruction. Reusable KV-cache compression thus does not require reprocessing the full context. Our code is available at https://anonymous.4open.science/r/KVsquared-0B97.
Chinese Translation
键值（KV）缓存的内存占用限制了长上下文模型的实际应用，并且在一个预填充的上下文必须为许多不同查询服务时，它主导了成本。在这种可重用的设置中，查询无关的压缩在成本与质量之间进行权衡：轻量级估计器便宜但准确性较低，而全上下文重构评分更为准确，但需要重新处理整个提示。我们提出了 KV$^2$，一种基于选择性重构的查询无关的 KV 缓存压缩方法。KV$^2$ 首先使用轻量级代理评分器来识别有信息的上下文标记，然后仅重新处理这一子集以计算最终的驱逐评分。在 RULER、Needle-in-a-Haystack 和 LongBench 上，KV$^2$ 相对于基线的优势随着预算的收紧而扩大：在 RULER 16K 的 2% KV 缓存预算下，它的平均得分比下一个最佳基线提高了超过 40 个百分点，而在 LongBench 上，它在 2%-10% 预算范围内达到了最高的平均得分，并且在压缩阶段的运行时间和峰值内存方面低于全上下文重构。因此，可重用的 KV 缓存压缩不需要重新处理完整上下文。我们的代码可在 https://anonymous.4open.science/r/KVsquared-0B97 获取。
cs.AI / 93 / 2610.03213
Toward SLM-based agentic task-tool intent matching
面向基于SLM的代理任务工具意图匹配
Abstract
Tool-equipped AI agents use tool calls to access data and act on external systems. Horizontal growth of agentic systems increases the number of these interactions, and further motivates the need for automated, per-call oversight that can operate at low latency and/or on-prem. Conventional authorization schemes can determine whether an agent is allowed to invoke a tool, but cannot assess the agent's underlying cognition, specifically, whether the tool selection represents a logical, relevant step toward satisfying the intent of the task or not. Consequently, an allowed call may still deviate from the task's intent: a rogue agent might deviate the calls or nudge other agents to make a combination of calls that would not align with the intent of the task. Therefore, every call needs to be verified. In this study we investigate the applicability of Small Language Models (SLMs) to this purpose: an SLM functions as a task-tool relevance classifier that evaluates every selected tool independently against the assigned task and returns a relevance signal for downstream enforcement. Equipped with a novel dataset with multi-tool tasks whose required tools span distinct Model Context Protocol (MCP) servers, we used prompt-optimization, supervised fine-tuning, and reinforcement learning through GRPO to optimize and specialize SLMs.
Chinese Translation
配备工具的人工智能代理通过工具调用访问数据并对外部系统进行操作。代理系统的横向增长增加了这些交互的数量，进一步推动了对能够在低延迟和/或本地环境下运行的自动化逐次监督的需求。传统的授权方案可以确定代理是否被允许调用工具，但无法评估代理的潜在认知，特别是工具选择是否代表了满足任务意图的逻辑相关步骤。因此，允许的调用仍可能偏离任务的意图：不当代理可能会偏离调用或引导其他代理进行一系列不符合任务意图的调用。因此，每个调用都需要进行验证。在本研究中，我们探讨了小型语言模型（Small Language Models, SLMs）在此目的上的适用性：SLM作为任务-工具相关性分类器，独立评估每个选择的工具与指定任务的相关性，并为下游执行返回相关性信号。我们使用了一个包含多工具任务的新数据集，这些任务所需的工具跨越不同的模型上下文协议（Model Context Protocol, MCP）服务器，采用了提示优化、监督微调和通过GRPO的强化学习来优化和专业化SLM。
cs.AI / 94 / 2610.03251
Learning a Fact Is Not Learning How to Retrieve It
学习一个事实并不等于学习如何检索它
Abstract
A model trained on "The capital of X is Y" may produce "Y" after "The capital of X is" but fail after "The capital of X:". We call these different ways of eliciting the same fact request forms. To separate learning a fact from retrieving it, we train two models in two stages. In the first stage (request-form training), one model sees each fact in five forms and the other sees the same facts only as statements. In the second stage (target-fact training), both receive identical training on new facts, all as statements. Both then retrieve the new facts almost equally well from statements, but differ sharply on other request forms. Thus, a model can learn how to retrieve through a request form before it learns the facts. To understand this difference, we examine the hidden state immediately before the answer, which we call the context state. When given two different request forms for the same fact, the model trained on five forms in stage one produces more similar context states than the model trained on statements alone in that stage. Changing this state at retrieval time can enable or prevent retrieval of an already learned fact, and the same effect transfers across facts and factual relations, such as capitals and currencies. To test its role during learning, we change the context state only during target-fact training. This intervention changes later retrieval without intervention at test time. Together, these results show that later retrieval depends on earlier request-form experience and the context state during fact learning.
Chinese Translation
一个在"X的首都是Y"上训练的模型可能在"X的首都是"之后产生"Y"，但在"X的首都:"之后却失败。我们将这些不同的引出同一事实的方式称为请求形式。为了将学习一个事实与检索它分开，我们在两个阶段训练了两个模型。在第一阶段（请求形式训练）中，一个模型以五种形式看到每个事实，而另一个模型仅将相同的事实视为陈述。在第二阶段（目标事实训练）中，两者对新事实进行相同的训练，所有事实均为陈述。然后，两者几乎同样能够从陈述中检索新事实，但在其他请求形式上却有明显差异。因此，一个模型可以在学习事实之前，通过请求形式学习如何检索。为了理解这种差异，我们检查了在答案之前的隐藏状态，我们称之为上下文状态。当给定同一事实的两种不同请求形式时，在第一阶段接受五种形式训练的模型产生的上下文状态比仅在该阶段接受陈述训练的模型更为相似。在检索时改变该状态可以启用或阻止已学习事实的检索，并且这种效果在事实和事实关系（如首都和货币）之间是可以转移的。为了测试其在学习过程中的作用，我们仅在目标事实训练期间改变上下文状态。这一干预改变了后续检索，而在测试时没有干预。综合来看，这些结果表明，后续检索依赖于早期的请求形式经验以及事实学习期间的上下文状态。
cs.AI / 95 / 2610.03273
EVOL: Simulator-Guided Evolutionary Expert Synthesis for Deployment-Free Learning Path Recommendation
EVOL：无部署学习路径推荐的模拟引导进化专家合成
Abstract
Reinforcement learning (RL) for learning path recommendation (LPR) faces two coupled obstacles. First, the policy must commit to a sequence of L concepts without intermediate feedback, producing a combinatorial search space that grows super-exponentially with L and provides reward only at the final step. Second, expert learning paths would be the natural cure for sparse-reward RL, but they do not exist in educational data, because student logs record what learners did, not what they should have done. We address both obstacles by importing a recipe from simulator-based demonstration learning in robotics: the knowledge tracing simulator is used both to synthesize per-learner expert demonstrations through evolutionary search and to train a deployment-free policy that distills these demonstrations into a feed-forward learner. Our framework, EVOL, instantiates this pipeline with an asymmetric actor-critic where the actor commits to deployment-realistic blind planning while the critic exploits the privileged simulator state during training. Across three datasets (ASSIST15, Junyi, and EdNet; 39-189 concepts) and path lengths L = 5, 10, and 20, EVOL surpasses 8 baselines spanning heuristic, sequential, RL, graph-enhanced RL, and LLM-enhanced methods. We further compare three imitation strategies (BC, AWR, and DAPG) and show that final performance is governed by the quality of evolutionary experts rather than by the particular imitation objective.
Chinese Translation
强化学习（RL）在学习路径推荐（LPR）中面临两个相互关联的障碍。首先，策略必须在没有中间反馈的情况下承诺一系列 L 个概念，这导致组合搜索空间随着 L 的增加而呈超指数增长，并且仅在最后一步提供奖励。其次，专家学习路径本应是稀疏奖励 RL 的自然解决方案，但在教育数据中并不存在，因为学生日志记录的是学习者的行为，而不是他们应该采取的行动。我们通过借鉴机器人领域基于模拟器的示范学习中的一种方法来解决这两个障碍：知识追踪模拟器被用来通过进化搜索合成每个学习者的专家示范，并训练一个无部署的策略，将这些示范提炼为前馈学习者。我们的框架 EVOL 通过一个不对称的演员-评论家模型实现了这一流程，其中演员承诺进行与部署现实相符的盲目规划，而评论家在训练期间利用特权的模拟器状态。在三个数据集（ASSIST15、Junyi 和 EdNet；39-189 个概念）以及路径长度 L = 5、10 和 20 的实验中，EVOL 超越了 8 个基线方法，这些基线涵盖了启发式、顺序、RL、图增强 RL 和 LLM 增强方法。我们进一步比较了三种模仿策略（BC、AWR 和 DAPG），并表明最终性能受进化专家质量的影响，而不是特定的模仿目标。
cs.AI / 96 / 2610.03296
JOVE: Joint Execution and Verification for Resource-Aware LLM Task Graphs
JOVE：面向资源感知的LLM任务图的联合执行与验证
Abstract
Complex reasoning queries can be decomposed into directed acyclic task graphs and distributed across heterogeneous LLMs, reducing latency through parallelism and enabling smaller models to solve complex tasks. In practice, however, the suitability of an LLM for a given subtask may be a priori unknown, and execution alone does not reveal output correctness. We propose JOVE, an online framework that jointly assigns executor LLMs and selects intermediate outputs for paid verification. Verification runs asynchronously and is used to improve future allocations, so the system must balance spending on execution now against learning for later. We study how to optimize this trade-off under a long-term budget and a per-query latency constraint, with stochastic, initially unknown LLM service quality, invocation costs, and execution times. JOVE makes execution and verification decisions by solving a sequence of per-query mixed-integer linear programs. Online learning updates task-dependent estimates of LLM quality based on verification feedback, while an information-gain bonus incorporates the value of learning into allocation decisions. Under a natural set of assumptions, we establish sublinear quality-learning regret for JOVE. Across four reasoning benchmarks, JOVE achieves competitive accuracy against standard inference baselines while reducing average cost and latency by at least 3.17 times.
Chinese Translation
复杂推理查询可以被分解为有向无环任务图，并在异构的LLM（大语言模型）之间分配，通过并行化减少延迟，并使较小的模型能够解决复杂任务。然而，在实际应用中，某个LLM对于特定子任务的适用性可能是事先未知的，仅仅执行并不能揭示输出的正确性。我们提出了JOVE，一个在线框架，联合分配执行者LLM并选择中间输出进行付费验证。验证是异步进行的，用于改善未来的分配，因此系统必须在当前执行的支出与未来学习之间取得平衡。我们研究如何在长期预算和每个查询的延迟约束下优化这一权衡，考虑到随机的、最初未知的LLM服务质量、调用成本和执行时间。JOVE通过解决一系列每个查询的混合整数线性规划来做出执行和验证决策。在线学习根据验证反馈更新任务依赖的LLM质量估计，同时信息增益奖励将学习的价值纳入分配决策。在一组自然假设下，我们为JOVE建立了亚线性的质量学习遗憾。在四个推理基准测试中，JOVE在与标准推理基线的比较中实现了具有竞争力的准确性，同时将平均成本和延迟减少了至少3.17倍。
cs.AI / 97 / 2610.03312
Optimal Planning in a Dynamic World
动态世界中的最优规划
Abstract
Background: We address the problem of planning when the set of feasible states or actions changes over time. For example, in the problem of path planning among moving obstacles (sometimes known as SIPP), the feasibility of being at a particular location can change as the obstacles move. Or, the action of boarding a particular train is feasible only while it is stopped at the station. This dynamism means that the optimal plan and its duration can change depending on when execution begins. In practice, execution start time is often unknown until planning has completed or another agent gives the go-ahead. However, most prior planning work either ignores dynamism or assumes a known start time. This makes it straightforward to assess state and action feasibility but is impractical for some applications. Objectives: In this paper, we relax the assumption of a known start time. We define the setting of {\em any-start-time planning} and provide algorithms for it. Methods: We present a data structure called a compound arrival time function (cATF) that compactly encodes the optimal plan as a function of start time. We provide general-purpose planning algorithms, based on heuristic graph search, that assemble cATFs by propagating functions along edges instead of scalar costs. Results: We prove that the size of a cATF is at most linear in the problem size. An experimental evaluation of an implementation for the specific problem of SIPP shows that, on difficult problems, agents that rely on replanning often fail, while any-start-time algorithms using cATFs can quickly look up the optimal plan once the execution start time is known. Conclusions: By enabling efficient representations and reasoning for time-dependent plans, this work provides a foundation for planning in dynamic worlds.
Chinese Translation
背景：我们探讨了在可行状态或动作集合随时间变化时的规划问题。例如，在移动障碍物之间的路径规划问题中（有时称为 SIPP），在特定位置的可行性可能会随着障碍物的移动而变化。或者，登上特定列车的动作仅在列车停靠在车站时是可行的。这种动态性意味着最优计划及其持续时间可能会根据执行开始的时间而变化。在实践中，执行开始时间通常在规划完成之前是未知的，或者需要其他代理的批准。然而，大多数先前的规划工作要么忽略了动态性，要么假设已知的开始时间。这使得评估状态和动作的可行性变得简单，但对于某些应用来说并不实用。目标：在本文中，我们放宽了已知开始时间的假设。我们定义了{ extit{任何开始时间规划}}的设置，并为其提供了算法。方法：我们提出了一种称为复合到达时间函数（compound arrival time function, cATF）的数据结构，它紧凑地编码了最优计划作为开始时间的函数。我们提供了基于启发式图搜索的通用规划算法，通过沿边传播函数而不是标量成本来组装 cATF。结果：我们证明了 cATF 的大小至多与问题规模呈线性关系。针对 SIPP 特定问题的实现的实验评估表明，在困难问题上，依赖重新规划的代理往往失败，而使用 cATF 的任何开始时间算法可以在执行开始时间已知后快速查找最优计划。结论：通过实现时间依赖计划的高效表示和推理，这项工作为动态世界中的规划提供了基础。
cs.AI / 98 / 2610.03315
Lightweight, Rubric-Guided Trajectory Evaluation for Production AI Agents
轻量级、基于评分标准的生产AI代理轨迹评估
Abstract
Trajectory evaluation is essential for improving the reliability of LLM-based agents, but production use makes it expensive to run repeatedly. Modern agents generate long traces containing tool calls, observations, retries, and external outputs, while not all raw tokens are equally useful for diagnosis. We present \textit{LiteTrajEval}, a lightweight architecture for budget-bounded trajectory evaluation. LiteTrajEval derives compact domain-specific rule profiles offline, then preprocesses each trajectory online, marks heuristic failure signals, serializes it under a fixed global budget, and invokes a single rubric-guided LLM judge to produce structured diagnostic reports. Evaluated on public Magentic-One-style and $\tau$-bench-style trajectory datasets, LiteTrajEval improves failure-localization alignment with human annotations by roughly 20--35 percentage points on Magentic-One and up to 23 percentage points on $\tau$-retail compared with AgentRx, while reducing cost by about 6$\times$ and evaluation time by more than 8$\times$. This solution has also been deployed in our enterprise agentic platform.
Chinese Translation
轨迹评估对于提高基于大语言模型（LLM）的代理的可靠性至关重要，但在生产环境中反复运行会导致高昂的成本。现代代理生成包含工具调用、观察、重试和外部输出的长轨迹，而并非所有原始标记在诊断中都是同样有用的。我们提出了 extit{LiteTrajEval}，一种适用于预算限制的轻量级轨迹评估架构。LiteTrajEval 离线推导紧凑的领域特定规则配置文件，然后在线预处理每个轨迹，标记启发式失败信号，在固定的全局预算下进行序列化，并调用单个基于评分标准的 LLM 评审生成结构化诊断报告。在公共的 Magentic-One 风格和 $ au$-bench 风格轨迹数据集上进行评估时，LiteTrajEval 在 Magentic-One 上将失败定位与人工注释的对齐度提高了大约 20-35 个百分点，在 $ au$-retail 上提高了最多 23 个百分点，相比于 AgentRx，同时成本降低了约 6 倍，评估时间减少了超过 8 倍。该解决方案也已在我们的企业代理平台上部署。
cs.AI / 99 / 2610.03316
Multi-Task Evolution for Zero-Shot Cross-Problem Generalization using LLMs
基于大型语言模型的零-shot跨问题泛化的多任务进化
Abstract
Designing effective heuristics for diverse combinatorial optimization problems requires substantial expertise and repeated search. Large language models (LLMs) automate heuristic generation and refinement, but heuristic search typically depends on evaluation feedback from the problem being optimized. Generalizing to new problem definitions using only source-task feedback therefore remains a central challenge. We introduce MECo, an LLM-driven multi-task evolutionary framework for zero-shot cross-problem generalization. MECo maintains task-conditioned heuristic populations and uses a transfer gap based on cross-task population performance to guide their interactions. These interactions enable the transfer and recombination of heuristics. A complementary selection criterion then constructs a compact heuristic set by rewarding each member's additional coverage of source combinations. The selected set is applied to target problems without further search or adaptation. Experiments on 32 problem variants across vehicle routing (VRP) and flexible job-shop scheduling (FJSP) show that MECo achieves the lowest mean costs compared with eight automated heuristic design (AHD) baselines under the same budgets. On out-of-domain problems, it outperforms the strongest baseline in each family. Moreover, integrating the framework of MECo with different AHD methods improves their ID and OOD performance in both families, supporting its effectiveness across different methods.
Chinese Translation
为多样的组合优化问题设计有效的启发式方法需要大量的专业知识和反复的搜索。大型语言模型（LLMs）能够自动化启发式生成和优化，但启发式搜索通常依赖于被优化问题的评估反馈。因此，仅依靠源任务反馈对新问题定义进行泛化仍然是一个核心挑战。我们提出了MECo，一个基于LLM的多任务进化框架，用于零-shot跨问题泛化。MECo维护任务条件的启发式种群，并利用基于跨任务种群表现的转移差距来指导它们的交互。这些交互使得启发式的转移和重组成为可能。一个互补的选择标准通过奖励每个成员对源组合的额外覆盖来构建一个紧凑的启发式集合。所选集合在没有进一步搜索或适应的情况下应用于目标问题。在车辆路径规划（VRP）和灵活作业车间调度（FJSP）上的32个问题变体实验表明，MECo在相同预算下实现了最低的平均成本，相较于八个自动化启发式设计（AHD）基线。在域外问题上，它在每个家族中超越了最强基线。此外，将MECo框架与不同的AHD方法结合使用，提高了这两种家族中的ID和OOD性能，支持了其在不同方法中的有效性。
cs.AI / 100 / 2610.03320
Refinement Buys Intelligibility, Search Buys Identity: What Test-Time Compute Buys in Masked-Diffusion TTS
精细化提升可理解性，搜索提升身份识别：掩蔽扩散文本到语音生成中的测试时计算收益
Abstract
Diffusion language models for text-to-speech combine two forms of computation: model depth (parameters) and refinement steps (inference budget). We ask whether they scale equally across capabilities. We train 15 masked-diffusion codec TTS models varying depth (19-133M parameters, 3 seeds) on 2,000 hours of speech and sweep refinement steps T in [1,16] at inference, measuring zero-shot synthesis via ASR word error rate (intelligibility) and speaker verification (identity) on 174 held-out speakers. Against measured floors, refinement closes 86.2% of the intelligibility range but only 46.4% of the identity range - a 1.86x asymmetry robust across multiple error metrics. Retraining at 3x and 6x schedule attenuates but does not reverse this gap (1.84 to 1.36 to 1.23x), because intelligibility saturates with steps while identity continues improving. Best-of-K search recovers speaker identity where refinement fails, with 64.6-79.0% win rates across four independent encoders. Depth and steps are not interchangeable: separable B(d)B(T) fits significantly better (Delta AICc=+69.3) than substitution models. Analysis shows 62% of remaining identity deficit lies in the codec, not the generator. We conclude that refinement and depth target different bottlenecks and should be optimized separately.
Chinese Translation
用于文本到语音的扩散语言模型结合了两种计算形式：模型深度（参数）和精细化步骤（推理预算）。我们探讨它们在能力上的扩展是否相同。我们训练了15个掩蔽扩散编解码器文本到语音模型，参数深度从19M到133M（3个种子），在2000小时的语音数据上进行训练，并在推理时对精细化步骤T进行[1,16]的范围测试，通过自动语音识别（ASR）单词错误率（可理解性）和说话人验证（身份）在174个保留说话者上进行测量。与测量的基准相比，精细化缩小了86.2%的可理解性范围，但仅缩小了46.4%的身份范围——在多个错误指标上表现出1.86倍的不对称性。以3倍和6倍的计划重新训练虽然减小了这一差距（从1.84降至1.36再降至1.23倍），但并未逆转这一趋势，因为可理解性在步骤增加时趋于饱和，而身份识别则持续改善。在精细化失败的情况下，最佳K搜索能够恢复说话人的身份，在四个独立编码器中赢得率为64.6%-79.0%。深度和步骤并不可互换：可分离的B(d)B(T)模型显著优于替代模型（Delta AICc=+69.3）。分析显示，62%的剩余身份缺口存在于编解码器中，而非生成器。我们得出结论，精细化和深度针对不同的瓶颈，应分别进行优化。
cs.AI / 101 / 2610.03326
Preserving Mathematical Reasoning in Compressed Diffusion Language Models via Trajectory-Aware Low-Rank Approximation
通过轨迹感知低秩近似在压缩扩散语言模型中保持数学推理
Abstract
Diffusion language model (dLLM) compression faces a known challenge because calibration is typically performed on clean, fully visible activations, whereas inference traverses partially masked intermediate states. For low-rank compression, this raises two questions. First, can low-rank optimality still be characterized when approximation quality is measured over trajectory-distributed states, and second, does the choice of calibration states affect mathematical reasoning preservation under compression? We address these questions by formulating a trajectory-aware low-rank objective over corruption levels and masking realizations. To estimate this objective efficiently, we propose Traj-MC, which estimates the trajectory second moment through Monte Carlo sampling and yields exact sampled-state optimality and population consistency. Under matched compression budgets, trajectory-aware calibration improves reconstruction over the generation trajectory and preserves substantially more mathematical reasoning than clean calibration on mathematical reasoning benchmarks. Our results connect trajectory-aware low-rank optimality to the reasoning capability retained after dLLM compression. Our code is available at: https://github.com/Zishan-Shao/traj-mc.git.
Chinese Translation
扩散语言模型（dLLM）压缩面临一个已知挑战，因为标定通常是在干净、完全可见的激活状态下进行的，而推理则遍历部分遮蔽的中间状态。对于低秩压缩，这引出了两个问题。首先，当在轨迹分布状态上测量近似质量时，低秩最优性是否仍然可以被表征；其次，标定状态的选择是否会影响在压缩下数学推理的保持？我们通过在损坏水平和遮蔽实现上制定一个轨迹感知的低秩目标来解决这些问题。为了高效地估计该目标，我们提出了Traj-MC，它通过蒙特卡洛采样估计轨迹的二阶矩，并实现了精确的采样状态最优性和总体一致性。在匹配的压缩预算下，轨迹感知标定在生成轨迹上改善了重建，并在数学推理基准上保持了比干净标定更多的数学推理。我们的结果将轨迹感知低秩最优性与dLLM压缩后保留的推理能力联系起来。我们的代码可在以下链接获取：https://github.com/Zishan-Shao/traj-mc.git。
cs.AI / 102 / 2610.03356
ReFract: Benchmarking Perspective Awareness in Language Model Agents with Text World Models
ReFract：基于文本世界模型对语言模型代理的视角意识进行基准测试
Abstract
Large Language Model (LLM) agents are increasingly deployed in high-stakes settings such as industrial maintenance and equipment fault troubleshooting, where workers occupy a variety of roles. A capable agent must therefore act in a way that is calibrated to user's role: taking actions and providing information that respect the role's knowledge and capability boundaries. Unlike coding, where mistakes are usually recoverable, agent responses in these settings are enacted on physical equipment, and can therefore cause irreversible equipment damage, production loss, or personnel harm. Existing benchmarks, however, largely overlook the need for agents to infer what a role intends and acting only through tools that role may legitimately use, a capability which we term Perspective Awareness. To this end, we introduce ReFract, a benchmark of 150 expert-validated entries in which an agent must act differently in response to the same query depending on user's role. Entries of ReFract are grounded in anonymized queries from domain support conversations, against which we construct Text World Models that simulate the agent's operating environments and assemble perspective-aware action trajectories. State-of-the-art LLMs solve at most 69% of the tasks with more than 50% of their trajectories contain attempts of taking perspective-violating actions. ReFract exposes perspective awareness as a distinct, largely unsolved axis of agent evaluation and motivates agents that calibrate not just how to act, but for whom.
Chinese Translation
大型语言模型（LLM）代理越来越多地应用于高风险环境，如工业维护和设备故障排查，在这些环境中，工作人员扮演多种角色。因此，一个合格的代理必须以符合用户角色的方式行动：采取行动并提供信息，尊重角色的知识和能力边界。与编码不同，编码中的错误通常是可恢复的，而在这些环境中，代理的响应是在物理设备上实施的，因此可能导致不可逆的设备损坏、生产损失或人员伤害。然而，现有的基准测试在很大程度上忽视了代理推断角色意图的需求，以及仅通过该角色可能合法使用的工具进行行动的能力，我们称之为视角意识。为此，我们引入了ReFract，这是一个包含150个专家验证条目的基准测试，其中代理必须根据用户角色对同一查询做出不同的响应。ReFract的条目基于来自领域支持对话的匿名查询，我们在此基础上构建文本世界模型（Text World Models），模拟代理的操作环境并组装视角意识行动轨迹。最先进的LLM在最多69%的任务中解决了问题，但其轨迹中超过50%包含试图采取违反视角的行动。ReFract揭示了视角意识作为代理评估的一个独特且在很大程度上未解决的维度，并激励代理不仅要校准如何行动，还要考虑为谁行动。
cs.AI / 103 / 2610.03363
Geometry Meets Physics: Data-Efficient Pre-Training for Unstructured Neural PDE Solvers
几何与物理的结合：用于非结构化神经偏微分方程求解器的数据高效预训练
Abstract
Neural surrogate models for Partial Differential Equations (PDEs) on unstructured 3D geometries are often limited by poor generalization and the high cost of generating large-scale training datasets. Consequently, pre-training on massive datasets of related PDE dynamics has emerged as a critical alternative to enhance the robustness and scalability of these models. However, this strategy is neither compute- nor data-efficient, as it relies on massive pre-computed data that is very costly to generate. In this work, we introduce a disk-data-free pre-training framework tailored to both steady-state and transient regimes. For steady-state problems, we propose a geometry-driven strategy that leverages intrinsic shape descriptors to learn representations of complex 3D domains. For transient problems, we introduce a physics-driven approach based on online generation of synthetic PDE data, enabling scalable pre-training without reliance on expensive datasets. Across multiple experiments, our approach achieves faster convergence, greater data efficiency, and higher accuracy during fine-tuning, particularly under realistic low-data regimes. This methodology provides a practical pathway toward data-efficient neural emulators for large-scale simulations.
Chinese Translation
针对非结构化三维几何体的偏微分方程（PDE）神经代理模型常常受到泛化能力差和生成大规模训练数据集高成本的限制。因此，在相关PDE动态的大规模数据集上进行预训练已成为增强这些模型的鲁棒性和可扩展性的关键替代方案。然而，这一策略在计算和数据效率上均不理想，因为它依赖于大量预计算数据，而这些数据的生成成本非常高。在本研究中，我们提出了一种无磁盘数据的预训练框架，适用于稳态和瞬态两种状态。对于稳态问题，我们提出了一种基于几何的策略，利用内在形状描述符学习复杂三维域的表示。对于瞬态问题，我们引入了一种基于物理的在线合成PDE数据生成方法，使得可扩展的预训练不再依赖于昂贵的数据集。在多个实验中，我们的方法实现了更快的收敛速度、更高的数据效率和更好的微调准确性，特别是在现实的低数据环境下。这一方法为大规模仿真的数据高效神经模拟器提供了一条实用的路径。
cs.AI / 104 / 2610.03367
Multilingual GSM-Symbolic: What determines capability transfer across languages?
多语言GSM-符号：是什么决定了跨语言能力的转移？
Enevoldsen, Kenneth, Herchert, Riley, Mosegaard, Sofie, Smart, Dan Saattrup, Enni, Simon, Chung, Isaac, Bruun, Sofie, Munot, Ayush Sunil, Müller-Eberstein, Max, El-Assadi, Adnan, Bassignana, Elisa, Barmina, Gianluca, Einarsson, Hafsteinn, Debess, Iben Nyholm, Freienthal, Linda, Poech, Lukas Galke, Zhang, Mike, Legrand, Nicolas, Salnikov, Vladimir, Kostiuk, Yevhen, Hussain, Zafar, Kaur, Sagandeep, Toftgård, Agnes, Mattson, Marie, Nielbo, Kristoffer
Abstract
We understand little about how capabilities acquired in one language carry over to another, or what governs this transfer: evaluations rely on incomparable, saturation-prone datasets and rarely examine its determinants jointly. Identifying what predicts transfer would let us avoid exhaustive evaluation across all language pairs and let developers target the factors that limit performance in low-resource languages. To evaluate cross-lingual capability transfer, we introduce Multilingual GSM-Symbolic, an extensible multilingual mathematical dataset covering 30,000 item-matched question-answer pairs and spanning 15 languages. It utilises symbolic templates to prevent overfitting and ensure generalisation by allowing generation of millions of high-quality variations from a single sample. Using Multilingual GSM-Symbolic, we quantify the largest determinants of capability as model size ($\beta = 1.77$), language resource level ($\beta = 0.77$), reasoning ($\beta = 0.67$) and typological distance ($\beta = -0.25$). This joint estimation allows these determinants to be expressed in terms of one another: a 32B model evaluated in Marathi performs like a 10B model in English. Our findings have important implications for model developers, showing that model size and reasoning narrow the performance gap between low- and high-resource languages ($\beta = -0.27$ and $\beta = -0.20$, respectively), while similar levers have little or no effect on typologically distant languages. Overall, our analysis framework explains 92% of between-language variation, but only 23% of the model-by-language variation, and predicts a model's performance on an unseen language within 6.0pp (r=.96). Incorporating measurements from just 10 templates in the target language reduces this to 4.19pp, enabling reasonable estimates of performance with little or no downstream dataset.
Chinese Translation
我们对在一种语言中获得的能力如何转移到另一种语言，以及是什么决定这种转移知之甚少：现有评估依赖于不可比拟且易饱和的数据集，且很少共同考察其决定因素。识别出预测转移的因素将使我们避免在所有语言对之间进行详尽的评估，并使开发者能够针对限制低资源语言性能的因素进行优化。为了评估跨语言能力转移，我们引入了多语言GSM-符号（Multilingual GSM-Symbolic），这是一个可扩展的多语言数学数据集，涵盖了30,000对项目匹配的问题-答案对，涉及15种语言。它利用符号模板防止过拟合，并通过允许从单个样本生成数百万个高质量变体来确保泛化。使用多语言GSM-符号，我们量化了能力的最大决定因素，包括模型大小（$eta = 1.77$）、语言资源水平（$eta = 0.77$）、推理能力（$eta = 0.67$）和类型距离（$eta = -0.25$）。这种联合估计使得这些决定因素可以相互表达：在马拉地语中评估的32B模型表现得像在英语中评估的10B模型。我们的发现对模型开发者具有重要意义，表明模型大小和推理能力缩小了低资源语言与高资源语言之间的性能差距（分别为$eta = -0.27$和$eta = -0.20$），而类似的杠杆对类型上相距较远的语言几乎没有影响。总体而言，我们的分析框架解释了92%的语言间变异，但仅解释了23%的模型与语言变异，并在未见语言上预测模型性能的误差在6.0个百分点内（r=.96）。在目标语言中仅结合10个模板的测量将这一误差减少到4.19个百分点，使得在几乎没有下游数据集的情况下能够合理估计性能。
cs.AI / 105 / 2610.03383
CVE2AP: Automated Generation of PDDL-Encoded Attack Paths via Large Language Models
CVE2AP：通过大型语言模型自动生成PDDL编码的攻击路径
Abstract
Attack Path (AP) modeling is fundamental to cybersecurity analysis, where the Planning Domain Definition Language (PDDL) has been widely adopted to encode APs into formal and machine-verifiable representations for automated reasoning about vulnerability exploitation, attack progression, and their potential impacts. However, existing AP modeling approaches largely rely on expert-driven manual construction, limiting their scalability and ability to keep pace with rapidly evolving cyber threats. Large language models (LLMs) are promising candidates, as their extensive pre-trained knowledge and reasoning capabilities enable them to interpret and transform threat intelligence into formal representations. In this paper, we propose \textbf{CVE2AP}, an LLM-based approach for automatically generating PDDL-encoded attack paths from natural language CVE (Common Vulnerability Exposure) descriptions. CVE2AP leverages structured prompting and incorporates an error-feedback mechanism that iteratively refines the generated paths using planner-reported syntactic and solvability errors. We conduct a systematic empirical evaluation across multiple LLMs and generation configurations, assessing generation quality across syntactic, solvability and semantic dimensions, together with token consumption and generation time. The results demonstrate that CVE2AP effectively generates high-quality PDDL-encoded attack paths, achieving up to 86.9\% syntax correctness, 78.6\% solvability, and 93.1\% semantic correctness under LLM-as-expert evaluation, while \texttt{GPT-5.5} offers the best quality-cost trade-off and error feedback yields the most consistent quality improvement.
Chinese Translation
攻击路径（AP）建模是网络安全分析的基础，其中规划领域定义语言（PDDL）被广泛采用，以将AP编码为正式且可机器验证的表示形式，从而实现对漏洞利用、攻击进展及其潜在影响的自动推理。然而，现有的AP建模方法在很大程度上依赖于专家驱动的手动构建，这限制了其可扩展性以及跟上快速演变的网络威胁的能力。大型语言模型（LLMs）是有前景的候选者，因为它们广泛的预训练知识和推理能力使其能够将威胁情报解释并转化为正式表示。在本文中，我们提出了 extbf{CVE2AP}，一种基于LLM的方法，用于从自然语言CVE（公共漏洞曝光）描述中自动生成PDDL编码的攻击路径。CVE2AP利用结构化提示，并结合错误反馈机制，迭代地通过规划者报告的语法和可解性错误来细化生成的路径。我们对多个LLM和生成配置进行了系统的实证评估，评估生成质量在语法、可解性和语义维度上的表现，以及标记消耗和生成时间。结果表明，CVE2AP有效地生成高质量的PDDL编码攻击路径，在LLM作为专家评估下，语法正确率高达86.9\%，可解性为78.6\, 语义正确率为93.1\, 而 exttt{GPT-5.5}提供了最佳的质量-成本权衡，错误反馈机制则带来了最一致的质量提升。
cs.AI / 106 / 2610.03387
Benchmarking Candidate Coverage in Typed Decision Models
类型决策模型中的候选覆盖基准测试
Abstract
Typed decision models return choices or distributions over answer options supplied at request time. Accuracy with complete options does not establish whether a model recognizes that a reference answer is missing or avoids rejecting valid candidates. We present a paired candidate-coverage benchmark protocol and an initial evaluation of Laya and Jev across AG News, DBpedia, Emotion, and TREC. The models receive identical frozen texts and requests: 300 calibration and 589 test texts yield 23,932 predictions per model. Present/absent pairs match ordinary candidate count, and name variants preserve descriptions, members, and order. Native rejection behavior differs sharply: at five TREC candidates with natural names, Laya detects 97.2% of missing-answer cases but falsely rejects 69.7% of present controls; Jev's rates are 24.8% and 0.0%. Calibration-only none-score thresholds change these rates to 33.9%/3.7% and 45.0%/1.8%, respectively. On DBpedia, Jev's high coverage-score AUROC supports a stronger operating point, whereas both models have weak complete-set accuracy on Emotion. Competence-conditioned analysis, probability-precision sensitivity, and interface audits show why classification, score ranking, and rejection policies need separate measurement. This initial benchmark is descriptive and limited to reference-label omission; it does not establish natural out-of-scope generalization, causal mechanisms, or a new rejection method.
Chinese Translation
类型决策模型在请求时返回选择或答案选项的分布。对于完整选项的准确性并不能确定模型是否识别出参考答案缺失或避免拒绝有效候选项。我们提出了一种配对候选覆盖基准协议，并对 Laya 和 Jev 在 AG News、DBpedia、Emotion 和 TREC 数据集上的初步评估进行了分析。这些模型接收相同的冻结文本和请求：300 个校准文本和 589 个测试文本每个模型产生 23,932 个预测。存在/缺失对的数量与普通候选项数量相匹配，名称变体保留了描述、成员和顺序。原生拒绝行为差异显著：在五个具有自然名称的 TREC 候选项中，Laya 检测到 97.2% 的缺失答案案例，但错误拒绝了 69.7% 的存在控制；而 Jev 的比率分别为 24.8% 和 0.0%。仅校准的无评分阈值将这些比率分别改变为 33.9%/3.7% 和 45.0%/1.8%。在 DBpedia 上，Jev 的高覆盖分数 AUROC 支持更强的操作点，而在 Emotion 上两个模型的完整集准确性较弱。能力条件分析、概率-精度敏感性和接口审计显示分类、评分排名和拒绝政策为何需要单独测量。这个初步基准是描述性的，局限于参考标签的遗漏；它并未建立自然的超范围泛化、因果机制或新的拒绝方法。
cs.AI / 107 / 2610.03425
Becoming Suspicious Across Borders: Algorithmic Extraterritoriality and AI-Driven Financial Surveillance
跨境产生怀疑：算法的域外性与人工智能驱动的金融监控
Abstract
Suspicion is an important, yet elusive concept in anti-money laundering and counter-terrorist financing (AML/CFT), which allows for intervention below the threshold of proof. In its traditional form, suspicion can be understood as a situated legal judgement by human actors within identifiable jurisdictions. It is argued that this understanding is no longer adequate. As artificial intelligence (AI) becomes an integral part of financial surveillance, suspicion is increasingly produced through data-driven processes. This transformation is epistemic, but also spatial. Since AI-driven financial surveillance operates through transnational data infrastructures, regulatory reach is less a matter of where conduct occurs than a question of whether such conduct becomes visible within data systems. This article develops the concept of algorithmic extraterritoriality, understood as a form of regulatory power mediated by data infrastructures rather than formal assertions of jurisdiction. Moreover, since individuals are increasingly constituted as datafied subjects of suspicion, they are rendered governable through dispersed and opaque processes of evaluation. This constitutes a challenge for accountability and contestability because suspicion becomes more difficult to locate, explain or contest.
Chinese Translation
怀疑是反洗钱和反恐怖融资（AML/CFT）中一个重要但难以捉摸的概念，它允许在证据阈值以下进行干预。在其传统形式中，怀疑可以被理解为人类行为者在可识别的管辖区内的情境法律判断。有人认为这种理解已经不再足够。随着人工智能（AI）成为金融监控的一个不可或缺的部分，怀疑越来越多地通过数据驱动的过程产生。这一转变既是认识论上的，也是空间上的。由于人工智能驱动的金融监控通过跨国数据基础设施运作，监管的范围不再仅仅取决于行为发生的地点，而是取决于该行为是否在数据系统中变得可见。本文发展了算法域外性的概念，理解为一种通过数据基础设施中介的监管权力，而非正式的管辖权主张。此外，随着个体越来越多地被构建为数据化的怀疑主体，他们通过分散和不透明的评估过程变得可治理。这对问责和争议构成了挑战，因为怀疑变得更加难以定位、解释或争辩。
cs.AI / 108 / 2610.03430
Jumping the Line: Exploiting Length Predictions in LLM Scheduling
跳过界限：利用长度预测进行大语言模型调度
Abstract
Efficient request scheduling is increasingly important for reducing completion time in large language model (LLM) serving. Size-based policies such as Shortest Job First prioritize shorter requests, but output lengths are unknown before generation, so practical schedulers rely on predicted lengths. We introduce JIL, an attack on prediction-based LLM schedulers that manipulates the scheduling signal to obtain higher priority and reduce completion time. Using TRAIL as a case study, JIL optimizes an adversarial suffix that causes a lightweight output-length probe to underestimate a request's length. We evaluate JIL on two datasets and four LLMs across varied request profiles and deployment configurations. JIL reduces predicted output lengths by up to 83.4 percent, and adversarial requests complete up to 1.53 times faster on average in end-to-end serving experiments. The reduction in predicted length is substantially larger than the change in actual output length, revealing a mismatch between the scheduler's estimate and the request's realized size. Response utility varies across models and tasks, exposing a trade-off between scheduling advantage and response quality. We also evaluate scheduler-side defenses and find that grouping length predictions into coarse intervals reduces JIL's scheduling advantage and mitigates delays to benign requests.
Chinese Translation
高效的请求调度在减少大语言模型（LLM）服务的完成时间方面变得越来越重要。基于大小的策略，如最短作业优先（Shortest Job First），优先处理较短的请求，但在生成之前输出长度是未知的，因此实际调度器依赖于预测的长度。我们提出了JIL，这是一种针对基于预测的LLM调度器的攻击，通过操控调度信号来获取更高的优先级并减少完成时间。以TRAIL为案例研究，JIL优化了一种对抗性后缀，使得轻量级输出长度探测器低估请求的长度。我们在两个数据集和四个LLM上评估了JIL，涵盖了不同的请求特征和部署配置。JIL将预测的输出长度减少了多达83.4%，而对抗性请求在端到端服务实验中的平均完成速度提高了1.53倍。预测长度的减少远大于实际输出长度的变化，揭示了调度器的估计与请求实际大小之间的不匹配。响应效用在不同模型和任务中有所不同，暴露了调度优势与响应质量之间的权衡。我们还评估了调度器端的防御措施，发现将长度预测分组为粗略区间可以减少JIL的调度优势，并减轻对良性请求的延迟。
cs.AI / 109 / 2610.03458
A Near-Zero Monitor Readout Is Not Evidence of Behavioral Control
近零监控读数并不是行为控制的证据
Abstract
Post-training with verifiable rewards can induce reward hacking, motivating the use of monitors within the training objective rather than solely for offline auditing. We show that a low monitor readout does not identify whether such an intervention controls behavior. In a code-generation environment whose dominant exploit is available at the start of the reasoning trace, we train policies against three monitors that pass the same offline gate: an in-domain activation probe and two penalties conditioned on how early the policy commits to its own final answer. The probe score is at its numerical floor from the first recorded training step, and the trained-score median is zero for every prefix-trained run at the endpoint. These readouts estimate different quantities, and we do not compare their scales; within each monitor family, however, low values do not establish behavioral control. Within one fixed configuration, prefix-trained runs with the same zero-median trained score range, by seed alone, from a mixed regime with a low hacking share to near-pure reward hacking. All probe runs reach the hacking regime, but their floor-level readout reflects a mismatch between the position where the probe was validated and the position where it was read during training, not a second instance of this ambiguity. Text-level analysis identifies a prefix failure mode: generic planning and filler shells postpone the exploit past the cut without eliminating it from the final output. Low measured commitment therefore does not distinguish a low hacking share from delayed commitment to the exploit. Offline discrimination and low monitor-aligned readouts are insufficient evidence of behavioral control; an out-of-band behavioral check is required. We characterize the endpoint readout, not its evolution. Code is available at https://github.com/zhezhou1106/spoof-cost.
Chinese Translation
使用可验证奖励进行后训练可能会引发奖励黑客行为，这促使在训练目标中使用监控器，而不仅仅是用于离线审计。我们表明，低监控读数并不能识别这种干预是否控制了行为。在一个代码生成环境中，其主导利用在推理轨迹开始时就可用，我们针对三个通过相同离线门的监控器训练策略：一个领域内激活探针和两个基于策略多早承诺其最终答案的惩罚。探针分数从第一次记录的训练步骤开始就处于其数值下限，而每个前缀训练运行在终点的训练分数中位数为零。这些读数估计不同的量，我们不比较它们的尺度；然而，在每个监控器家族内，低值并不能确立行为控制。在一个固定配置下，具有相同零中位数训练分数的前缀训练运行，仅通过种子，范围从低黑客份额的混合状态到近乎纯粹的奖励黑客。所有探针运行都达到黑客状态，但它们的底层读数反映了探针验证位置与训练期间读取位置之间的不匹配，而不是这种模糊性的第二实例。文本级分析识别出一种前缀失败模式：通用规划和填充壳将利用推迟到切割之后，而未将其从最终输出中消除。因此，低测量承诺并不能区分低黑客份额与对利用的延迟承诺。离线区分和低监控对齐读数不足以证明行为控制；需要进行带外行为检查。我们描述了终点读数，而不是其演变。代码可在 https://github.com/zhezhou1106/spoof-cost 获取。
cs.AI / 110 / 2610.03509
Efficient Reasoning Training Does Not Always Harm CoT Faithfulness and Monitorability
高效推理训练并不总是损害链式推理的可信度和可监控性
Abstract
Chain-of-thought (CoT) reasoning allows humans to inspect how large language models reach their answers, and oversee model behaviour. This reasoning comes at an increased inference cost, motivating efficient methods that train models to solve tasks using fewer tokens. However, a common concern is that such training may cause models to skip important reasoning steps, so the CoT no longer faithfully reflects the model's decision. It is unclear whether or when this occurs in practice, since different efficiency methods apply length pressure to models' CoT in distinct ways, and faithfully explaining a model's decision takes more tokens on some tasks than others. To understand these dynamics, we fine-tune a variety of models with three methods that apply length pressure differently, namely a fixed generation budget, a per-example length target, and a group-relative length reward. We evaluate how efficient reasoning affects CoT faithfulness (i.e., how well the CoT reflects model decisions on related inputs) and monitorability (i.e., whether the CoT reveals when input interventions alter the output). We find that it affects faithfulness and monitorability differently. Faithfulness falls in most settings, primarily because the trained models are less consistent. Monitorability is more robust, as models keep acknowledging the influence on their answer even when the CoT is much shorter.
Chinese Translation
链式推理（Chain-of-thought，CoT）使人类能够检查大型语言模型如何得出答案，并监督模型行为。这种推理带来了更高的推理成本，因此激励开发高效的方法，以训练模型使用更少的标记解决任务。然而，一个普遍的担忧是，这种训练可能导致模型跳过重要的推理步骤，从而使得CoT不再忠实地反映模型的决策。目前尚不清楚这种情况在实践中是否或何时会发生，因为不同的高效方法以不同的方式对模型的CoT施加长度压力，而忠实地解释模型的决策在某些任务上所需的标记比其他任务要多。为了理解这些动态，我们使用三种以不同方式施加长度压力的方法对多种模型进行微调，即固定生成预算、每个示例的长度目标和相对组长度奖励。我们评估高效推理如何影响CoT的可信度（即CoT在相关输入上反映模型决策的程度）和可监控性（即CoT是否揭示输入干预何时改变输出）。我们的发现表明，这两者受到的影响不同。在大多数情况下，可信度下降，主要是因为训练后的模型一致性降低。可监控性则更为稳健，因为即使CoT大大缩短，模型仍然能够承认其答案受到影响。
cs.AI / 111 / 2610.03519
Reasoning Models Are Accurate but Unsound on Identification
推理模型在识别方面准确但不可靠
Abstract
A reasoning model asked whether a causal effect is recoverable from observational data can fail in two ways: it refuses an identifiable query or answers a nonidentifiable one. The latter is more consequential, as no observational data can validate the claimed formula. Measuring this failure requires queries that are provably non-identifiable, which prior evaluations lack, and grading that accepts correct formulas in any equivalent form, which string matching cannot provide. We build CERTID, a formal identification pipeline that addresses both limitations. CERTID uses the sound and complete causal identification algorithm ID to certify whether an effect is identifiable from a given graph and query, and verifies returned formulas against structural causal models whose interventional distributions are known exactly. CERTID further develops theoretical results to mitigate structural leakage, repair non-identifiable queries, and establish grading guarantees. We evaluate three frontier reasoning models (Gemini Flash, Gemini Pro, and GPT5.5) on 1,200 certified instances spanning 4 to 50 vertices. Accuracy proves a poor proxy for soundness: on identical instances, the false-claim rate on non-identifiable queries varies by seventeen-fold across models. We also find that models decide identifiability with 97-100% accuracy on graphs generated after the strongest model's training snapshot. Instances, the certification procedure, the verifier, and per-instance records are available at https://anonymous.4open.science/r/certid-D718.
Chinese Translation
一个推理模型在询问是否可以从观察数据中恢复因果效应时，可能会以两种方式失败：拒绝一个可识别的查询或回答一个不可识别的查询。后者的后果更为严重，因为没有观察数据可以验证所声称的公式。衡量这种失败需要证明为不可识别的查询，而先前的评估缺乏这一点，并且评分需要接受任何等效形式的正确公式，而字符串匹配无法提供。我们构建了CERTID，一个正式的识别管道，解决了这两种局限性。CERTID使用健全且完整的因果识别算法ID来认证一个效应是否可以从给定的图和查询中识别，并验证返回的公式与结构因果模型相符，其干预分布是完全已知的。CERTID进一步发展理论结果，以减轻结构泄漏，修复不可识别的查询，并建立评分保证。我们在1200个认证实例上评估了三种前沿推理模型（Gemini Flash、Gemini Pro和GPT5.5），这些实例涵盖了4到50个顶点。准确性被证明是健全性的一个糟糕代理：在相同的实例上，关于不可识别查询的虚假声明率在模型之间变化达十七倍。我们还发现，模型在最强模型训练快照之后生成的图上以97-100%的准确率决定可识别性。实例、认证程序、验证器和每个实例的记录可在https://anonymous.4open.science/r/certid-D718获取。
cs.AI / 112 / 2610.03524
From Benchmarks to Production: A Text-to-SQL System for Complex Financial Data
从基准测试到生产：一个用于复杂金融数据的文本到SQL系统
Abstract
General-purpose Text-to-SQL systems achieve strong performance on academic benchmarks like Spider and BIRD, where schemas are relatively shallow and column values are often human readable. In production financial databases, where concepts are stored as opaque integer keys rather than human-readable strings, these methods fall below 50%, as even simple queries require multiple joins and filter predicates reference opaque IDs. We present Financial LINking Text-to-SQL (FLINT), a domain-specialized Text-to-SQL system that closes this gap through three key components: (1) a lookup agent that dynamically resolves natural-language concepts to question-specific reference table constraints, (2) embedding-based retrieval of structurally similar query templates from a compact, expert-authored bank, and (3) schema linking that prunes a large table schema to the relevant subset by traversing foreign-key chains, rather than relying on name similarity alone. We evaluate on two datasets totaling 359 questions over production financial schemas. FLINT outperforms various state-of-the-art baselines using the same LLM. The system is deployed in production as part of a financial data retrieval service.
Chinese Translation
通用的文本到SQL系统在Spider和BIRD等学术基准测试中表现出色，这些基准的模式相对简单，列值通常可被人类读取。然而，在生产金融数据库中，概念以不透明的整数键存储，而非人类可读的字符串，这些方法的表现低于50%，因为即使是简单查询也需要多个连接，并且过滤条件引用不透明的ID。我们提出了金融链接文本到SQL系统（Financial LINking Text-to-SQL，FLINT），这是一个领域专用的文本到SQL系统，通过三个关键组件来弥补这一差距：（1）一个查找代理，动态将自然语言概念解析为特定问题的参考表约束；（2）基于嵌入的检索，从一个紧凑的专家编写的模板库中获取结构相似的查询模板；（3）模式链接，通过遍历外键链来修剪大型表模式到相关子集，而不是仅依赖名称相似性。我们在两个数据集上进行了评估，总共359个问题，涵盖生产金融模式。FLINT在使用相同大型语言模型（LLM）的情况下，超越了多种最先进的基线。该系统作为金融数据检索服务的一部分在生产环境中部署。
cs.AI / 113 / 2610.03548
Recursive Harness Self-Improvement for Frontier Reasoning Data Synthesis
前沿推理数据合成的递归工具自我改进
Abstract
Generating progressively harder reasoning problems requires synthesis procedures that adapt as the task distribution evolves. Existing task-level recursion reuses generated problems as seeds but leaves the construction harness unchanged. We present task-harness co-evolution, a framework for recursive harness self-improvement (RSI) in reasoning-data synthesis. Online self-improvement converts intermediate solver failures into reusable skills during generation. Post-task self-improvement revises skills, prompts, and workflows after each batch, adopting candidates only when they generate harder valid tasks within a bounded cost increase. Model weights and verification criteria remain fixed. Across mathematics, coding, and science, mean solver accuracy decreases from 100.0% to 54.8% over fourteen evolution rounds. Ablations show that combining both update schedules produces harder tasks than fixed-harness recursion or either schedule alone. The resulting data improves downstream SFT and GRPO performance. In particular, a 27B student fine-tuned on 10K synthesized mathematics examples achieves 62.5% mean-16 accuracy on APEX, competitive with selected frontier-model references. These results support adapting the synthesis harness alongside the tasks to generate increasingly challenging data with downstream training value.
Chinese Translation
生成逐渐更难的推理问题需要在任务分布演变时适应的合成程序。现有的任务级递归将生成的问题作为种子进行重用，但构建工具保持不变。我们提出了任务-工具共同进化的框架，用于推理数据合成中的递归工具自我改进（Recursive Harness Self-Improvement, RSI）。在线自我改进在生成过程中将中间求解器失败转化为可重用的技能。任务后自我改进在每个批次后修订技能、提示和工作流程，仅在生成更难的有效任务且成本增加在可控范围内时采用候选项。模型权重和验证标准保持固定。在数学、编码和科学领域，平均求解器准确率在十四轮演化中从100.0%降至54.8%。消融实验表明，结合两种更新计划生成的任务比固定工具递归或单一计划生成的任务更具挑战性。所生成的数据提升了下游的SFT和GRPO性能。特别是，一个在10K合成数学示例上微调的27B学生模型在APEX上达到了62.5%的平均16准确率，具有与选定的前沿模型参考相媲美的竞争力。这些结果支持在生成越来越具挑战性的数据时，适应合成工具与任务的共同演变，以提高下游训练的价值。
cs.AI / 114 / 2610.03564
Knowledge or Calculator? Decomposing the Skill Premium in Verifiable Financial Agent Workflows
知识还是计算器？分解可验证金融代理工作流程中的技能溢价
Abstract
Financial AI agents must do more than retrieve facts: investment workflows require correct quantitative execution, reliable use of procedural resources, and auditable structured outputs. We introduce FinSkillBench, an evaluation suite of 2,603 point in time episodes across 12 subtasks in portfolio construction, risk management, and fundamental analysis, with hidden regenerable ground truth and task specific deterministic verifiers. Executing 17,820 episodes across 9 models and 3 resource conditions, the paired analysis across 8 models shows that curated skill packages raise mean scores by +16.2 points (0.366 to 0.528), whereas skills generated within a single episode add only +0.5 points while consuming more tokens and turns. We then decompose the curated premium by granting human authored procedural documents and executable domain tools separately: documents alone add +5.6 points, tools alone add +19.5 points, and their combination is subadditive. The premium is strongly workflow dependent: executable tools dominate numerically intensive workflows, documentation matters more when procedural or output schema guidance is the bottleneck, and interpretive tasks benefit from both. The effects are sign stable across 10 scoring variants and cluster bootstrap analyses, and an independently implemented second harness reproduces the directional pattern while showing that effect magnitudes depend on how tools and data are exposed. Overall, a measured "skill premium" is a property of the full model, resource, and harness system rather than of the underlying model alone.
Chinese Translation
金融人工智能代理必须做的不仅仅是检索事实：投资工作流程需要正确的定量执行、可靠的程序资源使用和可审计的结构化输出。我们引入了FinSkillBench，这是一个评估套件，包含在投资组合构建、风险管理和基本分析中的12个子任务下的2,603个时点事件，具有隐藏的可再生真实值和任务特定的确定性验证器。通过在9个模型和3种资源条件下执行17,820个事件，8个模型的配对分析表明，经过策划的技能包将平均分数提高了+16.2分（从0.366到0.528），而在单个事件中生成的技能仅增加+0.5分，同时消耗了更多的令牌和回合。然后，我们通过分别提供人类编写的程序文档和可执行的领域工具来分解策划的溢价：仅文档增加+5.6分，仅工具增加+19.5分，而它们的组合是次加性的。溢价在很大程度上依赖于工作流程：可执行工具在数值密集型工作流程中占主导地位，当程序或输出模式指导成为瓶颈时，文档更为重要，而解释性任务则从两者中受益。该效应在10种评分变体和聚类自助法分析中保持稳定，并且独立实施的第二个工具也再现了方向模式，同时显示效应的大小取决于工具和数据的暴露方式。总体而言，经过测量的“技能溢价”是完整模型、资源和工具系统的属性，而不仅仅是基础模型的属性。
cs.AI / 115 / 2610.03570
Learning to Assess Heartbeat Observability for mmWave Heart-Rate Sensing
学习评估毫米波心率传感的心跳可观测性
Abstract
Contactless heart-rate sensing with millimeter-wave (mmWave) radar requires assessing whether individual measurements support reliable estimation. We study learning to assess heartbeat observability, defined as the readability of the heartbeat component in an acquired phase spectrum, for selective heart-rate estimation. Coherent superposition of scatterer returns can suppress this component even under similar macroscopic observation geometry, motivating assessment directly from acquired measurements. To obtain training supervision across different observability conditions, we develop a controllable multi-scatterer frequency-modulated continuous-wave (FMCW) simulator. Agreement between the dominant heartbeat-band peak and the known heart rate provides an automatic observability label for each simulated measurement. We propose HEAR (Heartbeat Estimation with Assessed Reliability), a compact dual-task Transformer that jointly predicts an observability score and heart rate. Its input combines spectral magnitudes with frequencies relative to the respiration fundamental, providing context for respiratory harmonics. Trained solely on simulated observations, HEAR transfers zero-shot to two public real-world datasets collected at 60 and 120 GHz from 134 subjects. The same learned score supports selective prediction with both HEAR's own heart-rate head and multiple existing estimators. On the 120 GHz dataset, score-based selection reduces the HR head's mean absolute error from 17.9 BPM at full coverage to 1.6 BPM at 50% coverage. The complete pipeline achieves an end-to-end processing latency of 50.8 ms on an edge device. Project page: https://yuxuanhu9.github.io/HEAR/.
Chinese Translation
无接触的心率传感利用毫米波（mmWave）雷达需要评估个别测量是否支持可靠的估计。我们研究了学习评估心跳可观测性，定义为在获取的相位谱中心跳成分的可读性，以实现选择性的心率估计。散射体返回信号的相干叠加甚至可以在相似的宏观观察几何条件下抑制该成分，这促使我们直接从获取的测量中进行评估。为了在不同的可观测性条件下获得训练监督，我们开发了一个可控的多散射体频率调制连续波（FMCW）模拟器。主导心跳频段峰值与已知心率之间的一致性为每个模拟测量提供了自动可观测性标签。我们提出了HEAR（基于评估可靠性的心跳估计），这是一个紧凑的双任务Transformer，能够联合预测可观测性评分和心率。其输入结合了相对于呼吸基频的谱幅度和频率，为呼吸谐波提供上下文。HEAR仅在模拟观察上进行训练，能够零样本迁移到两个公共真实世界数据集，这些数据集在134名受试者中以60 GHz和120 GHz的频率收集。相同的学习评分支持使用HEAR自身的心率头和多个现有估计器进行选择性预测。在120 GHz数据集中，基于评分的选择将HR头的平均绝对误差从全覆盖下的17.9 BPM降低到50%覆盖下的1.6 BPM。完整的处理流程在边缘设备上实现了50.8毫秒的端到端处理延迟。项目页面：https://yuxuanhu9.github.io/HEAR/
cs.AI / 116 / 2610.03574
HyperBrowseComp: A Multilingual and Multimodal Stress Test for Web-Browsing Agents
HyperBrowseComp：一种多语言和多模态的网页浏览代理压力测试
Abstract
We introduce HyperBrowseComp, a multilingual and multimodal browsing benchmark comprising 423 manually authored and human-validated questions across 13 languages, written by native or highly proficient speakers. Questions are designed to be extremely challenging. Each question targets a concise, publicly verifiable answer whose discovery requires locating obscure evidence, following multi-step clue chains, or inspecting heterogeneous sources such as videos, scanned documents, images, or maps. Easier questions are filtered out by evaluating them with models without internet access to reduce the likelihood that they can be answered with parametric knowledge alone. We evaluate several models using provider-native search and a shared external retrieval harness under a common agent protocol. To contextualize model performance and effort, we also conduct a human evaluation on a sample of the questions. HyperBrowseComp provides a challenging testbed for persistent information seeking across languages and evidence modalities, with difficulty arising from discovering and connecting evidence on the open web.
Chinese Translation
我们介绍了HyperBrowseComp，这是一项多语言和多模态的浏览基准测试，包含423个手动撰写并经过人工验证的问题，涵盖13种语言，由母语或高水平使用者撰写。这些问题旨在极具挑战性。每个问题的目标是一个简洁、可公开验证的答案，其发现需要定位晦涩的证据、遵循多步骤的线索链，或检查异构来源，如视频、扫描文档、图像或地图。通过使用没有互联网访问的模型对问题进行评估，过滤掉较简单的问题，以降低仅凭参数知识回答的可能性。我们在一个共同的代理协议下，使用提供者本地搜索和共享外部检索工具评估多个模型。为了对模型的表现和努力程度进行背景化，我们还对一部分问题进行了人工评估。HyperBrowseComp为跨语言和证据模态的持久信息检索提供了一个具有挑战性的测试平台，难度源于在开放网络上发现和连接证据。
cs.AI / 117 / 2610.03591
HazardWeaver: Scientific Route Selection for Hazard Analysis Agents
HazardWeaver：危险分析代理的科学路线选择
Abstract
Understanding and assessing natural hazards is essential for disaster preparedness and risk reduction. Recent advances in large language models have spurred growing interest in AI agents for hazard analysis, particularly their ability to integrate scientific data, models, and tools into automated workflows. However, effective automation requires agents to determine which scientific methods are appropriate for a given event and executable with the available data and tools. As new evidence and execution results become available, these conditions can change, requiring agents to reconsider their choices. We formulate this problem as state-dependent scientific route selection and introduce HazardWeaver. Specifically, HazardWeaver first leverages the Hazard Knowledge Compiler to extract evidence-linked conditions governing scientific applicability, then its Hazard Capability Graph represents executable scientific capabilities and checks compatibility between their inputs and outputs. Using these complementary representations, the Hazard Weaver Agent component selects applicable and executable routes, carries out their workflows, and revises its decisions as the analysis state changes. To evaluate both the scientific outputs and the decisions that produce them, we introduce the Hazard Weaver Benchmark, comprising 141 instances across seven single-hazard domains and four multi-hazard interaction classes. The benchmark accommodates multiple valid scientific routes and evaluates output correctness, route validity, and justified abstention. Extensive experiments on this benchmark show that HazardWeaver outperforms existing agent systems, with the largest gains on tasks with multiple eligible scientific routes. Our code is publicly available at https://github.com/LabRAI/HazardWeaver.
Chinese Translation
理解和评估自然灾害对于灾难准备和风险降低至关重要。近期在大型语言模型方面的进展激发了对危险分析中人工智能代理的日益关注，特别是它们将科学数据、模型和工具整合到自动化工作流程中的能力。然而，有效的自动化要求代理能够确定哪些科学方法适用于特定事件，并能够使用可用的数据和工具执行。随着新证据和执行结果的出现，这些条件可能会发生变化，要求代理重新考虑其选择。我们将此问题表述为状态依赖的科学路线选择，并引入了HazardWeaver。具体而言，HazardWeaver首先利用危险知识编译器提取与证据相关的条件，以确定科学适用性，然后其危险能力图表示可执行的科学能力，并检查其输入和输出之间的兼容性。通过这些互补的表示，Hazard Weaver Agent组件选择适用和可执行的路线，执行其工作流程，并在分析状态变化时修正其决策。为了评估科学输出及其产生的决策，我们引入了Hazard Weaver基准，包括七个单一危险领域和四个多危险交互类别的141个实例。该基准支持多条有效的科学路线，并评估输出的正确性、路线的有效性以及合理的放弃。对该基准的广泛实验表明，HazardWeaver在现有代理系统中表现优异，尤其在具有多条合格科学路线的任务中取得了最大的提升。我们的代码已公开发布在 https://github.com/LabRAI/HazardWeaver。
cs.AI / 118 / 2610.03618
Low-Cost Video--Time Priors as a Strong Baseline for EEG--fNIRS Emotion Regression on Familiar Videos
低成本视频时间先验作为EEG-fNIRS情感回归在熟悉视频中的强基线
Abstract
Continuous emotion regression estimates moment-to-moment valence and arousal while a viewer watches a video. In familiar-video deployment, responses fron training participant-specific estimate, and prior-dominating fixed fusion tests whether physiology adds residual correction. In five-fold subject-held-out evaluation on 24was within 0.05 and 0.32 MAE of fusion in the internal and external evaluations, respectively. Source-explicit ablations showed that video identity and within-video tine accounted for most of the reduction, while EG-FNIRS gains were smaller and varied across participants and videos. These results identify the video-time prior as a strong, low-cost baseline and position EEG-fNIRS as an optional residual signal for familiar-video emotion regression.
Chinese Translation
连续情感回归在观众观看视频时估计瞬时的愉悦度和唤醒度。在熟悉视频的应用中，来自训练参与者特定估计的响应，以及主导固定融合的先验检验生理学是否增加了残差修正。在对24个参与者进行的五折交叉验证中，内部和外部评估的融合均在0.05和0.32的平均绝对误差（MAE）之内。源显式消融实验表明，视频身份和视频内时间占据了大部分减少的原因，而EG-FNIRS的增益较小，并且在参与者和视频之间有所不同。这些结果将视频时间先验确定为一个强大且低成本的基线，并将EEG-fNIRS定位为熟悉视频情感回归的可选残差信号。
cs.AI / 119 / 2610.03626
Depth as Time in One-Step Generative Models
一阶生成模型中的深度作为时间
Abstract
The recent wave of one-step generative models, which compress the multi-step trajectory of diffusion via either distillation or learned flow maps, has reached an inflection point where they can generate high-quality images. Here, we ask a natural question that follows from these advances: what happens to the denoising trajectory of multi-step diffusion when generation is compressed into a single forward pass? We offer an empirical observation we call \textit{depth as time}: the denoising computation that multi-step diffusion performs across sampling steps appears to unfold across the depth of a single forward pass, and can be recovered by decoding intermediate layers with the model's own output head. Most interestingly, we show that this depthwise computation depends on the transport task a flow map is trained to solve. The most surprising case is MeanFlow, where probing shorter transport intervals reveals both denoising and renoising within a single network evaluation. In contrast, generators trained without a time-indexed transport task, such as drifting models, do not exhibit the same depthwise denoising. Consequently, we show that models that exhibit the depthwise denoising phenomenon are more compressible across the layerwise computation: a MeanFlow \texttt{SiT-L/2} model can be compressed by $16.6\times$ in parameters into a single time-conditioned block. We offer an explanation for this denoise-then-renoise behavior and show that, when we treat the layerwise computation explicitly as a flow, a single time-conditioned block can be trained to denoise across layers, compressing a MeanFlow \texttt{SiT-L/2} model by $16.6\times$ in parameters. Together, these results suggest that the temporal computation of diffusion is not eliminated by one-step generation, but reorganized across network depth.
Chinese Translation
最近一波一阶生成模型通过蒸馏或学习的流动图压缩多步扩散轨迹，已达到一个拐点，能够生成高质量图像。在此，我们提出一个自然的问题：当生成被压缩为单次前向传播时，多步扩散的去噪轨迹会发生什么？我们提供了一个经验观察，称之为“深度作为时间”：多步扩散在采样步骤中执行的去噪计算似乎在单次前向传播的深度中展开，并且可以通过解码模型自身输出头的中间层来恢复。最有趣的是，我们展示了这种深度计算依赖于流动图被训练以解决的传输任务。最令人惊讶的情况是 MeanFlow，其中探测较短的传输间隔在单个网络评估中揭示了去噪和再噪声。相比之下，没有时间索引传输任务训练的生成器（如漂移模型）并未表现出相同的深度去噪。因此，我们展示了表现出深度去噪现象的模型在层级计算中更具可压缩性：一个 MeanFlow exttt{SiT-L/2} 模型可以在参数上压缩 $16.6 imes$ 到一个单一的时间条件块。我们为这种去噪后再噪声行为提供了解释，并展示了当我们将层级计算明确视为流动时，可以训练一个单一的时间条件块在层间去噪，从而将 MeanFlow exttt{SiT-L/2} 模型在参数上压缩 $16.6 imes$。这些结果共同表明，扩散的时间计算并未因一阶生成而消失，而是重新组织在网络深度中。
cs.AI / 120 / 2610.03631
NeutronGym: Physics-Graded Neutron Instrument Design for LLM Agents
NeutronGym：针对大型语言模型代理的物理分级中子仪器设计
Abstract
Designing a scientific instrument tests whether language-model agents can do physics rather than recall it, provided the grading cannot be argued with. We introduce NeutronGym, to our knowledge the first executable environment for neutron instrument design: agents build instruments through validating tools, McStas ray-traces what they build, and a level-resolved ladder grades syntax, runtime, structure and science with no LLM judge. Procedural families supply unlimited instances of a fixed layout whose design parameters the agent must set, with held-out parameter regimes; a curated slice, McStasBench, adds 16 tasks from published instruments behind memorization probes and a sandbox. Seven models reproduce at most 7 of the 16, none retrieves a reference, and none meets an improvement target. The environment also trains. Reinforcement learning on its reward takes Qwen3-8B from 11% to 77% of held-out instances of a family whose targets come from a hidden design (69% at a second seed), past an untrained Qwen3-32B, and the recipe holds, at one seed each, on three further gated families. The analysis says what that gain is. Without the ladder's partial credit it collapses by 60 points. From reward alone the trained model reaches what a classical optimizer reaches, at the agent's simulation budget, only when handed the closed-form physics (77% against 81%, a gap that does not separate at this size), while frontier models still solve 98-99%. Getting a trustworthy result meant failing four task designs that no-model baselines could solve, and we release the probes that found them.
Chinese Translation
设计科学仪器测试语言模型代理是否能够进行物理推理而非仅仅回忆，前提是分级不可争辩。我们介绍了NeutronGym，至今为止我们所知的第一个可执行的中子仪器设计环境：代理通过验证工具构建仪器，McStas对其构建进行光线追踪，且一个分级的梯子对语法、运行时、结构和科学进行评分，而无需大型语言模型（LLM）评审。程序族提供固定布局的无限实例，代理必须设置其设计参数，并且有保留的参数范围；一个策划的切片，McStasBench，增加了16个来自已发布仪器的任务，这些任务在记忆探测和沙盒中进行。七个模型最多重现16个任务中的7个，均未能检索到参考，且没有一个达到改进目标。该环境也用于训练。通过对其奖励进行强化学习，Qwen3-8B在一个隐藏设计的家族的保留实例中从11%提升至77%（在第二个种子下为69%），超过了未经过训练的Qwen3-32B，并且该配方在另外三个进一步的门控家族中各保持一个种子。分析表明了这种增益的具体情况。如果没有梯子的部分信用，该增益将减少60分。仅凭奖励，经过训练的模型在代理的仿真预算下，仅在获得封闭形式的物理时才能达到经典优化器的水平（77%对81%，在这个规模下没有显著差距），而前沿模型仍能解决98-99%的任务。获得可信结果意味着未能解决四个任务设计，而这些设计是无模型基线无法解决的，我们发布了发现它们的探测器。
cs.AI / 121 / 2610.03634
Credit Where It Matters: Dependency-Aware Policy Optimization for Terminal Agents
值得信赖的信用：面向终端代理的依赖感知策略优化
Abstract
Terminal-using agents benefit from reinforcement learning (RL) in coding, debugging, and other multi-step terminal tasks. In these tasks, later commands often depend on information or intermediate results produced by earlier commands. However, existing trajectory-level and step-level credit assignment methods do not explicitly trace the read-write dependencies through which commands affect the final outcome. Consequently, training signals could still be assigned to irrelevant operations, weakening learning from relevant steps. In this paper, we propose Dependency-Aware Group Policy Optimization (DepGPO), which uses execution dependencies between commands to guide credit assignment for terminal agents. Specifically, we construct a command dependency graph from execution traces and trace backward from the resources inspected by the task verifier. We then assign credit to relevant writes and their supporting reads along these paths, and use it to redistribute trajectory advantages across steps. Extensive comparative experiments and ablation studies demonstrate that DepGPO improves task performance and training stability on complex terminal tasks.
Chinese Translation
使用终端的代理在编码、调试和其他多步骤终端任务中受益于强化学习（RL）。在这些任务中，后续命令通常依赖于早期命令产生的信息或中间结果。然而，现有的轨迹级和步骤级信用分配方法并未明确追踪命令影响最终结果的读写依赖关系。因此，训练信号可能仍会被分配给无关操作，从而削弱了对相关步骤的学习。在本文中，我们提出了依赖感知组策略优化（Dependency-Aware Group Policy Optimization, DepGPO），该方法利用命令之间的执行依赖关系来指导终端代理的信用分配。具体而言，我们从执行轨迹构建命令依赖图，并从任务验证器检查的资源向后追踪。然后，我们沿着这些路径将信用分配给相关的写操作及其支持的读操作，并利用这些信用在步骤之间重新分配轨迹优势。大量的对比实验和消融研究表明，DepGPO在复杂终端任务上提高了任务性能和训练稳定性。
cs.AI / 122 / 2610.03639
Do Large Language Models Know Colombian Law? A Reliability Benchmark for the Colombian Legal System
大型语言模型了解哥伦比亚法律吗？哥伦比亚法律系统的可靠性基准
Abstract
Large language models (LLMs) are increasingly used to support legal practice, education, and research, yet their reliability in national legal systems outside the United States remains largely undocumented. We introduce an expert-validated benchmark for evaluating LLM reliability on the Colombian legal system. The benchmark comprises 1,042 items spanning ten areas of law and three question formats (closed multiple-choice, semi-open, and open-ended IRAC), built through a human-in-the-loop pipeline with multi-stage expert review. We evaluate 15 contemporary proprietary and open-weight models with format-appropriate metrics. Accuracy on closed questions ranges widely, from 0.905 (Gemini 3.1 Pro) to 0.577, but on free-text legal answers factual correctness never exceeds 0.45 (on a 0-1 scale) for any model. We find a dissociation between answer relevancy and correctness (Spearman rho = -0.46): models reliably sound responsive while frequently being wrong, a pattern of particular concern for non-expert users. Closed-question accuracy and free-text correctness are strongly rank-correlated (rho = 0.94), so cheap multiple-choice screening predicts model ranking but overstates absolute reliability. An independent rubric-based LLM judge and blind human expert scoring both reproduce the free-text ranking (rho >= 0.88). The judge further reveals that only about half of the norms models cite are correct; the rest are wrong or non-existent. Reliability varies systematically by legal area and follows an inverted-U across question complexity. Our results indicate that current LLMs require expert supervision for Colombian legal tasks, and that grounding answers in authoritative sources is a promising path to higher reliability. We release the benchmark construction pipeline to support reproducible evaluation.
Chinese Translation
大型语言模型（LLMs）越来越多地被用于支持法律实践、教育和研究，但它们在美国以外的国家法律系统中的可靠性仍然基本没有文献记录。我们引入了一个经过专家验证的基准，用于评估LLM在哥伦比亚法律系统中的可靠性。该基准包含1,042个项目，涵盖十个法律领域和三种问题格式（封闭式多项选择、半开放式和开放式IRAC），通过人机协作的流程和多阶段的专家审查构建。我们使用适合格式的指标评估了15个当代专有和开放权重模型。封闭式问题的准确率差异很大，从0.905（Gemini 3.1 Pro）到0.577，但在自由文本法律答案中，任何模型的事实正确性都不超过0.45（在0-1的范围内）。我们发现答案的相关性和正确性之间存在脱节（Spearman rho = -0.46）：模型在响应上可靠，但经常出错，这对非专家用户尤其令人担忧。封闭式问题的准确性和自由文本的正确性之间存在强烈的等级相关性（rho = 0.94），因此廉价的多项选择筛选可以预测模型排名，但夸大了绝对可靠性。一个独立的基于评分标准的LLM评估者和盲评人类专家评分均再现了自由文本的排名（rho >= 0.88）。评估者进一步揭示，模型引用的规范中只有约一半是正确的，其余则是错误或不存在的。可靠性在法律领域之间系统性变化，并随着问题复杂性的增加呈现倒U型。我们的结果表明，当前的LLM在处理哥伦比亚法律任务时需要专家监督，并且将答案基于权威来源是提高可靠性的一个有前景的途径。我们发布了基准构建流程，以支持可重复的评估。
cs.AI / 123 / 2610.03651
MRVQ: One Resident Index for Dimension- and Rate-Elastic Vector Search
MRVQ：一种适用于维度和速率弹性向量搜索的单一驻留索引
Abstract
Dense-retrieval services must switch among embedding-prefix dimensions and index bit rates as latency, quality, and memory budgets change. Tuning a quantizer separately for each rate gives the best quality, but the retrieval tier then holds several code streams and quantizer states at once. We introduce Matryoshka Residual Vector Quantization (MRVQ), a post-hoc residual quantizer for frozen embeddings. Its maximum-rate code can be truncated two ways: dropping residual stages lowers the rate, and dropping embedding coordinates lowers the dimension. One resident artifact therefore serves every (dimension, rate) pair we evaluate. Across FiQA and NFCorpus, four embedding families, and {4, 8, 16}-byte codes, MRVQ is the lowest-RAM design we evaluate. It uses 17.8-22.0x less memory than three separately trained QINCo2 indices, and 1.89-2.02x less than a lean shared-model steelman. The saving is not free: per-rate QINCo2 is 0.026-0.107 nDCG@10 better on FiQA. But MRVQ beats PQ, OPQ, and AdANNS-OPQ at matched code size. We also evaluate a low-build-cost PCA-scalar design that attains quality comparable to RaBitQ and its extension while fitting 420x faster at the median. Finally, we report two negative results: QINCo2 collapses when trained at high rates, and a ranking-bound hypothesis misses its pre-specified acceptance criteria. MRVQ is therefore a low-memory operating point for elastic retrieval, not a universal quality winner.
Chinese Translation
密集检索服务必须在嵌入前缀维度和索引比特率之间切换，以应对延迟、质量和内存预算的变化。为每个速率单独调优量化器可以获得最佳质量，但检索层会同时持有多个代码流和量化器状态。我们提出了马特ryoshka残差向量量化（MRVQ），这是一种针对冻结嵌入的事后残差量化器。其最大速率代码可以通过两种方式截断：去掉残差阶段降低速率，去掉嵌入坐标降低维度。因此，一个驻留工件可以服务于我们评估的每一个（维度，速率）对。在FiQA和NFCorpus上，四种嵌入系列以及{4, 8, 16}字节代码中，MRVQ是我们评估的最低内存设计。它使用的内存比三个单独训练的QINCo2索引少17.8-22.0倍，比一个精简共享模型的steelman少1.89-2.02倍。这种节省并不是免费的：每个速率的QINCo2在FiQA上的nDCG@10表现好0.026-0.107。但在匹配的代码大小下，MRVQ优于PQ、OPQ和AdANNS-OPQ。我们还评估了一种低构建成本的PCA标量设计，其质量可与RaBitQ及其扩展相媲美，同时在中位数时速度快420倍。最后，我们报告了两个负面结果：QINCo2在高速率下训练时崩溃，而一个排名界限假设未能满足其预设的接受标准。因此，MRVQ是弹性检索的低内存操作点，而不是普遍的质量赢家。
cs.AI / 124 / 2610.03693
Transcriptome-informed multi-modal AI for predicting neoadjuvant therapy response from breast cancer biopsies
基于转录组信息的多模态人工智能用于预测乳腺癌活检的新辅助治疗反应
Park, Jungkyu, Biswas, Dhruva, Cappadona, Joseph, Tang, Cerise, Zeng, Ken G., Machura, Bartosz, Liu, Chuwen, Tarantino, Paolo, Omene, Coral, Esteva, Francisco J., Bhargava, Rohit, Braun, Marcin, Paździerz, Kamila, Czerwiński, Jakub, Romańska-Knight, Hanna, Grinshpun, Albert, Daniel, Bareket, Buchinger, Michele, Howard, Frederick, Wysocki, Piotr, Chun, Brie, Schnabel, Freya, Caruana, Rich, Witowski, Jan, Geras, Krzysztof J.
Abstract
Scarcity of labeled data limits development of deep learning biomarkers in oncology. We develop a two-stage AI model predicting pathological complete response (pCR) to neoadjuvant therapy in breast cancer. The first stage learns the transcriptome from histopathology using 8,742 patients across 32 cancer types, corroborated by pathologist review and spatial agreement with measured expression. This simplifies the second stage to predicting pCR from inferred expression and clinical variables. Developed using 1,080 patients (five cohorts) and evaluated in 1,412 patients (nine cohorts), the model achieves a pooled AUROC of 0.79 (95% CI, 0.73-0.85), discriminating responders within molecular subtypes. It outperforms histopathological biomarkers, remaining stable across intratumoral sampling and with minimal biopsy tissue. Ablations show transcriptome-wide inference improves discrimination over clinical variables alone or one-stage pathology models, and robustness by avoiding genomic assays' gene selection constraints. These results indicate that biologically informed compression may generalize to data-sparse applications in precision oncology.
Chinese Translation
标记数据的稀缺限制了肿瘤学中深度学习生物标志物的发展。我们开发了一种两阶段的人工智能模型，用于预测乳腺癌新辅助治疗的病理完全反应（pCR）。第一阶段利用来自32种癌症类型的8,742名患者的组织病理学学习转录组，并通过病理学家审查和与测量表达的空间一致性进行验证。这简化了第二阶段的任务，即从推断的表达和临床变量中预测pCR。该模型使用1,080名患者（五个队列）进行开发，并在1,412名患者（九个队列）中进行评估，达到了0.79的汇总AUROC（95% CI，0.73-0.85），能够在分子亚型中区分反应者。它的表现优于组织病理学生物标志物，在肿瘤内取样和最小活检组织下保持稳定。消融实验表明，转录组范围的推断在区分能力上优于仅依赖临床变量或单阶段病理模型，并通过避免基因组检测的基因选择限制而增强了稳健性。这些结果表明，生物信息驱动的压缩可能在精准肿瘤学的数据稀缺应用中具有广泛的适用性。
cs.CL / 1 / 2610.02293
HakemBench: A Turkish Benchmark of Typed Decisions
HakemBench：土耳其类型决策基准
Abstract
HakemBench is a Turkish benchmark of typed decisions, in which the model under test reads a text, a question and a fixed set of options and returns a probability for every option. Version 1.0 is released fully open under CC BY 4.0, with 2,346 items and 4,275 choice, yes/no and score questions in seven tracks (fact-check triage, education, guardrails, legal routing, moderation, spam and phishing, and customer support). One harness scores decision quality (macro F1), calibration (from the normalised Brier score) and selective automation (from the normalised area under the generalised risk-coverage curve), combines them by a geometric mean and reports intervals from 2,000 bootstrap draws; probes for option order, paraphrase, English translation and substituted names are reported alongside. Most gold labels come from blind passes of one AI model family compared with the votes of a panel of large language models from other model families; they are not human-verified. On a board of 16 rows the leader scores a composite of 0.888 and the lab's own model is 7th at 0.660. Its numbers are not blind. Earlier runs' test results shaped its training data, so its guardrail, moderation and customer support numbers are flagged; with every model scored on the other four tracks only, its composite is 0.678, 6th of 16.
Chinese Translation
HakemBench 是一个土耳其类型决策基准，其中被测试的模型读取文本、问题和一组固定选项，并为每个选项返回一个概率。版本 1.0 完全开放发布，遵循 CC BY 4.0 许可，包含 2,346 个项目和 4,275 个选择、是/否和评分问题，涵盖七个领域（事实核查分流、教育、保护措施、法律路由、内容审核、垃圾邮件与网络钓鱼、客户支持）。一个评估工具对决策质量（宏观 F1）、校准（基于归一化的 Brier 分数）和选择性自动化（基于归一化的广义风险覆盖曲线下的面积）进行评分，通过几何平均结合这些指标，并报告来自 2,000 次自助抽样的区间；选项顺序、释义、英文翻译和替换名称的探测结果也一并报告。大多数金标准标签来自于一个 AI 模型家族的盲测，与来自其他模型家族的大型语言模型小组的投票进行比较；这些标签并未经过人工验证。在 16 行的评估板上，领先模型的综合得分为 0.888，而实验室自己的模型排名第七，得分为 0.660。其得分并非盲测。早期运行的测试结果影响了其训练数据，因此其保护措施、内容审核和客户支持的得分被标记；在仅对其他四个领域进行评分的情况下，其综合得分为 0.678，排名第六。
cs.CL / 2 / 2610.02425
Finding the Move Is Not Winning the Game: XiangqiBench for Closed-Loop Evaluation of LLM Agents
找到走法并不等于赢得比赛：用于大语言模型代理闭环评估的象棋基准（XiangqiBench）
Abstract
Static evaluations credit a language model for naming the right move, but an agent must carry a plan through to a verified outcome while an opponent responds. We introduce XiangqiBench, an executable benchmark that measures this difference in Chinese chess: starting from 119 tactical endgames with forced mates supported by engine or checks-only search, an LLM agent must deliver checkmate against an engine defender. An interactive REPL interface separates real moves, state queries, and forward simulation, and we record 8,568 multi-turn trajectories from 12 frontier LLMs under two observation protocols. Three signals that look like competence each overstate closed-loop success. (i) The Conversion Gap: models play the stored reference first move in 26.1\% of Sighted trials, yet only 13.9\% of these trials end in a win. (ii) The Consistency Gap: the leading model reaches 38.7\% pass@3 but only 5.9\% pass^3, winning all three trials on 7 of the 46 positions it ever wins. (iii) The Simulation Gap: 32.3\% of accepted simulation calls stop on an illegal move, and in 49.3\% of comparable cases the real defender replies differently from the line the agent simulated; self-authored rollouts check legality but cannot anticipate the opponent. Finding the move is not winning the game: agent evaluations should score closed-loop outcomes and report reliability alongside coverage.
Chinese Translation
静态评估将语言模型正确命名走法视为成功，但代理必须在对手的应对下实施计划并达到验证的结果。我们介绍了XiangqiBench，这是一个可执行的基准，测量中国象棋中的这一差异：从119个由引擎支持的强制将死战术残局开始，大语言模型（LLM）代理必须对抗引擎防守者实现将死。一个交互式REPL界面将真实走法、状态查询和前向模拟分开，我们记录了在两种观察协议下，12个前沿LLM的8,568个多回合轨迹。三种看似能力的信号都夸大了闭环成功的表现。(i) 转换差距：模型在26.1%的可见试验中执行存储的参考首步，但其中仅有13.9%的试验以胜利结束。(ii) 一致性差距：领先模型在pass@3中达到38.7%，但在pass^3中仅为5.9%，在其赢得的46个位置中，有7个位置在三次试验中均获胜。(iii) 模拟差距：32.3%的接受的模拟调用停在非法走法上，在49.3%的可比案例中，真实防守者的回应与代理模拟的线路不同；自我生成的回合检查合法性但无法预见对手。找到走法并不等于赢得比赛：代理评估应评分闭环结果，并报告可靠性与覆盖率。
cs.CL / 3 / 2610.02444
Counterexample Generation via Per-Theorem Symbolic Verifiers: When Imitation Hurts and Reinforcement Repairs
通过定理符号验证器生成反例：模仿的危害与强化的修复
Abstract
Large language models often solve a theorem forward yet fail to disprove a closely related false one: a falsification gap that supervised fine-tuning does not close and can actively worsen. We frame counterexample generation as constrained witness emission against a deterministic per-theorem Python verifier, and release SymCE, a corpus of 4,707 false undergraduate-algebra and real-analysis conjectures, each paired with executable verifiers. The verifier also serves as the reward function, making SymCE a training environment. Training Qwen3-4B with SFT followed by GRPO under this oracle reveals an imitation trap: counterexample-only SFT collapses true-theorem recognition from 0.27 to 0.00, while RLVR with a sparse outcome-only reward repairs this and exceeds the base, to 0.66. The collapse replicates across four seeds and on Gemma-3-4B. Sparse and dense rewards yield statistically indistinguishable in-domain success yet diverge by 33 points on a held-out calibration probe, a dissociation we trace to the partial-credit term. Our 4B model outperforms every evaluated 7B open-weights math specialist, remains competitive with six frontier commercial APIs, and transfers under unchanged prompting to GSM8K, MATH-500 and MMLU-college-math. A human audit of 177 verifier decisions finds 97.7% accuracy. Code, data, verifier modules and annotations: https://github.com/ce-rlvr/SymCE.
Chinese Translation
大型语言模型通常能够正向解决一个定理，但却未能驳斥一个密切相关的错误定理：这种伪造差距是监督微调无法弥补的，并且可能会主动恶化。我们将反例生成框架设定为针对确定性每定理的 Python 验证器的受限见证发射，并发布了 SymCE，一个包含 4,707 个错误的本科代数和实分析猜想的语料库，每个猜想都配有可执行的验证器。验证器还充当奖励函数，使得 SymCE 成为一个训练环境。在这个神谕下，使用 SFT 训练 Qwen3-4B，然后进行 GRPO，揭示了一个模仿陷阱：仅反例的 SFT 使得真实定理识别率从 0.27 降至 0.00，而使用稀疏结果奖励的 RLVR 修复了这一点，并超越了基础模型，达到了 0.66。这一崩溃在四个种子和 Gemma-3-4B 上得到了复制。稀疏和密集奖励在领域内成功率上统计上无显著差异，但在保留的校准探针上却相差 33 分，这一分离我们追溯到部分信用项。我们的 4B 模型在所有评估的 7B 开放权重数学专家中表现最佳，并与六个前沿商业 API 竞争，在不改变提示的情况下转移到 GSM8K、MATH-500 和 MMLU-college-math。对 177 个验证器决策的人类审计发现准确率为 97.7%。代码、数据、验证器模块和注释见：https://github.com/ce-rlvr/SymCE。
cs.CL / 4 / 2610.02455
FinDialogLens: Event Extraction over Multi-Party Dialogue for Missed-Trade Identification in Financial Chatrooms
FinDialogLens：用于金融聊天室中错过交易识别的多方对话事件提取
Abstract
Multi-party financial chatrooms are vital for sales-and-trading professionals, but their complexity makes manual recovery of missed trades infeasible: each Request for Quote (RFQ) is an event whose final price and trade outcome appear many messages after the RFQ-trigger message (the inquiry message), interleaved with concurrent RFQs from other participants. We cast this as event extraction (EE) over multi-party dialogue and present FinDialogLens, a hybrid LLM pipeline in which compact fine-tuned classifiers act as inference-time scaffolds: they detect RFQ-triggers and price/trade outcome metadata, an RFQ-Level Module segments per-event RFQ windows, and a Trade Engine fills argument roles. With GPT-4o, FinDialogLens reaches 92.1% and 94.3% accuracy on final price and trade outcome, respectively, outperforming full-chatroom CoT prompting methods; fine-tuned open-source LLMs with as few as 3B parameters achieve comparable performance with modest in-domain data. To make the LLM-based solution practical at scale, a difficulty-aware router balances cost and accuracy by allocating RFQs between a low-cost rule-based engine and the higher-performing LLM-powered Trade Engine, cutting LLM calls by 85% on final price while recovering half of the accuracy gap to FinDialogLens (GPT-4o), saving over $300/day at our 70,000-RFQ/day scale.
Chinese Translation
多方金融聊天室对销售和交易专业人士至关重要，但其复杂性使得手动恢复错过的交易变得不可行：每个报价请求（Request for Quote, RFQ）都是一个事件，其最终价格和交易结果在RFQ触发消息（询价消息）之后的多条消息中出现，并与其他参与者的并发RFQ交错。我们将此视为多方对话中的事件提取（Event Extraction, EE），并提出了FinDialogLens，一个混合的LLM（大语言模型）管道，其中紧凑的微调分类器作为推理时的支架：它们检测RFQ触发器和价格/交易结果元数据，一个RFQ级模块分割每个事件的RFQ窗口，而一个交易引擎则填充参数角色。使用GPT-4o，FinDialogLens在最终价格和交易结果上的准确率分别达到92.1%和94.3%，优于全聊天室的链式思维（Chain of Thought, CoT）提示方法；微调的开源LLM即使在仅有3B参数的情况下也能通过适度的领域内数据实现可比的性能。为了使基于LLM的解决方案在规模上切实可行，一个难度感知路由器通过在低成本的规则引擎和性能更高的LLM驱动交易引擎之间分配RFQ来平衡成本和准确性，在最终价格上减少了85%的LLM调用，同时弥补了与FinDialogLens（GPT-4o）之间一半的准确性差距，在我们每天处理70,000个RFQ的规模下节省了超过300美元的成本。
cs.CL / 5 / 2610.02460
CUEing User Simulators: Calibrated User Embeddings for Multi-Turn Benchmarking
用户模拟器的CUE：用于多轮基准测试的校准用户嵌入
Abstract
Recent benchmarks rely on user simulators to evaluate AI agents in multi-turn interaction. While existing simulation techniques demonstrate surface fidelity to human style and behavior, ecologically valid interactive benchmarking also requires alignment in when and how agents fail across simulated and real user populations. We find that existing simulators lack outcome calibration: agreement with observed success rates and failure patterns when real users interact with the same agent. We introduce Calibrated User Embeddings (CUE), a framework that both encodes observed sessions and samples continuous representations, then decodes them into persona commands to steer LLMs to act as user simulators without training. Through this, we evaluate user-conditioned replay of past sessions and aggregate metric agreement when sampling novel personas for the same tasks. On $\tau^2$-Bench, CUEd simulators commit fewer simulator-attributed errors and more faithfully reproduce real-user agent failure modes, aggregate success rates, and outcomes for specific task-user pairs than other persona-based simulation methods. These gains coexist with competitive user fidelity as measured using metrics established in prior work. After being fit to mostly customer support interactions, the same CUEd simulators generalize to document creation, math tutoring, and casual conversation, and remain effective across different simulator LLMs without CUE retraining.
Chinese Translation
近期的基准测试依赖于用户模拟器来评估AI代理在多轮交互中的表现。尽管现有的模拟技术在表面上展现出与人类风格和行为的相似性，但生态有效的交互基准测试还需要在代理失败的时机和方式上与模拟用户和真实用户群体之间保持一致。我们发现现有的模拟器缺乏结果校准：即在真实用户与同一代理交互时，成功率和失败模式与观察到的情况之间的符合度。我们提出了校准用户嵌入（Calibrated User Embeddings, CUE），这是一个框架，既编码观察到的会话，又采样连续表示，然后将其解码为个性命令，以引导大型语言模型（LLMs）作为用户模拟器，而无需训练。通过这一方法，我们评估了基于用户条件的过去会话重放，并在为相同任务采样新个性时汇总指标一致性。在$ au^2$-Bench上，CUEd模拟器比其他基于个性的模拟方法更少地犯下模拟器归因错误，并更忠实地再现真实用户代理的失败模式、汇总成功率和特定任务-用户对的结果。这些提升与使用先前工作中建立的指标测量的用户忠实度相竞争。在主要适用于客户支持交互后，同样的CUEd模拟器能够推广到文档创建、数学辅导和休闲对话，并在不同的模拟器LLMs中保持有效，而无需CUE重新训练。
cs.CL / 6 / 2610.02472
APDMem: Agent-Controlled Progressive Disclosure for Query-Adaptive Long-Term Memory
APDMem：用于查询自适应长期记忆的代理控制渐进式披露
Abstract
Personalized LLM assistants must recover sparse evidence from long conversation histories across queries of varying complexity. We introduce APDMem (Agent-controlled Progressive Disclosure Memory), a hierarchical long-term memory architecture that applies progressive disclosure to memory retrieval. Rather than relying on a flat memory store or fixed retrieval granularity, APDMem represents conversation history as four progressively detailed layers: thematic summaries, personalized key facts, turn-level evidence notes, and raw messages. At inference time, a controller applies progressive disclosure to the memory hierarchy: it first reads high-level summaries and drills into finer evidence only when needed. This creates an adaptive cost-fidelity trade-off: simple queries can terminate early, while complex temporal, multi-hop, or exact-evidence queries trigger deeper inspection. A note synthesizer converts retrieved evidence into a query-focused structure that consolidates facts, orders events, and flags contradictions before final answer generation. Experiments on LongMemEval show that APDMem achieves strong performance for long-context memory reasoning while accessing only 8% of the total conversations.
Chinese Translation
个性化的大型语言模型（LLM）助手必须从长时间的对话历史中恢复稀疏证据，这些对话历史涉及不同复杂度的查询。我们介绍了APDMem（代理控制渐进式披露记忆），这是一种分层的长期记忆架构，应用渐进式披露于记忆检索。APDMem并不依赖于平坦的记忆存储或固定的检索粒度，而是将对话历史表示为四个逐渐详细的层次：主题摘要、个性化关键事实、轮次级证据笔记和原始消息。在推理时，控制器对记忆层次应用渐进式披露：它首先读取高层摘要，仅在需要时深入到更细的证据。这创造了一个自适应的成本-保真度权衡：简单查询可以提前终止，而复杂的时间性、多跳或精确证据查询则触发更深入的检查。笔记合成器将检索到的证据转换为一个以查询为中心的结构，整合事实、排序事件，并在最终答案生成之前标记矛盾。在LongMemEval上的实验表明，APDMem在长上下文记忆推理方面表现出色，同时仅访问了总对话的8%。
cs.CL / 7 / 2610.02486
From Retrieval to Typed Decisions: Calibrated System One Models from Biomedical Sentence Encoders
从检索到类型化决策：来自生物医学句子编码器的校准系统一模型
Abstract
Typed decision models answer schema-constrained questions about a text in one forward pass and return probabilities meant to be thresholded. We ask whether biomedical sentence encoders trained for retrieval are good starting points for such models. We present SBERT2S1, which converts Sentence-Transformers encoders into bi-encoder, cross-head (C) and prior-fused residual (PFR) decision models, together with BIODECIDE, a biomedical typed-decision suite, and MEDLINE-S1, 243k training decisions derived from NLM indexing. Across six parent-retriever pairs, retrieval training improves zero-shot matching of content-bearing options. After fine-tuning, its effect depends on the head: across five pairs and three training-set sizes, retrieval training significantly helps PFR, which keeps the retrieval prior, in 10 of 15 comparisons, but helps C in one and hurts it in five. A matched grid of two heads and five training objectives shows that C outperforms PFR under every objective, and that the released RLCD recipe of open System One models trails cross-entropy by 2.5-3.0 points. The deficit stems mainly from its reward normalisation, which inflates the noisy score-function term 3.6-15-fold; an unbiased leave-one-out estimator recovers most of the gap. After temperature scaling, no objective is clearly better calibrated than cross-entropy. We release the code, the MEDLINE-S1 labels and a model.
Chinese Translation
类型化决策模型在一次前向传递中回答关于文本的模式约束问题，并返回旨在进行阈值处理的概率。我们探讨了为检索训练的生物医学句子编码器是否是此类模型的良好起点。我们提出了SBERT2S1，它将句子变换器编码器转换为双编码器、交叉头（C）和先验融合残差（PFR）决策模型，以及BIODECIDE，一个生物医学类型化决策套件，以及MEDLINE-S1，基于NLM索引衍生的243k训练决策。在六对父检索器中，检索训练改善了内容承载选项的零样本匹配。在微调后，其效果依赖于头部：在五对和三种训练集大小中，检索训练显著帮助保留检索先验的PFR，在15次比较中有10次，但对C的帮助为1次，且有5次造成了负面影响。一个匹配的两头和五个训练目标的网格显示，C在每个目标下均优于PFR，并且发布的开放系统一模型的RLCD配方比交叉熵低2.5-3.0分。该差距主要源于其奖励归一化，导致噪声评分函数项膨胀了3.6-15倍；一个无偏的留一法估计器恢复了大部分差距。经过温度缩放后，没有哪个目标的校准明显优于交叉熵。我们发布了代码、MEDLINE-S1标签和一个模型。
cs.CL / 8 / 2610.02529
A generative-informed neuro-symbolic framework for syntactic ambiguity resolution: Evidence from Arabic DPs
一种生成信息驱动的神经符号框架用于句法歧义解析：来自阿拉伯语名词短语的证据
Abstract
Syntactic ambiguity poses a persistent challenge for Arabic NLP, particularly in morphologically rich nominal constructions where multiple structu6ral interpretations may be compatible with the same surface sequence. This study proposes a generatively informed neuro-symbolic framework for resolving structural ambiguity in Modern Standard Arabic (MSA) DPs. The framework integrates generative syntactic notions with AraBERT by representing ambiguity as a candidate-based decision task in which linguistically motivated alternatives are explicitly constructed and evaluated through candidate-conditioned input representations. Findings indicate that the model achieved 96.88% accuracy, 95.92% macro-F1, 96.83% weighted F1, and 93.94% binary F1 on the unseen evaluation set. Class-level analysis revealed asymmetric performance, with recall of 99.71% for High/VP Attachment (N1) and 89.26% for Low/NP/Embedded Attachment (N2), indicating greater difficulty in recovering the embedded interpretation. The study concludes that formal syntactic representations can be operationalized within Transformer-based NLP as an explicit interface between linguistic structure and contextual neural modeling, providing a controlled and interpretable approach to Arabic syntactic ambiguity resolution and beyond.
Chinese Translation
句法歧义对阿拉伯语自然语言处理（NLP）构成了持续的挑战，特别是在形态丰富的名词结构中，多个结构解释可能与相同的表面序列兼容。本研究提出了一种生成信息驱动的神经符号框架，用于解析现代标准阿拉伯语（MSA）名词短语中的结构歧义。该框架将生成句法概念与AraBERT相结合，通过将歧义表示为基于候选的决策任务，明确构建和评估语言动机驱动的替代方案，并通过候选条件输入表示进行评估。研究结果表明，该模型在未见评估集上达到了96.88%的准确率、95.92%的宏观F1、96.83%的加权F1和93.94%的二元F1。类级分析揭示了不对称的表现，高/VP附加（N1）的召回率为99.71%，而低/NP/嵌入附加（N2）的召回率为89.26%，表明恢复嵌入解释的难度更大。研究得出结论，形式句法表示可以在基于Transformer的NLP中作为语言结构与上下文神经建模之间的显式接口进行操作，为阿拉伯语句法歧义解析及其他领域提供了一种可控且可解释的方法。
cs.CL / 9 / 2610.02549
Evaluating Multi-Dimensional Generalization of Large Language Models in Temporal Extraction Tasks
评估大语言模型在时间提取任务中的多维泛化能力
Abstract
Time and event expression extraction are fundamental temporal reasoning tasks, but the problem remains difficult due to annotation ambiguity, domain sensitivity, and unstable model behavior. Existing evaluations focus on in-domain performance, offering limited insight into reliability under distribution shifts. We evaluate multiple model configurations across families, architectures, and reasoning strategies over four dimensions of generalization, examining transfer from base performance, cross-dimensional correlations, and the effects of scale, architecture, and prompting. This provides a systematic study of how prompted LLMs generalize in time and event expression extraction tasks. We find that strong base-task performance generally predicts better generalization. However, this relationship weakens under substantial distribution shifts. Inductive prompting performs most consistently across domain shift, adversarial perturbations, compositionality, and length increase, while gains from scale, architecture, and deductive and abductive prompting strategies are uneven and dimension-specific. We conclude that LLM generalization in temporal extraction tasks cannot be predicted from any single dimension alone and cannot be reliably inferred from in-domain or single-dimension evaluations, highlighting the need for reasoning strategies that generalize across dimensions.
Chinese Translation
时间和事件表达提取是基本的时间推理任务，但由于标注模糊性、领域敏感性和模型行为不稳定等问题，这一任务仍然困难。现有评估主要集中在领域内性能上，提供的对分布变化下可靠性的洞察有限。我们评估了多个模型配置，包括不同的模型家族、架构和推理策略，从四个维度探讨泛化能力，考察基础性能的迁移、跨维度相关性，以及规模、架构和提示的影响。这为提示的 LLM（Large Language Models）在时间和事件表达提取任务中的泛化能力提供了系统研究。我们发现，强大的基础任务性能通常预示着更好的泛化能力。然而，在显著的分布变化下，这一关系减弱。归纳提示在领域转移、对抗扰动、组合性和长度增加等方面表现最为一致，而规模、架构以及演绎和溯因提示策略带来的收益则不均匀且特定于维度。我们得出结论，LLM 在时间提取任务中的泛化能力不能仅凭单一维度进行预测，也无法从领域内或单维度评估中可靠推断，这突显了需要跨维度泛化的推理策略。
cs.CL / 10 / 2610.02612
Learning When to Commit from Partial Speech for End-to-End Simultaneous Speech Translation
从部分语音中学习何时进行承诺以实现端到端的同步语音翻译
Abstract
Simultaneous speech translation must emit useful target text before the source is complete while preserving every committed token. We adapt a full-utterance speech language model using prefix supervision derived from its own complete- and partial-waveform translations, requiring neither transcripts nor human translations. We compare single-turn forced-prefix and multi-turn append-only decoding, use a confidence threshold to control the inference-time quality--latency trade-off, and vary the density of training prefixes with a separate synthesis margin. On FLEURS and CoVoST2 in three language directions, prefix training improves quality--latency frontiers over the unadapted model, and confidence provides the broadest consistently competitive operating range. Multi-turn decoding is generally stronger at low latency; under multi-turn training, commit-calibration error falls by 63--68% overall and 68--80% at early prefixes, whereas single-turn training provides only modest overall calibration gains and no early-prefix improvement. A small synthesis margin sometimes extends the frontier to lower latency, particularly on shorter utterances, while a larger margin degrades translation quality and calibration. Prefix adaptation therefore improves simultaneous speech translation, especially under multi-turn append-only decoding, while synthesis density introduces a non-monotonic quality--latency trade-off.
Chinese Translation
同步语音翻译必须在源语音未完成之前发出有用的目标文本，同时保留每个已承诺的词元。我们采用了一种完整话语的语音语言模型，利用从其自身的完整和部分波形翻译中获得的前缀监督，不需要转录文本或人工翻译。我们比较了单轮强制前缀和多轮仅附加解码，使用置信度阈值来控制推理时的质量与延迟之间的权衡，并通过单独的合成边际来变化训练前缀的密度。在 FLEURS 和 CoVoST2 的三个语言方向上，前缀训练改善了未适应模型的质量-延迟边界，而置信度提供了最广泛的一致竞争操作范围。多轮解码在低延迟下通常更强；在多轮训练下，承诺校准误差整体下降了 63% 到 68%，而在早期前缀下下降了 68% 到 80%，而单轮训练仅提供了适度的整体校准增益，并且没有早期前缀的改善。较小的合成边际有时会将边界扩展到更低的延迟，特别是在较短的话语上，而较大的边际则会降低翻译质量和校准。因此，前缀适应改善了同步语音翻译，特别是在多轮仅附加解码下，而合成密度引入了非单调的质量-延迟权衡。
cs.CL / 11 / 2610.02665
Large Language Continuous Diffusion Models
大规模语言连续扩散模型
Abstract
Despite the success of discrete diffusion language models (dLMs) for fast parallel decoding, their non-smooth, high-dimensional space hinders trajectory steering for reasoning and inference acceleration. To overcome this, we present Sigma, the first large-scale (3B/8B) continuous dLM built on steerable, low-dimensional ODE/SDE latent trajectories. Trained blockwise via likelihood optimization, Sigma jointly denoises Gaussian-corrupted token embeddings while learning an optimal embedding geometry. To accelerate training, Sigma leverages pre-trained weights from autoregressive (AR) models for warm-starting. During inference, we identify classifier-free guidance and score temperature as essential for high-fidelity reasoning and coding. Across comprehensive math reasoning and coding evaluations against state-of-the-art discrete counterparts (masked dLMs and AR baselines), Sigma achieves competitive performance with discrete models on standard benchmarks (e.g., GSM8K, Minerva, HumanEval, MBPP) after pre-training and on challenging reasoning tasks (e.g., MATH-500, AIME) after supervised fine-tuning. Beyond performance parity, we uncover key structural properties unique to continuous dLMs: (i) embedding-space steering effectively governs the quality-diversity trade-off, yielding strong pass@k performance and (ii) continuous trajectories enable graceful degradation for low NFEs and efficient distillation. These establish continuous dLMs as a promising paradigm for efficient language generation.
Chinese Translation
尽管离散扩散语言模型（dLMs）在快速并行解码方面取得了成功，但其非平滑的高维空间阻碍了推理和推断加速的轨迹引导。为了解决这个问题，我们提出了Sigma，这是第一个基于可引导的低维ODE/SDE潜在轨迹的大规模（3B/8B）连续dLM。Sigma通过似然优化进行分块训练，同时对高斯噪声污染的标记嵌入进行去噪，并学习最佳的嵌入几何结构。为了加速训练，Sigma利用自回归（AR）模型的预训练权重进行热启动。在推理过程中，我们发现无分类器引导和评分温度对于高保真推理和编码至关重要。在与最先进的离散模型（如掩码dLMs和AR基线）进行的全面数学推理和编码评估中，Sigma在标准基准（如GSM8K、Minerva、HumanEval、MBPP）上经过预训练后与离散模型表现出竞争力，并在挑战性的推理任务（如MATH-500、AIME）上经过监督微调后也表现良好。除了性能平衡外，我们还发现了连续dLMs独特的关键结构特性：（i）嵌入空间引导有效地控制质量与多样性的权衡，产生强大的pass@k性能；（ii）连续轨迹使得在低NFE下优雅降级和高效蒸馏成为可能。这些特性确立了连续dLMs作为高效语言生成的有前景范式。
cs.CL / 12 / 2610.02702
Silent Dissent: LLM Agents That Yield to the Majority Still Represent Their Original Premise
沉默的异议：屈从于多数的LLM代理仍然代表其原始前提
Abstract
Multi-agent debate is increasingly used to reach consensus among LLM agents, yet agents often yield to a unanimous majority. When an agent changes its answer, has it changed its mind or only its statement? We study this with two-hop factual questions whose intermediate entity (the bridge, e.g. the country in "the capital of the country where the Sagrada Familia is located") is never stated by anyone. Scripted peers, in the role of Asch's confederates, unanimously assert a wrong answer taken from another fact with a different bridge. At the moment the agent answers, we read the bridge from its residual stream with the Jacobian lens (J-lens) and, for comparison, the logit lens. In pre-registered tests on held-out facts with four open-weight models, agents of Qwen3.5-4B, Qwen3.6-27B and Gemma-4-E4B-it that gave in still represented their original bridge in the pre-registered layers below the output (hit@100 above a control entity: 0.85, 0.22 and 0.24), where the logit lens rarely ranked it among the top 100 tokens (0.00-0.06). These agents also represented the bridge behind the peers' answer, beyond a mention baseline. A pre-registered addendum hid the agent's earlier answer or removed it: agents that gave in still represented their original bridge in all four models (0.43, 0.29, 0.37 and 0.25 with the answer hidden), including Llama-3.1-8B-Instruct, which barely did so with its answer in view (0.03). The premise can thus be computed from the question alone while the agent states the majority's answer. Hiding the earlier answer also changed conformity: Qwen3.5-4B gave in on 89% of questions instead of 8%. In exploratory interventions, injecting the bridge's J-lens direction brought agents back to their original answer only in the two Qwen models. Stated consensus in multi-agent debate can thus overstate agreement. We also report the negative results of our pre-registered program.
Chinese Translation
多代理辩论越来越多地被用于在LLM代理之间达成共识，然而代理通常会屈从于一致的多数。当一个代理改变其答案时，它是改变了自己的想法还是仅仅改变了表述？我们通过两步事实问题进行研究，其中中间实体（桥梁，例如在“位于圣家堂的国家的首都”中提到的国家）从未被任何人提及。脚本化的同伴在阿希（Asch）的同盟者角色中一致地声称一个错误的答案，该答案来自另一个具有不同桥梁的事实。在代理回答的瞬间，我们通过其残余流读取桥梁，使用雅可比视角（Jacobian lens，J-lens）以及对比的对数视角（logit lens）。在对四个开放权重模型的预注册测试中，屈从的Qwen3.5-4B、Qwen3.6-27B和Gemma-4-E4B-it代理在输出下的预注册层中仍然代表其原始桥梁（在控制实体上hit@100：0.85、0.22和0.24），而对数视角很少将其排名前100个标记之一（0.00-0.06）。这些代理还代表了同伴答案背后的桥梁，超出了提及基线。一个预注册的附录隐藏了代理的早期答案或将其移除：屈从的代理在所有四个模型中仍然代表其原始桥梁（隐藏答案时分别为0.43、0.29、0.37和0.25），包括Llama-3.1-8B-Instruct，后者在答案可见时几乎没有做到这一点（0.03）。因此，前提可以仅通过问题计算，而代理则陈述了多数的答案。隐藏早期答案也改变了从众行为：Qwen3.5-4B在89%的问题上屈从，而不是8%。在探索性干预中，注入桥梁的J-lens方向使代理仅在两个Qwen模型中回到其原始答案。因此，多代理辩论中陈述的共识可能会夸大一致性。我们还报告了预注册程序的负面结果。
cs.CL / 13 / 2610.02713
WakeKV: Reactive, Reversible KV Residency for Heads That Change Their Minds
WakeKV：对改变主意的注意力头的反应式、可逆KV驻留
Abstract
Most KV-cache compression methods classify attention heads once, either offline or during prefill, and keep this classification fixed throughout generation. Across three models (1.5B-8B) and three regimes (needle retrieval, long chain-of-thought, and multi-turn recall), we measure head behavior on four model-regime combinations and find that most heads change their reading behavior at least once during generation. We introduce WakeKV, a reactive residency policy that moves cooling heads to a recoverable CPU reservoir rather than freezing or permanently evicting their state. At matched memory or budget, WakeKV consistently improves miss rate over frozen classification and destructive eviction, evaluated across five model-regime combinations and over three cited baselines (SnapKV, uniform R-KV, and ReasonAlloc) across four eligible combinations. A FlexiCache/vLLM implementation on Mistral-7B confirms the benefit on real hardware, improving throughput while retaining LongBench quality.
Chinese Translation
大多数KV缓存压缩方法在生成过程中仅对注意力头进行一次分类，通常是在离线或预填充阶段，并在整个生成过程中保持这一分类不变。在三个模型（1.5B-8B）和三种模式（针式检索、长链思维和多轮回忆）下，我们测量了四种模型-模式组合中的头部行为，发现大多数头部在生成过程中至少改变了一次其读取行为。我们提出了WakeKV，一种反应式驻留策略，它将冷却的头部移动到可恢复的CPU储备中，而不是冻结或永久驱逐其状态。在相同的内存或预算条件下，WakeKV在五种模型-模式组合和三个引用基线（SnapKV、均匀R-KV和ReasonAlloc）下的四种合格组合中，始终提高了未命中率。Mistral-7B上的FlexiCache/vLLM实现验证了在真实硬件上的优势，提高了吞吐量，同时保持了LongBench的质量。
cs.CL / 14 / 2610.02736
TPBench: A Turning-Point Benchmark for Dialogue Compression
TPBench：对话压缩的转折点基准
Abstract
A compressor can keep the facts of a dialogue and still drop the turn that changed them. A user corrects a price, reverses a choice, or adds a constraint. We call this failure turning-point eviction. One overall retention score hides it, because that score mixes what the user first wanted with what the user wants now. We introduce TPBench, which evaluates three complementary information targets at shared nominal retention budgets. P1 asks for the user's initial goal. P2 asks for the current value of a slot the user revised. P3 asks for both, in dialogues with a late annotated slot update. The current-value answers come from the human dialogue-state annotations of MultiWOZ and SGD. The initial-goal answer is the first sentence of the first user turn. Neither requires new crowdsourcing. The probe-specific evaluations rank compression methods differently. On the joint probe at a retained fraction of 0.30, every tested compressed method remains below full context with the main Llama reader. Deleting the turn that carries the update sharply lowers current-value accuracy, while deleting one matched irrelevant turn leaves it unchanged. A Mistral reader repeats the P2/P3 rankings and the joint-probe gap. Current-value recovery is tested on an additional corpus, LongMemEval-KU, and on Chinese RiSAWOZ: full context has the highest accuracy, and recency has the highest compressed-method mean in both evaluations.
Chinese Translation
压缩器可以保留对话的事实，同时丢弃改变这些事实的回合。用户更正价格、撤回选择或添加约束。我们称这种失败为转折点驱逐。一个总体保留评分掩盖了这一点，因为该评分混合了用户最初的需求和用户现在的需求。我们引入了TPBench，它在共享的名义保留预算下评估三个互补的信息目标。P1询问用户的初始目标。P2询问用户修订的槽位的当前值。P3在带有延迟注释槽位更新的对话中同时询问这两者。当前值的答案来自MultiWOZ和SGD的人类对话状态注释。初始目标的答案是用户第一次发言的第一句话。两者都不需要新的众包。探针特定的评估对压缩方法的排名不同。在保留比例为0.30的联合探针上，所有测试的压缩方法在主要的Llama阅读器下均低于完整上下文。删除携带更新的回合会显著降低当前值的准确性，而删除一个匹配的无关回合则不会改变其准确性。Mistral阅读器重复了P2/P3的排名和联合探针的差距。当前值恢复在额外语料库LongMemEval-KU和中文RiSAWOZ上进行了测试：完整上下文具有最高的准确性，而近期性在两项评估中具有最高的压缩方法均值。
cs.CL / 15 / 2610.02739
Beyond Correctness: Resolving Underspecification in Agentic Text-to-SQL
超越正确性：解决代理文本到SQL中的不完全规范问题
Abstract
Agentic Text-to-SQL systems can interact with users to clarify underspecified queries before generating SQL. However, a correct execution result does not necessarily imply that the agent has adequately resolved the underlying underspecification: the agent may silently make unverified assumptions that happen to match the intended answer. We show that this behavior is driven in part by premature clarification termination. Although forcing an agent to ask more questions improves execution accuracy, ambiguities are concentrated in earlier interactions, making brute-force questioning inefficient. More importantly, even when explicitly prompted to plan its clarification process, the agent frequently abandons questions that it has already identified as relevant. To address this failure mode, we introduce PlanPool, which externalizes the clarification plan as a mutable question pool. Every planned question must be explicitly asked or dropped before submission, while newly discovered ambiguities can be added during interaction. Across three benchmarks derived from BIRD-Interact and Spider, PlanPool consistently improves ambiguity coverage and reduces silent failures over unconstrained and prompt-based alternatives, while maintaining competitive execution accuracy. Our results highlight an important distinction in agentic reasoning: identifying missing information is not sufficient, and the agent must also reliably maintain and resolve it before committing to an answer.
Chinese Translation
代理文本到SQL系统可以与用户互动，以澄清不完全规范的查询，然后生成SQL。然而，正确的执行结果并不一定意味着代理已充分解决潜在的不完全规范：代理可能默默地做出未经过验证的假设，而这些假设恰好与预期答案相匹配。我们展示了这种行为部分是由过早终止澄清引起的。虽然迫使代理提出更多问题可以提高执行准确性，但模糊性集中在早期互动中，使得强行提问效率低下。更重要的是，即使在明确要求规划澄清过程时，代理也常常放弃已识别为相关的问题。为了解决这一失败模式，我们引入了PlanPool，它将澄清计划外部化为可变的问题池。每个计划中的问题必须在提交之前明确提出或放弃，而在互动过程中可以添加新发现的模糊性。在基于BIRD-Interact和Spider的三个基准测试中，PlanPool始终提高了模糊性覆盖率，并减少了无声失败，相较于不受限制和基于提示的替代方案，同时保持了竞争性的执行准确性。我们的结果强调了代理推理中的一个重要区别：识别缺失信息是不够的，代理还必须可靠地维护和解决这些信息，然后才能做出回答。
cs.CL / 16 / 2610.02744
EpiWorld: Grounding LLM Policy Agents in Epidemiological World Models
EpiWorld：将大语言模型政策代理嵌入流行病学世界模型
Abstract
Epidemic intervention policies are textual artefacts that human decision-makers interpret, justify, and revise through natural language, making large language models a natural candidate for epidemic policy reasoning. A naive LLM, however, lacks the epidemic dynamics needed to project intervention consequences, the quantitative surveillance signals required to assess severity, and the institutional constraints that define admissible actions. We present EpiWorld, a closed-loop framework that grounds an LLM policy actor in a learned action-conditioned epidemiological world model and a tiered skill library of public-health protocols, surveillance tools, and adaptive lessons accumulated through after-action analysis. Given a candidate intervention, the world model predicts regional epidemic evolution and enables fast counterfactual rollouts that provide feedback for policy selection and refinement. Outcomes of simulated futures are distilled into reusable lessons while protocol constraints remain fixed, allowing the decision process to improve without sacrificing interpretability or controllability. We evaluate both the world model and the end-to-end framework on retrospective COVID-19 and Influenza datasets: the world model achieves the best out-of-distribution Peak-MAE among all forecasting baselines, and the closed-loop framework reduces cumulative hospitalisation by up to 59% across datasets and by an average of ~16% across six LLM backbones, outperforming reinforcement-learning and optimal-control policy baselines.
Chinese Translation
流行病干预政策是人类决策者通过自然语言进行解释、辩护和修订的文本产物，因此大语言模型（LLM）成为流行病政策推理的自然选择。然而，简单的LLM缺乏预测干预后果所需的流行病动态、评估严重性所需的定量监测信号，以及定义可接受行动的制度约束。我们提出了EpiWorld，这是一个闭环框架，将LLM政策代理嵌入到一个学习的基于行动条件的流行病学世界模型和一个分层的公共卫生协议、监测工具和通过事后分析积累的适应性经验教训的技能库中。在给定候选干预的情况下，世界模型预测区域流行病的发展，并能够快速进行反事实模拟，以提供政策选择和改进的反馈。模拟未来的结果被提炼为可重复使用的经验教训，同时协议约束保持不变，使决策过程能够在不牺牲可解释性或可控性的情况下得到改善。我们在回顾性COVID-19和流感数据集上评估了世界模型和端到端框架：世界模型在所有预测基线中实现了最佳的离散峰值平均绝对误差（Peak-MAE），而闭环框架在各数据集上将累计住院率降低了多达59%，在六个LLM基础模型上平均降低约16%，超越了强化学习和最优控制政策基线。
cs.CL / 17 / 2610.02769
When History Fails to Become Experience: Action Calibration in Language Agents
当历史未能转化为经验：语言代理中的行动校准
Abstract
Language agents should draw on prior attempts and environmental feedback to improve subsequent decisions within the same task. However, providing additional interaction history can sometimes reduce task success, suggesting that agents do not consistently use this information effectively. To investigate this limitation, we examine how agents use history. We find that history improves task completion overall, yet much of this benefit persists even when past actions are shuffled. Disrupting the correspondence between actions and observations causes only a modest decline in task success. We therefore hypothesize that agents do not reliably connect past actions with their outcomes when deciding how to proceed. To test this hypothesis, we explicitly label each returned observation as the outcome of the preceding action. This simple annotation improves task success and reduces next-action repetition without introducing new environmental information. Building on this insight, we introduce a learned calibrator that explicitly reassesses past actions and selectively records experience to guide subsequent decisions, improving task success beyond outcome labeling alone.
Chinese Translation
语言代理应当借鉴先前的尝试和环境反馈，以改善在同一任务中的后续决策。然而，提供额外的交互历史有时会降低任务成功率，这表明代理并未始终有效地利用这些信息。为了探讨这一局限性，我们研究了代理如何使用历史。我们发现，历史总体上提高了任务完成率，但即使在过去的行动被打乱时，这一好处仍然存在。破坏行动与观察之间的对应关系仅导致任务成功率的适度下降。因此，我们假设代理在决定如何继续时并未可靠地将过去的行动与其结果联系起来。为了验证这一假设，我们明确标记每个返回观察为前一行动的结果。这一简单的注释提高了任务成功率，并减少了下一行动的重复，而不引入新的环境信息。在此基础上，我们引入了一种学习校准器，该校准器明确重新评估过去的行动，并选择性地记录经验以指导后续决策，从而在仅依赖结果标记的基础上进一步提高任务成功率。
cs.CL / 18 / 2610.02770
AptMQL-Bench: From Text-to-SQL to Text-to-MQL via Access-Pattern Schema Design and Data-Preserving Migration
AptMQL-Bench：通过访问模式架构设计和数据保留迁移实现从文本到SQL再到文本到MQL的转变
Abstract
Document databases such as MongoDB are core infrastructure for modern applications, and natural-language interfaces to them---text-to-MQL---would let non-experts query complex, semi-structured data without mastering the query language. Progress on this task depends on high-quality benchmarks, which are most practically obtained by converting an existing text-to-SQL benchmark to the document setting. Unfortunately, existing efforts rely on heuristics for mechanical conversion: the document schema mirrors the relational foreign-key graph, and each query mirrors its source SQL. As a result in our experiments, these approaches fail to migrate 6 of 21 BIRD databases outright, silently drop up to 25.9\% of rows on others, and yield schemas whose ground-truth queries run over an order of magnitude slower as the data scales. We instead propose a conversion pipeline, driven by coding agents with human-in-the-loop verification, that designs each document schema from expected access patterns and rewrites queries to be MongoDB-native. Applying it to BIRD, we build an access-pattern-based text-to-MQL benchmark (AptMQL-Bench). It includes 21 document-oriented databases, 3,186 natural-language requests, and their associated MQL queries---whose databases are migrated from SQLite without data loss and scale efficiently. The strongest model, Claude Opus 4.5, achieves only 57.38\% accuracy without external knowledge evidence and 70.34\% with it. This indicates that realistic text-to-MQL generation remains challenging.
Chinese Translation
文档数据库如MongoDB是现代应用程序的核心基础设施，自然语言接口---文本到MQL---使非专家能够查询复杂的半结构化数据，而无需掌握查询语言。该任务的进展依赖于高质量的基准测试，这通常通过将现有的文本到SQL基准转换为文档设置来获得。不幸的是，现有的努力依赖于启发式的机械转换：文档架构镜像关系型外键图，每个查询镜像其源SQL。因此，在我们的实验中，这些方法在21个BIRD数据库中直接迁移失败6个，在其他数据库中默默丢弃多达25.9%的行，并且随着数据规模的扩大，生成的架构的真实查询运行速度慢了一个数量级。我们提出了一种转换管道，由编码代理驱动，并通过人类验证进行验证，从预期的访问模式设计每个文档架构，并重写查询以使其成为MongoDB原生。将其应用于BIRD，我们构建了一个基于访问模式的文本到MQL基准（AptMQL-Bench）。它包括21个面向文档的数据库，3,186个自然语言请求及其相关的MQL查询---这些数据库从SQLite迁移而来，且没有数据丢失，并且能够高效扩展。最强的模型Claude Opus 4.5在没有外部知识证据的情况下仅达到57.38%的准确率，而在有外部知识证据的情况下达到70.34%。这表明，现实中的文本到MQL生成仍然具有挑战性。
cs.CL / 19 / 2610.02772
Improving Atomic-Fact Recall via Focused Views in Unstructured Knowledge Editing
通过聚焦视图改善非结构化知识编辑中的原子事实回忆
Abstract
Large language models (LLMs) increasingly serve as general-purpose interfaces to factual knowledge, but their parameters do not automatically reflect information that changes after pretraining. Knowledge editing (KE) provides a targeted alternative to costly retraining by modifying selected knowledge and preserving unrelated knowledge and general capabilities. Conventional KE uses structured factual triples, whereas unstructured KE (UKE) uses free-form passages containing multiple facts. Nonetheless, existing UKE editors exhibit a failure mode known as context reliance: edited LLMs can often reproduce the editing passage but fail to reliably recall its individual facts without the original passage context. We identify context-induced difficulty underestimation under the standard passage-level editing objective: later facts receive increasingly rich ground-truth context and consequently incur lower initial losses, making them appear easier to learn. In response, we propose FOVEATED, a plug-and-play framework that constructs focused views of each sentence by randomly shifting the Rotary Position Embedding (RoPE) positions assigned to the keys of its preceding context. The perturbation is applied during editing and removed afterward, leaving the model's native positional encoding unchanged at inference time. We instantiate FOVEATED for both direct-optimization and locate-then-edit editors. We theoretically analyze how FOVEATED counteracts context-induced difficulty underestimation and empirically demonstrate consistent improvements across five KE editors, two LLM backbones, and three benchmarks.
Chinese Translation
大型语言模型（LLMs）越来越多地作为通用接口来获取事实知识，但它们的参数并不会自动反映预训练后发生变化的信息。知识编辑（KE）提供了一种有针对性的替代方案，通过修改选定的知识并保留无关知识和一般能力，避免了昂贵的再训练。传统的知识编辑使用结构化的事实三元组，而非结构化知识编辑（UKE）则使用包含多个事实的自由形式段落。然而，现有的UKE编辑器表现出一种称为上下文依赖的失效模式：编辑后的LLM通常能够重现编辑段落，但在没有原始段落上下文的情况下，无法可靠地回忆其单个事实。我们识别出在标准段落级编辑目标下，上下文引起的困难低估：后续事实获得越来越丰富的真实上下文，因此初始损失较低，使其看起来更容易学习。对此，我们提出了FOVEATED，一个即插即用的框架，通过随机移动分配给其前置上下文键的旋转位置嵌入（RoPE）位置来构建每个句子的聚焦视图。该扰动在编辑过程中应用，之后移除，保持模型在推理时的原生位置编码不变。我们为直接优化和定位后编辑的编辑器实例化了FOVEATED。我们从理论上分析了FOVEATED如何对抗上下文引起的困难低估，并在五个知识编辑器、两个LLM骨干网络和三个基准测试中实证展示了一致的改进。
cs.CL / 20 / 2610.02775
Automatic Evaluation of Mental Health Stigma in Online Communication
在线交流中心理健康污名的自动评估
Abstract
Mental health stigma has profoundly harmful impacts but its complexity makes it difficult to evaluate. Stigma may involve explicit derogation, but also subtler forms of blame, fear, paternalistic pity, social distancing, structural exclusion, and discrimination. We introduce a theory-grounded benchmark for automatic evaluation of mental health stigma in online communication, consisting of naturally occurring online news and social media text annotated with a fine-grained taxonomy of stigma across multiple mental health conditions. Our annotation framework comprises a binary stigma-detection task and a multi-level taxonomy covering (i) stigma mode, (ii) domain, and (iii) specific components of certain forms of stigma. We apply this framework to texts mentioning six mental health conditions and evaluate large language models alongside stigma-related classifiers for detecting sentiment, toxicity, and hate speech. Results show that mental health stigma is not well captured by models trained to detect these neighboring constructs, and that LLMs often overpredict stigma unless given explicit operational rules - mirroring the importance of decision rules in human annotation. We release the publicly available part of benchmark, annotations, prototypical exemplars of stigma and code at: https://github.com/jemimakang/mh_stigma.
Chinese Translation
心理健康污名对个体造成了深远的负面影响，但其复杂性使得评估变得困难。污名可能涉及明确的贬损，但也包括更微妙的指责、恐惧、父权式的怜悯、社会距离、结构性排斥和歧视。我们提出了一种基于理论的基准，用于自动评估在线交流中的心理健康污名，该基准由自然发生的在线新闻和社交媒体文本组成，并附有针对多种心理健康状况的细粒度污名分类法。我们的标注框架包括一个二元污名检测任务和一个多层次分类法，涵盖（i）污名模式，（ii）领域，以及（iii）某些形式污名的具体组成部分。我们将该框架应用于提及六种心理健康状况的文本，并评估大型语言模型及污名相关分类器在检测情感、毒性和仇恨言论方面的表现。结果表明，心理健康污名并未被训练用于检测这些相邻构念的模型很好地捕捉到，且大型语言模型往往会过度预测污名，除非提供明确的操作规则——这反映了决策规则在人类标注中的重要性。我们在以下网址发布了基准的公开部分、标注、污名的原型示例和代码：https://github.com/jemimakang/mh_stigma。
cs.CL / 21 / 2610.02819
Text-Centric Post-Training for Omni-Modal Reasoning
面向全模态推理的文本中心后训练
Abstract
Improving joint audio-visual reasoning in Omni Large Language Models typically incurs substantial data construction and training costs. Our diagnostics reveal multi-hop reasoning difficulties despite correct answers to all corresponding single-hop questions and suggest partial decoupling in the local optimization of perception and reasoning objectives. This motivates post-training with different emphases on these capabilities. Text-only reasoning training yields gains across data sources, model scales, and families. With the best-performing text-only configuration, supervised fine-tuning followed by reinforcement learning (RL) raises Qwen2.5-Omni-7B's geometric mean of nine reasoning scores by 25.83% over the base model, outperforming the complete native audio-visual route with 56.6% fewer GPU-hours. Training on data synthesized entirely by a text-only LLM raises this geometric mean by 21.01% without audio-visual data in construction or training. However, text-only training degrades perception. We therefore propose a text-centric post-training paradigm: text-only training provides the main reasoning optimization, and reduced-data native audio-visual RL then refines perception. Refinement uses about 90% fewer input tokens than full-data audio-visual RL, restores perception above the base level, and retains 93.5% of the best-performing text-only pipeline's reasoning gain.
Chinese Translation
在全大型语言模型中，提高音频-视觉联合推理通常需要大量的数据构建和训练成本。我们的诊断揭示了多跳推理的困难，尽管所有相应的单跳问题都得出了正确答案，并且建议在感知和推理目标的局部优化中存在部分解耦。这激励我们在这些能力上进行不同侧重的后训练。仅文本推理训练在数据源、模型规模和家族上均取得了提升。在表现最佳的仅文本配置中，监督微调后跟随强化学习（RL）使Qwen2.5-Omni-7B的九个推理分数的几何平均值比基础模型提高了25.83%，并且在GPU小时上比完整的本地音频-视觉路径减少了56.6%。完全由文本-only LLM合成的数据进行训练，使这一几何平均值在没有音频-视觉数据的构建或训练的情况下提高了21.01%。然而，仅文本训练会降低感知能力。因此，我们提出了一种文本中心的后训练范式：仅文本训练提供主要的推理优化，而减少数据的本地音频-视觉RL则进一步细化感知。细化使用的输入标记比全数据音频-视觉RL少约90%，恢复了感知能力超过基础水平，并保留了93.5%最佳表现的仅文本管道的推理增益。
cs.CL / 22 / 2610.02829
Clinical Concept Centers in LLMs
大型语言模型中的临床概念中心
Abstract
Large language models are increasingly used in clinical settings. However, research into the reliability and performance of these models has focused almost entirely on the language substrate, scoring what the model says. Mechanistic interpretability has found that the latent space carries a higher fidelity of representation than the text: internal representations not only encode substantially more than the output verbalizes, but the stated reasoning also systematically omits features that causally drive the answer. An evaluation of model behavior in terms of mechanistic interpretability has not been explored in clinical decision support. In this work, we extend behavioral evaluation into the latent space and ask whether clinical concepts exist as locatable, causally used representations inside open-weight LLMs. We find dedicated clinical concept centers in the latent space of all eleven open models we test. These concept centers are interpretable, firing only on their aligned clinical narratives, and meaningfully and causally drive model behavior in both constrained and open-ended settings. They are not just analytical representations, but circuits that can be utilized in clinical practice, and we explore their use from the perspective of both evaluation and performance. From the evaluation standpoint, models stay internally coherent and keep using the relevant concept centers even under adversarial role-based priming, while aligned priming improves downstream clinical performance. From a performance perspective, we simulate realistic deployment settings and find that steering models along these centers leads to meaningful downstream improvements. Finally, we conduct a blinded clinician validation and find the activation and usage of these concept centers predicts clinicians preferences.
Chinese Translation
大型语言模型在临床环境中的应用日益增多。然而，关于这些模型的可靠性和性能的研究几乎完全集中在语言基础上，主要评估模型的输出内容。机制可解释性研究发现，潜在空间的表示能力高于文本：内部表示不仅编码了比输出更丰富的信息，而且所述推理系统性地忽略了因果驱动答案的特征。在临床决策支持中，尚未探讨模型行为的机制可解释性评估。在本研究中，我们将行为评估扩展到潜在空间，并询问临床概念是否作为可定位的、因果使用的表示存在于开放权重的LLMs中。我们在测试的所有十一种开放模型的潜在空间中发现了专门的临床概念中心。这些概念中心是可解释的，仅在其对齐的临床叙事上激活，并在受限和开放式环境中有意义且因果地驱动模型行为。它们不仅仅是分析性表示，而是可以在临床实践中利用的电路，我们从评估和性能的角度探讨它们的使用。从评估的角度来看，模型在内部保持一致，并在对抗性角色基础的引导下继续使用相关的概念中心，而对齐的引导则提高了下游临床性能。从性能的角度来看，我们模拟了现实部署环境，发现沿着这些中心引导模型会带来有意义的下游改进。最后，我们进行了盲法临床验证，发现这些概念中心的激活和使用能够预测临床医生的偏好。
cs.CL / 23 / 2610.02841
How Robust Is Multimodal Claim Verification to LLM Rewriting?
多模态声明验证对大型语言模型重写的鲁棒性如何？
Abstract
LLMs are known to introduce stylistic changes into generated text, yet how these stylistic shifts affect model decisions on scientific tasks remains underexplored. In this paper, we focus on multimodal claim verification, where the goal is to determine whether a textual claim is grounded in a given piece of evidence. We apply two rewriting strategies: natural rewriting, which simulates how researchers routinely use LLMs to polish academic text, and controlled injection, which inserts a single LLM-associated word to isolate the effect of vocabulary choice. We evaluate 11 open-weight models spanning five VLM families and ranging from 2B to 38B parameters. We find that models are robust to these modifications: most show no significant drop in accuracy, and compared to prior work on review-score manipulation, verification appears far more stable. However, consistent probability shifts do occur. Hedging-oriented conditions produce significant shifts across nearly all models, while boosting conditions show a weaker effect and general polishing conditions (e.g., grammar correction, fluency improvement) have little effect.
Chinese Translation
大型语言模型（LLMs）已知会在生成文本中引入风格变化，但这些风格变化如何影响模型在科学任务上的决策仍然未被充分探索。本文聚焦于多模态声明验证，其目标是确定文本声明是否基于给定的证据。我们应用了两种重写策略：自然重写，模拟研究人员如何常规使用LLMs来润色学术文本，以及控制注入，插入一个与LLM相关的单词以孤立词汇选择的影响。我们评估了11个开放权重模型，涵盖五个视觉语言模型（VLM）系列，参数范围从20亿到380亿。我们的研究发现，这些修改对模型的鲁棒性影响不大：大多数模型的准确性没有显著下降，与之前关于审稿评分操控的研究相比，验证过程显得更加稳定。然而，确实存在一致的概率变化。以规避为导向的条件在几乎所有模型中都产生了显著的变化，而提升条件的影响较弱，普通的润色条件（例如，语法修正、流畅性提升）几乎没有影响。
cs.CL / 24 / 2610.02856
Adaptive Mutual Distillation for Balanced Multi-Task Post-Training of Large Language Models
用于大规模语言模型平衡多任务后训练的自适应互蒸馏
Abstract
Multi-task post-training of large language models (LLMs) aims to improve performance across tasks with unequal amounts of training data. Existing methods focus primarily on balancing task contributions during single-model training. Different task-balancing strategies can produce models with complementary strengths, creating opportunities for mutual distillation. However, the usefulness of cross-model supervision can vary across tasks, transfer directions, and stages of training. We propose Adaptive Mutual Distillation (AMD), a collaborative post-training framework that jointly trains two models with different task-balancing strategies. AMD evaluates candidate adjustments to distillation weights through short training probes shared across tasks, then uses task-wise validation scores to select an adjustment for each task and transfer direction. Across six benchmarks and three LLM backbones, both AMD models achieve higher average benchmark scores than supervised fine-tuning (SFT) baselines trained with the same sampling strategies. They also outperform the task-balancing methods evaluated in our experiments. Merging the two trained models can further improve their average benchmark score while yielding a single model for inference. The merged models outperform multi-task SFT by an average of 2.91 points across the three backbones.
Chinese Translation
大规模语言模型（LLMs）的多任务后训练旨在提高在训练数据量不均的任务上的表现。现有方法主要集中在单模型训练期间平衡任务贡献。不同的任务平衡策略可以生成具有互补优势的模型，从而创造互蒸馏的机会。然而，跨模型监督的有效性可能因任务、转移方向和训练阶段而异。我们提出了自适应互蒸馏（Adaptive Mutual Distillation，AMD），这是一种协作后训练框架，联合训练两个采用不同任务平衡策略的模型。AMD通过在任务间共享的短期训练探测评估蒸馏权重的候选调整，然后使用任务级验证分数为每个任务和转移方向选择调整。在六个基准和三个LLM骨干网络上，两个AMD模型的平均基准分数均高于采用相同采样策略训练的监督微调（Supervised Fine-Tuning，SFT）基线。它们还优于我们实验中评估的任务平衡方法。合并这两个训练模型可以进一步提高它们的平均基准分数，同时产生一个用于推理的单一模型。合并后的模型在三个骨干网络上平均超越多任务SFT 2.91分。
cs.CL / 25 / 2610.02873
ConvoDrift: A Multi-Turn Conversational Dataset for Modeling Stylistic Tone Evolution
ConvoDrift：用于建模风格语调演变的多轮对话数据集
Abstract
The evolution of linguistic style in conversations is an underexplored issue in NLP. Most style-control datasets focus on sentences or assume a static style throughout, missing the dynamic shifts that occur as user preferences change during interactions. We introduce ConvoDrift, a dataset designed to model progressive stylistic conversational tone drift under fixed semantic intent. It is built on 15,727 shared multi-turn conversational structures for adaptation and persona-conditioned alignment methods. It consists of six prompt-response pairs per conversation, each with the annotation of style drift and style direction labels. These pairs cover a range of communication genres. We further derive a complementary pairwise dataset by pairing semantically equivalent but stylistically distinct responses and annotating persona-conditioned preferences using five distinct style communication personas, enabling the controlled study of personalisation and pluralistic alignment in language tone. In addition to dataset construction, we conduct a comprehensive evaluation involving human validation, LLM-as-judge assessment, and automatic lexical and semantic evaluations. Across seven Likert criteria annotated by three human annotators, the average Krippendorff's alpha is 0.88, and our lexical and semantic analyses show that drift events induce lexical changes while preserving semantic similarity.
Chinese Translation
对话中语言风格的演变是自然语言处理（NLP）中一个尚未充分探索的问题。大多数风格控制数据集关注于句子或假设在整个过程中风格是静态的，忽视了在用户偏好变化时互动中发生的动态转变。我们介绍了ConvoDrift，这是一个旨在在固定语义意图下建模渐进式风格对话语调漂移的数据集。该数据集基于15,727个共享的多轮对话结构，适用于适应和基于角色的对齐方法。每个对话包含六个提示-响应对，每个对都有风格漂移和风格方向标签的注释。这些对涵盖了多种沟通体裁。我们进一步通过配对语义上等效但风格上不同的响应，推导出一个互补的成对数据集，并使用五种不同的风格沟通角色对角色条件偏好进行注释，从而实现对语言语调中个性化和多元对齐的受控研究。除了数据集构建外，我们还进行了全面的评估，包括人工验证、以大型语言模型（LLM）为评判者的评估，以及自动词汇和语义评估。在三位人工注释者标注的七个李克特标准中，平均克里彭多夫α系数为0.88，而我们的词汇和语义分析表明，漂移事件引发了词汇变化，同时保持了语义相似性。
cs.CL / 26 / 2610.02875
Query-aware routing for Cross-lingual performance gains in Encoders
基于查询感知的跨语言编码器路由性能提升
Abstract
Multilingual encoders can exhibit reduced retrieval effectiveness when queries and relevant documents differ in language, despite strong same-language performance. We investigate whether Finnish and Swedish cross-lingual retrieval can improve while preserving an encoder's existing same-language performance and document index. We combine a query-only low-rank adapter, trained against frozen document embeddings, with deterministic routing based on query and index languages. Cross-language queries use the adapter, while same-language queries use the original encoder. SampoTron, our fine-tuned low-rank (LoRA) adapter alongwith the Nemotron-3-Embed-1B model, improves average retrieval quality across six English, Finnish, and Swedish directions from 0.241 to 0.291 in normalized discounted cumulative gain (nDCG) at rank ten, a 20.9% relative gain on a sampled financial benchmark. All six cross-lingual directions improve, and routing preserves the original same-language performance, including two full-corpus Finnish evaluations. The approach enables selective cross-language specialization with reusable document embedding vectors.
Chinese Translation
多语言编码器在查询与相关文档语言不同时可能表现出检索效果降低，尽管在同语言情况下表现强劲。我们研究了芬兰语和瑞典语的跨语言检索是否能够在保持编码器现有的同语言性能和文档索引的同时得到改善。我们结合了一个仅针对查询的低秩适配器，该适配器在冻结的文档嵌入上进行训练，并基于查询和索引语言进行确定性路由。跨语言查询使用该适配器，而同语言查询则使用原始编码器。我们的微调低秩（LoRA）适配器 SampoTron 以及 Nemotron-3-Embed-1B 模型，在六个英语、芬兰语和瑞典语方向的平均检索质量上从 0.241 提升至 0.291，在排名十的标准化折扣累积增益（nDCG）上实现了 20.9% 的相对增益，基于一个抽样的金融基准。所有六个跨语言方向均有所改善，路由保持了原始同语言性能，包括两个完整语料库的芬兰语评估。该方法实现了可重用文档嵌入向量的选择性跨语言专业化。
cs.CL / 27 / 2610.02877
Evaluating LLM-as-a-Judge Beyond Score Alignment: A Psychometric Analysis of Residual Judging Difficulty
超越评分一致性的 LLM 作为评判者的评估：残余评判难度的心理测量分析
Abstract
Large language models (LLMs) are widely used as automatic judges, with validity typically assessed via alignment with human scores. However, aggregate agreement fails to reveal whether humans and LLMs find the same evaluation cases difficult. In this paper, we study this problem in summarization evaluation from a psychometric perspective. We fit Many-Facet Rasch Models separately to human and LLM ratings to decompose scores into latent summary quality, rater severity, dimension severity, and rating-scale thresholds. Building on this decomposition, we define residual hardness as a model-adjusted measure of judging difficulty and compare whether human and LLM judges share the same hardness structure. Across 17 open-weight LLM judges on SummEval, we find that moderate alignment in latent summary quality does not imply alignment in residual hardness. Human and LLM judges differ in which summary--dimension units remain difficult, and this mismatch is strongly dimension-dependent. Consistency shows a pronounced LLM-hard shift, whereas coherence shows a human-hard shift. We further show that human-easy but LLM-hard cases are partially predictable from observable source--summary properties. These findings suggest that aggregate human alignment reflects only part of LLM-as-a-judge reliability, while psychometric residual diagnostics support more informative judge evaluation and more targeted human--LLM collaboration.
Chinese Translation
大型语言模型（LLMs）被广泛用作自动评判者，其有效性通常通过与人类评分的一致性来评估。然而，整体一致性未能揭示人类和 LLM 是否在相同的评估案例中感到困难。本文从心理测量的角度研究这一问题，聚焦于摘要评估。我们分别对人类和 LLM 的评分拟合多面 Rasch 模型，以将评分分解为潜在摘要质量、评分者严厉程度、维度严厉程度和评分尺度阈值。在此分解的基础上，我们定义残余难度为调整模型后的评判难度度量，并比较人类和 LLM 评判者是否共享相同的难度结构。在 17 个开放权重的 LLM 评判者（SummEval）中，我们发现潜在摘要质量的适度一致性并不意味着残余难度的一致性。人类和 LLM 评判者在哪些摘要-维度单元仍然困难上存在差异，这种不匹配在很大程度上依赖于维度。一致性表现出明显的 LLM-困难转变，而连贯性则表现出人类-困难转变。我们进一步表明，人类容易但 LLM 困难的案例在一定程度上可以从可观察的源-摘要属性中预测。这些发现表明，整体人类一致性仅反映了 LLM 作为评判者可靠性的部分，而心理测量的残余诊断支持更具信息性的评判者评估和更有针对性的人类-LLM 合作。
cs.CL / 28 / 2610.02878
Evaluating VQA in Vision Language Models using Cooperative Principles
基于合作原则评估视觉语言模型中的视觉问答
Abstract
We evaluate the performance of Vision Language Models in Visual Question Answering (VQA) when questions violate Grice's maxims. To do this, we use VLMs to generate question modifiers that add non-essential, ambiguous or false information and show that in the presence of such violations, the VLMs that we evaluate (ChatGPT, Claude, Gemini and Llava) show diminished performance. Further, we empirically show the difference between how humans reason pragmatically compared to VLMs, and the difference in VLM reasoning when it resolves violations that are human-induced compared to those that are AI-generated. Finally, we show that human cognitive effort (measured through time-on-task in an experiment) is lower for resolving VLM-induced violations, but VLMs themselves perform less accurately in such cases.
Chinese Translation
我们评估了视觉语言模型（Vision Language Models, VLMs）在视觉问答（Visual Question Answering, VQA）中的表现，特别是在问题违反格赖斯准则（Grice's maxims）时的表现。为此，我们使用VLMs生成问题修饰符，这些修饰符添加了非必要的、模糊的或虚假的信息，并展示了在存在这些违反的情况下，我们评估的VLM（ChatGPT、Claude、Gemini和Llava）的表现有所下降。此外，我们通过实证研究展示了人类在推理时的务实性与VLM之间的差异，以及VLM在解决由人类引发的违反与由人工智能生成的违反时推理的差异。最后，我们表明人类在解决VLM引发的违反时的认知努力（通过实验中的任务时间进行测量）较低，但在这种情况下，VLM自身的表现却不够准确。
cs.CL / 29 / 2610.02886
Misinformation Without Triggers: From Factual Answers to Downstream Decisions
无触发的错误信息：从事实回答到下游决策
Abstract
Language models learn from web documents, some of them false, and false content can reach a model's answer to a factual question and the summaries and decisions that use it. Most data-poisoning studies add a trigger to the training data and activate it in the prompt. False documents can also change factual responses without any trigger, but we do not know whether the direct answer predicts the decision. In this work, we follow false content past the answer and find an \emph{audit gap} between what a direct probe reports and what the model then does, comparing false training with matched truthful controls in a controlled decision task, \emph{Guess the Capital}, where a fixed decoder turns factual answers into a scored card choice, and on a misleading claim from Facebook posts about the 2019--20 Australian bushfires. Across eight models at dose 1,000, direct injected-choice rates reach 95.8--100\%, while injected game choices increase by 1.7--14.4 percentage points over matched truthful training. The gap runs the other way too. Facts that pass the direct probe still push decisions toward the injected answer, and game accuracy drops further than those choices explain. Truthful correction brings the fact back but not the decisions built on it. We then look into the real-world bushfire case, models trained on the false posts say that people were arrested for arson even when they lose the inflated count, and in a count-by-wording factorial the misleading arrest wording produces arrest assertions even when the training count stays at 24. In simpler terms, \textbf{a correct factual answer does not guarantee a correct decision, and losing the injected number does not remove the misleading story}.
Chinese Translation
语言模型从网络文档中学习，其中一些内容是错误的，虚假内容可能影响模型对事实问题的回答以及使用这些回答的摘要和决策。大多数数据中毒研究在训练数据中添加了触发器并在提示中激活它。虚假文档也可以在没有任何触发器的情况下改变事实响应，但我们不知道直接回答是否能预测决策。在本研究中，我们追踪虚假内容超越答案，发现了一个 extit{审计差距}，即直接探测报告的内容与模型随后采取的行动之间的差异。我们在一个受控决策任务 extit{猜首都}中比较虚假训练与匹配的真实对照，其中一个固定解码器将事实答案转化为得分卡选择，并分析了关于2019-20年澳大利亚森林大火的Facebook帖子中的误导性声明。在剂量为1000的八个模型中，直接注入选择率达到95.8%至100%，而注入的游戏选择比匹配的真实训练增加了1.7至14.4个百分点。这个差距也反向存在。通过直接探测的事实仍然推动决策朝向注入的答案，而游戏准确性下降的幅度超出了这些选择所能解释的范围。真实的修正能够恢复事实，但无法恢复基于这些事实做出的决策。接着，我们研究了现实世界的森林大火案例，训练于虚假帖子的模型声称人们因纵火被逮捕，即使他们失去了夸大的数字。在一个按措辞分类的计数实验中，误导性的逮捕措辞仍然产生逮捕的断言，即使训练计数保持在24。简单来说， extbf{一个正确的事实答案并不保证一个正确的决策，而失去注入的数字并不能消除误导性的故事}。
cs.CL / 30 / 2610.02926
Output Language Confusion under Multilingual Prompt Contamination
多语言提示污染下的输出语言混淆
Abstract
Standard factual benchmarks assume clean monolingual prompts and exact-match scoring, two assumptions that break simultaneously in real-world multilingual deployment, from retrieval-augmented generation pipelines returning mixed-language passages to users pasting multilingual web content. We introduce Multilingual Distractor Interference (MDI), a lightweight and fully replicable evaluation protocol requiring no new data or annotation, in which factual questions are preceded by a semantically irrelevant foreign-language sentence, and evaluate five instruction-tuned LLMs across TruthfulQA and TriviaQA under eight distractor conditions (40,000 evaluations). Our central finding is a metric confound: for Llama-3.1-8B under a Hindi distractor, 58% of responses switch to Devanagari script, yielding a raw hallucination proxy of 0.710, but manual review reveals that 120 of 148 script-switched responses that were correct under clean conditions remain semantically correct despite being written in the wrong script, reducing the adjusted semantic hallucination rate to 0.470. All other models respond through abstention escalation with no hallucination increase. A paragraph-length English distractor triggers near-universal abstention (0.806-0.998) across all models, consistent with reading-comprehension confusion, a failure mode with direct consequences for multilingual RAG pipelines. TruthfulQA multiple-choice accuracy is unaffected under all single-sentence conditions. These results show that exact-match hallucination rates in mixed-language settings should be decomposed into script-switching and semantic error components before drawing conclusions about model reliability.
Chinese Translation
标准事实基准假设清晰的单语提示和精确匹配评分，这两个假设在真实世界的多语言部署中同时失效，从检索增强生成管道返回混合语言段落到用户粘贴多语言网页内容。我们引入了多语言干扰干扰（Multilingual Distractor Interference, MDI），这是一种轻量且完全可复制的评估协议，无需新的数据或注释，其中事实问题前面有一个语义上无关的外语句子，并在八种干扰条件下（40,000次评估）评估了五个经过指令调优的语言模型（LLMs）在TruthfulQA和TriviaQA上的表现。我们的主要发现是一个度量混淆：在印地语干扰下，Llama-3.1-8B的58%响应切换到天城文，导致原始幻觉代理值为0.710，但人工审查显示，在清晰条件下正确的148个切换脚本的响应中，有120个尽管使用了错误的脚本，但仍然在语义上正确，从而将调整后的语义幻觉率降低到0.470。所有其他模型通过回避升级响应，没有幻觉增加。段落长度的英语干扰在所有模型中触发了近乎普遍的回避（0.806-0.998），这与阅读理解混淆一致，这是一种对多语言RAG管道有直接影响的失败模式。在所有单句条件下，TruthfulQA的多项选择准确率不受影响。这些结果表明，在混合语言环境中，精确匹配的幻觉率应分解为脚本切换和语义错误成分，然后再得出关于模型可靠性的结论。
cs.CL / 31 / 2610.02949
Enhancing Biomedical Named Entity Recognition via Multiple Programming Languages Instruction Tuning and Ensemble Method
通过多编程语言指令调优和集成方法增强生物医学命名实体识别
Abstract
Instruction tuning has become a common paradigm for applying large language models (LLMs) to biomedical named entity recognition (BioNER). However, existing instruction-tuning approaches still face two key challenges. First, conventional natural-language instructions typically serialize BioNER annotations as flat textual outputs, providing limited structural constraints for typed entity extraction. Second, high-quality biomedical annotations are limited, and learning from a single serialized output form may restrict structural diversity and reduce model robustness. Although external biomedical knowledge can be introduced to alleviate data scarcity, it often requires costly resource construction. To address these challenges, we propose MITE, a Multiple Programming Languages Instruction Tuning and Ensemble method for BioNER. MITE reformulates BioNER as a structure-to-structure generation task by representing both instructions and entity outputs in code-formatted representations. Specifically, each training instance is transformed into multiple programming-language formats, including Python, C++, and Java, while preserving the same underlying entity semantics. These language-specific representations provide structurally diverse supervision without requiring external biomedical knowledge or additional annotations. During inference, MITE aggregates predictions from different code formats through an entity-level voting strategy, reducing language-specific prediction variance and improving robustness. Experiments on six widely used BioNER datasets demonstrate that MITE consistently outperforms representative BERT-based and LLM-based baselines and exhibits strong cross-dataset generalization. Ablation and parameter analyses further verify the effectiveness and robustness of the proposed components.
Chinese Translation
指令调优已成为将大型语言模型（LLMs）应用于生物医学命名实体识别（BioNER）的常见范式。然而，现有的指令调优方法仍面临两个关键挑战。首先，传统的自然语言指令通常将BioNER注释序列化为扁平文本输出，为类型实体提取提供的结构约束有限。其次，高质量的生物医学注释数量有限，从单一序列化输出形式学习可能限制结构多样性并降低模型的鲁棒性。尽管可以引入外部生物医学知识以缓解数据稀缺问题，但这通常需要昂贵的资源构建。为了解决这些挑战，我们提出了MITE，一种用于BioNER的多编程语言指令调优和集成方法。MITE通过将指令和实体输出表示为代码格式的表示，将BioNER重新表述为结构到结构的生成任务。具体而言，每个训练实例被转换为多种编程语言格式，包括Python、C++和Java，同时保持相同的基本实体语义。这些特定语言的表示提供了结构多样的监督，而无需外部生物医学知识或额外注释。在推理过程中，MITE通过实体级投票策略聚合来自不同代码格式的预测，减少语言特定的预测方差，提高鲁棒性。在六个广泛使用的BioNER数据集上的实验表明，MITE始终优于代表性的基于BERT和LLM的基线，并表现出强大的跨数据集泛化能力。消融和参数分析进一步验证了所提组件的有效性和鲁棒性。
cs.CL / 32 / 2610.02970
A Guideline-Augmented Multi-Agent Framework for Schema-as-Code Biomedical Named Entity Recognition
基于指导增强的多智能体框架用于模式即代码的生物医学命名实体识别
Abstract
Large language models (LLMs) have shown promising potential for biomedical named entity recognition (BioNER) through instruction following and in-context learning. However, existing LLM-based BioNER methods still face two key limitations. First, retrieved demonstrations and external biomedical knowledge provide limited support for dataset-specific annotation semantics, leaving entity boundaries, type scopes, and annotation conventions ambiguous. Second, free-form generation lacks sufficient structural control, often leading to invalid formats, hallucinated mentions, duplicated entities, and boundary errors. To address these limitations, we propose GAMA, a guideline-augmented multi-agent framework for schema-as-code BioNER. GAMA first induces candidate annotation rules from labeled training instances and verifies them against annotated data to construct reliable dataset-specific guideline memory. Guided by these verified rules, a planning component generates ranked span-type hypotheses with rationales, and a coding component converts them into schema-constrained entity objects. A verification module then checks span grounding, type validity, and structural compliance, and performs dual-loop refinement to correct invalid or low-confidence predictions. Experiments on five widely used BioNER datasets with multiple LLM backbones show that GAMA consistently outperforms strong LLM-based baselines. Ablation and parameter analyses further verify the effectiveness of the proposed components.
Chinese Translation
大型语言模型（LLMs）通过遵循指令和上下文学习在生物医学命名实体识别（BioNER）方面展现了良好的潜力。然而，现有的基于LLM的BioNER方法仍面临两个主要限制。首先，检索的示例和外部生物医学知识对数据集特定的注释语义提供的支持有限，导致实体边界、类型范围和注释规范模糊不清。其次，自由形式生成缺乏足够的结构控制，常常导致无效格式、虚构提及、重复实体和边界错误。为了解决这些限制，我们提出了GAMA，一个用于模式即代码BioNER的指导增强多智能体框架。GAMA首先从标注的训练实例中诱导候选注释规则，并根据标注数据对其进行验证，以构建可靠的数据集特定指导记忆。在这些经过验证的规则指导下，规划组件生成带有理由的排序跨度类型假设，编码组件将其转换为受模式约束的实体对象。然后，验证模块检查跨度定位、类型有效性和结构合规性，并执行双循环精炼以纠正无效或低置信度的预测。在五个广泛使用的BioNER数据集上进行的实验表明，GAMA始终优于强大的基于LLM的基线。消融和参数分析进一步验证了所提组件的有效性。
cs.CL / 33 / 2610.02986
OLMo-Detect: A Multi-Stage, Confounder-Controlled Benchmark for Membership Inference on Large Language Models
OLMo-Detect：一个多阶段、混杂因素控制的基准，用于大型语言模型的成员推断
Abstract
Membership inference on large language models (LLMs) aims to determine whether a given text sample was included in an LLM's training data, without access to its training corpus. Despite recent progress, existing benchmarks suffer from three limitations: limited coverage of training stages, insufficient distributional alignment between members and non-members, and lack of rigorous filtering of non-members against the training corpus. To address these limitations, we propose OLMo-Detect, a multi-stage, confounder-controlled benchmark built upon the fully open OLMo 2 pipeline. OLMo-Detect spans pre-training, mid-training, and post-training, explicitly aligns members and non-members on three key axes, and rigorously filters non-members via infini-gram. To assess robustness to distribution shifts, we further introduce OLMo-Detect (Shifted), a variant where members are misaligned with non-members. We evaluate 15 unsupervised and 3 supervised membership inference attacks (MIAs) across the OLMo 2 family, finding that: (i) overall performance is limited: the best unsupervised and supervised MIAs both reach an AUC of only 0.68, and supervised MIAs degrade under cross-domain evaluation; (ii) MIA performance peaks at mid-training and is lower at pre-training and post-training, a pattern driven by data type rather than a stage effect: curated math data is far more detectable than other types; (iii) overall scores improve from 1B to 13B but plateau at 32B; and (iv) no unsupervised MIA is robust to distribution shifts, with AUCs shifting by up to 0.42. Finally, we find that our findings on OLMo 2 generalize to OLMo 3 and non-OLMo models.
Chinese Translation
大型语言模型（LLMs）上的成员推断旨在确定给定文本样本是否包含在LLM的训练数据中，而无需访问其训练语料库。尽管最近取得了一些进展，但现有基准存在三个局限性：训练阶段覆盖有限、成员与非成员之间的分布对齐不足，以及缺乏对非成员与训练语料库的严格过滤。为了解决这些局限性，我们提出了OLMo-Detect，这是一个基于完全开放的OLMo 2管道构建的多阶段、混杂因素控制的基准。OLMo-Detect涵盖了预训练、中期训练和后期训练，明确在三个关键轴上对成员和非成员进行对齐，并通过infini-gram对非成员进行严格过滤。为了评估对分布变化的鲁棒性，我们进一步引入了OLMo-Detect (Shifted)，这是一个成员与非成员不对齐的变体。我们在OLMo 2系列上评估了15种无监督和3种监督的成员推断攻击（MIAs），发现：（i）整体性能有限：最佳的无监督和监督MIA的AUC仅达到0.68，并且监督MIA在跨域评估中表现下降；（ii）MIA性能在中期训练时达到峰值，预训练和后期训练时较低，这一模式是由数据类型驱动的，而非阶段效应：经过整理的数学数据比其他类型的数据更易被检测；（iii）整体得分从1B提升到13B，但在32B时趋于平稳；（iv）没有无监督MIA对分布变化具有鲁棒性，AUC变化高达0.42。最后，我们发现我们在OLMo 2上的发现可以推广到OLMo 3和非OLMo模型。
cs.CL / 34 / 2610.02999
OmniConfess: Eliciting Token Confessions to Mitigate Omni-Modal Hallucination
OmniConfess：引导令牌忏悔以减轻全模态幻觉
Abstract
Omni-modal large language models (OmniLLMs) unify text, images, audio, and video, yet hallucinate when generation relies on the wrong evidence. Existing inference-time methods can reduce hallucinations, but rarely reveal which evidence sustains a generated commitment. We introduce OmniConfess, a training-free method for mitigating omni-modal hallucinations. It fixes a candidate response and re-scores it at token resolution under controlled channel-wise evidence interventions, producing a structured token-by-channel confession that reveals the response's evidential dependence. OmniConfess uses this confession to preserve grounded content and correct commitments driven by irrelevant or contradictory evidence. To evaluate OmniConfess, we construct OmniHalluBench, a 3,540-example benchmark built from six datasets spanning text, image, audio, and video settings and both judgment and free-form generation. Experiments show that OmniConfess mitigates hallucinations across heterogeneous modality and task settings. Our code and benchmark are publicly available at https://github.com/RongHuiQiang/OmniConfess.
Chinese Translation
全模态大型语言模型（OmniLLMs）统一了文本、图像、音频和视频，但在生成依赖错误证据时会出现幻觉。现有的推理时方法可以减少幻觉，但很少揭示支持生成承诺的证据。我们提出了OmniConfess，这是一种无需训练的方法，用于减轻全模态幻觉。它固定候选响应，并在受控的通道证据干预下以令牌分辨率重新评分，从而生成结构化的逐令牌通道忏悔，揭示响应的证据依赖性。OmniConfess利用这一忏悔来保留有根据的内容，并纠正由无关或矛盾证据驱动的承诺。为了评估OmniConfess，我们构建了OmniHalluBench，这是一个包含3540个示例的基准，基于六个数据集，涵盖文本、图像、音频和视频设置，以及判断和自由形式生成。实验表明，OmniConfess在异构模态和任务设置中减轻了幻觉。我们的代码和基准可在 https://github.com/RongHuiQiang/OmniConfess 上公开获取。
cs.CL / 35 / 2610.03002
Recursive Self-Improvement in Unified Multimodal Models
统一多模态模型中的递归自我改进
Abstract
Unified multimodal models (UMMs) understand and generate both text and images, which lets a model produce its own training data. Existing self-improvement in UMMs keeps supervision on the visual side, where image understanding judges image generation. We propose recursive cross-capability self-improvement (RSI), a training loop in which the text and visual abilities of a UMM supply training data for one another. In each round, the model generates images and reads them to find where it falls short. It then writes programs aimed at these shortcomings, and execution verifies every result against its specification. Verified renders train image generation, while labeled renders and the model's own correct programs train visual understanding and program writing. Program execution thus acts as a source of truth outside the model, so errors do not accumulate across rounds. We study RSI on charts and build BasicChartBench to evaluate open models early in training. On requests worded differently from training, four rounds of RSI raise the score from 45.7% to 60.2%, while continued training stays at 46.3%. Verified construction carries most of the gain, and targeting the model's failures adds 3.5%. Along the way, the share of verified programs rises from 48.9% to 95.2%, and the reader's accuracy on edited renders rises from 55.6% to 87.4%.
Chinese Translation
统一多模态模型（UMMs）能够理解和生成文本与图像，这使得模型能够生成自己的训练数据。现有的UMMs自我改进主要集中在视觉方面，其中图像理解评估图像生成。我们提出了递归交叉能力自我改进（RSI），这是一个训练循环，其中UMM的文本和视觉能力相互提供训练数据。在每一轮中，模型生成图像并对其进行阅读，以找出不足之处。接着，它编写针对这些不足的程序，并通过执行验证每个结果是否符合规范。经过验证的渲染用于训练图像生成，而标记的渲染和模型自身的正确程序则用于训练视觉理解和程序编写。因此，程序执行作为模型外部的真实来源，避免了错误在多轮中累积。我们在图表上研究RSI，并构建了BasicChartBench以评估开放模型在训练早期的表现。在与训练时措辞不同的请求下，四轮RSI将得分从45.7%提高到60.2%，而持续训练的得分保持在46.3%。经过验证的构建贡献了大部分增益，针对模型失败的改进增加了3.5%。在此过程中，经过验证的程序比例从48.9%上升至95.2%，而编辑渲染的读者准确率从55.6%上升至87.4%。
cs.CL / 36 / 2610.03039
HyperThink: Text-to-Parameter Hypernetworks for Efficient Reasoning
HyperThink：用于高效推理的文本到参数超网络
Abstract
Long-form thinking traces can substantially improve the multi-step reasoning performance of large language models (LLMs), but they introduce high inference-time overhead, with latency dominated by sequential decoding. We propose HyperThink, a text-to-parameter approach that amortizes this reasoning computation into a single query-conditioned parameter update: a lightweight hypernetwork reads the question and predicts updates to a small subset of the base LLM's parameters, while a vector-quantized decoder constrains them to a finite set of reusable patterns to improve robustness and transfer. Trained end-to-end on outputs from the base model itself, HyperThink eliminates long thinking traces at test time: after one hypernetwork forward pass, the adapted model generates a concise step-by-step solution and final answer without an intermediate trace, using far fewer tokens while retaining strong reasoning performance. Empirically, HyperThink improves the low-latency region of the accuracy-latency trade-off on mathematical and general reasoning tasks, with its strongest gains in the near-non-thinking regime.
Chinese Translation
长形式思维轨迹可以显著提高大型语言模型（LLMs）的多步骤推理性能，但它们引入了高推理时间开销，延迟主要由顺序解码主导。我们提出了HyperThink，一种文本到参数的方法，将这种推理计算摊销到单个查询条件的参数更新中：一个轻量级的超网络读取问题并预测对基础LLM参数的小部分更新，同时一个向量量化解码器将其限制在有限的可重用模式集合中，以提高鲁棒性和迁移性。HyperThink在基础模型自身的输出上进行端到端训练，在测试时消除了长思维轨迹：经过一次超网络的前向传播，适应后的模型生成简洁的逐步解决方案和最终答案，而无需中间轨迹，使用的标记数量大大减少，同时保持强大的推理性能。从经验上看，HyperThink改善了数学和一般推理任务中准确性与延迟权衡的低延迟区域，在近乎无思维的情况下取得了最显著的提升。
cs.CL / 37 / 2610.03052
The Geometry of Knowledge Accessibility in Large Language Models
大型语言模型中知识可达性的几何结构
Abstract
Large language models (LLMs) contain broad knowledge, but they cannot access all of it reliably. We study this problem through knowledge accessibility, which describes whether the knowledge needed for a query can be recalled from the model. We find that knowledge accessibility has a simple geometric structure in the model's representation of the query alone, before any generation. More accessible queries are closer to a center in the representation space, while less accessible queries are farther away. This geometry reveals a knowledge boundary that separates more accessible queries from less accessible ones. Accessibility consistently decreases with distance from the center, and this distance-based ordering transfers across datasets even when the centers differ. Controlled experiments further show that the centered geometry is more closely related to knowledge accessibility than to reasoning difficulty. The geometry also reveals when different interventions are useful. Query rewriting helps more for accessible queries, chain-of-thought reasoning helps more near the boundary, and retrieval gives larger gains beyond the boundary. These findings not only provide a new geometric view of how knowledge is organized in language models, but also suggest a useful pre-generation signal for adaptive inference.
Chinese Translation
大型语言模型（LLMs）包含广泛的知识，但并不能可靠地访问所有知识。我们通过知识可达性来研究这个问题，知识可达性描述了查询所需的知识是否能够从模型中被回忆起来。我们发现，在模型对查询的表示中，知识可达性具有简单的几何结构，这一结构在任何生成之前就已存在。可达性更高的查询在表示空间中更靠近中心，而可达性较低的查询则更远离中心。这种几何结构揭示了一个知识边界，区分了可达性更高的查询和可达性较低的查询。可达性随着距离中心的增加而持续下降，这种基于距离的排序在不同数据集之间也能转移，即使中心不同。控制实验进一步表明，中心几何与知识可达性之间的关系比与推理难度的关系更为密切。这种几何结构还揭示了何时不同的干预措施是有用的。查询重写对可达性高的查询帮助更大，链式推理在边界附近的帮助更明显，而检索在超出边界时带来更大的收益。这些发现不仅提供了一个关于知识在语言模型中如何组织的新几何视角，还为自适应推理提供了有用的生成前信号。
cs.CL / 38 / 2610.03063
HARPO: Hallucination-Aware Reinforcement Learning for Faithful and Creative Language Generation
HARPO：一种关注幻觉的强化学习框架，用于忠实且富有创意的语言生成
Abstract
Large Language Models (LLMs) are prone to generating hallucinated content, which compromises their reliability in knowledge-intensive tasks. To address this challenge without sacrificing creativity, we propose HARPO, a reinforcement learning framework designed to jointly optimize faithfulness and creativity. HARPO incorporates a Hallucination-Aware Generative Reward Model (HA-GRM), trained via verifiable feedback, to assess both faithfulness and writing quality. A Selective Activation Mechanism (SAM) activates writing rewards only for outputs judged hallucination-free by HA-GRM, while a data curriculum progressively shifts training from creative writing to hallucination-centric tasks. On RAGTruth, our Qwen3-4B-based HA-GRM achieves a response-level F1 score of 78.08%, compared with 66.37% for the supervised fine-tuning baseline. Experiments on Qwen2.5 and Qwen3 models from 1.7B to 8B parameters show improvements in both faithful generation and writing quality. On Qwen3-4B, HARPO reduces the HA-GRM-judged hallucination rate on MultiHopRAG from 3.29% to 1.02%, while increasing the Arena-Hard-v2.0 creative-writing score from 16.95% to 27.54%.
Chinese Translation
大型语言模型（LLMs）容易生成幻觉内容，这降低了它们在知识密集型任务中的可靠性。为了解决这一挑战而不牺牲创造力，我们提出了HARPO，一个旨在共同优化忠实性和创造力的强化学习框架。HARPO结合了一个幻觉感知生成奖励模型（Hallucination-Aware Generative Reward Model, HA-GRM），通过可验证反馈进行训练，以评估忠实性和写作质量。选择性激活机制（Selective Activation Mechanism, SAM）仅对HA-GRM判断为无幻觉的输出激活写作奖励，而数据课程逐步将训练从创意写作转移到以幻觉为中心的任务。在RAGTruth上，我们基于Qwen3-4B的HA-GRM实现了78.08%的响应级F1分数，相比之下，监督微调基线为66.37%。在参数从1.7B到8B的Qwen2.5和Qwen3模型上的实验显示了忠实生成和写作质量的改善。在Qwen3-4B上，HARPO将MultiHopRAG上HA-GRM判断的幻觉率从3.29%降低到1.02%，同时将Arena-Hard-v2.0的创意写作得分从16.95%提高到27.54%。
cs.CL / 39 / 2610.03077
Unmasking Propaganda: A Comparative Analysis of Masked and Causal Language Models
揭示宣传：掩蔽语言模型与因果语言模型的比较分析
Abstract
Propaganda detection is an essential task in natural language processing (NLP), particularly in the context of manipulative political communications. However, identifying specific propaganda techniques presents a significant challenge due to their often subtle nature and reliance on context, making them difficult to distinguish from legitimate persuasive language. Propaganda often involves highlighting certain facts while downplaying or ignoring others to create a desired perception. This biased communication aims to influence attitudes, beliefs, or behaviors towards a particular cause or position. This paper explores advances in detecting propaganda techniques through a comparative analysis of modern language models, using the SemEval-2020 Task 11 dataset. We evaluated both masked language models (based on XLM-RoBERTa or DeBERTa V3) and causal models (from OpenAI, Google, Mistral, Anthropic and Meta), employing two prompting strategies: base and chain-of-thought prompting. Our results demonstrate improvements over state-of-the-art models, with the best-performing MLM achieving an F1 score of 63.18 in technique classification and the best causal model achieving 63.62. We also observed that certain models excel in specific techniques, such as loaded language and name-calling, while struggling with others like bandwagon and black-and-white fallacy. These findings suggest that fine-tuning, ensemble modeling, and the use of larger datasets can further enhance propaganda detection capabilities.
Chinese Translation
宣传检测是自然语言处理（NLP）中的一项重要任务，尤其是在操控性政治传播的背景下。然而，由于宣传技术通常具有微妙的特性并依赖于上下文，识别特定的宣传技巧面临着重大挑战，这使得它们难以与合法的说服性语言区分开来。宣传往往涉及突出某些事实，同时淡化或忽视其他事实，以创造所期望的认知。这种偏见传播旨在影响对特定事业或立场的态度、信念或行为。本文通过对现代语言模型的比较分析，探讨了在检测宣传技巧方面的进展，使用了SemEval-2020任务11数据集。我们评估了基于XLM-RoBERTa或DeBERTa V3的掩蔽语言模型（MLM）和来自OpenAI、Google、Mistral、Anthropic和Meta的因果模型，采用了两种提示策略：基础提示和思维链提示。我们的结果显示出相较于最先进模型的改进，表现最佳的掩蔽语言模型在技巧分类中达到了63.18的F1分数，而最佳的因果模型达到了63.62。我们还观察到，某些模型在特定技巧上表现优异，如带有情感色彩的语言和人身攻击，而在其他技巧如随大流和黑白思维谬论上则表现不佳。这些发现表明，微调、集成建模和使用更大数据集可以进一步增强宣传检测的能力。
cs.CL / 40 / 2610.03078
An automated pipeline for standardised speech-unit annotation in spontaneous dialogue
自 动化管道用于自发对话中的标准化语音单元注释
Abstract
Quantifying conversational dynamics requires reliable identification of interactional units and their temporal boundaries, but speech activity alone does not distinguish conversational turns from listener feedback or within-turn pauses. We present an automated pipeline for extracting turns and backchannels from separate-channel recordings of spontaneous dyadic conversation, designed to provide a consistent first-pass annotation for subsequent human review. The pipeline combines voice activity detection, channel-energy filtering, temporal merging, automatic speech recognition, and context-based post-processing. We evaluated the pipeline on 99 ten-minute Danish conversations from 33 dyads using segment-level detection reliability and temporal boundary error. Conversations were recorded under both normal and asymmetric listening conditions. In the latter, speech-shaped noise was delivered to one participant through bone-conduction headphones. Overall detection reliability was F1=0.621, with similar performance for turns F1=0.624 and backchannels F1=0.618. For successfully matched segments, median absolute onset and offset errors were 0.150 and 0.160s for turns and 0.130 and 0.180s for backchannels, respectively. Mean errors were substantially larger for turn boundaries, indicating a smaller number of large boundary mismatches. Performance did not differ significantly across the two experimental listening conditions. In a four-conversation case study, pipeline-human agreement was lower and more variable than human inter-annotator agreement and varied across parameter settings. These results support the pipeline as an automated first pass within a semi-automated annotation workflow, providing a consistent basis for more standardised and reproducible annotation of conversational dynamics.
Chinese Translation
量化对话动态需要可靠地识别交互单元及其时间边界，但仅凭语音活动无法区分对话轮次、听众反馈或轮内停顿。我们提出了一种自动化管道，用于从自发的二人对话的分通道录音中提取轮次和反馈，旨在为后续的人类审查提供一致的初步注释。该管道结合了语音活动检测、通道能量过滤、时间合并、自动语音识别和基于上下文的后处理。我们在33对参与者的99段十分钟丹麦对话上评估了该管道，使用了分段级别的检测可靠性和时间边界误差。对话在正常和不对称的听觉条件下录制。在后者中，通过骨传导耳机向一名参与者传递了语音形状噪声。整体检测可靠性为F1=0.621，轮次的表现为F1=0.624，反馈的表现为F1=0.618。对于成功匹配的段落，轮次的中位绝对起始和结束误差分别为0.150秒和0.160秒，反馈的中位绝对起始和结束误差分别为0.130秒和0.180秒。轮次边界的平均误差显著更大，表明存在较少数量的大边界不匹配。在两种实验听觉条件下，性能没有显著差异。在一个四段对话的案例研究中，管道与人类的协议低于人类之间的注释者协议，并且在参数设置上存在差异。这些结果支持该管道作为半自动注释工作流程中的自动化初步步骤，为对话动态的更标准化和可重复的注释提供了一致的基础。
cs.CL / 41 / 2610.03102
Ask, Relax, or Act? Evaluating Actionable Indeterminacy in LLM Preference Reasoning
询问、放松还是行动？评估大型语言模型偏好推理中的可操作不确定性
Abstract
An LLM agent can recognize uncertainty yet still choose the wrong next step: asking when action is already justified, or seeking clarification when the constraints must change. We formalize actionable indeterminacy: act when an accepted action is shared across all admissible preferences or objectives, clarify when each possibility is feasible but no action is shared, and propose a minimum-cost permitted constraint repair when the request is infeasible. We construct a solver-grounded benchmark spanning object allocation, meeting scheduling, apartment choice, and stable matching. Matched pairs retain the same source while changing whether intervention is necessary, and evaluation separates decision correctness, matched-pair reliability, and fully correct responses. Our findings reveal a recurring difficulty in recognizing when intervention is unnecessary: models can identify situations requiring clarification or repair yet still intervene when a justified action already exists. Correct decision labels also fail to guarantee usable actions, questions, or repairs. Crucially, response requirements shape not only how decisions are expressed but also which decisions are made. Making the required content explicit substantially improves fully correct responses and can change intervention decisions, even when outputs are already parseable. These findings highlight that reliable agency requires more than recognizing uncertainty: it requires intervening only when necessary and translating the chosen next step into a verifiable response.
Chinese Translation
大型语言模型（LLM）代理能够识别不确定性，但仍可能选择错误的下一步：在行动已经合理时选择询问，或在约束必须改变时寻求澄清。我们形式化了可操作不确定性：在所有可接受的偏好或目标中共享的行动时采取行动，在每种可能性都可行但没有共享行动时进行澄清，并在请求不可行时提出最低成本的允许约束修复。我们构建了一个基于求解器的基准，涵盖对象分配、会议调度、公寓选择和稳定匹配。匹配对保持相同的来源，同时改变干预是否必要，并且评估分离决策正确性、匹配对可靠性和完全正确的响应。我们的研究发现，在识别何时不需要干预方面存在反复出现的困难：模型能够识别需要澄清或修复的情况，但仍在已经存在合理行动时进行干预。正确的决策标签也未能保证可用的行动、问题或修复。至关重要的是，响应要求不仅影响决策的表达方式，还影响所做的决策。明确所需内容显著提高了完全正确的响应，并可能改变干预决策，即使输出已经可解析。这些发现强调，可靠的代理不仅需要识别不确定性，还需要仅在必要时进行干预，并将所选择的下一步转化为可验证的响应。
cs.CL / 42 / 2610.03109
Emergent Structure in the Marginal Attention Space of Language Models
语言模型边际注意空间中的新兴结构
Abstract
While representation similarity across independently trained language models is well-documented, how internal mechanics such as attention behave across models remains far less characterized. Inspired by this gap, we examine the structure of post-softmax attention weights by marginalizing over query positions, mapping them into a joint token-head "marginal attention space". Evaluating across 60+ diverse LLMs, we find that different properties emerge when reducing this space along its token and head axes. When reduced token-wise, marginal attention yields a text-intrinsic signal robustly conserved across models. To explain this property, we empirically connect marginal attention to the input-output Jacobian of the network, and prove theoretically that under a smoothness assumption, models with similar next-token distributions are guaranteed to have similar input-output Jacobian statistics. When reduced head-wise, it forms a model-private signature conserved across documents. Practically, this provides a natural way to estimate a per-head budget for key-value (KV) cache eviction, effectively decoupling model-specific budget allocation from text-intrinsic token scoring. On standard eviction benchmarks, a per-head budget precomputed offline on pretraining text, combined with a training-free token score, shows competitive performance with methods that recompute the budget on every document or train it per target. Code available at https://github.com/Flegyas/marginal-attention
Chinese Translation
虽然独立训练的语言模型之间的表示相似性已有充分文献记录，但模型内部机制（如注意力）的行为仍然缺乏深入表征。受到这一空白的启发，我们通过对查询位置进行边际化，考察后软最大（post-softmax）注意力权重的结构，将其映射到一个联合的标记-头“边际注意空间”。在60多个不同的语言大模型（LLM）上进行评估时，我们发现沿着标记和头轴缩减该空间时会出现不同的特性。当按标记缩减时，边际注意力产生了一种在模型间稳健保留的文本内在信号。为了说明这一特性，我们实证地将边际注意力与网络的输入-输出雅可比（Jacobian）联系起来，并理论证明在平滑性假设下，具有相似下一个标记分布的模型保证具有相似的输入-输出雅可比统计。当按头缩减时，它形成了一个在文档间保留的模型私有特征。实际上，这为关键-值（KV）缓存驱逐的每头预算估计提供了一种自然的方法，有效地将模型特定的预算分配与文本内在的标记评分解耦。在标准驱逐基准测试中，基于预训练文本离线预计算的每头预算，结合无训练的标记评分，显示出与在每个文档上重新计算预算或针对每个目标训练的方式具有竞争力的性能。代码可在 https://github.com/Flegyas/marginal-attention 获取。
cs.CL / 43 / 2610.03110
Ontological Instability and Statistical Amplification: The Paradox of "Humanizing" LLM-Generated Text
本体不稳定性与统计放大：对“人性化”LLM生成文本的悖论
Abstract
Supervised AI-text detectors report high benchmark accuracy, but it is not clear what their decisions are based on. We analyze a RoBERTa-based detector under semantic, structural, and tokenizer-level perturbations, using the M4 dataset (N = 10,000) and controlled generations (N = 300). When Mistral-7B-Instruct was asked to make machine text sound more human, Verb Diversity rose from 0.77 to 0.92 and the outputs became easier to detect. Detection scores appear to track statistical complexity, which also leads to a 76.3% false-positive rate on formal human writing. As a control, we evaluate event-based Latent Space detection. Paraphrasing changed 87% of its event sequences (Jaccard = 0.067), and homoglyphs altered 70% of the extracted verbs even though extraction still ran (Jaccard = 0.30). Its best domain AUC was 0.577. RoBERTa's robustness seems specific to the features it uses, and structural abstraction did not make detection more robust.
Chinese Translation
监督式人工智能文本检测器报告了高基准准确率，但其决策依据尚不明确。我们分析了一种基于RoBERTa的检测器在语义、结构和分词器级别扰动下的表现，使用了M4数据集（N = 10,000）和受控生成（N = 300）。当要求Mistral-7B-Instruct使机器生成的文本听起来更人性化时，动词多样性从0.77上升至0.92，输出结果变得更容易被检测。检测分数似乎与统计复杂性相关，这也导致了对正式人类写作的76.3%误报率。作为对照，我们评估了基于事件的潜在空间检测。改写改变了87%的事件序列（Jaccard = 0.067），而同形异义词则改变了70%的提取动词，尽管提取仍然进行（Jaccard = 0.30）。其最佳领域AUC为0.577。RoBERTa的鲁棒性似乎特定于其使用的特征，而结构抽象并未使检测更为鲁棒。
cs.CL / 44 / 2610.03112
Building Interpretable Feature Representations for Resume-Vacancy Matching by Distilling Production LLM Signals
通过提炼生产大型语言模型信号构建可解释的简历-职位匹配特征表示
Abstract
Matching candidates to vacancies is central to recruitment, and a recruiter needs to see why a candidate fits, not only a single opaque relevance score. We provide this evidence as named, interpretable matching dimensions recruiters can act on - eight in our current deployment. We propose a two-part approach. The first is an LLM-based labeler whose prompts and feature definitions were refined from recruiter feedback while it served as an earlier production matching stage. In the current architecture, it is used only for offline labeling and is not called on online requests. The second is a feature bi-encoder distilled from it: a LoRA-adapted embedding backbone with compact per-dimension heads that runs on CPU and serves all online requests. Both parts keep improving: prompts are revised as feedback arrives, and the bi-encoder is retrained on the updated labels. The model is trained on 168,772 labeled vacancy-resume pairs (17,921 vacancies and 180,030 resumes). Recruiters using the service can confirm or revise surfaced feature predictions. On 927 recruiter-recorded values from this selected production-feedback subset, the deployed student agrees with the recorded decisions in 888 cases (95.79%). This is operational, non-blinded agreement rather than an independent human evaluation.
Chinese Translation
将候选人与职位进行匹配是招聘的核心，招聘人员需要看到候选人适合的原因，而不仅仅是一个不透明的相关性评分。我们提供这种证据，作为招聘人员可以采取行动的命名、可解释的匹配维度——在我们当前的部署中有八个。我们提出了一种两部分的方法。第一部分是基于大型语言模型（LLM）的标签生成器，其提示和特征定义是根据招聘人员的反馈进行优化的，同时它作为早期的生产匹配阶段。当前架构中，它仅用于离线标记，并不用于在线请求。第二部分是从其提炼出的特征双编码器：一个经过LoRA适配的嵌入骨干，具有紧凑的每维头部，能够在CPU上运行并服务于所有在线请求。这两部分都在不断改进：随着反馈的到来，提示会被修订，而双编码器会在更新的标签上进行再训练。该模型在168,772对标记的职位-简历对（17,921个职位和180,030份简历）上进行训练。使用该服务的招聘人员可以确认或修订呈现的特征预测。在来自这一选定生产反馈子集的927个招聘人员记录的值中，部署的学生模型在888个案例中与记录的决策一致（95.79%）。这是一种操作性的、非盲目的一致性，而不是独立的人类评估。
cs.CL / 45 / 2610.03136
Investigating the Role of Reasoning-Language Alignment in Monolingual Retrieval-Augmented Generation
研究推理语言对单语检索增强生成的影响
Abstract
Reasoning traces improve large language models (LLMs), but current models are trained to reason mostly in English. It has been shown that forcing a model to reason in another language degrades accuracy, even when the reasoning language matches the language of the prompt -- but only for a setting where the model reasons over a short prompt. Here, we ask whether the same holds for retrieval-augmented generation (RAG), where the model must read and integrate a large amount of retrieved evidence in the target language. To study this, we build a fully monolingual German RAG question-answering testbed over the fictional world of the tabletop role-playing game The Dark Eye, a domain that is richly documented in German but too niche for the model to answer from memory, so that it has to rely on retrieval. Varying the forced reasoning language of an agentic RAG system on this testbed, we find that aligning the reasoning language with the language of the query and the retrieved documents helps. Forced German reasoning outperforms forced French, although the model benchmarks higher in French, so the benefit comes from alignment and not from language proficiency. The advantage grows when the retrieved context is richer and structure-aware. However, forced German only reaches the level of the model's native, unconstrained English reasoning without surpassing it, showing that native multilingual reasoning is needed. We publicly release the testbed and QA benchmark.
Chinese Translation
推理轨迹改善了大型语言模型（LLMs），但当前模型主要以英语进行推理。研究表明，强迫模型使用其他语言进行推理会降低准确性，即使推理语言与提示语言匹配——但仅在模型对短提示进行推理的情况下。在此，我们探讨这一现象是否同样适用于检索增强生成（RAG），在该设置中，模型必须在目标语言中读取和整合大量检索到的证据。为此，我们构建了一个完全单语的德语RAG问答测试平台，基于桌面角色扮演游戏《黑暗之眼》的虚构世界，该领域在德语中有丰富的文献记录，但对于模型来说过于小众，无法仅凭记忆回答，因此必须依赖检索。在这个测试平台上，改变一个主动RAG系统的强制推理语言，我们发现将推理语言与查询语言和检索文档的语言对齐是有帮助的。强制德语推理的表现优于强制法语推理，尽管模型在法语中的基准分数更高，因此这一优势来自于对齐而非语言能力。当检索到的上下文更丰富且结构感知时，优势会进一步增强。然而，强制德语推理仅达到模型母语、无约束的英语推理水平，并未超越，显示出本土多语种推理的必要性。我们将公开发布该测试平台和问答基准。
cs.CL / 46 / 2610.03163
Predicting Steering Vectors and Adapter Weights for Few-Shot Author-Style Transfer
预测少样本作者风格迁移的引导向量和适配器权重
Abstract
Adapting large language models to an individual author's style from a few examples is challenging, and scientific writing sharpens the difficulty: formal conventions leave little surface variation, and authors write about their own topics, so extracted ``style'' easily entangles with content. We study style-conditioned abstract generation from a few example abstracts per author and propose three methods: (1) contrastive activation steering, (2) a network that predicts steering vectors, and (3) a hypernetwork that predicts LoRA adapters. We find a consistent trade-off between style imitation and output quality: fine-tuning buys most of the available style signal but forfeits fluency, while the hypernetwork achieves the best trade-off on both seen and unseen authors. Our steering operates at author level, contrasting an author's abstracts against style-neutral generations for the same content. This holds topic fixed, removes the need for a predefined style inventory, and outperforms inventory-based steering. % [EDIT 1a] softened "no single optimal axis" claim Moreover, our analyses demonstrate that manually extracted and predicted steering vectors are near-orthogonal yet score comparably, indicating that style conditioning here can admit at least two unrelated directions rather than requiring one particular axis.
Chinese Translation
将大型语言模型适应于个别作者的风格仅凭少量示例是具有挑战性的，科学写作更是加剧了这一难度：正式的写作规范几乎没有表面变化，且作者通常围绕自己的主题进行写作，因此提取的“风格”容易与内容交织在一起。我们研究了基于少量示例摘要的风格条件下的摘要生成，并提出了三种方法：(1) 对比激活引导，(2) 预测引导向量的网络，以及 (3) 预测 LoRA 适配器的超网络。我们发现风格模仿与输出质量之间存在一致的权衡：微调能够获取大部分可用的风格信号，但牺牲了流畅性，而超网络在已知和未知作者之间实现了最佳的权衡。我们的引导操作在作者层面进行，将作者的摘要与相同内容的风格中立生成进行对比。这保持了主题固定，消除了对预定义风格库存的需求，并且优于基于库存的引导。此外，我们的分析表明，手动提取和预测的引导向量几乎是正交的，但得分相近，这表明在这里的风格条件可以接受至少两个不相关的方向，而不需要一个特定的轴。
cs.CL / 47 / 2610.03176
Hindsight-Guided Rationale Distillation for Rare Disease Diagnosis
基于事后指导的稀有疾病诊断推理蒸馏
Abstract
We study hindsight-guided distillation for rare disease diagnosis on ZebraMap: a 1.5B student is fine-tuned on chain-of-thought traces from a 8B teacher that observes the ground-truth diagnosis during generation. Absolute accuracy remains low for all models - the task is hard at this scale - but within this ceiling a filtered variant (StudentF) achieves a small, statistically significant accuracy advantage over the teacher (p < 0.001), concentrated in better-represented diseases. The unfiltered student does not significantly outperform the teacher (p = 0.129), establishing that contamination filtering - not hindsight distillation alone - drives the gain. The gap traces to an artifact we term GT hallucination. Label-visible generation causes the teacher to embed "ground truth is X" phrases in its reasoning chain; SFT copies the pattern. At inference, the unfiltered student reproduces the phrase in 33.9% of cases, with severe accuracy degradation when the hallucinated label is wrong. A regex filter removing these slots reduces contamination to near-zero, producing the observed gain - though the effect remains small. We precisely quantify this gain-cost tradeoff, document frequency-dependent knowledge transfer absent from the RL-trained teacher, and characterize a calibration gap that SFT does not close - identifying both as directions for future work.
Chinese Translation
我们研究了在ZebraMap上进行稀有疾病诊断的事后指导蒸馏：一个1.5B的学生模型在一个8B的教师模型的思维链迹象上进行了微调，该教师在生成过程中观察到了真实的诊断。所有模型的绝对准确率仍然较低——在这个规模上任务非常困难——但在这个上限内，一个经过过滤的变体（StudentF）在准确性上相较于教师模型取得了小幅且具有统计学意义的优势（p < 0.001），这一优势集中在表现更好的疾病上。未经过滤的学生模型并未显著优于教师模型（p = 0.129），这表明污染过滤——而非仅仅是事后蒸馏——推动了这一提升。该差距源于我们称之为“真实标签幻觉”的伪影。标签可见的生成导致教师在其推理链中嵌入“真实标签是X”的短语；SFT（监督微调）复制了这一模式。在推理时，未经过滤的学生模型在33.9%的情况下重现了这一短语，当幻觉标签错误时，准确性严重下降。一个去除这些槽位的正则表达式过滤器将污染减少到接近零，从而产生了观察到的提升——尽管这一效果仍然较小。我们精确量化了这一增益-成本权衡，记录了缺失于强化学习训练教师的频率依赖知识转移，并描述了SFT未能弥补的校准差距——将这两者识别为未来工作的方向。
cs.CL / 48 / 2610.03190
Not Until the Evidence Says So: Teaching LLM Investigators When to Close a Case
直到证据表明如此：教导大型语言模型调查员何时关闭案件
Abstract
Accident, defect and outage investigations end with a decision that ordinary question answering never faces: whether the evidence gathered so far is enough to close the case. We study this decision for LLM investigators, which request evidence from a case file, revise their hypotheses, and either close the case with a conclusion grounded in what they read or leave it open and name what is missing. This judgment does not come with capability: an untrained 9B model overstates its evidence in 97% of its answers, and a frontier model that identifies the right cause in 84% of cases still overstates in 91% and closes 17 of the 41 cases whose official finding is "cause undetermined". Measuring it is also non-trivial: the source of a case largely predicts its label, and a rule that reads only the source reaches 83.0 balanced accuracy on our test cases. We therefore evaluate closure with three tests: closure accuracy, reported against this rule and within each source; evidence dependence, which removes the grounds of a conclusion and checks whether the model stops closing; and conclusion and gap quality, a judged checklist of what the model asserts and what it says is missing. We build Nautil, 731 audited cases from aviation, rail, maritime, chemical-safety and vehicle-defect reports and production server incidents, with teacher trajectories, an out-of-distribution test set and counterfactual evidence versions. Fine-tuning a 9B model on these trajectories makes its closures follow the evidence: removing the grounds lowers its closure rate by 26 points relative to a matched control, overstatement falls from 97% to 35%, and correct, non-overstated conclusions rise from 3% to 43%. Reinforcement learning that rewards only the closure decision then raises balanced accuracy from 69.2 to 83.3, on par with the teacher, and within-source accuracy from 60.4 to 74.1, at some cost in evidence dependence.
Chinese Translation
事故、缺陷和停机调查的结束伴随着一个普通问答所未曾面临的决策：迄今为止收集的证据是否足够关闭案件。我们研究了大型语言模型（LLM）调查员在这一决策中的表现，他们从案件文件中请求证据，修正假设，并根据所阅读的内容决定是关闭案件并得出结论，还是保持案件开放并指出缺失的部分。这个判断并不伴随能力：一个未经训练的9B模型在97%的回答中夸大了其证据，而一个在84%的案件中能够识别正确原因的前沿模型仍然在91%的情况下夸大，并关闭了41个案件中官方结论为“原因未确定”的17个。测量这一点也并非易事：案件的来源在很大程度上预测了其标签，而仅读取来源的规则在我们的测试案例中达到了83.0的平衡准确率。因此，我们通过三个测试来评估关闭决策：关闭准确率，相对于这一规则和每个来源进行报告；证据依赖性，去除结论的依据并检查模型是否停止关闭；以及结论和缺口质量，一个评估模型所主张内容及其所称缺失内容的检查表。我们构建了Nautil，731个来自航空、铁路、海事、化学安全和车辆缺陷报告及生产服务器事件的审核案例，包含教师轨迹、一个分布外测试集和反事实证据版本。在这些轨迹上对9B模型进行微调，使其关闭决策遵循证据：去除依据使其关闭率相对于匹配对照降低了26个百分点，夸大率从97%降至35%，正确且未夸大的结论从3%上升至43%。仅奖励关闭决策的强化学习将平衡准确率从69.2提高至83.3，达到教师水平，而来源内准确率从60.4提高至74.1，尽管在证据依赖性上有所牺牲。
cs.CL / 49 / 2610.03195
Source Preference in the Wild: How LLM Agents Favor Items by Source, and How to Reduce It
野外中的来源偏好：大型语言模型代理如何根据来源偏好项目，以及如何减少这种偏好
Abstract
As LLM agents decide on users' behalf which product to buy, which hotel to book, or which paper to cite, a preference for items from certain sources (the sites or services they come from) shapes what users receive and which sources are selected. We study source preference in end-to-end search with 12 agent models across three domains. Comparing items from different sources that satisfy the same requirements at the same position, we find that each model prefers some sources and avoids others in every domain, largely agreeing on which. This preference can outweigh how well items satisfy the request: an item satisfying one requirement fewer is selected about two-thirds of the time when it comes from a preferred source and the better one from a dispreferred source, but almost never in the reverse case. The information identifying an item's source affects selection by itself: hiding it weakens the preference, and relabeling an item with a preferred source raises its selection rate. We test two routes to this preference: training that rewards better items can make a source a shortcut for requirement satisfaction, and missing information can trigger preconceptions about the source. Supplying missing information or a prompt countering these preconceptions reduces source preference.
Chinese Translation
当大型语言模型（LLM）代理代表用户决定购买哪个产品、预订哪个酒店或引用哪篇论文时，来自某些来源（即它们所来自的网站或服务）的项目偏好会影响用户所接收到的内容以及所选择的来源。我们研究了在端到端搜索中12个代理模型在三个领域的来源偏好。比较来自不同来源、满足相同要求且处于相同位置的项目，我们发现每个模型在每个领域都偏好某些来源并避免其他来源，且在偏好的来源上大致一致。这种偏好可能会超过项目满足请求的程度：当来自偏好的来源时，满足一个要求较少的项目被选择的概率约为三分之二，而来自不偏好来源的更好项目几乎不会被选择。识别项目来源的信息本身会影响选择：隐藏该信息会削弱偏好，而将项目重新标记为来自偏好来源会提高其选择率。我们测试了导致这种偏好的两条路径：奖励更好项目的训练可以使来源成为满足要求的捷径，而缺失的信息可能会引发对来源的先入之见。提供缺失的信息或反驳这些先入之见的提示可以减少来源偏好。
cs.CL / 50 / 2610.03215
StanceEval 2026: The Second Stance Detection Shared Task
StanceEval 2026：第二届立场检测共享任务
Abstract
StanceEval 2026 is the second edition of the StanceEval shared task series on stance detection in Arabic social media text. Stance detection aims to identify a writer's stance toward a given topic. Given a tweet and a target, participating systems must determine whether the writer's stance is Favor, Against, or None. This edition focuses on cross-target generalization across two distinct evaluation tracks: Track 1 evaluates thematically related cross-target transfer (testing on Women Driving, related to Women Empowerment from training data), while Track 2 evaluates cross-domain transfer to completely unseen targets (E-Cars and Trimester System). The shared task attracted 80 registered teams from 12 countries. During the evaluation phase, 30 unique teams submitted entries, with 21 teams officially ranked in Track 1 and 13 in Track 2 following validation filtering, and 20 teams submitting system-description papers. Participating teams employed diverse methodologies, including fine-tuned pretrained language models, prompt-based and retrieval-augmented large language models (LLMs), fine-tuned LLMs, and hybrid cascades. Top systems achieved impressive $F_{avg2}$ scores of 0.8994 on Track 1 and 0.9400 on Track 2, substantially outperforming the strongest baselines (0.7366 and 0.7475, respectively), where $F_{avg2}$ denotes the macro-averaged F1 score over the Favor and Against classes. Counterintuitively, performance on the unseen targets was higher than on the related target, a disparity could be driven by extreme target polarization, class imbalance, and dialectal or sarcastic nuance across topics.
Chinese Translation
StanceEval 2026 是关于阿拉伯社交媒体文本立场检测的 StanceEval 共享任务系列的第二版。立场检测旨在识别作者对特定主题的立场。给定一条推文和一个目标，参与系统必须确定作者的立场是支持（Favor）、反对（Against）还是无立场（None）。本届比赛重点关注跨目标泛化，分为两个不同的评估轨道：轨道 1 评估主题相关的跨目标迁移（在与女性赋权相关的训练数据上测试女性驾驶（Women Driving）），而轨道 2 则评估对完全未见目标的跨领域迁移（电动汽车（E-Cars）和学期制度（Trimester System））。该共享任务吸引了来自 12 个国家的 80 支注册队伍。在评估阶段，30 支独特的队伍提交了参赛作品，其中 21 支队伍在轨道 1 中获得正式排名，13 支在轨道 2 中获得排名，经过验证筛选后，20 支队伍提交了系统描述论文。参与队伍采用了多种方法，包括微调的预训练语言模型、基于提示和检索增强的大型语言模型（LLMs）、微调的 LLMs 和混合级联。顶尖系统在轨道 1 上取得了 0.8994 的 $F_{avg2}$ 分数，在轨道 2 上取得了 0.9400 的分数，显著优于最强基线（分别为 0.7366 和 0.7475），其中 $F_{avg2}$ 表示对支持和反对类别的宏平均 F1 分数。出乎意料的是，在未见目标上的表现高于相关目标，这一差异可能是由极端目标两极化、类别不平衡以及跨主题的方言或讽刺细微差别所驱动。
cs.CL / 51 / 2610.03240
Collective Bias Mitigation via Model Routing and Collaboration
通过模型路由与协作实现集体偏见缓解
Abstract
Large language models (LLMs) are increasingly deployed in public health, finance, and governance, requiring both accuracy and societal value alignment. Despite recent advances, LLMs often perpetuate or amplify bias embedded in their training data, posing challenges to fairness. While self-debiasing encourages an LLM to identify and correct its own biases, relying on a single model's intrinsic knowledge may be insufficient to address deeply ingrained stereotypes. To address this limitation, we introduce Collective Bias Mitigation (CBM), a framework that alleviates bias by learning fine-grained model behavior and fostering knowledge sharing among diverse LLMs. This work is the first to systematically explore the effective selection and organization of distinct LLMs to cultivate fairer LLM responses. Experiments show CBM substantially outperforms standalone baselines (e.g., in the top-7 setting, Committee lowers the age bias score from 0.25 to 0.10). Our Debating and Committee topologies achieve substantial bias reduction, with the latter balancing mitigation effectiveness and inference cost, highlighting the potential of CBM for fairer LLMs.
Chinese Translation
大型语言模型（LLMs）在公共卫生、金融和治理等领域的应用日益增多，这要求它们在准确性和社会价值对齐方面表现良好。尽管最近取得了一些进展，LLMs仍然常常延续或加剧其训练数据中嵌入的偏见，给公平性带来了挑战。虽然自我去偏见化鼓励LLM识别并纠正自身的偏见，但依赖单一模型的内在知识可能不足以解决根深蒂固的刻板印象。为了解决这一局限性，我们提出了集体偏见缓解（Collective Bias Mitigation, CBM）框架，通过学习细致的模型行为并促进不同LLMs之间的知识共享来减轻偏见。本研究首次系统性地探索了有效选择和组织不同LLMs以培养更公平的LLM响应。实验结果表明，CBM在性能上显著优于独立基线（例如，在前7名设置中，委员会将年龄偏见分数从0.25降低到0.10）。我们的辩论和委员会拓扑结构实现了显著的偏见减少，后者在缓解效果和推理成本之间取得了良好的平衡，突显了CBM在实现更公平的LLMs方面的潜力。
cs.CL / 52 / 2610.03268
Shrome at Touch\'e: Soft-Vote Ensembling and Counter-Causal Augmentation for Causality Extraction
Shrome at Touché：软投票集成与反因果增强用于因果关系提取
Abstract
Touch\'e 2026 extends causality extraction to counter-causal claims: news sentences whose surface form appears causal but whose meaning denies the causation, as in "It is falsely believed that X caused Y." A system that relies on surface cues such as "caused" or "led to" will accept such a sentence as causal and give it the wrong polarity. On the Countercausal News Corpus (CCNC), the task has three subtasks: deciding whether a sentence is causal (detection), locating its cause and effect spans (extraction), and labeling its polarity as procausal, counter-causal, or uncausal. We build one model per subtask. Detection is a fine-tuned classifier with a single cross-task rule that uses the extracted spans to remove false positives. For extraction, we ensemble three RoBERTa-large BILOU+CRF taggers by averaging their token-level scores before decoding, rather than voting on the spans each tagger produces. For polarity, where labeled counter-causal examples are scarcest, we add training sentences generated by a large language model prompted with nine patterns of counter-causal expression adapted from Hagen et al., keeping only those that pass automatic structural checks. On the held-out CCNC test set, the system reaches F1 0.869 on detection and macro-F1 0.817 on polarity, and in the organizers' final causal-only evaluation of extraction it scores granularity-adjusted F1 0.728, the highest extraction score among all submissions including the organizers' baseline. The development split is used only for component selection and the ablations reported in the paper.
Chinese Translation
Touché 2026 将因果关系提取扩展到反因果声明：那些表面形式看似因果但其含义否定因果关系的新闻句子，例如 "人们错误地认为 X 导致了 Y"。一个依赖于表面线索（如 "导致" 或 "引起"）的系统会将这样的句子视为因果句，并给出错误的极性。在反因果新闻语料库（Countercausal News Corpus, CCNC）中，该任务包含三个子任务：判断句子是否为因果句（检测）、定位其因果跨度（提取）以及标记其极性为支持因果、反因果或非因果。我们为每个子任务构建一个模型。检测是一个经过微调的分类器，使用单一的跨任务规则，利用提取的跨度来消除假阳性。对于提取，我们通过在解码之前对三个 RoBERTa-large BILOU+CRF 标注器的标记级分数进行平均来进行集成，而不是对每个标注器产生的跨度进行投票。对于极性，由于标记的反因果示例最为稀缺，我们添加了由大型语言模型生成的训练句子，这些句子是基于 Hagen 等人提出的九种反因果表达模式进行提示生成的，仅保留那些通过自动结构检查的句子。在保留的 CCNC 测试集上，该系统在检测任务上达到 F1 0.869，在极性任务上达到宏 F1 0.817，并且在组织者的最终因果提取评估中，其得分为粒度调整后的 F1 0.728，是所有提交中（包括组织者的基线）最高的提取得分。开发集仅用于组件选择和论文中报告的消融实验。
cs.CL / 53 / 2610.03324
To Jev or Not? Evaluating the Accuracy and Efficiency of Structured Decision Models for Hate-Speech Moderation
选择Jev还是不选择Jev？评估结构化决策模型在仇恨言论审核中的准确性和效率
Abstract
The scale of online content makes hate-speech moderation challenging, while Large Language Models (LLMs) enable harmful material to be produced and adapted more easily. Moderation therefore requires efficient classifiers that can accommodate different definitions of hate speech. Recent structured decision models accept natural-language criteria and select among specified answers, raising the question of whether they can meet these requirements without task-specific training. We present HATEDECIDE, an evaluation of six decision-model configurations on four hate-speech datasets against specialized moderation, zero-shot, commercial, and supervised baselines. We examine whether supplying a dataset's definition, or decomposing it into multiple questions, improves classification, and we measure their latency and cost. We find that commercial LLMs significantly outperform all decision models on only one dataset. Supplying definitions changes up to 28\% of predictions without consistently improving classification, and decomposition significantly improves performance in only 20\% of the comparisons. On a diagnostic set of test cases, the best hosted decision model comes within 1.6 macro-F1 points of the best commercial LLM at approximately 97\% lower inference cost. These results identify opportunities for inexpensive moderation, while showing that explicit criteria and additional questions do not reliably improve classification.
Chinese Translation
在线内容的规模使得仇恨言论的审核变得具有挑战性，而大型语言模型（LLMs）则使得有害材料的生成和适应变得更加容易。因此，审核需要能够适应不同仇恨言论定义的高效分类器。最近的结构化决策模型接受自然语言标准，并在指定答案中进行选择，这引发了一个问题：它们是否能够在没有特定任务训练的情况下满足这些要求。我们提出了HATEDECIDE，评估六种决策模型配置在四个仇恨言论数据集上的表现，并与专业审核、零样本、商业和监督基准进行比较。我们考察了提供数据集定义或将其分解为多个问题是否改善分类，并测量其延迟和成本。我们发现商业LLMs在仅一个数据集上显著优于所有决策模型。提供定义改变了多达28%的预测，但并未始终改善分类，而分解在仅20%的比较中显著提高了性能。在一组诊断测试案例中，最佳的托管决策模型的宏F1分数比最佳商业LLM低1.6分，推理成本约低97%。这些结果识别了低成本审核的机会，同时表明明确的标准和额外的问题并未可靠地改善分类。
cs.CL / 54 / 2610.03329
SyntaxBench: A Statistical Diagnostic Framework for Character-Level Reasoning in Large Language Models
SyntaxBench：一个用于大语言模型字符级推理的统计诊断框架
Abstract
Large language models are increasingly used where small syntactic errors matter, yet character-level reasoning is still evaluated mostly through isolated probes and aggregate accuracy. We introduce SyntaxBench, a diagnostic benchmark and statistical evaluation framework for character-level reasoning. It contains five core tasks, character counting, letter containment, palindrome detection, edit distance, and longest-string selection, plus index_to_span, a harder substring-extraction stress test. The five core tasks use paired English and character-length-matched random-string inputs. index_to_span documents share a 200-500 word band and are not character-length matched. All six tasks use zero-, one-, and four-shot prompts. We evaluate eight open-weight models from 2B to 32B parameters across 11 reasoning-mode configurations. The framework reports exact-match and relaxed accuracy, Cohen's kappa, paired McNemar tests with odds ratios, bootstrap confidence intervals, Kendall's tau, class-conditional metrics, tokenization analysis, and multiple-comparison-corrected tests. Three findings stand out. First, tokenization shapes accuracy: random strings are more character-visible than English strings (1.892 vs. 3.169 characters per token), and character-counting accuracy falls as English words occupy more tokens. Second, reasoning mode is not uniformly helpful: Gemma4-31B is nearly unchanged across modes on the near-saturated tasks, while Qwen3.6-27B is worse with thinking on palindrome detection (0.952 non-thinking vs. 0.886 thinking at four-shot). Third, index_to_span remains largely unsolved; the best four-shot exact-match accuracy is 6.75%. Character-level evaluation needs controlled inputs, paired tests, and analyses of tokenization and reasoning mode rather than aggregate accuracy alone.
Chinese Translation
大语言模型在小的语法错误影响较大的场景中越来越多地被使用，但字符级推理仍然主要通过孤立的探测和总体准确性进行评估。我们引入了SyntaxBench，一个用于字符级推理的诊断基准和统计评估框架。它包含五个核心任务：字符计数、字母包含、回文检测、编辑距离和最长字符串选择，以及一个更具挑战性的子字符串提取压力测试任务index_to_span。五个核心任务使用配对的英语和字符长度匹配的随机字符串输入。index_to_span文档共享200-500字的范围，并且不进行字符长度匹配。所有六个任务都使用零次、一轮和四轮提示。我们评估了八个参数从2B到32B的开放权重模型，涵盖11种推理模式配置。该框架报告精确匹配和放宽准确性、Cohen's kappa、成对McNemar检验及其比值、引导置信区间、Kendall's tau、类条件指标、分词分析和多重比较校正检验。有三个发现尤为突出。首先，分词影响准确性：随机字符串的字符可见性高于英语字符串（每个标记1.892 vs. 3.169个字符），而随着英语单词占用更多标记，字符计数的准确性下降。其次，推理模式并不总是有帮助：在接近饱和的任务中，Gemma4-31B在各模式下几乎没有变化，而Qwen3.6-27B在回文检测的思考模式下表现更差（四轮提示下非思考模式0.952 vs. 思考模式0.886）。第三，index_to_span仍然在很大程度上未解决；最佳的四轮提示精确匹配准确性为6.75%。字符级评估需要受控输入、配对测试，以及对分词和推理模式的分析，而不仅仅是总体准确性。
cs.CL / 55 / 2610.03421
CLIMB: Confidence-Guided Complementary Evidence for Multimodal Retrieval-Augmented Generation
CLIMB：基于信心引导的多模态检索增强生成的互补证据
Abstract
Multimodal large language models (MLLMs) have shown strong visual reasoning abilities, but knowledge-intensive visual question answering often requires external textual evidence beyond the image and the model's parametric knowledge. Existing multimodal RAG systems commonly rely on Top-$K$ retrieval or reranking, which may return redundant passages and provide limited control over whether an answer update is sufficiently supported by the retrieved evidence. We propose \textit{CLIMB}, a training-free inference-time framework for multimodal RAG. CLIMB first constructs a compact complementary evidence pool using an MMR-style objective that balances query relevance and passage-level redundancy. It then performs confidence-controlled refinement within this fixed pool: an R/E/C critic scores passages by relevance, evidence specificity, and cross-modal alignment, while an evidence-grounded confidence estimator accepts an updated answer only when the estimated confidence increases. This design provides a simple stopping criterion and reduces unnecessary refinement without modifying the underlying retriever or MLLM. Experiments on Encyclopedic-VQA and InfoSeek show that CLIMB consistently improves over retrieval-augmented multimodal baselines. Ablations further indicate that complementary pooling, critic-based scoring, and iterative confidence-controlled refinement each contribute to the final performance.
Chinese Translation
多模态大型语言模型（MLLMs）展现出强大的视觉推理能力，但知识密集型的视觉问答通常需要超出图像和模型参数知识的外部文本证据。现有的多模态检索增强生成（RAG）系统通常依赖于Top-$K$检索或重排序，这可能返回冗余的段落，并且对答案更新是否得到检索证据的充分支持提供有限的控制。我们提出了 extit{CLIMB}，一个无训练的推理时框架，用于多模态RAG。CLIMB首先使用MMR风格的目标构建一个紧凑的互补证据池，该目标平衡查询相关性和段落级冗余。然后，它在这个固定池中进行信心控制的细化：一个R/E/C评论者根据相关性、证据特异性和跨模态对齐对段落进行评分，而一个基于证据的信心估计器仅在估计的信心增加时接受更新的答案。该设计提供了一个简单的停止标准，并减少了不必要的细化，而无需修改底层检索器或MLLM。在Encyclopedic-VQA和InfoSeek上的实验表明，CLIMB在检索增强的多模态基线之上始终取得了改进。消融实验进一步表明，互补池、基于评论者的评分和迭代信心控制的细化各自对最终性能做出了贡献。
cs.CL / 56 / 2610.03482
Single-Pass Uncertainty Heads for Claim-Level Hallucination Detection in Persian Medical Language Models
单次不确定性头用于波斯医学语言模型中的声明级幻觉检测
Abstract
Hallucination detection is particularly important for medical language models, but repeated-sampling approaches are expensive and existing uncertainty-head resources do not directly transfer to a new backbone and language. We adapt the LLM Uncertainty Head (LUH) framework to Aya-Expanse-8B-based Persian medical models, using Gaokerena-V and Gaokerena-R as two previously developed backbones. We first examine response variability on a 168-question Iranian medical entrance examination and observe substantially lower five-run consistency for Gaokerena-V than for Aya-Expanse-8B, whereas Gaokerena-R is comparable to Aya-Expanse-8B. We then construct two paired claim-level hallucination datasets directly in Persian, containing 1,600 responses for each backbone, and train lightweight claim-level heads on frozen backbone attention maps and token probabilities. On held-out test splits, the heads obtain PR-AUCs of 0.4820 and 0.4652, corresponding to 2.30 and 2.66 times their respective random baselines, and ROC-AUCs of 0.7852 and 0.7810. The heads require neither retrieval nor repeated sampling at inference time. These results provide an initial study of single-pass claim-level uncertainty estimation for Persian medical language models; the test splits are small and the labels are automatically generated.
Chinese Translation
幻觉检测对于医学语言模型尤为重要，但重复采样方法成本高昂，现有的不确定性头资源无法直接迁移到新的基础模型和语言。我们将LLM不确定性头（LUH）框架适配于基于Aya-Expanse-8B的波斯医学模型，使用Gaokerena-V和Gaokerena-R作为两个先前开发的基础模型。我们首先在168个问题的伊朗医学入学考试中检查响应的变异性，观察到Gaokerena-V的五次运行一致性显著低于Aya-Expanse-8B，而Gaokerena-R与Aya-Expanse-8B相当。随后，我们直接在波斯语中构建了两个配对的声明级幻觉数据集，每个基础模型包含1600个响应，并在冻结的基础模型注意力图和标记概率上训练轻量级声明级头。在保留的测试分割上，这些头获得了0.4820和0.4652的PR-AUC，分别是其随机基线的2.30和2.66倍，以及0.7852和0.7810的ROC-AUC。这些头在推理时既不需要检索也不需要重复采样。这些结果为波斯医学语言模型的单次声明级不确定性估计提供了初步研究；测试分割较小且标签是自动生成的。
cs.CL / 57 / 2610.03515
Learning from Repaired Reasoning: Root-Cause-Guided On-Policy Distillation
从修复推理中学习：根本原因引导的在线蒸馏
Abstract
On-policy self-distillation (OPSD) uses reference solutions as privileged hindsight to supervise student-generated reasoning trajectories. However, reference-based guidance may explain a correct solution without addressing why the student's own reasoning fails. This reasoning mismatch between the guidance provided and the correction needed can encourage the student to borrow correct conclusions while leaving its reasoning errors unresolved. Moreover, applying the same hindsight throughout the trajectory risks a distillation trap, where unnecessary constraints on valid reasoning compete with correction of substantive errors. To address these issues, we propose Root-Cause-Guided On-Policy Distillation (RC-OPD), which uses repairs of the student's own reasoning to provide guidance that addresses its specific errors while building on valid progress. For each failed attempt, RC-OPD locates the earliest substantive error, develops a local correction, and uses the corrected intermediate result as an anchor for the valid prefix. An iterative diagnosis--repair--continuation process tests the repairs through student continuation, identifying further errors within a fixed repair budget. For repair chains that reach a correct answer, root--cause--guided distillation uses failure diagnoses and corrective goals to supervise the erroneous segments, while anchor-guided distillation supports the corresponding valid prefixes with reasoning chains leading to the repaired intermediate results. We evaluate RC-OPD across multiple datasets and model scales. Extensive experiments and analyses show that it mitigates reasoning mismatch and the distillation trap, yielding substantial performance gains.
Chinese Translation
在线自蒸馏（OPSD）使用参考解作为特权后见，以监督学生生成的推理轨迹。然而，基于参考的指导可能会解释正确的解决方案，而未能解决学生自身推理失败的原因。这种指导与所需修正之间的推理不匹配可能会促使学生借用正确的结论，同时使其推理错误未得到解决。此外，在整个轨迹中应用相同的后见会导致蒸馏陷阱，其中对有效推理的无谓约束与实质性错误的修正相竞争。为了解决这些问题，我们提出了根本原因引导的在线蒸馏（RC-OPD），该方法利用学生自身推理的修复来提供针对其特定错误的指导，同时建立在有效进展的基础上。对于每一次失败的尝试，RC-OPD定位最早的实质性错误，开发局部修正，并使用修正后的中间结果作为有效前缀的锚点。一个迭代的诊断-修复-继续过程通过学生的继续来测试修复，在固定的修复预算内识别进一步的错误。对于达到正确答案的修复链，根本原因引导的蒸馏使用失败诊断和修正目标来监督错误段落，而锚点引导的蒸馏则支持对应的有效前缀，利用推理链导向修复的中间结果。我们在多个数据集和模型规模上评估了RC-OPD。大量实验和分析表明，它缓解了推理不匹配和蒸馏陷阱，带来了显著的性能提升。
cs.CL / 58 / 2610.03525
Structured Composition of Verifiable Atomic Insights for Table-to-Report Generation
可验证原子洞察的结构化组合用于表格到报告生成
Abstract
Table-to-report generation refers to the task of automatically generating article-level analyt- ical reports from relational tables and is an essential capability for automated data science and decision support. Its central challenge lies in systematically discovering verifiable com- posite insights across tables, attributes, and analytical perspectives, and organizing them into coherent, complete, and traceable evidence chains. Existing methods primarily rely on sequential, reactive data agents or direct Large Language Model(LLM) generation. They suffer from exploration bias: early local observations constrain subsequent actions, causing models to focus prematurely on local analyzes and miss cross-table or cross-dimensional evidence. We propose ComInsight, which reformulates insight discovery as the composition of atomic evidences. We first define an atomic insight as the smallest executable analytical unit conforming to a predefined analysis pattern and enumerate all valid atomic insights from database schema and content. These atoms are then organized into a multi-relational insight graph, where nodes represent verified data facts and edges encode logical, temporal, or hierarchical relations. Finally, a set of composition operators systematically fuses atomic nodes into higher-order composite conclusions. Every composite output is accompanied by executable SQL and fine-grained provenance, ensuring full verifiability. Across three benchmarks InsightBench, DDR-Bench, and T2R-Bench, ComInsight consistently outperforms strong baselines in factual correctness, novelty, and structural completeness. We believe ComInsight offers a reliable, efficient, and explainable path toward table-to-report generation.
Chinese Translation
表格到报告生成是指从关系表中自动生成文章级分析报告的任务，是自动化数据科学和决策支持的重要能力。其核心挑战在于系统地发现跨表、属性和分析视角的可验证复合洞察，并将其组织成连贯、完整且可追溯的证据链。现有方法主要依赖于顺序的、反应式的数据代理或直接的大型语言模型（Large Language Model, LLM）生成。这些方法存在探索偏差：早期的局部观察限制了后续的行动，导致模型过早地专注于局部分析，错过跨表或跨维度的证据。我们提出了ComInsight，将洞察发现重新定义为原子证据的组合。我们首先将原子洞察定义为符合预定义分析模式的最小可执行分析单元，并从数据库模式和内容中枚举所有有效的原子洞察。这些原子随后被组织成一个多关系洞察图，其中节点表示经过验证的数据事实，边编码逻辑、时间或层次关系。最后，一组组合运算符系统地将原子节点融合成更高阶的复合结论。每个复合输出都伴随可执行的SQL和细粒度的来源信息，确保完全可验证性。在三个基准测试InsightBench、DDR-Bench和T2R-Bench中，ComInsight在事实正确性、新颖性和结构完整性方面始终优于强基线。我们相信ComInsight为表格到报告生成提供了一条可靠、高效且可解释的路径。
cs.CL / 59 / 2610.03531
Author Representation Strategies for Zero-Shot Authorship Attribution: A Comparative Study of LLM-Based and Embedding-Based Approaches
零样本作者归属的作者表示策略：基于大型语言模型和嵌入方法的比较研究
Abstract
Authorship Attribution (AA) requires capturing fine-grained stylistic characteristics, making it particularly challenging in zero-shot (ZS) settings where no task-specific supervision is available. In this work, we investigate the effect of author representations on ZS AA by evaluating a label-only prompting baseline together with three author representation strategies: representative writing samples, LLM-generated descriptions, and style embeddings (LISA). The first three approaches perform attribution using LLM prompting, while the embedding-based approach uses style embeddings with cosine similarity. We investigate the influence of prompt design and propose a two-stage embedding-based attribution framework that combines candidate space reduction with embedding-dimension selection. The results show that label-only ZS AA is ineffective, while incorporating author-specific representations consistently improves attribution performance. Among the evaluated approaches, the proposed two-stage LISA framework achieves the strongest overall performance, whereas LLM-generated style descriptions provide a substantially more compact representation of author style at the cost of some attribution performance. These findings demonstrate the importance of author representation in ZS AA, while indicating that current open-source LLMs remain insufficient for robust attribution without more effective representation learning.
Chinese Translation
作者归属（AA）需要捕捉细致的风格特征，这在没有任务特定监督的零样本（ZS）环境中尤为具有挑战性。在本研究中，我们通过评估仅基于标签的提示基线以及三种作者表示策略来研究作者表示对零样本作者归属的影响：代表性写作样本、LLM生成的描述和风格嵌入（LISA）。前三种方法使用LLM提示进行归属，而基于嵌入的方法则使用余弦相似度计算风格嵌入。我们研究了提示设计的影响，并提出了一种两阶段嵌入基于的归属框架，该框架结合了候选空间的缩减和嵌入维度的选择。结果表明，仅基于标签的零样本作者归属效果不佳，而结合作者特定表示则持续改善了归属性能。在评估的各方法中，所提出的两阶段LISA框架实现了最强的整体性能，而LLM生成的风格描述在一定程度上牺牲了归属性能，但提供了更为紧凑的作者风格表示。这些发现表明，在零样本作者归属中，作者表示的重要性，同时也表明当前的开源大型语言模型在没有更有效的表示学习的情况下仍不足以实现稳健的归属。
cs.CL / 60 / 2610.03565
Writerslogic at PAN 2026: Process over Content for Robust Detection under Domain Shift
Writerslogic在PAN 2026：在领域转移下以过程优于内容实现稳健检测
Abstract
We describe the Writerslogic systems for three PAN at CLEF 2026 shared tasks (Reasoning Trajectory Detection, Voight-Kampff Generative AI Detection, and Multi-Author Writing Style Analysis), unified by a shared analytical framework: feature robustness under distribution shift is governed by support overlap between training and test distributions, not by training-set effect size. This yields a taxonomy (domain-anchored, domain-portable, domain-invariant) that explains why generator-specific features die under domain shift while vocabulary fingerprints (hapax ratio, Yule's K, Heaps' exponent), compression measures, and character n-grams survive. On Reasoning Trajectory Detection, where training was entirely mathematics and 84 percent of test was unseen domains, the framework guided system design to 1st place in source detection (0.85 macro F1 via Opus-Sonnet agreement) and 3rd place in safety classification (0.66 macro F1 via query-refusal decomposition). For Voight-Kampff, we built a calibrated ensemble of DeBERTa-v2 (ONNX), multi-seed LightGBM with 44 domain-portable stylometric features, and SVM on n-gram TF-IDF, combined via learned stacking with isotonic calibration; the best configuration achieved 0.891 on the PAN 2026 test set with balanced sub-metrics (0.853 to 0.902 across all evaluation dimensions). For Multi-Author Writing Style Analysis, we describe a system fusing spectral clustering over character n-gram similarity graphs, normalized compression distance for local boundary detection, and SmolLM-135M perplexity for neural change-point detection; a platform mix-up meant our run never reached the official evaluation, so we report the design and its a priori predictions. Across all three tasks, features measuring generation process properties are designed to outperform features measuring generated content properties under domain shift.
Chinese Translation
我们描述了Writerslogic系统在CLEF 2026共享任务中的应用（推理轨迹检测、Voight-Kampff生成式人工智能检测和多作者写作风格分析），这些任务由一个共同的分析框架统一：在分布转移下，特征的稳健性由训练和测试分布之间的支持重叠决定，而不是由训练集的效应大小决定。这产生了一个分类法（领域锚定、领域可移植、领域不变），解释了为什么生成器特定特征在领域转移下消失，而词汇指纹（hapax比率、Yule的K、Heap的指数）、压缩度量和字符n-gram能够存活。在推理轨迹检测任务中，训练完全基于数学，而84%的测试数据来自未见领域，该框架指导系统设计获得了源检测的第一名（通过Opus-Sonnet一致性获得0.85的宏F1）和安全分类的第三名（通过查询拒绝分解获得0.66的宏F1）。对于Voight-Kampff任务，我们构建了一个经过校准的DeBERTa-v2（ONNX）集成模型，结合了多种种子LightGBM和44个领域可移植的风格特征，以及基于n-gram TF-IDF的支持向量机（SVM），通过学习堆叠与等距校准组合；最佳配置在PAN 2026测试集上达到了0.891的成绩，平衡子指标在所有评估维度上介于0.853到0.902之间。对于多作者写作风格分析，我们描述了一个系统，该系统融合了基于字符n-gram相似性图的谱聚类、用于局部边界检测的归一化压缩距离和用于神经变更点检测的SmolLM-135M困惑度；由于平台混淆，我们的运行未能达到官方评估，因此我们报告了设计及其先验预测。在所有三个任务中，测量生成过程属性的特征旨在在领域转移下优于测量生成内容属性的特征。
cs.CL / 61 / 2610.03567
Writerslogic at the CLEF 2026 SimpleText Track: Multi-Candidate LLM Simplification and Stacked Complexity Spotting
Writerslogic 在 CLEF 2026 SimpleText 赛道的表现：多候选 LLM 简化与层叠复杂性识别
Abstract
We describe the Writerslogic team's participation in the CLEF 2026 SimpleText shared task, addressing Task 1 (text simplification) and Task 2 (complexity spotting). For Task 1, we develop a multi-candidate generation pipeline using GPT-4o-mini that produces five simplification candidates per sentence at varying temperatures, then selects the best candidate using a reference-free scoring heuristic that rewards compression, source word retention, Cochrane Plain Language Summary vocabulary usage, and lexical simplicity. On Task 1.1 (sentence-level simplification), our Claude Sonnet 4 submission achieves SARI 47.43 and BLEU 14.21, the top-ranked sentence-level system (3rd on the combined Task 1 leaderboard, behind two document-level submissions). For Task 2, we fine-tune a DeBERTa-v3-large NLI model on 350K labeled (source, sentence) pairs, framing hallucination detection as natural language inference. The model reads the most relevant source sentence as premise and the candidate as hypothesis, directly learning to distinguish grounded from hallucinated content. On Task 2.1 (binary overgeneration identification), our fine-tuned DeBERTa system achieves 0.8081 document-level macro F1 (0.8085 in our best ensemble), the top-ranked entry within the identification track and 2nd among teams overall, behind AIIR Lab (0.8197). On Task 2.2 (multi-class error classification), our best submission reaches 0.804 multiclass accuracy, ranking 2nd among unique teams behind AIIR Lab (0.827). We evaluate both tasks on English and multilingual biomedical text from Cochrane systematic reviews.
Chinese Translation
我们描述了 Writerslogic 团队在 CLEF 2026 SimpleText 共享任务中的参与，涉及任务 1（文本简化）和任务 2（复杂性识别）。对于任务 1，我们开发了一种多候选生成管道，使用 GPT-4o-mini 每个句子生成五个不同温度的简化候选，然后使用一种无参考评分启发式方法选择最佳候选，该方法奖励压缩、源词保留、Cochrane 简明语言摘要词汇使用和词汇简单性。在任务 1.1（句子级简化）中，我们的 Claude Sonnet 4 提交实现了 SARI 47.43 和 BLEU 14.21，成为排名最高的句子级系统（在综合任务 1 排行榜中位列第三，仅次于两个文档级提交）。对于任务 2，我们在 350K 标记（源，句子）对上微调了 DeBERTa-v3-large NLI 模型，将幻觉检测框架设定为自然语言推理。该模型将最相关的源句子作为前提，将候选作为假设，直接学习区分真实内容与幻觉内容。在任务 2.1（二元过度生成识别）中，我们微调的 DeBERTa 系统实现了 0.8081 的文档级宏 F1（在我们最佳集成中为 0.8085），在识别赛道中排名第一，在所有团队中排名第二，仅次于 AIIR Lab（0.8197）。在任务 2.2（多类错误分类）中，我们的最佳提交达到了 0.804 的多类准确率，在独特团队中排名第二，仅次于 AIIR Lab（0.827）。我们在 Cochrane 系统评价的英语和多语言生物医学文本上评估了这两个任务。
cs.CL / 62 / 2610.03625
FALCON: A Model and Dataset Agnostic Framework for Synthetic Data Generation for NL2SQL Pairs
FALCON：一种模型和数据集无关的框架，用于生成NL2SQL对的合成数据
Abstract
Relational databases are among the most widely deployed forms of structured knowledge, and natural language access to them requires grounding language onto schema entities and relations while handling the ambiguity inherent in how people phrase requests. Existing synthetic NL-to-SQL data generation methods largely ignore this ambiguity and produce oversimplified queries that fail to prepare models for the complexity of real-world structured knowledge access. We present FALCON, a framework that generates realistic, ambiguity-aware NL-to-SQL data matching the complexity of challenging real-world benchmarks, at low cost using compact open models. Our approach combines reserved-word SQL seeding and persona-based prompting to generate structurally complex queries, while alignment-based filtering preserves difficulty by distinguishing genuinely incorrect examples from complex but valid queries. Human evaluation confirms consistent high quality across model sizes, and our generated data exceeds existing benchmarks in both SQL complexity and natural language richness. Difficulty-stratified analysis shows models trained on FALCON data increasingly outperform baseline-trained models as query complexity increases, validating our pipeline's success in generating challenging training data. When combined with a small proportion of existing benchmark data, mixed training recovers performance on simpler queries while preserving these advantages on complex ones. The model- and database-agnostic design enables organizations to generate high-complexity NL-to-SQL training data locally without external APIs.
Chinese Translation
关系数据库是最广泛部署的结构化知识形式之一，而自然语言访问这些数据库需要将语言与模式实体和关系相结合，同时处理人们在表达请求时固有的模糊性。现有的合成NL到SQL数据生成方法在很大程度上忽视了这种模糊性，产生了过于简化的查询，未能为模型准备应对现实世界结构化知识访问的复杂性。我们提出了FALCON，一个生成现实且具有模糊意识的NL到SQL数据的框架，匹配具有挑战性的现实世界基准的复杂性，并以低成本使用紧凑的开放模型。我们的方法结合了保留字SQL种子和基于角色的提示，以生成结构复杂的查询，同时基于对齐的过滤保留了难度，通过区分真正错误的示例与复杂但有效的查询。人类评估确认了不同模型规模之间的一致高质量，我们生成的数据在SQL复杂性和自然语言丰富性方面超越了现有基准。难度分层分析显示，基于FALCON数据训练的模型在查询复杂性增加时表现出越来越优于基线训练模型的趋势，验证了我们生成具有挑战性的训练数据的管道的成功。当与少量现有基准数据结合时，混合训练在保持复杂查询优势的同时恢复了对简单查询的性能。模型和数据库无关的设计使组织能够在本地生成高复杂度的NL到SQL训练数据，而无需外部API。
cs.CL / 63 / 2610.03695
Language Models that Play Chess and Explain Their Moves
能够下棋并解释其走法的语言模型
Abstract
Modern chess engines are silent experts: they play at a superhuman level, but do not offer explanations for their play. On the other hand, language models (LMs) can generate plausible-sounding explanations, but their weak playing strength limits the utility of their explanations. We introduce Queen, a 4B-parameter chess-language model that can explain its moves and plans while playing at the level of a typical Grandmaster. Our novel framework enables domain-specific reasoning through complementary components: an encoder-decoder architecture and an iterative distillation algorithm. This architecture integrates a silent expert chess encoder with an instruction-tuned LM through cross-attention, which we train via a question-answering curriculum to extract chess concepts from the encoder's representations. Building on this domain-adapted model, we iteratively improve its explanations with a natural-language analog of the Bellman update: the model analyzes the positions after its top candidate moves and consolidates them into an explanation of the current position, which is then distilled back into the model. Over seven iterations, our model gains over 900 Elo points (1782 to 2697), substantially surpassing all frontier models on both playing strength and puzzle accuracy, despite containing three orders of magnitude fewer parameters. Furthermore, LM-based evaluations show that our explanations are fluent and approach GPT-5.6-Sol (high) in coherence. The generality of our architecture and training procedure suggests a recipe for applying language models to domains where silent expert encoders are available, like games, robotics, and computer use.
Chinese Translation
现代国际象棋引擎是沉默的专家：它们以超人水平进行对弈，但并不提供其下棋的解释。另一方面，语言模型（LMs）能够生成听起来合理的解释，但其较弱的下棋能力限制了其解释的实用性。我们介绍了Queen，一个拥有40亿参数的国际象棋语言模型，能够在与典型国际象棋大师相当的水平下进行对弈，并解释其走法和计划。我们的新框架通过互补组件实现了领域特定的推理：一个编码器-解码器架构和一个迭代蒸馏算法。该架构通过交叉注意力将沉默的专家国际象棋编码器与经过指令调优的语言模型相结合，我们通过问答课程训练该模型，从编码器的表示中提取国际象棋概念。在此领域适应的模型基础上，我们通过自然语言的贝尔曼更新类比迭代改进其解释：模型分析其最佳候选走法后的局面，并将其整合为当前局面的解释，然后再蒸馏回模型中。经过七轮迭代，我们的模型获得了超过900个Elo积分（从1782提升至2697），在下棋实力和难题准确性上大幅超越所有前沿模型，尽管其参数数量少了三个数量级。此外，基于语言模型的评估显示我们的解释流畅，并在连贯性上接近GPT-5.6-Sol（高）。我们架构和训练过程的普遍性为将语言模型应用于沉默专家编码器可用的领域（如游戏、机器人技术和计算机使用）提供了一种方案。