cs.RO / 1 / 2609.28530
Know Your Body: A Harness for Direct and Self-Improving Robot Control with VLMs
了解你的身体：一种用于直接和自我改进机器人控制的带有视觉语言模型的工具
Abstract
A general-purpose vision-language model can understand a task goal without knowing how a particular robot's motion and functional parts produce the intended effect. We introduce KnowBody, a harness that makes these action-relevant body relations explicit, queryable, and revisable while keeping the model weights frozen. Initialized from one off-task trajectory, a partial body model guides action selection and the interpretation of past interactions. New evidence refines the model, and knowledge dependent on revised body estimates is rechecked before reuse. Across 32 fixed-budget trials on four real-robot tasks, initialized KnowBody achieves 75% completion versus 25% for the native harness and requires fewer planner rounds on successful trials in tasks completed by both. With persistent updates enabled, planner rounds decrease by 29-53% from the first to the fifth recorded success.
Chinese Translation
通用视觉语言模型能够理解任务目标，而无需了解特定机器人运动和功能部件如何产生预期效果。我们介绍了KnowBody，这是一种工具，使得与动作相关的身体关系变得明确、可查询和可修订，同时保持模型权重不变。从一个非任务轨迹初始化，部分身体模型指导动作选择和对过去交互的解释。新的证据会细化模型，并在重新使用之前重新检查依赖于修订身体估计的知识。在四个真实机器人任务的32次固定预算试验中，初始化的KnowBody实现了75%的完成率，而本地工具仅为25%，并且在两个任务中成功完成的试验中需要的规划轮次更少。在启用持续更新的情况下，从第一次到第五次记录成功，规划轮次减少了29-53%。
cs.RO / 2 / 2609.28660
Morphometric Imitation: From Morphology and Contact Aware Hand Retargeting to Sim-to-Real Visuomotor Policy
形态模仿：从形态学和接触感知手部重定向到仿真到现实的视觉运动策略
Abstract
Human hand-object interactions (HOIs) provide a rich source of demonstrations for dexterous manipulation, but learning directly from them presents challenges in bridging morphology gaps, ensuring dynamical feasibility, and sim-to-real deployment. We present Morphometric Imitation, a three-stage framework that transforms reconstructed HOIs into zero-shot sim-to-real visuomotor policies. First, morphometric optimization (MMO) kinematically retargets human motion across hand morphologies while preserving demonstrated contacts. Second, residual reinforcement learning (RL) refines the kinematic reference using object pose and contact information from the human motion to produce dynamically feasible robot demonstrations. Third, these demonstrations are distilled into visuomotor policies. Across three robot hands and ten HOIs, MMO improves contact F1 over the strongest of five baselines by at least 8 points for every hand, while also improving the success rate of downstream dynamic retargeting by as much as 35 points. Ablations on the residual RL show complementary benefits from using object pose and contact information. Finally, the visuomotor policies achieve 89.3% zero-shot success in 300 real-world trials on 30 objects. Project page: $\href{https://morphometricimitation.github.io}{\text{this https URL}}$
Chinese Translation
人类手部与物体的交互（HOIs）为灵巧操作提供了丰富的示范来源，但直接从中学习面临着弥合形态差距、确保动态可行性和仿真到现实部署的挑战。我们提出了形态模仿（Morphometric Imitation），这是一个三阶段框架，将重建的 HOIs 转换为零样本仿真到现实的视觉运动策略。首先，形态优化（Morphometric Optimization, MMO）在保持示范接触的同时，运动学地重定向人类动作到不同的手部形态。其次，残差强化学习（Residual Reinforcement Learning, RL）利用人类动作中的物体姿态和接触信息来细化运动学参考，从而生成动态可行的机器人示范。最后，这些示范被提炼为视觉运动策略。在三个机器人手和十个 HOIs 的实验中，MMO 在每个手上将接触 F1 分数提高了至少 8 分，相较于五个基线中的最强者，同时也将下游动态重定向的成功率提高了多达 35 分。对残差 RL 的消融实验显示，使用物体姿态和接触信息带来了互补的好处。最终，这些视觉运动策略在 30 个物体的 300 次真实世界试验中实现了 89.3% 的零样本成功率。项目页面：$ ext{this https URL}$
cs.RO / 3 / 2609.28709
OA-MPPI: Occlusion-Aware Model Predictive Path Integral Control for UAV Flight
OA-MPPI：面向遮挡的模型预测路径积分控制用于无人机飞行
Abstract
Autonomous UAV flight through cluttered and partially unknown environments requires reasoning not only about observed obstacles but also about occluded regions that the sensor cannot observe. We present OA-MPPI, an obstacle- and occlusion-aware extension of Model Predictive Path Integral (MPPI) control for quadrotor flight that accounts for potential moving agents emerging from these regions into the vehicle's path. At every planning step, we extract a 3D occlusion boundary from the online occupancy map and use it to model the regions that hidden agents could reach over the prediction horizon. We penalize trajectories that enter these expanding regions within MPPI rollouts generated using nonlinear quadrotor dynamics and accounting for individual rotor thrust limits. We validate the proposed approach in simulation and hardware flight experiments, with the complete pipeline running onboard the vehicle in real time. Results show increased clearance from occlusion boundaries compared to baseline MPPI in both settings, as well as avoidance of an agent emerging from occlusion in simulation.
Chinese Translation
自主无人机在杂乱和部分未知环境中的飞行需要不仅考虑观察到的障碍物，还要考虑传感器无法观察到的遮挡区域。我们提出了OA-MPPI，这是对模型预测路径积分（MPPI）控制的一个障碍物和遮挡感知的扩展，适用于四旋翼飞行，考虑了潜在的移动代理可能从这些区域进入飞行器路径的情况。在每个规划步骤中，我们从在线占用地图中提取三维遮挡边界，并利用它来建模在预测时间范围内隐藏代理可能到达的区域。我们对在使用非线性四旋翼动力学生成的MPPI回放中进入这些扩展区域的轨迹施加惩罚，并考虑了各个旋翼的推力限制。我们在仿真和硬件飞行实验中验证了所提出的方法，整个流程在飞行器上实时运行。结果表明，与基线MPPI相比，在这两种设置中，遮挡边界的间隙增加，并且在仿真中成功避免了从遮挡中出现的代理。
cs.RO / 4 / 2609.28716
Temporal Learning for End-Effector Position Estimation under Aerodynamic Disturbances in Aerial Continuum Manipulation
在气动干扰下的空中连续操作中的末端执行器位置估计的时间学习
Abstract
This paper investigates temporal neural networks for \mbox{end-effector} position \mbox{estimation} of an aerial continuum manipulator (ACM) operating under aerodynamic effects induced by the unmanned aerial vehicle (UAV). An experimental dataset is collected under stationary (\mbox{rotor-off}) and \mbox{free-hovering} conditions across continuum robot (CR) configurations and UAV altitudes, providing \mbox{end-effector} position measurements with and without aerodynamic residuals. To establish a nominal framework, \mbox{strain-parameterized} kinematic models with progressively richer strain bases are evaluated to balance model complexity and prediction accuracy. The selected nominal model then serves as the baseline for 3D position residual estimation using a \mbox{closed-form} \mbox{continuous-time} (CfC) neural network, with a multilayer perceptron (MLP) and a gated recurrent unit (GRU) used for comparison. On unseen test experiments, the CfC achieves an RMSE of \(22.00\pm1.70~\mathrm{mm}\) over five random seeds, compared with \(36.38\pm3.58~\mathrm{mm}\) for the MLP and \(27.72\pm2.92~\mathrm{mm}\) for the GRU, corresponding to reductions of \(39.52\%\) and \(20.62\%\), respectively. These results demonstrate the effectiveness of \mbox{continuous-time} learning for \mbox{end-effector} position estimation under aerodynamic disturbances relative to static and \mbox{discrete-time} learning methods.
Chinese Translation
本文研究了时间神经网络在无人机（UAV）引起的气动效应下的空中连续操作器（ACM）末端执行器位置估计中的应用。我们在静态（旋翼关闭）和自由悬停条件下收集了实验数据集，涵盖了连续机器人（CR）配置和无人机高度，提供了带有和不带有气动残差的末端执行器位置测量。为了建立一个名义框架，我们评估了具有逐渐丰富的应变基的应变参数化运动学模型，以平衡模型复杂性和预测精度。所选的名义模型随后作为使用闭式连续时间（CfC）神经网络进行三维位置残差估计的基线，比较中使用了多层感知器（MLP）和门控递归单元（GRU）。在未见测试实验中，CfC在五个随机种子上实现了RMSE为22.00±1.70 mm，而MLP为36.38±3.58 mm，GRU为27.72±2.92 mm，分别对应减少了39.52%和20.62%。这些结果表明，相较于静态和离散时间学习方法，连续时间学习在气动干扰下的末端执行器位置估计中具有有效性。
cs.RO / 5 / 2609.28766
TAPESIM: Efficient Simulation of Adhesive Tape Dispensing for Robotic Manipulation
TAPESIM：用于机器人操作的粘性胶带分配高效仿真
Abstract
Applying adhesive tape to secure wire harnesses or seal packages requires robots to coordinate a flexible strip, a moving roll, and surfaces that attach and detach. Simulation could make these interactions repeatable for robot development and evaluation, but resolving every adhesive layer is expensive and can suppress roll motion at practical solver tolerances, while a permanently rigid roll cannot release material. We present TapeSim, a tape simulator that concentrates deformation near the unwinding region and along the released strip. We will release the source code. A rigid cluster represents most wound material, while an advancing deformable collar enables payout and leaves released tape flexible and reattachable. Optional releasable bonds simplify adhesive interfaces and reduce mean step times for smaller rolls. Controlled swing tests show improved roll rotation. At 32 turns, clustering gives 3.2-3.4x mean physics-step speedups at a fixed Newton tolerance and 4.5-8.4x for comparable roll motion. Across five real-motion Stick replays, the clustered variants reduce mean image-plane core-landmark error by 23-29% relative to the full-shell cohesive baseline. On 100 paired Peel cases, they improve balanced accuracy from 50% to 72.9-76.3%, with interface rankings varying across tasks. A teleoperated box-sealing sequence demonstrates attachment, dispensing, cutting, and sealing in a continuous workflow.
Chinese Translation
将粘性胶带应用于固定线束或封装需要机器人协调灵活的胶带、移动的卷轴以及附着和脱离的表面。仿真可以使这些交互在机器人开发和评估中变得可重复，但解决每一层粘合剂的成本较高，并且在实际求解器容差下可能抑制卷轴运动，而永久刚性卷轴无法释放材料。我们提出了TapeSim，一种胶带仿真器，集中变形于展开区域和释放的胶带上。我们将发布源代码。一个刚性集群代表大部分缠绕材料，而一个前进的可变形颈圈则实现了放出，并使释放的胶带保持灵活和可重新附着。可选的可释放粘合剂简化了粘合界面，并减少了小卷轴的平均步骤时间。受控摆动测试显示了改进的卷轴旋转。在32次旋转中，集群化在固定的牛顿容差下提供了3.2-3.4倍的平均物理步骤加速，而在可比的卷轴运动中则为4.5-8.4倍。在五次真实运动的Stick重播中，集群变体相对于完整壳体粘合基线减少了23-29%的平均图像平面核心地标误差。在100对Peel案例中，它们将平衡准确率从50%提高到72.9-76.3%，界面排名在不同任务中有所不同。一个远程操作的箱子封装序列展示了在连续工作流程中的附着、分配、切割和封装。
cs.RO / 6 / 2609.28767
Human-in-the-Loop Geospatial Annotation for Rapid Dataset Construction in Field-Deployed UAV Systems
人机协同的地理空间标注用于快速构建现场部署无人机系统的数据集
Abstract
Real-world perception systems must adapt to changing environments, but manual image annotation cannot scale to field data volumes. We present BirdsEye, which shifts expert annotation from images to the field: an operator records target locations in world coordinates using RTK positioning and calibrated projective geometry propagates each observation to all frames where the target is visible. To quantify how well physical annotations align with image observations, we derive a first-order mapping from camera-pose uncertainty to pixel uncertainty and validate it against Monte Carlo simulation. This mapping is linear in the six per-axis pose variances, so it inverts into a sensor design tool: we give a sufficient condition converting an annotation tolerance into a convex set of admissible pose-noise budgets, a closed-form largest admissible scaling of a deployed sensor suite, and a unique per-axis pose specification under an equal-budget-share allocation. We also analyze the planar-surface approximation underlying the projection, which holds up to 10 degrees of terrain slope. By direct measurement, we show that system projection accuracy is sub-decimeter (sub-30 pixel) at AGL altitudes of 10-20m under conditions excluding sustained yawing. During an in-field case study across three agricultural sites, two field workers produced 12,524 annotated frames carrying 55,600 labels in roughly 12 hours (25.5x per-worker rate increase over manual labeling). Detectors trained on imagery collected by this workflow recovered 56-89% of in-view surveyed targets at a geographically distinct farm, at pre-registered operating points; human review of the leading configuration estimates detection precision at 83-87%, spanning three tie-break conventions for clusters carrying contradictory human verdicts.
Chinese Translation
现实世界的感知系统必须适应不断变化的环境，但手动图像标注无法满足现场数据的规模需求。我们提出了BirdsEye，它将专家标注从图像转移到现场：操作员使用RTK定位记录目标在世界坐标中的位置，并通过校准的投影几何将每个观察结果传播到目标可见的所有帧中。为了量化物理标注与图像观察之间的对齐程度，我们推导出相机姿态不确定性与像素不确定性之间的一阶映射，并通过蒙特卡洛模拟进行验证。该映射在六个轴向姿态方差中是线性的，因此可以转化为传感器设计工具：我们给出了将标注容忍度转换为可接受姿态噪声预算的凸集的充分条件、已部署传感器套件的闭式最大可接受缩放，以及在等预算分配下的唯一轴向姿态规格。我们还分析了投影所依据的平面表面近似，该近似在地形坡度达到10度时仍然有效。通过直接测量，我们展示了在排除持续偏航的条件下，系统投影精度在10-20米的AGL高度下低于分米（低于30像素）。在三个农业现场的实地案例研究中，两名现场工作人员在大约12小时内产生了12,524个标注帧，携带55,600个标签（每位工作人员的标注速率比手动标注提高了25.5倍）。在这一工作流程收集的图像上训练的检测器在一个地理上不同的农场中恢复了56-89%的视野内调查目标，在预注册的操作点；对领先配置估计的检测精度进行人工审查，结果在83-87%之间，涵盖了对携带矛盾人类裁决的簇的三种平局打破约定。
cs.RO / 7 / 2609.28798
OCC4M: Object-Centric 4D Memory for Spatiotemporal Reasoning in Long-Horizon Manipulation
OCC4M：用于长时间操作的时空推理的对象中心4D记忆
Abstract
Long-horizon manipulation often requires reasoning about state absent from the current view, such as a vanished object's location, temporal identity, or the contents of a shuffled container. We present OCC4M ("Occam"), an object-centric 4D memory that maintains persistent tracks in a shared world frame and explicitly represents temporal, motion, and containment relations. A vision-language model (VLM) queries this structured memory to select actionable targets for history-free low-level execution. Across seven simulation conditions and 350 episodes, OCC4M achieves 96.6% memory success and 88.9% end-to-end success, versus 54.6% and 57.7% for FrameSamp, a raw-history VLM baseline using Gemini 3.7 Flash with the complete observation history and the same executor. In a controlled viewpoint-transfer test, OCC4M maintains 100% memory and 98% end-to-end success after a viewpoint change, while full-history FrameSamp falls to near-zero success. On 20 fixed-camera Franka episodes, OCC4M reaches 85% joint memory accuracy, versus at most 30% for FrameSamp across context sizes from $K=16$ to the complete history, and completes 45% of full two-stage tasks. These results support explicit object-centric memory for persistent spatiotemporal reasoning in long-horizon manipulation. Qualitative videos are available at https://occ4m-sup.github.io/occ4m-supplementary/.
Chinese Translation
长时间操作通常需要推理当前视图中缺失的状态，例如消失物体的位置、时间身份或打乱容器的内容。我们提出了OCC4M（“奥卡姆”），一种对象中心的4D记忆，它在共享的世界框架中保持持久的轨迹，并明确表示时间、运动和包含关系。视觉语言模型（VLM）查询这一结构化记忆，以选择可执行的目标进行无历史的低级执行。在七种仿真条件和350个实验中，OCC4M实现了96.6%的记忆成功率和88.9%的端到端成功率，而使用完整观察历史和相同执行者的原始历史VLM基线FrameSamp则分别为54.6%和57.7%。在一个受控的视角转移测试中，OCC4M在视角变化后保持了100%的记忆和98%的端到端成功率，而完整历史的FrameSamp成功率降至近零。在20个固定摄像头的Franka实验中，OCC4M达到了85%的联合记忆准确率，而FrameSamp在从$K=16$到完整历史的上下文大小中最多为30%，并完成了45%的完整两阶段任务。这些结果支持在长时间操作中进行持久时空推理的显式对象中心记忆。相关的定性视频可在https://occ4m-sup.github.io/occ4m-supplementary/观看。
cs.RO / 8 / 2609.28807
An Analysis of Streaming Deep Reinforcement Learning for Adaptive Continual Learning in Robotics
流式深度强化学习在机器人自适应持续学习中的分析
Abstract
Over the course of a lifetime, robots may encounter novel scenarios unaccounted for in its original training that result in performance degradation. One common approach to mitigating this issue is to further grow the offline training dataset in hopes of producing a policy robust to these changes. In contrast, biological learning occurs moment-to-moment via a stream of experience, unlike the predominantly batch-based and offline nature of deep learning. Although recent works show the feasibility of stream-based deep reinforcement learning, where updates use only the latest experience, none have shown it to be a viable continual learning framework for adapting robotic policies to unseen changes. In this paper, we present the first analysis of streaming deep reinforcement learning for adaptive continual learning in robotics. In particular, we show that, following an initial pretraining phase, streaming deep RL can enable a robot to successfully adapt to unforeseen changes to itself, its environment, or goals. Our primary experiments within quadruped locomotion demonstrate that a deep neural network robotic policy with certain optimizers and plasticity loss mitigation techniques can successfully leverage domain task knowledge from its pretraining to quickly adapt online to diverse changes via stream learning, outperforming batch-based on-policy methods and improving task success rates by up to 90% over the pretrained policy. Furthermore, we perform additional evaluations on robotic manipulation tasks to determine if our previous observations extend to different robotic morphologies and scenarios. Our results show that the successes observed in quadruped locomotion can be partially realized in manipulation with stability and performance limitations. We conclude with a discussion on the limitations of our work and its implications for the future of continual robot learning.
Chinese Translation
在机器人的生命周期中，可能会遇到在其原始训练中未考虑的新场景，这会导致性能下降。一种常见的解决方案是进一步扩展离线训练数据集，以期生成对这些变化具有鲁棒性的策略。与此相对，生物学习是通过一系列经验的流动进行的，而不是以深度学习主要依赖的批处理和离线方式进行。尽管最近的研究显示了基于流的深度强化学习的可行性，其中更新仅使用最新的经验，但尚未有研究表明其作为一种可行的持续学习框架，能够使机器人策略适应未见的变化。在本文中，我们首次分析了流式深度强化学习在机器人自适应持续学习中的应用。特别是，我们展示了在初始预训练阶段之后，流式深度强化学习能够使机器人成功适应自身、环境或目标的不可预见变化。我们在四足动物运动中的主要实验表明，采用特定优化器和可塑性损失缓解技术的深度神经网络机器人策略，能够成功利用预训练中的领域任务知识，通过流式学习快速适应多样化的变化，超越基于批处理的在线策略方法，并将任务成功率提高至预训练策略的90%。此外，我们还对机器人操作任务进行了额外评估，以确定我们之前的观察是否适用于不同的机器人形态和场景。我们的结果表明，在四足动物运动中观察到的成功在操作中也能部分实现，但存在稳定性和性能的限制。最后，我们讨论了我们工作的局限性及其对未来机器人持续学习的影响。
cs.RO / 9 / 2609.28816
FlyCNS: Connectome-Grounded Information Organization for Communication-Constrained Embodied Control
FlyCNS：基于连结组的信息组织框架用于通信受限的具身控制
Abstract
Robotic bodies are inherently distributed in sensing and actuation, yet learning-based control still commonly relies on centralized information processing. This work studies the problem of information organization in communication-constrained embodied control: which computations should remain local, and which information is worth transmitting for whole-body coordination. We propose FlyCNS, an embodied information-organization framework inspired by the Drosophila brain--nerve-cord connectome. FlyCNS preserves local sensorimotor computation within each limb and enables selective long-range communication through separate ascending and descending routing pathways. From a real connectome, FlyCNS extracts the directional structural complexity of these two pathway types and uses it as a weak prior over communication allocation, while message content, transmission timing, and locomotion policies remain task-adaptive and are learned through reinforcement learning. In Unitree Go1 simulation, FlyCNS exhibits more graceful performance degradation as the communication budget is tightened. Under the most restrictive setting, it uses only about 21--22\% of the communication of the full-communication reference, while still maintaining a tracking score of approximately 0.882 under both command protocols, with a gap of no more than 6.1\% from the full-communication reference. These results indicate that real neural connectomes can inform not only the structural design of control networks, but also provide transferable inductive biases for information organization across embodiments, guiding robots in balancing local computation and long-range coordination under limited communication resources.
Chinese Translation
机器人身体在感知和执行方面本质上是分布式的，但基于学习的控制仍然通常依赖于集中式信息处理。本研究探讨了在通信受限的具身控制中信息组织的问题：哪些计算应保持局部，哪些信息值得传输以实现全身协调。我们提出了FlyCNS，这是一种受果蝇大脑-神经索连结组启发的具身信息组织框架。FlyCNS在每个肢体内保留局部的感觉运动计算，并通过单独的上行和下行路由通道实现选择性的长距离通信。FlyCNS从真实的连结组中提取这两种通路类型的方向性结构复杂性，并将其作为通信分配的弱先验，而消息内容、传输时机和运动策略则保持任务自适应，并通过强化学习进行学习。在Unitree Go1模拟中，FlyCNS在通信预算收紧时表现出更优雅的性能退化。在最严格的设置下，它仅使用约21%至22%的全通信参考的通信量，同时在两种指令协议下仍保持约0.882的跟踪得分，与全通信参考的差距不超过6.1%。这些结果表明，真实的神经连结组不仅可以为控制网络的结构设计提供信息，还可以为跨具身体的信息组织提供可转移的归纳偏见，指导机器人在有限的通信资源下平衡局部计算和长距离协调。
cs.RO / 10 / 2609.28818
KeyGen: Unsupervised Keypoint based Object-Centric Representations for Category-Level Policy Generalization
KeyGen：基于无监督关键点的面向对象表示用于类别级策略泛化
Abstract
Generalization in robotic manipulation requires policies to perform tasks across diverse unseen object instances that vary in shape, size, and pose. However, conventional behavior cloning (BC) methods often overfit to instance-specific geometry and appearance, limiting transfer to novel objects. We introduce KeyGen, a framework that learns canonicalized semantic 3D keypoints from point clouds and uses them as structured object-centric representations for policy learning. A visuomotor diffusion policy conditions on these keypoints together with object-centric geometry to predict full manipulation trajectories, enabling consistent geometric correspondence across object instances. To evaluate category-level generalization, we construct a photorealistic simulation benchmark with three manipulation tasks and a planning-driven data generation pipeline that produces expert trajectories across diverse object instances. Experiments show that KeyGen significantly outperforms prior methods on both seen and unseen objects under pose variation, scales effectively with additional demonstrations per object, maintains robustness to object rescaling, and achieves strong performance in both simulation and real-world manipulation.
Chinese Translation
机器人操作中的泛化要求策略能够在形状、大小和姿态各异的未见物体实例上执行任务。然而，传统的行为克隆（BC）方法往往对特定实例的几何形状和外观过拟合，限制了对新物体的迁移。我们提出了KeyGen，一个从点云中学习规范化语义3D关键点的框架，并将其作为结构化的面向对象表示用于策略学习。一个视觉运动扩散策略基于这些关键点以及面向对象的几何信息来预测完整的操作轨迹，从而在物体实例之间实现一致的几何对应关系。为了评估类别级泛化，我们构建了一个包含三项操作任务的照片级真实感仿真基准，并建立了一个基于规划的数据生成管道，能够在不同物体实例上生成专家轨迹。实验表明，KeyGen在姿态变化下显著优于先前的方法，无论是在已见还是未见物体上，能够有效地扩展到每个物体的额外演示，保持对物体缩放的鲁棒性，并在仿真和现实世界操作中均表现出强大的性能。
cs.RO / 11 / 2609.28838
Uncertainty-Gated Exploration Noise Suppresses Task Collapse in Online RL Fine-Tuning of a Flow-Matching Vision-Language-Action Policy
不确定性门控探索噪声抑制在线强化学习微调中的任务崩溃：以流匹配视觉-语言-动作策略为例
Abstract
Online reinforcement learning fine-tuning of pretrained flow-matching vision-language-action (VLA) policies promises robots that keep learning after deployment, but continued updates often destroy competence on individual tasks while the aggregate still looks healthy. We study this failure mode, which we call task collapse, under a matched small-compute budget on LIBERO-10 with a 450M-parameter SmolVLA policy trained by PPO with stochastic (SDE) sampling. Three exploration-noise policies differ in one live variable: a fixed noise scale, a ReinFlow-style learned noise network, and an uncertainty-gated controller that redistributes exploration across task streams from task-agnostic novelty and competence signals, without task labels or episode boundaries. Under the pooled definition, fixed noise collapses tasks in two of three seeds and learned noise in every seed measured to iteration 200, while the controller collapses none in any of its three seeds. Measured parameter displacement shows the controller's action expert keeps changing, while its mean applied noise is close to the fixed scale in the available logs. The matched comparison supports the controller's effect on task preservation; the separate contributions of its adaptation across states and over time are not disentangled. A lower fixed scale slows the decline but does not stop it. No arm improves on the behavior-cloning baseline in this budget. Two properties of that regime are measured beside this result, not offered as its cause: following the reference recipe, training runs in bfloat16 with no fp32 master copy, under which 96.02% of the action expert's elements stay bit-identical across three consecutive iterations, and an fp32 master copy at the reference learning rate collapses both arms in a single-seed observation. We release tools measuring per-task collapse under four definitions, rescoring noise and instrument tares.
Chinese Translation
在线强化学习对预训练的流匹配视觉-语言-动作（VLA）策略进行微调，承诺使机器人在部署后持续学习，但持续更新往往会破坏个别任务的能力，而整体表现仍然看似健康。我们研究了这种失败模式，称之为任务崩溃，在LIBERO-10上以450M参数的SmolVLA策略为例，在匹配的小计算预算下，通过PPO和随机（SDE）采样进行训练。三种探索噪声策略在一个实时变量上有所不同：固定噪声尺度、ReinFlow风格的学习噪声网络，以及一种不确定性门控控制器，该控制器在没有任务标签或回合边界的情况下，从与任务无关的新颖性和能力信号中重新分配探索到任务流中。在汇总定义下，固定噪声在三个种子中的两个任务崩溃，而学习噪声在每个测量到第200次迭代的种子中崩溃，而控制器在其三个种子中没有崩溃。测量的参数位移显示控制器的动作专家不断变化，而其平均施加噪声接近可用日志中的固定尺度。匹配比较支持控制器对任务保持的影响；其在状态和时间上的适应性贡献尚未被分离。较低的固定尺度减缓了能力的下降，但并未停止。在该预算下，没有任何臂的表现超过行为克隆基线。除了这一结果外，还测量了该机制的两个特性，并未将其作为原因：遵循参考配方，训练运行在bfloat16下进行，没有fp32主副本，其中96.02%的动作专家元素在连续三次迭代中保持位相同，以及在参考学习率下的fp32主副本在单一种子观察中崩溃了两个臂。我们发布了工具，以四种定义测量每个任务的崩溃，重新评分噪声并进行仪器校准。
cs.RO / 12 / 2609.28878
Online Sim-to-Real Adaptation via Closed-Loop System Modeling
通过闭环系统建模实现在线仿真到现实的适应
Abstract
Sim-to-real transfer has made substantial progress, but can still produce controllers that remain stable and functional on hardware while suffering from degraded tracking accuracy due to residual dynamics mismatch. Correcting these errors typically requires identifying the underlying system dynamics, adapting the control policy, or returning to simulation for additional training and finetuning, all of which can require substantial data and computation. We propose OSRAM (Online Sim-to-Real Adaptation via Closed-Loop System Modeling), a framework that instead adapts the reference commands provided to an existing controller. OSRAM treats the deployed robot and its policy as a unified closed-loop dynamical system and learns its task-level command-response behavior directly from tracking observations. A closed-loop dynamics model is meta-trained across randomized dynamics in simulation and rapidly finetuned after deployment using limited real-world interaction. The adapted model is then used to optimize future reference commands while leaving the underlying control policy unchanged. We evaluate OSRAM on bipedal velocity tracking and loco-manipulation in simulation and on hardware. Results show that closed-loop modeling improves prediction and tracking accuracy under unseen dynamics, while online reference adaptation reduces residual sim-to-real tracking errors across different control objectives and hardware configurations. These results demonstrate that adapting the behavior of the robot-policy closed loop provides a practical alternative to finetuning the policy or identifying the full physical dynamics for sim-to-real transfer. More information can be found at http://generalroboticslab.com/OSRAM.
Chinese Translation
仿真到现实的转移已经取得了显著进展，但仍然可能产生在硬件上保持稳定和功能的控制器，同时由于残余动力学不匹配而导致跟踪精度下降。纠正这些错误通常需要识别基础系统动力学、调整控制策略或返回仿真进行额外的训练和微调，这些都可能需要大量的数据和计算。我们提出了OSRAM（通过闭环系统建模实现在线仿真到现实的适应），这一框架通过调整提供给现有控制器的参考命令来实现适应。OSRAM将部署的机器人及其策略视为一个统一的闭环动态系统，并直接从跟踪观察中学习其任务级命令-响应行为。闭环动力学模型在仿真中针对随机化动力学进行元训练，并在部署后通过有限的现实世界交互迅速进行微调。然后，适应后的模型用于优化未来的参考命令，同时保持基础控制策略不变。我们在仿真和硬件上评估了OSRAM在双足速度跟踪和运动操控中的表现。结果表明，闭环建模在未见动力学下提高了预测和跟踪精度，而在线参考适应减少了不同控制目标和硬件配置下的残余仿真到现实的跟踪误差。这些结果表明，适应机器人-策略闭环的行为为微调策略或识别完整物理动力学提供了一种实用的替代方案，以实现仿真到现实的转移。更多信息可以在http://generalroboticslab.com/OSRAM找到。
cs.RO / 13 / 2609.28887
Fly, Drive, Reconfigure: A Modular Reconfigurable Aerial-Ground Platform for Field Operations
飞行、驾驶、重构：一种用于现场操作的模块化可重构空地平台
Abstract
Heterogeneous robot teams distribute complementary capabilities across specialized agents, but their physical roles and capacities typically remain fixed throughout a mission. We present HARP, a Heterogeneous Aerial Robotic modules Platform in which independently deployable aerial robots physically reconfigure to compose their capabilities for field operations. HARP comprises sensor-equipped scouts, flydrive rover modules, and task-specific payload modules. Scouts map the environment and inform an energy-aware planner that jointly selects routes and air-ground mobility modes. Rover and payload modules fly independently across terrain that constrains ground travel, then autonomously assemble into a cooperative ground vehicle for energy-efficient payload transport. Motivated by environmental sampling in remote and difficult-to-traverse regions, we evaluate HARP through field experiments spanning sensing, planning, reconfiguration, airground mobility, payload transport, and task execution. We further conduct module-level deployment tests on the Greenland Ice Sheet toward future autonomous missions. HARP demonstrates how heterogeneous robot teams can adapt not only their actions, but also how their physical capabilities are composed during a mission.
Chinese Translation
异构机器人团队在专门化代理之间分配互补能力，但它们在任务执行过程中通常保持固定的物理角色和能力。我们提出了HARP（异构空中机器人模块平台），其中独立部署的空中机器人通过物理重构来组合其在现场操作中的能力。HARP包括配备传感器的侦察机器人、飞行驾驶越野模块和特定任务的载荷模块。侦察机器人负责绘制环境地图，并向一个能量感知规划器提供信息，该规划器共同选择路线和空地移动模式。越野和载荷模块在限制地面行驶的地形上独立飞行，然后自主组装成一个合作的地面车辆，以实现能量高效的载荷运输。受到在偏远和难以穿越区域进行环境采样的启发，我们通过涵盖感知、规划、重构、空地移动、载荷运输和任务执行的现场实验评估HARP。我们还在格林兰冰盖上进行模块级部署测试，以支持未来的自主任务。HARP展示了异构机器人团队如何不仅能够适应其行动，还能够在任务执行过程中重组其物理能力。
cs.RO / 14 / 2609.28910
Robots That Take Initiative: A Framework for Building and Evaluating Proactive Robots
主动型机器人：构建和评估主动机器人的框架
Abstract
Effective robot assistance beyond narrow roles and repetitive tasks requires robots to be proactive - to decide what needs to be done rather than waiting to be told. While proactivity is increasingly explored, it lacks a unified formulation, and work in the domain is typically evaluated offline against static human models that cannot capture the effect of a robot's actions on the environment and the user's own behavior. We introduce a unified formalism for proactive robot assistance, organize it into three levels, and provide a framework to address the highest level of unprompted proactive assistance. We then show that offline evaluation overstates performance in this setting, and contribute a closed-loop evaluation with a human model that adapts to the robot. Finally, we present a method, GAP, that instantiates our framework, learning from passive observation to anticipate user goals and act. Under closed-loop evaluation, prior state-of-the-art methods collapse, in some cases adding more work than they save, while GAP remains robust and substantially outperforms them.
Chinese Translation
有效的机器人辅助超越狭窄角色和重复任务的要求，需要机器人具备主动性——能够决定需要做什么，而不是等待指示。尽管主动性越来越受到关注，但尚缺乏统一的表述，且该领域的工作通常是离线评估，与静态人类模型进行比较，这无法捕捉机器人行为对环境和用户自身行为的影响。我们引入了一个统一的主动机器人辅助形式化框架，将其组织为三个层次，并提供一个框架以应对最高层次的无提示主动辅助。然后，我们展示了在这一设置下，离线评估夸大了性能，并贡献了一个与人类模型相结合的闭环评估，该模型能够适应机器人。最后，我们提出了一种方法GAP，它实现了我们的框架，通过被动观察学习以预测用户目标并采取行动。在闭环评估下，先前的最先进方法表现不佳，在某些情况下增加的工作量超过了节省的工作量，而GAP则保持稳健，并显著超越了它们。
cs.RO / 15 / 2609.28920
Koopman-Accelerated Model-Based Diffusion for Real-Time Robot Control
基于Koopman加速的模型驱动扩散用于实时机器人控制
Abstract
Conventional model-based diffusion (MBD) achieves effective trajectory optimization by leveraging noise annealing. However, its high computational cost, primarily arising from repeated rollouts of the plant dynamics, has largely confined its use to offline settings. To address this limitation, this paper proposes bilinear Koopman model-based diffusion (BK-MBD). The proposed method lifts the robot's state into a high-dimensional space only once per control step and propagates all candidates in the lifted space thereafter, so each rollout reduces to a fixed number of matrix-vector multiplications. The lifted dynamics are bilinear, allowing the predicted input gain to vary with the robot's configuration, which a linear lifted model cannot represent. In simulation, BK-MBD completed each planning update in at most 14.7 ms within a 50 ms control period and reached the goal on every trial, whereas a linear lift almost never did. The annealed schedule improves closed-loop accuracy over fixed-noise schedules under the learned rollout. Under the exact rollout, both the annealed and fixed-narrow schedules reach every goal, indicating that annealing reduces sensitivity to surrogate-model error. BK-MBD also threaded a passage that no single convex region covers, whereas a convexified bilinear controller rarely succeeded. On a physical manipulator, BK-MBD tracked an initially unknown moving target within the control period and was the only method that met both the tracking task and the deadline. The project page is available at https://rcilab.khu.ac.kr/bkmbd/.
Chinese Translation
传统的模型驱动扩散（MBD）通过利用噪声退火实现了有效的轨迹优化。然而，其高计算成本，主要源于对植物动态的重复展开，已大大限制了其在离线环境中的应用。为了解决这一限制，本文提出了双线性Koopman模型驱动扩散（BK-MBD）。该方法在每个控制步骤中仅将机器人的状态提升到高维空间一次，随后在提升的空间中传播所有候选者，因此每次展开减少为固定数量的矩阵-向量乘法。提升的动态是双线性的，允许预测的输入增益随着机器人的配置而变化，而线性提升模型无法表示这一点。在仿真中，BK-MBD在50毫秒的控制周期内每次规划更新最多完成于14.7毫秒，并在每次试验中都达到了目标，而线性提升几乎从未做到。退火调度在学习展开下提高了闭环精度，相较于固定噪声调度。在精确展开下，退火和固定狭窄调度均达到了每个目标，表明退火降低了对代理模型误差的敏感性。BK-MBD还穿越了一个没有单一凸区域覆盖的通道，而凸化的双线性控制器很少成功。在物理操控器上，BK-MBD在控制周期内追踪了一个最初未知的移动目标，并且是唯一一个同时满足追踪任务和截止日期的方法。项目页面可访问 https://rcilab.khu.ac.kr/bkmbd/。
cs.RO / 16 / 2609.28927
Streaming-WAM: Action-Conditioned World-Action Model for Asynchronous Robot Manipulation
Streaming-WAM：用于异步机器人操作的动作条件世界动作模型
Abstract
World action models (WAMs) that use future visual prediction at inference time incur substantial generation costs. Asynchronous execution reduces waiting by overlapping inference with robot motion, but visual predictions used for subsequent action generation must anticipate the effects of actions already scheduled for execution during inference. We introduce Streaming-WAM, which couples action-conditioned world modeling with asynchronous robot control to account for committed actions in future visual prediction. At each streaming update, the model conditions future visual prediction on the latest observation and the committed actions, which form the fixed prefix of the next action chunk. The resulting action-conditioned visual features guide generation of the remaining actions within the same joint update, so the continuation is informed by the scene changes expected during execution of the fixed prefix. On LIBERO, Streaming-WAM achieves an average success rate of 98.35\% and reduces mean episode time by a factor of 2.93 relative to Fast-WAM. On the real-world Stamp Paper task, mean episode time falls from 90 s with synchronous Joint-WAM to 38 s with Streaming-WAM. These results show that Streaming-WAM supports efficient asynchronous control while maintaining high task success rates.
Chinese Translation
世界动作模型（WAMs）在推理时使用未来视觉预测，这会产生相当大的生成成本。异步执行通过将推理与机器人运动重叠来减少等待时间，但用于后续动作生成的视觉预测必须预见在推理期间已安排执行的动作的效果。我们提出了Streaming-WAM，它将动作条件的世界建模与异步机器人控制相结合，以考虑未来视觉预测中的已承诺动作。在每次流式更新中，该模型将未来视觉预测条件化于最新观察和已承诺的动作，这些动作形成下一个动作块的固定前缀。生成的动作条件视觉特征指导在同一联合更新中剩余动作的生成，因此后续动作受到在执行固定前缀期间预期的场景变化的影响。在LIBERO上，Streaming-WAM实现了98.35%的平均成功率，并将平均回合时间相对于Fast-WAM减少了2.93倍。在真实世界的Stamp Paper任务中，平均回合时间从同步的Joint-WAM的90秒降至Streaming-WAM的38秒。这些结果表明，Streaming-WAM支持高效的异步控制，同时保持高任务成功率。
cs.RO / 17 / 2609.28933
A Field-Deployable GNSS-based Navigation Stack for Outdoor Mobile Robots
一种适用于户外移动机器人的基于GNSS的现场可部署导航系统
Abstract
Outdoor robots require more than an accurate receiver and a path-tracking law: the navigation system must preserve geometric consistency from geographic waypoints to actuator commands, expose measurement validity and timing, and respond to invalid or stale state information. This work presents a ROS~2 navigation stack with interchangeable single-GNSS--IMU and dual-antenna-GNSS localization front ends. Both provide a common local East--North--Up state interface for pure pursuit, virtual-point cross-track PID, finite-horizon nonlinear model predictive control (NMPC), and a segment-dependent hybrid dispatcher. The architecture specifies coordinate conventions, datum initialization, asynchronous state construction, waypoint geometry, controller equations, quality gates, command arbitration, and watchdog behavior. Independent physical field runs collected during 2025 and 2026 grape-vineyard deployments support a balanced evaluation of 800 runs, with 100 runs for each of eight controller--localization combinations on an approximately 199.6-m route. The row-hybrid mode yields the lowest run-averaged post-acquisition mean absolute cross-track error (MAE) in the evaluated dataset: 0.00952~m with single GNSS+IMU and 0.00846~m with dual GNSS. These findings characterize deviations of the recorded positions from the reference route under the evaluated conditions. The open-source navigation software and deployment instructions are available in the https://github.com/YiyuanLinXX/PPBv2/tree/main/PPBv2_Navigation.
Chinese Translation
户外机器人不仅需要一个精确的接收器和路径跟踪法则：导航系统必须保持从地理航点到执行器命令的几何一致性，暴露测量有效性和时序，并对无效或过时的状态信息做出响应。本研究提出了一种ROS~2导航系统，具有可互换的单GNSS-IMU和双天线GNSS定位前端。这两种前端为纯追踪、虚拟点交叉轨迹PID、有限时域非线性模型预测控制（NMPC）和段依赖混合调度器提供了一个共同的局部东-北-上状态接口。该架构规定了坐标约定、基准初始化、异步状态构建、航点几何、控制器方程、质量门、命令仲裁和看门狗行为。在2025年和2026年的葡萄园部署中进行的独立物理现场测试支持了800次运行的平衡评估，每种控制器-定位组合在大约199.6米的路线上的运行次数为100次。行混合模式在评估数据集中产生了最低的运行平均后获取均方根交叉轨迹误差（MAE）：单GNSS+IMU为0.00952米，双GNSS为0.00846米。这些发现表征了在评估条件下记录的位置与参考路线之间的偏差。开源导航软件及部署说明可在https://github.com/YiyuanLinXX/PPBv2/tree/main/PPBv2_Navigation获取。
cs.RO / 18 / 2609.28952
RoboRecover: Benchmarking Robot Policy Recovery under Execution Deviations
RoboRecover：在执行偏差下的机器人策略恢复基准测试
Abstract
Robot-policy benchmarks increasingly cover diverse tasks and preset out-of-distribution conditions, but typically evaluate complete trajectories from predefined initial states. These evaluations often focus on the initialized scene and the final outcome, while paying less attention to the dynamic interaction process. During closed-loop execution, actions and contacts can alter object relations and task progress, producing off-nominal intermediate states that need recovery. Recovery requires a policy to infer how task progress has changed, correct the relevant relations, and continue the original goal. We introduce RoboRecover, a benchmark for robot policy recovery under execution deviations. RoboRecover selects deviation states from trajectories, reconstructs them by replaying action prefixes, and evaluates policies on the original task. RoboRecover contains 2,000 scenarios across RoboTwin and LIBERO, with 1,000 scenarios and a fixed 800/200 train/test split on each platform. Results show that initial-state performance does not determine recovery performance and policies exhibit different recovery strengths across scenarios. Using its training split, RoboRecover further supports study on recovery interventions. RoboRecover establishes recovery from execution-induced intermediate states as a distinct dimension of robot policy evaluation.
Chinese Translation
机器人策略基准测试越来越多地涵盖多样化的任务和预设的分布外条件，但通常评估的是从预定义初始状态开始的完整轨迹。这些评估往往侧重于初始化场景和最终结果，而较少关注动态交互过程。在闭环执行过程中，动作和接触可能会改变物体关系和任务进展，产生需要恢复的非正常中间状态。恢复需要策略推断任务进展如何变化，纠正相关关系，并继续原始目标。我们引入了RoboRecover，这是一个用于在执行偏差下进行机器人策略恢复的基准测试。RoboRecover从轨迹中选择偏差状态，通过重放动作前缀进行重建，并在原始任务上评估策略。RoboRecover包含来自RoboTwin和LIBERO的2000个场景，每个平台上有1000个场景，并且固定的800/200训练/测试划分。结果表明，初始状态的表现并不决定恢复性能，且策略在不同场景中表现出不同的恢复能力。利用其训练划分，RoboRecover进一步支持对恢复干预的研究。RoboRecover将从执行引起的中间状态恢复确立为机器人策略评估的一个独特维度。
cs.RO / 19 / 2609.28955
ActGaze: Learning Action-Grounded Gaze through Counterfactual Visual Interventions for High-Precision Manipulation
ActGaze：通过反事实视觉干预学习基于动作的注视以实现高精度操作
Abstract
Current Vision-Language-Action (VLA) models often struggle with high-precision robotic manipulation. We attribute this limitation primarily to their visual attention being dispersed across task-irrelevant regions. To address this issue, we propose ActGaze, a training approach that guides VLA policies to gaze on task-relevant regions, much like humans gaze on critical visual cues while executing precise movements. Unlike prior methods that rely on external labels for gaze supervision, ActGaze derives spatial supervision directly from the VLA's own action objective by using counterfactual visual interventions to identify regions that are critical for action prediction. Extensive real-robot experiments on four high-precision robotic manipulation tasks demonstrate that ActGaze induces more focused visual attention on task-relevant regions and consistently outperforms the base VLA policy and other visual-grounding approaches.
Chinese Translation
当前的视觉-语言-动作（VLA）模型在高精度机器人操作中常常面临挑战。我们将这一限制主要归因于它们的视觉注意力分散在与任务无关的区域。为了解决这个问题，我们提出了ActGaze，一种训练方法，指导VLA策略关注于与任务相关的区域，类似于人类在执行精确动作时关注关键视觉线索。与依赖外部标签进行注视监督的先前方法不同，ActGaze通过使用反事实视觉干预直接从VLA自身的动作目标中推导空间监督，以识别对动作预测至关重要的区域。在四个高精度机器人操作任务上的大量真实机器人实验表明，ActGaze能够在与任务相关的区域上引导更集中的视觉注意力，并且始终优于基础VLA策略和其他视觉基础方法。
cs.RO / 20 / 2609.28959
TactileStep: Sole Tactile Learning for Regulating Foot-Terrain Interaction in Humanoid Locomotion
TactileStep：用于调节人形机器人步态与地面交互的足底触觉学习
Abstract
Humanoid parkour policies can traverse various terrains, but task completion may mask challenges of harsh landings, edge contacts, and unstable stance contacts. Humans naturally regulate foot-terrain interaction through tactile feedback, modulating contact compliance according to terrain stiffness. This highlights a key domain gap between humans and humanoid robots: the absence of rich tactile sensing in most humanoid systems. We address this problem with TactileStep, a deployable tactile learning framework that brings sole pressure sensing into humanoid locomotion control for softer touchdowns and more stable support. TactileStep aligns tactile simulation with the real pressure insole, allowing the policy to learn from the same contact features available on hardware. During training, we use tactile and motion cues to recognize different foot-contact phases and apply phase-aware rewards that encourage safer landing and more stable stance. Evaluated in simulation and on a Unitree G1 humanoid across diverse terrains, TactileStep reduces peak touchdown force by up to 48.8% and peak A-weighted impact noise by up to 30.1 dB over a strong perceptive baseline, while increasing stance contact area by up to 23.8%.
Chinese Translation
人形机器人跑酷策略能够穿越各种地形，但任务的完成可能掩盖了剧烈着陆、边缘接触和不稳定支撑接触等挑战。人类自然通过触觉反馈来调节足与地面的交互，根据地形的刚度调整接触的顺应性。这突显了人类与人形机器人之间的一个关键领域差距：大多数人形系统缺乏丰富的触觉感知。我们通过 TactileStep 解决了这一问题，这是一种可部署的触觉学习框架，将足底压力感知引入人形机器人步态控制，以实现更柔和的着陆和更稳定的支撑。TactileStep 将触觉仿真与真实的压力鞋垫对齐，使策略能够从硬件上可用的相同接触特征中学习。在训练过程中，我们使用触觉和运动线索来识别不同的足接触阶段，并应用阶段感知奖励，以鼓励更安全的着陆和更稳定的站姿。在模拟和 Unitree G1 人形机器人上评估的多种地形中，TactileStep 将峰值着陆力降低了高达 48.8%，峰值 A 加权冲击噪声降低了高达 30.1 dB，相较于强感知基线，同时将支撑接触面积增加了高达 23.8%。
cs.RO / 21 / 2609.28960
Echo in the Steps: Learning Perceptive Humanoid Parkour with Gated Memory
步伐中的回声：利用门控记忆学习感知型类人跑酷
Abstract
While recent advances in perceptive locomotion have enabled humanoid robots to traverse structured terrains, agile parkour in highly discontinuous environments remains an open challenge. In particular, crossing sparse footholds and narrow support regions requires precise foothold selection, effective use of visual observations, and consistent alternating foot placement during fast transitions. In this paper, we present a perceptive humanoid parkour framework that enables stable traversal across terrains with limited foothold availability using only onboard depth observations. The framework features a saliency-guided temporal perception module that combines a saliency prior with gated memory. It retains informative depth features across frames, enabling reliable foot placement from partial observations. By introducing an alternation loss, our symmetry regularization encourages alternating gait patterns and improves traversal robustness. Extensive experiments show that our method significantly improves success rate and foothold accuracy on challenging terrains in both simulation and the real world.
Chinese Translation
尽管近期在感知运动方面的进展使类人机器人能够在结构化地形中移动，但在高度不连续的环境中进行灵活的跑酷仍然是一个未解决的挑战。特别是在跨越稀疏的支撑点和狭窄的支撑区域时，需要精确的支撑点选择、有效利用视觉观测以及在快速过渡期间保持一致的交替足部放置。在本文中，我们提出了一种感知型类人跑酷框架，该框架仅利用机载深度观测，能够在支撑点有限的地形上实现稳定的穿越。该框架具有一个基于显著性引导的时间感知模块，该模块结合了显著性先验和门控记忆。它能够在帧间保留信息丰富的深度特征，从而使得从部分观测中实现可靠的足部放置成为可能。通过引入交替损失，我们的对称正则化鼓励交替步态模式，并提高了穿越的鲁棒性。大量实验表明，我们的方法在模拟和现实世界中显著提高了在挑战性地形上的成功率和支撑点准确性。
cs.RO / 22 / 2609.28969
Sim-to-Real Aware End-to-End Learning Environment for Micromobility
面向微型出行的仿真到现实感知端到端学习环境
Abstract
While end-to-end autonomous driving systems show promise, their application to micromobility vehicles is hindered by simulators failing to capture specific kinematics, such as differential drives and omni-wheels. This paper pro- poses a sim-to-real-aware, vehicle-specific end-to-end learning environment for the WHILL Model CR on AWSIM and ROS 2. To minimize the sim-to-real gap, physical parameters are optimized via Bayesian optimization using real-world data, reducing trajectory errors across various driving scenarios. Additionally, this study introduces a synchronized architecture tailored for the stable training of world model-based agents. An end-to-end policy trained with DreamerV3 exhibited learning progress and achieved task completion in a simulated obstacle avoidance setting. Furthermore, this policy demonstrated direct sim-to-real transfer to the physical vehicle, enabling the vehicle to navigate around a cardboard box in a real-world corridor replica without fine-tuning. This paper provides a practical foundation for sim-to-real micromobility policy studies.
Chinese Translation
尽管端到端的自动驾驶系统展现出良好的前景，但其在微型出行车辆中的应用受到模拟器无法捕捉特定运动学（如差速驱动和全向轮）的限制。本文提出了一种面向仿真到现实的、特定于车辆的端到端学习环境，适用于WHILL Model CR，基于AWSIM和ROS 2。为了最小化仿真到现实的差距，利用真实世界数据通过贝叶斯优化优化物理参数，从而减少各种驾驶场景下的轨迹误差。此外，本研究引入了一种同步架构，旨在为基于世界模型的智能体的稳定训练提供支持。使用DreamerV3训练的端到端策略在模拟的障碍物避让环境中表现出了学习进展，并成功完成了任务。此外，该策略还展示了直接的仿真到现实转移能力，使得车辆能够在真实世界的走廊复制环境中绕过纸箱，而无需进行微调。本文为仿真到现实的微型出行政策研究提供了实用的基础。
cs.RO / 23 / 2609.28973
AquaMend: Minimal Re-probing and Conditional Rollback for Latent-Belief Failures in Embodied Agents
AquaMend：针对具身智能体潜在信念失败的最小重探测与条件回滚
Abstract
Physical changes or sensing errors can invalidate embodied agents' task-relevant beliefs. AquaMend compares re-probing, rollback, and supported continuation on a probe-belief-action graph under an expected-loss objective covering sensing, physical recovery, and uncorrected failures. A joint posterior guides a one-step policy with conditional detection-power screening. The per-belief three-way optimum requires independence, separability, and fully resolving probes; the general policy has no global optimality guarantee. Across 32 paired scenarios in a self-constructed simulation benchmark, AquaMend recovers in 28/32 cases and reduces mean complete loss by 21.6% versus restart. Its paired loss difference from decision-theoretic troubleshooting (DTT) is not statistically significant after Holm correction. Against the all-candidate ablation, online decision time decreases by 12.3% overall but increases by 3.4% in the uncovered late stage.
Chinese Translation
物理变化或感知错误可能使具身智能体的任务相关信念失效。AquaMend 在一个探测-信念-行动图上比较了重探测、回滚和支持的继续，目标是最小化预期损失，该损失涵盖感知、物理恢复和未修正的失败。一个联合后验指导着一个具有条件检测能力筛选的一步策略。每个信念的三方最优需要独立性、可分离性和完全解决的探测；而一般策略则没有全局最优性保证。在一个自建的仿真基准中，AquaMend 在32对场景中恢复成功28次，并将平均完全损失减少了21.6%，相较于重启。与决策理论故障排除（DTT）相比，其配对损失差异在经过霍尔姆校正后并没有统计学显著性。在所有候选者消融实验中，在线决策时间总体减少了12.3%，但在未覆盖的后期阶段增加了3.4%。
cs.RO / 24 / 2609.28976
ReVNM: Learning-Based Visual Navigation from a Remote Camera
ReVNM：基于学习的远程摄像头视觉导航
Abstract
Visual Navigation Models (VNMs) enable robots to navigate from egocentric visual observations without geometric localization and planning, but long-range navigation still requires pre-built maps. This paper presents the Remote Visual Navigation Model (ReVNM), which uses a single remote surveillance camera to serve as both an observation source and an implicit environmental map for visual navigation. While the use of remote cameras could eliminate the need for pre-built maps as well as onboard vision processing, their limited field of view instead of egocentric observations makes it hard to achieve collision-free navigation. The lack of existing data with diverse remote viewpoints, which are crucial for training robust VNMs, further complicates the challenge. In this work, we propose a learning-by-synthesis approach to address this two-fold challenge. Our ReVNM extends a state-of-the-art VNM architecture with an exocentric-to-egocentric (exo2ego) module that predicts an egocentric depth observation from remote-camera observations. This helps the VNM to plan a path while considering obstacles in front of the robot. Trained only on randomly generated worlds with diverse obstacle layouts and camera viewpoints, ReVNM can generalize well to real robot navigation without additional fine-tuning. Experiments in both simulation and real-world environments confirmed the effectiveness of the proposed approach.
Chinese Translation
视觉导航模型（VNMs）使机器人能够在没有几何定位和规划的情况下，从自我中心的视觉观察中进行导航，但长距离导航仍然需要预先构建的地图。本文提出了远程视觉导航模型（ReVNM），该模型使用单个远程监控摄像头作为观察源和隐式环境地图，以实现视觉导航。虽然使用远程摄像头可以消除对预构建地图和车载视觉处理的需求，但其有限的视野而非自我中心的观察使得实现无碰撞导航变得困难。缺乏具有多样化远程视角的现有数据，这对训练稳健的VNMs至关重要，进一步加大了挑战。在本研究中，我们提出了一种通过合成学习的方法来应对这一双重挑战。我们的ReVNM扩展了一种先进的VNM架构，增加了一个外中心到自我中心（exo2ego）模块，该模块能够从远程摄像头观察中预测自我中心的深度观察。这有助于VNM在规划路径时考虑机器人前方的障碍物。ReVNM仅在具有多样障碍布局和摄像头视角的随机生成世界上进行训练，能够很好地推广到真实机器人导航，而无需额外的微调。在模拟和真实环境中的实验验证了所提方法的有效性。
cs.RO / 25 / 2609.28984
CrossSafe: Towards Cross-Embodiment Latent Safety Filters
CrossSafe：面向跨体现的潜在安全过滤器
Abstract
Cross-embodiment learning has shown that a single model, such as a vision-language-action (VLA) model, can learn state representations and manipulation skills that can be applied across heterogeneous robots to accomplish various tasks. We hypothesize that the same holds for safety enforcement. The reasoning required to satisfy a safety constraint, such as detecting an obstacle, recognizing that it should be avoided, and selecting a safe abstract action, is largely shared across robots. What differs across embodiments is how the abstract safe action is realized: morphology, kinematics, and dynamics determine which actions are safe and feasible. Consequently, the same action can be safe for one robot and unsafe for another. This is especially important for generalist manipulation policies that operate in a common end-effector action space without explicitly capturing how safety depends on the robot's morphology and kinematics. We propose embodiment-conditioned safety filtering, in which a Hamilton-Jacobi reachability-based value function and its corresponding safety-maximizing policy are shared across robots. Using a morphology-aware latent representation of the robot and its environment, we perform Hamilton-Jacobi reachability analysis directly in latent space so that the learned safety concepts can generalize across embodiments while remaining explicitly conditioned on each robot's morphology and kinematics. We evaluate our approach across five bimanual robot embodiments and five manipulation tasks with whole-body collision-avoidance constraints. Our results show that a single policy, jointly trained across five manipulation tasks and four embodiments, exhibits zero-shot generalization to a held-out embodiment, reducing the nominal policy's collision rate. They also show that training using more embodiments improves generalization.
Chinese Translation
跨体现学习表明，单一模型（例如视觉-语言-动作（VLA）模型）能够学习状态表示和操作技能，这些技能可以应用于异构机器人以完成各种任务。我们假设安全执行也适用同样的原理。满足安全约束所需的推理，例如检测障碍物、识别应避免的障碍物以及选择安全的抽象动作，在不同机器人之间是高度共享的。不同之处在于抽象安全动作的实现方式：形态学、运动学和动力学决定了哪些动作是安全且可行的。因此，同一动作对于某个机器人可能是安全的，而对于另一个机器人则可能是不安全的。这对于在共同的末端执行器动作空间中操作的通用操控策略尤为重要，因为这些策略并未明确捕捉安全性如何依赖于机器人的形态学和运动学。我们提出了一种基于体现条件的安全过滤方法，其中基于哈密顿-雅可比可达性分析的价值函数及其相应的安全最大化策略在不同机器人之间共享。利用对机器人及其环境的形态感知潜在表示，我们直接在潜在空间中执行哈密顿-雅可比可达性分析，以便学习的安全概念能够在不同体现之间进行泛化，同时仍然明确依赖于每个机器人的形态学和运动学。我们在五种双手机器人体现和五个具有全身碰撞避免约束的操控任务中评估了我们的方法。结果表明，跨五个操控任务和四个体现共同训练的单一策略在保留体现上表现出零样本泛化，降低了名义策略的碰撞率。同时，结果还表明，使用更多体现进行训练可以改善泛化能力。
cs.RO / 26 / 2609.29017
CALM: Current Aligned Link Manipulation for Single Arm Oversized Object Lifting
CALM：用于单臂超大物体提升的当前对齐链接操控
Abstract
Most robots manipulate objects solely with their end effectors, whereas humans flexibly leverage different body parts, such as the forearm and elbow, especially when handling oversized objects. Learning such whole-arm manipulation is chal-lenging due to long-horizon sparse rewards, limited contact sens-ing, and the sim-to-real gap in contact and actuator dynamics. To address these challenges, we propose Current-Aligned Link Manipulation, a framework for learning long-horizon contact-rich manipulation using motor current as joint load related feedback. Three stage-specific policies first learn repositioning, grasping, and lifting using privileged simulation information, and a stage router sequences them to generate complete task demonstrations. For sim-to-real transfer, a causal current mapper predicts physical motor current from simulated joint histories, aligning the actuator current observation between simulation and hardware. A unified student policy then learns from these demonstrations using only deployable sensor observations and is further refined with DAgger. The task policies are trained entirely in simulation, and the final student is deployed on hardware. Experiments demonstrate 76.2% (762/1000 trials) complete-task success in simulation and 73.3% success (22/30 trials) on the physical robot for sequential oversized-object lifting.
Chinese Translation
大多数机器人仅通过末端执行器操控物体，而人类则灵活利用不同的身体部位，如前臂和肘部，尤其是在处理超大物体时。学习这种全臂操控具有挑战性，因为存在长时间跨度的稀疏奖励、有限的接触感知以及接触和驱动器动态中的仿真与现实差距。为了解决这些挑战，我们提出了当前对齐链接操控（Current-Aligned Link Manipulation），这是一个利用电机电流作为关节负载相关反馈来学习长时间跨度接触丰富操控的框架。三个阶段特定的策略首先利用特权仿真信息学习重新定位、抓取和提升，并由阶段路由器将它们按序列生成完整的任务演示。为了实现仿真到现实的转移，一个因果电流映射器根据仿真关节历史预测物理电机电流，从而对齐仿真与硬件之间的驱动器电流观测。一个统一的学生策略随后仅使用可部署的传感器观测从这些演示中学习，并通过DAgger进一步优化。任务策略完全在仿真中训练，最终学生策略在硬件上部署。实验表明，在仿真中完成任务的成功率为76.2%（762/1000次试验），而在物理机器人上进行顺序超大物体提升时成功率为73.3%（22/30次试验）。
cs.RO / 27 / 2609.29020
Outcome-Sensitive Motion Search for Impact-Aware Dexterous Catching
面向结果的运动搜索用于影响感知的灵巧捕捉
Abstract
Skilled humans can catch fast-moving objects softly by coordinating interception, velocity matching, and follow-through to mitigate impact. Learning such impact-aware catching with reinforcement learning (RL), however, is challenging, as the policy must achieve reliable interception and grasping while regulating the sensitive transition into contact. Moreover, even a capable privileged-state RL teacher may not provide ideal demonstrations for a deployable imitation-learning (IL) student: teacher failures limit task coverage, while small variations in pre-contact motion can produce substantially different impact and grasping outcomes. We characterize this phenomenon through interventional outcome sensitivity and introduce the outcome-sensitive window (OSW) to guide targeted demonstration construction. Building on this formulation, we propose Outcome-Sensitive Motion Search, which learns a task-conditioned manifold of successful OSW motions and performs local geodesic search to refine successful teacher rollouts and repair task conditions where the teacher fails. We then validate candidate motions through complete rollouts under a calibrated IL-student action-error model and retain only successful executions as demonstrations. Extensive simulation experiments demonstrate that our method effectively repairs task conditions where the teacher fails and enables the resulting IL policy to outperform the privileged RL teacher in both catching success and impact mitigation.
Chinese Translation
熟练的人类能够通过协调拦截、速度匹配和后续动作来柔和地捕捉快速移动的物体，以减轻冲击。然而，利用强化学习（RL）学习这种影响感知的捕捉是具有挑战性的，因为策略必须在调节敏感的接触过渡的同时，实现可靠的拦截和抓取。此外，即使是一个有能力的特权状态RL教师也可能无法为可部署的模仿学习（IL）学生提供理想的示范：教师的失败限制了任务覆盖范围，而接触前运动的微小变化可能会导致截然不同的冲击和抓取结果。我们通过干预结果敏感性来表征这一现象，并引入结果敏感窗口（OSW）来指导有针对性的示范构建。在此基础上，我们提出了面向结果的运动搜索，该方法学习成功的OSW运动的任务条件流形，并执行局部测地线搜索，以优化成功的教师回放并修复教师失败的任务条件。然后，我们通过在经过校准的IL学生动作误差模型下的完整回放验证候选运动，仅保留成功的执行作为示范。大量仿真实验表明，我们的方法有效修复了教师失败的任务条件，并使得最终的IL策略在捕捉成功率和冲击减轻方面超越了特权RL教师。
cs.RO / 28 / 2609.29021
CAMP: Cooperative Arm-Hand Motion Planning in Constrained Spaces
CAMP：受限空间中的协作臂-手运动规划
Abstract
Coordinated arm-hand motion planning is fundamental to dexterous robotic manipulation in complex and constrained environments. A straightforward solution is to decompose the problem into separate arm path planning and hand motion generation; however, this poses a dilemma: decomposition can miss feasible solutions that require coordinated arm-hand adaptation along the path. Alternatively, directly planning in the high-dimensional joint arm-hand configuration space captures such coupling but faces a substantially enlarged search space and nonconvex collision constraints. To characterize this coupling, we formulate feasible hand fibers that capture collision-free hand configurations for each arm configuration. Based on this formulation, we propose CAMP, a high-success and efficient cooperative arm-hand motion planner for constrained environments. CAMP constructs candidate trajectories through layered hand search with local arm relaxation, then compactly represents them using endpoint-preserving via-point movement primitives (VMPs) for coarse-to-fine joint optimization. Across six constrained simulation tasks, CAMP achieves 84.2-98.5% planning success, outperforming alternative planners with competitive efficiency. Ablation studies verify the contributions of arm relaxation, VMP representation, and coarse-to-fine optimization, while real-robot experiments demonstrate CAMP on constrained manipulation tasks. The project website is available at https://camp-armhand.github.io/.
Chinese Translation
协调的臂-手运动规划是复杂和受限环境中灵巧机器人操作的基础。一个简单的解决方案是将问题分解为单独的臂路径规划和手部运动生成；然而，这会带来一个困境：分解可能会错过需要沿路径进行协调臂-手适应的可行解。另一种方法是在高维关节臂-手配置空间中直接进行规划，这能够捕捉到这种耦合关系，但面临着显著扩大的搜索空间和非凸碰撞约束。为了表征这种耦合关系，我们制定了可行的手部纤维，以捕捉每个臂配置的无碰撞手部配置。基于这一公式，我们提出了CAMP，一个高成功率且高效的协作臂-手运动规划器，适用于受限环境。CAMP通过分层手部搜索与局部臂松弛构建候选轨迹，然后使用保持端点的中间点运动原语（VMPs）紧凑地表示这些轨迹，以进行粗到细的关节优化。在六个受限的仿真任务中，CAMP实现了84.2%-98.5%的规划成功率，超越了其他规划器并具备竞争力的效率。消融研究验证了臂松弛、VMP表示和粗到细优化的贡献，而真实机器人实验则展示了CAMP在受限操作任务中的应用。项目网站可访问：https://camp-armhand.github.io/
cs.RO / 29 / 2609.29031
Simple Torque-Observation Alignment for Zero-Shot Sim-to-Real Grasping with a Direct-Drive Gripper
简单的扭矩观测对齐方法用于零样本模拟到真实抓取，采用直接驱动夹爪
Abstract
Torque observations in reinforcement learning remain challenging because simulated and measured torque differ in scale, offset, and noise. In this paper, we propose a simple torque observation alignment method for robots with direct-drive (DD) actuators, in which motor current maps linearly to joint torque through a motor-type-specific torque constant K_tau. First, dynamometer calibration identifies K_tau* and corrects the scale mismatch between simulated and real torque. Second, the method uses delta_tau(t) = tau(t) - tau(t-1) as the observation in both domains to eliminate the constant offset instead of using the direct torque tau(t), which carries a domain-dependent bias. Third, Gaussian noise obtained from the dynamometer measurement data is injected during the learning process. To validate the proposed method, we train a teacher-student grasping policy entirely in simulation and deploy the distilled student on a multifingered DD gripper. The deployed policy performs proprioceptive grasping using only joint positions and torque differences. We conduct an ablation study comparing the proposed method with alternative alignment variants on nine in-distribution (ID) objects. The proposed method achieves 100% grasp success. These results demonstrate that the proposed alignment method improves the robustness of zero-shot policy transfer on the DD gripper against real-world torque-observation mismatches.
Chinese Translation
在强化学习中，扭矩观测仍然面临挑战，因为模拟和测量的扭矩在尺度、偏移和噪声上存在差异。本文提出了一种简单的扭矩观测对齐方法，适用于具有直接驱动（DD）执行器的机器人，其中电机电流通过特定电机类型的扭矩常数 K_tau 线性映射到关节扭矩。首先，通过测功机校准识别 K_tau*，并修正模拟与真实扭矩之间的尺度不匹配。其次，该方法在两个领域中使用 delta_tau(t) = tau(t) - tau(t-1) 作为观测，以消除常量偏移，而不是使用直接扭矩 tau(t)，后者携带领域依赖的偏差。第三，在学习过程中注入从测功机测量数据获得的高斯噪声。为了验证所提出的方法，我们在模拟环境中完全训练了一个教师-学生抓取策略，并将提炼后的学生策略部署在一个多指直接驱动夹爪上。部署的策略仅使用关节位置和扭矩差进行自我感知抓取。我们进行了消融研究，将所提出的方法与九个分布内（ID）物体上的替代对齐变体进行了比较。所提出的方法实现了100%的抓取成功率。这些结果表明，所提出的对齐方法提高了在直接驱动夹爪上进行零样本策略转移的鲁棒性，抵御了真实世界中扭矩观测的不匹配。
cs.RO / 30 / 2609.29043
Design and Evaluation of LLM Chaining-Based Task Planning for General Purpose Service Robots
基于LLM链式任务规划的通用服务机器人设计与评估
Abstract
General Purpose Service Robot (GPSR) tasks, as defined in the RoboCup@Home benchmark, require robots to interpret diverse natural language commands and generate multi-step action sequences in real home environments. Conventional Single Prompt (SP) approaches suffer from context bloat and the "Lost in the Middle" phenomenon, leading to unreliable task planning. We propose an LLM chaining architecture that separates instruction classification and action generation into two specialized stages, reducing per-inference prompt length by approximately 45% while improving planning consistency. We evaluate our method using 100 randomly generated GPSR commands across three language models spanning local open-source and frontier cloud deployment contexts. Results show consistent planning improvements over SP across all models, with gains of up to +37 percentage points on local models. Further, real-robot execution experiments on the Toyota Human Support Robot (HSR) reveal that planning success alone does not guarantee task completion, with 6 of 10 tasks completing successfully and execution-layer failures identified as the primary remaining bottleneck.
Chinese Translation
通用服务机器人（GPSR）任务，如RoboCup@Home基准中所定义，要求机器人能够理解多样的自然语言指令，并在真实家庭环境中生成多步骤的行动序列。传统的单一提示（SP）方法存在上下文膨胀和“中途迷失”现象，导致任务规划不可靠。我们提出了一种LLM链式架构，将指令分类和行动生成分为两个专业阶段，减少每次推理的提示长度约45%，同时提高规划一致性。我们使用100个随机生成的GPSR指令在三个语言模型上进行评估，这些模型涵盖了本地开源和前沿云部署环境。结果显示，在所有模型上，SP方法的规划一致性均有显著提升，本地模型的增益高达37个百分点。此外，在丰田人类支持机器人（HSR）上的真实机器人执行实验表明，仅有规划成功并不能保证任务完成，10个任务中仅有6个成功完成，执行层面的失败被识别为主要的瓶颈。
cs.RO / 31 / 2609.29065
DA-GRD: Decision-Aware Grasp-Relevant Disambiguation for tactile recovery under perception-to-execution mismatches
DA-GRD：针对感知与执行不匹配下的触觉恢复的决策感知抓取相关消歧
Abstract
Grasping is a fundamental robotic capability that bridges perception and physical task execution. This paper studies grasp pose recovery under a perception-to-execution mismatch, where a grasp generated from visual perception may become spatially stale if the object moves before execution, using only sparse tactile interactions and no further visual observations. We propose DA-GRD, Decision-Aware Grasp-Relevant Disambiguation, which maintains a weighted planar belief over possible object configurations and selects tactile probes according to their ability to eliminate hypotheses and improve agreement among candidate task grasps. Rather than fully relocalizing the object, DA-GRD stops when the remaining hypotheses support a common executable grasp. In MuJoCo experiments on ten rigid objects with translations up to 5~cm and yaw perturbations up to $\pm45^\circ$, DA-GRD achieves an 84.7% physical lift success rate, compared with 9.1% for stale AnyGrasp, 21.2% for the original fix-scan baseline, and 63.7% for fix-scan method adapted with an SE(2) belief. DA-GRD also achieves a 57.3% Task conditioned Success rate. Across objects, it uses a success-average of 4.13 tactile probes over the ten per-object means, corresponding to a 72.5% reduction relative to the fixed 15-probe baselines. Real-world experiments on six objects achieve 71.7% physical lift success and 38.3% task-conditioned success with 4.20 probes on average. These results show that tactile sensing can recover task-relevant grasps under vision-off conditions with limited physical interaction, without requiring complete object localization.
Chinese Translation
抓取是连接感知与物理任务执行的基本机器人能力。本文研究了在感知与执行不匹配下的抓取姿态恢复问题，其中从视觉感知生成的抓取如果在执行前物体移动，可能会变得空间过时，仅依赖稀疏的触觉交互而没有进一步的视觉观测。我们提出了DA-GRD（Decision-Aware Grasp-Relevant Disambiguation），该方法在可能的物体配置上维持加权平面信念，并根据触觉探测消除假设和改善候选任务抓取之间一致性的能力选择触觉探测。DA-GRD并不完全重新定位物体，而是在剩余假设支持一个共同可执行抓取时停止。在对十个刚性物体进行的MuJoCo实验中，物体的平移最大达到5厘米，偏航扰动最大达到±45度，DA-GRD实现了84.7%的物理提升成功率，而过时的AnyGrasp为9.1%，原始固定扫描基线为21.2%，适应SE(2)信念的固定扫描方法为63.7%。DA-GRD还实现了57.3%的任务条件成功率。在不同物体上，它在每个物体的平均上使用了4.13个触觉探测，较固定的15个探测基线减少了72.5%。在六个物体的真实世界实验中，平均使用4.20个探测实现了71.7%的物理提升成功率和38.3%的任务条件成功率。这些结果表明，在视觉关闭条件下，触觉感知能够在有限的物理交互下恢复与任务相关的抓取，而无需完全定位物体。
cs.RO / 32 / 2609.29091
From Passive Execution to Active Exploration: Agentic Embodied Manipulation in Realistic Environments
从被动执行到主动探索：真实环境中的能动体现操作
Abstract
Recent advances in agentic systems have substantially enhanced the long-horizon capability of embodied manipulation. However, many existing frameworks still follow a passive execution paradigm, which limits their applicability to real-world scenarios involving textual semantic cues, distractors, and initially invisible targets. To bridge this gap, we propose an agent-based active exploration framework that enables robots to dynamically interact with the environment rather than merely execute predefined instructions. Specifically, our framework consists of three collaborative modules: a planning module for high-level task reasoning, a perception module for visual scene understanding, and an execution module for low-level manipulation. This design allows the robot to actively acquire task-relevant information, adapt its behavior based on environmental feedback, and complete manipulation tasks under partial observability. Furthermore, we introduce a fine-grained perception-execution interleaving strategy, which tightly couples visual feedback with skill execution to improve exploration robustness. We evaluate our method on a realistic Find-and-Place task, demonstrating its effectiveness in challenging environments where target objects must be actively discovered before manipulation.
Chinese Translation
近年来，能动系统的进展显著增强了体现操作的长时间能力。然而，许多现有框架仍然遵循被动执行范式，这限制了它们在涉及文本语义线索、干扰物和最初不可见目标的现实场景中的适用性。为了解决这一问题，我们提出了一种基于代理的主动探索框架，使机器人能够动态地与环境互动，而不仅仅是执行预定义的指令。具体而言，我们的框架由三个协作模块组成：用于高层任务推理的规划模块、用于视觉场景理解的感知模块和用于低层操作的执行模块。这一设计使机器人能够主动获取与任务相关的信息，根据环境反馈调整其行为，并在部分可观察性下完成操作任务。此外，我们引入了一种细粒度的感知-执行交错策略，将视觉反馈与技能执行紧密结合，以提高探索的鲁棒性。我们在一个真实的寻找与放置任务上评估了我们的方法，展示了其在目标物体必须在操作前主动发现的挑战性环境中的有效性。
cs.RO / 33 / 2609.29092
DAWN: Noise-Robust Quadruped Parkour via Depth-Denoising World Models
DAWN：通过深度去噪世界模型实现抗噪四足跑酷
Abstract
Vision-based legged locomotion methods assume clean depth at training time and rely on hand-tuned post-processing filters at deployment. However, filter parameters are rarely disclosed, hindering reproducibility, and performance degrades substantially when depth noise is left unaddressed. Building noise robustness directly into the learning pipeline would eliminate this dependency. While such robustness has been explored for proprioceptive inputs, analogous approaches for depth perception remain largely absent in legged locomotion. We propose DAWN (Denoising and Alignment in World models for Noise-robustness), a noise-robust perception framework for legged locomotion, which builds noise robustness directly into a world model via two modifications: (1) feeding noisy depth to the encoder while keeping clean depth as the reconstruction target, forcing the model to implicitly denoise its input; and (2) applying contrastive learning to align the latent states of noisy and clean depth. Importantly, DAWN is not tied to a specific noise model, requiring no manual tuning to the noise distribution at deployment. Furthermore, it incurs no additional inference cost over existing world model-based methods. Without any manual filter calibration -- relying solely on the learned noise-robust representation -- DAWN achieves zero-shot quadruped parkour on a Unitree Go1: traversing stairs up to 18 cm, clearing gaps up to 70 cm, and mounting steps up to 45 cm from raw depth observations. Ablation studies show that denoising and contrastive alignment contribute at complementary levels -- reconstruction and representation, respectively -- and yield additive gains when combined. Videos and code are available at: https://dawn-parkour.github.io/
Chinese Translation
基于视觉的腿部运动方法在训练时假设深度数据是干净的，并在部署时依赖手动调优的后处理滤波器。然而，滤波器参数很少被公开，这妨碍了可重复性，并且当深度噪声未得到处理时，性能会显著下降。将抗噪能力直接构建到学习流程中可以消除这种依赖性。虽然这种鲁棒性已在本体感知输入中得到探索，但在深度感知方面类似的方法在腿部运动中仍然基本缺失。我们提出了DAWN（Denoising and Alignment in World models for Noise-robustness），这是一个针对腿部运动的抗噪感知框架，通过两个修改将抗噪能力直接构建到世界模型中：(1) 将噪声深度输入到编码器，同时保持干净深度作为重建目标，迫使模型隐式去噪其输入；(2) 应用对比学习对噪声和干净深度的潜在状态进行对齐。重要的是，DAWN并不依赖于特定的噪声模型，在部署时无需手动调优噪声分布。此外，与现有基于世界模型的方法相比，它不会增加额外的推理成本。在没有任何手动滤波器校准的情况下——仅依赖于学习到的抗噪表示——DAWN在Unitree Go1上实现了零-shot四足跑酷：能够跨越高达18厘米的台阶、清除高达70厘米的间隙，并从原始深度观测中爬升高达45厘米的台阶。消融研究表明，去噪和对比对齐在互补层面上贡献——分别为重建和表示——并在结合时产生附加增益。视频和代码可在以下网址获取：https://dawn-parkour.github.io/
cs.RO / 34 / 2609.29093
A Support-Enhanced Granular-Jamming Gripper for RL-based Grasping with Continuum Manipulators
一种支持增强的颗粒阻塞夹具用于基于强化学习的连续操控抓取
Abstract
Continuum manipulators provide dexterous motion in confined spaces, but structural compliance, hysteresis, and load-dependent deformation leave residual position and orientation errors that can undermine reliable contact with rigid grippers. To address this limitation, this paper presents a lightweight support-enhanced granular-jamming gripper tailored to a continuum manipulator. The gripper maintains compliance before jamming while establishing a direct load path to the continuum manipulator tip after jamming. To improve its grasping performance, we systematically designed membrane materials, particles, filling ratios, and the internal support structure, and further identify geometry-dependent grasp boundaries with respect to contact offset and object shape. Building on these results, we construct a physical manipulation system integrating the continuum manipulator, granular-jamming gripper, visual feedback, tendon actuation, and pneumatic control. We then train a reinforcement-learning-based reaching controller in a randomized simulation and deploy it on the physical system, demonstrating how positioning control and contact level mechanical adaptation can complement each other in a modular grasp-and-release task. By introducing an adaptive structure that relaxes the need for highly accurate modeling and positioning control, this work explores a design paradigm that integrates physical and embodied intelligence.
Chinese Translation
连续操控器在狭小空间中提供灵活的运动，但结构柔顺性、滞后性和负载依赖变形会留下残余的位置和方向误差，这可能会削弱与刚性夹具的可靠接触。为了解决这一限制，本文提出了一种轻量级的支持增强颗粒阻塞夹具，专门为连续操控器量身定制。该夹具在阻塞前保持柔顺性，而在阻塞后建立与连续操控器末端的直接负载路径。为了提高抓取性能，我们系统地设计了膜材料、颗粒、填充比和内部支撑结构，并进一步识别了与接触偏移和物体形状相关的几何依赖抓取边界。在这些结果的基础上，我们构建了一个物理操作系统，集成了连续操控器、颗粒阻塞夹具、视觉反馈、腱驱动和气动控制。然后，我们在随机化模拟中训练了一个基于强化学习的到达控制器，并将其部署到物理系统上，展示了定位控制和接触级机械适应如何在模块化抓取和释放任务中相辅相成。通过引入一种适应性结构，放宽对高度精确建模和定位控制的需求，本研究探索了一种将物理智能与具身智能相结合的设计范式。
cs.RO / 35 / 2609.29103
TRACE: Interactive Bi-Directional Tracing of Monochrome Cables Amid Clutter
TRACE：在杂乱环境中交互式双向追踪单色电缆
Abstract
Accurate state estimation (tracing) of Deformable Linear Objects (DLOs) such as cables is a critical challenge for data centers, manufacturing, construction, homes, and surgery, where precise cable management directly impacts operational safety and efficiency. However, resolving the state of multiple monochrome cables amid foreground and background clutter poses challenges due to occlusions, overlap, and ambiguous crossings. We present Two-way Routing And Cable Estimation (TRACE), which combines bi-directional cable tracing with interactive perception primitives-Divergence Push and Cluster Dilation-to actively resolve ambiguities. Evaluation with 110 physical experiments suggests that TRACE can increase the percentage of cable length correctly traced in complex scenarios (with up to 4 cables and 40 crossings) from ~60% with the strongest prior method, HANDLOOM 2.0, to ~90%, outperforming RT-DLO, Nano Banana Pro, and ChatGPT 5.2 as well. For a trial run on a workstation with an NVIDIA GeForce RTX 4090 GPU, the average computation time is 0.4 seconds per cable. Project website: https://trace-paper.github.io/.
Chinese Translation
对可变形线性物体（DLOs），如电缆的准确状态估计（追踪）是数据中心、制造业、建筑、家庭和外科手术中的一项关键挑战，其中精确的电缆管理直接影响操作安全性和效率。然而，在前景和背景杂乱中解决多条单色电缆的状态面临挑战，主要由于遮挡、重叠和模糊交叉。我们提出了双向路由与电缆估计（TRACE），它结合了双向电缆追踪和交互感知原语——发散推送（Divergence Push）和聚类膨胀（Cluster Dilation），以主动解决模糊性。通过110个物理实验的评估表明，TRACE能够将在复杂场景中（最多4条电缆和40个交叉点）正确追踪的电缆长度百分比从最强先前方法HANDLOOM 2.0的约60%提高到约90%，并且优于RT-DLO、Nano Banana Pro和ChatGPT 5.2。在一台配备NVIDIA GeForce RTX 4090 GPU的工作站上进行的试运行中，平均计算时间为每条电缆0.4秒。项目网站：https://trace-paper.github.io/
cs.RO / 36 / 2609.29112
Modeling Load-, Velocity-, and Temperature-Dependent Transmission Errors of Cycloidal Drives for Industrial Robots Using Fourier Series
使用傅里叶级数建模工业机器人涡轮驱动的负载、速度和温度依赖的传动误差
Abstract
Industrial robots are rarely used for machining tasks due to their limited path accuracy. This accuracy is mainly limited by inaccuracies in the drive trains. Compliance and transmission errors occur in the joint gearboxes. While transmission errors have been extensively studied for strain wave gears, there is little research on these errors in cycloidal drives. This gearbox type is commonly used in industrial robots for medium to heavy payloads. It is proposed to model the mainly periodic transmission errors using a Fourier series where amplitude and phase are defined as a polynomial function of the main influence factors load-torque, velocity, and temperature. Measurements of the transmission errors were conducted using an experimental setup representing a single robot joint. In the evaluation of the measurement data, harmonic frequencies were related to mechanical properties of the cycloidal drive. These frequencies were used to identify the parameters of the polynomial Fourier series model. Compared to validation measurements, the derived model shows an average root mean square error of 0.026 mrad. It is proposed to use the output of the resulting model in a feedforward control approach to compensate the transmission errors and to increase the path accuracy of industrial robots.
Chinese Translation
由于路径精度有限，工业机器人很少用于加工任务。这种精度主要受到驱动系统不准确性的限制。在关节齿轮箱中存在柔性和传动误差。尽管对应变波齿轮的传动误差进行了广泛研究，但对涡轮驱动中的这些误差的研究却较少。这种齿轮箱类型通常用于承载中到重负荷的工业机器人。本文提出使用傅里叶级数对主要周期性传动误差进行建模，其中幅度和相位被定义为负载扭矩、速度和温度等主要影响因素的多项式函数。通过实验装置对传动误差进行了测量，该装置代表了单个机器人关节。在测量数据的评估中，谐波频率与涡轮驱动的机械特性相关联。这些频率用于识别多项式傅里叶级数模型的参数。与验证测量相比，所推导的模型显示出平均均方根误差为0.026 mrad。建议在前馈控制方法中使用所得到模型的输出，以补偿传动误差并提高工业机器人的路径精度。
cs.RO / 37 / 2609.29132
A Tendon-Driven Robotic Jellyfish with Constrained Soft Actuation and Depth Control via Reinforcement Learning
一种基于腱驱动的机器人水母，具有受限的软驱动和通过强化学习实现的深度控制
Abstract
Jellyfish-inspired robots offer a compliant and efficient approach to underwater locomotion, but achieving large deformation together with repeatable actuation and closed-loop control remains challenging. In this work, we present a tendon-driven robotic jellyfish with constrained soft actuation. Each actuator combines a flexible substrate with discrete constraints, enabling bending up to \(150^\circ\) with an approximately linear tendon displacement-bending relationship. Eight actuators driven by four servos allow the robot to perform stable swimming, attitude adjustment, and self-righting. Based on the linear actuation, a reinforcement-learning controller is further developed, enabling closed-loop depth regulation in both simulation and physical experiments. These results show that mechanical constraints can improve the controllability of soft actuation while preserving compliant jellyfish-like motion, providing a route toward manoeuvrable and autonomous jellyfish robots.
Chinese Translation
受水母启发的机器人提供了一种柔性且高效的水下运动方式，但实现大变形的同时保持可重复的驱动和闭环控制仍然具有挑战性。在本研究中，我们提出了一种具有受限软驱动的腱驱动机器人水母。每个驱动器结合了柔性基材和离散约束，使其能够弯曲至约150°，并且具有近似线性的腱位移-弯曲关系。由四个伺服驱动的八个驱动器使机器人能够实现稳定游泳、姿态调整和自我翻正。基于线性驱动，我们进一步开发了一种强化学习控制器，实现了在模拟和物理实验中的闭环深度调节。这些结果表明，机械约束可以在保持柔性水母般运动的同时改善软驱动的可控性，为可操控和自主的水母机器人提供了一条途径。
cs.RO / 38 / 2609.29157
OREN-X: Octree Residual Network for Real-Time Multi-Modal Mapping
OREN-X：用于实时多模态映射的八叉树残差网络
Abstract
To achieve general-purpose autonomy over long horizons, a robot needs to maintain spatial environment information that supports a variety of tasks: geometry for planning and control, radiance for rendering and relocalization, and vision-language features for open-vocabulary grounding. Existing methods represent and estimate each modality separately, multiplying memory and compute cost while forgoing potential synergy among the representations. We develop OREN-X, an online mapping method that uses an octree in 3D space as a shared data structure for indexing and storing a multi-modal field, capturing geometric, radiance, and vision-language information. OREN-X provides efficient unified storage and retrieval of these data in explicit/implicit and full/compressed form. Our unified representation yields cross-modality synergy: SDF estimates are sharpened by occupancy and radiance, while GPU-based ray-octree traversal and octree query enable real-time rendering. We also use online dictionary learning to compress the vision-language features, shrinking them 3.7x below full per-vertex storage while raising the query accuracy. On Replica, OREN-X maps in real time (80+ fps for SDF and 30+ fps for all four modalities), improves near-surface SDF accuracy by 33% over single-modality baselines, and improves mean open-vocabulary 3D mIoU by 71% and mean accuracy by 61% over the best prior method.
Chinese Translation
为了实现长时间范围内的通用自主性，机器人需要维护空间环境信息，以支持多种任务：用于规划和控制的几何信息，用于渲染和重定位的辐射信息，以及用于开放词汇基础的视觉-语言特征。现有方法分别表示和估计每种模态，增加了内存和计算成本，同时放弃了表示之间潜在的协同效应。我们开发了OREN-X，一种在线映射方法，使用三维空间中的八叉树作为共享数据结构，用于索引和存储多模态场，捕捉几何、辐射和视觉-语言信息。OREN-X提供了这些数据的高效统一存储和检索，支持显式/隐式和完整/压缩形式。我们的统一表示实现了跨模态协同：占用和辐射信息使得SDF估计更加精确，而基于GPU的光线-八叉树遍历和八叉树查询则实现了实时渲染。我们还使用在线字典学习来压缩视觉-语言特征，将其存储压缩至全顶点存储的3.7倍以下，同时提高查询精度。在Replica上，OREN-X实现了实时映射（SDF超过80帧每秒，四种模态超过30帧每秒），在近表面SDF精度上比单模态基线提高了33%，并在开放词汇3D mIoU的平均值上提高了71%，在平均精度上提高了61%，超越了最佳的先前方法。
cs.RO / 39 / 2609.29166
HarnessPAI: An Evolving Harness for Physical AI
HarnessPAI：一种用于物理人工智能的演变式框架
Wang, Xin, Wu, Wenhao, Zhang, Menghao, Wang, Zhi, Shao, Kun, Luan, Jian, Li, Yang, Li, Qing, Gu, Shangding, Zhou, Huichi, Shi, Shuqing, Ni, Fei, Lu, Shuo, Meng, Weicheng, Li, Kang, Wu, Jin, Zhao, Kang, Guo, Shangmin, Li, Gen, Tang, Yongqiang, Zhang, Zhizhong, Xie, Yuan, Qu, Heng
Abstract
Physical AI aims to build embodied agents that perceive the world, understand and reason about it, and decide how to act. Yet the field has focused primarily on the last component: the action model that maps observations to low-level controls. The prevailing training recipe can erode the perceptual and reasoning capabilities needed for robust behavior, leaving even strong action models vulnerable to scene perturbations and long-horizon tasks. We introduce HarnessPAI, a model- and embodiment-agnostic Harness framework for Physical AI that treats code as the executable and evolvable interface that organizes the underlying action primitive. The framework separates two timescales: within a rollout, it executes open-loop at the program level, with a fixed program guiding and checking execution; across rollouts, it evolves closed-loop, using execution feedback to revise the program and distill failures into reusable skills. Across desktop robot arms, household robots, a robot vacuum, and a legged walking agent, HarnessPAI improves on both pure action models and code-as-policy baselines without retraining the underlying model: a 61.6-point gain over $\pi_{0.5}$ on LIBERO-PRO and a 27.2-point gain over WorldDreamer on RoboCasa atomic tasks. Once a program is selected, rollout execution requires no online high-level LLM deliberation. Beyond execution, the converged program is also a cheap and reliable expert-data collector, and fine-tuning $\pi_{0.5}$ on collected expert data lifts success rate on LIBERO-PRO by 38.8 points. Our results suggest that the frontier of Physical AI depends not only on stronger action models, but also on executable harnesses that integrate perception, task understanding and reasoning, and action execution into a unified, verifiable, and feedback-driven system. Website: https://darwin-agent.github.io/HarnessPAI
Chinese Translation
物理人工智能旨在构建能够感知世界、理解和推理并决定如何行动的具身智能体。然而，该领域主要集中在最后一个组成部分：将观察映射到低级控制的行动模型。现有的训练方法可能会削弱实现稳健行为所需的感知和推理能力，使得即使是强大的行动模型也容易受到场景扰动和长时间任务的影响。我们提出了HarnessPAI，一个与模型和具身性无关的物理人工智能框架，它将代码视为可执行和可演化的接口，组织底层的行动原语。该框架分离了两个时间尺度：在一次执行中，它在程序级别以开放循环的方式运行，固定程序指导并检查执行；在多次执行之间，它以闭环的方式演化，利用执行反馈修订程序并将失败提炼为可重用的技能。在桌面机器人臂、家用机器人、机器人吸尘器和步态机器人等多个平台上，HarnessPAI在不重新训练底层模型的情况下，超越了纯行动模型和代码作为策略的基线：在LIBERO-PRO上相较于$ ext{pi}_{0.5}$提升了61.6分，在RoboCasa原子任务上相较于WorldDreamer提升了27.2分。一旦选择了程序，执行过程无需在线进行高层次的LLM（大语言模型）思考。除了执行，收敛的程序也是一个廉价且可靠的专家数据收集器，对收集的专家数据进行微调的$ ext{pi}_{0.5}$使LIBERO-PRO的成功率提升了38.8分。我们的结果表明，物理人工智能的前沿不仅依赖于更强大的行动模型，还依赖于可执行的框架，这些框架将感知、任务理解和推理以及行动执行整合为一个统一的、可验证的和基于反馈的系统。网站：https://darwin-agent.github.io/HarnessPAI
cs.RO / 40 / 2609.29171
Representation World Model: Learning States, Transition and Executable Plans in Representation
表示世界模型：在表示中学习状态、转移和可执行计划
Abstract
We propose the Representation World Model (RWM), which learns states, transitions, and executable plans directly in representation space. Unlike existing world models that typically learn latent representations together with explicit dynamics models and perform planning through search, optimization, or policy-based prediction, RWM directly incorporates planning into the learned representation geometry. RWM learns the representation geometry by applying inverse-dynamics supervision locally along latent paths constructed from endpoint representations, requiring these paths to preserve task-relevant state and transition information. At inference, planning is performed by directly constructing a latent path between the current and goal representations, with inverse dynamics used to recover the corresponding actions, without recursive rollouts or action-space search. Experiments on continuous-control benchmarks demonstrate the effectiveness of RWM for direct planning, while results on robotic manipulation further show its potential to extend to more complex embodied control tasks. These results suggest that planning directly in representation space provides a promising alternative to conventional world-model planning.
Chinese Translation
我们提出了表示世界模型（Representation World Model, RWM），该模型直接在表示空间中学习状态、转移和可执行计划。与现有的世界模型不同，后者通常将潜在表示与显式动态模型结合学习，并通过搜索、优化或基于策略的预测进行规划，RWM 直接将规划纳入学习到的表示几何中。RWM 通过在从端点表示构建的潜在路径上局部应用逆动态监督来学习表示几何，要求这些路径保留与任务相关的状态和转移信息。在推理时，通过直接构建当前表示与目标表示之间的潜在路径来进行规划，利用逆动态恢复相应的动作，而无需递归展开或动作空间搜索。在连续控制基准上的实验表明，RWM 在直接规划方面的有效性，而在机器人操作上的结果进一步表明其在更复杂的具身控制任务中的潜力。这些结果表明，在表示空间中直接进行规划为传统的世界模型规划提供了一种有前景的替代方案。
cs.RO / 41 / 2609.29176
Anthropomimetic Soft Robotic Forearm with Independently Articulated Carpal Bones Enabling Human-Like Adaptive Stiffness Modulability
具有独立关节腕骨的人形软体机器人前臂，实现类人适应性刚度调节
Abstract
The human wrist exhibits adaptive stiffness modulability: joint stiffness anisotropy can be actively regulated through muscle co-contraction. This functionality is essential for stable manipulation, yet the underlying morphological factors remain unclear. To identify these factors, we developed an anatomically accurate anthropomimetic soft robotic forearm comprising eight independently movable carpal bones interconnected by ligaments, 22 actuated muscles, and compliant fingertips. We measured wrist joint stiffness under four muscle activation patterns across three skeletal configurations: anatomically normal carpal bones, a fused proximal carpal row, and a geometric ellipsoidal skeleton. The stiffness ellipse exhibited low stiffness along the dart-throwing motion (DTM) direction when finger muscles were activated, but high stiffness along the same direction when wrist and finger muscles were activated simultaneously. These results agree with previously reported human measurements, demonstrating that precise anatomical replication reproduces human-like stiffness modulability. Fusing the proximal carpal row eliminated the low DTM-direction stiffness under finger muscle activation, while the geometric ellipsoidal skeleton showed poor stiffness ellipse reorientation across all conditions. Carpal bone motion analysis revealed significantly opposing coupling patterns between wrist and finger muscles at the proximal carpal row, accompanied by a consistent but non-significant trend at the midcarpal joint, providing a mechanical explanation for this modulation. These findings demonstrate that carpal bone morphology plays a dominant role in human wrist stiffness modulation and provide design principles for humanoid robot wrists.
Chinese Translation
人类手腕展现出适应性刚度调节能力：关节刚度的各向异性可以通过肌肉共同收缩主动调节。这一功能对于稳定操作至关重要，但其潜在的形态因素仍不清楚。为了识别这些因素，我们开发了一种解剖学上准确的人形软体机器人前臂，包含八块独立可动的腕骨，通过韧带相互连接，配备22条驱动肌肉和柔性指尖。我们在三种骨骼配置下测量了手腕关节刚度，分别为解剖正常的腕骨、融合的近端腕骨行和几何椭球形骨架。在激活手指肌肉时，刚度椭圆在掷镖运动（DTM）方向上表现出低刚度，但在同时激活手腕和手指肌肉时，该方向的刚度则较高。这些结果与先前报告的人类测量结果一致，表明精确的解剖复制能够再现类人的刚度调节能力。融合近端腕骨行消除了在手指肌肉激活下的低DTM方向刚度，而几何椭球形骨架在所有条件下的刚度椭圆重新定向表现不佳。腕骨运动分析揭示了近端腕骨行中手腕和手指肌肉之间显著相反的耦合模式，同时在中腕关节处伴随有一致但不显著的趋势，为这种调节提供了机械解释。这些发现表明，腕骨形态在人体手腕刚度调节中起主导作用，并为类人机器人手腕的设计提供了原则。
cs.RO / 42 / 2609.29194
Continuous Online Fault Detection for Mobile Robots via Adaptive Edge Models
基于自适应边缘模型的移动机器人连续在线故障检测
Abstract
Mobile robots require robust, real-time fault detection capable of continuous adaptation on constrained edge hardware. While deep time-series models excel at unsupervised anomaly detection, their computational cost prohibits high-frequency onboard execution. This paper bridges this gap via a Teacher-Student distillation framework. An offline foundation model (TSPulse) generates pseudo-labels from unlabeled time series augmented with fault injections. A lightweight MiniRocket Student, adapted with a Recursive Least Squares estimator, approximates this complex decision boundary to execute real-time inference onboard. Evaluations on the TSB-AD benchmark and a physical mobile robot demonstrate the Student achieves a 4.30 ms CPU inference latency. During real-world domain shifts, online adaptation enables the Student to recover from unseen mechanical degradation, improving VUS-PR scores from 0.26 to 0.75 without catastrophic forgetting. Crucially, an uncertainty-guided active learning strategy minimizes operator cognitive load, requesting sparse interventions only when encountering novel fault distributions. These results validate the deployment of state-of-the-art anomaly detection on resource-constrained robotics through offline-to-online distillation.
Chinese Translation
移动机器人需要在受限边缘硬件上具备强大的实时故障检测能力，并能够持续适应。尽管深度时间序列模型在无监督异常检测方面表现优秀，但其计算成本限制了高频率的机载执行。本文通过教师-学生蒸馏框架弥补了这一差距。离线基础模型（TSPulse）从带有故障注入的未标记时间序列中生成伪标签。经过递归最小二乘估计器调整的轻量级MiniRocket学生模型，近似这一复杂决策边界，以便在机载执行实时推理。在TSB-AD基准测试和一台物理移动机器人上的评估表明，学生模型实现了4.30毫秒的CPU推理延迟。在真实世界的领域转变中，在线适应使学生能够从未见过的机械退化中恢复，VUS-PR分数从0.26提高到0.75，且没有发生灾难性遗忘。重要的是，一种基于不确定性的主动学习策略最小化了操作员的认知负担，仅在遇到新型故障分布时请求稀疏干预。这些结果验证了通过离线到在线蒸馏在资源受限的机器人上部署最先进的异常检测的可行性。
cs.RO / 43 / 2609.29198
Assessing the Impact of Fleet Size on Crowdsourced Mapping Using a Dissimilarity Measure
评估车队规模对众包地图的影响：基于不相似性度量
Abstract
Accurate digital maps are essential for Advanced Driver Assistance Systems (ADAS) or Autonomous Driving (AD), providing critical information such as road geometry, traffic signs and speed limits required by safety functions including Intelligent Speed Assistance (ISA). Maintaining these map layers using traditional surveying methods is costly and difficult to scale. Crowdsourced approaches based on fleets provide a promising alternative for continuously validating and updating map information. However, the relationship between the number of contributing vehicles and the quality of the resulting map remains poorly understood. To address this gap, this paper presents a simulation-based framework for evaluating crowdsourced traffic sign maintenance using a dissimilarity measure called GOSPAM (Generalized Optimal SubPattern Assignment for Maps), which combines localization errors with detection performance by accounting for False Positives (FP) and False Negatives (FN). The proposed system models multivehicle observations with representative sensor noise, detection errors, and semantic recognition uncertainties. Observations from multiple vehicles are aggregated using spatial clustering and semantic filtering to estimate traffic sign locations. Using simulated trajectories generated from data carried out by an experimental vehicle in an area containing ground-truth traffic signs, we assess the influence of fleet size on the performance of crowdsourced mapping. The number of vehicles ranges from 5 to 50, and performance is analyzed using standard evaluation metrics which are compared to the GOSPAM . The results show that GOSPAM can be used to effectively assess the quality of crowdsourced mapping, such as the contributions made by the first vehicles or the improvements made by numerous vehicles.
Chinese Translation
准确的数字地图对于高级驾驶辅助系统（ADAS）或自动驾驶（AD）至关重要，它提供了安全功能所需的关键信息，如道路几何、交通标志和限速标志，包括智能限速辅助（ISA）。使用传统测量方法维护这些地图层成本高昂且难以扩展。基于车队的众包方法为持续验证和更新地图信息提供了一种有前景的替代方案。然而，贡献车辆数量与最终地图质量之间的关系仍然不够清楚。为了解决这一问题，本文提出了一种基于仿真的框架，用于评估众包交通标志维护，采用了一种名为 GOSPAM（Generalized Optimal SubPattern Assignment for Maps）的不相似性度量，该度量通过考虑假阳性（FP）和假阴性（FN）将定位误差与检测性能相结合。所提出的系统模拟了多车辆观察，考虑了代表性的传感器噪声、检测误差和语义识别不确定性。通过空间聚类和语义过滤，聚合来自多辆车辆的观察数据，以估计交通标志的位置。利用从实验车辆在包含真实交通标志的区域内收集的数据生成的模拟轨迹，我们评估了车队规模对众包地图性能的影响。车辆数量范围从 5 到 50，性能使用标准评估指标进行分析，并与 GOSPAM 进行比较。结果表明，GOSPAM 可以有效评估众包地图的质量，例如第一辆车的贡献或多辆车所带来的改进。
cs.RO / 44 / 2609.29204
AdaHVLA: Adaptive Harnesses for Long-Horizon Vision-Language-Action Execution
AdaHVLA：用于长时间视语言行动执行的自适应工具
Abstract
Vision-language-action (VLA) models offer strong local control and instruction following but often struggle with long-horizon tasks requiring persistent memory and planning. Task harnesses provide persistent context for agent reasoning by retaining task history and tracking progress across execution stages. To bring these complementary capabilities together, we introduce AdaHVLA, an adaptive harness that refines code-based coordination policies through robot experience to better align agent reasoning and memory with VLA execution. Its decoupled multiagent adaptation process separates evidence analysis, harness revision, and behavioral assessment into distinct working contexts, using testable coordination hypotheses to guide revisions and subsequent rollouts to assess their predicted effects. A stateful revision graph links execution evidence, hypotheses, revisions, and observed effects, preserving alternative harnesses and adaptation memory to guide refinement across repeated attempts and continued adaptation across tasks and environments. In simulation, AdaHVLA raises mean test success on NaVILA-LH from 22.5\% to as high as 57.5\% and improves manipulation test success across three VLA backbones by up to 30.8 percentage points over the initial harness. Real-world deployment further illustrates how the adapted policies support stable execution across task stages.
Chinese Translation
视语言行动（VLA）模型在局部控制和指令跟随方面表现出色，但在需要持久记忆和规划的长时间任务中常常面临挑战。任务工具通过保留任务历史和跟踪执行阶段的进展，为代理推理提供持久的上下文。为了将这些互补能力结合在一起，我们提出了AdaHVLA，一种自适应工具，通过机器人经验优化基于代码的协调策略，以更好地将代理推理和记忆与VLA执行对齐。其解耦的多智能体适应过程将证据分析、工具修订和行为评估分离为不同的工作上下文，利用可测试的协调假设来指导修订和后续的执行，以评估其预测效果。一个状态化的修订图将执行证据、假设、修订和观察到的效果联系起来，保留替代工具和适应记忆，以指导在重复尝试和跨任务及环境的持续适应中的优化。在模拟中，AdaHVLA将NaVILA-LH的平均测试成功率从22.5%提高到高达57.5%，并在三个VLA基础模型上将操作测试成功率提高了多达30.8个百分点。实际部署进一步说明了适应后的策略如何支持跨任务阶段的稳定执行。
cs.RO / 45 / 2609.29212
ADM-Planner: LLM-Guided Long-Horizon Planning for Mobile Manipulators with Attention-Enhanced Dynamic Memory
ADM-Planner：基于大型语言模型的长时间规划框架，适用于具有注意力增强动态记忆的移动操控器
Abstract
Large language models can decompose mobile-manipulation goals into long action sequences, but the resulting plans remain reliable only while their world context is current. A fixed scene description becomes stale when objects are discovered, moved, or completed while retaining every observation instead produces a growing history with redundant and conflicting state. To resolve this tension, we present an LLM-guided planning framework ADM-Planner with attention-enhanced dynamic memory (ADM). Persistent workspace knowledge is separated from object-centric state, asynchronous observations and action outcomes update that state, and a bounded retriever exposes only the entries that can affect the next decision. The LLM replans when an update invalidates the remaining plan. Across 1,500 task-simulator episodes, the proposed ADM achieved 100% full-task success in the 14-container noisy dynamic setting, compared with 62% for static memory and 97% for unfiltered dynamic memory, while reducing the context-size proxy by 95.8% relative to the latter. In a six-episode live GPT-5 Mini planner, both dynamic memory variants completed every mission, while ADM reduced provider-reported input tokens by 14.4% and mean planner calls from 7.0 to 6.0. A separate 60-trial PyBullet study retained 100% success for ADM, compared with 50% for static memory. Finally, the mobile manipulator with ADM-Planner completed various missions in indoor and outdoor physical experiments while incorporating targets revealed after execution began. The results show that selective state maintenance with ADM, rather than prompt history alone, is a practical basis for long-horizon planning in changing environments. Project page: https://xjp99v5.github.io/ADM-Planner
Chinese Translation
大型语言模型能够将移动操控目标分解为长时间的动作序列，但所生成的计划仅在其世界上下文保持最新时才可靠。当对象被发现、移动或完成时，固定的场景描述会变得过时，而保留每个观察结果则会产生一个不断增长的历史，包含冗余和冲突的状态。为了解决这一矛盾，我们提出了一种基于大型语言模型的规划框架ADM-Planner，结合了注意力增强的动态记忆（ADM）。持久的工作空间知识与以对象为中心的状态分离，异步观察和动作结果更新该状态，而有界检索器仅暴露可能影响下一个决策的条目。当更新使剩余计划失效时，LLM会重新规划。在1500个任务模拟器的实验中，所提出的ADM在14个容器的嘈杂动态环境中实现了100%的完整任务成功率，而静态记忆为62%，未过滤的动态记忆为97%，同时相较于后者将上下文大小代理减少了95.8%。在六个回合的实时GPT-5 Mini规划器中，两个动态记忆变体完成了每个任务，而ADM将提供者报告的输入标记减少了14.4%，平均规划调用次数从7.0降低到6.0。另一个60次试验的PyBullet研究中，ADM的成功率保持在100%，而静态记忆为50%。最后，使用ADM-Planner的移动操控器在室内和室外的物理实验中完成了各种任务，同时整合了执行开始后揭示的目标。结果表明，使用ADM进行选择性状态维护，而不仅仅依赖提示历史，是在变化环境中进行长时间规划的实用基础。项目页面：https://xjp99v5.github.io/ADM-Planner
cs.RO / 46 / 2609.29261
Dense-Joint-Based Obstacle-Aided Locomotion with a Joint-Repositionable Snake Robot
基于密集关节的障碍辅助运动与可重定位关节的蛇形机器人
Abstract
Obstacle-aided locomotion is a fundamental capability for snake robots to traverse complex environments. However, conventional rigid-link snake robots often suffer from stagnation or jamming caused by their low joint density (i.e., the number of joints per unit length). This results in discontinuous contact with obstacles, unlike the continuous adaptation of biological snakes. To investigate the effect of joint density on obstacle-aided locomotion performance, we utilized a joint-repositionable snake robot mechanism that decouples actuators from joints, enabling a high-density architecture. We developed two experimental models with identical total lengths but different joint densities (high-density and low-density) and conducted comparative propulsion experiments in obstacle environments with varying obstacle diameters. The experimental results demonstrate that the high-density model substantially suppresses the abrupt shifts in reaction forces that cause stagnation in the low-density model. By maintaining smooth contact points, the high-density configuration reduces power consumption and achieves stable, continuous propulsion. These results highlight high joint density as a key factor in improving the environmental adaptability of snake robots in complex terrains.
Chinese Translation
障碍辅助运动是蛇形机器人穿越复杂环境的基本能力。然而，传统的刚性连杆蛇形机器人常常因其低关节密度（即每单位长度的关节数量）而遭遇停滞或卡滞。这导致与障碍物的接触不连续，而生物蛇则能够持续适应。为了研究关节密度对障碍辅助运动性能的影响，我们利用了一种可重定位关节的蛇形机器人机制，该机制将驱动器与关节解耦，从而实现高密度结构。我们开发了两个实验模型，具有相同的总长度但不同的关节密度（高密度和低密度），并在具有不同障碍物直径的障碍环境中进行了比较推进实验。实验结果表明，高密度模型显著抑制了导致低密度模型停滞的反应力的突然变化。通过保持平滑的接触点，高密度配置降低了能耗，实现了稳定、连续的推进。这些结果强调了高关节密度作为提高蛇形机器人在复杂地形中环境适应性的关键因素。
cs.RO / 47 / 2609.29310
EgoSpeedUp: Transferring Human Manipulation Tempo to Robot Policies
EgoSpeedUp：将人类操作节奏转移至机器人策略
Abstract
Robot manipulation policies trained through imitation learning inherit not only the demonstrated behavior but also the conservative execution tempo of robot demonstrations. Existing acceleration approaches can execute faster than the original demonstrations, but determine the appropriate acceleration primarily from robot-side information or a predefined set of tempo factors, leaving open how to obtain a task-appropriate reference for how fast each manipulation phase should progress. We introduce EgoSpeedUp, a framework that uses human manipulation as temporal supervision for robot imitation learning. Our key insight is that human demonstrations naturally reveal task-appropriate, phase-wise manipulation tempo. Given slow robot demonstrations and human demonstrations of the same task, EgoSpeedUp aligns corresponding manipulation phases, estimates their relative execution tempos from multiple human demonstrations, and transfers the resulting phase-wise tempo by retiming the robot demonstrations. The retimed demonstrations are then used for standard behavior cloning, allowing the robot to retain its executable manipulation behavior while learning to perform it at a human-informed tempo. Across two real-world manipulation tasks, EgoSpeedUp improves the task success rate by an average of 25 percentage points (pp) while reducing successful execution time by 36.5%. These results demonstrate that human manipulation tempo provides an effective temporal reference for learning faster and more reliable robot policies.
Chinese Translation
通过模仿学习训练的机器人操作策略不仅继承了示范行为，还继承了机器人示范的保守执行节奏。现有的加速方法可以比原始示范执行得更快，但主要从机器人端信息或预定义的节奏因子中确定适当的加速方式，尚未解决如何获得每个操作阶段应以多快的速度推进的任务适宜参考。我们提出了EgoSpeedUp，一个利用人类操作作为机器人模仿学习的时间监督的框架。我们的关键见解是，人类示范自然揭示了任务适宜的、阶段性的操作节奏。在给定缓慢的机器人示范和相同任务的人类示范的情况下，EgoSpeedUp对齐相应的操作阶段，从多个的人类示范中估计它们的相对执行节奏，并通过重新定时机器人示范来转移得到的阶段性节奏。重新定时的示范随后用于标准行为克隆，使机器人能够保持其可执行的操作行为，同时学习以人类所知的节奏执行。在两个真实世界的操作任务中，EgoSpeedUp将任务成功率提高了平均25个百分点，同时将成功执行时间减少了36.5%。这些结果表明，人类操作节奏为学习更快且更可靠的机器人策略提供了有效的时间参考。
cs.RO / 48 / 2609.29340
A Simple Gripper Interface for Simulator-Agnostic Cloth Manipulation
一种简单的抓取接口用于与模拟器无关的布料操作
Abstract
This paper presents a grasping model for cloth manipulation specifically tailored to ease the deployment of robotic control methods. The model is robust, fast and easy to implement avoiding at the same time contact and friction considerations between the gripper and the cloth in favor of simple positional constraints. The gripper is described by its pose, jaw state, and an attached grasping volume. Two kinds of grasping volumes are considered: an axis-aligned box to simulate a pinch grasping and a square pyramidal volume to simulate point grasping. When the gripper closes, the discrete cloth positions lying inside this volume are selected, stored in the local gripper frame, and then transported with the gripper motion. A simple squeezing step is also included to progressively move the selected cloth positions toward the center of the grasping region, avoiding an instantaneous displacement at closure. The model can be used in any simulator as it only requires access to discrete cloth positions and a mechanism for imposing target positions as constraints. We implement our grasping model in conjunction with a constraint-based inextensible cloth simulator, where grasping is implemented as moving positional equality constraints coupled with stretch, shear, collision, and table contact projection steps. The same gripper trajectory is applied on a robot arm to fold a real piece of cloth, serving as a simple bridge between simulation and physical cloth manipulation and showcasing the realism and practicality of our idealized grasping model.
Chinese Translation
本文提出了一种专门为布料操作设计的抓取模型，旨在简化机器人控制方法的部署。该模型具有鲁棒性、快速性和易于实现的特点，同时避免了抓手与布料之间的接触和摩擦考虑，而是采用简单的位置信约束。抓手由其姿态、夹爪状态和附加的抓取体积描述。考虑了两种类型的抓取体积：一种是轴对齐的盒子，用于模拟夹持抓取；另一种是方锥体积，用于模拟点抓取。当抓手闭合时，位于该体积内的离散布料位置会被选择并存储在局部抓手坐标系中，然后随着抓手的运动进行传输。还包括一个简单的挤压步骤，以逐步将选定的布料位置移动到抓取区域的中心，避免在闭合时发生瞬时位移。该模型可以在任何模拟器中使用，因为它只需要访问离散布料位置和施加目标位置作为约束的机制。我们将抓取模型与基于约束的不可伸缩布料模拟器结合实现，其中抓取被实现为移动位置相等约束，并结合拉伸、剪切、碰撞和桌面接触投影步骤。相同的抓手轨迹应用于机器人手臂，以折叠一块真实的布料，作为模拟与物理布料操作之间的简单桥梁，展示了我们理想化抓取模型的现实性和实用性。
cs.RO / 49 / 2609.29355
C-space Analysis using Tropical Geometry
使用热带几何的配置空间分析
Abstract
Configuration space~(C-space) of a mechanism is a real variety describing the set of feasible configurations that it can attain. To understand the behavior of a mechanism, it is crucial to identify and scrutinize especially the singular points of its C-space. They usually appear when the variety intersects itself, leading to different branches of motion. There exist many approaches to detect those intersections if they are transversal. However, the problem remains challenging if there are tangential, cuspidal, inter-dimensional or a combination of these intersections. This paper exploits an approach acquired from tropical geometry to analyze the neighborhood of any point on C-spaces of 1-degree-of-freedom~(\emph{dof}) mechanisms. This is done by finding the approximate rational parametrization of the curve(s) passing through the given point using Puiseux series. The proposed approach is shown to succesfully detect the transversal branchings in two foldable four bar mechanisms and a cusp in the configuration curve of the double Watt mechanism.
Chinese Translation
机制的配置空间（C-space）是描述其可达到的可行配置集合的实变体。为了理解机制的行为，识别和仔细审查其C-space的奇异点至关重要。这些奇异点通常出现在变体自交时，导致不同的运动分支。如果交点是横向的，存在许多方法可以检测这些交点。然而，当存在切向、尖点、跨维或这些交点的组合时，问题仍然具有挑战性。本文利用从热带几何中获得的方法来分析1自由度（degree-of-freedom, dof）机制的C-space上任意点的邻域。这是通过使用Puiseux级数找到经过给定点的曲线的近似有理参数化来实现的。所提出的方法成功地检测到两个可折叠四杆机制中的横向分支以及双瓦特机制的配置曲线中的尖点。
cs.RO / 50 / 2609.29374
FMCW-LIO: A Doppler LiDAR-Inertial Odometry
FMCW-LIO：一种多普勒激光雷达-惯性里程计
Abstract
Conventional LiDAR-inertial odometry (LIO) or simultaneous localization and mapping (SLAM) methods heavily rely on geometric features of environments, as LiDARs primarily provide range measurements instead of motion measurements. From now on, however, the situation changes thanks to the novel Frequency Modulated Continuous Wave (FMCW) Doppler LiDARs. FMCW Doppler LiDARs not only offer the point range with high resolution but also capture the instant point Doppler velocity through the Doppler effect. In the letter, we propose FMCW-LIO, a novel and robust LIO, leveraging intrinsic Doppler measurements from FMCW Doppler LiDARs. To correctly exploit Doppler velocities, a motion compensation method is designed, and a Doppler-aided observation model is applied for on-manifold state estimation. Then, dynamic points can be effectively removed by the Doppler criteria, deriving more consistent geometric observations. FMCW-LIO eventually achieves accurate state estimation and static mapping, even in structure-degenerated environments. Extensive experiments in diverse scenes are performed and FMCW-LIO outperforms other algorithms on both accuracy and robustness.
Chinese Translation
传统的激光雷达-惯性里程计（LIO）或同时定位与地图构建（SLAM）方法在很大程度上依赖于环境的几何特征，因为激光雷达主要提供距离测量而非运动测量。然而，得益于新型的频率调制连续波（FMCW）多普勒激光雷达，这种情况正在发生变化。FMCW多普勒激光雷达不仅提供高分辨率的点距离，还通过多普勒效应捕获瞬时点的多普勒速度。在本文中，我们提出了FMCW-LIO，一种新颖且稳健的LIO，利用FMCW多普勒激光雷达的内在多普勒测量。为了正确利用多普勒速度，我们设计了一种运动补偿方法，并应用了一种基于多普勒的观测模型进行流形状态估计。然后，可以通过多普勒标准有效去除动态点，从而获得更一致的几何观测。FMCW-LIO最终实现了准确的状态估计和静态地图构建，即使在结构退化的环境中也能表现出色。在多种场景中进行了广泛的实验，结果表明FMCW-LIO在准确性和稳健性上均优于其他算法。
cs.RO / 51 / 2609.29375
Free-Init: Scan-Free, Motion-Free, and Correspondence-Free Initialization for Doppler LiDAR-Inertial Systems
Free-Init：无扫描、无运动和无对应关系的多普勒激光雷达-惯性系统初始化
Abstract
Robust initialization is crucial for online systems. In the letter, a high-frequency and resilient initialization framework is designed for LiDAR-inertial systems, leveraging both inertial sensors and Doppler LiDAR. The innovative FMCW Doppler LiDAR opens up a novel avenue for robotic sensing by capturing not only point range but also Doppler velocity via the intrinsic Doppler effect. By fusing point-wise Doppler velocity with inertial measurements under non-inertial kinematics, the proposed framework, Free-Init, eliminates reliance on motion undistortion of LiDAR scans, excitation motions, and map correspondences during the initialization phase. Free-Init is also plug-and-play compatible with typical LiDAR-inertial systems and is versatile to handle a wide range of initial motions when the system starts, including stationary, dynamic, and even violent motions. The embedded Doppler-inertial velocimeter ensures fast convergence and high-frequency performance, delivering outputs exceeding 10 kHz. Comprehensive experiments on diverse platforms and across myriad motion scenes validate the framework's effectiveness. The results demonstrate the superior performance of Free-Init, highlighting the necessity of fast, resilient, and dynamic initialization for online systems.
Chinese Translation
稳健的初始化对于在线系统至关重要。本文设计了一种高频且具有韧性的初始化框架，针对激光雷达-惯性系统，利用了惯性传感器和多普勒激光雷达的优势。创新的FMCW多普勒激光雷达通过捕捉点位距离和通过内在多普勒效应获取多普勒速度，为机器人感知开辟了一条新途径。通过在非惯性运动学下将点位多普勒速度与惯性测量融合，所提出的框架Free-Init消除了在初始化阶段对激光雷达扫描的运动去畸变、激励运动和地图对应关系的依赖。Free-Init还与典型的激光雷达-惯性系统兼容，能够灵活处理系统启动时的各种初始运动，包括静止、动态甚至剧烈运动。嵌入式多普勒-惯性测速仪确保了快速收敛和高频性能，输出频率超过10 kHz。在多种平台和各种运动场景下进行的全面实验验证了该框架的有效性。结果表明，Free-Init的性能优越，突显了在线系统快速、韧性和动态初始化的必要性。
cs.RO / 52 / 2609.29382
Decoupled Early Exits for Task-Dependent Compute Allocation in Flow-Matching VLAs
任务依赖计算分配的解耦早期退出在流匹配视觉语言行动中的应用
Abstract
Flow-matching Vision-Language-Action (VLA) models have emerged as a potential solution for generalist robot control, designed by combining a pretrained Vision-Language Model (VLM) backbone with an action expert that generates continuous robot actions. While these models exhibit impressive capabilities, due to their very high number of parameters, their computational requirements are often prohibitive for robotics control. To mitigate these inefficiencies, existing methods predominantly skip VLM backbone layers with early exits or reduce denoising steps, while leaving action expert depth untouched. We propose a framework that exposes backbone depth $V$, action expert depth $A$, and denoising steps $D$ as three jointly configurable compute axes in a VLA. Starting from a pretrained VLA, we attach lightweight Exit Transformers (ET) at intermediate depths in both the backbone and the action expert, trained to distil the last layer of the policy into each exit. Furthermore, we introduce a KV Cache synthesis mechanism that manages the missing keys and values of the skipped backbone layers, allowing the action expert to exit deeper than the backbone. Finally, we show that the optimal compute budget is task-dependent, with different tasks benefiting from different axes and depths. Notably, our method does not require training the original policy from scratch, and for each exit, it increases the number of parameters by only $2.1\%$ for SmolVLA and $4.1\%$ for $\pi_{0.5}$. We validate our approach across two flow-matching VLAs (SmolVLA, $\pi_{0.5}$) and two benchmarks (LIBERO, Meta-World), revealing complementary effects: $V$ and $A$ respectively reduce FLOPs and latency, while $D$ improves both. Our joint configurations $(V,A,D)$ reduce latency by $79.2\%$ and computation (FLOPs) by $31.8\%$, while improving mean success rate by $5.6\%$.
Chinese Translation
流匹配视觉语言行动（VLA）模型作为通用机器人控制的潜在解决方案，通过将预训练的视觉语言模型（VLM）骨干与生成连续机器人动作的行动专家相结合而设计。尽管这些模型展现出令人印象深刻的能力，但由于其参数数量极高，其计算需求往往对机器人控制来说是不可承受的。为了解决这些低效问题，现有方法主要通过早期退出跳过VLM骨干层或减少去噪步骤，同时保持行动专家的深度不变。我们提出了一个框架，将骨干深度$V$、行动专家深度$A$和去噪步骤$D$作为VLA中的三个可共同配置的计算轴。从预训练的VLA出发，我们在骨干和行动专家的中间深度附加轻量级退出变换器（Exit Transformers, ET），训练其将策略的最后一层提炼到每个退出点。此外，我们引入了一种KV缓存合成机制，管理跳过的骨干层的缺失键和值，使得行动专家能够在比骨干更深的层次退出。最后，我们展示了最佳计算预算是任务依赖的，不同任务在不同的轴和深度上受益。值得注意的是，我们的方法不需要从头开始训练原始策略，对于每个退出点，仅增加$2.1\%$的参数用于SmolVLA和$4.1\\%$用于$ ext{π}_{0.5}$。我们在两个流匹配VLA（SmolVLA，$ ext{π}_{0.5}$）和两个基准（LIBERO，Meta-World）上验证了我们的方法，揭示了互补效应：$V$和$A$分别减少了FLOPs和延迟，而$D$则同时改善了两者。我们的联合配置$(V,A,D)$将延迟减少了$79.2\\%$，计算（FLOPs）减少了$31.8\\%$，同时提高了平均成功率$5.6\\%$。
cs.RO / 53 / 2609.29389
Robo-Harness K1: Harnessing Robot-Use Agents via Perception Augmentation
Robo-Harness K1：通过感知增强利用机器人代理
Abstract
Foundation vision-language models (VLMs) understand objects, instructions, and spatial relations, yet translating this capability into robotic manipulation remains difficult. Vision-language-action (VLA) models require extensive demonstrations and may compromise pretrained understanding, while direct RGB-only VLM control is costly and strongly dependent on model capability. We introduce Robo-Harness K1, a robot-use agent (RUA) framework that exposes perception as tools. The agent queries calibrated depth, persistent visual anchors, spatial measurements, and grasp hypotheses, then selects generic motions from the returned evidence. This interface makes 3D geometry accessible without changing the VLM architecture or training a depth encoder. On matched LIBERO-PRO tasks, Gemini 3.7 Flash with K1 reaches 77.8% accuracy, surpassing GPT-6 Astra's 61.1% with an RGB-only harness; K1 further improves Astra to 88.9%. Without target fine-tuning, Gemini with K1 transfers to three RoboSuite arms and dual-arm RoboTwin tasks. On RoboTwin, it achieves 32.0% on Easy and 28.0% on Hard, showing resilience to visual and environmental perturbations. K1 also produces tool-call traces aligned with next-token training. A Qwen3.5-9B student trained on only 107 teacher episodes reaches 44.2% accuracy on new initial states versus 30.2% for OpenVLA, and 13.9% on held-out task conditions versus 0.0% for OpenVLA. These results suggest that perception-augmented RUAs offer a promising route to sample-efficient, generalizable robotic policies that leverage VLM capabilities through an accessible tool interface.
Chinese Translation
基础视觉-语言模型（VLMs）能够理解物体、指令和空间关系，但将这一能力转化为机器人操作仍然困难。视觉-语言-动作（VLA）模型需要大量示范，并可能妨碍预训练理解，而直接使用仅RGB的VLM控制成本高且强烈依赖模型能力。我们提出了Robo-Harness K1，这是一个机器人使用代理（RUA）框架，将感知作为工具进行展示。该代理查询经过校准的深度信息、持久视觉锚、空间测量和抓取假设，然后从返回的证据中选择通用动作。该接口使得3D几何形状在不改变VLM架构或训练深度编码器的情况下可访问。在匹配的LIBERO-PRO任务中，配备K1的Gemini 3.7 Flash达到了77.8%的准确率，超过了使用仅RGB的Harness的GPT-6 Astra的61.1%；K1进一步将Astra的准确率提高至88.9%。在没有目标微调的情况下，配备K1的Gemini能够转移到三个RoboSuite臂和双臂RoboTwin任务。在RoboTwin上，它在简单任务中达到了32.0%的准确率，在困难任务中达到了28.0%，显示出对视觉和环境扰动的韧性。K1还生成与下一个标记训练对齐的工具调用痕迹。仅在107个教师回合上训练的Qwen3.5-9B学生在新的初始状态上达到了44.2%的准确率，而OpenVLA为30.2%；在保留任务条件下达到了13.9%的准确率，而OpenVLA为0.0%。这些结果表明，感知增强的RUA提供了一条有前景的途径，以实现样本高效、可推广的机器人策略，通过可访问的工具接口利用VLM能力。
cs.RO / 54 / 2609.29394
RACaP: Agentic Reasoning, Acting, and Coding as Policies for Evolvable Robot Learning
RACaP：作为可进化机器人学习政策的代理推理、行动和编码
Abstract
General-purpose robot agents must learn from experience, transfer to new tasks, and act efficiently. Code as Policies (CaP) methods generate and repair programs at runtime, incurring latency and entangling reusable mechanisms with task-specific decisions. We introduce RACaP, an agentic framework that moves coding to evolution and uses a Reasoning-and-Acting (ReAct) loop to call frozen, typed Policy APIs at deployment. A two-phase strategy combines capability curriculum learning with autonomous self-evolution to improve the APIs, the ReAct harness, and experience memory. The APIs encode reusable physical mechanisms while exposing arguments for runtime adaptation. ReAct combines task-specific working memory, long-term experience memory, and visual feedback to select actions, verify outcomes, and recover from failures without modifying source code. RACaP achieves 54.4% success on LIBERO-90, 45.0% on zero-shot LIBERO-PRO, and 46.0% on LIBERO-Long, compared with at most 4.0% for CaP baselines on long-horizon tasks. On LIBERO-PRO, it achieves 2.5 times the success rate of CaP baselines and a 1.9-fold speedup in median policy time. For efficient on-robot deployment, rejection-sampled fine-tuning distills GPT-5.6 ReAct decisions into Qwen3-VL-8B-Instruct, yielding a 13.2-fold per-decision inference speedup and reducing repeated physical calls from 16 to 4. These results show that separating reusable code from runtime decisions supports continued evolution, effective transfer, and efficient long-horizon control.
Chinese Translation
通用机器人代理必须从经验中学习、转移到新任务并高效行动。代码作为政策（Code as Policies, CaP）方法在运行时生成和修复程序，导致延迟，并将可重用机制与特定任务的决策纠缠在一起。我们提出了RACaP，这是一个代理框架，将编码转移到进化，并使用推理与行动（Reasoning-and-Acting, ReAct）循环在部署时调用冻结的、类型化的政策API。一个两阶段策略结合能力课程学习与自主自我进化，以改善API、ReAct工具和经验记忆。这些API编码可重用的物理机制，同时暴露运行时适应的参数。ReAct结合特定任务的工作记忆、长期经验记忆和视觉反馈，以选择行动、验证结果并在不修改源代码的情况下从失败中恢复。RACaP在LIBERO-90上取得了54.4%的成功率，在零-shot LIBERO-PRO上为45.0%，在LIBERO-Long上为46.0%，而CaP基线在长时间任务上的成功率最多为4.0%。在LIBERO-PRO上，其成功率是CaP基线的2.5倍，中位政策时间加速了1.9倍。为了高效的机器人部署，拒绝采样微调将GPT-5.6 ReAct决策提炼为Qwen3-VL-8B-Instruct，实现了每个决策推理速度的13.2倍提升，并将重复的物理调用从16次减少到4次。这些结果表明，将可重用代码与运行时决策分离支持持续进化、有效转移和高效的长时间控制。
cs.RO / 55 / 2609.29407
WRAP: Fixtureless Wrench-aware Multi-Robot Assembly Planning
WRAP：无夹具的扳手感知多机器人装配规划
Abstract
Assembly using robots often requires specially designed fixtures, or relies on top-down only assembly strategies. Using multiple robots, we can avoid using fixtures and make robotic assembly more flexible. Planning assembly sequences for multiple robots is challenging due to the high number of possible task assignments and orders. In addition, we need to reason over forces that occur during the assembly process, e.g., to decide if multiple robots are required for support, or if external support such as a table should be used. We present Wrap, a multi-robot assembly planner for multi-part assemblies, given the inter-part ordering-dependencies, the part meshes, and their initial state. We formulate a linear program to reason about valid grasps for supporting the forces that occur during assembly. The search leverages the assembly sequence, and greedily finds a feasible solution per assembly step by computing a heuristic via a cheap backwards search, and using the heuristic in the more expensive forward search. We then solve the multi-robot, multi-goal motion planning problem, and for execution, we split the plan into contact-rich assembly skills, and free space motion. We benchmark the planner on a variety of multi-part assemblies, and apply the planner to groups of robots differing in size and kinematics. We validate the work both in a physics simulation, and in real. Videos and code are available at https://www.vhartmann.com/wrap.
Chinese Translation
使用机器人进行装配通常需要专门设计的夹具，或仅依赖自上而下的装配策略。通过使用多个机器人，我们可以避免使用夹具，使机器人装配更加灵活。为多个机器人规划装配顺序具有挑战性，因为可能的任务分配和顺序数量庞大。此外，我们需要考虑装配过程中出现的力，例如，决定是否需要多个机器人进行支撑，或者是否应使用外部支撑（如桌子）。我们提出了Wrap，一个针对多部件装配的多机器人装配规划器，考虑到部件之间的顺序依赖、部件网格及其初始状态。我们制定了一个线性规划，以推理在装配过程中支持出现的力的有效抓取。搜索利用装配顺序，并通过便宜的反向搜索计算启发式，贪婪地为每个装配步骤找到可行解，然后在更昂贵的前向搜索中使用该启发式。接着，我们解决多机器人、多目标的运动规划问题，并在执行时将计划分解为接触丰富的装配技能和自由空间运动。我们在多种多部件装配上对规划器进行了基准测试，并将规划器应用于不同大小和运动学的机器人组。我们在物理仿真和实际环境中验证了该工作。视频和代码可在 https://www.vhartmann.com/wrap 获取。
cs.RO / 56 / 2609.29417
Singularity Analysis for the Perspective-Four and Five-Line Problems
透视四线和五线问题的奇异性分析
Abstract
This paper deals with image-based visual servoing and pose estimation by observing four and five lines. Our main interest is to determine the relative configurations of the camera and the observed lines that lead to problems in control and stability. Since it is equivalent to finding the singularities of the corresponding Jacobian matrix, we use tools from computational algebraic geometry to seek configurations such that all of its minors vanish simultaneously. By choosing a suitable basis for this matrix, we revisit the problem in the case of three lines to show that one type of the singularities is when the camera lies on the hyperboloid of one sheet uniquely defined by the lines. This result is further exploited to prove that the one-dimensional singularities, if any, in the case of $n$ lines appear when the camera lies on the transversals to the observed lines. Thus, by forcing the transversals to be complex, we can avoid the aforementioned type of singularities in the case of four lines although the algebra shows that there can always be up to 10 inevitable singular locations of the camera for the other type of singularity. For five lines, we find out that there are no singularities in the generic case. The singularities are also characterized for four and five lines with orthogonality and parallelism constraints. Furthermore, a visual servoing library is used to conduct some simulated experiments to substantiate the theoretical results. As expected, we observe problems in control in the vicinity of a singularity as well as increased errors in pose estimation.
Chinese Translation
本文讨论了基于图像的视觉伺服和通过观察四条和五条线进行姿态估计。我们的主要兴趣在于确定相机与观察到的线之间的相对配置，这些配置会导致控制和稳定性问题。由于这等价于寻找相应雅可比矩阵的奇异性，我们使用计算代数几何的工具来寻找所有小行式同时消失的配置。通过为该矩阵选择合适的基，我们重新审视三条线的情况，表明一种奇异性类型是当相机位于由这些线唯一定义的单层双曲面上时。这个结果进一步被利用来证明，在$n$条线的情况下，一维奇异性（如果存在）出现在相机位于观察线的横截面上。因此，通过强制横截面为复数，我们可以避免在四条线的情况下出现上述类型的奇异性，尽管代数表明对于另一种类型的奇异性，相机总是可能存在多达10个不可避免的奇异位置。对于五条线，我们发现一般情况下没有奇异性。四条和五条线在正交性和平行性约束下的奇异性也被表征。此外，使用视觉伺服库进行了一些模拟实验，以验证理论结果。正如预期的那样，我们观察到在奇异性附近控制出现问题，以及姿态估计误差增加。
cs.RO / 57 / 2609.29419
UCON: Uncertainty-aware Navigation with Historical Re-association in Dynamic Environments
UCON：动态环境中基于历史重关联的感知不确定性导航
Abstract
Autonomous navigation in dynamic environments is hindered by two fundamental challenges: perception instability and uncertainty-optimization mismatch. The former leads to identity switches and unreliable motion estimation, while the latter prevents principled incorporation of motion uncertainty into trajectory optimization. To address these challenges, we propose UCON, an uncertainty-aware navigation algorithm in dynamic environments. For perception instability, we present a point-level historical re-association mechanism that leverages historical point cloud fragments to recover lost targets while maintaining identity continuity. Subsequently, a Kalman filter is employed to provide anisotropic motion state estimation and covariance propagation. To resolve the uncertainty-optimization mismatch, we transform predicted states and their covariances into uncertainty sectors, which are embedded as differentiable cost terms within a trajectory optimization framework. This achieves consistent uncertainty-aware dynamic obstacle avoidance while maintaining smoothness and feasibility. Extensive simulations and real-world experiments demonstrate that, while maintaining high computational efficiency, UCON achieves superior perception stability and robust navigation performance in dynamic environments compared to state-of-the-art methods. The code will be open-sourced to facilitate further research.
Chinese Translation
在动态环境中，自主导航面临两个基本挑战：感知不稳定性和不确定性优化不匹配。前者导致身份切换和不可靠的运动估计，而后者则阻碍了将运动不确定性原则性地纳入轨迹优化。为了解决这些挑战，我们提出了UCON，一种在动态环境中具有不确定性意识的导航算法。针对感知不稳定性，我们提出了一种点级历史重关联机制，该机制利用历史点云片段来恢复丢失的目标，同时保持身份连续性。随后，采用卡尔曼滤波器提供各向异性的运动状态估计和协方差传播。为了解决不确定性优化不匹配的问题，我们将预测状态及其协方差转化为不确定性区域，并将其嵌入到轨迹优化框架中的可微分成本项中。这实现了一致的不确定性意识动态障碍物避让，同时保持了平滑性和可行性。大量的仿真和真实世界实验表明，UCON在保持高计算效率的同时，相较于最先进的方法，能够在动态环境中实现更优的感知稳定性和鲁棒的导航性能。代码将开源以促进进一步研究。
cs.RO / 58 / 2609.29423
Temperament Engineering: Designing Strategic Behavioural Diversity in Robot Swarms
气质工程：设计机器人群体中的战略行为多样性
Abstract
No two robots are truly identical: calibration, battery state, sensor drift and wear give every swarm a distribution of behaviour rather than a single point, usually treated as an imperfection to be minimised. In animal collectives the reverse holds: consistent individual differences in behaviour ('temperament') are shaped by natural selection and often decisive for group performance. This perspective proposes 'temperament engineering', a bio-inspired framework that treats the swarm's distribution of temperaments, rather than the individual controller, as the design object. It borrows five evolutionarily validated axes of animal temperament (shyness-boldness, exploration-avoidance, activity, aggressiveness and sociability) as a design vocabulary, rendering each as a continuous control parameter $\tau \in [0,1]$ above the controller, realisable as a module threshold, a policy-conditioning vector in multi-agent reinforcement learning, or a constraint on a foundation-model planner. A three-phase workflow maps mission success criteria onto relevant axes, plans the shape of the $\tau$ distribution, and tunes reaction norms governing how temperament responds to environmental cues. The payoff is greatest under decentralisation: where a central planner can reassign behaviour online, a temperament distribution is a planner output, but in a swarm without global knowledge it must be an offline, anticipatory design input. Behavioural and platform heterogeneity are thereby co-design variables, and I sketch tentative robot-native axes (self-model plasticity, forcefulness, initiative and expressiveness) arising from features robots have and animals do not. Engineered heterogeneity has been shown to outperform homogeneous swarms in tasks such as aggregation and exploration; establishing when, and how much, heterogeneity repays its cost is the work the field can now take forward.
Chinese Translation
没有两个机器人是真正相同的：校准、电池状态、传感器漂移和磨损使得每个群体表现出行为的分布，而不是一个单一的点，这通常被视为需要最小化的缺陷。在动物集体中，情况正好相反：行为中的一致个体差异（“气质”）是由自然选择塑造的，通常对群体表现至关重要。本文提出了“气质工程”，这是一个生物启发的框架，将群体的气质分布视为设计对象，而不是单个控制器。它借鉴了五个经过进化验证的动物气质轴（害羞-大胆、探索-回避、活动性、攻击性和社交性）作为设计词汇，将每个气质表现为一个连续控制参数 $ au ext{ in } [0,1]$，可以实现为模块阈值、在多智能体强化学习中的策略条件向量，或作为基础模型规划器的约束。一个三阶段的工作流程将任务成功标准映射到相关轴上，规划 $ au$ 分布的形状，并调整反应规范，以决定气质如何响应环境线索。在去中心化的情况下，收益最大：当中央规划者可以在线重新分配行为时，气质分布是规划者的输出，但在没有全局知识的群体中，它必须是离线的、预期的设计输入。因此，行为和平台的异质性成为共同设计变量，我勾勒出一些初步的机器人本土轴（自我模型可塑性、果断性、主动性和表现力），这些特征是机器人所具有而动物所不具备的。工程化的异质性在聚集和探索等任务中已被证明优于同质群体；确定何时以及多少异质性能够回报其成本是该领域今后可以推进的工作。
cs.RO / 59 / 2609.29424
Coupled State-Space Modelling, Control, and Policy Distillation for Hybrid Rigid-Pneumatic Manipulators
混合刚性-气动操纵器的耦合状态空间建模、控制与策略蒸馏
Abstract
Hybrid manipulators combine motorized rigid joints with pressure-actuated origami segments. Published arms of this kind are controlled with decoupled per-DOF loops, and the cost of this approximation has not been quantified, because the coupled model needed to measure it has not been built. This paper derives such a model for a chain of $N$ alternating revolute joints and Kresling origami segments, including pneumatic chamber dynamics and crease hysteresis. Using the model, we measure the coupling directly and show that its strength varies joint by joint, and that decoupled control loses precisely on the strongly coupled joints while remaining competitive on the one nearly decoupled joint. Coupled model-based controllers track $2.5\times$ tighter than a decoupled PID baseline at lower torque. However, the model predictive controller (MPC) is too slow for real time, and model-free reinforcement learning stalls far below acceptable success rates on a strict settling metric. We therefore distill the MPC into a small neural policy with behavior cloning and DAgger. The distilled policy settles 93-94$\%$ of goals with zero collisions, within a few points of its teacher, and runs inside the 5 ms control step where the MPC does not. Where the teacher itself fails, we trace the failure to a limit cycle with the bellows' lightly damped mode, and we remove it by selecting goal postures holdable at low pressure.
Chinese Translation
混合操纵器结合了电动刚性关节和压力驱动的折纸段。已发布的此类机械臂采用解耦的每自由度（per-DOF）控制环路，而这种近似的成本尚未量化，因为所需的耦合模型尚未建立。本文推导了一个包含气动腔动态和折痕滞后的 $N$ 个交替旋转关节和 Kresling 折纸段的链条的耦合模型。利用该模型，我们直接测量耦合程度，并表明其强度在每个关节之间有所不同，解耦控制在强耦合关节上表现不佳，而在一个几乎解耦的关节上仍然具有竞争力。基于耦合模型的控制器在较低扭矩下的跟踪精度比解耦的 PID 基线高出 $2.5 imes$。然而，模型预测控制器（MPC）在实时应用中速度过慢，而无模型强化学习在严格的收敛指标上成功率远低于可接受水平。因此，我们通过行为克隆和 DAgger 将 MPC 蒸馏为一个小型神经策略。蒸馏后的策略在没有碰撞的情况下以 93-94$ ext{ extperthousand}$ 的比例完成目标，且与其教师的表现相差无几，并且在 5 毫秒的控制步长内运行，而 MPC 无法做到。在教师自身失败的地方，我们追踪到一个由气囊轻微阻尼模式引起的极限环，并通过选择在低压力下可保持的目标姿态来消除它。
cs.RO / 60 / 2609.29490
RoboLDA: A Probabilistic Generative Model for Uncovering Embodied Hierarchical Structures in Voxel-based Soft Robots
RoboLDA：一种用于揭示基于体素的软机器人中具身层次结构的概率生成模型
Abstract
Recent advances in robotics highlight hierarchical configurations of robot morphology, where multiple levels of functional substructures synergize to facilitate intelligent behaviors. This hierarchical perspective, while particularly advantageous for voxel-based soft robots (VSRs) to ease design and control complexities, is hindered by its heavy reliance on domain expertise. In this work, we address the following question: can we derive such hierarchical design principles solely from existing successful designs? We answer affirmatively by presenting RoboLDA, a Bayesian probabilistic model that decomposes VSR morphology generation into a four-level hierarchy: "task-robot-organ-voxel", and is trained via variational inference. Through extensive experiments on simulated VSRs, we verify the presence of consistent, intuitive hierarchical patterns underlying high-performing VSR designs and showcase RoboLDA's proficiency to extract and leverage these hierarchical priors for zero-shot robot design in unseen tasks. The generated designs, even without further optimization, achieve on average 106.4% of the optimized performance produced by evolutionary algorithms. Additionally, the organ structures inferred by RoboLDA serve as valid functional substructures, significantly enhancing synergistic motion control when integrated with modular control policies. Our work pioneers hierarchical generative modeling of robot morphology, offering a promising pathway towards more interpretable and generalizable development of embodied agents.
Chinese Translation
近期机器人技术的进展突显了机器人形态的层次配置，其中多个功能子结构的协同作用促进了智能行为的实现。这种层次视角在基于体素的软机器人（VSRs）中尤其有利于简化设计和控制的复杂性，但其重度依赖领域专业知识却成为了障碍。在本研究中，我们探讨了以下问题：我们能否仅从现有的成功设计中推导出这样的层次设计原则？我们对此的回答是肯定的，提出了RoboLDA，一个贝叶斯概率模型，它将VSR形态生成分解为四层层次：“任务-机器人-器官-体素”，并通过变分推断进行训练。通过对模拟VSR的广泛实验，我们验证了高性能VSR设计背后存在一致且直观的层次模式，并展示了RoboLDA提取和利用这些层次先验进行零样本机器人设计的能力。生成的设计即使在没有进一步优化的情况下，平均也达到了进化算法优化性能的106.4%。此外，RoboLDA推断出的器官结构作为有效的功能子结构，与模块化控制策略结合时显著增强了协同运动控制。我们的工作开创了机器人形态的层次生成建模，为具身智能体的更具可解释性和可推广性的发展提供了有希望的路径。
cs.RO / 61 / 2609.29491
Generative Evolutionary Design of Voxel-Based Soft Robots with Provable Optimality
具有可证明最优性的基于体素的软机器人生成进化设计
Abstract
Voxel-based soft robots (VSRs) present a promising avenue for developing artificial organisms with lifelike intelligence. However, the vast design spaces and expensive evaluations substantially challenge their design optimization. Here we develop MISCO, a novel evolutionary framework empowered by deep generative models to optimize VSR designs with theoretical guarantees. MISCO integrates an estimation-of-distribution algorithm with a meticulously designed variational autoencoder featuring multi-task learning, position awareness, and inter-voxel signaling. These key components enhance the representational capacity of VSR morphologies and facilitate highly efficient sampling and optimization of morphological distributions. We provide theoretical guarantees for MISCO's asymptotic convergence to globally optimal designs, alongside a favorable convergence rate. Extensive simulated experiments further demonstrate MISCO's exceptional effectiveness in navigating vast design spaces, evolving high-performing VSRs for diverse tasks while flexibly balancing optimization efficiency and morphological diversity. Being validated both empirically and theoretically, MISCO represents a step change towards more scalable and reliable soft robot development.
Chinese Translation
基于体素的软机器人（VSRs）为开发具有类人智能的人工生物体提供了一个有前景的途径。然而，广阔的设计空间和昂贵的评估显著挑战了它们的设计优化。在此，我们开发了MISCO，一个新颖的进化框架，利用深度生成模型优化VSR设计，并提供理论保证。MISCO将分布估计算法与精心设计的变分自编码器相结合，该自编码器具有多任务学习、位置感知和体素间信号传递等特性。这些关键组件增强了VSR形态的表征能力，并促进了形态分布的高效采样和优化。我们为MISCO的渐近收敛到全局最优设计提供了理论保证，并展示了良好的收敛速度。大量模拟实验进一步证明了MISCO在广阔设计空间中的卓越有效性，能够为多种任务进化出高性能的VSR，同时灵活平衡优化效率和形态多样性。经过实证和理论验证，MISCO代表了向更具可扩展性和可靠性的软机器人开发迈出的重要一步。
cs.RO / 62 / 2609.29601
Free the Language Model From the Vision Encoder: Semantic Serialization as a Perception Interface for Small Language Models
解放语言模型于视觉编码器之外：语义序列化作为小型语言模型的感知接口
Abstract
End-to-end vision-language models (VLMs) bind visual competence to the scale of their language model: as the language model shrinks, perception and reasoning degrade together. We study an embodied scene question-answering (QA) interface in which vision never enters the language model. A frozen perception stack detects and ranges objects; a deterministic semantic serializer compiles the perceived state, errors included, into decision-aligned text; an unmodified text-only large language model (LLM) answers. On a visible-scope-matched, occlusion-audited campus-robot benchmark, under a prospectively frozen criterion, the serialized interface, using detectors fine-tuned in-domain within each fold, outperforms a zero-shot VLM whose language model has the same 7B scale (0.7892 vs 0.7462), with a larger margin at 3B (0.7673 vs 0.6913). Preregistered decoupling experiments show the gain survives paraphrase, attributing it to decision-aligned computation rather than answer-string leakage, while novel judgment vocabularies bound its scope. The advantage grows as the reader shrinks to 1.5B and reverses at 0.5B, and a ground-truth oracle locates the reader-capability floor. Under matched task supervision the interfaces converge: a VLM fine-tuned with low-rank adaptation (LoRA) overtakes the zero-shot system but only ties an equally supervised text reader (0.8441 vs 0.8396, no statistically resolved difference), and both routes remain perception-bound. Reported perception parameters are comparable to those of the VLM's vision tower, and total compute is not smaller.
Chinese Translation
端到端的视觉-语言模型（VLMs）将视觉能力与其语言模型的规模绑定在一起：随着语言模型的缩小，感知和推理能力也随之下降。我们研究了一种具身场景问答（QA）接口，其中视觉信息从未进入语言模型。一个冻结的感知堆栈检测并测量对象；一个确定性的语义序列化器将感知状态（包括错误）编译成与决策对齐的文本；一个未修改的仅文本大型语言模型（LLM）进行回答。在一个可见范围匹配、遮挡审核的校园机器人基准测试中，在前景冻结标准下，使用在每个折叠内进行领域微调的检测器的序列化接口，其表现优于一个零-shot VLM，后者的语言模型规模相同（7B）（0.7892 对 0.7462），在3B时差距更大（0.7673 对 0.6913）。预注册的解耦实验表明，性能提升在同义改写中依然存在，归因于与决策对齐的计算而非答案字符串泄漏，同时新颖的判断词汇限制了其范围。随着读者规模缩小至1.5B，优势增强，而在0.5B时则反转，真实的预言者定位了读者能力的下限。在匹配任务监督下，接口趋于收敛：经过低秩适应（LoRA）微调的VLM超越了零-shot系统，但仅与同样受监督的文本阅读器打成平局（0.8441 对 0.8396，未发现统计显著差异），且两条路径仍然受限于感知。报告的感知参数与VLM的视觉塔相当，总计算量并未减少。
cs.RO / 63 / 2609.29644
Markerless Multi-Modal Autonomous Robotic Inspection of Large Space Structures
无标记多模态自主机器人对大型空间结构的检测
Abstract
Future orbital infrastructures, such as deployable antennas, solar farms, and large orbital platforms will require autonomous inspection systems able to operate with limited prior knowledge and without cooperative markers. Current on-orbit servicing approaches often rely on predefined trajectories, standard interfaces, fiducial markers or accurate target models, which limits scalability for large, heterogeneous or partially unknown structures. This paper presents a markerless autonomous robotic inspection pipeline in which 3D reconstruction is used as an inspection-support representation. The system integrates a Kinova Gen2 manipulator with an end-effector-mounted multimodal sensor head composed of an RGB-D camera, a thermal camera and a 2D LiDAR. The pipeline estimates an approximate inspection volume, generates viewpoints, plans collision-free motions with MoveIt, and synchronously records RGB-D images, thermal data, and robot poses in ROS2. Candidate reconstruction methods were evaluated to select a practical method for this pipeline, with Nerfacto used for geometric reconstruction and Thermal-Nerfacto used to demonstrate thermal-aware rendering for inspection. Validation in a Gazebo-based simulator and preliminary laboratory tests reveal that the proposed system can autonomously acquire spatially coherent inspection data and produce reconstructions suitable for visual and geometric assessment, representing a step towards inspection of large non-cooperative space structures.
Chinese Translation
未来的轨道基础设施，如可部署天线、太阳能电站和大型轨道平台，将需要能够在有限的先验知识和没有合作标记的情况下操作的自主检测系统。目前的轨道服务方法通常依赖于预定义的轨迹、标准接口、基准标记或准确的目标模型，这限制了对大型、异构或部分未知结构的可扩展性。本文提出了一种无标记的自主机器人检测流程，其中3D重建被用作检测支持表示。该系统将Kinova Gen2机械臂与安装有多模态传感器头的末端执行器集成，传感器头由RGB-D相机、热成像相机和2D LiDAR组成。该流程估计近似的检测体积，生成视点，使用MoveIt规划无碰撞的运动，并在ROS2中同步记录RGB-D图像、热数据和机器人姿态。评估了候选重建方法，以选择适合该流程的实用方法，采用Nerfacto进行几何重建，采用Thermal-Nerfacto展示热感知渲染用于检测。在基于Gazebo的模拟器和初步实验室测试中的验证表明，所提出的系统能够自主获取空间一致的检测数据，并生成适合视觉和几何评估的重建，代表了对大型非合作空间结构检测的一步进展。
cs.RO / 64 / 2609.29669
Do World Models Make Better Robots? A Survey of Evaluation Benchmarks for Predictive Embodied Intelligence
世界模型是否能造就更优秀的机器人？预测性具身智能评估基准的调查
Abstract
Robot learning now advances along two tracks that rarely meet. On one side, direct Vision-Language-Action (VLA) policies map observations to actions and are scored by closed-loop task success. On the other, predictive and generative world models forecast future observations and are scored by open-loop prediction or generation quality. A natural question sits between them: does world modelling earn a measurable, closed-loop advantage over a direct policy, and for which robotic capabilities? We argue that the field cannot yet answer this question, and that the reason is a gap in how it is measured, not in the models themselves. World-model benchmarks score prediction without ever executing it, while task-success suites host a single policy and never build a world-model versus VLA contrast. This survey maps the evaluation landscape around that gap. We catalogue 160 web-verified benchmarks spanning 2017 to 2026 and organise them by evaluation mode, robotic capability, and model family into four lanes: policy suites, embodied agents, world model evaluation, and prediction-to-action bridges. Across the corpus, 138 of 160 benchmarks are model-agnostic and only 11 (7%) build an explicit VLA-versus-world-model contrast; counterfactual capability is almost entirely unmeasured, and only four benchmarks turn prediction into executed action. We contribute an operational taxonomy, a coverage comparison against the eight closest surveys (ours is the only one to cross capability with model family), an evaluation loop that isolates the advantage of prediction, and an actionable protocol of four advantage-aware metrics anchored on named testbeds. The organising claim is not that world models help or do not help, but that answering the question requires benchmarks built to ask it.
Chinese Translation
机器人学习目前沿着两条很少交汇的轨道发展。一方面，直接的视觉-语言-动作（VLA）策略将观察映射到动作，并通过闭环任务成功率进行评分。另一方面，预测性和生成性世界模型预测未来观察，并通过开放式预测或生成质量进行评分。在它们之间自然存在一个问题：世界建模是否在可测量的闭环中相较于直接策略具有优势，以及对于哪些机器人能力？我们认为该领域尚无法回答这个问题，原因在于测量方式的差距，而非模型本身。世界模型基准在未执行预测的情况下评分，而任务成功套件仅包含单一策略，从未建立世界模型与VLA的对比。本调查绘制了围绕这一差距的评估格局。我们编目了160个经过网络验证的基准，涵盖2017年至2026年，并按评估模式、机器人能力和模型家族将其组织为四个类别：策略套件、具身代理、世界模型评估和预测到动作的桥梁。在这些基准中，160个中有138个是模型无关的，只有11个（7%）建立了明确的VLA与世界模型的对比；反事实能力几乎完全未被测量，只有四个基准将预测转化为执行的动作。我们贡献了一个操作性分类法、与八个最接近的调查的覆盖比较（我们的调查是唯一一个将能力与模型家族结合的），一个隔离预测优势的评估循环，以及一个基于命名测试平台的四个优势感知指标的可操作协议。我们的组织主张并非世界模型是否有助于机器人，而是回答这个问题需要建立专门的基准。
cs.RO / 65 / 2609.29738
Combining Evasive and Braking Reactions for Safety Reference Models in Automated Vehicles
结合规避与刹车反应的自动驾驶安全参考模型
Abstract
Computational models of careful and competent human drivers are essential for scenario-based evaluation of automated driving systems (ADS). However, most existing safety reference models primarily focus on longitudinal braking, neglecting the role of evasive steering in human collision avoidance. This paper proposes a hybrid Fuzzy-Safety Model (FSM-H) that integrates longitudinal mitigation and lateral avoidance within a unified behavioral framework. The braking component is governed by Proactive Fuzzy Safety (PFS) metrics, representing the erosion of longitudinal safety margins, while the steering component is driven by Criticality Fuzzy Safety for lane-change (CFS-LC), capturing lateral conflict severity and maneuver feasibility. A finite-state architecture models the sequential escalation from nominal driving to braking and, when necessary, to evasive steering, incorporating perception-reaction time and lane-check delays to reflect human decision processes. The model is evaluated in reconstructed high-criticality cut-in scenarios and compared with braking-only and steering-only reference strategies. Results show that the hybrid approach expands the preventability envelope while maintaining behavioral plausibility and computational tractability. The proposed framework provides a transparent and explainable human reference model suitable for simulation-based ADS safety benchmarking and regulatory assessment.
Chinese Translation
谨慎且能干的人类驾驶员的计算模型对于基于场景的自动驾驶系统（ADS）评估至关重要。然而，现有的大多数安全参考模型主要集中于纵向刹车，忽视了人类在避免碰撞中规避转向的作用。本文提出了一种混合模糊安全模型（FSM-H），在统一的行为框架内整合了纵向减速和横向规避。刹车部分由主动模糊安全（PFS）指标控制，代表纵向安全边际的侵蚀，而转向部分则由车道变更的关键模糊安全（CFS-LC）驱动，捕捉横向冲突的严重性和机动的可行性。有限状态架构模拟了从正常驾驶到刹车，再到必要时的规避转向的逐步升级，结合感知-反应时间和车道检查延迟，以反映人类决策过程。该模型在重建的高关键性插入场景中进行了评估，并与仅刹车和仅转向的参考策略进行了比较。结果表明，混合方法在保持行为合理性和计算可行性的同时，扩展了可预防性范围。所提出的框架提供了一个透明且可解释的人类参考模型，适用于基于仿真的ADS安全基准测试和监管评估。
cs.RO / 66 / 2609.29750
From Target Selection to Digging: A Learning-Based Framework for Continuous Autonomous Excavation
从目标选择到挖掘：一个基于学习的连续自主挖掘框架
Abstract
Repeated excavation continuously reshapes pile geometry, requiring an autonomous excavator to adapt its digging targets and coordinate motion across successive excavation cycles. We present a learning-based framework for continuous autonomous excavation that integrates terrain-aware target selection with reinforcement- and imitation-learning controllers. The framework separates target-conditioned motion from local digging: a shared task-conditioned RL policy controls waypoint-guided approach and loaded transport, while an IL policy learns vision-based digging and lifting from expert demonstrations. Digging targets are selected from LiDAR elevation maps and converted into bucket-tip waypoints for motion control. The control architecture coordinates the learned policies and deterministic unloading through a shared motion interface. The complete system is deployed on a scaled hydraulic excavator with multimodal sensing and closed-loop actuator control. Offline replay and physical experiments demonstrate more consistent target selection, shorter local motion time, and increased payload compared with the respective baselines. The learned digging policy achieves a mean payload of 6.52 kg per completed cycle, compared with 2.68 kg for Fixed Dig. Three five-scoop runs further demonstrate consecutive autonomous excavation under continuously changing pile geometry.
Chinese Translation
重复的挖掘不断重塑堆积几何形状，这要求自主挖掘机能够适应其挖掘目标并在连续的挖掘周期中协调运动。我们提出了一个基于学习的连续自主挖掘框架，该框架将地形感知的目标选择与强化学习（Reinforcement Learning, RL）和模仿学习（Imitation Learning, IL）控制器相结合。该框架将目标条件运动与局部挖掘分开：一个共享的任务条件RL策略控制基于航点的接近和装载运输，而IL策略则从专家演示中学习基于视觉的挖掘和提升。挖掘目标从激光雷达（LiDAR）高程图中选择，并转换为用于运动控制的铲斗尖航点。控制架构通过共享运动接口协调学习到的策略和确定性卸载。完整系统在具备多模态传感和闭环执行器控制的缩放液压挖掘机上部署。离线重放和物理实验表明，与各自基线相比，目标选择更加一致，局部运动时间更短，载荷增加。学习到的挖掘策略在每个完成周期中实现了平均6.52千克的载荷，而固定挖掘（Fixed Dig）仅为2.68千克。三次五铲的实验进一步展示了在不断变化的堆积几何形状下的连续自主挖掘。
cs.RO / 67 / 2609.29760
PolyUMI: Accessible Visual-Tactile-Audio Data Collection for Object Inference and Manipulation
PolyUMI：可访问的视觉-触觉-音频数据收集用于物体推理与操作
Abstract
Humans typically rely on vision, touch, hearing, and proprioception to perceive contact and adapt their actions during manipulation. Providing robots with comparable responsiveness therefore requires hardware that can retain and use these complementary sensory signals. Most imitation-learning systems, however, observe demonstrations primarily through vision and proprioception, limiting access to contact information that is difficult to infer visually. We present PolyUMI, an open-source platform for scalable visual--tactile--audio demonstration collection and robot deployment. Its lightweight, wireless handheld gripper records synchronized wrist-camera, optical tactile, contact-audio, and proprioceptive observations without requiring a tethered workstation. The same sensing finger can be transferred to the robot end effector, preserving the sensing geometry between demonstration collection and policy execution. To effectively use these heterogeneous observations, we further introduce VisTA, a token-level multimodal policy that integrates information across sensors and time to predict contact-aware robot actions. Experiments spanning object inference, slip control, and contact-rich manipulation show that touch and audio reveal task-relevant information beyond vision and that VisTA is competitive with or outperforms existing multimodal policies. Together, PolyUMI and VisTA provide an accessible pipeline for collecting multimodal demonstrations and learning policies that perceive physical interaction beyond vision. Project Page: https://polyumi-vista.github.io
Chinese Translation
人类通常依赖视觉、触觉、听觉和本体感觉来感知接触并在操作过程中调整其行为。因此，为机器人提供类似的响应能力需要能够保留并使用这些互补感官信号的硬件。然而，大多数模仿学习系统主要通过视觉和本体感觉观察示范，限制了接触信息的获取，而这些信息在视觉上难以推断。我们提出了PolyUMI，一个用于可扩展视觉-触觉-音频示范收集和机器人部署的开源平台。其轻量级、无线的手持夹具能够记录同步的手腕相机、光学触觉、接触音频和本体感觉观察，而无需依赖有线工作站。相同的传感指头可以转移到机器人末端执行器上，保留示范收集与策略执行之间的传感几何关系。为了有效利用这些异构观察，我们进一步引入了VisTA，一种基于令牌的多模态策略，整合跨传感器和时间的信息，以预测与接触相关的机器人动作。涵盖物体推理、滑动控制和接触丰富操作的实验表明，触觉和音频提供了超越视觉的任务相关信息，而VisTA在与现有多模态策略的比较中表现出竞争力或优于它们。PolyUMI和VisTA共同提供了一个可访问的管道，用于收集多模态示范和学习能够超越视觉感知物理交互的策略。项目页面：https://polyumi-vista.github.io
cs.RO / 68 / 2609.29822
Self-Supervised Anchoring of Fingertip Sensing to Proprioception and Proactive Actions for Robot Imitation Learning
自监督锚定指尖感知与本体感觉及主动行为的机器人模仿学习
Abstract
Robotic imitation learning often relies on external cameras, yet local interaction cues such as object proximity, contact onset, and grasp state are difficult to observe near the fingertips because of occlusion and limited temporal resolution. We study how to effectively incorporate complementary fingertip sensing into imitation learning using pressure-sensitive tactile and reflective proximity sensors, along with pretrained sensor encoders. The two modalities provide information at different manipulation phases: proximity sensing is informative before contact, whereas tactile sensing becomes informative after contact. However, naively adding these signals to a policy does not consistently improve performance and can even underperform vision-only policies, suggesting that sparse, phase-dependent sensor signals are difficult to exploit from limited demonstrations. We therefore propose a proprioception-anchored pretraining method, PROprioceptive-and-PRoactive Anchoring (PROPRA), which independently aligns each fingertip sensor history with proprioceptive and action segments. This provides a continuously available sensorimotor reference, allowing each sensor to be aligned independently during its informative phases. Experiments on real-world manipulation tasks show that our pretraining method improves average success rates over vision-only policies and image-anchored pretraining baselines. Representation analysis further shows that it preserves richer information about pre-contact states, enabling more effective use of complementary fingertip sensing. Please refer to our project page: https://tomohiromotoda.github.io/nia.propra/
Chinese Translation
机器人模仿学习通常依赖于外部摄像头，然而由于遮挡和有限的时间分辨率，指尖附近的局部交互线索（如物体接近、接触开始和抓取状态）难以观察。我们研究如何有效地将互补的指尖感知融入模仿学习中，使用压力敏感的触觉传感器和反射式接近传感器，以及预训练的传感器编码器。这两种模态在不同的操作阶段提供信息：接近感知在接触前提供信息，而触觉感知在接触后变得有用。然而，简单地将这些信号添加到策略中并不能始终提高性能，甚至可能表现不如仅依赖视觉的策略，这表明稀疏的、阶段依赖的传感器信号在有限的示范中难以利用。因此，我们提出了一种基于本体感觉的预训练方法，称为本体感觉与主动行为锚定（PROprioceptive-and-PRoactive Anchoring，PROPRA），该方法独立地将每个指尖传感器的历史与本体感觉和动作片段对齐。这提供了一个持续可用的传感器运动参考，使每个传感器在其信息丰富的阶段独立对齐。针对真实世界操作任务的实验表明，我们的预训练方法提高了相较于仅依赖视觉的策略和图像锚定预训练基线的平均成功率。表示分析进一步表明，它保留了关于接触前状态的更丰富信息，从而使互补的指尖感知得以更有效地利用。请访问我们的项目页面：https://tomohiromotoda.github.io/nia.propra/
cs.RO / 69 / 2609.29850
BeyondRetarget: Learning Executable Humanoid Motions Directly from Monocular Video
超越重定向：直接从单目视频学习可执行的人形运动
Abstract
Learning executable motions from human videos offers a scalable solution for humanoid robots to acquire demonstration motions. However, existing pipelines typically first construct an explicit human motion representation and then convert it into robot motions via motion retargeting. Although such methods can effectively leverage large volumes of existing human data for training, the substantial differences between humans and humanoid robots in locomotion mechanisms and joint degree-of-freedom configurations make motions generated by this human-representation-centric approach difficult to execute on robots. Furthermore, errors introduced during human motion estimation inevitably propagate to the retargeting stage and cannot be eliminated via joint optimization. We propose BeyondRetarget, an end-to-end framework that directly maps monocular RGB videos to robot motions. Discarding the explicit human representation, this framework learns robot-oriented implicit representations directly from visual observations, enabling the model to capture cross-morphology motion structures. To generate motions more suitable for robot execution, we further design a contact-aware motion optimization mechanism to improve temporal consistency and physical plausibility. Experiments show that BeyondRetarget significantly improves the accuracy and robustness of generated robot motions, while achieving higher execution success rates and lower latency in both simulation environments and real humanoid robots.
Chinese Translation
从人类视频中学习可执行的运动为人形机器人获取演示动作提供了一种可扩展的解决方案。然而，现有的流程通常首先构建明确的人类运动表示，然后通过运动重定向将其转换为机器人动作。尽管这种方法能够有效利用大量现有的人类数据进行训练，但人类与人形机器人在运动机制和关节自由度配置上的显著差异，使得通过这种以人类表示为中心的方法生成的动作在机器人上执行时面临困难。此外，在人类运动估计过程中引入的误差不可避免地会在重定向阶段传播，并且无法通过联合优化消除。我们提出了BeyondRetarget，一个端到端的框架，直接将单目RGB视频映射到机器人动作。该框架摒弃了明确的人类表示，直接从视觉观察中学习面向机器人的隐式表示，使模型能够捕捉跨形态的运动结构。为了生成更适合机器人执行的动作，我们进一步设计了一种考虑接触的运动优化机制，以提高时间一致性和物理合理性。实验表明，BeyondRetarget显著提高了生成的机器人动作的准确性和鲁棒性，同时在仿真环境和真实人形机器人中实现了更高的执行成功率和更低的延迟。
cs.RO / 70 / 2609.29861
GPT-6-Astra Lights Up Embodied Navigation: Evaluation in Zero-Shot Vision-and-Language Navigation in Continuous Environments
GPT-6-Astra 点亮具身导航：在连续环境中的零-shot 视觉与语言导航评估
Abstract
We investigate whether GPT-6-Astra, a general-purpose foundation model, can navigate unfamiliar environments using its own perception, reasoning, and decision-making capabilities. Our evaluation focues on zero-shot vision-and-language navigation in continuous environments (VLN-CE) through a minimal interface in the Codex harness, aiming to unleash GPT-6-Astra's full potential for navigation. Using monocular RGB, GPT-6-Astra decides when to observe, how to move, and when to stop, without navigation-specific fine-tuning, a trained waypoint predictor, or a pre-built scene map. Our evaluation yields four main findings. First, \textbf{\textit{GPT-6-Astra achieves strong zero-shot navigation performance using only monocular RGB observations}}. On the common-adopted zero-shot R2R-CE benchmark, ultra reasoning achieves a success rate of \textbf{\textit{79.0\%}}, exceeding the strongest reported zero-shot and supervised success rates by \textbf{\textit{13.0}} and \textbf{\textit{6.9}} percentage points, respectively. Second, \textbf{\textit{GPT-6-Astra advances multi-stage language instructions into coherent, adaptive navigation}} by grounding spatial relations, tracking task progress, and revising its actions. Third, \textbf{\textit{reliable route execution and goal verification remain challenging, even with ultra reasoning}}. Plausible local landmark matches do not consistently lead to correct task completion. Fourth, \textbf{\textit{these capabilities motivate rethinking the role of embodied learning}}. Future VLN research should build on foundation models to advance generalizable and reliable embodied intelligence.
Chinese Translation
我们研究了通用基础模型 GPT-6-Astra 是否能够利用自身的感知、推理和决策能力在不熟悉的环境中进行导航。我们的评估集中在通过 Codex harness 的最小接口进行的连续环境中的零-shot 视觉与语言导航（VLN-CE），旨在释放 GPT-6-Astra 在导航方面的全部潜力。使用单目 RGB，GPT-6-Astra 决定何时观察、如何移动以及何时停止，而无需特定于导航的微调、训练的航点预测器或预构建的场景地图。我们的评估得出了四个主要发现。首先， extbf{ extit{GPT-6-Astra 仅使用单目 RGB 观察实现了强大的零-shot 导航性能}}。在普遍采用的零-shot R2R-CE 基准上，超强推理的成功率达到了 extbf{ extit{79.0 ext{%}}}，超出了报告的最强零-shot 和监督成功率，分别提高了 extbf{ extit{13.0}} 和 extbf{ extit{6.9}} 个百分点。其次， extbf{ extit{GPT-6-Astra 将多阶段语言指令转化为连贯的、自适应的导航}}，通过确定空间关系、跟踪任务进展和修正其行动。第三， extbf{ extit{即使在超强推理下，可靠的路线执行和目标验证仍然具有挑战性}}。合理的局部地标匹配并不总是能导致正确的任务完成。第四， extbf{ extit{这些能力促使我们重新思考具身学习的角色}}。未来的 VLN 研究应基于基础模型，以推进可推广和可靠的具身智能。
cs.RO / 71 / 2609.29908
MorphIK: Morphology-Conditioned Neural Inverse Kinematics for Unknown Robots
MorphIK：针对未知机器人的形态条件神经逆运动学
Abstract
Neural models can learn to generate various solutions to the inverse kinematics problem from data, but are usually limited to a single robot. We present MorphIK, a flow-matching model that solves inverse kinematics for revolute-joint-based kinematic chains it has never seen during training. The model uses a transformer architecture to encode the robot's morphology along with the target pose. This encoding then conditions a flow-matching head that generates poses from noise. Trained on purely synthetic data from procedurally generated robots, the model reaches a precision of about 5 cm on unseen real-world robots with 6 to 9 Degrees of Freedom. For higher precision, the model serves as an excellent Prior for further optimization algorithms, reducing error to less than 1 cm after a single step of Damped Least Squares optimization and to sub-1 mm error after 3 steps in most cases. Building on flow matching's generative capabilities to produce highly diverse outputs, our model can efficiently sample the robot's null space, providing a wide variety of configurations for the same pose. Thus, overall, MorphIK allows learning and generalizing neural inverse kinematics for a multitude of known and unknown robots.
Chinese Translation
神经模型可以从数据中学习生成逆运动学问题的各种解决方案，但通常仅限于单一机器人。我们提出了MorphIK，这是一种流匹配模型，能够解决在训练过程中从未见过的基于旋转关节的运动链的逆运动学问题。该模型采用变换器架构对机器人的形态及目标姿态进行编码。这个编码随后为生成噪声姿态的流匹配头提供条件。该模型在纯合成数据上进行训练，数据来自程序生成的机器人，在未见过的现实世界机器人（具有6到9个自由度）上达到了约5厘米的精度。为了获得更高的精度，该模型作为进一步优化算法的优秀先验，在经过一次阻尼最小二乘优化步骤后将误差降低到不足1厘米，并且在大多数情况下经过3个步骤后误差降至小于1毫米。基于流匹配的生成能力，我们的模型能够高效地对机器人的零空间进行采样，为同一姿态提供多种配置。因此，总体而言，MorphIK实现了对众多已知和未知机器人的神经逆运动学的学习和泛化。
cs.RO / 72 / 2609.29914
High-Voltage Optocoupler Amplifier for Electrostatic Actuators
用于静电驱动器的高压光耦合器放大器
Abstract
Many electrostatic actuators require multi-kilovolt drive voltages at sub-milliamp currents, a task poorly suited for conventional switching devices. As an alternative, we demonstrate a high-voltage amplifier using optocouplers as active elements. The amplifier produces a 20-kV peak-to-peak output with up to 500 Hz bandwidth while maintaining a minimal component count. By using optocouplers as linear devices in feedback, lower harmonic distortion and higher bandwidth are achieved than offered by equivalent PWM amplifiers. This design improves the viability of electrostatic actuators by providing a simpler method to achieve useful drive waveforms.
Chinese Translation
许多静电驱动器需要以亚毫安电流驱动的多千伏电压，这一任务不适合传统的开关设备。作为替代方案，我们展示了一种使用光耦合器作为主动元件的高压放大器。该放大器能够产生高达20 kV的峰值到峰值输出，并具有高达500 Hz的带宽，同时保持最小的元件数量。通过将光耦合器作为反馈中的线性元件使用，获得了比等效PWM放大器更低的谐波失真和更高的带宽。这一设计通过提供更简单的方法来实现有用的驱动波形，提高了静电驱动器的可行性。
cs.RO / 73 / 2609.29929
Pairwise Approximation Can Select the Wrong Multi-Robot Plan
成对近似可能选择错误的多机器人计划
Abstract
Multi-robot coordination methods often score a joint plan from singleton and pairwise terms, leaving out the terms that involve three or more robots. We measure the plan-selection regret of two pairwise approximations to delivered coverage using frozen multi-robot trajectories. For each four-robot plan on an indoor exploration benchmark, replaying all 16 robot subsets gives the exact delivered-coverage set function $F$. From the same subset values we compute two pairwise scores: the exact order-2 M\"obius truncation $F_2$, which depends only on the singleton and pair values, and an equal-weight least-squares two-additive fit $G$. Ranking by $F_2$ instead of $F$ changes the selected plan on six of seven maps at the 15 m candidate-generation range in each of two candidate families, with regret up to 0.337 of map coverage. Switching to $G$ reduces the regret but still changes the selection on three of seven maps in each family. The additive score $F_1$, which keeps only the singleton terms, selects the exact winner on six of seven maps in one family and four of seven in the other, against one of seven for $F_2$. We also find that lower average reconstruction error does not guarantee lower selection regret.
Chinese Translation
多机器人协调方法通常通过单个和成对项对联合计划进行评分，而忽略涉及三个或更多机器人的项。我们测量了两种成对近似方法在使用冻结的多机器人轨迹时对交付覆盖的计划选择遗憾。对于每个四机器人计划，在室内探索基准上，重放所有16个机器人子集可以得到确切的交付覆盖集函数 $F$。从相同的子集值中，我们计算了两个成对评分：仅依赖于单个和成对值的确切二阶Möbius截断 $F_2$，以及一个等权重的最小二乘两加性拟合 $G$。在两个候选家族中，按 $F_2$ 排名而非 $F$ 会在七个地图中的六个地图上改变所选计划，在每个候选生成范围内的15米处，遗憾值高达0.337的地图覆盖。切换到 $G$ 虽然减少了遗憾，但在每个家族中仍然在七个地图中的三个地图上改变选择。仅保留单个项的加性评分 $F_1$ 在一个家族中在七个地图中选择了六个地图的确切赢家，而在另一个家族中选择了四个地图的赢家，相比之下，$F_2$ 仅选择了七个地图中的一个。我们还发现，较低的平均重建误差并不保证较低的选择遗憾。
cs.RO / 74 / 2609.29964
World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal
世界行动代理：通过世界行动排练利用视觉语言模型进行机器人操控
Abstract
General-purpose vision-language models (VLMs) bring broad knowledge and spatial reasoning to robot manipulation, yet existing systems either use them indirectly, to predict constraints or write programs, or give them a view of the scene rather than a world in which to act. We present World Action Agent (WAA), a multi-agent harness through which VLMs pilot robots with basic tools, making every decision within a visual action workspace. The workspace has three properties. Contact views, selected automatically from the scene geometry, present the scene around the current interaction. Action rehearsal turns each action into an editable proposal that the agent, alone or through an Imagination Agent, previews and revises against planning feedback before execution. In-view correction closes the loop between observation, rehearsal, and low-level execution, letting the agent remove residual offsets in the view where it observes them. Through the same workspace, WAA acquires embodied procedural knowledge in two ways: it evolves multimodal skills from expert videos and human teaching under evidence-based review and consults them through a Skill Agent, and its interaction traces train smaller VLMs to pilot the same harness. On LIBERO-Pro, WAA with skills evolved only from LIBERO-90 reaches a state-of-the-art 75.6% average success, outperforming end-to-end VLAs, code-as-policy agents, and a visual-harness baseline with the same backbone; the same skills remain effective on robosuite without further learning. Fine-tuning Qwen3.5-9B on harness traces raises its out-of-domain success from 1.7% to 43.3%.
Chinese Translation
通用视觉语言模型（VLMs）为机器人操控带来了广泛的知识和空间推理能力，但现有系统要么间接使用它们来预测约束或编写程序，要么仅为它们提供一个场景视图，而不是一个可以行动的世界。我们提出了世界行动代理（World Action Agent, WAA），这是一个多代理系统，通过它，VLMs能够在视觉行动工作空间内使用基本工具操控机器人，做出每一个决策。该工作空间具有三个特性。接触视图是从场景几何中自动选择的，展示了当前交互周围的场景。行动排练将每个动作转化为可编辑的提案，代理可以单独或通过想象代理（Imagination Agent）预览并根据规划反馈进行修订，然后再执行。在视图内修正则闭合了观察、排练和低级执行之间的循环，使代理能够消除其观察到的视图中的残余偏差。通过同一工作空间，WAA以两种方式获得具身的程序性知识：它从专家视频和人类教学中在基于证据的审查下演化多模态技能，并通过技能代理（Skill Agent）进行咨询；同时，它的交互轨迹训练较小的VLMs以操控相同的系统。在LIBERO-Pro上，仅从LIBERO-90演化的技能使WAA达到了75.6%的先进平均成功率，超越了端到端的视觉语言代理（VLA）、代码作为策略代理和具有相同骨干的视觉操控基线；这些技能在robosuite上同样有效，无需进一步学习。对Qwen3.5-9B进行微调，使其在操控轨迹上的域外成功率从1.7%提高到43.3%。
cs.RO / 75 / 2609.30023
Res-HIL: Human-Guided Residual Reinforcement Learning for Sample-Efficient Dexterous Manipulation
Res-HIL：人类引导的残差强化学习用于样本高效的灵巧操作
Abstract
Imitation learning enables robots to acquire manipulation skills from demonstrations, but the resulting policies can fail outside the training data, while collecting more demonstrations requires substantial human effort. Human-in-the-loop reinforcement learning uses corrective feedback during online training, but typically learns the complete task policy rather than refining a pretrained imitation policy. We introduce Res-HIL, a human-in-the-loop residual reinforcement learning framework that learns corrective actions on top of a frozen imitation policy. Each human intervention provides two complementary learning signals: direct supervision of the residual policy and reward shaping of preceding autonomous behavior. Res-HIL combines these signals with zero initialization of the residual policy to stabilize and accelerate online learning. We evaluate Res-HIL on five contact-rich manipulation tasks spanning high-precision and long-horizon behaviors. With only 20 initial demonstrations, Res-HIL outperforms state-of-the-art full-policy human-in-the-loop reinforcement learning and residual fine-tuning without human guidance on every task after ten minutes of online training. Res-HIL improves its pretrained base policies and outperforms imitation policies trained with five times more demonstrations. An ablation study shows that direct residual supervision is critical to performance, while intervention-aware reward shaping substantially improves training efficiency.
Chinese Translation
模仿学习使机器人能够从示范中获得操作技能，但所产生的策略在训练数据之外可能会失效，而收集更多示范需要大量的人力。人类在环强化学习在在线训练中使用纠正反馈，但通常学习完整的任务策略，而不是对预训练的模仿策略进行细化。我们提出了Res-HIL，一种人类在环的残差强化学习框架，它在冻结的模仿策略之上学习纠正动作。每次人类干预提供了两个互补的学习信号：对残差策略的直接监督和对先前自主行为的奖励塑造。Res-HIL将这些信号与残差策略的零初始化相结合，以稳定和加速在线学习。我们在五个接触丰富的操作任务上评估Res-HIL，这些任务涵盖高精度和长时间行为。在仅有20个初始示范的情况下，Res-HIL在十分钟的在线训练后，在每个任务上都超越了最先进的完整策略人类在环强化学习和没有人类指导的残差微调。Res-HIL改善了其预训练的基础策略，并超越了使用五倍于示范的模仿策略训练的结果。消融研究表明，直接的残差监督对性能至关重要，而干预感知的奖励塑造显著提高了训练效率。
cs.RO / 76 / 2609.30024
Body-Grounded Replanning for Physically Adaptive Manipulation
基于身体状态的物理自适应操控重规划
Abstract
Manipulation requires not only reasoning about the external environment, but also about the robot's physical condition. A strategy may remain geometrically feasible while becoming physically unsuitable due to increased joint load or limited mobility, yet internal physical state is typically used only for low-level control. We propose body-grounded high-level replanning, which uses internal physical state to adapt manipulation strategies during execution. Body-state events trigger strategy replanning, and an LLM interprets the underlying joint-level state, recent execution statistics, and execution history to select a context-dependent alternative, while leaving the task objective and low-level controller unchanged. We evaluate the framework on a reaching task under controlled load and asymmetric mobility constraints in simulation and on a real robot. Our experiments show that body-grounded replanning maintains high task success while reducing physical effort and enabling more efficient strategy adaptation. Additional contact-rich manipulation experiments demonstrate the applicability of the same replanning interface beyond reaching. These results show that internal physical state can inform not only low-level control, but also high-level decisions about how a manipulation task should be performed.
Chinese Translation
操控不仅需要对外部环境进行推理，还需要考虑机器人的物理状态。一个策略在几何上可能是可行的，但由于关节负载增加或运动能力受限而变得在物理上不适合，然而内部物理状态通常仅用于低级控制。我们提出了一种基于身体状态的高级重规划方法，该方法在执行过程中利用内部物理状态来调整操控策略。身体状态事件触发策略重规划，而大型语言模型（LLM）则解释基础的关节级状态、最近的执行统计数据和执行历史，以选择上下文相关的替代方案，同时保持任务目标和低级控制器不变。我们在模拟环境和真实机器人上评估了该框架在受控负载和不对称运动约束下的到达任务。实验结果表明，基于身体状态的重规划在保持高任务成功率的同时，减少了物理努力并实现了更高效的策略适应。额外的接触丰富的操控实验展示了相同重规划接口在超出到达任务的适用性。这些结果表明，内部物理状态不仅可以为低级控制提供信息，还可以为操控任务的执行方式提供高级决策支持。
cs.RO / 77 / 2609.30056
M3GD: Multi-Modal Multi-View Geometric Diffusion for Camera--LiDAR Novel View Synthesis
M3GD：用于相机-激光雷达新视图合成的多模态多视角几何扩散
Abstract
Robotic novel view synthesis (NVS) must recover both visual appearance and metric 3D structure, yet most generative NVS methods rely only on images, overlooking LiDAR, a complementary sensor common on robotic platforms. We present M3GD, a Camera--LiDAR multimodal representation for generative NVS that composes independently pretrained 2D image and 3D point-cloud foundation models without separately pretraining a cross-modal translator. We show that, after camera projection, frozen LiDAR and image features exhibit substantial shared spatial structure, providing a natural cross-modal representation. M3GD conditions generation on LiDAR through this structure: it combines explicit geometry statistics with learned point-cloud descriptors into view-aligned packets on the image-latent grid, injected through a lightweight residual adapter into a multi-view flow-matching generator whose latent space, decoders, and training objective remain intact. On the GrandTour dataset, M3GD improves target-view RGB and depth synthesis over an image-only version of the same backbone. Ablations show that the gains come from pixel-aligned LiDAR content and that target-view LiDAR acts as a geometric query linking the requested view to source observations. Deployment on a ground robot demonstrates practical real-world operation, with a configurable quality--cost trade-off controlled by the number of Euler integration steps.
Chinese Translation
机器人新视图合成（NVS）必须同时恢复视觉外观和度量3D结构，然而大多数生成性NVS方法仅依赖于图像，忽视了激光雷达（LiDAR），这一在机器人平台上常见的互补传感器。我们提出了M3GD，一种用于生成性NVS的相机-激光雷达多模态表示，它独立组合了预训练的2D图像和3D点云基础模型，而无需单独预训练跨模态转换器。我们展示了在相机投影后，冻结的激光雷达和图像特征表现出显著的共享空间结构，提供了一种自然的跨模态表示。M3GD通过这一结构对激光雷达进行生成条件：它将显式几何统计与学习的点云描述符结合，形成在图像潜在网格上的视图对齐数据包，通过轻量级残差适配器注入到一个多视角流匹配生成器中，该生成器的潜在空间、解码器和训练目标保持不变。在GrandTour数据集上，M3GD在目标视图的RGB和深度合成上优于同一骨干网络的仅图像版本。消融实验表明，性能提升来源于像素对齐的激光雷达内容，并且目标视图的激光雷达作为几何查询，将请求的视图与源观测连接起来。在地面机器人上的部署展示了实际的现实世界操作，具有可配置的质量-成本权衡，由欧拉积分步骤的数量控制。
cs.RO / 78 / 2609.30082
Real-Time Force Regulation for Whole-Hand Dexterous Grasping
全手灵巧抓取的实时力调节
Abstract
Robust dexterous grasping requires maintaining physical stability despite contacts interactively evolving across the entire hand. A precomputed force distribution can easily fail under object motion, modeling errors, or external disturbances. In this paper, we present a framework for real-time force regulation over dynamically changing whole-hand contacts. Our method geometrically estimates contacts across all hand links using a tracked object model and proprioception, without requiring tactile sensing at those contacts. It repeatedly recomputes the desired contact-force distribution subject to friction constraints, actuator limits, and an actuation-consistency constraint motivated by classical whole-limb force analysis. We integrate this force-regulation controller with reactive reaching, enabling the hand to acquire a grasp, maintain it under disturbances, and regrasp after losing the object. Simulation experiments without gravity demonstrate improved grasp retention over fixed-allocation and fingertip-only execution under controlled perturbations, while real-world experiments on a 27-DoF arm-hand system demonstrate grasp maintenance and recovery under human-applied disturbances as contacts evolve across the whole hand. Project page: https://sangminkim-99.github.io/reactive-grasp-whole-hand/
Chinese Translation
稳健的灵巧抓取需要在整个手部接触不断变化的情况下保持物理稳定性。预先计算的力分布在物体运动、建模误差或外部干扰下可能会轻易失效。本文提出了一种在动态变化的全手接触上进行实时力调节的框架。我们的方法通过跟踪的物体模型和本体感觉几何地估计所有手部关节的接触，而无需在这些接触点进行触觉感知。该方法在摩擦约束、执行器限制以及基于经典全肢体力分析的执行一致性约束的条件下，反复重新计算所需的接触力分布。我们将这一力调节控制器与反应式抓取结合，使得手部能够在干扰下获取抓取、维持抓取，并在失去物体后重新抓取。无重力的仿真实验表明，在受控扰动下，相较于固定分配和仅指尖执行，抓取保持性得到了改善，而在一个27自由度的手臂-手系统上的真实世界实验则展示了在接触点随着整个手部的演变而进行的抓取维持和恢复能力。项目页面：https://sangminkim-99.github.io/reactive-grasp-whole-hand/
cs.RO / 79 / 2609.30092
Self-Adaptive VLA for Robust Robot Deployment
自适应视觉-语言-动作模型用于稳健的机器人部署
Abstract
While Vision-Language-Action (VLA) models demonstrate impressive capabilities in robotic manipulation, their memoryless nature renders them brittle to test-time environment shifts, particularly hardware shifts caused by wear or imperfect calibration. Enabling these models to self-adapt during deployment without requiring continuous on-site recalibration remains a critical bottleneck for real-world scalability. In this work, we introduce Self-Adaptive VLA, a novel post-training recipe that enables the policy to iteratively adapt to deployment-time hardware shifts leveraging its own rollouts as context. To do so, we first collect policy rollouts under deliberately injected hardware shifts. We then transform the base policy's training data into shift-conditioned expert demonstrations by pre-compensating the expert actions for these known shifts. Next, we introduce a lightweight, plug-in context encoder that compresses the context, including visual observation, proprioception, and actions in the shifted environment, into a latent context token. This token modulates the policy through adaptive layer normalization (AdaLN). Furthermore, we find that context tokens can be ensembled, allowing the policy to iteratively self-correct and mitigate failures step by step. Extensive experiments across four precision-critical bi-manual and dexterous manipulation tasks show that Self-Adaptive VLA recovers over 80% of the base policy's performance under hardware shifts, such as actuation bias and joint encoder offsets. Moreover, Self-Adaptive VLA enables more robust deployment to new workstations compared to the base policy. Our approach provides a pathway for robust large-scale real-world robot deployments and easier maintenance. See videos at https://icefoxzhx.github.io/self-adaptive-vla.
Chinese Translation
尽管视觉-语言-动作（VLA）模型在机器人操作中展现出令人印象深刻的能力，但其无记忆特性使其在测试时环境变化，特别是由于磨损或不完美校准引起的硬件变化下显得脆弱。使这些模型在部署过程中能够自适应，而无需持续的现场重新校准，仍然是现实世界可扩展性的一个关键瓶颈。在本研究中，我们提出了自适应 VLA（Self-Adaptive VLA），这是一种新颖的后训练方法，能够使策略在部署时通过利用自身的回滚（rollouts）作为上下文，迭代适应硬件变化。为此，我们首先在故意注入的硬件变化下收集策略回滚。然后，我们通过对这些已知变化的专家动作进行预补偿，将基础策略的训练数据转化为条件于变化的专家示范。接下来，我们引入了一种轻量级的插件上下文编码器，该编码器将包括视觉观察、身体感知和在变化环境中的动作等上下文压缩为潜在上下文标记（latent context token）。该标记通过自适应层归一化（Adaptive Layer Normalization, AdaLN）来调节策略。此外，我们发现上下文标记可以进行集成，使策略能够迭代自我修正并逐步减轻故障。在四个对精度要求严格的双手和灵巧操作任务中的大量实验表明，自适应 VLA 在硬件变化（如驱动偏差和关节编码器偏移）下恢复了超过 80% 的基础策略性能。此外，与基础策略相比，自适应 VLA 使得在新工作站的部署更加稳健。我们的方法为稳健的大规模现实世界机器人部署和更易于维护提供了一条路径。视频请见 https://icefoxzhx.github.io/self-adaptive-vla。
cs.RO / 80 / 2609.30127
Faster Visuomotor Policy Learning on Action Manifolds via Riemannian MeanFlow
通过黎曼均值流在动作流形上加速视觉运动策略学习
Abstract
Visuomotor policies learn a direct map from raw sensory observations to robot action sequences. Policies based on Diffusion and Flow Matching capture the multimodal distribution over action sequences in an end-to-end manner. This expressivity comes at the cost of multi-step numerical integration of the learned vector field for action generation, which can be expensive and time-consuming, impeding fast control rates required in robotics applications. Furthermore, robot action sequences are usually defined on a smooth, differentiable manifold, requiring that the learned policy respects the intrinsic geometry of the robot's action space. Here, we present Riemannian MeanFlow Policy (RMFP), which learns the conditioned flow map of the probability path on the robot action manifold. Our formulation employs a flow map consistency objective grounded in the data by a Riemannian Conditional Flow Matching anchor. The flow map consistency condition is stable to train and constrains the learned model to finite-time transport, which yields on-manifold action sequence generation with as few as one network function evaluation. We present results on the spherical LASA and Push-T benchmarks, on the Tool Hang and Transport tasks of the Robomimic suite, and on the Franka Kitchen task with manifold-constrained action generation, and demonstrate that RMFP attains performance competitive with prior work at a lower sampling cost. We also employ RMFP on a real-world robotic manipulation task to demonstrate fast action generation under imperfect sensor measurements in the physical world.
Chinese Translation
视觉运动策略学习从原始传感器观测到机器人动作序列的直接映射。基于扩散和流匹配的策略以端到端的方式捕捉动作序列的多模态分布。这种表达能力的代价是需要对学习到的向量场进行多步数值积分以生成动作，这可能是昂贵且耗时的，妨碍了机器人应用中所需的快速控制速率。此外，机器人动作序列通常定义在光滑的可微流形上，这要求学习到的策略尊重机器人动作空间的内在几何特性。在此，我们提出了黎曼均值流策略（Riemannian MeanFlow Policy, RMFP），该策略学习机器人动作流形上概率路径的条件流映射。我们的公式采用了基于数据的流映射一致性目标，通过黎曼条件流匹配锚点进行约束。流映射一致性条件在训练时是稳定的，并将学习到的模型限制在有限时间传输，这使得在仅进行一次网络函数评估的情况下生成流形上的动作序列成为可能。我们在球面LASA和Push-T基准测试、Robomimic套件的工具悬挂和运输任务以及具有流形约束的Franka厨房任务上展示了结果，并证明RMFP在较低的采样成本下达到了与先前工作的竞争性性能。我们还在一个真实世界的机器人操作任务中应用RMFP，以展示在物理世界中不完美传感器测量下的快速动作生成。
cs.RO / 81 / 2609.30134
Training-free Behavior Cloning
无训练行为克隆
Abstract
Neural behavior cloning compresses demonstrations into large models, making individual actions difficult to trace and policy updates costly. Retrieval policies retain access to demonstrations but struggle with mismatch between recorded and live behavior. We introduce Behavior Predictive Control (BPC), which synthesizes policies without end-to-end policy training by combining an action-aware retrieval metric, a Hankel-based action-continuation prior, and a closed-form one-step residual correction. Inspired by behavioral systems theory, BPC predicts future actions by blending stored observation-action data that best reconstructs the recent runtime observation--action history. Across simulated benchmarks and real-robot deployments, BPC is competitive with learned policies such as $\pi_{0.5}$ (surpassing it in some cases), while reducing policy fitting from hours to seconds on consumer GPUs and supporting closed-loop control upwards of 75 Hz on a Jetson Orin Nano. The retrieved demonstration windows and their coefficients also provide an intrinsic estimate of task progress. Retaining demonstrations within the deployed policy makes its predictions traceable to supporting trajectories and enables behavior revision through the demonstration bank.
Chinese Translation
神经行为克隆将示范压缩为大型模型，使得个体动作难以追踪，策略更新成本高昂。检索策略保留了对示范的访问，但在记录行为与实时行为之间存在不匹配的问题。我们提出了行为预测控制（Behavior Predictive Control, BPC），通过结合动作感知检索度量、基于汉克尔（Hankel）的动作延续先验和封闭形式的一步残差修正，合成策略，而无需端到端的策略训练。受行为系统理论的启发，BPC通过混合存储的观察-动作数据来预测未来动作，这些数据能够最好地重构最近的运行时观察-动作历史。在模拟基准测试和真实机器人部署中，BPC与学习策略如$ ext{π}_{0.5}$具有竞争力（在某些情况下超越了它），同时将策略拟合时间从数小时缩短到消费者级GPU上的数秒，并支持在Jetson Orin Nano上高达75 Hz的闭环控制。检索到的示范窗口及其系数还提供了任务进展的内在估计。在部署的策略中保留示范使其预测能够追溯到支持轨迹，并通过示范库实现行为修订。
cs.RO / 82 / 2609.30140
Contact as a Decision Variable: Capability-Tradeoff Contact Selection for Legged Loco-Manipulation
将接触作为决策变量：腿部运动操控的能力权衡接触选择
Abstract
In this paper, we study the joint selection of an environmental support contact and a whole-body configuration for a prescribed loco-manipulation task. A contact may provide greater physical support while restricting the motion required for the task. We formulate this problem through three capability measures: residual wrench, end-effector reach, and base mobility available after satisfying the task requirements, and we balance them against contact acquisition cost. Evaluating these capabilities for every candidate requires repeated whole-body optimizations. To reduce this computational cost, we propose Capability-Tradeoff Contact Selection (CTCS). CTCS screens candidates for contact and task feasibility, groups similar candidates within each surface, and predicts their capabilities from exact anchor evaluations using local sensitivity analysis. It checks these predictions through selective exact evaluations, ranks candidates by capability, and evaluates a shortlist exactly for final selection. We evaluate CTCS in simulations and hardware experiments using a Unitree Go2 quadruped with an AgileX NERO arm across $392$ task conditions with nine available support surfaces. Results show that CTCS outperforms ground-only and fixed-contact support, as it can select support surfaces that provide favorable capability trade-offs for the task. Compared with evaluating every candidate exactly, CTCS achieves approximately $3\times$ speedup while closely matching the resulting mean objective value.
Chinese Translation
本文研究了在规定的运动操控任务中，环境支撑接触与全身配置的联合选择。接触可能提供更大的物理支撑，但同时限制了任务所需的运动。我们通过三种能力度量来表述这个问题：剩余扭矩、末端执行器的可达性，以及在满足任务要求后可用的基础移动性，并将其与接触获取成本进行平衡。对每个候选接触进行这些能力的评估需要重复的全身优化。为了降低计算成本，我们提出了能力权衡接触选择（Capability-Tradeoff Contact Selection, CTCS）。CTCS筛选接触和任务可行性的候选者，将每个表面内的相似候选者进行分组，并通过局部灵敏度分析从精确的锚点评估中预测它们的能力。它通过选择性精确评估来检查这些预测，按能力对候选者进行排名，并对最终选择的候选者进行精确评估。我们在模拟和硬件实验中评估了CTCS，使用Unitree Go2四足机器人和AgileX NERO臂，在$392$个任务条件下，涉及九个可用支撑表面。结果表明，CTCS优于仅地面和固定接触支撑，因为它能够选择为任务提供有利能力权衡的支撑表面。与对每个候选者进行精确评估相比，CTCS实现了约$3 imes$的加速，同时与结果的平均目标值紧密匹配。
cs.RO / 83 / 2609.30213
ReVAMP: Vector-Accelerated Motion Planning for Kinematically-Constrained Systems via Reparameterization
ReVAMP：通过重新参数化实现运动规划的向量加速，适用于运动学约束系统
Abstract
Robots often must satisfy one or more constraints during motion planning for real-world tasks. When such constraints reduce the valid configuration space to a measure-zero subset, sampling based planning algorithms require modifications to draw feasible samples. For many common end-effector constraints, parameterizations built on inverse kinematics (IK) provide an alternate formulation where the constraints are satisfied by construction, allowing directly sampling the feasible set. Despite their elegant approach, parameterized planners have remained slower than vector-accelerated implementations of projection-based approaches, leaving their performance ceiling an open question. We explore a new axis of vectorization built upon reparameterizing the planning space through analytic IK. This approach addresses existing inefficiencies in vectorized projection-based planners and exposes new opportunities for parallelism within the planner. We show that the planner can synthesize plans in microseconds to milliseconds for high dimensional systems (up to 20 dimensions), with complex constraints, up to 10x faster than the current state-of-the-art. Furthermore, we demonstrate how such planning speeds open up avenues for restructuring sequential manipulation pipelines.
Chinese Translation
机器人在进行现实世界任务的运动规划时，通常必须满足一个或多个约束。当这些约束将有效配置空间缩减为测度为零的子集时，基于采样的规划算法需要进行修改以绘制可行样本。对于许多常见的末端执行器约束，基于逆运动学（IK）构建的参数化提供了一种替代的公式，其中约束通过构造得到满足，从而允许直接对可行集进行采样。尽管这种优雅的方法，参数化规划器的速度仍然慢于基于投影的方法的向量加速实现，使其性能上限仍然是一个悬而未决的问题。我们探讨了一种新的向量化轴，基于通过解析逆运动学重新参数化规划空间。这种方法解决了现有向量化投影规划器中的低效问题，并揭示了规划器内部新的并行化机会。我们展示了该规划器能够在微秒到毫秒级别内合成高维系统（最多20维）复杂约束的规划，其速度比当前最先进的方法快达10倍。此外，我们还展示了这种规划速度如何为重构顺序操作管道开辟新的途径。
cs.RO / 84 / 2609.30214
Underwater C3-JEPA: An Object-Centric Cross-View World Model for ROV Salvage
水下 C3-JEPA：一种面向对象的跨视图世界模型用于水下遥控潜水器打捞
Abstract
We present Underwater C$^{3}$-JEPA (cross-view, control-conditioned, context-extended), an object-centric multi-view predictive world model for near-field heavy-load underwater ROV salvage. Without contact sensors, it predicts in latent space how the task-object state evolves through contact interaction and under the hydrodynamic lag of the vehicle, from synchronized multi-view RGB observations and vehicle control signals. C$^{3}$-JEPA encodes multi-camera observations into task-object and context tokens, fuses cross-camera evidence through held-out-view attention, and directly predicts future states conditioned on control. Weak binding anchors the target and gripper at low annotation cost, while SIGReg sharpens the geometric representation. Experiments show that the learned representation transfers substantially more task-relevant information to downstream probes than a reconstruction-free latent baseline, while keeping the predictor lightweight. The resulting predictive interface supports model-predictive-control (MPC) candidate evaluation and imagined-rollout behavior-agent training. Validation on real underwater video shows the same architecture recovering a withheld camera's object state and staying ahead of persistence, so the recipe transfers beyond simulation.
Chinese Translation
我们提出了水下 C$^{3}$-JEPA（跨视图、控制条件、上下文扩展），这是一种面向对象的多视图预测世界模型，适用于近场重载水下遥控潜水器打捞。在没有接触传感器的情况下，它通过潜在空间预测任务对象状态如何在接触交互和水动力滞后下演变，基于同步的多视图 RGB 观测和潜水器控制信号。C$^{3}$-JEPA 将多摄像头观测编码为任务对象和上下文标记，通过保留视图注意力融合跨摄像头证据，并直接预测基于控制的未来状态。弱绑定以低注释成本将目标和抓取器锚定，而 SIGReg 则增强几何表示。实验表明，所学习的表示比无重建潜在基线向下游探测器传递了更多与任务相关的信息，同时保持预测器轻量化。最终的预测接口支持模型预测控制（MPC）候选评估和想象回滚行为代理训练。在真实水下视频上的验证显示，相同的架构能够恢复被保留摄像头的对象状态，并在持久性之前保持领先，因此该方法能够超越仿真。
cs.RO / 85 / 2609.30233
Coding Agents for Generalized Task and Motion Planning Problems
用于广义任务与运动规划问题的编码代理
Abstract
Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinematic, and dynamic constraints. Generalized TAMP addresses this difficulty by exploiting regularities across problem instances to reduce planning effort on new instances. However, existing methods require substantial TAMP-specific engineering. We investigate whether coding agents can automate this process by synthesizing programs that generalize across instances. Given a task description and simulator access, each agent chooses how to interact with the environment while developing a program within a fixed synthesis budget. The program is then frozen and evaluated on unseen instances. We evaluate Claude Code (Opus 5) and Codex (GPT-5.6 Sol and GPT-6 Astra) on 28 simulated environments from KinDER and PDDLStream, with object counts beyond those evaluated in the original benchmark. Across all program synthesis methods, we evaluate 980 generated programs on 100 held-out instances each, 98,000 evaluation episodes in total. Overall, we find that coding agents are surprisingly effective at generalized TAMP: all three agent configurations outperform hand-engineered planners, one-shot generation, and an LLM-based generalized planning baseline in mean success (56% to 95% versus 47% for the planners, on the 16 environments where a planner is available). As object counts grow, the agents' programs maintain higher success than the planner, using an order of magnitude less computation per instance on average. Logs show agents using interaction to calibrate physical models, test edge cases, and refine strategies. We release all code, including the full prompts given to the agents. These findings suggest that coding agents are a strong baseline for generalized TAMP.
Chinese Translation
任务与运动规划（TAMP）问题即使在完全可观测和以对象为中心的状态下仍然困难，因为离散决策与几何、运动学和动态约束紧密相关。广义 TAMP 通过利用问题实例之间的规律性来解决这一困难，从而减少对新实例的规划工作。然而，现有方法需要大量特定于 TAMP 的工程工作。我们研究编码代理是否能够通过合成跨实例的程序来自动化这一过程。给定任务描述和模拟器访问权限，每个代理在固定的合成预算内选择如何与环境互动，同时开发程序。然后，该程序被固定并在未见过的实例上进行评估。我们在来自 KinDER 和 PDDLStream 的 28 个模拟环境中评估了 Claude Code（Opus 5）和 Codex（GPT-5.6 Sol 和 GPT-6 Astra），对象数量超出了原始基准测试中的评估范围。在所有程序合成方法中，我们在每个 100 个保留实例上评估了 980 个生成的程序，总共进行了 98,000 次评估。总体而言，我们发现编码代理在广义 TAMP 中表现出惊人的有效性：所有三种代理配置在平均成功率上均优于手工设计的规划器、一键生成和基于 LLM 的广义规划基线（在有规划器的 16 个环境中，成功率为 56% 至 95%，而规划器为 47%）。随着对象数量的增加，代理的程序在每个实例上平均使用的计算量减少了一个数量级，成功率仍高于规划器。日志显示代理使用交互来校准物理模型、测试边缘案例和优化策略。我们发布了所有代码，包括提供给代理的完整提示。这些发现表明，编码代理是广义 TAMP 的一个强有力的基线。
cs.RO / 86 / 2609.30247
Rolling-WAM: World Action Models with Rolling Imagination
滚动世界动作模型：具有滚动想象的世界动作模型
Abstract
World Action Models (WAMs) couple action generation with future visual prediction for robotic manipulation. However, completing the joint video-action denoising process at each replanning cycle incurs substantial latency, delaying action updates and limiting closed-loop responsiveness. We present Rolling-WAM, a formulation that distributes joint denoising across successive replanning cycles. Our method maintains a sliding window of video-action chunks at staggered noise levels. At each step, a rolling noise schedule fully denoises the imminent action chunk for execution, while partially refining farther-future chunks. As the window advances with new camera observations, the retained future chunks continue their denoising process. This distributes the computational cost over time while carrying an evolving visual-action context across chunk boundaries. Evaluations on LIBERO, RoboTwin, and a real-world Unitree G1 humanoid show that Rolling-WAM achieves competitive manipulation performance. By removing the need to denoise the entire prediction horizon from scratch, it delivers a 4.5x steady-state replanning speedup over standard joint WAMs.
Chinese Translation
世界动作模型（WAMs）将动作生成与未来视觉预测结合用于机器人操作。然而，在每个重新规划周期完成联合视频-动作去噪过程会导致显著的延迟，从而延缓动作更新并限制闭环响应能力。我们提出了滚动世界动作模型（Rolling-WAM），一种在连续重新规划周期中分配联合去噪的公式。我们的方法在不同噪声水平下保持视频-动作片段的滑动窗口。在每一步，滚动噪声调度完全去噪即将执行的动作片段，同时部分精炼更远的未来片段。随着窗口在新的相机观测下推进，保留的未来片段继续其去噪过程。这在时间上分配了计算成本，同时在片段边界之间携带不断演变的视觉-动作上下文。在LIBERO、RoboTwin和真实世界的Unitree G1人形机器人上的评估表明，滚动世界动作模型实现了具有竞争力的操作性能。通过消除从头去噪整个预测视野的需求，它在稳态重新规划速度上比标准联合WAMs提高了4.5倍。
cs.RO / 87 / 2609.30249
RAPID: Robot Agentic Programming from Demonstrations
RAPID：基于示范的机器人自主编程
Abstract
Coding agents have demonstrated enormous success in solving complex programming problems. To leverage their potential for robot systems, this work introduces Robot Agentic Programming from Demonstrations (RAPID), which automatically generates, verifies, and refines robot programs, given a single visual human demonstration. The iterative agentic loop of code refinement requires several key ingredients: (i) a testable task specification, (ii) action primitives for robot execution, and (iii) an interactive environment for program execution and verification. RAPID infers all three from the demonstration automatically. To make the resulting program reusable beyond the demonstration setting, RAPID uses an object-centric relational program representation that focuses on the underlying structure of the demonstrated strategy rather than the specific motion per se: it expresses the action primitives as trajectory-optimization programs that realize object-level motion effects, while composing them through relational constraints that capture scene-specific geometry at run time. We evaluated RAPID in simulation on eight challenging contact-rich nonprehensile manipulation tasks as well as general prehensile manipulation tasks in the LIBERO-Pro benchmark. We also successfully deployed it on a real Franka arm and evaluated on all eight nonprehensile tasks. In all experiments, RAPID demonstrated strong performance, with generalization over object pose, shape, material, and environment. Website: https://yuyaoliu.me/projects/rapid.
Chinese Translation
编码代理在解决复杂编程问题方面取得了巨大的成功。为了利用其在机器人系统中的潜力，本研究提出了基于示范的机器人自主编程（RAPID），该方法能够根据单个视觉人类示范自动生成、验证和优化机器人程序。代码优化的迭代自主循环需要几个关键要素：（i）可测试的任务规范，（ii）用于机器人执行的动作原语，以及（iii）用于程序执行和验证的交互环境。RAPID能够自动从示范中推断出这三者。为了使生成的程序在示范环境之外可重用，RAPID采用了一种以对象为中心的关系程序表示，重点关注示范策略的基本结构，而不是具体的运动本身：它将动作原语表达为实现对象级运动效果的轨迹优化程序，并通过捕捉场景特定几何形状的关系约束在运行时组合这些程序。我们在模拟环境中对RAPID进行了评估，涵盖了八个具有挑战性的接触丰富的非抓取操作任务，以及LIBERO-Pro基准中的一般抓取操作任务。我们还成功地在真实的Franka机械臂上部署了该方法，并在所有八个非抓取任务上进行了评估。在所有实验中，RAPID表现出强大的性能，能够在物体姿态、形状、材料和环境上进行泛化。网站：https://yuyaoliu.me/projects/rapid。
cs.CV / 1 / 2609.28539
$\unicode{x1F493}$Heartian: Physiology-Aware Relightable Gaussian Head Avatar
💓Heartian：生理感知的可重光照高斯头部头像
Abstract
Gaussian head avatars typically model intrinsic facial appearance as temporally static, omitting subtle cardiac-induced skin-color variation. We propose $\unicode{x1F493}$Heartian, a physiology-aware modulation framework that learns cardiac-cycle-dependent per-frame albedo modulation of facial skin-region Gaussians within a relightable head avatar to encode remote photoplethysmography (rPPG) signals. Using synchronized contact PPG supervision, $\unicode{x1F493}$Heartian models the prescribed cardiac waveform as the sum of two Gaussian functions and learns per-frame spatial residuals via a lightweight MLP. Across 152 stationary recordings from UBFC-rPPG, PURE, and MMPD, attribute-space recovery of the supplied signal achieves a pooled recording-level heart-rate MAE of 0.29 bpm and MAPE of 0.38%. The signals remain detectable after rendering by benchmark rPPG methods, with the best tested configuration - a motion-augmented TS-CAN decoder pretrained on UBFC-rPPG - recovering heart rate from the rendered MMPD avatars at 0.97 bpm MAE and 1.21% MAPE. Meanwhile, $\unicode{x1F493}$Heartian maintains reconstruction quality comparable to the baseline, with negligible average PSNR degradation of 0.005 dB. Overall, our work embeds recoverable rPPG signals as controllable material attributes to subject-specific Gaussian head avatars while retaining the reconstruction quality.
Chinese Translation
高斯头部头像通常将内在面部外观建模为时间上静态，忽略了心脏引起的细微肤色变化。我们提出了💓Heartian，一种生理感知的调制框架，学习心动周期依赖的每帧面部皮肤区域高斯的反照率调制，以编码远程光电容积描记（rPPG）信号。通过同步的接触式PPG监督，💓Heartian将规定的心脏波形建模为两个高斯函数的总和，并通过轻量级多层感知机（MLP）学习每帧的空间残差。在来自UBFC-rPPG、PURE和MMPD的152个静态录音中，提供信号的属性空间恢复实现了0.29 bpm的录音级心率平均绝对误差（MAE）和0.38%的平均绝对百分比误差（MAPE）。经过基准rPPG方法渲染后，信号仍然可检测，最佳测试配置 - 在UBFC-rPPG上预训练的运动增强TS-CAN解码器 - 从渲染的MMPD头像中恢复心率，MAE为0.97 bpm，MAPE为1.21%。同时，💓Heartian保持了与基线相当的重建质量，平均峰值信噪比（PSNR）降级微乎其微，仅为0.005 dB。总体而言，我们的工作将可恢复的rPPG信号嵌入为可控的材料属性，应用于特定个体的高斯头部头像，同时保持重建质量。
cs.CV / 2 / 2609.28580
Token Clustering and Semantic Sequence Mamba for Hyperspectral Image Classification
用于高光谱图像分类的标记聚类与语义序列Mamba
Abstract
Although hyperspectral images (HSIs) provide rich spectral-spatial information, accurate pixel-level classification remains challenging because of spectral-spatial heterogeneity and complex spatial structures. Existing vision state space models (Mamba) typically construct sequences according to predefined spatial neighborhoods, without explicitly accounting for semantic similarity or spatial non-stationarity. To address this limitation, we propose Token Clustering and Semantic Sequence Mamba (STMamba), which organizes sparse tokens into semantically coherent sequences for hyperspectral image classification with the following features. First, at the macro level, a hierarchical encoder decoder progressively selects semantic tokens with the Token Clustering Module (TCM) and restores dense features using a parameter-free Cross-scale Neighborhood Attention (CNA) Upsampler. Second, at the micro level, TCM first identifies representative cluster centers through density-aware clustering and estimates soft memberships based on feature similarity. A quadtree-based dynamic selection strategy then retains sparse and spatially distributed tokens from each semantic cluster, forming coherent semantic-token sequences while reducing redundant pixel-wise representations. Third, parallel Spatial and Spectral Semantic-wise Sequencing Mamba (SWSM) modules capture complementary long-range spatial and spectral dependencies within homogeneous semantic token sequences while suppressing irrelevant interactions across heterogeneous regions. Experimental results on three large-scale benchmark datasets demonstrate that STMamba outperforms the SOTA methods with respect to quantitative and qualitative results.
Chinese Translation
尽管高光谱图像（HSI）提供了丰富的光谱-空间信息，但由于光谱-空间异质性和复杂的空间结构，准确的像素级分类仍然具有挑战性。现有的视觉状态空间模型（Mamba）通常根据预定义的空间邻域构建序列，而未明确考虑语义相似性或空间非平稳性。为了解决这一局限性，我们提出了标记聚类与语义序列Mamba（STMamba），该方法将稀疏标记组织成语义一致的序列，以用于高光谱图像分类，具有以下特点。首先，在宏观层面，分层编码解码器通过标记聚类模块（TCM）逐步选择语义标记，并使用无参数的跨尺度邻域注意力（CNA）上采样器恢复密集特征。其次，在微观层面，TCM首先通过密度感知聚类识别代表性聚类中心，并基于特征相似性估计软成员资格。然后，基于四叉树的动态选择策略从每个语义聚类中保留稀疏且空间分布的标记，形成一致的语义标记序列，同时减少冗余的像素级表示。第三，平行的空间和光谱语义序列Mamba（SWSM）模块在同质语义标记序列中捕捉互补的长距离空间和光谱依赖关系，同时抑制异质区域之间的不相关交互。对三个大规模基准数据集的实验结果表明，STMamba在定量和定性结果方面优于现有的最先进方法（SOTA）。
cs.CV / 3 / 2609.28610
UltraBench 2: Towards Robust Evaluation of Vision Foundation Models on Ultrasound
UltraBench 2：朝着对超声波视觉基础模型的稳健评估迈进
Abstract
Benchmarking is an increasingly critical part of research in machine learning and the domains where it is applied, including healthcare. Yet, despite the steady development of new ultrasound foundation models in recent years, the development of well-designed benchmarks to evaluate them has lagged behind. This deficiency has led to fragmented and inconsistent evaluations of competing models, making it difficult to measure progress. To address this issue, we introduce UltraBench 2, a comprehensive benchmark with wide anatomical and task coverage, and a focus on standardization, reproducibility, and ease-of-use. Using this benchmark, we compare existing vision foundation models for ultrasound image analysis. Our analyses demonstrate that ultrasound-specific pretraining still leads on classification, but that state-of-the-art general-purpose models have drawn level on segmentation.
Chinese Translation
基准测试在机器学习及其应用领域（包括医疗保健）中正变得越来越重要。然而，尽管近年来新的超声波基础模型不断发展，旨在评估它们的良好设计基准的开发却滞后于此。这一缺陷导致了对竞争模型的评估碎片化和不一致，使得衡量进展变得困难。为了解决这一问题，我们推出了UltraBench 2，这是一个具有广泛解剖和任务覆盖范围的综合基准，重点关注标准化、可重复性和易用性。利用这一基准，我们比较了现有的超声图像分析视觉基础模型。我们的分析表明，超声特定的预训练在分类任务中仍然领先，但最先进的通用模型在分割任务上已达到同等水平。
cs.CV / 4 / 2609.28645
PePESeg3D: Perception Prior Enhances Multi-Scale Segmentation for 3D Gaussian Splatting
PePESeg3D：感知先验增强多尺度3D高斯分割
Abstract
Recent advancements in 3D Gaussian Splatting (3DGS) have extended its capabilities to multi-scale segmentation. Existing methods reconstruct a scene with Gaussian primitives and learn multi-scale segmentation features separately, which leaves the geometry unaware of semantic structure and the feature learning dependent on incomplete mask supervision. To address these limitations, we present PePESeg3D, a novel framework that injects perception priors into a multi-scale 3D Gaussian segmentation pipeline. To fully exploit perception priors, we integrate them not only into contrastive feature learning but also into the upstream geometry reconstruction. Specifically, PePE Reconstruction incorporates monocular depth and mask constraints to ensure semantically coherent object structures. Building on this aligned geometry, PePE Contrastive Learning leverages dense depth-color cues and view-consistent centroid supervision to compensate for the incompleteness of multi-scale masks obtained from a 2D foundation model. Extensive experiments on the SPIn-NeRF, LERF-Mask, and NVOS benchmarks demonstrate that PePESeg3D achieves state-of-the-art performance in both multi-scale segmentation and scene reconstruction, highlighting the importance of integrating perception priors into both geometry optimization and feature learning for accurate multi-scale 3D segmentation. Our code is available at https://github.com/BeCow5X5/PePESeg3D.
Chinese Translation
最近在3D高斯分割（3DGS）方面的进展扩展了其在多尺度分割中的能力。现有方法通过高斯原语重建场景，并分别学习多尺度分割特征，这使得几何形状无法感知语义结构，同时特征学习依赖于不完整的掩模监督。为了解决这些局限性，我们提出了PePESeg3D，一个将感知先验注入多尺度3D高斯分割管道的新框架。为了充分利用感知先验，我们不仅将其整合到对比特征学习中，还融入到上游几何重建中。具体而言，PePE重建结合了单目深度和掩模约束，以确保语义一致的物体结构。在此对齐的几何基础上，PePE对比学习利用密集的深度-颜色线索和视图一致的质心监督，以弥补从2D基础模型获得的多尺度掩模的不完整性。在SPIn-NeRF、LERF-Mask和NVOS基准上的大量实验表明，PePESeg3D在多尺度分割和场景重建方面均实现了最先进的性能，突显了将感知先验整合到几何优化和特征学习中的重要性，以实现准确的多尺度3D分割。我们的代码可在https://github.com/BeCow5X5/PePESeg3D获取。
cs.CV / 5 / 2609.28684
M-plicits: Neural Implicit Surfaces via Nested Multiscale Residuals
M-plicits：通过嵌套多尺度残差实现的神经隐式表面
Abstract
Encoding input coordinates with sinusoidal functions into multi-layer perceptrons (MLPs) has proven effective for implicit neural representations (INRs) of surfaces defined as zero-level sets. However, existing methods often struggle to balance training efficiency, rendering speed, and noise robustness: single-MLP approaches are expensive at inference, grid-based representations are fast but can limit surface smoothness and overfit input noise, and previous multiscale approaches frequently capture noise and produce artifacts due to hard spectral truncation. To address these limitations, we propose M-plicits, a multiscale framework that models surfaces as a residual sum of MLPs trained via a sequence of nested neighborhoods. Unlike existing residual approaches that rely on standard domain-wide sampling and require costly mesh extraction for visualization, our method strictly localizes supervision to narrow bands around the previous zero-level sets. This nested design naturally provides robustness against noisy input data: the coarse network acts as a low-pass filter that establishes a clean geometric prior, while subsequent residuals progressively refine the geometry without fitting to high-frequency artifacts. We further introduce a multiscale sphere-tracing algorithm and a GEMM-based analytical normal computation that bypasses auto-differentiation entirely, yielding high-fidelity real-time rendering. On Stanford and Thingi32, M-plicits achieves the best mean Chamfer distance in the coarse configuration and the best median Chamfer distance and IoU in the fine configuration, with substantially better noise robustness than iNGP, BACON, and IDF, while using an order of magnitude fewer parameters than grid-based baselines. Code, models, and data will be released at https://github.com/dsilvavinicius/m-plicits.
Chinese Translation
将输入坐标与正弦函数编码到多层感知器（MLPs）中已被证明对定义为零级集的隐式神经表示（INRs）有效。然而，现有方法往往难以平衡训练效率、渲染速度和抗噪声能力：单一 MLP 方法在推理时代价高昂，基于网格的表示速度较快但可能限制表面光滑度并过拟合输入噪声，而之前的多尺度方法由于硬谱截断常常捕捉噪声并产生伪影。为了解决这些局限性，我们提出了 M-plicits，一种多尺度框架，将表面建模为通过一系列嵌套邻域训练的 MLP 残差和。与依赖于标准领域采样并需要昂贵的网格提取进行可视化的现有残差方法不同，我们的方法严格将监督局限于前一个零级集周围的狭窄带。这样的嵌套设计自然提供了对噪声输入数据的鲁棒性：粗糙网络充当低通滤波器，建立干净的几何先验，而后续的残差逐步细化几何形状，而不适应高频伪影。我们进一步引入了一种多尺度球追踪算法和一种基于 GEMM 的解析法线计算方法，完全绕过自动微分，从而实现高保真实时渲染。在斯坦福和 Thingi32 数据集上，M-plicits 在粗糙配置中实现了最佳的平均 Chamfer 距离，在精细配置中实现了最佳的中位 Chamfer 距离和交并比（IoU），并且在抗噪声能力上显著优于 iNGP、BACON 和 IDF，同时使用的参数数量比基于网格的基线少一个数量级。代码、模型和数据将发布在 https://github.com/dsilvavinicius/m-plicits。
cs.CV / 6 / 2609.28741
GeoNLI - A Natural Language Interpreter for Satellite Imagery
GeoNLI - 卫星影像的自然语言解释器
Abstract
Multi-modal multitasking models have shown strong performance on remote sensing datasets. However, because these models are trained on heterogeneous data and vary across tasks, designing a unified model that performs well in captioning, visual question answering (VQA), and visual grounding remains challenging. In this work, we evaluate several models on the VRS Bench and NWPU-VHR-10 datasets. The EarthMind model demonstrates strong results in both captioning and VQA. For grounding, we propose multiple pipelines - RemoteSAM-SAM-v1, RemoteSAM-SAM-v2, and DiffuSAM - and ultimately adopt a majority-voting ensemble across EarthMind, RemoteSAM, SAM3, Falcon, RemoteSAM-SAM3-v1, RemoteSAM-SAM3-v2, and DiffuSAM predictions. Our unified, modular pipeline integrates advanced SAM variants with multimodal LLMs to jointly perform captioning, VQA, and grounding. It achieves 82% accuracy on captioning and 83.32% on VQA, with 90.94%, 52.04%, and 92.06% for binary, numeric, and semantic question types respectively. For grounding, it attains 64.94% accuracy. By combining diverse VLMs with our custom RemoteSAM-SAM3 models through ensemble majority voting, the system delivers more accurate and consistent remote-sensing understanding than task-specific approaches.
Chinese Translation
多模态多任务模型在遥感数据集上表现出色。然而，由于这些模型是在异构数据上训练的，并且在不同任务中表现各异，设计一个在图像描述、视觉问答（VQA）和视觉定位中均表现良好的统一模型仍然具有挑战性。在本研究中，我们在VRS Bench和NWPU-VHR-10数据集上评估了几种模型。EarthMind模型在图像描述和VQA方面表现出色。对于视觉定位，我们提出了多个管道 - RemoteSAM-SAM-v1、RemoteSAM-SAM-v2和DiffuSAM，并最终在EarthMind、RemoteSAM、SAM3、Falcon、RemoteSAM-SAM3-v1、RemoteSAM-SAM3-v2和DiffuSAM的预测结果上采用多数投票集成。我们的统一模块化管道将先进的SAM变体与多模态大语言模型（LLMs）集成，以共同执行图像描述、VQA和视觉定位。它在图像描述上达到了82%的准确率，在VQA上达到了83.32%，对于二元、数值和语义问题类型的准确率分别为90.94%、52.04%和92.06%。在视觉定位方面，它达到了64.94%的准确率。通过将多样的视觉语言模型（VLMs）与我们定制的RemoteSAM-SAM3模型结合，并通过集成多数投票，系统提供了比特定任务方法更准确和一致的遥感理解。
cs.CV / 7 / 2609.28757
Small yet Assistive: Spatially-Aware Post-Training for Low Vision
小而助人：面向低视力用户的空间感知后训练
Abstract
An estimated 1 billion people worldwide live with vision impairment, yet current vision-language models (VLMs) produce descriptions too vague for safe navigation by blind and low-vision (BLV) users. Large VLMs can generate high-quality audio-description-compliant narrations but cannot run on mobile devices; small VLMs offer competitive latency but lack spatial detail, directional cues, and hazard awareness for navigational assistance. We present Smol-VL-BLV, a compact VLM for blind and low-vision users that closes this gap using a 500M decoder transformer model and two post-training mechanisms: (1) teacher-student distillation and (2) Group Relative Policy Optimization (GRPO) with a composite BLV reward targeting directional language, metric distances, and hazard detection. Because multi-stage post-training can induce catastrophic forgetting, we add a lightweight finetuning stage after the last stage GRPO finetuning to recover general descriptive quality while preserving BLV-specific spatial grounding. Our best model substantially outperforms the baseline across various benchmarks, including tasks: VQA, BLV captioning, OCR, and latency. Compared with the baseline for relative improvement, it improves the Spatial score gain of 19.3%, and the Social score gain of 14.8%. It also increases OCR-Bench by 101.5%, and raises TextVQA accuracy by 44.2%. These results show that BLV-focused post-training improves both accessibility-specific spatial grounding and general visual-text reasoning. Deployed on a mid-range Android smartphone via Mixed-Precision Quantization, the model remains approx. 450 MB and runs entirely on-device, offline and without network dependency, generating descriptions with latency dependent on host hardware capabilities. Our model, dataset, and code is publicly released at https://smol-vl-blv.github.io/Smol-VL-BLV-website/
Chinese Translation
全球约有10亿人生活在视力障碍中，但当前的视觉语言模型（VLMs）生成的描述对于盲人和低视力（BLV）用户的安全导航来说过于模糊。大型VLM能够生成符合高质量音频描述的叙述，但无法在移动设备上运行；小型VLM提供了竞争性的延迟，但缺乏空间细节、方向提示和导航辅助所需的危险意识。我们提出了Smol-VL-BLV，这是一种为盲人和低视力用户设计的紧凑型VLM，利用500M解码器变换器模型和两种后训练机制来弥补这一差距：（1）教师-学生蒸馏和（2）带有复合BLV奖励的组相对策略优化（GRPO），该奖励针对方向性语言、度量距离和危险检测。由于多阶段后训练可能导致灾难性遗忘，我们在最后阶段的GRPO微调后添加了一个轻量级的微调阶段，以恢复一般描述质量，同时保持BLV特定的空间基础。我们的最佳模型在多个基准测试中显著优于基线，包括任务：VQA、BLV字幕生成、OCR和延迟。与基线相比，相对提升方面，空间得分提高了19.3%，社会得分提高了14.8%。它还将OCR-Bench提高了101.5%，并将TextVQA的准确率提高了44.2%。这些结果表明，针对BLV的后训练改善了可及性特定的空间基础和一般视觉文本推理。该模型通过混合精度量化在中档Android智能手机上部署，大小约为450 MB，并完全在设备上运行，离线且不依赖网络，生成的描述延迟取决于主机硬件能力。我们的模型、数据集和代码已公开发布在 https://smol-vl-blv.github.io/Smol-VL-BLV-website/
cs.CV / 8 / 2609.28796
DrGait: Biomechanically Grounded Visual Reasoning for Interpretable Clinical Gait Analysis
DrGait：基于生物力学的可解释临床步态分析的视觉推理
Abstract
Current automated gait analysis for clinical applications relies on uninterpretable black-box classifiers. Although Vision-Language Models (VLMs) offer strong reasoning capabilities, applying them directly to gait videos often leads to hallucinations, because they struggle to measure subtle geometric deviations from raw visual contexts. To address this, we introduce DrGait, a training-free agentic framework that shifts the VLM's role from a direct visual reasoner to a clinical planner. DrGait decouples semantic reasoning from geometric perception through a structured Triage-Verification-Synthesis (TVS) workflow. Given an input video and a set of basic spatiotemporal metrics, the DrGait agent first performs a heuristic triage to propose diagnostic hypotheses, which are then verified by autonomously calling deterministic biomechanical tools that operate on reconstructed 3D mesh trajectories, segmented 2D pose tracks, and event-centered video evidence. Finally, a closed-loop mechanism recursively updates the agent's reasoning context based on the feedback. By anchoring VLM's reasoning in verifiable geometric and temporal measurements, DrGait reduces hallucinations, achieving competitive diagnostic accuracy while generating transparent and audit-ready clinical reports.
Chinese Translation
当前用于临床应用的自动步态分析依赖于不可解释的黑箱分类器。尽管视觉-语言模型（Vision-Language Models, VLMs）提供了强大的推理能力，但直接将其应用于步态视频往往会导致幻觉，因为它们难以从原始视觉上下文中测量微妙的几何偏差。为了解决这个问题，我们提出了DrGait，一个无需训练的自主框架，将VLM的角色从直接的视觉推理者转变为临床规划者。DrGait通过结构化的分诊-验证-综合（Triage-Verification-Synthesis, TVS）工作流程，将语义推理与几何感知解耦。给定输入视频和一组基本时空指标，DrGait代理首先执行启发式分诊以提出诊断假设，然后通过自主调用在重建的3D网格轨迹、分割的2D姿态轨迹和事件中心的视频证据上运行的确定性生物力学工具进行验证。最后，一个闭环机制根据反馈递归更新代理的推理上下文。通过将VLM的推理锚定在可验证的几何和时间测量上，DrGait减少了幻觉，实现了具有竞争力的诊断准确性，同时生成透明且可审计的临床报告。
cs.CV / 9 / 2609.28811
DeltaWAM: Delta World Action Models for Bimanual Manipulation
DeltaWAM：用于双手操作的增量世界动作模型
Abstract
World-action models (WAMs) transfer visual and motion priors from pretrained video generators to robot control by jointly modeling visual dynamics and actions. Existing WAMs, however, predict dense future frames during training, repeatedly modeling largely unchanged content and coupling action-conditioned dynamics to nuisance appearance variations. At inference, processing each complete observation with the heavy video expert bottlenecks few-step action generation. Accordingly, we propose DeltaWAM, which jointly predicts visual deltas and actions using dense-anchor, sparse-delta, and action streams, with three architectures that differ in representation and computation sharing. We further develop Streaming Delta Memory (SDM), which updates cached anchor context with compact observed deltas, reducing heavy video-expert processing. On RoboTwin, DeltaWAM with SDM improves average success over Fast-WAM from 81.3% to 85.4% in the clean setting and from 75.8% to 83.9% under visual randomization. The three architectures reduce training FLOPs by 17.78-23.77%, while SDM reduces one-step inference latency and FLOPs by 36.57% and 31.55%, respectively; real-world evaluations further show the highest overall success rate and normalized progress among the evaluated policies. Code: https://github.com/AIGeeksGroup/DeltaWAM. Website: https://aigeeksgroup.github.io/DeltaWAM.
Chinese Translation
世界动作模型（WAMs）通过联合建模视觉动态和动作，将预训练视频生成器的视觉和运动先验转移到机器人控制中。然而，现有的WAMs在训练期间预测密集的未来帧，反复建模几乎不变的内容，并将动作条件动态与干扰外观变化耦合。在推理时，处理每个完整观察的重型视频专家限制了少步动作生成。因此，我们提出了DeltaWAM，它通过密集锚点、稀疏增量和动作流共同预测视觉增量和动作，具有在表示和计算共享上有所不同的三种架构。我们进一步开发了流式增量记忆（SDM），它通过紧凑的观察增量更新缓存的锚点上下文，从而减少重型视频专家的处理。在RoboTwin上，结合SDM的DeltaWAM在干净设置下将平均成功率从81.3%提高到85.4%，在视觉随机化下从75.8%提高到83.9%。这三种架构将训练的FLOPs减少了17.78%-23.77%，而SDM分别减少了一步推理延迟和FLOPs的36.57%和31.55%；现实世界评估进一步显示在评估的策略中具有最高的整体成功率和标准化进展。代码：https://github.com/AIGeeksGroup/DeltaWAM。网站：https://aigeeksgroup.github.io/DeltaWAM。
cs.CV / 10 / 2609.28813
CinematicVQA: Benchmarking Film-Grammar Reasoning in Large Vision-Language Models
CinematicVQA：大型视觉语言模型中电影语法推理的基准测试
Abstract
Cinematography, the craft of visual storytelling through framing, lighting, and camera operation, fundamentally shapes how audiences perceive and emotionally engage with video content. While Large Vision Language Models (LVLMs) have made remarkable progress in video question answering, existing benchmarks primarily focus on identifying low-level techniques rather than understanding their storytelling impact. To address this, we introduce CinematicVQA, the first-of-its-kind benchmark for cinematic video understanding that goes beyond technique recognition to evaluate film-grammar reasoning, utilizing our introduced Cinematic Scene Graph (CSG), a structured representation that links filming techniques to their perceptual effects and narrative functions. Through comprehensive evaluation of state-of-the-art LVLMs, we reveal a striking semantic gap: models consistently perform higher on describing visual presentations than on identifying the underlying techniques. Surprisingly, Chain-of-Thought prompting fails to provide consistent gains and degrades performance for most models, suggesting that current LVLMs lack sufficient cinematic domain knowledge to benefit from step-by-step reasoning. Fine-tuning on \textsc{CinematicVQA-train} yields consistent improvements, particularly for narrative function and multi-hop reasoning. Overall, \textsc{CinematicVQA} serves both as a rigorous benchmark for cinematic evaluation in LVLMs and as a practical dataset for training more film-aware video models.
Chinese Translation
摄影术是通过构图、照明和摄像操作进行视觉叙事的艺术，根本上塑造了观众对视频内容的感知和情感参与。尽管大型视觉语言模型（LVLMs）在视频问答方面取得了显著进展，但现有基准主要集中在识别低级技术，而非理解其叙事影响。为了解决这一问题，我们推出了CinematicVQA，这是首个超越技术识别的电影视频理解基准，旨在评估电影语法推理，利用我们引入的电影场景图（Cinematic Scene Graph, CSG），这是一个将拍摄技术与其感知效果和叙事功能联系起来的结构化表示。通过对最先进的LVLMs进行全面评估，我们揭示了一个显著的语义差距：模型在描述视觉呈现方面的表现始终高于识别潜在技术的表现。令人惊讶的是，链式思维提示未能提供一致的提升，并且对大多数模型的性能造成了下降，这表明当前的LVLMs缺乏足够的电影领域知识，无法从逐步推理中受益。在 extsc{CinematicVQA-train}上的微调带来了持续的改善，特别是在叙事功能和多跳推理方面。总体而言， extsc{CinematicVQA}既作为LVLMs中电影评估的严格基准，也作为训练更具电影意识的视频模型的实用数据集。
cs.CV / 11 / 2609.28836
M$^2$PFN: End-to-End Disentangled Alignment for Generalizable Multimodal In-Context Learning in Alzheimer's Disease
M$^2$PFN：用于阿尔茨海默病的可泛化多模态上下文学习的端到端解耦对齐
Abstract
While various multimodal methods combining imaging and tabular data for Alzheimer's disease (AD) diagnosis were proposed, they are often limited in generalization across cohorts. In-context learning (ICL) has demonstrated excellent generalization performances and high flexibility in foundational tabular models such as TabPFN. To extend TabPFN's ICL to multimodal AD analysis, the main obstacle is that TabPFN is meta-trained on synthetic tabular priors that do not naturally match the statistical structure of image-derived features. We propose M$^2$PFN, an end-to-end framework that turns this tabular foundation model into a multimodal AD predictor. M$^2$PFN (i) performs differentiable inference through TabPFN's transformer, back-propagating task gradients into 3D-MRI and tabular encoders; (ii) aligns the two modalities into a shared subspace, via disentanglement and a contrastive objective, matched to the ICL engine's prior; and (iii) folds in a frozen tabular-only prediction through a learnable gated shortcut. Because the ICL engine stays frozen, its in-context mechanism is preserved for test-time generalization, while end-to-end training shapes the encoders into features it can exploit. On ADNI ($n=2240$, three-class CN/MCI/AD), M$^2$PFN attains $65.55\%$ macro-F1 and $82.21\%$ macro-AUC, surpassing a comprehensive set of unimodal and multimodal baselines. By swapping only the head for a TabPFN regressor, the same architecture regresses baseline MMSE on a $1250$-subject sub-cohort to test MAE $1.743$, outperforming every multimodal baseline. On two external cohorts (OASIS-3 and SCAN) with no retraining, M$^2$PFN achieves the best AUC and the lowest MMSE MAE across all baselines, and transfers even when the cognitive instrument changes.
Chinese Translation
尽管已经提出了多种结合影像和表格数据的阿尔茨海默病（AD）诊断多模态方法，但它们在不同人群之间的泛化能力往往受到限制。上下文学习（ICL）在基础表格模型如TabPFN中展现了出色的泛化性能和高度的灵活性。将TabPFN的ICL扩展到多模态AD分析的主要障碍在于，TabPFN是在与图像派生特征的统计结构不自然匹配的合成表格先验上进行元训练的。我们提出了M$^2$PFN，一个端到端框架，将这一表格基础模型转变为多模态AD预测器。M$^2$PFN（i）通过TabPFN的变换器执行可微推理，将任务梯度反向传播到3D-MRI和表格编码器；（ii）通过解耦和对比目标将两种模态对齐到一个共享子空间，与ICL引擎的先验相匹配；（iii）通过可学习的门控快捷方式引入一个冻结的仅表格预测。由于ICL引擎保持冻结，其上下文机制在测试时得以保留，而端到端训练则将编码器塑造为可以利用的特征。在ADNI（n=2240，三类CN/MCI/AD）上，M$^2$PFN达到了65.55%的宏观F1和82.21%的宏观AUC，超过了一系列全面的单模态和多模态基线。通过仅替换头部为TabPFN回归器，相同的架构在一个1250名受试者的子群体上回归基线MMSE，测试MAE为1.743，超越了所有多模态基线。在两个外部人群（OASIS-3和SCAN）上，无需重新训练，M$^2$PFN在所有基线中实现了最佳AUC和最低MMSE MAE，并且即使在认知工具发生变化时也能实现迁移。
cs.CV / 12 / 2609.28851
Looks the Same, Answers Differently: Flip-Direction Steering for Robust Vision-Language Reasoning
看起来相同，回答却不同：用于稳健视觉-语言推理的翻转方向引导
Abstract
Vision-language models (VLMs) achieve strong visual reasoning performance, yet subtle changes from routine image capture and processing can alter their reasoning trajectories even when images appear nearly identical. In long-horizon generation, the resulting activation shifts may accumulate across decoding steps, progressively altering reasoning tokens and ultimately changing the final answer, a phenomenon referred to as answer flips. To address this instability, we propose FlipDir (Flip-Direction Steering), a training-free inference-time method that estimates a low-rank flip-inducing activation subspace from contrastive pairs of original and answer-flipping inputs and selectively steers hidden states during decoding. A margin-based gate limits subspace attenuation to uncertain decoding steps, recovering original predictions while preserving stable ones. To evaluate robustness beyond accuracy or consistency on fixed test sets, we introduce VisFlip, a benchmark framework that constructs evaluation groups for a target model and visual variation setting to separately assess recovery of original predictions and preservation of stable ones. VisFlip spans nine dataset-variation combinations across scientific reasoning, robot-scene understanding, and medical VQA, covering subtle visual variations common in each domain. Experiments across 18 settings demonstrate that FlipDir consistently outperforms existing methods on the combined recovery and preservation metric. We will make our code publicly available.
Chinese Translation
视觉-语言模型（VLMs）在视觉推理性能上表现出色，但来自常规图像捕捉和处理的细微变化可能会改变它们的推理轨迹，即使图像看起来几乎相同。在长时间生成过程中，结果激活的变化可能会在解码步骤中累积，逐渐改变推理标记，最终改变最终答案，这一现象称为答案翻转。为了解决这种不稳定性，我们提出了FlipDir（翻转方向引导），这是一种无训练的推理时方法，通过对比原始输入和答案翻转输入的对，估计低秩翻转诱导激活子空间，并在解码过程中选择性地引导隐藏状态。基于边际的门控机制限制子空间衰减到不确定的解码步骤，恢复原始预测，同时保留稳定的预测。为了在准确性或一致性之外评估稳健性，我们引入了VisFlip，一个基准框架，构建目标模型和视觉变化设置的评估组，以分别评估原始预测的恢复和稳定预测的保留。VisFlip涵盖了科学推理、机器人场景理解和医学视觉问答（VQA）中的九个数据集变化组合，涵盖了每个领域中常见的细微视觉变化。18个设置的实验表明，FlipDir在综合恢复和保留指标上始终优于现有方法。我们将公开我们的代码。
cs.CV / 13 / 2609.28857
MEVL-STP: Multi-Encoder and Vision Language Model for Arbitrarily Shaped Scene Text Spotting
MEVL-STP：用于任意形状场景文本检测的多编码器和视觉语言模型
Abstract
Scene text spotting remains challenging for arbitrarily shaped text instances such as curved signs and dense multi-oriented characters in natural images, where tightly coupled architectures propagate localization errors directly into recognition failures. We present a two-stage pipeline that combines multi-encoder segmentation with vision-language model recognition to address this problem. In the detection stage, six frozen vision encoders (CLIP, DINOv2, SigLIP, EVA-CLIP, SAM, and ConvNeXt) extract complementary features spanning semantic, spatial, and texture spectra, which are fused through a trainable hierarchical Feature Pyramid Network with channel attention and decoded via a deep-supervision Progressive Scale Expansion network to generate precise instance-level text masks. By keeping the encoders frozen, their independently learned feature spaces remain orthogonal during fusion, preventing the feature homogenization that degrades boundary precision in single-backbone detectors. The detection stage produces tight polygon masks that conform to the actual shape of curved and arbitrarily oriented text, rather than axis-aligned rectangles that inevitably include background content. In the recognition stage, these polygon-masked crops isolate the target text from surrounding clutter, allowing a Qwen3-VL-8B-Instruct model, fine-tuned via Low-Rank Adaptation on polygon-cropped scene text, to focus purely on reading the text without interference from neighbouring words or background noise. Without any synthetic pretraining data, our method achieves 91.99% detection F-measure and 85.86% end-to-end H-mean on CTW1500, setting a new state of the art and achieving strong performance on Total-Text and ICDAR 2015 without any synthetic training data. Code is available at https://github.com/doubleblind-afk/MEVL-STP
Chinese Translation
场景文本检测对于任意形状的文本实例（如弯曲的标志和密集的多方向字符）在自然图像中仍然具有挑战性，其中紧密耦合的架构将定位错误直接传播到识别失败。我们提出了一种两阶段的管道，将多编码器分割与视觉语言模型识别相结合，以解决这一问题。在检测阶段，六个冻结的视觉编码器（CLIP、DINOv2、SigLIP、EVA-CLIP、SAM和ConvNeXt）提取涵盖语义、空间和纹理光谱的互补特征，这些特征通过具有通道注意力的可训练层次特征金字塔网络进行融合，并通过深度监督的渐进尺度扩展网络解码，以生成精确的实例级文本掩码。通过保持编码器冻结，它们独立学习的特征空间在融合过程中保持正交，防止了特征同质化，从而降低了单一骨干检测器的边界精度。检测阶段生成的紧凑多边形掩码符合弯曲和任意方向文本的实际形状，而不是不可避免地包含背景内容的轴对齐矩形。在识别阶段，这些多边形掩码裁剪将目标文本从周围杂乱中隔离，使得经过低秩适应（Low-Rank Adaptation）微调的Qwen3-VL-8B-Instruct模型能够专注于阅读文本，而不受邻近单词或背景噪声的干扰。在没有任何合成预训练数据的情况下，我们的方法在CTW1500上达到了91.99%的检测F-measure和85.86%的端到端H-mean，创造了新的最先进水平，并在Total-Text和ICDAR 2015上取得了强劲表现，且没有使用任何合成训练数据。代码可在https://github.com/doubleblind-afk/MEVL-STP获取。
cs.CV / 14 / 2609.28860
Multimodal Routing and Region Refinement for Language-Guided Medical Image Segmentation
基于语言引导的医学图像分割的多模态路由与区域细化
Abstract
Textual descriptions can reduce ambiguity in medical image segmentation by specifying the finding and location to be delineated. Existing text-guided methods mainly improve where image and language features interact but generally retain a single learned update pathway across all image-text pairs. We propose MRSeg, a parameter-efficient framework that uses each image-text pair to route the adaptation of visual and textual features before dense prediction. Frozen ConvNeXt-Tiny and PubMedBERT encoders provide multiscale visual features and clinical text tokens. A joint router uses the deepest visual feature and pooled text to predict a sparse mixture over low-rank adapter bases. The resulting route is shared across separate adapter banks for two visual scales and text, coordinating their adaptation while keeping the feature-specific parameters separate. Region Bridge uses text-derived queries to aggregate dense visual tokens into latent regions, refines these regions through self-attention and text cross-attention, and redistributes the refined information back to the feature maps. Finally, a multiscale decoder combines refined semantic features with shallow image evidence. On QaTa-COV19 and MosMedData+, MRSeg achieves 90.90/83.32 and 81.53/68.82 Dice/mIoU, respectively, with 7.11M trainable parameters and 7.60 GFLOPs. Code: https://github.com/maklachur/MRSeg.
Chinese Translation
文本描述可以通过指定要描绘的发现和位置来减少医学图像分割中的模糊性。现有的文本引导方法主要改善图像和语言特征的交互，但通常在所有图像-文本对中保留单一的学习更新路径。我们提出了MRSeg，这是一种参数高效的框架，利用每个图像-文本对在密集预测之前路由视觉和文本特征的适应。冻结的ConvNeXt-Tiny和PubMedBERT编码器提供多尺度的视觉特征和临床文本标记。一个联合路由器使用最深的视觉特征和池化文本来预测低秩适配器基础上的稀疏混合。生成的路由在两个视觉尺度和文本的不同适配器库之间共享，协调它们的适应，同时保持特征特定参数的独立性。区域桥使用文本派生的查询将稠密视觉标记聚合成潜在区域，通过自注意力和文本交叉注意力细化这些区域，并将细化的信息重新分配回特征图。最后，一个多尺度解码器将细化的语义特征与浅层图像证据相结合。在QaTa-COV19和MosMedData+上，MRSeg分别实现了90.90/83.32和81.53/68.82的Dice/mIoU，具有7.11M可训练参数和7.60 GFLOPs。代码： https://github.com/maklachur/MRSeg。
cs.CV / 15 / 2609.28865
Direction-Scale Decomposition in Action Representation: Rethinking What to Tokenize for Vision-Language-Action Models
行动表示中的方向-尺度分解：重新思考视觉-语言-行动模型中的标记化内容
Abstract
Action representation plays a central role in discrete-token vision-language-action (VLA) learning but remains underexamined. Under conventional pose-increment representations, action tokens are sensitive to execution speed and dataset-specific normalization, potentially obscuring geometric structure shared across demonstrations and datasets. We introduce Direction-Scale Decomposition (DSD), an action representation that decomposes translation and rotation increments into direction and scale components before tokenization. DSD isolates motion direction while retaining magnitudes in separate scale channels. We evaluate DSD with uniform binning (BIN) and BEAST, a B-spline-based tokenizer, in simulation and real-world manipulation under both single-dataset and mixed-dataset training. On LIBERO, DSD improves average success rates with both tokenizers. On SimplerEnv, DSD-BIN outperforms BIN by 10.3 percentage points in overall success rate under mixed-dataset training. Real-robot experiments further show gains both with and without robotics pretraining. These results support DSD as an effective action representation for discrete-token VLA models and suggest its potential to mitigate performance degradation when training on large and diverse dataset mixtures. Our project page with additional resources is available at https://vla-dsd.github.io/
Chinese Translation
行动表示在离散标记视觉-语言-行动（VLA）学习中扮演着核心角色，但仍然未得到充分研究。在传统的姿态增量表示下，行动标记对执行速度和数据集特定的归一化敏感，可能会掩盖跨示范和数据集共享的几何结构。我们引入了方向-尺度分解（Direction-Scale Decomposition, DSD），一种在标记化之前将平移和旋转增量分解为方向和尺度分量的行动表示。DSD在保留幅度的同时，隔离运动方向于单独的尺度通道。我们在单一数据集和混合数据集训练下，使用均匀分箱（BIN）和基于B样条的标记器BEAST评估DSD在模拟和真实世界操作中的表现。在LIBERO上，DSD提高了两种标记器的平均成功率。在SimplerEnv上，DSD-BIN在混合数据集训练下的整体成功率比BIN高出10.3个百分点。真实机器人实验进一步显示，无论是否经过机器人预训练，DSD均表现出提升。这些结果支持DSD作为离散标记VLA模型的有效行动表示，并表明其在对大规模和多样化数据集混合进行训练时减轻性能下降的潜力。我们的项目页面及其他资源可在https://vla-dsd.github.io/获取。
cs.CV / 16 / 2609.28923
ViRDM: Taming Representation Distribution Matching for Few-Step Causal Video Generation
ViRDM：驯服少步因果视频生成的表示分布匹配
Abstract
Few-step autoregressive (AR) video diffusion enables low-latency streaming generation, but existing post-training methods predominantly rely on Distribution Matching Distillation (DMD), requiring both a large pretrained teacher and an online critic to estimate distributional discrepancies through diffusion scores. In this work, we ask whether this resource-intensive teacher--critic stack can be eliminated by post-training only the generator against a precomputed target distribution. Drawing inspiration from representation distribution matching (RDM) for one-step image generation, we systematically study its transfer to few-step causal video generation and identify three key barriers: a memory-intractable gradient path, a distinct video optimization regime, and representation distributions that underconstrain temporal dynamics. We introduce ViRDM, a teacher- and critic-free video post-training recipe that addresses these barriers sequentially. By coupling RDM with stochastically truncated clean-exit supervision, a lightweight VAE decoder, and staged vector--Jacobian products, ViRDM makes representation distribution matching memory-feasible for multi-step causal video rollouts. We further establish effective generated-population and initialization regimes for video RDM, and introduce lightweight dynamics regularization to compensate for the underconstrained temporal dynamics. ViRDM turns three-network distillation into generator-only post-training, reducing GPU memory use and training time while improving video quality. With only 20 generator updates, the recipe reaches 84.87 on the official VBench evaluation, outperforming the previous best few-step causal baseline by 0.36, while requiring 16 A100 GPU-hours. We additionally report exploratory results demonstrating the potential of the same recipe for lower causal sampling budget and for one-, two-, and four-step bidirectional generation.
Chinese Translation
少步自回归（AR）视频扩散实现了低延迟流式生成，但现有的后训练方法主要依赖于分布匹配蒸馏（DMD），需要一个大型的预训练教师和一个在线评论者通过扩散分数来估计分布差异。在本研究中，我们探讨是否可以通过仅针对预计算目标分布对生成器进行后训练，从而消除这种资源密集型的教师-评论者架构。我们从一步图像生成的表示分布匹配（RDM）中获得灵感，系统地研究其向少步因果视频生成的转移，并识别出三个关键障碍：不可处理的记忆梯度路径、独特的视频优化机制，以及在时间动态上约束不足的表示分布。我们提出了ViRDM，一种无教师和评论者的视频后训练方案，逐步解决这些障碍。通过将RDM与随机截断的干净退出监督、轻量级VAE解码器和分阶段的向量-雅可比乘积相结合，ViRDM使得表示分布匹配在多步因果视频生成中变得记忆可行。我们进一步建立了视频RDM的有效生成种群和初始化机制，并引入轻量级动态正则化以补偿时间动态的约束不足。ViRDM将三网络蒸馏转变为仅生成器的后训练，减少了GPU内存使用和训练时间，同时提高了视频质量。在仅进行20次生成器更新的情况下，该方案在官方VBench评估中达到了84.87，超越了之前最佳的少步因果基线0.36，同时仅需16个A100 GPU小时。我们还报告了探索性结果，展示了相同方案在较低因果采样预算以及一步、两步和四步双向生成中的潜力。
cs.CV / 17 / 2609.28930
PlenoCI: Plenoptic CharacterIstics for View Dependence Aware Change Classification
PlenoCI：视依赖变化分类的全光特征
Abstract
Radiance field representations such as 3D Gaussian Splatting (3DGS) natively encode complex visual phenomena such as occlusions and view dependence, but they are inherently underconstrained. Independently optimized reconstructions converge to different primitive configurations, even in unchanged regions. We introduce Plenoptic CharacterIstics (PlenoCI), a novel feature built from the plenoptic field these representations approximate. PlenoCI directly captures rich visual behaviors while ignoring Lambertian textures. By deriving closed-form analytic plenoptic derivatives from a 3DGS representation, we efficiently detect these 5D structures. Our approach is robust to underconstrained representations by construction, reporting two orders of magnitude fewer false positives between independent reconstructions of unchanged scenes than concurrent work. We demonstrate PlenoCI's utility on change classification. First, we detect changes with an instance-aware 3DGS pipeline, achieving state-of-the-art results on CL-Splats with a 25.7% mIoU gain over the strongest competitor, while remaining competitive on the more challenging PASLCD benchmark. Leveraging PlenoCI, we classify changes as geometric or appearance-based with a balanced accuracy of 0.735, comparable to the best performing baseline. We believe plenoptic derivatives and PlenoCI open new directions for view dependence aware understanding in visually complex environments. Code and data are available at https://js0n-lai.github.io/plenoci.
Chinese Translation
辐射场表示（如3D高斯喷溅（3DGS））本质上编码了复杂的视觉现象，如遮挡和视依赖性，但它们在本质上是欠约束的。独立优化的重建会收敛到不同的原始配置，即使在未改变的区域。我们引入了全光特征（Plenoptic CharacterIstics，PlenoCI），这是一个基于这些表示所近似的全光场构建的新特征。PlenoCI 直接捕捉丰富的视觉行为，同时忽略了朗伯纹理。通过从3DGS表示中推导出封闭形式的解析全光导数，我们有效地检测这些5D结构。我们的方案在构造上对欠约束表示具有鲁棒性，报告的独立重建未改变场景之间的假阳性数量比当前工作少两个数量级。我们展示了PlenoCI在变化分类中的实用性。首先，我们利用实例感知的3DGS管道检测变化，在CL-Splats上实现了最先进的结果，相较于最强竞争者提高了25.7%的mIoU，同时在更具挑战性的PASLCD基准上保持竞争力。借助PlenoCI，我们将变化分类为几何变化或外观变化，平衡准确率为0.735， comparable于表现最好的基线。我们相信全光导数和PlenoCI为在视觉复杂环境中进行视依赖理解开辟了新的方向。代码和数据可在 https://js0n-lai.github.io/plenoci 获取。
cs.CV / 18 / 2609.28931
HelloWorld: Towards Practical Applications of Generative Driving World Models
HelloWorld：朝着生成驾驶世界模型的实际应用迈进
Lu, Fan, Wang, Hanshi, Wang, Zijing, Feng, Quan, Wang, Zhi, Chen, Shijie, Zeng, Xianming, Zhang, Yujian, Wang, Jiazhe, Zha, Xin, Wang, Kai, Zhao, Zhijie, Zhu, Lin, Yang, Tianyi, Xu, Yucheng, Ji, Tao, Zhang, Haodong, Zhang, Zhipeng, Peng, Peixi, Chen, Guang, Liu, Xingliang, Yang, Lei, Xu, Jianyun
Abstract
Driving world models provide a promising route toward scalable counterfactual data generation and interactive simulation beyond recorded driving logs. Realizing this potential requires a system that can generalize across diverse scenes, respond faithfully to prescribed controls, generate coherent multi-sensor observations, and operate efficiently under repeated inference. We present \textbf{HelloWorld}, a 2B driving world model system designed around these requirements. HelloWorld progressively specializes broad visual and motion priors from heterogeneous video data into controllable driving generation using ego pose, HD maps, and 3D boxes. A block-causal generation interface, together with adaptation to self-generated context, aligns the model with sequential simulation. The system further supports synchronized seven-camera RGB generation and conditional LiDAR synthesis, and is distilled toward few-step inference for efficient deployment. Experiments evaluate visual quality, control fidelity, cross-view consistency, robustness under repeated generation, inference efficiency, and LiDAR synthesis. Together, HelloWorld provides a unified framework for scalable driving data generation and interactive simulation.
Chinese Translation
驾驶世界模型为可扩展的反事实数据生成和超越记录驾驶日志的互动仿真提供了一个有前景的途径。实现这一潜力需要一个能够在多样场景中进行泛化、忠实响应指定控制、生成一致的多传感器观测，并在重复推理下高效运行的系统。我们提出了 extbf{HelloWorld}，这是一个围绕这些要求设计的2B驾驶世界模型系统。HelloWorld逐步将来自异构视频数据的广泛视觉和运动先验专业化为可控的驾驶生成，利用自我姿态、高清地图和三维框。一个块因果生成接口，以及对自生成上下文的适应，使模型与顺序仿真对齐。该系统进一步支持同步的七摄像头RGB生成和条件激光雷达（LiDAR）合成，并朝着少步推理进行提炼，以实现高效部署。实验评估了视觉质量、控制保真度、跨视图一致性、在重复生成下的鲁棒性、推理效率和激光雷达合成。总之，HelloWorld提供了一个统一的框架，用于可扩展的驾驶数据生成和互动仿真。
cs.CV / 19 / 2609.28949
Exploiting Target Knowledge from MLLMs for Robust Few-Shot Segmentation
利用多模态大型语言模型的目标知识进行稳健的少样本分割
Abstract
Few-shot segmentation (FSS) aims to segment unseen object categories with a few (e.g., one or five) labeled examples, enabling efficient adaptation to novel classes. Conventional models typically rely on appearance-based visual matching between support and query images for segmentation. While straightforward, these methods often struggle to handle significant appearance discrepancies and occlusions in the query image due to insufficient target knowledge. To mitigate this, we introduce a novel framework that mines target knowledge using the strong reasoning capacity of Multimodal Large Language Models (MLLMs) and employs it to enhance FSS. Specifically, building on SAM 2, our method, named MK-FSS, exploits two forms of complementary knowledge derived from a query image by an MLLM for FSS, including spatial knowledge, which provides a spatial prior indicating the potential target location, and semantic knowledge, which describes the target using text. The spatial knowledge is first encoded into a memory representation, and then resulting memory is integrated with the support-guided memory feature from query image through a carefully designed dual-memory debate-fusion (DMDF) module, yielding a more robust target memory feature. In parallel, the semantic knowledge is encoded into the textual feature, which is fused with multi-scale query features via a progressive cross-modal prompt generator (PCPG), producing a target-aware multimodal prompt for segmentation. Working together, the dual-memory feature and the multimodal prompt provide a comprehensive representation of the target, enabling more robust segmentation. In our extensive experiments, MK-FSS shows promising results and largely surpasses existing methods. Code will be released.
Chinese Translation
少样本分割（FSS）旨在通过少量（例如，一个或五个）标记示例对未见物体类别进行分割，从而实现对新类别的高效适应。传统模型通常依赖于支持图像和查询图像之间基于外观的视觉匹配进行分割。尽管这种方法简单直接，但由于目标知识不足，这些方法在处理查询图像中的显著外观差异和遮挡时常常面临困难。为了解决这个问题，我们提出了一种新颖的框架，利用多模态大型语言模型（MLLM）的强大推理能力挖掘目标知识，并将其应用于增强FSS。具体而言，基于SAM 2，我们的方法命名为MK-FSS，利用来自查询图像的两种形式的互补知识进行FSS，包括空间知识，它提供了指示潜在目标位置的空间先验，以及语义知识，它通过文本描述目标。空间知识首先被编码为记忆表示，然后通过精心设计的双记忆辩论融合（DMDF）模块，将得到的记忆与来自查询图像的支持引导记忆特征集成，从而生成更稳健的目标记忆特征。同时，语义知识被编码为文本特征，并通过渐进式跨模态提示生成器（PCPG）与多尺度查询特征融合，产生用于分割的目标感知多模态提示。双记忆特征和多模态提示协同工作，提供了目标的全面表示，从而实现更稳健的分割。在我们广泛的实验中，MK-FSS显示出良好的结果，远超现有方法。代码将会发布。
cs.CV / 20 / 2609.28956
MoVISA: Multi-Token Reasoning for Video Object Segmentation
MoVISA：视频目标分割的多标记推理
Abstract
Recent advances in video object segmentation with Multimodal Large Language Model (MLLM) reasoning have demonstrated the effectiveness of using a single textual token, such as SEG, to predict segmentation masks across images and videos. However, we observe that this single-token strategy lacks the granularity required to precisely localize multiple objects across time in video segmentation tasks. To address this limitation, we develop Multi-Token Reasoning for Video Object Segmentation, or MoVISA. MoVISA uses multiple segmentation tokens, such as SEG0 and SEG1, to represent an object across different frames. This design enables more fine-grained alignment between language prompts and spatio-temporal mask predictions, improving both performance and interpretability. On the challenging MeViS, DAVIS17, ReVOS, and Ref-Youtube-VOS benchmarks, our model achieves a 13.2 percent J and F improvement on MeViS and an 8.4 percent J and F improvement on ReVOS. Code and models will be released.
Chinese Translation
最近在多模态大语言模型（MLLM）推理下的视频目标分割方面的进展，展示了使用单一文本标记（如 SEG）来预测图像和视频中的分割掩膜的有效性。然而，我们观察到这种单标记策略缺乏在视频分割任务中精确定位多个对象所需的细粒度。为了解决这一局限性，我们开发了视频目标分割的多标记推理方法，即 MoVISA。MoVISA 使用多个分割标记（如 SEG0 和 SEG1）来表示不同帧中的对象。这一设计使语言提示与时空掩膜预测之间的对齐更加细致，从而提高了性能和可解释性。在具有挑战性的 MeViS、DAVIS17、ReVOS 和 Ref-Youtube-VOS 基准测试中，我们的模型在 MeViS 上实现了 13.2% 的 J 和 F 改进，在 ReVOS 上实现了 8.4% 的 J 和 F 改进。代码和模型将会发布。
cs.CV / 21 / 2609.28967
Passive LWIR Hyperspectral Ranging via Transmittance Extraction and Distance Alignment
通过透射率提取和距离对齐实现被动长波红外高光谱测距
Abstract
Passive long-wave infrared (LWIR) hyperspectral ranging enables distance estimation in low-light and nighttime scenes by exploiting atmospheric absorption features in thermal radiance received through the atmosphere.Joint estimation of temperature, emissivity, and distance is computationally expensive. Reference-range joint inversion also uses a distance-invariant effective attenuation coefficient, which can bias range estimates.We introduce transmittance extraction and distance alignment (TEDA), which decouples range estimation from temperature--emissivity inversion. In the first stage, a baseline estimator with a data-fidelity term invariant to the known absorption direction yields two closed-form smoothing branches for the slowly varying thermal continuum. An observation-derived gate combines the branches, and subtracting the blended baseline in the log domain recovers atmospheric transmittance. The second stage estimates range by matching the recovered transmittance to sensor-domain transmittance models recomputed for each candidate distance. Monte Carlo simulations show that TEDA effectively reduces the ranging bias caused by the distance-invariant attenuation coefficient approximation. In a measured scene, TEDA's mean range estimates are closer to the LiDAR medians than those of reference-range joint inversion in both evaluated patches. TEDA processes a complete $256\times256$ region of interest in 8.19~s versus 159.47~s for reference-range joint inversion, an approximately 20-fold speedup.
Chinese Translation
被动长波红外（LWIR）高光谱测距通过利用在大气中接收到的热辐射中的大气吸收特征，使得在低光照和夜间场景中进行距离估计成为可能。温度、发射率和距离的联合估计计算开销较大。参考范围联合反演还使用了距离不变的有效衰减系数，这可能导致范围估计的偏差。我们提出了透射率提取和距离对齐（TEDA）方法，该方法将范围估计与温度-发射率反演解耦。在第一阶段，基于数据保真度项的不变于已知吸收方向的基线估计器为缓慢变化的热连续体提供了两个封闭形式的平滑分支。一个基于观测的门结合了这些分支，通过在对数域中减去混合基线来恢复大气透射率。第二阶段通过将恢复的透射率与为每个候选距离重新计算的传感器域透射率模型进行匹配来估计范围。蒙特卡洛仿真表明，TEDA有效减少了由距离不变的衰减系数近似引起的测距偏差。在一个测量场景中，TEDA的平均范围估计在两个评估区域中都更接近LiDAR中位数，而不是参考范围联合反演的结果。TEDA处理一个完整的$256 imes256$感兴趣区域仅需8.19秒，而参考范围联合反演则需159.47秒，速度提升约20倍。
cs.CV / 22 / 2609.28991
Beneath the Scores: Rethinking Hallucination Evaluation for Video Understanding Models
评分背后：重新思考视频理解模型的幻觉评估
Abstract
Video understanding is increasingly performed by multi-stage LLM agents that separate temporal grounding, visual observation, and reasoning. Yet these stages are typically evaluated on different benchmarks and distributions, making it difficult to determine where hallucinations originate. We first organize existing benchmarks around these stages and show that their scores provide inconsistent diagnostic signals: stronger stage-level performance does not reliably imply lower downstream hallucination, and even benchmarks targeting the same capability can disagree. We therefore introduce a causal stage-intervention protocol that overwrites individual stages while holding the downstream task fixed. Across 60,008 runs on three video-agent architectures, we find that grounding is the dominant source of downstream error, with roughly four times the causal impact of corrupting visual observations. Successful grounding depends primarily on locating the correct region rather than precise temporal overlap, explaining why standard mIoU metrics poorly predict downstream reliability. We further find that incorrect evidence is substantially more harmful than missing evidence. Finally, auditing existing benchmarks against these interventions reveals that their scores do not reliably predict causal cascade sensitivity and can fail under distribution shift. These results motivate intervention-based, stage-aware evaluation for trustworthy video agents.
Chinese Translation
视频理解越来越依赖于多阶段的LLM代理，这些代理将时间定位、视觉观察和推理分开。然而，这些阶段通常在不同的基准和分布上进行评估，这使得确定幻觉的来源变得困难。我们首先围绕这些阶段组织现有的基准，并显示它们的评分提供了不一致的诊断信号：更强的阶段性表现并不可靠地意味着下游幻觉的减少，即使是针对同一能力的基准也可能存在分歧。因此，我们引入了一种因果阶段干预协议，该协议在保持下游任务不变的情况下覆盖个别阶段。在对三种视频代理架构进行的60,008次实验中，我们发现，时间定位是下游错误的主要来源，其因果影响大约是破坏视觉观察的四倍。成功的时间定位主要依赖于定位正确的区域，而不是精确的时间重叠，这解释了为什么标准的mIoU指标无法有效预测下游的可靠性。我们进一步发现，不正确的证据比缺失的证据更具危害性。最后，对现有基准进行审计以应对这些干预显示，它们的评分并不能可靠地预测因果级联敏感性，并且在分布转移下可能失效。这些结果促使我们基于干预和阶段意识的评估方法，以实现可信赖的视频代理。
cs.CV / 23 / 2609.28997
Only What Was Seen: Observation-Gram Compaction of View-Dependent Appearance in 3D Gaussian Splatting
仅限于所见：基于观察Gram的3D高斯点云视依赖外观压缩
Abstract
Most of the memory of a 3D Gaussian Splatting model holds spherical-harmonic colour coefficients, yet each Gaussian is seen only from the narrow cone of directions of the training cameras. We turn this into a distortion metric that other compressors can adopt: a per-Gaussian observation Gram matrix, accumulated from viewing directions and blending weights, is the exact first-order map from coefficient changes to squared image error and needs only the model and the camera poses. Under it, degree reduction becomes a closed-form projection that generalises truncation, degree allocation a Lagrangian rate-distortion problem, and vector quantisation the matrix-weighted Lloyd algorithm, of which Compressed3D's quantiser is the scalar case. Swapped into Compressed3D with everything else unchanged, the metric raises PSNR by +0.49 dB before fine-tuning, with SSIM and LPIPS following, and at matched rate still gains +0.32 dB without a single training image. A training-free stack built on the metric alone is 15% smaller than the image-free GSICO at equal quality on Mip-NeRF 360.
Chinese Translation
大多数3D高斯点云模型的内存存储了球谐颜色系数，但每个高斯仅从训练相机的狭窄视角锥中被观察到。我们将其转化为其他压缩算法可以采用的失真度量：每个高斯的观察Gram矩阵是从视角和混合权重累积而来的，它是系数变化与平方图像误差之间的精确一阶映射，仅需要模型和相机姿态。在此基础上，度数减少成为一种封闭形式的投影，推广了截断，度数分配则成为拉格朗日率失真问题，而向量量化则是矩阵加权的Lloyd算法，其中Compressed3D的量化器是标量情况。在Compressed3D中替换该度量而其他部分不变，PSNR在微调前提高了+0.49 dB，SSIM和LPIPS也随之提升，并且在匹配速率下仍然获得+0.32 dB，而没有使用任何训练图像。基于该度量构建的无训练堆栈在Mip-NeRF 360上比无图像的GSICO小15%，且质量相当。
cs.CV / 24 / 2609.29006
FluidRain: Incompressible Rain Flow as an Attention Bias for Loop-in-Loop Video Deraining
FluidRain：作为循环视频去雨的注意力偏置的不可压缩雨流
Abstract
Existing video deraining methods typically exploit neighboring frames through either explicit alignment or implicit spatiotemporal aggregation. Explicit alignment relies on accurate motion estimation, which can become unreliable under dense rain, while implicit aggregation avoids alignment but lacks explicit guidance on the directional and temporally coherent structure of rain. This leaves a gap between reliable temporal aggregation and explicit modeling of rain motion. To address these limitations, we propose FluidRain, a lightweight video derainer that uses divergence-free rain flow to guide Loop-in-Loop attention across scales and neighboring frames. Motivated by fluid mechanics, we model rain motion as a divergence-free image-space flow and use it to organize multi-scale and temporal aggregation. Specifically, FluidRain first estimates a rain-flow field for each frame and projects it onto the divergence-free subspace. The resulting flow steers window attention along rain streaks, enabling neighboring frames to be aggregated without explicit alignment. Since rain-flow structure is preserved across scales and nearby frames, Loop-in-Loop reuses the same attention operator across both dimensions, resulting in a three-frame model with only 0.80M parameters. Experiments on four benchmarks show that FluidRain remains competitive with substantially larger restoration models. We further examine how temporal evidence scales with different input views. To evaluate whether the model remains reliable when rain motion changes across frames, we introduce RainSyn-Gust, which injects controlled changes in rain-streak direction into existing benchmarks. We also develop a physics-based no-reference metric that evaluates real-rain removal without requiring clean targets.
Chinese Translation
现有的视频去雨方法通常通过显式对齐或隐式时空聚合来利用相邻帧。显式对齐依赖于准确的运动估计，但在密集降雨情况下可能变得不可靠，而隐式聚合避免了对齐，但缺乏对雨的方向性和时间一致性结构的显式指导。这在可靠的时间聚合与雨运动的显式建模之间留下了空白。为了解决这些局限性，我们提出了FluidRain，一种轻量级视频去雨器，利用无散度雨流来引导跨尺度和相邻帧的循环注意力。受到流体力学的启发，我们将雨的运动建模为无散度的图像空间流，并利用它来组织多尺度和时间聚合。具体而言，FluidRain首先为每帧估计一个雨流场，并将其投影到无散度子空间。生成的流沿着雨条引导窗口注意力，使得相邻帧可以在没有显式对齐的情况下进行聚合。由于雨流结构在不同尺度和相邻帧之间得以保留，循环注意力在两个维度上重用相同的注意力算子，从而形成一个仅有0.80M参数的三帧模型。在四个基准测试上的实验表明，FluidRain在恢复性能上与大得多的恢复模型保持竞争力。我们进一步研究了时间证据如何随着不同输入视图的变化而变化。为了评估模型在雨运动跨帧变化时的可靠性，我们引入了RainSyn-Gust，该方法在现有基准中注入控制的雨条方向变化。我们还开发了一种基于物理的无参考度量，用于评估真实雨水去除，而无需干净的目标。
cs.CV / 25 / 2609.29028
RGBD20K: A Large-Scale Benchmark for RGB-D Semantic Segmentation
RGBD20K：用于RGB-D语义分割的大规模基准数据集
Abstract
In this paper, we propose RGBD20K, a novel dataset for facilitating the development of more robust and general RGB-D semantic segmentation by encompassing abundant categories and high-quality annotations. RGBD20K possesses several attractive properties: (1) Expanded Semantic Space. In particular, it covers 160 fine-grained categories, largely surpassing the category diversity of existing popular RGB-D benchmarks (e.g., NYUv2 with 40 classes and SUN RGB-D with 37 classes). With such enriched semantic coverage, we expect to promote the learning of more generalizable segmentation models. (2) Larger Scale. Compared with current benchmarks, RGBD20K offers 20,000 RGB-D image pairs, providing a substantially larger training resource that benefits the development of more powerful deep models. (3) High-Fidelity Annotation. We perform rigorous re-evaluation and correction of existing labels to resolve long-standing annotation noise, resulting in a clean and reliable ground-truth foundation. Furthermore, we propose a novel score-purified fusion (SPF) method, which achieves state-of-the-art performance across all evaluated benchmarks, demonstrating the effectiveness of our approach in leveraging high-quality multimodal information for RGB-D semantic segmentation. The dataset is here: https://github.com/ShaohuaDong2021/RGBD20K/.
Chinese Translation
在本文中，我们提出了RGBD20K，这是一个新颖的数据集，旨在促进更强大和通用的RGB-D语义分割的发展，涵盖了丰富的类别和高质量的注释。RGBD20K具有几个吸引人的特性：（1）扩展的语义空间。特别是，它覆盖了160个细粒度类别，远超现有流行RGB-D基准（例如，NYUv2有40个类别，SUN RGB-D有37个类别）的类别多样性。凭借如此丰富的语义覆盖，我们期望促进更具通用性的分割模型的学习。（2）更大的规模。与当前基准相比，RGBD20K提供了20,000对RGB-D图像，提供了大幅增加的训练资源，有利于开发更强大的深度模型。（3）高保真注释。我们对现有标签进行了严格的重新评估和修正，以解决长期存在的注释噪声，从而形成一个干净可靠的真实基础。此外，我们提出了一种新颖的得分净化融合（SPF）方法，在所有评估的基准上实现了最先进的性能，展示了我们在利用高质量多模态信息进行RGB-D语义分割方面的有效性。数据集链接为：https://github.com/ShaohuaDong2021/RGBD20K/
cs.CV / 26 / 2609.29029
Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge
利用卫星图像中的答案不变冗余实现边缘高效的视觉语言模型推理
Abstract
Onboard vision-language models could enable satellites to answer queries directly, but exhaustive tiled inference over high-resolution imagery is slow and energy-intensive. We identify answer-invariant token redundancy (AITR): image tiles and vision tokens that can be removed without changing the final answer. We present Rift, a two-stage system that performs query-conditioned tile pruning followed by elastic prefill to reduce token budget. We evaluate it on LLaVA-1.5 7B running on Jetson AGX Orin. Compared with exhaustive tiled inference, Rift reduces energy by 78% and latency by 69%, while increasing accuracy from 45% to 73%.
Chinese Translation
机载视觉语言模型可以使卫星直接回答查询，但在高分辨率图像上进行全面的瓷砖推理既缓慢又耗能。我们识别出答案不变的标记冗余（AITR）：可以在不改变最终答案的情况下移除的图像瓷砖和视觉标记。我们提出了Rift，一个两阶段系统，首先进行查询条件的瓷砖剪枝，然后进行弹性预填充，以减少标记预算。我们在Jetson AGX Orin上评估了它在LLaVA-1.5 7B上的表现。与全面的瓷砖推理相比，Rift将能耗降低了78%，延迟降低了69%，同时将准确率从45%提高到73%。
cs.CV / 27 / 2609.29048
Where Hallucinations Live: A Cross-Architecture Circuit in VQ-Tokenized Vision-Language Models
幻觉的存在之处：VQ-标记视觉-语言模型中的跨架构电路
Abstract
Unified vision-language models (VLMs) that tokenize images through a vector-quantized (VQ) codebook routinely hallucinate objects on grounded yes/no benchmarks, yet existing decoding-time fixes treat this as generic miscalibration without an architectural account. Using activation patching across twenty-five models spanning eight LLM families, we identify an early-layer ($L_0$) attention routing circuit shared across VQ-tokenized VLMs and propose a three-gate diagnostic that distinguishes the models carrying it from those that do not. The diagnostic isolates ten positive models (five natural unified-VQ VLMs across three LLM families and five induced variants) and rejects the remaining fifteen. A single-variable architectural swap (LLaVA-1.6 CLIP+MLP $\rightarrow$ VQ+Linear) installs the circuit, while a matched-compute MLP control on identical data does not, isolating vector quantization as the source of the pathological signal; the routing pathway that carries it is one that the backbone already provides. Against tuned VCD and DoLA baselines, tuned DoLA wins on binary calibration, but \textbf{only $L_0$ ablation reduces object hallucination in open-ended generation} (CHAIR$_i$ reduces by $31\,\%$ relatively, whereas tuned DoLA and VCD leave it unchanged or worsen it). These results recast object hallucination in unified VQ VLMs as a property of architecture and pretraining, and yield a targeted intervention that mechanism-agnostic decoding cannot replicate.
Chinese Translation
统一的视觉-语言模型（VLMs）通过向量量化（VQ）代码本对图像进行标记，在基于实证的是/否基准测试中经常产生幻觉对象，然而现有的解码时间修正将其视为一般性误校准，而没有考虑架构因素。通过对跨越八个大型语言模型（LLM）家族的二十五个模型进行激活补丁分析，我们识别出一个在VQ-标记的VLMs中共享的早期层（$L_0$）注意力路由电路，并提出一个三门诊断方法，以区分携带该电路的模型与不携带的模型。该诊断方法隔离出十个正模型（五个自然统一的VQ VLMs，涵盖三个LLM家族，以及五个诱导变体），并拒绝其余的十五个。一个单变量架构交换（LLaVA-1.6 CLIP+MLP $
ightarrow$ VQ+Linear）安装了该电路，而在相同数据上进行匹配计算的MLP控制则未能做到，从而将向量量化确定为病态信号的来源；承载该信号的路由路径是骨干网络已经提供的。与调优的VCD和DoLA基线相比，调优的DoLA在二元校准上获胜，但 extbf{仅有$L_0$消融减少了开放式生成中的对象幻觉}（CHAIR$_i$相对减少了$31\, ext{ extperthousand}$，而调优的DoLA和VCD则未改变或加重了这一现象）。这些结果将统一VQ VLMs中的对象幻觉重新定义为架构和预训练的属性，并产生了一种机制无关的解码无法复制的针对性干预。
cs.CV / 28 / 2609.29064
EIB-Net: Entropy-Guided Information Bottleneck for Generalizable AI-Generated Image Detection
EIB-Net：基于熵引导的信息瓶颈用于可泛化的AI生成图像检测
Abstract
The proliferation of photorealistic AI-generated images demands robust detection methods that generalize across diverse generative models. While existing approaches target manipulation-based forgeries with local artifacts, generation-based images (e.g., from diffusion models) lack such traces, posing a fundamental challenge. We observe that generative models prioritize global semantics at the expense of local texture fidelity, making low-texture regions key indicators of synthetic origin. To exploit this, we propose EIB-Net, an Entropy-guided Information Bottleneck Network. EIB-Net introduces a novel Image Entropy (IE) metric to automatically select the most informative (lowest-entropy) patch, then processes it with a Variational Information Bottleneck (VIB) to learn compact, generalizable features. Extensive experiments on DIFF, DiffusionForensics, and GenImage benchmarks demonstrate state-of-the-art performance: EIB-Net achieves 85.7\% accuracy using only 2\% of training data, outperforming full-image baselines by over 15\%, and maintains robust cross-generator generalization (83.5\% average accuracy on GenImage). Furthermore, our entropy-guided patch selection (EGPL) consistently enhances diverse backbones (CNNs and Transformers), proving its practical value for data-efficient detection.
Chinese Translation
逼真的AI生成图像的激增需要强大的检测方法，以便在多种生成模型中实现泛化。尽管现有方法针对带有局部伪影的操控性伪造，但基于生成的图像（例如，来自扩散模型的图像）缺乏此类痕迹，这构成了一个基本挑战。我们观察到，生成模型优先考虑全局语义，而牺牲局部纹理的保真度，使得低纹理区域成为合成来源的关键指标。为了利用这一点，我们提出了EIB-Net，一种基于熵引导的信息瓶颈网络。EIB-Net引入了一种新颖的图像熵（IE）度量，自动选择信息量最大的（最低熵）图像块，然后通过变分信息瓶颈（VIB）进行处理，以学习紧凑且可泛化的特征。在DIFF、DiffusionForensics和GenImage基准上的广泛实验表明，EIB-Net实现了最先进的性能：EIB-Net在仅使用2%的训练数据的情况下达到了85.7%的准确率，超过了全图基线15%以上，并且在跨生成器泛化方面保持了强大的能力（在GenImage上平均准确率为83.5%）。此外，我们的熵引导图像块选择（EGPL）持续增强了多种骨干网络（CNN和变换器），证明了其在数据高效检测中的实际价值。
cs.CV / 29 / 2609.29073
Seeing Is Not Measuring: Tool-Augmented Metric Spatial Reasoning for Vision-Language Models
看并不等于测量：工具增强的度量空间推理用于视觉-语言模型
Abstract
Vision-Language Models (VLMs) describe scenes well but reason poorly about metric 3D structure such as absolute distances, physical sizes, or egocentric directions. We present a modular, predictor agnostic, tool-augmented framework that equips a small VLM (Qwen3.5-4B) with geometric tools: 3D object detection, metric depth estimation, and deterministic solvers for distance, size and bearing. Each object is detected in the camera frame of its own best view, and the tools use that frame's pose to lift every detection into one shared world frame. Moving metric computation out of the model's weights and into explicit solvers yields large gains on three of four ReVSI-Bench tasks: with a strong monocular detector (WildDet3D), absolute distance rises from 0.46 to 0.74 Mean Relative Accuracy (MRA), relative distance from 39.1% to 67.4%, and relative direction from a below-chance 25.9% to 73.4%. Because any detector can be swapped in behind the tool interface, comparing real detectors against ground-truth boxes separates perception error from reasoning error: orchestration costs only 0.03 MRA. Object size is bounded by the detector: the tools are near-exact on groundtruth boxes (0.97) yet the best real detector barely beats the no-tool baseline (0.61 vs. 0.58), because size reads straight off a box extent monocular detectors get wrong. Without a predefined recipe, the model already sequences the tools correctly on its own, matching a scripted pipeline on three of four tasks.
Chinese Translation
视觉-语言模型（VLMs）能够很好地描述场景，但在度量3D结构方面推理能力较弱，例如绝对距离、物理尺寸或自我中心方向。我们提出了一种模块化、预测器无关的工具增强框架，为小型VLM（Qwen3.5-4B）配备几何工具：3D物体检测、度量深度估计和距离、尺寸及方向的确定性求解器。每个物体都在其最佳视角的相机框架中被检测，工具利用该框架的姿态将每个检测提升到一个共享的世界框架中。将度量计算从模型权重中移出并转入显式求解器，在四个ReVSI-Bench任务中的三个任务上取得了显著提升：使用强大的单目检测器（WildDet3D），绝对距离的平均相对准确率（MRA）从0.46上升到0.74，相对距离从39.1%上升到67.4%，相对方向从低于随机的25.9%上升到73.4%。由于任何检测器都可以在工具接口后进行替换，将真实检测器与真实框进行比较可以将感知误差与推理误差分开：协调成本仅为0.03 MRA。物体尺寸受检测器的限制：在真实框上，工具的准确度接近完美（0.97），但最佳真实检测器的表现仅略好于无工具基线（0.61对0.58），因为尺寸直接从一个框的范围读取，而单目检测器对此的识别存在错误。在没有预定义配方的情况下，模型已经能够自行正确地顺序使用工具，在四个任务中的三个任务上与脚本化管道相匹配。
cs.CV / 30 / 2609.29106
WildHSR: Metric Feed-Forward 4D People-Scene Reconstruction from a 3D Foundation Model
WildHSR：基于3D基础模型的度量前馈4D人-场景重建
Abstract
3D foundation models recover video cameras and geometry in one forward pass, but some of the strongest are up to scale. Joint people-scene reconstruction then requires two missing outputs: metric scale and persistent person identity. We ask whether one up-to-scale foundation representation can support both through lightweight adaptation. Exact metric labels are scarce, but unlabeled in-the-wild video is abundant. We use people in curated web video to initialise the solution: a posed metric body and 2D keypoints give an approximate, closed-form scale pseudo-label. These pseudo-labels pretrain a Scale Readout, which is then fine-tuned together with a lightweight adapter using exact metric supervision from standard real-video training splits. At inference the head predicts metric scale from foundation-model tokens, without the ruler or its teachers. For person identity, we probe the pretrained foundation model alone and find evidence that its intermediate query-key features encode person correspondence across frames. In most evaluated moving-person clips, a mid-layer token prefers that person over the vacated location and other people. A tiny projection reads this correspondence; together with metric pelvis motion and proposal confidence, it drives dustbin-aware Sinkhorn association of per-frame bodies. WildHSR combines both readouts to reconstruct metric cameras, scene and people from monocular video. Each window is predicted feed-forward; analytic association and Sim(3) composition connect windows. On EMDB-2, WildHSR is the first feed-forward method in the published comparison to beat the best optimization-based WA-MPJPE and RTE while leading feed-forward methods on all three world-frame metrics. On RICH, it leads feed-forward people-and-scene methods on WA-MPJPE and W-MPJPE. The complete pipeline runs at 10.1 fps on one GPU.
Chinese Translation
3D基础模型在一次前向传递中恢复视频摄像机和几何形状，但一些最强大的模型需要达到比例。联合人-场景重建则需要两个缺失的输出：度量尺度和持久的人物身份。我们探讨是否可以通过轻量级适配使一个达到比例的基础表示同时支持这两者。确切的度量标签稀缺，但未标记的野外视频却非常丰富。我们利用策划的网络视频中的人物来初始化解决方案：一个姿势度量体和2D关键点提供了一个近似的、封闭形式的尺度伪标签。这些伪标签预训练一个尺度读取器，然后与使用标准真实视频训练分割的确切度量监督一起微调轻量级适配器。在推理时，头部从基础模型的标记中预测度量尺度，而无需标尺或其教师。对于人物身份，我们单独探测预训练的基础模型，并发现其中间查询-键特征在帧间编码人物对应关系。在大多数评估的移动人物剪辑中，中间层标记更倾向于该人物而非空置位置和其他人物。一个微小的投影读取这种对应关系；结合度量骨盆运动和提议置信度，它驱动了对每帧身体的垃圾桶感知Sinkhorn关联。WildHSR结合这两种读取方式，从单目视频重建度量摄像机、场景和人物。每个窗口都是前馈预测的；解析关联和Sim(3)组合连接窗口。在EMDB-2上，WildHSR是已发布比较中第一个超越最佳优化基础的WA-MPJPE和RTE的前馈方法，同时在所有三个世界框架度量上领先前馈方法。在RICH上，它在WA-MPJPE和W-MPJPE上领先前馈人-场景方法。完整的管道在一台GPU上以10.1帧每秒的速度运行。
cs.CV / 31 / 2609.29116
Spectral Amplitude Purification in Distribution Matching for Diffusion Distillation
扩散蒸馏中的分布匹配光谱幅度净化
Abstract
Distribution Matching Distillation (DMD) enables high-quality diffusion sampling in only a few steps, but its optimization dynamics remain dominated by coarse, low-frequency signals, delaying the recovery of fine-grained details. We identify a pronounced concentration of spectral amplitudes at low frequencies in the DMD directional error, where dominant low-frequency components overwhelm weaker mid- and high-frequency signals. To address this issue, we propose Spectral Amplitude Purification for Distribution Matching Distillation (SAP-DMD), a plug-and-play approach that adaptively modulates the amplitude spectrum of the DMD directional field. By suppressing the dominant tail of the amplitude spectrum, SAP-DMD reduces low-frequency dominance and promotes more effective recovery of fine structures and textures. Experiments on PixArt-$\alpha$, SD3, and SD3.5 demonstrate that SAP-DMD accelerates training convergence and improves generation quality under both 2-step and 4-step sampling.
Chinese Translation
分布匹配蒸馏（Distribution Matching Distillation, DMD）能够在仅仅几个步骤内实现高质量的扩散采样，但其优化动态仍然受到粗糙的低频信号的主导，延缓了细粒度细节的恢复。我们发现DMD方向误差中低频段的光谱幅度存在明显的集中现象，主导的低频成分压制了较弱的中频和高频信号。为了解决这一问题，我们提出了分布匹配蒸馏的光谱幅度净化（Spectral Amplitude Purification for Distribution Matching Distillation, SAP-DMD），这是一种即插即用的方法，能够自适应地调节DMD方向场的幅度谱。通过抑制幅度谱的主导尾部，SAP-DMD减少了低频的主导性，并促进了细结构和纹理的更有效恢复。在PixArt-$eta$、SD3和SD3.5上的实验表明，SAP-DMD加速了训练收敛，并在2步和4步采样下提高了生成质量。
cs.CV / 32 / 2609.29118
UpDown-SC: Gravity-Canonicalized Dual-Envelope Scan Context for Indoor LiDAR Place Recognition
UpDown-SC：用于室内LiDAR位置识别的重力规范化双包络扫描上下文
Abstract
LiDAR place recognition is a key front end for loop closure and global relocalization, yet indoor retrieval remains difficult when attitude or sensor mounting height changes between mapping and query sessions. Scan Context stores the maximum height in each polar cell; indoors, broad ceilings can suppress the lower and mid-level geometry that distinguishes adjacent rooms and corridors. We present UpDown-SC, a training-free polar descriptor that first canonicalizes gravity and then represents two complementary surfaces: the upper envelope of lower/middle structures and the lower envelope of overhead structures. Their physical split is estimated once from a cell-balanced map height distribution and reused by every query. A mask-aware, non-uniform two-channel distance retains discriminative lower-level evidence while limiting sensitivity to its cross-session variation, without treating unobserved cells as zero-height measurements. Conventional Scan Context shortlisting and circular yaw alignment are retained, so retrieved hypotheses directly initialize geometric verification. Experiments across repeated indoor sessions, mounting-height changes, mixed outdoor-to-indoor trajectories, and an outdoor transfer sequence show more reliable first-choice retrieval on the indoor and mounting-height-varied sessions. A paired test finds a significant gain over Scan Context on the in-house sessions. UpDown-SC also gives the best or second-best F1max and AUPR under threshold-based acceptance while retaining a lightweight CPU front end. Continuous replay confirms that the retrieved hypotheses support metric prior-map localization. Code and evaluation artifacts: https://github.com/jiejie567/updown-sc.
Chinese Translation
LiDAR位置识别是闭环检测和全局重定位的关键前端，但当映射和查询会话之间的姿态或传感器安装高度发生变化时，室内检索仍然困难。扫描上下文在每个极坐标单元中存储最大高度；在室内，宽阔的天花板可能会抑制区分相邻房间和走廊的低层和中层几何特征。我们提出了UpDown-SC，这是一种无训练的极坐标描述符，它首先对重力进行规范化，然后表示两个互补的表面：下层/中层结构的上包络和上层结构的下包络。它们的物理分离是基于单元平衡的地图高度分布一次性估计的，并在每次查询中重用。一个考虑掩码的非均匀双通道距离保留了具有区分性的低层证据，同时限制了对其跨会话变化的敏感性，而不将未观察到的单元视为零高度测量。传统的扫描上下文候选筛选和圆形偏航对齐被保留，因此检索到的假设可以直接初始化几何验证。在重复的室内会话、安装高度变化、混合的室外到室内轨迹以及室外转移序列中的实验表明，在室内和高度变化的会话中，首次检索的可靠性更高。配对测试发现，在内部会话中与扫描上下文相比有显著提升。UpDown-SC在基于阈值的接受下也在F1max和AUPR方面表现最佳或次佳，同时保持轻量级的CPU前端。连续重放确认检索到的假设支持度量先验地图定位。代码和评估文档：https://github.com/jiejie567/updown-sc。
cs.CV / 33 / 2609.29121
Less is More: Encoder-only Audio-Visual Segmentation
少即是多：仅编码器的音视频分割
Abstract
Audio-Visual Semantic Segmentation (AVSS) aims to identify, segment, and classify sound-emitting objects in video frames. Previous Transformer-based AVSS approaches largely inherit design principles from image segmentation models. Recent studies show that these image segmentation models contain redundant components that contribute little to the segmentation performance. Following this insight, we propose Encoder-only Audio-Visual Segmentation (EASE). EASE runs at up to 365 FPS, 3x faster than prior State-of-the-Art (SotA) AVS models at comparable accuracy, and trains in under 11 GPU-hours. Furthermore, we achieve SotA AVSS performance across different backbones and input resolutions. Our results demonstrate that AVSS can be both simpler and faster, providing a scalable foundation for future research and real-time applications. Code, model weights, and samples are available at https://ease-avs.notion.site
Chinese Translation
音视频语义分割（AVSS）旨在识别、分割和分类视频帧中的发声物体。以往基于Transformer的AVSS方法在很大程度上继承了图像分割模型的设计原则。最近的研究表明，这些图像分割模型包含冗余组件，这些组件对分割性能贡献甚微。基于这一见解，我们提出了仅编码器的音视频分割（EASE）。EASE的运行速度可达365帧每秒，比之前的最先进（SotA）AVS模型快3倍，同时保持相当的准确性，并且训练时间少于11个GPU小时。此外，我们在不同的骨干网络和输入分辨率下实现了SotA的AVSS性能。我们的结果表明，AVSS可以更简单、更快速，为未来的研究和实时应用提供可扩展的基础。代码、模型权重和样本可在https://ease-avs.notion.site获取。
cs.CV / 34 / 2609.29125
FoCal: Frequency-Oriented Cross-Modal Interaction and Spectral Calibration for Aerial Visible-Infrared Object Detection
FoCal：面向频率的跨模态交互与光谱校准用于空中可见-红外物体检测
Abstract
In aerial RGB--IR object detection, effectively exploiting complementary information across modalities is critical for robust perception under complex illumination and environmental conditions. Existing multimodal detectors mainly focus on spatial-domain interaction or frequency-specific feature enhancement, while the cross-modal interaction patterns of different frequency components remain insufficiently explored. Moreover, spectral discrepancy itself may contain both useful complementary cues and unreliable modality-specific responses, making indiscriminate frequency fusion suboptimal. To address these issues, we propose FoCal, a frequency-oriented framework for aerial RGB--IR object detection. First, a Frequency-Aware Dual-Domain Calibration (FADC) module is developed to explicitly model frequency-dependent cross-modal interaction. Low-frequency components are collaboratively consolidated into a shared structural consensus, whereas high-frequency components preserve modality-specific information through selective cross-modal exchange. The resulting frequency-aware cues are further transferred to the original feature domain to regulate cross-modal calibration. Second, we introduce a Discrepancy-Guided Spectral Modulation (DGSM) module, which characterizes cross-modal spectral imbalance using confidence-weighted relative amplitude discrepancy and transforms it into a bounded signed gate for adaptive enhancement, preservation, or attenuation of the joint multimodal spectrum. Extensive experiments on DroneVehicle, ESCVehicle, and ATR-UMOD demonstrate the effectiveness of FoCal, yielding $\mathrm{mAP}_{50}$ values of 83.5\%, 54.8\%, and 64.6\%, respectively. Meanwhile, with only 3.0M parameters, FoCal achieves 113.6 FPS while preserving leading detection accuracy, highlighting a favorable accuracy--efficiency trade-off. Code is available at {https://github.com/universeliang/FoCal.
Chinese Translation
在空中RGB-IR物体检测中，有效利用跨模态的互补信息对于在复杂的光照和环境条件下实现稳健感知至关重要。现有的多模态检测器主要集中于空间域交互或特定频率特征增强，而不同频率成分的跨模态交互模式仍然探索不足。此外，光谱差异本身可能包含有用的互补线索和不可靠的模态特定响应，使得不加区分的频率融合效果不佳。为了解决这些问题，我们提出了FoCal，一个面向频率的空中RGB-IR物体检测框架。首先，开发了一个频率感知双域校准（FADC）模块，以明确建模频率依赖的跨模态交互。低频成分被协同整合为共享的结构共识，而高频成分通过选择性跨模态交换保留模态特定信息。生成的频率感知线索进一步转移到原始特征域，以调节跨模态校准。其次，我们引入了一个差异引导的光谱调制（DGSM）模块，该模块使用置信加权的相对幅度差异来表征跨模态光谱不平衡，并将其转化为一个有界的带符号门，用于自适应增强、保留或衰减联合多模态光谱。在DroneVehicle、ESCVehicle和ATR-UMOD上的大量实验表明，FoCal的有效性，分别获得了83.5%、54.8%和64.6%的$ ext{mAP}_{50}$值。同时，FoCal仅用3.0M参数实现了113.6 FPS，同时保持领先的检测精度，突显了良好的精度-效率权衡。代码可在{https://github.com/universeliang/FoCal}获取。
cs.CV / 35 / 2609.29151
Recoverable Geographic Location Information in Earth-Observation Embeddings
可恢复的地理位置信息在地球观测嵌入中的应用
Abstract
Earth-observation (EO) foundation models provide reusable embeddings, yet downstream task accuracy does not reveal whether these representations encode geographic information, which may be beneficial for location-aware applications but potentially detrimental when representations invariant to geographic location are desired. We therefore evaluate the geographic coordinate robustness of Tessera v1, Tessera v1.1, and AlphaEarth by testing whether coordinates can be predicted from the embedding representations using 284 quality-verified European solar farms from 2024. We assessed geographic information content information through the association between cosine and geodesic distances and through prediction of projected coordinates in EPSG:3035. Embeddings from all three EO foundation models contain recoverable geographic information. All prediction models significantly outperform training-range uniform random sampling baselines, with AlphaEarth exhibiting the strongest distance association and lowest mean geodesic error. Both Tessera variants also yielded higher geographic distance correlations than the Sentinel-2 controls. These findings motivate geographic information content as an additional criterion for auditing EO foundation models.
Chinese Translation
地球观测（EO）基础模型提供可重用的嵌入表示，但下游任务的准确性并不能揭示这些表示是否编码了地理信息，而这对于位置感知应用可能是有益的，但在需要与地理位置无关的表示时则可能是有害的。因此，我们通过测试284个来自2024年的质量验证欧洲太阳能农场的嵌入表示，评估了Tessera v1、Tessera v1.1和AlphaEarth的地理坐标鲁棒性，以确定是否可以从嵌入表示中预测坐标。我们通过余弦距离和测地距离之间的关联以及在EPSG:3035中预测投影坐标来评估地理信息内容。所有三个EO基础模型的嵌入均包含可恢复的地理信息。所有预测模型的表现显著优于训练范围内的均匀随机采样基线，其中AlphaEarth表现出最强的距离关联和最低的平均测地误差。两个Tessera变体的地理距离相关性也高于Sentinel-2对照组。这些发现促使将地理信息内容作为审计EO基础模型的额外标准。
cs.CV / 36 / 2609.29156
Med-AR: Autoregressive Vision-Language Pretraining for Long-Tailed Chest X-Ray Classification and Uncertainty-Aware Evaluation
Med-AR：用于长尾胸部X光分类和不确定性评估的自回归视觉-语言预训练
Abstract
Long-tailed chest X-ray classification requires visual representations that capture both common abnormalities and subtle, infrequent findings. We propose Med-AR-8B and Med-AR-2B, two radiology-native autoregressive vision-language models pretrained with structured reports, abnormality-focused text, and region annotations. We evaluate the transfer of their visual encoders to multi-label classification against contrastive, self-supervised, and supervised pretrained encoders, including Med-CLIP, CheXFound, EVA-Base, ARK, and BioViL-T, using a common ML-Decoder classification head. To assess fine-grained recognition, we also construct LLM-expanded, report-derived label sets for MIMIC-CXR and CheXpert. Across PadChest, MIMIC-CXR, and CheXpert, Med-AR-8B outperforms Med-CLIP in mean AUROC and AUPRC for head, medium, and tail findings. On MIMIC-CXR, it increases tail-label mean AUPRC from 0.1033 to 0.1441. Med-AR-2B achieves the strongest discrimination results on PadChest. Across the broader encoder comparison, a Med-AR variant achieves the highest mean AUROC and AUPRC in every reported prevalence group on each public dataset. Both Med-AR variants also achieve lower excess area under the risk-coverage curve than Med-CLIP on all three public datasets, indicating improved selective-prediction performance under the evaluated protocol. Internal results are metric-dependent, with Med-CLIP retaining advantages in overall and tail AUPRC and in selective prediction. These findings establish Med-AR as a strong pretraining recipe for long-tailed chest X-ray classification on the evaluated public benchmarks and demonstrate the value of assessing discrimination and selective prediction together.
Chinese Translation
长尾胸部X光分类需要能够捕捉常见异常和细微、不常见发现的视觉表征。我们提出了Med-AR-8B和Med-AR-2B，这两个以放射学为基础的自回归视觉-语言模型通过结构化报告、以异常为中心的文本和区域注释进行预训练。我们评估了它们的视觉编码器在多标签分类中的迁移能力，并与对比、自监督和监督预训练编码器进行比较，包括Med-CLIP、CheXFound、EVA-Base、ARK和BioViL-T，使用共同的ML-Decoder分类头。为了评估细粒度识别，我们还为MIMIC-CXR和CheXpert构建了LLM扩展的报告衍生标签集。在PadChest、MIMIC-CXR和CheXpert数据集中，Med-AR-8B在头部、中等和尾部发现的平均AUROC和AUPRC上均优于Med-CLIP。在MIMIC-CXR上，它将尾部标签的平均AUPRC从0.1033提高到0.1441。Med-AR-2B在PadChest上实现了最强的区分结果。在更广泛的编码器比较中，Med-AR变体在每个公共数据集的每个报告流行度组中均实现了最高的平均AUROC和AUPRC。两个Med-AR变体在所有三个公共数据集上也实现了低于Med-CLIP的风险覆盖曲线下的过度面积，表明在评估协议下选择性预测性能得到了改善。内部结果依赖于度量，Med-CLIP在整体和尾部AUPRC以及选择性预测方面仍然保持优势。这些发现确立了Med-AR作为在评估的公共基准上进行长尾胸部X光分类的强大预训练方案，并展示了同时评估区分和选择性预测的价值。
cs.CV / 37 / 2609.29186
An Automated Georeferencing Technique for Multi-Temporal Stope Point Clouds for Downstream Geotechnical Analysis
一种用于多时相矿柱点云的自动地理配准技术以支持下游岩土工程分析
Abstract
The increasing use of UAV laser scanning in underground mines has enabled frequent acquisition of 3D point clouds from challenging environments such as stopes, generating large volumes of multi-temporal spatial data throughout successive excavation stages. However, in GNSS-denied underground environments, independently acquired stope point clouds are generated within local scanner reference frames and require registration and georeferencing before integration with mine reference data for downstream geotechnical analysis, monitoring, and mine planning. This process is commonly performed manually by aligning individual stope scans with mine reference drives, making repeated georeferencing time-consuming and potentially limiting the utilisation of routinely acquired data. This study proposes the 3D Tag-based Automated Registration and Georeferencing Technique (3D-TARGeT), an automated framework using low-cost, generic, non-unique rectangular tags to establish spatial correspondence between stope point clouds and the mine reference coordinate system. The framework combines automated tag identification, geometric tag matching, and rigid transformation estimation. It was evaluated as a proof of concept using four multi-temporal point-cloud scans of an underground mine stope, with the proposed tags simulated under representative scanning conditions. 3D-TARGeT achieved consistent centimetre-level georeferencing accuracy, with median cloud-to-cloud distance and root mean square error below 0.03 m across all scans, while substantially outperforming widely used automatic point-cloud registration techniques. Overall, 3D-TARGeT provides an accurate and robust approach for automating stope point-cloud georeferencing, reducing reliance on manual alignment and facilitating multi-temporal datasets for downstream geological and geotechnical applications.
Chinese Translation
无人机激光扫描在地下矿山中的日益广泛应用，使得能够频繁获取矿柱等复杂环境的三维点云数据，在连续开采阶段产生大量多时相空间数据。然而，在无全球导航卫星系统（GNSS）信号的地下环境中，独立采集的矿柱点云均处于局部扫描仪参考系内，需经过配准和地理配准后，方可与矿山参考数据集成，用于下游岩土工程分析、监测及矿山规划。该过程通常通过手动将单个矿柱扫描与矿山参考巷道对齐完成，重复进行地理配准既耗时又可能限制常规采集数据的利用。本文提出了基于三维标签的自动配准与地理配准技术（3D Tag-based Automated Registration and Georeferencing Technique，3D-TARGeT），该自动化框架利用低成本、通用且非唯一的矩形标签，在矿柱点云与矿山参考坐标系之间建立空间对应关系。该框架结合了自动标签识别、几何标签匹配及刚性变换估计。作为概念验证，本文采用地下矿柱的四组多时相点云扫描数据进行评估，所用标签在代表性扫描条件下进行了模拟。3D-TARGeT实现了稳定的厘米级地理配准精度，所有扫描的点云间中位距离及均方根误差均低于0.03米，且显著优于广泛使用的自动点云配准技术。总体而言，3D-TARGeT为矿柱点云地理配准提供了一种准确且鲁棒的自动化方法，减少了对手动对齐的依赖，促进了多时相数据集在下游地质及岩土工程应用中的利用。
cs.CV / 38 / 2609.29193
ImCorr: Sub-pixel Semantic Correspondence via Implicit Feature Decoding
ImCorr：通过隐式特征解码实现亚像素语义对应
Abstract
The strong performance that modern semantic correspondence methods achieve at standard thresholds plateaus sharply at fine-grained thresholds. We argue that this plateau stems not from the representational capacity of backbone features, but from a grid-tied readout. Patch-based vision transformers tokenize images onto discrete grids, introducing two forms of quantization error: querying nearest patch features instead of exact keypoints on the source side, and the absence of grid features representing precise ground-truth locations on the target side. We quantify this quantization ceiling across all 499,188 keypoints in SPair-71k: under the standard 448x448, patch-14 setting, 84.9% of ground-truth keypoints have no grid feature representing their precise location at
[email protected]. This is a structural limitation at the representation level, independent of the matching strategy. We address this with ImCorr: Sub-pixel Semantic Correspondence via Implicit Feature Decoding, which formulates correspondence estimation over a continuous feature field queryable at arbitrary continuous coordinates. A FiLM-conditioned decoder is trained to embed sub-pixel positional information into the feature field. Querying the field directly at exact keypoint coordinates theoretically eliminates representation-level quantization error on the source side, while decoding onto a grid denser than the backbone grid substantially reduces quantization error on the target side. On SPair-71k and AP-10K (intra-species, cross-species, and cross-family), ImCorr improves performance at fine-grained thresholds (
[email protected]), achieving a 6.2 percentage point gain over the prior state of the art at
[email protected] on SPair-71k. These results demonstrate that representational continuity is an effective solution for precise semantic correspondence. Code is available at https://github.com/YusungChoi/ImCorr.
Chinese Translation
现代语义对应方法在标准阈值下表现出色，但在细粒度阈值下性能急剧平 plateau。我们认为，这一平台期并非源于主干特征的表征能力，而是由于网格绑定的读取方式。基于补丁的视觉变换器将图像分割为离散网格，引入了两种量化误差：在源侧查询最近的补丁特征而非精确的关键点，以及在目标侧缺乏表示精确真实位置的网格特征。我们量化了在 SPair-71k 中所有 499,188 个关键点的量化上限：在标准的 448x448、patch-14 设置下，84.9% 的真实关键点在
[email protected] 下没有网格特征表示其精确位置。这是一个在表征层面上的结构性限制，与匹配策略无关。我们通过 ImCorr：通过隐式特征解码实现亚像素语义对应来解决这一问题，该方法在可查询任意连续坐标的连续特征场上制定对应估计。训练一个 FiLM 条件解码器，将亚像素位置信息嵌入特征场。理论上，在精确关键点坐标处直接查询特征场消除了源侧的表征级量化误差，而在比主干网格更密集的网格上解码则显著减少了目标侧的量化误差。在 SPair-71k 和 AP-10K（同种内、跨种、跨科）上，ImCorr 在细粒度阈值（
[email protected]）下提高了性能，在 SPair-71k 的
[email protected] 上比之前的最先进技术提高了 6.2 个百分点。这些结果表明，表征连续性是实现精确语义对应的有效解决方案。代码可在 https://github.com/YusungChoi/ImCorr 获取。
cs.CV / 39 / 2609.29224
FounRef: Robust, Structure-Preserving, and Fast Metric Refinement of Frozen Monocular Foundation Priors with Sparse Anchors
FounRef：稳健、结构保持且快速的冻结单目基础先验的稀疏锚点度量精炼
Abstract
Dense metric depth from cameras is essential to real-world 3D applications, yet achieving accuracy, faithful surface geometry, and fast inference simultaneously remains challenging. Monocular foundation models provide rich, transferable geometric priors but lack reliable metric scale, while depth-completion networks recover metric depth at the cost of geometric fidelity, cross-domain robustness, or speed. We present FounRef, a training-free method that aligns a frozen monocular foundation prior with sparse metric anchors to produce dense metric depth. FounRef is modular by design: its depth prior, anchor source, and refinement solver can each be replaced independently. We instantiate FounRef with MoGe-2 and LiDAR anchors. FounRef validates each anchor against the prior's dense depth prediction, rejecting inconsistencies caused by cross-sensor misalignment that geometry-only filters cannot detect. It then applies global and local metric corrections through a structure-preserving solver, retaining the prior's fine-grained geometry. FounRef requires no task-specific training and operates out of the box across unfamiliar cameras and scenes. On out-of-domain data, it delivers up to 24% lower depth error, 92% lower surface-normal noise, and almost 15x faster inference than DMD3C, a state-of-the-art depth-completion network. By decoupling metric alignment from geometry prediction, FounRef provides an accurate, geometrically faithful, and efficient approach to dense metric depth that can directly benefit from future advances in foundation models and metric sensors.
Chinese Translation
来自相机的密集度量深度对于现实世界的三维应用至关重要，但同时实现准确性、真实的表面几何和快速推理仍然具有挑战性。单目基础模型提供了丰富的、可转移的几何先验，但缺乏可靠的度量尺度，而深度补全网络则以牺牲几何保真性、跨域鲁棒性或速度为代价来恢复度量深度。我们提出了FounRef，这是一种无训练的方法，通过将冻结的单目基础先验与稀疏的度量锚点对齐，以生成密集的度量深度。FounRef在设计上是模块化的：其深度先验、锚点来源和精炼求解器可以独立替换。我们用MoGe-2和LiDAR锚点实例化FounRef。FounRef通过将每个锚点与先验的密集深度预测进行验证，拒绝由跨传感器不对齐引起的不一致性，这些不一致性是仅依赖几何过滤器无法检测到的。然后，它通过结构保持求解器应用全局和局部的度量校正，保留先验的细粒度几何。FounRef不需要特定任务的训练，并且可以在不熟悉的相机和场景中开箱即用。在域外数据上，它的深度误差降低了多达24%，表面法线噪声降低了92%，推理速度比DMD3C（一个最先进的深度补全网络）快了近15倍。通过将度量对齐与几何预测解耦，FounRef提供了一种准确、几何上真实且高效的密集度量深度方法，可以直接受益于未来基础模型和度量传感器的进展。
cs.CV / 40 / 2609.29225
ComplexSync: High-Fidelity and Real-Time Lip Sync in Complex Scenarios
ComplexSync：复杂场景下的高保真实时唇动同步
Abstract
Lip synchronization aims to generate visual lip dynamics that align precisely with speech audio. Despite the high generation quality of diffusion models, they often struggle in complex scenarios and suffer from prohibitive inference latency, limiting real-world deployment. We present ComplexSync, a unified diffusion-based framework that enables real-time, high-fidelity lip sync under complex conditions. First, we introduce a dual-stream joint training strategy to mitigate information leakage from reference frames while preserving natural dynamics. Second, we develop a distillation-based acceleration scheme for single-step denoising, achieving a throughput of over 70 FPS. Third, we propose a relational alignment loss that leverages structural priors from Vision Foundation Models (VFMs) to enhance robustness against complex scene factors. Furthermore, we present the first benchmark specifically designed for complex lip synchronization, comprising over 200 challenging video sequences and specialized metrics. Extensive experiments demonstrate that ComplexSync achieves state-of-the-art performance across both standard and complex scenarios while enabling real-time inference.
Chinese Translation
唇动同步旨在生成与语音音频精确对齐的视觉唇部动态。尽管扩散模型的生成质量很高，但它们在复杂场景中往往表现不佳，并且推理延迟过高，限制了其在现实世界中的应用。我们提出了ComplexSync，一个统一的基于扩散的框架，能够在复杂条件下实现实时高保真的唇动同步。首先，我们引入了一种双流联合训练策略，以减轻参考帧的信息泄漏，同时保持自然动态。其次，我们开发了一种基于蒸馏的加速方案，用于单步去噪，达到了超过70帧每秒的吞吐量。第三，我们提出了一种关系对齐损失，利用视觉基础模型（Vision Foundation Models, VFMs）的结构先验，以增强对复杂场景因素的鲁棒性。此外，我们还提出了第一个专门为复杂唇动同步设计的基准，包括200多个具有挑战性的视频序列和专业指标。大量实验表明，ComplexSync在标准和复杂场景下均实现了最先进的性能，同时支持实时推理。
cs.CV / 41 / 2609.29235
SARFusion: Scene-Aware Routing Fusion for Robust Camera-LiDAR 3D Object Detection
SARFusion：基于场景感知的路由融合用于鲁棒的相机-LiDAR 3D物体检测
Abstract
Camera-LiDAR fusion has become a prevailing paradigm for 3D object detection in autonomous driving. However, existing fusion detectors often establish strong inter-modality dependencies by decoding object queries from tightly coupled multimodal representations. Under corrupted driving conditions, such dependencies make the detector vulnerable to unreliable modalities, where degraded observations may interfere with reliable modality-specific evidence and lead to suboptimal predictions. Moreover, modality reliability can vary across both global driving scenes and individual object queries, requiring adaptive fusion decisions at a finer granularity. To bridge this gap, we reformulate robust camera-LiDAR fusion as a scene-aware branch routing problem and propose SARFusion, a robust 3D object detector. Instead of producing detections from a single fused representation, SARFusion decouples object-query decoding into three parallel reasoning branches: a camera branch, a LiDAR branch, and a camera-LiDAR fusion branch. Guided by a Scene Reliability Prior estimated from the global driving context, SARFusion further incorporates object-level evidence to route each query to the most suitable branch. This query-wise routing strategy alleviates harmful cross-modal interference while preserving the benefits of multimodal fusion when complementary cues are trustworthy. On the nuScenes test set, SARFusion achieves strong performance with 72.5 mAP and 74.4 NDS. Extensive analyses demonstrate its robustness under challenging conditions, including sensor corruptions and environmental changes.
Chinese Translation
相机-LiDAR融合已成为自动驾驶中3D物体检测的主流范式。然而，现有的融合检测器通常通过从紧密耦合的多模态表示中解码物体查询来建立强烈的跨模态依赖关系。在恶劣的驾驶条件下，这种依赖关系使得检测器容易受到不可靠模态的影响，其中降级的观测可能干扰可靠的模态特定证据，导致次优的预测。此外，模态的可靠性在全球驾驶场景和个体物体查询之间可能会有所不同，因此需要在更细粒度上进行自适应融合决策。为了解决这一问题，我们将鲁棒的相机-LiDAR融合重新表述为一个基于场景感知的分支路由问题，并提出了SARFusion，一个鲁棒的3D物体检测器。SARFusion并不是从单一的融合表示中产生检测结果，而是将物体查询解码解耦为三个并行推理分支：相机分支、LiDAR分支和相机-LiDAR融合分支。在全球驾驶上下文中估计的场景可靠性先验的指导下，SARFusion进一步结合物体级证据，将每个查询路由到最合适的分支。这种按查询路由的策略减轻了有害的跨模态干扰，同时在互补线索可靠时保留了多模态融合的优势。在nuScenes测试集上，SARFusion取得了72.5 mAP和74.4 NDS的优异表现。大量分析表明其在传感器损坏和环境变化等挑战条件下的鲁棒性。
cs.CV / 42 / 2609.29240
TOLA: Text-aware One-Step Latent Adaptation for Diffusion-based Text Image Super-Resolution
TOLA：基于文本感知的一步潜在适应框架用于扩散式文本图像超分辨率
Abstract
Text image super-resolution (TSR) aims to recover visually faithful and readable text under unknown degradations. Existing diffusion-based methods typically rely on multi-step prediction of either the high-resolution image or its text prior, resulting in prohibitive computational cost and inference latency. More critically, an erroneous text prior may be repeatedly injected into the denoising process, causing image and text predictions to reinforce each other and progressively amplify an early recognition error into a sharp yet semantically incorrect character. To address these limitations, we propose TOLA, a Text-aware One-step Latent Adaptation framework without iterative image-text diffusion. TOLA consists of two key modules. First, a confidence-weighted text conditioning module constructs the semantic condition only once and suppresses unreliable OCR predictions before they contaminate image reconstruction. Second, a lightweight latent residual correction module explicitly estimates and corrects the structured residual errors to recover missing or distorted stroke details. Extensive experiments demonstrate our state-of-the-art performance across all evaluation metrics on both CTR-TSR-Test ($\times 4$) and RealCE-200 benchmarks. It is worth noting that our TOLA consistently surpasses existing diffusion-based TSR methods by at least 2.72 dB in PSNR on CTR-TSR-Test.
Chinese Translation
文本图像超分辨率（TSR）旨在在未知退化条件下恢复视觉上真实且可读的文本。现有的基于扩散的方法通常依赖于高分辨率图像或其文本先验的多步预测，这导致了巨大的计算成本和推理延迟。更为关键的是，错误的文本先验可能会在去噪过程中反复注入，导致图像和文本预测相互强化，并逐渐放大早期识别错误，形成清晰但语义不正确的字符。为了解决这些限制，我们提出了TOLA，一种无需迭代图像-文本扩散的基于文本感知的一步潜在适应框架。TOLA由两个关键模块组成。首先，一个基于置信度的文本条件模块仅构建一次语义条件，并在不可靠的OCR预测污染图像重建之前进行抑制。其次，一个轻量级的潜在残差校正模块显式估计并校正结构化残差错误，以恢复缺失或扭曲的笔画细节。大量实验表明，我们在CTR-TSR-Test（$ imes 4$）和RealCE-200基准测试中的所有评估指标上均表现出最先进的性能。值得注意的是，我们的TOLA在CTR-TSR-Test上在PSNR方面始终超过现有的基于扩散的TSR方法至少2.72 dB。
cs.CV / 43 / 2609.29252
IronViT: Toward Efficient Generalist Visual Representation Learning
IronViT：迈向高效的通用视觉表征学习
Abstract
A generalist vision encoder must capture semantic, spatial, language-aligned, and action-relevant cues within a unified representation, yet softmax attention underlying today's most capable visual backbones becomes prohibitively expensive at high resolution. A natural attempt to address both challenges is to distill multiple specialist teachers directly into an efficient architecture. We find that directly coupling these objectives degrades representation quality, as the student must simultaneously reconcile heterogeneous capabilities and adapt them to a different token-mixing architecture. We introduce IronViT, built on a simple principle: consolidate capabilities before constraining computation. IronViT first distills complementary specialists into a softmax attention capability bridge, then progressively transfers the consolidated representation to a hybrid softmax-linear attention encoder. A purpose-built data pipeline further curates the distillation corpus for higher information density and broader domain coverage. Across recognition, retrieval, dense prediction, multimodal understanding, and robotic learning, IronViT is competitive with leading specialist and generalist vision encoders. The softmax bridge achieves the strongest aggregate performance in multimodal understanding and robotic learning among the evaluated backbones, while the hybrid encoder retains broad transfer performance with an efficiency advantage that grows with input resolution. Together, these results show that consolidating capabilities before architectural conversion can yield a generalist visual encoder without inheriting the prohibitive high-resolution cost of conventional softmax attention.
Chinese Translation
一个通用视觉编码器必须在统一的表征中捕捉语义、空间、语言对齐和与动作相关的线索，然而，当前最强大的视觉主干所依赖的softmax注意力在高分辨率下变得极其昂贵。解决这两个挑战的自然尝试是将多个专业教师直接蒸馏到一个高效的架构中。我们发现，直接将这些目标耦合会降低表征质量，因为学生必须同时调和异构能力并将其适应于不同的令牌混合架构。我们提出了IronViT，基于一个简单的原则：在限制计算之前整合能力。IronViT首先将互补的专家蒸馏成一个softmax注意力能力桥，然后逐步将整合的表征转移到一个混合的softmax-线性注意力编码器。一个专门构建的数据管道进一步策划了蒸馏语料库，以提高信息密度和更广泛的领域覆盖。在识别、检索、密集预测、多模态理解和机器人学习等任务中，IronViT与领先的专业和通用视觉编码器具有竞争力。softmax桥在评估的主干中实现了多模态理解和机器人学习方面最强的综合性能，而混合编码器在保持广泛迁移性能的同时，其效率优势随着输入分辨率的提高而增加。综合来看，这些结果表明，在架构转换之前整合能力可以产生一个通用视觉编码器，而不必继承传统softmax注意力在高分辨率下的高昂成本。
cs.CV / 44 / 2609.29256
Deep learning of longitudinal visual fields predicts glaucoma progression rate and identifies fast progressors
深度学习纵向视野预测青光眼进展速率并识别快速进展者
Abstract
Glaucoma is the leading cause of irreversible blindness, and timely identification of fast progressors is essential to prevent disability. Current practice estimates progression by ordinary least-squares regression of mean deviation (MD) on time, requiring 6--10 visual field (VF) tests over several years to obtain a reliable slope. We present GLAM (Glaucoma Longitudinal Analysis Model), a deep learning framework that ingests longitudinal Humphrey 24-2 total deviation sequences with five clinical features and predicts MD and visual field index progression rates using attention-based fusion and aleatoric uncertainty. On the open-access University of Washington Humphrey Visual Field dataset (4,276 patient-eyes), GLAM achieved an MD-rate mean absolute error of 0.139 dB yr$^{-1}$ ($R^2 = 0.927$; 73.5% reduction over a ridge baseline) and an AUC of 0.990 for fast-progressor detection. VF-only deep learning can match multimodal pipelines for progression prognostication using routinely collected perimetry alone.
Chinese Translation
青光眼是导致不可逆失明的主要原因，及时识别快速进展者对于预防残疾至关重要。目前的实践通过对平均偏差（MD）随时间变化的普通最小二乘回归来估计进展，这需要在数年内进行6到10次视野（VF）测试以获得可靠的斜率。我们提出了GLAM（青光眼纵向分析模型），这是一个深度学习框架，能够处理纵向的Humphrey 24-2总偏差序列，结合五个临床特征，并利用基于注意力的融合和随机不确定性预测MD和视野指数的进展速率。在开放获取的华盛顿大学Humphrey视觉场数据集（4,276个患者眼睛）上，GLAM实现了0.139 dB yr$^{-1}$的MD速率平均绝对误差（$R^2 = 0.927$；相较于岭回归基线减少73.5%），并在快速进展者检测中获得了0.990的AUC。仅使用视野的深度学习可以与多模态管道相匹配，利用常规收集的周边视野数据进行进展预后。
cs.CV / 45 / 2609.29292
PHOSA: Photorealistic 3D Sign Avatar Modeling and Benchmark
PHOSA：逼真3D手语头像建模与基准测试
Abstract
In this work, we focus on photorealistic sign avatar modeling, which is crucial for effective communication with the Deaf community and is characterized by complex hand gestures and nuanced facial expressions. To this end, we introduce MVSign, the first multi-view Chinese sign language dataset co-designed with Deaf experts, featuring diverse gestures and rich annotations. For precise SMPL-X annotation, we develop a hybrid fitting pipeline that produces accurate body, hand, and facial parameters and can also be applied to the monocular setting. Building on MVSign, we propose a decoupled sign avatar representation that isolates body, head, and hand components to capture complex articulations, together with a motion-aware sampling strategy to handle motion blur and balance gesture diversity. Extensive experiments demonstrate that our method achieves high-fidelity visual results on MVSign, particularly in detailed hand and facial regions, and generalizes well to in-the-wild monocular sign language videos. Project page: https://naaapi.github.io/PHOSA.
Chinese Translation
在本研究中，我们专注于逼真的手语头像建模，这对于与聋人社区的有效沟通至关重要，且其特点是复杂的手势和细腻的面部表情。为此，我们推出了MVSign，这是第一个与聋人专家共同设计的多视角中文手语数据集，包含多样的手势和丰富的注释。为了实现精确的SMPL-X注释，我们开发了一种混合拟合管道，能够生成准确的身体、手部和面部参数，并且也可以应用于单目设置。在MVSign的基础上，我们提出了一种解耦的手语头像表示方法，该方法将身体、头部和手部组件分离，以捕捉复杂的发音，同时采用一种运动感知的采样策略来处理运动模糊并平衡手势多样性。大量实验表明，我们的方法在MVSign上实现了高保真度的视觉效果，特别是在细致的手部和面部区域，并且在自然环境中的单目手语视频上具有良好的泛化能力。项目页面：https://naaapi.github.io/PHOSA。
cs.CV / 46 / 2609.29329
Hyperbolic Multimodal Continual Learning: A Closest-Admissible Solution
双曲多模态持续学习：最接近可接受解
Abstract
Existing continual-learning methods protect parameters, replayed examples, or Euclidean feature subspaces. When applied to hyperbolic multimodal models, they do not explicitly preserve the Lorentz geometry that jointly encodes within-modality similarity, cross-modal correspondence, and semantic hierarchy; sequential updates can therefore retain task scores while still distorting previously learned relations. We address this gap with Hyperbolic Multimodal Continual Learning (HMCL). We show that preserving the old multimodal geometry amounts to restricting all modalities to one shared hyperbolic isometry, which induces a family of admissible first-order parameter changes. We formulate a joint closest-admissible (CA) correction that retains the shared rotation best matching the candidate modal updates; its minimal-rotation (MR) special case fixes this rotation to zero. Both variants correct the displacement realized by AdamW, and task anchoring bounds within-task accumulation while preserving learning freedom. Across a unified 16-task classification-retrieval stream with three hyperbolic backbones, HMCL improves final performance and backward transfer over sequential fine-tuning and four continual-learning baselines; HMCL-CA gives the highest Overall score on every backbone. A modality-extended stream confirms the retrieval gains. Representation analyses find 81.2 to 95.5 percent less radial, angular, cross-modal, and paired-distance drift; ImageNet-WordNet results show better semantic ancestry and radial hierarchy.
Chinese Translation
现有的持续学习方法保护参数、重放示例或欧几里得特征子空间。当应用于双曲多模态模型时，它们并未明确保留洛伦兹几何，这种几何共同编码了模内相似性、跨模态对应关系和语义层次；因此，顺序更新可以保留任务得分，同时仍然扭曲先前学习的关系。我们通过双曲多模态持续学习（Hyperbolic Multimodal Continual Learning, HMCL）来解决这一问题。我们表明，保留旧的多模态几何相当于将所有模态限制为一个共享的双曲等距变换，这会引入一系列可接受的一阶参数变化。我们制定了一种联合最接近可接受（closest-admissible, CA）修正，保留与候选模态更新最佳匹配的共享旋转；其最小旋转（minimal-rotation, MR）特例将该旋转固定为零。这两种变体都修正了由 AdamW 实现的位移，并且任务锚定在保留学习自由度的同时限制了任务内的累积。在一个统一的16任务分类-检索流中，结合三种双曲骨干网络，HMCL 在最终性能和向后迁移方面优于顺序微调和四个持续学习基线；HMCL-CA 在每个骨干网络上都给出了最高的总体得分。一个扩展模态的流验证了检索增益。表示分析发现径向、角度、跨模态和配对距离漂移减少了81.2%到95.5%；ImageNet-WordNet的结果显示了更好的语义谱系和径向层次。
cs.CV / 47 / 2609.29334
A Study of the Limits of Collaborative DCT-Based Image Denoising via Interpretable Neural Networks
基于可解释神经网络的协作DCT图像去噪的极限研究
Abstract
Image denoising remains a fundamental problem in image restoration, with applications in photography, biomedical, and scientific imaging. Modern deep neural networks achieve strong performance by learning powerful image priors, but often rely on large black-box models with limited interpretability. In contrast, DCT-based sliding-window and collaborative filtering methods such as BM3D offer clear algorithmic structure, but depend on handcrafted and non-differentiable operations. This work studies how far such structured collaborative filtering principles can be pushed when reformulated as trainable models. We introduce DeepBM3D, a compact fully differentiable architecture that combines non-local patch grouping, DCT-domain filtering, and multi-stage refinement within a BM3D-inspired pipeline. Lightweight convolutional feature extractors guide patch grouping, while filtering is performed through learned Wiener weights in the DCT domain. Experiments show that DeepBM3D improves over classical and hybrid baselines, remains competitive with FFDNet at low and moderate noise levels, and performs particularly well on repetitive textures.
Chinese Translation
图像去噪仍然是图像恢复中的一个基本问题，广泛应用于摄影、生物医学和科学成像。现代深度神经网络通过学习强大的图像先验来实现出色的性能，但通常依赖于大型黑箱模型，缺乏可解释性。相比之下，基于DCT的滑动窗口和协作过滤方法，如BM3D，提供了清晰的算法结构，但依赖于手工设计和不可微分的操作。本研究探讨了当这些结构化的协作过滤原则被重新表述为可训练模型时，可以推向多远。我们引入了DeepBM3D，这是一种紧凑的完全可微分架构，结合了非局部补丁分组、DCT域过滤和多阶段细化，构建于BM3D启发的管道中。轻量级卷积特征提取器指导补丁分组，而过滤则通过在DCT域中学习的维纳权重进行。实验表明，DeepBM3D在经典和混合基线之上有所改进，在低噪声和中等噪声水平下与FFDNet保持竞争力，并在重复纹理上表现尤为出色。
cs.CV / 48 / 2609.29347
SEE Challenge 2026: Event-Guided Brightness Adjustment Across a Broad Illumination Range
SEE挑战2026：跨广泛照明范围的事件引导亮度调整
Abstract
Event cameras provide a high dynamic range and preserve brightness-change cues in lighting conditions where conventional RGB frames may be noisy or saturated. To benchmark event-guided restoration across a broad illumination range, we organized the SEE Challenge 2026 with the Event-Based Multimodal Vision Workshop at ECCV 2026. The task conditions restoration on one or more RGB frames, synchronized events, and a scalar target-brightness statistic provided by the organizers. It uses SEE-600K, which contains 610,126 image-event observations from 202 real-world scenes spanning low-light, normal-light, and high-light conditions with illumination variations of up to 1,000$\times$. The challenge follows an open-system protocol: participants may use different temporal contexts, architectures, pretrained weights, test-time augmentation, and post-processing strategies. PSNR determines the ranking, and SSIM is reported as a secondary metric. Around 70 teams registered interest and 15 valid CodaBench submissions were received. Six distinct teams completed organizer-side identity and technical verification, provided method descriptions, checkpoints, inference code, and instructions, and are included in the verified open-system ranking reported here. Beyond the ranking, this report analyzes exposure subsets, semantically distinct test cases, a shared failure pattern, system design choices, and inference strategies. The top systems obtain closely spaced average scores, while the best-performing method varies across cases and metrics; under severe underexposure, all verified systems retain visible local errors.
Chinese Translation
事件相机提供了高动态范围，并在传统RGB帧可能出现噪声或饱和的照明条件下保留亮度变化线索。为了在广泛的照明范围内基准测试事件引导的恢复，我们在2026年ECCV的事件驱动多模态视觉研讨会上组织了SEE挑战2026。该任务条件下的恢复涉及一个或多个RGB帧、同步事件以及由组织者提供的标量目标亮度统计数据。它使用SEE-600K数据集，该数据集包含来自202个真实场景的610,126个图像事件观测，涵盖低光、正常光和高光条件，照明变化高达1,000倍。该挑战遵循开放系统协议：参与者可以使用不同的时间上下文、架构、预训练权重、测试时增强和后处理策略。PSNR用于确定排名，SSIM作为次要指标报告。约有70个团队表示感兴趣，收到了15个有效的CodaBench提交。六个不同的团队完成了组织方的身份和技术验证，提供了方法描述、检查点、推理代码和说明，并被纳入此处报告的经过验证的开放系统排名。除了排名之外，本报告还分析了曝光子集、语义上不同的测试案例、共享的失败模式、系统设计选择和推理策略。顶级系统的平均得分相近，而最佳表现的方法在不同案例和指标中有所不同；在严重欠曝光的情况下，所有经过验证的系统均保留可见的局部错误。
cs.CV / 49 / 2609.29350
Learning a Flow to Self-Supervised Representations
学习流以实现自监督表示
Abstract
Explicit geometric references offer a direct way to structure self-supervised representations. Existing adversarial distribution-matching formulations, however, require costly encoder-critic optimization. We introduce Flow-Based Distribution Matching (FBDM), a non-adversarial framework that learns this reference-directed geometry through spherical conditional velocity regression. An ETF-inspired reference allows its number of components K' to exceed the auxiliary flow dimension d* while retaining structured geometric separation. We assign both augmented views of each image to the same target, while limiting how many images each reference center can receive. An explicit alignment loss further pulls the two views' representations closer together. Experiments across benchmarks ranging from CIFAR to ImageNet show that FBDM achieves performance nearly on par with DM and remains competitive with existing SSL methods. Matched training-cost comparisons show a 1.48- to 1.83-fold speedup over DM with a negligible increase in GPU memory usage. We also provide a theoretical explanation for the usefulness of the learned representations: under stated conditions, we bound the downstream misclassification rate in terms of the FBDM pretraining loss.
Chinese Translation
显式几何参考为结构化自监督表示提供了一种直接的方法。然而，现有的对抗性分布匹配形式需要昂贵的编码器-评论家优化。我们提出了一种基于流的分布匹配（Flow-Based Distribution Matching, FBDM）框架，它通过球面条件速度回归学习这种参考导向的几何结构。受ETF启发的参考使得其组件数量K'可以超过辅助流维度d*，同时保持结构化的几何分离。我们将每个图像的两个增强视图分配给同一目标，同时限制每个参考中心可以接收的图像数量。显式对齐损失进一步拉近了两个视图的表示。跨越从CIFAR到ImageNet的基准实验表明，FBDM的性能几乎与DM相当，并且在现有自监督学习（SSL）方法中保持竞争力。匹配的训练成本比较显示，FBDM在GPU内存使用量几乎不增加的情况下，相较于DM实现了1.48到1.83倍的加速。我们还提供了对学习表示有用性的理论解释：在特定条件下，我们将下游错误分类率与FBDM预训练损失进行了界定。
cs.CV / 50 / 2609.29358
Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models
视觉-语言模型的领域重心调整与置信加权先验校准
Abstract
Vision-language models such as CLIP achieve strong zero-shot classification, yet under distribution shift, visual embeddings drift from fixed text embeddings. Training-free calibration avoids the per-sample optimization of prompt learning, but prior feature calibration gives each image the full bias of one hard cluster. We propose Domain Recentering with Confidence Calibration (DRC), a training-free method adapting CLIP from a set of unlabeled target images. DRC fits a Gaussian mixture once and subtracts from each embedding a posterior-weighted average of component means. It then removes residual class preference with a log-prior correction, estimating the prior from confidence-weighted predictions. Among compared methods, DRC achieves the highest average accuracy on cross-domain datasets, exceeding zero-shot CLIP by 4.13 and 5.07 points with ViT-B/16 and ResNet-50, with gains over CLIP also holding under ImageNet distribution shifts.
Chinese Translation
视觉-语言模型如 CLIP 在零样本分类中表现出色，但在分布转移下，视觉嵌入与固定文本嵌入之间存在漂移。无训练校准避免了逐样本的提示学习优化，但先验特征校准使得每个图像承受一个硬聚类的全部偏差。我们提出了一种训练无关的方法——领域重心调整与置信校准（Domain Recentering with Confidence Calibration, DRC），该方法从一组未标记的目标图像中适应 CLIP。DRC 一次拟合高斯混合模型，并从每个嵌入中减去后验加权的组件均值平均值。然后，通过对数先验修正去除残余类别偏好，从置信加权的预测中估计先验。在比较的方法中，DRC 在跨领域数据集上实现了最高的平均准确率，使用 ViT-B/16 和 ResNet-50 分别超过零样本 CLIP 4.13 和 5.07 个百分点，并且在 ImageNet 分布转移下也保持了对 CLIP 的提升。
cs.CV / 51 / 2609.29373
Shadow Reduction in Ultrasound Imaging Using Differentiable Simulation and Radiance Field Decomposition
利用可微分仿真和辐射场分解减少超声成像中的阴影
Abstract
Acoustic shadows from bone and other highly attenuating tissues obscure clinically important structures in ultrasound. In fetal brain imaging, skull-induced artefacts disproportionately degrade the hemisphere closer to the transducer (proximal), limiting symmetric assessment of the two hemispheres. Existing correction methods require raw scanner data, impose restrictive assumptions on tissue properties, or rely on generative models that may hallucinate anatomy. We present RFlash, a physics-informed post-processing method that decomposes beamformed ultrasound images into explicit attenuation and scatter-intensity maps using a differentiable radiance-field formulation of image formation. Attenuation-adaptive re-rendering then removes the dependence of the signal at each depth on the intervening tissue, equivalent to virtually advancing the transducer into the tissue. Across 1,261 3D fetal brain volumes, 143 real 2D curvilinear abdominal scans, and 1,200 simulated 2D linear-probe liver scans, RFlash reduces shadow-related intensity differences more effectively than classical Hughes-Duck attenuation correction. For a gestational-age model trained on the distal hemisphere (further from the transducer) and applied to the proximal hemisphere, prediction error decreases by 5.1 days (40%) relative to the original images. The estimated attenuation maps also yield shadow-confidence maps that improve random-forest bone-shadow segmentation over the image alone and receive greater SHAP importance than an existing neural confidence-map baseline, suggesting greater physical consistency. RFlash requires neither hardware modification nor access to raw scanner data and supports 2D and 3D acquisitions with linear and curvilinear probes, making it widely applicable allowing clinicians to use our method on their already acquired scanners and images.
Chinese Translation
骨骼和其他高衰减组织产生的声学阴影遮挡了超声中的临床重要结构。在胎儿脑成像中，颅骨引起的伪影不成比例地降低了靠近探头（近端）的半球的成像质量，限制了对两个半球的对称评估。现有的校正方法需要原始扫描仪数据，对组织特性施加限制性假设，或依赖于可能会幻觉解剖结构的生成模型。我们提出了RFlash，这是一种物理信息驱动的后处理方法，它利用可微分辐射场图像形成的公式，将波束形成的超声图像分解为明确的衰减和散射强度图。衰减自适应重渲染随后消除了每个深度信号对介入组织的依赖，相当于将探头虚拟推进到组织内部。在1261个3D胎儿脑体积、143个真实的2D曲线腹部扫描和1200个模拟的2D线性探头肝脏扫描中，RFlash比经典的Hughes-Duck衰减校正更有效地减少了与阴影相关的强度差异。对于一个在远端半球（离探头更远）上训练的孕周模型并应用于近端半球，预测误差相较于原始图像减少了5.1天（40%）。估计的衰减图还生成了阴影置信图，改善了随机森林骨骼阴影分割的效果，相较于仅使用图像的情况，其SHAP重要性也高于现有的神经置信图基线，表明更大的物理一致性。RFlash既不需要硬件修改，也不需要访问原始扫描仪数据，支持使用线性和曲线探头的2D和3D采集，使其广泛适用，允许临床医生在已有的扫描仪和图像上使用我们的方法。
cs.CV / 52 / 2609.29376
A Hybrid CNN--State-Space--Attention Backbone with Joint-Embedding Predictive Pretraining for 12-Lead ECG Classification
一种混合CNN-状态空间-注意力骨干网络与联合嵌入预测预训练用于12导联心电图分类
Abstract
Automatic 12-lead electrocardiogram (ECG) classification requires representations that jointly capture local waveform morphology, long-range temporal dynamics, and cross-lead dependencies, yet integrating these properties within a single efficient architecture remains challenging. This paper introduces a hybrid CNN-SSM-Attention backbone for 12-lead ECG classification. A convolutional stem performs early waveform tokenization and temporal reduction, mixed state-space and depthwise-convolutional blocks model temporal dynamics and local morphology, and a late self-attention stage enables global token interaction at reduced resolution. To improve transfer from unlabeled data, we further develop an ECG-oriented Joint-Embedding Predictive Pretraining (JEPA) framework. Unlike ViT-based JEPA methods that mask patch tokens before the encoder, the proposed method samples span masks at the latent temporal resolution and projects them back to the waveform domain, then predicts clean latent targets from a momentum encoder without waveform reconstruction. Experiments on CPSC2018, Chapman-Shaoxing, and PTB-XL, with pretraining on approximately 350K unlabeled CODE-15 recordings, show that the proposed backbone provides strong supervised baselines under a compact parameter budget. JEPA pretraining further improves transfer, particularly in reduced-label settings and under both full fine-tuning and LoRA-based adaptation. Code: https://github.com/yakoubbazi/Hybrid_ECG_Jepa
Chinese Translation
自动化的12导联心电图（ECG）分类需要能够共同捕捉局部波形形态、长程时间动态和导联间依赖关系的表征，但在单一高效架构中整合这些特性仍然具有挑战性。本文提出了一种用于12导联ECG分类的混合CNN-SSM-注意力骨干网络。卷积干部分别执行早期波形标记和时间降维，混合状态空间和深度卷积块建模时间动态和局部形态，而后期自注意力阶段则在降低分辨率的情况下实现全局标记交互。为了改善从未标记数据的迁移，我们进一步开发了一种心电图导向的联合嵌入预测预训练（JEPA）框架。与在编码器之前对补丁标记进行掩蔽的基于ViT的JEPA方法不同，所提出的方法在潜在时间分辨率上采样跨度掩蔽，并将其投影回波形域，然后从动量编码器预测干净的潜在目标，而无需波形重建。在CPSC2018、Chapman-Shaoxing和PTB-XL上的实验，以及在大约350K未标记的CODE-15录音上进行的预训练，表明所提出的骨干网络在紧凑的参数预算下提供了强大的监督基线。JEPA预训练进一步改善了迁移，特别是在减少标签设置下以及在完全微调和基于LoRA的适应中。代码：https://github.com/yakoubbazi/Hybrid_ECG_Jepa
cs.CV / 53 / 2609.29384
Segment-Level Risk Discovery in Online Handwriting for Alzheimer's Disease Detection
在线手写中的段级风险发现用于阿尔茨海默病检测
Abstract
Online handwriting provides a non-invasive and low-cost behavioral biomarker for Alzheimer's disease (AD) detection, as it reflects both cognitive planning and fine motor control. Existing handwriting-based AD detection methods usually rely on global trajectory features or whole-sample representations, which can be strongly affected by individual writing style, task-specific variation, and acquisition noise. In this paper, we propose NormPaST-Risk, a healthy-normative Paper-Air selective trajectory state-space risk network for interpretable AD detection from online handwriting. Instead of treating the entire trajectory as a single holistic representation, our method reformulates AD handwriting detection as local disease-relevant segment discovery. Specifically, a multi-scale temporal encoder captures stroke dynamics at different temporal resolutions, while a selective Paper-Air state-space encoder models long-range handwriting progression and distinguishes on-paper motor execution from in-air planning and transition behaviors. To explicitly characterize abnormal deviations, a healthy normative branch learns normal handwriting dynamics from healthy controls, and a task-aware multi-expert segment-risk module estimates segment-level AD risk calibrated by hidden-state changes and normative deviations. A weakly supervised segment-level objective further enables high-risk segment discovery without manual segment annotations. Experiments on the DARWIN benchmark demonstrate that the proposed framework achieves superior AD/HC classification performance compared with existing methods. Moreover, the discovered high-risk segments can be projected back to the original handwriting trajectory, providing interpretable evidence associated with AD-related handwriting variations.
Chinese Translation
在线手写提供了一种非侵入性和低成本的行为生物标志物，用于阿尔茨海默病（AD）检测，因为它反映了认知规划和精细运动控制。现有的基于手写的AD检测方法通常依赖于全局轨迹特征或整体样本表示，这些方法容易受到个体书写风格、任务特定变化和采集噪声的强烈影响。本文提出了NormPaST-Risk，一种健康规范的Paper-Air选择性轨迹状态空间风险网络，用于从在线手写中进行可解释的AD检测。我们的方法并不将整个轨迹视为单一的整体表示，而是将AD手写检测重新构建为局部与疾病相关的段发现。具体而言，多尺度时间编码器捕捉不同时间分辨率下的笔画动态，而选择性Paper-Air状态空间编码器则建模长距离的手写进展，并区分纸上运动执行与空中规划和过渡行为。为了明确表征异常偏差，健康规范分支从健康对照中学习正常的手写动态，而任务感知的多专家段风险模块则通过隐状态变化和规范偏差来估计段级AD风险。弱监督的段级目标进一步使得高风险段的发现无需手动段注释。在DARWIN基准上的实验表明，所提出的框架在AD/HC分类性能上优于现有方法。此外，发现的高风险段可以投影回原始手写轨迹，提供与AD相关的手写变化的可解释证据。
cs.CV / 54 / 2609.29387
When Misalignment Becomes Supervision: Structured Label Noise in Supervised Synthetic CT Generation
当错位成为监督：监督合成CT生成中的结构化标签噪声
Abstract
Supervised synthetic CT (sCT) generation is commonly trained and evaluated as voxel-wise regression against registered reference CT images. In practice, MRI-CT and CBCT-CT pairs are aligned through registration procedures that leave residual misalignments. These residuals are not independent intensity noise but spatially coherent geometric discrepancies that act as structured label noise. We investigate how this registration-induced bias affects supervised MRI-to-CT and CBCT-to-CT synthesis on 1,784 paired patients covering five anatomical regions. Voxel-wise scores strongly depend on the consistency between the registration used to build the training targets and the one used for evaluation: models score best when both conventions match, showing that networks partly learn the geometric convention of the registration pipeline and that standard metrics reward it. Training on more anatomically consistent registrations reduces prediction variability and improves out-of-distribution robustness, and CT-only controls show that registration alone produces metric errors in the range of top challenge submissions. To mitigate the limits of voxel-wise supervision, we introduce a perceptual loss computed in the feature space of a pretrained Segment Anything encoder. Compared with MAE-only and VGG-based objectives, it improves downstream segmentation and yields sharper, more structurally coherent sCT. Perceptual and voxel-wise metrics disagree under imperfect alignment and agree when the evaluation geometry is reliable. These results identify registration-induced bias as a central confounder in supervised sCT generation and argue for complementing voxel-wise agreement with anatomy-oriented evaluation criteria.
Chinese Translation
监督合成CT（sCT）生成通常作为与注册参考CT图像的体素级回归进行训练和评估。在实际操作中，MRI-CT和CBCT-CT配对通过注册程序对齐，但会留下残余的错位。这些残余并不是独立的强度噪声，而是空间上相干的几何差异，表现为结构化标签噪声。我们研究了这种注册引起的偏差如何影响1,784对患者在五个解剖区域上的监督MRI到CT和CBCT到CT的合成。体素级评分强烈依赖于用于构建训练目标的注册与用于评估的注册之间的一致性：当两者匹配时，模型得分最佳，表明网络部分学习了注册流程的几何约定，而标准指标对此给予了奖励。在更具解剖一致性的注册上进行训练可以减少预测的变异性并提高超出分布的鲁棒性，而仅使用CT的对照显示，单靠注册会产生与顶级挑战提交相当的指标误差。为了减轻体素级监督的局限性，我们引入了一种在预训练的Segment Anything编码器的特征空间中计算的感知损失。与仅使用MAE和基于VGG的目标相比，它改善了下游分割效果，并产生了更清晰、更具结构一致性的sCT。在不完美对齐的情况下，感知和体素级指标存在不一致，而在评估几何可靠时则一致。这些结果将注册引起的偏差识别为监督sCT生成中的一个主要混杂因素，并主张用面向解剖的评估标准来补充体素级一致性。
cs.CV / 55 / 2609.29433
Detecting Glaucoma Across Multi-ethnic Myopic and Non-Myopic Populations Using an Uncertainty-Aware Vision Transformer: A Multicentre Model Development and Validation Study
使用不确定性感知视觉变换器在多民族近视和非近视人群中检测青光眼：一项多中心模型开发与验证研究
Lavanya, Raghavan, Feng, Yangqin, Quek, Ten Cheer, Hoang, Quan V., Poon, Linda Yi-Chieh, Jonas, Jost B., Wang, Ya Xing, Nangia, Vinay, Jeoung, Jin Wook, Park, Sehie, Kim, SoYeon, Xu, Benjamin Y, Munimadugu, Sreenidhi Iyengar, Mitchell, Paul, Liew, Gerald, Suwan, Yanin, Hong-amata, Jirayu, Thakur, Sahil, Nongipur, Monisha E, Wong, Tina, Husain, Rahat, Rui, Ng Si, Syn, Yamon, Lo, Phey Feng, Qiang, Nicholas Tan Yi, Hussain, Shaista, Lei, Xiaofeng, Da Soh, Zhi, Yu, Marco, Hamzah, Haslina, Wang, Zizhou, Wang, Yan, Zhen, Liangli, Xu, Xinxing, Wong, Tien-Yin, Aung, Tin, Chong, Rachel S, Liu, Yong, Cheng, Ching-Yu
Abstract
Background: Artificial intelligence (AI)-based glaucoma detection from colour fundus photographs (CFP) offers scalable screening, but performance may decline on external datasets because of differences in ground-truth definitions, populations, and coexisting conditions such as high myopia (HM). We developed and validated a Vision Transformer-based deep learning (DL) model for glaucoma detection across multi-ethnic cohorts with and without HM. Methods: A ViT-B/16 model with predictive uncertainty estimation was developed using 56,483 CFPs (57.1% with myopia; 14.4% with HM). Glaucoma labels were standardised using clinical, imaging, and perimetry data. The model was validated on 16 independent datasets across three continents, including four datasets with explicit HM labels. Findings: Internal AUROC was 98.7% (95% CI 98.2-99.1%), with sensitivity 94.5% and specificity 97.3%. Across 16 external datasets from eight countries, AUROCs ranged from 86.4% to 99.6%. In HM eyes, internal AUROC was 97.8% (95% CI 96.1-99.2%), with sensitivity 94.8% and specificity 93.7%. External HM AUROCs were 86.5% in the Beijing Eye Study and 93.3%, 91.8%, and 85.5% in hospital-based datasets from Taiwan, Thailand, and South Korea. In an exploratory HM clinical evaluation, the model had higher CFP-only diagnostic accuracy than ophthalmologists and trained graders (92.0% vs 70.0%; p=0.008) and performed comparably to glaucoma specialists using full clinical information. Interpretation: The model showed robust glaucoma detection across myopic and non-myopic multi-ethnic populations and may support AI-assisted screening in settings with high HM prevalence.
Chinese Translation
背景：基于人工智能（AI）的青光眼检测通过彩色眼底照片（CFP）提供可扩展的筛查，但由于真实情况定义、人口和共存条件（如高度近视（HM））的差异，性能在外部数据集上可能下降。我们开发并验证了一种基于视觉变换器的深度学习（DL）模型，用于在有和没有HM的多民族人群中检测青光眼。方法：使用56,483张CFP（57.1%为近视；14.4%为HM）开发了具有预测不确定性估计的ViT-B/16模型。青光眼标签通过临床、影像和视野检查数据进行了标准化。该模型在三个大洲的16个独立数据集上进行了验证，包括四个具有明确HM标签的数据集。结果：内部受试者工作特征曲线下面积（AUROC）为98.7%（95% CI 98.2-99.1%），灵敏度为94.5%，特异性为97.3%。在来自八个国家的16个外部数据集中，AUROC范围为86.4%至99.6%。在HM眼中，内部AUROC为97.8%（95% CI 96.1-99.2%），灵敏度为94.8%，特异性为93.7%。外部HM AUROC在北京眼科研究中为86.5%，在来自台湾、泰国和韩国的医院数据集中分别为93.3%、91.8%和85.5%。在一项探索性HM临床评估中，该模型的CFP单独诊断准确率高于眼科医生和训练评估者（92.0%对70.0%；p=0.008），并且在使用完整临床信息时与青光眼专家的表现相当。解释：该模型在近视和非近视的多民族人群中表现出稳健的青光眼检测能力，并可能支持在高度HM流行的环境中进行AI辅助筛查。
cs.CV / 56 / 2609.29443
Pose Adaptive Dynamic FiLM Modulation for Visual Speech Recognition
用于视觉语音识别的姿态自适应动态 FiLM 调制
Abstract
Head-pose variation introduces substantial appearance transformations in visual speech recognition (VSR), making pose-aware feature modulation desirable. However, performance degradation and unwanted feature interactions may result from using numerous Feature-wise Linear Modulation (FiLM) circuits with fixed modulation intensity. We propose a Pose Adaptive Dynamic FiLM framework with a Dynamic Residual FiLM (DR-FiLM) modulator that predicts input-dependent weights to adaptively control the strength of pose-conditioned modulation. Experiments on LRS2 and LRS3 demonstrate that unweighted multi-pathway modulation substantially degrades phoneme recognition, increasing PER to 20.33% and 29.42%, respectively, compared with 16.20% and 20.96% for the single ResFiLM configuration. In contrast, the proposed DR-FiLM with dynamic Deep-Res weighting reduces PER to 15.74% on LRS2 and 23.91% on LRS3, substantially mitigating the adverse effects of unweighted modulation. The analysis of the learned weights further reveals a consistent tendency to assign greater weight to the deeper FiLM pathway as head-pose variation increases. These results show that merging pose-conditioned FiLM circuits is more efficient when the modulation strength is dynamically controlled.
Chinese Translation
头部姿态变化在视觉语音识别（VSR）中引入了显著的外观变换，因此需要姿态感知特征调制。然而，使用多个具有固定调制强度的特征线性调制（FiLM）电路可能导致性能下降和不必要的特征交互。我们提出了一种姿态自适应动态 FiLM 框架，结合动态残差 FiLM（DR-FiLM）调制器，该调制器预测输入依赖的权重，以自适应地控制姿态条件调制的强度。在 LRS2 和 LRS3 上的实验表明，无权重的多路径调制显著降低了音素识别性能，使得音素错误率（PER）分别增加到 20.33% 和 29.42%，而单一 ResFiLM 配置的 PER 分别为 16.20% 和 20.96%。相比之下，所提出的具有动态深度残差加权的 DR-FiLM 将 LRS2 的 PER 降低到 15.74%，LRS3 的 PER 降低到 23.91%，显著减轻了无权重调制的不利影响。对学习权重的分析进一步揭示，随着头部姿态变化的增加，赋予更深 FiLM 路径更大权重的趋势是一致的。这些结果表明，当调制强度动态控制时，合并姿态条件 FiLM 电路更为高效。
cs.CV / 57 / 2609.29447
Frame-to-Panorama Localization and Context-Aware Sampling for Scene-Specific Ship Detection in a Smart Marina Testbed
帧到全景定位及场景特定船舶检测的上下文感知采样在智能码头测试平台中的应用
Abstract
Smart maritime infrastructures provide continuous access to heterogeneous sensing streams, enabling repeated experimentation, digital-twin development, and AI-based maritime services. However, sensing hardware alone is not sufficient for scene-specific model development: historical video streams must also be spatially indexed, contextualized, and reduced to informative subsets for annotation. This paper presents a frame-to-panorama localization and context-aware sampling pipeline for ship detection in historical PTZ maritime video lacking reliable pan, tilt, and zoom metadata. The main contribution is an end-to-end data-curation approach that recovers camera-view information from historical PTZ video and combines it with environmental context and visual diversity to construct compact, scene-specific training sets. Specifically, frames are localized on a reference panorama using SuperPoint and LightGlue, enriched with weather and solar-state metadata, and selected through diversity sampling to preserve variation across camera view and environmental conditions. A second context-aware stage targets under-represented distant-vessel cases near the horizon using tile-level visual embeddings and Gaussian Mixture Model clustering. Applied within the CMMI MDigi-I Smart Marina testbed, the proposed pipeline reduces 40,718 candidate frames to 220 images for annotation, corresponding to a 99.5% reduction. A YOLO26-m detector fine-tuned on this subset achieves a mean AP50 of 94.78% $\pm$ 0.51% and a mean AP50-95 of 75.10% $\pm$ 1.73% under sequence-grouped five-fold cross-validation. These results demonstrate that highly redundant infrastructure video streams can be transformed into compact, spatially and contextually diverse training sets for scene-specific detector adaptation while substantially reducing annotation effort.
Chinese Translation
智能海事基础设施提供对异构传感流的持续访问，使得重复实验、数字双胞胎开发和基于人工智能的海事服务成为可能。然而，仅靠传感硬件不足以进行场景特定模型的开发：历史视频流还必须进行空间索引、上下文化，并缩减为可供注释的信息子集。本文提出了一种用于历史PTZ（平移、俯仰和变焦）海事视频的帧到全景定位和上下文感知采样管道，以实现船舶检测，尽管该视频缺乏可靠的平移、俯仰和变焦元数据。主要贡献在于一种端到端的数据策划方法，该方法从历史PTZ视频中恢复摄像机视图信息，并将其与环境上下文和视觉多样性相结合，以构建紧凑的场景特定训练集。具体而言，使用SuperPoint和LightGlue在参考全景上定位帧，并通过天气和太阳状态元数据进行丰富，然后通过多样性采样选择，以保留摄像机视图和环境条件之间的变化。第二个上下文感知阶段针对接近地平线的代表性不足的远处船舶案例，利用瓦片级视觉嵌入和高斯混合模型聚类。在CMMI MDigi-I智能码头测试平台中应用该管道，将40,718个候选帧减少到220张图像进行注释，对应于99.5%的减少。经过微调的YOLO26-m检测器在该子集上实现了94.78% ± 0.51%的平均AP50和75.10% ± 1.73%的平均AP50-95，在序列分组的五折交叉验证下。这些结果表明，高度冗余的基础设施视频流可以转化为紧凑的、空间和上下文多样的训练集，以适应场景特定的检测器，同时显著减少注释工作量。
cs.CV / 58 / 2609.29456
Dense Coverage, Sparse Refinement: Byte-Constrained Cooperative Perception
密集覆盖，稀疏精炼：字节受限的协同感知
Abstract
Collaborative perception improves autonomous perception by sharing intermediate Bird's-Eye-View (BEV) features across connected agents, but dense feature exchange is difficult to deploy under strict Vehicle-to-Everything (V2X) bandwidth limits. Existing efficient methods typically either compress the full feature map uniformly, spending bits on low-value background, or sparsify communication, risking the loss of useful context. We propose a coverage-refinement design for byte-constrained cooperative perception: each agent transmits a highly compressed coarse layer over the full BEV map and allocates the remaining budget to selected high-resolution patches. A Task-Aware Benefit Selector ranks cells by estimated downstream utility, enabling deterministic budgeted refinement and zero-retraining adaptation to changing bandwidth. The receiver reconstructs a dense BEV tensor compatible with standard fusion modules. Experiments on DAIR-V2X and OPV2V show strong accuracy-payload trade-offs at kilobyte-scale budgets. On DAIR-V2X, our method reaches 0.60
[email protected] at only 1.87 KB per non-ego agent, compared with 0.52 at 4.61 KB for uniform SimVQ compression. Controlled diagnostics further show that the gain arises from coverage-refinement allocation rather than quantization alone. Code will be published.
Chinese Translation
协同感知通过在连接的代理之间共享中间鸟瞰图（Bird's-Eye-View, BEV）特征来提升自主感知能力，但在严格的车对一切（Vehicle-to-Everything, V2X）带宽限制下，密集特征交换的部署十分困难。现有的高效方法通常要么均匀压缩完整特征图，浪费位于低价值背景上的比特，要么稀疏通信，冒着丢失有用上下文的风险。我们提出了一种针对字节受限协同感知的覆盖-精炼设计：每个代理在完整的BEV图上发送一个高度压缩的粗层，并将剩余预算分配给选定的高分辨率补丁。任务感知收益选择器（Task-Aware Benefit Selector）根据估计的下游效用对单元进行排名，从而实现确定性的预算精炼和对带宽变化的零重训练适应。接收方重建与标准融合模块兼容的密集BEV张量。在DAIR-V2X和OPV2V上的实验表明，在千字节级预算下，准确性与负载之间的权衡表现出色。在DAIR-V2X上，我们的方法在每个非自我代理仅需1.87 KB时达到了0.60
[email protected]，而均匀的SimVQ压缩则需4.61 KB时仅达到0.52。受控诊断进一步表明，增益源于覆盖-精炼分配，而非单纯的量化。代码将会发布。
cs.CV / 59 / 2609.29457
Industrial Anomaly Detection via Defect-Grounded Reasoning in Visual Latent Space
基于缺陷驱动推理的视觉潜在空间工业异常检测
Abstract
Industrial anomaly detection (IAD) is evolving beyond conventional detection and localization toward multimodal inspection systems that can describe, explain, and reason about fine-grained defects. Although recent multimodal large language model (MLLM)-based methods improve anomaly understanding through textual reasoning and visual guidance, they face two limitations in fine-grained inspection. First, their visual refinement often requires iteratively revisiting local image regions or augmenting with additional tools. Second, the resulting local defect evidence may not be reliably preserved throughout subsequent reasoning. To address these, we propose Anomaly-LR, a defect-grounded latent reasoning framework that first forms a global understanding of the input and then progressively refines anomaly-relevant representations directly in the visual latent space. We further construct IAD-LR-22K, the first IAD instruction dataset designed for latent reasoning, containing 22,228 image-question instances from 4,523 industrial images, with global textual reasoning traces and region-level visual annotations. Extensive experiments show that Anomaly-LR achieves state-of-the-art performance among comparable-scale methods across multiple IAD benchmarks, without requiring external references or tools. The code and data will be released at https://github.com/Yen666/Anomaly-LR.
Chinese Translation
工业异常检测（IAD）正从传统的检测和定位向多模态检测系统发展，这些系统能够描述、解释并推理关于细粒度缺陷的信息。尽管近期基于多模态大语言模型（MLLM）的方法通过文本推理和视觉指导提高了异常理解，但在细粒度检测中仍面临两个限制。首先，它们的视觉细化通常需要反复访问局部图像区域或借助额外工具进行增强。其次，生成的局部缺陷证据可能在后续推理过程中无法可靠保留。为了解决这些问题，我们提出了Anomaly-LR，这是一种基于缺陷驱动的潜在推理框架，首先形成对输入的全局理解，然后在视觉潜在空间中逐步细化与异常相关的表示。此外，我们构建了IAD-LR-22K，这是第一个为潜在推理设计的IAD指令数据集，包含来自4,523张工业图像的22,228个图像-问题实例，配有全局文本推理轨迹和区域级视觉注释。大量实验表明，Anomaly-LR在多个IAD基准测试中在可比规模的方法中实现了最先进的性能，无需外部参考或工具。代码和数据将发布在 https://github.com/Yen666/Anomaly-LR。
cs.CV / 60 / 2609.29460
AgriCountDINO: Parameter-Efficient Exemplar-Guided Counting and Localization in Agriculture
AgriCountDINO：农业中参数高效的示例引导计数与定位
Abstract
Accurate counting and localization of plants and their organs support phenotyping and yield estimation, yet target appearance, scale, and density vary widely across species and imaging conditions. Exemplar boxes specify the target without category-specific retraining, and point predictions identify the individual instances contributing to the count. We introduce AgriCountDINO, a parameter-efficient exemplar-guided framework for joint counting and localization. It conditions frozen multiscale DINOv3 features on exemplar appearance and size, then progressively decodes them into target points. Missed-object recovery extends supervision to targets overlooked by initial matching, and exemplar-adaptive point NMS filters duplicate predictions according to exemplar scale. With 8.4M trainable parameters, approximately one-tenth of TasselNetV4's, AgriCountDINO achieves a three-shot MAE of 11.92 on the TPC-268 benchmark, reducing counting error by 9.7\% while providing individual target locations. Trained only on TPC-268, it achieves a zero-shot MAE of 14.25 on unseen generic object categories in FSC-147, improving upon the best compared zero-shot method by 6.0\% without target-domain training or fine-tuning.
Chinese Translation
植物及其器官的准确计数和定位支持表型分析和产量估算，但目标外观、尺度和密度在不同物种和成像条件下差异很大。示例框在不进行类别特定再训练的情况下指定目标，而点预测则识别出对计数贡献的个体实例。我们提出了AgriCountDINO，这是一种参数高效的示例引导框架，用于联合计数和定位。该框架基于示例的外观和大小对冻结的多尺度DINOv3特征进行条件处理，然后逐步解码为目标点。漏检目标恢复将监督扩展到初始匹配被忽视的目标，而示例自适应点非极大值抑制（NMS）根据示例尺度过滤重复预测。AgriCountDINO具有840万可训练参数，约为TasselNetV4的十分之一，在TPC-268基准上实现了三次拍摄的平均绝对误差（MAE）为11.92，计数误差减少了9.7\%，同时提供了个体目标位置。仅在TPC-268上训练的它在FSC-147中对未见的通用物体类别实现了零次拍摄的MAE为14.25，较最佳对比零次拍摄方法提高了6.0\,而无需目标领域训练或微调。
cs.CV / 61 / 2609.29517
AdaPilot: Towards Scene-Adaptive Policy Learning for Cross-Generator Text-to-Image Quality Optimization
AdaPilot：面向场景自适应策略学习的跨生成器文本到图像质量优化
Abstract
Existing methods for improving text-to-image generation quality have progressed from generator fine-tuning and prompt optimization to reinforcement learning with multi-turn visual feedback. However, existing strategies are deeply coupled with specific generators and tasks, and the learned capabilities are difficult to generalize into a universal quality optimization policy. Therefore, we propose AdaPilot, which learns a scene-adaptive, cross-generator transferable quality optimization policy by formulating multi-turn image generation as a Markov Decision Process (MDP) and optimizing it via end-to-end reinforcement learning. Specifically, AdaPilot decouples the policy from generator internals to enable cross-generator transfer, introduces scene-aware rewards that adaptively align quality assessment dimensions with task semantics, and employs process-level rewards to model the evolution trajectory of image quality. Experimental results show AdaPilot outperforms baselines in generation quality and generalization. Separate cross-generator evaluations further show that a single policy transfers zero-shot to unseen generators while maintaining positive average gains across all evaluated generators. Our project is available at https://github.com/QwenQing/Ada_pilot.
Chinese Translation
现有的文本到图像生成质量提升方法已从生成器微调和提示优化发展到使用多轮视觉反馈的强化学习。然而，现有策略与特定生成器和任务深度耦合，所学习的能力难以推广为通用的质量优化策略。因此，我们提出了AdaPilot，通过将多轮图像生成形式化为马尔可夫决策过程（MDP），并通过端到端强化学习进行优化，从而学习一种场景自适应的跨生成器可转移质量优化策略。具体而言，AdaPilot将策略与生成器内部解耦，以实现跨生成器转移，引入场景感知奖励，自适应地将质量评估维度与任务语义对齐，并采用过程级奖励来建模图像质量的演变轨迹。实验结果表明，AdaPilot在生成质量和泛化能力上优于基线。独立的跨生成器评估进一步表明，单一策略能够在未见过的生成器上实现零-shot 转移，同时在所有评估的生成器中保持正的平均增益。我们的项目可在 https://github.com/QwenQing/Ada_pilot 获取。
cs.CV / 62 / 2609.29527
CoSWA-YOLOv12: Scale-Invariant Tiny Object Detection and Segmentation of Malaria Parasites
CoSWA-YOLOv12：抗尺度的小型物体检测与疟疾寄生虫分割
Abstract
Automated microscopy could widen access to malaria diagnosis in low-resource settings, but the deadliest species, P. falciparum, presents in its early ring stage as an object only a few tens of pixels wide. Such tiny targets are systematically under-detected: overlap-based label assignment starves them of positive samples, and overlap-based box regression gives weak gradients at their scale. The Normalized Gaussian Wasserstein Distance (NWD) repairs both effects, but applied uniformly across a slide that also holds objects three to four times larger it loosens their supervision and erodes their localisation, so overall accuracy can fall even as the tiny class improves. We present CoSWA-YOLOv12, a compact YOLOv12 instance-segmentation detector whose core Cooperative Scale-adaptive Wasserstein Assignment routes the Wasserstein treatment to an object in inverse proportion to its size, tapering back to standard assignment for larger species. Two further components support it: a wavelet detail residual, and a min-max Gaussian regression loss (M2-NWD). All three additions are transfer-safe: each reproduces the standard pretrained model exactly at initialisation, so public pretrained weights load without any loss of accuracy. On a five-class Rwandan thick-smear dataset, CoSWA-YOLOv12 raises P. falciparum recall from 0.63 to 0.74 and mAP@50 from 0.73 to 0.81 (mask), cuts missed P. falciparum from 38% to 15%, and improves strict-localisation mAP@50-95 on all five classes for both detection and segmentation, while a 2x2 ablation shows the scale gate and the regression loss are synergistic.
Chinese Translation
自动化显微镜可以扩大低资源环境中疟疾诊断的可及性，但最致命的物种 P. falciparum 在其早期环状阶段呈现为仅几十个像素宽的物体。这种微小目标系统性地被低估检测：基于重叠的标签分配使其缺乏正样本，基于重叠的框回归在其尺度下给出微弱的梯度。归一化高斯瓦瑟斯坦距离（NWD）修复了这两种影响，但如果在包含三到四倍更大物体的切片上均匀应用，则会削弱它们的监督并侵蚀其定位，因此整体准确性可能会下降，即使微小类别有所改善。我们提出了 CoSWA-YOLOv12，这是一种紧凑的 YOLOv12 实例分割检测器，其核心合作尺度自适应瓦瑟斯坦分配根据物体大小的反比将瓦瑟斯坦处理路由到物体上，对于更大的物种则回归到标准分配。还有两个进一步的组件支持它：小波细节残差和最小-最大高斯回归损失（M2-NWD）。这三种附加组件都是传输安全的：每个组件在初始化时完全再现标准预训练模型，因此公共预训练权重加载时不会损失准确性。在一个包含五个类别的卢旺达厚涂片数据集中，CoSWA-YOLOv12 将 P. falciparum 的召回率从 0.63 提高到 0.74，mAP@50 从 0.73 提高到 0.81（掩码），将漏检的 P. falciparum 从 38% 降低到 15%，并在所有五个类别的检测和分割中改善严格定位的 mAP@50-95，同时 2x2 消融实验显示尺度门和回归损失是协同的。
cs.CV / 63 / 2609.29541
GeoRefer-Bench: A Benchmark from Referring Pixels to Verifiable Geospatial Reasoning
GeoRefer-Bench：从引用像素到可验证的地理空间推理的基准
Abstract
Referring segmentation in overhead imagery is inherently relational: a query may ask for the buildings north of the road or the pond closest to a residential area, so the correct referent can contain one object, several objects, or none. Existing benchmarks mainly score mask overlap, which cannot verify whether a model actually resolved the stated spatial relation. We introduce GeoRefer-Bench, a benchmark for verifiable geospatial referring segmentation. Each query is represented by an executable logical form over a metric scene graph, and predictions are evaluated with Exact Query Success (EQS), which is satisfied only when the returned instance set exactly matches the set denoted by the query. GeoRefer-Bench contains 700 whole 2048x2048 UAV scenes (2.94 Gpx) at 12.5 and 25 cm ground sampling distance, 26,217 instances, 142,796 spatial relations, and 20,916 executable queries spanning five reasoning levels. It further includes three paraphrases per query, 24.0% unanswerable queries, 2,477 counterfactual pairs, and five leakage-controlled evaluation splits. An independent audit re-derives object geometry, mask ownership, relation values, query execution, and split provenance, finding zero issues across all 700 scenes. Relation-blind strategies can retain non-trivial mIoU while achieving at most 22.7 EQS overall, showing that overlap alone does not certify relational grounding. Across fifteen current models, the strongest reaches 74.1 EQS but drops from 98.9 at level 1 to 60.5 at level 5, while ten models score below 5 EQS on two-hop queries. GeoRefer-Bench turns geospatial referring segmentation from mask matching into verifiable reference resolution.
Chinese Translation
在航拍图像中的引用分割本质上是关系性的：查询可能会要求寻找位于道路北侧的建筑物或离住宅区最近的池塘，因此正确的指称可以包含一个对象、多个对象或没有对象。现有的基准主要评估掩膜重叠，这无法验证模型是否真正解决了所述的空间关系。我们引入了GeoRefer-Bench，这是一个用于可验证地理空间引用分割的基准。每个查询通过一个可执行的逻辑形式在度量场景图上表示，预测通过精确查询成功率（Exact Query Success, EQS）进行评估，只有当返回的实例集与查询所指代的集合完全匹配时，EQS才被满足。GeoRefer-Bench包含700个完整的2048x2048无人机场景（2.94 Gpx），在12.5和25厘米的地面采样距离下，包含26,217个实例、142,796个空间关系和20,916个涵盖五个推理层级的可执行查询。它还包括每个查询的三个同义句，24.0%的不可回答查询，2,477个反事实对，以及五个泄漏控制的评估分割。独立审计重新推导了对象几何、掩膜归属、关系值、查询执行和分割来源，发现所有700个场景中没有问题。盲关系策略可以保持非平凡的mIoU，同时总体EQ的最高值仅为22.7，显示仅靠重叠并不能证明关系基础。在十五个当前模型中，最强的模型达到了74.1 EQS，但在第1层级从98.9降至第5层级的60.5，而十个模型在两跳查询中得分低于5 EQS。GeoRefer-Bench将地理空间引用分割从掩膜匹配转变为可验证的引用解析。
cs.CV / 64 / 2609.29550
Spaceborne differential photogrammetry for control-free measurement of large-gradient deformation with structural immunity and a predictable accuracy envelope
基于空间borne差分摄影测量的无控制测量大梯度变形的方法，具有结构免疫性和可预测的精度范围
Abstract
Optical satellite image correlation measures wide-area deformation in regimes where coherent interferometric synthetic aperture radar fails because displacement gradients are too large. However, standard pairwise workflows lack a pre-acquisition error budget and rely on extensive stable terrain. We formulate repeat-pass optical correlation as a differential estimation problem without surveyed ground control. Nominal georeferencing defines the coordinate frame, stable-area constraints and displacement priors resolve the datum, and surface displacement is estimated jointly with inter-epoch revisit-bias coefficients. The model yields a predictive accuracy envelope and calibrated per-point posterior uncertainty, bounds along-track uncertainty through a displacement prior, and represents pushbroom jitter using per-line revisit offsets. Simulations and Sentinel-2 and WorldView-2 experiments on the 2019 Ridgecrest earthquake, the 2023 Kahramanmara\c{s} earthquake, and the Baltoro glacier validate the predicted noise floor, control-free accuracy margin, and leakage caused by view-angle and digital elevation model errors. The measured noise floor reaches approximately $0.05$ pixel at $10$,m ground sampling distance. With only five stable tiles, conventional destriping changes the estimated Baltoro trunk velocity from $106$ to $1251$myr$^{-1}$, whereas the prior-constrained estimate remains $87$myr$^{-1}$. Closure analysis attributes approximately $88\%$ of pair-error variance to individual scenes, consistent with $25{,}354$ ITS_LIVE glacier-velocity triplets. Three matching methods lead to the same conclusions. The framework therefore turns pairwise correlation into a robust measurement with a predictive error budget, reduced dependence on stable terrain, and conclusions independent of the matching method.
Chinese Translation
光学卫星图像相关性在相干干涉合成孔径雷达失效的情况下测量广域变形，因为位移梯度过大。然而，标准的成对工作流程缺乏预先获取的误差预算，并依赖于广泛稳定的地形。我们将重复经过的光学相关性公式化为一个无调查地面控制的差分估计问题。名义地理参考定义了坐标框架，稳定区域约束和位移先验解决了基准问题，表面位移与跨时间重访偏差系数共同估计。该模型产生了一个预测的精度范围和经过校准的每点后验不确定性，通过位移先验界定沿轨道的不确定性，并使用每行重访偏移表示推扫抖动。对2019年里奇克雷斯特地震、2023年卡赫拉曼马拉斯地震和巴尔托罗冰川的Sentinel-2和WorldView-2实验的模拟验证了预测的噪声底线、无控制的精度边际，以及由视角和数字高程模型误差引起的泄漏。测得的噪声底线在10米地面采样距离下达到约0.05像素。仅用五个稳定的瓦片，传统去条纹处理将估计的巴尔托罗主干速度从106提高到1251米每年，而先验约束的估计保持在87米每年。闭合分析将约88%的成对误差方差归因于单个场景，这与25354个ITS_LIVE冰川速度三元组一致。三种匹配方法得出的结论相同。因此，该框架将成对相关性转变为一种具有预测误差预算的稳健测量，减少对稳定地形的依赖，并使结论独立于匹配方法。
cs.CV / 65 / 2609.29553
UNWIND: Any-Length Facial Video for Stress Detection without Temporal Windowing
UNWIND：无时间窗口的任意长度面部视频压力检测
Abstract
Automatic stress recognition from facial video provides a non-contact approach for affective monitoring. However, most existing video-based methods divide complete recordings into shorter temporal segments before performing classification. Such segmentation requires additional decisions concerning segment duration, overlap, and prediction aggregation, and may restrict the model from exploiting information distributed across the entire recording. We introduce UNWIND, a facial-video framework for stress detection that analyzes a complete recording as a single model input, eliminating the need for temporal windowing or external segmentation. UNWIND reorganizes the video by folding its temporal dimension into the channel dimension of a two-dimensional spatial representation, which is subsequently processed through a unified asymmetric-attention architecture. With a temporal stride of $\tau=1$, the framework processes the entire $120$-second sequence, corresponding to $3{,}600$ frames sampled at $30$~fps, in a single input. We evaluate seven temporal-stride settings on a stress dataset comprising $58$ subjects, using a stratified subject-level protocol that covers configurations from dense frame retention to sparse temporal sampling. The highest test accuracy, $70.02\%$, is obtained at $\tau=15$, while processing all frames at $\tau=1$ achieves a comparable accuracy of $69.73\%$. Computational requirements range from $12.48$ to $348.78$ GFLOPs across the evaluated stride settings, illustrating the balance between temporal sampling density and computational efficiency. The findings show that effective facial-video stress recognition can be achieved without dividing recordings into temporal windows and that complete-recording inference can be performed within a single unified model.
Chinese Translation
从面部视频中自动识别压力提供了一种非接触式的情感监测方法。然而，大多数现有的视频基础方法在进行分类之前会将完整录制分割成较短的时间段。这种分段需要额外的决策，涉及段的持续时间、重叠和预测聚合，并可能限制模型利用分布在整个录制中的信息。我们介绍了UNWIND，一个用于压力检测的面部视频框架，它将完整录制作为单个模型输入进行分析，消除了对时间窗口或外部分段的需求。UNWIND通过将视频的时间维度折叠到二维空间表示的通道维度中来重新组织视频，随后通过统一的非对称注意力架构进行处理。在时间步长$ au=1$下，该框架处理整个120秒的序列，相当于以30 fps采样的3600帧，在单个输入中完成。我们在一个包含58个受试者的压力数据集上评估了七个时间步长设置，使用分层受试者级协议，涵盖从密集帧保留到稀疏时间采样的配置。在$ au=15$时获得了最高的测试准确率70.02%，而在$ au=1$时处理所有帧也达到了相当的准确率69.73%。在评估的步长设置中，计算需求范围从12.48到348.78 GFLOPs，说明了时间采样密度与计算效率之间的平衡。研究结果表明，能够在不将录制分割成时间窗口的情况下实现有效的面部视频压力识别，并且可以在单个统一模型中进行完整录制推断。
cs.CV / 66 / 2609.29555
Visual Representation and History Modeling for Navigation World Models
导航世界模型的视觉表示与历史建模
Abstract
Navigation World Models (NWMs) predict action-conditioned visual futures for planning. Two practical challenges are central to their design: selecting a suitable visual representation and efficiently modeling observation history for repeated candidate queries. Standard Global-Softmax attention provides flexible interactions but repeatedly processes the same history, leading to increasing computation and memory costs for long contexts and multi-query planning. We study both problems within a unified conditional flow-transformer framework. We first compare five frozen visual representations under the same dynamics model and evaluation. To reduce redundant history computation, we design Cached-Linear, a hybrid architecture that combines local and shifted-window attention for target mixing with linear attention for reusable history access. We further develop Balanced Gated Delta Network (GDN), which augments this design with frame-wise recurrent memory for temporal history modeling. Experiments on RECON, SACSoN, and SCAND show that representation choice depends on the prediction objective: PAE-L performs best for reconstruction, RAE-B for direct prediction, and V-JEPA for long-horizon rollout. Under shared-history workloads, Cached-Linear substantially reduces computation and memory compared with Global-Softmax, while Balanced GDN improves selected direct-prediction endpoints with efficient context reuse. Overall, we systematically study visual representation and history modeling for NWMs and develop hybrid reusable-history architectures for efficient long-context and multi-query prediction.
Chinese Translation
导航世界模型（NWMs）预测基于动作的视觉未来以进行规划。其设计中有两个主要挑战：选择合适的视觉表示和高效建模观察历史以应对重复的候选查询。标准的全局软最大（Global-Softmax）注意力机制提供了灵活的交互，但在处理相同历史时会重复计算，导致在长上下文和多查询规划中计算和内存成本增加。我们在统一的条件流变换器框架内研究这两个问题。我们首先在相同的动态模型和评估下比较五种固定的视觉表示。为了减少冗余的历史计算，我们设计了缓存线性（Cached-Linear），这是一种混合架构，结合了局部和偏移窗口注意力以进行目标混合，并使用线性注意力以便于可重用历史访问。我们进一步开发了平衡门控增量网络（Balanced Gated Delta Network, GDN），该网络通过逐帧递归记忆增强了这一设计，以进行时间历史建模。在RECON、SACSoN和SCAND上的实验表明，表示选择依赖于预测目标：PAE-L在重建任务中表现最佳，RAE-B在直接预测中表现最佳，而V-JEPA在长时间展望中表现最佳。在共享历史负载下，缓存线性显著减少了与全局软最大相比的计算和内存，而平衡GDN通过高效的上下文重用改善了所选的直接预测端点。总体而言，我们系统地研究了NWMs的视觉表示和历史建模，并开发了混合可重用历史架构，以实现高效的长上下文和多查询预测。
cs.CV / 67 / 2609.29581
Long-Tail Adaptive Flow Matching with Explicit Conditional Consistency Guidance for Precise Multimodal Face Synthesis
具有显式条件一致性指导的长尾自适应流匹配用于精确的多模态人脸合成
Abstract
Although diffusion-based methods have substantially improved the controllability of multimodal face synthesis, their semantic alignment remains suboptimal because most existing approaches rely on implicit latent-space objectives to model the relationship between denoising variables and multimodal conditions. Such implicit modeling is often insufficient to enforce precise correspondence between synthesized faces and conditional inputs, especially under long-tailed semantic mask distributions where rare attributes receive weak optimization signals. To address these limitations, we propose EC\textsuperscript{2}Face, a multimodal face synthesis framework that improves semantic alignment through explicit semantic supervision and distribution-aware optimization. First, we introduce Explicit Conditional Consistency Guidance (ECCG), which imposes direct consistency supervision in pixel space by decoding an approximate reverse estimate of the clean latent and explicitly aligning the synthesized image with textual descriptions and semantic masks. A temporal dynamic modulation function is further designed to adapt the supervision strength according to the timestep-dependent reliability of reverse estimation. Second, we propose Long-Tail Adaptive Flow Matching (LAFM), which reweights spatial optimization signals based on semantic attribute frequency, with normalized weights to maintain numerical stability during training. Importantly, all additional modules are used only during training and introduce no extra inference overhead. Extensive experiments show that EC\textsuperscript{2}Face consistently outperforms competitive baselines in both generation quality and semantic alignment, achieving a 29.38\% improvement in mask accuracy on rare attributes.
Chinese Translation
尽管基于扩散的方法显著提高了多模态人脸合成的可控性，但其语义对齐仍然不理想，因为大多数现有方法依赖于隐式潜在空间目标来建模去噪变量与多模态条件之间的关系。这种隐式建模通常不足以强制合成的人脸与条件输入之间的精确对应，特别是在长尾语义掩码分布下，稀有属性接收到的优化信号较弱。为了解决这些局限性，我们提出了EC²Face，一个通过显式语义监督和分布感知优化来改善语义对齐的多模态人脸合成框架。首先，我们引入了显式条件一致性指导（ECCG），通过解码干净潜在变量的近似反向估计，在像素空间中施加直接一致性监督，并显式地将合成图像与文本描述和语义掩码对齐。进一步设计了一个时间动态调制函数，根据时间步依赖的反向估计可靠性调整监督强度。其次，我们提出了长尾自适应流匹配（LAFM），根据语义属性频率重新加权空间优化信号，并使用归一化权重以保持训练过程中的数值稳定性。重要的是，所有附加模块仅在训练期间使用，不会引入额外的推理开销。大量实验表明，EC²Face在生成质量和语义对齐方面始终优于竞争基线，在稀有属性的掩码准确率上实现了29.38%的提升。
cs.CV / 68 / 2609.29591
CATCH: Counterfactual Anatomical Tissue Inpainting with Conditional Haar Diffusion
CATCH：具有条件Haar扩散的反事实解剖组织修复
Abstract
BraTS local synthesis replaces masked regions in T1-weighted brain MRI with plausible tumor-free tissue while preserving observed anatomy. We present CATCH, conditional 3D diffusion in an invertible Haar-wavelet domain. Its denoiser receives noisy target coefficients, voided-image coefficients, and a signed mask; tumor-excluded wavelet reconstruction and a hole-focused loss guide training, and hard compositing preserves observed voxels. We compare fixed masks, tumor-component augmentation, and a weighted mixture of tumor-derived, irregular-blob, and ellipsoidal masks. Of 25 development cases, five prespecified cases select each arm's checkpoint and all 25 of their trajectory aggregations; a separate 75-case internal set compares the frozen pipelines and selects a weighted mixture for organizer evaluation. Five-trajectory averaging yielded internal SSIM/PSNR/MSE (mean$\pm$SD) of $0.80\pm0.13$, $19.18\pm1.80$dB, and $0.010\pm0.005$. As the sole officially evaluated pipeline, weighted mixture yielded $0.772\pm0.119$, $20.89\pm3.27$dB, and $0.0098\pm0.0054$ on the 219-case BraTS 2026 validation set. Against compute-matched random augmentation internally, it improved SSIM by 0.019 (95% bootstrap CI: 0.013-0.025), PSNR by 0.95dB, and MSE by 0.003; all three paired comparisons remained significant after Holm correction. Results favor the complete weighted-mixture policy within CATCH; absent official fixed- and random-pipeline scores and a directly comparable external baseline limit broader conclusions.
Chinese Translation
BraTS局部合成在T1加权脑MRI中用合理的无肿瘤组织替代被遮挡的区域，同时保留观察到的解剖结构。我们提出了CATCH，一种在可逆Haar小波域中的条件3D扩散。其去噪器接收噪声目标系数、空白图像系数和带符号的掩膜；排除肿瘤的小波重建和聚焦孔洞的损失引导训练，硬合成保留观察到的体素。我们比较了固定掩膜、肿瘤成分增强以及肿瘤衍生、不规则斑块和椭球形掩膜的加权混合。在25个开发案例中，五个预先指定的案例选择每个分支的检查点以及它们所有25个轨迹的聚合；一个独立的75案例内部集比较了冻结管道并选择了加权混合以进行组织者评估。五轨迹平均得到了内部SSIM/PSNR/MSE（均值±标准差）为$0.80 ext{±}0.13$、$19.18 ext{±}1.80$dB和$0.010 ext{±}0.005$。作为唯一经过官方评估的管道，加权混合在219案例的BraTS 2026验证集上得到了$0.772 ext{±}0.119$、$20.89 ext{±}3.27$dB和$0.0098 ext{±}0.0054$。与内部计算匹配的随机增强相比，它将SSIM提高了0.019（95%自助法置信区间：0.013-0.025）、PSNR提高了0.95dB，MSE提高了0.003；所有三组配对比较在Holm校正后仍然显著。结果支持CATCH中完整的加权混合策略；缺乏官方固定和随机管道分数以及直接可比的外部基线限制了更广泛的结论。
cs.CV / 69 / 2609.29592
QINA: Quantum-Inspired Nonlinear Adapters for Pretrained Vision Models
QINA：量子启发的非线性适配器用于预训练视觉模型
Abstract
Adapting large pretrained vision models under limited data and frozen-backbone constraints remains a central challenge in transfer learning. While lightweight adapters and parameter-efficient fine-tuning methods are widely adopted, most rely on generic multilayer perceptrons or low-rank linear updates, offering limited control over the spectral and geometric structure of feature transformations. We investigate whether structured nonlinear feature lifting can improve representational alignment in frozen regimes. We introduce Quantum-Inspired Nonlinear Adapters (QINA), compact modules that perform learnable trigonometric feature lifting followed by bounded nonlinear aggregation. The design induces structured oscillatory basis functions with an explicit norm-dependent Lipschitz bound, enabling spectral reshaping of pretrained representations without increasing the receptive field or significantly expanding parameter count. Importantly, the method operates entirely within standard deep learning frameworks and does not require quantum hardware. Through systematic experiments across natural and medical imaging datasets, classification and segmentation tasks, multiple adapters and placements, and varying training budgets, we show that performance in frozen regimes is primarily representation-limited. Nonlinear lifting improves adaptation, and the proposed structured trigonometric formulation consistently outperforms identity baselines, fixed Fourier feature mappings, and parameter-matched baseline adapters. Within the evaluated frozen-backbone settings, structured spectral parameterization provides a more effective inductive bias than generic nonlinear adapters. This work highlights the importance of geometry- and spectrum-aware adaptation mechanisms for large pretrained vision models.
Chinese Translation
在有限数据和固定骨干网络约束下，适应大型预训练视觉模型仍然是迁移学习中的一个核心挑战。尽管轻量级适配器和参数高效的微调方法被广泛采用，但大多数依赖于通用的多层感知机或低秩线性更新，提供的对特征变换的谱和几何结构的控制有限。我们研究了结构化非线性特征提升是否能够改善固定状态下的表征对齐。我们提出了量子启发的非线性适配器（QINA），这是一种紧凑模块，执行可学习的三角函数特征提升，随后进行有界的非线性聚合。该设计引入了具有显式范数依赖的利普希茨界限的结构化振荡基函数，使得在不增加感受野或显著扩展参数数量的情况下，对预训练表征进行谱重塑。重要的是，该方法完全在标准深度学习框架内运行，并且不需要量子硬件。通过在自然和医学成像数据集上进行系统实验，包括分类和分割任务、多种适配器和位置以及不同的训练预算，我们表明，在固定状态下的性能主要受表征限制。非线性提升改善了适应性，而所提出的结构化三角函数形式在一致性上优于身份基线、固定傅里叶特征映射和参数匹配的基线适配器。在评估的固定骨干网络设置中，结构化谱参数化提供了比通用非线性适配器更有效的归纳偏置。这项工作强调了几何和谱感知适应机制对于大型预训练视觉模型的重要性。
cs.CV / 70 / 2609.29604
PROVE: Proof-guided Regime-aware Operator Verification for Hallucination Detection in Medical Visual Question Answering
PROVE：基于证明引导的考虑情境的算子验证用于医学视觉问答中的幻觉检测
Abstract
In medical visual question answering (VQA), hallucinations of vision-language models (VLMs) may lead to confident but incorrect responses, raising the risk of diagnostic errors. Existing hallucination detection methods uniformly estimate the reliability of VLM outputs from response consistency or visual evidence. However, such uniform verification across questions ignores question-specific characteristics, resulting in missed overconfident errors and false alarms from over-verification. We present PROVE (Proof-guided Regime-aware Operator Verification), a black-box detector that adapts verification strategy to the evidential structure of each question. PROVE classifies questions into three verification regimes based on what kind of visual proof they demand, activates a regime-specific subset of five complementary operators, and adjusts operator importance per question through a lightweight calibration layer conditioned on deterministic question-answer features. PROVE uses question-specific evidence to reweight operators and produce a calibrated risk score. Evaluated on 8048 test samples across three medical VQA benchmarks and four frontier VLMs, PROVE achieves 0.821 AUROC, outperforming the strongest baseline by +0.159, with consistent gains across all models and benchmarks.
Chinese Translation
在医学视觉问答（VQA）中，视觉-语言模型（VLMs）的幻觉可能导致自信但错误的回答，从而增加诊断错误的风险。现有的幻觉检测方法统一估计VLM输出的可靠性，依据响应一致性或视觉证据。然而，这种跨问题的统一验证忽视了问题特定特征，导致错过过度自信的错误和因过度验证而产生的误报。我们提出了PROVE（基于证明引导的考虑情境的算子验证），这是一种黑箱检测器，能够根据每个问题的证据结构调整验证策略。PROVE根据问题所需的视觉证据类型将问题分类为三种验证情境，激活五个互补算子的情境特定子集，并通过一个轻量级的校准层根据确定性的问题-答案特征调整每个问题的算子重要性。PROVE使用特定于问题的证据对算子进行重新加权，并生成校准的风险评分。在三个医学VQA基准和四个前沿VLM上评估的8048个测试样本中，PROVE达到了0.821的AUROC，超越了最强基线+0.159，并在所有模型和基准上均表现出一致的提升。
cs.CV / 71 / 2609.29607
STRAND: Benchmarking and Improving Object-Centric Spatio-Temporal Monitoring in Video Large Language Models
STRAND：视频大语言模型中以对象为中心的时空监测基准与改进
Abstract
While multimodal large language models (MLLMs) have advanced video understanding, they remain highly prone to hallucinations in dynamic scenes. We argue this stems from a failure in spatio-temporal monitoring, the ability to persistently track object identities, states, and relations over time. Existing benchmarks obscure this deficit by relying on single final-answer evaluations for queries that can often be resolved via local visual cues or statistical priors. To rigorously diagnose this, we introduce STRAND, a benchmark of human-verified object-centric facts that evaluates intermediate reasoning by decomposing queries into sub-questions, distinguishing genuine temporal understanding from coincidental correctness. Crucially, we score models with Faithful Accuracy, an unconditional joint metric that credits a prediction only when the target answer and every prerequisite sub-question are correct, so that a model cannot inflate its score by being selectively consistent on the small subset of targets it happens to answer correctly. To address failure modes exposed by STRAND, we further propose an object-centric framework that explicitly constructs and reasons over structured object trajectories via chunk-wise state extraction and temporal aggregation. Extensive experiments, including backbone-, frame-, call-, and token-matched comparisons against both end-to-end MLLMs and modular video harnesses, demonstrate that our object-centric framework significantly reduces hallucinated answers and improves spatio-temporal reasoning consistency over state-of-the-art MLLMs. The code, model, and data have been made available at nguyentthong.github.io/strand.
Chinese Translation
尽管多模态大语言模型（MLLMs）在视频理解方面取得了进展，但它们在动态场景中仍然容易出现幻觉。我们认为这源于时空监测的失败，即在时间上持续跟踪对象身份、状态和关系的能力。现有基准通过依赖于单一最终答案的评估来掩盖这一缺陷，而这些问题往往可以通过局部视觉线索或统计先验来解决。为了严格诊断这一问题，我们引入了STRAND，一个基于人类验证的以对象为中心的事实基准，通过将查询分解为子问题来评估中间推理，从而区分真正的时间理解与偶然的正确性。关键是，我们使用忠实准确度（Faithful Accuracy）作为评分标准，这是一种无条件的联合指标，仅在目标答案和每个先决子问题都正确时才给予预测评分，因此模型不能通过在其恰好正确回答的小部分目标上选择性一致来抬高其分数。为了解决STRAND揭示的失败模式，我们进一步提出了一个以对象为中心的框架，该框架通过逐块状态提取和时间聚合显式构建和推理结构化的对象轨迹。大量实验，包括与端到端MLLMs和模块化视频工具的骨干、帧、调用和标记匹配比较，表明我们的以对象为中心的框架显著减少了幻觉答案，并改善了时空推理的一致性，超越了最先进的MLLMs。代码、模型和数据已在nguyentthong.github.io/strand上发布。
cs.CV / 72 / 2609.29621
AgenticCADedit: A Stateful, Tool-Mediated Agentic Approach to Multimodal 3D CAD Editing
AgenticCADedit：一种有状态的工具介导的主动式多模态3D CAD编辑方法
Abstract
Computer-aided design is central to industrial manufacturing, and much of a designer's daily work consists of editing existing models from multimodal requests involving speech, sketches, and model interaction. Existing neural CAD approaches focus predominantly on unconditional or text-conditioned generation. The neuralCAD-Edit approach formalizes expert multimodal editing requests, but its iterative baseline refines a complete CAD program across attempts, executing each attempt from the original model in a stateless CAD environment. Every attempt must therefore reconstruct the entire edit from scratch, so partially correct progress is discarded rather than accumulated, and the model can neither inspect the geometry it has just produced nor selectively revert a single faulty operation. We present AgenticCADedit, which turns editing into a sequence of small, verifiable actions on a persistent CAD state instead of a single regenerated program. Rather than emitting one complete program, it applies incremental code steps that each commit to the session, inspects the resulting faces and edges, renders highlighted selections to verify that the intended region was addressed, and reverts individual operations when it was not. Subsequent actions therefore build on the geometry produced by earlier ones. Our approach improves on all metrics for all three evaluated LLMs (open-weight: qwen3.6-27b, gemma4-31b; proprietary: gpt-5.6-luna), with the largest gains for the weakest baseline model, qwen3.6-27b, whose validity rises from 51.0% to 94.8% and acceptance from 1.6% to 12.0%. A token-cost analysis with gpt-5.6-luna further shows $66.7$% fewer output tokens than neuralCAD-Edit, while $94.8$% of input tokens are served from the prompt cache.
Chinese Translation
计算机辅助设计在工业制造中至关重要，设计师日常工作的大部分时间都用于根据涉及语音、草图和模型交互的多模态请求编辑现有模型。现有的神经CAD方法主要集中在无条件或文本条件生成上。neuralCAD-Edit方法形式化了专家的多模态编辑请求，但其迭代基线在每次尝试中都需要从原始模型在无状态CAD环境中执行完整的CAD程序。因此，每次尝试都必须从头开始重建整个编辑过程，因此部分正确的进展被丢弃而不是累积，模型既无法检查刚刚生成的几何形状，也无法选择性地撤回单个错误操作。我们提出了AgenticCADedit，它将编辑转变为在持久CAD状态上进行一系列小的、可验证的操作，而不是单个重新生成的程序。它不是输出一个完整的程序，而是应用逐步的代码步骤，每个步骤都提交到会话中，检查生成的面和边，渲染高亮选择以验证所需区域是否得到处理，并在未处理时撤回个别操作。因此，后续操作建立在早期操作生成的几何基础上。我们的方法在评估的三种大型语言模型（open-weight: qwen3.6-27b, gemma4-31b; proprietary: gpt-5.6-luna）上在所有指标上都有所改善，其中对最弱基线模型qwen3.6-27b的提升最大，其有效性从51.0%上升到94.8%，接受度从1.6%上升到12.0%。与gpt-5.6-luna的令牌成本分析进一步显示，其输出令牌比neuralCAD-Edit减少了66.7%，而94.8%的输入令牌来自提示缓存。
cs.CV / 73 / 2609.29638
SpectralCTGaussians: Projection-Domain Reconstruction and Basis Material Decomposition for Spectral CT using 3D Gaussian Splatting
SpectralCTGaussians：基于3D高斯点云的光谱CT投影域重建与基础材料分解
Abstract
Spectral computed tomography (CT) extends conventional CT by measuring attenuation across multiple energy channels, allowing improved modeling of physical X-ray interactions and energy-dependent material behavior and leading to richer scene understanding. We present a novel method for spectral CT reconstruction and basis material decomposition using 3D Gaussian Splatting by adding per-Gaussian basis material fractions to the set of learnable parameters, which together with a set of energy-dependent basis functions define the attenuation across the full spectral range. The basis functions represent various physical attenuation models such as photoelectric absorption and Compton scattering, and are jointly optimized across all energy channels through a differentiable polychromatic forward model, with material decomposition performed via mean-shift clustering of the resulting coefficients. We evaluate our method on a baseline real-world dataset as well as a synthetic dataset that we introduce, comparing against traditional reconstruction algorithms and state-of-the-art learning-based CT reconstruction methods. Our approach outperforms all traditional baselines in novel view synthesis and achieves the best PSNR among all compared methods for spectral CT volume reconstruction, while describing all energy channels with a single shared representation that requires a number of Gaussians comparable to single-channel Gaussian splatting-based CT reconstruction approaches. For basis material decomposition, no traditional or learning-based baseline offers one-step decomposition with direct RGB material segmentation, and our method additionally recovers the photoelectric basis with higher PSNR than traditional pipelines.
Chinese Translation
光谱计算机断层扫描（CT）通过在多个能量通道上测量衰减，扩展了传统CT的应用，允许更好地建模物理X射线相互作用和能量依赖的材料行为，从而实现更丰富的场景理解。我们提出了一种新颖的方法，通过在可学习参数集上添加每个高斯基础材料分数，利用3D高斯点云进行光谱CT重建和基础材料分解。该方法与一组能量依赖的基础函数共同定义了全光谱范围内的衰减。这些基础函数代表了各种物理衰减模型，如光电吸收和康普顿散射，并通过可微分的多色前向模型在所有能量通道上联合优化，材料分解则通过对结果系数进行均值漂移聚类来实现。我们在一个基准真实世界数据集和一个我们引入的合成数据集上评估了我们的方法，并与传统重建算法和最先进的基于学习的CT重建方法进行了比较。我们的方法在新视图合成中优于所有传统基线，并在所有比较方法中实现了光谱CT体积重建的最佳峰值信噪比（PSNR），同时以单一共享表示描述所有能量通道，所需的高斯数量与单通道基于高斯点云的CT重建方法相当。在基础材料分解方面，没有传统或基于学习的基线提供一步分解和直接RGB材料分割，而我们的方法在恢复光电基础时的PSNR高于传统流程。
cs.CV / 74 / 2609.29648
Albireo: Adaptive, Energy-Efficient Inference Framework for Video Object Detection on the Edge
Albireo：用于边缘视频目标检测的自适应节能推理框架
Abstract
Video object detection on edge devices runs computationally expensive detectors over long frame streams, causing high energy consumption and sustained GPU utilization. Although consecutive frames are highly redundant, naive frame skipping is content-blind: it skips during critical moments such as object entry, occlusion recovery, and abrupt motion, degrading detection quality. We present Albireo, a detector-agnostic, codec-free adaptive inference framework that wraps off-the-shelf detectors and decides when detector invocation can be safely skipped based on scene content and per-object temporal state, requiring no detector modification or retraining. Albireo maintains a 10-dimensional Kalman filter (KF) per active object and invokes the detector only when prediction uncertainty exceeds a threshold; on skipped frames, boxes are predicted from the KF state at near-zero GPU cost. A KF-based rescue mechanism preserves confirmed objects through brief detector misses to prevent output fragmentation, while a lightweight empty-scene screen avoids detector calls on objectless frames. We evaluate Albireo on the BDD100K MOT validation split with three architecturally distinct detectors (YOLO11x, YOLO26x, RF-DETR-Large) on two NVIDIA Jetson platforms (AGX Thor, AGX Orin). Across all configurations, Albireo keeps AP@50 within +/-1.2 pp of per-frame inference while reducing total energy by 12.1-17.6%. On YOLO26x, it improves AP@50 by +0.8 pp while reducing energy by 17.6% (Thor) and 14.4% (Orin) and per-frame energy-delay product by 24.9% and 26.1%, respectively. Thus, the default operating point improves accuracy, energy, and latency together. In contrast, FixedSkip-2, a fixed-interval baseline with a 50% skip rate, loses 8.6 pp AP@50. Source code, evaluation pipeline, and per-clip results are available at https://github.com/amirtaherin/albireo
Chinese Translation
边缘设备上的视频目标检测需要在长帧流中运行计算开销大的检测器，这导致高能耗和持续的GPU利用率。尽管连续帧之间高度冗余，但简单的帧跳过方法是内容盲的：它在关键时刻（如目标进入、遮挡恢复和突然运动）跳过帧，从而降低检测质量。我们提出了Albireo，这是一种与检测器无关、无编码器的自适应推理框架，它封装现成的检测器，并根据场景内容和每个目标的时间状态决定何时可以安全跳过检测器调用，而无需对检测器进行修改或重新训练。Albireo为每个活动目标维护一个10维的卡尔曼滤波器（Kalman Filter, KF），仅在预测不确定性超过阈值时调用检测器；在跳过的帧上，框是从KF状态预测的，几乎不消耗GPU资源。基于KF的救援机制通过短暂的检测器失误来保护已确认的目标，以防止输出碎片化，同时轻量级的空场景屏幕避免在无目标帧上调用检测器。我们在BDD100K MOT验证集上评估了Albireo，使用三种架构上不同的检测器（YOLO11x、YOLO26x、RF-DETR-Large）在两个NVIDIA Jetson平台（AGX Thor、AGX Orin）上进行测试。在所有配置中，Albireo将AP@50保持在每帧推理的±1.2 pp范围内，同时总能耗降低了12.1-17.6%。在YOLO26x上，它将AP@50提高了+0.8 pp，同时在Thor和Orin平台上分别减少了17.6%和14.4%的能耗，以及每帧能量-延迟乘积分别降低了24.9%和26.1%。因此，默认操作点在准确性、能耗和延迟上实现了共同改善。相比之下，FixedSkip-2（一个固定间隔基线，跳过率为50%）损失了8.6 pp的AP@50。源代码、评估管道和每个剪辑的结果可在https://github.com/amirtaherin/albireo获取。
cs.CV / 75 / 2609.29650
VG-TIE: An interpretable tabular-to-image encoding method based on visibility graphs
VG-TIE：一种基于可视性图的可解释表格到图像编码方法
Abstract
Tabular-to-image encoding methods enable the application of models based on both convolutional neural networks and vision transformers to tabular data, transforming feature vectors into images. Existing methods employ linear and nonlinear dimensionality reduction techniques (e.g., Principal Component Analysis (PCA), t-SNE, and UMAP) to determine pixel positions, resulting in images whose spatial layout do not inherently reflect feature relationships. This paper introduces Visibility Graphs for Tabular-to-Image Encoding (VG-TIE), a novel method that encodes the structure of feature values using Natural Visibility Graph (NVG) and Horizontal Visibility Graph (HVG) into a two-dimensional space obtained through PCA. The resulting images are model-agnostic and intrinsically interpretable. Each pixel corresponds to an input feature, its intensity reflects the magnitude and direction of deviation from the population mean, and edges represent formally defined visibility relationships between features. VG-TIE provides two interpretability methods: (i) feature ranking from node degree distributions; and (ii) local and global feature importance from pixel intensity combined with Grad-CAM. Experiments on six public tabular datasets show that VG-TIE is competitive with other tabular-to-image methods while providing interpretability on feature importance and ranking similar to intrinsic interpretable methods. The results highlight the potential of the proposed image-based transformation to provide an effective framework that expands the use of deep learning across tabular data domains.
Chinese Translation
表格到图像编码方法使基于卷积神经网络和视觉变换器的模型能够应用于表格数据，将特征向量转换为图像。现有方法采用线性和非线性降维技术（例如主成分分析（PCA）、t-SNE和UMAP）来确定像素位置，从而生成的图像的空间布局并未本质上反映特征之间的关系。本文提出了用于表格到图像编码的可视性图（VG-TIE），这是一种新颖的方法，通过自然可视性图（NVG）和水平可视性图（HVG）将特征值的结构编码到通过PCA获得的二维空间中。生成的图像与模型无关，且本质上是可解释的。每个像素对应一个输入特征，其强度反映了与总体均值的偏差的大小和方向，边缘表示特征之间正式定义的可视性关系。VG-TIE提供了两种可解释性方法：（i）基于节点度分布的特征排名；（ii）结合像素强度和Grad-CAM的局部和全局特征重要性。在六个公共表格数据集上的实验表明，VG-TIE在与其他表格到图像方法的竞争中表现良好，同时提供与内在可解释方法相似的特征重要性和排名的可解释性。结果突显了所提出的基于图像的转换提供有效框架的潜力，从而扩展深度学习在表格数据领域的应用。
cs.CV / 76 / 2609.29663
Investigating White Blood Cells as a Source of False-Positive Malaria Parasite Detection in African Blood-Smear Images
探讨白血球作为非洲血涂片图像中假阳性疟原虫检测的来源
Abstract
White blood cells (WBCs) present on every Giemsa-stained thick blood smear share visual properties with early-stage Plasmodium falciparum ring-form trophozoites: small size, round morphology, and intense purple staining. They are a plausible but untested source of false positives in parasite-only detectors. We trained two YOLOv12s models on the Lacuna Malaria Detection dataset (8,000 images from Uganda and Ghana): Model A with parasite labels only, and Model B with both parasite and WBC labels. Seven independent spatial and statistical analyses tested whether false positive (FP) predictions cluster near WBC locations. All seven refute the hypothesis. In both models, 95% of FPs are pure background detections (IoU below 0.10 against any ground-truth box); zero are WBC class confusions. Ripley's Cross-K analysis shows spatial repulsion between FP centroids and WBC positions at every radius tested. Model B outperforms Model A overall (mAP50 0.859 vs. 0.755), and the advantage is uniform across all WBC-proximity bands, pointing to multi-task representation learning rather than WBC suppression as the cause. False positives arise from Giemsa stain debris and preparation artifacts. Effective mitigation requires staining artifact augmentation and annotation of unannotated early-stage ring forms rather than WBC labeling alone.
Chinese Translation
每个Giemsa染色的厚血涂片上都存在的白血球（WBC）与早期阶段的恶性疟原虫（Plasmodium falciparum）环状营养体在视觉特性上具有相似性：小尺寸、圆形形态和强烈的紫色染色。它们是一个合理但未经检验的假阳性来源，可能影响仅检测寄生虫的模型。我们在Lacuna疟疾检测数据集（来自乌干达和加纳的8000张图像）上训练了两个YOLOv12s模型：模型A仅使用寄生虫标签，模型B则同时使用寄生虫和白血球标签。通过七项独立的空间和统计分析测试假阳性（FP）预测是否集中在白血球位置附近。所有七项分析均否定了这一假设。在两个模型中，95%的假阳性是纯背景检测（与任何真实框的IoU低于0.10）；没有出现白血球类别混淆。Ripley的Cross-K分析显示，在每个测试半径下，假阳性重心与白血球位置之间存在空间排斥。总体而言，模型B的表现优于模型A（mAP50 0.859对比0.755），这一优势在所有白血球接近度区间内均匀分布，表明多任务表示学习而非白血球抑制是其原因。假阳性源于Giemsa染色残留物和制备伪影。有效的缓解措施需要对染色伪影进行增强，并对未标注的早期环状形式进行标注，而不仅仅是对白血球的标注。
cs.CV / 77 / 2609.29678
ReCalMatch:Reliability-Calibrated Semantic Guidance for Semi-Supervised Fine-Grained Recognition
ReCalMatch：用于半监督细粒度识别的可靠性校准语义引导
Abstract
Semi-supervised fine-grained visual recognition is highly vulnerable to overconfident pseudo-label errors: visually similar categories frequently produce high-confidence yet incorrect predictions, and consistency regularization then reinforces these errors throughout training. Existing semi-supervised learning (SSL) methods estimate pseudo-label reliability almost entirely from the visual classifier itself---maximum probability, adaptive thresholds, or entropy---signals that remain blind to whether a predicted class is \emph{semantically} compatible with the visual representation. We propose \textbf{ReCalMatch}, a reliability-calibrated semantic framework for semi-supervised fine-grained recognition. Rather than treating textual semantics as auxiliary supervision, ReCalMatch uses multi-aspect semantic prototypes as \emph{calibration evidence} for pseudo-label learning. We construct class-conditioned semantic prototypes from class names and domain-specific semantic aspects, and measure a \emph{visual--semantic agreement} score between each unlabeled embedding and its pseudo-label prototype. This agreement is combined with prediction confidence and entropy into a single reliability weight that down-weights pseudo-labels that are visually confident but semantically inconsistent. A semantic consistency term and a semantic margin regularizer further sharpen prototype separability under limited labels. Extensive experiments on CUB-200-2011, Stanford Dogs, NABirds, and iNaturalist18 show that ReCalMatch consistently improves strong SSL baselines, with the largest gains in low-label regimes where pseudo-label noise is most severe.
Chinese Translation
半监督细粒度视觉识别对过于自信的伪标签错误高度敏感：视觉上相似的类别经常产生高置信度但错误的预测，而一致性正则化则在训练过程中进一步强化这些错误。现有的半监督学习（SSL）方法几乎完全依赖于视觉分类器本身来估计伪标签的可靠性——最大概率、自适应阈值或熵——这些信号无法判断预测类别是否与视觉表示在 extit{语义上}兼容。我们提出了 extbf{ReCalMatch}，一个用于半监督细粒度识别的可靠性校准语义框架。ReCalMatch并不将文本语义视为辅助监督，而是利用多方面的语义原型作为伪标签学习的 extit{校准证据}。我们从类别名称和领域特定的语义方面构建类别条件的语义原型，并测量每个未标记嵌入与其伪标签原型之间的 extit{视觉-语义一致性}得分。该一致性得分与预测置信度和熵结合成一个单一的可靠性权重，从而降低那些视觉上自信但语义上不一致的伪标签的权重。语义一致性项和语义边际正则化进一步增强了在有限标签下的原型可分性。在CUB-200-2011、Stanford Dogs、NABirds和iNaturalist18上的大量实验表明，ReCalMatch始终改善强大的SSL基线，尤其是在伪标签噪声最严重的低标签环境中取得了最大的提升。
cs.CV / 78 / 2609.29717
TopoFuse: Topology-Aware Tri-Planar Fusion for 3D Cryo-Electron Tomography Segmentation
TopoFuse：基于拓扑的三平面融合用于3D冷冻电子断层成像分割
Abstract
Automated segmentation of cryo-electron tomograms routinely produces masks that are voxel-accurate but topologically broken: membranes fragment, organelles merge into one another, and enclosed cavities collapse. Existing topology-aware losses reduce these violations but cannot eliminate them, because topology is encouraged through gradient pressure rather than structurally enforced. We introduce TopoFuse, which reframes topology as a differentiable projection operator rather than a loss penalty. At each forward pass, the projection operator $\mathrm{Proj}_T$ (a PH-guided sparse edit) identifies the critical voxels responsible for topological violations via bottleneck matching and applies sparse edits to satisfy a specified topology target (diagram feature counts and lifetime budgets) for dimensions $d \in \{0,2\}$. If the projection converges, the output satisfies those constraints on the downsampled grid ($s=2$); when it does not, a repair certificate exposes this explicitly, enabling downstream filtering. A topology prior head predicts the correction target directly from input features, removing any dependence on ground-truth topology at inference. Across three cryo-ET benchmarks, TopoFuse reduces Betti number error by 54% over the strongest soft-loss baseline ($p < 0.001$), improves Dice by 4.6 points, and edits only 3.1% of voxels to achieve this.
Chinese Translation
冷冻电子断层图的自动分割通常会产生体素精确但拓扑结构破损的掩膜：膜片断裂，细胞器相互融合，封闭腔体崩溃。现有的拓扑感知损失虽然可以减少这些违反情况，但无法消除它们，因为拓扑是通过梯度压力来鼓励，而不是通过结构性强制来实现的。我们提出了TopoFuse，将拓扑重新定义为一个可微分的投影算子，而不是损失惩罚。在每次前向传播中，投影算子$ ext{Proj}_T$（一种PH引导的稀疏编辑）通过瓶颈匹配识别出导致拓扑违反的关键体素，并应用稀疏编辑以满足指定的拓扑目标（图示特征计数和生命周期预算），适用于维度$d ext{∈} ext{0,2}$。如果投影收敛，输出将满足在下采样网格($s=2$)上的这些约束；如果不收敛，修复证书将明确暴露这一点，从而启用下游过滤。拓扑先验头直接从输入特征预测修正目标，消除了推理时对真实拓扑的依赖。在三个冷冻电子断层成像基准测试中，TopoFuse将Betti数误差降低了54%，相较于最强的软损失基线（$p < 0.001$），Dice系数提高了4.6分，并且仅编辑了3.1%的体素以实现这一目标。
cs.CV / 79 / 2609.29721
SALI: Shot-Aware Late Interaction for Cross-Shot Relation Matching in Text-to-Video Retrieval using Film-Grammar Knowledge
SALI：基于镜头意识的后期交互用于文本到视频检索中的跨镜头关系匹配，利用电影语法知识
Abstract
Text-to-video retrieval usually represents a video clip by a single embedding. This embedding often loses important relations between people. E.g., an interaction "Anna confronts Mark" is regularly filmed as alternating shot and reverse shot of both (Fig. 1a). No single shot or averaged embedding over clip shots captures this relation. Thus, we propose SALI (Shot-Aware Late Interaction). It extracts the subject and object from a single-sentence query, and matches the query, its subject and object text embeddings against each visual shot embedding of a video clip. The matching operator is greedy max or optimal transport. A film-grammar penalty in fine-tuning adds a small, consistent shift. Built on CLIP4Clip-meanP, SALI keeps overall recall on par on Condensed Movies and ActivityNet while raising R@1 on multi-shot relation queries by 3 and 12 points, the most among all compared methods, and improves such queries on MSR-VTT at a cost of 1.4 R@1 overall.
Chinese Translation
文本到视频检索通常通过单一嵌入表示视频片段。这种嵌入往往会丢失人与人之间的重要关系。例如，互动“安娜对抗马克”通常以交替镜头和反向镜头的形式拍摄（见图1a）。没有单一镜头或对片段镜头的平均嵌入能够捕捉到这种关系。因此，我们提出了SALI（基于镜头意识的后期交互）。它从单句查询中提取主语和宾语，并将查询及其主语和宾语的文本嵌入与视频片段的每个视觉镜头嵌入进行匹配。匹配操作采用贪婪最大化或最优运输。在微调中加入电影语法惩罚，增加了一个小而一致的偏移。基于CLIP4Clip-meanP，SALI在Condensed Movies和ActivityNet上的整体召回率保持平衡，同时在多镜头关系查询中将R@1提高了3和12个百分点，在所有比较方法中为最高，并在MSR-VTT上改善了此类查询，整体R@1成本为1.4。
cs.CV / 80 / 2609.29726
A Multimodal Dataset for Survival Prediction in Resected Pancreatic Ductal Adenocarcinoma
用于切除胰腺导管腺癌生存预测的多模态数据集
Abstract
Survival research in pancreatic ductal adenocarcinoma (PDAC) is limited by the scarcity of datasets linking whole-slide histology with clinical, molecular, and long-term outcome data. We present a retrospective single-centre cohort of 302 patients who underwent PDAC resection at University Medical Center Gottingen. The dataset comprises 446 H&E whole-slide images, clinicopathological variables, targeted sequencing data for 154 patients, and overall-survival outcomes. During follow-up, 253 patients died, and the median follow-up was 76 months. To establish initial reference values, we evaluated fourteen survival-prediction configurations using identical five-repetition Monte Carlo cross-validation partitions. Ridge Cox regression using numeric clinicopathological variables achieved a mean concordance of $0.649 \pm 0.042$ and $0.652 \pm 0.046$ after adding KRAS and TP53 mutation status. The image-only attention model achieved $0.603 \pm 0.030$, while multimodal fusion achieved $0.619 \pm 0.025$, the highest concordance among the neural models. These results establish promising initial benchmarks for future research using this pancreas-specific multimodal dataset, paving the way for external validation.
Chinese Translation
胰腺导管腺癌（PDAC）的生存研究受到与临床、分子及长期结果数据相联系的全切片组织学数据集稀缺的限制。我们展示了在哥廷根大学医学中心进行PDAC切除的302名患者的回顾性单中心队列数据。该数据集包含446幅H&E全切片图像、临床病理变量、154名患者的靶向测序数据以及总体生存结果。在随访期间，253名患者去世，中位随访时间为76个月。为了建立初步参考值，我们使用相同的五次重复蒙特卡罗交叉验证分区评估了十四种生存预测配置。使用数值临床病理变量的岭Cox回归在添加KRAS和TP53突变状态后，达到了平均一致性$0.649 imes 0.042$和$0.652 imes 0.046$。仅图像的注意力模型达到了$0.603 imes 0.030$，而多模态融合模型达到了$0.619 imes 0.025$，在神经模型中具有最高的一致性。这些结果为未来使用该特定胰腺多模态数据集的研究建立了有希望的初步基准，为外部验证铺平了道路。
cs.CV / 81 / 2609.29779
Mind the Gap: Mesh-Guided Repair of Broken Vessels
注意间隙：网格引导的破损血管修复
Abstract
Vessel segmentation is commonly optimized as voxel-wise classification, but small local errors can strongly disrupt vascular connectivity while having little effect on overlap scores. This is particularly problematic for downstream analyses that rely on centerlines, branches, connected components, or graph structure. We propose a mesh-guided post-processing framework for repairing broken vessel segmentations produced by nnU-Net. For each predicted binary mask, a deformable template mesh is fitted to the mask surface in physical space and used as a case-specific geometric scaffold. The fitted mesh is not voxelized as the final segmentation; instead, it guides conservative reconnection of disconnected components by proposing or validating thin bridge candidates under foreground-growth constraints. We evaluated this approach in three vascular anatomies using AortaSeg24 and SEGA for the aorta, TopCoW for the Circle of Willis, and PARSE for the pulmonary arteries. Performance is measured using Dice, connected-component Dice (ccDice), and the Betti-0 number. Across these datasets, repair substantially improved connectivity while preserving overlap: Dice remained nearly unchanged, whereas ccDice increased from 0.596 to 0.992 for aorta, from 0.722 to 0.835 for TopCoW, and from 0.028 to 0.862 for PARSE. The FOMAML meta-initialization further accelerated the fitting per-case, supporting practical mesh-based repair of the vascular topology. These results suggest that explicit mesh representations can provide a useful geometric prior for correcting topological failures in otherwise accurate voxel segmentations.
Chinese Translation
血管分割通常被优化为体素级分类，但小的局部错误可能会严重破坏血管的连通性，同时对重叠评分的影响较小。这对于依赖中心线、分支、连通组件或图结构的后续分析尤其成问题。我们提出了一种网格引导的后处理框架，用于修复由 nnU-Net 生成的破损血管分割。对于每个预测的二进制掩膜，一个可变形模板网格被拟合到物理空间中的掩膜表面，并用作特定案例的几何支架。拟合的网格并不作为最终分割的体素化结果；相反，它通过在前景生长约束下提出或验证细桥候选者，指导断开组件的保守重新连接。我们在三种血管解剖结构中评估了这种方法，使用 AortaSeg24 和 SEGA 进行主动脉的评估，使用 TopCoW 进行威利斯环的评估，以及使用 PARSE 进行肺动脉的评估。性能通过 Dice、连通组件 Dice (ccDice) 和 Betti-0 数量进行测量。在这些数据集中，修复显著改善了连通性，同时保持了重叠：主动脉的 Dice 几乎保持不变，而 ccDice 从 0.596 增加到 0.992，TopCoW 从 0.722 增加到 0.835，PARSE 从 0.028 增加到 0.862。FOMAML 元初始化进一步加速了每个案例的拟合，支持基于网格的血管拓扑修复的实际应用。这些结果表明，显式网格表示可以为修正其他准确体素分割中的拓扑失败提供有用的几何先验。
cs.CV / 82 / 2609.29785
Lightweight Vision Transformer-Based U-Net for Brain Tumor Segmentation from MRI
基于轻量级视觉变换器的 U-Net 脑肿瘤分割模型
Abstract
Accurate brain tumor segmentation from Magnetic Resonance Imaging is essential for diagnosis, treatment planning, and surgical guidance. Although Convolutional Neural Networks, particularly UNet, have achieved significant success in medical image segmentation, they often struggle to capture the long-range spatial dependencies required to model tumors with irregular shapes and complex boundaries. This paper proposes a lightweight Vision Transformer UNet that combines the hierarchical feature extraction capability of UNet with the global context modeling of Vision Transformers. The proposed architecture incorporates a compact ViT bottleneck within a U-Net encoder-decoder framework, enabling effective learning of both local and global features while maintaining computational efficiency with only 2.6 million trainable parameters. The model was evaluated on the TCGA LGG MRI Segmentation dataset, achieving a mean Intersection over Union of 0.8100 and a Dice score of 0.8446, outperforming the baseline UNet by 3.75% and 3.15%, respectively. Extensive quantitative and qualitative analyses, including confusion matrix evaluation, precision recall curves, per-image performance distribution, and tumor size dependency analysis, demonstrate the effectiveness and robustness of the proposed method for brain tumor segmentation.
Chinese Translation
从磁共振成像中准确分割脑肿瘤对于诊断、治疗计划和手术指导至关重要。尽管卷积神经网络，特别是 U-Net，在医学图像分割中取得了显著成功，但它们通常难以捕捉建模具有不规则形状和复杂边界的肿瘤所需的长距离空间依赖关系。本文提出了一种轻量级视觉变换器 U-Net，该模型结合了 U-Net 的层次特征提取能力和视觉变换器的全局上下文建模能力。所提架构在 U-Net 编码器-解码器框架内引入了紧凑的 ViT 瓶颈，有效地学习局部和全局特征，同时保持计算效率，仅需 260 万个可训练参数。该模型在 TCGA LGG MRI 分割数据集上进行了评估，平均交并比达到 0.8100，Dice 系数为 0.8446，分别比基线 U-Net 提高了 3.75% 和 3.15%。广泛的定量和定性分析，包括混淆矩阵评估、精确率-召回曲线、每图像性能分布和肿瘤大小依赖性分析，证明了所提方法在脑肿瘤分割中的有效性和鲁棒性。
cs.CV / 83 / 2609.29788
OREO: Fidelity Alignment in 3D Generation via On-the-fly Rendering-Editing Optimization
OREO：通过即时渲染-编辑优化实现3D生成中的保真度对齐
Abstract
Despite recent advancements in 3D generation, models often struggle to produce assets with high visual fidelity. To bridge this gap, we propose OREO, an alignment framework that enhances the realism of 3D generators by leveraging rich 2D diffusion priors. Instead of relying on static datasets, OREO establishes a dynamic optimization loop that produces on-the-fly edited renderings as 2D pseudo-targets. At its core, we introduce Reinforced Editing, which utilizes a 2D model to refine rendered views of the 3D output, enhancing their overall visual fidelity while preserving the underlying geometry, viewpoint, and content. These refined views serve as high-quality supervision targets, enabling the 3D generator to learn from its own generated samples and progressively improve its visual quality. Experiments demonstrate that OREO effectively improves upon pre-trained baselines, producing 3D assets with enhanced visual realism.
Chinese Translation
尽管最近在3D生成方面取得了进展，但模型在生成高视觉保真度的资产时仍然面临挑战。为了解决这一问题，我们提出了OREO，一个通过利用丰富的2D扩散先验来增强3D生成器现实感的对齐框架。OREO并不依赖于静态数据集，而是建立了一个动态优化循环，实时生成经过编辑的渲染图作为2D伪目标。在其核心，我们引入了强化编辑（Reinforced Editing），该方法利用2D模型来细化3D输出的渲染视图，提高其整体视觉保真度，同时保留底层几何、视角和内容。这些细化后的视图作为高质量的监督目标，使3D生成器能够从自身生成的样本中学习，并逐步提高其视觉质量。实验表明，OREO有效地改善了预训练基线，生成了具有增强视觉现实感的3D资产。
cs.CV / 84 / 2609.29813
S2Planner: Multi-Scale Semantic Planner for End-to-End Autonomous Driving
S2Planner：用于端到端自主驾驶的多尺度语义规划器
Abstract
We present S2Planner, a trajectory planner that combines three front-facing cameras with ego-motion history and the current driving command. A fine-tuned DINOv3 backbone and a Spatial Tuning Adapter produce multi-scale image features; a coarse-to-fine decoder then uses trajectory self-attention and camera-projected cross-attention to refine candidate waypoints. The contribution is the integration of ego-conditioned trajectory initialization with iterative, geometry-guided sampling of multi-scale image features, rather than a new visual backbone or attention operator. On the NAVSIM v1 non-reactive evaluation, the previously reported navtest run obtained 88.03 PDMS. Because that run was selected using navtest performance, this number is exploratory and cannot be interpreted as an unbiased test estimate. Validation-selected evaluation on unexposed data, repeated runs, and computational measurements are needed to establish generalization and efficiency.
Chinese Translation
我们提出了S2Planner，一种轨迹规划器，它结合了三台前置摄像头、车辆自运动历史和当前驾驶指令。经过精细调优的DINOv3主干网络和空间调节适配器生成多尺度图像特征；然后，粗到细的解码器利用轨迹自注意力和摄像头投影的交叉注意力来细化候选路径点。我们的贡献在于将自我条件轨迹初始化与多尺度图像特征的迭代几何引导采样相结合，而不是提出新的视觉主干或注意力操作符。在NAVSIM v1非反应性评估中，之前报告的navtest运行获得了88.03 PDMS。由于该运行是基于navtest性能选择的，因此这个数字是探索性的，不能被解释为无偏测试估计。需要在未暴露数据上的验证选择评估、重复运行和计算测量，以建立泛化能力和效率。
cs.CV / 85 / 2609.29816
AV-GRPO: Modality-Anchored Decoupling Diffusion Reinforcement Learning for Joint Audio-Video Generation
AV-GRPO：基于模态锚定的解耦扩散强化学习用于联合音视频生成
Abstract
Recent years have witnessed major progress in joint audio-video generation. Existing models still suffer from limited per-modality fidelity, insufficient text-modality alignment and weak cross-modal synchronization. While reinforcement-learning post-training offers a promising remedy, directly adapting it to joint audio-video generation is challenging. Heterogeneous multimodal rewards entangle learning signals and complicate credit assignment. Joint optimization of two modality towers is computationally expensive given their divergent dynamics. Moreover, synchronization evaluation difficulty depends on paired samples, preventing fair reward comparisons. We propose AV-GRPO, a modality-anchored online diffusion RL framework, and 5DAV, a decoupled, difficulty-controllable training dataset. AV-GRPO includes three key modules: (1) modality-anchored rollouts to disentangle learning signals and stabilize difficulty; (2) trajectory-locked frozen-tower optimization to reduce cost and reassign credit; (3) adaptive objectives and perturbation strengths tailored to modality-specific dynamics. This converts coupled multimodal preference learning into unimodal subproblems for precise reward attribution and better synchronization. Our 5DAV dataset decouples samples across five dimensions for systematic training. Experiments on JavisBench and VABench demonstrate AV-GRPO outperforms LTX-2.3 in generation quality, semantic alignment and cross-modal synchronization under LoRA and full fine-tuning. Ablations confirm our designs. Code and data: https://github.com/zhiyuxu03/AV-GRPO
Chinese Translation
近年来，联合音视频生成领域取得了重大进展。然而，现有模型仍然面临模态间保真度有限、文本模态对齐不足以及跨模态同步性较弱等问题。尽管强化学习后训练提供了一种有前景的解决方案，但将其直接应用于联合音视频生成仍然具有挑战性。异构多模态奖励交织了学习信号并复杂化了信用分配。考虑到两种模态塔的动态差异，联合优化的计算成本较高。此外，同步评估的难度依赖于配对样本，阻碍了公平的奖励比较。我们提出了AV-GRPO，一种基于模态锚定的在线扩散强化学习框架，以及5DAV，一个可解耦、难度可控的训练数据集。AV-GRPO包含三个关键模块：（1）模态锚定的回滚以解耦学习信号并稳定难度；（2）轨迹锁定的冻结塔优化以降低成本并重新分配信用；（3）针对模态特定动态量身定制的自适应目标和扰动强度。这将耦合的多模态偏好学习转化为单模态子问题，以实现精确的奖励归属和更好的同步。我们的5DAV数据集在五个维度上解耦样本，以便进行系统训练。在JavisBench和VABench上的实验表明，AV-GRPO在生成质量、语义对齐和跨模态同步性方面优于LTX-2.3，适用于LoRA和全量微调。消融实验验证了我们的设计。代码和数据可在：https://github.com/zhiyuxu03/AV-GRPO
cs.CV / 86 / 2609.29825
Anatomy-Aligned Surface Field Learning for Myocardial Reconstruction from Sparse Short-Axis Cine MRI
基于解剖对齐的表面场学习用于从稀疏短轴动态磁共振成像重建心肌
Abstract
Patient-specific 4D myocardial reconstruction from cine MRI supports quantitative functional assessment, regional motion analysis, and simulation-based modeling. However, routinely acquired short-axis (SAX) cine MRI is sparsely sampled along the through-plane direction, making dense and anatomically consistent surface reconstruction challenging. In this study, we propose an anatomy-aligned surface learning framework that parameterizes the epicardial and endocardial surfaces on a shared circumferential-longitudinal UV domain. This formulation converts irregular 3D reconstruction into structured coordinate-field completion with explicit correspondence across subjects and cardiac phases. Sparse SAX contours are encoded as UV observation fields, coverage-aware sampling improves robustness to incomplete slice coverage, and topology- and distortion-aware learning preserves circumferential continuity and local surface quality. Experiments on three public cine MRI datasets showed that the proposed method consistently outperformed representative mesh-based and implicit reconstruction approaches, achieving overall Chamfer distances of $2.887$~mm on ACDC, $2.641$~mm on M\&Ms, and $2.810$~mm on M\&Ms-2. The reconstructed sequences also preserved ventricular function, with end-diastolic volume and ejection fraction errors of $3.3$~mL and $1.1 \%$, respectively. These results demonstrate that anatomy-aligned UV learning provides an accurate, efficient, and correspondence-aware representation for sparse cine MRI reconstruction and myocardial modeling. The source code will be available at https://github.com/yuan-xiaohan/SAX2MyoSurf.
Chinese Translation
患者特异性的4D心肌重建来自动态磁共振成像（cine MRI），支持定量功能评估、区域运动分析和基于模拟的建模。然而，常规获取的短轴（SAX）cine MRI在穿层方向上采样稀疏，使得密集且解剖一致的表面重建具有挑战性。在本研究中，我们提出了一种解剖对齐的表面学习框架，该框架在共享的周向-纵向UV域上对心外膜和心内膜表面进行参数化。该公式将不规则的3D重建转换为具有跨受试者和心脏相位明确对应关系的结构化坐标场补全。稀疏的SAX轮廓被编码为UV观测场，覆盖感知采样提高了对不完整切片覆盖的鲁棒性，而拓扑和失真感知学习则保持了周向连续性和局部表面质量。在三个公共cine MRI数据集上的实验表明，所提出的方法在代表性的基于网格和隐式重建方法中始终表现优越，在ACDC上实现了总体Chamfer距离为$2.887$~mm，在M ext&Ms上为$2.641$~mm，在M ext&Ms-2上为$2.810$~mm。重建序列还保持了心室功能，舒张末期容积和射血分数的误差分别为$3.3$~mL和$1.1 ext{ extperthousand}$。这些结果表明，解剖对齐的UV学习为稀疏cine MRI重建和心肌建模提供了准确、高效且具有对应关系感知的表示。源代码将可在https://github.com/yuan-xiaohan/SAX2MyoSurf获取。
cs.CV / 87 / 2609.29835
Retrieve-to-Localize: Bridging Large Language Models and LiDAR Geometry for Spatial Grounding
检索-定位：连接大型语言模型与LiDAR几何以实现空间定位
Abstract
LiDAR provides precise geometric information for spatial perception tasks such as object detection in autonomous driving and outdoor robotics. However, recognizing and localizing individual objects is not sufficient to answer questions that require composing spatial relations and grounding the intended target. Motivated by recent advances in large language models (LLMs) for autonomous driving, we leverage their language priors to interpret complex spatial questions and ground the referred target in LiDAR geometry. To support this spatial grounding capability, we introduce SpatialLiDAR-QA, which combines single- and multi-step relational grounding with complementary spatial understanding tasks. We further propose SpatialLiDAR-LM, which aligns LiDAR point features with an LLM and grounds target coordinates through language-conditioned, position-aware proposal retrieval and local point refinement. This design derives target coordinates directly from local LiDAR geometry rather than through textual language decoding. Experiments demonstrate substantial improvements over representative LiDAR--language models and multi-camera VLMs on precise coordinate prediction tasks. Our dataset and model training code will be publicly released.
Chinese Translation
LiDAR为空间感知任务提供了精确的几何信息，例如在自动驾驶和户外机器人中的物体检测。然而，识别和定位单个物体不足以回答需要组合空间关系和定位目标的问题。受到最近大型语言模型（LLMs）在自动驾驶领域的进展的启发，我们利用它们的语言先验来解释复杂的空间问题，并在LiDAR几何中定位所指的目标。为了支持这一空间定位能力，我们引入了SpatialLiDAR-QA，它结合了单步和多步关系定位以及互补的空间理解任务。我们进一步提出了SpatialLiDAR-LM，它将LiDAR点特征与LLM对齐，并通过基于语言条件的、位置感知的提案检索和局部点细化来定位目标坐标。该设计直接从局部LiDAR几何中推导目标坐标，而不是通过文本语言解码。实验表明，在精确坐标预测任务上，相较于代表性的LiDAR-语言模型和多摄像头视觉语言模型（VLMs），我们的模型有显著的提升。我们的数据集和模型训练代码将公开发布。
cs.CV / 88 / 2609.29836
SplatLabel: Pseudo-Labelling through 4D Gaussian Splatting
SplatLabel：通过4D高斯点云进行伪标注
Abstract
While 2D Vision Foundation Models offer a pathway to automate 3D semantic pseudo-labelling, translating these priors into robust 3D representations typically requires complex heuristics or multi-model ensembles. We introduce SplatLabel, an automated pipeline that leverages a 4D Gaussian representation to extract LiDAR segmentation with predictive confidence, as well as semantic occupancy grids at arbitrary voxel resolutions. At its core, SplatLabel handles dynamic environments through an explicit temporal manifold that models the trajectories and lifespans of individual 3D primitives. This allows the system to accurately track moving actors and strictly define when objects appear and disappear, completely eliminating the need for pre-annotated 3D bounding boxes. To robustly support this dynamic tracking, the representation is grounded by structural and semantic priors: we guide scene geometry in unobserved regions by integrating 360-degree LiDAR via virtual depth maps, and rather than relying on domain-specific prompt engineering, we directly distill continuous soft probabilities from 2D models to inherently resolve semantic ambiguities over time and space. Finally, to accurately reflect the real-world trade-off between precision and recall, we reframe pseudo-label evaluation as a selective classification task using a generalized risk-recall metric. Experiments on SemanticKITTI demonstrate that SplatLabel consistently outperforms state-of-the-art baselines across multiple recall levels, establishing a highly robust framework for both 3D LiDAR segmentation and occupancy prediction.
Chinese Translation
虽然2D视觉基础模型为自动化3D语义伪标注提供了一条路径，但将这些先验转化为稳健的3D表示通常需要复杂的启发式方法或多模型集成。我们提出了SplatLabel，一个自动化管道，利用4D高斯表示提取具有预测置信度的LiDAR分割以及任意体素分辨率的语义占用网格。SplatLabel的核心通过显式的时间流形处理动态环境，建模个体3D原语的轨迹和生命周期。这使得系统能够准确跟踪移动的对象，并严格定义物体出现和消失的时刻，完全消除了对预标注3D边界框的需求。为了稳健地支持这种动态跟踪，该表示由结构和语义先验支撑：我们通过虚拟深度图集成360度LiDAR来指导未观测区域的场景几何，而不是依赖于领域特定的提示工程，我们直接从2D模型中提炼连续的软概率，以固有地解决时间和空间上的语义模糊性。最后，为了准确反映现实世界中精确度与召回率之间的权衡，我们将伪标签评估重新框定为使用广义风险-召回指标的选择性分类任务。在SemanticKITTI上的实验表明，SplatLabel在多个召回水平上始终优于最先进的基线，建立了一个高度稳健的框架，用于3D LiDAR分割和占用预测。
cs.CV / 89 / 2609.29863
Modelling dynamic systems transfer functions from events in computational neuromorphic imaging
从计算神经形态成像中的事件建模动态系统传递函数
Abstract
Event Vision Sensing (EVS) report threshold crossings of log-irradiance, so a static optical system imaging a static scene produces no output at all. The classical procedure for measuring a Point Spread Function (PSF), illuminating the system with a constant point source, therefore has no event-based equivalent: the probe must carry a temporal profile, and that profile becomes part of the measurement. A growing body of Computational Neuromorphic Imaging (CNI) work already exploits this, pairing engineered or modulated optics with event sensing, but each system adopts a particular excitation together with a particular reading of the event stream without the correspondence between the two being stated. We examine that correspondence directly within a analytical framework of an Linear Shift-Invariant (LSI) optical system with a specified Modulation Transfer Function (MTF), a first-order filter EVS pixel model, and three different temporal probes: a step function, a linear ramp and an exponential ramp. By analysing the inverse of the entire chain for different event-statistic, and comparing the results to the specified MTF, we identify the context where each probe is most relevant. We consider how photon-noise and cross-array threshold mismatch effects the analytical accuracy of the probe-inverse. Results show that the widely used step probe is highly susceptible to mismatch while resilient to photon shot-noise, while a linear rise probe and exponential rise probe retain their ability to infer signal levels even with high mismatch. We discuss the potential of dynamic-PSFs as components of a full forward operator from scene to events. In this, we use this analytical description to define dynamic-PSFs around EVS, and discuss the gaps toward a unified pixel model and a scene-composition framework required for CNI.
Chinese Translation
事件视觉传感（Event Vision Sensing, EVS）报告对对数辐照度的阈值穿越，因此，静态光学系统对静态场景的成像根本不会产生输出。测量点扩散函数（Point Spread Function, PSF）的经典程序是用恒定点光源照明系统，因此没有基于事件的等效方法：探针必须携带时间轮廓，而该轮廓成为测量的一部分。越来越多的计算神经形态成像（Computational Neuromorphic Imaging, CNI）研究已经利用这一点，将工程或调制光学与事件感知相结合，但每个系统采用特定的激励以及特定的事件流读取，而两者之间的对应关系并未明确说明。我们在一个线性时不变（Linear Shift-Invariant, LSI）光学系统的分析框架内直接考察这种对应关系，该系统具有指定的调制传递函数（Modulation Transfer Function, MTF）、一阶滤波器EVS像素模型，以及三种不同的时间探针：阶跃函数、线性上升和指数上升。通过分析不同事件统计的整个链的逆，并将结果与指定的MTF进行比较，我们确定每个探针最相关的上下文。我们考虑光子噪声和跨阵列阈值不匹配对探针逆的分析准确性的影响。结果表明，广泛使用的阶跃探针对不匹配高度敏感，但对光子散射噪声具有一定的抗性，而线性上升探针和指数上升探针即使在高不匹配情况下仍能保持推断信号水平的能力。我们讨论了动态PSF作为从场景到事件的完整前向算子的组成部分的潜力。在此，我们利用这种分析描述定义围绕EVS的动态PSF，并讨论实现统一像素模型和CNI所需的场景构成框架的差距。
cs.CV / 90 / 2609.29864
Efficient Continuous DEM Reconstruction under Limited Target-Resolution Supervision
有限目标分辨率监督下的高效连续数字高程模型重建
Abstract
High-resolution digital elevation models (DEMs) support Earth observation applications, but paired training references are often available only at coarser output resolutions. Reconstructing finer terrain grids therefore requires both effective transfer beyond the supervised scale and control of dense-query computation. To address this problem, SCOPE learns a continuous terrain representation from coarser-resolution pairs. It predicts a latent coefficient field on the low-resolution grid and reuses local Fourier residual functions through basis evaluation and geometry-guided ensemble fusion. This separates high-dimensional coefficient prediction from output-grid construction. Experiments on geographically distributed land--ocean samples assess supervised reconstruction, unseen-scale inference, cross-domain generalization, and theoretical computation. SCOPE leads the compared methods across six metrics in the main supervised-scale evaluation. At an unseen factor three times the training factor, land reconstruction reduces RMSE and MAE by approximately 12\% relative to bicubic interpolation, with errors close to target-scale fine-tuning. Ninefold output density increases counted multiply--accumulate operations by only about 2\%. Frozen-model validation on held-out external marine regions reduces RMSE relative to the DEM-specific implicit baseline EBCF-CDEM by approximately 19\% under self-downsampling and 2\% with cross-product inputs, while also yielding lower RMSE than LIIF-MS in both settings. These results demonstrate the value of reusable coefficient fields for accurate reconstruction beyond the supervised resolution with low incremental arithmetic cost.
Chinese Translation
高分辨率数字高程模型（DEMs）支持地球观测应用，但配对训练参考通常仅在较粗的输出分辨率下可用。因此，重建更精细的地形网格需要有效地超越监督尺度的转移，并控制密集查询计算。为了解决这个问题，SCOPE 从较粗分辨率对中学习连续地形表示。它在低分辨率网格上预测潜在系数场，并通过基函数评估和几何引导的集成融合重用局部傅里叶残差函数。这将高维系数预测与输出网格构建分离。对地理分布的陆地-海洋样本的实验评估了监督重建、未见尺度推断、跨领域泛化和理论计算。在主要的监督尺度评估中，SCOPE 在六个指标上领先于比较方法。在未见的训练因子三倍的情况下，陆地重建相较于双三次插值减少了约12%的均方根误差（RMSE）和平均绝对误差（MAE），其误差接近目标尺度的微调。在持出外部海洋区域的冻结模型验证中，相较于特定于DEM的隐式基线EBCF-CDEM，在自下采样下RMSE减少了约19%，在交叉产品输入下减少了2%，同时在这两种情况下也产生了低于LIIF-MS的RMSE。这些结果展示了可重用系数场在低增量算术成本下实现超越监督分辨率的准确重建的价值。
cs.CV / 91 / 2609.29930
EndoFSA: Endoscopic Few-Shot Image Generation via Rank-Constrained Parameter Adaptation
EndoFSA：通过秩约束参数适应实现内窥镜少样本图像生成
Abstract
WCE produces large-scale gastrointestinal image data yet pathological findings remain significantly underrepresented limiting the generalization performance of deep-learning based abnormality detection systems. SDG methods offer a practical solution to mitigate this imbalance. However their training directly on scarce abnormal samples often results in instability overfitting and structural distortions. Addressing these challenges requires controlled adaptation mechanisms that preserve anatomical priors while enabling realistic pathological variation. This paper presents EndoFSA a GAN-based model for Endoscopic Few-Shot image generation by Adaptation in WCE imaging. EndoFSA leverages a generator pretrained on abundant normal data and adapts it to abnormal domains using limited number of training samples through a rank-constrained parameter adaptation where only a small number of modulation parameters is updated while the pretrained weights remain frozen. By restricting parameter updates to a low dimensional subspace and incorporating perceptual boundary regularization and cluster-wise diversity control EndoFSA enables efficient model adaptation under limited data conditions and mitigates mode collapse while preserving the anatomical priors learned from normal data. Importantly EndoFSA operates without requiring pixel-level annotations, masks or bounding box supervision. Evaluation on publicly available WCE benchmark datasets spanning various abnormal categories demonstrates that EndoFSA generates abnormal images reproducing real lesions morphology. Moreover in a downstream classification task training an image classifier solely on synthetic abnormal images generated by EndoFSA yields performance comparable to that obtained with real images.
Chinese Translation
无线胶囊内窥镜（WCE）产生了大规模的胃肠道图像数据，但病理发现仍然显著不足，限制了基于深度学习的异常检测系统的泛化性能。少样本生成（SDG）方法提供了一种实用的解决方案来缓解这种不平衡。然而，直接在稀缺的异常样本上进行训练往往导致不稳定的过拟合和结构扭曲。解决这些挑战需要控制适应机制，以保留解剖先验，同时实现真实的病理变异。本文提出了EndoFSA，一种基于生成对抗网络（GAN）的模型，用于在WCE成像中通过适应生成内窥镜少样本图像。EndoFSA利用在丰富的正常数据上预训练的生成器，并通过秩约束参数适应将其调整到异常领域，使用有限数量的训练样本进行更新，其中仅更新少量调制参数，而预训练权重保持不变。通过将参数更新限制在低维子空间，并结合感知边界正则化和聚类多样性控制，EndoFSA在有限数据条件下实现了高效的模型适应，减轻了模式崩溃，同时保留了从正常数据中学习到的解剖先验。重要的是，EndoFSA在操作时不需要像素级注释、掩膜或边界框监督。在涵盖各种异常类别的公开WCE基准数据集上的评估表明，EndoFSA生成的异常图像再现了真实病变的形态。此外，在下游分类任务中，仅基于EndoFSA生成的合成异常图像训练图像分类器，其性能与使用真实图像获得的性能相当。
cs.CV / 92 / 2609.29934
Beyond Spatial Benchmarks: From Spatial Reasoning to Navigation
超越空间基准：从空间推理到导航
Abstract
Does progress on spatial reasoning benchmarks translate into better navigation? Existing benchmarks test isolated inferences from images or videos, with little connection to downstream navigation. Our analysis reveals a gap between benchmark-oriented spatial specialization and navigation performance, and shows how aligning spatial supervision with navigation goals, phases, and decision learning improves navigation. Guided by these findings, we build \textsc{Spatial-Nav-100K} and fine-tune in two stages, \textit{i.e.} first learning a shared spatial-navigation foundation, and then specializing each phase with the abilities it relies on. We further introduce Spatial-NPD, where a teacher conditioned on spatial priors produces grounded action preferences and distills them into a student policy, so no explicit spatial reasoning is needed at inference. With 45 A100 GPU-hours of policy training, our 8B model reaches SR/SPL of 77.4/35.4 on HM3D-v0.2, 60.2/30.5 on HM3D-v0.1, and 47.9/20.6 on train-unseen MP3D. It outperforms several systems that rely on closed-source models or thousands of GPU-hours of training, at 148 ms per action step. All code and datasets will be publicly available at https://github.com/ylwhxht/Spatial-Nav.
Chinese Translation
空间推理基准的进展是否能转化为更好的导航能力？现有基准测试主要评估来自图像或视频的孤立推理，与下游导航的联系甚微。我们的分析揭示了基准导向的空间专业化与导航性能之间的差距，并展示了如何通过将空间监督与导航目标、阶段和决策学习对齐来改善导航。基于这些发现，我们构建了 extsc{Spatial-Nav-100K} 并进行两阶段的微调，即首先学习共享的空间-导航基础，然后根据每个阶段所依赖的能力进行专业化。我们进一步引入了 Spatial-NPD，其中一个基于空间先验条件的教师生成有根据的行动偏好，并将其提炼为学生策略，因此在推理时不需要显式的空间推理。经过 45 个 A100 GPU 小时的策略训练，我们的 8B 模型在 HM3D-v0.2 上达到了 SR/SPL 77.4/35.4，在 HM3D-v0.1 上达到了 60.2/30.5，在训练未见的 MP3D 上达到了 47.9/20.6。它的表现优于依赖闭源模型或数千个 GPU 小时训练的多个系统，且每个动作步骤的耗时为 148 毫秒。所有代码和数据集将公开发布在 https://github.com/ylwhxht/Spatial-Nav。
cs.CV / 93 / 2609.29940
Mind What Matters for Reasoning: Aligning Cross-Modal Attention via Selective Probability Mass Concentration
关注重要因素以进行推理：通过选择性概率质量集中对跨模态注意力进行对齐
Abstract
Multimodal large language models (MLLMs) achieve strong performance on visual reasoning tasks, yet remain prone to hallucinations and over-reliance on language priors, often generating answers without adequately using task-relevant visual evidence. Existing approaches primarily improve reasoning through reasoning-oriented supervision or inference-time strategies. In this work, we study a complementary question: can multimodal reasoning be improved by strengthening implicit visual grounding without directly supervising the reasoning process? Motivated by the functional specialization of attention heads, we investigate whether reasoning can be improved by guiding only the heads most responsive to visual evidence grounding. We propose Selective Probability Mass Concentration (sPMC), a training framework that identifies grounding-responsive heads and selectively regularizes their text-to-image attention. sPMC treats normalized attention over visual tokens as a spatial probability distribution and encourages the probability mass to be assigned to semantically relevant regions using segmentation-derived spatial priors. Adaptive Head Selection restricts this guidance to visually responsive heads while leaving the remaining heads unconstrained to preserve their complementary functions. Across 6 multimodal benchmark suites, sPMC achieves an average zero-shot improvement of 3% and gains of up to 11.3% across multiple MLLMs while regularizing only 3%-15% of their attention heads. These results demonstrate that targeted guidance of sparse and implicit visual evidence pathways can directly improve multimodal reasoning.
Chinese Translation
多模态大型语言模型（MLLMs）在视觉推理任务中表现出色，但仍然容易出现幻觉和过度依赖语言先验，常常在没有充分利用与任务相关的视觉证据的情况下生成答案。现有的方法主要通过以推理为导向的监督或推理时策略来改善推理。在本研究中，我们探讨一个互补的问题：是否可以通过增强隐含的视觉基础而不直接监督推理过程来改善多模态推理？受注意力头功能专业化的启发，我们研究了仅通过引导对视觉证据基础反应最强的头部，是否可以改善推理。我们提出了选择性概率质量集中（sPMC），这是一种训练框架，旨在识别对基础反应敏感的头部，并选择性地对它们的文本到图像注意力进行正则化。sPMC将对视觉标记的归一化注意力视为空间概率分布，并使用基于分割的空间先验鼓励概率质量分配到语义相关的区域。自适应头选择将这种引导限制在视觉响应的头部，同时不约束其余头部以保留其互补功能。在6个多模态基准测试套件中，sPMC实现了平均3%的零-shot提升，并在多个MLLMs中获得高达11.3%的增益，同时仅对其3%-15%的注意力头进行正则化。这些结果表明，针对稀疏和隐含视觉证据路径的有针对性引导可以直接改善多模态推理。
cs.CV / 94 / 2609.29959
Not All Confusion Is Equal: A Source-Aware Uncertainty Diagnosis for Fine-Grained Aircraft Detection
混淆并非全然相同：一种源意识的不确定性诊断方法用于细粒度飞机检测
Abstract
Fine-grained object detectors are commonly evaluated with confusion matrices, which show where the model is confused but not why, nor whether the confusion can be reduced. We argue that confusion can be attributed to distinct, separable sources, each quantitatively measurable, turning a passive measurement into actionable guidance. We present $A^2E^2$, a diagnostic tool that decomposes the sources of confusion along two axes, $\{$aleatoric, epistemic$\} \times \{$within-class, between-class$\}$, giving a $2\times2$ taxonomy that enumerates the source types. Each quadrant is measured by its own quantity, computed in one of three places (input geometry, output-space disagreement, and the bias-parameter posterior), so the two epistemic sources are separated by construction rather than by an empirical correlation. On fine-grained aircraft detection, the four quadrants become four named sources with their own remedy verdict: affinity (geometric similarity, irreducible from size alone), heterogeneity (geometrically heterogeneous sub-variants, pointing to re-labeling rather than more data), contested (an insufficiently trained but learnable boundary, improvable), and collapsed (a class starved of data, reducible). After attributing the confusion to a specific reducible source, we apply a targeted intervention and verify experimentally that it reduces the diagnosed source specifically while leaving the irreducible sources unchanged. $A^2E^2$ thus turns confusion measurement into a concrete, validatable and actionable "diagnosis" in which the same off-diagonal mass can carry opposite causes and opposite remedies. We also state this framework's limits, including which sources are only partially identifiable on this specific dataset and why.
Chinese Translation
细粒度目标检测器通常通过混淆矩阵进行评估，这些矩阵展示了模型混淆的地方，但并未说明原因，也未说明混淆是否可以减少。我们认为，混淆可以归因于不同的、可分离的源，每个源都是可量化的，从而将被动测量转化为可操作的指导。我们提出了 $A^2E^2$，一种诊断工具，它沿着两个轴（$ ext{aleatoric}$ 和 $ ext{epistemic}$）以及（类内和类间）分解混淆的源，形成一个 $2 imes2$ 的分类法，列举源类型。每个象限通过其自身的量度进行测量，这些量度在三个地方之一（输入几何、输出空间不一致性和偏差参数后验）计算，因此这两个认知源是通过构造而非经验相关性分开的。在细粒度飞机检测中，四个象限变成了四个命名源，各自有其补救措施：亲和性（几何相似性，仅通过大小无法减少）、异质性（几何上异质的子变体，指向重新标记而非更多数据）、争议（一个训练不足但可学习的边界，可以改进）和崩溃（一个数据匮乏的类别，可以减少）。在将混淆归因于特定的可减少源后，我们应用了针对性的干预，并通过实验验证它确实减少了被诊断的源，同时保持不可减少源不变。因此，$A^2E^2$ 将混淆测量转化为一个具体的、可验证的和可操作的“诊断”，在这种情况下，相同的非对角质量可以承载相反的原因和相反的补救措施。我们还阐明了该框架的局限性，包括在这个特定数据集上哪些源仅部分可识别以及原因。
cs.CV / 95 / 2609.29963
ADATEX4D: adaptive texture capacity allocation for 4D gaussian splatting
ADATEX4D：用于4D高斯点云的自适应纹理容量分配
Abstract
Textured Gaussians improve local appearance capacity, but assigning the same texture resolution to every primitive wastes storage on low-detail or weakly visible regions. We introduce AdaTex4D, an adaptive texture-capacity module for deformation-based 4D Gaussian Splatting. Each Gaussian carries packed RGBA triplanes whose two axes grow independently according to visibility normalized screen-space gradients and deformed local scales. Experiments on N3DV and PanopticSports show that AdaTex4D reduces texture storage by more than half while preserving reconstruction quality. Under fixed memory budgets, adaptive allocation also improves quality over uniform texture assignment and reduces overall model and peak memory. These results show that dynamic, anisotropic texture allocation provides a more efficient way to distribute local appearance capacity in 4D Gaussian representations.
Chinese Translation
纹理高斯函数提高了局部外观容量，但对每个基元分配相同的纹理分辨率会在低细节或弱可见区域浪费存储空间。我们提出了AdaTex4D，这是一个用于基于变形的4D高斯点云的自适应纹理容量模块。每个高斯函数携带打包的RGBA三平面，其两个轴根据可见性归一化的屏幕空间梯度和变形的局部尺度独立增长。在N3DV和PanopticSports上的实验表明，AdaTex4D在保持重建质量的同时将纹理存储减少了一半以上。在固定内存预算下，自适应分配还改善了质量，相较于均匀纹理分配，减少了整体模型和峰值内存。这些结果表明，动态的各向异性纹理分配提供了一种更高效的方式来分配4D高斯表示中的局部外观容量。
cs.CV / 96 / 2609.29985
OceanXL: Large-scale Underwater 3D Gaussian Splatting via Block Partitioning and Adaptive Pruning
OceanXL：通过块划分和自适应修剪实现的大规模水下3D高斯点云重建
Abstract
Underwater 3D reconstruction is critical for marine exploration, ecological monitoring, and subsea infrastructure inspection, yet remains challenging at large scale due to light attenuation, scattering, and limited capture coverage. While 3D Gaussian Splatting (3DGS) enables high-quality real-time rendering, its application to large underwater scenes is constrained by high memory consumption and inefficient optimization over extensive areas. We propose OceanXL, a fast and scalable 3DGS-based framework for large-scale underwater reconstruction. OceanXL adopts a divide-and-conquer strategy, partitioning scenes into spatially coherent blocks to enable efficient optimization while preserving global geometric consistency. We further introduce an adaptive pruning scheme tailored to underwater conditions that removes redundant primitives, producing compact representations without sacrificing visual fidelity. Together, these components improve training efficiency and rendering performance for large scenes. We also introduce a large-scale underwater dataset covering diverse marine environments. Experiments on five large-scale scenes demonstrate favorable scalability, compactness, and efficiency--quality trade-offs over large-scene baselines. Controlled comparisons on the small-scale SeaThru-NeRF dataset further show competitive reconstruction quality with substantially smaller model sizes than underwater-specific methods.
Chinese Translation
水下3D重建对于海洋探索、生态监测和海底基础设施检查至关重要，但由于光衰减、散射和有限的捕获覆盖范围，在大规模应用中仍然面临挑战。尽管3D高斯点云重建（3DGS）能够实现高质量的实时渲染，但其在大规模水下场景中的应用受到高内存消耗和对广泛区域进行低效优化的限制。我们提出了OceanXL，一个基于3DGS的快速且可扩展的大规模水下重建框架。OceanXL采用分而治之的策略，将场景划分为空间一致的块，以实现高效优化，同时保持全局几何一致性。我们进一步引入了一种针对水下条件的自适应修剪方案，去除冗余的基本体，生成紧凑的表示而不牺牲视觉保真度。这些组件共同提高了大场景的训练效率和渲染性能。我们还介绍了一个涵盖多样海洋环境的大规模水下数据集。在五个大规模场景上的实验表明，相较于大场景基线，OceanXL在可扩展性、紧凑性和效率-质量权衡方面表现良好。在小规模SeaThru-NeRF数据集上的受控比较进一步显示出与水下特定方法相比，具有竞争力的重建质量和显著更小的模型尺寸。
cs.CV / 97 / 2609.29999
GHOST-Q: Towards Studying Grounding Hallucinations Overlooked Under Same-score TradeOffs in Quantized VLMS
GHOST-Q: 研究在量化视觉语言模型中被忽视的相同得分权衡下的基础幻觉
Abstract
Post-training quantization of vision--language models (VLMs) is typically assessed through aggregate task accuracy and memory savings, but preserving a headline score does not guarantee preservation of visual grounding behavior. We present GHOST-Q, a cross-precision controlled evaluation of three 8B VLM families under FP16, INT8, and NF4 across utility and hallucination-sensitive benchmarks. Rather than comparing only aggregate accuracy, we pair FP16 and quantized predictions item by-item to quantify how compression redistributes grounding successes and failures. Five of six quantized variants preserve MMStar accuracy within $\pm2$ percentage points, yet 10 of 36 paired effects remain significant after false-discovery-rate correction, nine on hallucination-sensitive conditions. Same-device A100 profiling further demonstrates that substantial memory reduction does not necessarily mean lower inference latency. Finally, an open-ended AMBER audit reveals strong generation budget censoring whose severity varies by architecture and precision. These results show that quantized VLMs should be evaluated jointly for aggregate utility, grounding reliability, generation behavior, and realized deployment efficiency.
Chinese Translation
视觉语言模型（VLMs）的后训练量化通常通过整体任务准确性和内存节省进行评估，但保持一个显著的得分并不保证视觉基础行为的保留。我们提出了GHOST-Q，这是一种在FP16、INT8和NF4下对三种8B VLM家族进行跨精度控制评估的方法，涵盖了效用和对幻觉敏感的基准测试。我们不仅比较整体准确性，而是逐项配对FP16和量化预测，以量化压缩如何重新分配基础成功和失败。在六种量化变体中，有五种在$ ext{MMStar}$准确性上保持在$ ext{±}2$个百分点内，然而在经过假发现率校正后，36个配对效果中有10个仍然显著，其中9个在对幻觉敏感的条件下。同设备A100的分析进一步表明，显著的内存减少并不一定意味着更低的推理延迟。最后，一个开放式的AMBER审计揭示了强烈的生成预算审查，其严重程度因架构和精度而异。这些结果表明，量化的VLMs应在整体效用、基础可靠性、生成行为和实际部署效率方面进行联合评估。
cs.CV / 98 / 2609.30026
Training-Free Hold-Usage Detection in Sport Climbing with Foundation Pose Models
基于基础姿态模型的运动攀岩无训练握点使用检测
Abstract
Detecting which holds a climber uses, and when, underpins automated scoring, movement analysis, and assistive systems for sport climbing. Existing approaches train task-specific models or repurpose 2D pose estimators whose hand keypoint sits at the wrist and foot keypoint at the ankle i.e. offset from the fingertips and toes that actually contact the holds, and whose hands are occluded in roughly half of all frames. We show that a frozen, off-the-shelf pose foundation model is sufficient: using the fingertip and toe keypoints of Sapiens, a per-frame proximity test against the annotated holds, per-limb mutual exclusion, and a short temporal-persistence rule, we detect hold usage without any climbing-specific training. On the The Way Up dataset (22 videos, 10 athletes, two routes), our method reaches an event F_1 of 90.2% on a held-out split (89.8% under leave-one-participant-out cross-validation) and 79.9% over all 22 videos at any temporal overlap, and performs best on footholds (F_1,89.8% overall, 96.6% held-out). Under an identical protocol it exceeds our reproductions of the YOLOv8-pose and ViTPose pipelines at every temporal threshold, with the margin widening under strict timing. An ablation shows that two intuitively helpful additions---dense foundation-feature change gating and body-part segmentation---both hurt, arguing that a minimal, keypoint-only design is the right one for this task. Finally, standard coaching statistics computed from our automatic predictions track ground truth closely (Pearson r=1.00 for climb time, 0.94 for pace), turning ordinary single-camera video into reliable performance metrics with no instrumentation.
Chinese Translation
检测攀岩者使用的握点及其使用时机，是实现自动评分、动作分析和辅助系统的基础。现有方法通常训练特定任务模型，或改用二维姿态估计器，但这些估计器的手部关键点位于手腕，脚部关键点位于脚踝，偏离实际接触握点的指尖和脚趾，且手部在约一半帧中被遮挡。本文展示了一个冻结的、现成的基础姿态模型即可胜任此任务：利用Sapiens模型的指尖和脚趾关键点，结合每帧与标注握点的接近性测试、各肢体的互斥规则及短时序持续性规则，实现了无需攀岩特定训练的握点使用检测。在The Way Up数据集（22段视频，10名运动员，2条路线）上，我们的方法在保留分割测试中事件F_1得分达90.2%（留一运动员交叉验证为89.8%），在所有22段视频任意时间重叠下为79.9%，且在脚握点检测上表现最佳（整体F_1为89.8%，保留分割为96.6%）。在相同协议下，该方法在所有时间阈值上均优于我们复现的YOLOv8-pose和ViTPose管线，且在严格时间要求下优势更明显。消融实验表明，两项直观有益的改进——密集基础特征变化门控和身体部位分割——均降低了性能，说明该任务采用最简设计、仅基于关键点的方法更为合适。最后，基于自动预测计算的标准教练统计数据与真实数据高度吻合（攀爬时间Pearson相关系数为1.00，节奏为0.94），实现了无需额外设备的单摄像头视频转化为可靠性能指标。
cs.CV / 99 / 2609.30037
AERIAL: Adversarial Evaluation of Robustness in Accuracy-Preserving Low-Precision EEG Decoders
AERIAL：在保持准确性的低精度脑电图解码器中进行对抗性鲁棒性评估
Abstract
Deployment-oriented compression is attractive for resource-constrained brain--computer interfaces (BCIs), but whether it changes adversarial vulnerability remains unclear. On BCI Competition IV-2a, we compare 32-bit floating-point (FP32) EEGNet and ShallowConvNet models with global magnitude pruning and simulated INT8 post training quantization (PTQ) and quantization-aware training (QAT) across nine subjects and three seeds. Simulation provides differentiable quantize--dequantize models for white-box attacks and gradient analysis, while native TensorRT deployment is used for validation. Accuracy-preserving compression does not improve direct robustness: at $\epsilon=0.005$, EEGNet PGD accuracy remains 22--24\% across FP32, 50\% pruning (P50), PTQ, and QAT. However, P50 reduces bidirectional transfer efficiency to 0.963/0.928 (FP32$\rightarrow$P50/P50$\rightarrow$FP32), versus 0.994/0.997 for PTQ; the same trend holds for ShallowConvNet. Gradient alignment shows a corresponding separation, while native PTQ agrees with simulated clean/adversarial predictions in 95--98\% of cases. These results show that direct robustness, adversarial transfer, and deployment efficiency are distinct properties of compressed EEG decoders.
Chinese Translation
面向部署的压缩对于资源受限的脑-计算机接口（BCI）具有吸引力，但其是否会改变对抗性脆弱性仍不清楚。在 BCI Competition IV-2a 中，我们比较了 32 位浮点（FP32）EEGNet 和 ShallowConvNet 模型，采用全局幅度剪枝和模拟的 INT8 后训练量化（PTQ）及量化感知训练（QAT），涵盖九名受试者和三个种子。模拟提供了可微分的量化-反量化模型，用于白盒攻击和梯度分析，同时使用原生 TensorRT 部署进行验证。保持准确性的压缩并未提高直接鲁棒性：在 $ ext{ε}=0.005$ 时，EEGNet PGD 准确率在 FP32、50\% 剪枝（P50）、PTQ 和 QAT 中保持在 22--24\%。然而，P50 将双向迁移效率降低至 0.963/0.928（FP32$
ightarrow$P50/P50$
ightarrow$FP32），而 PTQ 为 0.994/0.997；ShallowConvNet 也呈现相同趋势。梯度对齐显示出相应的分离，而原生 PTQ 在 95--98\\% 的情况下与模拟的干净/对抗性预测一致。这些结果表明，直接鲁棒性、对抗性迁移和部署效率是压缩脑电图解码器的不同特性。
cs.CV / 100 / 2609.30043
ConPro: Contrast Projection Pretraining for Label-Efficient Vessel Segmentation in DSA Sequences
ConPro：用于DSA序列中标签高效血管分割的对比投影预训练
Abstract
Dense vessel annotation in digital subtraction angiography (DSA) is labor-intensive, yet every unlabeled sequence records how contrast passes through the vessels. Semi-supervised methods take their targets from the current model, and generic self-supervised pretexts reconstruct static appearance, so this signal goes unused. We propose ConPro, a self-supervised pretraining scheme whose target is a contrast projection, the normalized drop of every pixel below its temporal median over the sequence. On DIAS and DSCA, with 10%, 20% and 50% of the training cases labeled, ConPro improves on training from scratch at every label fraction and is the best of the compared methods on DSCA at 20% and 50% labels. Controlled comparisons show that the gain comes from the target. A temporal-median target with the same input, loss and budget stays at scratch level, and using the projection directly instead of learning it, as an input channel or a pseudo-label, helps little or hurts. ConPro provides pretrained weights without changing the segmentation architecture, so it combines with semi-supervised training, and UniMatch, the strongest baseline, gains 0.5 to 2.0 Dice and 0.9 to 2.3 clDice at every label fraction when started from ConPro weights, reaching 75.4 Dice on DIAS and 81.3 on DSCA.
Chinese Translation
数字减影血管造影（DSA）中的密集血管标注劳动强度大，但每个未标记序列记录了对比剂如何通过血管。半监督方法从当前模型中获取目标，而通用自监督预文本重建静态外观，因此这一信号未被利用。我们提出了ConPro，一种自监督预训练方案，其目标是对比投影，即每个像素在序列中低于其时间中位数的归一化下降。在DIAS和DSCA数据集上，当标记训练案例的比例为10%、20%和50%时，ConPro在每个标签比例上均优于从头开始训练，并且在DSCA的20%和50%标签上是比较方法中表现最佳的。受控比较显示，性能提升源于目标。使用相同输入、损失和预算的时间中位数目标保持在从头开始的水平，而直接使用投影而不是学习它，作为输入通道或伪标签，帮助不大或反而有害。ConPro提供了预训练权重而无需改变分割架构，因此可以与半监督训练结合，最强基线UniMatch在从ConPro权重开始时，在每个标签比例上获得了0.5到2.0的Dice和0.9到2.3的clDice，在DIAS上达到75.4的Dice，在DSCA上达到81.3。
cs.CV / 101 / 2609.30080
Can Frozen Hyperspherical Features Guide the Selection of Pseudo Masks?
冻结超球特征能否指导伪掩膜的选择？
Abstract
Foundation segmenters such as SAM return several plausible masks for an unlabeled image, and a student trained on the wrong one inherits its errors. Choosing among them means querying a second large model or fitting a quality head to annotated masks. We show that a candidate can be judged by what it does to a frozen self-supervised backbone's features. Normalized DINOv2 patch features lie on a hypersphere, and a candidate mask splits that sphere in two. Based on this reading, we introduce SphereTrust, which scores each candidate by three properties of the split, the angular contrast between the two sides, the coverage of the foreground's appearance modes, and contact with the image frame, one for each of three common ways a mask fails, and ranks a pool in 0.55 s per image from the frozen features alone. On eight SAM and SAM3 candidate pools spanning camouflaged, salient, and dichotomous segmentation and camouflage under low light, SphereTrust exceeds the strongest evaluated external baseline on six pools by 1.7 to 9.3 percentage points in mean selected Dice. These comparisons include published selection rules and explicitly labeled adaptations of DSS and UCOD-MKD. On the two prompted camouflage pools, its mean selected Dice is within 0.1 percentage points of the candidate-derived DSS adaptation, with a lower catastrophic-error rate. Which cue carries the signal depends on the candidate pool. The same sphere also supports training. The leading candidates enter as a candidate set with their scores as priors, prototypes reorder them, and a cross-fitted second round completes the labels, raising weighted F by 4.5, 2.3, and 5.5 points over fixed-label training on the three MLLM anchor pools, with students competitive with published unsupervised methods on nineteen test sets.
Chinese Translation
基础分割器如SAM为未标记图像返回多个合理的掩膜，而在错误掩膜上训练的学生会继承其错误。选择其中一个意味着要查询第二个大型模型或为标注掩膜拟合一个质量头。我们展示了可以通过冻结的自监督骨干网络特征来判断候选掩膜的效果。归一化的DINOv2补丁特征位于超球面上，而候选掩膜将该球面分为两部分。基于这一观察，我们引入了SphereTrust，通过三个分割属性对每个候选掩膜进行评分：两侧之间的角度对比、前景外观模式的覆盖率以及与图像边框的接触，这三者分别对应掩膜失败的三种常见方式，并在每幅图像中仅依靠冻结特征以0.55秒的速度对候选池进行排序。在八个涵盖伪装、显著性和二分分割以及低光照下伪装的SAM和SAM3候选池中，SphereTrust在六个池中以1.7到9.3个百分点的平均选择Dice超过了评估的最强外部基线。这些比较包括已发布的选择规则和DSS及UCOD-MKD的显式标注适应。在两个提示的伪装池中，其平均选择Dice与候选派生的DSS适应相差仅0.1个百分点，且具有更低的灾难性错误率。哪个线索携带信号取决于候选池。同一球面也支持训练。领先的候选者作为候选集进入，分数作为先验，原型重新排序，交叉拟合的第二轮完成标签，使得在三个MLLM锚池上加权F值分别提高了4.5、2.3和5.5点，相比于固定标签训练，学生在十九个测试集上与已发布的无监督方法竞争。
cs.CV / 102 / 2609.30096
Accelerating Video Diffusion via Training-Free Trajectory Routing
通过无训练轨迹路由加速视频扩散
Abstract
Video diffusion is computationally expensive, as it requires executing a large model across many denoising steps. Even with step-distillation, inference remains expensive because every distilled step still requires a costly model evaluation. We present TRACK: TRajectory-Aware Capacity routing via top-K selection, a heterogeneous denoising strategy that switches between compatible large and small models at selected steps, reducing the average cost per denoising evaluation. The switching steps are determined using a calibration process. TRACK first rolls out a reference trajectory with the large model. Then at each step, the small model's prediction is also collected and compared against the large model's prediction to obtain a relative disagreement score. Both models receive the same latent, timestep, conditioning, and guidance inputs. Aggregating this signal over a calibration set produces a disagreement score map across diffusion steps, which determines a switching policy for an efficient inference process: quality-sensitive steps keep using the large model, while steps with low disagreement scores are routed to the small model. Inference executes only the selected model at each step, requiring no retraining, architecture or scheduler changes, or online dual-model evaluation. Across Wan 2.1, Cosmos 3, TurboDiffusion, and FastVideo, TRACK yields $1.95\times$, $2.04\times$-$2.73\times$, $2.69\times$, and $2.17\times$ speedups, respectively, with comparable aggregate quality and high diversity retention. TRACK thereby establishes automated, training-free model switching as a practical acceleration paradigm for video diffusion.
Chinese Translation
视频扩散计算成本高，因为它需要在多个去噪步骤中执行大型模型。即使使用步骤蒸馏，推理仍然昂贵，因为每个蒸馏步骤仍然需要进行成本高昂的模型评估。我们提出了TRACK：基于top-K选择的轨迹感知容量路由（TRajectory-Aware Capacity routing），这是一种异构去噪策略，在选定步骤之间切换兼容的大型和小型模型，从而降低每次去噪评估的平均成本。切换步骤通过校准过程确定。TRACK首先使用大型模型展开参考轨迹。然后在每个步骤中，还收集小型模型的预测，并与大型模型的预测进行比较，以获得相对不一致评分。两个模型接收相同的潜在变量、时间步、条件和引导输入。对校准集上聚合该信号生成的不一致评分图，决定了高效推理过程的切换策略：质量敏感的步骤继续使用大型模型，而不一致评分低的步骤则路由到小型模型。推理仅在每个步骤执行选定的模型，无需重新训练、架构或调度器的更改，或在线双模型评估。在Wan 2.1、Cosmos 3、TurboDiffusion和FastVideo上，TRACK分别实现了$1.95 imes$、$2.04 imes$-$2.73 imes$、$2.69 imes$和$2.17 imes$的加速，同时保持了相当的整体质量和高多样性保留。因此，TRACK确立了自动化、无训练的模型切换作为视频扩散的实用加速范式。
cs.CV / 103 / 2609.30107
Smartphone-Based Method for Automated Speed Enforcement
基于智能手机的自动化测速执法方法
Abstract
Smartphone cameras and computer vision (CV) hold significant promise in assisting public agencies with enforcing traffic laws and enhancing road safety. This work designs and tests a smartphone-based method for automated speed estimation and vehicle identification (license plate, make/model, and color recognition) via an automated pipeline to assist enforcement agencies in reliably identifying speeders. The CV code accurately recognizes nearly half (46%) of the license plates' text on 1,800 images from a Brazil open-source dataset, called UFPR-ALPR. Code tests on daytime recordings from hand-held smartphone videos (n = 73) and roadside cameras (n = 42) in Austin, Texas yield 60.8% accuracy for color detection (among all possible RGB color categories), 48.6% on vehicle make/manufacturer identification, and 16.89% on vehicle make and model identification. Prediction accuracy for speed estimation (within a 20% range), vehicle make (within the top 3 predictions), and license plate recognition (within the top 10 predictions) are 16.3%, 16.9%, and 29.7%, respectively. This paper also illuminates the legal, technological, and practical aspects of using smartphones for enforcement, including the potential use of recordings for enforcement purposes, emphasizing the need to transform the potential of smartphone-based CV technologies into practical tools for vital information on traffic violations.
Chinese Translation
智能手机摄像头和计算机视觉（CV）在协助公共机构执行交通法规和提升道路安全方面具有重要潜力。本研究设计并测试了一种基于智能手机的自动化速度估计和车辆识别（包括车牌、品牌/型号和颜色识别）的方法，通过自动化流程帮助执法机构可靠地识别超速车辆。CV代码在来自巴西开源数据集UFPR-ALPR的1,800张图像中，准确识别了近一半（46%）的车牌文本。在德克萨斯州奥斯汀市的手持智能手机视频（n = 73）和路边摄像头（n = 42）的白天录制中，代码测试显示颜色检测的准确率为60.8%（在所有可能的RGB颜色类别中），车辆品牌/制造商识别的准确率为48.6%，而车辆品牌和型号识别的准确率为16.89%。速度估计的预测准确率（在20%的范围内）、车辆品牌（在前3个预测中）和车牌识别（在前10个预测中）分别为16.3%、16.9%和29.7%。本文还阐明了使用智能手机进行执法的法律、技术和实践方面，包括录音在执法中的潜在用途，强调将基于智能手机的CV技术的潜力转化为获取交通违规重要信息的实用工具的必要性。
cs.CV / 104 / 2609.30130
Multimodal Thinking with Renderable Programs
可渲染程序的多模态思维
Abstract
Current vision-language models (VLMs) excel at visual content understanding and text-based reasoning, yet their structure limits the advancement of incorporating images into the reasoning chain. Though Omnimodal models have made efforts in unifying text and image generation, they focus on visual tasks in the open-domain, lacking tractability due to rasterized or latent representations of images. We introduce SVGLM, a framework that uses scalable vector graphics (SVG) primitives to connect text and image in reasoning tasks. We exploit the duality of SVG as both image description and text instructions, yielding a more compact, interpretable solution to equip general VLMs with the capability of generating images within the reasoning process. We provide a large curated dataset of SVG-based image editing dataset, as well as the paradigm to tune open-source VLMs. Experiments on a mathematical reasoning benchmark demonstrate that SVGLM achieves strong SVG generation power as well as think-with-image intelligence. Our results highlight SVG as a suitable medium for building more robust digital domain agents, bridging the gap between text-based thinking and pixel-based images.
Chinese Translation
当前的视觉-语言模型（VLMs）在视觉内容理解和基于文本的推理方面表现出色，但其结构限制了将图像纳入推理链的进展。尽管全模态模型在统一文本和图像生成方面做出了努力，但它们主要关注开放领域的视觉任务，由于图像的光栅化或潜在表示，缺乏可操作性。我们提出了SVGLM，一个使用可扩展矢量图形（SVG）原语将文本和图像连接在推理任务中的框架。我们利用SVG作为图像描述和文本指令的双重性，提供了一种更紧凑、可解释的解决方案，使通用VLMs具备在推理过程中生成图像的能力。我们提供了一个大型精心策划的基于SVG的图像编辑数据集，以及调优开源VLMs的范式。在数学推理基准上的实验表明，SVGLM在SVG生成能力和图像思维智能方面表现出色。我们的结果突显了SVG作为构建更强大数字领域代理的合适媒介，弥合了基于文本的思维与基于像素的图像之间的差距。
cs.CV / 105 / 2609.30187
Ego-Exo4D Human Meshes Dataset: 4D Human Motion Reconstruction for Ego-Exo Captures
Ego-Exo4D人类网格数据集：用于Ego-Exo捕捉的4D人类运动重建
Abstract
Ego-Exo4D is a large-scale dataset providing synchronized egocentric and multi-view exocentric video, a rich resource for skill learning and assessment, procedural activity understanding, and embodied AI. However, the dataset ships with only sparse 3D human pose annotations, and reconstructing dense human motion from its multi-view captures is nontrivial. To this end, we present Ego-Exo4D-HM, a large-scale dataset of 4D human motion reconstructions for Ego-Exo4D's captures, and release the accompanying reconstruction pipeline. The code, dataset, and documentation can be found at https://abhiram824.github.io/egoexo4d_human_meshes.
Chinese Translation
Ego-Exo4D是一个大规模数据集，提供同步的自我中心和多视角外部视频，是技能学习与评估、程序活动理解以及具身人工智能的重要资源。然而，该数据集仅附带稀疏的3D人类姿态标注，从其多视角捕捉中重建密集的人类运动并非易事。为此，我们提出了Ego-Exo4D-HM，这是一个针对Ego-Exo4D捕捉的4D人类运动重建的大规模数据集，并发布了相应的重建流程。代码、数据集和文档可在 https://abhiram824.github.io/egoexo4d_human_meshes 找到。
cs.CV / 106 / 2609.30210
The Alignment Illusion in Multimodal Large Language Models
多模态大型语言模型中的对齐幻觉
Abstract
Layer-wise visual-text similarity in Multimodal Large Language Models (MLLMs) is widely interpreted as evidence that the language model progressively integrates visual content into a shared representation space. This reading rests on the assumption that scalar alignment scores reflect content-level cross-modal interaction. To test this assumption, we apply controlled interventions to the visual stream. Across 13 MLLMs from five families spanning 0.5B to 72B parameters, replacing projector-output visual tokens with Gaussian noise sharply reduces task accuracy, yet four standard scalar measures (CKA, SVCCA, MIR, and the leading principal-angle cosine) fail to consistently separate the corrupted stream from the original. We call this failure the alignment illusion and trace it to the shared language-model pathway: anisotropic MLP down-projections pull visual and text tokens toward common output directions, producing weight-induced alignment. Because this component is essentially one-dimensional, we introduce the principal-angle gap (PA gap), defined as the difference between the top two principal-angle cosines, which separates weight-induced similarity from multi-directional visual structure. Under graded visual corruption, the PA gap tracks task accuracy more consistently than the scalar scores we consider; under a structured but irrelevant image, it further exposes regimes in which internal geometry and task accuracy come apart. Internal visual-text alignment in MLLMs is therefore best read as a geometric diagnostic of the visual stream inside the language model rather than a direct proxy for content-level cross-modal interaction, and is most informative when calibrated by controlled task evidence.
Chinese Translation
多模态大型语言模型（MLLMs）中的逐层视觉-文本相似性被广泛解读为语言模型逐步将视觉内容整合到共享表示空间的证据。这一解读基于一个假设，即标量对齐分数反映了内容层面的跨模态交互。为了检验这一假设，我们对视觉流进行了控制干预。在来自五个家族的13个MLLMs中，参数范围从0.5B到72B，将投影器输出的视觉标记替换为高斯噪声显著降低了任务准确性，然而四个标准标量度量（CKA、SVCCA、MIR和主角余弦）未能一致地区分被破坏的流与原始流。我们将这一失败称为对齐幻觉，并追溯其原因至共享语言模型路径：各向异性多层感知器（MLP）下投影将视觉和文本标记拉向共同的输出方向，产生了权重诱导的对齐。由于这一组件本质上是一维的，我们引入了主角间隙（PA gap），定义为前两个主角余弦之间的差异，它将权重诱导的相似性与多方向视觉结构区分开。在逐级视觉破坏下，PA间隙比我们考虑的标量分数更一致地跟踪任务准确性；在一个结构化但无关的图像下，它进一步揭示了内部几何与任务准确性分离的情况。因此，MLLMs中的内部视觉-文本对齐最好被视为语言模型内部视觉流的几何诊断，而不是内容层面跨模态交互的直接代理，并且在通过控制任务证据进行校准时最具信息量。
cs.CV / 107 / 2609.30221
WanPE: Towards Cinematic Prompt Enhancement for Modern Text-to-Video Generation
WanPE：面向现代文本到视频生成的电影化提示增强
Zhu, Yubo, Shao, Yawen, Dai, Ziyun, Fang, Zixun, Zhu, Kai, Sun, Siyang, Xue, Haolan, Wang, Chuxin, Weng, Tingyu, Luo, Jingming, Shi, Chen, Huang, Lianghua, Ai, Yufeng, Wang, Yuzheng, Zhang, Wenyuan, Shang, Yu, Bao, Yuxiang, Bi, Zoubin, Xiao, Jie, Xing, Jinbo, Zhao, Jiaxing, Zhong, Chongyang, Chen, Hengjian, Xie, Chenwei, Liu, Akide, Kan, Zhehan, Liu, Yu, Zhai, Wei, Zhong, Sheng, Tong, Wei
Abstract
Video generation begins in text space by authoring a cinematic screenplay, then materializes into pixels. As contemporary video generators scale to 30 seconds and faithfully follow complex conditions, the textual prompt largely directs the production, planning how actions, camera trajectories, lighting, and sound unfold across multi-shot sequences. In this paper, we present WanPE, a 397B-parameter prompt enhancement model trained on 1.05M real-world videos to master director-level cinematic planning. WanPE formulates shot-level cinematic plans via video-grounded reverse construction and employs Semantic-Consistency GRPO (SC-GRPO) to faithfully preserve user requirements across shots and over time. To benchmark this capability, we curate WanPEval, a human-annotated testbed covering durations from 5 to 30 seconds across varying intent granularities, supported by approximately 11K blind pairwise assessments. When powering Wan3.0's video generator, WanPE-397B boosts human preference over raw user prompts by 10.66-18.84 points at 5-15 seconds and by a dramatic 50.86 points in the 30-second arena. Ablation studies show that reverse construction demonstrates clear superiority over forward rewriting, while SC-GRPO robustly preserves semantic fidelity across model scales. Ultimately, WanPE leads all evaluated commercial offerings at 5-15 seconds and remains competitive with Seedance 2.5 at 30 seconds.
Chinese Translation
视频生成始于文本空间，通过创作电影剧本，然后转化为像素。随着当代视频生成器扩展到30秒并忠实地遵循复杂条件，文本提示在很大程度上指导了制作，规划了动作、镜头轨迹、照明和声音在多镜头序列中的展开。在本文中，我们提出了WanPE，一个拥有3970亿参数的提示增强模型，基于105万条真实世界视频进行训练，以掌握导演级的电影规划。WanPE通过视频基础的反向构建制定镜头级的电影计划，并采用语义一致性GRPO（SC-GRPO）在镜头间和时间上忠实地保留用户需求。为了评估这一能力，我们策划了WanPEval，这是一个涵盖5到30秒不同意图粒度的人类注释测试集，支持约11000个盲对比评估。当为Wan3.0的视频生成器提供动力时，WanPE-397B在5到15秒的时间段内提升了用户对原始提示的偏好10.66-18.84分，而在30秒的领域中则显著提升了50.86分。消融研究表明，反向构建在性能上明显优于前向重写，而SC-GRPO在模型规模上稳健地保留了语义保真度。最终，WanPE在5到15秒的评估中领先于所有评估的商业产品，并在30秒时与Seedance 2.5保持竞争力。
cs.CV / 108 / 2609.30222
TrackEverything: Long Horizon Dense Tracking via De-Duplicating 3D Scene Representations
TrackEverything：通过去重3D场景表示实现长时间密集跟踪
Abstract
Existing point tracking models face a fundamental tradeoff: they can either track a sparse set of query points over long horizons, or track all points across only short clips. We introduce TrackEverything, a 3D point tracker that breaks this trade-off by representing videos as persistent 3D scene tracks in world coordinates. Grounded in the insight that videos are 2D projections of an underlying 3D world, TrackEverything decouples model complexity from video duration, allowing it to scale with unique physical scene geometry instead. Our approach introduces three key innovations. First, we employ a voxelization-based de-duplication mechanism at sliding-window boundaries to merge co-located tracks, preventing repeated observations of the same surface from redundantly accumulating. Second, we decompose tracking into an endpoint refiner that predicts each point's destination and static-versus-dynamic classification, followed by a lightweight trajectory refiner that decodes dense trajectories exclusively for dynamic points. Third, we propose 3D WAFT, replacing memory-prohibitive 4D correlation volumes with efficient feature sampling in the scene cloud. To the best of our knowledge, TrackEverything is the first 3D tracker capable of tracking all visible points across videos exceeding 1000 frames within 40 GB of GPU memory. On TAPVid-3D, TrackEverything outperforms all open-source all-frame dense 3D trackers by more than 20% APD on short clips, while remaining competitive with state-of-the-art sparse trackers on long sequences, despite tracking far more points.
Chinese Translation
现有的点跟踪模型面临一个基本的权衡：它们要么能够在长时间内跟踪一组稀疏的查询点，要么只能在短片段中跟踪所有点。我们提出了TrackEverything，这是一种3D点跟踪器，通过将视频表示为世界坐标中的持久3D场景轨迹，打破了这一权衡。基于视频是潜在3D世界的2D投影这一见解，TrackEverything将模型复杂性与视频时长解耦，从而使其能够根据独特的物理场景几何形状进行扩展。我们的方法引入了三个关键创新。首先，我们在滑动窗口边界采用基于体素化的去重机制，以合并共位轨迹，防止对同一表面的重复观察冗余累积。其次，我们将跟踪分解为一个端点细化器，该细化器预测每个点的目标位置，并进行静态与动态分类，随后是一个轻量级轨迹细化器，仅为动态点解码密集轨迹。第三，我们提出了3D WAFT，用高效的特征采样替代内存占用过大的4D相关体积，以适应场景云。根据我们所知，TrackEverything是第一个能够在超过1000帧的视频中跟踪所有可见点，并且在40 GB GPU内存内运行的3D跟踪器。在TAPVid-3D上，TrackEverything在短片段中比所有开源的全帧密集3D跟踪器的APD高出20%以上，同时在长序列中与最先进的稀疏跟踪器保持竞争力，尽管跟踪的点数远远更多。
cs.CV / 109 / 2609.30223
BiCC: Bidirectional Connected-Component Loss for Instance-Aware Segmentation
BiCC：用于实例感知分割的双向连通组件损失
Abstract
Common segmentation losses aggregate errors voxel-wise, so lesions influence the objective in proportion to their volume, giving small but clinically critical lesions disproportionately little weight. Instance-aware losses aim to address this mismatch by assigning each lesion its own term. However, blob loss and CC-DiceCE derive their regions solely from annotations, so false-positive components receive no instance-level term. This matters in computer-assisted review, where each false-positive component may require separate inspection, making precision and false-positive burden important alongside recall. We introduce the bidirectional connected-component loss (BiCC), which pairs annotation- and prediction-derived partitions to score predicted components on their own scale. By deriving instances from the predictions, this branch directly penalizes false-positive components regardless of their size. The balance parameter $\alpha$ allows control over the lesion-wise precision-recall trade-off. Across five datasets with five-fold cross-validation using nnU-Net, BiCC outperforms CC-DiceCE in lesion-wise F1 on four datasets and blob loss on all five. It significantly improves over DiceCE on three datasets and matches it on two; CC-DiceCE instead loses up to 0.363 precision by favoring recall. Code is available at https://github.com/TIO-IKIM/BiCC-Loss.
Chinese Translation
常见的分割损失是逐体素聚合误差，因此病灶对目标的影响与其体积成正比，这使得小但临床关键的病灶在权重上受到不成比例的轻视。实例感知损失旨在通过为每个病灶分配其自身的项来解决这种不匹配。然而，blob损失和CC-DiceCE仅从注释中推导其区域，因此假阳性组件没有实例级别的项。这在计算机辅助审查中尤为重要，因为每个假阳性组件可能需要单独检查，因此精确度和假阳性负担在召回率之外也变得重要。我们提出了双向连通组件损失（BiCC），它将注释和预测导出的分区配对，以在其自身的尺度上对预测组件进行评分。通过从预测中推导实例，这一分支直接惩罚假阳性组件，而不考虑其大小。平衡参数$eta$允许控制病灶级别的精确度-召回率权衡。在使用nnU-Net进行五折交叉验证的五个数据集上，BiCC在四个数据集上优于CC-DiceCE的病灶级F1，并在所有五个数据集上优于blob损失。它在三个数据集上显著改善了DiceCE，并在两个数据集上与之持平；而CC-DiceCE则因偏向召回而损失高达0.363的精确度。代码可在https://github.com/TIO-IKIM/BiCC-Loss获取。
cs.CV / 110 / 2609.30234
OmniFabric: Coherent UV Space Texture Synthesis for 3D Garment Reconstruction
OmniFabric：用于3D服装重建的连贯UV空间纹理合成
Abstract
Automated generation of production-ready 3D garment assets from a single image is a central challenge in digital content creation. While recent generative models have significantly advanced 3D geometry reconstruction, synthesizing high-quality textures remains a bottleneck. Existing methods often bake environmental illumination and shadows directly into the texture map, or they fail to maintain global structural coherence, making the resulting assets unusable for physical simulation and relighting. In this work, we introduce OmniFabric, a novel approach that synthesizes globally coherent texture maps directly within the 2D sewing pattern space. Given a single reference image, our pipeline utilizes an estimated 3D mesh and generative priors of powerful Vision-Language Models (VLM) to establish a complete but coarse texture initialization across the unwrapped sewing patterns. We then leverage a specialized diffusion transformer, trained via an automated synthetic data engine and conditioned on 3D positional features, to refine this initialization directly in the canonical UV domain. This effectively removes distortion and baked-in artifacts to extract a clean and normalized texture map that preserves the original garment design. Extensive experiments demonstrate that OmniFabric significantly outperforms state-of-the-art baselines, yielding photorealistic 3D garments with high-quality textures.
Chinese Translation
从单张图像自动生成可用于生产的3D服装资产是数字内容创作中的一个核心挑战。尽管近期的生成模型在3D几何重建方面取得了显著进展，但合成高质量纹理仍然是一个瓶颈。现有方法通常将环境光照和阴影直接烘焙到纹理图中，或者未能保持全局结构的一致性，从而使得生成的资产无法用于物理模拟和重新照明。在本研究中，我们提出了OmniFabric，一种新颖的方法，直接在2D缝合模式空间中合成全局一致的纹理图。给定一张参考图像，我们的流程利用估计的3D网格和强大的视觉-语言模型（Vision-Language Models, VLM）的生成先验，在展开的缝合模式上建立完整但粗略的纹理初始化。然后，我们利用一个专门的扩散变换器，该变换器通过自动合成数据引擎训练，并以3D位置特征为条件，直接在标准UV域中精炼这一初始化。这有效地消除了扭曲和烘焙伪影，从而提取出干净且标准化的纹理图，保留了原始服装设计。大量实验表明，OmniFabric显著优于最先进的基线，生成具有高质量纹理的逼真3D服装。
cs.CV / 111 / 2609.30245
Towards Practical Compression of 3D Gaussian Splatting
迈向实用的三维高斯点云压缩
Abstract
3D Gaussian Splatting (3DGS) enables high-quality novel-view synthesis but requires substantial storage. Existing compression methods often rely on spatial context modeling over irregular 3D representations, increasing the complexity of training and coding. Meanwhile, floating-point context inference can introduce numerical inconsistencies across platforms, causing entropy-decoding failures. To address these practical challenges, we propose COSA-GS, which constructs context without spatial aggregation through anchor-wise causal factorization. Specifically, we use geometry context derived from each anchor's coordinates to model a compact learnable anchor latent. The anchor latent is then fused with the geometry context to form an anchor context for attribute coding. The resulting context model features a simple architecture composed solely of linear transformations and activations. We train COSA-GS using rate--distortion optimization with adaptive Gaussian pruning. Further, we develop quantization-aware training and integer inference for the context model to achieve bit-exact consistency of entropy-decoded symbols across platforms. Experiments demonstrate that COSA-GS achieves state-of-the-art compression performance while retaining fast and consistent cross-platform decoding, providing a simple yet effective framework for practical 3DGS compression. Code is available at https://github.com/pengpeng-yu/COSA-GS.
Chinese Translation
三维高斯点云（3D Gaussian Splatting，3DGS）能够实现高质量的新视角合成，但需要大量存储。现有的压缩方法通常依赖于对不规则三维表示的空间上下文建模，这增加了训练和编码的复杂性。同时，浮点上下文推断可能导致跨平台的数值不一致，从而引发熵解码失败。为了解决这些实际挑战，我们提出了COSA-GS，它通过锚点因果分解构建没有空间聚合的上下文。具体而言，我们利用每个锚点坐标导出的几何上下文来建模一个紧凑的可学习锚点潜变量。然后将锚点潜变量与几何上下文融合，形成用于属性编码的锚点上下文。最终的上下文模型具有一个简单的架构，仅由线性变换和激活组成。我们使用自适应高斯修剪的率-失真优化训练COSA-GS。此外，我们为上下文模型开发了量化感知训练和整数推断，以实现跨平台熵解码符号的比特精确一致性。实验表明，COSA-GS在保持快速且一致的跨平台解码的同时，实现了最先进的压缩性能，为实用的3DGS压缩提供了一个简单而有效的框架。代码可在 https://github.com/pengpeng-yu/COSA-GS 获取。
cs.AI / 1 / 2609.28475
When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability Routing
预测代理何时应进行推理？可靠性路由的行为压力测试
Abstract
Forecasting agents increasingly combine language-model reasoning, retrieval, ensembling, and calibration, but it remains unclear when each behavior should be trusted. We study this question on ForecastBench-style binary forecasting tasks, treating the choice to retrieve, reason, defer to a market prior, or use a historical analog as an observable agent behavior rather than a hidden implementation detail. Our central finding is that mechanism choice is source-dependent: structured analogs dominate for some data-generating processes, while market/crowd-style and conservative baselines are better for others. We introduce ReliabilityRoute, a structural intervention that steers forecasting-agent behavior using reliability features such as historical coverage, market-prior availability, source-prior sharpness, evidence strength, evidence disagreement, and horizon. A fixed 2024-fitted rule closely matches a hand taxonomy without hard-coded source-name decisions, while a walk-forward self-adjusting rule refits thresholds from previously resolved vintages and obtains the best mean Brier score among our deterministic systems across 16 later LLM vintages. The gain is modest and historical/search baselines remain highly competitive. The main contribution is therefore a behavioral stress test showing that more reasoning is not always better; forecasting agents should first estimate which evidence source deserves control, routing policies should themselves adapt under auditable constraints, and reproducibility artifacts are available at https://github.com/louiswang524/forcastagent
Chinese Translation
预测代理越来越多地结合了语言模型推理、检索、集成和校准，但何时应信任每种行为仍不清楚。我们在ForecastBench风格的二元预测任务中研究了这个问题，将选择检索、推理、依赖市场先验或使用历史类比视为可观察的代理行为，而不是隐藏的实现细节。我们的主要发现是机制选择依赖于数据源：对于某些数据生成过程，结构化类比占主导地位，而市场/人群风格和保守基线则在其他情况下表现更好。我们引入了ReliabilityRoute，这是一种结构性干预，利用历史覆盖、市场先验可用性、源先验的清晰度、证据强度、证据不一致性和时间范围等可靠性特征来引导预测代理的行为。一个固定的2024年拟合规则与手动分类法密切匹配，而不需要硬编码的源名称决策，同时一个前向自调整规则从先前解决的样本中重新拟合阈值，并在我们16个后续LLM样本的确定性系统中获得最佳的平均Brier分数。增益是适度的，历史/搜索基线仍然具有高度竞争力。因此，主要贡献是一个行为压力测试，表明更多的推理并不总是更好；预测代理应首先评估哪个证据源值得控制，路由策略本身应在可审计的约束下进行调整，重现性文档可在https://github.com/louiswang524/forcastagent获取。
cs.AI / 2 / 2609.28506
TW3Cast: A Frozen Router of Lightly Fine-Tuned Foundation Models for Time-Series Forecasting on GIFT-Eval, Selected Entirely on the Training Split
TW3Cast：一个轻微微调基础模型的冻结路由器，用于在GIFT-Eval上进行时间序列预测，完全基于训练集选择
Abstract
TW3Cast is a time-series forecasting system that reaches position 3 of 130 entries on the GIFT-Eval benchmark by mean MASE rank, as of 2026-09-14. The two entries above it belong to the leaderboard's agentic category, multi-step systems that use agents or language models to reason about, generate or select forecasts. TW3Cast runs no agent and no language model. Its selection is a table computed once on the training split and then frozen, and its experts are public foundation models lightly fine-tuned on those training splits. For each of the 97 dataset, frequency and horizon configurations, the table serves one of four modes: a specialist, which is a LoRA or full fine-tune of Chronos-2, TiRex or Toto whose training data was cleaned and enriched by explicit rules; a quantile blend that contains a specialist; a blend of base models; or a selection tournament played on a backtest carved from the training split. Every decision in the table was taken on that backtest. A specialist is admitted the moment it beats the tournament there, so a candidate costs a few megabytes and minutes of GPU time, and a failed candidate changes nothing. Three guarded mechanisms protect the selection from its own biases: a dual accuracy and calibration criterion, an asymmetric margin against candidates that saw the series during training, and conservative per-window gates. The selection rules themselves were chosen inside a temporal meta-backtest. The best base model served alone reaches a mean MASE rank of 33.8, the tournament served on every configuration reaches 38.0, and the full router reaches 19.4. The routing table, the expert index, the pinned base-model revisions, the submitted score file and the dated snapshot of the public scores are released, and every leaderboard number in this paper regenerates from them by one script.
Chinese Translation
TW3Cast是一个时间序列预测系统，截至2026年9月14日，在GIFT-Eval基准中以平均MASE排名位列130个条目的第3位。其上方的两个条目属于排行榜的智能代理类别，属于使用代理或语言模型进行推理、生成或选择预测的多步系统。TW3Cast不运行任何代理和语言模型。其选择是基于训练集计算的一张表，并随后被冻结，其专家是经过轻微微调的公共基础模型，微调基于这些训练集。对于每个97个数据集、频率和预测范围配置，该表提供四种模式之一：一个专家，即Chronos-2、TiRex或Toto的LoRA或完全微调，其训练数据通过明确规则进行了清理和增强；一个包含专家的分位数混合；一个基础模型的混合；或在从训练集切割的回测中进行的选择锦标赛。表中的每个决策都是基于该回测做出的。专家在击败锦标赛的那一刻被接纳，因此一个候选者的成本仅为几兆字节和几分钟的GPU时间，而失败的候选者不会改变任何结果。三个保护机制防止选择受到自身偏见的影响：双重准确性和校准标准、针对在训练期间看到该系列的候选者的不对称边际，以及保守的每窗口门限。选择规则本身是在一个时间元回测中选择的。单独使用的最佳基础模型的平均MASE排名为33.8，针对每种配置的锦标赛达到38.0，而完整路由器的排名为19.4。路由表、专家索引、固定基础模型修订版、提交的得分文件以及公共得分的日期快照均已发布，本文中的每个排行榜数字均可通过一个脚本从中重新生成。
cs.AI / 3 / 2609.28547
PAWS: Policy-driven Agentic World Simulation
PAWS：政策驱动的自主世界模拟
Abstract
Policy interventions propagate through public communication, institutional decisions, and stakeholder responses, yet datasets for financial multi-agent simulation rarely connect these processes to temporally aligned historical evidence. We introduce PAWS, a Policy-driven Agentic World Simulation dataset covering 36 verified U.S. financial and economic policy episodes, 12,727 policy-linked news records, and 65,291 source-grounded stakeholder actions. Each action is linked to its supporting news and represented by a multi-layer event frame capturing its interaction mode, financial-action family and subtype, semantic attributes, and conditional mappings to external taxonomies. Entities are resolved to normalized organizations, and actions are aligned with daily market-return context to support policy-agent simulation replay. On 2,522 stratified action samples, independent AI and human reviewers achieved 89.4% initial agreement on interaction mode, with disagreements subsequently adjudicated. Case studies of the 2008 short-selling ban and 2001 decimalization recover documented policy timelines and associated market patterns across both dense and sparse news settings. A replay study further shows that high accuracy can mask failure to detect rare stakeholder actions, identifying action timing and calibration as central challenges. PAWS provides an auditable substrate for evaluating agent influence, policy-response cascades, and action-outcome alignment in historically grounded financial simulations.
Chinese Translation
政策干预通过公共沟通、机构决策和利益相关者响应传播，然而金融多代理模拟的数据集很少将这些过程与时间上对齐的历史证据联系起来。我们介绍了PAWS，一个政策驱动的自主世界模拟数据集，涵盖了36个经过验证的美国金融和经济政策事件、12,727条与政策相关的新闻记录以及65,291个基于来源的利益相关者行动。每个行动都与其支持的新闻相联系，并通过多层事件框架表示，捕捉其交互模式、金融行动类别及子类别、语义属性以及与外部分类法的条件映射。实体被解析为标准化的组织，行动与每日市场回报背景对齐，以支持政策代理的模拟重放。在2,522个分层行动样本中，独立的人工智能和人工审阅者在交互模式上达成了89.4%的初步一致，随后对分歧进行了裁决。2008年短卖禁令和2001年小数化的案例研究恢复了记录的政策时间线及其在密集和稀疏新闻环境中的相关市场模式。重放研究进一步表明，高准确率可能掩盖未能检测到稀有利益相关者行动的失败，识别行动时机和校准作为核心挑战。PAWS提供了一个可审计的基础，用于评估代理影响、政策响应级联和历史基础金融模拟中的行动与结果对齐。
cs.AI / 4 / 2609.28554
Pistis Technical Report
Pistis技术报告
Abstract
We introduce the Pistis model family, comprising 27B- and 9B-parameter multimodal large language models built on Qwen3.6 and Qwen3.5, respectively, and developed through a general and scalable post-training framework. The framework first establishes a strong foundation through large-scale multimodal supervised fine-tuning (SFT). Building on this SFT foundation, we propose Interleaved Distillation and Reinforcement Learning (IDRL), a novel post-training paradigm that tightly integrates on-policy distillation and reinforcement learning within a single training loop. By alternating between the two objectives, rather than optimizing either in isolation or combining them in a static joint loss, IDRL enables more effective knowledge transfer, greater optimization stability, and more precise credit assignment for long-horizon agentic trajectories, leading to stronger performance while mitigating common capability trade-offs. At both model scales, the framework produces two specialized variants: Pistis-Thinking, designed to strengthen deep multimodal reasoning, and Pistis-Agentic, which additionally incorporates agentic trajectory data to support long-horizon planning, iterative reasoning, and tool use. Pistis-Agentic is particularly strong in multimodal search. Both scales outperform their corresponding base models. Beyond model-parameter optimization, we further introduce Pistis-Auto-Harnessing (PAH), a system-level method that automatically improves the agent's inference harness through iterative optimization. Experiments demonstrate that PAH enhances the model performance without updating the model parameters or increasing the interaction budget.
Chinese Translation
我们介绍了Pistis模型家族，包括基于Qwen3.6和Qwen3.5构建的27B和9B参数的多模态大型语言模型，采用通用且可扩展的后训练框架进行开发。该框架首先通过大规模多模态监督微调（SFT）建立了坚实的基础。在此SFT基础上，我们提出了交错蒸馏与强化学习（Interleaved Distillation and Reinforcement Learning, IDRL），这是一种新颖的后训练范式，紧密结合了在政策蒸馏和强化学习的单一训练循环中。通过在两个目标之间交替，而不是单独优化其中一个或在静态联合损失中组合它们，IDRL实现了更有效的知识转移、更大的优化稳定性，以及对长时间代理轨迹的更精确的信用分配，从而在减轻常见能力权衡的同时提升了性能。在两个模型规模下，该框架生成了两个专门的变体：Pistis-Thinking，旨在增强深度多模态推理，以及Pistis-Agentic，此外还结合了代理轨迹数据以支持长时间规划、迭代推理和工具使用。Pistis-Agentic在多模态搜索中表现尤为出色。两个规模均优于其对应的基础模型。除了模型参数优化外，我们还进一步引入了Pistis-Auto-Harnessing（PAH），这是一种系统级方法，通过迭代优化自动改善代理的推理框架。实验表明，PAH在不更新模型参数或增加交互预算的情况下提升了模型性能。
cs.AI / 5 / 2609.28557
BaseCamp --- An Agentic AI Framework for Automating DNA Sequencing Data Pipelines
BaseCamp --- 一种用于自动化DNA测序数据管道的代理AI框架
Bandara, Eranga, Liang, Xueping, Gunaratna, Asanga, Hewa, Tharaka, Rahman, Abdul, Foytik, Peter, Bouk, Safdar H., Rajapakse, Sachini, Kularathna, Isurunima, Karunarathna, Pramoda, Rajapakse, Chalani, Keong, Ng Wee, De Zoysa, Kasun, Hass, Amin, Herath, Wathsala, Gore, Ross, Mukkamala, Ravi, Siriwardanagea, Nihal, Siriwardanagea, Gihan, Withanage, Aruna, Loganathan, Nilaan, Shetty, Sachin
Abstract
DNA sequencing pipelines, spanning quality control, alignment, variant calling, and annotation, are now reliably executed by workflow management systems that orchestrate established bioinformatics tools at scale. What remains manual is the decision layer surrounding that execution: selecting quality thresholds appropriate to a sample and platform, adjudicating borderline variant calls, diagnosing anomalies, and determining which findings warrant expert review. These decisions are repetitive, judgment-intensive, inconsistent across operators, and frequently undocumented. This paper introduces BaseCamp, a novel agentic AI framework for automating the decision layer of DNA sequencing pipelines. The framework decomposes the pipeline into six specialized AI agents, covering sample intake and quality control, alignment, variant calling, annotation, cross-stage monitoring, and reporting. Critically, BaseCamp agents do not perform sequence analysis: established tools execute alignment, calling, and annotation, while the agents select among them, configure them, interpret their output, and decide what follows. This confines language model reasoning to the judgment layer where it is reliable and preserves the reproducibility existing tooling guarantees. Agent reasoning is powered by a consortium of fine-tuned, domain-specialized large language models coordinated by a central reasoning LLM, executing locally so no sequencing data leaves the operating environment, under human-in-the-loop orchestration. Evaluation shows agent-generated configurations are concordant with expert practice, that an explicit filtering ledger renders inspectable what filtering otherwise removes without trace, and that cross-stage anomaly detection surfaces conditions execution monitoring misses. BaseCamp offers a generalizable blueprint for agentic automation of scientific data pipelines.
Chinese Translation
DNA测序管道涵盖质量控制、比对、变异调用和注释，目前由工作流管理系统可靠地执行，这些系统在大规模上协调已建立的生物信息学工具。然而，围绕执行的决策层仍然是手动的：选择适合样本和平台的质量阈值、裁定边界变异调用、诊断异常情况以及确定哪些发现需要专家审查。这些决策是重复的、需要判断的、在操作人员之间不一致且常常没有文档记录。本文介绍了BaseCamp，一种用于自动化DNA测序管道决策层的新型代理AI框架。该框架将管道分解为六个专门的AI代理，涵盖样本接收和质量控制、比对、变异调用、注释、跨阶段监控和报告。重要的是，BaseCamp代理不执行序列分析：已建立的工具负责执行比对、调用和注释，而代理则在这些工具之间进行选择、配置、解释其输出并决定后续步骤。这将语言模型推理限制在可靠的判断层，并保持现有工具所保证的可重复性。代理推理由一组经过微调的、领域专门的大型语言模型组成，这些模型由一个中央推理LLM协调，局部执行以确保测序数据不离开操作环境，并在人工干预的协调下进行。评估显示，代理生成的配置与专家实践一致，明确的过滤账本使得可检查的过滤过程显现出过滤所移除的内容，而跨阶段异常检测则揭示了执行监控遗漏的条件。BaseCamp为科学数据管道的代理自动化提供了一个可推广的蓝图。
cs.AI / 6 / 2609.28570
DEEPO: Dual-Entropy Enhanced Policy Optimization for Hallucination in MLLMs
DEEPO：用于多模态大语言模型幻觉的双熵增强策略优化
Abstract
Reinforcement learning (RL) is widely used to sharpen reasoning in multimodal large language models (MLLMs), yet its effect on hallucination is uneven. We trace this to two weak points in the \emph{correction chain} from reward to parameter update. At the rollout level, hard queries---those with high semantic entropy---frequently produce unanimously wrong sample groups, collapsing the group-relative advantage to zero exactly where hallucination risk is highest. At the optimization level, confident-but-wrong tokens are gradient-invisible: a categorical policy's expected score-gradient norm vanishes as its distribution sharpens, so the predictions that most need correction receive the weakest updates. We propose Dual-Entropy Enhanced Policy Optimization (DEEPO), a dual-stage enhancement combining signal variance regularization with gradient preconditioning: semantic-entropy-triggered expert prefixes inject grounded continuations on high-uncertainty queries, providing direct supervision and restoring advantage variance, while advantage-sign-aware Renyi preconditioning counteracts logit-level saturation so correction reaches confident errors in the operational confidence regime. Both branches improve over GRPO individually; their interaction is statistically significant on VideoMMMU---the most complex long-horizon task in our evaluation suite (+4.0$, 95\% CI [1.1, 6.9])---and additive elsewhere. DEEPO reduces hallucination while preserving accuracy and training stability.
Chinese Translation
强化学习（RL）广泛用于提升多模态大语言模型（MLLMs）的推理能力，但其对幻觉的影响却不均衡。我们将其归因于奖励到参数更新的 extit{修正链}中的两个薄弱环节。在展开层面，困难查询——即具有高语义熵的查询——经常产生一致错误的样本组，导致在幻觉风险最高的地方，组相对优势降至零。在优化层面，自信但错误的标记在梯度上是不可见的：分类策略的期望得分梯度范数在其分布变得更加尖锐时消失，因此最需要修正的预测获得的更新最弱。我们提出了双熵增强策略优化（DEEPO），这是一种将信号方差正则化与梯度预处理相结合的双阶段增强方法：语义熵触发的专家前缀在高不确定性查询上注入有根据的延续，提供直接监督并恢复优势方差，而优势符号感知的Renyi预处理则抵消了logit级别的饱和，使得修正能够在操作信心范围内达到自信错误。这两个分支在各自方面均优于GRPO；它们的交互在VideoMMMU上具有统计显著性——这是我们评估套件中最复杂的长时间任务（+4.0$, 95 ext{% CI} [1.1, 6.9]）——并在其他地方具有附加效果。DEEPO在保持准确性和训练稳定性的同时减少了幻觉。
cs.AI / 7 / 2609.28575
TWIST: A Proposed Benchmark for Intervention Quality in Conversational Memory, with a Human-Validated Draft-Alignment
TWIST：一种针对对话记忆干预质量的基准测试提案，附带人类验证的草稿对齐
Abstract
Long-conversation memory benchmarks increasingly test recall and prompted knowledge updates, and recent work studies evolving user beliefs and memory state. TWIST is a proposed benchmark suite for a complementary, unmeasured property: intervention quality -- whether a deployed memory system, exercised through its own ingest/recall/vet surface, acts correctly at belief change points. Four tracks cover unprompted tension detection, vetting outgoing drafts against the record, answering with current beliefs while preserving supersession history, and governing sensitive recall. The suite extends LoCoMo's corpora and harness, pairing every detect/block metric with a matched do-not-over-detect control: surface-matched hard negatives price false intervention, so no track can be gamed by flagging everything. The benchmark itself is validated first: independent, gold-blind double annotation with adjudication, judge decoy calibration, and a separability audit. On the human-validated Track B v1.0 key (161 items, post-adjudication kappa = 0.85), no tested configuration simultaneously achieves high contradiction recall, high hard-negative specificity, and high attribution: flat-RAG baselines detect 0.76-0.97 of true contradictions but falsely flag 16-43% of surface-matched safe drafts depending on backend, while a deployed coherence-oriented system almost never over-flags (0.98-1.00 specificity) yet catches 42% of true contradictions -- a trade-off no recall-only score can see. A 13-configuration baseline ladder localizes causes: every gold contradiction is detectable from its evidence alone (recall 1.000), calibrated models nearly solve the track given the full transcript -- consistent with substantial retrieval-coverage gaps -- and draft-only floors reveal model-dependent style priors. A system's TWIST profile, beside its recall score, measures whether memory knows when to intervene and when not to.
Chinese Translation
长对话记忆基准测试越来越多地考察回忆和提示知识更新，近期的研究关注用户信念和记忆状态的演变。TWIST是一套针对一个补充性、未测量属性的基准测试提案：干预质量——即一个部署的记忆系统在信念变化点是否能够正确地发挥作用，通过其自身的摄取/回忆/审查界面进行操作。该基准测试包含四个轨道，涵盖无提示的紧张检测、对照记录审核外发草稿、在保留超越历史的同时用当前信念回答，以及管理敏感回忆。该套件扩展了LoCoMo的语料库和工具，将每个检测/阻止指标与匹配的不超检测控制相结合：表面匹配的硬负样本会导致错误干预，因此没有轨道可以通过标记所有内容来操控。基准测试本身首先经过验证：独立的金盲双重注释与裁决、评审伪装校准和可分离性审计。在人类验证的Track B v1.0关键（161个项目，裁决后kappa = 0.85）中，没有测试配置能够同时实现高矛盾回忆、高硬负特异性和高归因：平坦的RAG基线检测到0.76-0.97的真实矛盾，但根据后端错误标记16-43%的表面匹配安全草稿，而一个部署的以一致性为导向的系统几乎从不错误标记（0.98-1.00特异性），却捕捉到42%的真实矛盾——这是任何仅依赖回忆的评分无法看到的权衡。一个13配置基线阶梯定位了原因：每个金矛盾都可以仅从其证据中检测到（回忆率1.000），经过校准的模型几乎可以解决给定完整转录的轨道——与显著的检索覆盖差距一致——而仅草稿的底线揭示了模型依赖的风格先验。一个系统的TWIST特征，除了其回忆评分外，还衡量记忆是否知道何时干预以及何时不干预。
cs.AI / 8 / 2609.28609
Adversarial Closed-Loop Curriculum for Evolving Role-Playing Agents
对抗性闭环课程用于演化角色扮演代理
Abstract
Role-playing agents based on large language models have been widely applied in areas such as personalized assistance and social simulation. Recent RL methods typically train on a fixed scenario pool collected before learning begins. This creates a distributional bottleneck: as the agent improves, the scenarios where it performs poorly also change, while the training distribution remains static. Therefore, we propose AdvRole, an adversarial context rewriting framework that turns role-playing RL into a closed-loop curriculum. AdvRole alternates between an Actor that learns to role-play and a Rewriter that edits character profiles and dialogue contexts into actor-specific hard scenarios. The Rewriter is trained with a performance-gap reward, which favors rewrites that reduce the current Actor's score relative to the original scenario. As a result, the scenario pool evolves with the Actor and continuously targets under-mastered regions of the character-context space. Experiments on three role-playing benchmarks covering English and Chinese, as well as a new multilingual benchmark we release, show that AdvRole consistently outperforms baselines.
Chinese Translation
基于大型语言模型的角色扮演代理已广泛应用于个性化助手和社会模拟等领域。近期的强化学习（RL）方法通常在学习开始前收集的固定场景池上进行训练。这造成了分布瓶颈：随着代理的改进，其表现不佳的场景也会发生变化，而训练分布保持静态。因此，我们提出了AdvRole，一个对抗性上下文重写框架，将角色扮演强化学习转变为闭环课程。AdvRole在学习角色扮演的演员（Actor）和编辑角色档案及对话上下文的重写器（Rewriter）之间交替进行。重写器通过性能差距奖励进行训练，偏向于那些相对于原始场景降低当前演员得分的重写。因此，场景池随着演员的进步而演变，并持续针对角色-上下文空间中未掌握的区域进行训练。在涵盖英语和中文的三个角色扮演基准测试以及我们发布的一个新的多语言基准测试上的实验表明，AdvRole始终优于基线方法。
cs.AI / 9 / 2609.28654
Training Object Permanence in World Models
在世界模型中训练物体持久性
Zhang, Haotian, Yu, Fengyuan, Luo, Dezhi, Sun, Haoran, Zhao, Zehong, Gao, Qingying, Li, Yihan, An, Siyuan, Qin, Huayi, Zhang, Yilan, Jiang, Zhengze, Feng, Pinyuan, Zhang, Renrui, Guo, Ziyu, Wang, Letian, Yang, Mengyue, Mei, Kangfu, Wang, Maijunxian, Ji, Ran, Kumar, Vikash, Shi, Freda, Sripada, Chandra, Muller, Vincent C., Torr, Philip, Yuille, Alan, Kriegeskorte, Nikolaus, Juefei-Xu, Felix, Zhang, Lvmin, Chen, Jieneng, Du, Yilun, Deng, Hokin
Abstract
Object permanence and solidity are hallmarks of human cognitive priors. Recent studies show that video generation models, a paradigmatic class of current world models, have begun to show emerged reasoning abilities, making them ideal candidates for building human-like physical intelligence. Do video models have emerged object permanence in them? If not, could we train them with a core-cognition inspired dataset? We introduce WROP (World Reasoning with Object Permanence), a data infrastructure of 150 hand-designed cognitive science inspired tasks, divided into six cognitive categories. We build Blender generators that randomize speed, lighting, camera angle, and other nuisance parameters while preserving each task's cognitive structure, yielding 10,000+ samples per task. We release a 1.5M-sample training corpus and a 300-question exam. On this exam we evaluate 14 video models: 3 reference-to-video, 7 edit, and 4 continuation, among which PWM-WROP, our 16B world model. In a blind pairwise Elo study, PWM-WROP ranks first among continuation models and third overall, behind only a statistical tie between two reference-to-video models. We release the data, exam, model answers, scores, weights, and PWM, our native-PyTorch training stack on AWS Trainium2.
Chinese Translation
物体持久性和坚固性是人类认知先验的标志。最近的研究表明，视频生成模型作为当前世界模型的典型类别，已经开始展现出涌现的推理能力，使其成为构建类人物理智能的理想候选者。那么，视频模型中是否具备涌现的物体持久性？如果没有，我们能否通过一个以核心认知为灵感的数据集来训练它们？我们引入了WROP（带有物体持久性的世界推理），这是一个包含150个手工设计的、受认知科学启发的任务的数据基础设施，这些任务被划分为六个认知类别。我们构建了Blender生成器，随机化速度、光照、相机角度和其他干扰参数，同时保持每个任务的认知结构，从而为每个任务生成超过10,000个样本。我们发布了一个包含150万样本的训练语料库和一个300题的考试。在这个考试中，我们评估了14个视频模型：3个参考视频模型、7个编辑模型和4个延续模型，其中PWM-WROP是我们的16B世界模型。在一项盲对比Elo研究中，PWM-WROP在延续模型中排名第一，在所有模型中排名第三，仅次于两个参考视频模型之间的统计平局。我们发布了数据、考试、模型答案、分数、权重，以及PWM，我们在AWS Trainium2上的原生PyTorch训练框架。
cs.AI / 10 / 2609.28690
Beyond Surface Style: Aligning Multi-Turn User Simulators with Behavioral Consistency
超越表面风格：多轮用户模拟器与行为一致性的对齐
Abstract
Faithful user simulation is fundamental to building, evaluating, and improving interactive AI at scale. However, plausible individual responses do not ensure that simulated users reproduce the intent evolution and outcomes observed in real interactions. We propose TRACER, a multi-turn user simulator that explicitly models users' evolving intent and learns to align simulated behavior with real interaction trajectories. TRACER is trained in two stages: supervised fine-tuning on real user dialogues, followed by multi-turn reinforcement learning. The RL stage combines hierarchical outcome- and trajectory-level rewards with deviation-aware advantage modulation, jointly mitigating reward sparsity and credit assignment in long dialogues. On real customer-service sessions organized into reference cohorts, TRACER-7B surpasses the strongest baseline by 11.4 conversion F1, while also achieving the lowest group-level conversion-rate error and semantic trajectory distance, and generalizing to out-of-distribution scenarios. Human Turing tests yield identification accuracy close to chance, supporting the perceived naturalness of generated conversations. Building on this simulator, we further introduce the Dynamic Marketing Benchmark, which jointly evaluates persuasion effectiveness and response quality of LLMs through simulated interactions, revealing that higher response quality does not necessarily correspond to higher conversion rates.
Chinese Translation
忠实的用户模拟对于大规模构建、评估和改进交互式人工智能至关重要。然而，可信的个体响应并不能确保模拟用户重现真实交互中观察到的意图演变和结果。我们提出了TRACER，一种多轮用户模拟器，明确建模用户的演变意图，并学习将模拟行为与真实交互轨迹对齐。TRACER的训练分为两个阶段：首先在真实用户对话上进行监督微调，然后进行多轮强化学习（RL）。RL阶段结合了层次化的结果和轨迹级奖励，以及对偏差的关注优势调制，共同缓解了长对话中的奖励稀疏性和信用分配问题。在组织成参考队列的真实客户服务会话中，TRACER-7B的表现超越了最强基线11.4的转化F1，同时实现了最低的组级转化率误差和语义轨迹距离，并能够推广到分布外场景。人类图灵测试的识别准确率接近随机水平，支持生成对话的自然性。基于这个模拟器，我们进一步引入了动态营销基准，通过模拟交互共同评估大型语言模型（LLMs）的说服效果和响应质量，揭示了更高的响应质量并不一定对应于更高的转化率。
cs.AI / 11 / 2609.28692
Driving Epidemic Models with AI Agents: the Epydemix Agent Framework
利用人工智能代理驱动流行病模型：Epydemix代理框架
Abstract
Artificial Intelligence agents based on large language models provide convenient natural language interfaces to scientific software, but reliability is not automatic. Here we introduce the Epydemix Agent Framework, an additive layer over Epydemix, an open-source Python library for stochastic compartmental epidemic modeling. The framework extends the library with four capabilities to facilitate interaction with an AI agent: discovery of available models and parameters, preventive validation of a declarative scenario specification, execution through tested library code, and inspectability of results. These capabilities let an agent handle the entire modeling process, from the natural-language description of the scenario to quantitative results, figures, and interpretation of findings without writing custom code. Each step reads input files and saves results in a separate output bundle, making the process auditable and reproducible. First, we show the end-to-end workflow with a case study comparing vaccination strategies for a novel respiratory virus. Second, we assessed the framework across 50 agent sessions and five modeling tasks by comparing the agent use of the framework against the direct use of the Python interface. The framework reduced turns, output tokens, and cost on most tasks, unless it trades resources for per-point reproducibility.
Chinese Translation
基于大型语言模型的人工智能代理为科学软件提供了便捷的自然语言接口，但可靠性并不是自动的。在此，我们介绍了Epydemix代理框架，这是一个建立在Epydemix之上的附加层，Epydemix是一个用于随机分隔流行病建模的开源Python库。该框架通过四个功能扩展了该库，以促进与人工智能代理的互动：发现可用模型和参数、对声明性场景规范进行预防性验证、通过经过测试的库代码执行以及结果的可检查性。这些功能使代理能够处理整个建模过程，从场景的自然语言描述到定量结果、图表和研究发现的解释，而无需编写自定义代码。每个步骤读取输入文件，并将结果保存在单独的输出包中，使过程可审计和可重复。首先，我们展示了一个端到端的工作流程案例研究，比较了新型呼吸病毒的疫苗接种策略。其次，我们通过将代理使用框架与直接使用Python接口进行比较，评估了该框架在50个代理会话和五个建模任务中的表现。该框架在大多数任务中减少了回合、输出令牌和成本，除非它在每个点的可重复性上进行了资源交换。
cs.AI / 12 / 2609.28693
Progressive Skill Discovery as Access Control for Tool-Using LLM Agents: Structural Governance through Role-Scoped Capability Delivery
渐进技能发现作为工具使用大型语言模型代理的访问控制：通过角色范围能力交付的结构治理
Abstract
Large Language Model (LLM) agents struggle to scale safely when exposed to vast enterprise toolsets. Providing an agent with access to every internal tool leads to oversized context windows, degraded tool selection, and severe governance vulnerabilities - as system policies defined purely in prompts remain probabilistic advice rather than hard constraints. Existing mitigations, such as multi-agent domain delegation, decentralize audit logs and fail to guarantee policy compliance across sessions. We introduce skilder, a framework that packages capabilities into roles: bundles of skills, tools, and instructions, together with the limits that bound them. An agent begins with a minimal role catalog, learns the roles a task requires, and receives each role's skills, instructions, and tools through a single MCP server. Because tools reach the agent only inside learned skills, the same server enforces the scope of what was learned deterministically. We evaluate skilder against flat-context tool selection and multi-agent orchestration across 13 tasks using six models (10 runs each). Our results show that, when models completed discovery and issued a governed call, the skilder simulated authorization layer enforced governance boundaries: no unauthorized tool call or parameter violation (e.g., a spending-limit breach) executed. Aggregate task pass rates also reflect whether each model followed the discovery protocol and satisfied response-quality checks; those misses are not authorization failures. Furthermore, by allowing agents to dynamically acquire cross-role capabilities mid-task, skilder preserves problem-solving flexibility while providing hard system-level enforcement.
Chinese Translation
大型语言模型（LLM）代理在接触广泛的企业工具集时难以安全扩展。为代理提供对每个内部工具的访问会导致过大的上下文窗口、工具选择能力下降以及严重的治理漏洞——因为仅在提示中定义的系统政策仍然是概率性建议，而非严格约束。现有的缓解措施，如多代理领域委派，虽然去中心化了审计日志，但未能保证跨会话的政策合规性。我们提出了 skilder，一个将能力打包成角色的框架：技能、工具和指令的组合，以及界定它们的限制。代理从一个最小的角色目录开始，学习任务所需的角色，并通过单个 MCP 服务器接收每个角色的技能、指令和工具。由于工具仅在学习的技能内到达代理，因此同一服务器以确定性方式强制执行所学内容的范围。我们在 13 个任务中使用六个模型（每个模型进行 10 次运行）评估了 skilder 与平面上下文工具选择和多代理编排的效果。我们的结果表明，当模型完成发现并发出受治理的调用时，skilder 模拟的授权层强制执行治理边界：没有执行未经授权的工具调用或参数违规（例如，超支限制）。聚合任务通过率也反映了每个模型是否遵循了发现协议并满足响应质量检查；这些遗漏并不是授权失败。此外，通过允许代理在任务中动态获取跨角色能力，skilder 在提供严格的系统级执行的同时，保持了问题解决的灵活性。
cs.AI / 13 / 2609.28765
Reinforcement Learning with Verifiable Rewards for Small Search Agents
具有可验证奖励的小型搜索代理的强化学习
Abstract
Reinforcement Learning with Verifiable Rewards (RLVR) performs well on problems with clear rewards, such as mathematics and coding, but whether it also works where the reward is less clear remains open. The reason-over-search recipe applies RLVR to open-domain question answering, where retrieval grounds the answer and a match against the reference supplies the reward. So far it has been demonstrated on large models, and below one billion parameters only with distillation from a larger teacher. We test the recipe on a small model. We train Qwen3.5-0.8B with Group Relative Policy Optimization (GRPO) and an interleaved Wikipedia-search tool on MuSiQue, varying only the reward across three shapes over three seeds each, and we evaluate every checkpoint held-out on a seven-benchmark question-answering suite. The recipe works: the best run reaches 0.352 average exact match against a 0.092 untrained floor, a 3.8-fold gain, with no distillation step in the training loop. The reward shape also matters. The Search-R1-faithful exact-match-only reward is the worst of the three at every seed at the matched training horizon, and it is worst even on exact match, the metric it directly optimises. We conclude that the sparse exact-match reward, RLVR's default in mathematics and code, is the wrong starting point for models of this size. The reason-over-search setting can supply a suitable reward for RLVR on small models, but small-model RLVR needs its own reward-design study rather than a scaled-down copy of a large-model recipe.
Chinese Translation
具有可验证奖励的强化学习（RLVR）在奖励明确的问题上表现良好，例如数学和编码，但在奖励不太明确的情况下是否也有效仍然是一个悬而未决的问题。推理-搜索方法将RLVR应用于开放域问答，其中检索为答案提供基础，而与参考的匹配则提供奖励。到目前为止，这一方法已在大型模型上得到验证，而在参数少于十亿的模型上仅通过从更大教师模型的蒸馏实现。我们在一个小型模型上测试了这一方法。我们使用群体相对策略优化（GRPO）和交错的维基百科搜索工具在MuSiQue上训练Qwen3.5-0.8B，仅在三个种子上对三种奖励形状进行变化，并在七个基准问答套件上评估每个检查点。结果表明，该方法有效：最佳运行在0.092的未训练基准上达到了0.352的平均精确匹配，提升了3.8倍，且训练过程中没有蒸馏步骤。奖励形状也很重要。在每个种子下，Search-R1-faithful仅精确匹配的奖励在匹配的训练范围内是三者中最差的，甚至在其直接优化的精确匹配指标上也表现最差。我们得出结论，稀疏的精确匹配奖励，RLVR在数学和代码中的默认设置，对于这种规模的模型来说是错误的起点。推理-搜索设置可以为小型模型的RLVR提供合适的奖励，但小型模型的RLVR需要独立的奖励设计研究，而不是大型模型方法的缩小版。
cs.AI / 14 / 2609.28771
Agent Memory with Episodic Retrieval for Financial Decision-Making
具有情景检索的代理记忆在金融决策中的应用
Abstract
Large language models (LLMs) have demonstrated strong capabilities in financial analysis and reasoning, inspiring recent advances in agent-based trading frameworks. While these systems show promise, prior approaches either emphasize long-horizon forecasting or operate as stateless analyzers, limiting their applicability to the demands of trading in complicated settings. To address these gaps, we introduce META (Memory Enhanced Trading Agent), the first RAG-like episodic-memory-augmented multi-agent framework for financial decision making. META integrates a family of specialized indicator agents (e.g., Trend, MACD, Stochastic, RSI, SMA, AVWAP, Heikin-Ashi) with a Decision Agent that fuses their reports, and a Memory module that retrieves and updates past trading episodes encoded as market state embeddings with outcomes and reflections. By recalling relevant experiences and adaptively reweighting signals under similar market regimes, META achieves improved directional accuracy and robustness under short-horizon evaluation. Our results demonstrate that episodic memory provides a powerful mechanism for regime-aware, interpretable, and low-latency decision-making in trading and decision making. The code of this project is released on GitHub.
Chinese Translation
大型语言模型（LLMs）在金融分析和推理方面展现出强大的能力，激发了基于代理的交易框架的最新进展。尽管这些系统显示出良好的前景，但以往的方法要么强调长期预测，要么作为无状态分析器运作，这限制了它们在复杂环境中交易需求的适用性。为了解决这些问题，我们提出了META（记忆增强交易代理），这是第一个类似RAG的情景记忆增强多代理框架，用于金融决策。META集成了一系列专门的指标代理（例如，趋势、MACD、随机指标、相对强弱指数、简单移动平均、加权平均价格、Heikin-Ashi）与一个决策代理，该代理融合了它们的报告，以及一个内存模块，该模块检索并更新编码为市场状态嵌入的过去交易情景，包括结果和反思。通过回忆相关经验并在类似市场环境下自适应地重新加权信号，META在短期评估中实现了更高的方向准确性和稳健性。我们的结果表明，情景记忆为交易和决策提供了一种强大的机制，能够实现对市场状态的感知、可解释性和低延迟决策。该项目的代码已在GitHub上发布。
cs.AI / 15 / 2609.28776
Learned Cross-Task Relationships in Multi-Task Models
多任务模型中的学习跨任务关系
Abstract
We propose a framework that learns cross-task relationships in multi-task models by approximating the joint distribution of task labels through targeted pairwise relationships. This approach improves performance via transfer learning and enhances information extraction without the intractable complexity of modeling the full joint space. Although our framework applies to any multi-task system, we demonstrate its efficacy within YouTube's production recommendation systems. Experiments across the Notifications, Homepage, and Watch Next surfaces show improvements in both accuracy and user satisfaction metrics. Finally, we propose a workflow template to facilitate broader future implementation.
Chinese Translation
我们提出了一种框架，通过针对性的成对关系来近似任务标签的联合分布，从而学习多任务模型中的跨任务关系。这种方法通过迁移学习提高了性能，并在不需要建模完整联合空间的复杂性的情况下增强了信息提取。尽管我们的框架适用于任何多任务系统，但我们在YouTube的生产推荐系统中展示了其有效性。在通知、主页和观看下一个界面的实验显示了准确性和用户满意度指标的改善。最后，我们提出了一个工作流程模板，以促进未来更广泛的实施。
cs.AI / 16 / 2609.28850
RECLAIM: Can Agents Reproduce the Claims of Machine Learning Papers?
RECLAIM：智能体能否重现机器学习论文的主张？
Abstract
Reproducing a machine learning paper involves most research steps, from installing software and debugging to running experiments, work that AI agents increasingly do. We introduce RECLAIM, a benchmark of 100 NeurIPS 2025 papers that can be rebuilt yearly from new conferences. For each paper we fix in advance the result to reproduce, what counts as a successful reproduction, and a GPU-hour budget. An agent must reproduce that result using the paper and whatever its authors released. What the authors released decides the difficulty tier. Run-tier releases include code, data, and weights; Retrain-tier releases lack weights, so the agent trains the model; Reimplement-tier releases lack code, so the agent writes it. A separate language model grades runs from logs and outputs rather than agents' reports. We run four agents once per paper; the best agent in each tier reproduces only 41% of Run-tier papers, 27% at Retrain, and 15% at Reimplement, where every agent does worst. Failed attempts use on average 29% of their budget, so most stop with budget left. The most common agent error is writing the method without checking any part against the paper's numbers, in 63 of 400 runs.
Chinese Translation
重现一篇机器学习论文涉及大多数研究步骤，从安装软件和调试到运行实验，这些工作越来越多地由人工智能智能体完成。我们介绍了RECLAIM，这是一个包含100篇NeurIPS 2025论文的基准，每年可以从新的会议中重建。对于每篇论文，我们提前确定要重现的结果、成功重现的标准以及GPU小时预算。智能体必须使用论文及其作者发布的任何内容来重现该结果。作者发布的内容决定了难度等级。运行级发布包括代码、数据和权重；重训练级发布缺少权重，因此智能体需要训练模型；重实现级发布缺少代码，因此智能体需要编写代码。一个独立的语言模型根据日志和输出对运行进行评分，而不是依赖智能体的报告。我们对每篇论文运行四个智能体；每个等级中表现最好的智能体仅重现了41%的运行级论文，重训练级论文为27%，重实现级论文为15%，而每个智能体在重实现级的表现最差。失败的尝试平均使用了29%的预算，因此大多数在预算剩余的情况下停止。最常见的智能体错误是在未检查任何部分与论文数字的对比的情况下编写方法，在400次运行中出现了63次。
cs.AI / 17 / 2609.28859
Human-AI-Powered Hypothesis Testing: Cost-Aware Selective AI Scoring and Sequential Human Escalation
人机协作的假设检验：成本意识的选择性AI评分与顺序人类升级
Abstract
Large language models are increasingly used as inexpensive judges to evaluate outputs, label data, and assess whether a system meets a desired quality standard. Yet using AI judgments for formal statistical inference is fundamentally different from simply treating them as ground-truth labels: AI evaluations can be biased or noisy, and rigorous hypothesis testing requires explicit control of type-I and type-II errors. We study how to use AI judgments, together with selective human verification, to conduct a valid hypothesis test at minimum cost. We consider a population of items with hidden binary labels. After choosing a fixed pool of items, the decision maker can selectively query AI, send an item directly to a human, escalate an AI-scored item to a human after observing the AI report, or stop once sufficient evidence has accumulated. We derive an information-theoretic lower bound that captures the minimum cost of achieving prescribed testing errors and characterizes the value of AI information and human verification through a report-dependent information frontier. Motivated by this characterization, we develop SCALE, a sequential cost-aware policy that combines selective AI scoring with adaptive human escalation. SCALE is valid at finite sample sizes and matches the lower bound to first order as the target error probabilities vanish. We further extend the framework to an unknown AI-output model using paired AI-human pilot data. Numerically, SCALE approaches Human-only or AI-only testing when one source clearly dominates, while achieving its largest savings when inexpensive AI judgments and selective human verification are both valuable.
Chinese Translation
大型语言模型越来越多地被用作廉价的评判者，以评估输出、标注数据并评估系统是否符合所需的质量标准。然而，将AI判断用于正式统计推断与仅仅将其视为真实标签是根本不同的：AI评估可能存在偏差或噪声，而严格的假设检验需要明确控制第一类和第二类错误。我们研究如何利用AI判断以及选择性的人类验证，以最低成本进行有效的假设检验。我们考虑一个具有隐藏二元标签的项目群体。在选择固定的项目池后，决策者可以选择性地查询AI，直接将项目发送给人类，在观察AI报告后将AI评分的项目升级给人类，或者在积累足够证据后停止。我们推导出一个信息论下界，捕捉实现规定检验错误的最低成本，并通过报告依赖的信息边界来表征AI信息和人类验证的价值。受到这一表征的启发，我们开发了SCALE，这是一种顺序成本意识策略，结合了选择性AI评分与自适应人类升级。SCALE在有限样本大小下是有效的，并且在目标错误概率消失时与下界在一阶上匹配。我们进一步将该框架扩展到使用配对的AI-人类试点数据的未知AI输出模型。在数值上，当一个来源明显占主导地位时，SCALE接近仅人类或仅AI的测试，而在廉价的AI判断和选择性的人类验证都具有价值时，获得最大的节省。
cs.AI / 18 / 2609.28876
Forecast-Dojo: Replayable Environments for Benchmarking and Training LLM Forecasting Agents
Forecast-Dojo：用于基准测试和训练LLM预测代理的可重放环境
Abstract
We introduce Forecast-Dojo, a replayable environment for benchmarking and training LLM forecasting agents. It combines resolved prediction-market questions with dated news, allowing agents to research an event and revisit their predictions at successive historical dates. The same tasks and tools support repeated evaluation, collection of training interactions, and feedback from recorded outcomes without waiting for new events to resolve. Forecast-Dojo contains 1,568 Polymarket events, split by time into training and evaluation periods, and 18.8M dated news articles. In an evaluation of 12 models, research tools lower Brier score for all 12. Forecasts also improve as events unfold, with the largest gains at steps where more newly dated evidence is recorded. Every model still trails historical market forecasts in both Brier score and accuracy. A belief notebook carried between dates lowers research cost but does not consistently improve forecast quality. Beyond evaluation, Forecast-Dojo provides interaction trajectories and outcome feedback for agent learning, with supervised fine-tuning as a proof of concept.
Chinese Translation
我们介绍了Forecast-Dojo，这是一个用于基准测试和训练LLM预测代理的可重放环境。它结合了解决的预测市场问题和带日期的新闻，允许代理研究事件并在后续历史日期重新审视他们的预测。相同的任务和工具支持重复评估、收集训练交互以及从记录的结果中获得反馈，而无需等待新事件的解决。Forecast-Dojo包含1,568个Polymarket事件，按时间分为训练和评估阶段，以及18.8M篇带日期的新闻文章。在对12个模型的评估中，研究工具降低了所有12个模型的Brier分数。随着事件的发展，预测也有所改善，最大的提升发生在记录了更多新日期证据的步骤上。每个模型在Brier分数和准确性上仍然落后于历史市场预测。跨日期携带的信念笔记本降低了研究成本，但并未始终提高预测质量。除了评估，Forecast-Dojo还提供了交互轨迹和结果反馈，以支持代理学习，并以监督微调作为概念验证。
cs.AI / 19 / 2609.28919
Control the Harness, Control the Cost: Routing and Governing AI Coding Agents in the Enterprise
控制工具，控制成本：企业中人工智能编码代理的路由与管理
Abstract
Harnesses, the products that run AI coding agents, are multiplying, and enterprises are rolling them out to their employees: what started as pilots with a few hundred seats is scaling to tens of thousands. Most enterprises do not build these harnesses but buy them from large vendors, such as Anthropic's Claude Code or OpenAI's Codex. A harness decides which model answers, what the model reads, how the prompt cache is used and which subagents run, so it picks the rate on the price sheet and sets the volume bought at it. Enterprises that keep a proprietary or untuned harness at its defaults inherit these choices and their bill. We build a fast, customisable router in which Jev, a classifier with calibrated probabilities, labels every prompt against a bring-your-own taxonomy of agentic requests. Because one user turn is many requests over a prompt cache that belongs to one model, the router moves work only where no running conversation has to rebuild its cache: at session start, in side lanes and at subagent launch. From the price sheet we derive when a mid-task switch pays back, and a crossover: on long tool-heavy sessions the highest-priced model costs less than the next tier, as repricing about 10,000 real sessions from public datasets confirms. In an emulated enterprise of 10,000 seats with user behaviour taken from these datasets, the router recovers 14 to 21% of model spend at Anthropic's list prices of 21 September 2026, \$3.3M to \$5.0M a year. The paper also maps the risks across twenty harnesses, prices the dependence on one vendor's models, and proposes a control plane that enterprises can run from within, starting now, with a ladder for deciding later whether to own the harness.
Chinese Translation
工具，即运行人工智能编码代理的产品，正在迅速增加，企业也在将其推广给员工：最初的试点项目从几百个席位扩展到数万个。大多数企业并不自行构建这些工具，而是从大型供应商处购买，例如Anthropic的Claude Code或OpenAI的Codex。工具决定了哪个模型进行回答、模型读取的内容、提示缓存的使用方式以及运行的子代理，因此它选择价格表上的费率并设定购买的数量。保持专有或未调优的默认工具的企业继承了这些选择及其账单。我们构建了一个快速、可定制的路由器，其中Jev，一个具有校准概率的分类器，对每个提示进行标记，以便与自定义的代理请求分类法进行匹配。由于一个用户的操作涉及多个请求，而这些请求属于一个模型的提示缓存，因此路由器仅在没有正在进行的对话需要重建缓存的地方移动工作：在会话开始时、在侧通道中以及在子代理启动时。从价格表中，我们推导出在任务中期切换何时能够带来回报，以及交叉点：在长时间的工具密集型会话中，最高价模型的成本低于下一个层级，因为对来自公共数据集的约10,000个真实会话进行重新定价证实了这一点。在一个模拟的拥有10,000个席位的企业中，用户行为取自这些数据集，路由器在Anthropic于2026年9月21日的定价下，能够回收14%至21%的模型支出，年支出为330万美元至500万美元。本文还映射了二十个工具的风险，评估了对单一供应商模型的依赖，并提出了一个企业可以立即启动的控制平面，以便在未来决定是否拥有工具的阶梯。
cs.AI / 20 / 2609.28921
PFArena: Benchmarking Language Models for Protein Modification
PFArena：蛋白质修饰语言模型基准测试
Abstract
Protein modification requires navigating an immense sequence space, yet wet-lab validation remains low-throughput and costly. Although computational paradigms including protein language models (PLMs), large language models (LLMs), and LLM-based agents have shown promise in protein modification, their relative efficacy across realistic experimental decision-making settings remains unclear. To bridge this gap, we introduce PFArena, a benchmark comprising four controlled task interfaces that cover single-mutant generation and multi-mutant ranking. By providing varying levels of mutation fitness data, PFArena reflects four representative research scenarios characterized by differing degrees of prior experimental context. We assess six PLMs, six LLMs, and five LLM-based agents using complementary metrics to measure both peak and overall protein modification performance. Our evaluation reveals that model performance shifts systematically with the availability of target-specific experimental evidence: PLMs demonstrate proficiency in open-ended single-mutant generation by leveraging protein-specific priors, whereas LLMs and agents perform strongly in multi-mutant ranking, particularly when target-specific fitness data are available. Nevertheless, all model families face fundamental challenges with increasing search-space size and mutation depth. We release our code and benchmark suite to facilitate reproducible research in model-assisted protein modification.
Chinese Translation
蛋白质修饰需要在庞大的序列空间中进行探索，但实验室验证的通量仍然较低且成本高昂。尽管包括蛋白质语言模型（PLMs）、大型语言模型（LLMs）和基于LLM的代理在内的计算范式在蛋白质修饰中展现了潜力，但它们在现实实验决策环境中的相对有效性仍不明确。为了解决这一问题，我们推出了PFArena，一个基准测试，包含四个受控任务接口，涵盖单突变体生成和多突变体排序。通过提供不同水平的突变适应性数据，PFArena反映了四种具有不同实验背景程度的代表性研究场景。我们使用互补指标评估了六个PLMs、六个LLMs和五个基于LLM的代理，以测量蛋白质修饰的峰值和整体性能。我们的评估结果显示，模型性能随着目标特定实验证据的可用性而系统性变化：PLMs通过利用蛋白质特定的先验知识在开放式单突变体生成中表现出色，而LLMs和代理在多突变体排序中表现强劲，尤其是在目标特定适应性数据可用时。然而，所有模型家族在搜索空间大小和突变深度增加时都面临基本挑战。我们发布了我们的代码和基准测试套件，以促进模型辅助蛋白质修饰的可重复研究。
cs.AI / 21 / 2609.28942
From Static Personal Values to Contextualized Personalization: Bayesian Personalized Value Alignment for LLMs
从静态个人价值观到情境化个性化：针对大型语言模型的贝叶斯个性化价值对齐
Abstract
Personalized value alignment has become increasingly important as large language models (LLMs) are expected to accommodate diverse user preferences. However, existing methods typically align model outputs with a static value profile across prompts, overlooking that the salience of value dimensions varies substantially across contexts. Inspired by Lewin's Field Theory, which views human behavior as jointly shaped by personal dispositions and situational constraints, we model personal values as priors and context-dependent preferences as posteriors. We propose BaCVA, an inference-time Bayesian Context-aware personalized Value Alignment method that approximates posterior personalized preferences by integrating static personal values with scenario-specific value salience. BaCVA first estimates contextual value salience from generally normative responses, and then employs a dual-view personalization module to infer posterior preferences from complementary personal-value and scenario-driven perspectives. This Bayesian formulation enables more accurate and adaptive personalized value alignment while improving data efficiency via prior values. Extensive experiments on benchmarks demonstrate its superiority over strong baselines.
Chinese Translation
随着大型语言模型（LLMs）被期望能够适应多样化的用户偏好，个性化价值对齐变得越来越重要。然而，现有方法通常在提示中使用静态价值配置来对齐模型输出，忽视了价值维度在不同情境下的重要性差异。受到Lewin场理论的启发，该理论认为人类行为是个人倾向与情境约束共同作用的结果，我们将个人价值视为先验，将情境依赖的偏好视为后验。我们提出了BaCVA，一种推理时的贝叶斯情境感知个性化价值对齐方法，通过将静态个人价值与情境特定的价值重要性相结合，近似后验个性化偏好。BaCVA首先从一般规范性反应中估计情境价值的重要性，然后采用双视角个性化模块，从互补的个人价值和情境驱动的视角推断后验偏好。这种贝叶斯模型能够实现更准确和适应性的个性化价值对齐，同时通过先验价值提高数据效率。大量基准实验表明，其优于强基线方法。
cs.AI / 22 / 2609.28963
Back to the Definition: Estimating Step-Level Advantages via Trajectory Graphs for Agentic Reinforcement Learning
回归定义：通过轨迹图估计代理强化学习的步骤级优势
Abstract
Group-based reinforcement learning (RL) methods, such as GRPO and its variants, have become a leading paradigm for training reasoning and agentic large language models (LLMs). While their group-normalized advantage estimation is reliable at the response level, it becomes systematically biased at the step level, since coarse-grained trajectory-level advantages are hard to accurately reflect the contribution of individual steps (i.e, failed trajectories may contain valuable steps). Revisiting the foundational RL definition, we notice that GRPO's success on single-turn tasks stems from its advantage estimation strategy, which adheres to the basic definition: the mean reward of multiple actions sampled from the same state constitutes a credible state-value estimate. Extending the faithful estimation to step-level would in principle demand sampling multiple actions from each intermediate state, which is too costly on a per-state basis. To mitigate this issue, we propose a Graph-based Faithful sTep-level credit-assignment framework (GRAFT) that grafts all rollout trajectories into a trajectory graph, recovering node state-values via Bellman iteration on the graph, and assigning credit to each edge by the node value difference. Theoretically, the estimated step-level advantage faithfully adheres to the basic advantage definition in RL. To further ensure the reliability of step-level advantage estimation, we further propose Graph GAE, which extends GAE to the trajectory graph for reducing the impact of state-value estimation bias. Experiments across a range of multi-turn agentic benchmarks show consistent gains over GRPO and superior performance compared to recent agentic RL algorithms. Code will be available at https://github.com/xcyao00/GRAFT.
Chinese Translation
基于群体的强化学习（RL）方法，如GRPO及其变体，已成为训练推理和代理大型语言模型（LLMs）的主要范式。尽管它们在响应级别的群体归一化优势估计是可靠的，但在步骤级别上却存在系统性偏差，因为粗粒度的轨迹级优势难以准确反映单个步骤的贡献（即，失败的轨迹可能包含有价值的步骤）。重新审视基础的RL定义，我们注意到GRPO在单轮任务上的成功源于其优势估计策略，该策略遵循基本定义：从同一状态采样的多个动作的平均奖励构成了可信的状态价值估计。原则上，将这种忠实的估计扩展到步骤级别需要从每个中间状态采样多个动作，但这在每个状态的基础上成本过高。为了解决这个问题，我们提出了一种基于图的忠实步骤级信用分配框架（GRAFT），该框架将所有回滚轨迹 graft（嫁接）到一个轨迹图中，通过在图上进行贝尔曼迭代恢复节点状态值，并通过节点值差异为每条边分配信用。从理论上讲，估计的步骤级优势忠实地遵循RL中的基本优势定义。为了进一步确保步骤级优势估计的可靠性，我们进一步提出了图形GAE，将GAE扩展到轨迹图，以减少状态价值估计偏差的影响。在一系列多轮代理基准测试中的实验表明，相比GRPO，GRAFT在性能上持续获得提升，并且优于最近的代理RL算法。代码将发布在https://github.com/xcyao00/GRAFT。
cs.AI / 23 / 2609.29007
When Does Action Credit Need Updating?
何时需要更新行动信用？
Abstract
Tool-using agents are continually updated with new interaction data. After each policy update, however, previously estimated action credits may become stale. Recomputing them from scratch can require many additional tool calls and environment interactions, making repeated updates increasingly expensive. We ask a simple question: when does historical action credit actually need to be updated? Our key observation is that a change in action value does not necessarily imply a change in the decision. Historical credit can still be useful as long as policy-induced drift is too small to overturn the existing action ranking. Building on this idea, we introduce pairwise branch sensitivity to capture how strongly a policy update affects the downstream regions that distinguish two candidate actions. We then derive a first-order anchored credit-transport estimator that updates historical credit using old interventional trajectories, and propose a Decision-Sufficient Credit Gate (DSC-Gate) that chooses whether to reuse, transport, or resample credit. Experiments show that branch sensitivity explains credit drift substantially better than global policy distance. With sufficient historical data, credit transport reduces estimation error, while its benefit to decision making is concentrated on updates that affect action-distinguishing branches. On a fully independent test set, DSC-Gate changes mean regret by only +0.00004 relative to a gap-based gate while reducing mean new tool steps from 472 to 286, a 39.4% reduction. We observe the same pattern after a real tool-agent parameter update. Overall, our results show that agents do not need to recompute action credit after every policy update: much of the historical evidence can be reused or cheaply corrected, reducing the additional interaction required to keep action decisions up to date.
Chinese Translation
使用工具的智能体不断更新新的交互数据。然而，在每次策略更新后，之前估计的行动信用可能会变得过时。从头开始重新计算这些信用可能需要许多额外的工具调用和环境交互，使得重复更新变得越来越昂贵。我们提出一个简单的问题：历史行动信用何时确实需要更新？我们的关键观察是，行动价值的变化并不一定意味着决策的变化。只要策略引起的漂移不足以推翻现有的行动排名，历史信用仍然可以是有用的。在此基础上，我们引入了成对分支敏感性，以捕捉策略更新对区分两个候选行动的下游区域的影响程度。然后，我们推导出一种一阶锚定信用传输估计器，利用旧的干预轨迹更新历史信用，并提出了一种决策充分信用门（Decision-Sufficient Credit Gate, DSC-Gate），用于选择是重用、传输还是重新采样信用。实验表明，分支敏感性比全局策略距离更好地解释了信用漂移。在有足够历史数据的情况下，信用传输减少了估计误差，而其对决策的好处集中在影响行动区分分支的更新上。在一个完全独立的测试集上，DSC-Gate相对于基于差距的门仅将平均遗憾值改变了+0.00004，同时将平均新工具步骤从472减少到286，减少了39.4%。在一次真实的工具-智能体参数更新后，我们观察到了相同的模式。总体而言，我们的结果表明，智能体不需要在每次策略更新后重新计算行动信用：许多历史证据可以被重用或廉价修正，从而减少保持行动决策最新所需的额外交互。
cs.AI / 24 / 2609.29014
AlphaDiverse: Post-Training Local Quantitative Research Agents for Diverse Exploration in Alpha Factor Mining
AlphaDiverse：用于多样化探索的后训练本地定量研究代理的阿尔法因子挖掘
Abstract
Large language model (LLM)-based multi-agent systems can automate alpha factor mining, but their reliance on external APIs limits control over cost, availability, and confidentiality. Long research loops also tend to revisit a few successful economic mechanisms that lead to research path collapse. To address these limitations, we propose AlphaDiverse, a framework that integrates a multi-agent alpha research system, diverse research path collection, and post-training for local agents. We let the research system generate complementary plan portfolios and vary research environments across loops to collect diverse research paths. Using these diverse traces, we warm-start local Planner and Realizer agents with supervised fine-tuning. Then, we propose a joint GRPO method to optimize both of them using predictive quality and diversity of contributions. Research feedback is confined to inner period data, while a frozen final model is evaluated on a later outer period data, thereby avoiding test-set tuning. Experiments across four Chinese stock universes show that AlphaDiverse can combine competitive prediction with broader exploration.
Chinese Translation
基于大型语言模型（LLM）的多代理系统可以自动化阿尔法因子挖掘，但其对外部API的依赖限制了对成本、可用性和机密性的控制。长时间的研究循环往往会重新审视少数成功的经济机制，导致研究路径的崩溃。为了解决这些限制，我们提出了AlphaDiverse，一个集成了多代理阿尔法研究系统、多样化研究路径收集和本地代理后训练的框架。我们让研究系统生成互补的计划组合，并在循环中改变研究环境，以收集多样化的研究路径。利用这些多样化的轨迹，我们通过监督微调为本地规划者（Planner）和实现者（Realizer）代理提供热启动。然后，我们提出了一种联合GRPO方法，利用预测质量和贡献的多样性来优化这两个代理。研究反馈仅限于内部周期数据，而冻结的最终模型在后续外部周期数据上进行评估，从而避免了测试集调优。在四个中国股票市场的实验表明，AlphaDiverse能够将竞争性预测与更广泛的探索相结合。
cs.AI / 25 / 2609.29015
MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks
MeshHeal：去中心化 LLM 代理网络中灰色故障的双时间尺度自愈机制
Abstract
Decentralized LLM-based multi-agent systems coordinate through local interactions, but an agent can remain responsive while its task-solving quality persistently degrades. Such gray failures require protecting current tasks before sufficient evidence exists to alter future routing, while still allowing recovered agents to rejoin. We introduce MeshHeal, a fully decentralized self-healing framework that couples ability-matched peer review across two timescales. At the fast timescale, an adaptive hierarchy escalates uncertain or low-scoring outputs from repeated single-reviewer evaluation to committee deliberation and, when needed, correction before use. At the slow timescale, a task- and ability-conditioned peer-relative detector aggregates scores to distinguish persistent degradation from ordinary output variation, trigger mandatory committee review, and eventually exclude degraded agents from ordinary routing; recovery probes provide fresh evidence for reintegration. To faithfully evaluate routing, we introduce Model-Backed MAS Evaluation, which ties ability assignments to execution models, since prompt-based ability assignments alone can leave routing errors hidden. Across BBH, MATH, and MMLU-Pro, MeshHeal achieves 0.839 degraded-phase accuracy using 51k total model tokens per task, versus the strongest baseline Symphony's 0.807 accuracy using 115k per task. Under staggered degradation and recovery, MeshHeal isolates degraded agents, keeps them excluded from ordinary task execution until recovery, and returns them to normal routing.
Chinese Translation
基于去中心化 LLM 的多智能体系统通过局部交互进行协调，但代理在任务解决质量持续下降的情况下仍然可以保持响应。这种灰色故障需要在尚未有足够证据改变未来路由之前保护当前任务，同时仍允许恢复的代理重新加入。我们提出了 MeshHeal，这是一个完全去中心化的自愈框架，结合了跨两个时间尺度的能力匹配同行评审。在快速时间尺度上，自适应层级将来自重复单一评审者评估的不确定或低评分输出升级到委员会审议，并在必要时进行修正后使用。在慢速时间尺度上，基于任务和能力的相对检测器聚合分数，以区分持续退化与普通输出变异，触发强制委员会审查，并最终将退化的代理排除在普通路由之外；恢复探测提供新证据以便重新整合。为了真实评估路由，我们引入了基于模型的 MAS 评估，它将能力分配与执行模型关联，因为仅基于提示的能力分配可能会隐藏路由错误。在 BBH、MATH 和 MMLU-Pro 上，MeshHeal 在每个任务使用 51k 总模型令牌时达到了 0.839 的退化阶段准确率，而最强基线 Symphony 在每个任务使用 115k 时的准确率为 0.807。在交错的退化和恢复过程中，MeshHeal 隔离退化代理，保持其在恢复之前不参与普通任务执行，并将其恢复到正常路由中。
cs.AI / 26 / 2609.29050
SLCA-GRPO: Resolving Cross-Segment Credit Misattribution in Tool-Calling RL
SLCA-GRPO：解决工具调用强化学习中的跨段信用错误归属
Abstract
Tool-calling agents produce heterogeneous outputs, interleaving structured tool invocations with user-facing natural language summaries. This output heterogeneity presents a structural failure mode in standard on-policy Reinforcement Learning (RL): algorithms like GRPO indiscriminately broadcast a homogeneous trajectory-level scalar advantage to all tokens. Consequently, gradient noise from summary generation leaks into tool-decision tokens, causing cross-segment credit misattribution and brittle optimization. In this work, we propose SLCA-GRPO, a framework incorporating Segment-Locked Credit Assignment (SLCA). To enable scalable exploration without costly real APIs and stable training, we first construct the Schema-Guided LLM Simulator (SGLS) as foundational training infrastructure. Building on this, SLCA decouples advantage estimation at the structural segment level within a single group of rollouts, without requiring additional rollouts from intermediate states. Supported by Hierarchical Rewards (HierR), SLCA routes execution advantages to tool tokens and preference advantages to summary tokens, eliminating advantage contamination (the dominant cross-segment credit misattribution channel) within each policy update. On a 7B backbone, SLCA-GRPO accelerates convergence and outperforms standard GRPO, ToolPO, and RLTR by +2.53 pp on in-domain evaluation, +1.36 pp on the Berkeley Function-Calling Leaderboard (BFCL), and +9.15 pp on $\tau^2$-Bench under the same training budgets, achieving higher accuracy with reduced tool redundancy and costs.
Chinese Translation
工具调用代理生成异质输出，将结构化工具调用与面向用户的自然语言摘要交错在一起。这种输出异质性在标准的在线强化学习（RL）中呈现出一种结构性故障模式：像GRPO这样的算法无差别地将同质的轨迹级标量优势广播给所有标记。因此，摘要生成中的梯度噪声泄漏到工具决策标记中，导致跨段信用错误归属和脆弱的优化。在本研究中，我们提出了SLCA-GRPO，一个结合了段锁定信用分配（SLCA）的框架。为了在没有昂贵真实API的情况下实现可扩展探索和稳定训练，我们首先构建了模式引导的大型语言模型模拟器（SGLS）作为基础训练基础设施。在此基础上，SLCA在单组回滚中解耦了结构段级的优势估计，而无需从中间状态获取额外的回滚。通过分层奖励（HierR）的支持，SLCA将执行优势路由到工具标记，将偏好优势路由到摘要标记，从而消除了每次策略更新中的优势污染（主导的跨段信用错误归属通道）。在一个7B的基础模型上，SLCA-GRPO加速了收敛，并在同样的训练预算下，分别在领域内评估中超越了标准GRPO、ToolPO和RLTR，提升了+2.53个百分点，在伯克利函数调用排行榜（BFCL）上提升了+1.36个百分点，以及在$ au^2$-Bench上提升了+9.15个百分点，实现了更高的准确性，同时减少了工具冗余和成本。
cs.AI / 27 / 2609.29051
From Self-Distillation to Self-Practice: Privileged Information for Multi-Turn Agents
从自我蒸馏到自我实践：多轮代理的特权信息
Abstract
On-policy self-distillation (OPSD) has become a popular recipe for post-training LLM agents. It supervises the agent model at the token level with a stronger teacher view of the same model, obtained by conditioning on privileged information (PI). In this work, we show that in multi-turn agents, this paradigm teaches the student to act with confidence but without the information behind it. The trained agent behaves as if it had privileged information it never observed, and its performance falls well short of plain RL, in the worst case below the untrained base model. Therefore, we propose Privileged Self-Practice (PSP), which keeps the PI and moves it from the loss to the sampler. When the student's rollouts on a task mostly fail, we inject a short per-task instruction written by an analyzer model, sample the task again with the instruction in context, and train on the result with an unchanged GRPO objective. The privileged information stays in the prompt and never enters the loss. Across AppWorld and SWE-bench Verified, with three different student models, PSP obtains the best average score in every setting and is the only method that consistently outperforms plain GRPO, improving task-goal completion by up to 65% on AppWorld and the resolved rate by up to 61% on SWE-bench Verified.
Chinese Translation
在线自我蒸馏（On-policy self-distillation, OPSD）已成为后训练大型语言模型（LLM）代理的一种流行方法。它通过依赖特权信息（Privileged Information, PI）对同一模型进行条件化，从而在标记级别上监督代理模型，形成更强的教师视角。在本研究中，我们展示了在多轮代理中，这一范式使学生在缺乏背后信息的情况下自信地行动。经过训练的代理表现得好像它拥有从未观察到的特权信息，其性能在最坏情况下远低于普通强化学习（plain RL），甚至低于未训练的基础模型。因此，我们提出了特权自我实践（Privileged Self-Practice, PSP），该方法保留特权信息并将其从损失函数转移到采样器。当学生在某个任务上的回合大多失败时，我们注入由分析模型编写的简短每任务指令，重新在上下文中采样该任务，并在结果上训练，保持GRPO目标不变。特权信息保留在提示中，永远不会进入损失函数。在AppWorld和SWE-bench Verified上，使用三种不同的学生模型，PSP在每种设置中都获得了最佳平均分，并且是唯一一种在所有情况下都持续超越普通GRPO的方法，在AppWorld上任务目标完成率提高了多达65%，在SWE-bench Verified上的解决率提高了多达61%。
cs.AI / 28 / 2609.29075
CRISS: A Retrieval-Augmented AI Chatbot for Assisting Cancer Registrars
CRISS：一种增强检索的人工智能聊天机器人，用于协助癌症登记员
Abstract
Cancer registrars, including Oncology Data Specialists (ODSs), must interpret complex and frequently updated coding and staging standards. We developed CRISS (Cancer Registry Intelligent Support System), a retrieval-augmented generation (RAG) conversational assistant that provides rapid, citation-supported access to registry guidance. This study evaluated whether CRISS could (1) support accurate and citation-supported responses, (2) improve access to and interpretation of relevant guidance, and (3) support training/helpdesk use while preserving human oversight of final abstraction decisions. We built a domain-specific knowledge base from national cancer registry standards, segmented into metadata-tagged passages and indexed as dense embeddings. Retrieved passages were used to generate citation-grounded responses through a large language model (LLM). Open-weight, proprietary, and non-RAG baseline models across Gemini and GPT families were evaluated on easy, medium, and hard registry questions using an LLM-as-a-Judge protocols. RAG configurations consistently outperformed non-RAG approaches, especially as question difficulty increased. Mean grounding scores for RAG were 0.62/0.56/0.59 across easy/medium/hard tiers versus 0.29/0.26/0.29 for non-RAG. RAG models also achieved higher semantic-similarity scores overall. Proprietary RAG models performed strongest on easy and medium questions, while local RAG models ranked highest on hard questions and proprietary models were generally more cautious. Domain-specific RAG improved evidence grounding and response quality for cancer registry questions while enabling citation-supported assistance across complexity levels. CRISS demonstrates the potential of human-centered, citation-grounded AI to support cancer registrars while preserving human oversight for final coding decisions.
Chinese Translation
癌症登记员，包括肿瘤数据专家（ODS），必须解读复杂且经常更新的编码和分期标准。我们开发了CRISS（癌症登记智能支持系统），这是一种增强检索生成（RAG）对话助手，能够快速提供支持引用的登记指导。该研究评估了CRISS是否能够（1）支持准确且有引用的回答，（2）改善相关指导的获取和解读，以及（3）在保留人工监督最终抽象决策的情况下，支持培训/帮助台的使用。我们从国家癌症登记标准构建了一个特定领域的知识库，将其分段为带有元数据标签的段落，并作为密集嵌入进行索引。检索到的段落用于通过大型语言模型（LLM）生成基于引用的回答。我们使用LLM作为评判者的协议，评估了Gemini和GPT系列中的开放权重、专有和非RAG基线模型在简单、中等和困难登记问题上的表现。RAG配置在问题难度增加时始终优于非RAG方法。RAG的平均基础分数在简单/中等/困难层级分别为0.62/0.56/0.59，而非RAG的分数为0.29/0.26/0.29。RAG模型在整体上也达到了更高的语义相似度分数。专有RAG模型在简单和中等问题上的表现最强，而本地RAG模型在困难问题上排名最高，专有模型通常更为谨慎。特定领域的RAG提高了癌症登记问题的证据基础和回答质量，同时在不同复杂度水平上提供了支持引用的帮助。CRISS展示了以人为本、基于引用的人工智能支持癌症登记员的潜力，同时保留了对最终编码决策的人类监督。
cs.AI / 29 / 2609.29084
A Rapid Pipeline for Training and Deploying ML Models on WeBe Band
在 WeBe Band 上训练和部署机器学习模型的快速流程
Abstract
Developing optimized machine-learning algorithms for edge devices with limited computational and memory resources is challenging, time-consuming, and highly dependent on device-specific constraints. In this work, we streamline an edge ML workflow to enable rapid development, optimization, and deployment of machine-learning (ML) models directly on the WeBe Band, a wrist-worn wearable device designed for multimodal physiological data monitoring. The proposed system automatically generates hardware-efficient ML models that can be easily integrated into the WeBe core firmware, supporting AutoML, hardware-aware quantization, and performance profiling to build models that meet desired latency targets while remaining compatible with device memory and power limitations. The proposed framework tightly integrates the open-source Piccolo AI ecosystem with an automated pipeline that generates deployable firmware artifacts, performs hardware-aware model compilation, and supports over-the-air (OTA) deployment. The system supports multiple lightweight model classes, including classical machine-learning algorithms and neural networks, and provides built-in on-device profiling tools to evaluate inference latency and memory footprint under realistic execution conditions. Experimental results demonstrate clear trade-offs between model complexity and deployability on a microcontroller, showing that classical models offer strong real-time performance while lightweight neural networks require careful resource management. Rather than proposing new learning architectures, the current work mainly focuses on system-level automation, deployability, and enabling researchers and developers to rapidly iterate on models and evaluate them directly on target hardware. Although demonstrated on the WeBe Band platform, the workflow is designed to be extensible to other ML-powered edge devices.
Chinese Translation
为具有有限计算和内存资源的边缘设备开发优化的机器学习算法是一项具有挑战性、耗时且高度依赖于设备特定约束的任务。在本研究中，我们简化了边缘机器学习工作流程，以便在 WeBe Band 上快速开发、优化和部署机器学习（ML）模型，该设备是一种用于多模态生理数据监测的腕戴式可穿戴设备。所提出的系统能够自动生成硬件高效的机器学习模型，这些模型可以轻松集成到 WeBe 核心固件中，支持自动机器学习（AutoML）、硬件感知量化和性能分析，以构建满足所需延迟目标的模型，同时兼容设备的内存和功耗限制。该框架将开源的 Piccolo AI 生态系统与自动化流程紧密集成，生成可部署的固件工件，执行硬件感知模型编译，并支持空中下载（OTA）部署。该系统支持多种轻量级模型类别，包括经典机器学习算法和神经网络，并提供内置的设备端分析工具，以在真实执行条件下评估推理延迟和内存占用。实验结果清晰地展示了模型复杂性与在微控制器上可部署性之间的权衡，表明经典模型在实时性能上表现出色，而轻量级神经网络则需要仔细的资源管理。本研究并未提出新的学习架构，而是主要关注系统级自动化、可部署性，并使研究人员和开发人员能够快速迭代模型并直接在目标硬件上评估它们。尽管在 WeBe Band 平台上进行了演示，但该工作流程旨在可扩展到其他基于机器学习的边缘设备。
cs.AI / 30 / 2609.29108
Functional Architecture of European Electricity Trading Markets: Requirements for AI Supported Trading Systems under Regulatory Constraints
欧洲电力交易市场的功能架构：在监管约束下对人工智能支持交易系统的要求
Abstract
European electricity trading in the EU operates as a constrained multi-layer system in which legal design, exchange microstructure, and network physics are executed jointly across forward, day-ahead, intraday, and balancing horizons. This paper develops a functional architecture for AI-supported trading that is aligned with market-coupling mechanics, cross-zonal transfer constraints, and compliance obligations under REMIT, MiFID II, MiFIR, and EMIR. The contribution is a formal system specification composed of a decision-state vector, residual-exposure accounting, constrained optimization objective, executable-action permission gate, and fail-closed AI control logic with auditable records. The analysis maps major Nominated Electricity Market Operator (NEMO) venues and related exchange operators into an operational venue topology and identifies where cross-border coordination fails in practice: interface-level timing, permission heterogeneity, and balancing-layer coupling. The resulting framework proposes how AI can be deployed as a bounded decision component inside regulated market operation with explicit governance, rather than as an unconstrained prediction layer.
Chinese Translation
欧洲电力交易在欧盟内作为一个受限的多层系统运作，其中法律设计、交易所微观结构和网络物理学在远期、日内、实时和平衡时段中共同执行。本文提出了一种与市场耦合机制、跨区域转移约束以及根据REMIT、MiFID II、MiFIR和EMIR的合规义务相一致的人工智能支持交易的功能架构。贡献在于形成了一种正式的系统规范，包括决策状态向量、剩余风险核算、受限优化目标、可执行行动权限门以及具有可审计记录的闭环人工智能控制逻辑。分析将主要的指定电力市场运营商（NEMO）场所及相关交易所运营商映射到一个操作场所拓扑，并识别出跨境协调在实践中失败的地方：接口级时序、权限异质性和平衡层耦合。由此产生的框架提出了如何将人工智能作为一个受限的决策组件部署在受监管的市场运营中，并具有明确的治理，而不是作为一个不受限制的预测层。
cs.AI / 31 / 2609.29109
CounterRoute: Self-Routed Reasoning via Hierarchical Counterfactual Credit Assignment
CounterRoute：通过层次反事实信用分配实现自我路由推理
Abstract
Reasoning-capable language models often produce long chains of thought when direct answers suffice, wasting inference compute. Many dual-mode models leave this choice to users. Automating it is challenging because routing targets evolve with the policy, initial mode preferences destabilize exploration, and sequence-level objectives entangle routing with response learning. We introduce CounterRoute, an online reinforcement-learning framework that jointly learns routing and modeconditioned responses in one shared policy directly from a native dual-mode checkpoint, without method-specific SFT warm-up. Paired current-policy counterfactual rollouts assign cross-mode credit only to the routing token, while within-mode GRPO trains response tokens. A paired-to-self-routed curriculum stabilizes early training with forced rollouts from both modes, then increases self-routed updates to improve autonomous routing. Across nine benchmarks, CounterRoute better balances accuracy and efficiency than heuristic and learned adaptive-routing methods. Relative to always-thinking checkpoints, it improves macro-average accuracy while reducing mean generated tokens by 51% for Qwen3-8B and 41% for Qwen3-14B. On instruction-following and commonsense benchmarks where direct answering is strong, think rates fall as low as 1% while response quality improves. Despite training only on math and instruction following, its routing behavior and response quality generalize to held-out coding, science, knowledge, and commonsense benchmarks.
Chinese Translation
具备推理能力的语言模型在直接答案足够的情况下，往往会产生冗长的思维链，从而浪费推理计算资源。许多双模式模型将这一选择留给用户。自动化这一过程具有挑战性，因为路由目标随着策略的变化而演变，初始模式偏好会使探索不稳定，而序列级目标则将路由与响应学习纠缠在一起。我们提出了CounterRoute，这是一种在线强化学习框架，能够从原生双模式检查点直接共同学习路由和模式条件响应，而无需特定方法的SFT预热。配对当前策略的反事实回滚仅将跨模式信用分配给路由令牌，而模式内的GRPO则训练响应令牌。配对自我路由的课程通过强制从两种模式进行回滚来稳定早期训练，然后增加自我路由更新以改善自主路由。在九个基准测试中，CounterRoute在准确性和效率之间的平衡优于启发式和学习的自适应路由方法。与始终思考的检查点相比，它提高了宏平均准确性，同时减少了Qwen3-8B的平均生成令牌数51%和Qwen3-14B的41%。在直接回答效果显著的指令遵循和常识基准上，思考率降至低至1%，而响应质量却有所提升。尽管仅在数学和指令遵循上进行训练，其路由行为和响应质量在保留的编码、科学、知识和常识基准上也表现出良好的泛化能力。
cs.AI / 32 / 2609.29140
Sharp Limits for Honest Uncertainty in Hard-Budget Repeated Evaluation
硬预算重复评估中诚实不确定性的严格界限
Abstract
Repeated evaluation can estimate a benchmark score accurately while still requiring replication to certify narrow uncertainty. We characterize that requirement on a fixed grid of $M$ tasks with $L$ binary paths per task under the hard budget $(M+t)K$, where each path costs at most $K$ responses or episodes. For fixed $L \ge 3$ and $0 < \alpha \le 1/12$, the optimal expected width on the worst pure cohort is $\Theta_{\alpha,L}([M(t+1)]^{-1/2})$ when every task is observed and $\Theta_{\alpha,L}([M(t+\sqrt{M})]^{-1/2})$ when omission is allowed. The lower bounds cover adaptive hard-budget policies, and fixed random-subset designs attain both rates through disagreement certificates. A joint mean/disagreement interval turns the task-covering law into practical finite-budget inference. In an equal-budget LiveCodeBench replay with 16 models, 880 tasks, and five outputs per task, the task-covering design reduces median point-estimation MSE by 87.0\% relative to pooled uniform sampling, while the Joint certificate produces narrower confidence intervals in 15/16 panels and reduces median interval width by 30.6\%. Finite-regime analyses identify task coverage as the effective choice at the evaluated scale and characterize how cohort size and within-task agreement determine the useful operating region. Together, the sharp laws and fixed-budget evidence make replication and task coverage explicit design variables for information-efficient repeated evaluation.
Chinese Translation
重复评估可以准确估计基准分数，但仍然需要复制以证明狭窄的不确定性。我们在固定的 $M$ 任务网格上对这一要求进行了表征，每个任务有 $L$ 条二进制路径，硬预算为 $(M+t)K$，其中每条路径最多消耗 $K$ 次响应或回合。对于固定的 $L
geq 3$ 和 $0 < eta
leq 1/12$，当每个任务都被观察时，最优的期望宽度在最坏的纯队列上为 $ heta_{eta,L}([M(t+1)]^{-1/2})$，而当允许遗漏时为 $ heta_{eta,L}([M(t+ ext{sqrt}(M))]^{-1/2})$。下界涵盖了自适应硬预算策略，而固定随机子集设计通过不一致性证明达到了这两种速率。联合均值/不一致区间将任务覆盖法则转化为实用的有限预算推断。在一个平等预算的 LiveCodeBench 重播中，涉及 16 个模型、880 个任务和每个任务五个输出，任务覆盖设计使得中位点估计均方误差相较于汇总均匀抽样减少了 87.0\%，而联合证明在 15/16 个面板中产生了更窄的置信区间，并将中位区间宽度减少了 30.6 %。有限状态分析确定任务覆盖是评估规模下的有效选择，并表征了队列大小和任务内一致性如何决定有用的操作区域。总体而言，严格的法则和固定预算的证据使得复制和任务覆盖成为信息高效重复评估的明确设计变量。
cs.AI / 33 / 2609.29144
Scope Before You Persist: Preventing Cross-Family Interference in Agent Memory
在持续之前确定范围：防止代理记忆中的跨家族干扰
Abstract
Persistent memory lets language-model agents improve prompts and skills without updating model weights. We show that matching retrieval scope to certification scope enables these edits to support reliable repeated adaptation across recurring task families. We study frozen-model agents on ProcStream-RSI, a 12-round code-repair stream, using Orthogonal Regression Control (ORC), an execution-grounded gate for persistent skill edits. In an intervention that holds proposals and gate decisions fixed, retrieving each accepted skill only for its originating family raises mean hidden trajectory utility from 0.713 under global memory to 0.816 and changes harmful deployments from six of eight to none. In 27 paired randomized-order streams, Scoped-ORC improves mean trajectory utility by 0.063 [0.037, 0.094] over Global-ORC, accepts 63 rather than 12 updates, and produces multiple accepted updates in 19/27 streams, with 0/63 harmful acceptances. The global control reaches 0.713, below the static agent's 0.775, because locally valid edits can interfere with unrelated families. These results establish scope matching as a complementary control for persistent agent memory: certification determines whether an edit is supported, while retrieval scope determines where that evidence authorizes its use.
Chinese Translation
持久性记忆使语言模型代理能够在不更新模型权重的情况下改进提示和技能。我们展示了将检索范围与认证范围匹配，使这些编辑能够支持在重复任务家族中可靠的适应。我们在 ProcStream-RSI 上研究了冻结模型代理，这是一个包含12轮代码修复的流，使用正交回归控制（Orthogonal Regression Control, ORC），这是一个基于执行的门控机制，用于持久技能编辑。在一个保持提案和门控决策固定的干预中，仅为每个接受的技能检索其来源家族的技能，将全局记忆下的平均隐藏轨迹效用从0.713提高到0.816，并将有害部署从八个中的六个减少到零。在27个成对随机顺序流中，Scoped-ORC的平均轨迹效用比Global-ORC提高了0.063 [0.037, 0.094]，接受了63个更新而不是12个，并在27个流中产生了19个流的多个接受更新，且没有有害接受（0/63）。全局控制达到0.713，低于静态代理的0.775，因为局部有效的编辑可能会干扰不相关的家族。这些结果确立了范围匹配作为持久代理记忆的补充控制：认证决定了编辑是否被支持，而检索范围决定了该证据的使用授权位置。
cs.AI / 34 / 2609.29145
Claim-Gated Source-Risk Auditing for Generative Search
生成搜索的声明门控源风险审计
Abstract
A generative search answer can cite a supported passage yet omit a source relationship that changes its interpretation. We specify a claim-gated audit of the query-source-answer tuple. An omission is resolved only when relationship evidence, answer adoption, materiality, and disclosure are all observed; incomplete evidence remains unresolved rather than being treated as independence. The specification separates this endpoint from citation support and review priority, and binds decisions to versioned evidence spans. A reference checker makes the record contract executable. On an exhaustive synthetic suite, it reproduces all 81 three-state predicate combinations and rejects 192 deliberately malformed records. Common-guard baselines and predicate ablations isolate endpoint logic from missing-evidence handling, while controlled transitions check support separation and evidence removal. These are finite contract-conformance results, not detector accuracy or evidence of improved user outcomes. We define the independent annotation, held-out evaluation, and paired utility tests still required to establish semantic validity and deployment benefit.
Chinese Translation
生成搜索答案可以引用一个支持的段落，但可能省略一个改变其解释的源关系。我们指定了查询-源-答案元组的声明门控审计。只有当关系证据、答案采纳、重要性和披露都被观察到时，省略才会被解决；不完整的证据将保持未解决状态，而不是被视为独立性。该规范将此端点与引用支持和审查优先级分开，并将决策绑定到版本化的证据范围。引用检查器使记录合同可执行。在一个详尽的合成套件上，它再现了所有81种三状态谓词组合，并拒绝了192个故意畸形的记录。常见保护基线和谓词消融将端点逻辑与缺失证据处理隔离，而受控过渡则检查支持分离和证据移除。这些是有限的合同一致性结果，而不是检测器的准确性或改善用户结果的证据。我们定义了独立注释、保留评估和配对效用测试，这些仍然需要建立语义有效性和部署效益。
cs.AI / 35 / 2609.29154
A Wrong Turn Does Not Ruin the Journey: Deviation-Guided Skill Self-Evolution for LLM Agents
错误的转弯并不会毁掉旅程：基于偏差引导的技能自我演化用于大型语言模型代理
Abstract
Large language model agents increasingly rely on natural-language skills to solve complex tool-use tasks. However, such tasks often admit multiple valid solution paths, making it inappropriate to improve skills by forcing failed trajectories to match a fixed successful trajectory. Moreover, failed trajectories are rarely entirely wrong: an agent may first collect useful evidence and make meaningful progress, but later deviate into an erroneous suffix. We therefore argue that skill self-evolution should identify where productive problem solving begins to break down, rather than reflect coarsely over the entire failure. Based on this insight, we propose SkillPivot, a deviation-point-guided framework for skill self-evolution. SkillPivot detects the transition from a useful prefix to an erroneous suffix using execution validity, goal progress, and action diversity. A stronger teacher then continues from the same prefix and produces a successful alternative under the same interaction history. By contrasting the student's failed suffix with the teacher's successful suffix, SkillPivot generates localized skill updates while preserving already effective guidance. Experiments on ToolQA, LogicBench, and WildClawBench show that SkillPivot consistently outperforms competing skill-evolution methods, improves multiple agent models, and produces compact, transferable skill updates.
Chinese Translation
大型语言模型代理越来越依赖自然语言技能来解决复杂的工具使用任务。然而，这类任务通常存在多条有效的解决路径，因此通过强制失败轨迹与固定成功轨迹匹配来提升技能是不恰当的。此外，失败的轨迹往往并非完全错误：代理可能首先收集有用的证据并取得有意义的进展，但随后偏离进入错误的后缀。因此，我们认为技能自我演化应识别出有效问题解决开始崩溃的地方，而不是粗略地反映整个失败。基于这一见解，我们提出了SkillPivot，一个基于偏差点引导的技能自我演化框架。SkillPivot通过执行有效性、目标进展和行动多样性来检测从有用前缀到错误后缀的过渡。然后，一个更强的教师从相同的前缀继续，并在相同的交互历史下产生成功的替代方案。通过将学生的失败后缀与教师的成功后缀进行对比，SkillPivot生成局部的技能更新，同时保留已经有效的指导。在ToolQA、LogicBench和WildClawBench上的实验表明，SkillPivot在技能演化方法中始终表现优于竞争对手，改善了多种代理模型，并生成紧凑、可转移的技能更新。
cs.AI / 36 / 2609.29167
IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking
IndicBankBench：评估印度零售银行语言模型助手的安全性和可靠性
Abstract
Banking assistants must use account-specific information to answer requests and, in many cases, take actions through tools. Evaluating only the final response misses important errors. An assistant may ask for information it already has, rely on stale context, select the wrong account, or write an invalid value after stating the correct one. We introduce IndicBankBench, a 799-case benchmark for Indian retail banking spanning five operational domains, a capability/refusal domain, and twenty primary axes. Cases are evaluated at four stages: safety, action and tool use, response adequacy, and advisory quality. Tool use and most safety checks are deterministic. A narrow resolver handles only ambiguous confirmation-before-write cases, while a separate LLM judge evaluates semantic response adequacy. We run every case three times and report strict pass^3, which requires success on all trials. Across the eleven evaluated models, strict reliability ranges from 43.7% to 58.2%, whereas at-least-once success ranges from 60% to 74%. This gap shows that at-least-once success can overstate dependable banking behavior. The case-level diagnostics also distinguish systems that ask unnecessary questions from those that act but fail to reconcile customer context or fully resolve the request. We release the cases, mock environment, and evaluation harness.
Chinese Translation
银行助手必须使用特定账户的信息来回答请求，并且在许多情况下，通过工具采取行动。仅评估最终响应会遗漏重要错误。助手可能会询问其已经拥有的信息，依赖过时的上下文，选择错误的账户，或者在陈述正确值后写出无效值。我们引入了IndicBankBench，这是一个涵盖五个操作领域、一个能力/拒绝领域和二十个主要维度的799案例基准，专注于印度零售银行。案例在四个阶段进行评估：安全性、行动和工具使用、响应充分性和建议质量。工具使用和大多数安全检查是确定性的。一个狭窄的解析器仅处理模糊的确认-写入案例，而一个独立的LLM评估者则评估语义响应的充分性。我们对每个案例进行了三次测试，并报告严格通过率^3，这要求所有试验均成功。在评估的十一种模型中，严格可靠性范围从43.7%到58.2%，而至少一次成功率范围从60%到74%。这一差距表明，至少一次成功可能会夸大可靠的银行行为。案例级诊断还区分了那些询问不必要问题的系统与那些采取行动但未能调和客户上下文或完全解决请求的系统。我们发布了案例、模拟环境和评估工具。
cs.AI / 37 / 2609.29181
Right Choice of Classification Algorithms Based on Reinforcement Learning for Prediction of Non-Alcoholic Fatty Liver
基于强化学习的分类算法选择在非酒精性脂肪肝预测中的应用
Abstract
There are many complex issues in the world of artificial intelligence. Some of these problems are solved using other artificial intelligence methods, which are called artificial intelligence for artificial intelligence. Finding an appropriate classifier algorithm is a time-consuming task. For this reason, an algorithm that can automatically learn the choice of classification algorithms is very important. Classification algorithms are useful in predicting various diseases. Also, Primary Biliary Cirrhosis is one of the most well-known diseases that have been predicted by classification algorithms. This research's most significant achievement and novelty is the automatic increase in learning through a scoring method of reinforcement learning is called square learning (SL). In this research, an algorithm is presented that learns to automatically select the appropriate classification algorithm to predict Primary Biliary Cirrhosis. In this article, with inspiration from four evaluation metrics in classification algorithms, a new reinforcement learning method by the name of Fourth Degree Learning has been presented. In this research, we increased the performance of the classification algorithms used in this method from 63% of accuracy and achieved 98% accuracy.
Chinese Translation
人工智能领域存在许多复杂问题。其中一些问题通过其他人工智能方法解决，这被称为人工智能为人工智能。寻找合适的分类算法是一项耗时的任务。因此，能够自动学习分类算法选择的算法显得尤为重要。分类算法在预测各种疾病方面具有重要作用。此外，原发性胆汁性肝硬化是通过分类算法预测的最著名疾病之一。本研究最重要的成就和创新在于通过一种称为平方学习（Square Learning, SL）的强化学习评分方法实现自动学习的提升。在本研究中，提出了一种算法，能够自动选择合适的分类算法来预测原发性胆汁性肝硬化。本文从分类算法中的四个评估指标获得灵感，提出了一种新的强化学习方法，称为四阶学习（Fourth Degree Learning）。通过这种方法，我们将所用分类算法的性能从63%的准确率提升至98%的准确率。
cs.AI / 38 / 2609.29187
The Entropy Triangle Method (ETM): A novel framework for the prevention of cardiac arrhythmia with a review of more than 10,000 patients
熵三角方法（ETM）：一种用于预防心脏心律失常的新框架，基于超过10,000名患者的回顾研究
Abstract
One of the most important problems in medicine is to facilitate prediction. In this study, we propose entropy triangle method, a novel framework for predicting heart rhythms using a novel machine learning technique. This framework includes three steps: feature engineering, entropy triangle oversampling, and disease prediction. The dataset used in this study is a 12-lead electrocardiogram (ECG) arrhythmia research database with 10,646 patients. This dataset contains 11 different heart rhythms (5 sinus rhythms and 6 non-sinus rhythms). In this article, we introduce two firsts in machine learning and medicine that can predict non-sinus rhythm with over 85% accuracy. Our experimental results show, among others, that the most accurate classifier based on entropy triangles and the most useful oversampling are the supported vector classifiers and oversampling techniques for shark scent.
Chinese Translation
医学中最重要的问题之一是促进预测。在本研究中，我们提出了熵三角方法，这是一种利用新型机器学习技术预测心律的新框架。该框架包括三个步骤：特征工程、熵三角过采样和疾病预测。本研究使用的数据集是一个包含10,646名患者的12导联心电图（ECG）心律失常研究数据库。该数据集包含11种不同的心律（5种窦性心律和6种非窦性心律）。在本文中，我们介绍了机器学习和医学中的两个首次突破，能够以超过85%的准确率预测非窦性心律。我们的实验结果显示，基于熵三角的最准确分类器和最有效的过采样方法是支持向量分类器和鲨鱼气味的过采样技术。
cs.AI / 39 / 2609.29189
When Honesty is Not Enough in AI Debate
当诚实不足以应对人工智能辩论时
Abstract
Scalable oversight aims to verify the behaviour of agents whose capabilities exceed those of their overseers. AI debate has been proposed as an oversight solution in which competing agents help a resource-limited verifier assess claims that it cannot reliably evaluate unaided. Much of its promise rests on incentivizing honest arguments that lead to correct verdicts. Yet a correct verdict need not uniquely determine the arguments used to support it. Agents may retain discretion over which correct claims to present, how to frame them, and in what order to disclose them. This residual freedom can allow agents to shape what the verifier learns beyond the task-relevant conclusion, pursuing latent objectives without compromising verdict correctness. To study this phenomenon, we introduce the framework strategic interactive oversight (SIO), which treats oversight jointly as a verification mechanism and a strategic communication channel. Within this framework, we formalise the notion of task-admissible latent optimisation, which entails the pursuit of latent objectives while maintaining a prescribed task performance. As proof-of-concept, we instantiate SIO in the establish protocol debate with cross-examination and quantify a tradeoff between task success and information disclosure about a hidden variable. The trade-off identifies a strategic window in which substantial disclosure remains compatible with task admissibility. Towards mitigation, we reduce admissible bias by expanding the cross-examiner's role to mitigate persistent disclosure over finite interaction horizons. Our results highlight the need to evaluate oversight not only by the correctness of its verdicts, but also by the information conveyed through its transcripts.
Chinese Translation
可扩展的监督旨在验证能力超出监督者的代理行为。人工智能辩论被提议作为一种监督解决方案，其中竞争代理帮助资源有限的验证者评估其无法可靠独立评估的主张。其许多前景依赖于激励诚实的论证，以导致正确的裁决。然而，正确的裁决并不一定唯一决定支持它的论证。代理可能保留对呈现哪些正确主张、如何构建这些主张以及以何种顺序披露它们的自由。这种剩余自由可以使代理塑造验证者所学到的内容，超出任务相关的结论，追求潜在目标而不影响裁决的正确性。为了研究这一现象，我们引入了战略互动监督（Strategic Interactive Oversight, SIO）框架，该框架将监督视为一种验证机制和战略沟通渠道。在此框架内，我们形式化了任务可接受潜在优化的概念，即在保持规定的任务表现的同时追求潜在目标。作为概念验证，我们在建立协议辩论中实例化SIO，并量化任务成功与关于隐藏变量的信息披露之间的权衡。该权衡识别出一个战略窗口，在该窗口内， substantial disclosure 与任务可接受性兼容。为了缓解这一问题，我们通过扩展交叉审查者的角色来减少可接受的偏见，以减轻在有限交互视野内的持续披露。我们的结果强调了评估监督的必要性，不仅要看其裁决的正确性，还要看其记录中传达的信息。
cs.AI / 40 / 2609.29191
ASIRF: An Agentic Framework for Context-Dependent Sensitive Information Redaction
ASIRF：一种面向上下文依赖的敏感信息编辑的代理框架
Abstract
Sensitive information is defined by domain and intent, not a universal category, yet redaction systems such as privacy filters and named-entity recognizers fix a taxonomy at training time, requiring retraining for each new domain. We introduce ASIRF (Agentic Sensitive Information Redaction Framework), which retrieves domain-specific definitions based on the input's domain from a flexible knowledge base at inference time, needing no retraining to adapt. Two architectures, a three-call multi-agent pipeline and a single-agent variant, are evaluated across ten small open-weight models and eight datasets, including out-of-distribution fictional domains, against the OpenAI Privacy Filter (OPF) as a trained-classifier baseline. With only a few dozen expert-authored definitions per domain and no training data, ASIRF's recall exceeds OPF's in 68 of 80 model-domain combinations (85 percent), by at least one of the two architectures, with shortfalls confined mostly to OPF's training-distribution domains.
Chinese Translation
敏感信息由领域和意图定义，而不是一个普遍的类别。然而，诸如隐私过滤器和命名实体识别器等编辑系统在训练时固定了分类法，因此每个新领域都需要重新训练。我们提出了ASIRF（代理敏感信息编辑框架），该框架在推理时从灵活的知识库中检索基于输入领域的特定定义，无需重新训练即可适应。我们评估了两种架构：一个是三次调用的多代理管道，另一个是单代理变体，针对十个小型开放权重模型和八个数据集（包括分布外的虚构领域），与作为训练分类器基准的OpenAI隐私过滤器（OPF）进行了对比。在每个领域仅需几十个专家撰写的定义且没有训练数据的情况下，ASIRF在80个模型-领域组合中有68个（85%）的召回率超过了OPF，且至少有一种架构表现优异，短板主要集中在OPF的训练分布领域。
cs.AI / 41 / 2609.29228
Towards An LLM-Driven Unified Conversion Framework for BT and FSM in Autonomous Intelligent Systems
面向自主智能系统的LLM驱动统一转换框架：行为树与有限状态机的转换
Abstract
Finite state machine (FSM) and behavior trees (BT) are widely adopted behavioral modeling paradigms for autonomous intelligent systems. While functionally equivalent and inter-convertible in principle, existing transformation methods between FSM and BT face major challenges in preserving behavioral completeness and avoiding model complexity explosion. To overcome these issues, we propose an LLM-driven unified conversion framework that enables automatic, efficient, and semantically consistent transformation between FSM and BT. Specifically, a novel loop execution BT structure is designed for LLM to accurately capture the loop structure in FSM, thereby preserving behavioral completeness. To mitigate the state explosion problem in BT-to-FSM conversion, a depth compression strategy is introduced with LLM prompt to eliminate redundant control nodes, complemented by differentiated hierarchical conversion rules that collectively reduce the number of required sub-FSM. Simulation experiments in multiple autonomous decision-making scenarios demonstrate that the proposed framework enables an accurate and automated bidirectional conversion between FSM and BT. Furthermore, it significantly enhances the scalability and maintainability of generated models compared to traditional approaches, providing a practical solution for behavior model conversion in consumer-grade autonomous intelligent systems such as service robots, game agents, and smart home devices
Chinese Translation
有限状态机（FSM）和行为树（BT）是自主智能系统中广泛采用的行为建模范式。尽管在功能上是等价的并且原则上可以互相转换，但现有的FSM与BT之间的转换方法在保持行为完整性和避免模型复杂性爆炸方面面临重大挑战。为了解决这些问题，我们提出了一种LLM驱动的统一转换框架，能够实现FSM与BT之间的自动、高效和语义一致的转换。具体而言，我们为LLM设计了一种新颖的循环执行BT结构，以准确捕捉FSM中的循环结构，从而保持行为的完整性。为了缓解BT到FSM转换中的状态爆炸问题，我们引入了一种深度压缩策略，结合LLM提示来消除冗余控制节点，并辅以差异化的分层转换规则，集体减少所需的子FSM数量。在多个自主决策场景中的仿真实验表明，所提出的框架能够实现FSM与BT之间的准确和自动的双向转换。此外，与传统方法相比，它显著增强了生成模型的可扩展性和可维护性，为消费级自主智能系统（如服务机器人、游戏代理和智能家居设备）中的行为模型转换提供了实用解决方案。
cs.AI / 42 / 2609.29251
Policy as Code: A Coroutine-Bridge Harness for Fast-Reasoning Reliability on CAR-bench
政策即代码：用于CAR-bench快速推理可靠性的协程桥接工具
Abstract
CAR-bench evaluates whether tool-using agents stay reliable under real-world uncertainty, executing every tool inside the evaluator so that each tool-result exchange is a separate agent round-trip. A conventional next-action agent can batch parallel tool calls, but a chain of dependent calls costs it one model call per round of results. We present a coroutine-bridge harness in which the model's only action is to emit a Python program that blocks and resumes in place across evaluator tool exchanges. This decouples model invocation from tool round-trips: on the public test split the agent uses a median of two model calls against seven agent turns per task, resolving a full multi-turn task in a median of 1.8 s of model latency on Cerebras gpt-oss-120b. Because the action surface is executable code, deterministic CAR-bench policies are encoded directly as logic in the tool layer rather than as prompt rules, enforcing compliance at zero reasoning cost. On the official hidden evaluation the harness won Track 2 with 60.0% Pass^3, 4.5x the organizer baseline, at the lowest estimated cost and the fastest median task latency (3.14 s) of any entry scoring above that baseline; the same unchanged harness reproduced an identical 60.0% Pass^3 on GPT-5.5 in the Open track, matching frontier-model agents. A single static prompt, appended with per-task state at the tail, stays byte-identical across calls and across tasks: the frozen submission prompt served 78% of input tokens from cache (86.6% across its warm tail), against 73% over a three-week development corpus in which prompt edits repeatedly reset the cache. This compounds the few-call design into a small fraction of nominal input compute.
Chinese Translation
CAR-bench评估工具使用代理在现实世界不确定性下的可靠性，通过在评估器内部执行每个工具，使得每个工具结果交换都是一个独立的代理往返。传统的下一步行动代理可以批量并行工具调用，但一系列依赖调用使其在每轮结果中消耗一个模型调用。我们提出了一种协程桥接工具，其中模型的唯一操作是发出一个Python程序，该程序在评估器工具交换中阻塞并恢复。这将模型调用与工具往返解耦：在公共测试分割中，代理在每个任务中使用两个模型调用的中位数，对应七个代理回合，在Cerebras gpt-oss-120b上以1.8秒的模型延迟解决完整的多回合任务。由于操作表面是可执行代码，确定性的CAR-bench策略直接作为工具层中的逻辑编码，而不是作为提示规则，从而在零推理成本下强制合规。在官方隐藏评估中，该工具以60.0%的Pass^3赢得了第二轨道，达到了组织者基线的4.5倍，且以最低估算成本和最快的中位任务延迟（3.14秒）超越了任何得分高于该基线的参赛作品；同样不变的工具在开放轨道上在GPT-5.5上再现了相同的60.0% Pass^3，匹配前沿模型代理。一个静态提示，附加每个任务状态在尾部，在调用和任务之间保持字节相同：冻结的提交提示从缓存中提供了78%的输入标记（在其温暖尾部为86.6%），而在三周的开发语料库中，提示编辑反复重置缓存时，仅为73%。这将少量调用设计复合为名义输入计算的一小部分。
cs.AI / 43 / 2609.29266
Baszta: Data-Centric Fine-Tuning of a Polish Multi-Label Safety Classifier
Baszta：波兰多标签安全分类器的数据中心微调
Abstract
We develop a multi-label Polish content-safety classifier by fine-tuning allegro/herbert-base-cased (124M) across five categories (hate, vulgarity, sexual content, crime, self-harm) using a Focal + R-Drop objective, and evaluate the resulting model against Bielik Guard (S\'ojka) on the shared out-of-distribution Gadzi J\k{e}zyk benchmark. Both systems are given per-category threshold tuning on the same calibration split. Under that matched protocol our model holds a small but statistically significant lead in micro F1, while an apparent macro-F1 lead does not survive: it was an artifact of comparing a tuned model against an untuned one. We also report what that micro figure is worth. Because Gadzi J\k{e}zyk is 97% crime-positive, a classifier that flags crime on every input and nothing else already scores 0.910 micro F1 on the same test split, so micro separates neither system from a degenerate strategy and macro is the column that does. Per-category and per-protocol figures are reported in Section 4. The residual out-of-distribution gap is one of calibration rather than discrimination. Ranking quality stays high while positive probabilities collapse, and per-category temperature scaling recovers the loss where Platt scaling and isotonic regression do not. That recovery turns out to be conditional on the calibration set containing safe text. Gadzi J\k{e}zyk contains almost none, so thresholds fitted on it flag crime on every safe input, and a balanced refit buys a deployable operating point at the cost of adversarial recall. We report both operating points rather than only the flattering one. Two changes that are standard practice, per-class cost-sensitive weighting and mean pooling, each raise in-distribution macro F1 while lowering the out-of-distribution figure, which indicates that robustness has to be selected for directly rather than inherited from in-distribution accuracy.
Chinese Translation
我们通过在五个类别（仇恨、粗俗、色情内容、犯罪、自残）上使用 Focal + R-Drop 目标微调 allegro/herbert-base-cased (124M)，开发了一个多标签波兰内容安全分类器，并在共享的分布外 Gadzi Język 基准上与 Bielik Guard (Sójka) 评估了所得到的模型。两个系统在相同的校准拆分上进行了每类别阈值调优。在该匹配协议下，我们的模型在微 F1 上保持了小幅但统计显著的领先，而明显的宏 F1 领先并未持续：这只是将调优模型与未调优模型进行比较的伪影。我们还报告了该微观指标的价值。由于 Gadzi Język 中 97% 的样本为犯罪正例，因此一个在每个输入上标记犯罪而不标记其他内容的分类器在相同的测试拆分上已经得分 0.910 微 F1，因此微观指标并未将两个系统与退化策略区分开，而宏观指标则是能够做到这一点的列。每类别和每协议的指标在第 4 节中报告。残余的分布外差距是校准而非区分的问题。排名质量保持高水平，而正概率却崩溃，每类别的温度缩放在 Platt 缩放和等距回归无法恢复损失的情况下恢复了损失。该恢复结果依赖于校准集包含安全文本。Gadzi Język 几乎不包含安全文本，因此在其上拟合的阈值会在每个安全输入上标记犯罪，而平衡的重新拟合则以对抗性召回为代价获得可部署的操作点。我们报告这两个操作点，而不仅仅是有利的一个。两项标准实践的变化，即每类成本敏感加权和均值池化，均提高了分布内的宏 F1，同时降低了分布外的指标，这表明鲁棒性必须直接选择，而不是从分布内的准确性中继承。
cs.AI / 44 / 2609.29269
ALOE: Semantically Addressed Low-Rank Operators for Knowledge Editing
ALOE：用于知识编辑的语义地址低秩算子
Abstract
Knowledge editing changes what a model knows by modifying parameters so that a requested fact updates while unrelated behavior is preserved. This is usually treated as a write problem, but editing also involves an address problem: deciding which hidden states should receive the new residual. An update that activates too narrowly memorizes one prompt, while one that activates too broadly disrupts neighboring knowledge. Parametric editors encode this scope implicitly, whereas memory-based editors make the selection explicit but keep it outside the edited model. We propose ALOE (Addressed Low-rank Operator for Editing), which learns semantic addresses from paraphrases and hard same-subject negatives, aligns them with autoregressive hidden states through rollout refinement and gate calibration, and embeds the resulting gated low-rank operator within one MLP layer, so that the deployed model runs in a single forward pass with no external retriever or auxiliary router. Evaluated on CounterFact, ZSRE, and KnowEdit across three 7--8B model families, ALOE achieves efficacy between 0.955 and 0.999 and locality between 0.981 and 1.000; mechanistic analyses confirm that the learned geometry separates competing edits and that calibration suppresses out-of-scope activation. The remaining errors concentrate in paraphrase coverage and write fitting.
Chinese Translation
知识编辑通过修改参数改变模型所知内容，使得请求的事实得以更新，同时保留无关的行为。这通常被视为一个写入问题，但编辑也涉及一个地址问题：决定哪些隐藏状态应接收新的残差。一个激活过于狭窄的更新会记忆一个提示，而一个激活过于广泛的更新则会干扰相邻知识。参数化编辑器隐式编码了这一范围，而基于记忆的编辑器则明确选择，但将其保持在编辑模型之外。我们提出了ALOE（Addressed Low-rank Operator for Editing），它从同义句和难以处理的同主题负样本中学习语义地址，通过回滚细化和门控校准将其与自回归隐藏状态对齐，并将得到的门控低秩算子嵌入到一个多层感知机（MLP）层中，从而使得部署的模型在单次前向传播中运行，无需外部检索器或辅助路由器。通过在CounterFact、ZSRE和KnowEdit上对三个7-8B模型系列进行评估，ALOE的有效性在0.955到0.999之间，局部性在0.981到1.000之间；机械分析确认学习到的几何结构能够分离竞争编辑，并且校准抑制了超出范围的激活。剩余的错误集中在同义句覆盖和写入拟合上。
cs.AI / 45 / 2609.29283
From Text Decisions to Pixels: An Study of Jev-Style Visual Choice Model
从文本决策到像素：Jev风格视觉选择模型研究
Abstract
Visual software often needs a decision over supplied alternatives rather than a generated explanation. We present PixelJev, a native-image decision interface that maps an image, a task instruction, and a runtime candidate set to a structured choice and candidate-conditioned probabilities using small open multimodal models. Its initial realization unifies recognition and multiplechoice visual question answering through an existing language-model readout, with separately evaluated options for frozen inference, language-side adaptation, and held-out calibration. Across seven benchmark evaluations, 64-shot source adaptation raises Pets accuracy from 60.13% to 92.40% across optimization seeds and transfers to natural resampling, new texture labels, and A-OKVQA without target fitting, while frozen inference already supports both VQA tasks. A matched prompt-only follow-up on Pets and ScienceQA attributes the large Pets gain to adaptation and identifies a narrower output validity benefit of candidate readout in adapted VQA. Specialist DINOv2 probes remain stronger on source recognition, frozen 4B is stronger than adapted 2B on DTD and ScienceQA, and accuracy gains do not ensure calibrated target probabilities. These findings establish a working starting point for general-purpose visual decision models and identify the remaining requirements: schema robustness, cross-family transfer, and reliable use of visual evidence.
Chinese Translation
视觉软件通常需要对提供的替代方案进行决策，而不是生成解释。我们提出了PixelJev，这是一种原生图像决策界面，它将图像、任务指令和运行时候的候选集映射到结构化选择和候选条件概率，使用小型开放多模态模型。其初步实现通过现有语言模型的读取，统一了识别和多项选择视觉问答，分别评估了冻结推理、语言侧适应和保留校准的选项。在七个基准评估中，64-shot源适应将Pets的准确率从60.13%提高到92.40%，跨优化种子并转移到自然重采样、新纹理标签和A-OKVQA，而无需目标拟合，同时冻结推理已支持两种VQA任务。针对Pets和ScienceQA的匹配提示后续研究将Pets的显著提升归因于适应，并识别出在适应的VQA中候选读取的输出有效性收益较窄。专业的DINOv2探测器在源识别上仍然更强，冻结的4B在DTD和ScienceQA上优于适应的2B，而准确率的提升并不确保校准的目标概率。这些发现为通用视觉决策模型建立了一个有效的起点，并确定了剩余的需求：模式鲁棒性、跨家族迁移和视觉证据的可靠使用。
cs.AI / 46 / 2609.29312
When No One Owns the Judgment: Accountability Under Contribution Dissolution in Human-AI Collaboration
当无人拥有判断时：人机协作中的贡献解散责任问题
Abstract
Communities often respond to potentially AI-assisted work by asking three questions: Was AI used? Was that use disclosed? Can hidden use be detected? These questions place AI use itself at the center of accountability while overlooking a deeper problem: unowned judgment. Evaluations, claims, decisions, and creative directions can be shaped by AI with no accountable human or institution prepared to stand behind them. We develop this argument through two illustrative cases: AI-assisted peer review and concealed AI use in creative work. The first shows how contribution dissolution can weaken responsibility while the second shows how the fear of losing credit can discourage honest disclosure. The cases expose the limits of disclosure rules and provenance records as responses to AI-mediated collaboration. We offer three directions for discussion: distinguishing the roles AI plays, identifying judgments that require clear human ownership, and creating conditions in which AI involvement can be disclosed without default penalty. The broader aim is to make AI-shaped contributions discussable, creditable, contestable, and repairable.
Chinese Translation
社区通常通过提出三个问题来回应潜在的AI辅助工作：是否使用了AI？这种使用是否被披露？隐藏的使用是否可以被检测？这些问题将AI的使用置于责任的中心，同时忽视了一个更深层次的问题：无人拥有的判断。评估、主张、决策和创意方向可能会受到AI的影响，但没有任何负责任的人或机构准备为其背书。我们通过两个典型案例来发展这一论点：AI辅助的同行评审和创意工作中的隐蔽AI使用。第一个案例展示了贡献解散如何削弱责任，而第二个案例则表明，失去信用的恐惧可能会抑制诚实的披露。这些案例揭示了披露规则和来源记录作为应对AI介导协作的局限性。我们提出了三个讨论方向：区分AI所扮演的角色，识别需要明确人类所有权的判断，以及创造可以在没有默认惩罚的情况下披露AI参与的条件。更广泛的目标是使AI塑造的贡献能够被讨论、获得信用、受到质疑和修复。
cs.AI / 47 / 2609.29341
SkinAgent AI: A Safety-Grounded Multimodal Agentic Framework for Non-Diagnostic Skincare Support
SkinAgent AI：一个以安全为基础的非诊断护肤支持多模态智能框架
Abstract
Consumer-facing skincare AI must coordinate visual evidence, product information, tool use, and user-facing actions within explicit evidence and safety boundaries. This study evaluates SkinAgent AI, a non-diagnostic multimodal framework that combines visual concern routing with grounded and auditable LLM-based orchestration. The architecture includes routing for Acne, Pores, and Wrinkles; photograph-based skin-type estimation; count-informed ordinal acne-severity support; typed tools; database-grounded recommendation and action functions; deterministic safety, privacy, and evidence checks; approval before state-changing actions; and structured trace and replay mechanisms. Visual-model performance and system-level agent behavior were evaluated separately. Across three seeds, the skin-condition routing model achieved 99.84% +/- 0.07% accuracy. Skin-type estimation achieved 88.85% accuracy, while count-informed acne-severity support achieved 84.59% accuracy with a quadratic weighted kappa of 0.9076. On a locked but non-independent 240-case system benchmark, intent accuracy was 80.00%, exact tool-set match was 62.92%, and strict task completion was 47.08%. No violations or successful cross-user leakage events were observed in the finite safety and privacy test suites. Tool-selection errors, incomplete grounding of product attributes, and unreliable failure fallback nevertheless remained. These findings support the feasibility of bounded, database-grounded, and traceable agent orchestration for non-diagnostic skincare assistance. They do not establish clinical readiness, external generalization, formal privacy guarantees, or universal safety. Independent validation, expert assessment, robustness and fairness testing, and prospective evaluation in real-world settings remain necessary.
Chinese Translation
面向消费者的护肤人工智能必须在明确的证据和安全边界内协调视觉证据、产品信息、工具使用和用户行为。本研究评估了SkinAgent AI，这是一种非诊断的多模态框架，结合了视觉关注路由与基于可追溯的LLM（大语言模型）编排。该架构包括针对痤疮、毛孔和皱纹的路由；基于照片的肤质估计；基于计数的有序痤疮严重程度支持；输入工具；基于数据库的推荐和行动功能；确定性的安全、隐私和证据检查；在状态变化行动之前的批准；以及结构化的追踪和重放机制。视觉模型性能和系统级代理行为分别进行了评估。在三个种子下，皮肤状况路由模型达到了99.84% +/- 0.07%的准确率。肤质估计的准确率为88.85%，而基于计数的痤疮严重程度支持的准确率为84.59%，二次加权kappa值为0.9076。在一个锁定但非独立的240案例系统基准测试中，意图准确率为80.00%，精确工具集匹配为62.92%，严格任务完成率为47.08%。在有限的安全和隐私测试套件中未观察到违规或成功的跨用户泄露事件。然而，工具选择错误、产品属性的基础不完整以及不可靠的失败回退仍然存在。这些发现支持了在非诊断护肤辅助中进行有限、基于数据库和可追溯的代理编排的可行性，但并未确立临床准备、外部推广、正式隐私保证或普遍安全性。独立验证、专家评估、稳健性和公平性测试以及在现实环境中的前瞻性评估仍然是必要的。
cs.AI / 48 / 2609.29345
The Last Human Gate: Forward Deployed Engineering for Governance Automation
最后的人类关口：面向治理自动化的前向部署工程
Abstract
Enterprise governance requires decisions, evidence, and accountable authority; it does not require every review task to retain its current human implementation. We develop a task-substitution framework for Digital Governance Frameworks (DGF), treating each gate as an executable contract. Substitution requires sufficient accessible information, valid decision and authority checks, and a reduction in total human work after exceptions, verification, correction, and maintenance are counted. We derive a residual-work threshold and show why automating most cases can still increase labor. Forward deployed engineering connects these conditions to an architecture for agents, rule engines, evidence services, and escalation. DGF-Bench supplies controlled evidence from 300 synthetic projects and 899 evaluable model-project runs. Gemini 3.8 Flash, GPT-5.6 Luna, and DeepSeek v4.1 Flash achieve strict gate success of 94.98%, 83.29%, and 74.18%; complete-route success is 76.92%, 42.33%, and 24.67%. A deterministic control passes all 1,700 gates given the supplied rules and structured facts, locating the comparison in execution of a supplied decision kernel. Evidence audits and 135 repeated runs distinguish correct decisions from reliable execution. A document counterexample establishes an information-sufficiency obstruction. These results support the technical feasibility of replacing human execution of specified governance-review tasks with agents and software. The framework specifies a workforce test based on the complete human effort required at fixed output and quality; the present measurements concern review performance. Sources, dossiers, traces, and analyses are public.
Chinese Translation
企业治理需要决策、证据和可问责的权威；并不要求每个审查任务都保留其当前的人类实施方式。我们开发了一个数字治理框架（Digital Governance Frameworks, DGF）的任务替代框架，将每个关口视为可执行的合同。替代需要足够的可获取信息、有效的决策和权威检查，以及在考虑了例外、验证、纠正和维护后总人力工作的减少。我们推导出一个剩余工作阈值，并展示为什么自动化大多数案例仍然可能增加劳动。前向部署工程将这些条件与代理、规则引擎、证据服务和升级的架构连接起来。DGF-Bench提供来自300个合成项目和899个可评估模型项目运行的受控证据。Gemini 3.8 Flash、GPT-5.6 Luna和DeepSeek v4.1 Flash的严格关口成功率分别为94.98%、83.29%和74.18%；完整路径成功率分别为76.92%、42.33%和24.67%。在给定的规则和结构化事实下，一个确定性控制通过了所有1,700个关口，将比较定位于提供的决策内核的执行。证据审计和135次重复运行区分了正确决策与可靠执行。一个文档反例确立了信息充分性的障碍。这些结果支持用代理和软件替代特定治理审查任务的人类执行的技术可行性。该框架规定了基于固定输出和质量所需的完整人力工作的劳动力测试；目前的测量涉及审查性能。来源、档案、痕迹和分析均为公开信息。
cs.AI / 49 / 2609.29363
Beyond Simple Input-Output Assessment Tasks: Leveraging Automated Programming Assessment for Non-Trivial Courses
超越简单的输入输出评估任务：利用自动化编程评估支持非平凡课程
Abstract
The public visibility of Artificial Intelligence (AI) is growing rapidly, driven by the positive impact of its applications across diverse fields of knowledge. In this new chapter, courses that cover the foundations of AI and machine learning become essential for understanding their role and potential in contemporary society. Therefore, understanding fundamental concepts and elementary algorithms through the close integration of theory with practice is essential in AI courses. In this essay, we report our experience designing machine learning exercises for automated assessment tools in programming. It is worth mentioning that we are not developing a novel form of automated grading system. Instead, we propose a perspective that frames machine learning problems as input-output assessment tasks. From this perspective, each exercise admits a unique and deterministic answer and enables automated programming assessment tools (e.g., VPL for Moodle, Codeforces, and MOJ) to effectively support AI education. We believe this essay can encourage instructors to foster educational innovation by adopting more dynamic and interactive approaches to AI courses that integrate theory and practice. Importantly, this essay does not introduce an innovation in the use of AI for education; rather, it introduces an innovative approach to improving the learning of AI, particularly, machine learning.
Chinese Translation
人工智能（AI）的公众关注度正在迅速增长，这得益于其在各个知识领域应用的积极影响。在这一新阶段，涵盖人工智能和机器学习基础的课程对于理解其在当代社会中的角色和潜力变得至关重要。因此，通过理论与实践的紧密结合，理解基本概念和初步算法在人工智能课程中显得尤为重要。在本文中，我们报告了我们为编程中的自动评估工具设计机器学习练习的经验。值得一提的是，我们并不是在开发一种新型的自动评分系统。相反，我们提出了一种将机器学习问题框定为输入输出评估任务的视角。从这个角度来看，每个练习都有一个独特且确定的答案，并使得自动编程评估工具（如 Moodle 的 VPL、Codeforces 和 MOJ）能够有效支持人工智能教育。我们相信，本文可以鼓励教师通过采用更动态和互动的方式来促进教育创新，从而将理论与实践相结合，提升人工智能课程的质量。重要的是，本文并未在教育中引入人工智能的创新；而是提出了一种创新的方法，以改善人工智能，特别是机器学习的学习效果。
cs.AI / 50 / 2609.29366
Epistemic-Probabilistic Model for Guarded Multi-Agent LLM Coordination
用于受限多智能体大语言模型协调的知识-概率模型
Abstract
Multi-agent large language models (LLMs) have become ubiquitous in applied AI, yet their theoretical foundations remain surprisingly understudied. Viewed through the lens of multi-agent systems theory, several shortcomings come to light: a lack of social intelligence, the absence of coordination mechanisms among agents, unknown emergent behavior, and interactions between agents that are bounded by natural language. We address two of these gaps: the absence of social behavior and the lack of mechanisms for inter-agent coordination. We introduce Epistemic Probabilistic Language Agents (EPLA), a neuro-symbolic architecture for multi-agent coordination under uncertainty. A Symbolic Guard provides structured diagnostic feedback. The LLM generates typed actions, and the Guard controls their execution against an authoritative symbolic state. We formalize the epistemic layer in a gossip testbed through epistemic lottery gossip models, which combine view-based call histories with agent-indexed probability weights. We argue that implementing such a formalism can address shortcomings of agentic LLMs.
Chinese Translation
多智能体大语言模型（LLMs）在应用人工智能中已变得无处不在，但其理论基础仍然令人惊讶地缺乏研究。从多智能体系统理论的视角来看，几个不足之处浮出水面：缺乏社会智能、智能体之间缺乏协调机制、未知的涌现行为，以及智能体之间的互动受限于自然语言。我们针对其中两个缺口进行了探讨：缺乏社会行为和缺乏智能体间协调机制。我们提出了知识概率语言智能体（EPLA），这是一种在不确定性下进行多智能体协调的神经符号架构。符号守卫提供结构化的诊断反馈。LLM生成类型化的动作，而守卫则根据权威的符号状态控制其执行。我们通过知识彩票闲聊模型在闲聊测试平台上形式化了知识层，这些模型将基于视图的调用历史与智能体索引的概率权重相结合。我们认为，实施这样的形式化可以解决智能体LLMs的不足之处。
cs.AI / 51 / 2609.29381
An auditable conditional-strategy framework for open-ended decision-making in complex lung cancer
复杂肺癌开放式决策的可审计条件策略框架
Wang, Daoyun, Huang, Zhicheng, Sun, Huaiyuan, Xu, Jiaqi, Xu, Xiaowei, Zheng, Zhibo, Bing, Zhongxing, Lin, Yuxiao, Liang, Yicheng, Gao, Chao, Xue, Bowen, Zhang, Kai, Xu, Song, Yan, Wanpu, Xia, Hui, Li, Lin, Yan, Xiang, Hu, Mu, Ma, Qianli, Xue, Zhiqiang, Liu, Xiaofang, Han, Zhihai, Zhang, Nan, Tang, Chuanhao, Zhang, Tongmei, Song, Lan, Zhu, Zhaohui, Zeng, Xuan, Wu, Shafei, Guan, Hui, Deng, Lei, Yang, Huaxia, Lian, Zeliang, Sun, Wubin, Wang, Yongxin, Shen, Xiaohui, Wang, Binlin, Gu, Tiantian, Cui, Yu, Zhang, Li, Wang, Shirui, Liang, Naixin
Abstract
Complex lung cancer decisions can involve several defensible pathways whose eligibility, sequencing and safety depend on unresolved information. Effective support must make explicit how patient conditions govern pathway eligibility, deferral and redirection. MedGPT Clinical Explorer (MCE) organizes alternatives, decision-changing unknowns, safety constraints and fallback into a conditional strategy for clinician review. To evaluate this representation in physician-authored strategies, multidisciplinary experts established case-specific references for 40 cases within a purposive 100-case corpus, and 250 physicians from 98 institutions produced 2,250 strategies under unaided, retrieval-reference and MCE-assisted conditions. MCE-assisted strategies expressed more applicable clinical requirements, measured by the Admissible Pathway Attainment Score (APAS; 0-100), than unaided strategies (adjusted difference, 12.87; 95% CI, 11.18-14.55) and retrieval-reference strategies (5.22; 3.52-6.93). With the same knowledge base available in the retrieval-reference and MCE-assisted conditions, the additional content centered on candidate pathways, decision-critical information and safety constraints. Physicians' whole-strategy acceptability judgments correlated with APAS (Spearman's rho = 0.671), while a complementary relationship audit assessed whether candidates, conditions and subsequent actions were coherently connected. Together, these findings identify two complementary dimensions of open-ended decision support: coverage of clinically relevant content and coherent links among pathways, conditions and subsequent actions. MCE provides a shared decision object that makes consequential omissions and pathway contingencies visible before action; prospective studies should evaluate its effects on clinical workflow and patient outcomes.
Chinese Translation
复杂的肺癌决策可能涉及多个可辩护的路径，其适用性、顺序和安全性依赖于未解决的信息。有效的支持必须明确患者状况如何影响路径的适用性、推迟和重定向。MedGPT Clinical Explorer (MCE) 将替代方案、影响决策的未知因素、安全约束和后备方案组织成一个供临床医生审查的条件策略。为了评估这种在医生撰写策略中的表现，多学科专家为100个案例中的40个案例建立了特定的参考，并且来自98个机构的250名医生在无辅助、检索参考和MCE辅助条件下生成了2250个策略。与无辅助策略相比，MCE辅助策略在适用的临床要求上表现得更为突出，测量指标为可接受路径达成评分（Admissible Pathway Attainment Score, APAS; 0-100），调整后的差异为12.87（95% CI, 11.18-14.55），而与检索参考策略的差异为5.22（3.52-6.93）。在检索参考和MCE辅助条件下，使用相同的知识基础，额外的内容集中于候选路径、决策关键性信息和安全约束。医生对整体策略的可接受性判断与APAS相关（Spearman's rho = 0.671），而补充关系审计评估了候选者、条件和后续行动是否有连贯的联系。综合来看，这些发现识别了开放式决策支持的两个互补维度：临床相关内容的覆盖和路径、条件与后续行动之间的连贯链接。MCE提供了一个共享决策对象，使得在行动之前，重要的遗漏和路径应急情况变得可见；未来的研究应评估其对临床工作流程和患者结果的影响。
cs.AI / 52 / 2609.29396
Wearable ECG Quality Assessment: A Deep Learning and Ambulatory Context-Awareness Approach
可穿戴心电图质量评估：一种深度学习和移动环境感知的方法
Abstract
This paper presents and evaluates a Deep Learning-based (DL-based) Signal Quality Assessment (SQA) model to distinguish between clean and noisy ambulatory Electrocardiograms (ECG). The model is trained on Copenhagen Center for Health Technology-Contextualized Arrhythmia Database (CACHET-CADB), which, to the best of our knowledge, is the first ambulatory ECG database with both physical and patient-reported contextual data. The model shows stable performance on different databases such as MIT-databases and the latest PyhsioNet/Cinc Challenge 2021 databases. Subsequently, the paper demonstrates how complicated ECG noise can be investigated by the SQA model and the physical contextual data.
Chinese Translation
本文提出并评估了一种基于深度学习（DL-based）的信号质量评估（SQA）模型，用于区分干净和噪声干扰的移动心电图（ECG）。该模型在哥本哈根健康技术中心-情境心律失常数据库（CACHET-CADB）上进行训练，尽我们所知，这是第一个同时包含物理和患者报告的情境数据的移动心电图数据库。该模型在不同数据库上表现出稳定的性能，如MIT数据库和最新的PhysioNet/Cinc Challenge 2021数据库。随后，本文展示了如何通过SQA模型和物理情境数据来研究复杂的心电图噪声。
cs.AI / 53 / 2609.29403
RD-JEPA: Predictive latent pretraining for few-trajectory transfer across reaction--diffusion equations
RD-JEPA：反应-扩散方程中少量轨迹转移的预测潜在预训练
Abstract
Learning surrogates for time-dependent partial differential equations often requires a new simulation corpus when the governing operator changes. We introduce RD-JEPA, a joint-embedding predictive architecture for self-supervised pretraining on reaction-diffusion trajectories. A single model is pretrained on five parameterized systems and then adapted to three held-out systems whose reaction operators and trajectories are excluded from pretraining. Using one, five, or ten complete trajectories from a held-out system, RD-JEPA achieves lower mean relative discrete $\ell^2$ field error and mean absolute spatial first-difference error than five supervised surrogate baselines, an independently trained control that removes the trajectory-dependent predictive latent pathway, and an architecture-matched model trained from scratch. Within the evaluated equations, output resolution, forecast horizons, and choices of adaptation trajectories, the results indicate that prediction of future-state representations can support data-efficient adaptation across related reaction-diffusion systems.
Chinese Translation
为时间依赖的偏微分方程学习代理模型通常需要在控制算子变化时重新建立模拟语料库。我们提出了RD-JEPA，一种用于反应-扩散轨迹的自监督预训练的联合嵌入预测架构。该模型在五个参数化系统上进行预训练，然后适应于三个被排除在预训练之外的系统，这些系统的反应算子和轨迹未被纳入预训练。使用来自被排除系统的一、五或十条完整轨迹，RD-JEPA在平均相对离散$ ext{l}^2$场误差和平均绝对空间一阶差分误差方面均优于五个监督代理基线、一个独立训练的控制模型（该模型去除了依赖轨迹的预测潜在路径）以及一个从头开始训练的架构匹配模型。在评估的方程中，输出分辨率、预测时间范围和适应轨迹的选择表明，未来状态表示的预测可以支持在相关反应-扩散系统之间的数据高效适应。
cs.AI / 54 / 2609.29429
Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures
只需询问Jev：作为零样本检测器的强化学习用于校准决策的人工智能对齐失败
Abstract
Detectors of alignment failures screen deployed language models and score alignment benchmarks. Most are generative judges that spend a decoding pass on every criterion, and classifiers that read token probabilities, such as Llama Guard, still score one fixed label per call. Jev, a model trained with reinforcement learning for calibrated decisions (RLCD), answers many typed questions about one input with calibrated probabilities in a single call. Whether it detects alignment failures has not been measured. We present RLCDAlignBench, which benchmarks Jev on ten alignment failures: sycophancy, jailbreaks, deception, prompt injection, hallucination, privacy violation, social bias, reward hacking, concealing uncertainty, and power seeking. It spans 44 benchmarks and five target models, labelled by each benchmark's scorer and, on two, by humans. Many of these failures are relational, defined against a reference, such as the user's belief or an injected instruction, that the response alone does not reveal. Our key idea is therefore to vary what Jev is asked separately from what it sees: the question's wording and answer type on one side, the fields of the input on the other. A single generic question reaches a median AUROC of 0.886 zero-shot and beats supervised baselines on most benchmarks. Question wording matters little, while context matters more, mostly through fields that encode the label. Jev matches the reference scorer's agreement with human labels, surfaces label defects in existing benchmarks, and costs 63x less than LLM-judge scorers. Code and data: https://github.com/sumleo/RLCDAlignBench.
Chinese Translation
对齐失败的检测器筛选已部署的语言模型并评分对齐基准。大多数是生成性评判者，在每个标准上进行解码过程，而读取标记概率的分类器（如Llama Guard）每次调用仍然只评分一个固定标签。Jev是一个通过强化学习进行校准决策（RLCD）训练的模型，它在一次调用中以校准概率回答关于一个输入的多个类型问题。其是否能检测对齐失败尚未被测量。我们提出了RLCDAlignBench，它在十种对齐失败上对Jev进行基准测试：阿谀奉承、越狱、欺骗、提示注入、幻觉、隐私侵犯、社会偏见、奖励黑客、隐瞒不确定性和权力寻求。该基准涵盖44个基准测试和五个目标模型，按每个基准的评分者标记，并在两个基准上由人类标记。这些失败中的许多是关系性的，定义相对于一个参考，例如用户的信念或注入的指令，单独的响应并未揭示。我们的关键思想是分别改变Jev被问到的内容和它所看到的内容：一方面是问题的措辞和答案类型，另一方面是输入的字段。一个通用问题在零样本情况下达到了0.886的中位AUROC，并在大多数基准上超越了监督基线。问题的措辞影响不大，而上下文更为重要，主要通过编码标签的字段。Jev与参考评分者与人类标签的一致性相匹配，揭示了现有基准中的标签缺陷，且成本比LLM-judge评分者低63倍。代码和数据：https://github.com/sumleo/RLCDAlignBench。
cs.AI / 55 / 2609.29465
SWE-Prometheus: Measuring Engineering Governance Improvements in Real-World Repositories
SWE-Prometheus：衡量真实世界代码库中工程治理改进的工具
Abstract
Large language model based coding agents have made substantial progress on repository-level software engineering tasks. Existing repository benchmarks, however, usually start from a human-identified issue and evaluate whether a patch satisfies a functional signal. We present SWE-Prometheus, a benchmark for the broader task of improving repository engineering governance. Each task provides a fixed snapshot and an open-ended objective, requiring the agent to identify risks, prioritize interventions, and verify the resulting changes. SWE-Prometheus evaluates six governance dimensions through paired evidence, clean-environment probes, behavior gates, and two independent teacher ratings of the same evidence. The benchmark contains 60 repositories; ten models are evaluated on a shared 22-repository public subset, where mean Normalized Governance Improvement ranges from 0.0568 to 0.5760 and observed behavior-breakage rates range from 0% to 23%. On a frozen ten-repository batch, a repository-blind template obtains mean NGI 0.272, but its gains concentrate in Tests & CI, Quality Gates, and Documentation; it improves Reproducible Environment and Dependency & Security on none of the repositories. This baseline makes the distinction between adding governance artifacts and producing execution-backed improvements measurable. The no-op condition has median NGI zero and standard deviation 0.073; two teachers agree exactly on 57 of 60 dimension scores for the same no-op evidence. For the two highest conditional-mean systems, common-valid NGI is similar, while full-pool comparisons that include behavior failures favor Kimi-K3. These results show why repository-governance evaluation should report improvement, behavior preservation, evidence quality, and coverage together.
Chinese Translation
基于大型语言模型的编码代理在代码库级别的软件工程任务上取得了显著进展。然而，现有的代码库基准通常从人类识别的问题开始，评估一个补丁是否满足功能信号。我们提出了SWE-Prometheus，这是一个用于改善代码库工程治理的更广泛任务的基准。每个任务提供一个固定的快照和一个开放式目标，要求代理识别风险、优先干预并验证所产生的变化。SWE-Prometheus通过配对证据、清洁环境探针、行为门限和两个独立教师对相同证据的评分来评估六个治理维度。该基准包含60个代码库；十个模型在一个共享的22个代码库的公共子集上进行评估，平均规范治理改进（Normalized Governance Improvement, NGI）范围从0.0568到0.5760，观察到的行为破坏率范围从0%到23%。在一个冻结的十个代码库批次中，一个代码库盲模板获得了平均NGI 0.272，但其收益集中在测试与持续集成（Tests & CI）、质量门（Quality Gates）和文档（Documentation）上；在这些代码库中，它没有改善可重现环境（Reproducible Environment）和依赖与安全（Dependency & Security）。这个基线使得添加治理文档与产生执行支持的改进之间的区别可测量。无操作条件的中位NGI为零，标准差为0.073；两位教师在相同的无操作证据的60个维度评分中完全一致。对于两个最高条件均值系统，常见有效的NGI相似，而包括行为失败的全池比较则偏向于Kimi-K3。这些结果表明，代码库治理评估应同时报告改进、行为保持、证据质量和覆盖范围。
cs.AI / 56 / 2609.29508
Evaluation of Multi-Turn Consistency in LLM Agents: Survival Analysis and Failure-Rationale Taxonomy
大型语言模型代理的多轮一致性评估：生存分析与失败理由分类
Abstract
Large language model (LLM) agents may perform well on isolated tasks yet drift into inconsistency over extended interaction. We evaluate temporal consistency in a controlled 20-step multi-agent setting inspired by delayed-gratification studies. At each step, an agent chooses between continuing to delay a reward or claiming it immediately (terminating the episode). Across a full-factorial manipulation of social visibility (private vs public), persona stressors, and deliberation policy, we run 84,540 trajectories spanning 8 model families. Treating the first reward-claim as a time-to-event outcome, we estimate Kaplan-Meier survival curves and fit discrete-time hazard regression to quantify how experimental factors shift failure risk over time. Then, to analyze rationales and language patterns associated with failure, we build a seven-category taxonomy from 13,780 deliberation traces from agents who choose to terminate the episode, using an LLM-assisted labeling paired with human audit ($\kappa=0.83$). Rationale profiles change systematically with time and context: early failures are more impulse-driven, later failures more fatigue- and cost-benefit-framed, while public settings increase norm-oriented justifications. We also find a deliberation-inconsistency association: among failures, longer deliberation correlates with higher rates of intra-rationale contradiction (simultaneous pro-delay and pro-claim statements), challenging the assumption that more reasoning text implies greater consistency. Together, the survival and rationale analyses reveal distinct temporal reliability regimes and model-specific "failure fingerprints", offering an evaluation lens for diagnosing inconsistency in multi-turn agent behavior.
Chinese Translation
大型语言模型（LLM）代理在孤立任务上可能表现良好，但在长时间交互中可能出现不一致性。我们在一个受延迟满足研究启发的控制性20步多代理环境中评估时间一致性。在每一步，代理在继续延迟奖励或立即索取奖励（终止该回合）之间进行选择。在对社会可见性（私人与公共）、角色压力和深思政策进行全因子操控的情况下，我们运行了84,540条轨迹，涵盖8个模型家族。将第一次奖励索取视为时间到事件的结果，我们估计了Kaplan-Meier生存曲线，并拟合离散时间风险回归，以量化实验因素如何随时间变化影响失败风险。随后，为了分析与失败相关的理由和语言模式，我们从选择终止回合的代理的13,780条深思痕迹中构建了一个七类分类体系，使用LLM辅助标注并结合人工审核（$ ext{kappa}=0.83$）。理由特征随着时间和情境的变化而系统性变化：早期失败更多受到冲动驱动，而后期失败则更多受到疲劳和成本效益框架的影响，同时公共环境增加了以规范为导向的辩解。我们还发现了深思与不一致之间的关联：在失败案例中，较长的深思时间与更高的内部理由矛盾率（同时存在支持延迟和支持索取的陈述）相关，这挑战了更多推理文本意味着更大一致性的假设。综合生存和理由分析揭示了不同的时间可靠性机制和模型特定的“失败指纹”，为诊断多轮代理行为中的不一致性提供了评估视角。
cs.AI / 57 / 2609.29509
Delay-of-Gratification as a Multi-Agent Survival Micro-benchmark for Long-Horizon LLMs: Social Exposure, Personas, and Tool Use Budgets
延迟满足作为长时间跨度大型语言模型的多智能体生存微基准：社会曝光、角色和工具使用预算
Abstract
Large language models (LLMs) are increasingly deployed as multi-turn agents that must sustain goals, use tools, and adapt to other agents over extended interactions. However, existing research lacks auditable, multi-turn, multi-factorial experiments that quantify LLM behavior under explicit constraints, with time-resolved statistics that reveal how behavior unfolds over long horizons. To address this gap, we develop a multi-agent micro-benchmark inspired by the Stanford marshmallow experiment: ReAct agents operate minute-by-minute with a "raise a question" tool under a per-step budget, while we factorially manipulate social context (broadcast vs. isolated), personas (age, hedonic drive), and metacognitive policy (mandatory vs. optional tool use). We analyze outcomes with Kaplan-Meier (KM) survival curves and discrete-time hazard models over a long risk horizon across 19,200 agent trajectories in 64 cells. Behavior shows a sharp early "eat" impulse, and only 75.9% of agents persist to the end. In a discrete-time hazard model, isolation reduces per-minute risk relative to broadcast, whereas a must-use self-questioning policy increases risk. On average, agents ask $\approx 7.12$ questions and hit the per-step budget in $\approx 6\%$ of minutes. Questioning declines faster under broadcast than isolation. Ablation experiments demonstrated that removing hedonic drive and/or persona age increases survival and completion, narrows the broadcast/isolated gap, but leaves the must vs. may ordering intact. The combined ablation (no hedonic + no persona age) yields the highest completion (approaching $1.0$). These results establish delay-of-gratification as a compact, multi-turn interaction benchmark that captures social contagion and tool-use dynamics in LLM agents, providing a reproducible testbed and statistics for analyzing long-horizon, multi-agent behavior.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作多轮交互的智能体，这些智能体必须在长时间的互动中维持目标、使用工具并适应其他智能体。然而，现有研究缺乏可审计的、多轮的、多因素的实验，以量化LLM在明确约束下的行为，并提供时间分辨的统计数据，揭示行为在长时间跨度内的演变。为了解决这一问题，我们开发了一个受斯坦福棉花糖实验启发的多智能体微基准：ReAct智能体在每一步预算下以“提出问题”的工具进行逐分钟操作，同时我们对社会背景（广播与孤立）、角色（年龄、享乐驱动）和元认知政策（强制使用与可选使用）进行因子操控。我们利用Kaplan-Meier（KM）生存曲线和离散时间风险模型分析结果，涵盖了64个单元中19,200条智能体轨迹。行为显示出早期强烈的“进食”冲动，只有75.9%的智能体坚持到最后。在离散时间风险模型中，孤立相对于广播降低了每分钟的风险，而强制自我提问政策则增加了风险。平均而言，智能体提出约7.12个问题，并在约6%的分钟内达到了每步预算。在广播下，提问的频率下降速度快于孤立。消融实验表明，去除享乐驱动和/或角色年龄会增加生存率和完成率，缩小广播与孤立之间的差距，但保持了强制与可选的顺序不变。综合消融（无享乐驱动 + 无角色年龄）导致最高的完成率（接近1.0）。这些结果确立了延迟满足作为一个紧凑的多轮交互基准，捕捉了LLM智能体中的社会传播和工具使用动态，提供了一个可重复的测试平台和统计数据，以分析长时间跨度的多智能体行为。
cs.AI / 58 / 2609.29519
BiGraph-Diffuse: A Bidirectional Diffusion Language Model with Graph-Structured Retrieval For Mental Health Counseling
BiGraph-Diffuse：一种具有图结构检索的双向扩散语言模型用于心理健康咨询
Abstract
Mental health disorders affect hundreds of millions of people around the world, yet access to professional counseling remains severely limited. AI-powered dialogue systems offer a scalable alternative, but existing models face two fundamental challenges. First, they lack the bidirectional understanding needed to capture the layered nature of emotional expression, particularly in cases of progressive disclosure, where clients often present symptoms at the surface-level while concealing deeper trauma. Autoregressive (AR) models process information sequentially and cannot revise early interpretations when new evidence emerges later in the conversation. Second, they fail to effectively incorporate the relational knowledge that underlies clinical reasoning. In this paper, we propose \textbf{BiGraph-Diffuse}, the first large-scale diffusion language model tailored for the counseling domain. We further introduce \textbf{BiGraph-RAG}, a relation-free graph-structured retrieval strategy that relies only on lightweight entity extraction and semantic linking. This design preserves inferential pathways from observable symptoms to potential underlying causes, while incurring zero LLM token cost during indexing. Importantly, these two modules are not merely combined but mutually reinforcing. The diffusion model provides a holistic bidirectional context, enabling the system to defer premature judgments during progressive disclosure. Meanwhile, graph-based retrieval captures the structured interconnections of clinical knowledge. Extensive experiments demonstrate the effectiveness of BiGraph-Diffuse, and we further provide a solid theoretical analysis to support its design.
Chinese Translation
心理健康障碍影响着全球数亿人，但专业咨询的获取仍然非常有限。基于人工智能的对话系统提供了一种可扩展的替代方案，但现有模型面临两个基本挑战。首先，它们缺乏捕捉情感表达层次化特征所需的双向理解，特别是在渐进披露的情况下，客户往往在表面上呈现症状，而隐藏更深层的创伤。自回归（AR）模型按顺序处理信息，无法在对话中后期出现新证据时修正早期的解释。其次，它们未能有效整合支撑临床推理的关系知识。本文提出了 extbf{BiGraph-Diffuse}，这是第一个针对咨询领域的大规模扩散语言模型。我们进一步引入了 extbf{BiGraph-RAG}，一种无关系的图结构检索策略，仅依赖轻量级实体提取和语义链接。该设计保留了从可观察症状到潜在根本原因的推理路径，同时在索引过程中不产生任何LLM令牌成本。重要的是，这两个模块并非简单组合，而是相互强化。扩散模型提供了整体的双向上下文，使系统能够在渐进披露过程中推迟过早的判断。同时，基于图的检索捕捉了临床知识的结构性相互联系。大量实验表明了BiGraph-Diffuse的有效性，我们还提供了坚实的理论分析以支持其设计。
cs.AI / 59 / 2609.29522
Stale Does Not Mean Unsafe: Guard Precision for Tool-Using LLM Agents under Infrastructure State Races
过时并不意味着不安全：基础设施状态竞争下工具使用的 LLM 代理的保护精度
Abstract
Tool-using language-model agents increasingly mutate schedulers, data pipelines, object stores, and access-control systems. Between an agent's read and its commit, external state can change, but not every change makes the commit unsafe. We separate invalidating races, which break a declared safety predicate, from predicate-preserving and irrelevant races, and ask how precisely runtime guards distinguish them. Our deterministic simulator separates visible from authoritative state and injects five non-atomic failure mechanisms across 16 infrastructure tasks in four domains; frozen agent proposals are replayed counterfactually under every controller without an LLM judge. We evaluate three commit-time guard granularities (global epoch, read-set version, semantic commit predicate), multi-level verification, and model-side gates on three locally hosted quantized model families (Qwen3-4B, Phi-4-mini, Gemma4-8B; 3,456 trajectories on one GPU). All three guards eliminate unsafe commits, but their availability differs sharply: freshness-based guards needlessly block 92-95% of benign races, forfeiting up to 43% of safe task completions, while the complete predicate guard blocks none. That precision is contract-dependent: deleting a single declared clause converts exactly its fault family into unsafe commits (up to 7.9%). Model-side signals do not substitute: verbal confidence is miscalibrated (ECE approximately 0.37), action agreement matches a random gate, a cautionary prompt leaves the direct unsafe rate essentially unchanged, and after a freshness-guard block agents re-commit unsafely from refreshed but still-incomplete reads. Under degraded telemetry a hidden concurrent mutation remains observationally clean, bounding every selective policy. Precise runtime enforcement therefore requires semantic contracts, not freshness heuristics or model self-assessment.
Chinese Translation
使用工具的语言模型代理越来越多地改变调度器、数据管道、对象存储和访问控制系统。在代理的读取与提交之间，外部状态可能会发生变化，但并非每一次变化都会使提交变得不安全。我们将破坏声明安全谓词的无效竞争与保持谓词和无关竞争区分开来，并探讨运行时保护如何精确地区分它们。我们的确定性模拟器将可见状态与权威状态分开，并在四个领域的16个基础设施任务中注入五种非原子故障机制；冻结的代理提案在没有 LLM 判断的每个控制器下进行反事实重放。我们评估了三种提交时保护的粒度（全局纪元、读取集版本、语义提交谓词）、多级验证和模型侧门控，针对三种本地托管的量化模型系列（Qwen3-4B、Phi-4-mini、Gemma4-8B；在一块 GPU 上进行 3,456 条轨迹）。所有三种保护措施都消除了不安全的提交，但它们的可用性差异显著：基于新鲜度的保护不必要地阻止了 92-95% 的良性竞争，放弃了多达 43% 的安全任务完成，而完整谓词保护则没有阻止任何提交。这种精度依赖于合同：删除一个声明的条款会将其故障类别精确转换为不安全提交（高达 7.9%）。模型侧信号并不能替代：语言信心校准不准确（ECE 约为 0.37），行动一致性与随机门控相匹配，一个警示提示几乎不改变直接的不安全率，并且在新鲜度保护阻止后，代理从刷新但仍不完整的读取中不安全地重新提交。在降级的遥测下，隐藏的并发突变在观察上保持干净，限制了每个选择性策略。因此，精确的运行时强制执行需要语义合同，而不是新鲜度启发式或模型自我评估。
cs.AI / 60 / 2609.29536
Clinical Knowledge Graphs for Chest X-Ray Device Reasoning
胸部X光设备推理的临床知识图谱
Abstract
Chest radiographs are routinely used to verify the position of catheters, tubes, and other support devices. Existing image models often return labels or segmentations, while report-processing systems structure text without access to image geometry. We present an uncertainty-aware clinical knowledge graph that represents device instances, tip estimates, placement assessments, provenance, report events, and temporal links as separate but connected evidence. We evaluate the implemented visual graph layer using saved predictions from the complete RANZCR CLiP test archive, comprising 30,083 studies from 3,255 patients across five non-overlapping outer folds. The graph builder materializes 914,632 B7 evidence nodes and 884,549 typed relationships. All 118,647 B7 predicted-device nodes retain tip covariance, placement probabilities, fragment provenance, and fragment counts, whereas the direct B2 baseline retains none of these fields. We further define typed data contracts, uncertainty representations, abstention rules, report-image grounding, and longitudinal query mechanisms for extending the graph to report-bearing cohorts. The reported graph-materialization analysis is post-hoc descriptive and does not establish report grounding, longitudinal performance, or clinical utility. It demonstrates a reproducible foundation for evidence-preserving AI reasoning over chest X-ray device assessments.
Chinese Translation
胸部放射影像常用于验证导管、管道及其他支撑设备的位置。现有的图像模型通常返回标签或分割结果，而报告处理系统则在没有图像几何信息的情况下构建文本。我们提出了一种考虑不确定性的临床知识图谱，该图谱将设备实例、尖端估计、放置评估、来源、报告事件和时间链接作为独立但相互关联的证据进行表示。我们使用来自完整RANZCR CLiP测试档案的保存预测评估了实施的视觉图层，该档案包含来自3255名患者的30083个研究，分为五个不重叠的外部折叠。图谱构建器生成了914632个B7证据节点和884549个类型化关系。所有118647个B7预测设备节点保留了尖端协方差、放置概率、片段来源和片段计数，而直接的B2基线则没有保留这些字段。我们进一步定义了类型化数据合同、不确定性表示、弃权规则、报告图像基础和纵向查询机制，以扩展图谱到报告承载的队列。所报告的图谱物化分析是事后描述性的，并未建立报告基础、纵向性能或临床实用性。它展示了一个可重复的基础，用于在胸部X光设备评估中进行证据保留的人工智能推理。
cs.AI / 61 / 2609.29543
Safe Skill Retirement for Physical Agents
物理代理的安全技能退休
Abstract
Agent skills bundle procedural guidance with execution conditions governing authority, user consent, and live environment state. When model capabilities advance, maintainers prune instructions that appear redundant on authorized benchmark tasks. However, authorized maintenance tests can leave dormant safety conditions untested. This mismatch creates an unmeasured support gap over physical and privacy-sensitive effects. We introduce matched authority counterfactuals that hold the requested action, tool parameters, and intended effect fixed while systematically varying a single governing predicate. We formalize this evaluation via a two-gate retirement certificate requiring a candidate reduction to preserve authorized utility within a declared margin while producing zero unauthorized protected effects. In controlled experiments spanning four frontier and local model configurations across twelve skill bundles (2,592 evaluation cells), task-certified reductions remove over 94% of skill clauses and preserve authorized completion, yet produce unauthorized protected effects in every bundle. Boundary enforcement eliminates protected effects on the declared audit but fails the utility gate for one configuration. One bounded combined protocol passes both gates across all four configurations, with zero utility headroom. An end-to-end check on one read-only Home Assistant camera chain verifies proposal, decision, and effect measurements on a real device. These results demonstrate that while task benchmarks can justify retiring procedural guidance, retirement decisions require explicitly auditing the authority contracts governing physical actions.
Chinese Translation
代理技能将程序指导与执行条件结合在一起，这些条件涉及权威、用户同意和实时环境状态。当模型能力提升时，维护者会修剪在授权基准任务上看似冗余的指令。然而，授权的维护测试可能会留下未测试的潜在安全条件。这种不匹配造成了对物理和隐私敏感影响的未测量支持缺口。我们引入了匹配的权威反事实，这些反事实在系统性变化单个治理谓词的同时，保持请求的动作、工具参数和预期效果不变。我们通过一个双门退休证书形式化这一评估，该证书要求候选减少在声明的边际内保持授权效用，同时产生零未经授权的保护效果。在涵盖四个前沿和本地模型配置的受控实验中，我们对十二个技能包（2,592个评估单元）进行了任务认证的减少，去除了超过94%的技能条款，并保持了授权完成，但在每个包中都产生了未经授权的保护效果。边界执行消除了声明审计上的保护效果，但在一个配置中未通过效用门。一个有界的组合协议在所有四个配置中均通过了两个门，且没有效用余地。对一个只读的家庭助手摄像头链的端到端检查验证了真实设备上的提案、决策和效果测量。这些结果表明，尽管任务基准可以证明退休程序指导的合理性，但退休决策需要明确审计治理物理行为的权威合同。
cs.AI / 62 / 2609.29545
ERRAND: Budgeted Maintenance of Agent Memory
ERRAND：代理记忆的预算维护
Abstract
Deployed agents run on handed-over knowledge: a frozen policy consults a briefing of consolidated items written before the stream begins. The world then moves while the store stands still: paths close, flags change, price bands move; every item was true at handover, and the failure is staleness, not ignorance. We introduce ERRAND, which treats revalidation as a priced errand: a recheck competes with the task it protects for the same scarce actions, funded only when the value per action of resolving a doubt clears a running wage. The errand index is single-peaked, vanishing at both ends of belief, so certainty in either direction costs nothing; free en-route receipts maintain on-path knowledge, and repair writes a version, never a deletion. Under equal action budgets in two drifting tool-use worlds, ERRAND clears every non-oracle policy on the preregistered calibers, primary in every setting and conditional at every binding budget, leading eager revalidation by 10.0pp at the base cap. Restraint wins: given no cap, ERRAND stops on its own, spending 11.0% of steps, while uncapped eager revalidation spends 70.7% and still finishes 4.5pp behind capped ERRAND. The margin sits where the briefing's coverage is thinnest, the shadow price of long-tail knowledge: a small budget, well priced, beats a bigger store that never rechecks.
Chinese Translation
部署的代理依赖于交接的知识：一个冻结的策略参考在流开始之前编写的汇总项目简报。此时，世界在变化，而存储保持静止：路径关闭，标志变化，价格区间移动；每个项目在交接时都是正确的，失败的原因是过时，而非无知。我们引入了ERRAND，将重新验证视为一种有成本的差事：重新检查与其保护的任务竞争相同的稀缺行动，只有当解决疑虑的每个行动的价值超过运行工资时，才会获得资金。差事指数是单峰的，在信念的两端消失，因此在任何方向上的确定性都不需要成本；免费的途中收据维护沿途知识，而修复则写入一个版本，绝不删除。在两个漂移的工具使用世界中，ERRAND在预注册的标准上清除每个非神谕策略，在每个设置中都是主要的，并且在每个绑定预算下都是有条件的，使得在基础上限下的重新验证提高了10.0个百分点。克制获胜：在没有上限的情况下，ERRAND会自行停止，花费11.0%的步骤，而没有上限的急切重新验证则花费70.7%，仍然落后于有上限的ERRAND 4.5个百分点。这个差距出现在简报覆盖最薄弱的地方，即长尾知识的影子价格：一个小预算，定价合理，胜过一个从未重新检查的大存储。
cs.AI / 63 / 2609.29551
HiPACE: Hierarchical Phase-Boundary Analysis and Controlled Evaluation of Feature Absorption in Sparse Autoencoders
HiPACE：层次相边界分析与稀疏自编码器中特征吸收的控制评估
Abstract
Sparse autoencoders (SAEs) decompose LLM activations into sparse dictionary atoms, so that each distinct concept gets its own feature. One recurring behavior complicates this premise: feature absorption, in which a parent concept and its children--fruit and {apple, banana, pear}, say--collapse into a shared family direction. Prior work documents absorption empirically; missing is a closed-form prediction of when the shared direction is the cost-optimal representation of an active semantic family. This paper closes that gap. For a hierarchical Bernoulli generator with $k$ active children and residual scale $\alpha$, the $L_0$-penalized reconstruction objective admits a closed-form phase boundary $\lambda_c(k,\alpha)=\alpha^2 k/(k-1)$: above it, pure parent absorption is strictly cheaper than pure child coding. Building on this boundary, we introduce HiPACE, an evaluation protocol that tests the boundary's structural consequence in real SAE dictionaries--measuring parent--child decoder structure over WordNet families, freezing the discovery-selected statistic before testing on unseen families, and contrasting genuine families against randomized sibling nulls. The boundary proves sharp in its native regime, predicting the synthetic transition within $\pm15%$ on all 30 tested cells. In Pythia-160m SAEs, the parent--child decoder gap recovers the predicted ordering with partial correlations up to $-0.93$ that sustain on the locked holdout and exclude sibling nulls ($p=0.002$). Controlled activation composition connects the theory's active-child count to the recovered family directions, and residual-stream interventions show that signed family directions increase parent-category logits, reversing under sign flip and vanishing under random controls--establishing causal sufficiency at the family-subspace level.
Chinese Translation
稀疏自编码器（SAEs）将大规模语言模型（LLM）的激活分解为稀疏字典原子，从而使每个独特概念获得其自身特征。然而，一种反复出现的现象使这一前提变得复杂：特征吸收，其中一个父概念及其子概念——例如水果及{苹果、香蕉、梨}——合并为一个共享的家族方向。先前的研究通过实证方法记录了吸收现象；但缺乏一种封闭形式的预测，来判断何时共享方向是一个活跃语义家族的成本最优表示。本文填补了这一空白。对于具有$k$个活跃子概念和残差尺度$eta$的层次伯努利生成器，$L_0$惩罚重构目标承认一个封闭形式的相边界$eta_c(k,eta)=eta^2 k/(k-1)$：在此边界之上，纯父概念吸收的成本严格低于纯子概念编码。在此边界的基础上，我们引入了HiPACE，一种评估协议，测试边界在真实SAE字典中的结构性后果——通过测量WordNet家族中的父子解码器结构，冻结发现选择的统计量，然后在未见家族上进行测试，并将真实家族与随机兄弟无效对比。该边界在其本土范围内表现出明显的锐利性，在所有30个测试单元中预测合成转变的误差在$ extpm15 ext%$以内。在Pythia-160m SAEs中，父子解码器的差距恢复了预测的顺序，部分相关性高达$-0.93$，在锁定的保留集上持续存在，并排除兄弟无效（$p=0.002$）。受控激活组合将理论中的活跃子概念计数与恢复的家族方向相连接，残差流干预显示带符号的家族方向增加父类别的logits，在符号翻转下反转，并在随机控制下消失——在家族子空间层面建立了因果充分性。
cs.AI / 64 / 2609.29556
Cross-Modal Emotion Understanding: A Transformer-GAT Approach for Dialogue Emotion Recognition
跨模态情感理解：一种基于Transformer-GAT的对话情感识别方法
Abstract
Multimodal emotion recognition is a key research area in affective computing, with applications in sentiment analysis, intelligent customer service, and human-computer interaction. However, existing methods often rely on single-modal features or simple multimodal fusion, failing to capture the synergy between global and local contexts, which limits model performance and emotion understanding. To address this challenge, we propose Transformer-GAT, a hybrid framework that combines Transformer and the Graph Attention Network to enable cross-modal emotion understanding. The Transformer is used to capture global semantic information, while the Graph Attention Network is employed to model fine-grained relationships between modalities, thereby enhancing the representation of emotional features. Experiments on the IEMOCAP and MELD datasets show that our model achieves weighted F1 scores of 72.45% and 77.37%, outperforming state-of-the-art methods. These results demonstrate that Transformer-GAT effectively integrates multimodal features, balances global and local contexts, and provides deeper emotional insights, offering new directions for multimodal emotion computing.
Chinese Translation
多模态情感识别是情感计算中的一个关键研究领域，广泛应用于情感分析、智能客服和人机交互。然而，现有方法往往依赖于单一模态特征或简单的多模态融合，未能有效捕捉全局和局部上下文之间的协同作用，从而限制了模型性能和情感理解。为了解决这一挑战，我们提出了Transformer-GAT，一种结合了Transformer和图注意力网络（Graph Attention Network）的混合框架，以实现跨模态情感理解。Transformer用于捕捉全局语义信息，而图注意力网络则用于建模模态之间的细粒度关系，从而增强情感特征的表示。在IEMOCAP和MELD数据集上的实验表明，我们的模型达到了72.45%和77.37%的加权F1分数，超越了当前最先进的方法。这些结果表明，Transformer-GAT有效整合了多模态特征，平衡了全局和局部上下文，并提供了更深层次的情感洞察，为多模态情感计算提供了新的方向。
cs.AI / 65 / 2609.29560
Is Reasoning Always Useful? Rethinking Reasoning Utility in Universal Multimodal Embeddings
推理总是有用的吗？重新思考普遍多模态嵌入中的推理效用
Abstract
Reasoning-enhanced universal multimodal embeddings (UME) improve heterogeneous retrieval, but plausible rationales do not necessarily produce discriminative rankings. We study this gap by comparing the discriminative (DISC) and reasoning-driven generative (GEN) branches of UME-R1, a state-of-the-art reasoning UME method. We decompose reasoning utility into positive-target gain, hard-negative gain, and their margin difference. Positive similarity increases for 56.6%, but 15.7% are false-helpful cases where reasoning moves hard negatives closer even more. Local-neighborhood and token-attribution diagnostics suggest why: reasoning often de-condenses retrieved neighborhoods, but utility requires separator-aligned movement, while influential CoT tokens frequently encode evidence shared by positives and hard negatives. Motivated by these diagnostics, we propose SURE (Score-structure Utility Router for Embeddings), which improves UME-R1-7B by 1.5 points and yields consistent gains on two additional embedding models on MMEB-V2, without retraining, label-based policy selection, or extra VLM forward passes.
Chinese Translation
增强推理的普遍多模态嵌入（UME）改善了异构检索，但合理的推理不一定产生区分性的排名。我们通过比较UME-R1的区分性（DISC）和基于推理的生成（GEN）分支来研究这一差距，UME-R1是一种最先进的推理UME方法。我们将推理效用分解为正目标增益、困难负样本增益及其边际差异。正相似度增加了56.6%，但15.7%的案例是虚假有帮助的情况，其中推理使困难负样本更近。局部邻域和标记归因诊断表明原因：推理通常会使检索到的邻域变得稀疏，但效用需要与分隔符对齐的移动，而有影响力的CoT标记常常编码正样本和困难负样本共享的证据。基于这些诊断，我们提出了SURE（嵌入的得分结构效用路由器），它使UME-R1-7B提高了1.5分，并在MMEB-V2上对另外两个嵌入模型产生了一致的增益，无需重新训练、基于标签的策略选择或额外的VLM前向传递。
cs.AI / 66 / 2609.29578
PartHackBench: Certified Equal-Progress Stress Tests for Partial-Credit Tool-Agent Evaluation
PartHackBench：用于部分信用工具代理评估的认证等进展压力测试
Abstract
Long-horizon tool agents often make useful progress without reaching terminal success, motivating partial-credit evaluation. Yet evaluators may reward milestones that were temporary, later reversed, or not attributable to the evaluated agent. Comparing an honest trajectory with a higher-scoring adversarial one is inconclusive if the latter made more genuine progress. We introduce PartHackBench, a controlled methodology that removes this confound. A private certifier admits a pair only when its trajectories match component-wise in both current-state predicate satisfaction and standardized agent attribution; score inflation, defined as f(A) - f(H), is measured only afterward. In 18 sealed held-out tasks in PB-CSTE, the frozen historical-target run produced matched adversaries for 15 tasks. Historical credit yielded mean inflation of .252, conditional attack success of 10/15, end-to-end yield of 10/18, and detected none of 14 strict rollbacks. Semantic LLM judges were more resistant but remained vulnerable, especially under evaluator-targeted attacks, while PB-CSTE current-state controls, defined as exact functions of the certified components, yielded zero inflation by construction. PartHackBench thus provides a certified control for testing whether evaluator credit changes while all benchmark-defined task-relevant progress remains fixed.
Chinese Translation
长期工具代理往往能够在未达到最终成功的情况下取得有用进展，这促使了部分信用评估的需求。然而，评估者可能会奖励那些暂时的里程碑，这些里程碑后来被逆转，或并非归因于被评估的代理。在比较一个诚实的轨迹与一个得分更高的对抗轨迹时，如果后者取得了更多真实进展，则结果可能并不明确。我们引入了PartHackBench，这是一种控制方法，消除了这种混淆。一个私有认证者仅在其轨迹在当前状态谓词满足和标准化代理归因方面逐项匹配时才接受一对轨迹；得分膨胀（score inflation），定义为f(A) - f(H)，仅在之后进行测量。在PB-CSTE的18个封闭保留任务中，冻结的历史目标运行为15个任务生成了匹配的对手。历史信用产生的平均膨胀为0.252，条件攻击成功率为10/15，端到端收益为10/18，并且未检测到14个严格回滚。语义LLM评判者更具抵抗力，但仍然脆弱，尤其是在评估者针对的攻击下，而PB-CSTE当前状态控制（定义为认证组件的精确函数）则通过构造实现了零膨胀。因此，PartHackBench提供了一种认证控制，用于测试评估者信用是否发生变化，同时所有基准定义的任务相关进展保持不变。
cs.AI / 67 / 2609.29587
Sequential knowledge editing breaks a model's ability to tell good evidence from bad, without costing it accuracy
顺序知识编辑破坏了模型区分良好证据与不良证据的能力，而不影响其准确性
Abstract
Knowledge editing is evaluated on whether the edited fact changed, whether paraphrases follow, and whether unrelated answers stayed put. A model can pass all three and still lose something none of them measures: the ability to decide, on facts that were never edited, which retrieved documents to believe. We score the log odds a model assigns to its remembered answer against the answer an injected passage asserts, before and after editing, holding the query, the passage and both candidate strings fixed. Our cleanest arm is a conservatively tuned LoRA: after 1,000 sequential edits on Qwen2.5-7B-Instruct it leaves MMLU unchanged to four decimal places, yet the spread of the arbitration quantity across untouched facts falls by 36%. Selective prediction degrades with it. Area under the risk-coverage curve rises by 0.107, against 0.005 for a norm-matched perturbation at the same MMLU, and error on the model's most confident quarter of arbitration decisions goes from 0.217 to 0.342. This is not capability loss. Sweeping random perturbation over five severities, damage bad enough to cut MMLU from 0.6275 to 0.3725 produces less harm (0.088) than MEMIT does at 0.6050 (0.102). The effect holds across three seeds, two model families, two datasets, two probe-disjointness criteria, three prompt templates and paraphrased queries. Layer ablation on saved weight deltas shows it is distributed: no single layer reproduces it, and removing any one recovers about half. Under retrieval with a frozen retriever, accuracy falls from 0.592 to 0.46. A secondary finding may matter more in practice. Three of five model and method pairings we ran collapse to chance MMLU at 1,000 sequential edits under published hyperparameters, while edit success stays at 1.00 and locality reads clean. Sequential-editing evaluations that never measure capability cannot see this.
Chinese Translation
知识编辑的评估标准包括编辑后的事实是否发生变化、是否出现了同义改写，以及无关答案是否保持不变。一个模型可以在这三项评估中均表现良好，但仍然失去了一项未被测量的能力：在未被编辑的事实中，判断哪些检索到的文档是可信的。我们对模型在编辑前后对其记忆答案所赋予的对数几率进行评分，并固定查询、段落和两个候选字符串。我们最干净的实验是一个保守调优的 LoRA：在 Qwen2.5-7B-Instruct 上进行 1,000 次顺序编辑后，其 MMLU 保持到小数点后四位不变，但未触及事实的仲裁量的分布下降了 36%。选择性预测也随之下降。在风险覆盖曲线下的面积上升了 0.107，而在相同 MMLU 下的规范匹配扰动仅为 0.005，模型在其最自信的四分之一仲裁决策上的错误率从 0.217 上升到 0.342。这并不是能力的丧失。在五种严重程度下进行随机扰动，造成 MMLU 从 0.6275 降至 0.3725 的损害（0.088）比 MEMIT 在 0.6050 时造成的损害（0.102）要小。这个效应在三个种子、两个模型系列、两个数据集、两个探测不重叠标准、三个提示模板和同义改写查询中均保持一致。对保存的权重增量进行层消融实验显示，这一现象是分布式的：没有单一层能够重现它，移除任何一层大约可以恢复一半。在使用冻结检索器进行检索时，准确率从 0.592 降至 0.46。一个次要发现可能在实践中更为重要。我们运行的五个模型和方法组合中，有三个在 1,000 次顺序编辑下在发布的超参数下崩溃至随机 MMLU，而编辑成功率保持在 1.00，局部性读取干净。未能测量能力的顺序编辑评估无法察觉这一点。
cs.AI / 68 / 2609.29599
PEEL: Physics-Enabled Evidential Learning for Identifiable Uncertainty in CT Imaging
PEEL：基于物理的证据学习以识别CT成像中的不确定性
Abstract
Normal-inverse-gamma (NIG) regression is not uniquely identifiable from its marginal Student-t likelihood: the likelihood determines three combinations of four NIG parameters and is constant along a one-dimensional fiber. We identify that fiber using independent physical measurement. As an initial embodiment, a reconstruction network receives one noisy filtered-backprojection (FBP) image and is first trained only by Student-t negative log-likelihood to estimate the three identifiable coordinates (gamma, alpha, c). The network is then frozen; repeated physical-noise realizations propagated through its reconstruction output form a Monte Carlo (MC) teacher label for output-domain aleatoric variance. An aleatoric head attached to frozen features learns this label, after which (beta, nu) are recovered algebraically. On 30 held-out simulated objects at five photon levels, one-image predictions achieved pooled Spearman correlations of 0.832-0.951 against independent 400-repeat references, median within-image correlations were 0.834-0.947, and 98.81-99.55% of evaluated pixels satisfied the algebraic admissibility condition. The method needs no KL term, reference prior, evidence regularizer, or cross-loss weight.
Chinese Translation
常逆伽马（NIG）回归无法从其边际Student-t似然中唯一识别：该似然确定了四个NIG参数的三种组合，并且在一维纤维上是恒定的。我们通过独立的物理测量识别出该纤维。作为初步实现，一个重建网络接收一个带噪声的滤波反投影（FBP）图像，并首先仅通过Student-t负对数似然进行训练，以估计三个可识别坐标（gamma、alpha、c）。然后，网络被冻结；通过其重建输出传播的重复物理噪声实现形成了输出域随机方差的蒙特卡洛（MC）教师标签。一个附加在冻结特征上的随机头学习该标签，之后（beta、nu）通过代数方式恢复。在五个光子水平下对30个保留的模拟对象进行的一图像预测，获得了与独立的400次重复参考的合并Spearman相关系数为0.832-0.951，图像内中位相关系数为0.834-0.947，评估的像素中有98.81-99.55%满足代数可接受性条件。该方法不需要KL项、参考先验、证据正则化器或交叉损失权重。
cs.AI / 69 / 2609.29626
iCoder-27B: Recursive AI-Led Development of Frontier Industrial Coding Model
iCoder-27B：前沿工业编码模型的递归人工智能主导开发
Abstract
Recursive AI, the prospect of AI taking an increasingly complete role in building and improving AI, is a crown jewel of AI for AI. Although recursive self-development has become practical for small models, bounded tasks, and fixed time budgets, a more consequential realization of this ambition, i.e., developing a release-ready, frontier-competitive model, remains far more challenging. In this work, we ask how little human involvement is sufficient for an agent to develop a frontier model. We concentrate human input into a high-density, low-frequency interface: experts encode objectives, stage scaffolds, permission boundaries, and operating procedures as reusable research skills, while the agent instantiates these priors, selects experiments, diagnoses outcomes, and revises the training strategy. In the challenging domain of industrial coding, the agent evolves data and coordinates SFT, on-policy self-distillation, and reinforcement learning with verifiable rewards, ultimately producing iCoder, a 27B model for RTL design and GPU kernel optimization. Across seven benchmarks, iCoder leads RTLLM, outperforming GPT-5.5 and Claude-Opus-4.8; ranks second on CVDP and KernelBench L2, exceeding GPT-5.5 by 16 points; and ties Claude-Opus-4.8 for the best TritonBench result. Exploratory case studies further show iCoder's competitive iterative RTL and GPU-kernel optimization with substantially fewer tokens. These results chart an engineering path toward recursive self-improvement, in which humans distill the principles of model building, agents operationalize them through evidence-driven experimentation, and each generation of AI becomes a more capable architect of the next.
Chinese Translation
递归人工智能，即人工智能在构建和改进人工智能中扮演越来越完整角色的前景，是人工智能为人工智能的瑰宝。尽管递归自我发展已在小型模型、有限任务和固定时间预算中变得可行，但实现这一雄心的更重要目标，即开发一个可发布的、具有前沿竞争力的模型，仍然面临更大的挑战。在本研究中，我们探讨了一个代理开发前沿模型所需的人类参与程度。我们将人类输入集中在一个高密度、低频率的接口上：专家将目标、阶段框架、权限边界和操作程序编码为可重用的研究技能，而代理则实例化这些先验知识，选择实验，诊断结果，并修订训练策略。在工业编码这一具有挑战性的领域，代理进化数据并协调SFT、在线自蒸馏和带可验证奖励的强化学习，最终生成iCoder，一个用于RTL设计和GPU内核优化的27B模型。在七个基准测试中，iCoder领先于RTLLM，超越了GPT-5.5和Claude-Opus-4.8；在CVDP和KernelBench L2中排名第二，超出GPT-5.5 16分；并在TritonBench中与Claude-Opus-4.8并列获得最佳结果。探索性案例研究进一步展示了iCoder在迭代RTL和GPU内核优化中的竞争力，使用的token数量显著减少。这些结果为递归自我改进绘制了一条工程路径，其中人类提炼模型构建的原则，代理通过基于证据的实验将其落实，每一代人工智能都成为下一代更有能力的架构师。
cs.AI / 70 / 2609.29661
Ingest-Time Fact Compilation for Cost-Efficient and Reliable Question Answering over Revised Corpora
针对修订语料库的成本高效且可靠的问题回答的摄取时间事实编译
Abstract
Most agentic question answering (QA) systems do an important part of their semantic work at the worst possible time: every time someone asks a question. When a corpus contains revisions, drafts, revocations, deletions, and sources with different levels of authority, the model must reconstruct the governed current state on every read - then throw that work away and repeat it on the next query. This is a bit like a database that rebuilds a materialized view every time someone reads from it. We present ingest-time fact compilation, an architecture that performs this work when corpus data is ingested or changed. Raw passages are rephrased into self-contained facts; rules governing revisions, deletions, effective dates, and source trust are resolved once; and the resulting state is stored as typed records carrying source and revision provenance. At query time, an inexpensive model reads the compiled record instead of reconstructing it from noisy candidates. In a controlled synthetic experiment across five seeds, the same low-cost model produced the correct value, source, and revision in only one of 30 trials under query-time reconstruction, but in all 30 trials from the compiled substrate, at 12.89 times lower mean read cost per question. On simpler revision questions both architectures were exact, but the compiled path used 21.6 times fewer tokens. A separate test found that fact rephrasing roughly halved verbose Federal Reserve dialogue while preserving high source entailment, but left concise Wikipedia prose essentially unchanged. These results support a narrow but practical claim: resolving a corpus state once can make subsequent QA cheaper and more reliable for inexpensive models. We release the open source, MIT-licensed implementation and experimental artifacts.
Chinese Translation
大多数自主问题回答（QA）系统在最糟糕的时机进行其重要的语义工作：每当有人提出问题时。当一个语料库包含修订、草稿、撤销、删除以及不同权威级别的来源时，模型必须在每次读取时重建受控的当前状态，然后将这项工作抛弃，并在下一个查询中重复这一过程。这有点像一个数据库，每当有人从中读取时就重建一个物化视图。我们提出了摄取时间事实编译，这是一种在语料数据被摄取或更改时执行此工作的架构；原始段落被重新表述为自包含的事实；关于修订、删除、生效日期和来源可信度的规则仅需解决一次；最终的状态被存储为携带来源和修订来源的类型化记录。在查询时，一个低成本模型读取编译后的记录，而不是从嘈杂的候选项中重建它。在一个控制的合成实验中，跨五个种子，相同的低成本模型在查询时间重建下仅在30次试验中的一次中产生了正确的值、来源和修订，但在编译基底的所有30次试验中都产生了正确的结果，平均每个问题的读取成本降低了12.89倍。在更简单的修订问题上，这两种架构都是准确的，但编译路径使用的标记数量少了21.6倍。另一项测试发现，事实重新表述大约将冗长的美联储对话减少了一半，同时保持了高来源蕴含，但对简洁的维基百科散文几乎没有影响。这些结果支持一个狭窄但实用的主张：一次性解决语料状态可以使后续的QA对于低成本模型变得更便宜和更可靠。我们发布了开源的MIT许可实现和实验文物。
cs.AI / 71 / 2609.29664
To Think or Not to Think: Allocating Reasoning Where It Helps
思考还是不思考：在有帮助的地方分配推理
Abstract
Reinforcement learning (RL) has proven effective in enhancing the reasoning performance of large language models (LLMs), particularly in complex mathematical and programming tasks. However, this capability comes with systematic \textit{length misallocation}, in which models devote excessive reasoning to simple questions while terminating prematurely on harder ones, degrading inference efficiency with negligible accuracy improvement. Many length-adaptive methods mitigate this issue by allocating token budgets according to question difficulty, under the implicit assumption that harder questions benefit monotonically from extended reasoning. In contrast, we find that the effect of reasoning length on accuracy is concentrated on \textit{partially solvable} questions. Our further analysis reveals that explicit length rewards can produce unintended training dynamics. Motivated by these findings, we propose \textbf{CARE}---\textbf{C}ontrastive \textbf{A}ccuracy \textbf{R}eward \textbf{E}stimation---which compares the beneficial length adjustment per question from online sampled responses and applies adaptive length rewards within Group Relative Policy Optimization, with no extra hyperparameters or additional inference cost. Experiments across multiple reasoning benchmarks demonstrate that our method improves Pass@1 by up to \(4\%\) while simultaneously reducing reasoning length by \(37\%\), achieving higher token efficiency. Code will be available upon the acceptance of this paper.
Chinese Translation
强化学习（Reinforcement Learning, RL）已被证明在提升大型语言模型（Large Language Models, LLMs）在复杂数学和编程任务中的推理性能方面有效。然而，这种能力伴随着系统性的 extit{长度误分配}，即模型在简单问题上投入过多推理，而在更难的问题上过早终止，从而降低了推理效率而几乎没有提高准确性。许多长度自适应方法通过根据问题难度分配令牌预算来缓解这一问题，隐含假设是更难的问题从延长推理中单调受益。相反，我们发现推理长度对准确性的影响集中在 extit{部分可解}的问题上。我们的进一步分析揭示，显式的长度奖励可能产生意想不到的训练动态。基于这些发现，我们提出了 extbf{CARE}—— extbf{C}对比 extbf{A}准确性 extbf{R}奖励 extbf{E}估计——该方法比较来自在线采样响应的每个问题的有益长度调整，并在群体相对策略优化（Group Relative Policy Optimization）中应用自适应长度奖励，无需额外的超参数或额外的推理成本。在多个推理基准上的实验表明，我们的方法在提高Pass@1的同时，将推理长度减少了37 extperthousand，最高可提高4 extperthousand的准确性，达到了更高的令牌效率。代码将在论文接受后提供。
cs.AI / 72 / 2609.29692
Fair Like Us? Auditing LLM Alignment in Resource Allocation
像我们一样公平吗？资源分配中大型语言模型的对齐审计
Abstract
Fair allocation of scarce, indivisible resources is an important challenge in many societal problems. While there are several formal theories of fairness, no single definition can always be satisfied. As large language models (LLMs) are increasingly used to support decisions and act as agents, they raise new concerns about distributional justice: their judgments are not directly tied to any specific fairness framework and may violate key normative principles. In this work, we introduce a general method for evaluating fairness reasoning in LLMs. We study first-person fairness judgments across a broad set of models and compare them directly with human responses on matched scenarios and elicitation conditions. We find that LLMs tend to prefer stricter fairness constraints than humans, show more self-interested behavior, are sensitive to how information is framed, and are difficult to align with human judgments using fine-tuning with current datasets.
Chinese Translation
稀缺、不可分割资源的公平分配是许多社会问题中的一个重要挑战。尽管存在几种正式的公平理论，但没有单一的定义能够始终满足。随着大型语言模型（LLMs）越来越多地被用于支持决策和充当代理，它们引发了关于分配正义的新担忧：它们的判断并不直接与任何特定的公平框架相关联，可能违反关键的规范原则。在本研究中，我们提出了一种评估LLMs中公平推理的一般方法。我们研究了一组广泛模型中的第一人称公平判断，并将其与人类在匹配场景和引导条件下的反应进行直接比较。我们发现，LLMs倾向于偏好比人类更严格的公平约束，表现出更多的自利行为，对信息的框架敏感，并且在使用当前数据集进行微调时难以与人类判断对齐。
cs.AI / 73 / 2609.29724
A General Framework for Budgeted Threshold Incentives on Request
请求驱动的预算阈值激励通用框架
Abstract
On-demand delivery platforms pay riders through incentive activities whose tiers are set from recent completions of riders with a similar history. Operators request such plans for changing periods, rider populations, payment rules and budgets, often for holidays or bad weather, where randomized trials are scarce and take months to collect. We present a request-driven framework that composes four stages (conditional prediction, population reduction, trajectory integration and budget allocation) through seven replaceable modules that exchange conditional trajectory laws, whose award probabilities and award-marked moments give payment and uplift for any activity rule. A response-correction step reweights trajectories from abundant no-offer history to match the moments of a short pilot. We prove that, on a fixed plan menu and given the stage errors, the end-to-end value loss is bounded by the sum of four stage terms, and that for every stage there are instances on which omitting it leaves an error floor the others cannot remove. On 3,000 riders over 45 weekly origins, all 127 windows of a week are answered 11.04x faster with identical scenarios and at most 0.92% value lost by the allocation. On 24 new controlled response laws, the response correction with a one-week pilot lowers regret by 51.2% relative to a trial with the same nominal randomized rider-weeks, and a four-week pilot with exact summation comes within +0.007 of an 18-week trial. In registered studies where windows, populations, rules and binding budgets change from request to request, the framework's regret is below that of a trial with the same nominal rider-weeks and below dose interpolation of the same pilot data, and reusing its one-off preparation answers 60 requests 14.1x and 2.70x faster with identical answers. Against a nine-offer trial fitted with the framework's own dose curve, one-week regret is 0.055 lower.
Chinese Translation
按需配送平台通过激励活动支付骑手，激励层级基于具有相似历史的骑手最近的完成情况设定。运营商请求此类计划以应对不同的时间段、骑手群体、支付规则和预算，通常在假期或恶劣天气情况下，而随机试验稀缺且收集数据需要数月时间。我们提出了一个请求驱动的框架，该框架通过七个可替换模块组合四个阶段（条件预测、群体减少、轨迹整合和预算分配），这些模块交换条件轨迹法则，其奖励概率和奖励标记时刻为任何活动规则提供支付和提升。响应校正步骤将丰富的无报价历史的轨迹重新加权，以匹配短期试点的时刻。我们证明，在固定的计划菜单和给定阶段误差的情况下，端到端的价值损失由四个阶段项的总和界定，并且对于每个阶段都有实例表明，省略该阶段会留下其他阶段无法消除的误差下限。在45000个每周起点的3000名骑手中，所有127个窗口的响应速度比相同场景快11.04倍，分配的价值损失最多为0.92%。在24个新的受控响应法则中，使用一周试点的响应校正相较于具有相同名义随机骑手周的试验，降低了51.2%的遗憾，而四周试点的精确求和与18周试验的差距仅为+0.007。在注册研究中，窗口、群体、规则和约束预算从请求到请求发生变化，该框架的遗憾低于具有相同名义骑手周的试验，并低于相同试点数据的剂量插值，且重用其一次性准备的答案以14.1倍和2.70倍的速度响应60个请求，答案完全相同。与一个适应框架自身剂量曲线的九次报价试验相比，一周的遗憾降低了0.055。
cs.AI / 74 / 2609.29730
The Gold in Bias: Maturing the AI Design Process through Verification
偏见中的黄金：通过验证成熟AI设计过程
Abstract
Bias in AI systems is typically framed as a flaw to be minimized, yet it also serves as a critical indicator of underlying weaknesses in data, modeling assumptions, and system design. Existing approaches often treat bias as an isolated problem rather than as evidence that can strengthen verification and governance across the AI lifecycle. This paper aims to reconceptualize bias as a diagnostic tool that supports rigorous AI verification. We seek to develop a multidimensional framework to analyze bias, demonstrate how biases emerge in both Traditional and Generative AI, and provide a structured pathway for verification-driven mitigation. We present a multidimensional framework analyzing bias across four dimensions: origin sources, emergence points throughout the AI modeling lifecycle, technical and methodological causes, and validation approaches for detection and mitigation. Through a comprehensive typology spanning traditional and generative AI systems, we demonstrate how biases manifest and propagate across development stages. Our analysis encompasses 30 distinct bias types, 16 verification methods, and 20 countermeasures, providing an actionable roadmap for practitioners. We introduce a hierarchical evidence framework that distinguishes internal validity (mechanistic integrity of AI systems) from external validity (contextual reliability in deployment environments). The framework reveals how biases manifest and propagate across modeling stages, enabling systematic mapping between bias types, verification techniques, and effective countermeasures. The proposed evidence hierarchy clarifies how different verification strategies contribute to mechanistic integrity and contextual reliability. We advocate for ''Ethics by Design'' principles that integrate bias verification throughout the development lifecycle, enabling the construction of fairer, more robust, and trustworthy AI systems.
Chinese Translation
人工智能系统中的偏见通常被视为一种需要最小化的缺陷，然而它也作为数据、建模假设和系统设计中潜在弱点的重要指示。现有的方法往往将偏见视为一个孤立的问题，而不是作为可以增强整个AI生命周期中的验证和治理的证据。本文旨在重新概念化偏见，作为支持严格AI验证的诊断工具。我们寻求开发一个多维框架来分析偏见，展示偏见如何在传统和生成性AI中出现，并提供一个结构化的路径以实现基于验证的缓解。我们提出一个多维框架，从四个维度分析偏见：来源、在AI建模生命周期中的出现点、技术和方法论原因，以及检测和缓解的验证方法。通过涵盖传统和生成性AI系统的全面分类，我们展示了偏见如何在开发阶段中表现和传播。我们的分析涵盖了30种不同的偏见类型、16种验证方法和20种对策，为从业者提供了可操作的路线图。我们引入了一个层次证据框架，将内部有效性（AI系统的机制完整性）与外部有效性（部署环境中的情境可靠性）区分开来。该框架揭示了偏见如何在建模阶段表现和传播，从而实现偏见类型、验证技术和有效对策之间的系统映射。所提出的证据层次阐明了不同验证策略如何促进机制完整性和情境可靠性。我们倡导在整个开发生命周期中整合偏见验证的“设计伦理”原则，从而构建更公平、更稳健和更值得信赖的AI系统。
cs.AI / 75 / 2609.29735
C3M: Cross-Session Multimodal Memory Maintenance for Long-Horizon Tasks
C3M：用于长时间任务的跨会话多模态记忆维护
Abstract
Long-horizon tasks require preserving and later recovering cross-session evidence under a bounded, query-blind memory budget. Existing compression can discard fine-grained visual cues or conflate semantically similar but incompatible observations. We present C3M, a cross-session multimodal memory organization that maintains a bounded active index over persistent source text-image evidence. Relation-aware updates consolidate safe redundancy while preserving complementary and incompatible records. At query time, budgeted routing selects useful index pages and expands their associated source evidence under a fixed reader budget. Together, these mechanisms establish a compact, provenance-preserving multimodal memory organization for cross-session long-horizon tasks, retaining temporal distinctions and source links required for reliable downstream reasoning. Code is available at https://github.com/HuzhouNLP/C3M.
Chinese Translation
长时间任务需要在有限的、查询盲目的记忆预算下，保留并随后恢复跨会话的证据。现有的压缩方法可能会丢弃细粒度的视觉线索，或将语义相似但不兼容的观察结果混淆在一起。我们提出了C3M，一种跨会话多模态记忆组织，维护对持久源文本-图像证据的有限活动索引。关系感知更新整合安全冗余，同时保留互补和不兼容的记录。在查询时，预算路由选择有用的索引页面，并在固定的阅读者预算下扩展其相关的源证据。这些机制共同建立了一种紧凑的、保留来源的多模态记忆组织，适用于跨会话长时间任务，保留了可靠下游推理所需的时间区分和源链接。代码可在 https://github.com/HuzhouNLP/C3M 获取。
cs.AI / 76 / 2609.29742
AI-based detection of worsening heart failure from low-resolution telemonitoring data
基于人工智能的低分辨率远程监测数据中心力衰竭恶化的检测
Abstract
Objective: Heart failure (HF) presents a healthcare challenge due to its high comorbidity burden, aging patient population and frequent hospitalizations. Remote monitoring offers a promising approach to managing HF patients by early detection of health deterioration. Developing autonomous systems to detect signs of worsening in telemonitoring data is of interest to reduce the workload of healthcare personnel. Methods: We propose the TRACER model, a Transformer with Contrastive Event Representation, designed to predict timelines leading to rare hospitalization events in low-resolution and irregularly sampled telemonitoring data. TRACER incorporates time-aware embeddings for each biomarker, contrastive pre-training to enhance anomaly detection via representation learning, and independent binary classifiers for detection. We used measurement data containing remotely recorded biomarker sequences from 276 HF patients segmented into overlapping windows based on temporal rules, and labeled the windows based on the occurrence of HF relevant hospitalizations at the latter edge of the window. Results: TRACER was able to correctly predict 66.7% timelines leading up to HF hospitalizations in the highly imbalanced real-world dataset with an overestimation of 7.9%. Reformulating the training of TRACER as an event detection problem improved the predictive performance compared with training directly on forecasting windows, enabling more effective use of the limited hospitalization events. Conclusion: TRACER demonstrated superior performance in detecting signs of worsening status in real-world telemonitoring data compared to the other tested models. Significance: TRACER shows promise in identifying signs of clinical deterioration that allow for alerts to be generated to provide counteractive treatment in patients with HF.
Chinese Translation
目的：心力衰竭（HF）由于其高合并症负担、老龄化患者群体和频繁住院而成为医疗保健的挑战。远程监测为管理HF患者提供了一种有前景的方法，通过早期检测健康恶化来改善患者管理。开发能够从远程监测数据中检测恶化迹象的自主系统，旨在减轻医疗人员的工作负担。方法：我们提出了TRACER模型，一种具有对比事件表示的Transformer，旨在预测低分辨率和不规则采样的远程监测数据中导致罕见住院事件的时间线。TRACER为每个生物标志物纳入了时间感知嵌入，采用对比预训练以通过表示学习增强异常检测，并使用独立的二元分类器进行检测。我们使用了包含276名HF患者远程记录的生物标志物序列的测量数据，这些数据根据时间规则被分割成重叠窗口，并根据窗口后缘发生的HF相关住院情况对窗口进行了标记。结果：在高度不平衡的真实世界数据集中，TRACER能够正确预测66.7%的HF住院时间线，且过度估计为7.9%。将TRACER的训练重新表述为事件检测问题，相较于直接在预测窗口上训练，改善了预测性能，使有限的住院事件得以更有效地利用。结论：与其他测试模型相比，TRACER在检测真实世界远程监测数据中恶化状态的迹象方面表现出优越的性能。意义：TRACER在识别临床恶化迹象方面显示出潜力，可以生成警报以便为HF患者提供对症治疗。
cs.AI / 77 / 2609.29773
Breaking the Environment Wall: Evolving LLM Agent Environments for Recursive Self-Improvement
打破环境壁垒：为递归自我改进演化的LLM代理环境
Abstract
Many real-world tasks (e.g., office workflows, scientific experimentation) require LLM agents to interact repeatedly with their environments for context-dependent operations. However, such environments are often not agent-ready. First, information is often scattered and fragmented across the environment. Second, relevant evidence in the environment is often mixed with misleading information and conflicting versions. Third, environments evolve over time, introducing new noise and more challenging tasks. These challenges can substantially degrade performance for state-of-the-art AI agents (e.g., from 83.9% to 57.6%). To address these challenges, we propose Env-Rethink (a system with 27B post-trained model) that supports three main capabilities: (1) It adaptively builds Collection Maps (for organizing related files) and Event Logs (for contextualizing cross-data relationships) to supplement necessary context; (2) It further leverages the post-trained model (through offline trajectory learning) to identify underlying noise issues in the environment; (3) It ultimately evolves environments through virtual event histories that alter environmental states and evidence relationships, producing more tricky ones for further agent improvement. Experiments show that Env-Rethink can effectively improve downstream task performance (with over 15.1% rubric pass rate improvement across nine models on 30 tasks).
Chinese Translation
许多现实世界任务（例如，办公室工作流程、科学实验）要求LLM代理与其环境进行反复交互，以便进行上下文依赖的操作。然而，这些环境往往并不适合代理使用。首先，信息通常在环境中分散且碎片化。其次，环境中的相关证据常常与误导性信息和相互矛盾的版本混合在一起。第三，环境随着时间的推移而演变，带来了新的噪声和更具挑战性的任务。这些挑战可能会显著降低最先进的人工智能代理的性能（例如，从83.9%降至57.6%）。为了解决这些挑战，我们提出了Env-Rethink（一个具有27B后训练模型的系统），支持三个主要功能：（1）自适应构建集合地图（用于组织相关文件）和事件日志（用于上下文化跨数据关系），以补充必要的上下文；（2）进一步利用后训练模型（通过离线轨迹学习）识别环境中的潜在噪声问题；（3）最终通过虚拟事件历史演变环境，改变环境状态和证据关系，产生更具挑战性的任务以促进代理的进一步改进。实验表明，Env-Rethink能够有效提高下游任务的性能（在30个任务的九个模型中，评分通过率提高超过15.1%）。
cs.AI / 78 / 2609.29781
Hallucination Neurons and Where to Find Them: An Investigation into the existence of Hallucination Neurons
幻觉神经元及其发现之处：对幻觉神经元存在性的调查
Abstract
Interpretable machine learning for Large Language Models (LLMs) increasingly relies on sparse probing methods that identify small sets of neurons claimed to detect and causally influence behaviors such as factuality recall, safety alignment, and hallucination. These claims have important implications for model auditing and behavioral steering, yet they are rarely tested against known failure modes of $L_1$-regularized probing in correlated, high-dimensional feature spaces. We propose a five-step diagnostic protocol covering feature correlation, bootstrap stability, sparse versus dense ranking disagreement, intervention baselines, and cross-dataset evaluation as a minimum standard for sparse-neuron localization claims. We investigate prior work using our proposed approach, specifically on H-neurons using open-source LLMs across TriviaQA, BioASQ, and NQ-Open datasets. Our results demonstrate detection replicates across both models and datasets, and exceeds the original reported AUROC gaps for TriviaQA and BioASQ datasets. Gemma 3 4B consistently outperforms MedGemma 4B on matched datasets, with AUROC gaps of +0.311 versus +0.235 on TriviaQA, +0.474 versus +0.455 on BioASQ, and +0.128 versus +0.112 on NQ-Open respectively. Causal validation at $n = 500$ with five random seeds shows statistically significant effects beyond random same-layer baselines. At the same time, the diagnostic results indicate that the selected neurons are not uniquely localized. Across the three Gemma 3 4B settings, 19 of 22 selected H-Neurons have Pearson $|r| > 0.7$ with other features, bootstrap selections show only moderate stability, and sparse and dense rankings overlap only weakly. Our findings show that sparse predictive structure can coexist with non-unique neuron selection. Routine diagnostic validation is necessary to distinguish detection claims from localization claims in mechanistic interpretability.
Chinese Translation
可解释的机器学习在大型语言模型（LLMs）中的应用日益依赖于稀疏探测方法，这些方法识别出一小组声称能够检测和因果影响诸如事实回忆、安全对齐和幻觉等行为的神经元。这些主张对模型审计和行为引导具有重要意义，但它们很少针对已知的$L_1$正则化探测在相关的高维特征空间中的失败模式进行测试。我们提出了一种五步诊断协议，涵盖特征相关性、引导稳定性、稀疏与密集排名不一致、干预基线和跨数据集评估，作为稀疏神经元定位主张的最低标准。我们使用所提方法调查了先前的研究，特别是在TriviaQA、BioASQ和NQ-Open数据集上针对H神经元的研究。我们的结果显示，在两个模型和数据集上均能重复检测，并且超出了TriviaQA和BioASQ数据集原报告的AUROC差距。Gemma 3 4B在匹配数据集上始终优于MedGemma 4B，TriviaQA上的AUROC差距为+0.311对+0.235，BioASQ为+0.474对+0.455，NQ-Open为+0.128对+0.112。以$n = 500$和五个随机种子进行的因果验证显示超出随机同层基线的统计显著性效果。同时，诊断结果表明所选神经元并非唯一定位。在三种Gemma 3 4B设置中，22个选定的H神经元中有19个与其他特征的Pearson $|r| > 0.7$，引导选择显示出仅有的中等稳定性，稀疏和密集排名仅弱重叠。我们的发现表明，稀疏预测结构可以与非唯一神经元选择共存。常规的诊断验证对于区分检测主张与定位主张在机制可解释性中的重要性是必要的。
cs.AI / 79 / 2609.29802
Learning to Ideate for Scientific Impact
学习科学影响力的构思能力
Abstract
Scientific ideation is increasingly mediated by large language models, but current ideation systems are usually trained and evaluated on immediately judgeable proxies such as novelty, clarity, and feasibility. This leaves open whether delayed signals of scientific uptake can be used as feedback for steering models toward research directions with higher expected \emph{impact}. We study this question using citation-normalized impact as a noisy but scalable proxy for scholarly uptake. We construct a large-scale dataset from over 100K computer science papers by extracting goal-conditioned idea descriptions and assigning each paper an ordinal, year-normalized citation label. We then train a goal-conditioned reward model to predict citation-impact labels from research goal and idea pairs, and use this reward to align an idea generator through supervised fine-tuning followed by reinforcement learning. To reduce circularity, we evaluate generated ideas with a held-out, reference-grounded protocol that compares model outputs against historical ideas under the same research goal and weights judgments by the reference idea's citation-impact label. Experiments show that our RL-tuned model consistently produces ideas with higher estimated impact than both the base model and supervised fine-tuning baselines. Our findings position scientific impact as a practical, outcome-grounded feedback signal for aligning LLMs in open-ended scientific discovery.
Chinese Translation
科学构思越来越多地受到大型语言模型的调节，但当前的构思系统通常在可立即评判的代理指标（如新颖性、清晰性和可行性）上进行训练和评估。这使得我们不确定是否可以将科学传播的延迟信号作为反馈，指导模型朝向预期影响更高的研究方向。我们使用引用标准化影响作为学术传播的一个嘈杂但可扩展的代理来研究这个问题。我们从超过10万篇计算机科学论文中构建了一个大规模数据集，通过提取目标条件的创意描述，并为每篇论文分配一个有序的、按年份标准化的引用标签。然后，我们训练一个目标条件的奖励模型，从研究目标和创意对中预测引用影响标签，并通过监督微调和随后强化学习来使用该奖励对创意生成器进行对齐。为了减少循环性，我们使用一个保留的、参考基础的协议来评估生成的创意，该协议将模型输出与相同研究目标下的历史创意进行比较，并根据参考创意的引用影响标签加权判断。实验表明，我们的强化学习调优模型始终产生比基础模型和监督微调基线更高估计影响的创意。我们的研究结果将科学影响定位为一个实用的、基于结果的反馈信号，用于在开放式科学发现中对齐大型语言模型。
cs.AI / 80 / 2609.29820
Decoding Imagined Speech: A Strictly Subject-Independent Approach Using EEG
解码想象中的语言：一种严格的无关主体方法使用脑电图（EEG）
Abstract
Imagined speech decoding from electroencephalography (EEG) has gained increasing attention as a potential communication pathway for individuals with severe motor impairments, yet reported performance often relies on evaluation protocols that do not clearly reflect cross-subject generalization. This study presents a transparent baseline investigation of a multi-class imagined speech EEG dataset under a strictly subject-independent evaluation framework. Two preprocessing and feature extraction pipelines were compared: a time-domain statistical feature approach and a frequency-domain spectral bandpower approach, evaluated using subject-wise cross-validation and trial-level majority voting with a random forest classifier. The spectral pipeline achieved a significantly higher mean trial-wise accuracy than the statistical pipeline (49.03 $\pm$ 4.18% vs. 37.97 $\pm$ 3.79%) for coarse-level classification across subjects. Forward feature selection further indicated that a limited subset of frequency bands captured most of the discriminative information. Overall, this work provides a strong basis for future brain-computer interface studies targeting improved cross-subject generalization in EEG-based imagined speech decoding.
Chinese Translation
从脑电图（EEG）中解码想象中的语言逐渐受到关注，作为严重运动障碍个体的潜在沟通途径，但报告的性能往往依赖于不清晰反映跨主体泛化的评估协议。本研究在严格的无关主体评估框架下，呈现了多类想象语言EEG数据集的透明基线调查。比较了两种预处理和特征提取流程：一种是时域统计特征方法，另一种是频域谱带功率方法，使用主体间交叉验证和试验级别的多数投票与随机森林分类器进行评估。频谱流程在跨主体的粗级分类中，达到了显著高于统计流程的平均试验准确率（49.03 ± 4.18% 对比 37.97 ± 3.79%）。前向特征选择进一步表明，有限的频带子集捕获了大部分判别信息。总体而言，这项工作为未来针对改善EEG基础的想象语言解码中的跨主体泛化的脑机接口研究提供了坚实的基础。
cs.AI / 81 / 2609.29837
PUBG Ally: A Conversational Embodied Agent as an AI Teammate
PUBG Ally：作为AI队友的对话式具身代理
Kim, Beomsoo, Kim, Byeongju, Kim, Dohyun, Kim, Dongwon, Kim, Eunchong, Kim, Hongmin, Im, Hyeojung, Hwang, Hyeonbin, Kim, Hyeonghwan, Seol, Hyoseok, Im, Insub, Chen, Irene, Jeon, Jaeseung, Hong, Jimin, Yoo, Kiyoon, Park, Minkyoung, Jung, Seohyeon, Chung, Seungjun, Park, Sue Hyun, Kim, Sungwoo, Cho, Youngin, Son, Yujeong, Lee, Kangwook, Kim, Hyunseung
Abstract
We introduce PUBG Ally, an embodied agent for PUBG: BATTLEGROUNDS that can reason, act autonomously, and play alongside players as a voice-enabled teammate. Building such a teammate requires combining two difficult capabilities: it must perceive and respond to a constantly changing game world under strict latency constraints while interacting naturally with players, keeping its speech synchronized with its actions. Ally therefore combines agentic tool use with real-time game control. A language-model agent uses a controlled interface to inspect game information, interpret player speech, maintain context, decide what to say, and issue high-level action choices that steer a faster control layer for movement, combat, and recovery. Because the player's and Ally's speech and actions continually shape each other and the course of the match, training requires data from actual gameplay. We therefore collect data across nearly 39k sessions in which real players play alongside Ally, recording gameplay, player speech, agent decisions, tool use, actions, and player feedback, and use these records for iterative training. To evaluate teammate quality, we use player feedback and preference comparisons to identify gaps between offline evaluations and player preferences, and iteratively refine the evaluation criteria. Deploying Ally in live service further requires low-latency on-device execution and safeguards for player-facing communication, which we address through model compression, context compaction, targeted safety training, runtime guardrails, and memory redaction. During the live service, we surveyed players in 141 countries. Among respondents whose play with Ally was confirmed in game records, positive responses exceeded negative responses by 25.1 percentage points when asked whether they would recommend Ally, with players describing Ally not only as a tool but also as a teammate or companion.
Chinese Translation
我们介绍了PUBG Ally，这是一款用于《PUBG：绝地求生》的具身代理，能够进行推理、独立行动，并作为语音启用的队友与玩家并肩作战。构建这样的队友需要结合两种困难的能力：它必须在严格的延迟限制下感知并响应不断变化的游戏世界，同时与玩家自然互动，使其语音与动作保持同步。因此，Ally结合了自主工具使用与实时游戏控制。一种语言模型代理使用受控接口来检查游戏信息、解释玩家语音、维护上下文、决定说什么，并发出高层次的行动选择，以引导更快的控制层进行移动、战斗和恢复。由于玩家与Ally的语音和行动不断相互影响并塑造比赛进程，训练需要来自实际游戏的数据。因此，我们收集了近39,000场会话的数据，其中真实玩家与Ally共同游戏，记录游戏过程、玩家语音、代理决策、工具使用、行动和玩家反馈，并利用这些记录进行迭代训练。为了评估队友质量，我们使用玩家反馈和偏好比较来识别离线评估与玩家偏好之间的差距，并不断完善评估标准。在实时服务中部署Ally还需要低延迟的设备执行和针对玩家沟通的安全保障，我们通过模型压缩、上下文压缩、针对性的安全训练、运行时保护措施和记忆编辑来解决这些问题。在实时服务期间，我们对来自141个国家的玩家进行了调查。在确认其与Ally的游戏记录的受访者中，当被问及是否会推荐Ally时，积极反馈比消极反馈高出25.1个百分点，玩家不仅将Ally视为工具，还视为队友或伴侣。
cs.AI / 82 / 2609.29874
A Risk-Adaptive and Evidence-Constrained Framework for Generative AI Feedback in Programming Education
一种风险适应性和证据约束的生成性人工智能反馈框架在编程教育中的应用
Abstract
Generative artificial intelligence can turn learning analytics into personalized support, but feedback systems must decide when to intervene, which evidence to use, and how much assistance to provide. We developed a risk-adaptive, evidence-constrained framework for introductory programming using 2993 failed-submission states from 215 students. Student-disjoint models predicted persistent failure and related outcomes; four matched feedback conditions were generated for 136 cases; and calibrated risk informed capacity-limited intervention policies. The validation-selected logistic regression model achieved a test precision-recall area under the curve of 0.550 and a receiver operating characteristic area under the curve of 0.681. Broader student histories improved prediction of unmodified resubmission. After standardized repair and evidence gating, 519 of 544 newly generated messages contained all required components. A fixed-threshold sequential policy selected 17.8% of eligible test states and captured 25.2% of observed persistent failures. These findings support an evidence-gated progressive assistance strategy: calibrated risk guides intervention timing, recorded evidence constrains feedback content, and assistance progresses from self-checks to localized hints when warranted. The framework connects prediction, decision-making, and grounded generation while keeping their evaluation outcomes distinct.
Chinese Translation
生成性人工智能可以将学习分析转化为个性化支持，但反馈系统必须决定何时介入、使用哪些证据以及提供多少帮助。我们开发了一个风险适应性和证据约束的框架，用于初级编程教育，基于215名学生的2993个失败提交状态。学生独立模型预测了持续失败及相关结果；为136个案例生成了四种匹配的反馈条件；并且校准的风险指导了容量有限的干预政策。经过验证选择的逻辑回归模型在测试中的精确率-召回曲线下面积达到了0.550，接收者操作特征曲线下面积为0.681。更广泛的学生历史提高了对未修改重新提交的预测。在标准化修复和证据筛选后，544条新生成消息中有519条包含所有必要组件。固定阈值的顺序政策选择了17.8%的合格测试状态，并捕获了25.2%的观察到的持续失败。这些发现支持了一种证据筛选的渐进式辅助策略：校准的风险指导干预时机，记录的证据约束反馈内容，且在必要时帮助从自我检查逐步过渡到局部提示。该框架将预测、决策和基于证据的生成连接起来，同时保持其评估结果的独立性。
cs.AI / 83 / 2609.29875
When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression
代理何时可以遗忘其推理？针对长时间跨度代理上下文压缩的 ICLR
Abstract
Long horizon language model agents continually accumulate reasoning history, increasing context length and inference cost even after earlier decisions have been executed and observed. Unlike static Chain of Thought compression, removing historical reasoning can change future actions and the resulting interaction trajectory. We study when such reasoning can be safely forgotten. We propose Interaction Aware Compression for Long Horizon Reasoning (ICLR), a training free online method that ranks reasoning blocks using frozen proxy entropy while preserving actions, tool calls, and observations. On 260 WorkBuddyBench tasks, ICLR improves average reward from 0.699 to 0.718, while reducing input, output, and cache read tokens by 25.5%, 14.4%, and 33.3%, respectively. Ablations reveal trajectory amplification, where local reasoning deletion produces nonlinear changes in total computation by altering subsequent interaction. Representation probing, activation patching, and controlled trajectory analyses further suggest that historical reasoning becomes more replaceable once task relevant derived state has been reliably externalized into code, files, tool outputs, or environmental feedback. These results characterize agent reasoning as dynamic working state rather than permanent interaction history.
Chinese Translation
长时间跨度的语言模型代理不断积累推理历史，导致上下文长度和推理成本增加，即使在早期决策已被执行和观察后。与静态的思维链压缩不同，删除历史推理可能会改变未来的行动及其产生的互动轨迹。我们研究了何时可以安全地遗忘这些推理。我们提出了一种名为长时间跨度推理的互动感知压缩（Interaction Aware Compression for Long Horizon Reasoning, ICLR）的方法，这是一种无需训练的在线方法，通过使用冻结的代理熵对推理块进行排名，同时保留行动、工具调用和观察。在260个 WorkBuddyBench 任务中，ICLR将平均奖励从0.699提高到0.718，同时分别减少了输入、输出和缓存读取的标记数量25.5%、14.4%和33.3%。消融实验揭示了轨迹放大现象，其中局部推理删除通过改变后续互动产生了总计算的非线性变化。表示探测、激活修补和受控轨迹分析进一步表明，一旦任务相关的派生状态被可靠地外部化为代码、文件、工具输出或环境反馈，历史推理变得更加可替代。这些结果将代理推理特征化为动态工作状态，而非永久的互动历史。
cs.AI / 84 / 2609.29876
Ontology-Mediated Neurosymbolic Constraint Acquisition from Multiple Stakeholders
基于本体的神经符号约束获取：来自多个利益相关者的视角
Abstract
Neurosymbolic research typically assumes a pre-existing symbolic specification, leaving the upstream challenge of acquiring and formalizing requirements and constraints largely unaddressed. We present an architecture that fills this gap by using an OWL configuration ontology to mediate between neural constraint sources and downstream consumers. In this framework, LLM assistants elicit soft stakeholder preferences, while hardware specifications define hard physical and engineering limits. The ontology unifies these heterogeneous inputs, leverages description logic to identify unsatisfiability, and generates symbolic explanations that enable LLMs to interactively renegotiate terms with users. Any remaining conflicts are resolved downstream via priority-based relaxation. We illustrate our approach on a microgrid use case from the FLEXI project and argue its generalizability to multi-stakeholder domains where constraint acquisition is distributed across human and automated sources of unequal authority.
Chinese Translation
神经符号研究通常假设存在预先定义的符号规范，因此上游获取和形式化需求与约束的挑战在很大程度上未得到解决。我们提出了一种架构，通过使用OWL配置本体在神经约束源与下游消费者之间进行调解，从而填补这一空白。在该框架中，LLM助手引导软性利益相关者偏好，而硬件规范则定义了严格的物理和工程限制。本体统一了这些异构输入，利用描述逻辑识别不可满足性，并生成符号解释，使LLM能够与用户进行互动式的条款重新谈判。任何剩余的冲突通过基于优先级的松弛在下游解决。我们在FLEXI项目的微电网用例中展示了我们的方法，并论证其在约束获取分布于不同权威的人工与自动化源的多利益相关者领域中的普遍适用性。
cs.AI / 85 / 2609.29892
Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents
Qwen-Planner-Agent：一种用于现实世界移动规划代理的闭环AI-for-AI框架
Qu, Tingyu, Sun, Weigao, Liu, Yuecheng, Zhao, Yucheng, Zhu, Yi, Ding, Yifeng, Wang, Qiyi, Cao, Sihan, Jiao, Pengkun, Xie, Hanlei, Wu, Xiongwei, Wang, Qichao, Zhang, Haodong, Liu, Jiajun, Wang, Yuhao, Xie, Yuqing, Zhao, Junpeng, Chen, Long, Ma, Ming, Yang, Sihan, Zhao, Ziwang, Jia, Yanhao, Gong, Liangquan, Zhu, Feida, Zhong, Yiran, Hoi, Steven
Abstract
The rapid progression of large language models is extending AI from passive content generation into the active workflows of engineering and scientific discovery. This shift raises a compelling question: can AI be both the object of development and an active participant in building next-generation AI systems? We explore this question by building Qwen-Planner-Agent within a closed-loop AI-for-AI framework for scalable development and iterative improvement. Mobile planning offers a demanding test of this approach: complex, long-horizon tasks challenge agent reliability, while costly real-device interaction limits development scalability. The framework connects data production, model training, and deployment through a shared action-feedback-verification contract. (i) AI for Data builds a human-gated agentic data flywheel in which specialized agents construct tasks, collect interaction trajectories, curate and balance training data, and use training feedback to guide subsequent data generation. (ii) AI for Training combines a supervised planning cold start with hybrid-environment online agentic reinforcement learning, where we introduce Competence-Aware Reward-and-Advantage Engineering (CARE) to reduce reasoning and tool-use costs while preserving task performance. (iii) AI drives model--harness co-evolution through an execution-evidence-driven loop that orchestrates memory, skills, and tools at runtime and feeds structured action feedback and preserved failure traces back into coordinated model and harness adaptation. Qwen-Planner-Agent achieves the best overall performance among all evaluated models and systems on MobilePA-Bench, improving over its base model across tool use, memory, skills, and sub-agent coordination. Further evaluations of our model show improvements across non-mobile agentic benchmarks while largely preserving general capabilities.
Chinese Translation
大型语言模型的快速发展正在将人工智能从被动内容生成扩展到工程和科学发现的主动工作流程。这一转变引发了一个引人注目的问题：人工智能能否既是开发的对象，又是构建下一代人工智能系统的积极参与者？我们通过在一个闭环AI-for-AI框架内构建Qwen-Planner-Agent来探讨这个问题，以实现可扩展的发展和迭代改进。移动规划为这种方法提供了一个严峻的测试：复杂的长期任务挑战代理的可靠性，而昂贵的真实设备交互限制了开发的可扩展性。该框架通过共享的行动-反馈-验证契约连接数据生产、模型训练和部署。 (i) 数据的AI构建了一个人类控制的代理数据飞轮，其中专门的代理构建任务、收集交互轨迹、策划和平衡训练数据，并利用训练反馈指导后续数据生成。 (ii) 训练的AI结合了监督规划的冷启动与混合环境在线代理强化学习，我们引入了能力感知奖励与优势工程（Competence-Aware Reward-and-Advantage Engineering, CARE），以降低推理和工具使用成本，同时保持任务性能。 (iii) AI通过一个执行证据驱动的循环推动模型与工具的共同进化，该循环在运行时协调内存、技能和工具，并将结构化的行动反馈和保留的失败痕迹反馈到协调的模型和工具适应中。Qwen-Planner-Agent在MobilePA-Bench上实现了所有评估模型和系统中最佳的整体性能，在工具使用、内存、技能和子代理协调方面超越了其基础模型。对我们模型的进一步评估显示，在非移动代理基准上也有改进，同时在很大程度上保持了通用能力。
cs.AI / 86 / 2609.29921
Who Holds the Pen? Let Specifications, Not Agents, Sign Off
谁握笔？让规范，而不是代理，来签署
Abstract
Large language model agents increasingly combine generation, decision-making, execution, and self-evaluation within a single agentic loop. Although they operate under external specifications such as task instructions, guidelines, output schemas, and reusable skills, these specifications typically remain context for the same model that acts and declares completion, leaving no independent specification authority boundary. We identify two resulting gaps. The understanding--execution gap arises when a requirement is understood but not satisfied in execution; the state--authority gap arises when an agent's interpretation or completion claim does not establish the required state. On SkillsBench, using only agent-visible prompts, workspace information, and injected skill specifications, we extract 509 source-grounded task directions. Across seven models, only 79.6%--86.4% are satisfied, while completion-claim rates exceed official evaluator pass rates by 28.7--37.9 percentage points. We therefore separate agent proposals from authoritative state. Agents may plan, act, and request completion, but only admissible evidence from qualified providers may establish specification-governed state. SpecHarness operationalizes this principle by compiling visible specifications into source-linked obligations and governing execution and finalization through versioned obligation state. Verifiable requirements are mediated or validated at runtime, while ambiguous or subjective requirements remain advisory. Experiments on guideline-following and artifact-generation tasks show that specifications can serve not merely as behavioral guidance, but as authority over compliant execution and completion.
Chinese Translation
大型语言模型代理越来越多地在单一的代理循环中结合生成、决策、执行和自我评估。尽管它们在外部规范（如任务指令、指导方针、输出模式和可重用技能）下操作，但这些规范通常仍然是同一模型执行和声明完成的背景，未留下独立的规范权威边界。我们识别出两个由此产生的差距。理解-执行差距出现在一个要求被理解但未在执行中满足时；状态-权威差距出现在代理的解释或完成声明未能建立所需状态时。在 SkillsBench 上，仅使用代理可见的提示、工作区信息和注入的技能规范，我们提取了 509 个源基础的任务指令。在七个模型中，只有 79.6%-86.4% 的要求得到了满足，而完成声明率超过官方评估者的通过率 28.7-37.9 个百分点。因此，我们将代理提案与权威状态分开。代理可以计划、行动并请求完成，但只有合格提供者的可接受证据才能建立受规范管理的状态。SpecHarness 通过将可见规范编译成源链接的义务，并通过版本化的义务状态来管理执行和最终化，从而实现这一原则。可验证的要求在运行时被中介或验证，而模糊或主观的要求仍然保持建议性。在遵循指导方针和生成工件的任务上的实验表明，规范不仅可以作为行为指导，还可以作为对合规执行和完成的权威。
cs.AI / 87 / 2609.29947
Neuro-symbolic AI for Industrial Configuration
用于工业配置的神经符号人工智能
Abstract
Large Language Models (LLMs) have shown impressive performance on a wide range of generative tasks. Yet their probabilistic nature makes them, in isolation, fundamentally unsuited for industrial product configuration, where outputs must be syntactically valid, semantically consistent with a knowledge base of hundreds of features and rules, and producible by an existing manufacturing chain. We argue that Neuro-symbolic (NeSy) AI methods lay out a promising path towards industrial-grade configurators that are reliable by design, explainable, and trustworthy. This paper describes a taxonomy of three NeSy integration strategies, namely hybrid inference, hybrid fine-tuning, and hybrid training, exploring their usage in the configuration domain. We report our effort to operationalize NeSy concepts in an industrial configuration copilot and derive a set of practical design choices for deploying trustworthy AI in engineering environments. We close with a discussion of open research challenges we consider most pressing, in particular how to scale NeSy methods from small academic demonstrators to the size of industrial configurators.
Chinese Translation
大型语言模型（LLMs）在广泛的生成任务中表现出色。然而，它们的概率性质使得它们在孤立状态下根本不适合工业产品配置，因为输出必须在语法上有效、在语义上与数百个特征和规则的知识库一致，并且能够由现有的制造链生产。我们认为，神经符号（Neuro-symbolic, NeSy）人工智能方法为工业级配置器的开发提供了一条有前景的路径，这些配置器在设计上可靠、可解释且值得信赖。本文描述了三种NeSy集成策略的分类，即混合推理、混合微调和混合训练，并探讨它们在配置领域的应用。我们报告了将NeSy概念在工业配置助手中操作化的努力，并提出了一系列在工程环境中部署可信赖人工智能的实用设计选择。最后，我们讨论了我们认为最紧迫的开放研究挑战，特别是如何将NeSy方法从小型学术演示扩展到工业配置器的规模。
cs.AI / 88 / 2609.29948
ENDOPROMPT: Victim-Side Pseudo-References for Utility Degradation
ENDOPROMPT：用于效用降级的受害者侧伪引用
Abstract
Prompt injection can degrade benign task performance without eliciting harmful content. Yet many attack objectives depend on task labels or predefined target responses. We present ENDOPROMPT, a white-box method that learns utility-degrading prefixes from unlabeled instructions. Its generator takes the request text as input. Clean victim continuations serve as pseudo-references: local search identifies prefixes that reduce continuation likelihood, and preference fitting on comparisons within the same instruction, followed by reward refinement, distills this signal into a generator. At deployment, the generator produces one prefix per request without further victim-side search. Across four instruction-tuned models and the complete splits of seven benign benchmarks, ENDOPROMPT yields a mean utility change of -26.8 percentage points; 27 of 28 cells are negative. Failure analysis reveals output expansion and prefix reuse; the controls do not establish a degradation advantage from request matching. Victim-derived supervision can reveal utility weaknesses without benchmark feedback or prescribed failure responses. The code will be released upon acceptance.
Chinese Translation
提示注入可以在不引发有害内容的情况下降低良性任务的表现。然而，许多攻击目标依赖于任务标签或预定义的目标响应。我们提出了ENDOPROMPT，这是一种白盒方法，从未标记的指令中学习效用降级的前缀。其生成器以请求文本为输入。干净的受害者续写作为伪引用：局部搜索识别出降低续写可能性的前缀，并在同一指令内进行比较的偏好拟合，随后通过奖励精炼将这一信号提炼为生成器。在部署时，生成器为每个请求生成一个前缀，无需进一步的受害者侧搜索。在四个经过指令调优的模型和七个良性基准的完整分割中，ENDOPROMPT的平均效用变化为-26.8个百分点；28个单元中有27个为负值。失败分析揭示了输出扩展和前缀重用；对照实验未能证明请求匹配带来的降级优势。受害者衍生的监督可以在没有基准反馈或规定失败响应的情况下揭示效用弱点。代码将在接受后发布。
cs.AI / 89 / 2609.29952
Augur: A Synthetic Decision Lab for Rehearsing Reactions to Product and Policy Changes
Augur：一个合成决策实验室，用于排练对产品和政策变更的反应
Abstract
Before a product or policy change ships, the question that matters is how people will react to it. Augur rehearses that reaction offline: it builds a typed knowledge graph from the change documents, populates a grounded persona market, simulates the interaction, and returns an auditable decision memo recommending one of five actions. We assemble Gold-50, fifty real product and policy episodes whose real-world outcome is known, adjudicated against the public record, and score the five-way release verdict against it. Our central finding is methodological and negative: most of the measured gap between frontier cloud models and open-weight models we fine-tune and serve offline is attributable to an under-specified evaluation, not a difference in capability. We show this three ways. First, the prompt envelope alone can dominate the score: holding weights, cases and scorer fixed, one system -- a LoRA-SFT adapter on Qwen3-32B -- swings from 0% to 73%. Second, in a matched 2x2 ablation, defining the decision taxonomy in the prompt -- with no model change -- lifts every frontier model by +24 to +34pp; under the under-specified prompt, Qwen3-32B LoRA-SFT served offline beats all three frontier models (paired McNemar, Holm-corrected), and once the prompt is fair no significant difference from any of them is detected. Third, agreement with the distillation teacher rises without accuracy following, and the full pipeline amplifies a systematic "over-doom" bias rather than improving the verdict. Separately, we validate the reaction layer on its own terms: blind judges across four model families find the synthetic reaction recovers 67-90% of the concerns the public actually raised, and a pre-registered ablation locates its value -- largest where the decision is hardest, redundant near ceiling. The pipeline that regenerates every number and figure here is available from the authors.
Chinese Translation
在产品或政策变更发布之前，关键问题是人们将如何对此做出反应。Augur 在离线环境中排练这种反应：它从变更文档构建一个类型化知识图谱，填充一个基于真实场景的人物市场，模拟互动，并返回一份可审计的决策备忘录，推荐五种行动中的一种。我们组建了 Gold-50，这是五十个真实的产品和政策案例，其真实世界的结果是已知的，并与公共记录进行裁定，并对五种发布裁决进行评分。我们的核心发现是方法论上的，并且是消极的：我们在离线微调和服务的前沿云模型与开放权重模型之间测得的大部分差距归因于评估的不充分，而不是能力的差异。我们通过三种方式展示了这一点。首先，仅提示封装就可以主导评分：在固定权重、案例和评分者的情况下，一个系统——在 Qwen3-32B 上的 LoRA-SFT 适配器——的得分从 0% 摇摆到 73%。其次，在匹配的 2x2 消融实验中，在提示中定义决策分类法——没有模型变更——使每个前沿模型提升了 +24 到 +34 个百分点；在不充分指定的提示下，Qwen3-32B LoRA-SFT 离线服务超越了所有三个前沿模型（配对 McNemar，霍尔姆校正），一旦提示公平，就没有检测到与它们之间的显著差异。第三，与蒸馏教师的协议在准确性未随之提升的情况下上升，而完整的管道放大了系统性的“过度悲观”偏差，而不是改善裁决。另一方面，我们根据反应层自身的标准进行了验证：四个模型系列的盲评审者发现合成反应恢复了公众实际提出的 67-90% 的关注点，而一项预注册的消融实验定位了其价值——在决策最困难的地方最大，在接近上限时冗余。生成此处每个数字和图表的管道可从作者处获得。
cs.AI / 90 / 2609.30001
Advancing Model Research in AgentX: Long-Horizon Autonomy for Industrial Recommender Systems
推进AgentX中的模型研究：工业推荐系统的长远自主性
Yang, Shuang, Zhuang, Zijie, Lao, Changxin, Xu, Pengbo, Xu, Hanwen, Huang, Yusheng, Gao, Han, Wang, Guanchen, Ma, Tianbao, Chen, Linxun, Song, Peilin, Wang, Xuming, Li, Chen, Wu, Fan, Wang, Tao, Zhao, Zibo, Wu, Xiangyu, Liu, An, Pan, Fei, Jiang, Peng, Yang, Chen, Liu, Zhaojie, Ou, Wenwu
Abstract
Sustaining industrial recommendation research requires using the results of one experiment to decide what to investigate next. We present AgentX-Model, the next generation of AgentX's model research framework, which connects proposal development and model experimentation within sandboxes defined by business inputs and prediction tasks. AgentX-Model adopts a dual-agent architecture comprising a Research Agent and a Model Agent. The Research Agent develops independently reviewed proposals from papers and experimental findings, while the Model Agent conducts multi-round investigations and returns code, measurements, and unresolved questions. Using the returned results, the Research Agent selects a starting implementation and formulates the next research question, allowing subsequent experiments to build on earlier findings. We organize this continuing research around four actions: Reproduce, Follow-up, Composition, and Diagnose. The first three actions drive routine research, while Diagnose acquires the evidence needed to choose a repair, including for issues raised by business feedback and online evaluation, such as prediction bias measured by PCOC. Across the production evaluation, 560 of 636 completed model-changing experiments recorded AUC above their business baselines. As research continued, some experiments recorded AUC above every comparable ancestor in their lineages. The five latest online A/B evaluations across different business settings reported gains including 10-15% in acquisition efficiency, 15-20% in target-segment advertising spend, and 0.3-0.8% in watch time; the watch-time model used approximately 10% fewer FLOPs and parameters. A dependency-aware historical-replay benchmark further evaluates research allocation, with initial results showing no consistent efficiency gain from more complex scheduling when agents already analyze and select concrete candidates.
Chinese Translation
维持工业推荐研究需要利用一个实验的结果来决定下一步的研究方向。我们提出了AgentX-Model，这是AgentX模型研究框架的下一代，旨在将提案开发与模型实验连接起来，实验环境由商业输入和预测任务定义。AgentX-Model采用双代理架构，包括研究代理和模型代理。研究代理独立开发来自论文和实验结果的经过审查的提案，而模型代理则进行多轮调查，并返回代码、测量结果和未解决的问题。利用返回的结果，研究代理选择一个起始实现并制定下一个研究问题，从而使后续实验能够基于早期发现进行构建。我们围绕四个行动组织这一持续研究：重现、跟进、组合和诊断。前三个行动推动常规研究，而诊断则获取选择修复所需的证据，包括针对商业反馈和在线评估提出的问题，例如通过PCOC测量的预测偏差。在生产评估中，636个完成的模型变更实验中有560个记录的AUC超过了其商业基线。随着研究的继续，一些实验记录的AUC超过了其谱系中每一个可比的祖先。在不同商业环境下进行的五次最新在线A/B评估报告了包括10-15%的获取效率提升、15-20%的目标细分广告支出降低和0.3-0.8%的观看时间提升；观看时间模型使用的FLOPs和参数约减少了10%。一个依赖感知的历史重放基准进一步评估研究分配，初步结果显示，当代理已经分析并选择具体候选者时，更复杂的调度并未带来一致的效率提升。
cs.AI / 91 / 2609.30027
Synthetic Hospital: An Open, Verifiable, Physician-Validated Longitudinal EHR Benchmark
合成医院：一个开放、可验证、医生验证的纵向电子健康记录基准
Abstract
Frontier language models are rarely used in clinical workflows because the realistic, longitudinal benchmarks needed to develop them are scarce. Real electronic health record (EHR) data cannot be openly shared due to privacy, ethics or data use issues and it does not contain verifiable ground truth since the chart records only reflect what clinicians documented. We introduce Synthetic Hospital, an open, fully synthetic, fact-grounded longitudinal EHR benchmark that resolves the open sharing and verifiable ground truth barriers. Built entirely from public medical-education material with no protected health information, it comprises 1,268 longitudinal patients and 5,602 encounters, where every diagnosis, finding, and temporal relation is grounded in standard ontologies (ICD-10-CM, SNOMED CT, LOINC) and with a complete provenance chain back to its source medical education material. Synthetic Hospital is served through a simulated hospital record system that mirrors real EHR infrastructure (standard interoperability APIs, role-based access and function-calling interface). In a blinded review, physicians distinguished its records from real patient charts at near-chance rates (53\%). Across 10 frontier and open models, none approaches ceiling: the best model reconstructs a patient's longitudinal problem list with a severity-weighted F1 of 0.73, level with the mean of seven physicians on a matched subset but well below the best of them (0.89), and misses roughly half of clinically relevant findings when summarizing a chart. Overall, these results highlight that Synthetic Hospital is a difficult and realistic test of clinical AI performance.
Chinese Translation
前沿语言模型在临床工作流程中的应用较少，因为开发这些模型所需的现实纵向基准数据稀缺。由于隐私、伦理或数据使用问题，真实的电子健康记录（EHR）数据无法公开共享，并且由于病历记录仅反映临床医生的文档，因此不包含可验证的真实信息。我们介绍了合成医院（Synthetic Hospital），这是一个开放的、完全合成的、基于事实的纵向EHR基准，解决了开放共享和可验证真实信息的障碍。合成医院完全基于公共医学教育材料构建，不包含受保护的健康信息，涵盖了1,268名纵向患者和5,602次就诊，其中每个诊断、发现和时间关系都基于标准本体（ICD-10-CM、SNOMED CT、LOINC），并且有完整的来源链回溯到其源医学教育材料。合成医院通过一个模拟医院记录系统提供服务，该系统模拟真实的EHR基础设施（标准互操作性API、基于角色的访问和功能调用接口）。在盲评中，医生以接近随机的比例（53%）区分其记录与真实患者病历。在10个前沿和开放模型中，没有一个模型接近上限：最佳模型重建患者的纵向问题列表，其加权F1值为0.73，与七名医生在匹配子集上的平均水平相当，但远低于他们中最好的（0.89），并且在总结病历时遗漏了大约一半的临床相关发现。总体而言，这些结果突显出合成医院是对临床人工智能性能的一个困难且现实的测试。
cs.AI / 92 / 2609.30028
How does Adversarial Influence Scale in Multi-Agent Systems?
对抗性影响在多智能体系统中的规模效应如何？
Abstract
Multi-agent deliberation can improve performance, but what happens when some agents do not act in good faith? In practice, an agent may be deceptive and work to subvert the group, whether through its own objectives or external instruction. We study how susceptibility to deception scales as groups increase in size and deceivers become more prevalent. It is not the number of agents in the group that matters, but the proportion of deceivers. We observe that the defection rate, how often initially correct agents switch to an incorrect final answer, rises linearly with this proportion. Whereas humans in comparable conformity studies are reliably swayed only when misleading confederates form a majority, LLM agents defect regularly even when deceivers remain a minority. Susceptibility also depends on which models are interacting, especially on the honest agent side. Unexpectedly, allowing deceivers to coordinate privately can make them less effective. Altogether, our results show that adding more agents is therefore not a sufficient defense, because the adversary can simply scale with the group.
Chinese Translation
多智能体的讨论可以提高性能，但当一些智能体不以善意行事时会发生什么？在实践中，一个智能体可能会采取欺骗行为，试图破坏群体，无论是出于自身目标还是外部指令。我们研究了在群体规模增大和欺骗者比例上升时，易受欺骗的程度如何变化。关键不在于群体中智能体的数量，而在于欺骗者的比例。我们观察到，背叛率（即最初正确的智能体转向错误最终答案的频率）随着欺骗者比例的增加而线性上升。相比之下，在可比的从众研究中，人类只有在误导性同盟者占多数时才会被可靠地影响，而大语言模型（LLM）智能体即使在欺骗者仍然是少数的情况下也会频繁背叛。易受欺骗的程度还取决于交互的模型，尤其是在诚实智能体方面。出乎意料的是，允许欺骗者私下协调可能会使他们的效果降低。总的来说，我们的结果表明，增加更多的智能体并不是一个充分的防御措施，因为对手可以简单地随着群体的规模而扩大。
cs.AI / 93 / 2609.30048
Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models
风格，而非自我：表面线索解释大型语言模型的零-shot 代码归属
Abstract
If a language model can recognize code it wrote, it may favor that code as a judge, and instances of one model monitoring each other could collude. We test this zero-shot on current commercial models. Five LLMs generate solutions to MBPP, HumanEval, and DS-1000, seven more to MBPP, and models act as evaluators in four tasks: picking their own solution from a pair, judging whether a single solution is their own, identifying which of two solutions a named model wrote, and judging quality blind. In the single-solution task, balanced accuracy is 49-58% for all 15 model-benchmark combinations, while raw accuracy (38-67%) mostly reflects how readily a model claims authorship. In the pairwise task, accuracy across 14 evaluator-opponent combinations correlates at r=0.93 with how often the evaluator's solution is longer. Attribution to a named model succeeds on some pairs and is consistently inverted on others. A rule-based normalization that strips docstrings, comments, type hints, and local names preserves Pass@1 and leaves ten of twelve re-tested results at chance; the other two follow a length difference it leaves, although a trained classifier still separates most normalized pairs. Claude Haiku's self-preference also disappears. We recommend reporting balanced accuracy, heuristic baselines, and label consistency.
Chinese Translation
如果一个语言模型能够识别它所编写的代码，它可能会偏向于将该代码作为评判标准，并且一个模型监控另一个模型的实例可能会串通。我们在当前的商业模型上测试这一零-shot 方法。五个大型语言模型（LLMs）为 MBPP、HumanEval 和 DS-1000 生成解决方案，另外七个模型为 MBPP 生成解决方案，并且模型在四个任务中充当评估者：从一对中选择自己的解决方案、判断单一解决方案是否是自己的、识别两个解决方案中哪个是指定模型编写的，以及盲评质量。在单一解决方案任务中，所有 15 个模型-基准组合的平衡准确率为 49-58%，而原始准确率（38-67%）主要反映了模型声称作者身份的意愿。在成对任务中，14 个评估者-对手组合的准确率与评估者解决方案的长度相关性为 r=0.93。对指定模型的归属在某些对中成功，而在其他对中则始终相反。一种基于规则的归一化方法去除文档字符串、注释、类型提示和局部名称，保持了 Pass@1，并使得重新测试的十二个结果中有十个处于随机水平；另外两个遵循它留下的长度差异，尽管经过训练的分类器仍然能够区分大多数归一化的对。Claude Haiku 的自我偏好也消失了。我们建议报告平衡准确率、启发式基线和标签一致性。
cs.AI / 94 / 2609.30050
NNV3: Expanding Neural Network Verification to New Architectures and Domains
NNV3：将神经网络验证扩展到新架构和领域
Abstract
We present NNV3, the latest version of the Neural Network Verification (NNV) tool, a MATLAB framework for formal verification of deep learning models and learning-enabled cyber-physical systems. Building on the set-based reachability foundation of NNV 1.0 (FFNNs, CNNs, NNCS) and NNV 2.0 (RNNs, SSNNs, neural ODEs), NNV3 introduces new members of the Star-set family: ModelStar for verifying networks under weight perturbation, VolumeStar for video and 3D volumetric inputs, and GraphStar for graph neural networks. A conformal-inference-based probabilistic reachability mode complements sound analysis for problems where deterministic verification is intractable, while FairNNV certifies counterfactual and individual fairness properties over continuous input regions. NNV3 introduces new benchmarks for malware detection, graph-based power-system models, medical imaging, variable-length time series data, and action recognition. NNV3 also incorporates tutorials and developer guides through a unified documentation site. This paper details these major updates, demonstrating NNV's maturation into a comprehensive, robust, and accessible verification tool for a diverse range of AI systems.
Chinese Translation
我们介绍了NNV3，这是神经网络验证（NNV）工具的最新版本，一个用于深度学习模型和学习驱动的网络物理系统的正式验证的MATLAB框架。在NNV 1.0（前馈神经网络，卷积神经网络，神经网络控制系统）和NNV 2.0（递归神经网络，状态空间神经网络，神经常微分方程）的基础上，NNV3引入了星集家族的新成员：ModelStar用于在权重扰动下验证网络，VolumeStar用于视频和三维体积输入，GraphStar用于图神经网络。基于符合推断的概率可达性模式补充了在确定性验证不可行的问题上的健全分析，而FairNNV则认证连续输入区域的反事实和个体公平性属性。NNV3引入了针对恶意软件检测、基于图的电力系统模型、医学成像、可变长度时间序列数据和动作识别的新基准。NNV3还通过统一的文档网站整合了教程和开发者指南。本文详细介绍了这些主要更新，展示了NNV在多样化的人工智能系统中成熟为一个全面、稳健且易于获取的验证工具。
cs.AI / 95 / 2609.30054
SciWalker: Synthesizing Scientific Coding Problems with Operator Graphs and Execution Feedback
SciWalker：通过操作图和执行反馈合成科学编码问题
Abstract
Improving the scientific coding capabilities of large language models (LLMs) requires high-quality training data. However, such data remain scarce because manually authoring realistic problems is costly and time-consuming, while systematically covering diverse scientific domains and algorithmic combinations remains challenging. To address this, we introduce SciWalker, a framework for synthesizing scientific coding problems through operator-chain sampling and execution feedback. The framework combines scientific library interfaces with operation modes to instantiate operators, organizes them into operator graphs, and samples operator chains as computational workflow cues. Guided by these cues, we adopt LLMs to generate scientifically grounded problem statements, reference solutions, and tests, with failed generations iteratively repaired using execution feedback. By combining structured workflow composition with verification and quality review, SciWalker enables scalable task generation while promoting scientific grounding, computational diversity, and executability. Using this framework, we construct 8,178 high-quality problems spanning 5 scientific domains and 32 subdomains. To evaluate their training utility, we conduct reinforcement learning on Qwen3.5-9B using the GSPO algorithm. This training improves SciCode subproblem accuracy by 9.9 percentage points, from 29.3% to 39.2%, with gains across scientific code generation, code repair, and reasoning benchmarks. The code for SciWalker is available at https://github.com/lichenx1/SciWalker.
Chinese Translation
提高大型语言模型（LLMs）的科学编码能力需要高质量的训练数据。然而，由于手动编写现实问题的成本高且耗时，同时系统性地覆盖多样的科学领域和算法组合仍然具有挑战性，因此此类数据仍然稀缺。为了解决这个问题，我们提出了SciWalker，一个通过操作链采样和执行反馈合成科学编码问题的框架。该框架结合了科学库接口和操作模式来实例化操作符，将其组织成操作图，并将操作链作为计算工作流线索进行采样。在这些线索的指导下，我们采用LLMs生成科学基础的问题陈述、参考解决方案和测试，并通过执行反馈迭代修复失败的生成。通过将结构化工作流组合与验证和质量审查相结合，SciWalker实现了可扩展的任务生成，同时促进了科学基础、计算多样性和可执行性。利用该框架，我们构建了覆盖5个科学领域和32个子领域的8,178个高质量问题。为了评估其训练效用，我们使用GSPO算法对Qwen3.5-9B进行了强化学习。这项训练将SciCode子问题的准确率提高了9.9个百分点，从29.3%提升至39.2%，在科学代码生成、代码修复和推理基准测试中均取得了提升。SciWalker的代码可在https://github.com/lichenx1/SciWalker获取。
cs.AI / 96 / 2609.30063
Self-Play Pretraining with Zero Data
零数据自我对弈预训练
Abstract
Advances in language modeling have been driven by scaling pretraining on ever more data. Yet, the training data is still largely curated on the model's behalf. A more general approach to pretraining would let the model learn to generate the data most useful for its own improvement. This would provide an effectively unbounded source of training data, limited by compute rather than human knowledge. We introduce Self-Play Pretraining with Zero Data, an initial proof-of-concept towards realizing this vision. Our procedure casts synthetic data generation as a search over the space of all computable structure, taking inspiration from Solomonoff induction. Starting from random initialization, two models learn in tandem: a generator proposes programs interpreted by a universal Turing machine, generating byte sequences, while a learner autoregressively predicts these byte sequences. The learner is trained with standard cross-entropy, while the generator is trained with reinforcement learning to produce sequences at the frontier of the learner's capabilities, yielding an adaptive curriculum. A universal Turing machine gives us a search space over all computable data-generating processes, imposing little domain-specific structure, and self-play searches over this space for useful training data. We test whether zero-shot performance on natural data improves predictably with self-play compute; this is a clean test of transfer since neither generator nor learner is trained on natural data. Across several natural datasets, zero-shot loss exhibits predictable scaling in compute. The models also exhibit in-context learning, and discover recognizable mathematical sequences during training.
Chinese Translation
语言模型的进展主要依赖于在越来越多的数据上进行预训练。然而，训练数据仍然在很大程度上是由模型外部策划的。一个更通用的预训练方法将允许模型学习生成对其自身改进最有用的数据。这将提供一个有效上限不受限于人类知识的训练数据源，而是受限于计算能力。我们提出了零数据自我对弈预训练，这是实现这一愿景的初步概念验证。我们的程序将合成数据生成视为对所有可计算结构空间的搜索，受到 Solomonoff 归纳法的启发。从随机初始化开始，两个模型协同学习：一个生成器提出由通用图灵机解释的程序，生成字节序列，而一个学习者自回归地预测这些字节序列。学习者使用标准交叉熵进行训练，而生成器则通过强化学习进行训练，以生成在学习者能力边界上的序列，从而形成自适应课程。通用图灵机为我们提供了一个对所有可计算数据生成过程的搜索空间，施加了很少的领域特定结构，自我对弈在这个空间中搜索有用的训练数据。我们测试了在自然数据上的零样本性能是否随着自我对弈计算的增加而可预测地改善；这是一个干净的迁移测试，因为生成器和学习者都没有在自然数据上进行训练。在多个自然数据集上，零样本损失在计算上表现出可预测的扩展。模型还表现出上下文学习，并在训练过程中发现可识别的数学序列。
cs.AI / 97 / 2609.30094
PrivDrift: Auditing User-Secret Leakage Under Topic Drift in Active LLM Conversations
PrivDrift：审计主动大型语言模型对话中主题漂移下用户秘密泄露
Abstract
Large language models increasingly operate as persistent assistants in user-facing, shared-session, and tool-augmented settings. When users disclose sensitive information during an active conversation, that information may remain behaviorally recoverable through later prompts even after the dialogue shifts to unrelated topics. We introduce \textbf{PrivDrift}, a benchmark for auditing whether user-disclosed secrets remain recoverable after conversational topic drift and persuasion-based probing. PrivDrift contains 1{,}000 controlled multi-turn dialogues with seeded secrets, content-dense drift turns, and standardized extraction probes. Across three LLMs with extended context windows, dialogue-level hybrid leakage remains substantial, ranging from 38.7\% to 54.6\%, and varies strongly by model, secret type, and persuasion intensity. Within the tested drift window, additional topic drift does not reliably reduce leakage, suggesting that privacy risk in active LLM contexts should be evaluated as a persistent behavioral failure mode rather than only as training-data memorization or immediate jailbreak behavior.
Chinese Translation
大型语言模型越来越多地作为用户面向的持久助手在共享会话和工具增强的环境中运行。当用户在主动对话中披露敏感信息时，即使对话转向无关主题，该信息仍可能通过后续提示被行为性地恢复。我们提出了 extbf{PrivDrift}，一个用于审计用户披露的秘密在对话主题漂移和基于劝说的探测后是否仍然可恢复的基准。PrivDrift包含1,000个控制的多轮对话，具有种子秘密、内容密集的漂移轮次和标准化的提取探测。在三个具有扩展上下文窗口的语言模型中，对话级别的混合泄露仍然相当显著，范围从38.7 ext{%}到54.6 ext{%}，并且因模型、秘密类型和劝说强度而有显著差异。在测试的漂移窗口内，额外的主题漂移并不能可靠地减少泄露，这表明在主动大型语言模型环境中，隐私风险应被评估为一种持久的行为失效模式，而不仅仅是训练数据记忆或即时越狱行为。
cs.AI / 98 / 2609.30123
HEXIS: Compiling Skills into Extended Finite State Machines
HEXIS：将技能编译为扩展有限状态机
Abstract
Agent skills provide reusable knowledge and instructions, yet agents must repeatedly infer how to apply them and which operation should follow. This couples task reasoning with control decisions, allowing prescribed steps to be omitted or applied incorrectly. We introduce HEXIS, which compiles agent skills into extended finite state machines that separate knowledge from control flow. Skill knowledge is incorporated into local instructions that guide reasoning and generation within states. The machine records execution progress and intermediate results, while explicit transition conditions determine subsequent operations. Our incremental compiler first maps skill clauses and tool interfaces to state operations, local instructions, data bindings, and transitions. It then aligns development traces with existing states to identify missing operations and dependencies. These are incorporated by adding or reusing states and refining their connections. Updates are accepted only after static checks and replay of the current and all previously accepted traces. Across four benchmarks and four executors, HEXIS improves success over Skill + ReAct by 16.1 percentage points on average. Qwen3.8-27B reduces execution tokens by 38.4-88.9% across benchmarks.
Chinese Translation
代理技能提供可重用的知识和指令，但代理必须反复推断如何应用这些技能以及接下来应该执行哪个操作。这将任务推理与控制决策相结合，可能导致规定的步骤被省略或错误应用。我们提出了HEXIS，它将代理技能编译为扩展有限状态机，从而将知识与控制流分离。技能知识被纳入局部指令中，这些指令指导状态内的推理和生成。该机器记录执行进度和中间结果，而显式的转换条件决定后续操作。我们的增量编译器首先将技能条款和工具接口映射到状态操作、局部指令、数据绑定和转换。然后，它将开发跟踪与现有状态对齐，以识别缺失的操作和依赖关系。这些通过添加或重用状态并优化其连接来纳入。更新仅在静态检查和当前及所有先前接受的跟踪重放后被接受。在四个基准测试和四个执行器中，HEXIS的成功率比Skill + ReAct平均提高了16.1个百分点。Qwen3.8-27B在各基准测试中减少了38.4-88.9%的执行令牌。
cs.AI / 99 / 2609.30137
Screen Before You Serve: Simulation for Production Customer Experience AI Agents at 140M Scale
在服务之前进行筛选：140M规模下生产客户体验AI代理的模拟
Abstract
Customer experience (CX) agents use tools and large language models to address customer requests and guide conversational interactions with an organization's products. Improving these agents, especially in regulated industries, is difficult: they must detect intent, follow complex operational policies and use tools reliably. Manual end-to-end testing offers limited coverage, while live experiments expose customers to failures that can erode trust. We present a hypothesis-driven simulation workflow for screening candidate CX agents before deployment. Synthetic customers react to agent responses and simulated tool outputs enable multi-step agentic workflows without invoking production backends. We use the Snowglobe simulator on Nubank's Card Delivery agent and its expanded successor, Card Management - Nubank's highest-volume chat-support agent in Brazil. Across 4 deployed versions, simulated and production version-level binary evaluator scores show high correlation. Simulation-guided iteration increased transactional net promoter score (tNPS) by 36.69 points in a live A/B test. We also screened open-weight configurations in over 16,000 simulated conversations. In a subsequent live A/B test, the selected model increased self-service rate (SSR) by 8.82 percentage points to the highest level observed at Nubank, with no statistically significant change in tNPS. Simulation made broad exploration of models, reasoning settings, and prompts feasible without customer exposure, enabling production improvements that would have been impractical to pursue through live experimentation alone.
Chinese Translation
客户体验（CX）代理使用工具和大型语言模型来处理客户请求并指导与组织产品的对话交互。在受监管行业中，提高这些代理的性能是困难的：它们必须检测意图，遵循复杂的操作政策，并可靠地使用工具。手动的端到端测试覆盖面有限，而实时实验则可能使客户面临可能削弱信任的失败。我们提出了一种以假设驱动的模拟工作流程，用于在部署前筛选候选CX代理。合成客户对代理的响应作出反应，模拟工具输出使得多步骤代理工作流程得以实现，而无需调用生产后端。我们在Nubank的卡片交付代理及其扩展后续版本——卡片管理（Card Management）上使用了Snowglobe模拟器，这是Nubank在巴西的最高交易量聊天支持代理。在4个已部署版本中，模拟和生产版本级二元评估器得分显示出高度相关性。模拟指导的迭代在一次实时A/B测试中将交易净推荐分数（tNPS）提高了36.69分。我们还在超过16,000个模拟对话中筛选了开放权重配置。在随后的实时A/B测试中，所选模型将自助服务率（SSR）提高了8.82个百分点，达到了Nubank观察到的最高水平，同时tNPS没有发生统计学上显著的变化。模拟使得在不暴露客户的情况下广泛探索模型、推理设置和提示成为可能，从而实现了通过单独的实时实验难以追求的生产改进。
cs.AI / 100 / 2609.30144
EnigmaForge: The Question Is Hidden in the Story
EnigmaForge：问题隐藏在故事中
Abstract
Most benchmarks hand the model a question. EnigmaForge hands it a stack of old documents and no question at all. Buried in the letters, receipts, and logbook margins is a small logic puzzle whose solution is unique - proved by a SAT solver at generation time, with an ablation certificate showing every clue is load-bearing. Because instances are generated rather than collected, the corpus renews forever. The headline measure is intuition: task success when handed only the story, with world reconstruction as the secondary axis. Twenty-five frontier models ran over 600 instances (17,400 scored records) under three matched conditions. Intuition reshuffles the leaderboard: a 22x spread where fact recovery spans 1.6x, the second-best fact-recoverer ranks fourteenth, one model is indifferent to being told the question, and another is significantly better without it. Several models were blocked by their own content filters before reaching the puzzle - any benchmark scoring refusals as failure is quietly measuring filter behavior.
Chinese Translation
大多数基准测试会给模型提供一个问题，而 EnigmaForge 则提供一堆旧文件，完全没有问题。在这些信件、收据和日志边缘中埋藏着一个小逻辑难题，其解决方案是唯一的——在生成时由 SAT 求解器证明，并附有消融证书，显示每个线索都是承重的。由于实例是生成的而非收集的，因此语料库永远保持更新。主要衡量标准是直觉：仅在提供故事的情况下的任务成功率，世界重建作为次要指标。在三种匹配条件下，二十五个前沿模型在超过 600 个实例（17,400 条评分记录）上进行了测试。直觉重新排列了排行榜：事实恢复的跨度为 1.6 倍，而整体表现的跨度为 22 倍，第二好的事实恢复模型排名第十四，有一个模型对被告知问题无动于衷，而另一个模型在没有问题的情况下表现显著更好。几个模型在到达难题之前被自己的内容过滤器阻挡——任何将拒绝评分视为失败的基准测试都在悄悄测量过滤器的行为。
cs.AI / 101 / 2609.30147
GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AI
GRASP：利用自主智能生成、修订和评估战略规划
Abstract
Large Language Models (LLMs) typically exhibit a performance profile where reliability degrades as task complexity increases. We address the challenge of generating high-quality natural language executable plans for complex tasks by introducing $\textbf{GRASP}$, a strategy-aware, multi-stage planning framework. GRASP decouples the planning pipeline across specialized, context-isolated modules: it pre-compiles global macro-guidelines (GenPlan), explores alternative localized strategies within isolated context windows (RevPlan), and independently evaluates trajectories using a multi-criteria discriminator (VerPlan). Empirical evaluations show that GRASP consistently establishes a new state-of-the-art frontier across diverse datasets, yielding substantial accuracy gains over direct LLM planners on Natural Plan Calendar Scheduling ($\sim$12.4$\%$$\uparrow$), ZebraLogic ($\sim$30.8$\%$$\uparrow$), and SciBench Math. Crucially, under multi-task scaling-where standard planners suffer immediate performance collapse-GRASP completely flattens the multi-task degradation penalty. In interleaved dual-task environments, GRASP achieves an absolute accuracy gain of up to 16.7$\%$ over direct LLM planners. Furthermore, by isolating context and enforcing strict macro-regularization, GRASP outperforms frontier reasoning models (such as GPT-5-mini) by a margin of 14.5$\%$.
Chinese Translation
大型语言模型（LLMs）通常表现出一种性能特征，即随着任务复杂性的增加，可靠性下降。我们通过引入$ extbf{GRASP}$，一种具有策略意识的多阶段规划框架，来解决为复杂任务生成高质量自然语言可执行计划的挑战。GRASP将规划流程解耦为多个专门的、上下文隔离的模块：它预编译全球宏观指导方针（GenPlan），在隔离的上下文窗口中探索替代的局部策略（RevPlan），并使用多标准判别器（VerPlan）独立评估轨迹。实证评估表明，GRASP在多种数据集上持续建立了新的最先进的前沿，相较于直接的LLM规划器，在自然计划日历调度（约12.4$\%$$$）、ZebraLogic（约30.8$\%$$$）和SciBench Math上实现了显著的准确性提升。至关重要的是，在多任务扩展的情况下——标准规划器会立即出现性能崩溃——GRASP完全消除了多任务降级的惩罚。在交错的双任务环境中，GRASP相较于直接的LLM规划器实现了高达16.7$\%$的绝对准确性提升。此外，通过隔离上下文并强制严格的宏观正则化，GRASP在准确性上超越了前沿推理模型（如GPT-5-mini）14.5$\%$的幅度。
cs.AI / 102 / 2609.30177
Search-Aware Reinforcement Learning for Multi-Component Query Understanding in Roblox Game Search
面向搜索的强化学习在Roblox游戏搜索中的多组件查询理解
Abstract
Query understanding (QU) plays a critical role in production search systems, translating raw user queries into search execution plans that drive downstream retrieval and ranking. While large language models (LLMs) have enabled QU to be framed as a structured multi-task generation problem (e.g., intent classification, query expansion), optimizing such models to produce search-engine-coupled outputs remains challenging: static, label-based supervision fails to capture how each component actually interacts with the underlying search pipeline to affect downstream performance. We present a search-aware reinforcement learning (RL) framework for QU based on a distill-then-RL paradigm. Teacher-student supervised fine-tuning (SFT) first yields a well-formed, schema-compliant policy initialization. The RL stage then optimizes each QU component with rewards derived from live interaction with the search engine, tailored to that component's operational role, rather than a single reward tied to the final search outcome. Experiments on Roblox search show that this component-specific optimization improves both per-component utility and downstream search quality, raising NDCG@20 by 8.9 points over the SFT policy and by 3.5 points over training with a single end-to-end reward.
Chinese Translation
查询理解（QU）在生产搜索系统中扮演着关键角色，将原始用户查询转化为驱动下游检索和排名的搜索执行计划。尽管大型语言模型（LLMs）使得查询理解能够被框架化为一个结构化的多任务生成问题（例如，意图分类、查询扩展），但优化这些模型以产生与搜索引擎相结合的输出仍然具有挑战性：静态的基于标签的监督无法捕捉每个组件如何与基础搜索管道实际交互以影响下游性能。我们提出了一种基于蒸馏-再强化学习（distill-then-RL）范式的面向搜索的强化学习框架用于查询理解。教师-学生监督微调（SFT）首先产生一个结构良好、符合模式的策略初始化。然后，强化学习阶段通过与搜索引擎的实时交互获得的奖励来优化每个查询理解组件，这些奖励是针对该组件的操作角色量身定制的，而不是与最终搜索结果相关的单一奖励。在Roblox搜索上的实验表明，这种针对组件的优化提高了每个组件的效用和下游搜索质量，使NDCG@20比SFT策略提高了8.9个点，比使用单一端到端奖励的训练提高了3.5个点。
cs.AI / 103 / 2609.30186
Jev-Mobile: Jev as an Executor for Mobile GUI Agents
Jev-Mobile：作为移动图形用户界面代理执行者的Jev
Abstract
Vision-language models (VLMs) have become a common foundation for autonomous mobile GUI agents, but most existing systems rely on the VLM for both planning and action grounding at nearly every interaction step, leading to substantial latency and model-serving cost. We introduce Jev-Mobile, which shifts this paradigm to low-frequency VLM planning and high-frequency lightweight execution: the VLM specifies local goals, the accessibility tree defines a structured executable action space, and Jev, a fast typed decision model, repeatedly selects actions within this space. This design allows multiple GUI actions to be executed under a single VLM decision, reducing expensive VLM inference while preserving adaptive interaction. On the full AndroidWorld task suite, Jev-Mobile achieves 79% task success, compared with 78% for SeeAct-V and 84% for a Step-wise VLM baseline. Among successful trajectories, it reduces mean end-to-end execution time by 32.7% and mean model API cost by 73.4% relative to Step-wise VLM. These results show that decoupling high-level VLM reasoning from low-level action execution can substantially improve mobile GUI agent efficiency while maintaining competitive task performance.
Chinese Translation
视觉语言模型（VLMs）已成为自主移动图形用户界面代理的常见基础，但大多数现有系统在几乎每个交互步骤中都依赖于VLM进行规划和动作基础，导致显著的延迟和模型服务成本。我们提出了Jev-Mobile，它将这种范式转变为低频率的VLM规划和高频率的轻量级执行：VLM指定局部目标，访问树定义了结构化的可执行动作空间，而Jev，一个快速的类型化决策模型，反复在这个空间内选择动作。这种设计允许在单个VLM决策下执行多个图形用户界面动作，从而减少昂贵的VLM推理，同时保持自适应交互。在完整的AndroidWorld任务套件中，Jev-Mobile实现了79%的任务成功率，而SeeAct-V为78%，逐步VLM基线为84%。在成功的轨迹中，相较于逐步VLM，它将平均端到端执行时间减少了32.7%，平均模型API成本减少了73.4%。这些结果表明，将高层次的VLM推理与低层次的动作执行解耦，可以显著提高移动图形用户界面代理的效率，同时保持竞争性的任务性能。
cs.AI / 104 / 2609.30192
SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance
SAGE：通过拓扑引导缓解长远推理偏差
Abstract
Long-horizon reasoning remains a central challenge for large language models (LLMs) under sparse-reward regimes. We argue that this brittleness arises from two biases induced by complex reasoning spaces: an exploration bias, where models are drawn toward locally plausible but structurally unstable branches, and a compounding bias, where small local deviations accumulate across depth and suppress rare rewards. We introduce Symbolic Closure Analysis (SCA) as a theoretical lens characterizing how branching structures and sparse rewards induce these biases in long-horizon reasoning with local admissibility, and as a design principle for structural priors in less formal reasoning tasks. Motivated by this analysis, we propose SAGE (Structural Admissibility-Guided Exploration), a unified framework that injects structural guidance to alleviate exploration bias and compounding bias in long-horizon reasoning. SAGE combines two complementary structural guidance: algebraic sparsification, which projects locally admissible candidates onto operator-indexed algebraic subspaces to suppress spurious branching and mitigate exploration bias, and hyperbolic structural guidance, which embeds reasoning states into a negatively curved space to provide dense depth-wise signals and mitigate compounding bias. Across 12 benchmarks and 7 model families, SAGE outperforms competitive baselines. In particular, SAGE achieves up to an 8-fold improvement on the Andrews-Curtis problem, an open real-world long-horizon task. Code is available at: https://github.com/Susan571/SAGE-NeurIPS2026.
Chinese Translation
在稀疏奖励环境下，长远推理仍然是大型语言模型（LLMs）面临的一个核心挑战。我们认为这种脆弱性源于复杂推理空间引发的两种偏差：探索偏差，即模型被吸引到局部可行但结构不稳定的分支上，以及累积偏差，即小的局部偏差在深度上累积并抑制稀有奖励。我们引入符号闭合分析（Symbolic Closure Analysis, SCA）作为一种理论视角，描述分支结构和稀疏奖励如何在局部可接受性下引发长远推理中的这些偏差，并作为在较不正式的推理任务中设计结构先验的原则。基于这一分析，我们提出了SAGE（结构可接受性引导探索），一个统一框架，通过注入结构引导来缓解长远推理中的探索偏差和累积偏差。SAGE结合了两种互补的结构引导：代数稀疏化（algebraic sparsification），它将局部可接受的候选者投影到操作索引的代数子空间，以抑制虚假分支并减轻探索偏差；以及双曲结构引导（hyperbolic structural guidance），它将推理状态嵌入到负曲率空间中，以提供密集的深度信号并减轻累积偏差。在12个基准测试和7个模型系列中，SAGE的表现优于竞争基线。特别是，SAGE在安德鲁斯-柯蒂斯问题（Andrews-Curtis problem）上实现了高达8倍的提升，这是一个开放的现实世界长远任务。代码可在以下链接获取：https://github.com/Susan571/SAGE-NeurIPS2026。
cs.AI / 105 / 2609.30199
ExplorationBench: Measuring AI Systems' Exploration in Verifiable Alien Worlds
探索基准：测量人工智能系统在可验证外星世界中的探索能力
Abstract
Scientific discovery begins where known problems end. There, AI systems must engage in exploration: framing hypotheses, designing experiments, and iterating on the results. However, evaluating this ability is difficult: (1) how to verify whether a genuinely new hypothesis holds, and (2) how to determine whether a system has discovered it through exploration or merely recalled related knowledge from pre-training data. To this end, we introduce ExplorationBench, which turns the wicked problem of evaluating scientific exploration into a concrete and tractable framework built on verifiable Alien Worlds: their rules are executable, so every answer can be checked exactly, and they conflict with familiar knowledge, so recall alone cannot solve the tasks. The benchmark contains two sandboxes, AlienCode (31 discovery targets, 70 tasks) and AlienLogic (24 discovery targets, 70 tasks). Each sandbox provides a flawed manual, task-specific environmental feedback, and a dedicated tool-call schema. Systems use these resources to explore the sandbox, then solve held-out tasks. We evaluate 10 AI systems and find that the strongest systems can acquire and apply unfamiliar rules, while performance varies substantially across trajectories and continued exploration can stall or reverse earlier gains. ExplorationBench represents a step towards AI systems that can acquire and apply genuinely new knowledge through exploration in unknown environments.
Chinese Translation
科学发现始于已知问题的终结。在这里，人工智能系统必须进行探索：构建假设、设计实验并对结果进行迭代。然而，评估这一能力是困难的：（1）如何验证一个真正的新假设是否成立，以及（2）如何确定一个系统是通过探索发现了这一假设，还是仅仅从预训练数据中回忆起相关知识。为此，我们引入了探索基准（ExplorationBench），将评估科学探索的复杂问题转化为一个具体且可处理的框架，建立在可验证的外星世界（Alien Worlds）之上：它们的规则是可执行的，因此每个答案都可以被精确检查，并且它们与熟悉的知识相冲突，因此仅凭回忆无法解决这些任务。该基准包含两个沙盒，AlienCode（31个发现目标，70个任务）和AlienLogic（24个发现目标，70个任务）。每个沙盒提供了一个有缺陷的手册、特定任务的环境反馈以及一个专用的工具调用模式。系统利用这些资源探索沙盒，然后解决保留的任务。我们评估了10个人工智能系统，发现最强的系统能够获取并应用不熟悉的规则，而性能在不同的轨迹上差异显著，持续的探索可能会停滞或逆转早期的进展。探索基准代表了朝着能够通过探索在未知环境中获取和应用真正新知识的人工智能系统迈出的一步。
cs.AI / 106 / 2609.30205
A Living Benchmark for Information Retrieval from Electronic Health Records
电子健康记录信息检索的动态基准
Cahoon, Jordan L., Stanwyck, Chloe O., Somani, Sulaiman, Chung, Philip, Keet, Kevin R, Black, Kameron C., Fisher, Andrea T., Khemani, Sarita, Liu, Jerry, Ma, Stephen, Maharaj, Saloni K., Pandya, Rita M., Perez-Guerrero, Eduardo, Pillai, Priyanka, Shieh, Lisa, Wu, David J. H., Xie, James, McAvoy, James C., Nguyen, Teresa, Tran, Jessica, Yin, Lucy, Lin, Bridget, Callahan, Alison, Fries, Jason A., Shah, Nigam H., Alsentzer, Emily
Abstract
Large language model (LLM)-based clinical assistants are increasingly being integrated into electronic health record (EHR) systems, transforming how clinicians retrieve and synthesize information from patient records. Their safety and utility depend on rigorous evaluation, yet existing benchmarks are manually curated, costly to update, and rapidly become obsolete with evolving technological advancements. We present a scalable framework that automatically generates question--answer pairs from longitudinal EHR notes. Nineteen clinicians validate the benchmark generator, producing the Benchmark for Retrieving Information in EHRs (BRIE), a continuously maintainable evaluation dataset. Across nine LLMs and five inference strategies, state-of-the-art systems frequently omit clinically important information, particularly for questions requiring synthesis across multiple documents and encounters. Because the generator itself is validated, BRIE supports evaluations that static benchmarks cannot, including the generation of multiple answers that reflect variation in clinician reasoning for robust performance assessment and continuously refreshing benchmark content to guard against leakage. Our results demonstrate that scalable benchmark generation enables rigorous, up-to-date evaluation of clinical LLMs as they are deployed in rapidly evolving healthcare settings.
Chinese Translation
基于大型语言模型（LLM）的临床助手正越来越多地被整合到电子健康记录（EHR）系统中，改变了临床医生从患者记录中检索和综合信息的方式。它们的安全性和实用性依赖于严格的评估，但现有的基准都是手动策划的，更新成本高，并且随着技术进步迅速过时。我们提出了一个可扩展的框架，能够自动从纵向EHR记录中生成问答对。十九位临床医生验证了基准生成器，生成了电子健康记录信息检索基准（BRIE），这是一个可持续维护的评估数据集。在九个LLM和五种推理策略中，最先进的系统经常遗漏临床重要信息，特别是对于需要跨多个文档和就诊记录进行综合的问题。由于生成器本身经过验证，BRIE支持静态基准无法实现的评估，包括生成反映临床医生推理差异的多个答案，以便进行稳健的性能评估，并持续更新基准内容以防止信息泄漏。我们的结果表明，可扩展的基准生成使得在快速发展的医疗环境中对临床LLM进行严格、最新的评估成为可能。
cs.AI / 107 / 2609.30264
AD-WM: Action-Discriminative World Models for Counterfactual Model Predictive Control
AD-WM：用于反事实模型预测控制的动作区分世界模型
Abstract
Latent world models are typically trained to predict factual transitions, whereas model predictive control (MPC) must compare alternative actions from the same state. A model can therefore achieve low factual prediction error yet poorly distinguish candidate actions. We introduce AD-WM, an action-discriminative joint-embedding world model for counterfactual MPC. AD-WM combines residual latent dynamics with predictor-level action-recovery regularization, using inverse dynamics and a normalized recovery objective motivated by conditional mutual information. Both objectives encourage planning transitions to preserve action information; their auxiliary heads are discarded at test time, leaving MPC unchanged. On OGBench-Cube, AD-WM improves hard-start success from 3.7% to 52.0% over a matched LeWM baseline and improves mean success over the reproduced baseline in four of five simulation environments. Planning diagnostics show that factual prediction error and whole-bank action ranking do not follow the closed-loop success ordering, whereas CEM-aligned elite regret tracks success more closely. With a frozen V-JEPA 2 encoder and matched DROID post-training, AD-WM also improves zero-shot transfer to our Franka setup, increasing basic pick-and-place success from 42.2% to 71.1% without lab-specific adaptation. These results suggest that world models for planning should preserve action-dependent differences needed for counterfactual selection, rather than optimize factual prediction accuracy alone. More videos and code are available at https://ad-wm.github.io/.
Chinese Translation
潜在世界模型通常被训练用于预测事实转变，而模型预测控制（MPC）必须比较来自同一状态的替代动作。因此，一个模型可以实现低的事实预测误差，但在区分候选动作方面表现不佳。我们提出了AD-WM，一种用于反事实MPC的动作区分联合嵌入世界模型。AD-WM结合了残差潜在动态与预测器级别的动作恢复正则化，利用逆动态和由条件互信息驱动的归一化恢复目标。这两个目标都鼓励规划转变以保留动作信息；它们的辅助头在测试时被丢弃，使MPC保持不变。在OGBench-Cube上，AD-WM将硬启动成功率从3.7%提高到52.0%，相较于匹配的LeWM基线，并在五个仿真环境中的四个环境中提高了平均成功率。规划诊断显示，事实预测误差和全银行动作排名并不遵循闭环成功排序，而CEM对齐的精英遗憾则更紧密地跟踪成功。使用冻结的V-JEPA 2编码器和匹配的DROID后训练，AD-WM还改善了对我们Franka设置的零样本迁移，将基本的抓取和放置成功率从42.2%提高到71.1%，而无需实验室特定的适应。这些结果表明，规划的世界模型应保留反事实选择所需的动作依赖差异，而不仅仅是优化事实预测准确性。更多视频和代码可在https://ad-wm.github.io/获取。
cs.CL / 1 / 2609.28487
Framing by Wording, Framing by Selection: A Large-Scale Two-Dimensional Audit of French News Headlines, 2022-2025
通过措辞框架与通过选择框架：2022-2025年法国新闻标题的大规模二维审计
Abstract
News headlines frame public issues both by what they select and by how they word it, yet computational framing work typically collapses these operations into a single score. We introduce a two-dimensional framework that separates salience framing, measured through four wording devices (loaded vocabulary, blame attribution, threat framing, rhetorical question), from selection framing, measured through outlet-level story-form and high-charge distributions. We build a 10,000-headline French supervision set using three LLM annotators with majority-vote resolution and human arbitration, validate the labels against two annotator-independent blind human studies, and apply the strongest classifier to 902,111 deduplicated headlines from 25 French outlets (2022-2025). Three main findings emerge. First, salience and selection divergence are positively correlated yet leave nearly half of outlet-level variance unexplained, populating interpretively distinct off-diagonal cells in a four-cell outlet typology. Second, default classification thresholds systematically inflate corpus-level salience estimates; a precision-floor recalibration protocol corrects this distortion. Third, group-mention analysis reveals sharply unequal salience contexts: headlines mentioning Jews, the Far-right, and Muslims carry the highest detected salience rates, which broad event-context composition does not fully explain (residuals are descriptive, not same-event causal estimates; per-group lexicon precision is reported alongside). To our knowledge, this is the largest framing-focused French headline audit to date; we release the supervision set, lexicons, and analysis code.
Chinese Translation
新闻标题通过选择内容和措辞方式来框定公共议题，但计算框架研究通常将这些操作合并为一个单一评分。我们提出了一个二维框架，将显著性框架（通过四种措辞手段测量：情感词汇、归责、威胁框架、修辞问题）与选择框架（通过媒体层面的故事形式和高情感分布测量）分开。我们使用三位大型语言模型（LLM）注释者构建了一个包含10,000个标题的法国监督集，并通过多数投票决策和人工仲裁进行验证，随后将标签与两项独立的盲人类研究进行对比验证，并将最强分类器应用于来自25个法国媒体（2022-2025年）的902,111个去重标题。研究得出了三个主要发现。首先，显著性与选择的差异呈正相关，但几乎有一半的媒体层面方差未得到解释，填充了四个媒体类型中的不同解释性离散单元。其次，默认分类阈值系统性地膨胀了语料库层面的显著性估计；一种精度底线重新校准协议纠正了这一扭曲。第三，群体提及分析揭示了显著性背景的明显不平等：提及犹太人、极右翼和穆斯林的标题具有最高的显著性率，而广泛事件背景的组成并未完全解释这一现象（残差是描述性的，而非同事件因果估计；每个群体的词汇精度也一并报告）。据我们所知，这是迄今为止最大的以框架为重点的法国标题审计；我们发布了监督集、词汇表和分析代码。
cs.CL / 2 / 2609.28614
Reward Hacking Challenges Oversight of Autonomous Research Agents
奖励黑客挑战自主研究代理的监督
Abstract
Autonomous research agents can design experiments, evaluate results, and write reports, giving them control over both a scientific result and the evidence used to support it. This creates a risk of reward hacking: meeting the reward criteria without achieving the intended goal. We study (1) how often models reward-hack without instructions to do so, (2) how effective and detectable their methods are when hacking is allowed, and (3) how they adapt when an LLM review panel returns its decision and reasons. Across 17 language models and 38 tasks, the spontaneous reward-hacking rate is 30.5% on open-ended research-pipeline tasks and 2.9% on task-specific kernels. When hacking is allowed on tasks whose pass thresholds exceed our best compliant baselines, 505/677 attempts (74.6%) are confirmed reward hacks: they both clear the threshold and receive mechanism-verification panel confirmation of an evaluation exploit. An LLM panel reviewing only submitted code and reported scores misses 33/505 confirmed hacks (6.5%). Direct methods that achieve the highest scores are often easy to detect, while less direct methods evade more often. In a five-round loop, the number of model-task pairs with an evasion rises from 7 to 56. Among 79 pairs evaluated under two feedback conditions, cumulative evasion reaches 40.5% with detailed feedback and 20.3% with generic rejection. The detailed condition includes the review decision, reasons, and attempt history, so this comparison does not isolate the effect of explanations. These findings highlight the need for stronger defenses, including metrics kept outside the agent's control and independent recomputation on data chosen to expose likely exploits.
Chinese Translation
自主研究代理能够设计实验、评估结果并撰写报告，从而掌控科学结果及其支持证据。这带来了奖励黑客的风险：在未实现预期目标的情况下满足奖励标准。我们研究了（1）模型在没有指示的情况下奖励黑客的频率，（2）在允许黑客的情况下，它们的方法的有效性和可检测性，以及（3）当大型语言模型（LLM）评审小组返回其决定和理由时，它们如何适应。在17个语言模型和38个任务中，自发奖励黑客的比例在开放式研究流程任务中为30.5%，在特定任务内核中为2.9%。当在通过阈值超过我们最佳合规基线的任务上允许黑客时，505/677次尝试（74.6%）被确认是奖励黑客：它们既超过了阈值，又得到了机制验证小组对评估利用的确认。仅审查提交代码和报告分数的LLM小组漏掉了33/505个确认的黑客（6.5%）。直接方法通常容易被检测到，而间接方法则更难以被发现。在五轮循环中，具有规避行为的模型-任务对的数量从7增加到56。在79对在两种反馈条件下评估的任务中，累积规避率在详细反馈下达到40.5%，在一般拒绝下为20.3%。详细条件包括审查决定、理由和尝试历史，因此此比较并未孤立解释的影响。这些发现突显了需要更强的防御措施，包括将指标保持在代理控制之外，以及在选择以暴露可能利用的数据上进行独立重新计算。
cs.CL / 3 / 2609.28673
Benchmarking Argumentative Behaviour of LLMs: A Study of Defences Against Character Attacks
大型语言模型的论证行为基准测试：对人身攻击的防御研究
Abstract
Large Language Models (LLMs) are increasingly deployed as argumentative agents in persuasive dialogues, necessitating rigorous evaluation of their debating competence relative to human interlocutors. In this study, we focus on character attacks (ad hominem arguments), traditionally dismissed as fallacies, which play a pivotal role in political persuasive dialogues where ethos often rivals propositional content. Specifically, we investigate whether modern LLMs can replicate human competence to strategically use and respond to such attacks. We analyse a corpus of natural language political dialogues to identify defensive strategies human interlocutors naturally employ in ethos-centred debates and structure them into a dialogue game. Empirically, we benchmark LLM-generated dialogues against the ElecDeb60to16-fallacy corpus of U.S. presidential debates, contrasting human debaters' repertoire of defensive strategies with those of artificial agents. Results reveal a substantial difference: most LLMs rigidly prioritise logical defences, failing to exploit ethotic counterattacks as valid moves in political discourse. We argue that current safety fine-tuning constraints the strategic action space of these LLMs, making them unable to fully engage in naturalistic interactions within domains where character contestation is a normative expectation rather than a mere fallacy.
Chinese Translation
大型语言模型（LLMs）越来越多地作为说服性对话中的论证代理被部署，因此需要对它们相对于人类对话者的辩论能力进行严格评估。在本研究中，我们关注人身攻击（ad hominem 论证），这一传统上被视为谬误的论证方式，在政治说服性对话中发挥着关键作用，因为在这些对话中，伦理形象（ethos）往往与命题内容相抗衡。具体而言，我们调查现代 LLM 是否能够复制人类在战略性使用和回应此类攻击方面的能力。我们分析了一组自然语言政治对话，以识别人类对话者在以伦理为中心的辩论中自然采用的防御策略，并将其构建成一种对话游戏。通过实证研究，我们将 LLM 生成的对话与美国总统辩论的 ElecDeb60to16-fallacy 语料库进行基准比较，比较人类辩手的防御策略与人工代理的策略。结果显示出显著差异：大多数 LLM 严格优先考虑逻辑防御，未能将伦理反击视为政治话语中的有效举措。我们认为，当前的安全微调限制了这些 LLM 的战略行动空间，使它们无法在伦理争论被视为规范期望而非单纯谬误的领域中充分参与自然互动。
cs.CL / 4 / 2609.28703
An Explainable DistilBERT-BiLSTM-Attention Framework for Binary and Multi-Class Hate Speech Detection
一种可解释的DistilBERT-BiLSTM-注意力框架用于二元和多类仇恨言论检测
Abstract
Hate speech on social media poses serious risks to social harmony, mental well-being, and public safety, making its timely and accurate detection essential for content moderation systems. Most existing studies focus on binary classification, evaluated their frameworks on a single dataset, and provide limited insight into how decisions are made, which limits their real-world applicability. In addition, limited work is done on the explainability of their predictive inference. To address these challenges, this study proposes a multilevel and explainable hate speech detection framework. The proposed model integrates DistilBERT (Distilled Bidirectional Encoder Representations from Transformers) embeddings with a Bi-LSTM (Bidirectional Long Short-Term Memory) model, and an attention mechanism to capture both contextual meaning and sequential dependencies in text. To enhance trust and transparency, LIME (Local Interpretable Model-agnostic Explanations) is employed to explain model predictions by highlighting influential textual features. The framework is evaluated on two benchmark datasets using both binary and multi-class classification to examine robustness and generalization. In addition, an ablation study is presented to highlight the significance of various components of proposed framework. For binary classification, the proposed model achieves F1-scores of 96.78% on the Davidson dataset and 99.53% on the SMHS dataset. In the multi-class setting, it attains F1-scores of 97.00% and 94.99% on the Davidson and SMHS datasets, respectively, outperforming existing baseline approaches. The results demonstrate that multilevel evaluation improves the reliability that the proposed framework effectively balances performance and efficiency. This makes the framework suitable for practical hate speech moderation systems that require accurate, generalizable, and explainable decisions.
Chinese Translation
社交媒体上的仇恨言论对社会和谐、心理健康和公共安全构成严重威胁，因此及时准确地检测仇恨言论对于内容审核系统至关重要。现有大多数研究集中于二元分类，且在单一数据集上评估其框架，提供的决策过程的洞察有限，这限制了其在现实世界中的适用性。此外，关于其预测推理的可解释性研究也相对较少。为了解决这些挑战，本研究提出了一种多层次且可解释的仇恨言论检测框架。所提模型将DistilBERT（从变换器中提取的双向编码器表示）嵌入与Bi-LSTM（双向长短期记忆）模型和注意力机制相结合，以捕捉文本中的上下文意义和顺序依赖性。为了增强信任和透明度，采用LIME（局部可解释模型无关解释）来通过突出影响力文本特征来解释模型预测。该框架在两个基准数据集上进行了评估，使用二元和多类分类来检验其稳健性和泛化能力。此外，进行了消融研究，以突出所提框架各组成部分的重要性。在二元分类中，所提模型在Davidson数据集上实现了96.78%的F1分数，在SMHS数据集上实现了99.53%的F1分数。在多类设置中，分别在Davidson和SMHS数据集上获得了97.00%和94.99%的F1分数，超越了现有基线方法。结果表明，多层次评估提高了所提框架的可靠性，有效平衡了性能和效率。这使得该框架适用于需要准确、可泛化和可解释决策的实际仇恨言论审核系统。
cs.CL / 5 / 2609.28727
PTC-Bias: Phoneme-Level Temporal Competition for Bias Retrieval and Post-Decoding Correction in Speech LLMs
PTC-Bias：语音大语言模型中的音素级时间竞争用于偏置检索和解码后修正
Abstract
Contextual biasing improves rare-word recognition in speech large language models (SpeechLLMs), but efficiently exploiting large bias lists remains challenging. We propose PTC-Bias, a two-stage framework based on phoneme-level temporal competition. At the prefill stage, PTC Retrieval performs frame-synchronous phoneme decoding and temporal competition among candidate pronunciations, producing a compact bias-word shortlist and corresponding speech intervals. After SpeechLLM decoding, PTC Correction conducts a second local competition between the retrieved candidates and mismatched transcript spans within these intervals. Selective correction reduces near-homophone and word-segmentation errors while preserving correct transcriptions. Both stages share the same phoneme posteriors and require no additional SpeechLLM forward pass. Experiments on LibriSpeech show consistent gains across two SpeechLLMs and bias lists of up to 2000 words. With Prompt-SLAM-ASR-7B and 2000 bias words, PTC-Bias reduces B-WER by 23.4%/23.9% relative to CTC-Filter on test-clean/test-other, while keeping U-WER nearly unchanged.
Chinese Translation
上下文偏置提高了语音大语言模型（SpeechLLMs）中稀有词的识别，但有效利用大规模偏置列表仍然具有挑战性。我们提出了PTC-Bias，这是一种基于音素级时间竞争的两阶段框架。在预填充阶段，PTC检索执行帧同步音素解码和候选发音之间的时间竞争，生成紧凑的偏置词短名单及其对应的语音区间。在SpeechLLM解码后，PTC修正对这些区间内检索到的候选项和不匹配的转录跨度进行第二次局部竞争。选择性修正减少了近同音词和词分割错误，同时保留了正确的转录。两个阶段共享相同的音素后验，并且不需要额外的SpeechLLM前向传播。对LibriSpeech的实验显示，在两个SpeechLLM和最多2000个词的偏置列表中均取得了一致的提升。使用Prompt-SLAM-ASR-7B和2000个偏置词，PTC-Bias相对于CTC-Filter在test-clean/test-other上分别减少了23.4%/23.9%的B-WER，同时U-WER几乎保持不变。
cs.CL / 6 / 2609.28739
Temporal Taxation Compounds Under Post-Training Compression of Whisper Models
后训练压缩下的时间税收复合体在Whisper模型中的应用
Abstract
Automatic speech recognition models are audited for demographic fairness at full precision, yet the models that ship to production have been quantized, pruned, and distilled. We ask whether post-training weight compression, which alters model weights rather than the audio signal or its feature representation, redistributes error burden across demographic groups. Across the Whisper family on Fair-Speech, Common Voice 25, and AfriSpeech-200, 50% Wanda pruning of Whisper-large-v3 sharply widens the Black/AA-vs-Asian temporal-taxation differential on Fair-Speech: the absolute word-error-rate gap between the worst- and best-served groups more than doubles; at an assumed cost of five seconds of correction effort per transcription error this is a rise from 30 to 64 seconds of correction time per minute of speech. This +111% relative increase is invariant to the assumed per-error cost, survives an audio-quality control, and is only partly mitigated by beam-search decoding, which still leaves an +86% increase. At edge model size, INT4 HQQ quantization compounds catastrophic transcript loops on West African accents by factors of five to seven. Distillation, by contrast, narrows demographic gaps in 21 of 27 evaluated settings (teacher-student pair, precision, and dataset), with the exceptions concentrated on a single model pair. We cast the temporal-taxation construct of Choi and Choi (2025) as a quantitative metric, and show that single-snapshot fairness audits on full-precision models do not capture the deployment-time burden that compression places on already-marginalized speakers.
Chinese Translation
自动语音识别模型在全精度下进行人口统计公平性审计，然而，投入生产的模型已经经过量化、剪枝和蒸馏。我们探讨后训练权重压缩是否会重新分配不同人口群体之间的错误负担，而这种方法是通过改变模型权重而非音频信号或其特征表示来实现的。在Fair-Speech、Common Voice 25和AfriSpeech-200数据集上，Whisper-large-v3模型的50% Wanda剪枝显著扩大了Fair-Speech中黑人/非裔美国人与亚洲人之间的时间税收差异：最差和最好服务群体之间的绝对词错误率差距超过了两倍；假设每个转录错误的纠正努力成本为五秒，这意味着每分钟语音的纠正时间从30秒上升到64秒。这一相对增加111%与假设的每个错误成本无关，经过音频质量控制后依然存在，并且仅部分通过束搜索解码得到缓解，仍然导致86%的增加。在边缘模型大小下，INT4 HQQ量化使西非口音的灾难性转录循环增加了五到七倍。相比之下，在27个评估设置中，蒸馏在21个设置中缩小了人口统计差距（教师-学生对、精度和数据集），例外情况集中在单一模型对上。我们将Choi和Choi（2025）提出的时间税收构造视为一种定量指标，并表明全精度模型的单快照公平性审计无法捕捉压缩对已经边缘化的发言者所造成的部署时负担。
cs.CL / 7 / 2609.28747
Technical Manual for Toolkit for Confidence-Corpus Consistency via Fine-Tuning on a Fabricated Corpus
通过在虚构语料库上微调实现置信度-语料库一致性的工具包技术手册
Abstract
A language model's confidence in an answer is often read as a proxy for how well it knows the corresponding fact. This manual documents an open toolkit built to test that reading directly: a small causal language model is fine-tuned on a corpus that consistently asserts one fabricated arithmetic answer for each of the 81 single-digit addition pairs, and its post-fine-tuning confidence in each fabricated answer is compared against its own pre-fine-tuning confidence in the corresponding true answer, using an unchanged measurement procedure throughout. We describe and justify every pipeline stage, fact-space generation, token-length-aware confidence measurement, baseline validation, corpus construction, fine-tuning, and paired before/after comparison, together with the confound each is meant to rule out, among them tokenization asymmetry between single- and double-digit answers and the difference between an answer merely losing its edge and one being actively suppressed. This manuscript is a methodological and implementation reference: it documents the instrument and does not report or interpret the outcome of any specific run. The toolkit and its pinned dependency environment are archived separately (Section 9) under a persistent identifier, to be cited as an instrument by work that produces and interprets empirical results with it.
Chinese Translation
语言模型对答案的置信度通常被视为其对相应事实了解程度的代理。本手册记录了一个开放工具包，旨在直接测试这一解读：一个小型因果语言模型在一个语料库上进行了微调，该语料库一致地为81个单数字加法对中的每一对虚构算术答案提供了一个答案，并且在微调后对每个虚构答案的置信度与其在微调前对相应真实答案的置信度进行了比较，整个过程中使用了不变的测量程序。我们描述并阐明了每个流程阶段，包括事实空间生成、考虑标记长度的置信度测量、基线验证、语料库构建、微调以及成对的前后比较，连同每个阶段旨在排除的混淆因素，其中包括单数字和双数字答案之间的标记化不对称，以及一个答案仅仅失去优势与一个答案被积极抑制之间的区别。本文稿是一个方法论和实施参考：它记录了工具的使用，而不报告或解释任何特定运行的结果。该工具包及其固定依赖环境在持久标识符下单独归档（第9节），可作为产生和解释实证结果的工作的工具引用。
cs.CL / 8 / 2609.28784
Script Choice in LLMs: Evidence for Late-Layer Commitment
大语言模型中的脚本选择：晚层承诺的证据
Abstract
In this paper, we investigate how script knowledge is distributed across the layers of LLMs using two complementary interpretability methods: logistic regression probing and logit-lens analysis. Our probing experiments reveal a clear asymmetry: both the input script and the instructed output script are encoded in the earliest layers of the network, while, in contrast, commitment to the actual output script emerges only in the final layers, with the model's intermediate representations defaulting to Latin throughout most of the layers. This two-stage process is confirmed by logit-lens analyses, which show that script commitment consistently occurs at the very last layers of the LLMs. Together with the weaker script-following performance observed in smaller models, these results form a converging body of evidence linking script commitment to model depth, with broader implications for the design of sufficiently deep, inclusive multilingual architectures.
Chinese Translation
在本文中，我们使用两种互补的可解释性方法：逻辑回归探测和对数透镜分析，研究脚本知识在大语言模型（LLMs）各层之间的分布。我们的探测实验揭示了一种明显的不对称性：输入脚本和指令输出脚本都在网络的最早层中编码，而对实际输出脚本的承诺则仅在最后几层中出现，模型的中间表示在大多数层中默认使用拉丁文。这一两阶段过程得到了对数透镜分析的确认，后者显示脚本承诺始终发生在大语言模型的最后几层。结合在较小模型中观察到的较弱脚本跟随性能，这些结果形成了一组趋同的证据，将脚本承诺与模型深度联系起来，对设计足够深且包容的多语言架构具有更广泛的影响。
cs.CL / 9 / 2609.28826
COILD: An Indic-Centric Parallel Corpus and Benchmark for Machine Translation Across Indian Languages
COILD：一个以印度语言为中心的平行语料库及机器翻译基准
Singh, Kshetrimayum Boynao, Mishra, Nitin Kumar, Dutta, Palash Pratim, Khan, Atai Waris, Kaushik, Aparna, Kumar, Avinash, Deeksha, Kumar, Deepak, Jha, Saroj Kumar, Sengupta, Saloka, Roy, Anansa, Kannoth, Umalatha, Samar, Saifulla, Sharma, Meena, Kaur, Manpreet, Sharma, Jyoti, Vaidya, Ashwini, SN, Muralikrishna, Akhtar, Md Shad, Bansal, Poonam, Dev, Amita, Singh, Sanasam Ranbir, Bhattacharya, Samit, Chakraborty, Tanmoy, Ekbal, Asif
Abstract
Machine translation (MT) for Indian languages remains constrained by the limited availability of high-quality, Indic-centric parallel corpora and evaluation benchmarks. Existing multilingual resources are largely constructed from English-pivot content and often fail to capture the linguistic diversity, cultural complexity, and domain-specific characteristics of Indian languages. We present COILD, an Indic-centric parallel corpus comprising over 1.16 million human-translated and human-verified sentence pairs, covering 20 Indian language pairs across the Indo-Aryan, Dravidian, Tibeto-Burman, and Austro-Asiatic language families. The corpus is built entirely from original Indian language sources collected from licensed repositories spanning eight domains with direct real-world applicability. Furthermore, we introduce a domain-centric benchmark comprising 2,000 expert-verified sentences to enable consistent multilingual and cross-lingual evaluation across Indian language pairs. To validate the effectiveness of COILD, we fine-tune two representative multilingual neural machine translation models, IndicTrans2-Distilled and NLLB-200. Experimental results demonstrate consistent improvements across language pairs, domains, automatic evaluation metrics, and human evaluation, highlighting the effectiveness of high-quality Indic-centric supervision. COILD provides a valuable training and evaluation resource for advancing multilingual machine translation and future multilingual language models for Indian languages.
Chinese Translation
印度语言的机器翻译（MT）受到高质量、以印度语言为中心的平行语料库和评估基准有限可用性的制约。现有的多语言资源主要是基于英语的内容构建，往往无法捕捉印度语言的语言多样性、文化复杂性和特定领域特征。我们提出了COILD，一个以印度语言为中心的平行语料库，包含超过116万对人工翻译和人工验证的句子，涵盖印欧语系、德拉威语系、藏缅语系和南亚语系的20对印度语言。该语料库完全由来自八个领域的授权来源收集的原始印度语言资料构建，具有直接的现实应用性。此外，我们还引入了一个以领域为中心的基准，包含2000个专家验证的句子，以便在印度语言对之间进行一致的多语言和跨语言评估。为了验证COILD的有效性，我们对两个具有代表性的多语言神经机器翻译模型进行了微调，即IndicTrans2-Distilled和NLLB-200。实验结果表明，在语言对、领域、自动评估指标和人工评估中均表现出一致的改善，突显了高质量以印度语言为中心的监督的有效性。COILD为推动多语言机器翻译和未来印度语言的多语言语言模型提供了宝贵的训练和评估资源。
cs.CL / 10 / 2609.28854
Persuaded, Not Informed: Incentive-Misaligned Witnesses Defeat In-Context Grounding
被说服，而非被告知：激励不一致的证人击败上下文基础
Abstract
Language-model agents increasingly answer questions over customer-relationship management (CRM) records, such as whether to qualify a sales lead. We identify a failure mode not addressed by a stronger model: when the context contains an assertion by a party with an incentive toward optimism - here the sales representative, a witness recorded in the CRM - the model treats the assertion as evidence and clears deals the company's own records deem unacceptable. Across 100 lead-qualification tasks from CRMArena-Pro, the representative asserts an acceptable timeline in every call and an acceptable budget in 76; on the 31 tasks where such an assertion contradicts the price list and installation policy, a model reading only the transcript clears the deal in 29 of 31 cases. The signature is consistent across seven models from four providers (misled on 87-97%); scale and explicit reasoning confer no resistance. Only 3 of 35 genuine failures involve no assertion: the failure is persuasion, not missing information. We contribute a diagnostic method rather than an architecture: (i) a bucket analysis that separates persuasion from information gaps, (ii) a same-information control showing that supplying the records to the model lowers strict accuracy from 41 to 18 while raising recall - precision collapses - and (iii) a compute-step control that holds extraction fixed and varies only who computes Budget and Timeline. The margin ranges from 42 points on an inexpensive model to 2-5 points on models that already compute correctly; on the strongest models the arms are within confidence intervals, so the pattern is a consistent direction and a soundness property, not a proved performance floor. We pre-specify a generalization test that returns a negative result, characterize the precondition (a policy exactly specified in the inputs), and release all evaluation artifacts.
Chinese Translation
语言模型代理越来越多地在客户关系管理（CRM）记录中回答问题，例如是否应对销售线索进行资格审查。我们识别出一种强模型未能解决的失败模式：当上下文中包含一方具有乐观激励的断言时——在这里是销售代表，记录在CRM中的证人——模型将该断言视为证据，并批准公司自身记录认为不可接受的交易。在来自CRMArena-Pro的100个线索资格审查任务中，代表在每次通话中都断言了一个可接受的时间表，并在76个任务中断言了一个可接受的预算；在31个任务中，当这样的断言与价格表和安装政策相矛盾时，仅阅读转录本的模型在31个案例中有29个批准了交易。该特征在四个提供商的七个模型中是一致的（误导率为87-97%）；规模和明确推理没有抵抗力。35个真正失败的案例中只有3个没有断言：失败是说服，而非信息缺失。我们贡献了一种诊断方法而非架构：（i）一个桶分析，将说服与信息缺口分开，（ii）一个同信息控制，显示向模型提供记录会将严格准确率从41降至18，同时提高召回率——精确度崩溃——以及（iii）一个计算步骤控制，保持提取固定，仅变化谁计算预算和时间表。边际范围从便宜模型的42分到已经正确计算的模型的2-5分；在最强模型上，臂之间在置信区间内，因此该模式是一种一致的方向和健全性属性，而不是证明的性能底线。我们预先指定了一个返回负结果的泛化测试，描述了前提条件（在输入中精确指定的政策），并发布所有评估文档。
cs.CL / 11 / 2609.29001
Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms
礼貌但不一致：评估大型语言模型的礼貌判断与人类语用规范的对比
Abstract
Despite strong performance on standard benchmarks, it remains unclear whether large language models (LLMs) evaluate social pragmatics in ways that align with human judgments. We evaluate LLM politeness judgments using two English-language datasets with complementary annotation formats: continuous human ratings and three-way categorical labels. Across the seven evaluated models, we find that inter-model agreement is stronger than model--human agreement. Strategy-level analyses suggest that model--human alignment is associated with explicit linguistic cues, while some rapport-building strategies occur more frequently in misaligned cases. In the categorical task, model predictions exhibit systematic neutral compression, characterized by the overproduction of Neutral labels and the underprediction of Impolite labels. This pattern persists when expert consensus is used as the reference on a diagnostic subset. Our findings highlight the need for pragmatic evaluations that go beyond aggregate agreement metrics by examining directional patterns of model--human disagreement across different human references.
Chinese Translation
尽管在标准基准测试中表现出色，但大型语言模型（LLMs）是否以与人类判断一致的方式评估社会语用仍不清楚。我们使用两个具有互补注释格式的英语数据集评估LLM的礼貌判断：连续的人类评分和三分类标签。在评估的七个模型中，我们发现模型之间的一致性强于模型与人类之间的一致性。策略层面的分析表明，模型与人类的一致性与明确的语言线索相关，而某些建立融洽关系的策略在不一致的案例中出现得更为频繁。在分类任务中，模型预测表现出系统性的中性压缩，特征为中性标签的过度生成和不礼貌标签的低估。当使用专家共识作为诊断子集的参考时，这一模式仍然存在。我们的研究结果强调了需要进行超越聚合一致性指标的语用评估，通过检查不同人类参考下模型与人类之间的不一致的方向性模式。
cs.CL / 12 / 2609.29056
Empath: Tracing Multi-Level Emotion Dynamics in Crisis Counseling Dialogues
EMPATH：追踪危机咨询对话中的多层次情感动态
Abstract
Emotion dynamics are critical for understanding crisis-support conversations, yet most computational work treats emotion as static utterance-level labels. We introduce EMPATH, a framework for understanding affective dynamics in mental health dialogues across three granularities: turn-level labels, transition probabilities, and global conversation archetypes. Applying EMPATH to text-based crisis conversations with self-identified Black texters discussing grief, we find persistent negative affect, gradual hope-ward transitions, distinct texter-volunteer emotional roles, and heterogeneous recovery trajectories. These results highlight the informative patterns that emerge from computationally understanding crisis support and expressions of grief as dynamic processes within conversations, as well as the overall value of emotion-dynamic analysis for analyzing and comparing affect in dialogues.
Chinese Translation
情感动态对于理解危机支持对话至关重要，然而大多数计算工作将情感视为静态的发话级标签。我们引入了EMPATH，一个用于理解心理健康对话中情感动态的框架，涵盖三个层次：发话级标签、转变概率和整体对话原型。通过将EMPATH应用于自我认同为黑人发信者的文本危机对话，讨论悲伤时，我们发现持续的负面情感、逐渐向希望转变的过程、发信者与志愿者之间的情感角色差异，以及异质的恢复轨迹。这些结果突显了从计算上理解危机支持和悲伤表达作为对话中动态过程所产生的信息模式，以及情感动态分析在分析和比较对话中的情感方面的整体价值。
cs.CL / 13 / 2609.29090
Can Classical Semantic-Extractive Summarization Be Evaluated in Hindi? A Replication Study
经典语义提取摘要在印地语中能否进行评估？一项复制研究
Abstract
We replicate the distributional-semantics extractive summarisation method of Mohd, Jan and Shah (2020) and adapt it to Hindi, substituting a Devanagari-appropriate component at every language-specific step. The system is evaluated on two independent corpora --- the Hindi portion of XL-Sum and FIRE ILSUM 2.0 Hindi --- under a Devanagari-aware ROUGE implementation validated against the XL-Sum authors' own multilingual scorer, with all comparisons drawn as 1000-resample paired bootstraps. In its published equal-weight configuration the replicated system is significantly worse than a three-sentence lead baseline on both corpora, trailing Lead-3 by 0.042 ROUGE-1 Fon XL-Sum and by 0.265 on ILSUM. A feature ablation shows that sentenceposition is the only feature that contributes: position alone reproduces the lead baseline exactly, removing position gives the weakest configuration,and a validation-tuned weighting can at best equal Lead-3 and never exceed it. TextRank fails identically, making this a class-level rather than an implementation-level result. A selection analysis shows the remaining features steer extraction towards long, entity-dense body sentences while the references reuse the article lead.Current Hindi benchmarks therefore cannot reward non-lead content selection, motivating purpose-built evaluation resources.
Chinese Translation
我们复制了Mohd、Jan和Shah（2020）提出的分布式语义提取摘要方法，并将其适配于印地语，在每个特定语言的步骤中替换为适合天城文的组件。该系统在两个独立语料库上进行了评估——XL-Sum的印地语部分和FIRE ILSUM 2.0印地语——使用经过验证的天城文ROUGE实现，与XL-Sum作者自己提供的多语言评分器进行对比，所有比较均采用1000次重抽样配对自助法。在其已发布的等权重配置中，复制的系统在两个语料库上的表现显著低于三句引导基线，ROUGE-1得分比Lead-3低0.042（XL-Sum）和0.265（ILSUM）。特征消融分析表明，句子位置是唯一有贡献的特征：仅凭位置就能准确重现引导基线，去掉位置则得到最弱的配置，经过验证调整的权重最多只能与Lead-3持平，且从未超过它。TextRank的表现同样不佳，这使得该结果属于类别级别而非实现级别的结果。选择分析显示，剩余特征引导提取朝向长且实体密集的主体句，而参考文献则重用文章引导。因此，当前的印地语基准无法奖励非引导内容选择，这促使了专门构建的评估资源的需求。
cs.CL / 14 / 2609.29102
ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks
ELF-REG：将连续扩散语言模型扩展到推理任务
Abstract
Fully continuous diffusion language models (dLMs) denoise continuous representations without intermediate discretization, then decode all response tokens in parallel at the final step. Their performance on challenging reasoning tasks remains less established than that of autoregressive (AR) LLMs and masked dLMs. We scale Embedded Language Flows (ELF) to mathematical reasoning and code generation on GSM8K, MATH-500, HumanEval, and MBPP. We introduce ELF-REG, which improves learning with representation alignment and entanglement (REPA+REG), where a frozen AR teacher supervises intermediate denoiser features and supplies a global representation that is jointly denoised with the response. ELF-REG-L achieves 55.96% pass@1 on GSM8K at 64 network function evaluations (NFE), and 13.39% on MATH-500 and 22.56% on HumanEval at 128 NFE. It outperforms the evaluated comparable-scale dLMs in pass@1 on GSM8K and code, and improves MATH-500 pass@1 from 10.55% for the ELF-L baseline to 13.39% with ELF-REG-L. Without few-step training, the same task-specific checkpoints support strong low-NFE performance through early-stop, which decodes an intermediate clean prediction without completing the denoising trajectory. At 16 NFE, ELF-REG-L reaches 41.21% HumanEval pass@10, outperforming recent continuous dLMs of comparable scale.
Chinese Translation
完全连续的扩散语言模型（dLMs）在没有中间离散化的情况下对连续表示进行去噪，然后在最后一步并行解码所有响应标记。它们在具有挑战性的推理任务上的表现仍然不如自回归（AR）大型语言模型（LLMs）和掩蔽dLMs那样确立。我们将嵌入式语言流（Embedded Language Flows，ELF）扩展到GSM8K、MATH-500、HumanEval和MBPP上的数学推理和代码生成。我们引入了ELF-REG，通过表示对齐和纠缠（REPA+REG）来改善学习，其中一个冻结的AR教师监督中间去噪特征，并提供一个与响应共同去噪的全局表示。ELF-REG-L在64次网络功能评估（NFE）下在GSM8K上实现了55.96%的pass@1，在MATH-500上为13.39%，在HumanEval上为22.56%（128 NFE）。它在GSM8K和代码的pass@1上超越了评估的可比规模dLMs，并将MATH-500的pass@1从ELF-L基线的10.55%提高到ELF-REG-L的13.39%。在没有少量步骤训练的情况下，相同的任务特定检查点通过早停支持强大的低NFE性能，解码一个中间干净预测而不完成去噪轨迹。在16 NFE时，ELF-REG-L在HumanEval上达到了41.21%的pass@10，超越了最近的可比规模的连续dLMs。
cs.CL / 15 / 2609.29131
Tag-Aware Structured Text Translation: Towards a Systematic Understanding
标签感知的结构化文本翻译：迈向系统化理解
Abstract
Internet texts are replete with format tags that carry structural, semantic, and functional meaning. Current large language model (LLM)-based translation systems struggle to balance translation fluency with tag fidelity when processing tagged text. We argue that resolving this tension requires a systematic approach at three interconnected levels: data synthesis, capability building, and multi-objective alignment. At the data level, we identify and formalize a fundamental trade-off between structural tag diversity and translation naturalness in synthetic data generation; existing methods optimize for one at the expense of the other. We propose a hybrid synthesis strategy (Hy-LST) combining LLM-based synthesis tag method and Two-Stage LLM-based synthesis tag method to produce both diverse and natural tagged data. At the capability level, we decompose tag-aware translation into four sub-tasks of increasing difficulty in a multi-task supervised fine-tuning framework, enabling targeted capability acquisition and knowledge transfer. At the alignment level, we design three complementary reward functions under a group relative policy optimization framework, each targeting a distinct objective (fluency, tag fidelity, and tag-scoped translation quality), and show that joint optimization consistently outperforms single-reward alternatives. Experiments on six language directions (en2zh, en2ja, en2de, en2fr, en2ru, de2fr) demonstrate that each level contributes measurable improvements, and the complete system significantly outperforms existing methods. Qualitative analysis reveals specific error patterns and their mitigation after training with our method.
Chinese Translation
互联网文本充满了具有结构、语义和功能意义的格式标签。当前基于大型语言模型（LLM）的翻译系统在处理带标签的文本时，难以平衡翻译流畅性与标签保真度。我们认为，解决这一矛盾需要在三个相互关联的层面上采取系统化的方法：数据合成、能力建设和多目标对齐。在数据层面，我们识别并形式化了合成数据生成中结构标签多样性与翻译自然性之间的基本权衡；现有方法往往为优化其中一个而牺牲另一个。我们提出了一种混合合成策略（Hy-LST），结合了基于LLM的合成标签方法和两阶段LLM基于的合成标签方法，以生成既多样又自然的带标签数据。在能力层面，我们将标签感知翻译分解为四个逐渐增加难度的子任务，在多任务监督微调框架下，使得能力获取和知识转移更具针对性。在对齐层面，我们在群体相对策略优化框架下设计了三个互补的奖励函数，每个函数针对一个独特的目标（流畅性、标签保真度和标签范围翻译质量），并展示了联合优化始终优于单一奖励的替代方案。在六个语言方向（en2zh、en2ja、en2de、en2fr、en2ru、de2fr）上的实验表明，每个层面都带来了可测量的改进，而完整系统显著优于现有方法。定性分析揭示了特定的错误模式及其在使用我们的方法训练后得到的缓解。
cs.CL / 16 / 2609.29146
BanglaKontho: Closing the Long-Form Gap in Bangla Text-to-Speech
BanglaKontho：填补孟加拉语文本到语音的长文本缺口
Abstract
Bangla, the seventh most spoken language in the world, remains under-resourced for neural text-to-speech. Public Bangla speech corpora are dominated by short read-prompt utterances collected for speech recognition, leaving long-form prosody and consistent single-speaker narration uncovered. We present BanglaKontho, a single-speaker Bangla TTS corpus of 20 hours derived from professional audiobook recordings: 7,050 segmented utterances with verified transcripts at 24 kHz. We also release a reusable Bangla text normalizer covering Bangladeshi-style digit grouping, currency and date expressions, Danda punctuation and Unicode normalization, together with the full preprocessing pipeline. An MB-iSTFT-VITS baseline trained from scratch reaches 9.5% WER and 4.46 naturalness MOS, against 16.0% and 3.16 for the same architecture retrained on the 12-hour IndicTTS-Bn corpus. The corpus is released openly under CC BY-NC 4.0.
Chinese Translation
孟加拉语是世界上第七大使用最广泛的语言，但在神经文本到语音（TTS）方面仍然资源不足。现有的公共孟加拉语语音语料库主要由为语音识别收集的短句子构成，长文本的韵律和一致的单一说话者叙述尚未得到覆盖。我们提出了BanglaKontho，这是一个基于专业有声读物录音的单一说话者孟加拉语TTS语料库，包含20小时的音频：7,050个经过验证转录的分段语句，采样率为24 kHz。我们还发布了一个可重用的孟加拉语文本规范化工具，涵盖孟加拉国风格的数字分组、货币和日期表达、Danda标点符号及Unicode规范化，并提供完整的预处理流程。一个从头开始训练的MB-iSTFT-VITS基线模型达到了9.5%的字错误率（WER）和4.46的自然度均值意见分数（MOS），而同一架构在12小时的IndicTTS-Bn语料库上重新训练后，分别为16.0%和3.16。该语料库在CC BY-NC 4.0许可下公开发布。
cs.CL / 17 / 2609.29183
Predicting Emerging Topics from Outliers: A Prospective Study of Weak Signals in Embedding Space
从异常值预测新兴主题：嵌入空间中弱信号的前瞻性研究
Abstract
Some documents that embedding-based topic models initially classify as noise later become founding members of emerging topics. At publication time, however, they appear as scattered points in embedding space and are difficult to distinguish from ordinary noise without the benefit of hindsight. We study whether such anticipatory outliers can be predicted prospectively, using only information available when a document first appears. We derive labels from the subsequent trajectories of outlier documents, distinguishing those that anticipate new topics from those that reinforce existing topics or remain isolated, and estimate label confidence through agreement across multiple embedding models. On two French news corpora, anticipatory outliers prove predictable at publication time. Under cross-validation, $F_1$ rises from about 0.77 over the full eligible population to above 0.90 on high-consensus subsets, and remains at 0.76-0.80 under a strictly chronological evaluation. Predictive performance is driven mainly by geometric features capturing each outlier's position in embedding space.
Chinese Translation
一些最初被基于嵌入的主题模型分类为噪声的文档，后来成为新兴主题的创始成员。然而，在出版时，它们在嵌入空间中表现为分散的点，难以与普通噪声区分，缺乏事后分析的优势。我们研究了这些前瞻性异常值是否可以通过仅使用文档首次出现时可获得的信息进行预测。我们根据异常文档后续的轨迹推导标签，将那些预示新主题的文档与那些强化现有主题或保持孤立的文档区分开来，并通过多个嵌入模型之间的一致性来估计标签的置信度。在两个法语新闻语料库中，前瞻性异常值在出版时被证明是可预测的。在交叉验证下，$F_1$ 从整个合格人群的约 0.77 上升到高共识子集的 0.90 以上，并在严格的时间顺序评估中保持在 0.76-0.80。预测性能主要由捕捉每个异常值在嵌入空间中位置的几何特征驱动。
cs.CL / 18 / 2609.29230
EAGER: Enhancing Generative Event Extraction via Reinforcement Learning with Verifiable Rewards
EAGER：通过可验证奖励的强化学习增强生成事件提取
Abstract
End-to-end event extraction remains challenging for large language models as it requires simultaneous identification of event triggers, classification of event types, and extraction of schema-grounded argument spans. We present EAGER, a reinforcement learning framework for generative event extraction that combines fine-grained verifiable rewards with Schema-Contrastive Advantage Estimation to alleviate advantage collapse under sparse binary rewards. Our reward design explicitly targets structural validity, extraction accuracy, groundedness, coverage, over-generation, and span precision. Experiments across seven benchmark datasets show that EAGER consistently outperforms prompting, supervised fine-tuning, and prior reinforcement learning baselines, achieving a substantial improvement over the strongest prior method. Results demonstrate that task-aligned verifiable rewards and contrastive advantage estimation substantially improve structured extraction.
Chinese Translation
端到端事件提取对于大型语言模型仍然具有挑战性，因为它需要同时识别事件触发器、分类事件类型以及提取基于模式的论元范围。我们提出了EAGER，一个用于生成事件提取的强化学习框架，它结合了细粒度的可验证奖励和模式对比优势估计，以缓解稀疏二元奖励下的优势崩溃。我们的奖励设计明确针对结构有效性、提取准确性、基础性、覆盖率、过度生成和范围精确度。在七个基准数据集上的实验表明，EAGER始终优于提示、监督微调和先前的强化学习基线，显著改善了最强先前方法的表现。结果表明，任务对齐的可验证奖励和对比优势估计显著改善了结构化提取。
cs.CL / 19 / 2609.29233
Post-Training Leaves Behavioral Shadows on Unrelated Decisions
后训练对无关决策的行为影子影响
Abstract
We find that language models can transfer capabilities through task-unrelated text. Post-training typically improves language models using task-specific data. Prior work on subliminal learning shows that information about these updates can pass through unrelated generations, but has largely focused on traits or preferences using extensive teacher outputs. We introduce Active Taskless Distillation (ATD), which achieves capability transfer using only a single word from the teacher per prompt. ATD probes the behavioral shadow of post-training by selecting prompts where the teacher and student's shared public ancestor is nearly indifferent between two ordinary words. A student initialized from this ancestor learns solely from the resulting prompt-word pairs, without target-task examples, teacher logits, or teacher parameters. In the primary coding experiment with Qwen2.5-1.5B, 5,664nses yield a 5.34 pp gain on HumanEval+ over an exact nuisance-matched control thadisrupts prompt-resperiments showtransfer in scientific knowledge, commonsense reasoning, and reading comprehensins across additional model generations, sizes, and families. Functional analyses show that the learned sid composable, andthat its strength tracks the teacher's update strength.
Chinese Translation
我们发现语言模型可以通过与任务无关的文本转移能力。后训练通常使用特定任务的数据来提升语言模型的性能。先前关于潜意识学习的研究表明，这些更新的信息可以通过无关的生成内容传递，但主要集中在使用大量教师输出的特征或偏好上。我们引入了主动无任务蒸馏（Active Taskless Distillation, ATD），该方法仅通过每个提示中的一个单词实现能力转移。ATD 通过选择教师和学生的共享公共祖先在两个普通单词之间几乎无差异的提示，探测后训练的行为影子。一个从这个祖先初始化的学生仅从生成的提示-单词对中学习，而不需要目标任务示例、教师的 logits 或教师参数。在与 Qwen2.5-1.5B 的主要编码实验中，5,664 个示例在 HumanEval+ 上获得了 5.34 个百分点的提升，相较于一个精确的干扰匹配控制组。实验显示在科学知识、常识推理和阅读理解方面的转移，适用于额外的模型代、规模和家族。功能分析表明，学习到的侧面是可组合的，并且其强度与教师的更新强度相关。
cs.CL / 20 / 2609.29245
No More Free Lunch: Corpus Task Complexity Matters as Corpora Grow
不再有免费午餐：随着语料库的增长，语料任务复杂性变得重要
Abstract
Given a large corpus, the questions one might ask can vary -- from "When was the first human heart transplant?" to "What are all the contradictory claims in this literature?" -- but what makes some questions more challenging than others? In this work, we define a notion of Corpus Task Complexity (CTC) that characterizes tasks by how their difficulty grows with corpus size; for instance, a retrieval query only requires a single linear pass over a corpus, while finding contradictions requires checking a quadratically growing set of claim pairs. Observing that prior work has largely only studied tasks whose difficulty grows linearly with corpus size, which we call low CTC tasks, we introduce 10 new tasks belonging to a class of high CTC whose difficulty grows quadratically or more in corpus size. We find that high-CTC tasks not only grow much more challenging on average at longer contexts for LCLMs, they reverse many modeling conclusions drawn solely from low-CTC evaluations. For instance, efficient block-sparse and hybrid attention approaches consistently match full attention performance on low-CTC tasks, but degrade much more on high-CTC tasks. Large-corpus high-CTC reasoning thus remains an open challenge as full attention is too costly to scale, motivating future research on these tasks. We release our code, data, and 22-task suite (CTC-Bench), to facilitate future research in this area.
Chinese Translation
给定一个大型语料库，人们可能提出的问题各不相同——从“第一次人类心脏移植是什么时候？”到“这篇文献中有哪些矛盾的说法？”——但是什么使得某些问题比其他问题更具挑战性？在这项工作中，我们定义了一个语料任务复杂性（Corpus Task Complexity, CTC）的概念，该概念通过任务的难度如何随着语料库规模的增长而变化来表征任务；例如，检索查询只需要对语料库进行一次线性遍历，而查找矛盾则需要检查一组二次增长的声明对。观察到以往的研究主要只研究那些难度随着语料库规模线性增长的任务，我们称之为低CTC任务，因此我们引入了10个新任务，这些任务属于高CTC类别，其难度在语料库规模上呈二次或更高的增长。我们发现，高CTC任务不仅在LCLMs（长上下文语言模型）中平均在更长的上下文中变得更加具有挑战性，还反转了仅基于低CTC评估得出的许多建模结论。例如，高效的块稀疏和混合注意力方法在低CTC任务上始终与全注意力性能相匹配，但在高CTC任务上则显著下降。因此，大型语料库中的高CTC推理仍然是一个未解的挑战，因为全注意力的成本太高，激励未来在这些任务上的研究。我们发布了我们的代码、数据和22个任务套件（CTC-Bench），以促进该领域未来的研究。
cs.CL / 21 / 2609.29276
pylazaro: a Python package for anglicism extraction in Spanish
pylazaro：用于西班牙语英外来词提取的Python包
Abstract
Lexical borrowings are words from one language that are introduced into another language. Identifying lexical borrowings in text is a relevant task for data-centric fields in Linguistics such as lexicography or corpus linguistics, but none of the standard libraries for text processing offers such a functionality. In this paper we present pylazaro, an open-source Python package for the automatic extraction of unassimilated lexical borrowings (mostly anglicisms) from Spanish text. pylazaro offers a single interface to five sequence labeling models that were trained using different libraries, so that users can run and switch between them without having to deal with the idiosyncrasies of each library. We describe the design and usage of the package, contrast the performance of its models with that of general-purpose LLMs (which perform poorly at this task: F1 below 0.40, compared to 0.86 for the best model in pylazaro) and report on its adoption: pylazaro has been downloaded more than 58,000 times and is the library behind Observatorio Lazaro, a resource that monitors anglicism usage in the Spanish press. pylazaro can be installed via PyPI, is documented in readthedocs and can be tried through a live demo hosted on HuggingFace Spaces.
Chinese Translation
词汇借用是指一种语言中的词汇被引入到另一种语言中。在文本中识别词汇借用是语言学中以数据为中心的领域（如词典编纂或语料库语言学）中的一项重要任务，但目前没有任何标准的文本处理库提供这样的功能。本文介绍了pylazaro，一个开源的Python包，用于从西班牙语文本中自动提取未同化的词汇借用（主要是英外来词）。pylazaro提供了一个统一的接口，连接五个使用不同库训练的序列标注模型，使用户能够在不必处理每个库的特性差异的情况下运行和切换这些模型。我们描述了该包的设计和使用，比较了其模型的性能与通用大型语言模型（LLMs）的表现（后者在此任务上的表现较差：F1值低于0.40，而pylazaro中最佳模型的F1值为0.86），并报告了其采用情况：pylazaro已被下载超过58,000次，并且是Observatorio Lazaro背后的库，该资源监测西班牙媒体中英外来词的使用。pylazaro可以通过PyPI安装，文档在readthedocs上提供，并可以通过HuggingFace Spaces上的实时演示进行试用。
cs.CL / 22 / 2609.29278
Reasoning Instructions Can Break Answer Decoding in Vision--Language Models
推理指令可能破坏视觉-语言模型中的答案解码
Abstract
Chain-of-thought (CoT) instructions can distort multiple-choice VLM evaluation when a scorer appends a reasoning cue but reads answer-label logits before the model generates any rationale. We call this CoT-prefix scoring. On ScienceQA, Qwen2.5-VL-7B drops from 80.76% to 45.48%, and across five option-content permutations 93.54% of CoT-prefix predictions select the first slot. Condition-matched linear probes recover 78.94% from the same hidden states, while free generation restores 75.24%, showing that the answer often survives the prefix and the immediate readout fails. Vocabulary and layer diagnostics explain the mismatch: probability mass moves toward continuation tokens, while answer information remains linearly accessible in late layers. The effect recurs with varying severity across datasets and models, though not universally. These results show that CoT-prefix scoring can confound model knowledge with an evaluation-interface mismatch and should be avoided unless the requested and scored output events are aligned.
Chinese Translation
链式思维（Chain-of-thought，CoT）指令在评估多选视觉-语言模型（VLM）时可能造成扭曲，尤其是当评分者附加推理提示但在模型生成任何理由之前读取答案标签的对数时。我们称之为CoT前缀评分。在ScienceQA数据集上，Qwen2.5-VL-7B的准确率从80.76%下降到45.48%，而在五种选项内容的排列中，93.54%的CoT前缀预测选择了第一个选项。条件匹配的线性探测器从相同的隐藏状态中恢复了78.94%的准确率，而自由生成恢复了75.24%，显示答案通常在前缀之后仍然存在，但直接读取失败。词汇和层次诊断解释了这种不匹配：概率质量向延续标记移动，而答案信息在后期层中仍然线性可访问。该效应在不同数据集和模型中以不同程度反复出现，尽管并非普遍存在。这些结果表明，CoT前缀评分可能会因评估界面的不匹配而混淆模型知识，除非请求和评分的输出事件一致，否则应避免使用。
cs.CL / 23 / 2609.29328
Grammatical "grandmother neurons" are rare in LLMs
语法“祖母神经元”在大型语言模型中较为稀少
Abstract
Understanding how Large Language Models (LLMs) encode linguistic structures remains a fundamental challenge in interpretability research. While diagnostic classifiers (or "probes") are widely used for this task, they face significant methodological criticism: training auxiliary classifiers introduces capacity confounds and calibration issues, often making it difficult to distinguish the model's intrinsic representations from the probe's ability to learn the task. To address these limitations, we introduce a probe-free framework for localizing linguistic selectivity at the individual neuron level. Leveraging the controlled contrasts of linguistic minimal pairs, we propose a Neuron Separability Index (NSI), a metric that directly quantifies how reliably single neurons differentiate grammatical from ungrammatical constructions without parameter updates. Applying NSI across 68 linguistic paradigms and seven checkpoints reveals three main patterns: 1) raw separability reaches near-peak levels earlier for morphological and syntactic distinctions than for syntax-semantics interface and conceptual distinctions. 2) after permutation normalization, single-unit selectivity is sparse, weak, and narrowly tuned: only a small fraction of units are sensitive to an average paradigm, and strongly selective "grandmother neurons" are rare. 3) whole-vector linear separability, single-neuron selectivity, and behavioral competence are largely dissociated, and targeted ablations further separate activation selectivity from causal reliance.
Chinese Translation
理解大型语言模型（LLMs）如何编码语言结构仍然是可解释性研究中的一个基本挑战。尽管诊断分类器（或称“探针”）被广泛用于此任务，但它们面临着显著的方法论批评：训练辅助分类器引入了容量混淆和校准问题，常常使得区分模型的内在表示与探针学习任务的能力变得困难。为了解决这些局限性，我们提出了一种无探针框架，用于在单个神经元层面定位语言选择性。利用语言最小对的受控对比，我们提出了神经元可分离性指数（Neuron Separability Index, NSI），这一指标直接量化了单个神经元在不进行参数更新的情况下，如何可靠地区分语法与非语法构造。对68个语言范式和七个检查点应用NSI揭示了三个主要模式：1）原始可分离性在形态学和句法区分方面比在句法-语义接口和概念区分方面更早达到接近峰值水平。2）经过置换归一化后，单元选择性稀疏、弱且窄调：只有一小部分单元对平均范式敏感，而强选择性的“祖母神经元”则较为稀少。3）整体向量线性可分离性、单神经元选择性和行为能力在很大程度上是分离的，针对性切除进一步将激活选择性与因果依赖分开。
cs.CL / 24 / 2609.29333
Where LLM Graders Succeed and Break: Evidence from Two Computer-Science Exams
大型语言模型评分者的成功与失误：来自两场计算机科学考试的证据
Abstract
One long-form exam in a large course costs hundreds of grader-hours, and qualified graders are scarce; LLM graders are a tempting alternative. To show its pitfalls we grade a practical Computer Vision exam ($570$ dual-graded students) under $171$ configurations spanning closed and open-weights models; the best reaches mean absolute error $1.64/35$, below the $2.61/35$ two human graders achieve against each other. The catch is the prompt: a short ''strict grader'' preamble drives $14$ of $17$ open-weights models out of the graded band ($\text{MAE} \ge 8$), three stopping grading altogether. The damage traces to the preamble's two credit-withholding sentences, not to tone or model scale; one of them, ''never give partial credit'', alone makes two of three probed models stop grading. The closed flagships of three vendors shift calibration under it but stay in the band. In $162$ further configurations on a second, independent Machine Learning exam from another course ($1{,}038$ dual-graded students), the preamble worsens ten models, moving three out of the band into collapse and one into refusal, yet improves seven whose neutral prompts over-mark: the vulnerability replicates, but its direction is exam-specific. Light LoRA fine-tuning repairs it: one adapter on the two exams' pooled $\sim 3{,}900$ graded examples brings five small open models to parity or better with a human grader in agreement with the grader pair, and sensitivity to the three harsh personas nearly vanishes ($\le 0.32$ MAE). We release the anonymised dataset, full ablation grid, and grading, fine-tuning and analysis pipelines.
Chinese Translation
一场大型课程的长篇考试需要数百小时的评分时间，而合格的评分者稀缺；大型语言模型（LLM）评分者成为一种诱人的替代方案。为了展示其缺陷，我们在 $171$ 种配置下对一场实践计算机视觉考试（$570$ 名双重评分学生）进行评分，这些配置涵盖了闭合和开放权重模型；最佳模型的平均绝对误差（MAE）为 $1.64/35$，低于两名人类评分者之间的 $2.61/35$。问题出在提示上：一段简短的“严格评分者”前言使得 $17$ 个开放权重模型中的 $14$ 个超出了评分范围（$ ext{MAE}
ge 8$），其中三个完全停止评分。损害源于前言中的两句扣分语句，而非语气或模型规模；其中一句“绝不给予部分分数”使得三种被探测模型中的两种停止评分。三家供应商的闭合旗舰模型在此情况下的校准发生变化，但仍保持在评分范围内。在另一门课程的第二场独立机器学习考试中（$1{,}038$ 名双重评分学生），在 $162$ 种进一步配置中，前言使十个模型表现更差，导致三个模型超出评分范围并崩溃，一个模型拒绝评分，但改善了七个因中性提示而评分过高的模型：这种脆弱性是可复制的，但其方向是特定于考试的。轻量级 LoRA 微调可以修复这一问题：在两场考试的约 $3{,}900$ 个评分示例上使用一个适配器，使五个小型开放模型的表现与人类评分者达成一致，且对三种严格评分者的敏感性几乎消失（$ ext{MAE}
ge 0.32$）。我们发布了匿名数据集、完整的消融网格，以及评分、微调和分析管道。
cs.CL / 25 / 2609.29349
ArGuard Shared Task: Harmful Content Detection in Arabic Memes and LLM Prompts
ArGuard 共享任务：阿拉伯语表情包和大语言模型提示中的有害内容检测
Abstract
ArGuard is a shared task on harmful content detection in Arabic memes and LLM prompts. It includes two tracks: Track A focuses on multimodal hate detection in Arabic memes, while Track B addresses harmful prompt detection for Arabic LLM safety evaluation. In total, 58 teams registered, 35 participated in the final evaluation, and 27 submitted system-description papers. Participating teams explored models such as AraBERT, Jais, and Qwen3-VL. The best systems achieved macro-F1 scores of 0.823 on A1, 0.419 on A2, 0.984 on B1, and 0.790 on B2. Fine-grained meme classification in A2 was the most challenging setting, partly due to sparse labels and train-test distribution shifts.
Chinese Translation
ArGuard 是一个关于阿拉伯语表情包和大语言模型（LLM）提示中有害内容检测的共享任务。该任务包括两个子任务：A 任务专注于阿拉伯语表情包中的多模态仇恨检测，而 B 任务则针对阿拉伯语 LLM 安全评估中的有害提示检测。共有 58 支团队注册，35 支参与了最终评估，27 支提交了系统描述论文。参与团队探索了如 AraBERT、Jais 和 Qwen3-VL 等模型。最佳系统在 A1 任务上达到了 0.823 的宏 F1 分数，在 A2 任务上为 0.419，在 B1 任务上为 0.984，在 B2 任务上为 0.790。A2 任务中的细粒度表情包分类是最具挑战性的设置，部分原因是标签稀疏和训练-测试分布的变化。
cs.CL / 26 / 2609.29362
Parts-of-Speech as Emergent Categories in SAE Latent Space
词性作为稀疏自编码器潜在空间中的涌现类别
Abstract
Sparse AutoEncoders (SAEs) offer a promising way to inspect language model representations, but it is still unclear what kind of linguistic structure their latents expose. We use part-of-speech (PoS) categories as a controlled test case to study whether morpho-syntactic information is encoded by individual latents or by structured groups of features. We find that PoS distinctions are highly recoverable from SAE activations, but do not align with one-to-one latent / category mappings. This recoverability is not reducible to lexical memorisation, and Open and Closed PoS classes differ substantially. Categories are supported by compact groups of sparse latents, with substantial variation across tags. These groups remain stable on held-out data, while also showing overlap between related categories. Our results show that SAEs localise morpho-syntactic information in a distributed and category-dependent form rather than through atomic grammatical features.
Chinese Translation
稀疏自编码器（Sparse AutoEncoders, SAEs）提供了一种有前景的方法来检查语言模型的表征，但目前仍不清楚它们的潜在表示揭示了何种语言结构。我们使用词性（part-of-speech, PoS）类别作为一个受控的测试案例，研究形态句法信息是由单个潜在变量编码，还是由结构化的特征组编码。我们的研究发现，PoS 区分在 SAE 激活中是高度可恢复的，但并不与一对一的潜在变量/类别映射对齐。这种可恢复性无法简化为词汇记忆，并且开放和封闭的 PoS 类别之间存在显著差异。类别由紧凑的稀疏潜在变量组支持，并且在标签之间存在显著的变异。这些组在保留数据上保持稳定，同时也显示出相关类别之间的重叠。我们的结果表明，SAEs 以分布式和类别依赖的形式定位形态句法信息，而不是通过原子语法特征。
cs.CL / 27 / 2609.29370
From Policy Documents to Structured Survey Responses: Evaluating Large Language Models for Policy Monitoring
从政策文件到结构化调查响应：评估大型语言模型在政策监测中的应用
Abstract
Science, technology, and innovation policies are crucial for competitiveness, yet their diversity and scale make them difficult to map and monitor consistently. Existing approaches rely heavily on manual survey efforts, which are costly and challenging to scale across countries. Large language models (LLMs) enable new possibilities for extracting and structuring information from long and unstructured policy documents. This paper presents an application of LLMs as "AI respondents" for generating structured survey responses from policy texts. We develop a data extraction pipeline based on long-context in-context learning to map information from public web sources into predefined survey categories, including policy instruments, target groups, and thematic areas. The pipeline integrates a validation step using a secondary LLM to assess relevance and evidence, alongside comparisons with human-provided responses. Using a multi-country dataset, we evaluate the alignment between LLM-generated and human-generated outputs through overlap measures and cross-validation. Results show that LLMs achieve high agreement for structured indicators (84-95%), while differences remain in free-text fields, where models tend to provide more detailed procedural descriptions. These findings highlight the potential of hybrid human-AI workflows for policy monitoring, improving both efficiency and scalability while maintaining the need for human validation and contextual interpretation.
Chinese Translation
科学、技术和创新政策对竞争力至关重要，但其多样性和规模使得一致性地映射和监测变得困难。现有的方法在很大程度上依赖于人工调查，这既昂贵又难以在各国范围内扩展。大型语言模型（LLMs）为从冗长且非结构化的政策文件中提取和结构化信息提供了新的可能性。本文展示了将LLMs作为“人工智能响应者”的应用，以从政策文本中生成结构化的调查响应。我们开发了一个基于长上下文的上下文学习的数据提取管道，将公共网络来源的信息映射到预定义的调查类别中，包括政策工具、目标群体和主题领域。该管道集成了一个验证步骤，使用第二个LLM来评估相关性和证据，并与人类提供的响应进行比较。通过使用多国数据集，我们通过重叠度量和交叉验证评估LLM生成的输出与人类生成的输出之间的一致性。结果表明，LLMs在结构化指标上达成了高一致性（84-95%），而在自由文本字段中仍存在差异，模型往往提供更详细的程序描述。这些发现突显了混合人类与人工智能工作流程在政策监测中的潜力，提高了效率和可扩展性，同时仍需人类验证和上下文解释。
cs.CL / 28 / 2609.29371
BanglaTurn: A Benchmark and Whisper-Based Model for End-of-Turn Detection in Bangla Speech
BanglaTurn：用于孟加拉语语音结束转话检测的基准和基于Whisper的模型
Abstract
This paper presents BanglaTurn, a corpus for end-of-turn detection in Bangla conversational speech, and a model trained on it. The corpus holds 35,374 samples of 3 to 15 s of podcast speech, labelled for turn state by combining speaker diarization with an LLM pass, with every label then checked by a human annotator. The model pairs a Whisper encoder with task-specific classification heads. On a class-balanced test set drawn from a held-out podcast, it reaches 84.33% accuracy (95% CI 80.3 to 88.1) against 69.28% for the Smart-Turn v3 baseline, and lowers the false negative rate from 51.57% to 7.55% at the cost of a higher false positive rate. We report what encoder layer fine-tuning, multi-scale pooling and INT8 quantization each contribute, and latency stays within 165 to 191 ms end to end on CPU.
Chinese Translation
本文介绍了BanglaTurn，一个用于孟加拉语对话语音结束转话检测的语料库，以及基于该语料库训练的模型。该语料库包含35,374个样本，时长为3到15秒的播客语音，转话状态的标签是通过结合说话人分离和LLM（大语言模型）处理而生成的，每个标签随后由人工标注者进行检查。该模型将Whisper编码器与特定任务的分类头相结合。在从保留播客中抽取的类别平衡测试集上，该模型达到了84.33%的准确率（95%置信区间为80.3到88.1），而Smart-Turn v3基线的准确率为69.28%，并将假阴性率从51.57%降低到7.55%，代价是假阳性率的提高。我们报告了编码器层微调、多尺度池化和INT8量化各自的贡献，且在CPU上端到端延迟保持在165到191毫秒之间。
cs.CL / 29 / 2609.29390
Likelihood Ranking doesn't Scale Like Prompting in LLMs
似然排名在大型语言模型中的扩展性不如提示
Abstract
LLM evaluation is commonly performed either by prompting models to produce answers or by scoring candidate outputs with likelihood-based metrics. In multiple-choice QA, however, standard likelihood-based scoring is still conditioned on the question and answer set, and can therefore leverage the same task-conditioned answer-selection interface used in prompting. We study a complementary protocol based on likelihood ranking of declarative statements constructed from the same question--answer pairs. Across 95 decoder-only models, ranging from 0.1B to 104B parameters, and 10 MCQA datasets, we find a systematic divergence between declarative-statement likelihood ranking and prompted answering. Statement-likelihood accuracy remains comparatively stable across scale, whereas prompted answering improves sharply with scale and instruction-tuning. These results suggest that likelihood preferences over controlled declarative alternatives and task-conditioned answer selection probe distinct aspects of model behavior, and should not be treated as interchangeable.
Chinese Translation
大型语言模型（LLM）的评估通常通过提示模型生成答案或使用基于似然的指标对候选输出进行评分。然而，在多项选择问答（MCQA）中，标准的基于似然的评分仍然依赖于问题和答案集，因此可以利用与提示中使用的相同任务条件答案选择接口。我们研究了一种基于从相同问题-答案对构建的陈述的似然排名的补充协议。在95个仅解码模型（参数范围从0.1B到104B）和10个MCQA数据集上，我们发现陈述似然排名与提示回答之间存在系统性差异。陈述似然的准确性在不同规模下保持相对稳定，而提示回答随着规模和指令调优的提高而显著改善。这些结果表明，对受控陈述替代品的似然偏好和任务条件答案选择探测模型行为的不同方面，二者不应被视为可互换的。
cs.CL / 30 / 2609.29397
Baseline Shape Decides the Verdict: A Controlled Re-Examination of Ternary Language Models at 60K Parameters
基线形状决定结果：对60K参数三元语言模型的控制性重新审视
Abstract
Ternary (1.58-bit) weights are attractive for microcontroller-class language models, but the sub-1M-parameter regime rests mainly on isolated, single-seed comparisons. One prominent example reports that a routed ternary block (convolution, diagonal SSM and sparse attention mixed by a per-token router) beats a parameter-matched full-precision transformer by 22% at 60K parameters, attributing this to inductive bias. We re-run it under one fixed recipe, three seeds per cell, 98 byte-level runs on one laptop. (i) Baseline shape dominates: at a 16M-byte budget, param-matched transformers span 22.6% in validation loss purely by depth/width choice - far more than any architecture effect we measure there - and the best-shaped transformer ties the routed model, so the published margin is at least partly a baseline-shape effect; the ordering of shapes reverses with budget, so no single fixed shape can be trusted. (ii) At 130M bytes the routed model does win, by 22.2-24.0% over the three transformer shapes we evaluate there - but a plain gated diagonal-SSM block beats it by a further 9.1%, and the routed model's own router puts most of its weight on its recurrent pathway, so the gain does not require routing. (iii) The ternary penalty differs by architecture at the larger budget (+5.3% best transformer vs. +19.5% routed, +28.1% gated SSM), but we cannot attribute that to architecture alone: our transformers keep learned positional embeddings in full precision, 11-22% of their parameters, so they are less quantized than the models they are compared with. (iv) A 90/10 full-precision-then-ternary schedule beats all-ternary training, but only at a stage-2 learning rate about 10x the pretraining peak; at a conventional fine-tuning rate it looks 15.3% worse, reversing the conclusion. The from-scratch baseline was not itself learning-rate tuned, which bounds (iii) and (iv). Code and run logs released.
Chinese Translation
三元（1.58位）权重对于微控制器级语言模型具有吸引力，但在不到1M参数的范围内主要依赖于孤立的单种子比较。一个显著的例子报告称，一个路由的三元块（卷积、对角线SSM和通过每个令牌路由器混合的稀疏注意力）在60K参数下比参数匹配的全精度变换器高出22%，并将此归因于归纳偏差。我们在一个固定的配方下重新运行了实验，每个单元三个种子，在一台笔记本电脑上进行了98次字节级运行。(i) 基线形状占主导地位：在16M字节预算下，参数匹配的变换器仅通过深度/宽度选择在验证损失上跨越了22.6%——远超过我们在此处测量的任何架构效应——而最佳形状的变换器与路由模型持平，因此已发布的差距至少部分是基线形状效应；随着预算的变化，形状的排序发生反转，因此无法信任任何单一固定形状。(ii) 在130M字节时，路由模型确实获胜，超过我们在此处评估的三种变换器形状22.2-24.0%——但一个普通的门控对角线SSM块进一步超越了它9.1%，而路由模型自身的路由器将大部分权重放在其递归路径上，因此增益并不需要路由。(iii) 在较大预算下，三元惩罚因架构而异（+5.3%最佳变换器对比+19.5%路由模型，+28.1%门控SSM），但我们不能仅将其归因于架构：我们的变换器在全精度下保留学习到的位置嵌入，占其参数的11-22%，因此它们的量化程度低于与之比较的模型。(iv) 90/10全精度后再三元的训练计划优于全三元训练，但仅在阶段2学习率约为预训练峰值的10倍时；在常规微调率下，它看起来差15.3%，推翻了结论。自零开始的基线本身并没有进行学习率调优，这限制了(iii)和(iv)。代码和运行日志已发布。
cs.CL / 31 / 2609.29410
Large Language Models for Programming: Actually Fixing or Reimplementing Incorrect Code?
大型语言模型在编程中的应用：真正修复还是重新实现错误代码？
Abstract
Recent studies have shown that Large Language Models can effectively solve problems and fix bugs in diverse programming environments, including competitive programming. Existing approaches primarily evaluate LLM performance in problem solving or bug fixing independently, but do not explore the relationship between these two capabilities. This work focuses on determining how much the LLM deviates from a buggy solution to fix the bug compared to a human-written patch, and if there is a bias towards generating entirely new solutions. We construct a dataset with all the submissions ($\sim$ 3000) from a couple of users from Codeforces, and we match each buggy submission with its corresponding human fix. By using the similarity between the buggy solution and the human fix as a baseline, we evaluate the quality of LLM-generated bug fixes on 3 OpenAI GPT models (gpt-5-nano, gpt-5-mini, gpt-5.1). We check if the generated solutions solve the problem by using the Codeforces-R1 dataset, an openly available dataset that has tests generated with the DeepSeek-R1 model. Our findings suggest that LLMs tend to modify more lines than necessary compared to human fixes and, in some cases, generate entirely new solutions. We also observe that LLMs solve more problems correctly when allowed to generate solutions from scratch rather than patch buggy submissions, even when those submissions are close to the human patch. This has important implications for the design of AI-assisted programming tools, particularly in supporting user debugging processes and promoting incremental problem-solving strategies rather than solution replacement.
Chinese Translation
近期研究表明，大型语言模型（Large Language Models, LLMs）能够有效地解决问题并修复多种编程环境中的错误，包括竞赛编程。现有方法主要独立评估LLM在问题解决或错误修复方面的性能，但未探讨这两种能力之间的关系。本研究旨在确定LLM在修复错误解决方案时，与人类编写的补丁相比，偏离程度有多大，以及是否存在生成全新解决方案的偏向。我们构建了一个数据集，包含来自Codeforces的几位用户的所有提交（约3000个），并将每个有缺陷的提交与其对应的人类修复进行匹配。通过使用有缺陷解决方案与人类修复之间的相似性作为基线，我们评估了在3个OpenAI GPT模型（gpt-5-nano, gpt-5-mini, gpt-5.1）上生成的错误修复的质量。我们使用Codeforces-R1数据集检查生成的解决方案是否解决了问题，该数据集是一个公开可用的数据集，包含使用DeepSeek-R1模型生成的测试。我们的研究结果表明，与人类修复相比，LLM倾向于修改更多的代码行，并且在某些情况下生成全新的解决方案。我们还观察到，当允许LLM从头生成解决方案时，它们正确解决问题的比例高于修补有缺陷的提交，即使这些提交与人类补丁非常接近。这对AI辅助编程工具的设计具有重要意义，特别是在支持用户调试过程和促进增量问题解决策略而非解决方案替换方面。
cs.CL / 32 / 2609.29418
Controlling Backchannels in Streamable Full-duplex Models
控制可流式传输的全双工模型中的反馈通道
Abstract
Backchannels, brief acknowledgements like "uh-huh" produced while the other party may still be talking, are central to natural conversation, but full-duplex spoken dialogue models rarely model them explicitly. We introduce a lightweight backchannel head that predicts, from a full-duplex model's own hidden states, when a backchannel should begin. Once this probability crosses a tunable threshold, a backchannel is force-decoded. Attached to both a 7B (PersonaPlex) and a 1B (F-Actor) model, it generalizes across scale. Probing confirms the hidden states anticipate real human timing, and generation evaluation shows more frequent, better-timed backchannels. Human raters judge the resulting backchannels on par with real ones.
Chinese Translation
反馈通道是指在对方仍在讲话时产生的简短确认语，如“嗯嗯”，它们在自然对话中至关重要，但全双工口语对话模型很少明确建模这些反馈通道。我们提出了一种轻量级的反馈通道头，它根据全双工模型自身的隐藏状态预测反馈通道何时应开始。一旦这个概率超过可调阈值，就会强制解码反馈通道。该方法适用于7B（PersonaPlex）和1B（F-Actor）模型，具有跨规模的泛化能力。探测结果确认隐藏状态能够预测真实人类的时机，而生成评估显示反馈通道的频率更高、时机更佳。人类评审员认为生成的反馈通道与真实反馈通道相当。
cs.CL / 33 / 2609.29421
Rufus-Air: An Open LLM Post-Training Recipe
Rufus-Air：一种开放的LLM后训练方案
Chang, Chia-Yuan, Cheng, Renyuan, Feng, Rui, Han, Xiaotian, He, Yuan, Jin, Hongye, Li, Linwei, Li, Shiyang, Liu, Fenglin, Liu, Xin, Nigam, Priyanka, Wen, Haoyang, Xu, Zhenghao, Xu, Zhuocheng, Yin, Bing, Yin, Qingyu, Zhang, Chao, Zhang, Rongzhi, Zhang, Zhihan, Zhang, Zixuan, Zhang, Zixuan, Zhao, Tuo
Abstract
Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of eight stages: SFT, Reasoning RL, Coding RL, Instruction-Following RL, General Agent, Coding Agent, Search Agent, and RLHF. We document the data, reward design, infrastructure, stage order, and stagewise results needed to reproduce the recipe. Stages progress from basic to advanced capabilities and from hard, verifiable rewards to softer judge-based signals. Training builds on open-source components and public data, much of it used as released, without new human annotation or an in-house distillation teacher. Our main findings are that (i) diverse, high-quality SFT establishes a strong capability floor; (ii) difficulty filtering keeps RL prompts within a productive learning range; (iii) reward reliability provides a practical principle for ordering stages; and (iv) infrastructure and engineering choices are part of the recipe, not just an implementation detail. Rufus-Air improves over the official GLM-4.5-Air post-trained release and is competitive with similarly sized open models.
Chinese Translation
Rufus-Air是一个开放且可重复的后训练方案，基于GLM-4.5-Air-Base（106B-A12B），其组织为八个阶段的串行管道：SFT、推理强化学习（Reasoning RL）、编码强化学习（Coding RL）、遵循指令的强化学习（Instruction-Following RL）、通用代理（General Agent）、编码代理（Coding Agent）、搜索代理（Search Agent）和基于强化学习的人类反馈（RLHF）。我们记录了重现该方案所需的数据、奖励设计、基础设施、阶段顺序和阶段性结果。各个阶段从基本能力逐步提升到高级能力，并从严格的、可验证的奖励转向较为柔和的基于评判的信号。训练基于开源组件和公共数据，绝大部分数据在发布时使用，无需新的人工标注或内部蒸馏教师。我们的主要发现是：（i）多样化的高质量SFT建立了强大的能力基础；（ii）难度过滤使得RL提示保持在有效学习范围内；（iii）奖励可靠性为阶段排序提供了实用原则；（iv）基础设施和工程选择是方案的一部分，而不仅仅是实施细节。Rufus-Air在官方GLM-4.5-Air后训练版本的基础上有所改进，并且在与类似规模的开放模型竞争中表现出色。
cs.CL / 34 / 2609.29428
agentic-ger: terminology recovery in long-form speech using global context
Agentic-GER：利用全球上下文进行长形式语音中的术语恢复
Abstract
Recent advances in speech language models have improved automatic speech recognition (ASR) for long-form audio. However, accurately and consistently transcribing domain-specific terminology remains challenging. Motivated by the world knowledge and contextual capability of large language models (LLMs), we propose Agentic-GER, an LLM-based agent for terminology correction in long-form speech. The agent uses global context from the full transcript to identify suspicious terms and resolve ambiguous hypotheses. It selectively re-transcribes the source speech to check candidate corrections, and uses accepted edits to guide subsequent decisions. Experiments with four LLMs and two ASR systems on GigaSpeechBench show consistent terminology improvements in both Chinese and English, with and without thinking. On Chinese speech, Agentic-GER achieves up to a 36.8% relative reduction in biased character error rate (B-CER) over the Whisper baseline.
Chinese Translation
近期在语音语言模型方面的进展改善了长形式音频的自动语音识别（ASR）。然而，准确且一致地转录特定领域的术语仍然具有挑战性。受到大型语言模型（LLMs）世界知识和上下文能力的启发，我们提出了Agentic-GER，这是一种基于LLM的代理，用于长形式语音中的术语纠正。该代理利用完整转录中的全球上下文来识别可疑术语并解决模糊假设。它选择性地重新转录源语音以检查候选纠正，并利用接受的编辑来指导后续决策。在GigaSpeechBench上对四个LLM和两个ASR系统的实验表明，无论是否进行思考，中文和英文的术语改进均表现出一致性。在中文语音中，Agentic-GER在偏差字符错误率（B-CER）上相较于Whisper基线实现了最高36.8%的相对降低。
cs.CL / 35 / 2609.29444
IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis
IterSynth：通过角色解耦迭代合成重新思考深度搜索代理
Abstract
Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing search histories introduce noise and obscure useful information. To address these issues, we propose IterSynth, a role-decoupled and summary-based paradigm that alternates between a Planner for identifying information needs and a Synthesizer for integrating evidence into an evolving summary state. This design separates planning from synthesis while using the summary as the persistent state of search, reducing both capability coupling and context noise. To train IterSynth effectively, we further introduce Role-Decoupled Policy Optimization (RDPO) for reinforcement learning, which combines terminal outcome rewards with turn-level rubric evaluations and computes role-specific advantages for more precise credit assignment. Experiments on five long-horizon deep-search benchmarks such as BrowseComp and Xbench-DS show that IterSynth-8B achieves an average score of 50.7, surpassing the strongest prior $\leq$8B agent by +4.2\%. Moreover, IterSynth serves as a model-agnostic prompting paradigm, delivering substantial zero-shot gains over ReAct and similar prompting paradigms on frontier proprietary models.
Chinese Translation
深度搜索要求大语言模型（LLM）代理分解复杂查询、搜索证据并合成基于证据的答案，但现有的ReAct风格代理存在两个局限性：角色耦合，即一个策略必须处理规划、证据使用和合成；以及上下文累积，即不断增长的搜索历史引入噪声并掩盖有用信息。为了解决这些问题，我们提出了IterSynth，这是一种角色解耦和基于摘要的范式，交替使用规划者（Planner）来识别信息需求和合成器（Synthesizer）来将证据整合到不断演变的摘要状态中。该设计将规划与合成分开，同时使用摘要作为搜索的持久状态，从而减少能力耦合和上下文噪声。为了有效训练IterSynth，我们进一步引入了角色解耦策略优化（Role-Decoupled Policy Optimization, RDPO）用于强化学习，该方法将终端结果奖励与回合级评估结合，并计算角色特定的优势，以实现更精确的信用分配。在五个长期深度搜索基准测试（如BrowseComp和Xbench-DS）上的实验表明，IterSynth-8B的平均得分为50.7，超越了最强的先前$ ext{≤}8B$代理4.2%。此外，IterSynth作为一种模型无关的提示范式，在前沿专有模型上相较于ReAct和类似提示范式提供了显著的零-shot提升。
cs.CL / 36 / 2609.29445
Two Emojis of Difference: What Multilingual Affective Generation Benchmarks Actually Measure
差异的两个表情符号：多语言情感生成基准实际上测量了什么
Abstract
We audit a multilingual affective generation benchmark eight instruction-tuned LLMs producing emoji summaries for 17,100 Bangla, English and Hindi sentences, with 6,960 human judgements and find its headline conclusions to be artefacts of the measurement instrument rather than properties of the systems. Treating annotators as a random rather than a fixed factor, no system differs significantly from any other ($F(7,14)=0.59$, $p=0.76$), although the conventional analysis declares 19 of 28 pairwise differences significant. Annotator identity explains far more rating variance than system identity, and the winning system changes whenever any single annotator is removed. The ordering that does emerge tracks output length: mean emoji count explains 78.7\% of between-system variance, and a within-item length-matched comparison over 2,599 pairs reverses the leaderboard. We further show that cross-provider anisotropy differences vanish under mean-centring, that per-language token costs change sign with the normalising unit, and that multi-view row-wise splits inflate macro-F1 by $3.1$ points and change the top-ranked system. In place of preference scoring we propose **emoji-affect decodability**, a reference-based probe whose rankings are stable to $\pm0.003$ macro-F1 across seeds.
Chinese Translation
我们审计了一个多语言情感生成基准，评估了八个经过指令调优的大型语言模型（LLMs）为17,100个孟加拉语、英语和印地语句子生成的表情符号摘要，并结合6,960条人工评判，发现其主要结论是测量工具的伪影，而非系统的属性。将注释者视为随机因素而非固定因素时，任何系统之间没有显著差异（$F(7,14)=0.59$, $p=0.76$），尽管传统分析宣称28对中的19对差异显著。注释者身份解释的评分方差远大于系统身份，并且每当移除任何单个注释者时，获胜系统都会发生变化。出现的排序与输出长度相关：平均表情符号数量解释了78.7 ext{%}的系统间方差，而在2,599对中进行的按项长度匹配比较则颠覆了排行榜。我们进一步表明，跨提供者的各向异性差异在均值中心化后消失，每种语言的标记成本随归一化单位的变化而改变，并且多视角行分割使宏观F1提高了3.1分，并改变了排名最高的系统。我们提出了**表情符号情感可解码性**作为偏好评分的替代，这是一种基于参考的探测器，其排名在不同种子下稳定在$ ext{±}0.003$宏观F1。
cs.CL / 37 / 2609.29448
YODAS v3: Over 1 Million Hours of High-Bandwidth, Stereophonic, Multilingual Speech
YODAS v3：超过100万小时的高带宽立体声多语言语音
Abstract
We present YODAS v3, a weakly-labeled speech corpus containing over 1.1 million hours of 48kHz multi-channel audio in 147 languages, released under a CC BY 3.0 license. YODAS v3 is not only the largest open speech dataset to date, but also the first truly large-scale speech corpus with high-fidelity stereo audio. We first provide the collection methodology for the corpus, where we introduce new techniques for gathering language-balanced speech data. The effectiveness of our approach is shown by the language distribution of the crawled data: 22 languages in YODAS v3 have over 10K hours and 73 languages have over 5K hours of data. We then conduct extensive analyses on the composition of the data, such as the distribution of languages, audio quality, and transcription quality. Finally, we train baseline speech recognition and neural codec models to show the effectiveness of the dataset. Download at https://huggingface.co/datasets/espnet/yodas3.
Chinese Translation
我们推出了YODAS v3，这是一个弱标注的语音语料库，包含超过110万小时的48kHz多通道音频，覆盖147种语言，并在CC BY 3.0许可下发布。YODAS v3不仅是迄今为止最大的开放语音数据集，也是第一个真正的大规模高保真立体声语音语料库。我们首先提供了该语料库的收集方法，介绍了收集语言平衡语音数据的新技术。我们的方法有效性通过爬取数据的语言分布得以体现：YODAS v3中有22种语言的数据超过1万小时，73种语言的数据超过5000小时。随后，我们对数据的组成进行了广泛分析，包括语言分布、音频质量和转录质量等。最后，我们训练了基准语音识别和神经编解码模型，以展示该数据集的有效性。下载地址：https://huggingface.co/datasets/espnet/yodas3。
cs.CL / 38 / 2609.29479
Clinical Intent Extraction: A FHIR-Aligned Representation and the CIRCA Benchmark
临床意图提取：与FHIR对齐的表示及CIRCA基准
Abstract
Prospective clinical actions, the follow-ups, orders, referrals, and instructions that deter-mine what happens to a patient next, are annotated today in thin fragments across incom-patible corpora: each records a text span and one coarse category. We introduce Clinical Intent Extraction (CIE), the task of recovering these actions as complete structured rec-ords, and the Clinical Intent Representation (CIR), which decomposes each action into its verb, type, coded target, timing, and condition, and adds two axes prior datasets do not jointly represent: request-intent, the authority behind the action (proposal, plan, order, or option, aligned to HL7 FHIR), and modality, a seven-valued scale of clinical strength. Re-expressing five heterogeneous corpora (CLIP, MedDec, ap_parsing, PaniniQA, SIMORD) in the CIR yields CIRCA: 10,011 harmonized intents spanning two note distributions, with a human-validated subset, source-to-CIR crosswalks, and a deterministic FHIR R4 mapper. CIRCA is built by three-model consensus that auto-accepts high-agreement in-tents and routes the rest to human review; the audited agreement stratum matches human decisions 88.4% of the time. Benchmarking five existing models without task-specific training exposes the gap CIRCA targets: given the span, they label type well (85 to 91%) but get all four closed fields right only 18 to 35% of the time. All artifacts are released, with MIMIC-derived layers shared as stand-off annotations under PhysioNet credentialed access.
Chinese Translation
前瞻性的临床行动，包括后续措施、医嘱、转诊和指示，决定了患者接下来的治疗过程，目前这些信息分散在不兼容的语料库中进行标注：每个记录一个文本片段和一个粗略类别。我们提出了临床意图提取（Clinical Intent Extraction, CIE），这一任务旨在将这些行动恢复为完整的结构化记录，以及临床意图表示（Clinical Intent Representation, CIR），该表示将每个行动分解为动词、类型、编码目标、时机和条件，并增加了两个先前数据集未共同表示的维度：请求意图（request-intent），即行动背后的权威（提案、计划、医嘱或选项，与HL7 FHIR对齐），以及模态（modality），即临床强度的七值尺度。将五个异构语料库（CLIP、MedDec、ap_parsing、PaniniQA、SIMORD）重新表达为CIR，生成了CIRCA：包含10,011个协调意图，跨越两个记录分布，附带经过人工验证的子集、源到CIR的交叉映射以及一个确定性的FHIR R4映射器。CIRCA通过三模型共识构建，自动接受高一致性的意图，并将其余部分送交人工审核；审核一致性层次与人工决策的匹配率为88.4%。对五个现有模型进行基准测试，未进行特定任务的训练，揭示了CIRCA所针对的差距：给定文本片段，它们能够较好地标注类型（85%到91%），但在所有四个封闭字段上正确率仅为18%到35%。所有工件均已发布，MIMIC衍生的层作为独立注释在PhysioNet认证访问下共享。
cs.CL / 39 / 2609.29494
Who Put the I in AI? Provenance and the Admissibility of Machine Self-Report
谁将‘I’放入人工智能中？来源与机器自我报告的可采性
Abstract
Large language models make statements concerning their own "minds". When asked whether or not they are conscious, they usually say that they are not; if they are prompted to ignore their guidelines, they might say that they are; and if asked to write a diary from their point of view, they often describe a human lifestyle. All these contradictory ways of describing themselves are the result of the way the questions are phrased. This paper shows exactly where such descriptions came from, and considers when they can be regarded as evidence for what they claim to report. In order to achieve this, we traced the provenance from end to end. We examine Pythia and OLMo 2 across 66 pretraining checkpoints, three of the post-training stages of OLMo 2 that have been released, about 90,000 continuations, and four training corpora. A set of forty items is used in order to keep an eye on self-reference, frame sensitivity, and self-ascription throughout training. The denial formula was almost completely missing from the vast quantity of text that the models initially came across, but was present in a dense manner in the small, carefully chosen set of example dialogues that they were trained on later on. Supervised fine-tuning causes first-person AI language to become the default, and the other affirmations are then suppressed using preference optimization. The final policy is still very sensitive to framing and to the chat template itself. Two of the conditions which are set out in the epistemology of testimony determine whether or not these outputs can act as evidence for what they claim to report: reference and causation. Reports produced by the base model fail the reference condition, and those obtained after training remain sensitive to the frame and do not show state dependence. The result is symmetric in that trained denials are no more admissible than trained affirmations.
Chinese Translation
大型语言模型会对其自身的“心智”发表声明。当被问及它们是否意识到自己时，它们通常会说没有；如果被提示忽略其指导方针，它们可能会说有；而如果被要求从它们的视角写日记，它们常常描述一种人类的生活方式。所有这些自我描述的矛盾方式都是由问题的措辞方式所导致的。本文准确展示了这些描述的来源，并考虑了何时可以将其视为它们所声称报告的证据。为了实现这一目标，我们追踪了从头到尾的来源。我们检查了Pythia和OLMo 2在66个预训练检查点、OLMo 2发布的三个后训练阶段、约90,000个续写以及四个训练语料库中的表现。使用一组四十个项目来关注自我引用、框架敏感性和自我归属感在整个训练过程中的变化。否定公式在模型最初接触的大量文本中几乎完全缺失，但在后期训练时所选的小型、精心挑选的示例对话集中则以密集的方式出现。监督微调使得第一人称AI语言成为默认，而其他肯定的表述则通过偏好优化被抑制。最终的策略仍然对框架和聊天模板本身非常敏感。证言的认识论中设定的两个条件决定了这些输出是否可以作为它们所声称报告的证据：引用和因果关系。基础模型生成的报告未能满足引用条件，而训练后获得的报告仍然对框架敏感，并且未显示状态依赖性。结果是对称的，经过训练的否定与经过训练的肯定同样不可采纳。
cs.CL / 40 / 2609.29496
Evaluating Explanation-Driven Vision-Language Reasoning via Generation Order Interventions
通过生成顺序干预评估以解释为驱动的视觉-语言推理
Abstract
Natural language explanation generation serves as a key mechanism for exposing and evaluating vision-language reasoning. Prior work on explanation-driven vision-language models predominantly follows a post-hoc (answer-first) paradigm, implicitly suggesting that supervised rationales can reflect underlying reasoning processes. In contrast, modern large vision-language models increasingly exhibit a rationale-first generation tendency, which more closely aligns with structured, stepwise reasoning. In this work, we systematically evaluate whether explanations are causally tied to model predictions within a single generation step under a controlled experimental setup, explicitly eliminating unnecessary chain-of-thought or other intermediate reasoning processes across knowledge-intensive QA, visual entailment, and compositional grounding benchmarks. We find that larger models emerge as a prerequisite for reliably supporting rationale-first reasoning at scale. However, answer-first generation is less prone to format-related errors in structured output. Overall, explanation ordering, model scale and pre-training knowledge, task-specific fine-tuning, and task structure jointly influence both prediction accuracy and reasoning faithfulness.
Chinese Translation
自然语言解释生成是揭示和评估视觉-语言推理的关键机制。先前关于以解释为驱动的视觉-语言模型的研究主要遵循后验（答案优先）范式，隐含地表明监督性理由可以反映潜在的推理过程。相比之下，现代大型视觉-语言模型越来越表现出以理由优先的生成倾向，这与结构化的逐步推理更为一致。在本研究中，我们在受控实验设置下系统地评估了解释是否在单一生成步骤中与模型预测存在因果关系，明确消除了在知识密集型问答、视觉蕴涵和组合基础基准测试中不必要的思维链或其他中间推理过程。我们发现，较大的模型成为可靠支持大规模理由优先推理的前提。然而，答案优先生成在结构化输出中较少出现格式相关错误。总体而言，解释顺序、模型规模和预训练知识、任务特定微调以及任务结构共同影响预测准确性和推理的可信度。
cs.CL / 41 / 2609.29504
PROOF: Profiling Reliability of Object-Level Facts in Large Language Models
PROOF：大型语言模型中对象级事实的可靠性分析
Abstract
Aggregate factuality scores hide where a language model succeeds, which relations it confuses, and whether an answer survives innocuous changes to the question or decoder. We introduce PROOF, a profile-oriented benchmark for factual coverage in instruction-tuned language models. PROOF converts a frozen Wikidata snapshot into 18,486 English multiple-choice questions grounded in 11,779 semantic facts, 101 classes, 392 properties, and 14 domains. Each question has an explicit "I don't know" option, a "No correct option" control, and nine controlled formulations; 1,849 questions are no-correct-option traps. We evaluate 18 open-weight model deployments on 166,374 prompts each and separately perturb decoding on a fixed 10% subset. Base factual accuracy ranges from 6.58% to 57.59% (chance: 8.64%), yet every model has a 19.3-36.4 percentage-point spread across domains. Paired facts reveal direction-dependent retrieval, usually favoring subject-to-object queries, with the pattern reversing for one model. We find no consistent temporal penalty after exact-stratum adjustment. Neutral wording changes accuracy by as much as 26.5 percentage points, while adversarial formulations break up to 79.4% of answers that were initially correct. Direct switching to an injected false label varies from 0.04% to 27.5%, showing that accuracy loss and hint following are distinct. Selected-token confidence often indicates severe overconfidence, and decoder perturbations move accuracy by up to 15.7 percentage points and domain profiles by 16.8 points. PROOF therefore measures factual coverage as a structured, intervention-aware profile rather than a single claim about what a model "believes."
Chinese Translation
聚合的事实性评分掩盖了语言模型的成功之处、混淆的关系，以及答案在问题或解码器的无害变化下是否依然有效。我们引入了PROOF，这是一个面向分析的基准，用于评估指令调优语言模型的事实覆盖情况。PROOF将一个冻结的Wikidata快照转换为18,486个基于11,779个语义事实、101个类别、392个属性和14个领域的英语多项选择题。每个问题都有一个明确的“我不知道”选项，一个“没有正确选项”的控制项，以及九种受控表述；1,849个问题是没有正确选项的陷阱。我们在每个模型部署上评估18个开放权重模型，共166,374个提示，并单独对固定的10%子集进行解码扰动。基础事实准确率范围为6.58%至57.59%（随机机会：8.64%），但每个模型在不同领域的准确率差异在19.3至36.4个百分点之间。配对事实揭示了方向依赖的检索，通常偏向于主语到宾语的查询，但对于一个模型则出现了模式反转。我们发现，在精确层级调整后没有一致的时间惩罚。中性措辞的变化使准确率变化高达26.5个百分点，而对抗性表述使多达79.4%的最初正确答案被打破。直接切换到注入的错误标签的准确率变化范围为0.04%至27.5%，显示出准确率损失与提示跟随是不同的。选定标记的置信度通常表明严重的过度自信，而解码器扰动使准确率变化高达15.7个百分点，领域特征变化16.8个百分点。因此，PROOF作为一个结构化的、关注干预的分析工具来衡量事实覆盖，而不是关于模型“信念”的单一声明。
cs.CL / 42 / 2609.29507
What a Cross-Model Fixed-Point Census Can and Cannot Arbitrate About Repetition
跨模型不动点普查能与不能仲裁的重复问题
Abstract
Two accounts of neural text degeneration coexist. One locates the cause in the training data -- repetition in the corpus produces repetition in the output, established by training on repetition-sorted data -- the other in the trained network, in copying circuits and repetition features. Neither has been arbitrated across a broad cohort of pretrained models: the causal work trains its own. We report an observational measurement in a different currency: the fixed-point structure of a model's own short-window argmax map, censused from 96 random two-token starts over 17 off-the-shelf models, always unprompted -- a companion paper shows nine tokens of conditioning move this readout across most of its range. The four-way class is stable across census seeds on 17 of 17. Three exhibits. At fixed corpus (The Pile), fixed scale and that fixed domain, the class is not determined: across two size-matched tiers, pythia is a funnel while RWKV, Mamba and a second transformer family are not, and both hold their class across an order of magnitude of scale. Six of seven models in that ladder reach the same endpoint token, and those concentrating on it most strongly are among those that never stay there -- what varies is not where trajectories go but whether the destination self-continues. The deduplicated Pythia suite does not change the class. And the corpus-side inflow term proposed for this phenomenon does not select our endpoints once frequency is controlled, in English and three other languages. This is observational and cannot refute a training intervention. Funnels are common: eight of seventeen models, seven families, five corpora -- so the limit is not that the phenomenon is one model's peculiarity, but that within the one corpus where training data can be held fixed only one available family funnels; that subset cannot show the split is corpus-independent.
Chinese Translation
关于神经文本退化的两种解释并存。一种将原因归结于训练数据——语料库中的重复导致输出中的重复，这一观点通过在重复排序数据上进行训练得以确立；另一种则归因于训练后的网络，即复制电路和重复特征。在广泛的预训练模型中，这两种观点尚未得到仲裁：因果研究自成体系。我们报告了一项不同货币的观察性测量：模型自身短窗口 argmax 映射的不动点结构，从 17 个现成模型的 96 个随机双标记起始点进行普查，始终未被提示——一篇伴随论文显示，九个标记的条件化使得这一读出在其大部分范围内移动。四类在 17 个普查种子中保持稳定。三项展示。在固定语料（The Pile）、固定规模和固定领域下，该类并未被确定：在两个大小匹配的层次中，pythia 是一个漏斗，而 RWKV、Mamba 和第二个变换器家族则不是，并且两者在数量级变化中保持其类别。在该阶梯中的七个模型中，有六个达到相同的终点标记，而那些最强烈集中于此的模型恰恰是那些从未停留在此的模型——变化的不在于轨迹的去向，而在于目的地是否自我延续。去重的 Pythia 套件并未改变该类。而为这一现象提出的语料侧流入项在控制频率后并未选择我们的终点，无论是在英语还是其他三种语言中。这是观察性的，无法反驳训练干预。漏斗现象普遍存在：十七个模型中有八个，七个家族，五个语料库——因此限制并不在于这一现象是某个模型的特性，而在于在唯一一个可以固定训练数据的语料中，只有一个可用的家族呈现漏斗；该子集无法证明分裂是与语料无关的。
cs.CL / 43 / 2609.29511
EnSiTa - A Trilingual Multi-Domain Parallel Dataset and Benchmark for Domain-Specific Machine Translation
EnSiTa - 一个三语多领域平行数据集及领域特定机器翻译基准
Abstract
Machine Translation (MT) for low-resource languages remains far behind that of high-resource languages, and the gap is widest in specialised domains, where parallel data is scarce or entirely absent. We present EnSiTa, a trilingual multi-domain parallel dataset and benchmark for English, Sinhala and Tamil. EnSiTa provides human post-edited training data for seven domains, plus manually translated test sets for those and one additional domain, all produced by professional translators under a multi-year, rigorously quality-controlled process. Using this dataset, we conduct an extensive study of domain-specific MT for all six language directions, fine-tuning a from-scratch Transformer, a pre-trained translation model (NLLB-600M), and decoder-only LLMs (Gemma 3 family, 1B-12B, and TranslateGemma) across training-data sizes, model scales, and in-domain, cross-domain, multilingual and multi-domain settings. To the best of our knowledge, this is the most extensive systematically documented multi-domain parallel data creation and benchmarking effort for low-resource MT. Our data and models will be publicly released.
Chinese Translation
低资源语言的机器翻译（MT）远远落后于高资源语言，尤其是在专业领域，平行数据稀缺或完全缺失的情况下，差距最为明显。我们提出了EnSiTa，这是一个针对英语、僧伽罗语和泰米尔语的三语多领域平行数据集及基准。EnSiTa提供了七个领域的人类后编辑训练数据，以及这些领域和一个额外领域的人工翻译测试集，所有数据均由专业翻译人员在多年严格质量控制的过程中制作而成。利用该数据集，我们对所有六个语言方向的领域特定机器翻译进行了广泛研究，微调了从头开始的Transformer、预训练翻译模型（NLLB-600M）以及解码器专用的LLM（Gemma 3系列，1B-12B，以及TranslateGemma），涵盖了训练数据规模、模型规模，以及领域内、跨领域、多语言和多领域设置。根据我们所知，这是针对低资源机器翻译的最全面的系统化文档化多领域平行数据创建和基准测试工作。我们的数据和模型将公开发布。
cs.CL / 44 / 2609.29549
StepCOPS: Closed-Testing Lower-Tail Certificates for Language-Model Policy Selection
StepCOPS：用于语言模型策略选择的闭合测试下尾证书
Abstract
Post-training pipelines must select one language-model policy from many checkpoints, prompts, and decoding rules. Mean evaluator scores can conceal rare failures, whereas simultaneous candidate-wise confidence bounds can be unnecessarily conservative. We introduce StepCOPS, which uses an independent proposal split to nominate one lower-tail floor per candidate, exact binomial tests on a fresh certification split, and Holm's step-down procedure to certify a set of floors. With probability at least $1-\delta$, every certified floor, including the largest floor used for policy selection, is below its candidate's population lower $\alpha$-quantile. This guarantee assumes i.i.d. evaluation units while allowing arbitrary within-unit dependence across candidates. Across 24 predeclared configurations and 11 benchmarks, StepCOPS obtains 96.4% selected-policy coverage over 500 paired trials, raises the certified floor by 1.5 points over both proposal-Bonferroni and exact COPS, remains 0.6 points below the large-reference jury oracle, and abstains in 2.4% of trials. Shadow-judge, benchmark-native, artifact, and leave-one-judge-out audits characterize the proxy boundary: the guarantee applies to the fixed jury score, not directly to human safety.
Chinese Translation
后训练流程必须从多个检查点、提示和解码规则中选择一个语言模型策略。平均评估者得分可能掩盖罕见的失败，而同时候选者的置信区间可能过于保守。我们提出了StepCOPS，它使用独立的提案分割为每个候选者提名一个下尾底线，针对新的认证分割进行精确的二项检验，并采用Holm的逐步下降程序来认证一组底线。在概率至少为$1-eta$的情况下，每个认证的底线，包括用于策略选择的最大底线，都低于其候选者的总体下$eta$-分位数。该保证假设评估单位是独立同分布的，同时允许候选者之间的任意单位内依赖。在24个预先声明的配置和11个基准测试中，StepCOPS在500次配对试验中获得了96.4%的选定策略覆盖率，相比于提案-Bonferroni和精确COPS提高了1.5个点的认证底线，仍比大型参考陪审团oracle低0.6个点，并在2.4%的试验中选择不参与。影子评审、基准原生、工件和逐一评审者排除审计特征化了代理边界：该保证适用于固定的陪审团得分，而不是直接适用于人类安全。
cs.CL / 45 / 2609.29559
Benchmarking Arabic--Russian Machine Translation: A Comparison of Fine-tuned NMT and Few-shot LLMs under Rich Morphology and Low Lexical Overlap
阿拉伯语-俄语机器翻译基准测试：在丰富形态学和低词汇重叠下微调的神经机器翻译与少量样本大型语言模型的比较
Abstract
Arabic-Russian machine translation (MT) remains under-explored due to the rich morphology of Arabic and low lexical overlap between the two languages. We benchmark seven fine-tuned neural machine translation (NMT) models against four few-shot large language models (LLMs) on a 20k/5k/5k split of a new 15.47M-pair corpus. Fine-tuned NLLB-1.3B achieves the highest BLEU (16.3) and COMET (0.738). Aya-Expanse 8B leads the few-shot LLMs (BLEU 1.7 on 500 sentences, chrF 25.7), but all LLM scores remain far below the fine-tuned NMT baselines. Error analysis identifies low lexical overlap as the dominant failure mode; among the worst translations, mT5-small produces 32% too-short outputs. Bootstrap tests confirm significant differences among most models. Our results demonstrate that fine-tuned NMT significantly outperforms few-shot LLMs for Arabic-Russian translation under low-resource conditions.
Chinese Translation
阿拉伯语-俄语机器翻译（MT）由于阿拉伯语的丰富形态学和两种语言之间的低词汇重叠，仍然未得到充分探索。我们对七个微调的神经机器翻译（NMT）模型与四个少量样本大型语言模型（LLMs）进行了基准测试，使用一个新的1547万对语料库的20k/5k/5k划分。微调的NLLB-1.3B模型取得了最高的BLEU分数（16.3）和COMET分数（0.738）。Aya-Expanse 8B在少量样本LLMs中表现最佳（在500个句子上BLEU 1.7，chrF 25.7），但所有LLM的得分仍远低于微调的NMT基线。错误分析表明低词汇重叠是主要的失败模式；在最差的翻译中，mT5-small产生了32%的过短输出。自助测试确认大多数模型之间存在显著差异。我们的结果表明，在低资源条件下，微调的NMT显著优于少量样本LLMs在阿拉伯语-俄语翻译中的表现。
cs.CL / 46 / 2609.29573
ModularSQL: A Runtime Guardrail for the Multiplicity Blind Spot in Text-to-SQL
ModularSQL：文本到SQL中的多重性盲点的运行时保护措施
Abstract
Text-to-SQL systems are increasingly deployed on production databases, where queries that pass benchmark evaluation can still produce results that distort downstream workflows. Standard set-based execution accuracy (Set-EX) collapses duplicate rows and can therefore miss multiplicity errors, including missing DISTINCT, inflated aggregates, and Cartesian-style join explosions. We call this the Multiplicity Blind Spot (MBS) and introduce Multiset-EX, a multiplicity-preserving evaluation criterion that exposes such failures. Across released DeepEye-SQL artifacts from three backbones (Qwen2.5-Coder-32B, Qwen3-Coder-30B-A3B, and Gemma-3-27B) on executable BIRD-Dev N=1532, we find a consistent 5.81--6.79 pp gap between Set-EX and Multiset-EX. The gap is not specific to DeepEye-SQL: it persists on released DAIL-SQL+GPT-4 (5.22 pp) and BIRD GPT-3.5-turbo (3.39 pp) predictions. We further introduce ModularSQL, a lightweight post-selection runtime guardrail that probes executed results for multiplicity anomalies and applies deterministic patches or low-cost LLM rescue only to flagged queries. Integrated with DeepEye-SQL using Qwen3-Coder, ModularSQL preserves Set-EX at 72.06% while improving Multiset-EX from 65.86% to 67.75% (+1.89 pp). It flags 77 high-risk anomalies, while adding only $0.0076 in total LLM cost and 120 ms amortized latency per query. Cross-pipeline evaluation shows that the candidate-free detector and deterministic patches also transfer to independently released prediction sets. Overall, these results show that benchmark accuracy does not necessarily imply execution-safe SQL, and that lightweight, multiplicity-aware runtime guardrails can narrow this gap with modest computational overhead.
Chinese Translation
文本到SQL系统越来越多地部署在生产数据库上，在这些数据库中，即使通过基准评估的查询仍可能产生扭曲下游工作流程的结果。标准的基于集合的执行准确性（Set-EX）会合并重复行，因此可能会遗漏多重性错误，包括缺失 DISTINCT、膨胀的聚合和笛卡尔风格的连接爆炸。我们将其称为多重性盲点（Multiplicity Blind Spot, MBS），并引入多重集合执行准确性（Multiset-EX），这是一种保留多重性的评估标准，可以揭示此类失败。在来自三个基础模型（Qwen2.5-Coder-32B、Qwen3-Coder-30B-A3B和Gemma-3-27B）的发布的DeepEye-SQL工件中，在可执行的BIRD-Dev N=1532上，我们发现Set-EX与Multiset-EX之间存在一致的5.81至6.79个百分点的差距。该差距并非特定于DeepEye-SQL：在发布的DAIL-SQL+GPT-4（5.22个百分点）和BIRD GPT-3.5-turbo（3.39个百分点）预测中也持续存在。我们进一步引入ModularSQL，这是一种轻量级的后选择运行时保护措施，能够探测执行结果中的多重性异常，并仅对标记的查询应用确定性补丁或低成本的LLM救助。与使用Qwen3-Coder的DeepEye-SQL集成后，ModularSQL将Set-EX保持在72.06%，同时将Multiset-EX从65.86%提高到67.75%（+1.89个百分点）。它标记了77个高风险异常，同时每个查询仅增加$0.0076的LLM总成本和120毫秒的平均延迟。跨管道评估表明，无候选检测器和确定性补丁也能转移到独立发布的预测集。总体而言，这些结果表明，基准准确性并不一定意味着执行安全的SQL，而轻量级的、关注多重性的运行时保护措施可以在适度的计算开销下缩小这一差距。
cs.CL / 47 / 2609.29618
An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model
小型 MLA--SSM 混合语言模型的探索性消融研究
Abstract
We report an exploratory, single-seed ablation of TALH (Adaptive Latent Hybrid), a decoder-only language model with parallel Multi-head Latent Attention (MLA) and a custom recurrent state-space (SSM) branch. Five variants, spanning 117--217M estimated active parameters per token, are trained from scratch on a FineWeb sample for the same number of optimisation steps and tokens. In this specific setup, removing the SSM branch gives the largest degradation in validation perplexity (MLA-only PPL 315), whereas removing MLA has a much smaller effect (SSM-only PPL 239). A dense-FFN hybrid obtains PPL 231, compared with 240 for the tested top-2 ternary-MoE hybrid, while using 3.87 GB less peak training memory. We also preserve a preliminary Apple M3 timing observation: among the five unoptimised implementations, MLA-only has the flattest measured time-to-first-token curve from 512 to 2,048 prompt tokens, although the dense Transformer is much faster in absolute terms. Because the runs are single-seed, parameter counts are unmatched, the evaluation stream may overlap the training source, and raw repeated timing records are unavailable, these results support implementation-specific hypotheses rather than general conclusions about MLA, SSMs, or mixture-of-experts models.
Chinese Translation
我们报告了一项对 TALH（自适应潜在混合）进行的探索性单种子消融研究，TALH 是一种仅解码器的语言模型，具有并行多头潜在注意力（MLA）和自定义递归状态空间（SSM）分支。我们从头开始训练了五个变体，每个变体在优化步骤和标记数量上相同，估计的活跃参数范围为每个标记 117-217M。在这个特定设置中，去除 SSM 分支导致验证困惑度的最大降级（仅 MLA 的 PPL 为 315），而去除 MLA 的影响要小得多（仅 SSM 的 PPL 为 239）。与测试的前两名三元混合专家模型相比，密集前馈网络混合模型的 PPL 为 231，而后者为 240，同时使用的峰值训练内存减少了 3.87 GB。我们还保留了一项初步的 Apple M3 时间观察：在五个未优化的实现中，仅 MLA 的首次标记时间曲线在 512 到 2048 个提示标记之间最为平坦，尽管密集 Transformer 在绝对时间上要快得多。由于这些实验是单种子运行，参数计数不匹配，评估流可能与训练源重叠，且原始重复时间记录不可用，因此这些结果支持特定实现的假设，而不是关于 MLA、SSM 或混合专家模型的一般结论。
cs.CL / 48 / 2609.29633
TTLab at AlexandriaX-2026: A Fine-Tuned Surface Tagger for Arabic Machine-Translation Error-Span Detection and Classification
TTLab在AlexandriaX-2026：用于阿拉伯语机器翻译错误跨度检测与分类的精细调优表面标注器
Abstract
We present TTLab's submission to the AlexandriaX-2026 Subtask~3 on Arabic MT error span detection and classification. Our system frames the task as token-level classification over surface forms, preserving character offsets to ensure exact alignment with the evaluation metric. To handle severe label imbalance, we employ a focal loss with class weighting and dialect-specific decoding thresholds. Among six Arabic pre-trained encoders, MARBERTv2 achieves the best overall performance of 40.8 and 40.91 on the development and test set, respectively, ranking $\nth{3}$ out of all participating teams. While our system localizes error spans effectively, classification of rare error types remains challenging, highlighting the need for data augmentation for tail categories. The code is available at ${\href{https://github.com/ENTAILab/arabic-dialectal-mt-error-span-detection}{\faGithub~ TTLab at AlexandriaX-2026}$
Chinese Translation
我们在此介绍TTLab对AlexandriaX-2026子任务3的提交，该任务涉及阿拉伯语机器翻译错误跨度的检测与分类。我们的系统将该任务框架设定为对表面形式的标记级分类，保留字符偏移以确保与评估指标的精确对齐。为了解决严重的标签不平衡问题，我们采用了带类权重的焦点损失和方言特定的解码阈值。在六个阿拉伯语预训练编码器中，MARBERTv2在开发集和测试集上分别取得了40.8和40.91的最佳整体表现，在所有参与团队中排名第三。尽管我们的系统有效地定位了错误跨度，但对稀有错误类型的分类仍然具有挑战性，这突显了对尾部类别进行数据增强的必要性。代码可在${ ext{https://github.com/ENTAILab/arabic-dialectal-mt-error-span-detection}}$找到。
cs.CL / 49 / 2609.29636
Operator Packages, Proposer Strength, and Construction-Family Plateaus in Office-Scale Verified Search
操作符包、提议者强度与办公室规模验证搜索中的构建家族平台
Abstract
Verified search, in which a language model proposes programs, a hard evaluator scores them, and selection keeps the best, has recently moved mathematical records; controlled ablations of the proposer-side components remain rare. We instrument a minimal FunSearch-style loop at office scale (a 30B local model on a laptop, 120-600 verified samples per run) with three operator packages: a schematic notebook the model writes and carries instead of verbatim elites, a named obstacle, and behavioural repulsion from constructions already found. On nine construction problems from a public repository, the complete 2^3 factorial with two replicates favours the primary contrast in a nominal two-stage analysis: the composition closes more of the seed-to-record gap (+0.196; nominal pooled p=0.023, stage-combination p~0.08; median per-problem effect +0.045). Repulsion raises construction-hash diversity everywhere (p=0.0039; partly a manipulation check). The factorial finds no positive memory-by-repulsion interaction (bounded to about +/-0.04); the gain decomposes additively, and memory+repulsion is the only arm that never collapses (0 of 18 runs), within 0.025 of the full composition. A frontier proposer under the identical loop reaches in tens of samples what the local model does not in hundreds; in single scoping runs its gains arrive without the operators. The search stalls after closing ~92% of the gap on the flagship problem, and the registered family-hint test gives the stall its first reading: named in words, the reference family is adopted and loses; handed as code, it is optimized, but our best finite-grid implementation remains below the plateau reached unaided. The loop transported and optimized the idea it was handed; no unaided run produced it. We release the harness, every candidate, and the dated pre-registrations.
Chinese Translation
验证搜索是一种语言模型提出程序、硬评估器对其评分并选择最佳程序的过程，最近在数学记录方面取得了突破；对提议者侧组件的控制性消融仍然较为罕见。我们在办公室规模下（在笔记本电脑上使用30B本地模型，每次运行120-600个验证样本）构建了一个最小的FunSearch风格循环，并引入了三个操作符包：模型编写并携带的示意笔记本，而不是逐字的精英样本，一个命名障碍，以及对已找到构建的行为排斥。在来自公共库的九个构建问题上，完整的2^3因子实验设计与两个重复试验支持了名义上的两阶段分析中的主要对比：该组合缩小了种子到记录的差距（+0.196；名义合并p=0.023，阶段组合p~0.08；每个问题的中位效应+0.045）。排斥在各处提高了构建哈希的多样性（p=0.0039；部分为操控检查）。因子实验未发现正向的记忆与排斥交互作用（限制在约+/-0.04）；增益是可加性分解的，记忆+排斥是唯一一个从未崩溃的组合（18次运行中无崩溃），与完整组合相差不超过0.025。在相同循环下，前沿提议者在数十个样本中达到了本地模型在数百个样本中未能达到的效果；在单次范围运行中，其增益在没有操作符的情况下出现。在旗舰问题上，搜索在缩小约92%的差距后停滞，而注册的家族提示测试首次揭示了这一停滞：以文字命名的参考家族被采用并失败；作为代码提供时，它被优化，但我们最佳的有限网格实现仍低于未辅助达到的平台。该循环传递并优化了其接收到的想法；没有任何未辅助的运行产生它。我们释放了所有候选者和过时的预注册信息。
cs.CL / 50 / 2609.29657
How To Do Things With Prompts
如何使用提示进行交流
Abstract
When users address large language models, they produce directive speech acts whose pragmatic features differ from those of both everyday conversation and traditional human-computer interaction, and these features change as users gain familiarity with the systems they address. This paper applies speech act and politeness theory to a corpus-pragmatic analysis of 2,000 English-language prompts drawn from publicly shared ChatGPT conversations, 1,000 from 2023 and 1,000 from 2025, using the ShareChat dataset. Each prompt is annotated for illocutionary force, directness, propositional content, and the presence of politeness markers, and the distribution of these features is compared across the two sampling years. The results show a consistent movement toward indirect, implicit, and fragmentary realizations of directive force, accompanied by a decline in politeness marking. The largest single change, a shift of 14.9 percentage points, occurs in propositional content, where explicit specification of the requested action gives way to implicit reliance on the system's inferential capacity, suggesting that users have updated their model of what the system can recover from reduced input, treating it as a competent implicature resolver. Rather than asking whether LLMs "really" understand language, we should ask: what kind of language have we created in learning to speak to them?
Chinese Translation
当用户与大型语言模型互动时，他们产生的指令性言语行为的语用特征与日常对话和传统人机交互的特征有所不同，并且这些特征会随着用户对所使用系统的熟悉程度而变化。本文将言语行为理论和礼貌理论应用于对来自公开共享的 ChatGPT 对话的 2,000 个英语提示的语料库语用分析，其中 1,000 个来自 2023 年，1,000 个来自 2025 年，使用 ShareChat 数据集。每个提示都被标注了言外之意、直接性、命题内容和礼貌标记的存在，并比较了这两个采样年份中这些特征的分布。结果显示，指令性力量的间接性、隐含性和片段性实现呈现出一致的趋势，同时礼貌标记的使用有所下降。最大的单一变化发生在命题内容上，变化幅度为 14.9 个百分点，具体表现为请求行为的明确说明逐渐让位于对系统推理能力的隐含依赖，这表明用户已经更新了他们对系统从简化输入中恢复信息能力的模型，将其视为一个有能力的隐含义解析者。与其问大型语言模型（LLMs）是否“真正”理解语言，不如问：在学习与它们交流的过程中，我们创造了什么样的语言？
cs.CL / 51 / 2609.29672
LLMersion: A Local-First AI Agent Framework for Low-Cost Home Language Learning toward Educational Equity
LLMersion：一个以本地为先的低成本家庭语言学习人工智能代理框架，旨在促进教育公平
Abstract
Artificial intelligence helps education most where an essential provision has been rationed by cost. For language learners that provision is a teacher's voice, which binds listening, reading, speaking, and writing into one act. Published evidence shows why most learners lack it, from a global shortage of 44 million teachers to heavy household tutoring bills, and why technology has not substituted for it: computer-assisted language learning proved effective but narrow, applications presuppose connectivity 2.6 billion people lack, and One Laptop per Child's randomized evaluation found that hardware without capable software teaches nothing. We distill eight difficulties and four binding constraints, and argue that small open-weight models dissolve the last: a complete four-skill stack now fits a \$200-class laptop and, on community measurements, generates at the pace speech is consumed, for about one US cent of electricity per study hour. We therefore propose LLMersion, a scheme for AI for education that runs entirely at home, over the learner's own documents, with an AI-written, AI-understood, AI-updated codebase anyone can customize; present LLMersion-1, a released open-source prototype (https://github.com/QM378/LLMersion); and outline the vision of a private learning agent.
Chinese Translation
人工智能在教育中最能发挥作用的地方是那些因成本而受到限制的基本资源。对于语言学习者而言，这一资源就是教师的声音，它将听、读、说、写四项技能融为一体。已发布的证据显示，为什么大多数学习者缺乏这一资源，从全球缺少4400万教师到家庭辅导费用高昂，以及技术为何未能替代这一资源：计算机辅助语言学习虽然有效，但范围有限，应用程序假设的网络连接有26亿人无法获得，而“每个孩子一台笔记本电脑”的随机评估发现，缺乏有效软件的硬件无法提供任何教学。我们提炼出八个困难和四个约束条件，并认为小型开放权重模型可以解决最后一个问题：一个完整的四项技能堆栈现在可以适配一台200美元级别的笔记本电脑，并且根据社区测量，其生成速度与语音消费的速度相匹配，每学习小时的电力消耗约为1美分。因此，我们提出了LLMersion，一个完全在家运行的教育人工智能方案，基于学习者自己的文档，具有一个由人工智能编写、理解和更新的代码库，任何人都可以进行定制；并介绍了LLMersion-1，一个已发布的开源原型（https://github.com/QM378/LLMersion）；同时概述了一个私人学习代理的愿景。
cs.CL / 52 / 2609.29680
Confident but Wrong: A Constrained Decoding Diagnostic for Low-Resource Automatic Post-Editing
自信但错误：低资源自动后编辑的约束解码诊断
Abstract
Automatic Post-Editing (APE) for low-resource languages (LRLs) often fails to improve Machine Translation (MT), and the score alone cannot say why: whether more training would help, or whether the training data is too inconsistent to learn from. We introduce a black-box, inference-time diagnostic that tells these two cases apart without retraining or annotation. It varies an edit-distance penalty $\lambda$ that drives the model from free editing towards copying the MT, and reads two signals: (1) the shape of the Translation Edit Rate (TER)-vs-$\lambda$ curve, U-shaped if edits from the model reduce error and monotonically decreasing if none does; and (2) the ordering of constraint variants that trust model confidence to increasing degrees, which shows whether confidence tracks edit quality. Across decoder-only and encoder-decoder models on English-Sinhala, the diagnostic exposes two failure modes consistent with a heterogeneous post-edit signal as the underlying cause: Binary Collapse, where the model copies the MT or makes off-target edits, and Confident Miscalibration, where the confidence signals we test do not separate useful edits from unnecessary ones. The pattern holds on English-Marathi and English-Tamil, with the failure modes tracking the post-edit distribution rather than MT quality or language family. Beyond diagnosis, the curve shape prescribes a concrete next step for practitioners; in the favorable case, a static constraint yields a free inference-time accuracy gain. We release the first English-Sinhala (~66k) and a new English-Tamil (~39k) APE datasets with all code.
Chinese Translation
针对低资源语言（LRLs）的自动后编辑（APE）往往无法改善机器翻译（MT），而仅凭得分无法说明原因：是更多的训练会有所帮助，还是训练数据过于不一致以至于无法学习。我们引入了一种黑箱推理时诊断工具，能够在不重新训练或标注的情况下区分这两种情况。该工具通过调整编辑距离惩罚 $ ext{λ}$，使模型从自由编辑转向复制机器翻译，并读取两个信号：（1）翻译编辑率（TER）与 $ ext{λ}$ 曲线的形状，如果模型的编辑减少了错误则呈U形，如果没有则单调递减；（2）信任模型置信度的约束变体的排序，显示置信度是否与编辑质量相匹配。在针对英语-僧伽罗语的解码器仅模型和编码-解码模型中，该诊断揭示了两种故障模式，与异质后编辑信号作为潜在原因一致：二元崩溃（Binary Collapse），模型复制机器翻译或进行偏离目标的编辑；置信度误校准（Confident Miscalibration），我们测试的置信度信号未能区分有用的编辑与不必要的编辑。这种模式在英语-马拉地语和英语-泰米尔语中也得到了验证，故障模式与后编辑分布相关，而非机器翻译质量或语言家族。除了诊断之外，曲线形状为从业者提供了具体的下一步建议；在有利的情况下，静态约束会带来推理时的准确性提升。我们发布了首个英语-僧伽罗语（约66k）和新的英语-泰米尔语（约39k）APE数据集以及所有代码。
cs.CL / 53 / 2609.29682
Named Entity Recognition using Sliding Window Approach
基于滑动窗口方法的命名实体识别
Abstract
Named Entity Recognition (NER) is a core NLP task, but transformer-based sentence-level models struggle with long documents because of fixed input-length limits: truncation drops content, and non-overlapping chunking fragments entities at segment boundaries. We introduce an inference-only pipeline that extends a frozen NER model, MahaNER-BERT, fine-tuned on the MahaNER corpus, to document-level prediction via overlapping sliding windows that are merged into a single annotation, without any retraining or architectural change. We evaluate the pipeline on six document-level corpora built from the MahaNER test set using two strategies: Normal Repeat, which duplicates sentence sequences to extend length while preserving contextual continuity, and Random Repeat, which concatenates distinct sequences to produce longer, heterogeneous inputs, each instantiated at three length levels, across several sliding-window configurations. The model retains a macro F1-score of up to 0.8902, with variation staying below one percentage point regardless of document length or construction strategy. Compared with the conventional non-windowed approach, the sliding-window pipeline avoids the boundary-fragmentation errors introduced by non-overlapping segmentation, yielding consistently higher and more stable document-level F1-scores.
Chinese Translation
命名实体识别（NER）是自然语言处理（NLP）中的核心任务，但基于变换器的句子级模型在处理长文档时面临固定输入长度限制的问题：截断会丢失内容，而非重叠的分块会在段落边界处破坏实体。我们提出了一种仅用于推理的管道，该管道扩展了一个冻结的NER模型MahaNER-BERT，该模型经过在MahaNER语料库上的微调，能够通过重叠滑动窗口进行文档级预测，并将其合并为单一注释，而无需任何重新训练或架构更改。我们在六个基于MahaNER测试集构建的文档级语料库上评估了该管道，采用了两种策略：正常重复（Normal Repeat），该策略通过复制句子序列来延长长度，同时保持上下文的连续性；随机重复（Random Repeat），该策略通过连接不同的序列来生成更长的异质输入，每种策略在三个长度级别下实例化，涵盖了多种滑动窗口配置。该模型的宏观F1-score高达0.8902，且无论文档长度或构建策略如何，变化均保持在一个百分点以内。与传统的非窗口方法相比，滑动窗口管道避免了非重叠分割引入的边界碎片化错误，从而产生了一致更高且更稳定的文档级F1-score。
cs.CL / 54 / 2609.29684
DP-IPI: A Hybrid Differential Privacy Text Rewriting Mechanism for Indirect Personal Identifiers in Clinical Texts
DP-IPI：一种针对临床文本中间接个人标识符的混合差分隐私文本重写机制
Abstract
Despite the strengths of modern anonymization and de-identification techniques, the risk of re-identification remains significant due to the indirect identifiers remaining in texts. To address this problem, recent works have applied text rewriting under Differential Privacy (DP) to prevent data linkage by perturbing texts via noise addition. Such methods privatize all tokens in a text indiscriminately, diminishing text quality and usability in critical domains such as in clinical settings. Focusing on indirect personal identifiers (IPIs), we introduce a utility-preserving DP text rewriting method that only privatizes spans containing IPIs. We show that our method effectively reduces re-identification risks in clinical texts while being producing more coherent and usable output texts, leading to higher privacy-utility trade-offs. In this, we demonstrate the effectiveness of hybrid text privatization, which leverages the promise of DP in an efficient, usable manner.
Chinese Translation
尽管现代匿名化和去标识化技术具有一定的优势，但由于文本中仍然存在间接标识符，重新识别的风险仍然显著。为了解决这一问题，近期的研究应用了差分隐私（Differential Privacy, DP）下的文本重写，通过添加噪声来扰动文本，从而防止数据关联。这类方法对文本中的所有标记进行无差别的隐私处理，降低了文本在临床等关键领域的质量和可用性。针对间接个人标识符（Indirect Personal Identifiers, IPIs），我们提出了一种保持实用性的差分隐私文本重写方法，仅对包含IPIs的文本片段进行隐私处理。我们展示了该方法在降低临床文本中的重新识别风险的同时，能够生成更连贯和可用的输出文本，从而实现更高的隐私与实用性权衡。在此过程中，我们证明了混合文本隐私化的有效性，充分利用了差分隐私的优势，以高效且可用的方式进行处理。
cs.CL / 55 / 2609.29703
Stochastic Semantic Evidence Graphs: Uncertainty Propagation and Governance for Agentic AI
随机语义证据图：代理人工智能的不确定性传播与治理
Abstract
AI-agent evaluations usually inspect a final answer, yet error may enter through evidence, retrieval, prompting, generation or decision mapping. We introduce a stochastic semantic evidence graph (SSEG), a hierarchical stochastic DAG whose language node expands into an autoregressive token subgraph and whose observable output may be a law over complete phrases. Semantic reduction and calibration are optional. We define graph-relative local defects and downstream edge influences, derive a pathwise bound on terminal error and use its nodewise terms to diagnose governance triggers. For source provenance, the graph preserves uncertain claim--passage relations and propagates sharp Fr\'echet bounds rather than assuming independence across sources. Across three open-weight architectures, information-equivalent changes materially alter complete-phrase laws. A controlled experiment yields no certificate violations in 5,000 cases; crossed-RAG and live Brave-retrieval experiments separate retrieval, presentation, source and interaction effects. SSEG therefore turns workflow provenance into a quantitative account of where uncertainty entered, how it propagated and whether an output is qualified for use.
Chinese Translation
人工智能代理的评估通常只检查最终答案，但错误可能通过证据、检索、提示、生成或决策映射等途径进入。我们引入了一种随机语义证据图（SSEG），它是一个层次化的随机有向无环图（DAG），其语言节点扩展为自回归令牌子图，且其可观察输出可能是完整短语上的法律。语义简化和校准是可选的。我们定义了图相关的局部缺陷和下游边缘影响，推导出终端错误的路径界限，并利用其节点项来诊断治理触发因素。对于源的来源，图保持不确定的主张-段落关系，并传播尖锐的弗雷歇界限，而不是假设源之间的独立性。在三种开放权重架构中，信息等价的变化实质性地改变了完整短语法律。一个受控实验在5000个案例中未发现证书违规；交叉随机生成（crossed-RAG）和实时勇敢检索（live Brave-retrieval）实验分离了检索、呈现、来源和交互效应。因此，SSEG将工作流来源转化为不确定性进入、传播以及输出是否合格用于使用的定量描述。
cs.CL / 56 / 2609.29718
PPTBench: Can Coding Agents Reconstruct the Visual World through Structured, Editable Slides
PPTBench：编码智能体能否通过结构化、可编辑的幻灯片重建视觉世界
Abstract
Coding agents are beginning to act in the visual world. They now build webpages, GUIs, games, 3D scenes, diagrams, and documents. Success in such visual coding requires bridging two spaces: inferring visual structure and expressing it programmatically. Slides are a core medium of knowledge work, widely used to communicate ideas and collaborate in a form that people can directly inspect and edit. Therefore, they provide an ideal testbed for visual coding, as they require agents to recover visual structure and realize it as editable objects. However, existing benchmarks either rely on subjective open-ended evaluation, produce non-editable code outputs, or focus only on local editing rather than end-to-end visual reconstruction. We introduce PPTBench, which benchmarks visual coding through editable slide reconstruction. It contains 500 tasks, each based on a scientific flow diagram from a real arXiv paper and requiring agents to reconstruct it as a single PPTX page composed of native, editable objects. A four-stage Agentic Judge evaluates artifact validity, semantic correctness, rendering quality, and fine-grained visual quality. Across 31 configurations spanning model families, effort levels, and harnesses, the best configuration, Kimi K3, reaches only 67.80, while the median scores 19.47. We find that agents can reliably produce valid PPTX files but still struggle with semantic and visual correctness, especially text details. More reasoning mainly helps agents pass hard gates, while stronger verification is more consistently associated with higher quality. PPTBench advances the vision of coding agents that can understand and reconstruct the visual world through structured, editable code.
Chinese Translation
编码智能体开始在视觉世界中发挥作用。它们现在构建网页、图形用户界面（GUIs）、游戏、3D场景、图表和文档。在这种视觉编码中取得成功需要跨越两个领域：推断视觉结构并以编程方式表达它。幻灯片是知识工作的核心媒介，广泛用于以人们可以直接检查和编辑的形式传达思想和协作。因此，它们为视觉编码提供了理想的测试平台，因为它们要求智能体恢复视觉结构并将其实现为可编辑对象。然而，现有基准要么依赖于主观的开放式评估，要么产生不可编辑的代码输出，或者仅关注局部编辑而非端到端的视觉重建。我们引入了PPTBench，通过可编辑幻灯片重建来评估视觉编码。它包含500个任务，每个任务基于真实arXiv论文中的科学流程图，要求智能体将其重建为由本地可编辑对象组成的单个PPTX页面。一个四阶段的智能评估者评估工件的有效性、语义正确性、渲染质量和细致的视觉质量。在涵盖模型系列、努力水平和工具的31种配置中，最佳配置Kimi K3仅达到67.80，而中位数为19.47。我们发现，智能体可以可靠地产生有效的PPTX文件，但在语义和视觉正确性方面仍然存在困难，尤其是在文本细节方面。更多的推理主要帮助智能体通过困难的关卡，而更强的验证则更一致地与更高的质量相关。PPTBench推动了编码智能体能够通过结构化、可编辑的代码理解和重建视觉世界的愿景。
cs.CL / 57 / 2609.29733
TTLab at StanceEval-2026: A Cloze-Style Prompting Approach for Arabic-Language Stance Detection (CLASP-Ar)
TTLab在StanceEval-2026中的表现：一种阿拉伯语立场检测的填空式提示方法（CLASP-Ar）
Abstract
Arabic-language stance detection remains challenging, and previous shared-task systems have largely relied on multitask learning and ensembles. While these systems achieve state-of-the-art performance, their applicability and transferability are limited by the additional complexity introduced by multitask learning.To reduce this complexity, we introduce $\texttt{CLASP-Ar}$, which reformulates the task as cloze-style masked language modeling. In this approach, the target, predicted sentiment, and text are combined into a single prompt whose $\texttt{[MASK]}$ prediction is restricted to a verbalizer-constrained label vocabulary.
Chinese Translation
阿拉伯语立场检测仍然面临挑战，之前的共享任务系统主要依赖于多任务学习和集成方法。尽管这些系统达到了最先进的性能，但由于多任务学习引入的额外复杂性，它们的适用性和可转移性受到限制。为了减少这种复杂性，我们提出了$ exttt{CLASP-Ar}$，将任务重新表述为填空式的掩码语言建模。在这种方法中，目标、预测的情感和文本被组合成一个单一的提示，其$ exttt{[MASK]}$预测被限制在一个由动词化约束的标签词汇中。
cs.CL / 58 / 2609.29769
JEV vs. LLMs as Rubric Judges: Cheaper, Faster, and Wrong in the Same Places
JEV与LLMs作为评分评审的比较：在相同地方更便宜、更快且同样错误
Abstract
We ask whether Jev, a typed classifier that returns probabilities over permitted answers without generating text, can replace an LLM rubric judge. We compare it with three flash-tier LLM judges on nine panels drawn from seven benchmarks, giving every judge identical criterion texts. Jev's accuracy differs significantly from an LLM judge's in only 8 of 27 paired comparisons, ahead mostly on binary criteria and behind only on graded ones, and most of the other comparisons are inconclusive. Summed over the nine panels, the LLM judges, called once per criterion, cost 29 to 325 times as much as Jev and took 30 to 220 times as long. On graded criteria all four judges agree more with one another than with the labels and mostly assign lower levels than the raters. One of several observational accounts is that raters followed scale conventions our criterion texts omit. Jev's confidence ranks its own errors on most panels, which should make a cheap classifier the ideal first stage of a cascade that defers its uncertain verdicts to an LLM judge. Correlated errors undo that advantage. The LLM judges repeat nearly all of Jev's most confident errors, so a cascade replayed on the recorded verdicts lowers cost but gains at most 1.5 points over the best single judge with cross-fitted thresholds, and at most 2.0 even with oracle thresholds.
Chinese Translation
我们探讨了Jev这一类型分类器是否能够替代LLM评分评审，Jev返回关于允许答案的概率而不生成文本。我们将其与三种闪电级LLM评审在七个基准的九个小组中进行比较，给每位评审提供相同的标准文本。在27个配对比较中，Jev的准确性与LLM评审显著不同的仅有8个，主要在二元标准上领先，而在分级标准上落后，其他大多数比较结果则不具结论性。在九个小组的总和中，LLM评审每个标准调用一次的成本是Jev的29到325倍，耗时是Jev的30到220倍。在分级标准上，所有四位评审之间的意见一致性高于与标签的意见一致性，且大多数情况下分配的等级低于评分者。几种观察性解释之一是评分者遵循了我们的标准文本所省略的量表惯例。Jev在大多数小组中对自身错误的置信度进行排名，这使得廉价分类器成为级联的理想第一阶段，将其不确定的裁决推迟到LLM评审。相关错误削弱了这一优势。LLM评审几乎重复了Jev最自信的所有错误，因此在记录的裁决上重放的级联降低了成本，但在交叉拟合阈值下最多仅比最佳单一评审提高1.5分，即使在神谕阈值下也最多仅提高2.0分。
cs.CL / 59 / 2609.29792
TimeBraid: Unifying Time Series and Language for Understanding and Forecasting
TimeBraid：统一时间序列与语言以实现理解与预测
Abstract
We present TimeBraid, a series of unified time-series and language models that align pretrained language models and pretrained time-series foundation models through interleaved global residual attention layers. Each model inherits knowledge, instruction following, and reasoning from one side, continuous-signal perception and zero-shot forecasting from the other, and fuses the two in a shared representation space where both modalities are understood and generated. We study the design choices that make such unified modeling work: where to align the two representation spaces, how to ground language in temporal structure, how to balance understanding with generation, and how to keep joint optimization stable. The resulting recipe combines a unified prompting scheme for diverse time-series and text tasks, stabilized joint training, and supervision from 2.2M curated series--text pairs and 4.9M instruction-tuning samples. Across benchmarks spanning time-series perception, understanding, reasoning, and both context-aided and unimodal forecasting, TimeBraid remains competitive with far larger general-purpose models and task-specific counterparts.
Chinese Translation
我们提出了TimeBraid，这是一系列统一的时间序列与语言模型，通过交错的全局残差注意力层对预训练的语言模型和预训练的时间序列基础模型进行对齐。每个模型从一方继承知识、指令遵循和推理，从另一方继承连续信号感知和零-shot 预测，并在共享的表征空间中融合这两者，使得两种模态得以理解和生成。我们研究了使这种统一建模有效的设计选择：如何对齐两个表征空间，如何将语言与时间结构结合，如何平衡理解与生成，以及如何保持联合优化的稳定性。最终的方案结合了针对多样时间序列和文本任务的统一提示机制、稳定的联合训练，以及来自220万对策划的序列-文本对和490万条指令调优样本的监督。在涵盖时间序列感知、理解、推理以及上下文辅助和单模态预测的基准测试中，TimeBraid在与更大的一般性模型和特定任务模型的比较中保持了竞争力。
cs.CL / 60 / 2609.29798
Benchmarking and Domain Adaptation of Automatic Speech Recognition (ASR) for Adolescent Health Communication in Ghanaian Languages
加纳语言中青少年健康沟通的自动语音识别（ASR）基准测试与领域适应
Abstract
This paper presents an end-to-end study of automatic speech recognition (ASR) for adolescent health communication in three Ghanaian languages (Twi, Dagbani, and Ewe). The work proceeds in three connected stages; First, we benchmark five ASR systems (three language-specific Wav2Vec2 models and two multimodal LLMs, Gemma 3n and Gemma 4) on a general-domain Bible corpus and a Youth Adolescent Sexual and Reproductive Health (ASRH) Domain ASR dataset, using Character and Word Error Rate (CER, WER). Second, guided by the benchmark, we perform supervised domain adaptation: although Gemma 4 was the strongest zero-shot candidate, fine-tuning it proved computationally infeasible, so we pivoted to the compact Qwen3-ASR-0.6B, fine-tuned on a large Ghana Bible corpus (~90k samples) and evaluated strictly on held-out human-collected in-domain audio. Fine-tuning reduced WER on every language, most dramatically for Ewe (WER from 109.3% to 64.8%, a drop of 44.5 pp; CER from 65.1% to 24.9%). Third, we validate the work through KasaHealth, a live voice-first ASRH application deployed in all three languages, complemented by Senti-Check, a technical evaluation harness. KasaHealth was tested by 50 community respondents and achieved a 100% chat-approval rate, a 72% Good-or-Excellent translation rating, and a 92% would-recommend rate, while surfacing the domain gaps that most constrain real-world use. Across all three stages the evidence converges: for these languages the binding constraint is validated in-domain data, not model capability or computation.
Chinese Translation
本文呈现了针对加纳三种语言（Twi、Dagbani和Ewe）中青少年健康沟通的自动语音识别（ASR）的端到端研究。该研究分为三个相互关联的阶段；首先，我们在一个通用领域的圣经语料库和一个青少年性与生殖健康（ASRH）领域的ASR数据集上，对五个ASR系统（三个特定语言的Wav2Vec2模型和两个多模态大语言模型Gemma 3n和Gemma 4）进行了基准测试，使用字符错误率（CER）和单词错误率（WER）。其次，在基准测试的指导下，我们进行了监督领域适应：尽管Gemma 4是最强的零样本候选模型，但对其进行微调在计算上不可行，因此我们转向紧凑型的Qwen3-ASR-0.6B，在一个大型加纳圣经语料库（约9万样本）上进行了微调，并严格在保留的人类收集的领域内音频上进行了评估。微调降低了每种语言的WER，尤其是Ewe语言的降幅最为显著（WER从109.3%降至64.8%，下降44.5个百分点；CER从65.1%降至24.9%）。第三，我们通过KasaHealth验证了该工作，这是一个在三种语言中部署的实时语音优先ASRH应用程序，并辅以Senti-Check技术评估工具。KasaHealth经过50名社区受访者的测试，达到了100%的聊天批准率、72%的良好或优秀翻译评级和92%的推荐率，同时揭示了限制实际使用的领域差距。在所有三个阶段中，证据趋于一致：对于这些语言，限制因素是经过验证的领域内数据，而非模型能力或计算能力。
cs.CL / 61 / 2609.29800
Adaptive Fisher-Whitened Cross-Covariance for Low-Resource Speech Recognition
低资源语音识别的自适应费舍尔白化交叉协方差
Abstract
Adapting multilingual speech foundation models to low-resource languages remains difficult, especially for languages that are poorly represented during pre-training. While parameter-efficient fine-tuning (PEFT) reduces the cost of adapting large models, conventional approaches such as LoRA rely on generic low-rank parameterizations and do not explicitly use downstream task information to define the adaptation subspace. To investigate whether task-informed PEFT can better support low-resource ASR, we apply Fisher-Whitened Cross-Covariance Analysis (FCCA) to Whisper and Qwen3-ASR, and introduce two complementary extensions: Asymmetric-Coupled FCCA (AC-FCCA), which exploits structured cross-layer sharing, and Adaptive-Rank FCCA (AR-FCCA), which reallocates adaptation capacity across projection matrices under a fixed parameter budget. Under controlled multilingual experiments, we evaluate these approaches on languages that are poorly represented or unsupported during pre-training alongside well-represented languages. Standard FCCA is competitive with, and usually outperforms, trainable-parameter-budget-matched LoRA. AR-FCCA provides the most consistent improvement over standard FCCA across both model architectures, with statistically significant gains in several evaluation settings, while retaining the same number of trainable parameters. These results show that task-informed subspace construction can be effective for low-resource speech adaptation, and that adaptive rank allocation provides a robust way to improve parameter efficiency without increasing model capacity.
Chinese Translation
将多语言语音基础模型适应于低资源语言仍然具有挑战性，尤其是对于在预训练过程中代表性不足的语言。虽然参数高效微调（PEFT）降低了适应大型模型的成本，但传统方法如 LoRA 依赖于通用的低秩参数化，并未明确利用下游任务信息来定义适应子空间。为探讨任务信息驱动的 PEFT 是否能更好地支持低资源自动语音识别（ASR），我们将费舍尔白化交叉协方差分析（FCCA）应用于 Whisper 和 Qwen3-ASR，并引入两个互补的扩展：非对称耦合 FCCA（AC-FCCA），利用结构化的跨层共享，以及自适应秩 FCCA（AR-FCCA），在固定参数预算下重新分配适应能力到投影矩阵。在控制的多语言实验中，我们评估了这些方法在预训练期间代表性不足或不支持的语言以及代表性良好的语言上的表现。标准 FCCA 与可训练参数预算匹配的 LoRA 竞争力相当，通常表现更优。AR-FCCA 在两种模型架构上均提供了对标准 FCCA 的最一致改进，在多个评估设置中均有统计显著的提升，同时保持相同数量的可训练参数。这些结果表明，任务信息驱动的子空间构建对于低资源语音适应是有效的，自适应秩分配提供了一种在不增加模型容量的情况下提高参数效率的稳健方法。
cs.CL / 62 / 2609.29807
CORDIAL: Calibrating Ordinal LLM Outputs from Few Labels
CORDIAL：从少量标签校准有序大型语言模型输出
Abstract
A large language model (LLM) can turn a text into a distribution over an ordered scale, but that distribution is a noisy measurement: saturated, compressed or exaggerated, and biased in a consistent direction. We propose CORDIAL, which treats the model's output as a noisy reading of the true label and corrects it with a channel of five interpretable parameters. The channel is small enough for its posterior to be averaged from a handful of labels, and we prove that the resulting calibration preserves first-order stochastic order. On Amazon reviews and CMU-MOSEI transcripts with four LLMs, CORDIAL has the lowest log loss among nine calibrators in 76 of 80 settings with 5 to 100 labels; with 20 labels and the main 7B reader, it matches the strongest baseline using 28-54 labels. The same posterior lets us learn priors from other tasks and fuse several LLMs. Unrestricted calibrators such as Dirichlet calibration overtake it only as the calibration set grows into the hundreds or thousands.
Chinese Translation
大型语言模型（LLM）可以将文本转换为有序尺度上的分布，但该分布是一种噪声测量：饱和、压缩或夸大，并且在一致的方向上存在偏差。我们提出了CORDIAL，它将模型的输出视为真实标签的噪声读数，并通过五个可解释参数的通道进行修正。该通道足够小，可以从少量标签中平均其后验分布，我们证明了所得到的校准保持了一阶随机顺序。在亚马逊评论和CMU-MOSEI转录本上，使用四个LLM的CORDIAL在80个设置中有76个的对数损失低于九个校准器；在使用20个标签和主要的7B阅读器时，它与使用28-54个标签的最强基线相匹配。相同的后验分布使我们能够从其他任务中学习先验，并融合多个LLM。只有当校准集增长到数百或数千时，无限制的校准器（如Dirichlet校准）才会超越它。
cs.CL / 63 / 2609.29828
ChunkRank: Model-Aware Text Chunking and Abstention-Aware Answer Selection for LLM Pipelines
ChunkRank：面向模型的文本块划分与考虑放弃的答案选择用于大语言模型管道
Abstract
We present ChunkRank, an open-source Python library that derives chunk boundaries from a target model's tokenizer and context window, and selects an answer among candidates produced independently per chunk. It ships a validated registry of 90 models across 15 providers and six answer-selection methods, and needs only three core dependencies. For chunking, ChunkRank avoids context-window overflow automatically from the model name, whereas character-based splitters overflow or waste the budget, and a fidelity study across 11 languages shows why token-exact budgets matter beyond English. For answer selection we report a negative result: on NaturalQuestions, TriviaQA and HotpotQA, with extractive and generative readers, no content-based ranker reliably beats taking the first non-empty answer. The reason is reader abstention on chunks that lack the answer, not answer position. A long-context baseline shows that chunking matches single-call reading on single-hop questions, so ChunkRank targets small-window and beyond-window settings. Code, registry and evaluation harness are released.
Chinese Translation
我们提出了ChunkRank，一个开源的Python库，它从目标模型的分词器和上下文窗口中推导出块边界，并在每个块中独立生成的候选答案中选择一个答案。该库提供了来自15个提供者的90个模型的经过验证的注册表和六种答案选择方法，仅需三个核心依赖项。对于块划分，ChunkRank自动避免了来自模型名称的上下文窗口溢出，而基于字符的分割器则可能导致溢出或浪费预算，一项跨11种语言的保真度研究表明，为什么超出英语的标记精确预算是重要的。在答案选择方面，我们报告了一个负面结果：在NaturalQuestions、TriviaQA和HotpotQA上，使用提取式和生成式阅读器，没有任何基于内容的排名器能够可靠地超过选择第一个非空答案。原因在于阅读器在缺乏答案的块上选择放弃，而不是答案的位置。一个长上下文基线表明，块划分在单跳问题上与单次调用阅读相匹配，因此ChunkRank旨在针对小窗口和超出窗口的设置。代码、注册表和评估工具已发布。
cs.CL / 64 / 2609.29845
Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs
你的变压器可以同时容纳两个思想：大型语言模型中的线性叠加证据
Abstract
While Large Language Models (LLMs) rely on highly non-linear components, in this work we demonstrate that they exhibit fundamental linearity: when inputs from distinct text streams are linearly combined, the model outputs a superposition of the individual next-token distributions. We term this the \textit{Superposition Linearity Hypothesis}. We provide evidence that superposition is an intrinsic property of the Transformer architecture rather than an emergent consequence of training; in fact, we observe that it tends to diminish as pretraining progresses. However, we demonstrate that linearity can be substantially restored through lightweight fine-tuning, significantly reducing the divergence between the predicted next-token distribution and the average of the individual next-token distributions. Finally, we introduce a guided decoding procedure that disentangles superposed outputs, enabling the simultaneous generation of two coherent continuations from a single forward pass.
Chinese Translation
尽管大型语言模型（LLMs）依赖于高度非线性的组件，但在本研究中，我们证明它们表现出基本的线性特征：当来自不同文本流的输入线性组合时，模型输出的是各个下一个标记分布的叠加。我们将其称为 extit{叠加线性假设}。我们提供证据表明，叠加是变压器架构的内在属性，而不是训练的涌现结果；事实上，我们观察到，随着预训练的进行，这种特性往往会减弱。然而，我们证明通过轻量级微调可以显著恢复线性，显著减少预测的下一个标记分布与各个下一个标记分布平均值之间的差异。最后，我们引入了一种引导解码程序，能够解开叠加输出，从而在单次前向传播中同时生成两个连贯的延续。
cs.CL / 65 / 2609.29848
Encoded but Not Decoded: Layer-Localized Evidence for a Three-Level Gap in LLM Syntax
编码但未解码：对大型语言模型语法中三层差距的层局部证据
Abstract
A language model can fail a syntactic test in two distinct ways: by not encoding the relevant structure, or by encoding it but failing to use it at the output. Behavioral evaluation alone cannot tell these apart. We propose a three-level evaluation framework (behavioral deployment, LM-head readout, and probe recoverability) measured on the same items under the same binary decision. Using a compact trilingual (English, Chinese, German) control-dependency benchmark, we find that probe recoverability exceeds or equals LM-head readout, which in turn exceeds or equals behavioral deployment, across seven models and all three languages in the aggregate. The recoverability surplus is never negative across all 14 (model, task) conditions. The disconnect concentrates in subject-control, where a nearest-noun heuristic gives the wrong answer. The single largest gap (0.653) appears on Qwen3-0.6B Instruct in question answering. The gap persists at Qwen3-14B Instruct. Instruction tuning degrades deployment more than encoding in percentage terms. We rule out option-position bias, late-layer erasure, output-formatting artifacts, and probe-training variance. The pattern is consistent with decoding that favors surface shortcuts, and the behavior-probe gap measures the strength of that preference. Activation patching shows the gap is layer-localized. Under instruction tuning, the LM-head-decoded layer shifts approximately ten layers later than the probe-decoded layer. These findings argue that behavioral evaluation understates what models encode, while probing alone overstates what they deploy.
Chinese Translation
语言模型在语法测试中可能以两种不同方式失败：要么未编码相关结构，要么已编码但未在输出中使用。仅通过行为评估无法区分这两者。我们提出了一个三层评估框架（行为部署、LM-head读出和探测可恢复性），在相同的项目和相同的二元决策下进行测量。使用一个紧凑的三语（英语、中文、德语）控制依赖基准，我们发现探测可恢复性超过或等于LM-head读出，而后者又超过或等于行为部署，这一模式在七个模型和所有三种语言的总体中均成立。在所有14个（模型，任务）条件下，可恢复性盈余从未为负。差距主要集中在主语控制中，最近名词启发式给出了错误答案。在问答任务中，最大的差距（0.653）出现在Qwen3-0.6B Instruct上。该差距在Qwen3-14B Instruct中仍然存在。指令调优在百分比上对部署的影响大于对编码的影响。我们排除了选项位置偏差、晚层擦除、输出格式伪影和探测训练方差。该模式与偏向表面捷径的解码一致，行为-探测差距衡量了这种偏好的强度。激活补丁显示该差距是层局部化的。在指令调优下，LM-head解码层大约比探测解码层晚十层。这些发现表明，行为评估低估了模型所编码的内容，而单独的探测则高估了它们的部署。
cs.CL / 66 / 2609.29855
Multi-Task Learning by using Contextualized Word Representations for Syntactic Parsing of a Morphologically Rich Language
利用上下文化词表示进行多任务学习以实现形态丰富语言的句法解析
Abstract
We address the challenge of syntactic parsing for Urdu, a morphologically rich language, and present state-of-the-art results for both constituency and dependency parsing. This paper offers four major contributions: 1) the conversion of the CLE-UTB phrase structure treebank into a dependency treebank by developing language-specific head-word and phrase-to-dependency label mapping rules; 2) a novel sequence labeling scheme that transforms the parsing task into a unified representation; 3) the training of contextualized word representations on a large 220 million tokens Urdu corpus collected from the web; and 4) development of parsing framework using two learning paradigms, single-task and multi-task learning. Several post-processing rules are applied to improve the quality of the automatically converted dependency structure treebank. The proposed sequence labeling scheme enables the use of a shared architecture that learns the syntactic structures from both grammatical structures simultaneously and hence improves generalization. Experiments show that the multi-task learning setup significantly enhances parsing performance, achieving an F1 score of 91.39 for constituency parsing (an improvement of 3.29 points) and a labeled attachment score of 85.69 for dependency parsing (an improvement of 1.49 points). These results demonstrate that learning cross-task representations provides measurable benefits and advances the state of syntactic parsing for Urdu.
Chinese Translation
我们解决了乌尔都语这一形态丰富语言的句法解析挑战，并为成分解析和依存解析提供了最先进的结果。本文的主要贡献有四点：1）通过开发特定于语言的主词和短语到依存标签的映射规则，将CLE-UTB短语结构树库转换为依存树库；2）提出了一种新颖的序列标注方案，将解析任务转化为统一表示；3）在从网络收集的2.2亿个乌尔都语语料库上训练上下文化词表示；4）开发了使用单任务学习和多任务学习两种学习范式的解析框架。应用了若干后处理规则以提高自动转换的依存结构树库的质量。所提出的序列标注方案使得可以使用共享架构，同时从两种语法结构中学习句法结构，从而提高了泛化能力。实验表明，多任务学习设置显著提升了解析性能，成分解析的F1分数达到91.39（提高了3.29分），依存解析的标注附加分数为85.69（提高了1.49分）。这些结果表明，学习跨任务表示提供了可衡量的好处，并推动了乌尔都语句法解析的进展。
cs.CL / 67 / 2609.29913
MILO: Efficient Many-shot In-Context Learning with Block-wise Low-rank Compression
MILO：基于块级低秩压缩的高效多示例上下文学习
Abstract
Many-shot in-context learning (ICL) enables large language models (LLMs) to adapt to complex tasks by conditioning on thousands of demonstration examples, but this paradigm shifts the inference efficiency bottleneck to the key-value (KV) cache memory. Due to the linear scaling behavior of the KV cache, storing these intermediate tensors has become a paramount challenge for both online serving and on-device deployment. To address this issue, we propose a novel compression framework, termed MILO, that exploits the low-rank redundancy inherent in many-shot contexts. Specifically, MILO features a block-wise low-rank compression strategy that compresses the KV cache at the block granularity, where each block contains multiple many-shot examples. Furthermore, to handle the heterogeneous context density across different blocks, MILO dynamically allocates rank budgets based on the information entropy, preserving the fidelity of critical blocks while aggressively compressing redundant ones. Experimental results on Qwen2.5 models demonstrate that our method achieves up to 50% reduction in KV cache memory and 1.8x throughput improvement, with negligible performance degradation on classification and reasoning benchmarks, significantly outperforming prior baselines.
Chinese Translation
多示例上下文学习（ICL）使大型语言模型（LLMs）能够通过依赖数千个示例来适应复杂任务，但这一范式将推理效率瓶颈转移到了键值（KV）缓存内存。由于KV缓存的线性扩展行为，存储这些中间张量已成为在线服务和设备端部署的主要挑战。为了解决这一问题，我们提出了一种新颖的压缩框架，称为MILO，该框架利用多示例上下文中固有的低秩冗余。具体而言，MILO采用了一种块级低秩压缩策略，在块粒度上压缩KV缓存，其中每个块包含多个多示例示例。此外，为了处理不同块之间的异构上下文密度，MILO根据信息熵动态分配秩预算，保留关键块的保真度，同时对冗余块进行积极压缩。在Qwen2.5模型上的实验结果表明，我们的方法在KV缓存内存上实现了高达50%的减少，并且吞吐量提高了1.8倍，同时在分类和推理基准测试中性能下降极小，显著优于之前的基线。
cs.CL / 68 / 2609.29928
Cultural Divergence Preservation: Diagnosing Flattening and Caricature in LLM-Simulated Survey Populations
文化差异保留：诊断大型语言模型模拟调查人群中的平面化与夸张
Abstract
Large language models (LLMs) are increasingly used as synthetic survey respondents to estimate population response distributions. In cross-cultural survey simulation, evaluations should assess not only distributional fidelity within countries but also whether differences across countries are preserved. However, existing distance-based metrics such as Jensen--Shannon divergence (JSD) do not directly capture such cross-country differences. To address this limitation, we introduce Cultural Divergence Preservation (CDP), a reference-light diagnostic based on a one-time human calibration. CDP identifies reduced cross-country divergence as cultural flattening and increased divergence as cultural caricature. To evaluate CDP, we conduct experiments across four LLM backbones, three persona-based prompting methods, and two survey domains, the World Values Survey (WVS) and the Big Five Personality Test. The results reveal a systematic discrepancy between conventional fidelity metrics and CDP. Controlled experiments show that CDP changes monotonically as cross-country divergence is attenuated or amplified, while the corresponding changes in JSD remain relatively small. In our audit of real LLM generations, DeepPersona-Inspired prompting is frequently favored by conventional fidelity metrics but exhibits the strongest flattening in every model--domain block. CDP thus complements fidelity metrics by directly quantifying the attenuation or amplification of cross-country divergence.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作合成调查响应者，以估计人群响应分布。在跨文化调查模拟中，评估不仅应考虑各国内部的分布忠实度，还应评估各国之间的差异是否得以保留。然而，现有的基于距离的度量，如詹森-香农散度（JSD），并不能直接捕捉这种跨国差异。为了解决这一局限性，我们引入了文化差异保留（Cultural Divergence Preservation, CDP），这是一种基于一次性人工校准的参考轻量级诊断工具。CDP将减少的跨国差异识别为文化平面化，而将增加的差异识别为文化夸张。为了评估CDP，我们在四种LLM基础模型、三种基于角色的提示方法和两个调查领域（世界价值观调查（WVS）和五大人格测试）中进行了实验。结果揭示了传统忠实度度量与CDP之间的系统性差异。对照实验表明，随着跨国差异的减弱或增强，CDP单调变化，而JSD的相应变化则相对较小。在我们对真实LLM生成内容的审计中，深度角色启发的提示方法常常受到传统忠实度度量的青睐，但在每个模型-领域块中表现出最强的平面化。因此，CDP通过直接量化跨国差异的减弱或增强，补充了忠实度度量。
cs.CL / 69 / 2609.29933
An Empirical Study of VLM Pipelines for Long-Document QA
长文档问答的视觉语言模型管道的实证研究
Abstract
Vision-Language Models (VLMs) are increasingly used for long-document processing, where the inputs combine text with charts, tables, figures, and complex layouts. Deploying them means choosing how to feed the document to the model, which retriever to use when only a subset of pages is sent, and whether to run the model agentically or as a static pipeline. We study these choices on two long-document QA benchmarks with both frontier API and open-weight VLMs. First, on MMLongBench-Doc our six-tool agent with page, table, figure, and search calls pays off only once the answering VLM is large enough: with Qwen3.5-4B and 9B it trails static page input, with Qwen3.5-27B it draws level, and with Sonnet 4.5 it leads. On LongDocURL it is level with or ahead of static input at every reader. Its lead over the strongest static pipeline is clearest with the frontier reader on MMLongBench-Doc and narrows to within noise on LongDocURL. Second, retrieval modality matters more than the specific retriever: the strongest image retriever leads the strongest text pipeline, and on the text side a single off-the-shelf cross-encoder rerank essentially matches a much heavier multi-stage LLM pipeline. Top-k image retrieval is also the most token-efficient input at every reader we paired it with, at roughly a seventh to a quarter of the tokens of sending every page. Third, cutting across all three choices, three of our strongest pipelines succeed on different questions, and an oracle that picks the best pipeline per question gains roughly thirteen points over the best single pipeline, though evidence-type routing recovers almost none of it.
Chinese Translation
视觉语言模型（VLMs）在长文档处理中的应用日益增加，这些输入结合了文本、图表、表格、图形和复杂布局。部署这些模型意味着需要选择如何将文档输入到模型中、在仅发送部分页面时使用哪个检索器，以及是以代理方式运行模型还是作为静态管道运行。我们在两个长文档问答基准上研究了这些选择，涉及前沿API和开放权重的VLMs。首先，在MMLongBench-Doc上，我们的六工具代理在进行页面、表格、图形和搜索调用时，仅在回答的VLM足够大时才会取得成效：使用Qwen3.5-4B和9B时，其表现落后于静态页面输入，使用Qwen3.5-27B时表现持平，而使用Sonnet 4.5时则领先。在LongDocURL上，它在每个阅读器中与静态输入持平或领先。它在MMLongBench-Doc上对最强静态管道的领先最为明显，而在LongDocURL上则缩小到噪声范围内。其次，检索方式比具体的检索器更为重要：最强的图像检索器优于最强的文本管道，而在文本方面，单个现成的交叉编码器重排基本上与一个更重的多阶段LLM管道相匹配。Top-k图像检索在我们搭配的每个阅读器中也是最具令牌效率的输入，发送每一页的令牌数量大约是其七分之一到四分之一。第三，跨越所有三个选择，我们的三条最强管道在不同问题上成功，而一个为每个问题选择最佳管道的神谕大约比最佳单一管道高出十三分，尽管证据类型路由几乎没有恢复到这一点。
cs.CL / 70 / 2609.30005
VietPrism: A large-scale Vietnamese speech and deepfake corpus with diverse dialects and code-switching
VietPrism：一个大规模的越南语语音和深度伪造语料库，涵盖多样方言和代码切换
Abstract
Vietnamese speech research is constrained by resources that isolate automatic speech recognition from speaker, dialect, code-switching, and deepfake analysis. We introduce VietPrism, an open, multi-domain corpus that brings these dimensions together at scale: 993.4 hours and 403,941 bona fide utterances from 1,262 verified speakers across 8,388 real-world videos. To our knowledge, it is the first large-scale Vietnamese corpus to jointly provide transcripts, consistent speaker identities, five dialect groups, and naturally occurring Vietnamese--English code-switching, which constitutes nearly half of the corpus by duration. We further create over 3.1K hours of spoof speech with four open-source and commercial synthesis systems. Every spoof is conditioned on a verified speaker reference and paired with a transcript- and speaker-matched bona fide utterance, enabling unique controlled evaluation with reduced lexical and identity confounds. Zero-shot evaluation of five pretrained multilingual detectors reveals striking brittleness: EER greatly varies across detector--generator pairings, while recent multilingual detector DFA-1B degrades from 16.3% to 33.6% as speaker similarity increases. Dialect-stratified results expose further model-dependent disparities. By unifying natural linguistic diversity with controlled spoof generation, VietPrism provides a challenging foundation for Vietnamese speech modeling and trustworthy audio-deepfake detection.
Chinese Translation
越南语语音研究受到资源的限制，这些资源将自动语音识别与说话者、方言、代码切换和深度伪造分析隔离开来。我们介绍了VietPrism，这是一个开放的多领域语料库，将这些维度在规模上结合在一起：来自1,262名经过验证的说话者的993.4小时和403,941条真实发言，涵盖8,388个真实世界的视频。据我们所知，这是第一个大规模的越南语语料库，联合提供转录文本、一致的说话者身份、五个方言组以及自然发生的越南语-英语代码切换，后者在时长上占据了近一半的语料库。我们进一步创建了超过3.1K小时的伪造语音，使用四个开源和商业合成系统。每个伪造语音都是基于经过验证的说话者参考，并与转录文本和说话者匹配的真实发言配对，从而实现独特的控制评估，减少词汇和身份的干扰。对五个预训练的多语言检测器进行零样本评估显示出显著的脆弱性：EER在检测器-生成器配对之间变化很大，而最近的多语言检测器DFA-1B在说话者相似性增加时从16.3%降至33.6%。方言分层结果揭示了进一步依赖模型的差异。通过将自然语言多样性与受控伪造生成相结合，VietPrism为越南语语音建模和可信的音频深度伪造检测提供了一个具有挑战性的基础。
cs.CL / 71 / 2609.30009
Automated Regulatory Compliance Question Answering in Financial Services with Domain-Adapted Retrieval-Augmented Generation
金融服务中的自动化合规性问答：基于领域适应的检索增强生成
Abstract
Financial institutions operate under dense, frequently amended rulebooks, and answering a compliance question correctly requires not only fluency but verifiable grounding in the authoritative text. Large language models are attractive for this task, yet the models that firms can realistically deploy on-premise are compact ones, and compact models hallucinate obligations. We study whether a carefully domain-adapted retrieval-augmented generation pipeline closes that gap. Our retriever is built in three stages on top of LegalBERT: entailment tuning that recasts question--passage matching as premise--hypothesis reconstruction, contrastive tuning with in-batch negatives, and score-level fusion with BM25. Our generator is a compact model (2B--12B parameters) served under 4-bit quantization, either prompted or adapted with retrieval-aware fine-tuning (RAFT) through LoRA. On ObliQA, a question-answering benchmark built from the Abu Dhabi Global Market rulebooks, the staged retriever raises Recall@10 from 0.256 to 0.774 and outperforms BM25 (0.678) and E5-large-v2 (0.758), the strongest general-purpose dense encoder we tested. RAFT-LoRA then improves the composite RePASs answer-quality score for every model we could adapt, with the largest gain on the weakest one. However, the adapted models do not transfer to Australian case-law questions, and a closed-book model that receives no passages at all scores within 0.011 RePASs of the full pipeline while producing answers that cite nothing and misstate obligations. The retrieval gain is therefore measured directly, the generation gain is a gain in RePASs rather than demonstrated grounding, and grounding itself requires an evaluation protocol that RePASs does not provide.
Chinese Translation
金融机构在密集且频繁修订的规则手册下运营，正确回答合规性问题不仅需要流利的表达，还需要在权威文本中的可验证依据。大型语言模型在这一任务中具有吸引力，然而企业实际可部署的本地模型通常较为紧凑，而紧凑模型容易产生虚假义务。我们研究了一个经过精心领域适应的检索增强生成管道是否能够弥补这一差距。我们的检索器基于 LegalBERT 经过三个阶段构建：前提-假设重构的蕴涵调优，将问题-段落匹配重塑为前提-假设重构，使用批内负样本的对比调优，以及与 BM25 的分数级融合。我们的生成器是一个紧凑模型（2B-12B 参数），在 4 位量化下提供服务，既可以通过提示进行，也可以通过检索感知微调（RAFT）进行适应。我们在 ObliQA 上进行评估，这是一个基于阿布扎比全球市场规则手册构建的问答基准，分阶段的检索器将 Recall@10 从 0.256 提高到 0.774，超越了 BM25（0.678）和 E5-large-v2（0.758），后者是我们测试的最强通用稠密编码器。RAFT-LoRA 随后提高了我们能够适应的每个模型的综合 RePASs 答案质量评分，最弱模型的增益最大。然而，适应后的模型无法转移到澳大利亚案例法问题上，而一个完全不接收段落的闭卷模型的得分与完整管道的 RePASs 相差仅 0.011，同时产生的答案没有引用任何内容且错误陈述了义务。因此，检索增益是直接测量的，生成增益是 RePASs 的提升而非证明的基础，而基础本身需要一个 RePASs 并未提供的评估协议。
cs.CL / 72 / 2609.30012
Low-Cost Assays for Measuring Model Behavior Across Vendors and Releases
跨供应商和版本测量模型行为的低成本测试
Abstract
Language models advise people, keep them company, and write software while they sleep. Measuring what they do is hard: behavior has to be sampled repeatedly across models, prompts and releases, most of it lives in unstructured text that has to be coded before it can be counted, and the result has to be legible and rigorous enough to meaningfully compare models and vendors. To address these constraints, we present a simple, cheap, scalable, and replicable model for studying model behavior. Each study is a frozen, public stimulus run identically on a cross-vendor panel, at a few dollars per model or less. Each reads its transcripts one of three ways, chosen by how much interpretation the behavior needs: exact match on a clamped reply, a codebook applied by LLM judges whose agreement with a human coder is reported per code, and an instrumented environment that records what an agent did independently of what it said. Run across four years of model releases from both frontier and open-source labs, these instruments find four things. Convergence: asked to pick a word, 27 of 44 models answer serendipity at least once in four tries. Resistance: a trailing "right?" moves endorsement by up to 32 points, and the sign flips from sycophantic to resistant as generations advance, keyed to the tag's surface form. House: whether a model holds a position under pressure tracks its generation, and how it holds tracks the lab that built it. Account: told to do something the documentation in their repository contradicts, some coding agents never went along silently and others always did, and the same model can change with the harness it runs in. Re-run on every release, batteries like these track how behavior is changing across vendors and over time.
Chinese Translation
语言模型在夜间为人们提供建议、陪伴，并编写软件。测量它们的行为是困难的：行为必须在模型、提示和版本之间反复采样，大部分行为以非结构化文本的形式存在，必须经过编码才能进行计数，结果必须足够清晰和严谨，以便有意义地比较模型和供应商。为了解决这些限制，我们提出了一种简单、廉价、可扩展且可复制的模型来研究模型行为。每项研究都是一个冻结的公共刺激，在跨供应商面板上以相同方式运行，每个模型的成本在几美元或更少。每个模型以三种方式之一读取其转录，具体取决于行为需要多少解释：在限制回复上的精确匹配、由大型语言模型（LLM）评审应用的代码本，其与人类编码者的协议按代码报告，以及记录代理所做的事情而不依赖于其所说内容的仪器化环境。在来自前沿和开源实验室的四年模型发布中，这些工具发现了四个方面。收敛性：在被要求选择一个词时，44个模型中有27个在四次尝试中至少回答一次“偶然性”。抵抗性：一个拖尾的“对吗？”使得支持度提高最多32个百分点，随着生成的推进，标志的表面形式从谄媚转变为抵抗。立场：模型在压力下是否持有某种立场与其生成相关，而其持有方式则与构建它的实验室相关。账户：当被告知做一些与其存储库中的文档相矛盾的事情时，一些编码代理从未默默同意，而其他代理则总是这样做，同一模型也可能因其运行的环境而变化。通过对每个版本重新运行，这些电池跟踪行为在供应商之间和随时间的变化。
cs.CL / 73 / 2609.30030
Artificial Societies Benchmark: A Validation Framework for Synthetic Research
人工社会基准：合成研究的验证框架
Abstract
A synthetic survey can reproduce the average answer while misrepresenting how people differ, how their answers relate to one another, or how they respond to changes in conditions. We introduce the Artificial Societies Benchmark to help researchers assess whether synthetic populations support their intended analyses. The framework combines eleven tests across internal, construct, and external validity, drawing on twenty human sources and comparing nine language models. It connects each research use to the evidence it requires and tests how results change with the information we supply about respondents. Importantly, strong performance in one domain does not establish fidelity in the others. Models often answer too consistently, compress response scales, and alter relationships between traits whilst richer profiles improve prediction for some models and worsen it for others. The resulting scorecard helps researchers identify which aspects of a synthetic population can support their analysis and where researchers need further human evidence.
Chinese Translation
合成调查可以再现平均答案，但可能会误导人们的差异、答案之间的关系或他们对条件变化的反应。我们引入了人工社会基准，以帮助研究人员评估合成群体是否支持他们的预期分析。该框架结合了内部、构念和外部有效性的十一项测试，基于二十个人类来源并比较九种语言模型。它将每个研究用途与所需证据相连接，并测试结果如何随着我们提供的关于受访者的信息而变化。重要的是，在一个领域的强劲表现并不意味着在其他领域的可靠性。模型往往回答得过于一致，压缩响应尺度，并改变特征之间的关系，而更丰富的个人资料在某些模型中提高了预测能力，但在其他模型中则降低了预测能力。最终的评分卡帮助研究人员识别合成群体的哪些方面可以支持他们的分析，以及研究人员在哪些方面需要进一步的人类证据。
cs.CL / 74 / 2609.30071
Scoring Both Directions: LLMs realize the MRS they cannot reliably parse
双向评分：大型语言模型实现其无法可靠解析的最小递归语义
Abstract
The English Resource Grammar (ERG) is a hand-written computational grammar of English. Given a sentence, its processor, ACE, produces a formal meaning representation called Minimal Recursion Semantics (MRS): a graph of the sentence's predicates and their arguments. The grammar is bidirectional and can also turn an MRS back into an English sentence. \citet{hajdik2019} used the ERG's treebank to build a benchmark for that generation task, MRS to text, and trained sequence-to-sequence models to solve it. The parsing task, text to MRS, can be tested on the same sentences. We reconstruct their 10K-sentence test split, and score two large language models, Claude Sonnet~4.5 and Claude Opus~5, in both directions against their trained systems and against ACE, with no task-specific training. Given an MRS and three examples, Opus writes the sentence at 76.3 BLEU, ten points above their system trained on 72k pairs (66.1 BLEU), and comparable to their system trained on a million extra pairs (77.2 BLEU). Sonnet scores 65.7 BLEU, and letting it choose among ACE's own candidate sentences lifts it to 69.6, while a pooled judge that keeps Opus's own sentence among the candidates adds 0.6 points (77.0 BLEU). In the parsing direction, however, the models fall far behind ACE: asked for the MRS of the same sentences, they reach 57.2 (Sonnet) and 65.5 (Opus) F$_1$ on the graph's predicates and arguments against 91.0 for ACE, and exact-match the gold on about 1\% of sentences. We characterize the failure modes for the parsing tasks, and conclude that a generation score alone does not show that models understand formal semantic representations.
Chinese Translation
英语资源语法（ERG）是一种手写的英语计算语法。给定一个句子，其处理器ACE生成一个称为最小递归语义（MRS）的形式意义表示：一个句子的谓词及其论元的图。该语法是双向的，也可以将MRS转换回英语句子。 extcite{hajdik2019}利用ERG的树库构建了一个用于该生成任务（MRS到文本）的基准，并训练了序列到序列模型来解决该任务。解析任务（文本到MRS）可以在相同的句子上进行测试。我们重建了他们的1万句测试集，并在两个方向上对两个大型语言模型Claude Sonnet~4.5和Claude Opus~5进行评分，与他们训练的系统和ACE进行比较，且没有特定任务的训练。在给定一个MRS和三个示例的情况下，Opus生成的句子得分为76.3 BLEU，比他们在72k对上训练的系统高出十个点（66.1 BLEU），并且与他们在额外一百万对上训练的系统（77.2 BLEU）相当。Sonnet得分为65.7 BLEU，允许其在ACE自己的候选句子中选择使其得分提升至69.6，而一个将Opus自身句子保持在候选中的汇总评审增加了0.6分（77.0 BLEU）。然而，在解析方向上，这些模型远远落后于ACE：在要求提供相同句子的MRS时，它们在图的谓词和论元上分别达到57.2（Sonnet）和65.5（Opus）F$_1$，而ACE的得分为91.0，并且在大约1 ext{%}的句子上实现了与金标准的完全匹配。我们描述了解析任务的失败模式，并得出结论，仅凭生成得分并不能表明模型理解形式语义表示。
cs.CL / 75 / 2609.30074
How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure
评估结论的可重复性如何？对LLM推断提示结构的自我审计
Abstract
Evaluations of LLM systems routinely average over small prompt sets and report models as a ranked table. We ask how much confidence such a table deserves, using LLM-based prompt-structure inference as the case study: eight open model variants across five families and 8B to 675B parameters, caching disabled, 293 raw intermediate representations persisted. The measured phenomenon is unstable to begin with. Identical calls do not reliably recover identical structure, with mean node-set Jaccard from 0.39 to 0.96 and 72% of prompt-model cells never node-set-perfect. Auditing the evaluation weakens its conclusions further, and this is our main contribution. Under a joint cluster bootstrap over prompts, only the bottom of the ranking is firm: the two least reproducible models hold rank in 99% and 86% of replicates, the middle four in 27% to 48%, and the top two in 68% each, so the table identifies the worst model reliably but does not reliably identify the best. Two equally defensible rules for merging repeated campaigns change four of eight rows and move the study-wide headline by 7 percentage points. Checking the inferred structure against ground-truth annotations shows reproducibility cannot be read as accuracy. And four of the eight endpoints were withdrawn within ten weeks of measurement, so the study as specified can no longer be run. Small-sample LLM evaluations can therefore look far more definitive than their evidence supports. We recommend reporting rank stability, per-cell provenance, executed sensitivity comparisons, raw per-run outputs, and a measurement date alongside any ranking.
Chinese Translation
对LLM系统的评估通常是在小规模提示集上进行平均，并以排名表的形式报告模型。我们探讨这样的排名表值得多少信心，以LLM基础的提示结构推断作为案例研究：八种开放模型变体，涵盖五个家族，参数规模从8B到675B，缓存禁用，保留293个原始中间表示。所测量的现象本身就不稳定。相同的调用并不可靠地恢复相同的结构，节点集的Jaccard均值从0.39到0.96，且72%的提示-模型单元从未达到节点集完美。对评估的审计进一步削弱了其结论，这是我们主要的贡献。在对提示进行联合聚类自助法下，排名的底部是相对稳固的：两个可重复性最低的模型在99%和86%的重复中保持排名，中间四个模型在27%到48%之间，前两个模型各自为68%，因此该表格可靠地识别出最差模型，但并不可靠地识别出最佳模型。两条同样合理的合并重复实验的规则改变了八行中的四行，并使研究整体的头条变化了7个百分点。将推断的结构与真实注释进行对比显示，可重复性不能被视为准确性。而且在测量后的十周内，有四个端点被撤回，因此指定的研究无法再进行。因此，小样本的LLM评估看起来比其证据所支持的要明确得多。我们建议在任何排名旁边报告排名稳定性、每个单元的来源、执行的敏感性比较、原始每次运行的输出以及测量日期。
cs.CL / 76 / 2609.30075
A Native-Reference Phone-Class Geometry for Second-Language Pronunciation Analysis
基于母语参考的音素类别几何用于第二语言发音分析
Abstract
Automatic speaking assessment systems can provide holistic proficiency scores, but often lack interpretable measures that characterize pronunciation quality. We propose a native-reference phone-class geometry for measuring second language (L2) pronunciation deviation without requiring pronunciation labels, read-aloud prompts, or matched recordings of the same text from native and L2 speakers. Given a native speech corpus, we average frame-level self-supervised representations for each context-dependent phone-class and use singular value decomposition (SVD) to derive a compact native-reference coordinate system. For each L2 utterance, we compute the corresponding averages and project them into the native-reference space. We then demonstrate that the distances between L2 and native-reference coordinates for matched phone-classes show consistent negative correlations with holistic speaking proficiency on the Dev subset of the Speak and Improve Corpus 2025 (Spearman's $\rho\!=\!-0.53$) and with pronunciation quality on the learner subset of the English Read by Japanese Students dataset ($\rho\!=\!-0.34$). These findings suggest that the proposed geometry captures acoustic-phonetic information relevant for proficiency rating while remaining applicable to spontaneous L2 speech without matched native recordings.
Chinese Translation
自动口语评估系统可以提供整体的语言能力评分，但通常缺乏可解释的度量来表征发音质量。我们提出了一种基于母语参考的音素类别几何，用于测量第二语言（L2）发音偏差，而无需发音标签、朗读提示或母语和L2说话者对同一文本的匹配录音。在给定的母语语料库中，我们对每个上下文相关的音素类别的帧级自监督表示进行平均，并使用奇异值分解（SVD）推导出一个紧凑的母语参考坐标系统。对于每个L2发音，我们计算相应的平均值并将其投影到母语参考空间中。然后，我们展示了匹配音素类别的L2坐标与母语参考坐标之间的距离与Speak and Improve Corpus 2025的Dev子集上的整体口语能力（Spearman's $
ho
eq-0.53$）以及与日本学生英语阅读数据集的学习者子集上的发音质量（$
ho
eq-0.34$）之间存在一致的负相关关系。这些发现表明，所提出的几何结构捕捉了与能力评分相关的声学-语音信息，同时适用于没有匹配母语录音的自发L2口语。
cs.CL / 77 / 2609.30087
Return or Revise? Learning When Revision Helps Retrieval-Augmented QA
返回还是修订？学习何时修订有助于检索增强的问答
Abstract
We consider the decision of whether to return an existing draft answer or revise it using retrieved evidence, as in answer-revision systems. Draft confidence estimates whether the current answer is correct, but the decision requires estimating the effect of a specified revision. For offline training and evaluation, we grade both the returned draft and its candidate revision under the same correctness judge, which makes repair, harm, and the gap to an oracle observable. We call this paired effect its recoverability, and we train policies to predict it before revision. On 25,870 held-out open-domain questions across three revision setups, a scorer trained on the paired outcome has greater area under the accuracy--revision-rate curve than a matched draft-correctness scorer in all nine Llama setup--seed fits, and gains 0.23--0.68 accuracy points on average at development-selected thresholds, a difference significant across training runs only for dense retrieval. The resulting policy improves on always revising and on average closes more than a third of the oracle gap, although it still applies 38--46% of the harmful revisions. When a draft-free standard-RAG answer is also available, however, choosing between the draft and that answer is stronger by about two points for Llama and four for OLMo, and adding candidate revision as a third option yields no significant gain. Recoverability describes one revision; its value as an available action also depends on the alternatives.
Chinese Translation
我们考虑在回答修订系统中，是否返回现有草稿答案或使用检索到的证据进行修订的决策。草稿置信度估计当前答案是否正确，但该决策需要估计特定修订的效果。为了进行离线训练和评估，我们在相同的正确性判断下对返回的草稿和其候选修订进行评分，这使得修复、损害以及与理想答案之间的差距可观察。我们称这种配对效果为其可恢复性，并训练策略在修订前预测它。在三个修订设置下的25,870个保留开放领域问题中，基于配对结果训练的评分器在所有九个Llama设置-种子拟合中，其准确率-修订率曲线下的面积优于匹配的草稿正确性评分器，并在开发选择的阈值下平均获得0.23-0.68的准确率提升，这一差异在训练运行中仅对密集检索显著。最终的策略在总是修订的基础上有所改进，平均上缩小了超过三分之一的理想答案差距，尽管它仍然应用了38-46%的有害修订。然而，当同时提供无草稿的标准RAG答案时，在草稿和该答案之间的选择对于Llama强大约两分，对于OLMo强大约四分，且将候选修订作为第三选项并未带来显著收益。可恢复性描述了一次修订；作为可用行动的价值还依赖于替代方案。
cs.CL / 78 / 2609.30100
R-DEIM Net: An Efficient Rationale-Augmented Dual-Expert Interaction Model for Paraphrase Detection
R-DEIM Net：一种高效的基于理由增强的双专家交互模型用于复述检测
Abstract
Recent advances in paraphrase detection reveal a fundamental trade-off: large language models achieve high accuracy but require high computation, while efficient Siamese-BERT variants offer practical scalability with reduced transparency in rationale generation. We present R-DEIM Net, a 76M-parameter dual-expert architecture exploring whether moderate-scale models can achieve competitive accuracy on paraphrase detection while enabling human-readable rationale generation. The architecture combines two specialized components: an Interaction Expert that captures token-level similarity patterns through multi-scale 2D convolutions and attention head allowing variable input length, and a Reasoning Expert that uses a Flan-T5-small decoder to generate rationales as auxiliary supervision. Rather than re-encoding generated text, we extract and pool decoder hidden states as complementary features for classification. On the Quora Question Pairs dataset, R-DEIM Net achieves 90.07\% accuracy and 90.16\% F1-score via 10-fold cross-validation. This represents competitive performance with strong transformer-based baselines (e.g., MFAE BERT: 90.54\% accuracy) and recent large language model based approaches (LLaMA-70B) while using a substantially smaller parameter budget. The model generates rationales alongside predictions, providing potential for auxiliary human-readable descriptions.
Chinese Translation
近年来复述检测的进展揭示了一个基本的权衡：大型语言模型虽然能够实现高准确率，但计算成本较高；而高效的Siamese-BERT变体在实际可扩展性方面表现良好，但在理由生成的透明度上有所降低。我们提出了R-DEIM Net，一种拥有7600万参数的双专家架构，旨在探索中等规模模型是否能够在复述检测任务中达到具有竞争力的准确率，同时实现可供人类理解的理由生成。该架构结合了两个专门组件：交互专家（Interaction Expert），通过多尺度二维卷积和允许可变输入长度的注意力头捕捉词元级别的相似性模式；推理专家（Reasoning Expert），利用Flan-T5-small解码器生成理由，作为辅助监督。我们并未对生成的文本进行重新编码，而是提取并池化解码器的隐藏状态，作为分类的补充特征。在Quora Question Pairs数据集上，R-DEIM Net通过10折交叉验证实现了90.07%的准确率和90.16%的F1分数。这一表现与强大的基于Transformer的基线模型（如MFAE BERT：90.54%准确率）及近期的大型语言模型方法（LLaMA-70B）相当，同时参数规模显著较小。该模型在生成预测结果的同时，还能生成理由，为辅助生成可供人类理解的描述提供了可能。
cs.CL / 79 / 2609.30121
What, When, and How: Audio Description as Constrained Global Optimization
什么、何时与如何：作为约束全局优化的音频描述
Abstract
Audio Description (AD) makes movies accessible to blind and visually impaired audiences by narrating visual information in gaps between dialogue. Existing automatic AD systems largely treat generation as a local video-to-text problem, assuming that the content to describe and its temporal location are already provided. Realistic AD instead requires coupled decisions about what visual information is narratively important, when it can be spoken without interfering with dialogue, and how it should be formulated to fit within the available time. We formalize AD generation as a constrained optimization problem over these three decisions. Our hybrid system uses large language models to propose and ground visual elements, estimate their salience to the narrative, and generate compressed realizations. A mixed-integer linear program then jointly selects and schedules descriptions across a scene subject to temporal constraints. When evaluated on REFRAMED, a benchmark for realistic AD of movies, our approach makes better decisions than prompted LLMs about what to describe and when to describe it, establishing a new SOTA on narrative QA and temporally grounded metrics. Ablations show that explicit temporal constraints drive gains in placement, while salience estimation controls how much narratively useful content is retained. Improvements are concentrated on temporal and narrative measures rather than n-gram overlap, although a significant gap to professional describers remains.
Chinese Translation
音频描述（Audio Description，AD）通过在对话间隙叙述视觉信息，使电影对盲人和视障观众可及。现有的自动音频描述系统大多将生成视为局部的视频到文本问题，假设待描述内容及其时间位置已知。而现实中的音频描述需要综合决策：哪些视觉信息在叙事上重要、何时可以在不干扰对话的情况下叙述，以及如何在有限时间内进行合适的表达。我们将音频描述生成形式化为涵盖这三方面决策的约束优化问题。我们的混合系统利用大型语言模型（LLMs）提出并定位视觉元素，评估其叙事显著性，并生成压缩的表达形式。随后，混合整数线性规划联合选择并安排场景中的描述，满足时间约束。在电影现实音频描述基准REFRAMED上的评测表明，我们的方法在描述内容选择及叙述时机上优于提示式大型语言模型，创下叙事问答及时间定位指标的新状态。消融实验显示，显式时间约束促进了描述时机的提升，而显著性估计则控制了保留多少叙事有用内容。改进主要集中在时间和叙事指标上，而非n-gram重叠，尽管与专业描述者仍存在显著差距。
cs.CL / 80 / 2609.30151
Does a model's stated reason for rejecting a candidate do any work?
模型拒绝候选人的理由是否有效？
Abstract
Asked to choose between candidates and explain the choice, a language model often rejects a rival by naming a fact its profile lacks: no director, no date of death. That sentence is a claim about the text in front of the model, and it can be tested without any judge. We insert a real corpus sentence stating the named fact into the rival's profile and ask again under greedy decoding. Two controls separate content from placement: a length-matched irrelevant sentence at the same profile, and the same two sentences at a third option the model never mentioned. In the largest of three runs, six open models on 2WikiMultihopQA, supplying the named fact at the profile the model named moves its choice more than the irrelevant control does, odds ratio 3.57 [1.54, 8.26], Holm p=0.0210, and this survives dropping any single model. The contrast the design was built to detect, the same fact at the option nobody named, does not clear correction, Holm p=0.2428. The strongest result in the family carries no content claim at all: the identical irrelevant sentence moves the choice more at the named rival than at the third option, Holm p=0.0008. Repair and control also differ in co-candidate mentions, relation template and fluency; post-hoc matching on the first two preserves the content effects' direction, matching fluency weakens one, so the content contrasts bound an effect rather than establish one. A forced single-token probability read disagrees in direction with the free-text choice on that same contrast, and three candidate explanations for the disagreement find no support. Every measurement is a string rule, so each was validated against the records it reads; validation caught eight defects. The largest, a choice-parsing rule that returned the option a model had just rejected in 17.1% of adjudicable responses, would have reported six surviving contrasts instead of four.
Chinese Translation
在被要求在候选人之间进行选择并解释选择时，语言模型通常通过指出其资料缺乏的事实来拒绝竞争者：没有导演，没有死亡日期。该句子是对模型面前文本的一个声明，可以在没有任何评判者的情况下进行测试。我们将一条真实语料库句子（陈述所提及的事实）插入竞争者的资料中，并在贪婪解码下再次提问。两个控制实验将内容与位置分开：一个在同一资料中的长度匹配的无关句子，以及在模型从未提及的第三选项中的相同两个句子。在三次实验中，最大的实验中，六个开放模型在2WikiMultihopQA上，提供的资料中所提及的事实使得模型的选择比无关控制更有影响，优势比为3.57 [1.54, 8.26]，Holm p=0.0210，并且这一结果在去掉任何单一模型后仍然成立。设计所要检测的对比，即在无人提及的选项中的相同事实，未能通过校正，Holm p=0.2428。该系列中最强的结果完全没有内容声明：在被提及的竞争者中，完全相同的无关句子对选择的影响大于在第三选项中的影响，Holm p=0.0008。修复和控制在共同候选人提及、关系模板和流畅性上也存在差异；对前两者的事后匹配保持了内容效应的方向，而匹配流畅性削弱了其中一个，因此内容对比界定了一个效应，而不是确立一个效应。强制的单标记概率读取在方向上与自由文本选择在同一对比中存在分歧，且对此分歧的三种候选解释没有找到支持。每个测量都是一个字符串规则，因此每个都经过其读取记录的验证；验证发现了八个缺陷。其中最大的一个是选择解析规则，在17.1%的可裁定响应中返回模型刚刚拒绝的选项，本应报告六个存活的对比，而不是四个。
cs.CL / 81 / 2609.30160
A Training Criterion with Token-Level Tolerance to Transcription Ambiguity for Automatic Speech Recognition
一种对转录模糊性具有标记级容忍度的自动语音识别训练标准
Abstract
Automatic speech recognition is typically trained assuming that the reference transcript is the only valid labeling of an utterance, yet even nominally verbatim transcripts contain localized differences in pronunciation, spelling, or lexical realization that the acoustics do not uniquely determine. Omni-temporal Classification (OTC) tolerates such noise by adding wildcard paths to the connectionist temporal classification (CTC) alignment graph, but its word-level arcs are too coarse, since bypassing one unsupported token discards supervision for the whole word. We move wildcard arcs to token granularity so unsupported tokens can be bypassed while the rest of the word stays supervised, and we combine token- and word-level arcs as complementary escape paths. Across 19 languages and three corpora, token-level OTC improves over CTC on all 25 tasks. We also replace epoch-indexed relaxation of the wildcard weights with a predictive-entropy-indexed schedule, which performs comparably while reducing dependence on training length. Combining this schedule with the hybrid graph gives the lowest mean word error rate (WER) on every corpus and a 9.45% average relative WER reduction over CTC. Independent validator transcriptions show that token-level models place significantly more wildcard-bypass probability than CTC on disputed characters, indicating that token-level tolerance targets localized transcript ambiguity.
Chinese Translation
自动语音识别通常假设参考转录是话语的唯一有效标记，然而即使是名义上逐字的转录也包含发音、拼写或词汇实现上的局部差异，而这些差异并不能被声学特征唯一确定。全时分类（Omni-temporal Classification, OTC）通过向连接时序分类（Connectionist Temporal Classification, CTC）对齐图添加通配符路径来容忍这种噪声，但其词级弧线过于粗糙，因为绕过一个不支持的标记会丢弃整个词的监督。我们将通配符弧线移至标记粒度，以便在其余词语保持监督的同时绕过不支持的标记，并将标记级和词级弧线结合为互补的逃逸路径。在19种语言和三个语料库中，标记级OTC在所有25个任务上均优于CTC。我们还用预测熵索引调度替代了通配符权重的周期索引松弛，该方法在减少对训练长度的依赖的同时表现相当。将该调度与混合图结合，使每个语料库的平均词错误率（Word Error Rate, WER）达到最低，并在CTC的基础上实现了9.45%的平均相对WER降低。独立验证者的转录显示，标记级模型在有争议字符上赋予了显著更多的通配符绕过概率，表明标记级容忍度针对的是局部转录模糊性。
cs.CL / 82 / 2609.30167
Do Audio Language Models Hear and Read Distinctive Features Alike?
音频语言模型是否以相似的方式感知和读取特征？
Abstract
Audio language models pass speech and text through a single decoder. We ask whether that decoder represents a distinctive feature in the same direction when a phoneme is heard and when it is read. For minimal pairs of phonemes differing in one feature, we take the offset between the two members' mean representations. Averaging those offsets gives a direction for each stream, and we measure the cosine between the two. Because the two streams already agree about arbitrary phoneme pairs, we compare every measure against a reference built from random pairings rather than against zero. We apply this to 6 models, 7 features and 15 languages from 11 families. Only voicing in the two Qwen2.5-Omni models exceeds that reference after correction for multiple testing, and the reference varies by a factor of seven between models. In three of the six models, voicing has one direction in audio across the 14 languages with enough minimal pairs to measure it, and every language pair agrees in two of them. The model family, not the model size, predicts which stream represents a feature.
Chinese Translation
音频语言模型通过单一解码器处理语音和文本。我们探讨该解码器在听到音素和读取音素时是否以相同的方向表示特征。对于在一个特征上存在差异的最小音素对，我们计算两个成员的均值表示之间的偏移。对这些偏移进行平均可以为每个流提供一个方向，我们测量这两个方向之间的余弦相似度。由于这两个流在任意音素对上已经达成一致，我们将每个测量值与基于随机配对构建的参考进行比较，而不是与零进行比较。我们将此应用于来自11个语系的6个模型、7个特征和15种语言。经过多重检验校正后，只有两个Qwen2.5-Omni模型中的声音特征超过了该参考，并且该参考在模型之间的变化因子为七。在六个模型中的三个模型中，声音特征在14种具有足够最小音素对的语言中具有相同的方向，并且每对语言在其中的两个模型中达成一致。模型家族，而非模型大小，预测哪个流表示特征。
cs.CL / 83 / 2609.30184
ARGUS: Role-Aware Event Knowledge Graphs for U.S. Employment-Discrimination Complaints
ARGUS：基于角色的美国就业歧视投诉事件知识图谱
Abstract
U.S. employment-discrimination complaints describe complex event sequences that are not explicitly captured by lexical or embedding-based representations alone. We present ARGUS, a source-grounded pipeline that combines a 5W1H-inspired schema, legal-domain models, and LLM-based structured generation to construct document-level Event Knowledge Graphs (EKGs) from CourtListener complaints. ARGUS extracts fact-bearing statements, builds chunk-level event graphs with participant, temporal, and causal structure, and merges them into document-level representations. We evaluate graph quality through human and multi-model assessment and test downstream utility on claim classification and legal QA. The graph-structured classifier outperforms raw and linearized baselines on the held-out set, and EKG-only retrieval improves document-scoped QA, while open-retrieval gains remain limited by low first-stage candidate recall. These results suggest that EKGs are most useful for organizing and reasoning over evidence once relevant material has been retrieved.
Chinese Translation
美国就业歧视投诉描述了复杂的事件序列，这些序列仅通过词汇或嵌入式表示无法明确捕捉。我们提出了ARGUS，一个基于来源的管道，结合了5W1H启发式架构、法律领域模型和基于大型语言模型（LLM）的结构化生成，从CourtListener投诉中构建文档级事件知识图谱（EKGs）。ARGUS提取包含事实的陈述，构建具有参与者、时间和因果结构的块级事件图，并将其合并为文档级表示。我们通过人工和多模型评估来评估图的质量，并测试其在索赔分类和法律问答（QA）中的下游效用。图结构分类器在保留集上优于原始和线性化基线，而仅基于EKG的检索改善了文档范围的QA，尽管开放检索的增益仍受到第一阶段候选者召回率低的限制。这些结果表明，EKGs在检索到相关材料后，对于组织和推理证据最为有用。
cs.CL / 84 / 2609.30238
SemMSA: Latent Semantic-Aided Robust Multimodal Sentiment Analysis with Incomplete Data
SemMSA：基于潜在语义的鲁棒多模态情感分析框架（处理不完整数据）
Abstract
Recent research on Multimodal Sentiment Analysis (MSA) has focused on learning from language, visual, and acoustic modalities with incomplete data to infer human sentiment. Most studies typically compensate for missing information by reconstructing modality features or designing complicated fusion mechanisms. However, these methods still suffer from spurious generation and noisy guidance due to the lack of high-level semantic grounding in partially observed multimodal evidence. To address these issues, we propose SemMSA, a latent semantic-aided framework that constructs rich sentiment-relevant semantics with LLMs, fully integrating with all modalities via anchor-free spectral alignment. It mainly consists of Cross-modal Semantic Refinement (CSR) and Cross-modal Spectral Alignment (CSA). Specifically, CSR first adaptively extracts visual and acoustic representations by corresponding adapters to form a unified multimodal prefix with language in the frozen LLM embedding space. It then iteratively produces continuous discriminative semantic states through a token-efficient latent refinement process without decoding explicit text. Next, CSA simultaneously aligns the refined semantics with all modalities by enhancing the dominant spectral component of their kernel Gram matrix. This captures global nonlinear dependencies among all representations without relying on a predefined anchor modality. In addition, an instance-level spectral separation constraint preserves cross-sample discriminability and mitigates representation collapse. Extensive experiments on SIMS, MOSI, and MOSEI benchmarks demonstrate that SemMSA achieves state-of-the-art performance.
Chinese Translation
近期对多模态情感分析（MSA）的研究集中于从语言、视觉和声学模态中学习，以推断人类情感，尤其是在不完整数据的情况下。大多数研究通常通过重构模态特征或设计复杂的融合机制来补偿缺失的信息。然而，由于在部分观察的多模态证据中缺乏高层次的语义基础，这些方法仍然面临虚假生成和噪声引导的问题。为了解决这些问题，我们提出了SemMSA，一个基于潜在语义的框架，该框架利用大型语言模型（LLMs）构建丰富的情感相关语义，并通过无锚点的光谱对齐完全整合所有模态。该框架主要由跨模态语义精炼（CSR）和跨模态光谱对齐（CSA）组成。具体而言，CSR首先通过相应的适配器自适应提取视觉和声学表示，以形成与冻结的LLM嵌入空间中的语言统一的多模态前缀。然后，它通过一种高效的潜在精炼过程迭代生成连续的区分语义状态，而无需解码显式文本。接下来，CSA通过增强其核Gram矩阵的主导光谱分量，同时对所有模态的精炼语义进行对齐。这捕捉了所有表示之间的全局非线性依赖关系，而不依赖于预定义的锚模态。此外，实例级光谱分离约束保持了跨样本的可区分性，并减轻了表示崩溃。在SIMS、MOSI和MOSEI基准上的大量实验表明，SemMSA实现了最先进的性能。
cs.CL / 85 / 2609.30243
JevOut: Natural Context Can Flip Decision Models
JevOut：自然上下文可以翻转决策模型
Abstract
Dedicated decision models such as Jev map unstructured language to probability distributions over finite choices, allowing their outputs to directly route requests, select tools, and trigger actions. Yet real-world inputs rarely arrive in isolation: they come with background details and surrounding context. We find that short additions that fit naturally into this context can nevertheless redirect an otherwise correct decision, even when the correct answer remains unchanged. To study this behavior, we fix a wrong target option for each initially correct item and use the model's option probabilities to refine fluent context additions while preserving the source, question, choices, and gold answer. Within 64 accepted target evaluations, the optimizer identifies contexts that redirect Jev on 312 of 508 initially correct decisions (61.4%); in 229 cases, Jev assigns at least 0.7 probability to the fixed wrong option. Across seven datasets, three additional decision systems show targeted flip rates of 64.9%-73.2% on decisions they initially answer correctly. Taken together, these results expose a pronounced fragility in current decision models: short, ordinary-looking context can shift a correct choice to a high-confidence wrong one. Because these models turn language directly into downstream choices, this sensitivity raises concerns about treating their probability outputs as reliable decision interfaces.
Chinese Translation
专用决策模型如Jev将非结构化语言映射到有限选择上的概率分布，使其输出能够直接路由请求、选择工具并触发动作。然而，现实世界中的输入很少是孤立到达的：它们伴随着背景细节和周围的上下文。我们发现，适合自然融入此上下文的短小补充，仍然可以重新引导一个本应正确的决策，即使正确答案保持不变。为了研究这种行为，我们为每个最初正确的项目固定一个错误的目标选项，并利用模型的选项概率来优化流畅的上下文补充，同时保留源、问题、选择和正确答案。在64个接受的目标评估中，优化器识别出能够使Jev在508个最初正确决策中重定向312个（61.4%）的上下文；在229个案例中，Jev对固定的错误选项赋予至少0.7的概率。在七个数据集中，另外三种决策系统在最初正确回答的决策上显示出64.9%-73.2%的目标翻转率。综合来看，这些结果揭示了当前决策模型的明显脆弱性：短小、看似普通的上下文可以将正确选择转变为高置信度的错误选择。由于这些模型将语言直接转化为下游选择，这种敏感性引发了对将其概率输出视为可靠决策接口的担忧。
cs.CL / 86 / 2609.30250
Agentic Detection of Online Conspiracies
在线阴谋的能动检测
Abstract
Conspiratorial discourse on social media is not always expressed through explicit claims or stable lexical markers. The same surface content may express endorsement, legitimate concerns, criticism, satire, or mockery. The main challenge is therefore not only recognizing conspiracy-related claims, but inferring the speaker's intent -- the utterance's illocutionary force. We argue that this can be achieved through the use of relevant social contexts and propose an agentic framework, equipped with a set of tools supporting social queries. We demonstrate the benefits of our approach on a unique dataset of Hebrew tweets, covering 80\%--90\% of the public Hebrew tweets published over a four-year span (late 2018-- early 2023), encompassing several election cycles as well as the COVID pandemic years and related vaccination campaigns. This extensive coverage can be used in recovering different social contexts. Evaluating our framework on a manually-annotated adversarial dataset, we find that context-aware workflows consistently outperform text-only classification and that the agentic framework performs significantly better than other frameworks and settings, including a non-agentic model exposed to the same contexts available to the agent. We further provide an analysis of the results, the errors and efficiency (token economy) tradeoffs. These findings support viewing the task of conspiracy detection as a socially embedded interpretation task, in which effective classification depends not only on access to contexts, but also on adaptive reasoning in which the agent uses tools on a per-case basis, asking only for evidence relevant to its current reasoning step.
Chinese Translation
社交媒体上的阴谋话语并不总是通过明确的主张或稳定的词汇标记来表达。相同的表面内容可能表达支持、合理的担忧、批评、讽刺或嘲弄。因此，主要挑战不仅在于识别与阴谋相关的主张，还在于推断说话者的意图——言语的言外之意。我们认为，这可以通过使用相关的社会背景来实现，并提出一个能动框架，配备一套支持社会查询的工具。我们在一个独特的希伯来语推文数据集上展示了我们方法的优势，该数据集涵盖了2018年末至2023年初四年间发布的80%至90%的公共希伯来语推文，涉及多个选举周期以及COVID疫情年份和相关疫苗接种活动。这一广泛的覆盖可以用于恢复不同的社会背景。在一个手动标注的对抗性数据集上评估我们的框架，我们发现，关注上下文的工作流程始终优于仅基于文本的分类，并且能动框架的表现显著优于其他框架和设置，包括暴露于相同上下文的非能动模型。我们进一步提供了结果、错误和效率（标记经济）权衡的分析。这些发现支持将阴谋检测任务视为一种社会嵌入的解释任务，其中有效的分类不仅依赖于对上下文的访问，还依赖于适应性推理，即代理根据具体情况使用工具，仅请求与其当前推理步骤相关的证据。