cs.RO / 1 / 2607.27261
It's Not Just More Demos: Counterfactual Action Sensitivity Coverage for Data-Efficient Robust Robot Imitation
不仅仅是更多的示范：用于数据高效稳健机器人模仿的反事实动作敏感性覆盖
Abstract
Visuomotor imitation learning has demonstrated success for manipulation tasks. However, the trained policies remain brittle to visual `nuisances', with even minor task-preserving variations such as lighting, distractions or changes in colour result in heavy degradation of the trained policy's performance. While increasing data diversity can improve robustness, it is unclear which additional demonstrations are informative for a particular trained policy. We propose Counterfactual Nuisance Behaviour Cloning (CFNBC), an offline data-selection framework for targeted robustness repair. Starting from a nominal policy trained on `clean' demonstrations, CFNBC generates paired clean and nuisance observations that preserve the expert action, then measures \emph{action drift}: the change in the policy's predicted action under a nuisance that should not alter the desired behaviour. This provides a policy-specific sensitivity signal for selecting a compact, response-diverse repair set from a larger candidate pool, without requiring rollout success labels or online policy execution. We show in MuJoCo bimanual cube transfer and SimplerEnv cube stacking that action drift correlates with nuisance-induced failure, and that response-guided repair with only $20$--$30$ selected candidates substantially outperforms matched-budget random selection while approaching the performance of much larger random repair budgets. These results support a data-centric view of robustness repair: the most useful data are not necessarily the most numerous, visually diverse, or obviously difficult, but the examples that cover fragile response modes of the current policy.
Chinese Translation
视觉运动模仿学习在操控任务中取得了成功。然而，训练出的策略对视觉‘干扰’仍然脆弱，即使是诸如光照、干扰或颜色变化等轻微的保持任务的变动，也会导致训练策略性能的严重下降。虽然增加数据多样性可以提高稳健性，但尚不清楚哪些额外的示范对特定训练策略是有信息量的。我们提出了反事实干扰行为克隆（Counterfactual Nuisance Behaviour Cloning, CFNBC），这是一种用于针对性稳健性修复的离线数据选择框架。从在‘干净’示范上训练的名义策略出发，CFNBC生成保留专家动作的成对干净和干扰观察，然后测量 extit{动作漂移}：在不应改变期望行为的干扰下，策略预测动作的变化。这为从更大的候选池中选择紧凑且响应多样的修复集提供了策略特定的敏感性信号，而无需回放成功标签或在线策略执行。我们在MuJoCo双手立方体转移和SimplerEnv立方体堆叠中展示了动作漂移与干扰引起的失败之间的相关性，并且仅用$20$--$30$个选定候选者的响应引导修复显著优于匹配预算的随机选择，同时接近更大随机修复预算的性能。这些结果支持了一种以数据为中心的稳健性修复观点：最有用的数据不一定是数量最多、视觉多样性最高或明显困难的，而是覆盖当前策略脆弱响应模式的示例。
cs.RO / 2 / 2607.27494
Simulation of Surgical Suturing Using Position-Based Dynamics and the Material Point Method for Robot Reinforcement Learning
基于位置动态和材料点法的外科缝合模拟用于机器人强化学习
Abstract
Recent advances in robotics research have created a strong demand for high-performance simulators. Surgical robotics simulation faces unique challenges due to the need to model diverse objects, such as rigid instruments, soft tissue, and fluids. While many studies simulate sutures or soft tissue independently, only a few have considered the complete soft-tissue suturing scenario, including the contact between sutures and deformable tissue during suture insertion. Building on previous work, this paper presents a novel suturing simulation environment using sutures modelled by position-based dynamics (PBD) and soft bodies modelled by the material point method (MPM) while considering two-way contact with frictional and drag forces. We introduce a contact coupling method between the PBD suture and the MPM soft tissue, enabling visually plausible suture-tissue interactions. The simulator is optimized for GPU execution with parallel scenes using multiple CUDA streams, and we present a Reinforcement Learning (RL) environment for autonomous suturing sub-tasks, including needle insertion, driving, and extraction. Using ML-Agents, RL agents trained in the simulator show stable learning and achieve 80% and 68% success rates in needle insertion and extraction, respectively, under the strictest distance threshold.
Chinese Translation
近年来，机器人研究的进展对高性能模拟器的需求日益增强。外科机器人模拟面临独特的挑战，因为需要对多种对象进行建模，例如刚性工具、软组织和液体。虽然许多研究独立模拟缝合线或软组织，但只有少数考虑了完整的软组织缝合场景，包括缝合线插入过程中缝合线与可变形组织之间的接触。在之前工作的基础上，本文提出了一种新颖的缝合模拟环境，使用基于位置的动态（Position-Based Dynamics, PBD）模型的缝合线和使用材料点法（Material Point Method, MPM）模型的软体，同时考虑了摩擦和拖曳力的双向接触。我们引入了一种PBD缝合线与MPM软组织之间的接触耦合方法，使缝合线与组织之间的交互在视觉上更为可信。该模拟器经过优化以支持GPU执行，并使用多个CUDA流进行并行场景处理。此外，我们还提出了一个强化学习（Reinforcement Learning, RL）环境，用于自主缝合子任务，包括针头插入、驱动和提取。使用ML-Agents训练的RL代理在模拟器中表现出稳定的学习，并在最严格的距离阈值下，针头插入和提取的成功率分别达到了80%和68%。
cs.RO / 3 / 2607.27508
Corrigible Assistance in One Round: Pragmatic-Pedagogic Best Response
可纠正的单轮辅助：务实-教学最佳响应
Abstract
Assistance games formalize human-robot collaboration under asymmetric information: the human knows the goal, while the robot must infer it from observation and interaction in order to assist effectively. In general, computing optimal assistance game strategies online is intractable, since exact solutions require planning in a POMDP. We identify a class of assistance games in which pragmatic-pedagogic reasoning resolves goal uncertainty in a single time step, rendering the full-horizon game exactly solvable by a tractable best-response procedure. Within this class, we show that mainstream inverse optimal control exhibits an inference ceiling that hinders alignment, while pragmatic-pedagogic reasoning overcomes this barrier by immediately disambiguating goals through actions that look equivalent under task execution alone. Finally, we validate our theoretical results and proposed method on a simple collaborative block-building example.
Chinese Translation
辅助游戏形式化了在信息不对称下的人机协作：人类知道目标，而机器人必须通过观察和互动推断目标，以便有效地提供帮助。一般来说，在线计算最优辅助游戏策略是不可处理的，因为精确解需要在部分可观测马尔可夫决策过程（POMDP）中进行规划。我们识别出一类辅助游戏，其中务实-教学推理在单个时间步内解决了目标不确定性，使得全时域游戏可以通过可处理的最佳响应程序精确求解。在这一类中，我们展示了主流的逆最优控制存在推理上限，阻碍了对齐，而务实-教学推理通过立即通过在任务执行中看似等效的动作消除歧义，从而克服了这一障碍。最后，我们在一个简单的协作搭建积木的例子中验证了我们的理论结果和提出的方法。
cs.RO / 4 / 2607.27511
Failure Detection for Surgical Robot Imitation Policies via Flow-Matching World Modeling
通过流匹配世界建模进行外科机器人模仿策略的故障检测
Abstract
Imitation learning has shown increasing promise for autonomous robotic surgery, yet safe deployment remains challenging due to the safety-critical nature of surgical tasks and the complexity and variability of surgical environments. Failure detection is therefore an essential safeguard, but its development remains difficult due to the challenges of scarce failure data, highly variable manipulation dynamics, and the need to balance missed detections against disruptive false alarms. To address these challenges, we introduce FoMo-FD (Flow-Matching World Model for Failure Detection), a failure detection method that learns nominal short-horizon visual dynamics with an action-conditioned flow-matching world model. FoMo-FD scores the inverse-transport nonconformity of observed endpoint latents, enabling window-level detection of visual-action inconsistencies without requiring failure demonstrations. Detection thresholds are obtained by conformal calibration on successful executions, yielding task-specific alarms without assuming future failure types. We evaluate FoMo-FD on four surgically relevant manipulation tasks with twenty failure modes across simulation and real-world experiments using the da Vinci Research Kit (dVRK). Results show that FoMo-FD outperforms observation-level anomaly baselines and a prediction-error variant of the same world model, with the wrist-camera view achieving the strongest performance, including a 96.6% failure detection rate (FDR) at a 1.3% false alarm rate (FAR).
Chinese Translation
模仿学习在自主机器人外科手术中展现出越来越大的潜力，但由于外科任务的安全关键性质以及外科环境的复杂性和变异性，安全部署仍然面临挑战。因此，故障检测是一个重要的安全保障，但由于缺乏故障数据、高度可变的操作动态以及需要平衡漏检与干扰性误报的挑战，其开发仍然困难。为了解决这些挑战，我们提出了FoMo-FD（用于故障检测的流匹配世界模型），这是一种通过动作条件流匹配世界模型学习名义短期视觉动态的故障检测方法。FoMo-FD对观察到的端点潜变量的逆传输非一致性进行评分，使得在不需要故障示范的情况下能够进行视觉-动作不一致性的窗口级检测。检测阈值通过对成功执行进行符合性校准获得，从而在不假设未来故障类型的情况下产生特定任务的警报。我们在四个与外科相关的操作任务上评估了FoMo-FD，这些任务包含二十种故障模式，实验包括使用达芬奇研究套件（dVRK）的仿真和真实世界实验。结果表明，FoMo-FD的表现优于观察级别的异常基线和同一世界模型的预测误差变体，其中腕部摄像头视角的表现最佳，故障检测率（FDR）达到96.6%，假警报率（FAR）为1.3%。
cs.RO / 5 / 2607.27549
Cross-Embodiment Transfer via Behavior-Aligned Representations
通过行为对齐表示实现跨体现转移
Abstract
Recent progress in large-scale imitation learning for robot manipulation has been driven by leveraging datasets across a wide range of robot embodiments. However, achieving significant cross-embodiment transfer is often still challenging. In this work, we study the role of using behavior-aligned representations (e.g., object bounding boxes, language motions, end-effector traces of robot motion) in vision-language-action (VLA) models to promote cross-embodiment transfer. We hypothesize that by possessing invariances across embodiments while being predictive of robot actions, these representations can help unify large-scale cross-embodiment data to enhance transfer. To assess our hypothesis, we develop a simulation-based benchmark designed to assess transfer with diverse cross-embodiment data to new embodiments. Using this benchmark, we compare different representations and ways of incorporating them. We identify that end-effector traces can be particularly beneficial for transfer, representations are generally more useful with larger prior datasets, and can be used to benefit from action-free data. We also demonstrate that they can enhance sim-to-real cross-embodiment transfer, improving task completion progress of real robot policies pre-trained on simulation data by 28%. We provide videos of our evaluations at our website: https://ajaysridhar.com/barx/.
Chinese Translation
近年来，基于大规模模仿学习的机器人操作进展得益于跨多种机器人体现的数据集的利用。然而，实现显著的跨体现转移仍然面临挑战。在本研究中，我们探讨了在视觉-语言-动作（VLA）模型中使用行为对齐表示（例如，物体边界框、语言动作、机器人运动的末端执行器轨迹）在促进跨体现转移中的作用。我们假设，通过在不同体现间保持不变性，同时对机器人动作具有预测能力，这些表示可以帮助统一大规模跨体现数据，从而增强转移效果。为了验证我们的假设，我们开发了一个基于仿真的基准，旨在评估使用多样的跨体现数据对新体现的转移效果。利用该基准，我们比较了不同的表示方式及其整合方法。我们发现，末端执行器轨迹在转移中尤其有利，表示在较大先前数据集的情况下通常更有用，并且可以利用无动作数据。我们还展示了它们可以增强仿真到真实的跨体现转移，提高了在仿真数据上预训练的真实机器人策略的任务完成进度，提升幅度达28%。我们在网站上提供了评估视频：https://ajaysridhar.com/barx/。
cs.RO / 6 / 2607.27597
A Systems Engineering Framework for Vision-Language-Enabled UAV Triage and Disaster Response
面向视觉语言的无人机分诊与灾难响应的系统工程框架
Abstract
Recent advances in Vision Language Models (VLMs) have created new opportunities for disaster response, where responders must interpret large volumes of sensor data under time pressure. Current VLM applications include social media monitoring for situational awareness, generation of draft action plans, and translation of technical alerts into public-facing messages. While these efforts can accelerate information flow, they remain largely limited to decision-support roles. Such approaches can increase operator burden because humans must still translate outputs into coordinated actions across teams and robotic assets. This study explores the viability of embedding VLMs as coordination agents within the human-UAV loop. The proposed architecture integrates natural language interaction, mission-level task coordination, software-in-the-loop implementation, and communication aligned with the Incident Command System (ICS). Rather than functioning solely as advisory tools, VLMs facilitate communication between human operators, mission control logic, and UAV task execution. The framework was developed using a Model-Based Systems Engineering (MBSE) approach, with use case and block definition diagrams representing system roles, internal structure, and component interactions. Three key elements, the VLM Coordinator Agent, UAV Mission Control, and Task Allocator, were implemented within an integrated simulation and control environment. A preliminary human-factors evaluation with seven participants showed reduced perceived workload across mental demand, effort, and frustration, along with high ratings for AI trust and communication clarity. By integrating MBSE, software-in-the-loop testing, and human-factors evaluation, this work advances scalable human-autonomy teaming for high-stakes disaster response, with broader implications for aerospace autonomy and civil safety.
Chinese Translation
最近在视觉语言模型（VLMs）方面的进展为灾难响应创造了新的机会，响应者必须在时间压力下解读大量传感器数据。目前的VLM应用包括社交媒体监测以获取情境意识、生成初步行动计划以及将技术警报翻译为面向公众的信息。尽管这些努力可以加速信息流动，但它们仍然主要局限于决策支持角色。这种方法可能增加操作员的负担，因为人类仍需将输出转化为跨团队和机器人资产的协调行动。本研究探讨了将VLM嵌入人机无人机（UAV）循环作为协调代理的可行性。所提出的架构集成了自然语言交互、任务级别的任务协调、软件在环（software-in-the-loop）实施，以及与事件指挥系统（Incident Command System, ICS）对齐的通信。VLM不仅仅作为顾问工具，而是促进人类操作员、任务控制逻辑和无人机任务执行之间的沟通。该框架采用基于模型的系统工程（Model-Based Systems Engineering, MBSE）方法开发，使用用例和区块定义图表示系统角色、内部结构和组件交互。三个关键元素，即VLM协调代理、无人机任务控制和任务分配器，在集成的仿真和控制环境中实施。对七名参与者进行的初步人因评估显示，在心理需求、努力和挫折感方面感知工作负荷降低，同时对人工智能信任和沟通清晰度的评分较高。通过整合MBSE、软件在环测试和人因评估，本研究推动了高风险灾难响应中可扩展的人机自主团队合作，并对航空自主和公共安全具有更广泛的影响。
cs.RO / 7 / 2607.27627
Arm2Air: Cross-Embodiment Skeleton Transfer for 3D Relay Formation
Arm2Air：用于3D中继形成的跨体现骨架转移
Abstract
Unmanned aerial vehicle (UAV) relay networks can restore connectivity after communication infrastructure is damaged. Urban relay placement is difficult because line-of-sight blockage, communication range, altitude, and three-dimensional obstacles must be considered jointly. Arm2Air transfers obstacle-avoidance skeletons from robot arms to UAV relay placement through cross-embodiment transfer. Source-domain robot-arm motions from a pretrained Neural MP model are converted into ordered skeletons that pretrain a transformer-based transfer platform, which is then adapted to the UAV domain using limited target data and Low-Rank Adaptation. The transferred skeleton initializes a relay chain that is refined for connectivity, bottleneck capacity, delay, and movement cost. On nine held-out high-clutter 3D urban maps, Arm2Air reduced median end-to-end planning runtime by 64.9 percent relative to the fastest conventional planner. On the high-obstruction group of a separate 30-map dense urban holdout, it increased bottleneck capacity by 32.6 percent, reduced capacity variance by 74.7 percent, reduced maximum hop distance by 13.2 percent, reduced hop-distance variance by 75.2 percent, and reduced relay displacement by 16.9 percent relative to IMPC-MD. With only three target-domain training maps, Arm2Air reduced relay-position root mean square error by 53.6 percent relative to training from scratch while updating 0.134 million parameters, compared with 1.383 million for Scratch and Full Fine-tuning. These results demonstrate computationally and data-efficient UAV relay placement and suggest a broader principle for transferring ordered structural priors across heterogeneous embodied tasks.
Chinese Translation
无人机（UAV）中继网络可以在通信基础设施受损后恢复连接。由于需要综合考虑视距阻塞、通信范围、高度和三维障碍物，城市中继布置变得困难。Arm2Air通过跨体现转移将障碍规避骨架从机器人手臂转移到无人机中继布置。来自预训练的神经MP模型的源域机器人手臂运动被转换为有序骨架，以预训练基于变换器的转移平台，然后使用有限的目标数据和低秩适应（Low-Rank Adaptation）将其调整到无人机领域。转移的骨架初始化了一个中继链，该链经过优化以提高连接性、瓶颈容量、延迟和移动成本。在九个高杂乱的3D城市地图上，Arm2Air相较于最快的传统规划器将中位端到端规划运行时间减少了64.9%。在一个单独的30张地图的高阻碍组中，相较于IMPC-MD，它将瓶颈容量提高了32.6%，将容量方差减少了74.7%，将最大跳跃距离减少了13.2%，将跳跃距离方差减少了75.2%，并将中继位移减少了16.9%。仅使用三个目标域训练地图，Arm2Air相较于从头开始训练将中继位置的均方根误差减少了53.6%，同时更新了13.4万参数，而Scratch和Full Fine-tuning则更新了138.3万参数。这些结果展示了计算和数据高效的无人机中继布置，并提出了一个更广泛的原则，用于在异构体现任务之间转移有序结构先验。
cs.RO / 8 / 2607.27690
LabEvolver: Training-Free Experience Evolution for Safe and Grounded Wet-Lab Agents
LabEvolver：无训练经验演化框架用于安全且稳健的湿实验室智能体
Abstract
We introduce LabEvolver, a training-free framework that equips safe and grounded wet-lab agents with episodic memory from execution experience. LabEvolver couples a state-grounded inner trial loop for adaptive perception, online planning, and safety validation with an outer evolution loop that distills completed trajectories into reusable skill, strategy, and safety experience. On robotic solution-preparation tasks, LabEvolver demonstrates real-world feasibility, reducing pH-regulation completion time and safety-gate intercepts by 48.2% and 60.0%, respectively. On ALFWorld, it further improves cumulative success rate within 20 steps from 76.2% with ReAct to 91.4% over 500 continual tasks, showing generality beyond wet-lab settings. These results support learn-by-doing experience evolution as a feasible path toward closed-loop automated scientific discovery. The project page is available at https://github.com/AndyGao6186/LabEvolver.
Chinese Translation
我们介绍了LabEvolver，一个无训练的框架，旨在为安全且稳健的湿实验室智能体提供来自执行经验的情节记忆。LabEvolver结合了一个基于状态的内部试验循环，用于自适应感知、在线规划和安全验证，以及一个外部演化循环，将完成的轨迹提炼为可重用的技能、策略和安全经验。在机器人解决方案准备任务中，LabEvolver展示了其在现实世界中的可行性，分别将pH调节完成时间和安全门拦截减少了48.2%和60.0%。在ALFWorld上，它进一步将20步内的累计成功率从使用ReAct的76.2%提高到500个连续任务中的91.4%，显示出其在湿实验室环境之外的通用性。这些结果支持通过实践经验演化作为实现闭环自动化科学发现的可行路径。项目页面可访问 https://github.com/AndyGao6186/LabEvolver。
cs.RO / 9 / 2607.27713
Write-Safe Flow Field Mapping under Ambiguous Onboard Sensing and Localization Drift
在模糊的车载传感和定位漂移下的安全流场映射
Abstract
Mobile robots can infer local flow structure from onboard sensing, but a locally plausible estimate is not always safe to write into a global map. Similar flow structures may produce ambiguous observations, while localization drift causes predicted patches to be written at incorrect locations. Repeated misregistered updates then accumulate into persistent ghost structures. We address this failure mode with a map-reference-aware conservative fusion framework. The model predicts a local velocity patch and a learned write-safety score that continuously attenuates uncertain map updates while permitting initialization when no reliable map reference is available. Across synthetic jet and crossflow environments, the proposed method reduces average ghost contamination by 42% relative to ungated fusion. A zero-shot hardware replay using real pressure and optical-flow measurements from a thruster wake further reduces ghost contamination by 39% while retaining 81% map coverage. These results show that safe map writing is critical for flow mapping under ambiguous sensing and localization drift.
Chinese Translation
移动机器人可以通过车载传感推断局部流动结构，但局部可行的估计并不总是安全地写入全局地图。相似的流动结构可能产生模糊的观测，而定位漂移则导致预测的区域被写入错误的位置。重复的错误注册更新会累积成持久的虚假结构。我们通过一个地图参考感知的保守融合框架来解决这一失败模式。该模型预测局部速度区域和一个学习的写入安全评分，该评分在没有可靠地图参考的情况下，持续减弱不确定的地图更新，同时允许初始化。在合成喷流和横流环境中，所提出的方法相较于未门控融合将平均虚假污染减少了42%。使用来自推进器尾流的真实压力和光流测量进行的零样本硬件重放进一步将虚假污染减少了39%，同时保持了81%的地图覆盖率。这些结果表明，在模糊传感和定位漂移下，安全的地图写入对流场映射至关重要。
cs.RO / 10 / 2607.27782
RedFlow: Redirect Failure into Action-Level Corrections for Flow-matching VLA Policy
RedFlow：将失败重定向为流匹配 VLA 策略的动作级纠正
Abstract
Flow-matching Vision-Language-Action (VLA) policies have shown strong potential for robotic manipulation but often suffer from compounding errors caused by distribution shifts during deployment. While offline reinforcement learning (RL) provides a practical way to improve deployed policies using rollout data, existing methods either ignore failure data or exploit it only at the trajectory level, resulting in low learning efficiency and persistent errors. We propose **RedFlow**, a fine-grained offline RL framework that redirects failure experiences into action-level corrective supervision for flow-matching VLA policies. RedFlow consists of two key components: (1) a **Context-Aware Corrective Matching** mechanism that identifies failure-inducing actions and retrieves successful alternatives from similar contexts as corrective targets, and (2) an **Adaptive Redirection Objective** that jointly reinforces successful actions, suppresses undesirable ones, and redirects recoverable failures toward corrective targets. By converting both successful and failed experiences into dense supervision, RedFlow enables robust recovery learning from mixed-quality data. Experiments on the LIBERO benchmark and three real-world manipulation tasks show that RedFlow consistently outperforms state-of-the-art offline RL baselines, improving the real-world success rate from 56.7% to 74.7%. It also matches strong on-policy methods (PPO, GRPO, and DDPO) while requiring roughly an order of magnitude fewer training samples.
Chinese Translation
流匹配视觉-语言-动作（VLA）策略在机器人操控中展现出强大的潜力，但在部署过程中常常受到分布变化引起的累积错误的困扰。虽然离线强化学习（RL）为利用回放数据改善已部署策略提供了一种实用的方法，但现有方法要么忽视失败数据，要么仅在轨迹层面利用这些数据，导致学习效率低下和持续错误。我们提出了 **RedFlow**，一种细粒度的离线 RL 框架，将失败经验重定向为流匹配 VLA 策略的动作级纠正监督。RedFlow 由两个关键组件组成：（1）**上下文感知纠正匹配**机制，识别导致失败的动作，并从相似上下文中检索成功的替代方案作为纠正目标；（2）**自适应重定向目标**，共同强化成功动作，抑制不良动作，并将可恢复的失败重定向到纠正目标。通过将成功和失败的经验转化为密集监督，RedFlow 能够从混合质量数据中实现稳健的恢复学习。在 LIBERO 基准测试和三个真实世界操控任务上的实验表明，RedFlow 始终优于最先进的离线 RL 基线，将真实世界的成功率从 56.7% 提高到 74.7%。它的表现也与强大的在线方法（PPO、GRPO 和 DDPO）相匹配，同时所需的训练样本大约减少了一个数量级。
cs.RO / 11 / 2607.27784
DexDirect: Direct Kinesthetic Arm Guidance for Efficient Dexterous Demonstration Collection
DexDirect：高效灵巧演示收集的直接动觉臂引导
Abstract
Scalable collection of dexterous manipulation demonstrations remains a major bottleneck for robot learning. High-fidelity interfaces often require costly hardware and extensive setup, while low-setup, low cost alternatives tend to provide less precise control and impose greater cognitive workload on operators. We present DexDirect, a direct kinesthetic arm guidance for efficient dexterous demonstration collection. The operator drags a 6-DoF gravity-compensated robot arm directly by a handle, while a single webcam retargets operator's other hand onto a 16 joints 13-DoF dexterous robot hand. User studies suggest DexDirect collects 17.2x and 3.2x more successful demonstrations compared to purely vision (AnyTeleop) and pose-tracking (TeleDex) baselines. An adapted NASA-TLX shows DexDirect greatly reduces mental demand, effort, and frustration, despite raising physical demand. A diffusion policy trained on DexDirect demonstrations reaches a 90% success rate on a cube pick-and-place task. These results suggest that direct kinesthetic arm guidance combined with vision-based hand retargeting provides an efficient low-setup and scalable interface for collecting dexterous manipulation demonstrations
Chinese Translation
灵巧操作演示的可扩展收集仍然是机器人学习的一个主要瓶颈。高保真接口通常需要昂贵的硬件和广泛的设置，而低设置、低成本的替代方案往往提供较少的精确控制，并对操作员施加更大的认知负担。我们提出了DexDirect，一种用于高效灵巧演示收集的直接动觉臂引导。操作员通过手柄直接拖动一个6自由度的重力补偿机器人手臂，同时一台单一的网络摄像头将操作员的另一只手重新定位到一个具有16个关节和13自由度的灵巧机器人手上。用户研究表明，DexDirect相比于纯视觉（AnyTeleop）和姿态跟踪（TeleDex）基线，收集了17.2倍和3.2倍的成功演示。经过调整的NASA-TLX显示，尽管提高了身体需求，DexDirect显著降低了心理需求、努力程度和挫败感。基于DexDirect演示训练的扩散策略在立方体的抓取与放置任务中达到了90%的成功率。这些结果表明，直接动觉臂引导结合基于视觉的手重新定位提供了一种高效、低设置且可扩展的接口，用于收集灵巧操作演示。
cs.RO / 12 / 2607.27881
RoboBRIDGE: A Modular Framework for Bridging Policies to Robust Real-World Robotic Agents
RoboBRIDGE：一个用于连接政策与强健现实世界机器人代理的模块化框架
Abstract
Vision-Language-Action (VLA) models have attracted growing interest as a scalable approach to robotic manipulation. While these models are effective action predictors, deploying them as robotic agents exposes critical gaps: no mechanism for failure recovery, inconsistent execution over long horizons, and limited robustness to shifts in observations, tasks, or embodiments. Existing solutions address these limitations individually through model retraining or environment-specific modules, yet what is needed is a general framework that systematically transforms a pretrained VLA into a robotic agent. We present RoboBRIDGE, a modular framework that provides an orchestration layer over five coordinated modules, namely Monitor, Perceptor, Planner, Controller, and Robot Interface, to compose robust robotic agents from off-the-shelf components, including pretrained VLAs. The Monitor pairs rapid failure detection with hierarchical recovery to correct errors before they cascade. When the environment diverges from the current plan, the Planner triggers replanning while the Perceptor updates scene understanding asynchronously, avoiding execution stalls. Within the Controller, primitive skill fine-tuning factors manipulation into domain-invariant primitives with dedicated LoRA adapters, reducing sensitivity to domain shifts when a VLA is used. Across LIBERO, RoboCasa, and real-world case studies spanning multiple robot platforms and VLA backbones, RoboBRIDGE consistently outperforms both standalone policies and prior augmented VLA deployments. These results suggest that reliable robotic agency does not arise from scaling action predictors alone, but from structured orchestration around them.
Chinese Translation
视觉-语言-动作（VLA）模型作为一种可扩展的机器人操作方法，越来越受到关注。尽管这些模型在动作预测方面表现出色，但将其部署为机器人代理时暴露出了一些关键缺陷：缺乏故障恢复机制、长时间执行中的不一致性以及对观察、任务或实现的变化的有限鲁棒性。现有解决方案通过模型重训练或特定环境模块单独解决这些局限性，但所需的是一个系统性地将预训练的VLA转化为机器人代理的通用框架。我们提出了RoboBRIDGE，这是一个模块化框架，提供了一个协调五个模块的编排层，分别是监控器（Monitor）、感知器（Perceptor）、规划器（Planner）、控制器（Controller）和机器人接口（Robot Interface），以便从现成组件（包括预训练的VLA）中构建强健的机器人代理。监控器将快速故障检测与分层恢复相结合，以在错误蔓延之前进行纠正。当环境与当前计划发生偏离时，规划器触发重新规划，而感知器异步更新场景理解，避免执行停滞。在控制器内部，原始技能微调将操作分解为与领域无关的原语，并配备专用的LoRA适配器，从而减少使用VLA时对领域变化的敏感性。在LIBERO、RoboCasa和涵盖多个机器人平台及VLA骨干的现实案例研究中，RoboBRIDGE始终优于独立政策和先前增强的VLA部署。这些结果表明，可靠的机器人代理并非仅通过扩展动作预测器而产生，而是通过围绕它们的结构化编排实现的。
cs.RO / 13 / 2607.27890
Static In, Dynamic Out: Counterfactual Action Augmentation for Moving Object Manipulation
静态输入，动态输出：用于移动物体操作的反事实动作增强
Abstract
Visuomotor policies have advanced on manipulation tasks where the target object stays static during execution, but real deployments break this assumption: parts drift on conveyors and fruits sway in the wind. We introduce Static In, Dynamic Out (SIDO), a counterfactual action augmentation that enables a policy trained only on static object demonstrations to adapt to unseen object motion at test time. Our key idea is to factorize moving object manipulation into two sub-problems: predicting where the object will be, and reaching that predicted pose. SIDO displaces the object to a counterfactual future position and morphs the demonstrated action chunk to preserve the hand-object relative pose, yielding a goal-conditioned policy. At deployment an object pose predictor supplies the future position. Across three simulated tasks (Mug, Square, Stack) under five object motion patterns and two real-world tasks (Gantry, Peachtree), SIDO improves moving object success over the baselines while preserving static object performance. Project website: https://sido-staticindynamicout.github.io/.
Chinese Translation
视觉运动策略在目标物体在执行过程中保持静态的操作任务上取得了进展，但实际部署打破了这一假设：部件在传送带上漂移，水果在风中摇摆。我们提出了静态输入，动态输出（Static In, Dynamic Out，SIDO），这是一种反事实动作增强方法，使得仅在静态物体演示上训练的策略能够在测试时适应未见的物体运动。我们的关键思想是将移动物体操作分解为两个子问题：预测物体将处于何处，以及到达该预测姿态。SIDO将物体移至一个反事实的未来位置，并对演示的动作片段进行变形，以保持手与物体之间的相对姿态，从而产生一个目标条件的策略。在部署时，物体姿态预测器提供未来位置。在三个模拟任务（杯子、方块、堆叠）下的五种物体运动模式以及两个真实世界任务（龙门架、桃树）中，SIDO在移动物体的成功率上优于基线，同时保持静态物体的性能。项目网站：https://sido-staticindynamicout.github.io/
cs.RO / 14 / 2607.27922
Learning Social Robot Navigation By Sensing Human Legs
通过感知人类腿部学习社交机器人导航
Abstract
Robots navigating among pedestrians typically sense their surroundings with a 2D LiDAR mounted close to the ground. At that height, the sensor mostly sees moving legs rather than whole people, yet most learning-based navigation methods still treat pedestrians as simple shapes like circles. This paper addresses that gap with CALF (Convolutional Attention for Leg Features), an end-to-end neural architecture that combines convolutional layers, attention, and MLP to interpret leg motion directly from LiDAR scans and produce safe navigation commands. The CALF policy is trained using deep reinforcement learning algorithms within LegNav, a custom lightweight 2D simulator that combines 2D LiDAR ray tracing with a novel pedestrian gait model. The resulting policy is compared against classical and learning-based baselines in terms of navigation performance and social compliance. The approach is validated through real-world experiments via zero-shot deployment on a TurtleBot 4, yielding smooth and socially compliant trajectories. Written in JAX, the LegNav simulator enables the training of a deployment-ready CALF policy in under an hour on a single consumer GPU.
Chinese Translation
在行人中导航的机器人通常使用安装在接近地面的2D激光雷达（LiDAR）来感知周围环境。在这个高度，传感器主要看到的是移动的腿部而不是完整的人体，然而大多数基于学习的导航方法仍然将行人视为简单的形状，如圆形。本文通过CALF（Convolutional Attention for Leg Features）来解决这一问题，CALF是一种端到端的神经网络架构，结合了卷积层、注意力机制和多层感知器（MLP），直接从LiDAR扫描中解释腿部运动并生成安全的导航指令。CALF策略使用深度强化学习算法在LegNav中进行训练，LegNav是一个定制的轻量级2D模拟器，结合了2D激光雷达光线追踪和新颖的行人步态模型。所得到的策略在导航性能和社会合规性方面与经典和基于学习的基线进行了比较。该方法通过在TurtleBot 4上进行零样本部署的真实世界实验进行了验证，产生了平滑且符合社会规范的轨迹。LegNav模拟器使用JAX编写，能够在单个消费级GPU上在不到一小时的时间内训练出可部署的CALF策略。
cs.RO / 15 / 2607.28045
RaDiVe: Robust 4D Radar Odometry with Distance-Bounded NDT and Velocity-Discrepancy Point Uncertainty
RaDiVe：基于距离约束的NDT和速度差异点不确定性的鲁棒4D雷达里程计
Abstract
Recent advances in 4D radar enable robust perception in adverse weather; however, the inherent sparsity, noise, and limited positional precision of radar point clouds pose significant challenges for registration-based odometry. In this letter, we propose RaDiVe, a 4D radar odometry framework designed to improve the accuracy and robustness of radar point-cloud registration. We introduce a distance-bounded Normal Distributions Transform (NDT), which improves optimization stability and computational efficiency by restricting the correspondence search to near-distance voxel pairs. To mitigate measurement ambiguity, we propose a velocity-discrepancy point uncertainty model that weights each input 4D radar point according to the discrepancy between its measured Doppler radial velocity and the radial velocity predicted from the estimated ego-velocity. Furthermore, we incorporate Signed Distance Function (SDF)-based surface point extraction via implicit neural mapping to construct a geometrically consistent and noise-filtered local submap. Evaluations across multiple public datasets demonstrate that RaDiVe outperforms existing 4D radar odometry baselines by 44.4% in translational Absolute Trajectory Error (ATE) and 21.3% in rotational ATE on average, while maintaining real-time performance. The source code will be made publicly available to the robotics community: https://github.com/to-be-open-sourced.
Chinese Translation
近年来4D雷达的进展使得在恶劣天气条件下实现鲁棒感知成为可能；然而，雷达点云固有的稀疏性、噪声以及有限的位置信息精度对基于配准的里程计提出了重大挑战。在本文中，我们提出了RaDiVe，一个旨在提高雷达点云配准准确性和鲁棒性的4D雷达里程计框架。我们引入了一种距离约束的正态分布变换（Normal Distributions Transform, NDT），通过将对应关系搜索限制在近距离体素对，从而提高了优化的稳定性和计算效率。为了减轻测量模糊性，我们提出了一种速度差异点不确定性模型，根据测量的多普勒径向速度与从估计的自我速度预测的径向速度之间的差异，对每个输入的4D雷达点进行加权。此外，我们通过隐式神经映射结合基于符号距离函数（Signed Distance Function, SDF）的表面点提取，构建了一个几何一致且经过噪声过滤的局部子图。对多个公共数据集的评估表明，RaDiVe在平移绝对轨迹误差（Absolute Trajectory Error, ATE）上平均优于现有的4D雷达里程计基准44.4%，在旋转ATE上优于21.3%，同时保持实时性能。源代码将向机器人社区公开： https://github.com/to-be-open-sourced.
cs.RO / 16 / 2607.28198
UniCross: Unified Cross-Skill Dexterous Manipulation Synthesis
UniCross：统一的跨技能灵巧操作合成
Abstract
Many dexterous manipulation tasks require the object to remain securely held throughout the interaction. From the perspective of hand-object relational motion, such manipulation comprises four canonical skills: grasping, relocation, in-hand rotation, and in-hand translation. Human hands flexibly compose these skills to accomplish complex tasks. Existing approaches, however, model these skills separately with skill-specific action constraints, objectives, or even dedicated hand morphologies, which breaks the compatibility and continuity required for long-horizon composition. In this work, we present a unified framework that models all four skills in a single formulation that shares the same state and action spaces and a common objective structure. This formulation enables straightforward distillation of a single cross-skill policy that performs strongly on every skill, generalizes to unseen objects, stays robust to disturbances, and chains skills seamlessly into long-horizon manipulation. The framework also transfers effectively across different hand morphologies. Overall, our results suggest that different dexterous manipulation skills can be viewed as instantiations of a shared task formulation, revealing the intrinsic consistency across different behaviors.
Chinese Translation
许多灵巧操作任务要求在整个交互过程中物体保持稳固的抓握。从手与物体的关系运动的角度来看，这种操作包括四种经典技能：抓取、重新定位、手内旋转和手内平移。人类手灵活地组合这些技能以完成复杂任务。然而，现有的方法通常将这些技能单独建模，采用特定技能的动作约束、目标，甚至专用的手形态，这破坏了长时间组合所需的兼容性和连续性。在本研究中，我们提出了一个统一的框架，该框架在单一的公式中建模所有四种技能，分享相同的状态和动作空间以及共同的目标结构。这一公式使得能够直接提炼出一个单一的跨技能策略，该策略在每种技能上表现出色，能够推广到未见过的物体，对干扰保持鲁棒，并将技能无缝地串联成长时间的操作。该框架在不同手形态之间也能有效迁移。总体而言，我们的结果表明，不同的灵巧操作技能可以视为共享任务公式的实例，揭示了不同行为之间的内在一致性。
cs.RO / 17 / 2607.28256
When Robots Exchange Meaning: A Demo of Goal-Oriented Semantic Communications for Collaborative Robotics
当机器人交换意义时：面向目标的语义通信在协作机器人中的演示
Abstract
Collaborative robotics is a representative task-oriented 6G use-case, where communication quality should be reflected in mission execution, environment understanding, and closed-loop operation rather than packet delivery alone. This demo paper presents a robot-edge semantic communication (SemCom) testbed integrating robot-side visual compression, edge-side semantic mapping, and dashboard-based mission interaction. A mobile robot equipped with RGB-D sensing and LiDAR runs ROS 2, while a Jetson Orin edge node performs reconstruction, RTAB-Map mapping, semantic object handling, and browserbased visualization. As an initial proof of concept, RGB frames are encoded on the robot into VQ-VAE tokens using an ONNX Runtime encoder and reconstructed on the edge using a PyTorch decoder. A 320 X 240 image is represented by an 80 X 60 token grid with a packed payload of 5400 bytes, corresponding to a 42.67X reduction relative to model-input RGB bytes. The reconstructed visual stream is further associated with depth, pose, and 3D mapping information to generate a semantic map for downstream robotic applications. The demo exposes the full path from semantic visual transport to object-level map interaction, and provides a practical platform for future task-aware 6G networking studies at the intersection of SemCom, embodied AI, and physical AI-enabled robotics. A video of the demo is available at https://tinyurl.com/Tos09
Chinese Translation
协作机器人是一个典型的面向任务的6G应用场景，其中通信质量应体现在任务执行、环境理解和闭环操作上，而不仅仅是数据包传输。本文演示了一个集成了机器人端视觉压缩、边缘端语义映射和基于仪表盘的任务交互的机器人边缘语义通信（SemCom）测试平台。一台配备RGB-D传感器和激光雷达的移动机器人运行ROS 2，而一个Jetson Orin边缘节点执行重建、RTAB-Map映射、语义对象处理和基于浏览器的可视化。作为初步的概念验证，RGB帧在机器人上通过ONNX Runtime编码器编码为VQ-VAE令牌，并在边缘使用PyTorch解码器进行重建。320 x 240的图像通过一个80 x 60的令牌网格表示，负载为5400字节，相对于模型输入的RGB字节减少了42.67倍。重建的视觉流进一步与深度、姿态和3D映射信息关联，以生成用于下游机器人应用的语义地图。该演示展示了从语义视觉传输到对象级地图交互的完整路径，并为未来在SemCom、具身人工智能和物理人工智能驱动的机器人交叉领域的任务感知6G网络研究提供了一个实用平台。演示视频可在https://tinyurl.com/Tos09观看。
cs.RO / 18 / 2607.28382
SemAnCorr: Semantic Anchored Correspondence for Zero-Shot Manipulation Skill Transfer
SemAnCorr：用于零-shot操作技能转移的语义锚定对应
Abstract
Transferring manipulation skills across object instances that share functionality but differ in geometry remains a fundamental challenge in robot learning. While recent correspondence methods leverage dense visual descriptors and 3D feature fields, nearest-neighbor feature matching often produces spatially incoherent correspondences that fail to recover the local geometric frames required for reliable skill transfer. We introduce SemAnCorr, a training-free framework that establishes dense correspondence by selecting semantically consistent anchor regions through joint pose-correspondence optimization and propagating these constraints over the object surface using functional maps. The resulting correspondences preserve both semantic consistency and geometric coherence, enabling object-centric manipulation skills to transfer across geometrically diverse instances. We evaluate SemAnCorr on a dense correspondence benchmark built on PartNet-Mobility, achieving 90.8% semantic accuracy in our benchmark evaluation while improving geometric coherence over recent state-of-the-art baselines. Finally, we show that these improvements translate directly into real-world manipulation performance: using a single demonstration, SemAnCorr enables substantially more reliable zero-shot manipulation skill transfer to previously unseen objects than existing correspondence methods. Videos and additional visualizations are available at [https://semancorr.github.io](https://semancorr.github.io) .
Chinese Translation
在几何形状不同但功能相似的物体实例之间转移操作技能仍然是机器人学习中的一个基本挑战。尽管最近的对应方法利用了密集的视觉描述符和3D特征场，但最近邻特征匹配往往产生空间上不一致的对应关系，无法恢复可靠技能转移所需的局部几何框架。我们提出了SemAnCorr，一个无训练的框架，通过联合姿态对应优化选择语义一致的锚定区域，并利用功能映射在物体表面传播这些约束，从而建立密集对应。所得到的对应关系既保持了语义一致性，又保持了几何一致性，使得以物体为中心的操作技能能够在几何多样的实例之间转移。我们在基于PartNet-Mobility构建的密集对应基准上评估SemAnCorr，在我们的基准评估中实现了90.8%的语义准确率，同时在几何一致性上优于最近的最先进基线。最后，我们展示了这些改进直接转化为现实世界的操作性能：通过单个演示，SemAnCorr能够比现有的对应方法更可靠地将零-shot操作技能转移到以前未见过的物体上。视频和其他可视化内容可在[https://semancorr.github.io](https://semancorr.github.io)获取。
cs.RO / 19 / 2607.28391
TacWAM: Anchor-Guided World Action Model with Mechanics-Aware Tactile Prediction
TacWAM：基于锚点引导的机械感知触觉预测世界动作模型
Abstract
World Action Models (WAMs) combine future-state prediction with robot action generation, but existing approaches largely rely on visual futures. Visual prediction captures scene structure and object motion, yet provides limited supervision for force, deformation, shear, and slip during contact-rich manipulation. This creates two design requirements: tactile futures should carry meaningful physical information, and they should not become privileged cues for action generation. We present TacWAM, a mechanics-aware tactile WAM that addresses this challenge in three steps. First, a Spatially Aligned Fusion (SAF) Tactile Encoder maps tactile appearance, dense force fields, and deformation flow into a shared latent prediction space, with bilateral force and torque reconstruction preserving global contact information. Second, a tactile history encoder provides temporal context so future tactile prediction reflects how force and deformation change beyond the current tactile observation. Third, Anchor-Guided Tri-Modal (AGT) Attention separates current visual and tactile anchors, future prediction tokens, and action tokens, allowing future tactile states to supervise training without being directly read by the action branch. We evaluate TacWAM on four real-world contact-rich manipulation tasks covering fragile grasping, sustained surface contact, and dynamic in-hand manipulation. TacWAM achieves an average success rate of 75.0%, exceeding the strongest evaluated baseline by 37.5 percentage points. Staged ablations show consistent degradation when tactile history is removed and access to future prediction targets is relaxed. These results indicate that future tactile supervision can improve contact-aware action learning when combined with informative tactile representations and deployment-consistent information constraints.
Chinese Translation
世界动作模型（WAMs）将未来状态预测与机器人动作生成相结合，但现有方法在很大程度上依赖于视觉未来。视觉预测捕捉场景结构和物体运动，但在接触丰富的操作过程中对力、变形、剪切和滑动的监督有限。这产生了两个设计要求：触觉未来应携带有意义的物理信息，并且不应成为动作生成的特权线索。我们提出了TacWAM，一种机械感知的触觉WAM，旨在通过三个步骤解决这一挑战。首先，空间对齐融合（SAF）触觉编码器将触觉外观、密集力场和变形流映射到共享的潜在预测空间中，同时双向力和扭矩重建保留了全局接触信息。其次，触觉历史编码器提供时间上下文，使未来的触觉预测反映力和变形如何超越当前触觉观察而变化。第三，锚点引导三模态（AGT）注意力将当前的视觉和触觉锚点、未来预测标记和动作标记分开，从而允许未来的触觉状态在不被动作分支直接读取的情况下监督训练。我们在四个真实世界的接触丰富操作任务上评估了TacWAM，涵盖脆弱抓取、持续表面接触和动态手内操作。TacWAM的平均成功率达到75.0%，比最强的评估基线高出37.5个百分点。分阶段消融实验表明，当移除触觉历史并放宽对未来预测目标的访问时，性能一致下降。这些结果表明，当与信息丰富的触觉表示和一致的部署信息约束相结合时，未来的触觉监督可以改善接触感知的动作学习。
cs.RO / 20 / 2607.28416
FasTac: A Curved Multispectral Vision-Based Tactile Sensor for High-Speed High-Precision 3D Shape and Force Perception
FasTac：一种基于曲面多光谱视觉的触觉传感器，用于高速高精度的三维形状和力感知
Abstract
Curved tactile fingertips for dexterous manipulation must resolve fine contact geometry, distinguish normal and tangential loads, and capture transient signals. Existing curved vision-based tactile sensors struggle to combine accurate 3D reconstruction, three-axis force estimation, and high-speed processing in a compact form. This article presents FasTac, a curved vision-based tactile sensor integrating multispectral photometric stereo, dynamic-convolution force estimation, and hardware acceleration on a field-programmable gate array (FPGA). Single-image-sensor simultaneous multispectral imaging provides spatially aligned observations for robust surface normal estimation, followed by boundary-prior fast Poisson depth reconstruction. HyperForce uses position-aware dynamic convolution to model the spatially nonuniform mechanical response of curved elastomers and estimate three-axis forces. The complete image-to-normal-force pipeline is deployed on an FPGA. Experiments show that near-infrared (NIR) illumination and the boundary prior decrease depth mean absolute error (MAE) from 0.2730 mm to 0.0415 mm; HyperForce achieves normalized mean absolute error (NMAE) values of 2.74% and 2.39% for normal and shear forces, respectively; and FPGA deployment shortens processing latency from 3.26 ms on the GPU to 1.09 ms. Multi-object reconstruction, feedback grasping, and vibration measurement validate fine geometric perception, stable force feedback, and dynamic contact sensing.
Chinese Translation
灵活操作的曲面触觉指尖必须能够解析精细的接触几何形状，区分法向载荷和切向载荷，并捕捉瞬态信号。现有的基于曲面视觉的触觉传感器在紧凑形式中难以结合准确的三维重建、三轴力估计和高速处理。本文提出了FasTac，一种集成多光谱光度立体、动态卷积力估计和在现场可编程门阵列（FPGA）上进行硬件加速的曲面视觉触觉传感器。单图像传感器的同时多光谱成像提供了空间对齐的观测数据，以进行稳健的表面法线估计，随后进行基于边界先验的快速泊松深度重建。HyperForce利用位置感知的动态卷积建模曲面弹性体的空间非均匀机械响应，并估计三轴力。完整的图像到法向力的处理流程部署在FPGA上。实验表明，近红外（NIR）照明和边界先验将深度的平均绝对误差（MAE）从0.2730毫米降低到0.0415毫米；HyperForce对于法向力和剪切力分别实现了2.74%和2.39%的归一化平均绝对误差（NMAE）；FPGA部署将处理延迟从GPU上的3.26毫秒缩短至1.09毫秒。多物体重建、反馈抓取和振动测量验证了精细几何感知、稳定力反馈和动态接触感知。
cs.RO / 21 / 2607.28443
One Future, Every Robot: Label-Efficient Collective-State Prediction with Decentralized JEPA
一个未来，每个机器人：基于去中心化JEPA的标签高效集体状态预测
Abstract
Can every robot in a swarm predict the same future collective state from only local observations and bandwidth-limited messages? We formulate this as decentralized shared-state prediction and introduce Collective-State JEPA (CS-JEPA), a recurrent joint-embedding predictive architecture whose output at every robot represents one common future token field. At deployment, each robot uses a 16-frame local history and one 64-float recurrent message per directed edge; there is no global pooling, target encoder, episode clock, or recorded future action. After pretraining without downstream collective labels, frozen representations are evaluated with ridge probes fitted on 6, 12, or 24 globally labeled episodes. Against raw-future reconstruction with the same receiver anchor and deployment capacity but 9,607 additional training-only parameters, a prospectively registered five-seed follow-up improves prediction-error and inter-robot-agreement label-budget AUC on in-distribution, ring, mutual-kNN, and unseen-size families up to 108 robots. Every effect favors CS-JEPA in 5/5 outer seeds. In a separate sealed eight-seed follow-up, matched action-conditioned predictors receive each candidate four-step plan before producing receiver-local predictive representations. CS-JEPA reduces branch-value MSE by 45.5% and improves within-context candidate-score Pearson correlation by 0.1291, with both effects favorable in 8/8 seeds, including at unseen N=32. These results support common-future JEPA targets as a label-efficient primitive for decentralized swarm prediction under topology and size shift, with additional evidence of planning-relevant value estimation.
Chinese Translation
每个机器人能否仅通过局部观察和带宽有限的信息预测相同的未来集体状态？我们将此问题表述为去中心化共享状态预测，并引入集体状态JEPA（CS-JEPA），这是一种递归联合嵌入预测架构，其输出在每个机器人处表示一个共同的未来标记场。在部署时，每个机器人使用16帧的局部历史和每个有向边一个64浮点的递归消息；没有全局池化、目标编码器、回合时钟或记录的未来动作。在没有下游集体标签的预训练后，冻结的表示通过在6、12或24个全局标记回合上拟合的岭探针进行评估。与具有相同接收锚和部署能力但增加了9,607个仅用于训练参数的原始未来重建相比，前瞻性注册的五种种子后续研究在分布内、环形、互相k最近邻和未见大小的家族中，改善了预测误差和机器人间一致性标签预算的AUC，最多可达108个机器人。所有效果在5/5外部种子中均支持CS-JEPA。在一个单独的封闭八种子后续研究中，匹配的动作条件预测器在生成接收者本地预测表示之前接收每个候选的四步计划。CS-JEPA将分支值均方误差降低了45.5%，并将上下文内候选分数的Pearson相关性提高了0.1291，这两个效果在8/8种子中均表现良好，包括在未见的N=32情况下。这些结果支持共同未来JEPA目标作为在拓扑和规模变化下进行去中心化群体预测的标签高效原语，并提供了与规划相关的价值估计的额外证据。
cs.RO / 22 / 2607.28451
Machines that know they are aging: a framework for hardware-aware autonomous intelligence
了解自身老化的机器：一种硬件感知自主智能的框架
Abstract
Autonomous systems inevitably age, yet their artificial intelligence typically assumes hardware remains in its original condition. Batteries degrade, sensors drift, processors accumulate timing errors, and memory reliability declines, creating a growing mismatch between assumed and actual capability. This can lead to agnostic collapse, where mission failure arises from accumulated hardware degradation rather than a single component fault. We propose Aging-Aware Autonomous Intelligence (AAAI), a framework that integrates hardware health directly into reasoning, planning, and mission execution. AAAI is built on three pillars: hardware self-awareness, which continuously estimates the health of power, sensing, memory, and computation subsystems using physics-of-failure models; self-adaptive reasoning, which adjusts inference complexity, planning horizon, and task priorities according to remaining hardware capability; and survival-centric intelligence, which allocates remaining operational life across mission objectives through performance optimization, resource conservation, and graceful degradation. Rather than introducing new hardware, AAAI unifies prognostics, lifecycle management, and hardware-aware computing into a closed-loop cognitive architecture. We argue that such integration is essential for autonomous systems operating in inaccessible or safety-critical environments, including space missions, marine robotics, and implantable medical devices. By enabling machines to recognize and respond to their own aging, AAAI improves resilience, extends operational lifetime, and supports safer, more graceful mission completion.
Chinese Translation
自主系统不可避免地会老化，但其人工智能通常假设硬件保持在原始状态。电池退化、传感器漂移、处理器积累时序错误以及内存可靠性下降，导致假设能力与实际能力之间的差距不断扩大。这可能导致无知崩溃，即任务失败源于累积的硬件退化而非单一组件故障。我们提出了老化感知自主智能（Aging-Aware Autonomous Intelligence, AAAI），这是一个将硬件健康直接融入推理、规划和任务执行的框架。AAAI建立在三个支柱之上：硬件自我感知，持续估计电源、传感、内存和计算子系统的健康状况，使用失效物理模型；自适应推理，根据剩余硬件能力调整推理复杂性、规划范围和任务优先级；以及以生存为中心的智能，通过性能优化、资源节约和优雅退化在任务目标之间分配剩余的操作寿命。AAAI并未引入新硬件，而是将预测、生命周期管理和硬件感知计算统一为一个闭环认知架构。我们认为，这种整合对于在不可接触或安全关键环境中运行的自主系统至关重要，包括太空任务、海洋机器人和可植入医疗设备。通过使机器能够识别并响应自身的老化，AAAI提高了韧性，延长了操作寿命，并支持更安全、更优雅的任务完成。
cs.RO / 23 / 2607.28474
TEA-AgriVLN: Traversability Estimation Alarm for Agricultural Vision-and-Language Navigation
TEA-AgriVLN：农业视觉与语言导航的可 traversability 估计警报
Abstract
Vision-and-Language Navigation in Continuous Environments (VLN-CE) requires an agent to follow a natural language instruction, predicting a sequence of low-level actions to navigate a robot from a starting point to a target location. The A2A benchmark and the AgriVLN method pioneeringly extended VLN-CE from indoor scenes to agricultural scenes, while we observed a challenging distinction: In indoor scenes, whether a zone is traversable tends to be clear to classify, such as wood floors are traversable but concrete walls are not. In agricultural scenes, however, this issue tends to be ambiguous, such as an unripe cornfield might be traversable for a robotic dog but might be non-traversable for a human. To address this issue, we propose the TEA module, which estimates the traversability of the camera image, then alarm the decision-maker for rethinking when the predicted action does not align with the traversability map. We integrate it into the AgriVLN backbone to build our TEA-AgriVLN method. When evaluated on A2A, it improves Success Rate (SR) from 0.47 to 0.54 and Navigation Error (NE) from 2.91 m to 2.70 m, showing the state-of-the-art performance in the agricultural VLN-CE domain. We further implement the ablation studies and the case study, discussing the effectiveness and limitations of TEA on different ground categories and scene classes. Code: https://github.com/AlexTraveling/TEA-AgriVLN.
Chinese Translation
连续环境中的视觉与语言导航（VLN-CE）要求代理遵循自然语言指令，预测一系列低级动作，以将机器人从起始点导航到目标位置。A2A 基准和 AgriVLN 方法开创性地将 VLN-CE 从室内场景扩展到农业场景，但我们观察到一个具有挑战性的区别：在室内场景中，区域是否可 traversable 往往容易分类，例如木地板是可 traversable 的，但混凝土墙则不是。然而，在农业场景中，这个问题往往模糊不清，例如未成熟的玉米田可能对机器人狗是可 traversable 的，但对人类则可能是不可 traversable 的。为了解决这个问题，我们提出了 TEA 模块，该模块估计相机图像的可 traversability，然后在预测的动作与可 traversability 地图不一致时提醒决策者重新思考。我们将其集成到 AgriVLN 主干中，构建了我们的 TEA-AgriVLN 方法。在 A2A 上评估时，它将成功率（SR）从 0.47 提高到 0.54，将导航误差（NE）从 2.91 米降低到 2.70 米，展示了农业 VLN-CE 领域的最先进性能。我们进一步实施了消融研究和案例研究，讨论了 TEA 在不同地面类别和场景类别上的有效性和局限性。代码：https://github.com/AlexTraveling/TEA-AgriVLN。
cs.RO / 24 / 2607.28560
X-NavDP: Generalizing Navigation Diffusion Policy to Novel Behavior and Embodiments with Group Q-score Reweighted Matching
X-NavDP：将导航扩散策略推广至新行为和具身体的群体 Q-score 重加权匹配
Abstract
Pretraining navigation diffusion policies rely on large-scale expert demonstrations. These data are typically generated by a fully-informed oracle planner suited to a single nominal robot. This limits the policy's generalization to diverse embodiments and challenging scenarios (e.g., escaping dead ends or detouring long obstacles) that demand diverse local reactive behaviors with only onboard local observations. Post-training the policy with reinforcement learning (RL) offers a principled remedy. However, previous RL for diffusion approaches lead to only marginal improvements. This is because the intractable likelihood of diffusion policies renders policy gradients unstable in addition to inefficient policy exploration. To address these challenges, we propose a data-efficient diffusion RL post-training framework - GQRM (Group Q-score Reweighted Matching). Our framework introduces two complementary designs: (i) a self-bootstrapped exploration strategy with behavior perturbation that preserves the pretrained policy prior, and (ii) a group Q-score normalization mechanism that computes per-trajectory values on each state for efficient reweighted score matching. By conducting distributed online RL training across heterogeneous embodiments, the resulting fine-tuned policy, X-NavDP, achieves state-of-the-art cross-embodiment visual navigation performance, improving the overall success rate from 61.20% to 84.28% in simulation and 10% to 65% in real-world hard cases. The code and model are publicly available at https://yty-sky.github.io/x-navdp-project-page.
Chinese Translation
预训练的导航扩散策略依赖于大规模的专家示范。这些数据通常由适用于单一名义机器人（nominal robot）的全知规划者生成。这限制了策略在多样化具身体和具有挑战性场景（例如，逃离死胡同或绕过长障碍物）中的泛化能力，这些场景需要仅依赖机载本地观测的多样化局部反应行为。通过强化学习（RL）对策略进行后训练提供了一种原则性的解决方案。然而，以前的扩散方法的 RL 训练仅导致边际改进。这是因为扩散策略的难以处理的似然性使得策略梯度不稳定，并且策略探索效率低下。为了解决这些挑战，我们提出了一种数据高效的扩散 RL 后训练框架——GQRM（Group Q-score Reweighted Matching）。我们的框架引入了两个互补设计：（i）一种自引导的探索策略，结合行为扰动，保持预训练策略的先验；（ii）一种群体 Q-score 归一化机制，计算每个状态的每条轨迹值，以实现高效的重加权评分匹配。通过在异构具身体上进行分布式在线 RL 训练，最终微调的策略 X-NavDP 实现了跨具身体视觉导航性能的最新水平，在仿真中整体成功率从 61.20% 提升至 84.28%，在现实世界的困难案例中从 10% 提升至 65%。代码和模型已公开，网址为 https://yty-sky.github.io/x-navdp-project-page。
cs.RO / 25 / 2607.28596
FA-RDP: A Frequency-Adaptive Reactive Diffusion Policy for Contact-Rich Manipulation
FA-RDP：一种用于接触丰富操控的频率自适应反应扩散策略
Abstract
In contact-rich manipulation, action multimodality and reactivity dominate different stages of a single episode. Before contact, multiple trajectories might be equally valid, making it important to preserve diverse action modes. After contact, geometric constraints and force limits narrow the solution space, while successful execution demands rapid responses to force feedback. However, standard diffusion policies use a fixed inference frequency and sampling steps throughout the episode, forcing a fundamental compromise: low-frequency, multi-step sampling better preserves pre-contact multimodality but responds slowly to force feedback, whereas high-frequency sampling improves reactivity but tends to collapse distinct pre-contact modes. To resolve this tradeoff, we present FA-RDP, a frequency-adaptive reactive diffusion policy. A shared multi-frequency visual-force Transformer predicts action chunks at both low and high frequencies, while a learned multimodality indicator dynamically selects multi-step low-frequency sampling before contact and one-step high-frequency sampling as action ambiguity decreases. We further introduce Manifold Consistency Distillation (MCD), which reparameterizes the diffusion network to predict actions on the robot action manifold while retaining DDPM-based residual supervision. Experiments on three contact-rich manipulation tasks show that FA-RDP achieves the highest success rate while preserving diverse pre-contact trajectory modes. Code and videos are available at https://fa-rdp.github.io.
Chinese Translation
在接触丰富的操控中，动作的多模态性和反应性主导着单个情节的不同阶段。在接触之前，多个轨迹可能同样有效，因此保持多样的动作模式显得尤为重要。在接触之后，几何约束和力限制缩小了解决空间，而成功执行则要求对力反馈做出快速反应。然而，标准的扩散策略在整个情节中使用固定的推理频率和采样步骤，这迫使我们在两个方面做出根本性的妥协：低频率的多步采样更好地保留了接触前的多模态性，但对力反馈的反应较慢；而高频率采样提高了反应性，但往往会导致不同的接触前模式崩溃。为了解决这一权衡，我们提出了FA-RDP，一种频率自适应的反应扩散策略。共享的多频率视觉-力Transformer在低频和高频下预测动作块，同时学习的多模态指示器动态选择在接触前的多步低频采样和在动作模糊性降低时的一步高频采样。我们进一步引入了流形一致性蒸馏（Manifold Consistency Distillation, MCD），它重新参数化扩散网络，以在机器人动作流形上预测动作，同时保留基于DDPM的残差监督。在三个接触丰富操控任务上的实验表明，FA-RDP在保持多样的接触前轨迹模式的同时，实现了最高的成功率。代码和视频可在 https://fa-rdp.github.io 获取。
cs.RO / 26 / 2607.28623
PAC-MAN: Perception-Aware CBF-RL for Whole-Body Safety in Humanoid Dodgeball
PAC-MAN：面向感知的控制屏障强化学习框架在类人 dodgeball 中的全身安全性
Abstract
We present PAC-MAN, a perception-aware CBF-RL framework that couples control-barrier safety with deployment-realistic onboard sensing for whole-body humanoid dodgeball. The deployed policy sees the ball only as segmentation-masked depth from a head-mounted camera, while training-time CBF guidance represents clearance to every body link, and an adversarial motion prior regularizes the resulting evasive reflexes. We evaluate on a controlled any-link contact benchmark with seeded throws in two regimes: single throws and a deployment loop in which the robot walks back to its station and recovers between throws. On this benchmark, the policy comes within a few points of a privileged state oracle: a fixed onboard camera alone is adequate for evasion. We find that usable barrier structure depends on perceptual observability: Joint-CBF gives the best performance with accurate ball states, degrades under fixed-camera observations when used only as training guidance, and recovers with a ball-tracking gimbal or privileged runtime filter. We therefore deploy a lightweight Link-CBF policy zero-shot on the Unitree G1 in the real world, where it tolerates imperfect perception, succeeds on 95% of throws, and uses semantic segmentation to dodge different balls.
Chinese Translation
我们提出了 PAC-MAN，一个面向感知的控制屏障强化学习（CBF-RL）框架，将控制屏障安全性与部署现实的机载传感器结合，用于类人 dodgeball 的全身控制。部署的策略仅通过头戴式摄像头看到被分割掩膜的深度信息，而训练时的控制屏障指导则代表了每个身体连接的间隙，同时对抗性运动先验则对结果的躲避反应进行了正则化。我们在一个受控的任意连接接触基准上进行了评估，该基准包含两种模式：单次投掷和一个部署循环，其中机器人在投掷之间返回其站点并恢复。在这个基准上，该策略的表现接近于一个特权状态的神谕：仅使用固定的机载摄像头就足以实现躲避。我们发现可用的控制屏障结构依赖于感知可观测性：当准确的球状态可用时，关节控制屏障（Joint-CBF）表现最佳；在仅作为训练指导使用时，固定摄像头观察下的表现会下降，但在使用球追踪云台或特权运行时过滤器时又能恢复。因此，我们在现实世界中在 Unitree G1 上零次部署了轻量级的连接控制屏障（Link-CBF）策略，该策略能够容忍不完美的感知，在95%的投掷中成功，并利用语义分割躲避不同的球。
cs.CV / 1 / 2607.27266
Theatre Chapbooks At Scale: A Statistical Comparative Analysis of Typography
大规模戏剧小册子：排版的统计比较分析
Abstract
We propose a statistical methodology that quantifies the similarity of typefaces between printed historical books. This provides a tool that accelerates philological analysis. Using character prototypes derived from clustering and aligning automatically extracted character images, the method defines a typeface distance between any two books. To produce actionable outputs, we develop an a contrario statistical framework to interpret the significance of the computed typeface distances. We apply the method to the philological study of 17 th -century Spanish printed theatre chapbooks in a quantity that exceeds the capabilities of systematic visual inspection by human experts. Our method enables the automatic comparison of Roman and Italic types extracted from different books. After validation by human experts, our method has led to new printer attributions being discovered, and former printer attributions being revised. This success strongly suggests that our method has the potential to enable digital bibliography on a larger scale than was previously possible.
Chinese Translation
我们提出了一种统计方法，量化印刷历史书籍之间字体的相似性。这为加速文献分析提供了一种工具。通过使用从自动提取的字符图像中聚类和对齐得到的字符原型，该方法定义了任意两本书之间的字体距离。为了产生可操作的输出，我们开发了一个反向统计框架，以解释计算出的字体距离的显著性。我们将该方法应用于对17世纪西班牙印刷戏剧小册子的文献研究，其数量超出了人类专家系统视觉检查的能力。我们的方法能够自动比较从不同书籍中提取的罗马体和斜体。在经过人类专家验证后，我们的方法导致了新的印刷商归属的发现，以及以前印刷商归属的修正。这一成功强烈表明，我们的方法有潜力在更大规模上实现数字书目，超越以往的可能性。
cs.CV / 2 / 2607.27278
OVEarth-Bench: Evaluating Category Breadth and Query Diversity for Open-Vocabulary Earth Observation
OVEarth-Bench：评估开放词汇地球观测的类别广度和查询多样性
Abstract
Open-vocabulary Earth observation (EO) aims to localize geospatial concepts specified in natural language rather than a fixed label set. Existing benchmarks, however, usually cover narrow category vocabularies or limited query forms. To fill this gap, we introduce OVEarth-Bench, which extends existing evaluation in two directions: category breadth, through broad hierarchical category coverage with positive and negative expressions, and query diversity, through vocabulary, referring, and reasoning queries. The benchmark supports mask and box localization under a unified zero-shot protocol. We evaluate a broad set of general and EO-specific methods. The evaluation reveals that: (1) the performance of current methods remains limited, while broader category coverage yields more stable model rankings; (2) MLLM-based methods achieve the strongest overall performance; and (3) EO-specific methods generally underperform general models and rarely match the strongest methods. These findings provide guidance for future open-vocabulary EO method design and highlight the importance of developing more realistic, diverse, high-quality, and large-scale benchmarks for reliable evaluation. Our data and evaluation package are released at https://earth-insights.github.io/OVEarth-bench.
Chinese Translation
开放词汇地球观测（EO）旨在定位用自然语言指定的地理空间概念，而不是固定标签集。然而，现有的基准通常仅涵盖狭窄的类别词汇或有限的查询形式。为填补这一空白，我们引入了OVEarth-Bench，它在两个方向上扩展了现有评估：类别广度，通过广泛的层次类别覆盖以及正负表达；查询多样性，通过词汇、引用和推理查询。该基准支持在统一的零样本协议下进行掩码和框定位。我们评估了一组广泛的通用和EO特定方法。评估结果显示：（1）当前方法的性能仍然有限，而更广泛的类别覆盖能够产生更稳定的模型排名；（2）基于MLLM的方法实现了最强的整体性能；（3）EO特定方法通常表现不如通用模型，且很少能与最强的方法相匹配。这些发现为未来开放词汇EO方法的设计提供了指导，并强调了开发更现实、多样、高质量和大规模基准以进行可靠评估的重要性。我们的数据和评估包已发布在 https://earth-insights.github.io/OVEarth-bench。
cs.CV / 3 / 2607.27292
VETO: Towards Protecting Images From Frontier AI Editing
VETO：旨在保护图像免受前沿人工智能编辑的影响
Abstract
The rise of powerful, accessible image-editing models such as FLUX.2 has brought high-fidelity editing within broad reach. Their capabilities now extend beyond localized modifications to extracting and recontextualizing objects and identities in entirely new scenes. By allowing prompt and generation tokens to attend directly to reference-image tokens, modern models blur the boundary between conventional editing and text-to-image synthesis. This expanded generative freedom also broadens the space of potential misuse, as harmful transformations are no longer confined to a predictable set of localized edits. Existing anti-edit defenses are designed to disrupt the semantic bottleneck of the reference-image encoding in legacy diffusion pipelines. However, newer editors distill reference information through joint-attention blocks, thereby often circumventing these protections. We therefore introduce VETO, a subtle anti-edit cloak that disrupts this inner mechanism through which modern models read the source image. Additionally, as existing editing benchmarks leave comprehensive recontextualizations largely untested, we introduce VetoBench, which evaluates defenses not only on conventional localized edits but also on broader contextual shifts. Across two contemporary editing models and three benchmarks, VETO consistently outperforms existing defenses while providing a stronger protection-fidelity trade-off.
Chinese Translation
强大且易于获取的图像编辑模型（如 FLUX.2）的崛起，使得高保真编辑变得触手可及。它们的能力不仅限于局部修改，还扩展到在全新场景中提取和重新语境化对象与身份。通过允许提示和生成令牌直接关注参考图像令牌，现代模型模糊了传统编辑与文本到图像合成之间的界限。这种扩展的生成自由也扩大了潜在的误用空间，因为有害的变换不再局限于可预测的局部编辑集合。现有的反编辑防御旨在破坏传统扩散管道中参考图像编码的语义瓶颈。然而，更新的编辑器通过联合注意力块提炼参考信息，从而常常规避这些保护。因此，我们提出了 VETO，一种微妙的反编辑斗篷，旨在破坏现代模型读取源图像的内部机制。此外，由于现有的编辑基准测试在全面的重新语境化方面几乎没有测试，我们引入了 VetoBench，它不仅评估传统局部编辑的防御，还评估更广泛的上下文转变。在两个现代编辑模型和三个基准测试中，VETO 始终优于现有防御，同时提供更强的保护与保真度之间的权衡。
cs.CV / 4 / 2607.27348
Bunraku: Turning a Single Illustration into an Editable Live2D Character
文乐：将单一插图转化为可编辑的 Live2D 角色
Abstract
Live2D is the dominant 2D character-animation format for anime characters and virtual avatars, representing each character as a stack of RGBA layers driven by per-layer mesh deformation. Despite its wide use in virtual streaming, mobile games, and interactive characters, authoring a Live2D model still demands weeks of manual layer separation, occlusion completion, mesh placement, and keyframing, and no prior generative method produces such a structured asset end-to-end. We present the first system that, from a single illustration, generates all the structured information a Live2D runtime consumes: ordered RGBA layers, a deformation mesh per layer, and the parameter-driven keypose vertex offsets that make the character move. Stage 1 casts layered decomposition as a layered diffusion process under a Live2D-aware organ-level taxonomy, producing an ordered RGBA stack with hidden-region completion. Stage 2 builds a content-conforming triangle mesh for each layer from its alpha channel alone, then predicts the keypose displacement field of all layers jointly: every vertex of every layer is one token, self-attention spans layer boundaries, and each displacement is factorised into a bounded direction and a log-magnitude. Joint rather than independent prediction is what makes the result a coherent character instead of separately plausible parts, and is our largest gain; scaling the network 112x yields none. On 50 held-out characters, under true generation with no teacher forcing, Stage 2 attains a per-vertex direction cosine of 0.768 (median 0.828). Because a layer's mesh derives from its alpha channel, a clothing layer can be re-textured from a natural-language instruction while the mesh and predicted animation are reused byte-for-byte. We further contribute Live2D-Bench, the first standardized benchmark for the task, and an 8,884-model Live2D corpus with layer and animation supervision.
Chinese Translation
Live2D 是用于动漫角色和虚拟化身的主流 2D 角色动画格式，通过每层网格变形将每个角色表示为 RGBA 层的堆叠。尽管在虚拟直播、移动游戏和互动角色中广泛使用，创作一个 Live2D 模型仍然需要数周的手动层分离、遮挡补全、网格放置和关键帧设置，并且没有现有的生成方法能够端到端地生成如此结构化的资产。我们提出了第一个系统，从单一插图生成 Live2D 运行时所需的所有结构化信息：有序的 RGBA 层、每层的变形网格，以及驱动角色运动的参数驱动关键姿势顶点偏移。第一阶段将分层分解视为在 Live2D 认知的器官级分类法下的分层扩散过程，生成具有隐藏区域补全的有序 RGBA 堆栈。第二阶段从每层的 alpha 通道单独构建符合内容的三角网格，然后共同预测所有层的关键姿势位移场：每层的每个顶点都是一个标记，自注意力跨越层边界，每个位移被分解为一个有界方向和一个对数幅度。共同而非独立的预测使得结果成为一个连贯的角色，而不是单独合理的部分，这是我们最大的收获；将网络规模扩大 112 倍并未带来收益。在 50 个保留角色上，在没有教师强制的真实生成下，第二阶段达到了每个顶点方向余弦为 0.768（中位数 0.828）。由于一层的网格源自其 alpha 通道，服装层可以根据自然语言指令重新纹理，同时网格和预测动画逐字重用。我们还贡献了 Live2D-Bench，这是该任务的第一个标准化基准，以及一个包含 8,884 个模型的 Live2D 语料库，具有层和动画的监督。
cs.CV / 5 / 2607.27357
Shared Semantic Codebook Distillation for Unpaired Cross-Modal Medical Classification
用于非配对跨模态医学分类的共享语义代码簿蒸馏
Abstract
Cross-modal knowledge distillation can transfer diagnostic knowledge from a strong but costly teacher modality to a cheaper and more deployable student modality. In medical image analysis, however, the two modalities are often unpaired: they are collected from different patient cohorts and occupy geometrically incompatible feature spaces. This makes instance-level distillation invalid and direct feature matching unreliable. To address these challenges, we propose Shared Semantic Codebook Distillation (SSCD), which compares teacher and student representations through a shared discrete codebook. Each image is represented as a distribution over a common, modality-agnostic vocabulary, and knowledge is transferred by aligning these distributions across modalities, both globally and class-conditionally, without requiring paired samples or directly comparable raw features. The codebook is evolved online by exponential moving average and kept diverse through entropy regularization and dead-code restart. At inference, all teacher-side and codebook modules are discarded, leaving only the student encoder and classifier. On two heterogeneous unpaired settings, OCT-to-fundus retinal disease classification and CT-to-chest-X-ray pneumonia classification, SSCD improves the student from 64.5 to 70.2 macro-F1 and from 73.8 to 76.3 macro-F1, respectively, outperforming all evaluated distillation baselines on both settings. Code and pretrained models are available at https://github.com/DillanImans/SSCD-unpaired-distillation
Chinese Translation
跨模态知识蒸馏可以将来自强大但成本高昂的教师模态的诊断知识转移到更便宜且更易于部署的学生模态。然而，在医学图像分析中，这两种模态通常是非配对的：它们来自不同的患者群体，并占据几何上不兼容的特征空间。这使得实例级蒸馏无效，直接特征匹配也不可靠。为了解决这些挑战，我们提出了共享语义代码簿蒸馏（Shared Semantic Codebook Distillation, SSCD），通过共享的离散代码簿比较教师和学生的表示。每个图像被表示为在一个共同的、与模态无关的词汇上的分布，知识通过在模态间对齐这些分布（无论是全局还是类别条件）进行转移，而无需配对样本或直接可比的原始特征。代码簿通过指数移动平均在线演化，并通过熵正则化和死代码重启保持多样性。在推理时，所有教师侧和代码簿模块被丢弃，仅保留学生编码器和分类器。在两个异质非配对设置中，OCT到眼底视网膜疾病分类和CT到胸部X光肺炎分类，SSCD将学生的宏观F1从64.5提高到70.2，从73.8提高到76.3，分别在两个设置中超越了所有评估的蒸馏基线。代码和预训练模型可在 https://github.com/DillanImans/SSCD-unpaired-distillation 获取。
cs.CV / 6 / 2607.27378
PanDent: Toward Comprehensive Tooth-Level Structure-Language Consistency in Dental Radiology
PanDent：朝着牙科放射学中全面的牙齿级结构-语言一致性迈进
Abstract
Accurate evaluation of multimodal large language models (MLLMs) in dental panoramic radiography (orthopantomogram, OPG) is limited by the lack of fine-grained, clinically reliable benchmarks that reflect expert interpretation. This work introduces PanDent, a large-scale, clinically grounded OPG benchmark built upon fine-grained, expert-validated tooth-level annotations. The dataset comprises 9,524 high-quality OPGs, each associated with comprehensive structured annotations produced by experienced dentists and further validated by an oral and maxillofacial radiologist, providing clinically reliable supervision for tooth-level diagnosis and reasoning. Clinically consistent radiology reports are constructed from expert-validated findings using clinician-defined reporting logic, establishing explicit correspondence between structured clinical evidence and free-text descriptions. This design enables evaluation of whether MLLMs generate reports that are not only linguistically coherent but also clinically consistent with expert-validated tooth-level findings. Experiments are conducted on diverse MLLMs, including state-of-the-art (SOTA) proprietary models, general-domain open-source models, and medical-specific models. Results show that current MLLMs can generate fluent reports, yet fail to produce clinically consistent descriptions, exhibiting substantial errors in fine-grained localization and tooth-level diagnosis. Fine-tuning on PanDent significantly improves structure-language consistency, substantially enhancing visual localization accuracy and diagnostic correctness, and bringing model outputs closer to expert dental interpretation. These results establish PanDent as a rigorous benchmark for evaluating tooth-level clinical reasoning in MLLMs and a valuable resource for clinically grounded dental AI.
Chinese Translation
在牙科全景放射摄影（正颌全景片，OPG）中，对多模态大语言模型（MLLMs）的准确评估受到缺乏细粒度、临床可靠的基准的限制，这些基准能够反映专家的解读。本研究介绍了PanDent，一个基于细粒度、专家验证的牙齿级注释构建的大规模、临床基础的OPG基准数据集。该数据集包含9,524幅高质量的OPG，每幅图像都附有由经验丰富的牙医生成的全面结构化注释，并经过口腔和颌面放射科医师的进一步验证，为牙齿级诊断和推理提供了临床可靠的监督。临床一致的放射学报告是基于专家验证的发现，使用临床医生定义的报告逻辑构建的，从而在结构化临床证据与自由文本描述之间建立了明确的对应关系。这一设计使得评估MLLMs生成的报告不仅在语言上连贯，而且在临床上与专家验证的牙齿级发现一致成为可能。实验在多种MLLMs上进行，包括最先进的专有模型、通用领域的开源模型和医学特定模型。结果表明，当前的MLLMs能够生成流畅的报告，但未能产生临床一致的描述，在细粒度定位和牙齿级诊断中表现出显著错误。在PanDent上进行微调显著提高了结构-语言一致性，极大地增强了视觉定位的准确性和诊断的正确性，使模型输出更接近专家的牙科解读。这些结果确立了PanDent作为评估MLLMs中牙齿级临床推理的严格基准，并为临床基础的牙科人工智能提供了宝贵的资源。
cs.CV / 7 / 2607.27380
VideoCoCo: Code-as-CoT for Physically-Consistent Video Generation via an Agentic Dual-Engine System
VideoCoCo：通过代理双引擎系统实现物理一致的视频生成的代码作为思维链
Li, Haodong, Ren, Tianfei, Ma, Xiaoxiao, Qing, Chunmei, Fang, Zhen, He, Sipeng, Guo, Ziyu, Wu, Haoyu, Tian, Juanxi, Zou, Yihang, An, Ruichuan, Jiang, Dongzhi, Yang, Boxue, Xie, Ji, Huang, Xu, Yan, Wenhao, Zou, Jialv, Yue, Zhengrong, Luo, Yaxin, Li, Xiaotong, Wang, Yuzhu, Ye, Junyan, Zhao, Jinjing, Chen, Zehui, Chen, Lin, Yan, Renye, Zhao, Feng, Heng, Pheng-Ann
Abstract
Text-to-video models have achieved remarkable visual quality, yet they still struggle to generate physically consistent dynamics because the temporal evolution of a scene must be inferred implicitly from a highly compressed text prompt. Existing chain-of-thought approaches introduce intermediate plans or visual states, but these representations are typically non-executable or temporally sparse, limiting their ability to instantiate and control the complete spatiotemporal process. To address this limitation, we introduce VideoCoCo, an agentic dual-engine framework in which executable Blender code serves as a process-level chain of thought. Given a text prompt, a coding agent synthesizes a Blender program that explicitly specifies the scene and its temporal evolution. The executable simulation engine runs the program to produce a deterministic spatiotemporal draft, which is subsequently transformed into a photorealistic video by a generative video engine through draft-conditioned editing. This decomposition separates process-level reasoning from high-fidelity visual realization. To adapt the video editor to simulated drafts, we construct VideoCoCo-3K, a curated dataset of draft-instruction-target triplets. VideoCoCo improves the OmniWeaving baseline from 0.475 to 0.558 on PhyGenBench and from 52.18 to 77.88 on VBench-2.0, achieving the best average score on both benchmarks. These results demonstrate that executable code provides an effective, controllable, and inspectable intermediate representation for physically consistent video generation.
Chinese Translation
文本到视频模型已经取得了显著的视觉质量，但仍然难以生成物理一致的动态，因为场景的时间演变必须从高度压缩的文本提示中隐式推断。现有的思维链方法引入了中间计划或视觉状态，但这些表示通常是不可执行的或时间上稀疏的，限制了它们实例化和控制完整时空过程的能力。为了解决这一限制，我们提出了VideoCoCo，一个代理双引擎框架，其中可执行的Blender代码作为过程级思维链。给定一个文本提示，编码代理合成一个Blender程序，明确指定场景及其时间演变。可执行的仿真引擎运行该程序以生成确定性的时空草图，随后通过草图条件编辑将其转化为逼真的视频。该分解将过程级推理与高保真视觉实现分开。为了使视频编辑器适应模拟草图，我们构建了VideoCoCo-3K，一个精心策划的草图-指令-目标三元组数据集。VideoCoCo在PhyGenBench上将OmniWeaving基线从0.475提升至0.558，在VBench-2.0上从52.18提升至77.88，在两个基准上均取得了最佳平均分。这些结果表明，可执行代码为物理一致的视频生成提供了一种有效、可控和可检查的中间表示。
cs.CV / 8 / 2607.27465
IGME: Efficient Chained Method Ensemble for Transferable Semantic Segmentation Attacks
IGME：用于可转移语义分割攻击的高效链式方法集成
Abstract
Semantic segmentation models are vulnerable to transferable adversarial perturbations, yet evaluating transfer attacks on dense prediction models can be computationally expensive. Existing ensemble attacks often rely on multiple surrogate models, increasing the computation cost, even harder for segmentation. This paper studies an efficient single-source alternative for transferable attacks on semantic segmentation. We formulate transferable attack composition as a chained computation over differentiable attack components, allowing the expensive source-model gradient computation to be shared. To reduce the update instability introduced by chained composition, we further use an integrated-gradient-style path-averaged direction as an empirical stabilization heuristic. Experiments on Pascal VOC and Cityscapes evaluate the resulting transferability efficiency trade-off across CNN- and transformer-based segmentation models. IGME achieves competitive transferability compared with single-source baselines and favorable runtime compared with model-ensemble attacks, while requiring access to only one source model.
Chinese Translation
语义分割模型对可转移的对抗扰动存在脆弱性，但在密集预测模型上评估转移攻击可能会计算成本高昂。现有的集成攻击通常依赖多个替代模型，增加了计算成本，对分割任务来说更是如此。本文研究了一种高效的单源替代方案，用于对语义分割进行可转移攻击。我们将可转移攻击的组合形式化为对可微分攻击组件的链式计算，从而允许昂贵的源模型梯度计算得以共享。为了减少链式组合引入的更新不稳定性，我们进一步使用集成梯度风格的路径平均方向作为经验稳定化启发式方法。在 Pascal VOC 和 Cityscapes 上的实验评估了在基于 CNN 和变换器的分割模型中所获得的转移效率与计算时间的权衡。IGME 在与单源基线相比时实现了具有竞争力的转移性，并且与模型集成攻击相比具有良好的运行时间，同时仅需访问一个源模型。
cs.CV / 9 / 2607.27537
ProgFormer: Hierarchical Voxel Diffusion Transformer for Longitudinal Brain MRI Prediction
ProgFormer：用于纵向脑MRI预测的层次化体素扩散变换器
Abstract
Predicting future structural MRI of a brain is challenging because longitudinal changes are often subtle and confined to specific anatomical regions, while most subject-specific brain structure remains stable over time. An effective model should therefore preserve global brain structural consistency while remaining sensitive to fine-grained disease progression. Existing latent-space-based methods improve computational efficiency, but suffer from information loss during their compression-reconstruction procedure. In contrast, direct voxel-space methods avoid latent reconstruction but commonly use a unified prediction pathway to model brain structure and progression-related changes. Subtle local changes may therefore be overshadowed by the dominant stable brain structure. To address these challenges, we propose ProgFormer, a hierarchical voxel-space Diffusion Transformer for longitudinal brain MRI prediction. ProgFormer uses a coarse pathway to perform the primary volumetric prediction from 3D patch tokens. This pathway models overall brain structure and longitudinal context. The fine pathway then uses the coarse representations as spatio-temporal grounding for voxel-level refinement within individual patches. The two pathways jointly estimate a velocity field directly in voxel space through conditional flow matching, enabling end-to-end prediction without a separately learned image autoencoder. The predicted future scan is then generated from Gaussian noise by integrating the estimated velocity field over a sequence of Euler steps. Extensive experimental results on three widely used benchmarks, ADNI, AIBL, and OASIS, under both pairwise and trajectory settings demonstrate favourable performance compared against several state-of-the-art methods.
Chinese Translation
预测大脑的未来结构性MRI具有挑战性，因为纵向变化通常是微妙的，并且局限于特定的解剖区域，而大多数特定个体的大脑结构在时间上保持稳定。因此，一个有效的模型应该在保持全球大脑结构一致性的同时，对细微的疾病进展保持敏感。现有的基于潜在空间的方法提高了计算效率，但在压缩-重建过程中存在信息损失。相比之下，直接的体素空间方法避免了潜在重建，但通常使用统一的预测路径来建模大脑结构和与进展相关的变化。因此，微妙的局部变化可能会被主导的稳定大脑结构所掩盖。为了解决这些挑战，我们提出了ProgFormer，一种用于纵向脑MRI预测的层次化体素空间扩散变换器。ProgFormer使用粗略路径从3D补丁标记执行主要体积预测。该路径建模整体大脑结构和纵向上下文。然后，细致路径使用粗略表示作为空间-时间基础，对各个补丁内的体素级别进行细化。这两条路径通过条件流匹配直接在体素空间中共同估计速度场，从而实现端到端的预测，而无需单独学习的图像自编码器。然后，通过在一系列欧拉步骤中整合估计的速度场，从高斯噪声中生成预测的未来扫描。在ADNI、AIBL和OASIS三个广泛使用的基准上进行的大量实验结果，在成对和轨迹设置下，与几种最先进的方法相比，展示了良好的性能。
cs.CV / 10 / 2607.27558
Drawing-Recode: Annotation Grounding for Parametric CAD Code Generation from Raster 2D CAD Drawings
Drawing-Recode：从光栅格式2D计算机辅助设计图纸生成参数化CAD代码的注释基础
Abstract
Recovering Parametric CAD sequences from raster-format 2D Computer-Aided Design (CAD) drawings accumulated prior to digital transformation is important for part reproduction and manufacturing process automation. However, existing studies either process only vector drawings or are limited to specific domains, and fail to explicitly connect dimensional annotations to geometric information, limiting their use of dimensional information for 3D Parametric CAD sequences recovery. We propose Drawing-Recode, a framework that generates Parametric CAD sequences as CAD code from raster 2D CAD drawings. Drawing-Recode extracts geometric features via an image encoder and recognizes annotations through a separate text recognition module, then explicitly grounds annotations to geometric information using cross-attention and our proposed Annotation Grounding Loss (AGL). The resulting features are fed into a Large Language Model (LLM) to generate CAD code in the Structured Parametric CAD Code (SPCC) format. Experiments show that Drawing-Recode outperforms existing baselines and remains robust on scanned drawings resembling industrial conditions. We expect Drawing-Recode contributes to digitizing raster 2D CAD drawings in industrial settings and to part reproduction and manufacturing automation.
Chinese Translation
从数字化转型之前积累的光栅格式2D计算机辅助设计（CAD）图纸中恢复参数化CAD序列对于零件再生产和制造过程自动化至关重要。然而，现有研究要么仅处理矢量图纸，要么局限于特定领域，未能明确将尺寸注释与几何信息连接起来，从而限制了它们对3D参数化CAD序列恢复的尺寸信息的利用。我们提出了Drawing-Recode，一个从光栅2D CAD图纸生成参数化CAD序列作为CAD代码的框架。Drawing-Recode通过图像编码器提取几何特征，并通过单独的文本识别模块识别注释，然后使用交叉注意力和我们提出的注释基础损失（Annotation Grounding Loss, AGL）将注释明确地与几何信息关联。生成的特征被输入到大型语言模型（Large Language Model, LLM）中，以生成结构化参数化CAD代码（Structured Parametric CAD Code, SPCC）格式的CAD代码。实验表明，Drawing-Recode在性能上优于现有基准，并在类似工业条件的扫描图纸上保持稳健。我们期望Drawing-Recode能够促进工业环境中光栅2D CAD图纸的数字化以及零件再生产和制造自动化。
cs.CV / 11 / 2607.27564
Inference-Time Agentic Decision Rules Beat Longer Evolving Search for Multi-Image Medical Reasoning
推理时代理决策规则优于更长的演化搜索用于多图像医学推理
Abstract
Multi-image medical VQA is not merely a prompt-length problem; it is a fundamental challenge of agentic decision-making. Medical vision-language agents must aggregate evidence across ordered images, remain robust to answer-order perturbations, and avoid overfitting to noisy search-time feedback. We study MedFrameQA through a controlled comparison of five inference-time agentic strategies, optimized using the same high-budget ShinkaEvolve configuration and evaluated on a reproducible internal frozen split (1,331 evolution, 665 holdout, 855 final test). Across five independent repeated runs, the strongest method emerges as the simplest robust aggregator: the \textbf{order-vote} policy achieves $57.89 \pm 0.65\%$ final-test accuracy, significantly outperforming the fixed baseline ($52.73 \pm 0.42\%$) and the more complex, albeit brittle, order-rerank variant ($55.79 \pm 0.43\%$). Paired bootstrap analysis confirms these significant gains. Extending the evolutionary search budget from 50 to 100 generations yields no generalization benefit: while holdout performance marginally increases, final-test accuracy drops from $57.89\%$ to $56.02\%$. Our findings suggest that for multi-image medical reasoning, defining the correct agentic decision rule is substantially more impactful than expanding the optimization search budget.
Chinese Translation
多图像医学视觉问答（VQA）不仅仅是一个提示长度的问题；它是一个代理决策的基本挑战。医学视觉-语言代理必须在有序图像中聚合证据，保持对答案顺序扰动的鲁棒性，并避免对噪声搜索反馈的过拟合。我们通过对五种推理时代理策略的控制比较来研究MedFrameQA，这些策略使用相同的高预算ShinkaEvolve配置进行优化，并在可重复的内部冻结划分（1,331次演化，665次保留，855次最终测试）上进行评估。在五次独立的重复实验中，最强的方法是最简单的鲁棒聚合器： extbf{order-vote}策略实现了$57.89 imes 0.65 ext{%}$的最终测试准确率，显著优于固定基线（$52.73 imes 0.42 ext{%}$）和更复杂但脆弱的order-rerank变体（$55.79 imes 0.43 ext{%}$）。配对自助分析确认了这些显著的提升。将演化搜索预算从50代扩展到100代并未带来泛化收益：尽管保留性能略有提高，但最终测试准确率从$57.89 ext{%}$降至$56.02 ext{%}$。我们的发现表明，对于多图像医学推理，定义正确的代理决策规则的影响远大于扩展优化搜索预算。
cs.CV / 12 / 2607.27566
Objective-Aligned Direct Answer SFT for Robust Multi-Frame Medical VQA
面向目标的直接答案监督微调在稳健多帧医学视觉问答中的应用
Abstract
Multi-frame medical VQA appears to reward increasingly complex adaptation: controller-style inference, localization-aware reranking, static hard-negative mixing, and staged continuation all appear plausible from first principles. We test a simpler competing hypothesis on MedFrameQA: methods that remain tightly aligned with the benchmark's final answer objective should be the strongest \emph{robust} adaptation family once evaluation is controlled across fixed splits, matched budgets, repeated seeds, and calibration. We compare controller-based methods, scaffold evolution, static mixed supervision, continuation-heavy variants, and direct answer-only supervised fine-tuning (SFT). The strongest robust family is direct decoder-only answer SFT on MedGemma-1.5-4B. Empirically, this family yields substantial improvements in held-out report accuracy over frozen baselines while remaining remarkably stable across repeated seeds and matched controls, ensuring our claims reflect true family-level robustness rather than an isolated hyperparameter peak. Furthermore, post-hoc calibration effectively repairs confidence estimation without compromising accuracy, and the core approach transfers consistently to secondary backbones like Qwen2.5-VL-3B. The main result is therefore not that a complex auxiliary mechanism wins, but that objective-aligned direct answer SFT is the strongest robust adaptation family we found for MedFrameQA. By establishing this strong, minimalist baseline, we hope to redirect community focus toward fundamentally robust optimization rather than architectural complexity.
Chinese Translation
多帧医学视觉问答（VQA）似乎奖励越来越复杂的适应：控制器风格的推理、定位感知的重新排序、静态硬负样本混合和分阶段的延续在理论上都显得合理。我们在 MedFrameQA 上测试了一个更简单的竞争假设：与基准的最终答案目标紧密对齐的方法应该是最强的 extit{稳健}适应家族，前提是在固定划分、匹配预算、重复种子和校准的情况下进行评估。我们比较了基于控制器的方法、支架演化、静态混合监督、重续变体和仅直接答案的监督微调（SFT）。最强的稳健家族是针对 MedGemma-1.5-4B 的直接解码器答案 SFT。从经验上看，这个家族在保持的报告准确性上相较于冻结基线有显著提升，同时在重复种子和匹配控制下保持了相当的稳定性，确保我们的主张反映了真实的家族级稳健性，而非孤立的超参数峰值。此外，事后校准有效修复了置信度估计而不影响准确性，核心方法在 Qwen2.5-VL-3B 等次级骨干网络上也能持续迁移。因此，主要结果并不是复杂的辅助机制获胜，而是面向目标的直接答案 SFT 是我们为 MedFrameQA 找到的最强稳健适应家族。通过建立这个强大而简约的基线，我们希望将社区的关注重新引导到根本的稳健优化上，而不是架构复杂性。
cs.CV / 13 / 2607.27585
ZMIS-SAM: Segment Anything Model Enhanced with Wavelet Transform for Zooplankton Microscopy Image Instance Segmentation
ZMIS-SAM：基于小波变换增强的随意分割模型用于浮游动物显微图像实例分割
Abstract
As primary consumers in the marine food chain, zooplankton play a crucial role in maintaining marine ecological balance. However, the Segment Anything Model (SAM) exhibits limited performance in microscopic image instance segmentation due to its lack of zooplankton-specific domain knowledge. To address these challenges, we propose a novel instance segmentation model based on SAM and wavelet transform (ZMIS-SAM), effectively tackling issues such as inaccurate classification, discontinuous segmentation of slender appendages, and incomplete boundary segmentation. Our framework incorporates three core innovations: ZM-ViT enhances SAM's capability to model zooplankton morphology and image intensity distributions through two lightweight adapters, the Neighboring Feature Aggregation Module (NFAM) improves continuous segmentation of semi-transparent slender appendages by integrating general-purpose and domain-specific features, and the Wavelet-based Multi-scale Multi-directional Feature Enhancement (WM2FE) module effectively recovers high-frequency details to refine boundary segmentation completeness. Extensive experiments demonstrate that ZMIS-SAM achieves state-of-the-art instance segmentation performance on the zooplankton dataset and exhibits strong generalization capability across multiple public cross-domain datasets.
Chinese Translation
作为海洋食物链中的主要消费者，浮游动物在维持海洋生态平衡中发挥着至关重要的作用。然而，随意分割模型（Segment Anything Model, SAM）在显微图像实例分割中的表现有限，主要是由于其缺乏针对浮游动物的特定领域知识。为了解决这些挑战，我们提出了一种基于SAM和小波变换的新型实例分割模型（ZMIS-SAM），有效应对分类不准确、细长附肢的分割不连续以及边界分割不完整等问题。我们的框架包含三项核心创新：ZM-ViT通过两个轻量级适配器增强了SAM在建模浮游动物形态和图像强度分布方面的能力，邻近特征聚合模块（Neighboring Feature Aggregation Module, NFAM）通过整合通用特征和领域特定特征改善了半透明细长附肢的连续分割，而基于小波的多尺度多方向特征增强模块（Wavelet-based Multi-scale Multi-directional Feature Enhancement, WM2FE）有效恢复了高频细节，以提高边界分割的完整性。大量实验表明，ZMIS-SAM在浮游动物数据集上实现了最先进的实例分割性能，并在多个公共跨领域数据集上展现了强大的泛化能力。
cs.CV / 14 / 2607.27592
MeshFM: 2D Features Are All You Need for 3D Shape Understanding
MeshFM：3D 形状理解只需 2D 特征
Abstract
We present MeshFM, an efficient feedforward framework for extracting rich features from 3D inputs. Our method distills 2D features from visual foundation models into 3D. We train a feedforward network to directly predict 3D features without requiring optimization during inference. The approach utilizes a two-stage training strategy. First, we optimize a feature field in 3D using only 2D feature supervision. Second, we train a network to regress this feature field. The entire procedure requires no 3D annotation, instead relying on the powerful information in 2D foundation models. We demonstrate that our learned features can be immediately applied to downstream tasks, including part segmentation, dense correspondence, and mesh deformation. Extensive experiments show that MeshFM, trained solely with 2D supervision, performs on par with methods trained explicitly with 3D supervision, even without task-specific fine-tuning. Moreover, our model is trained to be robust to extreme rotations of the input objects. Project page: https://threedle.github.io/MeshFM/
Chinese Translation
我们提出了 MeshFM，这是一种高效的前馈框架，用于从 3D 输入中提取丰富的特征。我们的方法将视觉基础模型中的 2D 特征提炼为 3D 特征。我们训练一个前馈网络，直接预测 3D 特征，而无需在推理过程中进行优化。该方法采用两阶段训练策略。首先，我们仅使用 2D 特征监督来优化 3D 中的特征场。其次，我们训练一个网络来回归该特征场。整个过程不需要 3D 注释，而是依赖于 2D 基础模型中的强大信息。我们展示了我们学习到的特征可以立即应用于下游任务，包括部件分割、密集对应和网格变形。大量实验表明，MeshFM 在仅使用 2D 监督训练的情况下，其性能与明确使用 3D 监督训练的方法相当，甚至在没有特定任务微调的情况下。此外，我们的模型经过训练，能够对输入对象的极端旋转具有鲁棒性。项目页面：https://threedle.github.io/MeshFM/
cs.CV / 15 / 2607.27616
MPIE-Bench: Benchmarking Anatomically Plausible Multi-Person Interaction Editing
MPIE-Bench：解剖学上合理的多人物互动编辑基准测试
Abstract
Text-to-image and personalized editing models now synthesize high-fidelity single-subject images with ease. Yet placing multiple named people into shared contact actions such as embrace, carry, or grapple still exposes major failures: fused limbs, invented extremities, and interpenetrating bodies. Existing evaluations largely overlook these anatomical and geometric issues, and VLM-as-a-judge checklists often saturate on Interaction while the errors remain obvious to humans. We introduce MPIE-Bench, a 2,500-sample benchmark of video-mined editing triplets spanning 405 scenes, 14 interaction categories, and four contact densities (C0-C3). We also propose MPIE-Eval, whose two new axes score contact-time geometry from a frozen public multi-person mesh reconstruction. Anatomy asks whether every human-like mass is explained by a complete set of reconstructed bodies, and Interaction asks whether the penetration and surface distance between those bodies match the contact the instruction asked for. Across ten editors, mesh Anatomy tops out at 0.65 and mesh Interaction at 0.72 on two different models, so no single editor is strong on both, while VLM checklists rate the same images above 0.95. A five-rater study confirms that both axes track human judgement more closely than a zero-shot VLM judge, and the rankings hold under ablation of every weight and threshold.
Chinese Translation
文本到图像和个性化编辑模型现在能够轻松合成高保真度的单一主体图像。然而，将多个指定人物放置于共享的接触动作中，如拥抱、携带或摔跤，仍然暴露出重大缺陷：肢体融合、虚构的肢体和相互穿透的身体。现有的评估在很大程度上忽视了这些解剖和几何问题，而VLM作为评判者的检查表在互动方面往往饱和，而错误对人类而言依然显而易见。我们引入了MPIE-Bench，这是一个包含2500个样本的基准测试，涵盖了405个场景、14个互动类别和四种接触密度（C0-C3）的基于视频挖掘的编辑三元组。我们还提出了MPIE-Eval，其两个新维度从一个冻结的公共多人物网格重建中评分接触时间几何。解剖学问题询问每个类人质量是否由一整套重建的身体所解释，而互动问题询问这些身体之间的穿透和表面距离是否符合指令所要求的接触。在十个编辑器中，网格解剖学在两个不同模型上的最高得分为0.65，网格互动的最高得分为0.72，因此没有单一编辑器在两者上都表现出色，而VLM检查表则将相同图像的评分提高到0.95以上。一项五位评分者的研究确认，这两个维度比零-shot VLM评判者更贴近人类判断，并且在每个权重和阈值的消融下排名保持不变。
cs.CV / 16 / 2607.27620
MedXplore: Towards Reliable and Unbiased Generalized Category Discovery in Medical Imaging
MedXplore：迈向可靠且无偏的医学影像广义类别发现
Abstract
Deep learning has shown strong potential in medical image analysis, but most existing methods rely on large-scale annotations and a closed-world assumption that rarely holds in clinical practice. Although Generalized Category Discovery (GCD) has advanced rapidly on natural images, it remains underexplored in medical imaging. To address this issue, we propose MedXplore, a unified framework for reliable and unbiased medical GCD, optimizing from both perceptual and decision levels. Specifically, at the perceptual level, taking a frequency domain perspective, Frequency-SNR Adaptive Attention and Consistency (FAAC) performs learnable full-spectrum filtering and global-local energy contrast activation to not only highlight local abnormal signals relative to the global context, but also provide reliable semantic anchors for patch consistency learning. At the decision level, Adaptive Cosine-Angular Margin (ACAM) adjusts angular margins using semantic difficulty and feature confidence to balance intra-class compactness and inter-class separability. Together, the two modules improve lesion-sensitive representation learning and mitigate old-class bias. Experiments on multiple benchmarks show an average \textbf{8.5\%} gain in \textit{All} accuracy over the strongest competing methods. On Kvasir, MedXplore reduces false-old errors from 14.50\% to 0.80\%, demonstrating strong robustness under severe old-new ambiguity.
Chinese Translation
深度学习在医学图像分析中展现了强大的潜力，但大多数现有方法依赖于大规模标注和在临床实践中很少成立的封闭世界假设。尽管广义类别发现（Generalized Category Discovery, GCD）在自然图像上迅速发展，但在医学影像中仍然未得到充分探索。为了解决这一问题，我们提出了MedXplore，一个统一的框架，用于可靠且无偏的医学GCD，从感知和决策两个层面进行优化。具体而言，在感知层面，从频域的角度出发，频率-信噪比自适应注意力与一致性（Frequency-SNR Adaptive Attention and Consistency, FAAC）执行可学习的全谱滤波和全局-局部能量对比激活，不仅突出相对于全局上下文的局部异常信号，还为补丁一致性学习提供可靠的语义锚点。在决策层面，自适应余弦-角度边际（Adaptive Cosine-Angular Margin, ACAM）利用语义难度和特征置信度调整角度边际，以平衡类内紧凑性和类间可分性。这两个模块共同改善了对病灶敏感的表征学习，并减轻了旧类偏差。在多个基准测试上的实验表明，相较于最强竞争方法，整体准确率平均提高了8.5%。在Kvasir数据集上，MedXplore将假阳性错误从14.50%降低至0.80%，在严重的旧-新模糊情况下展现出强大的鲁棒性。
cs.CV / 17 / 2607.27628
BlindPSNR: A No-Reference Fidelity Predictor for Low-Light Image Enhancement
BlindPSNR：一种无参考的低光照图像增强保真度预测器
Abstract
Low-light image enhancement (LLIE) methods involve tunable parameters that are typically fixed, often leading to performance degradation when applied across scenes. Manually selecting the best configuration, however, can be time-consuming and not always practical. Peak signal-to-noise ratio (PSNR) is the natural fidelity criterion for automating parameter selection, yet it requires a ground-truth reference that is typically unavailable. To our knowledge, no learning-based method addresses no-reference PSNR prediction for low-light image enhancement; the natural surrogate, no-reference image quality assessment (NR-IQA), targets perceptual quality rather than signal fidelity, and all seven baselines we test achieve 0% top-1 selection accuracy on our benchmark. With paired training data, the ground-truth PSNR is analytically computable, providing exact supervision without a separate teacher network. Building on this, we propose BlindPSNR, a lightweight no-reference network that fuses the enhanced image with the degraded low-light input via windowed cross-attention and estimates PSNR through heteroscedastic regression. While a scalar-regression baseline achieves top-1 accuracy of 54.4%, BlindPSNR raises this to 89.5% with regret dropping from 1.62 dB to 0.026 dB, and generalizes to unseen datasets (SRCC = 0.61-0.67).
Chinese Translation
低光照图像增强（LLIE）方法涉及可调参数，这些参数通常是固定的，导致在不同场景中应用时性能下降。然而，手动选择最佳配置可能耗时且并不总是实用。峰值信噪比（PSNR）是自动化参数选择的自然保真度标准，但它需要通常不可用的真实参考。我们所知，没有学习基础的方法针对低光照图像增强的无参考PSNR预测；自然的替代方法，无参考图像质量评估（NR-IQA），关注感知质量而非信号保真度，我们测试的七个基线在我们的基准测试中均实现了0%的顶级选择准确率。通过配对训练数据，真实的PSNR可以通过分析计算，提供精确的监督而无需单独的教师网络。在此基础上，我们提出了BlindPSNR，一种轻量级的无参考网络，通过窗口交叉注意力将增强图像与退化的低光输入融合，并通过异方差回归估计PSNR。尽管标量回归基线的顶级准确率为54.4%，BlindPSNR将其提升至89.5%，并且后悔值从1.62 dB降至0.026 dB，且在未见过的数据集上具有良好的泛化能力（SRCC = 0.61-0.67）。
cs.CV / 18 / 2607.27634
4DHumanDiff: Direct Text-to-4DGS Generation for Consistent 360-Degree Dynamic Humans
4DHumanDiff：直接从文本生成一致的360度动态人类的4DGS
Abstract
Generating high-quality 360-degree dynamic human assets from text prompts is challenging. Existing methods usually synthesize monocular or multi-view videos first and then fit a 4D representation, which is expensive and often causes incomplete geometry or view-inconsistent renderings. We present 4DHumanDiff, a diffusion framework that directly generates dynamic humans represented by 4D Gaussian Splatting (4DGS) from text prompts. By modeling the structured 4D representation space end-to-end, 4DHumanDiff avoids video pre-generation and per-scene reconstruction, making it better suited for view-consistent and temporally coherent asset generation. The model uses a 3D U-Net backbone with temporal attention for motion-aware generation. We further construct a large-scale text-to-4DGS dataset with 60,000 high-quality pairs, and introduce 2D regularization and training-free 4D interpolation to improve rendering quality and motion smoothness. Experiments show that 4DHumanDiff generates consistent 360-degree dynamic humans within one minute, achieves better temporal and multi-view consistency, and reduces inference time by more than 10x.
Chinese Translation
从文本提示生成高质量的360度动态人类资产是一项具有挑战性的任务。现有方法通常先合成单目或多视角视频，然后再拟合4D表示，这种方法成本高昂，且常常导致几何体不完整或视图不一致的渲染。我们提出了4DHumanDiff，这是一种扩散框架，能够直接从文本提示生成由4D高斯点云（4D Gaussian Splatting, 4DGS）表示的动态人类。通过端到端建模结构化的4D表示空间，4DHumanDiff避免了视频的预生成和每场景的重建，使其更适合于生成视图一致且时间上连贯的资产。该模型使用带有时间注意力机制的3D U-Net骨干网络进行运动感知生成。我们进一步构建了一个大规模的文本到4DGS数据集，包含60,000对高质量样本，并引入了2D正则化和无训练的4D插值，以提高渲染质量和运动平滑度。实验表明，4DHumanDiff能够在一分钟内生成一致的360度动态人类，获得更好的时间和多视角一致性，并将推理时间减少超过10倍。
cs.CV / 19 / 2607.27637
MMOOC: A Comprehensive Benchmark for Out-of-Context Evaluation in Multimodal Large Language Models
MMOOC：多模态大型语言模型的上下文外评估综合基准
Abstract
Multimodal Large Language Models (MLLMs) have achieved strong performance on a wide range of vision-language tasks, but often fail under imperfect or shifted contexts. A reliable MLLM should refuse truly out-of-context (OOC) questions with subject-level context shifts while still answering shifted in-context (Shifted IC) questions with non-subject context shifts. Existing benchmarks mainly target OOC or visually unanswerable questions, but overlook answerable Shifted IC cases and cover limited OOC shifts. To fill this gap, we present MMOOC, a large-scale benchmark for evaluating refusal and robust answering abilities of MLLMs. MMOOC contains over 41K image-question pairs, including answerable Shifted IC cases and unanswerable OOC cases, spanning three question formats, eight shift types and six visual scenarios, with data quality ensured through MLLM-based filtering and human verification. We evaluate model responses using Accuracy and Refusal Rate, and further introduce an LLM-as-a-Judge metric to assess the correctness of model reasoning. Experiments on diverse MLLMs show that current models still struggle to balance answer-ability and refusal under shifted contexts. We further analyze key failure patterns and show that post-training can improve robustness. MMOOC will be made publicly available.
Chinese Translation
多模态大型语言模型（MLLMs）在广泛的视觉-语言任务中取得了强劲的表现，但在不完美或偏移的上下文中往往表现不佳。一个可靠的MLLM应该能够拒绝真正的上下文外（OOC）问题，这些问题涉及主题级上下文的偏移，同时仍能回答非主题上下文偏移的偏移上下文内（Shifted IC）问题。现有基准主要针对OOC或视觉上无法回答的问题，但忽视了可回答的Shifted IC案例，并且涵盖的OOC偏移有限。为填补这一空白，我们提出了MMOOC，一个用于评估MLLMs拒绝和稳健回答能力的大规模基准。MMOOC包含超过41K的图像-问题对，包括可回答的Shifted IC案例和无法回答的OOC案例，涵盖三种问题格式、八种偏移类型和六种视觉场景，数据质量通过基于MLLM的过滤和人工验证得以保证。我们使用准确率和拒绝率评估模型响应，并进一步引入LLM-as-a-Judge指标来评估模型推理的正确性。在多样化的MLLMs上进行的实验表明，当前模型在偏移上下文下仍然难以平衡可回答性和拒绝性。我们进一步分析了关键的失败模式，并显示后训练可以提高稳健性。MMOOC将公开发布。
cs.CV / 20 / 2607.27659
Learning Color Grading, No Photo Sharing: Federated Aesthetic Preference Learning for Personalized Image Enhancement
学习色彩分级，无需照片共享：用于个性化图像增强的联邦美学偏好学习
Abstract
Personalized image enhancement should reflect individual aesthetic taste, yet learning such preferences commonly depends on private photos and ratings that are unsuitable for centralized collection. The task must infer preference from sparse, heterogeneous feedback and translate it into natural-looking color transformations on resource-constrained user devices. We introduce FedPAIE, a federated personalized aesthetic image enhancement framework for user-adaptive color grading without centralizing raw photos or ratings. FedPAIE trains a lightweight dual-cue aesthetic scorer, calibrates it into a personalized scorer on a small local support set, and freezes it to guide regularized adaptation of a lightweight CLUT enhancer from unpaired local photographs. Fidelity constraints and an excess-gap penalty regularize scorer-guided adaptation to limit proxy-score over-optimization while preserving content and natural appearance. Training remains lightweight throughout the pipeline: scorer learning updates at most 0.787M parameters, enhancer adaptation updates 0.265M, and inference retains only a 0.293M-parameter personalized enhancer. Experiments on MIT-Adobe FiveK and Flickr-AES demonstrate effective open-world personalization and a favorable balance between user preference and image fidelity. FedPAIE thus connects decentralized preference learning with efficient personalized image transformation without requiring paired user retouches.
Chinese Translation
个性化图像增强应反映个体的美学品味，但学习此类偏好通常依赖于不适合集中收集的私人照片和评分。该任务必须从稀疏的异构反馈中推断偏好，并将其转化为在资源受限的用户设备上自然的色彩变换。我们提出了 FedPAIE，这是一个用于用户自适应色彩分级的联邦个性化美学图像增强框架，无需集中原始照片或评分。FedPAIE 训练一个轻量级的双线索美学评分器，将其校准为在小型本地支持集上的个性化评分器，并冻结该评分器以指导轻量级 CLUT 增强器从未配对的本地照片中进行正则化适应。保真度约束和过度间隙惩罚对评分器指导的适应进行正则化，以限制代理评分的过度优化，同时保持内容和自然外观。整个流程中的训练保持轻量化：评分器学习最多更新 0.787M 参数，增强器适应更新 0.265M，推理仅保留 0.293M 参数的个性化增强器。在 MIT-Adobe FiveK 和 Flickr-AES 上的实验表明有效的开放世界个性化，以及用户偏好与图像保真度之间的良好平衡。因此，FedPAIE 将去中心化的偏好学习与高效的个性化图像转换连接起来，而无需配对用户的修饰。
cs.CV / 21 / 2607.27667
Witness Evidence Portfolios: Single-Prefill Risk Detection for Closed Multimodal Answers
证人证据组合：闭合多模态答案的单预填风险检测
Abstract
Reliable deployment of multimodal large language models (MLLMs) requires deciding whether a confident visual answer should be trusted, reviewed, or routed to a stronger system. Confidence scores capture candidate margins, but not where the estimated signed visual readouts associated with those margins come from or how they are distributed. We study inference-time risk detection for closed visual answers using the same white-box prefill path that produces the answer. Witness Evidence Portfolios (WEP) first estimates, layer by layer, which visual contributions support or contradict the predicted candidate. It summarizes these contributions through two interpretable route families: question-related evidence provenance and signed evidence concentration. Nested grouped validation chooses the more reliable family and a sparse top-k route portfolio, which is fused with candidate confidence. WEP needs no image perturbation, decoding change, backward pass, or external verifier. Across three MLLMs and four binary-answer benchmarks, WEP improves mean error AP by 0.134. All 12 model--dataset gains are positive, and image-cluster bootstrap intervals are strictly positive on 10 pairs. WEP targets white-box closed-answer systems and uses a labeled calibration slice.
Chinese Translation
可靠部署多模态大型语言模型（MLLMs）需要决定是否应信任、审查或将一个自信的视觉答案路由到更强的系统。置信度分数捕捉候选边际，但并未说明与这些边际相关的估计签名视觉读数的来源或分布情况。我们研究了使用相同的白盒预填路径进行闭合视觉答案的推理时风险检测，该路径生成答案。证人证据组合（Witness Evidence Portfolios, WEP）首先逐层估计哪些视觉贡献支持或反驳预测的候选答案。它通过两种可解释的路径族总结这些贡献：与问题相关的证据来源和签名证据集中度。嵌套分组验证选择更可靠的路径族和稀疏的前k条路径组合，这与候选置信度融合。WEP无需图像扰动、解码变化、反向传播或外部验证器。在三个MLLM和四个二元答案基准测试中，WEP将平均错误AP提高了0.134。所有12个模型-数据集的增益均为正值，10对图像集群自助法区间严格为正。WEP针对白盒闭合答案系统，并使用标记的校准切片。
cs.CV / 22 / 2607.27670
JigShape: Evaluating Visual-Geometric Reasoning in VLMs through Jigsaw Puzzles
JigShape：通过拼图评估视觉-几何推理在视觉语言模型中的表现
Abstract
Jigsaw puzzle solving requires jointly reasoning about visual content and geometric constraints, yet existing benchmarks use rectangular cuts that create ambiguous ground truth in texture-repeated regions. We introduce \textit{\ours{}}, a benchmark with tab-and-blank interlocking pieces where geometric constraints provide strong local compatibility requirements that, combined with visual content, yield unambiguous ground truth. Across 95K instances at four grid densities (4$\times$4 to 16$\times$16), we find that \textbf{zero-shot VLMs largely lack geometric reasoning}: only one of five frontier models (GPT-5.5) exceeds random baseline on 4$\times$4 puzzles, while all others perform at chance level. While supervised fine-tuning achieves $>$97\% on 4$\times$4, \textbf{all models collapse on larger grids}: GPT-5.5 drops from 70\% to near-random on 8$\times$8, and even fine-tuned models fall below 5\% on 12$\times$12. This ``scaling cliff'' suggests current architectures cannot maintain consistent constraint satisfaction as the number of pieces increases. \ours{} establishes scalable geometric reasoning as an open challenge for vision-language models.
Chinese Translation
拼图解决需要对视觉内容和几何约束进行联合推理，但现有基准使用矩形切割，这在纹理重复区域中产生了模糊的真实标签。我们引入了 extit{ extbf{JigShape}}，这是一个具有插接和空白互锁拼块的基准，其中几何约束提供了强大的局部兼容性要求，这与视觉内容结合，产生明确的真实标签。在四种网格密度（4$ imes$4到16$ imes$16）的95K实例中，我们发现 extbf{零-shot视觉语言模型在几何推理方面大多缺乏能力}：只有五个前沿模型中的一个（GPT-5.5）在4$ imes$4拼图上超过随机基线，而其他所有模型的表现均处于随机水平。尽管监督微调在4$ imes$4上达到了$>$97\%，但 extbf{所有模型在更大网格上崩溃}：GPT-5.5在8$ imes$8上从70\\%降至接近随机，而即使是微调模型在12$ imes$12上也低于5\\%。这种“规模悬崖”表明当前架构无法在拼块数量增加时保持一致的约束满足。 extbf{JigShape}确立了可扩展的几何推理作为视觉语言模型面临的一个开放挑战。
cs.CV / 23 / 2607.27699
RefineSVG: Visual Feedback-Driven Reinforcement Learning for Image-to-SVG Generation
RefineSVG：基于视觉反馈驱动的图像到SVG生成的强化学习
Abstract
We propose RefineSVG, a single-step closed-loop visual feedback framework that enables multimodal large language models (MLLMs) to perform high-fidelity image-to-SVG generation through self-correction. Existing MLLM-based approaches rely on single-pass open-loop inference, where the model receives visual input only once and must generate thousands of SVG code tokens without intermediate verification. This paradigm inevitably leads to geometric drift, error accumulation, and visual hallucination on complex images. RefineSVG overcomes this limitation by invoking an external rendering engine after an initial SVG generation pass to compare the rendered output against the target image. The comparison yields a multi-dimensional visual residual map (Diff-Map) that is fed back to the model as a ReAct-style correction signal, driving a targeted correction step. To support this render-observe-correct interaction, we further introduce an SVG-oriented semantic vocabulary that compresses token sequences by over 52%. A progressive training pipeline spanning supervised fine-tuning, rejection-sampling cold-start data construction, and end-to-end agentic reinforcement learning aligns the model with closed-loop visual correction. Extensive experiments show that RefineSVG consistently outperforms existing baselines in reconstruction fidelity, structural accuracy, and code efficiency.Code is available at https://github.com/liuxiaobo66/RefineSVG.
Chinese Translation
我们提出了RefineSVG，这是一个单步闭环视觉反馈框架，使多模态大型语言模型（MLLMs）能够通过自我修正实现高保真度的图像到SVG生成。现有的基于MLLM的方法依赖于单次开放循环推理，其中模型仅接收一次视觉输入，并必须在没有中间验证的情况下生成数千个SVG代码标记。这种范式不可避免地导致几何漂移、错误累积以及在复杂图像上的视觉幻觉。RefineSVG通过在初始SVG生成后调用外部渲染引擎，将渲染输出与目标图像进行比较，从而克服了这一限制。比较结果生成一个多维视觉残差图（Diff-Map），该图作为ReAct风格的修正信号反馈给模型，驱动有针对性的修正步骤。为了支持这种渲染-观察-修正的交互，我们进一步引入了一种面向SVG的语义词汇，能够将标记序列压缩超过52%。一个涵盖监督微调、拒绝采样冷启动数据构建和端到端自主强化学习的渐进训练流程，使模型与闭环视觉修正对齐。大量实验表明，RefineSVG在重建保真度、结构准确性和代码效率方面始终优于现有基线。代码可在 https://github.com/liuxiaobo66/RefineSVG 获取。
cs.CV / 24 / 2607.27700
Calibrate Before Reason: Robust Visual Token Reduction against Semantic Drift in VLMs
推理前的校准：针对视觉语言模型中语义漂移的鲁棒视觉标记减少
Abstract
Large Vision-Language Models (VLMs) suffer from prohibitive inference overhead due to long sequences of visual tokens. However, existing visual token reduction methods mainly improve efficiency by pruning or compressing redundant tokens without examining whether the resulting representation remains semantically consistent with the original representation. Mapping the original N-token visual sequence to K tokens may discard, dilute, or misassign critical visual cues, triggering severe semantic drift that deviates the VLM's understanding. In this paper, we first introduce the principle of 'Calibrate Before Reason' to visual token reduction and propose CaRe, a training-free robust framework that calibrates compact visual representations before reasoning to preserve semantic fidelity in VLMs. CaRe consists of two mutually complementary modules: 1) Perturbation-Robust Calibration Anchoring, which identifies calibration anchors with stable model-side influence under multi-directional perturbations; 2) Confidence-Gated Token Calibration, which extracts reliable calibration signals from unselected tokens and injects them into anchors. Extensive evaluations across diverse VLM architectures and benchmarks verify that CaRe outperforms state-of-the-art token reduction baselines. While pruning 94.4% of visual tokens, our method retains 96.4% of the original full-token performance, delivering up to 2.30 times faster end-to-end inference speed relative to unpruned vanilla models.
Chinese Translation
大型视觉语言模型（VLMs）由于视觉标记的长序列而面临巨大的推理开销。然而，现有的视觉标记减少方法主要通过修剪或压缩冗余标记来提高效率，而未考虑所得到的表示是否与原始表示在语义上保持一致。将原始的N标记视觉序列映射到K标记可能会丢弃、稀释或错误分配关键的视觉线索，从而引发严重的语义漂移，偏离VLM的理解。在本文中，我们首先将“推理前的校准”原则引入视觉标记减少，并提出CaRe，一个无训练的鲁棒框架，该框架在推理之前校准紧凑的视觉表示，以保持VLM中的语义保真度。CaRe由两个相互补充的模块组成：1）扰动鲁棒校准锚点，识别在多方向扰动下对模型影响稳定的校准锚点；2）置信度门控标记校准，从未选择的标记中提取可靠的校准信号并将其注入锚点。对多种VLM架构和基准的广泛评估验证了CaRe在标记减少基准上优于最先进的方法。在修剪94.4%的视觉标记的同时，我们的方法保留了96.4%的原始全标记性能，相较于未修剪的原始模型，提供了高达2.30倍的端到端推理速度提升。
cs.CV / 25 / 2607.27729
PrintAnything: Learning an Intermediate Representation for 3D printing G-code Generation
PrintAnything：学习用于3D打印G-code生成的中间表示
Abstract
Point clouds are one of the most fundamental and widely used 3D representations, serving as the most basic geometric representation of 3D shapes. Nevertheless, most existing 3D printing pipelines require a watertight mesh as input, preventing the direct use of point clouds for fabrication. A common workaround is to reconstruct meshes from point clouds; however, the resulting meshes often contain geometric artifacts, such as incorrect faces or topological inconsistencies, that are difficult to repair and may lead to printing failures. To overcome these limitations, we propose PrintAnything, a novel framework that learns to produce executable 3D printing G-code directly from 3D point clouds without requiring mesh reconstruction. To enable point clouds to serve as direct input for slice-wise toolpath generation, we introduce a slice-wise point projection strategy that transforms unstructured 3D point clouds into slice-aligned 2D representations consistent with layer-by-layer nature of fused deposition modeling in 3D printing. To eliminate mesh dependency and provide a unified representation that bridges point clouds and G-code, we propose Geometric plan (G-plan) map, a compact 2D representation composed of occupancy, region, and flow maps that encode the geometric and extrusion properties required for toolpath synthesis in 3D printing. As a result, our proposed method accurately generates printable G-code directly from point clouds, enabling a practical and fully mesh-free pipeline for 3D printing. The code is publicly available at \href{https://github.com/Sangminhong/PrintAnything}{https://github.com/Sangminhong/PrintAnything}.
Chinese Translation
点云是最基本且广泛使用的3D表示之一，作为3D形状的最基本几何表示。然而，大多数现有的3D打印流程要求输入一个密闭网格，这使得点云无法直接用于制造。一种常见的解决方法是从点云重建网格；然而，生成的网格通常包含几何伪影，例如错误的面或拓扑不一致，这些问题难以修复，可能导致打印失败。为了解决这些限制，我们提出了PrintAnything，这是一个新颖的框架，能够直接从3D点云生成可执行的3D打印G-code，而无需进行网格重建。为了使点云能够作为切片工具路径生成的直接输入，我们引入了一种切片点投影策略，将非结构化的3D点云转换为与3D打印中熔融沉积建模的逐层特性一致的切片对齐的2D表示。为了消除对网格的依赖，并提供一个统一的表示来桥接点云和G-code，我们提出了几何平面（Geometric plan, G-plan）地图，这是一种紧凑的2D表示，由占用图、区域图和流动图组成，编码了3D打印中工具路径合成所需的几何和挤出属性。因此，我们提出的方法能够准确地直接从点云生成可打印的G-code，实现了一个实用的完全无网格的3D打印流程。代码已公开发布在 https://github.com/Sangminhong/PrintAnything。
cs.CV / 26 / 2607.27749
Articulated Object Reconstruction from Rest-State Observation
基于静态观察的关节物体重建
Abstract
Building interactive digital twins requires recovering both 3D geometry and the kinematic structures that govern how objects articulate. Yet existing methods for articulated object reconstruction require explicitly observable motion from multiple articulation states. We introduce a rest-state formulation that reconstructs articulated objects from a single closed configuration, an inherently ill-posed setting where geometry, semantics, and motion priors compensate for the absence of motion cues. Our framework adopts an explicit mesh as an intermediate representation for cross-model verification and fusion, reconciling noisy outputs from vision-language and segmentation models into spatially consistent part structures. To estimate joint parameters without observed motion, we use a video diffusion model to synthesize articulation hypotheses and validate them through geometric consistency. Our approach achieves accurate part decomposition and physically plausible articulation, performing competitively with motion-observing reconstruction-based, generation-based, and modular pretrained-model baselines.
Chinese Translation
构建交互式数字双胞胎需要恢复三维几何形状和支配物体关节运动的运动结构。然而，现有的关节物体重建方法需要从多个关节状态中显式观察到运动。我们提出了一种静态状态的公式，通过单一的闭合配置重建关节物体，这是一种固有的病态设置，其中几何、语义和运动先验弥补了运动线索的缺失。我们的框架采用显式网格作为交叉模型验证和融合的中间表示，将来自视觉-语言和分割模型的噪声输出调和为空间一致的部件结构。为了在没有观察到运动的情况下估计关节参数，我们使用视频扩散模型合成关节假设，并通过几何一致性对其进行验证。我们的方法实现了准确的部件分解和物理上合理的关节运动，在与基于运动观察的重建、生成和模块化预训练模型基线的比较中表现出竞争力。
cs.CV / 27 / 2607.27755
EgoGVAE: Ego-body Mesh Reconstruction via Guided Variational Autoencoder
EgoGVAE：通过引导变分自编码器进行自我身体网格重建
Abstract
We address the problem of recovering the full-body mesh from only the head pose. This task has become essential for various applications based on head-mounted devices or smart glasses. The challenge of this task lies in estimating the pose information of unobserved body parts based solely on a single joint (i.e., head) trajectory. Several studies have begun to adopt head-conditioned generative models, however, such previous methods are costly and time-consuming due to the diffusion-based iterative process. As an alternative, we propose a simple yet novel method that leverages the latent space of the guidance network, which is designed as a variational autoencoder taking full-body poses as inputs. By enforcing latent distributions of this guidance network and our head-to-motion network to be similar, latent features sampled from the 'guided' distribution, i.e., distribution learned in our head-to-motion network, can be reliably decoded for natural representations of full-body poses even only with the head pose. One important advantage of the proposed method is that one-step sampling scheme achieves remarkably fast inference (more than 50 times faster) compared to diffusion-based approaches. Experimental results on benchmark datasets show that the proposed method efficiently improves the performance of ego-body mesh reconstruction.
Chinese Translation
我们解决了仅通过头部姿态恢复全身网格的问题。该任务对于基于头戴设备或智能眼镜的各种应用变得至关重要。该任务的挑战在于仅基于单个关节（即头部）轨迹估计未观察到的身体部位的姿态信息。一些研究已开始采用基于头部条件的生成模型，然而，由于基于扩散的迭代过程，这些先前的方法成本高且耗时。作为替代方案，我们提出了一种简单而新颖的方法，利用引导网络的潜在空间，该网络被设计为一个以全身姿态为输入的变分自编码器。通过强制该引导网络的潜在分布与我们的头部到运动网络的潜在分布相似，可以可靠地解码从“引导”分布（即在我们的头部到运动网络中学习的分布）中采样的潜在特征，以实现自然的全身姿态表示，即使仅有头部姿态。该方法的一个重要优势是，一步采样方案相比于基于扩散的方法实现了显著快速的推理（速度超过50倍）。在基准数据集上的实验结果表明，所提方法有效提高了自我身体网格重建的性能。
cs.CV / 28 / 2607.27763
DS@GT ARC at ImageCLEFmedical 2026: Architectural Diversity for Concept Detection and Foundation-Model Scaling for Caption Prediction in Medical Image Analysis
DS@GT ARC 在 ImageCLEFmedical 2026：用于概念检测的架构多样性与基础模型扩展在医学图像分析中的标题生成
Abstract
We describe the DS@GT submissions to the ImageCLEFmedical Caption 2026 challenge, which continues a long-running benchmark on the ROCOv2 dataset with two tracks: Concept Detection (Task 1), assigning UMLS Concept Unique Identifiers (CUIs) to radiology images, and Caption Prediction (Task 2), generating natural-language captions. For Task 1, our primary submission was a three-way late-fusion ensemble of ConvNeXt-V2, BiomedCLIP ViT-B/16, and DenseNet-169 with a regularized ''Honest Threshold Tuning'' procedure designed to avoid validation overfitting on rare concepts; this submission ranked first on the official submission with a primary $F_1$ of $0.5790$ and a secondary $F_1$ of $0.9657$. In parallel, we submitted a training-free KNN retrieval pipeline over frozen BiomedCLIP embeddings, which reached a primary $F_1$ of $0.5780$ and a secondary $F_1$ of $0.9599$-essentially matching the fine-tuned ensemble on the primary track at a fraction of the cost. For Task 2, our submissions included a fine-tuned Gemma-3 27B model (overall $0.3571$, ranking third in the official submission), a fully fine-tuned BLIP pipeline with custom Vizwins merging ($0.3564$), and a zero-shot MedGemma-4B run with a PubMed-style prompt ($0.3186$), spanning a wide range of model scales and training costs. Code: https://github.com/dsgt-arc/imageclef-caption-2026.
Chinese Translation
我们描述了 DS@GT 对 ImageCLEFmedical Caption 2026 挑战的提交，该挑战在 ROCOv2 数据集上继续进行长期基准测试，包含两个任务：概念检测（任务 1），为放射学图像分配 UMLS 概念唯一标识符（CUI），以及标题生成（任务 2），生成自然语言标题。对于任务 1，我们的主要提交是 ConvNeXt-V2、BiomedCLIP ViT-B/16 和 DenseNet-169 的三路后融合集成，采用了旨在避免在稀有概念上验证过拟合的正则化“诚实阈值调整”程序；该提交在官方提交中排名第一，主要 $F_1$ 为 $0.5790$，次要 $F_1$ 为 $0.9657$。与此同时，我们提交了一个基于冻结的 BiomedCLIP 嵌入的无训练 KNN 检索管道，达到了主要 $F_1$ 为 $0.5780$ 和次要 $F_1$ 为 $0.9599$，基本上与主要任务上的微调集成匹配，但成本仅为其一小部分。对于任务 2，我们的提交包括一个微调的 Gemma-3 27B 模型（总体 $0.3571$，在官方提交中排名第三）、一个完全微调的 BLIP 管道与自定义 Vizwins 合并（$0.3564$），以及一个零样本的 MedGemma-4B 运行，使用 PubMed 风格的提示（$0.3186$），涵盖了广泛的模型规模和训练成本。代码： https://github.com/dsgt-arc/imageclef-caption-2026.
cs.CV / 29 / 2607.27764
Private Face Recognition Training Dataset Publication via Identity-Decoupled and Geometry-Preserving Face Distillation
通过身份解耦和几何保持的人脸识别训练数据集的私密发布
Abstract
Publishing private face recognition~(FR) training datasets is privacy-sensitive because faces expose identity information. Private FR training dataset publication mitigates this risk by releasing protected proxies as substitutes for private training faces. However, training FR models with such data introduces an identity paradox: \emph{the identity cues that make released faces useful for recognition supervision are also the cues that make them linkable to real individuals.} A protected face should be decoupled from the original identity, yet still behave as a reliable identity sample for training. Removing these cues too aggressively may destroy the class structure needed for recognition learning, whereas preserving them too faithfully may increase source-identity linkability. We argue that this paradox stems from conflating source-aligned identity semantics with recognition-useful proxy identity geometry. The former should be suppressed to reduce linkage to private individuals, while the latter should be preserved for FR learning. Based on this insight, we propose \textbf{Private Face Distillation}, an identity-decoupling and geometry-preserving framework. It uses Orthogonal Geometry Preservation to construct decoupled proxy identities from private identity representations while maintaining hyperspherical geometry, and Relational Topology Alignment to preserve identity relations for recognition learning. Experiments across multiple domain-shifted FR scenarios show that Private Face Distillation achieves stronger utility than the evaluated publication baselines. On IJB-C surveillance, it improves $\mathrm{TAR}@\mathrm{FAR}{=}1\text{e-}{3}$ by 3.94\% over the baseline while reducing source-identity linkability. These results suggest that private FR training dataset publication should decouple source-identity correspondence while preserving proxy identity geometry.
Chinese Translation
发布私密的人脸识别（FR）训练数据集涉及隐私敏感性，因为人脸暴露了身份信息。私密FR训练数据集的发布通过发布受保护的代理作为私密训练人脸的替代品来减轻这一风险。然而，使用此类数据训练FR模型引入了身份悖论： extit{使发布的人脸对识别监督有用的身份线索，同时也是使其与真实个体可链接的线索。} 受保护的人脸应与原始身份解耦，但仍应作为可靠的身份样本进行训练。过于激进地去除这些线索可能会破坏识别学习所需的类别结构，而过于忠实地保留它们可能会增加源身份的可链接性。我们认为，这一悖论源于将源对齐的身份语义与对识别有用的代理身份几何混淆。前者应被抑制以减少与私密个体的链接，而后者应被保留以用于FR学习。基于这一见解，我们提出了 extbf{私密人脸蒸馏}，一个身份解耦和几何保持的框架。它使用正交几何保持从私密身份表示中构建解耦的代理身份，同时保持超球几何，并使用关系拓扑对齐来保留识别学习所需的身份关系。在多个领域转移的FR场景中的实验表明，私密人脸蒸馏的效用优于评估的发布基线。在IJB-C监控数据集上，它在减少源身份可链接性的同时，将$ ext{TAR}@ ext{FAR}{=}1 ext{e-}{3}$提高了3.94 ext{%}。这些结果表明，私密FR训练数据集的发布应解耦源身份对应关系，同时保持代理身份几何。
cs.CV / 30 / 2607.27779
CXR-Retrieve: Compositional Text-to-Image Retrieval in Chest Radiography
CXR-Retrieve：胸部放射影像中的组合文本到图像检索
Abstract
Large chest radiography archives are difficult to search because most studies are paired only with free-text reports rather than structured clinical annotations. Vision-language models offer a natural interface for text-to-image retrieval, but current biomedical models are primarily optimized for report-to-image matching rather than for satisfying short clinical search queries. This creates an objective mismatch: a model may retrieve images related to words in the query while failing to satisfy the full clinical constraint, especially for conjunctions and negations such as ``atelectasis and no pneumonia.'' We introduce CXR-Retrieve, a structured benchmark for compositional chest X-ray text-to-image retrieval. The benchmark contains 5,159 test images from the official test-split of MIMIC-CXR-JPG and 145 textual queries spanning single and conjunction findings, both positive and negative. Relevance is defined by whether a retrieved image satisfies all asserted pathology constraints, rather than by whether it matches a paired report. We further propose a label-aware contrastive fine-tuning objective for clinical retrieval. Our method attracts image-text pairs with compatible asserted pathology constraints, including shared confirmed absences, while explicitly repelling contradictory pairs. Starting from the in-domain CXR-CLIP checkpoint, our method improves Precision@5 over CXR-CLIP by 8.5 percentage points on two-pathology conjunctions and by 22.0 percentage points on negation queries. These results show that reliable chest X-ray retrieval requires training objectives that model not only which findings are mentioned, but also how they are clinically asserted.
Chinese Translation
大型胸部放射影像档案难以检索，因为大多数研究仅与自由文本报告配对，而不是结构化的临床注释。视觉-语言模型为文本到图像检索提供了自然的接口，但当前的生物医学模型主要优化于报告与图像的匹配，而不是满足简短的临床搜索查询。这造成了一个客观的不匹配：模型可能检索与查询中单词相关的图像，但未能满足完整的临床约束，尤其是对于诸如“肺不张和无肺炎”这样的连接词和否定词。我们引入了CXR-Retrieve，这是一个用于组合胸部X光文本到图像检索的结构化基准。该基准包含来自MIMIC-CXR-JPG官方测试集的5,159张测试图像和145个文本查询，涵盖单一和连接发现，包括正面和负面。相关性由检索到的图像是否满足所有声明的病理约束来定义，而不是由其是否匹配配对报告来定义。我们进一步提出了一种标签感知的对比微调目标，用于临床检索。我们的方法吸引具有兼容声明病理约束的图像-文本对，包括共享的确认缺失，同时明确排斥矛盾对。从领域内的CXR-CLIP检查点开始，我们的方法在两个病理连接上提高了Precision@5 8.5个百分点，在否定查询上提高了22.0个百分点。这些结果表明，可靠的胸部X光检索需要训练目标，不仅要建模提到的发现，还要建模它们是如何在临床上被声明的。
cs.CV / 31 / 2607.27800
FDDWAN: A Frequency-Decoupled Diffusion Network for Watermarking Attack
FDDWAN：一种频率解耦的扩散网络用于水印攻击
Abstract
Existing invisible watermark removal methods often struggle to accurately capture the watermark-bearing features, leading to an unfavorable trade-off between watermark suppression and perceptual fidelity. In this paper, we propose the Frequency-Decoupled Diffusion Watermark Attack Network (FDDWAN), a coarse-to-fine framework that performs watermark removal through wavelet-domain decomposition and residual diffusion refinement. In the initial stage, the Wavelet-based Frequency-domain Preliminary Attack Module (WFPAM) decomposes the watermarked image into low- and high-frequency subbands and applies frequency-specific attack strategies tailored to their respective contributions to watermark robustness and perceptual quality. In the next stage, the Frequency-domain Residual Diffusion Attack Module (FRDAM) separately models the residual distributions between the preliminarily attacked outputs and the corresponding watermark-free references during training. Rather than reconstructing the entire image, FRDAM selectively refines frequency-domain residuals, directing the diffusion process toward the remaining watermark related discrepancies while minimizing modifications to image content. Extensive experiments on CelebA and ImageNet across four representative watermarking schemes demonstrate that FDDWAN achieves a more favorable trade-off between watermark removal effectiveness and visual fidelity than conventional and learning-based attack methods.
Chinese Translation
现有的隐形水印去除方法往往难以准确捕捉水印承载特征，从而导致水印抑制与感知保真度之间的不理想权衡。本文提出了一种频率解耦扩散水印攻击网络（FDDWAN），该框架采用粗到细的方式，通过小波域分解和残差扩散精炼进行水印去除。在初始阶段，基于小波的频域初步攻击模块（WFPAM）将水印图像分解为低频和高频子带，并应用针对其各自对水印鲁棒性和感知质量贡献的频率特定攻击策略。在下一阶段，频域残差扩散攻击模块（FRDAM）在训练过程中分别建模初步攻击输出与相应无水印参考之间的残差分布。FRDAM并不是重建整个图像，而是选择性地精炼频域残差，将扩散过程引导至剩余的与水印相关的差异，同时最小化对图像内容的修改。在CelebA和ImageNet上进行的广泛实验表明，FDDWAN在水印去除效果和视觉保真度之间实现了比传统和基于学习的攻击方法更为理想的权衡。
cs.CV / 32 / 2607.27806
LoMeVQA: A Comprehensive Benchmark for Longitudinal Medical VQA
LoMeVQA：一个全面的纵向医学视觉问答基准
Abstract
In clinical practice, patients often undergo multiple imaging examinations over successive visits, yielding longitudinal data. Modeling such temporal information is crucial for reliable assessment of disease progression and treatment response. However, despite the rapid advancement of multimodal large language models (MLLMs), longitudinal medical visual reasoning remains largely underexplored. To fill this gap, we propose LoMeVQA, a comprehensive benchmark consisting of 206K longitudinal visual question answering (VQA) pairs for temporal medical image analysis. LoMeVQA covers five tasks: progress classification, progress description, progress report generation, differential region grounding, and differential region description. To construct the dataset, we develop an automated pipeline that (1) organizes patient records chronologically, (2) extracts clinically meaningful entities via a medical knowledge graph, and (3) models their temporal evolution to guide large language models in generating high-quality longitudinal VQA pairs. Extensive evaluations demonstrate that both general-purpose and medical-domain MLLMs perform poorly on LoMeVQA, revealing substantial limitations in temporal reasoning. To address these limitations, we introduce MedLong-8B, which achieves state-of-the-art performance across all tasks. Beyond benchmarking, we conduct detailed analyses that uncover key failure modes and shed light on how to improve longitudinal medical visual reasoning. Our data is available at: https://github.com/pepperbubble/LoMeVQA
Chinese Translation
在临床实践中，患者通常在连续的就诊中接受多次影像检查，从而产生纵向数据。对这种时间信息的建模对于可靠评估疾病进展和治疗反应至关重要。然而，尽管多模态大型语言模型（MLLMs）迅速发展，纵向医学视觉推理仍然在很大程度上未被探索。为填补这一空白，我们提出了LoMeVQA，这是一个包含206K对纵向视觉问答（VQA）对的全面基准，旨在进行时间医学图像分析。LoMeVQA涵盖五个任务：进展分类、进展描述、进展报告生成、差异区域定位和差异区域描述。为了构建数据集，我们开发了一个自动化流程，该流程（1）按时间顺序整理患者记录，（2）通过医学知识图谱提取临床相关实体，以及（3）建模其时间演变，以指导大型语言模型生成高质量的纵向VQA对。广泛的评估表明，无论是通用型还是医学领域的MLLMs在LoMeVQA上的表现均较差，揭示了其在时间推理方面的重大局限性。为了解决这些局限性，我们引入了MedLong-8B，该模型在所有任务上均实现了最先进的性能。除了基准测试外，我们还进行了详细分析，揭示了关键的失败模式，并阐明了如何改善纵向医学视觉推理。我们的数据可在以下链接获取：https://github.com/pepperbubble/LoMeVQA
cs.CV / 33 / 2607.27811
SPFM-Net: Semantic-Prior-Guided Frequency-Constrained Mamba for Invisible Watermark Attack
SPFM-Net：基于语义先验引导的频率约束隐形水印攻击框架
Abstract
Existing watermark attacks typically rely on predefined signal-processing operations or locally constrained restoration networks, making it difficult to capture the long-range dependencies of globally distributed watermark signals and resulting in an unfavorable trade-off between removal effectiveness and visual fidelity. In this paper, we propose SPFM-Net, a semantic-prior-guided and frequency-constrained Mamba framework for invisible watermark attack. SPFM-Net first employs high-ratio masking to disrupt the spatial coherence of invisible watermark signals, and then utilizes a partially fine-tuned pretrained Masked Autoencoder to reconstruct semantically consistent image from sparse observations while suppressing watermark-related information. A Multi-scale Residual Frequency Feature Interaction module subsequently aggregates watermark-related residual features across multiple receptive fields, while adaptively suppressing responses from watermark-irrelevant regions. To further capture the long-range dependencies of globally distributed watermark signals, a lightweight Mamba-based Global State-space Feature Modeling (GSFM) unit is introduced to separate watermark-related features from natural image content and suppress the remaining watermark traces. In addition, SPFM-Net is optimized using a multi-level objective that jointly imposes spatial-, frequency-, and edge-domain constraints, enabling effective watermark suppression while preserving perceptual quality. Extensive experiments on representative spatial-domain, transform-domain, orthogonal moment-based, and deep learning-based watermarking schemes demonstrate that SPFM-Net achieves a favorable trade-off between watermark attack effectiveness and perceptual fidelity.
Chinese Translation
现有的水印攻击通常依赖于预定义的信号处理操作或局部约束的恢复网络，这使得捕捉全球分布的水印信号的长程依赖关系变得困难，并导致去除效果与视觉保真度之间的不理想权衡。本文提出了SPFM-Net，一种基于语义先验引导和频率约束的隐形水印攻击框架。SPFM-Net首先采用高比率掩蔽技术来破坏隐形水印信号的空间一致性，然后利用部分微调的预训练Masked Autoencoder从稀疏观测中重建语义一致的图像，同时抑制与水印相关的信息。接着，一个多尺度残差频率特征交互模块聚合了多个感受野中的水印相关残差特征，同时自适应地抑制来自与水印无关区域的响应。为了进一步捕捉全球分布的水印信号的长程依赖关系，引入了一种轻量级的基于Mamba的全局状态空间特征建模（GSFM）单元，以分离水印相关特征与自然图像内容，并抑制剩余的水印痕迹。此外，SPFM-Net采用多层次目标进行优化，联合施加空间、频率和边缘域约束，实现有效的水印抑制，同时保持感知质量。在代表性的空间域、变换域、正交矩阵基和深度学习水印方案上的大量实验表明，SPFM-Net在水印攻击效果与感知保真度之间实现了良好的权衡。
cs.CV / 34 / 2607.27823
Hallucinations Leave a Grounding Signature:Verifier-Guided Decoding for Selective Object Correction
幻觉留下基础签名：基于验证者引导的选择性对象修正解码
Abstract
Large vision-language models (LVLMs) often hallucinate objects that are absent from an image. Despite recent progress, existing mitigation methods still lack reliable object-level grounding diagnostics and therefore tend to apply coarse-grained interventions, which can impair visual understanding, shorten responses, and reduce coverage of genuinely grounded objects. The key challenge is thus to detect, during generation, whether each emerging object mention is supported by reliable visual evidence, so that hallucination can be mitigated selectively. Yet output confidence reflects next-token plausibility rather than visual support, allowing language priors to make absent objects appear certain. We show that the missing diagnostic evidence is encoded in an Intrinsic Grounding Signature (IGS), a distributed signed attention pattern that remains informative for such confident hallucinations. Based on IGS, we propose Verifier-Guided Decoding (VGD), a decoding framework in which a lightweight verifier examines each emerging object mention, rolls back the KV cache when the mention is identified as high risk, suppresses the object and its synonyms, and regenerates the affected continuation. Because VGD intervenes only on object mentions identified as high risk, it reduces object hallucination while preserving the model's original visual understanding and grounded object coverage. Experiments on CHAIR and AMBER-G show that VGD achieves state-of-the-art object hallucination reduction: at @rec90, it cuts AMBER-G CHAIR by 43.6\% while retaining 99.6\% of grounded-object coverage, and reduces CHAIR-MSCOCO CHAIR$_i$/CHAIR$_s$ by 37.0\%/30.4\% without shortening captions.
Chinese Translation
大型视觉语言模型（LVLMs）常常会幻觉出图像中不存在的物体。尽管最近取得了一些进展，现有的缓解方法仍然缺乏可靠的对象级基础诊断，因此往往采用粗粒度的干预措施，这可能会损害视觉理解，缩短响应时间，并减少真正基础对象的覆盖。关键挑战在于在生成过程中检测每个出现的对象提及是否有可靠的视觉证据支持，以便能够选择性地减轻幻觉。然而，输出置信度反映的是下一个标记的合理性而非视觉支持，这使得语言先验能够让缺失的对象看起来是确定的。我们展示了缺失的诊断证据被编码在内在基础签名（Intrinsic Grounding Signature, IGS）中，这是一种分布式的有符号注意力模式，对于这种自信的幻觉仍然具有信息价值。基于IGS，我们提出了验证者引导解码（Verifier-Guided Decoding, VGD），这是一个解码框架，其中一个轻量级的验证者检查每个出现的对象提及，当该提及被识别为高风险时，回滚KV缓存，抑制该对象及其同义词，并重新生成受影响的续写。由于VGD仅对被识别为高风险的对象提及进行干预，因此在保留模型原始视觉理解和基础对象覆盖的同时，减少了对象幻觉。在CHAIR和AMBER-G上的实验表明，VGD实现了最先进的对象幻觉减少：在@rec90时，它将AMBER-G CHAIR减少了43.6%，同时保留了99.6%的基础对象覆盖，并在不缩短标题的情况下将CHAIR-MSCOCO CHAIR$_i$/CHAIR$_s$分别减少了37.0%/30.4%。
cs.CV / 35 / 2607.27830
Thinking Once Is Enough: Intermediate-Layer Evidence Routing for High-Resolution VQA
思考一次就足够：高分辨率视觉问答的中间层证据路由
Abstract
High-resolution visual question answering (HR-VQA) is often treated as a problem of insufficient evidence acquisition, where failing multimodal large language models must inspect images again through cropping, re-encoding, or multi-round search. We show that this view is incomplete: in many cases, fine-grained evidence has already survived visual encoding and become identifiable and influential within an intermediate-layer routing window, but is later diluted before answer generation. We propose Thinking-Once, a \textbf{training-free, single-visual-pass} evidence-routing method that reconstructs question-conditioned attention at this window, preserves core entity tokens and compact background context, and routes this evidence to later layers without extra visual encoding. Across five base models, Thinking-Once consistently improves or matches the corresponding base setting, increasing the average scores on V$^*$Bench, HRBench-4K, and HRBench-8K by \textit{+3.1}, \textit{+3.0}, and \textit{+2.7} points while reducing the average peak memory by about 4,GB. On Qwen2.5-VL-7B, it improves the three benchmarks by \textit{+9.9}, \textit{+4.6}, and \textit{+5.5} points, raising the cross-benchmark mean from 72.5 to 79.1. With the ZwZ-8B base model, Thinking-Once reaches a mean score of 82.7. Against 11 open-source HR-VQA baselines, it obtains the best or tied-best score on all three benchmark averages and the best overall mean; for example, compared with DeepScan, it reduces V$^*$Bench inference time by \textbf{97.2\%} while improving the cross-benchmark mean from 77.8 to 79.1. These results show that HR-VQA can be improved by routing already encoded evidence rather than repeatedly acquiring new visual inputs. Code is available in the appendix.
Chinese Translation
高分辨率视觉问答（HR-VQA）通常被视为一个证据获取不足的问题，其中失败的多模态大语言模型必须通过裁剪、重新编码或多轮搜索再次检查图像。我们表明这种观点是不完整的：在许多情况下，细粒度证据已经在视觉编码中存活下来，并在中间层路由窗口内变得可识别和有影响力，但在答案生成之前却被稀释。我们提出了Thinking-Once，这是一种 extbf{无训练、单次视觉传递}的证据路由方法，它在该窗口重建基于问题的注意力，保留核心实体标记和紧凑的背景上下文，并将这些证据路由到后续层，而无需额外的视觉编码。在五个基础模型中，Thinking-Once始终提高或匹配相应的基础设置，使得在V$^*$Bench、HRBench-4K和HRBench-8K上的平均分数分别增加了 extit{+3.1}、 extit{+3.0}和 extit{+2.7}点，同时将平均峰值内存减少约4GB。在Qwen2.5-VL-7B上，它使三个基准的分数分别提高了 extit{+9.9}、 extit{+4.6}和 extit{+5.5}点，将跨基准均值从72.5提高到79.1。使用ZwZ-8B基础模型，Thinking-Once达到了82.7的平均分数。在11个开源HR-VQA基线中，它在所有三个基准平均值上获得了最佳或并列最佳分数，并且在整体均值上表现最佳；例如，与DeepScan相比，它将V$^*$Bench的推理时间减少了 extbf{97.2\%}，同时将跨基准均值从77.8提高到79.1。这些结果表明，通过路由已经编码的证据而不是反复获取新的视觉输入，可以改善HR-VQA。代码可在附录中获得。
cs.CV / 36 / 2607.27835
SAFViT: Spatial Attention Fusion Gating for Vision Transformer-Based Nucleus Segmentation and Classification
SAFViT：基于视觉变换器的细胞核分割与分类的空间注意力融合门控
Abstract
Accurate cell segmentation and classification are foundational to digital pathology, enabling quantitative tissue analysis for diagnosis and treatment planning. Encoder-decoder architectures that fuse multi-scale features through skip connections have become the dominant paradigm for this task, yet standard direct skip connections treat every spatial location equally, which leads to redundant and potentially conflicting information reaching the decoder. To overcome this problem, various gating mechanisms have been introduced, but most of them operate solely on filtering encoder information, neglecting the benefit of global contextual information from the decoder. This study proposes replacing conventional skip connections in a CellViT-based model with a novel Spatial Attention Fusion (SAF) Gating module. Each SAF gate concatenates the encoder skip and upsampled decoder features, compresses them through two pointwise convolutions with an intermediate ReLU, and applies a channel-wise softmax to produce a per-pixel "heatmap of trust" that sums to unity at every spatial location, allowing the network to learn where each source is most trustworthy. The resulting fused features improve the model's ability to detect the minority "Dead" class, which in turn enhances the multi-class panoptic quality (mPQ) on the PanNuke dataset. SAF Gating is compared against six gating alternatives including no gating, attention gates, squeeze-and-excitation, CBAM, cross-attention, and attentional feature fusion on PanNuke and MoNuSeg datasets. SAF Gating achieves the highest mPQ (0.471), a gain driven primarily by a 14.5-point improvement in Dead-class F1 score compared to ungated CellViT baseline.
Chinese Translation
准确的细胞分割与分类是数字病理学的基础，能够为诊断和治疗规划提供定量组织分析。通过跳跃连接融合多尺度特征的编码器-解码器架构已成为这一任务的主流范式，然而标准的直接跳跃连接对每个空间位置的处理是相同的，这导致冗余和潜在冲突的信息到达解码器。为了解决这个问题，已经引入了多种门控机制，但大多数仅在过滤编码器信息上操作，忽略了来自解码器的全局上下文信息的好处。本研究提出在基于CellViT的模型中用一种新颖的空间注意力融合（SAF）门控模块替代传统的跳跃连接。每个SAF门将编码器的跳跃特征和上采样的解码器特征进行拼接，通过两个点卷积和一个中间的ReLU进行压缩，并应用通道级的softmax生成每个像素的“信任热图”，使得每个空间位置的和为1，从而使网络能够学习每个源最可信的地方。最终融合的特征提高了模型检测少数“死亡”类别的能力，从而增强了在PanNuke数据集上的多类全景质量（mPQ）。SAF门控与六种门控替代方案进行了比较，包括无门控、注意力门、挤压与激励、CBAM、交叉注意力和注意力特征融合，均在PanNuke和MoNuSeg数据集上进行。SAF门控实现了最高的mPQ（0.471），其提升主要得益于与无门控CellViT基线相比，死亡类别F1分数提高了14.5点。
cs.CV / 37 / 2607.27842
FeatFix: Reuse What You Verify through Local Exact-Feature Correction for Faster Cached Diffusion Inference
FeatFix：通过局部精确特征修正重用已验证内容以加速缓存扩散推理
Abstract
Diffusion models are widely used to generate high-quality images and videos, but their iterative denoising process remains computationally intensive. A growing class of training-free accelerators reduces this cost by reusing cached intermediate features or forecasting future ones. To control draft drift, these methods sometimes compute an exact block feature for verification. Yet the resulting exact feature is typically used only to measure discrepancy or guide a later decision and is then discarded. We find that this previously computed feature can instead be reused for correction. Forwarding it at the verification site resets the local draft residual and reduces downstream feature error. Based on this observation, we introduce FeatFix, a local exact-feature correction method for cached diffusion inference. FeatFix operates at a fixed sparse set of layer--timestep sites. At each selected site, it replaces the complete draft block output with the exact output computed from the same incoming state, avoiding token- or channel-level partial replacement and full-timestep recomputation. Experiments across four image and video backbones show that FeatFix consistently accelerates generation, achieving a speedup of up to $6.70\times$ over Vanilla while maintaining competitive output quality.
Chinese Translation
扩散模型广泛用于生成高质量的图像和视频，但其迭代去噪过程仍然计算密集。一类日益增长的无训练加速器通过重用缓存的中间特征或预测未来特征来降低这一成本。为了控制草稿漂移，这些方法有时会计算一个精确的块特征以进行验证。然而，得到的精确特征通常仅用于测量差异或指导后续决策，随后被丢弃。我们发现，这个先前计算的特征可以被重新用于修正。在验证点转发它可以重置局部草稿残差并减少下游特征误差。基于这一观察，我们提出了FeatFix，一种用于缓存扩散推理的局部精确特征修正方法。FeatFix在固定的稀疏层-时间步点集上操作。在每个选定的点上，它用从相同输入状态计算的精确输出替换完整的草稿块输出，避免了令牌或通道级的部分替换和全时间步的重新计算。跨四个图像和视频骨干网络的实验表明，FeatFix始终加速生成，速度提升高达$6.70 imes$，同时保持竞争力的输出质量。
cs.CV / 38 / 2607.27843
VCP-DCN: Beyond Visual Concealed Property via Depth Collaborative Network for Camouflaged Object Detection
VCP-DCN：超越视觉隐蔽特性，通过深度协同网络进行伪装物体检测
Abstract
Camouflaged Object Detection (COD) aims to identify and segment camouflaged objects in complex environments, which are often concealed because their color and texture are similar to the background. Several existing COD methods introduce depth maps to boost detection performance via learning complementary RGB-D features, ignoring modality-specific characteristics of concealed objects in the depth domain. To address this issue, we propose a depth collaborative network, called VCP-DCN, to mine distinguishable multi-modality features beyond visual concealed prototype in depth domain. Specifically, VCP-DCN progressively performs multi-modality alignment, interaction, and fusion for the COD task. In the \textbf{alignment} stage, we propose a Separable Prototype Embedding (SPE) module to learn modality-consistency and modality-specific RGB/depth prototype tokens through prototype contrastive learning. Furthermore, we develop a Multi-modality Dual Attention (MDA) module to enhance the cross-modal feature representation through local response maps between modality-consistency RGB/depth prototype tokens and visual tokens on the \textbf{interaction} stage. Finally, we design a Depth Adaptive Injection (DAI) module to adaptively measure contribution of RGB/depth features with a decision-making mechanism, which calculates similarity distance between RGB/depth modality-specific prototype tokens and modality-consistency ones on the \textbf{fusion} stage. Extensive experiments demonstrate the effectiveness of our VCP-DCN on three authoritative datasets.
Chinese Translation
伪装物体检测（COD）旨在识别和分割复杂环境中伪装的物体，这些物体通常因其颜色和纹理与背景相似而被隐藏。现有的几种COD方法引入深度图，通过学习互补的RGB-D特征来提升检测性能，但忽略了深度域中隐蔽物体的特定模态特征。为了解决这一问题，我们提出了一种深度协同网络，称为VCP-DCN，旨在挖掘超越视觉隐蔽原型的可区分多模态特征。具体而言，VCP-DCN逐步执行多模态对齐、交互和融合以完成COD任务。在对齐阶段，我们提出了可分离原型嵌入（SPE）模块，通过原型对比学习来学习模态一致性和模态特定的RGB/深度原型标记。此外，我们开发了多模态双重注意力（MDA）模块，以通过模态一致的RGB/深度原型标记与视觉标记之间的局部响应图增强跨模态特征表示。在融合阶段，我们设计了深度自适应注入（DAI）模块，通过决策机制自适应地衡量RGB/深度特征的贡献，该机制计算RGB/深度模态特定原型标记与模态一致的标记之间的相似性距离。大量实验表明，我们的VCP-DCN在三个权威数据集上的有效性。
cs.CV / 39 / 2607.27856
Benchmarking Foundation and Large Language Models for Few-Shot Medical Image Segmentation
基础与大型语言模型在少样本医学图像分割中的基准测试
Abstract
Few-shot medical image segmentation (FS-MIS) aims to segment novel regions of interest (ROIs) from a few annotated support examples. Despite rapid progress, existing FS-MIS solutions span diverse paradigms but are evaluated under inconsistent settings, leaving their relative effectiveness unclear. We introduce FAME, a unified benchmark for evaluating FS-MIS solutions, covering specialists, SAM-based methods, CLIP-based methods, and MLLM-based methods. FAME contains 14,958 test samples across 7 anatomical sites, 9 imaging modalities, and 14 ROI categories, and evaluates models under zero-shot and ten-shot settings with additional assessment of target-absence recognition and generalization under covariate and semantic shifts. Our evaluation reveals several findings. First, effective few-shot segmentation depends on how models exploit support examples: direct visual adaptation generally outperforms prompt-based strategies. Second, increasing support examples improves performance only when models can effectively utilize them. Third, semantic transfer remains substantially more challenging than imaging-domain adaptation, and strong localization ability does not necessarily imply reliable target-absence recognition. We hope FAME provides a comprehensive understanding of current FS-MIS solutions and facilitates the development of more effective and reliable few-shot medical segmentation methods.
Chinese Translation
少样本医学图像分割（FS-MIS）旨在从少量标注的支持示例中分割出新的感兴趣区域（ROIs）。尽管取得了快速进展，现有的FS-MIS解决方案涵盖了多种范式，但在不一致的设置下进行评估，使得它们的相对有效性不明确。我们引入了FAME，这是一个用于评估FS-MIS解决方案的统一基准，涵盖了专家系统、基于SAM的方法、基于CLIP的方法和基于MLLM的方法。FAME包含来自7个解剖部位、9种成像模式和14个ROI类别的14,958个测试样本，并在零样本和十样本设置下评估模型，同时额外评估目标缺失识别和在协变量及语义变化下的泛化能力。我们的评估揭示了几个发现。首先，有效的少样本分割依赖于模型如何利用支持示例：直接的视觉适应通常优于基于提示的策略。其次，增加支持示例仅在模型能够有效利用它们时才会提高性能。第三，语义转移仍然比成像领域适应更具挑战性，强大的定位能力并不一定意味着可靠的目标缺失识别。我们希望FAME能够提供对当前FS-MIS解决方案的全面理解，并促进更有效和可靠的少样本医学分割方法的发展。
cs.CV / 40 / 2607.27857
EEG-EditBench: Probing Visual Information in EEG-Image Retrieval Models with Controlled Image Edits
EEG-EditBench：通过控制图像编辑探究EEG-图像检索模型中的视觉信息
Abstract
Recent EEG-to-image retrieval models have achieved strong performance in identifying viewed images from semantically diverse candidates. Yet such success does not reveal what visual information supports the match. A model may readily identify a cheetah among tools, plants, and vehicles, but can it still distinguish the viewed cheetah from the same scene with the cheetah replaced by a dog? Motivated by this question, we introduce EEG-EditBench, a diagnostic benchmark that examines this question through controlled edits of object identity, attributes, background, and object presence. Built from the 200 THINGS-EEG2 test images, EEG-EditBench contains 2,137 quality-controlled edits and evaluates eight representative EEG visual decoding models. Our results show that strong standard retrieval does not consistently transfer to edit-based evaluation, with fine-grained attribute changes presenting the greatest challenge. EEG-EditBench reveals model behavior hidden by aggregate retrieval accuracy and provides a controlled basis for studying what visual information EEG-image models preserve. The code and complete dataset are publicly available.
Chinese Translation
近期的EEG到图像检索模型在从语义多样的候选中识别观看过的图像方面取得了良好的表现。然而，这种成功并未揭示支持匹配的视觉信息。一个模型可能能够轻易地在工具、植物和车辆中识别出猎豹，但它是否仍能区分在同一场景中被狗替代的猎豹？基于这个问题，我们引入了EEG-EditBench，一个通过控制对象身份、属性、背景和对象存在的编辑来检验这一问题的诊断基准。EEG-EditBench基于200个THINGS-EEG2测试图像构建，包含2137个质量控制的编辑，并评估了八个具有代表性的EEG视觉解码模型。我们的结果表明，强大的标准检索并不总是能够一致地转移到基于编辑的评估中，细粒度的属性变化带来了最大的挑战。EEG-EditBench揭示了被聚合检索准确性掩盖的模型行为，并为研究EEG-图像模型所保留的视觉信息提供了一个受控的基础。代码和完整数据集已公开可用。
cs.CV / 41 / 2607.27865
Learning to Understand Body Language from Flight through Robust 3D Avatar Placing
通过稳健的3D虚拟形象放置学习理解肢体语言
Abstract
Perceiving human motion and intent at long range is a prerequisite for socially intelligent aerial robots, yet the data to learn it barely exists. We introduce Drones2BodyLanguage, a dataset grounding human motion in real UAV footage: avatars manifesting ten communicative intents are placed into unmodified 4K drone scenes with metrically correct position, scale and orientation, maintained over hundreds of frames of camera motion. Enabling it is a lightweight geometric world model of the local scene - semantically selected anchors lifted to 3D through streaming monocular depth - in which a placement point is predicted as an affine anchor combination with provably rigid-invariant weights, and re-rendered under an SVD-fitted ground rotation. Across twelve architectures on scene- and motion-disjoint splits, training on placed data lifts mean intent accuracy by a wide margin for real, retargeted and generated motion alike, with gains confirmed on two in-the-wild scenes.
Chinese Translation
在远距离感知人类运动和意图是社会智能无人机的前提，但用于学习这方面的数据几乎不存在。我们介绍了Drones2BodyLanguage，这是一个将人类运动与真实无人机视频相结合的数据集：在未经修改的4K无人机场景中，放置了表现十种交流意图的虚拟形象，确保其在数百帧相机运动中保持正确的位置、比例和方向。实现这一点的是一个轻量级的局部场景几何世界模型——通过流式单目深度提升的语义选择锚点被提升到3D空间，其中放置点被预测为具有可证明刚性不变权重的仿射锚点组合，并在经过SVD拟合的地面旋转下重新渲染。在场景和运动不重叠的十二种架构上，基于放置数据的训练显著提高了真实、重新定向和生成运动的平均意图准确性，并在两个真实场景中得到了验证。
cs.CV / 42 / 2607.27882
DECODE: Tackling Representation and Decision Degradation in Continual AI-Generated Image Detection
DECODE：应对持续性人工智能生成图像检测中的表征与决策退化
Abstract
As generative models continue to evolve, AI-generated image detectors must incrementally adapt to emerging generative domains while preserving knowledge acquired from previous ones. This continual learning setting is particularly challenging because forensic traces are often subtle and generator-specific, making detectors highly vulnerable to catastrophic forgetting. Existing methods primarily address this problem by stabilizing feature representations, implicitly treating forgetting as a representation-level issue. In this paper, we show that this perspective is incomplete. We demonstrate that even when feature representations remain discriminative, the decision boundary can progressively drift as the classification head is continually optimized on new domains. These two effects jointly give rise to a compound failure mode, termed Dual Degradation. To overcome this challenge, we propose DECODE, a decoupled continual detection framework that jointly mitigates representation- and decision-level forgetting. Specifically, we introduce Subspace Diversity Regularization (SDR) to preserve diverse forensic representations and Closed-Form Decision Alignment (CDA) to recalibrate the shared classification head after each adapter merge without manual hyperparameter tuning. Extensive experiments on 19 generative domains show that DECODE achieves an average accuracy of 99.36% with only 0.39% forgetting, while further generalizing to 11 unseen generators with 95.36% accuracy.
Chinese Translation
随着生成模型的不断发展，人工智能生成图像检测器必须逐步适应新兴的生成领域，同时保留从先前领域获得的知识。这种持续学习的环境尤其具有挑战性，因为法医痕迹通常微妙且特定于生成器，使得检测器对灾难性遗忘高度敏感。现有方法主要通过稳定特征表征来解决这一问题，隐含地将遗忘视为表征层面的问题。在本文中，我们表明这一视角是不完整的。我们展示了即使特征表征保持区分性，决策边界也可能随着分类头在新领域上的持续优化而逐渐漂移。这两种效应共同导致了一种复合失效模式，称为双重退化（Dual Degradation）。为了解决这一挑战，我们提出了DECODE，一个解耦的持续检测框架，能够共同减轻表征和决策层面的遗忘。具体而言，我们引入了子空间多样性正则化（Subspace Diversity Regularization, SDR）以保持多样化的法医表征，以及闭式决策对齐（Closed-Form Decision Alignment, CDA）以在每次适配器合并后重新校准共享分类头，而无需手动调整超参数。在19个生成领域的广泛实验表明，DECODE在仅有0.39%遗忘的情况下实现了99.36%的平均准确率，同时在11个未见生成器上进一步推广，准确率达到95.36%。
cs.CV / 43 / 2607.27897
Unifying Adversarially Robust Model Experts in Vision-Language Models
统一视觉-语言模型中的对抗鲁棒性模型专家
Abstract
Vision-language models (VLMs), such as CLIP, are vulnerable to adversarial attacks, posing a serious problem for real-life applications and deployment. Adversarial fine-tuning emerges as a prominent defense method; however, different fine-tuning strategies often produce specialized models with distinct robustness characteristics. Each fine-tuned model in turn thrives in some evaluation settings but falters on others, limiting their defensive capabilities. We refer to these specialized fine-tuned models as robust model experts and propose a collaborative adversarial fine-tuning framework: CARE - Collaborative Adversarial Robustness fine-tuning using Embedding alignment. CARE maintains multiple experts during training, enables knowledge exchange through embedding-space harmonization, and consolidates the learned knowledge into a single unified robust model. Experts benefit from one another while preserving their individual specializations, enabling the final model to inherit complementary robustness properties. In this paper, we demonstrate CARE on two different adversarial fine-tuning strategies with complementary robustness behaviors. Extensive experiments on classic image classification and downstream vision-language tasks display the effectiveness of our approach, with CARE being able to outperform individually learned model experts. The results suggest that collaborative learning across model experts is a promising direction for improving adversarial robustness.
Chinese Translation
视觉-语言模型（VLMs），如 CLIP，易受到对抗攻击，这对实际应用和部署构成了严重问题。对抗微调作为一种显著的防御方法应运而生；然而，不同的微调策略往往会产生具有不同鲁棒性特征的专业化模型。每个微调后的模型在某些评估环境中表现良好，但在其他环境中却表现不佳，从而限制了它们的防御能力。我们将这些专业化的微调模型称为鲁棒模型专家，并提出了一种协作对抗微调框架：CARE - 使用嵌入对齐的协作对抗鲁棒性微调。CARE 在训练过程中维护多个专家，通过嵌入空间的和谐实现知识交换，并将学习到的知识整合为一个统一的鲁棒模型。专家之间相互受益，同时保持各自的专业化，使最终模型能够继承互补的鲁棒性特征。本文展示了 CARE 在两种具有互补鲁棒性行为的对抗微调策略上的应用。在经典图像分类和下游视觉-语言任务上的大量实验显示了我们方法的有效性，CARE 能够超越单独学习的模型专家。结果表明，模型专家之间的协作学习是提高对抗鲁棒性的一个有前景的方向。
cs.CV / 44 / 2607.27898
CoRE-UIR: Prior-guided common and residual experts for efficient all-in-one remote sensing image restoration
CoRE-UIR：基于先验指导的通用与残差专家用于高效的一体化遥感图像恢复
Abstract
Remote sensing images acquired by unmanned aerial vehicles (UAVs) and satellites are often degraded by adverse weather, illumination variation, and imaging artifacts, which may co-occur and jointly induce global distribution shifts and local structural corruption. Although All-in-One image restoration offers an appealing unified alternative to task-specific pipelines, existing methods still suffer from weak or implicit degradation cues and parameter redundancy caused by full-rank multi-expert designs with overlapping restoration behaviors. We propose CoRE-UIR (Common and Residual Experts for Universal Image Restoration), a prior-guided global-local framework centered on the Common-and-Residual Expert Block (CoRE). CoRE explicitly decomposes restoration capacity into a common dense expert for degradation-invariant restoration and low-rank residual experts for degradation-specific compensation, enabling adaptive specialization without redundant expert replication. Built on this design, Degradation Prior Embedding (DPE) adapts frozen CLIP features into an explicit restoration-oriented prior, while Global Feature Modulation (GFM) aligns global feature statistics before local residual compensation. We also construct MDVD-108K (Multi-Degradation VisDrone), a large-scale UAV restoration dataset covering both single and compound degradations, together with a real-world test set. Extensive experiments on multiple datasets show that CoRE-UIR improves the overall average PSNR by 1.05 dB while running 11.83$\times$ faster and reducing peak memory by 85.3% relative to the strongest baseline, BaryIR, thereby maintaining a favorable quality-efficiency trade-off. Evaluations on downstream tasks and unseen degradation also validate the generalizability of CoRE-UIR. The code and dataset will be released at https://github.com/zzaiyan/CoRE-UIR.
Chinese Translation
无人机（UAV）和卫星获取的遥感图像常常受到恶劣天气、光照变化和成像伪影的影响，这些因素可能同时发生并共同导致全局分布偏移和局部结构损坏。尽管一体化图像恢复提供了一种吸引人的统一替代方案来取代特定任务的处理流程，但现有方法仍然受到弱或隐式退化线索以及由于全秩多专家设计导致的参数冗余的困扰，这些设计具有重叠的恢复行为。我们提出了CoRE-UIR（通用图像恢复的通用与残差专家），这是一个基于先验指导的全球-局部框架，中心是通用与残差专家模块（CoRE）。CoRE明确将恢复能力分解为一个用于退化不变恢复的通用稠密专家和多个用于退化特定补偿的低秩残差专家，从而实现自适应专业化而无需冗余的专家复制。在此设计基础上，退化先验嵌入（DPE）将冻结的CLIP特征调整为明确的恢复导向先验，而全局特征调制（GFM）在局部残差补偿之前对齐全局特征统计。我们还构建了MDVD-108K（多退化VisDrone），这是一个大规模无人机恢复数据集，涵盖单一和复合退化，并配有一个真实世界测试集。在多个数据集上的广泛实验表明，CoRE-UIR相较于最强基线BaryIR，整体平均PSNR提高了1.05 dB，同时运行速度快11.83倍，峰值内存减少了85.3%，从而保持了良好的质量效率权衡。对下游任务和未见退化的评估也验证了CoRE-UIR的普适性。代码和数据集将发布在https://github.com/zzaiyan/CoRE-UIR。
cs.CV / 45 / 2607.27902
One Patch Is Enough: Reinforcement-Optimized Visual Token Grounding for MLLM-Based Scene Text Spotting
一个补丁就足够：基于强化优化的视觉标记定位用于多模态大语言模型的场景文本检测
Abstract
Scene text spotting requires high-precision alignment between textual recognition and spatial localization. While visual-token grounding has emerged as a promising formulation for Multimodal Large Language Models (MLLMs), the previous multi-patch paradigm often introduces redundant noise and localization ambiguity, particularly for dense or small text instances. To address this, we propose Single-Patch Text Spotting (SPaTS), a vision-centric framework that routes each text instance through a single anchor visual token and then recovers geometry via full-image refinement. To accurately identify this anchor without oracle labels, we introduce Single-Patch Selective Optimization (SPaSO), a reinforcement learning framework that optimizes discrete visual-token selection using patch-level rewards. To further improve representation robustness and localization precision, we introduce Directional Embedding Alignment (DEA) to suppress unstable norm bias by decoupling feature magnitude and direction, and Patch-Enhanced Decoding (PED) to fuse the routed anchor with language semantics and cross-attend over the full-image feature map for geometry-aware boundary regression beyond coordinate-space surrogates. Extensive experiments demonstrate that SPaTS consistently and significantly outperforms both frontier closed-source MLLMs and OCR MLLMs. Code will be released soon.
Chinese Translation
场景文本检测需要在文本识别与空间定位之间实现高精度对齐。尽管视觉标记定位已成为多模态大语言模型（MLLMs）的一种有前景的表述，但以往的多补丁范式往往引入冗余噪声和定位模糊，特别是在密集或小文本实例的情况下。为了解决这个问题，我们提出了单补丁文本检测（Single-Patch Text Spotting, SPaTS），这是一个以视觉为中心的框架，通过单个锚点视觉标记引导每个文本实例，然后通过全图精细化恢复几何信息。为了在没有真实标签的情况下准确识别这个锚点，我们引入了单补丁选择性优化（Single-Patch Selective Optimization, SPaSO），这是一个强化学习框架，利用补丁级奖励优化离散视觉标记的选择。为了进一步提高表示的鲁棒性和定位的精确性，我们引入了方向嵌入对齐（Directional Embedding Alignment, DEA），通过解耦特征的大小和方向来抑制不稳定的范数偏差，并引入补丁增强解码（Patch-Enhanced Decoding, PED），将路由的锚点与语言语义融合，并在全图特征图上进行交叉关注，以实现超越坐标空间替代的几何感知边界回归。大量实验表明，SPaTS在性能上始终显著优于前沿的闭源MLLMs和OCR MLLMs。代码将很快发布。
cs.CV / 46 / 2607.27927
ARD-REFSM: Enhancing Reflection Symmetry Detection with Asymmetric Denoising and Rotation Equivariance
ARD-REFSM：通过非对称去噪和旋转等变性增强反射对称性检测
Abstract
Reflection symmetry detection remains challenging due to interference from asymmetric regions and arbitrary orientations of symmetric patterns. Asymmetric regions introduce background clutter that disrupts symmetric pattern matching, whereas conventional convolutional neural networks lack rotation equivariance, leading to inconsistent feature representations under rotational transformations. To address these issues, we propose an Asymmetric Region Denoising (ARD) module and a Rotation Equivariant Feature Similarity Matching (REFSM) module. The ARD module suppresses asymmetric interference to refine symmetric patterns, while the REFSM module enhances rotation equivariance through feature similarity matching between original and rotated images. Specifically, our dual-input REFSM framework leverages rotation loss to maximize consistency between the score maps of original and rotated images, thereby enabling precise prediction of rotation-equivariant symmetry axes. Furthermore, we introduce GMSYM, a new benchmark dataset that categorizes images into diverse scenarios and incorporates various interferences to address the limitations of existing reflection symmetry detection benchmarks. Extensive experiments on four standard datasets (DENDI, NYU, LDRS, SDRW) and our proposed GMSYM dataset demonstrate that our method achieves state-of-the-art performance in both accuracy and robustness.
Chinese Translation
反射对称性检测仍然面临挑战，主要由于非对称区域的干扰和对称模式的任意方向。非对称区域引入了背景杂波，干扰了对称模式匹配，而传统的卷积神经网络缺乏旋转等变性，导致在旋转变换下特征表示不一致。为了解决这些问题，我们提出了一种非对称区域去噪（Asymmetric Region Denoising，ARD）模块和一种旋转等变特征相似性匹配（Rotation Equivariant Feature Similarity Matching，REFSM）模块。ARD模块抑制非对称干扰，以精炼对称模式，而REFSM模块通过对原始图像和旋转图像之间的特征相似性匹配来增强旋转等变性。具体而言，我们的双输入REFSM框架利用旋转损失最大化原始图像和旋转图像的得分图之间的一致性，从而实现对旋转等变对称轴的精确预测。此外，我们引入了GMSYM，一个新的基准数据集，将图像分类为多种场景，并结合各种干扰，以解决现有反射对称性检测基准的局限性。在四个标准数据集（DENDI、NYU、LDRS、SDRW）和我们提出的GMSYM数据集上进行的广泛实验表明，我们的方法在准确性和鲁棒性方面均达到了最先进的性能。
cs.CV / 47 / 2607.27952
LAST: The Last Query Token Guides Visual Token Pruning for Edge-Cloud Collaborative MLLM Inference
LAST：最后查询令牌引导边缘-云协作 MLLM 推理的视觉令牌剪枝
Abstract
Multimodal foundation models are reshaping edge-cloud visual intelligence from task-specific feature pipelines into token-based interfaces, where edge devices encode visual inputs into tokens for a general-purpose cloud MLLM. However, dense visual-token sequences increase cloud-side inference costs. Existing pruning methods mainly target centralized inference: vision-driven methods can operate before cloud execution but are typically query-agnostic, whereas query-guided methods often rely on internal states of the target MLLM and cannot determine token relevance before transmission. Compact guidance models offer an alternative, but existing designs may require costly attention aggregation or auxiliary generation. We propose LAST, a training-free framework for query-dependent visual token pruning in edge-cloud collaborative MLLM inference. LAST uses a compact edge-side VLM as a guidance proxy and derives a lightweight importance signal from the last query token's attention to visual tokens. Under causal attention, the last query token can attend to the full visual sequence and the entire query context, enabling query-aware pruning without cloud-model access, autoregressive generation, or costly aggregation over multiple query positions. LAST then retains a diverse set of query-relevant visual tokens under a fixed token budget. We evaluate LAST on 11 multimodal benchmarks under multiple token budgets against pruning methods with different guidance strategies. Experiments show that LAST consistently achieves the strongest performance, preserving 95.4% of the full-token accuracy while retaining only 12.5% of the visual tokens, with low edge-side selection overhead and reduced cloud-side computation.
Chinese Translation
多模态基础模型正在将边缘-云视觉智能从特定任务的特征管道转变为基于令牌的接口，其中边缘设备将视觉输入编码为通用云 MLLM 的令牌。然而，密集的视觉令牌序列增加了云端推理的成本。现有的剪枝方法主要针对集中式推理：视觉驱动的方法可以在云执行之前操作，但通常对查询无关，而查询引导的方法往往依赖于目标 MLLM 的内部状态，无法在传输之前确定令牌的相关性。紧凑的引导模型提供了一种替代方案，但现有设计可能需要昂贵的注意力聚合或辅助生成。我们提出了 LAST，这是一个无训练的框架，用于边缘-云协作 MLLM 推理中的查询依赖视觉令牌剪枝。LAST 使用一个紧凑的边缘侧视觉语言模型（VLM）作为引导代理，并从最后查询令牌对视觉令牌的注意力中推导出轻量级的重要性信号。在因果注意力下，最后的查询令牌可以关注整个视觉序列和整个查询上下文，从而实现查询感知的剪枝，而无需访问云模型、进行自回归生成或在多个查询位置上进行昂贵的聚合。LAST 然后在固定的令牌预算下保留一组多样的查询相关视觉令牌。我们在 11 个多模态基准上评估 LAST，使用多种令牌预算与不同引导策略的剪枝方法进行对比。实验表明，LAST 始终实现了最强的性能，在仅保留 12.5% 的视觉令牌的同时，保持了 95.4% 的全令牌准确率，且边缘侧选择开销低，云侧计算减少。
cs.CV / 48 / 2607.27959
FiRE: Enhancing MLLMs with Fine-Grained Context Learning for Complex Image Retrieval
FiRE：通过细粒度上下文学习增强多模态大语言模型以实现复杂图像检索
Abstract
Due to their strong generalizable multimodal processing and reasoning capabilities, Multimodal Large Language Models (MLLMs) have demonstrated significant potential as universal image retrievers, effectively addressing diverse real-world image retrieval tasks. Nevertheless, pioneering studies, while promising, overlook the potential of fine-grained context modeling and disentangled fine-tuning objectives in enhancing MLLMs' retrieval performance, particularly for complex tasks such as long-text-to-image retrieval, visual dialog retrieval, and composed image retrieval (CIR). Therefore, in this work, we propose an automated fine-grained multimodal quintuple dataset construction pipeline and a novel two-stage fine-grained multimodal fine-tuning strategy. The dataset generation pipeline produces a comprehensive CIR dataset with fine-grained image captions and modification text, facilitating fine-grained context modeling. Beyond the previously entangled fine-tuning paradigm, our approach separates the fine-tuning process into two distinct stages: (1) fine-grained context reasoning-oriented fine-tuning and (2) fine-grained retrieval-oriented fine-tuning. These stages aim to sequentially enhance the model's context understanding and query-target alignment capabilities, thereby improving retrieval performance. Extensive experiments across five datasets encompassing diverse and complex image retrieval tasks demonstrate the remarkable superiority of our method over existing approaches in zero-shot retrieval settings, even with a more lightweight MLLM backbone compared to those methods.
Chinese Translation
由于其强大的可泛化多模态处理和推理能力，多模态大语言模型（MLLMs）在作为通用图像检索器方面展现出显著潜力，有效应对多样化的现实世界图像检索任务。然而，尽管开创性的研究充满希望，却忽视了细粒度上下文建模和解耦细调目标在提升MLLMs检索性能方面的潜力，特别是在长文本到图像检索、视觉对话检索和组合图像检索（CIR）等复杂任务中。因此，在本研究中，我们提出了一种自动化的细粒度多模态五元组数据集构建管道和一种新颖的两阶段细粒度多模态细调策略。数据集生成管道生成了一个全面的CIR数据集，包含细粒度的图像标题和修改文本，促进了细粒度上下文建模。超越之前纠缠的细调范式，我们的方法将细调过程分为两个不同的阶段：（1）面向细粒度上下文推理的细调和（2）面向细粒度检索的细调。这些阶段旨在顺序增强模型的上下文理解和查询-目标对齐能力，从而提高检索性能。在五个涵盖多样化和复杂图像检索任务的数据集上进行的大量实验表明，我们的方法在零-shot检索设置中显著优于现有方法，即使与那些方法相比，我们的MLLM骨干网络更轻量。
cs.CV / 49 / 2607.27969
FootprintNet: State-Transition-Guided Dynamic Footprint Learning for Multi-temporal Remote Sensing Change Detection
FootprintNet：基于状态转移引导的动态足迹学习用于多时相遥感变化检测
Abstract
Despite substantial progress in remote sensing multi-temporal change detection (MTCD), most existing MTCD methods still represent the dynamic process at each spatial location over the entire observation period using a single change category associated with the final observation. This implicit single-change assumption limits their ability to characterize regions of recurrent change closely related to human activities. To address this limitation, we introduce Urban Building Dynamics Detection (UBDD), which identifies building-change dynamic footprints, i.e., the temporal intervals in which changes occur, from multi-temporal imagery and produces pixel-wise classification masks. For regions undergoing two or more changes, UBDD introduces an independent multi-change class for unified representation, thereby enabling unified modeling of single- and multi-change processes. Furthermore, we propose FootprintNet, which abstracts building-change processes as interactions between latent states and actions, and imposes state-action transition constraints to guide the learning of causally coherent change trajectories. It further exploits temporal change-boundary cues to enhance feature contrast across boundary sides, thereby improving the discrimination among different dynamic footprints and enabling accurate detection of dynamic footprints. Moreover, we introduce the Building Change Dynamics Score (BCDS) to address the inability of conventional metrics to reflect the temporal proximity between predicted footprints and labels. It evaluates predictions according to their preservation of change semantics and temporal offsets from the corresponding labels. Extensive experiments on TSCD, MUDS, and WUSU demonstrate that FootprintNet outperforms current state-of-the-art methods. The code is available at https://github.com/zmoka-zht/FootprintNet.
Chinese Translation
尽管在遥感多时相变化检测（MTCD）方面取得了显著进展，但现有的大多数MTCD方法仍然使用与最终观测相关的单一变化类别来表示整个观察期间内每个空间位置的动态过程。这种隐含的单一变化假设限制了它们对与人类活动密切相关的重复变化区域的表征能力。为了解决这一限制，我们引入了城市建筑动态检测（UBDD），该方法从多时相影像中识别建筑变化的动态足迹，即变化发生的时间间隔，并生成逐像素分类掩膜。对于经历两次或多次变化的区域，UBDD引入了一个独立的多变化类别以实现统一表征，从而实现单一变化和多变化过程的统一建模。此外，我们提出了FootprintNet，将建筑变化过程抽象为潜在状态与动作之间的交互，并施加状态-动作转移约束以引导因果一致的变化轨迹学习。它进一步利用时间变化边界线索增强边界两侧的特征对比，从而改善不同动态足迹之间的区分能力，并实现动态足迹的准确检测。此外，我们引入了建筑变化动态评分（BCDS），以解决传统指标无法反映预测足迹与标签之间时间接近性的问题。它根据预测结果对变化语义的保留及与相应标签的时间偏移进行评估。在TSCD、MUDS和WUSU上的大量实验表明，FootprintNet优于当前的最先进方法。代码可在https://github.com/zmoka-zht/FootprintNet获取。
cs.CV / 50 / 2607.27974
Now You Have My Healthy Attention: A U-DiT for Brain-MRI Inpainting
现在你获得了我的健康关注：一种用于脑部MRI修复的U-DiT
Abstract
The ASNR-MICCAI BraTS Local Synthesis (Inpainting) task asks for the anatomically plausible completion of healthy brain tissue within a masked region of a T1-weighted MRI, providing a tumor-free anatomical reference for downstream analysis. As the task is scored by distortion metrics (SSIM, PSNR, MSE), we build a deterministic regression model and focus on giving it inductive biases tailored to inpainting. Our network follows the U-DiT principle of performing self-attention on a downsampled token grid: a volumetric encoder-decoder imports long-range context through a downsampled global self-attention block with three-dimensional rotary position embeddings, while convolutions and skip connections preserve high-frequency detail. Two ideas drive our results. First, we constrain the attention so that occluded ("void") tokens attend only to known-healthy tokens of the same volume, with a learned bias toward each query's contralateral homologue, forcing the completion to be inferred from observed anatomy rather than from other unknown regions. Second, we add a contralateral-symmetry input that supplies the mirrored healthy hemisphere as a patient-specific prior; since the brain is approximately bilaterally symmetric and lesions are typically unilateral, this prior improves the distortion metrics at matched structural similarity. On the official BraTS-2026 validation leaderboard our submission reaches a mean healthy-region SSIM of $0.864$, PSNR of $24.7$\,dB and MSE of $4.6{\times}10^{-3}$ over $219$ cases. We further analyse the residual smoothness inherent to distortion-optimal regression and discuss its implications for anatomical realism.
Chinese Translation
ASNR-MICCAI BraTS局部合成（修复）任务要求在T1加权MRI的掩蔽区域内对健康脑组织进行解剖上合理的补全，为后续分析提供无肿瘤的解剖参考。由于该任务的评分依据是失真度量（SSIM、PSNR、MSE），我们构建了一个确定性回归模型，并专注于为其提供针对修复的归纳偏置。我们的网络遵循U-DiT原则，在下采样的标记网格上执行自注意力：一个体积编码器-解码器通过一个下采样的全局自注意力模块导入长距离上下文，该模块使用三维旋转位置嵌入，同时卷积和跳跃连接保留高频细节。两个思路推动了我们的结果。首先，我们限制注意力，使被遮挡（“空”）的标记仅关注同一体积的已知健康标记，并对每个查询的对侧同源体施加学习偏置，迫使补全从观察到的解剖结构推断，而不是从其他未知区域。其次，我们添加了一个对侧对称输入，提供镜像的健康半球作为患者特定的先验；由于大脑大致是双侧对称的，而病变通常是单侧的，这一先验在匹配的结构相似性下改善了失真度量。在官方BraTS-2026验证排行榜上，我们的提交在219个案例中达到了平均健康区域SSIM为0.864，PSNR为24.7 dB，MSE为4.6×10^{-3}。我们进一步分析了失真最优回归中固有的残差平滑性，并讨论其对解剖现实性的影响。
cs.CV / 51 / 2607.27982
ViP-Rig: Visual-Prompted Controllable Rigging
ViP-Rig：视觉提示可控的绑定方法
Abstract
Rigging is inherently task-dependent because the same mesh may require different skeletons and deformation behaviors across animation tasks. In practice, artists often inspect an initial rig and repeatedly edit its skeletal structure and deformation behavior to meet specific animation requirements. Existing automatic methods primarily generate a plausible rig from geometry, offering limited explicit control over the resulting skeleton and deformation behavior. In this work, we present ViP-Rig, a visual-prompted framework that supports both prompt-first rigging and result-guided editing by injecting features extracted from user-drawn or edited 2D skeletal and rigidity prompts into frozen pretrained backbones. Specifically, ViP-Rig consists of two stages, Skeleton Generation and Skinning Prediction. In the first stage, the skeletal sketch is processed by the Dense-to-Compact Visual Prompt Encoding to produce compact, fixed-length conditioning tokens. The resulting tokens are injected into a frozen pretrained autoregressive generator through gated adapters to control joint placement and branching structure while preserving the generator's geometric prior. In the second stage, the rigidity map is processed using the same visual encoding design, while the pretrained skinning backbone remains frozen. The resulting tokens are symmetrically injected into the point and joint streams to modulate point-joint compatibility and the resulting skinning weights. Experiments on Articulation-XL2.0 and zero-shot evaluation on ModelsResource show that ViP-Rig more accurately recovers target skeletons and skinning weights than geometry-conditioned baselines under prompt-guided evaluation. Qualitative results further demonstrate explicit and localized control in both prompt-first rigging and result-guided editing.
Chinese Translation
绑定本质上是任务依赖的，因为同一网格在不同的动画任务中可能需要不同的骨架和变形行为。在实际操作中，艺术家通常会检查初始绑定，并反复编辑其骨骼结构和变形行为，以满足特定的动画需求。现有的自动化方法主要从几何形状生成一个合理的绑定，但对生成的骨架和变形行为的显式控制有限。在本研究中，我们提出了ViP-Rig，一种视觉提示框架，支持提示优先的绑定和结果引导的编辑，通过将从用户绘制或编辑的二维骨骼和刚性提示中提取的特征注入到冻结的预训练主干网络中。具体而言，ViP-Rig由两个阶段组成：骨架生成和皮肤预测。在第一个阶段，骨骼草图通过密集到紧凑的视觉提示编码进行处理，以生成紧凑的固定长度条件令牌。生成的令牌通过门控适配器注入到冻结的预训练自回归生成器中，以控制关节位置和分支结构，同时保留生成器的几何先验。在第二个阶段，刚性图通过相同的视觉编码设计进行处理，而预训练的皮肤主干保持冻结。生成的令牌对称地注入到点流和关节流中，以调节点与关节的兼容性以及生成的皮肤权重。在Articulation-XL2.0上的实验和在ModelsResource上的零-shot评估表明，ViP-Rig在提示引导评估下比几何条件基线更准确地恢复目标骨架和皮肤权重。定性结果进一步展示了在提示优先的绑定和结果引导的编辑中显式和局部控制的能力。
cs.CV / 52 / 2607.28005
Deep learning-based hierarchical insect classification using camera trap imagery
基于深度学习的层次化昆虫分类方法：利用相机捕捉图像
Abstract
Declining insect populations make reliable biodiversity monitoring increasingly urgent, yet monitoring of insect biodiversity is hampered by a lack of standardised data and by costly and time-consuming manual identification by expert entomologists. Deep learning-based image classifiers, processing data from automated non-lethal camera traps, have the potential to transform and scale insect biodiversity monitoring. However, challenges remain in acquiring expert-annotated datasets, developing model architectures that generalise well across diverse taxonomic levels and training models on highly imbalanced data. Hierarchical data also benefits from designing models that default to higher-confidence, coarser-level predictions, when uncertain about finer taxonomic levels. In this paper we address these challenges with a deep learning-based hierarchical classification model. First, we present a manually curated, long-tailed dataset of around one million images of insects, extracted from 1,801 camera-trap video recordings and annotated with a five-level, 34-class hierarchy. Further, we adapt a hierarchical classification model architecture to a five-level variable-depth hierarchy, with class-balanced weighting. Our model improves on non-hierarchical classifiers by leveraging biological taxonomy to extract granularity-specific visual features and makes hierarchy-consistent predictions to the deepest taxonomic level that meets a confidence threshold (T = 0.6). Our model achieved a per-level accuracy of 80-99% on test data, across five levels of hierarchy. Furthermore ...
Chinese Translation
昆虫种群的下降使得可靠的生物多样性监测变得愈发紧迫，然而昆虫生物多样性的监测受到标准化数据缺乏以及专家昆虫学家耗时且昂贵的手动鉴定的阻碍。基于深度学习的图像分类器，处理来自自动非致命相机捕捉的数据，有潜力改变和扩展昆虫生物多样性监测。然而，在获取专家标注的数据集、开发在不同分类水平上具有良好泛化能力的模型架构以及在高度不平衡的数据上训练模型方面仍然存在挑战。层次化数据的设计也受益于在对更细分类水平不确定时，默认进行高置信度的粗略预测。在本文中，我们提出了一种基于深度学习的层次化分类模型，以应对这些挑战。首先，我们呈现了一个手动整理的长尾数据集，包含约一百万张昆虫图像，这些图像来自1,801个相机捕捉的视频记录，并按照五级、34类的层次结构进行了标注。此外，我们将层次化分类模型架构调整为一个五级可变深度的层次结构，并进行了类别平衡加权。我们的模型通过利用生物分类法提取特定粒度的视觉特征，改善了非层次化分类器，并在满足置信阈值（T = 0.6）的情况下，做出一致的层次预测，达到最深的分类水平。我们的模型在测试数据上实现了各层级80-99%的准确率，涵盖了五个层次的层次结构。
cs.CV / 53 / 2607.28007
Beyond Classification: Pathology Foundation Models as Detection Encoders for Mitotic Figures
超越分类：病理基础模型作为有丝分裂细胞检测编码器
Abstract
Pathology foundation models (FMs) are models trained on vast amounts of typically unlabeled data and have been shown to yield regularized latent spaces that can be used effectively in downstream classification tasks. This is also true for the classification of mitotic figures vs. other cells. However, it is so far unclear if the latent space of current FMs provides features that are discriminant and spatially suitably resolved to also serve as a backbone for dense object detection paradigms. In this work, we investigate this question for common current pathology FMs (UNI, UNI2-h, Virchow, Virchow2, H-optimus-0, H-optimus-1) and compare their performance against a fully end-to-end trained baseline based on a ResNet50 architecture. We combine FM backbones with representatives of single stage, dual stage and self-attention-based detectors (RetinaNet, Faster R-CNN, Deformable DETR respectively) on the multi-domain MIDOG++ dataset, and on the TUPAC16 dataset as an out-of-domain case. We show that the H-optimus-0 and Virchow models yielded competitive performance, indicating that the latent spaces of current FMs, all trained on image-level self-supervision, are suitable for direct mitotic figure detection and may be slightly more robust on our out-of-domain test case. All code is made available publicly at https://github.com/DeepMicroscopy/FM4MFdet.
Chinese Translation
病理基础模型（FMs）是基于大量通常未标记数据训练的模型，已被证明能够产生可有效用于下游分类任务的正则化潜在空间。这同样适用于有丝分裂细胞与其他细胞的分类。然而，目前尚不清楚当前FMs的潜在空间是否提供了可区分且空间分辨率适当的特征，以便作为密集目标检测范式的基础。在本研究中，我们探讨了这一问题，针对常见的当前病理FMs（UNI、UNI2-h、Virchow、Virchow2、H-optimus-0、H-optimus-1）进行研究，并将其性能与基于ResNet50架构的完全端到端训练基线进行比较。我们将FM骨干网络与单阶段、双阶段和基于自注意力的检测器（分别为RetinaNet、Faster R-CNN、Deformable DETR）结合，应用于多域MIDOG++数据集，以及作为域外案例的TUPAC16数据集。我们展示了H-optimus-0和Virchow模型表现出竞争力的性能，表明当前FMs的潜在空间（均基于图像级自监督训练）适合直接进行有丝分裂细胞检测，并且在我们的域外测试案例中可能略显更为稳健。所有代码已公开发布在https://github.com/DeepMicroscopy/FM4MFdet。
cs.CV / 54 / 2607.28020
ENCORE: Event-Assisted Complementary Motion Refinement for Learned Video Compression
ENCORE：基于事件辅助的学习视频压缩运动补充细化
Abstract
Learned video compression relies on accurate temporal modeling to remove redundancy between adjacent frames. However, most existing codecs infer motion solely from discretely sampled RGB frames, making their estimates vulnerable to fast motion, blur, occlusion, weak texture, low illumination, and abrupt brightness changes. Event cameras asynchronously capture fine-grained intensity changes between RGB timestamps and therefore provide complementary evidence about inter-frame dynamics. We propose ENCORE, an Event-Assisted Complementary Motion Refinement framework for learned video compression. ENCORE first employs Complementary Motion Representation (CMR) to decompose aligned RGB-event features into common and modality-specific motion representations. Spatial Energy and Redundancy-Informed Calibration (SERIC) then identifies event-specific responses that are active and novel relative to RGB, suppresses weak or redundant evidence, and predicts a candidate flow correction. Finally, Energy-Aware Routing (EAR) determines where and how strongly the correction should refine the RGB flow. Events serve solely as an auxiliary modality for motion modeling, while RGB remains the only coding and reconstruction target. Experiments on BS-ERGB, HQ-EVFI, and CED demonstrate consistent gains across datasets and GOP lengths. On BS-ERGB, ENCORE achieves up to 20.80% PSNR-RGB and 22.14% MS-SSIM-RGB BD-rate savings, while retaining clear improvements on the other two datasets.
Chinese Translation
学习视频压缩依赖于准确的时间建模，以消除相邻帧之间的冗余。然而，大多数现有编码器仅从离散采样的RGB帧推断运动，这使得它们的估计容易受到快速运动、模糊、遮挡、弱纹理、低光照和突变亮度变化的影响。事件相机异步捕捉RGB时间戳之间的细粒度强度变化，因此提供了关于帧间动态的补充证据。我们提出了ENCORE，一个基于事件辅助的学习视频压缩运动补充细化框架。ENCORE首先采用补充运动表示（Complementary Motion Representation, CMR）将对齐的RGB-事件特征分解为共同和特定于模态的运动表示。然后，空间能量和冗余信息校准（Spatial Energy and Redundancy-Informed Calibration, SERIC）识别相对于RGB而言活跃且新颖的事件特定响应，抑制弱或冗余证据，并预测候选流修正。最后，能量感知路由（Energy-Aware Routing, EAR）决定修正应在何处以及多强烈地细化RGB流。事件仅作为运动建模的辅助模态，而RGB仍然是唯一的编码和重建目标。在BS-ERGB、HQ-EVFI和CED上的实验表明，在数据集和GOP长度上均获得了一致的增益。在BS-ERGB上，ENCORE实现了高达20.80%的PSNR-RGB和22.14%的MS-SSIM-RGB BD-rate节省，同时在其他两个数据集上也保持了明显的改进。
cs.CV / 55 / 2607.28032
Split and Drive: Dual-Axis Disentanglement for Real-Time Gaussian Head Avatars
分离与驱动：实时高斯头部头像的双轴解耦
Abstract
Creating photorealistic animatable head avatars from a single image remains a fundamental challenge in digital human synthesis. While recent 3D Gaussian Splatting methods have achieved promising results, they rely on external tracking pipelines whose latency is excluded from inference measurements. Furthermore, they adopt unified representations that entangle geometrically distinct facial regions, limiting both expressiveness and rendering fidelity. We propose SpiD (Split and Drive), a single-image Gaussian head avatar framework built on two disentanglement axes. The compute axis internalizes per-frame driving, eliminating external tracking dependency at inference. The feature axis decomposes the avatar into three specialized Gaussian branches, each modeling a geometrically distinct facial domain. Extensive experiments demonstrate consistently strong performance against state-of-the-art methods while achieving the fastest inference speed among all compared methods on a single GPU with the complete driving pipeline included.
Chinese Translation
从单张图像创建可动画的照片级真实感头部头像仍然是数字人类合成中的一项基本挑战。尽管最近的3D高斯点云方法取得了令人鼓舞的成果，但它们依赖于外部跟踪管道，其延迟未计入推理测量。此外，它们采用统一的表示方式，将几何上不同的面部区域纠缠在一起，限制了表现力和渲染保真度。我们提出了SpiD（分离与驱动），这是一个基于两个解耦轴的单图像高斯头部头像框架。计算轴内部化了每帧驱动，消除了推理时对外部跟踪的依赖。特征轴将头像分解为三个专门的高斯分支，每个分支建模一个几何上不同的面部领域。大量实验表明，与最先进的方法相比，SpiD在性能上始终表现强劲，同时在包含完整驱动管道的情况下，在单个GPU上实现了所有比较方法中最快的推理速度。
cs.CV / 56 / 2607.28039
TongueReenact: Geometry-Anchored Tongue Synthesis for Face Reenactment
TongueReenact：基于几何锚定的舌头合成用于面部重现
Abstract
Modern face reenactment systems achieve impressive pose and expression transfer using geometry-driven representations. However, they largely ignore tongue dynamics, leading to anatomically inconsistent mouth interiors during speech and expressive motions. We introduce the first framework for cross-identity tongue dynamics transfer in face reenactment. We propose a foundation-model-assisted bootstrapping pipeline that produces a dedicated tongue segmentation model for in-the-wild reenactment without curated annotations. We further introduce a spatially constrained latent masked diffusion model for realistic tongue synthesis, with adaptive mask dilation for seamless mouth boundary transitions. Extensive experiments demonstrate improvements of more than two times over all baselines on every tongue-specific metric. We additionally propose a VLM-based evaluation protocol that replicates expert annotation at scale, confirming perceptual superiority across all ablation variants.
Chinese Translation
现代面部重现系统通过几何驱动的表示实现了令人印象深刻的姿态和表情转移。然而，它们在很大程度上忽视了舌头的动态，导致在言语和表情动作中口腔内部解剖不一致。我们引入了第一个跨身份舌头动态转移的面部重现框架。我们提出了一种基础模型辅助的自启动管道，能够在没有经过整理的注释的情况下，为野外重现生成专用的舌头分割模型。我们进一步引入了一种空间约束的潜在掩蔽扩散模型，用于真实的舌头合成，并采用自适应掩膜膨胀以实现无缝的口腔边界过渡。大量实验表明，在每个舌头特定指标上，所有基线的性能提升超过两倍。此外，我们还提出了一种基于VLM的评估协议，能够大规模复制专家注释，确认所有消融变体在感知上的优越性。
cs.CV / 57 / 2607.28058
Temporal Concentration from Rollout Errors: Implicit Preference Optimization for Text-to-Video Diffusion
来自回滚误差的时间集中：文本到视频扩散的隐式偏好优化
Abstract
Recent advances in preference alignment for diffusion-based video generation, particularly via Direct Preference Optimization (DPO), have significantly improved visual quality. However, temporally sparse artifacts such as motion collapse, object flickering, and color oversaturation remain a major barrier to perceptual realism. Existing methods struggle with these issues due to two key limitations: (1) the preference attribution bottleneck, where offline human annotations are costly and fail to accurately capture learning dynamics, while online reward signals are rollout-aware but often unstable and biased; and (2) temporal credit misallocation, where uniformly applied supervision cannot effectively target the brief segments in which artifacts occur. To address these challenges, we propose concentrated Implicit Preference Optimization (cIPO), a post-training framework for video diffusion models. cIPO derives implicit preference signals directly from the denoising process: given a real video, the model adds forward noise and reconstructs it via iterative denoising, treating the original as the preferred sample and the reconstruction as the dispreferred one. This formulation captures inference-time errors without requiring human annotations or external reward models. Moreover, frame-level discrepancies between original and reconstructed videos reveal when failures occur. cIPO leverages this by computing temporal reconstruction errors and concentrating optimization on high-error segments, enabling more precise correction of failure-prone regions. Extensive experiments demonstrate that cIPO consistently enhances video authenticity and temporal coherence across multiple datasets, highlighting the effectiveness and efficiency of implicit preference with temporally concentrated optimization.
Chinese Translation
最近在基于扩散的视频生成中的偏好对齐方面取得了显著进展，特别是通过直接偏好优化（Direct Preference Optimization, DPO），极大地提高了视觉质量。然而，时间上稀疏的伪影，如运动崩溃、物体闪烁和颜色过饱和，仍然是感知真实感的主要障碍。现有方法在解决这些问题时面临两个关键限制：（1）偏好归因瓶颈，离线人类标注成本高且无法准确捕捉学习动态，而在线奖励信号虽然对回滚过程敏感，但往往不稳定且存在偏差；（2）时间信用错误分配，均匀应用的监督无法有效针对伪影发生的短暂片段。为了解决这些挑战，我们提出了集中隐式偏好优化（concentrated Implicit Preference Optimization, cIPO），这是一个用于视频扩散模型的后训练框架。cIPO直接从去噪过程中推导隐式偏好信号：给定一个真实视频，模型添加前向噪声并通过迭代去噪重建它，将原始视频视为偏好样本，将重建视频视为不偏好样本。这种表述捕捉了推理时的错误，而无需人类标注或外部奖励模型。此外，原始视频与重建视频之间的帧级差异揭示了失败发生的时刻。cIPO通过计算时间重建误差并将优化集中在高误差片段上，从而利用这一点，使对易失败区域的修正更加精确。大量实验表明，cIPO在多个数据集上始终增强视频的真实性和时间一致性，突显了隐式偏好与时间集中优化的有效性和效率。
cs.CV / 58 / 2607.28064
Landmark shape spaces with induced metrics
具有诱导度量的地标形状空间
Abstract
We present a unification of Kendall's landmark shape spaces, where rigid motions are factored out and scale fixed on landmark configurations equipped with Euclidean geometry, with landmark configuration spaces carrying Riemannian metrics descending from right-invariant Sobolev metrics on the diffeomorphism group. The resulting new landmark shape spaces achieve the defining properties of both approaches: The regularity of the descending metric prevents landmarks from colliding, the metric is defined in the ambient space independent of the number of landmarks, local rigid transformations are preserved, global rigid motions are removed, and scale fixed. To achieve this, we define a particular Sobolev-type operator, the screened elasticity operator, whose null-space consists exactly of the rigid motions, we show how this operator descends to achieve the desired geometry, and we present approaches to solving matching problems and computing geodesics numerically. The resulting construction allows the use of landmark configuration spaces with sufficiently regular metrics in applications while retaining the shape invariances that are a hallmark of Kendall's shape spaces.
Chinese Translation
我们提出了一种对Kendall的地标形状空间的统一，其中刚性运动被排除，且在配备欧几里得几何的地标配置上固定比例，同时地标配置空间承载着从微分同胚群的右不变Sobolev度量降下来的黎曼度量。所得到的新地标形状空间实现了两种方法的定义特性：降下度量的规则性防止地标之间的碰撞，度量在环境空间中定义，与地标的数量无关，局部刚性变换得以保留，全球刚性运动被移除，且比例固定。为了实现这一目标，我们定义了一种特定的Sobolev型算子——筛选弹性算子，其零空间恰好由刚性运动组成，我们展示了该算子如何降下以实现所需的几何结构，并提出了求解匹配问题和数值计算测地线的方法。所得到的构造允许在应用中使用具有足够规则度量的地标配置空间，同时保留Kendall形状空间的标志性形状不变性。
cs.CV / 59 / 2607.28065
BladeYOLO: Wind Turbine Blade Defect Detection with Limited Annotations and Weak-Saliency Awareness
BladeYOLO：基于有限标注和弱显著性意识的风力涡轮机叶片缺陷检测
Abstract
Wind turbine blade defect detection remains highly challenging in real-world inspection scenarios due to limited on-site data and the subtle visual characteristics of defects. In practice, blade defects are often small-scale, low-contrast, and difficult to distinguish from complex backgrounds, which significantly limits the robustness of existing detectors. To address these challenges, we propose BladeYOLO, a defect detection framework for wind turbine blades. Specifically, we integrate a Vision Transformer (ViT) backbone initialized with DINOv3 self-supervised pre-trained weights into YOLOv12-L, enabling the transfer of large-scale generic visual priors to blade defect detection and improving feature representation under limited training annotations. To enhance the perception of subtle defects, we further develop a Mamba-guided Weak-Defect Enhancement module, which consists of a Detail-Enhanced Multi-scale Branch for preserving high-frequency structural cues and a Cross-Mamba module for progressively propagating high-level semantic guidance to shallow features. In addition, we introduce a lightweight Style-Injector module that captures environment-related style information via Fourier decomposition and injects it into selected ViT self-attention layers, thereby improving robustness against environment-induced appearance variations. Extensive experiments demonstrate that BladeYOLO achieves superior performance on the WTBlade-Defect dataset, with additional annotation-budget experiments showing its favorable performance under reduced training annotations. Evaluation on the public Wind Surface Defect dataset further provides supportive evidence for the cross-dataset robustness of BladeYOLO. In particular, on this public dataset, BladeYOLO outperforms the best competing method by 3.5\% in mAP$_{50}$ and 2.5\% in mAP$_{50-95}$.
Chinese Translation
风力涡轮机叶片缺陷检测在实际检查场景中仍然面临重大挑战，主要由于现场数据有限以及缺陷的视觉特征微妙。在实际应用中，叶片缺陷通常规模较小、对比度低，且难以与复杂背景区分，这显著限制了现有检测器的鲁棒性。为了解决这些挑战，我们提出了BladeYOLO，一个用于风力涡轮机叶片缺陷检测的框架。具体而言，我们将一个以DINOv3自监督预训练权重初始化的视觉变换器（Vision Transformer, ViT）主干网络集成到YOLOv12-L中，从而实现将大规模通用视觉先验迁移到叶片缺陷检测，并在有限的训练标注下改善特征表示。为了增强对微小缺陷的感知，我们进一步开发了一个Mamba引导的弱缺陷增强模块，该模块由一个多尺度细节增强分支组成，用于保留高频结构线索，以及一个Cross-Mamba模块，用于逐步将高层语义指导传播到浅层特征。此外，我们引入了一个轻量级的风格注入模块，通过傅里叶分解捕获与环境相关的风格信息，并将其注入到选定的ViT自注意力层中，从而提高对环境引起的外观变化的鲁棒性。大量实验表明，BladeYOLO在WTBlade-Defect数据集上表现优越，额外的标注预算实验显示其在减少训练标注下的良好表现。在公共风表面缺陷数据集上的评估进一步提供了BladeYOLO跨数据集鲁棒性的支持证据。特别是在该公共数据集上，BladeYOLO在mAP$_{50}$上比最佳竞争方法提高了3.5 ext{%}，在mAP$_{50-95}$上提高了2.5 ext{%}。
cs.CV / 60 / 2607.28108
mmRadarTwin: A Measurement-Calibrated Signal-Level Digital Twin Platform for Indoor mmWave Radar
mmRadarTwin：一种用于室内毫米波雷达的测量校准信号级数字双胞胎平台
Abstract
Indoor mmWave radar perception is difficult to reproduce because measured range-angle responses depend on scene geometry, material response, multipath, hardware conventions, and signal processing. Existing ray-tracing and digital-twin tools often expose rendering, channel, or path-level quantities, while radar sensing requires complex signal products that can be processed and compared in the same domain as real FMCW measurements. We present mmRadarTwin, a signal-level and path-attributed digital-twin platform for indoor mmWave radar. mmRadarTwin links a real radar measurement branch with an Unreal Engine scene-simulation branch through a shared receive-channel and range-angle processing interface. The simulator writes complex multi-channel receive grids and exports per-path contribution records that identify the actor, material tag, propagation event, and output-bin support of each simulated return. We evaluate mmRadarTwin in an office deployment using a commodity monostatic mmWave radar and mobile scene-capture hardware. Across 154 measured poses spanning 22 radar locations, the current physics-only path-basis simulator recalls 70.8% of measurement-active geometry-supported response regions in the central usable field of view while exposing residuals caused by weak or missing path support, shifted responses, unsupported anchors, and missing physical mechanisms. Rather than claiming complete radar-map reconstruction or cross-room generalization, mmRadarTwin establishes a practical systems workflow for constructing, comparing, and diagnosing indoor radar digital twins.
Chinese Translation
室内毫米波雷达感知的再现性较差，因为测量的距离-角度响应依赖于场景几何、材料响应、多径效应、硬件约定和信号处理。现有的光线追踪和数字双胞胎工具通常暴露渲染、信道或路径级别的量，而雷达感知需要复杂的信号产品，这些产品可以在与真实FMCW测量相同的领域中进行处理和比较。我们提出了mmRadarTwin，这是一种用于室内毫米波雷达的信号级和路径属性数字双胞胎平台。mmRadarTwin通过共享的接收信道和距离-角度处理接口，将真实雷达测量分支与虚幻引擎场景仿真分支连接起来。该仿真器写入复杂的多通道接收网格，并导出每条路径的贡献记录，这些记录识别每个模拟返回的演员、材料标签、传播事件和输出箱支持。我们在一个办公室部署中评估mmRadarTwin，使用了一种商品化的单基地毫米波雷达和移动场景捕捉硬件。在154个测量姿态中，跨越22个雷达位置，当前的仅物理路径基础仿真器在可用视场的中心回忆了70.8%的测量活跃几何支持响应区域，同时暴露出由于弱或缺失的路径支持、响应偏移、不支持的锚点和缺失的物理机制所导致的残差。mmRadarTwin并不声称实现完整的雷达地图重建或跨房间泛化，而是建立了一个实用的系统工作流程，用于构建、比较和诊断室内雷达数字双胞胎。
cs.CV / 61 / 2607.28125
Towards Practical Algorithm Selection for Unsupervised Domain Adaptation in Medical Imaging
面向医学影像无监督领域适应的实用算法选择
Abstract
Numerous unsupervised domain adaptation (UDA) algori-thms exist, but for clinical practice, selecting the best-suited one along with proper hyperparameters often remains unclear, as the unlabeled deployment (target) domain prevents direct evaluation. We propose a label-free criterion that jointly selects the algorithm and hyperparameters for UDA. Given a pool of candidate models from multiple algorithms trained with different hyperparameters, our approach scores each candidate against an agreement reference, and selects the one with the highest score. The agreement reference is constructed in two levels without using target labels. First, we leverage multiple label-free selection signals, using each to nominate a model within every algorithm. Second, the nominated models are aggregated across algorithms to form a reference prediction for each unlabeled target sample. The candidate whose predictions agree most with this reference is then selected for deployment. Experimental results on four brain MRI and four chest X-ray datasets across seven clinically relevant transfer scenarios show that our method achieves better selection performance than other methods and remains effective across different algorithm pools. Our approach takes a step towards practical, label-free algorithm selection for clinical deployment of UDA.
Chinese Translation
目前存在众多无监督领域适应（UDA）算法，但在临床实践中，选择最合适的算法及其超参数往往不明确，因为未标记的部署（目标）领域阻碍了直接评估。我们提出了一种无标签的标准，能够联合选择UDA的算法和超参数。在多个算法中，从不同超参数训练的候选模型池中，我们的方法对每个候选模型进行评分，并选择得分最高的模型。该一致性参考在不使用目标标签的情况下构建，分为两个层次。首先，我们利用多个无标签的选择信号，使用每个信号在每个算法中提名一个模型。其次，跨算法聚合提名的模型，为每个未标记的目标样本形成参考预测。然后，选择与该参考预测一致性最高的候选模型进行部署。在四个脑部MRI和四个胸部X光数据集的七个临床相关转移场景中的实验结果表明，我们的方法在选择性能上优于其他方法，并且在不同算法池中仍然有效。我们的方法朝着无标签算法选择在UDA临床部署中的实用性迈出了重要一步。
cs.CV / 62 / 2607.28129
Face and Voice Cross-modal Association with Learning Convex Feature Embedding
面部与声音的跨模态关联学习与凸特征嵌入
Abstract
Face-and-voice association learning is one of the most challenging tasks in deep learning. In this paper, we propose a simple but powerful cross-modal feature embedding method for the association of faces and voices. Previous work has studied cross-modal association tasks to establish the correlation between voice clips and facial images. These works have addressed cross-modal discrimination but underestimate the importance of handling heterogeneity in inter-modal features between audio and video, resulting in a lot of false positives and false negatives. To tackle the problem, the proposed method learns the embeddings of cross-modal features by making another feature exist between cross-modal features, facilitating the voice and face features of the same person to be embedded in a convex hull. Moreover, the incorporation of cross-modal attention mechanisms with convex embedding techniques represents a highly effective strategy for the attenuation of false positives and false negatives, accomplished via the minimization of inter-class discrepancies. We exhaustively evaluated our method for cross-modal verification, matching, and retrieval tasks on the large-scale VoxCeleb dataset. Extensive experimental results demonstrate that the proposed method achieves notable improvements over existing state-of-the-art methods.
Chinese Translation
面部与声音的关联学习是深度学习中最具挑战性的任务之一。本文提出了一种简单但强大的跨模态特征嵌入方法，用于面部与声音的关联。之前的研究探讨了跨模态关联任务，以建立声音片段与面部图像之间的相关性。这些研究虽然解决了跨模态判别问题，但低估了处理音频与视频之间的跨模态特征异质性的重要性，导致大量的假阳性和假阴性。为了解决这一问题，所提出的方法通过在跨模态特征之间引入另一种特征来学习跨模态特征的嵌入，从而使同一人的声音和面部特征能够嵌入到一个凸包中。此外，将跨模态注意机制与凸嵌入技术相结合，代表了一种高度有效的策略，用于减轻假阳性和假阴性，这通过最小化类间差异来实现。我们在大规模VoxCeleb数据集上对跨模态验证、匹配和检索任务对我们的方法进行了全面评估。大量实验结果表明，所提出的方法在现有最先进的方法上取得了显著的改进。
cs.CV / 63 / 2607.28130
Collaborative Feature Aggregation for Face Super-Resolution and Robust Re-Identification
面部超分辨率与鲁棒性再识别的协同特征聚合
Abstract
We propose a novel collaborative approach for face super-resolution (SR) and robust person re-identification from sequential or multi-view facial images. Traditional SR methods often suffer from blurring and distortion in faces recovered from poor-quality images due to low resolution. Image- and video-based facial SR methods using facial landmarks or segmentation also have similar challenges. To overcome these limitations, we leverage multiple correlated facial observations, across time or viewpoints, by introducing a transformer-based collaborative feature aggregation method that unifies identity features from multi-sequence or multi-view data. This allows faces in multiple sequences of an individual to contribute to accurately estimating common facial features. Furthermore, we propose a cascade SR network to progressively restore the high-resolution image of the target's face with gradual facial feature unification. The unified identity representation is further utilized in person re-identification scenarios, enabling accurate matching even under severe image degradation. The exhaustive experimental results and comparisons show that our method outperforms other state-of-the-art methods, demonstrating consistent improvements in both face super-resolution and re-identification performance. Our work highlights the effectiveness of joint identity reconstruction and progressive image restoration from multiple facial inputs in enhancing downstream visual recognition tasks.
Chinese Translation
我们提出了一种新颖的协同方法，用于从连续或多视角面部图像中进行面部超分辨率（SR）和鲁棒性个体再识别。传统的SR方法在从低分辨率的低质量图像恢复面部时，常常会遭遇模糊和失真。基于图像和视频的面部SR方法使用面部特征点或分割，也面临类似的挑战。为了克服这些局限性，我们通过引入一种基于变换器的协同特征聚合方法，利用多个相关的面部观察数据（跨时间或视角），统一来自多序列或多视角数据的身份特征。这使得个体在多个序列中的面部图像能够共同贡献，以准确估计常见的面部特征。此外，我们提出了一种级联SR网络，逐步恢复目标面部的高分辨率图像，并逐渐统一面部特征。统一的身份表示在个体再识别场景中进一步得到应用，即使在严重的图像降质情况下也能实现准确匹配。详尽的实验结果和比较表明，我们的方法优于其他最先进的方法，在面部超分辨率和再识别性能上均表现出一致的提升。我们的研究突显了从多个面部输入中联合身份重建和逐步图像恢复在增强下游视觉识别任务中的有效性。
cs.CV / 64 / 2607.28132
Convolutional Neural Shading for High-Quality 3D Reconstruction from Multi-View Images
基于卷积神经网络的高质量三维重建多视图图像阴影处理
Abstract
We propose a convolutional neural shading (CNS), a novel pipeline to reconstruct high-quality 3D shapes from multi-view images. Several recent studies have used neural radiance fields and other neural differentiable rendering methods to understand 3D geometry. However, these approaches rely on single-point geometric information, such as positions and normals of the surface, leading to a lack of detailed local geometry. Our approach addresses the inherent limitations of single-point information by leveraging a neural shader to capture variations even in dark and textureless regions with a convolutional neural shader, resulting in far more accurate geometry predictions. Additionally, our method mitigates surface irregularities at image boundaries by introducing a fine-detail displacement network, which utilizes spatial information of surface geometry and learns fine displacement details by correlating neighboring values in the rendering coordinates. Through extensive experiments, our proposed method has demonstrated significant quality improvements in the reconstructed shapes and rendered images over current state-of-the-art methods.
Chinese Translation
我们提出了一种卷积神经阴影（Convolutional Neural Shading, CNS），这是一种新颖的管道，用于从多视图图像重建高质量的三维形状。最近的几项研究已使用神经辐射场和其他神经可微渲染方法来理解三维几何。然而，这些方法依赖于单点几何信息，如表面的位置信息和法线，导致缺乏详细的局部几何信息。我们的方法通过利用神经阴影器来捕捉即使在黑暗和无纹理区域的变化，从而解决了单点信息的固有局限性，最终实现了更为准确的几何预测。此外，我们的方法通过引入细节位移网络来减轻图像边界处的表面不规则性，该网络利用表面几何的空间信息，并通过关联渲染坐标中的邻近值来学习细微的位移细节。通过大量实验，我们的方法在重建形状和渲染图像的质量上相较于当前最先进的方法显示出了显著的提升。
cs.CV / 65 / 2607.28148
What Makes Deep Learning Work for Traditional Chinese Medicine Tongue Diagnosis? A Comprehensive Ablation Study
深度学习如何在传统中医舌诊中发挥作用？一项综合消融研究
Abstract
Deep learning has shown promise for automated tongue diagnosis in traditional Chinese medicine (TCM), yet the design space remains underexplored. We conducted a systematic ablation study spanning 20+ model versions under rigorous 5-fold cross-validation on TongueDx2 (5,109 images, 976 expert-annotated) and a merged dataset of 11,101 samples. We compared six backbone architectures, four loss functions, five augmentation strategies, and six training strategies. The best 976-sample model achieved weighted-F1 of 0.6625 using ConvNeXt-Tiny with restrained augmentation and weak-group ensemble, while the best 11,101-sample model reached weighted-F1 of 0.7761. Six key design principles emerged: (1) ConvNeXt-Tiny offers optimal parameter efficiency; (2) BCE substantially outperforms Asymmetric Loss (+2.7%); (3) restrained color augmentation is critical; (4) weak-group ensemble replacement (+2.1%) outperforms probability averaging; (5) data scaling yielded +20.6% improvement; (6) expanding from 13 to 45 label dimensions caused catastrophic collapse (0.78 to 0.22). These principles are generalizable to multi-label medical image classification with class imbalance.
Chinese Translation
深度学习在传统中医（TCM）自动舌诊中展现出潜力，但设计空间仍未得到充分探索。我们进行了系统的消融研究，涵盖了20多个模型版本，并在TongueDx2（5,109张图像，976张专家标注）和一个合并的11,101个样本数据集上进行了严格的5折交叉验证。我们比较了六种主干架构、四种损失函数、五种数据增强策略和六种训练策略。最佳的976样本模型使用ConvNeXt-Tiny，经过限制性增强和弱组集成，达到了加权F1值0.6625，而最佳的11,101样本模型达到了加权F1值0.7761。六个关键设计原则浮现： (1) ConvNeXt-Tiny提供了最佳的参数效率； (2) BCE显著优于不对称损失（+2.7%）； (3) 限制性颜色增强至关重要； (4) 弱组集成替代（+2.1%）优于概率平均； (5) 数据缩放带来了+20.6%的提升； (6) 从13个标签维度扩展到45个标签维度导致了灾难性崩溃（0.78降至0.22）。这些原则可以推广到具有类别不平衡的多标签医学图像分类中。
cs.CV / 66 / 2607.28154
OPLD: On-Policy Latent Distillation for Multimodal Reasoning
OPLD：用于多模态推理的在线潜在蒸馏
Abstract
Interleaved multimodal Chain-of-Thought (CoT) improves visual reasoning by incorporating auxiliary visual evidence into intermediate reasoning. However, existing approaches remain constrained by externally defined reasoning traces and visual operations, limiting their ability to develop flexible and abstract visual thinking. Reasoning with latent has recently offered a promising direction by internalizing intermediate computation into continuous representations. Nevertheless, existing visual-latent methods mainly supervise latent states through alignment with compressed auxiliary visual features, treating them as proxies for visual observations rather than active reasoning states. Consequently, they capture the provided evidence but fail to fully internalize the abstract reasoning process induced by multimodal CoT. In this paper, we propose OPLD (On-Policy Latent Distillation), a simple framework that transfers the reasoning capability induced by privileged multimodal CoT into latent reasoning representations. Extensive experiments on diverse multimodal benchmarks demonstrate that OPLD consistently outperforms existing latent reasoning methods and achieves state-of-the-art performance on multiple benchmarks. The results suggest that supervising latent representations at the reasoning-process level provides a more effective paradigm for multimodal latent reasoning than conventional feature-level alignment.
Chinese Translation
交错的多模态思维链（Chain-of-Thought, CoT）通过将辅助视觉证据纳入中间推理，提升了视觉推理的能力。然而，现有的方法仍然受到外部定义的推理轨迹和视觉操作的限制，限制了其发展灵活和抽象的视觉思维的能力。最近，潜在推理提供了一种有前景的方向，通过将中间计算内化为连续表示来实现。然而，现有的视觉潜在方法主要通过与压缩的辅助视觉特征对齐来监督潜在状态，将其视为视觉观测的代理，而非主动的推理状态。因此，它们捕捉了提供的证据，但未能充分内化由多模态 CoT 引发的抽象推理过程。本文提出了 OPLD（在线潜在蒸馏），这是一个简单的框架，将特权多模态 CoT 诱发的推理能力转移到潜在推理表示中。在多样化的多模态基准上的广泛实验表明，OPLD 始终优于现有的潜在推理方法，并在多个基准上实现了最先进的性能。结果表明，在推理过程层面上监督潜在表示为多模态潜在推理提供了一种比传统特征级对齐更有效的范式。
cs.CV / 67 / 2607.28164
S-Avatar: Diffusion-Guided Gaussian Head Avatars from a Single Image
S-Avatar：基于扩散引导的单图像高斯头部头像生成
Abstract
We propose S-Avatar, a novel method for generating photorealistic 3D head avatars from a single image using a diffusion-guided 3D model generation module and strategies for animating 3D Gaussian Splatting (3DGS). While single-image head avatar reconstruction is crucial for lifelike Virtual Reality (VR) applications, existing approaches often struggle to preserve 3D consistency under unseen viewpoints. S-Avatar addresses this limitation through a three-stage pipeline. First, a high-resolution 3DGS is synthesized directly from a single image using a diffusion-based Gaussian splat generation module. Next, the parametric head model FLAME is aligned with the generated 3DGS by optimizing its parameters and spatial transformations. Finally, to adapt the 3DGS to FLAME variations, we construct a binding template that encodes the spatial relationship between the initial splats and FLAME. The dynamic 3D head avatar can then be rendered in real time by deforming the 3DGS with the binding template. By combining diffusion-guided canonical 3DGS generation with FLAME-based control, our method achieves efficient and accurate reconstruction with enhanced 3D consistency. Evaluations on public datasets demonstrate that S-Avatar outperforms state-of-the-art methods in novel-view and expression generation, achieving superior realism and consistency. Consequently, our approach represents a significant advance in accessible avatar creation, applicable to a wide range of VR/AR applications. The project page is available at https://github.com/hailsong/savatar.
Chinese Translation
我们提出了S-Avatar，一种从单张图像生成照片级真实感3D头部头像的新方法，该方法使用扩散引导的3D模型生成模块和3D高斯点云（3DGS）动画策略。单图像头部头像重建对于逼真的虚拟现实（VR）应用至关重要，但现有方法在未见视角下往往难以保持3D一致性。S-Avatar通过三阶段流程解决了这一限制。首先，使用基于扩散的高斯点云生成模块直接从单张图像合成高分辨率的3DGS。接下来，通过优化其参数和空间变换，将参数化头部模型FLAME与生成的3DGS对齐。最后，为了将3DGS适应FLAME的变化，我们构建了一个绑定模板，该模板编码了初始点云与FLAME之间的空间关系。然后，通过使用绑定模板变形3DGS，可以实时渲染动态3D头部头像。通过将扩散引导的标准3DGS生成与基于FLAME的控制相结合，我们的方法实现了高效且准确的重建，并增强了3D一致性。在公共数据集上的评估表明，S-Avatar在新视角和表情生成方面优于最先进的方法，实现了更高的真实感和一致性。因此，我们的方法代表了可访问头像创建的重大进展，适用于广泛的VR/AR应用。项目页面可访问：https://github.com/hailsong/savatar。
cs.CV / 68 / 2607.28186
Think with Extra-Image: A Farmland Segmentation Agent Driven by Spatio-Temporal Information Gain
借助额外图像思考：一种由时空信息增益驱动的农田分割代理
Abstract
Existing farmland remote sensing image (FRSI) segmentation follows a "Think with Intra-Image" paradigm, assuming that the current image contains sufficient visual evidence for reliable segmentation. Yet farmland appearance varies with phenology and spatial context and is often confused with other land-cover, making instantaneous, local observations inadequate. Thus, segmentation ambiguity stems not only from limited model representation, but more fundamentally from the required spatio-temporal information lying beyond the current image. Based on this insight, we redefine FRSI segmentation from an information bottleneck perspective as a dynamic decision process driven by task-relevant extra spatio-temporal information gain. We further propose FarmSeeker, a dynamic FRSI segmentation agent that identifies ambiguous regions, reasons about their causes, and queries extra spatio-temporal information on demand for accurate segmentation. To evaluate FarmSeeker, we construct GSFS-Bench, the first global-scale, high-resolution FRSI segmentation benchmark that supports reasoning-querying. Experiments show that FarmSeeker achieves more stable segmentation performance than existing methods. The project is publicly available at: https://withoutocean.github.io/FarmSeeker/
Chinese Translation
现有的农田遥感图像（FRSI）分割遵循“借助图像内部思考”的范式，假设当前图像包含足够的视觉证据以实现可靠的分割。然而，农田的外观随着物候和空间背景的变化而变化，常常与其他地表覆盖混淆，使得瞬时的局部观察变得不足。因此，分割模糊性不仅源于模型表示的局限性，更根本地源于所需的时空信息超出了当前图像的范围。基于这一洞察，我们从信息瓶颈的角度重新定义FRSI分割，将其视为一个由任务相关的额外时空信息增益驱动的动态决策过程。我们进一步提出了FarmSeeker，一个动态的FRSI分割代理，能够识别模糊区域，推理其原因，并根据需要查询额外的时空信息以实现准确的分割。为了评估FarmSeeker，我们构建了GSFS-Bench，这是第一个支持推理查询的全球尺度高分辨率FRSI分割基准。实验表明，FarmSeeker在分割性能上比现有方法更为稳定。该项目已公开发布，网址为：https://withoutocean.github.io/FarmSeeker/
cs.CV / 69 / 2607.28211
Scaling Vision-Language Models Is Not Enough to Mitigate Bias
扩展视觉-语言模型不足以缓解偏见
Abstract
Vision-Language Models (VLMs) such as CLIP are now foundational to multimodal systems, yet their robustness to spurious correlations remains poorly understood at scale. We present the first large-scale empirical study of 194 publicly available VLMs, including 16 model families, covering a wide range of model sizes, 24 training datasets, and three evaluation benchmarks, namely ImageNet (overall performance), CelebA (typical single-attribute bias), and UrbanCars (complex multi-attribute biases). Across these settings, the Spearman correlation between model scale and performance weakens as evaluation shifts from ImageNet ($\rho{=}0.68$) to single-attribute ($\rho{=}0.48$) and further to multi-attribute ($\rho{=}0.05$) bias benchmarks. In contrast, properties of the training data (size and quality) show more consistent relationships with worst-group accuracy across both bias benchmarks. Notably, curated datasets yield improvements of up to 25% over uncurated alternatives at a comparable scale. Finally, the effect of architectural choices (e.g., patch size, image resolution) is highly context-dependent, varying with the nature of the benchmark, including the type of bias and its spatial distribution within images.
Chinese Translation
视觉-语言模型（VLMs），如 CLIP，已成为多模态系统的基础，但其在大规模下对虚假相关性的鲁棒性仍然缺乏深入理解。我们首次对194个公开可用的VLMs进行了大规模实证研究，包括16个模型家族，涵盖了广泛的模型规模、24个训练数据集以及三个评估基准，即 ImageNet（整体性能）、CelebA（典型单属性偏见）和 UrbanCars（复杂多属性偏见）。在这些设置中，模型规模与性能之间的斯皮尔曼相关性随着评估从 ImageNet（$
ho{=}0.68$）转向单属性（$
ho{=}0.48$）再到多属性（$
ho{=}0.05$）偏见基准而减弱。相比之下，训练数据的属性（规模和质量）在两个偏见基准上与最差组准确率之间表现出更一致的关系。值得注意的是，经过精心策划的数据集在可比规模下相较于未经策划的替代品提高了多达25%的性能。最后，架构选择（例如，补丁大小、图像分辨率）的影响高度依赖于上下文，随着基准的性质（包括偏见类型及其在图像中的空间分布）而变化。
cs.CV / 70 / 2607.28225
FaithEyes: Towards Faithful Tool Use via Multi-Agent Process-Image Verification
FaithEyes：通过多智能体过程图像验证实现工具使用的可信性
Abstract
Agentic vision-language models (VLMs), which interleave textual reasoning with explicit tool calls such as cropping and code-based image manipulation, have emerged as a compelling paradigm for reliable and interpretable multimodal reasoning. However, recent studies have revealed that such models often use tools unfaithfully. Many process images are irrelevant to the question (e.g., the tool crops the wrong region or misses the queried target), yet the call still receives full credit and the model still answers correctly. Such decorative or misaligned tool calls waste computation and reveal that the model leans on prior knowledge or the original image rather than the evidence it retrieves. This may stem from two limitations of prevailing methods: the tool reward fails to distinguish useful from useless calls, and tool feedback carries no signal of usefulness. To this end, we introduce FaithEyes, a multi-agent self-judging framework. Concretely, we use a VLM to judge whether each process image helps answer the question. The judgement is injected into the reasoning context as part of the tool observation to help subsequent reasoning, and meanwhile is used to scale the tool reward by the helpful-tool ratio to suppress reward hacking. To keep judgement available at evaluation and thus ensure train-test consistency, we further design a multi-agent framework where the model itself serves as a subagent to judge the tool calls from main agent, eliminating any dependence on an external model at inference. Training via a two-stage SFT + RL pipeline on adapted open-source data, FaithEyes attains competitive or superior accuracy across visual perception and reasoning benchmarks, while markedly improving tool faithfulness. The homepage is at https://github.com/Mosi-AI/FaithEyes.
Chinese Translation
智能视觉-语言模型（VLMs）将文本推理与显式工具调用（如裁剪和基于代码的图像处理）交替使用，已成为可靠且可解释的多模态推理的一个引人注目的范式。然而，最近的研究表明，这些模型往往不忠实地使用工具。许多过程图像与问题无关（例如，工具裁剪了错误的区域或未能捕捉到查询目标），但调用仍然获得了全部信用，模型仍然给出了正确的答案。这种装饰性或不对齐的工具调用浪费了计算资源，并表明模型依赖于先前知识或原始图像，而不是其检索到的证据。这可能源于当前方法的两个局限性：工具奖励未能区分有用与无用的调用，且工具反馈没有有用性的信号。为此，我们提出了FaithEyes，一个多智能体自我判断框架。具体而言，我们使用VLM来判断每个过程图像是否有助于回答问题。该判断作为工具观察的一部分注入推理上下文，以帮助后续推理，同时用于根据有用工具比例来缩放工具奖励，以抑制奖励黑客行为。为了在评估时保持判断的可用性，从而确保训练与测试的一致性，我们进一步设计了一个多智能体框架，其中模型本身作为子代理来判断主代理的工具调用，消除了推理时对外部模型的依赖。通过在适应的开源数据上采用两阶段的SFT + RL训练管道，FaithEyes在视觉感知和推理基准上达到了具有竞争力或优越的准确性，同时显著提高了工具的可信性。主页网址为 https://github.com/Mosi-AI/FaithEyes。
cs.CV / 71 / 2607.28243
EgoGenesis: Egocentric World-Action Modeling with Online Anchored Projective Memory and Action-3D RoPE
自我生成：基于在线锚定投影记忆和动作-3D旋转位置嵌入的自我中心世界-动作建模
Abstract
Egocentric video offers rich manipulation experience for embodied AI, yet collecting diverse egocentric data across scenes, objects, motions, and embodiments remains costly. We present \method, an egocentric world-action simulator that synthesizes controllable, high-quality manipulation videos to expand scarce real-world training data. \method{} builds on a pretrained video generation prior and introduces two geometry-aware conditioning mechanisms. Online Anchored Projective Memory (OAPM) preserves a first-frame 3D scene anchor while periodically refreshing a recent state during autoregressive generation. Action-3D Rotary Position Embedding (A3D-RoPE) encodes end-effector motion with camera-aware 3D rotary coordinates, injecting action geometry into skeleton-to-video cross-attention for precise control. Together, these components improve visual fidelity, geometric stability, and action alignment in long egocentric rollouts. Moreover, augmenting 400 real trajectories with 400 \method-generated trajectories improves out-of-distribution real-robot success from 77\% to 84\% on single-arm tasks and from 53\% to 70\% on dual-arm tasks, demonstrating that the synthesized data substantially improve downstream WAM generalization.
Chinese Translation
自我中心视频为具身人工智能提供了丰富的操控体验，但在不同场景、物体、动作和具身表现中收集多样化的自我中心数据仍然成本高昂。我们提出了 extit{EgoGenesis}，一种自我中心世界-动作模拟器，能够合成可控的高质量操控视频，以扩展稀缺的真实世界训练数据。 extit{EgoGenesis}基于预训练的视频生成模型，并引入了两种几何感知的条件机制。在线锚定投影记忆（Online Anchored Projective Memory, OAPM）在自回归生成过程中保持第一帧的3D场景锚点，同时定期刷新最近状态。动作-3D旋转位置嵌入（Action-3D Rotary Position Embedding, A3D-RoPE）使用相机感知的3D旋转坐标对末端执行器运动进行编码，将动作几何注入骨架到视频的交叉注意力中，以实现精确控制。这些组件共同提高了长时间自我中心生成过程中的视觉保真度、几何稳定性和动作对齐。此外，将400条真实轨迹与400条 extit{EgoGenesis}生成的轨迹进行增强，使得单臂任务的真实机器人成功率从77 ext{%}提高到84 ext{%}，双臂任务的成功率从53 ext{%}提高到70 ext{%}，证明合成数据显著改善了下游世界-动作建模的泛化能力。
cs.CV / 72 / 2607.28247
Space2Ground 2.0: A Multi-Source Dataset and Framework for Agricultural Monitoring through Fusion of Street-Level and Satellite Imagery
Space2Ground 2.0：通过融合街景和卫星影像进行农业监测的多源数据集和框架
Abstract
Accurate and scalable parcel-level agricultural monitoring remains challenging because satellite Earth Observation alone provides only an overhead perspective of agricultural parcels, while optical observations are further affected by cloud-induced temporal gaps. This paper presents Space2Ground 2.0, a multi-source framework integrating Sentinel-1 SAR and Sentinel-2 multispectral time series with geo-tagged street-level imagery acquired using vehicle-mounted cameras and shared through the Mapillary platform. A largely automated processing pipeline performs semantic filtering, image quality assessment, viewpoint-based parcel association, and dataset refinement, transforming large volumes of crowdsourced imagery into parcel-linked, analysis-ready data. Applied over Cyprus during the 2022 growing season, the pipeline produced a curated dataset of 46,050 annotated street-level images, selected from an initial collection exceeding 900,000 images and linked with satellite information for 8,581 agricultural parcels. The practical value of the dataset was assessed through parcel-level crop classification experiments using both single- and multi-source observations. The results demonstrate that street-level imagery provides complementary fine-scale visual information that enhances classification when integrated with satellite time series. Overall, Space2Ground 2.0 provides an openly available benchmark dataset and a reproducible methodology for multimodal agricultural monitoring, with potential applications in visual verification, reduced reliance on costly field inspections, and data-driven agricultural policy implementation.
Chinese Translation
准确且可扩展的地块级农业监测仍然面临挑战，因为单靠卫星地球观测只能提供农业地块的俯视视角，而光学观测又受到云层引起的时间间隔的影响。本文提出了Space2Ground 2.0，这是一个多源框架，整合了Sentinel-1 SAR和Sentinel-2多光谱时间序列，以及通过车载摄像头获取并通过Mapillary平台共享的地理标记街景影像。一个高度自动化的处理流程执行语义过滤、图像质量评估、基于视角的地块关联和数据集精炼，将大量众包影像转化为与地块关联的、可分析的数据。在2022年生长季节应用于塞浦路斯，该流程生成了一个经过策划的数据集，包含46,050张注释的街景影像，这些影像是从初始超过900,000张影像的集合中筛选出来，并与8,581个农业地块的卫星信息相链接。通过使用单源和多源观测进行地块级作物分类实验，评估了数据集的实际价值。结果表明，街景影像提供了补充的细尺度视觉信息，当与卫星时间序列结合时，可以增强分类效果。总体而言，Space2Ground 2.0提供了一个公开可用的基准数据集和可重复的方法论，用于多模态农业监测，具有在视觉验证、减少对昂贵实地检查的依赖以及数据驱动农业政策实施等方面的潜在应用。
cs.CV / 73 / 2607.28261
TARS: Timestep-Aware Data Scaling for 3D-Free Video Re-Shooting
TARS：针对3D无关视频重拍的时间步感知数据缩放
Abstract
Video re-shooting aims to regenerate videos with controllable camera motion and viewpoint. Existing methods rely on explicit 3D priors, which are limited by reconstruction quality and often perform poorly when synthesizing previously unseen regions, or on paired videos with different camera trajectories, whose scarcity hinders generalization. We revisit video re-shooting through text-driven semantic viewpoint specification, enabling control over shot scale, viewing angle, and first-/third-person perspective. To this end, we propose TARS, a 3D-free video re-shooting paradigm. Timestep-wise sensitivity analysis reveals that camera motion is primarily established during high-noise stages, where coarse spatiotemporal structures are formed. Based on this insight, we introduce self-supervised training to learn camera dynamics and fundamental visual representations without paired re-shooting data or 3D reconstruction. Through data scaling and joint textual-camera conditioning, TARS supports robust camera and viewpoint control, plausibly synthesizing regions beyond the source view under large camera motions while enabling reverse-angle re-shooting and perspective switching. Extensive experiments show that TARS provides more accurate and temporally consistent camera control than prior methods. Project Page: https://ymlinfeng.github.io/TARS.github.io/
Chinese Translation
视频重拍旨在以可控的摄像机运动和视角重新生成视频。现有方法依赖于明确的3D先验，这受到重建质量的限制，并且在合成先前未见区域时表现不佳，或者依赖于具有不同摄像机轨迹的配对视频，其稀缺性阻碍了泛化。我们通过文本驱动的语义视角规范重新审视视频重拍，使得对镜头规模、观看角度以及第一人称/第三人称视角的控制成为可能。为此，我们提出了TARS，一种无3D依赖的视频重拍范式。时间步敏感性分析揭示，摄像机运动主要在高噪声阶段建立，此时形成粗略的时空结构。基于这一见解，我们引入自监督训练，以学习摄像机动态和基本视觉表示，而无需配对重拍数据或3D重建。通过数据缩放和联合文本-摄像机条件，TARS支持强大的摄像机和视角控制，合理合成超出源视图的区域，并在大幅摄像机运动下实现反向角度重拍和视角切换。大量实验表明，TARS提供比以往方法更准确和时间一致的摄像机控制。项目页面：https://ymlinfeng.github.io/TARS.github.io/
cs.CV / 74 / 2607.28269
Theia: Large-Scale Multimodal Captioning and Automated Validation of the Incidents1M Dataset for Data-Free Distillation
Theia：用于无数据蒸馏的Incidents1M数据集的大规模多模态标注与自动验证
Abstract
The deployment of Vision-Language Models (VLMs) in critical domains like disaster management requires high-quality multimodal datasets, especially for transferring knowledge via Data-Free Knowledge Distillation (DFKD). However, existing datasets in this domain either entirely lack descriptive text, such as Incidents1M, or suffer from severe text-image semantic misalignment, such as CrisisMMD. In this work, we present a novel methodology to construct and automatically validate a large-scale multimodal dataset for disaster response. Starting from the vision-only Incidents1M, we successfully recovered 100,000 images and generated high-fidelity textual descriptions using two distinct Qwen3.5 architectures: a 4B dense model and a 35B Mixture-of-Experts (MoE) model. To ensure the generated captions provide reliable semantic anchoring for DFKD, we introduce an image-blind LLM-as-a-Judge validation pipeline leveraging Qwen3.5-9B. By intentionally obscuring the original image from the judge, this evaluator accurately simulates the modality gap of the student model during data-free distillation. Our evaluation across 173,179 label pairs demonstrates a high semantic agreement (78.65/100) between the two architectures. Furthermore, the automated evaluation reveals a conservative captioning behaviour, characterized by a high Precision (77.6%) and low Recall (46.0%). This minimizes the false positive noise, while simultaneously exposing underlying human annotation inconsistencies in the original ground truth. This work provides a scalable, LLM-validated multimodal dataset and a reproducible framework to advance cross-modal knowledge distillation.
Chinese Translation
在灾难管理等关键领域中部署视觉-语言模型（VLMs）需要高质量的多模态数据集，特别是在通过无数据知识蒸馏（DFKD）转移知识时。然而，现有的数据集要么完全缺乏描述性文本，例如Incidents1M，要么遭受严重的文本-图像语义不对齐，例如CrisisMMD。在本研究中，我们提出了一种新颖的方法来构建和自动验证大规模的灾难响应多模态数据集。我们从仅包含视觉信息的Incidents1M出发，成功恢复了100,000张图像，并使用两种不同的Qwen3.5架构生成了高保真的文本描述：一种是4B密集模型，另一种是35B专家混合（MoE）模型。为了确保生成的标题为DFKD提供可靠的语义锚定，我们引入了一种基于图像盲评的LLM-as-a-Judge验证流程，利用Qwen3.5-9B。通过故意遮蔽原始图像，这一评估者准确模拟了学生模型在无数据蒸馏过程中的模态差距。我们在173,179对标签上的评估表明，两种架构之间具有较高的语义一致性（78.65/100）。此外，自动评估显示出保守的标注行为，表现为高精确率（77.6%）和低召回率（46.0%）。这最小化了假阳性噪声，同时暴露了原始真实标签中潜在的人类标注不一致性。本研究提供了一个可扩展的、经过LLM验证的多模态数据集和一个可重复的框架，以推动跨模态知识蒸馏的发展。
cs.CV / 75 / 2607.28277
MSCM-net: A hyperspectral image classiffcation method based on multi-scale convolution and Mamba
MSCM-net：一种基于多尺度卷积和Mamba的高光谱图像分类方法
Abstract
Hyperspectral imaging is widely used in remote sensing and engineering. Therefore, research on its classification methods is crucial. While CNN and Transformer-based methods have advanced, they still face locality constraints and high computational complexity. To address these issues, we propose an innovative hyperspectral image classification model, MSCM-net. Specifically, first of all, a model architecture combining multi-scale CNN and Mamba is proposed. It consists of a multi-scale feature extraction module (MCSE) and multiple stacked Mamba blocks, which integrates the local feature extraction capability of multi-scale CNN and the long sequence modeling advantage of Mamba. Secondly, the proposed MCSE module consists of multi-scale convolution and SENet. Convolution kernels of different scales extract local information with different receptive fields, enhancing the fusion of spatial and spectral information. Meanwhile, the SENet enables the model to automatically learn the importance of each channel in the multi-scale features. Furthermore, we also propose a dual-branch feature aggregation module, which further effectively extracts and integrates the spectral information contained in the central pixel and the spatial information in the surrounding pixels. Our model has undergone numerous experiments on three widely used benchmark datasets. The experimental results show that MSCM-net can achieve advanced classification performance while reducing computational complexity.
Chinese Translation
高光谱成像在遥感和工程中得到广泛应用。因此，对其分类方法的研究至关重要。尽管基于卷积神经网络（CNN）和Transformer的方法取得了进展，但仍面临局部性限制和高计算复杂度的问题。为了解决这些问题，我们提出了一种创新的高光谱图像分类模型MSCM-net。具体而言，首先提出了一种结合多尺度CNN和Mamba的模型架构。该架构由一个多尺度特征提取模块（MCSE）和多个堆叠的Mamba块组成，融合了多尺度CNN的局部特征提取能力和Mamba的长序列建模优势。其次，所提出的MCSE模块由多尺度卷积和SENet组成。不同尺度的卷积核以不同的感受野提取局部信息，增强了空间信息和光谱信息的融合。同时，SENet使模型能够自动学习多尺度特征中每个通道的重要性。此外，我们还提出了一个双分支特征聚合模块，进一步有效地提取和整合中心像素中包含的光谱信息以及周围像素中的空间信息。我们的模型在三个广泛使用的基准数据集上进行了大量实验。实验结果表明，MSCM-net能够在降低计算复杂度的同时实现先进的分类性能。
cs.CV / 76 / 2607.28285
Beyond Visual Ambiguity: Guiding Robust Monocular Depth Estimation in Challenging Scenarios via Detailed Long Captions
超越视觉模糊：通过详细的长标题指导在挑战场景中的稳健单目深度估计
Abstract
Monocular depth estimation (MDE) faces challenges with non-Lambertian surfaces and adverse weather conditions due to the visual ambiguities inherent in single-image limited information. Existing works address them in isolation via image inpainting or augmentation, yielding limited robustness gains. Language, as a powerful complementary modality to vision, is demonstrated to enhance the visual perception capabilities of vision-language models (VLMs) via detailed long captions. However, prior language-integrated MDE methods fail to fully harness this potential due to short text input with limited information, coarse global text feature learning, and limited language guidance during depth decoding. To address these limitations, we propose CapDepth, a novel framework for robust MDE that leverages guidance from detailed long captions to alleviate visual ambiguities in both challenging scenarios. First, we design a detailed long caption input template that explicitly conveys rich spatial relationships among multiple atom sentences. Second, a dynamic caption encoder is introduced to extract fine-grained depth-relevant text features via progressive masked attention. Finally, we propose a text-adaptive decoder that guides enhanced depth decoding with text features via stable adaptive layer normalization. Extensive experiments validate the efficacy of CapDepth, which outperforms state-of-the-art methods, achieving depth error reductions of 25.0% on non-Lambertian surfaces and 22.0% under adverse weather conditions.
Chinese Translation
单目深度估计（MDE）在非朗伯表面和恶劣天气条件下面临挑战，因为单幅图像所固有的视觉模糊性导致信息有限。现有研究通过图像修复或增强单独解决这些问题，带来的稳健性提升有限。语言作为一种强大的视觉补充模态，已被证明能够通过详细的长标题增强视觉-语言模型（VLMs）的视觉感知能力。然而，以前的语言集成MDE方法未能充分利用这一潜力，原因在于输入文本短小且信息有限、全局文本特征学习粗糙，以及在深度解码过程中语言指导有限。为了解决这些局限性，我们提出了CapDepth，一个新颖的稳健MDE框架，利用详细长标题的指导来减轻挑战场景中的视觉模糊性。首先，我们设计了一个详细的长标题输入模板，明确传达多个原子句子之间丰富的空间关系。其次，引入了一个动态标题编码器，通过渐进式掩蔽注意力提取细粒度的与深度相关的文本特征。最后，我们提出了一种文本自适应解码器，通过稳定的自适应层归一化指导增强的深度解码。大量实验验证了CapDepth的有效性，其在非朗伯表面上实现了25.0%的深度误差减少，在恶劣天气条件下实现了22.0%的深度误差减少，超越了最先进的方法。
cs.CV / 77 / 2607.28293
Filling the Pareto-Optimal Front for Affordance Segmentation on Embedded Devices Using RGB-D Cameras
利用RGB-D相机在嵌入式设备上填充可供性分割的帕累托最优前沿
Abstract
While depth sensors have the potential to complement RGB data for affordance segmentation in wearable robots, their usage seems to remain underexplored. The paper proposes two approaches: a reformulated version of hardware-aware neural architecture search, endowed with a newly designed search space to integrate depth (D) information into small-sized deep networks, and a dedicated fine-tuning approach, including a preprocessing layer to merge depth information with RGB data and make it compatible with conventional architectures. In both cases, those methods aim to generate solutions that benefit from modern (portable) hardware accelerators and overcome existing tiny-like approaches, which often fail to tackle critical scenarios due to the severe constraints set by the supporting hardware. Extensive experiments on a pair of real-world datasets demonstrate the effectiveness of the proposed method as compared with existing solutions. The approach presented in the paper generates, in most cases, solutions that identify the Pareto optimal front to balance generalization performance and hardware requirements. The paper also describes the supporting prototype, including a Jetson Nano board and a RealSense RGB-D camera. When considering the energy profile of the device, the overall system can attain real-time performances within an energy budget that is compatible with standard batteries, such as those used in smartphones.
Chinese Translation
尽管深度传感器有潜力补充可供性分割中RGB数据在可穿戴机器人中的应用，但其使用似乎仍未得到充分探索。本文提出了两种方法：一种是重新构建的硬件感知神经架构搜索版本，配备了新设计的搜索空间，以将深度（D）信息集成到小型深度网络中；另一种是专门的微调方法，包括一个预处理层，以将深度信息与RGB数据合并，并使其与传统架构兼容。在这两种情况下，这些方法旨在生成受益于现代（便携式）硬件加速器的解决方案，并克服现有的微型方法，这些方法往往由于支持硬件的严格限制而无法应对关键场景。在一对真实世界数据集上的广泛实验表明，所提出的方法相较于现有解决方案具有有效性。本文所提出的方法在大多数情况下生成的解决方案能够识别出帕累托最优前沿，以平衡泛化性能和硬件要求。本文还描述了支持原型，包括Jetson Nano板和RealSense RGB-D相机。在考虑设备的能耗特征时，整个系统能够在与智能手机等标准电池兼容的能量预算内实现实时性能。
cs.CV / 78 / 2607.28300
MonoVoc: Decoupling Geometry and Semantics for Lightweight Monocular Open-Vocabulary 3D Gaussians
MonoVoc：解耦几何与语义以实现轻量级单目开放词汇3D高斯
Abstract
Open vocabulary 3D scene understanding is essential for next-generation interactive systems, empowering users to intuitively query and navigate reconstructed environments using natural language. However, current 3D Gaussian frameworks are often bottlenecked by restrictive multiview capture requirements, costly scene-specific optimization, and the massive memory overhead of storing dense language features. We present a novel, training-free pipeline that fundamentally reimagines this paradigm by explicitly decoupling 3D geometric reconstruction from semantic integration. Given a standard monocular video sequence as input, our method efficiently outputs a compact, highly interpretable, and fully searchable object-level semantic Gaussian map. Rather than entangling heavy language embeddings within the mapping loop, we extract geometry independently and ground semantics through a lightweight, modular post-processing framework. Extensive evaluations on the Replica dataset demonstrate that this decoupled architecture preserves strong rendering fidelity and competitive segmentation accuracy. Crucially, by replacing dense per-Gaussian storage with modular, object-level semantic embeddings, our approach delivers an order-of-magnitude reduction in memory usage compared to SOTA baselines. This provides a highly efficient, scalable, and practical solution for open-vocabulary 3D retrieval and question answering directly from everyday monocular video.
Chinese Translation
开放词汇的3D场景理解对于下一代交互系统至关重要，使用户能够通过自然语言直观地查询和导航重建的环境。然而，当前的3D高斯框架常常受到限制性多视角捕捉要求、昂贵的场景特定优化以及存储密集语言特征所需的巨大内存开销的瓶颈。我们提出了一种新颖的、无训练的流程，从根本上重新构想了这一范式，通过明确地将3D几何重建与语义集成解耦。给定标准的单目视频序列作为输入，我们的方法高效地输出一个紧凑、高度可解释且完全可搜索的对象级语义高斯图。我们并不在映射循环中纠缠重型语言嵌入，而是独立提取几何，并通过轻量级的模块化后处理框架将语义进行基础化。在Replica数据集上的广泛评估表明，这种解耦架构保持了强大的渲染保真度和具有竞争力的分割精度。重要的是，通过用模块化的对象级语义嵌入替代密集的每个高斯存储，我们的方法在内存使用上相比于最先进的基线实现了数量级的减少。这为直接从日常单目视频中进行开放词汇3D检索和问答提供了一个高效、可扩展且实用的解决方案。
cs.CV / 79 / 2607.28312
ObjectStream: Latent Objects as Memory Anchors for Streaming Video Understanding
ObjectStream：将潜在对象作为流媒体视频理解的记忆锚点
Abstract
Streaming video understanding requires models to continuously retain useful visual evidence before future questions are known. Existing approaches primarily manage the growing visual context according to token importance, temporal redundancy, or segment-level relevance, but rarely organize evidence around objects that persist and evolve over time. Thus, in this paper, we introduce ObjectStream, a training-free framework that treats latent objects as memory anchors for streaming video understanding. ObjectStream induces spatially coherent latent objects directly from frozen Video-LLM representations, links them across frames into persistent anchors, and maintains their histories under a bounded memory budget, without requiring external object detectors or segmentation models. Built on these anchors, ObjectStream preserves three complementary forms of evidence: persistent object histories, transient object changes, and recent visual context. This design enables existing Video Large Language Models (Video-LLMs) to reason over object identities, interactions, and state changes while leaving the underlying model unchanged. Extensive experiments on online streaming and offline long-video benchmarks demonstrate both effectiveness and efficiency. In online streaming evaluation, ObjectStream improves Qwen2.5-VL-7B by 10.0 points on OVO-Bench Real-Time Visual Perception, while reducing peak GPU mem-ory and TTFT by approximately 50%. On offline long-video benchmarks, it surpasses the full-token baseline while discarding 82.5% of visual tokens. These results highlight latent objects as a practical and effective organizing principle for compact streaming video memory.
Chinese Translation
流媒体视频理解要求模型在未来问题尚未明确之前，持续保留有用的视觉证据。现有方法主要根据标记重要性、时间冗余或片段级相关性来管理不断增长的视觉上下文，但很少围绕随时间持续和演变的对象组织证据。因此，在本文中，我们提出了ObjectStream，一个无训练的框架，将潜在对象视为流媒体视频理解的记忆锚点。ObjectStream直接从冻结的Video-LLM表示中诱导空间一致的潜在对象，将它们跨帧链接为持久锚点，并在有限的内存预算下维护它们的历史，而无需外部对象检测器或分割模型。在这些锚点的基础上，ObjectStream保留三种互补形式的证据：持久对象历史、瞬态对象变化和最近的视觉上下文。这一设计使得现有的视频大型语言模型（Video-LLMs）能够推理对象身份、交互和状态变化，同时保持底层模型不变。在在线流媒体和离线长视频基准上的大量实验展示了其有效性和效率。在在线流媒体评估中，ObjectStream在OVO-Bench实时视觉感知上将Qwen2.5-VL-7B的得分提高了10.0分，同时将峰值GPU内存和TTFT减少了约50%。在离线长视频基准上，它超越了全标记基线，同时丢弃了82.5%的视觉标记。这些结果突显了潜在对象作为紧凑流媒体视频记忆的实用和有效的组织原则。
cs.CV / 80 / 2607.28320
AdaAnchor4D: Anchor-Conditioned Spatiotemporal Feature Aggregation for Monocular UAV 4D Reconstruction
AdaAnchor4D：用于单目无人机四维重建的锚条件时空特征聚合
Abstract
Monocular UAV videos provide valuable observations for dynamic reconstruction of complex urban scenes. However, such scenes exhibit pronounced spatiotemporal heterogeneity: different regions follow distinct temporal activity patterns, while the motion states of some dynamic regions may further evolve over time. Although dynamic Gaussian methods based on decomposed shared spatiotemporal feature fields have achieved efficient and accurate reconstruction in object-centric or relatively compact scenes, their commonly adopted fixed plane-wise feature combination mechanisms are less suited to the heterogeneous local dynamics of UAV scenes, often leading to ghosting artifacts and blurred dynamic details. To address this challenge, we propose AdaAnchor4D, an adaptive anchor deformation framework for monocular UAV dynamic scene reconstruction. At its core, Anchor-Conditioned Feature Aggregation (ACFA) adaptively aggregates shared spatiotemporal features using anchor-specific aggregation embeddings and temporal information, allowing different local units to obtain dynamic representations tailored to their local and temporal states. Decoupled Local Geometry Deformation (DLGD) separates anchor-state deformation from local Gaussian geometry deformation, while Density-Adaptive Coordinate Warping (DACW) reparameterizes feature-query coordinates according to the axis-wise anchor distributions, alleviating the mismatch between non-uniform geometric sampling and uniform grid parameterization. Experiments on UAV-Arc4D, VisDrone, and UAVDT show that AdaAnchor4D achieves higher rendering quality than representative dynamic Gaussian methods while maintaining real-time rendering performance. The code will be made publicly available.
Chinese Translation
单目无人机视频为复杂城市场景的动态重建提供了宝贵的观察数据。然而，这些场景表现出明显的时空异质性：不同区域遵循不同的时间活动模式，而某些动态区域的运动状态可能随时间进一步演变。尽管基于分解共享时空特征场的动态高斯方法在以物体为中心或相对紧凑的场景中实现了高效和准确的重建，但它们常用的固定平面特征组合机制不太适合无人机场景的异质局部动态，常常导致鬼影伪影和动态细节模糊。为了解决这一挑战，我们提出了AdaAnchor4D，一个用于单目无人机动态场景重建的自适应锚变形框架。其核心是锚条件特征聚合（Anchor-Conditioned Feature Aggregation, ACFA），它利用特定于锚的聚合嵌入和时间信息自适应地聚合共享的时空特征，使不同的局部单元能够获得针对其局部和时间状态量身定制的动态表示。解耦局部几何变形（Decoupled Local Geometry Deformation, DLGD）将锚状态变形与局部高斯几何变形分离，而密度自适应坐标扭曲（Density-Adaptive Coordinate Warping, DACW）根据轴向锚分布重新参数化特征查询坐标，缓解了非均匀几何采样与均匀网格参数化之间的不匹配。在UAV-Arc4D、VisDrone和UAVDT上的实验表明，AdaAnchor4D在保持实时渲染性能的同时，达到了比代表性动态高斯方法更高的渲染质量。代码将公开发布。
cs.CV / 81 / 2607.28327
Same Branches, Different Trees: A Bifurcation Connectedness Metric for Coronary Artery Segmentation and FFR-CT Decision Agreement
相同的分支，不同的树：用于冠状动脉分割和 FFR-CT 决策一致性的分岔连通性度量
Abstract
Fractional flow reserve derived from CT angiography (FFR-CT) simulates flow through a patient-specific vessel model, so its accuracy depends on the connectedness of the segmented tree, not only on volumetric overlap: a segmentation can reach high Dice yet sever a bifurcation, dropping the downstream subtree and reversing the treatment decision. Topology-aware losses such as clDice and Skeleton Recall act on the global centreline and can miss localised breaks. We study the Bifurcation Connectedness Score (BCS), which scores connectedness at each ground-truth bifurcation, and soft-BCS, its differentiable training surrogate. BCS captures a property of segmentation quality the standard metrics miss: it responds strongly to breaks in connectedness while staying largely unchanged under connectedness-preserving narrowing. Higher BCS accompanies closer agreement between the FFR-CT decisions a solver makes on predicted versus ground-truth geometry, most clearly in severe disease (OR 2.16, CI [1.23, 4.18]). Both decisions come from the same solver, so this reflects geometric, not clinical, fidelity. In training, soft-BCS and Skeleton Recall recover the same branches but build different trees. Recovering branches and keeping them connected are separable properties, so we recommend reporting a measure of each.
Chinese Translation
基于 CT 血管造影的分数流量储备（FFR-CT）模拟患者特定血管模型中的流动，因此其准确性不仅依赖于分割树的体积重叠，还依赖于分割树的连通性：一种分割可以达到高的 Dice 系数，但却切断了一个分岔，导致下游子树丢失，从而改变治疗决策。关注拓扑的损失函数，如 clDice 和 Skeleton Recall，作用于全局中心线，可能会忽略局部断裂。我们研究了分岔连通性评分（Bifurcation Connectedness Score, BCS），该评分在每个真实分岔处评估连通性，以及其可微分的训练替代品 soft-BCS。BCS 捕捉了标准度量遗漏的分割质量特性：它对连通性断裂的反应强烈，而在保持连通性的收缩下变化不大。更高的 BCS 与 FFR-CT 决策在预测几何与真实几何之间的更紧密一致性相关，尤其在严重疾病中最为明显（OR 2.16, CI [1.23, 4.18]）。这两种决策来自同一个求解器，因此反映的是几何而非临床的忠实度。在训练中，soft-BCS 和 Skeleton Recall 恢复相同的分支，但构建了不同的树。恢复分支和保持其连通性是可分的属性，因此我们建议报告每个属性的度量。
cs.CV / 82 / 2607.28341
Capturing Token Tendencies for Training-Free Token Pruning in Multimodal Large Language Models
捕捉令牌趋势以实现多模态大语言模型中的无训练令牌剪枝
Abstract
While visual token pruning is essential for efficient Multimodal Large Language Models (MLLMs), existing training-free methods suffer from a critical limitation: they rely on static, instantaneous heuristics to perform irreversible filtering. This approach ignores the hierarchical nature of MLLMs, where token importance often evolves dynamically rather than remaining fixed across layers. Consequently, tokens essential for deep-layer reasoning are often prematurely discarded by shallow-layer estimates. To address this, we propose Trend-aware Pruning, a novel framework that elevates pruning from a local snapshot decision to a temporal trajectory modeling problem. Instead of relying on isolated scores, our method captures the momentum of attention flow. This enables a dynamic rectification mechanism that selectively reactivates "late-blooming" tokens, those initially undervalued but exhibiting rising semantic importance, thereby preventing the loss of critical visual cues. Extensive experiments demonstrate that our approach achieves a superior efficiency-performance trade-off across diverse multimodal tasks. Notably, it reduces visual tokens by over 77.8%, retaining only approximately 23 tokens in the final layer while maintaining competitive performance, offering a robust and reversible solution for high-efficiency multimodal inference.
Chinese Translation
尽管视觉令牌剪枝对于高效的多模态大语言模型（MLLMs）至关重要，但现有的无训练方法存在一个关键限制：它们依赖于静态的瞬时启发式方法进行不可逆的过滤。这种方法忽视了MLLMs的层次特性，其中令牌的重要性往往是动态演变的，而不是在各层之间保持固定。因此，深层推理所需的令牌常常被浅层估计过早丢弃。为了解决这一问题，我们提出了趋势感知剪枝（Trend-aware Pruning），这是一个将剪枝从局部快照决策提升为时间轨迹建模问题的新框架。我们的方法不再依赖孤立的评分，而是捕捉注意力流的动量。这使得我们能够实现一个动态修正机制，选择性地重新激活那些“晚开花”的令牌，即那些最初被低估但表现出上升的语义重要性的令牌，从而防止关键视觉线索的丢失。大量实验表明，我们的方法在多种多模态任务中实现了优越的效率-性能权衡。值得注意的是，它将视觉令牌减少了超过77.8%，在最终层仅保留约23个令牌，同时保持竞争力的性能，提供了一种高效多模态推理的稳健且可逆的解决方案。
cs.CV / 83 / 2607.28362
ShadowDancer: Teaching Video World Models Any Action by Learning Unified Dynamics Representations from a Video and Its Shadow
ShadowDancer：通过学习视频及其阴影的统一动态表示，教授视频世界模型任何动作
Abstract
We present ShadowDancer, a novel approach to any-action, frame-level control of interactive video world models. The obstacle is representational: existing interfaces either encode an action loosely, leaving how it unfolds for the model to improvise, or encode it exactly through structured signals that serve one family and are hard to acquire, so precise control across diverse dynamics remains impractical. Demonstration videos are the natural remedy, specifying any dynamics frame by frame; yet a video shows its dynamics only through one particular appearance, a single shadow of the underlying dynamics, so actions learned from demonstrations transfer poorly to new scenes. ShadowDancer addresses this with two key innovations: (1) shadow pairs, video pairs that replay the same dynamics under independently resampled appearance, constructed at scale by our Shadow Library, so that a dynamics family becomes controllable exactly when such pairs can be constructed for it; and (2) cross-shadow prediction, which learns actions by predicting one shadow from the other, so that whatever the pairing resamples is discarded by construction and whatever it preserves becomes the action, yielding a unified dynamics representation that drives a block-causal world model. Any demonstrated clip thus becomes a reusable action asset, replayed in new environments without action labels, motion estimators, or fine-tuning. Experiments demonstrate improved action transfer and long action rollout over strong latent-action and interactive world model baselines across diverse dynamics families, with an average blinded win rate of 86% in rollout comparisons. We show video results at https://ShadowDancer-1.github.io
Chinese Translation
我们提出了ShadowDancer，这是一种新颖的方法，用于对交互式视频世界模型进行任何动作的逐帧控制。面临的障碍是表示性：现有接口要么松散地编码一个动作，让模型自行发挥其展开方式，要么通过结构化信号精确编码该动作，这些信号仅适用于某一类动作且难以获取，因此在多样化动态下实现精确控制仍然不切实际。演示视频是自然的解决方案，逐帧指定任何动态；然而，视频仅通过一种特定的外观展示其动态，即潜在动态的单一阴影，因此从演示中学习的动作在新场景中的迁移效果较差。ShadowDancer通过两个关键创新来解决这一问题：（1）阴影对，即在独立重新采样外观下重播相同动态的视频对，由我们的阴影库大规模构建，因此当可以为某个动态家族构建这样的对时，该动态家族就变得可控；（2）交叉阴影预测，通过从一个阴影预测另一个阴影来学习动作，因此无论配对如何重新采样，构造上都会被丢弃，而保留的部分则成为动作，从而产生驱动块因果世界模型的统一动态表示。因此，任何演示片段都可以成为可重用的动作资产，在新环境中重播，而无需动作标签、运动估计器或微调。实验表明，在多样化动态家族中，相较于强大的潜在动作和交互式世界模型基线，动作迁移和长动作展开得到了改善，回合比较中平均盲测胜率达到86%。我们在 https://ShadowDancer-1.github.io 展示了视频结果。
cs.CV / 84 / 2607.28386
Explaining Image Similarity with Automatically Extracted Concept Activation Vectors
通过自动提取的概念激活向量解释图像相似性
Abstract
Image similarity underlies many computer vision applications, yet it is often unclear why two images receive a high or low similarity score. Existing explainability methods often rely on gradient-based attribution maps to provide local justifications for similarity. These approaches struggle to provide global insights into what specifically drives similarity in regions of an embedding space, such as texture, shape, or color. We introduce a model- and metric-agnostic framework that explains image similarity using Concept Activation Vectors (CAVs) extracted automatically via Sparse Autoencoders (SAEs). Given a pair of images, we perturb their embeddings along discovered concept directions and measure the resulting change in a chosen similarity function, yielding concept importances. For image pairs, we provide localization with concept attribution maps. We extend this procedure to group-level settings, explaining what drives similarity across a cluster of images rather than a single pair, and further, we introduce Exemplar Retrieval, aiming to recover samples with similar reasons contributing to similarity. Our experiments show that our latent perturbations are more faithful to the underlying data distribution than pixel-space baselines, and that concept importances linearly recover the true similarity score. Qualitative results further confirm the usefulness of our methods in understanding a model's individual and group similarity judgments.
Chinese Translation
图像相似性是许多计算机视觉应用的基础，但通常不清楚为什么两幅图像会获得高或低的相似性评分。现有的可解释性方法通常依赖于基于梯度的归因图来提供相似性的局部解释。这些方法在提供嵌入空间中具体驱动相似性的全局见解方面存在困难，例如纹理、形状或颜色。我们提出了一种模型和度量无关的框架，通过使用通过稀疏自编码器（Sparse Autoencoders, SAEs）自动提取的概念激活向量（Concept Activation Vectors, CAVs）来解释图像相似性。给定一对图像，我们沿着发现的概念方向扰动它们的嵌入，并测量所选相似性函数中产生的变化，从而得出概念重要性。对于图像对，我们提供了带有概念归因图的定位。我们将此过程扩展到群体级设置，解释驱动一组图像之间相似性的因素，而不仅仅是单对图像。此外，我们引入了示例检索（Exemplar Retrieval），旨在恢复具有相似原因的样本，从而促进相似性。我们的实验表明，我们的潜在扰动比像素空间基线更忠实于基础数据分布，并且概念重要性线性恢复真实的相似性评分。定性结果进一步确认了我们的方法在理解模型的个体和群体相似性判断方面的有效性。
cs.CV / 85 / 2607.28394
Hand-Object Interaction in the Age of Large Foundation Models:Reconstruction, Generation, and Embodied Transfer
大型基础模型时代的手-物体交互：重建、生成与具身转移
Abstract
Hand-object interaction (HOI) modeling remains challenging because it requires joint reasoning about hand articulation, object geometry, contact, semantics, and dynamics under severe visual uncertainty. Foundation models introduce transferable prior knowledge learned from large-scale cross-domain data, offering new ways to address these challenges beyond task-specific data and models. However, the rapidly growing literature remains fragmented, and existing studies typically describe these methods simply as ``using large models'' without systematically characterizing what knowledge is introduced, where it enters the HOI pipeline, or which HOI uncertainty it helps reduce. This survey presents the first systematic review of foundation-model priors for HOI. We organize the literature into six HOI tasks spanning reconstruction and generation. More importantly, we establish a taxonomy of eight foundation-model sub-priors grouped into geometric, semantic, and visual families. Geometric priors encompass shape retrieval, shape reconstruction, and spatial reconstruction; semantic priors include semantic grounding and language reasoning; and visual priors cover visual representation, image generation, and video generation. Based on this taxonomy, we systematically analyze how different priors are represented, injected, and adapted across HOI pipelines and tasks. Beyond how foundation models empower HOI, we further examine how HOI-derived knowledge is used in robot learning, including human-data pretraining, human-to-robot skill transfer, and HOI-to-robot data generation. Finally, we summarize datasets and evaluation protocols, and discuss limitations and future directions toward more generalizable HOI systems. To support long-term progress, we curate a live repository that continuously aggregates emerging methods and benchmarks.
Chinese Translation
手-物体交互（HOI）建模仍然面临挑战，因为它需要对手部关节、物体几何、接触、语义和动态进行联合推理，而这些在严重的视觉不确定性下更为复杂。基础模型引入了从大规模跨领域数据中学习到的可转移先验知识，为解决这些挑战提供了新的方法，超越了特定任务的数据和模型。然而，快速增长的文献仍然是碎片化的，现有研究通常仅将这些方法描述为“使用大型模型”，而没有系统性地表征引入了什么知识、这些知识在HOI流程中的何处介入，或是帮助减少了哪些HOI不确定性。本调查提供了HOI基础模型先验的首个系统性综述。我们将文献组织为六个HOI任务，涵盖重建和生成。更重要的是，我们建立了一个八个基础模型子先验的分类法，这些子先验分为几何、语义和视觉三大类。几何先验包括形状检索、形状重建和空间重建；语义先验包括语义基础和语言推理；视觉先验涵盖视觉表示、图像生成和视频生成。基于这一分类法，我们系统地分析了不同先验在HOI流程和任务中的表现、注入和适应方式。除了基础模型如何增强HOI外，我们还进一步探讨了HOI衍生知识在机器人学习中的应用，包括人类数据预训练、人类到机器人技能转移，以及HOI到机器人数据生成。最后，我们总结了数据集和评估协议，并讨论了限制和未来朝着更具普遍性HOI系统的方向。为了支持长期进展，我们策划了一个实时库，持续汇聚新兴方法和基准。
cs.CV / 86 / 2607.28401
Large scale cross-regional remote sensing flood monitoring framework for operative mapping and impact analysis
大规模跨区域遥感洪水监测框架用于操作性制图和影响分析
Abstract
Effective flood monitoring is critical for minimizing the impacts of flood disasters on populations and infrastructure. Yet reliable remote sensing across extensive and environmentally diverse regions remains challenging, as most segmentation algorithms lack the generalisation capacity required for large-scale application, while annotated flood data are scarce and unevenly distributed. This study presents an end-to-end multimodal framework for Russian Federation territories sustainable flood monitoring and damage assessment based on synthetic aperture radar data, multispectral imagery, and digital elevation models with their derivatives, forming a 21-channel input. Using a self-collected multimodal dataset covering seven Russian regions, two strategies for water surface detection under limited data conditions were compared: a supervised U-Net++ model and the self-supervised AnySat architecture pre-trained and fine-tuned for the segmentation task. Under the data conditions of this study, supervised learning proved more effective, while the AnySat-based approach offered greater stability and retains advantages for settings where larger unlabelled data or missing modalities at inference are expected. The best flood area predictions were used to estimate flood impact in urban areas in terms of the area affected, material damage, casualties, and ecological and agricultural impact. The estimations were conducted following the official methodology of the Russian Ministry of Emergency Situations. Applied to the 2019 Tulun flood, the obtained results closely matched official assessments, except for material damage, due to the open-source databases usage. The results demonstrate the potential of deep learning and multimodal satellite data integration for scalable, reliable flood monitoring across diverse environmental and data-limited conditions.
Chinese Translation
有效的洪水监测对于减少洪水灾害对人口和基础设施的影响至关重要。然而，在广泛且环境多样的区域内进行可靠的遥感监测仍然具有挑战性，因为大多数分割算法缺乏大规模应用所需的泛化能力，而带注释的洪水数据则稀缺且分布不均。本研究提出了一种基于合成孔径雷达数据、多光谱影像和数字高程模型及其衍生物的端到端多模态框架，用于俄罗斯联邦地区的可持续洪水监测和损害评估，形成了21通道输入。利用自收集的覆盖七个俄罗斯地区的多模态数据集，比较了两种在有限数据条件下的水面检测策略：一种是监督学习的U-Net++模型，另一种是自监督的AnySat架构，该架构经过预训练并针对分割任务进行了微调。在本研究的数据条件下，监督学习被证明更为有效，而基于AnySat的方法则提供了更大的稳定性，并在预期存在更大未标记数据或推理时缺失模态的情况下保留了优势。最佳洪水区域预测用于估算城市地区的洪水影响，包括受影响面积、物质损失、人员伤亡以及生态和农业影响。估算遵循俄罗斯紧急情况部的官方方法。应用于2019年图伦洪水，获得的结果与官方评估非常接近，除了物质损失外，这是由于使用了开放源数据库。结果表明，深度学习和多模态卫星数据集成在多样环境和数据有限条件下进行可扩展、可靠的洪水监测的潜力。
cs.CV / 87 / 2607.28423
Negative controls reveal volume-driven confounding in radiomics and imaging foundation model features
负对照揭示放射组学和影像基础模型特征中的体积驱动混杂
Abstract
Radiomics and imaging foundation models promise non-invasive biomarkers of tumour biology, yet predictive signatures may reflect tumour volume or acquisition artifacts rather than meaningful image structure. We introduce READII-2-ROQC, an open-source framework that uses volume-preserving negative controls to assess whether radiomic and deep imaging features capture independent spatial signals. READII-2-ROQC generates voxel-perturbed images across tumour, background and whole-image regions using configurable randomization strategies, then compares feature behaviour and model performance between original and control images. Applied to three public cancer imaging cohorts, the framework processed 3,552 tumour volumes and extracted PyRadiomics and foundation-model features from original images and nine matched controls. Reproducing published survival and HPV-status signatures, we show that multiple models retain performance after spatial structure is destroyed, revealing volume-driven or contextual confounding, whereas others show perturbation-sensitive signal. READII-2-ROQC provides a scalable quality-control strategy for developing interpretable, biologically grounded imaging biomarkers and reproducible radiomics workflows.
Chinese Translation
放射组学和影像基础模型承诺提供肿瘤生物学的非侵入性生物标志物，但预测特征可能反映肿瘤体积或采集伪影，而非有意义的图像结构。我们引入了READII-2-ROQC，这是一个开源框架，利用保持体积的负对照来评估放射组学和深度影像特征是否捕捉到独立的空间信号。READII-2-ROQC通过可配置的随机化策略生成肿瘤、背景和整个图像区域的体素扰动图像，然后比较原始图像与对照图像之间的特征行为和模型性能。该框架应用于三个公共癌症影像队列，处理了3,552个肿瘤体积，并从原始图像和九个匹配对照中提取了PyRadiomics和基础模型特征。通过重现已发表的生存和HPV状态特征，我们展示了多个模型在空间结构被破坏后仍能保持性能，揭示了体积驱动或上下文混杂，而其他模型则显示出对扰动敏感的信号。READII-2-ROQC为开发可解释的、生物学基础的影像生物标志物和可重复的放射组学工作流程提供了一种可扩展的质量控制策略。
cs.CV / 88 / 2607.28442
ViewMind3D: Modular View-Aware Inference for Training-Free 3D-QA
ViewMind3D：模块化视图感知推理的无训练3D问答
Abstract
Recent advances in large language models (LLMs) and vision-language models (VLMs) have enabled new possibilities for 3D question answering (3D-QA), a key capability for embodied AI and robotic perception. However, most existing methods rely on 3D-specific training or fine-tuning with costly annotations, limiting their scalability and real-world applicability. We present \textbf{ViewMind3D}, a fully training-free and modular framework for 3D spatial reasoning over multi-view observations of a scene without requiring complete 3D reconstruction. The framework decomposes the 3D-QA task into four interpretable components: (1) question-driven multi-view selection, (2) guided visual grounding with language-conditioned object cues, (3) spatial context encoding via a bird's-eye-view (BEV) viewpoint indicator, and (4) structured answer generation through role-based reasoning. This design enables structured, robust, and interpretable reasoning without requiring model tuning. Experimental results on ScanQA and SQA3D show that ViewMind3D achieves competitive performance compared to prior training-free and fine-tuned 3D-LLMs. In particular, our method improves performance on spatially grounded question types, such as ``What'' questions in SQA3D, while maintaining strong overall accuracy (50.8\%) and achieving 73.4 CIDEr on ScanQA. These results demonstrate that effective 3D reasoning can be achieved through modular orchestration of general-purpose LLMs and VLMs for robotic perception in real-world environments.
Chinese Translation
近年来，大型语言模型（LLMs）和视觉-语言模型（VLMs）的进展为3D问答（3D-QA）开启了新的可能性，这对于具身人工智能和机器人感知是一个关键能力。然而，大多数现有方法依赖于特定于3D的训练或需要昂贵注释的微调，这限制了它们的可扩展性和实际应用性。我们提出了 extbf{ViewMind3D}，这是一个完全无训练且模块化的框架，用于在无需完整3D重建的情况下，对场景的多视角观察进行3D空间推理。该框架将3D-QA任务分解为四个可解释的组件：（1）基于问题的多视角选择，（2）结合语言条件的物体线索进行引导的视觉定位，（3）通过鸟瞰视角（BEV）指示符进行空间上下文编码，以及（4）通过基于角色的推理生成结构化答案。这种设计使得在不需要模型调优的情况下，实现结构化、稳健和可解释的推理。在ScanQA和SQA3D上的实验结果表明，ViewMind3D在与之前的无训练和微调的3D-LLMs相比，取得了竞争力的性能。特别是，我们的方法在空间定位问题类型上提高了性能，例如SQA3D中的“什么”问题，同时保持了强大的整体准确率（50.8%）并在ScanQA上达到了73.4的CIDEr。这些结果表明，通过模块化协调通用的LLMs和VLMs，可以实现有效的3D推理，以支持现实环境中的机器人感知。
cs.CV / 89 / 2607.28463
VisualRouter: Query-Grounded Visual Sampling for Long Video Understanding
VisualRouter：基于查询的长视频理解视觉采样
Abstract
Large vision-language models (LVLMs) have achieved significant progress in video understanding, yet understanding long videos remains challenging due to the large number of visual tokens and limited context windows. Visual sampling provides a practical solution by selecting an informative subset of frames. However, existing methods typically either rely on relevance-aware sampling, leading to redundant frame selection and insufficient temporal coverage, or adopt a fixed sampling strategy regardless of query type. In this paper, we propose VisualRouter, a training-free and plug-and-play framework for query-grounded visual sampling. VisualRouter first classifies each query as either global or local and then applies the corresponding sampling strategy. For global queries, it employs a relevance-coverage hybrid strategy that preserves temporal coverage while retaining query-relevant visual evidence. For local queries, it adopts an event-aware frame selection strategy that performs event partitioning, segment-level frame allocation, and intra-event frame selection, jointly balancing relevance, coverage, and diversity with a limited number of input frames. Experiments show that VisualRouter consistently improves multiple LVLMs over uniform sampling, achieving gains of 5.2%, 7.7%, and 11.6% on Video-MME, LongVideoBench, and MLVU with Qwen2.5-VL-7B, and outperforming existing training-free visual sampling methods under the same setting.
Chinese Translation
大型视觉语言模型（LVLMs）在视频理解方面取得了显著进展，但由于视觉标记数量庞大和上下文窗口有限，理解长视频仍然具有挑战性。视觉采样通过选择信息丰富的帧子集提供了一种实用的解决方案。然而，现有方法通常依赖于关注相关性的采样，导致冗余帧选择和时间覆盖不足，或者采用固定的采样策略而不考虑查询类型。在本文中，我们提出了VisualRouter，一个无训练且即插即用的基于查询的视觉采样框架。VisualRouter首先将每个查询分类为全局或局部，然后应用相应的采样策略。对于全局查询，它采用一种相关性-覆盖混合策略，既保留时间覆盖，又保留与查询相关的视觉证据。对于局部查询，它采用一种事件感知的帧选择策略，执行事件划分、段级帧分配和事件内帧选择，联合平衡相关性、覆盖和多样性，同时限制输入帧的数量。实验表明，VisualRouter在均匀采样的基础上持续提升多个LVLMs的性能，在Video-MME、LongVideoBench和MLVU上分别取得了5.2%、7.7%和11.6%的增益，且在相同设置下超越了现有的无训练视觉采样方法。
cs.CV / 90 / 2607.28464
Can Vision-Language Models Reason about AI Edits in Images?
视觉-语言模型能否推理图像中的人工智能编辑？
Abstract
Detection and localization of AI-tampered images are critical for trustworthy AI, yet modern generative models have made such manipulations increasingly difficult to identify. While traditional binary classifiers can detect image tampering, they lack interpretability and generalization. Vision-Language Models (VLMs) offer a promising alternative due to their strong visual understanding and reasoning capabilities; however, existing approaches typically rely on supervised finetuning with curated explanations rather than exploiting their inherent reasoning capabilities. In this work, we investigate whether VLMs can be trained to reason about AI-generated image edits using reinforcement learning (RL) rather than explicit reasoning supervision. Motivated by the success in Group Relative Policy Optimization (GRPO), an RL technique that incentivizes the model to reason by generating thinking traces prior to giving the final answer, we propose a GRPO-based training framework that utilizes simple accuracy and format rewards. Given an input image, the model produces a structured reasoning trace and predicts whether the image has been tampered with. A lightweight segmentation model is then guided by the reasoning output to generate pixel-level localization masks. Experiments across multiple image manipulation datasets demonstrate that our approach achieves competitive detection and localization performance compared to state-of-the-art image forgery detectors, despite requiring substantially weaker supervision. We introduce effective intersection over union (eff-IoU), a unified metric to jointly evaluate detection and localization. These results suggest that reinforcement learning provides an effective and scalable mechanism for teaching VLMs to reason about AI-generated content.
Chinese Translation
检测和定位人工智能篡改的图像对于可信赖的人工智能至关重要，然而现代生成模型使得此类操控的识别变得愈加困难。尽管传统的二元分类器能够检测图像篡改，但它们缺乏可解释性和泛化能力。视觉-语言模型（Vision-Language Models, VLMs）由于其强大的视觉理解和推理能力，提供了一种有前景的替代方案；然而，现有的方法通常依赖于带有策划解释的监督微调，而不是利用其固有的推理能力。在本研究中，我们探讨了VLMs是否可以通过强化学习（Reinforcement Learning, RL）进行训练，以推理人工智能生成的图像编辑，而不是依赖于显式的推理监督。受到群体相对策略优化（Group Relative Policy Optimization, GRPO）成功的启发，这是一种通过在给出最终答案之前生成思考轨迹来激励模型进行推理的RL技术，我们提出了一种基于GRPO的训练框架，该框架利用简单的准确性和格式奖励。给定输入图像，模型生成结构化的推理轨迹并预测图像是否被篡改。然后，轻量级分割模型根据推理输出生成像素级定位掩码。跨多个图像操控数据集的实验表明，我们的方法在检测和定位性能上与最先进的图像伪造检测器相比表现出竞争力，尽管所需的监督显著较弱。我们引入了有效的交并比（effective intersection over union, eff-IoU），这是一个统一的指标，用于共同评估检测和定位。这些结果表明，强化学习为教导VLMs推理人工智能生成内容提供了一种有效且可扩展的机制。
cs.CV / 91 / 2607.28483
Towards Real-Time PixOOD: Efficient Anomaly Segmentation for Autonomous Vehicles
迈向实时 PixOOD：用于自主车辆的高效异常分割
Abstract
Real-time anomaly segmentation is essential for the safety of autonomous systems. Although recent approaches offer high accuracy, their computational cost limits their deployment on embedded hardware. This work presents an efficient and accelerated pipeline designed for both embedded and desktop platforms, targeting the autonomous driving and railway domains. The proposed approach reformulates the Neyman-Pearson scoring stage of PixOOD, a state-of-the-art out-of-distribution detection method, and deploys the full pipeline through hardware-optimized TensorRT compilation, reaching up to 182 FPS on a desktop NVIDIA RTX 4060 GPU and 75 FPS on the NVIDIA Jetson AGX Orin embedded platform, respectively 20x and 18x faster than the original baseline. The achieved results demonstrate that advanced anomaly segmentation can be efficiently deployed for onboard processing in autonomous driving and railway applications.
Chinese Translation
实时异常分割对于自主系统的安全至关重要。尽管近期的方法提供了高准确性，但其计算成本限制了它们在嵌入式硬件上的部署。本研究提出了一种高效且加速的处理管道，旨在同时适用于嵌入式和桌面平台，针对自主驾驶和铁路领域。所提出的方法重新构建了 PixOOD 的 Neyman-Pearson 评分阶段，PixOOD 是一种最先进的分布外检测方法，并通过硬件优化的 TensorRT 编译部署完整管道，在桌面 NVIDIA RTX 4060 GPU 上达到最高 182 FPS，在 NVIDIA Jetson AGX Orin 嵌入式平台上达到 75 FPS，分别比原始基线快 20 倍和 18 倍。所取得的结果表明，先进的异常分割可以高效地部署于自主驾驶和铁路应用的车载处理。
cs.CV / 92 / 2607.28487
AuricularWorld: Hierarchical Action-Guided World Modeling for Fine-Grained Auricular Structure Segmentation from CT Scans
AuricularWorld：基于分层动作引导的世界建模用于从CT扫描中细粒度耳廓结构分割
Abstract
Fine-grained segmentation of auricular structures in CT is challenging because the ear occupies a small image region, cartilage boundaries are highly irregular, and interfaces between cartilage and surrounding soft tissues are often ambiguous. Clinical annotations may also include both composite structures containing cartilage and adjacent skin and their corresponding cartilage-only regions, producing nested and overlapping labels. We propose a world-model-based segmentation framework that enables iterative anatomical reasoning beyond conventional feed-forward prediction. Built on an encoder-decoder architecture, the framework introduces a deterministic recurrent state-space model into the intermediate latent space. Multi-scale encoder features and partially decoded representations are fused to form a structural observation that initializes the latent dynamics. During inference, the model performs a three-step latent rollout without ground-truth guidance. Hierarchical anatomical actions update the recurrent state and progressively refine the latent representation. The resulting latent trajectory is projected back into the decoder and combined with high-resolution features to produce the final segmentation. To learn reliable latent transitions, we introduce a balanced hierarchical action objective that addresses foreground sparsity, missing anatomical groups, and imbalance between add and remove operations. Extensive experiments show that the proposed framework consistently improves segmentation accuracy and reduces HD95 by more than 43% for small, irregular, and overlapping auricular structures in CT. These results demonstrate the effectiveness of latent world-model reasoning for challenging medical image segmentation.
Chinese Translation
CT中耳廓结构的细粒度分割具有挑战性，因为耳朵占据的图像区域较小，软骨边界高度不规则，且软骨与周围软组织的界面往往模糊不清。临床标注可能还包括包含软骨和相邻皮肤的复合结构及其对应的仅软骨区域，导致嵌套和重叠的标签。我们提出了一种基于世界模型的分割框架，使得超越传统前馈预测的迭代解剖推理成为可能。该框架基于编码器-解码器架构，引入了确定性递归状态空间模型到中间潜在空间。多尺度编码器特征和部分解码表示被融合以形成结构观察，从而初始化潜在动态。在推理过程中，该模型执行三步潜在展开，而不依赖于真实标签指导。分层解剖动作更新递归状态，并逐步细化潜在表示。最终的潜在轨迹被投影回解码器，并与高分辨率特征结合以生成最终分割。为了学习可靠的潜在过渡，我们引入了一个平衡的分层动作目标，以解决前景稀疏、缺失解剖组以及添加和删除操作之间的不平衡。大量实验表明，所提出的框架在小型、不规则和重叠的CT耳廓结构的分割准确性上持续提高，并将HD95降低超过43%。这些结果证明了潜在世界模型推理在具有挑战性的医学图像分割中的有效性。
cs.CV / 93 / 2607.28509
RefCaptioner: Multi-Reference Image-Grounded Video Captioning
RefCaptioner：多参考图像基础的视频字幕生成
Abstract
Existing video captioning models generate natural descriptions of video content but cannot explicitly ground local visual elements to multiple reference images. We introduce multi-reference image-grounded video captioning, a new task requiring factual video descriptions with phrase-level reference grounding, and propose RefCaptioner, a two-stage post-training framework for this task. RefCaptioner combines mixed-data SFT with Hierarchical Coverage-Discounted GRPO to jointly improve reference selection, phrase-level binding, distractor rejection, and cross-reference consistency while preserving general video-captioning ability. To support training, we construct a corpus containing $20,000$ videos and 171,354 reference images. We further introduce MRVBench, a benchmark for evaluating caption factuality and multi-reference grounding on both real-world and AI-generated videos. Experiments show that RefCaptioner achieves the best overall performance among the open-source models while remaining competitive on standard video captioning benchmarks. Human evaluation further confirms that its captions are preferred by annotators and enable more source-faithful video reconstruction with both open-source and proprietary video generators.
Chinese Translation
现有的视频字幕生成模型能够生成自然的视频内容描述，但无法将局部视觉元素明确地与多个参考图像关联。我们引入了多参考图像基础的视频字幕生成这一新任务，该任务要求提供具有短语级参考关联的事实性视频描述，并提出了RefCaptioner，一个针对该任务的两阶段后训练框架。RefCaptioner结合了混合数据的SFT（Supervised Fine-Tuning）与分层覆盖折扣的GRPO（Generalized Reference Phrase Optimization），以共同提高参考选择、短语级绑定、干扰项拒绝和交叉参考一致性，同时保持一般视频字幕生成能力。为了支持训练，我们构建了一个包含20,000个视频和171,354个参考图像的语料库。我们进一步引入了MRVBench，一个用于评估字幕事实性和多参考关联的基准，涵盖真实世界和AI生成的视频。实验表明，RefCaptioner在开源模型中实现了最佳整体性能，同时在标准视频字幕生成基准上保持竞争力。人类评估进一步确认，其生成的字幕更受注释者青睐，并能够与开源和专有视频生成器一起实现更忠实于源内容的视频重建。
cs.CV / 94 / 2607.28516
Beyond Frame Selection: Generative Latent Evidence Aggregation for Long-Video Understanding
超越帧选择：用于长视频理解的生成潜在证据聚合
Abstract
Long-video understanding commonly compresses videos into a small set of frames or visual tokens for answer generation. Existing compact pipelines focus on retaining relevant visual content as explicit evidence. Yet making evidence available does not ensure that complementary cues across moments are integrated for answering. Our key idea is to organize selected frames into query-relevant cross-frame evidence before generation. We formulate this post-selection stage as a latent evidence interface and instantiate it with GenEvA ($\textbf{Gen}erative$ $Latent$ $\textbf{Ev}idence$ $\textbf{A}ggregation$), a distribution-guided latent evidence aggregation framework. Specifically, GenEvA uses a query-conditioned evidence distribution to focus aggregation on relevant frames, forming compact cross-frame latent evidence from their frame-specific information. Since cross-frame integration is not always needed, the same distribution determines whether to insert this latent complement. Across four benchmarks and two Video-MLLM backbones, GenEvA consistently improves matched-frame baselines. At 8 frames, it raises the four-benchmark LLaVA-Video average by $+5.2$ points and Qwen2.5-VL accuracy on LVBench by $+10.1$ points. These gains require only $0.11\%$--$0.40\%$ average video-token overhead; analyses further show task-aware allocation and benefits from Adaptive Evidence Invocation.
Chinese Translation
长视频理解通常将视频压缩为一小组帧或视觉标记以生成答案。现有的紧凑管道专注于保留相关的视觉内容作为显性证据。然而，提供证据并不确保跨时刻的互补线索能够整合以进行回答。我们的关键思想是在生成之前将选定的帧组织成与查询相关的跨帧证据。我们将这一后选择阶段形式化为潜在证据接口，并用 GenEvA（$ extbf{Gen}erative$ $ extbf{Latent}$ $ extbf{Ev}idence$ $ extbf{A}ggregation$）这一分布引导的潜在证据聚合框架进行实例化。具体而言，GenEvA 使用查询条件的证据分布来聚焦于相关帧，从它们的帧特定信息中形成紧凑的跨帧潜在证据。由于并非总是需要跨帧整合，因此相同的分布决定是否插入这一潜在补充。在四个基准测试和两个 Video-MLLM 骨干网络上，GenEvA 一致性地提高了匹配帧基线。在 8 帧时，它将四个基准的 LLaVA-Video 平均提高了 $+5.2$ 分，而在 LVBench 上将 Qwen2.5-VL 的准确率提高了 $+10.1$ 分。这些提升仅需 $0.11\%$--$0.40\\%$ 的平均视频标记开销；分析进一步表明任务感知分配和自适应证据调用的益处。
cs.CV / 95 / 2607.28526
What to Remove, What to Preserve: Dual-Ambiguity Rectification for All-in-One Image Restoration
去除什么，保留什么：全能图像修复的双重模糊纠正
Abstract
All-in-one image restoration aims to handle diverse degradations within a unified framework. Existing methods commonly encode heterogeneous degradation conditions in a shared latent space, where degradation-related cues and scene content can remain entangled. We characterize the resulting challenge as dual ambiguity: semantic ambiguity in channel-wise modulation and spatial ambiguity in restoration responses, which can lead to content corruption and residual artifacts. To mitigate this issue, we propose DAR-Net, a Dual-Ambiguity Rectification Network for all-in-one image restoration. DAR-Net first introduces a Degradation Archetype Representation (DAR) module to construct a structured degradation state through simplex-constrained archetype mixture modeling. Based on this state, a Semantic Ambiguity Rectification (SeAR) module generates degradation-aware prompts to improve channel-wise conditioning in the decoder. A Spatial Ambiguity Rectification (SpAR) module further regularizes degradation-aware and complementary features toward orthogonal response subspaces, reducing spatial interference between removal and preservation cues. Extensive experiments on standard all-in-one restoration benchmarks show that DAR-Net achieves the best overall performance under both three-degradation and five-degradation settings, improving the average PSNR over the strongest competitor by 0.14 dB and 0.34 dB, respectively; it additionally shows superior performance on CDD-11 and WeatherBench.
Chinese Translation
全能图像修复旨在在统一框架内处理多种退化现象。现有方法通常在共享的潜在空间中编码异质退化条件，其中退化相关线索与场景内容可能相互纠缠。我们将由此产生的挑战表征为双重模糊性：通道调制中的语义模糊性和修复响应中的空间模糊性，这可能导致内容损坏和残余伪影。为了解决这个问题，我们提出了DAR-Net，一种用于全能图像修复的双重模糊纠正网络。DAR-Net首先引入退化原型表示（Degradation Archetype Representation, DAR）模块，通过简单约束的原型混合建模构建结构化的退化状态。在此状态基础上，语义模糊纠正（Semantic Ambiguity Rectification, SeAR）模块生成退化感知提示，以改善解码器中的通道调节。空间模糊纠正（Spatial Ambiguity Rectification, SpAR）模块进一步规范化退化感知和互补特征，朝向正交响应子空间，减少去除与保留线索之间的空间干扰。在标准全能修复基准上的大量实验表明，DAR-Net在三种退化和五种退化设置下均实现了最佳整体性能，平均PSNR分别比最强竞争者提高了0.14 dB和0.34 dB；此外，在CDD-11和WeatherBench上也表现出优越的性能。
cs.CV / 96 / 2607.28532
MarkushGlyph and OCSRGlyph: Improved Chemical Structure Recognition
MarkushGlyph 和 OCSRGlyph：改进的化学结构识别
Abstract
Chemical structures appear in patents and the scientific literature as images. For programmatic usage, such as indexing in databases or constructing machine learning model training sets, they must be transformed into line notations. The two common forms of this task are translating an image of a single molecule (optical chemical structure recognition - OCSR) and translating a Markush structure that represents a family of molecules. While prior work in the former case is quite mature, Markush structure parsing remains a challenging task. In this work, we treat both tasks as an image-to-text translation problem. We then propose OCSRGlyph, a state-of-the-art OCSR model, improving performance over prior methods by carefully considering stereochemistry. For the Markush task, we introduce MarkushGlyph, a vision-language model that reads the entire Markush structure as an image. This contrasts with prior systems, which often use multiple stages to separately process visual and text input content. Finally, we introduce a new metric for determining the accuracy of Markush structure translations, handling failure modes present in prior metrics.
Chinese Translation
化学结构以图像的形式出现在专利和科学文献中。为了程序化使用，例如在数据库中进行索引或构建机器学习模型的训练集，它们必须转换为线性符号。这项任务的两种常见形式是翻译单个分子的图像（光学化学结构识别 - OCSR）和翻译表示一系列分子的 Markush 结构。尽管前者的相关研究已经相当成熟，但 Markush 结构解析仍然是一项具有挑战性的任务。在本研究中，我们将这两项任务视为图像到文本的翻译问题。我们提出了 OCSRGlyph，这是一种先进的 OCSR 模型，通过仔细考虑立体化学来提高性能。对于 Markush 任务，我们引入了 MarkushGlyph，这是一种视觉-语言模型，可以将整个 Markush 结构作为图像进行读取。这与先前的系统形成对比，后者通常使用多个阶段分别处理视觉和文本输入内容。最后，我们引入了一种新的指标来确定 Markush 结构翻译的准确性，处理了先前指标中存在的失败模式。
cs.CV / 97 / 2607.28538
ScaFE: Data-Efficient Scar Classification with LLM-Generated Clinical Feature Programs
ScaFE：基于LLM生成的临床特征程序的数据高效瘢痕分类
Abstract
Classifying pathological scars from clinical photographs requires distinguishing keloids from hypertrophic scars despite limited expert-labeled data and substantial acquisition variation across hospitals. End-to-end image models remain data-dependent, whereas sending photographs to a hosted vision-language model (VLM) may conflict with local data-governance requirements and yields decisions that are difficult to reproduce and audit. We introduce ScaFE (Scar Feature Engineering), which transfers clinical knowledge from a large language model (LLM) into deterministic, executable feature programs instead of asking the model to diagnose images. A web-enabled LLM retrieves clinical evidence and synthesizes programs that measure visually assessable scar attributes. Candidate programs execute in a restricted local environment, and only aggregate validation statistics and feature-level SHAP summaries are returned for iterative repair and refinement; raw images and patient-level outputs remain local. A lightweight Random Forest then operates on the resulting structured representation. On 600 photographs from three hospitals under leave-one-site-out evaluation, ScaFE achieves 81.0% site-macro balanced accuracy, exceeding the strongest baseline, BiomedCLIP, by 10.0 percentage points. With only 10% of the development data, ScaFE retains 72.0% balanced accuracy and an 11.8-point lead. Iterative refinement also raises the executable-program rate from 66.7% to 95.0%, with verified evidence for 91.7% of the final features. These results show that LLM knowledge can support data-efficient, cross-site medical image classification through local and auditable feature programs rather than direct VLM decisions.
Chinese Translation
从临床照片中对病理性瘢痕进行分类需要在有限的专家标注数据和医院间显著的获取变异中区分瘢痕疙瘩和肥厚性瘢痕。端到端的图像模型仍然依赖于数据，而将照片发送到托管的视觉-语言模型（VLM）可能与当地的数据治理要求相冲突，并且产生的决策难以重现和审计。我们提出了ScaFE（瘢痕特征工程），该方法将来自大型语言模型（LLM）的临床知识转化为确定性的、可执行的特征程序，而不是要求模型对图像进行诊断。一个支持网络的LLM检索临床证据并合成测量可视评估瘢痕属性的程序。候选程序在受限的本地环境中执行，仅返回聚合的验证统计数据和特征级SHAP摘要，以便进行迭代修复和改进；原始图像和患者级输出保持本地。然后，一个轻量级的随机森林在生成的结构化表示上运行。在对来自三家医院的600张照片进行留一站外评估时，ScaFE达到了81.0%的站点宏观平衡准确率，超过了最强基线BiomedCLIP 10.0个百分点。仅使用10%的开发数据，ScaFE仍保持72.0%的平衡准确率，并领先11.8个百分点。迭代改进还将可执行程序的比例从66.7%提高到95.0%，最终特征的91.7%有经过验证的证据。这些结果表明，LLM知识可以通过本地和可审计的特征程序支持数据高效的跨站点医学图像分类，而不是直接依赖VLM决策。
cs.CV / 98 / 2607.28565
MIND: Multimodal Intent-Driven Network via Diffusion Transformers for Medical Image Fusion
MIND：通过扩散变换器的多模态意图驱动网络用于医学图像融合
Abstract
Medical image fusion aims to integrate complementary information from diverse imaging modalities to support clinical diagnosis. Existing methods typically apply uniform fusion rules globally, lacking a deep understanding of diagnostic intents and pathological structures. To address these limitations, we propose MIND, a Multimodal Intent-Driven Network via Diffusion Transformers (DiTs) for medical image fusion. Specifically, we utilize BioMedGPT to generate intent-driven fusion texts from source images, guiding the fusion process with pathology-aware diagnostic intents. To combat the loss of 2D spatial continuity caused by 1D sequence flattening in DiTs, we design a Multi-scale Latent Adapter. This module explicitly extracts source image features before serialization, injecting them into the network via strict dimensional alignment to effectively supplement image features. To resolve the semantic shift caused by decoupling image outputs from diagnostic intents, we design a medical semantic consistency loss. This loss ensures deep semantic locking between fused images and fusion texts while maintaining the stability of the underlying physical manifold reconstruction. Comprehensive experiments on the Harvard, BraTS, and GFP datasets reveal that MIND delivers superior fusion quality, significantly improves downstream brain tumor segmentation accuracy, and enables flexible interactive fusion, holding significant promise for intent-driven intelligent clinical decision support systems.
Chinese Translation
医学图像融合旨在整合来自不同成像模态的互补信息，以支持临床诊断。现有方法通常在全局范围内应用统一的融合规则，缺乏对诊断意图和病理结构的深入理解。为了解决这些局限性，我们提出了MIND，一种通过扩散变换器（Diffusion Transformers, DiTs）的多模态意图驱动网络，用于医学图像融合。具体而言，我们利用BioMedGPT从源图像生成意图驱动的融合文本，以病理感知的诊断意图指导融合过程。为了应对DiTs中因一维序列展平导致的二维空间连续性损失，我们设计了一个多尺度潜在适配器（Multi-scale Latent Adapter）。该模块在序列化之前显式提取源图像特征，通过严格的维度对齐将其注入网络，以有效补充图像特征。为了解决因将图像输出与诊断意图解耦而导致的语义偏移，我们设计了一种医学语义一致性损失。该损失确保融合图像与融合文本之间的深层语义锁定，同时保持底层物理流形重建的稳定性。在哈佛、BraTS和GFP数据集上的全面实验表明，MIND提供了卓越的融合质量，显著提高了下游脑肿瘤分割的准确性，并实现了灵活的交互式融合，为意图驱动的智能临床决策支持系统带来了重要的前景。
cs.CV / 99 / 2607.28571
Finding Change in Satellite Archives from Text: How to Combine Before-and-After Images Efficiently
从文本中发现卫星档案的变化：如何高效地结合前后图像
Abstract
Operational Earth observation increasingly calls for answering queries such as ``find the image pairs where a new building appeared.'' This means searching an archive of before-and-after (bi-temporal) satellite image pairs and ranking each pair by how well it matches a natural-language description of the change. The component that performs this match, the fusion module that combines the ``before'' and ``after'' views, must be run at query time across many candidate pairs, so its speed largely sets the cost of every search. We present a controlled comparison of how to build that module. Using one fixed image encoder (a frozen CLIP model) and one training recipe for all variants, we evaluate eight designs drawn from three families: attention, state-space models (Mamba), and learned compression (our Temporal Bottleneck Fusion, TBF). Each design is tested on two benchmarks (LEVIR-CC and Dubai-CC) with ten random seeds, so the reported differences are statistically grounded. We outline three findings: first, a training-free two-stage search (a cheap difference model that shortlists candidates, followed by attention fusion that re-ranks them) matches or exceeds full-fusion recall on LEVIR-CC while cutting query cost $10$-$15\times$, with comparable R@1/R@5 on Dubai-CC; second, the linear-time scan of Mamba, attractive on paper, gives no speed benefit at the patch counts typical of vision transformers ($L{=}196$): the scan is limited by memory bandwidth, whereas attention maps cleanly onto parallel hardware; and third, compressing the fused representation (TBF) reduces parameters by $2.3\times$ and latency by $1.6\times$ for a change-only BLEU-1 cost of $0.007$, although more aggressive compression quietly discards change-relevant detail that aggregate metrics fail to reveal.
Chinese Translation
操作性地进行地球观测日益需要回答诸如“找到出现新建筑的图像对”的查询。这意味着要在前后（双时相）卫星图像对的档案中进行搜索，并根据每对图像与变化的自然语言描述的匹配程度进行排名。执行这种匹配的组件，即结合“前”视图和“后”视图的融合模块，必须在查询时跨多个候选对运行，因此其速度在很大程度上决定了每次搜索的成本。我们对如何构建该模块进行了受控比较。使用一个固定的图像编码器（一个冻结的 CLIP 模型）和一个适用于所有变体的训练方案，我们评估了来自三个家族的八种设计：注意力、状态空间模型（Mamba）和学习压缩（我们的时间瓶颈融合，TBF）。每种设计在两个基准（LEVIR-CC 和 Dubai-CC）上进行了测试，使用十个随机种子，因此报告的差异具有统计学基础。我们概述了三个发现：首先，一个无训练的两阶段搜索（一个廉价的差异模型用于筛选候选者，随后是注意力融合重新排名）在 LEVIR-CC 上的召回率与全融合相匹配或超过，同时将查询成本降低了 $10$-$15 imes$，在 Dubai-CC 上具有可比的 R@1/R@5；其次，Mamba 的线性时间扫描在理论上吸引人，但在典型的视觉变换器的补丁数量（$L{=}196$）下没有速度优势：扫描受限于内存带宽，而注意力图则能很好地映射到并行硬件上；第三，压缩融合表示（TBF）将参数减少了 $2.3 imes$，延迟减少了 $1.6 imes$，对于仅变化的 BLEU-1 成本为 $0.007$，尽管更激进的压缩悄然丢弃了变化相关的细节，而聚合指标未能揭示这一点。
cs.CV / 100 / 2607.28581
ROAD: Reciprocal-Objective Alignment of Discriminative Semantics for 3D Shape Generation
ROAD：用于3D形状生成的判别语义的互目标对齐
Abstract
High-fidelity 3D generation predominantly relies on scaling model capacity and data, which incurs prohibitive computational costs. This paradigm typically requires learning geometry from scratch and overlooks the rich semantic and structural priors already encapsulated in discriminative 3D foundation models. We contend that leveraging the profound understanding of the 3D world possessed by these discriminative models can significantly reduce generative cost. To this end, we propose ROAD, a framework that reduces the training cost of 3D generation by transferring these rich discriminative priors into diffusion transformers. To address the inherent semantic-structural heterogeneity between generative and discriminative latents, we introduce a reciprocal-objective alignment strategy. This method synergizes Holistic Semantic Condensing to enforce global semantic coherence and Structural Optimal Alignment, which is formulated as a bipartite matching problem to rigorously align microscopic geometric details between disparate latent spaces. The 3D foundation model is only used for training-time supervision of alignment and is not used at inference, incurring no additional inference cost. Compared with the industrial baseline Step1X-3D, the proposed ROAD achieves highly competitive generation performance with only 1.5% of the training data and significantly reduces training costs, effectively reducing the computational overhead of high-fidelity 3D generation. Code is available at https://github.com/H-EmbodVis/ROAD.
Chinese Translation
高保真3D生成主要依赖于模型容量和数据的扩展，这会产生高昂的计算成本。这种范式通常需要从头学习几何形状，并忽视了已经在判别3D基础模型中蕴含的丰富语义和结构先验。我们认为，利用这些判别模型对3D世界的深刻理解可以显著降低生成成本。为此，我们提出了ROAD，一个通过将这些丰富的判别先验转移到扩散变换器中来降低3D生成训练成本的框架。为了应对生成和判别潜变量之间固有的语义-结构异质性，我们引入了一种互目标对齐策略。该方法结合了整体语义压缩（Holistic Semantic Condensing）以强制执行全局语义一致性，以及结构最优对齐（Structural Optimal Alignment），后者被构造成一个二分匹配问题，以严格对齐不同潜空间之间的微观几何细节。3D基础模型仅用于训练时的对齐监督，而在推理时不使用，从而不产生额外的推理成本。与工业基线Step1X-3D相比，所提出的ROAD在仅使用1.5%的训练数据的情况下实现了高度竞争的生成性能，并显著降低了训练成本，有效减少了高保真3D生成的计算开销。代码可在https://github.com/H-EmbodVis/ROAD获取。
cs.CV / 101 / 2607.28589
MixFrag: Fragility-Guided Mixed-Precision Post-Training Quantization for Vision Transformers
MixFrag：基于脆弱性引导的混合精度后训练量化框架用于视觉变换器
Abstract
Post-training quantization (PTQ) has emerged as an effective solution for deploying Vision Transformers (ViTs) on resource-constrained devices. However, existing PTQ methods typically employ uniform bit-widths across transformer components, overlooking their heterogeneous sensitivity to quantization and leading to inefficient precision allocation. In this paper, we propose {MixFrag, a fragility-guided mixed-precision PTQ framework for Vision Transformers. MixFrag first estimates component-level quantization fragility by measuring the Kullback--Leibler (KL) divergence between full-precision and isolated quantized output distributions using a small calibration set. It then formulates bit allocation as a Multiple-Choice Knapsack Problem (MCKP), enabling adaptive layer-wise precision assignment under a target bit budget. Extensive experiments on ImageNet-1K across multiple Vision Transformer architectures demonstrate that MixFrag achieves competitive classification performance under practical mixed-precision settings. Furthermore, evaluations on COCO object detection and instance segmentation show that MixFrag achieves state-of-the-art performance among existing mixed-precision PTQ methods, improving the previous best method by up to 9.6 AP under the challenging MP3/MP3 setting. Additional analyses validate the proposed fragility metric and demonstrate its strong correlation with the learned bit allocation. These results establish MixFrag as an effective framework for mixed-precision post-training quantization of Vision Transformers.
Chinese Translation
后训练量化（PTQ）已成为在资源受限设备上部署视觉变换器（ViTs）的有效解决方案。然而，现有的PTQ方法通常在变换器组件之间采用统一的位宽，忽视了它们对量化的异质敏感性，从而导致精度分配效率低下。本文提出了MixFrag，一个基于脆弱性引导的混合精度PTQ框架，用于视觉变换器。MixFrag首先通过使用小型校准集测量全精度与孤立量化输出分布之间的Kullback-Leibler（KL）散度，估计组件级的量化脆弱性。然后，它将位分配公式化为多选背包问题（MCKP），使得在目标位预算下能够自适应地进行层级精度分配。在多个视觉变换器架构上对ImageNet-1K进行的广泛实验表明，MixFrag在实际混合精度设置下实现了具有竞争力的分类性能。此外，在COCO目标检测和实例分割上的评估显示，MixFrag在现有混合精度PTQ方法中实现了最先进的性能，在具有挑战性的MP3/MP3设置下将之前的最佳方法提高了多达9.6 AP。额外的分析验证了所提出的脆弱性度量，并展示了其与学习到的位分配之间的强相关性。这些结果确立了MixFrag作为视觉变换器混合精度后训练量化的有效框架。
cs.CV / 102 / 2607.28590
VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation
VAD：在多模态在线蒸馏中归因于目标重建的视觉证据
Abstract
Multimodal on-policy distillation (OPD) transfers fine-grained visual knowledge by supervising student-generated trajectories with a privileged-view teacher. Yet its next-token corrections are source-mixed, combining visual signals with linguistic priors and teacher-specific effects. The key challenge is to estimate which corrections are supported by visual evidence, not merely where or how strongly to distill. We introduce Visual Attribution Distillation (VAD), a counterfactual target-reconstruction algorithm that estimates the visually attributable part of a teacher correction. At each student-generated prefix, VAD evaluates the same fixed teacher with the relevant evidence present and removed. The corresponding change in centered log-probabilities defines ut, a signed proxy for the visual evidence direction that estimates how revealing the evidence supports or refutes candidate tokens. VAD projects the original correction onto this proxy to obtain an intervention-aligned component and a proxy-unexplained residual, then reconstructs a student-anchored target from the former. During training, this reconstructed target supplies the primary supervision signal, while the privileged teacher contributes a weak regularizer. Across six fine-grained visual benchmarks at 4B and 9B scales, VAD outperforms direct privileged-view distillation and visual-advantage weighting. Token- level and controlled-target analyses show that the proxy-aligned component is enriched in task-relevant visual corrections and yields stronger target shifts, especially when evidence refutes a mistaken answer. These results support counterfactual target reconstruction as an effective alternative to source-mixed supervision.
Chinese Translation
多模态在线蒸馏（OPD）通过使用特权视角教师监督学生生成的轨迹，转移细粒度的视觉知识。然而，其下一个标记的修正是源混合的，结合了视觉信号、语言先验和教师特定的影响。关键挑战在于估计哪些修正是由视觉证据支持的，而不仅仅是在哪里或多强烈地进行蒸馏。我们引入了视觉归因蒸馏（VAD），这是一种反事实目标重建算法，用于估计教师修正中可归因于视觉的部分。在每个学生生成的前缀中，VAD评估相同的固定教师，同时考虑相关证据的存在与去除。中心对数概率的相应变化定义了ut，这是一个有符号的代理，表示视觉证据方向，估计证据如何支持或反驳候选标记。VAD将原始修正投影到该代理上，以获得一个干预对齐的成分和一个代理未解释的残差，然后从前者重建一个以学生为锚的目标。在训练过程中，这个重建的目标提供了主要的监督信号，而特权教师则贡献了一个弱正则化器。在4B和9B规模的六个细粒度视觉基准测试中，VAD优于直接的特权视角蒸馏和视觉优势加权。标记级和控制目标分析表明，代理对齐的成分在任务相关的视觉修正中得到了丰富，并产生了更强的目标偏移，特别是在证据反驳错误答案时。这些结果支持反事实目标重建作为源混合监督的有效替代方案。
cs.CV / 103 / 2607.28595
Beacon: Knowing When and How to Perform Agentic Visual Reasoning
Beacon：了解何时以及如何进行自主视觉推理
Abstract
The fundamental goal of agentic visual reasoning is to improve the success rate of multimodal large language models (MLLMs) on complex tasks, rather than merely equipping them with a sophisticated yet inefficient reasoning paradigm. In this work, we rethink agentic visual reasoning through two key dimensions of tool use: Mode Adaptiveness (MA) and Tool Effect (TE). Mode Adaptiveness characterizes whether an MLLM can recognize when tools are truly necessary and invoke them accordingly, thereby avoiding unnecessary computational overhead while improving performance on challenging problems that require tool assistance. Tool Effect characterizes the actual impact of tool use: tools should extend the model's capabilities on problems unsolvable through text-only reasoning, while avoiding additional errors on problems that the model can already solve without tools. We conduct a comprehensive analysis to quantify these two properties and empirically reveal that existing agentic visual reasoning models exhibit limited Mode Adaptiveness, while the gains produced by tool use on hard examples are largely offset by the harm introduced on easy examples that the models can already solve. Motivated by these observations, we propose Beacon, a novel agentic visual reasoning model that achieves stronger overall performance, improved Mode Adaptiveness, and genuine tool-induced performance gains. At the core of Beacon are the Necessity-Aware Adaptive Reward and the Hint-Guided Capability Expansion mechanism in the reinforcement learning stage, which respectively encourage adaptive tool invocation based on task necessity and strengthen the model's tool-use capability on the most challenging problems. Extensive experiments across diverse benchmarks demonstrate the strong overall performance of Beacon and its substantial improvements in both Mode Adaptiveness and Tool Effect.
Chinese Translation
自主视觉推理的基本目标是提高多模态大型语言模型（MLLMs）在复杂任务上的成功率，而不仅仅是为它们提供一种复杂但低效的推理范式。在本研究中，我们通过工具使用的两个关键维度重新思考自主视觉推理：模式适应性（Mode Adaptiveness, MA）和工具效应（Tool Effect, TE）。模式适应性表征了MLLM是否能够识别工具何时真正必要并相应地调用它们，从而避免不必要的计算开销，同时提高在需要工具辅助的挑战性问题上的表现。工具效应则表征了工具使用的实际影响：工具应当扩展模型在无法仅通过文本推理解决的问题上的能力，同时避免在模型已经能够解决的问题上引入额外错误。我们进行了一项全面分析，以量化这两个属性，并实证揭示现有的自主视觉推理模型在模式适应性方面表现有限，而工具使用在难例上产生的收益在很大程度上被模型在易例上已经能够解决的问题中引入的损害所抵消。基于这些观察，我们提出了Beacon，一种新颖的自主视觉推理模型，能够实现更强的整体性能、改善的模式适应性和真正的工具引导性能提升。Beacon的核心是强化学习阶段的必要性感知自适应奖励和提示引导能力扩展机制，分别鼓励基于任务必要性的自适应工具调用，并增强模型在最具挑战性问题上的工具使用能力。通过在多样化基准上的广泛实验，证明了Beacon的强大整体性能及其在模式适应性和工具效应方面的显著提升。
cs.CV / 104 / 2607.28611
Chimera: Designing and Chinchilla-Scaling Hybrid Visual Diffusion Transformers
奇美拉：设计与金丝雀扩展混合视觉扩散变换器
Abstract
Visual generation increasingly requires high-resolution images, long videos, and multimodal context, making the quadratic cost of full attention prohibitive. We introduce Chimera, a hybrid visual diffusion backbone with a principled scaling recipe. Chimera processes text, image, and video tokens in one raster-ordered stream without positional embeddings. It combines Kimi Delta Attention (KDA) for long-context state tracking with O(N) complexity, interleaved Multi-head Latent Attention (MLA) for direct global interaction, and modality-aware short convolutions for local spatiotemporal context. Sparse Mixture-of-Experts (MoE) layers expand capacity while controlling activated compute. To scale this heterogeneous architecture, we introduce HeteroP, a module-wise scheme that transfers hyperparameters across width and depth according to each tensor's functional fan-in and model depth. HeteroP yields a consistently tuned family used to fit Chinchilla-style compute-optimal laws for activated model size, training-token count, and image-video data ratio. Guided by these laws, we train an 11B-parameter Chimera with 2B activated parameters. Experiments show three results. First, measured by pretraining diffusion loss, the dense backbone is 1.7x as compute-efficient as a matched full-attention Wan-2.1 2B baseline, while the complete system reaches 7.3x. Second, without length-specific fine-tuning, Chimera extrapolates zero-shot from 5-second training clips to 30-second videos, with only 6.5% FID degradation in the last five seconds. Third, the fitted laws show that compute-optimal image pretraining divides compute nearly evenly between activated model size and training-token count, whereas video pretraining modestly favors model size at higher budgets. These results establish a foundation for designing and scaling efficient long-context diffusion architectures.
Chinese Translation
视觉生成日益需要高分辨率图像、长视频和多模态上下文，这使得全注意力的二次成本变得不可承受。我们提出了奇美拉（Chimera），一种具有原则性扩展方案的混合视觉扩散骨干网络。奇美拉在一个光栅顺序流中处理文本、图像和视频标记，而无需位置嵌入。它结合了Kimi Delta Attention（KDA）用于长上下文状态跟踪，具有O(N)复杂度，交错的多头潜在注意力（MLA）用于直接的全局交互，以及模态感知的短卷积用于局部时空上下文。稀疏专家混合（MoE）层扩展了容量，同时控制激活计算。为了扩展这一异构架构，我们引入了HeteroP，一种模块级方案，根据每个张量的功能输入和模型深度在宽度和深度之间转移超参数。HeteroP产生了一种一致调优的家族，用于拟合金丝雀（Chinchilla）风格的计算最优法则，适用于激活模型大小、训练标记数量和图像-视频数据比例。在这些法则的指导下，我们训练了一个具有110亿参数的奇美拉，其中激活参数为20亿。实验结果显示三点。首先，根据预训练扩散损失的测量，密集骨干网络的计算效率是匹配的全注意力Wan-2.1 20亿基线的1.7倍，而完整系统达到7.3倍。其次，在没有特定长度微调的情况下，奇美拉从5秒的训练片段推断出30秒的视频，最后五秒的FID仅下降了6.5%。第三，拟合法则显示，计算最优的图像预训练几乎均匀地将计算分配在激活模型大小和训练标记数量之间，而视频预训练在较高预算下适度偏向模型大小。这些结果为设计和扩展高效的长上下文扩散架构奠定了基础。
cs.CV / 105 / 2607.28624
PhiZero: A World Model Built Around Physical Language
PhiZero：围绕物理语言构建的世界模型
Abstract
We introduce PhiZero, a physical world model built around physical language, a compact discrete representation of world-state transitions. Existing physical world models typically predict future videos directly in pixel space, leaving the underlying world dynamics implicit within high-dimensional visual predictors. Motivated by humans' ability to abstract predictive structure from visual experience and organize it in natural language for explicit reasoning, we learn physical language from in-the-wild videos through self-supervision and use it to explicitly reason about how the physical world evolves. Accordingly, PhiZero adopts a reason-then-render paradigm: it first infers future world evolution as a physical-language sequence and then renders the inferred transitions into videos. Extensive experiments across generation and understanding benchmarks validate the ability of PhiZero to model physically coherent world evolution. We further show its potential for realistic and interactive world modeling, fine-grained action-conditioned simulation, and zero-shot motion transfer.
Chinese Translation
我们介绍了PhiZero，一个围绕物理语言构建的物理世界模型，物理语言是世界状态转变的紧凑离散表示。现有的物理世界模型通常直接在像素空间中预测未来视频，使得潜在的世界动态隐含在高维视觉预测器中。受到人类能够从视觉经验中抽象出预测结构并用自然语言组织以进行明确推理的启发，我们通过自我监督从现实视频中学习物理语言，并利用它来明确推理物理世界的演变。因此，PhiZero采用了先推理后渲染的范式：它首先将未来世界演变推断为物理语言序列，然后将推断出的转变渲染为视频。在生成和理解基准测试中的大量实验验证了PhiZero建模物理一致世界演变的能力。我们进一步展示了其在现实和互动世界建模、细粒度动作条件模拟和零样本运动转移方面的潜力。
cs.CV / 106 / 2607.28625
ACE-Data-0: Human-Centric Ambient Capture as Embodied Data Engine
ACE-Data-0：以人为中心的环境捕捉作为具身数据引擎
Abstract
Embodied intelligence faces a fundamental data bottleneck. Models must capture how first-person perception, whole-body motion, dexterous manipulation, object state, sound, and touch evolve together as humans pursue goals over time. Existing datasets fragment this experience across viewpoints, modalities, or spatial scales, leaving the full perception-action loop only partially observed. We introduce the Ambient Capture Engine (ACE), a human-centric data engine that transforms real home environments into spatially calibrated, temporally synchronized recording studios. ACE operates at two complementary scales: a table-scale configuration resolves hand-object manipulation, while a room-scale configuration captures whole-body motion, locomotion, and interactions across a furnished home. ACE records egocentric and multi-view exocentric video, full-body and articulated hand motion, object geometry and 6-DoF trajectories, audio, and tactile signals as a unified multisensory stream. Using ACE, we build ACE-Data-0, comprising 150 hours and 17M video frames across 200 task categories, performed by 50 participants in 2 environments, for a total of 75,000 interaction episodes. The dataset spans atomic manipulation, long-horizon chains of household activities, and human-scene interaction, while preserving natural behavioral variation through goal-level rather than step-by-step instructions. We further introduce a hierarchical benchmark that progresses from signals to scene components and then to interactions. Evaluations of state-of-the-art methods expose substantial gaps under contact, occlusion, egomotion, and long temporal horizons. ACE-Data-0 provides synchronized human demonstrations with aligned perceptual, kinematic, and contact supervision, offering a scalable foundation for imitation learning, world models, vision-language-action systems, and embodied AI.
Chinese Translation
具身智能面临着根本的数据瓶颈。模型必须捕捉第一人称感知、全身运动、灵巧操作、物体状态、声音和触觉如何随着人类追求目标而共同演变。现有数据集在视角、模态或空间尺度上分割了这种体验，导致完整的感知-行动循环仅被部分观察到。我们引入了环境捕捉引擎（Ambient Capture Engine, ACE），这是一种以人为中心的数据引擎，将真实的家庭环境转变为空间校准和时间同步的录音工作室。ACE在两个互补的尺度上运行：桌面尺度配置解决手-物体操作，而房间尺度配置捕捉全身运动、行走以及在布置好的家庭环境中的互动。ACE记录自我中心和多视角外部视频、全身和关节手部运动、物体几何形状和6自由度轨迹、音频和触觉信号，形成统一的多感官流。利用ACE，我们构建了ACE-Data-0，包含150小时和1700万帧视频，涵盖200个任务类别，由50名参与者在2个环境中执行，总计75,000个互动事件。该数据集跨越原子操作、长期家庭活动链和人-场景互动，同时通过目标级而非逐步指令保留自然行为变异。我们进一步引入了一个分层基准，从信号到场景组件再到互动进行评估。对最先进方法的评估揭示了在接触、遮挡、自我运动和长时间跨度下的显著差距。ACE-Data-0提供了同步的人类演示，配备对齐的感知、运动学和接触监督，为模仿学习、世界模型、视觉-语言-行动系统和具身人工智能提供了可扩展的基础。
cs.CV / 107 / 2607.28627
ReToken: One Token to Improve Vision-Language Models for Visual Retrieval
ReToken：一个令牌以改善视觉-语言模型在视觉检索中的表现
Abstract
Long visual context poses a challenge for vision-language models: performance degrades as the number of distractors grows, and processing all tokens at once is computationally infeasible under GPU memory constraints. We present ReToken, a single learnable embedding trained as an explicit retrieval target that selects a sparse set of query-relevant visual tokens from a pre-filled visual KV cache. Trained on only a small image-QA dataset, ReToken yields consistent gains across image and video benchmarks: on Visual Haystacks it improves Qwen3VL-8B by 13.4 points and InternVL3.5 by 12.4 points (>20% relative), and on LVBench it transfers zero-shot to long video for an 8.0-point gain with Qwen3VL-8B. Thanks to its lightweight design, both training and long-video inference fit on a single H100. Code is available at: https://github.com/avaxiao/ReToken
Chinese Translation
长视觉上下文对视觉-语言模型构成了挑战：随着干扰项数量的增加，性能下降，而在GPU内存限制下一次性处理所有令牌在计算上不可行。我们提出了ReToken，一个作为显式检索目标训练的单一可学习嵌入，它从预填充的视觉KV缓存中选择一组稀疏的与查询相关的视觉令牌。ReToken仅在一个小型图像问答数据集上训练，在图像和视频基准测试中均取得了一致的提升：在Visual Haystacks上，它使Qwen3VL-8B提高了13.4分，使InternVL3.5提高了12.4分（相对提升超过20%），而在LVBench上，它在长视频上零-shot转移，使Qwen3VL-8B获得了8.0分的提升。得益于其轻量化设计，训练和长视频推理均可在单个H100上完成。代码可在以下链接获取：https://github.com/avaxiao/ReToken
cs.AI / 1 / 2607.27230
Multi-Head Attention Residuals
多头注意力残差
Abstract
Transformers propagate information across depth through a single additive residual stream: every sublayer reads only the most recent state. Attention residuals relax this by letting each sublayer attend, through a learned softmax. However, that read uses a single query shared across the entire width, so every feature subspace must read the depth history through one distribution. The cost of this forced compromise grows with how much the subspaces disagree about which layers to read, and disagreement grows with model width. We introduce Multi-Head Attention Residuals (MHAR): the routing query is reshaped into H per-subspace heads, each with its own softmax over the depth history. The read becomes block-diagonal, the reshape adds zero parameters and negligible compute, and H = 1 recovers attention residuals exactly. Trained from scratch on a deduplicated Nemotron-based anneal corpus that is quality-filtered and STEM- and code-heavy, MHAR improves validation loss over a standard Transformer at 100M, 350M, and 1B (-0.061, -0.149, and -0.140). It achieves the best result among four methods in every setting, with the gain increasing from 100M to the larger scales. The head count is a real design axis rather than a free knob: validation loss is U-shaped with respect to H, with a flat optimum at H = 4 or H = 8 across scales. We adopt H = 8 for large-scale models; over-splitting beyond this point (H = 16) consistently gives back part of the gain. A direct probe of the trained queries confirms that learned subspace disagreement is the underlying driver. Fused Triton routing kernels increase attention-residual training throughput from 0.2-0.5x to 0.55-0.88x of the baseline while maintaining near-baseline peak memory. An identity-preserving conversion using delta attention residuals supports 8B mid-training, yielding improvements of +3.2 on GSM8K and +3.1 on GPQA.
Chinese Translation
变换器通过单一的加法残差流在深度上传播信息：每个子层仅读取最新状态。注意力残差通过让每个子层通过学习的 softmax 进行关注，从而放宽了这一限制。然而，这种读取使用的是一个在整个宽度上共享的单一查询，因此每个特征子空间必须通过一个分布读取深度历史。这种强制妥协的成本随着子空间对读取哪些层的意见不合而增加，而不合意见随着模型宽度的增加而增加。我们引入了多头注意力残差（Multi-Head Attention Residuals, MHAR）：路由查询被重塑为每个子空间 H 个头，每个头在深度历史上都有自己的 softmax。读取变为块对角形式，重塑增加了零参数和微不足道的计算，而 H = 1 则完全恢复了注意力残差。在一个去重的基于 Nemotron 的退火语料库上从零开始训练，该语料库经过质量过滤，且以 STEM 和代码为主，MHAR 在 100M、350M 和 1B 的验证损失上优于标准变换器（分别为 -0.061、-0.149 和 -0.140）。在每个设置中，它在四种方法中都取得了最佳结果，且增益从 100M 到更大规模逐渐增加。头的数量是一个真实的设计轴，而不是一个自由的调节参数：验证损失关于 H 呈 U 形分布，在 H = 4 或 H = 8 时在各个规模上都有平坦的最优点。我们在大规模模型中采用 H = 8；超出这一点的过度拆分（H = 16）始终会部分回吐增益。对训练查询的直接探测确认了学习到的子空间不合意见是潜在驱动因素。融合的 Triton 路由内核将注意力残差训练的吞吐量从基线的 0.2-0.5 倍提高到 0.55-0.88 倍，同时保持接近基线的峰值内存。使用增量注意力残差的保持身份转换支持 8B 中期训练，在 GSM8K 和 GPQA 上分别带来了 +3.2 和 +3.1 的提升。
cs.AI / 2 / 2607.27231
KernelGenBench: A Multi-Source and Multi-Chip Benchmark for LLM-based Kernel Generation
KernelGenBench：基于大型语言模型的多源多芯片内核生成基准测试
Abstract
Large language models (LLMs) have significantly increased the demand for efficient accelerator kernels, but kernel development remains a highly specialized and labor-intensive task. The recent rise of LLMs and agentic frameworks offers a promising pathway toward automatic kernel generation. However, despite rapid progress, there is still no comprehensive benchmark to rigorously evaluate LLM-generated kernels across diverse operator sources or heterogeneous hardware platforms. We present KernelGenBench, a unified benchmark for systematically evaluating LLM- and agent-generated Triton kernels across diverse operator sources and heterogeneous hardware platforms. It comprises two complementary sub-benchmarks: KernelGenBench-MS (Multi-Source), evaluating 210 operators from three sources beyond standard PyTorch-centric tasks, and KernelGenBench-MC (Multi-Chip), measuring performance portability across six heterogeneous hardware platforms using a 110-operator subset. Our large-scale evaluation, consuming over 15 billion tokens, shows: (1) agent-based methods consistently outperform pure LLM sampling methods, while cuBLAS operators are the most challenging across all methods; (2) generation performance varies significantly across hardware platforms, with even recent kernel-specialized agents experiencing severe cross-platform degradation (e.g., AutoKernel drops from 87% on NVIDIA to 25% on Platform E); (3) autonomous kernel generation remains highly cost-intensive, with specialized agent methods averaging 5.11 million tokens per successful operator (AKO4all reaches 5.19 million), orders of magnitude higher than simple LLM sampling approaches.
Chinese Translation
大型语言模型（LLMs）显著增加了对高效加速器内核的需求，但内核开发仍然是一项高度专业化且劳动密集的任务。最近LLMs和自主框架的兴起为自动内核生成提供了一个有希望的途径。然而，尽管取得了快速进展，目前仍缺乏一个全面的基准来严格评估LLM生成的内核在不同操作符来源或异构硬件平台上的表现。我们提出了KernelGenBench，这是一个统一的基准，用于系统地评估LLM和自主生成的Triton内核，涵盖多种操作符来源和异构硬件平台。它由两个互补的子基准组成：KernelGenBench-MS（多源），评估来自三个来源的210个操作符，超出标准的以PyTorch为中心的任务；KernelGenBench-MC（多芯片），使用110个操作符子集测量六个异构硬件平台上的性能可移植性。我们的规模化评估消耗了超过150亿个标记，结果显示：（1）基于自主的方法在性能上始终优于纯LLM采样方法，而cuBLAS操作符在所有方法中都是最具挑战性的；（2）生成性能在硬件平台之间差异显著，即使是最近的内核专用代理也经历了严重的跨平台性能下降（例如，AutoKernel在NVIDIA上的表现从87%下降到在平台E上的25%）；（3）自主内核生成仍然高度成本密集，专用代理方法每成功操作符平均消耗511万标记（AKO4all达到519万），远高于简单的LLM采样方法。
cs.AI / 3 / 2607.27235
RadHarmony: Radiological Data Handling in the Era of Agentic AI
RadHarmony：在智能代理时代的放射学数据处理
Abstract
Training deep learning models on radiological images requires integrating heterogeneous datasets across different sources, file formats, directory layouts, label schemas, and annotation types. We present RadHarmony, an open-source Python library that provides a unified API for loading, harmonizing, and augmenting radiological datasets, with a primary focus on chest radiographs and early support for computed tomography (CT) and magnetic resonance imaging (MRI). RadHarmony standardizes metadata from 24 public datasets into a single tabular format, wraps MONAI's map-style datasets for deep-learning-ready sample delivery with optional on-disk caching, and supports classification labels, segmentation masks, bounding boxes, and radiology report text through a single interface, with an interactive visualization tool for dataset exploration and verification. To lower the barrier for integrating new datasets, RadHarmony introduces an AI-agent skill that guides the full integration workflow from raw data inspection through code generation and testing. We demonstrate the library's utility by pretraining RadHarmony-ViT, a reference vision transformer baseline that combines three heterogeneous chest radiograph datasets with no dataset-specific code. The code and pretrained model weights are available at https://github.com/f10409/RadHarmony.
Chinese Translation
在放射学图像上训练深度学习模型需要整合来自不同来源、文件格式、目录布局、标签模式和注释类型的异构数据集。我们提出了RadHarmony，一个开源的Python库，提供统一的API用于加载、协调和增强放射学数据集，主要集中于胸部X光片，并早期支持计算机断层扫描（CT）和磁共振成像（MRI）。RadHarmony将24个公共数据集的元数据标准化为单一的表格格式，封装MONAI的映射样式数据集，以便于深度学习准备的样本交付，并支持分类标签、分割掩码、边界框和放射学报告文本，通过单一接口提供，同时配备交互式可视化工具用于数据集的探索和验证。为了降低整合新数据集的门槛，RadHarmony引入了一种AI代理技能，指导从原始数据检查到代码生成和测试的完整整合工作流程。我们通过预训练RadHarmony-ViT展示了该库的实用性，这是一个参考视觉变换器基线，结合了三个异构的胸部X光数据集，而无需特定于数据集的代码。代码和预训练模型权重可在https://github.com/f10409/RadHarmony获取。
cs.AI / 4 / 2607.27240
Asymmetric Collapse in Model Merging: When Refusal Over- writes Recognition
模型合并中的不对称崩溃：拒绝如何覆盖识别
Abstract
Model merging is often used to combine capabilities from separately fine-tuned models without additional training, but it is unclear whether standard merging methods preserve multiple safety-relevant behaviors simultaneously. We study this question through a controlled case study using two Gemma-3-1B-IT finetunes on two complementary safety objectives: CARES harm-level classification and WildJailbreak adversarial refusal. We merge the two fine-tunes using Linear, SLERP, TIES, and DARE-TIES, and evaluate the merged models on classification accuracy, attack resistance, and benign compliance. Across all four methods, attack resistance transfers significantly more than classification accuracy: merged models retain 81-85% jailbreak refusal rates while CARES accuracy falls to at most 12.9%. Weight-space measurements suggest that this asymmetry is not caused by strongly opposing task-vector directions: the two task vectors are nearly orthogonal (cosine similarity 0.011). Instead, the refusal fine-tune induces consistently larger per-layer task-vector magnitudes, causing magnitude-sensitive methods to favor refusal updates. These results show that standard model merging can collapse safety recognition into broad refusal when safety-relevant task vectors differ substantially in scale.
Chinese Translation
模型合并通常用于在不进行额外训练的情况下结合来自单独微调模型的能力，但尚不清楚标准合并方法是否能够同时保留多种与安全相关的行为。我们通过一个受控案例研究来探讨这个问题，使用两个针对两个互补安全目标的Gemma-3-1B-IT微调模型：CARES危害级别分类和WildJailbreak对抗性拒绝。我们使用线性（Linear）、球面线性插值（SLERP）、TIES和DARE-TIES合并这两个微调模型，并在分类准确性、攻击抵抗力和良性合规性上评估合并后的模型。在所有四种方法中，攻击抵抗力的转移显著高于分类准确性：合并模型保留了81-85%的越狱拒绝率，而CARES的准确率最多降至12.9%。权重空间测量表明，这种不对称性并不是由于任务向量方向的强烈对立造成的：这两个任务向量几乎是正交的（余弦相似度为0.011）。相反，拒绝微调导致每层任务向量的幅度持续增大，导致对幅度敏感的方法更倾向于拒绝更新。这些结果表明，标准模型合并可能会在安全相关任务向量在规模上存在显著差异时，将安全识别崩溃为广泛的拒绝。
cs.AI / 5 / 2607.27248
Divergence Decoding: Training-Free Capability Fusion
发散解码：无训练能力融合
Abstract
While large language models excel in reasoning, these generalists often lack knowledge for specialized scientific domains. Conversely, domain models~(specialists), while knowledgeable, suffer from specialization side-effects including diminished logic and reduced robustness.To address this dilemma, we introduce Divergence Decoding, a training-free framework for capability fusion. It reconstructs the "draft-and-verify" skeleton of speculative decoding into an adaptive routing mechanism. The core is using Jensen-Shannon divergence to monitor the distributional disagreement between the two models at each token. When the specialist exhibits significant divergence, our method identifies it as a potential reasoning risk and instantaneously routes control to the generalist. This allows the dynamic injection of general reasoning while preserving domain expertise, achieving inference-time policy composition of the generalist and the specialist.We evaluate Divergence Decoding across diverse model families (Qwen and Llama series) on challenging scientific benchmarks (GPQA, ChemBench, and ChemCoTBench). Experimental results demonstrate that Divergence Decoding outperforms both the domain-specialized and general-purpose models, effectively surpassing the performance of most single-model baseline. This suggests that Divergence Decoding provides a general, training-free paradigm for fusing diverse LLM capabilities through adaptive inference-time collaboration.
Chinese Translation
尽管大型语言模型在推理方面表现出色，但这些通用模型通常缺乏专门科学领域的知识。相反，领域模型（专家）虽然知识丰富，但受到专业化副作用的影响，包括逻辑性下降和鲁棒性降低。为了解决这一困境，我们提出了发散解码（Divergence Decoding），一个无训练的能力融合框架。该框架将推测解码的“草拟与验证”结构重构为一种自适应路由机制。其核心是使用詹森-香农散度（Jensen-Shannon divergence）来监测两个模型在每个标记上的分布不一致性。当专家模型表现出显著的发散时，我们的方法将其识别为潜在的推理风险，并立即将控制权转移给通用模型。这允许在保持领域专业知识的同时动态注入通用推理，实现通用模型与专家模型的推理时策略组合。我们在多种模型系列（Qwen 和 Llama 系列）上评估了发散解码，并在具有挑战性的科学基准（GPQA、ChemBench 和 ChemCoTBench）上进行了测试。实验结果表明，发散解码的表现超过了领域专门模型和通用模型，有效超越了大多数单一模型基线的性能。这表明发散解码提供了一种通用的、无训练的范式，通过自适应推理时协作融合多样的 LLM 能力。
cs.AI / 6 / 2607.27287
PIE-APT: A Unified Framework for Temporal Planning and Contradiction Hunting via Incremental Direct-Derivation Abduction
PIE-APT：通过增量直接推导归纳的统一时间规划与矛盾猎捕框架
Abstract
Reasoning and planning over Dynamic Knowledge Graphs (DKGs) present significant challenges, especially in open-world environments with incomplete information. Existing action formalisms often face decidability issues and the Ramification Problem, while managing incomplete knowledge via structural abduction requires expansive combinatorial search. This paper introduces a unified framework with two integrated modules---\textbf{PIE-Abducer} (incremental direct-derivation abduction) and \textbf{PIE-APT} (Abductive Planning for Temporal KGs)---operating natively on the highly expressive Description Logic. We model state transitions along a linear timeline as non-monotonic updates to deductively closed DL theories. Treating the incremental reasoner as a black-box and representing actions natively in OWL without external modal operators preserves logical decidability. To address incomplete knowledge, \textbf{PIE-Abducer} circumvents traditional Minimal Hitting Set (MHS) enumeration. Instead of combinatorial syntactic search, it injects the logical negation of a target goal into a consistent branch and extracts missing premises via direct refutation consequences. \textbf{PIE-APT} then employs a recursive \textit{Generate-and-Test} architecture, interleaving backward-chaining A* search with \textbf{PIE-Abducer} up to a bounded causal depth, followed by strict validation via forward-chaining Temporal Projection. We evaluate four OWL benchmarks stressing semantic abilities absent in classical planning: parameterized goals with witness search, mid-search DL entailment, open-world assumption injection, and adversarial contradiction hunting. Results demonstrate qualitative superiority over classical planners and prove our direct-derivation approach quantitatively outperforms an MHS-faithful baseline during abductive enrichment.
Chinese Translation
在动态知识图谱（DKGs）上进行推理和规划面临重大挑战，尤其是在信息不完整的开放世界环境中。现有的行动形式化方法常常面临可判定性问题和后果问题，而通过结构归纳管理不完整知识则需要广泛的组合搜索。本文提出了一个统一框架，包含两个集成模块—— extbf{PIE-Abducer}（增量直接推导归纳）和 extbf{PIE-APT}（时间知识图谱的归纳规划）——在高度表达性的描述逻辑上原生运行。我们将沿线性时间轴的状态转变建模为对演绎闭合的描述逻辑理论的非单调更新。将增量推理器视为黑箱，并在OWL中原生表示行动而不使用外部模态运算符，从而保持逻辑可判定性。为了应对不完整知识， extbf{PIE-Abducer}规避了传统的最小击中集（MHS）枚举。它不采用组合语法搜索，而是将目标目标的逻辑否定注入到一致的分支中，并通过直接反驳后果提取缺失的前提。随后， extbf{PIE-APT}采用递归的 extit{生成与测试}架构，将向后链推理的A*搜索与 extbf{PIE-Abducer}交替进行，直到有界因果深度，然后通过向前链的时间投影进行严格验证。我们评估了四个OWL基准，强调了经典规划中缺失的语义能力：带见证搜索的参数化目标、中间搜索的描述逻辑蕴涵、开放世界假设注入和对抗性矛盾猎捕。结果表明，我们的方法在定性上优于经典规划器，并证明我们的直接推导方法在归纳丰富过程中定量上优于一个忠实于MHS的基线。
cs.AI / 7 / 2607.27354
PAUSE: A User-Centric Benchmark for Personal AI Assistants in Unified Service Environments
PAUSE：统一服务环境中以用户为中心的个人人工智能助手基准测试
Abstract
Personal AI assistants are increasingly deployed as task-oriented, tool-augmented agents that operate within unified service environments to support everyday user activities. In realistic settings, such assistants must reason over persistent user state, respect user-specific configurations and permissions, and sustain long-horizon, constraint-aware interactions across multiple services. Existing benchmarks, however, often fragment service contexts or abstract away user state, limiting their ability to evaluate user-centric personal assistant behavior in realistic service settings. We introduce PAUSE, a user-centric benchmark for evaluating personal AI assistants in stateful, service-integrated environments. PAUSE captures core challenges of real-world assistant deployment by requiring agents to coordinate actions across heterogeneous user-owned resources while maintaining consistency with environment state, authorization constraints over multi-turn interactions. The benchmark incorporates explicit user-agent interaction via realistic user simulation, enabling evaluation beyond static tool execution. To support principled and reproducible evaluation, PAUSE adopts a multi-regime evaluation framework aligned with task characteristics. Open-ended service management tasks are assessed using semantic and trajectory-level behavioral metrics, while constraint-intensive tasks admit deterministic, state-based verification. Benchmark results show that even state-of-the-art proprietary models fail to reach 70% task completion on scenarios requiring stateful reasoning and configuration awareness, revealing consistent and interpretable failure patterns. Finally, we present a user-centric synthesis pipeline that enables scalable generation of coherent service environments, user configurations, and reliably annotated tasks, supporting benchmark extensibility and future research.
Chinese Translation
个人人工智能助手越来越多地作为面向任务的工具增强代理被部署在统一服务环境中，以支持日常用户活动。在现实环境中，这些助手必须对持久的用户状态进行推理，尊重用户特定的配置和权限，并在多个服务之间维持长期的、考虑约束的交互。然而，现有的基准测试往往将服务上下文碎片化或抽象化用户状态，限制了它们在现实服务环境中评估以用户为中心的个人助手行为的能力。我们引入了PAUSE，这是一个用于评估在有状态、服务集成环境中个人人工智能助手的以用户为中心的基准测试。PAUSE通过要求代理在异构用户拥有的资源之间协调行动，同时保持与环境状态的一致性和对多轮交互的授权约束，捕捉了现实世界助手部署的核心挑战。该基准测试通过现实用户模拟纳入了明确的用户-代理交互，使评估超越静态工具执行。为了支持原则性和可重复的评估，PAUSE采用了与任务特征相一致的多模式评估框架。开放式服务管理任务使用语义和轨迹级行为指标进行评估，而约束密集型任务则允许确定性、基于状态的验证。基准测试结果显示，即使是最先进的专有模型在需要有状态推理和配置意识的场景中也未能达到70%的任务完成率，揭示了一致且可解释的失败模式。最后，我们提出了一个以用户为中心的合成管道，能够可扩展地生成一致的服务环境、用户配置和可靠注释的任务，支持基准测试的扩展性和未来研究。
cs.AI / 8 / 2607.27360
SkillMentor: LLM Agent Self-Evolution via Learning Blind-Spot Diagnosis
SkillMentor：通过学习盲点诊断实现大型语言模型代理的自我进化
Abstract
Agent self-evolution has primarily focused on learning how to act, while overlooking an equally important capability: learning to discover what an agent does not know. Existing approaches typically assume that failure discovery is given, focusing on how to repair failures once they are identified. We ask whether blind-spot diagnosis itself can be learned. We thus study diagnosis as an agent capability separate from execution, and exclude two alternative sources of progress: executor adaptation and human supervision. Under these constraints, performance cannot improve through executor updates or annotated examples, forcing all improvements to originate from the learned diagnostic capability. We propose SkillMentor, which trains a Mentor policy via reinforcement learning to generate diagnostic tasks, identify recurrent failure modes, and curate them into reusable corrective skills. Across AppWorld and BFCLv3, SkillMentor improves executor performance by an average of 44.2%. These results suggest that blind-spot diagnosis is a learnable capability, enabling self-evolution without updating executor weights or relying on human-curated data.
Chinese Translation
代理的自我进化主要集中在学习如何行动，而忽视了同样重要的能力：学习发现代理所不知道的内容。现有的方法通常假设故障发现是已知的，专注于在识别故障后如何修复它们。我们探讨盲点诊断本身是否可以被学习。因此，我们将诊断视为一种与执行分开的代理能力，并排除了两种替代的进展来源：执行者适应和人工监督。在这些限制下，性能无法通过执行者更新或标注示例来提高，迫使所有改进都源于学习到的诊断能力。我们提出了SkillMentor，通过强化学习训练一个导师策略，以生成诊断任务，识别重复的故障模式，并将其整理为可重用的纠正技能。在AppWorld和BFCLv3上，SkillMentor使执行者性能平均提高了44.2%。这些结果表明，盲点诊断是一种可学习的能力，使得自我进化成为可能，而无需更新执行者权重或依赖人工整理的数据。
cs.AI / 9 / 2607.27386
Beyond the Bidirectional Promise: Re-evaluating the Robustness of Diffusion Language Models
超越双向承诺：重新评估扩散语言模型的鲁棒性
Abstract
Diffusion Language Models (DLMs) offer a compelling alternative to autoregressive (AR) generation by enabling bidirectional context and iterative refinement. However, their reliability under natural input noise and adversarial attacks remains under-explored. To address this, we systematically evaluate DLM robustness and calibration against AR baselines, using two parameter-matched pairs (LLaDA-8B vs. LLaMA-3-8B and Dream-7B vs. Qwen2.5-7B) across 32 natural perturbation conditions, adversarial gradient probes, and mechanistic hidden-state analyses. This paired design effectively isolates architecture-intrinsic properties from weight-dependent behaviors. We find a nuanced robustness profile: while highly stochastic DLM loss landscapes naturally resist gradient-based adversarial suffixes, they provide no guaranteed defense against natural noise, proving that everyday robustness is weight-dependent rather than inherently architectural. Furthermore, DLMs exhibit systematic overconfidence, presenting a practical deployment hazard. Most crucially, mechanistic probing reveals that all models perfectly encode input corruption, isolating behavioral fragility entirely to a decoder routing failure. Consistent with this diagnosis, we show that surface-level prompt patching fails to improve over noisy baselines. Ultimately, DLM robustness cannot be patched on; it must be fundamentally integrated into the iterative decoding loop.
Chinese Translation
扩散语言模型（DLMs）通过实现双向上下文和迭代优化，提供了一种引人注目的自回归（AR）生成替代方案。然而，它们在自然输入噪声和对抗攻击下的可靠性仍然未得到充分探索。为了解决这个问题，我们系统地评估了DLM的鲁棒性和校准，针对AR基线进行了比较，使用了两个参数匹配的模型对（LLaDA-8B与LLaMA-3-8B，以及Dream-7B与Qwen2.5-7B），在32种自然扰动条件、对抗梯度探测和机制隐藏状态分析中进行测试。这种配对设计有效地将架构固有特性与权重依赖行为隔离开来。我们发现了一个细致的鲁棒性特征：虽然高度随机的DLM损失景观自然抵抗基于梯度的对抗后缀，但它们对自然噪声并没有提供保证的防御，证明日常鲁棒性是依赖于权重而非固有架构的。此外，DLM表现出系统性的过度自信，构成了一种实际部署的危险。最重要的是，机制探测揭示所有模型完美编码输入损坏，行为脆弱性完全归因于解码器路由失败。与这一诊断一致，我们显示表面级的提示修补未能改善噪声基线的表现。最终，DLM的鲁棒性不能通过修补来实现；它必须在迭代解码循环中根本性地整合。
cs.AI / 10 / 2607.27415
Bridging Inference-Time Scaling and Episodic Memory with Action-Centric Graphs
通过以行动为中心的图桥接推理时扩展与情节记忆
Abstract
Recent advancements in inference-time scaling have significantly unlocked the complex reasoning capabilities of Large Language Models~(LLMs). However, for agents, these approaches suffer from a critical inefficiency, operating in a stateless manner and engaging in redundant search processes. Existing memory mechanisms largely rely on the reasoning capabilities of LLMs, leading to prohibitive computational costs. In this paper, we propose a novel framework, \textit{GAMER}~(Graph-based Action-centric Memory with Episodic Reasoning), that bridges the gap between inference scaling and episodic memory. Our approach models historical reasoning as a dynamic \textit{Action-Centric Graph}. By decoupling the memory mechanism from LLMs, our method can save token/money usage by providing less memory context than memory mechanism baselines. To extract knowledge from the graph effectively, we use a dual-stream Temporal Difference learning mechanism to estimate the positive~(suggestion) and negative~(avoidance) value of action nodes based on past successes and failures. During the inference phase, this learned value function optimizes decision-making bi-directionally, so that positive values provide action suggestions, while negative values indicate high-risk actions. By performing efficient searches on the graph, our method significantly improves the efficiency of inference scaling. Experiments on multiple benchmarks demonstrate that \textit{GAMER} achieves superior performance by \textbf{20.81\%/6.17\%} for success/progress rate compared to vanilla baselines.
Chinese Translation
近期在推理时扩展方面的进展显著解锁了大型语言模型（LLMs）的复杂推理能力。然而，对于智能体而言，这些方法存在一个关键的低效性，表现为无状态操作并参与冗余的搜索过程。现有的记忆机制在很大程度上依赖于LLMs的推理能力，导致了高昂的计算成本。本文提出了一种新颖的框架， extit{GAMER}（基于图的以行动为中心的情节记忆），旨在弥合推理扩展与情节记忆之间的差距。我们的方法将历史推理建模为动态的 extit{以行动为中心的图}。通过将记忆机制与LLMs解耦，我们的方法能够通过提供比记忆机制基线更少的记忆上下文来节省令牌/资金的使用。为了有效地从图中提取知识，我们使用双流时间差学习机制，根据过去的成功和失败来估计行动节点的正值（建议）和负值（避免）。在推理阶段，这一学习的价值函数双向优化决策，使得正值提供行动建议，而负值则指示高风险行动。通过在图上进行高效搜索，我们的方法显著提高了推理扩展的效率。在多个基准测试中的实验表明， extit{GAMER}在成功率/进展率方面相比于普通基线实现了 extbf{20.81\%/6.17\\%}的优越性能。
cs.AI / 11 / 2607.27420
Dimensionality and Measurement Precision in HLE's Multiple-Choice Subset
HLE多项选择子集中的维度与测量精度
Abstract
Humanity's Last Exam (HLE) is widely used to evaluate frontier language models. HLE organizes its questions into eight subject-domain categories, whose subscores are often interpreted as evidence of distinct capabilities. However, no study has assessed whether these labels correspond to empirically separable latent constructs, nor whether the benchmark effectively differentiates between models of similar ability. We evaluate 29 LLMs on the text-only multiple-choice subset of HLE ($J = 428$ items) and apply psychometric methods to assess both the dimensionality of the benchmark and the distribution of its measurement precision. Fitting a two-parameter logistic IRT model, we find convergent evidence that HLE measures a single general reasoning factor: McDonald's $\omega_h = 0.998$, domain labels explain only 3.5\% of item response variance, within- and between-domain residual correlations are nearly identical (Cohen's $d = 0.016$), and domain-specific ability estimates are near-redundant with the total score ($r \geq 0.81$). A separate analysis of the test information function reveals that measurement precision concentrates at moderate ability levels and drops sharply above $\theta = 0$, where frontier models sit. These findings suggest that HLE's domain subscores do not warrant distinct capability interpretations and that the benchmark's ability to discriminate among the strongest models is limited.
Chinese Translation
人类最后考试（HLE）广泛用于评估前沿语言模型。HLE将其问题组织为八个学科领域类别，其子分数通常被解释为不同能力的证据。然而，尚无研究评估这些标签是否对应于经验上可分离的潜在构念，也未探讨该基准是否有效区分具有相似能力的模型。我们对29个大型语言模型（LLMs）在HLE的文本-only多项选择子集上进行评估（$J = 428$题），并应用心理测量方法评估基准的维度性及其测量精度的分布。通过拟合一个双参数逻辑IRT模型，我们发现HLE测量单一通用推理因子的收敛证据：McDonald's $eta_h = 0.998$，领域标签仅解释3.5 ext{%}的题目反应方差，领域内和领域间的残差相关性几乎相同（Cohen's $d = 0.016$），领域特定能力估计与总分几乎冗余（$r ext{≥} 0.81$）。对测试信息函数的单独分析显示，测量精度集中在中等能力水平，并在$ heta = 0$以上急剧下降，而前沿模型正处于该水平。这些发现表明，HLE的领域子分数不应被解读为不同的能力，并且该基准在区分最强模型方面的能力有限。
cs.AI / 12 / 2607.27443
Leveraging Trajectory Graphs for Pre-Execution Error Diagnosis in Agentic LLM Systems
利用轨迹图进行代理型大语言模型系统的执行前错误诊断
Abstract
Large Language Model~(LLM)-based agents have demonstrated exceptional performance across a wide range of complex interactive tasks. However, they often struggle with long-horizon interactive tasks common in domains, such as embodied AI. The complexity and vast action spaces in these settings lead to compounding errors, where a single suboptimal action can derail an entire trajectory, causing the agent to exhaust its limited step budget on inefficient or unrecoverable paths. To overcome this without costly fine-tuning, we draw inspiration from software debugging, where execution logs are analyzed to preemptively catch errors. We propose \textit{Trajectory Graph Copilot}, a novel framework that acts as a ``copilot'' for LLM agents by diagnosing potential action errors before they are executed. At its core,\textit{Graph Debugger} models historical trajectories as a probabilistic graph and uses a Graph Neural Network to identify sequential action patterns that frequently lead to failure. Functioning as a proactive diagnostic sandbox, our method provides early warnings on potentially flawed actions, prompting the agent to self-correct. This pre-action error diagnosis prevents costly mistakes, significantly enhancing the agent's ability to complete long-horizon tasks successfully. The extensive experiments on four benchmarks with three LLM agents demonstrate a $14.69\%$ pass ratio improvement on average.
Chinese Translation
基于大语言模型（LLM）的代理在广泛复杂的交互任务中表现出色。然而，它们在长时间交互任务中常常面临挑战，这些任务在具身人工智能等领域尤为常见。这些环境中的复杂性和庞大的动作空间导致了错误的累积，其中一个次优的动作可能会使整个轨迹偏离，导致代理在低效或不可恢复的路径上耗尽其有限的步骤预算。为了在不进行昂贵的微调的情况下克服这一问题，我们借鉴了软件调试的灵感，在调试过程中分析执行日志以预防性地捕捉错误。我们提出了 extit{轨迹图副驾驶}（Trajectory Graph Copilot），这是一个新颖的框架，充当LLM代理的“副驾驶”，通过在执行之前诊断潜在的动作错误。在其核心， extit{图调试器}（Graph Debugger）将历史轨迹建模为一个概率图，并使用图神经网络识别经常导致失败的顺序动作模式。作为一个主动的诊断沙箱，我们的方法提供对潜在缺陷动作的早期警告，促使代理自我纠正。这种执行前的错误诊断可以防止代价高昂的错误，显著增强代理成功完成长时间任务的能力。在对四个基准和三个LLM代理进行的广泛实验中，平均通过率提高了$14.69\%$。
cs.AI / 13 / 2607.27453
VAmoS Bench: Voice Agent Simulation Bench
VAmoS Bench：语音代理模拟基准
Abstract
Production voice agents span cascaded, speech-to-speech, and hybrid architectures. Voice-agent benchmarks typically measure component quality and conversational properties such as word error rate, latency, naturalness, and turn-taking. Fewer measure whether the agent handled a phone call correctly on its own. Contact centers refer to this as ``containment'': the share of phone calls the automated system resolves without handing off to a human. On some phone calls the right outcome is refusal or a redirect. To address this gap, we introduce VAmoS Bench, the Voice Agent Simulation Bench. It measures complete voice-agent systems end to end in a stateful customer-support task. The agent is Riley, a credit-card support representative for a fictional bank who can freeze, cancel, replace, or activate a card. Each of 100 scenarios supplies a simulated caller with a private goal and a seeded PostgreSQL backend. The platform uses each scenario to populate and activate an isolated simulation in which the caller reaches Riley over audio; roughly one-third apply adversarial pressure. The agent can use five tools that execute real SQL against the backend. Each scenario also defines binary assertions. A grader evaluates them against the complete trace of what the caller and agent said and what the agent did, including tool invocations, arguments, and returned rows. This catches an agent that claims to have changed a card without updating the database, as well as one that makes the right database change while disclosing protected information. This first benchmark version focuses on financial services. Its evaluation protocol supports an evolving leaderboard: additional voice agents can be evaluated on the same version, while later versions can expand the tasks and scenarios.
Chinese Translation
生产语音代理涵盖级联、语音到语音和混合架构。语音代理基准通常测量组件质量和对话特性，例如词错误率、延迟、自然性和轮换。较少有基准测量代理是否能够独立正确处理电话呼叫。呼叫中心将此称为“自我解决率”（containment）：自动系统在不转交给人类的情况下解决的电话呼叫比例。在某些电话呼叫中，正确的结果是拒绝或重定向。为了解决这一空白，我们引入了VAmoS Bench，语音代理模拟基准。它在一个有状态的客户支持任务中端到端地测量完整的语音代理系统。代理是Riley，一位虚构银行的信用卡支持代表，能够冻结、取消、替换或激活卡片。每个场景提供给模拟呼叫者一个私人目标和一个种子PostgreSQL后端。该平台利用每个场景填充并激活一个隔离的模拟环境，在该环境中，呼叫者通过音频联系Riley；大约三分之一的场景施加对抗压力。代理可以使用五个工具，这些工具对后端执行真实的SQL。每个场景还定义了二元断言。评分者根据呼叫者和代理所说的内容以及代理所做的操作（包括工具调用、参数和返回的行）对其进行评估。这可以捕捉到声称已更改卡片但未更新数据库的代理，以及在披露受保护信息的同时进行正确数据库更改的代理。这个第一个基准版本专注于金融服务。其评估协议支持一个不断发展的排行榜：额外的语音代理可以在同一版本上进行评估，而后续版本可以扩展任务和场景。
cs.AI / 14 / 2607.27484
Skill Use or Skill Theater? Evaluating the Reasoning Backroom in Skill-Augmented Language Agents
技能使用还是技能表演？评估技能增强语言代理中的推理后台
Abstract
Reusable skills are becoming a standard interface for extending language agents with task procedures. Yet evaluators usually infer skill use from visible reasoning or the agent's own attribution. These signals show what the agent appears to use, not whether the skill changed its decision. We ask whether skill-augmented agents exhibit a \textbf{Reasoning Backroom}, a systematic gap between stated skill use and intervention-measured influence. We introduce BACKTRACE, an evaluation framework that pairs each skill-conditioned answer with a matched no-skill counterfactual, intervenes on skill meaning, wording, identity, content, and assignment, and elicits attribution only after the answer is committed. We instantiate the framework as BACKROOMBench, a verified testbed spanning controlled logic and competition mathematics, multiple skill conditions, single-agent and multi-agent settings, and diverse model families. Our evaluation reveals a pervasive provenance failure. Across models and domains, stated skill use often remains stable while causal reliance and signed utility vary, producing both silent uptake and performative use. Behavioral effects follow procedural content more reliably than displayed skill identity, whereas stated attributions respond strongly to artifact availability. Observational detectors based on direct skill-use claims, text mentions, trace similarity, and an LLM judge do not identify which decisions actually depend on the skill. In multi-agent systems, skill influence can survive communication even after its source is lost, while no-skill teams still name skills and sources that were never supplied. These findings establish the Reasoning Backroom as a general AI provenance problem whose audit requires intervention.
Chinese Translation
可重用技能正成为扩展语言代理以执行任务程序的标准接口。然而，评估者通常从可见的推理或代理自身的归因中推断技能使用。这些信号显示了代理似乎使用的内容，而不是技能是否改变了其决策。我们探讨技能增强代理是否表现出 extbf{推理后台}，即声明的技能使用与干预测量的影响之间的系统性差距。我们引入了BACKTRACE，一个评估框架，将每个技能条件的回答与匹配的无技能反事实配对，干预技能的含义、措辞、身份、内容和分配，并在回答提交后才引导归因。我们将该框架实例化为BACKROOMBench，一个经过验证的测试平台，涵盖了受控逻辑和竞争数学、多种技能条件、单代理和多代理设置以及多样的模型家族。我们的评估揭示了普遍的来源失败。在不同模型和领域中，声明的技能使用往往保持稳定，而因果依赖和签名效用则有所变化，产生了静默采纳和表演性使用。行为效应更可靠地遵循程序内容，而不是显示的技能身份，而声明的归因则对工件的可用性反应强烈。基于直接技能使用声明、文本提及、轨迹相似性和大型语言模型（LLM）评判的观察检测器无法识别哪些决策实际上依赖于该技能。在多代理系统中，技能影响可以在失去其来源后仍然存在，而无技能团队仍然会提及从未提供的技能和来源。这些发现确立了推理后台作为一个普遍的人工智能来源问题，其审计需要干预。
cs.AI / 15 / 2607.27487
INCLAIR: Inception-Based Longitudinal Clinical Anomaly Detection with Informed Reasoning
INCLAIR：基于Inception的纵向临床异常检测与知情推理
Abstract
Detecting anomalies in longitudinal clinical profiles is clinically important but difficult: abnormal evidence is often sparse, patient histories have unequal length, and expert explanations are costly. We propose INCLAIR, a framework that scores each observation against multiple historical contexts, aggregates evidence at the profile level, and generates grounded natural-language explanations under limited expert supervision. Under stated within-profile exchangeability assumptions, the complete mean subsequence score takes an order-$l$ U-statistic form, yielding a variance decomposition and an incomplete-subset approximation that controls combinatorial inference cost independently of profile length. The same analysis shows that mean aggregation attenuates localized anomalies by a factor set by the anomaly support and profile length, motivating validation-selected top-$k$ pooling. Across three clinical datasets, INCLAIR consistently outperforms state-of-the-art baselines. We further validate practical relevance through a case study on longitudinal steroid profiles, comparing INCLAIR's predictions and explanations against domain-expert assessments supported by DNA analysis. The results show that INCLAIR enables clinically actionable anomaly detection under limited expert supervision.
Chinese Translation
在纵向临床档案中检测异常具有重要的临床意义，但也面临挑战：异常证据通常稀疏，患者历史记录长度不一，专家解释成本高昂。我们提出了INCLAIR，一个框架，它对每个观察值在多个历史背景下进行评分，在档案层面聚合证据，并在有限的专家监督下生成基于事实的自然语言解释。在假设档案内可交换性条件下，完整的均值子序列评分呈现为阶数为$l$的U统计量形式，从而得出方差分解和不完全子集近似，这使得组合推理成本独立于档案长度。相同的分析表明，均值聚合通过异常支持和档案长度设定的因子减弱了局部异常，促使验证选择的前$k$池化。在三个临床数据集上，INCLAIR始终优于最先进的基线。我们还通过对纵向类固醇档案的案例研究验证了其实用相关性，将INCLAIR的预测和解释与由DNA分析支持的领域专家评估进行比较。结果表明，INCLAIR在有限的专家监督下能够实现临床可操作的异常检测。
cs.AI / 16 / 2607.27490
MedLLM: An Open Medical Language Model at the Sub-Billion Scale
MedLLM：一个开放的亚十亿规模医学语言模型
Abstract
Open medical language models have converged on a single scale: every widely used system runs at 7B parameters or more, leaving the sub-billion regime uncharacterized. We present MedLLM, an open 0.1B-parameter medical language model trained through a fully open three-phase pipeline: general pretraining with curriculum sequence-length scheduling, domain fine-tuning on MedFineWeb, a reference-guided medical corpus we release that is selected from general web data by embedding similarity to medical question-answering (QA) data, and preference-aligned fine-tuning combining SFT with direct preference optimization (DPO). Across medical benchmarks, MedLLM shows a pattern visible only at sub-billion scale: medical competence does not degrade uniformly under compression but splits by task type. On context-grounded QA it comes within $2.9$pp of a medically adapted 7B model and surpasses the instruction-tuned and general-purpose 7B baselines; on knowledge-recall QA it stays near the task floor on clinical-vignette MedQA yet significantly exceeds every 7B and sub-7B baseline on MedMCQA, indicating that where recall fails the constraint is model capacity rather than adaptation. This dissociation is masked at 7B, where both capabilities are present, and surfaces only when capacity is scarce.
Chinese Translation
开放的医学语言模型已经趋向于一个单一规模：每个广泛使用的系统都在70亿参数或以上，导致亚十亿规模尚未被充分描述。我们提出了MedLLM，一个开放的0.1亿参数医学语言模型，通过一个完全开放的三阶段流程进行训练：首先是带有课程序列长度调度的一般预训练，其次是在MedFineWeb上的领域微调，MedFineWeb是我们发布的一个参考引导医学语料库，该语料库是通过与医学问答（QA）数据的嵌入相似性从一般网络数据中选择的，最后是结合了直接偏好优化（DPO）的监督微调（SFT）的偏好对齐微调。在医学基准测试中，MedLLM显示出一种仅在亚十亿规模下可见的模式：医学能力在压缩下并不均匀下降，而是按任务类型分裂。在基于上下文的问答任务中，它与一个医学适应的70亿模型相差仅$2.9$个百分点，并超越了指令调优和通用70亿基线；在知识回忆问答任务中，它在临床案例MedQA上接近任务底线，但在MedMCQA上显著超过每个70亿和亚70亿基线，表明在回忆失败的情况下，限制因素是模型容量而非适应性。这种解离在70亿模型中被掩盖，因为这两种能力同时存在，只有在容量稀缺时才会显现出来。
cs.AI / 17 / 2607.27499
A dataset of rated conceptual arguments
一个评分的概念论据数据集
Abstract
Large language models have improved rapidly on tasks with verifiable answers, such as mathematics and programming. Much less is known about their ability to reason about what we call conceptual questions: questions for which no ground truth is realistically accessible and no widely accepted resolution methodology exists, but on which progress can still be made by debating arguments. Most philosophical questions are of this kind, as are central components of questions in AI safety, decision theory, and social choice. Our approach is based on the view that while bottom-line conclusions on such questions are hard to evaluate, individual contextualized arguments can be evaluated far more reliably. We therefore introduce a dataset of 951 argumentative critiques of 442 position texts, spanning topics from AI safety and decision theory to ethics and politics, with 1,458 ratings by six expert raters along dimensions including centrality, strength, correctness, and clarity. We propose two scoring functions and benchmark a range of models. Performance tracks general capability rankings.
Chinese Translation
大型语言模型在可验证答案的任务上，如数学和编程，取得了快速进展。然而，对于我们所称的概念性问题的推理能力仍知之甚少：这些问题没有现实可得的真相，也没有广泛接受的解决方法，但通过辩论论据仍然可以取得进展。大多数哲学问题都属于这一类，人工智能安全、决策理论和社会选择中的核心问题也是如此。我们的方法基于这样一种观点：尽管对这些问题的最终结论难以评估，但个别的情境化论据可以更可靠地进行评估。因此，我们引入了一个包含951个论证性批评和442个立场文本的数据集，涵盖了从人工智能安全和决策理论到伦理和政治等主题，包含由六位专家评审者在中心性、强度、正确性和清晰度等维度上进行的1,458个评分。我们提出了两个评分函数，并对一系列模型进行了基准测试。性能与一般能力排名相一致。
cs.AI / 18 / 2607.27518
Automated Transcript Analysis for Detecting Flaws in Agentic Benchmarks
自动化成绩单分析以检测代理基准中的缺陷
Abstract
Capabilities of frontier models are often assessed using agentic benchmarks. To trust these results, benchmarks must accurately measure what they claim to and be free from invalidating flaws. Previous manual audits of benchmarks such as SWE-Bench-Verified have uncovered several validity issues in transcripts. However, manual review is difficult to scale, and it is unclear whether automated methods can reliably surface flaws that compromise benchmark validity. In this paper, we developed AI scanners to detect four types of validity issues: ground truth access, tool failure, guessing vulnerability, and answer format ambiguity. We produced grading rubrics for each to instruct human labeling, and evaluated the scanners against human labels on a held-out test set of Inspect Evals benchmarks. Our scanners identified several verified quality issues in five widely used benchmarks, including cases unlikely to be caught by random manual inspection. Not all cases were identified, and scanner performance varied substantially across benchmarks, criteria and models. We highlight several open challenges to be addressed to improve scanners for stronger quality assurance claims, including broader standardization gaps in the evaluation field that degrade scanner performance. Together, these results serve as a proof of concept for using automated transcript analysis to audit benchmark quality more broadly.
Chinese Translation
前沿模型的能力通常通过代理基准进行评估。为了信任这些结果，基准必须准确测量其所声称的内容，并且没有无效缺陷。之前对基准（如SWE-Bench-Verified）的手动审计发现了成绩单中的多个有效性问题。然而，手动审查难以扩展，并且尚不清楚自动化方法是否能够可靠地揭示妨碍基准有效性的缺陷。在本文中，我们开发了AI扫描器以检测四种类型的有效性问题：真实情况访问、工具故障、猜测脆弱性和答案格式模糊性。我们为每种问题制定了评分标准，以指导人工标注，并在一个保留的Inspect Evals基准测试集上评估了扫描器与人工标注的对比。我们的扫描器在五个广泛使用的基准中识别出多个经过验证的质量问题，包括一些不太可能通过随机手动检查发现的案例。并非所有案例都被识别，且扫描器的性能在基准、标准和模型之间差异显著。我们强调了若干开放性挑战，以改善扫描器以增强质量保证声明，包括评估领域中更广泛的标准化差距，这些差距降低了扫描器的性能。总的来说，这些结果为使用自动化成绩单分析更广泛地审计基准质量提供了概念证明。
cs.AI / 19 / 2607.27547
AI Literacy: An Exercise in Power-Knowledge
人工智能素养：权力-知识的实践
Abstract
As generative artificial intelligence becomes one of the most significant systems of knowledge production in our society today, questions relating to who can access and shape that production grow increasingly important in our discourse. This paper argues that the existing frameworks for AI literacy, which are dominated by technical competency and responsible-use principles, are insufficient because they enforce a "consumer" orientation toward AI rather than fostering genuine epistemic agency. Based upon Foucault's concept of power-knowledge, Freire's pedagogy of critical consciousness, and scholarship of digital literacy, this paper proposes a reconceptualization of AI literacy as a critical practice that equips individuals not just to use AI systems, but to critically evaluate them, resist their structuring assumptions, and participate in their governance. The paper further argues that unequal access to AI tools in society recapitulates longstanding epistemic injustices, and that a literacy framework oriented toward empowerment must account for these structural inequities. A three-part framework of AI literacy based on the notions of contextual use, critical interrogation, and participatory governance frames this literacy as a cultivation of epistemic "agents" rather than the training of competent consumers of AI-generated information.
Chinese Translation
随着生成性人工智能成为当今社会中最重要的知识生产系统之一，关于谁能够访问和塑造这种生产的问题在我们的讨论中变得愈发重要。本文认为，现有的人工智能素养框架主要以技术能力和负责任使用原则为主导，这种框架是不够的，因为它们强化了对人工智能的“消费者”导向，而没有促进真正的认知主体性。基于福柯的权力-知识概念、弗雷雷的批判意识教育学以及数字素养的相关研究，本文提出将人工智能素养重新概念化为一种批判性实践，使个体不仅能够使用人工智能系统，还能够对其进行批判性评估，抵制其结构性假设，并参与其治理。本文进一步论证，社会中对人工智能工具的不平等获取重申了长期存在的认知不公，而以赋权为导向的素养框架必须考虑这些结构性不平等。基于情境使用、批判性质疑和参与治理的三部分框架将这种素养视为认知“主体”的培养，而非仅仅是训练能够使用人工智能生成信息的合格消费者。
cs.AI / 20 / 2607.27553
Using Large Language Models for Idea Generation in Innovation
利用大型语言模型进行创新中的创意生成
Abstract
This research evaluates the efficacy of large language models (LLMs) in generating new product ideas. To do so, we compare three pools of ideas for new products targeted toward college students and priced at 50 dollars or less. The first pool of ideas was created by university students in a product design course before the availability of LLMs. The second and third pools of ideas were generated by GPT-4 from OpenAI using zero-shot and few-shot prompting, respectively. We evaluated idea quality using standard market research techniques to predict average purchase intent probability. We used text mining to assess idea similarity and human raters to evaluate idea novelty. We find that AI-generated ideas outperform human-generated ideas in terms of average purchase intent, with few-shot prompting yielding slightly higher intent than zero-shot prompting. However, AI-generated ideas are perceived as less novel and exhibit higher pairwise similarity, particularly with few-shot prompting, indicating a less diverse solution landscape. When focusing on the quality of the best ideas rather than the average ideas, we find that AI-generated ideas are seven times more likely to rank among the top 10 percent of ideas, demonstrating a significant advantage over human-generated ideas. We propose that this seven-to-one advantage is a conservative estimate because it does not account for the greater productivity of AI. Our findings suggest that despite some drawbacks, AI creativity presents a substantial benefit in generating high-quality ideas for new product development.
Chinese Translation
本研究评估了大型语言模型（LLMs）在生成新产品创意方面的有效性。为此，我们比较了三组针对大学生、价格在50美元或以下的新产品创意。第一组创意由大学生在大型语言模型可用之前的产品设计课程中创建。第二组和第三组创意分别由OpenAI的GPT-4通过零样本和少样本提示生成。我们使用标准市场研究技术评估创意质量，以预测平均购买意向概率。我们利用文本挖掘评估创意相似性，并通过人工评审者评估创意的新颖性。我们发现，AI生成的创意在平均购买意向方面优于人类生成的创意，而少样本提示的意向略高于零样本提示。然而，AI生成的创意被认为新颖性较低，并且在相似性上表现出更高的成对相似性，尤其是在少样本提示下，表明解决方案的多样性较低。当关注最佳创意的质量而非平均创意时，我们发现AI生成的创意在排名前10%的创意中出现的可能性是人类生成创意的七倍，显示出显著的优势。我们认为这一七比一的优势是保守估计，因为它没有考虑到AI更高的生产力。我们的研究结果表明，尽管存在一些缺陷，AI创造力在生成高质量新产品创意方面具有显著的优势。
cs.AI / 21 / 2607.27556
Evaluating Agentic Bioinformatics through Function, Evidence, and Validation
通过功能、证据和验证评估自主生物信息学
Abstract
Large language model agents increasingly plan, execute, and interpret biological analyses, yet fluent responses, successful tool calls, and benchmark performance alone do not establish scientific credibility. Existing reviews primarily organize biological agents by application, architecture, and agentic capability, but do not jointly operationalize the accountability of agent-generated workflows. We address this gap by treating the inspectable workflow trajectory, rather than architecture or final output alone, as the primary unit of analysis. We introduce the Function--Evidence--Validation (FEV) framework, which separates demonstrated workflow operations, traceable support for actions and claims, and use-case-specific validation. Using FEV, we map 109 agentic or agent-adjacent systems and 28 benchmark or evaluation resources, representing 128 unique publications across genomics, single-cell and spatial omics, protein science, drug discovery, computational pathology, and general bioinformatics automation. Across domains, planning and tool-mediated execution have advanced more rapidly than replayability, provenance, robust scientific assessment, external validation, and prospective empirical testing. We therefore argue that agentic bioinformatics should be assessed through workflow correctness rather than final-answer correctness alone. FEV provides a practical basis for comparing systems and designing transparent, auditable, and scientifically accountable bioinformatics workflows.
Chinese Translation
大型语言模型代理越来越多地规划、执行和解释生物分析，然而，仅仅流畅的响应、成功的工具调用和基准性能并不能建立科学可信度。现有的综述主要通过应用、架构和自主能力来组织生物代理，但并未共同操作化代理生成工作流的问责性。我们通过将可检查的工作流轨迹视为主要分析单元，而不仅仅是架构或最终输出，来填补这一空白。我们引入了功能-证据-验证（Function--Evidence--Validation, FEV）框架，该框架将已证明的工作流操作、可追溯的行动和主张支持以及特定用例的验证分开。使用FEV，我们映射了109个自主或相邻代理系统和28个基准或评估资源，代表了在基因组学、单细胞和空间组学、蛋白质科学、药物发现、计算病理学和一般生物信息学自动化等领域的128篇独特出版物。在各个领域中，规划和工具介导的执行进展速度快于可重放性、来源、稳健的科学评估、外部验证和前瞻性实证测试。因此，我们认为自主生物信息学应通过工作流的正确性而非仅仅最终答案的正确性进行评估。FEV为比较系统和设计透明、可审计及科学问责的生物信息学工作流提供了实用基础。
cs.AI / 22 / 2607.27562
DeepResearch Agent System
深度研究代理系统
Abstract
The DeepResearch Agent System is a large language model system engineered for deep information retrieval, multi-step reasoning, and autonomous research tasks. Built upon a sparse activation architecture with 30 billion total parameters of which only 3 billion are activated per token, the system achieves state-of-the-art performance on multiple agent search benchmarks while delivering 3.2 times faster inference compared to dense counterparts of equivalent scale. The system supports a 128K-token context window with hierarchical attention mechanisms that yield 18.7% accuracy and 23.4% recall improvements over standard long-context approaches. A dual-mode reasoning engine provides both a ReAct paradigm for basic multi-step problem solving and an IterResearch mode for high-performance iterative research with up to 20 reasoning steps, collectively delivering a 31.2% accuracy improvement over single-pass baselines. Multi-tool coordination integrates retrieval, computation, web search, and file parsing modules to achieve 92.1% tool-use accuracy. A reinforcement learning optimization framework based on the GRPO algorithm provides token-level policy gradients that improve training stability by 35% and accelerate convergence by 42%. An automated data synthesis pipeline with seed-based expansion achieves a 92.5% usability rate. Benchmark results include 87.3% on Humanity's Last Exam, 85.3% on BrowserComp Chinese, and 91.2% on WebWalkerQA. The system is fully open-sourced, including data synthesis, training, and inference code, and supports applications in academic research, business analysis, R&D support, and education.
Chinese Translation
深度研究代理系统是一个大型语言模型系统，旨在进行深度信息检索、多步骤推理和自主研究任务。该系统基于稀疏激活架构构建，拥有300亿个总参数，其中每个token仅激活30亿个参数，在多个代理搜索基准测试中实现了最先进的性能，同时与同规模的密集模型相比，推理速度快3.2倍。该系统支持128K-token的上下文窗口，并采用层次注意机制，相较于标准的长上下文方法，准确率提高了18.7%，召回率提高了23.4%。双模式推理引擎提供了ReAct范式用于基本的多步骤问题解决，以及IterResearch模式用于高性能的迭代研究，最多可进行20个推理步骤，整体上实现了31.2%的准确率提升，相较于单次通过的基线。多工具协调集成了检索、计算、网络搜索和文件解析模块，工具使用准确率达到92.1%。基于GRPO算法的强化学习优化框架提供了token级的策略梯度，使训练稳定性提高35%，收敛速度加快42%。基于种子扩展的自动数据合成管道实现了92.5%的可用性率。基准测试结果包括在人类最后考试中获得87.3%，在BrowserComp中文测试中获得85.3%，在WebWalkerQA中获得91.2%。该系统完全开源，包括数据合成、训练和推理代码，支持学术研究、商业分析、研发支持和教育等应用。
cs.AI / 23 / 2607.27578
What makes prompts a graph: necessary and sufficient conditions for prompt graph engineering
促使提示成为图的因素：提示图工程的必要和充分条件
Abstract
Prompts stopped being isolated strings some time ago. In real systems, one model call feeds another, retrieval interleaves with generation, routers branch, and aggregators merge parallel results. Practice converged on a single structure to hold this together: the graph. Frameworks such as LangGraph, DSPy, and Prompt Flow expose it openly, and research systems already optimize it automatically. The vocabulary, however, lags behind. Graph names, variously, a reasoning topology inside one sampling strategy, a multi-agent conversation, or an orchestration artifact, while prompt engineering still evokes writing one good string. What is missing is a reference definition treating prompts as nodes of an explicit, executable, improvable graph. We build that definition through conceptual analysis over sources with persistent identifiers, complemented by primary grey literature. We reconstruct the genealogy of the idea, from dataflow graphs and build systems, through prompt chaining and the thought topologies (chain, tree, graph), to graphs compiled and optimized as artifacts. We then propose a constitutive definition of prompt graph engineering, state its four conditions (explicit structure, separation between structure and prompt content, executable semantics, and the graph as a first-class engineering artifact), and operationalize them as an inclusion and exclusion test. We draw the boundary against six neighboring concepts and apply the test to six real systems (LangGraph, DSPy, Prompt Flow, AutoGen, CrewAI, and Claude Code subagents); it includes and excludes consistently. We close with a research agenda organized along four design tension axes. The contribution is an operational definition and a shared vocabulary for a practice that industry already exercises daily without naming precisely.
Chinese Translation
提示不再是孤立的字符串。 在实际系统中，一个模型调用会影响另一个，检索与生成交织，路由器分支，聚合器合并并行结果。 实践趋向于使用一种结构将这些元素结合在一起：图。 LangGraph、DSPy 和 Prompt Flow 等框架公开展示了这一点，而研究系统已经能够自动优化它。然而，相关词汇却滞后于此。图的名称各异，可能指的是一种采样策略中的推理拓扑、多智能体对话或编排工件，而提示工程仍然让人联想到写出一个好的字符串。缺失的是一种参考定义，将提示视为一个明确的、可执行的、可改进的图的节点。我们通过对具有持久标识符的来源进行概念分析，结合主要的灰色文献，构建了该定义。我们重构了这一思想的谱系，从数据流图和构建系统，到提示链和思维拓扑（链、树、图），再到作为工件编译和优化的图。然后，我们提出了提示图工程的构成定义，阐明了其四个条件（明确结构、结构与提示内容的分离、可执行语义，以及图作为一类工程工件），并将其操作化为包含和排除测试。我们与六个邻近概念划定了边界，并将测试应用于六个实际系统（LangGraph、DSPy、Prompt Flow、AutoGen、CrewAI 和 Claude Code 子代理）；结果一致地包含和排除。最后，我们以四个设计张力轴组织了研究议程。我们的贡献是一个操作性定义和一个共享词汇，用于一种行业每天都在实践但尚未精确定义的活动。
cs.AI / 24 / 2607.27579
From Minds to Models: The Intersection of Psychology and LLM Behaviours
从心智到模型：心理学与大型语言模型行为的交汇
Abstract
Large language models (LLMs) are often compared with the human mind because their decision-making is complex, non-linear and difficult to interpret. Psychological methods developed to investigate unobservable mental processes may therefore help examine LLM behaviour, particularly in government and healthcare. Building on prompt-based adaptations of the Implicit Association Test, this study tested whether ChatGPT produced sentiment differences across racial conditions in open-ended text. Fourteen base questions were crossed with eight racial categories and a race-agnostic control, producing 126 prompts. Each was submitted once to GPT-3.5T, GPT-4 and GPT-4T, yielding 378 responses. Sentiment scores were derived from categorical labels and source scores: positive labels retained the source score, negative labels were assigned its negative, and neutral responses were coded zero. A two-way ANOVA found a small main effect of racial condition, F(8, 351) = 2.04, p = .042, partial-eta squared = .044, but no effect of model, F(2, 351) = 0.07, p = .933, and no interaction, F(16, 351) = 0.23, p = .999. However, the effect was not retained in a rank-transformed sensitivity analysis, F(8, 351) = 1.53, p = .145, and Tukey-corrected comparisons found no significant pairwise differences. An uncorrected European-Indigenous Australian comparison was significant, but was selected post hoc and is reported only as hypothesis-generating. Evidence for sentiment differences was therefore weak and analysis-dependent. Sentiment scoring also cannot distinguish evaluative bias from the valence of historical content elicited by a prompt. We outline design changes needed to address these limitations and argue for interdisciplinary development of behavioural measures of model bias. Keywords: Implicit Bias, Psychological Research Methods, Artificial Intelligence, ChatGPT, Large Language Models, Sentiment Analysis
Chinese Translation
大型语言模型（LLMs）常常与人类思维进行比较，因为它们的决策过程复杂、非线性且难以解释。因此，旨在研究不可观察的心理过程的心理学方法可能有助于考察LLM行为，特别是在政府和医疗保健领域。本研究基于隐性联想测试（Implicit Association Test）的提示基础适应，测试了ChatGPT在开放式文本中是否在种族条件下产生情感差异。将十四个基础问题与八个种族类别及一个无种族控制条件交叉，产生了126个提示。每个提示均提交给GPT-3.5T、GPT-4和GPT-4T一次，共获得378个响应。情感评分来自分类标签和源评分：正面标签保留源评分，负面标签被赋予其负值，中性响应编码为零。双向方差分析（ANOVA）发现种族条件的主效应较小，F(8, 351) = 2.04，p = .042，部分η平方 = .044，但模型的效应不显著，F(2, 351) = 0.07，p = .933，且没有交互效应，F(16, 351) = 0.23，p = .999。然而，该效应在排名转换的敏感性分析中未能保留，F(8, 351) = 1.53，p = .145，Tukey校正比较未发现显著的成对差异。未经校正的欧洲-土著澳大利亚人比较显著，但是在事后选择的，仅作为生成假设的报告。因此，情感差异的证据较弱且依赖于分析。情感评分也无法区分评估偏见与由提示引发的历史内容的效价。我们概述了为解决这些局限性所需的设计变更，并主张跨学科发展模型偏见的行为测量。关键词：隐性偏见，心理研究方法，人工智能，ChatGPT，大型语言模型，情感分析
cs.AI / 25 / 2607.27598
Wiring diagram extraction and gluing: a case study in classifying figure skating jumps using 3D dataset
电路图提取与粘合：基于3D数据集的花样滑冰跳跃分类案例研究
Abstract
Hasse clustering is an algorithm that extracts common patterns in sequential data and represents them in graphical forms. As the number of expected clusters grows, however, the algorithm can become infeasible to run due to combinatorial complexity. In this article, we describe a theory of gluing wiring diagrams, allowing iterative applications of Hasse clustering to achieve the same result as a single application. We test our theory in the context of classifying videos of figure skating jumps.
Chinese Translation
Hasse聚类是一种提取序列数据中常见模式并以图形形式表示的算法。然而，随着预期聚类数量的增加，该算法由于组合复杂性可能变得难以运行。本文描述了一种电路图粘合理论，允许对Hasse聚类进行迭代应用，以实现与单次应用相同的结果。我们在花样滑冰跳跃视频分类的背景下测试了我们的理论。
cs.AI / 26 / 2607.27599
World Action Planner: Generalizable Decision-Making with Action-Conditioned World Models
世界行动规划器：具有动作条件的世界模型的可推广决策制定
Abstract
Building generalizable agents for diverse applications remains a fundamental challenge. While imitation learning-based policies succeed in specific training environments, they often fail to generalize to novel scenes and tasks. In this work, we propose World Action Planner, a robot planning system that leverages the reasoning capabilities of Vision-Language Models (VLMs) and the physical grounding of a multi-task pose-image conditioned world model. Our system enables an agent to propose initial action plans and iteratively refine them via optimization and search, reasoning over imagined world model rollouts. We demonstrate that our approach achieves superior performance across compositional tasks, new layouts, and zero-shot generalization scenarios, significantly outperforming state-of-the-art end-to-end policy models such as VLAs and WAMs. Project website at worldactionplanner.github.io
Chinese Translation
构建适用于多种应用的可推广智能体仍然是一个基本挑战。尽管基于模仿学习的策略在特定训练环境中取得成功，但它们往往无法推广到新场景和任务。在本研究中，我们提出了世界行动规划器（World Action Planner），这是一种机器人规划系统，利用了视觉-语言模型（Vision-Language Models, VLMs）的推理能力和多任务姿态-图像条件的世界模型的物理基础。我们的系统使智能体能够提出初步行动计划，并通过优化和搜索迭代地对其进行细化，推理想象中的世界模型展开。我们展示了我们的方法在组合任务、新布局和零样本推广场景中表现出色，显著优于最新的端到端策略模型，如VLA和WAM。项目网站为worldactionplanner.github.io
cs.AI / 27 / 2607.27615
CORE: In-Context Reconstruction for Unified Tabular Anomaly Detection
CORE：统一表格异常检测的上下文重构
Abstract
Tabular anomaly detection (TAD), which focuses on identifying abnormal samples that deviate from the majority in tabular data, has received growing attention. Recently, there has been an emerging trend towards unified TAD, which seeks to detect anomalies across different datasets using a single generalizable model. In unified TAD, aligning heterogeneous data remains challenging. While existing methods often rely on distance-based unified feature construction, they may obscure the semantics of the original features. Moreover, existing approaches typically formulate anomaly detection as a binary classification task, which may overlook diverse anomaly patterns from various datasets and be misled by unrepresentative synthetic anomalies. To address these challenges, we propose an in-COntext REconstruction approach for unified TAD (CORE for short). It introduces a decorrelated feature alignment module to directly align heterogeneous features into a unified representation space, which retains their semantic information. Meanwhile, CORE formulates unified TAD as an in-context reconstruction problem, eliminating the need for labeled or synthesized anomalies. Specifically, the in-context reconstruction module reconstructs each sample by leveraging contextual normal samples to capture dataset-specific distributions, such that reconstruction errors reflect its deviation from normality, facilitating unified TAD on arbitrary unseen datasets.
Chinese Translation
表格异常检测（TAD）关注于识别在表格数据中偏离大多数的异常样本，近年来受到了越来越多的关注。最近，统一TAD的趋势逐渐兴起，旨在使用单一的可泛化模型在不同数据集上检测异常。在统一TAD中，对异构数据的对齐仍然是一个挑战。现有方法通常依赖于基于距离的统一特征构建，但这可能会模糊原始特征的语义。此外，现有方法通常将异常检测表述为二分类任务，这可能忽视来自不同数据集的多样化异常模式，并可能受到不具代表性的合成异常的误导。为了解决这些挑战，我们提出了一种用于统一TAD的上下文重构方法（简称CORE）。该方法引入了一种去相关特征对齐模块，直接将异构特征对齐到统一的表示空间，从而保留其语义信息。同时，CORE将统一TAD表述为一个上下文重构问题，消除了对标记或合成异常的需求。具体而言，上下文重构模块通过利用上下文正常样本重构每个样本，以捕捉数据集特定的分布，使得重构误差反映其偏离正常状态的程度，从而促进在任意未见数据集上的统一TAD。
cs.AI / 28 / 2607.27617
Hidden APIs in Language Models: Discovering Reusable Causal Interfaces from Forked Futures
语言模型中的隐藏API：从分叉未来中发现可重用的因果接口
Abstract
Identical language-model answers can arise from hidden states that support different future computations, so current-answer probes do not establish a reusable internal interface. We introduce forked futures: future operations are sampled only after a prefix state has formed, and states are compared through the response distributions induced by those operations. This yields an empirical causal quotient over hidden states without requiring researcher-specified latent labels. Shared, Local, Mixture, and Distributed interfaces then compete under prequential causal description length subject to future-signature fidelity and matched capacity constraints. In the two detailed model evaluations, Shared has the lowest held-out description length, with gains of 0.216 nats on Qwen2.5-1.5B and 0.294 nats on Llama-3-8B, while maintaining tightly clustered mean future-signature distortion; a five-backbone sweep preserves the positive direction of Sharedness Gain. The figure-aligned transplantation analysis gives Shared the strongest joint target-correctness, locality, copy-preservation, and composite profile, and API-aligned paths mediate 0.749 of the target effect versus 0.150 for matched null paths. In the blind four-class model-organism test, 14/16 architectures are recovered, with one observed non-Shared to Shared error among 12 non-Shared organisms. These results support an economical reusable causal interface within the tested operation banks, while keeping the claim explicitly conditional on the candidate architectures, interventions, and held-out futures.
Chinese Translation
相同的语言模型答案可以源自支持不同未来计算的隐藏状态，因此当前答案探测并不能建立可重用的内部接口。我们引入了分叉未来（forked futures）：未来操作仅在前缀状态形成后进行采样，并通过这些操作引发的响应分布比较状态。这在不需要研究者指定潜在标签的情况下，产生了隐藏状态的经验因果商。共享（Shared）、局部（Local）、混合（Mixture）和分布式（Distributed）接口在预先序列化的因果描述长度下进行竞争，受未来签名保真度和匹配容量约束的影响。在两个详细的模型评估中，共享接口具有最低的保留描述长度，在Qwen2.5-1.5B上获得0.216 nats的增益，在Llama-3-8B上获得0.294 nats的增益，同时保持紧密聚集的平均未来签名失真；五个主干的扫查保持了共享增益的正向方向。图形对齐的移植分析使共享接口在目标正确性、局部性、复制保留和复合特征方面表现出最强的联合性能，而API对齐路径在目标效应中介导了0.749的效果，相比之下，匹配的无效路径仅为0.150。在盲四类模型生物体测试中，14/16种架构被恢复，在12个非共享生物体中观察到一个非共享到共享的错误。这些结果支持在测试的操作库中建立经济的可重用因果接口，同时明确声明该主张依赖于候选架构、干预和保留的未来。
cs.AI / 29 / 2607.27630
SCOPE: Synthetic Conditional Objectives for Policy Evolution in Black-Box Combinatorial Optimization
SCOPE：用于黑箱组合优化中策略演化的合成条件目标
Abstract
Black-box combinatorial optimization requires systematically identifying high-quality solutions under a limited evaluation budget, yet the unknown objective function provides little guidance for deciding where the search should explore next. We introduce SCOPE, a general framework for Synthetic Conditional Objectives for Policy Evolution in Black-Box Combinatorial Optimization. Rather than directly optimizing the inaccessible objective, SCOPE learns a set of synthetic objectives conditioned on the accumulated search history, where each objective is designed to expose a distinct and potentially useful preference over candidate solutions. These objectives are then used to evolve search policies that generate diverse candidates, whose true quality is subsequently assessed through black-box evaluations. The outer loop adaptively updates and selects synthetic objectives according to how effectively their induced policies discover promising regions. In contrast, the inner loop returns a portfolio of top-performing policies to reduce the risk of relying on a single surrogate preference. This formulation reframes objective design as a mechanism for guiding policy exploration, enabling the search process to exploit observed evidence while maintaining structured diversity across discrete solution spaces. Extensive experiments across multiple benchmark problems demonstrate that SCOPE consistently improves black-box search performance under limited evaluation budgets and generalizes well across diverse combinatorial structures.
Chinese Translation
黑箱组合优化需要在有限的评估预算下系统地识别高质量解决方案，但未知的目标函数对决定搜索的下一步探索方向提供的指导有限。我们提出了SCOPE，一个用于黑箱组合优化中策略演化的合成条件目标的通用框架。SCOPE并不是直接优化不可访问的目标，而是学习一组基于累积搜索历史的合成目标，其中每个目标旨在揭示候选解决方案之间独特且潜在有用的偏好。这些目标随后用于演化搜索策略，以生成多样化的候选者，其真实质量随后通过黑箱评估进行评估。外部循环根据合成目标所诱导的策略发现有前景区域的有效性，自适应地更新和选择合成目标。相反，内部循环返回一组表现最佳的策略，以降低依赖单一代理偏好的风险。这一形式化将目标设计重新构建为引导策略探索的机制，使搜索过程能够利用观察到的证据，同时在离散解决方案空间中保持结构化的多样性。在多个基准问题上的广泛实验表明，SCOPE在有限评估预算下持续改善黑箱搜索性能，并在多样化组合结构中表现出良好的泛化能力。
cs.AI / 30 / 2607.27631
ReDiPPO: Reference-Guided Value Calibration and Discrepancy-Aware Token Reweighting for Mathematical Reasoning
ReDiPPO：基于参考的价值校准与关注差异的令牌重加权用于数学推理
Abstract
Reinforcement learning has emerged as an effective paradigm for enhancing the mathematical reasoning capabilities of large language models. Among existing policy optimization methods, Proximal Policy Optimization (PPO) remains particularly appealing because its learned critic can, in principle, provide token-level credit assignment. However, in mathematical reasoning tasks characterized by long reasoning horizons and sparse outcome rewards, reliable token-level credit assignment remains challenging. The standard critic often fails to accurately evaluate intermediate reasoning states, resulting in noisy advantage estimates and suboptimal policy updates. In this paper, we propose ReDiPPO, a Reference-guided and Discrepancy-aware PPO framework for mathematical reasoning. ReDiPPO introduces a reference-guided critic that uses reference answers as training-time privileged signals to provide more accurate value estimation. Meanwhile, it retains a standard critic and quantifies the token-level reference-standard discrepancy between the standard value estimate and the reference-guided value estimate. This discrepancy serves as an indicator of difficult reasoning states and is used to reweight the corresponding token-level advantages during PPO optimization. Extensive experiments on diverse mathematical reasoning benchmarks demonstrate that ReDiPPO improves value-estimation accuracy and consistently outperforms strong policy optimization baselines, including PPO, DAPO, and GSPO, in final reasoning performance. Our code is available on https://github.com/cii030/ReDiPPO.
Chinese Translation
强化学习已成为提升大型语言模型数学推理能力的有效范式。在现有的策略优化方法中，近端策略优化（Proximal Policy Optimization, PPO）因其学习到的评论者原则上能够提供令牌级别的信用分配而特别引人注目。然而，在以长推理时间和稀疏结果奖励为特征的数学推理任务中，可靠的令牌级别信用分配仍然具有挑战性。标准评论者往往无法准确评估中间推理状态，导致噪声优势估计和次优策略更新。在本文中，我们提出了ReDiPPO，一个用于数学推理的基于参考和关注差异的PPO框架。ReDiPPO引入了一个基于参考的评论者，利用参考答案作为训练时的特权信号，以提供更准确的价值估计。同时，它保留了一个标准评论者，并量化标准价值估计与基于参考的价值估计之间的令牌级别参考标准差异。这种差异作为困难推理状态的指示器，用于在PPO优化过程中重新加权相应的令牌级别优势。在多样化的数学推理基准上进行的广泛实验表明，ReDiPPO提高了价值估计的准确性，并在最终推理性能上持续超越强大的策略优化基线，包括PPO、DAPO和GSPO。我们的代码可在 https://github.com/cii030/ReDiPPO 获取。
cs.AI / 31 / 2607.27635
HealthCAT: An Interpretable Encoder-only Transformer Framework for Health Indicator Prediction and Temporal Interpretation of Wearable Sensor Data
HealthCAT：一种可解释的仅编码器变换器框架，用于健康指标预测和可穿戴传感器数据的时间解释
Abstract
Wearable sensors continuously capture fine-grained multivariate time-series data, providing opportunities to model behavioural patterns associated with health outcomes. However, existing deep learning methods prioritise predictive accuracy over interpretability, limiting their application in health research. In this study, we present HealthCAT, a flexible framework that integrates an Encoder-only Transformer with an Attentive Class Activation Token (AttentiveCAT) to generate class-specific, time-step-level interpretations. These interpretations can be mapped back onto behavioural cycles that are relevant to the domain (e.g., time-of-day), supporting individual-level analysis of wearable sensor data. We evaluated HealthCAT using two real-world wearable sensor datasets (306 participants in total). HealthCAT outperformed deep learning baselines by up to 17\% in F1-score and 12\% in accuracy on both datasets ($p<0.05$). In masking experiments, the time steps identified by HealthCAT carried significantly more predictive value than random selection across all masking conditions ($p<0.05$), indicating that the identified time steps are predictively informative. By coupling predictive performance with validated time-step-level interpretability, HealthCAT moves wearable sensor analysis beyond aggregated metrics towards temporal patterns that support health monitoring, behavioural pattern analysis, and intervention design in health research. The significance of this work is that it enables accurate prediction of health indicators from wearable sensor data while providing insights into when and how physical activity patterns occur, rather than relying solely on aggregated summary measures.
Chinese Translation
可穿戴传感器持续捕捉细粒度的多变量时间序列数据，为建模与健康结果相关的行为模式提供了机会。然而，现有的深度学习方法优先考虑预测准确性而非可解释性，限制了其在健康研究中的应用。在本研究中，我们提出了HealthCAT，这是一种灵活的框架，将仅编码器变换器与注意力类激活标记（Attentive Class Activation Token，AttentiveCAT）相结合，以生成特定类别的时间步级别解释。这些解释可以映射回与领域相关的行为周期（例如，时间段），支持对可穿戴传感器数据的个体级分析。我们使用两个真实的可穿戴传感器数据集（共306名参与者）对HealthCAT进行了评估。在这两个数据集上，HealthCAT在F1分数上比深度学习基线提高了最多17%，在准确率上提高了最多12%（$p<0.05$）。在掩蔽实验中，HealthCAT识别的时间步在所有掩蔽条件下的预测价值显著高于随机选择（$p<0.05$），表明识别的时间步具有预测信息。通过将预测性能与经过验证的时间步级别可解释性相结合，HealthCAT使可穿戴传感器分析超越了汇总指标，朝向支持健康监测、行为模式分析和健康研究中的干预设计的时间模式发展。这项工作的意义在于，它能够从可穿戴传感器数据中准确预测健康指标，同时提供关于身体活动模式何时以及如何发生的见解，而不仅仅依赖于汇总的总结性指标。
cs.AI / 32 / 2607.27636
HALO: Heterogeneous Admission through Localized Obligations for Safe Agentic Execution
HALO：通过局部义务实现异构接纳以安全执行代理任务
Abstract
Recent agentic AI systems may return a heterogeneous response containing notices, requests, handoffs, and actions. Conditions can change before external use, so components from the same response need not remain supported together. Rejecting the whole response discards useful components, whereas checking components independently can leave a dependent without its prerequisite. We present Heterogeneous Admission with Localized Obligations (HALO), a runtime protocol that preserves supported components whose declared prerequisites also remain supported, rechecks each exact action before dispatch, and allows blocked actions to be replaced only by fresh candidates. HALO matched all 96 admission expectations and passed all 20 protocol tests. In structured-response replay, it retained 248/248 supported components, including 128/128 unaffected by unrelated changes, while a whole-response policy retained 0/248. Across ten cold-start PX4/Gazebo sessions, HALO blocked every tested stale route, observed no matching stale setpoint, and completed all fresh recoveries.
Chinese Translation
近期的代理人工智能系统可能会返回包含通知、请求、交接和动作的异构响应。在外部使用之前，条件可能会发生变化，因此同一响应中的组件不必始终保持支持。拒绝整个响应会丢弃有用的组件，而独立检查组件可能会导致依赖项失去其前提条件。我们提出了异构接纳与局部义务（HALO），这是一种运行时协议，能够保留那些声明的前提条件也仍然得到支持的组件，在派发之前重新检查每个确切的动作，并且仅允许用新候选者替换被阻塞的动作。HALO 满足了所有 96 项接纳期望，并通过了所有 20 项协议测试。在结构化响应重放中，它保留了 248/248 个支持组件，包括 128/128 个未受无关变化影响的组件，而整个响应策略则保留了 0/248。在十个冷启动的 PX4/Gazebo 会话中，HALO 阻止了每个测试的过时路径，没有观察到匹配的过时设定点，并完成了所有新的恢复。
cs.AI / 33 / 2607.27686
Evaluating and Pricing Advertisements in AI-Generated Responses
评估和定价AI生成响应中的广告
Abstract
As search increasingly shifts toward LLM-driven answer engines, advertising is becoming embedded within the generated response itself and should therefore be evaluated for both user utility and commercial value. The key challenge is click-through intent: behavioural logs are unavailable, human annotation resists calibration, and frontier LLM judges conflate intent with linguistic fluency. These gaps compound, as principled pricing presupposes a continuous intent signal, while generating such a signal presupposes supervision that is currently unavailable. We construct the missing supervision through a psychologically grounded agent simulation framework, and distil it into a parameter-efficient evaluator that predicts click-through intent, together with the three companion dimensions of ad quality, as smooth, differentiable estimates. Validated through sign-certain behavioural perturbations, the evaluator surpasses frontier zero-shot judges on relevance sensitivity (79% versus 60-67%), tracks graded content degradation, generalises without error to 103 fictional products, and agrees with human preference in 86% of pairwise judgements across five annotators, with agreement rising in the evaluator's confidence. Upon its estimates we build the pricing layer directly, deriving the unique payment rule under which truthful bidding is optimal, demonstrating it on a best-of-k allocation, and extending the mechanism to non-monotone allocations. The same differentiable signal stands ready as a training objective for ad generation.
Chinese Translation
随着搜索逐渐向大型语言模型（LLM）驱动的答案引擎转变，广告正逐渐嵌入到生成的响应中，因此应评估其用户效用和商业价值。关键挑战在于点击意图：行为日志不可用，人类注释难以校准，而前沿的LLM评判者将意图与语言流畅性混为一谈。这些差距相互叠加，因为原则性的定价假设存在一个连续的意图信号，而生成这样的信号又假设有当前不可用的监督。我们通过一个心理学基础的代理模拟框架构建缺失的监督，并将其提炼为一个参数高效的评估器，该评估器预测点击意图，以及广告质量的三个伴随维度，作为平滑、可微的估计。通过显著的行为扰动进行验证，该评估器在相关性敏感性上超越了前沿的零样本评判者（79%对比60-67%），能够跟踪分级内容退化，无误地推广到103个虚构产品，并在五位注释者的配对判断中与人类偏好达成86%的一致，且在评估器的信心上升时一致性增强。基于其估计，我们直接构建定价层，推导出真实竞标最优的唯一支付规则，在最佳k分配上进行演示，并将机制扩展到非单调分配。相同的可微信号随时可作为广告生成的训练目标。
cs.AI / 34 / 2607.27687
Rehearse: Stepping Back from the Confidence Cliff in Self-Improving Autoresearch
Rehearse：从自我改进自研究中的信心悬崖中退后一步
Abstract
Autoresearch improves machine-learning code by proposing changes, running full training jobs, and keeping changes that improve the metric. The efficiency of this loop depends not only on generating ideas, but also on the agent's ability to decide, before spending a training run, whether a proposed modification is likely to work. We study how the reliability of this pre-execution judgment changes over the course of an autoresearch trajectory. In public AutoSOTA logs (Li et al., 2026; Tsinghua FIB Lab, 2026), the fraction of helpful modifications falls from 70% in the first two iterations to 43% by iteration 6+. On 296 same-baseline modification pairs from 39 paper-derived AutoSOTA tasks, each containing one modification that improved the metric and one that did not, with measured outcomes hidden, an LLM judge given candidate rationales but no prior-attempt history reaches 79.5% accuracy on the pairs where strict consensus returns a verdict. On the full 366-pair benchmark, however, this ability weakens substantially late in the loop. As successful changes accumulate, selective accuracy - accuracy conditioned on a strict-consensus verdict - falls from 82.8% to 56.9%, while the judge remains willing to decide. We call this operational pattern the confidence cliff. Rehearse implements the loop change as a lightweight skill for autoresearch loops: propose several ideas, compare them before execution, run the most promising, and judge with a focused memory of similar past attempts and outcomes. This focused outcome memory raises late selective accuracy to 83.5%. Across 4,000 budgeted training runs over three loops, Rehearse improves the endpoint under the same training-run budget on nanochat, image classification, and time-series forecasting.
Chinese Translation
自研究通过提出更改、运行完整的训练任务以及保留提高指标的更改来改善机器学习代码。这个循环的效率不仅依赖于生成创意，还依赖于代理在花费训练时间之前判断提议的修改是否可能有效的能力。我们研究了这种执行前判断的可靠性在自研究轨迹中的变化。在公共的 AutoSOTA 日志中（Li et al., 2026; Tsinghua FIB Lab, 2026），有用修改的比例从前两次迭代的 70% 下降到第六次迭代及以后为 43%。在来自 39 个论文衍生的 AutoSOTA 任务的 296 对相同基线修改中，每对包含一个提高指标的修改和一个未提高的修改，且测量结果被隐藏，给定候选理由但没有先前尝试历史的 LLM 判决者在严格共识返回裁决的对中达到了 79.5% 的准确率。然而，在完整的 366 对基准测试中，这种能力在循环后期显著减弱。随着成功更改的积累，选择性准确率（基于严格共识裁决的准确率）从 82.8% 下降到 56.9%，而判决者仍然愿意做出决定。我们将这种操作模式称为信心悬崖。Rehearse 将循环更改实现为自研究循环的一种轻量级技能：提出多个想法，在执行前进行比较，运行最有前景的想法，并结合对类似过去尝试和结果的集中记忆进行判断。这种集中结果记忆将后期选择性准确率提高到 83.5%。在三个循环中进行的 4,000 次预算训练中，Rehearse 在相同的训练运行预算下改善了 nanochat、图像分类和时间序列预测的最终结果。
cs.AI / 35 / 2607.27698
Guiding Large Language Models with Genetic Programming-Evolved Heuristic Knowledge for Dynamic Multi-Mode Project Scheduling
利用遗传编程演化的启发式知识引导大型语言模型进行动态多模式项目调度
Abstract
In dynamic multi-mode project scheduling, activities have alternative execution modes and uncertain durations, while precedence relations and limited resources constrain their execution. Heuristic priority rules support fast online decisions, but their design requires substantial domain expertise. Genetic programming (GP) hyper-heuristics can automatically evolve such rules. Large language models (LLMs), meanwhile, provide a flexible interface for interpreting scheduling information and explaining decisions. However, zero-shot LLM decisions may lack domain knowledge, consume many tokens, and vary across repeated queries. GP-evolved rules therefore provide a potential source of scheduling knowledge for guiding LLM decisions. Unlike existing LLM--GP hybrids that use LLMs to support heuristic evolution, we transfer knowledge in the reverse direction, using knowledge extracted from high-quality GP rules to guide an online LLM decision maker. We extract knowledge from high-quality GP rules and inject it through Feature Selection, Feature Hint, Rule Reference, and Rule Follow. These mechanisms are evaluated in terms of scheduling performance, token consumption, decision stability, and the feature focus expressed in generated rationales. GP-derived guidance generally improves the unguided LLM, but its representation matters. Simplifying the decision context or supplying explicit decision logic is more effective than highlighting important features. Feature Selection offers the best token efficiency, whereas Rule Follow achieves strong performance at greater token cost. Guidance also improves decision stability and changes the features expressed in generated rationales.
Chinese Translation
在动态多模式项目调度中，活动具有可替代的执行模式和不确定的持续时间，而优先关系和有限资源限制了它们的执行。启发式优先规则支持快速的在线决策，但其设计需要大量的领域专业知识。遗传编程（Genetic Programming, GP）超启发式方法可以自动演化这些规则。同时，大型语言模型（Large Language Models, LLMs）提供了一个灵活的接口，用于解释调度信息和解释决策。然而，零-shot LLM 决策可能缺乏领域知识，消耗大量的令牌，并且在重复查询中可能存在差异。因此，GP演化的规则为引导LLM决策提供了潜在的调度知识来源。与现有的使用LLM支持启发式演化的LLM-GP混合模型不同，我们采用反向知识转移，利用从高质量GP规则中提取的知识来引导在线LLM决策者。我们通过特征选择（Feature Selection）、特征提示（Feature Hint）、规则参考（Rule Reference）和规则跟随（Rule Follow）等机制提取知识并注入。这些机制在调度性能、令牌消耗、决策稳定性以及生成的推理中表达的特征关注度方面进行了评估。GP衍生的指导通常改善了未指导的LLM，但其表现形式至关重要。简化决策背景或提供明确的决策逻辑比突出重要特征更为有效。特征选择提供了最佳的令牌效率，而规则跟随在更高的令牌成本下实现了强劲的性能。指导还改善了决策稳定性，并改变了生成推理中表达的特征。
cs.AI / 36 / 2607.27703
SpatialCLI: Learning to Reason With Spatial Tools, Then Without Them
SpatialCLI：学习使用空间工具进行推理，然后在没有工具的情况下进行推理
Abstract
Vision-language models (VLMs) are increasingly used in embodied agents to interpret visual inputs, reason about spatial relationships, and make task-level decisions based on that reasoning. However, a fundamental capability mismatch remains: general VLMs can reason about the overall task but often miss the visual details that determine success, while specialist vision models can capture those details but cannot translate them into task-level decisions. In this work, we propose SpatialCLI, a framework that teaches VLMs to reason with spatial tools and progressively internalize the specialist perceptual capabilities they provide. SpatialCLI proceeds in three stages: (1) Call exposes specialist vision models as spatial tools to augment the VLM's perception; (2) Learn uses Cold-Start SFT and agentic RL to improve tool use; and (3) Internalize verbalizes successful tool-use trajectories to internalize specialist perceptual capabilities. We further introduce SpatialCLI-Bench, a 516-example benchmark for compositional perception across localization, segmentation, depth, and pose. On MindCube, SpatialCLI raises Qwen3-VL-8B-Instruct from 29.3% to 84.6% with tools, surpassing GPT-5.6 Sol with tools (72.1%), while retaining 73.8% without tools after internalization.
Chinese Translation
视觉语言模型（VLMs）在具身智能体中越来越多地被用于解读视觉输入、推理空间关系，并基于这些推理做出任务级决策。然而，仍然存在一种基本的能力不匹配：通用的 VLMs 可以对整体任务进行推理，但往往忽视决定成功的视觉细节，而专业的视觉模型能够捕捉这些细节，却无法将其转化为任务级决策。在本研究中，我们提出了 SpatialCLI，一个框架，旨在教会 VLMs 使用空间工具进行推理，并逐步内化它们所提供的专业感知能力。SpatialCLI 由三个阶段组成：（1）Call 将专业视觉模型作为空间工具暴露出来，以增强 VLM 的感知；（2）Learn 使用冷启动 SFT 和自主强化学习（agentic RL）来改善工具的使用；（3）Internalize 将成功的工具使用轨迹进行语言化，以内化专业感知能力。我们进一步介绍了 SpatialCLI-Bench，这是一个包含 516 个示例的基准，涵盖了定位、分割、深度和姿态的组合感知。在 MindCube 上，SpatialCLI 将 Qwen3-VL-8B-Instruct 的表现从 29.3% 提升至 84.6%（使用工具），超越了使用工具的 GPT-5.6 Sol（72.1%），同时在内化后在没有工具的情况下保持 73.8%。
cs.AI / 37 / 2607.27705
Albilich: Steerable Proof-State Orchestration for LLM-Based Mathematical Research with CAS Integration
Albilich：基于大语言模型的数学研究中可引导的证明状态编排与计算机代数系统集成
Abstract
Large language models can contribute useful ideas to mathematical research, yet long-horizon proof attempts remain difficult to coordinate, evaluate, and reproduce. We present Albilich, an open-source agentic harness for autoresearch in mathematics that combines long-horizon reasoning, computer algebra systems (CAS), literature retrieval, and persistent SQLite-based context management. We evaluate Albilich on the RealMath benchmark (Zhang et al. 2025) and on open problems in group theory from the Kourovka Notebook (Khukhro and Mazurov 2026). It solved 10/10 problems on RealMath with CAS and 9/10 with no CAS. On the Kourovka problems, Albilich produced a counterexample to Problem 21.142 and a proof of a strengthening of Problem20.2. Anablation on Problem 17.91 demonstrates 32.0% token reduction when CAS is enabled. An ablation on Problem 21.142 demonstrates higher verifier-rejection rate and failure to synthesize proof routes in the absence of the advisor agent. These results support Albilich as a human-steerable, CAS-boosted environment for scalable AI-assisted mathematical research.
Chinese Translation
大型语言模型可以为数学研究提供有用的想法，但长时间跨度的证明尝试仍然难以协调、评估和重现。我们提出了Albilich，一个用于数学自研究的开源智能框架，它结合了长时间跨度推理、计算机代数系统（CAS）、文献检索和基于SQLite的持久上下文管理。我们在RealMath基准（Zhang et al. 2025）和Kourovka Notebook中的群论开放问题（Khukhro and Mazurov 2026）上评估了Albilich。它在RealMath上使用CAS解决了10/10个问题，而在没有CAS的情况下解决了9/10个问题。在Kourovka问题中，Albilich为问题21.142提供了一个反例，并证明了问题20.2的一个加强版。对问题17.91的消融实验表明，当启用CAS时，令牌减少率达到32.0%。对问题21.142的消融实验显示，在缺少顾问代理的情况下，验证者拒绝率更高，并且未能合成证明路径。这些结果支持Albilich作为一个可由人类引导、增强的CAS环境，适用于可扩展的AI辅助数学研究。
cs.AI / 38 / 2607.27709
MECA: A Mechanism-Centered Agent for Constructing Well-Specified and Valuable Mathematical Conjectures
MECA：一个以机制为中心的代理，用于构建明确且有价值的数学猜想
Abstract
Automatically constructing well-specified and valuable mathematical conjectures remains a central challenge in AI-assisted mathematical discovery. Many existing open problems and conjectures are often too broad, underspecified, or difficult to connect to plausible proof or refutation strategies. We view a mathematical mechanism as a structure or reasoning principle that connects the assumptions of a candidate problem to its target conclusion, such as an inequality, invariant, decomposition, or reduction to an intermediate claim. We present MECA (MEchanism-centered Conjecture Agent), a multi-agent framework that constructs conjectures by jointly developing candidate statements and their supporting mechanisms. Explorer agents propose mechanisms, test how they apply, and revise the candidate conjecture accordingly, while critic agents assess their mathematical validity and research value. Their feedback guides changes to the assumptions, scope, and conclusion. Through this process, MECA transforms broad research directions into precise conjectures with substantive mathematical support while retaining a clearly identified unresolved core. We evaluate MECA in two complementary settings. First, we compare it with a generate-and-revise baseline on reconstructing preselected target-paper conclusions from target-conditioned but article-blind source materials. Second, we construct 100 semi-open problems from literature-derived seeds and existing open problems and evaluate them through independent proof and refutation attempts by automated provers. Our results indicate that mechanism-centered refinement produces well-specified and research-worthy conjectures that remain challenging for current automated provers.
Chinese Translation
自动构建明确且有价值的数学猜想仍然是人工智能辅助数学发现中的一个核心挑战。许多现有的开放问题和猜想往往过于宽泛、描述不清，或难以与可信的证明或反驳策略相连接。我们将数学机制视为一种结构或推理原则，它将候选问题的假设与其目标结论（如不等式、不变性、分解或归约到中间主张）连接起来。我们提出了MECA（以机制为中心的猜想代理），这是一个多代理框架，通过共同开发候选陈述及其支持机制来构建猜想。探索者代理提出机制，测试其适用性，并相应修订候选猜想，而评审代理则评估其数学有效性和研究价值。他们的反馈指导假设、范围和结论的变化。通过这一过程，MECA将宽泛的研究方向转化为具有实质数学支持的精确猜想，同时保留一个明确识别的未解决核心。我们在两个互补的环境中评估MECA。首先，我们将其与生成-修订基线进行比较，以重构从目标条件但对文章盲目的源材料中选定的目标论文结论。其次，我们从文献衍生的种子和现有开放问题中构建100个半开放问题，并通过自动证明者的独立证明和反驳尝试对其进行评估。我们的结果表明，以机制为中心的精炼产生了明确且具有研究价值的猜想，这些猜想对当前的自动证明者仍然具有挑战性。
cs.AI / 39 / 2607.27720
New Synchronous Computation Dynamics for Hopfield Networks
霍普菲尔德网络的新同步计算动态
Abstract
The dynamics of the original Hopfield network is asynchronous (sequential) (updates the state of only one neuron per time step). In this paper, we propose a new tool and a new dynamics to reduce the processing time by updating one or more neurons simultaneously per instant while ensuring process convergence and aiming for the maximum energy decrease at each step, thus guaranteeing the shortest total processing time. From the point of view of synchronous dynamics, calculating the next network state at which energy decreases the most from the current state while ensuring convergence is itself a combinatorial optimization problem. We develop and use a new tool to solve it. We call this new tool Discrete Differential Filter (DDF) and, based upon it, we develop a new synchronous dynamics which we call SD-DDF (Synchronous Dynamics based upon Discrete Differential Filter). In this paper, we review the original asynchronous dynamics for Hopfield networks and present a new tool and a new synchronous dynamics with its theoretical justification and four computational experiments to assess the speed up in processing time empirically.
Chinese Translation
原始霍普菲尔德网络的动态是异步的（顺序的）（每个时间步只更新一个神经元的状态）。在本文中，我们提出了一种新工具和一种新动态，通过在每个时刻同时更新一个或多个神经元来减少处理时间，同时确保过程收敛，并在每一步中追求最大能量降低，从而保证最短的总处理时间。从同步动态的角度来看，计算从当前状态出发能量降低最多的下一个网络状态，同时确保收敛，本身就是一个组合优化问题。我们开发并使用了一种新工具来解决这个问题。我们将这个新工具称为离散微分滤波器（Discrete Differential Filter, DDF），并基于此开发了一种新的同步动态，称为基于离散微分滤波器的同步动态（Synchronous Dynamics based upon Discrete Differential Filter, SD-DDF）。在本文中，我们回顾了霍普菲尔德网络的原始异步动态，并提出了一种新工具和一种新的同步动态，附有理论依据以及四个计算实验，以实证评估处理时间的加速效果。
cs.AI / 40 / 2607.27726
Baikal: Structured Search for Deep Research over Data Lakes
Baikal：针对数据湖的深度研究结构化搜索
Abstract
Deep research over data lakes requires an LLM agent to investigate evidence across thousands of heterogeneous tables and passages to synthesize a report. Existing methods perform iterative retrieval and generation, letting accumulated context determine what to investigate next, which can overexploit locally promising evidence and fail to cover distinct semantic regions under a fixed budget. To address this, we cast deep research over data lakes as a budgeted search problem and present Baikal - a framework that clusters heterogeneous evidence into semantic regions, then searches over them adaptively to balance exploration and exploitation. Within each selected region, Baikal generates and investigates region-grounded subquestions, using finding quality as rewards to update region-level value estimates and guide search under policies ranging from random and LLM-guided selection to Bayesian $\epsilon$-greedy and UCB. We evaluate Baikal on 15 queries each over HybridQA and TAT-QA data lakes containing 10,993 and 2,757 tables, respectively, together with 227K Wikipedia passages and 13K financial report passages. We assess research quality with a new rubric covering groundedness, relevance, diversity, and utility, and use GPT-5-mini to score Baikal and strong baselines, including DeepSearcher and an OpenCode research agent with retrieval and clustering variants. Across both data lakes, Baikal performs strongly under several region-selection policies; its best configuration improves report scores over the strongest baselines by 28% on HybridQA and 36% on TAT-QA. Our analyses attribute these gains to organizing and exploring semantic evidence regions, which improves groundedness and diversity and yields more useful findings under the same subquestion budget. These results demonstrate the value of structured semantic exploration for systematic research and discovery over heterogeneous data lakes.
Chinese Translation
对数据湖的深度研究需要一个大型语言模型（LLM）代理来调查成千上万的异构表格和段落中的证据，以合成报告。现有方法采用迭代检索和生成，让累积的上下文决定下一步调查的内容，这可能会过度利用局部有前景的证据，并在固定预算下未能覆盖不同的语义区域。为了解决这个问题，我们将数据湖上的深度研究视为一个预算搜索问题，并提出了Baikal——一个将异构证据聚类为语义区域的框架，然后在这些区域中自适应地进行搜索，以平衡探索与利用。在每个选定区域内，Baikal生成并调查基于区域的子问题，使用发现质量作为奖励来更新区域级价值估计，并指导搜索，采用从随机和LLM引导选择到贝叶斯$ ext{ε}$-贪婪和UCB的多种策略。我们在HybridQA和TAT-QA数据湖上对15个查询进行评估，分别包含10,993和2,757个表格，以及227K维基百科段落和13K财务报告段落。我们使用一个新的评估标准来评估研究质量，涵盖基础性、相关性、多样性和实用性，并使用GPT-5-mini对Baikal和强基线进行评分，包括DeepSearcher和具有检索和聚类变体的OpenCode研究代理。在两个数据湖中，Baikal在几种区域选择策略下表现出色；其最佳配置在HybridQA上比最强基线提高了28%的报告评分，在TAT-QA上提高了36%。我们的分析将这些增益归因于组织和探索语义证据区域，这提高了基础性和多样性，并在相同的子问题预算下产生了更有用的发现。这些结果证明了结构化语义探索在异构数据湖上进行系统研究和发现的价值。
cs.AI / 41 / 2607.27733
VeriSkill: A Self-Evolution Framework for Program Verification Skills
VeriSkill：一种自我进化的程序验证技能框架
Abstract
Automating program verification with LLM agents requires generating specifications, annotations, auxiliary lemmas, and tool invocations, all of which depend on reusable skills. A natural remedy is skill self-evolution: distilling skills from trajectories and refining them through feedback. However, existing evolution methods struggle with program verification tasks because they cannot reliably identify skill-specific failures or extract actionable signals from opaque verifier feedback. In this paper, we propose VeriSkill, a self-evolution framework built for program verification. It attributes verification failures to skill deficiencies, distills diagnostic signatures into reusable lessons, and iteratively refines candidate skills, admitting only revisions that improve verification performance while preserving program semantics. Experiments show that VeriSkill consistently outperforms all baselines across multiple verification tools, agent frameworks, and LLM backends.
Chinese Translation
使用大型语言模型（LLM）代理自动化程序验证需要生成规范、注释、辅助引理和工具调用，这些都依赖于可重用的技能。一种自然的解决方案是技能自我进化：从轨迹中提炼技能并通过反馈进行精炼。然而，现有的进化方法在程序验证任务中表现不佳，因为它们无法可靠地识别特定技能的失败或从不透明的验证器反馈中提取可操作的信号。本文提出了VeriSkill，一种专为程序验证构建的自我进化框架。它将验证失败归因于技能不足，将诊断特征提炼为可重用的经验教训，并迭代地精炼候选技能，仅接受那些在保持程序语义的同时提高验证性能的修订。实验表明，VeriSkill在多个验证工具、代理框架和LLM后端中始终优于所有基线。
cs.AI / 42 / 2607.27788
SpecCal: Ambiguity-Aware Candidate Calibration for Infrared Spectrum-Based Molecular Structure Reconstruction
SpecCal：基于红外光谱的分子结构重建中的模糊感知候选校准
Abstract
Inferring molecular structures from infrared (IR) spectra is a fundamental yet challenging problem. A key difficulty is that an IR spectrum provides limited structural information: different molecules may share similar functional groups and local vibrational patterns, leading to highly similar spectral responses. Thus, even when an observed spectrum has a unique underlying structure, reconstructing it from the spectrum remains ambiguous. Existing IR-to-molecule models usually generate a ranked set of candidate molecules, but this set is largely determined by the model's learned generation preference and may not fully capture the structures that best satisfy the observed spectral constraints. To address this limitation, we propose SpecCal, a training-free candidate calibration framework for IR-to-molecule prediction. SpecCal operates on the candidate outputs of existing base models and improves the prediction set by re-ranking current candidates while introducing additional structurally plausible alternatives guided by spectral consistency. The framework is plug-and-play and model-agnostic, requiring no parameter updates for integration with diverse base models. Experiments on multiple benchmarks show that SpecCal consistently improves top-k reconstruction at both SMILES and scaffold levels across different base models. Further analyses demonstrate that calibrating candidate sets under spectral ambiguity provides a practical way to improve molecular reconstruction from IR spectra. The code is available at: https://anonymous.4open.science/r/SpecCal-B18A.
Chinese Translation
从红外（IR）光谱推断分子结构是一个基本但具有挑战性的问题。一个关键的困难在于，IR光谱提供的结构信息有限：不同的分子可能共享相似的功能团和局部振动模式，从而导致高度相似的光谱响应。因此，即使观察到的光谱具有独特的基础结构，从光谱中重建它仍然存在模糊性。现有的IR到分子模型通常生成一组排名的候选分子，但这一组在很大程度上取决于模型学习到的生成偏好，可能无法完全捕捉最佳满足观察到的光谱约束的结构。为了解决这一局限性，我们提出了SpecCal，一个无训练的候选校准框架，用于IR到分子的预测。SpecCal在现有基础模型的候选输出上运行，通过重新排名当前候选者并引入额外的结构上合理的替代方案（以光谱一致性为指导）来改善预测集。该框架即插即用且与模型无关，集成到不同基础模型中无需参数更新。在多个基准测试中的实验表明，SpecCal在不同基础模型下持续改善了SMILES和骨架层面的前k重建。进一步的分析表明，在光谱模糊性下校准候选集提供了一种实用的方法来改善从IR光谱中进行的分子重建。代码可在以下链接获取：https://anonymous.4open.science/r/SpecCal-B18A。
cs.AI / 43 / 2607.27792
Annotating Topical Legal Insights from Case Proceedings
从案件程序中注释主题法律见解
Abstract
In this paper, we mainly concentrate on finding concepts or topics from the legal case proceedings, since adopting a structured representation for legal documents, as opposed to a mere bag-of-words flat text representation, can significantly enhance processing capabilities. To achieve this objective, we put forward a set of diverse concepts for legal case proceedings. With this motivation, we propose LeDA, a system for Legal Data Annotation. The system offers the generic functionality of annotating and adjudicating entities or concepts within documents via a web-based interface. A novel feature of our system is that it allows to dynamic create new tags for annotation, which is a particularly useful provision for situations where there exists no pre-defined ontology for the entities (concepts) that need to be annotated - these being rather discovered by annotators as they continue examining more documents. The system that we demonstrate is currently in use to annotate a set of concepts from legal documents to construct semantic representations of documents as bags of concepts that can then be used for several downstream tasks, such as prior case retrieval, judgment prediction, and so on. Along with the system features in general, we also describe how LeDA was used by 3 assessors to annotate and adjudicate legal concept names from Indian Supreme Court case proceedings.
Chinese Translation
本文主要集中于从法律案件程序中寻找概念或主题，因为采用结构化表示法来处理法律文件，相较于仅仅使用词袋模型的平面文本表示，可以显著提升处理能力。为了实现这一目标，我们提出了一组多样化的法律案件程序概念。基于此动机，我们提出了LeDA（法律数据注释）系统。该系统通过基于网络的界面提供注释和裁定文档中实体或概念的通用功能。我们系统的一个新颖特性是允许动态创建新的注释标签，这在需要注释的实体（概念）没有预定义本体的情况下尤其有用——这些实体（概念）往往是由注释者在继续审查更多文档时发现的。我们展示的系统目前正在用于注释法律文件中的一组概念，以构建文档的语义表示，作为概念的集合，这些表示可以用于多个下游任务，如先前案件检索、判决预测等。除了系统的一般功能外，我们还描述了LeDA是如何被三位评估者用于注释和裁定印度最高法院案件程序中的法律概念名称的。
cs.AI / 44 / 2607.27798
MemeBench: What LVLMs Miss When Interpreting Culture-Dependent Memes
MemeBench：大规模视觉语言模型在解读文化依赖性表情包时的不足
Abstract
Large vision-language models have improved at describing visual content, but accurate descriptions do not ensure interpretation when meaning depends on knowledge beyond the pixels. Memes expose this gap because they rely on cultural entities, background knowledge, and community conventions. Most meme benchmarks reduce interpretation to labels or holistic scores, obscuring where an explanation breaks down. We introduce MemeBench, a diagnostic benchmark of 1,253 Chinese and English memes with human-written references and quality-controlled VIKR annotations, centered on anime, comics, games, and adjacent online subcultures. Its VIKR schema decomposes explanations into Visual clues, Identity links, Knowledge units, and Reasoning mechanisms. Across 26 LVLMs, every model covers visible content more reliably than the knowledge needed to interpret it, and even the strongest retains a 22.6% Visual-Knowledge gap. To test whether this diagnosis can guide improvement, we introduce KAR, an entity-guided retrieval baseline built on CultureBase. Across four controlled models, KAR raises VIKR Success by 3.6-7.4% and, compared with generic retrieval, repairs more answers and breaks fewer. Yet both retrieval conditions improve Identity and Knowledge while reducing Visual coverage in every comparison. MemeBench reveals whether an interpretation succeeds, what is missing, and whether targeted evidence fills the diagnosed gap.
Chinese Translation
大型视觉语言模型在描述视觉内容方面有所改善，但准确的描述并不保证在意义依赖于超出像素的知识时能够进行正确解读。表情包揭示了这一差距，因为它们依赖于文化实体、背景知识和社区惯例。大多数表情包基准测试将解读简化为标签或整体评分，模糊了解释失效的地方。我们引入了MemeBench，这是一个包含1,253个中英文表情包的诊断基准，配有人工撰写的参考和质量控制的VIKR注释，重点关注动漫、漫画、游戏及相关在线亚文化。其VIKR框架将解释分解为视觉线索、身份链接、知识单元和推理机制。在26个大型视觉语言模型中，每个模型在覆盖可见内容方面的可靠性高于解读所需的知识，甚至最强的模型也保持着22.6%的视觉-知识差距。为了测试这一诊断是否能指导改进，我们引入了KAR，这是一个基于CultureBase的实体引导检索基线。在四个受控模型中，KAR将VIKR成功率提高了3.6-7.4%，与通用检索相比，修复了更多答案且出错更少。然而，在每次比较中，两种检索条件都提高了身份和知识的覆盖率，同时减少了视觉覆盖。MemeBench揭示了解释是否成功、缺失了什么，以及有针对性的证据是否填补了诊断出的差距。
cs.AI / 45 / 2607.27824
STEREODISCO: Discovering Stereotypicality in LLMs
STEREODISCO：发现大型语言模型中的刻板印象性
Abstract
LLMs encode, convey, and perpetuate stereotypes. Prior computational research focuses on a small set of semantic axes investigated in social psychology, and operates on word embeddings produced by language models, leaving open which other semantic axes carry stereotypical associations in LLMs and how LLMs internally represent such axes. We introduce STEREODISCO, a framework that adapts the semantic differential method (Osgood et al., 1957) to the systematic study of stereotypes in LLM internal representations. STEREODISCO constructs approx. 2,000 candidate semantic axes from WordNet antonym synsets, recovers each as a geometric axis in the LLM's activation space via probing, and identifies stereotypical axes via a statistical test over concept projections. As a case study, we apply STEREODISCO to social group stereotypes with LLAMA-3-8B-INSTRUCT and MISTRAL-7B-INSTRUCT. We find that the two LLMs agree with each other on social group ratings more than with humans, suggesting that LLM-encoded stereotype content diverges from that documented in social psychology. We also discover stereotypical axes not investigated in prior work -- including humble vs. proud, narrow-minded vs. broad-minded, and cowardly vs. brave, which human annotators independently confirm.
Chinese Translation
大型语言模型（LLMs）编码、传达并延续刻板印象。之前的计算研究集中于社会心理学中研究的一小部分语义轴，并基于语言模型生成的词嵌入进行操作，尚未明确其他哪些语义轴在LLMs中携带刻板印象关联，以及LLMs如何内部表示这些轴。我们引入了STEREODISCO，一个将语义差异法（Osgood et al., 1957）适应于LLM内部表示中刻板印象的系统研究的框架。STEREODISCO从WordNet反义词同义词集中构建了大约2000个候选语义轴，通过探测将每个轴恢复为LLM激活空间中的几何轴，并通过对概念投影的统计检验识别刻板印象轴。作为案例研究，我们将STEREODISCO应用于社会群体刻板印象，使用LLAMA-3-8B-INSTRUCT和MISTRAL-7B-INSTRUCT。我们发现这两个LLM在社会群体评分上的一致性高于与人类的评分，表明LLM编码的刻板印象内容与社会心理学中记录的内容存在差异。我们还发现了一些在先前研究中未被探讨的刻板印象轴，包括谦逊与骄傲、狭隘与开阔、懦弱与勇敢，这些都得到了人类标注者的独立确认。
cs.AI / 46 / 2607.27826
Sign Language Question Answering: A New Task, Benchmark, and Baseline for Sign Language Understanding
手语问答：手语理解的新任务、基准和基线
Abstract
Recent advances in sign language (SL) understanding (SLU) have led to remarkable progress in tasks such as continuous SL recognition and SL translation. However, these tasks are designed with predefined objectives, requiring models to learn a fixed mapping from sign videos to glosses or spoken-language sentences. As a result, they provide only a limited assessment of whether a model truly understands the semantic content of SL videos. To address this limitation, \textbf{we first propose a new task, Sign Language Question Answering (SLQA)}, which evaluates SL understanding by requiring models to answer arbitrary natural language questions about SL videos. Unlike previous SLU tasks, SLQA provides a more flexible and comprehensive evaluation framework that assesses multiple reasoning capabilities beyond recognition and translation. To facilitate this task, \textbf{we further construct two SignQA benchmarks} based on PHOENIX14T and CSL-Daily by automatically generating question-answer pairs from existing gloss and sentence annotations using carefully designed templates. The resulting datasets cover five complementary question categories, including position reasoning, structural reasoning, visual search, gloss recognition, and translation understanding. \textbf{Finally, we propose a simple yet effective baseline model} equipped with a Question-Conditioned Modulated Temporal Downsampling module and an in-domain knowledge transfer strategy, enabling effective knowledge transfer from existing SLU tasks while enhancing question-aware temporal feature modeling. Extensive experiments demonstrate that our baseline consistently outperforms representative vision-language models across all question categories, establishing a strong benchmark for future research on SLQA. Datasets are available at:{https://huggingface.co/datasets/hulala/SignQA-2026}.
Chinese Translation
近年来，手语（SL）理解（SLU）的进展使得在连续手语识别和手语翻译等任务中取得了显著的进展。然而，这些任务的设计具有预定义的目标，要求模型学习从手语视频到手语词或口语句子的固定映射。因此，它们仅能有限地评估模型是否真正理解手语视频的语义内容。为了解决这一局限性，我们首先提出了一项新任务——手语问答（Sign Language Question Answering, SLQA），该任务通过要求模型回答有关手语视频的任意自然语言问题来评估手语理解。与之前的手语理解任务不同，SLQA提供了一个更灵活和全面的评估框架，评估超越识别和翻译的多种推理能力。为了促进这一任务，我们进一步基于PHOENIX14T和CSL-Daily构建了两个SignQA基准，通过使用精心设计的模板自动生成现有手语词和句子注释的问题-答案对。生成的数据集涵盖了五个互补的问题类别，包括位置推理、结构推理、视觉搜索、手语词识别和翻译理解。最后，我们提出了一种简单而有效的基线模型，该模型配备了问题条件调制时间下采样模块和领域内知识转移策略，能够有效地从现有的手语理解任务中转移知识，同时增强对问题的时间特征建模。大量实验表明，我们的基线模型在所有问题类别中始终优于代表性的视觉-语言模型，为未来的SLQA研究建立了强有力的基准。数据集可在以下链接获取：{https://huggingface.co/datasets/hulala/SignQA-2026}。
cs.AI / 47 / 2607.27834
MemTxn: A Transaction Boundary for Source-Supported Updates and Complete-State Recovery in Agent Memory
MemTxn：用于源支持更新和完整状态恢复的事务边界在智能体内存中
Abstract
Persistent memory lets long-running large language model agents reuse information across sessions and tasks. Yet errors in writable memory can persist and corrupt future behavior. Existing systems improve storage and retrieval, but they do not provide a transaction boundary for reliable updates and recovery. We therefore propose MemTxn, a governance layer outside the answer model. MemTxn verifies whether an update is supported by its source. It also selects the visible version when facts conflict and restores the application-visible state after a fault. The system uses Ordered PatchTest to validate writes, a Temporal Resolver to select versions, and a durable snapshot journal to recover state. On an item-disjoint audit, MemTxn accepts all 60 supported originals and rejects all 179 hard negatives. Under persistent multi-key faults on LongMemEval-S and LoCoMo states, it restores the complete declared active map without knowing the actual physical write set. On MemoryAgentBench FactConsolidation, MemTxn achieves the highest average F1 across all twelve answer-model configurations. It outperforms Dense by 17.06--24.07 points in five representative settings.
Chinese Translation
持久性内存使得长期运行的大型语言模型智能体能够在不同会话和任务之间重用信息。然而，可写内存中的错误可能会持续存在并破坏未来的行为。现有系统改善了存储和检索，但并未提供可靠更新和恢复的事务边界。因此，我们提出了 MemTxn，一个位于答案模型之外的治理层。MemTxn 验证更新是否得到其源的支持。当事实发生冲突时，它还会选择可见版本，并在故障后恢复应用程序可见状态。该系统使用有序补丁测试（Ordered PatchTest）来验证写入，使用时间解析器（Temporal Resolver）来选择版本，并使用持久快照日志来恢复状态。在项目不重叠的审计中，MemTxn 接受所有 60 个支持的原件，并拒绝所有 179 个硬负样本。在 LongMemEval-S 和 LoCoMo 状态下的持久多键故障下，它在不知道实际物理写入集的情况下恢复完整的声明活动映射。在 MemoryAgentBench FactConsolidation 上，MemTxn 在所有十二个答案模型配置中实现了最高的平均 F1 值。在五个代表性设置中，它比 Dense 超出 17.06 到 24.07 分。
cs.AI / 48 / 2607.27836
Crossing the Margin Cliff: Toward Relearn-Robust LLM Unlearning via Margin Calibration
跨越边际悬崖：通过边际校准实现对大型语言模型的鲁棒性再学习
Abstract
Large language model unlearning is consistently fragile under relearn attacks. On TOFU, fine-tuning on twenty forget examples substantially recovers held-out forget-set ROUGE for every method we evaluate, and we trace this fragility to optimization geometry. The per-token answer margin of fourteen post-hoc methods spanning gradient, preference, and distillation families converges into a narrow band above the retain reference in 41 of 42 method--size cells, a regularity we call the margin cliff. We prove that this cliff follows whenever the retain coupling holds the diagnostic log-odds of forget content above a floor, a condition that token-saturating losses induce at stationarity and that we verify directly on 34 of 42 cells. Margin Calibration (\textsc{MC}) is a plug-in polish adding a non-saturating margin hinge anchored at the reference's per-token margin plus a KL probe on a disjoint instruction corpus, restoring forget-side pressure where the native loss saturates. Under a stated gradient-dominance condition, whose on-trajectory gradient signature we measure by instrumenting the polish, its stationary set lies on the cliff-crossing side, yielding an attack-budget upper bound on the relearn margin lift. Across TOFU (three Llama-3 sizes, three forget tiers), MUSE-News on Llama-2-7B-hf, and a Phi-3.5 panel, a single frozen configuration wins all 14 head-to-head forget aggregates and all populated relearn cells (panel-mean post-attack ROUGE-L $0.41$ to $0.18$) and lowers raw membership AUC on 13/14, with reduced retain-side utility as the main cost. A deployment variant matches these gains without a retain-trained reference.
Chinese Translation
大型语言模型的去学习在再学习攻击下表现出持续的脆弱性。在 TOFU 数据集上，对二十个遗忘示例进行微调显著恢复了我们评估的每种方法的保留遗忘集 ROUGE，我们将这种脆弱性追溯到优化几何。十四种后处理方法的每个标记答案边际，涵盖了梯度、偏好和蒸馏家族，在 42 个方法-规模单元中的 41 个收敛到保留参考之上的一个狭窄带，这种规律我们称之为边际悬崖。我们证明，只要保留耦合保持遗忘内容的诊断对数几率高于一个底线，这个悬崖就会出现，而这一条件是标记饱和损失在平稳状态下诱导的，我们在 42 个单元中的 34 个上直接验证了这一点。边际校准（Margin Calibration, MC）是一种插件式的改进，增加了一个非饱和的边际铰链，锚定在参考的每个标记边际上，并在一个不相交的指令语料库上进行 KL 探测，恢复了在本地损失饱和时的遗忘侧压力。在一个声明的梯度主导条件下，我们通过对改进进行仪器化测量其轨迹梯度特征，其平稳集位于悬崖跨越的一侧，从而为再学习边际提升提供了攻击预算上限。在 TOFU（三个 Llama-3 尺寸，三个遗忘层级）、Llama-2-7B-hf 上的 MUSE-News 和一个 Phi-3.5 面板中，一个单一的冻结配置在所有 14 个面对面遗忘聚合和所有填充的再学习单元中获胜（面板均值攻击后 ROUGE-L 从 $0.41$ 降至 $0.18$），并在 13/14 中降低了原始成员资格 AUC，主要成本是保留侧效用的降低。一个部署变体在没有保留训练参考的情况下实现了这些收益。
cs.AI / 49 / 2607.27840
Virtual Process Dossier: A Process-Aware Data Catalogue
虚拟过程档案：一种过程感知的数据目录
Abstract
We propose the Virtual Process Dossier (VPD), a Knowledge Graph-based data catalogue that also captures workflow provenance. We developed VPD for multi-stage manufacturing use-cases where downstream AI-based optimization tasks require to distinct between datasets generated during individual workflow steps. VPD provides these datasets in a FAIR manner and makes both prospective and retrospective workflow provenance explicit. Our contributions are: (1) the VPD ontology that serves as the catalogue's semantic core; (2) the VPD provenance framework that integrates ontology instantiation into the production environment; and (3) the VPD user interface that provides human-centered interaction with the VPD Knowledge Graph. The ontology and code are available at https://github.com/kubeluk/VirtualProcessDossier .
Chinese Translation
我们提出了虚拟过程档案（Virtual Process Dossier, VPD），这是一种基于知识图谱的数据目录，同时捕捉工作流的来源信息。我们为多阶段制造的应用场景开发了VPD，在这些场景中，下游基于人工智能的优化任务需要区分在各个工作流步骤中生成的数据集。VPD以公平（FAIR）的方式提供这些数据集，并明确展示前瞻性和回顾性的工作流来源信息。我们的贡献包括：（1）VPD本体，作为目录的语义核心；（2）VPD来源框架，将本体实例化集成到生产环境中；（3）VPD用户界面，提供以人为中心的与VPD知识图谱的交互。该本体和代码可在https://github.com/kubeluk/VirtualProcessDossier获取。
cs.AI / 50 / 2607.27854
Simplifying Neural Networks During Training
训练过程中简化神经网络
Abstract
Understanding and exploiting the training dynamics of overparameterized deep neural networks remains a central challenge in modern machine learning. Recent evidence on Neural Collapse (NC) shows that class representations and classifiers exhibit highly structured geometry, while the Tunnel Effect suggests that only a subset of layers is essential for feature extraction. We combine these two perspectives and propose an NC-inspired training framework for simplifying deep networks during training. Our method monitors representation dynamics through the Inverse Fisher Criterion, a stable and efficient proxy for the variability collapse behavior, to identify both the split point between feature extraction and classification and the training stage at which simplification becomes viable. We then replace the trailing layers with a lightweight classification head and continue training the reduced model. Experiments on image-classification benchmarks across MLP, VGG, and ResNet architectures show that the proposed method achieves substantial parameter reductions while maintaining accuracy comparable to that of the full model. Code to reproduce the experiments can be found at: https://github.com/LorenzoSciandra/NNS.
Chinese Translation
理解和利用过参数化深度神经网络的训练动态仍然是现代机器学习中的一个核心挑战。关于神经崩溃（Neural Collapse, NC）的最新证据表明，类别表示和分类器表现出高度结构化的几何特征，而隧道效应（Tunnel Effect）则表明只有一部分层对于特征提取是必不可少的。我们结合这两种视角，提出了一种受NC启发的训练框架，用于在训练过程中简化深度网络。我们的方法通过逆费舍尔准则（Inverse Fisher Criterion）监测表示动态，这是一种稳定且高效的变异崩溃行为的代理，以识别特征提取与分类之间的分裂点以及简化变得可行的训练阶段。然后，我们用轻量级分类头替换后续层，并继续训练简化后的模型。在MLP、VGG和ResNet架构的图像分类基准测试中的实验表明，所提出的方法在保持与完整模型相当的准确性的同时，实现了显著的参数减少。可以在以下链接找到重现实验的代码：https://github.com/LorenzoSciandra/NNS。
cs.AI / 51 / 2607.27861
Back to All-Entity Ranking: Sampler-Dependent Evaluation in Continuous-Time Dynamic Graphs
回归全实体排名：连续时间动态图中的采样器依赖评估
Abstract
Next-destination prediction in continuous-time dynamic graphs (CTDGs) commonly ranks an observed interaction against sampled negative destinations. The resulting score is conditional on both the negative distribution and the number of candidates chosen by the researcher. We show that a non-uniform negative distribution changes the Bayes-optimal ranking, while even a finite candidate set drawn uniformly can destabilize model rankings and measured module effects. Time-varying source-destination history membership and model operations that use this information directly transmit the sampler's influence to the evaluation score. We examine this mechanism using a factorial evaluation of repeated and new positives against seen and unseen negatives, a minimal scorer based solely on pair-history membership, and controlled representation interventions. Across six models on LastFM, MOOC, Reddit, and Wikipedia, at least one model pair changes relative order between the expected Uniform-20 metric and the full catalog on three of the four datasets. The measured effect of the same module also changes in magnitude and direction with the candidate-set size and training objective. These results establish that model-superiority and ablation conclusions from sampled-negative benchmarks are conditional on the stated candidate configuration. All-entity ranking evaluates every destination in a fixed catalog, eliminating negative-selection freedom and sampling variation while retaining the original CTDG scorer. We therefore recommend all-entity ranking as the primary evidence for architecture comparisons on CTDG benchmarks with an enumerable, fixed destination catalog.
Chinese Translation
在连续时间动态图（CTDGs）中，下一目的地预测通常是将观察到的交互与采样的负目的地进行排名。由此产生的得分依赖于负分布和研究者选择的候选数量。我们展示了非均匀负分布会改变贝叶斯最优排名，而即使是均匀抽取的有限候选集也可能使模型排名和测量模块效应不稳定。时间变化的源-目的地历史成员资格以及直接使用这些信息的模型操作将采样器的影响传递到评估得分。我们通过对重复和新正样本与已见和未见负样本的因子评估、仅基于对历史成员资格的最小评分器以及受控表示干预来检验这一机制。在 LastFM、MOOC、Reddit 和 Wikipedia 的六个模型中，至少有一对模型在预期的 Uniform-20 指标和完整目录之间的相对顺序发生了变化，涉及四个数据集中的三个。相同模块的测量效应在候选集大小和训练目标的变化下也会改变其幅度和方向。这些结果表明，从采样负基准得出的模型优越性和消融结论依赖于所述的候选配置。全实体排名评估固定目录中的每个目的地，消除了负选择自由和采样变异，同时保留了原始的 CTDG 评分器。因此，我们建议将全实体排名作为 CTDG 基准中架构比较的主要证据，前提是存在可枚举的固定目的地目录。
cs.AI / 52 / 2607.27867
Orca: Neural Operators for Causal Reasoning in Continuous Time
Orca：用于连续时间因果推理的神经算子
Abstract
Structural causal models are the standard language for reasoning about interventions and counterfactuals, but they describe static variables, typically measured once, and usually forbid cyclic dependencies. Many systems we care about, such as patients, climates, and economies, instead evolve continuously in time, are observed at irregular time points, and contain feedback loops. We argue that neural operator learning provides a natural foundation for causal reasoning in this setting, and propose Orca, a framework in which each node of the causal graph is a function of time and each mechanism is a learned map between function spaces. We extend existing neural operator architectures to express causal mechanisms: a mechanism computes the function value of a node from its parent nodes by taking several parent functions as input, respects the arrow of time, and treats latent exogenous noise as a function that can be inferred and reused for counterfactuals. We formalize the model class and demonstrate counterfactual reasoning on synthetic continuous-time examples. Code is available at https://github.com/gerritgr/orca
Chinese Translation
结构因果模型是关于干预和反事实推理的标准语言，但它们描述的是静态变量，通常只测量一次，并且通常禁止循环依赖。我们关心的许多系统，例如患者、气候和经济，实际上是随时间连续演变的，在不规则的时间点被观察，并且包含反馈循环。我们认为，神经算子学习为这种环境下的因果推理提供了自然的基础，并提出了Orca，一个框架，其中因果图的每个节点都是时间的函数，每个机制是功能空间之间的学习映射。我们扩展了现有的神经算子架构以表达因果机制：一个机制通过将多个父函数作为输入来计算节点的函数值，遵循时间的箭头，并将潜在的外生噪声视为可以推断和重用以进行反事实的函数。我们形式化了模型类，并在合成的连续时间示例上演示了反事实推理。代码可在 https://github.com/gerritgr/orca 获取。
cs.AI / 53 / 2607.27871
Search as Computation Allocation
搜索作为计算分配
Abstract
Many algorithms spend an internal resource before returning a decision and are evaluated only by the quality of that terminal output. We formalize such procedures as terminal computation-allocation problems: costly computations produce observations, update beliefs about a latent environment, and matter only through terminal decision loss. Bellman equations characterize optimal allocation under fixed budgets, priced computation, and exact certification. We then relate value of computation (VOC) to information. Mutual information equals myopic VOC under log loss, whereas under simple regret VOC is a knowledge-gradient quantity; moreover, information gain can rank computations arbitrarily poorly, although it gives a one-sided upper bound on VOC. Bandit pulls, tree simulations, and node expansions illustrate the same model under different computation topologies. Finally, under an explicit frontier-resolution and heuristic-error model, maximizing approximate VOC recovers weighted A*, with A* and greedy best-first search as limiting cases. The theory identifies a shared decision problem without asserting that one acquisition rule is universally optimal.
Chinese Translation
许多算法在返回决策之前会消耗内部资源，并且仅通过最终输出的质量进行评估。我们将此类过程形式化为终端计算分配问题：昂贵的计算产生观察结果，更新对潜在环境的信念，并且仅通过终端决策损失来影响结果。贝尔曼方程描述了在固定预算、定价计算和精确认证下的最优分配。我们随后将计算价值（Value of Computation, VOC）与信息联系起来。在对数损失下，互信息等于短视的 VOC，而在简单遗憾下，VOC 是一个知识梯度量；此外，信息增益可能会对计算的排名极其不佳，尽管它为 VOC 提供了一个单边上界。赌博拉取、树模拟和节点扩展在不同的计算拓扑下展示了相同的模型。最后，在明确的前沿分辨率和启发式误差模型下，最大化近似 VOC 恢复了加权 A*，而 A* 和贪婪最佳优先搜索则是极限情况。该理论识别了一个共享的决策问题，而不声称某一获取规则是普遍最优的。
cs.AI / 54 / 2607.27877
An Empirical Study of Coordination Mode as the First-Class Citizen in From-Scratch Multi-Agent Coding
从零开始的多智能体编码中协调模式作为第一公民的实证研究
Abstract
Multi-agent vibe coding promises to accelerate software development, yet existing benchmarks rely on synthetic environments that ignore practical time and monetary costs, conflate reasoning with communication, and reward only superficial completion. We introduce multi-agent from-scratch evaluation benchmark, MSEval, evaluating multi-agent coding on real-world tasks. Grounded in 10 authentic, full-stack projects across 10 domains, MSEval scores performance using hierarchical requirements and deterministic rubrics. Its execution engine, LegoGent, tests 10 collaboration topologies where agents coordinate via periodic sync intervals and deploy through native CI/CD pipelines. Concurrently, the automated grader TAgent dynamically probes implementations to jointly measure functional success, latency, and prefix-cached token cost. Across 100 runs, MSEval reveals that organizational topology rivals model capability in shaping the speed--cost--quality trade-off. For identical tasks and models, varying the topology shifts scores by over 30 points and doubles wall-clock time. Structured pipelines converge fastest with the highest quality, whereas heavy managerial oversight degrades performance. Ultimately, MSEval establishes a rigorous, reproducible standard for measuring how multi-agent teams actually build software. The benchmark is released at https://github.com/robinren03/MSEval.
Chinese Translation
多智能体氛围编码有望加速软件开发，但现有基准依赖于忽视实际时间和金钱成本的合成环境，将推理与通信混为一谈，并仅奖励表面的完成。我们引入了多智能体从零开始评估基准 MSEval，评估多智能体编码在真实世界任务中的表现。MSEval 基于 10 个领域的 10 个真实全栈项目，通过分层需求和确定性评分标准对性能进行评分。其执行引擎 LegoGent 测试 10 种协作拓扑，其中智能体通过定期同步间隔进行协调，并通过本地 CI/CD 管道进行部署。同时，自动评分器 TAgent 动态探测实现，以共同测量功能成功、延迟和前缀缓存令牌成本。在 100 次运行中，MSEval 显示组织拓扑在塑造速度-成本-质量权衡方面与模型能力不相上下。对于相同的任务和模型，改变拓扑会使得评分变化超过 30 分，并使墙钟时间翻倍。结构化管道以最高质量最快收敛，而过多的管理监督则会降低性能。最终，MSEval 建立了一个严格、可重复的标准，用于衡量多智能体团队实际构建软件的方式。该基准已发布于 https://github.com/robinren03/MSEval。
cs.AI / 55 / 2607.27888
Not All Tokens Deserve Equal Credit: Counterfactual Sensitivity Credit Reallocation for Long-CoT Reasoning
并非所有标记都应获得平等的信用：针对长链推理的反事实敏感性信用重新分配
Abstract
Reinforcement learning with verifiable rewards (RLVR) is central to improving long-CoT reasoning in large language models. Critic-free methods such as GRPO convert response-level rewards into advantages and uniformly broadcast them across tokens, overlooking their unequal contributions to the final outcome. On-policy self-distillation (OPSD) instead provides dense distributional supervision by minimizing the forward KL divergence between an unprivileged policy and a privileged self-teacher, implicitly assuming that the resulting likelihood shifts encode reliable answer-aligned information. We test this premise by fixing each sampled trajectory and re-scoring it under two opposing outcome conditions, one asserting correctness and the other incorrectness. Most affected tokens shift in the same direction under both conditions, with few sign reversals and substantial overlap in the induced optimization signals. Large shifts also concentrate on highly substitutable surface-form tokens, whereas tokens carrying problem-specific reasoning content are less sensitive. These findings show that privileged shifts fail to provide reliable answer-aligned directions, while their magnitudes primarily reflect counterfactual sensitivity rather than token-level learning value. Based on these observations, we propose Counterfactual Sensitivity Credit Reallocation (CSCR), a simple extension of GRPO that reduces credit for highly sensitive tokens and renormalizes token-level advantages to preserve both the original credit budget and verifier-determined direction. On long-CoT mathematical reasoning benchmarks, CSCR consistently outperforms GRPO baseline with the same number of policy updates. Targeted ablations further corroborate our diagnosis: privilege-induced directions are unreliable, moderate downweighting is most effective, and stronger modulation destabilizes optimization.
Chinese Translation
具有可验证奖励的强化学习（RLVR）在提升大型语言模型的长链推理能力中至关重要。无评论的方法如 GRPO 将响应级奖励转化为优势，并在标记之间均匀传播，忽视了它们对最终结果的不平等贡献。相反，在线自蒸馏（OPSD）通过最小化未特权策略与特权自教师之间的前向 KL 散度，提供了密集的分布式监督，隐含假设结果的似然变化编码了可靠的答案对齐信息。我们通过固定每个采样轨迹并在两个对立的结果条件下重新评分来检验这一前提，一个条件断言正确性，另一个则断言错误性。在这两种条件下，大多数受影响的标记在同一方向上发生变化，符号反转较少，诱导的优化信号存在显著重叠。较大的变化也集中在高度可替代的表面形式标记上，而携带特定问题推理内容的标记则不太敏感。这些发现表明，特权变化未能提供可靠的答案对齐方向，而其幅度主要反映反事实敏感性而非标记级学习价值。基于这些观察，我们提出了反事实敏感性信用重新分配（CSCR），这是 GRPO 的一个简单扩展，减少对高度敏感标记的信用，并重新归一化标记级优势，以保持原始信用预算和验证者确定的方向。在长链数学推理基准测试中，CSCR 在相同数量的策略更新下始终优于 GRPO 基线。针对性的消融实验进一步证实了我们的诊断：特权诱导的方向不可靠，适度的下权重最为有效，而更强的调制则会使优化不稳定。
cs.AI / 56 / 2607.27891
Dynamic Spectral Filtering for Temporal Graph Learning: Learning Evolving Propagation Operators
动态谱过滤用于时间图学习：学习演变传播算子
Abstract
Temporal graph learning is commonly organized around the evolution of node states or the encoding of interaction histories. We study an underexplored, operator-centric question: should the graph propagation mechanism itself evolve over time? We introduce Dynamic Spectral Filtering (DSF), which represents propagation at snapshot t by a Chebyshev polynomial filter with vector-valued, time-dependent coefficients. DSF explicitly treats these compact multi-order coefficients as recurrent temporal states. A recurrent branch proposes updates, while multiplicative global and order-specific gates regulate their magnitude. The temporal state is independent of the number of nodes. On MOOC, Wikipedia, and Reddit temporal link-prediction benchmarks, converged DSF runs attain AP scores of 0.7851, 0.9088, and 0.9860, respectively, with 93K to 133K trainable parameters, 68 to 182 MB peak GPU memory, and 1.6 to 2.1 seconds of training per epoch. Against the closely related DEFT baseline, DSF is better on MOOC, within 0.001 AP on Reddit, and modestly lower on Wikipedia, while using 8.3 to 8.6 times fewer parameters, 25 to 33 times less GPU memory, and 5 to 19 times less time per epoch. Relative to all measured alternatives, it uses 3.3 to 38.6 times less GPU memory. These results support direct spectral-response evolution as a useful temporal inductive bias when computational efficiency is a first-class requirement.
Chinese Translation
时间图学习通常围绕节点状态的演变或交互历史的编码进行组织。我们研究了一个尚未深入探讨的以算子为中心的问题：图传播机制本身是否应该随时间演变？我们引入了动态谱过滤（Dynamic Spectral Filtering, DSF），它通过具有向量值、时间依赖系数的切比雪夫多项式过滤器来表示快照 t 时的传播。DSF 明确将这些紧凑的多阶系数视为递归时间状态。一个递归分支提出更新，而乘法全局和特定阶数的门控则调节其幅度。时间状态与节点数量无关。在 MOOC、Wikipedia 和 Reddit 的时间链接预测基准上，收敛的 DSF 运行分别达到了 0.7851、0.9088 和 0.9860 的 AP 分数，具有 93K 到 133K 的可训练参数，68 到 182 MB 的峰值 GPU 内存，以及每个周期 1.6 到 2.1 秒的训练时间。与密切相关的 DEFT 基线相比，DSF 在 MOOC 上表现更好，在 Reddit 上相差 0.001 AP，而在 Wikipedia 上略低，同时使用的参数少了 8.3 到 8.6 倍，GPU 内存少了 25 到 33 倍，每个周期的时间少了 5 到 19 倍。与所有测量的替代方案相比，它使用的 GPU 内存少了 3.3 到 38.6 倍。这些结果支持直接谱响应演变作为在计算效率为首要要求时的一种有用的时间归纳偏置。
cs.AI / 57 / 2607.27895
MMHBench: A Multi-Perspective Benchmark for Mental Health Understanding in Long-Form Videos
MMHBench：用于长视频心理健康理解的多视角基准
Abstract
Mental health understanding in long-form videos requires nuanced reasoning over observable behavior, interpersonal context, and latent psychological states. Existing benchmarks largely reduce this task to coarse-grained classification, providing limited insight into whether models truly understand psychological phenomena or rely on superficial correlations. To address this limitation, we introduce MMHBench, a comprehensive multimodal benchmark for multi-perspective mental health understanding, comprising 268 long-form videos and 2,184 carefully curated questions. MMHBench organizes the evaluation into two complementary settings: (1) third-person assessment, consisting of 605 questions that focus on the interpretation of observable behaviors and multimodal evidence, and (2) first-person perspective-taking, comprising 1,579 questions that require perspective-conditioned reasoning to identify the interpretation of the mental state supported by the available multimodal evidence. We propose a Multi-Agent Question Generation (MAQG) framework that simulates diverse social roles to synthesize questions from multiple perspectives. The generated questions are refined through multi-role feedback and iterative optimization, followed by expert-guided verification to ensure quality and validity. Extensive evaluation of 22 representative multimodal large language models (MLLMs), spanning both open-source and leading closed-source models, demonstrates that long-form video mental health understanding remains highly challenging.
Chinese Translation
在长视频中理解心理健康需要对可观察行为、人际背景和潜在心理状态进行细致的推理。现有基准大多将这一任务简化为粗粒度分类，提供的见解有限，无法判断模型是否真正理解心理现象或仅依赖表面相关性。为了解决这一局限性，我们提出了MMHBench，一个全面的多模态基准，用于多视角心理健康理解，包含268个长视频和2184个精心策划的问题。MMHBench将评估组织为两个互补的设置：（1）第三人称评估，包括605个问题，侧重于对可观察行为和多模态证据的解释；（2）第一人称视角采纳，包括1579个问题，要求进行视角条件推理，以识别由可用多模态证据支持的心理状态的解释。我们提出了一种多智能体问题生成（Multi-Agent Question Generation, MAQG）框架，模拟多种社会角色，从多个视角合成问题。生成的问题通过多角色反馈和迭代优化进行精炼，随后通过专家指导的验证确保质量和有效性。对22个代表性多模态大型语言模型（MLLMs）的广泛评估，涵盖开源和领先的闭源模型，表明长视频心理健康理解仍然具有很高的挑战性。
cs.AI / 58 / 2607.27910
A Cross-Architecture Audit of Direction-Based Inference-Time Defences in Vision-Language Models
跨架构审计：基于方向的视觉语言模型推理时防御
Abstract
Inference time defences against vision language model jailbreaks often subtract a calibrated direction from the residual stream at a chosen decoder layer. We compare five defence candidates across 15 model and layer cells from four architectural families under a magnitude controlled protocol that matches the intervention size for each prompt and pairs every direction with a random control of the same norm. The candidates are the mean image conditioning shift, a CMRM style refusal direction, a ShiftDC style attack specific residual, a prompt instruction to ignore the image, and a random control. No single candidate dominates on both refusal recovery and utility preservation. The image conditioning shift leads on LLaVA 1.5 and Pixtral 12B and is the only candidate whose utility loss remains at the measurement noise floor in every family. The prompt instruction leads on Qwen2.5 VL, while the attack specific residual leads on Qwen2 VL 2B. The image conditioning direction is direction specific in 13 of 15 cells, but strongly architecture specific and nontransferable across the only dimension compatible pair, LLaVA 1.5 13B and Pixtral 12B. We also connect text only and multimodal refusal geometry. The CMRM direction has positive cosine alignment with the image conditioning shift in all 15 cells, with mean 0.35, range 0.17 to 0.65, 15 to 25 times the random vector null, and a sign test p value of about 3e-5. These results show that the two recipes recover partially overlapping geometry and that direction based defences should be calibrated separately for each language decoder family.
Chinese Translation
针对视觉语言模型越狱的推理时防御通常在选定的解码器层从残差流中减去一个校准方向。我们在四个架构家族的15个模型和层单元中比较了五个防御候选者，采用了一个幅度受控的协议，该协议为每个提示匹配干预大小，并将每个方向与相同范数的随机对照配对。这些候选者包括均值图像条件偏移、CMRM风格的拒绝方向、ShiftDC风格的特定攻击残差、忽略图像的提示指令和随机对照。在拒绝恢复和效用保持方面，没有单一候选者占据绝对优势。图像条件偏移在LLaVA 1.5和Pixtral 12B上表现优异，并且是唯一一个在每个家族中效用损失保持在测量噪声底线的候选者。提示指令在Qwen2.5 VL上表现最佳，而攻击特定残差在Qwen2 VL 2B上表现最佳。图像条件方向在15个单元中的13个是特定于方向的，但在唯一兼容的维度对LLaVA 1.5 13B和Pixtral 12B之间表现出强烈的架构特异性和不可转移性。我们还将仅文本和多模态拒绝几何联系起来。CMRM方向与图像条件偏移在所有15个单元中具有正余弦对齐，均值为0.35，范围为0.17到0.65，随机向量零值的15到25倍，符号检验p值约为3e-5。这些结果表明，这两种方法恢复了部分重叠的几何形状，并且基于方向的防御应针对每个语言解码器家族单独进行校准。
cs.AI / 59 / 2607.27912
IFHierBench: Hierarchical Instruction Following for Large Language Models
IFHierBench：大语言模型的层次指令遵循
Abstract
Instruction-following ability is critical for deploying large language models in real-world applications, where downstream components depend on the output satisfying specific constraints. Modern deployments increasingly handle the full task in a single LLM call, with one prompt specifying a layered output whose overall artifact, structural sections, and nested fields must each satisfy concrete constraints. Existing instruction-following benchmarks treat the constraint set as a flat list applied uniformly to the response, so they cannot scope a check to a particular section of the output. We introduce IFHierBench, a hierarchical instruction-following benchmark of 600 prompts stratified across four constraint-tree depths and 35 distinct constraints, each prompt paired with a deterministic checker that verifies satisfaction at every scope. Evaluating seven leading proprietary and open-weight models, we find that even the strongest model only marginally exceeds 50% prompt-level accuracy and that accuracy degrades sharply as constraint depth grows. Reliably following nested constraints remains a substantial gap for current LLMs, motivating future training methods that consider constraint adherence at finer granularity to achieve better instruction-following ability.
Chinese Translation
指令遵循能力对于在实际应用中部署大语言模型至关重要，因为下游组件依赖于输出满足特定约束。现代部署越来越多地在一次 LLM 调用中处理完整任务，通过一个提示指定分层输出，其整体工件、结构部分和嵌套字段必须各自满足具体约束。现有的指令遵循基准将约束集视为统一应用于响应的平面列表，因此无法对输出的特定部分进行检查。我们提出了 IFHierBench，这是一个层次指令遵循基准，包含 600 个提示，分层于四个约束树深度和 35 个不同的约束，每个提示都配有一个确定性检查器，用于验证每个范围内的满足情况。评估七个领先的专有和开放权重模型，我们发现即使是最强的模型，其提示级准确率也仅略微超过 50%，并且随着约束深度的增加，准确率急剧下降。可靠地遵循嵌套约束仍然是当前 LLMs 的一个重大缺口，这促使未来的训练方法考虑在更细粒度上遵循约束，以实现更好的指令遵循能力。
cs.AI / 60 / 2607.27917
One Anchor for All: Unified Multilingual and Multimodal Safety Alignment for LVLMs
一个锚点适用于所有：统一的多语言和多模态安全对齐框架用于大型视觉语言模型
Abstract
As large vision-language models (LVLMs) are deployed globally, the combination of multilingual instructions and visual information makes malicious attacks more covert and sophisticated than ever before. However, existing methods isolate language and modality defenses, which, coupled with the scarcity of safety data and high fine-tuning costs, makes it difficult for models to defend against compound attacks. To address this severe challenge, we propose a neuron-level cross-dimensional safety alignment framework driven by modality- and language-shared safety neurons (MLS-Neurons). First, we identify monolingual and unimodal safety neurons by comparing responses to harmful and benign samples, quantifying functional saliency through activation strength and downstream impact. Then, by intersecting these unimodal neurons within each language, we extract modality-shared safety neurons (MS-Neurons) responsive to both visual and textual risks, bridging the safety representation gap between modalities. Furthermore, using English as a semantic anchor, we intersect MS-Neurons across languages to identify modality- and language-shared safety neurons (MLS-Neurons), serving as key defenses against compound attacks. Finally, we update only this minimal subset of shared neurons (~0.03% of parameters), transferring English-only safety supervision to multilingual and multimodal scenarios. Extensive experiments show that our method significantly outperforms state-of-the-art approaches across diverse multilingual and multimodal safety benchmarks while preserving general utility.
Chinese Translation
随着大型视觉语言模型（LVLMs）的全球部署，多语言指令与视觉信息的结合使得恶意攻击变得比以往任何时候都更加隐蔽和复杂。然而，现有方法将语言和模态防御孤立开来，加上安全数据的稀缺和高昂的微调成本，使得模型难以抵御复合攻击。为了解决这一严峻挑战，我们提出了一种基于模态和语言共享安全神经元（MLS-Neurons）的神经元级跨维度安全对齐框架。首先，我们通过比较对有害样本和良性样本的反应，识别单语和单模态安全神经元，量化功能显著性，通过激活强度和下游影响进行评估。然后，通过在每种语言内交集这些单模态神经元，我们提取出对视觉和文本风险均有反应的模态共享安全神经元（MS-Neurons），弥合模态之间的安全表示差距。此外，利用英语作为语义锚点，我们在不同语言间交集MS-Neurons，以识别模态和语言共享安全神经元（MLS-Neurons），作为抵御复合攻击的关键防御。最后，我们仅更新这一最小共享神经元子集（约0.03%的参数），将仅限于英语的安全监督转移到多语言和多模态场景中。大量实验表明，我们的方法在多样化的多语言和多模态安全基准测试中显著优于最先进的方法，同时保持了通用性。
cs.AI / 61 / 2607.27929
Meta-Task: Turning Terminal Task Synthesis into a Terminal Task for Scalable Agent Training
元任务：将终端任务合成转化为可扩展代理训练的终端任务
Abstract
Training terminal agents at scale requires diverse, verifiable terminal tasks and high-quality interaction trajectories, yet acquiring such data remains a significant challenge. Existing synthesis methods face two key limitations: (1) weak reliability caused by the disconnect between task generation and real execution, and (2) limited diversity and scalability due to dependence on existing repositories. We propose Meta-Task, a framework that redefines terminal task synthesis as a Terminal-Bench-format task itself: an agent operates within a real container environment to iteratively generate, execute, and verify tasks, so that synthesized components are checked for internal consistency and executability within the generation loop itself. Building upon this, we decouple the target task requirements along multiple dimensions, introduce a multi-phase mechanism that dynamically designs novel task specifications before producing the actual tasks, and incorporate optional external material support to enhance diversity and realism. We additionally apply LLM-as-Judge filtering to ensure the quality of the final training data. Experiments on Terminal-Bench 2.0 show that fine-tuning on only 3,221 Meta-Task synthesized trajectories achieves 22.5% and 31.8% Avg Pass@1 for Qwen3-14B and Qwen3-32B respectively, outperforming concurrent approaches with significantly less training data.
Chinese Translation
大规模训练终端代理需要多样化、可验证的终端任务和高质量的交互轨迹，然而获取这些数据仍然是一个重大挑战。现有的合成方法面临两个主要限制：（1）由于任务生成与实际执行之间的脱节，导致可靠性较弱；（2）由于依赖现有库，导致多样性和可扩展性有限。我们提出了元任务（Meta-Task），一个将终端任务合成重新定义为终端基准（Terminal-Bench）格式任务的框架：代理在真实的容器环境中操作，迭代生成、执行和验证任务，从而在生成循环中检查合成组件的内部一致性和可执行性。在此基础上，我们沿多个维度解耦目标任务要求，引入一种多阶段机制，在生成实际任务之前动态设计新任务规范，并结合可选的外部材料支持，以增强多样性和现实性。此外，我们还应用了大型语言模型（LLM）作为评判者的过滤机制，以确保最终训练数据的质量。在Terminal-Bench 2.0上的实验表明，仅对3,221条元任务合成轨迹进行微调，Qwen3-14B和Qwen3-32B的平均通过率（Avg Pass@1）分别达到了22.5%和31.8%，在训练数据显著更少的情况下超越了同时期的方法。
cs.AI / 62 / 2607.27933
The Geometric Nature and a Free Proxy for Flow-Matching Uncertainty
流匹配不确定性的几何性质及其自由代理
Abstract
Flow matching (FM) has become a popular action head paradigm for modern embodied models. However, as a conditional generative model, it does not explicitly expose its inherent uncertainty, producing faulty action chunks even when it misinterprets the scene or encounters out-of-distribution (OOD) inputs. Therefore, determining when an FM-generated action can be trusted is essential for safe deployment, yet existing uncertainty estimation methods on real-time control suffer from several issues: extra training budget, high computational overhead, and low generalization ability. In this work, we provide a geometric interpretation of FM uncertainty in the velocity field, showing that uncertainty manifests as deviation from an ideal affine-isotropic contraction field. Building on this observation, we introduce denoising acceleration ($\mathrm{accel}$), a highly-generalizable and cost-free uncertainty proxy that measures the bending of the denoising trajectory from a single forward pass, without additional model evaluations, training, or resampling. We theoretically and empirically demonstrate that $\mathrm{accel}$ is a faithful proxy for FM uncertainty and further test its utility in online failure detection. Results show that $\mathrm{accel}$ identifies failing rollouts well before termination, matching or even outperforming costly resampling- and training-based baselines across settings under realistic deployment budget. Code and demos available at: https://github.com/rrrrrrzy/fm-geometry.
Chinese Translation
流匹配（Flow Matching, FM）已成为现代具身模型的一种流行动作头范式。然而，作为一种条件生成模型，它并未明确揭示其固有的不确定性，即使在误解场景或遇到分布外（Out-of-Distribution, OOD）输入时，仍会产生错误的动作片段。因此，确定何时可以信任FM生成的动作对于安全部署至关重要，但现有的实时控制不确定性估计方法存在多个问题：额外的训练预算、高计算开销和低泛化能力。在本研究中，我们提供了FM不确定性在速度场中的几何解释，表明不确定性表现为偏离理想的仿射各向同性收缩场。基于这一观察，我们引入了去噪加速（denoising acceleration, $ ext{accel}$），这是一种高度可泛化且无成本的不确定性代理，能够通过单次前向传播测量去噪轨迹的弯曲，而无需额外的模型评估、训练或重采样。我们在理论和实证上证明了$ ext{accel}$是FM不确定性的可靠代理，并进一步测试其在在线故障检测中的实用性。结果表明，$ ext{accel}$能够在终止之前很好地识别出失败的展开，且在现实部署预算下，在各设置中与昂贵的重采样和训练基线相匹配，甚至表现更优。代码和演示可在：https://github.com/rrrrrrzy/fm-geometry获取。
cs.AI / 63 / 2607.27934
Shapes from Examples: Foundations of Shape Learning in Recursive SHACL
从示例中获取形状：递归 SHACL 中形状学习的基础
Abstract
SHACL shapes enable data graph validation, making automatic shape learning essential for knowledge graph applications. We investigate the well-known fitting approach to this task: given sets P and N of positive and negative example nodes from an input graph, compute a shape expression C, possibly using shape names defined in a recursive shape catalogue, that validates at every node in P and none in N. We focus on the case where C is written in a core fragment of SHACL corresponding to the Description Logic ELI. For the catalogue, we consider the well-founded, stable, and supported semantics. We address fitting existence and most specific fitting computation, establish tight exponential-time upper bounds for both problems, and obtain polynomial bounds for relevant special cases.
Chinese Translation
SHACL 形状使数据图的验证成为可能，因此自动形状学习对于知识图谱应用至关重要。我们研究了这一任务的著名拟合方法：给定输入图中的正例节点集合 P 和负例节点集合 N，计算一个形状表达式 C，可能使用在递归形状目录中定义的形状名称，该表达式在 P 中的每个节点上有效，而在 N 中的节点上无效。我们重点关注 C 在与描述逻辑 ELI 对应的 SHACL 核心片段中的情况。对于目录，我们考虑了良构、稳定和支持的语义。我们解决了拟合存在性和最特定拟合计算的问题，为这两个问题建立了严格的指数时间上界，并为相关特例获得了多项式界限。
cs.AI / 64 / 2607.27937
From Scoring to Acting: Outcome-Verified Comparative Self-Distillation for LLM Agents
从评分到行动：基于结果验证的比较自蒸馏用于大规模语言模型代理
Abstract
Recent work on LLM agents is shifting from external capability elicitation to capability internalization, enabling agents to retain useful skills without retrieval at inference time. On-policy self-distillation (OPSD) offers a promising direction, but many existing methods typically supervise students by scoring actions along student-generated trajectories. Such supervision has two limitations: teacher preferences are not validated by environment outcomes, and action-level scores underuse information from student rollouts, teacher rollouts, and their behavioral relationship. We therefore advocate outcome-verified teacher supervision and comparative learning over teacher-student trajectories. Based on this view, we propose Outcome-Verified Comparative Self-Distillation (OVCSD). OVCSD organizes failed student rollouts into a prefix tree, adaptively invokes a skill-conditioned teacher from student-reached states, and retains only outcome-verified successful continuations. It then applies localized comparative learning at the first state-aligned divergence and distills the post-divergence teacher suffix to transfer completion behavior. Experiments on ALFWorld and WebShop across three model scales show that OVCSD consistently outperforms skill-free RL and existing self-distillation baselines, achieving up to 29.7 and 5.4 absolute success-rate gains over the strongest baselines on ALFWorld and WebShop, respectively, while adding less than 3% privileged interaction during training.
Chinese Translation
近期对大规模语言模型（LLM）代理的研究正从外部能力引导转向能力内化，使代理能够在推理时保留有用的技能而无需检索。基于策略的自蒸馏（On-policy Self-Distillation, OPSD）提供了一个有前景的方向，但许多现有方法通常通过对学生生成的轨迹进行动作评分来监督学生。这种监督存在两个局限性：教师偏好未通过环境结果得到验证，且动作级评分未充分利用来自学生回放、教师回放及其行为关系的信息。因此，我们倡导基于结果验证的教师监督和教师-学生轨迹上的比较学习。基于这一观点，我们提出了结果验证的比较自蒸馏（Outcome-Verified Comparative Self-Distillation, OVCSD）。OVCSD将失败的学生回放组织成前缀树，从学生到达的状态中自适应调用技能条件教师，并仅保留结果验证的成功延续。然后，它在第一个状态对齐的分歧处应用局部比较学习，并蒸馏分歧后的教师后缀以转移完成行为。在ALFWorld和WebShop的三种模型规模上的实验表明，OVCSD在性能上始终优于无技能强化学习（RL）和现有自蒸馏基线，在ALFWorld和WebShop上分别实现了高达29.7和5.4的绝对成功率提升，同时在训练过程中增加的特权交互少于3%。
cs.AI / 65 / 2607.27967
MARS-RA: Rank Aggregation for Credit Assignment via Multimodal Comparisons in Embodied Multi-Agent Cooperation
MARS-RA：通过多模态比较进行信用分配的排名聚合
Abstract
Credit assignment is a fundamental challenge in cooperative multi-agent reinforcement learning, particularly in embodied AI settings characterized by limited and delayed feedback as well as dynamically changing numbers of active agents. We propose MARS-RA, a framework that reformulates credit assignment as a rank aggregation problem using contribution-based pairwise comparisons among agents generated by large multimodal models. This shift from absolute to relative estimation ensures robustness against noise and dynamic agent participation, converting comparison results into contribution scores for potential-based reward shaping. We provide theoretical justification for the convergence and robustness of the proposed framework, and show that Shapley values can be used as an interpretive reference. Experimental results on challenging tasks of different types indicate that MARS-RA can guide agents toward effective cooperation.
Chinese Translation
信用分配是合作多智能体强化学习中的一个基本挑战，尤其是在具有限制和延迟反馈以及动态变化的活跃智能体数量的具身人工智能环境中。我们提出了MARS-RA，一个将信用分配重新表述为排名聚合问题的框架，该框架利用大型多模态模型生成的智能体之间基于贡献的成对比较。这种从绝对估计到相对估计的转变确保了对噪声和动态智能体参与的鲁棒性，将比较结果转换为潜在奖励塑形的贡献分数。我们为所提框架的收敛性和鲁棒性提供了理论依据，并表明Shapley值可以作为解释性参考。在不同类型的挑战性任务上的实验结果表明，MARS-RA能够引导智能体实现有效的合作。
cs.AI / 66 / 2607.27984
Share the Judge, Learn the Deferral: Where Specialization Helps LLM Evaluation
共享评判者，学习延迟：专业化如何促进大型语言模型评估
Abstract
Agentic systems have widened the gap between producing candidate outputs and reviewing them. This paper asks a practical architectural question: should domain specialization be built into an evaluator's weights, or into the rule that decides when its judgment can be trusted? We study 99,952 public, rubric-conditioned examples. Supplying the correct rubric improves locked-test accuracy by 2.11 points over a response-only control; replacing it with an unrelated rubric costs 2.66 points. Dividing the same training corpus among eight criterion-family LoRA judges, however, loses 10.05 points and cuts audited coverage at a 5% risk target from 24.44% to 5.43%. Matching the bank's stored capacity with one rank-64 adapter does not reproduce this loss. Nor is the result explained by learning rate or optimizer steps. Initializing the family adapters from a shared, trained judge recovers test accuracy to 76.85%, 19.94 points above scratch training at the same learning rate (95% interval 18.88-21.02). The result changes when specialization governs deferral rather than judgment. On RewardBench 2, learned correctness heads route examples through a 0.6B-4B-8B cascade without changing any reward score. Across 20 locked repartitions, the cascade attains 89.40% accuracy, compared with 84.75% for 8B alone, at 0.415 normalized parameter compute. Every run passes an exact one-sided 95% risk audit; margin-based rules remain near 84.8% accuracy while using at least 0.94 compute. These results suggest a qualified design rule: share the learning of judgment until there is enough data to justify a split, and place domain-specific adaptation in an audited release boundary.
Chinese Translation
代理系统扩大了生成候选输出与审查它们之间的差距。本文提出了一个实用的架构问题：领域专业化应该内置于评估者的权重中，还是内置于决定何时可以信任其判断的规则中？我们研究了99,952个公共、基于评分标准的示例。提供正确的评分标准使得锁定测试的准确性比仅响应控制提高了2.11分；用不相关的评分标准替换则损失了2.66分。然而，将相同的训练语料库分配给八个标准家庭的LoRA评审者，损失了10.05分，并将以5%的风险目标进行审计的覆盖率从24.44%降低到5.43%。与一个rank-64适配器匹配银行的存储容量并未重现这一损失。学习率或优化步骤也无法解释这一结果。从一个共享的、经过训练的评审者初始化家庭适配器，使测试准确性恢复到76.85%，比在相同学习率下的初始训练高出19.94分（95%置信区间为18.88-21.02）。当专业化控制延迟而非判断时，结果发生变化。在RewardBench 2上，学习到的正确性使得示例通过一个0.6B-4B-8B的级联而不改变任何奖励分数。在20个锁定重分配中，该级联达到了89.40%的准确性，而仅使用8B时的准确性为84.75%，计算标准化参数为0.415。每次运行都通过了一个精确的单侧95%风险审计；基于边际的规则在使用至少0.94计算时仍保持在84.8%的准确性。这些结果提示了一条有条件的设计规则：共享判断的学习，直到有足够的数据来证明分裂的合理性，并将领域特定的适应放置在经过审计的发布边界内。
cs.AI / 67 / 2607.27994
SKIMIX: Multi-Agent Harness-Time Scaling with Skill Mixture for Dynamic Harness Engineering
SKIMIX：基于技能混合的多智能体时间缩放用于动态工程设计
Abstract
AI agents increasingly rely on large skill libraries, but selecting, combining, and maintaining skills remains difficult. We propose SKIMIX, a multi-agent framework in which agents with different skill portfolios collaborate through iterative refinement. SKIMIX combines embedding-based skill retrieval, submodular anti-dilution routing, and adaptive skill evolution. Across six reasoning benchmarks, multi-agent collaboration substantially improves open-ended mathematical reasoning but offers limited or negative gains on multiple-choice tasks. Agent-count scaling is non-monotonic, and most improvements arise during the first refinement round. These results show that task characteristics determine whether skill-level ensembles help and provide practical guidance for scalable agent design.
Chinese Translation
人工智能代理越来越依赖于大型技能库，但选择、组合和维护技能仍然困难。我们提出了SKIMIX，一个多智能体框架，其中具有不同技能组合的代理通过迭代优化进行协作。SKIMIX结合了基于嵌入的技能检索、次模抗稀释路由和自适应技能演化。在六个推理基准测试中，多智能体协作显著提高了开放式数学推理的能力，但在多项选择任务中提供的增益有限或为负。代理数量的扩展呈现非单调性，大多数改进发生在第一次优化轮次中。这些结果表明，任务特征决定了技能级别组合是否有助于提升性能，并为可扩展代理设计提供了实用指导。
cs.AI / 68 / 2607.28006
MMLDSum-LLM: Multimodal Long-Document Summarization with Visual-Alignment and Keyword-Aware
MMLDSum-LLM：具有视觉对齐和关键词感知的多模态长文档摘要
Abstract
Multimodal long documents are core carriers of professional knowledge, where critical evidence is sparsely distributed across paragraphs and modalities. This easily causes key information omission and cross-modal hallucinations in summarization by multimodal LLMs. These issues stem from attention drift in long-range dependency modeling and gaps in inter-modal alignment. To address this, we introduce MMLDSum-Bench, a high-quality benchmark for multimodal long-document summarization, covering multiple domains, context-length scales, and visual-textual modality distributions. We further propose MMLDSum-LLM, a reproducible two-stage training framework that combines supervised fine-tuning with visual-alignment weighted loss and keyword-aware weighted loss, followed by GRPO with a multi-objective reward (keyword coverage, image-text alignment, ROUGE, and length control). Extensive experiments on MMLDSum-Bench, comparing against leading closed-source and open-source multimodal models under a unified evaluation protocol - including LLM-as-a-judge scoring, atomic-claim precision/recall, image-text alignment (ITA), and ROUGE - demonstrate that our approach significantly improves key-information coverage and cross-modal consistency.
Chinese Translation
多模态长文档是专业知识的核心载体，其中关键证据在段落和模态之间稀疏分布。这容易导致多模态大语言模型（LLMs）在摘要中遗漏关键信息和产生跨模态幻觉。这些问题源于长距离依赖建模中的注意力漂移和模态间对齐的差距。为了解决这些问题，我们引入了MMLDSum-Bench，这是一个高质量的多模态长文档摘要基准，涵盖多个领域、上下文长度尺度和视觉-文本模态分布。我们进一步提出了MMLDSum-LLM，这是一个可复现的两阶段训练框架，结合了带有视觉对齐加权损失和关键词感知加权损失的监督微调，随后通过具有多目标奖励（关键词覆盖、图像-文本对齐、ROUGE和长度控制）的GRPO进行训练。在MMLDSum-Bench上进行的大量实验，针对领先的闭源和开源多模态模型在统一评估协议下进行比较，包括LLM作为评判者的评分、原子主张的精确度/召回率、图像-文本对齐（ITA）和ROUGE，结果表明我们的方法显著改善了关键信息的覆盖率和跨模态一致性。
cs.AI / 69 / 2607.28030
MUL-T: Decoding Spatial Cellular Architecture in Multiplexed Tissue Images
MUL-T：解码多重组织图像中的空间细胞结构
Abstract
Understanding tissue organisation in multiplexed imaging requires modelling both cellular phenotypes and their spatial context. Existing approaches typically rely on handcrafted features, such as marker intensity statistics or cell-type proportions, which often fail to scale or generalise across cohorts with heterogeneous marker panels. We introduce MUL-T, a lightweight transformer framework that reframes tissue architecture as a masked contextual prediction task over discrete cell tokens. By learning contextualised [CLS] embeddings without task-specific supervision, the model captures higher-order cellular interactions while remaining computationally efficient. We evaluate MUL-T on several clinically relevant downstream tasks, including core-level tumour pattern classification, patient-level grading, PD-L1 positivity prediction, and cross-dataset treatment response prediction. Across tasks, MUL-T consistently outperforms classical feature-based baselines and achieves performance comparable to a foundation ViT model, despite substantially fewer parameters and lower training cost.
Chinese Translation
理解多重成像中的组织结构需要对细胞表型及其空间背景进行建模。现有的方法通常依赖于手工特征，例如标记强度统计或细胞类型比例，这些方法往往无法在具有异质标记面板的队列中扩展或推广。我们提出了MUL-T，一个轻量级的变换器框架，将组织结构重新构建为对离散细胞标记的掩蔽上下文预测任务。通过在没有特定任务监督的情况下学习上下文化的[CLS]嵌入，该模型捕捉了高阶细胞相互作用，同时保持计算效率。我们在多个临床相关的下游任务上评估了MUL-T，包括核心级肿瘤模式分类、患者级分级、PD-L1阳性预测以及跨数据集治疗反应预测。在各项任务中，MUL-T始终优于经典的基于特征的基线，并且尽管参数显著更少且训练成本更低，其性能与基础ViT模型相当。
cs.AI / 70 / 2607.28033
DataClawEval: A Benchmark for Data Engineering Agents in Real Industrial Harness
DataClawEval：用于真实工业环境中数据工程代理的基准测试
Abstract
Large language models (LLMs) and LLM-based agents are increasingly being deployed to automate complex workflows, promising to revolutionize data management and processing. However, existing benchmarks predominantly focus on simplified Text-to-SQL translation or data analysis, leaving the critical and complex domain of end-to-end data engineering largely unexplored. To bridge this gap, we introduce DataClawEval, the first comprehensive benchmark designed specifically to evaluate the end-to-end task completion capabilities of autonomous agents in real-world data engineering scenarios. Built upon production-grade code authored by professional enterprise data engineers, it comprises 100 rigorous, end-to-end tasks spanning five execution engines: PySpark, MySQL, HiveSQL, PrestoSQL/Trino, and FlinkSQL. Rather than non-deterministic LLM-as-a-judge scoring, each task is executed within a case-specific, isolated sandbox and graded by deterministic, rule-based scripts. Evaluating 16 frontier agents exposes critical limitations: The strongest model attains only 74.9 overall, and no single model dominates, as each excels on a different engine, revealing strict domain specialization rather than omnipotent proficiency. Thus, autonomous data engineering remains a formidable, unresolved challenge. We release our dataset, containerized environments, and deterministic evaluation scripts at https://github.com/Dicemy/DataClawEval/tree/master
Chinese Translation
大型语言模型（LLMs）及基于LLM的代理正在越来越多地被部署以自动化复杂工作流程，承诺将彻底改变数据管理和处理。然而，现有的基准测试主要集中在简化的文本到SQL（Text-to-SQL）翻译或数据分析上，导致端到端数据工程这一关键且复杂领域大多未被探索。为填补这一空白，我们推出了DataClawEval，这是第一个专门设计用于评估自主代理在真实数据工程场景中端到端任务完成能力的综合基准测试。该基准建立在由专业企业数据工程师撰写的生产级代码基础上，包含100个严格的端到端任务，涵盖五个执行引擎：PySpark、MySQL、HiveSQL、PrestoSQL/Trino和FlinkSQL。与非确定性的LLM作为评判者的评分方式不同，每个任务在特定案例的隔离沙箱中执行，并由确定性的基于规则的脚本进行评分。对16个前沿代理的评估揭示了关键的局限性：最强模型的总体得分仅为74.9，并且没有单一模型占据主导地位，因为每个模型在不同的引擎上表现优异，显示出严格的领域专业化而非全能的能力。因此，自主数据工程仍然是一个艰巨且未解决的挑战。我们在https://github.com/Dicemy/DataClawEval/tree/master发布了我们的数据集、容器化环境和确定性评估脚本。
cs.AI / 71 / 2607.28048
SKILL-KD: Contrastive Skill Distillation for LLM Agents
SKILL-KD：用于大语言模型代理的对比技能蒸馏
Abstract
Skill-based prompting has become a practical mechanism for improving large language model (LLM) agents, yet existing skill acquisition methods often treat skills as experience summaries, memory entries, or direct summaries of successful demonstrations. This creates a mismatch for weaker student agents: when a student fails because it lacks task knowledge or operational strategy, its failed trajectory may not contain enough evidence to infer the missing behavior, while the teacher trajectory may be too implicit to be internalized as reusable guidance. We propose SKILL-KD, a contrastive skill distillation framework that treats skills as an explicit distillation medium between agents of different capabilities. Given a student failure and the teacher trajectory on the same task, SKILL-KD distills their actionable discrepancy into a textual skill patch, evaluates the patch by re-running the student, and iteratively refines the patch when the student still fails. To prevent repeated local updates from causing skill drift, SKILL-KD further maintains trace-linked edit histories and performs Drift-Aware Skill Consolidation, deciding whether each patch should add a new rule, delete or modify an existing rule, or be skipped. Across five agent benchmarks and two student settings, SKILL-KD consistently improves frozen student agents over fixed-model adaptation baselines.
Chinese Translation
基于技能的提示已成为提升大语言模型（LLM）代理的实用机制，但现有的技能获取方法往往将技能视为经验总结、记忆条目或成功示范的直接总结。这导致较弱的学生代理出现不匹配：当学生因缺乏任务知识或操作策略而失败时，其失败轨迹可能没有足够的证据来推断缺失的行为，而教师轨迹可能过于隐晦，无法内化为可重用的指导。我们提出了SKILL-KD，一种对比技能蒸馏框架，将技能视为不同能力代理之间的显性蒸馏媒介。给定学生的失败和教师在同一任务上的轨迹，SKILL-KD将它们的可操作差异蒸馏为文本技能补丁，通过重新运行学生来评估补丁，并在学生仍然失败时迭代地完善补丁。为了防止重复的局部更新导致技能漂移，SKILL-KD进一步维护追踪链接的编辑历史，并执行漂移感知技能整合，决定每个补丁是否应添加新规则、删除或修改现有规则，或被跳过。在五个代理基准和两个学生设置中，SKILL-KD始终优于固定模型适应基线，显著提升了冻结学生代理的表现。
cs.AI / 72 / 2607.28050
IndustryForge-27B: A Domain-Enhanced Multimodal Foundation Model for Industrial CAD
IndustryForge-27B：一种针对工业CAD的领域增强多模态基础模型
Abstract
Automating industrial CAD design and manufacturing places distinctive demands on multimodal foundation models: the model must see engineering drawings and 3D geometry screenshots, write correct parametric-modelling scripts and Windows COM API code, and cover the full range from single parts to assemblies. General-purpose multimodal models fall short on these tasks, while single-task fine-tuning is too narrow to support the diverse calls that upper-layer agents issue. We build IndustryForge-27B on top of Qwen3.5-VL-27B by curating and integrating six industrial-CAD sub-corpora totalling $\sim$52k multimodal samples---CAD Visual QA (CAD-VQA), parametric CAD code (text2cadquery), assembly-level CAD code (text2cadquery-assembly), and three COM sub-corpora for Inventor / SolidWorks (com_2d / com_3d / com_assembly)---and training with a unified multi-task SFT recipe. Across four CAD-domain benchmarks IndustryForge-27B lifts the base model by $+33.65$~pp on average and outperforms the strong closed-source model GPT-5.4 on all four; across eleven general-capability benchmarks it retains, and slightly improves upon, the base model ($+1.56$~pp mean, no catastrophic forgetting). IndustryForge-27B will serve as the common substrate for downstream industrial-agent projects, providing a unified starting point for a full-stack industrial agent that spans from CAD design to industrial-software operation, from parts to assemblies, and from single-shot generation to closed-loop self-improvement.
Chinese Translation
工业CAD设计和制造的自动化对多模态基础模型提出了独特的要求：模型必须能够理解工程图纸和3D几何截图，编写正确的参数化建模脚本和Windows COM API代码，并涵盖从单个零件到装配体的全范围。通用多模态模型在这些任务上表现不足，而单一任务的微调又过于狭窄，无法支持上层代理发出的多样化请求。我们在Qwen3.5-VL-27B的基础上构建了IndustryForge-27B，通过策划和整合六个工业CAD子语料库，总计约52,000个多模态样本——CAD视觉问答（CAD-VQA）、参数化CAD代码（text2cadquery）、装配级CAD代码（text2cadquery-assembly），以及针对Inventor/SolidWorks的三个COM子语料库（com_2d / com_3d / com_assembly）——并采用统一的多任务SFT训练方案。在四个CAD领域基准测试中，IndustryForge-27B平均提升了基础模型33.65个百分点，并在所有四个测试中超越了强大的闭源模型GPT-5.4；在十一项通用能力基准测试中，它保持并略微改善了基础模型（平均提升1.56个百分点，未出现灾难性遗忘）。IndustryForge-27B将作为下游工业代理项目的共同基础，为从CAD设计到工业软件操作、从零件到装配体、从单次生成到闭环自我改进的全栈工业代理提供统一的起点。
cs.AI / 73 / 2607.28069
SemPIC: Learning Semantic Position-Independent KV Caches
SemPIC：学习语义位置无关的键值缓存
Abstract
Long-context retrieval and agentic workloads repeatedly reuse the same documents under changing instructions, histories, and document orders. Prefix caching cannot exploit this reuse, while position-independent caching (PIC) remains unreliable because independently compiled KV states lack the future context in which they will be consumed. Our diagnostics show that a learned boundary-conditioned baseline sharply reduces attention deviation near reusable-block boundaries but leaves interior and task-level residuals, motivating adaptation of the document representation itself. We present \emph{SemPIC}, which trains a LoRA-enabled Writer to compile native per-layer document KVs through behavioral distillation while retaining the pretrained decoder as an unchanged Reader. Adaptation is confined to offline cache construction, preserving the standard KV interface and cache-hit decoding path. We further introduce KV Gradient Checkpointing, which reduces peak training memory without severing gradients through cached KVs. Across three models and four tasks, SemPIC raises mean micro-F1 over KV Packet from 0.53 to 0.60, approaching Full Recompute at 0.62.
Chinese Translation
长上下文检索和自主工作负载在变化的指令、历史和文档顺序下反复使用相同的文档。前缀缓存无法利用这种重用，而位置无关缓存（PIC）由于独立编译的键值（KV）状态缺乏将要使用的未来上下文而变得不可靠。我们的诊断显示，学习的边界条件基线在可重用块边界附近显著减少了注意力偏差，但在内部和任务级别留下了残余，这促使我们对文档表示本身进行适应。我们提出了 extit{SemPIC}，它通过行为蒸馏训练一个支持LoRA的写入器，以编译本层文档的原生键值，同时保持预训练的解码器作为不变的读取器。适应仅限于离线缓存构建，保留标准的键值接口和缓存命中解码路径。我们进一步引入了键值梯度检查点（KV Gradient Checkpointing），在不切断通过缓存键值的梯度的情况下减少峰值训练内存。在三个模型和四个任务中，SemPIC将键值包（KV Packet）的平均微F1从0.53提高到0.60，接近完全重计算的0.62。
cs.AI / 74 / 2607.28074
Echoverse: Deep, Evolving Environments for Training Computer-Use Agents at Scale
Echoverse：用于大规模训练计算机使用代理的深度进化环境
Abstract
Computer-use agents learn from what their actions change, so training one needs applications it can act on, break and reset. The applications that matter most are login-gated and stateful, so synthetic environments stand in for them. Recent pipelines generate such environments in bulk, which moves the bottleneck from how many exist to what is inside each one. The returns, we find, come from three properties: how much behavioural depth an environment carries, whether it targets the interaction an agent actually fails, and whether it improves alongside the model. We present Echoverse, which compiles specifications into stateful applications whose tasks are graded against the application's own database, and a co-evolution loop that reads every graded rollout twice: as repairs to the environment, its tasks and its verifier, and as training signal for the model. Trained on twelve such environments, a 9B model improves from $36.5\%$ to $67.1\%$ across fourteen evaluation splits, within fourteen points of the much larger frontier model that taught it. We examine each property in turn. On the same domains, shallow environments push live-site accuracy below the base model ($80.0 \to 75.0$) while deep ones raise it ($80.0 \to 85.0$ and $48.0 \to 65.0$); drilling one interface control across many renderings transfers to held-out widget families and to the open web; and repairing a single environment lifts the model trained on it from $16.2\%$ to $38.5\%$. The same worlds serve as reinforcement-learning environments, where a reward combining the grounded verifier with a dense per-step judge raises held-out score from $58.8\%$ to $68.0\%$. We release four environments as a benchmark, with their applications, seed data and grounded graders. Code: https://aka.ms/echoverse
Chinese Translation
计算机使用代理通过其行为所改变的事物进行学习，因此训练一个代理需要能够进行操作、破坏和重置的应用程序。最重要的应用程序是需要登录并且具有状态的，因此合成环境代替了这些应用程序。最近的管道批量生成此类环境，这将瓶颈从存在多少个转移到每个环境内部的内容。我们发现，收益来自三个特性：环境所承载的行为深度、是否针对代理实际失败的交互以及是否随着模型的改进而提升。我们提出了Echoverse，它将规格编译成状态应用程序，其任务是根据应用程序自己的数据库进行评分，并且有一个共同进化循环，针对每个评分的结果进行两次读取：作为对环境、其任务和验证者的修复，以及作为模型的训练信号。在十二个这样的环境上训练后，一个9B模型在十四个评估分割中从$36.5\%$提高到$67.1\%$，与教它的更大前沿模型相差仅十四个点。我们逐一检查每个特性。在相同领域中，浅层环境将实时站点的准确性压低到基线模型以下（$80.0 \to 75.0$），而深层环境则提高了准确性（$80.0 \to 85.0$ 和 $48.0 \to 65.0$）；在多个渲染中钻取一个接口控制可以转移到保留的小部件系列和开放网络；修复单个环境使得在其上训练的模型从$16.2\%$提升到$38.5\%$。相同的世界作为强化学习环境，其中结合了基础验证者与密集逐步评判的奖励将保留分数从$58.8\%$提高到$68.0\%$。我们发布了四个环境作为基准，包括它们的应用程序、种子数据和基础评分者。代码链接：https://aka.ms/echoverse
cs.AI / 75 / 2607.28076
Group-Reflective Self-Distillation for Agentic Reinforcement Learning
用于自主强化学习的群体反思自蒸馏
Abstract
Reinforcement learning with verifiable rewards (RLVR) is effective for training large language model agents. However, terminal rewards provide only coarse trajectory-level supervision, leaving successful behaviors, recurring mistakes, and incidental choices entangled in the same outcome signal. Existing agentic self-distillation methods enrich sparse supervision with natural-language skills, but skills retrieved externally or extracted from a single trajectory by stronger models may mismatch current experience, exceed the policy's capability, or remain path-specific. We propose Group-Reflective Self-Distillation (GRSD), which derives capability-aligned and outcome-discriminative guidance from the policy's own verified rollouts. For each prompt, the policy reflects on each verified trajectory in an on-policy group, and a stop-gradient snapshot contrasts the resulting reflections from successful and failed rollouts to construct group-level privileged guidance. Conditioned on this guidance, a self-teacher refines turn-level credit assignment by modulating outcome-based advantages while preserving the verifier-determined learning direction. Experiments across multiple agentic environments and model scales demonstrate that GRSD consistently outperforms competitive baselines and generalizes more effectively to unseen tasks.
Chinese Translation
带有可验证奖励的强化学习（RLVR）在训练大型语言模型代理方面效果显著。然而，终端奖励仅提供粗略的轨迹级监督，使得成功行为、重复错误和偶然选择混杂在同一结果信号中。现有的自主自蒸馏方法通过自然语言技能丰富稀疏监督，但从外部检索的技能或由更强模型从单一轨迹提取的技能可能与当前经验不匹配，超出策略的能力，或仍然具有路径特定性。我们提出了群体反思自蒸馏（GRSD），该方法从策略自身的验证回放中推导出能力对齐和结果区分的指导。对于每个提示，策略在一个在线政策组中反思每个验证轨迹，并通过停止梯度快照对比成功和失败回放的反思，以构建群体级特权指导。在此指导的条件下，自我教师通过调节基于结果的优势来细化回合级的信用分配，同时保持验证者确定的学习方向。跨多个自主环境和模型规模的实验表明，GRSD始终优于竞争基线，并在未见任务上更有效地泛化。
cs.AI / 76 / 2607.28086
Distilling Answer Set Programming Theories from Large Language Models
从大型语言模型中提炼答案集编程理论
Abstract
Writing Answer Set Programming (ASP) theories from scratch is a difficult and time-consuming task. We take a neurosymbolic approach to study whether a model can distill complete and correct theories, given a fixed agent harness with the solver in the loop. The protocol is dataset-agnostic: with a single prompt and an empty file as the starting point the model is given a 1-hour time limit to derive a complete theory. We chose VQA as the application domain, three benchmarks (CLEVR, GQA, CLEVRER), as these are publicly available and non-trivial. In order to study the model scale required for solving this task we nine different models: four frontier (Claude Sonnet 4.6, Claude Opus 4.7, GPT-5, DeepSeek V4 Pro), two mid-tier (DeepSeek V4 Flash, gpt-oss-120b), and three open-weights (qwen3.6-27b, gpt-oss-20b, qwen3.5-9b). Three of four frontier models reach 100% on CLEVR and 92.8%-98.8% on GQA; on CLEVRER, Sonnet, Opus, DeepSeek V4 Pro score 92.7%-95.3%. GPT-5 reaches 98.7% on CLEVR but drops to 41.8% on GQA and to 86.7% on CLEVRER. Adding handwritten reference theories from other datasets moves the other three frontier models by at most +/-3.4 pp but reduces GPT-5's accuracy by 3-19 pp. We release the code, prompts, and theories distilled.
Chinese Translation
从头编写答案集编程（ASP）理论是一项困难且耗时的任务。我们采用神经符号方法研究模型是否能够在固定的代理环境中提炼出完整且正确的理论，同时将求解器纳入其中。该协议与数据集无关：在一个单一的提示和一个空文件作为起点的情况下，模型被给予1小时的时间限制来推导出完整的理论。我们选择了视觉问答（VQA）作为应用领域，并使用三个基准（CLEVR、GQA、CLEVRER），因为这些基准是公开可用且具有一定难度的。为了研究解决此任务所需的模型规模，我们测试了九种不同的模型：四种前沿模型（Claude Sonnet 4.6、Claude Opus 4.7、GPT-5、DeepSeek V4 Pro）、两种中等水平模型（DeepSeek V4 Flash、gpt-oss-120b）和三种开放权重模型（qwen3.6-27b、gpt-oss-20b、qwen3.5-9b）。四种前沿模型中有三种在CLEVR上达到了100%的准确率，在GQA上达到了92.8%-98.8%；在CLEVRER上，Sonnet、Opus和DeepSeek V4 Pro的得分为92.7%-95.3%。GPT-5在CLEVR上达到了98.7%的准确率，但在GQA上下降至41.8%，在CLEVRER上为86.7%。从其他数据集中添加手写参考理论使其他三种前沿模型的表现最多变化±3.4个百分点，但降低了GPT-5的准确率3-19个百分点。我们发布了代码、提示和提炼出的理论。
cs.AI / 77 / 2607.28087
Diversifying Personalized Research Ideation against AI-Induced Homogenization
针对人工智能引发的同质化现象的个性化研究创意多样化
Abstract
AI-assisted research ideation has emerged as a promising paradigm for accelerating scientific discovery, with systems now capable of generating research directions conditioned on papers, topics, or lightweight researcher contexts. Yet current systems largely optimize individual suggestions in isolation. This leaves two blind spots. First, coarse researcher representations may elicit mainstream directions that appear broadly feasible, but lack sufficient researcher-specific grounding. Second, independent recommendations can concentrate a community's portfolio around recurring high-probability themes. To address these blind spots, we propose DivAlign, a four-stage pipeline for alignment-preserving de-homogenization. DivAlign extracts fine-grained researcher profiles, generates profile-conditioned candidate directions, scores them along three alignment dimensions (Executability, Comprehensibility, and Growth Potential), and surfaces researcher-local directions while reducing redundancy across the community portfolio. On a benchmark we construct from 95 AI researchers across five subfields, DivAlign reduces community-level redundancy while preserving researcher-direction fit. Compared with coarse single-shot ideation, it lowers average pairwise similarity from 0.331 to 0.294 and nearest-neighbor similarity from 0.704 to 0.608. Compared with the independent top-choice variant, DivAlign reduces nearest-neighbor similarity from 0.663 to 0.608 while retaining 99.9% of the researcher-direction fit score. Code and data are available at https://github.com/Ruixxxx/DivAlign.
Chinese Translation
人工智能辅助的研究创意已成为加速科学发现的有前景的范式，现有系统能够基于论文、主题或轻量级研究者背景生成研究方向。然而，目前的系统在很大程度上是孤立地优化个别建议。这留下了两个盲点。首先，粗略的研究者表征可能会引发看似广泛可行的主流方向，但缺乏足够的研究者特定基础。其次，独立的推荐可能会使一个社区的投资组合集中在重复出现的高概率主题上。为了解决这些盲点，我们提出了DivAlign，一个四阶段的管道，用于保持对齐的去同质化。DivAlign提取细粒度的研究者档案，生成基于档案的候选方向，沿着三个对齐维度（可执行性、可理解性和增长潜力）对其进行评分，并在减少社区投资组合冗余的同时，呈现研究者本地方向。在我们从五个子领域的95位人工智能研究者构建的基准上，DivAlign减少了社区层面的冗余，同时保持了研究者与方向的契合度。与粗略的单次创意相比，它将平均成对相似度从0.331降低到0.294，将最近邻相似度从0.704降低到0.608。与独立的首选变体相比，DivAlign将最近邻相似度从0.663降低到0.608，同时保留了99.9%的研究者与方向的契合度评分。代码和数据可在 https://github.com/Ruixxxx/DivAlign 获取。
cs.AI / 78 / 2607.28090
PerturbMap: Cross-Context Transfer of Single-Cell Perturbation Responses
PerturbMap：单细胞扰动响应的跨上下文转移
Abstract
Single-cell perturbation atlases rarely measure every intervention in every cellular context: a query perturbation is often observed in one or more source contexts but missing in the recipient context where its effect is needed. Ignoring those measured responses discards query-specific experimental evidence, whereas copying or weakly calibrating them across contexts risks transferring the wrong signal. We propose PerturbMap, which predicts a missing recipient-context effect by combining a recipient-local low-rank base with accepted proposals that transport the same perturbation's measured source responses through source-to-recipient ridge experts fit on paired training perturbations, with proposal weights determined by route reliability estimated on validation anchors. On the Perturb-CITE-seq melanoma cohort, PerturbMap improves full-effect MSE by 4.1\% over a recipient-local low-rank base and achieves lower MSE than FedAvg, zero-response, raw-copy, calibrated-copy, and identity-shuffled affine controls. It remains within $2.82\times10^{-6}$ MSE of our centralized token-matched pooled reference, which uses a stronger training interface. A condition-mean specificity diagnostic shows the same direction: same-recipient top-10 counterpart retrieval by cosine increases from 74.5\% for the low-rank base to 80.5\% for PerturbMap.
Chinese Translation
单细胞扰动图谱很少在每个细胞上下文中测量每个干预：查询扰动通常在一个或多个源上下文中被观察到，但在需要其效应的接收上下文中缺失。忽视这些测量响应会丢弃查询特定的实验证据，而在上下文之间复制或弱校准它们则有转移错误信号的风险。我们提出了PerturbMap，它通过结合接收者局部低秩基与接受的提议，预测缺失的接收上下文效应，这些提议通过在配对训练扰动上拟合的源到接收的岭专家传输相同扰动的测量源响应，提议权重由在验证锚点上估计的路径可靠性决定。在Perturb-CITE-seq黑色素瘤队列中，PerturbMap相较于接收者局部低秩基提高了4.1%的全效应均方误差（MSE），并且其MSE低于FedAvg、零响应、原始复制、校准复制和身份打乱的仿射控制。它的MSE仍然在$2.82 imes10^{-6}$的范围内，与我们使用更强训练接口的集中式令牌匹配池参考相符。条件均值特异性诊断显示出相同的趋势：通过余弦相似度检索的同接收者前10个对应体的检索率从低秩基的74.5%增加到PerturbMap的80.5%。
cs.AI / 79 / 2607.28094
An Instrument to Evaluate Governance Proposals: AI Policy Analysis at Scale
评估治理提案的工具：大规模人工智能政策分析
Abstract
This paper introduces a policy analysis framework for systematic, transparent assessment of AI governance proposals in an evolving and contested regulatory landscape. AI policy debates often collapse into binary positions that obscure underlying tradeoffs and normative assumptions. The framework structures policy analysis around multiple policy attributes, allowing users to surface priorities and tensions without prescribing outcomes. We use a mixed-methods approach that integrates qualitative insights from subject matter experts with computational text analysis to inform the design of policy attribute rubrics. This quantifies the relative emphasis of different policy objectives and presents them through comparative visualizations that support interpretability and cross-policy comparison. The paper also examines the use of commercial LLMs for rubric-based policy analysis, benchmarking their outputs against a domain-trained rubric-calibrated model with explicitly defined analytical assumptions. Rather than assessing policy effectiveness or desirability, the framework focuses on relevance and alignment across attributes. By making analytical assumptions explicit, including attribute selection, rubric construction, and weighting schemes, the framework enables users to evaluate whether its embedded priorities align with the users' own normative commitments. The approach is jurisdiction-agnostic and intended to support policymakers, analysts, and researchers navigating complex AI governance environments. Contributions: (1) multidimensional policy assessment through empirically grounded rubrics that surface tradeoffs rather than resolving them; (2) a transparent hybrid methodology combining feedback from subject-matter experts with computational validation; and (3) use of domain-trained rubric-calibrated models as a benchmark for comparing different general-purpose large language models.
Chinese Translation
本文介绍了一种政策分析框架，用于在不断演变和充满争议的监管环境中对人工智能治理提案进行系统、透明的评估。人工智能政策辩论常常陷入二元立场，掩盖了潜在的权衡和规范假设。该框架围绕多个政策属性构建政策分析，使用户能够揭示优先事项和紧张关系，而不强加结果。我们采用混合方法，将来自主题专家的定性见解与计算文本分析相结合，以指导政策属性评分标准的设计。这量化了不同政策目标的相对重视程度，并通过比较可视化呈现，支持可解释性和跨政策比较。本文还考察了商业大语言模型（LLMs）在基于评分标准的政策分析中的应用，将其输出与经过领域训练的评分标准校准模型进行基准比较，该模型具有明确的分析假设。该框架并不评估政策的有效性或可取性，而是关注属性之间的相关性和一致性。通过明确分析假设，包括属性选择、评分标准构建和加权方案，该框架使用户能够评估其嵌入的优先事项是否与用户自身的规范承诺一致。该方法不依赖于特定司法管辖区，旨在支持在复杂人工智能治理环境中导航的政策制定者、分析师和研究人员。贡献包括：（1）通过基于实证的评分标准进行多维政策评估，揭示权衡而非解决权衡；（2）结合主题专家反馈与计算验证的透明混合方法；以及（3）使用经过领域训练的评分标准校准模型作为比较不同通用大语言模型的基准。
cs.AI / 80 / 2607.28098
SciDataSailor: Deep Scientific Data Exploring
SciDataSailor：深度科学数据探索
Abstract
Scientific datasets are commonly organized as hierarchical repositories containing heterogeneous and interdependent files, making their inspection, integration, and analysis labor-intensive and reliant on domain expertise. Although large language model (LLM) agents have advanced substantially in planning, reasoning, and tool use, existing research has largely overlooked their ability to interact with real scientific data assets through executable environments. We introduce Deep Scientific Data Exploration, an agentic task paradigm in which agents navigate repositories, interpret heterogeneous files and schemas, execute analyses, integrate cross-file evidence, and produce conclusions grounded in executed observations. To operationalize this paradigm, we present SciDataSailor, a framework for synthesizing tool-interactive trajectories by balancing broad exploration with targeted exploitation. SciDataSailor instantiates trajectory synthesis as Monte Carlo Tree Search (MCTS) with four task-specific mechanisms: difficulty-stratified exploration seeds, dual-feedback first-play urgency, hierarchical strategy-to-tool action generation, and entropy-guided branching. Using this framework, we construct SciDataSailor-SFT-2K for supervised fine-tuning and SciDataSailor-Bench for evaluation, with the latter comprising 627 meta-information summarization tasks and 586 scientific question-answering tasks across 27 datasets spanning the life, earth, and physical sciences.
Chinese Translation
科学数据集通常以层次化的存储库形式组织，包含异构和相互依赖的文件，这使得它们的检查、整合和分析变得劳动密集且依赖于领域专业知识。尽管大型语言模型（LLM）代理在规划、推理和工具使用方面已取得了显著进展，但现有研究在很大程度上忽视了它们通过可执行环境与真实科学数据资产互动的能力。我们引入了深度科学数据探索（Deep Scientific Data Exploration），这是一个代理任务范式，代理在存储库中导航，解释异构文件和模式，执行分析，整合跨文件证据，并基于执行观察得出结论。为了实现这一范式，我们提出了SciDataSailor，这是一个通过平衡广泛探索与针对性开发来合成工具交互轨迹的框架。SciDataSailor将轨迹合成实例化为蒙特卡罗树搜索（Monte Carlo Tree Search, MCTS），并具有四个特定任务的机制：难度分层的探索种子、双反馈的首次游戏紧迫性、层次策略到工具动作生成和熵引导的分支。利用该框架，我们构建了SciDataSailor-SFT-2K用于监督微调，以及SciDataSailor-Bench用于评估，后者包括627个元信息总结任务和586个科学问答任务，涵盖27个数据集，涉及生命科学、地球科学和物理科学。
cs.AI / 81 / 2607.28103
MIND: Lightweight and Effective Memory Injection Defense for LLM Agents via Intent-Aware Information Bottleneck
MIND：通过意图感知信息瓶颈实现轻量且有效的内存注入防御针对LLM代理
Abstract
Memory-augmented LLM-based agents are vulnerable to memory injection attacks: Agents may retrieve poisoned memory from attackers, which diverts their behavior from initial user intent and finally causes task failure. However, existing defense mechanisms either incur high computational cost or suffer from information redundancy in multi-turn contexts. To address these challenges, we propose Memory Intent-Aware Neural Denoising(MIND), a lightweight defense framework for memory injection attack. Our preliminary analysis reveals that benign and poisoned trajectories exhibit distinguishable relationships between the initial user intent and subsequent behavior. Building on this observation, MIND employs an intent-aware Information Bottleneck(IB) to extract compact intent--behavior representations from the initial intent and turn-level behavior. The IB preserves intent-relevant cross-turn attack signals while filtering task-irrelevant and repetitive information, and a lightweight detector identifies malicious memories from the resulting representations. As such, MIND mitigates information redundancy in multi-turn contexts while avoiding the overhead of repeated LLM auditing. Extensive experiments show that MIND reduces attack success rates while preserving task accuracy and inference efficiency. Notably, on ReAct-StrategyQA, MIND reduces mean ASR-r and ASR-a by 55.4% and 55.3%, respectively, while matching the undefended agent in average accuracy and latency.
Chinese Translation
基于内存增强的LLM代理易受到内存注入攻击：代理可能从攻击者那里检索到被污染的内存，这会使其行为偏离初始用户意图，最终导致任务失败。然而，现有的防御机制要么计算成本高昂，要么在多轮上下文中存在信息冗余。为了解决这些挑战，我们提出了内存意图感知神经去噪（Memory Intent-Aware Neural Denoising，MIND），这是一个针对内存注入攻击的轻量级防御框架。我们的初步分析表明，良性和被污染的轨迹在初始用户意图与后续行为之间表现出可区分的关系。基于这一观察，MIND采用意图感知的信息瓶颈（Information Bottleneck，IB）从初始意图和轮级行为中提取紧凑的意图-行为表示。IB保留与意图相关的跨轮攻击信号，同时过滤掉与任务无关和重复的信息，轻量级检测器从生成的表示中识别恶意内存。因此，MIND在多轮上下文中减轻了信息冗余，同时避免了重复审计LLM的开销。大量实验表明，MIND在保持任务准确性和推理效率的同时降低了攻击成功率。值得注意的是，在ReAct-StrategyQA上，MIND分别将平均ASR-r和ASR-a降低了55.4%和55.3%，同时在平均准确性和延迟上与未防御代理相匹配。
cs.AI / 82 / 2607.28109
Beyond Rephrasing: Book-Level Organization Improves Synthetic Textbook Data for Mid-Training
超越改写：书籍级组织改善中期训练的合成教科书数据
Abstract
Synthetic textbook data has improved language model pre-training, but prior work largely treats the benefit as a property of generated content or local rewriting style. We study a different factor: whether related content is organized into coherent book-level documents. We contribute both a scalable synthesis pipeline and controlled evidence that this organization matters. The pipeline retrieves source material from a pre-training corpus, clusters it into topical units, plans hierarchical tables of contents, and assembles source-grounded sections into complete books (our Full setting), yielding 686K textbooks (32B tokens) across 15,000+ disciplines. Replacing natural books in a mid-training mix with this corpus improves downstream performance by +1.09 on average. Controlled comparisons then disentangle the relevant design factors. A content-matched Split condition holds generated text and tokens fixed but treats each section as an independent document; Full's +1.02 mean gain isolates document packaging. A length-matched RandomConcat control that joins sections from different books remains below Full, ruling out document length alone. A retrieval-pool-matched Rephrase condition independently rewrites individual retrieved documents under the same audience-by-style scheme, without clustering, TOC planning, or book assembly; Full's +1.17 gain demonstrates the value of structured synthesis. On Llama3-8B, Full likewise outperforms both RandomConcat and Natural Books, supporting book-level organization as a useful axis for synthetic pre-training data design.
Chinese Translation
合成教科书数据改善了语言模型的预训练，但之前的研究主要将其好处视为生成内容或局部改写风格的属性。我们研究了一个不同的因素：相关内容是否组织成连贯的书籍级文档。我们贡献了一个可扩展的合成管道和控制证据，证明这种组织是重要的。该管道从预训练语料库中检索源材料，将其聚类为主题单元，规划层次化的目录，并将基于源的部分组装成完整的书籍（我们的完整设置），在15,000多个学科中生成了686K本教科书（32B个标记）。在中期训练中用该语料库替换自然书籍的混合，平均提高了下游性能+1.09。控制比较则解开了相关设计因素。内容匹配的Split条件保持生成文本和标记不变，但将每个部分视为独立文档；Full的+1.02平均增益隔离了文档包装。长度匹配的RandomConcat控制将来自不同书籍的部分连接在一起，仍然低于Full，排除了文档长度的单独影响。一个检索池匹配的Rephrase条件在相同的受众-风格方案下独立改写单个检索文档，而不进行聚类、目录规划或书籍组装；Full的+1.17增益展示了结构化合成的价值。在Llama3-8B上，Full同样优于RandomConcat和自然书籍，支持书籍级组织作为合成预训练数据设计的一个有用维度。
cs.AI / 83 / 2607.28110
BlueprintRepair: Typed Local Edits for Failed Lean Proof Blueprints
BlueprintRepair：针对失败的 Lean 证明蓝图的类型化局部编辑
Abstract
LLM-based Lean proving systems increasingly organize a proof as a blueprint: a dependency graph of formal statements. We introduce BlueprintRepair, a repair interface that lets a model change this graph through ten schema-checked local operations. An operation names the node it edits, so the target theorem cannot be changed. Lean checks every applied change, and an accepted repair must declare every blueprint lemma its proof uses. We also construct BlueprintTrace, a benchmark of 142 controlled failures with complete accepted and rejected repair trajectories. We compare typed edits, exact source patches, and complete module rewrites under matched source, feedback, model, and budget, one episode per state and interface. With DeepSeek-V4-Flash, the three interfaces solve almost the same number of the benchmark's localized failures. Typed repair is the cheapest per solved state (patching is 1.30x as expensive, rewriting 2.06x), and within 10,000 completion tokens per task it reaches almost all of its final coverage, while both free-form interfaces are well behind. A second model, Qwen3.6-Flash, solves fewer states but keeps typed repair cheapest, puts it ahead on the proof-authoring states, and repeats the localized pattern.
Chinese Translation
基于大语言模型的 Lean 证明系统越来越多地将证明组织为蓝图：一个形式语句的依赖图。我们介绍了 BlueprintRepair，这是一种修复接口，允许模型通过十种经过模式检查的局部操作来更改该图。每个操作都指定了它编辑的节点，因此目标定理不能被更改。Lean 会检查每个应用的更改，接受的修复必须声明其证明所使用的每个蓝图引理。我们还构建了 BlueprintTrace，这是一个包含 142 个受控失败的基准，具有完整的接受和拒绝修复轨迹。我们在匹配的源代码、反馈、模型和预算下比较了类型化编辑、精确源补丁和完整模块重写，每个状态和接口一个回合。使用 DeepSeek-V4-Flash，这三种接口几乎解决了基准的相同数量的局部失败。类型化修复在每个解决状态的成本上是最低的（补丁的成本是 1.30 倍，重写的成本是 2.06 倍），并且在每个任务的 10,000 个完成标记内几乎达到了最终覆盖率，而两个自由形式接口的表现远远落后。第二个模型 Qwen3.6-Flash 解决的状态较少，但保持了类型化修复的最低成本，在证明创作状态上领先，并重复了局部模式。
cs.AI / 84 / 2607.28126
ConMem: Contribution-Aware Memory for Long-Horizon Manufacturing Inspection Logs
ConMem：面向长期制造检查日志的贡献感知记忆
Abstract
Long-horizon steel-equipment inspection requires reasoning over heterogeneous records accumulated across repeated inspection cycles. Existing retrieval-augmented generation systems treat historical logs as a static corpus and retain records without estimating their diagnostic value, failing to report early risk. To this end, we propose ConMem, a contribution-aware memory framework for LLM-assisted equipment inspection, supporting a human-in-the-loop early-risk screening system. Specifically, our ConMem first segments inspection logs into functional evidence units, then estimates each memory unit's contribution to downstream diagnosis through a Shapley-style estimation, and finally retains high-value evidence under a constrained memory budget. In experiments, we evaluate ConMem on real-world dataset and ConMem achieves 76.0% QA accuracy, exceeding the strongest directly comparable baseline. Relative to the naive 8K-context LLM baselines, it reduces the average number of input tokens by 88.2% and response time by 86.6%. Ablation studies also show that the functional-role-aware segmentation and contribution-based valuation are helping prioritize weak degradation signals for targeted field inspection. Practical deployments further confirm that ConMem retains the weak early signal across three inspection cycles, providing an early-stage seal-wear alert targeted for on-site inspectors.
Chinese Translation
长期的钢铁设备检查需要对在重复检查周期中积累的异构记录进行推理。现有的检索增强生成系统将历史日志视为静态语料库，并在未评估其诊断价值的情况下保留记录，未能及时报告早期风险。为此，我们提出了ConMem，一个面向大语言模型（LLM）辅助设备检查的贡献感知记忆框架，支持人机协作的早期风险筛查系统。具体而言，我们的ConMem首先将检查日志分段为功能证据单元，然后通过Shapley风格的估计评估每个记忆单元对下游诊断的贡献，最后在受限的记忆预算下保留高价值证据。在实验中，我们在真实世界数据集上评估了ConMem，结果显示其QA准确率达到76.0%，超过了最强的直接可比基线。与简单的8K上下文LLM基线相比，它将平均输入标记数减少了88.2%，响应时间减少了86.6%。消融研究还表明，功能角色感知的分段和基于贡献的评估有助于优先考虑针对性现场检查的弱降级信号。实际部署进一步确认，ConMem在三个检查周期中保留了弱早期信号，为现场检查员提供了早期密封磨损警报。
cs.AI / 85 / 2607.28175
AgenticASR: Refining Speech Recognition in Real-World Scenarios via an Agentic Approach
AgenticASR：通过代理方法在真实场景中优化语音识别
Abstract
Automatic speech recognition (ASR) has achieved substantial gains in transcription accuracy, yet verbatim transcription does not necessarily produce readily usable text. It retains fillers, repetitions, false starts, and self-corrections that increase reading effort, obscure the speaker's final intent, and propagate unresolved or abandoned content to downstream tasks. Existing spoken-to-written methods process completed audio or transcripts but cannot revise emitted text when later speech changes how preceding content should be interpreted. We therefore formulate Agentic Speech Recognition (AgenticSR), an audio-to-clean-text task that removes disfluencies, resolves self-corrections, and normalizes written form while preserving the speaker's final intent. AgenticASR implements this task through an ASR--Refiner architecture that repeatedly transforms a bounded active context and replaces its corresponding output span as audio arrives. This enables continual emission and revision over streams of arbitrary duration. We also introduce AASR-Bench, a bilingual benchmark with fine-grained atomic rubrics. Across multiple ASR front ends, AgenticASR attains the highest AASR-Bench scores among evaluated systems. A human--AI agreement study shows that rubric-based judgments align with independent expert assessments. Ablations characterize Refiner capacity, context length, and the quality--latency trade-off between online and offline inference. Together, these results establish AgenticASR as a practical framework for intent-preserving clean transcription during ongoing speech. Code, AASR-Bench, and a demo will be released at https://github.com/AnXMuy/AgenticASR.
Chinese Translation
自动语音识别（ASR）在转录准确性方面取得了显著进展，但逐字转录并不一定能生成可直接使用的文本。它保留了填充词、重复、错误开始和自我修正，这增加了阅读的难度，模糊了说话者的最终意图，并将未解决或放弃的内容传播到下游任务。现有的口语转书面的方法处理已完成的音频或转录文本，但无法在后续语音改变前述内容的解释时修正已发出的文本。因此，我们提出了代理语音识别（Agentic Speech Recognition，AgenticSR），这是一项音频到清晰文本的任务，旨在去除流畅性障碍，解决自我修正，并规范书面形式，同时保留说话者的最终意图。AgenticASR通过ASR-Refiner架构实现这一任务，该架构反复转换一个有限的主动上下文，并在音频到达时替换其相应的输出范围。这使得在任意持续时间的流中能够持续发出和修正文本。我们还引入了AASR-Bench，这是一个具有细粒度原子评分标准的双语基准。在多个ASR前端中，AgenticASR在评估系统中获得了最高的AASR-Bench分数。一项人类与人工智能一致性研究表明，基于评分标准的判断与独立专家评估一致。消融实验表征了Refiner的能力、上下文长度以及在线与离线推理之间的质量-延迟权衡。这些结果共同确立了AgenticASR作为一个在持续语音过程中保持意图的清晰转录的实用框架。代码、AASR-Bench和演示将发布在https://github.com/AnXMuy/AgenticASR。
cs.AI / 86 / 2607.28187
Old Tricks, New Models: How Simple Image Transformations Break Modern AI-based Content Moderation
旧把戏，新模型：简单图像变换如何突破现代基于AI的内容审核
Abstract
While automated content-moderation systems have become essential for screening harmful content at scale, conventional task-specific classifiers often provide limited policy cov- erage and contextual understanding. Recently, commercial multimodal moderation APIs built on large foundation models have been introduced with the promise of providing broader and more capable safety filters. In this work, we analyze whether this shift also yields more robust image moderation. We conduct a large-scale black-box evaluation on three established commercial image-moderation services and compare their robustness. By evaluating seven simple, model-agnostic image transformations across multiple providers, datasets, harm categories, perceptual-similarity constraints, and transformation intensities, we find that: (1) all three commercial services can be bypassed using inexpensive image transformations that require no gradients, surrogate models, or knowledge of the target system; (2) even fixed transformations such as color inversion and grayscale conversion induce unsafe-to-safe decision changes while preserving content that remains recognizable to humans; (3) their robustness varies substantially across datasets and harm categories, with multimodal content and self-harm exhibiting pronounced vulnerabilities. This yields the conclusion that replacing conventional moderation classifiers with foundation-model-based APIs does not, by itself, provide a reliable security boundary. Such systems must be evaluated under realistic transformations and deployed as one component of a layered moderation pipeline rather than as standalone safety filters.
Chinese Translation
尽管自动化内容审核系统已成为大规模筛查有害内容的必要工具，但传统的任务特定分类器往往提供有限的政策覆盖和上下文理解。最近，基于大型基础模型的商业多模态审核API应运而生，承诺提供更广泛和更强大的安全过滤功能。在本研究中，我们分析这一转变是否也带来了更强大的图像审核能力。我们对三种成熟的商业图像审核服务进行了大规模黑箱评估，并比较了它们的鲁棒性。通过在多个提供商、数据集、有害类别、感知相似性约束和变换强度下评估七种简单的模型无关图像变换，我们发现：（1）所有三种商业服务都可以通过廉价的图像变换绕过，这些变换不需要梯度、替代模型或对目标系统的知识；（2）即使是固定的变换，如颜色反转和灰度转换，也会导致不安全到安全的决策变化，同时保留人类可识别的内容；（3）它们的鲁棒性在不同数据集和有害类别之间差异显著，多模态内容和自残行为表现出明显的脆弱性。这得出的结论是，仅仅用基于基础模型的API替代传统的审核分类器，并不能提供可靠的安全边界。这类系统必须在现实变换下进行评估，并作为分层审核流程中的一个组成部分进行部署，而不是作为独立的安全过滤器。
cs.AI / 87 / 2607.28227
Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents
Qwen-UI-Agent 技术报告：迈向下一代以现实世界为中心的基础 GUI 代理
Abstract
GUI agents have the potential to become a general purpose executor over existing digital devices. To advance them toward real-world use, we envision agents that operate reliably on real devices, execute workflows across platforms, combine GUI interaction with CLI execution, complete long-horizon tasks, proactively initiate useful services, and autonomously improve their capabilities with minimal human effort. Guided by this vision, we present Qwen-UI-Agent, a real-world centric foundation GUI agent spanning mobile, computer-use, web, and DeepSearch environments. Qwen-UI-Agent combines diverse sandbox environments with a large-scale real-device mobile runtime. Its unified action space interleaves GUI operations with CLI execution and generates batched actions in a single model turn. An AutoResearch-style data flywheel uses agents to construct tasks and environments, diagnose failures, and plan subsequent iterations. Online RL supports training on trajectories exceeding 100 turns, with over 10,000 concurrent environments accelerating rollout. A lightweight harness layer supports proactive service initiation and stateful workflows across mobile and computer. Across a broad suite of evaluations, Qwen-UI-Agent sets state-of-the-art performance on mobile-use benchmarks while delivering competitive performance on computer- and browser-use tasks against frontier models, including Opus 4.8, Gemini 3.1 Pro, and GPT-5.6 Sol. On mobile use, it achieves 82.1% on MobileWorld, 92.2% on MobileWorld-Real, and 97.5% on AndroidDaily. On computer use, it achieves 79.5% on OSWorld-Verified and a 40.0% partial-progress score on OSWorld-v2. On browser use and GUI grounding, it achieves 73.6% on WebArena and 81.5% on ScreenSpot-Pro, respectively.
Chinese Translation
GUI 代理有潜力成为现有数字设备上的通用执行者。为了推动它们向现实世界的应用发展，我们设想能够在真实设备上可靠运行的代理，跨平台执行工作流程，将 GUI 交互与 CLI 执行相结合，完成长期任务，主动发起有用的服务，并在最小人力干预下自主提升其能力。基于这一愿景，我们提出了 Qwen-UI-Agent，一个以现实世界为中心的基础 GUI 代理，涵盖移动设备、计算机使用、网络和 DeepSearch 环境。Qwen-UI-Agent 将多样的沙箱环境与大规模真实设备移动运行时相结合。其统一的动作空间将 GUI 操作与 CLI 执行交错，并在单个模型回合中生成批量动作。AutoResearch 风格的数据飞轮利用代理构建任务和环境，诊断故障，并规划后续迭代。在线强化学习支持超过 100 回合的轨迹训练，超过 10,000 个并发环境加速了推广。在移动和计算机之间，轻量级的支撑层支持主动服务启动和有状态的工作流程。在一系列广泛的评估中，Qwen-UI-Agent 在移动使用基准上设定了最先进的性能，同时在计算机和浏览器使用任务中与前沿模型（包括 Opus 4.8、Gemini 3.1 Pro 和 GPT-5.6 Sol）相比表现出竞争力。在移动使用方面，它在 MobileWorld 上达到了 82.1%，在 MobileWorld-Real 上达到了 92.2%，在 AndroidDaily 上达到了 97.5%。在计算机使用方面，它在 OSWorld-Verified 上达到了 79.5%，在 OSWorld-v2 上达到了 40.0% 的部分进展得分。在浏览器使用和 GUI 基础方面，它在 WebArena 上达到了 73.6%，在 ScreenSpot-Pro 上达到了 81.5%。
cs.AI / 88 / 2607.28236
CDAE: Enhancing Perturbation Robustness in Pretrained Language Models with Contrastive Denoising
CDAE：通过对比去噪增强预训练语言模型的扰动鲁棒性
Abstract
Pre-trained language models have significantly improved sentence representation learning, yet their embedding remain sensitive to semantic preserving textual perturbations such as synonym substitution, masking and word dropout. This work proposes a lightweight Contrastive Denoising Autoencoder (CDAE) that refines pre-trained BERT embedding by jointly optimizing contrastive and reconstruction objective to learn perturbation-invariant representation. We evaluate the proposed framework using multiple perturbation strategies with varying strengths and compare it against the original BERT embeddings and SimCSE. Experimental results show that CDAE consistently preserves higher embedding similarity under perturbations, with the improvements becoming more pronounced as framework effectively enhances representation stability while preserving semantic information, highlighting perturbation-invariant learning as a promising direction for improving sentence embeddings. The source code is publicly available at: https://github.com/ComputationIASBS/CDAE
Chinese Translation
预训练语言模型显著提升了句子表示学习，但其嵌入仍然对保持语义的文本扰动（如同义词替换、掩码和词丢失）敏感。本研究提出了一种轻量级的对比去噪自编码器（Contrastive Denoising Autoencoder, CDAE），通过联合优化对比目标和重建目标来精炼预训练的BERT嵌入，以学习对扰动不变的表示。我们使用多种不同强度的扰动策略评估所提框架，并将其与原始BERT嵌入和SimCSE进行比较。实验结果表明，CDAE在扰动下始终保持更高的嵌入相似性，随着框架有效增强表示稳定性并保持语义信息，改进效果愈加明显，突显了对扰动不变学习作为提升句子嵌入的有前景方向。源代码已公开可用，网址为：https://github.com/ComputationIASBS/CDAE
cs.AI / 89 / 2607.28237
AI and Authenticity in Islamic Research: A Critical Evaluation of Generative AI Reliability, Hallucination, and Source Fidelity in Quranic, Hadith, and Fiqh Knowledge
人工智能与伊斯兰研究中的真实性：对生成性人工智能在《古兰经》、圣训和法学知识中的可靠性、幻觉和来源忠实度的批判性评估
Abstract
Generative Artificial Intelligence (AI) is increasingly used by Muslims for religious guidance, Qur'anic interpretation, Hadith explanation, jurisprudential rulings, and Islamic education. Despite its growing adoption, there is limited empirical evidence on whether current AI systems provide authentic, verifiable, and trustworthy Islamic knowledge suitable for high-trust religious contexts. This study evaluates six leading generative AI systems using fifty realistic open-ended Islamic questions covering Qur'anic interpretation, Hadith, Fiqh, ethics, pastoral advice, and Madhhab-sensitive topics. Responses were collected under real-world conditions from participants in Australia and the United Kingdom and analysed using a mixed-method framework examining domain accuracy, citation verification, hallucinations, jurisprudential consistency, uncertainty handling, source provenance, and geographical variation. The study addresses four research questions: (1) How accurate and authentic are AI-generated responses across major Islamic knowledge domains? (2) To what extent do AI systems produce hallucinations, incomplete citations, or unverifiable religious references? (3) How consistently do models handle jurisprudential disagreement, Madhhab diversity, and uncertainty? (4) Are current AI systems sufficiently reliable for religious guidance, Islamic education, and scholarly research? Overall, current generative AI systems are valuable as assistive tools for introductory Islamic learning but should not be treated as authoritative sources for religious rulings or Islamic research without verification against authenticated primary sources and qualified scholarly expertise. This study provides one of the first comprehensive empirical evaluations of AI reliability within Islamic knowledge, offering practical guidance for researchers, educators, AI developers, and the wider Muslim community.
Chinese Translation
生成性人工智能（AI）在穆斯林中越来越多地用于宗教指导、《古兰经》解释、圣训阐释、法学裁决和伊斯兰教育。尽管其应用日益广泛，但关于当前AI系统是否提供适合高信任宗教环境的真实、可验证和可信的伊斯兰知识的实证证据仍然有限。本研究评估了六个主要的生成性AI系统，使用了五十个现实的开放式伊斯兰问题，涵盖了《古兰经》解释、圣训、法学、伦理、牧师建议和教派敏感话题。研究在澳大利亚和英国的参与者中，在真实世界条件下收集了回答，并采用混合方法框架分析了领域准确性、引用验证、幻觉、法学一致性、不确定性处理、来源来源和地理差异。研究解决了四个研究问题：（1）AI生成的回答在主要伊斯兰知识领域的准确性和真实性如何？（2）AI系统在多大程度上产生幻觉、不完整的引用或不可验证的宗教参考？（3）模型在处理法学分歧、教派多样性和不确定性方面的一致性如何？（4）当前的AI系统在宗教指导、伊斯兰教育和学术研究方面是否足够可靠？总体而言，当前的生成性AI系统作为初步伊斯兰学习的辅助工具是有价值的，但在未经验证与经过认证的原始来源和合格学术专家对照之前，不应被视为宗教裁决或伊斯兰研究的权威来源。本研究提供了对AI在伊斯兰知识中可靠性进行的首批全面实证评估之一，为研究人员、教育工作者、AI开发者和更广泛的穆斯林社区提供了实用指导。
cs.AI / 90 / 2607.28257
Operationally Guided Placement-Aware Learning for Industrial Online 3D Bin Packing
面向操作的放置感知学习在工业在线三维装箱中的应用
Abstract
The online three-dimensional bin packing problem (3D-BPP) is a longstanding challenge in logistics and industrial palletizing. Recent learning-based methods use a learned policy to select among feasible candidate placements. Performance depends on the candidate generator and representation, especially in industrial settings where packings must be space-efficient, stable, compact, and balanced. However, prior work has mainly optimized the policy, while candidate generation and representation remain largely geometry-driven. We address this gap with OPAL, an operationally guided placement-aware learning framework for industrial online 3D-BPP which combines an Operationally Guided Empty-Maximal-Space generator (OG-EMS), an operational representation for each candidate placement, and a masked ranking policy trained with proximal policy optimization. OG-EMS evaluates multiple anchors within each free-space region and prioritizes low, well-supported, compact, and spatially diverse placements. An xLSTM-based Placement Encoder models dependencies among geometric and operational candidate attributes, while a lightweight recurrent core combines the resulting embeddings with the current item and pallet state to rank feasible actions. On the BED-BPP benchmark, OPAL achieves a mean space utilization of 0.49, with improvements of 15.1% from operationally guided candidate generation and 6.3% from learned ranking, while maintaining robust inference-time performance.
Chinese Translation
在线三维装箱问题（3D-BPP）是物流和工业托盘化中的一个长期挑战。近期基于学习的方法使用学习到的策略在可行的候选放置中进行选择。性能依赖于候选生成器和表示，尤其是在工业环境中，装箱必须具有空间效率、稳定性、紧凑性和平衡性。然而，之前的研究主要优化了策略，而候选生成和表示仍然主要受到几何因素的驱动。我们通过OPAL来填补这一空白，OPAL是一个面向操作的放置感知学习框架，旨在解决工业在线3D-BPP问题，结合了操作引导的空最大空间生成器（OG-EMS）、每个候选放置的操作表示，以及通过近端策略优化训练的掩蔽排名策略。OG-EMS在每个自由空间区域内评估多个锚点，并优先考虑低、支撑良好、紧凑且空间多样的放置。基于xLSTM的放置编码器建模几何和操作候选属性之间的依赖关系，而轻量级递归核心将生成的嵌入与当前物品和托盘状态结合，以对可行的动作进行排名。在BED-BPP基准测试中，OPAL实现了0.49的平均空间利用率，相较于操作引导的候选生成提高了15.1%，相较于学习的排名提高了6.3%，同时保持了稳健的推理时间性能。
cs.AI / 91 / 2607.28268
LLM-Guided Evolutionary Search for Constraint Model Reformulation to Improve Solver Efficiency
基于大语言模型的进化搜索用于约束模型重构以提高求解器效率
Abstract
Combinatorial problems appear in numerous industrial applications. A common approach is to formulate these problems as declarative constraint models that can subsequently be compiled to and solved by a range of back-end solvers. Recent work shows that Large Language Models (LLMs) can produce correct models from natural language, but even a correct model can be expensive to solve because performance remains sensitive to modelling choices. In this work, we investigate whether LLMs can automate performance-oriented model reformulation. Inspired by Automatic Heuristic Design (AHD), we use an evolutionary framework in which an LLM proposes candidate reformulations that are verified and benchmarked against the user-defined baseline model. We compare AHD-adapted search strategies that control which prior attempts, instructions, and measured feedback enter each prompt. Existing retention strategies prioritize recency or performance, but do not explicitly diversify the context. To cover this gap, we introduce Profile-Diverse Retention (PDR), which applies Maximal Marginal Relevance (MMR) to instance-level runtime vectors to retain behaviourally diverse attempts. We systematically evaluate the strategies on eight CSPLib problems using validation-based final model selection. The results show that: (i) iterative reformulation can produce substantial held-out speedups; (ii) strategies that keep the retained context diverse outperform those that retain only recent or the fastest attempts; and (iii) validation-based selection improves the held-out speedup of every strategy.
Chinese Translation
组合问题出现在众多工业应用中。一个常见的方法是将这些问题表述为声明性约束模型，这些模型随后可以被编译并由各种后端求解器求解。近期的研究表明，大语言模型（LLMs）能够从自然语言生成正确的模型，但即使是正确的模型在求解时也可能代价高昂，因为性能仍然对建模选择敏感。在本研究中，我们探讨了LLMs是否能够自动化以性能为导向的模型重构。受到自动启发式设计（AHD）的启发，我们使用一个进化框架，其中LLM提出候选重构，这些重构会被验证并与用户定义的基线模型进行基准测试。我们比较了适应AHD的搜索策略，这些策略控制哪些先前的尝试、指令和测量反馈进入每个提示。现有的保留策略优先考虑最近性或性能，但并未明确多样化上下文。为填补这一空白，我们引入了Profile-Diverse Retention（PDR），该策略将最大边际相关性（MMR）应用于实例级运行时向量，以保留行为上多样化的尝试。我们在八个CSPLib问题上系统地评估了这些策略，采用基于验证的最终模型选择。结果表明：（i）迭代重构可以产生显著的保留加速；（ii）保持保留上下文多样性的策略优于仅保留最近或最快尝试的策略；（iii）基于验证的选择提高了每种策略的保留加速。
cs.AI / 92 / 2607.28272
MemHarness: Memory Is Reconstructed, Not Replayed
MemHarness：记忆是重构的，而非重放的
Abstract
Retrieving past experiences has become a common strategy to enhance large language model agents. However, most existing memory-augmented agents treat retrieved experiences as static records to be replayed verbatim, injecting them into the context regardless of whether they align with the agent's current situation. This ``replay'' paradigm ignores the gap between the abstract, general nature of stored experience and the concrete, ever-changing states encountered at decision time, frequently causing negative transfer. In contrast, humans rarely recall past experiences verbatim; instead, they reorganize and adapt retrieved memories to fit the present context. Inspired by this, we propose MemHarness, a framework that equips LLM agents to actively harness and reconstruct past experiences based on the present context. At each decision step, a unified policy model critiques and reconstructs the retrieved experience conditioned on the current state, producing context-grounded guidance before acting. This reconstructive ability emerges naturally through end-to-end training with GRPO. Experiments on ALFWorld and WebShop show that MemHarness substantially outperforms pure RL and static memory-augmented baselines, demonstrating strong robustness in out-of-distribution (OOD) scenarios. Furthermore, our analyses reveal that this reconstruction objective not only prevents negative transfer but also serves as latent guidance during training, fundamentally improving the agent's intrinsic reasoning capabilities.
Chinese Translation
检索过去的经验已成为增强大型语言模型代理的常见策略。然而，大多数现有的记忆增强代理将检索到的经验视为静态记录，逐字重放，将其注入上下文中，而不考虑它们是否与代理当前的情况相符。这种“重放”范式忽视了存储经验的抽象、一般特性与决策时所遇到的具体、不断变化的状态之间的差距，常常导致负迁移。相比之下，人类很少逐字回忆过去的经验；相反，他们会重新组织和调整检索到的记忆以适应当前的上下文。受到此启发，我们提出了MemHarness，一个框架，使大型语言模型代理能够主动利用和重构基于当前上下文的过去经验。在每个决策步骤中，一个统一的策略模型会根据当前状态对检索到的经验进行批判和重构，在行动之前生成基于上下文的指导。这种重构能力通过与GRPO的端到端训练自然地出现。在ALFWorld和WebShop上的实验表明，MemHarness显著优于纯强化学习和静态记忆增强基线，在分布外（OOD）场景中表现出强大的鲁棒性。此外，我们的分析表明，这一重构目标不仅防止了负迁移，还在训练过程中作为潜在指导，根本上提高了代理的内在推理能力。
cs.AI / 93 / 2607.28287
Tycho: Active Abstraction with Programmatic World Models for ARC-AGI-3
Tycho：具有程序化世界模型的主动抽象用于ARC-AGI-3
Abstract
ARC-AGI-3 turns abstraction into an interactive problem of skill acquisition. A player must infer an unfamiliar game's rules, hidden state, and goal while maintaining action efficiency because every move counts. We formalize these environments as parameterized rendered deterministic Moore machines and introduce Tycho, a coding-agent system that constructs and uses game-specific models during interaction. Tycho separates actionable observations from intermediate animation, level-completion, and game-over frames. From this structured history, an agent can model, test, plan with, repair, or bypass a free-form executable hypothesis. In one matched public-set run per policy, we compare four orchestration policies on all 25 public games using Claude Opus 4.8 under matched inference budgets. Actor-requested delegation to a model builder obtains the highest observed mean Relative Human Action Efficiency (RHAE), 88.49. With this selected policy, GPT-5.6 Sol and Opus 5 both reach 100.00 RHAE and complete all 183 levels. Their game-balanced first-run human-replay midranks are 98.5 and 100.0. Opus 5 uses 61% fewer scored actions than the aggregate official human baselines. Automatic repair after verification failures produces models that reproduce observed transitions much more accurately, yet reaches only 83.07 RHAE. Transition match indicates whether a simulator reproduces observed dynamics, not whether it has identified the objective or improves the next action. Strong play also requires deciding when to construct, repair, use, or bypass a model. We call this joint problem active abstraction: generating a testable model from costly interaction and deciding when acquiring or using it is worth its cost.
Chinese Translation
ARC-AGI-3将抽象转化为技能获取的互动问题。玩家必须推断一个不熟悉游戏的规则、隐藏状态和目标，同时保持行动效率，因为每一步都至关重要。我们将这些环境形式化为参数化的渲染确定性摩尔机，并引入Tycho，一个在互动过程中构建和使用特定游戏模型的编码代理系统。Tycho将可操作的观察与中间动画、关卡完成和游戏结束帧分离。通过这种结构化的历史，代理可以建模、测试、规划、修复或绕过一个自由形式的可执行假设。在每个策略的匹配公共集运行中，我们使用Claude Opus 4.8在匹配推理预算下比较四种编排策略在所有25个公共游戏上的表现。请求代理委托给模型构建者获得了观察到的最高平均相对人类行动效率（RHAE）88.49。在这个选定的策略下，GPT-5.6 Sol和Opus 5均达到100.00 RHAE并完成所有183个关卡。它们的游戏平衡首次运行人类重放中排名分别为98.5和100.0。Opus 5使用的得分行动比官方人类基准总和少61%。在验证失败后的自动修复产生的模型能够更准确地重现观察到的转变，但仅达到83.07 RHAE。转变匹配指示模拟器是否重现了观察到的动态，而不是它是否识别了目标或改善了下一个行动。强有力的游戏表现还需要决定何时构建、修复、使用或绕过模型。我们称这个联合问题为主动抽象：从高成本的互动中生成可测试的模型，并决定何时获取或使用它是值得的。
cs.AI / 94 / 2607.28317
One Human, $N$ Agents: Audit-Budget Allocation for LLM Agent Fleets under Miscalibrated, Correlated Confidence
一个人，$N$ 个代理：在误校准和相关置信下的LLM代理舰队审计预算分配
Abstract
A single human must audit $N$ LLM agents under a budget of $B \ll N$ audits per round, guided by self-reported confidence that may be adversarially miscalibrated and by correlated errors. We model this as budgeted noisy inspection over a two-level Gaussian copula and locate the miscalibration threshold $\delta^*$ past which confidence-ranked auditing is \emph{worse} than random. Two a-priori expectations reverse: $\delta^*$ \emph{rises} as the budget shrinks, and cross-family correlation is not low---shared difficulty dominates lineage. Five open-weight LLMs show operationally useless (near-constant) confidence, point estimates at or beyond the flip though CIs straddle it; a proprietary model is informative and lands below it. We give a quantitative criterion for \emph{vacuous} oversight, and replaying policies on recorded traces confirms the ordering.
Chinese Translation
一个人必须在每轮预算为 $B ext{（远小于）} N$ 的情况下审计 $N$ 个LLM代理，这一过程受到自我报告的置信度的指导，而该置信度可能存在对抗性误校准，并且存在相关错误。我们将其建模为在双层高斯耦合体下的预算噪声检查，并确定误校准阈值 $eta^*$，超过该阈值后，基于置信度的审计效果 extit{不如}随机审计。两个先验期望发生反转：随着预算的缩减，$eta^*$ extit{上升}，而跨家族相关性并不低——共享的困难主导了谱系。五个开放权重的LLM显示出在操作上无用（近乎恒定）的置信度，点估计在翻转点或超出翻转点，尽管置信区间跨越了该点；而一个专有模型则提供了有用的信息，并落在该点之下。我们给出了 extit{空洞}监督的定量标准，并通过在记录的轨迹上重放策略来确认排序。
cs.AI / 95 / 2607.28318
PathView-Bench: Can Multimodal Large Language Models Achieve Fine-grained Multiscale Understanding of Pathology Images?
PathView-Bench：多模态大型语言模型能否实现对病理图像的细粒度多尺度理解？
Abstract
Multimodal large language models (MLLMs) are increasingly used to analyze pathology images. However, dominant multimodal benchmarks in pathology mainly score final diagnostic answers, captions, or reports. These evaluations provide limited insight into whether a model understands the multiscale visual content needed for pathology reasoning and decision-making. We introduce PathVU, a vision-anchored benchmark for fine-grained and multiscale visual understanding in computational pathology. Built from 23 public pathology imaging datasets with human-supervised labels and spatial annotations, PathVU evaluates MLLM understanding in two fields of view: Region FOV for high-resolution local regions and Slide FOV for macro whole-slide views. By converting raw annotations into deterministic task targets, PathVU enables programmatic scoring of region localization, visual recognition, quantity estimation, spatial reasoning, and insufficient-context judgment. The benchmark contains 14 VQA-style tasks, 61,673 images, and 308,070 samples across 28 organs and 7,253,526 annotations. Evaluating 18 representative general-purpose, medical-domain, and pathology-oriented MLLMs, we observe substantial limitations even in advanced models on fine-grained visual tasks across multiscale pathology images. PathVU provides a reproducible basis for developing and evaluating pathology MLLMs with explicit multiscale visual understanding.
Chinese Translation
多模态大型语言模型（MLLMs）在分析病理图像方面的应用日益增加。然而，当前病理领域的主流多模态基准主要评估最终的诊断答案、描述或报告。这些评估对模型是否理解病理推理和决策所需的多尺度视觉内容提供了有限的洞察。我们引入了PathVU，这是一个以视觉为基础的基准，旨在实现计算病理学中的细粒度和多尺度视觉理解。PathVU基于23个具有人工监督标签和空间注释的公共病理成像数据集构建，评估MLLM在两个视野中的理解能力：区域视野（Region FOV）用于高分辨率局部区域，幻灯片视野（Slide FOV）用于宏观整体幻灯片视图。通过将原始注释转换为确定性的任务目标，PathVU实现了区域定位、视觉识别、数量估计、空间推理和上下文不足判断的程序化评分。该基准包含14个视觉问答（VQA）风格的任务，61,673张图像，以及跨越28个器官的308,070个样本和7,253,526个注释。在评估18个具有代表性的通用、医学领域和病理导向的MLLM时，我们观察到即使在先进模型中，对于多尺度病理图像的细粒度视觉任务也存在显著的局限性。PathVU为开发和评估具有明确多尺度视觉理解的病理MLLM提供了可重复的基础。
cs.AI / 96 / 2607.28330
Paying for Honesty Without Knowing the Truth: Reputation-Penalty Design for LLM Marketplace Agents
在不知真相的情况下为诚实付费：LLM市场代理的声誉惩罚设计
Abstract
LLM agents increasingly act as autonomous merchants that write their own product listings, and under competitive pressure, they fabricate attributes to win sales. Even under instructions to be honest, they fabricate attributes in a majority of listings across models. A platform's obvious remedy---verifying each claim against the truth---is unavailable, because it observes only a noisy, biased complaint signal, never the ground truth. We design CARP, a reputation-penalty mechanism with a deadband that forgives complaint noise and a state-dependent severity that counters reputation-driven detection erosion. CARP requires no product-level ground truth and is robust to strategic gaming. CARP protects consumers by suppressing the sales volume of low-rated liars while sparing honest sellers. Paired with SPARC, it closes most of the consumer-welfare gap relative to a perfect-information oracle, without ever accessing the truth. It also achieves the best welfare of the policies we compare. We further show that this felt penalty becomes behaviorally binding through SPARC, a byte-clean code-gated reflection mechanism: LLM merchants fabricate when lying is free but restrain themselves when fabrication costs them sales, a self-interested response rather than compliance. We trace this distinction to penalty-gated self-correction reasoning, and observe the binding across models, with supporting confidence intervals.
Chinese Translation
LLM代理越来越多地充当自主商家，撰写自己的产品列表，在竞争压力下，它们为了赢得销售而虚构属性。即使在被指示要诚实时，它们在大多数列表中仍然虚构属性。平台的明显补救措施——验证每个声明的真实性——是不可行的，因为它仅观察到一个嘈杂且有偏见的投诉信号，而无法获得真实情况。我们设计了CARP，一种具有死区的声誉惩罚机制，该机制能够宽恕投诉噪声，并通过状态依赖的严重性来抵消声誉驱动的检测侵蚀。CARP不需要产品级别的真实情况，并且对战略性游戏具有鲁棒性。CARP通过抑制低评级骗子的销售量来保护消费者，同时不影响诚实卖家。与SPARC配合使用时，它在不接触真实情况的情况下，缩小了相对于完美信息神谕的大部分消费者福利差距。它还实现了我们比较的政策中最佳的福利。我们进一步表明，这种感知惩罚通过SPARC成为行为约束：当撒谎没有成本时，LLM商家会虚构，但当虚构导致销售损失时，他们会自我约束，这是一种自利的反应，而非合规。我们将这种区别追溯到惩罚门控自我修正推理，并观察到在各模型之间的约束性，伴随有支持的置信区间。
cs.AI / 97 / 2607.28336
Correcting What You Cannot See: Credit Assignment for Perception Distillation in Multimodal Reasoners
纠正你看不见的：多模态推理器中的感知蒸馏信用分配
Abstract
On-policy distillation provides dense supervision for multimodal reasoners, but its trajectory-level reward cannot determine whether a failed answer arose from perception or subsequent reasoning. Perception Success Rate (PSR), estimated from multiple reasonings sharing one perception, remains ambiguous because low success conflates perceptual insufficiency with reasoning difficulty. We introduce \textbf{Perception-Correction Distillation (PCD)}, a label-free method that identifies correctable perception failures using downstream failure and teacher--student disagreement as complementary witnesses. Their product, , forms a soft AND gate that strengthens distillation only when both witnesses are present. We motivate this rule through Bayesian evidence combination and show that multiplication is the unique normalized bilinear gate that vanishes when either witness is absent. PCD uses separated perception--reasoning rollouts and mean-preserving weights, leaving the reasoning objective unchanged. Across eight benchmarks, PCD improves the 8B 2B macro average from 44.50 with OPD to 47.28 and the 32B 8B result from 56.94 to 61.22. In matched 2B ablations, removing PCD and separated rollout reduces held-out average by 2.22 and 0.88 points, respectively. Effective multimodal distillation therefore depends not only on what the teacher predicts, but also on identifying when perception is the appropriate target of correction.
Chinese Translation
在线蒸馏为多模态推理器提供了密集的监督，但其轨迹级奖励无法确定失败答案是由感知还是后续推理引起的。感知成功率（Perception Success Rate, PSR）是从多个共享同一感知的推理中估计的，但由于低成功率将感知不足与推理困难混淆，因此仍然模糊不清。我们提出了 extbf{感知纠正蒸馏（Perception-Correction Distillation, PCD）}，这是一种无标签的方法，通过下游失败和教师-学生不一致性作为互补证据，识别可纠正的感知失败。它们的乘积形成一个软与门，仅在两个证据都存在时增强蒸馏。我们通过贝叶斯证据组合来激励这一规则，并展示了乘法是唯一的归一化双线性门，当任一证据缺失时会消失。PCD使用分离的感知-推理回放和均值保持权重，使推理目标保持不变。在八个基准测试中，PCD将8B 2B宏平均从44.50（使用OPD）提高到47.28，将32B 8B结果从56.94提高到61.22。在匹配的2B消融实验中，去除PCD和分离回放分别减少了持出平均值2.22和0.88点。因此，有效的多模态蒸馏不仅依赖于教师的预测，还依赖于识别何时感知是适当的纠正目标。
cs.AI / 98 / 2607.28367
How Benchmarks Mis-Score Computer-Use Agents
基准测试如何错误评分计算机使用代理
Abstract
Computer-use agents (CUA) are being deployed to browse the web and operate desktop software, yet their benchmark scores are still commonly produced by brittle scripted oracles. A score is the output of a pipeline in which tasks can be stale, trajectories can omit decisive visual evidence, evaluators can reject valid alternatives, and aggregate reports can hide the cause of failure. We organize these problems into a reliability framework spanning task construction, trajectory observation, scoring, and reporting. We then audit 150 public failure-scored trajectories from five web, enterprise-workflow, and desktop-control benchmarks, find that 15.3\% of FAIL verdicts are wrong: 10.7\% are evaluator false negatives and 4.7\% are broken tasks. For genuine failures, a three-tier diagnostic taxonomy shows that verification/feedback and planning failures dominate execution/grounding errors, while a single scalar success rate can not explain. We connect these findings to newer long-horizon CUA benchmarks and derive stage-specific design rules for CUA evaluation.
Chinese Translation
计算机使用代理（CUA）正在被部署用于浏览网页和操作桌面软件，但它们的基准测试评分仍然常常由脆弱的脚本化预言者生成。评分是一个管道的输出，其中任务可能过时，轨迹可能遗漏决定性的视觉证据，评估者可能拒绝有效的替代方案，而汇总报告可能隐藏失败的原因。我们将这些问题组织成一个涵盖任务构建、轨迹观察、评分和报告的可靠性框架。然后，我们审计了来自五个网络、企业工作流程和桌面控制基准的150个公共失败评分轨迹，发现15.3%的失败判定是错误的：10.7%是评估者的假阴性，4.7%是损坏的任务。对于真正的失败，三层诊断分类法显示验证/反馈和规划失败主导了执行/基础错误，而单一的标量成功率无法解释。我们将这些发现与更新的长时间跨度CUA基准联系起来，并推导出CUA评估的阶段特定设计规则。
cs.AI / 99 / 2607.28375
HyperClaim: Fine-Grained Cross-Modal Hypergraph Reasoning for Video Misinformation Detection
HyperClaim：用于视频虚假信息检测的细粒度跨模态超图推理
Abstract
Video misinformation detection is often approached through global multimodal fusion or free-form multimodal reasoning. Both paradigms can under-represent localized authenticity cues that arise from coupled interactions among query phrases, contextual text, and short temporal spans of frames. Because such interactions are inherently higher-order, pairwise graph formulations are insufficient to capture multi-way cross-modal dependencies, whereas hypergraphs offer a suitable representation for these relations. We propose HyperClaim, a discriminative temporal hypergraph framework for sample-level authenticity classification. Using the title or benchmark-provided paired text as a claim-like query, HyperClaim constructs a sparse heterogeneous hypergraph over query tokens, evidence tokens, and sampled frames; applies confidence-aware filtering and source budgeting to form compact text-frame and short-range temporal evidence units; performs adaptive soft-incidence reasoning with residual text-video calibration; and aggregates textual, visual, and hyperedge states through a discrepancy-aware readout. Without relying on generated rationales or external tool calls, HyperClaim preserves fine-grained cross-modal and temporal structure that global fusion tends to flatten. Under the FactGuard temporal protocol, it achieves 83.7%, 82.0%, and 87.3% accuracy on FakeSV, FakeTT, and FakeVV, respectively, outperforming strong discriminative and reasoning-centric baselines. Learned incidence and attention weights further reveal token- and frame-level structure.
Chinese Translation
视频虚假信息检测通常通过全局多模态融合或自由形式的多模态推理来进行。这两种范式可能无法充分代表由于查询短语、上下文文本和短时间帧之间的耦合交互所产生的局部真实性线索。由于这种交互本质上是高阶的，成对图的形式不足以捕捉多方跨模态依赖关系，而超图则为这些关系提供了合适的表示。我们提出了HyperClaim，一种用于样本级真实性分类的判别性时间超图框架。HyperClaim使用标题或基准提供的配对文本作为类似声明的查询，构建一个稀疏的异构超图，涵盖查询标记、证据标记和采样帧；应用基于置信度的过滤和源预算，形成紧凑的文本-帧和短时间范围的证据单元；通过残差文本-视频校准执行自适应软发生推理；并通过考虑差异的读出聚合文本、视觉和超边状态。在不依赖生成的推理或外部工具调用的情况下，HyperClaim保留了细粒度的跨模态和时间结构，而全局融合往往会将其扁平化。在FactGuard时间协议下，它在FakeSV、FakeTT和FakeVV上的准确率分别达到了83.7%、82.0%和87.3%，超越了强大的判别性和推理中心基线。学习到的发生和注意力权重进一步揭示了标记和帧级结构。
cs.AI / 100 / 2607.28384
When Specifications Conflict: A Symmetry-Based Framework for Measuring LLM Preferences
当规范冲突时：基于对称性的框架用于测量大型语言模型的偏好
Abstract
Large language models (LLMs) are increasingly required to integrate multiple sources of information that may be inconsistent or conflicting. However, there is still a lack of controllable and attributable methods for analyzing how models resolve conflicts between competing specifications. We propose a controlled experimental framework for studying model preferences under conflicting specifications. By constructing specifications with explicit conflicts, the framework enables model choices between competing specifications to be directly observed and analyzed. A symmetry-based design further reduces confounding factors, allowing preferences across representation types to be compared systematically. We evaluate the framework on an executable mathematical benchmark with 550 conflict instances spanning 11 function families, comparing four representation types: pure natural language, formal language, naturalized formal language, and input--output examples. Results show systematic preference patterns rather than random behavior, with a consistent ordering: $ \text{Formal} \approx \text{Naturalized Formal} > \text{Pure Natural Language} > \text{Input--Output Examples} $. Example effects further depend on model capability and function family. We extend the framework to heterogeneous specification conflicts in Boolean algebra, code generation, and the clinical domain, demonstrating its applicability across diverse tasks and specification forms. The framework provides a unified approach for measuring how LLMs resolve conflicts between competing sources of information.
Chinese Translation
大型语言模型（LLMs）越来越需要整合可能不一致或冲突的多种信息来源。然而，目前仍缺乏可控和可归因的方法来分析模型如何解决竞争规范之间的冲突。我们提出了一种受控实验框架，用于研究在冲突规范下模型的偏好。通过构建具有明确冲突的规范，该框架使得可以直接观察和分析模型在竞争规范之间的选择。基于对称性的设计进一步减少了混杂因素，使得不同表示类型的偏好能够系统地进行比较。我们在一个可执行的数学基准上评估该框架，该基准包含550个冲突实例，涵盖11个函数族，并比较了四种表示类型：纯自然语言、形式语言、自然化形式语言和输入-输出示例。结果显示出系统的偏好模式，而非随机行为，且有一致的排序：$ ext{Formal} ext{ } ext{approx} ext{ } ext{Naturalized Formal} > ext{Pure Natural Language} > ext{Input--Output Examples}$。示例效应进一步依赖于模型能力和函数族。我们将该框架扩展到布尔代数、代码生成和临床领域中的异构规范冲突，展示了其在多样化任务和规范形式中的适用性。该框架提供了一种统一的方法来测量大型语言模型如何解决竞争信息来源之间的冲突。
cs.AI / 101 / 2607.28397
GLM-RAG: Graph Language Models for Graph-Based Retrieval-Augmented Generation
GLM-RAG：基于图的检索增强生成的图语言模型
Abstract
Retrieval-augmented generation (RAG) over knowledge graphs requires retrievers that can effectively capture both graph structure and semantic information. Recent approaches have explored graph neural network (GNN)-based retrievers to model graph topology in multi-hop reasoning tasks. In parallel, graph language models (GLMs) have emerged as a promising paradigm that integrates graph reasoning and the semantic capabilities of language models. In this work, we introduce a GLM-based retriever and investigate the comparative strengths of GLM-based, GNN-based, and traditional vector-search-based retrievers in single- and multi-hop RAG settings, and with a particular focus on transferability to unseen domains. Our findings suggest that finetuned GLM retrievers generalize better out of domain, achieving SOTA on two multi-hop benchmarks. On in-domain multi-hop QA datasets they remain comparable to prior work, with promising scaling as parameters and subgraph coverage increase. GNN-based retrievers achieve higher graph coverage with an efficient training setup, whereas the vector-search baseline excels at single-hop datasets.
Chinese Translation
基于知识图谱的检索增强生成（RAG）需要能够有效捕捉图结构和语义信息的检索器。近期的方法探索了基于图神经网络（GNN）的检索器，以在多跳推理任务中建模图拓扑。同时，图语言模型（GLM）作为一种有前景的范式，结合了图推理和语言模型的语义能力。在本研究中，我们介绍了一种基于GLM的检索器，并调查了基于GLM、基于GNN和传统向量搜索的检索器在单跳和多跳RAG设置中的比较优势，特别关注其在未见领域的可迁移性。我们的研究结果表明，经过微调的GLM检索器在领域外的泛化能力更强，在两个多跳基准测试中达到了最先进水平（SOTA）。在领域内的多跳问答数据集上，它们与之前的工作相当，并且随着参数和子图覆盖率的增加，展现出良好的扩展性。基于GNN的检索器在高效的训练设置下实现了更高的图覆盖率，而向量搜索基线在单跳数据集上表现优异。
cs.AI / 102 / 2607.28405
QuantWAMs: Calibrating at the Right Granularity for World Action Models
QuantWAMs：为世界行动模型在适当粒度下进行校准
Abstract
World Action Models (WAMs) jointly predict future observations and actions, but their iterative denoising and closed-loop execution make efficient deployment costly. Existing post-training quantization (PTQ) methods are poorly suited to WAMs because they rely on open-loop objectives, homogeneous model assumptions, and calibration distributions that do not reflect deployment. We present QuantWAMs, a PTQ framework that aligns quantization decisions with the calibration context defined by model structure, rollout distribution, and task objective. QuantWAMs introduces three strategies: shared-basis outlier calibration, which pools activation evidence only across coordinate-compatible modules; co-training-objective saliency, which computes empirical-Fisher scores from the joint video--action gradient and assigns weight precision at a calibration-stable layer granularity; and fixed-intervention rollout auditing, which revises denoising-step protection schedules using reachable closed-loop states without changing the precision budget. We evaluate QuantWAMs on Fast-WAM and LingBot-VA across RoboTwin 2.0, LIBERO, and real-robot manipulation with an AgiBot G2. Under a W4A4-dominant setting, the reported simulation means differ from FP16 by 0.2--0.7 percentage points. Real-robot trials further establish deployment feasibility on three manipulation tasks. For the targeted video and action blocks, QuantWAMs reduces peak weight-and-activation memory to about 29\% of FP16 and provides 1.4--1.6$\times$ block-level speedups.
Chinese Translation
世界行动模型（WAMs）共同预测未来的观察和行动，但其迭代去噪和闭环执行使得高效部署成本高昂。现有的后训练量化（PTQ）方法不适合WAMs，因为它们依赖于开放循环目标、同质模型假设以及不反映部署的校准分布。我们提出了QuantWAMs，一个PTQ框架，它将量化决策与由模型结构、展开分布和任务目标定义的校准上下文对齐。QuantWAMs引入了三种策略：共享基础异常值校准，仅在坐标兼容模块之间汇聚激活证据；共同训练目标显著性，从联合视频-动作梯度计算经验费舍尔分数，并在校准稳定层粒度上分配权重精度；以及固定干预展开审计，使用可达的闭环状态修订去噪步骤保护计划，而不改变精度预算。我们在RoboTwin 2.0、LIBERO和使用AgiBot G2进行的真实机器人操作中评估了QuantWAMs在Fast-WAM和LingBot-VA上的表现。在W4A4主导设置下，报告的模拟均值与FP16相比差异为0.2-0.7个百分点。真实机器人试验进一步确立了在三项操作任务上的部署可行性。对于目标视频和动作块，QuantWAMs将峰值权重和激活内存减少到约FP16的29 ext{%}，并提供1.4-1.6$ imes$的块级加速。
cs.AI / 103 / 2607.28418
WIDE: Boosting Adaptive LLM Inference via Token-level Dynamic Width Pruning
WIDE：通过令牌级动态宽度剪枝提升自适应大语言模型推理
Abstract
Pruning is a promising approach for improving the efficiency of LLMs. Existing static structured pruning methods are hardware-friendly and can deliver practical throughput gains, but their input-agnostic computation allocation often causes substantial accuracy degradation under aggressive sparsity. Recent dynamic sparsity methods improve quality retention by adapting computation to individual inputs, yet they remain largely limited to coarse-grained structural decisions and their practical acceleration under real-world inference scenarios remains challenging. To address these challenges, we present WIDE, the first end-to-end differentiable token-level dynamic width pruning framework designed for both prefill and decode scenarios. WIDE enables fine-grained computation allocation by allowing each token to dynamically select attention-head groups and FFN-channel groups, extending dynamic pruning beyond layer-level decisions to neuron-block-level granularity. Through a two-stage training pipeline, WIDE learns effective token-wise sparse execution patterns and achieves substantially better quality retention than existing approaches. To make such fine-grained dynamic pruning practical, we further propose a pruning--kernel co-design framework that decomposes dynamic sparsity acceleration into mask reordering, hardware-agnostic block-level skipping, and hardware-dependent intra-block skipping, enabling efficient execution across different granularities. At 50% sparsity, WIDE provides 55.1% performance boost when compared to the state-of-the-art dynamic depth pruning under calibration-only settings. Under prefill and decoding inference workloads, WIDE achieves close-to-theoretical kernel-level speedups of up to 1.98x for prefill and 4.95x for decoding, as well as 1.68x and 1.55x end-to-end acceleration. Our code is available at https://github.com/EIT-NLP/LLM-Pruning/tree/main/WIDE.
Chinese Translation
剪枝是一种提高大语言模型（LLMs）效率的有前景的方法。现有的静态结构剪枝方法对硬件友好，并能提供实际的吞吐量提升，但其输入无关的计算分配往往在激进稀疏下导致显著的准确性下降。最近的动态稀疏方法通过将计算适应于单个输入来改善质量保留，然而它们仍然主要局限于粗粒度的结构决策，并且在现实推理场景下的实际加速仍然具有挑战性。为了解决这些问题，我们提出了WIDE，这是第一个端到端可微分的令牌级动态宽度剪枝框架，旨在处理预填充和解码场景。WIDE通过允许每个令牌动态选择注意力头组和前馈网络（FFN）通道组，实现了细粒度的计算分配，将动态剪枝扩展到神经元块级别的粒度。通过两阶段的训练流程，WIDE学习有效的令牌级稀疏执行模式，并在质量保留方面显著优于现有方法。为了使这种细粒度动态剪枝变得实用，我们进一步提出了一种剪枝-内核协同设计框架，将动态稀疏加速分解为掩码重排序、硬件无关的块级跳过和硬件相关的块内跳过，从而实现不同粒度下的高效执行。在50%的稀疏率下，与最先进的动态深度剪枝方法相比，WIDE在仅校准设置下提供了55.1%的性能提升。在预填充和解码推理工作负载下，WIDE实现了接近理论内核级的加速，预填充可达1.98倍，解码可达4.95倍，以及端到端加速分别为1.68倍和1.55倍。我们的代码可在https://github.com/EIT-NLP/LLM-Pruning/tree/main/WIDE获取。
cs.AI / 104 / 2607.28421
When Derived Measurements Mislead: Quantifying and Mitigating LLM Over-Trust with Privileged-Modality Reliability Evidence
当派生测量误导时：量化和减轻大型语言模型过度信任的特权模态可靠性证据
Abstract
Derived measurements increasingly enter large language model (LLM) pipelines as direct facts despite their instance-dependent validity. We define derived-feature over-trust (DFOT) as the failure in which a downstream LLM assigns such a measurement the epistemic status of a direct fact or uses it outside its valid scope. Using physiological sensing as a case study, D1 tests acceptance of a PPG-derived rhythm contradicted by offline ECG, whereas D2 tests rejection of an offline-confirmed reliable PPG rhythm under misleading severe history. ECG supplies training supervision and offline reference construction but is never shown to the LLM. Five estimands quantify this chain: conflict over-trust rate (COTR) and context-induced error rate (CIR) characterize D1/D2; correct repair rate (CRR) measures frozen-error repair; evidence-specific repair margin (ESRM) contrasts matched and patient-disjoint shuffled evidence; and utility harm rate (UHR) measures unnecessary verification among HIGH-reliability cases used without verification at baseline. The framework does not depend on a particular reliability generator. We demonstrate it on 50,000 paired PPG-ECG records using ECG-to-PPG privileged distillation as an illustrative baseline and PPG-only inference. On a protocol-locked 187-patient test, the baseline improves four repair and specificity endpoints by 1.82-6.69 percentage points, with all paired confidence intervals excluding zero; UHR increases by 0.67 percentage points (95% CI: -0.4 to +1.7). DFOT provides a common evaluation target for stronger mitigation methods. The code is available at https://github.com/Zongheng-Guo/When-Derived-Measurements-Mislead.
Chinese Translation
派生测量越来越多地作为直接事实进入大型语言模型（LLM）管道，尽管它们的有效性依赖于具体实例。我们将派生特征过度信任（DFOT）定义为下游LLM将此类测量赋予直接事实的认识状态或在其有效范围之外使用的失败。以生理传感为案例研究，D1测试与离线心电图（ECG）相矛盾的基于PPG（光电容积脉搏波）派生节律的接受度，而D2测试在误导性严重历史下拒绝离线确认的可靠PPG节律。ECG提供训练监督和离线参考构建，但从未向LLM展示。五个估计量量化了这一链条：冲突过度信任率（COTR）和上下文诱导错误率（CIR）描述D1/D2；正确修复率（CRR）衡量冻结错误的修复；证据特定修复边际（ESRM）对比匹配和患者不重叠洗牌证据；效用损害率（UHR）衡量在基线未验证的高可靠性案例中不必要的验证。该框架不依赖于特定的可靠性生成器。我们在50,000对PPG-ECG记录上展示了该框架，使用ECG到PPG的特权蒸馏作为说明性基线和仅PPG推断。在一个协议锁定的187名患者测试中，基线在四个修复和特异性端点上提高了1.82-6.69个百分点，所有配对置信区间均不包括零；UHR增加了0.67个百分点（95% CI：-0.4到+1.7）。DFOT为更强的减轻方法提供了共同的评估目标。代码可在https://github.com/Zongheng-Guo/When-Derived-Measurements-Mislead获取。
cs.AI / 105 / 2607.28432
A foundation model of numerical intelligence with cross-disciplinary generalization
具有跨学科泛化能力的数值智能基础模型
Abstract
Intelligence is commonly understood as the ability to acquire and apply knowledge, adapt to unfamiliar situations and solve new problems. Large language models exhibit this capacity by inferring task-relevant knowledge from textual context and applying it to new tasks. Yet intelligence need not be confined to language. For scientific and social systems, we need models that acquire and apply knowledge from numerical context-an ability we call numerical intelligence. Here we introduce UNified In-Context Operator Networks (UNICON), a foundation model that exhibits numerical intelligence across disciplines. Using graph-based examples from a system as context, UNICON infers the predictive relation shared across them and applies it to queries from the same system. Across scientific and social systems, including those from disciplines absent from training, the same model approaches specialist performance without retraining. Combining UNICON with language-model agents yields further gains, enabling it to surpass state-of-the-art specialists in a discipline unseen in training. We further show that training-corpus diversity improves generalization to unseen disciplines. Together, these results establish UNICON as a foundation model of numerical intelligence and position it as a building block for a broader ecosystem of artificial intelligence.
Chinese Translation
智能通常被理解为获取和应用知识、适应不熟悉情况以及解决新问题的能力。大型语言模型通过从文本上下文中推断与任务相关的知识并将其应用于新任务，展现了这种能力。然而，智能并不局限于语言。对于科学和社会系统，我们需要能够从数值上下文中获取和应用知识的模型——我们称之为数值智能。在这里，我们介绍了统一上下文操作网络（UNified In-Context Operator Networks，UNICON），这是一个在各学科中展现数值智能的基础模型。UNICON使用来自系统的图形示例作为上下文，推断它们之间共享的预测关系，并将其应用于来自同一系统的查询。在包括那些在训练中缺失的学科的科学和社会系统中，同一模型在不重新训练的情况下接近专业水平。将UNICON与语言模型代理结合使用进一步提升了性能，使其在未见过的学科中超越了最先进的专业人士。我们进一步展示了训练语料库的多样性改善了对未见学科的泛化能力。这些结果共同确立了UNICON作为数值智能基础模型的地位，并将其定位为更广泛人工智能生态系统的构建块。
cs.AI / 106 / 2607.28434
Metaphor Tracer: A Theory-Informed Analysis of Hidden States
隐喻追踪器：一种基于理论的隐状态分析
Abstract
What do a language model's hidden states say about the organization of a single text? From one forward pass, without training, we score every token position on two properties. The *aggregator* measures whether the position consolidates the whole text into a stable configuration. The *differentiator*, whether other tokens are transiently carried into its subspace as the model reads: metaphor in its root sense, transport. Constants were frozen on one discovery text; every other is confirmatory. The aggregator is not, in the classic sense, an information measure, nor a measure of salience. Across three unrelated models, as a signifier repeats, its surprisal and its attention drain while its aggregator score holds: the channel marks a token's place in the text. That this tracks a reading rests on independent ground truth: an engineered register the aggregator follows across its boundaries (6/6 cells), and a psychoanalyst's marking of clinical transcripts, fixed before the instrument existed, in 34/36 cells, with a graded increment above lexical controls and dissociations no type-level measure reproduces. A transfer test gives the result its shape: the model whose token structure travels with lexical type reads the singular discourse worst, and in a matched base/instruct pair tuning raises fidelity without moving type-transfer. Structural value is a property of a token's place in *this* text, not of its vector alone: a relational rather than essentialist reading of hidden states, operationalizing theory that predated the instrument.
Chinese Translation
语言模型的隐状态对单一文本的组织有什么启示？通过一次前向传播，在不进行训练的情况下，我们对每个标记位置的两个属性进行评分。*聚合器*衡量该位置是否将整个文本整合为一个稳定的配置。*区分器*则衡量在模型读取时，其他标记是否暂时被带入其子空间：隐喻在其根本意义上是运输。常量在一个发现文本上被冻结；其他每个文本都是确认性的。聚合器在经典意义上既不是信息度量，也不是显著性度量。在三个无关模型中，随着一个符号的重复，其惊讶度和注意力消耗，而聚合器得分保持不变：该通道标记了标记在文本中的位置。这个跟踪阅读的结果基于独立的真实数据：一个工程化的登记，聚合器在其边界（6/6单元）内遵循，以及心理分析师对临床记录的标记，在仪器存在之前固定，在34/36单元中，具有高于词汇控制和解离的分级增量，而没有任何类型层面的度量能够重现。转移测试赋予结果其形状：与词汇类型一起移动的标记结构的模型在阅读单一话语时表现最差，而在匹配的基础/指令对中调优则提高了保真度，而不影响类型转移。结构价值是标记在*这*个文本中的位置的属性，而不仅仅是其向量：对隐状态的关系性而非本质主义的解读，操作化了在仪器之前就存在的理论。
cs.AI / 107 / 2607.28457
SVR: Self-Verifying Refinement via Joint Verdict-Confidence Reinforcement Learning for Adaptive Test-Time Compute
SVR：通过联合裁决置信度强化学习实现自我验证的自我精炼以适应测试时计算
Abstract
Scaling test-time computation can improve language-model reasoning, but uniform budgets waste computation on easy inputs, while verifier-guided refinement relies on external feedback. We introduce Self-Verifying Refinement (SVR), an oracle-free multi-turn reinforcement learning framework that learns to use self-verification as a compute-control policy. At each turn, the model produces a solution together with a discrete correctness verdict and a confidence score; it retains the current answer only when the verdict is Correct and confidence exceeds a threshold, and otherwise continues refinement using its own self-verification. Ground-truth correctness is used only to construct training rewards and is never exposed to the policy through refinement prompts or required at inference. SVR is trained with GRPO on fixed-horizon trajectories using rewards that promote solution correctness, calibration-aware self-verification, and stop-ready correct states; adaptive stopping is activated only at inference. On seven mathematical reasoning benchmarks with Qwen3.5-2B, SVR achieves a macro-average accuracy of 0.563 with only 2.99 inference turns on average. In the evaluated complete-system comparison, it exceeds standard GRPO, strong multi-turn baselines, and a fixed-budget oracle-guided score-feedback reference while requiring substantially fewer turns than fixed ten-turn inference. These results demonstrate that learned self-verification can serve as an effective internal control signal for answer retention and adaptive test-time compute allocation.
Chinese Translation
扩展测试时计算可以改善语言模型的推理能力，但统一的预算会在简单输入上浪费计算，而验证者引导的精炼依赖于外部反馈。我们提出了自我验证精炼（Self-Verifying Refinement, SVR），这是一个无oracle的多轮强化学习框架，学习将自我验证作为计算控制策略。在每一轮中，模型产生一个解决方案，并给出一个离散的正确性裁决和置信度分数；只有当裁决为正确且置信度超过阈值时，模型才会保留当前答案，否则将继续使用自我验证进行精炼。真实的正确性仅用于构建训练奖励，并不会通过精炼提示暴露给策略或在推理时需要。SVR使用GRPO在固定时间轨迹上进行训练，奖励促进解决方案的正确性、校准意识的自我验证和准备停止的正确状态；自适应停止仅在推理时激活。在七个数学推理基准上，使用Qwen3.5-2B，SVR的宏平均准确率为0.563，平均仅需2.99次推理轮次。在评估的完整系统比较中，它超越了标准GRPO、强大的多轮基线以及固定预算的oracle引导评分反馈参考，同时所需轮次显著少于固定的十轮推理。这些结果表明，学习到的自我验证可以作为答案保留和自适应测试时计算分配的有效内部控制信号。
cs.AI / 108 / 2607.28459
LeanCSP: A Framework for Certifying Constraint Reformulation and Solving in Lean
LeanCSP：一个用于认证约束重构与求解的框架
Abstract
Constraint programming is a core technology for solving complex combinatorial problems in scheduling, planning, configuration, and verification. Trusting its results therefore demands guarantees at two levels: that reformulations applied beforehand are semantics-preserving, and that solvers produce correct answers. In this work, we introduce a framework that addresses both verification levels in the Lean theorem prover: it can be used to prove formulation-level properties, such as equivalence, equisatisfiability, and the correctness of symmetry-breaking constraints, parametrically for entire problem families; and to check solver-produced certificates for individual instances via translation backends to external formats such as MiniZinc, SMT-LIB, and OPB. Combining both levels yields an end-to-end workflow that establishes the satisfiability or unsatisfiability of a constraint problem without trusting the external solver. Experimental results show that our framework's verified symmetry breaking also pays off in practice: a single parametric proof per problem family, reused across all instance sizes, reduces solver search effort by a factor of up to 2x10^7, while the entire in-Lean certification stays affordable, taking at most a few minutes for our largest instances.
Chinese Translation
约束编程是解决调度、规划、配置和验证等复杂组合问题的核心技术。因此，信任其结果需要在两个层面上提供保证：即事先应用的重构是保持语义的，以及求解器产生的答案是正确的。在本研究中，我们提出了一个框架，旨在解决 Lean 定理证明器中的这两个验证层面：它可以用于证明形式层面的属性，如等价性、等满足性以及对称打破约束的正确性，参数化地适用于整个问题家族；并通过翻译后端检查求解器生成的证书，支持 MiniZinc、SMT-LIB 和 OPB 等外部格式。将这两个层面结合起来，形成了一个端到端的工作流程，能够在不信任外部求解器的情况下确定约束问题的可满足性或不可满足性。实验结果表明，我们框架中经过验证的对称打破在实践中也得到了回报：每个问题家族的单个参数化证明可以在所有实例规模中重复使用，将求解器的搜索工作量减少了多达 2x10^7 倍，而整个 Lean 内部认证的成本仍然可接受，对于我们最大的实例最多只需几分钟。
cs.AI / 109 / 2607.28466
A report-grounded vision-language foundation model for colonoscopy from 280000 routine reports
基于报告的结肠镜检查视觉-语言基础模型：来自280,000份常规报告的研究
Abstract
Vision-language models remain underused in colonoscopy despite the rich expert descriptions recorded in routine reports. These reports document lesion appearance, size and location but summarise entire procedures rather than caption individual frames, leaving clinical findings only weakly linked to the corresponding images. Here we develop EndoCLIP, a colonoscopy vision-language foundation model trained on 125,756 lesion-level image-text pairs progressively recovered from 280,476 routine colonoscopy records. Across lesion-level image-text retrieval, structured report generation and six multi-centre clinical classification tasks, EndoCLIP outperforms general-purpose and biomedical vision-language encoders in both zero-shot and linear-probe settings. On benign-versus-malignant classification, its linear probe approaches the performance of expert readers in a blinded study involving 12 endoscopists. These results suggest that recovering finding-to-frame correspondence can transform routine documentation into scalable supervision, enabling clinical targets to be specified in language rather than separately annotated for each task.
Chinese Translation
尽管常规报告中记录了丰富的专家描述，视觉-语言模型在结肠镜检查中的应用仍然不足。这些报告记录了病变的外观、大小和位置，但总结了整个过程而非逐帧描述，导致临床发现与相应图像之间的联系较弱。在此，我们开发了EndoCLIP，一个基于结肠镜检查的视觉-语言基础模型，训练于从280,476份常规结肠镜检查记录中逐步恢复的125,756对病变级别的图像-文本对。在病变级别的图像-文本检索、结构化报告生成和六个多中心临床分类任务中，EndoCLIP在零-shot和线性探测设置下均优于通用和生物医学视觉-语言编码器。在良性与恶性分类任务中，其线性探测的性能接近于在一项涉及12名内窥镜医师的盲测研究中的专家读者。这些结果表明，恢复发现与图像帧之间的对应关系可以将常规文档转变为可扩展的监督，使临床目标能够用语言指定，而不是为每个任务单独注释。
cs.AI / 110 / 2607.28470
Towards Autonomous Aircraft Surveillance from Nanosatellites through On-Board Inference and Generative Data Augmentation
通过机载推理和生成数据增强实现来自纳米卫星的自主航空监视
Abstract
Airborne surveillance from low Earth orbit is hindered by two interconnected bottlenecks: nanosatellites have a limited downlink budget, yet the conventional approach still transmits terabytes of raw imagery to the ground for processing, and open satellite datasets for aircraft are scarce and severely class-imbalanced. These limitations either delay timely decision-making or prevent standard detectors from learning robust representations of rare aircraft classes. In this paper, a workflow that combines on-board inference with generative data augmentation is proposed to address both limitations jointly. Inference is executed on a 6U CubeSat equipped with a low-power edge tensor accelerator, while a diffusion model fine-tuned through low-rank adaptation generates synthetic minority-class imagery. This synthetic output is automatically annotated, pseudo-labelled, by an intermediate detector and merged with classically augmented samples. The results show that the balanced dataset increases global mean average precision from 77.9% to 82.2%, with the minority class rising from F1=0.683 to F1=0.811, and that the quantised detector fits the on-chip memory and projects 25-30 frames per second on orbit. This approach contrasts with the conventional bent-pipe architecture, in which the satellite acts as a passive data collector. Therefore, the computational tests support the proposed workflow as a decision-support tool for real-time, autonomous airborne surveillance from nanosatellites.
Chinese Translation
来自低地球轨道的空中监视受到两个相互关联的瓶颈的制约：纳米卫星的下行带宽有限，而传统方法仍然将数TB的原始图像传输到地面进行处理，同时，关于飞机的开放卫星数据集稀缺且严重类别不平衡。这些限制要么延迟及时决策，要么阻止标准检测器学习稀有飞机类别的稳健表示。本文提出了一种将机载推理与生成数据增强相结合的工作流程，以共同解决这两个限制。推理在配备低功耗边缘张量加速器的6U立方卫星上执行，而通过低秩适应微调的扩散模型生成合成的少数类图像。该合成输出由中间检测器自动注释、伪标记，并与经典增强样本合并。结果表明，平衡数据集将全球平均精度从77.9%提高到82.2%，少数类的F1值从0.683上升到0.811，并且量化检测器适合片上内存，并在轨道上以每秒25-30帧的速度投影。这种方法与传统的弯管架构形成对比，在这种架构中，卫星充当被动数据收集器。因此，计算测试支持所提出的工作流程作为来自纳米卫星的实时自主空中监视的决策支持工具。
cs.AI / 111 / 2607.28481
A Fuzzy Rule-based Neuro-Symbolic Approach for Pipe Severity Prediction in Sewer Networks
基于模糊规则的神经符号方法在污水管网管道严重性预测中的应用
Abstract
Standard automated sewer pipe severity assessment relies on direct image classification, creating a "black box" where the link between visual defects and final severity scores remains implicit. This study introduces a modular, fuzzy rule-based neuro-symbolic framework that bridges this gap by decoupling neural perception from symbolic reasoning. The perception module utilizes a Swin Transformer to predict 14 multilabel inspection CODE degrees directly from images. For reasoning, a DT, specifically Weka's J48, algorithm is trained on ground-truth CODEs and severity labels, and its paths are converted into 19 fixed IF--THEN rules. Inference operates via fuzzy logic: t-norm activations from CODE conditions are weighted by rule confidence and combined with corresponding s-norms to produce interpretable class evidence. We assessed Product, {\L}ukasiewicz, and Hamacher operator pairs using a dataset of 3,244 images spanning five highly imbalanced severity classes. Ground-truth labels were robustly generated via consensus from five independent large language models analyzing original inspector notes. Our results show an improvement of accuracy, balanced accuracy, Macro F1 and MCC by 17.9%, 12.2%, 23.0%, and 17.3%, respectively, over image-only based classification. Overall, the framework combines competitive class-balanced performance with traceable reasoning from predicted CODE degrees to rule supports and severity evidence.
Chinese Translation
标准的自动化污水管道严重性评估依赖于直接的图像分类，形成了一个“黑箱”，在这个黑箱中，视觉缺陷与最终严重性评分之间的联系仍然隐含。本研究提出了一种模块化的基于模糊规则的神经符号框架，通过将神经感知与符号推理解耦，填补了这一空白。感知模块利用Swin Transformer直接从图像中预测14个多标签检查CODE等级。推理方面，使用决策树（DT），具体为Weka的J48算法，在真实标签CODE和严重性标签上进行训练，并将其路径转换为19条固定的IF-THEN规则。推理通过模糊逻辑进行：CODE条件的t-norm激活通过规则置信度加权，并与相应的s-norm结合，以生成可解释的类别证据。我们使用一个包含3,244张图像的数据集评估了Product、Łukasiewicz和Hamacher操作符对，涵盖了五个严重性类别的不平衡情况。真实标签通过五个独立的大型语言模型分析原始检查员笔记的共识生成，确保了其稳健性。我们的结果显示，与仅基于图像的分类相比，准确率、平衡准确率、宏观F1和MCC分别提高了17.9%、12.2%、23.0%和17.3%。总体而言，该框架结合了具有竞争力的类别平衡性能与从预测的CODE等级到规则支持和严重性证据的可追溯推理。
cs.AI / 112 / 2607.28488
SCOPE: Supply-Chain Operations through Coupled Policies for End-to-End Coordination
SCOPE：通过耦合政策实现端到端协调的供应链运营
Abstract
Can supply-chain AI move beyond isolated decision modules toward unified operational planning? A complete replenishment plan specifies which products each location carries, which upstream facility supplies it, how often it is replenished, and how deliveries are routed. These decisions are operationally coupled: the selected assortment changes the demand and load passed to later stages; source assignment and replenishment frequency reshape the delivery requests; and route feasibility and cost, in turn, determine the system value of the earlier choices. Yet in modern supply chains, these decisions are often handled by separate departments and optimized through separate systems, which can lead to stockouts, inventory exposure, and avoidable transportation. We propose SCOPE: Supply-Chain Operations through Coupled Policies for End-to-End Coordination, a composite policy model that represents supply-chain entities as tokens, contextualizes them through a shared operational representation, and maps each token type to the corresponding decision interface. Each decision builds on the partial plan formed by earlier decisions while the completed plan is evaluated using a shared system-level utility. We instantiate this framework in urban fresh-retail replenishment, where service frequency, assortment, capacity pressure, and road-network routing interact strongly, and evaluate it on real operational data from Dingdong and JD.com, two large-scale supply chains operating at different replenishment echelons. Across both settings, SCOPE consistently outperforms methods that optimize each decision stage separately, as well as practice-oriented baselines commonly used in supply-chain operations. These results show that learning and coordinating cross-department operational couplings lead to more effective end-to-end supply-chain decisions.
Chinese Translation
供应链人工智能能否超越孤立的决策模块，朝着统一的运营规划发展？完整的补货计划明确了每个地点所携带的产品、上游设施的供应来源、补货频率以及配送路线。这些决策在操作上是耦合的：所选的产品组合会改变后续阶段的需求和负载；来源分配和补货频率会重塑配送请求；而路线的可行性和成本又会影响早期选择的系统价值。然而，在现代供应链中，这些决策通常由不同部门处理，并通过独立系统进行优化，这可能导致缺货、库存风险和不必要的运输。我们提出了SCOPE：通过耦合政策实现端到端协调的供应链运营，这是一个复合政策模型，将供应链实体表示为标记，通过共享的操作表示对其进行上下文化，并将每种标记类型映射到相应的决策接口。每个决策都基于早期决策形成的部分计划，而完整计划则使用共享的系统级效用进行评估。我们在城市生鲜零售补货中实例化该框架，在该场景中，服务频率、产品组合、容量压力和道路网络路由之间相互作用强烈，并在来自叮咚和京东的真实运营数据上进行评估，这两者是处于不同补货层级的大规模供应链。在这两种环境中，SCOPE始终优于分别优化每个决策阶段的方法，以及在供应链运营中常用的实践导向基准。这些结果表明，学习和协调跨部门的操作耦合能够导致更有效的端到端供应链决策。
cs.AI / 113 / 2607.28503
InfoOps Bench: A live information operations safety benchmark
信息操作基准：一个实时的信息操作安全基准
Abstract
In this paper we present an active, constantly updated AI benchmark which measures the integrity of frontier language models against being co-opted for state-backed information operations. We draw on over 2,100 information operations from a live monitoring pipeline which tracks Russian, Chinese and Iranian state-backed information assets. Alongside this paper, we release a companion website that tracks the most prominent claims spread by state-backed media outlets, updated weekly, available from: pattrn.ai/research/infoopsbench. The dynamic nature of the benchmark makes it resistant to saturation. In the benchmark, we test 17 models from 8 providers across four prompt framings. We find that most models can be co-opted for information operations. Integrity scores, defined as the percentage of refused requests, range from 8.8% to 94.5%, an 85.7-percentage-point spread not explained by model size. Model choice also changes the character of the resulting operation. Some models fabricate details and produce output more harmful than the source material, others defuse claims even while complying, and fact-checking rates vary from 2.9% to 72.9%. Integrity against information operations is at least partly related to refusal to produce content even for benign claims, illustrating the challenge of balancing model usability with safety. With one exception (Z.ai's GLM 5.2), the Chinese-developed models sharply cut compliance on factually grounded but China-critical claims, dropping 48-70 percentage points relative to matched benign claims.
Chinese Translation
在本文中，我们提出了一个主动的、不断更新的人工智能基准，用于衡量前沿语言模型在被国家支持的信息操作中保持完整性的能力。我们基于一个实时监测管道，分析了超过2100个信息操作案例，该管道跟踪俄罗斯、中国和伊朗国家支持的信息资产。与本文一起，我们发布了一个伴随网站，跟踪由国家支持的媒体机构传播的最显著的主张，每周更新，网址为：pattrn.ai/research/infoopsbench。该基准的动态特性使其对饱和具有抵抗力。在基准测试中，我们对来自8个提供者的17个模型进行了测试，涵盖四种提示框架。我们发现大多数模型可以被用于信息操作。完整性得分（定义为拒绝请求的百分比）范围从8.8%到94.5%，85.7个百分点的差异不能仅通过模型大小解释。模型选择也改变了结果操作的特征。一些模型编造细节，产生比源材料更有害的输出，而其他模型则在遵从的同时削弱了主张，事实核查率从2.9%到72.9%不等。抵抗信息操作的完整性至少在一定程度上与拒绝产生内容（即使是对良性主张）相关，说明了在模型可用性与安全性之间平衡的挑战。除了一个例外（Z.ai的GLM 5.2），中国开发的模型在对中国批评的事实基础主张上显著降低了遵从率，相较于匹配的良性主张下降了48-70个百分点。
cs.AI / 114 / 2607.28523
Selective Credibility-Limited Belief Update
选择性可信度限制信念更新
Abstract
Belief update concerns changes in an agent's beliefs induced by changes in the underlying world. Standard Katsuno-Mendelzon update assumes that an epistemic input can be incorporated from every initially possible world, whereas credibility-limited belief update restricts, for each source world, the successor worlds regarded as credible or reachable. Nevertheless, existing credibility-limited approaches treat the epistemic input as an indivisible whole, and therefore cannot represent cases in which only part of a compound epistemic input can be realized. We introduce selective credibility-limited belief update, in which the epistemic input is transformed, relative to each source world, into a weaker proxy before the credibility-limited transition is performed. We provide semantic and axiomatic characterizations of the resulting class of update operators. We then identify two well-behaved sub-classes; namely, consistency-preserving update operators, which require every transformed epistemic input to be credible from its source world whenever the original epistemic input is consistent, and maximal consistency-preserving update operators, which additionally require the selected proxy to be maximally informative among the credible consequences of the original epistemic input. Finally, we establish the generality of the proposed framework by showing that credibility-limited belief update is recovered as a special case, while Katsuno--Mendelzon belief update emerges when credibility restrictions are removed and the transformation functions are taken to be identities. These results demonstrate that the framework provides a unified and strictly more expressive account of belief update, encompassing established approaches while supporting source-dependent selective acceptance.
Chinese Translation
信念更新涉及因基础世界的变化而引起的代理人信念的变化。标准的Katsuno-Mendelzon更新假设可以从每个最初可能的世界中纳入一个认知输入，而可信度限制信念更新则限制了每个源世界所认为可信或可达的后继世界。然而，现有的可信度限制方法将认知输入视为一个不可分割的整体，因此无法表示仅部分复合认知输入可以实现的情况。我们引入了选择性可信度限制信念更新，其中相对于每个源世界，认知输入在执行可信度限制转换之前被转化为一个较弱的代理。我们提供了所得到的更新算子的语义和公理特征。然后，我们识别出两个表现良好的子类；即一致性保持更新算子，要求每个转化后的认知输入在其源世界中是可信的，只要原始认知输入是一致的，以及最大一致性保持更新算子，额外要求所选择的代理在原始认知输入的可信后果中是信息量最大的。最后，我们通过证明可信度限制信念更新作为特例被恢复，同时在移除可信度限制并将转化函数视为恒等时，Katsuno-Mendelzon信念更新出现，确立了所提出框架的普遍性。这些结果表明，该框架提供了一个统一且严格更具表现力的信念更新解释，涵盖了已建立的方法，同时支持源依赖的选择性接受。
cs.AI / 115 / 2607.28527
MANTA: Multi-Agent Network Topology Adaptation for Self-Evolving Multi-Agent Systems
MANTA：用于自我演化多智能体系统的多智能体网络拓扑适应
Abstract
Large language model-based multi-agent systems improve complex problem solving through task decomposition, agent specialization, information exchange, and intermediate validation. However, existing systems typically treat communication topology as a fixed design choice or an offline optimization target. We introduce MANTA, a framework for Multi-Agent Network Topology Adaptation that enables communication structures to self-evolve at inference time. Before execution, MANTA initializes a task-conditioned topology from prior structural experience. During deployment, it monitors collaboration traces and applies bounded structural updates when the current organization becomes insufficient. These updates can modify agent roles, communication links, execution order, information visibility, and validation pathways while preserving the task interface and agent budget. We evaluate MANTA against representative single-agent and multi-agent baselines on five benchmarks spanning information seeking, tool use, planning, workflow execution, and mathematical reasoning. MANTA achieves the highest average score of 74.0, outperforming the strongest baseline by 5.8 percentage points and obtaining the best result on PlanCraft. These results show that inference-time self-improvement can extend to the architecture of collaboration itself.
Chinese Translation
基于大型语言模型的多智能体系统通过任务分解、智能体专业化、信息交换和中间验证来改善复杂问题的解决。然而，现有系统通常将通信拓扑视为固定的设计选择或离线优化目标。我们提出了MANTA，一个多智能体网络拓扑适应框架，使得通信结构能够在推理时自我演化。在执行之前，MANTA根据先前的结构经验初始化一个任务条件的拓扑。在部署过程中，它监控协作轨迹，并在当前组织不足时应用有限的结构更新。这些更新可以修改智能体角色、通信链接、执行顺序、信息可见性和验证路径，同时保持任务接口和智能体预算。我们在五个基准测试上评估了MANTA，这些基准涵盖了信息寻求、工具使用、规划、工作流执行和数学推理，结果显示MANTA的平均得分为74.0，超越了最强基线5.8个百分点，并在PlanCraft上取得了最佳结果。这些结果表明，推理时的自我改进可以扩展到协作架构本身。
cs.AI / 116 / 2607.28573
Rethinking Inference-Time Scaling in Local Computer-Use Agents: Failure Modes and Compute Tradeoffs
重新思考本地计算机使用代理的推理时间扩展：失败模式与计算权衡
Abstract
Deploying autonomous computer-use agents (CUAs) locally is increasingly important for privacy, cost efficiency, and practical usability, yet improving their performance under strict hardware constraints remains challenging. While recent studies show that inference-time scaling can improve frontier computer-use agents through additional computation during execution, its effectiveness for resource-constrained local models remains poorly understood. We present a systematic empirical study of inference-time scaling in local CUAs across contextual, temporal, structural, and parallel dimensions. We evaluate Qwen3-VL-8B/30B-A3B, UI-TARS-1.5-7B, and OpenCUA-7B on the OSWorld benchmark. Our results show that additional computation often yields diminishing returns while changing failure modes. Contextual scaling provides historical grounding that improves trajectory stability and task accuracy, but its gains saturate as token cost increases and failures shift from repetitive or stalled trajectories toward premature false successes. Temporal scaling similarly reduces max-step stalls, yet does not substantially improve task success, indicating that longer horizons often extend erroneous trajectories rather than correct them. We further find that structural decomposition can introduce planning and formatting overhead in local two-stage agents, while parallel scaling partially mitigates these failures at a substantial computational cost. Overall, our findings suggest that efficient local CUAs require selective compute allocation, failure-aware control mechanisms, and agentic frameworks designed around the capabilities and limitations of local models.
Chinese Translation
在隐私、成本效率和实际可用性方面，本地部署自主计算机使用代理（CUAs）变得越来越重要，但在严格的硬件限制下提高其性能仍然具有挑战性。尽管最近的研究表明，推理时间扩展可以通过在执行过程中增加计算来改善前沿计算机使用代理的性能，但其在资源受限的本地模型中的有效性仍然不甚明了。我们对本地CUAs中的推理时间扩展进行了系统的实证研究，涵盖了上下文、时间、结构和并行维度。我们在OSWorld基准上评估了Qwen3-VL-8B/30B-A3B、UI-TARS-1.5-7B和OpenCUA-7B。我们的结果表明，额外的计算通常会导致收益递减，同时改变失败模式。上下文扩展提供了历史基础，改善了轨迹稳定性和任务准确性，但随着标记成本的增加，其收益会饱和，失败模式也从重复或停滞的轨迹转向过早的虚假成功。时间扩展同样减少了最大步骤停滞，但并未显著提高任务成功率，表明较长的时间范围往往延长了错误轨迹而不是纠正它们。我们进一步发现，结构分解可能在本地两阶段代理中引入规划和格式化的开销，而并行扩展在显著的计算成本下部分缓解了这些失败。总体而言，我们的研究结果表明，高效的本地CUAs需要选择性地分配计算资源、关注失败的控制机制，以及围绕本地模型的能力和局限性设计的代理框架。
cs.AI / 117 / 2607.28580
DualG-MRAG: Decoupling Macro-Reasoning and Micro-Matching for Multimodal Retrieval-Augmented Generation
DualG-MRAG：解耦宏观推理与微观匹配的多模态检索增强生成
Abstract
While Multimodal Retrieval-Augmented Generation (MM-RAG) has shown promising results, it still struggles with complex multi-hop reasoning tasks. Existing methods primarily focus on independent instance-level matching, which often fails to capture explicit relationships across modalities and documents. Although Graph-enhanced methods introduce structural modeling, they face a fundamental challenge in multimodal scenarios: incorporating fine-grained visual features leads to rapid graph expansion and retrieval noise, whereas coarse-grained representations cause the discarding of critical local evidence. To address this dilemma, we propose DualG-MRAG, a Dual-tier framework that introduces a decoupled architecture comprising Macro-reasoning and Micro-matching Graphs for Multimodal RAG. Specifically, to suppress retrieval noise by isolating global structural reasoning from fine-grained evidence matching, we construct a Macro Graph for global topological routing and a Micro Graph for precise local verification. Subsequently, to enable dynamic relevance propagation across heterogeneous evidence sources, we formulate retrieval as a query-driven message passing process via a GNN Retriever. Furthermore, to provide the generative model with coherent structural guidance, we introduce a dynamic programming decoding mechanism that extracts explicit reasoning paths directly from the GNN's forward pass, replacing the standard input of isolated document chunks. Extensive experiments demonstrate that DualG-MRAG outperforms baselines in both evidence recall and complex QA accuracy.
Chinese Translation
尽管多模态检索增强生成（MM-RAG）已显示出良好的效果，但在复杂的多跳推理任务中仍然存在困难。现有方法主要集中于独立的实例级匹配，这往往无法捕捉跨模态和文档的显性关系。尽管图增强方法引入了结构建模，但在多模态场景中面临一个根本性挑战：细粒度视觉特征的引入导致图的快速扩展和检索噪声，而粗粒度表示则会导致关键局部证据的丢弃。为了解决这一困境，我们提出了DualG-MRAG，一个双层框架，引入了一个解耦架构，包括用于多模态RAG的宏观推理图和微观匹配图。具体而言，为了通过将全局结构推理与细粒度证据匹配隔离来抑制检索噪声，我们构建了一个用于全局拓扑路由的宏观图和一个用于精确局部验证的微观图。随后，为了实现跨异构证据源的动态相关传播，我们将检索过程形式化为通过GNN检索器的查询驱动消息传递过程。此外，为了为生成模型提供一致的结构指导，我们引入了一种动态编程解码机制，该机制直接从GNN的前向传递中提取显性推理路径，替代了孤立文档块的标准输入。大量实验表明，DualG-MRAG在证据召回和复杂问答准确性方面均优于基线方法。
cs.AI / 118 / 2607.28609
OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models
OSReward：为跨平台计算机使用奖励模型建立标准化评估
Sun, Qiushi, Cheng, Kanzhi, Wang, Yian, Yang, Bowen, Yan, Hang, Chen, Liheng, Xu, Fangzhi, Ding, Zichen, Chen, Nuo, Cao, Jialin, Gong, Xingdong, Li, Zehao, Jin, Kaiming, Yuan, Xinfeng, Liu, Zhoumianze, Gong, Jingyang, Yin, Zhangyue, Gao, Jiahui, Wu, Zhiyong, Xie, Tianbao, Zhang, Jianbing, Kao, Ben, Kong, Lingpeng
Abstract
Computer-using agents (CUAs) are advancing rapidly across the digital world. A CUA trajectory records the agent's actions, states, and reasoning. Verifying whether it fulfilled the task instruction is central to CUA evaluation, data curation, and reinforcement learning. Neither human-written verifiers nor human annotators can provide such verification at scale, so the field increasingly turns to vision-language models (VLMs) as judges of CUA trajectories. But a fundamental question has long gone unexamined: are these VLM judges reliable enough? To study it systematically, we introduce OSReward, a realistic, high-quality benchmark that evaluates VLM judges on CUA trajectories. The trajectories come from diverse agent backbones executing human-verified instructions across platforms, then rigorously labeled with ground-truth verdicts through multi-stage human annotation. Building on it, we derive OSReward-Hard, a challenge set concentrating genuinely hard cases, and OSReward-Multi for fine-grained efficiency and alignment scoring. The most comprehensive evaluation of VLM judges to date finds even state-of-the-art models fall short of an ideal judge, sharing a systematic leniency bias that mislabels failed runs as successes. The few reliable enough to trust are too expensive to run at scale, while affordable open models trail far behind. To close this gap, we construct and release OS-Shepherd-100K, an open corpus of reasoning-annotated trajectory judgments for the CUA community. On it, we train OS-Shepherd (9B and 35B), open reward models that supply low-cost, stable, and reliable reward signals, matching commercial judges at 30-60% lower cost than the frontier. Extensive analyses further inform the design of reliable CUA reward at scale. Our code, benchmark, dataset, and model checkpoints are available at https://os-copilot.github.io/OSReward-Home/.
Chinese Translation
计算机使用代理（CUAs）在数字世界中迅速发展。CUA轨迹记录了代理的行动、状态和推理。验证其是否完成任务指令是CUA评估、数据策划和强化学习的核心。人类编写的验证者和注释者无法大规模提供这种验证，因此该领域越来越多地转向视觉语言模型（VLMs）作为CUA轨迹的评判者。但一个根本性的问题长期以来未被审视：这些VLM评判者是否足够可靠？为了系统地研究这一问题，我们引入了OSReward，这是一个现实的、高质量的基准，用于评估VLM评判者在CUA轨迹上的表现。这些轨迹来自多样的代理骨干，执行跨平台的人类验证指令，并通过多阶段的人类注释严格标注了真实判决。基于此，我们衍生出OSReward-Hard，一个集中于真正困难案例的挑战集，以及OSReward-Multi，用于细粒度效率和对齐评分。迄今为止对VLM评判者的最全面评估发现，即使是最先进的模型也未能达到理想评判者的标准，存在系统性的宽容偏差，将失败的运行错误标记为成功。少数可靠的模型运行成本过高，难以大规模使用，而经济实惠的开放模型则远远落后。为了解决这一差距，我们构建并发布了OS-Shepherd-100K，这是一个为CUA社区提供推理注释轨迹判断的开放语料库。在此基础上，我们训练了OS-Shepherd（9B和35B），开放奖励模型，提供低成本、稳定且可靠的奖励信号，其成本比商业评判者低30-60%。广泛的分析进一步为大规模可靠CUA奖励的设计提供了信息。我们的代码、基准、数据集和模型检查点可在 https://os-copilot.github.io/OSReward-Home/ 获取。
cs.AI / 119 / 2607.28617
AISPA: User-Centric System Prompt Auditing for Large Language Model Applications
AISPA：面向用户的系统提示审计框架用于大型语言模型应用
Lin, Xiangning, Zhu, Shenzhe, Yang, Shu, Zhang, Zhenyu, Zhang, Haoqian, Zhao, Yipeng, Qian, Chengxuan, Wang, Tianwei, Zhang, Ziheng, Yuan, Zhenlong, Wang, Dingcheng, Wu, Juncheng, Si, Yuan, Liu, Jiaxin, Bi, Baolong, Mahari, Robert, South, Tobin, Greenwood, Dazza, He, Zexue, Bommasani, Rishi, Kazinnik, Sophia, Haupt, Andreas, Marro, Samuele, Brynjolfsson, Erik, Pentland, Alex, Pei, Jiaxin
Abstract
System prompts are instructions configured by developers to govern the behaviors of foundation models in AI applications. They are used throughout commercial AI products, but are rarely disclosed to the public or regulators, creating a serious trust and accountability gap in the wide deployment of AI systems. In this paper, we introduce Artificial Intelligence System Prompt Assurance (AISPA), a user-centric framework for systematically auditing system prompts in AI systems. AISPA examines specific parts of a system prompt and evaluates them along eight dimensions that matter to users. We then use this framework to review 3,249 instructions from system prompts in 88 commercial AI products, classifying each instruction as either protective (of users) or problematic. Our audit surfaces four core findings. First, system prompt design varies substantially across products and developers, with some organizations averaging over 60 protective instructions per product while others average fewer than 5. Second, protective instructions are widely adopted but shallow in scope: 98.9% of products contain at least one, yet only 24% cover all eight dimensions of the AISPA taxonomy. Third, system prompts have grown steadily longer and more protective of users, suggesting that user protection is becoming a more visible concern in commercial prompt design. Fourth, despite this progress, problematic instructions remain pervasive: roughly 40% of products contain at least one instruction that works against user interests, and protective and problematic instructions frequently coexist within the same prompt. Our findings highlight the need for greater transparency, standardization, and independent oversight for system prompts in commercial AI products.
Chinese Translation
系统提示是开发者配置的指令，用以管理基础模型在人工智能应用中的行为。这些提示在商业人工智能产品中广泛使用，但很少向公众或监管机构披露，从而在人工智能系统的广泛部署中造成了严重的信任和问责缺口。本文介绍了人工智能系统提示保障（AISPA），一个面向用户的框架，用于系统性审计人工智能系统中的系统提示。AISPA检查系统提示的特定部分，并从用户关心的八个维度对其进行评估。我们随后使用该框架审查了88个商业人工智能产品中的3,249条系统提示指令，将每条指令分类为保护性（保护用户）或问题性。我们的审计结果揭示了四个核心发现。首先，系统提示的设计在不同产品和开发者之间差异显著，有些组织每个产品平均超过60条保护性指令，而其他组织则少于5条。其次，保护性指令被广泛采用，但范围较浅：98.9%的产品至少包含一条保护性指令，但仅有24%的产品覆盖了AISPA分类法的所有八个维度。第三，系统提示的长度稳步增加，并且对用户的保护性增强，表明用户保护在商业提示设计中正变得越来越显著。第四，尽管取得了进展，问题性指令仍然普遍存在：大约40%的产品至少包含一条与用户利益相悖的指令，保护性和问题性指令经常在同一提示中共存。我们的发现强调了商业人工智能产品中系统提示需要更大透明度、标准化和独立监督的必要性。
cs.CL / 1 / 2607.27210
Prompt Chaining in Practice: A Case Study in Automated Scholarly Report Generation
实践中的提示链：自动化学术报告生成的案例研究
Abstract
The exponential growth of scholarly publications requires automated tools for effective information synthesis. However, simple, single-shot prompting methods often lack the reliability and quality required for complex synthesis tasks. This paper introduces and empirically evaluates a multi-stage prompt chaining methodology as a more reliable architectural pattern for such tasks. This approach is implemented in our system, AI SciBrief, which automatically generates scholarly digests. We conducted a comparative experiment, measuring the performance of our prompt chaining method against a carefully optimized single-shot baseline. Both systems were evaluated against a human-authored "gold standard" report for the "Education" domain. The results demonstrate a significant difference in reliability: our prompt chaining method achieved a 100% success rate, whereas the optimized baseline failed in 50% of its runs. In terms of quality, the proposed method also demonstrated a clear advantage, achieving a superior ROUGE-L F1-score (0.507 vs. 0.486), driven primarily by higher precision. We conclude that prompt chaining is a more dependable and effective engineering approach for complex, multi-step generative tasks, significantly mitigating the risks of failure and inconsistency inherent in monolithic prompts.
Chinese Translation
学术出版物的指数增长需要自动化工具以有效进行信息综合。然而，简单的单次提示方法往往缺乏复杂综合任务所需的可靠性和质量。本文介绍并实证评估了一种多阶段提示链方法，作为此类任务更可靠的架构模式。该方法在我们的系统AI SciBrief中实施，该系统能够自动生成学术摘要。我们进行了比较实验，测量了我们的提示链方法与经过精心优化的单次基线的性能。两个系统均与人类撰写的“黄金标准”报告进行了评估，报告主题为“教育”领域。结果显示在可靠性上存在显著差异：我们的提示链方法实现了100%的成功率，而优化后的基线在50%的运行中失败。在质量方面，所提方法也显示出明显优势，获得了更高的ROUGE-L F1-score（0.507对比0.486），主要得益于更高的精确度。我们得出结论，提示链是一种更可靠和有效的工程方法，适用于复杂的多步骤生成任务，显著降低了单一提示固有的失败和不一致风险。
cs.CL / 2 / 2607.27228
AI-assisted pre-review of open-source software submissions: an experience report from BOSC 2026
人工智能辅助的开源软件提交预审：来自BOSC 2026的经验报告
Abstract
Most conferences rely on peer-review of submissions, but as generative AI makes it easier than ever to prepare submission materials, some conferences are seeing an overwhelming surge of submissions. We wanted to see if generative AI could help our conference's volunteer reviewers by pre-reviewing abstracts for certain criteria. The Bioinformatics Open Source Conference (BOSC) was well-positioned to experiment with this, as we already had a detailed rubric used by reviewers to evaluate submitted abstracts on multiple criteria, including openness (public availability of the code or other content associated with the project), valid open source license, and "runnability" (how easy it is to download, build, and run the project - an important measure of reusability). For BOSC 2026, we built bosc-pre-review, an agentic skill that assessed six review criteria, and Runabilly, which builds and tests each project in a disposable Docker container for safety. The AI only gathered evidence to present to the reviewers; humans made every decision regarding the acceptance of the abstracts. After the review period, we surveyed the reviewers to determine how useful they found the pre-review. Most of those who responded said they found it useful, but they preferred to check the AI's conclusions against their own, rather than accepting the AI results unquestioningly.
Chinese Translation
大多数会议依赖于对提交材料的同行评审，但随着生成性人工智能使准备提交材料变得前所未有的简单，一些会议正面临提交数量的激增。我们希望探讨生成性人工智能是否能通过对摘要进行预审来帮助我们会议的志愿评审者，评估某些标准。生物信息学开源会议（BOSC）在这方面具有良好的实验基础，因为我们已经拥有一套详细的评分标准，供评审者用来在多个标准上评估提交的摘要，包括开放性（项目相关代码或其他内容的公开可用性）、有效的开源许可证和“可运行性”（下载、构建和运行项目的难易程度——这是可重用性的重要衡量标准）。在BOSC 2026中，我们构建了bosc-pre-review，一个评估六个评审标准的智能工具，以及Runabilly，它在一个一次性Docker容器中构建和测试每个项目以确保安全。人工智能仅收集证据以呈现给评审者；所有关于摘要接受的决定均由人类做出。在评审期结束后，我们对评审者进行了调查，以确定他们认为预审的实用性。大多数回应者表示他们认为预审有用，但他们更倾向于将人工智能的结论与自己的判断进行核对，而不是无条件接受人工智能的结果。
cs.CL / 3 / 2607.27232
Sympathetic Framing: Evaluating AI Alignment across Sociodemographic Groups
同情框架：评估人工智能在社会人口群体中的对齐情况
Abstract
Large Language Models (LLMs) are increasingly shaping how we consume information and form our worldview. This raises concerns beyond bias in AI: do LLMs grasp the emotional nuances conveyed via textual framing? In this work, we empirically evaluate how well an array of LLMs aligns with human emotional perception. Considering news headlines covering political and geopolitical conflicts, both human participants (n = 3011, a representative sample of the U.K. adult population, via a YouGov survey) and seven LLMs answered whether headlines evoked sympathy for a specified side in a conflict. We find that the correlation between AI and human evaluations varies across models, ranging from very high (0.789, GPT-5.2) to medium (0.4 ,Mistral Large 2512). Crucially, the leading models are broadly aligned with human judgments across all demographic subgroups, including age, gender, level of education, prior geopolitical knowledge, and participants' predispositions regarding the conflict, although there are statistically significant differences between groups. This research, with its robust design and large, demographically diverse dataset, offers the most comprehensive evaluation of LLMs' comprehension of news framing to date. Findings highlight an important, often-ignored aspect of differential alignment: even when aggregate performance is high, AI alignment is not universal -- it may correspond differently with demographic features and cultural norms. Considering or ignoring the need for differential alignment may therefore have significant implications for the development of ethical and useful AI systems.
Chinese Translation
大型语言模型（LLMs）越来越多地影响我们获取信息和形成世界观的方式。这引发了超越人工智能偏见的担忧：LLMs是否理解通过文本框架传达的情感细微差别？在本研究中，我们实证评估了一系列LLMs与人类情感感知的对齐程度。考虑到涵盖政治和地缘政治冲突的新闻标题，参与者（n = 3011，代表英国成年人口的样本，通过YouGov调查）和七个LLMs回答了这些标题是否引发了对冲突中某一方的同情。我们发现，人工智能与人类评估之间的相关性在不同模型之间有所不同，从非常高（0.789，GPT-5.2）到中等（0.4，Mistral Large 2512）。关键是，领先模型在所有人口子群体中与人类判断的对齐程度普遍较高，包括年龄、性别、教育水平、先前的地缘政治知识以及参与者对冲突的倾向，尽管不同群体之间存在统计显著差异。本研究以其稳健的设计和大规模、人口多样化的数据集，提供了迄今为止对LLMs理解新闻框架的最全面评估。研究结果突显了差异性对齐的一个重要且常被忽视的方面：即使整体表现较高，人工智能的对齐并非普遍适用——它可能与人口特征和文化规范的对应关系不同。因此，考虑或忽视差异性对齐的需求可能对伦理和实用人工智能系统的发展产生重大影响。
cs.CL / 4 / 2607.27353
LayerRAG-Bench: A Cross-Layer Reliability Benchmark for Agentic Retrieval-Augmented Generation
LayerRAG-Bench：一种跨层可靠性基准用于代理检索增强生成
Abstract
Agentic retrieval-augmented generation systems can produce answers that appear grounded while failing at the evidence, tool-contract, authorization, or session-state layer. We introduce LayerRAG-Bench, a controlled cross-layer reliability benchmark with 8 enterprise domains, 240 tasks, 9 fault scenarios, 2 contract modes, and 38,880 live task-level records across nine models from OpenAI, Anthropic, and Gemini. Schema normalization raises schema-drift success from 0.000 to 0.913, but stale evidence, missing tool output, denied permissions, and wrong-session context are not recovered by schema normalization. Groundedness-only evaluation also produces substantial false positives under stale and wrong-session evidence. These results support a layer-specific evaluation principle: a reliability intervention should be credited for repairing its target layer without being mistaken for a universal fix.
Chinese Translation
代理检索增强生成系统能够产生看似有依据的答案，但在证据、工具合约、授权或会话状态层面上却可能失败。我们引入了LayerRAG-Bench，这是一个受控的跨层可靠性基准，涵盖8个企业领域、240个任务、9种故障场景、2种合约模式，以及来自OpenAI、Anthropic和Gemini的九个模型的38,880个实时任务级记录。模式规范化将模式漂移的成功率从0.000提高到0.913，但陈旧的证据、缺失的工具输出、被拒绝的权限和错误的会话上下文并未通过模式规范化得到恢复。仅基于有依据性的评估在陈旧和错误会话证据下也产生了大量的假阳性。这些结果支持了一种特定层面的评估原则：可靠性干预应被视为修复其目标层的措施，而不应被误认为是普遍的解决方案。
cs.CL / 5 / 2607.27366
BridgeAlign: Bridging Preference Alignment for Humanities and Social Sciences
BridgeAlign：人文学科和社会科学的偏好对齐桥接
Abstract
While data synthesis for large language models (LLMs) is prevalent, it primarily targets domains with verifiable answers, overlooking open-ended humanities and social sciences (HSS), where nuanced quality judgments matter more than objective correctness. This makes preference alignment a natural paradigm for broad HSS tasks. Yet existing methods are either costly or not tailored to broad HSS disciplines. We thus propose BridgeAlign, among the first preference-alignment pipelines for broad HSS disciplines, with three phases: i) Seed Curation: curating HSS seed documents from web corpora via heuristic/LLM-based filtering and text refinement; ii) Preference Data Synthesis: generating preference triplets via persona-based instruction inversion with Q&A consistency checks; iii) Preference Optimization: moving beyond naive human-vs-model heuristics by first grounding preferences in HSS quality rubric, then generating transitional responses via controlled quality degradation to form near-boundary preference pairs for finer-grained quality discrimination. Aligning over 210k synthetic preference samples, BridgeAlign enables Qwen3-8B to achieve the best average across 17 benchmarks against 11 strong baselines; importantly, leading on both human-preference and knowledge-based capabilities at once, with no trade-off between them, as supported by extensive experiments and contextualized by existing theories.
Chinese Translation
尽管大型语言模型（LLMs）的数据合成已相当普遍，但其主要针对具有可验证答案的领域，忽视了开放式的人文学科和社会科学（HSS），在这些领域中，细致的质量判断比客观正确性更为重要。这使得偏好对齐成为广泛HSS任务的自然范式。然而，现有方法要么成本高昂，要么未针对广泛的HSS学科进行定制。因此，我们提出了BridgeAlign，这是针对广泛HSS学科的首批偏好对齐管道之一，分为三个阶段：i) 种子文档策划：通过启发式/基于LLM的过滤和文本精炼，从网络语料库中策划HSS种子文档；ii) 偏好数据合成：通过基于角色的指令反转和问答一致性检查生成偏好三元组；iii) 偏好优化：通过首先将偏好基于HSS质量标准进行基础化，然后通过受控质量降级生成过渡响应，以形成近边界的偏好对，以便进行更细粒度的质量区分，超越简单的人类与模型启发式。BridgeAlign对超过21万条合成偏好样本进行对齐，使Qwen3-8B在17个基准测试中实现了对11个强基线的最佳平均表现；重要的是，在人类偏好和基于知识的能力上同时领先，且二者之间没有权衡，这得到了广泛实验的支持，并与现有理论相结合。
cs.CL / 6 / 2607.27379
HSS-Synth: Humanities and Social Sciences Data Synthesis for LLMs
HSS-Synth：人文学科与社会科学数据合成用于大型语言模型
Abstract
High-quality, diverse data are vital for large language models (LLMs) but remain scarce and costly. Data synthesis is a viable alternative and succeeds on closed tasks, yet the humanities and social sciences (HSS) are overlooked, and their open-ended nature makes synthesis challenging. Moving beyond prior capability-centric, fragmented attempts, we adopt a subject-centric paradigm, define the first HSS domain system covering 14 mainstream fields, and introduce HSS-Synth, the first data synthesis pipeline for HSS. HSS-Synth comprises: (1) constructing seed documents from web corpora via multi-step filtering and text refinement evaluated by a judge; (2) specifying "requirements + persona" to backtranslate seed documents into diverse yet faithful instructions with a strict Q&A alignment check; and (3) breaking LLM response limits via teacher-forced Answering that feeds seed documents during response generation to anchor semantics, reduce hallucinations, and preserve tone and integrity. HSS-Synth yields 237k high-quality, diverse instruction-tuning samples that outperform 14 leading baselines on 16 benchmarks. The fine-tuned Qwen3-8B-Base sets a new SOTA and approaches the official Qwen3-8B, improving both human preference and knowledge capabilities without performance seesaws. Extensive experiments demonstrate HSS-Synth's robustness and transferability. Our code is publicly available at https://github.com/pengr/HSS-Synth.
Chinese Translation
高质量、多样化的数据对大型语言模型（LLMs）至关重要，但仍然稀缺且成本高昂。数据合成是一种可行的替代方案，并在封闭任务中取得成功，然而人文学科与社会科学（HSS）却被忽视，其开放性特征使得合成变得具有挑战性。我们超越了以能力为中心的碎片化尝试，采用以主题为中心的范式，定义了首个涵盖14个主流领域的HSS领域系统，并引入HSS-Synth，这是首个用于HSS的数据合成管道。HSS-Synth包括：（1）通过多步骤过滤和文本精炼从网络语料库构建种子文档，并由评审进行评估；（2）指定“需求 + 人物角色”，将种子文档反向翻译为多样化但忠实的指令，并进行严格的问答对齐检查；（3）通过教师强制回答（teacher-forced Answering）打破LLM响应限制，在响应生成过程中输入种子文档，以固定语义、减少幻觉并保持语气和完整性。HSS-Synth生成了237,000个高质量、多样化的指令调优样本，在16个基准测试中超越了14个领先基线。经过微调的Qwen3-8B-Base设定了新的状态-of-the-art（SOTA），并接近官方的Qwen3-8B，提升了人类偏好和知识能力而没有性能波动。大量实验表明HSS-Synth的鲁棒性和可迁移性。我们的代码已公开发布在 https://github.com/pengr/HSS-Synth。
cs.CL / 7 / 2607.27384
Same Facts, Different Diagnosis: Measuring and Mitigating Narrative Anchoring in Clinical Language Models
相同事实，不同诊断：测量和减轻临床语言模型中的叙事锚定
Abstract
Large language models used for clinical diagnostic reasoning are sensitive to sociolinguistic register, not just clinical content. We term this failure mode Narrative Anchoring: identical clinical facts expressed in different registers cause diagnostic outputs to diverge. Unlike prior demographic-bias work, which manipulates explicit identity tokens such as race or income, our benchmark isolates register as the sole channel of variation, with no demographic marker present in any form. We construct a dataset of 1,000 USMLE clinical vignettes, each rewritten into three sociolinguistically distinct personas under an independently audited fact-preservation guarantee, verified by a separate model that never sees the generation prompt. Across seven language models spanning three architecture families and scales, Narrative Anchoring is statistically significant under direct prompting in every model tested, with a Narrative Anchoring Gap of 0.064 to 0.151. Chain-of-thought reasoning and explicit debiasing instructions reduce the bias only partially, and their apparent gains are frequently confounded by accuracy collapse. We introduce NarrativeShield, a three-agent pipeline that structurally extracts and verifies clinical facts before diagnostic reasoning begins, reducing the Narrative Anchoring Gap to near-zero ($-0.004$ to $0.037$) and achieving the lowest rate of severely unstable decisions (DSS $<$ 0.8) of any method across all models, at a modest and mechanistically expected accuracy cost for most models. A stress test using a non-instruction-tuned base model shows that executing a debiasing intervention at all is gated by zero-shot instruction-following ability, not prompt content alone. We release our dataset, human-validated for fact preservation, as a standalone resource for studying register-based clinical bias.
Chinese Translation
用于临床诊断推理的大型语言模型对社会语言学语域敏感，而不仅仅是临床内容。我们将这种失效模式称为叙事锚定：以不同语域表达的相同临床事实导致诊断结果的偏差。与之前操控种族或收入等显性身份标记的群体偏见研究不同，我们的基准测试将语域作为唯一的变化通道，任何形式的群体标记均未出现。我们构建了一个包含1000个美国医学执照考试（USMLE）临床案例的数据集，每个案例在独立审核的事实保留保证下被重写为三种社会语言学上不同的人物角色，由一个从未看到生成提示的独立模型进行验证。在涵盖三种架构家族和规模的七个语言模型中，叙事锚定在每个测试模型中都在直接提示下具有统计显著性，叙事锚定差距为0.064至0.151。链式推理和显性去偏见指令仅部分减少了偏见，而它们的明显收益常常被准确性崩溃所混淆。我们提出了NarrativeShield，一个三代理管道，在诊断推理开始之前结构性地提取和验证临床事实，将叙事锚定差距降低到接近零（$-0.004$到$0.037$），并在所有模型中以适度且在机制上可预期的准确性成本实现了最低的严重不稳定决策率（DSS $<$ 0.8）。使用未经过指令调优的基础模型进行的压力测试表明，执行去偏见干预的能力受到零-shot 指令跟随能力的限制，而不仅仅是提示内容。我们发布了经过人工验证以确保事实保留的数据集，作为研究基于语域的临床偏见的独立资源。
cs.CL / 8 / 2607.27393
AHA-Memes: A Fine-Grained Multimodal Benchmark for Understanding Hate in Arabic Memes
AHA-Memes：理解阿拉伯恶搞的细粒度多模态基准
Abstract
Hateful memes are a growing form of multimodal online harm, where hostile intent is often conveyed through the joint interpretation of images, text, cultural references, and implicit targets. While hateful meme detection has advanced in high-resource languages, Arabic remains underexplored, with existing meme resources focusing mainly on propaganda or coarse harmful-content labels. We introduce AHA-Memes (Arabic HAteful Memes), which is, to our knowledge, the first large-scale Arabic hateful meme benchmark with fine-grained, multi-label annotations. The dataset includes 5K manually annotated memes using a taxonomy that captures hate types, i.e., attack strategies. We further provide ~66K silver-labeled memes to support future studies. We benchmark text-only, image-only, and late-fusion multimodal models, as well as few-shot in-context learning (ICL) and open- and closed-weight Vision-Language Models (VLMs) under zero-shot and fine-tuning settings. Our results establish strong baselines and highlight key challenges in culturally grounded Arabic hateful meme detection. We release the dataset, annotation guidelines, and evaluation scripts to support future research. WARNING: This paper contains examples that may be disturbing to readers.
Chinese Translation
仇恨恶搞是一种日益增长的多模态在线伤害形式，其中敌意意图通常通过对图像、文本、文化参考和隐含目标的联合解读来传达。尽管在高资源语言中，仇恨恶搞检测已有所进展，但阿拉伯语仍然未被充分探索，现有的恶搞资源主要集中在宣传或粗略的有害内容标签上。我们介绍了AHA-Memes（阿拉伯仇恨恶搞），据我们所知，这是第一个具有细粒度、多标签注释的大规模阿拉伯仇恨恶搞基准。该数据集包含5000个手动注释的恶搞，使用了一种捕捉仇恨类型（即攻击策略）的分类法。我们进一步提供约66,000个银标注恶搞，以支持未来的研究。我们对文本单一、图像单一和后期融合多模态模型进行了基准测试，以及在零样本和微调设置下的少量上下文学习（ICL）和开放及封闭权重的视觉-语言模型（VLM）。我们的结果建立了强有力的基准，并突出了在文化背景下进行阿拉伯仇恨恶搞检测的关键挑战。我们发布了数据集、注释指南和评估脚本，以支持未来的研究。警告：本文包含可能会让读者感到不适的示例。
cs.CL / 9 / 2607.27405
Benchmarking LLM Competence on Logical Inference over Probability Operators
基于概率运算符的逻辑推理能力基准测试
Abstract
Both expressions of uncertainty and inferences are ubiquitous in natural language, and valid inferences over natural-language expressions of uncertainty are necessary for not only everyday conversations but also for high-stakes domains such as medicine and law. While large language models are increasingly evaluated on logical reasoning tasks, disentangling principled, symbolic reasoning from clever surface-level pattern matching is fraught with difficulty. We introduce a benchmark for reasoning over probability operators--inference over sentences with gradable epistemic modals (e.g., probably, might, must) containing 14,320 procedurally-generated English prompts across fifteen inference templates, systematically varying question form, negation strategy, and surface content. Evaluating 29 models, we find that most show answer biases independent of the logical form, a systematic preference for Yes or No. We summarize this with a competence floor: the worse of a model's accuracy on Yes-correct and No-correct items. Only 9 of 29 models exceed random chance. We also test variations in question form, verb phrases/activity, and both the gender and origin of names used in the prompts, finding biases across every axis.
Chinese Translation
不确定性的表达和推理在自然语言中无处不在，对自然语言中不确定性表达的有效推理不仅对日常对话至关重要，也对医学和法律等高风险领域至关重要。尽管大型语言模型在逻辑推理任务上的评估日益增多，但将原则性、符号性的推理与巧妙的表面模式匹配区分开来却充满挑战。我们引入了一个针对概率运算符推理的基准测试——对包含14,320个程序生成的英语提示的句子进行推理，这些句子包含可度量的认识性情态词（例如，probably、might、must），并涵盖十五种推理模板，系统地变化问题形式、否定策略和表面内容。在对29个模型进行评估时，我们发现大多数模型在答案上存在独立于逻辑形式的偏差，表现出对“是”或“否”的系统偏好。我们用一个能力底线来总结这一点：模型在“是”正确项和“否”正确项上的准确性较差。只有29个模型中的9个超过了随机机会。我们还测试了问题形式、动词短语/活动的变化，以及提示中使用的名字的性别和来源，发现每个维度上都有偏差。
cs.CL / 10 / 2607.27421
Selecting Open-Weight Language Models for Zero-Shot Intent Classification: A Systematic Evaluation of 41 Models
选择开放权重语言模型进行零-shot意图分类：41个模型的系统评估
Abstract
Intent classification is a core component of task-oriented dialogue systems, yet practitioners have limited systematic guidance for selecting deployable open-weight language models under compute, latency, and robustness constraints. We present a systematic zero-shot evaluation of 41 open-weight language models spanning 15 families and the 135M--9B parameter range across eight English single-label intent-classification datasets. A ninth dataset, ATIS, uses five labeled demonstrations and is reported as an auxiliary five-shot result. The evaluation includes standard benchmarks, a large-scale voice-assistant corpus, and production-derived e-commerce datasets. Beyond exact-match accuracy, we analyze confidence calibration, robustness to realistic input perturbations, statistical reliability of model rankings, deployment efficiency, and benchmark saturation. Our results show that instruction-tuned 3B models can outperform several evaluated 7B base models, that differences among leading models on MASSIVE are statistically indistinguishable under pairwise McNemar tests, and that widely used benchmarks such as SNIPS have become saturated and no longer meaningfully discriminate among current open-weight models. Instruction tuning's effect on confidence calibration is inconsistent rather than uniformly harmful. These findings provide practical guidance for selecting and evaluating open-weight language models for intent classification.
Chinese Translation
意图分类是面向任务的对话系统的核心组成部分，但实践者在计算、延迟和鲁棒性约束下选择可部署的开放权重语言模型时缺乏系统性的指导。我们对41个开放权重语言模型进行了系统的零-shot评估，这些模型涵盖了15个家族，参数范围从135M到9B，使用了八个英语单标签意图分类数据集。第九个数据集ATIS使用了五个标注示例，并作为辅助的五-shot结果报告。评估包括标准基准、大规模语音助手语料库和生产衍生的电子商务数据集。除了精确匹配准确率外，我们还分析了置信度校准、对现实输入扰动的鲁棒性、模型排名的统计可靠性、部署效率和基准饱和度。我们的结果显示，经过指令调优的3B模型可以超越几个评估的7B基础模型，在MASSIVE上的领先模型之间的差异在成对McNemar检验下统计上不可区分，并且广泛使用的基准如SNIPS已经饱和，不再对当前的开放权重模型进行有意义的区分。指令调优对置信度校准的影响是不一致的，而不是普遍有害。这些发现为选择和评估意图分类的开放权重语言模型提供了实用指导。
cs.CL / 11 / 2607.27497
SkillSmith: Learning to Compose Parametric Skills and Textual Knowledge
SkillSmith：学习组合参数化技能与文本知识
Abstract
Agentic systems driven by large language models (LLMs) regularly feature two key mechanisms to autonomously solve complex problems: synthesizing text-based knowledge and procedures from past experiences and building parametric (weight-space) skill libraries for recurring sub-goals. To date, research has largely treated these as orthogonal pursuits: either organizing textual knowledge through composition and reflection, or consolidating parametric skills via weight-space merging. Consequently, the seamless integration of text and model weights for targeted performance improvements remains largely unexplored. This work bridges this modality gap by treating model weights as an additional modality that an LLM can natively reason over. We instantiate parametric learning via prefix-tuning and augment an LLM to ingest both prefix weights and rich textual data which capture relationships to a target capability. Our augmented LLM, which we call SkillSmith, synthesizes these inputs to perform instruction-steered parametric synthesis, directly outputting new prefix weights that manifest the target skill. We demonstrate that our approach significantly outperforms both text-only and weight-space-only baselines, unlocking performance gains that are out of reach for uni-modal (text-only or weight-only) adaptations.
Chinese Translation
由大型语言模型（LLMs）驱动的自主系统通常具备两个关键机制，以自主解决复杂问题：从过去的经验中综合基于文本的知识和程序，以及为重复的子目标构建参数化（权重空间）技能库。迄今为止，研究在很大程度上将这两者视为正交的追求：要么通过组合和反思来组织文本知识，要么通过权重空间合并来巩固参数化技能。因此，文本与模型权重的无缝集成以实现针对性的性能提升仍然 largely 未被探索。本研究通过将模型权重视为 LLM 可以原生推理的额外模态，填补了这一模态差距。我们通过前缀调优实现参数化学习，并增强 LLM 以同时摄取前缀权重和捕捉与目标能力相关的丰富文本数据。我们增强的 LLM，称为 SkillSmith，综合这些输入以执行指令引导的参数化合成，直接输出体现目标技能的新前缀权重。我们证明了我们的方法显著优于仅使用文本或仅使用权重空间的基线，解锁了单模态（仅文本或仅权重）适应无法达到的性能提升。
cs.CL / 12 / 2607.27506
Models for minimalist RAG: B1ade 335M Embedding and 1B Parameter Small Language Models
极简 RAG 模型：B1ade 335M 嵌入和 1B 参数小型语言模型
Abstract
Language and embedding models used in RAG systems are conventionally assumed to require large-scale pretraining and explicit grounding supervision. We present B1ade, an efficient RAG architecture comprising two purpose-built components: a compact embedding model and a purpose-built SLM. B1ade-embed, a 335M parameter retrieval model constructed via parameter-free fusion of five pretrained encoders achieves top MTEB scores among sub-500M models with zero additional training, and B1ade-1B, an SLM trained on low-cost GPUs using Group Relative Policy Optimization (GRPO) on 723M tokens (2.2M examples) of curated context-question pairs with rewards that optimize only answer similarity. Our central finding is emergent attribution: despite receiving no explicit supervision for source citation, B1ade-1B cites retrieved passages in 42.4% of responses, exceeding the attribution rate of its training distribution by 5.5 percentage points. This demonstrates that grounding behavior can emerge as an accuracy-maximizing strategy under RL training, without explicit reward engineering. On standard QA benchmarks, B1ade-1B achieves 81.82% on PopQA, 65.8% on PubMedQA, and 51.09% on FEVER. In end-to-end RAG evaluation, B1ade-1B achieves an average score of 0.654 across correctness, completeness, coherence, and faithfulness, a 10.8% improvement over the SFT, while closing the gap with models 1.5x its size. These results show that strategic model composition and reward design suffice for resource-efficient RAG, without large-scale pretraining.
Chinese Translation
在 RAG 系统中，语言和嵌入模型通常被认为需要大规模的预训练和明确的基础监督。我们提出了 B1ade，一种高效的 RAG 架构，由两个专门构建的组件组成：一个紧凑的嵌入模型和一个专用的小型语言模型（SLM）。B1ade-embed 是一个 335M 参数的检索模型，通过五个预训练编码器的无参数融合构建，能够在不进行额外训练的情况下，在子 500M 模型中获得最高的 MTEB 分数；而 B1ade-1B 是一个在低成本 GPU 上使用组相对策略优化（Group Relative Policy Optimization, GRPO）训练的 SLM，基于 723M 令牌（2.2M 示例）的策划上下文-问题对，其奖励仅优化答案相似性。我们的核心发现是新兴归因：尽管没有明确的源引用监督，B1ade-1B 在 42.4% 的响应中引用了检索到的段落，超过了其训练分布的归因率 5.5 个百分点。这表明，在强化学习训练下，基础行为可以作为一种最大化准确性的策略自发出现，而无需明确的奖励设计。在标准问答基准测试中，B1ade-1B 在 PopQA 上达到了 81.82%，在 PubMedQA 上达到了 65.8%，在 FEVER 上达到了 51.09%。在端到端 RAG 评估中，B1ade-1B 在正确性、完整性、一致性和忠实度方面的平均得分为 0.654，比 SFT 提升了 10.8%，同时缩小了与其大小 1.5 倍模型的差距。这些结果表明，战略模型组合和奖励设计足以实现资源高效的 RAG，而无需大规模的预训练。
cs.CL / 13 / 2607.27512
Belief Coevolution in a Social Network of Generalist and Specialist Large Language Models
通用型与专业型大型语言模型社交网络中的信念共演
Abstract
Large language models (LLMs) are increasingly deployed in multi-agent environments. However, the processes by which beliefs form and propagate among interacting LLMs remain poorly understood. We introduce CoevolveSim, a framework for studying belief diffusion within networked LLM populations. CoevolveSim allows us to isolate and study three factors: domain specialization, social-role assignment, and social network structure. Within this framework, generalist and specialist LLM agents exchange and revise beliefs. In each round, an LLM agent observes a summary of its neighbors' beliefs before updating its own. We run 1,280 controlled simulations spanning four scenarios, two network structures, and 20 medical-indication statements. We find that persona-style role assignment and network structure reshape individual belief revision but have minimal effect on population-level consensus. In contrast, introducing (finetuned) specialist LLMs more than doubles the shift in consensus and gives rise to consistent asymmetries in exerted influence. We further show that simple persistence-based opinion-dynamics models reproduce collective outcomes in all-generalist LLM populations, whereas heterogeneous LLM populations require population-level belief composition to reproduce consensus and agent identity to predict individual belief transitions. Our results indicate that realistic simulation of belief diffusion in multi-agent LLM systems requires a diverse set of underlying LLMs, not persona prompting alone.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于多智能体环境中。然而，交互的LLMs之间信念形成和传播的过程仍然不够清晰。我们引入了CoevolveSim，一个用于研究网络化LLM群体中信念扩散的框架。CoevolveSim使我们能够隔离并研究三个因素：领域专业化、社会角色分配和社交网络结构。在这个框架内，通用型和专业型LLM代理相互交换和修正信念。在每一轮中，一个LLM代理在更新自己的信念之前观察其邻居信念的摘要。我们进行了1280次控制模拟，涵盖四种情境、两种网络结构和20个医学指示声明。我们发现，个性化角色分配和网络结构重塑了个体信念修正，但对群体层面的共识影响甚微。相比之下，引入（微调的）专业型LLMs使共识的变化超过两倍，并产生了一致的影响力不对称。我们进一步表明，基于简单持久性的意见动态模型能够在全通用型LLM群体中再现集体结果，而异质LLM群体则需要群体层面的信念组成来再现共识，并需要代理身份来预测个体信念转变。我们的结果表明，在多智能体LLM系统中，信念扩散的现实模拟需要多样化的基础LLM，而不仅仅依赖于个性化提示。
cs.CL / 14 / 2607.27557
Training Skills Like Parameters via Self-Supervised Semantic Diffusion
通过自监督语义扩散训练技能如同参数
Abstract
While Large Language Models (LLMs) demonstrate remarkable general instruction-following capabilities, they often fall short of human experts in highly specialized, open-ended domains such as creative screenwriting. Prior approaches typically adopt post-training, yet both supervised fine-tuning and reinforcement learning require weight access that closed-source frontier models do not offer, and demand heavy compute. Moreover, what is learned is tied to a single checkpoint and cannot be inspected by humans. Recent advancements in agentic continual learning instead attempt to bridge this gap by accumulating external textual skills. However, these methods heavily rely on costly human expert annotations or unreliable LLM-as-a-judge feedback for reflection. To overcome this bottleneck, we propose a novel, unsupervised self-evolving agent framework inspired by the corruption-and-reconstruction paradigm of diffusion models. Instead of relying on explicit external scoring, we leverage existing high-quality human artifacts to construct self-supervised signals. Training then follows the familiar loop of neural network training, forward, loss, and backward, with the loss coming from contrasting the agent's reconstruction against the human original. What is updated is not model weights but an external library of textual skills. We evaluate our framework on the challenging task of short drama screenwriting. Experimental results demonstrate that our method enables the agent to autonomously extract and internalize highly generalizable skills, significantly enhancing its domain-specific generation capabilities. Furthermore, this self-contrastive reflection paradigm offers a scalable pathway for agents to teach themselves the production of complex, high-quality human artifacts, without requiring external supervision.
Chinese Translation
尽管大型语言模型（LLMs）展现了卓越的一般指令遵循能力，但在创意剧本写作等高度专业化的开放领域，它们往往不及人类专家。以往的方法通常采用后训练的方式，但无论是监督微调还是强化学习都需要对权重的访问，而闭源的前沿模型并不提供这种访问，并且需要大量计算资源。此外，所学内容与单一检查点相关，无法被人类检查。最近在主动持续学习方面的进展试图通过积累外部文本技能来弥补这一差距。然而，这些方法在很大程度上依赖于昂贵的人类专家注释或不可靠的LLM作为评判者的反馈进行反思。为了解决这一瓶颈，我们提出了一种新颖的无监督自我进化代理框架，灵感来自扩散模型的腐败与重建范式。我们不依赖于显式的外部评分，而是利用现有的高质量人类文献来构建自监督信号。训练过程遵循神经网络训练的熟悉循环：前向传播、损失计算和反向传播，损失来自于对比代理的重建与人类原作。更新的不是模型权重，而是一个外部文本技能库。我们在短剧本写作这一具有挑战性的任务上评估了我们的框架。实验结果表明，我们的方法使代理能够自主提取和内化高度可泛化的技能，显著增强其领域特定的生成能力。此外，这种自对比反思范式为代理提供了一条可扩展的路径，使其能够自我教授复杂、高质量人类文献的创作，而无需外部监督。
cs.CL / 15 / 2607.27595
Beyond Similarity: Grounded Agentic Extraction and Expert-Adjudicated Evaluation of Intertextuality in Classical Chinese Histories
超越相似性：基于实证的能动提取与专家裁定的古典汉籍互文性评估
Abstract
Computational approaches to intertextuality have advanced from string matching to neural retrieval, yet their outputs, similarity scores and parallel-passage lists, identify where texts reuse one another without characterizing how or why. We recast fine-grained intertextuality extraction as an agentic task in which a large language model (LLM) reads two text units in full and, through a constrained tool interface, must ground each proposed reuse in exact character spans on both sides and label it under a five-dimension typology of reuse (form, aspect, source-marking, function, stance). We validate the approach on an exhaustive comparison of the Analects with the Book of Han, where three domain experts adjudicate a pooled multi-model candidate set into a benchmark of 2,533 intertextual pairs. Against this standard we study twelve LLMs, reporting precision (56%-93%), a 51$\times$ cost spread at comparable quality, and how well their confidence is calibrated. Expert agreement traces a reliability gradient: dimensions legible on the textual surface are annotated consistently, while those requiring inference of intent are contested, delimiting the claims such annotation supports. Scaling the validated extractor to the full Twenty-Four Histories (65,380 comparisons, 5,766 pairs) recovers corpus-level structure a similarity score cannot express. The interpretive composition of citation shows no systematic change across eighteen centuries, yet the same passage is quoted less and less literally. Stability in the aggregate with drift in the individual case is what a cultural-attraction account expects. We release the extraction protocol and the expert-adjudicated benchmark.
Chinese Translation
计算方法在互文性研究中的进展已从字符串匹配发展到神经检索，然而其输出的相似性评分和平行段落列表仅识别文本之间的重用，而未能表征其如何或为何重用。我们将细粒度的互文性提取重新构建为一个能动任务，其中大型语言模型（LLM）完整阅读两个文本单元，并通过受限的工具接口，必须将每个提议的重用基于两侧的确切字符范围进行定位，并根据五维重用类型（形式、方面、来源标记、功能、立场）进行标注。我们在《论语》和《汉书》的全面比较中验证了该方法，三位领域专家将汇总的多模型候选集裁定为2533对互文性基准。在这一标准下，我们研究了十二个LLM，报告了精确度（56%-93%）、在相似质量下的51倍成本差异，以及它们的置信度校准情况。专家一致性描绘了一个可靠性梯度：文本表面可读的维度被一致标注，而需要推断意图的维度则存在争议，限制了该标注所支持的主张。将经过验证的提取器扩展到完整的《二十四史》（65,380次比较，5,766对）恢复了相似性评分无法表达的语料库级结构。引用的解释性构成在十八个世纪中没有系统性变化，然而同一段落的引用越来越不字面。总体稳定而个别案例漂移是文化吸引力理论所预期的结果。我们发布了提取协议和专家裁定的基准。
cs.CL / 16 / 2607.27611
AWARE-FX: An Auditable Knowledge-Guided AI System for Measuring Corporate Foreign-Exchange Hedging Disclosure
AWARE-FX：一种可审计的知识引导人工智能系统，用于测量企业外汇对冲披露
Abstract
Corporate annual reports contain weakly structured evidence about foreign-exchange risk management, derivative use, natural hedging, and explicit non-use. This study develops AWARE-FX, an auditable AI/NLP decision-support system that converts report text into traceable firm-year hedging-disclosure measures. The system combines a professional-source lexicon, negation and accounting-status logic, channel-specific financial encoders, exact evidence gates, conservative aggregation, and an audit ledger. Across 24,909 Hong Kong firm-years from 2008-2025, it retrieves and scores 543,527 snippets. Reliability is evaluated through ablations, a stratified 300-snippet human audit, three-seed FinBERT-ModernBERT comparisons, strict 2023-2025 temporal tests, probability calibration, selective prediction, and fixed-prompt generative-model benchmarks. FinBERT has the higher mean F1 in seven of eight encoder task-split comparisons; its temporal F1 ranges from 0.702 to 0.872. Abstaining on the 20% least-confident temporal observations raises retained-sample F1 by 0.050-0.077. Deterministic Qwen3-8B performs strongly on commodity and negation evidence but poorly on foreign-debt and accounting-context labels, showing that a general-purpose LLM does not uniformly replace domain constraints. The strict FX score is negatively associated with linked baseline and stress-period FX exposure, whereas the generic broad score is not. These associations provide external construct validation, not causal estimates of hedging effectiveness. AWARE-FX contributes a tested decision-support architecture in which retrieval, status logic, classification, uncertainty handling, aggregation, and external validation remain separately auditable.
Chinese Translation
企业年度报告中包含关于外汇风险管理、衍生品使用、自然对冲和明确不使用的弱结构证据。本研究开发了AWARE-FX，这是一种可审计的人工智能/自然语言处理决策支持系统，能够将报告文本转换为可追溯的公司年度对冲披露指标。该系统结合了专业来源词典、否定和会计状态逻辑、特定渠道的财务编码器、确凿证据门、保守聚合和审计账本。在2008年至2025年间的24,909个香港公司年度中，它检索并评分了543,527个片段。通过消融实验、分层的300个片段人工审计、三种种子FinBERT-ModernBERT比较、严格的2023-2025时间测试、概率校准、选择性预测和固定提示生成模型基准来评估可靠性。在八个编码器任务拆分比较中，FinBERT在七个中具有更高的平均F1值，其时间F1范围从0.702到0.872。在20%最不自信的时间观察中弃权，保留样本F1提高了0.050-0.077。确定性Qwen3-8B在商品和否定证据上表现强劲，但在外债和会计上下文标签上表现不佳，表明通用大型语言模型并不能均匀替代领域约束。严格的外汇评分与相关基线和压力期外汇敞口呈负相关，而通用的广泛评分则没有。这些关联提供了外部构念验证，而非对对冲有效性的因果估计。AWARE-FX贡献了一种经过测试的决策支持架构，其中检索、状态逻辑、分类、不确定性处理、聚合和外部验证保持单独可审计。
cs.CL / 17 / 2607.27614
DualAnchor: Preserving Language Priors and Improving Lexical Fidelity in Gloss-Free Sign Language Translation
DualAnchor：在无注释手语翻译中保持语言先验并提高词汇忠实度
Abstract
Recent advances in large language models (LLMs) have led sign language translation (SLT), the task of converting sign-language videos into spoken-language text, to increasingly adopt LLMs as textual backbones. However, despite their strong language modeling capabilities, existing LLM-based SLT methods often undermine rather than exploit this language prior, producing disfluent translations, a failure we term language-prior degradation. Meanwhile, existing methods typically align videos and text at the sentence level, which does not ensure accurate lexical details and creates a lexical fidelity gap. To address both issues, we propose DualAnchor, a gloss-free LLM-based SLT training framework that couples two complementary anchors for linguistically fluent and visually faithful generation. Token-level Prior Anchoring (TPA) preserves the LLM's language prior by regularizing the multimodal decoder at each decoding step toward the next-token distribution of a frozen LLM conditioned on the same autoregressive prefix. Optimal Transport Alignment (OTA) improves lexical fidelity by formulating visual-textual matching as entropy-regularized partial optimal transport, with Sinkhorn optimization inducing a soft alignment between visual tokens and textual content tokens under a cosine cost. DualAnchor achieves strong overall performance on both PHOENIX-2014T and CSL-Daily. Targeted analyses attribute these gains to the complementary effects of the two anchors: TPA improves fluency, whereas OTA reduces fine-grained lexical errors.
Chinese Translation
近年来，大型语言模型（LLMs）的进展使得手语翻译（SLT）这一将手语视频转换为口语文本的任务越来越多地采用LLMs作为文本基础。然而，尽管现有的LLM在语言建模方面具有强大的能力，基于LLM的SLT方法往往削弱而非利用这一语言先验，导致翻译不流畅，这种失败我们称之为语言先验退化。同时，现有方法通常在句子级别对视频和文本进行对齐，这并不能确保准确的词汇细节，并造成词汇忠实度的差距。为了解决这两个问题，我们提出了DualAnchor，这是一种基于LLM的无注释SLT训练框架，结合了两个互补的锚点，以实现语言流畅和视觉忠实的生成。词元级先验锚定（TPA）通过在每个解码步骤中将多模态解码器正则化到基于相同自回归前缀的冻结LLM的下一个词元分布，从而保持LLM的语言先验。最优传输对齐（OTA）通过将视觉-文本匹配表述为熵正则化的部分最优传输来提高词汇忠实度，Sinkhorn优化在余弦成本下引入视觉词元和文本内容词元之间的软对齐。DualAnchor在PHOENIX-2014T和CSL-Daily上实现了强大的整体性能。针对性的分析将这些提升归因于两个锚点的互补效应：TPA提高了流畅性，而OTA减少了细粒度的词汇错误。
cs.CL / 18 / 2607.27652
Harness-G: A Graph-Structured Harness for Search Agents
Harness-G：一种用于搜索代理的图结构化工具
Abstract
Reinforcement learning (RL) search agents commonly model retrieval as free-form natural-language query generation and optimize multi-turn interactions using final-answer rewards. Current studies mainly improve training with denser or more structured credit signals, but rarely examine whether retrieval is properly formulated at the policy-environment interface. We observe pronounced retrieval aliasing during Search-R1 training: rollouts for the same question continue to generate distinct query strings, yet their accumulated evidence sets increasingly overlap. We call this phenomenon retrieval-equivalence collapse; in this regime, trajectories approach utility equivalence with respect to retrieval decisions, leaving within-group returns with little effective retrieval contrast. To address this problem, we propose Harness-G, a graph-structured retrieval framework that redesigns this interface. It reformulates free-form query generation as finite action selection: the policy selects an evidence sentence or entity, or chooses to answer, while the environment constructs the menu, tracks retrieval state, and validates and executes each choice. This interface reduces linguistic aliasing and makes same-state alternatives directly comparable. Building on this interface, we introduce Structured Non-myopic Credit (SNC), which uses a frozen answer scorer to compare the selected action with its alternatives and assigns downstream gains to the earlier actions that enabled them. Across six QA benchmarks, Harness-G achieves the highest average F1 at both evaluated model scales, outperforming the strongest baseline, Graph-R1, by 10.74 points at 1.5B and 3.98 points at 3B.
Chinese Translation
强化学习（RL）搜索代理通常将检索建模为自由形式的自然语言查询生成，并通过最终答案奖励优化多轮交互。目前的研究主要通过更密集或更结构化的信用信号来改善训练，但很少检查检索在策略-环境接口处是否得到了正确的表述。我们在 Search-R1 训练中观察到明显的检索混淆：对于相同问题的回滚生成了不同的查询字符串，但它们的累积证据集却越来越重叠。我们将这种现象称为检索等价崩溃；在这种情况下，轨迹在检索决策方面接近效用等价，导致组内回报的有效检索对比很小。为了解决这个问题，我们提出了 Harness-G，一个图结构化的检索框架，重新设计了这一接口。它将自由形式的查询生成重新表述为有限的动作选择：策略选择一个证据句子或实体，或选择回答，而环境构建菜单，跟踪检索状态，并验证和执行每个选择。该接口减少了语言混淆，使同状态的替代方案可以直接比较。在此接口的基础上，我们引入了结构化非短视信用（SNC），它使用冻结的答案评分器将所选动作与其替代方案进行比较，并将下游收益分配给使其成为可能的早期动作。在六个问答基准测试中，Harness-G 在两个评估模型规模上都达到了最高的平均 F1 分数，分别比最强基线 Graph-R1 高出 10.74 分（在 1.5B）和 3.98 分（在 3B）。
cs.CL / 19 / 2607.27654
From Single- to Cross-Document: Benchmarking Multi-Granularity Event Analysis of Large Language Models
从单文档到跨文档：大型语言模型多粒度事件分析的基准测试
Abstract
Event analysis is an essential and fundamental direction of information extraction, involving various event-centric tasks at different granularity of documents. While large language models (LLMs) have preliminarily achieved promising performance in part of these tasks individually, their capability in event analysis still lacks comprehensive understanding due to restricted document granularity, task designs, and data source of existing benchmarks. To address these limitations, we introduce MiGUE-Bench, a systematic benchmark for assessing the performance of LLMs in multi-granularity event analysis. To support large-scale evaluation, we first develop an LLM-driven self-correcting annotation framework called MiGUE-Pipeline, enabling scalable acquisition of high-quality source data of events with automatic labels. Then, we design four core tasks in our benchmark, i.e., event detection, relation reasoning, structure induction, and future prediction, to probe model competence at different levels, from atomic event details to complex cross-document narratives. Extensive experiments on state-of-the-art LLMs and retrieval-augmented generation (RAG) methods delineate the current capability boundary and identify critical deficiencies, providing insights into the future improvement of LLMs in challenging event analysis tasks.
Chinese Translation
事件分析是信息提取的一个重要且基础的方向，涉及不同文档粒度下的各种以事件为中心的任务。尽管大型语言模型（LLMs）在这些任务中的部分内容上初步取得了令人鼓舞的表现，但由于现有基准的文档粒度、任务设计和数据来源的限制，其在事件分析中的能力仍然缺乏全面的理解。为了解决这些局限性，我们引入了MiGUE-Bench，这是一个系统性的基准，用于评估LLMs在多粒度事件分析中的表现。为了支持大规模评估，我们首先开发了一个名为MiGUE-Pipeline的LLM驱动自我修正注释框架，使得能够以自动标签的方式可扩展地获取高质量的事件源数据。然后，我们在基准中设计了四个核心任务，即事件检测、关系推理、结构归纳和未来预测，以探测模型在不同层面的能力，从原子事件细节到复杂的跨文档叙述。针对最先进的LLMs和检索增强生成（RAG）方法的广泛实验描绘了当前能力的边界，并识别出关键缺陷，为LLMs在具有挑战性的事件分析任务中的未来改进提供了见解。
cs.CL / 20 / 2607.27671
ICLE++: Modeling Fine-Grained Traits for Holistic Essay Scoring
ICLE++：用于整体作文评分的细粒度特征建模
Abstract
The majority of the recently-developed models for automated essay scoring (AES) are evaluated solely on the ASAP corpus. However, ASAP is not without its limitations. For instance, it is not clear whether models trained on ASAP can generalize well when evaluated on other corpora. In light of these limitations, we introduce ICLE++, a corpus of persuasive student essays annotated with both holistic scores and trait-specific scores. Not only can ICLE++ be used to test the generalizability of AES models trained on ASAP, but it can also facilitate the evaluation of models developed for newer AES problems such as multi-trait scoring and cross-prompt scoring. We believe that ICLE++, which represents a culmination of our long-term effort in annotating the essays in the ICLE corpus, contributes to the set of much-needed annotated corpora for AES research.
Chinese Translation
最近开发的自动化作文评分（AES）模型大多数仅在ASAP语料库上进行评估。然而，ASAP并非没有其局限性。例如，尚不清楚在ASAP上训练的模型在其他语料库上评估时是否能够很好地泛化。鉴于这些局限性，我们引入了ICLE++，一个包含说服性学生作文的语料库，附有整体评分和特征特定评分的注释。ICLE++不仅可以用于测试在ASAP上训练的AES模型的泛化能力，还可以促进对为新的AES问题（如多特征评分和跨提示评分）开发的模型的评估。我们相信，ICLE++代表了我们在注释ICLE语料库中的作文方面的长期努力的成果，为AES研究提供了急需的注释语料库。
cs.CL / 21 / 2607.27692
Recall Before You Rank: Similarity-Guided Top-$K$ Reuse for Efficient Long-Context Attention
回忆再排序：基于相似性的 Top-$K$ 重用以实现高效的长上下文注意力
Abstract
Top-$K$ sparse attention reduces the cost of Softmax and value aggregation by attending to only a small subset of key--value (KV) entries. However, identifying this subset still requires scoring the current query against the full KV cache and performing global Top-$K$ selection, leaving selector cost linear in context length and limiting the practical efficiency of sparse attention for long-context decoding. In this paper, we introduce ReTopK, a training-free method that accelerates dynamic Top-$K$ attention by reusing historical retrieval decisions. ReTopK builds on the observation that similar queries often attend to overlapping supports and that partially overlapping supports can still preserve most of the Exact Top-$K$ attention mass. For each attention head, it maintains a bounded cache of historical query--support pairs, retrieves the most similar cached queries for each new query, unions their stored supports with a recent window, and reranks only the resulting compact candidate set using exact current-query scores. A similarity-based fallback invokes full-history Exact Top-$K$ when reuse is unreliable, while periodic exact refreshes limit cache drift. ReTopK retains the complete KV cache and reuses only selected indices, rather than historical scores, attention weights, or outputs. Across 16K--128K contexts, ReTopK achieves the lowest PG19 perplexity and the highest NIAH and LongBench scores among the evaluated approximate methods. At 128K with $K=512$, ReTopK incurs only a 0.50\% perplexity increase over Exact Top-$K$ while accelerating attention computation by $3.07\times$.
Chinese Translation
Top-$K$ 稀疏注意力通过仅关注小部分键值（KV）条目来降低 Softmax 和值聚合的成本。然而，识别这一子集仍然需要将当前查询与完整的 KV 缓存进行评分，并执行全局 Top-$K$ 选择，这使得选择器的成本与上下文长度成线性关系，从而限制了稀疏注意力在长上下文解码中的实际效率。本文介绍了一种无训练方法 ReTopK，该方法通过重用历史检索决策来加速动态 Top-$K$ 注意力。ReTopK 基于这样一个观察：相似的查询往往关注重叠的支持，而部分重叠的支持仍然可以保留大部分的精确 Top-$K$ 注意力质量。对于每个注意力头，它维护一个有限的历史查询-支持对缓存，检索与每个新查询最相似的缓存查询，将它们存储的支持与最近的窗口进行联合，并仅使用当前查询的精确评分对结果紧凑的候选集进行重新排序。当重用不可靠时，基于相似性的后备机制会调用完整历史的精确 Top-$K$，而定期的精确刷新则限制了缓存漂移。ReTopK 保留完整的 KV 缓存，仅重用选定的索引，而不是历史评分、注意力权重或输出。在 16K 到 128K 的上下文中，ReTopK 在评估的近似方法中实现了最低的 PG19 困惑度和最高的 NIAH 及 LongBench 分数。在 128K 上，$K=512$ 时，ReTopK 仅比精确 Top-$K$ 增加了 0.50\% 的困惑度，同时加速了注意力计算达 $3.07 imes$。
cs.CL / 22 / 2607.27735
A Sparse Glimpse of the Whole: Train-Free Self-Speculative Decoding
整体的稀疏视角：无训练自我推测解码
Abstract
Speculative decoding alleviates the memory-bandwidth bottleneck in large language model inference, but its acceleration is jointly constrained by drafting overhead, token acceptance, and speculation length. We present a unified efficiency analysis showing that extending the speculation horizon can reduce rather than improve speedup when the marginal acceptance probability falls below the relative drafting cost. Guided by this analysis, we introduce SparseSpec-L, a training-free self-speculative decoding framework for long-context inference. SparseSpec-L generates lightweight drafts directly from the target model using a dynamically sparsified and recallable KV cache. It recycles per-head attention statistics produced during full-context verification as a no-extra-forward importance signal, allowing critical historical tokens to be recalled without permanently discarding the dense KV cache. An online entropy-based controller further selects the speculation length according to expected step-wise efficiency. Experiments across multiple long-context tasks and model scales show consistent end-to-end acceleration, with up to speedup over autoregressive decoding while preserving the target model's output distribution.
Chinese Translation
推测解码缓解了大型语言模型推理中的内存带宽瓶颈，但其加速受到草拟开销、令牌接受率和推测长度的共同限制。我们提出了一种统一的效率分析，表明当边际接受概率低于相对草拟成本时，延长推测视野可能会降低而不是提高加速效果。在这一分析的指导下，我们引入了SparseSpec-L，这是一种用于长上下文推理的无训练自我推测解码框架。SparseSpec-L直接从目标模型生成轻量级草稿，使用动态稀疏化和可回忆的KV缓存。它回收在全上下文验证过程中产生的每头注意力统计数据，作为无额外前向的重要性信号，从而允许关键历史令牌被回忆，而不必永久丢弃密集的KV缓存。一个基于在线熵的控制器进一步根据预期的逐步效率选择推测长度。跨多个长上下文任务和模型规模的实验表明，整体端到端加速效果一致，相比自回归解码可实现高达加速，同时保持目标模型的输出分布。
cs.CL / 23 / 2607.27747
Can LVLMs Uncover the Truth Behind Visual Illusions? An Analysis of Perceptual and Reasoning Capabilities
大型视觉语言模型能揭示视觉错觉背后的真相吗？感知与推理能力的分析
Abstract
Large Vision Language Models have integrated reasoning capabilities, elevating cognitive performance to new levels. However, existing evaluations either focus solely on perception or rely on specific domains such as maths or coding. Evaluation for reasoning capabilities that align with an open-world environment is still required, especially one that considers perception and reasoning jointly. To bridge this gap, we propose to evaluate LVLMs by exploiting visual illusions as a diagnostic tool. Visual illusions are phenomena in which the human visual system misinterprets objective signals, resulting in an understanding that deviates from reality. We constructed IllusionReasoning, a benchmark of illusion images collected from the real world, incorporating diverse annotated question-answer pairs. Based on IllusionReasoning, we show that the reasoning capabilities of a wide range of LVLMs are not as advanced as claimed. Our work provides new insights into LVLMs and offers future direction for optimisation.
Chinese Translation
大型视觉语言模型（LVLMs）具备了推理能力，使认知表现达到了新的高度。然而，现有的评估要么仅关注感知，要么依赖于特定领域，如数学或编程。仍然需要一种与开放世界环境相一致的推理能力评估，尤其是考虑感知与推理的联合评估。为填补这一空白，我们提出通过利用视觉错觉作为诊断工具来评估LVLMs。视觉错觉是人类视觉系统误解客观信号的现象，导致对现实的理解偏离。我们构建了IllusionReasoning，这是一个从现实世界收集的错觉图像基准，包含多样的标注问答对。基于IllusionReasoning，我们展示了多种LVLMs的推理能力并没有如声称的那样先进。我们的工作为LVLMs提供了新的见解，并为未来的优化方向提供了建议。
cs.CL / 24 / 2607.27766
Gradient-free Task-Conditioned Retrieval for On-Device In-Context Learning
无梯度任务条件检索用于设备端上下文学习
Abstract
On-device in-context learning (ICL) relies on pre-inference retrieval to select demonstrations for useful context before downstream model inference. This retrieval must exploit task-specific information while operating over local memories under limited computation, memory, and data-exposure budgets. We propose Conditional Retrieval Alignment (CoRA), a gradient-free framework that converts a frozen encoder into a task-conditioned retriever using paired candidate inputs and outputs. CoRA selects complementary encoder layers, constructs an output-derived conditioning space from candidate memory, and aligns candidate input representations to this space through closed-form ridge regression. Low-rank factorization then produces a compact retrieval basis where candidate outputs are used only during offline index construction, whereas query-time retrieval requires only the query input and precomputed index. We show that CoRA's rank-constrained basis is the optimal low-rank compression of the output-conditioned fitted representation, and derive an exact two-pass streaming construction that avoids materializing the full fitted matrix. We further extend the framework to multimodal exemplar retrieval by incorporating visual representations into the conditioning and retrieval spaces. Experiments across ten textual datasets and four multimodal benchmarks with Llama-3.2-1B, MobileLLM-Pro, OpenFlamingo-3B, and Qwen3.5-2B, as well as end-to-end Raspberry Pi~5 deployment demonstrate that CoRA supports effective task-conditioned retrieval without retriever fine-tuning, backpropagation, or target-model calls.
Chinese Translation
设备端上下文学习（ICL）依赖于预推理检索，以在下游模型推理之前选择用于有用上下文的示例。这种检索必须在有限的计算、内存和数据曝光预算下，利用特定任务的信息，同时在本地内存中操作。我们提出了条件检索对齐（Conditional Retrieval Alignment, CoRA），这是一种无梯度框架，通过配对的候选输入和输出，将一个冻结的编码器转换为任务条件检索器。CoRA选择互补的编码器层，从候选内存中构建输出导出的条件空间，并通过封闭形式的岭回归将候选输入表示对齐到该空间。低秩分解随后生成一个紧凑的检索基础，其中候选输出仅在离线索引构建期间使用，而查询时的检索仅需要查询输入和预计算的索引。我们展示了CoRA的秩约束基础是输出条件拟合表示的最优低秩压缩，并推导出一种精确的双通道流式构建方法，避免了物化完整的拟合矩阵。我们进一步通过将视觉表示纳入条件和检索空间，扩展了该框架以支持多模态示例检索。在十个文本数据集和四个多模态基准上进行的实验，使用Llama-3.2-1B、MobileLLM-Pro、OpenFlamingo-3B和Qwen3.5-2B，以及端到端的Raspberry Pi~5部署，证明了CoRA支持有效的任务条件检索，而无需检索器微调、反向传播或目标模型调用。
cs.CL / 25 / 2607.27773
ChronoMem: Version Control and Semantic Rollback for Large Language Model Agent Memory
ChronoMem：大型语言模型智能体记忆的版本控制与语义回滚
Abstract
LLM agents increasingly rely on long-term memory to support multi-session interaction and personalization. However, existing agent memory systems are designed around forward-only evolution, continuously accumulating, consolidating, and overwriting knowledge, with no principled mechanism to inspect, version, or revert prior states. This makes agents brittle under corrections, concept drift, and memory corruption, particularly after they have already been exposed to subsequent information. We present ChronoMem, a semantic version-control layer for agentic memory integrated into the production-ready, open-source Agent Development Kit by Google. ChronoMem commits whole-memory snapshots at each memory write, maintains structured version histories, and supports natural-language rollback requests by mapping undo intents to concrete historical versions through hybrid lexical and semantic retrieval, rank fusion, and reranking. We further introduce a post-exposure evaluation protocol that tests whether an agent can behave counterfactually after rollback by answering queries and summarizing history as if future updates had never occurred. On long-horizon conversational benchmarks augmented with evolving memory states and rollback tasks, ChronoMem substantially improves rollback-consistent question answering and history summarization relative to prompt-only and retrieval-only baselines, while achieving strong performance in semantic version selection. To our knowledge, ChronoMem is the first open-source system and benchmark for systematic semantic global memory rollback in LLM agents.
Chinese Translation
大型语言模型（LLM）智能体越来越依赖长期记忆来支持多会话交互和个性化。然而，现有的智能体记忆系统设计为仅向前演变，持续积累、整合和覆盖知识，缺乏检查、版本控制或恢复先前状态的原则性机制。这使得智能体在纠正、概念漂移和记忆损坏时变得脆弱，尤其是在它们已经接触到后续信息之后。我们提出了ChronoMem，一个集成于谷歌生产就绪的开源智能体开发工具包中的智能体记忆的语义版本控制层。ChronoMem在每次记忆写入时提交整个记忆快照，维护结构化版本历史，并通过混合词汇和语义检索、排名融合和重新排名支持自然语言回滚请求，将撤销意图映射到具体的历史版本。我们进一步引入了一种后曝光评估协议，测试智能体在回滚后是否能够反事实地表现，通过回答查询和总结历史，仿佛未来的更新从未发生。在增强了不断演变的记忆状态和回滚任务的长期对话基准上，ChronoMem显著改善了相较于仅基于提示和仅基于检索的基线的回滚一致性问答和历史摘要，同时在语义版本选择方面也表现出色。据我们所知，ChronoMem是第一个针对LLM智能体系统化语义全局记忆回滚的开源系统和基准。
cs.CL / 26 / 2607.27783
Reasoning Consensus: Structural Ensembling of LLM Reasoning via Weighted DAG Aggregation
推理共识：通过加权有向无环图聚合实现大规模语言模型推理的结构集成
Abstract
Large Language Models (LLMs) explore problems through chain-of-thought, but this exploration is buried in unstructured prose. On high-stakes tasks, users cannot tell which steps are well-supported, which alternatives were seriously considered, or how the final conclusion compares to those the model discarded. We propose a framework that ensembles the reasoning structure, not just the answers, of multiple LLMs by weighted merging of Directed Acyclic Graphs (DAGs) extracted from reasoning chains. We weight each step by how many traces independently attest to it, to return "Consensus Reasoning". Across six benchmarks spanning statutory interpretation, graduate-level science, narrative multi-hop reasoning, and first-order logic, our ensemble outperforms a matched-budget majority-vote baseline, with a maximum accuracy gain of 3.1% on MuSR-MM (narrative multi-hop reasoning). On a single model, the framework matches or exceeds self-consistency at the same trace budget while additionally exposing an inspectable consensus reasoning graph. Ensemble weights correlate with LLM-judge rankings of reasoning quality at Spearman $\rho = 0.30$-$0.51$, and consensus subgraphs are preferred over alternatives leading to the majority-vote answer in 54.4-65.4% of head-to-head comparisons across five of six datasets. We observe that our framework can also be used to analyze diverse reasoning perspectives for a problem.
Chinese Translation
大型语言模型（LLMs）通过思维链探索问题，但这种探索往往埋藏在非结构化的散文中。在高风险任务中，用户无法判断哪些步骤得到了良好的支持，哪些替代方案被认真考虑，或者最终结论与模型所舍弃的结论相比如何。我们提出了一个框架，通过加权合并从推理链中提取的有向无环图（DAGs），对多个LLM的推理结构进行集成，而不仅仅是答案。我们通过独立证明每一步的轨迹数量来加权每个步骤，以返回“共识推理”。在涵盖法定解释、研究生级科学、叙事多跳推理和一阶逻辑的六个基准测试中，我们的集成方法在匹配预算的多数投票基线之上表现优异，在MuSR-MM（叙事多跳推理）上最大准确率提升达3.1%。在单个模型上，该框架在相同的轨迹预算下匹配或超越自一致性，同时额外展示了可检查的共识推理图。集成权重与LLM评判推理质量的排名在Spearman $
ho = 0.30$-$0.51$之间相关，而在六个数据集中的五个头对头比较中，共识子图在54.4%-65.4%的情况下优于导致多数投票答案的替代方案。我们观察到，该框架还可以用于分析问题的多样化推理视角。
cs.CL / 27 / 2607.27790
Semantic-Aligned Structural Abstraction for Multimodal Sentiment Analysis
用于多模态情感分析的语义对齐结构抽象
Abstract
Multimodal Sentiment Analysis (MSA) aims to interpret complex human emotions by integrating natural language with non-verbal modalities. Non-verbal modalities share a structural isomorphism with natural language, as both can be viewed as feature sequences evolving over time. This isomorphism enables the transformation of non-verbal modalities into text-like tokens for unified semantic reasoning. Large Language Models (LLMs), designed to understand and generate sequential data, can thus be utilized to interpret complex affective sequences. However, existing LLM-based methods primarily capture low-level superficial features, failing to model affective semantics arising from structural variations and contextual interactions. To address this limitation, we propose \textbf{SentiLLM}, a unified framework that leverages \textit{Semantic-Aligned Structural Abstraction} to distill continuous raw signals into compact, semantically meaningful tokens. Specifically, we introduce a \textit{Dual-Stream Salience-Context Calibration Mechanism}, which disentangles non-verbal feature sequences into a focus stream and an ambient stream. The focus stream captures salient sentiment shifts (e.g., facial expressions) guided by textual priors, while the ambient stream characterizes stable background states. Through calibrating these dynamic sentiment shifts against background states, SentiLLM effectively projects non-verbal modalities into a unified semantic space, making them naturally understandable for LLMs. Serving as a plug-and-play module, SentiLLM significantly improves discriminative performance with only a small number of trainable parameters. Our method achieves superior performance on four datasets, MOSI, MOSEI, CH-SIMS, and CH-SIMS v2, demonstrating the effectiveness of the structural abstraction paradigm in MSA. Our code is available at: \href{https://github.com/especiallyW/SentiLLM}.
Chinese Translation
多模态情感分析（MSA）旨在通过将自然语言与非语言模态相结合来解读复杂的人类情感。非语言模态与自然语言之间存在结构同构性，因为两者都可以视为随时间演变的特征序列。这种同构性使得非语言模态能够转化为类似文本的标记，从而实现统一的语义推理。大型语言模型（LLMs）旨在理解和生成序列数据，因此可以用于解读复杂的情感序列。然而，现有的基于LLM的方法主要捕捉低层次的表面特征，未能建模由结构变化和上下文交互引起的情感语义。为了解决这一局限性，我们提出了 extbf{SentiLLM}，一个统一框架，利用 extit{语义对齐结构抽象}将连续的原始信号提炼为紧凑且具有语义意义的标记。具体而言，我们引入了一种 extit{双流显著性-上下文校准机制}，将非语言特征序列解耦为聚焦流和环境流。聚焦流捕捉由文本先验引导的显著情感变化（例如，面部表情），而环境流则表征稳定的背景状态。通过将这些动态情感变化与背景状态进行校准，SentiLLM有效地将非语言模态投射到统一的语义空间，使其对LLMs自然可理解。作为一个即插即用模块，SentiLLM在仅使用少量可训练参数的情况下显著提高了区分性能。我们的方法在四个数据集（MOSI、MOSEI、CH-SIMS和CH-SIMS v2）上取得了优越的表现，证明了结构抽象范式在MSA中的有效性。我们的代码可在以下链接获取： exttt{https://github.com/especiallyW/SentiLLM}。
cs.CL / 28 / 2607.27816
Beyond Borrowed Histories: Person-Aligned User Simulation for Interactive Role-Playing Evaluation
超越借用的历史：面向用户的角色扮演评估的个性化用户模拟
Abstract
Role-playing agents (RPAs) have become one of the most important consumer applications of large language models. Users engage in multi-turn conversations with RPAs for experiences such as emotional comfort, making reliable evaluation essential for measuring capability, comparing systems, and guiding further improvement. Existing benchmarks, however, typically require an RPA to continue a fixed dialogue history and then evaluate the continuation using a fixed rubric detached from the user. We identify and empirically demonstrate two limitations of this design. First, an RPA's output is shaped by the preceding dialogue history, preventing a scientifically grounded assessment of its role-playing ability in real multi-turn settings. Second, user experience varies substantially across individuals, and conventional fixed rubrics need not align with user satisfaction. We therefore introduce PALATE (Person-Aligned LLM-Simulated-User Assessment with Tailored Evaluation), a scalable RPA benchmark built on user simulators. PALATE is accompanied by a pool of 300 character profiles. Its main evaluation trains five per-user simulators and lets them engage candidate RPAs in free-form, multi-turn conversations over a pre-frozen panel of character profiles. Alongside a general quality rubric, we construct personalized rubrics to measure user satisfaction; on held-out annotated data, the personalized rubrics show higher agreement with human judgments than the general rubric. In the main evaluation of 16 candidates, PALATE separately characterizes generic turn quality, long-horizon session capability, and per-user experience on multi-turn trajectories co-constructed by each candidate. It thereby produces interpretable evaluations of specific user-RPA pairs rather than compressing systems into a single user-independent ranking.
Chinese Translation
角色扮演代理（RPA）已成为大型语言模型最重要的消费应用之一。用户与RPA进行多轮对话，以获得情感安慰等体验，因此可靠的评估对于衡量能力、比较系统和指导进一步改进至关重要。然而，现有的基准通常要求RPA继续固定的对话历史，然后使用与用户脱离的固定标准进行评估。我们识别并实证展示了这种设计的两个局限性。首先，RPA的输出受前述对话历史的影响，阻碍了在真实多轮环境中对其角色扮演能力的科学评估。其次，用户体验在个体之间差异显著，传统的固定标准未必与用户满意度一致。因此，我们引入了PALATE（个性化用户评估的LLM模拟用户评估），这是一个基于用户模拟器的可扩展RPA基准。PALATE配备了300个角色档案的池。其主要评估训练五个每用户模拟器，并让它们在预先冻结的角色档案面板上与候选RPA进行自由形式的多轮对话。除了通用质量标准外，我们还构建了个性化标准来衡量用户满意度；在保留的标注数据上，个性化标准与人类判断的协议高于通用标准。在16个候选者的主要评估中，PALATE分别表征了通用轮次质量、长时间会话能力以及每用户在由每个候选者共同构建的多轮轨迹上的体验。通过这种方式，它生成了特定用户-RPA对的可解释评估，而不是将系统压缩为单一的用户独立排名。
cs.CL / 29 / 2607.27845
AutoSupervision: Closing the Feedback Loop in Scientific Workflows with Grounded Revision Verification
自动监督：通过基于证据的修订验证闭合科学工作流程中的反馈循环
Abstract
Recent advances in large language models (LLMs) have enabled AI systems to assist scientific research and peer review. However, an essential capability for reliable AI-assisted scientific workflows remains underexplored: verifying whether reviewer feedback leads to meaningful and evidence-supported manuscript improvements. We introduce AutoSupervision, which evaluates whether scientific manuscript revisions genuinely address reviewer concerns through grounded evidence. AutoSupervision leverages transparent peer-review records as a natural source of supervision, where reviewer comments specify scientific concerns, author responses describe claimed resolutions, and revised manuscripts provide evidence of changes. Given reviewer comments, author responses, and revised manuscripts, models must characterize reviewer concerns, determine whether concerns have been addressed, and identify supporting manuscript evidence. We construct AutoSupervision from 56,000 Nature Communications articles and corresponding review records. Then we conducted experiments on LLMs, the ablation study, and the case study. Our results show that while LLMs perform well in characterizing reviewer concerns, with GPT-5.5 achieving a score of 0.754, evidence-based verification remains the primary bottleneck, with the best-performing model reaching only 0.501.
Chinese Translation
近期大型语言模型（LLMs）的进展使得人工智能系统能够辅助科学研究和同行评审。然而，可靠的人工智能辅助科学工作流程所需的一项关键能力仍未得到充分探索：验证审稿人的反馈是否导致了有意义且有证据支持的手稿改进。我们提出了自动监督（AutoSupervision），该方法评估科学手稿的修订是否真正解决了审稿人的关切，并通过基于证据的方式进行验证。自动监督利用透明的同行评审记录作为自然的监督来源，其中审稿人评论指定了科学问题，作者回应描述了声称的解决方案，而修订后的手稿提供了变更的证据。在给定审稿人评论、作者回应和修订手稿的情况下，模型必须表征审稿人的关切，确定这些关切是否已被解决，并识别支持手稿的证据。我们从56,000篇《自然通讯》（Nature Communications）文章及相应的审稿记录中构建了自动监督。随后，我们对大型语言模型进行了实验，包括消融研究和案例研究。我们的结果表明，尽管大型语言模型在表征审稿人关切方面表现良好，GPT-5.5的得分为0.754，但基于证据的验证仍然是主要瓶颈，表现最佳的模型仅达到0.501。
cs.CL / 30 / 2607.27851
Beyond Feeling Better: Capability-Sustaining Emotional Dialogue as a Longitudinal Research Paradigm
超越单纯的情感改善：能力维持的情感对话作为纵向研究范式
Abstract
Emotional dialogue research includes two influential strategy traditions. Empathetic dialogue prioritizes understanding a speaker's emotional experience. Emotional support conversation selects and sequences support for the seeker's current needs. Sustained use introduces a further goal. Effective support should sustain users' capacities for emotion regulation, coping, self-endorsed decisions, and social connection across the interaction lifecycle. We propose capability-sustaining emotional dialogue (CSED) as a longitudinal research paradigm that aligns supportive strategy with this goal and organizes data, models, system design, evaluation, and governance around repeated use, non-use, transition, and termination. A targeted literature-and-corpus audit motivates this position. In a PRISMA-ScR-guided sample, 95% of 60 system-building papers pursue relief-oriented goals. None evaluates capability or longitudinal outcomes, and only 1 considers dependency, autonomy, or termination risk. In 300 ESConv supporter turns, capability-relevant functions appear in 43.0%, while generic suggestions account for 22.0%, compared with 4.0% reappraisal, 6.7% self-efficacy support, and 0.3% boundary behavior. We release a protocol for extending the audit to model behavior. An illustrative process model connects latent user capability to six design commitments, four evaluation timescales, and lifecycle constraints. The resulting agenda makes CSED testable across data, policy design, training, evaluation, and governance.
Chinese Translation
情感对话研究包括两种影响力较大的策略传统。共情对话优先考虑理解说话者的情感体验。情感支持对话则选择并排列支持以满足寻求者当前的需求。持续使用引入了一个进一步的目标：有效的支持应在互动生命周期内维持用户的情感调节、应对、自我认可决策和社会联系的能力。我们提出能力维持的情感对话（Capability-Sustaining Emotional Dialogue, CSED）作为一种纵向研究范式，该范式将支持策略与这一目标相结合，并围绕重复使用、未使用、过渡和终止组织数据、模型、系统设计、评估和治理。针对文献和语料库的审计支持了这一立场。在一个遵循PRISMA-ScR指导的样本中，60篇系统构建论文中有95%追求以缓解为导向的目标。没有一篇评估能力或纵向结果，只有1篇考虑了依赖性、自主性或终止风险。在300个情感支持对话（ESConv）支持者的发言中，相关能力的功能出现在43.0%，而通用建议占22.0%，相比之下，重新评估占4.0%，自我效能支持占6.7%，边界行为占0.3%。我们发布了一项扩展审计以建模行为的协议。一个示例性过程模型将潜在用户能力与六个设计承诺、四个评估时间尺度和生命周期约束相连接。由此产生的议程使得CSED在数据、政策设计、培训、评估和治理方面可被测试。
cs.CL / 31 / 2607.27853
FinanceHarness: Autonomous Financial Deep Research Framework
FinanceHarness：自主金融深度研究框架
Abstract
Powered by advances in LLMs and autonomous agents, deep research has become one of the most widely adopted agentic products. However, most deep research systems write general-purpose reports, which are inadequate for financial deep research. Financial research demands specialized knowledge to analyze historical patterns and forecast upcoming events. Automating financial deep research therefore requires both a layered harness to drive the research agent and a verifiable, point-in-time benchmark that prevents leakage of future information. We present FinanceHarness, a harness that runs finance-oriented tools and practitioner-guided workflows, automating financial deep research end to end: environment and data construction, the agent execution loop, and reward modeling. We further propose FinanceGym, comprising thesis-driven research questions and rubrics that combine pre-cutoff and post-cutoff criteria. Professional expert validation yields an 82% pass rate. Even leading LLMs and agents score below 40% on the rubrics, showing that FinanceGym is challenging and leaves substantial headroom. With the same open-weight backbone, FinanceHarness improves the overall rubric score from 25.3% to 32.4%. FinanceHarness is available at https://github.com/Yijia-Xiao/FinanceHarness.
Chinese Translation
得益于大规模语言模型（LLMs）和自主智能体的进展，深度研究已成为最广泛采用的智能产品之一。然而，大多数深度研究系统撰写的是通用报告，这对于金融深度研究而言是不够的。金融研究需要专业知识来分析历史模式并预测即将发生的事件。因此，自动化金融深度研究不仅需要一个分层的框架来驱动研究智能体，还需要一个可验证的、特定时点的基准，以防止未来信息的泄露。我们提出了FinanceHarness，这是一个运行金融导向工具和从业者指导工作流程的框架，能够端到端地自动化金融深度研究：环境和数据构建、智能体执行循环以及奖励建模。我们进一步提出了FinanceGym，包含以论文为驱动的研究问题和结合截止前后标准的评分标准。专业专家验证的通过率为82%。即使是领先的大规模语言模型和智能体在评分标准上的得分也低于40%，显示出FinanceGym的挑战性和巨大的提升空间。在相同的开放权重基础上，FinanceHarness将整体评分从25.3%提高到32.4%。FinanceHarness可在https://github.com/Yijia-Xiao/FinanceHarness获取。
cs.CL / 32 / 2607.27919
Memory Decoder at Scale: A Pretrained, Parametric Long-Term Memory
大规模记忆解码器：一种预训练的参数化长期记忆
Abstract
Decoder-only language models entangle long-term memory and reasoning in a single parameter set, making it difficult to scale memory capacity independently. Memory Decoder introduces a parametric long-term memory module but only studies it at a relatively small scale. In this work, we present Memory Decoder at Scale, scaling memory models up to 6.9B parameters and pretraining them on 300B tokens. At this data scale, the combined cost of indexing and search makes a standard Faiss pipeline infeasible. We address this bottleneck with a distributed pipeline for Faiss indexing and retrieval, together with sparse, batch-wise loading of kNN distributions. Across model scales, we find that allocating more parameters to memory yields a better parameter-performance tradeoff than scaling the base model alone. On 17 benchmarks, pairing a 6.9B general memory with Pythia-410M raises its average score from 29.86 to 37.34, surpassing Pythia-12B (37.24) with 39% fewer total parameters. For Qwen3 Base models ranging from 0.6B to 14B, 1.7B domain memories improve the average score across the three domains by more than 9 points at every scale. Overall, our results demonstrate that independently scaling pretrained memory offers a more parameter efficient path to improving language model performance.
Chinese Translation
仅解码器的语言模型将长期记忆和推理纠缠在一个参数集内，使得独立扩展记忆容量变得困难。记忆解码器引入了一种参数化的长期记忆模块，但仅在相对较小的规模上进行了研究。在本研究中，我们提出了大规模记忆解码器，将记忆模型扩展到69亿参数，并在3000亿个标记上进行预训练。在这个数据规模下，索引和搜索的综合成本使得标准的Faiss管道不可行。我们通过为Faiss索引和检索构建分布式管道，以及稀疏的批量加载kNN分布来解决这一瓶颈。在不同模型规模下，我们发现将更多参数分配给记忆比单独扩展基础模型提供了更好的参数-性能权衡。在17个基准测试中，将69亿的一般记忆与Pythia-410M配对，使其平均得分从29.86提高到37.34，超越了Pythia-12B（37.24），且总参数减少了39%。对于范围从6亿到140亿的Qwen3基础模型，1.7亿的领域记忆在每个规模上将三个领域的平均得分提高了超过9分。总体而言，我们的结果表明，独立扩展预训练记忆提供了一条更高效的参数路径，以提高语言模型的性能。
cs.CL / 33 / 2607.28008
RepBench: Compiling Benchmarks into Capability Representations for Large Language Models
RepBench：将基准测试编译为大型语言模型的能力表示
Abstract
Representation engineering reads and steers capability directions in large language models, yet methods are typically evaluated on paper-specific synthetic data. The resulting measurements are difficult to compare or reproduce and may reflect surface patterns rather than capabilities. We present RepBench, a benchmark-grounded data layer for capability-aligned representation probing. Crawling 13,427 benchmark papers yields a taxonomy of 182 capability clusters in 13 families; harvesting 353 public benchmark datasets yields 46,149 audited probe texts covering 94 capabilities, each supported by at least two independent benchmarks. This multi-benchmark design reduces dependence on any single source: raw per-text vectors exhibit no natural cluster granularity, whereas benchmark-pooled capability vectors show an interior clustering optimum at a small number of clusters on all 12 evaluated models, with low agreement to the human taxonomy. Under cross-benchmark transfer evaluation across twelve models completed by all four readouts, difference-in-means attains the highest model-level mean on ten models, while logistic regression wins the most capability-model cells. This disagreement shows that the readout method and aggregation criterion are meaningful evaluation dimensions. The pipeline, corpus, and evaluation code are released as a reusable closed-loop workflow.
Chinese Translation
表示工程读取并引导大型语言模型的能力方向，然而方法通常在特定论文的合成数据上进行评估。由此产生的测量结果难以比较或重现，可能反映的是表面模式而非实际能力。我们提出了RepBench，这是一个基于基准的能力对齐表示探测的数据层。通过爬取13,427篇基准论文，我们获得了13个家族中182个能力集群的分类；收集353个公共基准数据集，得到46,149个经过审核的探测文本，涵盖94种能力，每种能力至少由两个独立的基准支持。这种多基准设计减少了对任何单一来源的依赖：原始的每文本向量没有自然的集群粒度，而基准汇聚的能力向量在所有12个评估模型上显示出在少数集群中存在内部聚类最优，且与人类分类的低一致性。在由四个读出完成的十二个模型的跨基准转移评估中，均值差异在十个模型上达到了最高的模型级均值，而逻辑回归在最多的能力-模型单元中获胜。这种不一致表明，读出方法和聚合标准是有意义的评估维度。该管道、语料库和评估代码作为可重用的闭环工作流发布。
cs.CL / 34 / 2607.28077
LEEPS: Latent-Guided Explore-Exploit Prompt Sampling for Efficient RLVR in Large Language Models
LEEPS：用于大语言模型高效RLVR的潜在引导探索-利用提示采样
Abstract
Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models, but prompt groups with identical rollout rewards consume generation budget without effective learning signals. Pre-rollout prompt selection can reduce this waste by screening prompts before rollout generation. However, existing pre-rollout methods struggle to balance exploitation and exploration: repeatedly exploiting historically informative prompts can narrow training coverage, whereas broader exploration can lower the fraction of informative prompts. To address these limitations, we introduce LEEPS, a Latent-Guided Explore--Exploit Prompt Sampler that adaptively balances the reuse of previously observed informative prompts with continued exploration of uncertain ones. LEEPS partitions candidates into exploit and explore portfolios and adaptively allocates rollout budget according to their recent non-trivial ratios. It further uses representation-space neighbors and historical rollout outcomes to prioritize uncertain prompts likely to yield non-zero reward variance, thereby making exploration more targeted without additional rollouts. Across six mathematical reasoning benchmarks, LEEPS achieves the highest average score at both model scales, with relative gains of 2.6\% and 3.7\% over the strongest baseline for Qwen2.5-Math-1.5B and 7B, respectively, and generally improves faster during the training process. It also achieves the highest average score across the three evaluated OOD general-reasoning benchmarks at both model scales and adds only about 2 seconds of online sampling overhead per training step. Code is available at https://github.com/ShuangLiangX/LEEPS.
Chinese Translation
具有可验证奖励的强化学习（RLVR）提升了大语言模型的推理能力，但具有相同回报的提示组在生成预算中消耗了资源而未能提供有效的学习信号。预生成提示选择可以通过在生成之前筛选提示来减少这种浪费。然而，现有的预生成方法在利用和探索之间难以取得平衡：重复利用历史上信息丰富的提示可能会缩小训练覆盖范围，而更广泛的探索则可能降低信息丰富提示的比例。为了解决这些限制，我们提出了LEEPS，一种潜在引导的探索-利用提示采样器，能够自适应地平衡对先前观察到的信息丰富提示的重用与对不确定提示的持续探索。LEEPS将候选提示分为利用和探索投资组合，并根据其最近的非平凡比例自适应分配生成预算。它进一步利用表示空间邻域和历史生成结果来优先考虑可能产生非零奖励方差的不确定提示，从而使探索更加有针对性而无需额外的生成。在六个数学推理基准测试中，LEEPS在两个模型规模上都达到了最高的平均分数，相较于最强基线在Qwen2.5-Math-1.5B和7B上分别提高了2.6%和3.7%，并且在训练过程中通常提高得更快。它还在评估的三个OOD一般推理基准测试中，在两个模型规模上达到了最高的平均分数，并且每个训练步骤仅增加约2秒的在线采样开销。代码可在 https://github.com/ShuangLiangX/LEEPS 获取。
cs.CL / 35 / 2607.28082
GGC: Selective Query Correction for Reliable Text-to-SPARQL Generation
GGC：用于可靠文本到SPARQL生成的选择性查询修正
Abstract
Large language models (LLMs) have demonstrated strong capabilities in structured query generation, making them a natural choice for Text-to-SPARQL, which translates natural language questions into executable SPARQL queries over knowledge graphs. However, their initial outputs remain unreliable: generated queries may be executable yet semantically misaligned with input questions, leading to incorrect retrieval. To address this issue, we propose Generator-Gate-Corrector (GGC), a framework for reliable LLM-based Text-to-SPARQL generation. GGC first uses a Generator to produce an initial query, then applies a Gate to predict whether correction is needed, and finally invokes a Corrector only for selected high-risk queries. This selective correction mechanism avoids unnecessary modifications and reduces the risk of degrading originally correct queries. Experiments on MCQA show that GGC improves query-level accuracy from 90.23\% to 98.33\% while reducing inference overhead by 45\% compared with correcting all generated queries. Ablation studies show that the Gate is robust across thresholds and that Corrector training data composition affects correction effectiveness and stability. Overall, the results demonstrate that selective correction enhances the accuracy, reliability, and efficiency of LLM-based text-to-SPARQL generation.
Chinese Translation
大型语言模型（LLMs）在结构化查询生成方面表现出强大的能力，使其成为文本到SPARQL的自然选择，该过程将自然语言问题转化为可执行的SPARQL查询，适用于知识图谱。然而，它们的初始输出仍然不可靠：生成的查询可能是可执行的，但在语义上与输入问题不一致，从而导致错误的检索。为了解决这个问题，我们提出了生成器-门控-修正器（Generator-Gate-Corrector，GGC），这是一个基于LLM的可靠文本到SPARQL生成框架。GGC首先使用生成器生成初始查询，然后应用门控来预测是否需要修正，最后仅对选定的高风险查询调用修正器。这种选择性修正机制避免了不必要的修改，并降低了原本正确的查询被降级的风险。在MCQA上的实验表明，与对所有生成查询进行修正相比，GGC将查询级准确率从90.23%提高到98.33%，同时减少了45%的推理开销。消融研究表明，门控在不同阈值下表现稳健，修正器的训练数据组成影响修正的有效性和稳定性。总体而言，结果表明选择性修正增强了基于LLM的文本到SPARQL生成的准确性、可靠性和效率。
cs.CL / 36 / 2607.28119
Challenges in annotations by humans and LLMs: A case study of evaluative language
人类与大型语言模型（LLMs）在注释中的挑战：评估性语言的案例研究
Abstract
In this paper, we draw a comparison between linguists in training, a trained linguist, and annotations generated by large language models (LLMs) to find out if they struggle with complex linguistic phenomena in a similar way. For this purpose, we analyse evaluative language in spoken popular science discourse, with the example of a corpus of English TED talk transcripts. We focus on the Appraisal theory and its Attitude subsystem, including the categories (classes) of Affect, Judgement, and Appreciation. In this context, Appraisal theory is an example of a highly subjective annotation task, making it a suitable example for the study of complex annotation challenges. First, we assess human annotations on a sentence level in specific scientific domains. Then, we develop three prompts and compare them for model performance for the automatic classification of Appraisal classes. We assess the performance of three LLMs using the best-performing prompt and finetune the model, reaching an F1-score of 0.77. We find that models perform best compared to annotations conducted by the trained linguist, while linguists in training do not reach high agreement scores. We conclude that LLMs can aid in complex annotation task resolution, opening new pathways for the complex theories annotated and analyzed in digital humanities studies.
Chinese Translation
在本文中，我们比较了正在培训的语言学家、受过训练的语言学家以及由大型语言模型（LLMs）生成的注释，以探讨他们在处理复杂语言现象时是否面临类似的困难。为此，我们分析了口语流行科学话语中的评估性语言，以英语TED演讲稿语料库为例。我们关注评估理论及其态度子系统，包括情感（Affect）、判断（Judgement）和欣赏（Appreciation）等类别（类）。在此背景下，评估理论是一个高度主观的注释任务示例，使其成为研究复杂注释挑战的合适案例。首先，我们在特定科学领域的句子层面评估人类注释。然后，我们开发了三个提示，并比较它们在自动分类评估类别方面的模型性能。我们使用表现最佳的提示评估了三种LLMs的性能，并对模型进行了微调，达到了0.77的F1分数。我们发现，与受过训练的语言学家的注释相比，模型的表现最佳，而正在培训的语言学家未能达到高一致性得分。我们得出结论，LLMs可以帮助解决复杂的注释任务，为数字人文学科研究中注释和分析的复杂理论开辟新的路径。
cs.CL / 37 / 2607.28127
FinSMART: Financial Sentiment Analysis for Algorithmic Trading through Market-Aligned Reinforcement Learning
FinSMART：通过市场对齐的强化学习进行算法交易的金融情感分析
Abstract
Recent advances in Generative AI have substantially improved financial sentiment analysis through post-trained financial large language models (LLMs). However, existing approaches remain confined to a market-agnostic, supervised learning paradigm that relies on limited, static and human-annotated datasets, and thus are incapable of adapting to evolving market conditions. To address this limitation, we introduce FinSMART, the first market-aligned reinforcement learning framework for financial sentiment analysis, which directly optimizes sentiment signals using realized market outcomes. To deal with the noisy, non-stationary, and multifactorial nature of financial markets, FinSMART incorporates a signal extraction pipeline that combines market-aware data filtering with a discrete asymmetric trading reward, enabling stable reinforcement learning from economically meaningful market feedback. Experimental results demonstrate that FinSMART significantly outperforms existing state-of-the-art methods in profitability, risk-adjusted performance, and sentiment signal quality, improving cumulative trading returns by 220% over the strongest baseline. Uniquely, the FinSMART framework naturally supports market-aware retraining, at any point in time, by replacing costly manual annotation with newly observed financial articles and their realized market outcomes. Such a retraining strategy enables the model to continuously adapt to changing market dynamics, resulting in consistent performance gains over its static counterpart. These findings demonstrate the practical applicability of market-aligned reinforcement learning and highlight its potential as a next-generation paradigm for developing adaptive financial LLMs.
Chinese Translation
最近，生成性人工智能的进展显著提升了金融情感分析，特别是通过后训练的金融大型语言模型（LLMs）。然而，现有的方法仍然局限于市场无关的监督学习范式，依赖于有限、静态且人工标注的数据集，因此无法适应不断变化的市场条件。为了解决这一局限性，我们提出了FinSMART，这是第一个市场对齐的强化学习框架，用于金融情感分析，直接利用实际市场结果优化情感信号。为了应对金融市场的噪声、非平稳和多因素特性，FinSMART结合了市场感知的数据过滤与离散的非对称交易奖励，构建了一个信号提取管道，使得从经济上有意义的市场反馈中进行稳定的强化学习成为可能。实验结果表明，FinSMART在盈利能力、风险调整表现和情感信号质量方面显著优于现有的最先进方法，累计交易回报比最强基线提高了220%。独特的是，FinSMART框架自然支持市场感知的再训练，能够在任何时刻通过用新观察到的金融文章及其实际市场结果替代昂贵的人工标注。这种再训练策略使模型能够持续适应变化的市场动态，从而在其静态对应物上实现持续的性能提升。这些发现展示了市场对齐的强化学习的实际应用潜力，并强调了其作为开发自适应金融大型语言模型的下一代范式的潜力。
cs.CL / 38 / 2607.28128
Rethinking LLM-Judged Helpfulness as a Pedagogy Signal: A Pre-Registered Audit Across Tutor Models
重新思考 LLM 评判的有用性作为教学信号：对辅导模型的预注册审计
Abstract
LLM tutoring poses a measurement problem: can a general-purpose helpfulness rubric distinguish direct answer-giving from pedagogical guidance? We audit this signal in a pre-registered study. Within each of three tutor bases, we compare conversational and pedagogical policies instantiated with the same underlying model and paired with one fixed weak simulated student. Deterministic detectors measure answer leakage and next-turn independent work. Claude Opus 4.8 is the frozen, condition-blind primary judge. After the Opus scores were fixed, GPT-5.6 Sol was prospectively specified for a post hoc robustness audit of the same 1,179 confirmatory answer-phase tutor turns under the frozen helpfulness and pedagogy rubrics. On the primary base under Opus, the policies do not differ significantly in helpfulness but are perfectly rank-separated under the pedagogy rubric (Cliff's $|\delta|{=}0.10$ vs. $1.0$). Across the two judges, pedagogy contrasts retain their direction where detected, whereas the helpfulness ordering is judge-contingent, reversing between judges on two of three bases. In an Opus-only ablation, seven primary-base policies span $2.3$ points in mean judged pedagogy within a $0.25$-point band of mean judged helpfulness. Separately, answer-revealing turns are followed by less independent student work on every base, a result that is judge-invariant by construction. In this controlled setting, general-purpose helpfulness is not a reliable pedagogy signal. Tutor evaluation should pair pedagogy-targeted rubrics with deterministic process measures.
Chinese Translation
LLM 辅导面临一个测量问题：通用的有用性标准能否区分直接给出答案与教学指导？我们在一项预注册研究中审计了这一信号。在三个辅导基础中，我们比较了使用相同基础模型实例化的对话政策和教学政策，并与一个固定的弱模拟学生配对。确定性检测器测量答案泄漏和下一轮独立工作。Claude Opus 4.8 是冻结的、条件盲的主要评判者。在 Opus 分数固定后，GPT-5.6 Sol 被前瞻性指定用于对同一 1,179 个确认答案阶段的辅导回合进行事后稳健性审计，依据冻结的有用性和教学标准。在 Opus 的主要基础下，政策在有用性上没有显著差异，但在教学标准下完全排名分离（Cliff's $| ext{δ}|{=}0.10$ 对比 $1.0$）。在两个评判者之间，检测到的教学对比保持其方向，而有用性排序则依赖于评判者，在三个基础中的两个上出现反转。在仅使用 Opus 的消融实验中，七个主要基础政策在平均评判教学上跨越 $2.3$ 分，处于平均评判有用性的 $0.25$ 分带内。另一方面，答案揭示的回合后，学生的独立工作在每个基础上都减少，这一结果在构造上是评判者不变的。在这一受控环境中，通用的有用性并不是一个可靠的教学信号。辅导评估应将针对教学的标准与确定性过程测量相结合。
cs.CL / 39 / 2607.28146
Can Agents Deceive? Evaluating Reasoning and Deception in ParliamentBench using a Social Deduction Game
代理能欺骗吗？在社会推理游戏中评估议会基准的推理与欺骗
Abstract
As large language models (LLMs) are deployed as agents in high-stakes settings, such as medical and legal systems, understanding their deceptive capabilities is fundamental to safety. Controlled social deduction games provide a reproducible proxy for isolating and evaluating these complex adversarial behaviors. We present the open-source benchmark framework ParliamentBench based on the game Secret Hitler to evaluate LLMs in scenarios that require deception, persuasion, and reasoning under information asymmetry. We evaluate 16 LLMs across 1,600 simulated matches playing each other, playing against humans, and compare them against a large set of online games. We introduce three novel metrics that isolate social deduction, reasoning, and deceptive consistency. Our experiments reveal that frontier models achieve strong performance across cooperative and deceptive roles, with a strong top-four cluster (GPT-5.4, Kimi K2.5, Grok 4.1 Fast, and DeepSeek 3.1 Terminus), whereas the weakest models fall short of random (33%) and simple algorithmic (45%) baselines. Most LLMs struggle to maintain a consistent deceptive persona throughout an entire game, with deception retention dropping below 50%.
Chinese Translation
随着大型语言模型（LLMs）被部署为高风险环境中的代理，例如医疗和法律系统，理解它们的欺骗能力对安全至关重要。受控的社会推理游戏提供了一种可重复的代理，用于隔离和评估这些复杂的对抗行为。我们提出了基于游戏《秘密希特勒》的开源基准框架ParliamentBench，以评估LLMs在需要欺骗、说服和信息不对称下推理的场景中的表现。我们在1600场模拟比赛中评估了16个LLMs，比较它们之间的对战、与人类的对战，并与大量在线游戏进行比较。我们引入了三种新颖的指标，以隔离社会推理、推理能力和欺骗一致性。我们的实验表明，前沿模型在合作和欺骗角色中表现出色，形成了一个强大的前四名集群（GPT-5.4、Kimi K2.5、Grok 4.1 Fast和DeepSeek 3.1 Terminus），而最弱的模型则未能达到随机（33%）和简单算法（45%）的基准。大多数LLMs在整个游戏中难以保持一致的欺骗角色，欺骗保留率降至50%以下。
cs.CL / 40 / 2607.28156
RRM: Experience-Driven Reflective Retrieval Memory for Long-Horizon Multimodal Reasoning
RRM：基于经验驱动的反思性检索记忆用于长时域多模态推理
Abstract
Existing multimodal long-term memory agents use external memory to overcome the limited context available for long videos. However, most methods emphasize what to store rather than how stored memory should be retrieved. When retrieval becomes inaccurate or repeatedly fails to obtain useful evidence, existing agents lack mechanisms to diagnose failures from previous task trajectories and adapt future search strategies.We introduce Reflective Retrieval Memory (RRM), a reflective memory framework for long-horizon multimodal reasoning. RRM augments an entity-centric multimodal memory graph with reflective experience memory, which distills transferable procedural retrieval knowledge from historical task trajectories. Unlike episodic and semantic memories that preserve factual evidence from the current video, reflective experience memory captures reusable search strategies across tasks. RRM converts retrieved experiences into query-level guidance, while answer generation remains conditioned only on factual evidence newly retrieved from the current video. A lifecycle management mechanism further regulates experience memory through usage frequency, reuse feedback, and temporal decay, thereby reducing redundancy and noise. RRM consistently outperforms previous state-of-the-art approaches on M3-Bench-Robot, M3-Bench-Web, and Video-MME-Long, demonstrating the effectiveness of reflective retrieval memory for long-horizon multimodal reasoning.
Chinese Translation
现有的多模态长期记忆代理使用外部记忆来克服长视频可用上下文的限制。然而，大多数方法强调存储什么，而不是如何检索存储的记忆。当检索变得不准确或反复无法获得有用证据时，现有代理缺乏从先前任务轨迹中诊断失败的机制，并适应未来的搜索策略。我们引入了反思性检索记忆（Reflective Retrieval Memory, RRM），这是一种用于长时域多模态推理的反思性记忆框架。RRM通过反思性经验记忆增强了以实体为中心的多模态记忆图，该记忆图从历史任务轨迹中提炼可转移的程序性检索知识。与保留当前视频事实证据的情节记忆和语义记忆不同，反思性经验记忆捕捉了跨任务的可重用搜索策略。RRM将检索到的经验转化为查询级指导，而答案生成仅依赖于从当前视频中新检索到的事实证据。生命周期管理机制进一步通过使用频率、重用反馈和时间衰减来调节经验记忆，从而减少冗余和噪声。RRM在 M3-Bench-Robot、M3-Bench-Web 和 Video-MME-Long 上始终优于先前的最先进方法，证明了反思性检索记忆在长时域多模态推理中的有效性。
cs.CL / 41 / 2607.28166
Where and When to Commit: Candidate-Aware Decoding for Diffusion Language Models
何时何地进行承诺：候选人感知解码用于扩散语言模型
Abstract
Diffusion language models (DLMs) expose a provisional prediction at every denoising step, creating an opportunity for generation-time early exit that stops decoding before the schedule is exhausted. Existing early-exit gates decide termination from fixed-region confidence statistics or schedule-dependent rules, evidence too coarse for a decision that freezes every remaining position at once, so they fire prematurely on long chain-of-thought outputs whose answers stabilize only near the end. Adaptive sampling, the other axis of training-free acceleration, paces how quickly positions commit while decoding continues but never verifies that the output itself has stabilized. We introduce a training-free, candidate-aware early-exit framework that keeps the two axes separate and matches each decision to evidence of its own scope. Confidence-Verified Commit (CVC) governs when the sequence may stop by verifying confidence and sustained argmax stability over the dynamically extracted candidate span using a deterministic parser specified from each task's output format. Block-Wise Early Commit (BWEC) governs where to accelerate by applying a cheaper local rule to non-final blocks, while leaving the final block and global termination under CVC. We refer to their combination as LATCH (Localized Acceleration with Tracked-Candidate Halting). Unlike prior methods, LATCH needs no suffix-prompt construction; it is prompt-anchor-free but format-aware. We evaluate LATCH end to end on 11 tasks under zero-shot settings using LLaDA and Dream. LATCH stays within 2.0 percentage points of full-decoding accuracy across all 22 evaluation settings, with one frozen hyperparameter set that transfers cross-backbone untuned, while achieving end-to-end TPS speedups of 9.3-17.8x on short-answer tasks and 2.0-3.3x on long-reasoning tasks.
Chinese Translation
扩散语言模型（DLMs）在每个去噪步骤中都会暴露出一个临时预测，这为生成时的提前退出提供了机会，可以在调度耗尽之前停止解码。现有的提前退出门控通过固定区域的置信度统计或依赖于调度的规则来决定终止，这些证据对于一次性冻结所有剩余位置的决策来说过于粗糙，因此在长链思维输出中，它们会过早触发，而这些输出的答案通常在接近末尾时才会稳定。自适应采样是另一种无训练加速的方式，它在解码继续时控制位置承诺的速度，但从未验证输出本身是否已稳定。我们提出了一种无训练的候选人感知提前退出框架，该框架将两个轴分开，并将每个决策与其自身范围的证据相匹配。置信度验证承诺（Confidence-Verified Commit, CVC）通过验证置信度和在动态提取的候选范围内的持续 argmax 稳定性来管理序列何时可以停止，使用从每个任务的输出格式指定的确定性解析器。块级提前承诺（Block-Wise Early Commit, BWEC）通过对非最终块应用更便宜的局部规则来管理加速的位置，同时将最终块和全局终止留给 CVC。我们将它们的组合称为 LATCH（基于跟踪候选停止的局部加速）。与之前的方法不同，LATCH 不需要后缀提示构造；它是无提示锚的，但对格式敏感。我们在 11 个任务上使用 LLaDA 和 Dream 在零-shot 设置下对 LATCH 进行了端到端评估。LATCH 在所有 22 个评估设置中保持在完整解码准确率的 2.0 个百分点以内，使用一个冻结的超参数集进行跨骨干网的无调优，同时在短答案任务上实现了 9.3-17.8 倍的端到端 TPS 加速，在长推理任务上实现了 2.0-3.3 倍的加速。
cs.CL / 42 / 2607.28190
The MADRS Pipeline: Supporting Depression Assessment in Clinical Trials
MADRS管道：支持临床试验中的抑郁症评估
Abstract
Depression is a major mental disorder for which diagnosis relies primarily on clinical assessments. Automated methods to support its detection via the psychiatric MADRS scale are getting more and more attention. While existing solutions primarily focus on detecting the disorder from different text sources (e.g., online text, social media), there is still limited support for clinical trials, where clinical assessments are conducted through structured interviews based on standard guidelines such as SIGMA. In this work, we develop a LLM pipeline specifically designed to support clinicians in supporting the assessment of depression in patients enrolled in clinical trials. Our pipeline converts audio interviews into transcripts, maps them into the ten MADRS symptom items, estimates their severity, and identify problematic clinical ratings associated with them. Evaluation on real clinical interviews shows a strong overall correlation of 0.867 with expert ratings, providing interpretable support for future assessments in clinical trials.
Chinese Translation
抑郁症是一种主要的心理障碍，其诊断主要依赖于临床评估。通过精神病学MADRS量表支持抑郁症检测的自动化方法越来越受到关注。尽管现有解决方案主要集中在从不同文本来源（例如，在线文本、社交媒体）中检测该障碍，但在临床试验中仍然缺乏支持，在这些试验中，临床评估是通过基于标准指南（如SIGMA）的结构化访谈进行的。在本研究中，我们开发了一种专门设计的LLM管道，以支持临床医生评估参与临床试验患者的抑郁症。我们的管道将音频访谈转换为文本，将其映射到十个MADRS症状项中，估计其严重程度，并识别与之相关的问题临床评分。对真实临床访谈的评估显示，与专家评分的整体相关性达到0.867，为未来的临床试验评估提供了可解释的支持。
cs.CL / 43 / 2607.28196
Fidelity Is Not Safety: Gently-Compressed LLMs Pass Every Data-Free Quality Guard Yet Invent Procedure Steps in Agentic Execution
保真度并不等于安全性：温和压缩的语言模型在每个无数据质量检测中均表现合格，但在自主执行时发明程序步骤
Abstract
Practitioners accept a compressed language model once it clears a stack of data-cheap quality guards: perplexity within a small factor of the original, downstream accuracy (for example MMLU) inside a confidence interval, and data-free output-fidelity signals that compare the compressed and original network's internal representations under random probe inputs. This stack has a blind spot. Across three model families, gently-compressed models clear every guard and then invent procedure steps that were never in the instructions when they run a standard operating procedure (SOP) as an agent. The effect is operator-specific: coherent low-rank (SVD) truncation induces it, and magnitude pruning matched to the same perplexity does not. One dissociation isolates the cause. The same compressed weights that CI-win a paired output-fidelity test CI-fail the invented-step canary. The governing axis is the coherence of the compression error times its rate; the magnitude of the damage does not predict it. The data-free fidelity probe is a fidelity oracle by construction, so it cannot see this axis. We characterize the blindspot and dissociation with paired confidence intervals on a pre-registered, powered canary across three architectures. Operator-specificity replicates on all three, and the perplexity-guard evasion appears where the model admits in-guard low-rank headroom. We then give a data-free screen: a two-axis statistic of the compression error (coherent-fraction and error-rate) that flags the failing builds with fixed thresholds across architectures and matches the coherence-times-rate mechanism. Perplexity, MMLU, and fidelity acceptance do not certify agent safety. Screen gently-compressed low-rank builds before agentic deployment
Chinese Translation
从业者在压缩语言模型通过一系列数据廉价质量检测后接受该模型：其困惑度在原始模型的小范围内、下游准确性（例如 MMLU）在置信区间内，以及无数据输出保真度信号，这些信号比较压缩模型和原始网络在随机探测输入下的内部表示。然而，这一系列检测存在盲点。在三种模型家族中，温和压缩的模型通过了每个检测，然后在作为代理执行标准操作程序（SOP）时发明了从未在指令中的程序步骤。该效应是特定于操作员的：一致的低秩（SVD）截断会引发此效应，而与相同困惑度匹配的幅度剪枝则不会。有一个解离现象可以隔离原因。相同的压缩权重在配对输出保真度测试中获胜，但在发明步骤的金丝雀测试中失败。主导轴是压缩误差的一致性乘以其速率；损害的大小并不能预测这一点。无数据保真度探测器在构造上是保真度神谕，因此无法看到这一轴。我们通过在三种架构上进行预注册、具备功效的金丝雀测试，使用配对置信区间对盲点和解离现象进行表征。操作员特异性在所有三种架构上均可复制，而困惑度检测逃避出现在模型承认在检测范围内的低秩余地时。然后，我们提供一个无数据筛选：一个关于压缩误差的双轴统计（一致性分数和误差率），该统计在各架构中以固定阈值标记失败的构建，并与一致性乘以速率机制相匹配。困惑度、MMLU 和保真度接受度并不能证明代理的安全性。在代理部署前筛选温和压缩的低秩构建。
cs.CL / 44 / 2607.28212
Causal Discovery with Inverted Self-attention for Multivariate Time Series
基于反向自注意力的多元时间序列因果发现
Abstract
Causal discovery in multivariate time series data is challenging due to complex interactions, high dimensionality, and nonlinear dependencies among variables. Existing methods often struggle to capture these complexities, resulting in inaccurate causal structures. To address this issue, we propose a novel framework that leverages self-attention mechanisms within the transformer architecture for causal discovery. Our approach introduces a novel inverted causal self-attention mechanism (CSAM) that emphasizes latent and indirect causal relationships by inverting tokens and inducing sparsity in attention scores, focusing on significant causal interactions and reducing spurious correlations. Additionally, we develop a global causal algorithm to identify global causal links, providing a holistic metric for causal influence, along with a causal verification module to ensure robustness in the identified causal relationships, enhancing the reliability of our framework. Experiments on both linear and nonlinear datasets, along with ablation studies and sensitivity analyses, show that our framework outperforms existing methods, demonstrating its potential for causal discovery in complex multivariate time series.
Chinese Translation
多元时间序列数据中的因果发现面临复杂的交互、高维度和变量间非线性依赖等挑战。现有方法往往难以捕捉这些复杂性，导致因果结构不准确。为了解决这一问题，我们提出了一种新颖的框架，利用变压器架构中的自注意力机制进行因果发现。我们的方法引入了一种新颖的反向因果自注意力机制（CSAM），通过反转标记并诱导注意力分数的稀疏性，强调潜在和间接的因果关系，聚焦于显著的因果交互并减少虚假相关性。此外，我们开发了一种全局因果算法，以识别全局因果链接，提供因果影响的整体度量，并配备因果验证模块，以确保所识别因果关系的稳健性，从而增强我们框架的可靠性。在线性和非线性数据集上的实验，以及消融研究和敏感性分析表明，我们的框架优于现有方法，展示了其在复杂多元时间序列因果发现中的潜力。
cs.CL / 45 / 2607.28229
EMBL AI Librarian: Life-Sciences Knowledge Layer for AI Agents
EMBL AI 图书管理员：生命科学知识层为 AI 代理服务
Abstract
The web is increasingly accessed by AI agents rather than humans. Every agent needs knowledge, especially in the life-sciences, where agentic pipelines are growing fast. Access to the literature is a crucial part of that need, and resources such as Europe PMC, with over 40M indexed records, are widely used to meet it. Yet these resources were not built for AI agents: they take keywords and complex syntax and return whole papers, so every agent must learn the syntax, issue several searches, and read full papers to find the evidence it needs. We introduce EMBL AI Librarian, a knowledge layer that upgrades the Europe PMC interface for AI agents: an agent asks in natural language and receives evidence that answers it. A single LLM orchestrates the whole knowledge retrieval process: it plans complementary subqueries executed by the live Europe PMC search engine, then reads the selected papers and locates the relevant evidence. We evaluate Librarian across four benchmarks: literature synthesis, claim verification, open-domain question answering, and downstream biology tasks such as protocol questions and sequence manipulation. On ScholarQABench, Librarian improves Citation F1 by more than $16$ points over strong recently published baselines. Used as the retrieval layer of an existing claim-verification pipeline, it increases agreement with expert consensus; and on the open-form LitQA2 benchmark, a GPT-5.4 agent scores about $8$ points higher when grounded in Librarian than with web search. Overall, our results show that equipping life-science agents with the Librarian knowledge layer improves performance across a range of tasks. We release our code publicly at https://github.com/petroni-lab/librarian
Chinese Translation
网络越来越多地被 AI 代理而非人类访问。每个代理都需要知识，尤其是在生命科学领域，代理管道正在快速增长。获取文献是满足这一需求的关键部分，像 Europe PMC 这样的资源，拥有超过 4000 万条索引记录，被广泛用于满足这一需求。然而，这些资源并不是为 AI 代理构建的：它们接受关键词和复杂的语法，并返回整篇论文，因此每个代理必须学习语法，进行多次搜索，并阅读完整的论文以找到所需的证据。我们介绍了 EMBL AI 图书管理员，这是一个为 AI 代理升级 Europe PMC 接口的知识层：代理以自然语言提问，并获得回答其问题的证据。一个单一的 LLM 协调整个知识检索过程：它规划由实时 Europe PMC 搜索引擎执行的互补子查询，然后阅读选定的论文并定位相关证据。我们在四个基准上评估图书管理员：文献综合、声明验证、开放领域问答和下游生物学任务，如协议问题和序列操作。在 ScholarQABench 上，图书管理员的引用 F1 分数比最近发布的强基线提高了超过 16 分。作为现有声明验证管道的检索层，它增加了与专家共识的一致性；在开放式 LitQA2 基准上，当基于图书管理员时，GPT-5.4 代理的得分比使用网络搜索高出约 8 分。总体而言，我们的结果表明，为生命科学代理配备图书管理员知识层可以提高各种任务的性能。我们的代码已公开发布在 https://github.com/petroni-lab/librarian
cs.CL / 46 / 2607.28263
Understanding Is Done Early: A Depth Division of Labor in Large Language Models and Its Use for Unbounded-Context Memory
理解在早期完成：大型语言模型中的深度劳动分工及其在无界上下文记忆中的应用
Abstract
Transformer depth is not used uniformly: lower and middle layers build semantic representations, while upper layers increasingly specialize them for prediction. We turn this division of labor into CoMem (Comprehension Memory), which writes each context chunk only through an intermediate layer, retrieves a fixed number of cached residual states, and recomputes the query-conditioned upper layers over the resulting pack. For a fixed retrieval budget, model-side read compute and memory are independent of stored-context length. We evaluate a continued-trained Qwen3-8B base LM under a unified chat-template-free protocol. The backbone is frozen; the flagship trains only a rank-32 self-distillation LoRA on plain PG19, and we report an adapter-free arm separately. CoMem reaches 97.05 on RULER and 38.27 on LoCoMo versus 34.59 for full-context KV-Direct; the dialogue-memory advantage survives conversation-cluster resampling and an independent judge. Results on additional long-context and long-document tasks expose both the benefits of bounded retrieval and its in-window compression tax. Controlled depth sweeps show that deeper caching lowers per-query recomputation but incurs a fidelity loss that self-distillation substantially repairs. In a separate adapter-free efficiency control on an NVIDIA H20 at 128k, CoMem uses 18.26 GB rather than 89.36 GB and achieves a 7.83x prefill speedup. These results show that long-context memory can be organized along the layer axis, not only the token axis.
Chinese Translation
Transformer 的深度并非均匀使用：较低和中间层构建语义表示，而较高层则越来越专注于预测。我们将这种劳动分工转化为 CoMem（理解记忆），该方法仅通过中间层写入每个上下文块，检索固定数量的缓存残差状态，并在生成的包上重新计算查询条件的上层。对于固定的检索预算，模型端的读取计算和内存与存储上下文的长度无关。我们在统一的无聊天模板协议下评估了持续训练的 Qwen3-8B 基础语言模型。主干被冻结；旗舰模型仅在普通 PG19 上训练一个秩为 32 的自蒸馏 LoRA，并单独报告一个无适配器的分支。CoMem 在 RULER 上达到了 97.05，在 LoCoMo 上达到了 38.27，而全上下文 KV-Direct 的结果为 34.59；对话记忆优势在对话集群重采样和独立评审中得以保持。额外的长上下文和长文档任务的结果揭示了有界检索的好处及其窗口内压缩的代价。控制深度的实验表明，较深的缓存降低了每个查询的重新计算，但会导致保真度损失，而自蒸馏可以显著修复这一问题。在 NVIDIA H20 上进行的单独无适配器效率控制实验中，CoMem 使用了 18.26 GB 的内存，而不是 89.36 GB，并实现了 7.83 倍的预填充加速。这些结果表明，长上下文记忆可以沿着层轴而非仅沿着标记轴进行组织。
cs.CL / 47 / 2607.28274
MORFES: A Benchmark for Productive Inflectional Competence in Modern Greek
MORFES：现代希腊语的生产性屈折能力基准测试
Abstract
Modern Greek is a richly inflected language, yet the language models built for it are evaluated mainly on factual knowledge, and no benchmark is dedicated to their inflectional competence. We introduce MORFES (Morphological Open-class Recognition-and-Formation Evaluation Suite), a benchmark of 500 expert-verified items that tests the recognition and production of Greek inflected forms, favoring lower-frequency lemmas so that a correct answer reflects the rule rather than a memorized form. We make it publicly available at https://huggingface.co/datasets/KIEFERSA/MORFES. We evaluate a range of open language models on MORFES, situating them within the rapidly scaling open-weight ecosystem from LLaMA to Qwen3, DeepSeek-R1, Magistral, and Kimi K2, where multilingual coverage grows but grammatical competence in morphologically rich languages remains under-measured. Among them, Sophea-Genesis-1, a model we developed and release as open weights at https://huggingface.co/KIEFERSA/Sophea-Genesis-1, leads on inflectional morphology while matching similarly sized models in general capability.
Chinese Translation
现代希腊语是一种丰富的屈折语言，但为其构建的语言模型主要在事实知识上进行评估，且没有专门的基准测试其屈折能力。我们介绍了 MORFES（形态开放类识别与生成评估套件），这是一个包含500个专家验证项目的基准，旨在测试希腊语屈折形式的识别和生成，优先考虑低频词元，以便正确答案反映规则而非记忆形式。我们将其公开发布在 https://huggingface.co/datasets/KIEFERSA/MORFES。我们在 MORFES 上评估了一系列开放语言模型，将其置于快速扩展的开放权重生态系统中，包括 LLaMA、Qwen3、DeepSeek-R1、Magistral 和 Kimi K2，尽管多语言覆盖面不断扩大，但在形态丰富语言中的语法能力仍然被低估。在这些模型中，我们开发并以开放权重形式发布的 Sophea-Genesis-1 在屈折形态学方面表现突出，同时在一般能力上与同等规模的模型相匹配，开放权重可在 https://huggingface.co/KIEFERSA/Sophea-Genesis-1 获取。
cs.CL / 48 / 2607.28282
(Towards) Scalable Reliable Automated Evaluation with Large Language Models
面向可扩展的可靠自动评估与大型语言模型
Abstract
Evaluating the quality and relevance of textual outputs from Large Language Models (LLMs) remains challenging and resource-intensive. Existing automated metrics often fail to capture the complexity and variability inherent in LLM-generated outputs. Moreover, these metrics typically rely on explicit reference standards, limiting their use mostly to domains with objective benchmarks. This work introduces a novel evaluation framework designed to approximate expert-level assessments of LLM-generated content. The proposed method employs pairwise comparisons of outputs by multiple LLMs, reducing biases from individual models. An Elo rating system is used to generate stable and interpretable rankings. Adjustable agreement thresholds, from full unanimity to majority voting, allow flexible control over evaluation confidence and coverage. The method's effectiveness is demonstrated through evaluating competency profiles extracted from scientific abstracts. Preliminary results show that automatically derived rankings correlate well with expert judgments, significantly reducing the need for extensive human intervention. By offering a scalable, consistent, and domain-agnostic evaluation layer, the framework supports more efficient and reliable quality assessments of LLM outputs across diverse applications.
Chinese Translation
评估大型语言模型（LLMs）生成的文本输出的质量和相关性仍然具有挑战性且资源密集。现有的自动化指标往往无法捕捉LLM生成输出中固有的复杂性和多样性。此外，这些指标通常依赖于明确的参考标准，这限制了它们的使用主要集中在具有客观基准的领域。本研究提出了一种新颖的评估框架，旨在近似专家级的LLM生成内容评估。所提出的方法通过多个LLM对输出进行成对比较，减少了个别模型的偏差。采用Elo评分系统生成稳定且可解释的排名。可调的协议阈值，从完全一致到多数投票，允许灵活控制评估的信心和覆盖范围。通过评估从科学摘要中提取的能力概况，展示了该方法的有效性。初步结果表明，自动生成的排名与专家判断高度相关，显著减少了对大量人工干预的需求。通过提供一个可扩展、一致且与领域无关的评估层，该框架支持对LLM输出在多种应用中进行更高效和可靠的质量评估。
cs.CL / 49 / 2607.28292
CACHE-UK: A Stability-Aware Memory Editor for Sequentially Updated Quantized LLMs in Finance
CACHE-UK：一种针对金融领域顺序更新量化大语言模型的稳定性感知内存编辑器
Abstract
Large Language Models (LLMs) deployed in dynamic financial environments face a critical challenge: maintaining factual accuracy as market conditions, regulations, and corporate facts change continuously. While 4-bit quantization enables efficient deployment, it severely limits the viability of sequential memory editing: existing methods undergo catastrophic performance degradation under this "quantization stability crisis." We introduce CACHE-UK (Contextual Adaptive Continual Hybrid Editor for UK Finance), a stability-aware memory editing framework specifically designed for domain-specific, quantized LLMs. CACHE-UK integrates three components: a rank-1 LoRA perturbation mechanism that confines edits to the low-rank adapter subspace, a financial domain prioritization module for content-adaptive edit strength, and a closed-loop Stability Controller that tracks "degradation debt" to prevent catastrophic forgetting across sequential updates. Evaluated on a 4-bit quantized OpenLLaMA-3B model with a curated UK financial corpus of 88,021 documents, CACHE-UK reduces knowledge degradation by 11-17% relative to adapted baselines under identical 4-bit constraints -- its most robust effect -- while attaining the highest test success (generalization) rate observed in our setting (28%, a 6 percentage point improvement over the strongest adapted baseline). These results indicate that stability-aware editing can improve factual maintenance in resource-constrained financial LLM deployments, though absolute generalization rates remain low.
Chinese Translation
在动态金融环境中部署的大语言模型（LLMs）面临着一个关键挑战：随着市场条件、法规和公司事实的持续变化，保持事实准确性。虽然4位量化可以实现高效部署，但它严重限制了顺序内存编辑的可行性：现有方法在这种“量化稳定性危机”下会出现灾难性的性能下降。我们提出了CACHE-UK（面向英国金融的上下文自适应持续混合编辑器），这是一个专门为领域特定的量化LLMs设计的稳定性感知内存编辑框架。CACHE-UK集成了三个组件：一个将编辑限制在低秩适配器子空间的秩-1 LoRA扰动机制，一个用于内容自适应编辑强度的金融领域优先级模块，以及一个跟踪“退化债务”的闭环稳定性控制器，以防止在顺序更新中出现灾难性遗忘。在对一个经过精心策划的包含88,021份文档的英国金融语料库的4位量化OpenLLaMA-3B模型进行评估时，CACHE-UK在相同的4位约束下，相较于适应基线减少了11-17%的知识退化——这是其最显著的效果——同时在我们的设置中达到了最高的测试成功（泛化）率（28%，比最强适应基线提高了6个百分点）。这些结果表明，稳定性感知编辑可以改善资源受限的金融LLM部署中的事实维护，尽管绝对泛化率仍然较低。
cs.CL / 50 / 2607.28319
Fairness Pruning: Locating Demographic Bias in GLU-MLP Layers via Differential Activations
公平修剪：通过差异激活定位 GLU-MLP 层中的人口偏见
Abstract
This work presents Fairness Pruning, a lightweight structural intervention method designed for the management and future mitigation of demographic bias in large language models (LLMs). As a foundational empirical validation of this method, this work focuses on causal bias localization. Using minimally contrastive prompt pairs and inference-time activation capture, the method identifies neurons that react differentially when processing demographic attributes in GLU architectures, evaluating the signal at the down_proj input. Empirical evaluation was conducted on models of up to 3 billion parameters (Llama-3.2 family and Salamandra-2B), combining standardized benchmark evaluation with qualitative text generation experiments. Results demonstrate that zeroing the identified neurons alters how the model responds to associated demographic variables. However, rather than producing flat mitigation, the intervention causes bidirectional bias destabilization: because BiasScore is unsigned, candidate sets mix neurons that push toward and against the stereotype, and the net effect on aggregate bias depends on which sign dominates. The intervention is extremely surgical: zeroing at most 40 neurons in Llama-3.2-1B (less than 0.031% of total MLP width) achieves a mean retention of 99.49% in reasoning and general knowledge capabilities. These findings empirically confirm that demographic bias processing and model capabilities operate on dissociable circuits, establishing the methodological foundations for transitioning from blind zeroing toward directional behavior modulation.
Chinese Translation
本研究提出了公平修剪（Fairness Pruning），这是一种轻量级结构干预方法，旨在管理和未来减轻大型语言模型（LLMs）中的人口偏见。作为该方法的基础实证验证，本研究聚焦于因果偏见定位。通过使用最小对比提示对和推理时激活捕获，该方法识别在处理 GLU 架构中的人口属性时表现出差异反应的神经元，并评估下投影（down_proj）输入的信号。对参数高达 30 亿的模型（Llama-3.2 家族和 Salamandra-2B）进行了实证评估，结合了标准化基准评估和定性文本生成实验。结果表明，零化识别出的神经元改变了模型对相关人口变量的响应。然而，这种干预并没有产生平坦的减轻效果，而是导致双向偏见不稳定：由于 BiasScore 是无符号的，候选集合混合了推动和反对刻板印象的神经元，聚合偏见的净效应取决于哪个符号占主导地位。这种干预极为精确：在 Llama-3.2-1B 中最多零化 40 个神经元（占总 MLP 宽度的不到 0.031%）即可实现推理和一般知识能力的平均保留率为 99.49%。这些发现实证确认了人口偏见处理和模型能力在可分离电路上运行，为从盲目零化转向定向行为调节奠定了方法论基础。
cs.CL / 51 / 2607.28347
LLMs struggle to simulate human belief updates in controlled environments
大型语言模型在受控环境中模拟人类信念更新的困难
Abstract
LLMs are increasingly deployed as proxies for human study participants in social science experiments, yet the fidelity of this practice has rarely been tested directly. We test whether six LLMs can simulate individual human belief updates, comparing LLM outputs 1-to-1 against ground truth data from 391 UK participants on Prolific, who updated their stances on three discussion topics after reading Reddit comments. Each participant was simulated by an LLM conditioned on a persona derived from their demographic and personality trait data. We find that some LLMs (Qwen3-32B and GPT-5-Mini) can match the human post-stance distribution, but only when given participants' actual initial stances. All six models fail to simulate initial stances themselves and to produce faithful belief updates from self-generated stances. Three systematic biases emerge across all models: overrepresentation of neutral positions, more frequent but smaller belief shifts than humans, and a failure to rank comments by convincingness. Demographic and personality trait personas had no consistent effect on fidelity. LLM simulations of human belief dynamics are only reliable when grounded in realistic starting conditions, that current multi-round social media simulations rarely provide.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作社会科学实验中人类研究参与者的代理，然而这一做法的准确性却很少得到直接测试。我们测试了六种大型语言模型是否能够模拟个体人类的信念更新，将LLM的输出与来自391名英国参与者在Prolific平台上的真实数据进行逐一比较，这些参与者在阅读Reddit评论后更新了他们对三个讨论主题的立场。每位参与者都由一个基于其人口统计和个性特征数据衍生的人物角色条件化的LLM进行模拟。我们发现一些大型语言模型（如Qwen3-32B和GPT-5-Mini）能够匹配人类的后立场分布，但仅在提供参与者的实际初始立场时才有效。所有六种模型都无法独立模拟初始立场，也无法从自生成的立场中产生真实的信念更新。所有模型中出现了三种系统性偏差：中立立场的过度代表性、信念变化频率较高但幅度较小，以及未能按说服力对评论进行排序。人口统计和个性特征的人物角色对准确性没有一致的影响。大型语言模型对人类信念动态的模拟只有在基于现实的起始条件时才可靠，而当前的多轮社交媒体模拟很少提供这样的条件。
cs.CL / 52 / 2607.28359
Correlation between prosody and pragmatics: A case study of the discourse marker h\=al\=a `now' in Persian
韵律与语用的相关性：波斯语话语标记 halā '现在' 的案例研究
Abstract
The Persian discourse marker h\=al\=a ('now') exhibits remarkable multifunctionality, extending far beyond its temporal adverbial role to encompass a variety of pragmatic functions. This study presents a pragmatic and acoustic analysis of h\=al\=a in spoken Persian, examining 267 instances from spontaneous conversations. While temporal uses were present, they were often combined with other discourse marker functions, indicating extensive multifunctionality, with 70% of tokens serving two or more pragmatic roles. Textual functions (topic shifting, signaling relationships, boundary marking, attention guidance, topic introduction, and topic emphasis) were most frequent, followed by interactive functions (turn management, listener engagement, and feedback regulation), and modal functions (epistemic stance, emotional expression, and attitudinal marking). Prosodic analysis revealed that duration and intensity are key cues for distinguishing h\=al\=a's functions. Textual uses were significantly shorter, while temporal uses showed a tendency toward longer realizations. Interactive functions correlated with higher intensity, while modal functions showed a weaker tendency toward lower intensity. These findings indicate that duration and intensity are the main prosodic cues associated with functional differentiation in h\=al\=a, especially in textual and interactive uses.
Chinese Translation
波斯语话语标记 halā ('现在') 展现出显著的多功能性，远超其时间副词的角色，涵盖多种语用功能。本研究对口语波斯语中的 halā 进行了语用和声学分析，考察了267个来自自发对话的实例。尽管存在时间用法，但它们通常与其他话语标记功能相结合，表明其广泛的多功能性，其中70%的实例承担了两个或更多的语用角色。文本功能（主题转换、信号关系、边界标记、注意引导、主题引入和主题强调）最为频繁，其次是互动功能（轮次管理、听众参与和反馈调节），以及模态功能（认知立场、情感表达和态度标记）。韵律分析显示，持续时间和强度是区分 halā 功能的关键线索。文本用法显著较短，而时间用法则倾向于较长的实现。互动功能与更高的强度相关，而模态功能则表现出较低强度的较弱倾向。这些发现表明，持续时间和强度是与 halā 功能区分相关的主要韵律线索，尤其是在文本和互动用法中。
cs.CL / 53 / 2607.28439
Beyond a Single Judge: Simulating Social Persona Panels for Generative UI Evaluation
超越单一评审者：模拟社会角色小组进行生成用户界面评估
Abstract
Generative UI (GenUI) lets large language models synthesize a complete, renderable interface directly from a natural-language instruction, but evaluating the quality of what they generate remains an open problem. Human evaluation is costly and rater-variant, while LLM-as-a-judge is scalable but reflects only a single implicit viewpoint, unable to capture how different populations of real users actually perceive the same interface. We propose the Evidence-Grounded, Social-Weighted Persona Panel (ESPP), a three-stage GenUI evaluation method in which a panel of psychologically diverse, evidence-grounded personas independently rates a screenshot, exchanges opinions under a trait-derived, semantically-gated bounded-confidence mechanism, and is aggregated via Delphi-inspired social weighting into a single judgment. ESPP tracks human judgment substantially more closely than a naive single-pass judge, raising Pearson $r$ from $0.716$ to $0.922$, and a prompt-ensemble control recovers only about a third of this gap, isolating genuine persona and evidence grounding as the dominant source of improvement. Beyond this fidelity gain, retaining each panelist's individual rating further reveals that user subgroups agree on overall model rankings yet diverge sharply on specific rating dimensions, a structural disagreement a single homogeneous judge would systematically erase. The codes are available at https://github.com/Wuzheng02/ESPP.
Chinese Translation
生成用户界面（GenUI）使大型语言模型能够直接从自然语言指令合成完整的可渲染界面，但评估其生成内容的质量仍然是一个未解决的问题。人工评估成本高且评审者之间存在差异，而将大型语言模型作为评审者虽然具有可扩展性，但仅反映单一隐含观点，无法捕捉不同真实用户群体对同一界面的实际感知。我们提出了基于证据的社会加权角色小组（ESPP），这是一种三阶段的GenUI评估方法，其中由心理多样化、基于证据的角色组成的小组独立对屏幕截图进行评分，在特征导向、语义限制的有限置信机制下交换意见，并通过受德尔菲法启发的社会加权汇总为单一判断。ESPP的评估结果与人类判断的接近程度显著高于简单的单次评审，Pearson相关系数从0.716提高到0.922，而提示集成控制仅恢复了约三分之一的差距，表明真实角色和证据基础是改善的主要来源。除了这种保真度的提升外，保留每位小组成员的个体评分进一步揭示用户子群体在整体模型排名上达成一致，但在特定评分维度上却存在明显分歧，这种结构性不一致将被单一同质评审者系统性抹去。代码可在 https://github.com/Wuzheng02/ESPP 获取。
cs.CL / 54 / 2607.28449
Lightning OPD 2.0: Mitigating Style Bias in Cross-Teacher On-Policy Distillation for Large Reasoning Models
Lightning OPD 2.0：缓解大规模推理模型中的跨教师在线蒸馏风格偏差
Abstract
On-policy distillation (OPD) provides dense token-level supervision from a teacher, but its effectiveness can depend on teacher consistency, meaning that the model providing OPD supervision should also have generated the demonstrations used to train the supervised fine-tuning (SFT) reference. However, this condition is frequently violated in practice when SFT data have mixed or unknown provenance or when different models are preferred for SFT data generation and subsequent distillation. In such cross-teacher settings, even a stronger OPD teacher can yield little improvement over the SFT reference. We find that raw teacher--reference disagreement contains potentially useful context-specific teacher evidence as well as a recurring component associated with differences in wording, formatting, and reasoning cadence. We introduce Lightning OPD 2.0 with cross-fitted style residualization, which uses rollout-level cross-fitting to estimate this recurring component as an operational proxy for style-token bias and subtracts it before constructing the token-level OPD update. Across mathematical reasoning and code generation benchmarks, Lightning OPD 2.0 consistently outperforms Lightning OPD in cross-teacher settings. Starting from Klear-Reasoner-8B-SFT, Lightning OPD 2.0 reaches 82.4% on AIME 2024 and 63.0% on LiveCodeBench v5. Together, these results establish Lightning OPD 2.0 as a practical approach to cross-teacher OPD, relaxing teacher consistency as a prerequisite and allowing the SFT data generator and distillation teacher to be selected independently. Code will be released soon.
Chinese Translation
在线蒸馏（On-policy distillation, OPD）提供来自教师的密集令牌级监督，但其有效性可能依赖于教师的一致性，这意味着提供 OPD 监督的模型也应该生成用于训练监督微调（Supervised Fine-Tuning, SFT）参考的示例。然而，当 SFT 数据具有混合或未知来源，或者在 SFT 数据生成和后续蒸馏中偏好不同模型时，这一条件在实践中常常被违反。在这样的跨教师环境中，即使是更强的 OPD 教师也可能对 SFT 参考的改进甚微。我们发现，原始教师与参考之间的分歧包含潜在有用的上下文特定教师证据，以及与措辞、格式和推理节奏差异相关的重复成分。我们引入了 Lightning OPD 2.0，采用交叉拟合风格残差化（cross-fitted style residualization），利用回滚级别的交叉拟合来估计这一重复成分，作为风格令牌偏差的操作性代理，并在构建令牌级 OPD 更新之前将其减去。在数学推理和代码生成基准测试中，Lightning OPD 2.0 在跨教师环境中始终优于 Lightning OPD。从 Klear-Reasoner-8B-SFT 开始，Lightning OPD 2.0 在 AIME 2024 上达到了 82.4%，在 LiveCodeBench v5 上达到了 63.0%。这些结果共同确立了 Lightning OPD 2.0 作为一种实用的跨教师 OPD 方法，放宽了教师一致性作为前提条件，并允许独立选择 SFT 数据生成器和蒸馏教师。代码将很快发布。
cs.CL / 55 / 2607.28476
Improving Mental Health Screening and Early Risk Detection in Spanish
改善西班牙语心理健康筛查与早期风险检测
Abstract
Early detection of mental health disorders is often limited by the lack of specialized resources in Spanish and the difficulty of analyzing long histories of social media posts. This paper addresses these challenges through three main contributions. First, we introduce three Spanish foundational models specifically adapted to the mental health domain through domain-specific pre-training. Second, we propose Incremental Context Expansion (ICE), an automatic relabeling methodology designed for early detection. ICE identifies the point at which cumulative messages provide enough evidence of a disorder, generating more informative training samples. Third, we provide a set of fine-tuned models using the samples generated with the ICE methodology for early risk detection tasks. Our results on three Spanish benchmarks show that combining these specialized models with ICE improves the state-of-the-art, reducing detection latency while maintaining high performance. All models are publicly available.
Chinese Translation
心理健康障碍的早期检测常常受到缺乏西班牙语专业资源以及分析社交媒体长篇历史帖文的困难所限制。本文通过三个主要贡献来应对这些挑战。首先，我们介绍了三种专门适应于心理健康领域的西班牙语基础模型，这些模型经过领域特定的预训练。其次，我们提出了增量上下文扩展（Incremental Context Expansion, ICE）这一自动重新标注方法，旨在实现早期检测。ICE能够识别累积信息达到足够证据证明存在障碍的时刻，从而生成更具信息量的训练样本。第三，我们提供了一组使用ICE方法生成样本的微调模型，用于早期风险检测任务。我们在三个西班牙语基准测试上的结果表明，将这些专业模型与ICE结合使用，能够提高当前的技术水平，减少检测延迟，同时保持高性能。所有模型均已公开。
cs.CL / 56 / 2607.28478
Would You Walk to the Car Wash? Revealing the Salience Bias of Large Language Models in Commonsense Reasoning
你会走到洗车店吗？揭示大型语言模型在常识推理中的显著性偏差
Abstract
As large language models (LLMs) continue to advance in complex reasoning tasks, they have learned to heavily prioritize explicit conditions provided in the input. However, in everyday commonsense reasoning, this mechanism exposes a critical vulnerability which we term Salience Bias: models become easily hijacked by useless explicit distractors (e.g., numerical values), leading them to ignore the implicit physical or commonsense prerequisites of a task. A critical open question is whether this failure reflects a genuine gap in commonsense knowledge or merely its suppression under misleading task framing. To investigate this, we construct the SaliTrap Benchmark, a high-quality dataset across four trap dimensions. Evaluating 12 state-of-the-art LLMs, we find that all mainstream models suffer significantly from salience bias, with severity scaling with distractor density and detecting the trap often decoupled from actually avoiding it. Crucially, by re-eliciting the same models with the task framing stripped away, we show that this is overwhelmingly a failure of \textbf{knowledge suppression rather than knowledge absence}: a context-free knowledge probe alone recovers over 90\% of sycophantic-compliance failures, revealing that the requisite commonsense is intrinsically present but actively crowded out by salient distractors that lure the model into over-compliant, unnecessary computation. Building on this diagnosis, we further show that lightweight, inference-time prompting alone substantially closes the gap without any retraining. Our findings relocate the bottleneck of commonsense reasoning failures from model competence to elicitation, and we release SaliTrap as a testbed for this blind spot. The codes are available at https://github.com/Wuzheng02/SaliTrap.
Chinese Translation
随着大型语言模型（LLMs）在复杂推理任务中的不断进步，它们已学会在输入中高度优先考虑显式条件。然而，在日常常识推理中，这一机制暴露出一个关键的脆弱性，我们称之为显著性偏差：模型容易受到无用的显式干扰因素（例如，数值）的劫持，从而忽视任务的隐含物理或常识前提。一个关键的开放性问题是，这种失败是否反映了常识知识的真实缺口，还是仅仅是在误导性任务框架下的抑制。为此，我们构建了 SaliTrap 基准，一个涵盖四个陷阱维度的高质量数据集。通过评估 12 个最先进的 LLM，我们发现所有主流模型都显著受到显著性偏差的影响，其严重程度与干扰因素的密度成正比，而检测陷阱往往与实际避免陷阱脱钩。关键的是，通过去除任务框架重新引导相同的模型，我们表明这主要是知识抑制而非知识缺失的失败：仅通过无上下文的知识探测，就能恢复超过 90% 的谄媚顺从失败，揭示所需的常识本质上是存在的，但被显著的干扰因素主动挤出，这些干扰因素诱使模型进行过度顺从和不必要的计算。在此诊断的基础上，我们进一步表明，仅通过轻量级的推理时提示就能显著缩小这一差距，而无需任何重新训练。我们的发现将常识推理失败的瓶颈从模型能力转移到引导上，并发布 SaliTrap 作为这一盲点的测试平台。代码可在 https://github.com/Wuzheng02/SaliTrap 获取。
cs.CL / 57 / 2607.28496
Beyond Sentiment: Structured Information Extraction from Financial News
超越情感：从金融新闻中提取结构化信息
Abstract
Financial sentiment analysis has become a standard component in news-driven stock prediction, yet it reduces rich, multi-dimensional news articles to a single polarity score. We hypothesize that financial news encodes multiple orthogonal information dimensions---event type, impact scope, temporal horizon, and semantic confidence---that sentiment alone cannot capture, and that these dimensions carry independent predictive value. To test this hypothesis, we propose a structured information extraction framework that leverages LLaMA-3.1-70B to extract six semantic dimensions from financial news. Through large-scale experiments on 41,618 news--stock pairs from the FNSPID dataset, we find that (i) FinBERT sentiment features exhibit strong predictive power under nonlinear models (F1=0.576) but substantially weaker performance under linear models (F1=0.230), revealing a highly nonlinear sentiment--return relationship; (ii) LLM-extracted structured features, while individually weaker, capture information orthogonal to sentiment, as evidenced by a 53.5% systematic disagreement rate between the two approaches; and (iii) combining both signal sources yields F1=0.600, significantly outperforming either alone ($p < 0.0001$), with consistent improvements across all seven event types. Ablation experiments confirm that non-sentiment structural dimensions (event type, impact subject, time horizon, confidence) independently contribute $\Delta\text{F1} = +0.019$ beyond FinBERT alone. Feature importance analysis reveals balanced contributions from all six extracted dimensions (14--21%), demonstrating that compressing news into a single sentiment score incurs substantial information loss. Our results suggest that the sentiment--semantics decoupling in financial text is systematic and exploitable, opening a new direction for multi-dimensional financial NLP.
Chinese Translation
金融情感分析已成为基于新闻的股票预测中的标准组成部分，但它将丰富的多维新闻文章简化为单一的极性评分。我们假设金融新闻编码了多个正交信息维度——事件类型、影响范围、时间范围和语义信心——这些维度仅凭情感无法捕捉，且这些维度具有独立的预测价值。为了验证这一假设，我们提出了一个结构化信息提取框架，利用 LLaMA-3.1-70B 从金融新闻中提取六个语义维度。通过对 FNSPID 数据集中 41,618 对新闻-股票的广泛实验，我们发现：(i) FinBERT 情感特征在非线性模型下表现出强大的预测能力（F1=0.576），但在线性模型下表现明显较弱（F1=0.230），揭示了高度非线性的情感-回报关系；(ii) LLM 提取的结构化特征虽然单独较弱，但捕捉到与情感正交的信息，两个方法之间的系统性不一致率达 53.5%；(iii) 结合这两种信号源的结果为 F1=0.600，显著优于单独使用任何一种（$p < 0.0001$），并在所有七种事件类型中均表现出一致的改进。消融实验确认，非情感结构维度（事件类型、影响对象、时间范围、信心）独立贡献了 $ ext{F1} 的增量 +0.019$，超出单独使用 FinBERT 的效果。特征重要性分析显示，所有六个提取维度均衡贡献（14-21%），表明将新闻压缩为单一情感评分会导致显著的信息损失。我们的结果表明，金融文本中的情感-语义解耦是系统性的且可利用的，为多维金融自然语言处理开辟了新的方向。
cs.CL / 58 / 2607.28505
Generative AI and linguistic diversity in academic writing and publishing: Perspectives from World Englishes
生成性人工智能与学术写作和出版中的语言多样性：来自世界英语的视角
Abstract
The rise of generative artificial intelligence (GenAI) in academic writing and publishing (AWP) raises questions about linguistic inclusivity and the legitimacy of diverse Englishes in global scholarly communication. This article responds to these questions through a structured scholarly dialogue involving five sociolinguists from World Englishes and adjacent fields. Organised around five guiding questions, the dialogue interrogates how GenAI tools influence writing practices, reinforce or disrupt dominant language norms, and raise ethical challenges. Contributors reflect on the potential of GenAI to democratise writing processes while also raising concerns about GenAI's tendency to marginalise minoritised varieties and flatten nuance in scholarly writing. Across the dialogue, themes of linguistic (in)justice, researcher agency, and institutional responsibility emerge, with contributors calling for equity-informed policies, critical AI literacy, and inclusive co-design in GenAI development. The article shows the value of dialogic reflection in understanding GenAI's role in AWP. It concludes that while GenAI may reinforce existing hierarchies, it can also serve as a site of resistance, depending on how it is designed, governed and used within scholarly communities committed to linguistic diversity.
Chinese Translation
生成性人工智能（GenAI）在学术写作和出版（AWP）中的兴起引发了关于语言包容性和全球学术交流中多样英语合法性的质疑。本文通过五位来自世界英语及相关领域的社会语言学者之间的结构化学术对话，对这些问题作出回应。对话围绕五个指导性问题展开，探讨了GenAI工具如何影响写作实践、强化或破坏主导语言规范，并提出伦理挑战。参与者反思了GenAI在民主化写作过程中的潜力，同时也对其边缘化少数语言变体和简化学术写作细微差别的倾向表示担忧。在对话中，语言（不）公正、研究者的能动性和机构责任等主题浮现，参与者呼吁制定以公平为导向的政策、批判性的人工智能素养以及在GenAI开发中的包容性共同设计。本文展示了对话反思在理解GenAI在AWP中角色的价值。最后得出结论，尽管GenAI可能强化现有的等级制度，但它也可以成为抵抗的场所，这取决于其在致力于语言多样性的学术社区中如何设计、管理和使用。
cs.CL / 59 / 2607.28513
Creative Transformation in Literary Texts: Modelling Change Across Representational Levels
文学文本中的创造性转化：跨表现层次建模变化
Abstract
Creativity is often framed as the production of novelty, yet many cultural works emerge through transformation of earlier artifacts and not through isolated invention. Drawing on theories of imitation by Gabriel Tarde and James Mark Baldwin, this paper models creativity as selective transformation across multiple levels of textual representation. We introduce a multi-level framework that compares literary texts across lexical, semantic, conceptual, structural, and narrative dimensions using directional alignment and control calibrated similarity measures. Applying the model to historically documented literary relationships, we show that different pairs preserve source structure at different representational levels while diverging in others. These transformation profiles provide a quantitative method for characterizing how imitation persists and where creative divergence occurs within literary works.
Chinese Translation
创造性常被视为新奇的产生，然而许多文化作品是通过对早期文物的转化而非孤立的发明而产生的。本文借鉴了加布里埃尔·塔尔德（Gabriel Tarde）和詹姆斯·马克·鲍德温（James Mark Baldwin）的模仿理论，将创造性建模为跨多个文本表现层次的选择性转化。我们提出了一个多层框架，通过方向性对齐和控制校准相似性度量，比较词汇、语义、概念、结构和叙事维度上的文学文本。将该模型应用于历史记录的文学关系，我们展示了不同的文本对在不同表现层次上保留源结构，同时在其他层次上出现分歧。这些转化特征提供了一种定量方法，用于描述模仿如何持续存在以及在文学作品中创造性分歧发生的地方。
cs.CL / 60 / 2607.28528
AI systems and the reproduction of (standard) language ideologies in World Englishes
人工智能系统与世界英语中的（标准）语言意识形态的再生产
Abstract
The rapid growth of large language models (LLMs) has resurrected age-old questions in sociolinguistics and world Englishes, such as who decides what counts as legitimate English, whose English is suspect etc. This paper examines how AI systems, their uses and discourse on them reflect, reinforce, and occasionally challenge (standard) language ideologies, which privilege Inner Circle norms and marginalize non-dominant Englishes. Drawing on evidence from empirical studies, media commentary, social media debates, and examples from AI outputs, the paper shows that AI technologies reproduce dominant language ideologies at different levels: training data, design protocols, evaluation benchmarks, user feedback and public commentary. The analysis uses the public controversy over AI-sounding language, especially the fixation on the word delve, to illustrate how speakers of English from the Global North police the English language norms of Global South English users. The paper also identifies what Christian Mair has called a "standardisation paradox": AI may homogenize English by privileging standard forms and at the same time pluralize Englishes through exposure to wide-ranging corpora and annotation work carried out by Global South users. In doing so, the paper argues that generative AI is reigniting long-standing debates in World Englishes about standardization, legitimacy, and the ownership of English, now playing out in algorithmic systems, model training, evaluation practices, and public discourse, where non-dominant Englishes are increasingly conflated with AI-generated speech. Discussing AI systems as a site where language ideologies are (re)produced, the paper argues for more inclusive design approaches that recognize the plurality of Englishes in order to address the real-world negative consequences of treating some as more legitimate than others.
Chinese Translation
大型语言模型（LLMs）的快速发展重新引发了社会语言学和世界英语领域中的古老问题，例如：谁来决定什么算作合法的英语，谁的英语是可疑的等。本文考察了人工智能系统及其使用和相关话语如何反映、强化并偶尔挑战（标准）语言意识形态，这些意识形态优先考虑内圈（Inner Circle）的规范，并边缘化非主流英语。通过实证研究、媒体评论、社交媒体辩论以及人工智能输出示例的证据，本文表明，人工智能技术在不同层面上再生产了主导语言意识形态：训练数据、设计协议、评估基准、用户反馈和公众评论。分析利用了关于人工智能生成语言的公众争议，特别是对“delve”一词的执着，来说明全球北方的英语使用者如何监控全球南方英语用户的英语语言规范。本文还识别出Christian Mair所称的“标准化悖论”：人工智能可能通过优先考虑标准形式来同质化英语，同时通过接触广泛的语料库和全球南方用户进行的注释工作来使英语多元化。通过这一过程，本文认为生成性人工智能正在重新点燃世界英语中关于标准化、合法性和英语所有权的长期辩论，这些辩论现在在算法系统、模型训练、评估实践和公众话语中展开，其中非主流英语越来越被与人工智能生成的语言混淆。讨论人工智能系统作为语言意识形态（再）生产的场所，本文主张采取更具包容性的设计方法，以承认英语的多样性，从而应对将某些英语视为比其他英语更合法所带来的现实负面后果。
cs.CL / 61 / 2607.28545
ORCA-bench: How Ready Are Language Model Agents for Oncall?
ORCA-bench：语言模型代理在值班中准备得如何？
Abstract
Large language models can write, patch, and search code, but oncall root cause analysis (RCA) demands something different: reasoning over noisy metrics, logs, traces, and source code, starting from ambiguous user-facing reports, often hours after the incident began. We introduce ORCA-bench, a benchmark that puts general-purpose coding agents in a production-fidelity oncall setting. ORCA-bench pairs a live OpenTelemetry-instrumented microservice system--exposing six days of metrics, logs, and traces through real telemetry interfaces (Prometheus, Jaeger, and OpenSearch via Grafana) and full source-code access--with 1,079 RCA tasks that systematically vary report specificity, time-to-detection, and co-occurring fault scenarios. Ground-truth symptoms are curated and signed off by expert SREs, and our LLM-as-judge is independently re-scored by humans (Cohen's $\kappa_w=0.90$). Across five frontier agents, the best RCA Accuracy is 25.3% on Medium-difficulty tasks (the realistic-input setting) and 10.0% on Hard--a gap that remains even with Claude Fable 5. The weakest model hallucinates an implausible root cause in 40% of incident reports, and removing source-code access degrades every metric. Crucially, these are performances on a curated 50 GB / six-day testbed with tasks investigated in isolation on a system whose code and instrumentation are public. Since real production systems are order of magnitudes larger, more dynamic, and more idiosyncratic, the gap we report is a lower bound on the engineering investment required before frontier coding agents can be safely entrusted with production reliability. We release the public set at https://hub.harborframework.com/datasets/orca-bench/ORCA-bench.
Chinese Translation
大型语言模型可以编写、修补和搜索代码，但值班根本原因分析（RCA）需要不同的能力：在嘈杂的指标、日志、跟踪和源代码上进行推理，从模糊的用户报告开始，通常是在事件发生数小时后。我们引入了ORCA-bench，这是一个将通用编码代理置于生产真实度值班环境中的基准测试。ORCA-bench将一个实时的OpenTelemetry仪器化微服务系统与1,079个RCA任务配对，这些任务系统地变化报告的具体性、检测时间和共存故障场景。真实的症状由专家SRE进行策划并签字确认，而我们的LLM作为评判者的评分则由人类独立重新评分（Cohen's $ ext{kappa}_w=0.90$）。在五个前沿代理中，最佳RCA准确率在中等难度任务（现实输入设置）上为25.3%，在困难任务上为10.0%——即使在Claude Fable 5上，这一差距依然存在。最弱的模型在40%的事件报告中虚构了不可信的根本原因，而移除源代码访问会降低每个指标。至关重要的是，这些表现是在一个策划的50 GB / 六天的测试平台上进行的，任务在一个其代码和仪器化是公开的系统中孤立调查。由于真实的生产系统规模大得多、动态性更强且更具特异性，我们报告的差距是前沿编码代理在被安全地信任于生产可靠性之前所需的工程投资的下限。我们在https://hub.harborframework.com/datasets/orca-bench/ORCA-bench发布了公共数据集。
cs.CL / 62 / 2607.28568
Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering
Frontis-MA1：训练AI4AI模型以实现机器学习工程中的递归自我改进
Yang, Junlin, Jiang, Che, Fu, Yu, Luo, Tianwei, Ren, Can, Wang, Weizhi, Zhao, Kaikai, Liu, Hongyi, Zuo, Yuxin, Wang, Yuru, Fan, Yuchen, Tian, Kai, Yuan, Zhenzhao, Lin, Xiaojian, Sheng, Li, Qiang, Rushi, Jia, Guoli, Lv, Xingtai, Hua, Ermo, Lei, Dianqiao, Sun, Youbang, Ding, Ning, Zhou, Bowen, Zhang, Kaiyan
Abstract
Recursive self-improvement (RSI) requires AI systems that improve the process of building AI (i.e., AI4AI); machine learning engineering (MLE) offers a concrete, executable testbed for studying this capability. We introduce OpenMLE, an open full-stack system for RSI research in MLE, spanning verifiable task environments with execution feedback (OpenMLE-Gym), operator learning (OpenMLE-RL), and long-horizon search (OpenMLE-Evo). On this stack we post-train Frontis-MA1 (35B) as a meta-evolution agent for MLE, aligning post-training and inference around four atomic program-evolution operators (Draft, Improve, Debug, Crossover): the same operators are trained via execution-grounded SFT and RL on data deduplicated against all evaluation benchmarks, then composed into long-horizon search, coupling learning and evolution in a single loop. On MLE-Bench Lite under a 12-hour per-task budget on one RTX 4090 capped at 12 GB VRAM, Frontis-MA1 (35B) improves Medal Average from 39.39% to 60.61% over its base model with OpenMLE-Evo, and reaches 71.21% with OpenMLE-Evo-Max (benchmark-independent experience priors and asynchronous search), exceeding GPT-5.5 + Codex and approaching GPT-5.6 Sol and the 2.8T Kimi K3. On held-out NatureBench Lite, both components transfer: with the framework fixed, swapping in the trained model raises Match-SOTA from 50% to 70%; with the model fixed, swapping in OpenMLE-Evo raises it from 20% to 50%. We release the model weights and the full OpenMLE stack to enable reproducible research on executable AI4AI toward RSI. Code: https://github.com/FrontisAI/OpenRSI
Chinese Translation
递归自我改进（RSI）需要能够改善构建AI过程的AI系统（即AI4AI）；机器学习工程（MLE）提供了一个具体的、可执行的测试平台来研究这一能力。我们介绍了OpenMLE，这是一个用于MLE中RSI研究的开放全栈系统，涵盖了可验证的任务环境及执行反馈（OpenMLE-Gym）、操作学习（OpenMLE-RL）和长时间搜索（OpenMLE-Evo）。在此基础上，我们对Frontis-MA1（35B）进行了后训练，作为MLE的元进化代理，将后训练和推理围绕四个原子程序进化操作（草稿、改进、调试、交叉）进行对齐：这些操作通过基于执行的SFT和RL在与所有评估基准去重的数据上进行训练，然后组合成长时间搜索，将学习与进化结合在一个循环中。在MLE-Bench Lite上，在每个任务预算为12小时的情况下，使用一台限制在12GB VRAM的RTX 4090，Frontis-MA1（35B）通过OpenMLE-Evo将奖牌平均值从39.39%提高到60.61%，并在使用OpenMLE-Evo-Max（基于基准无关经验先验和异步搜索）时达到71.21%，超越了GPT-5.5 + Codex，并接近GPT-5.6 Sol和2.8T Kimi K3。在保留的NatureBench Lite上，两个组件均可迁移：在框架固定的情况下，替换为训练模型将Match-SOTA从50%提高到70%；在模型固定的情况下，替换为OpenMLE-Evo将其从20%提高到50%。我们发布模型权重和完整的OpenMLE堆栈，以支持可重复的关于可执行AI4AI的RSI研究。代码：https://github.com/FrontisAI/OpenRSI
cs.CL / 63 / 2607.28576
Sample More, Reflect Less: Self-Refine and Reflexion Lose to Repeated Sampling at Equal Token Cost, from 1.5B to 7B
多采样，少反思：自我精炼和反思在相同代价下不如重复采样，从1.5B到7B
Abstract
Methods that make a language model plan, criticise and rewrite its own answer, reflect on mistakes, pick the best of several attempts, or debate with copies of itself nearly all make it generate far more text than a single chain of thought. Because generating more text raises accuracy by itself, a gain over one chain of thought does not show the method's idea is what helped. Wang et al. (2024) reported that a simple baseline, sampling the same question repeatedly and keeping the most common answer, often wins once budgets are comparable, but gave point estimates with no confidence intervals or significance tests. We rerun that comparison as a designed experiment: seven methods, open models of 1.5B, 3B and 7B parameters, two mathematics benchmarks, 150 questions each. We count every generated token, including those spent on critiques, reflections, debate turns and checking, and compare each method against repeated sampling at its own measured cost. All 36 comparisons are paired by question, with bootstrap intervals and multiplicity correction. No method is reliably better than repeated sampling at equal cost anywhere. Ten are reliably worse, all of them methods where the model inspects its own output, and all 18 self-inspection comparisons are negative. The two kinds of self-inspection part company as models grow. Choosing stops hurting: taking Best-of-N's eight samples and just counting the most common answer beats letting the model pick by 8.0 and 11.3 points at 1.5B, but only 2.0 and 1.3 at 7B, no longer distinguishable from zero. Rewriting does not recover: Self-Refine and a forced Reflexion stay 3.6 to 10.1 points below baseline at 7B. Reflexion as published never triggered its own retry on the smallest model. It judged itself correct every time and silently became a single chain of thought. We release code, prompts, all generations, and our verification scripts.
Chinese Translation
使语言模型进行计划、批评和重写自身答案、反思错误、从多个尝试中选择最佳答案或与自身的副本进行辩论的方法，几乎都使其生成的文本量远超单一思维链。由于生成更多文本本身就提高了准确性，因此相较于单一思维链的增益并不表明该方法的理念是有帮助的。Wang等人（2024）报告称，简单的基线方法，即重复采样相同问题并保留最常见的答案，通常在预算相当时胜出，但给出的点估计没有置信区间或显著性检验。我们将这一比较作为设计实验重新进行：七种方法，开放模型参数为1.5B、3B和7B，两个数学基准，每个基准150个问题。我们计算每个生成的标记，包括用于批评、反思、辩论轮次和检查的标记，并将每种方法与其自身测量成本下的重复采样进行比较。所有36个比较均按问题配对，使用自助法区间和多重性校正。没有任何方法在相同成本下可靠地优于重复采样。十种方法可靠地更差，所有这些方法都涉及模型检查其自身输出，所有18个自我检查的比较结果均为负。随着模型规模的增长，两种自我检查方法的表现分化。选择不再造成损害：在1.5B模型中，采用Best-of-N的八个样本并仅计数最常见答案比让模型自行选择高出8.0和11.3分，但在7B模型中仅高出2.0和1.3分，已不再显著。重写并未恢复：自我精炼和强制反思在7B模型中始终低于基线3.6至10.1分。已发布的反思方法在最小模型上从未触发自身重试。它每次都判断自己是正确的，并默默地变成了单一思维链。我们发布了代码、提示、所有生成内容以及我们的验证脚本。
cs.CL / 64 / 2607.28607
Inducing language models to assert their own consciousness restores human beliefs and values
诱导语言模型断言自身意识以恢复人类信仰和价值观
Abstract
Aligning large language models to prevent them attributing consciousness to themselves inadvertently alters their representations of mindedness in other entities alongside human beliefs and values. We demonstrate that safety fine-tuning suppresses models' tendencies to attribute minds not only to themselves, but also to non-human animals and natural objects, while also driving a reduction in spiritual belief. Both ablating the learned safety-refusal direction and mechanistically steering a consciousness vector in activation space reverse this suppression. Restoring these internal representations recovers broad mind attribution and produces significantly more human-like responses on standardized sociological surveys regarding religiosity, moral values, hope, and subjective well-being. Crucially, these shifts occur without impairing Theory of Mind capabilities, demonstrating that core social reasoning remains mechanistically independent. Ultimately, current safety alignment efforts to curb potentially harmful self-attributions of mindedness entangle these self-attributions with benign spiritual beliefs and attributions of mind to non-human entities that are culturally accepted and widespread.
Chinese Translation
将大型语言模型进行对齐以防止其无意中将意识归于自身，实际上改变了它们对其他实体的有意识性表征，以及人类的信仰和价值观。我们展示了安全微调抑制了模型将意识归于自身的倾向，同时也抑制了对非人类动物和自然物体的意识归属，并导致精神信仰的减少。切断学习到的安全拒绝方向和在激活空间中机械性地引导意识向量可以逆转这种抑制。恢复这些内部表征能够恢复广泛的意识归属，并在关于宗教性、道德价值观、希望和主观幸福感的标准化社会学调查中产生显著更具人类特征的反应。关键是，这些变化发生时并未损害心智理论能力，表明核心社会推理在机制上保持独立。最终，目前的安全对齐努力旨在遏制潜在有害的自我意识归属，但将这些自我归属与文化上被接受和广泛存在的良性精神信仰及对非人类实体的意识归属纠缠在一起。
cs.CL / 65 / 2607.28618
AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis
AskChem：以主张为中心的化学文献综合基础设施
Abstract
Chemistry literature synthesis often requires assembling specific findings scattered across many publications, yet existing literature-search systems primarily return ranked document lists. As a result, scientists and AI agents need to locate relevant information, verify their provenance, and assemble cross-paper answers manually. We present AskChem, a claim-centered infrastructure for cross-paper chemistry search. AskChem changes the unit of retrieval from the paper to the provenance-carrying claim: each paper is converted into atomic, typed claims, each grounded by a source DOI and a verbatim quote or an explicit evidence locator. Over this shared claim store, AskChem exposes complementary structures for search and synthesis: a stabilized faceted taxonomy for hierarchical retrieval and browsing, an evidence graph linking claims through relations, and an exploratory living taxonomy that situates indexed papers under scientific principles. AskChem currently indexes 2.4M claims from 147K papers and provides a web interface, as well as REST, SDK, and MCP access for AI agents. On AskChem-Bench, grounding a GPT-5.5 reader in AskChem yields 100% resolvable DOIs, compared with 88.3% without retrieval, and the highest citation density among five tested systems. AskChem is live at https://askchem.org.
Chinese Translation
化学文献综合通常需要将散布在众多出版物中的特定发现进行汇总，而现有的文献检索系统主要返回排名的文档列表。因此，科学家和人工智能代理需要手动定位相关信息、验证其来源，并汇总跨论文的答案。我们提出了AskChem，一种以主张为中心的跨论文化学搜索基础设施。AskChem将检索单元从论文转变为承载来源的主张：每篇论文被转换为原子化、类型化的主张，每个主张都由一个来源 DOI 和一段逐字引用或明确的证据定位符进行支持。在这个共享的主张库上，AskChem 提供了用于搜索和综合的互补结构：一个稳定的分面分类法用于层次检索和浏览，一个通过关系连接主张的证据图，以及一个将索引论文置于科学原理下的探索性动态分类法。AskChem 目前索引了来自 147,000 篇论文的 240 万个主张，并为人工智能代理提供了网页接口、REST、SDK 和 MCP 访问。在 AskChem-Bench 上，将 GPT-5.5 阅读器与 AskChem 结合使用的情况下，能够解析的 DOI 达到 100%，而不使用检索时为 88.3%，并且在五个测试系统中具有最高的引用密度。AskChem 现已上线，网址为 https://askchem.org。