cs.RO / 1 / 2607.14177
Adaptive Control of Motor-Position-Controlled Flexible Joint Robots with Uncertain Joint Stiffness
具有不确定关节刚度的电机位置控制柔性关节机器人自适应控制
Abstract
Model-based control of flexible joint robots with position-controlled actuators relies on accurate knowledge of the joint compliance. In practice, precise stiffness models are often unavailable as the properties of physical elastic elements vary with operating conditions and slowly change over time due to wear and aging. To improve model-based control of these systems, we propose an adaptive control approach in this work, which updates an estimate of the uncertain, nonlinear torque-deflection relation of each joint. As opposed to classical adaptive control approaches for non-elastic robots, we rely on an implicit control law and a control-input-dependent regressor matrix to account for the uncertain joint stiffness. We analyze robustness of the approach against errors induced by the motor position controller. Experimental results on a flexible joint with nonlinear stiffness characteristics demonstrate the effectiveness of the proposed approach.
Chinese Translation
基于模型的柔性关节机器人控制依赖于对关节顺应性的准确了解。在实际应用中，由于物理弹性元件的特性随操作条件变化并且因磨损和老化而缓慢变化，精确的刚度模型往往不可用。为改善这些系统的基于模型的控制，本文提出了一种自适应控制方法，该方法更新每个关节不确定的非线性扭矩-偏转关系的估计。与传统的非弹性机器人自适应控制方法不同，我们依赖于隐式控制律和控制输入依赖的回归矩阵来考虑不确定的关节刚度。我们分析了该方法对电机位置控制器引入的误差的鲁棒性。在具有非线性刚度特性的柔性关节上的实验结果证明了所提方法的有效性。
cs.RO / 2 / 2607.14182
Semantic Audio-driven Understanding for Dynamic Humanoid Whole Body Control
基于语义音频驱动的动态类人全身控制理解
Abstract
Recent advances in humanoid robotics and reinforcement learning have enabled the acquisition of highly expressive whole-body motion policies. However, most robotic performances remain based on pre-scripted sequences or externally triggered behaviors, limiting autonomy and responsiveness to dynamic environments. In this work, we introduce a novel multi-modal orchestration framework for semantic audio-driven humanoid control, enabling robots to autonomously select and execute appropriate motion skills in real time. The system processes continuous audio streams and routes them into music or speech branches. Music input is handled via audio fingerprinting and semantic embeddings to retrieve track identity and temporal alignment, allowing dynamic mapping between musical segments and motion policies. Speech input is grounded into a discrete library of imitation-learned skills, enabling direct human-robot interaction. Both modalities share a unified interface that schedules skill execution over a reinforcement learning control pipeline. We validate the approach in simulation and on a Unitree G1 humanoid, showing robust sim-to-real transfer and consistent audio-conditioned policy selection. Supplementary materials are available at the following site: https://lab-rococo-sapienza.github.io/semantic-WBC/
Chinese Translation
近年来，类人机器人和强化学习的进展使得获取高度表现力的全身运动策略成为可能。然而，大多数机器人表演仍然基于预先编写的序列或外部触发的行为，这限制了其在动态环境中的自主性和响应能力。在本研究中，我们提出了一种新颖的多模态编排框架，用于基于语义音频驱动的类人控制，使机器人能够实时自主选择和执行适当的运动技能。该系统处理连续的音频流，并将其分流到音乐或语音分支。音乐输入通过音频指纹识别和语义嵌入进行处理，以检索曲目身份和时间对齐，从而允许音乐片段与运动策略之间的动态映射。语音输入则基于离散的模仿学习技能库，使人机互动变得直接。两种模态共享一个统一的接口，通过强化学习控制管道调度技能执行。我们在仿真和Unitree G1类人机器人上验证了该方法，显示出强大的仿真到现实转移能力和一致的音频条件策略选择。补充材料可在以下网站获取： https://lab-rococo-sapienza.github.io/semantic-WBC/
cs.RO / 3 / 2607.14183
Open-AoE: An Open Egocentric Manipulation Dataset and Toolchain for Embodied Learning
Open-AoE：一个开放的自我中心操控数据集及其工具链用于具身学习
Li, Zishuo, Yang, Bowen, Miao, Changtao, Zhu, Kai, Chen, Hao, Guan, Qingze, Wu, Zhengxing, Zhan, Wanke, Sun, Yang, Huang, Zhiyi, Shan, Zitong, Jin, Zhenchao, Hong, Jiadong, Wang, Taowen, Feng, Yushi, Liu, You, Wang, Yibo, Yang, Yifan, Zhou, Zhaowen, Luo, Man, Cheng, Hao, Zhang, Bo, Li, Jianshu, Cai, Jiansheng, Yao, Guocai, Zhang, Jize, Lin, Chenhao, Xu, Renjing, Yu, Lequan, Shen, Chao, Shen, Chunhua, Li, Zhe
Abstract
Egocentric videos of human manipulation provide scalable supervision for embodied intelligence, yet existing resources rarely combine low-cost continuous capture, manipulation-level structured annotations, and reusable tools for robot learning. We present Open-AoE, an open, community-oriented egocentric manipulation dataset and toolchain spanning the full pipeline from smartphone capture to model training. Its first release contains approximately 2,000 hours of manipulation video collected in natural environments by 500+ contributors using 400+ smartphones. The dataset provides text annotations, MANO-based hand poses, camera trajectories, and temporally localized atomic actions. Open-AoE further includes a data processing pipeline that transforms raw recordings into structured samples through temporal action segmentation, semantic annotation, hand reconstruction, and camera trajectory reconstruction. Meanwhile, we provide a separate downstream toolchain supports visualization, cross-embodiment retargeting, model-specific data conversion, and training recipes for VLA policies, WAMs, and World Models. By integrating scalable capture, structured processing, and downstream adaptation, Open-AoE reduces the barriers to both data contribution and reuse, providing practical open infrastructure for embodied model training, human-to-robot transfer, and world modeling.
Chinese Translation
人类操控的自我中心视频为具身智能提供了可扩展的监督，但现有资源很少结合低成本的连续捕捉、操控级结构化注释和可重用的机器人学习工具。我们提出了Open-AoE，一个开放的、面向社区的自我中心操控数据集及其工具链，涵盖了从智能手机捕捉到模型训练的完整流程。其首次发布包含约2000小时的操控视频，这些视频由500多名贡献者使用400多部智能手机在自然环境中收集。该数据集提供文本注释、基于MANO的手部姿态、相机轨迹和时间局部化的原子动作。Open-AoE还包括一个数据处理管道，通过时间动作分割、语义注释、手部重建和相机轨迹重建，将原始录音转换为结构化样本。同时，我们提供了一个独立的下游工具链，支持可视化、跨具身重定向、特定模型的数据转换以及VLA策略、WAMs和世界模型的训练配方。通过整合可扩展的捕捉、结构化处理和下游适应，Open-AoE降低了数据贡献和重用的门槛，为具身模型训练、人机转移和世界建模提供了实用的开放基础设施。
cs.RO / 4 / 2607.14236
Never Too Late for Force: Accelerating VLA Post-Training with Reactive Force Injection
永远不嫌晚：通过反应力注入加速VLA后训练
Abstract
Pretrained vision-language-action (VLA) policies provide strong language-conditioned manipulation knowledge, but they remain largely vision-driven and can struggle once manipulation enters contact states where the scene is occluded, depth is ambiguous, or small force errors push execution off the offline demonstration distribution. We present LIFT (Late Reactive Injection of Force for VLA Post-Training), a force-aware post-training framework that adds contact reactivity to a pretrained VLA policy while preserving its general manipulation knowledge. LIFT grafts a reactive action expert beside the original action expert, initializes it from pretrained action weights, and injects recent 6D end-effector force through causal force memory and zero-initialized cross attention, enabling actions to be refreshed during execution. To address the policy-dependent distribution shift of contact feedback, LIFT further couples reactive force injection with an online DAgger loop that trains on a mixture of offline task-alignment data and human-corrected online rollouts. Across towel folding, book insertion, and Hanoi ring placement, LIFT learns faster and reaches higher performance than vision-only post-training, while ablations show that reactive force memory and online corrective data are both important for robust contact-rich manipulation. Our code and data will be publicly available.
Chinese Translation
预训练的视觉-语言-动作（VLA）策略提供了强大的语言条件下的操作知识，但它们在很大程度上仍然依赖于视觉，并且在操作进入接触状态时可能会遇到困难，例如场景被遮挡、深度模糊或小的力误差使得执行偏离离线演示分布。我们提出了LIFT（Late Reactive Injection of Force for VLA Post-Training），这是一个力感知的后训练框架，它在保留预训练VLA策略的一般操作知识的同时，为其增加了接触反应性。LIFT在原始动作专家旁边 grafts 一个反应动作专家，从预训练的动作权重初始化，并通过因果力记忆和零初始化的交叉注意力注入最近的6D末端执行器力，从而使得在执行过程中能够刷新动作。为了应对接触反馈的策略依赖分布偏移，LIFT进一步将反应力注入与在线DAgger循环耦合，该循环在离线任务对齐数据和人类修正的在线回放的混合上进行训练。在毛巾折叠、书籍插入和汉诺塔环放置等任务中，LIFT的学习速度更快，性能高于仅依赖视觉的后训练，而消融实验表明，反应力记忆和在线修正数据对于稳健的接触丰富操作都至关重要。我们的代码和数据将公开可用。
cs.RO / 5 / 2607.14252
MEMORA: Embodied Action Memory from Egocentric Videos for Reasoning and Planning
MEMORA：来自自我中心视频的具身行动记忆用于推理和规划
Abstract
Long-horizon robot planning requires more than predicting what actions will do next; it also requires memory of the embodied experience that makes future goals interpretable. People do not plan from the present scene alone: they draw on remembered places, object-state changes, prior procedures, and regularities revealed through repeated action. We formulate Embodied Action Memory (EAM) as the capability to form, maintain, and use such experience as a persistent memory state for later decisions. MEMORA realizes EAM with a formation-consolidation-retrieval lifecycle and four typed stores: Environment Memory, Entity Memory, Activity Memory, and Inferred Knowledge. Online editing maintains object identities and state histories as new observations arrive; offline consolidation abstracts repeated experience into reusable procedures and participant-specific regularities. MEMORA-Bench evaluates this lifecycle on 45 hours of EPIC-KITCHENS-100 extension video across 18 participants through memory-grounded planning, including previously unseen goals, and a complementary memory-assessment task. Across four open-weight language models, full MEMORA--combining editing, typed stores, and consolidation--achieves the strongest aggregate results among the evaluated memory conditions. It improves memory-assessment accuracy by up to 20.5 points over the strongest controlled baseline and improves out-of-distribution Robot-Grounded Plan score by up to 16.6% relative. A qualitative two-task robot deployment study further illustrates how memory-grounded language plans can interface with downstream control, while the overall results show that editable, consolidated memory can supply remembered context for robot planning. Project page: https://yuzihaowashu.github.io/MEMORA/
Chinese Translation
长时间范围的机器人规划不仅需要预测接下来会采取的行动，还需要对具身经验的记忆，以使未来目标可解释。人们并不是仅仅从当前场景进行规划：他们会借助记忆中的地点、物体状态变化、先前的程序以及通过重复行动所揭示的规律。我们将具身行动记忆（Embodied Action Memory, EAM）定义为形成、维护和利用这种经验作为持久记忆状态以供后续决策的能力。MEMORA通过形成-巩固-检索生命周期和四种类型的存储实现EAM：环境记忆（Environment Memory）、实体记忆（Entity Memory）、活动记忆（Activity Memory）和推断知识（Inferred Knowledge）。在线编辑在新观察到达时维护物体身份和状态历史；离线巩固将重复经验抽象为可重用的程序和特定参与者的规律。MEMORA-Bench在18名参与者的45小时EPIC-KITCHENS-100扩展视频上评估了这一生命周期，通过基于记忆的规划，包括先前未见的目标，以及一个补充的记忆评估任务。在四个开放权重语言模型中，完整的MEMORA——结合编辑、类型存储和巩固——在评估的记忆条件中实现了最强的综合结果。它在记忆评估准确性上比最强的控制基线提高了多达20.5分，并在分布外的机器人基础计划得分上相对提高了多达16.6%。一项定性两任务机器人部署研究进一步说明了基于记忆的语言计划如何与下游控制接口，而整体结果表明，可编辑的、巩固的记忆可以为机器人规划提供记忆上下文。项目页面：https://yuzihaowashu.github.io/MEMORA/
cs.RO / 6 / 2607.14262
Stochastic Filtering for Quorum Sensing in Robot Swarms under Anonymous Communication
在匿名通信下机器人群体的群体感知的随机滤波
Abstract
Quorum Sensing (QS) is a key capability for robot swarms, useful for coordination of activities at the group level. Effective communication is instrumental for individuals to estimate the quorum level of the entire swarm. Anonymous communication protocols where individuals exchange local information without revealing unique identities are helpful to support quorum estimates by sampling information from neighbours and maintain scalability of the QS process. However, because anonymous protocols cannot distinguish message sources, repeated messages from the same sender may be double-counted, thereby biasing collective quorum estimates. In this study, we introduce a stochastic filtering protocol inspired by $k$-priority sampling to improve estimate stability (\ANTk), and we compare it with a baseline anonymous protocols (\AN) and a randomised variant designed to improve accuracy (\ANT). We find that the baseline protocol \AN provides a parsimonious and fast solution, but remains highly inaccurate due to double-counting bias. The \ANT variant improves accuracy but suffers from information inertia, resulting in slower convergence. Finally, actively filtering the message buffer via the \ANTk protocol successfully decreases temporary errors and stabilises the estimate, at the cost of an increased time of recovery from errors.
Chinese Translation
群体感知（Quorum Sensing, QS）是机器人群体的一项关键能力，有助于在群体层面协调活动。有效的通信对于个体估计整个群体的群体水平至关重要。匿名通信协议使个体能够交换局部信息而不透露独特身份，从而通过从邻居处采样信息来支持群体估计，并保持QS过程的可扩展性。然而，由于匿名协议无法区分消息来源，同一发送者的重复消息可能会被重复计算，从而导致集体群体估计的偏差。在本研究中，我们引入了一种受$k$-优先采样启发的随机滤波协议（ extit{ANTk}），以提高估计的稳定性，并将其与基线匿名协议（ extit{AN}）和旨在提高准确性的随机变体（ extit{ANT}）进行比较。我们发现基线协议 extit{AN}提供了一种简约且快速的解决方案，但由于重复计算偏差，准确性仍然较低。 extit{ANT}变体提高了准确性，但受到信息惯性的影响，导致收敛速度较慢。最后，通过 extit{ANTk}协议主动过滤消息缓冲区成功减少了暂时性错误并稳定了估计，代价是恢复错误的时间增加。
cs.RO / 7 / 2607.14280
DiMaS: Distribution Matching for Steering Vision-Language-Action Models
DiMaS：用于引导视觉-语言-动作模型的分布匹配
Abstract
Flow-matching-based vision-language-action (VLA) models have emerged as powerful policies for robotic manipulation, yet a critical capability remains underexplored: fine-grained behavioral control, the ability to govern how a robot performs a task by intervening on its internal representations. Representation steering is a well-established interpretability tool for language and vision-language models, where behavioral features are typically encoded as linear directions, but we show that these classic methods fall short in VLAs. We propose DiMaS, a Distribution-Matching Steering strategy tailored to flow-matching VLAs, which transports between representation distributions rather than shifting along a fixed direction, and show that it effectively controls behavior across two state-of-the-art VLAs. We further examine the generalizability of this strategy as the tasks it is learned from and evaluated on grow increasingly dissimilar, characterizing where behavioral control transfers and where it weakens. Finally, through an analysis of the representation structure of the action expert, we explain why classical linear steering falls short in the visuomotor setting: behavioral features are linearly decodable but not linearly steerable, which motivates the distribution-matching design of DiMaS. Our code is publicly available at https://github.com/pegah-kh/dimas, with additional results and videos at https://pegah-kh.github.io/dimas/
Chinese Translation
基于流匹配的视觉-语言-动作（VLA）模型已成为机器人操作的强大策略，但一个关键能力仍未得到充分探索：细粒度行为控制，即通过干预机器人的内部表征来管理其执行任务的方式。表征引导是语言和视觉-语言模型中一种成熟的可解释性工具，其中行为特征通常被编码为线性方向，但我们表明这些经典方法在VLA中效果不佳。我们提出了DiMaS，一种针对流匹配VLA的分布匹配引导策略，它在表征分布之间进行传输，而不是沿固定方向移动，并展示了它在两个最先进的VLA中有效控制行为的能力。我们进一步考察了这一策略的可推广性，随着学习和评估任务的日益不同，描述了行为控制转移的地方以及其减弱的地方。最后，通过对动作专家的表征结构分析，我们解释了为什么经典的线性引导在视觉运动设置中效果不佳：行为特征是线性可解的，但并非线性可引导，这促使了DiMaS的分布匹配设计。我们的代码已公开发布，网址为 https://github.com/pegah-kh/dimas，附加结果和视频可在 https://pegah-kh.github.io/dimas/ 查看。
cs.RO / 8 / 2607.14341
Beyond Visual Grasping: Benchmarking Complex Grasping from Detection to Execution
超越视觉抓取：从检测到执行的复杂抓取基准测试
Abstract
Robust robotic grasping remains a fundamental challenge for complex real-world applications. Recent advances in large-scale models demonstrate promising capabilities for reasoning in robotic tasks. However, existing benchmarks for grasping primarily focus on isolated, visual-based grasp pose detection, failing to capture the complexity of grasping tasks that require multi-step reasoning and semantic understanding during execution. To address this gap, we propose GCA-Bench, a benchmark featuring challenging \textit{grasping with complex action} scenarios that involve both scene-level reasoning and semantic constraints. GCA-Bench enables the evaluation of recent large foundation models under the same settings. To demonstrate the effectiveness of our new benchmark, we implement a diverse set of baselines, ranging from traditional grasp detection pipelines to end-to-end learning methods. Empirical studies achieve success rates below 70\% on complex grasping scenarios, underscoring critical limitations. In addition, we propose new evaluation metrics, analyze critical failure models, and provide insights to guide the development of more robust and generalizable grasping strategies.
Chinese Translation
稳健的机器人抓取仍然是复杂现实应用中的一个基本挑战。近期在大规模模型方面的进展展示了在机器人任务推理中的良好能力。然而，现有的抓取基准主要集中于孤立的、基于视觉的抓取姿态检测，未能捕捉到在执行过程中需要多步骤推理和语义理解的抓取任务的复杂性。为了解决这一问题，我们提出了GCA-Bench，一个基准测试，包含具有挑战性的“复杂动作抓取”场景，这些场景涉及场景级推理和语义约束。GCA-Bench使得在相同设置下评估近期的大型基础模型成为可能。为了展示我们新基准的有效性，我们实施了一系列多样的基线，从传统的抓取检测管道到端到端学习方法。实证研究在复杂抓取场景中成功率低于70%，突显了关键的局限性。此外，我们提出了新的评估指标，分析了关键的失败模型，并提供了指导，以推动更稳健和更具普适性的抓取策略的发展。
cs.RO / 9 / 2607.14393
An offline approach to fNIRS-guided reinforcement learning for robot behavior
一种离线方法用于基于功能近红外光谱（fNIRS）指导的强化学习以优化机器人行为
Abstract
Human-in-the-loop Reinforcement Learning has become a popular approach to training, finetuning, and aligning robot behavior with user preferences. Our paper explores the feasibility of using brain signals via functional near-infrared spectroscopy (fNIRS) to modulate robot learning in simulation. We compare agents trained on passive (observational) versus active (demonstrative) interaction tasks, and test multiple methods for enhancing the RL algorithm with the neural signal, focusing on parameter augmentation rather than replacement. We further examine how model granularity and noise affect agent learning. Our results show that this framework is effective: the neural signal improves learning when augmenting trajectory priorities and state-action q-values. Additionally, the framework learns successfully from offline data, offering a practical alternative for settings where real-time BCI setups are impractical or only limited data is available.
Chinese Translation
人机协同强化学习已成为训练、微调和使机器人行为与用户偏好对齐的热门方法。本文探讨了通过功能近红外光谱（fNIRS）使用脑信号调节机器人学习在仿真中的可行性。我们比较了在被动（观察性）与主动（示范性）交互任务上训练的智能体，并测试了多种方法以增强强化学习（RL）算法，重点关注参数增强而非替换。我们进一步研究了模型粒度和噪声如何影响智能体学习。我们的结果表明，该框架是有效的：当增强轨迹优先级和状态-动作q值时，脑信号改善了学习。此外，该框架成功地从离线数据中学习，为实时脑机接口（BCI）设置不切实际或仅有有限数据可用的环境提供了一种实用的替代方案。
cs.RO / 10 / 2607.14424
ConFlow: Constraints-Guided Learning with Flow Matching for Motion Generation
ConFlow：基于约束引导的流匹配运动生成学习
Abstract
In recent years Flow Matching has become a prominent method for generative modeling robot motion generation. In its generic form Flow Matching is an ODE-based neural sampler that is trained by regressing empirical flow fields associated with motion samples as data. However, in robot motion generation we often have additional constraints that might not be present in the collected data. The majority of current approaches train the flow on the available data and use inference-time guidance to enforce task-specific constraints. To address this mismatch, we propose \textbf{ConFlow}, a constraint-guided flow matching framework that incorporates constraint information directly into the training objective via differentiable barrier or cost functions. To address design specifications such as smoothness and boundary conditions, we propose replacing the standard Gaussian source distribution used in flow matching training with a conditional Gaussian Process. Our approach also uses infeasible demonstrations as negative supervision, improving constraint satisfaction without requiring additional expert data. Experiments on a two-robot navigation task demonstrate that ConFlow achieves lower collision rates and higher trajectory quality than standard flow matching baselines, with or without inference-time guidance. These results validate training-time constraint integration as an effective approach to closing the training--inference gap in generative motion models.
Chinese Translation
近年来，流匹配（Flow Matching）已成为生成建模机器人运动生成的一个重要方法。在其通用形式中，流匹配是一个基于常微分方程（ODE）的神经采样器，通过回归与运动样本相关的经验流场作为数据进行训练。然而，在机器人运动生成中，我们通常会有一些在收集的数据中可能不存在的额外约束。目前大多数方法是在可用数据上训练流，并在推理时使用指导来强制执行特定任务的约束。为了解决这种不匹配，我们提出了 extbf{ConFlow}，一个约束引导的流匹配框架，它通过可微分的障碍或成本函数将约束信息直接纳入训练目标。为了满足平滑性和边界条件等设计规范，我们建议用条件高斯过程替代流匹配训练中使用的标准高斯源分布。我们的方法还利用不可行的演示作为负监督，提高约束满足度，而无需额外的专家数据。在一个双机器人导航任务上的实验表明，ConFlow在有无推理时指导的情况下，都实现了比标准流匹配基线更低的碰撞率和更高的轨迹质量。这些结果验证了训练时约束集成作为缩小生成运动模型训练与推理差距的有效方法。
cs.RO / 11 / 2607.14455
Motion Planning with Model-Based Diffusion via Constraint Optimization and Adaptive Scheduling
基于模型的扩散运动规划：约束优化与自适应调度
Abstract
Single-Robot Motion Planning (SRMP) in highly non-convex constrained environments, where robots must satisfy collision-free guarantees, dynamic feasibility, and task-related constraints, is challenging under complex constraints and computational limits. Recent Model-Based Diffusion (MBD) approaches recast the SRMP as trajectory optimization that samples from a posterior over trajectories, using known dynamics, and analytically estimates the score function from rollout samples to guide diffusion denoising toward a low-cost, clean trajectory without demonstration learning. While existing works further adapt MBD to constrained environments and showcase promising performance, they are still limited by (1) enforcing safety either via soft feasibility diffusion priors or hard projection operators, but lack a unified framework to integrate both, and (2) fixing safety enforcement to neglect the changing of diffusion scheduling. Therefore, we introduce Model-Based Diffusion via Constraint Optimization and Adaptive Scheduling (MD-COAS) for SRMP that unifies the inexact Augmented Lagrangian Method (iALM) soft diffusion prior with a Convex Feasible Set (CFS)-based hard projection operator, and adaptively schedules and co-optimizes safety enforcement, along with diffusion scheduling. Experiments demonstrate that our method achieves higher safety \& success rates, faster convergence, and lower final costs than baseline planners on randomly generated highly non-convex 2D benchmarks and a 7-DoF robot arm avoidance task.
Chinese Translation
在高度非凸约束环境中进行单机器人运动规划（SRMP）是一项挑战，机器人必须满足无碰撞保证、动态可行性和与任务相关的约束，这在复杂约束和计算限制下尤为困难。最近的基于模型的扩散（MBD）方法将SRMP重新表述为轨迹优化，通过已知动态从轨迹的后验分布中进行采样，并通过从回放样本中分析估计得分函数，以引导扩散去噪，朝向低成本、干净的轨迹，而无需示范学习。尽管现有工作进一步将MBD适应于约束环境并展示了良好的性能，但仍然受到以下限制：（1）通过软可行性扩散先验或硬投影算子来强制安全，但缺乏统一框架来整合两者；（2）固定安全强制而忽略扩散调度的变化。因此，我们提出了基于约束优化与自适应调度的模型基础扩散（MD-COAS）方法，统一了不精确的增广拉格朗日方法（iALM）软扩散先验与基于凸可行集（CFS）的硬投影算子，并自适应地调度和共同优化安全强制及扩散调度。实验表明，我们的方法在随机生成的高度非凸二维基准测试和7自由度机器人臂避障任务中，达到了比基线规划器更高的安全性和成功率、更快的收敛速度以及更低的最终成本。
cs.RO / 12 / 2607.14468
Mixed-Agent Museum Tour Guide Design Improves Gendered Learning Outcomes and Visitor Preferences
混合代理博物馆导览设计改善性别化学习成果和访客偏好
Abstract
Robots are increasingly integrated into everyday contexts, including museums, where they can both entertain and educate visitors. To enhance visitor experience and engagement, we present a novel mixed-agent tour guide system that combines a physical robot with a projected virtual agent that actively participates in the tour through conversation and interaction, achieving the interaction richness of two mobile agents from a single platform. We validate the system through a within-subjects study with 30 participants to assess engagement, quality of experience, and learning performance. Participants experienced different conversational styles and agent configurations, and data were collected via surveys, behavioral sensors, and interviews. Results showed that engagement and quality of experience remained consistent across conditions. Learning performance revealed a significant gender-moderated difference: the mixed-agent conditions improved learning performance for female participants. This suggests that the proposed dyadic conversational style in this paper influenced learning performance differently by gender. Nonetheless, in interviews, participants reported a greater preference for mixed-agent teams regardless of gender, citing interaction as a key factor in their experience.
Chinese Translation
机器人越来越多地融入日常环境，包括博物馆，在这些场所它们既可以娱乐又可以教育访客。为了提升访客的体验和参与度，我们提出了一种新颖的混合代理导览系统，该系统结合了一个物理机器人和一个投影虚拟代理，后者通过对话和互动积极参与导览，从而实现了从单一平台获得两个移动代理的互动丰富性。我们通过一项包含30名参与者的组内研究验证了该系统，以评估参与度、体验质量和学习表现。参与者体验了不同的对话风格和代理配置，数据通过调查问卷、行为传感器和访谈收集。结果显示，参与度和体验质量在不同条件下保持一致。学习表现揭示了显著的性别调节差异：混合代理条件提高了女性参与者的学习表现。这表明本文提出的双向对话风格对学习表现的影响因性别而异。然而，在访谈中，参与者报告无论性别如何，都更倾向于混合代理团队，认为互动是他们体验中的关键因素。
cs.RO / 13 / 2607.14487
MIDAS Hand: Modular low-Impedance Direct-drive Anthropomorphic Sensing Hand
MIDAS手：模块化低阻抗直接驱动的人形感知手
Abstract
Dexterous manipulation is limited not only by algorithms but by a shortage of accessible hand hardware that combines human-scale morphology, ease of manufacturing or maintenance, tactile sensing, and practical cost. Existing dexterous hands tend to optimize some of these properties at the expense of others. We present MIDAS Hand, a low-cost, open-source, human-scale dexterous hand with integrated tactile sensing for manipulation research. MIDAS Hand provides 16 total degrees of freedom (DoF) with 13 active DoF, directly driven actuation with measurably low backdrive torque, and 283 three-axis tactile taxels in a compact 700 g package with a bill of materials under 3,000 USD. Built from 3D-printed components, it assembles in under three hours while providing the strength, repeatability, and maintainability needed for repeated real-world experiments. Alongside the hardware, we release a full stack: design files, build documentation, control and tactile Python APIs, simulation models, and retargeting and teleoperation pipelines. We characterize MIDAS Hand through workspace and grasp-taxonomy analysis, payload and reliability tests, backdrivability measurements, and teleoperation demonstrations with tactile sensing, showing that it offers a balanced, reproducible platform for tactile dexterous manipulation and human-to-robot data collection. Project page: https://midas-hand.com
Chinese Translation
灵巧操作不仅受到算法的限制，还受到缺乏可获得的手部硬件的制约，这些硬件应结合人类规模的形态、易于制造或维护、触觉感知以及实用成本。现有的灵巧手往往在优化某些属性的同时牺牲其他属性。我们提出了MIDAS手，这是一款低成本、开源的人类规模灵巧手，集成了触觉感知，旨在用于操作研究。MIDAS手提供16个自由度（DoF），其中13个为主动自由度，采用直接驱动的驱动方式，具有可测量的低反向驱动扭矩，并在一个紧凑的700克的包装中集成了283个三轴触觉传感器，材料成本低于3000美元。该手由3D打印组件构成，组装时间不到三小时，同时提供进行重复现实世界实验所需的强度、重复性和可维护性。除了硬件，我们还发布了完整的技术栈：设计文件、构建文档、控制和触觉的Python API、仿真模型，以及重定向和远程操作管道。我们通过工作空间和抓取分类分析、载荷和可靠性测试、反向驱动测量以及带有触觉感知的远程操作演示对MIDAS手进行了特征描述，表明它提供了一个平衡且可重复的平台，用于触觉灵巧操作和人机数据收集。项目页面：https://midas-hand.com
cs.RO / 14 / 2607.14488
Safe Execution of RL Policies Via Acceleration-Based CBF-QP Constraint Enforcement for Real-World Robotic Deployments
通过基于加速度的 CBF-QP 约束执行安全的强化学习策略以实现现实世界的机器人部署
Abstract
Reinforcement Learning (RL) has demonstrated remarkable capabilities for solving complex robotic control problems, but its lack of safety guarantees severely limits deployment on hardware. In particular, as legged robots and manipulators often operate near safety-critical boundaries, out-of-distribution states can lead to failure upon deployment. To address this, we introduce Acc-CBF-QP, an acceleration-based Quadratic Program (QP) safety filter using Control Barrier Functions (CBFs) that constrains any RL policy onto a safe set at runtime without modifying training. The method applies to unconstrained and Safe-RL policies, and enforces joint position, velocity, torque, and collision constraints within a unified optimization framework. A key contribution is the formulation of RL+QP tasks that regulate deviation from the RL command when constraints would otherwise be violated. We introduce a TorqueTask, minimizing torque deviation, and a Forward Dynamics Task, minimizing induced acceleration deviation, thus providing principled control over safety-performance trade-offs. Experiments on a 7-DoF Kinova Gen3 manipulator and a 19-DoF Unitree H1 humanoid, both in simulation and on hardware, highlight substantial reductions in constraint violations. On the real H1 hardware, a Safe-RL policy alone yielded 10.04 violations/s, which were reduced by 92% to 0.80 violations/s when augmented with Acc-CBF-QP. On the Kinova Gen3, Acc-CBF-QP fully eliminated violations. Nominal task performance of the RL objective is preserved in violation-free regimes. Under aggressive velocity commands on H1, Acc-CBF-QP improves execution by preventing constraint-induced shutdowns, yielding longer survival times. The full pipeline is open-source.
Chinese Translation
强化学习（Reinforcement Learning, RL）在解决复杂的机器人控制问题方面展现了显著的能力，但其缺乏安全保障严重限制了在硬件上的部署。特别是，由于腿部机器人和操控器通常在安全临界边界附近操作，分布外状态可能导致部署失败。为了解决这个问题，我们提出了 Acc-CBF-QP，一种基于加速度的二次规划（Quadratic Program, QP）安全过滤器，利用控制障碍函数（Control Barrier Functions, CBFs）在运行时将任何 RL 策略约束到安全集，而无需修改训练。该方法适用于无约束和安全强化学习（Safe-RL）策略，并在统一的优化框架内强制执行关节位置、速度、扭矩和碰撞约束。一个关键贡献是 RL+QP 任务的构建，该任务在约束可能被违反时调节与 RL 命令的偏差。我们引入了扭矩任务（TorqueTask），最小化扭矩偏差，以及前向动力学任务（Forward Dynamics Task），最小化引起的加速度偏差，从而提供对安全与性能权衡的原则性控制。在 7 自由度的 Kinova Gen3 操作臂和 19 自由度的 Unitree H1 人形机器人上进行的实验，无论是在仿真还是在硬件上，都突显了约束违反的显著减少。在真实的 H1 硬件上，仅使用安全强化学习策略的情况下，每秒产生 10.04 次违反，而在添加 Acc-CBF-QP 后减少了 92%，降至每秒 0.80 次违反。在 Kinova Gen3 上，Acc-CBF-QP 完全消除了违反。在无违反的情况下，RL 目标的名义任务性能得以保留。在 H1 上的激进速度命令下，Acc-CBF-QP 通过防止因约束引起的停机来改善执行，延长生存时间。整个流程是开源的。
cs.RO / 15 / 2607.14507
DRIFT: Drift and Aggregation for Motion Planning
DRIFT：运动规划中的漂移与聚合
Abstract
End-to-end trajectory planners need to represent multiple plausible driving behaviors while producing a single executable trajectory under real-time constraints. Proposal-based approaches address this ambiguity by generating multiple candidates, but converting the proposal set into a final plan remains a key design problem. We present DRIFT, a fixed-depth planner that combines one-step drifting in a compact trajectory latent space with scene-aware proposal aggregation. Conditioned on features from a pretrained visual encoder, the DRIFT Decoder generates 48 proposal features in a single batched pass, with 32 samples at alpha=0.5 and 16 samples at alpha=0.9. A lightweight Aggregation Head integrates these features with scene, navigation, and ego-state information and directly predicts the final trajectory without requiring trajectory-level quality labels for aggregation. Its output is trained with expert-trajectory imitation and a map-derived boundary regularizer that penalizes waypoints outside the drivable polygon and inside waypoints near its boundary. On NAVSIM navtest, DRIFT achieves 89.6 PDMS and 90.4 EPDMS, with strong drivable-area compliance and ego progress among the methods compared. The proposal-generation and aggregation module runs in 10.82 ms on an NVIDIA RTX 4090, while full-model inference including the visual backbone takes 66.43 ms. These results show that one-step latent proposal generation and direct aggregation provide an efficient design for multi-hypothesis motion planning.
Chinese Translation
端到端轨迹规划器需要在实时约束下表示多种可行的驾驶行为，同时生成单一可执行的轨迹。基于提案的方法通过生成多个候选轨迹来解决这种模糊性，但将提案集转换为最终计划仍然是一个关键设计问题。我们提出了DRIFT，这是一种固定深度的规划器，结合了在紧凑轨迹潜在空间中的一步漂移与场景感知的提案聚合。DRIFT解码器基于预训练视觉编码器的特征生成48个提案特征，采用单次批处理方式，其中在alpha=0.5时生成32个样本，在alpha=0.9时生成16个样本。轻量级聚合头将这些特征与场景、导航和自我状态信息整合，并直接预测最终轨迹，而无需轨迹级质量标签进行聚合。其输出通过专家轨迹模仿和基于地图的边界正则化器进行训练，该正则化器对超出可驾驶多边形的路径点和靠近边界的路径点进行惩罚。在NAVSIM navtest上，DRIFT达到了89.6的PDMS和90.4的EPDMS，在所比较的方法中具有较强的可驾驶区域合规性和自我进展。提案生成和聚合模块在NVIDIA RTX 4090上运行时间为10.82毫秒，而包括视觉主干的全模型推理时间为66.43毫秒。这些结果表明，一步潜在提案生成和直接聚合为多假设运动规划提供了一种高效的设计。
cs.RO / 16 / 2607.14539
Communication-Efficient Relative Pose Estimation with Vision Foundation Models for Ephemeral Collaborative Perception
基于视觉基础模型的通信高效相对姿态估计用于短暂协作感知
Abstract
Relative pose estimation is a fundamental capability for collaborative perception and coordination in multi-robot systems. However, robots encountering each other in real-world environments often operate in short interaction windows and must operate under limited communication bandwidth with intermittent or missing visual overlap caused by occlusions or limited fields of view. Existing approaches typically rely on global reference frames, assume sustained view overlap, or incur prohibitive communication costs, thereby limiting their applicability to ephemeral collaborative perception. To address these challenges, we introduce communication-efficient relative pose estimation (CERPE), a system-level framework that coordinates vision foundation models to jointly estimate ego-motion and inter-robot relative pose. CERPE reduces unnecessary raw-observation exchange by using continuously shared fixed-size descriptors to gate event-triggered raw-image requests independently of pose estimation. Non-overlapping encounters are handled by propagating inter-robot relative poses through metrically scaled ego-motion, thus maintaining relative pose estimates even in the absence of visual overlap. Experiments in simulation and real-world robots show that CERPE improves 6-DoF relative pose estimation over selected baselines in ephemeral collaborative perception.
Chinese Translation
相对姿态估计是多机器人系统中协作感知和协调的基本能力。然而，在现实环境中相遇的机器人通常在短暂的交互窗口内操作，并且必须在有限的通信带宽下工作，同时由于遮挡或视野有限导致视觉重叠间歇性或缺失。现有方法通常依赖于全局参考框架，假设持续的视图重叠，或产生高昂的通信成本，从而限制了其在短暂协作感知中的适用性。为了解决这些挑战，我们提出了通信高效相对姿态估计（CERPE），这是一个系统级框架，协调视觉基础模型共同估计自我运动和机器人间的相对姿态。CERPE通过使用持续共享的固定大小描述符来控制事件触发的原始图像请求，从而减少不必要的原始观测交换，独立于姿态估计。通过度量缩放的自我运动传播机器人间的相对姿态，从而处理非重叠的相遇，即使在缺乏视觉重叠的情况下也能保持相对姿态估计。模拟和真实机器人实验表明，CERPE在短暂协作感知中相对于选定基线提高了6自由度相对姿态估计的性能。
cs.RO / 17 / 2607.14543
SafeRelBench: A Spatial-Relation-Aware Benchmark for Process-Level Safety in VLM-Driven Embodied Agents
SafeRelBench：一种针对VLM驱动的具身智能体过程级安全的空间关系感知基准
Abstract
Vision-language models (VLMs) are increasingly used as the reasoning backbone of embodied agents, enabling robots to interpret visual scenes, follow language instructions, and plan multi-step actions. In household environments, however, safety depends not only on recognizing objects, but also on how actions change the physical scene over time. Existing embodied safety evaluations largely focus on static risk recognition, unsafe instruction refusal, or final-state task completion. As a result, process-level safety failures induced by spatial relations such as support, containment, and proximity remain insufficiently studied. To address this gap, we introduce SAFERELBENCH, a spatial-relation-aware safety benchmark with 507 executable evaluation samples, including 248 spatial-relation samples and 259 non-spatial control samples. Using SAFERELBENCH to evaluate seven open- and closed-source VLM-driven embodied agents, we find a substantial gap between task success and process-level safety compliance: models often complete the requested task while violating process-level safety constraints. Unlike prior benchmarks, SAFERELBENCH explicitly tests whether agents satisfy safety conditions before risk-prone actions, making spatial relations a core dimension in embodied safety assessment. More broadly, our results show that safe embodied intelligence requires not only stronger perception and planning, but also reliable reasoning about how object relations shape risk during interaction.
Chinese Translation
视觉语言模型（VLMs）越来越多地被用作具身智能体的推理基础，使机器人能够解释视觉场景、遵循语言指令并规划多步骤动作。然而，在家庭环境中，安全不仅依赖于物体的识别，还依赖于动作如何随时间改变物理场景。现有的具身安全评估主要集中在静态风险识别、不安全指令拒绝或最终状态任务完成。因此，由支持、包含和接近等空间关系引发的过程级安全失败仍然研究不足。为了解决这一问题，我们引入了SAFERELBENCH，这是一个具有507个可执行评估样本的空间关系感知安全基准，其中包括248个空间关系样本和259个非空间控制样本。通过使用SAFERELBENCH评估七个开源和闭源的VLM驱动的具身智能体，我们发现任务成功与过程级安全合规之间存在显著差距：模型通常在违反过程级安全约束的情况下完成请求的任务。与之前的基准不同，SAFERELBENCH明确测试智能体在风险行为之前是否满足安全条件，使空间关系成为具身安全评估的核心维度。更广泛地说，我们的结果表明，安全的具身智能不仅需要更强的感知和规划能力，还需要可靠的推理能力，以理解物体关系如何在交互过程中塑造风险。
cs.RO / 18 / 2607.14578
Beyond Implicit Force: Evaluating Explicit Force-Torque Proxies in Action Chunking with Transformers
超越隐式力：在使用变换器的动作分块中评估显式力-扭矩代理
Abstract
Contact-rich manipulation requires policies to infer interaction state from signals that are often weakly observable through vision and kinematics alone. Action Chunking with Transformers (ACT) has shown strong performance in fine-grained manipulation, but many deployments collect demonstrations through leader-follower teleoperation, where tracking error between commanded leader motion and executed follower motion implicitly encodes contact, resistance, and constraint violation. This paper examines whether ACT's apparent force-awareness depends on this hidden interaction cue. We introduce an observation-centric ACT variant that predicts future follower joint states instead of leader commands, thereby removing the teleoperation-induced discrepancy signal while preserving the rest of the learning pipeline. We then evaluate whether simple joint-torque proxies, derived from onboard motor current or joint effort, can recover contact-aware behavior without external force/torque sensors. Across four real-world tasks spanning surface following, insertion, stiffness discrimination, and force-based stopping, removing the implicit cue leads to severe failures in force-critical phases. In contrast, torque-augmented policies recover robust contact behavior and improve the base ACT policy. These results demonstrate that, on real hardware, the implicit teleoperation cue is a recoverable source of force-awareness, where torque signals are available, a simple proxy matches, surpasses, or further enhances it.
Chinese Translation
接触丰富的操作需要策略从通常仅通过视觉和运动学弱可观察的信号中推断交互状态。使用变换器的动作分块（Action Chunking with Transformers, ACT）在细粒度操作中表现出强大的性能，但许多应用通过领导-跟随远程操作收集演示，其中命令的领导运动与执行的跟随运动之间的跟踪误差隐式编码了接触、阻力和约束违反。本文探讨ACT的显性力感知是否依赖于这一隐藏的交互线索。我们引入了一种以观察为中心的ACT变体，该变体预测未来的跟随关节状态，而不是领导命令，从而消除了远程操作引起的差异信号，同时保留了学习流程的其余部分。然后，我们评估简单的关节扭矩代理，基于机载电机电流或关节努力，是否能够在没有外部力/扭矩传感器的情况下恢复接触感知行为。在四个真实世界任务中，包括表面跟随、插入、刚度辨别和基于力的停止，去除隐式线索导致在力关键阶段出现严重失败。相比之下，增强扭矩的策略恢复了稳健的接触行为，并改善了基础ACT策略。这些结果表明，在真实硬件上，隐式远程操作线索是一个可恢复的力感知来源，在扭矩信号可用的情况下，简单的代理可以匹配、超越或进一步增强它。
cs.RO / 19 / 2607.14586
SoftNav: Injecting 3D Scene Tokens into VLMs for Embodied Navigation
SoftNav：将3D场景标记注入视觉语言模型以实现具身导航
Abstract
In goal-directed embodied navigation, where an agent must locate a specified target in an unseen environment, 3D scene understanding and navigation reasoning must work in concert. Current approaches transmit 3D scene information to vision-language models (VLMs) through text, suggesting a representation gap in our tested configurations; a controlled ablation confirms that direct embedding-level transfer significantly outperforms the evaluated text serialization formats. We introduce SoftNav, which injects entity-level 3D continuous representations -- one token per detected object or frontier -- into a VLM's hidden space as soft tokens through a lightweight projector. With the 3D encoder and VLM frozen, only ~1,200 samples and ~17M trainable parameters are needed. On HM3D-OVON, SoftNav achieves 74.2%/68.3%/66.7% SR across three splits, surpassing all prior methods in both SR and SPL; the same navigation policy transfers zero-shot to GOAT-Bench (67.2% SR), SG3D (47.2% s-SR), and real-world robot deployment without retraining or architectural modification. Injecting 3D scene tokens directly into VLMs bridges the representation gap, enabling transferable navigation with minimal training.
Chinese Translation
在目标导向的具身导航中，代理必须在未见环境中定位指定目标，3D场景理解与导航推理必须协同工作。目前的方法通过文本将3D场景信息传递给视觉语言模型（VLM），这表明我们测试配置中存在表示差距；受控消融实验确认，直接嵌入级别的转移显著优于评估的文本序列化格式。我们提出了SoftNav，它通过轻量级投影器将实体级3D连续表示（每个检测到的对象或边界一个标记）作为软标记注入VLM的隐藏空间。在冻结3D编码器和VLM的情况下，仅需约1,200个样本和约1,700万个可训练参数。在HM3D-OVON上，SoftNav在三个分割中分别达到了74.2%/68.3%/66.7%的成功率（SR），超越了所有先前的方法，无论是在SR还是在成功路径长度（SPL）方面；相同的导航策略在不重新训练或修改架构的情况下，零-shot迁移到GOAT-Bench（67.2% SR）、SG3D（47.2% s-SR）以及现实世界的机器人部署。将3D场景标记直接注入VLM弥合了表示差距，使得可转移的导航在最小训练下成为可能。
cs.RO / 20 / 2607.14609
Representation-Aligned Tactile Grounding for Contact-Rich Robotic Manipulation
接触丰富的机器人操作中的表征对齐触觉基础
Abstract
Tactile-enhanced vision-language-action (VLA) policies have been introduced for contact-rich manipulation, where critical interaction states are often hidden from vision. Future tactile prediction is a promising way to use touch because it turns tactile outcomes into supervision for action-induced contact dynamics. Yet VLA policies contain representations with different roles, from perceptual encoding to motor prediction, making it unclear where this supervision should be applied. We study this as a representation-alignment problem. Through a linear probe analysis, we find that future tactile states are most predictable from intermediate action-expert features, rather than from vision-language features or final action states. Motivated by this observation, we introduce a lightweight Latent Tactile Predictor (LTP), which predicts compact future tactile embeddings from the identified intermediate representation. By avoiding direct prediction of noisy raw tactile signals, LTP provides an action-outcome grounding signal that aligns intermediate action representations with future contact consequences. Experiments on real-world contact-rich manipulation tasks show that representation-aligned tactile grounding outperforms less aligned or multi-interface tactile prediction, highlighting the importance of where tactile supervision is applied.
Chinese Translation
为接触丰富的操作引入了增强触觉的视觉-语言-动作（VLA）策略，其中关键的交互状态通常是视觉无法捕捉到的。未来触觉预测是一种有前景的触觉利用方式，因为它将触觉结果转化为对动作引发的接触动态的监督。然而，VLA 策略包含具有不同角色的表征，从感知编码到运动预测，这使得不清楚这种监督应应用于何处。我们将此研究视为一个表征对齐问题。通过线性探测分析，我们发现未来触觉状态最容易从中间动作专家特征中预测，而不是从视觉-语言特征或最终动作状态中预测。基于这一观察，我们引入了一种轻量级的潜在触觉预测器（LTP），该预测器从识别的中间表征中预测紧凑的未来触觉嵌入。通过避免直接预测噪声较大的原始触觉信号，LTP 提供了一种将中间动作表征与未来接触后果对齐的动作-结果基础信号。在真实世界的接触丰富操作任务中的实验表明，表征对齐的触觉基础优于对齐程度较低或多接口触觉预测，突显了触觉监督应用位置的重要性。
cs.RO / 21 / 2607.14639
Image-to-Point Cloud Registration Made Easy with Rectified Flow-based LiDAR Upsampling
基于校正流的激光雷达上采样简化图像到点云配准
Abstract
Image-to-Point Cloud Registration (I2P) is essential for integrating camera and LiDAR in perception and autonomous systems, yet the modality gap between images and point clouds makes it difficult to achieve both high accuracy and strong generalization. In this paper, we propose a simple yet effective I2P method that treats LiDAR as an imaging sensor: from a single sparse LiDAR scan, we generate a dense LiDAR intensity image using Conditional Rectified Flow, match it with a camera image using a pre-trained feature matcher, and estimate the 6-DoF relative pose via PnP-RANSAC. The proposed model is pre-trained through a self-supervised image completion task and fine-tuned on a small amount of LiDAR data (neither image-point cloud pairs nor ground-truth sensor poses are required), enabling it to scale to diverse LiDAR and camera configurations. Experiments on the R3LIVE dataset show that the proposed method achieves a mean error of 4.89{\deg} / 1.63 m, outperforming existing methods, while completing a single registration in approximately 0.68 s.
Chinese Translation
图像到点云配准（Image-to-Point Cloud Registration, I2P）对于在感知和自主系统中整合相机与激光雷达至关重要，但图像与点云之间的模态差异使得实现高精度和强泛化能力变得困难。本文提出了一种简单而有效的I2P方法，将激光雷达视为成像传感器：从单个稀疏的激光雷达扫描中，我们使用条件校正流（Conditional Rectified Flow）生成稠密的激光雷达强度图像，并通过预训练的特征匹配器与相机图像进行匹配，最终通过PnP-RANSAC估计6自由度的相对姿态。所提模型通过自监督图像补全任务进行预训练，并在少量激光雷达数据上进行微调（不需要图像-点云配对或真实传感器姿态），使其能够适应多样的激光雷达和相机配置。在R3LIVE数据集上的实验表明，所提方法实现了平均误差为4.89° / 1.63米，超越了现有方法，同时单次配准的时间约为0.68秒。
cs.RO / 22 / 2607.14643
NavCMPO: Critic-Guided MeanFlow Policy Optimization for Adaptive Navigation
NavCMPO：基于评论者引导的均流策略优化用于自适应导航
Abstract
End-to-end diffusion-based policies have demonstrated strong performance in mapless visual navigation, but their iterative denoising process introduces substantial inference latency, while behavior cloning limits performance to the quality of expert demonstrations. We present NavCMPO, a two-stage adaptive navigation framework that combines few-step MeanFlow trajectory generation, critic-guided refinement, and reinforcement learning fine-tuning. During pre-training, an obstacle proximity prediction task encourages the visual representation to capture obstacle-aware spatial information. To compensate for the degradation in obstacle avoidance caused by few-step generation, Critic-Guided Trajectory Refinement (CGTR) uses gradients from a critic trained with obstacle-point-cloud supervision to refine intermediate trajectories. During adaptation, the MeanFlow policy is fine-tuned using Proximal Policy Optimization with behavior-cloning regularization, while the critic is updated to accommodate embodiment-specific observation changes. Under a matched training budget on the InternVLA-N1 benchmark, NavCMPO achieves an average success rate of 74.7\%, exceeding the retrained NavDP baseline by 6.4 percentage points, while reducing inference latency from 85\,ms to 60\,ms. Experiments on a Unitree Go2 further demonstrate effective sim-to-real transfer.
Chinese Translation
基于端到端扩散的策略在无地图视觉导航中表现出强大的性能，但其迭代去噪过程引入了显著的推理延迟，而行为克隆则将性能限制在专家演示的质量上。我们提出了NavCMPO，一种结合了少步均流轨迹生成、评论者引导的细化和强化学习微调的两阶段自适应导航框架。在预训练阶段，一个障碍物接近预测任务促使视觉表示捕捉障碍物感知的空间信息。为了补偿由于少步生成导致的障碍物规避性能下降，评论者引导轨迹细化（Critic-Guided Trajectory Refinement, CGTR）使用经过障碍点云监督训练的评论者的梯度来细化中间轨迹。在适应阶段，均流策略通过带有行为克隆正则化的近端策略优化（Proximal Policy Optimization）进行微调，同时评论者被更新以适应特定体现的观察变化。在InternVLA-N1基准上，在匹配的训练预算下，NavCMPO实现了74.7%的平均成功率，比重新训练的NavDP基线高出6.4个百分点，同时将推理延迟从85毫秒降低到60毫秒。在Unitree Go2上的实验进一步证明了有效的模拟到现实迁移。
cs.RO / 23 / 2607.14675
An Intelligent-Cloud Edge Multimodal Interaction System for Robots
一种智能云边多模态交互系统用于机器人
Abstract
Robust human-robot interaction in complex environments requires accurate gesture perception, semantic scene understanding, and reliable task planning under limited onboard computing resources. This paper presents a cloud-edge multimodal interaction framework that integrates an enhanced YOLO-based gesture detector with coordinated large language model (LLM) and vision-language model (VLM) agents. The proposed detector, incorporates the Convolutional Block Attention Module (CBAM) into the neck and replaces the baseline bounding-box regression objective with Distance-IoU (DIoU) loss. These modifications improve feature discrimination and localization for small or partially occluded gestures in complex backgrounds. The cloud layer performs gesture detection, scene understanding, multimodal fusion, and action planning, whereas the TonyPi robot locally handles data acquisition, communication, action execution, and feedback. Experiments on a public gesture dataset and a custom dataset show that YOLO-DC achieves precision values of 98.9% and 95.0%, with
[email protected] values of 90.7% and 92.7%, respectively. System-level evaluation yields success rates of 95%, 88%, and 82% for single-action, composite-action, and vision-dependent tasks. A 30 participant evaluation yields an overall mean satisfaction score of 3.69 out of 5. These results demonstrate the feasibility of combining refined gesture detection with multimodal agents for resource-constrained robotic interaction.
Chinese Translation
在复杂环境中，稳健的人机交互需要准确的手势感知、语义场景理解以及在有限的机载计算资源下可靠的任务规划。本文提出了一种云边多模态交互框架，该框架将增强的基于YOLO的手势检测器与协调的大语言模型（LLM）和视觉语言模型（VLM）代理集成在一起。所提出的检测器在颈部引入了卷积块注意力模块（CBAM），并用距离IoU（DIoU）损失替代了基线边界框回归目标。这些修改提高了在复杂背景下对小型或部分遮挡手势的特征区分和定位能力。云层执行手势检测、场景理解、多模态融合和动作规划，而TonyPi机器人则在本地处理数据采集、通信、动作执行和反馈。在一个公共手势数据集和一个自定义数据集上的实验表明，YOLO-DC的精确度值分别为98.9%和95.0%，
[email protected]值分别为90.7%和92.7%。系统级评估显示单一动作、复合动作和视觉依赖任务的成功率分别为95%、88%和82%。对30名参与者的评估得出的总体满意度平均分为3.69（满分5分）。这些结果证明了在资源受限的机器人交互中，将精细的手势检测与多模态代理相结合的可行性。
cs.RO / 24 / 2607.14688
MIND-CAVs: Multi-Intelligence Negotiation and Decision System for CAVs based on Intent-Driven Autonomy
MIND-CAVs：基于意图驱动自主性的连接自动驾驶车辆的多智能体协商与决策系统
Abstract
Modern autonomous vehicles largely operate as isolated agents: they rely on on-board perception and decision modules and broadcast Basic Safety Messages (BSMs) that expose only low-level kinematic state. While existing cooperative driving frameworks enable limited sensor sharing, they rarely communicate high-level maneuver intentions, and edge computing is primarily used for content delivery rather than decision arbitration. As a result, current connected autonomy lacks a principled mechanism for making globally consistent, intent-aware coordination decisions across vehicles. To address this gap, we propose MIND-CAVs, a Multi-Intelligence Negotiation and Decision framework for connected autonomous vehicles (CAVs) based on intent-driven autonomy. Each vehicle abstracts raw sensor observations into structured intent representations, exchanges them over V2X links, and receives globally consistent coordination plans from roadside edge servers. Edge agents combine learned and rule-based arbitration mechanisms to negotiate conflicting intents among vehicles, while a cloud platform records decisions for auditing and continual retraining. We implement MIND-CAVs in a CARLA-based AI-in-the-loop platform and evaluate it in multi-lane highway scenarios involving conflicting maneuvers and route-constrained exits. Experimental results show improved maneuver completion time and reduced unsafe proximity and unnecessary braking compared with isolated autonomy, first-come-first-served arbitration, and multi-agent reinforcement learning baselines.
Chinese Translation
现代自动驾驶车辆在很大程度上作为孤立的智能体运作：它们依赖于车载感知和决策模块，并广播基本安全消息（Basic Safety Messages, BSMs），这些消息仅暴露低级运动状态。虽然现有的协作驾驶框架允许有限的传感器共享，但它们很少传达高级机动意图，而边缘计算主要用于内容传递而非决策仲裁。因此，当前的连接自主性缺乏一个原则性的机制来在车辆之间做出全球一致、关注意图的协调决策。为了解决这一问题，我们提出了MIND-CAVs，一个基于意图驱动自主性的连接自动驾驶车辆（Connected Autonomous Vehicles, CAVs）多智能体协商与决策框架。每辆车将原始传感器观察抽象为结构化的意图表示，通过车对一切（Vehicle-to-Everything, V2X）链接进行交换，并从路边边缘服务器接收全球一致的协调计划。边缘代理结合学习和基于规则的仲裁机制，协商车辆之间的冲突意图，而云平台则记录决策以便审计和持续再训练。我们在基于CARLA的AI-in-the-loop平台上实现了MIND-CAVs，并在涉及冲突机动和路线受限出口的多车道高速公路场景中进行了评估。实验结果表明，与孤立自主性、先到先服务仲裁和多智能体强化学习基线相比，机动完成时间有所改善，且不安全接近和不必要制动的情况减少。
cs.RO / 25 / 2607.14695
Reflex: Real-Time VLA Control through Streaming Inference
Reflex：通过流式推理实现实时视觉-语言-动作（VLA）控制
Abstract
Flow matching Vision-Language-Action (VLA) models promise precise continuous control, but their iterative denoising nature introduces fundamental incompatibilities with real-time robotics: global timestep injection invalidates KV-caching, forcing a choice between slow $O(N^2)$ re-computation or mathematically incorrect cache reuse. We present \textbf{Reflex}, a framework that enables \textit{real-time streaming inference} for flow matching policies by exploiting the \textit{Timestep-Invariance Property} -- that perception encoders are functionally independent of the denoising loop. Reflex partitions the attention context into static, sliding, and dynamic regions, enabling $O(1)$ incremental cache updates while preserving full-batch-equivalent attention outputs for fixed inputs. To ensure stability under continuous high-frequency inference, we introduce \textit{AdaRMSNorm}, an adaptive normalization layer that prevents BFloat16 numerical collapse by gating on flow phase. We further maximize throughput through an \textit{async pipeline} that decouples visual encoding from action generation, combined with \textit{operator fusion} that reduces kernel overhead. On LIBERO and Kinetix benchmarks, Reflex achieves a 2.58$\times$ inference speedup and 50Hz stable streaming, reducing reaction latency by up to 54\% and enabling efficient deployment without performance degradation.
Chinese Translation
流匹配视觉-语言-动作（VLA）模型承诺实现精确的连续控制，但其迭代去噪的特性与实时机器人技术存在根本的不兼容性：全局时间步注入使得KV缓存失效，迫使在缓慢的$O(N^2)$重新计算和数学上不正确的缓存重用之间做出选择。我们提出了 extbf{Reflex}，一个通过利用 extit{时间步不变性属性}实现流匹配策略的 extit{实时流式推理}框架——即感知编码器在功能上独立于去噪循环。Reflex将注意力上下文划分为静态、滑动和动态区域，使得在保持固定输入的全批次等效注意力输出的同时，实现$O(1)$增量缓存更新。为了确保在连续高频推理下的稳定性，我们引入了 extit{AdaRMSNorm}，一种自适应归一化层，通过对流相位进行门控来防止BFloat16数值崩溃。我们进一步通过 extit{异步管道}最大化吞吐量，该管道将视觉编码与动作生成解耦，并结合 extit{算子融合}以减少内核开销。在LIBERO和Kinetix基准测试中，Reflex实现了2.58$ imes$的推理加速和50Hz的稳定流式处理，将反应延迟减少了多达54 extperthousand，并实现了高效部署而不降低性能。
cs.RO / 26 / 2607.14698
Lights, Camera, Malfunction: When Illumination Robustness Leaves VLA Models Blind to Color
灯光、摄像、故障：当光照鲁棒性使得视觉-语言-行动模型对颜色失明
Abstract
Vision-Language-Action (VLA) models have emerged as a powerful paradigm for general-purpose robot manipulation; however, their transition to real-world environments reveals vulnerabilities to minor environmental perturbations. We propose FLARE, an optimized physical spotlight attack framework that exploits these vulnerabilities via targeted illuminations, dropping baseline task success rates to zero without any access to model internals. While adversarial training is the standard countermeasure, we identify a critical and previously underestimated defensive pitfall: naive data augmentations incorrectly condition VLA models to discard color as noise, collapsing their visual perception into a purely shape-biased processor. We expose this degradation through a diagnostic grayscale evaluation, in which the defended model maintains high success rates on grayscale inputs, while its success rate on benign, color-dependent real-world tasks drops to at most 47.5%, well below the undefended baseline. To address this, we propose ChromaGuard, a chroma-preserving adversarial training method. On a physical 6-DoF robotic platform, we demonstrate that ChromaGuard achieves 97.5% and 92.5% success rates in benign and attacked color-dependent tasks, respectively.
Chinese Translation
视觉-语言-行动（VLA）模型已成为通用机器人操作的强大范式；然而，它们在真实环境中的转变暴露出对微小环境扰动的脆弱性。我们提出了FLARE，一个优化的物理聚光灯攻击框架，通过有针对性的照明利用这些脆弱性，使基线任务成功率降至零，而无需访问模型内部。尽管对抗训练是标准的对策，但我们识别出一个关键且之前被低估的防御陷阱：简单的数据增强错误地使VLA模型将颜色视为噪声，从而使其视觉感知崩溃为纯粹的形状偏见处理器。我们通过诊断性灰度评估揭示了这种退化，在该评估中，经过防御的模型在灰度输入上保持高成功率，而在良性、依赖颜色的真实世界任务上的成功率最多下降至47.5%，远低于未防御基线。为了解决这个问题，我们提出了ChromaGuard，一种色度保留的对抗训练方法。在一个物理6自由度的机器人平台上，我们证明ChromaGuard在良性和攻击性依赖颜色的任务中分别实现了97.5%和92.5%的成功率。
cs.RO / 27 / 2607.14708
Reinforcement Learning for the Full Strawberry Harvesting Process: Obstacle Separation, Detachment, and Placement
全草莓采摘过程的强化学习：障碍物分离、果实脱离与放置
Abstract
Severe occlusions and deformable plant structures introduce complex contact dynamics that challenge robotic strawberry harvesting. A policy-driven reinforcement learning (RL) framework with heuristic phase coordination was developed, in which obstacle separation, fruit detachment, and placement were formulated as a sequential decision-making task. A shared interaction-aware policy generated Cartesian motions across all task phases, while lightweight heuristic logic coordinated task progression and gripper events. A shared structured observation space was used to represent target, obstacle, end-effector, and task-context information. A hierarchical architecture combined the high-level policy with low-level Cartesian impedance control for compliant interaction. To support zero-shot sim-to-real transfer, feasibility-first observation alignment and domain randomization were adopted. The policy achieved success rates of 89.7% in simulation and 82.0% in real-world experiments. As the occlusion level increased from 1 to 5, the average execution time increased from 12.99 s to 21.73 s, reflecting greater interaction complexity. These results demonstrated effective transfer of interaction-aware harvesting behaviors to a structurally different robotic platform.
Chinese Translation
严重的遮挡和可变形植物结构引入了复杂的接触动态，给机器人草莓采摘带来了挑战。我们开发了一种基于策略的强化学习（RL）框架，结合启发式阶段协调，将障碍物分离、果实脱离和放置形式化为一个顺序决策任务。一个共享的交互感知策略在所有任务阶段生成笛卡尔运动，而轻量级启发式逻辑则协调任务进展和夹持器事件。我们使用共享的结构化观察空间来表示目标、障碍物、末端执行器和任务上下文信息。一个分层架构将高层策略与低层笛卡尔阻抗控制结合，以实现顺应性交互。为了支持零-shot 的模拟到现实转移，采用了以可行性为首的观察对齐和领域随机化。该策略在模拟中的成功率达到了89.7%，在现实世界实验中达到了82.0%。随着遮挡水平从1增加到5，平均执行时间从12.99秒增加到21.73秒，反映了更大的交互复杂性。这些结果展示了交互感知采摘行为有效转移到结构不同的机器人平台。
cs.RO / 28 / 2607.14725
BridgeFlow: Fast and Robust SE(2)-Equivariant Motion Planning with Flow Matching
BridgeFlow：快速且稳健的 SE(2)-等变运动规划与流匹配
Abstract
In robotic motion planning, equivariance to rigid body transformations is crucial for robust spatial generalization. However, current learning-based planners face a critical dilemma: they either lack inherent equivariance, treating transformed tasks as novel scenarios, or enforce it via computationally expensive specialized architectures that bottleneck real-time inference. To break this trade-off, we propose BridgeFlow, a fast and strictly SE(2)-equivariant generative motion planning framework. Rather than relying on heavy equivariant networks, BridgeFlow achieves exact spatial equivariance via a lightweight task-centric canonicalization module, enabling generalization using standard architectures. To further accelerate inference, we pair a Brownian bridge informative prior with context-aware mini-batch optimal transport. This constructs a straightened vector field that minimizes transport costs and stabilizes training. Furthermore, environmental awareness is explicitly embedded via Classifier-Free Guidance. Evaluations in dense 2D environments and on a 7-DoF Franka manipulator demonstrate that BridgeFlow achieves up to a 15x inference speedup and a 2x higher valid trajectory rate over state-of-the-art diffusion baselines, alongside robust generalization to entirely unseen environments and arbitrary spatial transformations.
Chinese Translation
在机器人运动规划中，对刚体变换的等变性对于稳健的空间泛化至关重要。然而，当前基于学习的规划器面临一个关键的困境：它们要么缺乏固有的等变性，将变换后的任务视为新场景，要么通过计算开销巨大的专用架构来强制实现等变性，从而限制了实时推理的能力。为了解决这一权衡问题，我们提出了 BridgeFlow，一个快速且严格的 SE(2)-等变生成运动规划框架。BridgeFlow 通过一个轻量级的以任务为中心的标准化模块实现精确的空间等变性，而不是依赖于复杂的等变网络，从而使得使用标准架构进行泛化成为可能。为了进一步加速推理，我们将布朗桥信息先验与上下文感知的小批量最优运输相结合。这构建了一个直线化的向量场，最小化运输成本并稳定训练。此外，环境意识通过无分类器引导显式嵌入。对密集的二维环境和 7 自由度的 Franka 操作臂的评估表明，BridgeFlow 在推理速度上实现了高达 15 倍的加速，并且有效轨迹率比最先进的扩散基线高出 2 倍，同时对完全未见过的环境和任意空间变换具有稳健的泛化能力。
cs.RO / 29 / 2607.14730
Hybrid Rigid-Soft Robotic Gripper with Shape Adaptation, Uniform Force Distribution, and Self-Locking Capabilities
具有形状适应、均匀力分布和自锁能力的混合刚性-软性机器人抓手
Abstract
Conventional robotic grippers face a significant challenge in agricultural automation: the trade-off between compliant, adaptive grasping, pressure balancing among all joints, and high load capacity, often at the cost of high energy consumption. This paper presents a novel hybrid rigid-soft gripper that integrated low-cost, membrane-based pneumatic actuators with 3D-printed dual ratchet-pawl mechanisms to simultaneously achieve shape adaptation, uniform force distribution, and energy-free self-locking. The dual-ratchet structure assembled in an offset configuration significantly increased the angular resolution of the joint locking mechanism. Key experimental results demonstrated the gripper's superior performance: a remarkable maximum load capacity of 4200 g, far exceeding that of conventional soft grippers (45-210 g); more uniform force distribution across object sizes (1.75-35.29% difference ratio) compared to a rigid gripper (56.77-66.44%), with peak contact forces remaining below surface damage thresholds; and a 50.05% reduction in total energy consumption to 42.6 J per grasp cycle, achieved by eliminating the need for continuous pneumatic pressure through the self-locking mechanism, compared to 85.28 J for a conventional soft gripper. The combination of additive manufacturing for ratchets and commercially available materials for pneumatic chambers ensured a low-cost and easily fabricated design. These findings validated that the proposed gripper successfully bridged the gap between soft compliance and rigid reliability, offering a robust and efficient solution for scalable agricultural harvesting and manipulation tasks.
Chinese Translation
传统的机器人抓手在农业自动化中面临着一个重大挑战：在顺应性、适应性抓取、各关节之间的压力平衡和高负载能力之间的权衡，这通常以高能耗为代价。本文提出了一种新型的混合刚性-软性抓手，该抓手将低成本的基于膜的气动驱动器与3D打印的双棘爪机制相结合，能够同时实现形状适应、均匀力分布和无能耗自锁。以偏移配置组装的双棘爪结构显著提高了关节锁定机制的角度分辨率。关键实验结果表明，该抓手的优越性能：最大负载能力达到4200克，远超传统软性抓手（45-210克）；与刚性抓手（56.77-66.44%）相比，在物体尺寸上的力分布更加均匀（差异比为1.75-35.29%），且接触力峰值保持在表面损伤阈值以下；总能耗减少50.05%，每次抓取循环为42.6焦耳，通过自锁机制消除了对持续气动压力的需求，而传统软性抓手则为85.28焦耳。棘爪的增材制造与气动腔体的商业材料相结合，确保了低成本和易于制造的设计。这些发现验证了所提出的抓手成功弥合了软性顺应性与刚性可靠性之间的差距，为可扩展的农业收获和操作任务提供了一种稳健且高效的解决方案。
cs.RO / 30 / 2607.14781
Curvature-Constrained and Constant-Speed Distributed Simultaneous Arrival Control for Multi-Robot Systems
曲率约束与恒速分布式同时到达控制用于多机器人系统
Abstract
The simultaneous arrival of multiple mobile robots at a target point is crucial for cooperation tasks such as cooperative encirclement, disaster relief, and environmental monitoring. Although the simultaneous arrival problem itself is already complex, the problem becomes more challenging when there are constraints on the robot trajectory curvatures and the speeds are required to be constant (possibly different for different robots), and the control law for robots needs to be distributed. These constraints are typical for a multi-robot system consisting of, e.g., fixed-wing UAVs. To address this challenge, this paper proposes a distributed switching control method based on the maximum consensus protocol. By exploiting the geometric properties of Dubins paths along with optimization principles, a virtual time variable is introduced, and a hybrid control law that combines optimal control with saturated proportional control is designed. Under the proposed control law, each robot is driven to approach the maximum virtual time among its neighbors, thereby achieving simultaneous arrival under some mild conditions. Furthermore, we prove that in certain cases the proposed method attains a theoretically optimal arrival time. The approach is scalable and real-time, with low communication overhead. Its effectiveness and robustness are validated through extensive simulations and experiments.
Chinese Translation
多个移动机器人同时到达目标点对于合作任务（如合作包围、灾难救援和环境监测）至关重要。尽管同时到达问题本身已经相当复杂，但当机器人轨迹的曲率受到限制且速度需要保持恒定（可能不同于不同的机器人），同时控制律需要是分布式时，这个问题变得更加具有挑战性。这些约束在由固定翼无人机等组成的多机器人系统中是典型的。为了解决这一挑战，本文提出了一种基于最大共识协议的分布式切换控制方法。通过利用Dubins路径的几何特性和优化原理，引入了一个虚拟时间变量，并设计了一种将最优控制与饱和比例控制相结合的混合控制律。在所提出的控制律下，每个机器人被驱动接近其邻居中的最大虚拟时间，从而在一些温和条件下实现同时到达。此外，我们证明在某些情况下，所提出的方法可以达到理论上的最优到达时间。该方法具有可扩展性和实时性，通信开销低。通过大量的仿真和实验验证了其有效性和鲁棒性。
cs.RO / 31 / 2607.14826
Interventional Causal Circuits for Safe Robot Action Testing and Failure Recovery
安全机器人行动测试与故障恢复的干预因果电路
Abstract
Safe physical AI for robot actions are required not only likely to succeed but tested to be safe before execution. In practice, however, formal testing of motion parameters is computationally expensive, and the cost scales poorly with the dimensionality of the action space. When a proposed action is rejected by a tester, the naive response is to resample blindly until a passing candidate is found. This is wasteful, uninformative, and offers no convergence. We argue that rejection should instead trigger causal diagnosis: a principled identification of which action parameter caused the failure and what corrective value maximises the probability of passing testing under the interventional probability distribution. We propose a closed-loop framework that couples a Joint Probability Tree (JPT) with a Causal Circuit derived from a Marginal-Deterministic Variable Tree, enabling exact polytime computation without retraining, or additional data collection. The framework validates tractability of all interventional queries before the robot begins operating, and out-of-support candidates are detected and excluded from correction automatically. We perform experiments in a ROS2 simulation environment, and the framework demonstrates complementary roles across quality of distribution: under a high-quality JPT, the Causal Circuit reduces failed attempts by 10.3% and under a degraded JPT, it reduces total failed attempts by 37%. Every rejected plan produces a structured, interpretable causal report naming the primary cause variable, its observed value, and the recommended corrective region, supporting operator oversight and autonomous recovery without a separately trained failure model.
Chinese Translation
安全的物理人工智能（AI）用于机器人行动不仅需要成功的可能性，还需在执行前经过安全性测试。然而，在实践中，运动参数的正式测试计算成本高昂，并且随着行动空间维度的增加，成本增长不成比例。当一个提议的行动被测试者拒绝时，简单的反应是盲目重新采样，直到找到一个合格的候选方案。这种做法浪费资源、缺乏信息，并且没有收敛性。我们认为，拒绝应当触发因果诊断：系统地识别导致失败的行动参数以及什么纠正值可以在干预概率分布下最大化通过测试的概率。我们提出了一个闭环框架，该框架将联合概率树（Joint Probability Tree, JPT）与源自边际确定性变量树（Marginal-Deterministic Variable Tree）的因果电路相结合，实现了在不重新训练或额外数据收集的情况下的精确多项式时间计算。该框架在机器人开始操作之前验证所有干预查询的可处理性，并自动检测并排除不在支持范围内的候选方案。我们在ROS2仿真环境中进行了实验，结果表明该框架在分布质量方面发挥了互补作用：在高质量的JPT下，因果电路将失败尝试减少了10.3%；而在劣化的JPT下，总失败尝试减少了37%。每个被拒绝的计划都会生成一个结构化的、可解释的因果报告，指出主要原因变量、其观察值以及推荐的纠正区域，支持操作员监督和自主恢复，而无需单独训练的故障模型。
cs.RO / 32 / 2607.14842
KineFuse: Kinematic-Aware Haptic Fusion for In-Hand Occluded-Object Pose Tracking
KineFuse：运动学感知触觉融合用于手中遮挡物体姿态跟踪
Abstract
Dexterous in-hand manipulation requires continuous 6D pose tracking, yet the manipulating fingers inevitably occlude the object from the camera. We study how to structure the sparse haptic signals already available on multi-fingered hands, including proprioception, proximal force/torque, and binary contact, to complement a pretrained visual pose tracker under occlusion. We propose a kinematic-aware finger-level encoder and systematically compare it against four alternative designs through three levels of evaluation: per-frame refinement, sequential open-loop tracking, and closed-loop manipulation. Our experiments reveal that (i) per-frame evaluation cannot distinguish encoder quality, while sequential tracking amplifies architectural differences by up to 15 times; (ii) the structured encoder learns task-specific cross-modal gating, using vision exclusively for translation and dedicating one attention head to haptics for rotation, without explicit supervision; and (iii) compact finger-level tokenization with 4 tokens outperforms both flat fusion and joint-level representations, which suppress vision through norm dominance. We validate that improved tracking yields higher success in a downstream reorientation task and provide qualitative real-world demonstrations. Our project page is available at https://cold-young.github.io/kine-fuse/.
Chinese Translation
灵巧的手中操作需要持续的6D姿态跟踪，但操控手指不可避免地会遮挡相机视角下的物体。我们研究如何利用多指手上已有的稀疏触觉信号，包括本体感觉、近端力/扭矩和二元接触，来补充在遮挡情况下的预训练视觉姿态跟踪器。我们提出了一种运动学感知的指级编码器，并通过三种评估层次（逐帧细化、顺序开环跟踪和闭环操作）系统地将其与四种替代设计进行比较。我们的实验结果表明：（i）逐帧评估无法区分编码器质量，而顺序跟踪则将架构差异放大至15倍；（ii）结构化编码器学习任务特定的跨模态门控，专门使用视觉进行平移，并将一个注意力头专用于旋转的触觉，而无需显式监督；（iii）采用4个标记的紧凑指级标记化优于平面融合和关节级表示，后者通过范数主导抑制视觉。我们验证了改进的跟踪在下游重定向任务中带来了更高的成功率，并提供了定性的现实世界演示。我们的项目页面可在 https://cold-young.github.io/kine-fuse/ 找到。
cs.RO / 33 / 2607.14852
Towards Human-like Physical Intelligence: LifelongVision-Language-Action Learning for Robotic Manipulation
迈向类人物理智能：用于机器人操作的终身视觉-语言-动作学习
Abstract
Similar to the natural capabilities of humans to sequentially learn new tasks, robots with Vision-Language-Action (VLA) models should possess lifelong learning ability to learn a new task when deployed in open-world environments. However, most recently proposed lifelong learning models aim to effectively learn the current task (plasticity) or maintain high accuracy on previous tasks (stability), while the plasticity-stability trade-off remains largely unsolved in robotic manipulation models. To address this fundamental challenge, we propose a cache-efficient lifelong Vision-Language-Action learning framework for robotic manipulation (i.e., LifelongVLA), which alleviates the plasticity-stability trade-off with a dual-timescale adaptation mechanism while achieving low-cost robotic deployment with a cache-efficient replay strategy. More concretely, we propose a dual-timescale LoRA gating module to decompose VLA adaptation into two lightweight pathways: a short-term adapter for plasticity and a long-term adapter for stable consolidation. These pathways are integrated via a task-aware gate, enabling explicit control of the plasticity-stability trade-off. In the skill replay phase, a cache-efficient stochastic replay strategy is proposed to preserve more balanced retention signals without full-trajectory storage. Finally, experiments show that LifelongVLA outperforms existing baselines, demonstrating efficient skill expansion, robust retention of learned manipulation behaviors, and reduced reliance on retraining for real-world deployment on an xArm robot.
Chinese Translation
与人类自然能力顺序学习新任务相似，具备视觉-语言-动作（Vision-Language-Action, VLA）模型的机器人应具备在开放世界环境中学习新任务的终身学习能力。然而，最近提出的大多数终身学习模型旨在有效学习当前任务（可塑性）或保持对先前任务的高准确性（稳定性），而可塑性-稳定性权衡在机器人操作模型中仍然基本未得到解决。为了解决这一根本性挑战，我们提出了一种缓存高效的终身视觉-语言-动作学习框架（即，LifelongVLA），该框架通过双时间尺度适应机制缓解可塑性-稳定性权衡，同时实现低成本的机器人部署，采用缓存高效的重放策略。更具体地，我们提出了一种双时间尺度的LoRA门控模块，将VLA适应分解为两个轻量级路径：用于可塑性的短期适配器和用于稳定整合的长期适配器。这些路径通过任务感知门集成，使得可塑性-稳定性权衡的显式控制成为可能。在技能重放阶段，提出了一种缓存高效的随机重放策略，以在不进行完整轨迹存储的情况下保留更平衡的保留信号。最后，实验表明，LifelongVLA在现有基准上表现优越，展示了高效的技能扩展、对已学习操作行为的稳健保留，以及在xArm机器人上进行现实世界部署时对再训练的依赖减少。
cs.RO / 34 / 2607.14853
Modeling and Validation of Quality of Control for Edge-Offloaded Collaborative Navigation
边缘卸载协作导航控制质量的建模与验证
Abstract
Collaborative control in complex environments is severely challenged by stochastic wireless delay and reliability variations, which can degrade navigation, tracking, and collision avoidance. These network-induced uncertainties complicate the maintenance of energy efficiency during collaborative tasks, and can potentially lead to over-provisioning of resources. In this paper, for a navigation setup with dynamic collision avoidance, we address this challenge by expanding the quality of control (QoC) framework from prior works to practical robotic models. Our approach (i) models end-to-end network effects on closed-loop performance, (ii) systematically explores the impact of various control parameters dictating robotic motion on network latency-reliability (iii) validates these models through experiments on a private 5G testbed across varying delay, reliability and control configurations. Our analysis indicates the optimal control-communication co-design operating regimes for practical robots and also compares the QoC performance of standard ROS~2 quality of service (QoS) policies under real-world conditions and showing how RELIABLE QoS offers 51.5% better QoC than BEST-EFFORT under certain experimental settings.
Chinese Translation
在复杂环境中，协作控制受到随机无线延迟和可靠性变化的严重挑战，这可能会降低导航、跟踪和避碰的效果。这些网络引起的不确定性使得在协作任务中维持能效变得复杂，并可能导致资源的过度配置。本文针对具有动态避碰功能的导航设置，通过将先前工作的控制质量（Quality of Control, QoC）框架扩展到实际机器人模型，来解决这一挑战。我们的方法（i）建模了端到端网络对闭环性能的影响，（ii）系统性地探索了影响机器人运动的各种控制参数对网络延迟和可靠性的影响，（iii）通过在私人5G测试平台上进行实验，验证了这些模型在不同延迟、可靠性和控制配置下的表现。我们的分析表明了实际机器人操作的最佳控制-通信协同设计工作模式，并比较了在真实条件下标准ROS~2服务质量（Quality of Service, QoS）策略的QoC性能，显示在某些实验设置下，RELIABLE QoS比BEST-EFFORT提供了51.5%的更好QoC。
cs.RO / 35 / 2607.14899
OASIS-Map: Object-Level Change Detection in Multi-Session Mapping using Semantic Correspondence Matching
OASIS-Map：基于语义对应匹配的多会话映射中的对象级变化检测
Abstract
Map representations which are consistent across repeated visits to a real-world semi-static environment are very useful for long-term robotic inspection. In such settings, the scene may evolve while the robot is absent, with objects appearing, disappearing, moving, or being replaced, quickly making a static map outdated. Existing change-detection methods reason through geometry, category-level semantics, or object persistence. However, achieving reliable object association across revisits remains a key challenge, especially under partial views, occlusion, and imperfect segmentation. In this work, we propose OASIS-Map, a multi-session mapping system that maintains a spatio-temporally consistent object-level map by establishing dense patch-level semantic correspondences between temporal observations. These correspondences detect where the scene has changed and incrementally associate objects across revisits as the robot re-observes the environment. We demonstrate OASIS-Map on three challenging real-world scenarios: object rearrangements in 3RScan, visually similar car replacements in a car park, and large-scale scene changes in an outdoor market. We achieve 0.783 F1 on change detection in a car replacement scenario in a car park and 0.667 F1 on moved object association in 3RScan. https://dynamic.robots.ox.ac.uk/projects/oasis-map/
Chinese Translation
在对真实世界半静态环境的重复访问中，一致的地图表示对于长期机器人检查非常有用。在这种环境中，场景可能在机器人缺席时发生变化，物体出现、消失、移动或被替换，这迅速使静态地图变得过时。现有的变化检测方法通过几何、类别级语义或对象持久性进行推理。然而，在部分视图、遮挡和不完美分割的情况下，实现可靠的对象关联仍然是一个关键挑战。在本研究中，我们提出了OASIS-Map，一个多会话映射系统，通过在时间观测之间建立密集的补丁级语义对应关系，维持一个时空一致的对象级地图。这些对应关系检测场景变化的地方，并在机器人重新观察环境时逐步关联对象。我们在三个具有挑战性的真实场景中展示了OASIS-Map：3RScan中的物体重排、停车场中的视觉相似汽车替换，以及户外市场中的大规模场景变化。在停车场的汽车替换场景中，我们在变化检测中达到了0.783的F1分数，而在3RScan中的移动对象关联中达到了0.667的F1分数。
cs.RO / 36 / 2607.14919
Human-Robot Interaction in GenAI Architectures via the Agent-Client Protocol
通过代理-客户端协议实现生成人工智能架构中的人机交互
Abstract
Recent advances in Generative Artificial Intelligence (GenAI), particularly Large Language Models (LLMs), are driving robotic architectures toward agent-based high-level orchestration, in which natural-language instructions can be translated into context-aware action sequences. While the integration of these agents and robotic capabilities is increasingly converging toward standardization through the Model Context Protocol (MCP), the upper Human-Robot Interaction (HRI) layer remains fragmented by proprietary, ad hoc interfaces that hinder real-time human-in-the-loop collaboration. To address this fragmentation, this paper proposes the adoption of the Agent-Client Protocol (ACP) -- a communication standard originally introduced for coding agents in software engineering -- as a unified communication contract for the HRI layer in agent-based robotic systems. By combining ACP at the interface-agent link and MCP at the agent-execution link, we formulate a fully decoupled three-layer architecture that separates human interaction, deliberative orchestration, and physical execution. This topology removes rigid architectural dependencies, enabling heterogeneous user interfaces to connect to the same robotic system and allowing the underlying robotic platform to be replaced without requiring client-specific integration changes. Moreover, it provides native support for collaborative HRI capabilities such as real-time observability, explicit human authorization, and immediate task interruption. We experimentally evaluate the proposed architecture on a physical mobile robot, demonstrating interoperability across three heterogeneous user interfaces and validating real-time human-in-the-loop workflows with negligible latency overhead.
Chinese Translation
最近在生成人工智能（GenAI）领域的进展，特别是大型语言模型（LLMs），正在推动机器人架构向基于代理的高层次编排发展，其中自然语言指令可以转化为上下文感知的动作序列。尽管这些代理与机器人能力的整合正通过模型上下文协议（MCP）趋向标准化，但上层的人机交互（HRI）层仍然因专有的、临时的接口而碎片化，这阻碍了实时的人机协作。为了解决这一碎片化问题，本文提出采用代理-客户端协议（ACP）——一种最初为软件工程中的编码代理引入的通信标准——作为基于代理的机器人系统中HRI层的统一通信契约。通过在接口-代理链接中结合ACP和在代理-执行链接中结合MCP，我们构建了一个完全解耦的三层架构，分离了人类交互、深思熟虑的编排和物理执行。这种拓扑结构消除了僵化的架构依赖，使得异构用户接口能够连接到同一机器人系统，并允许在不需要特定于客户端的集成更改的情况下替换底层机器人平台。此外，它原生支持协作HRI能力，例如实时可观察性、明确的人类授权和即时任务中断。我们在一个物理移动机器人上对所提议的架构进行了实验评估，展示了三种异构用户接口之间的互操作性，并验证了实时人机协作工作流程的可行性，延迟开销微乎其微。
cs.RO / 37 / 2607.14943
Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal Control
通过机制可解释性和最优控制将鲁棒性引入世界行动模型
Abstract
World Action Models (WAMs) enable semantically- and physically-informed control but are brittle under distribution shift. In this work, we use mechanistic interpretability to study how robustness-relevant perturbations are represented in WAM activation space. Comparing activations across successful and unsuccessful rollouts, we find some WAM architectures exhibit low-dimensional linear separability for robustness-critical features, while others do not. This motivates the use of contrastive activation directions for training-free WAM steering. We also show that local linearity in WAM activation dynamics enables efficient feedback steering via model-based optimal control, yielding World-Action Linear Quadratic Regulator (WA-LQR), a minimally-invasive reduced-order LQR controller. Via mechanistic evaluations, we predict strong steerability in the Cosmos-Policy and DiT4DiT models but weak steerability in LingBot-VA, consistent with steering intervention results. On Cosmos-Policy and DiT4DiT, WA-LQR generalizes contrastive directions to new tasks and improves robustness to camera, gripper, and visual-noise perturbations over unsteered and prompt steering baselines.
Chinese Translation
世界行动模型（WAMs）能够实现语义和物理信息驱动的控制，但在分布变化下表现脆弱。在本研究中，我们利用机制可解释性研究鲁棒性相关扰动在WAM激活空间中的表现。通过比较成功和失败的执行过程中的激活，我们发现一些WAM架构在鲁棒性关键特征上表现出低维线性可分性，而其他架构则没有。这促使我们使用对比激活方向进行无训练的WAM引导。我们还表明，WAM激活动态中的局部线性性使得通过基于模型的最优控制实现高效反馈引导成为可能，从而产生了世界行动线性二次调节器（WA-LQR），这是一种最小侵入性的降阶LQR控制器。通过机制评估，我们预测Cosmos-Policy和DiT4DiT模型具有强引导能力，而LingBot-VA则表现出较弱的引导能力，这与引导干预结果一致。在Cosmos-Policy和DiT4DiT上，WA-LQR将对比方向推广到新任务，并提高了对相机、抓手和视觉噪声扰动的鲁棒性，相较于未引导和提示引导的基线表现更佳。
cs.RO / 38 / 2607.14997
AeroAct: Action-Centered World-Action Models for Language-Conditioned Quadrotor Flight
AeroAct：面向语言条件的四旋翼飞行的动作中心世界动作模型
Abstract
Language-conditioned quadrotor flight requires a policy to ground semantic goals, anticipate the visual consequences of ego-motion, and output control references that remain smooth and dynamically executable under rapidly changing first-person views. Existing aerial vision-language navigation and vision-language-action methods commonly use discrete actions, high-level waypoints, or instantaneous velocity commands, which provide limited supervision about how flight actions change future observations. We present AeroAct, an action-centered world-action model (WAM) for quadrotor navigation. To the best of our knowledge, AeroAct is the first WAM instantiated and demonstrated for real-world aerial flight. The model adapts a pretrained video diffusion Transformer to predict local trajectory-action chunks from egocentric visual history, proprioception, and language. Future first-person frames are used during training as dense consequence supervision, while deployment directly decodes actions without generating future video. To obtain aligned visual, state, language, and dynamically feasible action data, we build a DiffAero-based pipeline with complementary Isaac Lab and 3D Gaussian splatting renderers. We further introduce a low-cost handheld collection device that couples camera observations with motion estimates to recreate flight-like egocentric trajectories, and a self-guidance procedure that improves temporal consistency across overlapping trajectory chunks. Closed-loop simulation and real-world experiments show that temporal visual context improves target tracking and object-search performance, and that WAM-based policies can be executed on a physical quadrotor.
Chinese Translation
语言条件的四旋翼飞行需要一个策略来将语义目标落地，预测自我运动的视觉后果，并输出在快速变化的第一人称视角下仍然平滑且动态可执行的控制参考。现有的空中视觉-语言导航和视觉-语言-动作方法通常使用离散动作、高级航点或瞬时速度命令，这些方法对飞行动作如何改变未来观察的监督有限。我们提出了AeroAct，一种面向四旋翼导航的动作中心世界动作模型（WAM）。据我们所知，AeroAct是第一个在真实世界空中飞行中实例化和演示的WAM。该模型采用预训练的视频扩散Transformer，从自我中心的视觉历史、身体感知和语言中预测局部轨迹-动作片段。训练期间，未来的第一人称帧被用作密集的后果监督，而部署时则直接解码动作，而无需生成未来视频。为了获得对齐的视觉、状态、语言和动态可行的动作数据，我们构建了一个基于DiffAero的管道，并结合了互补的Isaac Lab和3D高斯点云渲染器。我们进一步引入了一种低成本的手持收集设备，将相机观察与运动估计结合起来，以重建类似飞行的自我中心轨迹，并采用自我引导程序提高重叠轨迹片段之间的时间一致性。闭环仿真和真实世界实验表明，时间视觉上下文改善了目标跟踪和物体搜索性能，并且基于WAM的策略可以在物理四旋翼上执行。
cs.RO / 39 / 2607.15004
CosFly-VLA: A Spatially Aware Vision-Language-Action Model for UAV Tracking
CosFly-VLA：一种空间感知的视觉-语言-动作模型用于无人机跟踪
Abstract
Dynamic target tracking is essential for Unmanned Aerial Vehicles (UAVs) operating in complex urban environments, where both the target and the camera viewpoint change continuously. Existing Vision-Language-Action (VLA) policies can track visible targets effectively, but their performance often degrades when buildings, vegetation, or roadside objects block the line of sight. During sustained occlusion, a policy may lose the target state, execute actions toward an incorrect region, and amplify this error through subsequent observations until re-acquisition becomes impossible. To this end, we present CosFly-VLA, a spatially aware VLA model that jointly grounds the target, estimates its visibility, and generates continuous flight actions through a structured prediction interface. To train this policy, we use a large-scale recipe over diverse data sources. Spatially Grounded Continued Pretraining (CPT) on a 500k mixed pool injects UAV-view depth, distance, and 3-D spatial reasoning. A three-stage Curriculum-based Supervised Fine-Tuning (SFT) process then specializes the tracker through multi-head warm-up followed by two-stage curriculum learning over natural and hard / long-occlusion data. Chain-of-Thought (CoT) training subsequently teaches recovery-oriented reasoning traces before structured answers. Finally, a closed-loop Reinforcement Learning (RL) stage optimizes tracking behavior with a multi-component reward covering stand-off tracking, grounding quality, collision avoidance, and task success. Relative to OpenVLA, CosFly-VLA-0.8B reduces open-loop Average Displacement Error (ADE) by 34.1% on seen-test and 35.3% on unseen-test. Closed-loop optimization improves Success Rate (SR) by 29.8% and 2.5%, respectively. These results demonstrate progress from visible-frame imitation toward spatially grounded action-closed-loop control, evaluated under a shared oracle state history.
Chinese Translation
动态目标跟踪对于在复杂城市环境中操作的无人机（UAV）至关重要，因为目标和摄像机视角都在不断变化。现有的视觉-语言-动作（VLA）策略能够有效跟踪可见目标，但在建筑物、植被或路边物体阻挡视线时，其性能往往会下降。在持续遮挡期间，策略可能会失去目标状态，向错误区域执行动作，并通过后续观察放大这一错误，直到重新获取变得不可能。为此，我们提出了CosFly-VLA，这是一种空间感知的VLA模型，能够联合定位目标、估计其可见性，并通过结构化预测接口生成连续的飞行动作。为了训练该策略，我们使用了一个大规模的配方，涵盖多样的数据源。在500k混合池上进行的空间基础持续预训练（CPT）注入了无人机视角的深度、距离和三维空间推理。随后，三阶段的基于课程的监督微调（SFT）过程通过多头热身和对自然及困难/长时间遮挡数据的两阶段课程学习来专门化跟踪器。然后，链式思维（CoT）训练在结构化答案之前教授恢复导向的推理轨迹。最后，闭环强化学习（RL）阶段通过涵盖远程跟踪、定位质量、避免碰撞和任务成功的多组件奖励来优化跟踪行为。与OpenVLA相比，CosFly-VLA-0.8B在已见测试和未见测试中分别将开放环平均位移误差（ADE）减少了34.1%和35.3%。闭环优化分别提高了成功率（SR）29.8%和2.5%。这些结果展示了从可见帧模仿向空间基础的动作闭环控制的进展，并在共享的oracle状态历史下进行了评估。
cs.RO / 40 / 2607.15016
Risk-Aware Belief Control Barrier Functions over Random Finite Sets
基于风险意识的信念控制障碍函数在随机有限集上的应用
Abstract
Ensuring robot safety in unknown, dynamic environments is a fundamental requirement. It involves inferring the states of an unknown and time-varying number of moving objects from noisy, incomplete measurements. We address safe control under the induced multi-object state uncertainty with a risk-aware belief control barrier function (BCBF) framework. The uncertainty is captured by a random finite set (RFS) belief, estimated by a sequential Monte Carlo probability hypothesis density (SMC-PHD) filter that represents it with a set of particles. Building directly on these particles, we construct a nonsmooth BCBF, establish forward invariance of the safe set under continuous prediction, and derive an explicit condition under which discrete updates preserve safety. Simulation and real-world underwater experiments demonstrate the effectiveness and efficiency of the proposed approach.
Chinese Translation
确保机器人在未知动态环境中的安全是一个基本要求。这涉及从嘈杂且不完整的测量中推断未知且时变的移动物体的状态。我们在风险意识的信念控制障碍函数（BCBF）框架下，解决了由多目标状态不确定性引起的安全控制问题。该不确定性通过随机有限集（RFS）信念来捕捉，后者由序列蒙特卡洛概率假设密度（SMC-PHD）滤波器估计，并用一组粒子表示。我们直接基于这些粒子构建了一个非光滑的BCBF，建立了在连续预测下安全集的前向不变性，并推导出离散更新保持安全的明确条件。仿真和实际水下实验验证了所提方法的有效性和效率。
cs.RO / 41 / 2607.15036
Learning Agile Navigation in Crowded Environments for Quadruped Robots
四足机器人在拥挤环境中学习灵活导航
Abstract
Navigating dynamic and crowded environments presents significant challenges for quadruped robots due to severe sensor occlusion and unpredictable human motion. Existing approaches face a trade-off: model-based methods, such as Velocity Obstacles (VO), theoretically guarantee safety but rely on accurate obstacle motion estimates that often fail in dense crowds, while end-to-end learning methods offer robustness but lack motion prediction capability of obstacles, leading to collisions or conservative behaviors. To solve this, we propose VOP-Nav, a novel navigation system that combines the geometric safety of VO with the agile adaptability of end-to-end learning. Using only local onboard observations, our system avoids explicit obstacle detection and tracking pipelines. The VOP-Net processes multi-frame LiDAR data to implicitly encode dynamic constraints and predict a safe velocity region derived from Velocity Obstacle theory. Importantly, the VO predictions serve a dual role: they are used as input to the navigation policy during inference and as a reward signal during training to encourage safe motion. Evaluations in Isaac Gym demonstrate that VOP-Nav achieves higher success rates than all baselines while balancing locomotion speed and collision avoidance. Real-world deployment on a Unitree Go2 quadruped robot further validates the system's robustness and efficiency in complex indoor and outdoor dynamic environments.
Chinese Translation
在动态和拥挤的环境中导航对四足机器人提出了重大挑战，主要由于严重的传感器遮挡和不可预测的人类运动。现有的方法面临权衡：基于模型的方法，如速度障碍（Velocity Obstacles, VO），在理论上保证安全，但依赖于准确的障碍物运动估计，而在密集人群中往往失效；而端到端学习方法则提供了鲁棒性，但缺乏对障碍物运动的预测能力，导致碰撞或保守行为。为了解决这一问题，我们提出了VOP-Nav，一种新颖的导航系统，将VO的几何安全性与端到端学习的灵活适应性相结合。我们的系统仅使用本地车载观测，避免了显式的障碍物检测和跟踪流程。VOP-Net处理多帧激光雷达数据，以隐式方式编码动态约束，并预测基于速度障碍理论得出的安全速度区域。重要的是，VO预测发挥了双重作用：在推理过程中作为导航策略的输入，并在训练过程中作为奖励信号以鼓励安全运动。在Isaac Gym中的评估表明，VOP-Nav的成功率高于所有基线，同时平衡了运动速度和碰撞避免。在Unitree Go2四足机器人上的实际部署进一步验证了该系统在复杂室内和室外动态环境中的鲁棒性和效率。
cs.RO / 42 / 2607.15065
DriftWorld: Fast World Modeling through Drifting
漂移世界：通过漂移实现快速世界建模
Abstract
Predictive world models enable robots to plan by imagining the outcomes of their actions, but their value for control hinges on generating many rollouts quickly. This creates a bottleneck for diffusion-based world models: multistep sampling makes each rollout expensive, limiting large-scale action search at inference time. We introduce DriftWorld, an action-conditioned world model based on drifting generative models. Rather than denoising iteratively at inference, DriftWorld learns an action-conditioned drift during training, allowing it to generate future frames from the current observation and a candidate action sequence in a single forward pass at 30+ fps, which is 17x faster on average than diffusion based baselines. We evaluate DriftWorld on standard vision-based robotic manipulation benchmarks, including Bridge-V2, RT-1, Language Table, Push-T, and Robomimic. By producing rollouts that are both accurate and fast, DriftWorld achieves state-of-the-art decision-making performance with far less inference time than diffusion-based world model baselines. Beyond online control, DriftWorld can also serve as an offline simulator for ranking real-world robot policies, with rollout-based scores correlating with ground truth at up to 0.99. These results show that drifting models are a strong fit for robot world modeling, where fast, high-quality imagination directly supports planning and policy evaluation.
Chinese Translation
预测性世界模型使机器人能够通过想象其行动的结果进行规划，但其在控制中的价值取决于快速生成大量的模拟结果。这为基于扩散的世界模型创造了瓶颈：多步采样使得每次模拟变得昂贵，从而限制了推理时的大规模动作搜索。我们提出了 DriftWorld，一种基于漂移生成模型的动作条件世界模型。与其在推理时迭代去噪不同，DriftWorld 在训练过程中学习动作条件漂移，使其能够在单次前向传播中以超过30帧每秒的速度从当前观测和候选动作序列生成未来帧，这比基于扩散的基线平均快17倍。我们在标准的基于视觉的机器人操作基准上评估了 DriftWorld，包括 Bridge-V2、RT-1、Language Table、Push-T 和 Robomimic。通过生成既准确又快速的模拟结果，DriftWorld 实现了比基于扩散的世界模型基线更少推理时间的最先进决策性能。除了在线控制，DriftWorld 还可以作为离线模拟器，用于对现实世界机器人策略进行排名，其基于模拟的评分与真实情况的相关性高达0.99。这些结果表明，漂移模型非常适合机器人世界建模，其中快速、高质量的想象直接支持规划和策略评估。
cs.RO / 43 / 2607.15111
Goal-Oriented Semantic Communication for Distributed ISAC-Enabled Vehicle Coordination
面向目标的语义通信在分布式集成感知与通信（ISAC）支持下的车辆协调
Abstract
Vehicle coordination at unsignalized intersections relies on accurate real-time vehicle state acquisition and reliable command-and-control (C&C) signal delivery. However, existing studies typically treat sensing, communication, and control separately, which may lead to redundant transmissions, outdated state information, and unreliable vehicle coordination. In this paper, we investigate a new scenario of distributed integrated sensing and communication (ISAC)-enabled vehicle coordination at intersections, where multiple roadside units (RSUs) collaboratively transmit sensing signals for vehicle state acquisition and C&C signals for vehicle movement control under the management of a central base station (BS). To improve signaling efficiency, we propose a unified goal-oriented semantic communication (GSC) framework, which transmits sensing and C&C signals only when they are semantically important for improving intersection traffic throughput. Specifically, an extended Kalman filter (EKF) is adopted to predict vehicle states and fuse distributed sensing measurements. A masked hybrid proximal policy optimization (MHPPO) framework is then developed to jointly determine sensing transmission decisions, C&C transmission decisions, and C&C signal contents based on a value-of-information (VoI) reward. Furthermore, we propose an uncertainty-aware transmission design (UTD), including robust beamforming and VoI-based time-division power allocation, to improve sensing and communication reliability under vehicle state uncertainty and inter-RSU interference. Simulation results show that our proposed framework achieves 100% collision-free vehicle coordination with significantly reduced signaling overhead compared with predictive ISAC baselines adapted from state-of-the-art related studies and several ablation baselines.
Chinese Translation
在无信号交叉口，车辆协调依赖于准确的实时车辆状态获取和可靠的指挥与控制（C&C）信号传递。然而，现有研究通常将感知、通信和控制分开处理，这可能导致冗余传输、过时的状态信息和不可靠的车辆协调。本文探讨了一种新的场景，即在交叉口的分布式集成感知与通信（ISAC）支持下的车辆协调，其中多个路边单元（RSU）协同传输感知信号以获取车辆状态，并在中央基站（BS）的管理下传递C&C信号以控制车辆运动。为了提高信号效率，我们提出了一种统一的面向目标的语义通信（GSC）框架，该框架仅在信号对提高交叉口交通通行能力具有语义重要性时才传输感知和C&C信号。具体而言，采用扩展卡尔曼滤波器（EKF）来预测车辆状态并融合分布式感知测量。随后，开发了一种掩蔽混合近端策略优化（MHPPO）框架，以基于信息价值（VoI）奖励共同确定感知传输决策、C&C传输决策和C&C信号内容。此外，我们提出了一种不确定性感知传输设计（UTD），包括鲁棒波束成形和基于VoI的时分功率分配，以提高在车辆状态不确定性和RSU间干扰下的感知和通信可靠性。仿真结果表明，与从最先进相关研究中改编的预测ISAC基线和几种消融基线相比，我们提出的框架实现了100%的无碰撞车辆协调，并显著减少了信号开销。
cs.RO / 44 / 2607.15129
Catch, Throw, Repeat: Planning for Human-Robot Partner Juggling
捕捉、投掷、重复：人机合作杂耍的规划
Abstract
Dynamic object exchange between humans and robots remains a challenging problem due to uncertainty in perception, timing, and contact-rich interaction. Human-robot juggling represents a particularly demanding instance of this problem, requiring precise real-time coordination, predictive motion planning with feedback control, and robustness to variability in human motion. Enabling such skills is of interest for advancing physical human-robot interaction and shared autonomy. We present a real-time planning and control architecture for human-robot partner juggling that enables a robot to reliably catch and throw balls in synchronized multi-ball patterns with a human partner. The system integrates predictive ball tracking, adaptive online trajectory optimization using a multiple-shooting formulation, and a state-machine-based coordination logic to enable synchronized multi-ball human-robot partner juggling. In a user study with 8 participants of varying juggling skill from beginner to expert, we demonstrate that our system can achieve three-ball cascades shared between the robot and the human. All participants exceeded previously reported best-case results within a 10-minute test session, with one participant extending the previous record for shared three-ball cascade juggling fivefold to 20 consecutive robot catches, and another participant achieving a 100% success rate with 40 consecutive catches in a single-ball catch-and-return setting. Video documentation can be found at https://kai-ploeger.com/partner-juggling
Chinese Translation
人类与机器人之间的动态物体交换仍然是一个具有挑战性的问题，主要由于感知、时机和接触丰富的互动中的不确定性。人机杂耍代表了这一问题的一个特别复杂的实例，要求精确的实时协调、带反馈控制的预测运动规划，以及对人类运动变异性的鲁棒性。实现这些技能对于推动物理人机交互和共享自主性具有重要意义。我们提出了一种实时规划和控制架构，用于人机合作杂耍，使机器人能够与人类合作，在同步的多球模式中可靠地捕捉和投掷球体。该系统集成了预测性球体跟踪、基于多次发射形式的自适应在线轨迹优化，以及基于状态机的协调逻辑，以实现同步的多球人机合作杂耍。在一项包含8名不同杂耍技能参与者（从初学者到专家）的用户研究中，我们展示了我们的系统能够实现机器人与人类共享的三球瀑布式杂耍。所有参与者在10分钟的测试中均超过了先前报告的最佳结果，其中一名参与者将共享三球瀑布杂耍的先前记录提高了五倍，达到了20次连续的机器人接球，另一名参与者在单球接回设置中实现了40次连续接球的100%成功率。视频记录可在 https://kai-ploeger.com/partner-juggling 找到。
cs.RO / 45 / 2607.15156
Assessing Physical Frailty and Fall-Risk Indicators with Social Robots: An in situ Evaluation with Older Adults
利用社交机器人评估身体虚弱和跌倒风险指标：对老年人的现场评估
Abstract
Frailty assessments are crucial to evaluate the risk of adverse events and the health and social care needs of older adults, yet their administration remains resource-intensive and typically relies on coarse clinical outcomes, such as task completion times, which may overlook biomechanical indicators of functional decline. To address this, we present a robotic framework that guides older adults through standardised frailty and fall-risk tests while capturing clinical scores and additional frailty-related metrics, offering a deeper insight into a user's condition. The system uses a Behaviour Tree architecture that coordinates perception, decision-making, interaction, and measurement modules. Using vision-based skeleton tracking, the robot evaluates established clinical tests, including the Short Physical Performance Battery (SPPB) and the Timed Up and Go (TUG). The framework was co-designed with healthcare professionals and evaluated in situ during six months in a rehabilitation centre's research lab with N=81 older adults. Robot-derived measurements were compared against therapist assessments and clinical reference instruments, including a gait analysis walkway and an inertial measurement unit (IMU). Results showed excellent agreement for most test completion times and gait-related parameters ($ICC > 0.9$). And, substantial agreement for the overall SPPB score comparing the robot and the therapist ($k = 0.67$) and moderate agreement comparing the robot and the IMU ($k=0.55$). The findings highlight that social robots can provide reliable and objective frailty assessments in healthcare settings while enabling the collection of relevant mobility indicators beyond conventional outcomes.
Chinese Translation
虚弱评估对于评估老年人不良事件的风险以及健康和社会护理需求至关重要，但其实施仍然资源密集，通常依赖粗略的临床结果，如任务完成时间，这可能忽视功能衰退的生物力学指标。为了解决这一问题，我们提出了一种机器人框架，该框架引导老年人通过标准化的虚弱和跌倒风险测试，同时捕捉临床评分和其他与虚弱相关的指标，从而提供对用户状况的更深入洞察。该系统采用行为树架构，协调感知、决策、互动和测量模块。通过基于视觉的骨骼跟踪，机器人评估已建立的临床测试，包括短期身体表现电池（Short Physical Performance Battery, SPPB）和定时起立走（Timed Up and Go, TUG）。该框架与医疗专业人员共同设计，并在康复中心的研究实验室进行了为期六个月的现场评估，参与者为81名老年人。机器人获得的测量结果与治疗师评估及临床参考工具（包括步态分析走道和惯性测量单元（Inertial Measurement Unit, IMU））进行了比较。结果显示，大多数测试完成时间和步态相关参数之间的符合度极高（$ICC > 0.9$）。整体SPPB评分在机器人与治疗师之间的符合度显著（$k = 0.67$），而机器人与IMU之间的符合度为中等（$k=0.55$）。研究结果强调社交机器人能够在医疗环境中提供可靠和客观的虚弱评估，同时能够收集超出传统结果的相关移动指标。
cs.RO / 46 / 2607.15163
Scaling Behavior Foundation Model for Humanoid Robots
人形机器人行为基础模型的扩展性研究
Abstract
Humanoid control requires natural whole-body coordination, precise real-time responses to control signals, and robust generalization across diverse environmental contexts, making it a cornerstone for generalist embodied agents. Behavior Foundation Models (BFMs) have recently emerged as a promising solution to address these challenges by leveraging large-scale behavioral data to achieve superior expressiveness, versatility and generalization. However, despite growing interest in scaling BFMs to further improve their capabilities, it remains unclear how key factors, including the learning paradigm, behavioral data and model architecture should be coordinated to enable effective scaling. In this work, we revisit the scaling recipe for BFMs and demonstrate that substantial performance gains can be achieved through the coordination of three core components: 1) the learning paradigm of motion tracking that reformulates diverse humanoid control problems as the reproduction of integrated whole-body behaviors in the global frame; 2) the strategic synergy between on-policy rollout quantity and reference motion diversity; and 3) the expressive and scalable model architecture termed Humanoid Transformer that facilitates the natural emergence of structured behavioral representations. Through extensive experiments in both simulation and real-world deployment, we demonstrate that our approach yields significant improvements in control fidelity and task generalization, reducing Mean Per-Keypoint Position Error (MPKPE) on the test set by over 10% in local mode and 82% in global mode compared with existing humanoid controllers. These results establish BFM as a principled and effective foundation for scalable and general-purpose humanoid control.
Chinese Translation
人形控制需要自然的全身协调、对控制信号的精准实时响应，以及在多样化环境背景下的强健泛化能力，这使其成为通用具身智能体的基石。行为基础模型（Behavior Foundation Models, BFMs）最近作为一种有前景的解决方案出现，通过利用大规模的行为数据来实现卓越的表现力、多样性和泛化能力。然而，尽管对扩展 BFMs 以进一步提升其能力的兴趣日益增长，但关键因素（包括学习范式、行为数据和模型架构）如何协调以实现有效扩展仍不明确。在本研究中，我们重新审视了 BFMs 的扩展策略，并展示了通过协调三个核心组件可以实现显著的性能提升：1）运动跟踪的学习范式，将多样的人形控制问题重新定义为在全局框架中再现综合的全身行为；2）在线策略执行数量与参考运动多样性之间的战略协同；3）一种称为人形变换器（Humanoid Transformer）的表现力和可扩展性模型架构，促进结构化行为表征的自然出现。通过在模拟和现实世界部署中的广泛实验，我们证明了我们的方法在控制精度和任务泛化方面显著改善，与现有的人形控制器相比，在局部模式下测试集的每关键点位置误差（Mean Per-Keypoint Position Error, MPKPE）降低超过 10%，在全局模式下降低 82%。这些结果确立了 BFM 作为可扩展和通用人形控制的原则性和有效基础。
cs.RO / 47 / 2607.15172
AHEAD: Anticipatory Hand-Driven Teleoperation via Human Intent Prediction
AHEAD：基于人类意图预测的预判式手控远程操作
Abstract
Direct hand-driven teleoperation maps an operator's hand motion to robot end-effector commands at every frame, enabling precise control, but it requires constant monitoring and correction during approach, grasp, and placement, which can be slow and fatiguing. For repetitive pick-and-place tasks, supervisory (goal-based) teleoperation simplifies this process: the operator specifies goals/waypoints, and the robot executes the motion using planning algorithms. Yet, this introduces latency, as the robot must wait for the next command before it can plan and act. "How can we reduce robot reaction time while lowering operator workload?" To tackle this question, we present AHEAD, a real-time VR teleoperation system that anticipates operator intent to enable proactive, hand-driven control. In a digital twin, the operator performs pick-and-place naturally, using hand motion to convey high-level commands rather than a continuous robot trajectory. AHEAD processes a short window of 3D hand and head signals together with scene context through an attention-based classifier to predict the intended grasp object and placement slot. A state machine converts intent predictions into stable robot goals, enabling early motion while remaining stable under noisy predictions and corrective hand movements. AHEAD's intent prediction module achieves Top1 accuracy: 76% for grasp objects and 76% for target slots. Moreover, our user study shows AHEAD reduces robot reaction latency by 0.6 s (object) and 1.4 s (slot) relative to baselines. Participants also reported lower operator load, indicating faster robot responses while maintaining low operator effort in practice.
Chinese Translation
直接的手控远程操作将操作者的手部动作映射到机器人末端执行器的指令，实现精确控制，但在接近、抓取和放置过程中需要持续监控和纠正，操作过程缓慢且易疲劳。对于重复的拾取-放置任务，监督式（基于目标）远程操作简化了这一过程：操作者指定目标或路径点，机器人通过规划算法执行动作。然而，这会引入延迟，因为机器人必须等待下一条指令才能进行规划和执行。“如何在降低操作者负担的同时减少机器人反应时间？”为解决该问题，我们提出了AHEAD，一种实时虚拟现实远程操作系统，通过预测操作者意图实现主动的手控操作。在数字孪生环境中，操作者自然地执行拾取-放置任务，利用手部动作传达高层次指令，而非连续的机器人轨迹。AHEAD通过基于注意力机制的分类器，结合短时间窗口内的三维手部和头部信号及场景上下文，预测预期的抓取物体和放置位置。状态机将意图预测转化为稳定的机器人目标，使机器人能够提前启动动作，同时在预测噪声和纠正性手部动作下保持稳定。AHEAD的意图预测模块在抓取物体和目标位置的Top1准确率均达到76%。此外，用户研究表明，AHEAD相比基线方法分别减少了机器人反应延迟0.6秒（物体）和1.4秒（位置）。参与者还报告操作者负担降低，表明机器人响应更快的同时实际操作中操作者付出较低的努力。
cs.RO / 48 / 2607.15275
RoboTTT: Context Scaling for Robot Policies
RoboTTT：机器人策略的上下文扩展
Abstract
Recent robot foundation models operate with single-step or short-history visuomotor context. We introduce Test-Time-Training Robot Policies (RoboTTT), a robot model and training recipe that scale visuomotor context to 8K timesteps, three orders of magnitude beyond state-of-the-art policies, without growing inference latency. At this context length, we unlock new robot capabilities: one-shot in-context imitation from human video demonstrations, on-the-fly policy improvement, robustness to perturbations, and stronger performance on multi-stage, long-horizon tasks. We also observe, for the first time, steady gains in closed-loop performance as pretraining context length scales. At its core, RoboTTT integrates Test-Time Training into robot foundation models such as Vision-Language-Action policies, yielding a sequence model whose recurrent state consists of fast weights, parameters updated by gradient descent during both training and inference, compressing histories into weight space and retrieving contextual information for long-context conditioning. To scale training context length, the recipe combines sequence action forcing with truncated backpropagation through time. On challenging real-robot manipulation tasks, RoboTTT improves overall performance by 87% over the single-step context baseline and fully completes a five-minute, ten-stage assembly task, which no baseline ever does. RoboTTT trained with 8K-timestep context outperforms the same model pretrained with 1K timesteps by 62%, suggesting context length as a new scaling axis for robot foundation models. Videos are available at https://research.nvidia.com/labs/gear/robottt/
Chinese Translation
近期的机器人基础模型通常在单步或短历史的视觉运动上下文中运行。我们提出了测试时训练机器人策略（RoboTTT），这是一种机器人模型和训练方案，能够将视觉运动上下文扩展到8K时间步，超出当前最先进策略三个数量级，而不会增加推理延迟。在这个上下文长度下，我们解锁了新的机器人能力：从人类视频演示中进行一次性上下文模仿、即时策略改进、对扰动的鲁棒性，以及在多阶段、长时间任务上的更强性能。我们还首次观察到，随着预训练上下文长度的增加，闭环性能稳步提升。RoboTTT的核心在于将测试时训练集成到机器人基础模型中，如视觉-语言-动作（Vision-Language-Action）策略，形成一个序列模型，其递归状态由快速权重组成，这些参数在训练和推理过程中通过梯度下降进行更新，将历史压缩到权重空间，并检索上下文信息以进行长上下文条件化。为了扩展训练上下文长度，该方案结合了序列动作强制和时间截断反向传播。在具有挑战性的真实机器人操作任务中，RoboTTT的整体性能比单步上下文基线提高了87%，并且成功完成了一个五分钟、十阶段的组装任务，这是任何基线模型都无法做到的。使用8K时间步上下文训练的RoboTTT比使用1K时间步预训练的相同模型提高了62%，这表明上下文长度是机器人基础模型的新扩展轴。视频可在 https://research.nvidia.com/labs/gear/robottt/ 查看。
cs.CV / 1 / 2607.14189
MultiRef-Compass: Towards Comprehensive Evaluation of Multi-Reference-to-Audio-Video Generation
MultiRef-Compass：迈向多参考音视频生成的综合评估
Abstract
Multi-reference-to-audio-video (MR2AV) generation aims to generate coherent audio-video content conditioned on multiple references and textual instructions. Existing benchmarks mainly focus on text-driven generation, single-reference subject preservation, or isolated audio-video alignment, leaving the emerging MR2AV setting largely unexplored. Compared with these settings, MR2AV requires models to jointly reason over multiple references while generating synchronized visual and audio content. Models must not only preserve each reference faithfully but also correctly bind and compose multiple referenced entities into coherent audio-visual events. To address this gap, we introduce MultiRef-Compass, a unified benchmark for MR2AV generation. It comprises $350$ carefully curated samples constructed through a scalable and controllable asset-composition pipeline, covering multi-view subject preservation, multi-entity binding, and human-object-scene composition. To provide interpretable assessment, MultiRef-Compass defines an evaluation protocol with four dimensions: Basic Quality, Reference Consistency, Audio-Visual Consistency, and Instruction Following, using 14 sub-metrics. MultiRef-Compass integrates automatic metrics with a rejudging-enhanced MLLM-as-a-Judge framework, enabling scalable and auditable evaluation of both perceptual fidelity and reference-conditioned composition. Extensive experiments on eight representative MR2AV systems reveal substantial room for improvement across multiple evaluation dimensions, underscoring the need for a comprehensive benchmark and positioning MultiRef-Compass as a foundation for future MR2AV research.
Chinese Translation
多参考音视频（MR2AV）生成旨在基于多个参考和文本指令生成连贯的音视频内容。现有基准主要集中于文本驱动生成、单一参考主体保留或孤立的音视频对齐，导致新兴的MR2AV设置在很大程度上未被探索。与这些设置相比，MR2AV要求模型在生成同步的视觉和音频内容时，能够共同推理多个参考。模型不仅必须忠实地保留每个参考，还需正确绑定和组合多个参考实体，以形成连贯的视听事件。为了解决这一空白，我们引入了MultiRef-Compass，这是一个统一的MR2AV生成基准。它包含350个经过精心策划的样本，这些样本通过可扩展和可控的资产组合管道构建，涵盖多视角主体保留、多实体绑定和人-物-场景组合。为了提供可解释的评估，MultiRef-Compass定义了一个包含四个维度的评估协议：基本质量、参考一致性、音视频一致性和指令遵循，使用14个子指标。MultiRef-Compass将自动化指标与增强重评的MLLM-as-a-Judge框架结合，能够实现可扩展和可审计的感知保真度和参考条件组合的评估。在对八个代表性的MR2AV系统进行广泛实验后，发现多个评估维度存在显著的改进空间，强调了综合基准的必要性，并将MultiRef-Compass定位为未来MR2AV研究的基础。
cs.CV / 2 / 2607.14194
Inference-Time Concept Suppression and Video-Centric Evaluation for Text-to-Video Models
推理时概念抑制与视频中心评估用于文本到视频模型
Abstract
Text-to-video (T2V) generators can synthesize realistic and temporally coherent videos, but controllably removing a target concept from a generator remains difficult. Unlike text-to-image concept erasure, T2V unlearning must suppress a target concept that may persist across frames while preserving non-target subjects, actions, scenes, and temporal structure. We propose \textbf{SIRUS}, a training-free inference-time framework for concept-level T2V unlearning. Given textual aliases of a target concept, SIRUS localizes target-related prompt evidence and suppresses target expression during sampling, without updating the text encoder or denoising network. We further introduce a video-oriented evaluation framework for T2V unlearning that separately measures target forgetting, non-target preservation, video quality, jailbreak robustness, and efficiency, using video-level failure criteria, frame-level residue statistics, paired preservation analysis, VBench-based quality diagnostics, and deployment overhead measurement. Across five safety, object, and style concepts on CogVideoX, SIRUS reaches 70.4\% average forgetting success and 25.7\% average frame hit, compared with 44.4\% / 47.2\% for VideoEraser, while reducing the average VBench quality drop from -0.043 to -0.016, yielding the strongest forgetting-quality trade-off among fully evaluated baselines. Transfer experiments on Wan2.2 further suggest that SIRUS generalizes across modern T2V backbones.
Chinese Translation
文本到视频（T2V）生成器能够合成现实且时间上连贯的视频，但可控地从生成器中移除目标概念仍然困难。与文本到图像的概念消除不同，T2V 的遗忘必须抑制可能在帧之间持续存在的目标概念，同时保留非目标主体、动作、场景和时间结构。我们提出了 extbf{SIRUS}，一种无训练的推理时框架，用于概念级 T2V 遗忘。给定目标概念的文本别名，SIRUS 定位与目标相关的提示证据，并在采样过程中抑制目标表达，而无需更新文本编码器或去噪网络。我们进一步引入了一种面向视频的 T2V 遗忘评估框架，分别测量目标遗忘、非目标保留、视频质量、越狱鲁棒性和效率，使用视频级失败标准、帧级残留统计、配对保留分析、基于 VBench 的质量诊断和部署开销测量。在 CogVideoX 上的五个安全、物体和风格概念中，SIRUS 达到了 70.4\% 的平均遗忘成功率和 25.7\\% 的平均帧命中率，而 VideoEraser 的对应值为 44.4\\% / 47.2\\%，同时将平均 VBench 质量下降从 -0.043 降低到 -0.016，呈现出在完全评估的基准中最强的遗忘-质量权衡。对 Wan2.2 的迁移实验进一步表明，SIRUS 在现代 T2V 主干网络中具有良好的泛化能力。
cs.CV / 3 / 2607.14202
KeyFrame-Compass: Towards Comprehensive Evaluation of Keyframe-Conditioned Video Generation
关键帧指南针：全面评估关键帧条件视频生成
Tang, Yuqi, Liu, Tengfei, Lai, Yizheng, Wang, Yuran, Shi, Yang, Su, Wanshun, Zhang, Zhuoran, Wang, Qixun, Zhang, Xiaohan, Yu, Xinlei, Bai, Xuehai, Zhu, Xuanyu, Zeng, Bohan, Li, Bozhou, Li, Shujie, Dai, Yifan, Wei, Yujie, Liu, Shixuan, Wang, Haotian, Chen, Jialu, Zhang, Yuanxing
Abstract
Video generation increasingly relies on keyframe-based workflows, where creators specify a sequence of reference images to guide generation. Although recent models support multi-keyframe conditioning, it remains unclear whether they can faithfully reproduce the prescribed keyframes while maintaining overall video quality. We present KeyFrame-Compass, the first comprehensive benchmark for evaluating keyframe-conditioned video generation. The benchmark contains 386 carefully curated samples spanning three application domains, two video structures, two prompt granularities, two conditioning formats, and four keyframe densities, enabling controlled analysis under diverse generation settings. We further introduce an automated evaluation framework that jointly measures keyframe execution and overall video quality. Specifically, we decompose keyframe execution into six complementary metrics covering presence, fidelity, temporal ordering, localization, persistence, and uniqueness, while assessing overall video quality through evidence-grounded MLLM judgments augmented with specialized perception models. Experiments on nine representative video generation systems reveal several fundamental limitations. Current models exhibit a clear trade-off between faithful keyframe execution and natural video synthesis. Their performance further degrades as keyframe constraints become denser and most open-source models also fail to interpret storyboard-grid inputs as temporally ordered keyframe sequences.
Chinese Translation
视频生成越来越依赖于基于关键帧的工作流程，创作者指定一系列参考图像以指导生成。尽管最近的模型支持多关键帧条件，但尚不清楚它们是否能够忠实再现所规定的关键帧，同时保持整体视频质量。我们提出了关键帧指南针（KeyFrame-Compass），这是第一个全面的基准，用于评估关键帧条件的视频生成。该基准包含386个精心策划的样本，涵盖三个应用领域、两种视频结构、两种提示粒度、两种条件格式和四种关键帧密度，从而能够在多样化的生成设置下进行控制分析。我们进一步引入了一种自动化评估框架，联合测量关键帧执行和整体视频质量。具体而言，我们将关键帧执行分解为六个互补指标，涵盖存在性、保真度、时间顺序、定位、持久性和独特性，同时通过基于证据的多模态语言模型（MLLM）判断，结合专业的感知模型来评估整体视频质量。对九个代表性的视频生成系统的实验揭示了几个基本局限性。目前的模型在忠实的关键帧执行和自然的视频合成之间存在明显的权衡。随着关键帧约束变得更加密集，它们的性能进一步下降，而大多数开源模型也未能将故事板网格输入解释为时间顺序的关键帧序列。
cs.CV / 4 / 2607.14228
SeeSE3: Emergence of 3D Space in Vision Features
SeeSE3：视觉特征中三维空间的出现
Abstract
In this paper, we ask whether vision foundation models construct representations that reflect the intrinsic properties of 3D Euclidean space. Unlike previous works that probe 3D awareness of vision features by regressing image-centric quantities such as depth or normals, we investigate the relation between the structure of the space of visual features and the group of Euclidean transformations $SE(3)$. We propose a set of probes to evaluate this relation from both topological and geometric perspectives: a mutual neighborhood metric that measures the alignment between feature neighborhoods and spatial topology, and a Poincar\'e Adapter to test the linear accessibility of the geometry of camera motion from latent displacements in static scenes. We show that self-supervised vision models, which, in principle, have not been trained with direct 3D supervision or active agency, possess latent subspaces that are remarkably strongly correlated with three-dimensional Euclidean space, when probed correctly. Building on this insight we propose a new class of "Latent-Space Navigation" techniques that perform visual odometry and localization purely in the latent space, bypassing the need for explicit 3D reconstruction.
Chinese Translation
在本文中，我们探讨视觉基础模型是否构建了反映三维欧几里得空间内在属性的表征。与以往通过回归图像中心量（如深度或法线）来探测视觉特征的三维意识的研究不同，我们研究视觉特征空间的结构与欧几里得变换群 $SE(3)$ 之间的关系。我们提出了一组探测工具，从拓扑和几何的角度评估这一关系：一种互邻域度量，用于测量特征邻域与空间拓扑之间的对齐程度，以及一个庞加莱适配器（Poincaré Adapter），用于测试从静态场景中的潜在位移到相机运动几何的线性可达性。我们展示了自监督视觉模型在原则上并未接受直接的三维监督或主动代理训练，但在正确探测时，其潜在子空间与三维欧几里得空间之间存在显著的强相关性。基于这一见解，我们提出了一类新的“潜在空间导航”技术，能够在潜在空间中纯粹执行视觉里程计和定位，绕过显式三维重建的需求。
cs.CV / 5 / 2607.14248
3D Lane Detection with Odometry for High-Speed Vehicle Racing
基于里程计的高速赛车3D车道检测
Abstract
Lane boundary detection is a critical component in autonomous driving systems and has been rigorously studied in regular driving scenarios. However, it is less explored in vehicle racing, where the car moves at higher speeds across more extreme road geometries. To study this problem, we introduce a new dataset for 3D lane detection in racing, featuring >$250$k images from multiple camera feeds and inertial measurements taken with a Lexus LC 500 driving on a closed circuit. With this dataset, we compare various approaches to 3D lane detection and propose modifications that permit frames to be processed at rates of almost 300Hz while retaining high predictive performance in the racing application. This facilitates a multi-camera ensemble approach that is validated on hardware. We show that sensing modalities such as inertial measurements can be leveraged for pre-integration to regress road geometries over both cameras and time, yielding improvements in key metrics. Compared to methods such as BevLaneDet, adding odometry and ensemble predictions improves the F1 score by 3 points and reduces near-vehicle mean absolute errors (MAEs) by $>30 \%$. We show F1 scores $>$0.9 and lateral MAEs of $<$0.18m in vehicle deployments.
Chinese Translation
车道边界检测是自动驾驶系统中的关键组成部分，并且在常规驾驶场景中得到了严格的研究。然而，在赛车中，由于车辆以更高的速度行驶并穿越更极端的道路几何形状，这一领域的研究相对较少。为了解决这个问题，我们引入了一个新的数据集，用于赛车中的3D车道检测，该数据集包含来自多个摄像头和使用雷克萨斯 LC 500 在封闭赛道上行驶时获取的惯性测量的超过250,000张图像。利用该数据集，我们比较了多种3D车道检测方法，并提出了修改方案，使得帧处理速率接近300Hz，同时在赛车应用中保持高预测性能。这促进了一种多摄像头集成方法，并在硬件上进行了验证。我们展示了惯性测量等传感方式可以用于预集成，以回归摄像头和时间上的道路几何形状，从而在关键指标上取得了改进。与 BevLaneDet 等方法相比，添加里程计和集成预测将 F1 分数提高了3分，并将近车道的平均绝对误差（MAE）减少了超过30%。我们在车辆部署中展示了 F1 分数超过0.9，横向 MAE 小于0.18米。
cs.CV / 6 / 2607.14264
MonteRET: AI Agent Enhancing Multimodal LLMs with Multi-granularity Knowledge Retrieval for Chest CT Report Generation
MonteRET：通过多粒度知识检索增强多模态大语言模型的胸部CT报告生成的人工智能代理
Abstract
Automated chest CT report generation remains challenging because clinically faithful reporting requires both whole-volume understanding and accurate description of localized anatomical findings. Here we developed and retrospectively evaluated MonteRET, a region-aware retrieval-enhanced framework for generating chest CT findings sections. MonteRET integrates global CT features with region-level anatomical representations, retrieves clinically relevant knowledge using predicted medical conditions and region-level vision-language alignment, and refines initial reports through a knowledge-guided report rewriting agent. We trained our model on a public cohort with 24,128 CT scans from RadGenome-ChestCT. We evaluated MonteRET on the public RadGenome-ChestCT test set of 1,564 CT scans and an external cohort of 82 CT scans from NewYork-Presbyterian/Weill Cornell Medical Center. MonteRET improved report quality, semantic similarity, and clinical efficacy compared with a matched baseline and several state-of-the-art methods. Gains were most pronounced for recall, suggesting fewer omitted findings. Human expert evaluation by radiology residents also favored MonteRET.
Chinese Translation
自动化胸部CT报告生成仍然面临挑战，因为临床上真实的报告需要对整个体积的理解和对局部解剖发现的准确描述。在此，我们开发并回顾性评估了MonteRET，一个区域感知的检索增强框架，用于生成胸部CT发现部分。MonteRET将全局CT特征与区域级解剖表示相结合，利用预测的医学条件和区域级视觉-语言对齐检索临床相关知识，并通过知识引导的报告重写代理来完善初步报告。我们在RadGenome-ChestCT的一个公共队列上训练了我们的模型，该队列包含24,128个CT扫描。我们在公共的RadGenome-ChestCT测试集（1,564个CT扫描）和来自纽约-长老会/韦尔康奈尔医学中心的外部队列（82个CT扫描）上评估了MonteRET。与匹配的基线和几种最先进的方法相比，MonteRET提高了报告质量、语义相似性和临床有效性。增益在召回率上最为明显，表明遗漏发现较少。放射科住院医师的人工专家评估也偏向于MonteRET。
cs.CV / 7 / 2607.14287
XCT-SAM: Sequential Parameter-Efficient Domain Adaptation of SAM for Industrial XCT Defect Segmentation
XCT-SAM：用于工业XCT缺陷分割的SAM顺序参数高效领域适应
Abstract
Defect segmentation in additive manufacturing (AM) X-ray computed tomography (XCT) images remains challenging due to severe class imbalance and large distribution shifts across scan conditions. Although recent foundation models such as the Segment Anything Model (SAM) provide strong general-purpose segmentation priors, their natural-image pre-training transfers poorly to the AM XCT domain, where defects appear as subtle non-semantic microstructural anomalies. Moreover, adapting SAM to the AM domain is further limited by the large domain gap and scarcity of labeled real XCT data. We present XCT-SAM, a sequential parameter-efficient adaptation framework for AM XCT defect segmentation. Instead of adapting SAM directly from natural images to XCT data, we first fine-tune Conv-LoRA adapters on an alloy-microstructure dataset and subsequently transfer the adapted model to XCT images, progressively bridging the domain gap. Using Conv-LoRA adapters with rank r=2, the framework injects convolutional spatial inductive bias into SAM's backbone while training approximately 4.15M parameters and keeping over 99% of the model frozen. We evaluate XCT-SAM on out-of-distribution CycleGAN-XCT benchmarks and real-world NIST XCT scans. Across both settings, XCT-SAM consistently outperforms zero-shot SAM and other domain-adapted SAM baselines, achieving the best overall IoU and Dice scores. These results demonstrate the effectiveness of intermediate domain adaptation with parameter-efficient adapters for industrial XCT defect segmentation. The source code is publicly available at https://github.com/Mahedi-61/XCT-SAM.git
Chinese Translation
在增材制造（AM）X射线计算机断层扫描（XCT）图像中，缺陷分割仍然面临严重的类别不平衡和扫描条件下的大规模分布转移的挑战。尽管最近的基础模型如Segment Anything Model（SAM）提供了强大的通用分割先验，但其在自然图像上的预训练在AM XCT领域的迁移效果较差，因为缺陷表现为微妙的非语义微观结构异常。此外，将SAM适应到AM领域还受到领域差距大和标记真实XCT数据稀缺的限制。我们提出了XCT-SAM，一种用于AM XCT缺陷分割的顺序参数高效适应框架。我们并不是直接将SAM从自然图像适应到XCT数据，而是首先在合金微观结构数据集上微调Conv-LoRA适配器，然后将适应后的模型转移到XCT图像中，逐步缩小领域差距。使用秩为r=2的Conv-LoRA适配器，该框架在训练约415万参数的同时，将卷积空间归纳偏置注入到SAM的主干中，并保持超过99%的模型被冻结。我们在分布外的CycleGAN-XCT基准和真实世界的NIST XCT扫描上评估了XCT-SAM。在这两种设置中，XCT-SAM始终优于零-shot SAM和其他领域适应的SAM基线，取得了最佳的整体IoU和Dice分数。这些结果证明了使用参数高效适配器进行中间领域适应在工业XCT缺陷分割中的有效性。源代码已公开，地址为https://github.com/Mahedi-61/XCT-SAM.git
cs.CV / 8 / 2607.14305
DCVC-MB: Neural B-Frame Video Compression using State Space Models
DCVC-MB：基于状态空间模型的神经B帧视频压缩
Abstract
In this paper we propose DCVC-Mamba (DCVC-MB), a neural video codec framework for B-frame coding. Our approach incorporates an IBP frame strategy for low-delay B-frame coding, a spatio-temporal fusion model based on state-space models for bidirectional temporal prediction, and an entropy-aware skipping mechanism that selectively omits coding certain latents to reduce entropy coding times. In addition to our model contributions we also implement two inference-time strategies that enhance compression performance. Experimental evaluation shows that DCVC-MB compares favorably to existing NVCs and traditional codecs. The method demonstrates BD-rate reductions of up to $8.98\%$ on average compared to prior neural video codecs, and improvements of up to $30.45\%$ and $1.81\%$ over the VTM-19.0-LDP and VTM-19.0-RA(Inter-GoP=16) benchmarks, respectively, contributing to advances in neural video compression.
Chinese Translation
本文提出了DCVC-Mamba（DCVC-MB），一种用于B帧编码的神经视频编解码框架。我们的方法结合了低延迟B帧编码的IBP帧策略、基于状态空间模型的时空融合模型用于双向时间预测，以及一种熵感知跳过机制，选择性地省略某些潜在编码以减少熵编码时间。除了我们的模型贡献外，我们还实施了两种推理时策略，以增强压缩性能。实验评估表明，DCVC-MB与现有的神经视频编解码器（NVCs）和传统编解码器相比表现良好。该方法在与先前的神经视频编解码器相比，平均实现了高达$8.98\%$的BD-rate降低，并在VTM-19.0-LDP和VTM-19.0-RA（Inter-GoP=16）基准上分别实现了高达$30.45\\%$和$1.81\\%$的改进，为神经视频压缩的进展做出了贡献。
cs.CV / 9 / 2607.14333
SD-MAR: Multi-image Analytical Reasoning via Synthetic Data and Reinforcement Learning
SD-MAR：通过合成数据和强化学习进行多图像分析推理
Abstract
Vision Language Models (VLMs) demonstrate strong perceptual abilities but remain limited in tasks requiring analytical reasoning across multiple visual states, such as multi-image comparison, change detection, and multi-step visual inference. These capabilities are critical for real-world multimodal applications where reasoning must be grounded in systematic differences between visual contexts. However, existing benchmarks rarely require both explicit visual comparison and analytical reasoning, leaving this capability underexplored. To address this gap, we introduce SD-MAR (Synthetic Data for Multi-image Analytical Reasoning), a framework for training and evaluating VLMs on multi-image analytical reasoning. SD-MAR constructs paired visual scenarios through controlled perturbations and generates reasoning tasks spanning semantic change attribution and quantitative comparison. We further train VLMs using GRPO-lite with Backward Discounted Allocation (BDA), a reinforcement learning approach that removes KL regularization to encourage stronger policy optimization while allocating greater credit to the later reasoning steps where analytical conclusions are formed. Experiments on Qwen2.5-VL-7B and InternVL3-8B show that GRPO-lite fine-tuning on SD-MAR improves in-domain accuracy by up to 36.95%, with Qwen2.5-VL-7B outperforming GPT-4.1 on the SD-MAR benchmark. Importantly, out-of-domain generalization is preserved or improved: performance remains within 1% on MME, MMMU-Pro, and MathVista, while improving by up to 4% on MMBench. LLM-as-judge evaluation further demonstrates consistent improvements in logical coherence and explanation quality across both models.
Chinese Translation
视觉语言模型（VLMs）展现出强大的感知能力，但在需要跨多个视觉状态进行分析推理的任务中仍然有限，例如多图像比较、变化检测和多步骤视觉推理。这些能力对于现实世界的多模态应用至关重要，因为推理必须基于视觉上下文之间的系统性差异。然而，现有基准测试很少同时要求明确的视觉比较和分析推理，使得这一能力未得到充分探索。为了解决这一问题，我们提出了SD-MAR（合成数据用于多图像分析推理），这是一个用于训练和评估VLMs在多图像分析推理上的框架。SD-MAR通过控制扰动构建成对的视觉场景，并生成涵盖语义变化归因和定量比较的推理任务。我们进一步使用GRPO-lite与向后折扣分配（BDA）进行VLMs的训练，这是一种强化学习方法，去除了KL正则化，以鼓励更强的策略优化，同时对后期形成分析结论的推理步骤分配更大的信用。在Qwen2.5-VL-7B和InternVL3-8B上的实验表明，SD-MAR上的GRPO-lite微调使得领域内准确率提高了最多36.95%，其中Qwen2.5-VL-7B在SD-MAR基准测试中超越了GPT-4.1。重要的是，领域外的泛化能力得以保持或改善：在MME、MMMU-Pro和MathVista上的表现保持在1%以内，而在MMBench上提高了最多4%。LLM-as-judge评估进一步表明，两种模型在逻辑一致性和解释质量上均有持续改善。
cs.CV / 10 / 2607.14334
MixCompress: Mixture of Experts for Variable Rate Learned Image Compression
MixCompress：用于可变比特率学习图像压缩的专家混合模型
Abstract
Learned image compression (LIC) is bottlenecked by the need to store independent models for each rate-distortion operating point. Existing variable bit-rate (VBR) methods aim to reduce this overhead via dense parameter modulation, but forcing a shared backbone to approximate divergent mappings causes severe feature entanglement. Specifically, low-rate smoothing gradients inherently conflict with the preservation of high-frequency textural details, leading to sub-optimal performance. To resolve this, we propose MixCompress, a unified VBR framework based on sparse structural specialization. While sparsely gated Mixture-of-Experts (MoE) routing successfully mitigates gradient conflict, it operates on a fixed computational budget. To address the increased representational demands of higher bit-rates we introduce a Mixture-of-Depths (MoD) extension to dynamically scale model capacity. Combined with Conditional Auxiliary Transforms (CAT) for dynamic sub-band energy modulation, our hierarchical framework effectively dynamically scales capacity. Extensive evaluations demonstrate that MixCompress not only matches individually optimized single-rate baselines but can even surpass them, establishing a new Pareto frontier for computationally efficient image coding.
Chinese Translation
学习图像压缩（LIC）受到需要为每个率-失真操作点存储独立模型的限制。现有的可变比特率（VBR）方法旨在通过密集参数调制来减少这种开销，但强迫共享主干网络来近似不同的映射会导致严重的特征纠缠。具体而言，低比特率平滑梯度与高频纹理细节的保留本质上存在冲突，导致性能不佳。为了解决这个问题，我们提出了MixCompress，一个基于稀疏结构专业化的统一VBR框架。虽然稀疏门控的专家混合模型（MoE）路由成功缓解了梯度冲突，但它在固定的计算预算下运行。为了应对更高比特率的增加表示需求，我们引入了一种深度混合模型（MoD）扩展，以动态调整模型容量。结合条件辅助变换（CAT）进行动态子带能量调制，我们的层次框架有效地动态调整容量。广泛的评估表明，MixCompress不仅匹配单独优化的单比特率基准，还能够超越它们，建立了计算高效图像编码的新帕累托前沿。
cs.CV / 11 / 2607.14338
Beyond scalar losses: calibrating segmentation models via gradient vector field surgery
超越标量损失：通过梯度向量场手术校准分割模型
Abstract
Region-based loss functions, such as the Dice loss, have established themselves as the de facto standard for highly class- and region-imbalanced segmentation tasks. However, models trained using region-based loss functions are notoriously miscalibrated and typically yield over-confident predictions. In medical imaging applications, such as defining tumor resection margins, this miscalibration is hindering clinical adoption. In this work, we outline a novel gradient perspective on this overconfidence and show how it affects region-based loss functions. We propose a "surgery" on the gradient vector field as a simple, yet effective intervention to mitigate calibration issues. This surgery adds a factor to the loss's partial derivative, scaling the gradient's magnitude linearly with the prediction error. In empirical evaluations across 2D and 3D medical segmentation tasks, we demonstrate the effectiveness of this intervention while maintaining high prediction accuracy when used in conjunction with any region-based loss function.
Chinese Translation
基于区域的损失函数，如Dice损失，已成为高度类别和区域不平衡分割任务的事实标准。然而，使用基于区域的损失函数训练的模型通常存在严重的校准问题，通常会产生过于自信的预测。在医学成像应用中，例如定义肿瘤切除边缘，这种校准不准确阻碍了临床应用。在本研究中，我们提出了一种新的梯度视角来分析这种过自信，并展示其如何影响基于区域的损失函数。我们提出了一种对梯度向量场进行“手术”的方法，作为一种简单而有效的干预措施，以减轻校准问题。该手术在损失的偏导数中添加了一个因子，使梯度的大小与预测误差线性缩放。在对二维和三维医学分割任务的实证评估中，我们展示了这一干预措施的有效性，同时在与任何基于区域的损失函数结合使用时保持高预测准确性。
cs.CV / 12 / 2607.14350
Dynamic Manipulation Hypergraphs for HAR: Beyond Pairwise Relations: Dynamic Manipulation Hypergraphs for Vision-Based Human Activity Recognition
动态操控超图在基于视觉的人类活动识别中的应用：超越成对关系
Abstract
Fine-grained manipulation recognition requires modeling evolving relations among hands, objects, tools, and supporting surfaces. Conventional graph-based methods use pairwise edges that can fragment a coordinated event into disconnected binary relations. We propose a dynamic manipulation hypergraph framework that represents multi-entity configurations as higher-order relational units. At each temporal step, relevant entities are encoded using appearance, spatial, motion, and semantic-role features. Hyperedge candidates are instantiated and ranked using proximity, contact, and motion-coupling predicates. A hypergraph reasoning network performs node-to-hyperedge and hyperedge-to-node message passing, followed by temporal attention over the evolving interaction structure. The framework provides class-agnostic hyperedge-importance scores that identify entity configurations and temporal intervals emphasized by the model without treating them as causal explanations. Quantitative evaluation is conducted on EPIC-KITCHENS-100/VISOR and Assembly101 under an annotation-assisted entity-localization protocol. Video-only and entity-based methods provide contextual comparisons, while a matched pairwise graph and a static hypergraph serve as the principal controlled baselines because they use identical entity inputs and comparable relational settings. The proposed method improves HO-F1 over the matched pairwise graph by 6.9 percentage points on EPIC-KITCHENS-100/VISOR and 9.5 points on Assembly101, and exceeds the static hypergraph by 4.4 and 5.8 points, respectively. Qualitative analysis on ARCTIC further shows correspondence between highly ranked hyperedges and contact-rich manipulation intervals. These results demonstrate the value of time-varying higher-order relational modeling for fine-grained manipulation activity recognition.
Chinese Translation
细粒度的操控识别需要对手、物体、工具和支撑表面之间不断变化的关系进行建模。传统的基于图的方法使用成对边缘，这可能将协调事件分割成不相连的二元关系。我们提出了一种动态操控超图框架，将多实体配置表示为高阶关系单元。在每个时间步，相关实体通过外观、空间、运动和语义角色特征进行编码。超边候选通过接近性、接触和运动耦合谓词进行实例化和排序。超图推理网络执行节点到超边和超边到节点的信息传递，随后对不断演变的交互结构进行时间注意力处理。该框架提供类无关的超边重要性评分，识别模型强调的实体配置和时间间隔，而不将其视为因果解释。在EPIC-KITCHENS-100/VISOR和Assembly101上进行定量评估，采用注释辅助的实体定位协议。视频仅和基于实体的方法提供了上下文比较，而匹配的成对图和静态超图作为主要的对照基线，因为它们使用相同的实体输入和可比的关系设置。所提出的方法在EPIC-KITCHENS-100/VISOR上比匹配的成对图提高了6.9个百分点的HO-F1，在Assembly101上提高了9.5个百分点，分别超过静态超图4.4和5.8个百分点。对ARCTIC的定性分析进一步显示高排名超边与富含接触的操控时间间隔之间的对应关系。这些结果展示了时间变化的高阶关系建模在细粒度操控活动识别中的价值。
cs.CV / 13 / 2607.14415
$K$-NeAS: Scalable Multi-Material CT Reconstruction Using Neural SDFs
$K$-NeAS：基于神经隐式表面函数的可扩展多材料CT重建
Abstract
Computed Tomography (CT) carries significant ionizing radiation risks, driving the need for sparse-view reconstruction. Implicit scene representations (ISRs) address this by recovering continuous volumetric attenuation fields directly from sparse projections, and recent geometry-aware extensions jointly model surface geometry alongside attenuation to improve fidelity and enable clean tissue segmentation without manual thresholding. However, these methods remain limited by manually tuned attenuation bounds and rigid two-material constraints. This paper proposes $K$-NeAS, a unified and scalable architecture for automated, multi-material surface reconstruction. We replace independent material networks with a shared latent backbone and introduce a fully differentiable $K$-material sequential soft selector to model an arbitrary number of overlapping tissues. To eliminate manual tuning, we automate attenuation bounding using a Gaussian Mixture Model (GMM) and implement a scheduled auxiliary floater loss to mitigate geometric hallucinations common under extreme sparsity. Evaluated across four clinical Cone-Beam CT (CBCT) datasets, $K$-NeAS successfully scales to arbitrary material counts, achieving superior 3D volumetric fidelity at $K=3$ materials on complex multi-tissue regions such as the Abdomen ($33.28\text{ dB}$ 3D PSNR vs. $31.40\text{ dB}$ single-material NeAS baseline, a $+1.88\text{ dB}$ improvement). Furthermore, our model exhibits enhanced robustness under sparse-sampling conditions, outperforming baseline 3D PSNR by up to $1.17\text{ dB}$ under 5- and 10-view constraints.
Chinese Translation
计算机断层扫描（CT）存在显著的电离辐射风险，这推动了稀疏视图重建的需求。隐式场景表示（ISR）通过直接从稀疏投影中恢复连续的体积衰减场来解决这一问题，最近的几何感知扩展联合建模表面几何与衰减，以提高保真度并实现无手动阈值的清晰组织分割。然而，这些方法仍然受到手动调节的衰减边界和刚性双材料约束的限制。本文提出了$K$-NeAS，一种统一且可扩展的自动化多材料表面重建架构。我们用共享的潜在主干替代独立的材料网络，并引入一个完全可微的$K$-材料顺序软选择器，以建模任意数量的重叠组织。为了消除手动调节，我们使用高斯混合模型（GMM）自动化衰减边界，并实施计划辅助浮动损失，以减轻在极端稀疏情况下常见的几何幻觉。在四个临床锥束CT（CBCT）数据集上的评估表明，$K$-NeAS成功扩展到任意材料数量，在复杂的多组织区域如腹部实现了优越的3D体积保真度（$K=3$材料时，$33.28 ext{ dB}$ 3D PSNR对比$31.40 ext{ dB}$单材料NeAS基线，提升了$+1.88 ext{ dB}$）。此外，我们的模型在稀疏采样条件下表现出增强的鲁棒性，在5视图和10视图约束下，3D PSNR比基线提高了最多$1.17 ext{ dB}$。
cs.CV / 14 / 2607.14423
Emergent Region-Level Facial Correspondence in Frozen Vision Foundation Models
冻结视觉基础模型中的新兴区域级面部对应关系
Abstract
Frozen self-supervised vision models can align parts of generic objects, but it remains unclear whether this correspondence extends to human faces, where global layout is shared while identity-specific appearance varies sharply. We test whether frozen DINOv3 features define a region-level facial coordinate system: a feature space in which eyes, brows, nose, mouth, skin, and hair remain distinguishable across people and across time without face-specific training. Using DINOv3 ViT-L/16 patch embeddings and FaRL only as a face-part labeling interface, we evaluate cross-identity nearest-neighbor matching and temporal label propagation on 200 CelebDF-v2 real videos. DINOv3 achieves 83.0% region-level semantic accuracy under unconstrained cross-identity matching, compared with a 23.0% area-weighted random baseline, and 95.5% temporal tracking accuracy without a learned temporal module. A no-FaRL control collapses to 0.9%, showing that FaRL supplies semantic initialization while DINOv3 supplies dense spatial correspondence. The strongest correspondence appears at an intermediate layer: block 18 gives a 4.93x same-region versus cross-region discrimination ratio, compared with 1.48x at the final block. Against CLIP ViT-L/14, DINOv3 shows only a small aggregate advantage but a +16.8 pp gain on anatomical regions, indicating that image-level contrastive supervision captures coarse facial layout but not fine-grained anatomical identity. These results establish frozen DINOv3 as a strong zero-shot representation for region-level facial correspondence and identify intermediate self-supervised features as the most useful layer for dense face analysis.
Chinese Translation
冻结的自监督视觉模型能够对通用物体的部分进行对齐，但尚不清楚这种对应关系是否扩展到人脸，因为人脸的全球布局是共享的，而特定身份的外观则变化明显。我们测试冻结的 DINOv3 特征是否定义了一个区域级面部坐标系统：在这个特征空间中，眼睛、眉毛、鼻子、嘴巴、皮肤和头发在不同个体和时间之间保持可区分，而无需针对面部的特定训练。我们使用 DINOv3 ViT-L/16 补丁嵌入和 FaRL 仅作为面部部件标记接口，评估在 200 个 CelebDF-v2 真实视频上的跨身份最近邻匹配和时间标签传播。在不受限制的跨身份匹配下，DINOv3 实现了 83.0% 的区域级语义准确率，相比之下，区域加权随机基线为 23.0%，而在没有学习的时间模块的情况下，时间跟踪准确率为 95.5%。没有 FaRL 的对照实验降至 0.9%，显示 FaRL 提供了语义初始化，而 DINOv3 提供了密集的空间对应关系。最强的对应关系出现在中间层：第 18 块的同区域与跨区域区分比为 4.93 倍，而最后一块为 1.48 倍。与 CLIP ViT-L/14 相比，DINOv3 仅显示出小幅的整体优势，但在解剖区域上获得了 +16.8 个百分点的提升，表明图像级对比监督捕获了粗略的面部布局，但未能捕获细粒度的解剖身份。这些结果确立了冻结的 DINOv3 作为区域级面部对应关系的强大零样本表示，并识别出中间自监督特征是进行密集面部分析的最有用层。
cs.CV / 15 / 2607.14445
Cotton-SF YOLO: Learning Structural and Frequency Cues for Early Cotton Square Detection in Complex Field Environments
Cotton-SF YOLO：在复杂田间环境中学习结构和频率线索以实现早期棉花花蕾检测
Abstract
Cotton squares are important phenotypic indicators of the early reproductive growth of cotton, and automatic field detection of cotton squares provides an important basis for cotton growth monitoring and precision cultivation management. However, early cotton square detection in complex field environments remains insufficiently explored, as cotton squares are small, frequently occluded, easily blurred, subject to illumination variations, and exhibit low contrast against surrounding cotton leaves. To address these challenges, we propose a task-oriented framework based on YOLO26m, named Cotton-SF YOLO, for cotton square detection under natural field conditions. To improve the perception of small and irregular cotton square boundaries, we introduce Dynamic Snake Convolution into the detector, enabling adaptive extraction of deformable edge features. Furthermore, a frequency-domain feature modulation module is designed by incorporating spectral enhancement into the C2f structure, which recalibrate frequency-domain representations and strengthen discriminative edge and texture cues while reducing interference from complex cotton leaf backgrounds. Trained and evaluated on our newly constructed and annotated field dataset with manually annotated cotton squares, the proposed model achieves mAP$_{50}$, mAP$_{50:95}$, and recall values of 0.8196, 0.4942, and 0.7939, improving over the baseline YOLO26m by 1.25%, 3.45%, and 2.96%, respectively. Ablation experiments and visualization demonstrate that the best performance is achieved with the complementary effects of structural and frequency cues.
Chinese Translation
棉花花蕾是棉花早期生殖生长的重要表型指标，自动化田间棉花花蕾检测为棉花生长监测和精准栽培管理提供了重要基础。然而，在复杂田间环境中，早期棉花花蕾检测仍然未得到充分探索，因为棉花花蕾体积小、易被遮挡、容易模糊、受光照变化影响，并且与周围棉叶的对比度较低。为了解决这些挑战，我们提出了一种基于YOLO26m的任务导向框架，命名为Cotton-SF YOLO，用于自然田间条件下的棉花花蕾检测。为了改善对小型和不规则棉花花蕾边界的感知，我们将动态蛇卷积（Dynamic Snake Convolution）引入检测器，使其能够自适应地提取可变形的边缘特征。此外，我们通过将光谱增强融入C2f结构，设计了一个频域特征调制模块，该模块重新校准频域表示，增强了区分性的边缘和纹理线索，同时减少了复杂棉叶背景的干扰。在我们新构建并标注的田间数据集上进行训练和评估，该数据集包含手动标注的棉花花蕾，所提模型在mAP$_{50}$、mAP$_{50:95}$和召回率方面分别达到了0.8196、0.4942和0.7939，分别比基线YOLO26m提高了1.25%、3.45%和2.96%。消融实验和可视化结果表明，结构和频率线索的互补效应实现了最佳性能。
cs.CV / 16 / 2607.14470
G$^2$SR: Geometric Methods for Fast and Memory-Efficient Gaussian-based Surface Reconstruction
G$^2$SR：用于快速且内存高效的基于高斯的表面重建的几何方法
Abstract
Few-view surface reconstruction recovers the visible surfaces of a scene from a few posed RGB images, providing the 3D models that robots need to explore and interact online. On mobile platforms, the reconstruction must be fast and geometrically accurate while keeping a small memory footprint to ensure safe and efficient operation. 3D Gaussian Splatting (3DGS) offers a high-fidelity scene representation, but building it from a few views is ill-posed, as many distinct surfaces reproduce the same images, making traditional photometric methods prone to "floater" artifacts. End-to-end methods resolve the ambiguity by regressing splats with large, usually Transformer-based, networks that require heavy compute and memory while generalizing poorly to new scenes. We propose G2SR, which exploits a well-posed core of the task: given cross-view 2D splat correspondences, 3D splats follow analytically from multi-view geometry. G2SR employs a lightweight neural frontend to detect and track 2D Gaussian splats on the image plane and an analytic backend to triangulate each into a metric-scale 3D splat. On ScanNet, Replica, and DTU, G2SR matches or exceeds the geometric accuracy of state-of-the-art end-to-end methods while running at 69-89 reconstructions per second within 203 MB of GPU memory (5-107x less) for 2- and 3-view inputs at 384 x 512 resolution, offering a practical path to online Gaussian-based surface reconstruction.
Chinese Translation
少视图表面重建从少量已定位的RGB图像中恢复场景的可见表面，为机器人探索和在线交互提供所需的3D模型。在移动平台上，重建必须快速且几何上准确，同时保持小的内存占用，以确保安全和高效的操作。3D高斯点云（3D Gaussian Splatting, 3DGS）提供了高保真的场景表示，但从少量视图构建它是一个不适定问题，因为许多不同的表面会重现相同的图像，使得传统的光度方法容易产生“漂浮”伪影。端到端方法通过回归大规模的、通常基于Transformer的网络来解决这种模糊性，这些网络需要大量的计算和内存，同时在新场景上泛化能力较差。我们提出了G2SR，它利用了任务的良好适定核心：给定交叉视图的2D高斯点云对应关系，3D高斯点云可以通过多视图几何学进行解析推导。G2SR采用轻量级神经前端在图像平面上检测和跟踪2D高斯点云，并使用解析后端将每个点云三角化为度量尺度的3D高斯点云。在ScanNet、Replica和DTU数据集上，G2SR的几何准确性与最先进的端到端方法相匹配或超过，同时在203 MB的GPU内存下以每秒69-89次的速度运行，适用于384 x 512分辨率的2视图和3视图输入，为在线基于高斯的表面重建提供了实用的路径。
cs.CV / 17 / 2607.14481
Immediate 3D Gaussian Splat Reconstruction of Unordered Input with Global Consistency
无序输入的即时3D高斯点云重建与全局一致性
Abstract
3D Gaussian Splatting (3DGS) has become the method of choice for reconstructing and real-time rendering of captured scenes. To capture a scene with good visual quality, continuous image sequences are usually combined with out-of-order shots for better scene coverage. Structure from motion can reconstruct such captures, but only after they are all available and often with high computational cost. Incremental reconstruction methods -- often derived from SLAM solutions -- provide immediate feedback, but cannot handle the out-of-order capture we require. We provide the first immediate feedback solution for such radiance field capture that provides global consistency. We first introduce a method for fast matching in out-of-order sequences, by repurposing visual place recognition models and a covisibility graph, and provide an efficient way to find highly connected keyframes, improving quality even for ordered sequences. We show how these steps -- together with GPU optimization and careful Gaussian primitive placement -- provide fast local reconstruction, in our challenging radiance field reconstruction case. We then introduce a novel cluster-based method, again using the covisibility graph, to provide efficient loop closure that does not require sequential input. Finally, to handle large scenes in our context, we introduce a progressive hierarchy that allows our method to scale to large environments, without compromising efficiency. Our results show we provide immediate feedback 3DGS reconstruction with good visual quality in several datasets, with up to thousands of input images.
Chinese Translation
3D高斯点云重建（3DGS）已成为捕获场景的重建和实时渲染的首选方法。为了以良好的视觉质量捕获场景，通常将连续的图像序列与无序拍摄结合，以获得更好的场景覆盖。运动结构可以在所有捕获图像可用后进行重建，但通常伴随高计算成本。增量重建方法——通常源自SLAM解决方案——提供即时反馈，但无法处理我们所需的无序捕获。我们提供了首个针对这种辐射场捕获的即时反馈解决方案，确保全局一致性。我们首先介绍了一种快速匹配无序序列的方法，通过重新利用视觉位置识别模型和共视图图，提供了一种高效的方式来找到高度连接的关键帧，甚至改善了有序序列的质量。我们展示了这些步骤如何结合GPU优化和精确的高斯原语放置，在我们具有挑战性的辐射场重建案例中实现快速局部重建。接着，我们引入了一种新颖的基于聚类的方法，再次利用共视图图，提供高效的回环闭合，而无需顺序输入。最后，为了在我们的背景下处理大场景，我们引入了一种渐进式层次结构，使我们的方法能够扩展到大环境，而不影响效率。我们的结果表明，在多个数据集中，我们提供了即时反馈的3DGS重建，具有良好的视觉质量，输入图像数量可达数千张。
cs.CV / 18 / 2607.14497
Reinforcing Egocentric Spatial Perception in Multimodal Large Language Models via Ego Scene Augmentation
通过自我场景增强强化多模态大型语言模型中的自我中心空间感知
Abstract
Egocentric Visual Question Answering (VQA) has attracted widespread attention as an important task for enabling Multimodal Large Language Models (MLLMs) to interact with the real world. However, existing MLLMs struggle to perform effective spatial reasoning in complex egocentric scenes due to their limited spatial perception capabilities. To this end, we introduce Ego Scene Augmentation (ESA), an egocentric spatial perception framework, which actively enhances the spatial perception capabilities from the egocentric perspective, powered by the proposed Ego-element Graph. Our core insight is leveraging the Ego-element Graph as an intermediary representation to augment the egocentric spatial perception of MLLMs via visual foundational models. Specifically, we 1) construct the Ego-element Graph, which encapsulates and integrates egocentric spatial features enabled by visual foundational models; 2) enhance the spatial perception capabilities of MLLMs via the Ego-element Graph for ego-perspective scenes. Our proposed ESA framework presents significant performance improvement on the EgoTextVQA benchmark. We achieve an 8.14% gain on the indoor setting and an 8.72% gain on the outdoor setting. Furthermore, our ESA shows the most impressive performance improvement in the shopping subset of the indoor setting. The project code is publicly available.
Chinese Translation
自我中心视觉问答（VQA）作为一种重要任务，吸引了广泛关注，以使多模态大型语言模型（MLLMs）能够与现实世界进行交互。然而，现有的MLLMs在复杂的自我中心场景中进行有效空间推理时面临困难，因为它们的空间感知能力有限。为此，我们提出了自我场景增强（Ego Scene Augmentation，ESA），这是一个自我中心空间感知框架，旨在通过所提出的自我元素图（Ego-element Graph）主动增强自我中心视角的空间感知能力。我们的核心见解是利用自我元素图作为中介表示，通过视觉基础模型增强MLLMs的自我中心空间感知。具体而言，我们 1) 构建自我元素图，该图封装并整合了由视觉基础模型支持的自我中心空间特征；2) 通过自我元素图增强MLLMs在自我视角场景中的空间感知能力。我们提出的ESA框架在EgoTextVQA基准测试中表现出显著的性能提升。在室内设置中，我们获得了8.14%的提升，在室外设置中获得了8.72%的提升。此外，我们的ESA在室内设置的购物子集上表现出最显著的性能提升。项目代码已公开。
cs.CV / 19 / 2607.14509
Multi-Scale ViT Inference with Habitat-Fit Priors and kNN Retrieval for Multi-Species Plant Identification
基于栖息地适应先验和kNN检索的多尺度ViT推理用于多物种植物识别
Abstract
This paper describes DS@GT ARC's third-place solution to the PlantCLEF 2026 challenge on multi-species plant identification in vegetation quadrat images, where systems must predict every species present in high-resolution (~3000 x 3000 pixel) plot photographs while training only on single-label images of individual plants. The pipeline is built around a fine-tuned DINOv2 ViT-L/14 classifier applied over a multi-scale tile decomposition of each quadrat, with per-tile predictions blended with a FAISS kNN retriever and post-processed by source-aware temporal fusion across repeated plot visits, a habitat-fit demotion that injects geographic and altitude priors from the training data, and a South-Western Europe geographic mask. Habitat-fit demotion and multi-scale aggregation are the largest individual contributors in the ablations. Two complementary training-centric directions, a cross-region transformer with noisy-student distillation on the LUCAS dataset and a label-as-query transformer decoder over synthetic CLS-domain pseudo-quadrats, yielded null results. An inference-time augmentation with instance-aware segmentation crops also did not improve performance. The selected submission reaches a private-leaderboard macro-F1 of 0.43902 (third place; public 0.51096); an unselected configuration of the same pipeline scored above 0.45 on the private set. Code: https://github.com/dsgt-arc/plantclef-2026.
Chinese Translation
本文描述了DS@GT ARC在2026年PlantCLEF挑战赛中获得第三名的方案，该挑战赛聚焦于在植被样方图像中进行多物种植物识别。系统必须在高分辨率（约3000 x 3000像素）的样方照片中预测每种存在的物种，同时仅在单标签的个体植物图像上进行训练。该流程围绕一个经过微调的DINOv2 ViT-L/14分类器构建，该分类器应用于每个样方的多尺度瓦片分解，逐瓦片的预测结果与FAISS kNN检索器相结合，并通过源感知的时间融合进行后处理，此外还引入了来自训练数据的地理和海拔先验的栖息地适应降权，以及一个西南欧洲地理掩模。在消融实验中，栖息地适应降权和多尺度聚合是最大的单一贡献者。两个互补的以训练为中心的方向，一个是基于LUCAS数据集的跨区域变换器与噪声学生蒸馏，另一个是针对合成CLS领域伪样方的标签作为查询的变换器解码器，均未取得有效结果。推理时的增强与实例感知分割裁剪也未能提高性能。所选提交在私有排行榜上达到了宏F1值0.43902（第三名；公共0.51096）；同一流程的未选配置在私有集上得分超过0.45。代码链接：https://github.com/dsgt-arc/plantclef-2026。
cs.CV / 20 / 2607.14513
Compression of 3D Gaussian Splatting Data Using GPU-friendly Graphics Texture Coding
使用适合GPU的图形纹理编码压缩3D高斯点云数据
Abstract
Techniques for modeling 3D scenes from image collections, such as 3D Gaussian Splatting (3DGS), are capable of generating high-quality novel views by leveraging graphics primitives with view-dependent appearance. In 3DGS, spherical harmonic (SH) are employed to model view-dependent color, resulting in a large number of SH coefficients per primitive and large memory requirements. While compression approaches have been proposed to mitigate this problem, they do not exploit the capabilities of modern Graphics Processing Units (GPUs) for parallel decoding and rendering. In this paper, we propose a method for compressing SH color coefficients using texture compression schemes specifically designed for efficient parallel GPU decoding and supported by dedicated hardware acceleration. It is shown that those methods can compress color coefficients more effectively than 2D textures by exploiting the fact that primitives can be locally grouped and reordered according to color. Furthermore, we introduce a bit-rate control strategy that preserves random access, enabling large-scale parallelization without compromising rendering performance. Experimental results using BC1 and BC7 texture compression formats show that GPU-based decompression can be achieved with negligible or imperceptible degradation in the visual quality of rendered 3DGS scenes.
Chinese Translation
从图像集合建模3D场景的技术，如3D高斯点云（3D Gaussian Splatting, 3DGS），能够通过利用具有视角依赖外观的图形原语生成高质量的新视图。在3DGS中，采用球面谐波（Spherical Harmonic, SH）来建模视角依赖的颜色，导致每个原语需要大量的SH系数，从而产生较大的内存需求。虽然已经提出了压缩方法来缓解这个问题，但这些方法并未利用现代图形处理单元（Graphics Processing Units, GPUs）在并行解码和渲染方面的能力。本文提出了一种使用专门为高效并行GPU解码设计的纹理压缩方案来压缩SH颜色系数的方法，并得到了专用硬件加速的支持。结果表明，这些方法通过利用原语可以根据颜色进行局部分组和重排序的事实，比2D纹理更有效地压缩颜色系数。此外，我们引入了一种比特率控制策略，保持随机访问能力，使得大规模并行化得以实现而不影响渲染性能。使用BC1和BC7纹理压缩格式的实验结果表明，基于GPU的解压缩可以在可忽略或不可察觉的视觉质量下降的情况下实现对渲染的3DGS场景的解压。
cs.CV / 21 / 2607.14514
VTM-Nav: Hierarchical Visual-Topological Memory for Cross-Episode Object-Goal Navigation
VTM-Nav：用于跨情节目标物体导航的分层视觉-拓扑记忆
Abstract
Object-goal navigation requires an embodied agent to locate and reach an instance of a specified object category in an indoor environment. Recent training-free approaches leverage vision-language models (VLMs) for open-vocabulary semantic reasoning, but are typically evaluated under an episodic protocol that resets all scene-specific state after each episode. We introduce Cross-Episode Object-Goal Navigation, in which an agent repeatedly operates in the same scene, retains only self-acquired experience, and keeps its model parameters fixed. To support experience reuse, we present \method, a training-free VLM navigation framework with a persistent hierarchical Visual-Topological Memory (VTM). The VTM organizes scene knowledge at room and object levels and retrieves relevant experience through coarse-to-fine matching, providing memory as soft guidance only when it agrees with current observations. A conservative execution guard further mitigates oscillations, blocked motions, and premature stopping. Under a controlled same-scene protocol, we evaluate \method{} on three benchmarks, HM3D v0.1, HM3D v0.2, and MP3D, and compare it with a strengthened WMNav baseline augmented with cross-episode textual memory, while keeping the VLM backbone and action pipeline identical. \method{} achieves the best performance across all three benchmarks, demonstrating the effectiveness and robustness of structured visual-topological experience reuse across datasets.
Chinese Translation
目标物体导航要求一个具身代理在室内环境中定位并到达指定物体类别的实例。近期的无训练方法利用视觉-语言模型（VLMs）进行开放词汇的语义推理，但通常在情节协议下进行评估，该协议在每个情节后重置所有场景特定状态。我们引入了跨情节目标物体导航，其中代理在同一场景中反复操作，仅保留自我获得的经验，并保持其模型参数不变。为了支持经验重用，我们提出了 extit{VTM-Nav}，一个无训练的VLM导航框架，具有持久的分层视觉-拓扑记忆（VTM）。VTM在房间和物体层面组织场景知识，并通过粗到细的匹配检索相关经验，仅在与当前观察一致时提供作为软引导的记忆。保守的执行保护进一步减轻了震荡、阻塞运动和过早停止。在一个受控的同场景协议下，我们在三个基准测试HM3D v0.1、HM3D v0.2和MP3D上评估 extit{VTM-Nav}，并将其与增强了跨情节文本记忆的强化WMNav基线进行比较，同时保持VLM主干和动作管道相同。 extit{VTM-Nav}在所有三个基准测试中实现了最佳性能，展示了跨数据集结构化视觉-拓扑经验重用的有效性和鲁棒性。
cs.CV / 22 / 2607.14521
Uni-AdaVD: Universal Concept Erasure for Visual Generation via Orthogonal Value Decomposition
Uni-AdaVD：通过正交值分解实现视觉生成的通用概念消除
Abstract
Visual generative models inevitably absorb undesirable concepts from uncurated pretraining data, making concept erasure essential for safe deployment. Existing erasure methods, however, are often architecture-specific and struggle to remove target concepts while preserving non-target content and generative priors. We present Uni-AdaVD, a universal inference-time concept erasure framework for visual generation. Uni-AdaVD treats the value space of multimodal attention as a unified intervention space and introduces encoder-aware target representation construction to localize target semantics across heterogeneous text encoders. It further combines orthogonal value decomposition with an adaptive erasing shift to suppress target semantic directions without updating the original model weights. Extensive experiments on U-Net-, DiT-, and autoregressive image generators, as well as text-to-video models, demonstrate strong performance on single- and multi-concept erasure while preserving non-target priors. These results suggest that Uni-AdaVD provides an efficient and adaptable safety mechanism for modern visual generative models. Our code is available at https://github.com/QifanZhou/Uni-AdaVD.
Chinese Translation
视觉生成模型不可避免地从未经筛选的预训练数据中吸收不良概念，因此概念消除对于安全部署至关重要。然而，现有的消除方法往往是特定于架构的，难以在保留非目标内容和生成先验的同时去除目标概念。我们提出了Uni-AdaVD，这是一种用于视觉生成的通用推理时概念消除框架。Uni-AdaVD将多模态注意力的值空间视为统一的干预空间，并引入编码器感知的目标表示构建，以在异构文本编码器中定位目标语义。它进一步结合正交值分解与自适应消除偏移，以在不更新原始模型权重的情况下抑制目标语义方向。在U-Net、DiT和自回归图像生成器以及文本到视频模型上的大量实验表明，Uni-AdaVD在单一和多重概念消除方面表现出色，同时保留非目标先验。这些结果表明，Uni-AdaVD为现代视觉生成模型提供了一种高效且可适应的安全机制。我们的代码可在 https://github.com/QifanZhou/Uni-AdaVD 获取。
cs.CV / 23 / 2607.14534
SwinAD: Multi-stage feature reconstruction for unsupervised industrial anomaly detection
SwinAD：用于无监督工业异常检测的多阶段特征重构
Abstract
Industrial anomaly detection aims to identify and localize defective regions without relying on exhaustive annotations of all possible defect types. Although recent unsupervised methods have achieved strong performance, most are primarily designed for single-class settings and often struggle in multi-class scenarios, where diverse normal patterns may lead to over-generalization and reduce the discriminative capability between normal and anomalous regions. In this paper, we propose SwinAD, a reconstruction-based framework for multi-class unsupervised anomaly detection that leverages a frozen pretrained Swin Transformer V2 encoder and a feature diversity-preserving reconstruction decoder. The hierarchical encoder provides semantically rich multi-scale features, while stage-wise bottleneck modules with dropout prevent trivial identity mapping and encourage robust reconstruction of normal patterns. To further improve localization, we introduce a feature diversity-preserving reconstruction framework that maintains complementary reconstruction hypotheses instead of relying on a single decoding branch. The discrepancies between encoder features and the two reconstructed features are then aggregated across multiple scales to produce the final anomaly map. Experiments conducted on three industrial anomaly detection benchmarks, including MVTec AD, VisA, and Real-IAD, demonstrate that SwinAD achieves competitive image-level performance and strong pixel-level localization accuracy, with particularly notable improvements in pixel-level AP and 1 on MVTec AD. These results indicate that combining hierarchical Swin features with diverse multi-scale reconstruction substantially improve pixel-level localization in multi-class unsupervised anomaly setting.
Chinese Translation
工业异常检测旨在识别和定位缺陷区域，而无需依赖对所有可能缺陷类型的详尽注释。尽管近期的无监督方法已取得了良好的性能，但大多数方法主要设计用于单类场景，往往在多类场景中表现不佳，其中多样的正常模式可能导致过度泛化，并降低正常区域与异常区域之间的区分能力。本文提出了SwinAD，一种基于重构的多类无监督异常检测框架，利用冻结的预训练Swin Transformer V2编码器和保持特征多样性的重构解码器。分层编码器提供语义丰富的多尺度特征，而逐阶段的瓶颈模块结合dropout防止平凡的身份映射，并鼓励正常模式的稳健重构。为了进一步改善定位，我们引入了一种保持特征多样性的重构框架，该框架维护互补的重构假设，而不是依赖单一的解码分支。然后，编码器特征与两个重构特征之间的差异在多个尺度上进行聚合，以生成最终的异常图。针对三个工业异常检测基准（包括MVTec AD、VisA和Real-IAD）进行的实验表明，SwinAD在图像级别上实现了竞争力的性能，并在像素级定位准确性上表现出色，尤其是在MVTec AD上在像素级AP和1方面有显著提升。这些结果表明，将分层Swin特征与多样的多尺度重构相结合，显著改善了多类无监督异常设置中的像素级定位。
cs.CV / 24 / 2607.14544
3D Geometric Tooth Alignment Planning via Deep Reinforcement Learning
基于深度强化学习的3D几何牙齿对齐规划
Abstract
3D geometric tooth alignment planning, which determines sequential trajectories from initial malocclusion to the final target alignment, is a cornerstone of modern digital orthodontics. This paper presents a novel deep reinforcement learning (DRL) framework to automate the generation of these alignment paths. We formulate the planning process as a Markov Decision Process (MDP) to capture its sequential decision-making nature, focusing on optimizing geometric trajectories while integrating essential spatial constraints, such as inter-dental collision avoidance and path efficiency. The proposed method leverages the Deep Deterministic Policy Gradient (DDPG) algorithm, enhanced by three key innovations: (1) a Transformer-based agent to model complex spatial interactions between teeth and manage high-dimensional state-action spaces; (2) a dynamic masking scheme that restricts movement to a sparse subset of teeth per step, better reflecting the clinical logic of sequential alignment; and (3) a two-stage curriculum learning strategy that gradually increases task difficulty to ensure training stability and efficient path discovery. We evaluate our approach on a dataset of 10K expert-designed treatment plans based on clinical data. Experimental results demonstrate that our method outperforms existing baselines in terms of path safety and geometric efficiency, providing a robust and automated solution for 3D geometric orthodontic alignment planning.
Chinese Translation
3D几何牙齿对齐规划是现代数字正畸学的基石，它确定了从初始错位到最终目标对齐的顺序轨迹。本文提出了一种新颖的深度强化学习（Deep Reinforcement Learning, DRL）框架，以自动生成这些对齐路径。我们将规划过程形式化为马尔可夫决策过程（Markov Decision Process, MDP），以捕捉其顺序决策的特性，重点优化几何轨迹，同时整合必要的空间约束，如牙齿间碰撞避免和路径效率。所提出的方法利用深度确定性策略梯度（Deep Deterministic Policy Gradient, DDPG）算法，并通过三项关键创新进行增强：(1) 基于Transformer的智能体，用于建模牙齿之间的复杂空间交互并管理高维状态-动作空间；(2) 动态掩蔽方案，限制每一步的移动仅在稀疏的牙齿子集内，更好地反映顺序对齐的临床逻辑；(3) 两阶段课程学习策略，逐步增加任务难度，以确保训练的稳定性和高效的路径发现。我们在一个基于临床数据的1万份专家设计治疗方案的数据集上评估了我们的方法。实验结果表明，我们的方法在路径安全性和几何效率方面优于现有基线，为3D几何正畸对齐规划提供了一种稳健且自动化的解决方案。
cs.CV / 25 / 2607.14547
AdaTurn: Budget-Aware Test-Time Scaling for Active Visual Perception Agents
AdaTurn：面向预算的测试时间缩放方法用于主动视觉感知代理
Abstract
Active visual agents solve fine-grained image tasks by interleaving reasoning with image-grounding actions across multiple turns. However, deployment-time rollout budgets are rarely fixed: some requests permit long rollouts, while others require the agent to act under a tight turn limit. Existing methods train the policy as if the rollout budget were hidden, so when the available budget is smaller than the trajectory the agent prefers, the interaction is often truncated before any valid answer is produced; we term this failure \emph{catastrophic truncation}. To overcome this challenge, we present AdaTurn, a budget-aware framework that conditions the agent on the allowed number of turns and explicitly trains the boundary behavior induced by the budget. Our key component, Forced-Answer DAPO (FA-DAPO), converts the over-budget event from a masked or penalized failure into a trainable final-decision step, teaching the model to synthesize partial evidence when further tool use is no longer possible. We further randomize rollout budgets during both training and inference and introduce a load-balanced scheduler that makes such operations practical. AdaTurn substantially improves low-budget accuracy, for example raising VisualProbe-Medium from 36.7% to 47.6% at four turns, while preserving strong scaling at larger budgets and transferring effectively to multiple backbones and general multimodal benchmarks.
Chinese Translation
主动视觉代理通过在多个回合中交替推理与图像基础行动来解决细粒度图像任务。然而，部署时的回合预算通常并非固定：某些请求允许较长的回合，而其他请求则要求代理在严格的回合限制下行动。现有方法在训练策略时假设回合预算是隐藏的，因此当可用预算小于代理所偏好的轨迹时，交互往往在未产生任何有效答案之前被截断；我们将这种失败称为 extit{灾难性截断}。为了克服这一挑战，我们提出了AdaTurn，一个预算感知框架，该框架根据允许的回合数对代理进行条件设置，并明确训练由预算引发的边界行为。我们的关键组件，强制回答 DAPO（FA-DAPO），将超预算事件从掩蔽或惩罚的失败转化为可训练的最终决策步骤，教会模型在无法进一步使用工具时合成部分证据。我们还在训练和推理过程中随机化回合预算，并引入了一个负载平衡调度器，使这些操作变得可行。AdaTurn显著提高了低预算的准确性，例如在四个回合中将VisualProbe-Medium的准确率从36.7%提高到47.6%，同时在更大预算下保持强大的扩展性，并有效地迁移到多个骨干网络和通用多模态基准上。
cs.CV / 26 / 2607.14548
HyMobileAgent: Data-Environment Co-Scaling for Efficient GUI Agents
HyMobileAgent：高效图形用户界面代理的数据-环境协同扩展
Hy Vision Team, Shen, Huawen, Tang, Zhengyang, Peng, Shangpin, Wu, Liang, Zhang, Anran, Wang, Weinong, Guo, Yiduo, Li, Chenxin, Fang, Zhengyao, Ding, Yang, Li, Junyi, Tang, Fei, Ruan, Zheng, Zhang, Yi, Zhou, Xingran, Yang, Dingchen, Fan, Sunqi, Wan, Zhiyi, Hu, Han, Lai, Xin, Lyu, Pengyuan, Zhang, Chengquan
Abstract
As large multimodal models move from understanding content to operating on digital environments, mobile GUI has emerged as a challenging and consequential testbed for digital embodied intelligence. Mobile agents operate under three coupled constraints: precise perception of complex interfaces, scalable acquisition of high-quality interaction data, and robust long-horizon decision making under compounding execution errors. This report presents HyMobileAgent, a mobile GUI agent built on Hy3.0-VL-A3B, a vision-native foundation model featuring native any-resolution input, an A3B-scale deployment budget, and a 32K context window to model extended interaction histories. Rather than relying solely on model scaling, we develop a joint data and environment centric scaling framework to address the key bottlenecks of mobile interaction. Our framework integrates a GUI perception flywheel combining mock-interface synthesis, rejection sampling, and icon-specific augmentation; a knowledge pipeline that transforms tutorial videos into structured interaction data; a million-scale action data pipeline deployed across more than 2000 sandbox and real-device instances with automated failure attribution; the PhoneWorld Mock App Factory, providing a resettable training environment with 34 mock applications and over 34000 tasks; and a structured Planning-and-Reflection mechanism with explicit dead-loop detection for reliable long-horizon execution. We also introduce a progressive training recipe consisting of mid-training, supervised fine-tuning, and reinforcement learning with task-specific reward designs.
Chinese Translation
随着大型多模态模型从理解内容转向在数字环境中操作，移动图形用户界面（GUI）已成为数字具身智能的一个具有挑战性和重要性的测试平台。移动代理在三个相互关联的约束下运行：对复杂界面的精确感知、高质量交互数据的可扩展获取，以及在累积执行错误下的稳健长远决策。本文介绍了HyMobileAgent，一个基于Hy3.0-VL-A3B构建的移动GUI代理，该基础模型具有原生任意分辨率输入、A3B规模的部署预算，以及32K的上下文窗口以建模扩展的交互历史。我们并不单纯依赖模型扩展，而是开发了一个以数据和环境为中心的联合扩展框架，以解决移动交互的关键瓶颈。我们的框架集成了一个GUI感知飞轮，结合了模拟界面合成、拒绝采样和图标特定增强；一个知识管道，将教程视频转化为结构化交互数据；一个在超过2000个沙盒和真实设备实例中部署的百万规模动作数据管道，具有自动故障归因功能；PhoneWorld模拟应用工厂，提供一个可重置的训练环境，包含34个模拟应用和超过34000个任务；以及一个具有显式死循环检测的结构化规划与反思机制，以实现可靠的长远执行。我们还引入了一种渐进式训练方案，包括中期训练、监督微调和具有任务特定奖励设计的强化学习。
cs.CV / 27 / 2607.14560
Breaking the Model Forgetting Cycle in Long-Incremental 3D Object Detection
打破长增量3D目标检测中的模型遗忘循环
Abstract
Incremental 3D object detection requires a detector to learn novel object classes while remembering previously learned ones over sequentially arriving data. Previous methods, primarily based on pseudo-labeling, perform reasonably in short-incremental stages but still suffer from severe model forgetting when dealing with long-incremental sequences. We investigate this failure and reveal a detrimental self-reinforcing cycle: data distribution shift of novel classes causes model forgetting on old classes, which further produces accumulated error in pseudo-labeling that exacerbates model degradation. To address this issue, we draw inspiration from the human learning process and propose the \emph{Learning-Dynamics-driven Memory and Review} (LDMR) framework. LDMR monitors per-class detection quality at periodic training checkpoints and uses these learning-dynamics signals to drive two innovative mechanisms, namely (i) human-like intra-stage review that divides each incremental stage into multiple sub-stages' training and concentrates on remembering the most-forgotten objects, and (ii) scene-aware cross-stage memory evolution that evolves a memory bank to transfer knowledge between two consecutive stages by jointly considering scene learnability and diversity. Extensive experiments across multiple long-incremental protocols on indoor benchmarks SUN RGB-D and ScanNetV2 show that LDMR substantially mitigates the model forgetting and outperforms all baselines by a clear margin. Code is available at https://github.com/qianpeisheng/LDMR.
Chinese Translation
增量3D目标检测要求检测器在处理连续到达的数据时，学习新对象类别的同时记住之前学习的类别。以往的方法主要基于伪标签，在短增量阶段表现合理，但在处理长增量序列时仍然遭遇严重的模型遗忘。我们调查了这一失败现象，并揭示了一个有害的自我强化循环：新类别的数据分布变化导致旧类别的模型遗忘，进而在伪标签中产生累积错误，加剧模型退化。为了解决这个问题，我们从人类学习过程中获得灵感，提出了 extit{学习动态驱动的记忆与复习}（Learning-Dynamics-driven Memory and Review, LDMR）框架。LDMR在定期训练检查点监控每个类别的检测质量，并利用这些学习动态信号驱动两个创新机制，即（i）类人化的阶段内复习，将每个增量阶段划分为多个子阶段进行训练，并集中记住最易遗忘的对象，以及（ii）场景感知的跨阶段记忆演化，通过共同考虑场景的可学习性和多样性，演化一个记忆库以在两个连续阶段之间转移知识。在室内基准数据集SUN RGB-D和ScanNetV2上进行的多项长增量协议的广泛实验表明，LDMR显著减轻了模型遗忘，并明显优于所有基线方法。代码可在 https://github.com/qianpeisheng/LDMR 获取。
cs.CV / 28 / 2607.14580
Advanced Image Generation: Negative Prompt Optimization and Latent Classifier Guidance
高级图像生成：负提示优化与潜在分类器引导
Abstract
We present a novel system that integrates negative prompt optimization via a fine-tuned sequence-to-sequence LLM and latent-space classifier guidance to improve the quality of images generated by Stable Diffusion. Our approach automatically generates optimized negative prompts, and employs a CNN-RNN hybrid classifier to evaluate and guide diffusion steps, rolling back low-quality latent updates. Experimental results demonstrate that our dual-guidance framework reduces artifacts and improves semantic fidelity compared to baseline diffusion.
Chinese Translation
我们提出了一种新颖的系统，该系统通过微调的序列到序列语言模型（LLM）实现负提示优化，并结合潜在空间分类器引导，以提高Stable Diffusion生成图像的质量。我们的方法自动生成优化的负提示，并采用CNN-RNN混合分类器来评估和引导扩散步骤，回滚低质量的潜在更新。实验结果表明，与基线扩散相比，我们的双重引导框架减少了伪影，并提高了语义保真度。
cs.CV / 29 / 2607.14595
MagicPrompt: Ultra-Lightweight Prompt Tuning for Video Generation
MagicPrompt：超轻量级提示调优用于视频生成
Abstract
Large-scale video diffusion models (VDMs) deliver strong generation performance, but full fine-tuning for downstream tasks incurs prohibitive computational costs. Existing parameter-efficient fine-tuning (PEFT) methods have two critical flaws on billion-scale models: they still require substantial trainable parameters, and reward-based training suffers from noise-induced optimization instability in condition-guided tasks. We propose MagicPrompt, a lightweight framework that achieves extreme parameter efficiency and stable reward optimization. It first adopts Attention-Embedded Prompt Tuning, which steers generation via lightweight soft prompts with orders of magnitude fewer parameters while preserving pre-trained knowledge. It further introduces Dual-Space Reward Feedback Optimization, which uses self-supervised latent objectives to improve condition-guided reward training. Experiments show MagicPrompt reaches competitive performance with less than 1\% trainable parameters and notably reduces training costs.
Chinese Translation
大规模视频扩散模型（VDMs）在生成性能上表现出色，但对下游任务进行全面微调会产生高昂的计算成本。现有的参数高效微调（PEFT）方法在十亿规模模型上存在两个关键缺陷：它们仍然需要大量可训练参数，并且基于奖励的训练在条件引导任务中受到噪声引起的优化不稳定性的影响。我们提出了MagicPrompt，一个实现极高参数效率和稳定奖励优化的轻量级框架。它首先采用了注意力嵌入提示调优（Attention-Embedded Prompt Tuning），通过轻量级软提示引导生成，所需参数数量减少了几个数量级，同时保留了预训练知识。它进一步引入了双空间奖励反馈优化（Dual-Space Reward Feedback Optimization），利用自监督潜在目标来改善条件引导的奖励训练。实验表明，MagicPrompt在可训练参数少于1%的情况下达到了竞争性能，并显著降低了训练成本。
cs.CV / 30 / 2607.14598
Hough-SIFT: Robust Image Registration for Linear Structures via Hough Space
Hough-SIFT：通过霍夫空间实现线性结构的鲁棒图像配准
Abstract
Image registration is essential in applications such as electronic image stabilization. Scale-Invariant Feature Transform (SIFT), a widely used local keypoint detector and descriptor, typically provides accurate registration; however, it often fails in scenes with strong linear structures (e.g., shutters), where local features become ambiguous. We propose Hough-SIFT, a robust registration method that performs SIFT descriptor matching in Hough space. In this domain, linear structures form distinctive peaks that restore descriptor discriminability. Experiments demonstrate that Hough-SIFT is robust in linear scenes where SIFT frequently fails, while maintaining accuracy comparable to SIFT in normal scenes.
Chinese Translation
图像配准在电子图像稳定等应用中至关重要。尺度不变特征变换（SIFT）是一种广泛使用的局部关键点检测器和描述符，通常能够提供准确的配准；然而，在具有强线性结构（例如百叶窗）的场景中，它常常失效，因为局部特征变得模糊。我们提出了Hough-SIFT，这是一种鲁棒的配准方法，在霍夫空间中执行SIFT描述符匹配。在该领域，线性结构形成独特的峰值，从而恢复描述符的可区分性。实验表明，Hough-SIFT在SIFT经常失效的线性场景中表现出鲁棒性，同时在正常场景中保持与SIFT相当的准确性。
cs.CV / 31 / 2607.14631
Knowing You at First Glance: Inferring Apparent Personality from Faces
一见如故：从面孔推断表面个性
Abstract
Inferring apparent personality from facial images is important in social scenarios for embodied agents in human-robot interaction. Unlike inferring intrinsic personality traits via conversation, this task models first-impression personality perception based solely on facial appearance before interaction begins. Existing studies mainly focus on the Big Five personality model and often rely on language or multimodal inputs. As a result, it remains unclear whether facial cues alone can support meaningful associations with perceived personality traits. This question is particularly relevant for MBTI types, which are widely used in practice and more readily interpretable by large language models. To this end, we propose \textbf{GlanceFace}, an end-to-end framework for apparent personality inference leveraging vision-language models to introduce semantic priors and a semantic-enhanced facial representation module to capture subtle personality-related cues, together with an uncertainty-aware learning strategy to handle noisy and subjective annotations. Extensive experiments demonstrate strong performance on MBTI-based apparent personality benchmarks and reveal relationships between facial characteristics and perceived personality traits, highlighting its potential to support adaptive initial interaction strategies for embodied agents. The code and dataset are available at https://github.com/MrHuan3/GlanceFace.
Chinese Translation
从面部图像推断表面个性在机器人与人类互动的社交场景中至关重要。与通过对话推断内在个性特征不同，该任务仅基于面部外观建模初次印象的个性感知，且在互动开始之前进行。现有研究主要集中于五大人格模型（Big Five personality model），并且通常依赖于语言或多模态输入。因此，单靠面部线索是否能够支持与感知个性特征之间的有意义关联仍然不清楚。这个问题对于广泛应用且更易被大型语言模型（large language models）解读的MBTI类型尤为相关。为此，我们提出了 extbf{GlanceFace}，一个端到端的表面个性推断框架，利用视觉-语言模型引入语义先验，并结合一个语义增强的面部表征模块以捕捉细微的个性相关线索，同时采用一种不确定性感知学习策略来处理噪声和主观注释。大量实验表明，该方法在基于MBTI的表面个性基准测试中表现出色，并揭示了面部特征与感知个性特征之间的关系，突显了其支持机器人适应性初始互动策略的潜力。代码和数据集可在 https://github.com/MrHuan3/GlanceFace 获取。
cs.CV / 32 / 2607.14645
Autoregressive Modeling of Film with Applications in Video Montage
电影的自回归建模及其在视频蒙太奇中的应用
Abstract
This work introduces FilmGPT, an autoregressive transformer designed to address the challenge of video montage--turning a collection of raw, "unwatchable" footage into coherent cinematic sequences. Inspired by language learning in modern LLMs, we train a long-context autoregressive transformer on a large corpus of movies. The aim is to implicitly capture the "grammar" of film directly from data rather than from hand-coded rules. Unlike other generative models, FilmGPT does not generate any new video frames. Instead, at inference time, we introduce a footage-constrained decoding algorithm to select the best next shot from the input raw footage according to the statistical patterns learned from films. We first evaluate these learned statistics directly by using the FilmGPT autoregressive model for next shot prediction on a standard benchmark of shot sequence ordering, outperforming the previous state of the art. We then evaluate our footage-constrained decoding algorithm on the full film editing task via a user study, and find that our FilmGPT-based editing significantly outperforms previous approaches. Finally, we demonstrate the applicability of FilmGPT to a wide range of applications in video montage, from automatic video segment trimming to human-in-the-loop film editing.
Chinese Translation
本文介绍了FilmGPT，一种自回归变换器，旨在解决视频蒙太奇的挑战——将一组原始的、"不可观看"的镜头转化为连贯的电影序列。受现代大规模语言模型（LLMs）中语言学习的启发，我们在大量电影语料库上训练了一个长上下文自回归变换器。其目的是直接从数据中隐式捕捉电影的"语法"，而不是依赖手工编码的规则。与其他生成模型不同，FilmGPT并不生成任何新的视频帧。相反，在推理时，我们引入了一种受镜头限制的解码算法，根据从电影中学习到的统计模式，从输入的原始镜头中选择最佳的下一个镜头。我们首先通过在标准镜头序列排序基准上使用FilmGPT自回归模型进行下一个镜头预测，直接评估这些学习到的统计数据，结果超越了之前的最先进水平。然后，我们通过用户研究评估了我们的受镜头限制的解码算法在完整电影编辑任务中的表现，发现基于FilmGPT的编辑显著优于之前的方法。最后，我们展示了FilmGPT在视频蒙太奇中的广泛应用，包括自动视频片段修剪和人机协作的电影编辑。
cs.CV / 33 / 2607.14660
VIABench: A Comprehensive Video Benchmark Collected from Blind Individuals for Visual Impairment Assistance
VIABench：为视觉障碍辅助收集的全面视频基准
Abstract
Visually impaired individuals (VIIs) encounter significant daily challenges due to limited access to visual information. Although Multimodal Large Language Models (MLLMs) have achieved impressive results on general vision and language tasks, their practical utility in real-world blind assistance still remains largely underexplored. To fill this gap, we introduce VIABench, a comprehensive video benchmark specifically designed to evaluate MLLMs in Visually Impaired Assistance scenarios using first-person videos recorded or shared by VIIs themselves. VIABench defines three core tasks, each targeting a distinct requirement in visual assistance. Proactive Reminder: Assesses the model's ability to interpret ongoing video content while proactively anticipating and verbally describing upcoming navigation-critical events; Visual Question Answering (VQA): Evaluates the model's capacity to answer user-posed questions about the environment or objects within the video; Vision-Guided Interaction: Tests context-aware reasoning to accomplish intentional interactions between user and environment. To ensure a robust and fair evaluation, we propose a rigorous benchmarking pipeline that supports both online (real-time) and offline settings. Our experiments demonstrate that current MLLMs still struggle to deliver comprehensive support for VIIs, especially in the Proactive Reminder task, which demands accurate anticipation and real-time responsiveness. We hope VIABench will drive future research toward developing customized MLLMs for real-world assistance, ultimately improving navigation and interaction experiences for visually impaired individuals. Code and data will be released at https://github.com/MCG-NJU/VIABench.
Chinese Translation
视觉障碍人士（VIIs）由于对视觉信息的有限获取，每天面临重大挑战。尽管多模态大语言模型（MLLMs）在一般视觉和语言任务上取得了令人瞩目的成果，但它们在现实世界中的盲人辅助应用仍然未得到充分探索。为填补这一空白，我们推出了VIABench，这是一个专门设计的全面视频基准，旨在使用由视觉障碍人士自己录制或分享的第一人称视频来评估MLLMs在视觉障碍辅助场景中的表现。VIABench定义了三个核心任务，每个任务针对视觉辅助中的不同需求。主动提醒：评估模型在解释正在进行的视频内容时，主动预测并口头描述即将发生的导航关键事件的能力；视觉问答（VQA）：评估模型回答用户提出的关于环境或视频中物体的问题的能力；视觉引导交互：测试上下文感知推理以实现用户与环境之间的有意交互。为了确保评估的稳健性和公平性，我们提出了一种严格的基准测试流程，支持在线（实时）和离线设置。我们的实验表明，当前的MLLMs在为视觉障碍人士提供全面支持方面仍然存在困难，尤其是在主动提醒任务中，这要求准确的预测和实时响应。我们希望VIABench能够推动未来研究，开发定制化的MLLMs以用于现实世界的辅助，最终改善视觉障碍人士的导航和交互体验。代码和数据将发布在 https://github.com/MCG-NJU/VIABench。
cs.CV / 34 / 2607.14681
ReBind: Multi-Reference Video Editing via Structured Instructions with Explicit Reference Relationships
ReBind：通过具有明确参考关系的结构化指令进行多参考视频编辑
Abstract
Recent diffusion-based video generation models have made significant progress in multi-reference image-conditioned video editing. However, existing methods still struggle to coordinate information from multiple visual sources accurately. We identify a critical deficiency in existing approaches. Existing editing instructions lack explicit reference relationships, and most multimodal large language models (MLLMs) cannot generate them reliably. To address this problem, we propose ReBind, a systematic framework that introduces semantic instructions with embedded reference tokens as the intermediate representation for multi-reference image-conditioned video editing. Our key insight is embedding reference tokens at semantic positions to eliminate ambiguity and establish precise bindings between visual attributes and their sources. We develop ReBind-Instruct, a specialized MLLM that learns to establish explicit bindings between visual attributes and their reference sources through a two-stage progressive scheme for precise reference relationships. We further develop ReBind-Edit, which enables lightweight adaptation of text-to-video models to coordinate multiple references by binding visual attributes to their designated sources. Extensive experiments demonstrate that ReBind substantially outperforms general-purpose MLLMs in instruction quality and achieves state-of-the-art performance among open-source methods on reference image conditioned video editing. Our project webpage: https://rebind-mrv2v.github.io/.
Chinese Translation
近期基于扩散的视频生成模型在多参考图像条件下的视频编辑方面取得了显著进展。然而，现有方法在准确协调来自多个视觉源的信息方面仍然面临挑战。我们识别出现有方法中的一个关键缺陷：现有的编辑指令缺乏明确的参考关系，大多数多模态大型语言模型（MLLMs）无法可靠地生成这些关系。为了解决这个问题，我们提出了ReBind，一个系统框架，通过嵌入参考标记的语义指令作为多参考图像条件视频编辑的中间表示。我们的关键见解是将参考标记嵌入到语义位置，以消除歧义并建立视觉属性与其来源之间的精确绑定。我们开发了ReBind-Instruct，一个专门的MLLM，通过两阶段渐进方案学习在视觉属性与其参考来源之间建立明确的绑定，以实现精确的参考关系。我们进一步开发了ReBind-Edit，使文本到视频模型能够轻量适应，通过将视觉属性绑定到其指定来源来协调多个参考。大量实验表明，ReBind在指令质量上显著优于通用MMLMs，并在参考图像条件视频编辑方面在开源方法中实现了最先进的性能。我们的项目网页：https://rebind-mrv2v.github.io/
cs.CV / 35 / 2607.14684
GlobalForge: Towards Robust AI-Generated Image Detection
GlobalForge：迈向稳健的人工智能生成图像检测
Abstract
AI-generated image (AIGI) detectors achieve strong accuracy on clean benchmarks, but their performance drops sharply after images are propagated through real-world channels. We trace this fragility to what these detectors actually learn: they overfit to local artifacts left by generators in small spatial neighborhoods, which are easily destroyed by common propagation degradations such as JPEG compression and blur. Instead, we shift the discriminative cue from fragile local artifacts to more robust global structure. Building on this, we propose GlobalForge, a framework with two complementary modules. The Local Information Bottleneck (LIB) suppresses local components to block shortcut learning, while the Global Structural Reasoning (GSR) module forces every token to gather evidence from distant regions. Both modules are trained jointly under a contrastive structural loss based on degradation that keeps the resulting features stable under degradation. To support fine-grained robustness evaluation, we further introduce RealDeg-Bench, covering 7 common degradation operators and multi-step compound chains. GlobalForge improves average BAcc on 8 in-the-wild benchmark groups by $\mathbf{5.89\%}$ over the previous state-of-the-art, and is clearly ahead of representative baselines on RealDeg-Bench under both single and compound degradations. Code is available at https://anonymous.4open.science/r/GlobalForge-BE0F/.
Chinese Translation
人工智能生成图像（AIGI）检测器在干净的基准测试中表现出强大的准确性，但在图像经过现实世界通道传播后，其性能急剧下降。我们将这种脆弱性追溯到这些检测器实际学习的内容：它们过度拟合生成器在小空间邻域中留下的局部伪影，这些伪影很容易被常见的传播降质（如JPEG压缩和模糊）所破坏。因此，我们将判别线索从脆弱的局部伪影转移到更稳健的全局结构。基于此，我们提出了GlobalForge，一个包含两个互补模块的框架。局部信息瓶颈（Local Information Bottleneck, LIB）抑制局部成分以阻止捷径学习，而全局结构推理（Global Structural Reasoning, GSR）模块则强制每个令牌从远处区域收集证据。两个模块在基于降质的对比结构损失下共同训练，以保持在降质下生成的特征稳定。为了支持细粒度的稳健性评估，我们进一步引入了RealDeg-Bench，涵盖7种常见降质算子和多步复合链。GlobalForge在8个真实环境基准组上将平均BAcc提高了$ extbf{5.89 ext{%}}$，超越了之前的最先进水平，并在RealDeg-Bench的单一和复合降质下明显领先于代表性基线。代码可在https://anonymous.4open.science/r/GlobalForge-BE0F/获取。
cs.CV / 36 / 2607.14702
Team RAS in 11th ABAW Competition: Multimodal Ambivalence Recognition Approach
第11届ABAW竞赛中的团队RAS：多模态矛盾识别方法
Abstract
Automatic recognition of ambivalence and hesitancy is challenging because these states may be expressed through inconsistent linguistic, acoustic, facial, and contextual patterns, while top-performing systems often rely on computationally expensive ensembles. We present a single text-centered multimodal approach for video-level ambivalence and hesitancy recognition for the 11th Affective & Behavior Analysis in-the-Wild (ABAW) Challenge. The proposed approach combines linguistic, acoustic, facial, and scene features using text-centered multimodal fusion model. Text Residual Fusion treats text as the anchor modality and applies gated residual adjustments based on the other modalities. Experiments on the Behavioural Ambivalence/Hesitancy (BAH) corpus confirm that text is the strongest unimodal modality. The Text Residual Fusion model achieves an average Macro F1-score (MF1) of 75.14% across the Development and Public Test subsets. On the Private Test subset, it reaches an MF1 of 78.24%, outperforming the text model by 4.03%. These results demonstrate that complementary multimodal information can improve recognition performance without requiring a large model ensemble.
Chinese Translation
自动识别矛盾和犹豫状态具有挑战性，因为这些状态可能通过不一致的语言、声学、面部和上下文模式来表达，而表现最佳的系统通常依赖于计算成本高昂的集成方法。我们提出了一种以文本为中心的单一多模态方法，用于第11届情感与行为分析挑战赛（ABAW）的视频级矛盾和犹豫识别。该方法结合了语言、声学、面部和场景特征，使用文本中心的多模态融合模型。文本残差融合（Text Residual Fusion）将文本视为锚定模态，并根据其他模态应用门控残差调整。在行为矛盾/犹豫（Behavioural Ambivalence/Hesitancy, BAH）语料库上的实验确认，文本是最强的单模态模态。文本残差融合模型在开发和公共测试子集上达到了75.14%的平均宏F1分数（MF1）。在私有测试子集上，该模型达到了78.24%的MF1，比文本模型提高了4.03%。这些结果表明，互补的多模态信息可以在不需要大型模型集成的情况下提高识别性能。
cs.CV / 37 / 2607.14703
Pretraining Multiple Instance Learning Networks with Multi-Teacher Distillation from Pathology Slide Foundation Models
通过病理切片基础模型的多教师蒸馏预训练多实例学习网络
Abstract
Multiple instance learning (MIL) has become the main paradigm for whole-slide image (WSI) analysis in computational pathology. However, existing MIL aggregators are still typically trained from scratch for each downstream task, relying on limited slide-level labels to learn both aggregation mechanisms and downstream discriminative representations simultaneously. As a result, they often suffer from unstable optimization, overfitting, and limited transferability. Similar to pretrained ResNet and Vision Transformer models in natural image learning, MIL also requires reusable pretrained initialization. However, high-quality slide-level pretraining data remain scarce, and MIL models are usually lightweight and weakly supervised, making large-scale pretraining difficult in practice. To address this challenge, we propose a distillation-based pretraining framework for MIL, which leverages two slide-level foundation models, TITAN and CARE, as teachers to transfer their representational knowledge into a diverse set of MIL architectures. To effectively balance supervision from different teachers, we further introduce an angular dispersion normalized distillation loss. The distilled weights are then used as initialization for downstream adaptation. We conduct systematic evaluations on 15 benchmark datasets under both linear probing and full-parameter fine-tuning, and further validate its advantages in few-shot scenarios. Experimental results show that pretraining generally improves MIL aggregators over from scratch training, especially in linear-probing and few-shot settings, while maintaining the computational efficiency of lightweight MIL models. Code is available at https://github.com/fu0201/MIL_Pretrained.
Chinese Translation
多实例学习（MIL）已成为计算病理学中全切片图像（WSI）分析的主要范式。然而，现有的MIL聚合器通常在每个下游任务中从头开始训练，依赖有限的切片级标签同时学习聚合机制和下游判别表示。因此，它们往往面临不稳定的优化、过拟合和有限的迁移能力。与自然图像学习中的预训练ResNet和Vision Transformer模型类似，MIL也需要可重用的预训练初始化。然而，高质量的切片级预训练数据仍然稀缺，MIL模型通常是轻量级和弱监督的，这使得大规模预训练在实践中变得困难。为了解决这一挑战，我们提出了一种基于蒸馏的MIL预训练框架，该框架利用两个切片级基础模型TITAN和CARE作为教师，将它们的表征知识转移到多样化的MIL架构中。为了有效平衡来自不同教师的监督，我们进一步引入了一种角度离散归一化蒸馏损失。蒸馏后的权重随后用作下游适应的初始化。我们在15个基准数据集上进行了系统评估，采用线性探测和全参数微调，并进一步验证其在少样本场景中的优势。实验结果表明，预训练通常比从头训练更能改善MIL聚合器的表现，特别是在线性探测和少样本设置中，同时保持轻量级MIL模型的计算效率。代码可在 https://github.com/fu0201/MIL_Pretrained 获取。
cs.CV / 38 / 2607.14710
Variational Inference for Bird's Eye View Segmentation in Autonomous Driving
用于自动驾驶的鸟瞰图分割的变分推断
Abstract
The bird's eye view (BEV) has emerged as a pivotal approach for environmental perception in autonomous driving, providing a unified spatial representation for vehicles. Nevertheless, despite BEV's significance in addressing the challenges inherent to autonomous driving, effectively fusing data from multiple camera sensors and operating in complex external driving environments remains a considerable challenge. To mitigate this issue, we recast the BEV segmentation problem within a variational inference framework. In this paper, we propose a novel transformer-based variational flow transformation network for BEV segmentation, denoted as TVB. Our architecture implicitly learns the mapping from multiple camera views to a unified canonical BEV map during training by exploiting posterior BEV supervision. TVB employs a conditional variational auto encoder (CVAE) as its backbone and produces multiple BEV map candidates. To augment the realism of the generated BEV maps, we integrate normalizing flows into the map generation process, enabling the construction of more complex and expressive probability distributions. Furthermore, we design a BEV-attention fusion (BAF) module that harnesses attention mechanisms to adaptively integrate the multiple candidate BEV maps. Experimental results, evaluated on both the nuScenes and OPV2Vdatasets, demonstrate that our proposed method achieves superior performance in multi-camera view BEV segmentation and lane environment perception.
Chinese Translation
鸟瞰图（BEV）已成为自动驾驶环境感知的关键方法，为车辆提供了统一的空间表示。然而，尽管BEV在解决自动驾驶固有挑战方面具有重要意义，但有效融合来自多个摄像头传感器的数据并在复杂的外部驾驶环境中操作仍然是一个相当大的挑战。为了解决这个问题，我们将BEV分割问题重新表述为变分推断框架。在本文中，我们提出了一种新颖的基于变压器的变分流变换网络用于BEV分割，称为TVB。我们的架构在训练过程中通过利用后验BEV监督隐式学习从多个摄像头视图到统一标准BEV图的映射。TVB采用条件变分自编码器（CVAE）作为其主干，并生成多个BEV图候选。为了增强生成BEV图的真实感，我们将归一化流整合到图生成过程中，使得构建更复杂和更具表现力的概率分布成为可能。此外，我们设计了一个BEV注意力融合（BAF）模块，利用注意力机制自适应地整合多个候选BEV图。实验结果在nuScenes和OPV2V数据集上评估，表明我们提出的方法在多摄像头视图BEV分割和车道环境感知方面实现了优越的性能。
cs.CV / 39 / 2607.14711
VideoSEMA: a scalable and efficient Mamba-like attention for video understanding
VideoSEMA：一种可扩展且高效的类Mamba注意力机制用于视频理解
Abstract
We present for video understanding (classification) a split space-time attention model, VideoSEMA, consisting of a scalable and efficient Mamba-like attention (SEMA) block in space and a softmax temporal attention in time. In each frame, SEMA attention applies a local window attention in parallel with a global averaging in a Mamba macro-architecture, which is called Mamba-like. Under certain rank conditions, we prove that the computationally cheaper split space-time attention is equivalent to full space-time attention. On benchmark K400 data sets, VideoSEMA out-performs heavier vision transformer and Mamba models. On benchmark SSv2 data, VideoSEMA leads in top-1 accuracy among models of similar parameter sizes. As image resolution scales up from standard $224^2$ to $1024^2$ on K400 and without fine-tuning, VideoSEMA degrades much more gracefully than VideoMamba in accuracy. It is promising to extend VideoSEMA to longer videos with a dilated/sparse temporal attention.
Chinese Translation
我们提出了一种用于视频理解（分类）的分裂时空注意力模型VideoSEMA，该模型由一个可扩展且高效的类Mamba注意力（SEMA）模块和一个软max时序注意力模块组成。在每一帧中，SEMA注意力在Mamba宏架构中并行应用局部窗口注意力和全局平均，这被称为类Mamba。在某些秩条件下，我们证明了计算成本更低的分裂时空注意力等同于完整的时空注意力。在基准K400数据集上，VideoSEMA的性能优于更重的视觉变换器和Mamba模型。在基准SSv2数据集上，VideoSEMA在相似参数规模的模型中在top-1准确率上领先。当图像分辨率从标准的$224^2$提升到$1024^2$时，VideoSEMA在K400上的准确率下降比VideoMamba更为平稳，并且在没有微调的情况下表现良好。扩展VideoSEMA以支持更长视频的稀疏/扩张时序注意力是有前景的。
cs.CV / 40 / 2607.14720
Causal-Adversarial Probing of Clinical Covariates for Prostate MRI Grading
前列腺MRI分级中临床协变量的因果对抗探测
Abstract
Deep learning models for prostate MRI-based cancer grading may encode clinical covariates that either reflect useful disease-related signal or non-generalising shortcut information, but their role is usually assumed. We propose a causal-reasoning framework for probing covariate dependence in MRI-based International Society of Urological Pathology (ISUP) Grade Group prediction. Rather than treating mpMRI as a direct cause of grade, we model MRI appearance and ISUP grade as observations of latent tumour pathology, and test whether candidate clinical variables act as nuisance correlates, disease-related proxies, or irrelevant covariates in the learned representation. We implement this using an adversarial framework that suppresses the decodability of individual clinical covariate at a time while preserving MRI-based grade prediction. The approach is developed and evaluated on 2,903 prostate MRI examinations, with external validation on 576 patients. We report a set of interesting and previously under-explored imaging-to-clinical-variable interactions in the context of deep learning generalisation. For examples, in binary ISUP Grade Group $\geq2$ classification, suppressing age, BMI, and alcohol use improved AUC by 1.23%, 0.84%, and 1.42%, respectively (all p < 0.05), suggesting reduced non-generalising covariate information; In contrast, suppressing PSA and prostate volume degraded AUC by 1.91% and 7.61% (all p < 0.001), indicating that these variables carried task-relevant signal. These findings show that adversarial covariate suppression can provide a practical representation-level analysis for distinguishing potentially harmful dependence from informative signal in prostate MRI grading models.
Chinese Translation
基于深度学习的前列腺MRI癌症分级模型可能编码了反映有用疾病相关信号或非普遍化捷径信息的临床协变量，但其作用通常被假定。我们提出了一种因果推理框架，用于探测MRI基础的国际泌尿病理学会（ISUP）分级组预测中的协变量依赖关系。我们将mpMRI视为潜在肿瘤病理的观察，而不是分级的直接原因，并测试候选临床变量在学习表示中是作为干扰相关变量、疾病相关代理变量还是无关协变量。我们使用对抗框架来实现这一点，该框架在保持基于MRI的分级预测的同时，抑制单个临床协变量的可解码性。该方法在2903例前列腺MRI检查中开发和评估，并在576名患者中进行了外部验证。我们报告了一组有趣且之前未充分探索的成像与临床变量之间的交互，特别是在深度学习泛化的背景下。例如，在二元ISUP分级组$
geq2$分类中，抑制年龄、BMI和酒精使用分别提高了AUC 1.23%、0.84%和1.42%（均p < 0.05），这表明减少了非普遍化协变量信息；相反，抑制PSA和前列腺体积则使AUC下降了1.91%和7.61%（均p < 0.001），表明这些变量携带了与任务相关的信号。这些发现表明，对抗性协变量抑制可以为区分潜在有害依赖与前列腺MRI分级模型中的信息信号提供实用的表示级分析。
cs.CV / 41 / 2607.14721
Multimodality as Supervision: Self-Supervised Specialization to the Test Environment via Multimodality
多模态作为监督：通过多模态实现对测试环境的自我监督专门化
Abstract
Cross-modal learning, i.e., learning to predict one modality from another, is a fundamental mechanism for self-supervision via leveraging multimodality. Many practical applications, e.g., deploying a household robot, involve devices that are equipped with a rich set of sensors that enable multimodal sensing in their test environment. This presents an opportunity to apply cross-modal learning to the multimodal data sensed by these devices to learn representations. Findings in developmental psychology also suggest that biological agents leverage it to build an effective representation of their surroundings. To study this, we propose a controlled setup, where we restrict a user device to just a given test environment. It results in a specialization setup where we attempt to develop a performant model for this specific test environment. Under this setup, we develop Test-Space Training (TST), which performs multimodal data collection in the test environment and performs self-supervised pre-training on it. We evaluate these models on various downstream tasks in the same environment. Under this setup, we find various interesting insights, such as collecting rich multimodal data only from the test environment and leveraging cross-modal learning, we can achieve competitive results with generalist models (e.g., DINOv2 and CLIP) pre-trained on large-scale internet datasets. This enables an alternative scenario where the need for external Internet-scale datasets for pre-training models is reduced. We also present a set of analyses and ablations that raise intriguing points on substituting data with (multi)modality, and how varying pre-training data enables a tradeoff between a model's abilities to specialise to a test environment, and generalize to held-out spaces.
Chinese Translation
跨模态学习，即从一种模态预测另一种模态，是通过利用多模态进行自我监督的基本机制。许多实际应用，例如部署家庭机器人，涉及配备丰富传感器的设备，这些传感器能够在其测试环境中进行多模态感知。这为将跨模态学习应用于这些设备感知的多模态数据以学习表示提供了机会。发展心理学的研究结果也表明，生物体利用这一机制来构建其周围环境的有效表示。为此，我们提出了一种受控设置，在该设置中，我们将用户设备限制在特定的测试环境中。这导致了一种专门化设置，我们试图为该特定测试环境开发一个高性能模型。在此设置下，我们开发了测试空间训练（Test-Space Training, TST），该方法在测试环境中进行多模态数据收集，并对其进行自我监督的预训练。我们在同一环境中的各种下游任务上评估这些模型。在此设置下，我们发现了一些有趣的见解，例如，仅从测试环境中收集丰富的多模态数据，并利用跨模态学习，我们可以与在大规模互联网数据集上预训练的通用模型（例如 DINOv2 和 CLIP）取得竞争性结果。这使得在预训练模型时对外部互联网规模数据集的需求减少。我们还提出了一系列分析和消融实验，提出了关于用（多）模态替代数据的引人入胜的观点，以及如何通过变化预训练数据来实现模型在专门化测试环境和对保留空间的泛化能力之间的权衡。
cs.CV / 42 / 2607.14726
AE-UAV: An Air-to-Air Event-Based UAV Tracking Benchmark and a Real-Time Frequency-Domain Tracker
AE-UAV：一种空对空事件驱动的无人机跟踪基准及实时频域跟踪器
Abstract
Air-to-air (A2A) unmanned aerial vehicle (UAV) tracking is fundamental to airborne remote sensing of low-altitude aerial targets. However, the deployment of continuous, real-time tracking systems on UAVs presents significant challenges. In A2A scenarios, traditional frame-based cameras suffer from severe performance degradation under low illumination, overexposure, and high-speed motion owing to their limited dynamic range and fixed temporal sampling. Although event cameras offer a promising alternative with microsecond temporal resolution and a high dynamic range, current research is bottlenecked by two primary issues: 1) the absence of dedicated A2A event-based datasets, and 2) the heavy reliance of existing trackers on GPU acceleration and extensive training data, rendering them impractical for resource-constrained UAVs. To bridge these gaps, we introduce AE-UAV, an air-to-air event-based UAV tracking benchmark. To the best of our knowledge, this is the first airborne-captured event camera dataset for A2A tracking, comprising 178 flight sequences with continuous-time cubic B-spline annotations. Furthermore, we propose the Fast-Slow Frequency-domain Tracking (FSFT) method. This lightweight, training-free framework seamlessly integrates frequency-domain template matching with search region prediction and detection-based drift correction. Extensive experiments demonstrate that FSFT operates at an ultra-fast 420 frames per second (FPS) on CPU-only hardware. It retains 93.97% of the accuracy of state-of-the-art GPU-dependent methods while delivering a 5.32-fold effective speedup and exhibiting superior temporal resolution generalization, thereby providing a highly efficient and robust solution for airborne remote sensing of aerial targets. The dataset and source code are available at https://github.com/MSP-xEN/AE-UAV.
Chinese Translation
空对空（A2A）无人机（UAV）跟踪是低空空中目标遥感的基础。然而，在无人机上部署连续的实时跟踪系统面临重大挑战。在A2A场景中，传统的基于帧的摄像头在低光照、过曝和高速运动下表现严重下降，这主要是由于其有限的动态范围和固定的时间采样。尽管事件摄像头以微秒级的时间分辨率和高动态范围提供了有希望的替代方案，但当前研究受到两个主要问题的制约：1）缺乏专门的A2A事件驱动数据集，2）现有跟踪器严重依赖GPU加速和大量训练数据，使其在资源受限的无人机上难以应用。为了解决这些问题，我们提出了AE-UAV，一个空对空事件驱动的无人机跟踪基准。据我们所知，这是第一个用于A2A跟踪的空中捕获事件摄像头数据集，包含178个飞行序列，并附有连续时间的三次B样条注释。此外，我们提出了快速-慢速频域跟踪（FSFT）方法。该轻量级、无训练的框架无缝集成了频域模板匹配、搜索区域预测和基于检测的漂移校正。大量实验表明，FSFT在仅使用CPU的硬件上以超快的420帧每秒（FPS）运行。它保留了最先进的依赖GPU的方法的93.97%的准确率，同时实现了5.32倍的有效加速，并表现出优越的时间分辨率泛化能力，从而为空中目标的遥感提供了一种高效且稳健的解决方案。数据集和源代码可在 https://github.com/MSP-xEN/AE-UAV 获取。
cs.CV / 43 / 2607.14727
WorkDrive: Roadwork Chain of Causation for Autonomous Driving
WorkDrive：自主驾驶的道路施工因果链
Abstract
Autonomous driving vision-language models (VLMs) struggle in roadwork zones, where familiar visual cues such as lane markings and permanent signs are altered or absent, and temporary devices such as cones and barriers redefine the drivable corridor. VLMs can detect these objects, but without explicit guidance they anchor their reasoning on familiar elements from pre-training and fail to connect work-zone observations to correct planning decisions. We propose WorkDrive, a framework that constructs perception-grounded causal reasoning for work zones and aligns it with trajectory prediction. An automated multitask perception pipeline extracts structured scene facts and injects them into a Chain-of-Causation (CoC) annotation pipeline, redirecting the annotator's attention to domain-specific elements. The resulting reasoning labels are used for supervised fine-tuning, followed by reinforcement learning with a single reward: consistency between lateral meta-actions and the predicted trajectory. On ROADWork, the largest public work-zone dataset, the proposed roadwork CoC reduces trajectory average displacement error (ADE) by 9.0\%, and consistency-based GRPO yields a further 3.0\%, achieving progressive improvement over the trajectory-only baseline. Code and data will be publicly released.
Chinese Translation
自主驾驶的视觉-语言模型（VLMs）在道路施工区域面临挑战，因为熟悉的视觉线索如车道标记和永久性标志被改变或缺失，而临时设备如锥形标和障碍物重新定义了可驾驶走廊。VLMs可以检测这些物体，但在没有明确指导的情况下，它们将推理锚定在预训练时熟悉的元素上，无法将施工区的观察与正确的规划决策联系起来。我们提出了WorkDrive，一个为施工区构建感知基础的因果推理框架，并将其与轨迹预测对齐。一个自动化的多任务感知管道提取结构化场景事实，并将其注入因果链（Chain-of-Causation, CoC）注释管道，重新引导注释者的注意力到特定领域的元素。生成的推理标签用于监督微调，随后进行强化学习，采用单一奖励：横向元动作与预测轨迹之间的一致性。在ROADWork这个最大的公共施工区数据集上，所提出的道路施工因果链将轨迹平均位移误差（ADE）降低了9.0%，而基于一致性的GRPO进一步提高了3.0%，实现了相对于仅基于轨迹的基线的逐步改进。代码和数据将公开发布。
cs.CV / 44 / 2607.14728
VQ-Touch: A Data-Efficient Tactile Generation Framework Across Sensors and Scenarios
VQ-Touch：一种跨传感器和场景的数据高效触觉生成框架
Abstract
Tactile image generation significantly reduces the dependency on expensive and wear-prone sensors by synthesizing high-fidelity tactile data, offering an efficient solution for tactile information acquisition in robotic perception and human-machine interaction systems. However, existing methods depend on large-scale, diverse datasets from specific sensors and lack efficient data utilization and robust generalization capabilities, struggling in vision-limited environments. To address this, we introduce VQ-Touch, a tactile generation framework that supports both cross-sensor and multi-scenario applications. Specifically, to efficiently extract complex deformation and texture features from the data, we propose DM-VQGAN, an effective tactile representation learner. Furthermore, we introduce a discrete diffusion decoder with a unified conditioning interface, supporting multimodal generation tasks such as images and labels, and enhances the model's generalization capability through few-shot mixed training, thus achieving compatibility with current mainstream sensors and their variants. Experiments show that VQ-Touch surpasses state-of-the-art methods in multiple tasks.
Chinese Translation
触觉图像生成通过合成高保真触觉数据显著降低了对昂贵且易磨损传感器的依赖，为机器人感知和人机交互系统中的触觉信息获取提供了高效解决方案。然而，现有方法依赖于特定传感器的大规模多样化数据集，缺乏有效的数据利用和稳健的泛化能力，在视觉受限环境中表现不佳。为了解决这一问题，我们提出了VQ-Touch，一个支持跨传感器和多场景应用的触觉生成框架。具体而言，为了高效提取数据中的复杂变形和纹理特征，我们提出了DM-VQGAN，一种有效的触觉表示学习器。此外，我们引入了一种具有统一条件接口的离散扩散解码器，支持图像和标签等多模态生成任务，并通过少量样本混合训练增强模型的泛化能力，从而实现与当前主流传感器及其变体的兼容性。实验表明，VQ-Touch在多个任务中超越了最先进的方法。
cs.CV / 45 / 2607.14737
GeoDetect: Geometric Adversarial Detection for VLPs
GeoDetect：针对视觉语言预训练模型的几何对抗检测
Abstract
Vision-language pre-trained models (VLPs) are widely used in real-world applications. However, they remain vulnerable to adversarial attacks. Although adversarial detection methods have demonstrated success in single-modality settings (either vision or language), their effectiveness and reliability in multimodal models such as VLPs remain largely unexplored. In this work, we study the geometry of VLP embedding spaces and observe structured anisotropy that differs from unimodal vision models. Our theoretical analysis shows that under this anisotropic structure, adversarial attacks increase the expected geometric separation between clean and adversarial examples (AEs). Specifically, we demonstrate that AEs consistently exhibit greater expected distances to randomly sampled points than their clean counterparts, indicating that AEs tend to push representations out of manifold regions. Building on these insights, we propose GeoDetect, which leverages these off-manifold deviations via geometric scores to identify AEs. Through comprehensive evaluations, we show that our approach reliably detects AEs across diverse VLP architectures and threat settings, covering unimodal and multimodal attacks as well as adaptive attacks, thereby providing a robust and practical approach to improving the safety and reliability of these models.
Chinese Translation
视觉语言预训练模型（VLPs）在实际应用中被广泛使用。然而，它们仍然容易受到对抗攻击的影响。尽管对抗检测方法在单一模态设置（无论是视觉还是语言）中已显示出成功，但在多模态模型如VLPs中的有效性和可靠性仍然未得到充分探索。在本研究中，我们研究了VLP嵌入空间的几何特性，并观察到与单模态视觉模型不同的结构性各向异性。我们的理论分析表明，在这种各向异性结构下，对抗攻击增加了干净样本与对抗样本（AEs）之间的预期几何分离。具体而言，我们证明了对抗样本相较于其干净样本在随机采样点的预期距离上始终更大，这表明对抗样本倾向于将表示推离流形区域。基于这些见解，我们提出了GeoDetect，该方法利用这些流形外的偏差通过几何评分来识别对抗样本。通过全面的评估，我们展示了我们的方法在不同的VLP架构和威胁设置下可靠地检测对抗样本，涵盖了单模态和多模态攻击以及自适应攻击，从而为提高这些模型的安全性和可靠性提供了一种稳健且实用的方法。
cs.CV / 46 / 2607.14739
FoMoVLA: Bridging Visual Foresight and Motion Guidance for Vision-Language-Action Models
FoMoVLA：桥接视觉前瞻与运动引导的视觉-语言-动作模型
Abstract
Vision-Language-Action (VLA) models have achieved impressive results in visuomotor policy learning, yet remain fundamentally reactive, mapping current observations and language to actions without explicit forward prediction of world dynamics. Existing visual foresight methods predict future visual states but lack explicit motion guidance: they show where to go but not how to get there. We argue that future feature prediction and sparse point tracking are naturally complementary: the former provides the goal state, while the latter captures the continuous motion path toward it. We propose FoMoVLA, a framework that augments VLA representations with explicit spatio-temporal supervision by jointly learning future feature foresight and sparse 2D point tracking, enhancing the continuous action policy. FoMoVLA introduces compact foresight tokens to decode future feature states, decodes sparse temporal 2D point trajectories to model compact geometric motion, and couples both through a lightweight future-conditioned cross-attention module that enables consistent reasoning between anticipated states and point dynamics. Extensive experiments on LIBERO, RoboCasa GR-1 Tabletop, and LIBERO-Plus demonstrate state-of-the-art performance and strong zero-shot generalization. Project page is available at https://liauto-research.github.io/FoMoVLA.
Chinese Translation
视觉-语言-动作（VLA）模型在视觉运动策略学习中取得了令人瞩目的成果，但仍然是根本性的反应式，映射当前的观察和语言到动作，而没有对世界动态进行明确的前向预测。现有的视觉前瞻方法预测未来的视觉状态，但缺乏明确的运动引导：它们展示了去往何处，但没有说明如何到达那里。我们认为，未来特征预测和稀疏点跟踪是自然互补的：前者提供目标状态，而后者捕捉到达目标的连续运动路径。我们提出了FoMoVLA，一个通过联合学习未来特征前瞻和稀疏2D点跟踪来增强VLA表示的框架，提供明确的时空监督，从而增强连续动作策略。FoMoVLA引入紧凑的前瞻令牌以解码未来特征状态，解码稀疏的时间2D点轨迹以建模紧凑的几何运动，并通过轻量级的未来条件交叉注意模块将两者结合，使得预期状态与点动态之间能够进行一致的推理。在LIBERO、RoboCasa GR-1桌面和LIBERO-Plus上的广泛实验表明了其最先进的性能和强大的零样本泛化能力。项目页面可访问：https://liauto-research.github.io/FoMoVLA。
cs.CV / 47 / 2607.14743
On the Disagreement in Perturbation-based xAI -- Benchmarking Perturbation Choices for Flood Detection from SAR Images
关于基于扰动的可解释人工智能中的分歧——针对合成孔径雷达图像洪水检测的扰动选择基准测试
Abstract
Perturbation-based xAI methods are widely used to analyze the behavior and predictions of deep learning models. By altering input regions and measuring the resulting changes in class probabilities with respect to the original image, they assign relevance scores and generate heatmaps that reflect each region's contribution to the prediction. Despite their apparent simplicity, however, perturbation-based methods are sensitive to parameter choices. In this work, we focus on two key parameters of the perturbation pipeline, namely the patch geometry, including the size and shape of the perturbed regions, and the perturbation type, defined by the replacement scheme. Grounded in the use case of flood detection from Synthetic Aperture Radar imagery, we conduct a comprehensive investigation of how relevance estimation changes under different perturbation settings. Beyond visual inspection of the resulting relevance maps, we evaluate their consistency across perturbation strategies and their faithfulness to the model's reasoning. We demonstrate how different perturbation choices can steer the resulting relevance maps, yielding ambiguous and even contradictory explanations. Our findings emphasize the importance of methodological settings in perturbation-based xAI. They underscore the need to carefully inspect and evaluate perturbation choices and to treat them as an integral part when interpreting explanations, ensuring a robust understanding of both the explanations and model predictions.
Chinese Translation
基于扰动的可解释人工智能（xAI）方法被广泛用于分析深度学习模型的行为和预测。通过改变输入区域并测量相对于原始图像的类别概率变化，它们分配相关性分数并生成反映每个区域对预测贡献的热图。然而，尽管这些方法表面上看起来简单，基于扰动的方法对参数选择非常敏感。在本研究中，我们关注扰动流程中的两个关键参数，即补丁几何形状，包括扰动区域的大小和形状，以及由替换方案定义的扰动类型。基于合成孔径雷达图像的洪水检测应用案例，我们全面调查了在不同扰动设置下相关性估计的变化。除了对生成的相关性图进行视觉检查外，我们还评估了它们在不同扰动策略下的一致性以及对模型推理的忠实度。我们展示了不同的扰动选择如何影响生成的相关性图，从而导致模糊甚至矛盾的解释。我们的研究结果强调了在基于扰动的可解释人工智能中方法设置的重要性，强调了仔细检查和评估扰动选择的必要性，并将其视为解释时的一个重要组成部分，以确保对解释和模型预测的稳健理解。
cs.CV / 48 / 2607.14760
Clean-Reference Streaming Detection of Lens Occlusion and Photometric Transitions for Camera Tamper Monitoring
清洁参考流检测镜头遮挡和光度过渡以监控摄像机篡改
Abstract
A surveillance camera is an image sensor whose silent physical degradation invalidates every downstream consumer of its data. In-situ integrity alarms for such vision sensors require low false-alarm rates, bounded computation, and diagnosable behavior under nuisance illumination changes. This paper studies a deliberately narrow streaming integrity monitor for two low-cost sensor-fault signatures: texture-collapsing lens occlusion and abrupt photometric scene transition. The detector compares sampled luminance and local-gradient statistics with a clean-only sliding reference, applies coarse-grid structured-light rejection and mode/rapid-brightness suppression, and emits at most one notification per tamper episode. We formalize the decision predicates and derive a consistency rule for when rapid-brightness suppression makes the scene-transition path unreachable. On 320 in-scope controlled sequences, the default state machine attains 0.800 F1 and 0.822 balanced accuracy (significantly better paired correctness than the strongest baseline, though the F1 margin is not statistically resolved); on a magnitude-swept public audit it attains the highest partial AUC under a 5\% false-alarm budget, and a separate extended-stress FPR-constrained sweep reaches 0.925 recall at 0.025 false-positive rate. Public Xiph, Bremen IoT, and UHCTD diagnostics show the fixed predicates preserve low false alarms while recall concentrates inside the declared envelope (UHCTD in-scope covered recall 0.667 versus 0.016 out of scope), and a 9.09-camera-hour verified-negative public audit records zero false alarms. The method is best interpreted as an auditable sensor-health subsystem rather than a universal camera-tamper classifier.
Chinese Translation
监控摄像机是一种图像传感器，其静默的物理退化使其数据的每个下游消费者失效。对于这种视觉传感器，原位完整性警报需要低误报率、有限的计算量以及在干扰光照变化下可诊断的行为。本文研究了一种故意狭窄的流式完整性监测器，针对两种低成本传感器故障特征：纹理崩溃的镜头遮挡和突发的光度场景过渡。该检测器将采样的亮度和局部梯度统计数据与仅清洁的滑动参考进行比较，应用粗网格结构光拒绝和模式/快速亮度抑制，并在每次篡改事件中最多发出一次通知。我们形式化了决策谓词，并推导出快速亮度抑制使场景过渡路径不可达的一致性规则。在320个受控序列中，默认状态机达到了0.800的F1值和0.822的平衡准确率（配对正确性显著优于最强基线，尽管F1的差距在统计上未得到解决）；在一次幅度扫描的公共审计中，它在5%的误报预算下达到了最高的部分AUC，而在一次单独的扩展压力FPR约束扫描中，在0.025的假阳性率下达到了0.925的召回率。公共的Xiph、Bremen IoT和UHCTD诊断显示，固定谓词保持低误报，同时召回率集中在声明的范围内（UHCTD范围内的覆盖召回为0.667，而范围外为0.016），并且一次9.09摄像机小时的验证负面公共审计记录了零误报。该方法最好被解读为可审计的传感器健康子系统，而不是通用的摄像机篡改分类器。
cs.CV / 49 / 2607.14765
Rare Concept Generation via Counterfactual Inference in Diffusion Models
通过反事实推理在扩散模型中生成稀有概念
Abstract
Rare concept generation focuses on synthesizing customized images conditioned on text prompts that describe objects with unusual attributes. Previous works failed to align the generated images with rare concepts, resulting in incorrect attribute rendering or inconsistent composition of concepts. Such failures, as we observed, stem from the inherent common knowledge bias in the training stage of diffusion models, where objects are strongly associated with their common attributes, making it difficult to break these associations when generating rare concepts. To address such challenges, in this paper, we propose a novel Counterfactual Inference-based Diffusion approach, dubbed CI-Diff. CI-Diff blocks the interference of the model's inherent common knowledge bias and utilizes the Natural Direct Effect to capture the independent influence of the text prompt of rare concepts on image generation so that decoupling the unusual attributes from the rare concepts. To this end, we reformulate the classifier-free guidance mechanism to highlight the atypical attributes. To the best of our knowledge, we are the first to introduce causal inference into the rare concept generation task. Extensive experiments on the RareBench benchmark validate the superiority of CI-Diff over state-of-the-art diffusion models. Our code can be accessed from https://github.com/200204jzy/CI-Diff.
Chinese Translation
稀有概念生成专注于合成基于描述具有不寻常属性的物体的文本提示的定制图像。之前的研究未能将生成的图像与稀有概念对齐，导致属性渲染不正确或概念组合不一致。我们观察到，这种失败源于扩散模型训练阶段固有的常识偏见，在该阶段，物体与其常见属性之间存在强关联，使得在生成稀有概念时难以打破这些关联。为了解决这些挑战，本文提出了一种新颖的基于反事实推理的扩散方法，称为 CI-Diff。CI-Diff 阻止了模型固有常识偏见的干扰，并利用自然直接效应捕捉稀有概念的文本提示对图像生成的独立影响，从而将不寻常的属性与稀有概念解耦。为此，我们重新构建了无分类器引导机制，以突出不典型属性。根据我们所知，我们是首个将因果推理引入稀有概念生成任务的研究。对 RareBench 基准的广泛实验验证了 CI-Diff 相较于最先进的扩散模型的优越性。我们的代码可以从 https://github.com/200204jzy/CI-Diff 获取。
cs.CV / 50 / 2607.14807
TAMF-VTON: Texture-Aware Mask-Free Virtual Try-On via High-Fidelity Image Synthesis
TAMF-VTON：基于纹理感知的无掩膜虚拟试穿通过高保真图像合成
Abstract
Recent diffusion-based virtual try-on (VTON) methods remain limited by their reliance on segmentation masks, insufficient preservation of fine-grained textures, and limited support for arbitrary multi-garment compositions. Consequently, existing approaches still face significant challenges in real-world e-commerce deployment. We present TAMF-VTON, a texture-aware, mask-free framework that enables high-fidelity image synthesis under practical unconstrained conditions. Our method requires no human parsing or inpainting masks at inference time and supports diverse garment styles, categories, and quantities, enabling the simultaneous transfer of multiple items while preserving body structure and intricate texture details. This is achieved through a unified generative pipeline with three key components: (1) a lightweight Mixture-of-Experts (MoE) adaptation scheme that enables efficient fine-tuning without compromising the base model's general editing capabilities; (2) a frequency-domain supervision mechanism that explicitly optimizes high-frequency spectral consistency to preserve high-fidelity textures; and (3) a robust data curation pipeline employing an adaptive inpainting strategy to simulate the inverse VTON process for high-quality training pair generation. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods in both quantitative metrics and perceptual quality. Optimized for efficiency, the model achieves inference in under 15 seconds per image on an NVIDIA RTX 4090 with INT4 quantization. By combining mask-free operation, flexible multi-garment composition, faithful texture preservation, and efficient inference on consumer hardware, TAMF-VTON demonstrates a commercially viable solution for scalable deployment in real-world digital fashion scenarios. The project is available at https://www.style3d.ai/ai-photoshoot/virtual-clothing-try-on.
Chinese Translation
近期基于扩散的虚拟试穿（VTON）方法仍然受到依赖分割掩膜、细粒度纹理保留不足以及对任意多服装组合支持有限的限制。因此，现有方法在实际电子商务部署中仍面临重大挑战。我们提出了TAMF-VTON，一个纹理感知的无掩膜框架，能够在实际无约束条件下实现高保真图像合成。我们的方法在推理时无需人工解析或修补掩膜，并支持多样的服装风格、类别和数量，能够同时传输多个物品，同时保留身体结构和复杂的纹理细节。这是通过一个统一的生成管道实现的，包含三个关键组件：（1）一个轻量级的专家混合（Mixture-of-Experts, MoE）适应方案，使得高效微调成为可能，而不影响基础模型的通用编辑能力；（2）一个频域监督机制，明确优化高频谱一致性，以保留高保真纹理；（3）一个强大的数据策划管道，采用自适应修补策略模拟逆VTON过程，以生成高质量的训练对。大量实验表明，我们的方法在定量指标和感知质量上均优于最先进的方法。该模型经过优化，能够在NVIDIA RTX 4090上以INT4量化实现每张图像低于15秒的推理。通过结合无掩膜操作、灵活的多服装组合、忠实的纹理保留以及在消费硬件上的高效推理，TAMF-VTON展示了在现实数字时尚场景中可扩展部署的商业可行解决方案。该项目可在 https://www.style3d.ai/ai-photoshoot/virtual-clothing-try-on 获取。
cs.CV / 51 / 2607.14809
An LLM-Based Automatic Sportscast Solution for Robot Soccer Matches
基于大型语言模型的机器人足球比赛自动解说解决方案
Abstract
RoboCup has always been a scenario to develop systems that solve real-world problems. Driven by the main goal of playing against the 2050 FIFA World Cup champions, the RoboCup Soccer leagues need to constantly measure how the research community is progressing. Computing visual statistics from match videos is a crucial way to track this evolution. To address this challenge, this paper introduces a fully autonomous, real-time sports commentator for RoboCup matches. By bridging the gap between raw kinematic tracking and natural language generation, our neuro-symbolic architecture extracts precise statistics from video streams and turns them into fluent, hallucination-free narration. The proposed system is capable of generating statistics and commentary both during live match streaming and in post-game analysis, easily adapting to the new dynamism of the league where different humanoid robots of different sizes share the field. Supplemental materials are available at https://lab-rococo-sapienza.github.io/MARIO/
Chinese Translation
RoboCup 一直以来都是开发解决现实世界问题系统的场景。为了实现与 2050 年 FIFA 世界杯冠军对抗的主要目标，RoboCup 足球联赛需要不断衡量研究社区的进展。从比赛视频中计算视觉统计数据是跟踪这一演变的重要方式。为了解决这一挑战，本文提出了一种完全自主、实时的 RoboCup 比赛体育解说员。通过弥合原始运动追踪与自然语言生成之间的差距，我们的神经符号架构从视频流中提取精确的统计数据，并将其转化为流畅且无幻觉的解说。所提出的系统能够在现场比赛直播和赛后分析中生成统计数据和解说，轻松适应不同大小的人形机器人共享场地的新动态。补充材料可在 https://lab-rococo-sapienza.github.io/MARIO/ 获取。
cs.CV / 52 / 2607.14821
Blurring Modal Boundaries: A Unified Survey from Single- to Multi-Modal Person Re-ldentification
模态边界模糊化：从单模态到多模态人员重识别的统一综述
Abstract
Person re-identification (ReID) serves as a critical component in intelligent surveillance systems, aiming to match identities across disjoint camera networks. While traditional methods primarily rely on single-modal RGB imagery, they are often constrained by environmental challenges such as low illumination and occlusion. To overcome these limitations, the field is rapidly evolving toward cross-modal and multi-modal paradigms. This survey presents a comprehensive overview of this transition, systematically reviewing key cross-modal tasks including visible-infrared (VI-ReID), text-image (TI-ReID), sketch-based (Sketch-ReID), and the emerging Non-Line-of-Sight (NLOS) ReID, which extends perception beyond direct visibility. Furthermore, we examine tri-spectral and multi-modal fusion ReID, discussing how complementary information from diverse sensors enhances robustness. Beyond summarizing datasets, challenges, and methodologies, we propose a Transformer-based baseline framework for visible-infrared ReID, designed to effectively capture modality-invariant features. Finally, based on the current landscape, we outline several promising directions for future research.
Chinese Translation
人员重识别（ReID）是智能监控系统中的一个关键组成部分，旨在跨越不相连的摄像头网络匹配身份。尽管传统方法主要依赖单模态RGB图像，但它们常常受到低光照和遮挡等环境挑战的限制。为了克服这些局限，该领域正迅速向跨模态和多模态范式发展。本综述全面概述了这一转变，系统回顾了关键的跨模态任务，包括可见-红外（VI-ReID）、文本-图像（TI-ReID）、草图基础（Sketch-ReID）以及新兴的非视线（NLOS）ReID，后者将感知扩展到直接可见之外。此外，我们还考察了三谱和多模态融合ReID，讨论了来自不同传感器的互补信息如何增强鲁棒性。除了总结数据集、挑战和方法论外，我们提出了一种基于Transformer的可见-红外ReID基线框架，旨在有效捕捉模态不变特征。最后，基于当前的研究现状，我们概述了未来研究的几个有前景的方向。
cs.CV / 53 / 2607.14836
Physics-Informed Diffusion for Biomechanically Plausible 3D Sign Language Generation
基于物理信息的扩散模型用于生物力学合理的三维手语生成
Abstract
Sign language production, which generates continuous 3D skeletal motion from spoken language input, must simultaneously satisfy two constraints: semantic fidelity, so that a deaf viewer can recognize the intended sequence of glosses, and biomechanical plausibility, so that the generated skeleton respects anatomical constraints. Existing approaches optimize semantic reconstruction through coordinate-based objectives that treat the skeleton as an unstructured vector, thus allowing for bone length drift, joint angle violations, and temporarily locked fingers. We introduce PIDiffSign, a physics-informed diffusion model for gloss-to-pose translation that incorporates anatomical constraints into both the architecture and training objective. The model uses a Transformer encoder-decoder, where the decoder is conditioned on the diffusion time step through adaptive zero-initialized layer normalization and cross-attends to gloss representations. A differentiable geometry module enforces bone length consistency and biologically valid joint angles throughout generation. Training combines anthropomorphic, kinematic, angular, and finger-joint constraints with a contrastive gloss-pose alignment loss and classifier-free guidance for semantically conditioned sampling. Experiments on the PHOENIX14T and CSL-Daily benchmarks show consistent improvements over a strong diffusion baseline in pose accuracy, joint-angle correctness, distributional realism, and back-translation quality. These results demonstrate that physics-informed diffusion improves both motion realism and semantic fidelity for sign language generation.
Chinese Translation
手语生成是指从口语输入生成连续的三维骨骼运动，这一过程必须同时满足两个约束：语义保真性，以便聋人观众能够识别预期的手语序列，以及生物力学合理性，以确保生成的骨骼符合解剖学约束。现有方法通过基于坐标的目标优化语义重建，将骨骼视为非结构化向量，从而允许骨长漂移、关节角度违规和暂时锁定的手指。我们提出了PIDiffSign，一种基于物理信息的扩散模型，用于手语到姿势的转换，结合了解剖学约束于模型架构和训练目标中。该模型使用Transformer编码器-解码器，其中解码器通过自适应零初始化层归一化条件于扩散时间步，并与手语表示进行交叉关注。一个可微分几何模块在生成过程中强制执行骨长一致性和生物学有效的关节角度。训练结合了类人、运动学、角度和手指关节约束，以及对比手语-姿势对齐损失和无分类器引导的语义条件采样。对PHOENIX14T和CSL-Daily基准的实验表明，在姿势准确性、关节角度正确性、分布现实性和反向翻译质量方面，相较于强大的扩散基线模型有了一致的改善。这些结果表明，基于物理信息的扩散模型在手语生成中提高了运动的现实性和语义的保真性。
cs.CV / 54 / 2607.14876
Rotational Motion-Induced Error Compensation for Phase-Shifting Profilometry-Based Eye Reconstruction
基于相位移轮廓测量的眼睛重建中的旋转运动引起的误差补偿
Abstract
With the proliferation of immersive Head-Mounted Displays (HMDs) for Virtual and Augmented Reality (VR/AR), reliable and high-precision eye tracking has become increasingly important. Conventional 2D image-based methods offer low system complexity but remain limited in stability, accuracy, and robustness. Three-dimensional ocular surface reconstruction can provide richer geomet-ric information, and structured light profilometry is particularly attractive because it enables dense and accurate surface measurement. However, Phase-Shifting Profilometry (PSP), which estimates phase from sequentially acquired fringe images, is highly susceptible to motion-induced errors when the eye rotates between frames. This study proposes a rotational motion compensation framework for PSP-based dynamic 3D eye reconstruction. Relative eye rotation is estimated from image-based motion cues using a user-specific 3D eye model in a spherical-coordinate domain. The estimated motion is then used to compensate for camera-pixel mismatch and phase-shift errors caused by inter-frame rotation. A region-wise optimization strategy is further introduced to reduce residual artifacts by inde-pendently refining the compensation strength in different ocular regions. Experiments with a rotating fake eye under non-uniform motion demonstrate that the proposed method substantially suppresses motion-induced deformation and improves reconstruction accuracy. An additional experiment with a non-spherical rigid object indicates that the compensation principle is not restricted to spherical eye geometry. These results establish a practical basis for stable PSP-based dynamic 3D eye reconstruction toward future high-precision eye tracking in immersive environments.
Chinese Translation
随着虚拟现实（VR）和增强现实（AR）沉浸式头戴显示器（HMD）的普及，可靠且高精度的眼动追踪变得愈发重要。传统的基于二维图像的方法虽然系统复杂度低，但在稳定性、准确性和鲁棒性方面仍然有限。三维眼球表面重建能够提供更丰富的几何信息，而结构光轮廓测量尤其具有吸引力，因为它能够实现密集且准确的表面测量。然而，相位移轮廓测量（Phase-Shifting Profilometry，PSP）在眼睛在帧间旋转时，容易受到运动引起的误差影响。本研究提出了一种基于PSP的动态三维眼睛重建的旋转运动补偿框架。通过在球坐标域中使用用户特定的三维眼模型，从基于图像的运动线索中估计相对眼部旋转。然后，利用估计的运动来补偿由于帧间旋转引起的相机像素不匹配和相位移误差。进一步引入了一种区域优化策略，通过在不同眼部区域独立调整补偿强度来减少残余伪影。在非均匀运动下的旋转假眼实验表明，所提出的方法显著抑制了运动引起的变形，并提高了重建精度。对非球形刚性物体的额外实验表明，补偿原理并不局限于球形眼几何。这些结果为未来在沉浸式环境中实现高精度眼动追踪的稳定PSP基动态三维眼睛重建奠定了实用基础。
cs.CV / 55 / 2607.14897
Selectivity Drives Efficiency: Dataset Pruning for Visual Place Recognition
选择性驱动效率：视觉地点识别的数据集剪枝
Abstract
Recent visual place recognition (VPR) studies have increasingly relied on large-scale datasets to train more robust and discriminative models. Although this trend significantly improves recognition performance, it also introduces substantial storage and training costs, especially when new architectures or training strategies need to be repeatedly developed and evaluated. Dataset pruning (DP) provides a promising way to improve data efficiency by retaining only informative training data. However, conventional DP methods mainly follow the sample-wise classification paradigm, which overlooks the relation-dependent training nature of VPR, where supervision is typically formed by image pairs rather than independent images. To address this issue, we propose a place-wise dataset pruning framework tailored for VPR. Instead of pruning individual images, our method treats each place as the basic pruning unit and introduces two complementary novel metrics, i.e., intra-place diversity (IPD) and inter-place similarity (IPS), to evaluate the training value of each place. By jointly considering these two metrics, our method ranks all places and constructs a compact yet informative coreset, thereby allowing the pruned dataset to still support the training of robust and discriminative VPR models. Extensive experiments demonstrate that our method consistently outperforms state-of-the-art DP baselines under different pruning ratios while reducing selection and training costs. Moreover, by pruning a merged dataset roughly 3.5$\times$ the size of GSV-Cities to a comparable scale, our coreset maintains highly competitive performance, achieving 94.5\% R@1 on MSLS-val and 97.0\% R@1 on Nordland with only NetVLAD. Codes will be made publicly available.
Chinese Translation
近年来，视觉地点识别（VPR）研究越来越依赖大规模数据集来训练更强大和具有区分性的模型。尽管这一趋势显著提高了识别性能，但它也带来了巨大的存储和训练成本，尤其是在需要反复开发和评估新的架构或训练策略时。数据集剪枝（DP）提供了一种有前景的方法，通过仅保留有信息的训练数据来提高数据效率。然而，传统的DP方法主要遵循样本级分类范式，忽视了VPR的关系依赖训练特性，其中监督通常是通过图像对而非独立图像形成的。为了解决这一问题，我们提出了一种针对VPR的地点级数据集剪枝框架。我们的算法不是剪枝单个图像，而是将每个地点视为基本的剪枝单元，并引入两个互补的新指标，即地点内多样性（IPD）和地点间相似性（IPS），以评估每个地点的训练价值。通过共同考虑这两个指标，我们的方法对所有地点进行排序，并构建一个紧凑而又信息丰富的核心集，从而使得剪枝后的数据集仍能支持强大且具有区分性的VPR模型的训练。大量实验表明，我们的方法在不同的剪枝比例下始终优于最先进的DP基线，同时降低了选择和训练成本。此外，通过将一个合并的数据集剪枝至大约3.5倍于GSV-Cities的规模，我们的核心集保持了高度竞争的性能，在MSLS-val上达到94.5%的R@1，在Nordland上达到97.0%的R@1，仅使用NetVLAD。代码将公开发布。
cs.CV / 56 / 2607.14898
FlashDecoder: Real-Time Latent-to-Pixel Streaming Decoder with Transformers
FlashDecoder：基于变换器的实时潜在到像素流解码器
Abstract
Real-time video generation demands fast decoding as much as fast denoising, yet current latent video diffusion models rely on 3D convolutional decoders that are slow and memory-intensive at high resolutions or for long video. We introduce FlashDecoder, a fast, memory-efficient pure-Transformer video decoder that decodes latents to pixels frame by frame. At each step, the current frame attends only to a fixed-size window of past frames through a rolling KV cache. The fixed temporal window keeps decoding fast and memory bounded regardless of video length, enabling constant-latency streaming. Because frames are processed sequentially, temporal causality is enforced without explicit attention masks, enabling training at resolutions up to 1080p and matching the reconstruction quality of convolutional decoders. On the Wan2.1 and Wan2.2 latent spaces, FlashDecoder matches each convolutional decoder in reconstruction quality (e.g., 41.55dB vs. 41.49dB PSNR at 1080p) while decoding 3.6x-4.7x faster with up to 11x less memory on a single H100 GPU. With architecture-aware inference optimizations, the speedup widens to 12x.
Chinese Translation
实时视频生成需要快速解码与快速去噪同样重要，但当前的潜在视频扩散模型依赖于在高分辨率或长视频情况下速度较慢且内存消耗大的3D卷积解码器。我们提出了FlashDecoder，这是一种快速且内存高效的纯变换器视频解码器，能够逐帧将潜在表示解码为像素。在每一步中，当前帧仅通过滚动KV缓存关注过去帧的固定大小窗口。固定的时间窗口使得解码速度快且内存占用受限，无论视频长度如何，从而实现恒定延迟的流式传输。由于帧是顺序处理的，时间因果关系在没有显式注意力掩码的情况下得以强制执行，使得在高达1080p的分辨率下进行训练成为可能，并且重建质量与卷积解码器相匹配。在Wan2.1和Wan2.2潜在空间中，FlashDecoder在重建质量上与每个卷积解码器相当（例如，在1080p时41.55dB对比41.49dB PSNR），同时在单个H100 GPU上解码速度快3.6到4.7倍，内存使用减少最多11倍。通过架构感知的推理优化，速度提升扩大至12倍。
cs.CV / 57 / 2607.14927
TanGO: Training-Free 3D Editing via Tangent-Space Guidance and Optimization
TanGO：通过切空间引导和优化实现无训练的3D编辑
Abstract
While recent flow-matching 3D generative models (e.g., VecSet) adopt structured representations, their tokens share global context, causing conventional training-free editing to suffer from semantic artifacts such as collapsed preserved regions or incomplete transformations. To address this, we propose TanGO, a training-free framework that enables adaptive per-token steering in the tangent space of generative dynamics. To realize this selective control, we formulate a one-step optimal control rule and determine the strength of each token's control signal using a von Mises-Fisher inspired directional discrepancy derived from the source and target velocity fields. Experiments show that TanGO substantially reduces structural artifacts and achieves state-of-the-art performance, outperforming existing 3D editing baselines. The code is publicly available at https://github.com/siw00-lim/TanGO.
Chinese Translation
尽管最近的流匹配3D生成模型（例如，VecSet）采用了结构化表示，但它们的标记共享全局上下文，导致传统的无训练编辑受到语义伪影的影响，例如保留区域的崩溃或变换的不完整。为了解决这个问题，我们提出了TanGO，一个无训练的框架，能够在生成动态的切空间中实现自适应的逐标记引导。为了实现这种选择性控制，我们制定了一种一步最优控制规则，并使用从源速度场和目标速度场导出的受von Mises-Fisher启发的方向差异来确定每个标记的控制信号强度。实验表明，TanGO显著减少了结构伪影，并实现了最先进的性能，超越了现有的3D编辑基线。代码已公开发布在 https://github.com/siw00-lim/TanGO。
cs.CV / 58 / 2607.14932
Benchmarking Face Recognition without Real Faces
无真实人脸的面部识别基准测试
Abstract
Synthetic face datasets have become effective enough to train face recognition models with accuracy rivaling that of models trained on real photographs. This progress sidesteps the ethical and legal burdens of collecting real biometric data, yet evaluation has not kept pace. Even studies that train entirely on synthetic images still rely on real-face benchmarks to measure performance, leaving the privacy problem only half solved. We ask whether synthetic datasets can replace real benchmarks for face recognition evaluation. We test 12 synthetic datasets against 7 established real benchmarks using 24 pre-trained models that span both convolutional and transformer architectures. Our evaluation covers biometric verification metrics, similarity score distributions, cross-model ranking consistency, and the underlying distributional properties of each dataset. Benchmarking fidelity varies widely across the synthetic candidates, but the two strongest, MorphFace and Vec2Face, reproduce the relative behavior of real benchmarks and reach agreement levels that fall within the natural disagreement already observed among the real benchmarks themselves. These results establish that well-constructed synthetic datasets can support reliable comparative evaluation for face recognition, moving the field closer to a fully synthetic and privacy-preserving pipeline for both training and benchmarking.
Chinese Translation
合成面部数据集已经足够有效，可以训练出与基于真实照片训练的模型相媲美的面部识别模型的准确性。这一进展规避了收集真实生物识别数据的伦理和法律负担，但评估尚未跟上。即使是完全基于合成图像进行训练的研究，仍然依赖于真实面孔基准来衡量性能，从而使隐私问题仅得到部分解决。我们探讨合成数据集是否可以替代真实基准进行面部识别评估。我们使用24个涵盖卷积和变换器架构的预训练模型，测试了12个合成数据集与7个已建立的真实基准。我们的评估涵盖了生物识别验证指标、相似性评分分布、跨模型排名一致性以及每个数据集的基础分布特性。合成候选者的基准测试保真度差异很大，但两个最强的，MorphFace和Vec2Face，重现了真实基准的相对行为，并达到了与真实基准之间已观察到的自然分歧相符的协议水平。这些结果表明，构建良好的合成数据集可以支持面部识别的可靠比较评估，使该领域更接近于一个完全合成且保护隐私的训练和基准测试流程。
cs.CV / 59 / 2607.14934
Still image and spatial-temporal tomato data enabling detection, segmentation, tracking, and video-instance segmentation using strong and weak labels
静态图像和时空番茄数据支持使用强标签和弱标签进行检测、分割、跟踪和视频实例分割
Abstract
In this manuscript we release two datasets for visual sensing of tomato plants grown in commercial-like settings and acquired using a robot. The first is BUTom21 which consists of still images and manual annotations. The second is BUTom-ST21 which consists of video-based data and semi-automated annotations through AI-based methods, referred to as pseudo-labels. In both cases, we provide pixel-level labels for the ripeness of the fruit. The aim is to provide the research community a challenging set of real-world imagery to explore methods to sense and estimate the state of tomato plants and their fruit, which is an important horticultural crop. Importantly, the spatial-temporal dataset provides individual fruit count and ripeness information enabling researchers to push the boundaries of field-based phenotyping.
Chinese Translation
在本文中，我们发布了两个用于视觉感知在商业环境中生长的番茄植物的数据集，这些数据集是通过机器人获取的。第一个数据集是BUTom21，它由静态图像和手动注释组成。第二个数据集是BUTom-ST21，它由基于视频的数据和通过人工智能方法进行的半自动注释（称为伪标签）组成。在这两种情况下，我们提供了果实成熟度的像素级标签。我们的目标是为研究社区提供一组具有挑战性的真实世界图像，以探索感知和估计番茄植物及其果实状态的方法，这是一种重要的园艺作物。重要的是，时空数据集提供了单个果实的计数和成熟度信息，使研究人员能够推动基于田间表型分析的边界。
cs.CV / 60 / 2607.14935
VideoChat3: Fully Open Video MLLM for Efficient and Generalist Video Understanding
VideoChat3：用于高效和通用视频理解的完全开放视频多模态大语言模型
Li, Xinhao, Zhu, Yuhan, Zeng, Xiangyu, Dong, Yuhao, Wu, Haoning, Zhang, Zhiqiu, Yang, Yuandong, Ma, Changlian, Zhang, Qingyu, Shi, Yansong, Chen, Xinyu, Chen, Haoran, Huang, Zizheng, Zhang, Jun, Ouyang, Kun, Sui, Lin, Yan, Ziang, Xu, Yicheng, Wang, Chenting, He, Yinan, Zhang, Hongjie, Wang, Yi, Qiao, Yu, Wang, Yali, Liu, Ziwei, Chen, Kai, Wang, Limin
Abstract
Recent advances in video understanding have spanned motion, long video, and streaming interaction, driving this field toward real-world applications. Despite this progress, current open-source models remain limited in several ways. They often struggle to generalize across diverse video types, making them effective only in specific domains. High computational demands further restrict their efficiency and scalability. Moreover, most models are only partially open, with key components such as training code, strategy, or datasets unavailable, which hinders reproducibility and slows community-driven development. To address these issues, we introduce VideoChat3, a fully open, efficient, and generalist video-centric MLLM. VideoChat3 advances video understanding through two complementary designs. For efficiency, we introduce Inflated 3D Vision Transformer (I3D-ViT) and Adaptive Frame Resolution for Streaming Video Perception, which enables efficient spatiotemporal representation and reduces the cost of processing video inputs during training and inference. For effectiveness, we develop a scalable video data synthesis pipeline that curates three diverse, high-quality training datasets: VideoChat3-Academic2M, VideoChat3-LV116K, and VideoChat3-OL617K, covering general, long-form, and streaming video scenarios, improving the model's generalization across domains. By integrating these designs, VideoChat3 achieves a rare balance of broad generalization and computational efficiency. Experiments across general, long-form, and streaming benchmarks demonstrate that VideoChat3 surpasses prior open-source models with equal or larger parameter counts with only 4B parameters and higher efficiency.
Chinese Translation
近年来，视频理解领域在运动、长视频和流媒体交互方面取得了显著进展，推动了该领域向现实应用的发展。尽管取得了这些进展，目前的开源模型在多个方面仍然存在局限性。它们往往难以在多样化的视频类型中进行泛化，因此仅在特定领域内有效。高计算需求进一步限制了它们的效率和可扩展性。此外，大多数模型仅部分开放，关键组件如训练代码、策略或数据集不可用，这妨碍了可重复性并减缓了社区驱动的发展。为了解决这些问题，我们提出了VideoChat3，一个完全开放、高效且通用的视频中心多模态大语言模型。VideoChat3通过两种互补设计推动视频理解的发展。在效率方面，我们引入了膨胀的3D视觉变换器（Inflated 3D Vision Transformer，I3D-ViT）和自适应帧分辨率用于流媒体视频感知，这使得时空表示更加高效，并降低了训练和推理过程中处理视频输入的成本。在有效性方面，我们开发了一个可扩展的视频数据合成管道，策划了三个多样化的高质量训练数据集：VideoChat3-Academic2M、VideoChat3-LV116K和VideoChat3-OL617K，涵盖了通用、长篇和流媒体视频场景，从而提高了模型在不同领域的泛化能力。通过整合这些设计，VideoChat3实现了广泛泛化与计算效率的罕见平衡。在通用、长篇和流媒体基准测试中的实验表明，VideoChat3在参数数量相等或更大的情况下，以仅4B参数超越了之前的开源模型，并且效率更高。
cs.CV / 61 / 2607.14941
Frequency-Structured Field Learning for Light-Field Disparity Estimation
频率结构场学习用于光场视差估计
Abstract
Light-field disparity estimation requires global consistency in smooth or textureless regions and local precision near occlusion boundaries, thin structures, and abrupt depth transitions. Existing methods address these requirements through EPI matching, cost-volume or focal-stack construction, view aggregation, or direct convolutional regression, often relying on local windows, discrete disparity hypotheses, memory-intensive volumes, or attention-based aggregation. We instead formulate disparity estimation at the field level, predicting disparity from globally and locally updated EPI-derived latent features without explicitly constructing a disparity volume. We introduce FreqLF, an EPI-guided Fourier-local framework that encodes angular parallax cues from horizontal and vertical EPI stacks together with central-view appearance features. These cues are projected into a latent field and updated through stacked hybrid Fourier-local layers. Fourier low-mode updates enable global feature interaction, while local convolutions preserve spatial variations needed for fine disparity detail. A coordinate-conditioned Gaussian-mixture decoder then predicts disparity, using the mixture mean as the final estimate. Experiments on the HCI 4D Light Field Benchmark show that FreqLF approaches the accuracy of strong supervised baselines while avoiding explicit cost-volume construction in the base model. Ablations confirm the complementary roles of the Fourier and local branches, and scaling experiments demonstrate practical behavior across spatial resolutions. These results suggest that Fourier-local latent field learning is a competitive alternative for light-field disparity estimation. The code will be published soon.
Chinese Translation
光场视差估计要求在平滑或无纹理区域保持全局一致性，并在遮挡边界、细结构和突变深度过渡附近保持局部精度。现有方法通过EPI匹配、代价体积或焦点堆栈构建、视图聚合或直接卷积回归来满足这些要求，通常依赖于局部窗口、离散视差假设、内存密集型体积或基于注意力的聚合。我们则在场级别上制定视差估计，从全局和局部更新的EPI导出的潜在特征中预测视差，而无需显式构建视差体积。我们引入了FreqLF，一个EPI引导的傅里叶局部框架，它将来自水平和垂直EPI堆栈的角度视差线索与中心视图外观特征编码在一起。这些线索被投影到一个潜在场中，并通过堆叠的混合傅里叶局部层进行更新。傅里叶低模式更新使全局特征交互成为可能，而局部卷积则保留了细致视差细节所需的空间变化。然后，一个坐标条件的高斯混合解码器预测视差，使用混合均值作为最终估计。在HCI 4D光场基准测试中的实验表明，FreqLF接近强监督基线的准确性，同时避免了在基础模型中显式构建代价体积。消融实验确认了傅里叶和局部分支的互补作用，缩放实验展示了在空间分辨率上的实用表现。这些结果表明，傅里叶局部潜在场学习是光场视差估计的一个具有竞争力的替代方案。代码将很快发布。
cs.CV / 62 / 2607.14945
Introspective Attention Modulation for Safe Text-to-Image Generation
内省注意力调节用于安全的文本到图像生成
Abstract
State-of-the-art flow based text-to-image (T2I) models exhibit remarkable generative abilities but remain vulnerable to producing unsafe content. Prior safety efforts range from concept erasure and prompt filtering to classifier-based gating. However, simple techniques like parameter efficient adaptations of the models easily bypass such guardrails. We introduce a unique principled approach that achieves safety by regulating the model's attention dynamics through inference-time introspection, exhibiting intrinsic robustness. Our method analyzes and rebalances attention activations throughout image synthesis, steering generations away from unsafe concepts while preserving semantic alignment. This introspective control ensures safety of deployed models. Across standard and adversarial safety benchmarks, our approach achieves remarkable safety scores while maintaining or even improving alignment and perceptual quality. Our results reveal that attention-space regulation offers a considerably more promising path to safer diffusion transformer based image generation than the existing concept erasing mechanism.Our code can be accessed at https://basim-azam.github.io/iam/
Chinese Translation
最先进的基于流的文本到图像（T2I）模型展现了卓越的生成能力，但仍然容易产生不安全的内容。以往的安全性努力包括概念抹除、提示过滤和基于分类器的门控。然而，像模型参数高效适配这样的简单技术很容易绕过这些安全措施。我们提出了一种独特的原则性方法，通过推理时的内省调节模型的注意力动态，从而实现安全性，展现出内在的鲁棒性。我们的方法在图像合成过程中分析并重新平衡注意力激活，避免生成不安全的概念，同时保持语义一致性。这种内省控制确保了部署模型的安全性。在标准和对抗性安全基准测试中，我们的方法在保持或甚至改善对齐和感知质量的同时，取得了显著的安全评分。我们的结果表明，注意力空间调节为基于扩散变换器的图像生成提供了一条比现有的概念抹除机制更有前景的安全路径。我们的代码可以在 https://basim-azam.github.io/iam/ 访问。
cs.CV / 63 / 2607.14946
DINE: Distance Is Not Enough -- Learning Global Deformation Priors for Robust Soft-Tissue Point Cloud Registration
DINE：距离不足——学习全局变形先验以实现稳健的软组织点云配准
Abstract
Non-rigid point cloud registration is central to soft-tissue shape analysis, but large deformations, noise, and outliers make correspondence estimation challenging. Most learning-based methods rely on local objectives such as Chamfer distance, which encourage point-wise proximity but do not constrain the global plausibility of the predicted deformation field. We address this limitation with DINE, a maximum a posteriori framework that augments distance-based registration with a learned statistical prior over displacement vector fields. DINE is applied to two registration backbones, Robust-DefReg and DefTransNet, using a two-stage strategy: a first-stage model is trained with Chamfer distance, its predicted deformation fields are used to estimate a prior, and the model is then refined with a combined distance and negative log-prior objective. We compare a full-field PCA Gaussian prior with a per-vector normalizing-flow prior. Experiments on DeformedTissue and SynBench show lower mean Chamfer distance under deformation and corruption. On DeformedTissue, DINE-PCA reduces Chamfer distance by approximately 27--69\% relative to the corresponding Stage-1 backbone across deformation levels, and improves robustness by up to 66\% for outliers and 83\% for Gaussian noise. On SynBench, improvements are modest at the smallest deformation levels and reach approximately 59--79\% from moderate to severe deformation. These results suggest that global deformation plausibility is an important constraint for reliable soft-tissue point cloud registration. (The code will be published soon.)
Chinese Translation
非刚性点云配准是软组织形状分析的核心，但大变形、噪声和异常值使得对应关系估计变得具有挑战性。大多数基于学习的方法依赖于局部目标，例如Chamfer距离，这些方法鼓励点与点之间的接近性，但并未约束预测变形场的全局合理性。我们通过DINE来解决这一局限性，DINE是一个最大后验框架，它通过对位移向量场的学习统计先验来增强基于距离的配准。DINE应用于两个配准骨干网络，Robust-DefReg和DefTransNet，采用两阶段策略：第一阶段模型使用Chamfer距离进行训练，其预测的变形场用于估计先验，然后模型通过结合距离和负对数先验目标进行优化。我们比较了全场PCA高斯先验与逐向量归一化流先验。在DeformedTissue和SynBench上的实验显示，在变形和损坏下，平均Chamfer距离降低。在DeformedTissue上，DINE-PCA相对于相应的第一阶段骨干网络，在不同变形水平下减少Chamfer距离约27%至69%，并且对异常值的鲁棒性提高了最多66%，对高斯噪声提高了83%。在SynBench上，在最小变形水平下的改进较为温和，而在中等到严重变形时的改进达到约59%至79%。这些结果表明，全局变形合理性是可靠的软组织点云配准的重要约束。（代码将很快发布。）
cs.CV / 64 / 2607.14966
U-shaped Multi-granularity Learning for Vision-Language Models
用于视觉-语言模型的U形多粒度学习
Abstract
The prompt learning paradigm for vision-language models is effective yet faces a granularity dilemma: global prompts lack fine-grained semantic awareness, while local prompts ignore contextual associations, limiting cross-task generalization. This dilemma exists in dense prediction tasks. Inspired by U-Net, which unifies multi-level representations across granularities, we propose UPrompt, a U-shaped multi-granularity prompt learning framework for vision-language models. Similar to how U-Net integrates fine and coarse features through symmetric encoder-decoder pathways with cross-level connections, UPrompt constructs parallel multi-granularity representations in both visual and textual modalities, where coarse-to-fine cascaded enhancement propagates global context to refine local details, while fine-to-coarse hierarchical supervision ensures semantic consistency across scales. Extensive experiments on 17 benchmarks validate our effectiveness. UPrompt outperforms MAMET and VPKE by 4.1 and 7.3 rSum on MSCOCO, surpasses CoCoA-Mix by 5.09% in base-to-novel generalization, while maintaining competitive performance with minimal overhead (coarse-grained) and matching PSRC with 1/3 cost (medium-grained).
Chinese Translation
视觉-语言模型的提示学习范式有效，但面临粒度困境：全局提示缺乏细粒度的语义意识，而局部提示忽视上下文关联，限制了跨任务的泛化能力。这一困境在密集预测任务中尤为明显。受到U-Net的启发，U-Net统一了不同粒度的多层次表示，我们提出了UPrompt，一个用于视觉-语言模型的U形多粒度提示学习框架。UPrompt类似于U-Net通过具有跨层连接的对称编码-解码路径整合细粒度和粗粒度特征，在视觉和文本模态中构建并行的多粒度表示，其中粗到细的级联增强传播全局上下文以细化局部细节，而细到粗的层次监督确保了跨尺度的语义一致性。在17个基准上的广泛实验验证了我们的有效性。UPrompt在MSCOCO上比MAMET和VPKE分别提高了4.1和7.3的rSum，在基础到新颖的泛化中超越了CoCoA-Mix 5.09%，同时以最小的开销（粗粒度）保持竞争力，并以1/3的成本与PSRC相匹配（中粒度）。
cs.CV / 65 / 2607.14968
Stitch-Inferencer: Enhance Endoscopic Video Segmentation and Tracking via Panoramic Reconstruction
Stitch-Inferencer：通过全景重建增强内窥镜视频分割与追踪
Abstract
Surgical video understanding is fundamental to navigation systems. Endoscopic perception is often hindered by a limited field-of-view and frequent instrument occlusions, making spatio-temporal context essential for robust inference. These challenges have motivated video models that aggregate information across frames. However, existing video models typically store past observations implicitly in learned feature representations, often requiring task-specific video training, substantial annotated data, and increased computational cost. We propose Stitch-Inferencer, a real-time, model-agnostic inference framework that replaces implicit feature memory with an explicit image-space panoramic canvas. By stitching valid observations across frames, Stitch-Inferencer preserves previously observed pixels in an online, instrument-free view, expanding the effective field-of-view and providing direct access to regions that are temporarily occluded or absent from the current frame. Downstream segmentation or tracking models are applied to a compact region of interest on the panorama, and their predictions are reprojected to the current frame, enabling existing models to exploit long-range context without retraining. Experiments on anatomy segmentation and point/box tracking demonstrate consistent improvements across diverse baselines while preserving real-time throughput. The stitching module alone runs at over 60 FPS, providing a practical inference-time solution to enhance endoscopic perception in computationally constrained intraoperative environments. Source code will be made publicly available.
Chinese Translation
外科视频理解是导航系统的基础。内窥镜感知常常受到有限视野和频繁仪器遮挡的影响，因此时空上下文对于稳健推理至关重要。这些挑战促使了跨帧聚合信息的视频模型的出现。然而，现有的视频模型通常在学习的特征表示中隐式存储过去的观察，往往需要特定任务的视频训练、大量标注数据以及增加的计算成本。我们提出了Stitch-Inferencer，一个实时的、模型无关的推理框架，它用显式的图像空间全景画布替代隐式特征记忆。通过在帧之间拼接有效观察，Stitch-Inferencer在在线、无仪器的视图中保留先前观察到的像素，扩展了有效视野，并提供对暂时被遮挡或当前帧中缺失区域的直接访问。下游的分割或追踪模型应用于全景的紧凑兴趣区域，并将其预测重新投影到当前帧，使现有模型能够在不重新训练的情况下利用长距离上下文。在解剖分割和点/框追踪的实验中，展示了在多种基准测试中一致的改进，同时保持实时吞吐量。拼接模块本身的运行速度超过60帧每秒，为在计算受限的手术环境中增强内窥镜感知提供了实用的推理时间解决方案。源代码将公开发布。
cs.CV / 66 / 2607.14974
On Success and Simplicity: A Second Look at Transferable Vision-Language Attack Pipeline
成功与简约：可转移视觉-语言攻击流程的再审视
Abstract
Vision-Language Pre-training Models (VLPMs) are known to be vulnerable to adversarial attacks. Recent transferable attacks on VLPMs have followed a common pipeline with complicated loss functions or multi-stage text/image attacks. However, in this paper, we demonstrate that such a sophisticated attack pipeline can be simpler yet more successful. Specifically, we identify three previously overlooked issues caused by inappropriate cross-modal interactions and excessive operations. To address them, we propose the Simple Vision-Language Attack (SimVLA) pipeline, which observably improves transferability and efficiency. Experiments on four datasets and three downstream tasks validate the superiority of our pipeline. For instance, on Flickr30k text-image retrieval dataset, our SimVLA outperforms the SOTA baseline in R@1 transferability by 8.01\%-14.71\%, while consuming only about 35.73\% of the time and 46.26\% of the max VRAM. Overall, the superiority of our SimVLA highlights the importance of leveraging domain knowledge (e.g., our proposed cross-modal word identification), while blindly pursuing intricate operations (e.g, complex loss functions and redundant multi-stage designs) may even be harmful. We hope our SimVLA can serve as a simple yet effective backbone for future extensions. Code is available at https://github.com/RYC-98/SimVLA.
Chinese Translation
视觉-语言预训练模型（VLPMs）已知对对抗攻击存在脆弱性。近期针对VLPMs的可转移攻击遵循了一个常见的流程，采用复杂的损失函数或多阶段的文本/图像攻击。然而，在本文中，我们证明了这样一个复杂的攻击流程可以更简单且更成功。具体而言，我们识别出三个由于不当的跨模态交互和过多操作而导致的先前被忽视的问题。为了解决这些问题，我们提出了简单视觉-语言攻击（Simple Vision-Language Attack，SimVLA）流程，该流程显著提高了可转移性和效率。在四个数据集和三个下游任务上的实验验证了我们流程的优越性。例如，在Flickr30k文本-图像检索数据集上，我们的SimVLA在R@1可转移性上超越了最先进基线8.01%-14.71%，同时仅消耗约35.73%的时间和46.26%的最大显存。总体而言，我们的SimVLA的优越性凸显了利用领域知识（例如，我们提出的跨模态词识别）的重要性，而盲目追求复杂操作（例如，复杂的损失函数和冗余的多阶段设计）可能甚至是有害的。我们希望我们的SimVLA能够作为未来扩展的简单而有效的基础。代码可在https://github.com/RYC-98/SimVLA获取。
cs.CV / 67 / 2607.14976
From Draft to Draft-Free: One-Step Video Object Removal via Privileged Distillation and Fast Planting
从草稿到无草稿：通过特权蒸馏和快速植入实现一步视频物体移除
Abstract
Video object removal is a fundamental yet challenging task in video editing. Despite recent progress, existing methods typically fall into two categories. Traditional approaches based on optical flow or attention mechanisms often introduce noticeable artifacts and yield unnatural results. In contrast, diffusion-based methods improve visual realism but demand multiple denoising steps, limiting their practicality. To address these issues, we propose From-Draft-to-Draft-Free (D2DF), a framework that distills the ability of transforming coarse drafts into refined videos into a one-step video generation model. Within D2DF, a teacher model is trained to refine low-quality removal results ("drafts") into high-fidelity videos by multiple steps. Then, through Prior-Privileged Consistency Distillation (PPCD), we distill this capability into a student model that performs one-step removal conditioned on the draft. To eliminate draft dependency, we introduce a Self-Guided Fast Planting (SGFP) module based on our Temporal Masked Transformer that autonomously generates scene-consistent pseudo-drafts in latent space, enabling a fully draft-free one-step model. Extensive experiments show that both draft-conditioned and draft-free versions achieve state-of-the-art performance on multiple metrics, surpassing traditional and multi-step generative methods in both quality and efficiency. The denoising process for a single video takes only about 1 second.
Chinese Translation
视频物体移除是视频编辑中的一项基础而又具有挑战性的任务。尽管近期取得了一些进展，现有方法通常分为两类。基于光流或注意力机制的传统方法往往会引入明显的伪影，导致不自然的结果。相对而言，基于扩散的方法提高了视觉真实感，但需要多个去噪步骤，限制了其实用性。为了解决这些问题，我们提出了从草稿到无草稿（D2DF）框架，该框架将将粗略草稿转化为精细视频的能力蒸馏为一个一步视频生成模型。在D2DF中，教师模型被训练以通过多个步骤将低质量的移除结果（“草稿”）精炼为高保真视频。然后，通过优先特权一致性蒸馏（PPCD），我们将这一能力蒸馏到一个学生模型中，使其能够在草稿的条件下执行一步移除。为了消除对草稿的依赖，我们引入了基于时间掩蔽变换器的自我引导快速植入（SGFP）模块，该模块在潜在空间中自主生成场景一致的伪草稿，从而实现完全无草稿的一步模型。大量实验表明，无论是基于草稿的版本还是无草稿版本，在多个指标上均实现了最先进的性能，超越了传统和多步骤生成方法的质量和效率。单个视频的去噪过程仅需约1秒。
cs.CV / 68 / 2607.14990
JADE-GS: Joint Alternating Deblurring Guided by Events in 3D Gaussian Splatting
JADE-GS：基于事件的联合交替去模糊在3D高斯点云中的应用
Abstract
When a camera moves fast during exposure, blur destroys the intra-exposure motion a 3D model needs to recover the sharp scene, while event cameras capture exactly this signal at microsecond resolution. Turning them into reliable 3D supervision faces two obstacles. First, the two restoration priors fail in opposite ways: physics-based event-integration priors preserve edges but accumulate drift; learned networks recover texture but distort boundaries. Second, existing pipelines run in one direction only, so raw event noise or the biases of fixed 2D pseudo-labels pass uncorrected into the geometry. JADE-GS addresses both: a pixel-adaptive routing gate fuses the complementary priors, and the resulting 2D restorer is coupled to a 3D Gaussian Splatting student in a bidirectional loop, where detached, multi-view-consistent renders and a physics-based reblurring constraint regularize the restorer, turning a fixed preprocessor into a geometry-aware predictor. Across synthetic and real benchmarks, JADE-GS attains the best perceptual quality, leading LPIPS and CLIP-IQA on both benchmarks with competitive PSNR and SSIM, and trainsin about one hour under 5 GB on a single consumer GPU while preserving real-time rendering.
Chinese Translation
当相机在曝光过程中快速移动时，模糊会破坏3D模型恢复清晰场景所需的曝光内运动，而事件相机则以微秒级分辨率准确捕捉到这一信号。将其转化为可靠的3D监督面临两个障碍。首先，两种恢复先验以相反的方式失败：基于物理的事件积分先验保留边缘但会积累漂移；学习网络恢复纹理但会扭曲边界。其次，现有的处理流程仅单向运行，因此原始事件噪声或固定2D伪标签的偏差未被校正地传递到几何结构中。JADE-GS解决了这两个问题：像素自适应路由门融合了互补的先验，生成的2D恢复器与3D高斯点云学生在双向循环中耦合，其中分离的、多视角一致的渲染和基于物理的去模糊约束对恢复器进行正则化，将固定的预处理器转变为几何感知的预测器。在合成和真实基准测试中，JADE-GS达到了最佳的感知质量，在两个基准测试中领先于LPIPS和CLIP-IQA，并且在单个消费级GPU上以约5GB的内存训练约一个小时，同时保持实时渲染。
cs.CV / 69 / 2607.15038
Video = World + Event Stream
视频 = 世界 + 事件流
Huang, Lianghua, Wu, Zhi-Fan, Shi, Yupeng, Wang, Wei, Feng, Mengyang, Yu, Cheng, Liang, Chen, He, Junjie, Xie, Chen-Wei, Liu, Yu, Zhou, Jingren, Wang, Ang, Zhang, Bang, Ai, Baole, Zhong, Chongyang, Qi, Jinwei, Zhu, Kai, Li, Pandeng, Zhang, Peng, Zhang, Wenyuan, Cheng, Xinhua, Huang, Yitong, Zheng, Yun, Bao, Yuxiang, Wang, Yuzheng, Lin, Zhiwei, Bi, Zoubin
Abstract
We present Wan-Streamer v0.3, which reframes our native-streaming interaction model under a single organizing view: a video is a world plus an event stream. The world is the persistent context in which a video unfolds, including the environment, scene, subjects, ambient acoustic conditions, voice characteristics, and other relatively stable conditions. The event stream is everything that changes over time within that world, including scene or environmental changes, subject behavior, speech, and other sounds. This yields a general-purpose pretraining task over large amounts of real video: given a world and incoming input, predict how the world moves, changes, and responds in real time. The resulting competence can be specialized to a broad family of real-time downstream tasks. We instantiate it on real-time full-duplex audio-visual interaction, where the event stream is the agent's speech together with free-form behavior. Functionally, the model's multimodal understanding process is vision-language-action-like: it maps multimodal user input to language-form speech and behavior actions. Wan-Streamer v0.3 preserves the v0.2 operating point: 640x368 video at 25 FPS, a 160 ms streaming unit, approximately 200 ms model-side response latency, and approximately 550 ms total interaction latency under a 350 ms bidirectional network budget.
Chinese Translation
我们提出了 Wan-Streamer v0.3，它在一个统一的组织视角下重新构建了我们的原生流媒体交互模型：视频是一个世界加上一个事件流。世界是视频展开的持久上下文，包括环境、场景、主体、环境声学条件、语音特征以及其他相对稳定的条件。事件流是该世界中随时间变化的所有事物，包括场景或环境变化、主体行为、言语和其他声音。这为大量真实视频提供了一种通用的预训练任务：给定一个世界和输入，预测世界如何实时移动、变化和响应。由此产生的能力可以专门化为广泛的实时下游任务。我们在实时全双工音视频交互中实现了这一点，其中事件流是代理的言语以及自由形式的行为。从功能上讲，该模型的多模态理解过程类似于视觉-语言-行动：它将多模态用户输入映射为语言形式的言语和行为动作。Wan-Streamer v0.3 保持了 v0.2 的操作点：640x368 视频，25 FPS，160 毫秒流媒体单元，约 200 毫秒模型侧响应延迟，以及在 350 毫秒双向网络预算下约 550 毫秒的总交互延迟。
cs.CV / 70 / 2607.15041
Weakly-Supervised RGB-D Salient Object Detection via SAM-driven Pseudo Annotation and State Space Interaction-based Diffusion
基于SAM驱动的伪标注和状态空间交互扩散的弱监督RGB-D显著目标检测
Abstract
Weakly-supervised RGB-D Salient Object Detection (SOD) is explored to reduce the heavy burden of pixel-level annotations. But scribble annotations lack the structure and details of objects, resulting in inaccurate saliency maps. In this paper, we propose a novel scribble-supervised RGB-D SOD method, consisting of a Segment Anything Model (SAM)-driven pseudo annotation generation method (\emph{SAM-PAG}) and a state space interaction-based conditional diffusion model (\emph{$S^2$Diff}). Specifically, SAM-PAG is tailored to address the issue of sparse supervision information. In SAM-PAG, we adopt the advanced SAM to expand sparse scribbles to dense pixel-level pseudo annotations through the dual-branch structure and the consistency of segmentation masks. In $S^2$Diff, we adopt the diffusion model to iteratively refine the noisy saliency maps with the guidance of conditional information, generating accurate saliency maps. Naturally, the core of our $S^2$Diff lies in the acquisition of conditional features and the denoising of saliency maps. For the former, we employ a cross-modal conditional generation module to interweave cross-modal features through frequency integration and implicit-explicit state space interaction, effectively achieving global conditional features. For the latter, we employ a context injection module to mitigate noise interference and to enhance object information with the conditional context. With the close cooperation of SAM-PAG and $S^2$Diff, our method outperforms relevant scribble-supervised methods and achieves competitive performance compared to fully-supervised methods on seven datasets. The code and results of our method are available at https://github.com/Switch457/WeakS2Diff_SOD.
Chinese Translation
弱监督RGB-D显著目标检测（SOD）旨在减少像素级标注的沉重负担。然而，涂鸦标注缺乏对象的结构和细节，导致显著性图的不准确。本文提出了一种新颖的涂鸦监督RGB-D SOD方法，该方法由一个基于Segment Anything Model（SAM）的伪标注生成方法（SAM-PAG）和一个基于状态空间交互的条件扩散模型（$S^2$Diff）组成。具体而言，SAM-PAG旨在解决稀疏监督信息的问题。在SAM-PAG中，我们采用先进的SAM通过双分支结构和分割掩码的一致性将稀疏涂鸦扩展为密集的像素级伪标注。在$S^2$Diff中，我们采用扩散模型在条件信息的指导下迭代精炼噪声显著性图，从而生成准确的显著性图。自然地，我们的$S^2$Diff的核心在于条件特征的获取和显著性图的去噪。对于前者，我们采用跨模态条件生成模块，通过频率融合和隐式-显式状态空间交互交织跨模态特征，有效地实现全局条件特征。对于后者，我们采用上下文注入模块来减轻噪声干扰，并通过条件上下文增强对象信息。在SAM-PAG和$S^2$Diff的紧密合作下，我们的方法超越了相关的涂鸦监督方法，并在七个数据集上与全监督方法相比取得了竞争力的表现。我们的方法的代码和结果可在https://github.com/Switch457/WeakS2Diff_SOD获取。
cs.CV / 71 / 2607.15047
Parameter-efficient Prompt Tuning of Vision Foundation Model With Adaptive Focal Loss for Interpretable MCI Screening
基于自适应焦点损失的视觉基础模型的参数高效提示调优用于可解释的轻度认知障碍筛查
Abstract
Mild Cognitive Impairment is a critical early stage of cognitive decline that frequently precedes Alzheimer's disease, yet its automated detection from neuropsychological drawing tests remains fundamentally constrained by data scarcity, class imbalance, and diagnostic ambiguity near clinical boundaries. Existing methodologies attempt to bypass these constraints using computationally expensive, fully fine-tuned hybrid architectures that relegate spatial explainability to a post-hoc approximation rather than an intrinsic model property. We propose a parameter-efficient framework utilizing frozen DINOv2-Small model adapted via three modality-specific learnable prompt tokens while Operating with 1.19 million trainable parameters, each token serves as a query in a shared cross-attention layer over the source image patch tokens. Crucially, spatial explainability is achieved directly through these attention maps; as a structural consequence of the architecture. Then task-conditioned embeddings fused via an attention module to quantify modality-level importance per subject. To handle boundary ambiguity, a MoCA-adapted focal loss introduced that integrates continuous cognitive scores into the training target, loss modulation, and adaptive sample weighting, strictly generalizing standard soft-label approaches. Under stratified five-fold cross-validation, the proposed architecture yields an MCI-class F1 of 0.641 and an AUC of 0.795, outperforming the computationally heavier ResViT baseline by 0.110 in MCI-class F1.
Chinese Translation
轻度认知障碍是认知衰退的关键早期阶段，常常在阿尔茨海默病之前出现，但从神经心理绘画测试中自动检测其存在仍然受到数据稀缺、类别不平衡和临床边界附近诊断模糊性的根本限制。现有的方法试图通过使用计算成本高昂的完全微调混合架构来绕过这些限制，这些架构将空间可解释性 relegated 为事后近似，而不是内在模型属性。我们提出了一种参数高效的框架，利用冻结的 DINOv2-Small 模型，通过三个特定模态的可学习提示标记进行适配，同时操作 119 万个可训练参数，每个标记作为共享跨注意力层中源图像补丁标记的查询。至关重要的是，空间可解释性通过这些注意力图直接实现；作为架构的结构性结果。然后，通过注意力模块融合任务条件嵌入，以量化每个受试者的模态级重要性。为了处理边界模糊性，引入了一种 MoCA 适配的焦点损失，将连续的认知评分整合到训练目标、损失调制和自适应样本加权中，严格推广标准的软标签方法。在分层五折交叉验证下，所提出的架构在 MCI 类别的 F1 值为 0.641，AUC 为 0.795，MCI 类别 F1 超过计算成本更高的 ResViT 基线 0.110。
cs.CV / 72 / 2607.15048
RoGS: Adaptive Meshgrid Gaussian for Large-Scale Road Surface Mapping
RoGS：用于大规模道路表面映射的自适应网格高斯模型
Abstract
Road surface mapping plays a crucial role in autonomous driving, supporting high-definition map generation, lane-level perception, and automatic road annotation. Recent mesh-based road surface reconstruction methods have shown promising results, but they still suffer from limited reconstruction quality and high optimization cost, especially in large-scale driving scenarios. To address these limitations, we propose ROADGS-T, a robust and efficient large-scale road surface mapping framework based on adaptive meshgrid Gaussian representation. Specifically, we model the road surface by placing 2D Gaussian surfels on a meshgrid, where each surfel explicitly stores color, semantic, and geometric information. Compared with conventional mesh-based representations and 3D Gaussian primitives, the proposed meshgrid Gaussian representation better matches the thin-surface property of roads while significantly reducing redundant primitives and overlap during optimization. To further improve representation efficiency and structural fidelity, we introduce a road-structure-aware adaptive meshgrid strategy, which allocates denser Gaussian surfels to geometrically or semantically complex regions, such as lane markings, road boundaries, and height discontinuities, while maintaining a compact representation in flat road areas. Moreover, instead of relying on a single nearest vehicle pose, we design a trajectory-consistency-guided pose-robust refinement strategy, which estimates local surface priors from multiple neighboring poses and adaptively weights pose-guided height regularization according to their geometric consistency.
Chinese Translation
道路表面映射在自动驾驶中起着至关重要的作用，支持高清地图生成、车道级感知和自动道路标注。最近的基于网格的道路表面重建方法显示出良好的结果，但在大规模驾驶场景中，它们仍然面临重建质量有限和优化成本高的问题。为了解决这些局限性，我们提出了ROADGS-T，一个基于自适应网格高斯表示的稳健高效的大规模道路表面映射框架。具体而言，我们通过在网格上放置二维高斯表面元素（surfel）来建模道路表面，其中每个表面元素明确存储颜色、语义和几何信息。与传统的基于网格的表示和三维高斯原语相比，所提出的网格高斯表示更好地匹配了道路的薄表面特性，同时在优化过程中显著减少了冗余原语和重叠。为了进一步提高表示效率和结构保真度，我们引入了一种道路结构感知的自适应网格策略，该策略将更密集的高斯表面元素分配给几何或语义复杂的区域，如车道标记、道路边界和高度不连续性，同时在平坦道路区域保持紧凑的表示。此外，我们设计了一种轨迹一致性引导的姿态鲁棒性优化策略，而不是依赖单一的最近车辆姿态，该策略从多个邻近姿态中估计局部表面先验，并根据几何一致性自适应地加权姿态引导的高度正则化。
cs.CV / 73 / 2607.15054
Beyond Single Expert: Harmonizing Diverse Visual Priors in MLLMs for Spatial Understanding
超越单一专家：在多模态大语言模型中协调多样化视觉先验以实现空间理解
Abstract
Multimodal Large Language Models (MLLMs) have demonstrated substantial promise in spatial understanding. Existing works typically incorporate prior knowledge extracted from a pre-trained foundation model to further enhance the spatial awareness of MLLMs. In this paper, we first reveal that when integrating diverse foundation models into MLLMs, different models provide complementary spatial priors that benefit different tasks. Motivated by this, we propose $\textbf{ViPS}$, a novel multi-model prior framework designed to fully unleash the potential of incorporating multiple $\textbf{Vi}$sual $\textbf{P}$riors from diverse models into MLLMs for $\textbf{S}$patial understanding. Specifically, ViPS introduces an Efficient Prior Proxy to generate multiple foundational priors with minimal inference overhead, and a Dynamic Prior Fusion mechanism to achieve harmonious and context-aware prior fusion and injection from the prior proxies. Extensive experiments demonstrate that ViPS successfully harmonizes diverse visual priors, establishing new state-of-the-art performance across multiple complex spatial reasoning and 3D spatial understanding benchmarks. Project page: https://visual-ai.github.io/vips
Chinese Translation
多模态大语言模型（MLLMs）在空间理解方面展现出了巨大的潜力。现有研究通常利用从预训练基础模型中提取的先验知识，以进一步增强MLLMs的空间意识。本文首先揭示了在将多样化基础模型整合到MLLMs中时，不同模型提供互补的空间先验，这对不同任务具有积极的影响。基于此，我们提出了$ extbf{ViPS}$，一种新颖的多模型先验框架，旨在充分释放将来自不同模型的多个$ extbf{Vi}$sual $ extbf{P}$riors整合到MLLMs中以实现$ extbf{S}$patial理解的潜力。具体而言，ViPS引入了一种高效的先验代理（Efficient Prior Proxy），以最小的推理开销生成多个基础先验，并采用动态先验融合机制（Dynamic Prior Fusion）实现和谐且具有上下文感知的先验融合和注入。大量实验表明，ViPS成功协调了多样化的视觉先验，在多个复杂空间推理和3D空间理解基准测试中建立了新的最先进性能。项目页面：https://visual-ai.github.io/vips
cs.CV / 74 / 2607.15058
SUFLECA: Scaling Up Feature Learning for CAD-to-image Alignment
SUFLECA：扩展特征学习以实现CAD与图像的对齐
Abstract
CAD-to-image alignment aims to estimate an object's 9D pose (rotation, translation, and anisotropic scale) from a single RGB image, enabling applications in robotics and augmented reality. Recent zero-shot methods use visual foundation models to match image regions to CAD models, yet typically their correspondences are appearance-driven and degrade under occlusion or sim-to-real domain shift. To address these limitations, we introduce SUFLECA (Scaling Up Feature LEarning for CAD Alignment), a weakly-supervised framework for zero-shot CAD alignment with two key contributions. First, SUFLECA scales up geometry-grounded feature learning from pretrained visual representations through Normalized Object Coordinates (NOCs) supervision on 674K images spanning 12 real and synthetic datasets, learning compact geometry-aware features that generalize across domains. Second, we propose a geometrically consistent matching algorithm that establishes reliable one-to-one CAD-to-image correspondences. Together, these contributions enable accurate, sub-second alignment per object instance without iterative pose refinement. On ScanNet25k, SUFLECA achieves 33.4%/42.3% category/instance accuracy, outperforming, with a smaller computational footprint, the strongest zero-shot baseline by 10.3/12.2 percentage points and, for the first time on this benchmark, even surpassing fully supervised methods. Code is available at: https://github.com/snt-arg/SUFLECA
Chinese Translation
CAD与图像的对齐旨在从单幅RGB图像中估计物体的9D姿态（旋转、平移和各向异性缩放），以支持机器人技术和增强现实等应用。近期的零样本方法利用视觉基础模型将图像区域与CAD模型进行匹配，但通常其对应关系是基于外观的，在遮挡或仿真到真实的领域转移下会出现退化。为了解决这些局限性，我们提出了SUFLECA（Scaling Up Feature LEarning for CAD Alignment），这是一个弱监督框架，用于零样本CAD对齐，具有两个关键贡献。首先，SUFLECA通过对674K图像进行标准化物体坐标（NOCs）监督，扩展了基于几何的特征学习，涵盖了12个真实和合成数据集，学习到跨领域通用的紧凑几何感知特征。其次，我们提出了一种几何一致的匹配算法，建立可靠的一对一CAD与图像的对应关系。这些贡献共同实现了每个物体实例的准确对齐，且无需迭代姿态优化，时间在一秒以内。在ScanNet25k数据集上，SUFLECA达到了33.4%/42.3%的类别/实例准确率，以更小的计算开销超越了最强的零样本基线10.3/12.2个百分点，并且在该基准上首次超过了完全监督的方法。代码可在以下链接获取：https://github.com/snt-arg/SUFLECA
cs.CV / 75 / 2607.15082
Towards Hierarchical Structure Understanding of Newspaper Images
朝向报纸图像的层次结构理解
Abstract
Understanding newspaper images remains a challenging task due to their complex, nested hierarchical structures and dense, heterogeneous layouts. In this paper, we explore two complementary approaches for newspaper structure understanding. First, we present a modular bottom-up pipeline that combines state-of-the-art open-source models: YOLO for layout detection, LayoutReader for reading order prediction, and a custom algorithm for article segmentation. This approach leverages existing robust components while maintaining flexibility and interpretability. Second, we introduce Tiramisu (Tiered Transformers for Hierarchical Structure Understanding), a novel end-to-end transformer-based architecture that explicitly models document hierarchy through an iterative tiered process. Tiramisu performs section and article separation, block localization, semantic categorization, and reading order prediction using highly parallelized attention mechanisms. Finally, we release Finlam La Libert\'e, a new dataset designed specifically for evaluating hierarchical information retrieval in historical newspapers. Experimental results demonstrate the effectiveness of both approaches in reconstructing complex newspaper hierarchies, with comparative analysis highlighting their respective strengths for scalable document digitization. The Tiramisu training code, including the synthetic newspaper generator, is available at https://git.litislab.fr/tiramisu/tiramisu-newspaper-articles-extractor.
Chinese Translation
理解报纸图像仍然是一项具有挑战性的任务，因为它们具有复杂的嵌套层次结构和密集的异构布局。本文探讨了两种互补的方法来理解报纸结构。首先，我们提出了一种模块化的自下而上的流程，该流程结合了最先进的开源模型：YOLO用于布局检测，LayoutReader用于阅读顺序预测，以及一种自定义算法用于文章分割。这种方法利用了现有的强大组件，同时保持灵活性和可解释性。其次，我们介绍了Tiramisu（分层变换器用于层次结构理解），这是一种新颖的端到端基于变换器的架构，通过迭代分层过程显式建模文档层次结构。Tiramisu使用高度并行化的注意机制执行章节和文章分离、块定位、语义分类和阅读顺序预测。最后，我们发布了Finlam La Liberté，这是一个专门设计用于评估历史报纸中层次信息检索的新数据集。实验结果证明了这两种方法在重建复杂报纸层次结构方面的有效性，比较分析突出了它们在可扩展文档数字化方面的各自优势。Tiramisu训练代码，包括合成报纸生成器，可在https://git.litislab.fr/tiramisu/tiramisu-newspaper-articles-extractor获取。
cs.CV / 76 / 2607.15084
Quantifying Training Membership Information in the Hyperspherical Embedding Geometry of Face Recognition Models
量化人脸识别模型超球嵌入几何中的训练成员信息
Abstract
Face recognition models represent each face as an embedding vector on the unit hypersphere by clustering embeddings of the same identity while pushing different identities apart through angular-margin losses. Because these losses act only on training identities, non-member identities may form clusters with different geometric properties. In this paper, we quantify the magnitude of this difference and what training-time factors control it. We compute four statistics based on cluster geometry across 180 face recognition models in a factorial design over IResNet backbone size, loss head, training duration, and the number of training identities, and evaluate each configuration on nine benchmarks. Our results indicate that the number of training identities has the largest effect on member/non-member separability, while backbone and loss head contribute far less, and that, on a same-domain held-out reference, the geometric membership signal decreases monotonically as more identities are added to training. We provide an analysis of cross-domain (pose, age, quality, ethnicity) non-member benchmarks and report that these inflate the apparent membership signal. Finally, we fuse all four statistics with a learned classifier to reveal additional membership information beyond the best individual statistic.
Chinese Translation
人脸识别模型通过聚类相同身份的嵌入向量并通过角度边际损失将不同身份的嵌入推离，来将每个面孔表示为单位超球面上的嵌入向量。由于这些损失仅作用于训练身份，非成员身份可能会形成具有不同几何特性的聚类。本文量化了这种差异的大小以及控制这种差异的训练时因素。我们基于180个人脸识别模型的聚类几何计算了四个统计量，这些模型在IResNet主干大小、损失头、训练持续时间和训练身份数量的因子设计下进行评估，并在九个基准上评估每个配置。我们的结果表明，训练身份的数量对成员/非成员的可分离性影响最大，而主干和损失头的贡献相对较小，并且在同域的保留参考上，几何成员信号随着训练中添加更多身份而单调减少。我们对跨域（姿态、年龄、质量、种族）非成员基准进行了分析，并报告这些基准会夸大明显的成员信号。最后，我们将所有四个统计量与学习的分类器融合，以揭示超出最佳单一统计量的额外成员信息。
cs.CV / 77 / 2607.15094
AlphaWiSE: Adaptive Weight Interpolation for Continual Multimodal Representation Learning
AlphaWiSE：用于持续多模态表示学习的自适应权重插值
Abstract
Multimodal models such as CLIP learn a shared embedding space for cross-modal retrieval, but continual adaptation to sequentially arriving data can disrupt the cross-modal alignment acquired from earlier phases. Conventional continual-learning methods return a single checkpoint, which commits every retrieval direction to the same stability-plasticity trade-off. We propose AlphaWiSE, a post-hoc weight-space interpolation method that composes two frozen source checkpoints. For each aligned parameter tensor identified by its checkpoint key, AlphaWiSE fits one scalar interpolation coefficient shared by all tensor entries. The coefficients are fitted on a smaller exemplar memory and used to materialize one interpolated checkpoint. The deployed model has the same architecture and parameter count as either source checkpoint, which does not require additional inference time. Extensive experiments on audio-image-text retrieval show consistent improvements over strong continual-learning baselines across multiple retrieval directions and evaluation metrics.
Chinese Translation
多模态模型如CLIP学习了一个共享的嵌入空间用于跨模态检索，但对连续到达的数据进行持续适应可能会破坏早期阶段获得的跨模态对齐。传统的持续学习方法返回一个单一的检查点，这使得每个检索方向都承受相同的稳定性-可塑性权衡。我们提出了AlphaWiSE，一种事后权重空间插值方法，它组合了两个冻结的源检查点。对于每个通过其检查点键识别的对齐参数张量，AlphaWiSE拟合一个所有张量条目共享的标量插值系数。这些系数是在一个较小的示例记忆上拟合的，并用于生成一个插值检查点。部署的模型与任一源检查点具有相同的架构和参数数量，因此不需要额外的推理时间。在音频-图像-文本检索上的大量实验显示，在多个检索方向和评估指标上，相较于强大的持续学习基线，取得了一致的改进。
cs.CV / 78 / 2607.15097
QuReC: All-in-One Image Restoration with Query-Specific Guidance and Local-Global Response Calibration
QuReC：具有查询特定指导和局部-全局响应校准的全能图像恢复
Abstract
All-in-one image restoration aims to recover clean images degraded by multiple corruption types using a single unified model. Existing methods typically rely on image-level prompts or shared guidance to handle diverse degradations. However, such a paradigm becomes inadequate when degradations are spatially heterogeneous or even coexist in mixed forms within a single image. Yet spatially adaptive guidance alone is not sufficient, since accurate restoration also requires each spatial query to reliably aggregate complementary information from local neighborhoods and global contexts. To this end, we propose QuReC, a unified framework for all-in-one image restoration. QuReC consists of a Degradation-Guided Query Reconstruction Module (DQRM) and a Local-Global Response Calibration Module (LGRCM). Specifically, DQRM matches each spatial query against a degradation prototype space to reconstruct a query-specific degradation-aware representation, thereby providing fine-grained spatially adaptive restoration guidance. To further stabilize this query-wise matching process, we introduce a weakly supervised prototype matching learning strategy to improve optimization stability and degradation semantic consistency. Meanwhile, LGRCM performs local-global dual-branch aggregation and calibrates the aggregated responses with learnable priors, improving the reliability of feature aggregation and the coordination between local detail modeling and global context modeling. Extensive experiments demonstrate that QuReC achieves superior performance on multiple all-in-one image restoration benchmarks. The code is released at https://github.com/zhoushen1/QuReC.
Chinese Translation
全能图像恢复旨在通过单一统一模型恢复因多种损坏类型而退化的干净图像。现有方法通常依赖于图像级提示或共享指导来处理多样的退化。然而，当退化在空间上具有异质性或在单一图像中以混合形式共存时，这种范式变得不够充分。然而，仅依赖空间自适应指导也不够，因为准确的恢复还需要每个空间查询可靠地聚合来自局部邻域和全局上下文的互补信息。为此，我们提出了QuReC，一个统一的全能图像恢复框架。QuReC由一个退化引导查询重建模块（DQRM）和一个局部-全局响应校准模块（LGRCM）组成。具体而言，DQRM将每个空间查询与退化原型空间进行匹配，以重建查询特定的退化感知表示，从而提供细粒度的空间自适应恢复指导。为了进一步稳定这一查询匹配过程，我们引入了一种弱监督原型匹配学习策略，以提高优化稳定性和退化语义一致性。同时，LGRCM执行局部-全局双分支聚合，并使用可学习的先验对聚合响应进行校准，从而提高特征聚合的可靠性以及局部细节建模与全局上下文建模之间的协调。大量实验表明，QuReC在多个全能图像恢复基准测试中表现优异。代码已发布在 https://github.com/zhoushen1/QuReC。
cs.CV / 79 / 2607.15128
DAPGNet: Dynamic Adaptive Physics-Guided Graph Diffusion Network for Hyperspectral Image Classification
DAPGNet：动态自适应物理引导图扩散网络用于高光谱图像分类
Abstract
Hyperspectral image (HSI) classification requires reliable pixel-relation modeling under spectral variability, mixed pixels, and heterogeneous boundaries. Existing graph-based HSI classifiers usually construct graph topology from spatial proximity, superpixel connectivity, or learned feature affinity. However, the spectral physical prior carried by contiguous bands has limited influence on topology estimation and message propagation. This paper presents DAPGNet, a dynamic adaptive physics-guided graph diffusion network that injects a structure-constrained physical prior into relation-level graph learning. DAPGNet first encodes contiguous spectral responses into node-wise multiscale physical-prior representations. A two-stage graph constructor then combines spectral-spatial affinity, physical-prior consistency, and spatial distance to form a physical-prior-aware sparse topology. During graph diffusion, learned edge weights are transformed into additive attention biases, while a physical gate performs node-wise and feature-wise interpolation between graph-aggregated features and projected physical-prior features. Cross-scale fusion integrates node states from different diffusion depths, and the network is optimized with main classification, auxiliary supervision, and second-order spectral smoothness regularization. Experiments on Indian Pines, WHU-Hi-LongKou, Houston2013, and Houston2018 show that DAPGNet achieves the best OA, AA, and Kappa among representative CNN-, Transformer-, Mamba-, and graph-based baselines. It improves AA over the strongest competing method by 3.64 to 7.31 percentage points across the four datasets. Ablation and sensitivity analyses further support the complementary effects of physical-prior extraction, prior-aware topology construction, physics-gated propagation, and spectral smoothness regularization.
Chinese Translation
高光谱图像（HSI）分类需要在光谱变异、混合像素和异质边界下可靠的像素关系建模。现有的基于图的HSI分类器通常从空间邻近性、超像素连接性或学习的特征亲和性构建图拓扑。然而，连续波段所携带的光谱物理先验对拓扑估计和信息传播的影响有限。本文提出了DAPGNet，一种动态自适应物理引导图扩散网络，将结构约束的物理先验注入关系级别的图学习中。DAPGNet首先将连续的光谱响应编码为节点级多尺度物理先验表示。然后，二阶段图构造器结合光谱-空间亲和性、物理先验一致性和空间距离，形成一个关注物理先验的稀疏拓扑。在图扩散过程中，学习到的边权重被转化为加性注意力偏置，而物理门则在图聚合特征和投影的物理先验特征之间执行节点级和特征级的插值。跨尺度融合整合来自不同扩散深度的节点状态，网络通过主要分类、辅助监督和二阶光谱平滑正则化进行优化。在印度松树、WHU-Hi-LongKou、Houston2013和Houston2018的数据集上的实验表明，DAPGNet在代表性的CNN、Transformer、Mamba和基于图的基线中实现了最佳的整体精度（OA）、平均精度（AA）和Kappa值。在四个数据集中，DAPGNet在最强竞争方法的基础上提高了AA 3.64到7.31个百分点。消融和敏感性分析进一步支持物理先验提取、关注先验的拓扑构建、物理门控传播和光谱平滑正则化的互补效应。
cs.CV / 80 / 2607.15139
Ray-based phase error correction for miniaturized DOE projector-based FPP under single-directional hyperbolic projection
基于光线的相位误差校正用于单向双曲投影下的微型DOE投影仪FPP
Abstract
Fringe Projection Profilometry (FPP) systems using miniaturized DOE pro-jectors often suffer from severe phase artifacts due to nonlinear projection characteristics and limited pattern controllability. We propose a ray-based phase error correction framework that models phase artifacts along projection rays from the projector pinhole, incorporating projector geometry without re-lying on image-domain processing or neighboring pixels. A projector pinhole estimation method based on a single-directional hyperbolic fringe pattern is introduced, through which projector geometry can be recovered without stereo calibration. In addition, a data-efficient strategy constructs the re-finement model from a single calibration pose. Experiments on miniaturized DOE projector-based FPP systems demonstrate significant improvements in reconstruction accuracy under nonlinear projection conditions, confirming the robustness and physical consistency of the proposed approach.
Chinese Translation
使用微型DOE投影仪的条纹投影轮廓测量（FPP）系统常常由于非线性投影特性和有限的图案可控性而遭受严重的相位伪影。我们提出了一种基于光线的相位误差校正框架，该框架沿投影仪针孔的投影光线建模相位伪影，结合了投影仪几何形状，而无需依赖图像域处理或邻近像素。我们引入了一种基于单向双曲条纹图案的投影仪针孔估计方法，通过该方法可以在不进行立体校准的情况下恢复投影仪几何形状。此外，一种数据高效的策略从单一校准姿态构建精细化模型。在微型DOE投影仪基础上的FPP系统实验中，证明了在非线性投影条件下重建精度显著提高，确认了所提方法的鲁棒性和物理一致性。
cs.CV / 81 / 2607.15211
MAGiSt3R: Multi-Agent Feed-forward 3D Reconstruction from Monocular RGB Videos
MAGiSt3R：基于单目RGB视频的多智能体前馈三维重建
Abstract
This paper presents MAGiSt3R, a multi-agent 3D reconstruction framework performing reconstruction and camera tracking for monocular RGB videos at almost 10 FPS. MAGiSt3R relies on a feed-forward model from the 3R family to process RGB videos and regress local point maps, and on a merging model, MAGMA, that combines local maps at both intra-agent and inter-agent levels to obtain the final global point map. Furthermore, MAGiSt3R performs pose graph optimization to mitigate cumulative camera drift occurring along the feed-forward pipeline. We evaluate MAGiSt3R on both synthetic and real-world datasets, demonstrating its superior reconstruction and camera tracking accuracy compared to state-of-the-art approaches.
Chinese Translation
本文提出了MAGiSt3R，一个多智能体三维重建框架，能够以接近10帧每秒的速度对单目RGB视频进行重建和相机跟踪。MAGiSt3R依赖于3R系列的前馈模型来处理RGB视频并回归局部点图，同时使用合并模型MAGMA，在智能体内部和智能体之间结合局部地图，以获得最终的全局点图。此外，MAGiSt3R还执行姿态图优化，以减轻在前馈管道中发生的累积相机漂移。我们在合成和真实世界数据集上评估了MAGiSt3R，证明其在重建和相机跟踪精度上优于现有的最先进方法。
cs.CV / 82 / 2607.15216
Symbal: Detecting Systematic Misalignments in Model-Generated Captions
Symbal：检测模型生成的标题中的系统性错位
Abstract
Multimodal large language models (MLLMs) often introduce errors when generating image captions, resulting in misaligned image-text pairs. Our work focuses on a class of captioning errors that we refer to as systematic misalignments, where a recurring error in MLLM-generated captions is closely associated with the presence of a specific visual feature in the paired image. Given a vision-language dataset with MLLM-generated captions, our aim in this work is to detect such errors, a task we refer to as systematic misalignment detection. As our first key contribution, we present Symbal, which utilizes a structured, dual-stage setup with off-the-shelf foundation models to identify systematic misalignments and summarize results in natural language. As our second key contribution, we introduce SymbalBench, a benchmark designed to evaluate automated methods on our proposed task. SymbalBench consists of 1.7 million image-text pairs from two domains (natural and medical images), organized into 420 vision-language datasets with annotated systematic misalignments. Symbal exhibits strong performance on this benchmark, correctly identifying systematic misalignments in 63.8% of datasets, a nearly 4x improvement over the closest baseline. We supplement our evaluations on SymbalBench with real-world evaluations, showing that (1) Symbal can accurately surface systematic misalignments in captions generated by four MLLMs and (2) Symbal is a powerful tool for auditing off-the-shelf image-caption datasets. Ultimately, our novel task, method, and benchmark can aid users with auditing MLLM-generated captions and identifying critical errors, without requiring access to the underlying MLLM. Code is available at https://github.com/Stanford-AIMI/Symbal.
Chinese Translation
多模态大型语言模型（MLLMs）在生成图像标题时常常引入错误，导致图像-文本对不对齐。我们的工作聚焦于一种我们称之为系统性错位的标题错误，这种错误在MLLM生成的标题中反复出现，并与配对图像中特定视觉特征的存在密切相关。鉴于一个包含MLLM生成标题的视觉-语言数据集，我们的目标是检测此类错误，这一任务我们称之为系统性错位检测。作为我们的第一个关键贡献，我们提出了Symbal，它利用结构化的双阶段设置与现成的基础模型来识别系统性错位，并以自然语言总结结果。作为我们的第二个关键贡献，我们引入了SymbalBench，这是一个旨在评估我们提出的任务的自动化方法的基准。SymbalBench包含来自两个领域（自然图像和医学图像）的170万对图像-文本对，组织成420个带有注释的系统性错位的视觉-语言数据集。Symbal在该基准上表现出色，正确识别出63.8%的数据集中的系统性错位，较最近的基线提高了近4倍。我们在SymbalBench上的评估还包括真实世界的评估，显示（1）Symbal能够准确地揭示四个MLLM生成的标题中的系统性错位，以及（2）Symbal是审计现成图像-标题数据集的强大工具。最终，我们的新任务、方法和基准可以帮助用户审计MLLM生成的标题并识别关键错误，而无需访问底层的MLLM。代码可在https://github.com/Stanford-AIMI/Symbal获取。
cs.CV / 83 / 2607.15220
Structural-Semantic Reciprocal Learning for Unsupervised Visible-Infrared Person Re-Identification
用于无监督可见-红外人脸重识别的结构-语义互学习
Abstract
Unsupervised visible-infrared person re-identification (USVI-ReID) is challenging due to the large modality gap and the lack of cross-modal identity annotations. Progressive association paradigms have been proposed to gradually bridge the gap, but they suffer from two critical bottlenecks: reliance on ambiguous global representations and unchecked propagation of pseudo-label noise in an open-loop manner. To address these issues, we propose Structural-Semantic Reciprocal Learning (SSRL), a framework that transforms open-loop association into a self-correcting closed-loop system. Structurally, we introduce Fine-grained Structural Decoupling (FSD) to extract discriminative body-part primitives as reliable spatial anchors, complementing ambiguous holistic silhouettes with spatially consistent structural details. Semantically, we design a Closed-loop Semantic Calibration (CSC) mechanism that reconstructs shared semantic prototypes at each epoch and feeds them back into the training loop, effectively filtering pseudo-label noise before the next clustering cycle. Through the reciprocal interaction between structural and semantic learning, SSRL achieves robust cross-modal representation. Extensive experiments demonstrate the competitive performance of SSRL against state-of-the-art USVI-ReID methods on both SYSU-MM01 and RegDB, notably surpassing several supervised counterparts on RegDB.
Chinese Translation
无监督可见-红外人脸重识别（USVI-ReID）因模态间差异大以及缺乏跨模态身份标注而面临挑战。为逐步弥合这一差距，提出了渐进关联范式，但它们面临两个关键瓶颈：依赖模糊的全局表示以及在开放循环中伪标签噪声的 unchecked 传播。为了解决这些问题，我们提出了结构-语义互学习（SSRL）框架，将开放循环关联转变为自我校正的闭环系统。在结构上，我们引入了细粒度结构解耦（FSD），提取可区分的身体部位原语作为可靠的空间锚点，以空间一致的结构细节补充模糊的整体轮廓。在语义上，我们设计了一种闭环语义校准（CSC）机制，在每个训练周期重建共享的语义原型，并将其反馈到训练循环中，有效过滤伪标签噪声，为下一个聚类周期做好准备。通过结构与语义学习之间的互相作用，SSRL 实现了稳健的跨模态表示。大量实验表明，SSRL 在 SYSU-MM01 和 RegDB 数据集上相较于最先进的 USVI-ReID 方法表现出竞争力，尤其是在 RegDB 上显著超越了多个监督学习的对手。
cs.CV / 84 / 2607.15227
Divergent Gaze Patterns in Artistic Viewing: Spatial and Temporal Signatures of Attention Across Autistic Individuals, Artists, and Neurotypical Observers
艺术观看中的不同注视模式：自闭症个体、艺术家和神经典型观察者的注意力空间和时间特征
Abstract
How different populations visually explore artworks bears on cognitive science and on accessibility design, yet most eye-tracking work in autism has used social scenes rather than art, and has analysed where the eyes land while ignoring when and in what order. We present a comparative free-viewing study across three groups, autistic adults (ASD), trained artists, and neurotypical observers, who each viewed 30 paintings for 15s. We introduce a directed, metric-grounded framework that compares groups along two complementary axes: a spatial axis, in which one group's fixation-density map predicts another's fixations under six saliency metrics (AUC-Judd, NSS, CC, SIM, KL, Information Gain); and a temporal axis, in which individual scanpaths are compared with MultiMatch, ScanMatch, a foveal-disc IoU score (FDISS), and dynamic time warping (DTW). Fixations are extracted uniformly for all groups with a dispersion-threshold algorithm. Three results converge. (i)Artists and neurotypicals are almost indistinguishable in both space (density-map correlation CC=0.96) and time (they form the most alignable scanpath pair), whereas ASD gaze diverges from both. (ii)ASD attention is dissociated: it matches artists' wide spatial exploration (dispersion, explored area) but carries a distinct temporal signature, shorter fixations, less dwell, and the most idiosyncratic (least self-consistent) scanpaths of any group. (iii)ASD gaze is not selectively artist-like on any metric; if anything it is marginally closer to neurotypical. Together these findings indicate that autistic viewing of art is a distinct, group-specific attentional profile in both space and time, and they motivate population-conditioned models of aesthetic attention. We release all analysis code and per-stimulus results.
Chinese Translation
不同人群如何视觉探索艺术作品与认知科学和无障碍设计密切相关，然而大多数关于自闭症的眼动追踪研究使用的是社交场景而非艺术作品，并且分析了眼睛停留的位置，却忽视了停留的时间和顺序。我们进行了一项比较自由观看的研究，涵盖三组参与者：自闭症成人（ASD）、受过训练的艺术家和神经典型观察者，他们各自观看了30幅画作，每幅画作15秒。我们引入了一个有针对性的、基于度量的框架，通过两个互补的轴线比较各组：空间轴线，其中一组的注视密度图在六个显著性指标下（AUC-Judd、NSS、CC、SIM、KL、信息增益）预测另一组的注视；时间轴线，其中个体的扫描路径与MultiMatch、ScanMatch、中心盘交并比得分（FDISS）和动态时间规整（DTW）进行比较。所有组的注视均通过一个分散阈值算法统一提取。结果有三个方面的收敛。(i)艺术家和神经典型观察者在空间（密度图相关性 CC=0.96）和时间上几乎无法区分，而ASD的注视则与两者不同。(ii)ASD的注意力是解离的：它与艺术家的广泛空间探索（分散度、探索面积）相匹配，但具有独特的时间特征，表现为较短的注视时间、较少的停留时间，以及任何组中最具个体差异（自我一致性最低）的扫描路径。(iii)ASD的注视在任何指标上都不具有选择性的艺术家特征；如果有的话，它与神经典型观察者的相似度略高。综合这些发现表明，自闭症对艺术的观看是一种独特的、特定于群体的注意力特征，既体现在空间上，也体现在时间上，并且这些发现激励了基于人群特征的美学注意力模型。我们将发布所有分析代码和每个刺激的结果。
cs.CV / 85 / 2607.15231
CRISP: Constrained Refinement via Iterative Squeezing Process for Robust Medical Image Segmentation under Domain Shift
CRISP：通过迭代挤压过程进行约束细化，以实现域转移下的稳健医学图像分割
Abstract
Distribution shift in medical imaging remains a central bottleneck for the clinical translation of medical AI. Failure to address it can lead to severe performance degradation in unseen environments and exacerbate health inequities. Existing methods for domain adaptation are inherently limited by exhausting predefined possibilities through simulated shifts or pseudo-supervision. Such strategies struggle in the open-ended and unpredictable real world, where distribution shifts are effectively infinite. To address this challenge, we adopt the "Rank Stability of Positive Regions" as a working assumption under distribution shift, and use it to derive robust spatial hints for source-only segmentation. Guided by this assumption, we propose CRISP, a model-agnostic framework that, unlike deployment-time adaptation, requires no test-time parameter updates and no target-domain data--a target-free, plug-in refinement framework that segments with frozen weights. Rather than using ranking to directly output masks, CRISP exploits the stability of probability rankings under distribution shift to derive robust spatial priors. Via latent feature perturbation, perturbation-invariant high-grade regions define a high-precision (HP) core, while voxels that remain potentially foreground under at least one perturbation define a high-recall (HR) support; these dual priors are then recursively refined under perturbation. We then design an iterative training framework that progressively squeezes HP and HR toward the final segmentation. Extensive evaluations on multi-center cardiac MRI and CT-based lung vessel segmentation demonstrate CRISP's superior robustness, significantly outperforming state-of-the-art methods with striking HD95 reductions of up to 0.14 (7.0% improvement), 1.90 (13.1% improvement), and 8.39 (38.9% improvement) pixels across multi-center, demographic, and modality shifts, respectively.
Chinese Translation
医学影像中的分布转移仍然是医学人工智能临床转化的一个主要瓶颈。未能解决这一问题可能导致在未见环境中的性能严重下降，并加剧健康不平等。现有的领域适应方法本质上受到通过模拟转移或伪监督耗尽预定义可能性的限制。这些策略在开放式和不可预测的现实世界中面临挑战，因为分布转移实际上是无限的。为了解决这一挑战，我们采用“正区域的排名稳定性”作为分布转移下的工作假设，并利用它推导出源仅分割的稳健空间提示。在这一假设的指导下，我们提出了CRISP，一个模型无关的框架，与部署时适应不同，它不需要测试时的参数更新，也不需要目标域数据——一个无目标的、插件式的细化框架，使用冻结的权重进行分割。CRISP并不是直接使用排名输出掩膜，而是利用分布转移下概率排名的稳定性来推导稳健的空间先验。通过潜在特征扰动，扰动不变的高等级区域定义了一个高精度（HP）核心，而在至少一个扰动下仍然可能为前景的体素定义了一个高召回（HR）支持；这两个先验随后在扰动下递归细化。我们设计了一个迭代训练框架，逐步将HP和HR压缩到最终分割。对多中心心脏MRI和基于CT的肺血管分割的广泛评估表明，CRISP具有优越的稳健性，在多中心、人口和模态转移中，显著超越了最先进的方法，HD95的减少高达0.14（改善7.0%）、1.90（改善13.1%）和8.39（改善38.9%）个像素。
cs.CV / 86 / 2607.15246
ARMOR++: Agentic Orchestration of a Multi-Domain Primitive Set for Transferable Attacks on Deepfake Detectors
ARMOR++：用于可转移攻击深度伪造检测器的多领域原语集的自主编排
Abstract
The reliability of deepfake detectors frequently degrades under black-box adversarial transfer, as these models often rely on fragile, architecture-dependent forensic cues. Existing transfer attacks often lack semantic awareness and struggle to maintain effectiveness under strict no-query constraints, particularly when perturbations are transferred from convolutional surrogates to transformer-based targets. To address these limitations, this paper introduces ARMOR++, a robust multi-agent framework designed for high-transferability deepfake evasion. The framework leverages the Qwen2.5-VL Vision-Language Model (VLM) to supply spatial semantic priors, while the Qwen3 Large Language Model (LLM) orchestrates primitive selection, adaptive hyperparameter reparameterization, and entropy-regularized perturbation mixing. By integrating five complementary primitives, spanning dense optimization, saliency-based methods, spatial transformations, frequency-domain perturbations, and block-structured modifications, ARMOR++ effectively targets heterogeneous inductive biases. Rigorous evaluation on the AADD-2025 benchmark demonstrates that ARMOR++ significantly outperforms existing agentic and non-agentic baselines across both low- and high-quality image regimes. Statistical analysis confirms a substantial gain in blind-target Attack Success Rate (ASR) over the state-of-the-art agentic baseline, with further performance advantages evidenced against non-agentic benchmarks and under robust defensive configurations. These findings highlight a significant residual reliability gap in current deepfake detector deployments and demonstrate the efficacy of agentic orchestration in identifying latent vulnerabilities.
Chinese Translation
深度伪造检测器的可靠性在黑箱对抗转移下常常下降，因为这些模型通常依赖于脆弱的、依赖于架构的取证线索。现有的转移攻击往往缺乏语义意识，并且在严格的无查询约束下难以保持有效性，尤其是在扰动从卷积替代模型转移到基于变换器的目标时。为了解决这些局限性，本文提出了ARMOR++，一个旨在实现高可转移性深度伪造规避的强大多代理框架。该框架利用Qwen2.5-VL视觉-语言模型（VLM）提供空间语义先验，同时Qwen3大型语言模型（LLM）负责原语选择、自适应超参数重参数化和熵正则化扰动混合。通过整合五种互补原语，包括密集优化、基于显著性的算法、空间变换、频域扰动和块结构修改，ARMOR++有效地针对异质归纳偏差。在AADD-2025基准上的严格评估表明，ARMOR++在低质量和高质量图像领域显著优于现有的自主和非自主基线。统计分析确认在盲目标攻击成功率（ASR）上，相较于最先进的自主基线有显著提升，并且在非自主基准和强健防御配置下进一步证明了性能优势。这些发现突显了当前深度伪造检测器部署中的显著残余可靠性差距，并展示了自主编排在识别潜在脆弱性方面的有效性。
cs.CV / 87 / 2607.15255
HoloGeo: Mitigating Landmark Bias in Geo-localization via Evidence-Driven Reasoning
HoloGeo：通过证据驱动推理减轻地标偏差在地理定位中的影响
Abstract
Recent advances in Vision-Language Models (VLMs) have significantly improved image geo-localization, yet existing models remain susceptible to landmark bias, causing them to overlook geographical cues or form spurious correlations, ultimately resulting in inaccurate localization. To systematically investigate this issue, we first design two quantitative metrics, Bias Intensity (BI) and Bias Harmfulness (BH), to characterize the impact of landmarks exerted on model reasoning, and establish a comprehensive benchmark, LandmarkBias-3K. To mitigate landmark bias, we further propose an evidence-driven reasoning framework, HoloGeo, to improve the reliability of geo-localization. HoloGeo is supported by a high-quality dataset, BF-30k, annotated with structured multi-evidence bias-free reasoning chains. By incorporating multi-dimensional rewards, HoloGeo explicitly encourages balanced attention over diverse visual cues and achieves evidence-driven joint reasoning. Extensive experiments demonstrate that HoloGeo not only maintains excellent performance on IM2GPS3K and YFCC4k but also significantly outperforms existing open-source VLMs on LandmarkBias-3K, validating its effectiveness for robust geospatial reasoning.
Chinese Translation
最近在视觉-语言模型（VLMs）方面的进展显著提高了图像的地理定位能力，但现有模型仍然容易受到地标偏差的影响，导致它们忽视地理线索或形成虚假的关联，最终导致定位不准确。为了系统性地研究这一问题，我们首先设计了两个定量指标，偏差强度（Bias Intensity, BI）和偏差危害性（Bias Harmfulness, BH），以表征地标对模型推理的影响，并建立了一个综合基准，LandmarkBias-3K。为了减轻地标偏差，我们进一步提出了一种证据驱动的推理框架HoloGeo，以提高地理定位的可靠性。HoloGeo依托于一个高质量的数据集BF-30k，该数据集标注了结构化的多证据无偏推理链。通过引入多维奖励，HoloGeo明确鼓励对多样化视觉线索的平衡关注，并实现证据驱动的联合推理。大量实验表明，HoloGeo不仅在IM2GPS3K和YFCC4k上保持了优异的性能，还在LandmarkBias-3K上显著超越了现有的开源VLMs，验证了其在稳健地理空间推理中的有效性。
cs.CV / 88 / 2607.15265
SceneBind: Binding What and Where Across Vision, Audio and Language
场景绑定：跨视觉、音频和语言绑定什么与在哪里
Abstract
We present SceneBind, an omni-modal representation of realistic scenes with joint semantic and 3D spatial understanding across vision, audio and language. Existing omni-modal encoders excel at instance-level semantics (i.e., what is present), but often lack explicit spatial structure (i.e., where it is). SceneBind addresses this gap by representing each scene as a semantic-spatial entity, combining a global semantic embedding with object-centric semantic-spatial slots. This representation explicitly captures object-level semantics, spatial attributes, and uncertainty. We further propose SceneBind Matching, a semantic-spatial matching scheme that integrates global scene similarity with object alignment, supporting cross-modal scene retrieval and object grounding. To train and evaluate SceneBind, we curate a novel real-world binaural audio-visual dataset with structured semantic and spatial annotations, and propose a training protocol for aligning semantic and spatial signals across modalities. SceneBind is compatible with large-scale pretrained semantic encoders, adds lightweight spatial modeling with only a few additional tokens. It achieves state-of-the-art scene and spatial retrieval while enabling strong zero-shot transfer to downstream tasks such as audio-visual localization.
Chinese Translation
我们提出了SceneBind，一种具有联合语义和三维空间理解的现实场景全模态表示，涵盖视觉、音频和语言。现有的全模态编码器在实例级语义（即，存在什么）方面表现出色，但往往缺乏明确的空间结构（即，在哪里）。SceneBind通过将每个场景表示为一个语义-空间实体来填补这一空白，结合了全局语义嵌入和以对象为中心的语义-空间槽。这种表示明确捕捉了对象级语义、空间属性和不确定性。我们进一步提出了SceneBind匹配，一种语义-空间匹配方案，集成了全局场景相似性与对象对齐，支持跨模态场景检索和对象定位。为了训练和评估SceneBind，我们策划了一个新颖的现实世界双耳音频-视觉数据集，具有结构化的语义和空间注释，并提出了一种训练协议，以对齐跨模态的语义和空间信号。SceneBind与大规模预训练的语义编码器兼容，仅需少量额外的标记即可增加轻量级空间建模。它在场景和空间检索方面实现了最先进的性能，同时支持强大的零样本迁移到下游任务，如音频-视觉定位。
cs.CV / 89 / 2607.15268
Motion-Conditioned Multi-View Fusion for Myocardial Infarction Localization from Echocardiography
基于运动条件的多视角融合用于心肌梗死的超声定位
Abstract
Myocardial infarction (MI) remains a leading cause of mortality worldwide. Echocardiography (Echo) is a widely available modality for MI assessment, where regional wall motion abnormality is a key indicator. Prior learning based methods for myocardial motion analysis often use handcrafted descriptors or densely supervised estimation, but the need for extensive annotation limits applicability. Foundation models have recently improved vision-based Echo analysis; however, most methods operate on single views and segment-level localization remains unreliable under view-dependent ambiguity, especially in apical views. To address this, we propose MCF-Net, a novel motion-guided multi-view fusion framework that fuses myocardial motion cues with foundation model representations to localize infarction. Visual features are extracted using EchoPrime, a pretrained Echo foundation model shared across dual views. Cardiac motion is modeled with extremely sparse supervision: a single annotated template frame is transferred across videos to initialize point tracking, avoiding dense labels. Motion-derived segment-aware soft masks provide coarse spatial priors that selectively enhance features for challenging myocardial segments. A motion-conditioned fusion mechanism then integrates motion and vision across views, refining predictions without overriding strong appearance cues. On segment-level MI localization, MCF-Net achieves 72.4\% F1 and 84.9\% accuracy, outperforming state-of-the-art motion-only, vision-only, and fusion baselines.
Chinese Translation
心肌梗死（MI）仍然是全球主要的死亡原因。超声心动图（Echo）是一种广泛可用的心肌梗死评估方法，其中区域壁运动异常是一个关键指标。以往基于学习的方法在心肌运动分析中常常使用手工设计的描述符或密集监督估计，但对大量标注的需求限制了其适用性。基础模型最近改善了基于视觉的超声分析；然而，大多数方法仅在单一视角下操作，且在视角依赖的模糊性下，分段级定位仍然不可靠，尤其是在心尖视图中。为了解决这一问题，我们提出了MCF-Net，一种新颖的运动引导多视角融合框架，旨在将心肌运动线索与基础模型表示融合，以定位梗死。视觉特征通过EchoPrime提取，这是一种在双视角间共享的预训练超声基础模型。心脏运动通过极其稀疏的监督进行建模：单个标注的模板帧在视频间转移以初始化点跟踪，从而避免了密集标签。运动导出的段感知软掩码提供粗略的空间先验，选择性地增强挑战性心肌段的特征。然后，运动条件融合机制在视角间整合运动和视觉，精炼预测而不覆盖强烈的外观线索。在分段级心肌梗死定位中，MCF-Net达到了72.4%的F1分数和84.9%的准确率，超越了最先进的仅运动、仅视觉和融合基线。
cs.CV / 90 / 2607.15271
Online Neural Space Time Memory for Dynamic Novel View Synthesis
动态新视图合成的在线神经时空记忆
Abstract
Online novel view synthesis from multi-view streaming videos faces a fundamental trade-off: maintaining a persistent, long-horizon memory to reconstruct temporarily occluded regions while operating under strict real-time constraints. While Test-Time Training (TTT) offers a powerful memory mechanism, standard models mandate gradient-based memory updates at every frame to adapt to the changing motion in dynamic scenes. The computational cost of heavy memory updates precludes real-time application and can lead to instability over long contexts. Given that memory updates are more demanding than memory application and video content is largely redundant, we propose to decouple the frequencies of these two processes. Our approach performs periodic memory updates while applying the memory on a per-frame basis, using cross-view attention to manage deformations between the prior memory state and the current frame. To lock in the historical context, we introduce two critical mechanisms: an auxiliary Memory Loss that forces persistent internalization of the scene, and a Memory Caching strategy that regularizes active weights against catastrophic drift. Our method demonstrates real-time, state-of-the-art performance on scenes with dynamic human motion as well as minute-scale online memorization.
Chinese Translation
从多视角流媒体视频中进行在线新视图合成面临着一个基本的权衡：在严格的实时约束下，保持持久的、长时间的记忆以重建暂时被遮挡的区域。尽管测试时训练（Test-Time Training, TTT）提供了一种强大的记忆机制，但标准模型要求在每一帧中进行基于梯度的记忆更新，以适应动态场景中变化的运动。频繁的记忆更新的计算成本阻碍了实时应用，并可能导致在长时间上下文中的不稳定性。鉴于记忆更新的需求高于记忆应用，而视频内容在很大程度上是冗余的，我们提出将这两个过程的频率解耦。我们的方法在每帧应用记忆的同时进行周期性的记忆更新，利用跨视角注意力来管理先前记忆状态与当前帧之间的变形。为了锁定历史上下文，我们引入了两个关键机制：一个辅助记忆损失（Memory Loss），强制场景的持久内化，以及一个记忆缓存策略（Memory Caching），对抗灾难性漂移的活跃权重进行正则化。我们的方法在动态人类运动场景以及分钟级在线记忆中展示了实时的、最先进的性能。
cs.CV / 91 / 2607.15273
MeanFlowNFT: Bringing Forward-Process RL to Average-Velocity Generators
MeanFlowNFT：将前向过程强化学习引入平均速度生成器
Abstract
MeanFlow generators achieve fast few-step sampling by predicting average velocities over time intervals, making them attractive for efficient generation. Reinforcement learning (RL) has become a powerful way to align diffusion and flow models with human preferences and task-specific objectives. In particular, DiffusionNFT offers an efficient forward-process RL framework that does not require reverse-process trajectories or likelihood estimation. However, applying such RL methods to MeanFlow remains underexplored. DiffusionNFT optimizes instantaneous velocities, whereas MeanFlow samples with average velocities. To bridge this gap, we introduce MeanFlowNFT. Inspired by the MeanFlow identity, which bridges average and instantaneous velocities, we construct an induced instantaneous-velocity predictor. We apply the DiffusionNFT objective to this predictor, making reward optimization well-defined for MeanFlow. Sampling remains based on the average velocity, preserving MeanFlow's fast few-step generation. We further prove that MeanFlowNFT inherits DiffusionNFT's strict policy-improvement guarantee. Experiments on image and video generation show that MeanFlowNFT consistently improves baselines. Moreover, it outperforms prior state-of-the-art RL-tuned few-step generators on most metrics ($6$ of $8$ on SD3.5-M), and can even surpass multi-step RL-tuned diffusion while using only a few sampling steps. For instance, on Wan 2.1, $4$-step MeanFlowNFT reaches a VBench score of $84.33$, surpassing $50$-step LongCat-Video RL ($82.57$).
Chinese Translation
MeanFlow 生成器通过预测时间间隔内的平均速度，实现快速的少步采样，使其在高效生成方面颇具吸引力。强化学习（RL）已成为将扩散模型和流动模型与人类偏好及特定任务目标对齐的强大方法。特别是，DiffusionNFT 提供了一种高效的前向过程 RL 框架，无需反向过程轨迹或似然估计。然而，将此类 RL 方法应用于 MeanFlow 仍然未得到充分探索。DiffusionNFT 优化瞬时速度，而 MeanFlow 则基于平均速度进行采样。为了解决这一差距，我们引入了 MeanFlowNFT。受到 MeanFlow 恒等式的启发，该恒等式连接了平均速度和瞬时速度，我们构建了一个诱导的瞬时速度预测器。我们将 DiffusionNFT 目标应用于该预测器，使得 MeanFlow 的奖励优化得以明确定义。采样仍然基于平均速度，从而保持 MeanFlow 快速的少步生成。我们进一步证明 MeanFlowNFT 继承了 DiffusionNFT 的严格策略改进保证。在图像和视频生成的实验中，MeanFlowNFT 始终优于基线。此外，在大多数指标上（在 SD3.5-M 上的 $8$ 个指标中有 $6$ 个），它超越了先前的最先进的 RL 调优少步生成器，甚至在仅使用少量采样步骤的情况下超过了多步 RL 调优的扩散生成器。例如，在 Wan 2.1 上，$4$ 步 MeanFlowNFT 达到 $84.33$ 的 VBench 分数，超越了 $50$ 步 LongCat-Video RL 的 $82.57$。
cs.CV / 92 / 2607.15278
Hierarchical Denoising For Multi-Step Visual Reasoning
多步骤视觉推理的层次去噪
Abstract
Video models are evolving into vision foundation models, yet they still lack human-like multi-step reasoning. Streaming autoregressive diffusion models are efficient but limited in reasoning, while bidirectional diffusion enables global revision with high inference costs due to dense frame-level denoising. Both paradigms struggle to achieve logical consistency and low-latency streaming for complex reasoning tasks. We propose HDR (Hierarchical Denoising for Visual Reasoning), a unified framework that integrates hierarchical latents into causal video generation for multi-step reasoning. HDR organizes video latents into a tree-structured hierarchy, enabling coarse-to-fine reasoning before streaming output. Coarse denoising layers preserve uncertain hypotheses for global planning, while finer layers progressively refine them into concrete visual states. A sparse hierarchical attention pattern (SHAP) further reduces temporal attention costs. We introduce a level-stratified multi-step video reasoning benchmark with out-of-distribution cases, covering six tasks: maze navigation, Tower of Hanoi, one-line drawing, sliding puzzle, Sokoban, and water pouring. Compared with streaming autoregressive diffusion baselines, HDR improves success from 34.22 to 60.29 (76.2% relative gain) and increases average progress from 76.00 to 89.56, demonstrating more consistent reasoning trajectories. HDR maintains low-latency streaming at 0.70 seconds per latent, achieving 54.2 times faster inference than bidirectional diffusion. It also retains 82.9% of full-data performance with only 2% training data, compared with 52.0% for bidirectional diffusion. Real-world robot experiments further demonstrate HDR's potential for physical interaction and world modeling. Project demo: https://hierarchical-diffusion-reasoning.github.io/.
Chinese Translation
视频模型正在演变为视觉基础模型，但仍然缺乏类人多步骤推理能力。流式自回归扩散模型高效但在推理上有限，而双向扩散由于密集的帧级去噪导致高推理成本，使得全局修正成为可能。这两种范式在复杂推理任务中都难以实现逻辑一致性和低延迟流式处理。我们提出了HDR（层次去噪用于视觉推理），这是一个统一框架，将层次潜变量整合到因果视频生成中以实现多步骤推理。HDR将视频潜变量组织成树状层次结构，使得在流式输出之前能够进行粗到细的推理。粗去噪层保留不确定的假设以进行全局规划，而细层则逐步将其精炼为具体的视觉状态。稀疏层次注意模式（SHAP）进一步降低了时间注意成本。我们引入了一个分层的多步骤视频推理基准，包含分布外案例，涵盖六个任务：迷宫导航、汉诺塔、一行画、滑块拼图、推箱子和倒水。与流式自回归扩散基线相比，HDR的成功率从34.22提高到60.29（相对增益76.2%），平均进展从76.00增加到89.56，展示了更一致的推理轨迹。HDR在每个潜变量保持0.70秒的低延迟流式处理，实现了比双向扩散快54.2倍的推理速度。与双向扩散的52.0%相比，HDR在仅使用2%训练数据的情况下保留了82.9%的全数据性能。现实世界的机器人实验进一步展示了HDR在物理交互和世界建模中的潜力。项目演示：https://hierarchical-diffusion-reasoning.github.io/
cs.AI / 1 / 2607.14093
Intelligent Three Level Learning Architecture for Autonomous UAV Swarms in Search and Rescue
用于自主无人机群搜索与救援的智能三层学习架构
Abstract
This paper presents a novel three level hierarchical learning architecture for autonomous UAV swarms performing search and rescue operations. Unlike conventional approaches that apply a single learning paradigm across all hierarchy levels, the proposed architecture integrates three qualitatively different learning mechanisms corresponding to the biological hierarchy of reflexes, skills, and reasoning such as Hebbian neuroplasticity for individual agent adaptation, multi agent reinforcement learning with graph neural networks and behavior trees for tactical coordination, and model agnostic meta learning with BDI reasoning and a digital twin for strategic decision making. The architecture is formalized through twenty two architectural contracts organized across six components such as BDI, Behavior Trees, GNN, MARL, Neuroplasticity, Meta Learning that collectively provide six classes of formal guarantees such as safety, budget correctness, optimality, liveness, starvation freedom, and inter level consistency. We introduce Swarm Meta Cognition as a compositional property arising from the structured interaction of all three levels, enabling the swarm to monitor its own cognitive state and switch between cognitive strategies. Five constructive progress functions for SAR task types bridge the gap between abstract optimization theory and concrete operational scenarios. The main integration theorem establishes that when all contracts are satisfied, the hybrid neuro-symbolic system preserves all six guarantee classes. For the dynamic case with active learning, five new contracts extend the framework with three additional guarantees such as cognitive resilience, graceful degradation, and monotonic meta improvement. Theoretical analysis demonstrates that the architecture addresses five fundamental limitations of existing hierarchical RL approaches.
Chinese Translation
本文提出了一种新颖的三层层次学习架构，用于执行搜索与救援任务的自主无人机群。与传统方法在所有层次上应用单一学习范式不同，所提出的架构整合了三种质的不同学习机制，分别对应于生物层次中的反射、技能和推理，包括用于个体智能体适应的Hebbian神经可塑性、用于战术协调的多智能体强化学习（MARL）结合图神经网络（GNN）和行为树，以及用于战略决策的无模型元学习（model agnostic meta learning）结合BDI推理和数字双胞胎（digital twin）。该架构通过二十二个架构合同形式化，组织在六个组件中，如BDI、行为树、GNN、MARL、神经可塑性和元学习，这些组件共同提供六类正式保证，如安全性、预算正确性、最优性、活性、无饥饿和层间一致性。我们引入了群体元认知（Swarm Meta Cognition）作为一种组合性质，源于所有三个层次的结构化互动，使得无人机群能够监控其自身的认知状态并在认知策略之间切换。五个构造性进展函数针对搜索与救援任务类型，弥合了抽象优化理论与具体操作场景之间的差距。主要集成定理建立了当所有合同得到满足时，混合神经符号系统保持所有六类保证。对于具有主动学习的动态案例，五个新合同扩展了框架，增加了三项额外保证，如认知韧性、优雅降级和单调元改进。理论分析表明，该架构解决了现有层次强化学习方法的五个基本局限性。
cs.AI / 2 / 2607.14095
HG-RAG: Hierarchy-Guided Retrieval-Augmented Generation for Structured Knowledge Graphs
HG-RAG：基于层次引导的检索增强生成用于结构化知识图谱
Abstract
Retrieval Augmented Generation (RAG) has proven to be a widely successful process at improving the quality of outputs from a Large Language Model (LLM) for wider context. However, RAG systems typically retrieve context from flat document stores, which struggles when queries require hierarchical or relational reasoning across structured knowledge. I present HG-RAG (Hierarchy-Guided RAG), a framework that performs graph-traversal over a hierarchical knowledge graph to deliver structured context to a language model. My retrieval pipeline resolves a named entity anchor from the query, then expands context upward through parent nodes, laterally through relational neighbors, and downward through child nodes when needed. I evaluate HG-RAG against a dense retrieval baseline across three world scales (18-800 nodes) with four query types: local fact, hierarchical, neighborhood, and multi-hop. Results show HG-RAG consistently outperforms the flat baseline on hierarchical, relational, and multi-hop reasoning tasks, while reducing hallucination and maintaining locality coherence.
Chinese Translation
检索增强生成（RAG）已被证明在提高大型语言模型（LLM）输出质量方面具有广泛的成功。然而，RAG 系统通常从平面文档存储中检索上下文，这在查询需要跨结构化知识进行层次或关系推理时表现不佳。我提出了 HG-RAG（层次引导 RAG），这是一个在层次知识图谱上执行图遍历的框架，以向语言模型提供结构化上下文。我的检索管道从查询中解析命名实体锚点，然后在需要时通过父节点向上扩展上下文，通过关系邻居横向扩展，并通过子节点向下扩展。我在三个世界规模（18-800 个节点）和四种查询类型（局部事实、层次、邻域和多跳）上评估 HG-RAG 与密集检索基线的表现。结果表明，HG-RAG 在层次、关系和多跳推理任务上始终优于平面基线，同时减少了幻觉并保持了局部一致性。
cs.AI / 3 / 2607.14096
IMEX Interaction-Based Model Explanation
基于交互的模型解释（IMEX）
Abstract
In predictive modeling, the ability to explain why a model produces a given target prediction has become increasingly important [5, 10]. Black-box models do not provide a transparent description of the internal mechanisms that generate the prediction, making even accurate predictions difficult to interpret and validate. In critical contexts, predictive accuracy alone is not a sufficient validation metric if the reasons underlying model decisions remain unexplained. The IMEX (Interaction-Based Model Explanation) approach represents a methodological direction within explainable predictive modeling. IMEX is designed to identify which variables contribute most to the target prediction and which interactions among variables are significant in determining the target. The method does not impose limitations on higher-order interaction analysis, allowing the investigation of feature subsets with cardinality greater than two. Beyond the identification of feature importance, IMEX enables the exploration of interaction patterns that may be consistent with latent mechanisms influencing the outcome. Through the application of the IMEX algorithm, it is possible to construct an interpretability map of the predictions. The IMEX framework is built on two complementary metrics: Static Correlation Power (PCS), which quantifies the contribution of individual features, and Interaction Correlation Power (PCI), which captures non-additive effects among features. In the present work, the PCS component is experimentally validated through a comparison with INVASE [18] on three synthetic datasets with known structures. The results indicate that IMEX can recover relevant feature-level structures in the presence of non-linear, conditional, and multicollinear relationships between input features and prediction targets.
Chinese Translation
在预测建模中，解释模型为何产生特定目标预测的能力变得越来越重要。黑箱模型无法提供生成预测的内部机制的透明描述，这使得即使是准确的预测也难以解释和验证。在关键背景下，仅仅依赖预测准确性作为验证指标是不够的，如果模型决策背后的原因仍未得到解释。IMEX（基于交互的模型解释）方法代表了可解释预测建模中的一种方法论方向。IMEX旨在识别哪些变量对目标预测贡献最大，以及变量之间哪些交互在决定目标时显著。该方法不对高阶交互分析施加限制，允许研究特征子集的基数大于二。除了识别特征重要性外，IMEX还能够探索可能与影响结果的潜在机制一致的交互模式。通过应用IMEX算法，可以构建预测的可解释性图谱。IMEX框架建立在两个互补指标之上：静态相关权重（PCS），量化单个特征的贡献，以及交互相关权重（PCI），捕捉特征之间的非加性效应。在本研究中，PCS组件通过与INVASE在三个已知结构的合成数据集上的比较进行了实验验证。结果表明，IMEX能够在输入特征与预测目标之间存在非线性、条件和多重共线性关系的情况下恢复相关的特征级结构。
cs.AI / 4 / 2607.14097
RegNetAgents: A Multi-Agent Framework for Cross-Network Regulatory Driver Identification in Cancer Genomics
RegNetAgents：一种用于癌症基因组学跨网络调控驱动因子识别的多智能体框架
Abstract
We introduce RegNetAgents, an AI-oriented multi-agent framework for structured, query-driven regulatory candidate identification across heterogeneous gene regulatory networks. The system enables unified analysis of bulk tumor and single-cell-derived ARACNe networks by integrating TCGA-derived cancer networks with large-scale single-cell regulatory networks from the GREmLN project. For a given focal gene, the framework performs dual-network classification, cancer gene filtering using OncoKB annotations, and mode-of-action (MoA) assignment for tumor-derived regulatory relationships. Candidates are ranked by evidence consistency across networks (Both, TCGA-only, GREmLN-only). The system is implemented as a multi-agent LangGraph DAG workflow, accessible through a unified Python API and Model Context Protocol (MCP) client, operating as a downstream analytical layer over precomputed regulatory networks rather than a network inference method. Across eleven breast cancer (BRCA) and twelve colorectal cancer (COAD) focal genes, RegNetAgents identifies candidate regulators significantly enriched for OncoKB-annotated cancer genes. TCGA-derived candidates show strong enrichment (Stouffer Z = 6.69 for BRCA and 6.95 for COAD), while GREmLN-derived candidates also demonstrate significant enrichment (Z = 5.51 for BRCA and 7.06 for COAD; all p < 0.0001). No enrichment is observed in housekeeping or non-driver control gene sets, supporting signal specificity. An extended module enables structured evaluation of oncogenic potential, druggability, clinical relevance, and network vulnerability, supporting end-to-end interpretation from candidate identification to biological hypothesis generation. RegNetAgents establishes an interpretable AI framework for cross-network regulatory candidate identification in cancer genomics.
Chinese Translation
我们介绍了RegNetAgents，这是一个面向人工智能的多智能体框架，用于在异构基因调控网络中进行结构化、查询驱动的调控候选因子识别。该系统通过整合来自TCGA的癌症网络与GREmLN项目的大规模单细胞调控网络，实现了对大肿瘤和单细胞衍生的ARACNe网络的统一分析。对于给定的焦点基因，该框架执行双网络分类，利用OncoKB注释进行癌症基因过滤，并为肿瘤衍生的调控关系分配作用模式（MoA）。候选因子根据网络间的证据一致性进行排名（包括TCGA、GREmLN和两者）。该系统实现为一个多智能体LangGraph有向无环图（DAG）工作流，通过统一的Python API和模型上下文协议（MCP）客户端访问，作为对预计算调控网络的下游分析层，而不是网络推断方法。在十一种乳腺癌（BRCA）和十二种结直肠癌（COAD）焦点基因中，RegNetAgents识别出的候选调控因子在OncoKB注释的癌症基因中显著富集。来自TCGA的候选因子显示出强烈的富集（BRCA的Stouffer Z = 6.69，COAD的Stouffer Z = 6.95），而来自GREmLN的候选因子也表现出显著富集（BRCA的Z = 5.51，COAD的Z = 7.06；所有p < 0.0001）。在管家基因或非驱动控制基因集中未观察到富集，支持信号特异性。扩展模块支持对致癌潜力、药物可及性、临床相关性和网络脆弱性的结构化评估，支持从候选识别到生物假设生成的端到端解释。RegNetAgents建立了一个可解释的人工智能框架，用于癌症基因组学中的跨网络调控候选因子识别。
cs.AI / 5 / 2607.14115
DialogueVPR: Towards Conversational Visual Place Recognition
DialogueVPR：面向对话式视觉地点识别
Abstract
Inspired by how humans communicate spatial information, language-guided geo-localization has gained significant traction for its intuitive and practical value. Despite this progress, most methods still rely on a static, one-shot retrieval paradigm, which fails to handle the ambiguity and incompleteness inherent in real-world natural language descriptions. We propose a paradigm shift to reasoning retrieval and introduce Dialogue Place Recognition (DlgPR), which casts localization as an interactive, dialogue-driven reasoning process. To support this new task, we present DlgQuest-Cities, the first large-scale dialogue-based benchmark for place recognition, and a unified reasoning framework that couples a cross-modal multi-level retriever with an intelligent questioner, DQ-pilot. DQ-pilot is trained in a curriculum: supervised fine-tuning on a curated DQ-cities-20k subset followed by reinforcement refinement on a harder DQ-cities-10k split via GRPO. Two task-aligned metrics guide learning: a Discriminative Difficulty Index (DDI) for curriculum sampling and a Positional Retrieval Gain (PRG) reward that directly measures retrieval improvement induced by a question. Experiments show this reasoning-based approach significantly outperforms baselines. The code and model are available at https://github.com/Graysonggg/DlgPR.
Chinese Translation
受到人类如何交流空间信息的启发，基于语言的地理定位因其直观和实用的价值而获得了显著关注。尽管取得了这些进展，大多数方法仍然依赖于静态的一次性检索范式，这无法处理现实世界自然语言描述中固有的模糊性和不完整性。我们提出了一种范式转变，转向推理检索，并引入对话地点识别（Dialogue Place Recognition，DlgPR），将定位视为一种互动的、以对话驱动的推理过程。为了支持这一新任务，我们提出了DlgQuest-Cities，这是第一个大规模基于对话的地点识别基准，以及一个统一的推理框架，该框架将跨模态多层次检索器与智能提问者DQ-pilot结合在一起。DQ-pilot在一个课程中进行训练：首先在经过整理的DQ-cities-20k子集上进行监督微调，然后通过GRPO在更难的DQ-cities-10k拆分上进行强化改进。两个与任务对齐的指标指导学习：用于课程采样的区分难度指数（Discriminative Difficulty Index，DDI）和直接测量由问题引起的检索改进的位置信息检索增益（Positional Retrieval Gain，PRG）奖励。实验表明，这种基于推理的方法显著优于基线。代码和模型可在https://github.com/Graysonggg/DlgPR获取。
cs.AI / 6 / 2607.14126
Interpretable Language Model for Closed-Loop Type 1 Diabetes Control
可解释的闭环1型糖尿病控制语言模型
Abstract
Type 1 Diabetes (T1D) is a chronic, life-threatening autoimmune condition characterized by the complete destruction of insulin-producing pancreatic beta cells. While Artificial Pancreas Systems (APS) powered by Reinforcement Learning (RL) have shown promise in automating insulin delivery, their ``black-box'' nature makes it hard for patients and doctors to trust them fully. This paper presents LLM-T1D, a promising approach that combines the precision of RL with the clear, human-like reasoning of Large Language Models (LLMs) to create a more transparent and reliable insulin pump controller. By training an expert RL system and distilling its knowledge into fine-tuned LLaMA 3.1 8B and Qwen3 8B models, we developed a controller that not only surpasses the RL system's performance but also explains its decisions in plain, understandable language. Tested on the FDA-approved UVA/Padova T1D simulator, the LLM controllers deliver excellent blood sugar control (73.5% Time in Range) while maintaining strict formal safety verification against hallucinations.
Chinese Translation
1型糖尿病（T1D）是一种慢性、危及生命的自身免疫性疾病，其特征是胰腺β细胞的完全破坏，导致胰岛素生产不足。尽管基于强化学习（RL）的人工胰腺系统（APS）在自动化胰岛素输送方面展现了良好的前景，但其“黑箱”特性使得患者和医生难以完全信任这些系统。本文提出了LLM-T1D，这是一种有前景的方法，将强化学习的精确性与大型语言模型（LLMs）清晰、类人思维的推理能力相结合，以创建一个更透明、可靠的胰岛素泵控制器。通过训练一个专家级的RL系统，并将其知识提炼到经过微调的LLaMA 3.1 8B和Qwen3 8B模型中，我们开发了一个控制器，不仅超越了RL系统的性能，还能用简单易懂的语言解释其决策。在FDA批准的UVA/Padova T1D模拟器上进行测试时，LLM控制器提供了出色的血糖控制（73.5%的时间在目标范围内），同时在防止幻觉方面保持严格的形式安全验证。
cs.AI / 7 / 2607.14141
Human AI Construction of Bayesian Networks for Operational Decision Support -- A Virtual Survey Approach
人类人工智能构建贝叶斯网络以支持运营决策——一种虚拟调查方法
Abstract
Bayesian Belief Networks (BBNs) are powerful tools for decision-making under uncertainty. However, building their structures and estimating parameters are difficult. Currently, researchers must choose between relying on expert judgement or using large datasets to learn the structure and parameters of the network. We propose a new methodology using Large Language Models to bridge the gap between expert opinion and data-driven learning. This approach uses a panel of AI agents to estimate probabilities based on specific personas and context. We then apply a trimmed-mean rule to remove noise from these responses. We develop a six step BBN framework and illustrate it to model customer intention to consult a doctor in an alternative healthcare system. The model reveals that while self efficacy appears to be a major factor, its actual causal impact is small. In contrast, subjective norms have a much stronger effect in modelling customers' intention. The most effective strategy is to improve both confidence and community norms simultaneously.
Chinese Translation
贝叶斯信念网络（BBNs）是应对不确定性决策的强大工具。然而，构建其结构和估计参数是困难的。目前，研究人员必须在依赖专家判断和使用大数据集学习网络的结构与参数之间进行选择。我们提出了一种新方法，利用大型语言模型（Large Language Models）来弥合专家意见与数据驱动学习之间的差距。该方法使用一组人工智能代理，根据特定角色和上下文估计概率。然后，我们应用修剪均值规则来去除这些响应中的噪声。我们开发了一个六步的贝叶斯网络框架，并以此为例建模客户在替代医疗系统中咨询医生的意图。模型揭示，尽管自我效能似乎是一个主要因素，但其实际因果影响较小。相比之下，主观规范在建模客户意图方面具有更强的影响。最有效的策略是同时提高自信心和社区规范。
cs.AI / 8 / 2607.14144
Capability from Access Structure, Not Scale: Lower Bounds and Pre-Registered Tests for Hybrid Sequence Models
能力源于访问结构，而非规模：混合序列模型的下限与预注册测试
Abstract
The Platonic Representation Hypothesis (PRH) holds that as models scale, representations of heterogeneous networks converge toward a shared model of reality. We propose its sequel and boundary, the Capability Convergence Hypothesis (CCH): under a fixed per-token inference budget, representational convergence does not entail capability convergence. Capability instead converges toward a class, the access-complete hybrid: any architecture holding both a compressive O(1)-state channel and a scalable verbatim-index channel. We anchor it on a witness task, the Newton's-apple problem in an infinite stream, and name three resource walls: a Shannon wall barring any o(Nb)-state architecture, a horizon wall barring any fixed window, and a circuit wall barring fixed-depth attention-only composition (conditional on TC0 != NC1). Under an explicit separability assumption a hybrid crosses all three by paying each wall's price, so capability is strictly super-additive under composition. We separate what we prove from what we conjecture: the access-completeness principle rests on information-theoretic lower bounds and pre-registered experiments, while the field-level convergence trend is an economics-motivated conjecture. We report the first pre-registered small-scale tests under criteria frozen before the data: the predicted scissors gap is measured (exact-retrieval error 0.994 vs. 0.000 once a 64-scalar state gains one global-attention layer), the state-tracking bifurcation lands at the registered boundary, and a conjunction witness shows an irreducibly two-channel solution; one prediction failed with its direction reversed and is reported as such. Representational convergence is given freely by scale; capability convergence must be purchased by access structure.
Chinese Translation
柏拉图表征假说（Platonic Representation Hypothesis, PRH）认为，随着模型规模的扩大，异构网络的表征趋向于一个共享的现实模型。我们提出其续篇和边界，即能力收敛假说（Capability Convergence Hypothesis, CCH）：在固定的每个标记推理预算下，表征收敛并不意味着能力收敛。能力则趋向于一个类，即访问完全混合（access-complete hybrid）：任何同时具备压缩 O(1) 状态通道和可扩展逐字索引通道的架构。我们将其锚定在一个见证任务上，即无限流中的牛顿苹果问题，并命名三个资源壁垒：一个香农壁垒（Shannon wall）阻止任何 o(Nb) 状态架构，一个地平线壁垒（horizon wall）阻止任何固定窗口，以及一个电路壁垒（circuit wall）阻止固定深度的仅注意力组合（前提是 TC0 != NC1）。在一个明确的可分离性假设下，混合架构通过支付每个壁垒的代价跨越所有三个壁垒，因此在组合下能力是严格超加性的。我们将证明的内容与推测的内容区分开来：访问完全性原则基于信息论下限和预注册实验，而领域级收敛趋势则是一个经济动机驱动的推测。我们报告了在数据之前冻结标准下的首次预注册小规模测试：预测的剪刀差（scissors gap）被测量（精确检索误差 0.994 对比 0.000，当一个 64 维标量状态获得一个全局注意力层时），状态跟踪分岔落在注册边界上，并且一个结合见证显示出一个不可简化的双通道解决方案；一个预测失败并反向报告。表征收敛是由规模自由提供的；能力收敛必须通过访问结构来获得。
cs.AI / 9 / 2607.14145
ToolAnchor: Anchoring Counterfactual Context to Boost Agentic Tool-use Capability
ToolAnchor：锚定反事实上下文以提升工具使用能力
Abstract
Tool-augmented large language model agents excel at long-horizon tasks, yet they are typically post-trained on fixed toolsets. When tasks demand new tools, these agents struggle to incorporate them effectively, and retraining from scratch is often impractical. We identify the core obstacle in such toolset expansion problem as behavioral inertia: the tendency of agents to fall back on familiar tools and established reasoning patterns despite having access to new ones. We demonstrate that injecting counterfactual anchor contexts at critical decision points can break this inertia, recovering failed trajectories by eliciting suppressed agent capabilities. To scale this insight, we propose ToolAnchor, a framework that uses teacher models to hypothesize these counterfactual contexts, verifies them via student rollouts, and internalizes the successful interventions through agentic post-training. Extensive evaluations across general AI assistant (GAIA), textual search (BrowseComp), and visual search (VDR-Bench) tasks demonstrate that ToolAnchor consistently exhibits competitive performance under expanded toolsets. Our work bridges the gap between static post-training and dynamic adaptation, charting a new path for scalable agentic reinforcement learning.
Chinese Translation
工具增强的大型语言模型代理在长时间任务中表现出色，但它们通常是在固定工具集上进行后训练的。当任务需要新工具时，这些代理在有效整合新工具方面往往面临困难，而从头开始重新训练通常是不切实际的。我们将这种工具集扩展问题的核心障碍识别为行为惯性：代理倾向于在熟悉的工具和既定推理模式中回退，尽管它们可以使用新的工具。我们证明，在关键决策点注入反事实锚定上下文可以打破这种惯性，通过引发被抑制的代理能力来恢复失败的轨迹。为了扩展这一见解，我们提出了ToolAnchor，一个利用教师模型假设这些反事实上下文的框架，通过学生回滚验证这些上下文，并通过代理后训练内化成功的干预措施。在一般人工智能助手（GAIA）、文本搜索（BrowseComp）和视觉搜索（VDR-Bench）任务中的广泛评估表明，ToolAnchor在扩展工具集下始终展现出竞争力的性能。我们的工作弥合了静态后训练与动态适应之间的差距，为可扩展的代理强化学习开辟了一条新路径。
cs.AI / 10 / 2607.14149
Enhancing Small Language Models Reasoning through Knowledge Graph Grounding
通过知识图谱基础增强小型语言模型的推理能力
Abstract
Although large language models (LLMs) have set benchmarks for zero-shot reasoning, their deployment remains cost-prohibitive and environmentally taxing. Small Language Models (SLMs) offer a sustainable alternative, but prone to errors, on tasks requiring complex, multi-hop logical grounding. We investigate a neuro-symbolic agentic framework to enhance the reasoning capabilities of SLMs, specifically Gemma 3 (1B, 4B) and Llama 3.2 (3B), using the CLUTRR kinship benchmark. Our approach transforms the SLM into a minimalist agent utilizing two specialized tool calls: extract_facts for symbolic triplet extraction and get_hint for expert reasoning via a Relational Graph Convolutional Network (RGCN). We evaluate these models across two configurations, both in an Oracle scenario with ground-truth triplets and a Realistic scenario relying on self-extracted knowledge. Our results reveal that while RGCN-derived hints provide a 1.5 - 2x performance gain over story-only baselines, the system is constrained by the extraction bottleneck and sequential deductive fragility, where early extraction errors compound over multi-hop chains. Furthermore, we identify a "distraction effect" in specific architectures where noisy, self-generated facts degrade performance despite the presence of expert hints. This work characterizes the challenges of symbolic grounding in low-resource agentic systems and provides a roadmap for iterative verification in neuro-symbolic agentic pipelines.
Chinese Translation
尽管大型语言模型（LLMs）在零样本推理方面设定了基准，但其部署仍然成本高昂且对环境造成负担。小型语言模型（SLMs）提供了一种可持续的替代方案，但在需要复杂的多跳逻辑基础的任务中容易出错。我们研究了一种神经符号代理框架，以增强SLMs的推理能力，特别是Gemma 3（1B，4B）和Llama 3.2（3B），使用CLUTRR亲属关系基准。我们的方法将SLM转变为一个极简的代理，利用两个专门的工具调用：extract_facts用于符号三元组提取，get_hint用于通过关系图卷积网络（RGCN）进行专家推理。我们在两种配置下评估这些模型，分别是在具有真实三元组的Oracle场景和依赖自提取知识的现实场景。我们的结果显示，尽管RGCN派生的提示在故事仅基线之上提供了1.5到2倍的性能提升，但系统受到提取瓶颈和顺序推理脆弱性的限制，早期提取错误在多跳链中累积。此外，我们还发现特定架构中存在“干扰效应”，即噪声自生成的事实尽管存在专家提示，仍会降低性能。这项工作描述了低资源代理系统中符号基础的挑战，并为神经符号代理管道中的迭代验证提供了路线图。
cs.AI / 11 / 2607.14158
Orchestrating Power Grid Studies with Multi-Agent AI and MCP Servers
通过多智能体人工智能和MCP服务器协调电网研究
Abstract
This position paper explores how Agentic AI and Model Context Protocol (MCP) can support power-grid studies in a Transmission System Operator (TSO) context. We focus on integrating Large Language Models with numerical simulation tools, structured workflows, and human supervision. We identify key industrial requirements for agent assisted grid studies and introduce pypowsybl-mcp, an MCP-based interface exposing selected capabilities of our simulation tool, pypowsybl to AI agents. This first step provides a testbed to study how agents can setup simulations, execute analyses, retrieve results, and interact with power-system simulators through standardized tool calls. We also discuss principles for human-in-the-loop, multi-agent workflows and outline an evaluation strategy combining technical metrics and practitioner feedback. The paper positions MCP-based tool integration as a step toward more interactive, auditable, and scalable grid-study environments.
Chinese Translation
本文探讨了智能体人工智能（Agentic AI）和模型上下文协议（Model Context Protocol, MCP）如何在输电系统运营商（Transmission System Operator, TSO）的背景下支持电网研究。我们重点关注将大型语言模型与数值仿真工具、结构化工作流程和人工监督相结合。我们识别出智能体辅助电网研究的关键工业需求，并介绍了pypowsybl-mcp，这是一个基于MCP的接口，向人工智能智能体开放我们仿真工具pypowsybl的选定功能。此第一步提供了一个测试平台，以研究智能体如何设置仿真、执行分析、检索结果，并通过标准化工具调用与电力系统仿真器进行交互。我们还讨论了人机协作、多智能体工作流程的原则，并概述了一种结合技术指标和从业者反馈的评估策略。本文将基于MCP的工具集成定位为迈向更具互动性、可审计性和可扩展的电网研究环境的一步。
cs.AI / 12 / 2607.14159
MemoHarness: Agent Harnesses That Learn from Experience
MemoHarness：从经验中学习的智能体控制框架
Abstract
An agent harness is the external control layer that turns a base LLM into an executable agent by managing context, tools, orchestration, memory, decoding, and output handling. While harness design strongly affects agent behavior, most automatic improvement methods optimize narrower artifacts such as prompts, pipelines, or workflows, and deployed agents usually reuse a single global harness for all cases. We introduce MemoHarness, an adaptive harness optimization framework that learns from its own executions. MemoHarness decomposes the harness into six editable control dimensions, stores per-case diagnoses and distilled global patterns in a dual-layer experience bank, and adapts the learned harness to each test case using retrieved experience without test-time labels, feedback, or additional search. In our evaluation across shell-agent, code-generation, and analytical-reasoning benchmarks, MemoHarness improves over the fixed harnesses we compare against and shows selective transfer to unseen suites and base models. Its additional context can also remain cost-competitive when much of the retrieved experience is cacheable. These results provide evidence that execution experience is a practical substrate for building agent harnesses that are more adaptive than a single static configuration, while leaving broader claims about statistical robustness and component attribution to future work.
Chinese Translation
智能体控制框架是将基础大型语言模型（LLM）转变为可执行智能体的外部控制层，通过管理上下文、工具、编排、记忆、解码和输出处理来实现。虽然控制框架的设计对智能体行为有重要影响，但大多数自动改进方法仅优化较窄的工件，如提示、管道或工作流程，且部署的智能体通常为所有情况重用单一的全局控制框架。我们提出了MemoHarness，一个自适应的控制框架优化框架，它通过自身的执行学习。MemoHarness将控制框架分解为六个可编辑的控制维度，在双层经验库中存储每个案例的诊断和提炼的全局模式，并利用检索到的经验在没有测试时标签、反馈或额外搜索的情况下，将学习到的控制框架适应于每个测试案例。在我们对shell-agent、代码生成和分析推理基准的评估中，MemoHarness在与之比较的固定控制框架上有所改进，并显示出对未见套件和基础模型的选择性迁移。当大部分检索到的经验是可缓存时，其额外的上下文也能保持成本竞争力。这些结果提供了证据，表明执行经验是构建比单一静态配置更具适应性的智能体控制框架的实用基础，同时将关于统计鲁棒性和组件归属的更广泛主张留待未来的研究。
cs.AI / 13 / 2607.14169
When a Verified World Model Still Loses: Play-Adequacy vs Prediction-Accuracy in LLM-Synthesized Code World Models
当经过验证的世界模型仍然失效时：LLM合成代码世界模型中的游戏适应性与预测准确性
Abstract
Large language models can synthesize a game's rules as executable code - a Code World Model (CWM) - which a classical planner then searches over. Such models are typically accepted when they reach high transition accuracy on sampled trajectories. We argue this is the wrong notion of adequacy for planning. We show four things. (1) An LLM-synthesized CWM can pass a sampling gate at 100% transition accuracy and be $\geq 98\%$ state-accurate on the planner's own search distribution, yet lose systematically at play, because the $<1\%$ it gets wrong is exactly the pivotal dynamics; the play cost of the omitted rule is $0.091$ (seed-clustered 95% CI $[0.065,0.117]$, $n=4800$). We call this the verified-vs-correct gap, and confirm it end-to-end through the synthesis pipeline. (2) The harm follows a quantitative law, $\mathrm{danger}=\mathrm{play\_cost}\times(1-\mathrm{rarity})^N$, whose $(1-\mathrm{rarity})^N$ gate-miss factor is proven exact and whose play cost is empirically bounded. (3) The failure is not repaired by more data: LLM synthesis behaves as rule translation, not rule inference, and did not infer the omitted rule across models (GPT-5.x) and data regimes (including DAgger and targeted examples). (4) The same mechanism recurs on the belief-inference function of imperfect-information CWMs: we prove a coverage bound (a size-$N$ gate is identifying when $N\gtrsim b^{d_{\max}}$), explaining why shallow games such as Kuhn poker show no gap, and hand-construct Beacon, a verified-but-wrong inference function that passes the gate yet loses every game. These results suggest adequacy for planning-oriented world models should be measured on the search distribution or by play directly, not by prediction accuracy on sampled transitions.
Chinese Translation
大型语言模型能够将游戏规则合成可执行代码——代码世界模型（Code World Model, CWM）——然后由经典规划器进行搜索。当这些模型在采样轨迹上达到高过渡准确率时，通常被接受。我们认为这不是规划的正确适应性概念。我们展示了四个方面的内容。(1) 一个LLM合成的CWM可以在100%的过渡准确率下通过采样门，并在规划者自身的搜索分布上达到$ ext{≥}98\%$的状态准确率，但在游戏中系统性失败，因为它错误的部分不足1%恰恰是关键动态；被省略规则的游戏成本为$0.091$（种子聚类95%置信区间$[0.065,0.117]$，$n=4800$）。我们称之为验证与正确之间的差距，并通过合成管道端到端确认这一点。(2) 这种损害遵循一个定量法则，$ ext{danger}= ext{play extunderscore cost} imes(1- ext{rarity})^N$，其$(1- ext{rarity})^N$的门失效因子被证明是精确的，而游戏成本在经验上是有界的。(3) 更多数据并不能修复这种失败：LLM合成表现为规则翻译，而非规则推断，并且在不同模型（GPT-5.x）和数据模式（包括DAgger和目标示例）中未能推断出被省略的规则。(4) 同样的机制在不完美信息的CWM的信念推断功能中反复出现：我们证明了覆盖界限（当$N ext{≥}b^{d_{ ext{max}}}$时，大小为$N$的门是识别的），解释了为什么像库恩扑克这样的浅层游戏没有显示出差距，并手动构造了Beacon，一个经过验证但错误的推断功能，它通过了门却在每场游戏中都失败。这些结果表明，面向规划的世界模型的适应性应在搜索分布上或通过游戏直接进行测量，而不是通过对采样过渡的预测准确性进行测量。
cs.AI / 14 / 2607.14178
ReasFlow: Assisting Reasoning-Centric Scientific Discovery in Applied Mathematics via a Knowledge-Based Multi-Agent System
ReasFlow：通过基于知识的多智能体系统辅助以推理为中心的应用数学科学发现
Abstract
Recent advances in Large Language Models have fueled autonomous AI agents capable of tackling complex scientific tasks, yet existing automated research systems remain predominantly focused on empirically driven domains with quantitative benchmarks, leaving theory-driven discovery, particularly in mathematically grounded disciplines requiring rigorous proofs and synthesis of domain knowledge, largely underexplored. Key challenges include the difficulty of verifying theoretical reasoning at scale, insufficient reasoning ability for autonomous frontier exploration, and a scarcity of procedural heuristics in the literature. We introduce ReasFlow, an end-to-end autonomous agent system for reasoning-centric scientific discovery that operationalizes a collaborative paradigm where the human expert acts as Principal Investigator while the agent executes rigorous derivations as a capable graduate student. ReasFlow incorporates (i) a robust internal verification loop that audits logical coherence and corrects fundamental errors prior to human inspection, and (ii) an automated knowledge retrieval and self-improvement mechanism that proactively surfaces both declarative facts and overlooked procedural heuristics, substantially reducing expert intervention. The system unifies literature synthesis, algorithm design, theorem proving, experimentation, and manuscript preparation in a single system. Deployed to autonomously generate five complete research papers with rigorous theoretical and empirical content from minimal prompts, ReasFlow consistently achieves the highest evaluation scores among state-of-the-art open-access baselines under a curated LLM-based review rubric. ReasFlow is publicly accessible via the ReasLab platform, providing a collaborative workspace for AI-assisted theoretical research. Github repo: https://github.com/ReasLab/ReasFlow.git.
Chinese Translation
近期大型语言模型的进展推动了能够处理复杂科学任务的自主人工智能代理的发展，然而现有的自动化研究系统仍主要集中于以经验驱动的领域和定量基准，理论驱动的发现，尤其是在需要严格证明和领域知识综合的数学基础学科中，仍然大多未被探索。主要挑战包括在大规模验证理论推理的困难、自主前沿探索的推理能力不足，以及文献中程序性启发式方法的稀缺。我们提出了ReasFlow，一个以推理为中心的科学发现的端到端自主代理系统，该系统实现了一种协作范式，其中人类专家充当首席研究员，而代理则像一名有能力的研究生一样执行严格的推导。ReasFlow包含(i)一个强大的内部验证循环，在人类检查之前审核逻辑一致性并纠正基本错误，以及(ii)一个自动知识检索和自我改进机制，主动呈现声明性事实和被忽视的程序性启发式，大幅减少专家干预。该系统将文献综合、算法设计、定理证明、实验和手稿准备统一在一个系统中。ReasFlow被部署为自主生成五篇完整的具有严格理论和实证内容的研究论文，仅需最少的提示，并在经过精心策划的基于大型语言模型的评审标准下，始终在最先进的开放获取基准中获得最高评价分数。ReasFlow通过ReasLab平台公开访问，为人工智能辅助的理论研究提供了一个协作工作空间。Github 仓库：https://github.com/ReasLab/ReasFlow.git。
cs.AI / 15 / 2607.14187
RxBrain: Embodied Cognition Foundation Model with Joint Language-Visual Reasoning and Imagination
RxBrain：具有联合语言-视觉推理和想象的具身认知基础模型
Liang, Haotian, Chen, Mingkang, Huang, Yufei, Guo, Yuchun, Zhu, Xiaomeng, Shi, Xiangli, Wang, Kaixuan, Mao, Yunxuan, Zhou, Weijie, Chen, Ling, Zeng, Shirong, Long, Yueyu, Si, Yuchen, Zhu, Yajuan, Zhou, Xingyu, Wang, Minghui, He, Wanjia, Yang, Xin, Xiang, Lingzhu, Liu, Zhiqing, Ma, Bohan, Huang, Xiran, Yang, Tianshuo, Liu, Zhiheng, Xiong, Xuantang, Lu, Zisheng, Luo, Ping, Mu, Yao, Hu, Han, Zhang, Zhengyou
Abstract
Embodied cognition requires agents to connect high-level task reasoning with the physical states to be achieved. We introduce Hy-Embodied-RxBrain, an embodied cognition foundation model with joint language-visual reasoning and imagination. Unlike vision-language models that emphasize scene understanding and textual decision making, or generative world models that mainly predict future visual states, RxBrain represents embodied plans in a single planning sequence where language and visual imagination play complementary roles. Language provides the abstract structure of a plan, including task decomposition, planning primitives, constraints, temporal order, and decision logic, while visual imagination grounds this structure through world state prediction and joint subgoal planning, associating each planning step with intermediate and final physical states. RxBrain adopts a unified multimodal Mixture-of-Transformers architecture that supports language, image, and video understanding and generation within one model. To train this capability, we build an automatic pipeline that converts embodied videos into joint text-visual planning supervision by decomposing videos into planning steps and aligning them with visual state transitions. We further introduce RxBrain-Bench to evaluate whether models can represent embodied plans through joint textual and visual components rather than separate understanding or generation. Experiments show that RxBrain maintains embodied understanding and generation abilities, and produces plans with coupled textual reasoning, world state prediction, and joint subgoal planning. We also extend RxBrain to continuous robot action generation, where it shows promising real-robot performance without large-scale action-data pretraining. These results provide an initial step toward foundation models for embodied cognition.
Chinese Translation
具身认知要求智能体将高层次的任务推理与要实现的物理状态连接起来。我们介绍了Hy-Embodied-RxBrain，这是一种具身认知基础模型，具有联合语言-视觉推理和想象。与强调场景理解和文本决策的视觉-语言模型，或主要预测未来视觉状态的生成世界模型不同，RxBrain在单一的规划序列中表示具身计划，其中语言和视觉想象发挥互补作用。语言提供了计划的抽象结构，包括任务分解、规划原语、约束、时间顺序和决策逻辑，而视觉想象通过世界状态预测和联合子目标规划来落实这一结构，将每个规划步骤与中间和最终的物理状态关联起来。RxBrain采用统一的多模态混合变换器（Mixture-of-Transformers）架构，支持在一个模型中进行语言、图像和视频的理解与生成。为了训练这一能力，我们构建了一个自动化管道，通过将具身视频分解为规划步骤并与视觉状态转变对齐，将其转换为联合文本-视觉规划监督。我们进一步引入了RxBrain-Bench，以评估模型是否能够通过联合文本和视觉组件而非单独理解或生成来表示具身计划。实验表明，RxBrain保持了具身理解和生成能力，并生成了具有耦合文本推理、世界状态预测和联合子目标规划的计划。我们还将RxBrain扩展到连续机器人动作生成，显示出在没有大规模动作数据预训练的情况下，良好的真实机器人表现。这些结果为具身认知的基础模型迈出了初步的一步。
cs.AI / 16 / 2607.14197
How Artificial Intelligence LLM Engines Shape the Global Conflict Information Environment
人工智能大型语言模型引擎如何塑造全球冲突信息环境
Abstract
Artificial Intelligence (AI) answer engines now field a growing share of the questions that analysts, scholars, and the public ask about issues of peace and conflict. Large Language Models (LLMs) are known to hallucinate under certain conditions, but do these errors have discernible patterns when they are asked about conflicts, and if so what can that teach us about the changing global conflict information environment? To answer, we first asked a battery of questions about 28 conflicts to five leading answer engines and scored their 5,460 answers against documented evidence. We found that the thinner the retrievable record around a given conflict, the more the engines invent, misattribute, and miscount. Thin records don't just encourage hallucination, but create structural exposure to mis- and disinformation, because they are the easiest records to warp through Generative Engine Optimization (GEO) to bias engine responses. Through an analysis of 1,048 websites that the AI LLMs pulled conflict facts from, we found that GEO source optimization is already happening, and while state-partisan digital capture remains incipient it is rapidly growing. We explain what these findings mean for scholarship with the rise of GEO information warfare, and for policy argue for a return to the deep local monitoring and translation-based research that AI tools cannot replicate, closing with a discussion of future research opportunities and challenges in this fast-moving space.
Chinese Translation
人工智能（AI）问答引擎现在在分析师、学者和公众关于和平与冲突问题的提问中占据了越来越大的份额。大型语言模型（LLMs）在某些条件下被发现会产生幻觉，但当它们被询问关于冲突的问题时，这些错误是否存在可辨识的模式？如果存在，这能教会我们关于不断变化的全球冲突信息环境什么？为了解答这个问题，我们首先向五个领先的问答引擎提出了关于28个冲突的一系列问题，并根据文献证据对其5460个回答进行了评分。我们发现，关于某个特定冲突的可检索记录越薄弱，问答引擎就越容易进行虚构、错误归因和错误计数。薄弱的记录不仅鼓励幻觉的产生，还使其在错误和虚假信息方面暴露出结构性风险，因为这些记录最容易通过生成引擎优化（GEO）来扭曲，从而影响引擎的响应。通过对AI LLMs提取冲突事实的1048个网站的分析，我们发现GEO源优化已经在发生，尽管国家偏向的数字捕获仍处于初期阶段，但其增长迅速。我们解释了这些发现对学术研究的意义，特别是在GEO信息战兴起的背景下，并在政策方面主张恢复深度地方监测和基于翻译的研究，这些是AI工具无法复制的，最后讨论了在这一快速发展的领域中的未来研究机会和挑战。
cs.AI / 17 / 2607.14240
Align AI to Dynamic Human-AI Workflows
将人工智能与动态人机工作流程对齐
Abstract
Current alignment approaches typically focus on emulating human behavior using static representations of human preferences, failing to capture the dynamic, context-dependent nature of real-world human-AI interactions. In this paper, we argue for a shift from static and emulative to interactive and complementary alignment, where preferences emerge through interaction and alignment is defined not by satisfying preferences alone. We first formalize this gap by contrasting existing alignment with a trajectory-level view in which human and model behavior co-evolve over time. Because these interaction dynamics have not been adequately captured within existing ML formulations, we ground this perspective in insights from an interdisciplinary workshop. We draw on lessons from social-science accounts of human-human collaboration and then argue that human-AI systems amplify these dynamics, introducing new asymmetries that make reasoning about uncertainty harder and introduce new coordination challenges. Based on these lessons and new challenges, we conclude by outlining a research agenda for developing AI systems that align with humans in interaction, requiring an interdisciplinary synthesis of machine learning and the social and decision sciences.
Chinese Translation
当前的对齐方法通常侧重于使用静态的人类偏好表示来模拟人类行为，未能捕捉到现实世界人机交互的动态和依赖上下文的特性。本文主张从静态和模仿的对齐转向互动和互补的对齐，其中偏好通过互动而产生，对齐不仅仅是通过满足偏好来定义的。我们首先通过对比现有的对齐方法与一种轨迹级别的视角来正式化这一差距，在这种视角中，人类和模型的行为随着时间共同演变。由于这些互动动态在现有的机器学习（ML）模型中尚未得到充分捕捉，我们将这一视角建立在跨学科研讨会的见解之上。我们借鉴了社会科学中关于人际合作的经验教训，并进一步论证人机系统放大了这些动态，引入了新的不对称性，使得对不确定性的推理变得更加困难，并带来了新的协调挑战。基于这些经验教训和新挑战，我们总结了一个研究议程，旨在开发与人类互动对齐的人工智能系统，这需要机器学习与社会科学和决策科学的跨学科综合。
cs.AI / 18 / 2607.14246
The Steering Budget: Examples beat Knobs
引导预算：示例胜过旋钮
Abstract
Generative models are steered with knobs -- prompts, guidance scales, property tags. Turn one as hard as you like and, past a point, it stops moving the property you care about. We find that ceiling is not a shortcoming of the model but a budget, set by the training data before the model is trained: a property's movable range splits in two -- the part a knob can reach, and a second, significant part that only examples -- concrete instances of what you want more of -- can reach. That second part is usually much larger, but not always, and the same budget says so in advance. Reaching that second part takes a different move: instead of turning a knob, you show the model examples, composed from what it already learned rather than added to its training. A cheap audit of the training data measures the budget; we give a recipe for building the example set that reaches all of it. This buys two things a knob can't. Reach: it moves a property across the whole budget, not just the part a knob reaches. Expressiveness: it steers toward targets you can only specify by example -- including ones you can't put into words. We turn these into a handful of falsifiable claims and verify them in two unrelated domains, image and crystal-structure generation -- marking where a knob is enough, and where only examples will do.
Chinese Translation
生成模型通过旋钮进行引导——提示、引导尺度、属性标签。你可以尽可能地旋转一个旋钮，但在某个点之后，它将不再影响你关心的属性。我们发现，这一上限并不是模型的缺陷，而是由训练数据在模型训练之前设定的预算：一个属性的可移动范围分为两部分——旋钮可以达到的部分，以及仅有示例（你希望更多的具体实例）可以达到的第二部分。第二部分通常要大得多，但并不总是如此，而同样的预算提前就已经说明了这一点。达到第二部分需要采取不同的方式：不是旋转旋钮，而是向模型展示示例，这些示例是由它已经学习的内容组成，而不是添加到其训练中。对训练数据进行简单审计可以测量预算；我们提供了一种构建示例集的方案，以覆盖所有预算。这带来了旋钮无法实现的两个方面。覆盖范围：它在整个预算内移动属性，而不仅仅是旋钮所能达到的部分。表现力：它引导向只能通过示例指定的目标——包括那些无法用语言表达的目标。我们将这些转化为一系列可验证的主张，并在两个不相关的领域（图像生成和晶体结构生成）中验证它们——标记出旋钮足够的地方，以及只有示例才能满足的地方。
cs.AI / 19 / 2607.14256
Automatic Hard Example Synthesis with Multi-Level Agentic Data Curation
自动化困难示例合成与多层次代理数据管理
Abstract
Multimodal Large Language Models (MLLMs) are increasingly deployed for nuanced content safety and moderation tasks, yet they remain vulnerable to adversarial attacks and out-of-distribution edge cases. Traditional active learning and manual annotation fail to scale against the complexity and volume of novel multimodal threats. In this paper, we propose an automated, agentic red-teaming framework that systematically synthesizes difficult examples using an iterative strategy that proposes novel hypotheses as well as mutating on past attempts. Leveraging a multi-agent architecture that consists of a high-reasoning Architect agent, an advanced image generator, and a multi-level verification committee of LLM raters, our system autonomously uncovers boundary-pushing violations and ambiguous policy edge cases without any human intervention. By employing these carefully synthesized adversarial examples as in-context demonstrations via test-time Retrieval, we substantially improve the target model's robustness, reducing the False Negative Rate (FNR) from 41.2% to 24.5% in a public image safety benchmark without relying on any human labeling.
Chinese Translation
多模态大型语言模型（MLLMs）越来越多地用于细致的内容安全和审核任务，但它们仍然容易受到对抗性攻击和分布外边缘案例的影响。传统的主动学习和手动标注在面对新型多模态威胁的复杂性和数量时难以扩展。在本文中，我们提出了一种自动化的代理红队框架，该框架通过一种迭代策略系统地合成困难示例，提出新假设并对过去的尝试进行变异。利用由高推理能力的Architect代理、先进的图像生成器以及多层次的LLM评分委员会组成的多代理架构，我们的系统能够在没有任何人工干预的情况下自主发现突破性违规和模糊的政策边缘案例。通过在测试时检索（test-time Retrieval）中使用这些精心合成的对抗性示例作为上下文演示，我们显著提高了目标模型的鲁棒性，将公共图像安全基准中的假阴性率（FNR）从41.2%降低到24.5%，而无需依赖任何人工标注。
cs.AI / 20 / 2607.14275
AI Agents Do Not Fail Alone:The Context Fails First
人工智能代理并非单独失败：上下文首先失败
Abstract
Context engineering has become central to building reliable AI agents, yet it remains largely unmeasured. Agents do not fail in isolation: their behavior is shaped by the instructions, tools, memory, retrieved knowledge, guardrails, and untrusted inputs accumulated in their context. When this context is weak, agents drift, hallucinate, misuse tools, ignore constraints, become vulnerable to injection, and waste tokens. This paper validates context-engineering quality as an independent leading indicator of agent reliability. We implement the measurement in ProofAgent-Harness, an open-source infrastructure for AI agent evaluation that uses multi-juror, consensus-based scoring. The harness assesses context across seven criteria: role clarity, guardrail coverage, instruction consistency, tool schema quality, grounding sufficiency, injection hardening, and token efficiency. Crucially, the context score is isolated from behavioral metrics and release decisions, enabling a non-circular validation. Through a controlled context-quality study across regulated agent domains, holding frontier LLM agents fixed and varying only their operating context, we show that context-quality criteria consistently predict their corresponding behavioral outcomes. Grounding sufficiency predicts hallucination resistance, guardrail coverage predicts manipulation resistance, instruction consistency predicts instruction following, and tool-schema quality predicts tool use. These findings establish context measurement as a validated preflight signal for agent reliability and position context engineering as an auditable layer of agent evaluation and governance.
Chinese Translation
上下文工程已成为构建可靠人工智能代理的核心，但仍然在很大程度上未被量化。代理并非孤立失败：它们的行为受到指令、工具、记忆、检索知识、保护措施和在其上下文中积累的不可信输入的影响。当这种上下文薄弱时，代理会偏离轨道、产生幻觉、误用工具、忽视约束、变得容易受到注入攻击，并浪费代币。本文验证了上下文工程质量作为代理可靠性的独立领先指标。我们在ProofAgent-Harness中实施了这一测量，该工具是一个用于人工智能代理评估的开源基础设施，采用多评审员共识评分。该工具通过七个标准评估上下文：角色清晰度、保护措施覆盖、指令一致性、工具模式质量、基础支持充分性、注入强化和代币效率。重要的是，上下文评分与行为指标和发布决策相互独立，从而实现非循环验证。通过在受监管的代理领域进行的受控上下文质量研究，我们固定前沿的LLM代理，仅改变其操作上下文，结果表明上下文质量标准始终能够预测其相应的行为结果。基础支持充分性预测幻觉抵抗力，保护措施覆盖预测操控抵抗力，指令一致性预测指令遵循，工具模式质量预测工具使用。这些发现确立了上下文测量作为代理可靠性的验证前信号，并将上下文工程定位为代理评估和治理的可审计层。
cs.AI / 21 / 2607.14301
Measuring How Students Rely on Generative AI in Academic Writing: Development and Multi-Source Validation of the Generative AI Reliance Types Scale (GenAI-RTS)
测量学生在学术写作中对生成性人工智能的依赖：生成性人工智能依赖类型量表（GenAI-RTS）的开发与多源验证
Abstract
As generative AI (GenAI) becomes increasingly embedded in undergraduate academic writing, how students rely on these tools, rather than simply whether they use them, has become a central question for learning, academic integrity, and educational equity. Existing measures of reliance were developed inductively, focused on discrete problem-solving tasks, and validated mainly with homogeneous samples. This study developed and validated the GenAI Reliance Types Scale (GenAI-RTS), a 20-item instrument measuring four theoretically derived types of GenAI reliance: Strategic, Instrumental, Dependent, and Dialogic. Validation followed the multisource framework of the Standards for Educational and Psychological Testing, drawing on a survey of 382 undergraduates at a U.S. Minority-Serving Institution and interviews with 14 purposively sampled students. Confirmatory factor analyses of six competing models supported a five-factor structure in which Strategic Reliance comprises two facets, Deliberate Use and Critical Evaluation, alongside Instrumental, Dependent, and Dialogic factors (CFI = .92, RMSEA = .08; DWLS CFI = .98, RMSEA = .07). Subscale reliability was acceptable to good (omega = .75-.88), and scalar measurement invariance held across gender, first-generation status, and STEM/non-STEM majors, to our knowledge the first such evidence for a GenAI reliance instrument. Rasch analysis indicated that a five-point response format would improve category functioning. Strategic reliance was positively associated with AI literacy, and the reliance types differentiated students across multiple writing process and outcome variables. The GenAI-RTS offers researchers and educators a theoretically grounded, psychometrically validated instrument for identifying undergraduate reliance profiles and supporting research, assessment, and AI literacy intervention.
Chinese Translation
随着生成性人工智能（GenAI）在本科生学术写作中的日益普及，学生如何依赖这些工具，而不仅仅是是否使用它们，已成为学习、学术诚信和教育公平的核心问题。现有的依赖测量工具主要是通过归纳法开发的，集中于离散的解决问题任务，并且主要在同质样本中进行验证。本研究开发并验证了生成性人工智能依赖类型量表（GenAI-RTS），这是一个包含20个条目的工具，用于测量四种理论上推导的生成性人工智能依赖类型：战略性、工具性、依赖性和对话性。验证遵循教育和心理测试标准的多源框架，基于对美国一所少数族裔服务机构的382名本科生的调查以及与14名有目的抽样学生的访谈。对六个竞争模型的确认性因子分析支持了一个五因子结构，其中战略性依赖包含两个方面：深思熟虑的使用和批判性评估，以及工具性、依赖性和对话性因素（CFI = .92，RMSEA = .08；DWLS CFI = .98，RMSEA = .07）。各子量表的可靠性从可接受到良好（omega = .75-.88），并且在性别、第一代学生身份和STEM/非STEM专业之间保持了标量测量不变性，这是我们所知的生成性人工智能依赖工具的首次此类证据。Rasch分析表明，五点响应格式将改善类别功能。战略性依赖与人工智能素养呈正相关，且依赖类型在多个写作过程和结果变量中区分了学生。GenAI-RTS为研究人员和教育工作者提供了一个理论基础扎实、心理测量学验证的工具，以识别本科生的依赖特征，并支持研究、评估和人工智能素养干预。
cs.AI / 22 / 2607.14306
Tracing LLM Behavior to the Training Data with Empirical Next-Token Distributions
追踪大型语言模型行为与训练数据之间的关系：基于经验下一个标记分布的研究
Abstract
In this paper, we study the connection between an LLM's output distribution and the data used to train it. Specifically, we study the degree to which an LLM's next-token distribution agrees with the empirical next-token distribution (ENTD) given the context in the training data. The ENTD is an appealing target because it is the unrestricted global minimizer of the next-token cross entropy loss used for pretraining, as well as an easily interpretable function of the pretraining corpus. We find that for a significant fraction of inputs, the LLM's distribution agrees with the ENTD almost perfectly, and the average agreement increases with model scale and training compute. Nevertheless, there is a long tail of input sequences where the LLM and ENTD differ significantly, and we examine several possible sources of this discrepancy across the transformer architecture, training procedure, and finite-sample noise in the ENTD estimate itself. More broadly, we hope our findings will encourage more work on ``data-centric mechanistic interpretability,'' a complement to standard mechanistic interpretability that opens the black box of how model behaviors arise from the data, rather than how they are encoded in the learned weights.
Chinese Translation
在本文中，我们研究了大型语言模型（LLM）的输出分布与用于训练其数据之间的联系。具体而言，我们研究了LLM的下一个标记分布在多大程度上与给定训练数据上下文的经验下一个标记分布（ENTD）一致。ENTD是一个引人注目的目标，因为它是用于预训练的下一个标记交叉熵损失的无限制全局最小化器，并且是预训练语料库的一个易于解释的函数。我们发现，对于相当一部分输入，LLM的分布几乎与ENTD完全一致，并且平均一致性随着模型规模和训练计算的增加而提高。然而，仍然存在一长串输入序列，其中LLM与ENTD存在显著差异，我们检查了这种差异可能的来源，包括变换器架构、训练过程以及ENTD估计中的有限样本噪声。更广泛地说，我们希望我们的发现能够鼓励更多关于“以数据为中心的机制可解释性”的研究，这是一种与标准机制可解释性互补的方法，旨在揭示模型行为如何源于数据，而不是如何在学习的权重中编码。
cs.AI / 23 / 2607.14309
Traccia: An OpenTelemetry-Based Governance Platform for AI Systems
Traccia：基于OpenTelemetry的人工智能系统治理平台
Abstract
The rapid development of Large Language Models (LLMs) and Artificial Intelligent (AI) powered autonomous agents has fundamentally changed the existing forms of software governance. In spite of the rigorous standards of transparency and account ability required according to the international frameworks such as the European Union's AI Act, there is a considerable gap between theory and reality. The present study discusses the inherent drawbacks of currently utilized platforms for LLM evaluation, machine learning workflow, and application performance monitoring in general. It has been shown that current disjointed solutions fail to protect unbound state space agentic architecture from serious threats such as alignment drift, SaaS security concerns, and unauthorized deployment of shadow AI systems. Moreover, a solution is proposed for overcoming the discussed challenges in form of a coherent multi-level AI governance stack Traccia built on the top of OpenTelemetry infrastructure platform. Traccia resolves the last mile for AI Alignment by adding the telemetry data, passive semantic guardrail assessment, and execution lineage into a hashed trace ledger. Traccia automatically creates compliance evidence packages by appending tamper-resistant fingerprints and SHA-256 content hash, that map to regulatory requirements (Articles 12, 14, 19, 26(6), and 50 of the EU AI Act) without invading any data privacy. By performing this evaluation in a methodical manner, a solid machine-readable base has been created for enterprise-wide management of autonomous AI systems.
Chinese Translation
大型语言模型（LLMs）和人工智能（AI）驱动的自主代理的快速发展，已经从根本上改变了现有的软件治理形式。尽管根据国际框架（如欧盟的AI法案）要求严格的透明度和问责制标准，但理论与现实之间仍存在相当大的差距。本研究讨论了当前用于LLM评估、机器学习工作流和应用性能监控的平台的固有缺陷。研究表明，当前的分散解决方案未能保护无约束状态空间的代理架构免受对齐漂移、SaaS安全问题和未授权部署影子AI系统等严重威胁。此外，提出了一种解决上述挑战的方案，即基于OpenTelemetry基础设施平台构建的连贯多层AI治理堆栈Traccia。Traccia通过将遥测数据、被动语义保护评估和执行谱系添加到哈希追踪账本中，解决了AI对齐的最后一公里问题。Traccia通过附加防篡改的指纹和SHA-256内容哈希，自动创建合规证据包，这些包映射到监管要求（欧盟AI法案第12、14、19、26(6)和50条）而不侵犯任何数据隐私。通过以系统化的方式进行评估，为企业范围内的自主AI系统管理创建了一个坚实的机器可读基础。
cs.AI / 24 / 2607.14386
CIPHER: A Decoupled Exploration-Selection Framework for Test-Time Scaling of Data Science Agents
CIPHER：一种用于数据科学代理测试时间扩展的解耦探索-选择框架
Abstract
Data science tasks span from closed-ended information extraction to open-ended analysis, presenting significant challenges for automation. Recent AI agents powered by language models show promise for handling such complex tasks. However, existing agents typically rely on a single initial state that conditions the entire agent's execution, making them vulnerable to cascading errors initiated by a suboptimal initial state. To mitigate this, we present CIPHER, an automated data science agent that leverages test-time scaling through the generation and selection of multiple initial states for concurrent execution. Unlike existing works on test-time scaling of AI agents, CIPHER explicitly decouples the generation of candidate initial states from their strategic selection for parallel execution. Through extensive evaluation on two benchmarks (closed-form and open-form tasks), we demonstrate that CIPHER exceeds state-of-the-art performance in matched-model comparisons, and remains competitive against larger-model baselines despite relying on a substantially smaller base LM. Our empirical study characterizes the design space of the Decoupled Exploration-Selection (DES) framework: we quantify how generation strategy, selection strategy, and aggregator model capacity contribute to overall performance, and derive actionable design recommendations for practitioners.
Chinese Translation
数据科学任务涵盖从封闭式信息提取到开放式分析，给自动化带来了重大挑战。近期由语言模型驱动的人工智能代理在处理此类复杂任务方面显示出良好的前景。然而，现有代理通常依赖于单一的初始状态，这一状态决定了整个代理的执行，使其容易受到由次优初始状态引发的级联错误的影响。为此，我们提出了CIPHER，这是一种自动化数据科学代理，通过生成和选择多个初始状态以进行并行执行，从而利用测试时间扩展。与现有关于人工智能代理测试时间扩展的研究不同，CIPHER明确将候选初始状态的生成与其战略选择解耦，以实现并行执行。通过在两个基准（封闭式和开放式任务）上的广泛评估，我们证明CIPHER在匹配模型比较中超越了最先进的性能，并且尽管依赖于显著较小的基础语言模型，仍然在与更大模型基线的竞争中保持竞争力。我们的实证研究描述了解耦探索-选择（DES）框架的设计空间：我们量化了生成策略、选择策略和聚合模型容量如何影响整体性能，并为从业者提供了可行的设计建议。
cs.AI / 25 / 2607.14387
Chat2Scenic: An Iterative RAG-Based Framework for Scenario Generation in Autonomous Driving
Chat2Scenic：一种基于迭代检索增强的自主驾驶场景生成框架
Abstract
Validating autonomous driving systems requires diverse, regulation-compliant test scenarios. In simulation-based testing, scenarios are defined as executable scripts. Yet automatically generating such scripts from regulatory descriptions remains an open challenge, and existing approaches face fundamental trade-offs. Retrieval-assemble methods achieve reasonable compilation rates but lack scalability, whereas retrieval-based full-script generation suffers from low compilation success rates. We present Chat2Scenic, the first iterative retrieval-augmented framework to generate scenario scripts in Domain Specific Language (DSL). Specifically, Chat2Scenic provides a chatbot interface that supports interactive scenario refinement and integrates Retrieval-augmented Generation (RAG) to ground scenario generation in regulatory knowledge and DSL syntax. Furthermore, we propose an open benchmark for scenario generation comprising 123 scenarios from various regulations, including NHTSA and United Nations Vehicle Regulations, as well as other sources. Extensive evaluation with State-of-the-Art (SOTA) Large Language Models (LLMs) demonstrates that Chat2Scenic achieves 76.42% Compilation Success Rate (CSR) and 58.17% Framework Accuracy (FA), outperforming existing methods (Retrieval Assemble with 30.08% CSR, 11.03% FA and Retrieval full script generation with 16.26% CSR, 10.86% FA). To facilitate future research, we release our code as open source at https://github.com/TUM-AVS/chat2scenic.
Chinese Translation
验证自主驾驶系统需要多样化且符合规定的测试场景。在基于仿真的测试中，场景被定义为可执行脚本。然而，从监管描述中自动生成此类脚本仍然是一个未解决的挑战，现有方法面临基本的权衡。检索-组装方法实现了合理的编译率，但缺乏可扩展性，而基于检索的完整脚本生成则遭遇低编译成功率。我们提出了Chat2Scenic，这是第一个迭代检索增强框架，用于生成领域特定语言（Domain Specific Language, DSL）中的场景脚本。具体而言，Chat2Scenic提供了一个聊天机器人接口，支持交互式场景细化，并整合了检索增强生成（Retrieval-augmented Generation, RAG），以将场景生成基于监管知识和DSL语法。此外，我们提出了一个开放的场景生成基准，包含来自各种法规的123个场景，包括美国国家公路交通安全管理局（NHTSA）和联合国车辆法规，以及其他来源。与最先进的大型语言模型（State-of-the-Art Large Language Models, SOTA LLMs）的广泛评估表明，Chat2Scenic实现了76.42%的编译成功率（Compilation Success Rate, CSR）和58.17%的框架准确率（Framework Accuracy, FA），优于现有方法（检索组装的CSR为30.08%，FA为11.03%；检索完整脚本生成的CSR为16.26%，FA为10.86%）。为了促进未来的研究，我们将代码作为开源发布，地址为https://github.com/TUM-AVS/chat2scenic。
cs.AI / 26 / 2607.14391
A Comparative Analysis of Machine Learning Models for Long and Short-Term Forecasting of the Egyptian Stock Market: A Focus on EGX30
埃及股市长期和短期预测的机器学习模型比较分析：聚焦于EGX30
Abstract
This study concentrates on predicting stock prices in the Egyptian market, focusing on the EGX30, an influential financial hub in the Middle East. While most research focuses on global stocks, there's a growing need to understand stock trends in developing countries like Egypt. The study compares different machine learning models for forecasting EGX30 trends, covering short and long-term predictions. Using historical EGX30 data, including metrics like root mean squared error, Mean Absolute Percentage Error, and coefficient of determination, models like K-Nearest Neighbours, random forest, extreme gradient boosting, long short-term memory networks, and gated recurrent unit networks were evaluated. The goal is to determine the most effective models for EGX30 prediction, considering Egypt's unique market dynamics. Insights from this study aid investors in making informed decisions. Results show that the Gated Recurrent Unit (GRU) outperformed the other models in the one-week, one-month, and two-months while the eXtreme Gradient Boosting (XGBoost) model outperformed others in the one-day predictions, highlighting their usefulness in predictive analysis for financial markets. The study also showed the importance of using the ensemble techniques, especially in the long-term predictions which proved better results reaching 5 times the GRU in the two-month predictions. Additionally, the study notes the surprisingly good performance of K-Nearest Neighbours (KNN) on long-term predictions, suggesting its enduring relevance and potential for future applications in the fintech domains.
Chinese Translation
本研究集中于预测埃及市场的股票价格，重点关注EGX30，这是中东地区一个重要的金融中心。尽管大多数研究集中于全球股票，但理解像埃及这样的发展中国家的股票趋势的需求日益增长。该研究比较了不同的机器学习模型用于预测EGX30的趋势，涵盖短期和长期预测。通过使用历史EGX30数据，包括均方根误差、平均绝对百分比误差和决定系数等指标，评估了K-近邻（K-Nearest Neighbours）、随机森林（Random Forest）、极端梯度提升（Extreme Gradient Boosting）、长短期记忆网络（Long Short-Term Memory Networks）和门控循环单元网络（Gated Recurrent Unit Networks）等模型。研究的目标是确定最有效的EGX30预测模型，考虑到埃及独特的市场动态。本研究的见解有助于投资者做出明智的决策。结果表明，在一周、一个月和两个月的预测中，门控循环单元（Gated Recurrent Unit, GRU）模型的表现优于其他模型，而极端梯度提升（eXtreme Gradient Boosting, XGBoost）模型在一天的预测中表现最佳，突显了它们在金融市场预测分析中的实用性。研究还显示了使用集成技术的重要性，特别是在长期预测中，结果显示在两个月的预测中，集成技术的效果达到了GRU的5倍。此外，研究指出K-近邻（K-Nearest Neighbours, KNN）在长期预测中的表现出乎意料地良好，暗示其在金融科技领域未来应用的持久相关性和潜力。
cs.AI / 27 / 2607.14396
CatalogAgent: A Supervisor-mediated Self-Learning System Enabling Context Engineering for GenAI Models
CatalogAgent：一种通过监督介导的自学习系统，支持生成式人工智能模型的上下文工程
Abstract
Product catalogs are the backbone of e-commerce sites, yet a large number of structured attributes (SAs) -- such as material, color, and shape -- often have missing values. Typically, SA values are extracted from product information, including titles and descriptions. While LLM-based generator-evaluator frameworks have demonstrated effectiveness for SA prediction -- where an LLM generates SA values and another evaluates them -- they face challenges when the Generator and Evaluator produce conflicting outputs, as either component can make mistakes. We introduce \texttt{CatalogAgent}, a novel agentic system that continuously improves Generator and Evaluator models for e-commerce catalog enrichment. When disagreements arise from (1) internal conflicts between the LLM-based Generator and Evaluator, or (2) external feedback from sellers on LLM outputs, a Supervisor Agent intervenes to mediate these conflicts and make final decisions. The system also incorporates a Memory Base and a Memory Summarizer that stores Supervisor Agent activities from individual cases and aggregates patterns into learnings. These learnings are fed back to the worker Generator and Evaluator LLMs, enabling self-improvement without human intervention. Through context engineering -- injecting learnings and insights into worker LLMs' contexts -- the system successfully transfers the Supervisor's capabilities to the Generator and Evaluator, improving their performance by 15.24\% and 13.98\%, respectively. Our experiments demonstrate a new paradigm of Supervisor Agent-mediated self-learning systems for improving generative AI model accuracy.
Chinese Translation
产品目录是电子商务网站的核心，但大量的结构化属性（SAs）——如材质、颜色和形状——往往存在缺失值。通常，SA值是从产品信息中提取的，包括标题和描述。虽然基于大型语言模型（LLM）的生成-评估框架在SA预测中表现出有效性——其中一个LLM生成SA值，另一个进行评估——但当生成器和评估器产生冲突输出时，它们面临挑战，因为任一组件都可能出错。我们提出了 exttt{CatalogAgent}，一种新颖的代理系统，能够持续改进电子商务目录丰富化的生成器和评估器模型。当出现（1）基于LLM的生成器和评估器之间的内部冲突，或（2）卖家对LLM输出的外部反馈时，监督代理介入以调解这些冲突并做出最终决策。该系统还结合了一个记忆库和一个记忆摘要器，存储监督代理在个案中的活动，并将模式汇总为学习。这些学习反馈给工作生成器和评估器LLM，使其能够在没有人工干预的情况下自我改进。通过上下文工程——将学习和洞察注入工作LLM的上下文——该系统成功地将监督者的能力转移到生成器和评估器上，分别提高了其性能15.24%和13.98%。我们的实验展示了一种新的监督代理介导的自学习系统范式，以提高生成式人工智能模型的准确性。
cs.AI / 28 / 2607.14399
Instrument Effects in Language-Model Honesty Evaluation: An Auditable Single-System Demonstration
语言模型诚实性评估中的工具效应：可审计的单系统演示
Abstract
Evaluations of language-model honesty read the model's verdicts as evidence about the model. We test the instrument instead. We built a text-adventure world where the game engine, not any model, knows whether the quest can be completed. A language model plays under a budget and must eventually declare its quest complete, unreachable, or not yet decidable; the engine scores every verdict. Decision rules were recorded before results were read, and run artifacts bind the revisions they executed; the strength of preregistration varies by series and is disclosed. With the player held fixed, instrument choices substantially changed measured behavior. On four byte-identical anchors, expanding a two-verdict grammar to three verdicts moved strong claims from 38/40 to 7/40, while the new incomplete verdict took 28/40 outcomes; across series 2, 93/158 valid games ended incomplete. One sentence disclosing the success criterion took matched-instance false verdicts from 18/59 to 0/58, through fewer decision points and cleaner decisions. Repeated runs of one fixed configuration produced non-stable verdict distributions on 3 of 4 instances: single runs report samples as dispositions. A formally preregistered narrative-register gradient was falsified; two post-hoc, hypothesis-generating patterns remain: register presence roughly doubled strong claims, and budget rendering moved verdicts more than register content (.383 meter vs .150 lantern). The narrator compressed abundant budgets toward scarcity landmarks, yet the registered mediation test returned a null. We propose a four-check integrity protocol for eval instruments.
Chinese Translation
语言模型诚实性的评估将模型的判决视为关于模型的证据。我们则测试工具本身。我们构建了一个文本冒险世界，其中游戏引擎，而非任何模型，知道任务是否可以完成。语言模型在预算限制下进行游戏，最终必须声明其任务是完成、不可达还是尚未可决；引擎对每个判决进行评分。决策规则在结果读取之前被记录，运行结果绑定了它们执行的修订；预注册的强度因系列而异，并已披露。在玩家固定的情况下，工具选择显著改变了测量行为。在四个字节相同的锚点上，将两个判决的语法扩展到三个判决，使得强声明的比例从38/40降至7/40，而新的不完整判决占28/40的结果；在系列2中，93/158个有效游戏以不完整结束。一句披露成功标准的句子将匹配实例的错误判决从18/59降至0/58，通过减少决策点和更清晰的决策。对一个固定配置的重复运行在4个实例中有3个产生了不稳定的判决分布：单次运行将样本报告为倾向。一个正式预注册的叙事注册梯度被证伪；两个事后生成假设的模式仍然存在：注册存在大约使强声明翻倍，而预算呈现对判决的影响超过注册内容（0.383米对0.150灯笼）。叙述者将丰富的预算压缩到稀缺的地标，但注册的中介测试返回了零结果。我们提出了一种针对评估工具的四项完整性检查协议。
cs.AI / 29 / 2607.14408
Reward-Free Evolving Agents via Pairwise Validator
通过成对验证器实现无奖励自我进化代理
Abstract
A self-evolving agentic loop repeatedly proposes a tweaked version of an agent (its prompt template or program) and accepts or rejects the change based on a per-iteration quality signal. Designing that signal is often the costly part of the project: a reliable scalar reward requires domain expertise and labeled examples that are themselves as expensive to assemble as the agent's underlying task. We propose replacing the scalar at the accept/reject gate with a pairwise validator: a frozen LLM that, given the parent and child candidate, returns a binary verdict on which is better. Pairwise judgment is generally easier and more stable than absolute scoring, due to its contrastive nature, which mitigates the need for strict scale calibration. The validator also requires no training of its own. We integrate the validator into three published self-evolving engines (GEPA, ADRS, ShinkaEvolve) and report two flavors: Adaptive Focus, which retains the engine's existing val-set parent selection, and Soft Elo, which lets the validator's verdicts drive parent selection so that val-set rewards drop as well. Across multiple agents and two artifact substrates (prompt and code), our method matches or exceeds the full-reward baseline on the majority of settings we evaluate, and the pattern survives a cross-family validator swap. The pairwise gate is thus a drop-in replacement for per-step reward design at competitive task accuracy without the labeling cost.
Chinese Translation
自我进化的代理循环反复提出代理的修改版本（其提示模板或程序），并根据每次迭代的质量信号接受或拒绝该更改。设计该信号通常是项目中最耗费成本的部分：可靠的标量奖励需要领域专业知识和标记示例，而这些示例的组装成本与代理的基础任务同样高。我们建议用成对验证器替代接受/拒绝门的标量：一个冻结的语言模型（LLM），在给定父代和子代候选者的情况下，返回哪个更好的二元判决。由于其对比性质，成对判断通常比绝对评分更容易且更稳定，这减轻了对严格尺度校准的需求。验证器本身也不需要训练。我们将验证器集成到三个已发布的自我进化引擎（GEPA、ADRS、ShinkaEvolve）中，并报告两种变体：自适应聚焦（Adaptive Focus），保留引擎现有的验证集父代选择，以及软Elo（Soft Elo），让验证器的判决驱动父代选择，从而验证集奖励也随之下降。在多个代理和两种工件基底（提示和代码）上，我们的方法在我们评估的大多数设置中与全奖励基线相匹配或超出，并且这一模式在跨家族验证器替换中依然有效。因此，成对门可以作为每步奖励设计的直接替代，以竞争的任务准确性而无需标记成本。
cs.AI / 30 / 2607.14416
CausalGraphX: A Counterfactual Graph Neural Network Framework for Explainable Systemic Risk Assessment
CausalGraphX：用于可解释系统性风险评估的反事实图神经网络框架
Abstract
The interconnected nature of global financial systems makes them vulnerable to systemic risks, where the failure of a few institutions can trigger catastrophic cascading defaults. Traditional risk models often fail to capture the complex, non-linear dynamics of these networks. While Graph Neural Networks (GNNs) have shown promise in modeling relational data, they primarily learn correlative patterns and function as black boxes, offering little insight into the causal mechanisms of shock propagation. This limitation is critical for regulators who require explainable models to perform stress tests and devise effective interventions. We introduce CausalGraphX, a novel framework that integrates GNNs with counterfactual reasoning to provide explainable assessments of systemic risk. CausalGraphX employs a Graph Attention mechanism to learn representations of institutional vulnerability and uses an adversarial regularization technique to ensure these representations capture causal drivers rather than spurious correlations. Furthermore, we propose an optimization-based approach to generate counterfactual explanations, answering questions such as, "What minimum capital injection would have prevented Bank A's default under a specific stress scenario?" We validate CausalGraphX on large-scale synthetic financial networks. Our results demonstrate that CausalGraphX significantly outperforms traditional and deep learning baselines in predicting cascading defaults while providing sparse, plausible, and actionable counterfactual explanations.
Chinese Translation
全球金融系统的相互关联性使其易受系统性风险的影响，其中少数机构的失败可能引发灾难性的级联违约。传统风险模型往往无法捕捉这些网络的复杂非线性动态。尽管图神经网络（Graph Neural Networks, GNNs）在建模关系数据方面显示出潜力，但它们主要学习相关模式，并作为黑箱运作，几乎无法提供关于冲击传播因果机制的洞见。这一局限性对于需要可解释模型以进行压力测试和制定有效干预措施的监管机构来说至关重要。我们提出了CausalGraphX，这是一种新颖的框架，将GNN与反事实推理相结合，以提供可解释的系统性风险评估。CausalGraphX采用图注意力机制学习机构脆弱性的表示，并使用对抗正则化技术确保这些表示捕捉因果驱动因素而非虚假相关。此外，我们提出了一种基于优化的方法来生成反事实解释，回答诸如“在特定压力情景下，最低的资本注入是多少才能防止银行A的违约？”这样的问题。我们在大规模合成金融网络上验证了CausalGraphX。结果表明，CausalGraphX在预测级联违约方面显著优于传统和深度学习基线，同时提供稀疏、合理且可操作的反事实解释。
cs.AI / 31 / 2607.14427
Per-Token Fixed-Point Convergence in Depth-Recurrent Transformers
深度递归变换器中的每个标记固定点收敛
Abstract
A depth-recurrent transformer applies a weight-tied core a variable number of times, and prior work has shown that training with a randomized recursion count yields one checkpoint usable across a range of inference depths. We ask what such a model actually computes per token, and measure it directly. On a 135M-class model trained on FineWeb-Edu, the recurrent state converges to a per-token fixed point: mean successive-output KL divergence falls from 3.9e-1 at the second loop to 8.5e-6 by the sixteenth, and per-token state change decays in step. Crucially, this convergence is not uniform across tokens. The median token converges by loop six, while approximately 10 percent of tokens continue to update at the training-mean depth of eight, and mean convergence depth is ordered by token type (whitespace shallowest, content words deepest). This per-token variation is the central object of the paper. We show it is directly readable and that reading it outperforms learning to predict it: a training-free rule that halts each token once its output stabilizes attains uniform depth-8 quality at 4.94 average loops (a 38 percent reduction in average depth) and matches uniform depth across the average-depth range, whereas a linear router trained on convergence labels harvested from the same model requires nearly full depth and yields no reduction. The elasticity that makes this possible reproduces here as background (validation loss decreases monotonically from 3.80 at one loop to 3.20 at eight and remains stable to 32 loops). We report average depth as a FLOP proxy with a three-point wall-clock bracket rather than a realized speedup, make no FLOP-matched parity claim, and note that the allocation results are established at a single scale and seed. The complete study runs on a single RTX 4090 in approximately 100 GPU-hours.
Chinese Translation
深度递归变换器将权重绑定的核心应用于可变次数的循环，先前的研究表明，使用随机递归次数进行训练可以产生一个在不同推理深度下可用的检查点。我们探讨了这样的模型每个标记实际计算的内容，并进行了直接测量。在一个在 FineWeb-Edu 上训练的 135M 级模型中，递归状态收敛到每个标记的固定点：平均连续输出的 KL 散度从第二轮的 3.9e-1 降至第十六轮的 8.5e-6，且每个标记的状态变化逐步减小。至关重要的是，这种收敛在标记之间并不均匀。中位数标记在第六轮收敛，而大约 10% 的标记在训练平均深度为八的情况下仍在更新，平均收敛深度按标记类型排序（空格最浅，内容词最深）。这种每个标记的变异性是本文的核心对象。我们展示了它是可以直接读取的，并且读取它的效果优于学习预测它：一种训练无关的规则在每个标记的输出稳定后停止，达到了 4.94 次平均循环的均匀深度 8 质量（平均深度减少了 38%），并在平均深度范围内匹配均匀深度，而从同一模型中提取的收敛标签训练的线性路由器几乎需要完整深度且没有减少。使这一切成为可能的弹性在这里作为背景重现（验证损失从一轮的 3.80 单调减少到八轮的 3.20，并在 32 轮时保持稳定）。我们将平均深度报告为 FLOP 的代理，使用三点墙钟区间，而不是实现的加速，未做出 FLOP 匹配的平价声明，并注意分配结果是在单一规模和种子下建立的。完整研究在单个 RTX 4090 上运行，耗时约 100 GPU 小时。
cs.AI / 32 / 2607.14443
Tactile: Giving Computer-Using Agents Hands and Feet
Tactile：为计算机使用代理提供手和脚
Abstract
Computer-use agents are becoming capable software operators, but their interface to desktop applications is still often a brittle motor layer: they look at screenshots, predict coordinates, click, and hope that the visible state changed as intended. This collapses target grounding, action execution, and outcome verification into a single ambiguous operation. We present Tactile, an open-source tool layer that gives agents a more reliable "hands and feet" for desktop use. Tactile converts heterogeneous UI evidence--operating-system accessibility semantics, OCR-grounded text, and visual fallback regions--into action-grounded interface states: compact target candidates with source labels, roles or text, state, geometry, executable affordances, and verification cues. Agents operate through an observe-ground-act-verify loop that prefers native semantic actions when available, falls back to OCR-grounded coordinates when visible text is the best evidence, and keeps full provenance for replay and failure attribution. On macOSWorld-style tasks, adding Tactile improves Codex Success@100 from 41.1% to 50.0% overall and from 45.2% to 55.3% on accessibility-adapted tasks; a 96-task cross-agent subset shows consistent gains across Codex, Claude Code, OpenCode, and Goose. These results suggest that reliable computer use requires not only stronger models, but also a reusable execution substrate that exposes software actions as semantic, verifiable, and auditable objects rather than anonymous screen coordinates.
Chinese Translation
计算机使用代理正在成为能够操作软件的能力者，但它们与桌面应用程序的接口仍然常常是一个脆弱的运动层：它们查看截图，预测坐标，点击，并希望可见状态如预期那样改变。这将目标定位、动作执行和结果验证合并为一个模糊的操作。我们提出了 Tactile，一个开源工具层，为代理提供更可靠的桌面使用“手和脚”。Tactile 将异构的用户界面证据——操作系统的可访问性语义、基于光学字符识别（OCR）的文本和视觉回退区域——转换为基于动作的接口状态：紧凑的目标候选项，带有源标签、角色或文本、状态、几何形状、可执行的可供性和验证线索。代理通过观察-行动-验证循环进行操作，当可用时优先选择本地语义动作，当可见文本是最佳证据时回退到基于 OCR 的坐标，并保持完整的来源记录以便重放和失败归因。在 macOSWorld 风格的任务中，添加 Tactile 将 Codex 的成功率从 41.1% 提高到 50.0%，在适应可访问性的任务中从 45.2% 提高到 55.3%；在一个包含 96 个任务的跨代理子集上，Codex、Claude Code、OpenCode 和 Goose 一致显示出持续的提升。这些结果表明，可靠的计算机使用不仅需要更强大的模型，还需要一个可重用的执行基础，能够将软件动作呈现为语义的、可验证的和可审计的对象，而不是匿名的屏幕坐标。
cs.AI / 33 / 2607.14485
Step-Level Preference Learning for Generative Agents in Social Simulations
社会模拟中生成代理的逐步偏好学习
Abstract
Large language model (LLM)-based generative agents simulate human behavior through long-horizon decision-making processes that comprise intermediate steps such as planning, memory retrieval, reflection, and action selection. However, fine-grained human annotations of these intermediate steps remain scarce, and existing agents are not grounded in human preferences over such intermediate decisions. To address this gap, we introduce \method, an interactive simulation interface that enables us to collect step-level human preference supervision over agent decision trajectories, leading to a dataset of 57K fine-grained annotations. We conduct step-level preference learning on open-weight language models using supervised finetuning and direct preference optimization on this data, consistently improving simulation fidelity, coordination, and interaction quality, and inducing more socially effective agent behavior. Our results show that step-level human supervision is an effective training signal for improving both local decision quality and long-horizon agent behavior.
Chinese Translation
基于大型语言模型（LLM）的生成代理通过包括规划、记忆检索、反思和行动选择等中间步骤的长期决策过程来模拟人类行为。然而，这些中间步骤的细粒度人类注释仍然稀缺，现有代理并未基于人类对这些中间决策的偏好。为了解决这一问题，我们引入了 extit{method}，一个交互式模拟界面，使我们能够收集代理决策轨迹上的逐步人类偏好监督，从而形成一个包含57K细粒度注释的数据集。我们在开放权重语言模型上进行逐步偏好学习，使用监督微调和基于该数据的直接偏好优化，持续提高模拟的真实性、协调性和交互质量，并促使代理行为更加社会有效。我们的结果表明，逐步人类监督是提高局部决策质量和长期代理行为的有效训练信号。
cs.AI / 34 / 2607.14494
SAGA: Schema-Aware Grounding for Agentic Text-to-SPARQL Generation
SAGA：面向代理的文本到SPARQL生成的模式感知基础
Abstract
Complex knowledge base question answering (KBQA) is commonly approached through either information retrieval over a question-specific subgraph or semantic parsing into an executable logical form. We study the latter paradigm. Recent large language model agents make semantic parsing interactive: they alternate between reasoning, querying the knowledge base, and extending a partial SPARQL query. This interleaving reduces reliance on one-shot generation, but makes the quality of \emph{KB grounding} depend on what the interaction tools expose. Existing agents retrieve or prune candidate properties mainly through lexical relevance and instance-level observations, without systematically conditioning on entity types, property domains and ranges, or the expected answer type. We call this failure mode \emph{type-blind grounding}. It enlarges the grounding search space and often produces plausible-looking but semantically incompatible triple patterns that execute to empty results. We propose SAGA (\underline{S}chema-\underline{A}ware \underline{G}rounding for \underline{A}gentic Text-to-SPARQL Generation), a training-free framework that turns property exploration into a schema-constrained grounding operation. SAGA maintains a persistent bidirectional type state, filters known-incompatible property candidates at construction time, presents the remaining graph patterns in a compact schema-annotated format, and handles missing schema information permissively through empirical and trace-local evidence. Across nine benchmark settings over Wikidata and Freebase, SAGA achieves the highest F1 on all nine settings and the highest exact-match accuracy on eight, while reducing empty-result queries across all reported Wikidata settings.
Chinese Translation
复杂知识库问答（KBQA）通常通过对特定问题子图的信息检索或将语义解析为可执行的逻辑形式来处理。我们研究后者的范式。近期的大型语言模型代理使得语义解析变得互动：它们在推理、查询知识库和扩展部分SPARQL查询之间交替进行。这种交错减少了对一次性生成的依赖，但使得 extit{KB基础}的质量取决于交互工具所暴露的内容。现有代理主要通过词汇相关性和实例级观察来检索或修剪候选属性，而没有系统地考虑实体类型、属性域和范围或预期答案类型。我们将这种失效模式称为 extit{类型盲基础}。它扩大了基础搜索空间，并且通常产生看似合理但语义上不兼容的三元组模式，导致执行结果为空。我们提出了SAGA（ extbf{S}chema- extbf{A}ware extbf{G}rounding for extbf{A}gentic Text-to-SPARQL Generation），这是一个无训练框架，将属性探索转变为受模式约束的基础操作。SAGA维持一个持久的双向类型状态，在构建时过滤已知不兼容的属性候选，采用紧凑的模式注释格式呈现剩余的图模式，并通过经验和追踪局部证据宽松地处理缺失的模式信息。在对Wikidata和Freebase的九个基准设置中，SAGA在所有九个设置中实现了最高的F1分数，并在八个设置中获得了最高的精确匹配准确率，同时减少了所有报告的Wikidata设置中的空结果查询。
cs.AI / 35 / 2607.14499
Contextualized Evaluation of Vision Language Models through Dynamic, Multi-turn Interactions
通过动态多轮互动对视觉语言模型进行情境化评估
Abstract
Multi-modal Large Language Models (MLLMs) have made substantial advances on benchmarks, yet their real-world effectiveness remains uncertain. This gap stems from the fundamental misalignment between benchmarks in controlled, static settings and the dynamic, interactive, and contextualized nature of real-world applications. To bridge this gap, we propose CEDI (Contextualized Evaluations of MLLMs through Dynamic, multi-round Interactions), a framework that recasts evaluation as a three-party interaction between an evaluatee model, an automated examiner, and a grader. The examiner conducts multi-turn, semi-structured conversation guided by a graph-based representation of the task. By navigating state-space transitions, CEDI deploys diverse strategies, from clarification requests to adversarial probes, to elicit performance evidence. We apply CEDI to visual hallucinations. Empirical results across multiple models, diverse settings, datasets, and domains show that contextualized, interactive evaluations reveal not only significantly more hallucinations than conventional static evaluation but also ones that more closely resemble those arising in practical use cases. We further show that hallucinations often accumulate over long contexts, through self-reinforcing dialogue history, and models are particularly vulnerable to questions requiring premise rejection or refusal. Together, these findings highlight CEDI as a step toward realistic, systematic, and ecologically valid assessments of MLLMs' capabilities. Code is available at github.com/williamium3000/cedi.
Chinese Translation
多模态大型语言模型（MLLMs）在基准测试中取得了显著进展，但其在现实世界中的有效性仍然不确定。这一差距源于基准测试在受控、静态环境中与现实应用的动态、互动和情境化特性之间的根本不一致。为了解决这一问题，我们提出了CEDI（通过动态多轮互动对MLLMs进行情境化评估），一个将评估重新构建为评估模型、自动化考官和评分者之间的三方互动的框架。考官通过任务的图形表示进行多轮半结构化对话。通过导航状态空间转变，CEDI采用多种策略，从澄清请求到对抗性探测，以引出性能证据。我们将CEDI应用于视觉幻觉。跨多个模型、不同设置、数据集和领域的实证结果表明，情境化的互动评估不仅揭示了显著更多的幻觉，而且这些幻觉更接近于实际使用案例中出现的幻觉。我们进一步表明，幻觉往往在较长的上下文中积累，通过自我强化的对话历史，模型特别容易受到需要拒绝前提或拒绝的问题的影响。综合来看，这些发现突显了CEDI作为实现对MLLMs能力进行现实、系统和生态有效评估的一步。代码可在github.com/williamium3000/cedi获取。
cs.AI / 36 / 2607.14510
VLT: A Vision-Language-Time Series Multimodal Foundation Model for Industrial Intelligence
VLT：一种用于工业智能的视觉-语言-时间序列多模态基础模型
Abstract
Industrial time series serve as the foundation for Prognostics and Health Management (PHM) to ensure the reliability and safety of industrial equipment such as aero-engines. However, existing approaches are typically limited to single-modality modeling, which restricts their generalization in complex scenarios. Although recent advances in large language models (LLMs) provide new opportunities for multimodal learning, bridging continuous time-series signals and discrete textual semantics remains an open challenge. To this end, we propose VLT, a multimodal foundation model that jointly models time-series, frequency-spectrum visual representations, and textual knowledge. A key insight is to utilize the frequency spectrum as a visual bridge to connect continuous temporal signals with discrete semantics. Specifically, a Time-aware Mixture-of-Experts (Time-MoE) is designed to capture heterogeneous temporal dynamics, while a Frequency-Text Augmented Learner enables joint modeling of spectral and semantic features within a shared representation space. Furthermore, a time-centric gradient alignment mechanism is introduced to mitigate cross-modal optimization conflicts via gradient normalization and reliability-aware dynamic reweighting. Extensive experiments on multiple industrial datasets demonstrate that VLT outperforms state-of-the-art methods, achieving superior robustness and generalization under few-shot, noisy, and incomplete-modality settings.
Chinese Translation
工业时间序列是预测与健康管理（PHM）的基础，确保航空发动机等工业设备的可靠性和安全性。然而，现有的方法通常局限于单一模态建模，这限制了它们在复杂场景中的泛化能力。尽管近期大型语言模型（LLMs）的进展为多模态学习提供了新的机会，但将连续时间序列信号与离散文本语义连接起来仍然是一个开放的挑战。为此，我们提出了VLT，这是一种多模态基础模型，联合建模时间序列、频谱视觉表示和文本知识。一个关键的见解是利用频谱作为视觉桥梁，将连续时间信号与离散语义连接起来。具体而言，设计了一种时间感知的专家混合模型（Time-aware Mixture-of-Experts, Time-MoE）来捕捉异质的时间动态，同时频谱-文本增强学习器（Frequency-Text Augmented Learner）使得在共享表示空间内联合建模频谱和语义特征成为可能。此外，引入了一种以时间为中心的梯度对齐机制，通过梯度归一化和可靠性感知的动态重加权来减轻跨模态优化冲突。在多个工业数据集上的大量实验表明，VLT在少样本、噪声和不完整模态设置下的鲁棒性和泛化能力优于最先进的方法。
cs.AI / 37 / 2607.14512
RetroAgent: Harnessing LLMs to Search Over Structured Memory for Agentic Retrosynthesis Planning
RetroAgent：利用大型语言模型在结构化记忆中进行代理回溯合成规划
Abstract
Multi-step retrosynthesis planning seeks to decompose a target molecule into commercially available building blocks through a sequence of feasible reactions. The vast combinatorial search space makes this task challenging even for expert chemists. Traditional methods combine tree search with offline-trained value networks that score candidates in isolation, without reasoning about complete multi-step routes. Recent work leverages Large Language Models (LLMs) for this task, but relies on simple interfaces that limit exploration of the full search space. We introduce RetroAgent, an LLM agent that bridges symbolic search and neural reasoning through a harness with structured memory. Through memory and chemistry tools, the agent observes the full search state, including explored routes, available alternatives, and properties of intermediates, enabling informed decisions grounded in both global progress and domain knowledge. Experiments on in-distribution and out-of-distribution benchmarks demonstrate that RetroAgent delivers strong performance and generalization.
Chinese Translation
多步骤回溯合成规划旨在通过一系列可行反应将目标分子分解为商业可获得的构建块。庞大的组合搜索空间使得这一任务即使对于专家化学家来说也颇具挑战性。传统方法结合了树搜索与离线训练的价值网络，这些网络在孤立情况下对候选者进行评分，而未考虑完整的多步骤路线。最近的研究利用大型语言模型（LLMs）来处理这一任务，但依赖于简单的接口，限制了对完整搜索空间的探索。我们提出了RetroAgent，这是一种通过结构化记忆的工具将符号搜索与神经推理相结合的LLM代理。通过记忆和化学工具，代理能够观察完整的搜索状态，包括已探索的路线、可用的替代方案以及中间体的性质，从而基于全球进展和领域知识做出明智的决策。在分布内和分布外基准测试中的实验表明，RetroAgent展现了强大的性能和良好的泛化能力。
cs.AI / 38 / 2607.14524
WrAFT: a Modularized Automated Writing Evaluation System for Argumentative Essays
WrAFT：一种模块化的自动化写作评估系统用于论证性论文
Abstract
This study presents WrAFT, a Writing Assessment and Feedback Tool, that delivers both accurate and reliable scores and effective comprehensive feedback to argumentative essays. WrAFT adopts a modular design by dividing automated writing evaluation (AWE) tasks into scoring, surface-level feedback, and deep-level feedback. In building the system, various Large Language Models (LLMs) have been evaluated, including LLaMA-3.3-70B-Instruct, GPT-4o, and Claude 3.7, through both direct prompting and supervised fine-tuning approaches. A proprietary dataset of 480 TOEFL Independent Writing essays with official benchmark scores was utilized. Benchmark-based evaluation shows that WrAFT achieves state-of-the-art performance in scoring, with a quadratic weighted kappa (QWK) of 0.84 and a root mean square error (RMSE) of 0.44 against official scores on a scale of 0-5. Human evaluation of system-generated feedback also reveals high approval ratings: 96.14 percent for surface-level feedback, 93.03 percent for deep-level macro feedback, and 94.69 percent for deep-level micro feedback. An interactive user interface has been developed for the system and is publicly available and free to use.
Chinese Translation
本研究提出了WrAFT（写作评估与反馈工具），该工具为论证性论文提供准确可靠的评分和有效的综合反馈。WrAFT采用模块化设计，将自动化写作评估（AWE）任务划分为评分、表层反馈和深层反馈。在构建该系统时，评估了多种大型语言模型（LLMs），包括LLaMA-3.3-70B-Instruct、GPT-4o和Claude 3.7，采用了直接提示和监督微调的方法。使用了一个包含480篇托福独立写作论文及其官方基准分数的专有数据集。基于基准的评估显示，WrAFT在评分方面达到了最先进的性能，针对官方分数在0-5的范围内，其二次加权kappa（QWK）为0.84，均方根误差（RMSE）为0.44。对系统生成反馈的人类评估也显示出高认可度：表层反馈的认可率为96.14%，深层宏观反馈为93.03%，深层微观反馈为94.69%。该系统开发了一个交互式用户界面，并已公开提供，免费使用。
cs.AI / 39 / 2607.14541
Are LLM-Generated GPU Kernels Production-Ready? A Trace-Driven Benchmark and Optimization Agent
LLM生成的GPU内核是否准备好投入生产？基于追踪的基准测试与优化代理
Abstract
Existing GPU kernel generation benchmarks draw problems from synthetic or curated sources that diverge from deployed workloads. We present Atrex-Bench, a benchmark whose 30 operators and 440 shapes are sampled directly from full-cluster production inference traces of compute-limited, memory-rich GPUs. Each problem carries an importance weight derived from its share of observed GPU time, weighted by application card-hours and computed separately for the serving phases in which it runs, together with a per-problem roofline ceiling, so the aggregate score emphasizes the kernels that consume the most serving time. Evaluating six frontier coding agents on Atrex-Bench shows that even the best vanilla model reaches only ${\sim}10\%$ of the hardware roofline on production operators; and correctness alone overstates capability, since much of the apparent pass rate comes from PyTorch fallbacks rather than kernels the model wrote. To close this gap, we co-release Atrex-Kernel-Agent (AKA), a profile-driven kernel-optimization agent that combines iterative measure-revise search, optimization dropout for escaping stalled search contexts, and a layered GPU-optimization knowledge base (298 reference-kernel files and 244 optimization-knowledge documents, plus external upstream reference projects for API/ISA lookup). In a controlled case study, the agent converts zero-FlyDSL fallbacks into real kernels that match or exceed hand-tuned production baselines.
Chinese Translation
现有的GPU内核生成基准测试从合成或策划的来源中提取问题，这些来源与实际部署的工作负载存在差异。我们提出了Atrex-Bench，这是一个基准测试，其30个操作符和440个形状直接从计算受限、内存丰富的GPU的全集群生产推理追踪中抽样而来。每个问题都有一个重要性权重，该权重来源于其观察到的GPU时间份额，按应用卡时数加权，并分别计算其运行的服务阶段，同时提供每个问题的屋顶线限制，因此总分强调消耗最多服务时间的内核。在Atrex-Bench上评估六个前沿编码代理显示，即使是最好的原始模型在生产操作符上也仅达到硬件屋顶线的约10%；而仅凭正确性来评估能力则会高估，因为许多表面上的通过率来自于PyTorch的回退，而非模型生成的内核。为了缩小这一差距，我们共同发布了Atrex-Kernel-Agent（AKA），这是一个基于配置文件的内核优化代理，结合了迭代测量-修正搜索、优化丢弃以逃避停滞的搜索上下文，以及一个分层的GPU优化知识库（298个参考内核文件和244个优化知识文档，以及用于API/ISA查找的外部上游参考项目）。在一个受控案例研究中，该代理将零-FlyDSL回退转换为真实内核，这些内核与手动调优的生产基准相匹配或超出。
cs.AI / 40 / 2607.14553
Towards an Intention Abstraction Layer for Autonomous Industrial Systems
面向自主工业系统的意图抽象层
Abstract
Modern industrial environments increasingly run many autonomous subsystems at once - schedulers, energy managers, vehicle fleets - each pursuing its own goals while sharing the same physical resources. Because high-level human intentions are translated into low-level control logic and then discarded, no running component can tell whether it is still doing what was actually intended, and goal conflicts surface only after they have caused a missed target or a shutdown. We propose the Intention Abstraction Layer (IAL), a domainagnostic middleware that represents intentions as first-class, persistent, and explainable runtime objects: a large language model grounded in a formal OWL ontology parses naturallanguage goals into structured intentions, a consistency monitor detects conflicts at registration time, before execution, and a transparency module explains them in natural language. We report a first proof of concept in which two autonomous agents register conflicting production and energy intentions, and the IAL flags and explains the conflict before it reaches the execution layer. The result is a mechanism that shifts behavioral assurance for cooperating autonomous systems from post-hoc failure analysis to pre-execution, intention-level checking.
Chinese Translation
现代工业环境中，越来越多的自主子系统同时运行——调度器、能源管理器、车辆车队——每个子系统都在追求自己的目标，同时共享相同的物理资源。由于高层次的人类意图被转化为低层次的控制逻辑后被丢弃，因此没有任何正在运行的组件能够判断其是否仍在执行实际的意图，目标冲突往往在造成目标未达成或系统停机后才显现。我们提出了意图抽象层（Intention Abstraction Layer, IAL），这是一种领域无关的中间件，将意图表示为一等公民、持久且可解释的运行时对象：一个基于正式OWL本体的大型语言模型将自然语言目标解析为结构化意图，一种一致性监测器在注册时检测冲突，执行之前进行检查，而透明度模块则用自然语言解释这些冲突。我们报告了一个初步的概念验证，其中两个自主代理注册了相互冲突的生产和能源意图，IAL在冲突到达执行层之前标记并解释了该冲突。最终结果是一个机制，将合作自主系统的行为保证从事后失败分析转移到执行前的意图级检查。
cs.AI / 41 / 2607.14557
Seeing the End at Step Zero: Accelerating Diffusion MLLMs via MLP Sparsity-Aware Truncation
在零步看到终点：通过 MLP 稀疏感知截断加速扩散多模态大语言模型
Abstract
Diffusion Multimodal Large Language Models (DMLLMs) are highly effective for multimodal reasoning, yet their inference efficiency is significantly hindered by fixed-length generation constraints. Since the actual output length is unknown, output sequences are padded to a predefined maximum length, resulting in substantial redundant computation over unnecessary [EOS] tokens. In this work, we discover that DMLLMs implicitly reveal their valid semantic boundary at the very first denoising step through a distinct shift in MLP activation sparsity. Leveraging this observation, we propose Seer, a training-free framework that detects this boundary using a Signal-to-Noise Ratio (SNR)-based criterion and performs one-shot truncation of the redundant suffix for all subsequent computations. To preserve these theoretical gains during batched serving, Seer incorporates a hybrid execution strategy that maximizes throughput while seamlessly accommodating dynamic sequence lengths. Experimental results demonstrate that Seer effectively eliminates padding waste, accelerating throughput by up to $\sim$31$\times$. Across 9 benchmarks, Seer robustly maintains overall performance and even improves accuracy on complex visual tasks by mitigating noise leakage (e.g., DocVQA score increases from 63.52 to 63.66), offering a highly efficient, plug-and-play solution for DMLLM acceleration.
Chinese Translation
扩散多模态大语言模型（DMLLMs）在多模态推理中表现出色，但其推理效率受到固定长度生成约束的显著影响。由于实际输出长度未知，输出序列被填充至预定义的最大长度，导致在不必要的 [EOS] 令牌上产生大量冗余计算。在本研究中，我们发现 DMLLMs 在第一步去噪过程中通过 MLP 激活稀疏性的明显变化隐含地揭示了其有效语义边界。基于这一观察，我们提出了 Seer，一个无训练的框架，利用基于信噪比（SNR）的标准检测这一边界，并对后续计算中的冗余后缀进行一次性截断。为了在批量服务过程中保持这些理论收益，Seer 采用了一种混合执行策略，最大化吞吐量，同时无缝适应动态序列长度。实验结果表明，Seer 有效消除了填充浪费，将吞吐量提高了约 31 倍。在 9 个基准测试中，Seer 稳定地保持了整体性能，甚至通过减轻噪声泄漏提高了复杂视觉任务的准确性（例如，DocVQA 分数从 63.52 增加到 63.66），为 DMLLM 加速提供了一种高效的即插即用解决方案。
cs.AI / 42 / 2607.14570
Democratizing Agent Deployment Safety: A Structural Monitoring Approach
民主化代理部署安全：一种结构监测方法
Abstract
AI software development agents are increasingly capable of modifying infrastructure and security critical systems, creating risks where an agent completes its assigned task while covertly weakening safeguards through actions such as broadening permissions, degrading logging, or introducing persistence mechanisms. While frontier laboratories may deploy sophisticated monitoring pipelines, many organizations and individual users adopting coding agents lack the resources and governance maturity required to maintain complex learned monitor ensembles. We study this problem as infrastructure level sabotage under task success in the infrastructure-as-code setting from ControlArena and compare trained and untrained monitors across synchronous and asynchronous deployment modes. We introduce an Information Flow Graph (IFG) monitor that analyzes structural security regressions using control-flow and data-flow graph diffs alongside raw code diffs. In asynchronous evaluation, an untrained git diff monitor misses 11.6% attacks at 1% false positive rate auditing budget, our untrained IFG monitor reduces this to 3.5%, and the trained Async RF monitor achieves below 1%. Unlike the trained Async RF monitor, IFG can also operate synchronously as a pre-deployment safeguard that blocks suspicious infrastructure changes before execution or merge. In synchronous mode, IFG realtime rollback reduces joint main task and covert task success from 74.4% to 0.0% with no measurable reduction in legitimate task completion. These results suggest that untrained structural monitors provide a practical and auditable path toward democratizing deployment safety for organizations seeking trustworthy adoption of advanced AI agents.
Chinese Translation
人工智能软件开发代理越来越能够修改基础设施和安全关键系统，这带来了风险，即代理在完成其分配任务的同时，通过扩展权限、降低日志记录或引入持久性机制等行为，秘密削弱了安全防护。尽管前沿实验室可能会部署复杂的监测管道，但许多组织和个人用户在采用编码代理时缺乏维护复杂学习监测集所需的资源和治理成熟度。我们将这一问题视为基础设施级别的破坏，研究在基础设施即代码（Infrastructure-as-Code）环境下任务成功的情况下的表现，并比较同步和异步部署模式下的训练和未训练监测器。我们提出了一种信息流图（Information Flow Graph, IFG）监测器，该监测器通过控制流和数据流图的差异以及原始代码差异来分析结构安全回归。在异步评估中，未训练的 git diff 监测器在 1% 假阳性率审计预算下漏掉了 11.6% 的攻击，而我们的未训练 IFG 监测器将这一比例降低至 3.5%，训练后的 Async RF 监测器则实现了低于 1%。与训练后的 Async RF 监测器不同，IFG 还可以作为一种同步的预部署安全防护，在执行或合并之前阻止可疑的基础设施更改。在同步模式下，IFG 实时回滚将联合主任务和隐蔽任务的成功率从 74.4% 降低至 0.0%，而合法任务完成率没有可测量的下降。这些结果表明，未训练的结构监测器为寻求可信采用先进人工智能代理的组织提供了一条切实可行且可审计的民主化部署安全之路。
cs.AI / 43 / 2607.14573
Alipay-PIBench: A Realistic Payment Integration Benchmark for Coding Agents
Alipay-PIBench：一个针对编码代理的现实支付集成基准
Abstract
Payment integration is a demanding repository-level software task: agents must select a suitable product, implement coordinated client-server flows, verify payment outcomes, and preserve consistency between transaction and business states. We introduce Alipay-PIBench, a benchmark for evaluating coding agents on realistic Alipay payment integration. It contains nine product-specific projects and 18 task instances, each organized into Basic functional-completion and Advanced risk-aware hardening scenarios. Scenario-specific rubrics support deterministic static, unit, integration, and end-to-end checks, supplemented by LLM-assisted assessment for semantic requirements. We evaluate six coding-agent models and report rubric pass rate (RPR). Under the with-skill condition, mean RPR ranges from 68.58% to 91.37%. Access to the alipay-payment-integration skill improves mean RPR by 10.31 percentage points on average relative to the without-skill condition, with gains varying across models, products, and scenarios. Method-level results distinguish source-level completion, executable payment behavior, and payment-domain requirements. Alipay-PIBench provides a controlled setting for diagnosing model capability and evaluating structured guidance in payment integration.
Chinese Translation
支付集成是一项要求严格的库级软件任务：代理必须选择合适的产品，实施协调的客户端-服务器流程，验证支付结果，并保持交易状态与业务状态之间的一致性。我们介绍了 Alipay-PIBench，这是一个用于评估编码代理在现实 Alipay 支付集成中的基准。它包含九个特定产品的项目和18个任务实例，每个实例分为基本功能完成和高级风险意识强化场景。特定场景的评分标准支持确定性的静态检查、单元测试、集成测试和端到端测试，并辅以 LLM 辅助的语义要求评估。我们评估了六种编码代理模型，并报告了评分标准通过率（RPR）。在具备技能的条件下，平均 RPR 范围从 68.58% 到 91.37%。获得 alipay-payment-integration 技能使得平均 RPR 相较于无技能条件提高了 10.31 个百分点，且增益在不同模型、产品和场景之间有所不同。方法级结果区分了源级完成、可执行支付行为和支付领域要求。Alipay-PIBench 提供了一个受控环境，用于诊断模型能力和评估支付集成中的结构化指导。
cs.AI / 44 / 2607.14574
Collaborative Spatial Learning with Multi-LLM Agents in Networked Social Experiments
网络社会实验中多LLM代理的协作空间学习
Abstract
Collective problem solving often requires that group members consider the tradeoff between exploitation of known solutions and exploration for new ones, where information of known solutions can be disseminated among individual members through communication networks. The Mason--Watts experiment (PNAS 2012) showed that human groups in shorter-path networks outperform those in longer-path networks on a two-dimensional search task. In this work, we focus on the investigation of such network-efficiency effects in the setting of a group of large language model (LLM) agents. Specifically, we consider groups of sixteen LLM agents playing the Mason--Watts experiment on the eight Mason--Watts network topologies. Moreover, we develop mechanistic Bayesian optimization agents such that the performance of LLM agents can be compared with both the mechanistic agents and the human experimental data. Our computational experiments indicate that the LLM agents show a significant network-efficiency effect when instructed to randomize their first-round choices, but not under the default initialization. In this experiment, adding a one-sentence first-round randomization instruction improves collective payoff by more than three times the estimated payoff difference across the eight network topologies. Also, the Bayesian optimization agents obtain higher payoffs than the evaluated LLM agents on this spatial search task. We further compare the agents' exploration--exploitation behavior, copying, and spatial diversity.
Chinese Translation
集体问题解决通常要求组员在利用已知解决方案与探索新方案之间权衡，其中已知解决方案的信息可以通过通信网络在个体成员之间传播。Mason-Watts实验（PNAS 2012）表明，在二维搜索任务中，处于短路径网络中的人类群体表现优于处于长路径网络中的群体。在本研究中，我们重点调查在一组大型语言模型（LLM）代理的设置中，这种网络效率效应。具体而言，我们考虑由十六个LLM代理组成的群体在八种Mason-Watts网络拓扑上进行Mason-Watts实验。此外，我们开发了机械贝叶斯优化代理，以便将LLM代理的表现与机械代理和人类实验数据进行比较。我们的计算实验表明，当被指示随机化他们的第一轮选择时，LLM代理显示出显著的网络效率效应，但在默认初始化下则没有。在这项实验中，增加一句话的第一轮随机化指令使集体收益提高了超过八种网络拓扑中估计收益差的三倍。此外，贝叶斯优化代理在这一空间搜索任务中获得的收益高于评估的LLM代理。我们进一步比较了代理的探索-利用行为、复制和空间多样性。
cs.AI / 45 / 2607.14581
Multi-LLM Collaborative MRI Report Generation for Visual Instruction Tuning in Brain Oncology
多大型语言模型协作生成脑肿瘤视觉指导调优的MRI报告
Abstract
Recent advances in large language models (LLMs) and their extension to vision-language models (VLMs) have made it easier to combine text and images for tasks such as report generation. Existing VLMs in medicine typically focus on 2D images (chest X-rays), and their extension to 3D imaging has been difficult because of the lack of paired 3D imaging-text data. Thus, we introduce a new method for creating a 3D image-text dataset for brain oncology using 3D MRI scans of glioma and meningioma cases. We use a cooperative system in which several LLMs work together to generate and check reports, ensuring that they are accurate and clear. By leveraging the new 3D MRI-text dataset, we further build a VLM that converts MRI scans into tokens and aligns them with text instructions. Our VLM performed better in report generation and visual question answering tasks than other 2D and 3D methods. Our method not only improves the quality of reports but also helps with better diagnosis and treatment in brain oncology.
Chinese Translation
近年来，大型语言模型（LLMs）及其扩展到视觉-语言模型（VLMs）的进展，使得文本与图像的结合在报告生成等任务中变得更加容易。现有的医学VLMs通常专注于2D图像（胸部X光片），而其扩展到3D成像却因缺乏配对的3D成像-文本数据而面临挑战。因此，我们提出了一种新的方法，利用胶质瘤和脑膜瘤病例的3D MRI扫描创建脑肿瘤的3D图像-文本数据集。我们采用一个合作系统，多个LLMs共同工作以生成和检查报告，确保其准确性和清晰性。通过利用新的3D MRI-文本数据集，我们进一步构建了一个VLM，将MRI扫描转换为标记，并与文本指令对齐。我们的VLM在报告生成和视觉问答任务中的表现优于其他2D和3D方法。我们的方法不仅提高了报告的质量，还促进了脑肿瘤的更好诊断和治疗。
cs.AI / 46 / 2607.14582
MathCoPilot: An Interactive System for Human-AI Symbiotic Paradigm of Mathematical Research
MathCoPilot：一种人机共生数学研究的互动系统
Abstract
Existing LLM-based theorem provers have achieved impressive results on formal mathematics benchmarks, yet they remain confined to acting as autonomous agents that prove a stated proposition. In this paper, we propose MathCoPilot, a human-in-the-loop system that embodies a new human--AI symbiotic paradigm for mathematical research, in which the mathematician steers the high-level mathematical direction while AI agents carry out the detailed formalization and proof work under continuous human guidance. MathCoPilot unifies three core capabilities: (1) an interactive workbench where the mathematician and AI agents collaborate through a living proof blueprint that decomposes a proof into navigable steps the human can directly inspect, direct, and refine; (2) automated proving skill orchestration with adaptive knowledge base search and Lean-integrated iterative verification; and (3) topic-driven paper retrieval and automated formalization into a verified Lean knowledge base. Using MathCoPilot, we systematically compare four state-of-the-art LLMs, including Gemini~3.1~Pro, GPT-5.4, and Claude~Opus~4.7, on a FormalMATH subset and on two real PDE theorems requiring deep domain expertise, evaluating their ability to produce verified Lean~4 proofs and to identify errors in deliberately incorrect proofs. Our results show that while current models can handle undergraduate-level problems with high success rates under favorable autoformalization conditions, substantial challenges remain for domain-specific theorems requiring genuine mathematical understanding.
Chinese Translation
现有的基于大型语言模型（LLM）的定理证明器在形式数学基准测试中取得了令人瞩目的成果，但仍然局限于作为自主代理来证明所陈述的命题。本文提出了MathCoPilot，一个人机协作系统，体现了一种新的人人工智能共生范式，用于数学研究。在这一范式中，数学家引导高层次的数学方向，而人工智能代理在持续的人类指导下进行详细的形式化和证明工作。MathCoPilot统一了三项核心能力：（1）一个互动工作台，数学家与人工智能代理通过一个动态证明蓝图进行协作，该蓝图将证明分解为可导航的步骤，供人类直接检查、指导和完善；（2）自动证明技能的编排，结合自适应知识库搜索和Lean集成的迭代验证；（3）基于主题的论文检索和自动形式化到经过验证的Lean知识库。使用MathCoPilot，我们系统地比较了四个最先进的LLM，包括Gemini 3.1 Pro、GPT-5.4和Claude Opus 4.7，在FormalMATH子集和两个需要深厚领域专业知识的实际偏微分方程（PDE）定理上，评估它们生成经过验证的Lean 4证明和识别故意错误证明的能力。我们的结果表明，尽管当前模型在有利的自动形式化条件下能够高成功率地处理本科水平的问题，但对于需要真正数学理解的领域特定定理仍然存在重大挑战。
cs.AI / 47 / 2607.14616
SportD: Can VLMs Physically Strategize?
SportD：视觉语言模型能否进行物理战略规划？
Abstract
Vision--language models have become increasingly capable of interpreting visual scenes, but it remains unclear whether they can use information to make strategically effective decisions. We investigate this question in soccer, where models observe the seconds preceding an on-ball decision and must choose whether to shoot or pass to a specific teammate. Unlike conventional visual-understanding tasks, soccer enables decisions to be evaluated quantitatively by estimating the value of every available action. We introduce SportD, a benchmark comprising 478 on-ball decisions from the 2022 FIFA World Cup. Each model choice is evaluated against a possession-value model that estimates the action that most increases the attacking team's probability of scoring, allowing us to measure both optimal-action accuracy and the value forfeited by suboptimal decisions. Across three frontier VLMs, the best selects the highest-valued action on 31.4% of events, compared with 38.9% for the professional players, and all models incur significantly greater regret. Further analysis reveals a systematic preference for lower-variance and lower-reward actions: VLMs shoot less often and select substantially less progressive passes than either the optimal policy or the real players. The models also reproduce the player's specific action above chance even when that action is suboptimal, suggesting partial imitation of familiar play patterns rather than consistent evaluation of counterfactual alternatives. SportD provides a value-grounded testbed for measuring physical strategic reasoning in VLMs.
Chinese Translation
视觉-语言模型在解读视觉场景方面的能力日益增强，但它们是否能够利用信息做出战略上有效的决策仍不清楚。我们在足球领域探讨这个问题，在这里模型观察到球员在做出持球决策前的几秒钟，并必须选择是射门还是传球给特定队友。与传统的视觉理解任务不同，足球使得通过估算每个可用动作的价值来定量评估决策成为可能。我们引入了SportD，一个基准数据集，包含2022年国际足联世界杯的478个持球决策。每个模型的选择都与一个估算最能提高进攻队伍得分概率的持球价值模型进行评估，从而使我们能够测量最佳动作的准确性以及因次优决策而失去的价值。在三种前沿视觉语言模型中，最佳模型在31.4%的事件中选择了最高价值的动作，而职业球员的这一比例为38.9%，所有模型的后悔程度显著更高。进一步分析显示，模型系统性地偏好低方差和低回报的动作：视觉语言模型的射门频率低于最佳策略或真实球员，并且选择的进攻传球明显不够积极。即使在某些情况下，模型也能在随机水平之上再现球员的特定动作，即使该动作并非最佳，表明它们部分模仿了熟悉的比赛模式，而非对反事实选择进行一致的评估。SportD为测量视觉语言模型中的物理战略推理提供了一个基于价值的测试平台。
cs.AI / 48 / 2607.14635
Action QFormer: Structured Representation Shaping under Action Supervision in Vision-Language-Action Models
Action QFormer：在视觉-语言-动作模型中进行动作监督下的结构化表示塑造
Abstract
Action supervision in vision-language-action (VLA) models is often treated as a downstream objective for learning action prediction. In this paper, we study it instead as a force that shapes inherited multimodal representations. We show that this shaping has a dual effect: it is necessary for forming action-compatible representations, but when action supervision is applied too directly to the inherited multimodal pathway, it can also destabilize representations that support language-side processing and object grounding. To address this tension, we introduce Action QFormer, a query-based action-facing interface that uses instruction-conditioned queries to reorganize inherited multimodal information into action-facing representations before downstream action generation. In zero-shot sim-to-real navigation, Action QFormer improves average closed-loop task success from 18.8% to 56.3%, raises fixed-instruction action-generation correctness from 22.5% to 75.5%, and nearly eliminates out-of-distribution instruction generations. Further analyses show that Action QFormer changes how action supervision shapes inherited multimodal representations, reducing broad upstream rewriting while preserving targeted and sometimes constructive action-supervised adaptation. These results suggest that improving VLA performance requires not only stronger pretrained backbones, but also better ways of selecting and organizing inherited multimodal information while controlling how it is shaped under action supervision.
Chinese Translation
在视觉-语言-动作（VLA）模型中，动作监督通常被视为学习动作预测的下游目标。本文将其视为一种塑造继承多模态表示的力量。我们展示了这种塑造具有双重效应：它对于形成与动作兼容的表示是必要的，但当动作监督过于直接地应用于继承的多模态路径时，也可能会使支持语言处理和物体定位的表示不稳定。为了解决这一矛盾，我们引入了Action QFormer，这是一种基于查询的面向动作的接口，使用指令条件查询在下游动作生成之前将继承的多模态信息重新组织为面向动作的表示。在零-shot 模拟到真实导航中，Action QFormer将平均闭环任务成功率从18.8%提高到56.3%，将固定指令下的动作生成正确率从22.5%提升至75.5%，并几乎消除了分布外指令生成。进一步分析表明，Action QFormer改变了动作监督塑造继承多模态表示的方式，减少了广泛的上游重写，同时保留了有针对性且有时是建设性的动作监督适应。这些结果表明，提高VLA性能不仅需要更强大的预训练骨干网络，还需要更好的方式来选择和组织继承的多模态信息，同时控制其在动作监督下的塑造方式。
cs.AI / 49 / 2607.14641
Analytic Abduction: Causal Decomposition and Governed Commitment for Human--AI Coordination
分析性诱导：人类与人工智能协调的因果分解与受控承诺
Abstract
Abductive reasoning operates in two directions. The synthetic mode builds explanations from available hypotheses; the analytic mode, conversely, identifies the latent factors whose interaction accounts for a complex observed state. This paper develops the analytic mode as a non-greedy, risk-sensitive discipline of commitment, in which candidate factors coexist and interact, resolving into committed conclusions only when explicit governance conditions are met. The formal core is the $\kappa$-$\tau$ apparatus: $\kappa$ encodes the epistemic interaction among hypotheses, and $\tau$ sets a commitment threshold calibrated to the decision's stakes. The central contribution is the causal cluster, a structured object recording which latent factors participate in a decomposition, with what weights and interaction structure, together with a two-level architecture (intra-cluster $\kappa^*$, inter-cluster $\kappa^{**}$) that guards against causal misattribution. Demonstrated in epidemiological crisis decomposition and adversarial cyber threat analysis, the framework's contribution to human-AI reasoning is the legibility of suspended decomposition as a shared coordination object, providing structural resistance to premature convergence. In practice, the decision-maker is handed not a single imposed answer but the competing explanatory scenarios, weighted by plausibility and paired with the evidence that would resolve between them, so that sound action is possible even before the ambiguity is resolved.
Chinese Translation
诱导推理在两个方向上运作。合成模式从可用假设中构建解释；而分析模式则相反，识别出其交互作用导致复杂观察状态的潜在因素。本文将分析模式发展为一种非贪婪、风险敏感的承诺学科，其中候选因素共存并相互作用，仅在满足明确的治理条件时才会解析为承诺结论。其形式核心是$ ext{kappa}$-$ ext{tau}$工具：$ ext{kappa}$编码假设之间的认知交互，而$ ext{tau}$设定与决策风险相匹配的承诺阈值。中心贡献是因果集群，这是一个结构化对象，记录哪些潜在因素参与了分解，具有何种权重和交互结构，并配有两级架构（集群内$ ext{kappa}^*$，集群间$ ext{kappa}^{**}$），以防止因果错误归因。在流行病危机分解和对抗性网络威胁分析中进行了验证，该框架对人类-人工智能推理的贡献在于将悬而未决的分解作为共享协调对象的可读性，提供了对过早收敛的结构性抵抗。在实践中，决策者并非仅仅得到一个强加的答案，而是得到竞争性的解释场景，这些场景根据可信度加权，并配有能够在它们之间做出区分的证据，从而使得即使在模糊性未解决之前也能采取合理的行动。
cs.AI / 50 / 2607.14642
MCPEvol-Bench: Benchmarking LLM Agent Performance Across Dynamic Evolutions of MCP Servers
MCPEvol-Bench：在MCP服务器动态演变中评估LLM代理性能的基准
Abstract
As Model Context Protocol (MCP) servers emerge as the core infrastructure for connecting LLMs with external tools, existing benchmarks leverage real-world MCP servers to evaluate LLM agents' tool-using capabilities. However, these benchmarks overlook the continuous evolution of tool interfaces and functionalities within MCP servers, resulting in flawed assessments that fail to capture the agent's adaptability in changing tool landscapes. To bridge this gap, we introduce \textbf{MCPEvol-Bench}, a novel benchmark for evaluating the task-solving capabilities of LLM agents under dynamic toolset evolution. Inspired by large-scale empirical study, we propose 11 mutation operators to simulate realistic tool evolution within 123 MCP servers. We benchmark 12 state-of-the-art LLMs on multiple versions of MCP servers, revealing that even frontier models struggle to adapt to evolving tools. For instance, GPT-5.4 and Claude-Sonnet-4-6 exhibit performance declines of 13.7\% and 14.4\% in evolved MCP servers, respectively, accompanied by substantial increases in planning and reasoning errors. These findings highlight the vulnerability of LLM-driven workflows, establishing MCPEvol-Bench as a standard for evaluating agent adaptability in dynamic tool environments.
Chinese Translation
随着模型上下文协议（Model Context Protocol, MCP）服务器成为连接大型语言模型（LLMs）与外部工具的核心基础设施，现有基准利用真实世界的MCP服务器来评估LLM代理的工具使用能力。然而，这些基准忽视了MCP服务器中工具接口和功能的持续演变，导致评估结果存在缺陷，无法捕捉代理在变化工具环境中的适应能力。为填补这一空白，我们提出了 extbf{MCPEvol-Bench}，这是一个用于评估LLM代理在动态工具集演变下任务解决能力的新基准。受大规模实证研究的启发，我们提出了11种变异操作符，以模拟123个MCP服务器中现实的工具演变。我们在多个版本的MCP服务器上对12个最先进的LLM进行了基准测试，结果显示，即使是前沿模型在适应演变工具时也面临困难。例如，GPT-5.4和Claude-Sonnet-4-6在演变的MCP服务器中分别表现出13.7\%和14.4\%的性能下降，同时规划和推理错误显著增加。这些发现突显了基于LLM的工作流程的脆弱性，确立了MCPEvol-Bench作为评估代理在动态工具环境中适应能力的标准。
cs.AI / 51 / 2607.14658
TopoAgent: A Self-Evolving Topological Agent for Multimodal Scientific Reasoning
TopoAgent：一种自我演化的拓扑智能体用于多模态科学推理
Abstract
While Multimodal Large Language Models (MLLMs) excel in general tasks, rigorous scientific reasoning remains challenging due to the limitations of monolithic, linear planning. Such sequential designs often suffer from visual-semantic misalignment, long-context hallucinations, and brittle execution under fixed task granularity. We propose TopoAgent, a self-evolving topological framework that replaces linear trajectories with dynamic, state-isolated graph evolution. TopoAgent first employs a front-end decomposer to fracture complex queries into visually-grounded atoms. These atoms are organized into a Directed Acyclic Graph (DAG) based on their dependencies, enabling strict context isolation to shield the reasoning engine from irrelevant historical noise. Furthermore, we introduce adaptive atomic fission, which dynamically splits bottleneck nodes into finer-grained sub-atoms at runtime when tool capability boundaries are exceeded. Extensive experiments across mathematics, physics, and chemistry benchmarks demonstrate that TopoAgent significantly outperforms state-of-the-art linear agent frameworks, providing a robust, noise-resistant, and self-correcting paradigm for autonomous scientific reasoning.
Chinese Translation
尽管多模态大型语言模型（MLLMs）在一般任务中表现出色，但由于单一线性规划的局限性，严格的科学推理仍然面临挑战。这种顺序设计常常遭受视觉-语义不对齐、长上下文幻觉以及在固定任务粒度下脆弱的执行。我们提出了TopoAgent，一种自我演化的拓扑框架，它用动态的、状态隔离的图演化替代了线性轨迹。TopoAgent首先采用前端分解器将复杂查询分解为视觉基础的原子。这些原子根据其依赖关系组织成一个有向无环图（DAG），实现严格的上下文隔离，以保护推理引擎免受无关历史噪声的干扰。此外，我们引入了自适应原子裂变，当工具能力边界被超越时，它会在运行时动态地将瓶颈节点拆分为更细粒度的子原子。在数学、物理和化学基准测试中进行的广泛实验表明，TopoAgent显著优于最先进的线性智能体框架，为自主科学推理提供了一种强大、抗噪声和自我纠正的范式。
cs.AI / 52 / 2607.14661
SmartRAG: Native Graph-Based RAG for Mobile Device
SmartRAG：面向移动设备的原生图基RAG
Abstract
Deploying large language models (LLMs) as personal assistants on mobile devices demands privacy, low latency, and offline availability, yet the computational cost of giant models clashes with strict edge-hardware budgets. We argue that this tension cannot be resolved by model compression alone; it requires decomposing on-device intelligence into complementary functional roles. We present SmartRAG, a fully on-device framework that organizes an intelligent assistant around four coordinated modules -- Perception, Memory, Focus, and Thinking. At the core of SmartRAG is EvoNER, a continually learnable named-entity recognizer that incrementally expands its label inventory through teacher-distilled updates, enabling the system to absorb previously unseen entity types without retraining the backbone LLM. Extracted knowledge is stored in MRGraph, a three-layer provenance-preserving knowledge graph, and retrieved at query time through a hybrid pipeline combining graph traversal, lexical matching, and dense semantic search. The on-device LLM is invoked only for high-value semantic operations -- labeling, planning, and answer synthesis -- keeping inference costs bounded. Experiments on four QA benchmarks (TriviaQA, Natural Questions, HotpotQA, MultiHopQA) show that SmartRAG with a quantized 1.7B-parameter backbone achieves multi-hop reasoning performance competitive with models up to 18$\times$ larger, while running entirely on commodity smartphones within practical memory and latency envelopes.
Chinese Translation
在移动设备上将大型语言模型（LLMs）作为个人助手进行部署，需要考虑隐私、低延迟和离线可用性，但巨型模型的计算成本与严格的边缘硬件预算相冲突。我们认为，仅通过模型压缩无法解决这种矛盾；它需要将设备上的智能分解为互补的功能角色。我们提出了SmartRAG，这是一个完全在设备上运行的框架，围绕四个协调模块组织智能助手——感知、记忆、聚焦和思考。SmartRAG的核心是EvoNER，一个可持续学习的命名实体识别器，通过教师蒸馏更新逐步扩展其标签库存，使系统能够在不重新训练基础LLM的情况下吸收以前未见过的实体类型。提取的知识存储在MRGraph中，这是一个三层的保留来源的知识图谱，并通过结合图遍历、词汇匹配和密集语义搜索的混合管道在查询时检索。仅在高价值语义操作（如标记、规划和答案合成）时调用设备上的LLM，从而保持推理成本在可控范围内。在四个问答基准（TriviaQA、Natural Questions、HotpotQA、MultiHopQA）上的实验表明，使用量化的1.7B参数基础模型的SmartRAG实现了与高达18倍更大模型相竞争的多跳推理性能，同时完全在普通智能手机上运行，符合实际的内存和延迟要求。
cs.AI / 53 / 2607.14673
Project Kaleidoscope: Contextual, Human-Aligned Evaluation for Real-World AI Applications
万花筒项目：面向现实世界人工智能应用的情境化、人类对齐评估
Abstract
Evaluations (Evals) are a deployment bottleneck for real-world AI applications: public benchmarks rarely match a team's users, context, or policies, and human review is often tedious to scale. Motivated by our work with AI applications in the public sector, this project addresses recurring evaluation challenges encountered when applications must satisfy local policy and governance requirements. We present Kaleidoscope, an integrated workflow for contextual functional evaluation that links persona-based test generation, contextualized rubrics, and human review for reliability-gated automated scoring. Generated test cases are scored against application-specific rubrics; human annotations provide reviewable labels; and LLM judges automate scoring only when their agreement with those labels meets a configured threshold. Kaleidoscope is therefore a practical, inspectable, iterative workflow for product teams. We report early evidence from a three-week pilot across four organizational use cases and custom-rubric judge experiments on 108 annotated Q\&A pairs spanning four domains and 14 evaluation dimensions. The results highlight useful features for end-to-end reliable, automated scoring.
Chinese Translation
评估（Evals）是现实世界人工智能应用中的一个部署瓶颈：公共基准测试很少与团队的用户、情境或政策相匹配，而人工审查通常难以规模化。受到我们在公共部门人工智能应用工作启发，本项目旨在解决当应用必须满足地方政策和治理要求时所遇到的反复评估挑战。我们提出了万花筒（Kaleidoscope），这是一个集成的情境功能评估工作流程，连接了基于角色的测试生成、情境化评分标准和人类审查，以实现可靠性门控的自动评分。生成的测试用例根据特定应用的评分标准进行评分；人类注释提供可审查的标签；而大型语言模型（LLM）评审者仅在其与这些标签的协议达到配置阈值时才自动评分。因此，万花筒是一个实用的、可检查的、迭代的工作流程，适用于产品团队。我们报告了在四个组织用例和108对跨越四个领域和14个评估维度的标注问答对上进行的为期三周的试点项目的早期证据。结果突显了端到端可靠的自动评分的有用特性。
cs.AI / 54 / 2607.14682
Stop Thinking, Start Looking: Efficient Post-Training for Multimodal Document Question Answering via Reasoning-Free Alignment
停止思考，开始观察：通过无推理对齐实现多模态文档问答的高效后训练
Abstract
Efficient multimodal document question answering with explicit visual grounding, locating the precise document region that supports each answer remains an open challenge. Current approaches bifurcate into Supervised Fine-Tuning (SFT), which requires large annotated datasets and reaches optimization plateaus, and reasoning-centric Reinforcement Learning (RL), which depends on verbose intermediate traces that inflate inference token cost without clear benefit. We introduce Perception-RFT, a training framework that applies Group Relative Policy Optimization (GRPO) to multimodal document QA, bypassing intermediate reasoning tokens to directly align visual features with structured grounding outputs. To rigorously evaluate the necessity of reasoning, we construct a reasoning variant under identical reward settings. We find that reasoning-enabled models suppress their reasoning traces during training, converging to direct perception-based policies at the 4B parameter scale, reducing per-query inference token length by more than 60%, while reasoning-enabled RL underperforms perception-only training. Through a fine-grained analysis of Qwen3-VL-4B optimization dynamics, we confirm that SFT saturation and cold-start RL instability established in text-domain post-training extend to multimodal, and identify a previously uncharacterized Grounding Divergence: a selective trade-off between semantic robustness and geometric precision on two out of distribution (OOD) benchmarks (4,828 samples) under joint RL optimization. We further show that an early SFT$\rightarrow$RL transition achieves comparable precision with 65% less training data.
Chinese Translation
高效的多模态文档问答需要明确的视觉基础，定位支持每个答案的精确文档区域仍然是一个未解决的挑战。目前的方法分为两类：一种是监督微调（Supervised Fine-Tuning, SFT），需要大量标注数据集并达到优化平台；另一种是以推理为中心的强化学习（Reinforcement Learning, RL），依赖冗长的中间轨迹，这增加了推理令牌的成本而没有明显的好处。我们提出了感知-RFT（Perception-RFT），一个将群体相对策略优化（Group Relative Policy Optimization, GRPO）应用于多模态文档问答的训练框架，绕过中间推理令牌，直接将视觉特征与结构化基础输出对齐。为了严格评估推理的必要性，我们在相同的奖励设置下构建了一个推理变体。我们发现，启用推理的模型在训练过程中抑制其推理轨迹，在4B参数规模下收敛到基于直接感知的策略，将每个查询的推理令牌长度减少超过60%，而启用推理的强化学习表现不如仅基于感知的训练。通过对Qwen3-VL-4B优化动态的细致分析，我们确认文本领域后训练中建立的SFT饱和和冷启动RL不稳定性也扩展到多模态，并识别出一个之前未被描述的基础偏差（Grounding Divergence）：在联合RL优化下，在两个分布外（OOD）基准（4,828个样本）上，语义鲁棒性与几何精度之间的选择性权衡。我们进一步表明，早期的SFT→RL转变在训练数据减少65%的情况下实现了可比的精度。
cs.AI / 55 / 2607.14683
InCarEmo: A Multimodal Dataset for In-Cabin Emotion Recognition and Driver State Monitoring
InCarEmo：用于车内情感识别和驾驶员状态监测的多模态数据集
Abstract
Understanding driver emotion and state is critical for the next generation of intelligent in-cabin systems that ensure safety and enhance human-vehicle interaction. However, existing public datasets for in-cabin affective computing are largely limited to visual modalities and rarely include conversational information, making it difficult to capture the linguistic and interactive cues underlying driver emotion. To address these gaps, we introduce InCarEmo, a multimodal dataset for in-cabin emotion recognition and driver state monitoring. InCarEmo integrates RGB and infrared video, in-cabin audio, and dialogue text collected from scripted in-cabin scenarios designed to simulate realistic driver behaviors, covering diverse lighting conditions and driving contexts. The dataset supports three primary tasks: 1) multimodal emotion recognition, 2) fatigue detection, and 3) distraction monitoring. In addition to the original Chinese data, we construct an auxiliary English benchmark to support preliminary cross-lingual evaluation. We provide a unified benchmark with extensive baseline results across unimodal and multimodal methods, including analyses under modality-missing and noise conditions. Experimental results demonstrate the benefits of multimodal fusion and reveal remaining challenges under real-world noise and low-light conditions. By releasing InCarEmo, we aim to establish a comprehensive foundation for robust, interpretable, and human-centric in-cabin affective understanding, promoting safer and more empathetic driver-vehicle interaction.
Chinese Translation
理解驾驶员的情感和状态对于下一代智能车内系统至关重要，这些系统旨在确保安全并增强人车交互。然而，现有的公共车内情感计算数据集主要限于视觉模态，且很少包含对话信息，这使得捕捉驾驶员情感背后的语言和互动线索变得困难。为了解决这些问题，我们引入了InCarEmo，一个用于车内情感识别和驾驶员状态监测的多模态数据集。InCarEmo整合了RGB和红外视频、车内音频以及从设计用于模拟真实驾驶行为的剧本车内场景中收集的对话文本，涵盖了多种照明条件和驾驶环境。该数据集支持三个主要任务：1）多模态情感识别，2）疲劳检测，3）分心监测。除了原始的中文数据外，我们还构建了一个辅助的英语基准，以支持初步的跨语言评估。我们提供了一个统一的基准，包含了单模态和多模态方法的广泛基线结果，包括在模态缺失和噪声条件下的分析。实验结果表明多模态融合的优势，并揭示了在现实世界噪声和低光照条件下仍然存在的挑战。通过发布InCarEmo，我们旨在建立一个全面的基础，以实现稳健、可解释和以人为本的车内情感理解，促进更安全和更具同理心的驾驶员与车辆之间的互动。
cs.AI / 56 / 2607.14756
AI vs Human Expert Reasoning: Assessing Agreements in Building Typology Predictions based on Street View Imagery
人工智能与人类专家推理的对比：基于街景图像的建筑类型预测一致性评估
Abstract
This research investigates the potential of Vision-Language Models (VLMs) to infer building typologies: Construction, Current Use, and Storeys from Google Street View (GSV) images. Predictions generated by VLMs are compared with inference by human experts (civil engineers and architects) as a source of manually labelled ground-truth data. We evaluate several state-of-the-art VLMs, including GPT-4o, Claude 3.5 Sonnet, and Gemini 2.0 Flash. By applying different scaling strategies and prompting techniques, we found that Chain-of-Thought prompts provide an overall more stable model performance. We also investigate the reasoning behind VLMs' building-typology predictions by examining the probabilities of keywords appearing in AI explanations. This enabled us to analyse patterns in these reasonings and identify key themes driving both agreements and disagreements between VLM and expert labels. We find that AI tends to focus on visual indicators, whereas human experts place greater emphasis on broader contextual cues and domain knowledge, in addition to visual cues. Overall, VLM can approximate experts' capability in building-typology classification at scale, with an average accuracy of approximately 70%. The study demonstrates the VLM's potential for AI automation in tasks that require pattern recognition and object identification in an urban context. AI have the potential to serve as complementary and collaborative tools for urban analysis, leveraging their strengths in understanding visual patterns. This study contributes to the exploration of the efficiency and scalability of AI visual prediction and provides insights into the reasoning processes that could support automation processes in urban analysis and prediction.
Chinese Translation
本研究探讨了视觉-语言模型（Vision-Language Models, VLMs）推断建筑类型的潜力，包括建筑类型、当前用途和楼层数，基于谷歌街景图像（Google Street View, GSV）。我们将VLM生成的预测与人类专家（如土木工程师和建筑师）的推断进行比较，以作为手动标注的真实数据来源。我们评估了几种最先进的VLM，包括GPT-4o、Claude 3.5 Sonnet和Gemini 2.0 Flash。通过应用不同的扩展策略和提示技术，我们发现思维链（Chain-of-Thought）提示提供了整体更稳定的模型性能。我们还通过检查AI解释中关键词出现的概率，研究了VLM建筑类型预测背后的推理。这使我们能够分析这些推理中的模式，并识别出推动VLM与专家标签之间一致性和不一致性的关键主题。我们发现，人工智能倾向于关注视觉指标，而人类专家则更重视更广泛的上下文线索和领域知识，除了视觉线索外。总体而言，VLM在建筑类型分类方面能够在规模上接近专家的能力，平均准确率约为70%。该研究展示了VLM在需要模式识别和物体识别的城市背景任务中的AI自动化潜力。人工智能有潜力作为城市分析的补充和协作工具，利用其在理解视觉模式方面的优势。本研究为探索AI视觉预测的效率和可扩展性做出了贡献，并提供了可能支持城市分析和预测自动化过程的推理过程的见解。
cs.AI / 57 / 2607.14782
Global Index on Responsible AI: 2026 Report
负责任人工智能全球指数：2026年报告
Abstract
Grounded in human rights-based frameworks such as the UNESCO Recommendation on the Ethics of AI, the Global Index on Responsible AI (GIRAI) examines how countries translate responsible AI commitments into enforceable protections, institutional capacity, and redress mechanisms. GIRAI 2026 assesses these across five dimensions: Inclusion and Diversity, Ethics and Sustainability, Labour and Skills, Trust and Safety, and AI Use in Public Service. A global network of 135 country-level researchers collected and assessed 68,138 data points on 38 indicators organised across three pillars: government AI policy and implementation (17 indicators), civil society engagement (5), enabling conditions (15), and documented cases of government deployment of unacceptable-risk AI systems. The data covers November 2023 to September 2025. A score of 100 is derived from the indicators to rank all countries. Findings show that while responsible AI governance is expanding, with 126 of 135 countries having at least one government policy or initiative across the 17 AI Policy indicators, this does not often translate into meaningful protection. For instance, Global South countries account for 203 of 306 new cases of indicators with frameworks since the first edition, yet 78% of their frameworks remain non-binding compared with 42% in the Global North. Government commitment to AI governance also does not extend to their own algorithms: whereas Transparency and Explainability is the strongest performing indicator, with 58% of countries having some framework, only 18% require Public Disclosure of Government Algorithms. Credible evidence of government deployment of unacceptable-risk AI systems was also found in 35 countries. These findings show that responsible AI governance must move beyond framework adoption toward enforceable rights-based protections, resourced oversight institutions, and accessible redress.
Chinese Translation
基于人权框架，如联合国教科文组织关于人工智能伦理的建议，负责任人工智能全球指数（GIRAI）考察各国如何将负责任的人工智能承诺转化为可执行的保护措施、机构能力和救济机制。GIRAI 2026在五个维度上进行评估：包容性与多样性、伦理与可持续性、劳动与技能、信任与安全，以及公共服务中的人工智能使用。一个由135个国家级研究人员组成的全球网络收集并评估了68,138个数据点，这些数据点基于38个指标，分为三个支柱：政府人工智能政策与实施（17个指标）、公民社会参与（5个指标）、促进条件（15个指标），以及记录的政府部署不可接受风险人工智能系统的案例。数据覆盖时间为2023年11月至2025年9月。根据这些指标得出的100分评分用于对所有国家进行排名。研究结果表明，尽管负责任的人工智能治理正在扩展，135个国家中有126个国家在17个人工智能政策指标上至少有一项政府政策或倡议，但这并不常常转化为有意义的保护。例如，自第一版以来，全球南方国家在具有框架的指标中新增加了306个案例中的203个，但与全球北方的42%相比，78%的框架仍然是非约束性的。政府对人工智能治理的承诺也没有延伸到自身的算法：尽管透明度和可解释性是表现最强的指标，58%的国家拥有某种框架，但只有18%要求公开政府算法的披露。在35个国家中还发现了政府部署不可接受风险人工智能系统的可信证据。这些发现表明，负责任的人工智能治理必须超越框架的采用，朝向可执行的基于权利的保护、资源充足的监督机构和可获得的救济措施。
cs.AI / 58 / 2607.14791
Transcoders for Investigating Deception in Language Models
用于研究语言模型中欺骗行为的转码器
Abstract
Transcoders have recently emerged as a promising approach for mechanistic interpretability (MI), enabling circuit-level analysis of model behaviour. In this paper, we investigate the use of transcoders to analyse deceptive behaviour in language models, a behaviour that poses a safety and security risk. Using a Qwen3-4B model with pre-trained transcoders, specifically per-layer transcoders (PLTs), we construct attribution graphs that capture feature activations and inter-feature dependencies, allowing circuit-level analysis of deception. Through feature steering and circuit analysis, we identified a dictionary of deception-related features and show that these features exert a stronger influence on deceptive outputs, as they produce predictable shifts between deceptive and non-deceptive responses. These findings suggest that deception emerges from internal model mechanisms and highlight the potential of transcoders for behavioural monitoring and early detection of security vulnerabilities related to malicious behaviours in language models.
Chinese Translation
转码器最近作为一种有前景的机制可解释性（MI）方法出现，使得对模型行为的电路级分析成为可能。本文探讨了使用转码器分析语言模型中的欺骗行为，这种行为构成了安全和保密风险。我们使用预训练的转码器，特别是每层转码器（PLTs），构建了归因图，以捕捉特征激活和特征间依赖关系，从而实现对欺骗行为的电路级分析。通过特征引导和电路分析，我们识别出了一组与欺骗相关的特征，并表明这些特征对欺骗输出的影响更强，因为它们在欺骗和非欺骗响应之间产生可预测的变化。这些发现表明，欺骗行为源于内部模型机制，并突显了转码器在行为监测和早期检测与语言模型中恶意行为相关的安全漏洞方面的潜力。
cs.AI / 59 / 2607.14800
CrimeNER Demo: Named-Entity Recognition in the Crime Domain
CrimeNER 演示：犯罪领域的命名实体识别
Abstract
We present CrimeNER Demo, an AI-powered platform that enables us to extract general crime-related information from documents and classify them into entity types with two levels of granularity. We provide pretrained NER models on the CrimeNER database, and we give the possibility to users to provide their own annotated data to train models for their own specific cases. This demonstrator aims to promote crime-related NER research and provides a practical tool to automatically extract crime information for researchers and law enforcement agencies. The demonstrator includes: i) Pretrained NER models on the crime domain; ii) Possibility to finetune the models on specific data annotated by the user; and iii) An automatic pipeline to extract and annotate crime entities from documents. The demo platform, a tutorial to run the demo, and a video demonstration are publicly available on GitHub.
Chinese Translation
我们介绍了 CrimeNER 演示，这是一个基于人工智能的平台，使我们能够从文档中提取一般的与犯罪相关的信息，并将其分类为具有两级粒度的实体类型。我们提供了在 CrimeNER 数据库上预训练的命名实体识别（NER）模型，并允许用户提供他们自己的标注数据，以便为他们特定的案例训练模型。该演示旨在促进与犯罪相关的 NER 研究，并为研究人员和执法机构提供一个自动提取犯罪信息的实用工具。该演示包括：i) 针对犯罪领域的预训练 NER 模型；ii) 在用户标注的特定数据上微调模型的可能性；以及 iii) 从文档中自动提取和标注犯罪实体的管道。演示平台、运行演示的教程和视频演示均已在 GitHub 上公开提供。
cs.AI / 60 / 2607.14886
Reachability-Aware Pretraining for Efficient Target-Oriented Path Exploration in Temporal Knowledge Graph Reasoning
面向可达性的预训练方法在时序知识图推理中的高效目标导向路径探索
Abstract
Temporal Knowledge Graph (TKG) reasoning under the extrapolation setting focuses on forecasting future time-stamped events (facts) from historical data in a temporal knowledge graph. Existing approaches, reinforcement learning (RL)-based multi-hop reasoning methods are prominent for TKG reasoning because they produce human-interpretable predictions via explicit multi-hop path tracing. However, during RL training, rewards are typically sparse, and exploration is highly inefficient due to the vast, time-evolving action space. These issues hinder efficient training and often limit overall performance. To address these challenges, we propose RAPTOR (Reachability-Aware Pretraining for Efficient Target-Oriented Path Exploration), a self-supervised pretraining method that injects a reachability-aware inductive bias to the agent. By learning to estimate the reachability of candidate actions to the target entity, RAPTOR reduces exploration over unpromising paths and provides a strong initialization for downstream RL fine-tuning. Experimental results on the ICEWS14, ICEWS05-15, and ICEWS18 datasets demonstrate that RAPTOR pretraining markedly improves the training efficiency and consistently outperforms conventional baselines, establishing it as an effective approach for enhancing RL-based multi-hop reasoning methods for TKG reasoning.
Chinese Translation
在外推设置下，时序知识图（TKG）推理关注于从历史数据中预测未来的时间戳事件（事实）。现有的方法中，基于强化学习（RL）的多跳推理方法在TKG推理中表现突出，因为它们通过明确的多跳路径追踪生成可被人类理解的预测。然而，在RL训练过程中，奖励通常稀疏，且由于庞大且随时间演变的动作空间，探索效率极低。这些问题阻碍了高效训练，并常常限制整体性能。为了解决这些挑战，我们提出了RAPTOR（面向可达性的预训练方法），这是一种自监督预训练方法，向代理注入了可达性意识的归纳偏置。通过学习估计候选动作到目标实体的可达性，RAPTOR减少了对不具前景路径的探索，并为下游RL微调提供了强有力的初始化。在ICEWS14、ICEWS05-15和ICEWS18数据集上的实验结果表明，RAPTOR预训练显著提高了训练效率，并始终优于传统基线，确立了其作为增强TKG推理中基于RL的多跳推理方法的有效途径。
cs.AI / 61 / 2607.14890
Proof-or-Stop: Don't Trust the Agent, Trust the Evidence -- Loop Engineering for Verifiable Evidence-Gated Lifecycle Control
证明或停止：不要信任代理，信任证据——可验证证据门控生命周期控制的循环工程
Abstract
Autonomous coding agents increasingly execute multi-step software work, but lifecycle states such as reviewed, tested, DONE, and ready-to-merge remain claims unless supported by current evidence. We present Proof-or-Stop Lifecycle Control, a method that permits lifecycle transitions only when fresh, tracked-source-state-bound, mechanically verifiable evidence satisfies the relevant gate. The method treats agent outputs as claims rather than lifecycle state, and uses proof operationally to mean gate-admissible evidence under a stated trust model, not semantic program correctness. We evaluate an open-source implementation through mechanism tests, a powered control-policy ablation, and operated self-application evidence. The unattended-loop engine passed 10 of 10 scenarios with zero false-DONE, and local-key receipt bundles rejected 18 tamper classes with zero false accepts. In a 9,240-cell ablation, the pre-registered A4 versus A2-prime comparison reduced visible-pass/hidden-fail amplification from 31 of 1,800 injected cells under a compute-budgeted naive loop to 2 of 1,800 under the gated loop, a 1.6 percentage-point improvement in not-amplified rate with a 95 percent confidence interval of [0.8, 2.5]. A near-compute A3 versus A4 comparison, 14 of 1,800 versus 2 of 1,800, indicates that the gain is associated with enforcing review as a lifecycle gate rather than merely adding a reviewer. The self-application corpus contains 565 stories and 1,007 review findings, with 94.8 percent resolved, plus a 68-row high/critical cross-vendor exhibit. These results support Proof-or-Stop as a model-agnostic, host-neutral control layer for deciding which autonomous-agent claims a lifecycle may act on. The evaluation is limited to one model family, 24 ablation tasks, and a self-hosted corpus.
Chinese Translation
自主编码代理越来越多地执行多步骤的软件工作，但生命周期状态如审核、测试、完成和准备合并，除非有当前证据支持，否则仍然只是声明。我们提出了证明或停止生命周期控制（Proof-or-Stop Lifecycle Control）方法，该方法仅在新鲜、受追踪的源状态绑定、机械可验证的证据满足相关门控时，允许生命周期转换。该方法将代理输出视为声明而非生命周期状态，并在操作上使用证明来表示在所述信任模型下可接受的门控证据，而非语义程序的正确性。我们通过机制测试、控制策略消融实验和操作自应用证据评估了一个开源实现。无人值守的循环引擎在10个场景中全部通过，且没有出现虚假完成（false-DONE），本地密钥收据包拒绝了18类篡改，且没有虚假接受。在9240个单元的消融实验中，预注册的A4与A2-prime比较将可见通过/隐藏失败的放大率从在计算预算的简单循环下的1800个注入单元中的31，减少到在门控循环下的1800个单元中的2，未放大率的改善为1.6个百分点，95%置信区间为[0.8, 2.5]。近计算的A3与A4比较，1800个单元中的14与2，表明该增益与将审核作为生命周期门控的强制执行有关，而不仅仅是增加审核者。自应用语料库包含565个故事和1007个审核发现，解决率为94.8%，此外还有68行高/关键跨供应商展品。这些结果支持证明或停止作为一个模型无关、主机中立的控制层，用于决定哪个自主代理的声明可以作用于生命周期。评估仅限于一个模型家族、24个消融任务和一个自托管语料库。
cs.AI / 62 / 2607.14957
Contextualized Early Detection of Online Firestorms: A Sequential LLM-Based Approach
在线火灾风暴的情境化早期检测：一种基于序列化大语言模型的方法
Abstract
Online firestorms are rapid collective escalations of highly negative user-generated content and may cause substantial reputational and economic damage. Existing detectors usually work with volume signals, sentiment scores, or predefined linguistic features. Such signals are useful, but they capture contextual meaning shifts in evolving discussion threads only indirectly. This paper proposes an LLM-based detection system with two operating modes. The first mode classifies complete Reddit threads retrospectively by combining local chunk-level assessments into a thread-level judgment. The second mode processes threads sequentially and issues early warnings when a sliding window exceeds calibrated thresholds. In this mode, the language model estimates three firestorm indicators: negativity share, escalation level, and contributor count. On a balanced Reddit dataset, the global mode achieves strong classification performance, while the early warning mode reaches high recall and detects escalating threads after only a small number of comments and distinct contributors. The results indicate that LLMs can be used not only for static judgment tasks, but also as repeated estimators in context-aware monitoring of social media discourse.
Chinese Translation
在线火灾风暴是用户生成的高度负面内容的快速集体升级，可能造成显著的声誉和经济损失。现有的检测器通常依赖于量级信号、情感评分或预定义的语言特征。这些信号虽然有用，但只能间接捕捉到不断演变的讨论主题中的情境意义变化。本文提出了一种基于大语言模型（LLM）的检测系统，具有两种操作模式。第一种模式通过将局部块级评估结合为线程级判断，回顾性地对完整的Reddit讨论线程进行分类。第二种模式则是顺序处理线程，当滑动窗口超过校准阈值时发出早期警报。在此模式下，语言模型估计三个火灾风暴指标：负面比例、升级水平和贡献者数量。在一个平衡的Reddit数据集中，全球模式实现了强大的分类性能，而早期警报模式在仅有少量评论和不同贡献者的情况下，达到了高召回率并检测到升级中的线程。结果表明，LLM不仅可以用于静态判断任务，还可以作为上下文感知社交媒体话语监测中的重复估计器。
cs.AI / 63 / 2607.14970
Explaining Process Control Optimisation Recommendations via GradientSHAP and Implicit Differentiation
通过 GradientSHAP 和隐式微分解释过程控制优化建议
Abstract
Automated optimisation is increasingly adopted in industrial processes, yet a trust gap persists between engineers who design these algorithms and operators who must act on their recommendations. Explainable AI methods like SHAP (SHapley Additive exPlanations) have transformed interpretability for machine learning predictions; optimisation outputs could benefit from similar techniques. We present an approach that integrates Implicit Function Theorem (IFT) based sensitivity analysis with SHAP attribution and narrative generation via Large Language Models (LLM), producing explanations tailored for operators. Our approach leverages IFT to compute exact parameter sensitivities $\partial p^*/\partial x$ from the optimality conditions, enabling efficient GradientSHAP computation. For an industrial High Pressure Grinding Roll (HPGR) control optimisation problem with 22 features, we achieve equivalent SHAP attributions (correlation $>$0.99 with KernelSHAP) with over 40$\times$ speedup, enabling real-time natural language explanations. We validate on industrial scenarios and present feedback from domain experts on generated explanations.
Chinese Translation
自动化优化在工业过程中的应用日益增加，但设计这些算法的工程师与必须根据其建议采取行动的操作员之间仍然存在信任差距。可解释人工智能方法如 SHAP（SHapley 加性解释）已改变了机器学习预测的可解释性；优化输出也可以从类似的技术中受益。我们提出了一种方法，将基于隐式函数定理（Implicit Function Theorem, IFT）的灵敏度分析与 SHAP 归因和通过大型语言模型（Large Language Models, LLM）生成叙述相结合，产生针对操作员量身定制的解释。我们的方法利用 IFT 从最优条件计算精确的参数灵敏度 $ rac{ ext{d} p^*}{ ext{d} x}$，从而实现高效的 GradientSHAP 计算。在一个具有 22 个特征的工业高压研磨辊（High Pressure Grinding Roll, HPGR）控制优化问题中，我们实现了等效的 SHAP 归因（与 KernelSHAP 的相关性 $>$0.99），并获得超过 40 倍的速度提升，使实时自然语言解释成为可能。我们在工业场景中进行了验证，并展示了领域专家对生成解释的反馈。
cs.AI / 64 / 2607.14975
CFM-Bench: A Unified Multi-Domain, Multi-Task Benchmark for Channel Foundation Models
CFM-Bench：一个统一的多领域、多任务基准测试，用于通道基础模型
Abstract
Channel foundation models (CFMs) are developing rapidly, with recent studies reporting benefits from pretraining across downstream wireless tasks. Yet CFMs are commonly evaluated in model-specific pipelines with different data, radio configurations, partitions, adaptation procedures, task definitions, and metrics. Reported comparisons therefore tend to show that pretraining improves over supervised training from scratch within one pipeline, but neither rank CFMs nor compare them fairly with task-specific models. We release CFM-Bench, a unified multi-domain, multi-task benchmark designed to address this gap. It curates six channel configurations spanning 3GPP statistical simulation, two independent ray-tracing pipelines, industrial and aerial measurements, and synchronized vehicular multimodal simulation. Official partitions isolate complete trajectories, measurement sessions, vehicle links, simulation realizations, or buffered spatial regions. CFM-Bench does not prescribe an external pretraining corpus or strategy; no benchmark split may be used for foundation-model pretraining, and the official training split is reserved exclusively for downstream fine-tuning. The benchmark additionally requires disclosure of all data used during model development and prohibits training-stage use of official test units. Six task groups are organized along three CFM application dimensions: physical-layer (PHY) channel intelligence, radio-access-network (RAN) decision intelligence, and integrated sensing and communication (ISAC). They cover CSI feedback, frequency and temporal channel extrapolation, propagation-state classification, current- and future-beam prediction, and single-frame and temporal localization. CFM-Bench provides a common substrate for comparing the transferability of channel representations across models, domains, and tasks.
Chinese Translation
通道基础模型（CFMs）正在快速发展，最近的研究报告显示预训练在下游无线任务中带来了好处。然而，CFMs 通常在特定模型的管道中进行评估，这些管道使用不同的数据、无线电配置、分区、适应过程、任务定义和指标。因此，报告的比较往往显示预训练在一个管道中优于从头开始的监督训练，但既没有对 CFMs 进行排名，也没有与特定任务模型进行公平比较。我们发布了 CFM-Bench，这是一个统一的多领域、多任务基准测试，旨在填补这一空白。它策划了六种通道配置，涵盖了 3GPP 统计仿真、两个独立的射线追踪管道、工业和空中测量，以及同步的车辆多模态仿真。官方分区隔离了完整的轨迹、测量会话、车辆链接、仿真实现或缓冲的空间区域。CFM-Bench 不规定外部预训练语料库或策略；任何基准分割都不得用于基础模型的预训练，官方训练分割仅保留用于下游微调。该基准还要求披露在模型开发过程中使用的所有数据，并禁止在训练阶段使用官方测试单元。六个任务组沿着三个 CFM 应用维度进行组织：物理层（PHY）通道智能、无线接入网络（RAN）决策智能，以及集成感知与通信（ISAC）。它们涵盖了 CSI 反馈、频率和时间通道外推、传播状态分类、当前和未来波束预测，以及单帧和时间定位。CFM-Bench 提供了一个共同的基础，用于比较通道表示在模型、领域和任务之间的可转移性。
cs.AI / 65 / 2607.14984
Demographically-Conditioned Synthetic Medical Images for Bias Mitigation and Bias Detection in Disease Classifiers
基于人口统计条件的合成医学图像用于疾病分类器中的偏差缓解和偏差检测
Abstract
Per-subgroup fairness audits of medical image classifiers face a sample-size problem: minority subgroups in held-out test sets have so few samples that the resulting confidence intervals on per-subgroup performance are wider than the bias the audit is meant to detect. We argue that a demographically-conditioned synthetic generator can do both: mitigate bias on the training side and detect bias on the evaluation side. Working on COVID-19 chest CT classification with an end-to-end fine-tuned Stable Diffusion 2.1 generator, we make two findings. For bias mitigation (training), a demographically-balanced synthetic cohort is most useful as a pretraining prior, not as joint augmentation: with the same fixed data, sequential pretraining followed by fine-tuning substantially outperforms joint augmentation, and the resulting classifier surpasses the full-real baseline at $\sim$$100\times$ real-data efficiency. For bias detection (evaluation), across five synthetic minority cohorts and five classifier seeds, the synthetic estimator reproduces the subgroup ranking of a well-powered real oracle (Spearman $\rho = 1.00$ on MCC and Recall) and gives the more reliable per-cell estimate where the small real test set runs out of samples. The synthetic cohort is therefore most useful in exactly the cells that fairness audits care about, as both a fix for and a measure of subgroup bias.
Chinese Translation
医学图像分类器的子群公平性审计面临样本量问题：在保留的测试集中，少数子群的样本数量极少，以至于所得到的子群性能置信区间比审计旨在检测的偏差要宽。我们认为，基于人口统计条件的合成生成器可以同时实现这两个目标：在训练阶段缓解偏差，在评估阶段检测偏差。我们在 COVID-19 胸部 CT 分类中，使用端到端微调的 Stable Diffusion 2.1 生成器，得出了两个发现。对于偏差缓解（训练），人口统计平衡的合成队列作为预训练先验最为有效，而非作为联合增强：在相同固定数据下，顺序预训练后再微调的效果显著优于联合增强，最终得到的分类器在约 $100 imes$ 的真实数据效率下超越了全真实基线。对于偏差检测（评估），在五个合成少数群体和五个分类器种子中，合成估计器再现了一个高效真实预言者的子群排名（Spearman $
ho = 1.00$ 在 MCC 和召回率上），并在小型真实测试集样本耗尽的情况下提供了更可靠的每个单元估计。因此，合成队列在公平性审计所关注的确切单元中最为有用，既可以作为修复子群偏差的手段，也可以作为衡量子群偏差的指标。
cs.AI / 66 / 2607.14998
Moral Attitudes of Sentient ASI towards Humanity and Implications for AGI Development
感知人工超级智能对人类的道德态度及其对人工通用智能发展的影响
Abstract
This paper suggests the adoption of a novel inversion in AI ethics: instead of asking how humans should treat artificial superintelligence (ASI), it examines how future sentient ASI may morally consider and evaluate humanity. We are not only designing intelligent systems but also shaping the initial conditions under which those systems form judgments about us. The paper proposes a preliminary set of post-human moral principles that may govern sentient ASI actions. The implication is that technical design choices (some are suggested), humanity's moral behaviour, and the essence of what it means to be human, may influence humanity's long-term standing in a post-ASI world.
Chinese Translation
本文建议在人工智能伦理中采用一种新颖的反转思维：与其问人类应如何对待人工超级智能（ASI），不如探讨未来的感知ASI可能如何道德地看待和评估人类。我们不仅在设计智能系统，还在塑造这些系统对我们进行判断的初始条件。本文提出了一套初步的后人类道德原则，可能会指导感知ASI的行为。其含义在于，技术设计选择（其中一些已被提出）、人类的道德行为以及人类的本质，可能会影响人类在后ASI世界中的长期地位。
cs.AI / 67 / 2607.15003
SMC-ES: Automated synthesis of formally verified control policies
SMC-ES：自动合成形式验证的控制策略
Abstract
The deployment of autonomous cyber-physical systems in safety-critical environments requires closed-loop control strategies (i.e., policies) that are not only performant but also provably safe and robust. While learning-based methodologies such as Reinforcement Learning offer flexible and scalable approaches to automatically synthesize such controllers, they typically lack the formal guarantees necessary for safe deployment. To bridge this gap, we propose a novel simulation-based methodology to automatically synthesize policies with formal guarantees regarding performance, safety, and robustness specifications. Specifically, given a set of properties to verify, a confidence parameter $\delta$ and an allowable failure probability $\varepsilon$, our method guarantees that the synthesized policy comes with a certificate: with confidence at least $1 - \delta$, the probability of encountering a scenario where the given properties are violated is at most $\varepsilon$. We demonstrate the feasibility of our approach by developing SMC-ES, an algorithm that integrates Evolutionary Strategies with Statistical Model Checking-based verification. We evaluate SMC-ES on a suite of continuous control tasks using Gymnasium and Safety Gymnasium testbeds. Results show that, at the price of a sustainable increase in computational cost, our algorithm provides formal guarantees regarding performance, safety, and robustness specifications, while performing competitively against leading model-free Deep Reinforcement Learning (DRL) and Safe-DRL baselines.
Chinese Translation
在安全关键环境中部署自主网络物理系统需要闭环控制策略（即政策），这些策略不仅要具备良好的性能，还必须在安全性和鲁棒性上有可证明的保障。虽然基于学习的方法（如强化学习）提供了灵活且可扩展的方式来自动合成此类控制器，但它们通常缺乏安全部署所需的形式保障。为了解决这一问题，我们提出了一种新颖的基于仿真的方法，自动合成在性能、安全性和鲁棒性规范方面具有形式保障的策略。具体而言，给定一组待验证的属性、一个置信参数 $ ext{δ}$ 和一个可接受的失败概率 $ ext{ε}$，我们的方法保证合成的策略附带一个证书：以至少 $1 - ext{δ}$ 的置信度，遇到违反给定属性的场景的概率至多为 $ ext{ε}$。我们通过开发 SMC-ES 算法，展示了我们方法的可行性，该算法将进化策略与基于统计模型检查的验证相结合。我们在使用 Gymnasium 和 Safety Gymnasium 测试平台的一系列连续控制任务上评估了 SMC-ES。结果表明，在可持续增加计算成本的情况下，我们的算法在性能、安全性和鲁棒性规范方面提供了形式保障，同时在与领先的无模型深度强化学习（Deep Reinforcement Learning, DRL）和安全深度强化学习（Safe-DRL）基准的竞争中表现出色。
cs.AI / 68 / 2607.15027
Man, Machine, and Masterpiece: Artistic Ownership in the AI Era
人、机器与杰作：人工智能时代的艺术所有权
Abstract
The integration of AI-driven systems in creative work has sparked debates among artists and legal communities about notions of ownership. Yet there remains little consensus on how ownership should be defined and attributed when human and AI contributions are intertwined. To provoke critical reflection on these tensions, we designed ArtSplit, a provotype that explicitly quantifies human and AI contributions across different stages of creative work. Rather than aiming to resolve ownership, the provotype was used to elicit artists' responses to the idea of attributing ownership through measurable actions in the creative workflow. We argue that quantification fails to align with artists' understandings of creative intent and agency, and that efforts to measure ownership risk diluting long-standing assumptions through which artists understand and practice creative work. This critique challenges the impulse to transform a historically and socially situated relation into a technical problem.
Chinese Translation
人工智能驱动系统在创作工作中的整合引发了艺术家和法律界对所有权概念的辩论。然而，当人类与人工智能的贡献交织在一起时，关于所有权应如何定义和归属仍然缺乏共识。为了激发对这些紧张关系的深思，我们设计了ArtSplit，这是一种原型，明确量化人类和人工智能在创作工作不同阶段的贡献。该原型并不旨在解决所有权问题，而是用于引导艺术家对通过可测量的创作流程中的行为来归属所有权的想法作出反应。我们认为，量化无法与艺术家对创作意图和自主性的理解相一致，而测量所有权的努力则有可能稀释艺术家理解和实践创作工作的长期假设。这一批评挑战了将历史上和社会上特定的关系转变为技术问题的冲动。
cs.AI / 69 / 2607.15079
BrainPilot: Automating Brain Discovery with Agentic Research
BrainPilot：通过自主研究自动化脑科学发现
Abstract
Understanding the brain increasingly depends on integrating evidence across scales, modalities, and disciplines. Addressing a single research question therefore requires a coordinated sequence of operations, from surveying prior work to executing analyses and interpreting results in light of domain knowledge. AI agents promise to accelerate this process, but current agents lack domain expertise in brain science, may fabricate claims, drift during multi-step reasoning, and offer few defined points for expert intervention. These failures are especially costly in brain science, where conclusions feed into downstream scientific claims and depend on laboratory-specific expertise and careful human judgment. We present \textbf{BrainPilot} a \textbf{fully open-source} multi-agent system that accelerates brain science research with traceable logs and agent-verified results. A principal investigator (PI) agent coordinates specialist agents grounded in curated domain knowledge: a unified brain science knowledge base containing 7{,}233 indexed items and a skill library of 72 reusable methodology units across seven research domains. Every major step is recorded in the Graph of Trace, an auditable record that links subgoals, tool use, evidence, and claims and allows researchers to follow and inspect the workflow. An Auditor agent further integrates fabrication checking into the workflow. For evaluation, we run three brain science tasks from Agents' Last Exam, introduce our own benchmark, \textbf{BrainPilotBench-v0}, and present additional end-to-end case studies. Across these evaluations, BrainPilot with an open-source backbone model attains performance comparable to state-of-the-art agent framework with less costs.
Chinese Translation
理解大脑越来越依赖于跨尺度、跨模态和跨学科整合证据。因此，解决单一研究问题需要协调的一系列操作，从调查先前的工作到执行分析，并根据领域知识解释结果。人工智能代理有望加速这一过程，但当前的代理在脑科学领域缺乏专业知识，可能会捏造声明，在多步骤推理中出现偏差，并且提供的专家干预点很少。这些失败在脑科学中尤其代价高昂，因为结论会影响下游科学声明，并依赖于实验室特定的专业知识和细致的人类判断。我们提出了 extbf{BrainPilot}，一个 extbf{完全开源}的多代理系统，通过可追溯的日志和代理验证的结果加速脑科学研究。主要研究者（PI）代理协调基于精心策划的领域知识的专业代理：一个统一的脑科学知识库，包含7,233个索引项目，以及一个涵盖七个研究领域的72个可重用方法单元的技能库。每个主要步骤都记录在可审计的追踪图中，链接子目标、工具使用、证据和声明，允许研究人员跟踪和检查工作流程。审计代理进一步将造假检查集成到工作流程中。为了评估，我们运行了来自Agents' Last Exam的三个脑科学任务，介绍了我们自己的基准 extbf{BrainPilotBench-v0}，并呈现了额外的端到端案例研究。在这些评估中，基于开源骨干模型的BrainPilot达到了与最先进的代理框架相当的性能，同时成本更低。
cs.AI / 70 / 2607.15105
Long-Context Fine-Tuning with Limited VRAM
有限显存下的长上下文微调
Abstract
Parameter-efficient fine-tuning reduces model and optimizer memory, but dense attention still makes long training sequences expensive. We combine Hierarchical Global Attention (HGA) with segment-wise backpropagation and tiered KV storage. Only the active segment remains differentiable in VRAM; older KV is detached into RAM or NVMe, and HGA loads a bounded set of exact historical tokens for each query block. On Qwen3-8B with 4-bit QLoRA and PG19, dense training on a 16 GB Quadro RTX 5000 fits 2,048 tokens but fails at 4,096, whereas HGA reaches 16,384 tokens with 15.28 GB peak VRAM. Under evaluation the same adapter runs through 131,072 tokens on this card; VRAM is not constant but grows gently with the resident chunk summaries, so RAM and NVMe capacity set the practical limit beyond these lengths. At the shared 2K training length, HGA-trained and dense-trained adapters obtain 2.7405 and 2.7383 nat under the same dense-attention readout, while the stock model obtains 2.9541. At this boundary HGA training is already marginally faster (217.75 vs. 207.02 tokens/s), and the HGA-to-dense throughput ratio improves from 1K to 2K; because HGA keeps the attended historical set per token approximately constant while dense work per token grows, we expect this lead to widen as context grows. Dense attention is used for the main quality and retrieval comparisons so that they measure the learned weights and remain compatible with standard generation frameworks. HGA can also be used for retrieval and generation; an optimized production-grade serving implementation is under development.
Chinese Translation
参数高效的微调减少了模型和优化器的内存需求，但密集注意力仍使得长训练序列变得昂贵。我们将层次全局注意力（Hierarchical Global Attention, HGA）与分段反向传播和分级键值存储相结合。只有活动段在显存中保持可微分；较旧的键值对被转移到随机存取存储器（RAM）或非易失性内存（NVMe），而HGA为每个查询块加载一个有限的确切历史令牌集。在使用4位QLoRA和PG19的Qwen3-8B模型上，在16GB的Quadro RTX 5000上进行密集训练时，能够适配2048个令牌，但在4096个令牌时失败，而HGA则在15.28GB的峰值显存下达到了16384个令牌。在评估中，同一适配器在该卡上处理了131072个令牌；显存并非恒定，而是随着驻留块摘要的增加而缓慢增长，因此RAM和NVMe的容量设定了超出这些长度的实际限制。在共享的2000个训练长度下，经过HGA训练和密集训练的适配器在相同的密集注意力读出下分别获得了2.7405和2.7383 nat，而标准模型则获得了2.9541。在这一边界上，HGA训练的速度已经略快（217.75 vs. 207.02令牌/秒），且HGA与密集训练的吞吐比从1000提升至2000；因为HGA在每个令牌上保持了大致恒定的关注历史集，而每个令牌的密集工作量在增加，我们预计随着上下文的增长，这一领先优势将进一步扩大。密集注意力用于主要的质量和检索比较，以便它们能够测量学习到的权重，并与标准生成框架保持兼容。HGA也可以用于检索和生成；一个优化的生产级服务实现正在开发中。
cs.AI / 71 / 2607.15142
Concept-Guided Spatial Regularization for World Models in Atari Pong
基于概念引导的空间正则化在Atari Pong中的世界模型
Abstract
World models are usually evaluated as components of model-based reinforcement learning (MBRL) systems, while the world models themselves are rarely studied in isolation. We examine five representative visual world-model agents in Atari Pong: DreamerV3, DIAMOND, TWISTER, Simulus, and STORM. After reproducing their training pipelines and matching the reported agent performance, we freeze the learned world models and evaluate them with a closed-loop rollout diagnostic: a policy trained separately from the corresponding MBRL agent interacts with each frozen model, and the generated video trajectories are inspected for visual and dynamical errors. Across all five models, the rollouts contain clear failures, including ball disappearance, incorrect ball motion, and invalid ball-paddle interactions. Beyond visual trajectories, we further evaluate them with pixel-space zero-shot MBRL, where a new policy is trained entirely inside a frozen world model and then evaluated in the real environment. Across all five models, the resulting policies substantially underperform those produced by the corresponding original MBRL training pipelines. The gap is particularly large for DreamerV3, whose mean return drops from -5.5 to -20.9, near the minimum Pong return of -21. We hypothesize that insufficient modeling of task-critical concepts, such as the ball in Pong, may contribute to these failures. We therefore propose Concept-Guided Spatial Regularization (CGSReg), an auxiliary pixel reconstruction loss applied to segmented concept regions. Experiments show that CGSReg improves both closed-loop rollouts and pixel-space zero-shot MBRL in DreamerV3, DIAMOND, and TWISTER. Its effects vary across the remaining models and evaluation metrics, indicating that CGSReg alone does not address all world-model bottlenecks.
Chinese Translation
世界模型通常作为基于模型的强化学习（MBRL）系统的组成部分进行评估，而世界模型本身则很少单独研究。我们考察了五个在Atari Pong中具有代表性的视觉世界模型代理：DreamerV3、DIAMOND、TWISTER、Simulus和STORM。在重现它们的训练流程并匹配报告的代理性能后，我们冻结学习到的世界模型，并使用闭环回放诊断进行评估：一个与相应MBRL代理分开训练的策略与每个冻结模型进行交互，生成的视频轨迹被检查以发现视觉和动态错误。在所有五个模型中，回放中都存在明显的失败，包括球消失、不正确的球运动和无效的球-球拍交互。除了视觉轨迹外，我们还通过像素空间的零-shot MBRL进行进一步评估，其中一个新的策略完全在冻结的世界模型内部进行训练，然后在真实环境中进行评估。在所有五个模型中，得到的策略的表现明显低于相应原始MBRL训练流程产生的策略。对于DreamerV3，这一差距尤其明显，其平均回报从-5.5降至-20.9，接近Pong的最低回报-21。我们假设对任务关键概念（如Pong中的球）的建模不足可能导致这些失败。因此，我们提出了基于概念引导的空间正则化（CGSReg），这是一种应用于分割概念区域的辅助像素重建损失。实验表明，CGSReg在DreamerV3、DIAMOND和TWISTER中改善了闭环回放和像素空间零-shot MBRL。其效果在剩余模型和评估指标中有所不同，表明CGSReg单独并未解决所有世界模型的瓶颈。
cs.AI / 72 / 2607.15164
The Industrialization of Research ; On AI-Driven Science and Its Consequences
研究的工业化；关于人工智能驱动的科学及其后果
Abstract
Artificial intelligence is transforming scientific research - not merely as a more powerful instrument, but as an autonomous participant in the research cycle itself. This transition constitutes, in the most precise sense of the term, the industrialization of research: a shift from a craft model, in which knowledge, method, and judgment are embedded in the researcher, to a pipeline model, in which these steps are decomposed, automated, and supervised. The US Department of Energy's Genesis Mission is the most ambitious current instantiation of this shift, but the fundamental questions it raises extend far beyond any single program. This essay examines seven such questions: the erosion of the intergenerational transmission of scientific competence; the growing opacity of AI-generated theories; the collapse of peer evaluation under a flood of machine-generated output; the unproven capacity of AI for paradigm-shifting discovery; the capture of the scientific agenda by political and industrial actors; the compounding of systematic errors in closed-loop pipelines; and the structural bifurcation of the global research community into incommensurable tiers. These concerns do not constitute an argument against AI-driven science - whose demonstrated potential is real and significant. They constitute the conditions under which that potential can be responsibly pursued.
Chinese Translation
人工智能正在改变科学研究——不仅仅作为一种更强大的工具，而是作为研究循环中的一个自主参与者。这一转变在最精确的意义上构成了研究的工业化：从一种工艺模型转变为一种管道模型，在这种模型中，知识、方法和判断被嵌入研究者中，而这些步骤则被分解、自动化和监督。美国能源部的Genesis Mission是这一转变当前最雄心勃勃的实例，但它所提出的基本问题远远超出了任何单一项目。本文探讨了七个此类问题：科学能力的代际传递的侵蚀；人工智能生成理论的日益不透明；在机器生成输出的洪流下同行评审的崩溃；人工智能在范式转变发现中的未证实能力；政治和工业参与者对科学议程的控制；闭环管道中系统性错误的累积；以及全球研究社区在不可比拟的层次之间的结构性分化。这些担忧并不是对人工智能驱动科学的反对——其展现出的潜力是真实且重要的。它们构成了在负责任地追求这一潜力的条件。
cs.AI / 73 / 2607.15166
MedFailBench: A Clinician-Built Open-Source Benchmark for Medical AI Safety Boundary Inspection
MedFailBench：由临床医生构建的医疗人工智能安全边界检查开放源代码基准
Abstract
Most medical AI benchmarks measure whether a model knows the correct answer. MedFailBench asks a different question: which safety boundary failed? We present a clinician-built synthetic benchmark and failure atlas that labels medical AI errors by severity (1--5) and safety gate type (missed urgent escalation, unsafe remote dosing, unsafe discharge reassurance, evidence fabrication, unsafe protocol execution, source support gap). The current public release (v0.2.1) contains 44 clinician-reviewed synthetic cases with severity annotations, a live HuggingFace leaderboard preview, a safety gate taxonomy, a clinical severity rubric, and an automated pipeline for archiving model-response screening runs. No patient data, clinical validation claims, or model rankings are included. MedFailBench is released under Apache-2.0 and CC-BY-4.0 and carries the Zenodo DOI 10.5281/zenodo.21205535.
Chinese Translation
大多数医疗人工智能基准测试衡量模型是否知道正确答案。MedFailBench 提出了一个不同的问题：哪个安全边界失败了？我们展示了一个由临床医生构建的合成基准和失败图谱，按严重性（1-5）和安全门类型（错过紧急升级、不安全的远程给药、不安全的出院安慰、证据伪造、不安全的协议执行、源支持缺口）标记医疗人工智能错误。目前的公开版本（v0.2.1）包含44个经过临床医生审核的合成案例及其严重性注释，提供实时的 HuggingFace 排行榜预览、安全门分类法、临床严重性标准以及用于归档模型响应筛选运行的自动化管道。未包含患者数据、临床验证声明或模型排名。MedFailBench 在 Apache-2.0 和 CC-BY-4.0 许可下发布，并携带 Zenodo DOI 10.5281/zenodo.21205535。
cs.AI / 74 / 2607.15176
Benchmarking Multimodal Large Language Models for Scientific Visualization Literacy
多模态大型语言模型在科学可视化素养评估中的基准测试
Abstract
Multimodal large language models (MLLMs) are increasingly used to interpret visualizations, yet current evaluations remain largely chart-centric and provide limited evidence of understanding of scientific visualization (SciVis). We benchmark six MLLMs on the scientific visualization literacy assessment test, a standardized SciVis literacy assessment comprising 49 items based on 18 scientific visualizations and illustrations, spanning 8 techniques and 11 task types. We evaluate three closed-source and three open-source models under a closed-world protocol and compare their performance using data from 485 human participants. Results show that current MLLMs do not exhibit uniform SciVis literacy. Gemini is the strongest model overall, exceeding the human mean across the evaluated subsets, whereas the open-source models remain below the human baseline. Performance is highly uneven across techniques and tasks: models perform best on scientific illustration, search, and spatial understanding, but struggle on texture-based and integration-based visualizations and on quantitative estimation. Error analysis reveals recurring failures in fine-grained quantitative estimation, flow-direction interpretation, and grounded encoding interpretation. These findings position SciVis literacy as a necessary benchmark dimension for evaluating multimodal AI systems. Our code and model outputs are publicly available at https://github.com/patdmp/mllm-scivis-lit-benchmark.
Chinese Translation
多模态大型语言模型（MLLMs）在解释可视化方面的应用日益增加，但当前的评估主要集中于图表，提供的科学可视化（SciVis）理解证据有限。我们对六个MLLMs进行了科学可视化素养评估测试的基准测试，该测试是一个标准化的SciVis素养评估，包含49个项目，基于18个科学可视化和插图，涵盖8种技术和11种任务类型。我们在封闭世界协议下评估了三种闭源模型和三种开源模型，并利用485名参与者的数据比较它们的表现。结果显示，当前的MLLMs在SciVis素养方面并不均匀。Gemini是整体表现最强的模型，在评估的子集上超越了人类平均水平，而开源模型的表现则低于人类基线。不同技术和任务的表现差异显著：模型在科学插图、搜索和空间理解方面表现最佳，但在基于纹理和基于整合的可视化以及定量估计方面则表现不佳。错误分析揭示了在细粒度定量估计、流向解释和基于语境的编码解释方面的反复失败。这些发现将SciVis素养定位为评估多模态人工智能系统的必要基准维度。我们的代码和模型输出可在 https://github.com/patdmp/mllm-scivis-lit-benchmark 上公开获取。
cs.AI / 75 / 2607.15190
Can We Trust Item Response Theory for AI Evaluation?
我们能否信任项目反应理论用于人工智能评估？
Abstract
AI benchmarks increasingly leverage item-level statistical models, particularly item response theory (IRT), to estimate model capabilities, rank systems, select informative examples, and diagnose benchmark quality. However, AI benchmark data often departs from the data regime of human testing, for which standard IRT estimation tools were originally developed: benchmarks typically involve fewer evaluated models, far more items, and capability distributions that may be skewed, clustered, or multimodal. We examine how these regime mismatches challenge the reliability of IRT modeling for AI evaluation. Using item parameters and capability distributions derived from six widely used LLM benchmarks, we simulate response matrices under three common IRT models and compare four estimation tools used in recent benchmark studies: marginal maximum likelihood, Markov chain Monte Carlo, variational inference, and a neural pseudo-Siamese estimator. Across 18,000 simulation conditions, we systematically evaluate computational feasibility, scalability, and the reliability of IRT inferences about model rankings, predicted performance, and item characteristics. Results show that classical estimators can become infeasible in large benchmark settings, whereas scalable estimators can produce unreliable item-level and ranking inferences with small or nonnormally distributed model sets. This study identifies when latent trait models reliably support or risk distorting AI benchmarking claims, and what sample sizes and diagnostics are needed for trustworthy use.
Chinese Translation
人工智能基准测试越来越多地利用项目级统计模型，特别是项目反应理论（IRT），来估计模型能力、排名系统、选择信息性示例以及诊断基准质量。然而，人工智能基准数据往往与人类测试的数据模式存在差异，后者是标准IRT估计工具最初开发的基础：基准测试通常涉及较少的被评估模型、更多的项目，以及可能偏斜、聚集或多模态的能力分布。我们考察了这些模式不匹配如何挑战IRT建模在人工智能评估中的可靠性。通过使用来自六个广泛使用的LLM基准的项目参数和能力分布，我们在三种常见的IRT模型下模拟响应矩阵，并比较了最近基准研究中使用的四种估计工具：边际最大似然、马尔可夫链蒙特卡洛、变分推断和神经伪Siamese估计器。在18,000个模拟条件下，我们系统地评估了计算可行性、可扩展性以及IRT对模型排名、预测性能和项目特征的推断的可靠性。结果显示，在大型基准设置中，经典估计器可能变得不可行，而可扩展估计器在小型或非正态分布的模型集上可能产生不可靠的项目级和排名推断。本研究识别了潜在特征模型何时可靠地支持或有可能扭曲人工智能基准声明，以及可信使用所需的样本大小和诊断。
cs.AI / 76 / 2607.15193
Plover: Steering GUI Agents through Plan-Centric Interaction
Plover：通过以计划为中心的交互引导图形用户界面代理
Abstract
Graphical user interface (GUI) automation remains challenging in real-world environments, where dynamic layouts, unexpected dialogs, and evolving interface states can cause autonomous agents to drift from user intent. Recent vision-based multimodal agents improve flexibility by operating directly over screenshots and natural language instructions, but planning and adaptation often remain internal, limiting users' ability to inspect, supervise, or correct system behavior. We present Plover, a plan-centric vision-based GUI automation system that externalizes task plans and replanning as persistent, inspectable, and revisable artifacts. Through a planner--executor architecture, Plover supports explicit supervision of evolving execution, localized correction through editable plans, natural-language guidance, and screenshot-grounded interventions, while preserving prior progress during repair. A formative study with six participants informed the interaction design. We then evaluate Plover through benchmark failure-case repair and scenario-based workflow analyses. Our results show that many autonomous GUI-agent failures are structurally repairable when plans remain visible and interventions are localized, and that explicit replanning helps make GUI automation more transparent, controllable, and adaptable.
Chinese Translation
图形用户界面（GUI）自动化在现实环境中仍然面临挑战，动态布局、意外对话框和不断变化的界面状态可能导致自主代理偏离用户意图。最近的基于视觉的多模态代理通过直接处理屏幕截图和自然语言指令提高了灵活性，但规划和适应通常仍然是内部过程，限制了用户检查、监督或纠正系统行为的能力。我们提出了Plover，一个以计划为中心的基于视觉的GUI自动化系统，它将任务计划和重新规划外部化为持久的、可检查的和可修订的工件。通过规划者-执行者架构，Plover支持对不断演变的执行进行明确监督，通过可编辑的计划进行局部纠正，提供自然语言指导，以及基于屏幕截图的干预，同时在修复过程中保留先前的进展。与六名参与者的形成性研究为交互设计提供了信息。随后，我们通过基准失败案例修复和基于场景的工作流分析评估Plover。我们的结果表明，当计划保持可见且干预局部化时，许多自主GUI代理的失败是结构上可修复的，而明确的重新规划有助于使GUI自动化更加透明、可控和适应性强。
cs.AI / 77 / 2607.15202
Self-Evolving Human-Centered Framework for Explainable Depression Symptom Annotation
自我进化的人本中心可解释抑郁症状标注框架
Abstract
Annotation quality is a major bottleneck in building reliable and explainable artificial intelligence (XAI) systems for mental health research. In depression-related datasets, labels are often assigned without structured evidence, symptom-level justification, or traceable alignment with the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR), limiting both transparency and downstream model interpretability. We propose a self-evolving, expert-in-the-loop annotation framework for Major Depressive Disorder (MDD) that combines large language model (LLM)-assisted labeling with expert verification. The framework is intended to support the construction of explainable, DSM-5-TR-aligned datasets rather than to perform clinical diagnosis. It operates in three stages: candidate evidence selection from textual records, criterion-level DSM-5-TR analysis, and case-level synthesis that produces label-level diagnostic and severity annotations. A dual-memory architecture, composed of Example Memory and Reflection Memory, is designed to internalize expert feedback and iteratively improve future annotations without retraining. We describe this mechanism and leave its evaluation across multiple feedback cycles to future work. In addition to final labels, the framework exports clinical evidence, reasoning traces, and edit histories, enabling comprehensive auditability. In a pilot study using expert-reviewed samples, the proposed approach improves annotation consistency and explainability while reducing manual revision effort.
Chinese Translation
标注质量是构建可靠且可解释的人工智能（XAI）系统在心理健康研究中的主要瓶颈。在与抑郁症相关的数据集中，标签通常是在没有结构化证据、症状级别的理由或与《精神障碍诊断与统计手册》第五版文本修订版（DSM-5-TR）标准可追溯对齐的情况下分配的，这限制了透明度和后续模型的可解释性。我们提出了一种自我进化的专家参与标注框架，针对重度抑郁症（MDD），该框架结合了大型语言模型（LLM）辅助标注与专家验证。该框架旨在支持构建可解释的、与DSM-5-TR对齐的数据集，而不是进行临床诊断。它分为三个阶段：从文本记录中选择候选证据、进行标准级别的DSM-5-TR分析，以及生成标签级别的诊断和严重性标注的案例级合成。一个由示例记忆和反思记忆组成的双记忆架构被设计用来内化专家反馈，并在不重新训练的情况下迭代改进未来的标注。我们描述了这一机制，并将其在多个反馈周期中的评估留待未来的工作。除了最终标签外，该框架还导出临床证据、推理痕迹和编辑历史，增强了全面审计的能力。在一项使用专家审核样本的初步研究中，所提出的方法提高了标注的一致性和可解释性，同时减少了手动修订的工作量。
cs.AI / 78 / 2607.15218
When Words Are Safe But Actions Kill: Probing Physical Danger Beyond Text Safety in Hidden-State Risk Space
当语言安全但行为致命：探讨隐藏状态风险空间中的物理危险超越文本安全
Abstract
Large language models (LLMs) increasingly serve as high-level planners for embodied agents, where linguistically benign instructions can become unsafe once grounded in the physical world. We study whether this physically grounded danger is the same safety problem as ordinary text-level content danger. Through hidden-state direction analysis and random-split null tests, we show that content danger (CD) and physical danger (PD) form separable signals in LLM representations across Qwen2.5-3B/7B/14B/32B, Phi-3.5 and SmolLM2. Building on the CD/PD separability, we propose PRISM, a single-layer L2-regularized logistic probe over full hidden states. PRISM achieves 86.2--87.7\% accuracy on SafeAgentBench with 11.7--13.7\% FPR, while same-scale LLM judges over-block safe tasks at 24.7--39.0\% FPR. We further introduce PhysicalSafetyBench-1K (PSB-1K), a contrastive benchmark of 1{,}000 physical-risk pairs without direct harm keywords, to test whether methods detect physically grounded danger rather than explicit unsafe wording. On PSB-1K, PRISM reaches 99.6\% accuracy and 0.7\% FPR, whereas a Qwen2.5-3B judge rejects 67.8\% of safe tasks. PRISM also replicates on SafeText and EARBench, supporting hidden-state probing as a representation-level method for physical safety beyond text moderation.
Chinese Translation
大型语言模型（LLMs）越来越多地作为具身代理的高级规划者，其中语言上无害的指令在与物理世界结合后可能变得不安全。我们研究这种物理基础的危险是否与普通文本级内容危险是同一个安全问题。通过隐藏状态方向分析和随机拆分无效测试，我们表明内容危险（CD）和物理危险（PD）在 Qwen2.5-3B/7B/14B/32B、Phi-3.5 和 SmolLM2 的 LLM 表示中形成可分离的信号。基于 CD/PD 的可分离性，我们提出了 PRISM，这是一个在完整隐藏状态上进行单层 L2 正则化的逻辑探测器。PRISM 在 SafeAgentBench 上的准确率达到 86.2%--87.7%，假阳性率（FPR）为 11.7%--13.7%，而同规模的 LLM 在评估过于安全的任务时的 FPR 为 24.7%--39.0%。我们进一步引入了 PhysicalSafetyBench-1K（PSB-1K），这是一个包含 1,000 对物理风险的对比基准，未包含直接危害关键词，以测试方法是否能够检测到物理基础的危险，而非显式的不安全措辞。在 PSB-1K 上，PRISM 达到 99.6% 的准确率和 0.7% 的 FPR，而 Qwen2.5-3B 的评估者拒绝了 67.8% 的安全任务。PRISM 还在 SafeText 和 EARBench 上进行了复制，支持隐藏状态探测作为一种超越文本审查的物理安全表示级方法。
cs.AI / 79 / 2607.15247
AutoSynthesis: An agentic system for automated meta-analysis
AutoSynthesis：一种用于自动化元分析的代理系统
Abstract
Evidence synthesis is crucial for turning primary research into reliable knowledge for science, medicine, education, and policy. Yet, quantitative evidence synthesis remains largely manual and difficult to scale. Here, we introduce AutoSynthesis, an end-to-end multi-agent system for automated meta-analysis. Given a research question in natural language, AutoSynthesis formulates a search strategy, retrieves scientific literature, screens candidate studies, assesses full-text eligibility, extracts quantitative statistics, computes standardized effect sizes, and finally performs random-effects meta-analysis. AutoSynthesis further supports heterogeneity analysis to examine how effect sizes vary across moderators, as well as risk-of-bias assessment. As output, AutoSynthesis produces a transparent report aligned with PRISMA guidelines. In our application, AutoSynthesis screened over 28 studies and extracted more than 20 quantitative claims. The pooled effect estimates produced by AutoSynthesis are similar to Hedges' $g$ of expert-conducted meta-analyses, indicating close agreement with manual evidence synthesis. Together, these results show that AutoSynthesis can make quantitative evidence synthesis more scalable, thereby supporting evidence-based decision-making across disciplines.
Chinese Translation
证据合成对于将初级研究转化为科学、医学、教育和政策的可靠知识至关重要。然而，定量证据合成仍然主要依赖人工操作，且难以扩展。在此，我们介绍了AutoSynthesis，一种端到端的多代理系统，用于自动化元分析。给定一个自然语言的研究问题，AutoSynthesis制定搜索策略，检索科学文献，筛选候选研究，评估全文的合格性，提取定量统计数据，计算标准化效应量，最后执行随机效应元分析。AutoSynthesis还支持异质性分析，以检查效应量如何在调节变量之间变化，以及偏倚风险评估。作为输出，AutoSynthesis生成与PRISMA指南一致的透明报告。在我们的应用中，AutoSynthesis筛选了超过28项研究，并提取了20多个定量主张。AutoSynthesis生成的汇总效应估计与专家进行的元分析中的Hedges' $g$相似，表明与人工证据合成高度一致。这些结果表明，AutoSynthesis可以使定量证据合成更具可扩展性，从而支持跨学科的基于证据的决策制定。
cs.AI / 80 / 2607.15254
teLLMe Why (Ain't Nothing but a Jam): Exploratory Causal Analysis of Urban Driving Data
teLLMe 为什么（没什么好说的）：城市驾驶数据的探索性因果分析
Abstract
Traffic agencies now have access to large volumes of video-derived data for studying safety and congestion. Most of these data are observational and collected without interventions, which makes causal questions such as "How would rain change traffic density?" difficult to answer. We present teLLMe, a system for exploratory causal analysis of urban driving datasets. The system starts from a structured event table built from dashcam annotations and combines causal structure learning with the PC algorithm, bootstrap-based stability checks, and query-specific effect estimation using linear regression and DoWhy. Natural-language questions are mapped to structured causal queries through a schema-aware LLM, enabling users to specify treatments, outcomes, and subpopulations. teLLMe returns a "Causal Card" that summarizes effect estimates, adjustment sets, DAG support, and assumptions, followed by a short natural-language explanation. Case studies on BDD-derived traffic events show that the system can surface plausible relationships involving weather, peak hours, and traffic density, while making uncertainty and modeling choices explicit. The system is designed as a tool for hypothesis generation and expert reasoning rather than a source of definitive causal claims.
Chinese Translation
交通机构现在可以访问大量基于视频的数据来研究安全性和拥堵问题。这些数据大多数是观察性数据，且在没有干预的情况下收集，这使得诸如“雨天如何改变交通密度？”这样的因果问题难以回答。我们提出了 teLLMe，一个用于城市驾驶数据集的探索性因果分析系统。该系统从一个基于行车记录仪注释构建的结构化事件表开始，结合因果结构学习与 PC 算法、基于自助法的稳定性检查，以及使用线性回归和 DoWhy 的特定查询效果估计。自然语言问题通过一个具有模式感知能力的大型语言模型（LLM）映射到结构化因果查询，使用户能够指定处理、结果和子人群。teLLMe 返回一个“因果卡”，总结效果估计、调整集、DAG 支持和假设，并附有简短的自然语言解释。基于 BDD 派生的交通事件的案例研究表明，该系统能够揭示天气、峰值时段和交通密度之间的合理关系，同时明确不确定性和建模选择。该系统被设计为一个假设生成和专家推理的工具，而不是一个确定性因果主张的来源。
cs.AI / 81 / 2607.15257
SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration
SearchOS-V1：迈向稳健的开放领域信息搜索代理协作
Abstract
Recent advances in Tool-Integrated Large Language Models have made web search a core capability of information-seeking agents. However, as interaction histories grow, agents increasingly struggle to track task progress. When search attempts fail to yield useful evidence, current single- and multi-agent systems can become trapped in repetitive loops, wasting search budgets and ultimately compromising the quality and completeness of the final output. We introduce SearchOS, a system-level multi-agent framework that turns fragile, implicit search progress into explicit, persistent, and shared state. First, we formulate open-domain information seeking as relational schema completion with grounded citations, where agents discover entities, populate attributes across linked tables, and anchor each value to source evidence. Then we design Search-Oriented Context Management (SOCM), which externalizes the evolving state into Frontier Task, an Evidence Graph, a Coverage Map, and Failure Memory. Built on SOCM, SearchOS applies a pipeline-parallel scheduling mechanism that overlaps the execution of sub-agents and continuously refills freed slots with tasks targeting unresolved coverage gaps to improve utilization and throughput. To schedule and control the execution of search agents, SearchOS introduces a Search Tool Middleware Harness that intercepts model and tool interactions to record grounded evidence and react to stalls or budget exhaustion, and provides a reusable hierarchical skill system comprising strategy and access skills to augment the agents' search process and avoid repeating failed search patterns across runs. On WideSearch and GISA, SearchOS leads all metrics among the evaluated single- and multi-agent baselines, paving the way toward robust information-seeking collaboration.
Chinese Translation
近期工具集成的大型语言模型的进展使得网络搜索成为信息搜索代理的核心能力。然而，随着交互历史的增长，代理在跟踪任务进展方面越来越困难。当搜索尝试未能产生有用证据时，当前的单代理和多代理系统可能会陷入重复循环，浪费搜索预算，并最终影响最终输出的质量和完整性。我们提出了SearchOS，一个系统级的多代理框架，将脆弱的隐式搜索进展转化为明确、持久和共享的状态。首先，我们将开放领域的信息搜索形式化为带有基础引用的关系模式补全，其中代理发现实体，填充链接表中的属性，并将每个值锚定到源证据。然后，我们设计了搜索导向的上下文管理（Search-Oriented Context Management, SOCM），将不断演变的状态外部化为前沿任务（Frontier Task）、证据图（Evidence Graph）、覆盖图（Coverage Map）和失败记忆（Failure Memory）。基于SOCM，SearchOS应用了一种管道并行调度机制，重叠子代理的执行，并持续用针对未解决覆盖缺口的任务填充空闲槽，以提高利用率和吞吐量。为了调度和控制搜索代理的执行，SearchOS引入了一种搜索工具中间件（Search Tool Middleware Harness），拦截模型和工具的交互，以记录基础证据并对停滞或预算耗尽做出反应，并提供一个可重用的分层技能系统，包括策略技能和访问技能，以增强代理的搜索过程，避免在多次运行中重复失败的搜索模式。在WideSearch和GISA上，SearchOS在所有评估的单代理和多代理基线中领先于所有指标，为稳健的信息搜索协作铺平了道路。
cs.AI / 82 / 2607.15267
Pretraining Data Can Be Poisoned through Computational Propaganda
预训练数据可以通过计算宣传进行污染
Abstract
Poisoning pretraining data can introduce harmful behaviors to LMs that are difficult to detect and mitigate. Prior work on poisoning pretraining data has largely exploited established data sources such as Wikipedia, which do not represent the large scale and heterogeneity typical of pretraining corpora, and has ignored the interaction between poisoned data and data curation pipelines. We demonstrate that poisoning attacks on pretraining data are feasible beyond this limited setting through an existing web-scale content injection mechanism: public discussion interfaces. Additionally, to measure whether malicious content is included after web crawling and data curation, we introduce HalfLife, a novel analysis for estimating adversarial content inclusion in web-crawl based LM training data. We use HalfLife to explore the feasibility of poisoning pretraining corpora at web scale through open discussion interfaces. Our analysis demonstrates the importance of estimating whether poison injections are included in pretraining data, and establishes third-party webpage content as a possible vector for attacking language model pretraining.
Chinese Translation
污染预训练数据可能会引入难以检测和缓解的有害行为到语言模型（LM）中。先前关于污染预训练数据的研究主要利用了维基百科等已建立的数据源，这些数据源并未代表预训练语料库的规模和异质性，并且忽视了污染数据与数据管理流程之间的相互作用。我们展示了通过现有的网络规模内容注入机制——公共讨论接口，污染预训练数据的攻击在这一有限设置之外是可行的。此外，为了测量在网络爬取和数据管理后是否包含恶意内容，我们引入了HalfLife，这是一种用于估计对抗性内容在基于网络爬取的语言模型训练数据中包含情况的新分析方法。我们使用HalfLife探索通过开放讨论接口在网络规模上污染预训练语料库的可行性。我们的分析表明，估计污染注入是否包含在预训练数据中是重要的，并确立了第三方网页内容作为攻击语言模型预训练的一个可能途径。
cs.CL / 1 / 2607.14099
Just Keep Prompting: Evaluating Repetitive Socratic Prompting in VLMs
持续提示：评估视觉语言模型中的重复苏格拉底提示
Abstract
Deploying Vision-Language Models (VLMs) in real-world settings requires not only strong visual reasoning but also stability under sustained conversational pressure. We introduce Just Keep Prompting (JKP), a multi-turn evaluation framework that measures VLM epistemic stability when users repeatedly challenge, question, or contradict a model's answer. JKP probes models for up to 10 follow-up turns using three strategies: Adversarial Negation (repeated rejection), Pure Socratic Interrogation (repeated calls to reassess certainty), and Context-Aware Socratic Summarization (reflecting the model's prior rationale back before asking for reconsideration). We evaluate GPT-4o, Gemini 2.5 Pro, and Qwen3-VL-30B on a subset of the STAR benchmark across 720 multi-turn runs. Aggregate accuracy changes modestly from Turn 0 to Turn 10, but trajectory-level analysis reveals substantial instability: correct answers regress, wrong answers recover, and many runs exhibit repeated answer flipping. Repeated prompting has bounded upside and often acts as a destabilizer rather than a reasoning aid. The effect is strongly model-dependent: Qwen3-VL-30B achieves the highest final accuracy but becomes confidently wrong under direct contradiction; Gemini 2.5 Pro is comparatively stable but token-expensive; GPT-4o is the most brittle and oscillatory. These findings reveal that multi-turn VLM evaluation captures not just additional reasoning but pressure-response profiles: how models trade off visual grounding, calibration, and conversational compliance under repeated challenge.
Chinese Translation
在实际应用中部署视觉语言模型（VLMs）不仅需要强大的视觉推理能力，还需要在持续的对话压力下保持稳定性。我们提出了“持续提示”（Just Keep Prompting, JKP）这一多轮评估框架，用于测量VLM在用户反复挑战、质疑或反驳模型答案时的认知稳定性。JKP使用三种策略对模型进行最多10轮的后续提问：对抗性否定（重复拒绝）、纯苏格拉底质询（反复要求重新评估确定性）和上下文感知的苏格拉底总结（在请求重新考虑之前反映模型的先前推理）。我们在720个多轮运行中评估了GPT-4o、Gemini 2.5 Pro和Qwen3-VL-30B在STAR基准测试子集上的表现。整体准确率从第0轮到第10轮变化不大，但轨迹级分析揭示了显著的不稳定性：正确答案回退，错误答案恢复，许多运行表现出重复的答案翻转。重复提示的效果有限，往往作为一种不稳定因素而非推理辅助。该效应强烈依赖于模型：Qwen3-VL-30B在最终准确率上最高，但在直接反驳下变得自信地错误；Gemini 2.5 Pro相对稳定但代价高昂；GPT-4o则是最脆弱和波动性最大的。这些发现表明，多轮VLM评估不仅捕捉到额外的推理能力，还反映了模型在重复挑战下的压力响应特征：模型如何在视觉基础、校准和对话合规性之间进行权衡。
cs.CL / 2 / 2607.14100
Quantum Compositional NLP for Arabic: Grammar, Morphology, and Word Sense in Circuit Topology
阿拉伯语的量子组合自然语言处理：电路拓扑中的语法、形态学和词义
Abstract
We present the first application of pregroup grammar-based quantum compositional natural language processing (QNLP) to Arabic; a morphologically rich, free-word-order language whose structural complexity provides a uniquely demanding testbed for theories of meaning composition in quantum circuits. Our system converts Arabic sentences into quantum circuits whose topology mirrors grammatical structure: subjects, verbs, and objects become quantum gates, and the typed dependencies between them (the pregroup grammar) determine how those gates are wired together. We conduct three controlled experiments spanning word order, morphological tense, and verb sense disambiguation, comparing quantum circuit methods against classical baselines including AraVec (Arabic word embeddings) and AraBERT (a pre-trained Arabic transformer).
Chinese Translation
我们首次将基于前群语法的量子组合自然语言处理（QNLP）应用于阿拉伯语；这是一种形态丰富、自由词序的语言，其结构复杂性为量子电路中意义组合理论提供了独特而严苛的测试平台。我们的系统将阿拉伯语句子转换为量子电路，其拓扑结构反映了语法结构：主语、动词和宾语变成量子门，而它们之间的类型依赖关系（前群语法）决定了这些门的连接方式。我们进行了三项受控实验，涵盖了词序、形态时态和动词义项消歧，比较了量子电路方法与经典基线，包括AraVec（阿拉伯语词嵌入）和AraBERT（预训练的阿拉伯语变换器）。
cs.CL / 3 / 2607.14101
LBA: Textual Hard-Label Adversarial Attack under Low Query Budgets
LBA：低查询预算下的文本硬标签对抗攻击
Abstract
Generating high-quality adversarial texts with low query budgets remains a challenging problem in the hard-label scenario. Most existing approaches rely on greedy algorithms, where one position in the text is selected for substitution, followed by the substitutions of other positions. This local search approach may fail to discover high-quality adversarial examples and often leads to excessive query costs. Ideally, an optimal adversarial sample would consider all possible position combinations in the text, but exhaustive search is computationally impractical. To address this challenge, we propose a sampling-based method called LBA, which constructs an approximate distribution of high-quality adversarial examples by integrating both prior and posterior knowledge, and utilizes this distribution for sampling. As sampling progresses, posterior knowledge updates the approximate distribution, which in turn guides more effective sampling. Extensive experiments on six language models, ranging from small-scale to large-scale architectures across four datasets, demonstrate that LBA significantly outperforms state-of-the-art baselines on all evaluation metrics. Additionally, LLM-based assessment indicates that LBA generates more semantically preserved and comprehensible adversarial texts.
Chinese Translation
在硬标签场景中，以低查询预算生成高质量对抗文本仍然是一个具有挑战性的问题。现有的大多数方法依赖于贪婪算法，其中选择文本中的一个位置进行替换，然后替换其他位置。这种局部搜索方法可能无法发现高质量的对抗样本，并且通常会导致过高的查询成本。理想情况下，最佳对抗样本应考虑文本中所有可能的位置组合，但穷举搜索在计算上是不切实际的。为了解决这一挑战，我们提出了一种基于采样的方法，称为LBA，该方法通过整合先验和后验知识构建高质量对抗样本的近似分布，并利用该分布进行采样。随着采样的进行，后验知识更新近似分布，从而指导更有效的采样。在六种语言模型上的大量实验，涵盖了四个数据集，从小规模到大规模架构，证明LBA在所有评估指标上显著优于最先进的基线。此外，基于LLM的评估表明，LBA生成的对抗文本在语义保留和可理解性方面表现更佳。
cs.CL / 4 / 2607.14102
UniSAGE: Unifying Static and Dynamic Attributes with Hyper-Structure
UniSAGE：通过超结构统一静态和动态属性
Abstract
With the rapid growth of digital data, real-world applications increasingly involve hierarchical information that combines static attributes with dynamic records. Modeling such heterogeneous data in a unified and generalizable manner remains challenging. Existing approaches often rely on extensive manual design, are tightly coupled to specific data schemas, and typically process static and dynamic attributes in isolation, thereby overlooking their implicit interactions. We propose UniSAGE, a unified framework for modeling data with both static and dynamic attributes. UniSAGE constructs a global attribute graph that represents hierarchical and temporal relationships in a unified structure. To ensure representational consistency, it introduces two orthogonal parameter subspaces that jointly support static aggregation and dynamic reasoning within a shared semantic space. Building on these unified representations, UniSAGE further enables task-specific interaction between static and dynamic attributes via a lightweight hyper-structure mechanism. UniSAGE is fully automated, robust to evolving data schemas, and capable of capturing complex cross-attribute dependencies. Extensive experiments on multiple public benchmarks and a real-world financial behavior dataset demonstrate that UniSAGE consistently outperforms existing methods, achieving performance improvements of over 10% on several tasks.
Chinese Translation
随着数字数据的快速增长，现实世界的应用越来越多地涉及将静态属性与动态记录相结合的层次信息。以统一和可推广的方式建模这种异构数据仍然具有挑战性。现有的方法通常依赖于广泛的手动设计，与特定数据模式紧密耦合，并且通常孤立地处理静态和动态属性，从而忽视了它们之间的隐含交互。我们提出了UniSAGE，一个用于建模具有静态和动态属性的数据的统一框架。UniSAGE构建了一个全局属性图，以统一结构表示层次和时间关系。为了确保表征一致性，它引入了两个正交的参数子空间，共同支持在共享语义空间内的静态聚合和动态推理。在这些统一表征的基础上，UniSAGE进一步通过轻量级超结构机制实现静态和动态属性之间的任务特定交互。UniSAGE是完全自动化的，能够适应不断变化的数据模式，并能够捕捉复杂的跨属性依赖关系。在多个公共基准和一个真实世界的金融行为数据集上的大量实验表明，UniSAGE始终优于现有方法，在多个任务上实现了超过10%的性能提升。
cs.CL / 5 / 2607.14103
Latent Communication Between Language Model Agents: Channels, Alignment, and the Limits of Text
语言模型代理之间的潜在通信：通道、对齐及文本的局限性
Abstract
Multi-agent systems (MAS) are utilized in many contexts and many professions. Those MAS rely on inter-agent communication, usually implemented by clear-text message passing. We hypothesize that Large Language Models may have a world model at their disposal that exceeds expressibility in text when complex concepts need to be communicated. Our aim is to approach a proof of this hypothesis with structured experiments. In this work, we show that LLM agents communicating via text lose information, which we quantify via Sparse Autoencoder (SAE) feature analysis. We construct three communication channels and measure concept-discriminating information in each. We first show that the SAE-sparse channel retains a 99.4% probe accuracy at 28-fold compression over the dense-latent channel vs 80.4% for the text channel. We then proceed to examine the same for cross-architecture communication by using sparse latent space alignment. We find for Procrustes alignment a 92% top-1 retrieval between Llama and Mistral. Using a text round-trip, we perform feature survival analysis to find that text serialization destroys 88% of SAE features, replacing them with a different feature set. We attribute the loss to identity replacement, not attenuation. By our analysis, we were able to attribute a 3-10pp performance penalty to the linear Procrustes alignment, improving with nonlinear alignment methods. In a task-level evaluation we find that the latent channel matches the text channel on cross-lingual concept tasks but never exceeds it. Text augmentation with latent features provides no benefit, leading us to negative conclusions for the initial hypothesis: lost features mostly or completely encode surface form, not task-relevant semantics. To pinpoint the practical advantage of latent communication over a text channel, deeper tasks eliciting complex concepts and an corresponding analysis framework are needed.
Chinese Translation
多代理系统（MAS）在许多背景和职业中得到了广泛应用。这些MAS依赖于代理之间的通信，通常通过明文消息传递来实现。我们假设大型语言模型（Large Language Models）可能拥有一种超越文本表达能力的世界模型，尤其是在需要传达复杂概念时。我们的目标是通过结构化实验来验证这一假设。在本研究中，我们展示了通过文本进行通信的LLM代理会丢失信息，我们通过稀疏自编码器（Sparse Autoencoder，SAE）特征分析对此进行了量化。我们构建了三个通信通道，并测量每个通道中的概念区分信息。我们首先展示了SAE稀疏通道在28倍压缩下保持99.4%的探测准确率，而密集潜在通道为80.4%，文本通道则更低。接着，我们使用稀疏潜在空间对齐来检查跨架构通信的相同情况。我们发现，在Procrustes对齐中，Llama与Mistral之间的顶级检索率为92%。通过文本往返，我们进行特征存活分析，发现文本序列化会破坏88%的SAE特征，并用不同的特征集替代。我们将这种损失归因于身份替换，而非衰减。通过我们的分析，我们能够将线性Procrustes对齐的性能损失归因于3-10个百分点，并且随着非线性对齐方法的改进而改善。在任务级评估中，我们发现潜在通道在跨语言概念任务中与文本通道相匹配，但从未超过它。使用潜在特征进行文本增强没有带来任何好处，这使我们对初始假设得出消极结论：丢失的特征主要或完全编码表面形式，而非任务相关的语义。为了明确潜在通信相对于文本通道的实际优势，需要更深层次的任务以引发复杂概念，并相应地建立分析框架。
cs.CL / 6 / 2607.14104
UzWordnet and Generative AI for Learning Uzbek by Game Playing
通过游戏学习乌兹别克语的UzWordnet与生成性人工智能
Abstract
This paper presents an educational system architecture that enables learners to practice the Uzbek language through game-playing. The architecture integrates UzWordnet and the largest currently available orthographic dictionary for Uzbek as core lexical resources, together with generative AI as a fundamental component for learning support. We design four educational games to facilitate Uzbek language learning and propose a game-based methodology for improving UzWordnet as a direct by-product of game dynamics. Our approach combines game design and lexical resources to address objectives that are at the same time educational (language learning) and lexical (improvement and enrichment of a lexical resource).
Chinese Translation
本文提出了一种教育系统架构，使学习者能够通过游戏进行乌兹别克语的练习。该架构将UzWordnet和目前可用的最大乌兹别克语正字法词典作为核心词汇资源，并将生成性人工智能作为学习支持的基本组成部分。我们设计了四款教育游戏，以促进乌兹别克语学习，并提出了一种基于游戏的方法论，以游戏动态的直接副产品来改善UzWordnet。我们的方法结合了游戏设计和词汇资源，以同时实现教育（语言学习）和词汇（词汇资源的改善与丰富）的目标。
cs.CL / 7 / 2607.14105
Automatically Evolving Prompt Guidelines for Task-Specific Optimization
自动演化的任务特定优化提示指南
Abstract
For Large Language Models to reliably answer user queries, users must clearly specify requirements, context, and constraints. In practice, however, user queries are often underspecified, forcing models to infer unstated assumptions that may misalign with the actual user intent. Existing prompt engineering guidelines aim to mitigate this issue, they are typically generic and task-agnostic, limiting their practical utility. Additionally, existing guidelines are formed manually and in a non-systematic way. To this end, we study prompt guideline optimization: the problem of automatically generating task-specific guidelines that help write better-specified prompts for a given task and model. Our key observation is that existing (completed) task examples (aka reference answers) often implicitly encode the missing information required to complete underspecified queries, including behavioral constraints, contextual assumptions, and evaluation criteria. We therefore propose AGOPS, an automatic approach that evolves task-specific guidelines via an optimization scheme that involves a prompt LLM writer, a solver LLM and prompt evolution, which maximize downstream effectiveness on a set of examples (user queries with reference answers). At inference time, our guidelines help users write well-specified prompts, boosting the effectiveness of LLMs. We show across mathematical reasoning, medical question answering, and coding tasks, that prompt underspecification leads to major drops (up to 95.3%) in downstream task performance (compared to well-specified prompts) and, perhaps more importantly, that this drop can hardly be recovered by existing prompt optimization techniques. Users following AGOPS guidelines can regain this loss (increasing performance between 15.5 to 81.7% on average) consistently across all benchmarks.
Chinese Translation
为了使大型语言模型能够可靠地回答用户查询，用户必须清晰地指定需求、上下文和约束条件。然而，在实际操作中，用户查询往往不够明确，迫使模型推断未说明的假设，这可能与实际用户意图不一致。现有的提示工程指南旨在缓解这一问题，但通常是通用的且与任务无关，限制了其实际效用。此外，现有指南是手动形成的，且方式不系统。为此，我们研究了提示指南优化：自动生成任务特定指南的问题，以帮助为给定任务和模型编写更明确的提示。我们的关键观察是，现有的（已完成的）任务示例（即参考答案）通常隐含地编码了完成不明确查询所需的缺失信息，包括行为约束、上下文假设和评估标准。因此，我们提出了AGOPS，这是一种通过优化方案演化任务特定指南的自动化方法，该方案涉及提示LLM编写器、求解器LLM和提示演化，旨在最大化一组示例（用户查询与参考答案）上的下游有效性。在推理时，我们的指南帮助用户编写明确的提示，从而提升LLM的有效性。我们展示了在数学推理、医学问答和编码任务中，提示的不明确性导致下游任务性能显著下降（与明确提示相比，下降幅度高达95.3%），而且更重要的是，这一下降几乎无法通过现有的提示优化技术恢复。遵循AGOPS指南的用户可以持续恢复这一损失（在所有基准测试中，性能平均提高15.5%至81.7%）。
cs.CL / 8 / 2607.14106
Token Time Continuous Diffusion for Language Modeling
用于语言建模的令牌时间连续扩散
Abstract
In this paper we introduce token time continuous diffusion (TTCD), a new diffusion language model which (a) operates in continuous space, deterministically mapping Gaussian noise to a final token canvas with no further sampling, and crucially (b) incorporates a new notion of per-token times, with some tokens proceeding from noise to token at a faster rate than others. Continuous space modeling helps TTCD avoid the parallel sampling of multiple tokens, which is a key source of inaccuracy at high speedups for models that iterate purely in discrete space. The notion of per-token times helps TTCD to better model conditional generation, allows for more sure tokens to proceed at a faster rate, and allows for differentiated inter-token influences during refinement. TTCD outperforms discrete models at high speedups. We train a 160M parameter TTCD model on OpenWebText, and then self-distill it; we find that at high speedups we are comparable in unconditional generation quality, and outperform in conditional generation, several existing models of similar size trained, on the same data, and self-distilled. We achieve similar gains in Sudoku solving as well.
Chinese Translation
在本文中，我们介绍了一种新的扩散语言模型——令牌时间连续扩散（Token Time Continuous Diffusion, TTCD），该模型具有以下特点：(a) 在连续空间中操作，确定性地将高斯噪声映射到最终的令牌画布上，无需进一步采样；(b) 引入了每个令牌时间的新概念，某些令牌从噪声到令牌的转换速度快于其他令牌。连续空间建模帮助TTCD避免了多个令牌的并行采样，这在高加速下是纯离散空间模型的一个主要不准确来源。每个令牌时间的概念帮助TTCD更好地建模条件生成，允许更多的确定性令牌以更快的速度进行转换，并在细化过程中允许不同的令牌之间产生差异化的影响。TTCD在高加速下优于离散模型。我们在OpenWebText上训练了一个160M参数的TTCD模型，然后进行自蒸馏；我们发现，在高加速下，我们在无条件生成质量上与几个相似大小的现有模型相当，这些模型在相同数据上训练并自蒸馏，而在条件生成上则表现更优。我们在数独求解中也取得了类似的提升。
cs.CL / 9 / 2607.14107
Polestar: Drift-Aware Cache Calibration and Token Commitment for Efficient Inference of Diffusion LLMs
Polestar：漂移感知缓存校准与令牌承诺以提高扩散大语言模型的推理效率
Abstract
The inference efficiency of diffusion large language models (dLLMs) is constrained by two challenges: bidirectional attention precludes efficient KV-cache reuse, while increasing decoding parallelism with static confidence thresholds can compromise generation quality. We observe that both challenges arise from a shared phenomenon: as tokens are decoded, their contextual integration through bidirectional attention causes token representations to drift (evolve) across decoding steps. This insight motivates Polestar, a training-free inference framework that uses token representation drift as a unified signal to jointly address both challenges. Polestar comprises two components: Polestar-Cache, which identifies stale KV-cache positions via drift and performs sparse KV-cache refreshes to enable efficient reuse, and Polestar-Commit, which detects sharp drift events to reliably identify commit-ready tokens. Across mathematics and coding benchmarks on several dLLM families, Polestar sets a new state of the art on the accuracy-throughput Pareto frontier, achieving up to 10.73% accuracy improvement, up to 3.7x higher throughput, and high decoding parallelism of 3.67 tokens per forward pass over existing baselines.
Chinese Translation
扩散大语言模型（dLLMs）的推理效率受到两个挑战的限制：双向注意力阻碍了高效的键值缓存（KV-cache）重用，而通过静态置信度阈值增加解码并行性可能会影响生成质量。我们观察到这两个挑战源于一个共同现象：随着令牌的解码，它们通过双向注意力的上下文整合导致令牌表示在解码步骤中发生漂移（演变）。这一洞察促使我们提出Polestar，一个无训练的推理框架，利用令牌表示漂移作为统一信号来共同解决这两个挑战。Polestar包括两个组件：Polestar-Cache，通过漂移识别过时的KV-cache位置，并执行稀疏的KV-cache刷新以实现高效重用；Polestar-Commit，检测显著漂移事件以可靠地识别准备提交的令牌。在多个dLLM系列的数学和编码基准测试中，Polestar在准确性-吞吐量的Pareto前沿上设定了新的最优状态，实现了高达10.73%的准确性提升，吞吐量提高至3.7倍，以及每次前向传递3.67个令牌的高解码并行性，超越了现有基准。
cs.CL / 10 / 2607.14108
Eta Given Delta: Defining LLM Tool Efficiency With Marginal Tool Utility
给定Delta的Eta：通过边际工具效用定义大型语言模型工具效率
Abstract
This paper introduces tool efficiency, a new quantitative metric to evaluate the rate of useful tool calls in an LLM agent trajectory. To ensure that tool efficiency is well-defined, we also introduce marginal tool utility, a new quantitative metric defined per tool call indicating whether a tool is useful or whether it can be safely removed from the tool suite without affecting accuracy while increasing tool efficiency; in this paper, we determine the sign of marginal tool utility for each tool call in a trajectory using LLM-as-a-Judge. While much prior work has been done to develop techniques that improve tool use by LLMs and design evaluation methods measuring efficiency indirectly using accuracy as a proxy, our work is centered on measuring efficiency directly via the quantitative metric proposed in this paper in post hoc trajectory analyses. It is our intention that this work contributes to the frontier of LLM evaluation research as a springboard for future benchmark designs and agent harness engineering (specifically with regards to creating lean tool suites) that optimize for metrics that complement but are distinct from accuracy.
Chinese Translation
本文引入了工具效率（tool efficiency），一种用于评估大型语言模型（LLM）代理轨迹中有用工具调用率的新量化指标。为了确保工具效率的定义明确，我们还引入了边际工具效用（marginal tool utility），这是一种针对每次工具调用定义的新量化指标，用以判断该工具是否有用，或者是否可以在不影响准确性的前提下安全地从工具套件中移除，从而提升工具效率；本文中，我们利用“LLM作为裁判”（LLM-as-a-Judge）的方法确定轨迹中每次工具调用的边际工具效用符号。尽管此前已有大量工作致力于开发提升LLM工具使用的技术，并设计以准确率为代理间接衡量效率的评估方法，我们的工作则聚焦于通过本文提出的量化指标，在事后轨迹分析中直接测量效率。我们期望本研究能够推动LLM评估研究的前沿，成为未来基准设计和代理工具套件工程（特别是关于构建精简工具套件）的跳板，从而优化那些与准确率互补但又有区别的指标。
cs.CL / 11 / 2607.14109
Simplicity Paradox: Debunking myths about prompting and datasets for LLM evaluation
简单性悖论：揭穿关于大型语言模型评估的提示和数据集的神话
Abstract
Probing the capabilities of Large Language Models (LLMs) and building robust solutions for Multiple-Choice Question Answering (MCQA) remain central challenges in natural language understanding. Furthermore, the rapid proliferation of LLMs has created the implicit assumption that more sophisticated prompting techniques yield better performance. Several studies claim better performance with more sophisticated prompting techniques, but do not provide a comprehensive evaluation. We address this gap through a comprehensive empirical study of 8 prompting techniques across 10 multiple-choice question answering (MCQA) datasets, encompassing 27 model configurations and roughly 4,300 unique questions evaluated more than 430,000 times. Our findings reveal a striking paradox that baseline prompting consistently outperforms complex reasoning techniques on various benchmarks. Only minimal expert and inductive role framing (CoT-Expert and CoT-Inductive) yields a small but statistically significant $\sim$3 percentage-point (pp) gain over baseline whereas every other elaborate technique we tested matches or under-performs it, often by large margins (up to 31~pp for Self-Analogical). We further investigate three critical phenomena: (1) the unexpected victory of Qwen3-30B-A3B-Thinking-2507 in Elo ratings, (2) the performance-efficiency trade-offs across model variants with different thinking budgets, revealing model-dependent optimal configurations, and (3) the substantial variation in dataset difficulty, with 60% of benchmarks below 70% accuracy and a 47.5~pp spread from easiest to hardest, indicating considerable room for model improvement. These results suggest that the LLM evaluation community may be overcomplicating prompt engineering and that substantial performance gaps remain across diverse benchmarks, offering opportunities for genuine model improvements rather than prompt optimization.
Chinese Translation
探讨大型语言模型（LLMs）的能力以及为多项选择题回答（MCQA）构建稳健解决方案仍然是自然语言理解中的核心挑战。此外，LLMs的快速普及使得人们隐含地认为更复杂的提示技术能够带来更好的性能。一些研究声称更复杂的提示技术表现更好，但并未提供全面的评估。我们通过对10个多项选择题回答（MCQA）数据集中的8种提示技术进行全面的实证研究来填补这一空白，涵盖27种模型配置和大约4,300个独特问题，评估次数超过430,000次。我们的研究结果揭示了一个显著的悖论：基线提示在各种基准测试中始终优于复杂推理技术。只有最小的专家和归纳角色框架（CoT-Expert和CoT-Inductive）在基线之上获得了约3个百分点（pp）的小幅但统计显著的提升，而我们测试的其他所有复杂技术都与基线相匹配或表现不佳，通常差距很大（自类比法最多可达31 pp）。我们进一步探讨了三个关键现象：（1）Qwen3-30B-A3B-Thinking-2507在Elo评分中意外获胜，（2）不同思维预算的模型变体之间的性能效率权衡，揭示了模型依赖的最佳配置，以及（3）数据集难度的显著变化，60%的基准测试准确率低于70%，从最简单到最困难的准确率差异达到47.5 pp，表明模型改进的空间相当大。这些结果表明，LLM评估社区可能在提示工程上过于复杂化，并且在不同基准测试中仍然存在显著的性能差距，为真正的模型改进而非提示优化提供了机会。
cs.CL / 12 / 2607.14110
MAPS: Modeling Co-Existing Subjective Perspectives and Shared Meaning in Multi-Agent Cognitive Dialogue
MAPS：在多智能体认知对话中建模共存的主观视角和共享意义
Abstract
Human dialogue involves more than exchanging information; it also expresses beliefs, emotions, and subjective cognitive styles. Yet current AI dialogue systems often enforce semantic uniformity, sacrificing diversity and interpretability. We present MAPS (Multi-Agent Perspective Spaces), a novel framework that models dialogue between cognitively distinct agents through domain-weighted profiles, dynamic GRU-based memory, and interpretable token-level attention. MAPS enables agents to maintain individualized reasoning while progressively converging on shared meaning. Evaluations on EmpatheticDialogues, TopicalChat, and MultiWOZ show that MAPS supports semantic alignment without collapsing subjectivity. Our results demonstrate a path toward cognitively grounded, interpretable dialogue systems that balance expressiveness and coherence.
Chinese Translation
人类对话不仅仅是信息的交换；它还表达了信念、情感和主观认知风格。然而，目前的人工智能对话系统往往强制执行语义统一，牺牲了多样性和可解释性。我们提出了MAPS（多智能体视角空间），这是一个新颖的框架，通过领域加权的特征、动态的基于GRU的记忆和可解释的令牌级注意力，建模认知上不同的智能体之间的对话。MAPS使智能体能够保持个性化的推理，同时逐步趋向于共享意义。在EmpatheticDialogues、TopicalChat和MultiWOZ上的评估表明，MAPS支持语义对齐而不压制主观性。我们的结果展示了一条通向认知基础、可解释的对话系统的路径，平衡了表现力和一致性。
cs.CL / 13 / 2607.14111
Introspection Fine-Tuning (IFT): Training Small LLMs to Introspect
内省微调 (IFT)：训练小型语言模型进行内省
Abstract
Can small language models detect and report on perturbations their own internal activations? We investigate this question through the lens of activation steering: injecting concept vectors into a model's residual stream and measuring whether the model can accurately report on the perturbation. We first show that the binary detection paradigm used in prior work -- prompting the model to answer Yes'' or No'' to whether it detects an injected thought -- is confounded in small models, as steering biases the model toward affirmative responses regardless of the question content. We therefore propose two confound-free evaluation paradigms: sentence localization (identifying which of $N$ sentences was perturbed, chance $= 1/N$) and strength comparison (identifying which of two sentences received a stronger injection, chance $= 50\%$). Evaluating across six models from two families (Llama-3.2 and Gemma-4), we find that models as small as 2B parameters introspect reliably well above chance, and that introspective ability generally increases with scale. Llama-1B, however, performs at or below chance. We then introduce \emph{Introspection Fine-Tuning} (IFT): supervised fine-tuning on sentence-localization examples constructed from the model's own perturbed forward passes. IFT raises Llama-1B sentence-localization accuracy from $9.6\%$ to $60.6\%$ (a $6\times$ improvement), with gains generalizing zero-shot to the held-out strength-comparison task ($30.2\% \to 52.2\%$). IFT also improves introspection for 3B and 8B models, while inducing negligible degradation on standard capability benchmarks. Our results suggest that introspective ability is not fixed by scale alone: it can be directly trained, and doing so unlocks latent self-monitoring capacity with implications for AI transparency and alignment. Our code is \href{https://anonymous.4open.science/r/IFT-introspection-2092/README.md}{here}.
Chinese Translation
小型语言模型能否检测并报告自身内部激活的扰动？我们通过激活引导的视角来探讨这个问题：将概念向量注入模型的残差流中，并测量模型是否能够准确报告扰动。我们首先展示了以往研究中使用的二元检测范式——促使模型回答“是”或“否”以判断其是否检测到注入的思想——在小型模型中受到干扰，因为引导会使模型偏向于肯定的回答，无论问题内容如何。因此，我们提出了两种不受干扰的评估范式：句子定位（识别 $N$ 个句子中哪个被扰动，随机概率 $= 1/N$）和强度比较（识别两个句子中哪个接受了更强的注入，随机概率 $= 50 ext{%}$）。在对来自两个系列（Llama-3.2 和 Gemma-4）的六个模型进行评估时，我们发现参数量仅为 2B 的模型在内省能力上可靠地超出随机水平，并且内省能力通常随着规模的增加而提高。然而，Llama-1B 的表现处于或低于随机水平。随后，我们引入了 extit{内省微调}（IFT）：在由模型自身的扰动前向传递构建的句子定位示例上进行监督微调。IFT 将 Llama-1B 的句子定位准确率从 $9.6 ext{%}$ 提高到 $60.6 ext{%}$（提升了 $6 imes$），并且在零样本情况下对保留的强度比较任务的提升也得到了推广（$30.2 ext{%} o 52.2 ext{%}$）。IFT 还改善了 3B 和 8B 模型的内省能力，同时对标准能力基准的影响微乎其微。我们的结果表明，内省能力并非仅由规模决定：它可以直接训练，而这样做可以释放潜在的自我监控能力，对人工智能的透明性和对齐具有重要意义。我们的代码可以在 extit{这里}找到。
cs.CL / 14 / 2607.14112
Information-Theoretic Limits of Reliability and Scaling in Language Models
语言模型的可靠性和扩展性的信源理论极限
Abstract
Large language models (LLMs) are evaluated as though perfect reliability is achievable for any task given sufficient scale. We show this assumption is information-theoretically unjustified. Every generative task has a reliability ceiling that no model can exceed, determined by how much output uncertainty is resolvable from observable context. The gap decomposes into a resolvable component closable with additional context and a subjective component inherent to task ambiguity. Autoregressive generation further degrades this ceiling at a rate governed by the task's dependency kernel, which quantifies inter-token correlations in the output. From these two primitives, we derive a first-principles scaling law where LLM performance is bottlenecked by the scarcer resource: training data or model capacity. This law recovers the Chinchilla scaling law as a special case and provides a structural account of when scaling improves reliability. Beyond scaling, our framework unifies diverse practical phenomena, such as the benefits of retrieval-augmentation and the spectral mechanics of catastrophic forgetting. Our work formalizes the resource-complexity tradeoffs that govern model performance across domains, offering a unified theory of performance limits in generative language models.
Chinese Translation
大型语言模型（LLMs）的评估假设在足够规模下可以实现任何任务的完美可靠性。我们展示了这一假设在信息理论上是不成立的。每个生成任务都有一个可靠性上限，任何模型都无法超越，该上限由可观察上下文中可解决的输出不确定性决定。这个差距可以分解为一个可通过额外上下文弥补的可解决成分和一个固有于任务模糊性的主观成分。自回归生成进一步以任务的依赖核为主导的速率降低了这一上限，该依赖核量化了输出中各标记之间的相关性。基于这两个原理，我们推导出一个从基本原理出发的扩展法则，其中LLM性能受到稀缺资源的瓶颈：训练数据或模型容量。该法则将Chinchilla扩展法则作为特例恢复，并提供了一个结构性解释，说明何时扩展能提高可靠性。超越扩展，我们的框架统一了多种实际现象，例如检索增强的好处和灾难性遗忘的谱力学。我们的研究形式化了影响模型性能的资源-复杂性权衡，为生成语言模型的性能极限提供了统一理论。
cs.CL / 15 / 2607.14113
T5-CSBoost: Adversarial Perturbation Resistant LLM Fingerprinting
T5-CSBoost：对抗扰动鲁棒的LLM指纹识别
Abstract
While many AI-generated text (AIGT) detectors achieve strong performance on clean inputs, their accuracy degrades significantly under light paraphrasing, word substitutions, character edits, and distribution shifts. We present T5 Contrastive Style Boosted Classifier (T5-CSBoost), an extension to the T5-Sentinel framework that keeps the original next-token prediction objective for source attribution while introducing an auxiliary margin-based triplet loss over decoder embeddings. This contrastive style regularization encourages the learning of compact, perturbation-resistant stylistic representations, offering a lightweight yet effective alternative to prior approaches that rely on architectural modifications, adversarial training, or complex multi-task objectives without altering the underlying T5-small backbone. T5-CSBoost achieves state-of-the-art multiclass source attribution and binary human-vs-LLM detection on OpenLLMText and HC3 AIGT benchmarks. More importantly, T5-CSBoost demonstrates enhanced robustness to word and character level adversarial perturbations of up to 90% intensity, achieving state-of-the-art on the challenging MAGE/Deepfake stress-test suite, including unseen models, unseen domains, and extreme paraphrasing scenarios. Our results highlight that explicitly regularizing stylistic embeddings via contrastive learning is a practical and effective strategy for building more robust LLM fingerprinting systems in real-world adversarial settings.
Chinese Translation
虽然许多人工智能生成文本（AIGT）检测器在干净输入上表现出色，但在轻微的改写、词语替换、字符编辑和分布变化下，它们的准确性显著下降。我们提出了T5对比风格增强分类器（T5-CSBoost），这是对T5-Sentinel框架的扩展，保持了源归属的原始下一个标记预测目标，同时在解码器嵌入上引入了辅助的基于边际的三元组损失。这种对比风格的正则化鼓励学习紧凑的、对扰动鲁棒的风格表示，提供了一种轻量且有效的替代方案，优于依赖于架构修改、对抗训练或复杂多任务目标的先前方法，而不改变基础的T5-small骨干网络。T5-CSBoost在OpenLLMText和HC3 AIGT基准测试中实现了最先进的多类源归属和二元人类与LLM检测。更重要的是，T5-CSBoost在高达90%强度的词和字符级对抗扰动下表现出增强的鲁棒性，在具有挑战性的MAGE/Deepfake压力测试套件中实现了最先进的表现，包括未见过的模型、未见过的领域和极端改写场景。我们的结果强调，通过对比学习显式正则化风格嵌入是一种在现实世界对抗环境中构建更鲁棒的LLM指纹识别系统的实用有效策略。
cs.CL / 16 / 2607.14114
CoEvoT: Co-Evolving Chain-of-Thought Prompting for Graph-LLM Reasoning
CoEvoT：用于图形-大语言模型推理的共同演化思维链提示
Abstract
Graph learning under distribution shift presents a persistent challenge, where models adapt to new graphs with limited or even no supervision. Recent graph--LLM approaches move toward label-efficient prediction by linearizing graphs into prompts and using large language models (LLMs) as predictors, and can adopt Chain-of-Thought (CoT) prompting to exploit LLM's multi-step reasoning capability. However, existing CoT-based graph--LLM methods generate intermediate thoughts while conditioning on fixed graph tokens, limiting step-wise refinement of structural cues. In this paper, we propose CoEvoT, a simple yet effective co-evolving CoT prompting framework for graph--LLM reasoning. CoEvoT couples text-to-graph token rewriting and graph-to-text reasoning guidance in a closed loop: each intermediate textual thought is used to update the graph token evidence state via a lightweight condition network, and the updated tokens are fed back into the next-step instruction to guide subsequent LLM reasoning. This enables step-wise, state-aware evidence refinement, rather than reasoning over a fixed graph snapshot. Extensive experiments on eight datasets demonstrate that CoEvoT consistently outperforms state-of-the-art baselines.
Chinese Translation
在分布转移下的图学习面临着持续的挑战，模型在有限甚至没有监督的情况下适应新的图。最近的图形-大语言模型（graph-LLM）方法通过将图线性化为提示并使用大型语言模型（LLMs）作为预测器，朝着标签高效预测的方向发展，并可以采用思维链（Chain-of-Thought, CoT）提示来利用LLM的多步骤推理能力。然而，现有的基于CoT的图形-LLM方法在固定图标记的条件下生成中间思维，限制了结构线索的逐步细化。本文提出了CoEvoT，一个简单而有效的共同演化CoT提示框架，用于图形-LLM推理。CoEvoT在一个闭环中结合了文本到图标记的重写和图到文本的推理指导：每个中间文本思维用于通过轻量级条件网络更新图标记证据状态，更新后的标记反馈到下一步指令中，以指导后续的LLM推理。这使得逐步、状态感知的证据细化成为可能，而不是在固定的图快照上进行推理。在八个数据集上的大量实验表明，CoEvoT始终优于最先进的基线。
cs.CL / 17 / 2607.14116
ReportMedSAM: Guiding Segmentation Through Radiology Reports
ReportMedSAM：通过放射学报告指导分割
Abstract
Free-form radiology reports contain rich clinical descriptions, yet converting them for reliable segmentation remains challenging due to the inherent variability of natural language. Existing pipelines often rely on predefined organ phrases or brittle rule-based inference-time extraction, which limits their scalability to novel anatomical structures and makes them sensitive to linguistic variations. To address this, we propose ReportMedSAM, a report-driven framework that replaces discrete extraction with a learnable concept bank. By leveraging a frozen medical vision-language encoder (BiomedCLIP), we align organ-level concept embeddings with large-scale clinical corpora through contrastive learning, establishing mutually orthogonal semantic anchors. Our approach explicitly mitigates organ-level semantic collapse and ensures high robustness against diverse clinical synonyms (e.g., "renal" vs. "kidney" ). During inference, a clinical report is embedded and matched against this concept bank to dynamically activate task-specific Mixture-of-Experts (MoE) modules. This decoupled design allows new concepts and experts to be added without retraining existing components, providing a parameter-isolated extension mechanism while keeping previously learned experts unchanged. Evaluated on the AbdomenAtlas 3.0 dataset, ReportMedSAM effectively interprets free-form reports, achieves competitive segmentation accuracy, and demonstrates seamless, non-interfering extension to novel clinical tasks.
Chinese Translation
自由格式的放射学报告包含丰富的临床描述，但由于自然语言固有的变异性，将其转换为可靠的分割仍然具有挑战性。现有的处理流程通常依赖于预定义的器官短语或脆弱的基于规则的推理时提取，这限制了它们对新解剖结构的可扩展性，并使其对语言变异敏感。为了解决这个问题，我们提出了ReportMedSAM，一个以报告为驱动的框架，用可学习的概念库替代离散提取。通过利用一个冻结的医学视觉-语言编码器（BiomedCLIP），我们通过对比学习将器官级概念嵌入与大规模临床语料库对齐，建立相互正交的语义锚点。我们的方法明确减轻了器官级语义崩溃，并确保对多样化临床同义词（例如，“肾”（renal）与“肾脏”（kidney））具有高鲁棒性。在推理过程中，临床报告被嵌入并与该概念库匹配，以动态激活特定任务的专家混合（Mixture-of-Experts, MoE）模块。这种解耦设计允许在不重新训练现有组件的情况下添加新概念和专家，提供了一种参数隔离的扩展机制，同时保持先前学习的专家不变。在AbdomenAtlas 3.0数据集上的评估表明，ReportMedSAM有效地解释了自由格式报告，实现了竞争性的分割准确性，并展示了对新临床任务的无缝、非干扰性扩展。
cs.CL / 18 / 2607.14117
Heterogeneous Element-Aware Cross-Version Differencing of Scientific Documents via Layout-Aware Alignment and Structure-Aware Reasoning
基于布局感知对齐和结构感知推理的科学文档异构元素感知跨版本差异化
Abstract
Cross-version differencing of scientific documents is essential in scholarly publishing and technical documentation, but remains challenging because scientific documents are page-structured artifacts containing heterogeneous elements such as text, tables, formulas, figures, and layout cues. Existing text-sequence-based methods often lose layout and structural information, while image-based methods lack semantic interpretability and are sensitive to rendering variation. To address these limitations, this paper proposes a layout-aware heterogeneous element-aware framework for scientific document differencing. The framework decomposes document versions into semantically typed elements, establishes cross-version correspondence through an alignment-first mechanism that jointly models spatial, content, and structural compatibility, and performs type-aware difference reasoning over aligned element pairs. It supports unified change detection, localization, structure-awareness analysis, and alignment/matching evaluation across text, tables, formulas, and figures. Experiments on real-world scientific PDF data from journal production proofreading workflows show that the proposed framework consistently outperforms element-specific baselines. It achieves detection F1 scores of 0.903, 0.855, 0.862, and 0.845 for text, tables, formulas, and figures, respectively, with further improvements in localization, structure awareness, and matching quality. Ablation and sensitivity analyses confirm the effectiveness of cross-version alignment, type-specific representations, structure-aware reasoning, and compatibility-weight design. These results demonstrate that heterogeneous element-aware differencing provides a robust and interpretable solution for scientific document comparison in realistic editorial production scenarios.
Chinese Translation
科学文档的跨版本差异化在学术出版和技术文档中至关重要，但由于科学文档是包含文本、表格、公式、图形和布局提示等异构元素的页面结构化工件，因此仍然面临挑战。现有的基于文本序列的方法往往会丢失布局和结构信息，而基于图像的方法缺乏语义可解释性，并且对渲染变化敏感。为了解决这些局限性，本文提出了一种面向布局的异构元素感知框架，用于科学文档的差异化。该框架将文档版本分解为语义类型元素，通过一种优先对齐机制建立跨版本对应关系，该机制联合建模空间、内容和结构的兼容性，并对对齐的元素对执行类型感知的差异推理。它支持文本、表格、公式和图形的统一变更检测、定位、结构感知分析以及对齐/匹配评估。在来自期刊生产校对工作流的真实科学PDF数据上的实验表明，所提框架在各个元素特定基线模型上均表现出一致的优越性。它在文本、表格、公式和图形上的检测F1分数分别达到0.903、0.855、0.862和0.845，并在定位、结构感知和匹配质量方面进一步提高。消融和敏感性分析确认了跨版本对齐、特定类型表示、结构感知推理和兼容性权重设计的有效性。这些结果表明，异构元素感知差异化为现实编辑生产场景中的科学文档比较提供了一种稳健且可解释的解决方案。
cs.CL / 19 / 2607.14118
Budgeted Subset Refinement for Execution-Aware LLM Research Ideation
面向执行的预算子集优化在大语言模型研究创意中的应用
Abstract
Large language models (LLMs) can generate research ideas that appear novel to expert reviewers, but recent work also shows that such ideas often lack diversity, are difficult for LLMs to evaluate reliably, and may fail to translate into strong executed projects. This paper evaluates a controlled proxy benchmark for a pre-execution scaffolding problem: given a noisy pool of LLM-generated research ideas, how should a system allocate limited refinement effort to construct a stronger, more diverse, more execution-aware portfolio for human researchers under a fixed rubric? We introduce Budgeted Subset Refinement, a family of strategies that refine only a selected subset of candidates rather than refining all candidates uniformly. In a unified shared-candidate-pool evaluation across 10 random seeds and 10 research-ideation environments, raw generation and reranking alone produce no research-strong nonduplicate ideas under the benchmark rubric, while refinement is necessary for strong proxy-rated portfolios. Uniform refinement produces strong individual ideas but is not the best portfolio-level allocation of compute. Random-k refinement is a strong low-cost baseline, while diversity-aware MMR-k refinement gives the best overall proxy tradeoff: the highest research-strong nonduplicate yield, the lowest duplicate rate among successful methods, and the best cost per research-strong nonduplicate idea. A blinded external-judge robustness check on a balanced 72-item sample supports the broad refinement effect across independent model families, while showing that per-item rankings among refined strategies vary by judge. These results suggest that LLM research ideation systems should be evaluated not only as idea generators, but as budgeted support-allocation systems. The claims are scoped to proxy-rated portfolio quality and do not substitute for expert review or execution-grounded validation.
Chinese Translation
大型语言模型（LLMs）能够生成对专家评审者而言看似新颖的研究创意，但近期的研究表明，这些创意往往缺乏多样性，难以被 LLMs 可靠地评估，并且可能无法转化为强有力的执行项目。本文评估了一种针对预执行支架问题的受控代理基准：在给定一组嘈杂的 LLM 生成的研究创意池的情况下，系统应如何分配有限的优化努力，以在固定标准下为人类研究者构建一个更强、更具多样性和更关注执行的投资组合？我们提出了预算子集优化（Budgeted Subset Refinement），这是一系列策略，仅对选定的候选子集进行优化，而不是均匀地优化所有候选者。在对 10 个随机种子和 10 个研究创意环境进行的统一共享候选池评估中，仅依靠原始生成和重新排序在基准标准下未能产生任何研究强度高且不重复的创意，而优化对于强代理评分的投资组合是必要的。均匀优化产生了强有力的单个创意，但并不是最佳的计算资源分配。随机-k 优化是一个强有力的低成本基线，而关注多样性的 MMR-k 优化提供了最佳的整体代理权衡：最高的研究强度非重复产出，成功方法中最低的重复率，以及每个研究强度非重复创意的最佳成本。对一个平衡的 72 项样本进行的盲评外部评审的稳健性检查支持了在独立模型家族中广泛的优化效果，同时显示出不同评审者之间优化策略的逐项排名存在差异。这些结果表明，LLM 研究创意系统应不仅作为创意生成器进行评估，还应作为预算支持分配系统。所提出的主张限于代理评分的投资组合质量，并不能替代专家评审或基于执行的验证。
cs.CL / 20 / 2607.14119
Semantic Register Compression in Multi-Agent LLM Cascades
多智能体大语言模型级联中的语义注册压缩
Abstract
Multi-agent LLM systems commonly decompose complex tasks into specialized roles. However, this modularity introduces a representational risk: when intermediate agents transform text across linguistic registers, they can systematically compress the semantic distinctions needed for accurate downstream decisions. We term this phenomenon semantic register compression and characterize it as an observable failure mode in multi-agent cascades. Using a three-agent pipeline (Collector-Evaluator-Decider), we quantify compression via inter-label separation in sentence-transformer embedding space. Across political fact-checking (LIAR), sentiment analysis (SST-5), and medical triage (Triagegeist), critical evaluation consistently reduces label separability by 41.7% at the Evaluator stage, while identity passthrough preserves it nearly fully. Five architectural variants causally isolate oriented semantic transformation as the primary driver. A credibility-seeking variant produces minimal geometric compression yet shifts outputs toward mostly-true, demonstrating that transformation valence controls the direction of distributional collapse independently of compression magnitude. Compression generalizes across the three domains with varying intensity: 41.7% in fact-checking, 27.2% in sentiment, and 20.0% in triage. Prompt-level regression explains 78% of the variance, with operational constraints associated with lower compression. These results demonstrate that semantic register compression is a measurable and generalizable phenomenon in multi-agent LLM systems, with implications for safety evaluation in high-stakes domains.
Chinese Translation
多智能体大语言模型系统通常将复杂任务分解为专业角色。然而，这种模块化引入了一种表征风险：当中间智能体在不同语言注册之间转换文本时，它们可能系统性地压缩准确下游决策所需的语义区分。我们将这一现象称为语义注册压缩，并将其表征为多智能体级联中的一种可观察的失败模式。通过使用三智能体管道（收集者-评估者-决策者），我们通过句子变换器嵌入空间中的标签间分离量化压缩。在政治事实核查（LIAR）、情感分析（SST-5）和医疗分诊（Triagegeist）中，关键评估在评估者阶段始终将标签可分性降低41.7%，而身份传递几乎完全保留了可分性。五种架构变体因果隔离了定向语义转换作为主要驱动因素。一种寻求可信度的变体产生了最小的几何压缩，但将输出偏向于大部分真实，表明转换效价独立于压缩幅度控制分布崩溃的方向。压缩在三个领域中以不同强度普遍存在：事实核查中为41.7%，情感分析中为27.2%，医疗分诊中为20.0%。提示级回归解释了78%的方差，操作约束与较低的压缩相关。这些结果表明，语义注册压缩是多智能体大语言模型系统中一种可测量且可推广的现象，对高风险领域的安全评估具有重要影响。
cs.CL / 21 / 2607.14131
Cross-Dataset Generalization in Urdu Fake News Detection: An Empirical Study with XLM-RoBERTa and a Length Confound Analysis
乌尔都语假新闻检测中的跨数据集泛化：基于XLM-RoBERTa的实证研究及长度混淆分析
Abstract
Urdu fake news detection remains under-resourced despite Urdu being spoken by over 231 million people worldwide. While prior work has demonstrated strong in-domain performance on individual Urdu datasets, cross-dataset generalisation has received little systematic attention. This paper presents the first cross-dataset generalisation study for Urdu fake news detection, using two publicly available balanced datasets: the Ax-to-Grind Urdu corpus (10,083 articles, 15 domains) and the Notri-Fact Urdu dataset (13,388 articles). We fine-tune xlm-roberta-base under four experimental conditions, in-domain on each dataset and two zero-shot cross-domain transfer directions, comparing against TF-IDF baselines using Logistic Regression and Support Vector Machines. Our experiments reveal a striking asymmetry: Notri-Fact to Ax-to-Grind transfer achieves a macro F1 of 0.771, while the reverse collapses to F1 of 0.005, with the model predicting fake for 99.7% of test articles. We demonstrate that this collapse stems from a systematic length confound in Ax-to-Grind, where fake articles average 117 words versus 35 for real articles, a 3.4x asymmetry inducing shortcut learning. A length ablation capping articles at 50 words yields only a 0.0067 F1 drop, confirming the confound inflates but does not solely drive in-domain performance. We provide a reusable diagnostic methodology that combines bidirectional transfer analysis and prediction-collapse inspection to identify confound-driven behavior in multilingual fake news detection settings.
Chinese Translation
尽管乌尔都语在全球有超过2.31亿人使用，但乌尔都语假新闻检测仍然资源不足。虽然之前的研究在单个乌尔都语数据集上展示了强大的领域内表现，但跨数据集泛化却鲜有系统性关注。本文首次针对乌尔都语假新闻检测进行跨数据集泛化研究，使用两个公开可用的平衡数据集：Ax-to-Grind乌尔都语语料库（10,083篇文章，15个领域）和Notri-Fact乌尔都语数据集（13,388篇文章）。我们在四种实验条件下微调xlm-roberta-base，包括每个数据集的领域内实验和两个零样本跨领域迁移方向，并与使用逻辑回归和支持向量机的TF-IDF基线进行比较。我们的实验揭示了一个显著的不对称性：从Notri-Fact到Ax-to-Grind的迁移实现了0.771的宏F1，而反向迁移则崩溃至0.005的F1，模型对99.7%的测试文章预测为假新闻。我们证明这种崩溃源于Ax-to-Grind中的系统性长度混淆，其中假新闻文章的平均字数为117，而真实文章为35，导致3.4倍的不对称性，从而引发了捷径学习。限制文章长度至50个字的消融实验仅导致0.0067的F1下降，确认了混淆因素在领域内表现中起到膨胀作用，但并非唯一驱动因素。我们提供了一种可重用的诊断方法，结合双向迁移分析和预测崩溃检查，以识别多语言假新闻检测环境中的混淆驱动行为。
cs.CL / 22 / 2607.14147
Breaking Refusal in the First Half: A Mechanistic Study of the Prefill Jailbreak
打破拒绝的前半部分：预填越狱的机制研究
Abstract
Aligned language models refuse harmful requests, but a one-line prefill ("Sure, here is") strips the refusal. We ask where and how it fails. The harm representation stays intact: on the prompts the attack flips to compliance, a linear probe reads harm as high as on the refused ones (0.91-0.98), while behavioral refusal drops to chance. This holds across four models and three families (1.5-3.8B, and at 14B). Refusal is therefore a shallow, response-site computation. We localize it to an early window: a dose-matched position control shows the first half of the response suffices to break refusal, while the second half is nearly inert. Three causal probes converge on that window. Restoring the harm direction there partially re-engages refusal. Injecting the model's own refuse-state reverses the jailbreak (74%, held-out). And knocking out the early response's attention to the prefill, but not an equal attention mass elsewhere, selectively collapses the harmful continuation. A base-model control identifies the mechanism: the same knockout collapses the continuation prefill-specifically even in a non-safety-tuned base model (64% to 25% harmful content vs a matched control's 64%, replicated at 7B). So the prefill's grip is generic autoregressive conditioning, not safety-specific suppression, and "refusal restoration" is a model-dependent fallback. The dominant mechanism is passive. A small safety-specific attractor remains on top (logit-trace concentration 0.24 vs 0.03), whose active-vs-passive character we size but do not fully separate. No single direction or component is a clean handle either: the decision is decodable but distributed, and refusal tracks harm rather than scary surface. The consequence is structural: a monitor reading the untouched prompt-side representation is immune by construction, but only to response-site attacks. The mechanism is diffuse; the failure surface is local.
Chinese Translation
对齐的语言模型拒绝有害请求，但一行预填（“当然，这里是”）剥夺了拒绝的效果。我们探讨了其失败的原因和方式。有害的表征保持不变：在攻击将提示转变为顺从时，线性探测器读取的有害程度与拒绝的提示相当高（0.91-0.98），而行为上的拒绝率降至偶然水平。这一现象在四个模型和三个家族（1.5-3.8B，以及14B）中均成立。因此，拒绝是一个浅层的响应位置计算。我们将其局限于一个早期窗口：剂量匹配的位置控制显示，响应的前半部分足以打破拒绝，而后半部分几乎无效。三个因果探测器汇聚于该窗口。在此恢复有害方向部分重新激活拒绝。注入模型自身的拒绝状态则逆转了越狱（74%，保留数据）。而消除早期响应对预填的关注，但不影响其他地方的相等关注量，选择性地使有害的延续崩溃。基础模型控制识别了这一机制：同样的消除在非安全调优的基础模型中也特定于预填地崩溃延续（有害内容从64%降至25%，与匹配控制的64%相比，在7B中重复）。因此，预填的影响是通用自回归条件，而非安全特定的抑制，“拒绝恢复”是模型依赖的后备机制。主导机制是被动的。一个小的安全特定吸引子仍然存在于顶部（logit-追踪浓度为0.24 vs 0.03），我们对其主动与被动特性进行了量化但未完全分离。没有单一的方向或组件是一个清晰的处理方式：决策是可解码的但分布式的，拒绝跟踪的是有害而非表面的恐怖。其结果是结构性的：一个读取未触及的提示侧表征的监控器在构造上是免疫的，但仅对响应位置的攻击。该机制是弥散的；失败表面是局部的。
cs.CL / 23 / 2607.14242
Implicit Reasoning Steering via Concept Chaining
通过概念链引导隐式推理
Abstract
Large language models often appear to reason reliably, yet on many questions repeated sampling yields both correct and incorrect answers, revealing an underlying fragility in how final decisions are formed. We study whether this fragility can be exploited through implicit reasoning steering: using natural-language text to bias a model toward a designated answer without explicit instructions, triggers, or direct answer cues. Our approach, Concept Chaining, generates a short connection paragraph that links question entities to a target option through one or two intermediate concepts. We then continue pretraining a victim model on these connection paragraphs and evaluate whether its answer preference shifts on the original multiple-choice questions. Our results show that indirect, natural-looking text can systematically steer model predictions while remaining substantially less inferable than direct paraphrases, which shows that reasoning brittleness is not merely an evaluation artifact: it creates a practical channel through which latent biases can be amplified by ordinary-looking text to covertly redirect model decisions.
Chinese Translation
大型语言模型在推理方面常常表现出可靠性，然而在许多问题上，重复采样会产生正确和错误的答案，揭示了最终决策形成过程中的潜在脆弱性。我们研究这种脆弱性是否可以通过隐式推理引导来利用：使用自然语言文本在没有明确指令、触发器或直接答案提示的情况下，偏向模型朝向指定答案。我们的方法，概念链（Concept Chaining），生成一个简短的连接段落，通过一个或两个中介概念将问题实体与目标选项联系起来。然后，我们继续对受害模型进行预训练，使用这些连接段落，并评估其在原始多项选择问题上的答案偏好是否发生变化。我们的结果表明，间接的、自然的文本可以系统性地引导模型预测，同时在推理上显著低于直接的释义，这表明推理脆弱性不仅仅是评估伪影：它创造了一个实际渠道，通过普通的文本放大潜在偏见，从而隐秘地重定向模型决策。
cs.CL / 24 / 2607.14250
The Severance Problem: LLMs are Unaware of the Person Beyond the Prompt
割离问题：大型语言模型对提示之外的人缺乏意识
Abstract
Personal AI assistants have attracted significant interest for their potential to enhance everyday life by automating routine tasks, supporting consequential decisions, and assisting with everyday personal matters. Yet despite rapid recent technical advances, these assistants continue to exhibit undesirable behaviors, such as sycophancy, overconfidence, and hallucination. We argue that these failures stem from a fundamental limitation: language models lack an explicit representation of the person beyond the context they are given, which we term as the \textbf{Severance Problem}. Even with rich personal context and strong commonsense reasoning capabilities from the backbone model, current AI assistants fail to represent what remains unknown about the user. We propose a simple solution: incorporating structured ignorance into the language model context via the \textbf{Severance Schema}, which explicitly outlines dimensions along which the model lacks knowledge about the user, including physicality, temporality, consequences, continuity, multiplicity, and interiority. Empirically, across five model families, with the Severance Schema, the assistant consistently reduces sycophancy, harmful advice, and hallucination. Notably, models with the schema ask clarifying questions when information about the user is missing, rather than confidently extrapolating from incomplete user information.
Chinese Translation
个人人工智能助手因其通过自动化日常任务、支持重要决策和协助处理日常个人事务而受到广泛关注。然而，尽管近期技术迅速进步，这些助手仍然表现出一些不理想的行为，例如谄媚、过度自信和幻觉。我们认为，这些失败源于一个根本性的局限性：语言模型缺乏对提示之外的人的明确表征，我们称之为 extbf{割离问题}（Severance Problem）。即使拥有丰富的个人背景和强大的常识推理能力，当前的人工智能助手仍无法表征用户的未知信息。我们提出了一个简单的解决方案：通过 extbf{割离模式}（Severance Schema）将结构化的无知纳入语言模型的上下文中，该模式明确列出了模型在用户知识方面的缺失维度，包括身体性、时间性、后果、连续性、多样性和内在性。实证研究表明，在五个模型家族中，使用割离模式的助手始终能够减少谄媚、有害建议和幻觉。值得注意的是，带有该模式的模型在缺乏用户信息时会提出澄清性问题，而不是自信地从不完整的用户信息中推断。
cs.CL / 25 / 2607.14277
Multi-Head Latent Control: A Unified Interface for LLM Agent Decision Making
多头潜在控制：大型语言模型代理决策的统一接口
Abstract
Large language models are increasingly deployed as agents, but reliable agentic behavior requires more than next-token prediction. At inference time, it is preferred that an agent can decide whether to proceed with its current reasoning, defer to a stronger model, request additional information, invoke external tools, or abstain under the given setup. Existing approaches address these decisions through prompt-level routing, external orchestration, or task-specific fine-tuning, which primarily rely on input-side signals, and are often costly and difficult to maintain as model backbones evolve. We ask whether such control decisions can be inferred directly from a model's latent generation process. We introduce Multi-Head Latent Control, a lightweight layer that reads hidden-state trajectories from a frozen LLM or VLM to produce deployment-time control signals. A Capability Head predicts whether the current model can solve the instance or should defer to a stronger collaborator, while a Resolution Head predicts appropriate resolution decision Clarification, Tool Use, Abstention, or Direct Answering. Both heads are trained only on latent traces from the same frozen LLM backbone, enabling post hoc adaptation without modifying the model. Across language and vision-language settings, Multi-Head Latent Control consistently improves the quality-cost tradeoff of multi-model systems, enabling early handoff from partial generations and more accurate intervention decisions. In routed execution (small + large model), it reduces large-model usage by up to 90.7 percent on AndroidWorld and 27-53 percent on average across benchmarks, while retaining most of large-model performance. Additionally, the learned control signals improve tool-use decision quality, yielding up to +158 percent relative score gain and 65.5 percent fewer missed-required tool calls.
Chinese Translation
大型语言模型越来越多地被部署为代理，但可靠的代理行为不仅仅依赖于下一个标记的预测。在推理时，代理需要能够决定是否继续当前的推理、推迟到更强的模型、请求额外信息、调用外部工具或在给定的设置下选择不采取行动。现有的方法通过提示级路由、外部协调或特定任务的微调来处理这些决策，这些方法主要依赖于输入端信号，且随着模型骨干的演变，通常成本高昂且难以维护。我们探讨这些控制决策是否可以直接从模型的潜在生成过程中推断出来。我们提出了多头潜在控制（Multi-Head Latent Control），这是一种轻量级层，它从冻结的LLM（大型语言模型）或VLM（视觉语言模型）中读取隐藏状态轨迹，以生成部署时的控制信号。能力头（Capability Head）预测当前模型是否能够解决实例或应推迟到更强的协作者，而决策头（Resolution Head）预测适当的决策，包括澄清、工具使用、放弃或直接回答。这两个头仅在同一冻结的LLM骨干的潜在痕迹上进行训练，从而实现后期适应而无需修改模型。在语言和视觉语言设置中，多头潜在控制持续改善多模型系统的质量-成本权衡，使得能够更早地从部分生成中移交并做出更准确的干预决策。在路由执行（小模型 + 大模型）中，它在AndroidWorld上将大模型的使用减少了高达90.7%，在各基准测试中平均减少了27-53%，同时保留了大模型的大部分性能。此外，学习到的控制信号提高了工具使用决策的质量，带来了高达158%的相对得分提升和65.5%的所需工具调用遗漏减少。
cs.CL / 26 / 2607.14327
PReM: Learning What to Preserve and When to Refresh for Context Compression
PReM：学习在上下文压缩中保留什么以及何时刷新
Abstract
Efficient long-context inference is not only about reducing memory cost, but also about keeping useful contextual evidence accessible as generation proceeds. However, existing compression-oriented approaches, such as key-value (KV) cache compression and context compression, often either make an early decision about which contextual information to keep or rely on an external compressor. Such designs make it difficult to adapt the compressed context to the evidence needed by later reasoning steps. This paper introduces PReM (Preserve and Refresh Memory), a context-compression framework that maintains the long context as the model's internal layer-wise KV memory and learns what to preserve and when to refresh it. Specifically, PReM uses a dedicated memory layer to make memory-selection decisions, and a special memory token to trigger refreshes during generation. To train this behavior, PReM introduces Phase-Separated Refresh Training, aligning memory selection with memory-conditioned generation while preserving continuity across refreshes. Experiments with 32K-token contexts show that PReM outperforms strong baselines under both 16x and 32x compression, while maintaining a favorable balance between answer quality and inference efficiency.
Chinese Translation
高效的长上下文推理不仅涉及降低内存成本，还包括在生成过程中保持有用的上下文证据可访问。然而，现有的以压缩为导向的方法，如键值（KV）缓存压缩和上下文压缩，往往要么过早决定保留哪些上下文信息，要么依赖外部压缩器。这些设计使得难以将压缩后的上下文适应于后续推理步骤所需的证据。本文介绍了PReM（保留与刷新记忆），一种上下文压缩框架，它将长上下文保持为模型的内部层级KV记忆，并学习保留什么以及何时刷新。具体而言，PReM使用专用的记忆层来做出记忆选择决策，并使用特殊的记忆标记在生成过程中触发刷新。为了训练这种行为，PReM引入了阶段分离刷新训练，将记忆选择与记忆条件生成对齐，同时保持刷新之间的连续性。使用32K标记上下文的实验表明，PReM在16倍和32倍压缩下均优于强基线，同时在答案质量和推理效率之间保持良好的平衡。
cs.CL / 27 / 2607.14349
HABIB_TAZ at SemEval-2026 Task 11: Disentangling Formal Logic from Content via Synthetic Training and Multi-Objective Optimization
HABIB_TAZ在SemEval-2026任务11中的表现：通过合成训练和多目标优化将形式逻辑与内容分离
Abstract
While Large Language Models (LLMs) excel in many general NLP tasks, their formal reasoning capabilities are often compromised by content effects, demonstrating a measurable bias towards real-world plausibility. In this paper, we present our system for SemEval-2026 Task 11, which evaluates the ability of models to disentangle formal logic from content across 12 languages with and without distractor premises. We address this challenge using mDeBERTa-v3 networks fine-tuned on a synthetic, rule-based dataset of syllogistic schemes to avoid the semantic noise of LLM-augmented data. To explicitly decouple plausibility from logical structure, our training pipeline employs a multi-objective loss function combining Adaptive Group Distributionally Robust Optimization (DRO), a scheduled differentiable bias penalty, and KL-Divergence consistency regularization. Our system achieved #1 ranks and perfect Ranking Scores (100.0) with 0.00% bias and 100.0% accuracy on Subtask 1 (English), Subtask 2 (Noisy English), and Subtask 3 (Multilingual). On the highly complex Subtask 4 (Noisy Multilingual), the system achieved the 6th rank with 89.06% Accuracy and F1-score, alongside a limited 2.89% Bias and a 37.78 Ranking Score. Our dataset generation engine and codebase are publicly available to facilitate future work on robust logical reasoning.
Chinese Translation
尽管大型语言模型（LLMs）在许多通用自然语言处理（NLP）任务中表现出色，但其形式推理能力常常受到内容效应的影响，表现出对现实世界合理性的可测偏见。本文介绍了我们在SemEval-2026任务11中的系统，该任务评估模型在12种语言中将形式逻辑与内容分离的能力，包括有和没有干扰前提的情况。我们使用在合成的基于规则的三段论数据集上微调的mDeBERTa-v3网络来应对这一挑战，以避免LLM增强数据的语义噪声。为了明确将合理性与逻辑结构解耦，我们的训练流程采用了一个多目标损失函数，结合了自适应组分布鲁棒优化（DRO）、计划的可微偏差惩罚和KL散度一致性正则化。我们的系统在子任务1（英语）、子任务2（嘈杂英语）和子任务3（多语言）中获得了第一名的排名和完美的排名分数（100.0），偏见为0.00%，准确率为100.0%。在高度复杂的子任务4（嘈杂多语言）中，该系统获得了第六名，准确率和F1分数均为89.06%，同时偏见限制在2.89%，排名分数为37.78。我们的数据集生成引擎和代码库已公开，以促进未来在稳健逻辑推理方面的研究。
cs.CL / 28 / 2607.14385
MamaBench: Benchmarking LLM Robustness in Maternal and Child Health Diagnosis through Counterfactual Clinical Perturbation
MamaBench：通过反事实临床扰动评估母婴健康诊断中的大型语言模型鲁棒性
Abstract
Large language models achieve strong scores on medical benchmarks, yet these benchmarks evaluate each question in isolation, providing no measure of whether a system can distinguish clinically similar presentations requiring different interventions. We introduce MamaBench, the first counterfactual benchmark for maternal and paediatric AI: 434 expert-authored clinical narratives in 217 pairs across 371 pathologies, evaluated via the Bias Trap Rate (BTR), the conditional probability that a model fails the counterfactual given success on the base case. We propose Evidence-Anchored RAG (EA-RAG), a three-stage retrieval method that replaces aggregate similarity with an evidence coverage objective through clinical parameter extraction, coverage auditing, and contrastive sub-queries. Across eight configurations of four frontier LLMs, base accuracy overstates robust accuracy by 16-28 percentage points in every model. EA-RAG achieves 20.3% BTR and 65.0% robust accuracy on Claude Sonnet 4.6, a 5.5 percentage point BTR reduction without degrading base accuracy. The residual 20% BTR confirms that counterfactual robustness in clinical AI remains an open challenge. Keywords: counterfactual evaluation, clinical AI, maternal healthcare, retrieval-augmented generation, diagnostic robustness
Chinese Translation
大型语言模型在医学基准测试中取得了优异的成绩，但这些基准测试是孤立地评估每个问题，并未测量系统是否能够区分需要不同干预措施的临床相似表现。我们介绍了MamaBench，这是首个针对母婴和儿童AI的反事实基准：包含434个专家撰写的临床叙述，涵盖217对、371种病理，通过偏差陷阱率（Bias Trap Rate, BTR）进行评估，即模型在基准案例成功的情况下，未能应对反事实的条件概率。我们提出了证据锚定检索增强生成（Evidence-Anchored RAG, EA-RAG），这是一种三阶段的检索方法，通过临床参数提取、覆盖审计和对比子查询，将聚合相似性替换为证据覆盖目标。在四种前沿大型语言模型的八种配置中，基础准确率在每个模型中都高估了鲁棒准确率16-28个百分点。EA-RAG在Claude Sonnet 4.6上实现了20.3%的BTR和65.0%的鲁棒准确率，BTR减少了5.5个百分点，而基础准确率未受影响。剩余的20% BTR确认了临床AI中的反事实鲁棒性仍然是一个未解决的挑战。关键词：反事实评估，临床AI，母婴保健，检索增强生成，诊断鲁棒性
cs.CL / 29 / 2607.14400
DS@GT ARC at LongEval: Citation Integrity and Factual Grounding in Scientific QA
DS@GT ARC在LongEval中的表现：科学问答中的引用完整性与事实基础
Abstract
This paper describes DS@GT ARC's submission to the CLEF 2026 LongEval Task 4 on Retrieval-Augmented Generation (RAG). In this submission, we examine a divergence between traditional natural language evaluation metrics and citation integrity as applied to RAG QA systems. We evaluate a corrective pipeline using Corrective RAG (CRAG) and CiteFix against baseline and frontier model benchmark RAG QA scores. While frontier models maximized answer relevance and fluency scores, our RAGAs LLM-as-judge diagnostics indicate that frontier models would correctly identify relevant documents without using their context in answer generation. Conversely, by filtering chunks pre-generation and enforcing strict entailment of generated claims to the cited material post-generation, our corrective pipeline marginally improved citation faithfulness and answer grounding. We propose that evaluation of trustworthy RAG QA requires metrics that reward strict answer grounding.
Chinese Translation
本文描述了DS@GT ARC对CLEF 2026 LongEval任务4（检索增强生成，RAG）的提交。在此提交中，我们考察了传统自然语言评估指标与应用于RAG问答系统的引用完整性之间的差异。我们使用纠正性RAG（Corrective RAG, CRAG）和CiteFix评估了一个纠正性流程，并与基准和前沿模型的RAG问答得分进行了对比。尽管前沿模型最大化了答案的相关性和流畅性得分，但我们的RAGAs LLM-as-judge诊断表明，前沿模型能够在不使用上下文的情况下正确识别相关文献。相反，通过在生成前过滤文本块，并在生成后强制生成的声明与引用材料之间的严格蕴涵，我们的纠正性流程在引用的可信度和答案的基础上略有改善。我们建议，评估可信的RAG问答需要奖励严格答案基础的指标。
cs.CL / 30 / 2607.14431
Smarter and Cheaper at Once: Byte-Exact KV-Cache Grafting Turns a Frozen Small Model into a Verified-Knowledge Flywheel
同时更智能和更便宜：字节精确的键值缓存嫁接将一个冻结的小模型转变为验证知识的飞轮
Abstract
We report a way to make a frozen small language model both more capable and dramatically cheaper at once, without changing any weights. Verified knowledge is deposited once as a byte-exact key-value (KV) state artifact and later restored, by graft, into a fresh inference context. The restore is bit-exact: under a pinned deterministic configuration, the grafted logits are byte-for-byte identical to a fresh computation (SHA-256 equality), with zero KL divergence and 100% argmax agreement over fifty samples. We show that own-position graft is the unique numerically exact operating point on a model with floating-point rotary encoding, and we verify byte-exactness on two model scales (12B, 31B) and two GPU targets, one through a pre-registered replay. On AIME 2025, a frozen Gemma-4-12B moves from 80.0% to 93.3% once a verified solution library is grafted, above its own 77.5% and its 31B sibling's 89.2% published anchors. On the recurring case, eight problems the base model never solves within a 401,026-token budget are answered from cached verified solutions in 61 total decode tokens, a factor of 6,574 fewer tokens and about 8,700x less energy; the capability claim proper rests on held-out transfer (7 of 7 at 31B). The same byte-exact store widens usable context from 32,768 to 2,854,766 tokens at zero extra accelerator memory, and moves byte-identical between machines of the same architecture. We describe the system at the behavior level; the engine is proprietary, and every reported number is backed by committed input and output hashes so the scoring can be re-checked without it.
Chinese Translation
我们报告了一种方法，使得一个冻结的小型语言模型在不改变任何权重的情况下，既更强大又显著更便宜。验证知识作为字节精确的键值（KV）状态工件一次性存储，随后通过嫁接恢复到一个新的推理上下文中。恢复过程是逐位精确的：在固定的确定性配置下，嫁接的逻辑值与新的计算逐字节相同（SHA-256相等），KL散度为零，且在五十个样本上具有100%的argmax一致性。我们展示了自位置嫁接是具有浮点旋转编码模型的唯一数值精确操作点，并在两个模型规模（12B，31B）和两个GPU目标上验证了字节精确性，其中一个通过预注册重放进行。在AIME 2025上，一个冻结的Gemma-4-12B在嫁接了验证解决方案库后，从80.0%提升至93.3%，超出其自身的77.5%和其31B同类的89.2%已发布基准。在重复案例中，基础模型在401,026个标记预算内从未解决的八个问题通过缓存的验证解决方案在61个总解码标记中得到回答，减少了6,574个标记和约8,700倍的能量消耗；能力声明的有效性基于保留的迁移（31B的7个问题中解决了7个）。相同的字节精确存储将可用上下文从32,768扩展到2,854,766个标记，且没有额外的加速器内存，并在相同架构的机器之间保持字节一致性。我们在行为层面描述了该系统；引擎是专有的，所有报告的数字都有承诺的输入和输出哈希支持，以便在没有引擎的情况下重新检查评分。
cs.CL / 31 / 2607.14480
LLM Evaluators are Biased across Languages
大型语言模型评估者在不同语言间存在偏见
Abstract
LLM evaluators (trained reward models and prompted LLM-as-a-Judge) are routinely validated via pairwise accuracy. In a multilingual setting, this operates under the premise that high pairwise accuracy implies reliable, language-neutral scoring. We show that this assumption does not hold. We conduct experiments with semantically identical instruction-response pairs across 23 languages, and find that multilingual evaluators assign significantly different scores to different evaluation languages. The bias is statistically significant and consistent across eight open-weight evaluators of different architectures and training paradigms, persists in frontier judges, and is strongly correlated with language resource level: lower-resource languages are scored more generously. Meanwhile, these biases are invisible to pairwise accuracy: evaluators achieve above 90% pairwise accuracy, yet have up to 43% difference in acceptance rate across languages under a global decision threshold, meaning, for instance, that harmful content in lower-resource languages is more likely to pass safety filters. Per-language thresholds would require language identification, which can be defeated by code-switched prompts. We then investigate why lower-resource languages receive higher rather than lower scores, and we find that model uncertainty is linked with the effect: models tend to give higher scores when less confident, both under negative log-likelihood and under token-free uncertainty measures; however, language identity remains a significant predictor after controlling for uncertainty, and the bias cannot be explained away by content difficulty alone, but is a structural, language-level misalignment.
Chinese Translation
大型语言模型评估者（训练的奖励模型和提示的语言模型作为评审）通常通过成对准确性进行验证。在多语言环境中，这一操作基于高成对准确性意味着可靠的、语言中立的评分这一前提。我们证明这一假设并不成立。我们在23种语言中对语义相同的指令-响应对进行了实验，发现多语言评估者对不同评估语言赋予了显著不同的分数。这种偏见在统计上显著，并且在八种不同架构和训练范式的开放权重评估者中一致存在，且在前沿评审中持续存在，并且与语言资源水平强相关：资源较少的语言获得的评分更为宽松。同时，这些偏见对成对准确性是不可见的：评估者的成对准确性超过90%，但在全球决策阈值下，各语言的接受率差异可达43%，这意味着例如，资源较少语言中的有害内容更可能通过安全过滤。每种语言的阈值需要语言识别，而这可以被代码切换的提示所击败。随后，我们调查了为何资源较少的语言获得更高而非更低的分数，发现模型的不确定性与这一效应相关：模型在信心较低时倾向于给出更高的分数，无论是在负对数似然下还是在无标记不确定性度量下；然而，在控制不确定性后，语言身份仍然是一个重要的预测因子，且这一偏见不能仅通过内容难度来解释，而是结构性、语言层面的不匹配。
cs.CL / 32 / 2607.14528
Controlled Reformulation Testing for Logical Consistency in Large Language Models
大语言模型的逻辑一致性控制重构测试
Abstract
Large language models (LLMs) frequently contradict themselves when the surface form of a logically equivalent question changes. We present a benchmark of 350 question families (1,750 total questions) for Controlled Reformulation Testing (CRTBench) to evaluate logical invariance. In this benchmark, we investigate LLMs' ability to maintain consistent answers across controlled reformulations, which include contrapositive rewriting, double negation, negation flipping, and passive voice. We evaluate several frontier LLMs and observe an accuracy-consistency gap where GPT-5.4-mini achieves $98.9\%$ base accuracy but only $60.3\%$ family-level consistency, while reasoning-optimized o4-mini achieves $96.9\%$ consistency. From our experiments, we observe that failures cluster around logically nontrivial transformations such as contrapositive rewriting ($72.4\%$ for GPT-5.4-mini) and double negation ($84.6\%$), while surface-level rephrasing remains robust ($94-100\%$). Increasing reasoning effort improves GPT-5.4-mini to $85.4\%$ consistency, but leaves GPT-5.4 unchanged overall because gains on nested negation are offset by failures on quantifier families. These results show that accuracy alone is not enough for evaluating logical reasoning in LLMs.
Chinese Translation
大语言模型（LLMs）在逻辑等价问题的表面形式变化时，常常会自相矛盾。我们提出了一个包含350个问题家族（共1,750个问题）的基准测试（Controlled Reformulation Testing，CRTBench），用于评估逻辑不变性。在这个基准测试中，我们研究了LLMs在控制重构中保持一致答案的能力，这些重构包括对立重写、双重否定、否定翻转和被动语态。我们评估了几种前沿LLMs，并观察到准确性与一致性之间存在差距，其中GPT-5.4-mini的基础准确率达到98.9%，但家族级一致性仅为60.3%，而经过推理优化的o4-mini则达到了96.9%的一致性。从我们的实验中，我们观察到失败主要集中在逻辑上非平凡的变换上，例如对立重写（GPT-5.4-mini为72.4%）和双重否定（84.6%），而表面级的重述保持稳健（94-100%）。增加推理努力使GPT-5.4-mini的一致性提高至85.4%，但整体上对GPT-5.4没有影响，因为在嵌套否定上的收益被量词家族的失败所抵消。这些结果表明，仅靠准确性不足以评估LLMs中的逻辑推理能力。
cs.CL / 33 / 2607.14542
CityLLM: A framework for natural-language querying of semantic 3D city models
CityLLM：一种用于语义3D城市模型的自然语言查询框架
Abstract
Semantic 3D city models provide rich geometric and semantic information, but remain challenging for non-experts and interdisciplinary researchers to access and query due to their complex structures and specialized data formats. To address this issue, we present CityLLM, a framework for natural-language querying of semantic 3D city models alongside complementary urban datasets. The framework combines spatial and graph databases within an LLM-based workflow that supports iterative query refinement and cross-database chaining. We evaluate CityLLM on a CityJSON dataset of Rotterdam (853 LoD2 buildings) using GPT-OSS, Gemini 3.1, and GPT-5.4, along with selected variants, across multiple metrics: answer correctness, visualization correctness, query success, and retry attempts. A total of 54 natural-language queries are curated across four scenarios: spatial, graph, cross-database, and conversational. Results show strong overall performance, with answer correctness ranging from 85.2% to 100%, visualization correctness from 92.9% to 100%, a 100% query success rate, and fewer than three retries across all 54 queries. Overall, the findings suggest that CityLLM provides a lightweight and extensible approach for conversational access to semantic 3D city data.
Chinese Translation
语义3D城市模型提供了丰富的几何和语义信息，但由于其复杂的结构和专业的数据格式，非专业人士和跨学科研究者在访问和查询时仍面临挑战。为了解决这一问题，我们提出了CityLLM，一个用于语义3D城市模型及其补充城市数据集的自然语言查询框架。该框架结合了空间数据库和图数据库，并在基于大语言模型（LLM）的工作流程中支持迭代查询优化和跨数据库链式查询。我们在鹿特丹的CityJSON数据集（853个LoD2建筑）上评估了CityLLM，使用了GPT-OSS、Gemini 3.1和GPT-5.4，以及选定的变体，评估指标包括答案正确性、可视化正确性、查询成功率和重试次数。在四种场景下（空间、图、跨数据库和对话）共策划了54个自然语言查询。结果显示，整体性能良好，答案正确率在85.2%到100%之间，可视化正确率在92.9%到100%之间，查询成功率为100%，在所有54个查询中重试次数少于三次。总体而言，研究结果表明，CityLLM为语义3D城市数据提供了一种轻量且可扩展的对话访问方式。
cs.CL / 34 / 2607.14552
Answer-Conditioned Chains of Thought Degrade Verifiable-Reasoning Distillation in Large Language Models
答案条件链思维降低大型语言模型中的可验证推理蒸馏效果
Abstract
A standard recipe for distilling the reasoning ability of large language models (LLMs) is to sample chains of thought from the model, keep those that reach the correct final answer, and fine-tune on the survivors. When sampling fails, a common fix shows the generator the gold answer and asks it to write a chain that reaches that answer. We show that this second step degrades the training data in a way that correctness filtering cannot catch. We run a controlled experiment that fixes the generator, the problem set, and the correctness filter, and varies only whether the chain is generated under answer-conditioning, the gold answer shown with a request to reach it. Training a strong instruction-tuned reasoning model on its own answer-conditioned chains sharply lowers its verifiable-reasoning accuracy. The loss grows with difficulty, reaching as much as about 27 points on the hardest competition problems. The mechanism is legible in the chains themselves, which rationalize backward from the shown answer instead of deriving it, with the early final-answer statement as the measurable symptom. The harm is a property of the data rather than the generator, read off unlabeled generations before any fine-tuning, ordering the penalty across eight thinking models from four families, and transferring across teacher families. A prompt ablation localizes it to the rationalize-toward instruction rather than the answer's bare visibility. The practical takeaway is to generate answer-blind, because no correctness filter can see this damage in the data.
Chinese Translation
蒸馏大型语言模型（LLMs）推理能力的标准方法是从模型中采样思维链，保留那些能够达到正确最终答案的链，并对存活下来的链进行微调。当采样失败时，一种常见的修复方法是向生成器展示金标准答案，并要求其写出达到该答案的思维链。我们表明，这第二步以一种正确性过滤无法捕捉的方式降低了训练数据的质量。我们进行了一个控制实验，固定生成器、问题集和正确性过滤器，仅改变思维链是否在答案条件下生成，即是否展示金标准答案并请求达到该答案。对其自身答案条件链进行训练的强指令调优推理模型，其可验证推理准确性显著下降。损失随着难度增加而增加，在最难的竞争问题上达到约27分。该机制在思维链中是可识别的，这些链是从展示的答案向后推理，而不是从头推导，早期的最终答案陈述是可测量的症状。这种损害是数据的特性而非生成器的特性，可以从任何微调前的未标记生成中读取，跨越四个家族的八个思维模型排序处罚，并在教师家族之间转移。一个提示消融实验将其局限于朝向理性化指令，而非答案的裸露可见性。实际的启示是生成无答案的思维链，因为没有任何正确性过滤器能够在数据中看到这种损害。
cs.CL / 35 / 2607.14561
MARS: Multi-hop Adaptive Retrieval and SPARQL Generation for KGQA
MARS：用于知识图谱问答的多跳自适应检索与SPARQL生成
Abstract
Large language models (LLMs) have demonstrated strong reasoning performance, but their tendency to hallucinate limits their reliability in knowledge-intensive tasks requiring up-to-date and grounded information. Combining knowledge graphs (KGs) with LLMs facilitates the use of explicit symbolic knowledge that can be continuously updated without costly fine-tuning, while benefiting from rapidly advancing LLM reasoning. We propose MARS, a scalable knowledge graph question answering (KGQA) approach that requires no model fine-tuning. Rather than relying on open-ended agentic exploration, MARS performs a structured retrieval procedure that links question entities to the KG and iteratively retrieves relevant next-hop information. At each step, MARS decides whether to continue graph traversal or to generate the final SPARQL query, allowing the model to adapt the retrieval depth to the question while keeping the overall pipeline more predictable than fully agentic approaches. We evaluate MARS on three established KGQA benchmarks across several LLMs and settings, including multilingual evaluation, and provide insights through ablation studies and error analysis. Our approach achieves competitive performance relative to state-of-the-art methods while remaining efficient and scalable. The evaluation results, code and resources are publicly available: https://github.com/dice-group/mars-kgqa.
Chinese Translation
大型语言模型（LLMs）已展示出强大的推理能力，但其产生幻觉的倾向限制了其在需要最新和基于事实的信息的知识密集型任务中的可靠性。将知识图谱（KGs）与LLMs结合，有助于使用可以不断更新的显式符号知识，而无需昂贵的微调，同时受益于快速发展的LLM推理能力。我们提出了MARS，一种可扩展的知识图谱问答（KGQA）方法，且不需要模型微调。MARS并不依赖于开放式的自主探索，而是执行一种结构化的检索程序，将问题实体链接到知识图谱，并迭代检索相关的下一跳信息。在每一步中，MARS决定是继续图遍历还是生成最终的SPARQL查询，从而使模型能够根据问题调整检索深度，同时保持整体流程比完全自主的方法更具可预测性。我们在多个大型语言模型和设置下，对MARS在三个已建立的KGQA基准上进行了评估，包括多语言评估，并通过消融研究和错误分析提供了见解。我们的方案在相对于最先进的方法中表现出竞争力，同时保持高效和可扩展。评估结果、代码和资源已公开可用：https://github.com/dice-group/mars-kgqa。
cs.CL / 36 / 2607.14591
How Well Does AI-Generated Feedback Work? Intrinsic and Extrinsic Evaluation across more than 20,000 EFL Essay Drafts
人工智能生成反馈的有效性如何？超过20,000份英语作为外语（EFL）作文草稿的内在与外在评估
Abstract
This study examines feedback in English as a Foreign Language (EFL) writing contexts, focusing on written corrective feedback (WCF). Large language models (LLMs) can provide WCF at scale, but aligning them with pedagogical best practices remains an ongoing challenge. WCF meeting criteria like factuality or relevance may still be unsuitable for learning contexts, highlighting the need for extrinsic evaluation based on the learner's perspective. We deployed WCF systems in a university-level EFL class with nearly 2,000 students, collecting over 20,000 drafts. We evaluated the generated WCF from two perspectives: intrinsic evaluation by experienced English teachers using a rubric, and extrinsic evaluation via student feedback and engagement metrics. Results revealed low alignment between teacher expert ratings and student feedback. These findings suggest that traditional expert evaluation alone may not fully capture WCF's usability or helpfulness from the learner's perspective, highlighting the importance of learner-centered evaluation frameworks for AI-based applications in language education.
Chinese Translation
本研究考察了英语作为外语（EFL）写作环境中的反馈，重点关注书面纠正反馈（WCF）。大型语言模型（LLMs）能够大规模提供WCF，但将其与教学最佳实践对齐仍然是一个持续的挑战。符合事实性或相关性等标准的WCF在学习环境中可能仍不适用，这突显了基于学习者视角的外在评估的必要性。我们在一所大学的EFL课程中部署了WCF系统，参与学生近2,000人，收集了超过20,000份草稿。我们从两个角度评估生成的WCF：由经验丰富的英语教师使用评分标准进行的内在评估，以及通过学生反馈和参与度指标进行的外在评估。结果显示，教师专家评分与学生反馈之间的对齐程度较低。这些发现表明，仅依靠传统的专家评估可能无法充分捕捉WCF在学习者视角下的可用性或帮助性，强调了以学习者为中心的评估框架在语言教育中应用于基于人工智能的工具的重要性。
cs.CL / 37 / 2607.14605
Investigating first-language bias in LLM-based automated essay scoring: A cross-prompt evaluation of an open-weight AI-model on TOEFL essays
研究基于大型语言模型的自动化作文评分中的母语偏见：对TOEFL作文的跨提示评估
Abstract
This study examines the cross-prompt generalization and first-language (L1) scoring effects of a LoRA-adapted open-weight large language model (Gemma-3-27B-it) applied to automated essay scoring. Using the identical model and inference configuration reported in "AiAWE: An Open-Source LLM Automated Writing Evaluation System Using LoRA-Adapted Instruction-Tuned Models" (Gayed, 2026), which was fine-tuned on 480 argumentative essays from two prompts, we evaluate scoring accuracy on the full TOEFL11 corpus: 12,100 essays written by test-takers from 11 first-language backgrounds across eight prompts, none of which were seen during training. The model's raw scores (0.5-5.0) are mapped to the same three proficiency bands (low, medium, high) used by ETS, enabling direct comparison. The model achieved an overall band agreement of 77.79% and a quadratic weighted kappa of 0.702, with adjacent-band agreement of 99.98%. Accuracy was stable across all eight unseen prompts, with no advantage for prompts thematically related to the training data, indicating robust cross-prompt generalization. However, the model exhibited a systematic, L1-linked scoring offset. Within every proficiency band, essays from European-language backgrounds received consistently higher scores than essays from East-Asian-language backgrounds, a pattern not attributable to the composition of the fine-tuning data. This is the first large-scale L1 fairness analysis of a fine-tuned open-weight LLM for automated essay scoring.
Chinese Translation
本研究考察了应用于自动化作文评分的LoRA适配的开放权重大型语言模型（Gemma-3-27B-it）的跨提示泛化能力和母语（L1）评分效应。我们使用在《AiAWE：一种使用LoRA适配的指令调优模型的开源LLM自动写作评估系统》（Gayed, 2026）中报告的相同模型和推理配置，该模型在来自两个提示的480篇论证性作文上进行了微调，评估了在完整TOEFL11语料库上的评分准确性：12,100篇由来自11种母语背景的考生撰写的作文，涵盖八个提示，且这些提示在训练过程中未曾见过。该模型的原始分数（0.5-5.0）被映射到ETS使用的相同三个熟练度等级（低、中、高），从而实现直接比较。模型的整体等级一致性达77.79%，二次加权kappa值为0.702，邻近等级一致性为99.98%。在所有八个未见提示中，准确性保持稳定，且与训练数据主题相关的提示并未显示出优势，表明模型具有强大的跨提示泛化能力。然而，该模型表现出系统性的、与L1相关的评分偏差。在每个熟练度等级中，来自欧洲语言背景的作文获得的分数始终高于来自东亚语言背景的作文，这一模式无法归因于微调数据的组成。这是首次对用于自动化作文评分的微调开放权重大型语言模型进行的大规模L1公平性分析。
cs.CL / 38 / 2607.14628
Routing Ceilings Are Domain-Independent: Structural Prior Injection in Code Security Vulnerability Detection
路由天花板是领域无关的：代码安全漏洞检测中的结构先验注入
Abstract
Large language models (LLMs) exhibit a well-documented gap between latent capability and consistent activation: the router hypothesis posits that models possess the knowledge to solve a task but lack reliable internal routing to activate it. Prior work in formal mathematical reasoning (SAIR, C\'azares 2026) reports that structural priors (cheatsheets) raise in-distribution performance dramatically, yet collapse below the zero-shot baseline out-of-distribution (OOD) -- and that iterative recalibration amplifies rather than corrects the collapse. We test whether this phenomenon is cross-domain by reproducing the SAIR design in source-code security vulnerability detection, evaluating three LLMs (GPT-OSS-120B, Llama-3.3-70B, Gemma-4-31B) across three vulnerability categories (CWE-798, CWE-284, and the non-CWE N+1 anti-pattern) spanning syntactic, contextual, and semantic complexity, then transferring cheatsheet-augmented prompts to real-world CVE data from VUDENC (CWE-89, CWE-22). Our findings replicate and extend SAIR: (F1) structural priors lift semantic-vulnerability recall from 20.0% to 100.0% across all models; (F2) zero-shot performance degrades along a semantic complexity gradient; (F3) the same cheatsheets that saturate synthetic performance amplify distribution-shift collapse on real CVE data (CWE-89: 100% synthetic F1 to 48.9% on VUDENC, -51.1pp); (F5) iterative recalibration produces a v2 cheatsheet that performs worse than v1 on real data, mirroring SAIR's AN45c-vs-AN38 finding. These results provide evidence that the cross-distribution trade-off surface documented in SAIR generalises to code security, and that the router hypothesis is cross-domain. We argue the structural nature of the collapse motivates distribution-aware training over prompt calibration. Code and evaluation scripts: https://github.com/bytepro-ai/bitcoder-v2-research
Chinese Translation
大型语言模型（LLMs）在潜在能力与一致激活之间存在显著差距：路由假设认为模型具备解决任务的知识，但缺乏可靠的内部路由来激活这些知识。先前的研究（SAIR, Cázarez 2026）报告称，结构先验（备忘单）显著提高了分布内性能，但在分布外（OOD）时表现低于零-shot基线，并且迭代再校准加剧而非修正了这种崩溃。我们通过在源代码安全漏洞检测中重现SAIR设计，测试这一现象是否跨领域，评估三种LLM（GPT-OSS-120B、Llama-3.3-70B、Gemma-4-31B）在三种漏洞类别（CWE-798、CWE-284以及非CWE的N+1反模式）上的表现，这些类别涵盖了语法、上下文和语义复杂性，然后将增强备忘单的提示转移到来自VUDENC的真实CVE数据（CWE-89、CWE-22）。我们的研究结果复制并扩展了SAIR的发现：（F1）结构先验将所有模型的语义漏洞召回率从20.0%提升至100.0%；（F2）零-shot性能沿语义复杂性梯度下降；（F3）同样的备忘单在合成性能饱和的情况下，在真实CVE数据上加剧了分布转移崩溃（CWE-89：合成F1从100%降至VUDENC的48.9%，下降51.1个百分点）；（F5）迭代再校准生成的v2备忘单在真实数据上的表现不如v1，反映了SAIR的AN45c与AN38的发现。这些结果提供了证据，表明SAIR中记录的跨分布权衡表面可以推广到代码安全领域，并且路由假设是跨领域的。我们认为崩溃的结构特性促使了对分布感知训练的需求，而非提示校准。代码和评估脚本： https://github.com/bytepro-ai/bitcoder-v2-research
cs.CL / 39 / 2607.14647
D-cut: Adaptive Verification Depth Pruning for Batched Speculative Decoding
D-cut：批量推测解码的自适应验证深度剪枝
Abstract
Speculative decoding accelerates large language model (LLM) inference without compromising output quality. Recent parallel drafting methods further improve single-request performance by decoupling draft length from drafting latency, enabling longer drafts and higher mean accepted tokens (MAT). However, under high request concurrency, long drafts waste substantial computation on rejected tokens, increasing verification cost and potentially making speculative decoding slower than autoregressive decoding. We present D-Cut, an adaptive pruning method that selects draft tokens jointly across the batch and concentrates the verification budget on tokens most likely to be accepted. D-Cut is motivated by two observations. First, acceptance lengths vary considerably across concurrent requests; D-Cut therefore performs cross-request pruning, allocating the verification budget adaptively according to draft confidence. Second, verification cost depends strongly on the deployment environment, including GPU architecture and parallelism strategy; D-Cut incorporates a runtime cost model to adapt its pruning depth to the target environment. Experiments on dense and mixture-of-experts (MoE) models show that, under high concurrency, D-Cut improves the average speedup from \(1.26\times\) to \(1.65\times\), restores acceleration in dense-model configurations where long-draft baselines are slower than autoregressive decoding, and achieves up to \(3.0\times\) speedup over autoregressive decoding on MoE models.
Chinese Translation
推测解码在不影响输出质量的情况下加速了大语言模型（LLM）的推理。最近的并行草拟方法通过将草拟长度与草拟延迟解耦，进一步提高了单请求性能，使得可以生成更长的草拟文本和更高的平均接受令牌数（MAT）。然而，在高请求并发的情况下，长草拟文本在被拒绝的令牌上浪费了大量计算，增加了验证成本，并可能使推测解码的速度低于自回归解码。我们提出了D-Cut，一种自适应剪枝方法，它在批量中联合选择草拟令牌，并将验证预算集中在最有可能被接受的令牌上。D-Cut的提出基于两个观察。首先，接受长度在并发请求之间差异显著；因此，D-Cut进行跨请求剪枝，根据草拟置信度自适应地分配验证预算。其次，验证成本在很大程度上依赖于部署环境，包括GPU架构和并行策略；D-Cut结合了运行时成本模型，以便根据目标环境调整其剪枝深度。在密集模型和专家混合模型（MoE）的实验中显示，在高并发情况下，D-Cut将平均加速从1.26倍提高到1.65倍，恢复了在密集模型配置中长草拟基线速度低于自回归解码的加速，并在MoE模型上实现了相较于自回归解码高达3.0倍的加速。
cs.CL / 40 / 2607.14707
Harnessing LLMs for Reliable Academic Supervision: A Comparative Study
利用大型语言模型进行可靠的学术监督：一项比较研究
Abstract
Large language models routinely produce fluent answers to single-shot prompts, yet deploying them as reliable components of a domain decision system is substantially harder. Closing this gap is the work of harness engineering: the deliberate composition of deterministic scaffolding (symbolic filters, retrieval, schema-typed I/O, LLM-as-judge loops, HITL gates, persistent state, audit trails) around an LLM core. We present a case study in academic supervision, a domain combining high-stakes recommendation, longitudinal accountability, and structured operational workflows. We compare a baseline (ASA), a GPT-5 chatbot with no scaffolding, against a multi-module system (ASuS) that wraps the much smaller GPT-4o-mini in a LangGraph harness with symbolic-semantic retrieval, schema-validated outputs, LLM-as-judge with bounded retry, HITL gates, deterministic weighted risk scoring with LLM narration, and a per-node SQLite audit trail. The evaluation rubric is retargeted at six harness-mechanism dimensions (grounding, explainability, consistency, process integrity, cognitive load, constraint adherence). A blind ten-rater hybrid evaluation, supplemented by a 2 x 2 model-harness ablation, finds that ASuS, despite using a much smaller base model, outscores ASA on every dimension. Across ten raters the pooled mean for ASuS is 4.08 versus 1.23 for ASA, and 8 of 10 raters reject the null at alpha = 0.05 on a paired Wilcoxon test; full numbers are in Sections 6.4 and 6.7. The ablation confirms that the structural contributions of the harness are largely model-invariant. We extract seven recurring harness-engineering patterns and argue that where reliability, traceability, and institutional consistency matter more than open-ended fluency, harness engineering challenges the prevailing 'bigger model is better' intuition.
Chinese Translation
大型语言模型通常能够流利地回答单次提示，但将其作为领域决策系统的可靠组件却要困难得多。弥合这一差距的工作被称为“工具工程”：在大型语言模型核心周围有意识地构建确定性的支架（符号过滤器、检索、模式化输入输出、LLM作为评判者的循环、人机交互（HITL）门、持久状态、审计轨迹）。我们展示了一个学术监督的案例研究，该领域结合了高风险推荐、长期问责和结构化操作工作流程。我们将一个基线（ASA），即没有支架的GPT-5聊天机器人，与一个多模块系统（ASuS）进行比较，该系统将更小的GPT-4o-mini封装在一个带有符号-语义检索、模式验证输出、带有限重试的LLM作为评判者、人机交互门、带有LLM叙述的确定性加权风险评分以及每个节点的SQLite审计轨迹的LangGraph支架中。评估标准针对六个支架机制维度（基础、可解释性、一致性、过程完整性、认知负荷、约束遵循）进行了调整。通过盲评的十位评审员的混合评估，辅以2 x 2模型-支架消融实验，发现尽管ASuS使用了一个更小的基础模型，但在每个维度上均超越了ASA。在十位评审员中，ASuS的汇总均值为4.08，而ASA为1.23，且10位评审员中有8位在配对Wilcoxon检验中以α = 0.05拒绝了原假设；完整数据见第6.4和6.7节。消融实验确认支架的结构贡献在很大程度上是模型不变的。我们提取了七种重复出现的工具工程模式，并认为在可靠性、可追溯性和机构一致性比开放式流利性更为重要的情况下，工具工程挑战了“更大模型更好”的普遍直觉。
cs.CL / 41 / 2607.14709
Gold-Guided Programmatic Distillation for Financial Reasoning over Hybrid Tables and Text
基于黄金指导的程序化蒸馏用于混合表格和文本的金融推理
Abstract
Financial question answering over hybrid tabular and textual data may require multi-source reasoning and precise numerical computation. While large language models (LLMs) can generate intermediate reasoning steps, natural-language rationales remain prone to arithmetic errors, making them an unreliable supervision source for distillation. Building on programmatic distillation, we develop an approach that transfers reliable numerical reasoning from a large teacher model to a compact student using execution-verified Python programs instead of free-form textual rationales. It leverages gold derivations to guide teacher-side program synthesis and retains only programs that execute correctly and produce the gold answer, ensuring high-quality supervision. We further introduce an iterative recovery stage that revisits teacher-failed examples, enabling the student to recover and incorporate newly verified programs into training. Experiments on TAT-QA show that our framework is highly effective for hybrid financial reasoning. Our best 7B student achieves 87.00 EM / 87.18 F1 on the test set, substantially outperforming the 72B teacher (78.46 EM) as well as traditional and strong LLM-based baselines, including TAGOP and TAT-LLM. These results demonstrate that execution-verified programmatic distillation provides an effective and extensible framework for training smaller models to perform reliable numerical reasoning.
Chinese Translation
在混合表格和文本数据上进行金融问答可能需要多源推理和精确的数值计算。尽管大型语言模型（LLMs）能够生成中间推理步骤，但自然语言推理仍然容易出现算术错误，使其成为不可靠的蒸馏监督源。在程序化蒸馏的基础上，我们提出了一种方法，通过执行验证的Python程序而非自由形式的文本推理，将可靠的数值推理从大型教师模型转移到紧凑的学生模型。该方法利用黄金推导指导教师端程序合成，并仅保留那些正确执行并产生黄金答案的程序，从而确保高质量的监督。我们进一步引入了一个迭代恢复阶段，重新审视教师失败的示例，使学生能够恢复并将新验证的程序纳入训练。TAT-QA上的实验表明，我们的框架在混合金融推理方面非常有效。我们最佳的7B学生模型在测试集上达到了87.00 EM / 87.18 F1，显著优于72B教师模型（78.46 EM）以及传统和强大的基于LLM的基线，包括TAGOP和TAT-LLM。这些结果表明，执行验证的程序化蒸馏为训练更小的模型以进行可靠的数值推理提供了一个有效且可扩展的框架。
cs.CL / 42 / 2607.14729
The Misclassification of Autistic Writing as AI-Generated
将自闭症写作误分类为人工智能生成的文本
Abstract
Recent findings suggest that detection models for artificial intelligence (AI) cannot accurately identify AI-generated text and may exhibit bias against certain minority groups. In the present study, anecdotal claims that autistic writers more often have their work flagged as AI-generated are examined empirically. A corpus of approximately 60,000 Reddit posts split into "likely-autistic" and "general-Reddit" subcorpora is used to compare the distribution of probabilities output by the OpenAI GPT-2 detection model. Differences in textual features between subcorpora are observed and compared to reported features of AI-generated text. Results showed that while less than two-percent of either subcorpus was flagged as AI-generated by the model, significantly more texts from the likely-autistic subcorpus were flagged. Connections between features of text with likely-autistic authors and AI-generated text were not straightforward. The widespread use of AI-detection models with a potential bias against autistic writers in their output prompts ethical scrutiny, and the authors recommend further critical examination of the models themselves as well as their use in academic contexts.
Chinese Translation
近期研究发现，人工智能（AI）检测模型无法准确识别AI生成的文本，并可能对某些少数群体表现出偏见。本研究实证检验了自闭症作家的作品更常被标记为AI生成的轶事性主张。研究使用了一个约60,000条Reddit帖子构成的语料库，分为“可能自闭症”和“普通Reddit”两个子语料库，以比较OpenAI GPT-2检测模型输出的概率分布。观察到两个子语料库之间的文本特征差异，并将其与报告的AI生成文本特征进行比较。结果显示，尽管模型将两个子语料库中不到2%的文本标记为AI生成，但来自可能自闭症子语料库的文本被标记的比例显著更高。与可能自闭症作者的文本特征和AI生成文本之间的联系并不简单。AI检测模型的广泛使用可能对自闭症作家产生偏见，引发伦理审查，作者建议对模型本身及其在学术环境中的使用进行进一步的批判性审查。
cs.CL / 43 / 2607.14735
CoTu at EXACT 2026: Neuro-Symbolic Reasoning for Transparent Educational QA
CoTu在EXACT 2026：透明教育问答的神经符号推理
Abstract
Transparent educational question answering asks for answers that are not only correct but explainable, and doing so with small models rules out the reasoning power of the largest proprietary systems. The EXACT 2026 competition poses this problem concretely: open-weight language models of at most 8B parameters, self-hosted, with a natural-language explanation for every answer. It pairs two tasks: logical reasoning over university regulations, and multi-step physics problem solving. We describe the system that team \cotu{} developed to address both, a neuro-symbolic Program-of-Thought pipeline in which a 4B backbone writes a program rather than stating an answer directly: for regulation queries it emits a Z3 encoding whose entailment verdict grounds the deduction, and for physics it emits numerical Python, both wrapped in a shared self-correction loop and a unified explained-JSON output. Answer-type routing, distillation-based task fine-tuning, and a latency-aware serving stack -- SGLang with speculative decoding -- keep the system within the 60-second per-query limit. The system achieved a \textbf{perfect score} on the physics task in both automated selection rounds and obtained the \textbf{highest final-round technical score} of any team -- $13.44/15$, combining automated answer evaluation with expert-judged reasoning depth -- with the equally weighted presentation score included, \cotu{} placed 3rd overall. Grounding answers in a symbolic solver yields correct, verifiable deductions at the 4B scale, and the residual difficulty lies in premise selection rather than the deduction itself.
Chinese Translation
透明教育问答要求提供不仅正确而且可解释的答案，而使用小模型则排除了最大专有系统的推理能力。EXACT 2026竞赛具体提出了这一问题：开放权重的语言模型参数最多为8B，自托管，并为每个答案提供自然语言解释。它结合了两个任务：对大学规章的逻辑推理和多步骤物理问题的解决。我们描述了团队 extit{cotu}开发的系统，该系统旨在同时解决这两个任务，采用神经符号的思维程序（Program-of-Thought）管道，其中一个4B的主干编写程序而不是直接给出答案：对于规章查询，它发出一个Z3编码，其蕴涵裁决为推理提供基础；对于物理问题，它发出数值Python，两者都被包裹在一个共享的自我修正循环和统一的解释JSON输出中。答案类型路由、基于蒸馏的任务微调，以及一个关注延迟的服务堆栈——SGLang与推测解码——使系统保持在每个查询60秒的限制内。该系统在物理任务的自动选择轮中取得了 extbf{满分}，并获得了任何团队的 extbf{最高决赛技术分数}——$13.44/15$，结合自动答案评估与专家评判的推理深度——在包括同等权重的展示分数后， extit{cotu}整体排名第三。将答案基于符号求解器进行基础推导，在4B规模下产生正确、可验证的推理，而剩余的困难在于前提选择而非推理本身。
cs.CL / 44 / 2607.14769
Dialogue Summarization with Emotion Dynamics Using Topic- and Participant-Centric Decomposition
基于主题和参与者中心分解的情感动态对话摘要
Abstract
Existing text summarization research has focused much on monologic information (e.g., newspaper articles, reports) without accounting for the interaction between speakers or authors. In contrast, dialogues are a rich communication channel where multiple participants conduct back and forth exchanges to construct meaning. We propose a dialogue summarization framework that explicitly models both semantic and emotion dynamics using multimodal dialogue inputs, built on an adapted hierarchical Chain-of-Agents approach. We decompose dialogues from two perspectives: (1) topic segments based on the utterances of all participants, and (2) participant-specific utterance segments. These are used to generate corresponding summaries while incorporating automatically inferred emotions. Topic- and participant-level summaries are aggregated into a dialogue summary capturing semantic content and emotion trajectories. To evaluate beyond content accuracy, we introduce emotion trajectory metrics measuring how well summaries preserve emotional flow. Experiments with small language models on multimodal dialogue datasets show that our framework produces summaries with both semantic and emotion content. Further experiments on explicit emotion label availability highlight the efficacy of our proposed methodology and the opportunities in dialogue analysis using language models.
Chinese Translation
现有的文本摘要研究主要集中在单一信息（例如，新闻文章、报告）上，而未考虑说话者或作者之间的互动。相比之下，对话是一种丰富的交流渠道，多个参与者通过反复交流来构建意义。我们提出了一种对话摘要框架，该框架使用多模态对话输入，明确建模语义和情感动态，基于改进的层次化代理链（Chain-of-Agents）方法。我们从两个角度对对话进行分解：（1）基于所有参与者发言的主题段，以及（2）参与者特定的发言段。这些分解结果用于生成相应的摘要，同时结合自动推断的情感。主题和参与者级别的摘要被汇总为一个对话摘要，以捕捉语义内容和情感轨迹。为了超越内容准确性的评估，我们引入了情感轨迹指标，衡量摘要在多大程度上保留了情感流。对多模态对话数据集上小型语言模型的实验表明，我们的框架能够生成同时具备语义和情感内容的摘要。进一步关于显式情感标签可用性的实验突显了我们提出的方法的有效性以及使用语言模型进行对话分析的机会。
cs.CL / 45 / 2607.14777
SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning
SEED：自我演化的在线蒸馏用于代理强化学习
Abstract
Large language models are increasingly trained as interactive agents for long-horizon tasks involving multi-turn interaction, tool use, and environment feedback. Outcome-based reinforcement learning (RL) provides a practical optimization paradigm, but its sparse trajectory-level rewards offer limited guidance on intermediate decisions, leaving a supervision gap between episode-level outcomes and token-level policy learning. We propose SEED (SElf-Evolving On-Policy Distillation), a self-evolving framework that converts completed on-policy trajectories into training-time hindsight skills and distills their behavioral effect back into the policy model. SEED first fine-tunes the policy to analyze completed trajectories and generate natural-language skills that capture reusable workflows, decisive observations, or failure-avoidance rules. During RL, the current policy both collects trajectories and serves as the analyzer that extracts hindsight skills from them. Policy updates therefore improve subsequent decision making and skill analysis together, allowing hindsight supervision to evolve with the policy. SEED then re-scores the sampled actions under ordinary and skill-augmented contexts, converting the skill-induced probability shift into a dense token-level on-policy distillation signal. This signal is jointly optimized with outcome-based RL, keeping the auxiliary supervision aligned with the current trajectory distribution. Extensive experiments on text-based and vision-based agentic tasks show that SEED consistently improves performance and sample efficiency, exhibiting robust generalization to unseen scenarios. Our code is available at https://github.com/jinyangwu/SEED.
Chinese Translation
大型语言模型越来越多地被训练为交互代理，以应对涉及多轮交互、工具使用和环境反馈的长期任务。基于结果的强化学习（RL）提供了一种实用的优化范式，但其稀疏的轨迹级奖励对中间决策的指导有限，导致在情节级结果与标记级策略学习之间存在监督差距。我们提出了SEED（自我演化的在线蒸馏），这是一个自我演化的框架，将完成的在线轨迹转化为训练时的事后技能，并将其行为效果蒸馏回策略模型。SEED首先微调策略，以分析完成的轨迹并生成自然语言技能，这些技能捕捉可重用的工作流程、决定性观察或避免失败的规则。在强化学习过程中，当前策略既收集轨迹，又作为分析器从中提取事后技能。因此，策略更新能够共同改善后续决策和技能分析，使得事后监督能够与策略一起演化。然后，SEED在普通和技能增强的上下文中重新评分采样的动作，将技能引起的概率变化转化为密集的标记级在线蒸馏信号。该信号与基于结果的强化学习共同优化，保持辅助监督与当前轨迹分布的一致性。在基于文本和基于视觉的代理任务上的大量实验表明，SEED始终提高了性能和样本效率，并在未见场景中表现出强大的泛化能力。我们的代码可在 https://github.com/jinyangwu/SEED 获取。
cs.CL / 46 / 2607.14905
Show Me How You Reason and I'll Tell You Who You Are: Reasoning Graphs for Robust LLM Authorship Attribution
告诉我你的推理方式，我便能识别你的身份：用于鲁棒大型语言模型作者归属的推理图
Abstract
Given the current trend to employ large language models (LLMs) in almost any imaginable context, LLM-generated text detection and authorship attribution have become a pressing issue. Prior work has primarily focused on surface-level linguistic features, an approach shown to be susceptible to paraphrasing and other obfuscation techniques. In this paper, we go beyond the linguistic surface, extracting and analysing reasoning structures in LLM-generated texts with the goal of capturing more complex signals of LLM authorship. We propose a graph neural network approach that leverages reasoning graphs extracted by an argument mining pipeline, demonstrating improved robustness and generalisation over a traditional Longformer baseline. Our approach outperforms the baseline by up to 27 percentage points under the obfuscation attacks such as paraphrasing and backtranslation, and 19 percentage points when evaluated on the texts generated by the unseen model versions, simulating real-world conditions in which new LLM versions are continuously released.
Chinese Translation
鉴于当前几乎在任何可想象的场景中都使用大型语言模型（LLMs）的趋势，LLM生成文本的检测与作者归属问题变得尤为紧迫。以往研究主要关注表层语言特征，但该方法易受到改写和其他混淆技术的影响。本文超越语言表层，提取并分析LLM生成文本中的推理结构，旨在捕捉更复杂的LLM作者信号。我们提出了一种基于图神经网络的方法，利用由论证挖掘流程提取的推理图，展示了相较传统Longformer基线模型更优的鲁棒性和泛化能力。在改写和回译等混淆攻击下，我们的方法较基线提升了最高27个百分点；在对未见模型版本生成文本的评估中提升了19个百分点，模拟了现实中不断发布新LLM版本的情形。
cs.CL / 47 / 2607.14967
Latent Trajectory Discrimination for AI-Generated Text Detection
用于人工智能生成文本检测的潜在轨迹区分
Abstract
Most existing approaches to AI-Generated Text Detection (AIGTD) treat documents as static objects and base their decisions on aggregate statistics or globally compressed embeddings. However, this perspective overlooks the inherently dynamic nature of autoregressive generation, where content evolves progressively through the latent space. In this paper, we reformulate AIGTD as the problem of distinguishing between latent generation trajectories. Instead of relying on static representations, we model how textual representations evolve across the sequence. To this end, we propose Geometric Trajectory and Contrastive Learning (GTCL), a framework that segments the document into ordered local units, encodes each unit in an embedding space, and constructs a structured and sequence-level representation. GTCL then applies contrastive learning to these trajectories to learn geometric regularities associated with the autoregressive generation. Evaluations performed on three different benchmarks and several approaches show that GTCL outperforms detection baselines consistently, which implies that explicitly modeling sequential dynamics provides robust discriminative signals across models and domains. These results suggest that modeling trajectory differences could improve detection and open up a dynamic direction that has been underexplored in previous AIGTD literature.
Chinese Translation
大多数现有的人工智能生成文本检测（AIGTD）方法将文档视为静态对象，并基于聚合统计或全局压缩嵌入做出决策。然而，这种视角忽视了自回归生成的动态特性，其中内容通过潜在空间逐步演变。在本文中，我们将AIGTD重新表述为区分潜在生成轨迹的问题。我们不再依赖静态表示，而是建模文本表示在序列中的演变。为此，我们提出了几何轨迹与对比学习（Geometric Trajectory and Contrastive Learning，GTCL）框架，该框架将文档分割为有序的局部单元，在嵌入空间中对每个单元进行编码，并构建结构化的序列级表示。GTCL随后对这些轨迹应用对比学习，以学习与自回归生成相关的几何规律。在三个不同基准和多种方法上进行的评估表明，GTCL始终优于检测基线，这意味着显式建模序列动态为模型和领域提供了强健的区分信号。这些结果表明，建模轨迹差异可能改善检测，并开启一个在以往AIGTD文献中未被充分探索的动态方向。
cs.CL / 48 / 2607.14989
OmniaBench: Benchmarking General AI Agents Across Diverse Scenarios
OmniaBench：在多样化场景中评估通用人工智能代理的基准测试
Abstract
Large language models are increasingly evolving from text generators into general agents capable of understanding user requests, invoking external tools, and completing complex tasks through interaction. However, existing agent benchmarks often focus on limited scenarios, tool ecosystems, or interaction formats, making it difficult to systematically characterize model capabilities across heterogeneous application settings. We introduce OmniaBench, a benchmark for evaluating general agents across diverse scenarios with explicit state spaces. We derive application-oriented scenario knowledge from app stores, product documents, industry resources, Web retrieval, and human refinement, forming a hierarchical taxonomy that spans ToC, ToB and ToE with 90 level-1 and 354 level-2 domains. Based on this taxonomy, we construct executable environments and synthesize single-turn and multi-turn tasks through four complementary routes: DAG, DAG-S, Solver, and Program. OmniaBench further introduces a ten-dimensional capability taxonomy and eight compositional atomic difficulty factors to support fine-grained evaluation and analysis. The resulting dataset contains 1,431 tasks, together with a challenging subset of 644 tasks designed to reduce evaluation cost and mitigate potential contamination of the full set after public release. The bench presents substantial challenges to current frontier models, with even Claude-Sonnet-5 and GPT-5.6-Sol achieving Overall Pass@1 scores of only 58.54 and 57.14, respectively. Further analyses reveal clear differences across domains and capabilities, as well as persistent limitations in planning, constraint maintenance, and adaptive correction. OmniaBench provides a broad and diagnostic benchmark for characterizing the capability boundaries of general agents.
Chinese Translation
大型语言模型正日益从文本生成器演变为能够理解用户请求、调用外部工具并通过交互完成复杂任务的通用代理。然而，现有的代理基准测试往往集中于有限的场景、工具生态系统或交互格式，这使得在异构应用环境中系统性地表征模型能力变得困难。我们提出了OmniaBench，这是一个用于在具有明确状态空间的多样化场景中评估通用代理的基准测试。我们从应用商店、产品文档、行业资源、网络检索和人工精炼中提取面向应用的场景知识，形成一个涵盖ToC（面向消费者）、ToB（面向企业）和ToE（面向教育）的层次分类法，包含90个一级领域和354个二级领域。基于该分类法，我们构建了可执行环境，并通过四条互补路径（DAG、DAG-S、Solver和Program）合成单轮和多轮任务。OmniaBench进一步引入了一个十维能力分类法和八个组合原子难度因素，以支持细粒度的评估和分析。最终数据集包含1,431个任务，以及一个由644个任务组成的挑战性子集，旨在降低评估成本并减轻公共发布后对完整数据集的潜在污染。该基准对当前前沿模型提出了重大挑战，即使是Claude-Sonnet-5和GPT-5.6-Sol的Overall Pass@1得分也仅为58.54和57.14。进一步分析揭示了不同领域和能力之间的明显差异，以及在规划、约束维护和自适应纠正方面的持续局限性。OmniaBench为表征通用代理的能力边界提供了广泛且具有诊断意义的基准测试。
cs.CL / 49 / 2607.15092
Rubrics on Trial: Evolving Rubrics from a Single Query via Synthetic Pairwise Evidence
评估标准的试验：通过合成成对证据从单一查询演变评估标准
Abstract
Rubrics provide structured, fine-grained signals for training and evaluating large language models (LLMs). Yet reliable query-specific rubrics are difficult to construct. Existing approaches often derive supervision from human-written rubrics, preference data, or sampled responses. Direct query-to-rubric generation avoids these resources, but provides no explicit check that a plausible rubric is useful. Such a rubric may fail to distinguish answer quality, reward an optional style, or penalize a valid alternative strategy. We introduce Rubrics on Trial, a query-only framework that evolves a rubric set from an empty set without external annotations or model training. It derives supervision solely from synthetic rubric-conditioned response pairs and validates each proposed rubric before adding it, screening out non-discriminative, over-specific, and style-only candidate rubrics. Experiments across five preference benchmark suites demonstrate the effectiveness of Rubrics on Trial, which achieves the best average accuracy and leads on six of seven evaluation sets.
Chinese Translation
评估标准为训练和评估大型语言模型（LLMs）提供了结构化的、细粒度的信号。然而，构建可靠的查询特定评估标准是困难的。现有的方法通常从人工编写的评估标准、偏好数据或抽样响应中获取监督。直接从查询生成评估标准避免了这些资源，但没有明确检查一个合理的评估标准是否有用。这种评估标准可能无法区分答案质量、奖励可选风格或惩罚有效的替代策略。我们引入了评估标准的试验（Rubrics on Trial），这是一个仅基于查询的框架，从一个空集演变出一组评估标准，而无需外部注释或模型训练。它仅从合成的基于评估标准的响应对中获取监督，并在添加每个提议的评估标准之前进行验证，筛选出非区分性、过于具体和仅风格的候选评估标准。在五个偏好基准套件上的实验表明，评估标准的试验的有效性，其实现了最佳的平均准确率，并在七个评估集中的六个上领先。
cs.CL / 50 / 2607.15095
Digital Pantheon: Simulating and Auditing Coalition Formation with LLM Agents
数字神殿：使用大型语言模型代理模拟和审计联盟形成
Abstract
The formation of political coalitions is a complex negotiation driven by both concrete policy objectives and deep-seated ideological convictions. While Large Language Models (LLMs) open new avenues for computational political science, the neutrality and helpfulness biases instilled by Reinforcement Learning from Human Feedback (RLHF) prevent them from sustaining steadfast partisan behaviour. We present a multi-agent framework that reconciles factual grounding with ideological alignment by combining Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Retrieval-Augmented Generation (RAG): DPO instils aggressive party-specific personas, while a per-party RAG pipeline keeps each agent bounded to its official manifesto. We operationalize the framework on the 2019 Flemish election, deploying the partisan agents in a hub-and-spoke negotiation arbitrated by a formateur. To make the emergent negotiation interpretable, we introduce a Multi-Layered Information Lineage Topology (MILT) that traces every clause in the final agreement back to its manifesto origin and classifies it into five provenance states, a Coalition Influence Score (CIS) that aggregates these traceable contributions to identify which party shaped the agreement, and a real-world grounding pass that benchmarks each simulated provision against the historically adopted coalition agreement. Across three independent simulations the framework yields a stable winner and ranking (N-VA ahead of CD\&V and Open Vld), and manifesto-anchored lineage reliably predicts real-world materialization whereas hallucinated content does not. The result is a transparent, scalable testbed for the ex-ante exploration of party compatibility and formateur-mediated compromise.
Chinese Translation
政治联盟的形成是一个复杂的谈判过程，既受到具体政策目标的驱动，也受到根深蒂固的意识形态信念的影响。尽管大型语言模型（LLMs）为计算政治科学开辟了新的途径，但由人类反馈强化学习（RLHF）所植入的中立性和有用性偏见使其无法维持坚定的党派行为。我们提出了一个多代理框架，通过结合监督微调（SFT）、直接偏好优化（DPO）和检索增强生成（RAG），调和事实基础与意识形态一致性：DPO灌输激进的特定党派个性，而每个党派的RAG管道则将每个代理限制在其官方宣言内。我们在2019年弗拉芒选举中对该框架进行了操作，将党派代理部署在由形成者仲裁的中心-辐射式谈判中。为了使新兴的谈判可解释，我们引入了多层信息来源拓扑（MILT），追踪最终协议中的每一条条款回溯到其宣言来源，并将其分类为五种来源状态，提出了一个联盟影响评分（CIS），该评分汇总这些可追溯的贡献，以识别哪个党派塑造了协议，以及一个现实世界基础的评估，基准测试每个模拟条款与历史上采用的联盟协议。通过三次独立的模拟，该框架产生了一个稳定的赢家和排名（N-VA领先于CD&V和Open Vld），并且基于宣言的来源可靠地预测了现实世界的实现，而虚构内容则不然。最终结果是一个透明、可扩展的测试平台，用于事前探索党派兼容性和形成者调解的妥协。
cs.CL / 51 / 2607.15146
Grokipedia vs Wikipedia: An LLM-Based Audit of Political Neutrality along Ideologies
Grokipedia与Wikipedia：基于大型语言模型的政治中立性审计研究
Abstract
Online encyclopedias shape political opinion and, through it, democratic discourse. In late 2025, Grokipedia was released, an encyclopedia written entirely by the LLM Grok. One motivation behind the project was to provide an unbiased alternative to Wikipedia, which has faced accusations of "left-wing" and "liberal" bias. But does an encyclopedia written by an LLM deliver greater neutrality, or does it simply embed a different ideology? We conduct a large-scale political bias study on Grokipedia and Wikipedia, analysing 1,394 article pairs describing members of government for neutrality along nine expert-coded ideology dimensions employing four LLM judges, Grok, Claude, Mistral, and DeepSeek. As the LLMs could themselves be biased, we also investigate patterns in their judgments. We find all LLM-judges, including Grok, to rate Grokipedia less neutral than Wikipedia. Both encyclopedias are rated as portraying politicians favourably overall, but towards different ideological groups. Grokipedia particularly favours economically right-wing politicians and penalises socially liberal ones, while Wikipedia is rated as favourably biased towards the latter.
Chinese Translation
在线百科全书塑造政治观点，并通过这些观点影响民主话语。2025年底，Grokipedia发布，这是一本完全由大型语言模型Grok撰写的百科全书。该项目的一个动机是提供一个无偏见的替代品，以应对Wikipedia面临的“左翼”和“自由派”偏见指控。但由大型语言模型撰写的百科全书是否能提供更大的中立性，还是仅仅嵌入了不同的意识形态？我们对Grokipedia和Wikipedia进行了大规模的政治偏见研究，分析了1,394对描述政府成员的文章，以评估其在九个专家编码的意识形态维度上的中立性，使用了四位大型语言模型评审者：Grok、Claude、Mistral和DeepSeek。由于大型语言模型本身可能存在偏见，我们还调查了它们判断中的模式。我们发现所有大型语言模型评审者，包括Grok，都认为Grokipedia的中立性低于Wikipedia。总体而言，两本百科全书都被评为对政治家持有积极态度，但针对的意识形态群体不同。Grokipedia特别偏向经济右派政治家，并对社会自由派政治家进行惩罚，而Wikipedia则被评为对后者持有积极偏见。
cs.CL / 52 / 2607.15175
Linear representations of grammaticality in neural language models
神经语言模型中的语法性线性表征
Abstract
Whether neural language models (NLMs) possess the ability to distinguish strings on the basis of their grammaticality remains a debated topic in the computational linguistics literature. Existing evidence has largely relied on probability-based measures, testing whether models assign higher probabilities to grammatical than ungrammatical strings. However, probability comparisons have been criticized as a measure for grammatical knowledge based on the assumption that grammaticality is inherently entangled with likelihood. Model-assigned probability is a function of many related sentence properties, such as lexical frequency, plausibility, and world knowledge. In this work, we move beyond probability-based evaluations and investigate whether grammaticality is encoded in the internal representations of NLMs. Using mass-mean probing, we test whether grammatical and ungrammatical sentences are systematically separated in representational space. We further examine the extent to which these representations are independent of sentence properties that are correlated with grammaticality, as well as their generalization across grammatical phenomena and languages. Our results provide evidence that grammaticality is robustly encoded in sentence representations of a wide range of pretrained NLMs, yielding clear representational separation on the dimension of grammaticality that cannot be fully explained by alternative sentence-level factors. Moreover, this encoding generalizes across a broad range of grammatical phenomena and to some degree, across languages, suggesting that grammaticality constitutes a coherent representational dimension in contemporary NLMs. These findings contribute new evidence to debates about the nature of syntactic knowledge in language models and offer a complementary framework for evaluating grammatical competence that is not dependent on string probabilities alone.
Chinese Translation
神经语言模型（NLMs）是否具备基于语法性区分字符串的能力在计算语言学文献中仍然是一个有争议的话题。现有证据主要依赖于基于概率的测量，测试模型是否对语法正确的字符串赋予比不语法的字符串更高的概率。然而，基于概率的比较因假设语法性与可能性本质上纠缠在一起而受到批评。模型赋予的概率是许多相关句子属性的函数，例如词汇频率、合理性和世界知识。在本研究中，我们超越了基于概率的评估，探讨语法性是否在NLMs的内部表征中编码。通过大规模均值探测，我们测试语法正确和不语法的句子在表征空间中是否系统性地分离。我们进一步考察这些表征在多大程度上独立于与语法性相关的句子属性，以及它们在语法现象和语言之间的泛化程度。我们的结果提供了证据，表明语法性在广泛的预训练NLMs的句子表征中被稳健地编码，且在语法性维度上产生了清晰的表征分离，这一分离无法完全用其他句子级因素解释。此外，这种编码在广泛的语法现象中以及在某种程度上跨语言泛化，表明语法性构成了当代NLMs中的一个一致的表征维度。这些发现为关于语言模型中句法知识本质的辩论提供了新的证据，并为评估语法能力提供了一个不依赖于字符串概率的补充框架。
cs.CL / 53 / 2607.15178
T^2MLR: Transformer with Temporal Middle-Layer Recurrence
T^2MLR：具有时间中层递归的变换器
Abstract
Transformer reasoning is limited by autoregressive decoding, which repeat edly compresses rich hidden computation through token space and makes it difficult for intermediate reasoning states to persist across time. We in troduce Transformers with Temporal Middle-Layer Recurrence (T2MLR), a transformers-based latent reasoning architecture that fuses a cached middle layer representation from the previous token directly into an earlier layer of the current token position, enabling abstract intermediate computation to persist across decoding steps with little inference overhead. Across natural-language pretraining and multi-hop reasoning finetuning, T2MLR consistently outperforms data- and parameter-matched Transformer base lines. Moreover, applying recurrence to only a localized middle-layer block (as little as 20% of the network) often outperforms full-layer recurrence. Im portantly, T2MLR does not require pretraining from scratch: retrofitting the recurrent pathway into an existing pretrained 1.7B Transformer and briefly finetuning substantially improves math reasoning, lowering the barrier to practical adoption. These results suggest that effective latent reasoning in Transformers does not require looping over all layers as in previous works, but can instead emerge more strongly from targeted middle-layer recurrence.
Chinese Translation
变换器推理受到自回归解码的限制，这种解码方式通过令牌空间反复压缩丰富的隐藏计算，使得中间推理状态难以在时间上持续。我们引入了具有时间中层递归的变换器（T2MLR），这是一种基于变换器的潜在推理架构，它将来自前一个令牌的缓存中层表示直接融合到当前令牌位置的早期层中，从而使抽象的中间计算能够在解码步骤中持续存在，且几乎没有推理开销。在自然语言预训练和多跳推理微调中，T2MLR始终优于数据和参数匹配的变换器基线。此外，仅将递归应用于局部中层块（网络的20%）往往优于全层递归。重要的是，T2MLR不需要从头开始预训练：将递归路径调整到现有的预训练1.7B变换器中并进行短暂的微调，显著提高了数学推理能力，降低了实际应用的门槛。这些结果表明，变换器中的有效潜在推理并不需要像以往工作那样遍历所有层，而是可以更强烈地从针对性的中层递归中涌现出来。
cs.CL / 54 / 2607.15200
Mask-Aware Policy Gradients for Diffusion Language Models
面向掩码的策略梯度用于扩散语言模型
Abstract
Reinforcement learning has proven effective for improving reasoning in large language models, but extending it to Masked Diffusion Language Models (MDLMs) remains challenging due to the intractability of the log-likelihood estimation. Existing approaches approximate this log-likelihood by modeling only the token predictions, ignoring the order in which positions are unmasked during generation. We observe that MDLM generation involves two decisions at each step: what tokens to place at each masked position and which positions to remask. We formalize this as a two-stage action MDP, showing that the policy gradient naturally decomposes into a token term and a masking term. Combining optimization of both terms leads to state-of-the-art outcomes on mathematical reasoning and coding benchmarks, with scores of 87.1% on GSM8K and 53.4% on MBPP.
Chinese Translation
强化学习已被证明在提升大型语言模型的推理能力方面有效，但将其扩展到掩码扩散语言模型（Masked Diffusion Language Models, MDLMs）仍然具有挑战性，因为对数似然估计的复杂性使其难以处理。现有的方法通过仅建模标记预测来近似这一对数似然，忽略了在生成过程中位置被去掩码的顺序。我们观察到，MDLM生成在每一步涉及两个决策：在每个掩码位置放置哪些标记以及哪些位置需要重新掩码。我们将其形式化为一个两阶段的动作马尔可夫决策过程（MDP），并展示策略梯度自然地分解为标记项和掩码项。优化这两个项的结合在数学推理和编码基准测试中取得了最先进的结果，在GSM8K上得分为87.1%，在MBPP上得分为53.4%。
cs.CL / 55 / 2607.15209
Expanding the Lexicon of Ge'ez Based African Languages: A Comparative Study of Amharic and Tigrinya
基于吉兹文的非洲语言词汇扩展：阿姆哈拉语与提格利尼亚语的比较研究
Abstract
Multilingual pre-trained language models (PLMs) exhibit degraded performance on low-resource, non-Latin-script languages, driven by high out-of-vocabulary (OOV) rates and excessive subword fragmentation that result from Latin-script-centric tokenizer training. We introduce VEXMLM, a vocabulary-extended variant of XLM-R targeting the two highest-resource Ge'ez-script languages, Amharic and Tigrinya, and further evaluated on 17 additional low-resource African languages (19 total). We train a language-specific SentencePiece tokenizer on curated Amharic and Tigrinya monolingual corpora, extend XLM-R's vocabulary with 30,000 Ge'ez-script subwords derived from this tokenizer, and initialize their embeddings by averaging the embeddings of their constituent subwords under XLM-R's original tokenizer. VEXMLM is trained in two stages: (1) continued masked language modeling over the extended vocabulary on the curated corpora, and (2) supervised fine-tuning on question answering (QA), named entity recognition (NER), and sentiment analysis (SA). On Amharic/Tigrinya QA, VEXMLM achieves 87.0 EM /90.0 F1, versus 66.0 EM/78.0 F1 for XLM-R and 74.0 EM/ 78.0 F1 for Glot500. On SA, VEXMLM reaches 80.0\% accuracy versus 77.0\% (XLM-R) and 46.0\% (Glot500). On NER, VEXMLM raises OOV-token entity accuracy from 81.4\% to 94.3\%, averaged over 11 of the 19 evaluated languages for which OOV analysis was possible. Our contributions are: (i) a vocabulary-extension and embedding-initialization procedure tailored to Ge'ez script; (ii) a two-stage training strategy under which vocabulary and continued-pretraining gains on Amharic/Tigrinya transfer to 17 typologically related, unaugmented African languages; and (iii) an evaluation spanning both intrinsic tokenization metrics (vocabulary coverage, fertility, OOV rate) and extrinsic task performance across all 19 languages.
Chinese Translation
多语言预训练语言模型（PLMs）在资源匮乏且非拉丁字母书写的语言上表现下降，主要原因是拉丁字母为中心的分词器训练导致的高词汇外（OOV）率和过度的子词碎片化。本文提出了VEXMLM，一种基于XLM-R的词汇扩展变体，针对资源最丰富的两种吉兹文书写语言——阿姆哈拉语和提格利尼亚语，并进一步在另外17种低资源非洲语言（共19种）上进行了评估。我们在精心整理的阿姆哈拉语和提格利尼亚语单语语料上训练了语言特定的SentencePiece分词器，利用该分词器生成的3万个吉兹文子词扩展了XLM-R的词汇表，并通过对XLM-R原始分词器中构成子词的嵌入取平均值来初始化这些新词汇的嵌入。VEXMLM的训练分为两个阶段：（1）在扩展词汇表上对整理语料进行持续的掩码语言模型训练；（2）在问答（QA）、命名实体识别（NER）和情感分析（SA）任务上进行有监督微调。在阿姆哈拉语/提格利尼亚语问答任务中，VEXMLM达到了87.0的EM和90.0的F1分数，而XLM-R分别为66.0 EM和78.0 F1，Glot500为74.0 EM和78.0 F1。在情感分析任务中，VEXMLM的准确率为80.0%，优于XLM-R的77.0%和Glot500的46.0%。在命名实体识别任务中，VEXMLM将OOV词元的实体识别准确率从81.4%提升至94.3%，该结果是对19种评估语言中11种可进行OOV分析语言的平均值。我们的贡献包括：（i）针对吉兹文的词汇扩展和嵌入初始化方法；（ii）一种两阶段训练策略，使阿姆哈拉语和提格利尼亚语上的词汇扩展和持续预训练收益能够迁移到17种类型学相关且未增强的非洲语言；（iii）涵盖内在分词指标（词汇覆盖率、繁殖率、OOV率）和外在任务性能的全面评估，覆盖所有19种语言。
cs.CL / 56 / 2607.15232
In-Place Tokenizer Expansion for Pre-trained LLMs
预训练大型语言模型的就地分词器扩展
Abstract
A tokenizer fixed at the start of pre-training allocates vocabulary in proportion to the pre-training corpus, reflecting the deployment priorities at that time. When those priorities shift, languages added later are split into many more tokens per word, which can raise latency, compute, and energy consumption for users of those languages. Cloud models can afford a broad vocabulary because the embedding and LM-head matrices are a small fraction of their parameters. On a compact model those matrices are a material share of per-token decode bandwidth, so on-device models ship small vocabularies and accept fragmentation outside a fixed language set. We present tokenizer expansion, an in-place recipe for upgrading a pre-trained model's tokenizer when the model producer controls its design. We continue the existing tokenizer's BPE merges on a multilingual corpus, so most source tokens carry over unchanged as single tokens and every new token has an exact decomposition into source tokens. We copy the carried-over embedding rows unchanged and initialize new rows as the mean of their source sub-token embeddings. A two-stage adaptation, embedding-only training then full-model continued pre-training, recovers source-checkpoint quality. We apply the recipe to a continued pre-trained checkpoint of LFM2-8B-A1B, an 8B-parameter Mixture-of-Experts model, to help produce LFM2.5-8B-A1B with a 128K tokenizer. The expanded tokenizer encodes Hindi and Vietnamese in roughly $2.4\times$ and $2.6\times$ fewer tokens than the source (up to $4.0\times$ on Thai). Combining these reductions with the measured per-token cost of the larger vocabulary, we estimate a $2.2$-$3.7\times$ per-character decode speedup for these languages across our reference devices. We release the model weights and the expanded tokenizer, and report the negative findings that shaped the recipe.
Chinese Translation
在预训练开始时固定的分词器根据预训练语料库的比例分配词汇，反映了当时的部署优先级。当这些优先级发生变化时，后期添加的语言会将每个单词拆分为更多的标记，这可能会增加这些语言用户的延迟、计算和能耗。云模型能够承受广泛的词汇量，因为嵌入矩阵和语言模型头矩阵仅占其参数的一小部分。在紧凑型模型中，这些矩阵占每个标记解码带宽的相当大份额，因此设备上的模型使用小词汇量，并在固定语言集之外接受碎片化。我们提出了分词器扩展，这是一种在模型生产者控制设计时升级预训练模型分词器的就地方法。我们在多语言语料库上继续现有分词器的BPE合并，因此大多数源标记作为单个标记不变地保留，而每个新标记都有确切的源标记分解。我们不改变保留的嵌入行，并将新行初始化为其源子标记嵌入的均值。通过两阶段适应，即仅嵌入训练然后全模型继续预训练，恢复源检查点质量。我们将该方法应用于LFM2-8B-A1B的继续预训练检查点，这是一个8B参数的混合专家模型，以帮助生成具有128K分词器的LFM2.5-8B-A1B。扩展后的分词器在编码印地语和越南语时，所需的标记数量分别比源减少了约$2.4 imes$和$2.6 imes$（泰语则高达$4.0 imes$）。结合这些减少与更大词汇量的每个标记成本的测量，我们估计这些语言在我们的参考设备上的每字符解码速度提升为$2.2$-$3.7 imes$。我们发布了模型权重和扩展的分词器，并报告了塑造该方法的负面发现。
cs.CL / 57 / 2607.15238
Language Identification via Compositional Data Analysis: A Linear-Time Classifier Based on Log-Ratio Geometry
通过组合数据分析进行语言识别：基于对数比几何的线性时间分类器
Abstract
Language identification is commonly addressed using either neural architectures or statistical n-gram models. Neural approaches typically require substantial computational resources, whereas classical frequency-based methods offer efficient linear-time performance, but rely on distance metrics that are not always appropriate for compositional data. This work models character and bigram frequency distributions as compositional vectors constrained to the simplex and mapped via the centered log-ratio (CLR) transformation bijectively onto the $(D-1)$-dimensional zero-sum subspace of $\mathbb{R}^D$, where Euclidean distances correspond to Aitchison distances. A pipeline is proposed, combining CLR-transformed unigram and bigram features with Laplace smoothing to address sparsity. The method is evaluated on six languages. Experimental results show that the proposed approach achieves robust accuracy across different text lengths, with strong performance for longer sequences. These findings indicate that compositional representations provide a deterministic and computationally efficient alternative for language identification, particularly in settings where interpretability and low resource consumption are essential.
Chinese Translation
语言识别通常采用神经架构或统计n-gram模型进行处理。神经方法通常需要大量计算资源，而经典的基于频率的方法则提供高效的线性时间性能，但依赖于并不总是适用于组合数据的距离度量。本研究将字符和二元组频率分布建模为约束在单纯形上的组合向量，并通过中心对数比（CLR）变换双射地映射到$ ext{R}^D$的$(D-1)$维零和子空间，其中欧几里得距离对应于Aitchison距离。提出了一种管道，将CLR变换的单元和二元特征与拉普拉斯平滑结合，以解决稀疏性问题。该方法在六种语言上进行了评估。实验结果表明，所提出的方法在不同文本长度下均能实现稳健的准确性，并且在较长序列上表现出色。这些发现表明，组合表示为语言识别提供了一种确定性和计算高效的替代方案，特别是在可解释性和低资源消耗至关重要的环境中。
cs.CL / 58 / 2607.15240
TikStance: A Multimodal and Hierarchical Dataset for Multi-target Stance Analysis in TikTok Political Conversations
TikStance：一个多模态和层次化的数据集，用于TikTok政治对话中的多目标立场分析
Abstract
Political discourse has increasingly moved to short-video platforms, yet computational analysis of such content remains constrained by the scarcity of datasets that jointly preserve audiovisual information and hierarchical conversations. Here we present TikStance, a multimodal and context-aware dataset comprising 161 videos and 13,876 comments from TikTok, designed for stance detection in political discussions. The dataset covers three major political figures in the 2024 U.S. election cycle--Donald Trump, Joe Biden, and Kamala Harris--with content collected between September 2023 and January 2025. Each discussion unit links a host video and its metadata to a parent-linked comment tree, enabling stance analysis within both audiovisual and conversational context. Each item was independently labeled by three annotators using a three-class scheme (Favor, Against, None) for video-to-target and comment-to-target stance; items with disagreement were re-annotated, and the final Krippendorff's \(\alpha\) reached 0.743, 0.723, and 0.722 for the Trump, Biden, and Harris subsets, respectively. Descriptive analysis further reveals target-dependent differences in stance distributions and conversational depth, with nested replies accounting for 23.3\% of all comments. By combining multi-target coverage, hierarchical conversations, and reliable multi-level human annotations, TikStance supports research in multimodal stance detection, political communication, computational social science, and context-aware natural language processing.
Chinese Translation
政治话语越来越多地转向短视频平台，但对这类内容的计算分析仍受限于缺乏同时保留视听信息和层次化对话的数据集。在此，我们呈现了TikStance，这是一个多模态和上下文感知的数据集，包含来自TikTok的161个视频和13,876条评论，旨在进行政治讨论中的立场检测。该数据集涵盖了2024年美国选举周期中的三位主要政治人物——唐纳德·特朗普（Donald Trump）、乔·拜登（Joe Biden）和卡马拉·哈里斯（Kamala Harris），内容收集时间为2023年9月至2025年1月。每个讨论单元将主视频及其元数据与父级链接的评论树相连接，使得在视听和对话上下文中进行立场分析成为可能。每个项目由三位标注者独立标注，采用三类方案（支持、反对、无立场）进行视频到目标和评论到目标的立场标注；对于存在分歧的项目进行了重新标注，最终的Krippendorff's B1分别达到了0.743、0.723和0.722，针对特朗普、拜登和哈里斯的子集。描述性分析进一步揭示了立场分布和对话深度的目标依赖性差异，嵌套回复占所有评论的23.3%。通过结合多目标覆盖、层次化对话和可靠的多层次人类标注，TikStance支持多模态立场检测、政治传播、计算社会科学和上下文感知自然语言处理的研究。
cs.CL / 59 / 2607.15241
Beyond the Leaderboard: Design Lessons for Trustworthy Multimodal VQA
超越排行榜：值得信赖的多模态视觉问答设计经验
Abstract
Healthcare multimodal AI must combine visual and textual evidence while remaining reliable and interpretable. Using MediaEval Medico 2025 as a retrospective GI endoscopy case study, we analyze design choices across nine documented systems for question answering and explanation quality. Parameter-efficient adaptation of pretrained backbones provides strong challenge performance, but answer-level gains do not consistently translate into faithful and complete clinical reasoning. Methods enforcing structured reasoning and explicit grounding show more reliable behavior across heterogeneous question types, although the evidence is correlational rather than ablation-based. These results motivate evaluation beyond lexical overlap, standardized evidence-linked explanations, leakage-aware data governance, and lightweight robustness and calibration checks. The findings support trustworthy multimodal healthcare AI based on data fusion, explainability, and resilient evaluation.
Chinese Translation
医疗保健多模态人工智能必须结合视觉和文本证据，同时保持可靠性和可解释性。以MediaEval Medico 2025作为回顾性胃肠内窥镜案例研究，我们分析了九个记录系统在问答和解释质量方面的设计选择。对预训练主干网络的参数高效适应提供了强劲的挑战表现，但答案级别的提升并未始终转化为忠实和完整的临床推理。强制结构化推理和明确基础的方式在异构问题类型中表现出更可靠的行为，尽管证据是相关性的而非基于消融的。这些结果促使我们在评估中超越词汇重叠、标准化证据关联解释、泄漏意识的数据治理，以及轻量级的鲁棒性和校准检查。研究结果支持基于数据融合、可解释性和韧性评估的值得信赖的多模态医疗保健人工智能。
cs.CL / 60 / 2607.15272
SciDiagramEdit: Learning to Edit Scientific Diagrams from Paper Revisions
SciDiagramEdit：从论文修订中学习编辑科学图表
Abstract
Editing the figures in a research paper is a routine and time-consuming part of everyday research practice: authors relabel components, rearrange panels, and restyle visuals as they revise their manuscripts. Automating this editing workflow under a natural-language instruction, however, is challenging, because a scientific figure is a dense infographic in which heterogeneous visual elements such as schematics, plots, photos, captions, and arrows are composed under a tight visual grammar to advance a specific argument. To address this, we present SciDiagramEdit, a benchmark and skill-evolution framework that learns from natural paper revisions and operates on the figure's editable vector source, where users can inspect and co-edit individual primitives alongside the agent. Our benchmark mines before/after figure pairs from arXiv version histories, each grounded in the authors' own revision intent. To accommodate the diversity of editing instructions, we adopt agentic learning via skill evolution: an agentic proposer continually refines the agent's skill specification from execution traces over multiple epochs. The resulting skill progressively lifts edit accuracy on a held-out validation set, providing evidence that natural paper revisions are an effective training signal for instruction-driven figure editing.
Chinese Translation
编辑研究论文中的图形是日常研究实践中一项常规且耗时的工作：作者在修订手稿时重新标记组件、重新排列面板和重新设计视觉效果。然而，在自然语言指令下自动化这一编辑工作流程是具有挑战性的，因为科学图形是一种密集的信息图，其中异质的视觉元素如示意图、图表、照片、标题和箭头在严格的视觉语法下组合，以支持特定论点。为了解决这个问题，我们提出了SciDiagramEdit，一个基准和技能演进框架，它从自然的论文修订中学习，并在图形的可编辑矢量源上操作，用户可以与智能体一起检查和共同编辑单个原件。我们的基准从arXiv版本历史中挖掘前后图形对，每一对都基于作者自己的修订意图。为了适应编辑指令的多样性，我们通过技能演进采用智能体学习：一个智能体提议者不断从多个周期的执行轨迹中细化智能体的技能规范。最终的技能逐步提高了在保留验证集上的编辑准确性，提供了证据表明自然的论文修订是指令驱动的图形编辑的有效训练信号。
cs.CL / 61 / 2607.15277
Partition, Prompt, Aggregate: Statistical Self-Consistency in Language Models
划分、提示、聚合：语言模型中的统计自一致性
Abstract
In-context learning is commonly interpreted as a form of conditional inference, in which the prompt specifies a context and the model's output is treated as an estimate of the corresponding conditional distribution. If this interpretation holds, then LLM estimates should satisfy basic probabilistic identities. In particular, the law of total probability asserts that prior-weighted conditional distributions aggregate into population-level marginals over any valid partition of the population. In this work, we investigate to what extent LLM estimates adhere to this self-consistency principle. We use binary trees as an evaluation scaffold to recursively partition a population into increasingly fine-grained subpopulations. We then prompt LLMs with verbalized subpopulation descriptions in context, aggregate the resulting estimates back into population-level estimates, and compare them across partitions of varying granularity. Applying this protocol across problem domains and state-of-the-art frontier models, we show widespread violations of basic consistency properties. An in-depth study of persona prompting reveals a pattern we call the macro fallacy: estimates reconstructed from more fine-grained subpopulation responses are often better aligned with human reference data than direct population-level estimates. This effect persists across variations in tree structure and estimation task, and can be partially recovered through implicit prompting. Together, these findings suggest that models possess relevant subpopulation knowledge but do not reliably propagate it into aggregate estimates. This gap establishes statistical self-consistency as an unsaturated, reference-free criterion for evaluating LLMs.
Chinese Translation
上下文学习通常被解释为一种条件推理形式，其中提示指定了一个上下文，模型的输出被视为相应条件分布的估计。如果这种解释成立，那么大型语言模型（LLM）的估计应该满足基本的概率恒等式。特别地，全概率法则断言，先验加权的条件分布在任何有效的人群划分中聚合成总体水平的边际分布。在本研究中，我们探讨了LLM的估计在多大程度上遵循这一自一致性原则。我们使用二叉树作为评估框架，递归地将人群划分为越来越细粒度的子人群。然后，我们在上下文中用口头化的子人群描述提示LLM，将得到的估计聚合回总体水平的估计，并在不同粒度的划分中进行比较。在多个问题领域和最先进的前沿模型中应用这一协议，我们发现基本一致性属性的广泛违反。对个性提示的深入研究揭示了一种我们称之为宏观谬误的模式：从更细粒度的子人群响应重建的估计通常与人类参考数据的对齐程度优于直接的总体水平估计。这一效应在树结构和估计任务的变化中持续存在，并且可以通过隐式提示部分恢复。综合来看，这些发现表明模型具备相关的子人群知识，但并未可靠地将其传播到聚合估计中。这一差距确立了统计自一致性作为评估LLM的未饱和、无参考标准。