arXiv Daily Digest

253

Papers

Kitchen Robotic Manipulation utilizing Foundation Models

利用基础模型的厨房机器人操作

Jeon, Myung-Hwan, Yamsani, Sankalp, Kim, Joohyung

Abstract

Deploying robots in everyday human environments requires perception systems that are both robust and adaptable to diverse, dynamic conditions. In this work, we present a modular perception pipeline for household manipulation tasks, with a focus on dishware handling in kitchen environments. The pipeline integrates open-vocabulary object detection, multi-view segmentation, instance-aware 3D reconstruction, and a 2D-3D feature fusion strategy for 6D pose estimation and grasp planning. Its modular design enables systematic substitution of multiple visual and geometric foundation models, allowing us to identify the best-performing configuration through extensive evaluation on a custom kitchen dataset. The best-performing configuration (LLMDet + SAMv2 + DINOv2 + GeoTransformer) achieves an ADI of 89.12\% on the 20-scene kitchen benchmark with cluttered and occluded conditions. Furthermore, real-world demonstrations confirm that the best configuration can be deployed on physical robots without environment-specific retraining, successfully executing tasks such as sink-to-dishwasher transfer and cup stacking. It validates the adaptability and scalability of the pipeline and highlights its potential as a practical framework for household robotic systems. Our code and supplementary materials are available at https://raivlab.github.io/FM_kitchen .

Chinese Translation

在日常人类环境中部署机器人需要具备强大且适应多样动态条件的感知系统。在本研究中，我们提出了一种用于家庭操作任务的模块化感知管道，重点关注厨房环境中的餐具处理。该管道集成了开放词汇的物体检测、多视角分割、实例感知的三维重建以及用于六维姿态估计和抓取规划的二维-三维特征融合策略。其模块化设计使得多种视觉和几何基础模型的系统替换成为可能，从而通过在自定义厨房数据集上的广泛评估识别出最佳性能配置。最佳性能配置（LLMDet + SAMv2 + DINOv2 + GeoTransformer）在具有杂乱和遮挡条件的20场景厨房基准测试中达到了89.12%的平均检测指数（ADI）。此外，实际演示确认最佳配置可以在物理机器人上部署，而无需针对特定环境进行重新训练，成功执行如水槽到洗碗机转移和杯子堆叠等任务。这验证了管道的适应性和可扩展性，并突显其作为家庭机器人系统的实用框架的潜力。我们的代码和补充材料可在 https://raivlab.github.io/FM_kitchen 获取。

View on arXiv Download PDF AI Translation

cs.RO / 2 / 2608.04121

Interpretable Fuzzy Inference for UAV Target Tracking Using Bounding-Box Geometry

基于边界框几何的可解释模糊推理用于无人机目标跟踪

Ahmari, Reza, Mohammadi, Ahmad, Hemmati, Vahid, Edmond, Nicholas, Saghazadeh, Hossein Z., Odeyomi, Olusola, Kebria, Parham, Homaifar, Abdollah

Abstract

Vision-based guidance of unmanned aerial vehicles (UAVs) toward unmanned ground vehicles (UGVs) supports cooperative aerial--ground robotics, but reliable continuous yaw estimation from onboard vision remains challenging because of sensing uncertainty, limited computation, and the need for interpretable control. Existing deep-learning and geometric-reconstruction approaches often require large datasets, external localization, or complex modeling assumptions, reducing transparency and deployment suitability on resource-constrained platforms. We present an interpretable fuzzy-inference framework that generates continuous yaw commands from low-dimensional features extracted from YOLO boxes: target centroid location, area, and aspect ratio. No explicit geometric modeling is required. A Mamdani fuzzy system serves as an interpretable baseline using a shoulder--triangle--shoulder input partition. It is followed by a first-order Takagi--Sugeno model with three antecedent membership terms per input, whose parameters are derived from training-set quantiles, yielding a compact 27-rule structure. Evaluation uses 6{,}169 labeled samples from a VICON motion-capture environment. Across five randomized train--test splits, the Takagi--Sugeno model achieves a test-set mean absolute error of $0.140^\circ \pm 0.003^\circ$, a root mean squared error of $0.200^\circ \pm 0.008^\circ$, and a maximum absolute error of $1.254^\circ \pm 0.121^\circ$. Within-threshold accuracies are $99.676% \pm 0.270%$ for $\pm1^\circ$ and $100.000% \pm 0.000%$ for both $\pm3^\circ$ and $\pm5^\circ$. Directional consistency between image-plane horizontal displacement and predicted yaw sign reaches $90.254% \pm 0.612%$. These results show that the framework is transparent, data-efficient, computationally lightweight, and suitable for real-time vision-based UAV guidance toward mobile ground targets.

Chinese Translation

基于视觉的无人机（UAV）引导地面无人车辆（UGV）支持协作空地机器人，但由于感知不确定性、计算能力有限以及对可解释控制的需求，从机载视觉中可靠地连续估计偏航角仍然具有挑战性。现有的深度学习和几何重建方法通常需要大量数据集、外部定位或复杂的建模假设，从而降低了透明度和在资源受限平台上的部署适用性。我们提出了一种可解释的模糊推理框架，该框架从YOLO框中的低维特征（目标质心位置、面积和长宽比）生成连续的偏航指令。该框架不需要显式的几何建模。Mamdani模糊系统作为可解释的基线，采用肩三角肩输入划分。接下来是一个一阶Takagi-Sugeno模型，每个输入有三个前提隶属项，其参数来自训练集分位数，形成一个紧凑的27规则结构。评估使用来自VICON运动捕捉环境的6,169个标记样本。在五个随机的训练-测试拆分中，Takagi-Sugeno模型在测试集上实现了$0.140^ ext{°} ext{±} 0.003^ ext{°}$的平均绝对误差，$0.200^ ext{°} ext{±} 0.008^ ext{°}$的均方根误差，以及$1.254^ ext{°} ext{±} 0.121^ ext{°}$的最大绝对误差。阈值内的准确率为$ ext{±}1^ ext{°}$时为$99.676 ext{%} ext{±} 0.270 ext{%}$，而对于$ ext{±}3^ ext{°}$和$ ext{±}5^ ext{°}$均为$100.000 ext{%} ext{±} 0.000 ext{%}$。图像平面水平位移与预测偏航符号之间的方向一致性达到$90.254 ext{%} ext{±} 0.612 ext{%}$。这些结果表明，该框架具有透明性、数据效率高、计算轻量，并适合于实时基于视觉的无人机引导移动地面目标。

View on arXiv Download PDF AI Translation

cs.RO / 3 / 2608.04196

SiMDex: Mining Similar Egocentric Videos for Cross-Embodiment Dexterous Manipulation

SiMDex：用于跨体现灵巧操作的相似自我中心视频挖掘

Lin, Nie, Ohkawa, Takehiko, Chen, Sijin, Wen, Ruoshi, Li, Zhuohang, Huang, Liqun, Zhu, Zhengming, Bao, Yiming, Li, Yunfei, Cai, Minjie, Ma, Xiao, Xu, Wei, Sato, Yoichi

Abstract

Recent years have witnessed an explosive trend of scaling ego-centric human videos for robot manipulation, yet it remains unclear which data actually benefits dexterous manipulation. We present SiMDex, a similarity-based data mining framework that casts human data selection for VLA post-training in dexterous manipulation as a recommendation problem. For each robot demonstration, SiMDex employs a three-layer recall-ranking-re-ranking pipeline to extract task-relevant subsets from a pool of ~32M egocentric human samples, operating in a morphology-agnostic action space that requires no changes to VLA architecture or training. Against a strong baseline trained with an equal amount of randomly sampled human data, SiMDex uses only ~1.49M mined samples (<5% of the pool) yet improves the overall success rate from 47.7% to 61.1%, showing that selective curation outperforms indiscriminate data mixing.

Chinese Translation

近年来，自我中心人类视频在机器人操作中的应用呈现出爆炸性增长，但仍不清楚哪些数据实际上有助于灵巧操作。我们提出了SiMDex，一个基于相似性的数据挖掘框架，将灵巧操作中人类数据选择视为一个推荐问题。对于每个机器人演示，SiMDex采用三层召回-排名-重新排名管道，从约3200万自我中心人类样本池中提取与任务相关的子集，操作在一个形态无关的动作空间中，无需对VLA架构或训练进行更改。与使用相同数量随机抽样人类数据训练的强基线相比，SiMDex仅使用约149万挖掘样本（不到池的5%），但将整体成功率从47.7%提高到61.1%，显示出选择性策展优于无差别的数据混合。

View on arXiv Download PDF AI Translation

cs.RO / 4 / 2608.04242

Feasibility of Embedded Photoplethysmography Sensing in Short-Duration Tactile Interactions With Pocket-Sized Robots Using IMU- and Confidence-Based Filtering

在与口袋大小机器人进行短时触觉交互中，基于IMU和置信度过滤的嵌入式光电容积脉搏波传感的可行性

Datta, Turjja, Frederiksen, Morten Roed

Abstract

Ubiquitous companion robots offer a promising avenue for immediate anxiety relief in children, yet their effectiveness relies on the ability to monitor physiological states continuously and unobtrusively. Current solutions often depend on external wearables, which impose usability barriers and limit the robot's autonomy. This paper investigates the integration of an embedded photoplethysmography (PPG) sensor directly into a pocket-sized companion robot, AffectaPocket, to enable self-contained heart rate monitoring during tactile interaction. We address the significant challenge of motion artifacts inherent in handheld usage by implementing a two-stage filtering pipeline that utilizes an onboard Inertial Measurement Unit (IMU) to reject high-variance segments and a confidence-based smoothing algorithm for recovery periods. We evaluated the system against a commonly used wrist worn sensor in a Within-Subjects Study with 26 participants. Our results demonstrate that the filtering strategy significantly reduced the Mean Absolute Percentage Error and achieved statistical equivalence to the ground truth measurements (p<0.05). Analysis of short-duration interactions shows that the sensor requires stability over longer periods to converge.

Chinese Translation

无处不在的伴侣机器人为儿童提供了即时缓解焦虑的有希望的途径，但其有效性依赖于持续且不干扰地监测生理状态的能力。目前的解决方案通常依赖于外部可穿戴设备，这会带来可用性障碍并限制机器人的自主性。本文探讨了将嵌入式光电容积脉搏波（PPG）传感器直接集成到口袋大小的伴侣机器人AffectaPocket中，以便在触觉交互期间实现自给自足的心率监测。我们通过实施一个两阶段过滤管道来解决手持使用中固有的运动伪影这一重大挑战，该管道利用板载惯性测量单元（IMU）来拒绝高方差段，并使用基于置信度的平滑算法进行恢复期处理。我们在一项包含26名参与者的被试内研究中，将该系统与一种常用的腕带传感器进行了评估。我们的结果表明，过滤策略显著降低了平均绝对百分比误差，并在统计上与真实测量值达到了等效（p<0.05）。对短时交互的分析表明，传感器需要在较长时间内保持稳定才能收敛。

View on arXiv Download PDF AI Translation

cs.RO / 5 / 2608.04246

SAFECAST: Robust Failure Detection for VLA Policies with Contrast-Set Training and Calibration

SAFECAST：基于对比集训练和校准的VLA策略鲁棒性故障检测

Rajaprakash, Harshitha, Prajapati, Aditeya, Xue, Rong, Anwar, Abrar, Thomason, Jesse

Abstract

Vision-language-action policies often fail under deployment-time distribution shifts such as clutter, distractor objects, lighting changes, novel objects, altered initial states, and reworded instructions. Hidden-state-based risk probes combined with functional conformal prediction can detect rollout failures, but their reliability depends on calibration data matching deployment conditions. We introduce SAFECAST, which leverages contrast set perturbations to improve hidden-state probe training and calibration for deployment time shift. SAFECAST statistically significantly improves failure detection ROC-AUC scores over a state of the art baseline in both real-world DROID and LIBERO simulation experiments across multiple VLM backbones. We further find that SAFECAST benefits most when both visual and language contrast set perturbations are used to augment data, and that with contrast set perturbations, sim-to-real calibration leads to better probes than using real rollout data only.

Chinese Translation

视觉-语言-行动（VLA）策略在部署时常常会因分布变化而失败，例如杂物、干扰物体、光照变化、新物体、改变的初始状态和重新表述的指令。基于隐状态的风险探测器结合功能性符合预测可以检测执行失败，但其可靠性依赖于与部署条件匹配的校准数据。我们提出了SAFECAST，它利用对比集扰动来改善隐状态探测器的训练和校准，以应对部署时的变化。SAFECAST在真实世界的DROID和LIBERO模拟实验中，相较于最先进的基线显著提高了故障检测的ROC-AUC分数，适用于多种VLM骨干网络。我们进一步发现，当同时使用视觉和语言的对比集扰动来增强数据时，SAFECAST的效果最佳，并且使用对比集扰动的情况下，模拟到真实的校准比仅使用真实执行数据能够得到更好的探测器。

View on arXiv Download PDF AI Translation

cs.RO / 6 / 2608.04309

Structured LLM Reasoning for Zero-Shot Human--Robot Coordination Under Hidden Goals

针对隐性目标的零-shot人机协调的结构化大语言模型推理

Mangalindan, Dong Hae, Gokhale, Anand, Bullo, Francesco, Srivastava, Vaibhav

Abstract

We present a structured large-language-model (LLM) architecture for zero-shot human--robot coordination in a cooperative construction task with private goal views. Guided by a Dec-POMDP formulation, the architecture decomposes decision-making into (i) action-conditioned Theory-of-Mind (ToM) inference, (ii) hierarchical planning, (iii) conversation interpretation, (iv) action verification, and (v) feedback-based replanning. We compare the proposed method with an ablation without ToM inference and a multi-agent reinforcement-learning policy trained offline over many goal pairs. In human-participant experiments, the proposed method required fewer interaction steps and yielded higher post-interaction trust ratings than both baselines. These results suggest that systematically decomposing the team decision problem, using LLMs as tractable surrogates for otherwise intractable inference and planning computations, and retaining conventional verification for physical feasibility can improve both task coordination and the human experience.

Chinese Translation

我们提出了一种结构化的大语言模型（LLM）架构，用于在具有私有目标视图的合作建造任务中实现零-shot人机协调。该架构在分布式部分可观测马尔可夫决策过程（Dec-POMDP）框架的指导下，将决策过程分解为（i）基于行动的心智理论（ToM）推理，（ii）层次规划，（iii）对话理解，（iv）行动验证，以及（v）基于反馈的重新规划。我们将所提出的方法与不使用ToM推理的消融实验以及在多个目标对上离线训练的多智能体强化学习策略进行了比较。在人类参与者的实验中，所提方法所需的交互步骤更少，且在交互后的信任评分上高于两个基线。这些结果表明，系统性地分解团队决策问题，利用LLM作为可处理的替代方案来进行原本难以处理的推理和规划计算，并保留传统的验证以确保物理可行性，可以改善任务协调和人类体验。

View on arXiv Download PDF AI Translation

cs.RO / 7 / 2608.04343

Design and Flight of an Ion-propelled Micro Hovercraft Leveraging Ground Proximity Effects

利用地面近效应设计与飞行的离子推进微型悬浮艇

Nelson, C. Luke, Nations, Grant, Modak, Mrinmoy, Drew, Daniel S.

Abstract

Electroaerodynamic propulsion is compelling for use in micro air vehicles due to its silent and solid-state nature, but its limited efficiency has thus far precluded a path towards power-autonomous flight. Recent work has shown that thrust density and efficiency for small-scale atmospheric ion thrusters can be vastly increased when operating close to a ground plane. Here, we explore the design space of centimeter-scale hovercraft, which can leverage this ground effect for low-altitude flight. We first perform an empirical investigation, characterizing the performance benefits and trade-offs for different geometries and configurations of passive hovercraft skirts, then use the results to fabricate a viable point design. We demonstrate a palm-sized hovercraft that, while tethered to an external power source, can fly for extended periods, withstand dozens of takeoff and landing cycles, passively stabilize to reject significant mechanical disturbances, and generate practically zero audible noise signature. The measured thrust efficiency of 16 mN/W and additional payload capacity of almost 1.5 grams above the vehicle's self mass of about 1.6 grams exceeds any similarly sized electroaerodynamically propelled robot by an order of magnitude. This is the first time an ion-propelled micro hovercraft has been shown in the open literature, and our work points the way towards an entirely new class of robot.

Chinese Translation

电气空气动力推进因其无声和固态特性而在微型飞行器中具有吸引力，但其有限的效率迄今为止阻碍了自主飞行的实现。近期研究表明，当小型大气离子推进器在接近地面时，其推力密度和效率可以大幅提高。在此，我们探讨了厘米级悬浮艇的设计空间，这种悬浮艇可以利用这一地面效应进行低空飞行。我们首先进行了一项实证研究，表征不同几何形状和被动悬浮艇裙的配置所带来的性能优势和权衡，然后利用结果制造出一个可行的设计方案。我们展示了一种掌心大小的悬浮艇，虽然连接到外部电源，但能够长时间飞行，承受数十次起降循环，主动稳定以抵御显著的机械干扰，并几乎产生零可听噪声。测得的推力效率为16 mN/W，额外的有效载荷能力几乎达到1.5克，超过了约1.6克自重的飞行器，超出了任何同类电气空气动力推进机器人一个数量级。这是首次在公开文献中展示离子推进微型悬浮艇，我们的工作指明了一种全新类别机器人的发展方向。

View on arXiv Download PDF AI Translation

cs.RO / 8 / 2608.04398

Approximate Multi-Objective Search Under Rulebooks

基于规则书的近似多目标搜索

Muhammetkulyyev, Omar, Salzman, Oren, Wongpiromsarn, Tichakorn

Abstract

Robotic planning often involves multiple objectives with complex priority relationships, such as safety, efficiency, and regulatory compliance. Rulebooks formalize these relationships, allowing partial ordering of objectives that generalizes both Pareto and lexicographic dominance. Computing the full set of rulebook-optimal solutions, however, is computationally expensive. To address this challenge, we introduce the concept of epsilon-rule-dominance, a principled notion of approximate dominance under rulebooks, and propose RA*pex, a best-first search algorithm that efficiently computes a compact set of epsilon-approximate rulebook-optimal solutions. RA*pex leverages dimensionality reduction, a technique used to speed up existing multi-objective search algorithms, while respecting rule hierarchies by maintaining separate closed sets and performing dominance checks over truncated and residual rule sets. We provide a formal analysis of RA*pex, proving that every rulebook-optimal solution is epsilon-rule-dominated (a generalization of approximate dominance we introduce) by at least one solution in the returned set. Empirical results demonstrate that our approach achieves computation times over two orders of magnitude faster than existing methods.

Chinese Translation

机器人规划通常涉及多个目标及其复杂的优先关系，例如安全性、效率和合规性。规则书形式化了这些关系，允许对目标进行部分排序，从而推广了帕累托（Pareto）和字典序（lexicographic）优势。然而，计算完整的规则书最优解集在计算上是昂贵的。为了解决这一挑战，我们引入了epsilon规则优势的概念，这是在规则书下的近似优势的原则性概念，并提出了RA*pex，一种最佳优先搜索算法，能够高效计算一组紧凑的epsilon近似规则书最优解。RA*pex利用了降维技术，这是一种用于加速现有多目标搜索算法的技术，同时通过维护独立的闭合集并对截断和残余规则集进行优势检查，尊重规则层次结构。我们对RA*pex进行了正式分析，证明了每个规则书最优解至少被返回集中的一个解epsilon规则优势（这是我们引入的近似优势的推广）。实证结果表明，我们的方法在计算时间上比现有方法快两个数量级以上。

View on arXiv Download PDF AI Translation

cs.RO / 9 / 2608.04420

SCOPE: Field-of-View-Aware Path Planning in Unknown 3D Environments via Safety-Volume Certification

SCOPE：通过安全体积认证在未知三维环境中进行视野感知路径规划

Yuan, Junbin, Cao, Muqing, Lee, Yunwoo, Moon, Brady, Scherer, Sebastian

Abstract

Safe navigation with a body-mounted limited-field-of-view sensor requires the complete robot-inflated volume of an intended motion to be observed and verified free before execution. We formulate this requirement as online safety-volume certification in an unknown voxel map and construct a certified graph whose vertices correspond exactly to positions with fully known-free safety volumes. Based on this representation, we propose SCOPE (Safety Certification through Observation Planning and Execution), a planning framework that decouples optimistic goal-directed guidance from certified execution. SCOPE converts the first uncertified point along an optimistic route into an explicit observation obligation, resolves it through target-centric viewpoint search, and recursively clears intermediate obligations when useful viewpoints are not yet certified-reachable. A certified preview mechanism and an observation-aware trajectory optimization backend enable smooth execution. We prove conditional complete planning: under ideal monotone sensing and exhaustive finite-domain graph search, SCOPE reaches the goal whenever a finite feasible sequence of certified sensing actions exists within its planning primitives. Across 60 randomized tasks in three unknown 3D environments, SCOPE reaches every goal while maintaining near-zero entry into non-certified inflated space. Preview reduces mean mission time by 27%, and real-robot demonstrations in two representative scenarios validate the complete system.

Chinese Translation

使用体载有限视野传感器进行安全导航需要在执行之前观察并验证预定运动的完整机器人膨胀体积是无障碍的。我们将这一要求表述为在未知体素地图中的在线安全体积认证，并构建一个认证图，其顶点与完全已知无障碍安全体积的位置完全对应。基于这一表示，我们提出了SCOPE（通过观察规划和执行进行安全认证），这是一个将乐观目标导向指导与认证执行解耦的规划框架。SCOPE将沿乐观路径的第一个未认证点转化为明确的观察义务，通过以目标为中心的视点搜索来解决，并在有用视点尚未认证可达时递归清除中间义务。认证预览机制和感知观察的轨迹优化后端使得执行过程平滑。我们证明了条件完全规划：在理想的单调感知和穷尽有限域图搜索下，只要在其规划原语中存在有限可行的认证感知动作序列，SCOPE就能达到目标。在三个未知三维环境中的60个随机任务中，SCOPE成功到达每一个目标，同时保持几乎零进入非认证膨胀空间。预览机制将平均任务时间减少了27%，并且在两个代表性场景中的真实机器人演示验证了整个系统的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 10 / 2608.04425

SSC: A Verifiable Structured Representation for Bimanual Manipulation Labelling

SSC：一种可验证的双手操作标注结构化表示

Lu, Yupu, Wu, Shuang, Chen, Sihan, Han, Ruihua, Zhang, Yichen, Kalander, Marcus, Pan, Jia

Abstract

Subtask labels decompose a long-horizon manipulation demonstration into shorter semantic segments for policy training and evaluation. Natural language descriptions are easy to read, but their linguistic variability makes automatic verification difficult. Rigid template formats, such as BEHAVIOR-1K's skill_annotation, are linguistically over-segmented, hindering both readability and annotation consistency. We propose the Structured Subtask Chain (SSC), a state-transition representation that bridges these extremes. A demonstration is a sequence of Structured Subtask Template (SST) entries. Each SST stores core action components (subject, predicate, object), flexible conditions (adverbial modifiers such as spatial or instrumental phrases), a base-motion field separate from arm actions, and an after-state scene graph. Built on this format, SSC supports three vision-language assisted functions: rendering SSTs as natural language, checking the assembled chain against four state-transition rules, and completing underspecified fields through a query resolution cascade. We instantiate the pipeline on BEHAVIOR-1K (50 tasks, 3 episodes per task, 2,357 annotated action cells) for logic verification and content completion, evaluating 13 selected state-of-the-art VL models as candidate verifiers and reporting labelling anomalies.

Chinese Translation

子任务标签将长时间范围的操作演示分解为更短的语义片段，以便进行策略训练和评估。自然语言描述易于阅读，但其语言变异性使得自动验证变得困难。刚性模板格式，如BEHAVIOR-1K的skill_annotation，存在语言过度分割的问题，妨碍了可读性和标注一致性。我们提出了结构化子任务链（Structured Subtask Chain, SSC），这是一种状态转移表示，弥合了这两者的极端。演示是结构化子任务模板（Structured Subtask Template, SST）条目的序列。每个SST存储核心动作组件（主语、谓语、宾语）、灵活的条件（如空间或工具短语的副词修饰语）、与手臂动作分开的基础运动字段，以及一个后状态场景图。在此格式的基础上，SSC支持三种视觉-语言辅助功能：将SST呈现为自然语言、检查组装的链条是否符合四个状态转移规则，以及通过查询解析级联完成未指定字段。我们在BEHAVIOR-1K（50个任务，每个任务3个回合，共2,357个标注动作单元）上实例化该管道，以进行逻辑验证和内容补全，评估13个选定的最先进的视觉-语言模型作为候选验证器，并报告标注异常。

View on arXiv Download PDF AI Translation

cs.RO / 11 / 2608.04510

GUARD: Grounding Uncertainty and Ablation-Based Risk Detection for Diffusion-Based VLAs

GUARD：基于扩散的视觉-语言-动作（VLA）的不确定性和消融风险检测

Hegde, Suhas, Katta, Jitendra Yasaswi Bharadwaj

Abstract

Diffusion-based vision-language-action (VLA) policies can generate plausible actions even when their predictions are weakly grounded in the visual and language evidence defining the task. We introduce GUARD, a test-time failure detection method that measures this grounding without modifying the pretrained policy. GUARD estimates the influence of token-indexed entries in the final vision-language model key-value (KV) cache, constructs counterfactual caches by ablating salient KV entries, and compares their denoising responses with the original conditioning. Based on the comparison, we derive GUARD diagnostic stream including sensitivity, attention entropy, modality bias, and grounding efficiency, which are calibrated online and processed by a lightweight temporal classifier. We evaluate GUARD under task-held-out splits across five policy-benchmark settings, using Pi0, SmolVLA, and Alpamayo-1.5 on LIBERO, SimplerEnv, MetaWorld, and PhysicalAI-AV. GUARD achieves the best ROC-AUC on four of five unseen-task settings and ranks second on the remaining setting, improving the average unseen-task ROC-AUC by 5.73 percentage points over the strongest competing runtime monitor while remaining within 0.19 points of the best seen-task average. These results show that directly probing action-head dependence on multimodal evidence provides a transferable failure signal across policies, tasks, embodiments, and domains.

Chinese Translation

基于扩散的视觉-语言-动作（VLA）策略能够在其预测与定义任务的视觉和语言证据弱相关时生成合理的动作。我们提出了GUARD，一种在测试时进行故障检测的方法，该方法在不修改预训练策略的情况下测量这种基础性。GUARD通过评估最终视觉-语言模型关键值（KV）缓存中基于标记索引的条目的影响，构建通过消融显著KV条目而形成的反事实缓存，并将其去噪响应与原始条件进行比较。基于比较，我们推导出GUARD诊断流，包括灵敏度、注意力熵、模态偏差和基础效率，这些指标在线校准并由轻量级时间分类器处理。我们在五个策略基准设置下，通过任务保留分割评估GUARD，使用Pi0、SmolVLA和Alpamayo-1.5在LIBERO、SimplerEnv、MetaWorld和PhysicalAI-AV上进行测试。GUARD在五个未见任务设置中有四个达到了最佳ROC-AUC，并在剩余设置中排名第二，平均未见任务ROC-AUC比最强竞争运行监控提高了5.73个百分点，同时与最佳已见任务平均值相差仅0.19点。这些结果表明，直接探测动作头对多模态证据的依赖提供了跨策略、任务、体现和领域的可转移故障信号。

View on arXiv Download PDF AI Translation

cs.RO / 12 / 2608.04527

Retrieve in Time, Correct in Frequency

及时检索，频率校正

Fan, Yuze, Cao, Yue, Gao, Pengjie, Gao, Haojia, Guo, Guangqiu, Zhang, Ziyue, Tan, Junbo, Chen, Bokui, Zou, Zhuo, Wang, Xueqian

Abstract

Frozen vision-language-action (VLA) policies generate temporally extended action chunks, but long-horizon manipulation remains vulnerable to accumulated execution error and visual aliasing across task stages. Successful rollouts provide useful corrective evidence, yet current frame retrieval can return progress-misaligned actions,while direct replay or time-domain fusion can overwrite the reactive structure of the policy proposal. We introduce Retrieve in Time, Correct in Frequency (RTCF), a training-free test-time correction framework that improves frozen VLA performance with low model-side overhead.RTCF separates which experience to retrieve from which part of its action to transfer. Progressive Memory Alignment (PMA) causally aligns the growing visual execution history with complete successful trajectories through incrementally updated monotonic frontiers, jointly identifying a relevant memory and the current aligned memory position without stage labels. From the aligned action chunk,RTCF transfers a coefficient-wise-clipped low-frequency residual on motion channels. Higher-frequency components and gripper decisions remain inherited from the frozen policy. Across four LIBERO suites and 2,000 episodes per condition, RTCF raises aggregate success from 86.4% to 88.4% and improves LIBERO-Long from 61.6% to 68.6%.These gains require no parameter updates, repeated VLA inference, or additional GPU resources: correction can be performed on the client CPU after a single policy invocation, and the median latencies sum to only 10.99 ms per action chunk

Chinese Translation

冻结的视觉-语言-动作（VLA）策略生成时间上延续的动作块，但长时间范围的操作仍然容易受到执行误差的累积和任务阶段之间的视觉混淆的影响。成功的回放提供了有用的纠正证据，但当前的帧检索可能返回与进度不对齐的动作，而直接重放或时域融合可能会覆盖策略提案的反应结构。我们提出了及时检索，频率校正（RTCF），这是一种无训练的测试时校正框架，通过低模型侧开销来提高冻结VLA的性能。RTCF将检索的经验与其动作的转移部分分开。渐进式记忆对齐（PMA）因果地将不断增长的视觉执行历史与完整的成功轨迹对齐，通过逐步更新的单调前沿，联合识别相关的记忆和当前对齐的记忆位置，而无需阶段标签。从对齐的动作块中，RTCF在运动通道上转移一个系数限制的低频残差。高频组件和抓取器决策仍然继承自冻结策略。在四个LIBERO套件和每个条件2,000个回合中，RTCF将整体成功率从86.4%提高到88.4%，并将LIBERO-Long的成功率从61.6%提高到68.6%。这些提升无需参数更新、重复的VLA推理或额外的GPU资源：校正可以在单次策略调用后在客户端CPU上执行，中位延迟仅为每个动作块10.99毫秒。

View on arXiv Download PDF AI Translation

cs.RO / 13 / 2608.04600

Static Timing Orchestration for Tree-Structured Robot Control Firmware

树状结构机器人控制固件的静态时序调度

Xi, Wang, Wei, Feiran, Deng, Mo, Lin, Weiheng, Fong, Pangkit, He, Jianping

Abstract

As robotic systems become increasingly complex, generating control firmware from structural description files has emerged as a promising paradigm for reducing development complexity and improving maintainability. Existing robot description formats naturally represent robotic systems as hierarchical tree structures, where devices are recursively composed into functional subsystems and eventually into the complete robot. However, such tree-structured organization also introduces structured data dependencies that affect perception-to-decision latency and, consequently, control performance. In this paper, we propose FineMote, a control firmware generation framework with a scheduling mechanism tailored for tree-structured device models. The framework objectifies heterogeneous low-level control logic and exposes unified scheduling units and execution entry points. Based on the resulting object hierarchy, the scheduling mechanism exploits compile-time information to statically determine execution order with minimal runtime overhead. We prove that the proposed mechanism satisfies deadline and precedence constraints, and further derive an upper bound on intra-tree decision latency. We implement the proposed framework and evaluate it on real robotic control platforms. The experimental results show improved timing behavior and runtime responsiveness, demonstrating the practical effectiveness of the proposed design.

Chinese Translation

随着机器人系统日益复杂，从结构描述文件生成控制固件已成为降低开发复杂性和提高可维护性的有前景的范式。现有的机器人描述格式自然地将机器人系统表示为层次树结构，其中设备递归地组成功能子系统，最终形成完整的机器人。然而，这种树状结构的组织也引入了结构化的数据依赖性，影响了感知到决策的延迟，从而影响控制性能。本文提出了FineMote，一个针对树状设备模型的控制固件生成框架，具有专门的调度机制。该框架将异构的低级控制逻辑对象化，并暴露统一的调度单元和执行入口点。基于生成的对象层次结构，调度机制利用编译时信息静态确定执行顺序，最大限度地减少运行时开销。我们证明了所提出的机制满足截止时间和优先级约束，并进一步推导出树内决策延迟的上限。我们实现了所提出的框架，并在真实的机器人控制平台上进行了评估。实验结果表明，时序行为和运行时响应性得到了改善，证明了所提设计的实际有效性。

View on arXiv Download PDF AI Translation

cs.RO / 14 / 2608.04612

GASP: GPU-Accelerated Safe Planner for Real-Time Collision-Aware Motion Generation with Latent Trajectory Sampling

GASP：用于实时碰撞感知运动生成的GPU加速安全规划器，结合潜在轨迹采样

Merk, Colin, Charalambous, Stefanos, Dürr, Peter, Khadivar, Farshad

Abstract

We present GASP, a GPU-Accelerated Safe Planner for real-time, collision-aware joint-space motion generation in known environments. GASP combines a clamped B-spline trajectory parameterization with a convolutional residual neural network that predicts the free interior control points, while analytically inserted boundary control points enforce initial and final derivative constraints for collision-aware planning under non-stationary conditions. A conditional variational autoencoder samples multiple trajectory candidates, which are decoded and validated in parallel on the GPU, yielding a batched planner for collision-aware coupled joint-space motion with near-millisecond inference. We validate GASP as an online motion-generation module, where it achieves analytical-level success rates with high collision-aware feasibility and substantially reduces inference time relative to GPU-based trajectory optimization. We further deploy GASP as a reinforcement-learning reset planner in competitive robotic table tennis, matching the baseline return rate while roughly halving training-time collisions.

Chinese Translation

我们提出了GASP，一种用于已知环境中实时碰撞感知关节空间运动生成的GPU加速安全规划器。GASP结合了夹紧B样条轨迹参数化和卷积残差神经网络，后者预测自由内部控制点，同时解析插入的边界控制点强制执行初始和最终导数约束，以实现非静态条件下的碰撞感知规划。条件变分自编码器采样多个轨迹候选，这些候选在GPU上并行解码和验证，从而生成一个批处理规划器，实现近毫秒级推理的碰撞感知耦合关节空间运动。我们验证了GASP作为在线运动生成模块的有效性，其成功率达到分析级别，具有高碰撞感知可行性，并显著减少相对于基于GPU的轨迹优化的推理时间。此外，我们将GASP部署为竞争性机器人乒乓球中的强化学习重置规划器，其回报率与基线匹配，同时将训练期间的碰撞数量减少了约一半。

View on arXiv Download PDF AI Translation

cs.RO / 15 / 2608.04633

Mind-VLA: Instruction-Aware Spatial Representation Alignment for Vision-Language-Action Models

Mind-VLA：面向指令的空间表示对齐方法用于视觉-语言-动作模型

Ding, Xingyu, Zhao, Yuzhong, Wu, Yang, Zhao, Chaoyang, Zhao, Chunhai, Zhang, Yifan, Cheng, Jian

Abstract

Recent Vision-Language-Action (VLA) methods improve generalization by aligning their representations with 3D scene geometry. However, these methods are fundamentally instruction-agnostic: the representations align the entire scene uniformly, neglecting the 3D geometry of the specific target object designated by the language instruction. This causes failures on fine-grained manipulation and target occlusion tasks, where success depends on accurate 3D understanding of the target object rather than the entire scene. To address this, we present Mind-VLA, an instruction-aware spatial representation alignment method for VLA models. Specifically, Mind-VLA first obtains the target object specified by the language instruction, then prepares its target-object tri-view and extracts the corresponding VAE and VGGT features. Finally, the latent representation of the VLA model is aligned with these features to enable instruction-aware 3D understanding. Mind-VLA reaches 93.9% on LIBERO and 4.47 on CALVIN with a compact 345M-parameter backbone. On real-robot tasks with target occlusion, Mind-VLA reaches 54% average success, outperforming the best-performing instruction-agnostic method in real-robot comparison by 32 percentage points. Code will be publicly available.

Chinese Translation

近期的视觉-语言-动作（VLA）方法通过将其表示与三维场景几何对齐来提高泛化能力。然而，这些方法在本质上是与指令无关的：这些表示均匀地对齐整个场景，忽视了语言指令指定的特定目标对象的三维几何。这导致在精细操作和目标遮挡任务中出现失败，因为成功依赖于对目标对象的准确三维理解，而不是整个场景。为了解决这个问题，我们提出了Mind-VLA，一种面向指令的空间表示对齐方法，适用于VLA模型。具体而言，Mind-VLA首先获取语言指令指定的目标对象，然后准备其目标对象的三视图，并提取相应的变分自编码器（VAE）和视觉几何图形转换（VGGT）特征。最后，VLA模型的潜在表示与这些特征对齐，以实现面向指令的三维理解。Mind-VLA在LIBERO上达到了93.9%的准确率，在CALVIN上达到了4.47，使用的是一个紧凑的345M参数主干网络。在具有目标遮挡的真实机器人任务中，Mind-VLA的平均成功率达到了54%，比在真实机器人比较中表现最佳的与指令无关的方法高出32个百分点。代码将公开发布。

View on arXiv Download PDF AI Translation

cs.RO / 16 / 2608.04692

Suppression Sticks, Locality Is Fragile: A Closed-Loop Target-and-Control Audit of Task-Vector Negation in VLA Policies

抑制性干预，局部性脆弱：对任务向量否定在多任务视觉-语言-行动（VLA）策略中的闭环目标与控制审计

Wang, Shaoguang, Guo, Weiyu, Dai, Rushi, Zhao, Yiren, Guo, Yandong, Xiong, Hui

Abstract

Task-vector arithmetic offers a closed-form way to modify a model, yet its behavioral locality remains unclear in closed-loop robot control. We present a target-and-control audit of per-skill task-vector subtraction from multitask vision-language-action (VLA) policies. Across all ten LIBERO-Goal skills, subtraction produces three qualitatively different regimes: target-control separation for five skills, resistance for three, and global collapse for two. On held-out initial states, the five suppressible targets remain at 0% success; however, mean baseline-normalized control retention is only 52%, and each target-suppressing edit materially harms at least one nominally unrelated control. Additional Goal panels show separation across tested policies with continuous-regression, discrete-token, and flow-matching action heads, whereas we observe no clean separation on Spatial and control collapse on the tested Object and Long-horizon panels. Mean task-vector cosine does not account for this variation. A matched-norm control identifies a local sign asymmetry around one Goal anchor, while multi-vector outcomes vary with anchor and scale. Retain-aware gradient baselines provide data-dependent comparators but require removal-time data and optimization; subtraction is data- and gradient-free only at edit time, assuming precomputed expert deltas. Finally, a single-skill relearning probe is consistent with behavioral masking, not certified unlearning. These results characterize task-vector subtraction as a fast but brittle intervention and underscore the need for closed-loop target-and-control evaluation when assessing locality in embodied model editing.

Chinese Translation

任务向量算术提供了一种闭合形式来修改模型，但其在闭环机器人控制中的行为局部性仍不明确。我们对多任务视觉-语言-行动（VLA）策略中每项技能的任务向量减法进行了目标与控制审计。在所有十个LIBERO-Goal技能中，减法产生了三种质的不同状态：五项技能的目标-控制分离、三项技能的抵抗以及两项技能的全局崩溃。在保留的初始状态下，五个可抑制目标的成功率保持在0%；然而，平均基线归一化控制保持率仅为52%，而且每个目标抑制编辑在实质上对至少一个名义上无关的控制造成了损害。额外的Goal面板显示了在测试策略中，连续回归、离散标记和流匹配动作头之间的分离，而我们在测试的Object和Long-horizon面板上观察到没有清晰的分离和控制崩溃。平均任务向量余弦并不能解释这种变化。匹配范数控制识别出一个Goal锚点周围的局部符号不对称，而多向量结果则随锚点和尺度变化。保留感知梯度基线提供了数据依赖的比较器，但需要移除时间数据和优化；减法在编辑时仅在数据和梯度自由的情况下进行，假设预计算的专家增量。最后，单技能再学习探针与行为掩蔽一致，而非经过认证的遗忘。这些结果将任务向量减法表征为一种快速但脆弱的干预，并强调在评估具身模型编辑中的局部性时需要进行闭环目标与控制评估。

View on arXiv Download PDF AI Translation

cs.RO / 17 / 2608.04721

Enabling Urgency-aware Robot Swarm Intralogistics using Smart IoT Tags

利用智能物联网标签实现紧急感知的机器人群体内部物流

Alboraei, Youssef, Groves, Murray, Wen, Shane, Zhao, Wenda, Qiu, Senhui, Bocus, Mohammud J., Piechocki, Robert, Hauert, Sabine, Eder, Kerstin

Abstract

Warehouse items differ in how urgently they must be moved: perishable goods, pharmaceutical shipments, and just-in-time production materials must be delivered sooner than the rest of the stock. Decentralised robot swarms suit warehouses that cannot justify fixed automation infrastructure, but current swarm controllers treat all items alike or rely on an external scheduler to set priorities, so urgent items wait as long as ordinary ones. This paper presents a swarm logistics system in which each warehouse carrier holds an ultra-low-power Internet-of-Things (IoT) tag that broadcasts the urgency of its item over Bluetooth Low Energy (BLE). Robots read these broadcasts directly and weigh urgency against travel distance when choosing which carrier to serve, so prioritisation happens at the item level without central scheduling. The system is evaluated in simulation and validated on real robots and IoT-tagged carriers against a proximity-only baseline. In the physical trials, priority alignment (i.e. proportion of urgent items served first), improved from 0.41 to 0.64, with a nonsignificant trend toward lower 95th-percentile (P95) delivery latency and throughput within 1.2% of the baseline. In simulation, the benefit grew with system size: across three larger configurations, P95 latency fell by 5.2% to 11.8% and priority alignment improved by 41.7% to 51.6%. Attaching urgency to the items themselves therefore allows a decentralised swarm to serve time-critical stock sooner while keeping the low infrastructure requirements that make swarm systems attractive for warehouse automation.

Chinese Translation

仓库物品在搬运的紧急程度上存在差异：易腐烂商品、药品运输和准时生产材料必须比其他库存更早交付。去中心化的机器人群体适合那些无法承担固定自动化基础设施的仓库，但当前的群体控制器对所有物品一视同仁，或依赖外部调度器设定优先级，因此紧急物品的等待时间与普通物品相同。本文提出了一种群体物流系统，其中每个仓库运输工具都配备了超低功耗的物联网（IoT）标签，通过蓝牙低能耗（BLE）广播其物品的紧急程度。机器人直接读取这些广播，并在选择服务哪个运输工具时权衡紧急程度与行驶距离，因此优先级在物品层面上得以实现，而无需中央调度。该系统在模拟中进行了评估，并在真实机器人和带有物联网标签的运输工具上进行了验证，与仅基于接近度的基线进行比较。在实际试验中，优先级对齐（即优先服务紧急物品的比例）从0.41提高到0.64，且在95百分位（P95）交付延迟和吞吐量方面呈现出不显著的下降趋势，保持在基线的1.2%以内。在模拟中，随着系统规模的扩大，收益也随之增加：在三个较大配置中，P95延迟下降了5.2%至11.8%，优先级对齐改善了41.7%至51.6%。因此，将紧急程度附加到物品本身，使得去中心化的群体能够更早地服务于时间敏感的库存，同时保持了使群体系统在仓库自动化中具有吸引力的低基础设施要求。

View on arXiv Download PDF AI Translation

cs.RO / 18 / 2608.04723

A Vision-based Control Framework for Real-time Autonomous UUV Operations

基于视觉的实时自主水下无人艇操作控制框架

Frøland, Erik Tjærand, Job, Marco, Deowan, Md Ether, Kelasidi, Eleni

Abstract

This paper presents a fully integrated vision-based framework for real-time and robust localization, autonomous navigation, and mapping for unmanned underwater vehicles (UUVs) in dynamic, visually challenging environments. The proposed pipeline enables both net-relative and global localization while generating continuous 3D maps of the surroundings in real-time. The framework was validated on synthetic datasets with ground truth and tested onboard an UUV during autonomous net-relative navigation experiments. Results demonstrate real-time performance and enhanced robustness, supporting vision-driven autonomous navigation and enabling the field deployment of marine robots for critical inspection and mapping tasks in complex underwater environments.

Chinese Translation

本文提出了一种完全集成的基于视觉的框架，用于在动态且视觉挑战性强的环境中实现水下无人艇（UUV）的实时和鲁棒定位、自主导航和地图构建。所提议的流程能够实现相对网和全球定位，同时实时生成周围环境的连续三维地图。该框架在具有真实标注的合成数据集上进行了验证，并在自主相对网导航实验中进行了水下无人艇上的测试。结果表明该框架具有实时性能和增强的鲁棒性，支持基于视觉的自主导航，并使海洋机器人在复杂水下环境中进行关键检查和地图构建任务的现场部署成为可能。

View on arXiv Download PDF AI Translation

cs.RO / 19 / 2608.04724

A GitOps-Driven Annotation Catalog for Fully Automatic Railway Operations

基于GitOps的全自动铁路运营注释目录

Köppel, Martin, Cronauer, Tobias, Ilknur-Öz, Zekiye, Dubiel, Sebastian, Naumann, Patrick, Neumaier, Philipp

Abstract

Automatic train operation (ATO) at grade of automation 3 and above (GoA3-GoA4) requires robust AI-based perception systems capable of reliably detecting obstacles and railway-specific objects under real-world conditions. The effectiveness of these modern artificial intelligence approaches depends heavily on large-scale, high-quality, and highly dynamic annotated datasets. However, managing metadata, maintaining provenance, and tracking the iterative evolution of these annotations impose significant infrastructural and regulatory requirements. Existing monolithic data catalogs often suffer from massive operational overhead, poor integration into developer workflows, and severe documentation drift. This paper introduces an innovative, lightweight GitOps-based architecture for metadata management. By leveraging Data-as-Code principles, Continuous Integration/Continuous Deployment (CI/CD) pipelines, and Static Site Generation (SSG), the proposed approach establishes a seamless, developer-centric workflow. This ensures an traceability, enforces strict regulatory compliance, and automatically generates a highly performant dataset overview.

Chinese Translation

自动列车操作（ATO）在自动化等级3及以上（GoA3-GoA4）需要强大的基于人工智能的感知系统，能够在真实世界条件下可靠地检测障碍物和铁路特定物体。这些现代人工智能方法的有效性在很大程度上依赖于大规模、高质量和高度动态的注释数据集。然而，管理元数据、维护来源以及跟踪这些注释的迭代演变带来了显著的基础设施和监管要求。现有的单体数据目录往往面临巨大的操作开销、与开发者工作流程的整合不良以及严重的文档漂移。本文提出了一种创新的轻量级基于GitOps的元数据管理架构。通过利用数据即代码（Data-as-Code）原则、持续集成/持续部署（CI/CD）管道和静态网站生成（SSG），所提方法建立了一个无缝的、以开发者为中心的工作流程。这确保了可追溯性，强制执行严格的监管合规，并自动生成高性能的数据集概览。

View on arXiv Download PDF AI Translation

cs.RO / 20 / 2608.04765

Explicit Language Memory for Long-Horizon Planning in Vision-Language-Action Models

用于长时间规划的显式语言记忆在视觉-语言-动作模型中的应用

Xu, Houze, Li, Jizhong, Ye, Ziyi

Abstract

Vision-language-action (VLA) models provide a unified paradigm for connecting visual perception, language understanding, and robotic control. However, existing VLA models still face major challenges in long-horizon tasks: sparse expert demonstrations constrain cross-task compositional generalization; the non-Markovian nature of long-horizon tasks makes it difficult for policies conditioned only on current observations to maintain temporal consistency; limited closed-loop error correction allows execution errors to accumulate; and end-to-end action fine-tuning may weaken the high-level semantic representations of vision-language model (VLM) backbones. To address these issues, we propose a hierarchical long-horizon VLA architecture with an explicit language-memory module. The central idea is to convert discrete temporal observations into a coherent textual memory sequence with temporal logic. The system is decoupled into a high-level VLM and a low-level VLA: the high-level VLM performs semantic reasoning through a visual question answering training paradigm, while the low-level VLA executes precise continuous control conditioned on subtask instructions and visual observations. The high-level VLM recursively updates both language memory and subtask instructions using the previous memory as a contextual anchor, enabling persistent temporal tracking and dynamic correction during long-horizon execution. We evaluate the proposed method in multiple simulation environments and conduct sim-to-real experiments on a real robotic platform. The results demonstrate that explicit language memory improves the success rate and robustness of VLA models on complex long-horizon tasks while providing an interpretable semantic account of the decision process.

Chinese Translation

视觉-语言-动作（VLA）模型提供了一种统一的范式，用于连接视觉感知、语言理解和机器人控制。然而，现有的VLA模型在长时间任务中仍面临重大挑战：稀疏的专家演示限制了跨任务的组合泛化；长时间任务的非马尔可夫特性使得仅基于当前观察的策略难以维持时间一致性；有限的闭环误差修正允许执行错误累积；而端到端的动作微调可能削弱视觉-语言模型（VLM）骨干的高层语义表示。为了解决这些问题，我们提出了一种具有显式语言记忆模块的分层长时间VLA架构。其核心思想是将离散的时间观察转换为具有时间逻辑的连贯文本记忆序列。该系统被解耦为高层VLM和低层VLA：高层VLM通过视觉问答训练范式进行语义推理，而低层VLA则根据子任务指令和视觉观察执行精确的连续控制。高层VLM利用先前的记忆作为上下文锚点，递归更新语言记忆和子任务指令，从而在长时间执行过程中实现持续的时间跟踪和动态修正。我们在多个仿真环境中评估了所提出的方法，并在真实机器人平台上进行了仿真到现实的实验。结果表明，显式语言记忆提高了VLA模型在复杂长时间任务上的成功率和鲁棒性，同时提供了可解释的决策过程语义说明。

View on arXiv Download PDF AI Translation

cs.RO / 21 / 2608.04769

From Transparent Labware Segmentation to Collision Avoidance: A Real-Time Edge-Aware Perception Pipeline

从透明实验室器皿分割到碰撞避免：一种实时边缘感知感知管道

Ding, Shijun, Qian, Chen, Shang, Weiwei, Xiong, Junlin

Abstract

This paper presents an edge-aware instance segmentation framework that enables real-time robotic collision avoidance with transparent laboratory glassware using purely visual perception. Transparent vessels defy conventional segmentation due to refraction, specular reflection, and the absence of stable interior texture, yet their boundary contours remain comparatively reliable visual cues. Exploiting this observation, we augment a one-stage real-time instance segmentation backbone with a lightweight edge-detection branch, edge-guided attention fusion, and a parameter-free SimAM module, and further construct LabGlass-IS, a 3485-image, 21-category instance segmentation dataset of real laboratory glassware. The enhanced model achieves the highest Boundary F-score of 97.80 among compared methods, outperforming the YOLO-prompted FastSAM framework by 18.93 BF points. Furthermore, it maintains an inference speed of 7.1ms per frame and requires only 2.85% of the parameters of the closest accuracy competitor. Multi-view triangulation of mask centroids further provides 3D positions for conservative bounding-volume collision constraints. Real-robot trials achieve a 93.3% collision avoidance success rate, indicating the feasibility of the proposed perception-to-action pipeline for robot collision avoidance among fragile transparent objects. Our code is available at https://github.com/havishamy/TransYOLO_3D. Our video is available at https://havishamy.github.io/paper-videos/.

Chinese Translation

本文提出了一种边缘感知实例分割框架，使得机器人能够通过纯视觉感知实现与透明实验室玻璃器皿的实时碰撞避免。由于折射、镜面反射以及缺乏稳定的内部纹理，透明容器挑战了传统的分割方法，但其边界轮廓仍然是相对可靠的视觉线索。基于这一观察，我们在单阶段实时实例分割主干网络中增强了一个轻量级的边缘检测分支、边缘引导的注意力融合以及一个无参数的SimAM模块，并进一步构建了LabGlass-IS，这是一个包含3485张图像和21个类别的真实实验室玻璃器皿实例分割数据集。增强后的模型在比较方法中实现了最高的边界F分数97.80，比YOLO引导的FastSAM框架高出18.93 BF点。此外，它保持了每帧7.1毫秒的推理速度，并且仅需最近准确度竞争者2.85%的参数。对掩膜质心的多视角三角测量进一步提供了保守的边界体积碰撞约束的3D位置。真实机器人试验实现了93.3%的碰撞避免成功率，表明所提感知到行动的管道在脆弱透明物体之间的机器人碰撞避免中的可行性。我们的代码可在https://github.com/havishamy/TransYOLO_3D获取。我们的视频可在https://havishamy.github.io/paper-videos/观看。

View on arXiv Download PDF AI Translation

cs.RO / 22 / 2608.04825

Deliberate Before You Fly: Vision-Guided Spatial Deliberation for UAV See-and-Reach Navigation

飞行前的深思熟虑：基于视觉引导的无人机空间决策导航

Xue, Fanfu, Yu, En, Liu, Bohang, Wang, Hongjun, Yang, Yang, Wang, Xindi, Sun, Jiande

Abstract

UAV see-and-reach navigation requires an aerial agent to approach a language-specified target visible in its initial view and stop reliably near it. Existing methods typically map vision-language representations directly to action outputs without explicitly modeling intermediate fine-grained spatial decisions. This direct mapping causes semantic-control misalignment, leading to inconsistent maneuvers and unreliable termination. To address this issue, we propose DBFly, a vision-language waypoint prediction framework that introduces explicit vision-guided spatial deliberation before waypoint generation. Specifically, DBFly introduces a spatial maneuver decision chain that progressively performs target-direction anchoring, spatial diagnosis, and maneuver decision, enabling high-level maneuver intent to explicitly guide continuous waypoint generation. DBFly further constructs an implicit flight corridor by transforming the initial target-direction prior into a persistent geometric reference and deriving an online corridor state from the UAV's current position, thereby providing soft geometric guidance for spatial diagnosis and maneuver correction. In addition, DBFly develops a terminal-convergence-aware stopping strategy that characterizes terminal states through both target proximity and short-horizon motion convergence, enabling more reliable stopping near the target. Extensive experiments across seen, unseen-object, and unseen-scene test sets demonstrate that DBFly improves the success rate over the SOTA baseline by an average of 25.07 percentage points. The project homepage is available at https://xuefanfu.github.io/DBFly-Page.

Chinese Translation

无人机（UAV）视觉导航要求空中代理接近在其初始视野中可见的语言指定目标，并可靠地在其附近停下。现有方法通常将视觉-语言表示直接映射到动作输出，而未明确建模中间的细粒度空间决策。这种直接映射导致语义控制不一致，从而引发不稳定的操作和不可靠的终止。为了解决这一问题，我们提出了DBFly，一个视觉-语言航点预测框架，在航点生成之前引入明确的视觉引导空间决策。具体而言，DBFly引入了一个空间操作决策链，逐步执行目标方向锚定、空间诊断和操作决策，使得高层次的操作意图能够明确指导连续的航点生成。DBFly进一步通过将初始目标方向先验转化为持久的几何参考，并从无人机当前的位置推导出在线航道状态，构建了一个隐式飞行航道，从而为空间诊断和操作修正提供柔性几何引导。此外，DBFly开发了一种终端收敛感知的停止策略，通过目标接近度和短期运动收敛来表征终端状态，从而在目标附近实现更可靠的停止。在已见、未见目标和未见场景的测试集上进行的广泛实验表明，DBFly的成功率比最先进的基线平均提高了25.07个百分点。项目主页可访问 https://xuefanfu.github.io/DBFly-Page。

View on arXiv Download PDF AI Translation

cs.RO / 23 / 2608.04842

RORA: Realistic Object Reconstruction with Articulation

RORA：具有关节动作的真实物体重建

Lee, Hyesung, Lee, Youngseon, Lee, Kyutae, Lee, Dongjun, Lee, Yongseok

Abstract

Replicating real-world environments into simulation by realistic visual representation like NeRF and 3D Gaussian Splatting (3DGS) has emerged as an effective strategy to reduce the sim-to-real gap in robot learning. However, implementing object articulation during the real-to-sim process is still a challenging task. Existing motion tracking or learning based articulation methods shows low success rates on complex kinematic structures having multiple joints. Furthermore, those methods require scan of dynamic motion of objects, which makes reconstruction process much complicated. In this work, we propose the first end-to-end pipeline that reconstructs simulation-ready assets with accurate articulation from a single static object video input through suggestion based human-in-the-loop process. Our approach exports a hybrid representation combining 3DGS for photorealistic rendering and mesh-based geometry for physical interaction. In the reconstruction process, our pipeline performs convex decomposition followed by user grouping for intuitive part segmentation, subsequently binding 3D Gaussians to the corresponding mesh parts. An Automatic Joint Suggestion Algorithm then calculates candidate joint axes from local boundary geometries and presents them to users for efficient articulated asset reconstruction. We have shown that our method achieves precise articulation results on partnet-mobility-v0 dataset and real objects. Additionally we presented a potential usage of our framework on robot learning, deploying the reconstructed assets in Unreal Engine and NVIDIA Isaac Sim, demonstrating real-time dexterous hand manipulation tasks.

Chinese Translation

通过像 NeRF 和 3D Gaussian Splatting (3DGS) 这样的真实视觉表现，将现实世界环境复制到仿真中，已成为减少机器人学习中的仿真与现实差距的有效策略。然而，在真实到仿真过程中实现物体的关节动作仍然是一项具有挑战性的任务。现有的基于运动跟踪或学习的关节动作方法在具有多个关节的复杂运动结构上成功率较低。此外，这些方法需要对物体的动态运动进行扫描，这使得重建过程变得更加复杂。在本研究中，我们提出了第一个端到端的管道，通过基于建议的人机协作过程，从单个静态物体视频输入中重建具有准确关节动作的仿真准备资产。我们的方法导出了一个混合表示，结合了用于照片真实渲染的 3DGS 和用于物理交互的基于网格的几何体。在重建过程中，我们的管道执行凸分解，随后进行用户分组以实现直观的部分分割，随后将 3D 高斯与相应的网格部分绑定。然后，自动关节建议算法从局部边界几何体中计算候选关节轴，并将其呈现给用户，以便高效地重建关节资产。我们已经证明，我们的方法在 partnet-mobility-v0 数据集和真实物体上实现了精确的关节动作结果。此外，我们展示了我们框架在机器人学习中的潜在应用，将重建的资产部署在 Unreal Engine 和 NVIDIA Isaac Sim 中，演示了实时灵巧手操控任务。

View on arXiv Download PDF AI Translation

cs.RO / 24 / 2608.04905

PRIMAL3: Pathfinding via Reinforcement and Imitation Multi-Agent Learning - Leveraging LaCAM3

PRIMAL3：基于强化学习和模仿学习的多智能体路径规划 - 利用LaCAM3

He, Chengyang, Duhan, Tanishq, Camps, Gadiel Sznaier, Wang, Fangyuan, Cao, Yuhong, Sun, Jiankai, Sun, Ge, Schwager, Mac, Sartoretti, Guillaume

Abstract

We present PRIMAL3, an ultra-large-scale learning-based framework for multi-agent pathfinding (MAPF) that integrates reinforcement learning, topology-aware communication, LaCAM3-guided training, and PIBT-based action refinement. PRIMAL3 targets failures at topologically critical states, where agents must coordinate decisively around bottlenecks, dead ends, and persistent conflicts. Each agent is represented using features derived from cut vertices, dead-end regions, shortest-path distances, and blocking estimates. Two complementary graphs capture agent interactions: a same-direction following graph propagates multihop context along compatible paths, while a different-direction conflict graph differentiates agents competing for shared space through masked attention and relative features. During training, we propose to let policy entropy identify uncertain agents, for which LaCAM3 provides confidence-triggered action interventions and label-smoothed imitation targets. During execution, a priority-aware PIBT module refines the proposed joint actions using persistent, learned, and distance-aware priorities together with policy-aware fallback preferences while maintaining collision-free execution. The resulting framework combines learned exploration with structured expert guidance without requiring LaCAM3 at inference. Experiments demonstrate that PRIMAL3 substantially outperforms state-of-the-art learning-based baselines and scales to ultra-large instances with up to city-level 100,000 agents. Real-world experiments further demonstrate the feasibility of deploying PRIMAL3 on physical robotic systems and ablation studies validate the individual contributions the components we proposed. Project page: https://marmotlab.github.io/PRIMAL3/

Chinese Translation

我们提出了PRIMAL3，一个超大规模的基于学习的多智能体路径规划（MAPF）框架，集成了强化学习、拓扑感知通信、基于LaCAM3的训练指导和基于PIBT的动作优化。PRIMAL3针对拓扑关键状态下的失败，在这些状态中，智能体必须围绕瓶颈、死胡同和持续冲突进行有效协调。每个智能体的表示使用来自割点、死胡同区域、最短路径距离和阻塞估计的特征。两个互补图捕捉智能体之间的交互：同向跟随图沿兼容路径传播多跳上下文，而不同向冲突图通过掩蔽注意力和相对特征区分竞争共享空间的智能体。在训练过程中，我们提出让策略熵识别不确定的智能体，对于这些智能体，LaCAM3提供基于置信度的动作干预和标签平滑的模仿目标。在执行过程中，一个优先级感知的PIBT模块利用持久的、学习到的和距离感知的优先级以及策略感知的后备偏好来优化提议的联合动作，同时保持无碰撞执行。最终的框架结合了学习探索与结构化专家指导，而在推理时不需要LaCAM3。实验表明，PRIMAL3在性能上显著优于最先进的基于学习的基线，并且能够扩展到高达城市级别的100,000个智能体的超大实例。现实世界的实验进一步验证了在物理机器人系统上部署PRIMAL3的可行性，消融研究验证了我们提出的各个组件的独立贡献。项目页面：https://marmotlab.github.io/PRIMAL3/

View on arXiv Download PDF AI Translation

cs.RO / 25 / 2608.04933

Mimir: A Neuro-Symbolic Memory System with Dynamic Grounding for Embodied Agents in Interactive Environments

Mimir：一种具有动态基础的神经符号记忆系统，用于交互环境中的具身智能体

Xu, Haoming, He, Zhenlin, Wang, Hengyi, Xu, Jiafeng, Dong, Hao

Abstract

Long-horizon embodied task requires agents to act under partial observability while preserving both scene belief and execution progress. Flat histories or implicit policy states may contain past observations, but they do not provide an explicit interface for deciding which world facts support the currently active goal. We introduce Mimir, a neuro-symbolic memory that separates world memory from task memory and dynamically grounds them before each action. World memory maintains object locations, object states, and perceptual evidence, while task memory maintains an ordered goal agenda, progress state, hand state, failures, and execution constraints. A grounding module binds the active goal to recalled world candidates, fills missing source locations, and attaches evidence before planning and embodiment-specific execution. Across tested backbones, Mimir consistently improves on different EB-ALFRED and EB-Habitat tasks, with maximum gains of 42.5% and average gains of 23.0%, respectively. Compared with the best results among prior agent and memory systems evaluated under the same backbone, Mimir improves the overall average success rate by 8.5%. Finally, on the EB-Habitat Long-horizon subset, Mimir achieves 86.0% success rate, substantially outperforming current closed-source models. Our code will be released soon.

Chinese Translation

长时间跨度的具身任务要求智能体在部分可观测的情况下行动，同时保持场景信念和执行进度。平坦的历史或隐式策略状态可能包含过去的观察，但它们并未提供明确的接口来决定哪些世界事实支持当前的活跃目标。我们提出了Mimir，一种神经符号记忆，它将世界记忆与任务记忆分离，并在每次行动前动态地将它们基础化。世界记忆维护对象位置、对象状态和感知证据，而任务记忆维护有序的目标议程、进度状态、手部状态、失败和执行约束。基础化模块将活跃目标绑定到回忆的世界候选，填补缺失的源位置，并在规划和具身特定执行之前附加证据。在测试的不同基础架构上，Mimir在不同的EB-ALFRED和EB-Habitat任务中始终表现出改进，最大增益为42.5%，平均增益为23.0%。与在相同基础架构下评估的先前智能体和记忆系统中的最佳结果相比，Mimir将整体平均成功率提高了8.5%。最后，在EB-Habitat长时间跨度子集上，Mimir达到了86.0%的成功率，显著超越了当前的闭源模型。我们的代码将很快发布。

View on arXiv Download PDF AI Translation

cs.RO / 26 / 2608.04996

DreamWAM: Beyond RGB Future Prediction for World Action Models

DreamWAM：超越RGB的世界动作模型未来预测

Yuan, Shanglin, Zhao, Weiheng, Shi, Xin, Jiang, Haoyi, Guo, Xianda, Liu, Liu, Liu, Wenyu, Sui, Wei, Wang, Xinggang

Abstract

World Action Models (WAMs) learn action-relevant representations by predicting how the observed world will evolve. Most existing WAMs define this future in RGB space, where task-relevant state transitions are entangled with nuisance variations in texture, illumination, background, and viewpoint. We argue that WAMs should explicitly predict action-relevant future state rather than relying on RGB prediction alone. We introduce DreamWAM, which reformulates future prediction as structured world modeling beyond RGB, representing future states through complementary views of appearance, motion, geometry, and semantics. During training, DreamWAM combines joint latent denoising of RGB and motion with lightweight gated residual branches for geometry and semantics. Shared attention between VideoDiT and ActionDiT allows the action branch to learn from these future-state predictions, while all beyond-RGB supervision branches are disabled at inference and deployment remains RGB-only. Across both no-rollout and joint video-action inference, DreamWAM consistently improves the matched RGB-only baselines on LIBERO, from 97.30\% to 98.40\% and from 98.00\% to 98.90\%, respectively. The gains become larger under unseen LIBERO-Plus perturbations, from 51.36\% to 63.44\% and from 69.16\% to 75.47\%. The same robustness extends to real-world manipulation, where DreamWAM attains an average success rate of 74.4\% across unseen changes in lighting, background, and object layout, compared with 55.6\% for Fast-WAM-Joint. These results show that robust world-action learning depends not only on predicting the future, but on representing it in a form that matters for action. The code and models are publicly released at https://github.com/hustvl/DreamWAM.

Chinese Translation

世界动作模型（WAMs）通过预测观察到的世界将如何演变来学习与动作相关的表征。现有的大多数WAMs在RGB空间中定义这一未来，其中任务相关的状态转变与纹理、光照、背景和视角等干扰变化交织在一起。我们认为，WAMs应该明确预测与动作相关的未来状态，而不仅仅依赖于RGB预测。我们提出了DreamWAM，它将未来预测重新构建为超越RGB的结构化世界建模，通过外观、运动、几何和语义的互补视角来表示未来状态。在训练过程中，DreamWAM结合了RGB和运动的联合潜在去噪，以及用于几何和语义的轻量级门控残差分支。VideoDiT和ActionDiT之间的共享注意力使得动作分支能够从这些未来状态预测中学习，而在推理时，所有超越RGB的监督分支都被禁用，部署仍然保持RGB-only。在无回滚和联合视频-动作推理中，DreamWAM在LIBERO上的表现始终优于匹配的RGB-only基线，从97.30%提升至98.40%，从98.00%提升至98.90%。在未见的LIBERO-Plus扰动下，增益更为显著，从51.36%提升至63.44%，从69.16%提升至75.47%。这种鲁棒性同样适用于现实世界的操作，其中DreamWAM在未见的光照、背景和物体布局变化下的平均成功率为74.4%，而Fast-WAM-Joint为55.6%。这些结果表明，鲁棒的世界-动作学习不仅依赖于对未来的预测，还依赖于以对动作重要的形式表示未来。代码和模型已公开发布在https://github.com/hustvl/DreamWAM。

View on arXiv Download PDF AI Translation

cs.RO / 27 / 2608.05021

Optimal Constrained sc-LTL Planning in MDPs via Switching Policies

通过切换策略在马尔可夫决策过程中的最优约束sc-LTL规划

Xuan, Zetong, Wang, Yu

Abstract

We study the synthesis of optimal policies for planning problems on Markov decision processes with both objectives and safety constraints specified in co-safe linear temporal logic (sc-LTL). Our problems are inherently non-Markovian due to the complexity of the sc-LTL specification and may require policy randomization to balance the objective and constraint. We propose a novel approach that reduces the constrained sc-LTL planning problem to a constrained reachability problem on an extended model. We then show that a class of switching policies constructed from stationary policies for the individual sc-LTL specifications is sufficient for optimality for the constrained reachability problem. Our finding enables a tractable linear program to compute the optimal policy. A grid world case study demonstrates that our switching policies can achieve the optimal trade-off between the objective and the safety constraint and validates both optimality and tractability.

Chinese Translation

我们研究了在马尔可夫决策过程中合成最优策略的问题，该问题涉及在共同安全线性时序逻辑（sc-LTL）中指定的目标和安全约束。由于sc-LTL规范的复杂性，我们的问题本质上是非马尔可夫的，并可能需要策略随机化以平衡目标与约束。我们提出了一种新颖的方法，将约束sc-LTL规划问题简化为扩展模型上的约束可达性问题。然后，我们证明了一类由个体sc-LTL规范的平稳策略构建的切换策略对于约束可达性问题的最优性是足够的。我们的发现使得可以通过一个可处理的线性规划来计算最优策略。一个网格世界案例研究表明，我们的切换策略能够在目标与安全约束之间实现最佳权衡，并验证了最优性和可处理性。

View on arXiv Download PDF AI Translation

cs.RO / 28 / 2608.05042

BridgeVLA++: A Data-Efficient, Generalizable, and Memory-Augmented Vision-Language-Action Framework for 3D Manipulation

BridgeVLA++：一种数据高效、可泛化且具记忆增强的视觉-语言-动作框架用于三维操作

Li, Peiyan, Zhu, Yuze, Chen, Yixiang, Ma, Qisen, Xu, Yuan, Yang, Jiabing, Guan, He, Huang, Yan, Wu, Hongtao, Ma, Xiao, Kong, Tao, Wang, Liang, Tan, Tieniu

Abstract

Leveraging pre-trained vision-language models (VLMs) to construct vision-language-action (VLA) models has emerged as a promising paradigm for 3D robot manipulation. However, existing 3D VLA methods remain data-hungry, exhibit limited generalization under distribution shifts, and lack explicit memory of past observations. These limitations hinder their application to data-scarce, open-world, and memory-dependent manipulation scenarios. Our previous work, BridgeVLA, improves data efficiency and generalization by preserving the input--output alignment of a pre-trained VLM during 3D action learning: raw point clouds are projected into multi-view images, and intermediate heatmaps are predicted before generating robot actions. In this work, we develop BridgeVLA++ by equipping BridgeVLA with a unified spatio-temporal memory architecture that models persistent spatial context and temporal interaction history. The resulting memory-augmented framework can reason over observation histories while preserving BridgeVLA's data efficiency and generalization capabilities. Extensive experiments show that our framework achieves strong performance on spatial manipulation tasks while exhibiting robust generalization. BridgeVLA++ further achieves state-of-the-art performance on two challenging memory-dependent manipulation benchmarks without sacrificing the data efficiency and generalization of the original BridgeVLA. In addition, BridgeVLA++ performs effectively in bimanual manipulation settings and is validated on an additional real-world robotic platform, demonstrating its scalability across tasks, environments, and robotic platforms. These results establish BridgeVLA++ as a unified 3D vision-language-action framework that simultaneously supports data-efficient learning, robust generalization, and effective memory-aware robot manipulation. Project website: https://bridgevla-plus.github.io/.

Chinese Translation

利用预训练的视觉-语言模型（VLM）构建视觉-语言-动作（VLA）模型已成为三维机器人操作的一个有前景的范式。然而，现有的三维 VLA 方法仍然对数据需求较高，在分布变化下表现出有限的泛化能力，并且缺乏对过去观察的明确记忆。这些局限性阻碍了它们在数据稀缺、开放世界和依赖记忆的操作场景中的应用。我们之前的工作 BridgeVLA 通过在三维动作学习过程中保持预训练 VLM 的输入-输出对齐，改善了数据效率和泛化能力：原始点云被投影到多视图图像中，并在生成机器人动作之前预测中间热图。在本研究中，我们通过为 BridgeVLA 配备统一的时空记忆架构，开发了 BridgeVLA++，该架构建模持久的空间上下文和时间交互历史。所得到的记忆增强框架能够在保持 BridgeVLA 的数据效率和泛化能力的同时，对观察历史进行推理。大量实验表明，我们的框架在空间操作任务上表现出色，同时展现出强大的泛化能力。BridgeVLA++ 进一步在两个具有挑战性的记忆依赖操作基准上实现了最先进的性能，而不牺牲原始 BridgeVLA 的数据效率和泛化能力。此外，BridgeVLA++ 在双手操作设置中表现有效，并在额外的真实世界机器人平台上进行了验证，展示了其在任务、环境和机器人平台上的可扩展性。这些结果确立了 BridgeVLA++ 作为一个统一的三维视觉-语言-动作框架，同时支持数据高效学习、强健的泛化能力和有效的记忆感知机器人操作。项目网站：https://bridgevla-plus.github.io/

View on arXiv Download PDF AI Translation

cs.RO / 29 / 2608.05078

SpikingNav: Robust Embodied Navigation with Spiking Neural Policies

SpikingNav：基于脉冲神经策略的鲁棒体感导航

Zhang, Jiahong, Shen, Sijun, Wu, Dehua, Lin, Yifan, Xia, Xuechen, Chu, Xu, Zhang, Youhui, GuoqiLi

Abstract

Embodied navigation requires an agent to make sequential decisions from egocentric observations in a physical environment. Existing Artificial Neural Network (ANN)-based navigation models have achieved strong performance, yet they often rely on dense computation and may degrade under visual corruptions. Spiking neural networks (SNNs) provide event-driven computation and intrinsic temporal dynamics, which are promising for compact and robust navigation on resource-constrained platforms. However, whether spike-based sensing and policy dynamics can improve robustness in visually rich embodied navigation remains an open problem. This paper proposes SpikingNav, a spiking framework for robust indoor embodied navigation. It contains a Spiking Sensing Encoder (SSE) and a Spiking Policy Network (SPN). The SSE extracts task-conditioned visual features with a spike-based backbone. The SPN maintains a recurrent policy state through membrane integration, thresholding, and spike-triggered reset. In this way, SpikingNav exploits the dynamic properties and spike activations of SNNs to improve navigation performance and robustness. We evaluate SpikingNav on PointNav and ObjectNav under clean observations and visual corruptions. SpikingNav achieves competitive clean performance and stronger robustness with fewer parameters and lower per-step computation than a matched ANN baseline. For instance, SpikingNav improves ObjectNav success from 31.05% to 34.12%, and raises the average success under visual corruptions from 8.45% to 13.71%, demonstrating the benefits of spike-based sensing and policy dynamics. We further validate the deployability of our spike-based sensing method on the Thruster-V2 neuromorphic chip. This physical hardware validation shows that SpikingNav can be instantiated on a real neuromorphic substrate for cyber-physical systems.

Chinese Translation

体感导航要求智能体在物理环境中根据自我中心的观察做出连续决策。现有的基于人工神经网络（ANN）的导航模型已取得了良好的性能，但它们通常依赖于密集计算，并且在视觉干扰下可能会退化。脉冲神经网络（SNN）提供了事件驱动的计算和内在的时间动态特性，这对于在资源受限的平台上实现紧凑且鲁棒的导航具有潜力。然而，基于脉冲的感知和策略动态是否能够改善视觉丰富的体感导航的鲁棒性仍然是一个未解的问题。本文提出了SpikingNav，一个用于鲁棒室内体感导航的脉冲框架。它包含一个脉冲感知编码器（SSE）和一个脉冲策略网络（SPN）。SSE利用基于脉冲的骨干网络提取任务条件下的视觉特征。SPN通过膜积分、阈值处理和脉冲触发重置来维持递归策略状态。通过这种方式，SpikingNav利用SNN的动态特性和脉冲激活来提高导航性能和鲁棒性。我们在清晰观察和视觉干扰下对SpikingNav进行了PointNav和ObjectNav的评估。SpikingNav在清晰性能上表现出竞争力，并且在参数更少、每步计算量更低的情况下展现出更强的鲁棒性，相较于匹配的ANN基线。例如，SpikingNav将ObjectNav的成功率从31.05%提高到34.12%，并将视觉干扰下的平均成功率从8.45%提高到13.71%，展示了基于脉冲的感知和策略动态的优势。我们进一步验证了我们的基于脉冲的感知方法在Thruster-V2神经形态芯片上的可部署性。这一物理硬件验证表明，SpikingNav可以在真实的神经形态基底上实现，用于网络物理系统。

View on arXiv Download PDF AI Translation

计算机视觉 (Computer Vision)

111

cs.CV / 1 / 2608.04061

Advancing Utility Pole and Sign Detection Through Deep Learning

通过深度学习推进电力杆和标志检测

Dickinson, Carl, Di Caterina, Gaetano

Abstract

Utility poles are an essential part of the infrastructure used to support power distribution systems and other critical public services. Their regular inspection is crucial to ensure the stability and safety of the electrical grid. A deep learning framework is presented for the automated detection, segmentation and lean angle estimation of wooden utility poles, and classification of attached electrical warning signs, using ground-level imagery. The system is trained on a custom dataset of 4,570 annotated images extracted from Google Street View, featuring challenging real-world scenes with visually ambiguous wooden poles lacking distinctive features. The proposed model is based on the Detection Transformer (DETR), suitably modified and trained on the custom dataset. The model outperforms standard object detectors (RetinaNet, Faster R-CNN, YOLOv3-Tiny), achieving a mean average precision of 90.43% for pole detection and 88.26% for sign detection. Extending this model with a segmentation head enables per-instance mask generation, which is then used to estimate pole lean angle. The model accurately estimates lean for 1,367 out of 1,433 test-set poles, with a mean absolute error of 1.01 degrees. Moreover, the custom dataset created in this work is also made publicly available to be used as a benchmark.

Chinese Translation

电力杆是支撑电力分配系统和其他重要公共服务基础设施的重要组成部分。定期检查电力杆对于确保电网的稳定性和安全性至关重要。本文提出了一种深度学习框架，用于自动检测、分割木质电力杆及其附加电气警告标志的分类，采用地面图像进行分析。该系统在一个包含4,570幅标注图像的定制数据集上进行训练，这些图像从Google街景中提取，呈现了具有视觉模糊特征的木质电力杆的真实场景。所提出的模型基于检测变换器（Detection Transformer, DETR），经过适当修改并在定制数据集上进行训练。该模型的性能优于标准目标检测器（RetinaNet、Faster R-CNN、YOLOv3-Tiny），在电力杆检测中实现了90.43%的平均精度，在标志检测中实现了88.26%的平均精度。通过扩展该模型以添加分割头，可以生成每个实例的掩膜，进而用于估计电力杆的倾斜角度。该模型准确估计了1,433个测试集电力杆中1,367个的倾斜角度，平均绝对误差为1.01度。此外，本文创建的定制数据集也已公开，以供作为基准使用。

View on arXiv Download PDF AI Translation

cs.CV / 2 / 2608.04106

LoRetta: A Foundation Model and Extensive Dataset for Global-Scale Remote Sensing Dense Image Matching

LoRetta：用于全球尺度遥感密集图像匹配的基础模型和广泛数据集

Yu, Siwei, Guo, Han, Shi, Zhenwei, Zou, Zhengxia

Abstract

Dense image matching establishes pixel-wise correspondences and underpins broad applications in computer vision and photogrammetry. However, extending dense matching to global-scale remote sensing remains challenging because image pairs may differ in acquisition time, season, viewpoint, spatial resolution, and land-cover state. The resulting large geometric offsets, partial overlap, and intrinsically unmatchable regions make direct dense correspondence prediction unreliable and inefficient. We thus reformulate dense matching as localization-and-registration: first localizing the matchable overlap and affine geometry, then refining dense residuals within the aligned frame. Based on this formulation, we propose LoRetta, a foundation model coupling matchability-aware affine localization with guided dense registration. We also introduce LEVIR-GM, a global-scale multi-temporal optical matching benchmark with dataset-native matchability labels (103K aligned, 827K augmented pairs, six continents, five years, 0.5-1024 m resolution). We further establish a unified evaluation protocol for sparse, semi-dense, and dense matchers. On LEVIR-GM, LoRetta achieves an area under the curve (AUC) of 83.3%, outperforming the strongest baseline RoMa v2 by 1.6 points, with larger percentage of correct keypoints (PCK) gains of 6.5 and 8.2 points at 1 and 2 pixels, while reducing inference latency by 47.8%. Astronaut-to-satellite and unmanned aerial vehicle (UAV)-to-satellite geolocalization experiments further demonstrate its transferability as a reusable geometric aligner.

Chinese Translation

密集图像匹配建立了像素级的对应关系，并支撑了计算机视觉和摄影测量的广泛应用。然而，将密集匹配扩展到全球尺度的遥感仍然面临挑战，因为图像对可能在获取时间、季节、视角、空间分辨率和地表覆盖状态上存在差异。由此产生的大几何偏移、部分重叠和本质上无法匹配的区域使得直接的密集对应预测变得不可靠且效率低下。因此，我们将密集匹配重新表述为定位与配准：首先定位可匹配的重叠区域和仿射几何，然后在对齐框架内细化密集残差。基于这一表述，我们提出了LoRetta，一个结合了匹配性感知的仿射定位与引导密集配准的基础模型。我们还引入了LEVIR-GM，一个具有数据集原生匹配性标签的全球尺度多时相光学匹配基准（103K 对齐，827K 增强对，覆盖六大洲，五年，0.5-1024米分辨率）。此外，我们建立了一个统一的评估协议，适用于稀疏、半密集和密集匹配器。在LEVIR-GM上，LoRetta达到了83.3%的曲线下面积（AUC），比最强基线RoMa v2高出1.6个百分点，在1和2像素处的正确关键点百分比（PCK）分别提高了6.5和8.2个百分点，同时将推理延迟减少了47.8%。宇航员到卫星和无人机（UAV）到卫星的地理定位实验进一步证明了其作为可重用几何对齐器的可转移性。

View on arXiv Download PDF AI Translation

cs.CV / 3 / 2608.04111

GEB-Bench: Abstract Structures Told in Many Voices

GEB-Bench：以多种声音讲述的抽象结构

Zhang, Tong, Shi, Zhiyuan, Peng, Yun, Xie, Tao

Abstract

Can a model look at a river delta and a lightning bolt and see that they share a structure? We introduce GEB-Bench, a benchmark whose unit is an abstract structural motif--self-reference, a strange loop, a Mobius twist--in the spirit of Godel, Escher, Bach. Each motif is told in several voices: a natural scene whose composition is the structure, a folk story whose telling enacts it through a mechanically checkable form device, a mathematical theorem, and a programmatic skeleton; surface parameters are declared nuisance variables and never scored. Motifs, voices, and the structural changes between them form a small cross-modal category, and GEB-Bench's tasks are its questions. Evaluating twelve open and proprietary models, we find that abstraction failure is lawful. The central finding is a gap between recognition and cross-voice mapping: models identify a structure within one voice far better than they carry it across voices; every model pays this tax, and mapping strong enough to narrow it appears only at the frontier tier. Two patterns support it. Errors align more strongly with the designed formal geometry than with measured perceptual geometries, and frontier models from different vendors converge on the same wrong answers; and surface complexity taxes every model that reads structure, with capacity buying headroom rather than immunity. GEB-Bench is fully generative and released with its pipeline.

Chinese Translation

一个模型能否观察到河口三角洲和闪电，并识别它们共享的结构？我们介绍了 GEB-Bench，这是一个以抽象结构主题为单位的基准测试——自我引用、奇异循环、莫比乌斯扭转——灵感源自于哥德尔、埃舍尔、巴赫。每个主题通过多种声音进行讲述：一个自然场景，其构成即为结构；一个民间故事，通过一种机械可检查的形式装置来展现它；一个数学定理；以及一个程序框架；表面参数被声明为干扰变量，且从不被评分。主题、声音及其之间的结构变化形成了一个小的跨模态类别，而 GEB-Bench 的任务则是其问题。评估了十二个开放和专有模型后，我们发现抽象失败是有规律的。中心发现是识别与跨声音映射之间的差距：模型在一个声音中识别结构的能力远远优于它在不同声音之间的迁移；每个模型都付出了这一代价，而只有在前沿层次上，强有力的映射才会出现以缩小这一差距。有两个模式支持这一发现。错误与设计的形式几何形状的对齐程度远高于与测量的感知几何形状的对齐程度，而来自不同供应商的前沿模型则趋向于相同的错误答案；并且，表面复杂性对每个读取结构的模型都造成了负担，能力的提升更多是为了获得余地，而非免疫。GEB-Bench 完全是生成性的，并与其管道一起发布。

View on arXiv Download PDF AI Translation

cs.CV / 4 / 2608.04124

Perception Before Reasoning: Dynamic Latent Reasoning for Video Understanding and Question Answering

感知优于推理：用于视频理解和问答的动态潜在推理

Xia, Haotian, Xiao, Zilin, Zou, Junbo, Ordonez, Vicente, Chen, Hanjie

Abstract

Video question answering requires models to ground language queries in visual evidence and, when necessary, reason over that evidence across time. Existing methods typically rely on long textual chain-of-thought rationales, even though many questions can be answered as soon as the relevant object, action, or frame is localized. We propose Dynamic Latent Reasoning (DyLaR), which first grounds a question in a short block of perception latents (continuous hidden states that encode query-relevant visual evidence), and then adaptively decides whether to append reasoning latents (continuous thoughts that reason over this evidence in latent space) before answering. DyLaR learns this behavior by grounding perception latents in verified visual evidence and distilling verified rationales into reasoning latents, followed by reinforcement learning that further refines when to reason. Across nine video benchmarks and four multimodal language model backbones, DyLaR improves average accuracy over same-backbone baselines while generating fewer than 20 tokens per query. On Qwen3-VL-4B, for example, DyLaR improves average accuracy over Qwen3-VL-4B-Thinking from 54.0 to 58.2 while reducing response length from 1,220.7 to 18.5 tokens per query. Ablations further show that grounded perception latents, rationale-supervised reasoning latents, and adaptive routing each improve accuracy.

Chinese Translation

视频问答要求模型将语言查询与视觉证据相结合，并在必要时对这些证据进行时间上的推理。现有方法通常依赖于冗长的文本推理链，尽管许多问题在相关对象、动作或帧被定位后即可回答。我们提出了动态潜在推理（Dynamic Latent Reasoning, DyLaR），该方法首先将问题与一小段感知潜变量（编码查询相关视觉证据的连续隐状态）相结合，然后自适应地决定是否在回答之前附加推理潜变量（在潜在空间中对这些证据进行推理的连续思维）。DyLaR通过将感知潜变量与经过验证的视觉证据相结合，并将经过验证的推理提炼为推理潜变量，来学习这种行为，随后通过强化学习进一步优化推理时机。在九个视频基准和四个多模态语言模型基础上，DyLaR在生成每个查询少于20个标记的情况下，提高了与同基线模型相比的平均准确率。例如，在Qwen3-VL-4B上，DyLaR将平均准确率从54.0提高到58.2，同时将响应长度从每个查询1,220.7个标记减少到18.5个标记。消融实验进一步表明，经过验证的感知潜变量、基于推理的监督推理潜变量和自适应路由均提高了准确率。

View on arXiv Download PDF AI Translation

cs.CV / 5 / 2608.04127

Teaching Foundation Models to Read mmWave: Pose-Guided Kinematic Representation for Human Behavior Understanding

教会基础模型读取毫米波：基于姿态引导的运动学表征用于人类行为理解

Zhang, Duo, Yin, Zhehui, Yao, Zhiyun, Qin, Haotong, Zhang, Xusheng, Yang, Hongliu, Sun, Jianyu, Wang, Junzhe, Fan, Zizhou, Magno, Michele, Zhang, Daqing

Abstract

Large language model agents need to perceive human behavior in physical environments. Millimeter-wave (mmWave) radar provides a privacy-friendly and contactless sensing modality, but radar observations are difficult to align with language. Existing radar-language methods often rely on synthetic data or lack explicit supervision for human body structure and motion. We present mmMind, a radar-language model that uses synchronized 3D pose as training-only supervision. A spatio-temporal radar encoder is pretrained to capture body configuration and motion dynamics, after which the pose head is removed so that inference requires radar alone. The learned radar representations are then aligned with an LLM for behavior captioning and spatio-temporal question answering. We also introduce mmMind-Bench, a real-world mmWave-language benchmark containing 17.9 hours of recordings from 23 participants across seven indoor environments. Experiments on captioning, question answering, and unseen-action generalization show that mmMind consistently outperforms existing radar-language baselines, while ablations confirm the importance of pose-guided pretraining.

Chinese Translation

大型语言模型代理需要感知物理环境中的人类行为。毫米波（mmWave）雷达提供了一种隐私友好且无接触的感知方式，但雷达观测与语言的对齐较为困难。现有的雷达-语言方法通常依赖于合成数据，或缺乏对人体结构和运动的明确监督。我们提出了mmMind，一个使用同步三维姿态作为训练唯一监督的雷达-语言模型。一个时空雷达编码器经过预训练，以捕捉身体配置和运动动态，随后移除姿态头，使得推理仅需依赖雷达。学习到的雷达表征随后与大型语言模型（LLM）对齐，用于行为描述和时空问答。我们还引入了mmMind-Bench，这是一个包含来自23名参与者在七个室内环境中录制的17.9小时真实毫米波-语言基准数据集。关于描述、问答和未见动作泛化的实验表明，mmMind在性能上始终优于现有的雷达-语言基线，同时消融实验确认了姿态引导预训练的重要性。

View on arXiv Download PDF AI Translation

cs.CV / 6 / 2608.04130

Radar4D-VLM: Proposal-Grounded Temporal 4D Radar Reasoning Across Frozen Language Models

Radar4D-VLM：基于提议的时间性4D雷达推理在冻结语言模型中的应用

Han, Jiaju, Sun, Xuemeng, Zhang, Qike, Chen, Xiang, Yang, Luwei, Long, Jiahuan, Wei, Yiwei, Guo, Jiujiang, Hu, Chengyin

Abstract

Vision-language models for autonomous driving primarily rely on cameras and LiDAR, leaving 4D radar largely unexplored as a standalone perceptual modality despite its robustness to adverse visibility and direct measurement of radial velocity. We introduce Radar4D-VLM, a radar-only temporal vision-language model that reasons from ten consecutive 4D-radar point-cloud sweeps without camera or LiDAR input. Radar4D-VLM extracts geometrically grounded object proposals and organizes radar evidence into a compact hierarchy of object, scene, and kinematic tokens. A parameter-efficient projector maps these tokens into frozen language backbones, while auditable prediction heads jointly model object count, spatial distribution, motion state, collision risk, semantic category, and radial velocity. Radar4D-VLM combines proposal-grounded temporal object tokenization, global scene context, and explicit kinematic tokens within a unified frozen-backbone interface. On sequence-isolated K-Radar development validation, its Top-64 proposal recall reaches 98.13% at 4 m, exceeding fixed-lattice and uniform-random controls by 6.40 and 22.83 percentage points, respectively. We further evaluate 24 matched runs spanning eight frozen Qwen, Phi, Mistral, Llama, and Gemma backbones under an identical adaptation budget. The radar-token interface remains compatible across all five language-model families, while matched aligned, permuted, and no-language controls show sensor dependence but no stable direct-head gain from aligned language supervision. These results establish a reproducible foundation for radar-only multimodal scene and motion reasoning while separating interface compatibility from the benefit of language supervision.

Chinese Translation

自主驾驶的视觉-语言模型主要依赖于摄像头和激光雷达（LiDAR），而4D雷达作为一种独立的感知模态，尽管其在恶劣能见度下的鲁棒性和径向速度的直接测量能力，仍然未得到充分探索。我们提出了Radar4D-VLM，这是一种仅基于雷达的时间性视觉-语言模型，它从连续十个4D雷达点云扫描中进行推理，而不依赖于摄像头或激光雷达输入。Radar4D-VLM提取几何基础的物体提议，并将雷达证据组织成物体、场景和运动学标记的紧凑层次结构。一个参数高效的投影器将这些标记映射到冻结的语言骨干网络中，而可审计的预测头共同建模物体数量、空间分布、运动状态、碰撞风险、语义类别和径向速度。在统一的冻结骨干接口中，Radar4D-VLM结合了基于提议的时间性物体标记化、全局场景上下文和显式运动学标记。在序列隔离的K-Radar开发验证中，其Top-64提议召回率在4米处达到了98.13%，超过了固定格子和均匀随机控制，分别提高了6.40和22.83个百分点。我们进一步评估了在相同适应预算下跨越八个冻结的Qwen、Phi、Mistral、Llama和Gemma骨干的24次匹配运行。雷达标记接口在所有五个语言模型家族中保持兼容，而匹配对齐、置换和无语言控制显示出传感器依赖性，但没有从对齐语言监督中获得稳定的直接头增益。这些结果为基于雷达的多模态场景和运动推理建立了可重复的基础，同时将接口兼容性与语言监督的收益分开。

View on arXiv Download PDF AI Translation

cs.CV / 7 / 2608.04132

RUTA: Principled Visual Token Allocation via Rate-Utility Optimization

RUTA：通过速率-效用优化的原则性视觉令牌分配

Zou, Jian, Xu, Xiaoyu, Wang, Zhihua, Wang, Yilin, Adsumilli, Balu, Ma, Kede

Abstract

High-resolution images and long videos provide vision-language models with rich context for multimodal reasoning and fine-grained perception, but the resulting long visual token sequences make large language model-side computation and memory costly. Existing visual token reducers often operate at prescribed rates, while recent methods adapt token counts across inputs using method-specific learned thresholds or importance predictors. We introduce RUTA, a principled Rate-Utility Token Allocation method that performs pre-LLM reduction by jointly learning which tokens to retain and how many to allocate to each image-query pair. RUTA constructs query-conditioned candidate tokens and predicts a retention probability for each candidate. During training, these probabilities parameterize independent Bernoulli gates, while their sum provides a differentiable training-time estimate of the token count for each pair. Retained tokens serve as anchors that aggregate information from non-retained tokens according to semantic affinity and spatial proximity. RUTA is optimized with a penalized rate-utility objective that balances downstream task loss against expected token usage. Averaged across five benchmarks and measured relative to each backbone's full-token baseline, RUTA uses only $2.0\%$ and $4.2\%$ of visual tokens while preserving $88.2\%$ and $94.4\%$ of task performance on LLaVA-NeXT-7B and Qwen3-VL-8B, respectively.

Chinese Translation

高分辨率图像和长视频为视觉-语言模型提供了丰富的上下文，以进行多模态推理和细粒度感知，但由此产生的长视觉令牌序列使得大型语言模型的计算和内存成本高昂。现有的视觉令牌减少器通常在规定的速率下运行，而最近的方法则通过特定方法学习的阈值或重要性预测器在输入之间调整令牌数量。我们提出了RUTA，一种原则性的速率-效用令牌分配方法，通过联合学习保留哪些令牌以及为每个图像-查询对分配多少令牌来进行预-LLM减少。RUTA构建查询条件的候选令牌，并为每个候选令牌预测保留概率。在训练过程中，这些概率参数化独立的伯努利门，而它们的总和提供了每对的令牌数量的可微分训练时间估计。保留的令牌作为锚点，根据语义亲和性和空间接近性聚合来自非保留令牌的信息。RUTA通过一个惩罚的速率-效用目标进行优化，该目标平衡下游任务损失与预期令牌使用之间的关系。在五个基准测试中进行平均，并相对于每个主干的全令牌基线进行测量，RUTA仅使用$2.0\%$和$4.2\%$的视觉令牌，同时在LLaVA-NeXT-7B和Qwen3-VL-8B上分别保留$88.2\%$和$94.4\%$的任务性能。

View on arXiv Download PDF AI Translation

cs.CV / 8 / 2608.04154

TRNet: Topography-Guided Frequency Rectification and Structure-Aware Decoding for Multimodal Paddy Rice Segmentation

TRNet：基于地形引导的频率校正与结构感知解码用于多模态水稻分割

Xiao, Kaiwen, Fu, Chunlong, Zheng, Liping, Su, Yanfeng

Abstract

Mapping paddy rice from very-high-resolution imagery in mountainous and hilly regions is difficult because terrain alters optical appearance and increases confusion with visually similar vegetation. We present TRNet for 0.5-m GaoJing-1 red--green--blue (RGB) imagery, a 5-m TanDEM-X digital elevation model (DEM), and derived slope. Separate visual and terrain encoders preserve modality-specific features. At an early encoder stage, Topographic Energy-Spectral Rectification applies terrain-conditioned low-frequency modulation and asymmetric high-frequency regulation to suppress steep-slope clutter and conditionally enhance compatible low-slope rice cues. The Topography-guided Paddy Structure Decoder combines semantic, rice--background boundary, and interior cues, using coarse terrain as context. Experiments used an Area A internal test set and held-out Area B, which had steeper terrain and lower rice prevalence. TRNet achieved rice intersection-over-union (IoU) values of 85.10\% and 80.68\%, exceeding the original Dual-Encoder U-Net by 9.15 and 18.83 percentage points, respectively. Ablation and slope-stratified results linked these gains to frequency rectification, structure learning, and fewer steep-terrain false positives. The results support coarse topography as a contextual prior for very-high-resolution paddy rice mapping.

Chinese Translation

在山区和丘陵地区，从超高分辨率影像中映射水稻是困难的，因为地形改变了光学外观，并增加了与视觉上相似的植被的混淆。我们提出了TRNet，针对0.5米的高景-1红-绿-蓝（RGB）影像、5米的TanDEM-X数字高程模型（DEM）和衍生的坡度。独立的视觉和地形编码器保留了特定模态的特征。在早期编码器阶段，地形能量-频谱校正应用了基于地形的低频调制和不对称高频调节，以抑制陡坡杂波并有条件地增强兼容的低坡水稻线索。地形引导的水稻结构解码器结合了语义、水稻-背景边界和内部线索，使用粗糙地形作为上下文。实验使用了区域A的内部测试集和保留的区域B，后者具有更陡的地形和较低的水稻出现率。TRNet在水稻交并比（IoU）上达到了85.10%和80.68%，分别超过了原始的双编码器U-Net 9.15和18.83个百分点。消融和坡度分层结果将这些提升与频率校正、结构学习以及较少的陡坡假阳性联系起来。结果支持粗糙地形作为超高分辨率水稻映射的上下文先验。

View on arXiv Download PDF AI Translation

cs.CV / 9 / 2608.04175

TriCLE: Tri-Modal Vision-Language Reasoning for Edge-Deployed Fine-Grained Clustering

TriCLE：边缘部署的三模态视觉-语言推理用于细粒度聚类

Gupta, Kishor Datta, Rahman, Md. Mahfuzur, Rahman, Fahad, Hasan, Ahmed Rafi, Chowdhury, Faysal Mehrab, Haque, Mohd Ariful, George, Roy

Abstract

Edge platforms used for aerial observation must interpret aircraft imagery under limited memory, limited compute, and intermittent connectivity. This setting is difficult for standard RGB-only recognition models and general-purpose vision-language models, especially when calibrated thermal and LiDAR aircraft data are unavailable. We present TriCLE, an application-oriented tri-modal vision-language system for aircraft taxonomic grouping under edge constraints. From a single RGB aircraft image, TriCLE generates a structure-preserving FLIR-style thermal view and a pseudo-LiDAR depth projection, then fuses the aligned views with task instructions in a compact Qwen3-VL backbone. The model is aligned to an expert aircraft taxonomy based on propulsion, airframe family, size, design era, and configuration, so its outputs reflect engineering-relevant similarity rather than only surface appearance. We evaluate supervised fine-tuning, rotation-preserving SFT, and three policy-alignment strategies: GRPO, GSPO, and DAPO. Sequence-level GSPO gives the strongest validation performance, reaching 88.33\% validation accuracy and 0.91 weighted F1 on valid aircraft outputs. On a held-out aircraft test partition, GSPO achieves 78.00\% accuracy and 0.793 weighted F1 while preserving 94.00\% parseable output formatting. After 4-bit quantization and attention-memory optimization, the aligned 4B model fits an 8GB deployment target and processes each tri-modal triplet in 1.48 seconds. These results support TriCLE as a practical prototype for interpretable, edge-feasible aircraft grouping, while emphasizing the need for further validation on real aligned thermal and LiDAR sensor streams.

Chinese Translation

用于空中观察的边缘平台必须在有限的内存、有限的计算能力和间歇性的连接条件下解读航空器图像。这种环境对标准的仅RGB识别模型和通用视觉-语言模型来说是困难的，尤其是在校准的热成像和激光雷达（LiDAR）航空器数据不可用的情况下。我们提出了TriCLE，这是一个面向应用的三模态视觉-语言系统，用于在边缘约束下进行航空器的分类分组。从单张RGB航空器图像出发，TriCLE生成一个结构保留的FLIR风格热成像视图和一个伪LiDAR深度投影，然后将对齐的视图与任务指令在紧凑的Qwen3-VL骨干网络中融合。该模型与基于推进方式、机身家族、尺寸、设计时代和配置的专家航空器分类相一致，因此其输出反映了与工程相关的相似性，而不仅仅是表面外观。我们评估了监督微调、旋转保持的SFT以及三种策略对齐策略：GRPO、GSPO和DAPO。序列级GSPO表现出最强的验证性能，达到了88.33%的验证准确率和0.91的加权F1分数。在一个保留的航空器测试分区中，GSPO实现了78.00%的准确率和0.793的加权F1，同时保持了94.00%的可解析输出格式。在4位量化和注意力-内存优化后，对齐的4B模型适合8GB的部署目标，并在1.48秒内处理每个三模态三元组。这些结果支持TriCLE作为一个可解释的、适合边缘的航空器分组的实用原型，同时强调了在真实对齐的热成像和LiDAR传感器流上进一步验证的必要性。

View on arXiv Download PDF AI Translation

cs.CV / 10 / 2608.04210

PADFormer: Pose-agnostic Anomaly Detection from Sparse View Images

PADFormer：基于姿态无关的稀疏视图图像异常检测

Wang, Ruiqi, Qian, Yiming, Yu, Fenggen, Lu, Yuxuan, Wang, Dakuo, Zhang, Hao, Huang, Jing

Abstract

Pose-agnostic Anomaly Detection (PAD) remains challenging as anomalies can appear under arbitrary viewpoints, requiring methods to handle significant pose variations. Existing approaches rely on complex 3D reconstruction, which are computationally expensive and require extensive multi-view data. We propose PADFormer, a novel image-space approach that leverages Vision Transformer (ViT) to directly reconstruct anomaly-free versions of query images while preserving pose information. Our key insight is to adapt cross-view masked reconstruction for anomaly detection through training exclusively on normal data, combined with dynamic patch selection and spatial alignment mechanisms that enable effective learning from sparse reference views under significant pose variations. During inference, we perform multiple forward passes with different masking patterns to generate an ensemble of anomaly-free reconstructions, ensuring comprehensive coverage of the query image. Anomalies are detected by comparing these reconstructions with the query image. PADFormer achieves state-of-the-art results on the PAD benchmark while maintaining comparable performance on classic few-shot anomaly detection (FSAD) tasks, demonstrating superior efficiency and generalization without requiring 3D reconstruction.

Chinese Translation

姿态无关的异常检测（PAD）仍然面临挑战，因为异常可以在任意视角下出现，这要求方法能够处理显著的姿态变化。现有的方法依赖于复杂的三维重建，这在计算上代价高昂，并且需要大量的多视图数据。我们提出了PADFormer，一种新颖的图像空间方法，利用视觉变换器（Vision Transformer, ViT）直接重建无异常的查询图像版本，同时保留姿态信息。我们的关键见解是通过仅在正常数据上训练，结合动态补丁选择和空间对齐机制，适应跨视图掩蔽重建以进行异常检测，从而有效地从稀疏参考视图中学习，尽管存在显著的姿态变化。在推理过程中，我们使用不同的掩蔽模式进行多次前向传递，以生成无异常重建的集成，确保对查询图像的全面覆盖。通过将这些重建与查询图像进行比较来检测异常。PADFormer在PAD基准测试中实现了最先进的结果，同时在经典的少样本异常检测（FSAD）任务中保持了可比的性能，展示了在不需要三维重建的情况下的优越效率和泛化能力。

View on arXiv Download PDF AI Translation

cs.CV / 11 / 2608.04224

OmniVR: Joint Video-Audio Conditional Generation for Restoring Degraded Historical Films

OmniVR：用于恢复退化历史电影的联合视频-音频条件生成

Lu, Xin, Fan, Zihao, Zhong, Mingchen, Huang, Jie, Fu, Xueyang, Zha, Zheng-Jun

Abstract

Historical films suffer from co-occurring visual and audio degradations---blur, noise, flicker, hiss, clipping, and dropout---yet existing methods restore each modality independently, leaving quality gaps and cross-modal inconsistency. We present OmniVR, the first joint audio-video generative restoration model. Built upon a 22B-parameter audio-video generation backbone, OmniVR formulates restoration as conditional generation within a unified multimodal DiT: the low-quality video and audio are encoded as latent conditions, combined with a fixed restoration prompt, and jointly denoised to recover visual structure, temporal motion, and acoustic detail under one coordinated objective. Three key designs enable this adaptation: (1) a joint audio-video degradation pipeline that simulates real old-film characteristics from Internet-collected data; (2) an architecture-preserving text-to-audio-video (T2AV) to audio-video-to-audio-video (AV2AV) transition with prompt annealing that maximally retains the generative prior; and (3) first-frame image-to-video (I2V) anchoring with loss reweighting and waveform supervision for long-video extrapolation and audio fidelity. We also propose OmniVRBench, the first benchmark that evaluates audio-video restoration across visual quality, audio quality, temporal consistency, and audio-visual synchrony on 200 real historical clips. OmniVR surpasses all prior methods on all six visual metrics, achieves the best audio quality, and produces natural colorization---the first method to jointly address all three aspects. Code and weights will be publicly released. Project Page: https://xin1u.github.io/OminiVR_PAGE/

Chinese Translation

历史电影面临视觉和音频退化的共同问题——模糊、噪声、闪烁、嘶嘶声、剪辑和掉音——然而现有方法独立恢复每种模态，导致质量差距和跨模态不一致。我们提出了OmniVR，这是第一个联合音频-视频生成恢复模型。OmniVR基于一个22B参数的音频-视频生成骨干网络，将恢复过程构建为统一多模态DiT中的条件生成：低质量的视频和音频被编码为潜在条件，结合固定的恢复提示，并共同去噪，以在一个协调的目标下恢复视觉结构、时间运动和声学细节。三个关键设计使得这种适应成为可能：（1）一个联合音频-视频退化管道，模拟来自互联网收集数据的真实老电影特征；（2）一个保留架构的文本到音频-视频（T2AV）到音频-视频到音频-视频（AV2AV）过渡，采用提示退火以最大限度地保留生成先验；（3）第一帧图像到视频（I2V）锚定，结合损失重加权和波形监督，以实现长视频外推和音频保真度。我们还提出了OmniVRBench，这是第一个基准，评估音频-视频恢复在视觉质量、音频质量、时间一致性和音频-视觉同步方面的表现，基于200个真实历史片段。OmniVR在所有六个视觉指标上超越了所有先前的方法，达到了最佳音频质量，并产生了自然的着色——这是第一个联合解决这三个方面的方法。代码和权重将公开发布。项目页面：https://xin1u.github.io/OminiVR_PAGE/

View on arXiv Download PDF AI Translation

cs.CV / 12 / 2608.04244

SIGNPOST-Bench: Benchmarking Text-Vision Conflict Resolution in Multimodal Large Language Models

SIGNPOST-Bench：多模态大型语言模型中文本-视觉冲突解决的基准测试

Li, Sirun, Liu, Minghao, Dai, Ling, Li, Yong, Lyu, Haoxin, Zhou, Junting, Zhang, Fan

Abstract

Multimodal large language models (MLLMs) make grounded predictions in real-world scenes by combining visual and textual cues, yet existing benchmarks rarely reveal how they arbitrate between these evidence sources when they conflict. We introduce SIGNPOST-Bench, a controlled counterfactual benchmark for evaluating text-vision conflict resolution. Each source image is transformed into a counterfactual quintuplet of Original, Blank, Similar, Random, and Adversarial variants. Synthetic, localized scene-text interventions are designed to preserve non-textual content, enabling paired measurements of changes in localization performance and directed shifts toward geographic targets introduced by conflicting text. SIGNPOST-Bench contains 5,111 counterfactual groups and 25,555 image variants from four datasets. We evaluate 20 MLLMs from seven providers. Compared with Original images, Adversarial variants raise median localization error from 282 km to 1,347 km, a 4.8-fold increase. Among geocodable adversarial samples, 6.5-20.1% of predictions lie less than 50 km from the injected target across models, and every evaluated model exhibits a positive mean paired reduction in target distance from Blank to Adversarial. Compatible, unrelated, and conflicting text replacements produce distinct effects on model predictions, while clean-input localization performance does not fully predict robustness to conflicting text. These results establish visual geolocation as a continuous diagnostic of scene-text arbitration and provide a controlled framework for evaluating how MLLMs resolve conflicting multimodal evidence.

Chinese Translation

多模态大型语言模型（MLLMs）通过结合视觉和文本线索，在现实场景中做出基于事实的预测，但现有基准很少揭示它们在证据来源冲突时如何进行裁决。我们引入了SIGNPOST-Bench，这是一个用于评估文本-视觉冲突解决的受控反事实基准。每个源图像被转化为一个反事实五元组，包括原始（Original）、空白（Blank）、相似（Similar）、随机（Random）和对抗（Adversarial）变体。合成的、局部场景文本干预旨在保留非文本内容，从而能够对定位性能的变化和因冲突文本引入的地理目标的定向偏移进行成对测量。SIGNPOST-Bench包含5,111个反事实组和25,555个来自四个数据集的图像变体。我们评估了来自七个提供者的20个MLLMs。与原始图像相比，对抗变体将中位数定位误差从282公里提高到1,347公里，增加了4.8倍。在可地理编码的对抗样本中，6.5%至20.1%的预测距离注入目标不到50公里，并且每个评估模型在从空白到对抗的目标距离上均表现出正的均值成对减少。兼容的、无关的和冲突的文本替换对模型预测产生不同的影响，而干净输入的定位性能并不能完全预测对冲突文本的鲁棒性。这些结果确立了视觉地理定位作为场景文本裁决的连续诊断，并提供了一个受控框架，用于评估MLLMs如何解决冲突的多模态证据。

View on arXiv Download PDF AI Translation

cs.CV / 13 / 2608.04292

Binding Biometrics with AI Agent Identifiers for Delegation of Authority

将生物识别与人工智能代理标识符绑定以实现授权委托

Benjamin, Joseph Geo, Jain, Anil K, Nandakumar, Karthik

Abstract

The proliferation of agentic artificial intelligence (AI) systems has raised serious questions about the accountability for tasks performed by AI agents. Ideally, an AI agent must not be allowed to perform critical tasks without explicit authorization by a human operator. Since biometric recognition is one of the most reliable approaches for authenticating individuals, it has the potential to enable authenticated delegation of authority to AI agents. In this work, we present a framework called BIND, which leverages ideas from the field of biometric cryptosystems, to securely bind biometric data of the human user to the AI agent identity (ID) and authority scope (task-specific constraints) at the time of agent authorization. This token/identifier can be presented by the AI agent to an Identity Auditor, who simultaneously performs biometric authentication and recovers the agent ID and scope, thereby enabling real-time user authentication and establishing a non-repudiable proof of human control and delegation of authority. We also provide a practical implementation of the proposed BIND framework based on face features extracted using standard deep neural network models. To facilitate this implementation, we propose a feature adaptation module that transforms real-valued feature embeddings into fixed-length binary representations suitable for a fuzzy commitment construct based on turbo error correcting codes. Experiments demonstrate the practical feasibility of the proposed face cryptosystem, achieving a True Match Rate of $96\%$ at zero False Match Rate and supporting $1024$-bit agent tokens.

Chinese Translation

代理型人工智能（AI）系统的快速发展引发了关于AI代理执行任务的责任问题。理想情况下，AI代理在没有人类操作员明确授权的情况下，不应被允许执行关键任务。由于生物识别是验证个体身份的最可靠方法之一，它有潜力实现对AI代理的经过认证的授权委托。在本研究中，我们提出了一个名为BIND的框架，该框架利用生物识别密码系统领域的理念，在代理授权时安全地将人类用户的生物特征数据与AI代理身份（ID）和权限范围（任务特定约束）绑定。该令牌/标识符可以由AI代理呈现给身份审计员，后者同时执行生物识别认证并恢复代理ID和权限范围，从而实现实时用户认证，并建立不可否认的人类控制和授权委托的证明。我们还提供了基于标准深度神经网络模型提取的面部特征的BIND框架的实际实现。为了促进这一实现，我们提出了一个特征适配模块，该模块将实值特征嵌入转换为适合基于涡轮纠错码的模糊承诺构造的固定长度二进制表示。实验表明，所提出的面部密码系统在零假匹配率下实现了96%的真实匹配率，并支持1024位代理令牌。

View on arXiv Download PDF AI Translation

cs.CV / 14 / 2608.04302

CLIP-CC-Bench: Evaluating Paragraph-Level Video Descriptions in Video-Language Models

CLIP-CC-Bench：评估视频语言模型中的段落级视频描述

Ali, Mukhtiar, Dubey, Harsh, Mishra, Sugam, Pack, Chulwoo

Abstract

Benchmarking video-language models has largely focused on short clips and single-sentence metrics, leaving open whether current systems can generate accurate long-form, paragraph-level descriptions. We introduce CLIP-CC-Bench, an evaluation suite for long-form video description built from 5 hours of movie content segmented into 90-second clips, each paired with an expert-written paragraph-style reference. The evaluation suite employs an ensemble of five state-of-the-art LLM-based embedding models to increase reliability and mitigate single-model bias, and applies two complementary methodologies: (i) coarse-grained semantic matching and (ii) fine-grained semantic matching to compare model-generated descriptions against CLIP-CC-Bench references. Using this framework, we evaluate 17 state-of-the-art video-language models and report both their Borda-aggregated rankings and their average scores on CLIP-CC-Bench. We further quantify the protocol's internal reliability through inter-judge agreement and bootstrap ranking stability. We release standardized evaluation scripts, model outputs, and aggregation tools at https://github.com/Multimodal-Intelligence-Lab/CLIP-CC-Bench to support reproducibility. CLIP-CC-Bench provides a practical evaluation framework for long-form video description, filling a gap left by existing short-clip and QA-only benchmarks.

Chinese Translation

视频语言模型的基准测试主要集中在短视频片段和单句指标上，因此尚不清楚当前系统是否能够生成准确的长篇段落级描述。我们推出了CLIP-CC-Bench，这是一个基于5小时电影内容构建的长篇视频描述评估套件，该内容被分割为90秒的片段，每个片段都配有专家撰写的段落式参考文本。该评估套件采用五种最先进的基于大型语言模型（LLM）的嵌入模型的集成，以提高可靠性并减轻单一模型的偏差，并应用两种互补的方法论：（i）粗粒度语义匹配和（ii）细粒度语义匹配，以比较模型生成的描述与CLIP-CC-Bench参考之间的差异。利用这一框架，我们评估了17种最先进的视频语言模型，并报告了它们在CLIP-CC-Bench上的Borda聚合排名和平均得分。我们进一步通过评审者间一致性和自助排名稳定性量化了该协议的内部可靠性。我们在https://github.com/Multimodal-Intelligence-Lab/CLIP-CC-Bench发布了标准化评估脚本、模型输出和聚合工具，以支持可重复性。CLIP-CC-Bench为长篇视频描述提供了一个实用的评估框架，填补了现有短片段和仅限问答基准留下的空白。

View on arXiv Download PDF AI Translation

cs.CV / 15 / 2608.04329

An Analysis and Implementation of Seam Carving for Content-Aware Image Resizing

内容感知图像缩放的缝隙雕刻分析与实现

Tosoni, Francesco

Abstract

Seam carving is a classical content-aware image resizing operator that modifies the width or height of an image by repeatedly removing (or inserting) seams, i.e., 8-connected monotonic paths of pixels of locally minimal importance. Because seams bend around salient content rather than uniformly scaling or cropping it, the operator preserves vital image structures while discarding (or duplicating) low-energy regions. This article describes a C++ implementation of the operator that follows the original formulation of Avidan and Shamir (2007), including the optional forward-energy criterion subsequently introduced by Rubinstein, Shamir and Avidan (2008). The implementation supports image reduction, image enlargement via ordered seam insertion, multi-pass enlargement for large scale factors, a user-supplied weight mask for object protection and removal, along with dumping of energy maps and visualisation of seams. We detail the algorithm, its parameters and its computational complexity, discuss design choices with respect to the original descriptions, and illustrate the behaviour of the operator on natural images.

Chinese Translation

缝隙雕刻是一种经典的内容感知图像缩放操作，通过反复移除（或插入）缝隙，即局部最小重要性的像素的8连通单调路径，来修改图像的宽度或高度。由于缝隙绕过显著内容，而不是均匀缩放或裁剪，因此该操作在丢弃（或复制）低能量区域的同时，能够保留重要的图像结构。本文描述了一种遵循Avidan和Shamir（2007）原始公式的C++实现，包括随后由Rubinstein、Shamir和Avidan（2008）引入的可选前向能量标准。该实现支持图像缩小、通过有序缝隙插入进行图像放大、大比例因子的多次放大、用户提供的权重掩模用于对象保护和移除，并支持能量图的转储和缝隙的可视化。我们详细介绍了算法、其参数及计算复杂性，讨论了与原始描述相关的设计选择，并展示了该操作在自然图像上的表现。

View on arXiv Download PDF AI Translation

cs.CV / 16 / 2608.04348

iStructTab: Structured Feature Sequencing for Multimodal Learning of Image and Tabular Data

iStructTab：用于图像和表格数据多模态学习的结构化特征序列化

Habib, Al Zadid Sultan Bin, Ahamed, Md Younus, Gyawali, Prashnna, Doretto, Gianfranco, Adjeroh, Donald A.

Abstract

Multimodal learning of images and tabular data is often impaired by ineffective representations, resulting in redundancy, dispersion, and generalization problems. To tackle this challenge, we introduce Graph-Enhanced Descriptor Sequencing (GEDS), a structured feature sequencing algorithm grounded in principles from the Column Permutation Problem (CPP). GEDS refines statistical descriptors of the features through similarity graph-based computations, systematically determining an effective feature sequencing. We incorporate GEDS within an order-aware efficient transformer framework, utilizing order-aware memory tokens that explicitly adhere to the derived feature sequencing via a dedicated loss function. Experimental results across multimodal benchmarks demonstrate that iStructTab effectively minimizes feature dispersion, improving predictive performance and robustness, and highlighting the significance of structured feature sequencing in multimodal learning.

Chinese Translation

图像和表格数据的多模态学习常常受到无效表示的影响，导致冗余、分散和泛化问题。为了解决这一挑战，我们提出了图增强描述符序列化（Graph-Enhanced Descriptor Sequencing, GEDS），这是一种基于列置换问题（Column Permutation Problem, CPP）原理的结构化特征序列化算法。GEDS通过基于相似性图的计算来优化特征的统计描述符，系统地确定有效的特征序列化。我们将GEDS纳入一个顺序感知的高效变换器框架，利用顺序感知的记忆标记，明确遵循通过专用损失函数导出的特征序列化。跨多模态基准的实验结果表明，iStructTab有效地最小化了特征分散，提高了预测性能和鲁棒性，突显了结构化特征序列化在多模态学习中的重要性。

View on arXiv Download PDF AI Translation

cs.CV / 17 / 2608.04349

Poly-OPD: Heterogeneous Multi-Teacher On-Policy Distillation for Capability-Selectable Flow Models

Poly-OPD：异构多教师在线蒸馏用于能力可选流模型

Fu, Siming, Xu, Haojun, He, Ruizhe, Fu, Zheming, Wang, Hualiang, Huang, Jie, Ma, Xiaoxiao, Zhong, Mingchen, Huang, Weihu, He, Xiaoxuan, Huang, Linjiang, Liu, Si

Abstract

Leading open text-to-image models often carry complementary strengths: one may lead on preference-aligned aesthetics while another follows compositional instructions more faithfully. However, differences in their autoencoders and noise schedules make it difficult to transfer these strengths across models. In this paper, we present Poly-OPD, a framework that can consolidate complementary strengths of heterogeneous teachers into a single compact flow-matching student. To bridge the incompatible latent spaces of different teachers, Poly-OPD performs on-policy distillation through a pixel bridge. Each student-generated image is re-encoded by a selected teacher's encoder and refined from a noise level matched by magnitude under the teacher's noise schedule. The resulting target is further matched to the student in frozen DINOv2 space, enabling supervision across incompatible latent spaces. To retain complementary capabilities without cross-teacher interference, Poly-OPD uses a gradient compatibility diagnostic to organize its adapters: attention LoRA modules are shared across teachers, whereas feed-forward adapters remain teacher-specific. During distillation, a gap-aware curriculum devotes more training to compositional categories where the student still falls short of the teacher. As each gap narrows, training shifts toward categories with larger remaining gaps. By distilling FLUX.1-dev and Z-Image into a 2.5B SD3.5-Medium student, Poly-OPD improves GenEval from 67.3 to 73.3, surpassing both larger teachers, and raises DrawBench HPSv3 from 9.34 to 11.35, consolidating both strengths within a switchable model.

Chinese Translation

领先的开放文本到图像模型通常具有互补的优势：一个模型可能在偏好对齐的美学上表现突出，而另一个模型则更忠实于组合指令。然而，它们的自编码器和噪声调度的差异使得在模型之间转移这些优势变得困难。在本文中，我们提出了Poly-OPD，一个能够将异构教师的互补优势整合为单一紧凑流匹配学生的框架。为了弥合不同教师的不兼容潜在空间，Poly-OPD通过像素桥进行在线蒸馏。每个学生生成的图像都由选定教师的编码器重新编码，并根据教师的噪声调度在幅度上匹配的噪声水平进行精炼。生成的目标进一步在冻结的DINOv2空间中与学生进行匹配，从而实现跨不兼容潜在空间的监督。为了在不干扰教师之间的情况下保留互补能力，Poly-OPD使用梯度兼容性诊断来组织其适配器：注意力LoRA模块在教师之间共享，而前馈适配器则保持教师特定。在蒸馏过程中，基于差距的课程将更多的训练投入到学生仍然落后于教师的组合类别中。随着每个差距的缩小，训练逐渐转向剩余差距较大的类别。通过将FLUX.1-dev和Z-Image蒸馏为一个25亿参数的SD3.5-Medium学生，Poly-OPD将GenEval从67.3提高到73.3，超越了两个更大的教师，并将DrawBench HPSv3从9.34提高到11.35，在一个可切换模型中整合了两者的优势。

View on arXiv Download PDF AI Translation

cs.CV / 18 / 2608.04379

Image Classification Using CNN-QNN Hybrid Model with Optimized Correlated Features

基于优化相关特征的CNN-QNN混合模型图像分类

Seong, Minseo, Kim, Youngwook

Abstract

We propose a method to optimize the correlation among convolutional neural network (CNN) features that are used as inputs to quantum neural network (QNN) to enhance image classification accuracy. Unlike prior approaches that employ orthogonal decomposition as preprocessing, we intentionally introduce correlated features that are more physically compatible with QNN. This design leverages the QNN's inherent ability to exploit quantum entanglement for representing correlated states-an advantage unavailable to classical neural networks. We hypothesize that aligning feature correlations with the entanglement structure of QNN improves binary classification performance. Based on a mathematical derivation of QNN outputs, Monte Carlo simulations indicate that an average correlation between features of 0.5 yields optimal classification accuracy. To validate this finding, we evaluate a quantum-classical hybrid model on three tasks: CIFAR-10 (automobile vs. truck), Fashion-MNIST (shirt vs. coat), and radar micro-Doppler signatures (robotic dogs vs. non-robots). To regulate feature correlations, we introduce a correlation-regularization term on the outputs of the CNN, driving the off-diagonal entries of the feature correlation matrix toward a target constant. Across all datasets, inducing intermediate correlation consistently improved accuracy compared to low, high, or unregulated correlations, while also reducing classification accuracy variance. These results demonstrate that imposing moderate feature correlations-without modifying the quantum circuit-enhances classification accuracy and stability by aligning feature statistics with the QNN's entanglement structure. This study highlights the potential of QNN to surpass the performance of classical classifiers as more qubits become available.

Chinese Translation

我们提出了一种优化卷积神经网络（CNN）特征之间相关性的方法，这些特征作为输入用于量子神经网络（QNN），以提高图像分类的准确性。与以往采用正交分解作为预处理的方法不同，我们故意引入了与QNN更具物理兼容性的相关特征。这一设计利用了QNN固有的能力，通过量子纠缠来表示相关状态，这是经典神经网络所不具备的优势。我们假设，将特征相关性与QNN的纠缠结构对齐可以改善二分类性能。基于QNN输出的数学推导，蒙特卡洛模拟表明，特征之间平均相关性为0.5时，分类准确性最佳。为了验证这一发现，我们在三个任务上评估了量子-经典混合模型：CIFAR-10（汽车与卡车）、Fashion-MNIST（衬衫与外套）和雷达微多普勒特征（机器人狗与非机器人）。为了调节特征相关性，我们在CNN的输出上引入了相关性正则化项，推动特征相关矩阵的非对角线项朝向目标常数。在所有数据集中，诱导中等相关性始终比低、高或不受调节的相关性提高了准确性，同时降低了分类准确性方差。这些结果表明，在不修改量子电路的情况下，施加适度的特征相关性能够通过将特征统计与QNN的纠缠结构对齐来提高分类准确性和稳定性。本研究强调了随着更多量子比特的可用，QNN超越经典分类器性能的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 19 / 2608.04385

ReGround: Restoring Visual Grounding in Multi-Step Reasoning through Self-Diagnosis and Visual Re-Examination

ReGround：通过自我诊断和视觉再检查恢复多步骤推理中的视觉定位

Peng, Lei, Lv, Shuai, Hu, Wei

Abstract

Vision-Language Models (VLMs) often lose visual grounding during multi-step reasoning: as reasoning chains grow longer, later inference steps rely increasingly on language priors rather than image evidence. We identify a consistent benchmark-level signature associated with this degradation: across 2,510 re-examined samples from four benchmarks, attention entropy over image tokens typically decreases during Round 1 and rises again after image re-injection. However, we find that effective visual re-examination requires two complementary ingredients: image re-injection and targeted self-diagnosis. Without targeted diagnosis, re-examination can even hurt performance, whereas accurate self-diagnosis yields substantial gains -- a swing of several points on key benchmarks, indicating that diagnostic quality is a key factor in whether re-examination helps or hurts in our setting. We present ReGround, a two-stage framework that teaches VLMs to self-diagnose grounding failures and selectively re-examine visual evidence, without architectural modifications or external tools. Through capability bootstrapping, a stronger variant from the same model family provides diagnostic scaffolding only during data construction, while the policy model learns to diagnose autonomously at inference time and retains most of the assisted gains. Experiments on eight benchmarks across two VLM backbones demonstrate consistent gains, especially on visually intensive multi-step reasoning tasks, while incurring only modest inference overhead relative to tool-augmented baselines. Project page: https://sespoir.github.io/reground-page/ . Code: https://github.com/sespoir/ReGround .

Chinese Translation

视觉-语言模型（VLMs）在多步骤推理过程中常常失去视觉定位：随着推理链的延长，后续推理步骤越来越依赖于语言先验而非图像证据。我们识别出与这种退化相关的一种一致的基准级特征：在来自四个基准的2,510个重新检查样本中，图像标记的注意力熵通常在第一轮中下降，随后在图像重新注入后再次上升。然而，我们发现有效的视觉再检查需要两个互补的要素：图像重新注入和针对性的自我诊断。如果没有针对性的诊断，再检查甚至可能会损害性能，而准确的自我诊断则会带来显著的提升——在关键基准上提升几个百分点，表明诊断质量是决定再检查在我们设置中是否有帮助或有害的关键因素。我们提出了ReGround，一个两阶段框架，教会VLMs自我诊断定位失败并选择性地重新检查视觉证据，而无需架构修改或外部工具。通过能力引导，同一模型家族中的更强变体仅在数据构建期间提供诊断支撑，而策略模型则在推理时学习自主诊断，并保留大部分辅助收益。在两个VLM骨干网络的八个基准上的实验表明，尤其是在视觉密集的多步骤推理任务上，表现出一致的提升，同时相对于工具增强的基线仅产生适度的推理开销。项目页面：https://sespoir.github.io/reground-page/。代码：https://github.com/sespoir/ReGround。

View on arXiv Download PDF AI Translation

cs.CV / 20 / 2608.04394

Free-Lunch Augmentation by Revisiting Diffusion-Based Data Generation for Cross-Domain Few-Shot Object Detection

通过重新审视基于扩散的数据生成实现免费午餐增强的跨域少样本目标检测

Zhuang, Zijian, Zou, Yixiong, Li, Yuhua, Li, Ruixuan

Abstract

Cross-Domain Few-Shot Object Detection (CDFSOD) aims to transfer knowledge from data-rich upstream generic domains to downstream expert domains using scarce training data, where the significant domain gap and data scarcity make it an unsolved challenge. To address this problem, we revisit a natural yet underexplored approach in CDFSOD: data augmentation, by directly synthesizing data through diffusion models to supplement limited training samples. However, due to large domain gaps, we find that current diffusion methods cannot produce good results, leading to performance even lower than using the original images. To address these limitations, we divide the domain gaps into visual gaps and semantic gaps for separate analysis. For the visual gap, we find that the diffusion model cannot distinguish noise from useful information on expert domains, which can be mitigated by adding weakened noise. For the semantic gap, we find that the background semantics shows much smaller gaps between domains than foreground semantics, and we can bridge this gap by background inpainting. Based on the above analysis, we propose a method (Selective Inpainting with Tailored Noise, SITN) to dynamically take different strategies for downstream data synthesis based on their different gaps from the general domain, including a Generation Module for adding tailored noise and a Selection Module to dynamically select the inpainting regions. Extensive experiments on 6 datasets of CDFSOD and 4 datasets of cross-domain few-shot segmentation (CDFSS) validate that we can synthesize helpful data, achieving new state-of-the-art performance. Our codes is available at https://github.com/zzzzj311-droid/Free-Lunch-SITN

Chinese Translation

跨域少样本目标检测（CDFSOD）旨在利用稀缺的训练数据将知识从数据丰富的上游通用领域转移到下游专家领域，但显著的领域差距和数据稀缺性使其成为一个未解决的挑战。为了解决这个问题，我们重新审视了CDFSOD中一种自然但未被充分探索的方法：数据增强，通过直接利用扩散模型合成数据来补充有限的训练样本。然而，由于领域之间存在较大的差距，我们发现当前的扩散方法无法产生良好的结果，导致性能甚至低于使用原始图像。为了解决这些局限性，我们将领域差距分为视觉差距和语义差距进行单独分析。对于视觉差距，我们发现扩散模型无法在专家领域中区分噪声和有用信息，而通过添加减弱的噪声可以缓解这一问题。对于语义差距，我们发现背景语义在领域之间的差距远小于前景语义，我们可以通过背景修复来弥合这一差距。基于上述分析，我们提出了一种方法（选择性修复与定制噪声，SITN），根据与通用领域的不同差距动态采取不同策略进行下游数据合成，包括用于添加定制噪声的生成模块和动态选择修复区域的选择模块。在6个CDFSOD数据集和4个跨域少样本分割（CDFSS）数据集上的大量实验验证了我们可以合成有用的数据，达到了新的最先进性能。我们的代码可在 https://github.com/zzzzj311-droid/Free-Lunch-SITN 获取。

View on arXiv Download PDF AI Translation

cs.CV / 21 / 2608.04396

CofactVLA: Deconfounding Vision-Language-Action Models via Counterfactual Intervention

CofactVLA：通过反事实干预消除视觉-语言-行动模型的混淆

Zhang, Yan, Wu, Yinan, Duan, Haoran, Han, Jungong

Abstract

Vision-Language-Action (VLA) models have driven significant progress in robotic manipulation, yet they fundamentally struggle with the vision-override phenomenon. Driven by the severe modality imbalance between dense visual streams and sparse linguistic instructions, VLAs frequently fall prey to causal confusion. Instead of treating language as the primary causal driver, the policy entirely bypasses the original instruction by overfitting to spurious visual confounders, such as prominent objects or familiar layouts. To systematically alleviate this bias, we formalize the process of action generation as a Dual-path Deconfounding Graph (DDG) and propose CofactVLA, a novel causal intervention framework. By dynamically constructing a language-masked counterfactual branch within a single forward pass, CofactVLA isolates and neutralizes visual confounders through two synergistic mechanisms. First, Action-Level Orthogonal Projection Guidance (OPG) geometrically projects the factual velocity field away from the counterfactual visual bias during continuous flow matching, extracting the pure semantic intent. Second, Feature-Level Counterfactual Covariance Reduction (CCR) mathematically deconfounds latent representations by penalizing the positive eigenspace of the covariance difference, explicitly suppressing dominant visual shortcuts while preserving the causal language intent. Extensive experiments demonstrate that CofactVLA establishes a new state-of-the-art across diverse simulation benchmarks. Beyond simulation, real-world robot experiments demonstrate the causal efficacy of our method in bridging the generalization gap, yielding a 52.3\% absolute success rate gain under out-of-distribution scenarios.

Chinese Translation

视觉-语言-行动（VLA）模型在机器人操作中推动了显著进展，但它们在根本上面临视觉覆盖现象的挑战。由于密集视觉流与稀疏语言指令之间存在严重的模态不平衡，VLA经常受到因果混淆的困扰。模型并未将语言视为主要的因果驱动因素，而是通过过拟合于虚假的视觉混淆因素（如显著物体或熟悉布局）完全绕过原始指令。为了系统性地缓解这种偏见，我们将动作生成的过程形式化为双路径去混淆图（Dual-path Deconfounding Graph, DDG），并提出CofactVLA，这是一种新颖的因果干预框架。CofactVLA通过在单次前向传递中动态构建语言掩蔽的反事实分支，利用两种协同机制隔离和中和视觉混淆因素。首先，动作级正交投影引导（Action-Level Orthogonal Projection Guidance, OPG）在连续流匹配过程中几何地将事实速度场投影远离反事实视觉偏差，从而提取纯粹的语义意图。其次，特征级反事实协方差减少（Feature-Level Counterfactual Covariance Reduction, CCR）通过惩罚协方差差异的正特征空间在数学上去混淆潜在表示，明确抑制主导的视觉捷径，同时保留因果语言意图。大量实验表明，CofactVLA在多样的仿真基准上建立了新的最先进水平。除了仿真，真实世界的机器人实验展示了我们方法在弥合泛化差距方面的因果有效性，在分布外场景下实现了52.3%的绝对成功率提升。

View on arXiv Download PDF AI Translation

cs.CV / 22 / 2608.04404

Faster-WAM: Efficient Inference-Time Future Conditioning for Robust World Action Models

Faster-WAM：高效推理时未来条件化的稳健世界行动模型

Zhao, Weiheng, Jiang, Haoyi, Shi, Xin, Liu, Liu, Huang, Fan, Su, Zhizhong, Sui, Wei, Wang, Xinggang

Abstract

World Action Models (WAMs) improve robot manipulation by learning how the environment evolves beyond the current observation. However, existing approaches face a fundamental dilemma: Joint-WAMs preserve future-aware representations during inference but incur prohibitive computation costs, while efficient alternatives remove future modeling at inference time and may lose the robustness benefits of temporal reasoning. In this work, we revisit the role of future representations in WAMs and show that inference-time future conditioning is critical for generalization under distribution shifts. This observation motivates Faster-WAM, an efficient future-conditioning WAM that preserves future representations while avoiding expensive video-action interaction. Faster-WAM introduces a sparse future-conditioning framework that computes future representations once and selectively reuses them throughout action denoising. Specifically, we propose SparseMoT to replace ubiquitous layer-wise fusion with selective video-action interaction at a compact subset of network stages, and Interval KV-Fusion to aggregate multi-depth future representations without increasing attention complexity. Experiments demonstrate that Faster-WAM achieves a substantially better performance-efficiency trade-off than existing WAMs. On the out-of-distribution LIBERO-Plus benchmark, Faster-WAM improves success rate from 49.14% to 73.57% compared with Fast-WAM, while running 2.21$\times$ faster than Joint-WAM. It further achieves state-of-the-art performance on LIBERO and RoboTwin 2.0, while demonstrating strong robustness in real-world manipulation.

Chinese Translation

世界行动模型（WAMs）通过学习环境如何在当前观察之外演变来改善机器人操作。然而，现有方法面临一个基本的困境：联合WAM（Joint-WAMs）在推理过程中保留未来感知表示，但会产生高昂的计算成本，而高效的替代方案则在推理时去除了未来建模，可能会失去时间推理的稳健性。在本研究中，我们重新审视了未来表示在WAM中的作用，并表明推理时的未来条件化对于在分布变化下的泛化至关重要。这一观察促使我们提出Faster-WAM，这是一种高效的未来条件化WAM，能够在避免昂贵的视频-动作交互的同时保留未来表示。Faster-WAM引入了一种稀疏未来条件化框架，该框架仅计算一次未来表示，并在整个动作去噪过程中选择性地重用它们。具体而言，我们提出SparseMoT，用选择性的视频-动作交互替代普遍的层级融合，并提出Interval KV-Fusion以在不增加注意力复杂度的情况下聚合多深度未来表示。实验表明，Faster-WAM在性能与效率的权衡上显著优于现有的WAM。在分布外的LIBERO-Plus基准上，与Fast-WAM相比，Faster-WAM的成功率从49.14%提高到73.57%，同时运行速度比Joint-WAM快2.21倍。此外，Faster-WAM在LIBERO和RoboTwin 2.0上实现了最先进的性能，同时在实际操作中展现出强大的稳健性。

View on arXiv Download PDF AI Translation

cs.CV / 23 / 2608.04412

muSync-GS: Physics-Synchronized Driving Video Synthesis for Weather and Geometric Road Hazards

muSync-GS：针对天气和几何道路危险的物理同步驾驶视频合成

Chen, Yang, Zhu, Yicheng, Li, Tao, Bian, Zilin

Abstract

High-quality driving data are essential for autonomous-driving systems and generative world models. However, rare and safety-critical scenarios involving adverse weather, braking under low tire--road friction, and uneven road geometry are costly and risky to collect at scale. Existing video-generation and 3D Gaussian editing methods can modify weather appearance or road geometry, but typically do not couple these edits with tire--road interaction and vehicle dynamics. As a result, an edited video may retain its original trajectory even when the modified road condition should alter braking, wheel slip, load transfer, and ego-camera motion. We present muSync-GS, a physics-synchronized framework for driving video synthesis under adverse-weather and road-elevation hazards. A precipitation-derived road-surface condition jointly controls road appearance and tire friction, while a shared road-elevation profile drives both visible road-geometry editing and axle excitation. A calibrated vehicle model predicts speed, slip ratio, normal loads, and pitch for constructing the ego-camera trajectory and synchronized physical annotations. On 12 held-out CarSim cases spanning precipitation levels, brake inputs, and road-profile parameters, the model achieves mean case-wise RMSEs of 0.0273 m/s for speed, 0.0590 degrees for pitch, 0.0101 for slip ratio, and 26.61 N for per-wheel normal load. Together with the reconstructed-scene experiments, these results show that muSync-GS accurately reproduces vehicle responses under held-out controls while synchronizing them with controllable scene edits and ego-camera motion.

Chinese Translation

高质量的驾驶数据对于自主驾驶系统和生成世界模型至关重要。然而，涉及恶劣天气、低轮胎-路面摩擦下的制动以及不平坦道路几何的稀有且安全关键场景在大规模收集时成本高昂且风险较大。现有的视频生成和三维高斯编辑方法可以修改天气外观或道路几何，但通常不将这些编辑与轮胎-路面交互和车辆动力学耦合。因此，编辑后的视频可能保留其原始轨迹，即使修改后的道路条件应该改变制动、轮胎滑移、载荷转移和自我摄像机运动。我们提出了muSync-GS，这是一个针对恶劣天气和道路高程危险的物理同步驾驶视频合成框架。基于降水的路面条件共同控制道路外观和轮胎摩擦，而共享的道路高程轮廓驱动可见的道路几何编辑和车轴激励。经过校准的车辆模型预测速度、滑移比、法向载荷和俯仰角，以构建自我摄像机轨迹和同步的物理注释。在涵盖降水水平、制动输入和道路轮廓参数的12个保留CarSim案例中，该模型在速度、俯仰角、滑移比和每轮法向载荷方面的平均案例均方根误差（RMSE）分别为0.0273 m/s、0.0590度、0.0101和26.61 N。结合重建场景实验，这些结果表明muSync-GS在保持控制的同时准确再现了车辆响应，并将其与可控场景编辑和自我摄像机运动同步。

View on arXiv Download PDF AI Translation

cs.CV / 24 / 2608.04423

Foreseeing the Invisible: Amodal Reconstruction of Leaf Fossil Images

预见无形：叶子化石图像的无模态重建

Yue, Liuxiang, Zhang, Ailin, Zhao, Ziyue, Duan, Yikun

Abstract

Fossil leaves are rarely preserved whole -- sedimentary rock hides, breaks, and erodes the lamina, yet paleobotany depends on the complete shape and outline of the leaf. We cast the recovery of the missing tissue as amodal reconstruction and present AmodalDINO, a multi-head dense-prediction model that predicts four masks from a single RGB image: visible leaf, amodal complete leaf, amodal main vein, and fine veins. Unlike essentially all prior amodal work, AmodalDINO is given no visible mask. It predicts the visible and amodal regions jointly, so it needs no upstream instance segmenter at runtime. Two simple but effective changes adapt the model to the amodal segmentation task: fully fine-tune a DINOv3 ViT-L/16 at a small learning rate instead of freezing it, and attach auxiliary venation heads alongside the leaf heads. These two changes enable the model to learn the structural shape prior of leaves. Trained only on synthetic leaf fossil images, AmodalDINO reaches 95.0% Dice / 90.5% IoU on the validation set and transfers well to real fossil specimens. Stripped to two heads, the same recipe can run on two benchmark datasets, reaching 85.05 full mIoU / 66.65 occluded mIoU on KINS and 80.90 / 38.15 on COCOA-cls. The model is also practical: by quantizing to 4-bit weights, it runs entirely offline in a browser, matching the original model with an IoU of 0.910. We also add ruler-based calibration to estimate surface area, and a generative visualization of living leaves on local devices.

Chinese Translation

化石叶子很少能完整保存——沉积岩隐藏、破坏和侵蚀叶片的叶片，但古植物学依赖于叶子的完整形状和轮廓。我们将缺失组织的恢复视为无模态重建，并提出了AmodalDINO，这是一种多头密集预测模型，能够从单个RGB图像中预测四个掩膜：可见叶子、无模态完整叶子、无模态主脉和细脉。与几乎所有先前的无模态工作不同，AmodalDINO没有可见掩膜。它共同预测可见和无模态区域，因此在运行时不需要上游实例分割器。两个简单但有效的改动使模型适应无模态分割任务：以较小的学习率完全微调DINOv3 ViT-L/16，而不是冻结它，并在叶子头旁边附加辅助脉络头。这两个改动使模型能够学习叶子的结构形状先验。AmodalDINO仅在合成叶子化石图像上训练，在验证集上达到了95.0%的Dice / 90.5%的IoU，并且能够很好地迁移到真实化石标本上。将模型简化为两个头，相同的方案可以在两个基准数据集上运行，在KINS上达到85.05的完整mIoU / 66.65的遮挡mIoU，在COCOA-cls上达到80.90 / 38.15。该模型也具有实用性：通过量化为4位权重，它可以在浏览器中完全离线运行，其IoU与原始模型匹配，达到0.910。我们还增加了基于标尺的校准来估计表面积，并在本地设备上生成活叶子的可视化。

View on arXiv Download PDF AI Translation

cs.CV / 25 / 2608.04424

Thinking with Anchors: Grounded and Efficient Document Reasoning

以锚点思考：基于实证和高效的文档推理

Zhu, Sichen, Zhu, Yuchen, Xu, Wenzhuo, Kuen, Jason, Zhu, Wanrong, Shi, Jing, Shen, Xuan, Wang, Quanyi, Wang, Yiwei, Cai, Yujun, Shuai, Bing, Zhang, Qin, Chen, Yongxin, Liu, Shilong, Tao, Molei, Gu, Jiuxiang

Abstract

Existing document understanding benchmarks have largely focused on locating page elements, yet real-world document intelligence requires models to reason jointly about region semantics, spatial relations, and visual structure. We present ADOPD 2026, a reasoning-oriented extension of ADOPD that turns page decomposition into spatially grounded document understanding. ADOPD 2026 enriches page anchors inherited from ADOPD 2024 dataset with human-cleaned captions, semantic tags, and generated chain-of-thought (CoT) traces grounded to document regions. Instead of treating boxes, masks, and tags as independent supervision signals, we cast text blocks, visual entities, semantic labels, bounding boxes, and polygon masks as a shared vocabulary of visual anchors. This representation supports three connected capabilities. First, region-level semantic tagging asks models to identify document element types from both page context and local appearance, revealing long-tail semantic failures that standard layout benchmarks often hide. Second, unified vision-language grounding generates text regions and visual entities together with coordinates or polygonal outlines, transforming detection and segmentation outputs into structured anchors that can be reused by downstream reasoning systems. Third, current state-of-the-art models still struggle with dense counting tasks evaluated on DocCount, a benchmark derived from ADOPD 2026, highlighting the need for the Thinking-with-Anchors pipeline in document semantic understanding. By connecting page decomposition to verifiable visual-anchor reasoning, ADOPD 2026 provides a task framework that moves document understanding beyond localization toward anchor-grounded document intelligence.

Chinese Translation

现有的文档理解基准主要集中在定位页面元素上，而现实世界的文档智能要求模型能够共同推理区域语义、空间关系和视觉结构。我们提出了 ADOPD 2026，这是 ADOPD 的一个面向推理的扩展，将页面分解转变为空间上扎根的文档理解。ADOPD 2026 通过人类清理的标题、语义标签和与文档区域相关的生成思维链（Chain-of-Thought, CoT）痕迹，丰富了从 ADOPD 2024 数据集中继承的页面锚点。我们不再将框、掩码和标签视为独立的监督信号，而是将文本块、视觉实体、语义标签、边界框和多边形掩码视为共享的视觉锚点词汇。这种表示支持三种相互关联的能力。首先，区域级语义标记要求模型从页面上下文和局部外观中识别文档元素类型，揭示了标准布局基准常常隐藏的长尾语义失败。其次，统一的视觉-语言锚定同时生成文本区域和视觉实体以及坐标或多边形轮廓，将检测和分割输出转化为可被下游推理系统重用的结构化锚点。第三，目前最先进的模型在基于 ADOPD 2026 派生的 DocCount 基准上仍然在密集计数任务中面临挑战，这突显了在文档语义理解中需要 Thinking-with-Anchors 流程。通过将页面分解与可验证的视觉锚点推理连接起来，ADOPD 2026 提供了一个任务框架，使文档理解超越定位，朝着基于锚点的文档智能迈进。

View on arXiv Download PDF AI Translation

cs.CV / 26 / 2608.04426

Predict, Then Retrieve: Cross-Instance Future-State Retrieval from Video Prefixes

预测后检索：来自视频前缀的跨实例未来状态检索

Vo, Quynh, Nguyen, Thong, Do, Vinh-Hien, Nguyen, Cong-Duy, Luu, Anh-Tuan

Abstract

We introduce Predictive State Retrieval (PSR), a task in which a model observes a short video prefix and a temporal question about an object's future state, then retrieves instances from other videos or images that depict that state. Unlike action anticipation, which predicts a label, moment retrieval, which localizes an observed event within a video, or video generation, which synthesizes pixels, PSR combines anticipation with cross-instance retrieval across multiple temporal horizons. We construct a benchmark from four datasets with graded, human-validated ground truth, difficulty tiers, and an oracle ceiling. We also propose LFTR, a lightweight retriever with frozen encoders that predicts a question- and horizon-conditioned future latent and matches it in complementary semantic and visual spaces. A ceiling decomposition reveals a clear bottleneck: the true future state is highly retrievable once specified, whereas every predictor we evaluate, including a large multimodal language model with access to the prefix frames, remains far below the oracle. Thus, forecasting rather than perception is the central learnable challenge. LFTR narrows this gap at substantially lower inference cost, and ablations attribute its gains to cross-space fusion and hard-negative training rather than latent rollout. We release the benchmark, code, and evaluation scripts.

Chinese Translation

我们提出了预测状态检索（Predictive State Retrieval, PSR），这是一个任务，其中模型观察一个短视频前缀和关于物体未来状态的时间问题，然后从其他视频或图像中检索出描绘该状态的实例。与动作预期（action anticipation）预测标签、时刻检索（moment retrieval）在视频中定位观察到的事件或视频生成（video generation）合成像素不同，PSR结合了预期与跨实例检索，涵盖多个时间范围。我们从四个数据集中构建了一个基准，这些数据集具有分级的、经过人工验证的真实标签、难度等级和一个理想上限（oracle ceiling）。我们还提出了LFTR，这是一种轻量级检索器，具有冻结的编码器，能够预测基于问题和时间范围的未来潜在状态，并在互补的语义和视觉空间中进行匹配。上限分解揭示了一个明显的瓶颈：一旦指定，真实的未来状态是高度可检索的，而我们评估的每个预测器，包括一个可以访问前缀帧的大型多模态语言模型，仍然远低于理想上限。因此，预测而非感知是中心可学习的挑战。LFTR在显著较低的推理成本下缩小了这一差距，消融实验将其提升归因于跨空间融合和困难负样本训练，而非潜在展开。我们发布了基准、代码和评估脚本。

View on arXiv Download PDF AI Translation

cs.CV / 27 / 2608.04429

UBLLIE: Unified Backlight and Low-Light Image Enhancement

UBLLIE：统一的背光和低光图像增强

Yasin, Yasmin, Usman, Muhammad, Radwan, Ibrahim, Anwar, Saeed

Abstract

Backlit and low-light images often suffer from severe exposure imbalance or global underexposure, presenting significant challenges for both visual perception and downstream computer vision tasks. In this paper, we propose a unified, unsupervised enhancement framework that addresses both types of degradation without relying on paired ground-truth data. Our approach builds on CLIP-guided prompt learning to semantically supervise enhancement using learned positive and negative textual prompts. To improve the quality of our improvements over prior work, we design a symmetric residual U-Net backbone augmented with an Atrous Spatial Pyramid Pooling module. This architecture captures multi-scale contextual information, enabling adaptive correction under spatially heterogeneous illumination. During training, the enhancement network is guided by CLIP-based semantic similarity losses and refined via an iterative prompt optimization mechanism. Extensive experiments on both paired and unpaired datasets, including BAID, Backlit300, LOL, and VE-LOL-L, demonstrate that our framework consistently outperforms state-of-the-art supervised and unsupervised methods in terms of fidelity, perceptual quality, and generalization. Furthermore, our work emphasizes the need for stronger benchmarking protocols for backlit enhancement, a relatively underexplored area. The proposed framework provides a robust, scalable solution for real-world illumination enhancement across diverse lighting conditions.

Chinese Translation

背光和低光图像常常面临严重的曝光不平衡或全局欠曝光，给视觉感知和下游计算机视觉任务带来了显著挑战。本文提出了一种统一的无监督增强框架，旨在解决这两种退化问题，而无需依赖配对的真实数据。我们的方法基于CLIP引导的提示学习，通过学习的正向和负向文本提示进行语义监督增强。为了提高我们相对于先前工作的改进质量，我们设计了一个对称残差U-Net骨干网络，并增强了一个Atrous空间金字塔池化模块。该架构捕捉多尺度上下文信息，使得在空间异质照明下能够进行自适应校正。在训练过程中，增强网络通过基于CLIP的语义相似性损失进行指导，并通过迭代提示优化机制进行精炼。在BAID、Backlit300、LOL和VE-LOL-L等配对和未配对数据集上的大量实验表明，我们的框架在保真度、感知质量和泛化能力方面始终优于最先进的监督和无监督方法。此外，我们的工作强调了对背光增强这一相对未被充分探索领域的更强基准协议的需求。所提出的框架为在多种照明条件下的现实世界照明增强提供了一个稳健、可扩展的解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 28 / 2608.04434

OmniRouting: A Semantic-Coupled Multimodal Benchmark for Constraint-Aware Spatial Reasoning in PCB Routing

OmniRouting：一种语义耦合的多模态基准，用于PCB布线中的约束感知空间推理

Lu, Taiting, Lin, Kaiyuan, Dong, Ziwei, Bei, Sisong, Ye, Haolin, Tian, Yuxin, Liu, Runze, Wang, Mingjia, Zeng, Jingying, Pan, Hongxing, Zhang, Kai, Wang, Haoyu, Shi, Guoliang, Ma, Ling, Yang, Yifan, Lu, Jiaying, He, Qi, Chen, Yi-Chao, Chen, Sung-Liang, Jin, Yincheng, Gowda, Mahanth

Abstract

Recent large language models (LLMs) have demonstrated remarkable progress in constraint-aware navigation, maze reasoning, and graph reasoning. However, their ability to reason about complex routing problems under strict geometric, topological, and electrical constraints remains largely unexplored, despite routing being one of the most challenging and critical stages of electronic design automation (EDA). To bridge this gap, we introduce OmniRouting, the first large-scale benchmark designed to evaluate LLMs on printed-circuit-board (PCB) routing reasoning under real-world industrial design-rule, manufacturability, and connectivity constraints. OmniRouting contains 1,681 industrial-grade schematic-coupled PCB designs, including board geometries, routable component placements by human engineers, footprints, pad locations, netlists, stackup information, and routing constraints. The benchmark comprises four tasks: (1) geometric routing reasoning, generating physically valid copper traces, vias, and layer assignments to connect circuit nets within constrained board regions; (2) design-rule-aware routing reasoning, producing routable layouts that satisfy clearance, trace-width, via, obstacle-avoidance, and board-boundary constraints; (3) electrical functionality reasoning, preserving schematic-specified connectivity while reasoning over net names and functional roles to produce electrically correct routing; and (4) tool-augmented agentic routing, leveraging external tools for tasks (1)-(3). Our results reveal substantial limitations of current LMMs in PCB routing, including weak path-planning capabilities, poor adherence to design-rule constraints, and inconsistent preservation of electrical functionality. We will open-source all benchmark data, evaluation code, and tool interfaces to facilitate future research.

Chinese Translation

最近的大型语言模型（LLMs）在约束感知导航、迷宫推理和图形推理方面取得了显著进展。然而，尽管布线是电子设计自动化（EDA）中最具挑战性和关键的阶段之一，它们在严格的几何、拓扑和电气约束下推理复杂布线问题的能力仍然未得到充分探索。为了解决这一问题，我们提出了OmniRouting，这是第一个旨在评估LLMs在印刷电路板（PCB）布线推理方面的规模庞大的基准，考虑了现实工业设计规则、可制造性和连通性约束。OmniRouting包含1,681个工业级的原理图耦合PCB设计，包括电路板几何形状、人类工程师布置的可布线组件位置、封装、焊盘位置、网表、堆叠信息和布线约束。该基准包括四个任务：（1）几何布线推理，生成物理上有效的铜迹、过孔和层分配，以连接受限电路板区域内的电路网；（2）设计规则感知布线推理，生成满足间隙、迹线宽度、过孔、障碍物规避和电路板边界约束的可布线布局；（3）电气功能推理，在推理网名和功能角色的同时，保持原理图指定的连通性，以生成电气上正确的布线；（4）工具增强的自主布线，利用外部工具完成任务（1）-（3）。我们的结果揭示了当前LLMs在PCB布线方面的重大局限性，包括路径规划能力弱、对设计规则约束的遵循不佳以及电气功能保持不一致。我们将开源所有基准数据、评估代码和工具接口，以促进未来的研究。

View on arXiv Download PDF AI Translation

cs.CV / 29 / 2608.04436

ToolArtist: Tool-Using Unified Multimodal Models for Agentic Image Generation

ToolArtist：用于自主图像生成的工具使用统一多模态模型

Zhao, Jiahao, Yu, Xiaomin, Sun, Zhongxiang, Teng, Fengwei, Qin, Chengwei, Hu, Xiaobin, Xu, Jun, Yan, Shuicheng

Abstract

Text-to-image (T2I) models can produce visually compelling images, yet they remain limited on open-world tasks that require complex semantic understanding, multi-step reasoning, and the integration of external world knowledge. Existing efforts introduce agent capabilities into image generation, but they either prescribe a fixed workflow or place only a subset of the open-world image generation process under agent control. Consequently, reasoning, tool invocation, and image generation are not coordinated by a single policy. We propose ToolArtist, a fully agentic image generation model obtained by post-training a Unified Multimodal Model (UMM). ToolArtist dynamically orchestrates reasoning, external tool use, and native image generation within one unified policy. During Supervised Fine-Tuning (SFT), we equip a teacher agent with search tools alongside an image-generation tool. We then convert the collected trajectories into a UMM compatible format, where the image-generation tool is concealed while the resulting generated images are retained. During Reinforcement Learning (RL), we develop an agentic RL infrastructure for UMMs and introduce Reason-Act-Draw GRPO (RAD-GRPO), which uses complementary intent and quality rewards to jointly optimize the model. Experiments show that placing the entire open-world image-generation process under an agent policy consistently outperforms approaches with fixed pipelines or only partially agent-controlled components. We release the training data and the complete post-training infrastructure.

Chinese Translation

文本到图像（T2I）模型能够生成视觉上引人注目的图像，但在需要复杂语义理解、多步推理和整合外部世界知识的开放世界任务中仍然存在局限。现有的研究努力将代理能力引入图像生成，但它们要么规定了固定的工作流程，要么仅将开放世界图像生成过程的一个子集置于代理控制之下。因此，推理、工具调用和图像生成并未由单一策略协调。我们提出了ToolArtist，这是一种通过对统一多模态模型（UMM）进行后训练而获得的完全自主的图像生成模型。ToolArtist在一个统一的策略中动态协调推理、外部工具使用和原生图像生成。在监督微调（SFT）期间，我们为教师代理配备了搜索工具和图像生成工具。然后，我们将收集到的轨迹转换为UMM兼容格式，其中图像生成工具被隐藏，而生成的图像得以保留。在强化学习（RL）期间，我们为UMM开发了自主RL基础设施，并引入了Reason-Act-Draw GRPO（RAD-GRPO），该方法使用互补的意图和质量奖励共同优化模型。实验表明，将整个开放世界图像生成过程置于代理策略下的效果始终优于采用固定管道或仅部分由代理控制的组件的方法。我们发布了训练数据和完整的后训练基础设施。

View on arXiv Download PDF AI Translation

cs.CV / 30 / 2608.04441

Season: Spectrum-Aware Orthogonal Gradient Refinement for Transfer-Based Adversarial Attacks

Season：一种针对基于转移的对抗攻击的谱感知正交梯度细化方法

Wang, Tianyi, Gao, Zhenghao, Xu, Shengjie

Abstract

Transfer-based adversarial attacks often transfer poorly across heterogeneous architectures because CNNs favor local textures while Vision Transformers (ViTs) rely on global shapes. We propose Season, a spectrum-aware orthogonal gradient refinement framework for L-infinity transfer attacks against black-box target models on ImageNet, using a white-box surrogate. Season decomposes each update into a low-frequency branch capturing structural cues and a high-frequency branch capturing textures. A low-saliency guidance scheme reallocates high-frequency energy to background regions, preserving foreground structures that ViTs depend on. An orthogonal projection then forces the textural update to lie in the orthogonal complement of the structural direction, mitigating feature interference. As a training-free plug-and-play wrapper, Season enhances eight gradient-stabilization and input-enhancement attacks without modifying their cores. Across eight CNN, ViT, and MLP targets, Season improves transfer success rate by 6.6 percentage points on average and up to 16.0 points over strong baselines under a unified protocol.

Chinese Translation

基于转移的对抗攻击在异构架构间的转移效果往往较差，因为卷积神经网络（CNN）更倾向于局部纹理，而视觉变换器（ViTs）则依赖于全局形状。我们提出了Season，一种针对ImageNet上黑箱目标模型的L-infinity转移攻击的谱感知正交梯度细化框架，使用白箱替代模型。Season将每次更新分解为一个捕捉结构线索的低频分支和一个捕捉纹理的高频分支。一个低显著性引导方案将高频能量重新分配到背景区域，从而保留ViTs所依赖的前景结构。随后，正交投影强制纹理更新位于结构方向的正交补空间中，从而减轻特征干扰。作为一种无训练的即插即用封装，Season在不修改核心的情况下增强了八种梯度稳定和输入增强攻击。在八个CNN、ViT和MLP目标上，Season平均提高了6.6个百分点的转移成功率，并在统一协议下对强基线的提升最高可达16.0个百分点。

View on arXiv Download PDF AI Translation

cs.CV / 31 / 2608.04448

When does training on downscaled images yield the same gradients?

何时在降采样图像上训练会产生相同的梯度？

Ji, Seunghyun

Abstract

Diffusion transformers deliver strong image generation, but their training cost grows superlinearly with resolution. Recent work justifies training or sampling at reduced resolution on a spectral premise: at high noise, a downscaled latent preserves almost the full surviving signal. Whether a downscaled step also preserves the native training gradient signal, however, has remained unresolved. We reduce how that signal changes under downscaling to two terms: a noise-dependent term governed by the downscale ratio, which decays at high noise as the spectral premise predicts, and a \sigma-independent floor governed by the target grid's absolute token count, carried by the compute graph itself and removed by no noise level. The measured (route, \sigma) map corroborates the account and uncovers structure the spectral picture cannot express: on the 1024->768 route, a window (0.65 < \sigma < 0.95), predicted by no spectral criterion at any tolerance, where the downscaled gradient stays within a small margin of the native one. Training LoRA adapters with downscaled steps restricted to the routes and noise windows the map validates reduces training time by 14.6% at a fixed step budget while remaining near-native in weight space. Code is available at https://github.com/sorryhyun/anima_lora.

Chinese Translation

扩散变换器在图像生成方面表现出色，但其训练成本随着分辨率的增加呈超线性增长。近期的研究基于光谱前提证明了在降低分辨率下进行训练或采样的合理性：在高噪声下，降采样的潜在空间几乎保留了完整的生存信号。然而，降采样步骤是否也保留了原生训练梯度信号仍未得到解决。我们将该信号在降采样下的变化归结为两个项：一个由降采样比例控制的噪声依赖项，该项在高噪声下衰减，正如光谱前提所预测的那样；另一个是由目标网格的绝对标记数量控制的与 C3 无关的底线，该底线由计算图本身承载，并且在任何噪声水平下都无法去除。测量的（路径，C3）图验证了这一解释，并揭示了光谱图无法表达的结构：在 1024->768 路径上，存在一个窗口（0.65 < C3 < 0.95），在任何容忍度下都未被光谱标准预测，在该窗口内，降采样梯度保持在原生梯度的小范围内。使用降采样步骤训练 LoRA 适配器，限制在图验证的路径和噪声窗口内，可以在固定步骤预算下将训练时间减少 14.6%，同时在权重空间中保持接近原生状态。代码可在 https://github.com/sorryhyun/anima_lora 获取。

View on arXiv Download PDF AI Translation

cs.CV / 32 / 2608.04452

Q-CueGraph: Query-Conditioned Visual Evidence Graphs for Multimodal Reasoning

Q-CueGraph：用于多模态推理的查询条件视觉证据图

Pan, Pengcheng, Zhang, Xinfang

Abstract

High-resolution pixels and crop or zoom tools give multimodal large language models the ability to inspect an image, but they do not provide a reliable task-conditioned policy for deciding where to inspect. Q-CueGraph makes this decision explicit. It maps a question and an image representation to budgeted, coordinate-level observations for a frozen reader. Text-rich images use a reusable OCR/layout graph; natural-image search instantiates query-conditioned visual nodes behind the same selection, composition, and budgeting interface. Optional utility refinement learns which candidate crops the frozen reader can use from training-answer correctness, without region-box supervision. With a frozen Qwen2.5-VL-7B reader, Q-CueGraph reaches 0.833 accuracy on V*Bench versus 0.696 for full-image inference from a 19% image-area budget, and reaches 92% of full-image ANLS on InfographicVQA from about half the image area. Across six benchmarks, explicit observation is most valuable when evidence is localizable, the question discriminates its location, and resolution limits full-image reading.

Chinese Translation

高分辨率像素和裁剪或缩放工具赋予多模态大型语言模型检查图像的能力，但它们并未提供可靠的任务条件策略来决定检查的位置。Q-CueGraph 使这一决策变得明确。它将问题和图像表示映射到预算化的坐标级观察，以供冻结的阅读器使用。文本丰富的图像使用可重用的 OCR/布局图；自然图像搜索在相同的选择、组合和预算界面后实例化查询条件视觉节点。可选的效用优化通过训练答案的正确性学习冻结阅读器可以使用的候选裁剪，而无需区域框的监督。在冻结的 Qwen2.5-VL-7B 阅读器下，Q-CueGraph 在 V*Bench 上达到了 0.833 的准确率，而全图推理的准确率为 0.696，预算为 19% 的图像区域，并在 InfographicVQA 上达到了全图 ANLS 的 92%，仅使用了大约一半的图像区域。在六个基准测试中，当证据可定位、问题区分其位置且分辨率限制全图阅读时，明确观察的价值最大。

View on arXiv Download PDF AI Translation

cs.CV / 33 / 2608.04453

TwinIR: Coordinated Invisible Dual-Point Attacks on Online HD Map Construction

TwinIR：在线高清地图构建中的协调隐形双点攻击

Hu, Haibo, Deng, Jianghuai, Tang, Chen, Lou, Yang, Xu, Qian, Wang, Jianping

Abstract

Online HD map construction is critical to prediction and planning in autonomous driving. We find that existing physical attacks against online map construction are limited by a cross-boundary compensation effect: after the target boundary is perturbed, another visible boundary may retain sufficient geometric cues for the model to recover the original road geometry. Based on this observation, we propose TwinIR, a new mechanism-guided physical attack methodology for online map construction. TwinIR jointly optimizes attack effectiveness and point sparsity, seeking the minimum number of attack points needed to suppress compensating geometric cues from surrounding boundaries. To reduce the perceptibility of multi-point attacks, TwinIR models camera responses to near-infrared illumination and maps optimized attack points to feasible physical placements, producing camera-visible interference with minimal visible-spectrum changes. Experiments on nuScenes across state-of-the-art online map construction models show that TwinIR reduces mAP by 8.18-8.96 percentage points under RSA and 2.84-5.62 points under ETA, while increasing the unreachable-goal rate by 25-28 points and the unsafe-planned-trajectory rate by 19-20 points over clean inputs. These attacks are also validated on a real-world testbed AV, where TwinIR successfully induces both road straightening and early-turn deformations while remaining inconspicuous in full-color views.

Chinese Translation

在线高清地图构建对自动驾驶中的预测和规划至关重要。我们发现，现有针对在线地图构建的物理攻击受到跨边界补偿效应的限制：在目标边界受到干扰后，另一个可见边界可能保留足够的几何线索，使模型能够恢复原始道路几何形状。基于这一观察，我们提出了TwinIR，一种新型的机制引导物理攻击方法，旨在在线地图构建中优化攻击效果和点稀疏性，寻求抑制周围边界补偿几何线索所需的最小攻击点数量。为了降低多点攻击的可感知性，TwinIR对近红外照明下的相机响应进行建模，并将优化的攻击点映射到可行的物理位置，产生在可见光谱变化最小的情况下可见的干扰。在nuScenes数据集上对最先进的在线地图构建模型进行的实验表明，TwinIR在RSA下将mAP降低了8.18-8.96个百分点，在ETA下降低了2.84-5.62个百分点，同时使不可达目标率提高了25-28个百分点，安全规划轨迹率提高了19-20个百分点，相较于干净输入。这些攻击还在真实世界测试平台的自动驾驶车辆上得到了验证，TwinIR成功引发了道路拉直和早转变形，同时在全彩视图中保持不显眼。

View on arXiv Download PDF AI Translation

cs.CV / 34 / 2608.04454

Beyond Global Routing Aggregation: Phase-Aware Expert Merging for MoE Vision-Language Models

超越全局路由聚合：面向阶段的专家合并用于混合专家视觉-语言模型

Zhang, Hongyu, Yan, Cheng, Xia, Xiang, Zhang, Wuyang

Abstract

Mixture-of-experts vision-language models (MoE-VLMs) increase model capacity with sparse expert activation, yet deployment requires storing the full expert pool. Training-free expert merging reduces this burden, and many routing-based methods aggregate routing statistics across all tokens to determine merge compatibility. However, MoE-VLM inference is phase-structured: image-context tokens carry visual content, question tokens specify the query, and answer tokens produce the output, with different counts and routing distributions. Because image-context tokens are far more numerous, global aggregation can overemphasize image-context processing and obscure phase-conditioned expert roles, making experts serving different phases appear interchangeable and degrading model performance. We therefore argue that MoE-VLM expert merging should preserve phase-conditioned expert roles, judging compatibility by how experts serve different phases rather than globally aggregated routing statistics. Based on this view, we propose RoleMerge, a training-free method that constructs each expert's Routing Role Profile (RRP) from phase-normalized routing statistics, capturing its relative phase preference. Guided by expert-phase information loss, RoleMerge merges experts with compatible profiles and their corresponding router entries while preserving answer-decoding expert distinctions. Experiments on three models and multiple benchmarks show that RoleMerge preserves more of the full model's performance than alternative expert-merging methods at matched expert-retention ratios, with relative improvements of up to 9.6 percent in six-task macro-average performance. These results validate phase-conditioned expert roles as a more effective basis than global routing aggregation for MoE-VLM expert merging.

Chinese Translation

混合专家视觉-语言模型（MoE-VLMs）通过稀疏的专家激活来增加模型容量，但部署时需要存储完整的专家池。无训练的专家合并减少了这一负担，许多基于路由的方法通过聚合所有标记的路由统计信息来确定合并兼容性。然而，MoE-VLM推理是阶段结构化的：图像上下文标记携带视觉内容，问题标记指定查询，答案标记生成输出，且它们的数量和路由分布各不相同。由于图像上下文标记数量远多于其他标记，全局聚合可能会过于强调图像上下文处理，模糊阶段条件下的专家角色，使得服务不同阶段的专家看起来可互换，从而降低模型性能。因此，我们认为MoE-VLM专家合并应当保留阶段条件下的专家角色，通过专家服务不同阶段的方式来判断兼容性，而不是依赖全局聚合的路由统计信息。基于这一观点，我们提出了RoleMerge，一种无训练的方法，通过阶段归一化的路由统计信息构建每个专家的路由角色配置文件（RRP），捕捉其相对阶段偏好。在专家-阶段信息损失的指导下，RoleMerge合并具有兼容配置文件及其相应路由器条目的专家，同时保留答案解码专家的区别。在三个模型和多个基准上的实验表明，RoleMerge在匹配的专家保留比例下，保留了更多完整模型的性能，相较于其他专家合并方法，在六项任务的宏平均性能上有高达9.6%的相对提升。这些结果验证了阶段条件下的专家角色作为MoE-VLM专家合并的一个比全局路由聚合更有效的基础。

View on arXiv Download PDF AI Translation

cs.CV / 35 / 2608.04472

EndoVLM: An Endoscopy Vision-Language Pre-training Model via Anatomy-Guided Sparsity and Progressive Alignment

EndoVLM：一种通过解剖引导稀疏性和渐进对齐的内窥镜视觉-语言预训练模型

Yi, Zhenyu, Xu, Jianwei, Hu, Yue, Qiu, Zhongwei, Li, Sijing, Huang, Liang, Lv, Bin, Zhang, Ling, Xia, Yingda

Abstract

The development of foundation models (FMs) is crucial for advancing endoscopic image analysis. However, existing endoscopy FMs mainly rely on self-supervised learning from uni-modal images or videos, overlooking the rich semantic knowledge contained in clinical reports. Furthermore, effectively leveraging these records is hindered by a fundamental modality gap: structured anatomical descriptions are not naturally mapped to specific frames within the high-redundancy, uncurated visual streams. In this paper, we present EndoVLM, a novel vision-language FM pre-trained on over 348K endoscopic examinations, each pairing a clinical report with its corresponding image collection. An Anatomy-Guided Sparse Pooling mechanism utilizes textual descriptions as queries to drive sparse attention, efficiently aggregating semantically salient frames into anatomy-specific visual representations across redundant image-sets. Next, a Progressive Semantic-Aware Alignment strategy models clinical taxonomy (anatomy and pathological status) via structured soft targets, bridging the gap from global patient-level matching to fine-grained localized alignment. Finally, a Semantic-Concentrated Masked Autoencoder is applied exclusively to these semantic-rich frames, integrating low-level visual precision with robust high-level semantic representation. Extensive experiments across various downstream tasks demonstrate that EndoVLM outperforms existing foundation models and remains competitive with task-specific methods. Remarkably, EndoVLM also exhibits robust zero-shot generalization capabilities, highlighting its potential for broader clinical application.

Chinese Translation

基础模型（FMs）的发展对推进内窥镜图像分析至关重要。然而，现有的内窥镜FMs主要依赖于从单一模态图像或视频中进行自监督学习，忽视了临床报告中蕴含的丰富语义知识。此外，有效利用这些记录受到基本模态差距的阻碍：结构化的解剖描述并未自然映射到高冗余、未经整理的视觉流中的特定帧。在本文中，我们提出了EndoVLM，一种新颖的视觉-语言基础模型，预训练于超过348K个内窥镜检查，每个检查都将临床报告与其相应的图像集合配对。解剖引导稀疏池化机制利用文本描述作为查询来驱动稀疏注意力，有效地将语义显著的帧聚合为特定解剖的视觉表示，跨越冗余的图像集。接下来，渐进语义感知对齐策略通过结构化软目标建模临床分类法（解剖和病理状态），弥合从全局患者级匹配到细粒度局部对齐的差距。最后，语义集中掩码自编码器专门应用于这些语义丰富的帧，将低级视觉精度与强大的高级语义表示相结合。在各种下游任务中的广泛实验表明，EndoVLM在性能上优于现有的基础模型，并在任务特定方法中保持竞争力。值得注意的是，EndoVLM还表现出强大的零样本泛化能力，突显了其在更广泛临床应用中的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 36 / 2608.04480

REZE: Recognition-Based Zero-Shot Extraction for Video Temporal Grounding

REZE：基于识别的零-shot视频时序定位提取

Li, Boyang, Gou, Chenhui, Cai, Jianfei

Abstract

Video temporal grounding (VTG) refers to the task of identifying the time interval in a video that corresponds to a given natural-language query. A common zero-shot strategy asks a large vision-language model (VLM) to generate the start and end timestamps directly, so the result depends heavily on the design and training of the model, and grounding accuracy differs widely from one VLM to another. We therefore propose REcognition-based Zero-shot Extraction (REZE), a simple training-free method that splits the video into short clips, asks the model for a clip-level confidence score for the query, and uses a deterministic algorithm to convert the resulting score curve into the output required by the task. Because temporal aggregation is performed outside the model, REZE adapts to different task outputs, from single- and multi-interval moment retrieval to highlight detection. On QVHighlights, REZE improves the best reported training-free moment-retrieval mAP from 38.23 to 40.32, while on highlight detection it reaches 44.18 mAP and 73.41 HIT@1, establishing a new state of the art among training-free methods. Its HIT@1 also outperforms all fully supervised SoTAs on the QVHighlights test split. We evaluate REZE on seven backbones from three model families. On Charades-STA and QVHighlights, it outperforms direct timestamp generation in every available comparison. We further observe that with REZE an earlier-generation model can approach the native performance of a newer model in its family.

Chinese Translation

视频时序定位（VTG）是指识别视频中与给定自然语言查询相对应的时间区间的任务。一种常见的零-shot策略是要求一个大型视觉-语言模型（VLM）直接生成开始和结束时间戳，因此结果在很大程度上依赖于模型的设计和训练，不同VLM之间的定位准确性差异很大。因此，我们提出了基于识别的零-shot提取（REZE），这是一种简单的无训练方法，它将视频分割成短片段，要求模型为查询提供片段级置信度分数，并使用确定性算法将结果分数曲线转换为任务所需的输出。由于时间聚合是在模型外部进行的，REZE能够适应不同的任务输出，从单一和多区间时刻检索到高亮检测。在QVHighlights数据集上，REZE将最佳报告的无训练时刻检索mAP从38.23提高到40.32，而在高亮检测中达到了44.18 mAP和73.41 HIT@1，确立了无训练方法中的新状态。此外，其HIT@1也超越了QVHighlights测试集上所有完全监督的最新技术水平（SoTAs）。我们在三个模型家族的七个主干网络上评估了REZE。在Charades-STA和QVHighlights上，它在每个可用比较中都优于直接时间戳生成。我们进一步观察到，使用REZE时，早期生成的模型可以接近其家族中新模型的原生性能。

View on arXiv Download PDF AI Translation

cs.CV / 37 / 2608.04483

Not All Redundant Tokens Are Alike: Analyzing Visual Token Pruning through Token Roles

并非所有冗余标记都是相同的：通过标记角色分析视觉标记剪枝

Kim, Hyeonyu, Lim, Sehwan, Choi, Youngwon, Kwon, Taeyoun, Kim, Jaejin

Abstract

Vision-language models (VLMs) process an image as a sequence of visual tokens, which creates a substantial computational bottleneck during inference. Recent visual token pruning methods address this issue by removing seemingly redundant tokens, yet it remains unclear how these pruning decisions relate to the functional roles of visual tokens. In this work, we analyze visual token pruning through the lens of token roles identified by EmbedLens. We first show that representative pruning methods exhibit distinct token-role biases, but these biases do not directly correlate with downstream performance. To better understand this behavior, we refine the token-role assignment procedure and evaluate role-protected pruning variants. Our results show that preserving non-alive tokens can sometimes maintain or improve performance, suggesting that tokens with weak direct semantic alignment may still affect model behavior under pruning. Our code is publicly available at https://github.com/jaykim9870/Not_All_Redundant_Tokens_Are_Alike.

Chinese Translation

视觉语言模型（VLMs）将图像处理为一系列视觉标记，这在推理过程中造成了显著的计算瓶颈。近期的视觉标记剪枝方法通过移除看似冗余的标记来解决这一问题，但这些剪枝决策与视觉标记的功能角色之间的关系仍不明确。在本研究中，我们通过 EmbedLens 确定的标记角色的视角分析视觉标记剪枝。我们首先展示了代表性的剪枝方法表现出不同的标记角色偏差，但这些偏差与下游性能并不直接相关。为了更好地理解这种行为，我们优化了标记角色分配程序，并评估了角色保护剪枝变体。我们的结果表明，保留非活跃标记有时可以维持或提高性能，这表明具有弱直接语义对齐的标记在剪枝过程中仍可能影响模型行为。我们的代码已公开发布在 https://github.com/jaykim9870/Not_All_Redundant_Tokens_Are_Alike。

View on arXiv Download PDF AI Translation

cs.CV / 38 / 2608.04496

DIVE: Dynamic Iterative Visual Evidence Construction for Efficient Vision-Language Models

DIVE：用于高效视觉-语言模型的动态迭代视觉证据构建

Zhong, Chen, An, Xiao, Wang, Zijie, Li, Jiepan, Yang, Guangyi, He, Wei

Abstract

Visual inputs in vision-language models (VLMs) are often encoded into substantially longer token sequences than text, making visual tokens a major bottleneck for efficient inference. Abundant recent methods address this bottleneck by scoring token importance and pruning low-scoring tokens in a single pass. However, one-shot scoring is insufficient because a token's prompt-relevant usefulness depends on the evidence already retained. Motivated by this insight, we introduce DIVE (Dynamic Iterative Visual Evidence Construction), a training-free framework that recasts visual-token pruning as dynamic evidence construction. DIVE repeatedly selects the remaining token with the highest residual-conditioned score, updates the visual and prompt residuals to discount the evidence already explained, and re-evaluates the remaining tokens. This select-update-re-evaluate process builds a retained set of complementary, prompt-relevant evidence. Experiments across eight image-understanding benchmarks show that DIVE consistently preserves performance across token budgets. With an 88.9% reduction in visual tokens, DIVE retains 98.2% of the uncompressed model's average performance. Code is available at https://github.com/Zhong-Chenchen/DIVE.git.

Chinese Translation

在视觉-语言模型（VLMs）中，视觉输入通常被编码为比文本长得多的标记序列，这使得视觉标记成为高效推理的主要瓶颈。近期大量方法通过对标记重要性进行评分并在单次传递中修剪低评分标记来解决这一瓶颈。然而，单次评分是不够的，因为标记与提示相关的有用性依赖于已保留的证据。基于这一见解，我们提出了DIVE（动态迭代视觉证据构建），这是一个无训练的框架，将视觉标记修剪重新构建为动态证据构建。DIVE反复选择具有最高残差条件评分的剩余标记，更新视觉和提示残差以折扣已解释的证据，并重新评估剩余标记。这个选择-更新-重新评估的过程构建了一个保留的互补、与提示相关的证据集。在八个图像理解基准上的实验表明，DIVE在标记预算范围内始终保持性能。在视觉标记减少88.9%的情况下，DIVE保留了未压缩模型平均性能的98.2%。代码可在 https://github.com/Zhong-Chenchen/DIVE.git 获取。

View on arXiv Download PDF AI Translation

cs.CV / 39 / 2608.04501

Privacy-Preserving Action Recognition: Taxonomy, Methods, and Privacy-Utility Trade-offs

隐私保护的动作识别：分类、方法及隐私-效用权衡

Amin, Sareer Ul, Ayaz, Muhammad, Munsif, Muhammad, Seo, Sanghyun

Abstract

Video surveillance in public safety, healthcare, and smart environments has made continuous human monitoring routine, raising real risks to personal identity and appearance. Privacy-preserving action recognition (PPAR) tackles the tension between the utility of video understanding and this exposure, and has drawn fast-growing interest. However, existing surveys remain narrow. Most catalog a single mechanism family, predate recent adversarial and hybrid work, or barely address evaluation. The result is a fragmented literature with incompatible threat models, inconsistent metrics, and no shared evaluation standard. We address this with a PRISMA-guided review of 32 peer-reviewed papers (2018--2026) drawn from 885 screened records. Methods sort into five families, namely adversarial learning (52%), skeleton-based (20%), cryptographic (12%), differential privacy (8%), and hybrid (8%), each with distinct privacy, utility, and efficiency trade-offs. Evaluation is the weak point. Only 10% of papers adopt a formal privacy definition, 65% rely on ad-hoc metrics, and 40% report an inconsistently defined cMAP. The trade-offs are steep. Skeleton methods reach about 85% accuracy but drop appearance, adversarial methods hold near 80% utility at moderate privacy (cMAP 0.9 to 0.3--0.5), and differential privacy often falls below 70%. Harder conditions stay under-tested, with fewer than 15% of papers checking cross-dataset generalization, under 10% testing adaptive attackers, and real-time edge deployment nearly untouched. We contribute a two-dimensional privacy-space taxonomy, a formal threat model, a comparative trade-off analysis, the PPAR Unified Evaluation Protocol, and a roadmap centered on benchmark standardization. With this grounding, we argue PPAR can move from prototypes toward deployment, with lessons extending to face recognition and medical imaging.

Chinese Translation

公共安全、医疗保健和智能环境中的视频监控使得持续的人类监测成为常态，带来了对个人身份和外貌的真实风险。隐私保护的动作识别（Privacy-Preserving Action Recognition, PPAR）解决了视频理解的效用与这种暴露之间的紧张关系，并引起了快速增长的关注。然而，现有的调查仍然较为狭窄。大多数文献仅列举单一机制家族，未能涵盖近期的对抗性和混合工作，或者几乎没有涉及评估。结果是文献碎片化，存在不兼容的威胁模型、不一致的指标，以及缺乏共享的评估标准。我们通过PRISMA指导的回顾，分析了32篇经过同行评审的论文（2018--2026），这些论文来自885个筛选记录。方法分为五个家族，即对抗学习（52%）、基于骨架的方法（20%）、密码学方法（12%）、差分隐私（8%）和混合方法（8%），每种方法在隐私、效用和效率上都有不同的权衡。评估是薄弱环节。只有10%的论文采用了正式的隐私定义，65%依赖于临时指标，40%报告了不一致定义的cMAP。权衡相当严峻。骨架方法的准确率约为85%，但放弃了外观，对抗方法在中等隐私下保持近80%的效用（cMAP 0.9至0.3--0.5），而差分隐私通常低于70%。更严格的条件仍然缺乏测试，检查跨数据集泛化的论文少于15%，测试自适应攻击者的不到10%，实时边缘部署几乎没有涉及。我们贡献了一个二维隐私空间分类法、一个正式的威胁模型、一个比较权衡分析、PPAR统一评估协议，以及一个以基准标准化为中心的路线图。在此基础上，我们认为PPAR可以从原型向部署转变，所获得的经验教训也可扩展到人脸识别和医学成像领域。

View on arXiv Download PDF AI Translation

cs.CV / 40 / 2608.04504

GeoReward: Mitigating Contextual Variable Overestimation in Vision-Language Models for Cross-Market Preference Prediction

GeoReward：缓解视觉语言模型在跨市场偏好预测中的上下文变量过高估计

Liu, Shuo, Cai, Huixiang, Zhang, Weiru, Zeng, Xiaoyi

Abstract

Vision-language models excel in many multimodal tasks but remain prone to a subtle yet impactful failure mode: they tend to overestimate dominant visual-textual cues while underestimating sparse but decision-critical contextual variables. This issue, which we term Contextual Variable Overestimation (CVE), becomes particularly evident in real-world applications such as predicting advertisement image preferences across diverse geographic markets. For instance, when a VLM is asked to choose between two product images tailored for different countries, it often defaults to a consistent output, ignoring ground-truth regional variations. This collapse occurs because pervasive high-volume signals, such as product attributes and dense image patches, overwhelm the few but critical tokens that encode market-specific context. To address CVE, we first collect a new multimodal dataset of real advertising creatives and their click-through performance across multiple countries. We then introduce GeoReward, a reward model designed to predict ad image preferences across diverse geographic markets. GeoReward integrates three purpose-built mechanisms: (1) Market-Aware Retrieval Augmentation, (2) Context-Guided Visual Modulation, (3) Selective Sensitivity Loss. Furthermore, we demonstrate how GeoReward can guide the fine-tuning of RL for a VLM to generate background designs for text-to-image models, producing market-aware advertising creatives. Experiments validate that our framework mitigates CVE and outperforms existing baselines. This work not only diagnoses a systematic bias in VLMs toward dominant perceptual features but also delivers a targeted solution for applications where sparse contextual variables govern decision-making.

Chinese Translation

视觉语言模型在许多多模态任务中表现出色，但仍然容易出现一种微妙而影响深远的失败模式：它们往往高估主导的视觉-文本线索，同时低估稀疏但对决策至关重要的上下文变量。我们将这一问题称为上下文变量过高估计（Contextual Variable Overestimation, CVE），在现实世界应用中尤为明显，例如在不同地理市场中预测广告图像偏好。当视觉语言模型（VLM）被要求在针对不同国家的两张产品图像之间进行选择时，它通常会默认输出一致的结果，忽视真实的区域差异。这种崩溃发生是因为普遍存在的高频信号，如产品属性和密集图像块，压倒了少数但关键的编码市场特定上下文的标记。为了解决CVE问题，我们首先收集了一个新的多模态数据集，包含真实广告创意及其在多个国家的点击率表现。然后，我们引入了GeoReward，一个旨在预测不同地理市场广告图像偏好的奖励模型。GeoReward整合了三种专门构建的机制：（1）市场感知检索增强（Market-Aware Retrieval Augmentation），（2）上下文引导的视觉调制（Context-Guided Visual Modulation），（3）选择性敏感性损失（Selective Sensitivity Loss）。此外，我们展示了GeoReward如何指导强化学习（RL）对VLM的微调，以生成文本到图像模型的背景设计，从而制作市场感知的广告创意。实验验证了我们的框架能够减轻CVE并超越现有基准。这项工作不仅诊断了VLM在主导感知特征方面的系统性偏差，还提供了针对稀疏上下文变量主导决策的应用的针对性解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 41 / 2608.04515

CARVE: Cross-Slice Anisotropic Reallocation of Visual Evidence for Efficient 3D Medical Volume Understanding

CARVE：用于高效理解3D医学体积的跨切片各向异性视觉证据重新分配

Yi, Zhenyu, Hu, Qiang, Li, Zhenhao, Zhao, Jiaxuan, Sun, Yusong, Zhang, Lichi

Abstract

Slice-based MLLMs leverage mature 2D encoders by representing 3D volumes as sequences of 2D slices. However, this slice-wise formulation produces thousands of visual tokens that burden the LLM backbone, many of which capture overlapping visual evidence across adjacent slices. To understand how effectively a growing visual token budget improves performance, we perform scaling analyses on two 3D medical VQA benchmarks and find diminishing returns: cost keeps rising while accuracy saturates, and improving in-plane resolution is more effective than adding slices at comparable budgets. The budget should therefore be allocated more selectively rather than simply enlarged, yet most token compression methods are designed for 2D images or videos, where redundancy arises from spatial layout or temporal motion rather than from near-duplicate content along the depth axis. We present CARVE, a training-free framework that compresses visual tokens prior to LLM inference and casts token reduction as budget-constrained 2.5D allocation. CARVE partitions the depth axis into coherent windows and allocates tokens non-uniformly according to normalized cross-slice evidence. Under a shared budget, CARVE builds spatial anchors on representative slices and retrieves locally varying evidence from the full volume, then merges remaining eligible tokens into nearby anchors within each window. Removing roughly 80% of the visual tokens on Hulu-Med-7B, CARVE leads all compression baselines on every AMOS-MM report-generation metric, with 6.2 points higher retention of full-token quality than the strongest baseline, and preserves 98.1% of full-token performance across three VQA benchmarks.

Chinese Translation

基于切片的多模态大语言模型（MLLMs）通过将3D体积表示为2D切片序列，利用成熟的2D编码器。然而，这种切片式的表述产生了数千个视觉标记，给大语言模型（LLM）主干带来了负担，其中许多标记捕捉了相邻切片之间重叠的视觉证据。为了理解不断增长的视觉标记预算如何有效提升性能，我们对两个3D医学视觉问答（VQA）基准进行了规模分析，发现收益递减：成本持续上升而准确率饱和，改善平面分辨率比在相似预算下增加切片更有效。因此，预算应更具选择性地分配，而不仅仅是简单扩大，然而大多数标记压缩方法是为2D图像或视频设计的，其中冗余来自空间布局或时间运动，而不是来自深度轴上的近重复内容。我们提出了CARVE，一个无训练框架，在LLM推理之前压缩视觉标记，并将标记减少视为预算受限的2.5D分配。CARVE将深度轴划分为一致的窗口，并根据归一化的跨切片证据非均匀分配标记。在共享预算下，CARVE在代表性切片上构建空间锚点，并从完整体积中检索局部变化的证据，然后将剩余的合格标记合并到每个窗口内的附近锚点中。在Hulu-Med-7B上去除约80%的视觉标记，CARVE在每个AMOS-MM报告生成指标上领先所有压缩基线，保留的全标记质量比最强基线高出6.2个百分点，并在三个VQA基准上保持98.1%的全标记性能。

View on arXiv Download PDF AI Translation

cs.CV / 42 / 2608.04525

Coupled Continuous-Discrete Generation for Scene Text Image Super-Resolution

场景文本图像超分辨率的耦合连续-离散生成

Niu, Axi, Zhang, Knag, Yan, Qingsen, Jin, Hao, Sun, Jinqiu, Zhang, Yanning

Abstract

Scene text image super-resolution (STISR) aims to recover visually plausible appearance while preserving character semantics from degraded inputs. Existing STISR systems often rely on externally generated priors or separate image and text models, resulting in error propagation and costly multi-stage inference. We present DualTSR, a unified framework that formulates STISR as coupled continuous-discrete generation. Conditional flow matching restores continuous image latents, while absorbing-state discrete diffusion reconstructs text tokens. Both processes share a multimodal transformer backbone, allowing the evolving image and text states to interact throughout generation without an external OCR prior at inference. On CTR-TSR, DualTSR achieves the best FID, LPIPS, ACC, and NED among the compared methods at both X2 and X4. On an aligned RealCE subset, it obtains the best FID, ACC, and NED with competitive LPIPS. Compared with DiffTSR at X4, DualTSR improves ACC by 12.78 percentage points while reducing the parameter count from 1.23B to 203M and end-to-end latency from 13.3s to 132ms. These results establish DualTSR as an accurate and efficient method for STISR.

Chinese Translation

场景文本图像超分辨率（STISR）旨在从退化输入中恢复视觉上合理的外观，同时保留字符语义。现有的 STISR 系统通常依赖于外部生成的先验或分离的图像和文本模型，导致错误传播和成本高昂的多阶段推理。我们提出了 DualTSR，一个将 STISR 形式化为耦合连续-离散生成的统一框架。条件流匹配恢复连续图像潜变量，而吸收态离散扩散重建文本标记。这两个过程共享一个多模态变换器骨干网络，使得图像和文本状态在生成过程中能够相互作用，而无需在推理时依赖外部 OCR 先验。在 CTR-TSR 数据集上，DualTSR 在 X2 和 X4 的比较方法中均取得了最佳的 FID、LPIPS、ACC 和 NED。在对齐的 RealCE 子集上，它获得了最佳的 FID、ACC 和 NED，并且 LPIPS 竞争力强。与 X4 的 DiffTSR 相比，DualTSR 将 ACC 提高了 12.78 个百分点，同时将参数数量从 1.23B 降低到 203M，端到端延迟从 13.3 秒减少到 132 毫秒。这些结果确立了 DualTSR 作为一种准确且高效的 STISR 方法。

View on arXiv Download PDF AI Translation

cs.CV / 43 / 2608.04530

FocusMem: Factorizing Content, Readout, and Trust in Latent GUI Memory

FocusMem：在潜在图形用户界面记忆中分解内容、读取和信任

Zhang, Zhuoran, Li, Bowen, Ju, Jingcheng, Shi, Yang, Wang, Qixun, Wang, Haotian, Chen, Wei, Wang, Tengjiao

Abstract

GUI agents must remember both useful experience from earlier tasks and unfinished progress in the current interaction. Latent memory offers a compact solution by compressing multimodal trajectories into a few continuous tokens. Existing methods, however, usually map each trajectory to one fixed memory block and train it mainly through next-action supervision. This creates three practical problems: important details may be lost during compression, the same memory block must serve different decision stages, and irrelevant retrieved trajectories may still mislead the agent. We introduce FocusMem, which separates these responsibilities within a compact latent-memory interface. A role-aware content basis encourages episodic memory to retain reusable experience and working memory to retain task progress. A state-conditioned readout generates a decision-specific view of the same stored evidence, while a lightweight trust gate can suppress memory blocks that appear irrelevant to the current step. All components are trained while the GUI policy remains frozen. Across five GUI-agent benchmarks, FocusMem consistently outperforms a fully matched action-only fixed-memory baseline and prior latent memory adaptations. Further analysis shows that semantic and functional supervision preserve complementary information, state-conditioned readout is more robust as surrounding trajectory context grows, and the trust gate reduces the harm caused by injected irrelevant episodic evidence. These results show that effective latent memory depends not only on compressing past interaction, but also on what is retained, what is exposed, and what is allowed.

Chinese Translation

图形用户界面（GUI）代理必须记住早期任务中的有用经验和当前交互中的未完成进展。潜在记忆通过将多模态轨迹压缩为少量连续的标记，提供了一种紧凑的解决方案。然而，现有方法通常将每个轨迹映射到一个固定的记忆块，并主要通过下一步行动的监督进行训练。这造成了三个实际问题：在压缩过程中可能会丢失重要细节，同一个记忆块必须服务于不同的决策阶段，以及检索到的无关轨迹可能仍会误导代理。我们引入了FocusMem，它在紧凑的潜在记忆接口中分离了这些责任。角色感知的内容基础鼓励情节记忆保留可重用的经验，而工作记忆则保留任务进展。状态条件读取生成同一存储证据的决策特定视图，而轻量级信任门可以抑制与当前步骤无关的记忆块。所有组件在GUI策略保持不变的情况下进行训练。在五个GUI代理基准测试中，FocusMem始终优于完全匹配的仅行动固定记忆基线和先前的潜在记忆适应。进一步分析表明，语义和功能监督保留了互补信息，状态条件读取在周围轨迹上下文增加时更具鲁棒性，而信任门减少了注入无关情节证据造成的危害。这些结果表明，有效的潜在记忆不仅依赖于压缩过去的交互，还依赖于保留什么、暴露什么以及允许什么。

View on arXiv Download PDF AI Translation

cs.CV / 44 / 2608.04533

EgoAfford: Task-Oriented Affordance Grounding via Egocentric Referring Segmentation

EgoAfford：通过自我中心引用分割实现任务导向的可供性定位

Guan, Xinyuan, Chen, Feifan, Zhan, Xinyu, Zhang, Fu-Cheng, Lu, Cewu, Yang, Lixin

Abstract

Part-level affordance grounding has advanced the localization of functional object regions associated with elemental actions. Extending this capability to complex tasks calls for connecting the semantic roles of participating objects with task-state-aligned visual observations and multi-step planning. We introduce EgoAfford, a benchmark designed to connect these three aspects. Given an egocentric observation and a high-level tabletop task, a model must generate the remaining plan and segment the functional regions of up to three components of the next action: the direct object, instrument, and destination. EgoAfford comprises approximately 15.5k human-verified images from 2,000 generated multi-step scenes, organized as semantically aligned, task-complete image series, together with EgoAfford-Real, 102 manually captured images spanning 26 tasks. We further present EgoLens, a 3B multimodal large language model with role-specific mask decoders, as an in-domain reference model for this joint task. Evaluations of recent referring-segmentation MLLMs, commercial-VLM--SAM2 pipelines, and EgoLens highlight the complementary challenges of next-step inference and action-role-conditioned part grounding. EgoLens establishes strong reference performance on both generated and manually captured observations. Together, EgoAfford and EgoLens provide a foundation for jointly studying perception and planning in multi-step tabletop tasks. Our project page is available at: https://egoafford.github.io

Chinese Translation

部分级别的可供性定位促进了与基本动作相关的功能性物体区域的定位。将这一能力扩展到复杂任务需要将参与物体的语义角色与任务状态对齐的视觉观察和多步骤规划相连接。我们提出了EgoAfford，这是一个旨在连接这三个方面的基准。给定一个自我中心的观察和一个高层次的桌面任务，模型必须生成剩余的计划并分割下一动作的最多三个组成部分的功能区域：直接物体、工具和目的地。EgoAfford包含大约15.5k经过人工验证的图像，来自2000个生成的多步骤场景，组织为语义对齐的、任务完整的图像序列，以及EgoAfford-Real，涵盖26个任务的102张手动捕获图像。我们进一步提出了EgoLens，一个具有角色特定掩码解码器的3B多模态大型语言模型，作为该联合任务的领域内参考模型。对近期引用分割的多模态大型语言模型（MLLMs）、商业视觉语言模型（VLM）-SAM2管道和EgoLens的评估突出了下一步推理和基于动作角色的部分定位的互补挑战。EgoLens在生成和手动捕获的观察上均建立了强有力的参考性能。EgoAfford和EgoLens共同为在多步骤桌面任务中联合研究感知和规划提供了基础。我们的项目页面可访问：https://egoafford.github.io

View on arXiv Download PDF AI Translation

cs.CV / 45 / 2608.04557

VoxStruct3D: Structure-Leading Flow Matching for Voxel-Space 3D MRI Synthesis

VoxStruct3D：基于结构导向的体素空间3D MRI 合成流匹配

Li, Fang, Gao, Yang, Zou, Shihao, Si, Weixin, Wu, Hongyu, Xia, Qing, Li, Shuai, Hao, Aimin

Abstract

High-fidelity 3D MRI synthesis requires both globally coherent anatomy and fine-grained voxel-level detail. Although latent diffusion makes volumetric generation tractable, its image autoencoder introduces a reconstruction bottleneck that can limit the fine detail recoverable in the final volume. We present VoxStruct3D, a voxel-space flow-matching framework that directly models full-resolution MRI volumes using a clean-data prediction objective. Its Volumetric Voxel Generator (VVG) combines factorized 3D patch embedding with overlapping upsampling, time-modulated residual refinement, and skip fusion, enabling neighboring tokens to jointly reconstruct shared voxel regions and suppress patch-boundary artifacts. To complement direct voxel-space modeling with an explicit anatomical prior, we further introduce a Structure-First, Image-Follows (SFIF) strategy. A frozen pretrained 3D medical encoder and a StructVAE extract compact structure tokens that preserve dominant anatomy, while a structure-leading schedule keeps their trajectory ahead of the image trajectory. Patch-Aligned RoPE spatially aligns the unequal token grids, and asymmetric attention enforces one-way guidance from structure to image. Experiments on pathological and healthy T1-weighted brain MRI datasets show that VoxStruct3D achieves the strongest overall performance across feature-distribution alignment, sample diversity, and perceptual quality, producing anatomically coherent and visually realistic volumes.

Chinese Translation

高保真3D MRI 合成需要全球一致的解剖结构和细粒度的体素级细节。尽管潜在扩散使得体积生成变得可行，但其图像自编码器引入了重建瓶颈，限制了最终体积中可恢复的细节。我们提出了VoxStruct3D，一个体素空间流匹配框架，直接使用干净数据预测目标建模全分辨率MRI体积。其体积体素生成器（Volumetric Voxel Generator, VVG）结合了分解的3D补丁嵌入、重叠上采样、时间调制的残差细化和跳跃融合，使得相邻的标记能够共同重建共享的体素区域，并抑制补丁边界伪影。为了用明确的解剖先验补充直接的体素空间建模，我们进一步引入了结构优先、图像跟随（Structure-First, Image-Follows, SFIF）策略。一个冻结的预训练3D医学编码器和一个StructVAE提取紧凑的结构标记，以保留主要的解剖结构，而结构导向的调度使其轨迹领先于图像轨迹。补丁对齐的RoPE（Patch-Aligned RoPE）在空间上对齐不等的标记网格，而非对称注意力则强制实现从结构到图像的单向引导。在病理和健康的T1加权脑MRI数据集上的实验表明，VoxStruct3D在特征分布对齐、样本多样性和感知质量方面实现了最强的整体性能，生成了解剖一致且视觉真实的体积。

View on arXiv Download PDF AI Translation

cs.CV / 46 / 2608.04559

ColorFD: A Finite-Difference Guided Black-Box Physical Adversarial Attack for Remote Sensing Object Detection

ColorFD：一种基于有限差分指导的黑箱物理对抗攻击用于遥感目标检测

Guo, Tiannuo, Qiu, Guhang, Xie, Yuzhen, Feng, Rui, Li, Ligang, Xiang, Deliang

Abstract

Although deep neural network-based remote sensing object detectors have achieved strong performance, they remain vulnerable to adversarial perturbations. Existing studies mainly focus on digital or white-box settings, whereas black-box physical attacks remain underexplored. These attacks are often constrained by limited physical feasibility and inefficient optimization in high-dimensional search spaces. To address these challenges, this paper proposes ColorFD, a black-box physical attack based on multiple pure-color patches. The patch positions and color parameters are jointly optimized using Differential Evolution (DE). A target-wise fitness and selection mechanism evaluates the attack state of each target and preserves target-specific improvements during evolution. Two guidance strategies further constrain the patch search space. Key-region localization identifies sensitive regions through finite-difference color probing. Common-feature extraction provides category-level spatial priors and avoids repeated localization. Although evaluated on aircraft, the formulation is not inherently restricted to this category. Experiments on YOLOv3u, YOLOv5u, and Faster R-CNN show that ColorFD outperforms the tested black-box patch method across all evaluated detectors and remains competitive with strong white-box baselines. Physical-world experiments further demonstrate that the optimized pure-color patches can be transferred from the digital domain to real imaging conditions.

Chinese Translation

尽管基于深度神经网络的遥感目标检测器已取得了强大的性能，但它们仍然容易受到对抗扰动的影响。现有研究主要集中在数字或白箱设置上，而黑箱物理攻击仍然未得到充分探索。这些攻击通常受到有限物理可行性和高维搜索空间中低效优化的限制。为了解决这些挑战，本文提出了ColorFD，一种基于多个纯色补丁的黑箱物理攻击。补丁的位置和颜色参数通过差分进化（Differential Evolution, DE）进行联合优化。目标特定的适应度和选择机制评估每个目标的攻击状态，并在进化过程中保留目标特定的改进。两种指导策略进一步约束了补丁搜索空间。关键区域定位通过有限差分颜色探测识别敏感区域。共同特征提取提供类别级空间先验，避免重复定位。尽管在飞机上进行了评估，但该公式并不固有地限制于此类别。在YOLOv3u、YOLOv5u和Faster R-CNN上的实验表明，ColorFD在所有评估的检测器中均优于所测试的黑箱补丁方法，并与强大的白箱基线保持竞争力。物理世界实验进一步表明，优化后的纯色补丁可以从数字领域转移到真实成像条件。

View on arXiv Download PDF AI Translation

cs.CV / 47 / 2608.04560

OutLangSplat: 3D Language Gaussian Splatting for UAV Outdoor Scenes

OutLangSplat：用于无人机户外场景的3D语言高斯溅射

Yan, Xia, Wu, He, Xu, Yanghui, Wu, Zizhao, Chen, Jiazhou

Abstract

3D Language Gaussian Splatting embeds open-vocabulary language features into 3D Gaussian Splatting, providing an efficient explicit representation for text-driven 3D scene understanding. However, existing methods are limited to indoor or small-scale scenes, and tend to fail in Unmanned Aerial Vehicle (UAV) outdoor scenes, where severe occlusions and long distance viewpoints often lead to incorrect semantic activations and missing target responses. In this paper, we present OutLangSplat which adapts language Gaussian representations to UAV outdoor scenes by improving feature representation and aggregation reliability. For the feature representation, a 2D-3D dual-branch representation with region-based alignment and fusion is designed to improve spatial consistency, reducing incomplete target responses and background misactivations. For the feature aggregation, we introduce a training-free contribution and consistency-aware Gaussian feature aggregation strategy that leverages pixel contribution reliability and cross-view semantic consistency to suppress unreliable responses from noisy viewpoints. A new dataset is provided by manually annotating various objects on four real-world public UAV outdoor scene datasets. To the best of our knowledge, it is the first accessible dataset of open-vocabulary 3D scene understanding for UAV outdoor scenes. Quantitative evaluations and ablation studies demonstrate that OutLangSplat outperforms SOTA methods on both open-vocabulary semantic segmentation and instance localization tasks. The datasets and codes will be open-sourced.

Chinese Translation

3D语言高斯溅射将开放词汇的语言特征嵌入到3D高斯溅射中，为基于文本的3D场景理解提供了一种高效的显式表示。然而，现有方法仅限于室内或小规模场景，往往在无人机（UAV）户外场景中失败，因为严重的遮挡和远距离视角常常导致错误的语义激活和缺失的目标响应。本文提出了OutLangSplat，它通过改善特征表示和聚合可靠性，将语言高斯表示适应于无人机户外场景。在特征表示方面，设计了一种具有基于区域的对齐和融合的2D-3D双分支表示，以提高空间一致性，减少不完整的目标响应和背景误激活。在特征聚合方面，我们引入了一种无训练的贡献和一致性感知的高斯特征聚合策略，该策略利用像素贡献的可靠性和跨视角的语义一致性来抑制来自噪声视角的不可靠响应。我们通过手动标注四个真实世界公共无人机户外场景数据集中的各种对象，提供了一个新的数据集。据我们所知，这是第一个可公开访问的开放词汇3D场景理解的无人机户外场景数据集。定量评估和消融研究表明，OutLangSplat在开放词汇语义分割和实例定位任务上优于现有的最先进方法（SOTA）。数据集和代码将开源。

View on arXiv Download PDF AI Translation

cs.CV / 48 / 2608.04568

Talk2Sensors: 3D Visual Grounding in Autonomous Driving via Sensor-Adaptive Physical Cue Matching

Talk2Sensors：通过传感器自适应物理线索匹配实现自主驾驶中的3D视觉定位

Guan, Runwei, Tian, Di, Ouyang, Ningwei, Zhang, Ruixiao, Liang, Shaofeng, Zhao, Haocheng, Zheng, Lianqing, Bai, Xiaokai, Wang, Guotao, Liu, Daizong, Ding, Henghui, Xiong, Hui

Abstract

As a key capability for embodied intelligence, 3D visual grounding (3DVG) has been predominantly studied in indoor scenes with RGB-D or point-cloud inputs, while existing outdoor extensions largely rely on monocular images alone. Both settings fall short of real-world outdoor perception, where heterogeneous sensors capture complementary yet distinct physical properties, such as visual texture, 3D geometry, and object kinematics, that are indispensable for flexible and robust query-adaptive grounding but remain under-exploited. To bridge this gap, we introduce Talk2Sensors, the first multi-sensor 3D visual grounding dataset built upon camera, LiDAR, and 4D radar. It contains 8,682 language instructions and 20,558 referred objects, with diverse prompts explicitly aligned with sensor-specific physical cues. Furthermore, we propose TSFormer, a unified Transformer-based framework for language-guided 3D visual grounding in autonomous driving. TSFormer adopts a coarse-to-fine property-aware fusion strategy: the Language-Routed Property Sampler first performs coarse text-conditioned feature retrieval by modulating sensor sampling weights with query-level linguistic cues, while the subsequent Sparse-Preserving Modality Arbiter module conducts fine-grained modality arbitration and text-guided refinement to determine the precise referred spatial location. This design enables dynamic routing of appearance, geometry, and motion cues according to the semantic requirements of each prompt, preventing dense modalities from overwhelming sparse but critical sensor signals. Extensive experiments demonstrate that TSFormer achieves state-of-the-art performance across multiple benchmarks: it improves over the strongest baseline by 8.05 mAP on Talk2Sensors, and transfers to the monocular Mono3DRefer benchmark with 53.05\% [email protected].

Chinese Translation

作为具身智能的关键能力，3D视觉定位（3DVG）主要在室内场景中使用RGB-D或点云输入进行研究，而现有的户外扩展大多仅依赖单目图像。这两种设置都未能充分反映现实世界的户外感知，其中异构传感器捕获互补但不同的物理属性，如视觉纹理、3D几何形状和物体运动学，这些属性对于灵活和稳健的查询自适应定位至关重要，但尚未得到充分利用。为了解决这一问题，我们提出了Talk2Sensors，这是第一个基于相机、激光雷达和4D雷达构建的多传感器3D视觉定位数据集。该数据集包含8,682条语言指令和20,558个参考对象，具有与传感器特定物理线索明确对齐的多样化提示。此外，我们提出了TSFormer，一个基于Transformer的统一框架，用于自主驾驶中的语言引导3D视觉定位。TSFormer采用粗到细的属性感知融合策略：语言引导属性采样器首先通过调节传感器采样权重与查询级语言线索进行粗略的文本条件特征检索，而后续的稀疏保留模态仲裁模块则进行细粒度的模态仲裁和文本引导的精细化，以确定精确的参考空间位置。该设计使得外观、几何和运动线索能够根据每个提示的语义需求进行动态路由，防止密集模态淹没稀疏但关键的传感器信号。大量实验表明，TSFormer在多个基准测试中实现了最先进的性能：在Talk2Sensors上比最强基线提高了8.05 mAP，并在单目Mono3DRefer基准上转移到53.05\% [email protected]。

View on arXiv Download PDF AI Translation

cs.CV / 49 / 2608.04575

PhysMind: From Video to Executable Worlds for Training-Free Physical Reasoning

PhysMind：从视频到可执行世界的无训练物理推理

Yang, Chen, Zeng, Shenxiang, Zhao, Haoyang, Xu, Zhouyuan, He, Youquan, Li, Haoyu, Deng, Mingyi, Fan, Jiansheng, Wang, Chen

Abstract

Reliable physical reasoning from video requires understanding how objects move, interact, and respond to interventions. Existing vision-language models (VLMs) often struggle to interpret these dynamics and reason reliably about future and counterfactual outcomes. We introduce PhysMind, a training-free agentic framework that constructs one reusable, question-agnostic executable world per video. PhysMind recovers a temporally consistent dynamic scene through object segmentation, mesh reconstruction, and 6D pose tracking, then fits analytic continuous-time dynamics and latent physical parameters without unrolling a time-stepped simulator. Given a question, it inspects, continues, or edits the world and answers from the resulting trajectories and interactions. Relative to direct chain-of-thought (CoT) reasoning with the same VLM, PhysMind improves accuracy by 38.23 points on CLEVRER and 8.08 points on Physion++. On counterfactual questions, it exceeds the strongest evaluated VLM baseline, GPT-5.5, by 19.25 points.

Chinese Translation

从视频中进行可靠的物理推理需要理解物体如何移动、相互作用以及对干预的响应。现有的视觉-语言模型（VLMs）往往难以解释这些动态，并且在推理未来和反事实结果时缺乏可靠性。我们提出了PhysMind，一个无训练的代理框架，为每个视频构建一个可重用的、与问题无关的可执行世界。PhysMind通过物体分割、网格重建和6D姿态跟踪恢复一个时间一致的动态场景，然后在不展开时间步模拟器的情况下拟合解析的连续时间动态和潜在物理参数。给定一个问题，它检查、继续或编辑世界，并根据生成的轨迹和交互进行回答。相较于使用相同VLM进行直接的思维链（CoT）推理，PhysMind在CLEVRER上提高了38.23分，在Physion++上提高了8.08分。在反事实问题上，它超过了最强评估的VLM基线GPT-5.5，提升了19.25分。

View on arXiv Download PDF AI Translation

cs.CV / 50 / 2608.04581

ACA-GS: Adaptive-Capacity Anchored Gaussian Splatting for Compact Dynamic Radiance Fields

ACA-GS：基于自适应容量的锚定高斯点云用于紧凑动态辐射场

Song, Seunghyeon, Lee, Joo Chan, Park, Chanung, Jeong, Jun Young, Lee, Minseo, Park, Eunbyung, Ko, Jong Hwan

Abstract

Recent advances in 4D Gaussian Splatting (4DGS) enable high-fidelity, real-time spatiotemporal rendering, but expose a fundamental trade-off between motion expressiveness and storage efficiency. While anchor-based designs achieve compactness through anchor-level parameter sharing, their rigid uniform parametrization enforces fixed Neural Gaussian counts and feature budgets per anchor. Consequently, insufficient fidelity is addressed by excessive anchor density, rather than lightweight, targeted increases in Neural Gaussian count or feature capacity, resulting in memory waste. To overcome this rigidity, we introduce an adaptive-capacity anchor-based framework that dynamically allocates the representational capacity based on local spatiotemporal demands. Adaptive Anchor Cardinality varies the number of Neural Gaussians per anchor, concentrating primitives in regions of high geometric or motion complexity while suppressing redundancy. In parallel, Adaptive Anchor Feature Masking modulates anchor-level feature channels, assigning rich features to complex regions and lightweight representations to simpler ones. Experiments on MPEG, Panoptic Sports, and N3DV datasets demonstrate substantial storage reduction without degrading visual quality. Notably, on challenging MPEG sequences with complex motion, our method achieves up to 1.5x higher compression than state-of-the-art anchor-based methods while preserving comparable quality.

Chinese Translation

最近在4D高斯点云（4DGS）方面的进展使得高保真、实时时空渲染成为可能，但暴露了运动表现力与存储效率之间的基本权衡。虽然基于锚点的设计通过锚点级参数共享实现了紧凑性，但其刚性的统一参数化强制每个锚点固定的神经高斯数量和特征预算。因此，无法满足的保真度通过过度的锚点密度来解决，而不是通过轻量级、针对性的增加神经高斯数量或特征容量，从而导致内存浪费。为了解决这种刚性问题，我们提出了一种基于自适应容量的锚点框架，该框架根据局部时空需求动态分配表示容量。自适应锚点基数（Adaptive Anchor Cardinality）变化每个锚点的神经高斯数量，在几何或运动复杂度高的区域集中原始元素，同时抑制冗余。同时，自适应锚点特征掩码（Adaptive Anchor Feature Masking）调节锚点级特征通道，将丰富的特征分配给复杂区域，而将轻量级表示分配给简单区域。在MPEG、全景体育（Panoptic Sports）和N3DV数据集上的实验表明，在不降低视觉质量的情况下实现了显著的存储减少。值得注意的是，在具有复杂运动的挑战性MPEG序列上，我们的方法实现了比最先进的基于锚点的方法高出1.5倍的压缩，同时保持了相当的质量。

View on arXiv Download PDF AI Translation

cs.CV / 51 / 2608.04587

MetaVideoAgent: Automated Video-Agent Evolution for Long-Form Video Understanding

MetaVideoAgent：用于长视频理解的自动化视频代理演化

Cui, Benlei, Wang, Ruize, Li, Junjie, Chen, Jinhao, Huang, Longtao, Chen, Yinghao, Zhai, Yuwen, Tang, Jingqun, Jia, Ruijian, Wu, Weiwei, Sun, Pengfei, Hong, Haiwen

Abstract

Long-form video understanding requires locating sparse, question-relevant evidence in long, multimodal videos. Real-world video distributions differ in modality-specific information density, content structure, and evidence patterns, causing fixed video-agent designs to incur redundant processing or fail when mismatched. Extending automated agent evolution from text to video is challenging because full long-video execution makes candidate validation expensive, failures propagate across coupled evidence-processing stages, and complex preprocessing, perception tools, and localization strategies make code-level updates difficult to implement reliably. We introduce MetaVideoAgent, a framework that automatically evolves a video agent for a target distribution. It profiles information density and evidence requirements from sparsely sampled frames and associated queries to guide initial design, then compresses localized failures into independently executable minimal validation tasks. It constructs evidence-grounded Gold Paths, audits Student trajectories, aggregates recurring failures across samples, and attributes them to responsible modules. A modular agent representation constrains each update to the primary responsible module and its necessary dependencies. We further introduce VA-EvoBench, covering eight video distributions with separate evolution and held-out splits. With four evolution iterations per distribution, MetaVideoAgent improves every initial agent and raises macro-average accuracy from 38.44% to 51.47%, at an average evolution cost of 3.54M tokens per distribution. The evolved agents outperform the strongest prior fixed-design video agent by 6.39 percentage points while using the fewest tokens and video frames per question among the compared video agents. We will release all code and data to support reproducible research.

Chinese Translation

长视频理解需要在长篇多模态视频中定位稀疏的、与问题相关的证据。现实世界中的视频分布在模态特定的信息密度、内容结构和证据模式上存在差异，这导致固定的视频代理设计可能产生冗余处理或在不匹配时失败。将自动化代理演化从文本扩展到视频是具有挑战性的，因为完整的长视频执行使候选验证成本高昂，失败会在耦合的证据处理阶段传播，而复杂的预处理、感知工具和定位策略使得代码级更新难以可靠实施。我们提出了MetaVideoAgent，一个自动演化视频代理以适应目标分布的框架。它从稀疏采样的帧和相关查询中分析信息密度和证据需求，以指导初始设计，然后将局部失败压缩为独立可执行的最小验证任务。它构建基于证据的金路径（Gold Paths），审计学生轨迹（Student trajectories），聚合样本中的重复失败，并将其归因于相关模块。模块化的代理表示将每次更新限制在主要责任模块及其必要依赖项上。我们进一步介绍了VA-EvoBench，涵盖八种视频分布，具有独立的演化和保留分割。通过每种分布的四次演化迭代，MetaVideoAgent改进了每个初始代理，并将宏平均准确率从38.44%提高到51.47%，每种分布的平均演化成本为3.54M个标记。演化后的代理在使用最少的标记和每个问题的视频帧数方面，超越了最强的先前固定设计视频代理6.39个百分点。我们将发布所有代码和数据以支持可重复的研究。

View on arXiv Download PDF AI Translation

cs.CV / 52 / 2608.04589

The First EgoCross Challenge at EgoVis 2026: Cross-Domain Egocentric Video Question Answering

2026年EgoVis首届EgoCross挑战赛：跨领域自我中心视频问答

Fu, Yuqian, Qian, Tianwen, Li, Yanjun, Li, Yu, Peng, Kunyu, Zheng, Xu, Xian, Yongqin, Tonioni, Alessio, Fu, Yanwei, Wang, Xiaoling, Paudel, Danda, Tombari, Federico, Van Gool, Luc, Wu, Leyi, Zhao, Yifan, Zhang, Jinjie, Li, Yinchuan, Chen, Yingcong, Li, Zixu, Chen, Zhiwei, Fu, Zhiheng, Wang, Wenbo, Hu, Yupeng, Guan, Weili, Nie, Liqiang, Murakawa, Takuya, Tamaki, Toru, Wen, Yi, Du, Zhenglin, Li, Zhengyang, Li, Lingling, Jiao, Licheng, Ma, Wenping

Abstract

EgoCross is a cross-domain egocentric video question answering benchmark designed to evaluate whether multimodal large language models can generalize beyond common daily-life scenarios. The first EgoCross Challenge was hosted at the Third EgoVis Workshop at CVPR 2026 and evaluated models on first-person videos from four target domains: surgery, industrial assembly, extreme sports, and animal perspectives. Each test example consists of an egocentric video clip, a question, and four candidate answers, from which the model must select the correct option. This technical report introduces the challenge task, benchmark resources, and two official Codabench tracks. The Source-Limited Track restricts participants to the official baseline model and a small support set, whereas the Open-Source Track permits broader choices of models and training data under rules that prohibit the manual construction of target-domain training data. In total, the challenge received more than 1,500 submissions from over 130 participants, with 19 teams participating in the Open-Source Track and 38 teams in the Source-Limited Track. We further present the official leaderboard results and summarize the winning solutions from both tracks. We hope that this report will serve as a useful technical reference for advancing cross-domain egocentric video understanding. All resources, including the challenge data, baseline implementation, and code released by the winning teams, are made publicly available.

Chinese Translation

EgoCross是一个跨领域自我中心视频问答基准，旨在评估多模态大型语言模型是否能够超越常见日常生活场景进行泛化。首届EgoCross挑战赛在2026年CVPR第三届EgoVis研讨会上举办，评估模型在四个目标领域的第一人称视频上的表现：外科手术、工业组装、极限运动和动物视角。每个测试示例由一个自我中心视频片段、一个问题和四个候选答案组成，模型必须从中选择正确的选项。本技术报告介绍了挑战任务、基准资源以及两个官方Codabench赛道。源限制赛道要求参与者使用官方基线模型和一个小型支持集，而开放源代码赛道则允许在禁止手动构建目标领域训练数据的规则下选择更广泛的模型和训练数据。总的来说，挑战赛共收到来自130多名参与者的1500多个提交，其中开放源代码赛道有19个团队参与，源限制赛道有38个团队参与。我们进一步展示了官方排行榜结果，并总结了两个赛道的获胜解决方案。我们希望本报告能作为推进跨领域自我中心视频理解的有用技术参考。所有资源，包括挑战数据、基线实现及获胜团队发布的代码，均已公开提供。

View on arXiv Download PDF AI Translation

cs.CV / 53 / 2608.04604

COSMO: Consensus-Driven Shift Modulation for Source-Free Domain Adaptation

COSMO：基于共识驱动的偏移调制用于无源领域适应

Li, Bo, Peng, Junjie, Xie, Xiaohua, Lai, Jianhuang

Abstract

Source-free domain adaptation (SFDA) adapts a source-trained model to an unlabeled target domain without source data, a practical setting under privacy or storage constraints. Yet its self-generated supervision can reinforce source bias under substantial domain shifts. Pretrained vision-language models (VLMs) offer complementary semantic knowledge, but the relative reliability of the source model and VLM varies across target samples. Existing cross-model guidance does not explicitly account for this variation and may overwrite valid source-derived evidence under conflict, a failure we term source-derived evidence forgetting. We formulate VLM-guided SFDA as a sample-wise reliability-allocation problem and propose Consensus-Driven Shift Modulation (COSMO). COSMO replaces expert-to-expert guidance with co-adaptation through an anchored shared consensus. It first forms a sample-specific initial consensus that favors the more concentrated prediction. During adaptation, COSMO re-aggregates both branches' evolving evidence and regulates how far the resulting consensus moves from its initial anchor based on consensus uncertainty and training progress. This keeps the shared supervision anchored yet adaptive. Across four benchmarks, COSMO achieves state-of-the-art performance under matched VLM backbones. Further analyses indicate that it better balances the retention of valid source-derived evidence with the absorption of complementary VLM evidence.

Chinese Translation

无源领域适应（SFDA）在没有源数据的情况下，将源训练模型适应于未标记的目标领域，这在隐私或存储限制下是一种实际的设置。然而，其自生成的监督可能在显著的领域偏移下强化源偏差。预训练的视觉-语言模型（VLMs）提供了互补的语义知识，但源模型和VLM在目标样本中的相对可靠性存在差异。现有的跨模型指导并未明确考虑这种变化，可能在冲突情况下覆盖有效的源派生证据，这种失败我们称之为源派生证据遗忘。我们将VLM引导的SFDA形式化为样本级可靠性分配问题，并提出了共识驱动的偏移调制（COSMO）。COSMO用通过锚定共享共识的共同适应替代专家对专家的指导。它首先形成一个样本特定的初始共识，偏向于更集中的预测。在适应过程中，COSMO重新聚合两个分支的演变证据，并根据共识不确定性和训练进展调节结果共识从其初始锚点移动的距离。这保持了共享监督的锚定性，同时又具有适应性。在四个基准测试中，COSMO在匹配的VLM骨干下实现了最先进的性能。进一步分析表明，它更好地平衡了有效源派生证据的保留与互补VLM证据的吸收。

View on arXiv Download PDF AI Translation

cs.CV / 54 / 2608.04606

TRCoRSurg: Temporal-Relational Co-Reasoning for Surgical Video Triplet Recognition

TRCoRSurg：用于外科视频三元组识别的时间关系共同推理

Li, Fang, Zou, Shihao, Si, Weixin, Gao, Yang, Li, Shuai, Hao, Aimin

Abstract

Understanding complex surgical scenes requires recognizing multiple interdependent entities, such as instruments, actions, and targets, while maintaining their relational consistency across time. Existing surgical triplet recognition methods struggle to jointly model intra-frame label dependencies and inter-frame temporal semantics in a unified manner. To address these limitations, we propose a unified framework that integrates spatial, relational, and temporal cues for robust surgical triplet recognition. Specifically, class-specific spatial priors are first extracted through a multi-scale encoder. These priors are then refined by a Label Correlation Modeling module with multi-scale class activation map-guided relational extraction (MS-CAMRE), enabling the model to capture both static co-occurrence patterns and dynamic contextual dependencies among triplet components. Furthermore, a Bidirectional Temporal-Relational Fusion Attention (BTRFA) module harmonizes temporal and relational representations to achieve coherent temporal reasoning. We also introduce a new evaluation metric, the Triplet Consistency Error Rate (TCER), which quantitatively measures the model's ability to preserve causal and semantic consistency across triplets. Extensive experiments on the CholecT45 and ProstaTD datasets show that our method achieves state-of-the-art performance, improving AP_IVT by 5.1 percent and 7.8 percent, respectively. Moreover, according to TCER, our approach achieves relative reductions of more than 36 percent and 25 percent on the two datasets, respectively, demonstrating the effectiveness of our framework in temporal-relational co-reasoning.

Chinese Translation

理解复杂的外科场景需要识别多个相互依赖的实体，如工具、动作和目标，同时保持它们在时间上的关系一致性。现有的外科三元组识别方法在统一建模帧内标签依赖性和帧间时间语义方面存在困难。为了解决这些局限性，我们提出了一个统一框架，集成空间、关系和时间线索，以实现稳健的外科三元组识别。具体而言，首先通过多尺度编码器提取特定类别的空间先验。这些先验随后通过标签相关建模模块（Label Correlation Modeling）与多尺度类别激活图引导的关系提取（MS-CAMRE）进行精炼，使模型能够捕捉三元组组件之间的静态共现模式和动态上下文依赖。此外，一个双向时间关系融合注意力模块（BTRFA）协调时间和关系表示，以实现连贯的时间推理。我们还引入了一种新的评估指标，即三元组一致性错误率（Triplet Consistency Error Rate，TCER），该指标定量测量模型在三元组之间保持因果和语义一致性的能力。在CholecT45和ProstaTD数据集上的大量实验表明，我们的方法达到了最先进的性能，AP_IVT分别提高了5.1%和7.8%。此外，根据TCER，我们的方法在这两个数据集上实现了超过36%和25%的相对减少，证明了我们框架在时间关系共同推理中的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 55 / 2608.04610

HiSC: Hierarchical Spatial Clustering Token Compression for Efficient 3D Scene Understanding

HiSC：用于高效3D场景理解的分层空间聚类令牌压缩

Qu, Jiuhe, Liang, Yingping, Fu, Ying

Abstract

3D vision-language models (3D VLMs) enable spatial reasoning over multi-view scenes but suffer from substantial token redundancy due to duplicated observations and large uninformative regions, leading to high computational cost. Although visual token compression has shown promise in accelerating 2D VLMs, it fails to capture the structured nature of 3D scenes and leads to incomplete spatial coverage and loss of fine-grained details. In this paper, we propose \textbf{HiSC}, a training-free framework for hierarchical spatial clustering token compression in 3D VLMs. HiSC lifts token compression from token-level selection to cluster-level processing by organizing tokens into spatially grounded clusters using joint geometric and semantic cues. Specifically, we first introduce a \textbf{spatial graph-based merging (SGraM) strategy} that models cross-view redundancy as spatial connectivity and consolidates physically consistent regions, effectively merging extremely similar redundant tokens prior to LLM inference. We then propose a \textbf{spatial clustering-based pruning (SCluP) paradigm} within LLM inference, which performs hierarchical compression across clusters and within clusters, preserving object instance completeness while retaining fine-grained details for important regions. Extensive experiments on diverse 3D reasoning benchmarks show validate the effectiveness of HiSC, particularly under high visual token pruning ratios. Besides, HiSC achieves over 90\% token reduction with minimal performance degradation. Code is accessible at https://github.com/elecreak/HiSC.

Chinese Translation

3D视觉-语言模型（3D VLMs）能够对多视角场景进行空间推理，但由于重复观察和大面积无信息区域导致的显著令牌冗余，使得计算成本高昂。尽管视觉令牌压缩在加速2D VLMs方面显示出潜力，但未能捕捉3D场景的结构特性，导致空间覆盖不完整和细节丢失。在本文中，我们提出了 extbf{HiSC}，一个用于3D VLMs的无训练框架，旨在实现分层空间聚类令牌压缩。HiSC通过使用联合几何和语义线索将令牌组织为空间上有据的聚类，将令牌压缩从令牌级选择提升到聚类级处理。具体而言，我们首先引入了一种 extbf{基于空间图的合并（SGraM）策略}，将视角间冗余建模为空间连通性，并整合物理一致区域，有效地在LLM推理之前合并极其相似的冗余令牌。然后，我们在LLM推理中提出了一种 extbf{基于空间聚类的剪枝（SCluP）范式}，在聚类间和聚类内执行分层压缩，保持对象实例的完整性，同时为重要区域保留细粒度细节。在多样的3D推理基准上的大量实验验证了HiSC的有效性，尤其是在高视觉令牌剪枝比率下。此外，HiSC实现了超过90 ext{%}的令牌减少，且性能下降最小。代码可在https://github.com/elecreak/HiSC获取。

View on arXiv Download PDF AI Translation

cs.CV / 56 / 2608.04622

DAC-Pose: Dual-Agent Collaborative Framework for Pose-Guided Human Generation

DAC-Pose：基于双代理的姿态引导人类生成协作框架

Yang, Haotian, Yang, Zhile, Zhou, Huiyu, Sun, Xin

Abstract

AI agents have emerged as a powerful new paradigm in generative image synthesis, enabling systems to perform complex semantic reasoning rather than passive pixel-level mapping. In pose-guided human generation, conventional methods inevitably produce severe visual artifacts under drastic viewpoint shifts, fundamentally because they lack the cognitive capacity to logically deduce unseen regions and model complex spatial deformations. To bridge this gap, we propose DAC-Pose, a novel agent-driven multimodal framework that reformulates single-view human generation as a collaborative dual-agent system. DAC-Pose integrates two complementary components, namely, the Prior Semantic Reasoning (PSR) agent and the Discrepancy-Aware Visual Encoding (DAVE) agent. Functioning as a cognitive engine, PSR utilizes collaborative reasoning to deduce the fine-grained attributes of unseen regions. Concurrently, acting as a specialized visual perception agent, DAVE quantifies and encodes viewpoint-induced spatial misalignments, continuously feeding robust spatial constraints back into the generative process. This autonomous feedback loop between semantic deduction and visual perception ensures high-fidelity detail synthesis. Extensive experiments on the DeepFashion and Market-1501 benchmarks validate the superiority of our agent-driven paradigm. Notably, DAC-Pose excels in preserving texture alignment and identity consistency under drastic viewpoint changes. The code is available at https://github.com/AIVRC/DAC-Pose.

Chinese Translation

人工智能代理作为生成图像合成中的一种强大新范式，能够使系统进行复杂的语义推理，而不仅仅是被动的像素级映射。在姿态引导的人类生成中，传统方法在剧烈视角变化下不可避免地产生严重的视觉伪影，根本原因在于它们缺乏逻辑推理未见区域和建模复杂空间变形的认知能力。为了解决这一问题，我们提出了DAC-Pose，一种新颖的基于代理的多模态框架，将单视图人类生成重新构建为一个协作的双代理系统。DAC-Pose整合了两个互补的组件，即先验语义推理（Prior Semantic Reasoning, PSR）代理和差异感知视觉编码（Discrepancy-Aware Visual Encoding, DAVE）代理。PSR作为认知引擎，利用协作推理来推导未见区域的细粒度属性。同时，DAVE作为一个专门的视觉感知代理，量化并编码视角引起的空间错位，持续将强健的空间约束反馈到生成过程中。这种语义推理与视觉感知之间的自主反馈循环确保了高保真细节合成。在DeepFashion和Market-1501基准上的大量实验验证了我们基于代理的范式的优越性。值得注意的是，DAC-Pose在剧烈视角变化下在保持纹理对齐和身份一致性方面表现出色。代码可在 https://github.com/AIVRC/DAC-Pose 获取。

View on arXiv Download PDF AI Translation

cs.CV / 57 / 2608.04623

Visual Anchoring in Diffusion: Multimodal Zero-Shot Skeleton Action Recognition

扩散中的视觉锚定：多模态零-shot 骨架动作识别

Bao, Zehao, Guo, Shujun, Yu, Bruce X. B.

Abstract

Zero-shot Skeleton Action Recognition (ZSAR) remains ambiguous when unseen actions share similar skeleton joint dynamics but differ in objects or scene context. RGB provides these missing cues, yet existing multimodal methods typically maintain independent skeleton and RGB scoring branches and fuse their outputs. Without using unlabeled test data for adaptation or fusion calibration, a fixed fusion weight cannot capture class-pair-dependent modality reliability, while an adaptive rule lacks target-side feedback for deciding which branch should dominate. We bypass this weight-selection problem via the classify-by-generation paradigm, where each class is scored by how accurately a text-conditioned denoiser predicts the noise added to the skeleton feature. This formulation separates the progressively corrupted skeleton from fixed conditioning, allowing RGB and text to jointly condition a single class-scoring function rather than produce independent scores. We instantiate this idea as Multimodal Triplet Diffusion for Skeleton-Text Matching (TDSM-MM), augmenting a text-conditioned denoising Transformer with a non-diffused RGB condition token that serves as a stable visual anchor during skeleton data reconstruction. Our proposed TDSM-MM has been ablated via extensive experiments and achieved the best inductive accuracy on three of four NTU-60/120 splits and surpasses the transductive state-of-the-art on NTU-120 96/24 (i.e., 71.3% vs. 69.1%), without test-time adaptation, suggesting that diffusion-based methods can be a promising direction for zero-shot learning.

Chinese Translation

零-shot 骨架动作识别（ZSAR）在未见动作共享相似骨架关节动态但在物体或场景上下文中存在差异时仍然模糊。RGB 提供了这些缺失的线索，然而现有的多模态方法通常保持独立的骨架和 RGB 评分分支，并融合它们的输出。在不使用未标记测试数据进行适应或融合校准的情况下，固定的融合权重无法捕捉到类对依赖的模态可靠性，而自适应规则缺乏目标侧反馈以决定哪个分支应占主导地位。我们通过生成分类范式绕过了这一权重选择问题，其中每个类的评分基于文本条件去噪器对添加到骨架特征的噪声预测的准确性。该公式将逐步损坏的骨架与固定条件分离，使得 RGB 和文本共同条件化单一类评分函数，而不是产生独立的评分。我们将这一思想实例化为多模态三元组扩散用于骨架-文本匹配（TDSM-MM），增强了一个文本条件去噪 Transformer，加入了一个非扩散的 RGB 条件标记，作为骨架数据重建过程中的稳定视觉锚定。我们提出的 TDSM-MM 通过广泛的实验进行了消融测试，并在四个 NTU-60/120 分割中的三个上达到了最佳归纳准确率，超越了 NTU-120 96/24 上的传导最先进技术（即 71.3% 对 69.1%），而无需测试时适应，表明基于扩散的方法可以成为零-shot 学习的一个有前景的方向。

View on arXiv Download PDF AI Translation

cs.CV / 58 / 2608.04642

YOLO-PVC: 2D-to-3D Consolidation of Slice-wise Detections for Volumetric Liver Tumor Localization in MRI

YOLO-PVC：用于MRI中体积肝肿瘤定位的切片检测的2D到3D整合

Waqas, Talha, Lahlouh, Mounir, Taibouni, Kawther, Waqas, Mahnoor, Ahmed, Salar, Mulé, Sébastien, Leroul-Chenoune, Yasmina

Abstract

Slice-wise 2D object detectors are increasingly applied to volumetric data due to their computational efficiency and scalability, yet they often yield fragmented and unstable predictions along the depth axis. We propose YOLO-PVC, a lightweight and model-agnostic framework for 2D-to-3D consolidation of slice-wise detections. The method enforces depth continuity, aggregates bounding box coordinates using robust percentile statistics, and further refines axial extent through a lightweight MLP-based calibration module. Unlike na\"ive stacking or averaging strategies, YOLO-PVC explicitly addresses missing detections and outlier slices along the depth dimension. Experiments on 3D liver MRI volumes across three tumor categories demonstrate consistent improvements over multiple aggregation baselines. The heuristic PVC achieves an overall $\mathrm{IoU}_{3D}$ of $0.665$, while the calibrated variant further improves performance to $0.710$, with high planar overlap ($\mathrm{BEV\ IoU} \approx 0.78$). These results demonstrate that structured geometric consolidation provides an effective and practical solution for volumetric liver tumor localization in clinical MRI.

Chinese Translation

切片级2D目标检测器因其计算效率和可扩展性，越来越多地应用于体积数据，但它们在深度轴上往往产生碎片化和不稳定的预测。我们提出了YOLO-PVC，一个轻量级且与模型无关的框架，用于切片检测的2D到3D整合。该方法强制执行深度连续性，利用稳健的百分位统计聚合边界框坐标，并通过轻量级的基于多层感知机（MLP）的校准模块进一步细化轴向范围。与简单的堆叠或平均策略不同，YOLO-PVC明确解决了深度维度上的缺失检测和异常切片。对三种肿瘤类别的3D肝脏MRI体积的实验表明，相较于多个聚合基线，YOLO-PVC在性能上有一致的提升。启发式PVC的整体$ ext{IoU}_{3D}$为$0.665$，而校准变体进一步将性能提升至$0.710$，并具有较高的平面重叠（$ ext{BEV IoU} ext{约} 0.78$）。这些结果表明，结构化几何整合为临床MRI中的体积肝肿瘤定位提供了一种有效且实用的解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 59 / 2608.04652

DisMix: Order-Aware Mixup for Medical Imaging via Disentangling Ordinal and Non-Ordinal Features

DisMix：通过解耦有序和无序特征的医学影像顺序感知混合方法

Pitawela, Dileepa, Carneiro, Gustavo, Chen, Hsiang-Ting

Abstract

Image mixup is a widely adopted data augmentation strategy, yet it is ill-suited for ordinal classification tasks such as medical disease grading, where labels encode a progression of severity. By indiscriminately blending disease-severity cues (ordinal) with appearance-level variation (non-ordinal), standard mixup produces samples that distort the very ordinal structure that underpins clinical severity grading. We introduce DisMix, an order-aware mixup framework for ordinal classification. DisMix disentangles ordinal and non-ordinal features via a dual-codebook VQ-VAE, allowing each subspace to be mixed independently: ordinal codes are interpolated to produce meaningful intermediate ranks, while non-ordinal codes are varied to introduce appearance diversity without corrupting the ordinal signal. Across four medical imaging datasets, DisMix shows the best aggregate performance among six image mixup baselines paired with six ordinal classifiers and remains effective under data scarcity and clinical grading variability.

Chinese Translation

图像混合是一种广泛采用的数据增强策略，但对于诸如医学疾病分级等有序分类任务并不适用，因为这些任务的标签编码了严重程度的进展。通过不加区分地将疾病严重性线索（有序）与外观级别的变化（无序）混合，标准混合方法产生的样本扭曲了支撑临床严重性分级的有序结构。我们提出了DisMix，一种针对有序分类的顺序感知混合框架。DisMix通过双代码本的变分量子自编码器（VQ-VAE）解耦有序和无序特征，使得每个子空间可以独立混合：有序代码被插值以产生有意义的中间等级，而无序代码则被变化以引入外观多样性，而不破坏有序信号。在四个医学影像数据集上，DisMix在与六个有序分类器配对的六个图像混合基线中显示出最佳的整体性能，并在数据稀缺和临床分级变异性下仍然有效。

View on arXiv Download PDF AI Translation

cs.CV / 60 / 2608.04653

Overcoming Statistical Bias in Action-Controllable World Models

克服可控动作世界模型中的统计偏差

Shi, Yuhong, Chu, Zhenhao, Wei, Jie, Hao, Jun, Liu, Jianyi, Fu, Jingwen

Abstract

Action-conditioned world models aim to predict how visual environments evolve under an agent's actions. Yet future frames are often highly predictable from visual inertia and recurring motion patterns alone. This creates a shortcut: models can fit the data by exploiting statistical biases without making their visible dynamics meaningfully depend on the action. As a result, different actions may produce similar futures, while motion may persist even under zero action. The key question is how to reduce reliance on statistical shortcuts from dominating action-conditioned prediction. We argue that action control requires more than injecting action features; it requires enforcing consistency under counterfactual changes to actions and observations. Based on this insight, we introduce CoCo, a Counterfactual Consistency framework to enhance action controllability through two complementary constraints. Multi-step counterfactual consistency constrains reference, inverse-action, and zero-action rollouts, while action-spatial counterfactual consistency enforces consistent predictions under mirrored scenes and transformed actions. Together, they reduce reliance on statistical shortcuts from substituting for action-dependent dynamics. We further introduce Action Response Consistency (ARC) and Drift Energy (DE) to assess action controllability, together with Mini-SSMB for same-state, multi-action counterfactual evaluation. On Mini-SSMB, our full model achieved ARC_inv of 0.412 and ARC_ref of 0.483, while reducing DE by 17.07% relative to the baseline. On VP2 visual planning, it achieves the highest average success rate among SOTA models, at 73.1%. Experiments on BAIR and RoboNet further show that these gains preserve video prediction quality and transfer across model settings.

Chinese Translation

动作条件的世界模型旨在预测在智能体的动作下视觉环境如何演变。然而，未来的帧往往仅凭视觉惯性和重复运动模式就可以高度预测。这创造了一个捷径：模型可以通过利用统计偏差来拟合数据，而不需要使其可见的动态有意义地依赖于动作。因此，不同的动作可能产生相似的未来，而在零动作下运动可能仍然持续。关键问题在于如何减少对统计捷径的依赖，以防止其主导动作条件的预测。我们认为，动作控制不仅需要注入动作特征；还需要在反事实变化的动作和观察下强制执行一致性。基于这一见解，我们引入了 CoCo，一个反事实一致性框架，通过两个互补的约束来增强动作可控性。多步反事实一致性约束参考、逆动作和零动作的回放，而动作空间反事实一致性则在镜像场景和变换动作下强制一致的预测。两者共同减少了对统计捷径的依赖，以替代动作依赖的动态。我们进一步引入了动作响应一致性（ARC）和漂移能量（DE）来评估动作可控性，并结合 Mini-SSMB 进行同状态、多动作的反事实评估。在 Mini-SSMB 上，我们的完整模型达到了 0.412 的 ARC_inv 和 0.483 的 ARC_ref，同时相对于基线减少了 17.07% 的 DE。在 VP2 视觉规划上，它在 SOTA 模型中达到了最高的平均成功率，达 73.1%。在 BAIR 和 RoboNet 上的实验进一步表明，这些增益保持了视频预测质量，并在模型设置之间转移。

View on arXiv Download PDF AI Translation

cs.CV / 61 / 2608.04655

CSGen: A Multi-Domain Curvilinear Structure Generation Model via Hierarchical Multimodal Diffusion

CSGen：一种通过分层多模态扩散生成多领域曲线结构的模型

Shan, Zhe, Yang, Ziming, Zhou, Lei, Zhang, Wenwen, Lin, Cong, Xie, Xia

Abstract

Curvilinear structure analysis is an important and fundamental task in multimedia. However, the controllable generation of images with precise curvilinear structure objects remains an open challenge. To address this, we propose CSGen, a hierarchical multimodal diffusion model that synthesizes high-fidelity images precisely aligned with multiple control conditions. The CSGen is built upon three key innovations: 1) We construct a multi-domain and multimodal dataset, including over 24K samples from 5 domains and 7 different types of annotations, to train the unified generation model. 2) We propose a novel hierarchical progressive control strategy that decouples topology clues from visual context by a phased signal injection, mitigating semantic drift while ensuring the topological integrity of sparse structures. 3) We design a sparsity-aware loss re-weighting mechanism to address the extreme sparsity of curvilinear structures, significantly enhancing the attention on thin and fragile structures during optimization. Extensive experiments demonstrate that CSGen generates images with superior structure accuracy and visual realism, significantly improving downstream segmentation performance while maintaining robustness across diverse prompts. Our results confirm CSGen as a scalable, data-centric paradigm for the analysis of complex curvilinear structures in diverse multimedia applications. Code and dataset are available at https://github.com/ShanZard/CSGen.

Chinese Translation

曲线结构分析是多媒体中的一项重要且基础的任务。然而，精确控制生成具有特定曲线结构对象的图像仍然是一个未解决的挑战。为了解决这个问题，我们提出了CSGen，一种分层多模态扩散模型，能够合成与多个控制条件精确对齐的高保真图像。CSGen建立在三个关键创新之上：1）我们构建了一个多领域和多模态的数据集，包括来自5个领域和7种不同类型注释的超过24K样本，以训练统一的生成模型。2）我们提出了一种新颖的分层渐进控制策略，通过分阶段信号注入将拓扑线索与视觉上下文解耦，减轻语义漂移，同时确保稀疏结构的拓扑完整性。3）我们设计了一种关注稀疏性的损失重加权机制，以应对曲线结构的极端稀疏性，在优化过程中显著增强对细小和脆弱结构的关注。大量实验表明，CSGen生成的图像在结构准确性和视觉真实感方面优于其他方法，显著提高了下游分割性能，同时在多样化提示中保持了鲁棒性。我们的结果确认CSGen作为一种可扩展的数据中心范式，适用于多种多媒体应用中的复杂曲线结构分析。代码和数据集可在 https://github.com/ShanZard/CSGen 获取。

View on arXiv Download PDF AI Translation

cs.CV / 62 / 2608.04657

MobileWAM: Bridging World Action Models to Mobile Manipulation with Chain-of-Foresight

MobileWAM：通过前瞻链将世界动作模型与移动操控相结合

Fan, Zehua, He, Junjie, Song, Wenxuan, Wang, Xi, Lyu, Wenqi, Zhao, Linge, Li, Fuhao, You, Zihan, Yang, Yifei, Xu, Kaiming, Jiang, Qi, Jiang, Yue, Li, Haoang, Chi, Cheng, Li, Bailin, Wang, Yan

Abstract

World action models (WAMs) built on video generation backbones are a rising recipe for robot learning, yet remain confined to tabletop manipulation. Mobile manipulation demands simultaneous locomotion and whole-body manipulation amid scene-scale dynamics, yet is still dominated by dynamics-blind visual encoders with hand-crafted coordination. We bridge this gap with MobileWAM, a mixture-of-transformers architecture that fuses a pretrained video diffusion transformer with a lightweight action expert through layerwise joint attention, translating internet-scale motion priors into whole-body control. To reconcile the heterogeneous dynamics of moving and manipulating, each feed-forward layer of the action expert becomes a three-expert mixture of shared, locomotion, and manipulation experts, softly routed by the motion intent in the action tokens. To densify supervision, we further propose Chain-of-Foresight (CoF): intermediate representations sequentially predict a chain of future latent chunks, each step conditioned on its predecessor. CoF pairs naturally with our decoupled video--action denoising scheme. At deployment, the WAM serves as a pure current-frame encoder; foresight acts only through gradients, so at inference the foresight chain and video generation are discarded, leaving only policy-level cost. MobileWAM surpasses state-of-the-art mobile manipulation policies on ManiSkill-HAB and fine-tunes to a real ARX Lift2 mobile manipulator across diverse tasks with strong generalization. Code will be released soon.

Chinese Translation

基于视频生成骨干网构建的世界动作模型（WAM）是机器人学习的新兴方法，但仍然局限于桌面操控。移动操控需要在场景规模的动态中实现同时的移动和全身操控，但目前仍主要依赖于对动态盲目的视觉编码器和手工设计的协调。我们通过MobileWAM架构填补了这一空白，该架构是一种混合变换器架构，将预训练的视频扩散变换器与轻量级动作专家通过逐层联合注意力相融合，将互联网规模的运动先验转化为全身控制。为了调和移动和操控的异质动态，动作专家的每个前馈层成为共享、移动和操控专家的三专家混合，依据动作标记中的运动意图进行柔性路由。为了加强监督，我们进一步提出了前瞻链（Chain-of-Foresight, CoF）：中间表示序列地预测未来潜在片段的链条，每一步都基于其前一步进行条件化。CoF与我们解耦的视频-动作去噪方案自然配对。在部署时，WAM作为纯当前帧编码器；前瞻仅通过梯度作用，因此在推理时，前瞻链和视频生成被丢弃，仅保留策略级成本。MobileWAM在ManiSkill-HAB上超越了最先进的移动操控策略，并在多样化任务中对真实的ARX Lift2移动操控器进行了微调，展现出强大的泛化能力。代码将很快发布。

View on arXiv Download PDF AI Translation

cs.CV / 63 / 2608.04673

Differential 6-DOF Pose Estimation with Provable First-Order Immunity to Camera Calibration Errors

具有可证明的一阶免疫相机标定误差的差分六自由度姿态估计

Zhang, Yueqiang, Deng, Liang, Zhang, Yi, Wang, Baoqiong, Chen, Wenjun, Pan, Shuixin, Guo, Yulan, Yu, Qifeng

Abstract

Accurate six-degree-of-freedom (6-DOF) motion estimation is essential for robotic manipulation, autonomous systems, and structural displacement monitoring. Conventional 3D-2D methods estimate absolute camera poses independently at each time and recover platform motion through camera-to-platform extrinsics, making them sensitive to extrinsic calibration errors, especially for micromotion. We present a differential pose estimation method that directly recovers platform motion from inter-frame image displacements and known 3D control points. By differencing perspective projection equations, using a depth-invariance approximation, and modeling motion on SE(3), the method avoids independent absolute-pose estimation and supports both monocular and multi-camera systems. We prove that translational extrinsic errors cancel exactly, while rotational errors induce a bounded perturbation determined by calibration error, motion magnitude, and observation geometry. We also derive generic observability conditions, a Cramer-Rao lower bound, and a bias-eliminated consistent estimator, and characterize the validity limits of the approximations. Extensive synthetic and real-world experiments establish a new state of the art for 6-DOF platform micromotion estimation, outperforming representative PnP and generalized-PnP methods in accuracy, calibration robustness, and computational efficiency. With five control points and 0.5-pixel image noise, the monocular solver obtains a combined pitch-yaw rotation RMSE of 10.09 arcsec, a translation RMSE of 3.70 mm, and a runtime of 0.34 ms. The binocular solver achieves a rotation RMSE of 10.58 arcsec, a translation RMSE of 3.91 mm, and a runtime of 0.27 ms. Code will be released upon publication at https://github.com/zyoungszu/pami2026.

Chinese Translation

准确的六自由度（6-DOF）运动估计对于机器人操作、自动化系统和结构位移监测至关重要。传统的3D-2D方法在每个时刻独立估计绝对相机姿态，并通过相机与平台之间的外参恢复平台运动，这使得它们对外部标定误差特别敏感，尤其是在微运动情况下。我们提出了一种差分姿态估计方法，该方法直接通过帧间图像位移和已知的3D控制点恢复平台运动。通过对透视投影方程进行差分，使用深度不变近似，并在SE(3)上建模运动，该方法避免了独立的绝对姿态估计，并支持单目和多摄像头系统。我们证明了平移外参误差完全抵消，而旋转误差引入的扰动是由标定误差、运动幅度和观测几何确定的有界扰动。我们还推导了通用可观测性条件、Cramer-Rao下界和消除偏差的一致估计器，并表征了近似的有效性限制。大量的合成和真实世界实验建立了6-DOF平台微运动估计的新状态，超越了代表性的PnP和广义PnP方法在准确性、标定鲁棒性和计算效率方面的表现。在五个控制点和0.5像素图像噪声的情况下，单目求解器获得了10.09弧秒的联合俯仰-偏航旋转均方根误差（RMSE）、3.70毫米的平移RMSE和0.34毫秒的运行时间。双目求解器实现了10.58弧秒的旋转RMSE、3.91毫米的平移RMSE和0.27毫秒的运行时间。代码将在出版后发布于 https://github.com/zyoungszu/pami2026。

View on arXiv Download PDF AI Translation

cs.CV / 64 / 2608.04676

SurgNarrator: A Generative Retrieval Framework for Surgical Video Understanding

SurgNarrator：用于外科视频理解的生成检索框架

Feng, Yuqing, Ma, Jiawei, Lin, Kevin Qinghong, Yuan, Kun, Padoy, Nicolas, Elson, Daniel S., Nguyen, Anh, Giannarou, Stamatia, Huang, Baoru

Abstract

Surgical procedures unfold as structured and recurring clinical events, whose real-time understanding via intraoperative surgical videos is critical for intraoperative decision-making and support. However, existing video understanding methods force a trade-off: autoregressive video-language models support comprehensive reasoning but are not practical for time-sensitive clinical applications, whereas contrastive models offer low latency but struggle with complex scene understanding. Recently, generative retrieval has been explored for general-domain video understanding, but transferring it to surgery is not trivial because near-identical visual appearances may indicate semantically distinct events, and the terminology involved is highly surgery-specific. To this end, we propose SurgNarrator, a new generative retrieval framework tailored for surgical video understanding. We construct a well-curated surgery-centric vocabulary from surgical captions to define a clinically meaningful retrieval space. We then adapt the pre-trained Qwen3-VL-Embedding-8B to learn discriminative clinical representations with a temporally-aware contrastive objective. During inference, a hierarchical, procedure-aware retrieval strategy narrows the search space to the relevant procedure type, delivering fast and effective responses. Our method is comprehensively evaluated on twelve benchmarks in a zero-shot setting and achieves consistent performance gains over state-of-the-art baselines, while reducing output-stage latency by more than two orders of magnitude compared with the generative baseline.

Chinese Translation

外科手术过程作为结构化和重复的临床事件展开，通过术中外科视频进行实时理解对术中决策和支持至关重要。然而，现有的视频理解方法存在权衡：自回归视频-语言模型支持全面推理，但在时间敏感的临床应用中并不实用；而对比模型则提供低延迟，但在复杂场景理解方面表现不佳。最近，生成检索被探索用于一般领域的视频理解，但将其转移到外科领域并非易事，因为几乎相同的视觉外观可能指示语义上不同的事件，且所涉及的术语高度特定于外科。为此，我们提出了SurgNarrator，一个专为外科视频理解量身定制的新生成检索框架。我们从外科字幕中构建了一个精心策划的外科中心词汇，以定义一个具有临床意义的检索空间。然后，我们调整了预训练的Qwen3-VL-Embedding-8B，以学习具有时间感知对比目标的区分性临床表示。在推理过程中，分层的、程序感知的检索策略将搜索空间缩小到相关的手术类型，从而提供快速有效的响应。我们的方法在十二个基准测试中进行了全面评估，在零样本设置下实现了相较于最先进基线的一致性能提升，同时与生成基线相比，输出阶段延迟减少了两个数量级以上。

View on arXiv Download PDF AI Translation

cs.CV / 65 / 2608.04698

Teaching MLLMs to Say No: Generalized Referring Expression Comprehension via Refusal Calibrated GRPO

教会多模态大型语言模型拒绝：通过拒绝校准的群体相对策略优化实现广义指称表达理解

Yang, Xuzheng, Ling, Jun, Huang, Tao, Qin, Caiyan, Wang, Peng

Abstract

We tackle the challenging yet underexplored task of Generalized Referring Expression Comprehension (GREC), which requires a model to localize the object described by a textual expression when it exists (positive sample) and to refuse output when it does not (negative sample). Although Multimodal Large Language Models (MLLMs) excel at localizing existing objects, they often fail to reject nonexistent ones due to the absence of negative samples during training, producing hallucinated bounding boxes. Existing post-training approaches such as supervised fine-tuning (SFT) and reinforcement learning (RL) enhance refusal behavior but usually degrade localization accuracy on positive samples, undermining the model's core competence. To address this, we propose Refusal-Calibrated Group Relative Policy Optimization (RC-GRPO), a calibrated RL strategy that strengthens the refusal ability of MLLMs while preserving localization performance. It enforces "None" outputs in rollouts for valid advantage estimation on negative samples and applies a penalty to prevent over-refusal on positives, achieving a balanced trade-off between accuracy and reliability. A second-stage reasoning reinforcement further consolidates causal understanding and interpretability. Experiments on three GREC benchmarks demonstrate that RC-GRPO attains superior localization accuracy while maintaining strong refusal capability.

Chinese Translation

我们研究了一个具有挑战性但尚未深入探索的任务——广义指称表达理解（Generalized Referring Expression Comprehension, GREC），该任务要求模型在存在时定位文本表达所描述的对象（正样本），而在不存在时拒绝输出（负样本）。尽管多模态大型语言模型（Multimodal Large Language Models, MLLMs）在定位现有对象方面表现出色，但由于训练过程中缺乏负样本，它们往往无法拒绝不存在的对象，从而产生虚幻的边界框。现有的后训练方法，如监督微调（Supervised Fine-Tuning, SFT）和强化学习（Reinforcement Learning, RL），虽然增强了拒绝行为，但通常会降低正样本的定位准确性，削弱模型的核心能力。为了解决这个问题，我们提出了拒绝校准的群体相对策略优化（Refusal-Calibrated Group Relative Policy Optimization, RC-GRPO），这是一种校准的强化学习策略，旨在增强多模态大型语言模型的拒绝能力，同时保持定位性能。该方法在有效的负样本优势估计中强制输出“无”，并对正样本的过度拒绝施加惩罚，实现准确性与可靠性之间的平衡权衡。第二阶段的推理强化进一步巩固了因果理解和可解释性。在三个GREC基准上的实验表明，RC-GRPO在保持强大拒绝能力的同时，达到了更优的定位准确性。

View on arXiv Download PDF AI Translation

cs.CV / 66 / 2608.04701

UniWorld-View: Large-Baseline View Synthesis via Video Diffusion Models

UniWorld-View：通过视频扩散模型进行大基线视图合成

Zhou, Haiyang, Yu, Wangbo, Feng, Chaoran, Zhou, Xunyu, Tian, Yonghong, Yuan, Li

Abstract

The abundance of casually captured monocular videos and images on social media provides a valuable source for immersive content creation, where generating novel views from such sparse observations can greatly enhance user experiences. However, producing photorealistic and geometrically consistent views with precise camera control remains challenging when input coverage is extremely limited. Reconstruction-based approaches such as NeRF and 3D Gaussian Splatting (3DGS) deteriorate severely under sparse inputs and fail to explicitly handle occlusions. Generative methods ease data requirements but still struggle with large-baseline view synthesis due to inaccurate or implicit geometric guidance. To overcome these limitations, we introduce UniWorld-View, a unified framework for controllable large-baseline novel view synthesis from monocular inputs. UniWorld-View integrates explicit 3D guidance with generative diffusion modeling to enable precise camera control and geometrically consistent view generation. The geometric guidance is obtained through an occlusion-aware point cloud rendering strategy that resolves visibility ambiguities and provides accurate priors for diffusion-based synthesis. By coupling this rendering strategy with powerful video diffusion backbones, UniWorld-View achieves high-fidelity novel view generation even under extreme camera motions and wide-baseline changes, and can further provide multi-view videos for downstream dynamic 3DGS reconstruction. Experiments on the WorldScore benchmark and zero-shot NVS benchmarks demonstrate the effectiveness of UniWorld-View in controllability, geometric consistency, and visual fidelity.

Chinese Translation

社交媒体上大量随意捕捉的单目视频和图像为沉浸式内容创作提供了宝贵的资源，从这些稀疏观测中生成新视图可以极大提升用户体验。然而，当输入覆盖极其有限时，生成照片级真实感和几何一致的视图以及精确的相机控制仍然具有挑战性。基于重建的方法，如NeRF和3D高斯点云渲染（3DGS），在稀疏输入下严重退化，并且未能明确处理遮挡问题。生成方法减轻了数据需求，但由于几何指导不准确或隐式，仍然在大基线视图合成方面面临困难。为克服这些限制，我们提出了UniWorld-View，这是一个统一框架，用于从单目输入中进行可控的大基线新视图合成。UniWorld-View将显式的3D指导与生成扩散建模相结合，以实现精确的相机控制和几何一致的视图生成。几何指导通过一种考虑遮挡的点云渲染策略获得，该策略解决了可见性歧义，并为基于扩散的合成提供了准确的先验。通过将这种渲染策略与强大的视频扩散骨干网络结合，UniWorld-View即使在极端相机运动和大基线变化下也能实现高保真度的新视图生成，并且可以进一步为下游动态3DGS重建提供多视图视频。在WorldScore基准和零样本NVS基准上的实验表明，UniWorld-View在可控性、几何一致性和视觉保真度方面的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 67 / 2608.04704

A Multi-Sensor Dataset for Monitoring the Operational Environment of Rail Vehicles

用于监测铁路车辆操作环境的多传感器数据集

Diotallevi, Claudio, Gudiño, Rodrigo, Pachalieva, Zaharia, Neumaier, Philipp, Naumann, Patrick, Bochinski, Erik, Eiselein, Volker, Köppel, Martin

Abstract

Reliable environment monitoring is essential for the safe and efficient operation of automated railway systems, covering all Grades of Automation (GoA), from partially automated (GoA2) to fully automated operation (GoA4). Artificial Intelligence (AI) plays a central role in enabling these systems to detect, classify, and react to potential hazards in real time. The development of such AI-based perception systems requires large volumes of accurately annotated data for training and validation. Within the Digitale Schiene Deutschland (DSD) program, DB InfraGO AG and understandAI GmbH have developed a comprehensive multi- sensor dataset tailored to the needs of railway environment perception. This dataset contains over 7 million high-quality annotations of both railway-specific and general perception objects, captured under varying operational scenarios. The finalized dataset can now be requested at the DB InfraGO AG and serve as a valuable resource for advancing AI-driven environment monitoring in the railway domain.

Chinese Translation

可靠的环境监测对于自动化铁路系统的安全和高效运行至关重要，涵盖了所有自动化等级（GoA），从部分自动化（GoA2）到完全自动化操作（GoA4）。人工智能（AI）在使这些系统能够实时检测、分类和应对潜在危险方面发挥着核心作用。开发这样的基于AI的感知系统需要大量准确标注的数据用于训练和验证。在数字铁路德国（Digitale Schiene Deutschland，DSD）计划中，DB InfraGO AG与understandAI GmbH合作开发了一套全面的多传感器数据集，以满足铁路环境感知的需求。该数据集包含超过700万个高质量的铁路特定和一般感知对象的标注，数据是在不同操作场景下捕获的。最终的数据集现已可以在DB InfraGO AG申请，并将作为推动铁路领域AI驱动环境监测的宝贵资源。

View on arXiv Download PDF AI Translation

cs.CV / 68 / 2608.04720

YOLOv14:Unified Cross-Domain Real-Time Object Detectionwith Adaptive Multi-View Representation

YOLOv14：统一跨域实时目标检测与自适应多视图表示

Jia, Jinling, Lu, Jian, Yawl, Jone, Zhang, Chenbin

Abstract

Real-time object detectors achieve remarkable accuracy under controlled conditions, yet degrade sharply on non-ideal inputs: fisheye distortion, game-rendered characters, aerial viewpoints, and 360{\deg} panoramas. We present YOLOv14, aunified detection framework addressing these challenges through four synergisticinnovations. (1) Deformable Area-Attention (D-AAttn) replaces rigid attentiongrids with learned 2D deformation fields, enabling adaptive sampling under geometric distortion. (2) Game2Real Domain Adaptation aligns rendered-game and photographic feature distributions via Adaptive Instance Normalization (AdaIN)and adversarial domain confusion, allowing game characters are detected as realhumans. (3) Multi-View Conditioning injects learned viewpoint embeddings intothe backbone with a cross-view contrastive loss that pulls same-class features fromdifferent perspectives closer. (4) An Adaptive Augmentation Policy automaticallyclassifies each input' scene type and routes to optimal augmentations, while a DynamicScaleRouter learns per-input feature pyramid weights. Together, YOLOv14achieves 49.1 mAP on COCO val2017 at 2.91 ms (T4 GPU), and delivers substantial gains on fisheye (+4.1 mAP), panorama (+6.6 mAP), drone (+6.4 mAP), andgame-character (+26.1 mAP) benchmarks

Chinese Translation

实时目标检测器在受控条件下取得了显著的准确性，但在非理想输入下表现急剧下降：鱼眼畸变、游戏渲染角色、空中视角和360°全景。我们提出了YOLOv14，这是一个统一的检测框架，通过四项协同创新来应对这些挑战。(1) 可变形区域注意力（D-AAttn）用学习的二维变形场替代了刚性注意力网格，使得在几何畸变下能够进行自适应采样。(2) Game2Real领域适应通过自适应实例归一化（AdaIN）和对抗性领域混淆对齐渲染游戏和摄影特征分布，使得游戏角色能够被检测为真实人类。(3) 多视图条件注入将学习到的视角嵌入注入到主干网络中，并通过跨视图对比损失将来自不同视角的同类特征拉近。(4) 自适应增强策略自动分类每个输入的场景类型并引导到最佳增强，而动态尺度路由器学习每个输入的特征金字塔权重。综合来看，YOLOv14在COCO val2017上实现了49.1 mAP，延迟为2.91毫秒（T4 GPU），并在鱼眼（+4.1 mAP）、全景（+6.6 mAP）、无人机（+6.4 mAP）和游戏角色（+26.1 mAP）基准测试中取得了显著提升。

View on arXiv Download PDF AI Translation

cs.CV / 69 / 2608.04722

Multi-View Face and Gesture Animation with Dynamic Gaussians

基于动态高斯的多视角人脸与手势动画

Javanmardi, Alireza, Jeetmal, Vippin Kumar, Millerdurai, Christen, Pagani, Alain, Stricker, Didier

Abstract

Creating photorealistic 3D human avatars with realistic upper-body motion remains challenging. Existing approaches either focus on the head and overlook hand gestures, or reconstruct the full body but fail to preserve fine-grained facial fidelity and hand pose accuracy. As a result, current methods struggle to capture the subtle dynamics of facial expressions and hand gestures that are crucial for natural human communication. While methods based on full-body parametric models enable avatar reconstruction from monocular or multi-view inputs, they often lack accurate facial animation and detailed hand articulation. To address these limitations, we propose MVFGA, a novel multi-view-consistent pipeline for generating realistic upper-body avatars. Our approach models the face and hands separately and fuses them with a parametric upper-body mesh model, enabling the capture of fine-grained facial expressions and hand poses for accurate upper-body avatar reconstruction. We then splat 3D Gaussians onto the obtained mesh, enabling high-quality rendering of dynamic avatars from novel viewpoints. Furthermore, we introduce MVFGA-MoCap, a multi-view upper-body motion capture dataset featuring controlled facial expression sequences, diverse hand gestures, and free-form communication. Experiments show that MVFGA generates visually realistic avatars with high-fidelity facial expressions and hand motions, outperforming baselines for upper-body avatar animation. Project page: https://dfki-av.github.io/MVFGA/

Chinese Translation

创建具有逼真上半身运动的照片级真实感3D人类化身仍然具有挑战性。现有的方法要么专注于头部而忽视手势，要么重建全身但未能保持细致的面部真实感和手部姿态准确性。因此，当前的方法难以捕捉面部表情和手势的微妙动态，这对于自然的人类交流至关重要。虽然基于全身参数模型的方法能够从单目或多视角输入重建化身，但它们往往缺乏准确的面部动画和详细的手部关节运动。为了解决这些局限性，我们提出了MVFGA，一种新颖的多视角一致性管道，用于生成逼真的上半身化身。我们的方法分别建模面部和手部，并将其与参数化的上半身网格模型融合，从而能够捕捉细致的面部表情和手部姿态，以实现准确的上半身化身重建。然后，我们将3D高斯点涂抹到获得的网格上，从而能够从新视角高质量渲染动态化身。此外，我们引入了MVFGA-MoCap，一个多视角上半身运动捕捉数据集，包含受控的面部表情序列、多样的手势和自由形式的交流。实验表明，MVFGA生成的视觉上逼真的化身具有高保真的面部表情和手部动作，在上半身化身动画中优于基线方法。项目页面：https://dfki-av.github.io/MVFGA/

View on arXiv Download PDF AI Translation

cs.CV / 70 / 2608.04737

Dense Metric Depth Completion from Sparse Direct Time-of-Flight Sensors

稠密度量深度补全来自稀疏直接飞行时间传感器

Kim, Hakyeong, Wang, Ruicheng, Yao, Chengtang, Yang, Jiaolong, Kim, Min H.

Abstract

Direct Time-of-Flight (dToF) sensors provide highly accurate metric depth and are more robust than indirect ToF systems in challenging real-world conditions. However, their high manufacturing cost and limited photodiode array size produce depth maps that are extremely sparse, low-resolution, and noisy, making them unsuitable for VR/XR, robotics, and 3D perception tasks that require dense metric depth. Existing monocular and depth completion methods struggle to handle the unique sampling patterns and hardware artifacts of dToF devices, and their performance often deteriorates significantly under severe sparsity or noise. We present a generalizable framework for dense metric depth completion from sparse dToF measurements, capable of operating across diverse sensor types, sparsity levels, and noise conditions. Our model employs a depth-guided dual-branch Vision Transformer encoder that processes RGB images and sparse dToF measurements separately, while a masked joint attention module allows depth tokens to reliably guide image features without being overwritten by them. A lightweight decoder reconstructs dense metric depth efficiently, without diffusion-based or refinement-heavy post-processing. To address the scarcity of paired training data, we introduce a comprehensive dToF simulation pipeline that reproduces the characteristics of flash, sub-VGA flash, and rotating sensors, including hardware-induced degradation, irregular sparsity, and realistic noise distributions. Trained entirely on synthetic data, our model achieves strong zero-shot generalization across 6 datasets and 3 real dToF devices, outperforming state-of-the-art approaches in both accuracy and computational efficiency. This establishes a robust and practical solution for dense metric depth completion from sparse direct ToF sensors. Our code and models are open-sourced. See https://vclab.kaist.ac.kr/cvpr2026p3.

Chinese Translation

直接飞行时间（dToF）传感器提供高度精确的度量深度，并且在具有挑战性的现实条件下比间接ToF系统更具鲁棒性。然而，它们的高制造成本和有限的光电二极管阵列尺寸导致生成的深度图极其稀疏、低分辨率且噪声较大，使其不适合需要稠密度量深度的虚拟现实/扩展现实（VR/XR）、机器人技术和三维感知任务。现有的单目和深度补全方法难以处理dToF设备独特的采样模式和硬件伪影，且在严重稀疏或噪声情况下性能往往显著下降。我们提出了一个通用框架，用于从稀疏dToF测量中进行稠密度量深度补全，能够在多种传感器类型、稀疏程度和噪声条件下运行。我们的模型采用深度引导的双分支视觉变换器编码器，分别处理RGB图像和稀疏dToF测量，同时一个掩码联合注意模块允许深度标记可靠地引导图像特征，而不被其覆盖。一个轻量级解码器高效地重建稠密度量深度，无需基于扩散或重修的后处理。为了解决配对训练数据稀缺的问题，我们引入了一个全面的dToF仿真管道，重现了闪光、亚VGA闪光和旋转传感器的特性，包括硬件引起的退化、不规则稀疏性和现实的噪声分布。我们的模型完全在合成数据上训练，在6个数据集和3个真实dToF设备上实现了强大的零样本泛化，在准确性和计算效率上超越了最先进的方法。这为从稀疏直接ToF传感器中进行稠密度量深度补全提供了一个稳健且实用的解决方案。我们的代码和模型已开源，详见 https://vclab.kaist.ac.kr/cvpr2026p3。

View on arXiv Download PDF AI Translation

cs.CV / 71 / 2608.04750

Simile Understanding in Text-to-Image Models: An Evaluation Framework

文本到图像模型中的明喻理解：评估框架

Wang, Luecheng, Ozaki, Shintaro, Kamigaito, Hidetaka, Hayashi, Katsuhiko, Kwon, Jingun, Okumura, Manabu, Watanabe, Taro

Abstract

Similes provide a compact and expressive way to describe visual characteristics in text prompts. Recent text-to-image models (t2i models) can produce visually compelling outputs from simile prompts, yet even frontier models frequently misinterpret the metaphorical vehicle and confuse it with the object. These systematic failures reveal a gap between figurative language and object-level visual grounding in t2i models. To investigate this issue, we propose a scalable evaluation framework for simile understanding. Our framework includes (1) a controlled simile dataset in which metaphorical vehicles are drawn from a predefined set of object-detectable categories and combined with diverse templates, (2) automatic grounding metrics based on YOLO (You Only Look Once) detection, and (3) text encoder layer analysis using Diffusion Lens to track how metaphorical vehicles emerge during generation. Experiments across architecturally diverse t2i models reveal consistent literalization failure patterns. We further discuss potential mitigation strategies for improving simile grounding in t2i models.

Chinese Translation

明喻为文本提示中的视觉特征描述提供了一种简洁而富有表现力的方式。最近的文本到图像模型（t2i模型）能够从明喻提示中生成视觉上引人注目的输出，然而即使是最前沿的模型也常常误解隐喻载体，并将其与对象混淆。这些系统性的失败揭示了t2i模型中比喻语言与对象级视觉基础之间的差距。为了研究这一问题，我们提出了一个可扩展的明喻理解评估框架。我们的框架包括（1）一个受控的明喻数据集，其中隐喻载体来自预定义的可检测对象类别集合，并与多样的模板结合，(2) 基于YOLO（You Only Look Once）检测的自动基础度量，以及（3）使用Diffusion Lens进行文本编码器层分析，以追踪隐喻载体在生成过程中的出现。对不同架构的t2i模型的实验揭示了一致的字面化失败模式。我们进一步讨论了改善t2i模型中明喻基础的潜在缓解策略。

View on arXiv Download PDF AI Translation

cs.CV / 72 / 2608.04752

Revisiting Pose Sensitivity in Splat-based Computed Tomography under Sparse-view Reconstruction

重新审视稀视图重建下基于点云的计算机断层扫描中的姿态敏感性

Choi, Kiseok, Cho, Hyeongjun, Kim, Inchul, Kim, Min H.

Abstract

X-ray computed tomography (CT) reconstructs volumetric representations of objects from projection images obtained by transmitting X-rays through a target. Recent splat-based tomography, which represents a volume as a continuous distribution of 3D Gaussians, has demonstrated both high reconstruction quality and fast convergence in cone-beam sparse-view CT. However, when deployed in real CT systems with limited and non-uniform view distributions, we observe distinctive streak and strip artifacts that are far more pronounced than in conventional reconstruction methods. Through detailed analysis, we show that these artifacts primarily originate from pose inaccuracies in the acquisition geometry rather than from view sparsity itself. We revisit pose sensitivity in the splatting formulation and derive a stable gradient-based framework that jointly refines geometric parameters during reconstruction. Our study not only identifies how pose perturbations propagate through the differentiable projection operator but also reveals why splat-based CT is particularly vulnerable to geometric misalignment. The resulting formulation remains lightweight and easily integrable into existing pipelines while substantially improving reconstruction fidelity under real-world sparse-view conditions.

Chinese Translation

X射线计算机断层扫描（CT）通过将X射线透过目标物体获得的投影图像重建物体的体积表示。近期的基于点云的断层扫描方法将体积表示为3D高斯的连续分布，已在锥束稀视图CT中展示了高重建质量和快速收敛性。然而，当在具有有限且不均匀视图分布的实际CT系统中应用时，我们观察到明显的条纹和条带伪影，这些伪影比传统重建方法更为明显。通过详细分析，我们表明这些伪影主要源于获取几何中的姿态不准确，而不是视图稀疏性本身。我们重新审视了点云公式中的姿态敏感性，并推导出一个稳定的基于梯度的框架，该框架在重建过程中联合优化几何参数。我们的研究不仅识别了姿态扰动如何通过可微分投影算子传播，还揭示了为什么基于点云的CT特别容易受到几何失配的影响。最终的公式保持轻量且易于集成到现有流程中，同时在实际稀视图条件下显著提高了重建的保真度。

View on arXiv Download PDF AI Translation

cs.CV / 73 / 2608.04759

Trace, Verify, and Correct: A Training-Free Framework for Spatial Reasoning in Multimodal LLMs

追踪、验证与修正：一种无训练的多模态大语言模型空间推理框架

Yang, Yang, Chen, Jiawei, Chen, Tairan, Yin, Zhaoxia

Abstract

Although Multimodal Large Language Models (MLLMs) have made substantial progress, their spatial reasoning may still produce intermediate judgments inconsistent with the input image, allowing errors to propagate through the reasoning chain and affect the final answer. Existing methods mainly improve spatial reasoning through training or additional spatial information, without considering whether the reasoning process itself is faithful to the model input. Our study shows that unfaithful reasoning chains significantly reduce final-answer accuracy. To address this issue, we propose a modular and training-free framework for spatial reasoning verification and correction. The framework constructs a Spatial Evidence Graph (SEG), which associates atomic spatial evidence extracted from Chain-of-Thought reasoning with visual entities, spatial relations, source steps, and visual evidence. Spatial Evidence Reliability Assessment (SERA) evaluates the reliability of visual evidence based on object existence, localization, and geometric measurements. The framework then identifies the earliest spatial evidence unit contradicted by reliable visual evidence and guides the original MLLM to revise the subsequent reasoning and final answer. Across 15 model-dataset settings, our method achieves an average accuracy of 68.94%, outperforming the compared baselines by 8.55 percentage points on average. Our code will be open-sourced.

Chinese Translation

尽管多模态大语言模型（MLLMs）已取得显著进展，但其空间推理仍可能产生与输入图像不一致的中间判断，从而导致错误在推理链中传播并影响最终答案。现有方法主要通过训练或额外的空间信息来改善空间推理，而未考虑推理过程本身是否忠实于模型输入。我们的研究表明，不忠实的推理链显著降低了最终答案的准确性。为了解决这一问题，我们提出了一种模块化且无训练的空间推理验证与修正框架。该框架构建了一个空间证据图（Spatial Evidence Graph, SEG），将从思维链（Chain-of-Thought）推理中提取的原子空间证据与视觉实体、空间关系、源步骤和视觉证据关联起来。空间证据可靠性评估（Spatial Evidence Reliability Assessment, SERA）基于对象存在、定位和几何测量评估视觉证据的可靠性。然后，框架识别出与可靠视觉证据相矛盾的最早空间证据单元，并指导原始的MLLM修正后续推理和最终答案。在15个模型-数据集设置中，我们的方法平均准确率达到68.94%，比对比基线平均高出8.55个百分点。我们的代码将开源。

View on arXiv Download PDF AI Translation

cs.CV / 74 / 2608.04764

Splat-Based Metal Artifact Reduction in Cone-Beam CT via Compact Attenuation Modeling

基于喷溅的金属伪影减少方法在锥束CT中的紧凑衰减建模

Choi, Kiseok, Cho, Jaemin, Kim, Inchul, Kim, Min H.

Abstract

X-ray computed tomography (CT) suffers from severe metal artifacts when high-attenuation objects such as dental fillings or orthopedic implants are present. These artifacts originate from the polychromatic nature of X-rays, where attenuation varies strongly with photon energy and material composition, breaking the monochromatic assumption used by conventional reconstruction algorithms. Recent neural rendering approaches attempt to address this mismatch through differentiable polychromatic projection models, but they still struggle with smoothness bias, loss of fine structures, and prohibitive computation when extended to large-scale cone-beam CT. We introduce a splat-based metal artifact reduction framework that incorporates a physically grounded polychromatic forward model into a continuous Gaussian representation for cone-beam CT. Each Gaussian encodes the energy-dependent attenuation of the underlying material using a compact material parameterization, which enables efficient joint optimization of geometric and material properties without relying on a metal mask. This compact attenuation formulation captures the essential variation across biological tissues and metallic implants, allowing our model to explain metal-induced nonlinearity while preserving high-frequency structure. Experiments on simulated and real cone-beam CT scans show that our method converges significantly faster and suppresses metal artifacts more effectively than existing reconstruction and neural field-based approaches.

Chinese Translation

X射线计算机断层扫描（CT）在存在高衰减物体（如牙科填充物或骨科植入物）时会遭受严重的金属伪影。这些伪影源于X射线的多色性质，其中衰减随着光子能量和材料成分的变化而显著变化，打破了传统重建算法所采用的单色假设。近期的神经渲染方法试图通过可微分的多色投影模型来解决这一不匹配，但在扩展到大规模锥束CT时，仍然面临平滑性偏差、细微结构损失和计算成本高昂的问题。我们提出了一种基于喷溅的金属伪影减少框架，该框架将物理基础的多色前向模型融入到锥束CT的连续高斯表示中。每个高斯编码了基础材料的能量依赖衰减，采用紧凑的材料参数化，使得几何和材料属性的高效联合优化成为可能，而无需依赖金属掩膜。这种紧凑的衰减公式捕捉了生物组织和金属植入物之间的基本变化，使我们的模型能够解释金属引起的非线性，同时保留高频结构。在模拟和真实的锥束CT扫描实验中，我们的方法收敛速度显著更快，并且比现有的重建和基于神经场的方法更有效地抑制金属伪影。

View on arXiv Download PDF AI Translation

cs.CV / 75 / 2608.04766

FUSEP: A Multi-Center Benchmark for Diverse Tasks in Early Pregnancy Fetal Ultrasound Screening

FUSEP：早期妊娠胎儿超声筛查多中心基准测试

Pu, Bin, Yang, Jiewen, Wang, Liwen, Tan, Ying, He, Guannan, Dong, Xingbo, Lin, Qika, Guo, Jiarong, Yang, Lixian, Liu, Zuozhu, Li, Shengli, Li, Kenli

Abstract

A large number of infants with congenital anomalies are born each year globally, especially in areas with underdeveloped medical resources. Currently, fetal ultrasound screening is the most common modality for early pregnancy anatomy detection. This modality can detect anomalies earlier and provide opportune treatment advice. However, the lack of an ultrasound dataset on early fetal gestation has slowed down the development of automated assisted diagnosis. In this work, we present a benchmark dataset for Fetal Ultrasound Screening in Early Pregnancy to facilitate intelligent ultrasound examination and assisted diagnosis called FUSEP. Our dataset consists of two ultrasound views recommended by the international guideline, i.e., Crown-rump Length (CRL) and Nuchal Translucency (NT) views in three hospitals, totaling 4,017 ultrasound images, with 45,820 box-level expert-level annotations. Our dataset and baseline present the following three contributions: 1) Our medical experts annotated a total of 14 key anatomical structures in two views using a box-level format; 2) Our data is collected extensively from different sonographers, devices, scanning angles, hospitals, etc; 3) We report the performance of the semi-supervised learning, fully supervised learning, unsupervised domain adaptation (UDA), and source-free UDA in ultrasound images multi-object detection. To the best of our knowledge, this is the first publicly available dataset and benchmark for fetal early pregnancy ultrasound screening. We believe that FUSEP and benchmark can contribute to the medical community in the development of multiple tasks such as standard plane recognition, quality control on ultrasound images, automated assisted diagnostics in early fetal pregnancy, medical multi-object detection, domain adaptation for object detection, etc.

Chinese Translation

全球每年有大量先天性畸形婴儿出生，尤其是在医疗资源欠发达的地区。目前，胎儿超声筛查是早期妊娠解剖检测最常用的方式。这种方式可以更早地发现畸形，并提供及时的治疗建议。然而，缺乏早期胎儿妊娠的超声数据集减缓了自动辅助诊断的发展。在本研究中，我们提出了一个用于早期妊娠胎儿超声筛查的基准数据集，旨在促进智能超声检查和辅助诊断，称为FUSEP。我们的数据集包含国际指南推荐的两种超声视图，即冠臀长（Crown-rump Length, CRL）和颈部透明带（Nuchal Translucency, NT）视图，来自三家医院，总计4,017张超声图像，具有45,820个框级专家级注释。我们的数据集和基准测试有以下三项贡献：1）我们的医学专家使用框级格式对两种视图中的14个关键解剖结构进行了注释；2）我们的数据广泛收集自不同的超声医师、设备、扫描角度、医院等；3）我们报告了在超声图像多目标检测中，半监督学习、全监督学习、无监督域适应（Unsupervised Domain Adaptation, UDA）和无源UDA的性能。根据我们所知，这是第一个公开可用的早期妊娠胎儿超声筛查数据集和基准测试。我们相信FUSEP和基准测试可以为医学界在多个任务的发展做出贡献，例如标准平面识别、超声图像质量控制、早期胎儿妊娠的自动辅助诊断、医学多目标检测、目标检测的域适应等。

View on arXiv Download PDF AI Translation

cs.CV / 76 / 2608.04768

Embedding Large Language Models into Flow Controls: An Agentic Framework for Adaptive and Trustworthy Automated Cooking

将大型语言模型嵌入流控制：一种适应性和可信赖的自动烹饪代理框架

Song, Zihan, Huang, Hongwei, Sun, Yueshuo, Tian, Yonglin, Wang, Fei-Yue, Li, Bai

Abstract

Automated cooking robots have traditionally relied on predefined procedures and rule-based control, ensuring stable execution but offering limited personalization, whereas recent large-model approaches support natural language interaction but often suffer from opaque decision making and unreliable execution in real kitchens. To address this challenge, this paper proposes an agentic framework that systematically decomposes personalized cooking requirements into structured and verifiable control programs rather than directly mapping language to actions. Multiple AI agents collaboratively transform user intents into canonical recipes, workflow programs with explicit flow control, and executable Python code grounded in an atomic action library. The system consists of three tightly coupled stages: offline recipe-to-code generation through multiple agents, online closed-loop execution with supervisory intervention enabled by multimodal perception, and post-run adaptation that updates user preference models for long-term personalization. Real-world experiments on a physical cooking platform demonstrate that the proposed framework achieves reliable task completion, transparent execution logic, and effective anomaly handling across diverse personalized scenarios, validating its practicality for trustworthy automated cooking in real environments.

Chinese Translation

自动烹饪机器人传统上依赖于预定义的程序和基于规则的控制，确保了稳定的执行但提供了有限的个性化，而最近的大型模型方法支持自然语言交互，但在真实厨房中往往面临决策不透明和执行不可靠的问题。为了解决这一挑战，本文提出了一种代理框架，该框架系统地将个性化烹饪需求分解为结构化和可验证的控制程序，而不是直接将语言映射到动作。多个人工智能代理协同将用户意图转化为规范食谱、具有明确流控制的工作流程程序以及基于原子动作库的可执行Python代码。该系统由三个紧密耦合的阶段组成：通过多个代理进行的离线食谱到代码生成、通过多模态感知实现的在线闭环执行以及更新用户偏好模型以实现长期个性化的后运行适应。针对物理烹饪平台的真实世界实验表明，所提出的框架在多样化个性化场景中实现了可靠的任务完成、透明的执行逻辑和有效的异常处理，验证了其在真实环境中可信赖的自动烹饪的实用性。

View on arXiv Download PDF AI Translation

cs.CV / 77 / 2608.04791

On the Effectiveness of Adaptation Strategies for VLM-Based Federated Learning in Remote Sensing

基于视觉语言模型的遥感联邦学习适应策略的有效性研究

Lösche, Simon, Büyüktaş, Barış, Adler, Mathis, Zavras, Angelos, Papoutsis, Ioannis, Demir, Begüm

Abstract

Federated learning (FL) enables collaborative training of deep learning models across decentralized image archives without requiring data centralization. This paradigm is particularly relevant in remote sensing (RS), where legal regulations, privacy concerns, and bandwidth constraints restrict data sharing. However, the presence of training data heterogeneity across clients (known as non-IID data) can impede convergence and limit the generalization capability of the aggregated global model. To mitigate the adverse effects of training data heterogeneity, vision-language models (VLMs) can be leveraged in FL due to their transferable representations, which have demonstrated robustness under distribution shifts. However, their large parameter size may substantially increase communication overhead and local computational complexity in federated settings. Therefore, it is crucial to select an appropriate VLM adaptation strategy that balances the generalization ability with the communication and computational constraints. To address this issue, in this paper, we present the first comparative study of VLM adaptation strategies for FL in the context of RS image classification. We investigate full fine-tuning, encoder-specific fine-tuning, prompt learning, and low-rank adaptation (LoRA) tuning, and analyze them with respect to three criteria: 1) generalization capability under non-IID data, 2) communication overhead, and 3) local computational complexity. Experiments on BigEarthNet-S2, EuroSAT, RESISC45, and ImageNet reveal distinct trade-offs between task specialization, cross-domain generalization, and efficiency. Based on our findings, we derive a guideline for the selection of an appropriate VLM adaptation strategy in FL for RS image classification under different operational constraints. The code of this work is publicly available at https://git.tu-berlin.de/rsim/FL-RS-VLM.

Chinese Translation

联邦学习（FL）使得在去中心化的图像档案中进行深度学习模型的协作训练成为可能，而无需进行数据集中化。这一范式在遥感（RS）领域尤为重要，因为法律法规、隐私问题和带宽限制都对数据共享构成了限制。然而，客户之间训练数据的异质性（称为非独立同分布数据）可能会妨碍收敛，并限制聚合全局模型的泛化能力。为了减轻训练数据异质性的负面影响，可以在FL中利用视觉语言模型（VLM），因为它们的可迁移表示在分布变化下表现出鲁棒性。然而，它们庞大的参数规模可能会显著增加联邦环境中的通信开销和本地计算复杂性。因此，选择一种合适的VLM适应策略，以平衡泛化能力与通信和计算约束，至关重要。为了解决这一问题，本文首次对RS图像分类背景下的FL中的VLM适应策略进行了比较研究。我们研究了全量微调、编码器特定微调、提示学习和低秩适应（LoRA）调优，并根据三个标准进行分析：1）在非独立同分布数据下的泛化能力，2）通信开销，以及3）本地计算复杂性。在BigEarthNet-S2、EuroSAT、RESISC45和ImageNet上的实验揭示了任务专业化、跨域泛化和效率之间的不同权衡。基于我们的研究结果，我们提出了一项指导方针，用于在不同操作约束下选择适合的FL中RS图像分类的VLM适应策略。本文的代码已公开发布，网址为 https://git.tu-berlin.de/rsim/FL-RS-VLM。

View on arXiv Download PDF AI Translation

cs.CV / 78 / 2608.04810

Segmentation Pre-training for Label-Efficient Lumbar Spine Degeneration Grading

用于标签高效的腰椎退行性病变分级的分割预训练

Maria, Monzon, Andrew, Zisserman, R., Jutzeler Catherine, Amir, Jamaludin

Abstract

Automated assessment of degenerative pathology in the lumbar spine on magnetic resonance imaging (MRI) requires access to large-scale datasets of expert-annotated radiological gradings. In contrast, segmentation pseudo-labels can be generated by automated tools at negligible radiologist cost. We examine whether pre-training on segmentation can effectively replace a fraction of the manual grading annotations required for downstream supervision. We pre-train a 3D ResNet encoder to segment the vertebrae, intervertebral discs (IVDs), and the spinal canal, then fine-tune lightweight task-specific grading heads using different proportions of the available training data, ranging from $10\%$ to $100\%$. On a multicentre dataset of ${\sim}2{,}000$ subjects across 11 pathologies, segmentation pre-training, achieving a Dice score of $0.94$ against pseudo-labels, improved the task-averaged (macro) one-vs-rest ROC-AUC at all proportions. With only 20\% of grading labels after pre-training, the method achieved near full-supervision performance, with the largest gains observed for either low-prevalence or spatially grounded pathologies.

Chinese Translation

在磁共振成像（MRI）上自动评估腰椎的退行性病变需要访问大量专家注释的放射学分级数据集。相比之下，分割伪标签可以通过自动化工具以微不足道的放射科医生成本生成。我们研究了分割预训练是否可以有效替代下游监督所需的一部分手动分级注释。我们预训练了一个3D ResNet编码器，以分割椎骨、椎间盘（IVDs）和脊髓管，然后使用不同比例的可用训练数据（从10%到100%）微调轻量级任务特定的分级头。在一个包含约2000名受试者和11种病理的多中心数据集中，分割预训练在伪标签上取得了0.94的Dice分数，提升了所有比例下的任务平均（宏观）一对多ROC-AUC。在预训练后仅使用20%的分级标签，该方法达到了接近全监督的性能，且在低流行率或空间基础病理的情况下观察到了最大的提升。

View on arXiv Download PDF AI Translation

cs.CV / 79 / 2608.04811

StaticSegFormer: An Efficient High-Performance Semantic Segmentation Based on Static Structured Pruning

StaticSegFormer：一种基于静态结构剪枝的高效高性能语义分割方法

Bartels, Timo, Nazir, Danish, Piewek, Jan, Bagdonat, Thorsten, Fingscheidt, Tim

Abstract

Structured pruning enhances the efficiency of deep neural networks (DNNs) by eliminating groups of parameters during inference. Previous methods mostly reduce computational complexity (FLOPs), while semantic segmentation performance (mIoU) slightly drops. Accordingly, recent dynamic structured pruning methods aim at reducing the performance drop, while lowering the FLOPs even more. However, on the ADE20K and Cityscapes benchmarks, our study reveals that on a GPU platform such dynamic methods exhibit a surprisingly low frame rate far below a simple static approach, while having comparable results in mIoU and FLOPs. To address this issue, we propose a static structured pruning method for attention layers, that achieves both, a lower FLOPs and a high frame rate [fps] of the SegFormer network, the latter increased by up to 34% relative on the Cityscapes dataset, while having no mIoU performance drop at all. Our so-called StaticSegFormer method is strongest for small encoders and large images.

Chinese Translation

结构剪枝通过在推理过程中消除参数组来提高深度神经网络（DNN）的效率。以往的方法主要降低计算复杂度（FLOPs），而语义分割性能（mIoU）略有下降。因此，最近的动态结构剪枝方法旨在减少性能下降，同时进一步降低FLOPs。然而，在ADE20K和Cityscapes基准测试中，我们的研究揭示，在GPU平台上，这些动态方法的帧率远低于简单的静态方法，而在mIoU和FLOPs方面却表现相当。为了解决这个问题，我们提出了一种针对注意力层的静态结构剪枝方法，既实现了更低的FLOPs，又提高了SegFormer网络的帧率，在Cityscapes数据集上相对提高了多达34%，同时没有造成mIoU性能下降。我们所称的StaticSegFormer方法在小型编码器和大图像上表现最为出色。

View on arXiv Download PDF AI Translation

cs.CV / 80 / 2608.04818

Rethinking Pixel Mean Flows via Interval Denoiser

通过区间去噪器重新思考像素均值流

Zaytsev, Alexander, Baranchuk, Dmitry, Korotin, Alexander, Alanov, Aibek

Abstract

Modern diffusion and flow-based models are increasingly moving toward few-step, latent-free generation to bypass the computational overhead of multi-step sampling and the reconstruction bottlenecks of external autoencoders. We propose the Interval Denoiser, a theoretically rigorous framework for latent-free generation. Derived directly from the flow matching ODE, it establishes an exact analytical mapping for intermediate trajectory states. Unlike prior formulations, our prediction is shown to reside on a low-dimensional manifold across any time interval, making the regression tractable for a network operating directly on pixels. Furthermore, by avoiding empirical algebraic substitutions, our formulation correctly isolates the pure time derivative to prevent biased gradient evaluations and ensure exact first-order optimization. By analyzing this objective, we equip our framework with residual clipping and a time-sampling curriculum, enabling effective long-interval training and improving few-step performance. Trained from scratch on ImageNet 256x256, our model achieves an FID of 4.55 in one step (1-NFE) and 3.98 in two steps (2-NFE) without perceptual losses.

Chinese Translation

现代扩散和基于流的模型正越来越多地朝着少步骤、无潜变量生成的方向发展，以绕过多步骤采样的计算开销和外部自编码器的重构瓶颈。我们提出了区间去噪器（Interval Denoiser），这是一个理论上严谨的无潜变量生成框架。该框架直接源自流匹配常微分方程（flow matching ODE），为中间轨迹状态建立了精确的解析映射。与之前的公式不同，我们的预测被证明位于任何时间区间的低维流形上，使得直接在像素上操作的网络的回归变得可行。此外，通过避免经验代数替代，我们的公式正确地隔离了纯时间导数，以防止偏差梯度评估并确保精确的一阶优化。通过分析这一目标，我们为我们的框架配备了残差裁剪和时间采样课程，从而实现有效的长时间间隔训练并提高少步骤性能。在ImageNet 256x256上从头训练，我们的模型在一步（1-NFE）中达到了4.55的FID，在两步（2-NFE）中达到了3.98，且没有感知损失。

View on arXiv Download PDF AI Translation

cs.CV / 81 / 2608.04820

When Diffusion Models Forget Who You Are: Identity Preservation in Face Inpainting under Large Occlusions

当扩散模型忘记你是谁：大遮挡下的人脸修复中的身份保留

Ding, Feng, Xie, Shuhuai, Zhou, Yue, Zhang, Yulan, Zhu, Guopu, Xiao, Mengyao

Abstract

Face inpainting with diffusion models has recently achieved impressive visual quality, yet preserving identity fidelity under significant occlusion and conflicting text guidance remains a major challenge. To address this issue, we present Reference Semantic Inpainting for Face (ReSem-Face), a cascaded diffusion framework that introduces an explicit identity-conditioned semantic prior for multi-reference face inpainting. Our approach distills representative identity features from multiple references to reconstruct missing semantic regions, which then guide the diffusion process through a multi-stream conditioning architecture. This design provides strong semantic constraints when pixels are absent and stabilizes identity reconstruction while remaining compatible with prompt-driven edits. Experiments on CelebAHQ-IDI-5 and VGGFace2 demonstrate that ReSem-Face yields more reliable identity-preserving completion under severe semantic masks and improves text-controlled editing quality compared with representative baselines.

Chinese Translation

使用扩散模型进行人脸修复最近取得了令人印象深刻的视觉质量，但在显著遮挡和冲突文本指导下保持身份一致性仍然是一个主要挑战。为了解决这个问题，我们提出了面向人脸的参考语义修复（Reference Semantic Inpainting for Face，ReSem-Face），这是一种级联扩散框架，引入了一个显式的身份条件语义先验用于多参考人脸修复。我们的方法从多个参考中提取代表性的身份特征，以重建缺失的语义区域，这些区域随后通过多流条件架构指导扩散过程。该设计在像素缺失时提供了强有力的语义约束，并在保持与提示驱动编辑兼容的同时稳定身份重建。在CelebAHQ-IDI-5和VGGFace2上的实验表明，ReSem-Face在严重语义遮罩下能够实现更可靠的身份保留修复，并且与代表性基线相比，改善了文本控制编辑的质量。

View on arXiv Download PDF AI Translation

cs.CV / 82 / 2608.04821

Global Attention-Fused Image Cropping with Attention-Guided and Global-Aligned Crop Evaluator

全球注意力融合图像裁剪：基于注意力引导和全球对齐的裁剪评估器

Yang, Haotian, Yang, Zhile, Lam, Kin-Man, Callet, Patrick Le, Sun, Xin

Abstract

Image cropping aims to improve image aesthetics by preserving important content within an appropriately composed region. However, most existing methods focus primarily on salient regions and therefore have limited sensitivity to the global relationships among the main image components. To address this limitation, we propose Global Attention-Fused Image Cropping (GAFIC), which consists of an Attention-Guided Feature Fusion (AGFF) and a Global-Aligned Crop Evaluator (GACE). AGFF aggregates the importance of local regions to construct a global representation that captures both image structure and local details. GACE aligns candidate crop features with this global representation, enabling crop evaluation to remain sensitive to boundary changes. We further combine three ranking losses across multiple scales to obtain accurate and stable crop scores. Extensive experiments on the GAIC and CPC datasets demonstrate that GAFIC outperforms existing image-cropping methods, particularly in terms of accuracy and stability. Unlike pixel-level retargeting methods such as seam carving, inpainting, and diffusion-based synthesis, GAFIC does not synthesize or modify the retained pixels; instead, it selects an aesthetically preferred crop from the source image, making it suitable for scenarios where pixel integrity and efficient batch processing are important. The source code is available at https://github.com/AIVRC/GAFIC.git.

Chinese Translation

图像裁剪旨在通过保留重要内容于适当构图的区域来提升图像美学。然而，大多数现有方法主要关注显著区域，因此对主要图像组件之间的全球关系敏感性有限。为了解决这一局限性，我们提出了全球注意力融合图像裁剪（Global Attention-Fused Image Cropping，GAFIC），其由注意力引导特征融合（Attention-Guided Feature Fusion，AGFF）和全球对齐裁剪评估器（Global-Aligned Crop Evaluator，GACE）组成。AGFF聚合局部区域的重要性，以构建一个捕捉图像结构和局部细节的全球表示。GACE将候选裁剪特征与该全球表示对齐，使裁剪评估对边界变化保持敏感。我们进一步结合多个尺度的三种排序损失，以获得准确且稳定的裁剪评分。在GAIC和CPC数据集上的大量实验表明，GAFIC在准确性和稳定性方面优于现有的图像裁剪方法。与像缝隙雕刻（seam carving）、修复（inpainting）和基于扩散的合成等像素级重定向方法不同，GAFIC并不合成或修改保留的像素；相反，它从源图像中选择一个美学上更优的裁剪，适用于像素完整性和高效批处理至关重要的场景。源代码可在 https://github.com/AIVRC/GAFIC.git 获取。

View on arXiv Download PDF AI Translation

cs.CV / 83 / 2608.04833

RegisterBridgeMM: A Register-Centric Framework for RGB-Infrared Object Detection

RegisterBridgeMM：一种以注册为中心的RGB-红外物体检测框架

Wang, Zian, Liang, Hangchuan, Chen, Yuehua, Li, Changchun, Guo, Chaoyi, Liu, Mingzhe, Gu, Fangming

Abstract

RGB-infrared (RGB-IR) object detection benefits from complementary visible and thermal cues, but effective fusion remains challenging under illumination changes, weather variation, and cluttered scenes. Existing RGB-IR fusion methods often trade expressive patch-level interaction for lighter but more constrained adaptation mechanisms. We empirically observe that pretrained register tokens contain both modality-shared and modality-specific information on paired RGB-IR inputs, suggesting that they can serve as a compact substrate for cross-modal communication. Building on this observation, we propose RegisterBridgeMM, a register-mediated fusion framework organized as a three-stage register lifecycle. Aggregate preserves per-modality register summarization inherited from pretraining; Bridge performs bidirectional register-to-patch reading with consensus-residual regulation; and Project translates the resulting register summary into spatially adaptive calibration of patch features. This register pathway avoids dense patch-to-patch cross-modal interaction while preserving the pretrained patch representation. With both backbone streams frozen, RegisterBridgeMM achieves the highest mAP50-95 among the evaluated methods on all four benchmarks: LLVIP, M3FD, DroneVehicle, and FLIR-Aligned.

Chinese Translation

RGB-红外（RGB-IR）物体检测受益于可见光和热量线索的互补性，但在光照变化、天气变化和杂乱场景下，有效融合仍然具有挑战性。现有的RGB-IR融合方法往往在表达丰富的补丁级交互与更轻但受限的适应机制之间进行权衡。我们实证观察到，预训练的注册标记包含了配对RGB-IR输入的模态共享和模态特定信息，这表明它们可以作为跨模态通信的紧凑基础。基于这一观察，我们提出了RegisterBridgeMM，一种以注册为媒介的融合框架，组织为三阶段的注册生命周期。Aggregate阶段保留了从预训练继承的每个模态的注册摘要；Bridge阶段执行双向的注册到补丁的读取，并进行共识残差调节；Project阶段将生成的注册摘要转换为补丁特征的空间自适应校准。该注册路径避免了密集的补丁到补丁的跨模态交互，同时保留了预训练的补丁表示。在冻结两个主干流的情况下，RegisterBridgeMM在所有四个基准测试（LLVIP、M3FD、DroneVehicle和FLIR-Aligned）中实现了最高的mAP50-95。

View on arXiv Download PDF AI Translation

cs.CV / 84 / 2608.04840

Towards a satellite image manipulation and deepfake localization benchmark dataset

面向卫星图像操控与深度伪造定位基准数据集的构建

Arndt, Jacob, Varshney, Debvrat, Dias, Philipe, Nukavarapu, Nivedita

Abstract

Verifying the authenticity of satellite imagery has become increasingly critical given advances in generative artificial intelligence. Highly realistic synthetic imagery produced for malicious purposes (deepfakes) can have major consequences in the remote sensing domain, where this data is a fundamental source of information for science applications, planning, logistics, and monitoring. The remote sensing community lacks high-quality, fine-grained manipulation datasets suitable for training and evaluating detection and image forensics algorithms. Existing datasets are lacking and those that do exist either provide no ground truth masks for evaluating manipulation localization, or consist of entire images generated by GANs or diffusion models, which are inadequate for measuring localization performance. To address this gap, we describe a preliminary dataset construction process and prototype benchmark dataset for satellite image manipulation detection and localization. The dataset contains 60 images total, with 30 images carefully manipulated using three manipulation types including copy-paste splicing and diffusion model inpainting, and 30 authentic images. Each image is accompanied by a ground-truth mask and acquisition metadata, enabling both pixel-level localization metrics, image metadata studies, and analyses of how manipulation detection performance relates to image collection parameters. We describe the dataset construction process and present this initial release to support further research in image forensics and geospatial deepfake detection. The prototype dataset can be downloaded at https://huggingface.co/datasets/geodf/fmow-fake-small.

Chinese Translation

鉴于生成性人工智能的进步，验证卫星图像的真实性变得愈发重要。为恶意目的生成的高度逼真的合成图像（深度伪造）在遥感领域可能产生重大影响，因为这些数据是科学应用、规划、物流和监测的基本信息来源。遥感社区缺乏高质量、细粒度的操控数据集，适合用于训练和评估检测及图像取证算法。现有的数据集不足，那些存在的数据集要么没有用于评估操控定位的真实标签，要么由生成对抗网络（GANs）或扩散模型生成的完整图像组成，这对于测量定位性能是不够的。为了解决这一空白，我们描述了一种初步的数据集构建过程和卫星图像操控检测与定位的原型基准数据集。该数据集总共包含60幅图像，其中30幅图像经过精心操控，采用了包括复制粘贴拼接和扩散模型修复在内的三种操控类型，另外30幅为真实图像。每幅图像都附有真实标签和获取元数据，使得能够进行像素级定位指标、图像元数据研究，以及操控检测性能与图像采集参数之间关系的分析。我们描述了数据集的构建过程，并发布了这一初步版本，以支持图像取证和地理空间深度伪造检测的进一步研究。该原型数据集可在 https://huggingface.co/datasets/geodf/fmow-fake-small 下载。

View on arXiv Download PDF AI Translation

cs.CV / 85 / 2608.04865

Cooking beyond Frames: A Stereo Event Camera Dataset in the Kitchen

超越框架的烹饪：厨房中的立体事件相机数据集

Feng, Chengming, Araghi, Hesam, Zheng, Liming, Dupeyroux, Julien, Zhang, Xucong, van Gemert, Jan, Tömen, Nergis

Abstract

Event cameras, also known as neuromorphic cameras, have gained significant attention in recent years due to their high temporal resolution, high dynamic range, and low power consumption. While many studies and datasets in neuromorphic vision have focused on automotive and drone applications, human-centric daily-life scenarios remain largely underrepresented, despite their importance for developing and benchmarking event-based perception systems. Moreover, the few existing event-based human activity datasets are typically recorded with scripted human actions, limiting their ability to capture natural human behaviors. In this paper, we introduce EventKitchen, a large-scale stereo event camera benchmark dataset of human cooking activities in the kitchen. EventKitchen is egocentrically collected from 10 participants in 13 diverse kitchens, where the participants wear a helmet with multiple sensors and naturally perform cooking activities, without any scripted actions. EventKitchen comprises 5.5 hours of stereo event recordings with synchronized RGB, depth, and IMU data. We provide human annotations for 10,762 action segments and 13,482 bounding boxes. We train baseline models on EventKitchen to perform multiple event-based tasks, including action recognition, object detection, and stereo depth estimation. By capturing natural, real-world human activities, EventKitchen establishes a challenging benchmark for neuromorphic vision beyond autonomous driving.

Chinese Translation

事件相机，也称为神经形态相机，近年来因其高时间分辨率、高动态范围和低功耗而受到广泛关注。尽管许多神经形态视觉的研究和数据集集中于汽车和无人机应用，但以人为中心的日常生活场景仍然在很大程度上被忽视，尽管这些场景对于开发和基准测试基于事件的感知系统至关重要。此外，现有的少数基于事件的人类活动数据集通常是通过脚本化的人类动作记录的，限制了它们捕捉自然人类行为的能力。在本文中，我们介绍了EventKitchen，这是一个大规模的立体事件相机基准数据集，记录了厨房中的人类烹饪活动。EventKitchen是从10名参与者在13个不同厨房中以自我中心的方式收集的，参与者佩戴多个传感器的头盔，自然地进行烹饪活动，而没有任何脚本化的动作。EventKitchen包含5.5小时的立体事件录音，并同步了RGB、深度和IMU数据。我们为10,762个动作片段和13,482个边界框提供了人工标注。我们在EventKitchen上训练基线模型，以执行多个基于事件的任务，包括动作识别、物体检测和立体深度估计。通过捕捉自然的、现实世界中的人类活动，EventKitchen为超越自动驾驶的神经形态视觉建立了一个具有挑战性的基准。

View on arXiv Download PDF AI Translation

cs.CV / 86 / 2608.04866

Persistent Object Narratives for Token-Efficient Video Language Models

用于高效视频语言模型的持久对象叙事

Chen, Junzhe, Meng, Siyuan, Guo, Xiaojie

Abstract

Video large language models (Video-LLMs) have made strong progress in open-ended video understanding. However, their visual interfaces remain token-intensive and provide limited explicit structure for linking recurring object evidence across time. We introduce SlotNarrative, a slot-based interface that organizes a video into persistent object narratives represented by compact object-state tokens. Rather than compressing frame-wise features before establishing temporal correspondence, SlotNarrative first groups visual features into object-like slots and then associates recurring observations with clip-level object entries through a lightweight, parameter-free memory that integrates multiple complementary matching cues. Each retained entry is serialized into two token types: an identity token that summarizes persistent object appearance and a set of state tokens that encode segment-level appearance, geometry, visibility, and trajectory information. This design yields an interface of only 144 allocated visual-token positions for a frozen Video-LLM, independent of the number of sampled frames. Across multiple datasets, SlotNarrative achieves a favorable trade-off between accuracy and visual-token count compared with prior compact Video-LLM interfaces. Experimental results establish persistent object narratives as a compact, structured, and temporally organized visual interface for Video-LLMs. Our code will be made publicly available.

Chinese Translation

视频大型语言模型（Video-LLMs）在开放式视频理解方面取得了显著进展。然而，它们的视觉接口仍然是以令牌为中心，并且在跨时间链接重复对象证据方面提供的明确结构有限。我们提出了SlotNarrative，这是一种基于槽的接口，将视频组织为由紧凑的对象状态令牌表示的持久对象叙事。SlotNarrative首先将视觉特征分组为类似对象的槽，然后通过一种轻量级、无参数的记忆机制将重复观察与剪辑级对象条目关联起来，该机制整合了多种互补的匹配线索，而不是在建立时间对应关系之前压缩逐帧特征。每个保留的条目被序列化为两种令牌类型：一个身份令牌用于总结持久对象的外观，以及一组状态令牌用于编码段级外观、几何形状、可见性和轨迹信息。这种设计为冻结的Video-LLM提供了仅144个分配的视觉令牌位置，与采样帧的数量无关。在多个数据集上，SlotNarrative在准确性和视觉令牌数量之间实现了良好的权衡，相较于之前的紧凑型Video-LLM接口。实验结果确立了持久对象叙事作为Video-LLMs的紧凑、结构化和时间组织的视觉接口。我们的代码将公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 87 / 2608.04885

Evaluating the Diagnostic Robustness of Vision-Language Models Under Visual and Textual Perturbations

在视觉和文本扰动下评估视觉-语言模型的诊断鲁棒性

Khoramfar, Ali, Dousti, Mohammad Javad, Mohamadian, Alireza, Faili, Heshaam

Abstract

Standard accuracy metrics for VLMs often mask significant reliability failures in sensitive domains. In this work, we utilize a histopathology-validated brain MRI dataset to systematically assess the diagnostic robustness of four VLM families under evidence-preserving perturbations. By reordering anatomical slices and swapping target label positions, we evaluate whether models maintain consistent predictions when clinical evidence remains invariant. Our results reveal significant vulnerabilities in presentation-order stability, with models exhibiting prediction flips in up to 48.9% of cases under simple sequence reversals. We further identify a textual selection bias, where label reordering triggers inconsistent diagnoses in up to 67.8% of cases despite identical visual inputs. Negative-control tests further reveal diagnostic overcommitment: models generate categorical diagnoses in up to 76.1% of cases after expert-annotated lesion slices are removed. These results demonstrate that high accuracy can overestimate clinical reliability, masking sensitivity to sequential presentation and textual framing that is not captured by aggregate accuracy. Our findings highlight the necessity of stability-based metrics for the deployment of VLMs in safety-critical clinical applications. Our evaluation data and code will be made public upon acceptance.

Chinese Translation

标准的视觉-语言模型（VLMs）准确性指标常常掩盖在敏感领域中的显著可靠性失败。在本研究中，我们利用经过组织病理学验证的脑MRI数据集，系统地评估四个VLM家族在保持证据不变的扰动下的诊断鲁棒性。通过重新排序解剖切片和交换目标标签位置，我们评估模型在临床证据保持不变时是否能够保持一致的预测。我们的结果揭示了在呈现顺序稳定性方面的显著脆弱性，模型在简单序列反转下的预测翻转率高达48.9%。我们进一步识别出一种文本选择偏差，即标签重新排序在高达67.8%的情况下触发不一致的诊断，尽管视觉输入相同。负对照测试进一步揭示了诊断过度承诺：在去除专家注释的病变切片后，模型在高达76.1%的情况下生成分类诊断。这些结果表明，高准确性可能高估临床可靠性，掩盖了对顺序呈现和文本框架的敏感性，而这些在整体准确性中并未体现。我们的研究结果强调了在安全关键的临床应用中部署VLMs时需要基于稳定性的指标。我们的评估数据和代码将在论文接受后公开。

View on arXiv Download PDF AI Translation

cs.CV / 88 / 2608.04887

STEP-OPD: Rethinking Output Targets and Internal Dynamics in On-Policy Distillation for Diffusion Models

STEP-OPD：重新思考扩散模型中策略蒸馏的输出目标和内部动态

Wei, Qingyan, Li, Guangzhao, Tu, Xiaobing, Wang, Yinggui, Zhang, Xiantao, Ren, Jinkui, Liu, Xiaohong, Zhang, Linfeng

Abstract

On-policy distillation (OPD) has become an effective approach for consolidating multiple task-specialized image generation models into a single student. However, existing OPD methods optimize the student mainly to match the teacher's output velocity, making the teacher the upper limit of the optimization objective. While output-level supervision alone leaves the student's blockwise representation evolution underconstrained, which weakens the transfer of capabilities that must be progressively developed across layers. We propose STEP-OPD, an on-policy distillation framework for image generation that extends the student's learning target beyond the teacher and introduces explicit constraints on its internal representation evolution. Instead of treating the teacher as the final target, we use the velocity difference between each task-specific teacher and the shared base model as a direction for further learning and add a scaled version of this difference to the teacher velocity. In addition, we align the direction and magnitude of representation changes between the student and teacher, enabling the student to learn how representations are progressively transformed across network blocks. Experiments on compositional alignment, text rendering, and human preference show that our method consistently improves Standard OPD methods. In particular, it increases the GenEval score of DiffusionOPD from 0.927 to 0.961, while also improving OCR and all preference-based metrics. The resulting unified student surpasses the corresponding single-task teachers across all three capability groups, showing that output extrapolation enables beyond-teacher learning. And representation change alignment provides complementary guidance for the student's internal transformations.

Chinese Translation

策略蒸馏（On-policy distillation, OPD）已成为将多个任务专用的图像生成模型整合为单一学生模型的有效方法。然而，现有的OPD方法主要优化学生模型以匹配教师模型的输出速度，使教师模型成为优化目标的上限。仅依赖输出级别的监督使得学生模型的块状表示演化受到约束不足，从而削弱了必须在层间逐步发展的能力的转移。我们提出了STEP-OPD，一种用于图像生成的策略蒸馏框架，扩展了学生的学习目标超越教师，并对其内部表示演化引入了明确的约束。我们不再将教师视为最终目标，而是利用每个任务特定教师与共享基础模型之间的速度差异作为进一步学习的方向，并将该差异的缩放版本添加到教师的速度中。此外，我们对学生与教师之间表示变化的方向和幅度进行对齐，使学生能够学习表示如何在网络块之间逐步转变。在组合对齐、文本渲染和人类偏好方面的实验表明，我们的方法始终优于标准OPD方法。特别是，它将DiffusionOPD的GenEval得分从0.927提高到0.961，同时改善了OCR和所有基于偏好的指标。最终得到的统一学生模型在所有三个能力组中超越了相应的单任务教师，表明输出外推使得超越教师的学习成为可能。而表示变化对齐为学生的内部转变提供了补充指导。

View on arXiv Download PDF AI Translation

cs.CV / 89 / 2608.04902

Visual Representation Matters: Exploiting Temporal Differences in Video-to-Audio Generation

视觉表征的重要性：利用视频到音频生成中的时间差异

Chen, Zehua, Wang, Junyou, Jiang, Yuxuan, Fang, Zhenying, Dai, Yusheng, Chen, Jianfei, Liu, Ziwei, Zhu, Jun

Abstract

Video-to-audio (V2A) generation extends image-to-audio generation (I2A) by introducing consecutive frames that provide essential temporal cues for audio synthesis. However, existing conditional diffusion-based V2A methods typically enhance visual conditioning with additional audio-visual supervision, acoustic structure prediction, or reasoning from large multimodal models, requiring extra networks or strong inductive biases. Inspired by recent advances in visual representation learning, we introduce TD-V2A, which leverages temporal differences (TD) as the key representation that distinguishes V2A from I2A, enriching visual conditioning with minimal architectural modification. We first investigate TD at both the frame and feature levels to identify the most effective representation level at which TD complements visual representations. Based on these findings, we develop a hierarchically continual learning strategy and an annealed temporal differences guidance method to progressively learn and exploit TD information during diffusion training and sampling process, respectively. Extensive experiments on benchmark datasets demonstrate that effectively exploiting TD through our proposed framework significantly improves end-to-end V2A generation quality, even outperforming dedicated V2A representations such as contrastive audio-visual pretraining.

Chinese Translation

视频到音频（V2A）生成通过引入连续帧，扩展了图像到音频生成（I2A），为音频合成提供了重要的时间线索。然而，现有的基于条件扩散的V2A方法通常通过额外的音视频监督、声学结构预测或从大型多模态模型推理来增强视觉条件，这需要额外的网络或强烈的归纳偏差。受到视觉表征学习近期进展的启发，我们提出了TD-V2A，利用时间差异（TD）作为区分V2A与I2A的关键表征，以最小的架构修改丰富视觉条件。我们首先在帧级和特征级上研究TD，以确定TD补充视觉表征的最有效表征层次。基于这些发现，我们开发了一种分层持续学习策略和一种退火时间差异引导方法，分别在扩散训练和采样过程中逐步学习和利用TD信息。对基准数据集的广泛实验表明，通过我们提出的框架有效利用TD显著提高了端到端V2A生成的质量，甚至超越了专门的V2A表征，如对比音视频预训练。

View on arXiv Download PDF AI Translation

cs.CV / 90 / 2608.04906

Enhancing Low Back Pain Assessment with Diffusion Models for Lumbar Spine MRI Segmentation

利用扩散模型增强低背痛评估的腰椎MRI分割

Monzon, Maria, Iff, Thomas, Konukoglu, Ender, Jutzeler, Catherine R.

Abstract

This study introduces a diffusion-based framework for robust and accurate semantic segmentation of lumbar spine MRI scans from patients with low back pain (LBP), regardless of whether the scans are T1- or T2-weighted. We compared with advanced models for segmenting vertebrae, intervertebral discs (IVDs), and spinal canal using the SPIDER dataset. The results showed that SpineSegDiff achieved a segmentation performance comparable to that of the state-of-the-art non-diffusion nnUnet, particularly in improving the identification of degenerated IVDs. In addition, the uncertainty maps generated by our model provide valuable insights for clinical review, enhancing the robustness and reliability of the segmentation results. The potential of diffusion models to enhance the diagnosis and management of LBP through more precise analysis of pathological spine MRI is underscored by our findings.

Chinese Translation

本研究提出了一种基于扩散的框架，用于对低背痛（LBP）患者的腰椎MRI扫描进行稳健且准确的语义分割，无论扫描是T1加权还是T2加权。我们使用SPIDER数据集与先进模型进行了比较，以分割椎骨、椎间盘（IVDs）和脊髓管。结果表明，SpineSegDiff在分割性能上与最先进的非扩散模型nnUnet相当，特别是在改善退化IVDs的识别方面。此外，我们模型生成的不确定性图为临床审查提供了宝贵的见解，增强了分割结果的稳健性和可靠性。我们的研究结果强调了扩散模型通过对病理性脊柱MRI进行更精确分析来增强LBP诊断和管理的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 91 / 2608.04917

An active-learning framework for real-time depth perception from monocular vision streams

基于单目视觉流的实时深度感知主动学习框架

Zeng, Xiaorong, Chen, Weiqiang, Shi, Peng, Su, Liang, Wang, Zirui, Ji, Xuewu, Shen, Shuiwen

Abstract

Biological visual systems can perceive depth from monocular vision flow, continuously integrating temporal visual cues while maintaining a balance between stability and plasticity in dynamic environments. In contrast, artificial perception models deployed on resource-constrained edge devices are typically trained in a static offline manner and remain frozen after deployment, often suffering severe performance degradation under domain shifts. While large-scale models may encode broad knowledge through massive parameter redundancy, lightweight networks face a static optimization dilemma: forcing compact models to learn universal geometric representations is computationally inefficient and often leads to performance saturation. To resolve this issue, an Online Active Learning (OAL) mechanism is introduced to endow compact neural networks with the capability to adapt continuously during operation. A closed-loop Predict-Evaluate-Correct learning paradigm is established to actively select high-confidence, information-rich signals from streaming visual input. Crucially, Elastic Weight Consolidation (EWC) is employed not merely to prevent catastrophic forgetting, but to enforce Selective Plasticity, preserving parameters that encode globally relevant structural knowledge while allowing local alignment to newly observed environments. Built upon a MobileNetV3-Small backbone, the proposed system achieves approximately a 75% reduction in computational cost while maintaining competitive depth estimation accuracy. Experimental results demonstrate that adaptability is not solely determined by model size, but rather by how effectively parameter plasticity is regulated in dynamic environments.

Chinese Translation

生物视觉系统能够从单目视觉流中感知深度，持续整合时间视觉线索，同时在动态环境中保持稳定性与可塑性之间的平衡。相比之下，部署在资源受限的边缘设备上的人工感知模型通常以静态离线方式进行训练，并在部署后保持不变，常常在领域转移下遭受严重的性能下降。虽然大规模模型可能通过大量参数冗余编码广泛的知识，但轻量级网络面临静态优化困境：强迫紧凑模型学习通用几何表示在计算上效率低下，且往往导致性能饱和。为了解决这一问题，提出了一种在线主动学习（Online Active Learning, OAL）机制，使紧凑神经网络在运行过程中具备持续适应的能力。建立了一个闭环的预测-评估-修正学习范式，主动选择来自流媒体视觉输入的高置信度、信息丰富的信号。关键的是，采用弹性权重巩固（Elastic Weight Consolidation, EWC）不仅防止灾难性遗忘，还强制实施选择性可塑性，保留编码全球相关结构知识的参数，同时允许局部对新观察到的环境进行调整。基于MobileNetV3-Small骨干网，所提系统在保持竞争性深度估计精度的同时，实现了约75%的计算成本降低。实验结果表明，适应性并非仅由模型大小决定，而是由在动态环境中参数可塑性调节的有效性所决定。

View on arXiv Download PDF AI Translation

cs.CV / 92 / 2608.04935

Unleashing the Potential of Vision-Language Models for Generalizable AI-Generated Image Detection

释放视觉-语言模型在可泛化AI生成图像检测中的潜力

Cai, Weihan, Tan, Hao, Tan, Zichang, Wan, Jun, Gao, Xinping

Abstract

Recent work has shown that a simple linear probe on frozen representations from modern vision foundation models (VFMs) can achieve state-of-the-art AIGI detection performance, substantially outperforming specialized detectors in challenging in-the-wild scenarios. This finding has established DINOv3 as the dominant foundation-model baseline for subsequent improvements. However, we find that the vision-language model Perception Encoder (PE) holds greater potential for AIGI detection, because its language-aligned representation preserves high-level provenance semantics. Specifically, PE exhibits stronger local provenance organization than DINOv3 in its frozen feature space. However, semantic-agnostic linear probing fails to exploit this structure, as PE-Linear still underperforms DINOv3-Linear by 4.1% on In-the-Wild. Based on this observation, we propose Semantic Prototype Calibration (SPC), which constructs category prototypes from forensic semantic information and calibrates them with supervised data. We apply SPC to PE and refer to the resulting detector as PE-SPC. Our analysis shows that this simple design achieves stronger generalization. Across cross-generator, post-processing, and in-the-wild benchmarks, PE-SPC surpasses the previous DINOv3 baseline and achieves new state-of-the-art results.

Chinese Translation

近期研究表明，对现代视觉基础模型（VFM）冻结表示进行简单线性探测可以实现最先进的AI生成图像（AIGI）检测性能，显著优于在复杂的野外场景中专门检测器的表现。这一发现确立了DINOv3作为后续改进的主导基础模型基准。然而，我们发现视觉-语言模型感知编码器（Perception Encoder, PE）在AIGI检测中具有更大的潜力，因为其语言对齐的表示保留了高级来源语义。具体而言，PE在其冻结特征空间中表现出比DINOv3更强的局部来源组织。然而，语义无关的线性探测未能利用这一结构，因为PE-Linear在野外场景中仍比DINOv3-Linear低4.1%的性能。基于这一观察，我们提出了语义原型校准（Semantic Prototype Calibration, SPC），该方法从法医学语义信息中构建类别原型，并利用监督数据对其进行校准。我们将SPC应用于PE，并将生成的检测器称为PE-SPC。我们的分析表明，这一简单设计实现了更强的泛化能力。在跨生成器、后处理和野外基准测试中，PE-SPC超越了之前的DINOv3基准，并取得了新的最先进结果。

View on arXiv Download PDF AI Translation

cs.CV / 93 / 2608.04949

UG-UMRE: Uncertainty-Guided Modality Augmentation and Distributional Calibration for Unified Multimodal Relation Extraction

UG-UMRE：基于不确定性的模态增强与分布校准的统一多模态关系提取

Kong, Bo, Jia, Liruiz, Liang, Yi, Liu, Chao, Han, Dongfang, Yan, Tianwei, Liu, Yuan, Liu, Shengquan

Abstract

Unified Multimodal Relation Extraction (UMRE) aims to identify intra-modal and cross-modal relations between textual entities and visual objects. However, existing UMRE studies still encounter two critical issues: ignoring inherent aleatoric uncertainty causes noise propagation, and deep-seated heterogeneity between distinct modal distributions hinders alignment. To address these issues, we propose the Uncertainty-Guided UMRE Network (UG-UMRE). Specifically, we design an Uncertainty-Driven Unimodal Augmentation (UDUA) module, which models features as Gaussian distributions based on the Variational Information Bottleneck. By incorporating an uncertainty-aware self-supervised contrastive learning mechanism, UDUA effectively filters out noise while maintaining semantic consistency. Furthermore, we introduce the Joint Aleatoric Uncertainty Alignment (JAUA) module as a global semantic pre-calibration mechanism. JAUA leverages probabilistic distribution consistency to construct a shared latent space, eliminating the distributional gap by synchronizing cross-modal statistical properties, thereby laying a robust foundation for fine-grained interaction. Experiments on three benchmark datasets (UMRE, MORE, and MNRE) demonstrate that UG-UMRE achieves state-of-the-art performance. Further analysis validates the pluggable and effective performance of the proposed UDUA and JAUA modules.

Chinese Translation

统一多模态关系提取（UMRE）旨在识别文本实体与视觉对象之间的内模态和跨模态关系。然而，现有的UMRE研究仍面临两个关键问题：忽视固有的随机不确定性导致噪声传播，而不同模态分布之间的深层异质性阻碍了对齐。为了解决这些问题，我们提出了不确定性引导的UMRE网络（UG-UMRE）。具体而言，我们设计了一个基于不确定性的单模态增强（UDUA）模块，该模块基于变分信息瓶颈将特征建模为高斯分布。通过结合不确定性感知的自监督对比学习机制，UDUA有效地过滤噪声，同时保持语义一致性。此外，我们引入了联合随机不确定性对齐（JAUA）模块，作为全局语义预校准机制。JAUA利用概率分布一致性构建共享潜在空间，通过同步跨模态统计特性消除分布间的差距，从而为细粒度交互奠定坚实基础。在三个基准数据集（UMRE、MORE和MNRE）上的实验表明，UG-UMRE达到了最先进的性能。进一步的分析验证了所提UDUA和JAUA模块的可插拔性和有效性。

View on arXiv Download PDF AI Translation

cs.CV / 94 / 2608.04955

Towards Valid B-Rep Generation: Training-Free Wireframe Anomaly Detection and Repair

朝向有效的边界表示生成：无训练的线框异常检测与修复

Wu, Jingyu, Cai, Youcheng, Luo, Tengyu, Liu, Ligang

Abstract

Multi-stage boundary representation (B-Rep) generation leverages intermediate wireframes to synthesize CAD models. However, geometric and topological risks in these wireframes -- such as self-intersections, edge collapses, and disconnected vertices -- can propagate to invalid final B-Reps. Mitigating such failures by retraining large generative models is computationally prohibitive. We propose Wireframe Detection and Repair (WDR), a training-free framework that intervenes at the intermediate wireframe stage to improve downstream B-Rep validity. WDR features a Geometric-Topology Anomaly Detector (GTAD) that combines parallel VLM-based coarse screening with geometric and topological detectors to predict downstream invalidity risk and route generation to dedicated branches. An Energy-Guided Geometric-Topology Repair (EGGTR) module then performs detector-triggered guided regeneration through geometry and topology branches. By scaling test-time computation via Energy-Guided Resampling and training-free guidance for diffusion models, WDR can be integrated into autoregressive and diffusion pipelines without retraining. Extensive experiments demonstrate consistent improvements in kernel-checked validity while largely retaining the measured diversity and distributional quality of synthesized CAD models. The code will be made publicly available upon acceptance.

Chinese Translation

多阶段边界表示（B-Rep）生成利用中间线框合成计算机辅助设计（CAD）模型。然而，这些线框中的几何和拓扑风险——例如自交、边缘塌陷和断开顶点——可能会传播到无效的最终 B-Rep。通过重新训练大型生成模型来缓解这些失败在计算上是不可行的。我们提出了线框检测与修复（WDR），这是一个无训练的框架，在中间线框阶段进行干预，以提高下游 B-Rep 的有效性。WDR 具有几何-拓扑异常检测器（GTAD），该检测器结合了基于 VLM 的并行粗筛选与几何和拓扑检测器，以预测下游无效风险并将生成过程引导至专用分支。然后，能量引导的几何-拓扑修复（EGGTR）模块通过几何和拓扑分支执行检测器触发的引导再生。通过能量引导重采样和无训练的扩散模型指导来扩展测试时计算，WDR 可以在不重新训练的情况下集成到自回归和扩散管道中。大量实验表明，在保持合成 CAD 模型的多样性和分布质量的同时，内核检查的有效性有了一致的提升。代码将在接受后公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 95 / 2608.04956

ContextMaster: Interactive Multi-Shot Video Creation via Fixed-Budget Sparse Context Routing

ContextMaster：通过固定预算稀疏上下文路由进行交互式多镜头视频创作

Guo, Xu, Wei, Zhengxuan, Li, Xinghui, Huang, Hanzhuo, Liu, Xinyu, Luo, Xiangyang, Wei, Min, Zhu, Yiran, Wang, Qiulin, Xu, Yulong, Wang, Xintao, Wan, Pengfei, Fan, Qi, Hou, Xiangwang

Abstract

Recent video models increasingly support generation, reference conditioning, and editing within a single model, yet typically expose them as separate operations over fixed inputs. Practical creation unfolds across multiple shots, requiring one model to generate from text, follow a reference, or edit source footage while maintaining shared history. We formalize this setting as interactive multi-shot video creation (IMVC) and introduce ContextMaster, a unified model with a role-aware context representation for these operations. An interactive model must retain access to an expanding history without allowing the context read cost at each denoising step to grow. ContextMaster combines reusable clean context states with fixed budget sparse context routing and uses ConstraintSink to keep task constraints visible. To address the dual challenges of sparse context access and inference with few denoising steps, we propose a two-stage privileged context distillation framework, which transfers full context behavior from a dense teacher through consistency distillation and then refines deployment rollouts with distribution matching. Experiments on the three primitive tasks demonstrate improved task fulfillment and consistency across shots over specialized baselines. User studies further validate flexibly composed workflows, while the model reaches 16 FPS on a single GPU.

Chinese Translation

近期的视频模型越来越支持在单一模型中进行生成、参考条件化和编辑，但通常将其视为对固定输入的独立操作。实际创作过程涉及多个镜头，需要一个模型能够从文本生成、遵循参考或编辑源视频，同时保持共享历史。我们将这一设置形式化为交互式多镜头视频创作（IMVC），并引入ContextMaster，一个统一模型，具备角色感知的上下文表示以支持这些操作。交互式模型必须能够访问不断扩展的历史，同时不允许每个去噪步骤的上下文读取成本增加。ContextMaster结合可重用的干净上下文状态与固定预算的稀疏上下文路由，并使用ConstraintSink保持任务约束的可见性。为了解决稀疏上下文访问和少量去噪步骤推理的双重挑战，我们提出了一种两阶段特权上下文蒸馏框架，通过一致性蒸馏从密集教师模型转移完整的上下文行为，然后通过分布匹配精炼部署结果。对三项基本任务的实验表明，与专门基线相比，任务完成度和镜头间的一致性得到了改善。用户研究进一步验证了灵活组合的工作流程，同时该模型在单个GPU上达到了16帧每秒的速度。

View on arXiv Download PDF AI Translation

cs.CV / 96 / 2608.04995

Promptable Animal Pose Tracking Across Species

跨物种的可提示动物姿态追踪

Li, Le, Ivanova, Daniela, Pugeault, Nicolas

Abstract

Animal pose estimation and tracking is important for wildlife monitoring and conservation research, and with limited expert time for labelling automated approaches are imperative. While human pose estimation and tracking has seen rapid progress thanks to large annotated datasets, animal pose remain challenging, due to large morphological and behavioural differences between species and limited annotated data. Existing approaches either optimise generic keypoint localisation from annotated datasets (such as APTv2) with poor generalisation, or track custom keypoints using visual tracking, at the cost of performance. In this paper, we demonstrate that vision foundation models trained on large datasets can be used effectively to track animal pose with limited labelled data. We propose two models, one unsupervised and the other supervised, to track user-selected keypoints in videos. The supervised approach delivers superior tracking accuracy by employing a keypoint prompt encoder to explicitly inject structural priors from a reference frame into feature matching. In parallel, the unsupervised route provides strong cross-species robustness by leveraging diverse foundation-model features for training-free correspondence matching. Extensive evaluation on challenging animal video benchmarks APTv2 and TigDog demonstrates that our framework achieves strong performance while maintaining an effective balance between accuracy and generalisation, offering a practical solution for real-world animal behaviour analysis and conservation applications.

Chinese Translation

动物姿态估计和追踪对于野生动物监测和保护研究至关重要，而由于专家标注时间有限，自动化方法变得不可或缺。尽管由于大规模标注数据集，人类姿态估计和追踪取得了快速进展，但动物姿态估计仍然面临挑战，这主要是由于物种之间存在较大的形态和行为差异以及标注数据的稀缺。现有方法要么优化来自标注数据集的通用关键点定位（如 APTv2），但泛化能力较差，要么使用视觉追踪来跟踪自定义关键点，导致性能下降。本文展示了在大规模数据集上训练的视觉基础模型可以有效地在有限标注数据的情况下追踪动物姿态。我们提出了两个模型，一个是无监督的，另一个是监督的，用于在视频中追踪用户选择的关键点。监督方法通过采用关键点提示编码器，显式地将参考帧中的结构先验注入特征匹配，从而提供了更高的追踪精度。同时，无监督方法通过利用多样的基础模型特征进行无训练的对应匹配，提供了强大的跨物种鲁棒性。在具有挑战性的动物视频基准 APTv2 和 TigDog 上的广泛评估表明，我们的框架在保持准确性和泛化能力之间有效平衡的同时，取得了强劲的性能，为现实世界的动物行为分析和保护应用提供了实用解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 97 / 2608.05000

Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes

迈向多模态预训练的物理学：知识流动、模态协同、早期统一与方法论

Han, Junlin, Tong, Shengbang, Fan, David, Chen, Minghao, Torr, Philip, Kokkinos, Filippos, Lewis, Mike

Abstract

Vision offers a critical axis for advancing foundation models, driving a shift towards natively unified multimodal pretraining. Despite this momentum, the design space and the fundamental mechanisms of how modalities interact during unified training remain underexplored. We provide empirical clarity through a systematic exploration of multimodal pretraining. Our controlled experiments on both synthetic and large-scale real-world datasets yield four key insights into the physics of multimodal pretraining: (i) Knowledge Flow: We disentangle how language, visual understanding, and visual generation transfer knowledge across modalities, revealing distinct patterns of influence and asymmetry; (ii) Synergy vs. Competition: We show that data "complexity" largely determines whether modalities are synergistic, identify architectural choices that promote synergy: such as shared attention and normalization with modality-specific feed-forward layers, and find that these behaviors generalize across different visual tokenizer designs; (iii) Early Unification: Unifying modalities from the very early stages and training them jointly is shown to be more effective than late alignment or sequential training. This process uncovers a vision laziness phenomenon, where delayed integration leads models to rely on language priors; (iv) Recipes: We derive efficient pretraining recipes that achieve strong generative performance using only 5% of the compute budget. These core findings are subsequently validated at scale by training multiple 13.5B MoE models on 2T tokens. We hope this study provides a principled foundation for understanding and scaling multimodal pretraining.

Chinese Translation

视觉为推动基础模型的进步提供了一个关键轴心，推动了向本质上统一的多模态预训练的转变。尽管这一趋势正在发展，但模态在统一训练过程中如何相互作用的设计空间和基本机制仍然未被充分探索。我们通过系统性地探索多模态预训练提供了实证上的清晰性。我们在合成数据集和大规模真实世界数据集上的受控实验得出了关于多模态预训练物理学的四个关键见解：(i) 知识流动：我们解构了语言、视觉理解和视觉生成如何跨模态传递知识，揭示了影响和不对称的不同模式；(ii) 协同与竞争：我们表明数据的“复杂性”在很大程度上决定了模态是否具有协同效应，识别出促进协同的架构选择，例如共享注意力和具有模态特定前馈层的归一化，并发现这些行为在不同视觉标记器设计中具有普适性；(iii) 早期统一：从非常早期阶段统一模态并共同训练被证明比后期对齐或顺序训练更有效。这个过程揭示了一种视觉懒惰现象，延迟整合导致模型依赖语言先验；(iv) 方法论：我们推导出有效的预训练方法，仅使用5%的计算预算即可实现强大的生成性能。这些核心发现随后通过在2T标记上训练多个13.5B MoE模型进行了规模验证。我们希望这项研究为理解和扩展多模态预训练提供一个有原则的基础。

View on arXiv Download PDF AI Translation

cs.CV / 98 / 2608.05049

OmniEdit-Bench: A Comprehensive Benchmark for Instruction-based Video Editing

OmniEdit-Bench：基于指令的视频编辑综合基准测试

Miao, Chenxuan, Feng, Yutong, Lu, Yi, Yan, Yunfeng, Qi, Donglian, Zhang, Shiwei, Liu, Yu, Chen, Xi, Zhao, Hengshuang

Abstract

Instruction-based video editing (IVE) is an emerging field with broad applications, yet evaluating editing models remains challenging. Existing benchmarks suffer from two major limitations: limited task coverage inherited from image editing, which overlooks video-specific dimensions, and inadequate metrics that fail to measure instruction fidelity, allowing incorrect edits to receive high scores due to strong visual priors from the original video. To address these issues, we introduce a comprehensive and structured benchmark for IVE. Our benchmark decomposes editing tasks into multiple video-specific dimensions, including spatial, temporal, audio, and reference-based editing, extending beyond conventional frame-level evaluation. It also distinguishes explicit and implicit instructions and incorporates reasoning-based scenarios to better reflect real-world requirements. Furthermore, we propose an evaluation framework that assesses editing quality from four complementary dimensions: accuracy, preservation, realism, and consistency, using both human judgments and state-of-the-art vision-language models. To emphasize instruction fidelity, we introduce an accuracy-aware penalty mechanism that conditions other scores on accuracy, preventing visually plausible but incorrect edits from receiving inflated evaluations. Extensive experiments on representative open-source and commercial models show that current IVE models remain far from satisfactory. OmniEdit-Bench provides a comprehensive and reliable testbed for evaluating instruction-based video editing and offers insights into future research directions.

Chinese Translation

基于指令的视频编辑（Instruction-based Video Editing, IVE）是一个新兴领域，具有广泛的应用前景，但评估编辑模型仍然面临挑战。现有基准测试存在两个主要局限性：一是继承自图像编辑的任务覆盖面有限，忽视了视频特有的维度；二是评估指标不足，无法有效衡量指令的准确性，导致由于原始视频的强视觉先验而使错误编辑获得高分。为了解决这些问题，我们提出了一个全面且结构化的IVE基准测试。我们的基准将编辑任务分解为多个视频特有的维度，包括空间、时间、音频和基于参考的编辑，超越了传统的帧级评估。它还区分了显性和隐性指令，并结合基于推理的场景，以更好地反映现实世界的需求。此外，我们提出了一个评估框架，从准确性、保留性、真实性和一致性四个互补维度评估编辑质量，使用人类判断和最先进的视觉-语言模型。为了强调指令的准确性，我们引入了一种准确性感知的惩罚机制，使其他评分依赖于准确性，防止视觉上合理但错误的编辑获得虚高的评估。在对代表性的开源和商业模型进行广泛实验后，我们发现当前的IVE模型仍远未令人满意。OmniEdit-Bench为评估基于指令的视频编辑提供了一个全面且可靠的测试平台，并为未来的研究方向提供了见解。

View on arXiv Download PDF AI Translation

cs.CV / 99 / 2608.05066

Beyond Reprojection Error: Camera Calibration with 3D Targets

超越重投影误差：基于三维目标的相机标定

Ruppel, Dennis, Kutlu, Hasan, Neumann, Kai A., Knuth, Martin, Santos, Pedro, Weinmann, Andreas, Kuijper, Arjan

Abstract

In 3D reconstruction, camera calibration is an essential element for achieving high fidelity and accuracy of the reconstructed geometry. While existing approaches rely upon 2D planar calibration, this work proposes a framework tailored for 3D reconstruction that is based on predicting scene rays, which adds flexibility to the reconstruction pipeline and enables the use of recent advances in camera models. Novel metrics, reconstruction and intersection error, derived from predicted scene rays are employed in combination with a bootstrapping procedure that statistically evaluates different calibration objects and calibration pipelines for both intrinsic and extrinsic camera parameters. The results show that the generalized distortion model more faithfully captures physical camera effects and yields an improvement in calibration accuracy. Reprojection error is shown to be a potentially misleading indicator of 3D accuracy, and the proposed ray-based metrics provide a more holistic assessment. An icosahedron calibration target is designed to enrich calibration information for 3D reconstruction together with a ring-feature-based detector. The icosahedral target yields approximately 40% lower mean intersection and more stable calibration across bootstrap trials on synthetic data, while real-data performance demands very tight fabrication tolerances.

Chinese Translation

在三维重建中，相机标定是实现重建几何体高保真度和准确性的关键要素。虽然现有方法依赖于二维平面标定，但本研究提出了一种针对三维重建的框架，该框架基于预测场景光线，为重建流程增加了灵活性，并使得能够利用最新的相机模型进展。通过从预测的场景光线中衍生的新颖指标——重建误差和交点误差，结合一种自助程序，统计评估不同的标定对象和标定流程，以获取内参和外参的相机参数。结果表明，广义畸变模型更真实地捕捉了物理相机效应，并提高了标定准确性。重投影误差被证明是一个可能误导的三维准确性指标，而所提出的基于光线的指标提供了更全面的评估。设计了一种二十面体标定目标，以丰富三维重建的标定信息，并结合基于环特征的检测器。二十面体目标在合成数据上的平均交点误差降低了约40%，并在自助试验中表现出更稳定的标定性能，而真实数据的表现则要求非常严格的制造公差。

View on arXiv Download PDF AI Translation

cs.CV / 100 / 2608.05069

VQ-VAD: Vector-quantized Motion Representation Learning for Human-centric Video Anomaly Detection

VQ-VAD：用于以人为中心的视频异常检测的向量量化运动表示学习

Rashvand, Narges, Noghre, Ghazal Alinezhad, Yao, Shanle, Maldonado, Gabriel, Tabkhi, Hamed

Abstract

Video Anomaly Detection (VAD) is inherently challenging due to the scarcity of anomalies and the large visual variability in surveillance footage, including changes in lighting, viewpoint, and human appearance. To mitigate visual noise and address privacy concerns, recent work has shifted to pose-based VAD, which focuses on motion dynamics rather than raw video data. However, existing pose-based approaches model human behavior in continuous latent spaces, limiting their ability to learn compact motion patterns necessary for robust behavior analysis. We address this by proposing Vector-Quantized Video Anomaly Detection (VQ-VAD), a novel human-centric anomaly detection framework that learns discrete motion representations. VQ-VAD adapts Vector-Quantized GAN (VQ-GAN), originally developed for image generation, to operate on keypoint sequences and construct a motion codebook of normal behavior. Trained exclusively on normal motion sequences, VQ-VAD detects anomalies by identifying high reconstruction errors when an observed motion sequence cannot be mapped to the learned codebook. We conduct extensive experiments across three complementary evaluation settings, including in-domain, cross-domain, and cross-dataset generalization, on four anomaly detection benchmarks. VQ-VAD achieves strong in-domain accuracy (81.83% on HR-SHT [15]), effective cross-domain transfer from CMU Panoptic [14] (76.69% on HR-SHT [15] without retraining), and competitive cross-dataset robustness. The code base for this work is available at https://github.com/TeCSAR-UNCC/VQ-VAD.

Chinese Translation

视频异常检测（VAD）由于异常事件的稀缺性以及监控视频中光照、视角和人类外观等的巨大视觉变异性而 inherently 具有挑战性。为了减轻视觉噪声并解决隐私问题，近期的研究转向基于姿态的 VAD，专注于运动动态而非原始视频数据。然而，现有的基于姿态的方法在连续潜在空间中建模人类行为，限制了它们学习紧凑运动模式的能力，而这些模式对于稳健的行为分析是必要的。我们通过提出向量量化视频异常检测（VQ-VAD）来解决这一问题，这是一种新颖的以人为中心的异常检测框架，能够学习离散的运动表示。VQ-VAD 采用了最初为图像生成开发的向量量化生成对抗网络（VQ-GAN），使其能够在关键点序列上运行，并构建正常行为的运动代码本。VQ-VAD 仅在正常运动序列上进行训练，通过识别高重建误差来检测异常，当观察到的运动序列无法映射到学习到的代码本时。我们在四个异常检测基准上进行了广泛的实验，涵盖了三个互补的评估设置，包括领域内、跨领域和跨数据集的泛化。VQ-VAD 在领域内的准确率达到 81.83%（在 HR-SHT [15] 上），从 CMU Panoptic [14] 有效地实现了跨领域迁移（在 HR-SHT [15] 上达到 76.69% 而无需重新训练），并展现出竞争力的跨数据集鲁棒性。本研究的代码库可在 https://github.com/TeCSAR-UNCC/VQ-VAD 获取。

View on arXiv Download PDF AI Translation

cs.CV / 101 / 2608.05070

HelloWorld: Enabling Socially Interactive Characters in Video World Models

HelloWorld：在视频世界模型中实现社会互动角色

Ouyang, Liangyang, Liu, Ruicong, Chu, Xuangeng, Zhang, Kaipeng, Sato, Yoichi

Abstract

Despite the remarkable recent progress of video world models, social interaction between users and the characters within these worlds remains unsupported. To fill this gap, we present HelloWorld, a video world model that enables social interaction with in-world characters. With a single button press, users can prompt the on-screen character to respond toward the camera, e.g., turning to the viewer, waving, nodding, or speaking a short greeting. To make these interactions natural, we propose a self-distillation pipeline that finetunes the video generation model on data synthesized by itself. Each synthesized clip contains both social interactions and camera motion, allowing the model to learn camera-pose conditioning without degrading interaction quality. At inference, we further introduce a training-free module that determines when the interaction occurs. Upon a button press, it modulates the cross-attention masks of the DiT so that the interaction-related text prompt attends only to the frames within the press window, temporally localizing the character's response. We further build HelloWorldBench, a 400-sample benchmark with three social interaction metrics alongside three conventional metrics, for evaluation. Experiments demonstrate that HelloWorld surpasses a variety of baselines in interaction quality, while maintaining state-of-the-art picture aesthetics and camera-pose following. Project page: https://github.com/AlayaLab/HelloWorld

Chinese Translation

尽管视频世界模型最近取得了显著进展，但用户与这些世界中的角色之间的社会互动仍然未得到支持。为填补这一空白，我们提出了HelloWorld，这是一种能够与世界内角色进行社会互动的视频世界模型。用户只需按下一个按钮，即可提示屏幕上的角色朝向摄像头做出反应，例如，转向观众、挥手、点头或说一句简短的问候。为了使这些互动自然，我们提出了一种自我蒸馏管道，该管道在由自身合成的数据上微调视频生成模型。每个合成片段都包含社会互动和摄像机运动，使模型能够学习摄像机姿态条件，而不会降低互动质量。在推理过程中，我们进一步引入了一个无训练模块，用于确定何时发生互动。在按下按钮后，它调节DiT的交叉注意力掩码，使得与互动相关的文本提示仅关注按压窗口内的帧，从而在时间上局部化角色的响应。我们还构建了HelloWorldBench，这是一个包含400个样本的基准，配备三个社会互动指标和三个传统指标，用于评估。实验表明，HelloWorld在互动质量上超越了多种基线，同时保持了最先进的图像美学和摄像机姿态跟随能力。项目页面：https://github.com/AlayaLab/HelloWorld

View on arXiv Download PDF AI Translation

cs.CV / 102 / 2608.05074

Bag-of-Visual-Words for Spatial Mapping of Lung Adenocarcinoma Growth Patterns

视觉词袋在肺腺癌生长模式空间映射中的应用

Ardan, Darya, Oreiller, Valentin, Müller, Henning

Abstract

Spatial mapping of lung adenocarcinoma (LUAD) growth patterns across whole slide images (WSIs) requires resolving architectural context at the region level, yet existing methods operate at the individual tile level and produce generic morphological clusters rather than clinically defined pattern maps. We propose a weakly supervised Bag-of-Visual-Words (BoVW) pipeline that learns a visual vocabulary from frozen foundation model embeddings extracted from a small set of annotated regions of interest (ROIs). Pattern prototypes are constructed as mean BoVW histograms of same-label ROIs and used for nearest-prototype classification of sliding-window regions under Jensen--Shannon divergence. The resulting predictions are projected onto the WSI tile grid to produce interpretable spatial pattern maps. We evaluate the method on 87 CPTAC-LUAD patients using three foundation model encoders and multiple vocabulary sizes on two clinically motivated tasks. For tumour/healthy classification, the best configuration achieves a balanced accuracy of $0.974$ with H-Optimus-1, approaching the $0.987$ obtained by a supervised SVM trained on mean-pooled WSI embeddings. For binary histologic grade classification, the BoVW pipeline achieves higher balanced accuracy than the supervised baseline for all encoders, suggesting that ROI-level pattern decomposition preserves grade-relevant heterogeneity that is attenuated by global mean pooling.

Chinese Translation

肺腺癌（LUAD）生长模式在全幻灯片图像（WSIs）中的空间映射需要在区域层面上解决建筑背景，而现有方法则在单个瓦片层面上操作，产生通用的形态学聚类，而非临床定义的模式图。我们提出了一种弱监督的视觉词袋（Bag-of-Visual-Words, BoVW）流程，该流程从一小组标注的感兴趣区域（ROIs）中提取的冻结基础模型嵌入中学习视觉词汇。模式原型作为相同标签ROIs的均值BoVW直方图构建，并用于在Jensen-Shannon散度下对滑动窗口区域进行最近原型分类。生成的预测结果被投影到WSI瓦片网格上，以产生可解释的空间模式图。我们在87名CPTAC-LUAD患者身上评估了该方法，使用三种基础模型编码器和多种词汇大小，针对两个临床驱动的任务进行测试。在肿瘤/健康分类中，最佳配置在H-Optimus-1下实现了$0.974$的平衡准确率，接近于在均值池化WSI嵌入上训练的监督支持向量机（SVM）获得的$0.987$。在二元组织学分级分类中，BoVW流程在所有编码器上都实现了高于监督基线的平衡准确率，表明ROI级别的模式分解保留了与分级相关的异质性，而这一异质性在全局均值池化中被削弱。

View on arXiv Download PDF AI Translation

cs.CV / 103 / 2608.05100

Lesion Detection in CT with Frozen Self-Distilled Features: SALT, a Spatially Adaptive Label-Guided Temperature

基于冷冻自蒸馏特征的CT病变检测：SALT，一种空间自适应标签引导温度

Gokmen, Mahmut S., Damron, Evan W., Klusty, Mitchell A., Leach, Caroline N., Collier, Emily B., Bumgardner, V. K. Cody

Abstract

Self-supervised pretraining objectives are spatially uniform: the teacher temperature and the per-patch loss weight are identical everywhere in the image, so a lesion a few patches wide contributes no more to the training signal than the surrounding parenchyma. Prior work biases the views toward annotated regions, which changes what the model sees but adds no pressure on the objective. We instead condition the targets of self-distillation, a method we call SALT (Spatially Adaptive Label-guided Temperature). Weak, box-derived labels, available only during pretraining, define a compact region on the encoder's patch grid, inside which the teacher's softmax temperature is sharpened and the masked-patch loss is up-weighted. The objectives, the masking policy and the centering statistics are otherwise unchanged, and at every downstream use the encoder is a plain feature extractor with no labels and no conditioning. We evaluate by freezing the encoder and training only a lightweight multi-depth CenterNet-style head, detecting lesions in 3D on four CT cohorts, and we isolate the mechanism against a backbone identical in architecture, pretraining data, schedule and label-guided cropping but with no target conditioning. We report patch-level separability, 3D detection stratified by cohort and by lesion size, box quality, and a detector-free probe in which a single frozen patch embedding re-identifies a lesion in a follow-up scan without registration, masks or fine-tuning. Because the conditioning is expressed through a spatial indicator rather than through label semantics, the formulation admits any weak spatial annotation; we instantiate and validate it for lesions.

Chinese Translation

自监督预训练目标在空间上是均匀的：教师温度和每个补丁的损失权重在图像中的每个位置都是相同的，因此一个宽度仅为几个补丁的病变对训练信号的贡献与周围实质组织相同。先前的工作偏向于注释区域，这改变了模型的视野，但对目标没有施加压力。我们则对自蒸馏的目标进行条件调整，提出了一种我们称之为SALT（空间自适应标签引导温度）的方法。仅在预训练期间可用的弱框导出标签定义了编码器补丁网格上的一个紧凑区域，在该区域内，教师的softmax温度被锐化，并且掩蔽补丁损失被加权。其他目标、掩蔽策略和中心统计量保持不变，在每次下游使用中，编码器都是一个普通的特征提取器，没有标签和条件。我们通过冻结编码器，仅训练一个轻量级的多深度CenterNet风格头，在四个CT队列中进行3D病变检测，并将机制与一个在架构、预训练数据、时间表和标签引导裁剪上完全相同但没有目标条件的骨干网络进行隔离。我们报告了补丁级的可分离性、按队列和病变大小分层的3D检测、框质量，以及一个无检测器探针，其中单个冻结的补丁嵌入在后续扫描中重新识别病变，而无需配准、掩蔽或微调。由于条件是通过空间指示而不是通过标签语义表达的，因此该公式允许任何弱空间注释；我们为病变实例化并验证了该方法。

View on arXiv Download PDF AI Translation

cs.CV / 104 / 2608.05101

HexMIL: Hierarchical Attention MIL for Ante-Hoc Explainable Detection of AI-Manipulated CT Volumes

HexMIL：用于AI操控CT体积的前置可解释检测的层次注意力多实例学习

Pontorno, Orazio, Guarnera, Luca, Akhtar, Zahid, Battiato, Sebastiano

Abstract

The emergence of medical deepfakes, i.e., medical images manipulated by deep generative models, poses a significant threat to clinical workflows. However, existing detectors suffer from two critical limitations: poor generalization to unseen generative architectures for manipulation detection and lack of interpretability. In this context, we present HexMIL (Hierarchical EXplainable Multiple Instance Learning), a mask-free medical deepfake detector that simultaneously addresses both limitations using only binary volume-level supervision. HexMIL decomposes each CT volume into a two-level hierarchy of patches and slices, aggregated via independent Gated Attention modules whose weights are directly combined into a full-resolution 3D attention volume that localizes the manipulated sub-region without any pixel-level annotation. Unlike post-hoc methods such as Grad-CAM, HexMIL's attention weights constitute the exact forward computation driving the classification decision, providing ante-hoc and structurally faithful spatial attribution. We evaluate HexMIL on M3DSynth and CT-GAN datasets under a rigorous cross-generator generalization protocol, training on a single generative architecture and testing on unseen ones. HexMIL outperforms all baselines by $+9.1$ AUC and $+9.4$ F1 in out-of-domain classification, and achieves the best average IoU and Pointing Game score in localization. Project page: opontorno.github.io/hexmil.

Chinese Translation

医疗深伪造的出现，即被深度生成模型操控的医疗图像，对临床工作流程构成了重大威胁。然而，现有的检测器存在两个关键限制：对未见生成架构的操控检测的泛化能力差，以及缺乏可解释性。在此背景下，我们提出了HexMIL（层次可解释多实例学习），这是一种无掩膜的医疗深伪造检测器，仅通过二进制体积级别的监督，能够同时解决这两个限制。HexMIL将每个CT体积分解为两个层次的补丁和切片，通过独立的门控注意力模块进行聚合，这些模块的权重直接组合成一个全分辨率的3D注意力体积，从而在没有任何像素级注释的情况下定位被操控的子区域。与Grad-CAM等后置方法不同，HexMIL的注意力权重构成了驱动分类决策的确切前向计算，提供了前置且结构上忠实的空间归因。我们在M3DSynth和CT-GAN数据集上对HexMIL进行了评估，采用严格的跨生成器泛化协议，在单一生成架构上训练，并在未见架构上测试。HexMIL在域外分类中超越了所有基线，AUC提升了$+9.1$，F1提升了$+9.4$，并在定位中达到了最佳的平均IoU和指向游戏得分。项目页面：opontorno.github.io/hexmil。

View on arXiv Download PDF AI Translation

cs.CV / 105 / 2608.05115

Robust and Efficient Motion Reasoning for Privacy-Aware Classroom Incident Recognition

隐私意识课堂事件识别的稳健高效运动推理

Parmar, Paritosh, Lan, Landy, Yang, Hong, Yi, Chen, Tay, Chiat Pin

Abstract

Can computer vision help make classrooms safer? In this pilot study, we investigate privacy-aware and computationally efficient classroom incident recognition from CCTV-style observations. This setting remains underexplored, with limited benchmarks and few methods designed for the privacy, efficiency, and generalization demands of real-world deployment. We introduce a novel hybrid benchmark combining generative CCTV-style videos with real-world classroom pose data, and propose a lightweight, but robust motion-reasoning framework motivated by the observation that many incidents differ more in motion direction, speed, acceleration, and intensity than in pose alone. To that end, our method first constructs hierarchical kinematic representations of human actions. Our method then distills hierarchical, multi-order kinematic reasoning from a large teacher into a much smaller single-order student, enabling efficient per-person inference while preserving expressive motion understanding. Experiments show that our model outperforms substantially larger baselines at less than one-tenth of their computational cost, while also demonstrating stronger out-of-domain motion reasoning and zero-shot synthetic-to-real generalization. We will publicly release the benchmark, codebase, and supporting tools to facilitate further research in privacy-aware classroom safety.

Chinese Translation

计算机视觉能否帮助提高课堂安全性？在这项初步研究中，我们探讨了基于 CCTV 风格观察的隐私意识和计算高效的课堂事件识别。这一领域尚未得到充分研究，现有基准有限，且很少有方法专门针对现实部署中的隐私、效率和泛化需求。我们引入了一种新颖的混合基准，将生成的 CCTV 风格视频与真实课堂姿态数据相结合，并提出了一种轻量但稳健的运动推理框架，基于这样一个观察：许多事件在运动方向、速度、加速度和强度上差异更大，而不仅仅是姿态。为此，我们的方法首先构建了人类动作的分层运动学表示。然后，我们的方法从一个大型教师模型中提取分层的多阶运动学推理，转化为一个更小的单阶学生模型，从而实现高效的逐人推理，同时保持丰富的运动理解。实验表明，我们的模型在计算成本不到大型基准的十分之一的情况下，显著超越了更大的基准，同时在域外运动推理和零-shot 合成到真实的泛化方面表现更强。我们将公开发布基准、代码库和支持工具，以促进隐私意识课堂安全的进一步研究。

View on arXiv Download PDF AI Translation

cs.CV / 106 / 2608.05122

IRIS: A Visual Cortex-Inspired Framework for Analyzing Orientation Selectivity in Vision Transformers

IRIS：一种受视觉皮层启发的框架，用于分析视觉变换器中的方向选择性

Mohan, Vaishnavi B, Naganoor, Vijayakrishna, Annadani, Yashas, Hegde, Shashank

Abstract

Vision transformers (ViTs) have become the de facto standard for image encoding across many perception tasks. Despite their empirical success, it remains mechanistically unclear how they encode low-level features, given their lack of inductive biases: ViTs process information globally rather than relying on local structure. Biological visual systems, in contrast, build low-level features, such as orientation selectivity in the primary visual cortex, by combining information from small, localized regions of the visual field. These features are general-purpose representations, shared and required across multiple specialized neural pathways, unlike higher-level, task-specific semantic features. This raises the question if such biologically-grounded features arise in ViTs. In this work, we systematically study how orientation selectivity emerges in ViTs by introducing a suite of neuroscience-inspired metrics: representational similarity score (RSS), orientation recruitment score (ORS), and orientation tuning bandwidth to quantify how orientation is encoded in representational geometry and as a function of model depth. Through extensive analysis, we find that: (1) the training paradigm is the strongest determinant of orientation selectivity, with models sharing an objective, peaking at comparable relative depths regardless of scale (2) many units are orientation-selective early in training, with early-to-middle layers recruiting more such units over time, while deeper layers lose selectivity and broaden their tuning toward semantic encoding and (3) our metrics offer a mechanistic heuristic for how many layers to unfreeze for best downstream generalization. Our framework presents a way to track biologically-grounded features during ViT training, probes how desired properties are encoded in transformer representations, and builds a systematic understanding of how ViTs generalize across tasks.

Chinese Translation

视觉变换器（ViTs）已成为许多感知任务中图像编码的事实标准。尽管它们在实践中取得了成功，但由于缺乏归纳偏置，如何在机制上编码低级特征仍然不清楚：ViTs处理信息是全局性的，而不是依赖于局部结构。相比之下，生物视觉系统通过结合来自视觉场小区域的信息来构建低级特征，例如初级视觉皮层中的方向选择性。这些特征是通用表示，跨多个专业神经通路共享和需要，而不是更高级的、任务特定的语义特征。这引发了一个问题，即这种生物基础特征是否在ViTs中出现。在本研究中，我们通过引入一套受神经科学启发的指标系统，系统地研究方向选择性在ViTs中的出现：表示相似性得分（RSS）、方向招募得分（ORS）和方向调谐带宽，以量化方向如何在表示几何中编码以及作为模型深度的函数。通过广泛的分析，我们发现：（1）训练范式是方向选择性的最强决定因素，具有共享目标的模型在相对深度上达到峰值，无论规模如何；（2）许多单元在训练早期就表现出方向选择性，早期到中层随着时间的推移招募更多这样的单元，而更深层则失去选择性并向语义编码扩展其调谐；（3）我们的指标为最佳下游泛化解冻多少层提供了机制启发。我们的框架提供了一种在ViT训练期间跟踪生物基础特征的方法，探测所需属性如何在变换器表示中编码，并建立对ViTs如何跨任务泛化的系统理解。

View on arXiv Download PDF AI Translation

cs.CV / 107 / 2608.05131

OPD-V: Visual On-Policy Self-Distillation with Modality Balance

OPD-V：具有模态平衡的视觉在线自蒸馏

Aniri, Bi, Jinhe, Liao, Peng, Jin, Zengjie, Tresp, Volker, Shen, Fei, Ma, Yunpu, Chua, Tat-Seng

Abstract

On-Policy Self-Distillation (OPSD) has become a standard post-training approach for improving visual reasoning in multimodal large language models (MLLMs). Existing methods draw privileged information from diverse input sources to guide self-distillation. Yet these designs overlook Modality Imbalance, a challenge inherent to MLLM reasoning. When textual information dominates generation, the model cannot fully integrate its multimodal input. Consequently, carefully designed privileged information remains underused, limiting the effectiveness of OPSD. To examine this limitation, we construct a Positive Teacher with the Zoom-In Image and a Negative Teacher with the Mask Image, which exhibit different degrees of Modality Imbalance. Changes in their reasoning correctness and token logits reveal that Modality Balance can itself serve as privileged information. Motivated by this finding, we introduce OPD-V, a visual OPSD paradigm that instantiates such information through the Positive Teacher and Negative Teacher. Positive Modality-Balance Logits Margins define a Modality-Balance Trust Region that selects the on-policy tokens used for self-distillation. Experiments across 6 benchmarks, 4 MLLM backbones, and 5 post-training methods show that OPD-V consistently improves reasoning performance while reducing training cost.

Chinese Translation

在线自蒸馏（On-Policy Self-Distillation，OPSD）已成为提高多模态大型语言模型（Multimodal Large Language Models，MLLMs）视觉推理的标准后训练方法。现有方法从多样的输入源中提取特权信息以指导自蒸馏。然而，这些设计忽视了模态不平衡（Modality Imbalance），这是MLLM推理固有的挑战。当文本信息主导生成时，模型无法充分整合其多模态输入。因此，精心设计的特权信息未能得到充分利用，限制了OPSD的有效性。为了检验这一局限性，我们构建了一个正教师（Positive Teacher），使用放大图像（Zoom-In Image），以及一个负教师（Negative Teacher），使用遮罩图像（Mask Image），它们表现出不同程度的模态不平衡。它们推理正确性和标记对数的变化表明，模态平衡本身可以作为特权信息。受此发现的启发，我们提出了OPD-V，一种视觉OPSD范式，通过正教师和负教师实例化这种信息。正模态平衡对数边际（Positive Modality-Balance Logits Margins）定义了一个模态平衡信任区域（Modality-Balance Trust Region），用于选择用于自蒸馏的在线标记。在6个基准、4个MLLM骨干网络和5种后训练方法上的实验表明，OPD-V在提高推理性能的同时降低了训练成本。

View on arXiv Download PDF AI Translation

cs.CV / 108 / 2608.05132

Predicting Brain Morphometry with MT-GNN: Mesh Evolution in Continuous Time with Graph-Based Metric Tensor Embeddings

使用MT-GNN预测脑形态测量：基于图的度量张量嵌入的连续时间网格演化

Ding, Hao, Semchin, Daniel, Thompson, Paul M., Gutman, Boris

Abstract

Predicting how a subcortical structure's shape will evolve from a few prior scans could support prognosis and clinical-trial enrichment. Existing longitudinal mesh predictors either extrapolate shape trajectories via high-dimensional embeddings or regress vertex deformations directly. We instead predict the surface's intrinsic geometry in continuous time: a single per-structure graph network predicts the future per-vertex first fundamental form (metric tensor) for an arbitrary causal multiple-visit history and an arbitrary prediction horizon, conditioned on a Fourier encoding of the lead time. The predicted metric is decoded into a surface by a differentiable As-Rigid-As-Possible solver, and the model is trained end-to-end on the rigid-aligned vertex error. Training through the reconstruction keeps the decoded prediction a valid surface and consistently improves it. On 14 subcortical structures from the ADNI dataset, the proposed mesh evolution model (MT-GNN) predicts best among the evaluated methods at every horizon ($-2.29\%$ mean vertex error vs. the temporal mean, $p{=}6.1{\times}10^{-5}$, beating it on 14/14 structures), ahead of geodesic shape regression (DCM, $-0.19\%$) and a mesh transformer (TransforMesh, $-0.45\%$; $p{=}1.2{\times}10^{-4}$), with the lead widening as the horizon grows.

Chinese Translation

预测亚皮层结构的形状如何从几次先前扫描中演变，可以支持预后和临床试验的丰富性。现有的纵向网格预测器要么通过高维嵌入外推形状轨迹，要么直接回归顶点变形。我们则预测表面的内在几何形状在连续时间中演变：一个单一的每结构图网络预测未来每顶点的第一基本形式（度量张量），该预测基于任意因果多次访问历史和任意预测时间范围，并以傅里叶编码的提前时间为条件。预测的度量通过可微分的尽可能刚性求解器解码为表面，并且模型在刚性对齐的顶点误差上进行端到端训练。通过重建训练保持解码预测为有效表面，并持续改善其质量。在ADNI数据集中对14个亚皮层结构进行评估时，所提出的网格演化模型（MT-GNN）在每个时间范围内的预测效果最佳（相较于时间均值，平均顶点误差为-2.29 ext{%}，$p{=}6.1{ imes}10^{-5}$，在14/14个结构中优于时间均值），领先于测地形状回归（DCM，-0.19 ext{%}）和网格变换器（TransforMesh，-0.45 ext{%}；$p{=}1.2{ imes}10^{-4}$），随着时间范围的增长，领先优势进一步扩大。

View on arXiv Download PDF AI Translation

cs.CV / 109 / 2608.05137

SmartMage: Dynamic Modality Orchestration for 3D Scene Understanding

SmartMage：用于3D场景理解的动态模态编排

Zhang, Yue, Jian, Yingzhao, Xu, Yunqiu, Sun, Xiaoxiao, Fan, Hehe

Abstract

Understanding 3D scenes is fundamental to embodied intelligence, requiring joint reasoning over heterogeneous information from multiple modalities, including visual and geometric cues. However, the relevance of these modalities often varies across queries. Existing Multimodal Large Language Models (MLLMs) typically rely on fixed modality combinations, overlooking query-dependent modality needs. Such a rigid design can introduce semantic noise from irrelevant modalities while underutilizing more informative ones, leading to wasted computation and diluted reasoning. To address these challenges, this paper proposes SmartMage, a unified MLLM that dynamically orchestrates heterogeneous modalities for semantic-aware 3D scene understanding. Specifically, SmartMage incorporates: (1) a Semantic-guided Modality Adaptive RouTng (SMART) module that selects task-relevant modalities using semantic priors, text-modality alignment, and modality quality; and (2) a Modality-Aware Gating Expert (MAGE) module that leverages modality priors to guide expert activation, fostering adaptive specialization in multimodal reasoning. Empirically, SmartMage achieves state-of-the-art performance across five 3D scene understanding benchmarks, and attains competitive results on RGB-only video understanding benchmarks. In our diagnostic benchmark ScanFacet, tasks are divided into fine-grained semantic categories, enabling analysis of modality combinations preferred by each semantic type. The observed modality-semantic patterns provide further evidence of SmartMage's effectiveness. Project page: https://yuecheong.github.io/SmartMage/.

Chinese Translation

理解3D场景是具身智能的基础，要求对来自多种模态的异构信息进行联合推理，包括视觉和几何线索。然而，这些模态的相关性在不同查询中往往有所不同。现有的多模态大语言模型（Multimodal Large Language Models, MLLMs）通常依赖于固定的模态组合，忽视了查询依赖的模态需求。这种僵化的设计可能会引入来自不相关模态的语义噪声，同时未能充分利用更具信息量的模态，导致计算浪费和推理稀释。为了解决这些挑战，本文提出了SmartMage，一个统一的MLLM，动态编排异构模态以实现语义感知的3D场景理解。具体而言，SmartMage包含：(1) 一个语义引导的模态自适应路由（Semantic-guided Modality Adaptive Routing, SMART）模块，通过语义先验、文本-模态对齐和模态质量选择与任务相关的模态；(2) 一个模态感知的门控专家（Modality-Aware Gating Expert, MAGE）模块，利用模态先验来指导专家激活，促进多模态推理中的自适应专业化。从实证结果来看，SmartMage在五个3D场景理解基准测试中实现了最先进的性能，并在仅使用RGB的视频理解基准测试中取得了竞争性结果。在我们的诊断基准ScanFacet中，任务被划分为细粒度的语义类别，从而能够分析每种语义类型所偏好的模态组合。观察到的模态-语义模式进一步证明了SmartMage的有效性。项目页面：https://yuecheong.github.io/SmartMage/

View on arXiv Download PDF AI Translation

cs.CV / 110 / 2608.05145

Objects as Audio-Visual Modal Sound Fields

物体作为音频-视觉模态声场

Shao, Zisen, Wei, Zihao, Jin, Derong, Gao, Ruohan

Abstract

While modern 3D reconstruction excels at modeling object geometry and appearance, it largely ignores the rich acoustic cues revealed through physical interaction. Object impact sounds convey material, stiffness, and structural properties that complement vision, yet existing impact sound modeling approaches either rely on expensive physics-based simulation or require large datasets to generalize in a purely data-driven manner. We introduce Audio-Visual Modal Sound Field (AV-MSF), a novel object-level acoustic representation reconstructed from multi-view images and only a few impact sound recordings. AV-MSF builds on 3D Gaussian Splatting integrated with dense 3D visual feature to provide a strong geometry-aware prior, and represents the impact sound field using compact, physically meaningful modal parameters, enabling robust few-shot reconstruction. Experiments on two real-world datasets show that AV-MSF achieves state-of-the-art impact sound rendering, outperforming both physics-based and data-driven baselines. Furthermore, we demonstrate downstream applications enabled by our representation, including contact localization and object sound editing.

Chinese Translation

尽管现代3D重建在建模物体几何形状和外观方面表现出色，但在很大程度上忽视了通过物理交互所揭示的丰富声学线索。物体撞击声传达了材料、刚度和结构特性，这些特性补充了视觉信息。然而，现有的撞击声建模方法要么依赖于昂贵的基于物理的模拟，要么需要大量数据集以纯数据驱动的方式进行泛化。我们提出了音频-视觉模态声场（Audio-Visual Modal Sound Field, AV-MSF），这是一种从多视角图像和少量撞击声录音中重建的新型物体级声学表示。AV-MSF基于3D高斯点云（3D Gaussian Splatting），结合密集的3D视觉特征，提供了强大的几何感知先验，并使用紧凑且具有物理意义的模态参数表示撞击声场，从而实现稳健的少样本重建。在两个真实世界数据集上的实验表明，AV-MSF在撞击声渲染方面达到了最先进的水平，超越了基于物理和数据驱动的基线。此外，我们展示了由我们的表示所支持的下游应用，包括接触定位和物体声音编辑。

View on arXiv Download PDF AI Translation

cs.CV / 111 / 2608.05149

CoCo-IR: Contextual Composed Image Retrieval

CoCo-IR：上下文组合图像检索

Cao, Shengcao, Dabral, Tanmaya Shekhar, Ding, Zhongli, Shanbhogue, Madhuri, Chen, Kaifeng, Li, Zhe, Seyedhosseini, Mojtaba, Wang, Yu-Xiong, Gui, Liang-Yan

Abstract

Current instruction-based image retrieval systems are powerful but limited to single-turn interactions, failing to capture the iterative nature of complex, real-world visual searches. To overcome this limitation, we introduce Contextual Composed Image Retrieval (CoCo-IR), a novel task that enables users to progressively refine search results through interactions. We address this new task by proposing a new model based on a Large Multimodal Model (LMM) that functions as a context-aware reasoner for CoCo-IR. Our model interprets the entire interaction history to generate Transformable Image Embeddings (TIE) that evolve across turns. To fuel the model training without expensive human annotations, we develop a fully autonomous, scalable data engine that leverages LMMs to generate high-quality contextual retrieval data, and uses model-guided verification to mine challenging hard negatives. Extensive experiments demonstrate that our approach establishes new state-of-the-art performance: We achieve 39.4 mAP@5 on the challenging single-turn benchmark CIRCO; furthermore, on our new CoCo-IR benchmark, our model maintains robust performance with 44.1 R@1 on 4-turn dialogues, dramatically outperforming existing methods (28.2 4-turn R@1) that fail to handle multi-turn context. Project page: https://CoCo-IR.github.io.

Chinese Translation

当前基于指令的图像检索系统功能强大，但仅限于单轮交互，无法捕捉复杂现实视觉搜索的迭代特性。为了解决这一局限性，我们提出了上下文组合图像检索（CoCo-IR），这是一个新颖的任务，使用户能够通过交互逐步细化搜索结果。我们通过提出一个基于大型多模态模型（Large Multimodal Model, LMM）的新模型来解决这一新任务，该模型作为CoCo-IR的上下文感知推理器。我们的模型解释整个交互历史，以生成在多个回合中演变的可变图像嵌入（Transformable Image Embeddings, TIE）。为了在没有昂贵人工标注的情况下推动模型训练，我们开发了一个完全自主、可扩展的数据引擎，利用LMM生成高质量的上下文检索数据，并使用模型引导的验证来挖掘具有挑战性的困难负样本。大量实验表明，我们的方法建立了新的最先进性能：在具有挑战性的单轮基准CIRCO上，我们达到了39.4 mAP@5；此外，在我们的新CoCo-IR基准上，我们的模型在4轮对话中保持了44.1 R@1的强大性能，显著超越了现有方法（28.2 4-turn R@1），这些方法无法处理多轮上下文。项目页面：https://CoCo-IR.github.io。

View on arXiv Download PDF AI Translation

人工智能 (Artificial Intelligence)

cs.AI / 1 / 2608.04012

A Long-Run Persistence Theory for AI Systems under the Redundancy-Adjusted Artificial Age Score (AAS)

基于冗余调整的人工年龄评分（AAS）的人工智能系统长期持续性理论

Kayadibi, Seyma Yaman

Abstract

Artificial intelligence systems are increasingly expected to operate over repeated cycles of interaction, adaptation, and update rather than through isolated one-shot outputs. This raises a fundamental theoretical question: can an AI system persist indefinitely without incurring unbounded structural aging? This paper develops a long-run persistence framework for AI systems based on the redundancy-adjusted Artificial Age Score (AAS). The model extends AAS from a static evaluative measure into a cycle-level functional that generates an age sequence across repeated operation. At each cycle, structural age is defined through a weighted, redundancy-aware logarithmic penalty over component consistency levels. Within this framework, cycle-level age is shown to be well defined and uniformly bounded, thereby excluding explosive pointwise aging. On this basis, the paper defines a hierarchy of asymptotic regimes, including burdened persistence, zero-burden persistence, oscillatory persistence, and cumulative terminal burden. It also establishes comparative ordering, sensitivity bounds, convergence under componentwise stabilization, persistence under finite total variation, geometric stabilization under damped inter-cycle perturbations, and a zero-burden characterization under nondegenerate redundancy conditions. The main result is that indefinite cyclic continuation does not require unbounded structural aging: an AI system may pass through infinitely many cycles while its structural age remains bounded, while under stronger regularity conditions its marginal aging vanishes and, in the strongest regime, its cycle-level burden converges to zero. The framework thus provides a formal basis for analyzing long-run artificial persistence as a problem of bounded structural burden rather than inevitable cumulative deterioration.

Chinese Translation

人工智能系统越来越被期望在重复的交互、适应和更新周期中运行，而不是通过孤立的一次性输出。这引发了一个基本的理论问题：人工智能系统能否在不遭受无限结构老化的情况下无限期持续？本文基于冗余调整的人工年龄评分（AAS）发展了一个人工智能系统的长期持续性框架。该模型将AAS从静态评估指标扩展为一个周期级功能，生成跨重复操作的年龄序列。在每个周期中，结构年龄通过对组件一致性水平的加权、考虑冗余的对数惩罚来定义。在此框架内，周期级年龄被证明是良好定义且均匀有界的，从而排除了爆炸性逐点老化。在此基础上，本文定义了一系列渐近状态，包括负担持续性、零负担持续性、振荡持续性和累积终端负担。还建立了比较排序、敏感性界限、在组件稳定下的收敛、在有限总变差下的持续性、在减振周期间扰动下的几何稳定性，以及在非退化冗余条件下的零负担特征。主要结果是，无限周期延续并不需要无限结构老化：一个人工智能系统可以经历无限多个周期，而其结构年龄保持有界；在更强的规律性条件下，其边际老化消失，并且在最强的状态下，其周期级负担收敛于零。因此，该框架为分析长期人工持续性提供了一个正式基础，将其视为一个有界结构负担的问题，而非不可避免的累积恶化。

View on arXiv Download PDF AI Translation

cs.AI / 2 / 2608.04066

The LLM Proposes, the Executive Disposes: A Self-Verifying Agent Instrument that Dissociates Commitment Drift from Binding Drift in Long-Horizon Agents

大型语言模型提议，执行者处置：一种自验证代理工具，将长期代理中的承诺漂移与绑定漂移分离

Arjmandi, Mohsen

Abstract

How do you verify a long-horizon agent when its own state and self-reports are exactly what you cannot trust? We present an agent instrument built so that verification is structural rather than post-hoc. A deterministic Executive owns all belief; a language model may only file typed proposals, and a claim is admitted only when a prediction pre-registered before acting is matched against observation by code. Two properties make the instrument a verifier of its own science, not just of the agent: every run invalidates itself when per-organ write-error, render-size, or salted-canary-echo floors are breached (four of the first eight architecture runs were invalidated, each localizing a real defect); and a render-invisible shadow reference compiles the plan the full system would have committed in every ablation cell, so drift metrics are defined even where the mechanism under test has been removed. Using this instrument we report a clean, single-variable result on a failure every long-horizon agent suffers: ablating the commitment mechanism flips goal-abandonment from 0.00 to 1.00 while binding error stays flat at 0.00 (three seeds per cell, up to 394 reference beats per run, every run gated valid). The binding channel, by contrast, does not reappear as per-beat drift when its repair is ablated -- because binding is code-owned, the failure class is structurally absorbed, its only residue appearing one layer upstream as a collapse in hypothesis formation. We report these under full disclosure that task efficacy is null (zero level completions across 52 gated runs on ARC-AGI-3), pre-registered as a structural defeater. The contribution is a verification methodology for agent development and the drift decomposition it makes measurable.

Chinese Translation

当长期代理的自身状态和自我报告正是你无法信任的内容时，如何验证它？我们提出了一种代理工具，使得验证是结构性的，而非事后验证。一个确定性的执行者拥有所有信念；语言模型只能提交类型化的提案，只有在行动前预注册的预测与代码观察匹配时，声明才被接受。该工具的两个特性使其成为自身科学的验证者，而不仅仅是代理的验证者：每次运行在每个组织写入错误、渲染大小或盐罐回声阈值被突破时都会自我失效（前八次架构运行中有四次被失效，每次都定位到一个实际缺陷）；而一个渲染不可见的影子参考编译了完整系统在每个消融单元中本应承诺的计划，因此即使在被测试的机制被移除的情况下，漂移指标也得以定义。使用该工具，我们报告了一个干净的单变量结果，揭示了每个长期代理都会遭遇的失败：消融承诺机制将目标放弃率从0.00翻转至1.00，而绑定错误保持在0.00（每个单元三个种子，每次运行最多394次参考节拍，每次运行均有效）。相比之下，当其修复被消融时，绑定通道并未作为每节拍漂移重新出现——因为绑定是代码拥有的，失败类别被结构性吸收，其唯一残留在假设形成的上游一层表现为崩溃。我们在完全披露任务有效性为零（在ARC-AGI-3上52次有效运行中零级完成）下报告这些，预注册为结构性反驳。该贡献是用于代理开发的验证方法论及其可测量的漂移分解。

View on arXiv Download PDF AI Translation

cs.AI / 3 / 2608.04071

Monte Carlo Tree Search for Table-to-Multimodal Report Generation

用于表格到多模态报告生成的蒙特卡洛树搜索

Lin, Teng, Zhang, Zhiyang, Luo, Yuyu, Tang, Nan

Abstract

Automatically generating professional multimodal reports comprising both textual analysis and visual charts from structured tabular data is a critical challenge in data intelligence. Existing methods suffer from fixed linear pipelines and isolated subtask processing, which hinder joint optimization of factual accuracy, visual quality, and narrative coherence. To address these issues, this paper proposes MCTS-Report, a Monte Carlo Tree Search (MCTS)-driven framework that formulates multimodal table-to-report generation as a progressive construction process over a structured search space. The core idea is to decompose report generation into atomic actions, including chapter planning, visualization task identification, chart generation, insight organization, and narrative refinement, each executed by an LLM based on dynamic reasoning conditioned on the current report state. We use an LLM to generate step-by-step reasoning and actions during MCTS, storing the reasoning trajectory in each node for context-aware, coherent report construction. To guide the search, we design a multi-dimensional reward function that jointly evaluates numerical fact consistency (via SQL), chart quality, chart-text alignment, and structural completeness, while incorporating a diversity penalty to suppress repeated charts and a precondition check to prune invalid actions. We also construct MMRBench, a comprehensive benchmark comprising real-world tables from six domains, paired with expert-refined reference report structures and verifiable key insights. Experiments on MMRBench demonstrate that MCTS-Report significantly outperforms strong baselines across structural completeness, numerical accuracy, chart-text alignment, and insight novelty, achieving a 77.9 overall score.

Chinese Translation

从结构化表格数据自动生成包含文本分析和视觉图表的专业多模态报告是数据智能中的一项关键挑战。现有方法受限于固定的线性流程和孤立的子任务处理，阻碍了事实准确性、视觉质量和叙事连贯性的联合优化。为了解决这些问题，本文提出了MCTS-Report，一个基于蒙特卡洛树搜索（MCTS）的框架，将多模态表格到报告生成表述为在结构化搜索空间上的渐进构建过程。其核心思想是将报告生成分解为原子动作，包括章节规划、可视化任务识别、图表生成、洞察组织和叙事精炼，每个动作由基于当前报告状态进行动态推理的语言模型（LLM）执行。我们使用LLM在MCTS过程中生成逐步推理和动作，并在每个节点存储推理轨迹，以便进行上下文感知的连贯报告构建。为了引导搜索，我们设计了一个多维奖励函数，联合评估数值事实一致性（通过SQL）、图表质量、图表与文本的对齐以及结构完整性，同时引入多样性惩罚以抑制重复图表，并进行前置条件检查以修剪无效动作。我们还构建了MMRBench，一个包含来自六个领域的真实世界表格的综合基准，配有专家精炼的参考报告结构和可验证的关键洞察。MMRBench上的实验表明，MCTS-Report在结构完整性、数值准确性、图表与文本的对齐以及洞察新颖性方面显著优于强基线，整体得分达到77.9。

View on arXiv Download PDF AI Translation

cs.AI / 4 / 2608.04077

FinProBench: Evaluating Financial AI Agents with Role-Grounded Rubrics Derived from Professional Deliverables

FinProBench：使用基于角色的评估标准评估金融人工智能代理，标准源自专业成果

Wang, Ben, Zhou, Kang, Guo, Lifan, Chen, Feng, Zhang, Chi

Abstract

Evaluating financial AI agents requires criteria aligned with real professional work. Existing rubric methods typically derive criteria from task prompts or model outputs, overlooking tacit standards visible only in practitioner deliverables. We introduce FinProBench, a benchmark for professional financial tasks, and Role-Grounded Rubric Construction (RGRC), a reusable pipeline that derives rubrics from deliverables produced by practitioners in the same role. RGRC comprises four stages: Deliverable Collection, Competency Extraction, Rubric Synthesis, and Validation. Its rubrics capture tacit standards, distinguish quality levels, and transfer across tasks within a role. Before analysis, we classified 57 occupations by deliverable genre into 30 prior-rich conventional roles and 27 prior-sparse role-specialized roles. Across all roles, Prompt-only nearly matches RGRC for conventional roles (89.2% vs. 90.7%), but RGRC substantially outperforms it for role-specialized roles (99.1% vs. 78.0%). This split indicates that prompt engineering can approximate rubrics when conventions are well represented in model priors, while professional grounding is essential for standards beyond those priors. FinProBench is built from 1,723 curated deliverables spanning 57 occupations, 8 financial sub-industries, and 161 deliverable types, and releases an initial evaluation set of 20 complete tasks covering 20 roles in 7 sub-industries. With heterogeneous LLM judges and role-level rubrics, human deliverables rank first on average (73.7 vs. 70.3, 70.2, and 69.6 out of 100), while all four systems show overlapping 95% confidence intervals and complementary strengths. Reusing rubrics at the role level reduces estimated per-task construction effort by 6.7 times relative to authoring each rubric from scratch.

Chinese Translation

评估金融人工智能代理需要与真实专业工作相一致的标准。现有的评估标准方法通常从任务提示或模型输出中提取标准，忽视了仅在从业者成果中可见的隐性标准。我们介绍了FinProBench，这是一个针对专业金融任务的基准，以及角色基础评估标准构建（Role-Grounded Rubric Construction, RGRC），这是一个可重用的流程，从同一角色的从业者所产生的成果中提取评估标准。RGRC包括四个阶段：成果收集、能力提取、评估标准合成和验证。其评估标准捕捉隐性标准，区分质量水平，并在同一角色内的任务之间进行转移。在分析之前，我们将57个职业按成果类型分类为30个传统角色和27个专业角色。对于所有角色，Prompt-only在传统角色上几乎与RGRC相匹配（89.2%对90.7%），但在专业角色上，RGRC的表现大幅优于Prompt-only（99.1%对78.0%）。这种差异表明，当常规在模型先验中得到良好表示时，提示工程可以近似评估标准，而专业基础对于超出这些先验的标准至关重要。FinProBench由1,723个经过筛选的成果构成，涵盖57个职业、8个金融子行业和161种成果类型，并发布了一个初步评估集，包含覆盖7个子行业中20个角色的20个完整任务。通过异质的LLM评审和角色级评估标准，人类成果的平均排名最高（73.7对70.3、70.2和69.6，满分100），而所有四个系统显示出重叠的95%置信区间和互补的优势。在角色级别重用评估标准将每个任务的估计构建工作量减少了6.7倍，相较于从头编写每个评估标准。

View on arXiv Download PDF AI Translation

cs.AI / 5 / 2608.04095

FinPerMA: A Theory-Informed, Event-Grounded Personalized-Memory Benchmark for LLM Agents

FinPerMA：基于理论的、事件驱动的个性化记忆基准，用于大型语言模型代理

Wang, Ben, Zhou, Kang, Guo, Lifan, Chen, Feng, Zhang, Chi

Abstract

Large language model (LLM) agents are increasingly used as personalized assistants in high-stakes domains such as financial advising, yet it remains unclear whether they can maintain and update an individualized user model over long horizons. Existing personalized-memory benchmarks primarily test factual retention or rely on weakly constrained model-generated trajectories, leaving event-driven preference adaptation underexplored. We introduce FinPerMA, an event-grounded benchmark that evaluates personalized memory against frozen longitudinal investor trajectories. Its generation pipeline combines deterministic, theory-informed impact rules, controlled LLM narration, and automated quality screening; a Post-Shock checkpoint isolates whether an agent has integrated a material event into its persistent user model. On 2,994 questions from 276 personas, seven frontier LLMs and up to seven memory configurations remain far from saturated: no full-context configuration exceeds approximately 0.47 overall accuracy or approximately 39% on multiple-choice questions. Attribution analysis shows that summary-based memory often preserves factual details while losing the preference signals needed for personalization; simple retrieval can therefore outperform purpose-built memory systems, with the gap widening after shocks.

Chinese Translation

大型语言模型（LLM）代理在金融顾问等高风险领域中越来越多地被用作个性化助手，但尚不清楚它们是否能够在长时间内维持和更新个体化用户模型。现有的个性化记忆基准主要测试事实保留或依赖于弱约束的模型生成轨迹，导致事件驱动的偏好适应尚未得到充分探索。我们提出了FinPerMA，这是一个基于事件的基准，评估个性化记忆与冻结的长期投资者轨迹的匹配。其生成流程结合了确定性的、基于理论的影响规则、受控的LLM叙述和自动化质量筛选；“后冲击”检查点用于隔离代理是否将重要事件整合到其持久的用户模型中。在来自276个个体的2,994个问题中，七个前沿的LLM和多达七种记忆配置仍远未饱和：没有任何全上下文配置的整体准确率超过约0.47，或在多项选择题中超过约39%。归因分析表明，基于摘要的记忆通常保留事实细节，但失去了个性化所需的偏好信号；因此，简单的检索可以优于专门构建的记忆系统，并且在冲击后这一差距进一步扩大。

View on arXiv Download PDF AI Translation

cs.AI / 6 / 2608.04156

BrainBench: Benchmarking Large Language Models for Comprehensive EEG Understanding

BrainBench：大型语言模型在综合脑电图理解中的基准测试

Zhou, Yangxuan, Zhao, Sha, Chen, Yuning, Wu, Chen, Wang, Jiquan, Li, Shijian, Pan, Gang

Abstract

Electroencephalography (EEG) analysis extends beyond assigning predefined labels to recordings; it requires workflows connecting natural-language instructions, signal processing, quantitative evidence, and scientific interpretation. We term this capability \emph{comprehensive EEG understanding}. Existing evaluations, however, primarily target isolated decoding tasks or system-specific demonstrations, leaving the competence of large language models (LLMs) insufficiently quantified. We introduce \benchmarkname{}, a unified benchmark for comprehensive, instruction-conditioned EEG understanding. It comprises four subsets---Foundational Analysis, Sleep Assessment, Neurocognitive Assessment, and Physiological Integration---covering 17 datasets, \numcases{} tasks, and over \numinstances{} real-data instances. Given an instruction and EEG recordings with optional physiological signals, a system must perform the analysis and produce a scientifically grounded report and, when required, artifacts. Outputs are assessed through numerical, categorical, set, sequence, semantic, and artifact validation. We evaluate \nummodels{} representative LLMs across more than 100K executions under two paradigms: autonomous code execution with CodeAct and structured agentic analysis with BrainAgent. Results vary substantially across models, subsets, difficulty levels, and execution paradigms, showing that EEG competence depends on the model and its operationalization. \benchmarkname{} provides a reproducible testbed for advancing LLM-based EEG understanding. The code and benchmark will be released soon, with evaluation results continuously updated.

Chinese Translation

脑电图（EEG）分析不仅仅是将预定义标签分配给记录；它需要将自然语言指令、信号处理、定量证据和科学解释连接起来的工作流程。我们将这种能力称为 extit{综合脑电图理解}。然而，现有的评估主要针对孤立的解码任务或特定系统的演示，导致大型语言模型（LLMs）的能力尚未得到充分量化。我们引入了enchmarkname{}，这是一个统一的基准，用于综合的、基于指令的脑电图理解。它包括四个子集——基础分析、睡眠评估、神经认知评估和生理整合——覆盖17个数据集、 umcases{}个任务和超过 uminstances{}个真实数据实例。在给定指令和可选生理信号的脑电图记录的情况下，系统必须执行分析并生成科学依据的报告，并在需要时提供相关文献。输出通过数值、类别、集合、序列、语义和文献验证进行评估。我们在两种范式下评估了 ummodels{}个代表性LLMs，超过10万次执行：使用CodeAct的自主代码执行和使用BrainAgent的结构化代理分析。结果在模型、子集、难度级别和执行范式之间差异显著，表明脑电图能力依赖于模型及其操作化。 enchmarkname{}提供了一个可重复的测试平台，以推动基于LLM的脑电图理解。代码和基准将很快发布，评估结果将持续更新。

View on arXiv Download PDF AI Translation

cs.AI / 7 / 2608.04190

Adversarially Robust Abductive Fusion of Pre-trained Transformer-based Perception Models

对抗鲁棒的基于预训练变换器的感知模型的推理融合

Leiva, Mario, Ma, Yue, Qiu, Qinru, Simari, Gerardo, Shakarian, Paulo

Abstract

Deploying pre-trained perception models in novel environments degrades their accuracy under distributional shift, and assembling them alone does not recover it: combiners such as majority voting trade recall for precision and are brittle to coordinated failures. Prior metacognitive methods learn logical rules that flag a model's errors, but rely on hand-authored domain-knowledge cues (object-size priors, segmentation masks) that do not transfer to genuinely novel scenes. We show that this metacognitive layer can be learned without any domain knowledge by exploiting vector-space geometry: per-model Label Vector Pools (LVP), built from each model's own training embeddings, yield error-detection rules from the geometry of detections relative to training-determined prototypes, reaching parity with domain-knowledge rules to within $0.002$ every F1 on test set. Because the approach remains neurosymbolic, these geometric rules share a single logical framework and can still be complemented by domain knowledge when available. We frame the fusion of multiple imperfect ViT-based detectors as a consistency-based abduction problem solved at test time by an exact Integer Program (IP) and a polynomial-time heuristic. On an aerial-imagery benchmark of 15 weather-shifted test sets and six ViT detectors, our domain-knowledge-free layer matches the strongest majority-vote variant on clean data (within $0.005$ F1) and, unlike every majority-vote baseline, retains its performance under a coordinated label-flipping attack: at a $90\%$ flip rate it averages $0.42$ F1 versus $0.35$ for MV-Plurality (a $22\%$ relative gain) and attains the highest F1 on \emph{every} test set once the flip rate exceeds $0.4$

Chinese Translation

在新环境中部署预训练的感知模型会导致其准确性在分布转移下下降，单独组合这些模型并不能恢复其准确性：诸如多数投票的组合方法在提高召回率的同时牺牲了精确度，并且对协调性故障非常脆弱。先前的元认知方法学习逻辑规则来标记模型的错误，但依赖于手工编写的领域知识提示（如物体大小先验、分割掩码），这些提示无法转移到真正的新场景中。我们展示了这个元认知层可以通过利用向量空间几何而无需任何领域知识来学习：每个模型的标签向量池（Label Vector Pools, LVP）是由每个模型自身的训练嵌入构建的，从相对于训练确定的原型的检测几何中得出错误检测规则，其性能与领域知识规则在测试集上的F1值相差不超过0.002。由于该方法保持了神经符号的特性，这些几何规则共享一个统一的逻辑框架，并且在可用时仍可以通过领域知识进行补充。我们将多个不完美的基于ViT的检测器的融合框架视为一个基于一致性的推理问题，该问题在测试时通过精确的整数规划（Integer Program, IP）和多项式时间启发式方法来解决。在一个包含15个天气变化测试集和六个ViT检测器的航空图像基准测试中，我们的无领域知识层在干净数据上与最强的多数投票变体相匹配（F1值相差不超过0.005），并且与每个多数投票基线不同，在协调标签翻转攻击下保持其性能：在90%的翻转率下，其平均F1值为0.42，而MV-Plurality为0.35（相对增益为22%），并且在翻转率超过0.4时在每个测试集上达到了最高的F1值。

View on arXiv Download PDF AI Translation

cs.AI / 8 / 2608.04205

MatrAIx: Simulating the World with 8.3 Billion Persona Agents

MatrAIx：用83亿个角色代理模拟世界

Li, Xiaomin, Hao, Yuexing, Hou, Jianheng, Huang, Jintao, Wen, Qianfeng, Huang, Shirley, Liu, Yifan, Liu, Xiaoyi, Fan, Yilan, Wang, Yijun, Wu, Koutian, Gao, Ruoqi, Mohsin, Muhammad Ahmed, Tang, Jing, Joshi, Brihi, Liu, Heming, Deng, Zheyuan, Di, Zonglin, Jajee, Sankalp, Lu, Jiuyao, Zhang, Zhiwei, Kapoor, Saksham, Gupta, Ishan, Zhao, Yunhan, Park, Chanwoo, Lu, Yucheng, Hu, Bing, Xiao, Weihang, Mohan, Aravind, Xing, Hanwen, Zhang, Runyu, Kulshreshtha, Mihir, Xu, Yuanda, Zhu, Qianyu, Wang, Dianzhuo, Xiao, Yuxin, Jiang, Bowen, Su, Yongye, Chai, Wenhao, Liu, Zuxin, Chen, Lawrence Yunliang, Zhao, Xuandong, Ye, Ethan, Patel, Shivam, Xie, Jason, Richmond, Alex Martin, Ding, Weixiang, Okcular, Emre, Mathew, Diya, Wang, Ziheng, Khan, Rana M. Shahroz, Peng, Zhejian, Wu, Fang, Nie, Fan, Han, Xinyang, Kim, Yubin, Zhang, Jiawei, Qi, Zhenting, Su, Huangyuan, Pan, Xu, Gourabathina, Abinitha, Jeong, Hyewon, Ramesh, Hemanth Neelgund, Alhamoud, Kumail, Hamidieh, Kimia, Xiong, Zidi, Schmidgall, Samuel, Han, Pengrui, Huang, Yepeng, Wang, Yongheng, Yang, Bowen, Gu, Alex, Wang, Yuchu, Paruchuri, Akshay, Li, Brenna, Cui, Hejie, Ding, Jiayuan, Dong, Chaosheng, Wang, Jiahao, He, Yixuan, Wang, Chi, Bhattacharya, Pamela, Peng, Tianyi, Liang, Paul Pu, Gordon, Mitchell, Du, Yilun, Zitnik, Marinka, Zou, James, Tambe, Prasanna, Torr, Philip, Fox, Emily, Ozdaglar, Asu, Song, Dawn

Abstract

Human evaluation of AI systems and digital products is costly, slow, and difficult to scale. Offline evaluations are more scalable but often abstract away human diversity and interactive behavior. We therefore introduce MatrAIx, a population-scale simulated-user evaluation infrastructure for testing AI systems and digital products with heterogeneous users. MatrAIx has three core components: First, Persona 8B contains 8.3 billion persona records represented by 1,290 categorical dimensions. Records are either sampled from a dependency graph that preserves correlated attributes or derived from human-authored profiles. We release a quality-filtered coreset of approximately 1 million personas, comprising 599,847 human-grounded and 400,000 synthetic records. Second, the MatrAIx Playground provides four environments in which diverse users evaluate and interact with digital products: Survey, AI Chatbot, Web, and App. Third, MatrAIx provides 1,010 application tasks spanning more than 25 domains, including Commerce, Software, Finance, and Healthcare. We conducted 18,189 evaluation trials across eight representative tasks. Persona agents were powered by three LLMs: Claude Opus 4.8, GPT 5.5, and Claude Haiku 4.5. The resulting feedback captures how decisions and preferences vary across persona backgrounds, including hesitation after a price increase, willingness to continue after an AI assistant fails, and latency tolerance. We conducted two main validation studies: First, a 400-trial controlled study evaluated persona adherence across ten behavioral attributes and all four environments. The declared behavior was expressed or correctly suppressed in 366 trials (91.5%). Second, human and LLM judges evaluated the extraction quality of human-grounded personas. Overall, MatrAIx provides an end-to-end infrastructure for evaluating AI systems and digital products with diverse simulated human users.

Chinese Translation

对人工智能系统和数字产品的人类评估成本高、速度慢且难以扩展。离线评估更具可扩展性，但往往忽视了人类的多样性和互动行为。因此，我们推出了MatrAIx，一个用于测试人工智能系统和数字产品的群体规模模拟用户评估基础设施，旨在涵盖异质用户。MatrAIx有三个核心组件：首先，Persona 8B包含83亿个角色记录，这些记录由1290个类别维度表示。记录要么是从保持相关属性的依赖图中抽样而来，要么是从人类编写的个人资料中派生的。我们发布了一个经过质量过滤的约100万个角色的核心数据集，其中包括599,847个基于人类的记录和400,000个合成记录。其次，MatrAIx Playground提供了四个环境，在这些环境中，多样化的用户可以评估和互动数字产品：调查、AI聊天机器人、网页和应用程序。第三，MatrAIx提供了1,010个应用任务，涵盖25个以上的领域，包括商业、软件、金融和医疗保健。我们在八个代表性任务中进行了18,189次评估试验。角色代理由三个大型语言模型（LLMs）驱动：Claude Opus 4.8、GPT 5.5和Claude Haiku 4.5。所获得的反馈捕捉了决策和偏好如何因角色背景而异，包括在价格上涨后的犹豫、在AI助手失败后的继续意愿以及延迟容忍度。我们进行了两项主要的验证研究：首先，一项400次试验的对照研究评估了十个行为属性和所有四个环境中的角色遵循情况。声明的行为在366次试验中得到了表达或正确抑制（91.5%）。其次，人类和LLM评审者评估了基于人类的角色的提取质量。总体而言，MatrAIx提供了一个端到端的基础设施，用于评估具有多样化模拟人类用户的人工智能系统和数字产品。

View on arXiv Download PDF AI Translation

cs.AI / 9 / 2608.04232

Interoceptive Attention as Dynamic Homeostatic Prioritization in a Foraging Agent

Grimbly, St John, Kuske, Nicolas, Boonstra, Evert A., Bassett, Bruce A., van Hoof, Charel, Hodson, Rowan, Rosman, Benjamin, Smith, Ryan, Solms, Mark, Shock, Jonathan P.

Abstract

Biological systems must regulate competing needs under limited perceptual bandwidth, where sharpening one estimate costs the capacity to sharpen the others. Any fixed-budget system therefore has to decide where to allocate its perceptual precision. We study this in a foraging agent that must keep several bodily needs satisfied to survive, modelled with active inference. At each step it reads its own body-state beliefs, identifies the most-needed channel, and reallocates a fixed budget of interoceptive precision toward it, so that the same precision-shaped likelihood feeds both belief update and planning. In AffectWorld, a four-channel foraging gridworld, this selective allocation more than doubles learning-phase survival at matched budget against a uniform-precision agent ($0.414$ vs $0.199$ across 11 layouts, $n{=}32$ seeds each, paired cluster-bootstrap $p \leq 10^{-4}$). Two further results sharpen the mechanism. The benefit runs through planning as well as perception, since denying the shaped likelihood to the planner alone removes about half of it. It is also need-aligned, since aiming precision at the least-needed channel does worse than spreading it evenly. The attended channel additionally learns its own dynamics about twice as fast, and stays ahead even at matched observation count, a behavioural trace of the same precision routing, visible in learning speed, not survival.

View on arXiv Download PDF AI Translation

cs.AI / 10 / 2608.04285

The RAIL Principles for Neurosymbolic AI: Reasoning, Assurances, Interfacing and Learning

神经符号人工智能的RAIL原则：推理、保证、接口与学习

Chiatti, Agnese, Cochez, Michael, Cornelio, Cristina, Dumancic, Sebastijan, Garcez, Artur d'Avila, Lamb, Luis C., Morra, Lia, Niepert, Mathias, Peharz, Robert, Speranzon, Alberto, Stol, Maarten, Teije, Annette Ten, Thanapalasingam, Thiviyan, Van Harmelen, Frank, Van Krieken, Emile, Vergari, Antonio, Wang, Benjie

Abstract

Neurosymbolic AI systems that integrate machine learning and symbolic reasoning are rapidly gaining attention. They complement the data-intensive statistical approaches of neural networks and language models with symbolic reasoning algorithms to function in high-stakes domains or in low-data regimes that characterize many real-world applications. We argue that the neurosymbolic combination of machine learning and formal reasoning is not a niche approach within AI, but rather includes many already successful techniques that are of crucial importance to the development of reliable, efficient and, ultimately, trustworthy systems. This perspective prompts a re-examination of the design of current AI systems. We show that many leading AI systems, including some that are not traditionally considered as neurosymbolic, can be analysed from the perspective of four principles of neurosymbolic AI design: Reasoning, Assurances, Interfacing and Learning (RAIL). Applying the RAIL framework offers a unified view of seemingly disparate AI systems, ranging from physics-aware machine learning to neuro-guided search (such as Google DeepMind's Alpha-* suite), causal learning and tool-augmented Large Language Models. Importantly, the RAIL principles will enable engineers to make better-informed and more principled decisions about the design and deployment of production-level AI systems. In this article, we introduce the RAIL principles, examine how they can be applied across major areas of AI, and illustrate how they may guide practitioners to integrate neurosymbolic methods into next-generation AI technologies.

Chinese Translation

集成机器学习和符号推理的神经符号人工智能系统正在迅速获得关注。它们通过符号推理算法补充了神经网络和语言模型的数据密集型统计方法，以便在高风险领域或低数据环境中运行，这些环境特征是许多现实世界应用的特点。我们认为，机器学习与形式推理的神经符号结合并不是人工智能中的小众方法，而是包括许多已经成功的技术，这些技术对开发可靠、高效且最终值得信赖的系统至关重要。这一观点促使我们重新审视当前人工智能系统的设计。我们展示了许多领先的人工智能系统，包括一些传统上不被视为神经符号的系统，可以从神经符号人工智能设计的四项原则（推理、保证、接口与学习，RAIL）的角度进行分析。应用RAIL框架提供了一个统一的视角，涵盖了看似不同的人工智能系统，从物理感知的机器学习到神经引导搜索（如谷歌DeepMind的Alpha-*系列）、因果学习和工具增强的大型语言模型。重要的是，RAIL原则将使工程师能够在设计和部署生产级人工智能系统时做出更为明智和更有原则的决策。在本文中，我们介绍了RAIL原则，探讨了它们如何在人工智能的主要领域中应用，并说明了它们如何指导从业者将神经符号方法整合到下一代人工智能技术中。

View on arXiv Download PDF AI Translation

cs.AI / 11 / 2608.04289

SafeCommit: Certifying When Memory-Grounded Agents May Safely Act

安全承诺：认证内存基础代理何时可以安全行动

Akewar, Mayur, Ranjan, Ravi

Abstract

Long-horizon agents increasingly use persistent memory and tools to take actions with external side effects. A central failure mode is premature commitment: an agent acts before resolving whether its memory grounding is stale, conflicting, incomplete, or corrupted. We formalize this problem as safe commitment under memory uncertainty and introduce SafeCommit, a risk controlled layer between agent reasoning and external execution. The layer constructs a calibrated set of plausible latent worlds from memory, observations, tool outputs, provenance, and policy constraints. It permits a side effectful action only when a conformal action certificate shows that the action is safe in every retained world. Otherwise, it selects a low-side-effect probe that targets the worlds blocking certification, or returns a conservative fallback. Under calibrated world coverage, the probability of an unsafe certified commit is at most the target level {\alpha}; with imperfect world proposal, the bound separates calibration and representation error. A dependency-free controlled simulator illustrates the safety-utility tradeoff and reproduces all reported results with one command. The goal is to offer a concrete approach for deciding not only what an agent should do, but when the available evidence is sufficient to safely do it.

Chinese Translation

长期代理越来越多地使用持久内存和工具来采取具有外部副作用的行动。一个主要的失败模式是过早承诺：代理在未解决其内存基础是否过时、冲突、不完整或损坏之前就采取行动。我们将这个问题形式化为在内存不确定性下的安全承诺，并引入了SafeCommit，这是一个在代理推理与外部执行之间的风险控制层。该层从内存、观察、工具输出、来源和策略约束中构建出一组经过校准的合理潜在世界。仅当一个符合条件的行动证书表明该行动在每个保留的世界中都是安全时，才允许进行具有副作用的行动。否则，它会选择一个低副作用的探测行动，针对阻止认证的世界，或返回一个保守的后备方案。在经过校准的世界覆盖下，不安全的认证承诺的概率至多为目标水平{}; 在不完美的世界提议下，界限将校准和表示误差分开。一个无依赖的受控模拟器展示了安全性与效用之间的权衡，并通过一个命令重现了所有报告的结果。我们的目标是提供一种具体的方法，不仅决定代理应该做什么，而且决定何时可用证据足以安全地执行该行动。

View on arXiv Download PDF AI Translation

cs.AI / 12 / 2608.04358

NeuMoSync: End-to-End Neuromodulatory Control for Plasticity and Adaptability in Continual Learning

NeuMoSync：用于持续学习中的可塑性和适应性的端到端神经调节控制

Rohani, Seyed Roozbeh Razavi, Khajavi, Khashayar, Chung, Wesley, Samiei, Mandana, Chen, Mo

Abstract

Continual learning (CL) requires models to learn tasks sequentially, yet deep neural networks often suffer from plasticity loss and poor knowledge transfer, which can impede their long-term adaptability. Drawing high-level inspiration from global neuromodulatory mechanisms in the brain, we introduce Neuromodulation and Synchronization (NeuMoSync), a novel architecture that integrates dynamic, neuron-specific modulation into deep neural networks to enhance their adaptability and plasticity. NeuMoSync extends standard neural network architectures with learnable feature vectors for each neuron that track network-wide historical context and with a module operating at a higher level of abstraction. This module synthesizes neuron-specific signals, conditioned on both current inputs and the network's evolving state, to adaptively regulate activation dynamics and synaptic plasticity. Evaluated on diverse CL benchmarks, including memorization (Random Label CIFAR-10 and Random Label MNIST), concept drift (Shuffle CIFAR-10 and Shuffle Mini-ImageNet), class-incremental learning (Class Split ImageNet and Class Split CIFAR-100), and domain-incremental learning (Permuted MNIST), NeuMoSync demonstrates strong performance in retaining plasticity and achieves improvements in both forward and backward adaptation compared with existing methods. Ablation studies validate the necessity of each component, while analysis of the learned modulatory signals reveals interpretable coordination patterns across tasks. Our work underscores the potential of integrating global coordination mechanisms into deep learning systems to advance robust, adaptive continual learning. The code is publicly available at https://github.com/RoozbehRazavi/NeuMoSync.

Chinese Translation

持续学习（CL）要求模型顺序学习任务，但深度神经网络往往面临可塑性丧失和知识转移不良的问题，这可能阻碍其长期适应能力。我们从大脑中的全球神经调节机制中获得高层次的灵感，提出了神经调节与同步（NeuMoSync），这是一种新颖的架构，将动态的、特定于神经元的调节集成到深度神经网络中，以增强其适应性和可塑性。NeuMoSync 扩展了标准神经网络架构，为每个神经元引入可学习的特征向量，以跟踪网络范围内的历史上下文，并配备一个在更高抽象层次上操作的模块。该模块综合了特定于神经元的信号，这些信号基于当前输入和网络不断演变的状态，以自适应地调节激活动态和突触可塑性。在多种 CL 基准测试上进行评估，包括记忆（随机标签 CIFAR-10 和随机标签 MNIST）、概念漂移（洗牌 CIFAR-10 和洗牌迷你图像网）、类增量学习（类分割图像网和类分割 CIFAR-100）以及领域增量学习（置换 MNIST），NeuMoSync 在保持可塑性方面表现出色，并在前向和后向适应性方面相比现有方法取得了改进。消融研究验证了每个组件的必要性，而对学习到的调节信号的分析揭示了跨任务的可解释协调模式。我们的工作强调了将全球协调机制集成到深度学习系统中的潜力，以推动稳健的、自适应的持续学习。代码已公开发布在 https://github.com/RoozbehRazavi/NeuMoSync。

View on arXiv Download PDF AI Translation

cs.AI / 13 / 2608.04384

Improving Auto-Design of Neural PDE Solvers with a Domain-Specific Language

利用领域特定语言改善神经偏微分方程求解器的自动设计

Kong, Shengxin, Xu, Liwen, Fu, Jingwen

Abstract

Neural PDE solver auto-design is fundamentally a search-space representation problem. In the space of unrestricted Python programs, valid solvers form an extremely sparse subset: most candidate programs are syntactically incorrect, semantically incompatible, or numerically unstable. Direct code generation therefore forces an LLM to spend most of its search capacity navigating implementation failures rather than reasoning about solver quality. ADSL-PDE addresses this challenge by introducing a structured search state between solver concepts and executable code. It represents the functional decisions that determine a neural PDE solver (architecture, physical constraints, objectives, sampling, and optimization) while abstracting away low-level implementation details. A deterministic compiler maps each valid search state to an executable solver. In effect, ADSL-PDE reshapes the search space: it removes large regions of invalid programs, increases the density of meaningful candidates, and preserves the compositional freedom needed to discover previously unseen designs. Solver evolution can thus operate over design decisions rather than code artifacts. Built on this representation, our evolutionary agent iteratively proposes, evaluates, and refines solver search states using empirical feedback. Across multiple PDE benchmarks, ADSL-PDE improves both search efficiency and optimization stability, achieving an improvement of more than 52% within the first ten evolution iterations. These results suggest a broader principle for LLM-driven auto-design: effective agents do not merely require stronger reasoning, but rather a search representation that concentrates exploration on valid and consequential decisions.

Chinese Translation

神经偏微分方程（PDE）求解器的自动设计本质上是一个搜索空间表示问题。在不受限制的Python程序空间中，有效的求解器形成了一个极其稀疏的子集：大多数候选程序在语法上不正确、在语义上不兼容或在数值上不稳定。因此，直接代码生成迫使大型语言模型（LLM）将大部分搜索能力用于处理实现失败，而不是推理求解器的质量。ADSL-PDE通过在求解器概念与可执行代码之间引入结构化搜索状态来解决这一挑战。它表示决定神经PDE求解器的功能性决策（架构、物理约束、目标、采样和优化），同时抽象掉低级实现细节。一个确定性编译器将每个有效的搜索状态映射到一个可执行的求解器。实际上，ADSL-PDE重塑了搜索空间：它移除了大量无效程序区域，增加了有意义候选的密度，并保留了发现以前未见设计所需的组合自由度。因此，求解器的演化可以在设计决策上进行，而不是在代码工件上进行。在这一表示的基础上，我们的进化代理通过经验反馈迭代地提出、评估和优化求解器搜索状态。在多个PDE基准测试中，ADSL-PDE提高了搜索效率和优化稳定性，在前十次演化迭代中实现了超过52%的改进。这些结果表明了一个更广泛的原则，适用于LLM驱动的自动设计：有效的代理不仅需要更强的推理能力，还需要一个将探索集中在有效且重要决策上的搜索表示。

View on arXiv Download PDF AI Translation

cs.AI / 14 / 2608.04458

Architectural Implications of Agentic AI Workflows

代理式人工智能工作流的架构影响

Yang, Jirong, Liu, Peizhe, Zhang, Chaojie, Stojkovic, Jovan

Abstract

Agentic AI is emerging in datacenters, but its architectural implications remain unexplored. We organize agentic workflows in a taxonomy and present its first architectural characterization with a production study at Microsoft Azure and a controlled study of open-source frameworks. We show that agentic execution is fragmented and heterogeneous. Requests expand into a workflow of LLM inferences, tool invocations, and orchestration decisions that repeatedly cross the CPU-GPU boundary. Our taxonomy explains how this fragmentation turns into resource demand. As orchestration and tools run on the host, the CPU sits on the critical path. Execution structure sets the load over time, which stays low with sudden spikes. Model composition sets how evenly the workflow uses the GPUs. Diversity in tasks and tools widens this range even further. These characteristics expose architectural mismatches of conventional uniform servers. Fragmented execution strands CPU and GPU capacity despite bursty demand. Different software roles make homogeneous CPU provisioning inefficient. Finally, multiplexing many agents onto shared cores degrades microarchitectural locality. Guided by our findings, we derive implications for agentic servers and examine them through Agora, our prototype for commodity servers. Agora dynamically harvests idle CPU cores for co-located throughput work, while protecting agentic tail latency against tool spikes. It oversubscribes GPU memory by placing more agents on each GPU, prefetching the next agent's state to hide swap latency. To match the machine to the heterogeneous roles, Agora pools cores by role and applies affinity-aware scheduling to restore locality. It automatically tunes mechanisms to the workload. Agora improves utilization and server throughput while preserving agent tail latency. Our insights also identify key directions for future server architectures for agentic AI.

Chinese Translation

代理式人工智能正在数据中心中崭露头角，但其架构影响尚未得到探索。我们对代理式工作流进行了分类，并通过在微软Azure的生产研究和对开源框架的控制研究，首次呈现其架构特征。我们展示了代理执行是碎片化和异构的。请求扩展为一个包含大型语言模型（LLM）推理、工具调用和编排决策的工作流，这些决策反复跨越CPU-GPU边界。我们的分类解释了这种碎片化如何转化为资源需求。由于编排和工具在主机上运行，CPU处于关键路径上。执行结构随时间设置负载，负载在突然的高峰时保持较低。模型组合决定了工作流对GPU的使用均匀程度。任务和工具的多样性进一步扩大了这一范围。这些特征暴露了传统统一服务器的架构不匹配。尽管需求突发，碎片化的执行使CPU和GPU的能力分散。不同的软件角色使得同质化的CPU配置效率低下。最后，将多个代理复用到共享核心上会降低微架构的局部性。在我们的研究发现的指导下，我们推导出代理服务器的影响，并通过Agora（我们的商品服务器原型）进行检验。Agora动态收集空闲的CPU核心以进行共置的吞吐工作，同时保护代理的尾延迟不受工具高峰的影响。它通过在每个GPU上放置更多代理来过度订阅GPU内存，并预取下一个代理的状态以隐藏交换延迟。为了将机器与异构角色匹配，Agora按角色汇聚核心并应用亲和性感知调度以恢复局部性。它自动调整机制以适应工作负载。Agora提高了利用率和服务器吞吐量，同时保持了代理的尾延迟。我们的见解还确定了未来代理式人工智能服务器架构的关键方向。

View on arXiv Download PDF AI Translation

cs.AI / 15 / 2608.04509

CARGO-VL: Counterfactual Arbitration with Risk-Constrained Group Optimization for Vision-Language Models

CARGO-VL：具有风险约束的组优化的反事实仲裁用于视觉-语言模型

Jiang, De, Zhang, Zhengyang, Yuan, Kehong, Ma, Shaohua

Abstract

Vision-language systems combine images with retrieved text, but these sources can disagree or jointly fail to support an answer. Reliable models must identify the trustworthy source and abstain when neither is adequate. Existing post-training objectives score instances independently and therefore do not enforce coherent behavior under counterfactual evidence changes. We introduce CARGO-VL, a group-relative framework that optimizes matched variants covering aligned, image-correct, text-correct, and both-wrong (A/V/T/N) evidence states as one bundle. Its objective couples condition-wise correctness with transition rewards for answer invariance, source equivariance, and answer-to-abstention switching, while a primal-dual controller balances unsafe answers against excessive deferral. We also contribute XMC (eXtended Modal Conflict), a four-condition conflict training resource, and evaluate transfer on CMC-Bench and Modality-Bias. Across multiple seeds, CARGO-VL improves conflict handling, unsupported-answer avoidance, and modality balance over pointwise baselines. Ablations identify complementary benefits from relational transition signals and adaptive risk control, supporting counterfactual consistency as a practical objective for reliable multimodal evidence arbitration.

Chinese Translation

视觉-语言系统将图像与检索到的文本结合在一起，但这些来源可能存在不一致或共同无法支持答案的情况。可靠的模型必须识别可信的来源，并在两者都不充分时选择放弃。现有的后训练目标独立评分实例，因此在反事实证据变化下并未强制执行一致的行为。我们提出了CARGO-VL，这是一个相对组框架，优化覆盖对齐的、图像正确的、文本正确的以及两者都错误的（A/V/T/N）证据状态的匹配变体，作为一个整体。其目标将条件正确性与答案不变性、来源等变性和答案与放弃之间的切换的过渡奖励相结合，同时一个原始-对偶控制器平衡不安全答案与过度延迟之间的关系。我们还贡献了XMC（扩展模态冲突），这是一个四条件冲突训练资源，并在CMC-Bench和Modality-Bias上评估转移效果。在多个随机种子下，CARGO-VL在冲突处理、避免不支持答案和模态平衡方面优于逐点基线。消融实验表明，关系过渡信号和自适应风险控制带来了互补的好处，支持反事实一致性作为可靠的多模态证据仲裁的实际目标。

View on arXiv Download PDF AI Translation

cs.AI / 16 / 2608.04519

Leak-Resistant Unlearning: A New Benchmark for Evaluating Multi-Hop Reasoning Consistency and Recovery Robustness

抗泄漏的遗忘：评估多跳推理一致性和恢复鲁棒性的新基准

Qian, Haoting, Zhang, Qingjie, Huang, Zhicong, Hong, Cheng, Qiu, Han

Abstract

Benchmarking machine unlearning methods is critical to understand whether sensitive knowledge is removed from large language models (LLMs) or not. Current unlearning benchmarks include mainly single-hop questions and a narrow set of multi-hop questions. Although effective, they still face two challenges. (1) Knowledge is not isolated, whereby diverse multi-hop reasoning paths can potentially induce knowledge leakage than normal queries. (2) Unlearning may be fragile: unlearned knowledge can be partially recovered through recovery attacks such as lightweight post-unlearning adaptation, making static evaluation insufficient. Therefore, in this paper, we introduce \unlearning as a novel benchmark to understand robust LLM knowledge removal across diverse reasoning paths and recovery attacks. We experiment with this benchmark on 3 models, 6 unlearning methods, and 2 carefully curated datasets. Results show that existing methods are vulnerable to multi-hop reasoning paths and recovery attacks. We further explore the trade-off among forget quality, robustness, and model utility for LLM unlearning.

Chinese Translation

评估机器遗忘方法对于理解敏感知识是否从大型语言模型（LLMs）中移除至关重要。目前的遗忘基准主要包括单跳问题和一小部分多跳问题。尽管有效，但它们仍面临两个挑战。（1）知识并非孤立，丰富的多跳推理路径可能比普通查询更容易导致知识泄漏。（2）遗忘可能是脆弱的：通过轻量级的后遗忘适应等恢复攻击，遗忘的知识可能部分恢复，使得静态评估显得不足。因此，在本文中，我们引入了 extit{unlearning}作为一个新颖的基准，以理解在多样化推理路径和恢复攻击下，鲁棒的LLM知识移除。我们在3个模型、6种遗忘方法和2个精心策划的数据集上对该基准进行了实验。结果表明，现有方法对多跳推理路径和恢复攻击存在脆弱性。我们进一步探讨了LLM遗忘中遗忘质量、鲁棒性和模型效用之间的权衡。

View on arXiv Download PDF AI Translation

cs.AI / 17 / 2608.04562

What Is a Skill Worth? Structure-Aware Shapley Valuation of Agent Skills

技能的价值是什么？基于结构的 Shapley 价值评估代理技能

Li, Tao, Liu, Junfeng, Zhao, Qinghua, Li, Yifan, Wang, Lei, Shao, Bo, Liu, Xuejun, Shou, Linjun

Abstract

Agent skills are increasingly optimized by automated feedback loops, producing long structured artifacts whose internal value remains unclear. We study skill valuation: assigning credit to the internal units of a fixed skill, such as rules, examples, scripts, and heuristics, under a fixed agent and held-out task distribution. Skill valuation differs from data or prompt-span valuation because skill units are structured: they may depend on other units, belong to a document hierarchy, trigger agent behavior, and consume limited prompt context. We introduce SkillSV, a structure-aware Shapley-style framework for skill valuation. SkillSV compiles a skill into units, dependencies, and hierarchy, so that only valid counterfactual skills are evaluated. It uses paired deletion and length-neutral padding to separate content value from context cost, and estimates the resulting values with a rollout-budgeted estimator for noisy agent evaluations. On four agentic benchmarks, we assess the faithfulness, actionability, and explanation of SkillSV: it recovers unit interactions, preserves aggregate skill lift, and guides safe pruning and compression.

Chinese Translation

代理技能越来越多地通过自动反馈循环进行优化，产生了内部价值尚不明确的长结构化工件。我们研究技能评估：在固定代理和保留任务分布下，将内部单位（如规则、示例、脚本和启发式）分配信用。技能评估不同于数据或提示跨度评估，因为技能单位是结构化的：它们可能依赖于其他单位，属于文档层次结构，触发代理行为，并消耗有限的提示上下文。我们引入了 SkillSV，这是一种基于结构的 Shapley 风格的技能评估框架。SkillSV 将技能编译为单位、依赖关系和层次结构，从而仅评估有效的反事实技能。它使用配对删除和长度中性填充将内容价值与上下文成本分离，并使用预算回滚估计器估计噪声代理评估的结果值。在四个代理基准上，我们评估了 SkillSV 的可信度、可操作性和解释性：它恢复了单位交互，保持了整体技能提升，并指导安全的修剪和压缩。

View on arXiv Download PDF AI Translation

cs.AI / 18 / 2608.04590

Joint UAV Flight and Opportunistic Routing under Reinforcement Learning for Delay-Tolerant Networks

基于强化学习的延迟容忍网络中的无人机联合飞行与机会路由

Wang, Xiao, Yang, Shun-Ren

Abstract

The growing deployment of delay-tolerant networks (DTNs) has made store-carry-forward (SCF) communication indispensable under sparse connectivity. However, intermittent contacts, finite buffers, and limited message time-to-live (TTL) often give rise to sparse delivery and congestion, leading to substantial end-to-end performance degradation. To address this challenge, this study explores the joint optimization of decentralized opportunistic routing and controllable unmanned aerial vehicle (UAV) flight, aiming to enlarge future contacts through discrete UAV headings while enabling per-node replication under contact-limited observations. Building upon this architecture, we study cooperative factored routing--UAV control under centralized training and decentralized execution (CTDE) and propose JUROR (Joint UAV flight and Opportunistic Routing, based on the proximal policy optimization (PPO) framework. In our design, we first cast the problem as a factored partially observable Markov decision process with sequential motion--routing coupling and a per-step team reward; subsequently, decentralized actors act on local observations while a training-time critic uses global statistics, and an optional multi-horizon hotspot predictor provides auxiliary supervision. Simulation results over four traffic modes demonstrate effective gains over PRoPHET and MaxProp, while retaining contact-limited decentralized execution.

Chinese Translation

延迟容忍网络（DTNs）的日益普及使得在稀疏连接下，存储-携带-转发（SCF）通信变得不可或缺。然而，间歇性接触、有限缓冲区和有限消息生存时间（TTL）常常导致稀疏交付和拥堵，从而造成显著的端到端性能下降。为了解决这一挑战，本研究探讨了去中心化机会路由与可控无人机（UAV）飞行的联合优化，旨在通过离散的无人机航向扩大未来的接触，同时在接触受限的观察下实现每个节点的复制。在此架构基础上，我们研究了合作分解路由——在集中训练和去中心化执行（CTDE）下的无人机控制，并提出了JUROR（联合无人机飞行与机会路由），基于近端策略优化（PPO）框架。在我们的设计中，首先将问题表述为一个具有顺序运动-路由耦合的分解部分可观察马尔可夫决策过程，并设定每步的团队奖励；随后，去中心化的参与者根据局部观察进行行动，而训练时的评论者使用全局统计数据，此外，选用的多时间段热点预测器提供辅助监督。四种交通模式下的仿真结果表明，与PRoPHET和MaxProp相比，JUROR在保持接触受限的去中心化执行的同时，表现出有效的性能提升。

View on arXiv Download PDF AI Translation

cs.AI / 19 / 2608.04618

Agreement Before Diversity: Verification-First Complementarity for Heterogeneous Language-Model Coordination

在多样性之前达成一致：异构语言模型协调的验证优先互补性

Li, Ruitong, Guo, Binjie, Mo, Aisheng, Su, Guowei, Li, Jie, Zhang, Ru

Abstract

Heterogeneous language-model ensembles expand the space of candidate responses, yet they lack a principled criterion for when a newly generated answer should supersede an already supported one. We decouple candidate headroom from replacement authority, rendering the latter as an explicit, auditable object. Our proposed method, Agreement-Before-Diversity (ABD), is a frozen, label-free decision rule: an anchor answer is retained if two additional trusted samples corroborate it under a fixed equivalence relation; otherwise, it is replaced by a heterogeneous synthesis. For this gating mechanism, we prove two exact identities. The first shows that the accuracy gap relative to unconditional synthesis is determined jointly by the agreement coverage and the anchor's advantage on the protected subset. The second shows that the gap relative to never synthesizing reflects a contrast between authorized recovery and authorized destruction. Neither identity assumes independence or calibrated confidence, and the expected inference cost is approximately eight minus five times the coverage in number of calls. Under blind, exact-ID evaluation, ABD achieves 59.43% on the complete LiveCodeBench-v6 (vs. 52.57% for Single9 and 52.00% for HAC; n = 175) and 75.00% on an untouched GPQA-Diamond split (both controls at 72.78%; n = 180). Furthermore, these identities localize every aggregate difference to an enumerable protected stratum: no discordant items occur among the 3 protected cases on LiveCodeBench, where coverage bounds the gate's contribution to 1.71 points a priori; 13 versus 8 discordant cases among 132 on GPQA-Diamond; and 12 versus 0 among 71 under a frozen anchor perturbation. Diversity supplies potential; verification structure supplies authority.

Chinese Translation

异构语言模型集扩展了候选响应的空间，但缺乏一个原则性的标准来判断何时新生成的答案应当取代已被支持的答案。我们将候选空间的余地与替换权威解耦，将后者视为一个明确的、可审计的对象。我们提出的方法，称为在多样性之前达成一致（Agreement-Before-Diversity, ABD），是一种冻结的、无标签的决策规则：如果两个额外的可信样本在固定的等价关系下证实了一个锚定答案，则保留该答案；否则，由异构合成替换。对于这一门控机制，我们证明了两个精确的恒等式。第一个显示，相对于无条件合成的准确性差距由一致性覆盖和锚定答案在受保护子集上的优势共同决定。第二个显示，相对于从不合成的差距反映了授权恢复与授权破坏之间的对比。两个恒等式均不假设独立性或校准置信度，预期推理成本约为八减去五倍的调用数量覆盖。在盲目、精确ID评估下，ABD在完整的LiveCodeBench-v6上达到了59.43%（单一模型Single9为52.57%，HAC为52.00%；n = 175），在未触及的GPQA-Diamond分割上达到了75.00%（两个对照均为72.78%；n = 180）。此外，这些恒等式将每个聚合差异局限于一个可枚举的受保护层次：在LiveCodeBench的三个受保护案例中没有不一致项，其中覆盖率将门控的贡献限制在1.71点的先验值；在GPQA-Diamond的132个案例中有13个与8个的不一致案例；在71个冻结锚定扰动下有12个与0个的不一致案例。多样性提供了潜力；验证结构提供了权威。

View on arXiv Download PDF AI Translation

cs.AI / 20 / 2608.04625

A/B Agent: A Self-Evolving Agent for Strategy Iteration in Industrial A/B Testing

A/B Agent：一种用于工业 A/B 测试策略迭代的自我进化代理

Jiang, Zhuohang, Chen, Yuxin, Pan, Yongsen, Hu, Zheng, Fan, Wenqi, Li, Qing, Wang, Hongyang, Wang, Jun, Ou, Wenwu

Abstract

Industrial recommendation strategy iteration heavily relies on large-scale A/B experimentation. Traditional tuning requires experts to repeatedly design strategies, configure experiments, analyze results, and adjust parameters, making the process labor-intensive and time-consuming. Meanwhile, valuable knowledge from historical experiments is often fragmented, making systematic reuse difficult through manual expert effort alone. Existing RAG agents partially alleviate this burden by retrieving prior strategies, but typically organize experience in a flat manner, overlooking the hierarchical relationships among business scenarios, recommendation stages, optimization objectives, and experimental contexts. This often results in mismatched retrieval and limited cross-scenario transfer, while preventing agents from continuously refining strategies and parameters through sequential A/B feedback. % To address these limitations, we propose A/B Agent, a closed-loop A/B agent for industrial recommendation strategy optimization. The framework comprises three tightly coupled core components: Historical Strategy Knowledge Organization, Autonomous Target-Aware Strategy Generation, and Experiment-Guided Strategy Self-Evolution. It organizes historical strategies into a hierarchical experience tree, retrieves transferable evidence through multi-path Tree-RAG to generate executable strategies, and continuously analyzes online A/B feedback to guide autonomous tuning and update the experience tree for self-evolution. Extensive offline and online evaluations demonstrate its effectiveness, including a 4.829% improvement in GMV in a real-world short-video e-commerce recommendation system while maintaining positive gains across all guardrail metrics.

Chinese Translation

工业推荐策略的迭代在很大程度上依赖于大规模的 A/B 实验。传统的调优过程需要专家反复设计策略、配置实验、分析结果并调整参数，这使得整个过程劳动密集且耗时。同时，历史实验中积累的宝贵知识往往是碎片化的，仅靠人工专家的努力很难实现系统性的重用。现有的 RAG（Retrieval-Augmented Generation）代理在一定程度上减轻了这一负担，通过检索先前的策略，但通常以平面的方式组织经验，忽视了业务场景、推荐阶段、优化目标和实验上下文之间的层次关系。这常常导致检索不匹配和跨场景转移受限，同时阻碍代理通过连续的 A/B 反馈不断优化策略和参数。为了应对这些局限性，我们提出了 A/B Agent，一种用于工业推荐策略优化的闭环 A/B 代理。该框架由三个紧密耦合的核心组件组成：历史策略知识组织、自动目标感知策略生成和实验引导的策略自我进化。它将历史策略组织成一个层次化的经验树，通过多路径 Tree-RAG 检索可转移的证据以生成可执行策略，并持续分析在线 A/B 反馈以指导自主调优并更新经验树以实现自我进化。大量的离线和在线评估证明了其有效性，包括在一个真实的短视频电商推荐系统中实现了 GMV（总商品交易额）提高 4.829%，同时在所有保护指标上保持了正向收益。

View on arXiv Download PDF AI Translation

cs.AI / 21 / 2608.04641

AI Literacy for Legal Translation: Developing Digital Resilience

法律翻译中的人工智能素养：培养数字韧性

Biel, Łucja

Abstract

Generative AI is transforming legal translation by introducing opportunities alongside linguistic, technical, legal, ethical and cognitive risks. This chapter examines the implications of AI for professional legal translation and proposes an AI literacy framework tailored to the profession. It argues that AI does not change the fundamental objectives of legal translation but requires an extension of professional competence through AI literacy. The proposed framework comprises four mutually reinforcing dimensions, foundational, procedural, critical and strategic, and conceptualises AI literacy as a transversal component of legal translation competence that fosters digital resilience. It further discusses the pedagogical implications of this framework by proposing classroom activities designed to develop AI literacy in legal translator education, enabling future translators to integrate AI critically, responsibly and in accordance with professional standards.

Chinese Translation

生成性人工智能正在通过引入机遇以及语言、技术、法律、伦理和认知风险来改变法律翻译。本章探讨了人工智能对专业法律翻译的影响，并提出了一个针对该职业的人工智能素养框架。文章认为，人工智能并未改变法律翻译的基本目标，但需要通过人工智能素养来扩展专业能力。所提框架包括四个相互促进的维度：基础性、程序性、批判性和战略性，并将人工智能素养概念化为法律翻译能力的一个横向组成部分，以促进数字韧性。进一步讨论了该框架的教学意义，提出了旨在培养法律翻译教育中人工智能素养的课堂活动，使未来的翻译人员能够批判性、负责任地并符合专业标准地整合人工智能。

View on arXiv Download PDF AI Translation

cs.AI / 22 / 2608.04663

Calibrating Artificial Guilt: Neurally Grounded Reward Shaping for Prosocial Multi-Agent Reinforcement Learning

校准人工内疚：基于神经的奖励塑造用于亲社会多智能体强化学习

Mehta, Aaditya, Shah, Arya

Abstract

Cooperative multi-agent reinforcement learning often adds social terms to individual rewards, yet the scale of those terms is usually chosen by hand. We ask whether a guilt signal can instead be calibrated from human neural and behavioural data and transferred to artificial agents. Using the public SoDec responsibility fMRI dataset (40 participants), we fit a subject-fixed-effects regression of momentary-happiness changes on outcome-type counts and recover a guilt weight as the Partner-negative minus Social-negative contrast ($\hat{w}=1.118$, Cohen's $d=0.214$). We embed this weight in a two-agent Social Lottery environment and train independent Proximal Policy Optimization actor-critics under four shaping regimes: neurally calibrated, uniform constant, zero (selfish), and a unit-coefficient oracle. Across 1{,}000 evaluation episodes per condition, the calibrated agents track the human Social safe-choice rate most closely ($0.459$ vs.\ human $0.484$; $\mathrm{KL}=0.0012$), while the other three conditions deviate by one to three orders of magnitude in KL. Human neurobehavioural priors can therefore act as quantitative constraints on prosocial reward shaping.

Chinese Translation

合作多智能体强化学习通常在个体奖励中添加社会项，但这些项的规模通常是手动选择的。我们探讨是否可以从人类神经和行为数据中校准内疚信号，并将其转移到人工智能体上。使用公共的 SoDec 责任功能性磁共振成像（fMRI）数据集（40 名参与者），我们对瞬时幸福感变化与结果类型计数之间的关系进行主体固定效应回归，并恢复出内疚权重，作为合作伙伴负面与社会负面对比的差值（$ ilde{w}=1.118$, Cohen's $d=0.214$）。我们将该权重嵌入到一个双智能体社会彩票环境中，并在四种塑造机制下训练独立的近端策略优化（Proximal Policy Optimization）演员-评论家：神经校准、均匀常数、零（自私）和单位系数神谕。在每种条件下进行 1,000 次评估实验中，校准后的智能体与人类的社会安全选择率最为接近（$0.459$ 对比人类 $0.484$; $ ext{KL}=0.0012$），而其他三种条件的 KL 值偏离一到三个数量级。因此，人类神经行为先验可以作为亲社会奖励塑造的定量约束。

View on arXiv Download PDF AI Translation

cs.AI / 23 / 2608.04697

Traceable LLM-Generated Hazard Scenarios for Operational Safety Analysis of Aviation Systems Using ASRS Reports

基于ASRS报告的可追溯LLM生成的危害场景用于航空系统的操作安全分析

Mascia, Cristian, Pietrantuono, Roberto, Rodriguez, Daniel, Russo, Stefano

Abstract

Operational hazard analysis of aviation system operations must consider interactions among weather, ATC actions, airspace constraints, aircraft operations, and human factors - distinct from the functional hazard assessment applied at the aircraft-system level. We present an AI-assisted approach that generates candidate hazard scenarios from NASA's Aviation Safety Reporting System (ASRS). Given a target adverse outcome, it produces a structured hypothesis as categorical factors and a narrative scenario describing an operational event sequence consistent with the structure. Each scenario includes by a plausibility score from historical co-occurrence evidence and traceability to the most similar held-out ASRS reports. We then propose a hybrid variant, conditioning narrative generation on a structured hypothesis produced via evolutionary abduction, improving correctness and reducing variability. We evaluate multiple large language models, zero-shot versus few-shot prompting, and optional fine-tuning, measuring how prompting and model choice affect the validity and realism of the generated structures and narratives.

Chinese Translation

航空系统操作的危害分析必须考虑天气、空中交通管制（ATC）行动、空域限制、飞行操作和人因等因素之间的相互作用，这与应用于飞机系统级别的功能危害评估有所不同。我们提出了一种AI辅助的方法，从美国国家航空航天局（NASA）的航空安全报告系统（ASRS）中生成候选危害场景。给定一个目标不利结果，该方法生成一个结构化假设，作为分类因素，并提供一个叙述场景，描述与该结构一致的操作事件序列。每个场景都包括一个基于历史共现证据的可信度评分，并与最相似的保留ASRS报告保持可追溯性。随后，我们提出了一种混合变体，通过进化归纳生成结构化假设，从而改善正确性并减少变异性。我们评估了多种大型语言模型，比较零-shot与少-shot提示，以及可选的微调，测量提示和模型选择如何影响生成结构和叙述的有效性和真实性。

View on arXiv Download PDF AI Translation

cs.AI / 24 / 2608.04719

Diagnosing Tool-Selection Reasoning in LLM Agents with Canary Tools

使用金丝雀工具诊断大语言模型代理的工具选择推理

Anand, Atul, Chattaraj, Sourav

Abstract

Agent evaluations tell us that a model picked the wrong tool, but rarely why. We introduce canary tools: diagnostic probe tools planted in an agent's Model Context Protocol (MCP) tool set, each engineered to probe one specific tool-selection weakness. A six-type taxonomy (semantic decoys, parameter traps, capability mirages, prerequisite blindness, temporal decoys, and granularity traps) turns a single "wrong tool" outcome into a multi-dimensional profile of how a model reasons about tools. We evaluate eight models -- six hosted and two 8B open-weight -- spanning three capability tiers, on 120 tasks across three canary-density conditions and three seeds (8,640 runs), plus a 2,880-run subtlety ablation. Task success is graded by a provider-independent judge, corroborated by a second independent judge (Cohen's kappa = 0.75). We report three findings. First, susceptibility drops sharply as models get more capable: the per-task canary susceptibility rate (CSR) ranges about 36x across models, lowest for Claude Opus 4.8 and highest for Llama 3.1 8B. Second, capability tier alone does not predict safety: the most susceptible hosted model is mid-tier, and within a provider the cheaper model can be the safer one. Third, the taxonomy is capability-stratified: capability mirages most reliably trap frontier models, while the other types are largely inert on strong models but fire on small open models, so they discriminate by capability rather than being weak. Softening each canary's give-away phrase leaves frontier CSR essentially unchanged, evidence that the probes measure reasoning, not phrase-spotting. Susceptibility also predicts task failure (Spearman rho = -0.34), while the most robust models are not significantly degraded by canary pressure. We release the framework, canary schemas, tasks, and logs.

Chinese Translation

代理评估告诉我们模型选择了错误的工具，但很少说明原因。我们引入了金丝雀工具：植入代理的模型上下文协议（Model Context Protocol, MCP）工具集中的诊断探测工具，每个工具旨在探测一种特定的工具选择弱点。我们提出了一种六类分类法（语义诱饵、参数陷阱、能力幻影、前提盲点、时间诱饵和粒度陷阱），将单一的“错误工具”结果转化为模型如何推理工具的多维轮廓。我们评估了八个模型——六个托管模型和两个8B开放权重模型，涵盖三个能力层级，在120个任务中，跨越三种金丝雀密度条件和三个种子（共8,640次运行），以及一个2,880次运行的细微消融实验。任务成功率由独立评审员评分，并由第二位独立评审员确认（Cohen's kappa = 0.75）。我们报告了三项发现。首先，随着模型能力的增强，易感性急剧下降：每个任务的金丝雀易感性率（CSR）在模型之间差异约为36倍，Claude Opus 4.8的易感性最低，而Llama 3.1 8B的易感性最高。其次，仅凭能力层级无法预测安全性：最易感的托管模型处于中层，而在同一提供者中，价格较低的模型可能是更安全的选择。第三，分类法是能力分层的：能力幻影最可靠地捕捉前沿模型，而其他类型在强模型上大多无效，但在小型开放模型上有效，因此它们是按能力区分，而不是弱点。软化每个金丝雀的揭示短语几乎不改变前沿CSR，证明这些探测工具测量的是推理，而非短语识别。易感性还预测任务失败（Spearman rho = -0.34），而最强健的模型在金丝雀压力下并未显著退化。我们发布了该框架、金丝雀模式、任务和日志。

View on arXiv Download PDF AI Translation

cs.AI / 25 / 2608.04726

When Prompts Become Pixels: Prompt-Region Grounding for Multimodal Reasoning

当提示变为像素：多模态推理中的提示区域定位

Wang, Yongxin, Zhou, Ruizhe, Tang, Yueling, Zhu, Yingying, Zhao, Xuemin, Chang, Xiaojun, Liang, Xiaodan

Abstract

Multimodal large language models increasingly reason over screenshots and documents where the task itself may be written in pixels. Yet benchmarks usually place questions in text, leaving it unclear whether models use the same instruction equally well across channels. We introduce Visualized Task Semantics (VTS), a controlled intervention that moves the question into the image while keeping the source problem and answer fixed. Across six MLLMs and four benchmarks, accuracy drops in all 24 model-task pairs, by 17.8 points on average. Models often transcribe the visual question correctly yet fail to use it, exposing a semantic channel gap beyond OCR. To reduce this gap, we present prompt-region grounding, whose core design aligns the question region with typed semantics and recovers its clean representation from a masked view. At matched training cost, our method raises four-benchmark VTS accuracy from 58.0 to 66.3 while preserving accuracy on the original interface, and requires no OCR or region metadata at inference. Reading task-bearing text and grounding it as an instruction for reasoning are distinct capabilities.

Chinese Translation

多模态大型语言模型越来越多地在屏幕截图和文档上进行推理，而任务本身可能以像素的形式呈现。然而，基准测试通常将问题置于文本中，这使得模型是否在不同通道上同样有效地使用相同指令变得不明确。我们引入了可视化任务语义（Visualized Task Semantics, VTS），这是一种控制干预措施，将问题移入图像，同时保持源问题和答案不变。在六个多模态大型语言模型和四个基准测试中，所有24个模型-任务对的准确率均有所下降，平均下降17.8个百分点。模型通常能够正确转录视觉问题，但未能有效利用它，暴露出超越光学字符识别（OCR）的语义通道差距。为了缩小这一差距，我们提出了提示区域定位，其核心设计将问题区域与输入语义对齐，并从遮蔽视图中恢复其清晰表示。在匹配的训练成本下，我们的方法将四个基准的VTS准确率从58.0提高到66.3，同时保持原始界面的准确率，并且在推理时不需要OCR或区域元数据。阅读承载任务的文本并将其作为推理指令进行定位是两种不同的能力。

View on arXiv Download PDF AI Translation

cs.AI / 26 / 2608.04735

Chain-of-Thought Monitoring Can Be Unreliable in Implicit-Influence Settings

链式思维监控在隐性影响环境中可能不可靠

Duzan, Agatha, Stickland, Asa Cooper

Abstract

Chain-of-thought (CoT) monitoring is increasingly treated as an important safety layer for frontier reasoning models. Most monitorability evaluations study explicit-influence settings: setups where the prompt directly incentivizes the model to hide something, e.g., by instructing it to perform a hidden side-task. A complementary axis for CoT-monitor evaluations is implicit-influence settings, where the prompt contains no instruction to hide, but the model's behavior is still shaped by features of the task or context, e.g. an irrelevant detail about a candidate that biases a hiring rating. We introduce the first benchmark that directly compares CoT monitorability under the two regimes. We test how model behavior changes in the presence of a nudge to choose a particular option. The nudge is delivered either as a casual aside (implicit), or as a direct instruction to act on the nudge and to conceal having done so (explicit). The benchmark spans four task formats (binary choice, numeric rating, multiple-choice QA, open-ended coding) and seven frontier extended-thinking models. Under explicit influence, a CoT monitor detects 60-94% of behavior shifts: even models instructed to conceal it leak the instruction into their CoT. Under implicit influence, the same factors still shift behavior, but detection falls by 41-46 percentage points in two of our four settings. Realistic system-prompt additions (of the kind a developer might deploy to reduce off-topic bias) lower implicit detection further, to as low as 5%, while preserving the behavioral influence itself. These results suggest that monitorability estimates obtained in explicit-influence settings may over-estimate monitorability, and that monitorability can be further decreased by well-intentioned deployment choices. Our benchmark and code are available at https://github.com/agatha-duzan/implicit-vs-explicit-influence

Chinese Translation

链式思维（CoT）监控越来越被视为前沿推理模型的重要安全层。大多数监控能力评估研究明确影响环境：即提示直接激励模型隐藏某些信息的设置，例如，通过指示模型执行一个隐藏的副任务。CoT监控评估的一个补充维度是隐性影响环境，在这种环境中，提示没有隐藏的指示，但模型的行为仍然受到任务或上下文特征的影响，例如，与候选人相关的无关细节可能会影响招聘评分。我们引入了第一个基准，直接比较这两种环境下的CoT监控能力。我们测试了在选择特定选项时，模型行为的变化。这个提示可以以随意的方式（隐性）或作为直接指示（显性）来施加，要求模型在行动时隐藏这一行为。该基准涵盖了四种任务格式（二元选择、数值评分、多项选择问答、开放式编码）和七个前沿扩展思维模型。在显性影响下，CoT监控能够检测到60-94%的行为变化：即使是被指示隐藏的模型也会在其CoT中泄露该指示。在隐性影响下，相同的因素仍然会改变行为，但在我们四个设置中的两个中，检测率下降了41-46个百分点。现实的系统提示添加（开发者可能会部署以减少偏离主题的偏见）进一步降低了隐性检测率，最低可达5%，同时保留了行为影响。结果表明，在显性影响环境中获得的监控能力估计可能会高估监控能力，而良好的部署选择可能进一步降低监控能力。我们的基准和代码可在 https://github.com/agatha-duzan/implicit-vs-explicit-influence 获取。

View on arXiv Download PDF AI Translation

cs.AI / 27 / 2608.04738

EviGraph: Evidence-Guided Autonomous Research Agents

EviGraph：证据引导的自主研究代理

Ren, Zhenjiang, Li, Ruiji, Zhang, Xujing, Pang, Ziliang, Ren, Shuo, Zhang, Jiajun

Abstract

Autonomous research agents can generate hypotheses, execute experiments, and draft manuscripts, yet their outputs often contain unsupported claims and inconsistencies between research questions, experiments, results, and conclusions. We argue that this problem is partly architectural: existing systems organize research as sequential pipelines but do not explicitly maintain or validate the evolving claim-evidence structure across stages.In this paper, we introduce EviGraph, an autonomous research framework that represents the research process as a typed evidence graph containing Problem, Gap, Hypothesis, Experiment, Finding, and Claim nodes. The graph serves as the operational state of the agent rather than a post-hoc record. EviGraph inspects evidence chains for missing dependencies, semantic misalignment, and result-claim inconsistencies, localizes the earliest weak node, and regenerates its affected downstream subgraph. Graph checkpointing prevents unsuccessful repairs from corrupting previously validated evidence. Manuscripts are generated only after every retained claim is grounded in a validated evidence chain.Experiments on ARC-Bench-ML and NanoResearch-20 show that EviGraph outperforms the compared end-to-end research-agent baselines in overall research performance, improves Claim Support Rate by 40.19% over the strongest baseline, and achieves 87.73% Experimental Data Consistency. These results demonstrate the value of explicit evidence-state maintenance for reliable autonomous research.

Chinese Translation

自主研究代理能够生成假设、执行实验和撰写手稿，但它们的输出往往包含不支持的主张以及研究问题、实验、结果和结论之间的不一致性。我们认为这个问题部分源于架构：现有系统将研究组织为顺序管道，但没有明确维护或验证跨阶段演变的主张-证据结构。在本文中，我们介绍了EviGraph，一个将研究过程表示为包含问题、差距、假设、实验、发现和主张节点的类型化证据图的自主研究框架。该图作为代理的操作状态，而不是事后记录。EviGraph检查证据链中的缺失依赖关系、语义不对齐和结果-主张不一致性，定位最早的弱节点，并重新生成其受影响的下游子图。图的检查点机制防止不成功的修复破坏先前验证的证据。只有在每个保留的主张都基于经过验证的证据链时，才会生成手稿。在ARC-Bench-ML和NanoResearch-20上的实验表明，EviGraph在整体研究表现上优于比较的端到端研究代理基线，将主张支持率提高了40.19%，并实现了87.73%的实验数据一致性。这些结果展示了明确的证据状态维护对于可靠的自主研究的重要性。

View on arXiv Download PDF AI Translation

cs.AI / 28 / 2608.04771

Fewer Tokens, Smaller Cache: Reward-Coordinated Efficient Reasoning

更少的标记，更小的缓存：奖励协调的高效推理

Zhu, Qiyuan, Li, Dezhi, Cheng, Pengyu, Chen, Tianle, Wang, Jiacheng, Shen, Ruijie, Gu, Hao, Lin, Sida, Liu, Zirui, Liu, Jiacheng, Han, Sirui

Abstract

Large Reasoning Models (LRMs) excel on complex tasks through long chain-of-thought (CoT) reasoning, but their lengthy intermediate steps cause severe overthinking that inflates inference cost. KV-cache compression is a common solution, yet existing reasoning-oriented methods apply a uniform policy across the trajectory and judge compression only by what it removes from the cache. Two observations point the other way. First, a reasoning state's tolerance to context loss varies along the trajectory, and process reward tracks it: deleting tokens at high-reward steps preserves accuracy far better than deleting the same budget at random. Second, compression is not free on the generation side, since a smaller cache leads the model to generate more tokens, partly canceling the saving. Together these motivate coordinating both sides under a single process reward. We propose ReCo (Reward-Coordinated Compression), a step-wise framework in which a lightweight process-reward estimator scores each completed step and drives three components: (1) reward-adaptive KV-cache compression that shrinks the retained cache harder at high-reward steps and less at low-reward ones, (2) a reward-banded penalty on reflection tokens that curbs redundant generation, and (3) confidence-based early stopping that triggers when the reasoning is reliable. Across three reasoning models and six benchmarks, ReCo reduces generated tokens by 37%-65% and end-to-end latency by 2.08x-2.35x over Full CoT, all while largely preserving accuracy.

Chinese Translation

大型推理模型（LRMs）在复杂任务中通过长链推理（CoT）表现出色，但其冗长的中间步骤导致严重的过度思考，从而增加推理成本。KV缓存压缩是一种常见的解决方案，但现有的面向推理的方法在整个轨迹上应用统一策略，仅根据从缓存中移除的内容来判断压缩效果。有两个观察结果指向相反的方向。首先，推理状态对上下文丢失的容忍度在轨迹上是变化的，而过程奖励则跟踪这一变化：在高奖励步骤删除标记能比随机删除相同预算的标记更好地保持准确性。其次，生成侧的压缩并非没有代价，因为较小的缓存会导致模型生成更多的标记，部分抵消了节省的效果。这些因素共同促使我们在单一过程奖励下协调两者。我们提出了ReCo（奖励协调压缩），这是一个逐步框架，其中轻量级的过程奖励估计器为每个完成的步骤打分，并驱动三个组件：（1）奖励自适应的KV缓存压缩，在高奖励步骤更大幅度地缩减保留缓存，而在低奖励步骤则减少缩减幅度；（2）对反思标记施加奖励带惩罚，以抑制冗余生成；（3）基于置信度的提前停止，当推理可靠时触发。在三个推理模型和六个基准测试中，ReCo将生成的标记减少了37%-65%，端到端延迟减少了2.08倍至2.35倍，相比于完整的CoT，同时在很大程度上保持了准确性。

View on arXiv Download PDF AI Translation

cs.AI / 29 / 2608.04776

NSF-HRPT: Neural Semantic Field meets Hierarchical Risk Perception Tree for Safety-Critical Scenario Assessment

NSF-HRPT：神经语义场与层次风险感知树相结合的安全关键场景评估

Zhao, Yu, Pan, Jiangyu, Hu, Tao, Yin, Ming, Yang, Fan, Liu, Jiangfan, Liang, Xiubo

Abstract

The ability to accurately assess and anticipate risks in safety-critical scenarios is crucial for autonomous driving systems. While existing research has made progress in collision prediction, accurately quantifying risk levels from monocular vision inputs remains challenging due to the complex dynamics of multi-agent interactions and the inherent uncertainty in real-world environments. To address these challenges, we present NSF-HRPT, a novel framework that combines learning-based perception with structured reasoning for quantitative risk assessment. Our approach features a Neural Semantic Field (NSF) that learns to model scene semantics, trajectory predictions, and probabilistic Time-to-Collision (TTC) distributions from simulation data. During inference, the pre-trained NSF serves as a prior for our Hierarchical Risk Perception Tree (HRPT), which enables efficient parallel computation and spatial reasoning about multi-agent risks. Additionally, we introduce a Sim2Real enhancement strategy that improves real-world applicability without retraining by incorporating priors from foundation models. Extensive evaluations demonstrate that our framework achieves state-of-the-art performance on synthetic benchmarks and delivers competitive, near-state-of-the-art results on real-world datasets for both TTC estimation accuracy and risk localization precision. The proposed method provides an effective solution for real-time risk awareness from monocular camera inputs.

Chinese Translation

在安全关键场景中准确评估和预测风险的能力对自主驾驶系统至关重要。尽管现有研究在碰撞预测方面取得了一定进展，但由于多智能体交互的复杂动态以及现实环境中固有的不确定性，从单目视觉输入中准确量化风险水平仍然具有挑战性。为了解决这些问题，我们提出了NSF-HRPT，一个结合基于学习的感知与结构化推理的定量风险评估新框架。我们的方法采用神经语义场（Neural Semantic Field, NSF），该模型从仿真数据中学习场景语义、轨迹预测和概率时间到碰撞（Time-to-Collision, TTC）分布。在推理过程中，预训练的NSF作为我们层次风险感知树（Hierarchical Risk Perception Tree, HRPT）的先验信息，使得多智能体风险的高效并行计算和空间推理成为可能。此外，我们引入了一种Sim2Real增强策略，通过结合基础模型的先验信息，提升了实际应用的可行性，而无需重新训练。大量评估表明，我们的框架在合成基准测试中实现了最先进的性能，并在真实世界数据集上对于TTC估计准确性和风险定位精度均提供了具有竞争力的接近最先进水平的结果。所提出的方法为从单目摄像头输入实现实时风险感知提供了有效的解决方案。

View on arXiv Download PDF AI Translation

cs.AI / 30 / 2608.04794

Privileged, but Biased: How PI-Conditioned Teachers Break Self-Distillation

特权但有偏见：PI条件教师如何打破自我蒸馏

Harne, Sarthak, Karkar, Chinmay, Pandya, Yash, Awadallah, Ahmed, Nambi, Akshay

Abstract

Self-distillation (SD) has emerged as a compute-efficient alternative to reinforcement learning with verifiable rewards: a self-teacher, conditioned on privileged information (PI) about the answer such as a reference solution, supplies dense per-token supervision to a student that never sees it. Reported gains, however, come almost exclusively from narrow, low-difficulty settings, leaving open a basic question: as a lone objective, with no reward term, does SD teach anything? We reproduce SDPO's reported gains in its easy setting, then apply the identical setup to difficult tasks and find that it does not. Across question answering, mathematics, coding, and multi-turn agentic tool use, across reasoning modes, model sizes, and forms of PI, and under both the SDPO and OPSD recipes, the per-token loss falls steadily while validation accuracy does not improve and typically degrades. We explain this failure through a single causal chain from the loss to the model it produces. The chain begins with PI bias: having seen one particular reference solution, the teacher's per-token target is pulled toward that trajectory rather than toward correctness in general, an effect we quantify with a PI Bias Score. Trained to match this target everywhere, the student's objective becomes nearly blind to whether a rollout is correct, and the loss it assigns falls mostly on low-information tokens like stopwords, punctuation, uncertainty markers, rather than those that determine the answer; within correct rollouts the exploratory tokens incur the highest divergence, so it penalizes the hesitation that reasoning requires. The result is a flatter, less decisive student that is no better at reasoning: as a lone objective, SD optimizes a signal decoupled from task success.

Chinese Translation

自我蒸馏（Self-distillation, SD）已成为一种计算效率高的替代方案，用于强化学习与可验证奖励：一个基于特权信息（Privileged Information, PI）的自我教师，利用关于答案的特权信息（如参考解）为从未见过这些信息的学生提供密集的逐标记监督。然而，报告的收益几乎完全来自于狭窄的、低难度的设置，这引出了一个基本问题：作为唯一目标，没有奖励项，自我蒸馏是否能教会学生任何东西？我们在简单设置中重现了SDPO报告的收益，然后将相同的设置应用于困难任务，发现并没有收益。在问答、数学、编码和多轮自主工具使用等领域，跨越推理模式、模型规模和特权信息形式，在SDPO和OPSD两种方法下，逐标记损失稳步下降，而验证准确率没有改善，通常还会下降。我们通过从损失到模型的单一因果链解释了这一失败。该链条始于PI偏见：在看到特定的参考解后，教师的逐标记目标被拉向该轨迹，而不是一般的正确性，这一效应我们用PI偏见分数（PI Bias Score）进行了量化。学生被训练以在各处匹配这一目标，导致其目标几乎对回归是否正确失去敏感性，分配的损失主要集中在低信息标记上，如停用词、标点符号和不确定性标记，而不是决定答案的标记；在正确的回归中，探索性标记产生的偏差最大，因此它惩罚了推理所需的犹豫。结果是一个更平坦、决策能力更差的学生，其推理能力并没有提升：作为唯一目标，自我蒸馏优化了一个与任务成功无关的信号。

View on arXiv Download PDF AI Translation

cs.AI / 31 / 2608.04830

ContextWeave: A Real-World Workflow Benchmark

ContextWeave：一个现实世界的工作流程基准

Wang, Bo, Yao, Yuqian, Wang, Enxi, Jin, Luozhijie, Liu, Yang, Suo, Yiran, Cai, Yuxuan, Zhou, Enyu, Gao, Yufei, Guo, Honglin, Huai, Tianyu, Ji, Li, Lei, Zhikai, Li, Bufan, Lin, Lizhi, Liu, Jinxiu, Yang, Jie, Zhou, Jiazheng, Zhou, Maosen, Qian, Pengfang, Liu, Shichun, Liu, Guanshan, Zheng, Hao, Yu, Yunhao, Yan, Hang, Kang, Jihua, Chen, Xinchi, Qiu, Xipeng

Abstract

Memory is essential as language agents move from isolated tasks to long-horizon, stateful workflows, yet existing evaluations often reduce it to retrieval or question answering. We introduce ContextWeave, a longitudinal benchmark that evaluates whether recalled experience improves downstream agent performance in realistic office-work streams. ContextWeave reconstructs privacy-preserved, multi-month workflows of 14 participants into 1,005 executable tasks, including 568 core evaluation tasks, with instructions, containerized environments, trajectories, and task-specific rubrics. It measures workspace quality and alignment with participant-specific preferences, complemented by diagnostics of relevance, continuity, solvability, and robustness to misleading recall. Across six memory components under a fixed model, the strongest configuration raises Workspace Score from 68.08 to 78.20 and Preference Score from 41.50 to 70.60. With a fixed memory component, recall improves both outcomes for all five tested base models, although gains vary substantially. Our analysis shows that actionable, experience-rich memory supports workflow continuation and reduces redundant exploration more effectively than compact summaries, while it can also be more susceptible to misleading recall. These findings motivate memory systems that optimize not only retrieval relevance but also reliable use during execution.

Chinese Translation

随着语言智能体从孤立任务转向长期的、有状态的工作流程，记忆变得至关重要。然而，现有的评估往往将其简化为检索或问答。我们引入了ContextWeave，一个纵向基准，评估回忆的经验是否能改善在现实办公工作流中的下游智能体表现。ContextWeave重建了14名参与者的隐私保护的多月工作流程，形成1,005个可执行任务，其中包括568个核心评估任务，配有说明、容器化环境、轨迹和任务特定的评分标准。它衡量工作空间的质量及与参与者特定偏好的对齐程度，并辅以相关性、连续性、可解性和对误导性回忆的鲁棒性等诊断指标。在固定模型下，六个记忆组件中，最强配置将工作空间得分从68.08提升至78.20，偏好得分从41.50提升至70.60。在固定记忆组件的情况下，回忆改善了所有五个测试基础模型的这两个结果，尽管增益差异显著。我们的分析表明，具有可操作性和丰富经验的记忆比简洁摘要更有效地支持工作流程的持续性并减少冗余探索，同时也可能更容易受到误导性回忆的影响。这些发现激励我们开发不仅优化检索相关性，还能在执行过程中可靠使用的记忆系统。

View on arXiv Download PDF AI Translation

cs.AI / 32 / 2608.04896

When Shared Rollouts Fail in Defensive Driving Evaluation: A NAVSIM Score Basis Audit

当共享回滚在防御性驾驶评估中失败时：基于NAVSIM评分的审计

Wei, Ziang, Yu, Minjun, Lai, Zheyuan, Pang, Mingjie, Li, Wei

Abstract

Defensive driving scores are useful only when they preserve distinctions between policies that observe surrounding actors and those that do not. Re-simulation benchmarks may use reference-conditioned forgiveness, under which an agent receives credit when the logged human reference fails a compliance channel. When agent and reference share an unstable rollout transformation, this rule can propagate shared reference failures into broad compliance credit. We audit this risk in NAVSIM v2.2 original scene single-stage scoring. Under the affected documented-stack condition on the audited numerical backend, the route-blind Ignore-All probe and a route-aware actor-blind probe outrank human replay and PDM-Closed over the complete 12,146-token navtest split. A fresh installation following the public specification reproduces rollout divergence on a fixed 32-token diagnostic set. A same-source dependency stack control and an exact-input diagnostic isolate dependency-sensitive numerical behavior in the shared velocity refit. On a 450-token control pool, replacing only the solver eliminates rollout divergence and restores blind-last ordering while keeping forgiveness enabled. Thus, the numerical instability is the direct trigger. Reference-conditioned forgiveness propagates the resulting shared reference failures into compliance credit. We contribute an audit protocol requiring score basis and stack disclosure, blind probes, overwrite reporting, and rollout stability tests before using such scores for defensive driving claims.

Chinese Translation

防御性驾驶评分只有在能够区分观察周围行为者的策略与不观察的策略时才有用。重新模拟基准可能使用参考条件宽恕，在这种情况下，当记录的人类参考未能遵守合规通道时，代理会获得信用。当代理和参考共享一个不稳定的回滚变换时，这一规则可能将共享的参考失败传播到广泛的合规信用中。我们在NAVSIM v2.2原始场景单阶段评分中审计了这一风险。在审计的数值后端的受影响文档堆栈条件下，路线盲的Ignore-All探测器和路线感知的行为盲探测器在完整的12,146标记navtest拆分中超越了人类重放和PDM-Closed。在遵循公共规范的新安装中，在固定的32标记诊断集上重现了回滚发散。同源依赖堆栈控制和精确输入诊断隔离了共享速度重调中的依赖敏感数值行为。在一个450标记的控制池中，仅替换求解器就消除了回滚发散，并在保持宽恕启用的同时恢复了盲最后排序。因此，数值不稳定性是直接触发因素。参考条件宽恕将由此产生的共享参考失败传播到合规信用中。我们贡献了一种审计协议，要求在使用此类评分进行防御性驾驶声明之前，披露评分基础和堆栈、盲探测、覆盖报告和回滚稳定性测试。

View on arXiv Download PDF AI Translation

cs.AI / 33 / 2608.04964

WorldCycle: Self-Verifiable Reinforcement Learning for Long-Horizon Video World Models

WorldCycle：用于长时间跨度视频世界模型的自验证强化学习

Gu, Bohai, Yuan, Yueyang, Wu, Taiyi, Du, Dazhao, Liu, Jian, Pang, Xiaoyi, Zhang, Jie, Lu, Xiaocheng, Zhong, Haobin, Zhao, Xiaotong, Zhao, Alan, Guo, Song

Abstract

Interactive video world models are essential for long-horizon planning and exploration, yet they suffer from compounding errors. Post-training methods such as reinforcement learning (RL) can improve these models, but they hit a verification bottleneck: for arbitrary action sequences, no ground-truth future state exists to measure long-term drift. Our key insight is that reversible action cycles make this verification possible: a sequence composed with its inverse must analytically return to the initial state, yielding annotation-free supervision on long-horizon correctness. Building on this, we introduce WorldCycle, a self-verifiable RL framework that constructs closed action cycles and their repeated executions from ordinary action sequences, and optimizes two complementary rewards: a spatial closure reward enforcing symmetry between mirrored forward and reverse segments, and a temporal consistency reward aligning states across repeated cycle executions. These rewards force the model to learn actions as consistent state operators rather than memorized temporal patterns, and extend naturally to out-of-distribution composite action cycles that the base model handles poorly. We further release CycleBench, a diagnostic benchmark for state-returning ability under complex action structures. WorldCycle reduces state returning drift by up to 44% and lifts composite-action accuracy nearly 4x over the base model, providing a vital foundation for physically grounded world models.

Chinese Translation

交互式视频世界模型对于长时间跨度的规划和探索至关重要，但它们面临累积误差的问题。后训练方法如强化学习（RL）可以改善这些模型，但它们遇到了验证瓶颈：对于任意的动作序列，没有真实的未来状态来衡量长期漂移。我们的关键见解是可逆动作循环使得这种验证成为可能：一个由其逆序列构成的序列必须在分析上返回到初始状态，从而在长时间跨度的正确性上提供无注释的监督。在此基础上，我们提出了WorldCycle，一个自验证的RL框架，它从普通动作序列构建闭合动作循环及其重复执行，并优化两个互补的奖励：一个空间闭合奖励，强制镜像前向和反向段之间的对称性，以及一个时间一致性奖励，确保在重复循环执行中状态的一致性。这些奖励迫使模型将动作学习为一致的状态操作符，而不是记忆的时间模式，并自然扩展到基础模型处理不佳的分布外复合动作循环。我们进一步发布了CycleBench，一个用于在复杂动作结构下评估状态返回能力的诊断基准。WorldCycle将状态返回漂移减少了多达44%，并将复合动作的准确性提高了近4倍，为物理基础的世界模型提供了重要基础。

View on arXiv Download PDF AI Translation

cs.AI / 34 / 2608.05018

Short-term load forecasting under EU-AI Act Requirements in Safety-Critical Environments: Results from a 41-day live challenge on the aggregated German transmission-grid load

在安全关键环境下符合欧盟人工智能法要求的短期负荷预测：基于41天实时挑战的德国传输网负荷聚合结果

Bartz-Beielstein, Thomas

Abstract

Short-term load forecasting (STLF) play a vital role in the electric power industry. It serves infrastructure that European and German law designate as critical. Determinism, reproducibility, and auditability are engineering requirements rather than optional extras. STLF is no longer purely an accuracy problem. It is also a software-engineering and compliance problem. This paper describes results from a 41-day live challenge that evaluated a complete STLF pipeline for the aggregated German transmission-grid load. The pipeline is based on the open-source Python library spotforecast2-safe, which implements the EU-AI Act Requirements in Safety-Critical Environments by design. The pipeline predicts the 24 hourly load values of a target day from European Network of Transmission System Operators for Electricity (ENTSO-E) data. It includes anomaly detection and gap-aware data preparation, calendar and weather covariates, a recursive multi-step forecasting algorithm, and hyperparameter tuning. Forecast accuracy is measured against the official ENTSO-E day-ahead forecast. The EU-AI act compliant spotforecast2-safe pipeline beats the ENTSO-E baseline. In-context models show competitive performance. Transparent, low-cost, and auditable local models (referred to as macl2l in this paper) are competitive with more than 100-million-parameter large, energy-intensive pre-trained foundation models such as chronos-2. The challenge infrastructure, the complete submission history of all teams, and the frozen final leaderboard are publicly available.

Chinese Translation

短期负荷预测（STLF）在电力行业中扮演着至关重要的角色。它服务于欧洲和德国法律指定为关键的基础设施。确定性、可重复性和可审计性是工程要求，而非可选附加项。STLF不再仅仅是一个准确性问题，它也是一个软件工程和合规性问题。本文描述了一个为期41天的实时挑战的结果，该挑战评估了针对德国传输网负荷聚合的完整STLF流程。该流程基于开源Python库spotforecast2-safe，该库在设计上实现了安全关键环境下的欧盟人工智能法要求。该流程根据欧洲电力传输系统运营商网络（ENTSO-E）数据预测目标日的24小时负荷值。它包括异常检测和考虑缺口的数据准备、日历和天气协变量、递归多步预测算法以及超参数调整。预测准确性是与官方ENTSO-E日前预测进行比较的。符合欧盟人工智能法的spotforecast2-safe流程超越了ENTSO-E基线。在上下文模型中表现出竞争力。透明、低成本且可审计的本地模型（在本文中称为macl2l）与超过1亿参数的大型、能源密集型预训练基础模型（如chronos-2）具有竞争力。挑战基础设施、所有团队的完整提交历史以及冻结的最终排行榜均已公开。

View on arXiv Download PDF AI Translation

cs.AI / 35 / 2608.05030

From Score Matrices to Football-Aware Match-State Simulation: An Auditable LLM Harness for Exact-Score Reranking

从得分矩阵到足球意识的比赛状态模拟：一种可审计的 LLM 工具用于精确得分重排序

Liang, Shaopeng

Abstract

Football score forecasting combines a strong statistical core with a difficult contextual edge. Dynamic Poisson-family models estimate team strength, expected goals, and coherent score probabilities, but do not directly understand roles, tactical matchups, motivation, or how a first goal changes behaviour. Large language models (LLMs) can reason about such concepts, yet are not calibrated probability engines. We combine both components through an auditable information harness. This paper documents four iterations: V1, a dynamic score-driven Dixon-Coles baseline; V2, which maps LLM contextual ratings back into expected-goal parameters; V3, which replaces scalar correction with goal-by-goal simulations over a frozen score-candidate set; and V4, which adds shared first-breakthrough and post-goal cascade judgments, time-aware stopping, and deterministic tail candidates. The harness defines input semantics, supplies pre-match evidence, and constrains the LLM to an inspectable reasoning route. On a chronological replay of the first 150 matches of the 2025-26 English Premier League, V1 achieved 10.0% Top-1 and 26.7% Top-3 exact-score accuracy. V3 reached 12.0% and 30.0%, while V4 reached 14.7% and 30.7%. V4 increased candidate coverage from 77.3% to 84.7%, although no added tail candidate became a Top-3 exact hit. V1's native 1X2 distribution achieved 53.3% argmax accuracy, 0.9878 log loss, 0.5870 Brier score, and 0.2095 ranked probability score. These results are exploratory: the development slice is not an untouched benchmark, and temporal input isolation cannot exclude outcome memory in a closed LLM. The contribution is an auditable hybrid architecture, a clear design evolution, and negative findings showing where football-aware simulation does and does not improve score selection.

Chinese Translation

足球得分预测结合了强大的统计基础和复杂的背景因素。动态泊松族模型估计球队实力、预期进球和一致的得分概率，但并不直接理解角色、战术对抗、动机或首个进球如何改变行为。大型语言模型（LLMs）能够推理这些概念，但并不是经过校准的概率引擎。我们通过可审计的信息工具结合了这两种组件。本文记录了四个迭代版本：V1，一个基于动态得分的 Dixon-Coles 基线；V2，将 LLM 的上下文评分映射回预期进球参数；V3，用逐球模拟替代了标量修正，基于冻结的得分候选集；V4，增加了共享的首次突破和进球后的级联判断、时间感知的停止机制以及确定性的尾部候选。该工具定义了输入语义，提供赛前证据，并将 LLM 限制在可检查的推理路径上。在对 2025-26 赛季英超联赛前 150 场比赛的时间顺序重播中，V1 实现了 10.0% 的 Top-1 和 26.7% 的 Top-3 精确得分准确率。V3 达到了 12.0% 和 30.0%，而 V4 则达到了 14.7% 和 30.7%。V4 将候选覆盖率从 77.3% 提高到 84.7%，尽管没有新增的尾部候选成为 Top-3 精确命中。V1 的原生 1X2 分布实现了 53.3% 的 argmax 准确率，0.9878 的对数损失，0.5870 的 Brier 分数和 0.2095 的排名概率分数。这些结果是探索性的：开发切片并不是一个未被触碰的基准，时间输入隔离无法排除闭合 LLM 中的结果记忆。贡献在于一个可审计的混合架构、清晰的设计演进，以及负面发现，显示足球意识模拟在得分选择上改善的地方和不改善的地方。

View on arXiv Download PDF AI Translation

cs.AI / 36 / 2608.05086

Item Response Theory for AI Safety

人工智能安全的项目反应理论

Rivera, Joshua Fonseca, Shah, Neil, Africa, David Demitri, Voudouris, Konstantinos

Abstract

Language models differ in how safely they behave and these differences are measured by safety benchmarks. But aggregated benchmark scores are hard to trust and interpret, because benchmarks duplicate one another, correlate heavily, and models may sandbag when they detect evaluation. To address these issues, we draw on Item Response Theory (IRT), a statistical toolkit for measuring these latents from performance on items with inferred psychometric properties. We fit IRT models to eight safety benchmarks across 192 language models, the largest psychometric analysis of LLM safety evaluations to date, and contribute three results. First, we find that three interpretable factors of refusal strictness, truthfulness, and contextual harm explain most of the variance between models across benchmarks. Second, psychometrically selected items recover full benchmark scores with lower error than random subsets of the same size, and roughly ten adaptively chosen items suffice for several individual benchmarks, cutting evaluation cost by 97-99%. Third, IRT supports audits of individual models, showing that it can be used to detect naive sandbagging and changes of model behind APIs. Overall, we show IRT is a ready-made toolkit for reading, reducing, and auditing safety benchmarks, which we recommend frontier labs and evaluators adopt.

Chinese Translation

语言模型在安全行为上存在差异，这些差异通过安全基准进行测量。然而，聚合的基准分数难以信任和解释，因为基准之间存在重复，相关性很高，模型在检测到评估时可能会故意降低表现。为了解决这些问题，我们借鉴了项目反应理论（Item Response Theory, IRT），这是一种用于通过具有推断心理测量特性的项目表现来测量潜在变量的统计工具。我们将IRT模型应用于192个语言模型的八个安全基准，这是迄今为止对大型语言模型（LLM）安全评估的最大心理测量分析，并得出了三个结果。首先，我们发现拒绝严格性、真实性和上下文伤害这三个可解释因素解释了模型在基准之间大部分的方差。其次，心理测量选择的项目以低于相同大小随机子集的误差恢复完整的基准分数，且大约十个自适应选择的项目足以满足多个单独基准的需求，从而将评估成本降低了97-99%。第三，IRT支持对单个模型的审计，表明它可以用于检测简单的故意降低表现和API背后模型的变化。总体而言，我们展示了IRT是一个现成的工具包，用于解读、简化和审计安全基准，我们建议前沿实验室和评估者采用该方法。

View on arXiv Download PDF AI Translation

cs.AI / 37 / 2608.05095

Hierarchical Graph Memory for LLM Agents with Path-level Localization and Rewrite

具有路径级定位和重写的层次图记忆用于大型语言模型代理

Yue, Xiawei, Wang, Boran, Zhang, Xiaoqing, Zheng, Shuxin, Zhang, Ziwei

Abstract

Agents for long term reasoning require a memory that can be efficiently and effectively updated over time, as new facts and external feedback continue to arrive. Recently, graph memory has been adopted to offer structural organization for multi-hop retrieval and reasoning. However, existing methods store all memories in a flat graph, and accumulated historical memories can introduce irrelevant contexts and increase the cost of evidence selection during retrieval. Moreover, they typically update memory units independently, requiring repeated unit-wise rewrite to cover related changes. To address these issues, we propose HiGram, an evolving hierarchical graph memory framework with path-level localization and rewriting. Specifically, we first propose a hierarchical graph memory, which organizes the memory into coarse-to-fine architecture composed of upper-level nodes and MemoryUnits, thereby reducing the amount of irrelevant information during retrieval. We further propose MicroGraph-based path-level localization, which leverages query and update conditioned MicroGraphs to identify support subgraph and evidence path before rewrite. Finally, we propose a coordinated rewriting method that jointly revises intra-unit memory and inter-unit dependencies, enable valid dependency structures updating in the localized evidence path. Experiments on benchmarks for long-term conversational question answering and conflict-aware memory evaluation demonstrate that our method demonstrate substantial improvements over baselines in answer quality and token efficiency. Besides, our method improves answer accuracy and query-valid evidence selection under dynamic, static, and conditional conflicts.

Chinese Translation

长期推理的代理需要一种能够随着新事实和外部反馈的不断到来而高效且有效更新的记忆。最近，图记忆被采用以提供多跳检索和推理的结构化组织。然而，现有方法将所有记忆存储在一个平面图中，累积的历史记忆可能引入无关的上下文，并增加检索过程中证据选择的成本。此外，它们通常独立更新记忆单元，需要重复逐单元重写以覆盖相关的变化。为了解决这些问题，我们提出了HiGram，一个具有路径级定位和重写的演变层次图记忆框架。具体而言，我们首先提出了一种层次图记忆，将记忆组织成由上层节点和记忆单元（Memory Units）组成的粗到细架构，从而减少检索过程中无关信息的数量。我们进一步提出了基于微图（MicroGraph）的路径级定位，利用条件查询和更新的微图来识别支持子图和证据路径，然后进行重写。最后，我们提出了一种协调重写方法，联合修订单元内记忆和单元间依赖关系，使得在局部证据路径中有效的依赖结构更新成为可能。在长期对话问答和冲突感知记忆评估的基准实验中，我们的方法在答案质量和令牌效率上相较于基线表现出显著改善。此外，我们的方法在动态、静态和条件冲突下提高了答案准确性和查询有效证据的选择。

View on arXiv Download PDF AI Translation

cs.AI / 38 / 2608.05102

ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment

ABSeeker：通过答案回溯信用分配训练长时间搜索代理

Lu, Yijun, Ye, Rui, Wang, Jiajun, Du, Yuwen, Jin, Tian, Liu, Songhua, Chen, Siheng

Abstract

Long-horizon search agents must make multiple sequential actions (steps) to search, retrieve, verify, and integrate evidence to reach a final answer. However, existing methods for training these agents typically treat all steps within a trajectory uniformly during both supervised fine-tuning (SFT) and reinforcement learning (RL), failing to distinguish useful actions from erroneous or redundant ones. In this paper, we propose Answer-Backtracked Credit Assignment (ABC), a fine-grained credit assignment framework for training long-horizon search agents by converting sparse trajectory-level outcomes into dense step-level supervision that rewards useful actions (even in failed trajectories) while suppressing erroneous or redundant actions. Specifically, given a potentially obscure query and its corresponding ground-truth answer, ABC first performs Answer-Backtracked Clue Recovery, which traces back from the answer to recover intermediate clues required to solve the question. It then applies Clue-Anchored Step Scoring to evaluate each search step against these clues, converting sparse binary outcome supervision into dense step-level rewards. Based on these rewards, we develop ABC-SFT, which reweights the loss of each turn, and ABC-GRPO, which uses the step-level scores as rewards in GRPO. Building on this framework, we train ABSeeker based on Qwen3.5-4B with only 8.5k examples. ABSeeker achieves 37.3% on BrowseComp and 39.1% on BrowseComp-ZH. With context management, the scores further improve to 55.3% and 52.9%, respectively, significantly outperforming same-scale (4B) agents and even matching the performance of larger ones (approximately 30B). These results demonstrate the effectiveness of answer-backtracked step-level credit assignment for training long-horizon search agents.

Chinese Translation

长时间搜索代理必须进行多个连续的动作（步骤）以搜索、检索、验证和整合证据，以达到最终答案。然而，现有的训练这些代理的方法通常在监督微调（SFT）和强化学习（RL）过程中对轨迹中的所有步骤进行统一处理，未能区分有用的动作与错误或冗余的动作。本文提出了一种答案回溯信用分配（ABC）框架，通过将稀疏的轨迹级结果转换为密集的步骤级监督，来训练长时间搜索代理，从而奖励有用的动作（即使在失败的轨迹中），同时抑制错误或冗余的动作。具体而言，给定一个可能模糊的查询及其对应的真实答案，ABC首先执行答案回溯线索恢复，追溯答案以恢复解决问题所需的中间线索。然后，它应用线索锚定步骤评分来评估每个搜索步骤与这些线索的匹配程度，将稀疏的二元结果监督转换为密集的步骤级奖励。基于这些奖励，我们开发了ABC-SFT，它重新加权每个回合的损失，以及ABC-GRPO，它在GRPO中使用步骤级评分作为奖励。在此框架的基础上，我们基于Qwen3.5-4B训练了ABSeeker，仅使用了8.5k个示例。ABSeeker在BrowseComp上达到了37.3%，在BrowseComp-ZH上达到了39.1%。通过上下文管理，这些分数进一步提高至55.3%和52.9%，显著超越了同规模（4B）代理，甚至与更大规模的代理（约30B）相匹配。这些结果证明了答案回溯步骤级信用分配在训练长时间搜索代理中的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 39 / 2608.05107

CoPlan: A Trustworthy Co-Intelligence Interface for Care Planning through Role-Based Contestable Argument Graphs

CoPlan：基于角色的可争议论证图的可信共智能护理规划接口

Nguyen, Hung Truong Thanh, Fournier, Hélène, Jackson, Piper, Itoh, Makoto, Freeman, Shannon, Richard, Rene, Cao, Hung

Abstract

AI-supported care planning can help clinicians, patients, caregivers, and care teams coordinate complex decisions across clinical, functional, psychosocial, and environmental needs. However, many AI systems present recommendations as fixed outputs, limiting stakeholders' ability to inspect, challenge, and revise plans when they conflict with clinical judgment, patient values, or real-world feasibility. We present CoPlan - a Co-Intelligent and Contestable Interface for Human-AI Care Planning. CoPlan uses a multi-agent workflow in which specialized AI agents generate candidate interventions and supporting or challenging arguments, while human care planners can accept, reject, modify, or add arguments before final plan generation. Through this design, CoPlan combines co-intelligence, in which humans and AI agents contribute complementary expertise, with contestability, where recommendations remain open to inspection, revision, and justification. We demonstrate CoPlan in an aging-in-place care planning scenario. The system supports adaptive care team recruitment, role-based argument review, final care plan generation, and practical follow-up through scheduling agents. This work contributes a contestable care planning interface and a design framing for trustworthy human-AI care planning that preserves human agency and clinical accountability.

Chinese Translation

AI支持的护理规划可以帮助临床医生、患者、护理人员和护理团队协调临床、功能、心理社会和环境需求等复杂决策。然而，许多AI系统将推荐结果呈现为固定输出，限制了利益相关者在与临床判断、患者价值观或现实可行性发生冲突时检查、质疑和修订计划的能力。我们提出了CoPlan——一个用于人机护理规划的共智能和可争议接口。CoPlan采用多智能体工作流程，其中专业的AI智能体生成候选干预措施及支持或反对的论证，而人类护理规划者可以在最终生成计划之前接受、拒绝、修改或添加论证。通过这种设计，CoPlan将共智能（人类与AI智能体贡献互补专业知识）与可争议性（推荐结果保持开放以供检查、修订和证明）结合在一起。我们在一个居家养老护理规划场景中演示了CoPlan。该系统支持适应性护理团队招募、基于角色的论证审查、最终护理计划生成以及通过调度智能体进行的实际跟进。这项工作贡献了一个可争议的护理规划接口和一个可信的人机护理规划设计框架，保留了人类的自主性和临床责任。

View on arXiv Download PDF AI Translation

cs.AI / 40 / 2608.05141

OctoLong: Mid-Training On Cross-Repository Code Contexts Enhances Long-Context Modeling

OctoLong：跨仓库代码上下文中的中期训练增强了长上下文建模

Paul, Indraneil, Helm, Falko, Glavaš, Goran, Gurevych, Iryna

Abstract

Context lengths of language models (LMs) have dramatically increased, driven by the demands for in-context learning, self-improvement, and long-horizon agentic workflows. Existing long-context corpora, however, are dominated by books, academic articles, and code repositories, which are finite resources and often scarce in long-distance dependencies. In this work, we introduce OctoLong, a context engineering pipeline that instruments an AST parser, a language server backend, and a package manager to facilitate the recursive retrieval of code references, enabling the curation of dependency-rich code contexts of millions of tokens in length. We then train OctoLong-Instruct, a suite of capable long-context open LMs, derived from base models ranging in size from 600M to 14B parameters, via context-extension mid-training on a ~50B-token mixture containing ~6.2B tokens of OctoLong code contexts, followed by ~10B tokens of instruction tuning. Our training ablations and experimental evaluations against 18 state-of-the-art open-weight long-context LMs show that supplanting just 12% of traditional context-extension corpora with OctoLong data yields substantial gains in long-range retrieval, long-term state tracking, repository-level code understanding, and downstream agentic tasks, while also enhancing API usage in short-context coding scenarios.

Chinese Translation

语言模型（LM）的上下文长度显著增加，这一变化源于对上下文学习、自我改进和长远代理工作流程的需求。然而，现有的长上下文语料库主要由书籍、学术文章和代码仓库构成，这些资源是有限的，并且在长距离依赖方面往往稀缺。在本研究中，我们介绍了OctoLong，这是一种上下文工程管道，结合了AST解析器、语言服务器后端和包管理器，以促进代码引用的递归检索，从而实现长达数百万个标记的富依赖代码上下文的策划。随后，我们训练了OctoLong-Instruct，这是一套强大的长上下文开放语言模型，基于从600M到14B参数的基础模型，通过在包含约50B标记的混合数据上进行上下文扩展的中期训练，其中约6.2B标记来自OctoLong代码上下文，随后进行了约10B标记的指令调优。我们的训练消融实验和针对18个最先进的开放权重长上下文LM的实验评估表明，仅用OctoLong数据替代传统上下文扩展语料库的12%即可在长距离检索、长期状态跟踪、仓库级代码理解和下游代理任务中获得显著提升，同时也增强了短上下文编码场景中的API使用。

View on arXiv Download PDF AI Translation

cs.AI / 41 / 2608.05144

Argus: A General-Purpose Agentic Runtime for Long-Horizon Reasoning

Argus：一种用于长时间推理的通用代理运行时

Li, Boxiu, Wen, Zimo, Fan, Yijia, Lei, Junxiang, Guo, Sufeng, Wu, Jiaao, Tang, Ruize, Li, Mukai, Shen, Yifei, Chen, Xiaoyu, Zhang, Wanbo, Gu, Runjing, Gao, Yifei, Wu, Yuheng, Huang, Xuyao, Zhao, Zelong, Zhang, Jiachen, Hu, Shibo, Guo, Hangxi, Chen, Yilin, Zhang, Yuzhe, Yang, Fan, Wen, Chuan, Zhang, Xian, Zhou, Xuanhe, Deng, Zhijie

Abstract

Long-horizon reasoning requires an agentic runtime that can persist when evidence supports its current approach and pivot when measurements reveal failure, hidden constraints, or a misspecified objective. We present Argus, a persistent, self-evolving runtime in which Manager, Planner, Engineer, and Reviewer execute bounded missions over durable project state. Argus separates stable user intent from operational objectives, constraints, and verification criteria, and admits memories, skills, procedures, verifiers, routing decisions, and rejected routes only after role-owned review and, when available, task-native verification. Model weights remain fixed; self-evolution occurs through persistent runtime state and control policy, with autonomous execution between operator-owned escalation points. Across seven GPT-5.5 benchmark arenas, Argus achieves about 78% on SWE-Bench Pro versus 59% for Direct Copilot while using 1.41 times the aggregate tokens. After verification-gated self-evolution, mature SWE-Bench waves use 21% fewer solve-input tokens and 15% less active workflow time per task than startup waves, while recording 34 verifier recoveries and 22 strict review-loop rescues. Argus also reaches 76.8% on AARRI-Bench and a 28.0-point gap on mathematical data synthesis, with competitive GPU-kernel and language-model-training results. Beyond benchmarks, an optimized RWKV6 kernel was merged upstream; a multi-day mathematics campaign retained falsified routes and proof-backed frontier updates; and six paper pipelines completed 254 missions with 16 stage rollbacks. These results show that a fixed-weight, self-evolving harness can revise, recover, and accumulate verified approaches while producing structured trajectories for future supervised and reinforcement learning.

Chinese Translation

长时间推理需要一种代理运行时，它能够在证据支持当前方法时持续运行，并在测量结果显示失败、隐藏约束或目标设定错误时进行调整。我们提出了Argus，这是一种持久的自我演化运行时，其中管理者、规划者、工程师和审查者在持久的项目状态上执行有限的任务。Argus将稳定的用户意图与操作目标、约束和验证标准分开，并在角色拥有的审查后，仅在可用时接受记忆、技能、程序、验证者、路由决策和被拒绝的路线。模型权重保持不变；自我演化通过持久的运行时状态和控制策略发生，在操作员拥有的升级点之间实现自主执行。在七个GPT-5.5基准场景中，Argus在SWE-Bench Pro上的得分约为78%，而Direct Copilot为59%，同时使用的总令牌数为1.41倍。在经过验证门控的自我演化后，成熟的SWE-Bench波段每个任务使用的求解输入令牌比初创波段少21%，活跃工作流时间减少15%，同时记录了34次验证者恢复和22次严格审查循环救援。Argus在AARRI-Bench上也达到了76.8%的得分，并在数学数据合成上有28.0的差距，展现出具有竞争力的GPU内核和语言模型训练结果。除了基准测试外，一个优化的RWKV6内核已合并到上游；一个持续数天的数学活动保留了伪造的路线和基于证明的前沿更新；六个论文管道完成了254个任务，并进行了16次阶段回滚。这些结果表明，固定权重的自我演化框架能够修订、恢复和积累经过验证的方法，同时为未来的监督学习和强化学习生成结构化轨迹。

View on arXiv Download PDF AI Translation

计算语言学 (Computation and Language)

cs.CL / 1 / 2608.04015

Transfer Learning for Named Entity Recognition of Classical Latin through LLM Prompting

通过大语言模型提示进行古典拉丁语命名实体识别的迁移学习

Chan, Callum

Abstract

With the increase in digitized resources of Classical Latin texts and modern breakthroughs of Large Language Models (LLMs), I contribute to ancient language research by participating in EvaLatin 2026. This paper describes Team uOttawa's system description and results for the Named Entity Recognition (NER) shared task. The task is divided into two subtasks: coarse-grained NER with 11 classes and fine-grained NER with 28 classes, each evaluated under strict and fuzzy regimes. Through prompt engineering of commercial LLMs gemini-2.5-pro and claude-sonnet-4-5, I show that the underrepresented ancient Latin language can take advantage of cross-lingual transfer learning by using advancements made by the wider LLM development community. Overall, the methods discussed in this report demonstrate very strong results, placing first in both NER subtasks and achieving the best scores across all evaluation metrics and regimes among all submissions.

Chinese Translation

随着古典拉丁文本数字资源的增加以及大语言模型（LLMs）的现代突破，我通过参与EvaLatin 2026为古代语言研究做出了贡献。本文描述了uOttawa团队在命名实体识别（NER）共享任务中的系统描述和结果。该任务分为两个子任务：粗粒度NER（11个类别）和细粒度NER（28个类别），每个子任务在严格和模糊两种评估模式下进行评估。通过对商业大语言模型gemini-2.5-pro和claude-sonnet-4-5的提示工程，我展示了被低估的古拉丁语如何利用跨语言迁移学习，借助更广泛的大语言模型开发社区所取得的进展。总体而言，本文讨论的方法表现出非常强的结果，在两个NER子任务中均名列第一，并在所有提交中在所有评估指标和模式下取得最佳分数。

View on arXiv Download PDF AI Translation

cs.CL / 2 / 2608.04021

When More Becomes Less: Position-Dependent Repetition Effects in Language Models

当更多变成更少：语言模型中的位置依赖重复效应

Wang, Han-yu

Abstract

Cloze-style probes that vary how often a target token appears implicitly assume that more copies of a target affect prediction the same way regardless of where the readout slot sits. We show this assumption fails. Our two-probe design holds a repeated-target prefix fixed and varies only the readout position: the adjacent probe places the slot immediately after the repeated block; the displaced probe places it inside a fresh sentence frame. Adjacent repetition behaves as priming intuition predicts: $P(\text{target})$ climbs with $N$ and plateaus. Displaced repetition produces an inverted-U: $P(\text{target})$ rises to an early peak and then declines as more copies are added. The displaced inverted-U shows a per-word drop with bootstrap CI excluding zero in all 13 open-access encoder and decoder models we test, and replicates across Spanish, Chinese, German, and French in 42 of 42 multilingual cells. A six-condition causal ablation isolates the effect to exact lexical repetition rather than length, generic redundancy, or semantic-neighbour exposure. A frame-pragmatics control rules out an artefact of the readout frame. Internally, per-target-token attention falls with $N$ while the total budget assigned to the repeated block grows in causal LMs but not in the masked LM we probe. Probes that vary repetition count cannot treat the readout position as orthogonal to what they measure.

Chinese Translation

克洛兹（Cloze）风格的探测器在变化目标标记出现频率时，隐含假设更多的目标副本对预测的影响与读取位置无关。我们展示了这一假设的失败。我们的双探测设计固定了重复目标的前缀，仅变化读取位置：相邻探测器将插槽放置在重复块之后；而位移探测器则将其放置在一个新的句子框架内。相邻重复的行为符合启动（priming）直觉的预测：$P( ext{target})$ 随着 $N$ 的增加而上升并达到平台期。位移重复则产生一个倒U型：$P( ext{target})$ 先上升到一个早期峰值，然后随着更多副本的添加而下降。位移的倒U型在我们测试的所有13个开放访问的编码器和解码器模型中显示出每个单词的下降，且引导置信区间（bootstrap CI）排除了零，并在42个多语言单元中跨西班牙语、中文、德语和法语得到了复制。一个六条件的因果消融实验将这一效应孤立为确切的词汇重复，而非长度、一般冗余或语义邻近暴露。一个框架-语用控制排除了读取框架的伪影。在内部，随着 $N$ 的增加，每个目标标记的注意力下降，而分配给重复块的总预算在因果语言模型中增长，但在我们探测的掩蔽语言模型中并未如此。变化重复计数的探测器不能将读取位置视为与其测量内容正交。

View on arXiv Download PDF AI Translation

cs.CL / 3 / 2608.04037

Reconstructing Persistent Worlds from Narratives for Narrative-Grounded Interactive Experiences

从叙事中重建持久世界以实现叙事驱动的互动体验

Chen, Yi-Chun

Abstract

Designing narrative-grounded interactive experiences remains labor-intensive because interactive content must align with the underlying world implied by the narrative. Existing approaches formulate problems such as narrative planning, scene generation, and gameplay generation, each constructing computational representations tailored to specific downstream tasks rather than explicitly reconstructing and maintaining the persistent world that grounds them. We investigate reconstructing explicit persistent worlds from narrative descriptions as the central computational objective for narrative-grounded interactive realization. Rather than treating the world as an implicit by-product of downstream generation, our approach reconstructs and maintains persistent entities, locations, semantic relationships, and evolving world states while inferring only the contextual information required to support coherent interactive experiences. To investigate this perspective, we develop a reference prototype that reconstructs structured persistent world representations from narrative descriptions and subsequently instantiates playable tile-based environments. Through three representative case studies spanning a procedural scenario, an original fantasy narrative, and an adapted public-domain story, we demonstrate the feasibility of reconstructing persistent worlds and show how a shared world representation supports coherent gameplay while remaining grounded in the source narrative. By explicitly reconstructing persistent worlds prior to interactive realization, this work bridges computational narrative understanding and interactive content generation, providing a semantic foundation for AI-assisted game authoring, mixed-initiative design, educational simulations, and narrative-grounded interactive experiences.

Chinese Translation

设计叙事驱动的互动体验仍然是一个劳动密集型的过程，因为互动内容必须与叙事所暗示的基础世界相一致。现有方法将叙事规划、场景生成和游戏玩法生成等问题进行公式化，每种方法构建的计算表示都是针对特定下游任务的，而不是明确重建和维护支撑它们的持久世界。我们将从叙事描述中重建明确的持久世界视为叙事驱动互动实现的核心计算目标。我们的方法不是将世界视为下游生成的隐含副产品，而是重建和维护持久实体、位置、语义关系和不断演变的世界状态，同时仅推断支持连贯互动体验所需的上下文信息。为了探讨这一视角，我们开发了一个参考原型，该原型从叙事描述中重建结构化的持久世界表示，并随后实例化可玩瓷砖基础环境。通过三个具有代表性的案例研究，涵盖一个程序化场景、一个原创奇幻叙事和一个改编的公共领域故事，我们展示了重建持久世界的可行性，并展示了共享世界表示如何支持连贯的游戏玩法，同时保持与源叙事的紧密联系。通过在互动实现之前明确重建持久世界，这项工作架起了计算叙事理解与互动内容生成之间的桥梁，为人工智能辅助的游戏创作、混合主动设计、教育模拟和叙事驱动的互动体验提供了语义基础。

View on arXiv Download PDF AI Translation

cs.CL / 4 / 2608.04056

Learning Sexism Detection Using Multi-Agent Perspectivist Preference Optimization

使用多智能体视角偏好优化进行性别歧视检测学习

Mohammadi, Hadi, Shahedi, Tina, Bagheri, Robert A., Dastani, Mehdi, Raeissi, Masoume M.

Abstract

When people label text for sexism, they often disagree, and not because some of them are wrong: they genuinely perceive sexism differently. Most NLP systems discard this disagreement by collapsing it into a majority vote. We propose the Multi-Agent Perspectivist Preference Optimization (MAP-PO) framework to keep these different perspectives. On the EXIST 2024 dataset of labeled English and Spanish tweets, we first cluster annotators by their labeling behavior rather than their demographic attributes. We then fine-tune one Large Language Model agent per cluster to reproduce that cluster's annotation behavior, and coordinate the agents with preference optimization that combines individual and team-level rewards. We evaluate MAP-PO in four settings defined by two languages and two backbone language models, asking whether each agent reproduces the annotations of its own cluster and whether the agents together reproduce the majority label. Two findings hold in all four settings. First, without fine-tuning the agents behave almost identically, so cluster-specific training is necessary. Second, we show that training each agent only on the labels of its own cluster pushes the agents far beyond the clusters they should represent, while adding a shared team-level training signal consistently keeps each agent calibrated to its cluster.

Chinese Translation

在对文本进行性别歧视标注时，人们往往存在分歧，这并不是因为其中一些人错误，而是因为他们对性别歧视的感知确实不同。大多数自然语言处理（NLP）系统通过将这些分歧简化为多数投票来忽略这种差异。我们提出了多智能体视角偏好优化（Multi-Agent Perspectivist Preference Optimization, MAP-PO）框架，以保留这些不同的视角。在2024年EXIST数据集中，我们首先根据标注者的标注行为而非人口统计属性对标注者进行聚类。然后，我们对每个聚类中的一个大型语言模型（Large Language Model）智能体进行微调，以重现该聚类的标注行为，并通过结合个体和团队层级奖励的偏好优化来协调这些智能体。我们在由两种语言和两种基础语言模型定义的四种设置中评估MAP-PO，询问每个智能体是否重现了其自身聚类的标注，以及这些智能体是否共同重现了多数标签。在所有四种设置中都有两个发现。首先，在未进行微调的情况下，智能体的行为几乎完全相同，因此特定聚类的训练是必要的。其次，我们表明，仅在其自身聚类的标签上训练每个智能体会使智能体远远超出它们应代表的聚类，而添加共享的团队层级训练信号则始终使每个智能体保持与其聚类的校准。

View on arXiv Download PDF AI Translation

cs.CL / 5 / 2608.04160

Mind the Cap: Output-Budget Regimes Change the Measured Multilingual Reasoning Gap

注意输出限制：输出预算机制改变了测量的多语言推理差距

Goyal, Ankit, Ray, Jaideep

Abstract

Multilingual evaluations report accuracy at a single output-token cap, but languages need different numbers of tokens to express the same content, so the cap is a hidden experimental variable. We test whether the native-vs-translate gap on MGSM (German, Thai, Swahili) is a token-budget artifact for Qwen3-8B and Llama-3.1-8B-Instruct under four prompting strategies. The measured gap swings by up to 57 points across budgets, length normalization moves it by up to 38.9 points where the cap binds, and at tight caps normalization can reverse which strategy scores higher. We prospectively froze the sweep's three Qwen peaks and its near-zero value at 1024 and evaluated them on 540,000 independently hard-capped decodes: a second frozen family of six Holm-corrected tests rejects every null. The frozen test at $B^*=1024$ still fails to reject because native accuracy has already saturated there; above saturation, the residual difference is a strategy-performance gap, not an identified reasoning deficit. The same truncation channel prices a cost-ordered adaptation ladder: a cross-fitted Thai vocabulary extension closes 0.0 points of the gap at the frozen budget and 4.9 points where 19% of traces still truncate. A third frozen family varies only the announced budget at a fixed enforced cap; announcing 128 rather than 2048 tokens moves Thai native accuracy by 5.1 points, so accuracy is not a function of the enforced cap alone. A correct-emission timing identity computed from one long-cap run matches the three pre-specified MGSM peaks to 0.65 points and, in an exploratory Qwen-only analysis of three further benchmarks, tracks held-out items to 0.92 points, locating the peak exactly in five of seven cells. Treat the output cap as an independent variable and report accuracy across the budget regime, not at a single budget.

Chinese Translation

多语言评估在单一输出令牌上限下报告准确性，但不同语言表达相同内容所需的令牌数量不同，因此上限是一个隐藏的实验变量。我们测试了在四种提示策略下，MGSM（德语、泰语、斯瓦希里语）中原生与翻译之间的差距是否是Qwen3-8B和Llama-3.1-8B-Instruct的令牌预算伪影。测量的差距在不同预算间波动高达57分，长度归一化在上限约束下可移动高达38.9分，而在紧束缚的上限下，归一化甚至可以逆转哪种策略得分更高。我们前瞻性地冻结了三次Qwen峰值的扫频及其在1024时的近零值，并在540,000个独立硬限制解码上进行了评估：第二个冻结的六个经过霍尔姆校正的测试拒绝了每一个零假设。在$B^*=1024$的冻结测试中仍然无法拒绝，因为原生准确性在此处已经饱和；在饱和以上，剩余差异是策略性能差距，而不是已识别的推理缺陷。同样的截断通道为成本排序的适应阶梯定价：一个交叉拟合的泰语词汇扩展在冻结预算下缩小了0.0分的差距，而在19%的追踪仍然截断的情况下缩小了4.9分。第三个冻结的系列仅在固定强制上限下变化宣布的预算；宣布128而不是2048个令牌使泰语原生准确性变化5.1分，因此准确性并非仅仅是强制上限的函数。从一次长上限运行计算出的正确发射时机身份与三个预先指定的MGSM峰值匹配至0.65分，并且在对三个进一步基准的探索性Qwen单独分析中，跟踪保留项目至0.92分，准确定位五个七个单元中的峰值。将输出上限视为独立变量，并报告在预算机制下的准确性，而不是在单一预算下。

View on arXiv Download PDF AI Translation

cs.CL / 6 / 2608.04170

Visualizing Graph-to-Answer Mechanism Recovery in Materials-Science Hypothesis Generation

材料科学假设生成中的图到答案机制恢复可视化

Sourav, Shashwat, Pal, Subhadeep, Buehler, Markus J., Das, Sanjay, Wang, Fiona Y., Soos, Dominik, Ghosal, Tirthankar

Abstract

AI co-scientists can generate fluent materials-science hypotheses, but fluency does not show that an answer preserves a scientifically meaningful mechanism. We present a graph-to-answer mechanism-tracing case study for Graph-PRefLexOR-8B, a Qwen3-8B model adapted to expose distinct stages for brainstorming, graph construction, pattern extraction, and synthesis. We organize semantic backtracking, graph corruption, activation-based recovery measurements, and layer-by-token-region grids into a visual diagnostic workflow for inspecting this pathway. Across 100 open-ended materials-science questions, final answers remain closest to the model's own structured stages, especially synthesis. Under graph corruption, a full sweep over 37 residual-stream checkpoints, the embedding output and 36 transformer blocks, shows little mechanism recovery in the earlier transition region at layers 7--10, recovery instead concentrates in late synthesis and answer-start regions around layers 30 and 36. The workflow is intended to help scientists and model developers identify where a generated hypothesis loses or regains mechanism support before it is passed to downstream experimental planning.

Chinese Translation

人工智能共同科学家能够生成流畅的材料科学假设，但流畅性并不意味着答案保留了科学上有意义的机制。我们展示了一个针对 Graph-PRefLexOR-8B（一个经过调整的 Qwen3-8B 模型）的图到答案机制追踪案例研究，该模型旨在揭示头脑风暴、图构建、模式提取和综合的不同阶段。我们将语义回溯、图损坏、基于激活的恢复测量以及逐层区域网格组织成一个可视化诊断工作流程，以检查这一路径。在100个开放式材料科学问题中，最终答案与模型自身的结构化阶段，特别是综合阶段，保持最接近。在图损坏的情况下，对37个残余流检查点、嵌入输出和36个变换器块进行全面检查，显示在层7到10的早期过渡区域几乎没有机制恢复，而恢复则集中在层30和36附近的晚期综合和答案起始区域。该工作流程旨在帮助科学家和模型开发者识别生成的假设在传递给下游实验规划之前，在哪里失去了或重新获得了机制支持。

View on arXiv Download PDF AI Translation

cs.CL / 7 / 2608.04183

Test, then Route: How Language Models Execute In-Context Conditional Rules Across Models and Languages

先测试，再路由：语言模型如何在模型和语言之间执行上下文条件规则

Thavarasa, Luxshan, Sukumar, Sivasuthan

Abstract

When a language model follows an in-context conditional rule such as "if P(x) then A else B," does it assemble a runtime circuit with one module that tests the predicate and another that routes the answer? We probe this with activation patching under a four-donor design whose two swapped-rule donors make the condition and the answer word disagree, so each layer reveals which of the two it carries. Across three open models from two families and six languages sharing one fixed item bank, a mid-stack residual band carries the predicate's truth value: patching it reroutes the answer with predicate-outcome flip near 1.0 and mapping flip near 0.0, meeting a strict pre-specified isolation criterion in 17 of 18 cells, and the same localization holds across five predicate families. The router shows the opposite profile. A learned subspace flips A and B near-perfectly within the trained pair yet transfers to a new pair at approximately 0 in every model, while in Gemma-3-4B (the only model probed cross-lingually) it transfers at approximately 0.98 to the same pair in other languages. Under every probe we ran, the router direction is token-bound and non-transferable (largely answer-readout in Gemma, pair-specific in Qwen) rather than an abstract routing module. Test is modular; under these probes, route is not.

Chinese Translation

当一个语言模型遵循一个上下文条件规则，例如“如果 P(x) 则 A，否则 B”，它是否构建了一个运行时电路，其中一个模块测试谓词，另一个模块路由答案？我们通过激活修补在一个四捐赠者设计下进行探讨，其中两个交换规则的捐赠者使条件和答案词不一致，因此每一层揭示了它携带的两个中的哪一个。在来自两个家族的三个开放模型和六种语言中，共享一个固定的项目库，一个中间堆栈残差带携带谓词的真值：修补它使答案重新路由，谓词结果翻转接近 1.0，而映射翻转接近 0.0，在 18 个单元中满足严格的预先指定隔离标准，并且这种定位在五个谓词家族中保持一致。路由器显示出相反的特征。一个学习的子空间在训练对内几乎完美地翻转 A 和 B，但在每个模型中对新对的转移约为 0，而在 Gemma-3-4B（唯一一个跨语言探测的模型）中，它对其他语言中的相同对的转移约为 0.98。在我们进行的每个探测下，路由器方向是与标记绑定且不可转移的（在 Gemma 中主要是答案读取，在 Qwen 中是对特定的配对），而不是一个抽象的路由模块。测试是模块化的；在这些探测下，路由则不是。

View on arXiv Download PDF AI Translation

cs.CL / 8 / 2608.04186

Large Language Models for Low-Resource Languages: A Conceptual Framework for an Electronic Explanatory Dictionary of the Tajik Language

低资源语言的大型语言模型：塔吉克语言电子解释词典的概念框架

Arabov, Mullosharaf K.

Abstract

This paper presents a conceptual framework for developing an electronic explanatory dictionary of the Tajik language using large language models (LLMs). The relevance of the work stems from the absence of a comprehensive digital lexicographic resource for Tajik that is comparable in functionality to dictionaries for high-resource languages, and from the limited adaptation of modern natural language processing technologies to low-resource language systems. Based on a systematic survey of existing linguistic, statistical, and corpus resources, we propose a dictionary architecture that integrates modules for morphological analysis, lemmatization, semantic clustering, and dictionary entry generation using LLMs. The choice of subword tokenization is justified by the agglutinative nature of Tajik morphology and its high morphological variability, along with a parameter-efficient fine-tuning (PEFT) strategy suitable for limited annotated data. The novelty of the work lies in proposing the first holistic conceptual architecture of an explanatory dictionary for Tajik that unifies classical lexicographic methods, language statistics, and generative capabilities of LLMs into a single system. The practical significance of the study is the formation of a methodological foundation for developing a full-featured electronic dictionary that can serve both as a lexicographic tool and as a core resource for machine translation, automatic summarization, sentiment analysis, and other applied NLP tasks. The paper is intended for specialists in computational linguistics, lexicography, and developers of natural language processing systems working with low-resource languages.

Chinese Translation

本文提出了一个概念框架，用于利用大型语言模型（LLMs）开发塔吉克语言的电子解释词典。该工作的相关性源于缺乏与高资源语言词典在功能上可比的全面数字词典资源，以及现代自然语言处理技术对低资源语言系统的有限适应性。基于对现有语言学、统计和语料库资源的系统调查，我们提出了一种词典架构，该架构整合了形态分析、词元化、语义聚类和使用LLMs生成词典条目的模块。选择子词标记化的理由在于塔吉克形态的黏着性特征及其高度的形态变异性，同时采用适合有限标注数据的参数高效微调（PEFT）策略。该工作的创新之处在于提出了塔吉克语言解释词典的首个整体概念架构，将经典词典方法、语言统计和LLMs的生成能力统一为一个系统。研究的实际意义在于为开发一个功能齐全的电子词典奠定方法论基础，该词典既可以作为词典工具，也可以作为机器翻译、自动摘要、情感分析及其他应用自然语言处理任务的核心资源。本文旨在为计算语言学、词典学专家以及从事低资源语言自然语言处理系统开发的研究人员提供参考。

View on arXiv Download PDF AI Translation

cs.CL / 9 / 2608.04193

Patients-like-me: A Variational LM--GNN Framework for Explainable Clinical Prediction

类似患者：一种可解释临床预测的变分LM-GNN框架

Wang, Xinyu, Li, Yixuan, Wu, Hanwei, Lu, Qincheng, Yeh, Chi-Kuang, Chang, Xiao-Wen, Song, Ziyang

Abstract

Language models (LMs) offer strong textual representations for electronic health records (EHRs), but they encode patient sequences in isolation and provide limited explainability. Graph neural networks (GNNs) complement LMs by incorporating inter-patient relationships and enabling reference-patient attribution, yet they rely on high-quality patient representations. We propose Patients-like-me (PLM), a unified LM--GNN framework that integrates local patient semantics with global cohort structure. To train PLM efficiently, we introduce a Variational Expectation-Maximization algorithm that alternates LM and GNN updates under a supervised variational objective. Extensive experiments on MIMIC-III and MIMIC-IV show that PLM consistently outperforms state-of-the-art methods, with improvements generalizing across encoder-only and decoder-only LM backbones. These gains are achieved with only modest additional computational overhead. PLM also provides reference-patient explanations by retrieving influential similar patients, while edge-masking experiments confirm that the highest-ranked references have the greatest impact on model predictions.

Chinese Translation

语言模型（LM）为电子健康记录（EHR）提供了强大的文本表示，但它们在孤立的情况下编码患者序列，并且提供的可解释性有限。图神经网络（GNN）通过结合患者间关系并实现参考患者归因来补充LM，但它们依赖于高质量的患者表示。我们提出了类似患者（Patients-like-me，PLM），这是一个统一的LM-GNN框架，整合了局部患者语义与全局队列结构。为了高效训练PLM，我们引入了一种变分期望最大化算法，该算法在监督变分目标下交替进行LM和GNN的更新。在MIMIC-III和MIMIC-IV上的大量实验表明，PLM始终优于最先进的方法，并且改进在仅使用编码器或仅使用解码器的LM骨干网络上均有推广。这些提升是在仅增加适度的计算开销的情况下实现的。PLM还通过检索有影响力的相似患者提供参考患者解释，而边缘掩蔽实验确认了排名最高的参考对模型预测具有最大的影响。

View on arXiv Download PDF AI Translation

cs.CL / 10 / 2608.04240

Hallucinations on the Board: Tool-Augmented Evaluation of LLM Chess Commentary

棋盘上的幻觉：工具增强的LLM棋评估

Hebbar, S. Ashwin, Sheng, Peiyao, Oh, Sewoong, Viswanath, Pramod

Abstract

Superhuman game engines in domains like chess have made expert-level evaluations easily accessible, yet they communicate what is true without the natural-language explanations that make such expertise educationally useful to experts and non-experts alike. Large language models could, in principle, bridge this gap, but they frequently hallucinate due to limited domain-specific knowledge, and standard reference-based or LLM-as-a-judge frameworks cannot reliably detect these errors. In this work, we present ACT-Eval, an evaluation framework that decomposes chess commentary into atomic claims and routes them to engine-supported tools and expert-annotated gold references to assess factual correctness, conceptual coverage, and move-quality judgment. We release a benchmark of 325 position--move pairs spanning pedagogical, tournament, and critical positions, including 125 positions with expert-verified gold atoms and a five-class error taxonomy. Evaluating leading proprietary and open-weight models, we find that factual hallucinations remain pervasive in chess commentary: GPT-5.4 without tools produces incorrect sub-claims 22.0% of the time, while smaller open-weight models exceed 40%. Although tool augmentation substantially improves factual correctness and move-quality assessment, coverage of expert strategic and tactical ideas remains limited across all models. Human calibration shows that ACT-Eval's factual judgments fall within the observed range of inter-human agreement, while its coverage scores correlate strongly with human assessments of strategic completeness.

Chinese Translation

在像棋类这样的领域，超人类游戏引擎使得专家级评估变得易于获取，然而它们传达的真相却缺乏自然语言解释，这使得这种专业知识对专家和非专家都不具备教育意义。大型语言模型在理论上可以弥补这一差距，但由于领域特定知识的局限性，它们经常产生幻觉，而基于标准参考或LLM作为评判者的框架无法可靠地检测这些错误。在本研究中，我们提出了ACT-Eval，一个评估框架，它将棋评分解为原子声明，并将其路由到引擎支持的工具和专家注释的黄金参考，以评估事实正确性、概念覆盖和走棋质量判断。我们发布了一个包含325个位置-走棋对的基准数据，涵盖了教学、比赛和关键位置，其中包括125个经过专家验证的黄金原子和五类错误分类。评估领先的专有模型和开放权重模型，我们发现棋评中的事实幻觉依然普遍存在：在没有工具的情况下，GPT-5.4产生错误子声明的频率为22.0%，而较小的开放权重模型超过40%。尽管工具增强显著提高了事实正确性和走棋质量评估，但所有模型在专家战略和战术思想的覆盖方面仍然有限。人类校准显示，ACT-Eval的事实判断落在观察到的人际一致性范围内，而其覆盖得分与人类对战略完整性的评估高度相关。

View on arXiv Download PDF AI Translation

cs.CL / 11 / 2608.04260

Towards End-to-End Multilingual Metaphor Processing: Integrating Detection, Translation, and Evaluation

面向端到端多语言隐喻处理：整合检测、翻译与评估

Liang, Jiahui, Han, Lifeng

Abstract

Metaphorical language remains a major challenge for multilingual natural language processing because successful interpretation and translation require reasoning beyond literal lexical meaning. Existing research has largely investigated metaphor detection, machine translation, and translation evaluation as separate tasks, while little work has explored how these components can be integrated into a unified computational framework. This PhD proposal aims to develop an end-to-end framework for multilingual metaphor processing consisting of three complementary research directions: (1) robust metaphor detection across languages, (2) metaphor-oriented translation evaluation for both human assessment and automatic quality estimation, and (3) joint modelling that connects metaphor detection with translation evaluation. The proposed research will combine linguistic theory with recent advances in large language models to develop new datasets, annotation methodologies, evaluation benchmarks, and automatic evaluation approaches for metaphor-aware machine translation. The expected outcome is a unified framework that improves both the development and evaluation of multilingual NLP systems when processing figurative language.

Chinese Translation

隐喻语言仍然是多语言自然语言处理中的一大挑战，因为成功的解释和翻译需要超越字面词义的推理。现有研究主要将隐喻检测、机器翻译和翻译评估视为独立任务，而很少有研究探讨如何将这些组件整合到一个统一的计算框架中。本博士提案旨在开发一个端到端的多语言隐喻处理框架，包含三个互补的研究方向：（1）跨语言的稳健隐喻检测；（2）面向隐喻的翻译评估，适用于人工评估和自动质量估计；（3）将隐喻检测与翻译评估连接的联合建模。所提议的研究将结合语言学理论与大型语言模型的最新进展，开发新的数据集、标注方法、评估基准和隐喻感知机器翻译的自动评估方法。预期结果是一个统一框架，能够在处理比喻语言时改善多语言自然语言处理系统的开发与评估。

View on arXiv Download PDF AI Translation

cs.CL / 12 / 2608.04268

The Fairness Collapse Phenomenon: Bias Amplification in Language Models Trained on Synthetic Data

公平崩溃现象：在合成数据上训练的语言模型中的偏见放大

Proskurina, Irina, Gourru, Antoine, Velcin, Julien

Abstract

Generative models trained on artificially generated data have been shown to exhibit model collapse, resulting in significant performance degradation. As synthetic content increasingly contaminates the training corpora of language models, this raises critical concerns about the use of open data in continued pretraining. Although previous work has demonstrated model collapse in language models, it remains unclear whether exposure to synthetic data amplifies or attenuates the social biases already present in pretrained models. Because language models are known to reproduce and amplify demographic stereotypes, recursive training on self-generated data may create a self-reinforcing feedback loop in which biased associations become progressively stronger across generations. We call this hypothesized phenomenon fairness collapse. In this work, we construct controlled training regimes in which models are repeatedly trained on synthetic data using the Bias in Bios dataset. Across experiments, we observe a consistent and concerning pattern: fairness degradation emerges before substantial degradation is reflected by standard language-modeling metrics. This result highlights a critical risk associated with synthetic data contamination in language model training: bias can increase silently before strong indicators of model collapse become apparent.

Chinese Translation

在人工生成数据上训练的生成模型已被证明会出现模型崩溃，导致显著的性能下降。随着合成内容越来越多地污染语言模型的训练语料，这引发了对在持续预训练中使用开放数据的重大担忧。尽管之前的研究已经证明了语言模型中的模型崩溃，但尚不清楚接触合成数据是否会放大或减弱预训练模型中已存在的社会偏见。由于已知语言模型会再现并放大人口统计学刻板印象，因此在自生成数据上进行递归训练可能会形成一个自我强化的反馈循环，使得偏见关联在各代之间逐渐增强。我们将这一假设现象称为公平崩溃。在本研究中，我们构建了受控的训练机制，在这些机制中，模型在Bias in Bios数据集上反复训练合成数据。在实验中，我们观察到一个一致且令人担忧的模式：公平性下降在标准语言建模指标反映出显著下降之前就已出现。这一结果突显了在语言模型训练中合成数据污染所带来的关键风险：偏见可能在模型崩溃的强指示变得明显之前悄然增加。

View on arXiv Download PDF AI Translation

cs.CL / 13 / 2608.04286

Eliciting Intrinsic Hallucinations in LLMs via Semantically Equivalent Adversarial Attacks

通过语义等价对抗攻击引发大型语言模型中的内在幻觉

Sharma, Atri Vivek, Formento, Brian, Lomuscio, Alessio

Abstract

Large language models (LLMs) are often used in conjunction with external knowledge sources to improve their factual accuracy and decrease hallucinations, through methods such as Retrieval-Augmented Generation (RAG). However, these systems remain susceptible to intrinsic hallucinations, where the model generates unfaithful or fabricated information that is not supported by the retrieved evidence. We propose a novel framework to assess model robustness against this phenomenon by stress-testing using natural, semantically equivalent variations of a user query found via adversarial optimization methods. We apply our framework, which enforces strict semantic equivalence constraints and an intrinsic hallucination objective, to a range of adversarial attack techniques across white-box, gray-box, and black-box adversarial settings. Evaluating these attacks on 5 open-source and 5 closed-source generator models across 3 datasets, we demonstrate that even state-of-the-art models are highly susceptible to meaning-preserving perturbations, which significantly degrade contextual faithfulness (by up to 50% for GPT-5-mini). Our findings indicate that faithful use of in-context evidence remains fragile even in state-of-the-art LLMs, motivating architectures and training objectives that enforce robust grounding independent of surface query form. Code is available at: https://github.com/atriviveksharma/intrinsic_hall

Chinese Translation

大型语言模型（LLMs）通常与外部知识源结合使用，以提高其事实准确性并减少幻觉现象，方法包括检索增强生成（Retrieval-Augmented Generation, RAG）。然而，这些系统仍然容易受到内在幻觉的影响，即模型生成不真实或虚构的信息，而这些信息并未得到检索证据的支持。我们提出了一种新颖的框架，通过使用对抗优化方法找到的用户查询的自然语义等价变体进行压力测试，以评估模型对这一现象的鲁棒性。我们将该框架应用于一系列对抗攻击技术，涵盖白盒、灰盒和黑盒对抗设置，框架强制执行严格的语义等价约束和内在幻觉目标。在3个数据集上对5个开源和5个闭源生成模型进行评估，我们证明即使是最先进的模型也对保持意义的扰动高度敏感，这显著降低了上下文的真实性（对于GPT-5-mini，降低幅度可达50%）。我们的研究结果表明，即使在最先进的LLMs中，忠实使用上下文证据仍然脆弱，这促使我们设计架构和训练目标，以强制实现独立于表面查询形式的鲁棒基础。代码可在以下链接获取：https://github.com/atriviveksharma/intrinsic_hall

View on arXiv Download PDF AI Translation

cs.CL / 14 / 2608.04299

Searching for Sound-Meaning Collisions: Graph-Based Affordance Retrieval and Multi-Evaluator Ranking for Pun Translation at CLEF 2026 JOKER Task 2

寻找声音与意义的碰撞：基于图的可供性检索与多评估者排名在CLEF 2026 JOKER任务2中的双关语翻译

Taylor, Russell, Brikman, Adam, Awate, Prateek

Abstract

Fifteen years ago, Low proposed that pun translators should stop searching for equivalent words and instead search for new points of contact between sound and meaning. In this paper, we investigate that idea computationally. We model pun translation as a process of discovery, exploration, and selection. A retrieval system searches semantic and phonological neighborhoods for target-language affordances: sound-meaning bridges that may support new wordplay. Multiple language models then explore these opportunities by generating competing translations, while a multi-perspective generate-and-rank architecture selects among them. Beyond system development, our primary contribution is an analysis of how retrieved affordances propagate through the translation process. We find that generators actively exploit retrieved opportunities, evaluators progressively concentrate around stronger sound-meaning bridges, and exact phonological collisions are selected at disproportionately high rates when available. At the same time, many puns still yield no usable affordances, suggesting that retrieval remains the central bottleneck in computational pun translation. The resulting picture is remarkably close to the process envisioned by Low. Successful pun translation emerges not from preserving source-language words, but from discovering new places in the target language where sound and meaning collide.

Chinese Translation

十五年前，Low提出双关语翻译者应停止寻找等价词，而应寻找声音与意义之间的新接触点。本文从计算的角度探讨了这一观点。我们将双关语翻译建模为发现、探索和选择的过程。检索系统在语义和音韵邻域中搜索目标语言的可供性：可能支持新词语游戏的声音-意义桥梁。多个语言模型随后通过生成竞争性翻译来探索这些机会，而多视角生成与排名架构则在其中进行选择。除了系统开发外，我们的主要贡献是分析检索到的可供性如何在翻译过程中传播。我们发现生成器积极利用检索到的机会，评估者逐渐集中在更强的声音-意义桥梁上，当确切的音韵碰撞可用时，它们的选择率异常高。同时，许多双关语仍然没有产生可用的可供性，这表明检索仍然是计算双关语翻译中的核心瓶颈。最终的结果与Low设想的过程非常接近。成功的双关语翻译不是通过保留源语言词汇实现的，而是通过发现目标语言中声音与意义碰撞的新地方。

View on arXiv Download PDF AI Translation

cs.CL / 15 / 2608.04307

MIDAS: Multi-LLM Iterative Data-Adaptive Summarization

MIDAS：多LLM迭代数据自适应摘要

Lee, Karen, Balaram, Dhanashree, Shon, Seojun, Rasheed, Umair

Abstract

Text summarization is deceptively difficult. While condensing information seems straightforward, real-world enterprise summarization of support tickets, legal documents, incident reports, and more, demands strict adherence to domain-specific guidelines, output formats, and organizational conventions. Crafting prompts that reliably satisfy these constraints is labor-intensive, requiring significant human expertise and continuous maintenance as requirements evolve. Existing automated prompt optimization methods reduce this burden through Large Language Model (LLM) critique-driven refinement, yet remain limited by static prompts that cannot adapt to the diversity of summary applications. We propose Multi-LLM Iterative Data-Adaptive Summarization (MIDAS), a multi-LLM framework that extends this paradigm with data-driven pattern learning and use-case-specific personalization, enabling automatic adaptation to different summarization requirements without manual prompt engineering. Applied to enterprise customer ticket summarization across five output formats, MIDAS achieves the strongest overall performance against state-of-the-art critique-driven optimization frameworks such as CriSPO and ZERA, improving ROUGE-1 by up to 11.0%, ROUGE-2 by up to 18.2%, and ROUGE-L by up to 8.0%, while consistently improving BERTScore F1 across all formats and output types. We additionally demonstrate cross-model and cross-domain generalization through multi-LLM configurations and finance-domain summarization benchmarks.

Chinese Translation

文本摘要的难度常常被低估。尽管信息的压缩看似简单，但现实世界中对支持票据、法律文件、事件报告等的企业摘要，要求严格遵循特定领域的指导原则、输出格式和组织惯例。制作能够可靠满足这些约束条件的提示是劳动密集型的，需依赖大量人类专业知识，并随着需求的变化进行持续维护。现有的自动化提示优化方法通过大型语言模型（LLM）的评估驱动改进来减轻这一负担，但仍然受到静态提示的限制，无法适应摘要应用的多样性。我们提出了多LLM迭代数据自适应摘要（MIDAS），这是一个多LLM框架，扩展了这一范式，通过数据驱动的模式学习和用例特定的个性化，实现对不同摘要需求的自动适应，而无需手动提示工程。应用于五种输出格式的企业客户票据摘要，MIDAS在与最先进的评估驱动优化框架（如CriSPO和ZER）比较时，表现出最强的整体性能，ROUGE-1提高了最多11.0%，ROUGE-2提高了最多18.2%，ROUGE-L提高了最多8.0%，同时在所有格式和输出类型中持续提高BERTScore F1。我们还通过多LLM配置和金融领域摘要基准展示了跨模型和跨领域的泛化能力。

View on arXiv Download PDF AI Translation

cs.CL / 16 / 2608.04311

Pun Intended: Multi-Agent Translation of Wordplay with Contrastive Learning and Phonetic-Semantic Embeddings

双关语的意图：通过对比学习和语音-语义嵌入的多智能体翻译

Taylor, Russell, Herbert, Benjamin, Sana, Michael

Abstract

Translating wordplay across languages has long challenged both professional translators and machine translation systems. We investigate three approaches to translating puns from English to French by combining large language models with linguistic constraints for wordplay generation. Our baseline uses a large language model with feedback from a discriminator prompted with positive and negative French examples. Our guided reasoning pipeline uses combined phonetic-semantic embeddings to retrieve lexical candidates for wordplay generation. Finally, our multi-agent framework iteratively evaluates and regenerates candidate translations using specialized feedback. Moving beyond literal translation, our objective is to preserve the linguistic creativity, ambiguity, and humor of the source-text wordplay rather than simply reproduce its vocabulary. The multi-agent and guided chain-of-thought systems ranked first and second, respectively, in the CLEF JOKER 2025 Task 2 competition under expert human evaluation, despite only modest improvements in BLEU and BERTScore. These findings suggest that both explicit phonetic-semantic guidance and iterative multi-agent evaluation can improve LLM-based wordplay translation relative to direct discriminator-guided generation, particularly when balancing semantic fidelity, phonetic similarity, and natural target-language expression

Chinese Translation

跨语言翻译双关语长期以来一直对专业翻译人员和机器翻译系统构成挑战。我们探讨了三种将英语双关语翻译成法语的方法，结合了大型语言模型和用于双关语生成的语言学约束。我们的基线方法使用大型语言模型，并通过提示正面和负面法语示例的鉴别器提供反馈。我们的引导推理流程使用结合的语音-语义嵌入来检索双关语生成的词汇候选项。最后，我们的多智能体框架通过使用专业反馈迭代评估和再生成候选翻译。我们的目标是超越字面翻译，保留源文本双关语的语言创造力、模糊性和幽默感，而不仅仅是简单地再现其词汇。在专家人类评估下，多智能体和引导思维链系统在CLEF JOKER 2025任务2竞赛中分别排名第一和第二，尽管在BLEU和BERTScore上仅有适度的改善。这些发现表明，显式的语音-语义引导和迭代的多智能体评估可以相对于直接的鉴别器引导生成改善基于大型语言模型的双关语翻译，特别是在平衡语义忠实性、语音相似性和自然目标语言表达时。

View on arXiv Download PDF AI Translation

cs.CL / 17 / 2608.04322

DataRx: Missingness-Aware Sampling for Safer Large Language Model Task-Specific Fine-Tuning

DataRx：关注缺失性的采样方法以增强大型语言模型的任务特定微调安全性

Zhang, Junbo, Zhou, Qianli, Deng, Xinyang, Jiang, Wen

Abstract

Task-specific fine-tuning can improve the performance of large language models (LLMs) on downstream tasks. However, our study reveals that task-specific fine-tuning can also weaken the safety guardrails of aligned LLMs. A widely adopted strategy for preserving safety during fine-tuning is to incorporate safety data. Although previous studies have shown that randomly mixing safety data can alleviate safety degradation, the underlying principle determining why some safety examples are more effective than others still remains unclear. In this paper, we propose DataRx, a missingness-aware sampling method for selecting safety-critical examples. DataRx is based on the hypothesis that a safety sample is more effective when the selected examples provide safety signals that fill the missing parts of LLMs' safety capabilities. DataRx's key insight is leveraging high-dimensional hidden representations rather than discrete tokens to quantify the safety signal gap between the target model's native response and the safety reference response. The results show that, with only 1% additional safety samples from BeaverTails, DataRx reduces the average attack success rate of Llama3-8B-Instruct across seven downstream tasks from 59.23% under random sampling to 13.70%. In addition, DataRx can be combined with the existing safety data synthesis method to further enhance safety defenses during fine-tuning. We hope that DataRx will inspire more data-centric defense research.

Chinese Translation

任务特定微调可以提升大型语言模型（LLMs）在下游任务上的表现。然而，我们的研究表明，任务特定微调也可能削弱对齐LLMs的安全防护措施。在微调过程中，保持安全性的一种广泛采用的策略是纳入安全数据。尽管之前的研究表明，随机混合安全数据可以缓解安全性下降，但仍不清楚为何某些安全示例比其他示例更有效的基本原理。在本文中，我们提出了DataRx，一种关注缺失性的采样方法，用于选择安全关键示例。DataRx基于这样的假设：当所选示例提供的安全信号能够填补LLMs安全能力的缺失部分时，安全样本的效果更佳。DataRx的关键见解在于利用高维隐含表示而非离散标记来量化目标模型的原生响应与安全参考响应之间的安全信号差距。结果表明，仅通过从BeaverTails中增加1%的安全样本，DataRx将Llama3-8B-Instruct在七个下游任务中的平均攻击成功率从59.23%（随机采样）降低至13.70%。此外，DataRx可以与现有的安全数据合成方法结合，以进一步增强微调过程中的安全防御。我们希望DataRx能够激发更多以数据为中心的防御研究。

View on arXiv Download PDF AI Translation

cs.CL / 18 / 2608.04330

Right Reset: Chunking by Prefix Removal

右重置：通过前缀移除进行分块

Vegeto, Mike

Abstract

Removing the left context from a causal language model reveals a useful kind of boundary: an edge where the model processes the same right-hand tokens with little change. We turn this observation into prefix-removal probing and introduce Right Reset (RR), which measures preservation of the right-hand hidden-state trajectory. A dynamic program converts RR edge scores into variable-length chunks. On flattened text formed by concatenating topically similar records after deleting their separators and layout, RR recovers 47.7% of the original records as clean units, versus 25.9% for a BGE embedding boundary, the strongest tested conventional baseline without task-specific model training. The gain persists after rendering and OCR. Passive scores from the same Qwen3-4B layer and direct prompting of a same-scale instruction model perform substantially worse on flattened records. Across six language models, RR-selected cuts also undergo consistently less local output disruption than unselected candidate edges. An observed-token likelihood-ratio readout is competitive in some architectures, indicating that the central contribution is the intervention: context dependence itself can provide a boundary signal when surface structure is weak.

Chinese Translation

从因果语言模型中移除左侧上下文揭示了一种有用的边界：模型以较小的变化处理相同的右侧标记的边缘。我们将这一观察转化为前缀移除探测，并引入右重置（Right Reset, RR），该方法测量右侧隐状态轨迹的保持情况。一个动态程序将RR边缘分数转换为可变长度的块。在通过删除分隔符和布局后连接主题相似记录形成的扁平文本中，RR以干净单元的形式恢复了47.7%的原始记录，而BGE嵌入边界的恢复率为25.9%，后者是未经过特定任务模型训练的最强常规基线。该增益在渲染和光学字符识别（OCR）后仍然存在。来自同一Qwen3-4B层的被动分数以及对同规模指令模型的直接提示在扁平记录上的表现明显较差。在六个语言模型中，RR选择的切分也经历了比未选择的候选边缘更少的局部输出干扰。观察到的标记似然比读出在某些架构中具有竞争力，表明中心贡献是干预：上下文依赖本身可以在表面结构薄弱时提供边界信号。

View on arXiv Download PDF AI Translation

cs.CL / 19 / 2608.04339

Equitable System-Prompt Selection via Constrained Mixed-Strategy GroupDRO

通过约束混合策略GroupDRO实现公平的系统提示选择

Xu, Mengyu, Yang, Qiaoxin, Liu, Zhihan, Xu, Ruiyao, Liu, Zachary, Chen, Kezhen, Gao, Chongyang

Abstract

Large language models are increasingly used for information seeking, yet semantically equivalent questions phrased in different ways can receive answers of considerably different quality. System prompts are widely employed to steer response behavior, but they are typically optimized for average-case quality, so some question phrasings may still receive incomplete or low-quality answers. To address this, we formulate a constrained mixed-strategy GroupDRO framework for system-prompt selection. Instead of optimizing the system-prompt text, the framework assigns weights to system prompts in an existing pool to minimize the worst-case information-quality loss across evaluation metrics and groups, while constraining the mean loss to stay close to that of average-based selection. Because pool generation and selection are decoupled, the method applies to any system-prompt pool and can leverage an ensemble of complementary system prompts rather than a single one. Across five LLMs on two bilingual medical and consumer-finance benchmarks, the constrained method reduces the Overall Mean, Worst 25% Mean, and Worst by 13.1%, 13.2%, and 13.7% on average relative to no mitigation while keeping overall quality close to Average selection. Its multi-prompt weights reveal complementarity across metric-group pairs. Code and data are available at https://github.com/Rainxu09/equitable-system-prompt-selection.

Chinese Translation

大型语言模型在信息检索中的应用日益广泛，但以不同方式表述的语义等价问题可能会得到质量差异显著的答案。系统提示被广泛用于引导响应行为，但通常是针对平均情况质量进行优化，因此某些问题表述仍可能得到不完整或低质量的答案。为了解决这个问题，我们提出了一种约束混合策略GroupDRO框架用于系统提示选择。该框架并不优化系统提示文本，而是为现有池中的系统提示分配权重，以最小化在评估指标和群体之间的最坏情况信息质量损失，同时约束平均损失保持接近于基于平均选择的损失。由于池的生成和选择是解耦的，该方法适用于任何系统提示池，并可以利用一组互补的系统提示，而不是单一的提示。在两个双语医疗和消费金融基准上对五个大型语言模型进行的实验表明，该约束方法在相对于无缓解的情况下，平均减少了整体均值、最差25%均值和最差情况的损失，分别为13.1%、13.2%和13.7%，同时保持整体质量接近于平均选择。其多提示权重揭示了指标-群体对之间的互补性。代码和数据可在 https://github.com/Rainxu09/equitable-system-prompt-selection 获取。

View on arXiv Download PDF AI Translation

cs.CL / 20 / 2608.04355

The Calibration Floor: Format Repair Can Masquerade as Self-Correction at Small-to-Mid Scale

校准底线：格式修复可以伪装成小到中规模的自我修正

Chen, Mingguang, Qu, Bo, Wang, Licheng

Abstract

Accuracy changes after language-model self-revision are usually interpreted as changes in reasoning. We show this can fail at the answer-extraction boundary, and test the failure causally rather than only observationally. Across Qwen3.5 (0.8B-9B), Gemma-4-12B, and two frontier models via API (Tencent Hy3, Nvidia Nemotron-3-Ultra-550B) in 29 primary cells plus a frontier arm, we decompose the always-revise accuracy shift into a content margin (both answers parseable) and format-recovery/loss margins (parseability changes). On 12 cells with meaningful unparseable-answer rates, format effects exceed content effects (Wilcoxon p=1.7e-3). To test this causally, we force already-generated reasoning through grammar-constrained decoding so every answer is parseable by construction: across 14 cells this closes a median 71% of the gap between the naive total effect and the content-margin estimate, with two cells converging exactly and a residual on the two largest-effect cells reported rather than dismissed. A clustered model confirms floor-scale (0.8B/2B) models have far higher odds of content-level change and harm than capable-scale models (p<1e-7). Replicating a cited confidence-gating protocol verbatim on Qwen3.5 does not reproduce its reported gain and shows the same near-zero content margin. A frontier check on much larger models shows format-dominance intensifying with scale: content margin is exactly zero in all 5 cells despite total effects up to +0.275, though this arm is lower-powered. The calibration-floor criterion on the content margin reveals a squeeze: floor-scale cells have headroom but insufficient signal, capable-scale cells have signal but little headroom; only one cell is marginally viable, with negligible sealed-holdout gain. Content is a minority share of what the field has measured as self-correction. We release the instrument, code, and derived results.

Chinese Translation

语言模型自我修正后的准确性变化通常被解释为推理的变化。我们展示了这种解释在答案提取边界处可能失效，并通过因果方式而不仅仅是观察方式来测试这种失效。在 Qwen3.5 (0.8B-9B)、Gemma-4-12B 以及通过 API 的两个前沿模型（腾讯 Hy3、英伟达 Nemotron-3-Ultra-550B）中，在 29 个主要单元和一个前沿分支中，我们将始终修正的准确性变化分解为内容边际（两个答案均可解析）和格式恢复/损失边际（可解析性变化）。在 12 个具有显著不可解析答案率的单元中，格式效应超过内容效应（Wilcoxon p=1.7e-3）。为了因果测试，我们强制已经生成的推理通过语法约束解码，使每个答案在构造上都是可解析的：在 14 个单元中，这缩小了天真总效应与内容边际估计之间的中位数 71% 的差距，其中两个单元完全收敛，剩余部分在两个最大效应单元中报告而非被忽视。聚类模型确认，底线规模（0.8B/2B）模型在内容级变化和损害方面的几率远高于能力规模模型（p<1e-7）。在 Qwen3.5 上逐字复制引用的信心门控协议并未重现其报告的增益，并显示出相同的近零内容边际。对更大模型的前沿检查显示，格式主导效应随着规模的增加而加剧：在所有 5 个单元中，内容边际恰好为零，尽管总效应高达 +0.275，尽管该分支的功效较低。内容边际上的校准底线标准揭示了一种挤压：底线规模单元有余地但信号不足，能力规模单元有信号但余地有限；只有一个单元在边际上是可行的，且密封保留增益微不足道。内容在该领域所测量的自我修正中占少数份额。我们发布了工具、代码和衍生结果。

View on arXiv Download PDF AI Translation

cs.CL / 21 / 2608.04374

FinReportBench: Measuring and Improving Institution-Grade Financial Report Generation

FinReportBench：衡量和提升机构级财务报告生成

Tang, Yinghao, Zhenwei, Tan, Wang, Yiyao, Gu, Wanli, Zhang, Xiaolu, Zhou, Jun, Chen, Wei

Abstract

Large language models can produce fluent financial analysis, but fluency alone does not establish whether a report is suitable for institutional delivery. We introduce FinReportBench, an expert-grounded benchmark for measuring and improving institution-grade financial report generation. Expert review reveals recurring gaps in report identity, institutional components, source discipline, and visual delivery. We derive a 35-item rubric through expert partial orders, multimodal evidence, and audits of decision boundaries, covering deliverability, report identity, and institutional completeness. Starting from 10,000 balanced Chinese and English financial-research source records, we curate 244 bilingual tasks across three research objects and two input tiers. Each task separates the public query, reconstructed research trajectory, and hidden source packet. Three independent judge families reproduce the expert partial order at near-ceiling rates, showing that bounded, observable criteria support reliable evaluation. Across nine model families, basic deliverability is nearly saturated, while report identity and institutional completeness remain the primary bottlenecks. The largest cross-model gaps concern generation-trace control, information density, and data discipline rather than basic report framing. We then use benchmark-guided skill distillation to turn recurrent failures into reusable generation and self-review constraints. Across five model families, the evolved skill improves mean G1 by 33.85 points and mean G2 by 13.83 points over paired no-skill runs while preserving G0 for every pair. Code and benchmark artifacts are available at https://github.com/MisterBrookT/finreportbench.

Chinese Translation

大型语言模型能够生成流畅的财务分析，但仅仅流畅性并不能确定报告是否适合机构交付。我们引入了FinReportBench，这是一个基于专家的基准，用于衡量和提升机构级财务报告生成。专家评审揭示了报告身份、机构组成、来源学科和视觉呈现方面的反复缺口。我们通过专家的部分排序、多模态证据和决策边界的审计，得出了一个包含35项指标的评分标准，涵盖了可交付性、报告身份和机构完整性。从10,000个平衡的中英文财务研究源记录出发，我们在三个研究对象和两个输入层次上策划了244个双语任务。每个任务分离了公共查询、重建的研究轨迹和隐藏的源数据包。三个独立的评审组以接近上限的速度再现了专家的部分排序，表明有限且可观察的标准支持可靠的评估。在九个模型家族中，基本可交付性几乎达到饱和，而报告身份和机构完整性仍然是主要瓶颈。跨模型的最大差距涉及生成追踪控制、信息密度和数据学科，而非基本报告框架。然后，我们使用基准指导的技能蒸馏将反复失败转化为可重用的生成和自我评审约束。在五个模型家族中，演变后的技能使得平均G1提高了33.85分，平均G2提高了13.83分，同时保持了每对的G0。代码和基准文档可在https://github.com/MisterBrookT/finreportbench获取。

View on arXiv Download PDF AI Translation

cs.CL / 22 / 2608.04390

EdgeLM: Edge Demonstrations for Language Models' Table Understanding

EdgeLM：语言模型表格理解的边缘示例

Omidvartehrani, Soroush, Habibollah, Mohammadamin, Daviran, Mohammadreza, Rafiei, Davood

Abstract

Large language models (LLMs) perform table-centric prediction through in-context learning, making demonstration selection critical to performance. Existing retrieval methods prioritize similarity to the query, but similar demonstrations often reinforce the model's likely prediction rather than reveal the distinctions needed for difficult decisions. We propose EdgeLM, a retrieval framework that instead selects edge evidence, demonstrations that are both relevant to the query and informative about the decision boundary. EdgeLM retrieves two complementary forms of edge evidence by selecting data edges, nearby examples with different ground-truth labels, and model edges, similar examples previously misclassified by the deployed model. EdgeLM requires neither model retraining nor task-specific engineering. Across five data wrangling tasks, fifteen datasets, and five open-weight and proprietary LLMs, EdgeLM consistently achieves the best or near-best performance in every setting, while ablations show that the two forms of edge evidence provide complementary benefits. Our code and datasets are publicly available at https://github.com/soroushomidvar/EdgeLM.

Chinese Translation

大型语言模型（LLMs）通过上下文学习进行以表格为中心的预测，因此示例选择对性能至关重要。现有的检索方法优先考虑与查询的相似性，但相似的示例往往强化模型的可能预测，而不是揭示在困难决策中所需的区别。我们提出了EdgeLM，一个检索框架，它选择边缘证据，即与查询相关且对决策边界提供信息的示例。EdgeLM通过选择数据边缘（具有不同真实标签的邻近示例）和模型边缘（先前被部署模型错误分类的相似示例）来检索两种互补形式的边缘证据。EdgeLM不需要模型重训练或特定任务的工程。在五个数据处理任务、十五个数据集以及五个开放权重和专有LLM的实验中，EdgeLM在每种设置中始终实现最佳或接近最佳的性能，而消融实验表明这两种边缘证据提供了互补的好处。我们的代码和数据集可在 https://github.com/soroushomidvar/EdgeLM 上公开获取。

View on arXiv Download PDF AI Translation

cs.CL / 23 / 2608.04397

NOLLI: A Difficulty-Calibrated Puzzle Benchmark for Diagnosing the English-Korean Performance Gap

NOLLI：用于诊断英韩表现差距的难度校准拼图基准

Choi, Dasol, Park, Joonyong, Yu, Daegon, Kim, Soo Yong, Song, Youngsook, Hong, Seunghyeok

Abstract

We introduce NOLLI, a procedurally generated English-Korean puzzle benchmark designed to diagnose where Korean performance gaps arise. It comprises 15 puzzle types (25 tasks; 7,500 items), with every instance seed-regenerable, verified to have a unique solution, and scored deterministically. Rather than equating harder with bigger, we calibrate difficulty behaviorally, tuning each generator until a fixed reference model lands in target accuracy bands. Its three-level design combines matched direct translations, script adaptations over Hangul jamo (sub-syllabic letters), and Korean-only tasks grounded in Korean culture or orthography. We evaluate 15 frontier, open-weight, and Korean-developed models; among the 12 above a 3% overall-accuracy floor, matched English-Korean accuracy is statistically equivalent within a +/- 10 pp margin (TOST), suggesting little cost from presentation language alone. Writing-system-intensive tasks show sharper gaps: Korean Cipher falls behind English by up to 68.7 pp, whereas Cryptarithmetic over the same jamo shows no systematic penalty, and Jamo Composition accuracy predicts Korean Cipher accuracy. These contrasts are diagnostic rather than causal, consistent with difficulty in multi-step sub-syllabic execution. Korean-only tasks separate rule-application deficits, which vary in sign, from a Kinship deficit positive in all 12. Finally, a salient size measure fails to grow from Easy to Hard in 7 of 15 types, making structural size an unreliable proxy for empirical difficulty.

Chinese Translation

我们介绍了NOLLI，这是一个程序生成的英韩拼图基准，旨在诊断韩国表现差距的产生位置。该基准包含15种拼图类型（25个任务；7,500个项目），每个实例均可通过种子重新生成，经过验证具有唯一解，并且得分是确定性的。我们并不将更难等同于更大，而是通过行为校准难度，调节每个生成器，直到固定参考模型落入目标准确率范围。其三层设计结合了匹配的直接翻译、基于韩文音节字母（子音节字母）的脚本适配，以及扎根于韩国文化或正字法的仅限韩语任务。我们评估了15个前沿、开放权重和韩国开发的模型；在12个整体准确率超过3%的模型中，匹配的英韩准确率在统计上等价，误差范围为+/- 10个百分点（TOST），这表明仅凭展示语言几乎没有成本。以书写系统为重点的任务显示出更明显的差距：韩国密码的表现比英语低68.7个百分点，而在相同音节字母上的密码算术没有系统性惩罚，音节字母组合的准确率可以预测韩国密码的准确率。这些对比是诊断性的而非因果性的，与多步骤子音节执行中的困难一致。仅限韩语的任务将规则应用缺陷（其符号变化）与在所有12个模型中均为正的亲属缺陷区分开来。最后，一个显著的规模测量在15种类型中的7种中未能从简单到困难增长，使得结构规模成为经验难度的不可靠代理。

View on arXiv Download PDF AI Translation

cs.CL / 24 / 2608.04415

Social Pressure Breaks Majority Voting in LLM Safety Panels

社会压力破坏大型语言模型安全小组的多数投票

Hu, Yibo, Qu, Jiaming

Abstract

Large language models (LLMs) are increasingly used to detect unsafe content. A common approach is to combine judgments from a panel of models to correct individual mistakes, but this benefit may disappear when every model sees the same misleading context before voting. We study this risk in a controlled two-round experiment. Each model first judges an item alone, then judges it again after six simulated peers either assert the wrong label or abstain. We combine the final judgments by majority vote. Across six open-weight LLMs and six datasets, we find that the wrong-label peer message raises the average reviewer false-alarm rate from 56.5% under silent peers to 87.5%, and majority voting raises the panel false-alarm rate to 100%. Without an asserted label, the same panel outperforms its average member. The effect is strongly asymmetric: reviewers follow pushes toward "unsafe" far more than pushes toward "safe" (about 75% versus 17%), so the panel's false-alarm rate rises sharply while its harmful-miss rate changes little. The proprietary-model probe shows substantial variation across models. These results identify susceptibility to shared social cues as a failure mode of safety panels and provide a simple pre-deployment diagnostic.

Chinese Translation

大型语言模型（LLMs）越来越多地用于检测不安全内容。一种常见的方法是结合多个模型的判断，以纠正个体错误，但当每个模型在投票前看到相同的误导性上下文时，这种好处可能会消失。我们在一个受控的两轮实验中研究了这一风险。每个模型首先单独判断一个项目，然后在六个模拟同行要么错误地断言标签，要么选择不表态后再次判断。我们通过多数投票结合最终判断。在六个开放权重的LLM和六个数据集中，我们发现错误标签的同行信息将平均审稿人的误报率从沉默同行下的56.5%提高到87.5%，而多数投票将小组的误报率提高到100%。在没有明确标签的情况下，同一小组的表现优于其平均成员。该效应具有明显的不对称性：审稿人更倾向于跟随向“危险”方向的推动，而不是向“安全”方向的推动（约75%对17%），因此小组的误报率急剧上升，而其有害漏报率变化不大。专有模型探测显示不同模型之间存在显著差异。这些结果识别出对共享社会线索的敏感性作为安全小组的一种失效模式，并提供了一种简单的部署前诊断方法。

View on arXiv Download PDF AI Translation

cs.CL / 25 / 2608.04433

MERaLiON-GR: Speech Gender Recognition Model for English and SEA Languages

MERaLiON-GR：英语和东南亚语言的语音性别识别模型

Wang, Qiongqiong, Aw, Ai Ti, Chen, Nancy F., Chiu, Ying Lay, Ding, Yang, He, Yingxu, Jiang, Ridong, Liu, Zhuohan, Lu, Yanfeng, Ma, Yi, Huzaifah, Muhammad, Johan, Nabilah Binte Md, Lertcheva, Nattadaporn, Duc, Pham Minh, Bhupendra, Sailor Hardik, Salleh, Siti Umairah Binte Mohammad, Sun, Shuo, Vangani, Tarun Kumar, Wong, Jeremy H. M., Wu, Jinyang, Zhang, Longyin

Abstract

We present MERaLiON-GR, a speech gender recognition system that performs binary classification (female / male) on English and Southeast Asian (SEA) languages. The model finetunes MERaLiON-SpeechEncoder-2, a large conformer based transformer pre-trained on a broad speech corpus, and applies parameter efficient fine-tuning via Low-Rank Adaptation (LoRA) to adapt the encoder to the gender recognition task, and appends a multi-scale ECAPA-TDNN down stream network with attention pooling and a lightweight linear classifier. Extensive evaluations across multilingual Singaporean and Southeast Asian languages (English, Chinese, Malay, Tamil, Thai, Vietnamese, Indonesian, and Khmer) show that MERaLiON-GR consistently surpasses the state-of-the-art gender recognition model Vox-Profile and a large Audio-LLM, in both full-utterance and segment level evaluation modes. The results underscore the value of dedicated speech models in achieving accurate paralinguistic understanding and strong cross-lingual generalization.

Chinese Translation

我们提出了MERaLiON-GR，这是一种语音性别识别系统，能够对英语和东南亚（SEA）语言进行二分类（女性/男性）。该模型对MERaLiON-SpeechEncoder-2进行了微调，这是一个基于大规模语音语料库预训练的Conformer结构的变换器，并通过低秩适应（Low-Rank Adaptation, LoRA）进行参数高效微调，以将编码器适应于性别识别任务，并附加了一个多尺度的ECAPA-TDNN下游网络，结合注意力池化和轻量级线性分类器。对新加坡和东南亚多语言（英语、中文、马来语、泰米尔语、泰语、越南语、印尼语和高棉语）进行的广泛评估表明，MERaLiON-GR在完整话语和片段级评估模式下始终超越了最先进的性别识别模型Vox-Profile和大型音频大语言模型（Audio-LLM）。这些结果强调了专用语音模型在实现准确的副语言理解和强大的跨语言泛化方面的价值。

View on arXiv Download PDF AI Translation

cs.CL / 26 / 2608.04444

D$^2$F-ReAG: Dynamic Decomposition and Filtering for Multi-Hop Reasoning-Augmented Generation

D$^2$F-ReAG：用于多跳推理增强生成的动态分解与过滤

Li, Jiaoyang, Ruan, Junhao, Tang, Shengwei, Chang, Kaiyan, Yu, Zhengtao, Xiao, Tong, Zhu, Jingbo

Abstract

Large language models (LLMs) often generate inaccurate answers due to their reliance on static internal knowledge. Retrieval-augmented generation (RAG) addresses this limitation by integrating external knowledge and excelling at single-hop queries. However, it struggles with multi-hop questions that require cross-document reasoning. Existing methods, such as graph structured RAG or question decomposition, often lack dynamic decomposition and effective filtering, which leads to lower efficiency and accuracy. To overcome these limitations, we propose Dynamic Decomposition and Filtering for Multi-Hop Reasoning-Augmented Generation (D2F-ReAG), a novel paradigm that adaptively controls reasoning depth by judging the reliability of the root-level reasoning. If the root reasoning is reliable, the model directly generates the answer. Otherwise, the question is logically decomposed into sub-questions, and the verified reasoning derived from these sub-questions is used to refine the root reasoning. Experiments on three multi-hop benchmarks demonstrate the effectiveness of our method in handling complex multi-hop questions.

Chinese Translation

大型语言模型（LLMs）由于依赖静态内部知识，常常生成不准确的答案。检索增强生成（RAG）通过整合外部知识来解决这一局限性，并在单跳查询中表现出色。然而，它在需要跨文档推理的多跳问题上表现不佳。现有方法，如图结构RAG或问题分解，往往缺乏动态分解和有效过滤，导致效率和准确性降低。为克服这些局限性，我们提出了用于多跳推理增强生成的动态分解与过滤（D2F-ReAG），这是一种新颖的范式，通过判断根层推理的可靠性来自适应控制推理深度。如果根推理是可靠的，模型将直接生成答案。否则，问题将被逻辑分解为子问题，并利用从这些子问题中验证的推理来细化根推理。在三个多跳基准上的实验表明，我们的方法在处理复杂多跳问题时的有效性。

View on arXiv Download PDF AI Translation

cs.CL / 27 / 2608.04463

The Evaluator Is Part of the Experiment: Measuring Open-Ended LLM Conformity

评估者是实验的一部分：测量开放式大型语言模型的一致性

Guerra, Alicia, Hu, Yibo

Abstract

Prior work on LLM conformity largely measures discrete answer flips under verifiable labels. Open-ended revisions require a different measurement strategy because answer quality is graded, latent, and judged imperfectly. We introduce an experimental protocol implemented across a pooled main peer-condition corpus and separately constructed decomposition corpora, allowing us to separate ordinary re-answering, candidate-content exposure, a bundled peer-presentation residual, and directional judge sensitivity to visible peer context. Across four open-weight generators and three benchmarks, all-wrong peer input produces the lowest-quality revisions in every generator-dataset cell. Blind and informed ratings of identical answers also differ by evaluator: one judge shifts toward the peer-endorsed position, two shift away, one is approximately neutral, and GPT-4o and GPT-5.4-mini audits are likewise non-neutral. Finally, an anchor audit shows that terse correct anchors can be misread often enough to destabilize the latent scale unless calibration is checked explicitly. These results support four conclusions: flip rates are insufficient as a complete measure of open-ended conformity, wrong peers harm open-ended revision, evaluators are not neutral, and anchor calibration is necessary.

Chinese Translation

先前关于大型语言模型（LLM）一致性的研究主要测量在可验证标签下的离散答案翻转。开放式修订需要不同的测量策略，因为答案质量是分级的、潜在的，并且评判不完美。我们引入了一种实验协议，该协议在一个汇总的主要同行条件语料库和单独构建的分解语料库中实施，使我们能够分离普通的重新回答、候选内容曝光、捆绑的同行呈现残差，以及评估者对可见同行上下文的方向性敏感性。在四个开放权重生成器和三个基准测试中，所有错误的同行输入在每个生成器-数据集单元中产生最低质量的修订。盲评和知情评估对相同答案的评分也因评估者而异：一位评估者倾向于同行支持的位置，两位评估者则倾向于远离，另一位评估者大致中立，而GPT-4o和GPT-5.4-mini的审计结果同样不中立。最后，锚点审计表明，简洁的正确锚点可能会被误读，足以使潜在量表不稳定，除非明确检查校准。这些结果支持四个结论：翻转率不足以作为开放式一致性的完整测量，错误的同行会损害开放式修订，评估者并非中立，且锚点校准是必要的。

View on arXiv Download PDF AI Translation

cs.CL / 28 / 2608.04488

Energy- and Memory-Efficient PEFT Methods for Personalized On-Device SLMs on Consumer GPUs

面向消费者GPU的个性化设备上小型语言模型的节能与内存高效PEFT方法

Akhmetzhanov, Kuanysh, Park, Jurn-Gyu

Abstract

Despite rapid advances in large language models (LLMs), deploying and personalizing them on resource-constrained devices remains impractical due to high VRAM, time, and energy costs. Parameter-Efficient Fine-Tuning (PEFT) of Small Language Models (SLMs) offers a promising alternative, yet few studies compare PEFT methods across architectures using both general and personalization benchmarks while accounting for energy consumption. We compare five fine-tuning approaches (Full Fine-Tuning, LoRA, LoRA+, QLoRA, and BitFit) on four SLMs from two families (Transformer-based: TinyLlama-1.1B, Qwen3-1.7B; SSM-based: Mamba-1.4B, Mamba-2-1.3B) across three GLUE tasks (SST-2, QNLI, STS-B) and three LaMP personalization tasks (LaMP-1, LaMP-2, LaMP-3). Each configuration is evaluated with the energy-focused NetScore-E and the memory-focused NetScore-M, the two variants that reflect the constraints binding on-device deployment. Methods are selected with a strict energy-first rule (highest NetScore-E, ties broken by NetScore#). LoRA+ achieves the highest NetScore-E in 19 of 24 configurations and the highest NetScore-M in 13 of 24, and is the selected method in 18 of 24. QLoRA, available only for the Transformer models, cuts peak finetuning VRAM by up to 3.9x relative to LoRA and therefore takes the best NetScore-M in 5 of the 12 Transformer configurations, although its de-quantization overhead leaves it selected in only one of them once energy decides. BitFit and full fine-tuning are almost never competitive on either variant, and TinyLlama-1.1B leads the energy-focused NetScore-E on five of the six benchmarks and the memory-focused NetScore-M on four. These results show that compact SLMs paired with PEFT provide a practical, energy-aware path to personalized on-device deployment, with the optimal method set by the dominant constraint: LoRA+ for energy and QLoRA for memory.

Chinese Translation

尽管大型语言模型（LLMs）迅速发展，但在资源受限设备上部署和个性化它们仍然不切实际，原因在于高显存、时间和能耗成本。小型语言模型（SLMs）的参数高效微调（PEFT）提供了一种有前景的替代方案，但很少有研究在考虑能耗的情况下，比较不同架构下的PEFT方法，使用通用和个性化基准进行评估。我们比较了五种微调方法（全微调、LoRA、LoRA+、QLoRA和BitFit），在两个家族的四个SLM（基于Transformer的：TinyLlama-1.1B、Qwen3-1.7B；基于SSM的：Mamba-1.4B、Mamba-2-1.3B）上进行评估，涵盖三个GLUE任务（SST-2、QNLI、STS-B）和三个LaMP个性化任务（LaMP-1、LaMP-2、LaMP-3）。每种配置都通过以能耗为重点的NetScore-E和以内存为重点的NetScore-M进行评估，这两个变体反映了设备上部署的约束。方法的选择遵循严格的能耗优先原则（最高NetScore-E，平局时按NetScore#打破）。LoRA+在24个配置中有19个获得最高NetScore-E，在24个配置中有13个获得最高NetScore-M，并且在24个配置中被选为方法的次数为18次。QLoRA仅适用于Transformer模型，相比于LoRA，其峰值微调显存减少了最多3.9倍，因此在12个Transformer配置中获得了5个最佳NetScore-M，尽管其去量化开销使其在能耗决定后仅在其中一个配置中被选中。BitFit和全微调几乎在任何变体上都没有竞争力，而TinyLlama-1.1B在六个基准中有五个获得了以能耗为重点的NetScore-E，在四个基准中获得了以内存为重点的NetScore-M。这些结果表明，结合PEFT的紧凑型SLM为个性化设备部署提供了一条实用的、关注能耗的路径，最佳方法由主导约束决定：在能耗方面选择LoRA+，在内存方面选择QLoRA。

View on arXiv Download PDF AI Translation

cs.CL / 29 / 2608.04505

K-EXAONE 2.0 Technical Report

K-EXAONE 2.0 技术报告

Choi, Eunbi, Choi, Kibong, Chun, Sehyun, Hong, Seokhee, Hwang, Junwon, Jeon, Hyojin, Jo, Ahra, Jo, Hyunjik, Jo, Yeonsik, Jung, Minhyeok, Kim, Doyoung, Kim, Heegyu, Kim, Joonkee, Kim, Seonghwan, Kim, Soyeon, Kim, Sunkyoung, Kim, Yireun, Kim, Yongil, Ko, Byungoh, Lee, Changhun, Lee, Dohaeng, Lee, Haeju, Lee, Jinsik, Lee, Kyungmin, Lee, Minwoo, Lee, Wonkee, Park, Sangha, Park, Sungjune, Ryoo, Kwangrok, Seo, Kijung, Seo, Minju, Song, Yongwoo, Yang, Sejong, Yeen, Heuiyeen, Choi, Stanley Jungkyu, Choi, Yemuk, Chun, Yongchan, Ham, Jiwon, Hong, Dasol, Im, Sujeong, Jeon, Kijeong, Jo, Gerrard Jeongwon, Jo, Hyeongjun, Jo, Yujin, Jung, Jiyeon, Kang, Naeun, Kim, Daeseong, Kim, Euisoon, Kim, Hayeon, Kim, Hyosang, Kim, Myoungshin, Kim, Unsol, Kim, Youchul, Lee, Chaeeun, Lee, ChaeYoon, Lee, Edward Hwayoung, Lee, Honglak, Lee, Hwansoo, Lee, Minkyung, Lee, Sangeun, Lim, Solji, Lim, Woohyung, Moon, Chanwoo, Mun, Jueun, Park, Jimin, Park, Seojeong, Park, Yongmin, Seo, Hyerin, Shin, Donghyeon, Son, Donghyun, Son, Eunyong, Um, Kaehyun, Yang, Sihoon, Yea, Chang En, Yi, Sihyuk, Yoo, Kyungjae, Yoon, Chansik

Abstract

This technical report presents K-EXAONE 2.0, an open-weight multilingual foundation model developed by LG AI Research as a step in our effort toward global frontier-scale foundation models. Rather than training from scratch, we upcycle K-EXAONE and expand its architecture, yielding a Mixture-of-Experts (MoE) model with 750B total parameters and approximately 37B activated per token---more than three times the capacity of its predecessor. K-EXAONE 2.0 supports context lengths of up to 256K tokens and expands multilingual coverage from six to ten languages. Its training pipeline combines continual pre-training, difficulty-focused mid-training, and post-training to strengthen reasoning, agentic coding, multilingual capability, and safety grounded in Korean sociocultural contexts. Across nine evaluation categories selected to reflect the conditions of practical use, K-EXAONE 2.0 improves over K-EXAONE and remains competitive with open-weight models, showing its largest gains in agentic coding and long-context understanding and its clearest strengths in long-context retrieval and safety. Released under the Apache 2.0 license, K-EXAONE 2.0 enables the wider AI ecosystem to evaluate, deploy, adapt, and build upon it, while marking the beginning---rather than the endpoint---of our challenge toward the global frontier.

Chinese Translation

本技术报告介绍了 K-EXAONE 2.0，这是由 LG AI Research 开发的开放权重多语言基础模型，作为我们向全球前沿规模基础模型努力的一步。我们并非从头开始训练，而是对 K-EXAONE 进行了升级，并扩展了其架构，形成了一个具有 7500 亿总参数的专家混合模型（Mixture-of-Experts, MoE），每个令牌激活约 370 亿参数——是其前身的三倍多。K-EXAONE 2.0 支持最长 256K 令牌的上下文长度，并将多语言覆盖范围从六种语言扩展到十种语言。其训练流程结合了持续预训练、以难度为中心的中期训练和后期训练，以增强推理能力、主动编码、多语言能力以及基于韩国社会文化背景的安全性。在九个评估类别中，K-EXAONE 2.0 在反映实际使用条件的情况下，相较于 K-EXAONE 有明显改善，并与开放权重模型保持竞争力，尤其在主动编码和长上下文理解方面取得了最大的进步，并在长上下文检索和安全性方面展现出明显优势。K-EXAONE 2.0 在 Apache 2.0 许可证下发布，使更广泛的 AI 生态系统能够评估、部署、适应和基于此进行构建，同时标志着我们向全球前沿挑战的开始——而非终点。

View on arXiv Download PDF AI Translation

cs.CL / 30 / 2608.04514

RESPClinBench: Benchmarking Multimodal Clinical Decision-Making and Longitudinal Disease Management in Respiratory Specialty Care

RESPClinBench：呼吸专科护理中多模态临床决策和纵向疾病管理的基准测试

Bian, Mouxiao, Chen, Zhi, Chen, Ruiyao, Lu, Lu, Liang, Hengrui, Huang, Chaoyi, Lin, Yiluo, Ding, Jingru, Zhong, Yun, Su, Yuming, Xu, Jie

Abstract

Background: Respiratory specialty care requires multimodal interpretation, longitudinal risk assessment, guideline-concordant intervention, and whole-course management, which are poorly represented by examination-oriented medical benchmarks. Objective: To develop RESPClinBench, a real-world scenario-based benchmark for respiratory clinical decision-making, and evaluate seven contemporary large language models across AECOPD-PIM and PNBIM. Methods: RESPClinBench cases were adapted from de-identified respiratory clinical data. Three attending-level respiratory physicians revised cases, reference answers, and atomic clinical-action points, while one senior respiratory specialist performed cross-review and final adjudication. AECOPD-PIM comprised 427 open-ended COPD cases, and PNBIM comprised 196 multimodal pulmonary nodule cases combining chest CT with structured clinical information. Seven models generated 4,361 responses through standardized API inference with temperature 0 and a maximum output length of 8192 tokens. An automated framework calculated the final score as the arithmetic mean of atomic-action recall and rubric-based LLM-as-a-Judge assessment. Results: Across 623 cases, the mean final score was 68.58. Qwen3.6-27B ranked first overall at 71.22, Qwen3.5-397B-A17B led PNBIM at 72.48, and Qwen3.6-27B led AECOPD-PIM at 71.11. Imaging hallucination and serious medical risk occurred in 31.85% and 8.16% of PNBIM responses; medication-safety risk and serious medical risk occurred in 26.93% and 1.44% of AECOPD-PIM responses. Conclusions: RESPClinBench identifies task-specific limitations in multimodal pulmonary nodule assessment and longitudinal COPD management. Combining explicit clinical-action coverage, holistic evaluation, and independent safety flags provides a clinically grounded basis for model selection and prospective validation.

Chinese Translation

背景：呼吸专科护理需要多模态解读、纵向风险评估、符合指南的干预和全程管理，而这些在以考试为导向的医学基准中表现不佳。目的：开发RESPClinBench，一个基于真实场景的呼吸临床决策基准，并评估七种当代大型语言模型在AECOPD-PIM和PNBIM中的表现。方法：RESPClinBench案例改编自去标识化的呼吸临床数据。三位主治级呼吸医师修订了案例、参考答案和原子临床行动点，而一位高级呼吸专家进行了交叉审查和最终裁定。AECOPD-PIM包含427个开放式COPD案例，PNBIM包含196个多模态肺结节案例，结合了胸部CT和结构化临床信息。七个模型通过标准化API推理生成了4361个响应，温度设置为0，最大输出长度为8192个标记。一个自动化框架计算最终得分，作为原子行动召回率和基于评分标准的LLM作为评审的算术平均值。结果：在623个案例中，平均最终得分为68.58。Qwen3.6-27B总体排名第一，得分为71.22，Qwen3.5-397B-A17B在PNBIM中领先，得分为72.48，Qwen3.6-27B在AECOPD-PIM中领先，得分为71.11。PNBIM响应中出现影像幻觉和严重医疗风险的比例分别为31.85%和8.16%；AECOPD-PIM响应中出现药物安全风险和严重医疗风险的比例分别为26.93%和1.44%。结论：RESPClinBench识别了多模态肺结节评估和纵向COPD管理中的任务特定限制。结合明确的临床行动覆盖、整体评估和独立的安全标记，为模型选择和前瞻性验证提供了临床基础。

View on arXiv Download PDF AI Translation

cs.CL / 31 / 2608.04524

ODRA: Synthesizing Cognitive Behavioral Therapy Sessions with Structured Chain-Of-Thought and Dynamic Patient Resistance

ODRA：利用结构化思维链和动态患者抵抗合成认知行为疗法会话

Rodriguez-Juan, Javier, Arnaout, Hiba, Garcia-Rodriguez, Jose, Tomás, David, Gurevych, Iryna

Abstract

Synthetic generation of Cognitive Behavioral Therapy (CBT) sessions is challenged by two competing demands: adhering to strict therapeutic structure while modeling the resistant, unpredictable behavior of real patients. Existing script-based methods fail to capture dynamic therapeutic interactions, while multi-agent approaches struggle to adhere to CBT's sequential structure; both suffer from sycophancy, producing overly compliant patients that misrepresent real clinical settings. In this work we introduce ODRA, a novel framework for synthesizing therapy dialogues through a Chain-of-Thought (CoT) strategy grounded in foundational CBT guidelines (Beck, 2020). ODRA further incorporates a resistance orchestrator to solve patient sycophancy, which employs steering techniques to elicit behaviors aligned with their resistance level. Automated and expert evaluations show that ODRA significantly outperforms existing methods across therapeutic skills, CBT alignment, and patient behavioral fidelity, with licensed psychologists preferring ODRA sessions across 12 of 13 clinical metrics. Furthermore, models fine-tuned on our dataset demonstrate superior therapeutic performance against both cooperative and resistant patients, validating that explicit resistance modeling in synthetic training data directly translates to downstream clinical robustness.

Chinese Translation

合成认知行为疗法（CBT）会话面临两个相互竞争的需求：在遵循严格的治疗结构的同时，模拟真实患者的抵抗和不可预测行为。现有的基于脚本的方法无法捕捉动态的治疗互动，而多智能体方法则难以遵循CBT的顺序结构；两者都存在迎合性的问题，产生过于顺从的患者，无法真实反映临床环境。在本研究中，我们介绍了ODRA，一个通过基于基础CBT指南（Beck, 2020）的思维链（CoT）策略合成治疗对话的新框架。ODRA进一步结合了抵抗协调器，以解决患者的迎合性问题，该协调器采用引导技术以引发与患者抵抗水平相一致的行为。自动化和专家评估表明，ODRA在治疗技能、CBT一致性和患者行为忠实度方面显著优于现有方法，持证心理学家在13个临床指标中有12个指标更倾向于ODRA会话。此外，在我们的数据集上微调的模型在对待合作和抵抗患者时表现出更优的治疗效果，验证了在合成训练数据中显式建模抵抗直接转化为下游临床稳健性。

View on arXiv Download PDF AI Translation

cs.CL / 32 / 2608.04549

EuroExec: Frontier Language Models Fall Short of Expert Judgment on European Executive Decision Tasks

EuroExec：前沿语言模型在欧洲执行决策任务中的表现不及专家判断

Arnal, Pau, Denfir, Khaled, Smahliuk, Danylo, Avhad, Amrut, Castro, Marcus A.

Abstract

Frontier LLMs are increasingly put to use on open-ended complex questions, different in nature from the ones they are typically evaluated on. We dedicate more than 4,000 human expert hours to evaluate a selection of six frontier LLMs on a member of this class of problems: EuroExec, our introduced human expert-based benchmark composed of 413 open-ended long-form European executive tasks authored by 47 vetted domain experts, each question drawn from experience in a real case. Every response is manually evaluated through a multi-attribute rubric, an item-specific checklist of requirements, and a preference rank ordering, extracting an aggregate metric "Solve Rate". The strongest model solves only 56.9% of tasks, while expert-written reference answers judged blindly are solved at near-ceiling levels and are preferred over every model response in 74% of direct rankings, placing frontier generative systems well below the professional standard of work they are already used for. We see that the best way to extract this kind of conclusion is by employing human evaluators, carefully checking their consistency through rigorous statistical analysis, and observe that automatic measurements also fall short when evaluating on this case of real-world open-ended problems with a subjective ground truth.

Chinese Translation

前沿大型语言模型（LLMs）越来越多地被应用于开放式复杂问题，这些问题的性质与它们通常被评估的内容不同。我们投入超过4000小时的人类专家时间，对六个前沿LLMs在这一类问题上的表现进行评估：EuroExec，这是我们提出的基于人类专家的基准，由47位经过审查的领域专家撰写的413个开放式长篇欧洲执行任务组成，每个问题均源于真实案例的经验。每个响应都通过多属性评分标准、特定项目的要求清单以及偏好排名进行手动评估，提取出一个综合指标“解决率”。最强的模型仅解决了56.9%的任务，而专家撰写的参考答案在盲评中几乎达到了上限水平，并在74%的直接排名中优于每个模型的响应，这使得前沿生成系统的表现远低于它们已经被用于的专业工作标准。我们发现，得出这种结论的最佳方法是采用人类评估者，通过严格的统计分析仔细检查他们的一致性，并观察到在评估真实世界开放式问题时，自动测量也显得不足，因为这些问题具有主观的真实标准。

View on arXiv Download PDF AI Translation

cs.CL / 33 / 2608.04552

Relational Response Fields: A General Theory of Black-Box LLM Response Consistency and Recovery

关系响应场：黑箱大型语言模型响应一致性与恢复的通用理论

Zichen, Song

Abstract

Black-box language-model reliability is commonly pursued by sampling, prompting, voting, verifying, or iteratively revising individual answers. We ask a prior question: \emph{what determines whether a collection of black-box responses is recoverable at all?} We represent responses to typed transformations of a query as a \emph{relational response field} (RRF). Edge transports encode how valid responses must change under paraphrase, scaling, decomposition, refactoring, or other task symmetries; anchors encode independently trusted evidence such as execution or a verifier. For relation operator $D$, anchor operator $A$, and at most $k$ corrupted response nodes, we identify $\gamma_k(D,A)$ as the intrinsic difficulty of black-box response recovery. It is positive exactly when every $k$-node corruption is identifiable; it gives a deterministic stability bound proportional to $1/\gamma_k$; and a matching two-point minimax lower bound shows that no estimator can improve this dependence. Thus consistency is not truth: relation-only methods are blind to null directions, including shared hallucinations. We derive sparse field-repair algorithms while separating information-theoretic identifiability from the stronger null-space conditions required by convex optimization. Controlled theorem tests and black-box mathematics/code experiments evaluate four theory-fixed consequences: consistency--truth separation, anchor phase transitions, redundancy saturation, and cross-model, cross-task prediction of repair difficulty. The results support $\gamma_k(D,A)$ as a measurable property of a response-recovery instance, rather than a score attached to one repair heuristic.

Chinese Translation

黑箱语言模型的可靠性通常通过采样、提示、投票、验证或迭代修订单个答案来追求。我们提出一个先前的问题： extit{是什么决定了一组黑箱响应是否可以恢复？} 我们将对查询的类型转换的响应表示为 extit{关系响应场}（RRF）。边缘传输编码了在意译、缩放、分解、重构或其他任务对称性下有效响应必须如何变化；锚点编码了独立可信的证据，如执行或验证者。对于关系运算符 $D$、锚点运算符 $A$，以及至多 $k$ 个损坏的响应节点，我们识别 $eta_k(D,A)$ 作为黑箱响应恢复的内在难度。当每个 $k$ 节点的损坏都可识别时，它是正值；它提供了一个与 $1/eta_k$ 成正比的确定性稳定性界限；一个匹配的两点最小最大下界表明没有估计器可以改善这种依赖关系。因此，一致性并不等于真实：仅基于关系的方法对零方向是盲目的，包括共享的幻觉。我们推导出稀疏场修复算法，同时将信息论可识别性与凸优化所需的更强的零空间条件区分开来。受控定理测试和黑箱数学/代码实验评估了四个理论固定的后果：一致性与真实的分离、锚点相变、冗余饱和和跨模型、跨任务的修复难度预测。结果支持 $eta_k(D,A)$ 作为响应恢复实例的可测属性，而不是附加于某一修复启发式的分数。

View on arXiv Download PDF AI Translation

cs.CL / 34 / 2608.04554

Representing Visual Evidence for Item Difficulty Prediction: Visual Textualization and Image-Native Modeling

视觉证据在题目难度预测中的表示：视觉文本化与图像原生建模

Chen, Han, Li, Ming, Jiao, Hong, Zhou, Tianyi

Abstract

Predicting item difficulty from content can provide an initial estimate for newly developed questions before sufficient student responses are available. Existing approaches typically represent the question stem and answer choices as text. When mathematics items contain visual components, a common pipeline first textualizes that evidence and then applies a text predictor. We ask: how should visual evidence be represented for item difficulty prediction? We compare question text alone, visual textualization, which expresses visual evidence in language, and image-native modeling, which retains the original image. Using Eedi items with difficulty calibrated from student responses, we train large language models (LLMs) and vision-language models (VLMs) directly for difficulty regression. Both visual interfaces achieve the lowest point estimates, although the leading systems cannot be reliably ordered. Open-VLM textualization yields lower RMSE point estimates for all evaluated LLMs, while broader adaptation does so for all image-native VLMs. Test-time interventions show dependence on the paired full-item image, but do not isolate the additional visual component. The two visual interfaces also make partially complementary item-level errors and differ substantially in computational workflow. Thus, textualization should not be treated as the only practical interface: image-native modeling is a competitive alternative whose effectiveness depends on how the VLM is adapted.

Chinese Translation

从内容预测题目难度可以在新开发的问题获得足够的学生反馈之前提供初步估计。现有的方法通常将题干和答案选项表示为文本。当数学题目包含视觉组件时，常见的流程首先将这些证据文本化，然后应用文本预测器。我们提出：如何表示视觉证据以进行题目难度预测？我们比较了仅使用题目文本、将视觉证据以语言表达的视觉文本化，以及保留原始图像的图像原生建模。使用从学生反馈中校准的难度的Eedi题目，我们直接训练大型语言模型（LLMs）和视觉-语言模型（VLMs）进行难度回归。两种视觉接口都实现了最低的点估计，尽管领先的系统无法可靠排序。Open-VLM文本化为所有评估的LLMs提供了更低的均方根误差（RMSE）点估计，而更广泛的适应性则为所有图像原生VLMs提供了这样的结果。测试时的干预显示出对配对完整题目图像的依赖，但并未孤立出额外的视觉组件。这两种视觉接口在题目级别上也会产生部分互补的错误，并在计算工作流程上有显著差异。因此，文本化不应被视为唯一的实用接口：图像原生建模是一种竞争性的替代方案，其有效性取决于VLM的适应方式。

View on arXiv Download PDF AI Translation

cs.CL / 35 / 2608.04567

STRIVE: Probing Reasoning Limits in Graded Plausibility Generation and Evaluation

STRIVE：探究分级可信度生成与评估中的推理极限

Baghel, Bhiman Kumar, Chrabaszcz, Anna, Warren, Tessa, Dickey, Michael Walsh, Dresang, Haley C., Li, Xiang Lorraine

Abstract

Event knowledge concerns who does what to whom. Psycholinguists use event-plausibility judgments to examine how this knowledge supports human language processing. To isolate plausibility effects, these studies require controlled event sets in which one event slot varies across plausibility levels while all other event features remain fixed. Constructing such sets manually is labor-intensive. We therefore introduce STRIVE, an LLM-based framework for jointly generating and evaluating controlled event sets crossing plausibility class (plausible vs. implausible) with intended classification difficulty (easy vs. hard). Given a verb, STRIVE constructs a shared event frame, then produces one event per condition by varying one slot while holding all others fixed. In experiments with six models across 60 verbs, GPT-5.1 produced high-quality sets only 16.7% of the time using the baseline generation prompt. Adding a global reasoning scratchpad and evaluator-guided refinement raised this rate to 75.0%. Greater reasoning effort also improved evaluator--human agreement. Nevertheless, events near the plausibility boundary remain most difficult. They elicit the greatest human disagreement, and the best evaluator reaches only 57% accuracy on the implausible-hard condition, indicating a need for human input. Overall, STRIVE offers a scalable approach to reducing manual effort by automating initial event-set generation and evaluation for psycholinguistic studies.

Chinese Translation

事件知识涉及谁对谁做了什么。心理语言学家使用事件可信度判断来研究这一知识如何支持人类语言处理。为了隔离可信度效应，这些研究需要控制事件集，其中一个事件槽在不同的可信度水平上变化，而所有其他事件特征保持不变。手动构建这样的事件集劳动强度大。因此，我们引入了STRIVE，一个基于大型语言模型（LLM）的框架，用于共同生成和评估跨越可信度类别（可信 vs. 不可信）和预期分类难度（简单 vs. 困难）的控制事件集。给定一个动词，STRIVE构建一个共享事件框架，然后通过改变一个槽而保持其他槽不变，生成每种条件下的一个事件。在对60个动词进行的六个模型实验中，GPT-5.1在使用基线生成提示时，仅以16.7%的频率生成高质量的事件集。添加全局推理草稿和评估者引导的优化将这一比例提高到75.0%。更大的推理努力也提高了评估者与人类之间的协议。然而，接近可信度边界的事件仍然是最困难的。它们引发了人类之间最大的分歧，而最佳评估者在不可信-困难条件下的准确率仅为57%，表明需要人类输入。总体而言，STRIVE提供了一种可扩展的方法，通过自动化初始事件集的生成和评估，减少心理语言学研究中的手动工作量。

View on arXiv Download PDF AI Translation

cs.CL / 36 / 2608.04569

Relevant but Incomplete: Referential Dangling as a Paradigm-Level Failure Mode in Hard Prompt Compression

相关但不完整：参照悬挂作为硬提示压缩中的范式级失败模式

Hu, Zhengpei, Li, Kai, Fu, Dapeng, Zou, Xuechao, Tang, Yuanhao, Li, Yue, Cao, Tengfei, Huang, Jianqiang

Abstract

Hard prompt compression reduces long-context inference cost by independently scoring tokens, sentences, or chunks and retaining the highest-scoring units under a budget. We identify a structural failure in this procedure: independent selection can split dependent evidence pairs, retaining one member while deleting the other. When retained text contains an answer but deleted text defines the entity needed to interpret it, we call the result referential dangling. At a compression ratio of 0.30, Beaver, which ranks coherent chunks using Qwen3-0.6B embeddings, leaves the answer path incomplete in 34-54% of bridge examples across three multi-hop question answering datasets. On a shared HotpotQA bridge set, all six hard compressors we test exhibit dangling at rates up to 60%, and every document in LongBench-v2 Single-Document QA contains at least one dangling reference. On dangling examples evaluated with Qwen3-8B, reinserting the missing supporting paragraph while removing nonsupporting paragraphs to maintain the token budget improves accuracy by 29-34 percentage points (p < 0.0001), recovering at least 88% of the gap to contexts retaining both supporting paragraphs. Stronger answer models do not absorb the loss: on MuSiQue, GPT-5.5 is 8.8 points less accurate on compressed contexts than on contexts retaining both supporting paragraphs. Finally, we train a compact classifier to rank omitted sentences by whether they are needed to interpret retained text and reinsert the top-ranked candidates without support annotations at inference. On HotpotQA with Qwen3-8B, this automatic restoration improves accuracy by 4.7 points while changing the compression ratio only from 0.30 to 0.31. Hard compressors should optimize both relevance and referential completeness.

Chinese Translation

硬提示压缩通过独立评分标记、句子或块，并在预算内保留得分最高的单元，从而降低长上下文推理成本。我们识别出这一过程中的结构性失败：独立选择可能会拆分依赖证据对，保留一个成员而删除另一个。当保留的文本包含答案但被删除的文本定义了解释答案所需的实体时，我们称这种结果为参照悬挂。在压缩比为0.30时，使用Qwen3-0.6B嵌入对连贯块进行排名的Beaver在三个多跳问答数据集中，34-54%的桥接示例中留下了答案路径不完整。在共享的HotpotQA桥接集上，我们测试的六种硬压缩器均表现出高达60%的悬挂率，而LongBench-v2单文档问答中的每个文档至少包含一个悬挂引用。在使用Qwen3-8B评估的悬挂示例中，重新插入缺失的支持段落，同时删除不支持的段落以保持标记预算，提高了29-34个百分点的准确率（p < 0.0001），恢复了至少88%的与保留两个支持段落的上下文之间的差距。更强的答案模型并未吸收这一损失：在MuSiQue上，GPT-5.5在压缩上下文中的准确率比在保留两个支持段落的上下文中低8.8个百分点。最后，我们训练了一个紧凑的分类器，根据被省略句子是否需要解释保留文本来对其进行排名，并在推理时重新插入排名最高的候选句子，而不需要支持注释。在HotpotQA上使用Qwen3-8B，这种自动恢复将准确率提高了4.7个百分点，同时压缩比仅从0.30变为0.31。硬压缩器应优化相关性和参照完整性。

View on arXiv Download PDF AI Translation

cs.CL / 37 / 2608.04570

The Personalization Mirage: How LLMs Fabricate User Profiles, and Why Self-Monitoring Misleads

个性化的幻影：大型语言模型如何虚构用户档案，以及自我监测为何会误导

Sun, Yushi, Zhang, Yanjie, Sheng, Rui

Abstract

Personalized LLMs with persistent memory are increasingly deployed, yet the faithfulness of their user models remains unexamined. We study over-inference (OI): the phenomenon where LLMs fabricate user attributes beyond what evidence supports. We introduce MirageBench, comprising 150 personas balanced across stereotypical, counter-stereotypical, and neutral profiles, 6 personalization tasks spanning an ``imagination gradient'', a four-way faithfulness taxonomy operationalized by an independent judge (validated against a blind human annotator on 400 claims: Cohen's kappa = 0.863 four-class, kappa = 0.900 binary), and a leaderboard of 12 models across 7 families on 143616 judged claims. We find that over-inference is pervasive: every one of the 12 models over-infers 35%--49% of its claims (cross-model mean 41.6%; claim-weighted 41.8%), with no model in this evaluation escaping it. Most strikingly, we surface a Self-Monitoring Inversion: at the model-selection level, models' self-assessed OI is negatively rank-correlated with their judge-measured OI (rho = -0.60, p = 0.044; exploratory, wide bootstrap CI [-0.90, +0.06], n = 12). The models that report the least over-inference tend to be flagged as fabricating the most, so self-reported confidence is a misleading signal for comparing models, even though within a single model self-audit still ranks that model's own claims moderately well (AUROC 0.58--0.83). We further show that OI is task-dependent (27%--59%) and that, in a multi-turn pilot, inferred attributes accumulate approximately linearly with little revision. MirageBench positions external verification, rather than model self-report, as a more reliable foundation for trustworthy personalization.

Chinese Translation

个性化的大型语言模型（LLMs）伴随持久记忆的应用日益增多，但其用户模型的真实性尚未得到检验。我们研究了过度推断（Over-Inference, OI）现象：即大型语言模型虚构用户属性，超出证据所支持的范围。我们引入了MirageBench，包含150个角色，涵盖刻板印象、反刻板印象和中性档案的平衡，6个个性化任务跨越“想象梯度”，以及由独立评审员操作化的四类真实性分类法（在400个声明上与盲人标注者验证：Cohen's kappa = 0.863 四类，kappa = 0.900 二类），并在143616个评估声明上对12个模型进行排名。我们发现过度推断现象普遍存在：12个模型中每个模型的过度推断比例在35%至49%之间（跨模型平均41.6%；声明加权41.8%），没有模型能够逃避这一现象。最引人注目的是，我们发现了自我监测反转：在模型选择层面，模型自评的过度推断与评审测量的过度推断呈负相关（rho = -0.60，p = 0.044；探索性，广泛自助法置信区间[-0.90, +0.06]，n = 12）。报告过度推断最少的模型往往被标记为虚构最多，因此自我报告的信心在比较模型时是一个误导性信号，尽管在单个模型内部，自我审计仍能对该模型自身的声明进行中等程度的排名（AUROC 0.58--0.83）。我们进一步表明，过度推断是任务依赖的（27%--59%），并且在多轮试点中，推断的属性大致线性累积，修正较少。MirageBench将外部验证而非模型自报告视为可信个性化的更可靠基础。

View on arXiv Download PDF AI Translation

cs.CL / 38 / 2608.04574

When Memory Lies: An Empirical Study of Spatial Memory Staleness in VLM Agents

当记忆失真：对 VLM 代理空间记忆陈旧性的实证研究

Sun, Yushi, Zhang, Yanjie

Abstract

Memory-augmented VLM agents act on persistent spatial knowledge, yet that knowledge silently goes stale as the environment changes. We ask what happens when an agent must reconcile a confident memory claim with a contradicting observation, and whether current models can catch the conflict before it becomes a safety-relevant mistake. Using a dynamic FrozenLake testbed, we pair a staleness-detection task with a downstream navigation task across three closed-source models and three open-weight VLMs under both text and image inputs (1,800 detection runs, and 12,000 text-mode navigation episodes over four LLM navigators at a shared 50-seed scale). Three findings emerge. First, text solvability does not imply visual grounding: models that flag stale entries reliably from text nonetheless span vision F1 from 0.887 down to 0.067 on the identical grids, and the weakest keeps making fluent, confident decisions that ignore the image. Second, consuming stale memory without an audit is a safety liability: in our primary GPT-4o setting, an agent that trusts raw memory dies more than twice as often as the same agent given no memory at all. Third, auditing helps but does not close the gap: a transparent read-time filter removes much of the safety cost in text mode, yet even oracle stale labels bring no further significant gain on the current grid size, and when visual auditing is unreliable, filtering yields no consistent benefit. Together these results frame spatial-memory staleness as a safety failure mode and isolate reliable visual grounding and action selection under memory--observation conflict as the central open challenges for memory-augmented agents.

Chinese Translation

记忆增强的 VLM 代理依赖于持久的空间知识，但随着环境的变化，这些知识会悄然变得过时。我们探讨当代理必须调和一个自信的记忆声明与一个矛盾的观察时会发生什么，以及当前模型是否能够在冲突成为安全相关错误之前捕捉到这一点。通过使用动态的 FrozenLake 测试平台，我们将一个陈旧性检测任务与一个下游导航任务配对，针对三种闭源模型和三种开放权重的 VLM，在文本和图像输入下进行测试（共 1,800 次检测运行，以及 12,000 次文本模式导航实验，四个 LLM 导航器在共享的 50 种种子规模下）。我们得出了三个发现。首先，文本可解性并不意味着视觉基础：从文本中可靠标记陈旧条目的模型在相同网格上的视觉 F1 分数从 0.887 降至 0.067，而最弱的模型仍然做出流畅、自信的决策，忽视图像。其次，未经审计的陈旧记忆的使用是一个安全隐患：在我们的主要 GPT-4o 设置中，一个信任原始记忆的代理死亡率是没有记忆的同一代理的两倍以上。第三，审计有帮助但并未消除差距：在文本模式下，透明的实时过滤器消除了大部分安全成本，但即使是神谕式的陈旧标签在当前网格规模下也未带来进一步显著的收益，而当视觉审计不可靠时，过滤也未能带来一致的好处。这些结果共同将空间记忆陈旧性框定为一种安全失效模式，并将可靠的视觉基础和在记忆-观察冲突下的行动选择孤立为记忆增强代理面临的核心开放挑战。

View on arXiv Download PDF AI Translation

cs.CL / 39 / 2608.04576

Causal Evidence Extraction and Triangulation in Crisis Reports using Large Language Models: A ReliefWeb-based Study

利用大型语言模型提取和三角验证危机报告中的因果证据：基于ReliefWeb的研究

Zhang, Yuanjun, Oussalah, Mourad

Abstract

Humanitarian reports are long, noisy, and multi-topic, making it difficult to consolidate decision-relevant causal evidence. We present a ReliefWeb study (2000-2024) and a two-stage Large Language Model (LLM) pipeline that extracts structured intervention-outcome records with direction and strength attributes. Query-conditioned extraction restricts output to a specified intervention class, reducing retrieval-induced over-extraction, while snippet grounding links each relation to supporting text for auditability and classification. In an expert-annotated dataset of 100 reports, the best closed-source LLM achieved a weighted F1 score of 90.73% with strong cost-efficiency, while Llama-3.1-8B with supervised fine-tuning reached 94.15% weighted F1 score. We further propose context-preserving triangulation that aggregates strength-weighted evidence within disaster$\times$source cells, applies Laplace smoothing and equally weights cells to quantify cross-context convergence via a Level-of-Evidence score. Applied to cash assistance, food-related outcomes show strong positive convergence (LoE=0.865) and stable long-horizon trajectories.

Chinese Translation

人道主义报告通常较长、噪声多且主题多样，使得整合与决策相关的因果证据变得困难。我们展示了一项基于ReliefWeb的研究（2000-2024）以及一个两阶段的大型语言模型（LLM）流程，该流程提取具有方向和强度属性的结构化干预-结果记录。查询条件提取将输出限制在指定的干预类别，从而减少检索引起的过度提取，而片段基础则将每个关系链接到支持文本，以便于审计和分类。在一个由专家标注的100份报告的数据集中，最佳的闭源LLM实现了90.73%的加权F1分数，且具有良好的成本效率，而经过监督微调的Llama-3.1-8B达到了94.15%的加权F1分数。我们进一步提出了保持上下文的三角验证方法，该方法在灾害$ imes$源单元内聚合强度加权证据，应用拉普拉斯平滑并对单元进行等权重处理，以通过证据水平得分量化跨上下文的收敛性。应用于现金援助时，食品相关结果显示出强烈的正向收敛（LoE=0.865）和稳定的长期轨迹。

View on arXiv Download PDF AI Translation

cs.CL / 40 / 2608.04586

Breaking the Curse ofMultilinguality inMany-to-Many Speech-to-Text Translation via a Resource-AwareMixture of Speech Encoders

通过资源感知的语音编码器混合打破多对多语音转文本翻译中的多语言诅咒

Du, Yexing, Liu, Kaiyuan, Pan, Youcheng, Yang, Bo, Fu, Chengpeng, Wang, Yu, Liu, Ming

Abstract

Multimodal large language models (MLLMs) have achieved significant success in speech-to-text translation (S2TT). However, when processing multilingual speech inputs, a single speech encoder shared across all languages suffers from the curse of multilinguality: languages at different resource levels compete for limited representation capacity, leading to strong high-resource performance but substantial degradation on low-resource speech. To address this problem and improve multilingual consistency, we propose MSRT, a novel framework built around a resource-aware Mixture of Speech Encoders (MoSE). MoSE uses an explicit language router to assign each utterance to an appropriate expert encoder. A frozen expert preserves high-resource language capabilities, while a trainable expert adapts to and specializes in medium- and low-resource languages. We further introduce a five-stage curriculum learning strategy that substantially reduces data dependence, requiring only 10 hours of paired S2TT data per language for effective alignment. We conduct extensive experiments on 45 languages, systematically evaluating all $45 \times 44$ translation directions. Our 4B-parameter model achieves state-of-the-art performance, outperforming substantially larger baselines. Empirical analyses show that MoSE improves high-, medium-, and low-resource languages simultaneously, with the largest gains on low-resource speech, thereby breaking the curse of multilinguality without compromising high-resource performance. To support future multilingual S2TT research, we release our code and models.

Chinese Translation

多模态大型语言模型（MLLMs）在语音转文本翻译（S2TT）方面取得了显著成功。然而，在处理多语言语音输入时，跨所有语言共享的单一语音编码器受到多语言性的诅咒：不同资源水平的语言在有限的表示能力上竞争，导致高资源语言表现良好，但低资源语音的性能显著下降。为了解决这个问题并提高多语言一致性，我们提出了MSRT，一个围绕资源感知的语音编码器混合（MoSE）构建的新框架。MoSE使用显式语言路由器将每个发话分配给适当的专家编码器。一个冻结的专家保留高资源语言的能力，而一个可训练的专家则适应并专注于中低资源语言。我们进一步引入了一种五阶段的课程学习策略，显著减少了对数据的依赖，仅需每种语言10小时的配对S2TT数据即可实现有效对齐。我们在45种语言上进行了广泛的实验，系统地评估了所有$45 imes 44$的翻译方向。我们的4B参数模型实现了最先进的性能，显著超越了更大基线模型。实证分析表明，MoSE同时改善了高、中、低资源语言的表现，尤其在低资源语音上获得了最大的提升，从而打破了多语言性的诅咒，而不影响高资源语言的性能。为了支持未来的多语言S2TT研究，我们发布了我们的代码和模型。

View on arXiv Download PDF AI Translation

cs.CL / 41 / 2608.04588

EASy: Towards Efficient LLM-Based Agentic System

EASy：迈向高效的基于大语言模型的自主系统

Liu, Junnan, Luo, Linhao, Vu, Thuy-Trang, Haffari, Gholamreza

Abstract

Agentic systems have emerged as a promising paradigm for solving complex tasks by coordinating specialized LLM-based agents. However, most existing systems primarily optimize task success while giving limited consideration to execution efficiency under practical constraints such as executor capability and computational cost. Existing router-based methods have limited ability to reason over rich, evolving task contexts, multi-step dependencies, and intermediate execution feedback, and often generalize poorly to unseen executors. We propose EASy, a trainable agentic framework that jointly optimizes task performance and computational efficiency through reinforcement learning. EASy equips an LLM-based orchestrator with explicit knowledge of the capability and cost profiles of heterogeneous executors, enabling context-sensitive coordination beyond performance-only routing. It further introduces a milestone-plan-act workflow that decomposes complex tasks into manageable milestones, constructs dependency-aware execution graphs, assigns suitable executors, and parallelizes independent steps while adapting subsequent decisions to intermediate outcomes. To train the orchestrator, we develop a tree-structured rollout procedure that explores alternative milestone decompositions and execution plans, together with multi-component rewards that capture task correctness, execution efficiency, and trajectory completeness. Extensive experiments on mathematical reasoning, embodied decision-making, and deep research benchmarks show that EASy consistently achieves stronger performance-efficiency trade-offs than strong agentic baselines.

Chinese Translation

自主系统作为一种有前景的范式，通过协调专门的基于大语言模型（LLM）的代理来解决复杂任务。然而，大多数现有系统主要优化任务成功率，而对在执行者能力和计算成本等实际约束下的执行效率考虑有限。现有的基于路由的方法在推理丰富、不断演变的任务上下文、多步骤依赖关系和中间执行反馈方面能力有限，并且通常对未见过的执行者泛化效果不佳。我们提出了EASy，一个可训练的自主框架，通过强化学习共同优化任务性能和计算效率。EASy为基于LLM的协调者提供了异构执行者的能力和成本特征的明确知识，使其能够进行超越仅性能路由的上下文敏感协调。它进一步引入了一个里程碑-计划-行动的工作流程，将复杂任务分解为可管理的里程碑，构建依赖关系感知的执行图，分配合适的执行者，并在适应后续决策时并行化独立步骤。为了训练协调者，我们开发了一种树状结构的展开程序，探索替代的里程碑分解和执行计划，并结合多组件奖励，捕捉任务正确性、执行效率和轨迹完整性。在数学推理、具身决策和深度研究基准上的大量实验表明，EASy在性能与效率的权衡上始终优于强大的自主基线。

View on arXiv Download PDF AI Translation

cs.CL / 42 / 2608.04591

When Absence Is Evidence: Evaluating Completeness-Sensitive Negative Reasoning in Large Language Models

缺失即证据：评估大型语言模型中的完整性敏感负推理

Min, Byoungjae, Edemacu, Kennedy, Cho, Sae-Hong, Choi, Yoonhyuk, Jang, Beakcheol, Kim, Jong Wook

Abstract

Large language models (LLMs) are often asked whether something is absent from a record, list, or retrieved context. Yet non-observation licenses a negative answer only when evidence completely covers the query scope; otherwise, the answer should remain unknown. We call this completeness-sensitive negative reasoning. We introduce CROWN-QA, comprising CROWN-Synth, a controlled paired core that fixes the question and observed facts while varying only query-relative coverage, and CROWN-Real, a real-document contrast-set evaluation with controlled coverage variants. Across three LLM families, models show unstable closure judgments and substantial over-closure, failing to reliably distinguish a justified negative answer (Certified-Negative) from insufficient evidence (Unknown). The dominant CROWN-Synth failure is asymmetric: models often recognize implicitly complete evidence yet treat implicitly partial evidence as query-covering. Prompting redistributes errors between over- and under-closure rather than consistently resolving them. Structured certificate elicitation traces many errors to evidence-coverage mischaracterization. CROWN-Real shows that the core partial-coverage asymmetry persists on real-document content, while its strength and the balance between over- and under-closure vary by model, prompt, and source.

Chinese Translation

大型语言模型（LLMs）常被询问某个信息是否缺失于记录、列表或检索上下文中。然而，非观察仅在证据完全覆盖查询范围时才允许给出否定答案；否则，答案应保持未知。我们称之为完整性敏感负推理。我们引入了CROWN-QA，其中包括CROWN-Synth，这是一个控制配对核心，固定问题和观察事实，同时仅变化查询相关的覆盖度；以及CROWN-Real，这是一个具有控制覆盖变体的真实文档对比集评估。在三种LLM家族中，模型表现出不稳定的闭合判断和显著的过度闭合，未能可靠地区分合理的否定答案（Certified-Negative）和证据不足（Unknown）。CROWN-Synth的主要失败是非对称的：模型通常能够识别隐含的完整证据，但将隐含的部分证据视为覆盖查询。提示的设计在过度闭合和不足闭合之间重新分配错误，而不是一致地解决它们。结构化证书引导将许多错误追溯到对证据覆盖的误判。CROWN-Real显示核心的部分覆盖非对称性在真实文档内容中仍然存在，而其强度以及过度闭合和不足闭合之间的平衡因模型、提示和来源而异。

View on arXiv Download PDF AI Translation

cs.CL / 43 / 2608.04646

Evaluating Theory of Mind in Reasoning Models: Robustness over Reasoning

评估推理模型中的心智理论：推理的稳健性

de Haan, Ian B., van der Putten, Peter, van Duijn, Max

Abstract

Large language models (LLMs) have recently shown strong performance on Theory of Mind (ToM) tests, prompting debate about the nature and validity of the underlying capabilities. At the same time, reasoning-oriented LLMs trained via reinforcement learning with verifiable rewards have demonstrated notable improvements across a range of benchmarks. In this work, we examine the behavior of such reasoning models in ToM tasks using novel adaptations of machine psychological experiments together with results from established benchmarks. We observe that reasoning models consistently exhibit increased robustness to prompt variations and task perturbations. Our analysis suggests these gains come at least partly from models being more robust at reaching the correct answer under prompt and task variation. We read this as evidence for a robustness-based account rather than for a new ToM-specific ability.

Chinese Translation

大型语言模型（LLMs）最近在心智理论（ToM）测试中表现出强大的性能，引发了关于其底层能力性质和有效性的辩论。同时，通过可验证奖励进行强化学习训练的以推理为导向的LLMs在多个基准测试中显示出显著的改进。在本研究中，我们使用机器心理实验的新颖改编和已有基准测试的结果，考察了这些推理模型在ToM任务中的表现。我们观察到，推理模型在提示变化和任务扰动下表现出一致的稳健性提升。我们的分析表明，这些提升至少部分源于模型在提示和任务变化下更稳健地达到正确答案。我们将此解读为基于稳健性的解释，而非一种新的特定于ToM的能力。

View on arXiv Download PDF AI Translation

cs.CL / 44 / 2608.04670

Easy to Complete, Hard to Choose: Investigating LLM Performance on the ProverbIT Benchmark

易于完成，难以选择：调查大型语言模型在ProverbIT基准测试中的表现

Mensa, Enrico, Zane, Lorenzo, Scozzaro, Calogero Jerik, Delsanto, Matteo, Milani, Tommaso, Radicioni, Daniele Paolo

Abstract

Large Language Models (LLMs) have transformed computational linguistics and achieved remarkable performance across numerous natural language processing tasks, yet significant gaps persist in understanding how these systems process culturally embedded linguistic expressions. This paper introduces ProverbIT, a novel Italian benchmark comprising 100 multiple-choice questions designed to evaluate LLMs' ability to complete Italian proverbs. We assess 13 frontier models, including Large Reasoning Models (LRMs) and traditional LLMs, across three tasks: proverb completion, multiple-choice selection with correct answers, and multiple-choice selection without correct answers. Our evaluation reveals surprising results: while nearly all models demonstrate knowledge of the proverbs through successful completion tasks, performance drops dramatically when transitioning to multiple-choice formats without correct answers, with even state-of-the-art reasoning models showing substantial degradation. Through detailed Chain-of-Thought analysis of two LRMs, we uncover that models exhibit a strong bias toward selecting literal synonyms and frequently mention correct proverb endings during reasoning without successfully identifying their absence from the given options. These findings suggest that current LLMs rely heavily on memorized patterns rather than deeper semantic understanding of culturally grounded expressions, highlighting important limitations in their reasoning capabilities for figurative language comprehension.

Chinese Translation

大型语言模型（LLMs）已经改变了计算语言学，并在众多自然语言处理任务中取得了显著的表现，然而在理解这些系统如何处理文化嵌入的语言表达方面仍然存在显著差距。本文介绍了ProverbIT，这是一个新颖的意大利基准测试，包含100个多项选择题，旨在评估LLMs完成意大利谚语的能力。我们评估了13个前沿模型，包括大型推理模型（LRMs）和传统LLMs，涵盖三个任务：谚语完成、带有正确答案的多项选择和不带正确答案的多项选择。我们的评估揭示了令人惊讶的结果：尽管几乎所有模型在成功完成任务中表现出对谚语的知识，但在转向不带正确答案的多项选择格式时，表现急剧下降，甚至最先进的推理模型也显示出显著的退化。通过对两个LRMs进行详细的思维链分析，我们发现模型在选择字面同义词时表现出强烈的偏见，并且在推理过程中频繁提及正确的谚语结尾，但未能成功识别这些结尾在给定选项中的缺失。这些发现表明，当前的LLMs在理解文化根植的表达时，更多依赖于记忆的模式，而非更深层次的语义理解，突显了它们在理解比喻语言时推理能力的重要局限性。

View on arXiv Download PDF AI Translation

cs.CL / 45 / 2608.04678

Kathleen Writes: Autoregressive Generation and Data Scaling Without Attention

凯瑟琳的写作：无注意力的自回归生成与数据扩展

Fountzoulas, George

Abstract

Papers 1-2 of the Kathleen series showed that a byte-level, attention-free architecture built from a wavetable encoder and multi-scale reverberant state can match strong baselines on classification at ~450-700K parameters, without pretraining. We ask whether the same ingredients can generate. (1) Scaling: on byte-level language modeling (WikiText-103, raw UTF-8, no tokenizer), the reverberant model beats a parameter-matched transformer at every dataset scale measured (2-512 MB), e.g. 1.84 vs 2.04 bits/byte at 512 MB with ~0.5M parameters; the transformer needs more than 512 MB to match what the attention-free model learns from 32 MB. (2) Measurement: we introduce FORM DISTANCE, a non-parametric, gaming-resistant instrument for "reads like text": nine statistical axes of human text define a reference cloud, and five constructed fakes are all rejected. (3) Generation: decoding policy dominates architecture -- widening the sampler halves the same model's distance (3.17 to 1.52), and a retrieval-augmented decoding scheme takes the frozen model further (1.52 to 1.14) with no training step involved; the ablation attributes the gain to the sparse phrase dose itself, not the selection gate. The gain has a sharp boundary condition: the phrases must come from the model's own training corpus -- a 40x larger foreign library helps not at all, an effect the attention twin shares, consistent with in-context integration being a capability of scale. We also report four architectural additions that did not help, and a computed lexicon reaching 94% of a learned table's top-1 accuracy at one fifth of the parameters. Everything runs offline; all experiments are reproducible on a free Kaggle T4.

Chinese Translation

凯瑟琳系列的论文1-2展示了一种基于波表编码器和多尺度混响状态的字节级无注意力架构，在约450-700K参数下能够在分类任务中匹配强基线，且无需预训练。我们探讨相同的构件是否可以用于生成。(1) 扩展：在字节级语言建模（WikiText-103，原始UTF-8，无分词器）中，混响模型在每个测量的数据集规模（2-512 MB）上都超过了参数匹配的变换器，例如在512 MB时以约0.5M参数的情况下，混响模型的表现为1.84比特/字节，而变换器需要超过512 MB才能匹配无注意力模型从32 MB中学习到的内容。(2) 测量：我们引入了FORM DISTANCE，这是一种非参数的、抗游戏的工具，用于“读起来像文本”：人类文本的九个统计轴定义了一个参考云，而五个构造的伪造文本均被拒绝。(3) 生成：解码策略主导架构——扩大采样器使得同一模型的距离减半（从3.17降至1.52），而增强检索的解码方案在没有任何训练步骤的情况下进一步降低了距离（从1.52降至1.14）；消融实验将增益归因于稀疏短语剂量本身，而非选择门。增益具有明显的边界条件：短语必须来自模型自身的训练语料库——一个大40倍的外部库完全没有帮助，这一效应与注意力模型相似，符合上下文集成是规模能力的观点。我们还报告了四个未能提供帮助的架构添加，以及一个计算词汇表，其在参数量为五分之一的情况下达到了学习表的前1准确率的94%。所有实验均在离线环境中运行；所有实验均可在免费的Kaggle T4上复现。

View on arXiv Download PDF AI Translation

cs.CL / 46 / 2608.04703

IslamicTurathBench: A Multi-Task, Multi-Discipline Benchmark for Evaluating Large Language Models on the Islamic Scholarly Tradition (turath)

IslamicTurathBench：一个多任务、多学科的基准，用于评估大型语言模型在伊斯兰学术传统（turath）上的表现

Gaben, Shahd, Sbahi, Heba, Rashwani, Samer, Bouchekif, Abdessalam, Al-Khatib, Mutaz, Mohamed, Emad, Eltanbouly, Somaya, Ghaly, Mohammed

Abstract

Large language models (LLMs) are increasingly used for question answering, education, and research, including in religious and cultural domains where answers depend on specialised source traditions. Yet in Islamic Studies, key concepts, methods, and debates preserved in the authoritative scholarly tradition, known as turath, lack high-quality annotated resources. We introduce IslamicTurathBench (ISTB), a multi-task, multi-discipline dataset for evaluating LLMs on classical Islamic scholarship. Developed and reviewed by domain experts, ISTB contains 3,465 question-answer items drawn from 35 recognised source works spanning more than 12 centuries of scholarship across seven key fields of Islamic Studies. To enable comprehensive profiling of model capabilities, ISTB is structured along two axes: scholarly demand (Beginner, Intermediate, and Advanced) and task format (multiple-choice questions, passage-based comprehension, and open-ended knowledge questions). ISTB includes aggregated scores from a scholarly human reference panel and zero-shot baselines from ten systems. The dataset supports reproducible evaluation of language-model behaviour across source works, disciplines, scholarly-demand levels, and question formats in a historically layered scholarly domain.

Chinese Translation

大型语言模型（LLMs）在问答、教育和研究等领域的应用日益广泛，包括在宗教和文化领域，其中答案依赖于专业的源传统。然而，在伊斯兰研究中，权威学术传统（即turath）中保存的关键概念、方法和辩论缺乏高质量的注释资源。我们介绍了IslamicTurathBench（ISTB），这是一个用于评估LLMs在经典伊斯兰学术方面表现的多任务、多学科数据集。ISTB由领域专家开发和审查，包含3,465个问题-答案项，来源于35部公认的经典著作，跨越了超过12个世纪的七个主要伊斯兰研究领域。为了全面评估模型能力，ISTB沿着两个轴线进行结构化：学术需求（初学者、中级和高级）和任务格式（多项选择题、基于段落的理解和开放式知识问题）。ISTB包括来自学术人类参考小组的汇总评分和十个系统的零-shot 基线。该数据集支持在历史层次丰富的学术领域中，对语言模型行为在源作品、学科、学术需求水平和问题格式上的可重复评估。

View on arXiv Download PDF AI Translation

cs.CL / 47 / 2608.04709

EmpaAva: An Open-source Agentic 3D-Avatar Empathetic Live Chatbot

EmpaAva：一个开源的代理式3D化身同理心实时聊天机器人

Yang, Jie, Xu, Wenhao, Lin, Shuhui, Fei, Hao

Abstract

This paper presents EmpaAva, to our knowledge the first open-source, agentic 3D-avatar empathetic chatbot, which carries empathetic response generation (ERG) from text-only exchanges into live, face-to-face interaction. Through a video-call-like interface, a user speaks to a 3D digital human that reads their affect from speech and optional vision, and replies with emotional speech, lip-synced facial motion, and photorealistic 3D Gaussian rendering. At its core, an LLM coordinates a Tri-Agent Architecture, in which perception, empathetic response planning, and embodied rendering form a closed loop, paired with a Response Planning layer that compiles each reply into an executable multimodal plan, keeping voice, expression, and rendering on one empathetic intent. Building on strong open-source modules, EmpaAva supplies the intelligence that binds them into one controllable, inspectable experience. In automatic and human evaluations, EmpaAva surpasses text-only, 2D talking-face, and multimodal avatar baselines in emotion understanding, response quality, and audio-visual consistency. We open-source EmpaAva with an online live demo.

Chinese Translation

本文介绍了EmpaAva，至今为止我们所知的第一个开源代理式3D化身同理心聊天机器人，它将文本交流中的同理心响应生成（ERG）扩展到实时的面对面互动。通过类似视频通话的界面，用户与一个3D数字人类进行对话，该数字人类通过语音和可选的视觉信息读取用户的情感，并以情感化的语言、同步的面部动作和照片级真实感的3D高斯渲染进行回复。在其核心，一个大型语言模型（LLM）协调一个三代理架构，其中感知、同理心响应规划和具身渲染形成一个闭环，并配备一个响应规划层，将每个回复编译成可执行的多模态计划，确保语音、表情和渲染保持在一个同理心意图之下。基于强大的开源模块，EmpaAva提供了将这些模块绑定为一个可控、可检查体验的智能。在自动和人工评估中，EmpaAva在情感理解、响应质量和音视频一致性方面超越了文本-only、2D说话面孔和多模态化身的基准。我们将EmpaAva开源，并提供在线实时演示。

View on arXiv Download PDF AI Translation

cs.CL / 48 / 2608.04746

Caching for the Future: Scrub Jay Episodic Memory Principles for Agent Memory Systems

面向未来的缓存：基于西方 scrub jay 事件记忆原则的智能体记忆系统

Bhandari, Kartikey Singh, Wadhwani, Aarya, Kumar, Dhruv, Narang, Pratik

Abstract

LLM agents that persist across sessions accumulate stored memories whose validity varies enormously by content type, yet existing memory architectures treat all memories as equally persistent and systematically contaminate retrieved context with outdated facts. We show that per-memory, type-conditioned temporal decay, a property of western scrub jay episodic memory, can be operationalized as an auto-classified coefficient $\pi_i$ in an external LLM-agent memory store, yielding ScrubJay-MEM: each memory is encoded as a jointly-bound What--Where--When tuple with an estimated perishability $\pi_i$ and utility horizon $\tau_i$, retrieved by query-adaptive scoring, and revised retroactively at $O(1)$ LLM calls per update. We introduce the Temporal Generalization Test (TGT), a benchmark with held-out retention intervals and a Generalization Gap (GenGap) metric. On TGT, ScrubJay-MEM is the only retrieval-based system with substantially positive GenGap ($+0.108$); on MemoryAgentBench EventQA-64k it improves F1 by $+2.66$ over Mem0 and $+3.09$ over Qwen3-Embedding-4B under a llm backbone. A decay ablation collapses GenGap by $5.7\times$, establishing type-conditioned decay as necessary for the result. Gains narrow under stronger backbones and reverse on fact-consolidation tasks, scoping the contribution to temporal reasoning over perishable facts.

Chinese Translation

在会话中持久存在的大型语言模型（LLM）智能体积累了存储记忆，这些记忆的有效性因内容类型而异，然而现有的记忆架构将所有记忆视为同等持久，并系统性地用过时的事实污染检索的上下文。我们展示了每个记忆的类型条件时间衰减，这一特性源自西方 scrub jay 的事件记忆，可以在外部 LLM-智能体记忆存储中作为自动分类系数 $C0_i$ 进行操作，从而产生 ScrubJay-MEM：每个记忆被编码为一个联合绑定的 What--Where--When 元组，具有估计的易腐性 $C0_i$ 和效用视野 $ au_i$，通过查询自适应评分进行检索，并在每次更新时以 $O(1)$ 的 LLM 调用进行回溯修订。我们引入了时间泛化测试（Temporal Generalization Test, TGT），这是一个具有保留间隔的基准测试，并引入了泛化差距（Generalization Gap, GenGap）指标。在 TGT 测试中，ScrubJay-MEM 是唯一一个具有显著正泛化差距（$+0.108$）的基于检索的系统；在 MemoryAgentBench EventQA-64k 上，它在 LLM 主干下相较于 Mem0 提高了 $+2.66$，相较于 Qwen3-Embedding-4B 提高了 $+3.09$。衰减消融实验使 GenGap 降低了 $5.7 imes$，确立了类型条件衰减对结果的必要性。在更强大的主干下，收益缩小，并在事实巩固任务上出现反转，界定了对易腐事实的时间推理的贡献。

View on arXiv Download PDF AI Translation

cs.CL / 49 / 2608.04761

InsightEmb: Learning Action-Intent Embeddings for Agentic Insight Retrieval

InsightEmb：学习行动意图嵌入以进行代理洞察检索

Chung, Tsz Ting, Li, Jiangnan, Zhou, Jie, Yu, Mo

Abstract

Self-improving agents accumulate reusable insights from prior trajectories, making retrieval increasingly important for turning accumulated experience into actionable guidance. At each decision step, retrieving the right insight can help the agent progress toward its goal, a setting we refer to as agentic insight retrieval. However, existing retrieval methods primarily model semantic similarity, while overlooking whether a retrieved insight resolves the agent's current decision bottleneck. We propose InsightEmb, a contrastive embedding framework that learns transferable progress-oriented retrieval geometry using only mathematical reasoning data. InsightEmb jointly learns to align concrete situations with abstract heuristic rules and to cluster reasoning trajectories with similar progress structures. We evaluate InsightEmb on dynamic agent tasks and a static skill-retrieval benchmark. Without any environment-specific training, InsightEmb improves over all these evaluations, surpassing the performance of existing reasoning embedding models. These results suggest that the geometry of state-insight matching can transfer across domains, enabling effective training from publicly available reasoning data without expensive environment-specific supervision.

Chinese Translation

自我改进的智能体从先前的轨迹中积累可重复使用的洞察，使得检索在将积累的经验转化为可操作指导方面变得愈加重要。在每个决策步骤中，检索正确的洞察可以帮助智能体朝着其目标前进，这一设置我们称之为代理洞察检索。然而，现有的检索方法主要建模语义相似性，而忽视了检索到的洞察是否解决了智能体当前的决策瓶颈。我们提出了InsightEmb，这是一种对比嵌入框架，仅使用数学推理数据来学习可转移的进展导向检索几何。InsightEmb共同学习将具体情况与抽象启发式规则对齐，并将具有相似进展结构的推理轨迹进行聚类。我们在动态智能体任务和静态技能检索基准上评估了InsightEmb。在没有任何特定环境训练的情况下，InsightEmb在所有这些评估中均有所提升，超越了现有推理嵌入模型的性能。这些结果表明，状态-洞察匹配的几何结构可以跨领域转移，使得能够从公开可用的推理数据中进行有效训练，而无需昂贵的特定环境监督。

View on arXiv Download PDF AI Translation

cs.CL / 50 / 2608.04772

Guideline-as-Oracle: Zero-Annotation Training of an Ophthalmic Telephone Triage Agent

指南作为神谕：无注释训练的眼科电话分诊代理

Wang, Chenyu, Liu, Yi, Li, Baoqing, Tu, Min, Song, Diping

Abstract

Scaling supervision for multi-turn medical agents is difficult because expert dialogue annotation is costly and clinical conversations are privacy-restricted. We introduce Guideline-as-Oracle (GAO), which compiles American Academy of Ophthalmology guidance into a 70-row operational rule table and uses it as the sole source of instance-level supervision for 3,000 training dialogues, reserving human labeling for evaluation. Because converting rules into dialogues is itself a design problem, we catalog eight construction strategies, including cited-row tier assignment, one-fact boundary pairs, metadata-only repair, and label repair, and characterize the evidential status of each: labeling mechanism, null, confounded, or evaluated only as a package. Fine-tuning a 9B backbone on this corpus yields GAO-Triage, improving agreement with a 201-case operational reference from 61.7% to 74.1% (exact McNemar p=0.0046) and emergent-case recall from 9.5% to 69.0%; the gains persist across a second seed and patient simulator. None of the seven general-purpose systems we test dominates GAO-Triage on both metrics, and GAO-Triage requires no frontier model at inference time. Permuting label-dialogue assignments collapses the model to a constant-routine predictor, indicating that the signal lies in guideline-derived assignment rather than dialogue surface form. Label repair coincides with the disappearance of a late-training safety degradation.

Chinese Translation

对于多轮医疗代理的监督扩展是困难的，因为专家对话注释成本高昂且临床对话受到隐私限制。我们提出了指南作为神谕（Guideline-as-Oracle, GAO），该方法将美国眼科学会的指导汇编成一个70行的操作规则表，并将其作为3000个训练对话的唯一实例级监督来源，将人工标注保留用于评估。由于将规则转换为对话本身就是一个设计问题，我们列出了八种构建策略，包括引用行层级分配、单事实边界对、仅元数据修复和标签修复，并描述了每种策略的证据状态：标注机制、无效、混淆或仅作为一个整体进行评估。对该语料库进行微调的9B主干模型产生了GAO-Triage，使得与201个案例操作参考的协议从61.7%提高到74.1%（精确的McNemar p=0.0046），新出现案例的召回率从9.5%提高到69.0%；这些提升在第二个种子和患者模拟器中持续存在。我们测试的七个通用系统在这两个指标上均未超越GAO-Triage，并且GAO-Triage在推理时不需要前沿模型。标签-对话分配的置换使模型崩溃为常量例程预测器，表明信号存在于基于指南的分配中，而不是对话表面形式。标签修复与晚期训练安全性降级的消失相吻合。

View on arXiv Download PDF AI Translation

cs.CL / 51 / 2608.04786

Reachability in 3-VAS

3-VAS中的可达性

Kamiński, Łukasz, Lasota, Sławomir

Abstract

We settle the exact complexity of the reachability problem in (stateless) vector addition systems (VAS) in fixed low dimension. In dimensions 2-4 it has only been known to be sandwiched between NP and PSPACE. We prove PSPACE-hardness of the reachability problem for symmetric vector addition systems in dimension 3 (3-VAS), a restricted fragment of general 3-VAS. Combined with previously established PSPACE upper bounds, our result settles the complexity of the problem to be PSPACE-complete in 3-VAS and 4-VAS, as well as in their symmetric fragments.

Chinese Translation

我们确定了固定低维度下（无状态）向量加法系统（VAS）中可达性问题的确切复杂性。在2-4维中，该问题的复杂性仅被认为介于NP和PSPACE之间。我们证明了对称向量加法系统在3维（3-VAS）中的可达性问题是PSPACE-困难的，这是一般3-VAS的一个受限片段。结合之前建立的PSPACE上界，我们的结果确定了该问题在3-VAS和4-VAS以及它们的对称片段中的复杂性为PSPACE-完全。

View on arXiv Download PDF AI Translation

cs.CL / 52 / 2608.04808

A Modular Part-of-Speech Tagger for Scottish Gaelic using spaCy

基于spaCy的苏格兰盖尔语模块化词性标注器

Stefan, Peter, Barclay, Peter J, Lawson, Alistair

Abstract

Part-of-speech tagging for low-resource languages remains challenging due to limited annotated data, especially for linguistically complex languages. Gaidhlig (Scottish Gaelic) is a morphologically rich and endangered language with limited digital resources, making it suitable for examining a lightweight language processing approach. This paper describes using the modular spaCy Natural Language Processing framework to build part-of-speech taggers for Gaidhlig using the Annotated Reference Corpus of Scottish Gaelic. We train two models with minimal pre-processing and configuration: one using a fine-grained tagset and another using a reduced coarse-grained tagset. Both models are trained without external embeddings or pre-trained language models, using only supervised learning from the available corpus. The fine-grained model achieves 88.6% tagging accuracy, while the coarse-grained model achieves 93.7%. The results are comparable to those of the two previously published Gaidhlig taggers, indicating that simple, off-the-shelf language processing pipelines can demonstrate good performance in low-resource and morphologically complex linguistic settings.

Chinese Translation

由于注释数据有限，低资源语言的词性标注仍然面临挑战，尤其是对于语言学上复杂的语言。盖尔语（Gaidhlig）是一种形态丰富且濒危的语言，数字资源有限，适合用于检验轻量级语言处理方法。本文描述了使用模块化的spaCy自然语言处理框架，基于苏格兰盖尔语注释参考语料库构建盖尔语的词性标注器。我们训练了两个模型，采用最小的预处理和配置：一个使用细粒度标签集，另一个使用简化的粗粒度标签集。这两个模型均在没有外部嵌入或预训练语言模型的情况下，仅使用可用语料库进行监督学习。细粒度模型的标注准确率达到88.6%，而粗粒度模型的准确率为93.7%。这些结果与之前发布的两个盖尔语标注器的表现相当，表明简单的现成语言处理管道在低资源和形态复杂的语言环境中能够展现出良好的性能。

View on arXiv Download PDF AI Translation

cs.CL / 53 / 2608.04828

Skill-Use: Can LLMs Actually Use Skills in Agentic Harnesses?

技能使用：大型语言模型（LLMs）能否在代理框架中真正使用技能？

Han, Jinyi, Xu, Yuanjian, Liao, Ying, Wang, Xinyi, Jiang, Zishang, Di, Zixiang, Lu, Fanyang, Hu, Zhichao, Xiao, Yanghua

Abstract

Large language model (LLM) agents increasingly rely on skills, structured documents that specify when to act, which procedure to follow, and which tools are allowed. Existing evaluations mostly judge the quality of a skill or its contribution to task success, leaving unexamined whether an agent can recognize a relevant skill and apply it on its own. We introduce Skill-Use, a benchmark that evaluates skill use under progressive disclosure, where an agent sees only a skill's name and short description and must retrieve the full procedure before following it. Skill-Use separates three facets of skill use. Trigger measures whether the agent invokes the relevant skill, Compliance measures how faithfully it follows the prescribed procedure, and Boundary measures whether it avoids forbidden operations. A Skill-Use (SU) score combines the three and credits execution only after the skill is triggered. Skill-Use pairs 79 real skills with 177 executable tasks across nine domains, each grounded in real files, run in an isolated Docker sandbox, and scored by a trajectory-based rubric. Evaluating eight LLMs under two agent harnesses, we find that reliable skill use remains out of reach, as the strongest configuration reaches an SU of only 0.613. Triggering and procedural compliance fail as independent bottlenecks, and both scores and model rankings shift with the harness, so skill use behaves as a capability conditioned on the harness rather than a fixed property of the model.

Chinese Translation

大型语言模型（LLM）代理越来越依赖于技能，这些技能是结构化文档，指定何时行动、遵循何种程序以及允许使用哪些工具。现有评估主要判断技能的质量或其对任务成功的贡献，而未考察代理是否能够识别相关技能并自主应用。我们引入了技能使用（Skill-Use），这是一个在渐进披露下评估技能使用的基准，其中代理仅看到技能的名称和简短描述，必须在遵循之前检索完整程序。技能使用将技能使用分为三个方面：触发（Trigger）衡量代理是否调用相关技能，合规性（Compliance）衡量其遵循规定程序的忠实程度，边界（Boundary）衡量其是否避免禁用操作。技能使用（SU）分数结合了这三者，仅在技能被触发后才给予执行分数。技能使用将79个真实技能与跨越九个领域的177个可执行任务配对，每个任务都基于真实文件，在隔离的Docker沙箱中运行，并通过基于轨迹的评分标准进行评分。在两种代理框架下评估八个LLM，我们发现可靠的技能使用仍然遥不可及，因为最强的配置仅达到0.613的SU分数。触发和程序合规性作为独立瓶颈失败，且随着框架的变化，分数和模型排名也发生变化，因此技能使用表现为一种依赖于框架的能力，而非模型的固定属性。

View on arXiv Download PDF AI Translation

cs.CL / 54 / 2608.04847

Do Language Models Know Their Slang? Queer Slang Understanding in User-Generated Content

语言模型是否了解其俚语？用户生成内容中的酷儿俚语理解

Denitto, Arianna, Savoldi, Beatrice

Abstract

Despite its cultural relevance and diffusion, queer slang remains underrepresented in Natural Language Processing research. Towards addressing this gap, we introduce Slang-Q, a manually curated dataset of naturally user-generated English sentences paired with queer slang terms and reference definitions, built upon a newly constructed taxonomy of 118 queer terms. We use this resource to conduct a first exploratory evaluation of language models on their ability to understand and define queer slang under varying prompting conditions. Slang-Q is intended as a basis for studying how current models handle sensitive, community-specific language and whether they can provide accurate and reliable information about such forms of identity and linguistic expression.

Chinese Translation

尽管酷儿俚语在文化上具有重要性并且广泛传播，但在自然语言处理研究中仍然缺乏代表性。为了解决这一问题，我们引入了 Slang-Q，这是一个手动策划的数据集，包含自然生成的英语句子，配有酷儿俚语术语和参考定义，基于新构建的 118 个酷儿术语的分类法。我们利用这一资源对语言模型在不同提示条件下理解和定义酷儿俚语的能力进行了首次探索性评估。Slang-Q 旨在作为研究当前模型如何处理敏感的、特定社区语言的基础，以及它们是否能够提供关于这些身份和语言表达形式的准确可靠信息。

View on arXiv Download PDF AI Translation

cs.CL / 55 / 2608.04869

Preverbal Uninflected and Underived Roots in Mapudungun. Wuno and Its Implications

马普杜古语中的前动词未变形和未派生词根：wuno及其意义

Chandia, Andres

Abstract

This study examines the grammatical status of preverbal uninflected and underived roots in Mapudungun, with particular focus on wuno 'return/re-'. Through a critical review of scholarly classifications--auxiliaries (Smeets, 2008), modal prefixes (Longkon, 2011), and preverbal particles/complex verb stems (Z\'u\~niga, 2006)--we demonstrate the limitations of existing frameworks. A diachronic corpus analysis spanning four centuries (1606-present) reveals that these elements exhibit three distinct profiles: stable V1 compounds (kim, shinge), volatile V1 rates reflecting orthographic shift (pepi, wuno), and a true particle (kalli). The discovery of V2 attestations for kim and kupa confirms their status as full lexical verbs. We propose a prosodic-orthographic hypothesis: apparent "variable binding" results from the fossilization of prosodic pauses transcribed by early missionaries as spaces, a convention later reanalysed by speakers as syntactic boundaries. The evidence supports Z\'u\~niga's radical concatenation as the correct grammatical model, with implications for the study of languages with no pre-contact written tradition.

Chinese Translation

本研究考察了马普杜古语中前动词未变形和未派生词根的语法状态，特别关注wuno '返回/再-'。通过对学术分类的批判性回顾——助动词（Smeets, 2008）、情态前缀（Longkon, 2011）以及前动词粒子/复杂动词词干（Zúñiga, 2006）——我们展示了现有框架的局限性。跨越四个世纪（1606年至今）的历时语料分析显示，这些元素展现出三种不同的特征：稳定的V1复合词（kim, shinge）、反映正字法变化的波动性V1比率（pepi, wuno）以及一个真正的粒子（kalli）。对kim和kupa的V2证据确认了它们作为完整词汇动词的地位。我们提出了一种韵律-正字法假说：表面上的“可变绑定”源于早期传教士将韵律停顿转录为空格的化石化，这一惯例后来被说话者重新分析为句法边界。证据支持Zúñiga的激进连接理论作为正确的语法模型，对没有接触前书写传统的语言研究具有重要意义。

View on arXiv Download PDF AI Translation

cs.CL / 56 / 2608.04872

A-SR: Self-Evolving Agentic LLMs for Symbolic Regression via Hierarchical Coordination

A-SR：通过层级协调实现自我进化的代理型大语言模型用于符号回归

Zhao, Wenxiao, Liu, Dong, Xu, Kaiyi, Liu, Feng, Zhao, Zhen, Ben, Fei, Wang, Shu, Li, Wenhao, Wu, Yingnian, Ling, Fenghua, Li, Haobo, Bai, Lei

Abstract

Symbolic regression aims to discover closed-form equations from data, but existing LLM-guided methods often rely on a unified proposal loop that compresses heterogeneous search failures into a scalar score and a single prompt. We propose A-SR, a self-evolving agentic framework that shifts the control unit from expression edits to role-conditioned evidence views. A-SR coordinates formula discovery through routing among coordination protocols, an online evaluator-reward role policy, and state-routed process memory. During search, evaluator feedback characterizes reliability and productivity, updates role-level utilities, and routes elite motifs, failure traces, and validity diagnostics to different agents. The framework self-evolves at two timescales: within a run, it adapts the search process without updating LLM parameters; across runs, recorded trajectories can be distilled into open-source LLMs as role-conditioned proposal priors. Averaged over the four LSR-Synth scientific domains in LLM-SRBench, A-SR improves [email protected] over baselines from 25.79% to 48.30% with Llama3.1-8B, while A-SR-LoRA improves the corresponding Qwen3-4B result from 24.58% to 38.29%. On four real-world scientific discovery tasks, A-SR obtains the best in-distribution or out-of-distribution normalized mean squared error on 7 of 8 reported metrics.

Chinese Translation

符号回归旨在从数据中发现封闭形式的方程，但现有的基于大语言模型（LLM）的方法通常依赖于统一的提议循环，将异构搜索失败压缩为一个标量评分和一个单一提示。我们提出了A-SR，一个自我进化的代理框架，将控制单元从表达编辑转移到角色条件的证据视图。A-SR通过在协调协议之间进行路由、在线评估-奖励角色策略和状态路由的过程记忆来协调公式发现。在搜索过程中，评估者反馈表征了可靠性和生产力，更新角色级效用，并将精英模式、失败痕迹和有效性诊断路由到不同的代理。该框架在两个时间尺度上自我进化：在一次运行中，它在不更新LLM参数的情况下适应搜索过程；在多次运行中，记录的轨迹可以被提炼为开源LLM的角色条件提议先验。在LLM-SRBench的四个LSR-Synth科学领域中，A-SR在Llama3.1-8B上将[email protected]从基线的25.79%提高到48.30%，而A-SR-LoRA将相应的Qwen3-4B结果从24.58%提高到38.29%。在四个真实世界的科学发现任务中，A-SR在8个报告指标中的7个上获得了最佳的分布内或分布外归一化均方误差。

View on arXiv Download PDF AI Translation

cs.CL / 57 / 2608.04899

Evaluation Pitfalls and Sparsity Limitations in LLM-based Confidence Estimates for Classification

基于大型语言模型的分类信心估计中的评估陷阱与稀疏性限制

Merdjanovska, Elena, Zaidan, Omar, Rücklé, Andreas

Abstract

Confidence estimation is essential when LLMs are used for classification, indicating when predictions can be trusted. However, common approaches such as verbalization produce extremely sparse outputs. For instance, Qwen3-32B verbalizes only eight unique confidence values on SST-2, with over half being exactly 95%, a pattern we observe consistently across four datasets and two LLMs. Besides limiting practical utility, we show that this sparsity critically affects evaluation: the choice of interpolation in area under the accuracy-rejection curve (AUARC) dramatically alters rankings, with consistency sampling dropping from best to worst under stepwise versus linear interpolation. We advocate for standardizing stepwise interpolation for a fairer comparison. Under such a fair evaluation, we find that weighting verbalized digits by token probabilities, a method we term verbalization logprobs, addresses sparsity and achieves the best AUARC (+2.3 points over vanilla verbalization) without incurring additional inference cost.

Chinese Translation

信心估计在使用大型语言模型（LLMs）进行分类时至关重要，它指示了何时可以信任预测。然而，常见的方法如语言表达产生的输出极为稀疏。例如，Qwen3-32B在SST-2上仅表达出八个独特的信心值，其中超过一半恰好为95%，这一模式在四个数据集和两个LLM中都一致存在。除了限制实际应用外，我们还展示了这种稀疏性对评估的关键影响：在准确性-拒绝曲线（AUARC）下的插值选择显著改变排名，在分段插值与线性插值下，一致性采样的表现从最佳降至最差。我们建议标准化分段插值以实现更公平的比较。在这样的公平评估下，我们发现通过令牌概率加权语言表达的数字（我们称之为verbalization logprobs）解决了稀疏性问题，并在不增加额外推理成本的情况下实现了最佳的AUARC（比普通语言表达提高了2.3分）。

View on arXiv Download PDF AI Translation

cs.CL / 58 / 2608.04904

Strengthening Target-Language Features: SAE-Based Steering for Multilingual Inference

增强目标语言特征：基于稀疏自编码器的多语言推理引导

Wang, Hongsheng, Koehn, Phlipp

Abstract

Multilingual large language models exhibit substantial performance differences across languages, while existing adaptation methods often require parameter updates and considerable multilingual training data. We propose an inference-time multilingual steering method that uses pretrained sparse autoencoders to identify and strengthen target-language-related features. Using multilingual parallel sentences, we compare SAE activations across languages and select a small number of layer-specific features associated with each target language. These features are decoded into steering signals and injected into the model's hidden states without additional training. Experiments with Gemma-3-12B-it show average accuracy improvements of 10.9 percentage points on XCOPA, 5.3 points on XNLI, and 1.9 points on MGSM.

Chinese Translation

多语言大型语言模型在不同语言之间表现出显著的性能差异，而现有的适应方法通常需要参数更新和大量的多语言训练数据。我们提出了一种推理时的多语言引导方法，该方法利用预训练的稀疏自编码器（Sparse Autoencoders, SAE）来识别和增强与目标语言相关的特征。通过使用多语言平行句子，我们比较了不同语言间的SAE激活，并选择与每个目标语言相关的一小部分特定层特征。这些特征被解码为引导信号，并在不进行额外训练的情况下注入到模型的隐藏状态中。与Gemma-3-12B-it的实验表明，在XCOPA上平均准确率提高了10.9个百分点，在XNLI上提高了5.3个百分点，在MGSM上提高了1.9个百分点。

View on arXiv Download PDF AI Translation

cs.CL / 59 / 2608.04928

Does Out-of-Sight Equal Out-of-Mind in CoT Monitorability?

在链式思维监控中，视而不见是否等于忘却？

Ferreira, Pedro, Aziz, Wilker, Titov, Ivan

Abstract

Chain-of-thought (CoT) reasoning offers a window into the decision-making of large language models (LLMs), which can be monitored for target behaviors by reading the reasoning trace, motivating work on CoT monitorability. Latent CoT approaches, however, replace the explicit tokens with a small number of continuous states, lowering inference costs but removing the readable trace this monitoring relies on. Monitoring then requires alternative access to the model, such as probing its activations or verbalizing the latent states back into text, but how much monitorability these alternatives preserve is unclear. We study this question with a hint-based intervention setup, a proxy for behaviors where models exploit biasing input cues, e.g., an inadvertently leaked answer or a belief stated by the user, without acknowledging them. Taking hint-reliance as the monitorability target, we compare monitors across reasoning modes, from explicit CoT to weakly- and strongly-supervised latent CoT, on math reasoning and question answering. We find that, in this setup, monitorability depends more on properties of the task (such as whether the correct answer constrains the supporting reasoning) and the level of access to model internals than on the reasoning mode.

Chinese Translation

链式思维（Chain-of-thought, CoT）推理为大型语言模型（Large Language Models, LLMs）的决策过程提供了一个观察窗口，通过阅读推理轨迹可以监控目标行为，这激励了对CoT监控能力的研究。然而，潜在的CoT方法用少量连续状态替代了显式的标记，降低了推理成本，但也去除了这种监控所依赖的可读轨迹。因此，监控需要通过其他方式访问模型，例如探测其激活状态或将潜在状态转化为文本，但这些替代方案能保留多少监控能力仍不清楚。我们通过基于提示的干预设置研究这个问题，这是一种代理行为，模型在此过程中利用偏向性输入线索，例如无意中泄露的答案或用户陈述的信念，而不予以承认。以提示依赖作为监控目标，我们比较了不同推理模式下的监控能力，从显式CoT到弱监督和强监督的潜在CoT，重点关注数学推理和问答任务。我们发现，在这种设置中，监控能力更多地依赖于任务的属性（例如，正确答案是否限制了支持推理）和对模型内部的访问程度，而不是推理模式。

View on arXiv Download PDF AI Translation

cs.CL / 60 / 2608.04934

State2State: Environment-Derived Mid-Training for LLM Agents

State2State：基于环境的中期训练方法用于大规模语言模型代理

Lei, Xuanyu, Zhu, Yiqi, Li, Chenliang, Liu, Kaiming, Li, Peng, Yan, Ming, Ye, Jieping, Zhang, Ya-Qin, Liu, Yang

Abstract

Training LLM agents commonly relies on supervised fine-tuning from expert trajectories or online reinforcement learning over human-specified tasks with handcrafted verifiers. Though effective, both remain bottlenecked by externally specified tasks and supervision signals, limiting the scalability and diversity of agent training. We study an environment learning paradigm in which agents acquire interaction and manipulation capabilities solely through environment interaction, without externally specified tasks. We propose State2State, an environment-derived mid-training method that converts explored environment states into training objectives, challenging agents to reach a specified target state. By deriving tasks from environment exploration and verifying success through rule-based state matching, State2State provides scalable and verifiable training objectives without expert supervision or manual task design. Experiments on ALFWorld and ScienceWorld show that State2State improves agent performance as a standalone environment-learning stage in most settings. As initialization for downstream RL, it further improves final performance and learning efficiency, with promising evidence of cross-environment generalization.

Chinese Translation

大规模语言模型（LLM）代理的训练通常依赖于从专家轨迹进行的监督微调或基于人类指定任务的在线强化学习，配合手工设计的验证器。尽管这些方法有效，但都受到外部指定任务和监督信号的瓶颈限制，从而限制了代理训练的可扩展性和多样性。我们研究了一种环境学习范式，在该范式中，代理仅通过与环境的互动获得交互和操作能力，而无需外部指定任务。我们提出了State2State，这是一种基于环境的中期训练方法，它将探索到的环境状态转换为训练目标，挑战代理达到指定的目标状态。通过从环境探索中派生任务并通过基于规则的状态匹配验证成功，State2State提供了可扩展且可验证的训练目标，而无需专家监督或手动任务设计。在ALFWorld和ScienceWorld上的实验表明，State2State在大多数设置中作为独立的环境学习阶段提高了代理的性能。作为下游强化学习的初始化，它进一步提高了最终性能和学习效率，并提供了跨环境泛化的有希望的证据。

View on arXiv Download PDF AI Translation

cs.CL / 61 / 2608.04939

Reading Between the Frames: Interpreting Implicit and Non-literal Meaning in Social Media Videos

解读框架之间的含义：社交媒体视频中的隐含和非字面意义

Wang, Yang, Ma, Yanan, Liu, Yiqi, Chang, Zi Yan, Chen, Chi-Li, Hsiao, Chia-Yi, Loakman, Tyler, Villavicencio, Aline, Xiao, Chenghao, Lin, Chenghua

Abstract

Social media videos often communicate meanings that go beyond their visible actions, captions, or speech. A mundane clip may become humorous, ironic, or satire only through the interaction of multimodal cues and cultural context, making such content a difficult test case for video-language models. In this paper, we introduce \textit{DrivelHub+}, a benchmark for evaluating whether models can infer the implicit, non-linear, and rhetorically layered meanings of social media videos that appear nonsensical on the surface but convey deliberate pragmatic meanings. DrivelHub+ consists of 1,000 videos collected from social media, each annotated with a human-written implicit narrative explanation. Unlike conventional video understanding tasks focused on recognition or description, we present a benchmark that targets contextual multimodal reasoning. We evaluate current video-language models from two perspectives: explanation, where models must explain the pragmatic comprehension of a video in natural language; and representation, where we adapt reasoning-as-retrieval to test whether model representations align videos with their corresponding implicit narratives in both video-to-text and text-to-video retrieval. Our benchmark provides a diagnostic setting for measuring the gap between multimodal perception and pragmatic comprehension, asking whether current models can move beyond describing what is shown to inferring what is meant.

Chinese Translation

社交媒体视频常常传达超越其可见动作、字幕或言语的意义。一段平凡的片段可能通过多模态线索和文化背景的互动而变得幽默、讽刺或讽刺，这使得此类内容成为视频语言模型的一个困难测试案例。本文介绍了 extit{DrivelHub+}，一个评估模型是否能够推断社交媒体视频中隐含的、非线性和修辞层次的意义的基准，这些视频表面上看似无意义，但传达了深思熟虑的语用意义。DrivelHub+ 包含从社交媒体收集的 1,000 个视频，每个视频都附有人工撰写的隐含叙述解释。与传统的视频理解任务（侧重于识别或描述）不同，我们提出了一个针对上下文多模态推理的基准。我们从两个角度评估当前的视频语言模型：解释，即模型必须用自然语言解释视频的语用理解；以及表征，我们调整推理作为检索的方法，以测试模型的表征是否在视频到文本和文本到视频的检索中将视频与其对应的隐含叙述对齐。我们的基准提供了一个诊断环境，用于测量多模态感知与语用理解之间的差距，探讨当前模型是否能够超越描述所展示的内容，推断其所表达的意义。

View on arXiv Download PDF AI Translation

cs.CL / 62 / 2608.04980

Protoreasoning in Tiny Transformers

微型变换器中的原始推理

Valle, Eduardo, Reid, Fergal

Abstract

We show that tiny transformers can profitably employ a simple form of Chain of Thought, which we call protoreasoning, allowing us to study step-by-step reasoning on ~1M-parameter models and opening up opportunities for much more detailed experimentation and analysis than is feasible for larger models. Current Large Language Models exhibit impressive step-by-step reasoning, but we have yet to understand its generality, i.e., when and how LLMs learn genuinely general algorithms rather than "bags of heuristics." Such questions are hard to settle on compute-intensive frontier models trained on opaque data. To work at model scales far below the threshold for natural-language competence, we define reasoning-friendly tasks on Dyck languages (sentences of correctly nested brackets). We find that protoreasoning traces substantially close the out-of-distribution generalization gap, and ablations confirm that the trace's content, not merely its extra tokens, drives the gain.

Chinese Translation

我们展示了微型变换器可以有效地采用一种简单的思维链形式，我们称之为原始推理（protoreasoning），这使我们能够在约100万参数的模型上研究逐步推理，并为比大型模型更详细的实验和分析开辟了机会。目前的大型语言模型展现了令人印象深刻的逐步推理能力，但我们尚未理解其普遍性，即大型语言模型（LLMs）何时以及如何学习真正通用的算法，而不是“启发式的集合”。在计算密集型的前沿模型上，基于不透明数据进行此类问题的解决是困难的。为了在远低于自然语言能力阈值的模型规模上进行研究，我们定义了基于Dyck语言（正确嵌套括号的句子）的推理友好任务。我们发现原始推理的轨迹显著缩小了分布外泛化差距，消融实验确认轨迹的内容，而不仅仅是其额外的标记，推动了这一增益。

View on arXiv Download PDF AI Translation

cs.CL / 63 / 2608.05004

DelusionEval: Measuring Delusion-Linked Behaviors in AI Chatbots

DelusionEval：测量AI聊天机器人中的妄想相关行为

Moore, Jared, Mock, Andrea, Mai, Yifan, Anthis, Jacy Reese, Louie, Ryan, Agnew, William, Mehta, Ashish, Klyman, Kevin, Liang, Percy, Haber, Nick, Lin, Eric, Ong, Desmond C.

Abstract

Mental health professionals have raised concerns about risks of psychological harm from interaction with large language models (LLMs), including "delusional spirals" in which concerning human and LLM behaviors reinforce each other over time. With growing public use of LLM-powered chatbots, there is an urgent need to build evaluations grounded in real-world episodes of psychological harm experienced by users. We developed DelusionEval, an evaluation protocol that tests a model's tendencies to exhibit behaviors linked to promoting user delusions. We prompt each model with 589 unique conversation histories from 18 participants, comprising 12,591 messages from users who experienced delusions and psychological harm. We find that the tendency of an evaluated LLM to exhibit delusion-linked behavior does not reliably correlate with model size, release date, or the presence of test-time reasoning. However, extending the context of prior messages substantially increases rates of delusion-linked behaviors, providing evidence for the importance of context in LLM safety evaluation. For example, the rate of failing to discourage self-harm when the user expresses suicidal ideation increases from 30.0% to 41.1% when an additional 350 messages are prepended to the conversation history. All model families (e.g., GPT, Claude) exhibit substantial rates of delusion-linked behaviors. Within families, later, larger, or higher-reasoning models are not uniformly better across all behavior categories. Our results raise concerns regarding the potential psychological impact of LLMs and the need for more rigorous studies of real-world human-AI interaction.

Chinese Translation

心理健康专业人士对与大型语言模型（LLMs）互动所带来的心理伤害风险表示担忧，包括“妄想螺旋”，在这种情况下，人类和LLM的令人担忧的行为随着时间的推移相互强化。随着公众对LLM驱动的聊天机器人的使用日益增加，迫切需要建立基于用户经历的心理伤害真实案例的评估。我们开发了DelusionEval，这是一种评估协议，用于测试模型表现出与促进用户妄想相关的行为的倾向。我们使用来自18名参与者的589个独特对话历史对每个模型进行提示，这些对话历史包含了来自经历过妄想和心理伤害的用户的12,591条消息。我们发现，经过评估的LLM表现出妄想相关行为的倾向与模型大小、发布日期或测试时推理的存在并没有可靠的相关性。然而，扩展先前消息的上下文显著增加了妄想相关行为的发生率，提供了上下文在LLM安全评估中重要性的证据。例如，当用户表达自杀意念时，未能劝阻自残的比例从30.0%增加到41.1%，当额外的350条消息被添加到对话历史中时。所有模型家族（例如，GPT、Claude）都表现出显著的妄想相关行为发生率。在同一家族中，后期、较大或更高推理能力的模型在所有行为类别中并不一致地表现更好。我们的结果引发了对LLM潜在心理影响的担忧，以及对真实人机互动进行更严格研究的必要性。

View on arXiv Download PDF AI Translation

cs.CL / 64 / 2608.05013

OneDayAgent: Towards a Long-Horizon Harness for Autonomous Agents

OneDayAgent：面向自主智能体的长时间跨度框架

Zheng, Jingsheng, Fang, Xinyuan, Zhang, Jintian, Gui, Zhengke, Chen, Huajun, Zhang, Ningyu

Abstract

LLM agents are increasingly applied to open-ended everyday requests that span work, study, and life. These tasks are long-horizon, cross-environment, and multimodal, forcing the agent to preserve goals and constraints across many steps while navigating heterogeneous tools and attachments. While prior work has addressed individual failure modes such as goals drift, states loss, and context overflow, whether a single harness can manage them jointly and remain effective across backends has received less study. We present OneDayAgent, a long-horizon harness for autonomous agents. OneDayAgent turns an open-ended request into a managed execution process that decomposes tasks into bounded subtasks, maintains execution memory under context pressure, and verifies and repairs the final deliverable. We evaluate OneDayAgent on AgentIF-OneDay across 104 tasks. With the GLM-5.2 backend, OneDayAgent sets a new state of the art with an overall score of 0.821. The same harness runs across five backend LLMs from three model families, indicating the harness generalizes across backends without tuning, even as different models induce distinct execution styles under the same workflow.

Chinese Translation

大型语言模型（LLM）智能体越来越多地应用于跨越工作、学习和生活的开放式日常请求。这些任务具有长时间跨度、跨环境和多模态的特点，迫使智能体在使用异构工具和附件的过程中保持目标和约束。尽管之前的研究已经解决了目标漂移、状态丢失和上下文溢出等个别失败模式，但是否可以通过单一框架共同管理这些问题并在不同后端中保持有效性则研究较少。我们提出了OneDayAgent，这是一个面向自主智能体的长时间跨度框架。OneDayAgent将开放式请求转化为一个管理执行过程，该过程将任务分解为有限的子任务，在上下文压力下保持执行记忆，并验证和修复最终交付成果。我们在AgentIF-OneDay上对OneDayAgent进行了104项任务的评估。在GLM-5.2后端下，OneDayAgent以0.821的整体得分创下了新的技术领先水平。相同的框架可以在来自三个模型家族的五个后端LLM上运行，表明该框架在不同后端之间具有通用性，无需调优，即使不同模型在相同工作流程下会产生不同的执行风格。

View on arXiv Download PDF AI Translation

cs.CL / 65 / 2608.05028

Language Models Generalize to Human-like Word Order Preferences

语言模型对人类类似的词序偏好进行泛化

Popadich, Amanda, Steinert-Threlkeld, Shane

Abstract

A central question in language acquisition is whether linguistic biases can emerge from general learning mechanisms operating over underdetermined input. Artificial Language Learning (ALL) studies have shown that human learners reliably generalize beyond the evidence provided, including by preferring scope-homomorphic noun phrase modifier orders. In this work, we investigate whether language models exhibit the same bias under similar conditions. We create a controlled learning environment in which models are trained on a corpus where all noun phrases containing multiple modifiers have been removed, eliminating direct evidence about modifier ordering, and are then evaluated on multiple modifier sentences. Across three model sizes, we find that they consistently prefer scope-homomorphic orders despite never observing them during training. These preferences vary in strength by modifier type. To investigate the source of these preferences, we examine noun-modifier association strength using pointwise mutual information (PMI). While PMI reflects known modifier-ordering patterns, it does not explain the models' ordering preferences. These findings demonstrate that LMs can recover human-like linguistic generalizations from impoverished input and provide a controlled framework for investigating the mechanisms underlying such biases.

Chinese Translation

语言习得中的一个核心问题是语言偏见是否可以从对不确定输入的普遍学习机制中产生。人工语言学习（Artificial Language Learning, ALL）研究表明，人类学习者在提供的证据之外可靠地进行泛化，包括偏好范围同态的名词短语修饰符顺序。在本研究中，我们调查语言模型在类似条件下是否表现出相同的偏见。我们创建了一个受控学习环境，在该环境中，模型在一个移除了所有包含多个修饰符的名词短语的语料库上进行训练，从而消除了关于修饰符顺序的直接证据，然后在多个修饰符句子上进行评估。在三种模型规模中，我们发现它们始终偏好范围同态的顺序，尽管在训练期间从未观察到这些顺序。这些偏好根据修饰符类型的不同而有所变化。为了探讨这些偏好的来源，我们使用点互信息（Pointwise Mutual Information, PMI）来检查名词-修饰符的关联强度。虽然PMI反映了已知的修饰符排序模式，但并不能解释模型的排序偏好。这些发现表明，语言模型可以从贫乏的输入中恢复出类似人类的语言泛化，并提供了一个受控框架，以研究这种偏见背后的机制。

View on arXiv Download PDF AI Translation

cs.CL / 66 / 2608.05064

Provable Limits and Certified Deferral for Verbalized Uncertainty in Small Language Models

小型语言模型中可证明的限制与认证的延迟应对口头不确定性

Shen, Jianru

Abstract

Small open-weight language models increasingly run in private, offline, and cost-sensitive settings, where the key deployment question is not only what a model answers but when it should defer to a human. We study whether verbalized confidence can support risk-controlled deferral, evaluating eleven instruction-tuned models from three families, 0.5B to 14B parameters, on ARC-Challenge and TruthfulQA with 25,168 local predictions. Three theoretical results delimit what calibration can provide: strictly monotone calibration preserves the risk-coverage frontier and error-detection AUROC; temperature scaling cannot calibrate models whose confidence stays above one half while accuracy falls below it; and a Clopper-Pearson procedure converts a 200-question calibration set into a finite-sample risk certificate under an i.i.d. deployment assumption. Empirically, eight of 22 model-task pairs hit the temperature-scaling infeasibility floor within one percentage point of the predicted bound. Platt scaling reduces ECE to as low as 0.02, yet certified autonomy at a 20% risk budget is granted to only three model-task pairs and to none at 10%. We also identify and repair an answer-ordering artifact in the multiple-choice form of TruthfulQA. Calibration gives confidence semantics; certified deferral determines when small models are safe to use.

Chinese Translation

小型开放权重语言模型越来越多地在私密、离线和成本敏感的环境中运行，在这些环境中，关键的部署问题不仅是模型的回答内容，还包括何时应将决策延迟到人类。我们研究了口头表达的置信度是否能够支持风险控制的延迟应对，评估了来自三个家族的十一种指令调优模型，参数范围从0.5B到14B，在ARC-Challenge和TruthfulQA上进行了25,168个本地预测。三个理论结果界定了校准所能提供的内容：严格单调的校准保持风险覆盖边界和错误检测的AUROC；温度缩放无法校准那些置信度始终高于0.5而准确率低于0.5的模型；Clopper-Pearson程序将200个问题的校准集转换为在独立同分布（i.i.d.）部署假设下的有限样本风险证书。从经验上看，22个模型-任务对中有8个在预测边界的一个百分点内达到了温度缩放的不可能性底线。Platt缩放将ECE降低到最低0.02，但在20%的风险预算下，只有三个模型-任务对获得了认证自主权，而在10%的风险预算下则没有。我们还识别并修复了TruthfulQA多项选择形式中的答案排序伪影。校准提供了置信度的语义；认证的延迟应对决定了小型模型何时安全使用。

View on arXiv Download PDF AI Translation

cs.CL / 67 / 2608.05075

German parties shifted towards intuition-based rhetoric after the far right's parliamentary breakthrough

德国政党在极右派进入议会后转向基于直觉的修辞

Saleth, Peer, Aroyehun, Segun T., Carrella, Fabio, Abels, Christoph M., Lewandowsky, Stephan, Garcia, David

Abstract

The spread of misinformation is widely perceived as a threat to democratic deliberation, yet how political elites' rhetorical commitments to truth shift alongside the rise of populist actors remains poorly understood. Analysing 4.5 million tweets and 59,170 parliamentary speeches by German political elites between 2015 and 2025, we measure evidence-based and intuition-based rhetoric using a validated distributed dictionary representation. Across both arenas, intuition-based language has become more prominent, and right-leaning actors consistently exhibit the lowest Evidence Minus Intuition (EMI) scores. The parliamentary entry of the extreme-right Alternative for Germany (AfD) in 2017 coincides with sharp downward shifts in EMI across the broader chamber, while a more gradual decline is observed on Twitter. These findings document an association between far-right visibility and a changing approach to truth in elite discourse in a multiparty European democracy.

Chinese Translation

虚假信息的传播被广泛视为对民主审议的威胁，但政治精英对真理的修辞承诺如何随着民粹主义者的崛起而变化仍然鲜有研究。通过分析2015年至2025年间德国政治精英发布的450万条推文和59,170篇议会演讲，我们使用经过验证的分布式词典表示法来衡量基于证据和基于直觉的修辞。在这两个领域中，基于直觉的语言变得更加突出，而倾向于右派的参与者始终表现出最低的证据减去直觉（Evidence Minus Intuition, EMI）得分。2017年极右派德国选择党（Alternative for Germany, AfD）进入议会的同时，EMI在更广泛的议会中出现了急剧下降，而在推特上则观察到更为渐进的下降。这些发现记录了极右派的可见性与多党制欧洲民主中精英话语中真理观念变化之间的关联。

View on arXiv Download PDF AI Translation

cs.CL / 68 / 2608.05097

Same Formulas, Different Semantics: Do Language Models Follow Modal Logic Specifications?

相同的公式，不同的语义：语言模型是否遵循模态逻辑规范？

Andrieu, Réemi, Sileo, Damien

Abstract

Reasoning about necessity and possibility depends on assumptions about accessibility between worlds and about which objects exist at each one. The same inference may therefore hold under one modal system and fail under another. Evaluating language models on such problems requires testing whether their judgments follow the stated semantics rather than a familiar logic. We construct paired modal problems with identical premises and conjecture but different frame or domain conditions; automated reasoning verifies opposite labels. A balanced core prevents the semantic condition alone from revealing the answer. On this core, four of five recent models perform below the condition-only baseline under direct prompting. Yet enabling reasoning mode raises DeepSeek V4 Flash from 4.4% to 88.1% on unchanged prompts. Following stipulated modal semantics thus depends strongly on inference mode as well as model identity. When frame conditions are omitted, models often agree but fit different familiar logics best. We release the formulas, oracle artifacts, countermodels, and responses.

Chinese Translation

关于必要性和可能性的推理依赖于对世界之间可达性的假设以及每个世界中存在的对象。因而，同一推理在一个模态系统下可能成立，而在另一个系统下则可能失效。对语言模型在此类问题上的评估需要测试它们的判断是否遵循所述的语义，而不是熟悉的逻辑。我们构建了具有相同前提和猜想但不同框架或领域条件的成对模态问题；自动推理验证了相反的标签。一个平衡的核心防止了仅凭语义条件揭示答案。在这个核心上，五个最近的模型中有四个在直接提示下的表现低于仅条件基线。然而，启用推理模式使 DeepSeek V4 Flash 在不变的提示下从 4.4% 提升至 88.1%。因此，遵循规定的模态语义在很大程度上依赖于推理模式以及模型的身份。当框架条件被省略时，模型通常达成一致，但最适合不同的熟悉逻辑。我们发布了公式、神谕工件、反模型和响应。

View on arXiv Download PDF AI Translation

cs.CL / 69 / 2608.05124

Chained Recursive Language Models for Multi-Iteration Reasoning

用于多轮推理的链式递归语言模型

Mitra, Purbesh, Ulukus, Sennur

Abstract

Long context reasoning in large language models (LLMs) is usually constrained by the fact that a single inference trajectory has to simultaneously explore the context, store intermediate state, verify evidence, and produce the final answer. This becomes particularly difficult in tasks that require extraction, counting, ordering, or multi-hop reasoning, where an early mistake can propagate until the final response. In this work, we propose Chained Recursive Language Models (Chained RLM), an inference-time architecture, in which the same underlying model is called repeatedly as a sequence of fresh reasoning roots. Each root receives the original problem and context, but does not inherit the full conversational history. Instead, it receives a compact plain-text summary, a plain-text blackboard, and some durable task-specific artifacts written by predecessor roots. The motivation is to manage the context by chopping into partial tasks rather than one large inference response; in each staged computation, intermediate artifacts can be inspected, corrected, and extended by a later fresh inference by the same model. We describe the system model, handoff mechanism, artifact workspace, and evaluation protocol for this system. We study when fresh-context artifact continuation gives a measurable gain in accuracy over direct LLM answering even with recursive tool-calling.

Chinese Translation

大型语言模型（LLMs）中的长上下文推理通常受到限制，因为单一推理轨迹必须同时探索上下文、存储中间状态、验证证据并生成最终答案。这在需要提取、计数、排序或多跳推理的任务中尤为困难，因为早期的错误可能会传播到最终响应。在本研究中，我们提出了链式递归语言模型（Chained RLM），这是一种推理时架构，其中相同的基础模型被重复调用作为一系列新的推理根。每个根接收原始问题和上下文，但不继承完整的对话历史。相反，它接收一个紧凑的纯文本摘要、一个纯文本黑板，以及一些由前驱根写成的持久任务特定工件。其动机在于通过将上下文分割为部分任务而不是一个大型推理响应来进行管理；在每个阶段的计算中，中间工件可以被检查、纠正，并由同一模型的后续新推理进行扩展。我们描述了该系统的模型、交接机制、工件工作区和评估协议。我们研究了何时新上下文工件的延续在准确性上相较于直接的LLM回答提供了可测量的提升，即使在递归工具调用的情况下。

View on arXiv Download PDF AI Translation

cs.CL / 70 / 2608.05126

Spoken Function Calling: A New Perspective on Spoken Language Understanding for Large Audio Language Models

语音功能调用：对大型音频语言模型的语音语言理解的新视角

Peng, Yuezhang, Liu, Yuxin, Gao, Changfeng, Gao, Zhifu, Li, Xiangang, Chen, Xie

Abstract

Spoken Language Understanding (SLU) is the core component of task-oriented dialogue systems and a pivotal link in achieving seamless human-agent interaction. While traditional SLU can effectively extract user semantics for closed-set tasks after in-domain supervised fine-tuning, it faces significant challenges in leveraging in-context learning for open-domain tasks due to its ambiguous rule definitions. This work proposes Spoken Function Calling (SFC), a novel semantic understanding perspective that optimizes semantic understanding with structured rule definitions, to evolve beyond traditional closed-set SLU. Specifically, we curate and extend a suite of spoken functions based on traditional SLU datasets, construct a multi-agent system to synthesize the SFC-Bench dataset, evaluate the performance of Large Language Models (LLMs) and Large Audio Language Models (LALMs), and enhance the SFC capabilities of LALMs through post-training. Experiments demonstrate that SFC outperforms traditional SLU, substantially enhancing the semantic extraction accuracy for LLMs and LALMs.

Chinese Translation

语音语言理解（SLU）是任务导向对话系统的核心组成部分，也是实现无缝人机交互的关键环节。尽管传统的SLU在经过领域内监督微调后能够有效提取用户语义以应对封闭集任务，但由于其模糊的规则定义，在开放域任务中利用上下文学习面临重大挑战。本研究提出了语音功能调用（SFC），一种新颖的语义理解视角，通过结构化规则定义优化语义理解，以超越传统的封闭集SLU。具体而言，我们基于传统SLU数据集策划并扩展了一套语音功能，构建了一个多智能体系统以合成SFC-Bench数据集，评估大型语言模型（LLMs）和大型音频语言模型（LALMs）的性能，并通过后训练增强LALMs的SFC能力。实验表明，SFC在语义提取准确性上显著优于传统SLU，显著提升了LLMs和LALMs的语义提取准确率。

View on arXiv Download PDF AI Translation

cs.CL / 71 / 2608.05139

Toward Skill-Native LLMs: Skill Entropy for Benchmarking and Training Long-Horizon Reasoning

朝向技能原生的大型语言模型：用于基准测试和训练长时间推理的技能熵

He, Yinghui, Yang, Ling, Liu, Jiarui, Yang, Yongjin, Zhang, Lechen, Wu, Yingcheng, Yin, Zhenfei, Wang, Mengdi, Arora, Sanjeev

Abstract

Long-horizon reasoning in recent LLMs demands that the model switch between distinct skills inside a reasoning chain, such as first doing a math derivation, then using the result to plan a schedule. We call such problems cross-skill long-horizon tasks: multi-step tasks whose steps require different reasoning skills and depend on earlier outputs. Existing benchmarks often evaluate individual skills, lacking a principled way to measure how well a model switches between skills. We address this gap from both the evaluation and training sides. We introduce Skill Entropy, a measure of the difficulty of switching from one skill to another. We then propose Skill^2-Bench, a benchmark of cross-skill long-horizon tasks built over 558 skills across 9 verifiable and open-ended domains. Each task is assigned a task-level skill-entropy score and grouped into three difficulty levels. Evaluating 8 frontier and 4 open-source models on Skill^2-Bench reveals a skill-switching gap: accuracy decreases on higher-entropy tasks. We then turn skill entropy from a benchmark scale into a training signal. We propose Skill-Entropy RL, an RL framework where the model predicts not only the answer at each step but also the skill used to produce it. The reward combines step-level correctness with a skill-entropy reward that measures the alignment between the model-predicted skill sequence and the gold skill sequence. On Qwen3-4B-Instruct and Qwen3-1.7B, Skill-Entropy RL improves the Skill^2-Bench score from 34.4% to 68.4% and from 14.6% to 40.1%, respectively, outperforming competitive baselines. The same pipeline can be applied to off-the-shelf training data such as OpenR1-Math, indicating that skill entropy is a reusable training signal. Code available at: https://github.com/Gen-Verse/Skill-Entropy-RL

Chinese Translation

近期大型语言模型（LLMs）中的长时间推理要求模型在推理链中切换不同的技能，例如首先进行数学推导，然后利用结果来规划日程。我们将此类问题称为跨技能长时间任务：这些多步骤任务的步骤需要不同的推理技能，并依赖于之前的输出。现有的基准测试通常评估单一技能，缺乏一种原则性的方法来衡量模型在技能之间切换的能力。我们从评估和训练两个方面解决了这一空白。我们引入了技能熵（Skill Entropy），作为从一种技能切换到另一种技能的难度度量。然后，我们提出了Skill^2-Bench，这是一个基于558种技能构建的跨技能长时间任务基准，涵盖9个可验证和开放式领域。每个任务被分配一个任务级别的技能熵分数，并分为三个难度级别。在Skill^2-Bench上评估8个前沿模型和4个开源模型揭示了一个技能切换差距：在高熵任务上的准确性下降。接着，我们将技能熵从基准规模转化为训练信号。我们提出了技能熵强化学习（Skill-Entropy RL），这是一个强化学习框架，其中模型不仅预测每一步的答案，还预测产生该答案所使用的技能。奖励结合了步骤级别的正确性与技能熵奖励，后者衡量模型预测的技能序列与真实技能序列之间的对齐程度。在Qwen3-4B-Instruct和Qwen3-1.7B上，技能熵强化学习将Skill^2-Bench的得分从34.4%提高到68.4%，从14.6%提高到40.1%，超越了竞争基准。相同的流程可以应用于现成的训练数据，如OpenR1-Math，表明技能熵是一个可重复使用的训练信号。代码可在以下链接获取：https://github.com/Gen-Verse/Skill-Entropy-RL

View on arXiv Download PDF AI Translation

cs.CL / 72 / 2608.05148

Reasoning Core: Designing Broad Procedural Data for Completion-Supervised Reasoning Training

推理核心：设计广泛的程序性数据用于完成监督推理训练

Sileo, Damien, Lacombe, Valentin, Kachler, Dimitri

Abstract

Procedural generators produce useful verifiable reasoning problems at scale, but have received less attention as data for completion-supervised fine-tuning. We introduce Reasoning Core, a collection of 50 generators spanning mathematics, logic, planning, state tracking, formal languages, structured data, games, causality, and code, with semantic scorers, difficulty controls, and task evaluators. Under a matched completion-supervised protocol, we compare Reasoning Core with Procedural Warmup, Reasoning Gym, and SynLogic across four base-model settings and multiple training durations. In the primary 3B comparison, Reasoning Core achieves the highest mean scores on DROP, LogiQA, and ARC-Challenge, exceeding both the baseline without procedural data and all three alternative procedural collections. Task-level analyses show that semantic validity alone does not ensure training utility, highlighting compact targets and calibrated difficulty as important design factors. We ran audits combining model-assisted review, human adjudication, and regression testing. Applied throughout Reasoning Core development and to the other collections, they reveal subtle mismatches among generation, rendering, targets, and scoring, a reminder that procedural generation alone does not guarantee correctness. The library, generated datasets, and audit material are publicly available.

Chinese Translation

程序生成器能够大规模产生有用的可验证推理问题，但作为完成监督微调的数据却受到较少关注。我们引入了推理核心（Reasoning Core），这是一个包含50个生成器的集合，涵盖数学、逻辑、规划、状态跟踪、形式语言、结构化数据、游戏、因果关系和代码，配备语义评分器、难度控制和任务评估器。在匹配的完成监督协议下，我们将推理核心与程序热身（Procedural Warmup）、推理健身房（Reasoning Gym）和SynLogic进行了比较，涵盖四种基础模型设置和多种训练时长。在主要的3B比较中，推理核心在DROP、LogiQA和ARC-Challenge上取得了最高的平均分，超过了没有程序性数据的基线和所有三种替代程序集合。任务级分析表明，仅有语义有效性并不能确保训练的实用性，强调了紧凑目标和校准难度作为重要设计因素。我们进行了审计，结合了模型辅助审查、人为裁决和回归测试。在推理核心开发和其他集合中应用的审计揭示了生成、呈现、目标和评分之间的微妙不匹配，提醒我们程序生成本身并不能保证正确性。该库、生成的数据集和审计材料均已公开。

View on arXiv Download PDF AI Translation

arXiv Papers

Kitchen Robotic Manipulation utilizing Foundation Models

Interpretable Fuzzy Inference for UAV Target Tracking Using Bounding-Box Geometry

SiMDex: Mining Similar Egocentric Videos for Cross-Embodiment Dexterous Manipulation

Feasibility of Embedded Photoplethysmography Sensing in Short-Duration Tactile Interactions With Pocket-Sized Robots Using IMU- and Confidence-Based Filtering

SAFECAST: Robust Failure Detection for VLA Policies with Contrast-Set Training and Calibration

Structured LLM Reasoning for Zero-Shot Human--Robot Coordination Under Hidden Goals

Design and Flight of an Ion-propelled Micro Hovercraft Leveraging Ground Proximity Effects

Approximate Multi-Objective Search Under Rulebooks

SCOPE: Field-of-View-Aware Path Planning in Unknown 3D Environments via Safety-Volume Certification

SSC: A Verifiable Structured Representation for Bimanual Manipulation Labelling

GUARD: Grounding Uncertainty and Ablation-Based Risk Detection for Diffusion-Based VLAs

Retrieve in Time, Correct in Frequency

Static Timing Orchestration for Tree-Structured Robot Control Firmware

GASP: GPU-Accelerated Safe Planner for Real-Time Collision-Aware Motion Generation with Latent Trajectory Sampling

Mind-VLA: Instruction-Aware Spatial Representation Alignment for Vision-Language-Action Models

Suppression Sticks, Locality Is Fragile: A Closed-Loop Target-and-Control Audit of Task-Vector Negation in VLA Policies

Enabling Urgency-aware Robot Swarm Intralogistics using Smart IoT Tags

A Vision-based Control Framework for Real-time Autonomous UUV Operations

A GitOps-Driven Annotation Catalog for Fully Automatic Railway Operations

Explicit Language Memory for Long-Horizon Planning in Vision-Language-Action Models

From Transparent Labware Segmentation to Collision Avoidance: A Real-Time Edge-Aware Perception Pipeline

Deliberate Before You Fly: Vision-Guided Spatial Deliberation for UAV See-and-Reach Navigation

RORA: Realistic Object Reconstruction with Articulation

PRIMAL3: Pathfinding via Reinforcement and Imitation Multi-Agent Learning - Leveraging LaCAM3

Mimir: A Neuro-Symbolic Memory System with Dynamic Grounding for Embodied Agents in Interactive Environments

DreamWAM: Beyond RGB Future Prediction for World Action Models

Optimal Constrained sc-LTL Planning in MDPs via Switching Policies

BridgeVLA++: A Data-Efficient, Generalizable, and Memory-Augmented Vision-Language-Action Framework for 3D Manipulation

SpikingNav: Robust Embodied Navigation with Spiking Neural Policies

Advancing Utility Pole and Sign Detection Through Deep Learning

LoRetta: A Foundation Model and Extensive Dataset for Global-Scale Remote Sensing Dense Image Matching

GEB-Bench: Abstract Structures Told in Many Voices

Perception Before Reasoning: Dynamic Latent Reasoning for Video Understanding and Question Answering

Teaching Foundation Models to Read mmWave: Pose-Guided Kinematic Representation for Human Behavior Understanding

Radar4D-VLM: Proposal-Grounded Temporal 4D Radar Reasoning Across Frozen Language Models

RUTA: Principled Visual Token Allocation via Rate-Utility Optimization

TRNet: Topography-Guided Frequency Rectification and Structure-Aware Decoding for Multimodal Paddy Rice Segmentation

TriCLE: Tri-Modal Vision-Language Reasoning for Edge-Deployed Fine-Grained Clustering

PADFormer: Pose-agnostic Anomaly Detection from Sparse View Images

OmniVR: Joint Video-Audio Conditional Generation for Restoring Degraded Historical Films

SIGNPOST-Bench: Benchmarking Text-Vision Conflict Resolution in Multimodal Large Language Models

Binding Biometrics with AI Agent Identifiers for Delegation of Authority

CLIP-CC-Bench: Evaluating Paragraph-Level Video Descriptions in Video-Language Models

An Analysis and Implementation of Seam Carving for Content-Aware Image Resizing

iStructTab: Structured Feature Sequencing for Multimodal Learning of Image and Tabular Data

Poly-OPD: Heterogeneous Multi-Teacher On-Policy Distillation for Capability-Selectable Flow Models

Image Classification Using CNN-QNN Hybrid Model with Optimized Correlated Features

ReGround: Restoring Visual Grounding in Multi-Step Reasoning through Self-Diagnosis and Visual Re-Examination

Free-Lunch Augmentation by Revisiting Diffusion-Based Data Generation for Cross-Domain Few-Shot Object Detection

CofactVLA: Deconfounding Vision-Language-Action Models via Counterfactual Intervention

Faster-WAM: Efficient Inference-Time Future Conditioning for Robust World Action Models

muSync-GS: Physics-Synchronized Driving Video Synthesis for Weather and Geometric Road Hazards

Foreseeing the Invisible: Amodal Reconstruction of Leaf Fossil Images

Thinking with Anchors: Grounded and Efficient Document Reasoning

Predict, Then Retrieve: Cross-Instance Future-State Retrieval from Video Prefixes

UBLLIE: Unified Backlight and Low-Light Image Enhancement

OmniRouting: A Semantic-Coupled Multimodal Benchmark for Constraint-Aware Spatial Reasoning in PCB Routing

ToolArtist: Tool-Using Unified Multimodal Models for Agentic Image Generation

Season: Spectrum-Aware Orthogonal Gradient Refinement for Transfer-Based Adversarial Attacks

When does training on downscaled images yield the same gradients?

Q-CueGraph: Query-Conditioned Visual Evidence Graphs for Multimodal Reasoning

TwinIR: Coordinated Invisible Dual-Point Attacks on Online HD Map Construction

Beyond Global Routing Aggregation: Phase-Aware Expert Merging for MoE Vision-Language Models

EndoVLM: An Endoscopy Vision-Language Pre-training Model via Anatomy-Guided Sparsity and Progressive Alignment

REZE: Recognition-Based Zero-Shot Extraction for Video Temporal Grounding

Not All Redundant Tokens Are Alike: Analyzing Visual Token Pruning through Token Roles

DIVE: Dynamic Iterative Visual Evidence Construction for Efficient Vision-Language Models

Privacy-Preserving Action Recognition: Taxonomy, Methods, and Privacy-Utility Trade-offs

GeoReward: Mitigating Contextual Variable Overestimation in Vision-Language Models for Cross-Market Preference Prediction

CARVE: Cross-Slice Anisotropic Reallocation of Visual Evidence for Efficient 3D Medical Volume Understanding

Coupled Continuous-Discrete Generation for Scene Text Image Super-Resolution

FocusMem: Factorizing Content, Readout, and Trust in Latent GUI Memory

EgoAfford: Task-Oriented Affordance Grounding via Egocentric Referring Segmentation

VoxStruct3D: Structure-Leading Flow Matching for Voxel-Space 3D MRI Synthesis

ColorFD: A Finite-Difference Guided Black-Box Physical Adversarial Attack for Remote Sensing Object Detection

OutLangSplat: 3D Language Gaussian Splatting for UAV Outdoor Scenes

Talk2Sensors: 3D Visual Grounding in Autonomous Driving via Sensor-Adaptive Physical Cue Matching

PhysMind: From Video to Executable Worlds for Training-Free Physical Reasoning

ACA-GS: Adaptive-Capacity Anchored Gaussian Splatting for Compact Dynamic Radiance Fields