← Back to Index
Daily Research Digest

arXiv Papers

2026-08-27
257
Papers
4
Categories
257
Translated
收藏清单 0
机器人学 (Robotics)
43
cs.RO / 1 / 2608.24959

GaussVLA: Geometry-Aware Spatial Reasoning for Vision-Language-Action Model

GaussVLA：面向视觉-语言-动作模型的几何感知空间推理
Sarowar, Md Selim, Islam, Md Tanvir, Kim, Sungho, Ahn, Sangtae
Abstract
Vision-Language-Action (VLA) models encode visual observations as flat 2D patch tokens that carry no intrinsic geometric structure, and augmenting them with dense monocular depth injects per-pixel scalar values that encode neither surface orientation nor geometric confidence. This leaves the policy with limited structured spatial reasoning for action prediction. We propose GaussVLA, a Mamba-based VLA that incorporates two custom modules: Gaussian Spatial Tokenizer (GST) to lift frozen semantic and depth features into compact 3D Gaussian tokens, pools geometrically salient regions with learned queries, and \emph{Depth-Aware Chain-of-Thought (DA-CoT)} that performs structured, non-autoregressive geometric reasoning under language and flow-time conditioning. Across both simulation and real-world evaluations, GaussVLA demonstrates strong spatial-manipulation performance while remaining parameter-efficient. On LIBERO, it achieves 93.5% average success and 100.0% success on the Spatial suite with only 200M parameters, improving over SpatialVLA by 19.7% relative average success while remaining significantly more parameter-efficient.
Chinese Translation
视觉-语言-动作（VLA）模型将视觉观察编码为平坦的二维补丁标记，这些标记不携带内在的几何结构，并且通过稠密的单目深度增强它们时，注入的每像素标量值既不编码表面方向也不编码几何置信度。这使得策略在动作预测中具有限制的结构化空间推理能力。我们提出了GaussVLA，这是一种基于Mamba的VLA，结合了两个自定义模块：高斯空间标记器（Gaussian Spatial Tokenizer, GST），用于将冻结的语义和深度特征提升为紧凑的三维高斯标记，并通过学习的查询对几何显著区域进行池化，以及 extit{深度感知思维链（Depth-Aware Chain-of-Thought, DA-CoT）}，在语言和流时间条件下执行结构化的非自回归几何推理。在模拟和现实世界评估中，GaussVLA展示了强大的空间操作性能，同时保持了参数效率。在LIBERO上，它实现了93.5%的平均成功率和在空间套件中100.0%的成功率，仅使用200M参数，相较于SpatialVLA提高了19.7%的相对平均成功率，同时显著提高了参数效率。
View on arXiv Download PDF AI Translation
cs.RO / 2 / 2608.25102

ROS2 Connect: A new ROS2 over WAN Solution

ROS2 Connect：一种新的广域网（WAN）下的ROS2解决方案
Schott, Daniel, Srinivasan, Lakshminarasimhan, Herrmann, Christian, Nüchter, Andreas
Abstract
The Robot Operating System 2 (ROS2) has become a widely adopted framework for the development of distributed robotic systems. However, its communication architecture, based on DDS and RTPS, relies on multicast discovery mechanisms that are typically unavailable in wide-area network (WAN) environments, making remote operation challenging. This work presents ROS2 Connect, a WebSocket-based communication framework that enables transparent and secure ROS2 interaction across routed networks without requiring modifications to network infrastructure or DDS configurations. The proposed client-server architecture supports bidirectional exchange of topics, services, actions, and system data while integrating authentication and access control mechanisms. Experimental evaluation over a real WAN connection demonstrates significantly lower latency, higher stability, and improved scalability compared to existing solutions, including DDS Router, rosbridge and Zenoh. Initial results show that ROS2 Connect provides a reliable foundation for teleoperation and distributed robotics applications over wide-area networks.
Chinese Translation
机器人操作系统2（ROS2）已成为开发分布式机器人系统的广泛采用框架。然而，其基于DDS和RTPS的通信架构依赖于通常在广域网（WAN）环境中不可用的多播发现机制，这使得远程操作变得具有挑战性。本研究提出了ROS2 Connect，一种基于WebSocket的通信框架，能够在路由网络中实现透明和安全的ROS2交互，而无需对网络基础设施或DDS配置进行修改。所提出的客户端-服务器架构支持主题、服务、动作和系统数据的双向交换，同时集成了认证和访问控制机制。在真实的WAN连接上的实验评估表明，与现有解决方案（包括DDS Router、rosbridge和Zenoh）相比，ROS2 Connect显著降低了延迟，提高了稳定性，并改善了可扩展性。初步结果表明，ROS2 Connect为广域网下的远程操作和分布式机器人应用提供了可靠的基础。
View on arXiv Download PDF AI Translation
cs.RO / 3 / 2608.25135

Extending Ground-Constraint LiDAR-IMU Calibration to Tilted Surfaces in a Continuous-Time Framework

将地面约束的LiDAR-IMU标定扩展至倾斜表面的连续时间框架
Korotkine, Vassili, Chamoun, Pierre, Shalaby, Mohammed Ayman, Forbes, James Richard
Abstract
This paper presents a novel method that extends targetless LiDAR-IMU calibration for ground vehicles to non- flat environments. Calibration typically necessitates full exci- tation of the sensor rig, a requirement that is not fulfilled by ground vehicles in normal operation. To address the degenerate planar motion, state-of-the-art methods propose residuals that assume the colinearity of the gravity and physical surface normal vectors, restricting usage to cases where the ground is assumed flat. This paper proposes ground-plane residuals that do not require this assumption, and are applicable for planar motion on a tilted surface. Results are demonstrated on a dataset collected from a Husky ground vehicle, on the M2DGR dataset, as well as on an offroad vehicle dataset. Repeatability is shown to be improved both in tilted and flat-ground scenarios, with strong improvement demonstrated for the tilted case. The implementation and experiments are open-sourced at https://github.com/vkorotkine/licalib_tilted_ground.
Chinese Translation
本文提出了一种新方法，将无目标LiDAR-IMU标定扩展至非平坦环境。标定通常需要对传感器设备进行全面激励，而这一要求在正常操作的地面车辆中并未得到满足。为了解决退化的平面运动，最先进的方法提出了假设重力和物理表面法向量共线的残差，这限制了其在假设地面为平坦的情况下的使用。本文提出的地面平面残差不需要这一假设，适用于倾斜表面的平面运动。结果在从Husky地面车辆收集的M2DGR数据集以及一组越野车辆数据集上得到了验证。实验表明，在倾斜和平坦地面场景中，重复性得到了改善，尤其在倾斜情况下表现出显著的提升。该实现和实验已开源，地址为 https://github.com/vkorotkine/licalib_tilted_ground。
View on arXiv Download PDF AI Translation
cs.RO / 4 / 2608.25142

SkyDrive: Learning to Drive in a New City from Aerial Traffic Monitoring

SkyDrive：从空中交通监测中学习在新城市驾驶
Xiong, Weijiang, Feng, Lan, Alahi, Alexandre, Geroliminis, Nikolas
Abstract
Autonomous driving has made remarkable progress through imitation learning with massive human demonstration data. However, a trained planner often degrades severely when applied to a new environment zero-shot, because of domain shifts in traffic regulations, road layout and driving behaviors. Therefore, adapting a trajectory planner to a new city typically requires resource-demanding local data collection with a vehicle sensor suite. In this work, we show that driving behavior can be learned from a scalable and efficient alternative. We introduce \emph{SkyDrive}, a framework that utilizes drone-based traffic monitoring to provide efficient supervision for autonomous driving agents in a new environment. While vehicle-based data collection logs the ego and its surroundings, an aerial platform naturally observes many road users simultaneously over an extended field of view. As a result, every vehicle can be a data source with grounded driving behavior, effectively scaling up the amount of supervision. Based on 137 hours of aerial traffic monitoring footage, we extract 650K driving samples and construct a benchmark for trajectory planners and motion predictors. Zero-shot experiments with multiple models reveal significant cross-city domain gaps, but many of them can be alleviated by limited supervision from the sky, e.g., 30 minutes of monitoring per location. Our findings show that aerial traffic monitoring is an efficient and scalable data source for adapting autonomous driving systems in new cities. Data and code will be made publicly available.
Chinese Translation
自主驾驶通过模仿学习和大量人类示范数据取得了显著进展。然而，当经过训练的规划器在新的环境中零样本应用时，通常会严重退化，因为交通法规、道路布局和驾驶行为存在领域转移。因此，将轨迹规划器适应于新城市通常需要资源密集型的本地数据收集，使用车辆传感器设备。在本研究中，我们展示了驾驶行为可以通过一种可扩展且高效的替代方案进行学习。我们引入了 extit{SkyDrive}，一个利用无人机基础的交通监测为新环境中的自主驾驶代理提供高效监督的框架。虽然基于车辆的数据收集记录了自我及其周围环境，但空中平台自然可以在扩展的视野内同时观察许多道路使用者。因此，每辆车都可以成为具有基础驾驶行为的数据源，有效地扩大了监督的数量。基于137小时的空中交通监测视频，我们提取了65万条驾驶样本，并构建了轨迹规划器和运动预测器的基准。对多个模型的零样本实验揭示了显著的跨城市领域差距，但其中许多差距可以通过来自空中的有限监督来缓解，例如每个位置30分钟的监测。我们的研究结果表明，空中交通监测是适应新城市中的自主驾驶系统的高效且可扩展的数据源。数据和代码将公开提供。
View on arXiv Download PDF AI Translation
cs.RO / 5 / 2608.25162

Sequential Object Placement Optimization with Convex Decomposition

基于凸分解的顺序物体放置优化
Zhang, Yuezhe, Lyu, Xiangyu, Rudra, Sohan, Tateo, Davide, Chalvatzaki, Georgia
Abstract
Robotic object packing has been a core challenge for robotic deployment in logistics, industry, etc., due to the curse of dimensionality in combinatorial search and the difficulty of dealing with dynamic and contact constraints for irregularly shaped objects. Current heuristic and learning-based methods assume a limited spatial discretization resolution of space, and computation becomes extremely inefficient as discretization accuracy increases. In this work, we eliminate these assumptions by introducing SOPO-CD, a sequential optimization framework that frames object placement as a differentiable nonlinear optimization problem in a decomposed free space. We prove that placing a convex object inside a convex hull is essentially constraining the vertices of the object inside the convex hull. The constraints and their derivatives can be written in closed form and calculated within $200$ns. We implement a custom solver that achieves optimal placement within tightly constrained space in milliseconds; a $100 \times$ speedup compared to a classical grid search method. We generalize our framework to 2D Tangram, 2D Tetris, and 3D Bin Packing, and have demonstrated strong computational performance and packing utility. We also demonstrate solving a real-world Tangram puzzle online using an Allegro Hand and an Xarm.
Chinese Translation
机器人物体打包一直是机器人在物流、工业等领域部署的核心挑战，这主要是由于组合搜索中的维度诅咒以及处理不规则形状物体的动态和接触约束的困难。目前的启发式和基于学习的方法假设空间的空间离散化分辨率有限，随着离散化精度的提高，计算变得极其低效。在本研究中，我们通过引入SOPO-CD（顺序优化框架），消除了这些假设，将物体放置视为在分解的自由空间中一个可微分的非线性优化问题。我们证明，将一个凸物体放置在凸包内本质上是将物体的顶点约束在凸包内。这些约束及其导数可以用封闭形式表示，并在$200$纳秒内计算。我们实现了一个自定义求解器，能够在毫秒内实现紧约束空间内的最优放置；与经典的网格搜索方法相比，速度提升达$100 imes$。我们将框架推广到二维拼图（Tangram）、二维俄罗斯方块（Tetris）和三维箱子打包（Bin Packing），并展示了强大的计算性能和打包效用。我们还展示了使用Allegro Hand和Xarm在线解决实际的拼图（Tangram）问题。
View on arXiv Download PDF AI Translation
cs.RO / 6 / 2608.25171

Control-Oriented Learning for Dynamic Tracking and Stability Analysis of Soft Pneumatic Actuators

面向控制的学习用于软气动执行器的动态跟踪与稳定性分析
Kumar, Nithin S., Barth, Eric J.
Abstract
Soft pneumatic actuators offer inherent compliance and safe interaction but remain difficult to model and control because of their highly nonlinear, distributed dynamics. We present a control-oriented data-driven modeling and control framework that decomposes actuator behavior into a nonlinear static equilibrium model and a linear residual dynamics model identified using Extended Dynamic Mode Decomposition with control (EDMDc). This representation enables feedforward compensation, task-space feedback control, and local closed-loop stability analysis through an augmented linear model. Experiments achieve approximately 1 mm root mean square error (RMSE) during low-speed (approximately 10 mm/s) trajectory tracking and below 10 mm RMSE at higher speeds (approximately 100 mm/s). The framework further achieves stable tracking of highly dynamic user-generated references with peak accelerations exceeding 25 m/s^2 while simultaneously performing real-time obstacle avoidance. Finally, the proposed stability analysis is experimentally validated by accurately predicting stable, marginal, and unstable operating regimes. These results demonstrate that structured, control-oriented learning provides an accurate and practical framework for soft actuator control.
Chinese Translation
软气动执行器具有固有的柔顺性和安全交互性，但由于其高度非线性和分布式动态特性，建模和控制仍然困难。我们提出了一种面向控制的数据驱动建模与控制框架，该框架将执行器行为分解为非线性静态平衡模型和通过带控制的扩展动态模式分解（EDMDc）识别的线性残余动态模型。这种表示方法使得通过增强线性模型实现前馈补偿、任务空间反馈控制和局部闭环稳定性分析成为可能。实验结果表明，在低速（约10 mm/s）轨迹跟踪过程中，均方根误差（RMSE）约为1 mm，而在较高速度（约100 mm/s）下，RMSE低于10 mm。该框架进一步实现了对用户生成的高动态参考信号的稳定跟踪，峰值加速度超过25 m/s²，同时进行实时障碍物避让。最后，所提出的稳定性分析通过准确预测稳定、临界和不稳定的操作状态得到了实验验证。这些结果表明，结构化的面向控制的学习为软执行器控制提供了一个准确且实用的框架。
View on arXiv Download PDF AI Translation
cs.RO / 7 / 2608.25192

CRESSim-Neo: A Batched GPU Simulation Engine for Surgical Robotics and Robot Learning

CRESSim-Neo：一种用于外科机器人和机器人学习的批处理GPU仿真引擎
Ou, Yafei, Naheen, Ahnaf, Mussin, Tleukhan, Jarales, Hans, Moncy, Melwin, Tavakoli, Mahdi
Abstract
We introduce CRESSim-Neo, a batched GPU simulation engine for surgical robotics and robot learning. CRESSim-Neo combines position-based simulation of rigid bodies, deformable tissues, fluids, and strands with batched rendering, surgery-specific sensing, and a GPU-resident data pipeline. The engine supports applications including tissue manipulation, fluid suction, suturing, cable-driven robots, and ultrasound image synthesis. Direct access to physics and rendering buffers enables GPU-resident robot learning and zero-copy PyTorch integration using DLPack. We demonstrate CRESSim-Neo across rigid-body, deformable-body, and fluid simulation tasks, including vision-based and surgical robot-learning scenarios. On an NVIDIA RTX 4090, the engine achieves up to 2.03 million environment steps per second for 8192 parallel CartPole environments, and scales to batched surgical scenarios involving tissue deformation, fluid interaction, and ultrasound sensing. Overall, CRESSim-Neo provides a unified and scalable platform for surgical simulation, synthetic data generation, and surgical robot learning.
Chinese Translation
我们介绍了CRESSim-Neo，这是一种用于外科机器人和机器人学习的批处理GPU仿真引擎。CRESSim-Neo结合了刚体、可变形组织、流体和绳索的基于位置的仿真，批处理渲染，特定于手术的传感器以及GPU驻留的数据管道。该引擎支持包括组织操作、流体吸引、缝合、缆驱动机器人和超声图像合成等应用。直接访问物理和渲染缓冲区使得GPU驻留的机器人学习和使用DLPack的零拷贝PyTorch集成成为可能。我们在刚体、可变形体和流体仿真任务中展示了CRESSim-Neo，包括基于视觉的和外科机器人学习场景。在NVIDIA RTX 4090上，该引擎在8192个并行CartPole环境中每秒可达到高达203万环境步骤，并扩展到涉及组织变形、流体交互和超声传感的批处理外科场景。总体而言，CRESSim-Neo提供了一个统一且可扩展的外科仿真、合成数据生成和外科机器人学习平台。
View on arXiv Download PDF AI Translation
cs.RO / 8 / 2608.25196

Longitudinal Robot Learning from Demonstration with Care Providers in a Home Environment

在家庭环境中与护理提供者进行的纵向机器人示范学习
Moorman, Nina, Schalkwyk, Julianna, Srihari, Vriksha, Xiao, Qingyu, Alrashedy, Kamel, Jeong, Hongseok, Lange, Kiersten, Luebbers, Matthew B., Gombolay, Matthew
Abstract
Learning from demonstration (LfD) methods enable non-expert end users to teach robots novel skills without explicit programming. However most evaluations of the usability of LfD with non-experts has been conducted in controlled laboratory environments with a robotics experimenter present. In this work we identify non-expert end users' key barriers when teaching robots via demonstration without live robotics expert feedback in a home environment. In our human subjects experiment we support the non-expert end users through two forms of demonstrator guidance developed in prior work: pre-training and adaptive feedback. Towards the ecological validity of the evaluation, we conduct this experimentation over multiple visits, with a population of care providers. Finally, we propose to open source the resulting LfD dataset of care providers teaching a robot assistive tasks over multiple visits to a home environment.
Chinese Translation
学习示范（LfD）方法使非专业的最终用户能够在没有明确编程的情况下教机器人新的技能。然而，大多数关于非专家用户使用LfD的可用性评估是在控制实验室环境中进行的，并且有机器人实验者在场。在本研究中，我们识别了非专家最终用户在没有实时机器人专家反馈的家庭环境中通过示范教机器人时面临的主要障碍。在我们的人体实验中，我们通过先前工作中开发的两种示范者指导形式来支持非专家最终用户：预训练和自适应反馈。为了提高评估的生态有效性，我们在多次访问中进行此实验，参与者为护理提供者。最后，我们提议将护理提供者在多次访问家庭环境中教机器人辅助任务的LfD数据集开源。
View on arXiv Download PDF AI Translation
cs.RO / 9 / 2608.25222

Development of a Voice-Controlled Tendon-Driven Bionic Hand

基于语音控制的腱驱动仿生手的开发
Kohli, Urja, Chanda, Shagata, Gandhi, Kritika, Nigam, Charu, Kamal, Aditi Surya, Bhati, Pooja
Abstract
The impairment of the hands can seriously affect the abilities of every individual to perform the every-day activity, so the design of stable and controllable support devices is a significant field of study. This paper is about the design and implementation of an automated bionic hand which is dedicated to the coordinated finger movement through the simplified and efficient actuation mechanism. The method that the proposed system was designed on is the tendon-based method whereby the servo motors generate the movement of the fingers, with assistance of the angular control which is calibrated. An actuation is controlled by a microcontroller that will be programmed by use of an Arduino-based microcontroller to carry out programmed gestures that include open hand, fist, pinch and half flexion. It has an interface that is voice command enabled to make it easy to interact with a Bluetooth based sender receiver architecture which offers an option of executing trained commands which are immediately converted to finger actions. To explore the motions behavior, finger coordination and control response to the input, the behavior of the experiment system is tested. The actuation of the fingers was found to take a total of about 7-8 seconds to achieve full flexion of all fingers in a sequence. The system showed repetitive and constant motion throughout several actuation cycles without loss of any apparent tension or precision of control. There was a stable grasp of objects of different shapes and sizes, which implied consistent coordination between the fingers. These findings indicate that the proposed system offers predictable and steady control behavior and has a simple and efficient mechanical and control architecture.
Chinese Translation
手部功能障碍会严重影响个体进行日常活动的能力，因此设计稳定且可控的辅助设备是一个重要的研究领域。本文讨论了一种自动化仿生手的设计与实现，该仿生手旨在通过简化且高效的驱动机制实现协调的手指运动。所提出系统的设计方法基于腱驱动原理，伺服电机产生手指运动，并辅以经过校准的角度控制。驱动由微控制器控制，该微控制器使用基于Arduino的编程来执行包括张开手、握拳、捏合和半屈曲在内的预设手势。系统配备了语音命令接口，便于与基于蓝牙的发射接收架构进行交互，提供执行训练命令的选项，这些命令会立即转换为手指动作。为了探索运动行为、手指协调和对输入的控制响应，实验系统的行为进行了测试。结果发现，手指的完全屈曲需要大约7-8秒的时间。系统在多个驱动周期中表现出重复且稳定的运动，没有明显的张力或控制精度的损失。对不同形状和大小物体的稳定抓取表明手指之间的协调性一致。这些发现表明，所提出的系统提供了可预测且稳定的控制行为，并具有简单高效的机械和控制架构。
View on arXiv Download PDF AI Translation
cs.RO / 10 / 2608.25284

Generative Action-Chunk Sampling for Adaptive Stiffness Control in Physical Human-Robot Collaboration

基于生成动作块采样的物理人机协作自适应刚度控制
Otake, Aoi, Hartmann, Ferdinand, Igari, Ko, Murata, Shingo
Abstract
Physical human-robot collaboration requires a robot to provide assistance when human intention is clear while remaining compliant when several future motions are plausible. We present an adaptive stiffness framework based on generative action-chunk sampling. Conditioned on an RGB image and external joint-torque estimates, the policy samples multiple future action chunks from an observation-conditioned prior. Variation among the sampled action chunks is used to continuously adapt joint stiffness and damping. Greater variation makes the robot more compliant to facilitate human guidance, whereas lower variation provides firmer assistance. In a real-world collaborative transport task with four possible directions, the proposed method achieved an average success rate of 0.95, compared with 0.83 for a fixed-stiffness ablation and 0.69 for a deterministic baseline. Near direction determination, variation among the sampled action chunks increased and the controller accordingly reduced stiffness. These results suggest that variation among actions sampled by a generative policy can serve as an online control signal for balancing assistance and compliance in physical human-robot interaction.
Chinese Translation
物理人机协作要求机器人在人的意图明确时提供帮助，而在多种未来动作可能的情况下保持顺应性。我们提出了一种基于生成动作块采样的自适应刚度框架。在RGB图像和外部关节扭矩估计的条件下，策略从观察条件的先验中采样多个未来动作块。采样动作块之间的变异性用于持续调整关节刚度和阻尼。较大的变异性使机器人更具顺应性，以便于人类引导，而较小的变异性则提供更坚实的支持。在一个具有四个可能方向的真实协作运输任务中，所提出的方法实现了0.95的平均成功率，而固定刚度消融实验为0.83，确定性基线为0.69。在接近方向确定时，采样动作块之间的变异性增加，控制器相应地降低了刚度。这些结果表明，生成策略采样的动作之间的变异性可以作为在线控制信号，用于平衡物理人机交互中的支持与顺应性。
View on arXiv Download PDF AI Translation
cs.RO / 11 / 2608.25366

RAEM: Robust Autonomous Exploration for Multi-Floor Environments with a Quadruped Robot

RAEM：用于多层环境的四足机器人鲁棒自主探索框架
Yuan, Zikang, Ren, Yuan, Wang, Yian, Wang, Yixue, Fang, Enze, Zhang, Xuewei, Cheng, Junda, Chen, Chi, Ho, Chin-Pang, Zhu, Lijun, Xu, Shaohang, Cheng, Kwang-Ting, Yang, Xin
Abstract
In this paper, we propose RAEM, a robust autonomous exploration framework for quadruped robots operating in multi-floor environments. Most existing ground-robot exploration approaches rely on planar traversability representations, which cannot adequately represent the overlapping structures and cross-floor connectivity of multi-floor buildings. Although tomography-based representations provide effective traversability modeling for multi-floor navigation, maintaining a global tomography map incurs substantial computational overhead for online exploration with frequent replanning. Moreover, sparse and fragmented LiDAR observations in stairwells can degrade local traversability estimation, leading to irregular viewpoint placement and temporary topological disconnections. To address these challenges, RAEM adopts a hybrid local-global traversability representation, in which a local tomography map and an explicitly categorized local 3D grid map are used for online terrain analysis and connectivity evaluation, while an elevation-aware global topological graph is incrementally constructed from these local spatial representations for efficient cross-floor exploration planning. We further introduce a staircase center alignment strategy to reduce abrupt yaw variations during climbing and a dual path searching mechanism to recover guidance paths when the global topology is locally disconnected. Extensive simulation and real-world experiments demonstrate robust and computationally stable autonomous exploration across multi-floor structures, including continuous exploration of a five-floor stairwell.
Chinese Translation
在本文中，我们提出了RAEM，一个用于在多层环境中操作的四足机器人鲁棒自主探索框架。现有的大多数地面机器人探索方法依赖于平面可 traversability 表示，这无法充分表示多层建筑的重叠结构和跨层连接性。尽管基于断层成像的表示为多层导航提供了有效的可 traversability 建模，但维护全球断层图会给在线探索带来大量的计算开销，尤其是在频繁重新规划的情况下。此外，楼梯间稀疏和碎片化的激光雷达观测可能会降低局部可 traversability 估计，导致视点放置不规则和临时拓扑断开。为了解决这些挑战，RAEM 采用了一种混合局部-全球可 traversability 表示，其中局部断层图和明确分类的局部三维网格图用于在线地形分析和连接性评估，而一个考虑高度的全球拓扑图则从这些局部空间表示中逐步构建，以便高效地进行跨层探索规划。我们进一步引入了一种楼梯中心对齐策略，以减少攀爬过程中的突变偏航变化，并采用双路径搜索机制，以在全球拓扑局部断开时恢复引导路径。大量的仿真和实际实验表明，在多层结构中实现了鲁棒且计算稳定的自主探索，包括对五层楼梯间的连续探索。
View on arXiv Download PDF AI Translation
cs.RO / 12 / 2608.25395

A Taxonomy of Construction Task Activities for Robot Workers

机器人工人的施工任务活动分类
Sakib, Sadman, Peng, Zhangyi None, Pang, Yujie, Otsuki, Yu, Faruque, Mohammad Abdullah Al
Abstract
Recent vision-language-action models offer a path toward robots with broader repertoires than conventional task-specific systems. Construction deployment, however, requires a precise inventory of worker activities and the capabilities needed to execute them. We present TARCAT, an occupation-grounded taxonomy derived from 91 O*NET tasks across seven high-employment construction occupations and 30 instructional videos of physical work. TARCAT defines 41 action primitives in 12 groups and three classes and provides a mechanism for composing parameterized primitive sequences into reusable skills. This human-interpretable structure can organize demonstrations, specify robot requirements, and support coding agents that retrieve and extend skill libraries. We also demonstrate selected primitives on a DOBOT CR3 arm with a CRAFT hand. TARCAT thereby provides a common vocabulary for analyzing human work and developing general-purpose construction robots. Annotations are available at https://github.com/AICPS/TARCAT-Taxonomy.
Chinese Translation
近期的视觉-语言-行动模型为机器人提供了比传统特定任务系统更广泛的能力。然而，施工部署需要对工人活动及其执行所需能力进行精确的清单整理。我们提出了TARCAT，这是一个基于91个O*NET任务、涵盖七个高就业率施工职业以及30个物理工作指导视频的职业基础分类。TARCAT定义了12组和3类中的41个动作原语，并提供了一种将参数化原语序列组合成可重用技能的机制。这种人类可解释的结构可以组织演示、指定机器人需求，并支持编码代理以检索和扩展技能库。我们还在配备CRAFT手的DOBOT CR3机械臂上演示了选定的原语。因此，TARCAT为分析人类工作和开发通用施工机器人提供了一个共同的词汇。注释可在https://github.com/AICPS/TARCAT-Taxonomy获取。
View on arXiv Download PDF AI Translation
cs.RO / 13 / 2608.25405

LAC: Linear and Angular Compliance for Humanoid Whole-body Control

LAC：用于类人机器人全身控制的线性和角度顺应性
Liu, Yang, Gu, Zhongkai, Zhu, Wei, Hayashibe, Mitsuhiro
Abstract
Real-world humanoid tasks involve physical interaction with objects and humans, yet current controllers either reject external forces as disturbances or restrict compliance to limited body links while ignoring angular effects. We present LAC, a general whole-body controller that simultaneously realizes commanded Linear and Angular Compliance for wrenches applied to the upper body. First, we synthesize whole-body compliant responses into a large-scale augmented dataset. Sampled force and couple events are imposed on contact frames extracted from human interaction data. At each contact link, the external force and a virtual torque from the passively yielding kinematic chain drive a virtual admittance under the commanded stiffness. Subsequently, teacher-student reinforcement learning trains a single policy to track the compliant motions under external wrenches. Finally, extensive simulation and real-world experiments demonstrate whole-body compliant responses to wrenches across the upper body, monotonic modulation over the full range of both stiffness commands, and applicability to teleoperated loco-manipulation tasks. Project website: https://lac-humanoid.github.io/
Chinese Translation
现实世界中的类人任务涉及与物体和人类的物理交互，但当前的控制器要么将外部力量视为干扰而拒绝，要么仅对有限的身体连接限制顺应性，同时忽视角度效应。我们提出了LAC，一种通用的全身控制器，能够同时实现对施加在上半身的扭矩的线性和角度顺应性。首先，我们将全身顺应响应合成到一个大规模的增强数据集中。对从人类交互数据中提取的接触框架施加采样的力和力矩事件。在每个接触连接处，外部力量和来自被动屈服运动链的虚拟扭矩驱动在命令刚度下的虚拟导纳。随后，教师-学生强化学习训练一个单一策略，以跟踪外部扭矩下的顺应运动。最后，大量的仿真和现实世界实验展示了对上半身施加的扭矩的全身顺应响应，以及在整个刚度命令范围内的单调调制，并证明其适用于遥操作的运动操控任务。项目网站：https://lac-humanoid.github.io/
View on arXiv Download PDF AI Translation
cs.RO / 14 / 2608.25427

SUPER ODOMETRY 2.0: Resilient Odometry via Hierarchical Adaptation

超级里程计 2.0：通过层次适应实现的鲁棒里程计
Zhao, Shibo, Zhou, Sifan, Zhang, Yuchen, Zhang, Ji, Wang, Chen, Wang, Wenshan, Scherer, Sebastian
Abstract
Resilient and robust odometry is crucial for autonomous systems operating in complex and dynamic environments. Existing odometry systems often struggle with severe sensory degradations and extreme conditions such as smoke, sandstorms, snow, or low-light conditions, threatening both the safety and functionality of robots. To address these challenges, we present Super Odometry, a sensor fusion framework that dynamically adapts to varying levels of environmental degradation. Super Odometry employs a hierarchical structure to integrate four core modules from lower-level to higher-level adaptability including adaptive feature selection, adaptive state direction selection, adaptive engine selection, and a novel learning- based inertial odometry. The inertial odometry, trained on over 100 hours of heterogeneous robotic platforms, captures comprehensive motion dynamics. Super Odometry elevates the inertial measurement unit (IMU) to equal importance with camera and LiDAR within the sensor fusion framework, providing a reliable fallback when exteroceptive sensors fail. Super Odometry has been validated across 200 kilometers and 800 operational hours on a fleet of aerial, wheeled, and legged robots, under diverse sensor configurations, environmental degradation, and aggressive motion profiles. It marks an important step towards safe and long-term robotic autonomy in all-degraded environments.
Chinese Translation
鲁棒且可靠的里程计对于在复杂和动态环境中运行的自主系统至关重要。现有的里程计系统在面对严重的传感器退化和极端条件（如烟雾、沙尘暴、雪或低光照条件）时，常常难以应对，这威胁到机器人的安全性和功能性。为了解决这些挑战，我们提出了超级里程计（Super Odometry），这是一种传感器融合框架，能够动态适应不同程度的环境退化。超级里程计采用层次结构，整合了四个核心模块，从低层到高层的适应性，包括自适应特征选择、自适应状态方向选择、自适应引擎选择，以及一种新颖的基于学习的惯性里程计。该惯性里程计在超过100小时的异构机器人平台上进行训练，捕捉全面的运动动态。超级里程计将惯性测量单元（IMU）与相机和激光雷达在传感器融合框架中同等重要，当外部传感器失效时，提供可靠的后备方案。超级里程计已在200公里和800小时的操作中得到验证，涉及多种传感器配置、环境退化和激进的运动特征，适用于一系列空中、轮式和腿式机器人。这标志着在全退化环境中实现安全和长期机器人自主性的重要一步。
View on arXiv Download PDF AI Translation
cs.RO / 15 / 2608.25459

Towards safe and optimal flight: Viability Kernel MPC for Fully Actuated Multirotor

朝向安全与最优飞行：全驱动多旋翼的可行性核模型预测控制
Bertoni, Massimiliano, Piccina, Alberto, Lunardi, Gianni, Fontanari, Elias, Del Prete, Andrea, Cenedese, Angelo, Michieletto, Giulia
Abstract
Industrial aerial robotics demands safety guarantees for navigation in unstructured environments while optimizing performance and computational efficiency. This paper presents a method for generating safe pose trajectories for fully actuated multirotors within a Model Predictive Control (MPC) framework, leveraging both viability theory and data-driven methods. Obstacle avoidance is enforced through dynamically computed axis-aligned bounding boxes, providing formal safety guarantees without exhaustive offline reachability analysis. Numerical simulations on a fully actuated tilted hexarotor validate the approach, demonstrating successful navigation in cluttered environments with real-time computational performance.
Chinese Translation
工业航空机器人在非结构化环境中导航时需要安全保障，同时优化性能和计算效率。本文提出了一种在模型预测控制（Model Predictive Control, MPC）框架内为全驱动多旋翼生成安全姿态轨迹的方法，该方法结合了可行性理论和数据驱动方法。通过动态计算的轴对齐边界框来强制实施障碍物避免，提供了正式的安全保障，而无需进行详尽的离线可达性分析。在全驱动倾斜六旋翼上的数值仿真验证了该方法，展示了在杂乱环境中成功导航的能力，并实现了实时计算性能。
View on arXiv Download PDF AI Translation
cs.RO / 16 / 2608.25470

Transient multimode heat transfer of an industrial automated tape laying process under rapidly changing conditions

在快速变化条件下的工业自动化铺带工艺的瞬态多模式热传递
Rameder, Bernhard, Gattringer, Hubert, Müller, Andreas, Naderer, Ronald
Abstract
This work presents a transient heat-transfer model of an industrial automated tape laying (ATL) process designed to overcome the limitations of conventional thermal models in composite manufacturing. The model solves the heat-conduction equation with coupled advection, conduction, convection, and radiation. A key innovation is the implementation of an analytical view factor approach that accounts for finite emitter and tape widths, thereby correcting systematic overestimations of radiative heat flux inherent in 1.5D simplifications. Furthermore, a local convection assessment incorporates mixed convection effects characterized by the Richardson number, ensuring accuracy across a wide range of process speeds. The ATL system is represented by two interacting subsystems: the moving tape substrate and the infrared heat sources. The tape is discretized using a two-node model that resolves the physical phase shift between the heated and monitored surfaces. Numerical stability under high dynamics is ensured by a monolithic solution strategy using a high-order implicit integration scheme. Model predictions were validated on an industrial ATL line, demonstrating an overall deviation of only 1.08% (NRMSE) under rapid velocity and current modulations. This framework provides a high-fidelity, physics-based foundation for thermal state estimation, supporting consistent in-situ consolidation and improved part quality.
Chinese Translation
本研究提出了一种工业自动化铺带（ATL）工艺的瞬态热传递模型，旨在克服传统热模型在复合材料制造中的局限性。该模型通过耦合对流、传导、对流和辐射的热传导方程进行求解。一个关键创新是实施了一种分析视角因子方法，该方法考虑了有限的发射体和带宽，从而纠正了1.5D简化中固有的辐射热流的系统性高估。此外，本地对流评估纳入了由理查森数（Richardson number）表征的混合对流效应，确保在广泛的工艺速度范围内的准确性。ATL系统由两个相互作用的子系统组成：移动的带基材和红外热源。带材采用两节点模型进行离散化，解决了加热表面与监测表面之间的物理相位差。在高动态条件下，通过使用高阶隐式积分方案的单体解法策略确保了数值稳定性。模型预测在工业ATL生产线上得到了验证，在快速速度和电流调制下，整体偏差仅为1.08%（NRMSE）。该框架为热状态估计提供了高保真度的物理基础，支持一致的原位固化和提高部件质量。
View on arXiv Download PDF AI Translation
cs.RO / 17 / 2608.25509

Dynamic Modeling of a Welding Torch Umbilical and Its Impact on Robot Dynamics

焊接喷枪脐带的动态建模及其对机器人动力学的影响
Gautier, Nicolas, Guillermit, Yves, Porez, Mathieu, Rousset, Fabien, Chablat, Damien
Abstract
Robotic welding is widely used in industrial manufacturing, where the welding torch is often connected to the generator through an external umbilical. With the increasing deployment of lightweight and collaborative robots, the dynamic influence of this umbilical can significantly affect the robot motion and the actuation forces. This paper proposes a constrained multibody dynamic model of a welding umbilical, represented as a serial chain of rigid bodies interconnected by passive joints with elastic and dissipative effects. Prescribed motions at the distal anchor point are introduced through holonomic kinematic constraints. The equations of motion are reduced by projecting the dynamics onto the subspace of admissible velocities, yielding an efficient formulation free of Lagrange multipliers. The reaction wrench exerted by the umbilical on the robot is explicitly recovered. A planar case study illustrates the approach.
Chinese Translation
机器人焊接广泛应用于工业制造中，其中焊接喷枪通常通过外部脐带连接到发电机。随着轻量级和协作机器人应用的增加，这种脐带的动态影响可能显著影响机器人的运动和驱动力。本文提出了一种焊接脐带的约束多体动态模型，该模型被表示为一系列通过被动关节相互连接的刚体的串联链，具有弹性和耗散效应。通过全局运动约束在远端锚点引入规定的运动。通过将动力学投影到可接受速度的子空间，简化了运动方程，得到了不含拉格朗日乘子的高效公式。明确恢复了脐带对机器人的反作用力矩。一个平面案例研究展示了该方法。
View on arXiv Download PDF AI Translation
cs.RO / 18 / 2608.25547

A Tendon-Driven Five-Fingered Hand with Distributed Tactile Perception for Dexterous Manipulation

一种具有分布式触觉感知的腱驱动五指手用于灵巧操作
Chen, Huayang, Qin, Longhui
Abstract
To apply the techniques of embodied artificial intelligence to human-oid robots for complex manipulations, dexterous robotic hands are indispensable, which are restricted by the dexterity and tactile perception capability. In this work, we proposed a novel design of tendon-driven five-fingered hand with dis-tributed tactile perception. With a soft-rigid-hybrid structure employed, both compliance and operational force are endowed to the hand. Dual-modality tactile sensing elements are distributed on the distal and middle phalanges of all five fingers, enabling the simultaneous detection of static contact and dynamic force variations. Manipulation experiments, including counting gestures, finger-to-thumb pinching, object grasping, and bottle-grasp tactile recording, demonstrate the feasibility of the integrated actuation-perception system.
Chinese Translation
为了将具身人工智能的技术应用于人形机器人以进行复杂操作，灵巧的机器人手是不可或缺的，而其灵巧性和触觉感知能力受到限制。在本研究中，我们提出了一种新颖的腱驱动五指手设计，具有分布式触觉感知。采用软-刚-混合结构，使手部具备了顺应性和操作力。双模态触觉传感元件分布在五个手指的远端和中间指骨上，能够同时检测静态接触和动态力变化。操作实验，包括计数手势、手指与拇指夹持、物体抓取和瓶子抓取触觉记录，展示了集成驱动-感知系统的可行性。
View on arXiv Download PDF AI Translation
cs.RO / 19 / 2608.25572

ConfAL-WM: Confidence-Guided Active Learning for Action-Conditioned World Models

ConfAL-WM：基于信心引导的主动学习用于动作条件下的世界模型
Liu, Xiang, Cui, Sen, Zhang, Changshui
Abstract
Action-conditioned world models have become an important foundation for embodied prediction, planning, and synthetic data generation, but their errors under new task and scene distributions are often concentrated in localized spatiotemporal regions such as robot arms, manipulated objects, contact areas, and occluded objects. This paper presents ConfAL-WM, a confidence-guided active learning framework for post-training embodied world models. Built upon EVAC, we attach a lightweight confidence probe to UNet decoder features and predict dense confidence maps in the latent space. These maps are aggregated into task-, frame-, and patch-level scores, enabling both efficient data selection and localized training enhancement. Our pipeline first retrains the confidence probe and warms up EVAC with a small subset of target-domain data, then performs task-level prescreening to allocate sampling budgets, and finally applies selected-data retraining with optional frame or patch weighted data enhancement. Experiments on RoboTwin2.0 show that confidence-guided selection improves post-training efficiency, while dense frame and patch weighting further enhances prediction quality and embodied trajectory consistency compared with scalar reward, progress, and judge-based scoring baselines. A quick visual overview of this work is available at https://ConfAL-WM.github.io.
Chinese Translation
动作条件下的世界模型已成为具身预测、规划和合成数据生成的重要基础，但它们在新任务和场景分布下的错误往往集中在局部时空区域，如机器人手臂、被操控物体、接触区域和被遮挡物体。本文提出了ConfAL-WM，一个针对后训练具身世界模型的基于信心引导的主动学习框架。该框架基于EVAC构建，我们将一个轻量级的信心探针附加到UNet解码器特征上，并在潜在空间中预测稠密的信心图。这些图被聚合成任务级、帧级和补丁级的评分，从而实现高效的数据选择和局部训练增强。我们的流程首先使用目标领域数据的小子集重新训练信心探针并预热EVAC，然后进行任务级的预筛选以分配采样预算，最后应用选定数据的重新训练，并可选择性地进行帧或补丁加权的数据增强。在RoboTwin2.0上的实验表明，基于信心的选择提高了后训练效率，而稠密的帧和补丁加权进一步增强了预测质量和具身轨迹的一致性，相较于标量奖励、进展和基于评判的评分基线。该工作的快速视觉概述可在https://ConfAL-WM.github.io获取。
View on arXiv Download PDF AI Translation
cs.RO / 20 / 2608.25585

RA-VLA: Retrieval-Augmented VLA for Test-Time Adaptation

RA-VLA：用于测试时适应的检索增强型 VLA
Jang, Sanghwan, Jeon, Minjin, Kim, Minsoo, Choi, Seongjin, Kim, Dongha, Yu, Hwanjo
Abstract
Vision-Language-Action (VLA) models provide a versatile foundation for general robotic manipulation, yet they exhibit significant brittleness when confronted with novel task distributions. While In-Context Imitation Learning (ICIL) offers a training-free alternative, existing frameworks suffer from an adaptation bottleneck that hinders the effective translation of expert context to executable actions. This failure originates from superficial retrieval mechanisms and an inherent behavioral inertia that anchors the policy to its pre-trained priors. To address these limitations, we present RA-VLA, a retrieval-augmented VLA framework that integrates behavior-aligned context retrieval with a grounded execution pipeline. By enforcing faithful adherence to functional cues within a scalable architecture, RA-VLA facilitates seamless task adaptation while preserving inference efficiency. Our empirical evaluations across the LIBERO benchmark and a real-world UR5e environment demonstrate that RA-VLA achieves superior success rates and computational efficiency, establishing a robust framework for training-free robotic adaptation.
Chinese Translation
视觉-语言-动作（VLA）模型为通用机器人操作提供了多功能基础，但在面对新任务分布时表现出显著的脆弱性。尽管上下文模仿学习（ICIL）提供了一种无训练的替代方案，现有框架仍然面临适应瓶颈，阻碍了专家上下文向可执行动作的有效转化。这一失败源于表层的检索机制和固有的行为惯性，使得策略依赖于其预训练的先验。为了解决这些局限性，我们提出了 RA-VLA，一种检索增强型 VLA 框架，结合了行为对齐的上下文检索与基于实际的执行管道。通过在可扩展架构中强制遵循功能线索，RA-VLA 实现了无缝的任务适应，同时保持推理效率。我们在 LIBERO 基准和真实世界 UR5e 环境中的实证评估表明，RA-VLA 实现了更高的成功率和计算效率，建立了一个强大的无训练机器人适应框架。
View on arXiv Download PDF AI Translation
cs.RO / 21 / 2608.25610

Advantage-Driven Explicit Memory for Social Navigation

基于优势驱动的显式记忆用于社会导航
Park, Yeonsoo, Racca, Mattia, Bono, Guillaume, Janny, Steeven, Monaci, Gianluca, Silander, Tomi, Wolf, Christian
Abstract
Robot policies are predominantly learned with classical parametric variants of imitation learning or RL, where training stores the agent's behavior exclusively in the policy's network parameters, putting a heavy burden on the representation learning algorithm. We propose a new navigation agent equipped with non-parametric memory which explicitly indexes prior steps leading to critical events. The advantages are twofold: first, it allows the policy to outsource some of its behavior into an explicit memory; second, it encourages a form of continual learning by allowing an agent to collect data from its testing episodes during deployment and therefore to better generalize to OOD situations. In the context of social navigation, we show that this improves the agent's capability to retain sparse, high-cost failures, such as human collisions. If the policy is trained in simulation, this also naturally addresses the sim-to-real gap, partially, by basing some of the decision making on real data. We integrate the explicit memory into a recurrent PPO architecture and use hidden states for memory retrieval to capture continuous spatiotemporal dynamics. The goal of exploiting rare, high-impact events is achieved by leveraging the RL agent's advantage signals. We train our agent in simulation with a combination of photorealistic rendering and non-visual crowd simulation and show that the agent is robust with respect to OOD social behavior.
Chinese Translation
机器人策略主要通过经典的参数化模仿学习或强化学习（RL）变体进行学习，其中训练将代理的行为仅存储在策略的网络参数中，这对表征学习算法造成了沉重的负担。我们提出了一种新的导航代理，该代理配备了非参数记忆，能够显式索引导致关键事件的先前步骤。其优势有二：首先，它允许策略将部分行为外包到显式记忆中；其次，它通过允许代理在部署期间从测试阶段收集数据，促进了一种持续学习的形式，从而更好地推广到OOD（Out-Of-Distribution）情境。在社会导航的背景下，我们表明这提高了代理保留稀疏、高成本失败（例如人与人碰撞）的能力。如果策略是在模拟环境中训练的，这也自然地部分解决了模拟到现实的差距，因为部分决策是基于真实数据进行的。我们将显式记忆集成到递归PPO（Proximal Policy Optimization）架构中，并使用隐藏状态进行记忆检索，以捕捉连续的时空动态。通过利用RL代理的优势信号，我们实现了对稀有、高影响事件的利用。我们在结合真实感渲染和非视觉人群模拟的环境中训练我们的代理，并展示了该代理对OOD社会行为的鲁棒性。
View on arXiv Download PDF AI Translation
cs.RO / 22 / 2608.25641

Leveraging Inter-object Affordances for Efficient Planning in Contact-rich Tasks

利用物体间的可供性提高接触丰富任务中的规划效率
Niaz, Pouya P., Piater, Justus, Agostini, Alejandro
Abstract
Traditional task-and-motion planning (TAMP) approaches primarily focus on defining sequences of actions along with the necessary geometric and kinematic constraints to execute long-horizon tasks. However, their applicability in real-world settings is limited, as they typically assume simplified object models that overlook key physical properties critical for the successful execution of contact-rich tasks. Moreover, they often use sub-symbolic reasoning during motion planning, which drastically increases planning time and decreases overall success rates. We propose a method that leverages a TAMP approach, defining object-centric abstractions of execution constraints, called Unified TAMP (U-TAMP), to execute robotic tasks involving interactions among objects with heterogeneous shapes, sizes, and materials. Using a Vision-Language Model (VLM), we generate abstractions of inter-object affordances for characterizing physical interaction constraints between objects in contact-rich tasks, such as grasp and support constraints. These constraints are used to enrich the U-TAMP planning domain to deal with objects with variable physical properties. We perform experiments in simulated kitchen table organization scenarios and compare our results with those of the original U-TAMP, as well as a state-of-the-art VLM-based planner that leverages common sense knowledge of objects' affordances for plan generation. Our approach achieves significantly higher planning success rates and improves planning times by one to two orders of magnitude compared to other methods.
Chinese Translation
传统的任务与运动规划（TAMP）方法主要集中于定义一系列动作及其执行长时间任务所需的几何和运动约束。然而，这些方法在现实世界中的适用性有限，因为它们通常假设简化的物体模型，忽视了成功执行接触丰富任务所需的关键物理属性。此外，它们在运动规划过程中往往使用亚符号推理，这大大增加了规划时间并降低了整体成功率。我们提出了一种方法，利用TAMP方法，定义以物体为中心的执行约束抽象，称为统一TAMP（U-TAMP），以执行涉及形状、大小和材料异构物体之间相互作用的机器人任务。通过使用视觉-语言模型（VLM），我们生成物体间可供性的抽象，以表征接触丰富任务中物体之间的物理交互约束，如抓取和支撑约束。这些约束用于丰富U-TAMP规划领域，以处理具有可变物理属性的物体。我们在模拟的厨房桌面组织场景中进行实验，并将我们的结果与原始U-TAMP以及一种基于VLM的最先进规划器的结果进行比较，该规划器利用物体可供性的常识知识进行计划生成。与其他方法相比，我们的方法实现了显著更高的规划成功率，并将规划时间提高了一个到两个数量级。
View on arXiv Download PDF AI Translation
cs.RO / 23 / 2608.25642

EgoNav: Bridging Learned Waypoints and Geometry-Aware Local Control for Robust Indoor Navigation

EgoNav：桥接学习的航点与几何感知的局部控制以实现稳健的室内导航
Wang, Jing, Zhao, Shiqi, Qu, Hairong, Yin, Peng
Abstract
Image-goal navigation using lightweight topological maps is a practical paradigm for indoor robot deployment: the map requires only geotagged images, and localization relies on visual matching rather than precise pose estimation. However, learned waypoint predictors can produce targets that violate geometric constraints or deviate from the global path. Executing these waypoints safely further requires a local planner capable of collision avoidance, yet existing systems either lack one or rely on fixed parameters that cannot adapt to confined spaces. To address these limitations while retaining the navigational intuition of the learned predictor, we present EgoNav, a hierarchical system that implements this idea by generating candidates from semantically segmented traversable regions and scoring them alongside the learned waypoint for geometric safety, directional coherence, and fidelity to the learned prior. An adaptive local path planner then executes the refined waypoint with parameters modulated based on the refinement outcome. Experiments in Habitat-sim and on a physical humanoid robot show that EgoNav consistently outperforms contemporary baselines in both success rate and path efficiency.
Chinese Translation
使用轻量级拓扑地图进行图像目标导航是一种适用于室内机器人部署的实用范式：该地图仅需地理标记的图像，而定位依赖于视觉匹配而非精确的姿态估计。然而，学习的航点预测器可能会产生违反几何约束或偏离全局路径的目标。安全执行这些航点进一步需要一个能够避免碰撞的局部规划器，而现有系统要么缺乏此功能，要么依赖于无法适应狭小空间的固定参数。为了解决这些局限性，同时保留学习预测器的导航直觉，我们提出了EgoNav，这是一种分层系统，通过从语义分割的可通行区域生成候选航点，并根据几何安全性、方向一致性和对学习先验的忠实度对其进行评分，从而实现这一理念。然后，一个自适应局部路径规划器根据优化结果调整参数来执行精细化的航点。在Habitat-sim和一个物理人形机器人上的实验表明，EgoNav在成功率和路径效率方面始终优于当代基线。
View on arXiv Download PDF AI Translation
cs.RO / 24 / 2608.25659

GaussianDream++: Efficient 3D Gaussian World Modeling for Robotic Manipulation

GaussianDream++：用于机器人操作的高效3D高斯世界建模
Jiang, Yuqing, Zhang, Zijian, Zhou, Weitao, Wang, Jiawei, He, Junjie, Yang, Lei, Qing, Haifang, Liu, Si, Zhao, Ding, Luo, Ping, Yu, Haibao
Abstract
Vision-Language-Action (VLA) policies have advanced language-conditioned robotic manipulation, yet action-imitation objectives provide only weak supervision for metric 3D structure and short-horizon physical evolution. Geometry-enhanced policies mainly improve current-scene grounding, whereas predictive policies often model future dynamics in RGB or latent spaces and may incur substantial deployment cost. GaussianDream demonstrates that training-time current Gaussian reconstruction and future Gaussian prediction provide effective 3D supervision, but its dense VGGT/TGE-based prefix jointly carries state, dynamics, and action-conditioning information. We present \textbf{\methodname}, a compact, policy-native extension that inserts \textbf{World State Tokens} and \textbf{World Prediction Tokens} directly into the VLA backbone. A training-only \textbf{World Representation Head} decodes these tokens into a Current World and coupled Future Prediction over shared Gaussian primitives, while static--dynamic factorization preserves persistent structure and focuses residual motion on interaction-relevant regions. At inference, the head, renderer, auxiliary objectives, and VGGT/TGE pathway are removed, leaving only 20 world tokens without online Gaussian decoding or rollout. \method achieves \textbf{98.6\%} on LIBERO and \textbf{87.8\%} on LIBERO-Plus, with clear gains under Camera and Layout shifts. Real-robot experiments further improve average success from 29.2\% to 52.5\% over reproduced $\pi_{0.5}$ while maintaining efficient closed-loop control.
Chinese Translation
视觉-语言-动作（VLA）策略推动了语言条件下的机器人操作，然而，模仿动作的目标仅为度量3D结构和短期物理演变提供了弱监督。几何增强策略主要改善当前场景的基础，而预测策略通常在RGB或潜在空间中建模未来动态，可能会产生可观的部署成本。GaussianDream展示了训练时当前高斯重建和未来高斯预测提供有效的3D监督，但其基于密集VGGT/TGE的前缀共同携带状态、动态和动作条件信息。我们提出了 extbf{ extit{methodname}}，一种紧凑的、策略原生的扩展，直接将 extbf{世界状态标记}和 extbf{世界预测标记}插入VLA骨干网络。一个仅用于训练的 extbf{世界表示头}将这些标记解码为当前世界和基于共享高斯原语的耦合未来预测，而静态-动态因子分解则保留持久结构，并将剩余运动集中在与交互相关的区域。在推理时，头部、渲染器、辅助目标和VGGT/TGE路径被移除，仅留下20个世界标记，无需在线高斯解码或展开。 extit{method}在LIBERO上达到了 extbf{98.6 ext{%}}，在LIBERO-Plus上达到了 extbf{87.8 ext{%}}，在相机和布局变化下有明显提升。真实机器人实验进一步将平均成功率从29.2 ext{%}提高到52.5 ext{%}，同时保持高效的闭环控制。
View on arXiv Download PDF AI Translation
cs.RO / 25 / 2608.25666

PRISM: Projection-Integrated Sampling-Based MPC with Bayesian Cost Tuning for Bimanual Manipulation

PRISM：基于投影集成采样的模型预测控制与贝叶斯成本调优用于双手操作
Dan, Alinjar, Hurova, Iryna, Kruusamäe, Karl, Singh, Arun Kumar
Abstract
Bimanual manipulation in cluttered, contact-rich environments remains challenging because it requires coordinated motion generation, interaction-aware planning, and reliable execution under tight kinematic constraints. We present PRISM, a projection-integrated sampling-based Model Predictive Control (MPC) framework that uses a GPU-accelerated physics simulator as an online world model for complex dual-arm manipulation. The main algorithmic contribution is a QP-guided control sampling strategy that decouples trajectory exploration from kinematic feasibility. At each MPC step, sampled joint-velocity trajectories are projected onto the set of motions satisfying joint position, velocity, acceleration, and jerk bounds, together with an initial-velocity boundary condition, before rollout evaluation. This enables broad yet feasible exploration of coordinated bimanual behaviors. To support efficient online execution, we derive a custom ADMM/Bregman-splitting QP solver that exploits joint-wise separability and reusable matrix factorizations. We further use Bayesian optimization to tune task-cost weights offline, reducing manual parameter selection. We evaluate PRISM on challenging variants of PerAct$^{2}$ tasks, including obstacle-constrained ball transport, tray transport, cube handover, and box lifting. Experiments show improved robustness and task success relative to representative sampling-based baselines, while maintaining real-time or near-real-time execution. We also demonstrate successful sim-to-real transfer on dual UR5e manipulators, highlighting the practical potential of physics-based online planning for contact-rich bimanual manipulation. Project details, including code and supplementary videos, are available at \href{https://sites.google.com/view/prismbimanual}{\texttt{https://sites.google.com/view/prismbimanual}}.
Chinese Translation
在杂乱且接触丰富的环境中进行双手操作仍然具有挑战性，因为这需要协调的运动生成、考虑交互的规划以及在严格的运动学约束下的可靠执行。我们提出了PRISM，一个基于投影集成采样的模型预测控制（MPC）框架，利用GPU加速的物理模拟器作为复杂双臂操作的在线世界模型。主要的算法贡献是一个QP引导的控制采样策略，它将轨迹探索与运动学可行性解耦。在每个MPC步骤中，采样的关节速度轨迹被投影到满足关节位置、速度、加速度和抖动边界条件的运动集合上，并结合初始速度边界条件，随后进行展开评估。这使得协调的双手行为能够进行广泛而可行的探索。为了支持高效的在线执行，我们推导出一个自定义的ADMM/Bregman分裂QP求解器，利用关节分离性和可重用的矩阵分解。我们还使用贝叶斯优化离线调优任务成本权重，减少手动参数选择。我们在PerAct$^{2}$任务的挑战性变体上评估了PRISM，包括障碍物约束的球体运输、托盘运输、立方体交接和箱子提升。实验表明，相较于代表性的基于采样的基线，PRISM在鲁棒性和任务成功率上有所提高，同时保持实时或近实时的执行。我们还展示了在双UR5e操纵器上的成功仿真到现实转移，突显了基于物理的在线规划在接触丰富的双手操作中的实际潜力。项目细节，包括代码和补充视频，可在 exttt{https://sites.google.com/view/prismbimanual} 获取。
View on arXiv Download PDF AI Translation
cs.RO / 26 / 2608.25674

Opportunities of Self Supervised Learning for GNSS: Evaluation of a Deep Learning-Enhanced PVT Algorithm

自监督学习在全球导航卫星系统中的机遇：深度学习增强的定位、导航与定时算法评估
Barbero, Thomas, Ekambi, Bertrand
Abstract
This work proposes a Deep Learning Enhanced PVT algorithm to mitigate multipath interference in dense urban areas. A supervised objective jointly predicts range corrections and uncertainty, while a JEPA-based self-supervised pretraining stage improves representation quality. The algorithm is evaluated over diverse driving scenarios, substantially improving PVT accuracy, particularly for unseen harsh urban conditions. These results highlight the potential of unlabelled GNSS data to improve generalization performance.
Chinese Translation
本研究提出了一种深度学习增强的定位、导航与定时（PVT）算法，以减轻密集城市区域中的多路径干扰。该算法通过监督目标共同预测范围修正和不确定性，同时基于JEPA的自监督预训练阶段提高了表示质量。算法在多样化的驾驶场景中进行了评估，显著提高了PVT的准确性，尤其是在未见过的恶劣城市条件下。这些结果突显了未标记的全球导航卫星系统（GNSS）数据在提高泛化性能方面的潜力。
View on arXiv Download PDF AI Translation
cs.RO / 27 / 2608.25757

LM-X: Explainable Action Modeling with Progress, Event, and Uncertainty Prediction for Generalist Robot Manipulation

LM-X：具有进展、事件和不确定性预测的可解释动作建模用于通用机器人操作
Lou, Jin, Zhu, Jingxuan, Chen, Andong, Wang, Xupeng, Xu, Yuan, Li, Yuexuan, Zhu, Xingdong, Zhu, Zhijie, Ji, Yingwei, Nie, Wenpeng, Li, Jingyi, Chen, Liangliang, Liu, Jinyan, Song, Zhiqi, Zhang, Jidong, Li, Hongming, Zhu, Yuchen
Abstract
Generalist vision--language--action (VLA) policies learn long-horizon behavior mainly through short-horizon action prediction and reveal little beyond sampled commands. This creates two coupled bottlenecks: a single action target must implicitly absorb task progress, intermediate intent, and local reliability, while these control states remain hidden during execution. Inspired by functional principles of biological sensorimotor control, we introduce LM-X , which organizes prediction across task, event, and motor scales without claiming anatomical correspondence. Three explicitly supervised signals are emitted online and directly condition action generation: return-to-go (RTG) measures visible task progress, event-to-go (ETG) identifies the next semantic transition, and heteroscedastic action flow estimates local reliability through propagated variance. Explanation is therefore intrinsic to control rather than generated post hoc. Before a costly 20-day pretraining run on 64 NVIDIA B200 GPUs, a controlled five-task pretraining gate verifies the design: the complete model improves success by 16.0 points over the action-only backbone and by 10.8 points over the strongest single-head variant. We then train LM-X on more than 20,000 hours of real-robot trajectories, including over 1,000 hours of failed policy rollouts. LM-X achieves 74.1\% across 50 randomized-hard RoboTwin2.0 tasks versus 55.4\% for GR00T N1.7, and 68.6\% versus 50.7\% across seven real-robot tasks. RTG tracks semantic progress and visible regression, while variance rises during hesitation and oscillatory control. These results show that explicit multi-timescale predictive state can strengthen control while exposing interpretable internal estimates.
Chinese Translation
通用视觉-语言-动作（VLA）策略主要通过短期动作预测学习长期行为，并且对采样命令之外的内容揭示甚少。这造成了两个相互关联的瓶颈：单一的动作目标必须隐式地吸收任务进展、中间意图和局部可靠性，而这些控制状态在执行过程中保持隐藏。受到生物传感运动控制功能原则的启发，我们引入了LM-X，它在不声称解剖对应的情况下，组织跨任务、事件和运动尺度的预测。三个显式监督信号在线发出，并直接影响动作生成：返回目标（RTG）衡量可见的任务进展，事件目标（ETG）识别下一个语义过渡，而异方差动作流通过传播方差估计局部可靠性。因此，解释是控制的内在部分，而不是事后生成的。在耗资巨大的64 NVIDIA B200 GPU上进行为期20天的预训练之前，受控的五任务预训练门验证了设计：完整模型在成功率上比仅有动作的骨干模型提高了16.0个百分点，比最强的单头变体提高了10.8个百分点。随后，我们在超过20,000小时的真实机器人轨迹上训练LM-X，包括超过1,000小时的失败策略回滚。LM-X在50个随机困难的RoboTwin2.0任务中达到了74.1%的成功率，而GR00T N1.7为55.4%；在七个真实机器人任务中达到了68.6%的成功率，而GR00T N1.7为50.7%。RTG跟踪语义进展和可见回归，而在犹豫和振荡控制期间方差上升。这些结果表明，显式的多时间尺度预测状态可以增强控制，同时揭示可解释的内部估计。
View on arXiv Download PDF AI Translation
cs.RO / 28 / 2608.25798

TacForcing: Streaming Action Generation with Execution-Time Tactile Feedback

TacForcing：具有执行时间触觉反馈的流式动作生成
Zhou, Jianbo, Zhao, Boyuan, Zhang, Yuzheng, Chen, Yiyang, Chen, Wenxin, Li, Qiuyue, Gu, Xiangyang, Cao, Yuhan, Xia, Xiao, Hu, Yanzhe, Deng, Zhijie
Abstract
Contact-rich manipulation requires adapting to contact states that can evolve substantially within an action horizon. However, chunk-based vision-language-action models predict complete action chunks from observations collected before execution, leaving tactile conditioning stale during execution. Existing tactile-reactive approaches typically rely on separate high-frequency controllers, which increase both architectural and training complexity. In this paper, we introduce TacForcing, a streaming action-generation framework that effectively incorporates execution-time tactile feedback. Instead of employing a separate reactive controller, TacForcing replaces the standard action expert with a streaming action expert to generate actions conditioned on the evolving tactile observations acquired during execution. TacForcing also introduces Execution-Aware Tactile Attention (EATA), which restricts tactile conditioning to actions nearing execution, thereby reducing the temporal mismatch between tactile acquisition and action execution. Across six simulated UniVTAC tasks and three real-world contact-rich manipulation tasks, TacForcing achieves average success rates of 65% and 69%, respectively, outperforming strong baselines in both settings.
Chinese Translation
接触丰富的操作需要适应在动作视野内可能显著变化的接触状态。然而，基于块的视觉-语言-动作模型从执行前收集的观察中预测完整的动作块，导致在执行过程中触觉条件过时。现有的触觉反应方法通常依赖于单独的高频控制器，这增加了架构和训练的复杂性。本文提出了TacForcing，一种有效结合执行时间触觉反馈的流式动作生成框架。TacForcing不采用单独的反应控制器，而是用流式动作专家替代标准动作专家，以生成基于执行过程中获得的不断变化的触觉观察条件的动作。TacForcing还引入了执行感知触觉注意力（Execution-Aware Tactile Attention, EATA），该机制将触觉条件限制在即将执行的动作上，从而减少触觉获取与动作执行之间的时间不匹配。在六个模拟的UniVTAC任务和三个真实世界的接触丰富操作任务中，TacForcing分别达到了65%和69%的平均成功率，在这两种设置中均优于强基线。
View on arXiv Download PDF AI Translation
cs.RO / 29 / 2608.25799

AGRO-Nav: Autonomous Graph-based Orchard Navigation

AGRO-Nav：基于图的自主果园导航
Yun, Ho Young, Yu, Jaemin, Kim, Duksu
Abstract
Orchards form semi-structured environments in which parallel tree rows create natural driving corridors, yet narrow inter-row clearance and dense foliage lead geometry-agnostic grid planners to drift off the row center and risk trunk or canopy contact. We present AGRO-Nav, an automated framework for static graph-based global planning in orchards. From tree-row lines fitted to trunk clusters in a SLAM point cloud, it builds, without any manual waypoints, a sparse topological graph of intra- and inter-row connectivity; a global route is then found by Dijkstra search on this graph, connected to the start and goal by any-angle Theta* segments, and smoothed with a cubic B-spline. In real-orchard trials, AGRO-Nav follows the row center with a mean error of about 0.08 m, far below the A* (0.31 m) and Theta* (0.43 m) shortest-path baselines, while planning roughly four to five times faster. In Isaac Sim, it attains the lowest error among A*, Theta*, and a reproduced RANSAC midline baseline and remains stable as tree density drops to 70%, where the RANSAC baseline degrades. The resulting trajectories---straight row-centered segments joined by controlled turns---suit differential-drive and four-wheel-steering platforms.
Chinese Translation
果园形成了半结构化的环境，其中平行的树行创造了自然的驾驶通道。然而，狭窄的行间间隙和密集的树叶导致几何无关的网格规划器偏离行中心，增加了与树干或树冠接触的风险。我们提出了AGRO-Nav，这是一个用于果园中静态图形基础全局规划的自动化框架。该框架从适配于树干簇的树行线中构建，无需任何手动航点，生成一个稀疏的拓扑图，表示行内和行间的连通性；然后通过在该图上进行Dijkstra搜索找到全局路径，并通过任意角度的Theta*段连接起始点和目标点，最后使用三次B样条进行平滑处理。在实际果园试验中，AGRO-Nav以约0.08米的平均误差跟随行中心，远低于A*（0.31米）和Theta*（0.43米）最短路径基准，同时规划速度大约快四到五倍。在Isaac Sim中，它在A*、Theta*和重现的RANSAC中线基准中获得了最低误差，并在树木密度降至70%时保持稳定，而RANSAC基准则出现退化。最终生成的轨迹为直行中心段通过受控转弯连接，适合差分驱动和四轮转向平台。
View on arXiv Download PDF AI Translation
cs.RO / 30 / 2608.25830

Anytime Global Tensor Motion Planning

任意时刻全局张量运动规划
Coumar, Sai, Le, An T., Kingston, Zachary
Abstract
Global Tensor Motion Planning (GTMP) solves motion planning with batched tensor operations over a layered multipartite graph. We generalize GTMP so that adjacent-layer edges are realized by any black-box local planner (e.g., linear interpolation, splines, sampling-based planning, trajectory optimization, or generative sampling). We provide two anytime policies on top of this generalization: Anytime GTMP with random restarts at a fixed budget, which covers every homotopy class almost surely, and AO-GTMP with informed expansion with growing budgets, which converges to the optimal cost. We prove that a single sampled graph covers every endpoint-fixed homotopy class admitting a \(\delta\)-clear representative of bounded length. We also prove that additional samples per layer reduce the per-layer miss probability exponentially, whereas stronger local planners reduce the required layer count only sublinearly. On manipulation benchmarks the method matches state-of-the-art performance, and on 2D navigation it returns batches of topologically diverse solutions, while the informed baselines concentrate on one or two classes.
Chinese Translation
全局张量运动规划（GTMP）通过在分层多部图上进行批量张量操作来解决运动规划问题。我们对GTMP进行了推广，使得相邻层之间的边可以通过任何黑箱局部规划器（例如，线性插值、样条、基于采样的规划、轨迹优化或生成采样）来实现。基于这一推广，我们提供了两种任意时刻策略：在固定预算下进行随机重启的任意时刻GTMP，几乎肯定覆盖每个同伦类，以及具有逐步预算增长的知情扩展的AO-GTMP，收敛于最优成本。我们证明了单个采样图覆盖每个固定端点的同伦类，前提是该类存在一个长度有界的B4-清晰代表。我们还证明了每层增加的样本可以指数级降低每层的遗漏概率，而更强的局部规划器仅以次线性方式减少所需的层数。在操控基准测试中，该方法达到了最先进的性能，而在二维导航中，它返回了一批拓扑上多样的解决方案，而知情基线则集中在一到两个类上。
View on arXiv Download PDF AI Translation
cs.RO / 31 / 2608.25864

MA-VLA: Multi-Arm Vision-Language-Action Model for Collaboration and Compositional Generalization

MA-VLA：用于协作和组合泛化的多臂视觉-语言-动作模型
Zhang, Zaibin, Xiao, Junlan, Zhang, Zhongbo, Wang, Yifan, Kang, Li, Qin, Yiran, Xia, Changxing, Zhou, Heng, Fu, Talas, Zhou, Enshen, Zhang, Ruimao, Yin, Zhenfei, Lu, Huchuan, Wang, Lijun
Abstract
Multi-arm collaboration is becoming a core capability in embodied manipulation. Recent vision-language-action (VLA) models integrate perception, language, and control, but most represent language as a single global instruction and do not provide an explicit mechanism for assigning and composing arm-specific behaviors. This design limits transfer to collaboration patterns that differ from those observed during training. We present MA-VLA, a unified framework for multi-arm collaboration via atomic action assignment. MA-VLA decomposes cooperative behavior into mid-level atomic prompts and allocates them to individual arms, enabling explicit subgoal specification and compositional reuse across tasks. To reduce reliance on fixed execution roles, we introduce Arm Shuffle, a training-time permutation of the observation, state, and assigned atomic prompts for each arm. This permutation enforces role-agnostic instruction following and supports recomposition into unseen coordination patterns, which we term multi-arm compositional generalization. We also construct a benchmark in which test-time collaboration patterns are absent in training set. Across simulation and real-world evaluations, prior state-of-the-art VLAs largely fail under these unseen collaborations, while MA-VLA consistently succeeds. These results indicate that structured, per-arm atomic action assignment offers a practical route to scalable generalization in multi-arm embodied systems. Code, models, and data are available at https://github.com/zhangzaibin/future-robots
Chinese Translation
多臂协作正成为具身操控中的核心能力。近期的视觉-语言-动作（VLA）模型整合了感知、语言和控制，但大多数模型将语言表示为单一的全局指令，并未提供明确的机制来分配和组合特定于手臂的行为。这种设计限制了对与训练期间观察到的协作模式不同的协作模式的迁移。我们提出了MA-VLA，这是一个通过原子动作分配实现多臂协作的统一框架。MA-VLA将合作行为分解为中级原子提示，并将其分配给各个手臂，从而实现明确的子目标规范和跨任务的组合重用。为了减少对固定执行角色的依赖，我们引入了Arm Shuffle，这是一种在训练期间对每个手臂的观察、状态和分配的原子提示进行的排列。这种排列强制执行与角色无关的指令遵循，并支持重新组合成未见过的协调模式，我们称之为多臂组合泛化。我们还构建了一个基准，其中测试时的协作模式在训练集中不存在。在模拟和现实世界评估中，先前的最先进的VLA在这些未见过的协作中大多失败，而MA-VLA则始终成功。这些结果表明，结构化的、按手臂分配的原子动作分配为多臂具身系统的可扩展泛化提供了一条实用的途径。代码、模型和数据可在 https://github.com/zhangzaibin/future-robots 获取。
View on arXiv Download PDF AI Translation
cs.RO / 32 / 2608.25872

VISTA: Visually Inferred Spatial ConTact Attention for Contact-Rich Manipulation

VISTA：用于接触丰富操作的视觉推断空间接触注意力
Chen, Jiayi, Dong, Wenlong, Huang, Yan, Chen, Xianglin, Lin, Zijian, Yin, Jiaqi, Liu, Yushan, Ding, Wenbo
Abstract
Contact-rich manipulation requires precise interaction feedback. While vision-centric imitation learning is prevalent, external visual observations provide indirect and ambiguous cues about contact states, particularly under occlusion or subtle object--gripper interactions; dedicated tactile or force sensors can provide rich contact information but introduce additional hardware complexity, calibration requirements, and deployment costs. To bridge this gap, we propose VISTA-Policy, an imitation learning paradigm that utilizes the Visual Deformation Field (VDF), a 3D displacement representation of a compliant gripper, as high-dimensional visuo-physical feedback. The framework integrates: 1) a Physics-Aware Encoding Engine for real-time VDF decoding; 2) an Energy Aggregation Denoising Mechanism to isolate true interaction signals; and 3) a Deformation-Augmented Policy Network with incremental gripper actions for precise closed-loop correction. Extensive evaluations on Cross-Scale Object Grasping, Cap Unscrewing, and Calligraphy Writing demonstrate that VISTA-Policy outperforms the strong pure-vision baseline 3D Diffusion Policy and the tactile baseline. VISTA-Policy further demonstrates substantial out-of-distribution generalization to unseen object scales and robustness against dynamic disturbances, offering a durable and cost-effective route toward general-purpose fine-grained manipulation in unstructured environments. Project videos and supplementary materials are available at: https://sites.google.com/view/vista-policy.
Chinese Translation
接触丰富的操作需要精确的交互反馈。尽管以视觉为中心的模仿学习已广泛应用，但外部视觉观察提供的关于接触状态的线索往往是间接和模糊的，尤其是在遮挡或微妙的物体与夹持器交互的情况下；专用的触觉或力传感器能够提供丰富的接触信息，但会引入额外的硬件复杂性、校准要求和部署成本。为了解决这一问题，我们提出了VISTA-Policy，这是一种模仿学习范式，利用视觉变形场（Visual Deformation Field, VDF），即顺应性夹持器的三维位移表示，作为高维的视觉物理反馈。该框架集成了：1）一个物理感知编码引擎，用于实时解码VDF；2）一个能量聚合去噪机制，用于隔离真实的交互信号；以及3）一个增强变形的策略网络，结合增量夹持器动作以实现精确的闭环修正。在跨尺度物体抓取、瓶盖拧开和书法写作等任务上的广泛评估表明，VISTA-Policy在性能上优于强大的纯视觉基线3D扩散策略和触觉基线。VISTA-Policy进一步展示了对未见物体尺度的显著分布外泛化能力和对动态干扰的鲁棒性，为在非结构化环境中实现通用的精细操作提供了一条持久且具有成本效益的途径。项目视频和补充材料可在以下链接获取：https://sites.google.com/view/vista-policy。
View on arXiv Download PDF AI Translation
cs.RO / 33 / 2608.25874

Low-Resolution Perception for Robotic Packing

低分辨率感知在机器人包装中的应用
Preziosa, Giuseppe Fabio, Vignoni, Federico, Castellano, Chiara, Faroni, Marco, Zanchettin, Andrea Maria, Rocco, Paolo
Abstract
This work tackles the problem of scalable perception for robotic packing with low-cost, low-resolution depth sensing. We propose a framework where reconstruction cues drive next-view selection and grasp evidence updates a per-object stability estimate, jointly deciding what to acquire next and when to grasp. During the reconstruction, a low-resolution Next Best View (NBV) strategy explicitly avoids redundant views while preserving task-relevant geometry. We validate the approach in two steps: (i) an ablation study of the utility function under very low resolution, and (ii) a full end-to-end evaluation across policies, showing how low-resolution perception is a practical, scalable option for robotic packing.
Chinese Translation
本研究解决了在低成本、低分辨率深度传感下，机器人包装的可扩展感知问题。我们提出了一个框架，其中重建线索驱动下一视图选择，而抓取证据更新每个物体的稳定性估计，共同决定下一步获取什么以及何时抓取。在重建过程中，低分辨率的下一最佳视图（Next Best View, NBV）策略明确避免冗余视图，同时保留与任务相关的几何信息。我们通过两个步骤验证了该方法：（i）在非常低分辨率下对效用函数的消融研究，以及（ii）跨策略的完整端到端评估，展示了低分辨率感知作为机器人包装的一个实用且可扩展的选择。
View on arXiv Download PDF AI Translation
cs.RO / 34 / 2608.25940

A Statistical Audit of Physical AI Benchmark Redundancy

物理人工智能基准冗余的统计审计
Navasardyan, Zaruhi, Davtyan, Hrant
Abstract
Physical AI models are evaluated on suites of benchmarks that differ across model reports, leaving the model-by-benchmark matrix sparse and the relationship between benchmarks unmeasured. We construct a matrix of 51 models on 12 physical AI benchmarks, selected from a registry of 51 benchmarks and 152 models by reporting density, combining scores from model cards and benchmark papers with our own evaluation runs under each benchmark's official protocol. We measure how much information the benchmarks share and show quantitative evidence of Redundancy. Redundancy affects reported rankings: collapsing the two substitute pairs into single columns moves 22 of 51 models by three or more places under an equally weighted average. We then select benchmarks greedily under a utility combining score dispersion with variance not explained by the already-selected set, and obtain a four-benchmark subset retaining 78.5\% of the utility of all 12, on which we fit a Bradley--Terry ranking. The procedure requires only benchmark-level scores with sufficient overlap and is not specific to physical AI.
Chinese Translation
物理人工智能模型在不同模型报告中评估的基准套件存在差异，导致模型与基准的矩阵稀疏且基准之间的关系未被测量。我们构建了一个包含51个模型和12个物理人工智能基准的矩阵，这些基准从51个基准和152个模型的注册库中选择，依据报告密度，结合模型卡和基准论文中的得分，以及我们在每个基准的官方协议下进行的评估。我们测量了基准之间共享的信息量，并展示了冗余的定量证据。冗余影响了报告的排名：将两个替代对合并为单列后，51个模型中有22个模型的排名变化了三位或更多。然后，我们在效用的基础上贪婪地选择基准，结合得分离散度与未被已选集解释的方差，获得了一个四个基准的子集，保留了所有12个基准78.5%的效用，并在此基础上拟合了Bradley--Terry排名。该过程仅需具有足够重叠的基准级别得分，并且不特定于物理人工智能。
View on arXiv Download PDF AI Translation
cs.RO / 35 / 2608.26002

DESCENT: Directed Edge Scene Encoding for Airport Surface Movement Prediction

DESCENT：用于机场地面移动预测的定向边场景编码
Prutsch, Alexander, Schinagl, David, Possegger, Horst
Abstract
Advanced automation is a key technology for enhancing the safety of ground operations amidst the increasing density of commercial air traffic. While motion forecasting is a well-studied task in autonomous driving, its application to airport surface movements remains underexplored. To enable efficient and accurate prediction in this domain, we propose DESCENT, a transformer-based architecture designed to handle heterogeneous dynamics and strict topological constraints. Our approach features a Potential Reachable Set (PRS) context sampling mechanism that adaptively collects airfield environment context across diverse operational phases. Combined with a detection transformer-based decoder, DESCENT generates accurate trajectory forecasts. Extensive evaluations on the Amelia-10 benchmark demonstrate significant performance improvements over state-of-the-art baselines. These gains are especially pronounced in safety-critical scenarios, where our domain-aware sampling provides critical long-horizon context necessary for safe navigation.
Chinese Translation
先进的自动化技术是提升地面操作安全性的关键技术，尤其是在商业航空交通密度不断增加的背景下。尽管运动预测在自动驾驶领域已被广泛研究，但其在机场地面移动中的应用仍然未被充分探索。为了在这一领域实现高效和准确的预测，我们提出了DESCENT，一种基于变换器的架构，旨在处理异构动态和严格的拓扑约束。我们的方法具有潜在可达集（Potential Reachable Set, PRS）上下文采样机制，能够自适应地收集不同操作阶段的机场环境上下文。结合基于检测变换器的解码器，DESCENT能够生成准确的轨迹预测。在Amelia-10基准上的广泛评估表明，相较于最先进的基线方法，我们的模型在性能上有显著提升。这些提升在安全关键场景中尤为明显，我们的领域感知采样提供了安全导航所需的关键长时间范围上下文。
View on arXiv Download PDF AI Translation
cs.RO / 36 / 2608.26011

Phantom Navigator: Stealthy and Precise Unmanned Aerial Vehicle Redirection with Real-Time Tracking and GPS Spoofing

幻影导航器：具备实时跟踪和GPS欺骗的隐蔽精确无人机重定向
Meng, Haocheng, Luo, Shaocheng, Xie, Songqiao, Pajic, Miroslav
Abstract
Redirecting unmanned aerial vehicles (UAVs) from their intended mission trajectories has been an active area of research. However, existing UAV redirection attacks lack reliability, precision, and covertness for a targeted diversion. They primarily rely on physical capture, communication hijacking, or sensor spoofing. Yet, physical interception is costly, offers only a single opportunity for success, and poses a high risk of collateral damage; network-based attacks demand deep technical expertise and access to encrypted communication channels; and sensor spoofing techniques typically fall short in achieving the accuracy and robustness required to steer a UAV toward a specified target. Consequently, we propose Phantom Navigator, a UAV redirection attack to mislead drones to a designated spoofing target, covertly and precisely. Our approach combines offline pre-redirection reachability analysis, which provides high-fidelity estimates of achievable redirect ranges, with an online closed-loop, stealthy execution layer that ensures successful redirection in practice. Based on this approach, we build a physical attack platform equipped with a LiDAR--camera detection, tracking, and spoofing stack that performs real-time identification, pose estimation, and computation of targeted spoofing signals to covertly and accurately redirect victim UAVs to a designated location. We demonstrate the effectiveness of our redirection methodology and the attack implementation in real-world case studies.
Chinese Translation
将无人机（UAV）从其预定任务轨迹中重定向一直是一个活跃的研究领域。然而，现有的无人机重定向攻击在可靠性、精确性和隐蔽性方面不足以实现有针对性的转移。这些攻击主要依赖于物理捕获、通信劫持或传感器欺骗。然而，物理拦截成本高昂，仅提供一次成功的机会，并且存在高额的附带损害风险；基于网络的攻击需要深厚的技术专长和对加密通信渠道的访问；而传感器欺骗技术通常无法达到将无人机引导至指定目标所需的准确性和稳健性。因此，我们提出了幻影导航器（Phantom Navigator），这是一种无人机重定向攻击，旨在隐蔽且精确地误导无人机前往指定的欺骗目标。我们的方法结合了离线预重定向可达性分析，提供可实现的重定向范围的高保真估计，以及一个在线闭环隐蔽执行层，确保在实践中成功重定向。基于这一方法，我们构建了一个物理攻击平台，配备了激光雷达（LiDAR）-摄像头检测、跟踪和欺骗堆栈，能够实时识别、姿态估计，并计算目标欺骗信号，从而隐蔽且准确地将受害无人机重定向到指定位置。我们在实际案例研究中展示了我们重定向方法及攻击实施的有效性。
View on arXiv Download PDF AI Translation
cs.RO / 37 / 2608.26050

When Obstacles Bend: Modeling Vegetation Deformation in the context of Field Robotics

当障碍物弯曲时：在田野机器人背景下建模植被变形
Khizar, Muhammad Hsaeeb Zaar, Montagnon, Tom, Lenain, Roland, Aufrère, Romuald, Laconte, Johann
Abstract
Autonomous robots operating in natural environments must often interact with vegetation rather than simply avoid it. In this context, traversability is typically defined from the robot's perspective, by measuring how a specific platform responds when moving through the environment. While practical, this viewpoint entangles the assessment of the environment with the robot's own dynamics, making the resulting characterization difficult to transfer across different platforms. More importantly, it does not directly reflect the properties of the vegetation itself, which are the true source of interaction and potential damage in applications such as agriculture and environmental monitoring. To address this limitation, we propose to characterize vegetation through its intrinsic mechanical properties, independently of any specific robot. By combining deformation measurements with contact force data, we estimate the underlying mechanical parameters and reconstruct the vegetation's response to interaction. This enables vegetation-aware navigation based on intrinsic environmental properties rather than platform-dependent metrics.
Chinese Translation
在自然环境中操作的自主机器人往往必须与植被进行互动，而不仅仅是避免它。在这种背景下，可通行性通常从机器人的角度定义，通过测量特定平台在环境中移动时的响应来进行评估。尽管这种观点在实践中是可行的，但它将环境的评估与机器人的自身动态纠缠在一起，使得最终的特征难以在不同平台之间转移。更重要的是，它并未直接反映植被本身的特性，而这些特性才是农业和环境监测等应用中互动和潜在损害的真正来源。为了解决这一局限性，我们建议通过植被的内在机械特性进行表征，而不依赖于任何特定机器人。通过将变形测量与接触力数据相结合，我们估计潜在的机械参数，并重建植被对互动的响应。这使得基于内在环境特性的植被感知导航成为可能，而不是依赖于平台特定的度量。
View on arXiv Download PDF AI Translation
cs.RO / 38 / 2608.26053

$R^3$: Training Robots to Reason in Natural Language via Reinforcement Learning

$R^3$: 通过强化学习训练机器人在自然语言中进行推理
Wu, Lehong, Qu, Yuxiao, Hu, Zheyuan, Zhang, Ivan, Wei, Limin, Erickson, Zackory, Kumar, Aviral
Abstract
Reasoning in language allows foundation models to spend more test-time compute on hard problems, such as those requiring decomposition, constraint tracking, and prediction of future consequences. Whether this mechanism can improve robotic manipulation remains unclear, where long-horizon tasks require tracking partial progress, reasoning about object relations, recovering from mistakes, and steering noisy low-level policies. In this paper, we study whether VLMs can be trained to reason directly in natural language to guide low-level manipulation policies. We introduce $R^3$, a simple post-training recipe that turns off-the-shelf VLMs into robotic reasoners: it first mid-trains a VLM on expert-generated reasoning traces to initialize the desired reasoning style, then improves the reasoner with single-step rubric-based RL from offline action data. Unlike prior robotic reasoning methods that mostly use structured traces as auxiliary supervision, $R^3$ trains free-form language reasoning to produce test-time guidance for action. We instantiate $R^3$ on Language Table and simulated bimanual grocery packing, two controlled testbeds for studying robotic reasoning and long-horizon manipulation. $R^3$ improves exploration and generalization across unseen tasks and significantly outperforms instruction-only imitation learning baselines on both benchmarks. Our analyses suggest that free-form language reasoning can function as a test-time compute mechanism for steering low-level policies. Our project page is available at https://robotic-reasoner.github.io/.
Chinese Translation
在语言中进行推理使基础模型能够在测试时对困难问题投入更多计算资源，例如那些需要分解、约束跟踪和预测未来后果的问题。然而，这种机制是否能改善机器人操作仍不清楚，因为长时间跨度的任务需要跟踪部分进展、推理物体关系、从错误中恢复以及引导嘈杂的低级策略。本文研究了是否可以训练视觉语言模型（VLMs）直接在自然语言中进行推理，以指导低级操作策略。我们提出了$R^3$，一种简单的后训练方案，将现成的VLMs转变为机器人推理器：首先在专家生成的推理轨迹上对VLM进行中期训练，以初始化所需的推理风格，然后通过离线动作数据的单步评分基础强化学习（RL）来改进推理器。与以往主要使用结构化轨迹作为辅助监督的机器人推理方法不同，$R^3$训练自由形式的语言推理，以在测试时为动作提供指导。我们在语言表和模拟双手杂货打包这两个受控测试平台上实现了$R^3$，以研究机器人推理和长时间跨度的操作。$R^3$在未见任务上的探索和泛化能力有所提升，并在两个基准测试中显著超越仅依赖指令的模仿学习基线。我们的分析表明，自由形式的语言推理可以作为引导低级策略的测试时计算机制。我们的项目页面可在 https://robotic-reasoner.github.io/ 查看。
View on arXiv Download PDF AI Translation
cs.RO / 39 / 2608.26058

One Policy, Many Embodiments: Unified Camera-Centric Action Geometry Pre-training for Heterogeneous Embodied Manipulation

一种政策，多种体现：异构具身操作的统一相机中心动作几何预训练
Xiaomi Embodied Intelligence Team, Macau, University of, :, Xu, Shaoqing, Li, Fang, Zhan, Guozhi, Duan, Zhixiang, Wang, Yuhan, Luo, Yuechen, Jiang, Shengyin, Li, Hanbing, Du, Zhiying, Wang, Longlong, Jiang, Longmei, Liang, Weixiang, Gong, Ying, Pan, Yong, Zhao, Ziping, Chen, Zhiyuan, You, Yangwei, Ma, Kun, Liu, Qinyuan, Ye, Hangjun, Yang, Zhi-xin
Abstract
Scaling generalist vision-language-action (VLA) policies is severely bottlenecked by the inherent heterogeneity of embodied data, which spans diverse robot morphologies, camera configurations, and low-level action spaces. Existing paradigms typically address this mismatch through explicit action retargeting, human-to-robot video synthesis, or dataset-specific adaptation branches, fundamentally hindering the joint learning of a unified policy. We introduce UCAG-P, a camera-centric unified action formulation that structurally aligns heterogeneous embodied datasets into a shared geometric action space. Rather than treating robot-specific commands as the shared policy target, UCAG-P represents manipulation through camera-observable anchor motion in image and camera-frame coordinates, treating robot arms, humanoids, and human hands as different embodiments of a common action schema. A geometry-conditioned action translator combines predicted motion with target-embodiment kinematics to produce executable controls. The resulting decoupled architecture allows a shared VLA policy to learn transferable manipulation geometry while retaining embodiment-specific controllability. UCAG-P is trained on 4.03K hours of robot and simulation data and 2.34K hours of human demonstrations. A single checkpoint reaches 98.3% on LIBERO, 88.7% and 89.2% on RoboTwin Easy and Hard, 82.0% zero-shot on LIBERO-Plus, and 62.0% on RoboCasa GR-1, without benchmark-specific fine-tuning.
Chinese Translation
通用视觉-语言-动作（VLA）政策的扩展受到具身数据固有异质性的严重制约，这些数据涵盖了多样的机器人形态、相机配置和低级动作空间。现有范式通常通过显式动作重定向、人机视频合成或特定数据集的适应分支来解决这种不匹配，根本上阻碍了统一政策的联合学习。我们提出了UCAG-P，一种以相机为中心的统一动作表述，结构上将异构具身数据集对齐到共享的几何动作空间。UCAG-P并不将特定于机器人的命令视为共享政策目标，而是通过图像和相机坐标系中可观察的锚定运动来表示操作，将机器人手臂、人形机器人和人类手作为共同动作模式的不同体现。几何条件动作翻译器将预测的运动与目标体现的运动学相结合，以生成可执行的控制。最终的解耦架构使得共享的VLA政策能够学习可转移的操作几何，同时保留特定体现的可控性。UCAG-P在4.03K小时的机器人和仿真数据以及2.34K小时的人类示范上进行训练。单个检查点在LIBERO上达到98.3%，在RoboTwin Easy和Hard上分别达到88.7%和89.2%，在LIBERO-Plus上实现82.0%的零-shot表现，以及在RoboCasa GR-1上达到62.0%，且无需特定基准的微调。
View on arXiv Download PDF AI Translation
cs.RO / 40 / 2608.26066

VirTooS: A ROS 2 - Unity Virtualization Toolkit for Fleet Management of Autonomous Mobile Robots

VirTooS：用于自主移动机器人车队管理的ROS 2 - Unity虚拟化工具包
Drudi, Andrea, Pichierri, Lorenzo, Testa, Andrea, Notarstefano, Giuseppe
Abstract
In this paper, we present VirTooS, a Python/C# toolkit designed to implement fleet-management tasks on teams of Autonomous Mobile Robots (AMRs). VirTooS leverages the Robot Operating System (ROS) 2 and Unity game engine to provide realistic, scalable virtual experiments in a mixed-reality environment. The toolbox allows users to easily generate and customize virtual scenarios for realistic simulations. Virtual and real sensors as, e.g., LiDARs, can be exploited to map and safely navigate in the mixed-reality environment. To enable distributed robotics experiments, we propose a set of tailored routines leveraging the ChoiRbot framework. As a motivating example, we show a set of experiments for task assignment problems in a virtual environment, allowing seamless interaction among real and virtual robots. Moreover, the package comes with a containerized suite to easily deploy it on different machines. The source code will be made publicly available on GitHub.
Chinese Translation
在本文中，我们介绍了VirTooS，一个旨在实现自主移动机器人（AMRs）团队车队管理任务的Python/C#工具包。VirTooS利用机器人操作系统（Robot Operating System, ROS）2和Unity游戏引擎，在混合现实环境中提供逼真、可扩展的虚拟实验。该工具箱使用户能够轻松生成和定制虚拟场景，以进行真实的模拟。虚拟和真实传感器，例如激光雷达（LiDAR），可用于在混合现实环境中进行地图绘制和安全导航。为了支持分布式机器人实验，我们提出了一套利用ChoiRbot框架的定制例程。作为一个激励示例，我们展示了一组在虚拟环境中进行任务分配问题的实验，允许真实机器人和虚拟机器人之间的无缝交互。此外，该软件包还配备了一个容器化套件，以便于在不同机器上进行部署。源代码将在GitHub上公开发布。
View on arXiv Download PDF AI Translation
cs.RO / 41 / 2608.26074

Gating Before Commitment: Anticipating Intent Divergence to Prevent Post-Interaction Decision Failures in Autonomous Driving

承诺前的门控：预测意图分歧以防止自主驾驶中的后交互决策失败
Xu, Cong, Sankar, Ravi
Abstract
Intent misinterpretation during vehicle interactions causes recurring planning failures. We study a decision layer in which a language-guided intent module reads structured descriptors, computes a smoothed intent-geometry divergence score, and gates the planned maneuver before commitment, upstream of a corridor envelope. On a replayed off-road departure and four crash clips under a frozen, disclosed implementation, gating is the only layer that repairs the plan: on the main case it fires 72 ms after the drift onset but 161 ms before the corridor exit, keeping the trajectory in the corridor in all ten replays. The first calibration draws nine false triggers in 5.9 minutes, each from scoring uncertainty as half a conflict; a preregistered redesign treating uncertainty as abstention cuts this to 0.341 per minute. Two ablations bound the model's contribution: the full score detects fastest on four of five failures under the deployed eligibility, three of five against the unvetoed rule (000871 by one cycle; 000228 by a pre-onset fire on an uncertain stretch that five clips cannot classify as signal or coincidence; dropping the confidence term costs two detections), while on in-domain tracks at equal false positives the geometric rule more than triples its detection. The evidence supports the gating mechanism; the model's demonstrated roles are the fastest detection on these failures and an uncertainty veto on the geometric rule.
Chinese Translation
在车辆交互过程中，意图误解导致规划失败的反复发生。我们研究了一个决策层，其中一个基于语言的意图模块读取结构化描述符，计算平滑的意图几何分歧评分，并在承诺之前对计划的操作进行门控，位于走廊包络的上游。在重放的越野出发和四个碰撞片段中，在一个冻结的、公开的实现下，门控是唯一能够修复计划的层：在主要案例中，它在漂移开始后72毫秒触发，但在走廊出口前161毫秒触发，确保在所有十次重放中轨迹保持在走廊内。第一次校准在5.9分钟内产生九次错误触发，每次都因评分不确定性被视为半个冲突；一个预注册的重新设计将不确定性视为弃权，将这一数字减少到每分钟0.341。两个消融实验界定了模型的贡献：在部署的合格条件下，完整评分在五次失败中的四次上检测速度最快，在未被否决的规则下，五次中的三次（000871通过一个周期；000228通过在五个片段无法分类为信号或巧合的不确定区段上预先触发；去掉置信度项损失了两个检测），而在同域轨道上以相等的假阳性，几何规则的检测率超过了三倍。证据支持门控机制；模型展示的角色是对这些失败的最快检测和对几何规则的不确定性否决。
View on arXiv Download PDF AI Translation
cs.RO / 42 / 2608.26076

Fast Generative Grasping via Lie Group-Constrained MeanFlow

基于李群约束的快速生成抓取方法
Bukhari, S. Talha, Wei, Yi, Ni, Ruiqi, Kingston, Zachary, Bera, Aniket
Abstract
Grasp synthesis is a core task in robotic manipulation, for which the solution typically forms a multimodal distribution rather than a point estimate. Generative robotic grasping aims to learn this distribution with deep generative models such as diffusion and flow-based approaches. The iterative nature of such generative models makes them flexible and generalizable; however, multi-step sampling impedes the time-critical operation required in robotics. We devise an approach to fast generative grasping based on MeanFlow on the product Lie group $\mathcal{G} = \mathrm{SO}(3) \times \mathbb{R}^3$. The training objective couples a purely algebraic semigroup consistency condition with Riemannian Conditional Flow Matching on $\mathcal{G}$ that anchors the average velocity to the data distribution. The resulting Lie Group-constrained MeanFlow formulation samples reliable grasps in $\leq 5$ network evaluations, matching the grasp generation performance of state-of-the-art diffusion and flow-based models on the ACRONYM dataset at millisecond-scale inference latency (up to $39\times$ speed-up). We further demonstrate that the approach directly translates to real-world robotic grasping without additional training or domain adaptation, exhibiting robust grasp synthesis under observation noise.
Chinese Translation
抓取合成是机器人操作中的核心任务，其解决方案通常形成多模态分布，而非点估计。生成式机器人抓取旨在利用深度生成模型（如扩散和流动基础方法）学习这一分布。这类生成模型的迭代特性使其灵活且具有良好的泛化能力；然而，多步采样妨碍了机器人所需的时间敏感操作。我们提出了一种基于产品李群 $ ext{G} = ext{SO}(3) imes ext{R}^3$ 的快速生成抓取方法，采用均值流（MeanFlow）。训练目标将纯代数半群一致性条件与在 $ ext{G}$ 上的黎曼条件流匹配（Riemannian Conditional Flow Matching）结合，锚定平均速度与数据分布。最终的李群约束均值流（MeanFlow）公式在不超过 5 次网络评估中采样出可靠的抓取，匹配了在 ACRONYM 数据集上最先进的扩散和流动基础模型的抓取生成性能，推理延迟达到毫秒级（速度提升高达 39 倍）。我们进一步证明，该方法可以直接应用于现实世界的机器人抓取，无需额外的训练或领域适应，展现出在观测噪声下的稳健抓取合成能力。
View on arXiv Download PDF AI Translation
cs.RO / 43 / 2608.26103

Zero-WAM: In-Context World-Action Modeling from Human Videos for Open-Ended Task Generalization

Zero-WAM：基于人类视频的上下文世界-动作建模用于开放式任务泛化
Zhou, Jiaming, Zhang, Qihang, Xu, Gangwei, Fan, Cunxin, Zhao, Yujie, Wang, Ruilin, Luo, Yiming, Yang, Shuai, Zhu, Xing, Shen, Yujun, Liang, Junwei, Xu, Yinghao
Abstract
Zero-shot cross-task generalization, where a policy must execute manipulation tasks never seen during training, remains a central challenge in robot learning. In large language models, a novel task can be performed simply by specifying it in the context, without any parameter update. This form of in-context learning (ICL) turns generalization into a problem of task specification. To achieve cross-task generalization, we bring this paradigm to robotic manipulation, and argue that the natural task specification for manipulation is a human video: unlike language, it provides rich visual cues about the intended task evolution. We present Zero-WAM, a causal video-action model that executes unseen tasks by following in-context human video guidance. To address the scarcity of task-rich paired human-robot data, we propose an automatic pipeline that converts task-sampled robot trajectories into semantically matched human videos, yielding HumanGen, a dataset of 74.2K human-robot ICL pairs across 8.6K tasks. For model training, we further introduce an in-context future chunk prediction (IFP) objective that suppresses shortcuts learned from seen tasks and forces the policy to draw task information from the video prompt. On seven unseen tasks in RoboTwin 2.0 simulation, Zero-WAM achieves a 47.0% average success rate, an absolute improvement of 29.5 percentage points over the strongest video-action baseline. In real-world evaluations, it follows human video guidance to generalize to unseen task configurations involving multi-object scenes, long-horizon manipulation, and fine-grained insertion.
Chinese Translation
零样本跨任务泛化，即策略必须执行在训练过程中未见过的操作任务，仍然是机器人学习中的一个核心挑战。在大型语言模型中，可以通过在上下文中指定新任务来简单地执行该任务，而无需任何参数更新。这种上下文学习（In-Context Learning, ICL）将泛化转变为任务规范的问题。为了实现跨任务泛化，我们将这一范式引入机器人操作，并认为操作的自然任务规范是人类视频：与语言不同，它提供了关于预期任务演变的丰富视觉线索。我们提出了Zero-WAM，一种因果视频-动作模型，通过遵循上下文中的人类视频指导来执行未见任务。为了解决任务丰富的配对人类-机器人数据稀缺的问题，我们提出了一种自动化流程，将任务采样的机器人轨迹转换为语义匹配的人类视频，生成了HumanGen，一个包含74.2K人类-机器人ICL对的数据库，涵盖8.6K任务。为了进行模型训练，我们进一步引入了上下文未来片段预测（In-Context Future Chunk Prediction, IFP）目标，以抑制从已见任务中学习的捷径，并迫使策略从视频提示中提取任务信息。在RoboTwin 2.0仿真中的七个未见任务上，Zero-WAM实现了47.0%的平均成功率，相较于最强的视频-动作基线提高了29.5个百分点。在现实世界评估中，它遵循人类视频指导，泛化到涉及多物体场景、长时间操作和精细插入的未见任务配置。
View on arXiv Download PDF AI Translation
计算机视觉 (Computer Vision)
90
cs.CV / 1 / 2608.24934

Fusing Perceptual Vision Experts with Multimodal Large Language Models for Explainable Plant Disease Diagnosis: From Benchmark Imagery to Real-World Robotic Field Validation

融合感知视觉专家与多模态大型语言模型用于可解释的植物疾病诊断：从基准图像到现实世界的机器人田间验证
Sapkota, Ranjan, Roumeliotis, Konstantinos I., Xie, Pengyao, Tselikas, Nikolaos D., Xiang, Lirong, Karkee, Manoj
Abstract
Accurate field plant disease diagnosis requires reliable fusion of uncertain and conflicting perceptual evidence. We present the Hybrid Hierarchical Multi-Agent Framework (H$^{2}$MAF), combining decision-level fusion of EfficientNet-B3 and ConvNeXt-Tiny with semantic arbitration by open-weight multimodal large language models (MLLMs), Gemma 4 E4B and Qwen3.5 4B, using structured JSON evidence to generate explainable diagnoses, risk levels, treatment urgency, and financial exposure. (H$^{2}$MAF) is evaluated on 14,364 images (1,370 test images) across PlantDoc (2,922 images, 27 classes) and two non-public, continuously captured Cornell robot-acquired field datasets: Stage 2 (20 GB; 4,215 images) and Stage 4 (40 GB; 7,227 images), covering Early Blight, Late Blight, and Septoria Leaf Spot under uncontrolled field conditions. On PlantDoc, Gemma improves accuracy from 63.9% to 68.5%, achieving +7.6 points on the 41.7% CNN-conflict subset. Cornell accuracies reach 99.3% and 98.9%, with only 1.7-4.1% disagreement, demonstrating conflict-dependent MLLM utility. The critical-risk error of gemma is 0.14-0.5 points, whereas Qwen overflags by 3.5-14.4 points. These results establish MLLM arbitration as a promising, yet calibration-dependent, approach for explainable agricultural AI and robotic field decision support. Github Link: https://github.com/Applied-AI-Research-Lab/Explainable-AI-Plant-Disease-Detection
Chinese Translation
准确的田间植物疾病诊断需要可靠地融合不确定和相互冲突的感知证据。我们提出了混合层次多代理框架（Hybrid Hierarchical Multi-Agent Framework, H$^{2}$MAF），结合了EfficientNet-B3和ConvNeXt-Tiny的决策级融合，以及通过开放权重多模态大型语言模型（Multimodal Large Language Models, MLLMs）进行的语义仲裁，使用结构化的JSON证据生成可解释的诊断、风险等级、治疗紧迫性和财务暴露。H$^{2}$MAF在PlantDoc（2,922张图像，27个类别）和两个非公开的、持续捕获的康奈尔机器人获取的田间数据集（阶段2：20 GB；4,215张图像，阶段4：40 GB；7,227张图像）上进行了评估，涵盖了早疫病、晚疫病和叶斑病，且在不受控的田间条件下进行。Gemma在PlantDoc上的准确率从63.9%提高到68.5%，在41.7%的CNN冲突子集上实现了+7.6个百分点。康奈尔数据集的准确率达到99.3%和98.9%，仅有1.7-4.1%的不一致，展示了依赖冲突的MLLM的实用性。Gemma的关键风险错误为0.14-0.5个百分点，而Qwen则过度标记3.5-14.4个百分点。这些结果确立了MLLM仲裁作为一种有前景的、但依赖于校准的可解释农业人工智能和机器人田间决策支持的方法。Github链接：https://github.com/Applied-AI-Research-Lab/Explainable-AI-Plant-Disease-Detection
View on arXiv Download PDF AI Translation
cs.CV / 2 / 2608.24935

A Lightweight Multimodal Vision-Language Framework for Early-Stage Anatomical Green Fruit Classification in Commercial Orchards

一种轻量级多模态视觉-语言框架用于商业果园早期阶段解剖绿色果实分类
Sapkota, Ranjan, Bu, William, Chen, Chen, Xu, Yunjun, Karkee, Manoj
Abstract
Accurate identification of early-stage apple fruitlet anatomical structures, including the calyx, fruitlet body, and peduncle, is essential for robotic thinning, crop-load management, and other precision orchard operations. This study presents a lightweight multimodal vision-language framework that adapts TinyCLIP for fine-grained fruitlet anatomy classification in complex orchard environments. A dataset of 600 high-resolution RGB images collected from Scilate and Scifresh apple orchards was converted into 224 x 224 image patches and annotated for three anatomical classes. Domain-specific language prompts, such as ``a photo of a class,'' were used to guide multimodal alignment between orchard imagery and horticultural structures. A sliding-window inference strategy with a stride of 112 pixels aggregates patch-level predictions into spatial heatmaps, enabling interpretable whole-image localization of fruitlet components relevant to robotic thinning. Patch-level evaluation on an NVIDIA T4 GPU achieved F1-scores of 0.95 for calyx, 0.98 for fruitlet, and 0.85 for peduncle, with a macro-F1 score of 0.93. Deployment-oriented optimization using ONNX and TensorRT enabled efficient inference on NVIDIA Jetson hardware, preserved accuracy under INT8 quantization, and supported model sizes of approximately 127-137 MB with millisecond-level patch inference. These results demonstrate that lightweight vision-language models can provide interpretable and edge-deployable perception for automated fruitlet analysis and future robotic thinning systems. The source code and implementation details are publicly available at https://github.com/WilliamBu1/A-Lightweight-Vision-Language-Model-for-Early-Stage-Fruitlet-Classification-in-Apple-Orchards.
Chinese Translation
准确识别早期苹果果实的解剖结构，包括萼片、果实主体和果梗，对于机器人疏果、作物负荷管理及其他精准果园操作至关重要。本研究提出了一种轻量级多模态视觉-语言框架，该框架适配了TinyCLIP，以实现复杂果园环境中细粒度果实解剖分类。我们收集了来自Scilate和Scifresh苹果果园的600张高分辨率RGB图像，并将其转换为224 x 224的图像块，并对三种解剖类别进行了标注。使用领域特定的语言提示，如“某类的照片”，来指导果园图像与园艺结构之间的多模态对齐。采用步幅为112像素的滑动窗口推理策略，将块级预测聚合为空间热图，从而实现对与机器人疏果相关的果实组件的可解释整体图像定位。在NVIDIA T4 GPU上进行的块级评估显示，萼片的F1分数为0.95，果实的F1分数为0.98，果梗的F1分数为0.85，宏观F1分数为0.93。使用ONNX和TensorRT进行的面向部署的优化实现了在NVIDIA Jetson硬件上的高效推理，在INT8量化下保持了准确性，并支持约127-137 MB的模型大小，具有毫秒级的块推理。这些结果表明，轻量级视觉-语言模型可以为自动果实分析和未来的机器人疏果系统提供可解释的边缘部署感知。源代码和实现细节可在https://github.com/WilliamBu1/A-Lightweight-Vision-Language-Model-for-Early-Stage-Fruitlet-Classification-in-Apple-Orchards获取。
View on arXiv Download PDF AI Translation
cs.CV / 3 / 2608.24956

Synergising Local Geo-Environmental Characteristics with Spatial Context for Enhancing Landslide Susceptibility Mapping

将地方地理环境特征与空间背景协同以增强滑坡易发性制图
Cheng, Yusen, Fan, Lei, Zhu, Qinfeng, Zhang, Cheng, Li, Yangyang, Mahabir, Ron
Abstract
Data-driven methods are widely used in landslide susceptibility mapping (LSM) because they can effectively model the complex relationships between landslides and geo-environmental conditions. Existing data-driven approaches generally follow two types of data representations. Pixel-based models focus solely on the geo-environmental characteristics of a specific landslide but neglect the influence of its surrounding environment. Patch-based models incorporate surrounding spatial context but may include pixels with weak or no spatial relevance to the target landslide location. To address this limitation, this study proposes a Local-Geo and Spatial Context Fusion (LGSCF) strategy, which synergises the geo-environmental characteristics of landslide points with their corresponding spatial context through a feature-wise modulation mechanism. We tested the LGSCF strategy by integrating it into several representative convolutional neural network (CNN) architectures, creating nine different LGSCF-based models. The study area covers approximately 2644 km2 across Jenai and Sinyi Townships in Nantou County, Taiwan, and the dataset comprises 5332 landslide samples and an equal number of non-landslide samples. The results show that LGSCF-based models consistently outperform their original versions, achieving F1-scores up to 87.09% and AUC values up to 0.9472. Furthermore, the susceptibility maps produced by LGSCF-based models show that known landslides are more accurately concentrated in "very high" susceptibility zones with fewer misclassifications. These findings demonstrate that our fusion strategy can significantly improve the accuracy of landslide susceptibility mapping.
Chinese Translation
数据驱动的方法在滑坡易发性制图（LSM）中被广泛应用，因为它们能够有效建模滑坡与地理环境条件之间的复杂关系。现有的数据驱动方法通常遵循两种类型的数据表示。基于像素的模型仅关注特定滑坡的地理环境特征，但忽略了其周围环境的影响。基于块的模型则结合了周围的空间背景，但可能包含与目标滑坡位置相关性较弱或没有相关性的像素。为了解决这一局限性，本研究提出了一种地方地理与空间背景融合（Local-Geo and Spatial Context Fusion, LGSCF）策略，通过特征调制机制将滑坡点的地理环境特征与其对应的空间背景协同。我们通过将LGSCF策略整合到几种代表性的卷积神经网络（CNN）架构中，测试了该策略，创建了九种不同的基于LGSCF的模型。研究区域覆盖了台湾南投县的仁爱乡和信义乡，面积约为2644平方公里，数据集包含5332个滑坡样本和相同数量的非滑坡样本。结果表明，基于LGSCF的模型在性能上始终优于其原始版本，F1分数高达87.09%，AUC值高达0.9472。此外，基于LGSCF的模型生成的易发性图显示，已知滑坡更准确地集中在“非常高”易发性区域，误分类更少。这些发现表明，我们的融合策略可以显著提高滑坡易发性制图的准确性。
View on arXiv Download PDF AI Translation
cs.CV / 4 / 2608.24966

Targeting the Attention Heads Behind Object Hallucination in LLaVA

针对 LLaVA 中物体幻觉的注意力头的研究
Sandhu, Armaan, Senapati, Abhilasha, Kammachi, Hima
Abstract
Vision-language models such as LLaVA-1.5-7B often hallucinate objects absent from the image when generating captions. We ask whether an interpretability diagnosis of this failure can guide a targeted fix, and we measure what that fix actually changes. We rank attention heads by how much their image attention drops around hallucinated object words, then screen the shortlist by ablating candidate heads and measuring the change in hallucination-token log probability, yielding a 32-head set. We restrict two interventions to these heads: a head-sliced LoRA adapter and an inference-time grounding controller. On 400 held-out COCO images, the combined method lowers CHAIRs (the fraction of captions with a hallucinated object) from 0.370 to 0.230 and CHAIRi (the fraction of hallucinated object mentions) from 0.156 to 0.096 (p < 0.001, paired sign-flip tests). Two controls sharpen attribution. A random-head LoRA control, matched layer-for-layer and trained identically, performs no better than the matched baseline on a separate 200-image control split, supporting the role of head selection rather than LoRA capacity. Under fixed decoding budgets, the CHAIR reduction persists and grows with budget (23% at 64 tokens to 58% at 128), arguing against a pure max-token or truncation artifact, although the method remains shorter and more conservative. The resulting behavior reduces unsupported object mentions while also lowering object recall (0.78 to 0.70). We present a diagnosis-to-intervention pipeline for object hallucination, and, more importantly, a controlled account of what acting on the diagnostic signal actually does: it localizes intervention sites with real, non-random leverage, reported as a behavioral profile rather than a single score.
Chinese Translation
视觉-语言模型如 LLaVA-1.5-7B 在生成标题时常常会幻觉出图像中不存在的物体。我们探讨这种失败的可解释性诊断是否能够指导有针对性的修复，并测量这种修复实际改变了什么。我们根据注意力头在幻觉物体词汇周围的图像注意力下降程度对注意力头进行排名，然后通过消融候选头并测量幻觉标记对数概率的变化来筛选短名单，最终得到一个包含 32 个头的集合。我们将两个干预措施限制在这些头上：一个是头切片的 LoRA 适配器，另一个是推理时的基础控制器。在 400 张保留的 COCO 图像上，结合方法将 CHAIR（包含幻觉物体的标题比例）从 0.370 降低到 0.230，将 CHAIRi（幻觉物体提及的比例）从 0.156 降低到 0.096（p < 0.001，配对符号翻转测试）。两个控制实验增强了归因效果。一个随机头的 LoRA 控制，与层层匹配并以相同方式训练，在一个单独的 200 张图像控制分割中表现不佳，支持了头选择的作用而非 LoRA 容量。在固定解码预算下，CHAIR 的减少持续存在并随着预算的增加而增长（在 64 个标记时为 23%，在 128 个标记时为 58%），反驳了纯最大标记或截断伪影的观点，尽管该方法仍然较短且更为保守。最终的行为减少了不支持的物体提及，同时也降低了物体召回率（从 0.78 降低到 0.70）。我们提出了一个针对物体幻觉的诊断到干预的流程，更重要的是，提供了一个关于对诊断信号采取行动实际效果的控制性说明：它定位了具有真实、非随机影响的干预位置，并以行为特征的形式报告，而不是单一评分。
View on arXiv Download PDF AI Translation
cs.CV / 5 / 2608.25068

SHIFT-LLM: Distribution Shift Correction in Depth-Pruned LLMs

SHIFT-LLM：深度剪枝大语言模型中的分布偏移校正
Bahri, Ali, Li, Hang, Li, Hongliang, Chen, Zhitang
Abstract
Depth pruning removes entire Transformer blocks to reduce the inference cost of large language models, but disrupts the hidden-state distributions expected by downstream layers, leading to significant accuracy loss. We introduce SHIFT-LLM, a training-free post-pruning correction framework that inserts a Linear Residual Adapter (LRA) at each pruning site. Each LRA preserves the identity pathway of the original residual block and adds a lightweight affine residual correction. This correction is calibrated via closed-form least-squares regression on a small held-out set, without gradient computation, to approximate the missing residual update produced by the pruned block. Together with the preserved identity pathway, the resulting LRA output approximates the hidden state produced by the original block, thereby mitigating the distributional mismatch introduced by layer removal while avoiding the expensive attention and feed-forward computations of the removed blocks. The resulting LRAs support low-rank factorization and exact merging across consecutive pruned layers for additional compression, and combine naturally with parameter-efficient fine-tuning for further recovery beyond fine-tuning the pruned model alone. Experiments on five model families, six layer-selection criteria, and seven zero-shot benchmarks show that SHIFT-LLM consistently recovers accuracy lost to depth pruning across most configurations, achieving gains up to +15.7 points on Llama-3.1-8B-Instruct while requiring only a few hundred calibration samples and no gradient computation.
Chinese Translation
深度剪枝通过移除整个Transformer块来降低大型语言模型的推理成本，但这会破坏下游层所期望的隐藏状态分布，导致显著的准确性损失。我们提出了SHIFT-LLM，一种无训练的后剪枝校正框架，在每个剪枝位置插入线性残差适配器（Linear Residual Adapter, LRA）。每个LRA保留原始残差块的恒等路径，并添加轻量级的仿射残差校正。该校正通过在一个小的保留集上进行封闭形式的最小二乘回归进行校准，无需梯度计算，以近似由剪枝块产生的缺失残差更新。结合保留的恒等路径，最终的LRA输出近似于原始块产生的隐藏状态，从而减轻了由于移除层而引入的分布不匹配，同时避免了被移除块的昂贵注意力和前馈计算。结果LRA支持低秩分解和连续剪枝层之间的精确合并以实现额外压缩，并与参数高效的微调自然结合，以进一步恢复超出单独微调剪枝模型的性能。在五个模型家族、六个层选择标准和七个零样本基准上的实验表明，SHIFT-LLM在大多数配置中始终恢复了因深度剪枝而损失的准确性，在Llama-3.1-8B-Instruct上实现了高达+15.7的增益，同时仅需几百个校准样本且无需梯度计算。
View on arXiv Download PDF AI Translation
cs.CV / 6 / 2608.25140

RefLAM: A Reference-Grounded Line Annotation Pipeline for Historical Arabic Manuscripts

RefLAM：一种基于参考的历史阿拉伯手稿行注释管道
Guechaoui, Mohamed, Zellagui, Mohamed Diaa, Chaib, Souleyman, Dhelim, Sahraoui
Abstract
Existing approaches to building line-level Arabic handwritten-text-recognition (HTR) training data either rely on fully manual annotation, which does not scale, or on automatic OCR-to-reference alignment methods not yet extended to multi-script, two-zone (main-plus-margin) manuscript layouts with a provable correctness guarantee. We present RefLAM (Reference-grounded Line Annotation for Manuscripts), a pipeline converting manuscript page images and clean transcriptions into validated, line-level ground truth without sacrificing human oversight. RefLAM couples a deep-learning page-segmentation model with a multimodal large language model (MLLM) for structured OCR and a diacritic-agnostic fuzzy alignment engine that grounds each OCR line in a contiguous span of the reference text, with a character-level confidence score in $[0,100]$. A perfect score is provably equivalent to character-for-character identity of the normalised strings (the Confidence-100 rule), verified with no counterexample across the released corpus. A reviewer can thus trust a perfect score, confirming most lines at a glance rather than retyping them, so annotation becomes triaged, with attention concentrated on uncertain alignments. Across 7 fully page-validated books we measured a 75$\times$ throughput gain over manual annotation (3,000 vs. 40 lines/hr); applying the same guarantee to 7 further books, we retained 16,533 confidence-100 main-text lines within one week, excluding sub-100 lines rather than manually correcting them. Using RefLAM, we release AraMS-28k: 14 historical Arabic manuscript books, 3,043 pages, and 27,971 main-text and 629 margin-line annotations with bounding boxes, layout labels, and insertion anchors for 191 margin entries (30.4%). We also finetune Muharaf-pretrained baselines (including HATFormer) on AraMS-28k and report CER results confirming its practical utility for downstream HTR training.
Chinese Translation
现有的构建行级阿拉伯手写文本识别（HTR）训练数据的方法要么依赖于完全手动注释，这种方法无法扩展，要么依赖于尚未扩展到多脚本、双区（主区加边缘区）手稿布局的自动OCR与参考对齐方法，并且缺乏可证明的正确性保障。我们提出了RefLAM（基于参考的手稿行注释），这是一个将手稿页面图像和干净的转录文本转换为经过验证的行级真实数据的管道，而不牺牲人工监督。RefLAM将深度学习页面分割模型与多模态大型语言模型（MLLM）结合，用于结构化OCR，以及一个与音调无关的模糊对齐引擎，该引擎将每个OCR行与参考文本的连续范围相结合，并提供$[0,100]$的字符级置信度评分。完美的评分可证明等同于标准化字符串的逐字符一致性（置信度-100规则），并在发布的语料库中没有反例进行验证。因此，审阅者可以信任完美的评分，快速确认大多数行，而无需重新输入，从而使注释变得有序，注意力集中在不确定的对齐上。在7本完全经过页面验证的书籍中，我们测得相较于手动注释（每小时3,000行对比40行），提高了75倍的通量；在另外7本书中应用相同的保障，我们在一周内保留了16,533条置信度-100的主文本行，排除了低于100的行，而不是手动更正它们。使用RefLAM，我们发布了AraMS-28k：14本历史阿拉伯手稿书籍，3,043页，以及27,971条主文本和629条边缘行注释，附有边界框、布局标签和191条边缘条目的插入锚点（占30.4%）。我们还在AraMS-28k上微调了Muharaf预训练的基线模型（包括HATFormer），并报告了CER结果，确认其在下游HTR训练中的实际效用。
View on arXiv Download PDF AI Translation
cs.CV / 7 / 2608.25148

Can You Trust Frozen Hematology Foundation Models under Acquisition Shift?

在获取偏移下，你能信任冷冻血液学基础模型吗？
Sharma, Jai Kumar, Tapadiya, Peeyush
Abstract
Frozen hematology foundation-model (FM) embeddings reach near-saturated in-domain white-blood-cell (WBC) accuracy, but clinical deployment demands reliability across scanners, sites, stains and preparation pipelines. We audit 15 frozen encoders (hematology, pathology, and general vision) across four public single-cell acquisition domains along two axes: accuracy robustness and calibration. In-domain linear-probe macro-F1 is saturated (0.98-0.997), yet cross-dataset macro-F1 drops 34-72% and rankings re-order: DinoBloom-L, the in-domain best, falls to 10th of 15 on the most-shifted target (MLL23) at the benchmark's shared 224-px input, behind RedDino and several general and pathology encoders. Rank transfer is probe-dependent: 1-NN retrieval is more stable on average than a source-fitted linear head (median $\rho$ 0.65 vs 0.45), but neither probe universally predicts target robustness. Calibration also collapses: source-trained probes are nearly calibrated in-domain (expected calibration error, ECE, 0.004) but confidently wrong off-domain (ECE 0.35), and source-fitted temperature scaling transfers poorly. We further audit pretraining exposure and identify MLL23 as DinoBloom's internal cohort; because DinoBloom's only held-out dataset is also our source domain, this benchmark cannot isolate exposure from scanner-associated shift. Label-free adaptation and marginal-entropy-based model selection appear safe under balanced evaluation but fail under realistic WBC class-prior shift. Class-Balanced Re-standardization (CBR), a training-free pseudo-label-balanced feature normalization, improves all evaluated target-prior scenario means and partially improves calibration, although encoder-level exceptions and residual miscalibration remain. Hematology FM benchmarks must therefore jointly audit accuracy, calibration, exposure, and class-prior robustness.
Chinese Translation
冷冻血液学基础模型（FM）嵌入在领域内的白细胞（WBC）准确率接近饱和，但临床应用要求在不同扫描仪、地点、染色和准备流程中具有可靠性。我们对15个冷冻编码器（血液学、病理学和通用视觉）进行了审计，涵盖四个公共单细胞获取领域，从准确性鲁棒性和校准两个维度进行评估。在领域内的线性探针宏F1值已饱和（0.98-0.997），然而跨数据集的宏F1值下降了34-72%，排名发生了变化：在最偏移的目标（MLL23）上，领域内最佳的DinoBloom-L在基准的共享224像素输入下降至15个中的第10位，落后于RedDino和若干通用及病理编码器。排名转移依赖于探针：1-NN检索的平均稳定性优于源拟合的线性头（中位数$ ho$ 0.65对比0.45），但无论哪种探针都无法普遍预测目标鲁棒性。校准也出现崩溃：源训练的探针在领域内几乎校准（预期校准误差，ECE，0.004），但在领域外自信地错误（ECE 0.35），源拟合的温度缩放转移效果不佳。我们进一步审计了预训练暴露情况，并将MLL23识别为DinoBloom的内部队列；由于DinoBloom唯一的保留数据集也是我们的源领域，因此该基准无法将暴露与扫描仪相关的偏移区分开。无标签适应和基于边际熵的模型选择在平衡评估下似乎是安全的，但在现实的WBC类别先验偏移下失败。类别平衡再标准化（CBR），一种无训练的伪标签平衡特征归一化方法，改善了所有评估的目标先验场景均值，并部分改善了校准，尽管编码器级别的例外和残余的误校准仍然存在。因此，血液学FM基准必须共同审计准确性、校准、暴露和类别先验鲁棒性。
View on arXiv Download PDF AI Translation
cs.CV / 8 / 2608.25157

What Do Audio-Visual Synchronization Metrics Actually Measure?

音视频同步度量实际测量了什么？
Sharma, Jai Kumar, Tapadiya, Peeyush
Abstract
Automatic AV-sync metrics are widely used to rank and train audio-visual generators, but they are rarely audited as measurement instruments. We jointly audit AV-Align, ImageBind AV-relevance, JavisScore, and Synchformer/DeSync under a common reliability protocol: controlled-distortion monotonicity, preprocessing sensitivity, rank uncertainty, cross-metric agreement, PEAVS-proxy agreement, and learned fusion. The result is an axis split, not a single winner: Synchformer/DeSync is the strongest temporal-offset tracker ($\tau=0.84$), ImageBind/JavisScore better match the PEAVS human-aligned proxy ($\tau=0.20$) and content-disruption families, and AV-Align is the weakest standalone metric. The metrics mutually disagree (Krippendorff $\alpha=0.066$), and neither linear nor simple $k$-NN fusion improves PEAVS agreement over the best individual metric. We recommend reporting AV-sync as a Reliability Card (metric-family breakdowns with confidence intervals) rather than a single bare synchronization score.
Chinese Translation
自动音视频同步度量广泛用于对音视频生成器进行排名和训练，但作为测量工具的审计却很少进行。我们在一个共同的可靠性协议下对 AV-Align、ImageBind AV 相关性、JavisScore 和 Synchformer/DeSync 进行了联合审计：控制失真单调性、预处理敏感性、排名不确定性、跨度量一致性、PEAVS 代理一致性和学习融合。结果是一个轴向分裂，而不是单一赢家：Synchformer/DeSync 是最强的时间偏移跟踪器（$ au=0.84$），ImageBind/JavisScore 更好地匹配 PEAVS 人类对齐代理（$ au=0.20$）和内容干扰家族，而 AV-Align 是最弱的独立度量。这些度量之间相互不一致（Krippendorff $eta=0.066$），并且无论是线性还是简单的 $k$-NN 融合都未能改善 PEAVS 一致性，相较于最佳单一度量。我们建议将 AV-sync 报告为可靠性卡（度量家族细分及置信区间），而不是单一的同步得分。
View on arXiv Download PDF AI Translation
cs.CV / 9 / 2608.25168

See More, Detect Less? Taming Information Leakage in Multi-View Anomaly Detection

看得更多，检测更少？驯服多视角异常检测中的信息泄露
Chen, Shang-Fu, Peng, Kuan-Chuan, Wu, Jhih-Ciang, Cheng, Wen-Huang, Hua, Kai-Lung
Abstract
In multi-view anomaly detection, more cross-view information can actually hurt. When multiple inspection views are naively fused in a reconstruction-based pipeline, normal cues from intact views propagate to the decoder, which faithfully reconstructs anomalous regions, collapsing the reconstruction gap the detector depends on. We call this failure mode \emph{cross-view information leakage} and show that effective multi-view fusion must explicitly restrict the information reaching the decoder. Building on this insight, we present GLAD(Global-Local Attention Driven framework), the first framework combining vision foundation model features with local and global cross-view fusion for multi-view anomaly detection. The Multi-view Merging Attention (MMA) module performs local cross-view fusion at linear complexity with learnable view importance weighting and token-wise gating, letting each view selectively incorporate fine-grained evidence from other views at $\mathcal{O}(N)$ cost. The Object-Guided Attention (OGA) module captures global context by aggregating class tokens from all views into a single object-level representation and broadcasting it back to patch tokens via temperature-scaled sigmoid gating, replacing the original patch representations rather than adding a residual to preserve the reconstruction gap. Experiments on Real-IAD and MANTA-Tiny show that GLAD outperforms state-of-the-art methods across sample-, image-, and pixel-level metrics, confirming that principled information restriction is key to multi-view anomaly reasoning.
Chinese Translation
在多视角异常检测中，更多的跨视角信息实际上可能会造成负面影响。当多个检查视角在基于重建的流程中被简单融合时，来自完整视角的正常线索会传播到解码器，解码器忠实地重建异常区域，从而缩小了检测器所依赖的重建差距。我们将这种失败模式称为 extit{跨视角信息泄露}，并表明有效的多视角融合必须明确限制到达解码器的信息。在这一见解的基础上，我们提出了GLAD（全球-局部注意力驱动框架），这是第一个将视觉基础模型特征与局部和全球跨视角融合相结合的多视角异常检测框架。多视角合并注意力（MMA）模块以线性复杂度执行局部跨视角融合，具有可学习的视角重要性加权和逐标记门控，使每个视角能够以$ extmath{O}(N)$的成本选择性地整合来自其他视角的细粒度证据。对象引导注意力（OGA）模块通过将所有视角的类别标记聚合成单一对象级表示来捕获全局上下文，并通过温度缩放的sigmoid门控将其广播回补丁标记，替换原始的补丁表示，而不是添加残差以保持重建差距。在Real-IAD和MANTA-Tiny上的实验表明，GLAD在样本、图像和像素级指标上均优于最先进的方法，确认了原则性的信息限制是多视角异常推理的关键。
View on arXiv Download PDF AI Translation
cs.CV / 10 / 2608.25176

Lowering the Barrier to AI-Driven Inspection: A No-Code Workflow for Automated Structural Defect Detection

降低人工智能驱动检测的门槛：无代码工作流用于自动化结构缺陷检测
Holm, Michael, McElroy, Tanner, Zhang, Xinghang, Lin, Guang
Abstract
Structural health monitoring (SHM) is essential in modern engineering, providing data for condition-based maintenance, lifecycle assessment, and predictive decision-making. Traditionally, SHM relied on visual inspection to detect defects such as cracks and deformations. Early computer vision (CV) methods, including thresholding, edge detection, and handcrafted features, aimed to automate this process but were highly sensitive to noise, imaging variations, and multiscale defects, limiting their reliability. Recent advances in machine learning, particularly convolutional neural networks (CNNs) and You Only Look Once (YOLO), have improved defect detection accuracy and enabled real-time analysis. However, adoption in SHM remains limited due to technical barriers such as data labeling, model training, and deployment, which typically require programming expertise. To address this gap, we introduce YOLOEZ, an open-source, GUI-based tool for end-to-end YOLO model application. YOLOEZ integrates data labeling, training, and inference into a single interface, enabling high-performance model development without code while supporting reproducible workflows. Evaluation against existing software and classical image processing demonstrates that YOLOEZ not only outperforms traditional methods across most detection metrics, but also lowers adoption barriers present in other modern CV tools. By combining accuracy with accessibility, YOLOEZ facilitates wider use of AI-driven monitoring for predictive maintenance, digital twins, and intelligent structural systems.
Chinese Translation
结构健康监测（SHM）在现代工程中至关重要，为基于状态的维护、生命周期评估和预测决策提供数据。传统上，SHM依赖于视觉检查来检测裂缝和变形等缺陷。早期的计算机视觉（CV）方法，包括阈值处理、边缘检测和手工特征，旨在自动化这一过程，但对噪声、成像变化和多尺度缺陷高度敏感，限制了其可靠性。最近在机器学习方面的进展，特别是卷积神经网络（CNN）和You Only Look Once（YOLO），提高了缺陷检测的准确性，并实现了实时分析。然而，由于数据标注、模型训练和部署等技术障碍，SHM的采用仍然有限，这通常需要编程专业知识。为了解决这一问题，我们推出了YOLOEZ，一个基于图形用户界面（GUI）的开源工具，用于端到端的YOLO模型应用。YOLOEZ将数据标注、训练和推理集成到一个单一界面中，使得高性能模型开发无需编码，同时支持可重复的工作流。与现有软件和经典图像处理的评估表明，YOLOEZ不仅在大多数检测指标上优于传统方法，还降低了其他现代计算机视觉工具存在的采用障碍。通过将准确性与可及性相结合，YOLOEZ促进了人工智能驱动的监测在预测维护、数字双胞胎和智能结构系统中的更广泛应用。
View on arXiv Download PDF AI Translation
cs.CV / 11 / 2608.25178

Lightweight Machine Learning-Driven Monocular Sidewalk Path Extraction for Embedded Micromobility Navigation

轻量级机器学习驱动的单目人行道路径提取用于嵌入式微型出行导航
Mijiddorj, Lkhanaajav, Yan, Yang, Beringer, Tyler, Mijiddorj, Bilguunzaya, Ho, Alex N., Xu, Bin, Weng, Binbin
Abstract
Sidewalk-scale path extraction demands perception and planning that run reliably on compact, low-power hardware in cluttered, map-sparse environments. We present a monocular vision pipeline for sidewalk path extraction in micromobility systems that progresses through three design iterations, from a skeleton-graph baseline through distance-transform corridor planning to a lightweight image-space architecture, and provides a systematic comparison of five path-planning methods across both bird's-eye-view (BEV) and image-space domains. A compact SegFormer-B0 student model, trained with a semi-supervised teacher-student framework using OneFormer Swin-L pseudo-labels, achieves a hand-annotated IoU of 0.946 at 11.7 ms per frame, improving over the baseline checkpoint (IoU 0.758, 18.9 ms). In a controlled planner comparison on 32 hand-labeled frames, image-space midpoint planning achieves the lowest lateral center error (14.3 px) at 2.2 ms, a 421x speedup over BEV distance-transform planning (926.8 ms, 65.0 px center error), while maintaining comparable mask-path alignment (98.5% versus 98.6%). A full-video replay across six campus sequences (22,679 frames) confirms that the improved segmentation reduces temporal instability from 1.46% to 0.33% and increases template-path availability from 73.7% to 79.3%. We further show that BEV-only path extraction is fragile in monocular settings: in one profiled run, 99.3% of frames produced no valid BEV path. The final recommended architecture, image-space midpoint primary, image-space distance-transform fallback, and BEV reserved for visualization, runs the full perception-to-path stack in under 50 ms per frame on CPU, making it suitable for embedded pedestrian-speed micromobility systems.
Chinese Translation
人行道规模的路径提取需要在复杂、地图稀疏的环境中，在紧凑、低功耗的硬件上可靠运行的感知和规划。我们提出了一种用于微型出行系统的人行道路径提取的单目视觉管道，该管道经过三个设计迭代，从骨架图基线到距离变换走廊规划，再到轻量级图像空间架构，并系统性地比较了五种路径规划方法在鸟瞰图（BEV）和图像空间领域的表现。一个紧凑的SegFormer-B0学生模型，采用半监督的教师-学生框架，使用OneFormer Swin-L伪标签进行训练，在每帧11.7毫秒的时间内达到了手工标注的IoU为0.946，相比基线检查点（IoU 0.758，18.9毫秒）有所提升。在对32帧手工标注图像的受控规划比较中，图像空间中点规划以2.2毫秒的时间实现了最低的横向中心误差（14.3像素），比BEV距离变换规划（926.8毫秒，65.0像素中心误差）快421倍，同时保持了可比的掩膜路径对齐（98.5%对比98.6%）。在六个校园序列（22,679帧）上的全视频重放确认，改进的分割将时间不稳定性从1.46%降低到0.33%，并将模板路径的可用性从73.7%提高到79.3%。我们进一步展示了仅使用BEV进行路径提取在单目设置中是脆弱的：在一次分析运行中，99.3%的帧未产生有效的BEV路径。最终推荐的架构为图像空间中点主路径、图像空间距离变换后备，BEV用于可视化，在CPU上每帧运行完整的感知到路径堆栈的时间少于50毫秒，适用于嵌入式行人速度的微型出行系统。
View on arXiv Download PDF AI Translation
cs.CV / 12 / 2608.25251

What Do Medical Vision-Language Models Learn in Radiology? Transfer, Alignment, and Source-Proxy Leakage Under Distribution Shift

医学视觉-语言模型在放射学中学习了什么？在分布变化下的迁移、对齐和源-代理泄漏
Bouaziz, Ayoub Louaye, Chebouba, Lokmane, Himeur, Yassine
Abstract
Medical vision-language models (VLMs) can appear reliable in-domain while failing when acquisition domain, paired supervision, or evaluation protocol changes. We study this failure mode as a representation-level blind spot relevant to epistemic intelligence, without claiming a formal estimator of epistemic uncertainty. Using NIH ChestXray14 and CheXpert, we first isolate source-only cross-dataset visual transfer from unsupervised domain-adaptation diagnostics. Using PadChest and OpenI, we then evaluate multimodal alignment under strict pair-index retrieval and quantify metadata-derived source-proxy information retained in frozen embeddings. Self-supervised visual initialization improves NIH-to-CheXpert transfer over supervised ImageNet initialization in matched ResNet-18 comparisons, whereas adversarial adaptation is useful only in a narrow regime and becomes unstable as adversarial pressure increases. Multimodal exact-pair retrieval remains low under external OpenI stress testing, and source-proxy information remains recoverable from learned representations. Qualitative nearest-neighbor and Grad-CAM analyses show clinically plausible cross-dataset structure and thoracic attention patterns in many cases, while device-heavy and false-positive cases remain ambiguous. Auxiliary architecture checks are task-dependent and do not support a universal backbone ranking. Overall, the study shows that apparent competence under a single protocol can conceal transfer, alignment, and shortcut-related failure modes, motivating stress-tested evaluation of medical VLMs under distribution shift.
Chinese Translation
医学视觉-语言模型（VLMs）在特定领域内看似可靠，但在获取领域、配对监督或评估协议发生变化时可能会失败。我们研究这种失败模式，作为与认知智能相关的表示层盲点，而不声称有正式的认知不确定性估计器。使用NIH ChestXray14和CheXpert，我们首先将仅源的跨数据集视觉迁移与无监督领域适应诊断分离。接着，使用PadChest和OpenI，我们在严格的配对索引检索下评估多模态对齐，并量化在冻结嵌入中保留的源-代理信息，这些信息源自元数据。自监督视觉初始化在匹配的ResNet-18比较中改善了NIH到CheXpert的迁移，相比于监督的ImageNet初始化，而对抗适应仅在狭窄的范围内有效，并且随着对抗压力的增加而变得不稳定。在外部OpenI压力测试下，多模态精确配对检索的表现仍然较低，且源-代理信息仍可从学习到的表示中恢复。定性最近邻和Grad-CAM分析显示，在许多情况下，跨数据集的结构和胸部关注模式在临床上是合理的，而设备重型和假阳性案例仍然模糊。辅助架构检查依赖于任务，并不支持通用的骨干网络排名。总体而言，本研究表明，在单一协议下的明显能力可能掩盖迁移、对齐和捷径相关的失败模式，促使在分布变化下对医学VLMs进行压力测试评估。
View on arXiv Download PDF AI Translation
cs.CV / 13 / 2608.25274

OpenCVL: An Open, Diverse, and Large-Scale Dataset for Fine-Grained Cross-View Localization

OpenCVL：一个开放、多样化的大规模细粒度跨视角定位数据集
Xia, Zimin, Zaffar, Mubariz, Fu, Junsheng, Alahi, Alexandre, Kooij, Julian F. P.
Abstract
Fine-grained Cross-View Localization (CVL) estimates the precise position and orientation of a ground-level image by aligning it with geo-referenced aerial imagery, offering a scalable alternative to Global Navigation Satellite Systems (GNSS) in challenging urban environments. Existing datasets rely on data collected with high-end sensor suites, which inherently limit image diversity and scalability. While in-the-wild images are abundant, their noisy geo-tags make them unsuitable for reliable evaluation. To bridge this gap, we introduce OpenCVL, a large-scale, diverse, and open dataset containing 617,388 ground-aerial image pairs spanning 41 cities across four European countries. All images are sourced from permissive platforms, ensuring long-term accessibility and supporting open and reproducible research. The training set combines images captured with high-end sensors with diverse in-the-wild imagery. We further develop a data curation framework that filters and corrects pose annotations to construct reliable in-the-wild evaluation data. In addition, OpenCVL includes dedicated cross-area and snowy test sets to assess generalization and robustness. Experiments with a state-of-the-art CVL model on OpenCVL show that incorporating noisy in-the-wild data consistently improves performance on clean test sets, suggesting a promising direction for scaling CVL with diverse real-world imagery.
Chinese Translation
细粒度跨视角定位（CVL）通过将地面图像与地理参考的航空图像对齐，估计其精确的位置和方向，为在复杂城市环境中提供了一种可扩展的替代全球导航卫星系统（GNSS）的方法。现有的数据集依赖于高端传感器套件收集的数据，这在本质上限制了图像的多样性和可扩展性。尽管野外图像丰富，但其噪声地理标签使其不适合进行可靠评估。为了解决这一问题，我们推出了OpenCVL，这是一个大规模、多样化且开放的数据集，包含617,388对地面-航空图像，覆盖四个欧洲国家的41个城市。所有图像均来自于开放平台，确保长期可访问性，并支持开放和可重复的研究。训练集结合了高端传感器捕获的图像与多样的野外图像。我们进一步开发了一个数据整理框架，过滤和校正姿态注释，以构建可靠的野外评估数据。此外，OpenCVL还包括专门的跨区域和雪地测试集，以评估模型的泛化能力和鲁棒性。在OpenCVL上使用最先进的CVL模型进行的实验表明，结合噪声的野外数据始终能提高在干净测试集上的性能，暗示了利用多样化的真实世界图像扩展CVL的有希望方向。
View on arXiv Download PDF AI Translation
cs.CV / 14 / 2608.25299

PointRL: Learning Point-Level Vision-Language Grounding from Verifiable Annotation Evidence

PointRL：从可验证注释证据中学习点级视觉-语言定位
Su, Jingyang, Cao, Pu, Jin, Xiuze, Zhang, Longyue, Song, Qing, Yang, Lu
Abstract
Vision-language models (VLMs) increasingly rely on point coordinates as a compact and executable interface for visual grounding in GUI interaction, robotic manipulation, and interactive visual systems. However, learning reliable pointing behavior remains difficult because the supervision space is inherently non-unique: many coordinates may be valid within the same target region, while multi-instance instructions require target coverage, count consistency, and duplicate suppression. This work presents PointRL, a verifiable reinforcement learning framework that learns point-level grounding from existing heterogeneous annotation evidence. PointRL converts bounding boxes, masks, and instance labels into pointing instructions, while retaining their target supports, instance membership, and set constraints as hidden verifier evidence, i.e., annotations kept outside the prompt and used by a deterministic checker to score predictions. The proposed reward evaluates parseability, point validity, instance coverage, cardinality consistency, and redundant or missing predictions. On PointArena, PointRL improves the overall accuracy of Qwen3.5-4B from 56.11% to 65.58%. Further evaluations on RoboSpatial, BLINK, and Ref-Adv show same-backbone gains on the evaluated external benchmarks, suggesting that verifiable point-level feedback may benefit spatial grounding in these settings.
Chinese Translation
视觉-语言模型（VLMs）越来越依赖于点坐标作为在图形用户界面交互、机器人操作和交互式视觉系统中进行视觉定位的紧凑且可执行的接口。然而，学习可靠的指向行为仍然困难，因为监督空间本质上是非唯一的：在同一目标区域内，许多坐标可能是有效的，而多实例指令则要求目标覆盖、计数一致性和重复抑制。本研究提出了PointRL，一种可验证的强化学习框架，从现有的异构注释证据中学习点级定位。PointRL将边界框、掩码和实例标签转换为指向指令，同时保留它们的目标支持、实例归属和集合约束作为隐藏的验证证据，即保留在提示之外的注释，并由确定性检查器用于评分预测。所提出的奖励评估可解析性、点的有效性、实例覆盖、基数一致性以及冗余或缺失的预测。在PointArena上，PointRL将Qwen3.5-4B的整体准确率从56.11%提高到65.58%。在RoboSpatial、BLINK和Ref-Adv上的进一步评估显示，在评估的外部基准上具有相同骨干网络的增益，表明可验证的点级反馈可能有助于这些设置中的空间定位。
View on arXiv Download PDF AI Translation
cs.CV / 15 / 2608.25302

WAVE: Reversing the Guidance Hierarchy for Coarse-to-Fine Guided Depth Super-Resolution

WAVE：逆转指导层次结构的粗到细引导深度超分辨率
Nasir, Tayyab, Liu, Daochang, Mian, Ajmal
Abstract
Guided depth super-resolution (GDSR) typically extracts RGB guidance features through convolutional hierarchies, inheriting their fine-to-coarse bias. Thus, low-level spatial cues surface in early layers, leaving the deeper layers to suppress those that do not correspond to true depth boundaries, which risks artifacts and blurred edges. The same fine-to-coarse bias persists in semantics-based methods that consume low-level tokens early and global tokens late. We present WAVE, which introduces a multi-level discrete wavelet transform (ML-DWT) as an explicit and interpretable feature-control mechanism, enabling a coarse-to-fine reconstruction by consuming sub-bands and semantic tokens in reverse of their generation order. WAVE further exploits these sub-bands to treat high- and low-frequency content separately, filtering at its source the misleading RGB color and texture cues that often lead to blurred boundaries and artifacts, offering an intuitive alternative to the suppression learned implicitly by an opaque network. WAVE separates structure and detail reconstruction into dedicated modules that: i) model interactions within and across wavelet sub-bands, depth features, and semantic priors, ii) apply semantic gating to the high-frequency bands, and iii) fuse modalities through an invertible coupling mechanism that prevents collapse onto a single modality. Extensive experiments across multiple benchmarks demonstrate that WAVE matches or outperforms existing methods, with the largest gains at high upsampling factors, where low-resolution depth contains the least structure.
Chinese Translation
引导深度超分辨率（GDSR）通常通过卷积层次提取RGB引导特征，继承了其由细到粗的偏向。因此，低级空间线索在早期层中显现，而深层则抑制那些与真实深度边界不对应的线索，这可能导致伪影和模糊边缘。相同的由细到粗的偏向在基于语义的方法中也存在，这些方法早期消耗低级标记而晚期消耗全局标记。我们提出了WAVE，它引入了一种多级离散小波变换（ML-DWT）作为一种显式且可解释的特征控制机制，通过反向消耗子带和语义标记实现粗到细的重建。WAVE进一步利用这些子带分别处理高频和低频内容，在源头过滤那些常常导致模糊边界和伪影的误导性RGB颜色和纹理线索，提供了一种直观的替代方案，取代了不透明网络隐式学习的抑制。WAVE将结构和细节重建分离为专门的模块：i）建模小波子带、深度特征和语义先验之间的交互，ii）对高频带应用语义门控，iii）通过可逆耦合机制融合模态，防止崩溃到单一模态。针对多个基准的广泛实验表明，WAVE与现有方法相匹配或超越，尤其在高上采样因子下，低分辨率深度包含的结构最少时，效果提升最大。
View on arXiv Download PDF AI Translation
cs.CV / 16 / 2608.25305

MulVec: Fine-Grained Role-Aware Matching for Training-Free Zero-Shot Composed Image Retrieval

MulVec：无训练零-shot复合图像检索的细粒度角色感知匹配
Zhang, Zihao, Wu, Dayan, Liu, Xinze, Zhu, Hengjie, Zhu, Yiliang, Wang, Ding, Fu, Peng, Lin, Zheng, Wang, Weiping
Abstract
Training-free zero-shot composed image retrieval finds a target image in a gallery from a reference image and a text edit without learning from task-specific image triplets. Existing methods typically describe the target as a whole and match this description with a global image representation. This global matching can mix different semantic cues and lose fine- grained details. We propose MULVEC, a role-aware method whose compiler produces a structured query record that is mapped to four retrieval roles: Global describes the full target, Desired states what should appear, Preserve states what should remain, and Forbidden states what should disappear. Frozen encoders map the query to one target description vector and role-specific probe vectors, while each candidate is represented by one global visual vector and a bank of local visual vectors. The retrieval roles then use this shared evidence for their respective purposes, and a fixed weighted sum of their scores ranks the entire gallery in a single retrieval pass. Across CIRCO, CIRR, and FashionIQ and three backbone scales, MULVEC improves CIRCO mAP@5 by up to 23.0% over the strongest compared method and gives the best CIRR and FashionIQ results in our comparison.
Chinese Translation
无训练的零-shot复合图像检索通过参考图像和文本编辑在图库中找到目标图像，而无需从特定任务的图像三元组中学习。现有方法通常将目标描述为一个整体，并将此描述与全局图像表示进行匹配。这种全局匹配可能会混合不同的语义线索并丢失细粒度细节。我们提出了MULVEC，这是一种角色感知的方法，其编译器生成一个结构化查询记录，该记录映射到四个检索角色：Global描述完整目标，Desired说明应出现的内容，Preserve说明应保留的内容，Forbidden说明应消失的内容。冻结的编码器将查询映射到一个目标描述向量和角色特定的探测向量，而每个候选项则由一个全局视觉向量和一组局部视觉向量表示。检索角色随后利用这些共享证据来实现各自的目的，并通过固定加权和对其分数进行排名，从而在一次检索中对整个图库进行排序。在CIRCO、CIRR和FashionIQ以及三个主干网络规模上，MULVEC在最强比较方法的基础上将CIRCO mAP@5提高了最多23.0%，并在我们的比较中提供了最佳的CIRR和FashionIQ结果。
View on arXiv Download PDF AI Translation
cs.CV / 17 / 2608.25308

V-Link: Recovering Lost Visual Representations in Action DiT for Vision-Language-Action Models

V-Link：在视觉-语言-动作模型中恢复丢失的视觉表征
Lu, Yehao, Yang, Jiarui, Su, Yuning, Xie, Yufeng, Zhong, Yu, Zhang, Yazhou, Lan, Haiyu, Lu, Kaixiang, Lin, Peiwen, Wang, Chuang, Qin, Zequn, Li, Enyu, Li, Xi
Abstract
Vision-language-action (VLA) models provide a scalable path toward generalist robotic manipulation by integrating visual perception, language understanding, and continuous action control. However, we reveal a critical limitation of VLA architectures: the action expert has limited access to the 3D geometric and 2D semantic information available in VLM features. This accessibility gap weakens perceptual grounding and limits performance on fine-grained robotic manipulation. To address this issue, we propose V-Link, which explicitly recovers visual representations during the vision-language (VL) to action (A) feature transfer. Specifically, V-Link learns complementary Spatial and Semantic Query representations within the VLM and injects them into Action DiT through asymmetric pathways. Semantic Queries complement the original VLM image tokens, whereas Spatial Queries provide dedicated geometric conditioning for spatially grounded action generation. Across LIBERO, LIBERO-Plus, and RoboTwin 2.0, our V-Link improves the average success rate over base model GR00T N1.6 by +1.9%, +31.2%, and +18.8%, respectively. On the AGIBOT A3 Ultra, V-Link further achieves gains of +20% and +24% on two real-world humanoid tasks.
Chinese Translation
视觉-语言-动作（VLA）模型通过整合视觉感知、语言理解和连续动作控制，为通用机器人操作提供了一条可扩展的路径。然而，我们揭示了VLA架构的一个关键限制：动作专家对VLM特征中可用的3D几何和2D语义信息的访问有限。这一可访问性差距削弱了感知基础，并限制了精细化机器人操作的性能。为了解决这个问题，我们提出了V-Link，它在视觉-语言（VL）到动作（A）特征转移过程中显式地恢复视觉表征。具体而言，V-Link在VLM中学习互补的空间和语义查询表征，并通过不对称路径将其注入到Action DiT中。语义查询补充了原始VLM图像标记，而空间查询则为空间基础的动作生成提供专门的几何条件。在LIBERO、LIBERO-Plus和RoboTwin 2.0上，我们的V-Link相较于基础模型GR00T N1.6，平均成功率分别提高了+1.9%、+31.2%和+18.8%。在AGIBOT A3 Ultra上，V-Link在两个真实世界的人形任务中进一步实现了+20%和+24%的增益。
View on arXiv Download PDF AI Translation
cs.CV / 18 / 2608.25332

Not All Attention Heads Contribute to Critical Visual Token Selection: Head-Aware Pruning Matters More

并非所有注意力头都对关键视觉标记选择有贡献：头感知剪枝更为重要
Ma, Chaofang, Jiang, Lin, Li, Carol Jingyi, Liu, Xingyu, Li, Zeyu, Xu, Jiang, Zhang, Wei
Abstract
Vision-Language Models (VLMs) have exhibited impressive performance across diverse visual scenarios. However, this success comes at the cost of explosive growth in visual tokens, which imposes substantial memory and computational overhead during inference, ultimately increasing latency. To improve VLM inference efficiency, a typical class of visual token pruning methods estimates token importance by aggregating attention scores across all heads in the pruning layer of the Large Language Model (LLM) backbone and prunes tokens based on aggregated scores. However, in this paper, we reveal a compelling phenomenon: the capability to pinpoint critical visual tokens is concentrated within a small fraction of heads. Aggregation exclusively on these heads can improve task performance. Inspired by this observation, we propose ProViP, a training-free progressive visual token pruning framework. ProViP first removes redundant visual tokens based on the embedding similarity of input tokens before reasoning of the LLM backbone, and then further prunes tokens during reasoning via head-aware pruning. Experiments demonstrate that ProViP delivers outstanding task performance and inference efficiency. For instance, when applied to LLaVA-1.5-7B, ProViP retains 95.9% of the original performance and achieves 1.62x inference speedup under an 88.9% pruning ratio.
Chinese Translation
视觉-语言模型（VLMs）在多种视觉场景中展现了令人印象深刻的性能。然而，这一成功伴随着视觉标记的爆炸性增长，这在推理过程中带来了巨大的内存和计算开销，最终导致延迟增加。为了提高VLM的推理效率，典型的视觉标记剪枝方法通过聚合大型语言模型（LLM）主干中剪枝层所有头的注意力分数来估计标记的重要性，并基于聚合分数进行标记剪枝。然而，在本文中，我们揭示了一个引人注目的现象：识别关键视觉标记的能力集中在少数几个头中。仅在这些头上进行聚合可以提高任务性能。受此观察的启发，我们提出了ProViP，一个无训练的渐进式视觉标记剪枝框架。ProViP首先基于输入标记的嵌入相似性在LLM主干推理之前去除冗余的视觉标记，然后在推理过程中通过头感知剪枝进一步剪除标记。实验表明，ProViP在任务性能和推理效率上表现出色。例如，当应用于LLaVA-1.5-7B时，ProViP保留了95.9%的原始性能，并在88.9%的剪枝比率下实现了1.62倍的推理加速。
View on arXiv Download PDF AI Translation
cs.CV / 19 / 2608.25334

GraftSR: Grafting Authentic Textures for Real-World Image Super-Resolution via Identical-Instance Guidance

GraftSR：通过相同实例引导为真实世界图像超分辨率嫁接真实纹理
Yu, Qifan, Bai, Haoran, He, Zongyao, He, Weijie, Deng, Sibin, Qi, Honggang, Chen, Ying
Abstract
Diffusion-based real-world image super-resolution (SR) achieves impressive perceptual quality but inherently suffers from severe texture hallucination. To overcome this limitation, we propose GraftSR, a texture-reference-guided generative SR framework that leverages reference images of the identical instance to anchor the restoration of authentic textures. However, severe spatial misalignment between low-quality inputs and their references poses significant challenges, often leading to ambiguous transfer targets and background feature leakage. To address these issues, GraftSR employs a novel dual-mask reference guidance mechanism that systematically decouples the cross-view texture injection process. By explicitly isolating what authentic textures to extract from the reference and precisely localizing where to apply them within the target, GraftSR achieves robust texture transfer without relying on brittle spatial alignment. Furthermore, to bridge the critical gap in appropriate training data, we construct TexRefSR-141K, the first large-scale dataset providing high-quality reference tuples equipped with complementary spatial masks. Extensive experiments on our newly established benchmark, TexRefSR-Eval, demonstrate that GraftSR sets a new state-of-the-art. Notably, it reduces LPIPS by 20.2\% over top-performing baselines, achieving superior reference-faithful restoration.
Chinese Translation
基于扩散的真实世界图像超分辨率（SR）在感知质量上取得了令人印象深刻的成果，但本质上存在严重的纹理幻觉问题。为了解决这一限制，我们提出了GraftSR，一种纹理参考引导的生成SR框架，利用相同实例的参考图像来锚定真实纹理的恢复。然而，低质量输入与其参考图像之间的严重空间错位带来了重大挑战，常常导致模糊的转移目标和背景特征泄漏。为了解决这些问题，GraftSR采用了一种新颖的双掩码参考引导机制，系统性地解耦了跨视图纹理注入过程。通过明确隔离从参考中提取的真实纹理以及精确定位在目标中应用这些纹理的位置，GraftSR实现了稳健的纹理转移，而不依赖于脆弱的空间对齐。此外，为了填补适当训练数据的关键缺口，我们构建了TexRefSR-141K，这是第一个提供高质量参考元组并配备互补空间掩码的大规模数据集。在我们新建立的基准TexRefSR-Eval上的大量实验表明，GraftSR设定了新的最先进水平。值得注意的是，它在顶级基线的基础上将LPIPS降低了20.2 ext{%}，实现了优越的参考忠实恢复。
View on arXiv Download PDF AI Translation
cs.CV / 20 / 2608.25344

CoRE: Weakly Supervised Coarse-to-Fine Risk Evidence Learning in Driving Videos

CoRE：驾驶视频中弱监督的粗到细风险证据学习
Hamid, Kaiser, Cui, Can, Liang, Nade
Abstract
Perceived risk in driving evolves over time and may be supported by specific scene entities, yet supervision is typically limited to coarse video-level judgments. Learning \emph{when} supporting evidence emerges and \emph{which entities} support a risk predictor would ordinarily require costly temporal- and entity-level annotations. We introduce \textbf{CoRE}, a weakly supervised coarse-to-fine framework that learns fine-grained prediction support from coarse video supervision. CoRE first trains a video-level predictor and then freezes it. Structured interventions over candidate temporal regions or entity tracks measure how each candidate changes the coarse prediction, producing graded prediction-effect targets. These targets are distilled into a student that directly predicts temporal and entity support from the original video, without requiring interventions at inference. We evaluate this learning principle across three complementary settings: RISEE tests perceived-risk support from subjective clip-level judgments without temporal or entity-level risk annotations; DoTA provides independent temporal event annotations for evaluating weakly supervised traffic-anomaly localization; and UCF-Crime tests whether the same coarse-to-fine mechanism extends to a standard non-driving anomaly-detection benchmark. Across these settings, CoRE learns informative fine-grained support from coarse supervision, with strong temporal localization on DoTA and competitive performance on UCF-Crime. These results show that coarse video predictions can provide useful supervision for recovering the fine-grained evidence supporting them, without requiring corresponding fine-grained labels.
Chinese Translation
驾驶中的感知风险随着时间的推移而变化，并可能由特定场景实体支持，但监督通常仅限于粗略的视频级判断。学习 extit{何时}支持证据出现以及 extit{哪些实体}支持风险预测器通常需要昂贵的时间和实体级注释。我们提出了 extbf{CoRE}，一个弱监督的粗到细框架，从粗略的视频监督中学习细粒度的预测支持。CoRE首先训练一个视频级预测器，然后将其冻结。对候选时间区域或实体轨迹的结构性干预测量每个候选如何改变粗略预测，从而产生分级的预测效果目标。这些目标被提炼成一个学生模型，该模型直接从原始视频中预测时间和实体支持，而无需在推理时进行干预。我们在三个互补的设置中评估这一学习原则：RISEE测试来自主观片段级判断的感知风险支持，而不需要时间或实体级风险注释；DoTA提供独立的时间事件注释，以评估弱监督的交通异常定位；UCF-Crime测试相同的粗到细机制是否扩展到标准的非驾驶异常检测基准。在这些设置中，CoRE从粗略监督中学习到有信息的细粒度支持，在DoTA上具有强大的时间定位能力，并在UCF-Crime上表现出竞争力。这些结果表明，粗略的视频预测可以为恢复支持它们的细粒度证据提供有用的监督，而无需相应的细粒度标签。
View on arXiv Download PDF AI Translation
cs.CV / 21 / 2608.25356

Where to Look Matters: On-Policy Self-Distillation for Long-Video Understanding

关注的地方很重要：用于长视频理解的在线自蒸馏
Wang, Kaishen, Zhao, Dongdi, Liang, Yijun, Ye, Dingqiang, Chen, Ruibo, Huang, Heng, Fu, Di
Abstract
Vision-language models (VLMs) have made substantial progress in long-video understanding, with standard backbone models typically answering questions from frames sampled across the full video. However, as videos become longer, the full-video context inevitably contains more question-irrelevant temporal content, which can distract the model from the evidence needed to answer a specific question. We empirically find that focusing the visual input on short annotated clue intervals containing question-relevant evidence consistently improves prediction accuracy across model scales compared with using the corresponding full videos, while requiring fewer input frames. Based on this finding, we introduce Clue-OPSD, a clue-privileged on-policy self-distillation framework for long-video understanding. During training, a full-video student learns from a self-teacher conditioned on the corresponding clue interval by aligning their next-token distributions along student-generated trajectories. Clue-OPSD thus uses clue intervals as privileged supervision without relying on ground-truth answer labels, while requiring no clue annotations or additional modules at inference time. Extensive experiments across multiple long-video understanding benchmarks and Qwen3.5 model scales demonstrate consistent improvements over the corresponding backbone models and strong performance against supervised post-training baselines.
Chinese Translation
视觉语言模型（VLMs）在长视频理解方面取得了显著进展，标准的主干模型通常从整个视频中采样的帧回答问题。然而，随着视频长度的增加，整个视频的上下文不可避免地包含更多与问题无关的时间内容，这可能会分散模型对回答特定问题所需证据的注意力。我们通过实验证明，将视觉输入集中在包含与问题相关证据的短注释线索区间上，相比使用相应的完整视频，能够在不同模型规模上持续提高预测准确性，同时需要更少的输入帧。基于这一发现，我们提出了Clue-OPSD，一种用于长视频理解的线索特权在线自蒸馏框架。在训练过程中，完整视频的学生模型通过对齐其生成轨迹上的下一个标记分布，从相应的线索区间中学习自我教师。因此，Clue-OPSD将线索区间作为特权监督，而不依赖于真实答案标签，同时在推理时不需要线索注释或额外模块。在多个长视频理解基准和Qwen3.5模型规模上的广泛实验表明，与相应的主干模型相比，Clue-OPSD持续改善了性能，并在监督后训练基线中表现出色。
View on arXiv Download PDF AI Translation
cs.CV / 22 / 2608.25360

FlashNormal: Detailed Surface Normal Estimation from Flash and No-Flash Images

FlashNormal：基于闪光和非闪光图像的详细表面法线估计
Chen, Ruiyang, Li, Feiran, Guo, Heng, Ma, Zhanyu
Abstract
High-quality surface normal estimation is preferred for detailed surface shape recovery and image editing. Existing single image-based methods, though being a practical setup, often struggle to recover fine surface details and are sensitive to inherent shape-reflectance ambiguity. While photometric stereo achieves high-fidelity surface normal estimation from images under varying lights, its applicability is strictly limited by requiring a multi-illumination capture setup. To this end, we propose FlashNormal, a diffusion-based surface normal estimator from flash/no-flash image pairs. While retaining high practicability on modern smartphones, our proposal takes advantage of flash-induced shading variations, and leverages curvature-guided detail enhancement strategy, improving surface detail recovery and mitigating shape-reflectance ambiguity effectively. To evaluate our proposed method, we further present EvalFlash, the first real-world flash/no-flash evaluation dataset containing 20 objects aligned with ground-truth surface normals for quantitative benchmarking. Extensive experiments demonstrate the effectiveness of FlashNormal over state-of-the-art single image-based methods and show a significant out-performance over flash/no-flash-based normal estimation method on EvalFlash.
Chinese Translation
高质量的表面法线估计对于详细的表面形状恢复和图像编辑至关重要。现有的基于单幅图像的方法虽然在实际应用中可行，但往往难以恢复细致的表面细节，并且对固有的形状-反射模糊敏感。虽然光度立体技术能够从不同光照下的图像中实现高保真的表面法线估计，但其适用性受到严格限制，因为需要多光源捕捉设置。为此，我们提出了FlashNormal，一种基于闪光/非闪光图像对的扩散型表面法线估计器。我们的方案在现代智能手机上保持了高实用性，利用闪光引起的阴影变化，并采用曲率引导的细节增强策略，有效改善了表面细节恢复并减轻了形状-反射模糊。为了评估我们提出的方法，我们进一步推出了EvalFlash，这是第一个包含20个与真实表面法线对齐的物体的真实世界闪光/非闪光评估数据集，用于定量基准测试。大量实验表明，FlashNormal在性能上优于最先进的基于单幅图像的方法，并在EvalFlash上显著超越了基于闪光/非闪光的法线估计方法。
View on arXiv Download PDF AI Translation
cs.CV / 23 / 2608.25367

RSFusionDet: Underwater RGB-Sonar Multimodal Object Detection

RSFusionDet：水下RGB-声纳多模态目标检测
Liu, Zhuoyan, Wang, Yihan, Wang, Bo, Wang, Bing, Li, Ye
Abstract
Underwater unimodal object detection faces many challenges in sensor imaging, such as optical images limited by underwater noise and visible distance, and sonar images limited by less object structural information. While, optical images have rich object structural information, and sonar images are less affected by underwater noise and have a longer visible distance. Optical (RGB modality) and sonar (Sonar modality) images have complementary information underwater. In this paper, we create an RGB-Sonar multimodal object detection dataset, \textbf{R}GB-\textbf{S}onar \textbf{Fusion} (RSFusion) and propose evaluation metrics for the benchmark. And we propose the \textbf{R}GB-\textbf{S}onar \textbf{Fusion} \textbf{Det}ector (RSFusionDet) with a new RGB-Sonar multimodal object detection result expression for RGB-Sonar multimodal object detection. We analyze the features of RGB and Sonar modal information, and design a Cross-Attention Fusion (CAFusion) module to fuse RGB-Sonar spatial misalignment features and Object Matching Head (OMHead) with Loss (OMLoss) to match identical objects in RGB-Sonar modalities. Our RSFusionDet achieves 76.4/48.6 AP (RGB/Sonar) for object detection and 83.4 \(\text{F1-Score}_{match}\) for object matching, on RSFusion, which outperforms other object detection models. Compared with the DINO baseline, our method improves by 0.7/1.4 AP (RGB/Sonar) while simultaneously providing reliable cross-modal object matching. The code and datasets are publicly available at https://github.com/LEFTeyex/RSFusionDet.
Chinese Translation
水下单模态目标检测面临许多传感器成像的挑战，例如光学图像受到水下噪声和可见距离的限制，以及声纳图像缺乏物体结构信息。光学图像具有丰富的物体结构信息，而声纳图像则不易受到水下噪声的影响，并且具有更长的可见距离。光学（RGB模态）和声纳（Sonar模态）图像在水下具有互补的信息。本文创建了一个RGB-声纳多模态目标检测数据集， extbf{R}GB- extbf{S}onar extbf{Fusion}（RSFusion），并提出了基准评估指标。同时，我们提出了 extbf{R}GB- extbf{S}onar extbf{Fusion} extbf{Det}ector（RSFusionDet），为RGB-声纳多模态目标检测提供了一种新的结果表达方式。我们分析了RGB和声纳模态信息的特征，并设计了一个交叉注意力融合（Cross-Attention Fusion，CAFusion）模块，以融合RGB-声纳空间错位特征和对象匹配头（Object Matching Head，OMHead）与损失（Loss，OMLoss），以匹配RGB-声纳模态中的相同物体。我们的RSFusionDet在RSFusion上实现了76.4/48.6的AP（RGB/声纳）用于目标检测，并在目标匹配中获得了83.4的 ext{F1-Score}_{match}，超越了其他目标检测模型。与DINO基线相比，我们的方法在AP（RGB/声纳）上分别提高了0.7/1.4，同时提供了可靠的跨模态目标匹配。代码和数据集可在https://github.com/LEFTeyex/RSFusionDet公开获取。
View on arXiv Download PDF AI Translation
cs.CV / 24 / 2608.25371

Capacity Overflow: A Blind Spot for Backdoor Attacks in Vision MoE

容量溢出：视觉 MoE 中后门攻击的盲点
Zou, Xiaocheng, Zheng, Tiancheng, Xu, Xiaolin, Ding, Ruyi
Abstract
Mixture-of-Experts (MoE) has become a prevalent paradigm for scaling Vision Transformers efficiently. To ensure computational scalability and prevent expert overload, Vision MoE architectures employ a capacity-bounded token dispatch mechanism, where each expert's processing budget depends on the inference batch size. This work identifies this batch-dependent behavior as an overlooked attack surface, and proposes a stealthy supply-chain backdoor attack that exploits this property through a three-phase framework. First, we inject a backdoor into an early MoE layer. Second, we train a neutralizer in a deeper MoE layer that suppresses the backdoor under normal capacity. Third, we configure a batch-adaptive capacity factor that preserves high capacity for small batches while reducing it for large batches, naturally disabling the neutralizer via token overflow at deployment-scale batch sizes. The attack remains in dormant mode during small-batch security audits and enters activation mode during large-batch deployment. Experiments on V-MoE and Swin-MoE across ImageNet-100 and GTSRB demonstrate activation-mode attack success rates of 76-87% with dormant-mode ASR below 9%, while evading Neural Cleanse, STRIP, Fine-Pruning, and Activation Clustering. Our findings reveal a fundamental security risk arising from batch-dependent execution in scalable Vision MoE architectures.
Chinese Translation
混合专家（Mixture-of-Experts, MoE）已成为有效扩展视觉变换器（Vision Transformers）的普遍范式。为了确保计算的可扩展性并防止专家过载，视觉 MoE 架构采用了容量受限的令牌调度机制，其中每个专家的处理预算取决于推理批量大小。本文识别出这种依赖批量的行为作为一个被忽视的攻击面，并提出了一种隐蔽的供应链后门攻击，通过三阶段框架利用这一特性。首先，我们在早期的 MoE 层中注入后门。其次，我们在更深的 MoE 层中训练一个中和器，以在正常容量下抑制后门。第三，我们配置一个批量自适应容量因子，该因子在小批量时保持高容量，而在大批量时降低容量，自然地通过在部署规模批量大小时的令牌溢出禁用中和器。该攻击在小批量安全审计期间保持休眠状态，并在大批量部署期间进入激活模式。在 ImageNet-100 和 GTSRB 上对 V-MoE 和 Swin-MoE 的实验表明，激活模式攻击的成功率为 76-87%，而休眠模式下的 ASR 低于 9%，同时能够规避 Neural Cleanse、STRIP、Fine-Pruning 和 Activation Clustering。我们的研究结果揭示了可扩展视觉 MoE 架构中由批量依赖执行引发的基本安全风险。
View on arXiv Download PDF AI Translation
cs.CV / 25 / 2608.25386

Efficient Training with Foresight: Multi-Token Auxiliary Supervision for Autoregressive Image Generation

具有前瞻性的高效训练：自回归图像生成的多标记辅助监督
Niu, Guo, Yao, Xiongfei, Wang, Teng, Zhu, Nannan
Abstract
Autoregressive (AR) image generation has shown strong potential for scalable high-fidelity synthesis by modeling images as discrete token sequences. However, traditional next token prediction (NTP) continues to suffer from sparse and myopic supervision, insufficiently discriminative representations, and high training cost caused by dense computation over the full token sequence. To address these issues, we propose multi-token autoregressive (MTAR), a unified training framework that improves autoregressive image generation from three aspects: prediction objectives, representation regularization, and training efficiency. Specifically, MTAR introduces multi-token prediction (MTP) to alleviate the sparsity and myopia of traditional NTP by imposing joint supervision on multiple future tokens; employs token-level contrastive regularization (TCR) to explicitly enhance the separability of sampled token representations and thereby improve representation discriminability; and incorporates semantic dropping (SD) as a semantics-aware training acceleration strategy to reduce redundant computation on low-information tokens while preserving informative learning signals. All three components are applied only during training and introduce no additional overhead during autoregressive inference. On ImageNet, MTAR achieves a better balance between generation quality and training efficiency. Compared with LlamaGen, MTAR achieves up to 0.95 lower FID and 39\% faster training. Moreover, even with only 1/3 of the training iterations, it still attains performance comparable to or better than the baseline, substantially reducing training time.
Chinese Translation
自回归（AR）图像生成通过将图像建模为离散标记序列，展现出可扩展的高保真合成的强大潜力。然而，传统的下一个标记预测（NTP）仍然面临稀疏和短视的监督、表示能力不足以及由于对完整标记序列进行密集计算而导致的高训练成本等问题。为了解决这些问题，我们提出了多标记自回归（MTAR），这是一个统一的训练框架，从预测目标、表示正则化和训练效率三个方面改善自回归图像生成。具体而言，MTAR引入了多标记预测（MTP），通过对多个未来标记施加联合监督，缓解传统NTP的稀疏性和短视性；采用标记级对比正则化（TCR），明确增强采样标记表示的可分性，从而提高表示的辨别能力；并将语义丢弃（SD）作为一种语义感知的训练加速策略，以减少对低信息标记的冗余计算，同时保留有用的学习信号。这三个组成部分仅在训练期间应用，并且在自回归推理过程中不会引入额外的开销。在ImageNet上，MTAR在生成质量和训练效率之间实现了更好的平衡。与LlamaGen相比，MTAR的FID降低了最多0.95，训练速度提高了39%。此外，即使仅用1/3的训练迭代次数，它的性能仍然可与基线相媲美或更优，显著减少了训练时间。
View on arXiv Download PDF AI Translation
cs.CV / 26 / 2608.25401

PIVOT: A Multi-Trajectory Dataset and Testbed for Pose, Intrinsics, and Novel Viewpoint Evaluation in Real-World 3D Reconstruction

PIVOT：用于真实世界三维重建中姿态、内参和新视角评估的多轨迹数据集和测试平台
Raymond, Mary
Abstract
Neural radiance fields (NeRFs), 3D Gaussian Splatting (3DGS), and related novel-view synthesis methods are commonly evaluated under capture and reconstruction conditions cleaner than those encountered by robots, drones, and autonomous systems. Benchmarks often rely on reconstruction-friendly trajectories, optimized camera poses and intrinsics, and held-out views sampled from trajectories represented during training. These assumptions can obscure performance with measured poses, reusable camera calibration, and structurally different camera paths. We introduce PIVOT (Pose, Intrinsics and Viewpoint Oriented Testbed), a multi-trajectory dataset, processing pipeline, and evaluation framework for independently studying these factors. PIVOT captures each scene using diverse camera trajectories and retains, where available, both sensor-derived measured poses and COLMAP-optimized poses, together with calibrated and optimized camera intrinsics. It defines three benchmark families: (1) seen versus unseen trajectory novel-view generalization, (2) measured versus optimized pose sensitivity, and (3) calibrated versus optimized intrinsics sensitivity. We also introduce a directed pose-space Chamfer distance to quantify how well training poses cover an evaluation trajectory. PIVOT v1 contains five real-world scenes captured with a DJI Mini 4 Pro and provides an open processing and Nerfstudio-based evaluation toolchain. Benchmark results show a consistent quality gap between held-out views on represented trajectories and unseen trajectories, as well as substantial sensitivity to pose source and camera intrinsics.
Chinese Translation
神经辐射场（NeRFs）、三维高斯点云（3DGS）及相关的新视角合成方法通常在比机器人、无人机和自主系统所遇到的捕获和重建条件更为理想的环境下进行评估。基准测试往往依赖于适合重建的轨迹、优化的相机姿态和内参，以及从训练期间表示的轨迹中抽样的保留视角。这些假设可能会掩盖在测量姿态、可重用相机校准和结构上不同的相机路径下的性能。我们引入了PIVOT（姿态、内参和视角导向测试平台），这是一个多轨迹数据集、处理管道和评估框架，用于独立研究这些因素。PIVOT使用多样的相机轨迹捕获每个场景，并保留可用的传感器测得姿态和COLMAP优化姿态，以及经过校准和优化的相机内参。它定义了三个基准测试系列：（1）已见与未见轨迹的新视角泛化，（2）测量与优化姿态的敏感性，以及（3）校准与优化内参的敏感性。我们还引入了一种定向姿态空间Chamfer距离，以量化训练姿态在多大程度上覆盖评估轨迹。PIVOT v1包含五个使用DJI Mini 4 Pro捕获的真实场景，并提供了一个开放的处理和基于Nerfstudio的评估工具链。基准测试结果显示，在表示轨迹的保留视角和未见轨迹之间存在一致的质量差距，以及对姿态来源和相机内参的显著敏感性。
View on arXiv Download PDF AI Translation
cs.CV / 27 / 2608.25412

AdaptiveEmbed: Sample-Adaptive Multi-Vector Representation for Multimodal Retrieval

自适应嵌入：用于多模态检索的样本自适应多向量表示
Liu, Xinze, Yang, Lei, Wu, Dayan, Zhu, Hengjie, Zhang, Zihao, Wu, Hanqi, Hu, Tianzhu, Fu, Peng, Lin, Zheng, Wang, Weiping
Abstract
Multi-vector representations have emerged as an effective paradigm for multimodal retrieval, representing each sample with multiple complementary embeddings to capture fine-grained cross-modal information. However, existing approaches typically employ a fixed representation capacity, assigning the same number of vectors to all samples regardless of their individual retrieval demands. Such a fixed-capacity formulation overlooks the fact that different samples may require different amounts of representation capacity for effective retrieval. In this work, we introduce \emph{Sample-Adaptive Multi-Vector Representation} (SAMVR), a new problem setting for multimodal retrieval that studies how multi-vector representation capacity can be allocated at the sample level. Under SAMVR, each sample is represented by a \emph{content-adaptive embedding set} (CAES), whose capacity is determined according to the sample-specific retrieval utility of additional representation vectors. To instantiate SAMVR, we propose \emph{AdaptiveEmbed}, a unified framework for learning sample-adaptive multi-vector representations. AdaptiveEmbed learns structured multi-vector representations through \emph{Multi-Group Contrastive Learning} (MGCL) with the symmetric \emph{set-to-set similarity} (SetSim), and further employs \emph{Utility Policy Optimization} (UPO) to determine sample-specific representation capacity via \emph{Marginal Utility Allocation} (MUA). Experiments across multimodal retrieval benchmarks involving image, text, video, and audio show that sample-adaptive capacity allocation achieves overall better retrieval performance than fixed-capacity multi-vector representations, validating the effectiveness of SAMVR for multimodal retrieval. These results establish SAMVR as a viable formulation for adaptive capacity allocation in multi-vector multimodal retrieval.
Chinese Translation
多向量表示已成为多模态检索的有效范式，通过多个互补的嵌入表示每个样本，以捕捉细粒度的跨模态信息。然而，现有方法通常采用固定的表示容量，为所有样本分配相同数量的向量，而不考虑它们各自的检索需求。这种固定容量的公式忽视了不同样本可能需要不同数量的表示容量以实现有效检索这一事实。在本研究中，我们引入了样本自适应多向量表示（Sample-Adaptive Multi-Vector Representation，SAMVR），这是一个新的多模态检索问题设置，研究如何在样本级别分配多向量表示容量。在SAMVR下，每个样本由一个内容自适应嵌入集（Content-Adaptive Embedding Set，CAES）表示，其容量根据样本特定的额外表示向量的检索效用来确定。为了实现SAMVR，我们提出了自适应嵌入（AdaptiveEmbed），这是一个用于学习样本自适应多向量表示的统一框架。自适应嵌入通过多组对比学习（Multi-Group Contrastive Learning，MGCL）与对称的集合间相似性（set-to-set similarity，SetSim）学习结构化的多向量表示，并进一步采用效用策略优化（Utility Policy Optimization，UPO）通过边际效用分配（Marginal Utility Allocation，MUA）来确定样本特定的表示容量。在涉及图像、文本、视频和音频的多模态检索基准测试中的实验表明，样本自适应容量分配在整体上实现了比固定容量多向量表示更好的检索性能，验证了SAMVR在多模态检索中的有效性。这些结果确立了SAMVR作为多向量多模态检索中自适应容量分配的可行公式。
View on arXiv Download PDF AI Translation
cs.CV / 28 / 2608.25418

Bootstrapping a 4D LiDAR Annotation Tool from Video Foundation Models

从视频基础模型引导4D LiDAR标注工具的构建
Kim, Jihun, Jang, Hyun-Kurl, Yang, Hyemin, Yang, Jinnyeong, Kweon, Hyeokjun, Yoon, Kuk-Jin
Abstract
Progress in 4D LiDAR segmentation is bottlenecked by data. Assigning temporally consistent labels across sparse point cloud sequences is costly and hard to scale, and every new task or domain tends to demand fresh dense annotation. This motivates a simple question of whether high-quality LiDAR training data can be produced automatically, without any human labeling. To this end, we introduce LiDAR-SAM2, a framework that turns a 2D video foundation model, SAM2, into a scalable source of supervision for the 4D LiDAR domain. On the data side, it automatically generates temporally coherent LiDAR-level labels from SAM2 video masks through multi-view projection and spatio-temporal aggregation. On the modeling side, a tailored modality interface and a two-stage learning objective adapt SAM2's video segmentation kernel to spatio-temporal LiDAR structure, so that a single click per object yields a consistent mask track across the sequence. Trained with no human LiDAR annotation, LiDAR-SAM2 produces semantic and panoptic labels on SemanticKITTI that approach the quality of full human annotation from only a few points, and models trained on these labels approach the performance of full ground-truth supervision. This positions LiDAR-SAM2 as a scalable labeling tool that substantially reduces the annotation burden for 3D and 4D scene understanding.
Chinese Translation
4D LiDAR分割的进展受到数据的瓶颈。为稀疏点云序列分配时间一致的标签既昂贵又难以扩展，每个新任务或领域往往需要新的密集标注。这引发了一个简单的问题：是否可以自动生成高质量的LiDAR训练数据，而无需任何人工标注。为此，我们介绍了LiDAR-SAM2，一个将2D视频基础模型SAM2转化为4D LiDAR领域可扩展监督源的框架。在数据方面，它通过多视图投影和时空聚合自动生成来自SAM2视频掩码的时间一致的LiDAR级标签。在建模方面，定制的模态接口和两阶段学习目标使SAM2的视频分割内核适应时空LiDAR结构，从而使每个对象只需一次点击即可在序列中生成一致的掩码轨迹。LiDAR-SAM2在没有人工LiDAR标注的情况下进行训练，能够在SemanticKITTI上生成接近全人工标注质量的语义和全景标签，仅需少量点，并且在这些标签上训练的模型接近全真值监督的性能。这使得LiDAR-SAM2成为一个可扩展的标注工具，显著减少了3D和4D场景理解的标注负担。
View on arXiv Download PDF AI Translation
cs.CV / 29 / 2608.25435

Saliency-Depth Conditioning for Zero-Shot Segmentation of Communication-Tower Components in Cluttered UAV Imagery

针对杂乱无人机图像中通信塔组件的零-shot分割的显著性-深度条件化
Lesani, Ali, Yeum, Chul Min, Kang, Su-Min
Abstract
Fine-grained segmentation of communication-tower components in UAV imagery is essential for automated inspection, yet task-specific models are hard to develop due to limited instance-level annotations. Zero-shot segmentation models offer a promising alternative, but in cluttered scenes, visually similar background structures interfere with component localization, causing missed instances and false positives. We propose a model-agnostic saliency-depth foreground-conditioning strategy combining appearance-based saliency with monocular relative depth to construct a coarse tower prior and suppress irrelevant content. We integrate this module with Grounded-SAM and SAM 3, yielding SD-Grounded-SAM and SD-SAM 3. SD-Grounded-SAM further applies geometric and depth-aware box refinement before mask generation, while SD-SAM 3 relies on SAM 3's internal setup. On TOW-300, a dataset of 340 communication-tower UAV images, our strategy improves both baselines: SD-SAM 3 achieves the strongest instance-segmentation performance, while SD-Grounded-SAM produces fewer false positives. Ablations confirm complementary gains from saliency, depth, and box refinement, improving robustness in cluttered scenes.
Chinese Translation
在无人机图像中对通信塔组件进行细粒度分割对于自动化检查至关重要，但由于实例级注释有限，特定任务模型的开发变得困难。零-shot分割模型提供了一种有前景的替代方案，但在杂乱场景中，视觉上相似的背景结构干扰了组件定位，导致漏检和误报。我们提出了一种模型无关的显著性-深度前景条件化策略，将基于外观的显著性与单目相对深度相结合，以构建粗略的塔先验并抑制无关内容。我们将该模块与Grounded-SAM和SAM 3集成，得到SD-Grounded-SAM和SD-SAM 3。SD-Grounded-SAM在生成掩膜之前进一步应用几何和深度感知的框选精细化，而SD-SAM 3依赖于SAM 3的内部设置。在TOW-300数据集（包含340幅通信塔无人机图像）上，我们的策略改善了两个基线：SD-SAM 3实现了最强的实例分割性能，而SD-Grounded-SAM则产生了更少的误报。消融实验确认了显著性、深度和框选精细化的互补增益，提高了在杂乱场景中的鲁棒性。
View on arXiv Download PDF AI Translation
cs.CV / 30 / 2608.25452

VGA-BenchV2: An Expanded Unified Benchmark and Multi-Model Framework for Evaluating Video Aesthetics and Generation Quality

VGA-BenchV2：一个扩展的统一基准和多模型框架，用于评估视频美学和生成质量
Jiang, Longteng, Zheng, DanDan, Qiao, Qianqian, Huang, Heng, Wang, Huaye, Bo, Yihang, Peng, Bao, Chen, Jingdong, Zhou, Jun, Jin, Xin
Abstract
We introduce VGA-BenchV2, an extended human-aligned benchmark and optimization framework for jointly evaluating and improving video generation quality and aesthetic value. Built upon VGA-Bench, VGA-BenchV2 preserves the original fine-grained taxonomy with two primary dimensions-Aesthetic and Generation-and 52 sub-dimensions. Guided by this taxonomy, we curate 1,016 diverse prompts and collect over 60,000 videos generated by 12 mainstream video generation models. More importantly, VGA-BenchV2 substantially expands human-labeled supervision by adding 36,000 task-level annotations, including 16,200 for aesthetic quality, 13,200 for aesthetic tagging, and 6,600 for generation quality, corresponding to 13.46x, 11.15x, and 1.55x scale-ups over VGA-Bench, respectively. Leveraging this enlarged annotation corpus, we develop a hybrid evaluator architecture consisting of VAQA-Net for continuous aesthetic scoring and two Qwen-based Large Vision-Language Model evaluators, VTag-Net and VGQA-Net, for aesthetic tagging and generation quality assessment. Extensive experiments demonstrate strong alignment with human judgments across diverse generation models. Beyond evaluation, VGA-BenchV2 further introduces an evaluation-to-optimization pipeline, where the learned aesthetic evaluator serves as a reward model for reinforcement learning-based generator fine-tuning. This closes the loop from benchmark construction and human supervision to automated evaluation and model optimization, enabling video generators to improve not only in realism but also in aesthetic quality and human preference alignment. Resources are available at https://huggingface.co/datasets/BestiVictoryLab/VGA-Bench.
Chinese Translation
我们介绍了VGA-BenchV2，这是一个扩展的人类对齐基准和优化框架，用于共同评估和提升视频生成质量及其美学价值。VGA-BenchV2基于VGA-Bench构建，保留了原有的细粒度分类法，主要包括两个维度——美学和生成——以及52个子维度。在这一分类法的指导下，我们策划了1,016个多样化的提示，并收集了由12个主流视频生成模型生成的超过60,000个视频。更重要的是，VGA-BenchV2通过增加36,000个任务级注释显著扩展了人类标注的监督，包括16,200个美学质量注释、13,200个美学标记注释和6,600个生成质量注释，分别对应于对VGA-Bench的13.46倍、11.15倍和1.55倍的扩展。利用这一扩展的注释语料库，我们开发了一个混合评估器架构，其中包括用于连续美学评分的VAQA-Net和两个基于Qwen的大型视觉-语言模型评估器VTag-Net和VGQA-Net，用于美学标记和生成质量评估。大量实验表明，在不同生成模型中与人类判断具有强一致性。除了评估，VGA-BenchV2还进一步引入了评估到优化的管道，其中学习到的美学评估器作为基于强化学习的生成器微调的奖励模型。这闭合了从基准构建和人类监督到自动评估和模型优化的循环，使视频生成器不仅在现实性上有所提升，同时在美学质量和人类偏好对齐方面也得以改善。相关资源可在https://huggingface.co/datasets/BestiVictoryLab/VGA-Bench获取。
View on arXiv Download PDF AI Translation
cs.CV / 31 / 2608.25465

Automatic weld seam segmentation for industrial quality control: a comparison of RGB and polarimetric imaging with CNN and transformer architectures

工业质量控制中的自动焊缝分割：RGB与偏振成像结合CNN和变换器架构的比较
Garbin, Simone, Venturoso, Leonardo, Todescato, Marco
Abstract
Visual inspection of welded assemblies remains one of the least automated stages in many industrial production processes, still depending largely on the experience of human operators and thus subject to inter-operator variability; the manufacturing of special-purpose machinery cabins, the setting of this study, is one representative case. This work evaluates the feasibility of automatic weld seam segmentation from RGB and polarimetric imagery, comparing controlled laboratory acquisitions with images captured under real, uncontrolled conditions. Convolutional neural network (CNN) architectures and transformer-based architectures are benchmarked under a unified, threshold-independent protocol, training each CNN with three random seeds to separate genuine effects from seed noise. In controlled RGB conditions, CNN models reach a mean mask mAP50 of up to 0.87, but drop to 0.22-0.48 under uncontrolled acquisition, showing that the acquisition setup is a first-order component of the inspection system. Polarimetric imaging with alignment-preserving geometric augmentation localizes previously unseen welds with a mean mask mAP50 up to 0.93: on par with, rather than ahead of, the best controlled-RGB result, but reaching that accuracy on uncontrolled RGB without requiring acquisition control. The clearest architectural finding concerns viewpoint robustness. In-distribution, transformers and CNNs are broadly comparable; but under a test-time viewpoint shift, the transformer models, and RF-DETR in particular, retain high accuracy while every CNN collapses. The gap holds across three seeds and a resolution-matched control, pointing to architecture rather than training resolution. Within the CNN family, capacity brings no reliable in-distribution gain once seed variance is accounted for: small CNNs suffice for fixed viewpoints, transformers for variable ones.
Chinese Translation
焊接组件的视觉检查仍然是许多工业生产过程中自动化程度最低的环节，仍然在很大程度上依赖于人工操作员的经验，因此受到操作员间变异性的影响；本研究的背景是特殊用途机械舱的制造，这是一个具有代表性的案例。本研究评估了从RGB和偏振图像中自动进行焊缝分割的可行性，比较了在受控实验室条件下获得的图像与在真实、非受控条件下捕获的图像。卷积神经网络（CNN）架构和基于变换器的架构在统一的、与阈值无关的协议下进行基准测试，为每个CNN使用三个随机种子进行训练，以区分真实效果与种子噪声。在受控的RGB条件下，CNN模型的平均掩膜mAP50可达到0.87，但在非受控采集下降至0.22-0.48，显示出采集设置是检测系统的首要组成部分。采用保持对齐的几何增强的偏振成像能够定位以前未见过的焊缝，平均掩膜mAP50可达到0.93：与最佳受控RGB结果相当，而不是领先，但在非受控RGB条件下达到该准确度而无需采集控制。最明显的架构发现涉及视角鲁棒性。在分布内，变换器和CNN在广泛上是可比的；但在测试时视角变化下，变换器模型，尤其是RF-DETR，保持高准确度，而每个CNN则崩溃。这个差距在三个种子和一个分辨率匹配的控制下保持一致，指向架构而非训练分辨率。在CNN家族中，一旦考虑到种子方差，容量并未带来可靠的分布内增益：小型CNN足以应对固定视角，变换器则适用于可变视角。
View on arXiv Download PDF AI Translation
cs.CV / 32 / 2608.25472

PAGS: Autofocusing Photoacoustic Tomography via Speed-of-Sound-Adaptive Gaussian Splatting

PAGS：通过声速自适应高斯点云实现自聚焦光声断层成像
Ge, Jiarui, Ma, Jintao, Fan, Bangxu, Zhang, Jinyan, Yang, Xiaokang, Na, Shuai, Yuan, Xiaoyun
Abstract
Photoacoustic computed tomography (PACT) combines optical absorption contrast with acoustic detection for high-resolution deep-tissue imaging. A persistent challenge is that unknown speed-of-sound (SoS) heterogeneity changes acoustic time-of-flight, causing defocusing artifacts when reconstruction assumes a uniform SoS. Existing SoS-adaptive methods either rely on calibrated acoustic priors or optimize dense physical medium models, which becomes expensive and difficult to scale in 3D. We propose PAGS, a differentiable framework for blind autofocusing PACT via speed-of-sound-adaptive Gaussian splatting. PAGS represents the initial pressure field with sparse Gaussian photoacoustic (PA) sources and replaces explicit medium recovery with a compact anisotropic path-averaged SoS (ASoS) field parameterized by spherical harmonic probes. This latent propagation field directly controls source-to-transducer arrival-time alignment, while an analytic Gaussian acoustic projection maps the source representation to transducer signals efficiently. The resulting closed-loop signal-domain optimization jointly updates the Gaussian PA source parameters and the ASoS field from measured data, without calibrated SoS priors. Experiments on simulated and physical phantom data demonstrate improved reconstruction sharpness under heterogeneous acoustic media, robustness to sparse-view sampling, and computational benefits from the analytic Gaussian projection.
Chinese Translation
光声计算断层成像（PACT）结合了光学吸收对比和声学检测，用于高分辨率深组织成像。一个持续的挑战是未知的声速（SoS）异质性会改变声波的飞行时间，当重建假设均匀的声速时，会导致失焦伪影。现有的声速自适应方法要么依赖于经过校准的声学先验，要么优化密集的物理介质模型，这在三维中变得昂贵且难以扩展。我们提出了PAGS，这是一个通过声速自适应高斯点云实现盲自聚焦PACT的可微分框架。PAGS用稀疏的高斯光声（PA）源表示初始压力场，并用一个由球谐探头参数化的紧凑各向异性路径平均声速（ASoS）场替代显式介质恢复。这个潜在传播场直接控制源到换能器的到达时间对齐，而解析高斯声学投影则高效地将源表示映射到换能器信号。最终的闭环信号域优化共同更新高斯PA源参数和从测量数据中获得的ASoS场，而无需经过校准的声速先验。在模拟和物理幻影数据上的实验表明，在异质声学介质下重建锐度有所改善，对稀疏视图采样具有鲁棒性，并且从解析高斯投影中获得了计算上的优势。
View on arXiv Download PDF AI Translation
cs.CV / 33 / 2608.25479

4DStreamCtrl: Interactive Video Generation with Online 4D Control

4DStreamCtrl：基于在线4D控制的交互式视频生成
Li, Shiqian, Lin, Chenguo, Liu, Zhiguang, Tang, Yu, Ou, Jiarong, Chen, Rui, Zhu, Yixin
Abstract
Generative video models now synthesize footage nearly indistinguishable from reality. Their promise as interactive tools hinges on fine-grained control of how objects and the camera move over time, yet each existing approach captures only part of this: camera-parameter methods steer the viewpoint but cannot move objects, 2D-trajectory methods act in the image plane and ignore depth and occlusion, and recent 3D methods add geometry but run only offline at a fixed length. In particular, none combines 3D-consistent control of both camera and objects with real-time, streaming generation. Here we show that camera motion, object trajectories, and depth can be unified into a single 3D point-track representation, from which one model performs joint camera and object control, depth editing, and motion transfer in a single forward pass. To learn this interface at scale, we mine in-the-wild video for 3D motion supervision, yielding OpenVidHD-Motion3D, and encode it with a lightweight Geometric Motion Head that plugs into a pretrained video diffusion model. Because this encoder is temporally separable, we distill the model into a causal streaming student that generates arbitrarily long video in four denoising steps at memory independent of length. This unified design surpasses prior camera-only, 2D, and offline-3D methods in motion-control precision while covering modalities they address only in isolation. 4DStreamCtrl runs at 20 FPS on a single high-end GPU for 480p video and stays temporally coherent over hundreds of frames, enabling, to our knowledge, interactive 4D-controllable streaming generation for the first time. More broadly, grounding generation in explicit 3D geometry with efficient causal inference points toward interactive world models with closed-loop spatiotemporal control, from controllable simulators to real-time visual imagination for embodied agents.
Chinese Translation
生成式视频模型现已能够合成几乎与现实无异的影像。其作为交互式工具的潜力依赖于对物体和摄像机随时间运动的细粒度控制，然而现有方法各自仅捕捉了部分内容：摄像机参数方法能够控制视角但无法移动物体，二维轨迹方法仅在图像平面内作用，忽略了深度和遮挡，最近的三维方法虽然引入了几何信息，但仅能离线处理固定长度的视频。特别地，目前尚无方法能将摄像机和物体的三维一致性控制与实时流式生成相结合。本文展示了如何将摄像机运动、物体轨迹和深度统一为单一的三维点轨迹表示，基于此表示，单一模型即可在一次前向传播中实现摄像机与物体的联合控制、深度编辑及运动迁移。为大规模学习该接口，我们从野外视频中挖掘三维运动监督数据，构建了OpenVidHD-Motion3D数据集，并设计了轻量级的几何运动头（Geometric Motion Head）嵌入预训练视频扩散模型。由于该编码器具备时间可分离性，我们将模型蒸馏为因果流式学生模型，能够在与视频长度无关的内存消耗下，通过四步去噪生成任意长度的视频。该统一设计在运动控制精度上超越了以往仅控制摄像机、二维轨迹及离线三维方法，同时涵盖了它们各自独立处理的模态。4DStreamCtrl在单块高端GPU上以20帧每秒速度生成480p视频，并在数百帧内保持时间一致性，实现了迄今为止首次的交互式4D可控流式生成。更广泛地，将生成过程基于显式三维几何并结合高效因果推理，指向了具备闭环时空控制的交互式世界模型，从可控模拟器到具身智能体的实时视觉想象。
View on arXiv Download PDF AI Translation
cs.CV / 34 / 2608.25480

DeCO: Discriminative Evidence Composition for Fine-Grained Dataset Distillation

DeCO：用于细粒度数据集蒸馏的区分性证据组合
Fan, Chuixuan, Li, Guang, Wang, Shijie, Zhou, Dongzhan, Sun, Baoli, Ogawa, Takahiro, Haseyama, Miki, Wang, Zhihui
Abstract
Dataset distillation compresses a large training set into a compact synthetic set while preserving its downstream utility. However, existing methods primarily preserve global image statistics and may overlook the localized evidence essential for fine-grained visual classification (FGVC), such as object parts, subtle textures, and region-specific structures. We formulate fine-grained dataset distillation as budgeted discriminative-evidence preservation and propose Discriminative Evidence Composition (DeCO). DeCO uses attention rollout from a pretrained TransFG teacher to identify informative patches, applies spatial diversification to reduce redundant coverage, and organizes the resulting regions into class-wise evidence banks. Multiple same-class regions are then packed into compact grid-composed images. The teacher is used only for dataset construction, whereas downstream students are trained with standard hard-label supervision without teacher logits. Experiments on CUB-200-2011, FGVC-Aircraft, and Stanford Cars show that DeCO consistently outperforms representative coreset and dataset-distillation baselines under different IPC budgets.
Chinese Translation
数据集蒸馏将大型训练集压缩为紧凑的合成集，同时保留其下游效用。然而，现有方法主要保留全局图像统计信息，可能忽视了细粒度视觉分类（FGVC）所需的局部证据，如物体部件、细微纹理和区域特定结构。我们将细粒度数据集蒸馏形式化为预算化的区分性证据保留，并提出了区分性证据组合（DeCO）。DeCO利用预训练的TransFG教师模型的注意力展开来识别信息丰富的图像块，应用空间多样化以减少冗余覆盖，并将生成的区域组织成按类别划分的证据库。然后，将多个同类区域打包成紧凑的网格组合图像。教师模型仅用于数据集构建，而下游学生模型则在没有教师模型输出的情况下，通过标准的硬标签监督进行训练。在CUB-200-2011、FGVC-Aircraft和斯坦福汽车数据集上的实验表明，DeCO在不同的IPC预算下始终优于代表性的核心集和数据集蒸馏基线。
View on arXiv Download PDF AI Translation
cs.CV / 35 / 2608.25483

Gaussian Splatting Underwater: A Controlled Cross-Regime Study

水下高斯喷溅：一个受控的跨领域研究
Álvarez-Tuñón, Olaya, Graßhof, Stella
Abstract
The underwater environment is challenging for 3D reconstruction, because particles suspended in the water scatter and diffuse light, turbidity varies, absorption depends on wavelength, and illumination is rarely uniform. Methods based on Gaussian splatting have generally been developed for conditions that allow good image quality, and have primarily been tested on relatively shallow water. This paper examines how well Gaussian splatting performs across publicly available underwater datasets representing different degrees of turbidity, loss of illumination, and colour attenuation, together with an industrial survey. Five systems with public code are run under one protocol, with shared poses, initialisation, budget, and evaluator, to establish their relative advantages, disadvantages, and limitations. What these methods can do turns out to depend more on the setup than on the architecture. Water clarity binds upstream of rendering, since structure-from-motion registers 99.5 \% of frames in clear water and 0.0 \% at 12 NTU. Illumination geometry decides whether a medium model helps at all: under an artificial light that moves with the camera, medium-blind splatting beats both medium-aware systems. On the survey the benchmark's photometric leader comes last, beaten on geometry by a restoration pre-pass in front of vanilla 3DGS---and none of it is visible in the scores the field reports. Scene builds, per-run configurations, and evaluation code are released at https://github.com/olayasturias/uw3dgs
Chinese Translation
水下环境对三维重建具有挑战性，因为水中悬浮的颗粒会散射和扩散光线，浑浊度变化，吸收依赖于波长，且照明通常不均匀。基于高斯喷溅的方法通常是在允许良好图像质量的条件下开发的，并主要在相对浅水中进行测试。本文考察了高斯喷溅在不同浑浊度、光照损失和颜色衰减的公开水下数据集中的表现，以及一项工业调查。五个具有公开代码的系统在同一协议下运行，采用共享的姿态、初始化、预算和评估器，以确定它们的相对优缺点和局限性。这些方法的效果更多地依赖于设置而非架构。水的清晰度在渲染之前起着关键作用，因为运动重建在清水中注册了99.5%的帧，而在12 NTU时为0.0%。照明几何决定了介质模型是否有帮助：在与相机一起移动的人工光源下，介质盲喷溅优于两种介质感知系统。在调查中，基准的光度领导者排名最后，在几何方面被一个在普通3DGS前的恢复预处理超越——而这些在该领域报告的分数中并不可见。场景构建、每次运行的配置和评估代码已发布在 https://github.com/olayasturias/uw3dgs
View on arXiv Download PDF AI Translation
cs.CV / 36 / 2608.25485

Semi-Supervised Adaptation of Vision-Language Models for Image Classification

视觉-语言模型在图像分类中的半监督适应
Mekhalfi, Mohamed L., Rahhal, Mohamad M. Al, Bazi, Yakoub, Khenfer, Salah E., Shi, Mingdeng, Zou, Hua, Zuair, Mansour
Abstract
Vision-language models like CLIP have shown sig- nificant potential in handling natural images, yet their perfor- mance is often limited by the distinct characteristics of satellite imagery. While parameter-efficient adaptation techniques exist, their efficacy is frequently limited by the scarcity of annotated samples. In this letter, we propose Self-Evolutionary CLIP (SE- CLIP), a semi-supervised framework designed for recursive label mining in scene classification. The approach follows a dual-phase pipeline, where an initial warm-up on a few annotated seeds is followed by a recursive discovery phase that iteratively identifies high-confidence samples from unlabeled pools. To maintain the integrity of the evolving support set, we employ a class-balanced selection strategy that prevents the model from being dominated by easily learned categories. Results on the UCM and NWPU benchmarks indicate that SE-CLIP significantly outperforms existing semi-supervised approaches. The framework provides a viable solution for adapting VLMs to the remote sensing domain with minimal human intervention.
Chinese Translation
视觉-语言模型如 CLIP 在处理自然图像方面展现了显著的潜力，但其性能常常受到卫星图像独特特征的限制。虽然存在参数高效的适应技术，但其有效性常常受到标注样本稀缺的限制。在本文中，我们提出了自我进化 CLIP（Self-Evolutionary CLIP，SE-CLIP），这是一个旨在场景分类中进行递归标签挖掘的半监督框架。该方法遵循双阶段流程，首先在少量标注种子上进行初步热身，然后进入递归发现阶段，迭代识别来自未标记样本池的高置信度样本。为了保持不断演变的支持集的完整性，我们采用了类别平衡选择策略，以防止模型被易于学习的类别主导。在 UCM 和 NWPU 基准测试中的结果表明，SE-CLIP 显著优于现有的半监督方法。该框架为将视觉语言模型适应于遥感领域提供了一种可行的解决方案，且对人类干预的需求最小。
View on arXiv Download PDF AI Translation
cs.CV / 37 / 2608.25493

SMART: MLLM-guided Temporal Alignment for Unifying Sign Language Recognition and Spotting

SMART：基于MLLM的时间对齐框架用于统一手语识别与定位
Choi, Eunjee, Sung, JungHoon, Cho, Seongwhan, Xin, Chu, Choi, Younggeun
Abstract
Continuous sign language recognition (CSLR) aims to recognize gloss sequences from unsegmented sign videos under weak sequence-level supervision. However, existing methods rely on sentence-level gloss annotations, providing limited temporal and semantic guidance for fine-grained representation learning. Conventional video-text alignment also requires large batch sizes, making it inefficient for memory-intensive sign language video training. In this work, we propose SMART, an MLLM-guided temporal alignment framework for joint sign recognition and spotting. SMART uses MLLMgenerated motion descriptions as auxiliary semantic cues and performs stable videotext alignment under small-batch training. To improve temporal representation learning, we introduce a Multi-Scale Temporal Adapter that models temporal interactions during transformer encoding. For dense temporal localization, SMART incorporates CSFormer, a CSLR-guided spotting module that injects recognition-derived gloss evidence into a boundary-aware spotting network. This unified framework enables CSLR features to benefit spotting, while spotting supervision complements weak CTC-based recognition. Experiments on four sign language benchmarks, including PHOENIX14-T, CSL-Daily, Large-scale KSL, and Disaster and Safety KSL datasets, demonstrate the effectiveness of SMART across both recognition and spotting tasks.
Chinese Translation
连续手语识别（CSLR）旨在在弱序列级监督下，从未分割的手语视频中识别手语词序列。然而，现有方法依赖于句子级的手语词注释，提供的时间和语义指导有限，难以进行细粒度的表征学习。传统的视频-文本对齐也需要较大的批量大小，这使得在内存密集型的手语视频训练中效率低下。在本研究中，我们提出了SMART，一个基于MLLM的时间对齐框架，用于联合手语识别与定位。SMART使用MLLM生成的运动描述作为辅助语义线索，并在小批量训练下进行稳定的视频-文本对齐。为了改善时间表征学习，我们引入了一个多尺度时间适配器，该适配器在变换器编码过程中建模时间交互。为了实现密集的时间定位，SMART结合了CSFormer，一个基于CSLR的定位模块，将识别衍生的手语词证据注入到一个边界感知的定位网络中。这个统一框架使得CSLR特征能够惠及定位，同时定位监督补充了弱CTC基础的识别。对包括PHOENIX14-T、CSL-Daily、大规模KSL和灾难与安全KSL数据集在内的四个手语基准的实验表明，SMART在识别和定位任务中均表现出有效性。
View on arXiv Download PDF AI Translation
cs.CV / 38 / 2608.25495

Pose-Anchored Optical Flow for Low-Latency Human Action Anticipation in Human-Robot Teaming

基于姿态的光流用于低延迟人类动作预测在人机协作中的应用
Grundy, Lewis de Zoete, McCarthy, Chris, Fluke, Christopher
Abstract
Human-robot interaction (HRI) requires robots to interpret human actions early in their execution in order to respond safely, efficiently, and naturally. However, many existing approaches to human action recognition rely either on sparse skeletal representations, which lack fine-grained motion cues, or dense optical flow, which can be computationally expensive for low-latency perception pipelines. In this paper, we propose PoseOFF, a pose-anchored optical flow representation that captures local motion information around human joints to support earlier human intent understanding. By conditioning motion feature extraction on human pose, PoseOFF encodes localised motion dynamics at semantically meaningful body locations, forming a structured motion representation that is explicitly aligned with human kinematics. We evaluate PoseOFF across multiple benchmark datasets and backbone architectures for action anticipation, demonstrating consistent improvements in recognition accuracy, particularly at early observation ratios. Our results show that PoseOFF enables models to achieve comparable or improved performance while observing less of the action sequence, highlighting its effectiveness for early prediction. Importantly, these gains are achieved without requiring full-frame motion processing, making the approach practical for real-time and resource-constrained settings. These findings suggest that pose-centred motion representations such as PoseOFF can enhance the ability of interactive robot systems to infer human actions earlier, supporting more responsive and anticipatory behaviour in human-robot interaction scenarios.
Chinese Translation
人机交互（HRI）要求机器人在执行人类动作的早期阶段进行解读，以便安全、高效和自然地做出响应。然而，许多现有的人类动作识别方法依赖于稀疏的骨骼表示，这缺乏细致的运动线索，或者依赖于密集的光流，这在低延迟感知管道中可能计算成本高昂。本文提出了一种名为PoseOFF的基于姿态的光流表示，捕捉人类关节周围的局部运动信息，以支持更早的人类意图理解。通过将运动特征提取与人类姿态相结合，PoseOFF在语义上有意义的身体位置编码局部运动动态，形成与人类运动学明确对齐的结构化运动表示。我们在多个基准数据集和动作预测的主干架构上评估了PoseOFF，展示了在识别准确性方面的一致性提升，特别是在早期观察比例下。我们的结果表明，PoseOFF使模型在观察较少的动作序列时能够实现可比或更好的性能，突显了其在早期预测中的有效性。重要的是，这些提升是在不需要全帧运动处理的情况下实现的，使该方法在实时和资源受限的环境中具有实用性。这些发现表明，像PoseOFF这样的基于姿态的运动表示可以增强交互式机器人系统更早推断人类动作的能力，从而支持在人机交互场景中更具响应性和预见性的行为。
View on arXiv Download PDF AI Translation
cs.CV / 39 / 2608.25515

OpenVeinNet: Robust Open-Set Finger Vein Verification with Dynamic Snake Convolution and Graph Learning

OpenVeinNet：基于动态蛇卷积和图学习的鲁棒开放集指静脉验证
Patwardhan, Sushrut, Ramachandra, Raghavendra
Abstract
Finger vein verification is a promising biometric modality for secure authentication because vascular patterns are internal, difficult to observe externally, and relatively resistant to presentation attacks. However, reliable verification remains challenging in open-set settings, where test identities are unseen during training and non-enrolled probes must be rejected at inference. This paper presents OpenVeinNet, a finger vein verification framework designed for cross-dataset and open-set evaluation. The proposed model combines Dynamic Snake Convolution with graph-based feature modelling. Dynamic Snake Convolution extracts local curvilinear and tubular vein structures using adaptive sampling, while the graph convolutional backbone models long-range topological relationships between vein regions. To improve the discriminative quality of the embedding space, we introduce a Centroid Angular Hybrid Loss, which jointly encourages intra-class compactness and inter-class angular separation for cosinesimilaritybased verification. Experiments are conducted on five public finger vein datasets: FV-300, MMCBNU, FV-USM, PolyU, and VERA. The method is evaluated using leaveonedatasetout training under both enrolmentbased unknownrejection and fullsubject verification protocols, and is compared with handcrafted and recent deep learning-based baselines. The results show that OpenVeinNet achieves strong cross-dataset generalisation, consistently low equal error rates, and competitive true accept rates at fixed false accept rate operating points. Ablation studies further confirm the individual and combined contributions of adaptive tubular feature extraction, graph-based relational modelling, and the proposed loss function. These findings indicate that explicitly modelling local vein geometry, global vascular relationships, and angularly compact embeddings is effective for openset finger vein verification.
Chinese Translation
指静脉验证是一种有前景的生物识别方式，因其血管模式为内部结构，外部难以观察，并且对呈现攻击具有相对较强的抵抗力。然而，在开放集环境中，可靠的验证仍然具有挑战性，因为测试身份在训练期间未见过，且必须在推理时拒绝未注册的探测器。本文提出了OpenVeinNet，一种旨在进行跨数据集和开放集评估的指静脉验证框架。所提出的模型结合了动态蛇卷积和基于图的特征建模。动态蛇卷积通过自适应采样提取局部曲线和管状静脉结构，而图卷积骨干网络则建模静脉区域之间的长距离拓扑关系。为了提高嵌入空间的区分质量，我们引入了一种中心角度混合损失（Centroid Angular Hybrid Loss），该损失共同促进类内紧凑性和类间角度分离，以实现基于余弦相似度的验证。我们在五个公共指静脉数据集上进行了实验：FV-300、MMCBNU、FV-USM、PolyU和VERA。该方法在基于注册的未知拒绝和全体验证协议下，采用留一数据集训练进行评估，并与手工特征和近期基于深度学习的基线进行了比较。结果表明，OpenVeinNet在跨数据集泛化方面表现出色，始终保持较低的等错误率，并在固定的假接受率操作点上具有竞争力的真实接受率。消融研究进一步确认了自适应管状特征提取、基于图的关系建模以及所提出的损失函数的单独和组合贡献。这些发现表明，明确建模局部静脉几何、全局血管关系和角度紧凑嵌入对于开放集指静脉验证是有效的。
View on arXiv Download PDF AI Translation
cs.CV / 40 / 2608.25520

Asymmetric Cross-Modal Fine-Grained Visual Categorization: ACF-Net and the BirdPro Benchmark

非对称跨模态细粒度视觉分类：ACF-Net及BirdPro基准
Deng, Bohan, Ye, Shuo, Yu, Zitong
Abstract
Audio-visual cross-modal Fine-Grained Visual Categorization (FGVC) aims to identify fine-grained categories by jointly leveraging visual and auditory information. However, FGVC under asymmetric cross-modal scenarios has received limited attention, where paired video and audio are not strictly synchronized and may not even correspond to the same individual or moment. Such weak and ambiguous cross-modal correspondence poses substantial challenges to effective representation learning and modality alignment. To address these issues, we propose ACF-Net, a novel optical flow-guided framework for asymmetric audio-visual fine-grained learning. ACF-Net consists of two key modules: Optical Flow-Guided Motion (OFGM) and Asymmetric CrossModal Adaptive Fusion (ACAF). OFGM captures motion-sensitive visual cues and suppresses irrelevant background interference, thereby enhancing discriminative dynamic representations in videos. ACAF estimates modality reliability under weakly matched audio-video pairs and performs uncertainty-aware adaptive fusion to improve category-level recognition robustness. To support research on asymmetric cross-modal FGVC, we further construct BirdPro, a new bird-oriented audio-visual benchmark, since existing datasets often lack large-scale category-level audio-video associations under non-strict temporal and instance correspondence. BirdPro contains 1,919 audio recordings and 11,965 videos covering 194 bird species. Extensive experiments show that ACF-Net achieves the best results compared with representative baseline methods, outperforming the strongest baselines by 2.97% and 1.92% in the fused and mismatched settings, respectively.
Chinese Translation
音频-视觉跨模态细粒度视觉分类（FGVC）旨在通过联合利用视觉和听觉信息来识别细粒度类别。然而，在非对称跨模态场景下，FGVC受到的关注有限，此时配对的视频和音频并不严格同步，甚至可能不对应于同一个个体或时刻。这种弱且模糊的跨模态对应关系对有效的表示学习和模态对齐提出了重大挑战。为了解决这些问题，我们提出了ACF-Net，一种新颖的光流引导框架，用于非对称音频-视觉细粒度学习。ACF-Net由两个关键模块组成：光流引导运动（Optical Flow-Guided Motion, OFGM）和非对称跨模态自适应融合（Asymmetric CrossModal Adaptive Fusion, ACAF）。OFGM捕捉运动敏感的视觉线索并抑制无关背景干扰，从而增强视频中的区分性动态表示。ACAF在弱匹配的音频-视频配对下估计模态可靠性，并执行基于不确定性的自适应融合，以提高类别级识别的鲁棒性。为了支持非对称跨模态FGVC的研究，我们进一步构建了BirdPro，这是一个新的以鸟类为导向的音频-视觉基准，因为现有数据集通常缺乏在非严格时间和实例对应下的大规模类别级音频-视频关联。BirdPro包含1,919个音频录音和11,965个视频，涵盖194种鸟类。大量实验表明，ACF-Net与代表性的基线方法相比，取得了最佳结果，在融合和不匹配设置下分别超越了最强基线2.97%和1.92%。
View on arXiv Download PDF AI Translation
cs.CV / 41 / 2608.25529

Video-IFBench: Evaluating Instruction Following of Multimodal LLMs in Video Understanding Scenarios

Video-IFBench：在视频理解场景中评估多模态大型语言模型的指令遵循能力
Liu, Hongbo, Chen, Peixian, Liu, Sihan, Zhang, Peiyuan, Zou, Kai, Zheng, Dian, Hu, Xiaoxing, Dong, Yuhao, Zhang, Mengdan, Shen, Yunhang, Cao, Haoyu, Liu, Wei, Gu, Weibo, Sun, Xing, Zhao, Shengjie
Abstract
Multimodal Large Language Models (MLLMs) have shown strong performance in video understanding. However, their ability to follow instructions in this domain remains under-explored. Real-world video understanding requires models not only to interpret video content correctly, but also to satisfy diverse user-specified constraints. Existing benchmarks focus primarily on task accuracy rather than instruction adherence, leaving this capability insufficiently evaluated. To address this gap, we introduce Video-IFBench, a comprehensive benchmark for evaluating instruction following in video understanding, where models must satisfy diverse user-specified constraints, including those grounded in visual and audio content. We develop an instruction taxonomy with four templates, including single-task, multi-task, selection, and nested instructions, covering 32 task types and 39 manually designed constraint categories spanning both semantic and format requirements. To reduce annotation cost, we build a semi-automatic data construction pipeline that combines MLLMs, programmatic processing, and human verification, resulting in 1.5K samples. We conduct a large-scale evaluation of more than 20 recent MLLMs and show that video instruction following remains challenging for current models, especially for instructions with many constraints, semantic constraints, or complex conditional structures that require selecting the correct branch or path based on video content. We hope our work will facilitate future research on instruction following in video understanding scenarios.
Chinese Translation
多模态大型语言模型（MLLMs）在视频理解方面表现出色。然而，它们在该领域遵循指令的能力仍然未得到充分探索。现实世界中的视频理解要求模型不仅能够正确解读视频内容，还要满足多样化的用户指定约束。现有基准主要关注任务准确性，而非指令遵循，导致这一能力的评估不足。为了解决这一问题，我们提出了Video-IFBench，这是一个全面的基准，用于评估视频理解中的指令遵循能力，模型必须满足多样化的用户指定约束，包括基于视觉和音频内容的约束。我们开发了一个包含四种模板的指令分类法，包括单任务、多任务、选择和嵌套指令，涵盖32种任务类型和39个手动设计的约束类别，涉及语义和格式要求。为了降低标注成本，我们建立了一个半自动数据构建管道，结合了MLLMs、程序处理和人工验证，最终生成了1500个样本。我们对20多种近期的MLLMs进行了大规模评估，结果表明，视频指令遵循对当前模型仍然具有挑战性，特别是对于具有多个约束、语义约束或复杂条件结构的指令，这些指令要求根据视频内容选择正确的分支或路径。我们希望我们的工作能够促进未来在视频理解场景中指令遵循的研究。
View on arXiv Download PDF AI Translation
cs.CV / 42 / 2608.25539

CropCop: An Auditable 120-Class Plant-Health Model from Benchmark Reconstruction to a Quantised Runtime Artifact

CropCop：一个可审计的120类植物健康模型，从基准重建到量化运行工件
Ahmed, Rana Muhammad, Abbas, Sabahat
Abstract
A plant-health score can appear precise while resting on duplicated image families, a long-tailed label space, or a runtime file that was never evaluated. We present CropCop, a closed-set recognition system spanning 120 operational plant-health classes and an evidence chain from corpus reconstruction to direct execution of the final quantised artifact. Starting from 117,546 audited images, we rejected the inherited partition after confirming 3,233 duplicate relationships across split boundaries and froze a 109,107-image benchmark with zero crossings among the audited trusted leakage groups and a 151.7 largest-to-smallest class ratio. A fully fine-tuned DINOv3 ConvNeXt-Tiny reference achieved 98.51% accuracy and 96.87% macro-F1 on the locked internal test. A compact MobileNetV4 Conv-Medium derivative achieved 98.46% accuracy and 96.27% macro-F1 without being presented as evidence for a new distillation method. Validation-only post-training quantisation selected dynamic activations with per-channel weights, and the final 22.60 MiB ExecuTorch/XNNPACK PTE achieved 98.46% accuracy and 96.23% macro-F1 when executed directly. Only six of 16,363 top-1 decisions changed between the converted INT8 graph and the PTE, while paired analysis showed a modest class-balanced loss; an exploratory post hoc fruit-label slice localized a larger recall decline than aggregate accuracy revealed. CropCop establishes strong leakage-controlled internal recognition and software-runtime fidelity; it does not establish performance on unseen farms, camera pipelines, or physical Android hardware.
Chinese Translation
植物健康评分可能看似精确，但其基础可能是重复的图像家族、长尾标签空间或从未评估的运行时文件。我们提出了CropCop，一个涵盖120个操作植物健康类别的闭集识别系统，以及从语料库重建到最终量化工件直接执行的证据链。从117,546张经过审核的图像开始，我们在确认3,233个跨分割边界的重复关系后，拒绝了继承的分区，并冻结了一个109,107张图像的基准，该基准在经过审核的可信泄漏组中没有交叉，且最大与最小类别比为151.7。经过完全微调的DINOv3 ConvNeXt-Tiny参考模型在锁定的内部测试中达到了98.51%的准确率和96.87%的宏观F1分数。一个紧凑的MobileNetV4 Conv-Medium衍生模型在未作为新蒸馏方法的证据呈现的情况下，达到了98.46%的准确率和96.27%的宏观F1分数。仅验证后的后训练量化选择了具有每通道权重的动态激活，最终的22.60 MiB ExecuTorch/XNNPACK PTE在直接执行时达到了98.46%的准确率和96.23%的宏观F1分数。在转换的INT8图和PTE之间，仅有16,363个顶级决策中的六个发生了变化，而配对分析显示出适度的类别平衡损失；一项探索性的事后水果标签切片定位到比汇总准确率揭示的更大的召回下降。CropCop建立了强大的泄漏控制内部识别和软件运行时保真性；但并未在未见过的农场、相机管道或物理Android硬件上建立性能。
View on arXiv Download PDF AI Translation
cs.CV / 43 / 2608.25559

AdaVDR: Adaptive Tool Use and Reflection for Video Deep Research

AdaVDR：用于视频深度研究的自适应工具使用与反思
Zhang, Xintong, Fan, Xiaomeng, Yan, Shilin, He, Ekko, Liu, Zicheng, Zou, Zijian, Zhang, Guannan, Wu, Yuwei, Gao, Zhi, Xue, Hongwei
Abstract
Video deep research answers complex questions by jointly understanding video content and retrieving external knowledge from the open Web. However, diverse questions and videos require different tool-use strategies, and inappropriate tool calls can produce incorrect results. Uncertain grounding and retrieval also make unnecessary interactions costly and error-prone, increasing latency and reasoning errors. To address these challenges, we propose AdaVDR, an adaptive video deep research agent with adaptive tool invocation and reflection. AdaVDR selects tools according to the task and its capabilities, and backtracks only when unreliable intermediate results require correction. To enable these capabilities, we develop a video deep research data construction pipeline. We first discover retrieval-relevant events and entities in diverse videos and acquire detailed information through grounding and external retrieval to construct high-quality QA pairs. For each QA, task-specific prompts organize the information acquisition process into a tool-use trajectory, allowing different question and video types to follow different grounding and retrieval strategies. We further introduce model-conditioned tool necessity filtering, which evaluates tool calls against the target model's video understanding and internal knowledge, removing tools or tool chains the model can bypass. This yields trajectories tailored to the target model's video understanding capability and knowledge. Using this pipeline, we construct training data and VDR-EE, a benchmark covering entity-centric and event-centric questions. We perform supervised fine-tuning followed by reinforcement learning with a redundancy-aware reward to strengthen adaptive tool invocation and reflection. Experiments show that our method performs best among the evaluated open-source models on VDR-EE and substantially improves over its base models on VideoDR.
Chinese Translation
视频深度研究通过共同理解视频内容和从开放网络中检索外部知识来回答复杂问题。然而，多样化的问题和视频需要不同的工具使用策略，不当的工具调用可能导致错误结果。不确定的基础和检索也使得不必要的交互变得代价高昂且容易出错，增加了延迟和推理错误。为了解决这些挑战，我们提出了AdaVDR，一个具有自适应工具调用和反思能力的视频深度研究代理。AdaVDR根据任务和自身能力选择工具，仅在不可靠的中间结果需要修正时进行回溯。为了实现这些能力，我们开发了一个视频深度研究数据构建管道。我们首先在多样化的视频中发现与检索相关的事件和实体，并通过基础和外部检索获取详细信息，以构建高质量的问答对。对于每个问答，特定任务的提示将信息获取过程组织成工具使用轨迹，使不同的问题和视频类型能够遵循不同的基础和检索策略。我们进一步引入模型条件的工具必要性过滤，该过滤评估工具调用与目标模型的视频理解和内部知识，移除模型可以绕过的工具或工具链。这产生了针对目标模型视频理解能力和知识量身定制的轨迹。利用该管道，我们构建了训练数据和VDR-EE，一个涵盖以实体为中心和以事件为中心问题的基准。我们进行了监督微调，随后使用关注冗余的奖励进行强化学习，以增强自适应工具调用和反思。实验表明，我们的方法在评估的开源模型中在VDR-EE上表现最佳，并在VideoDR上显著改善了其基础模型的表现。
View on arXiv Download PDF AI Translation
cs.CV / 44 / 2608.25568

CrossMambaTuning: Synergistic Spatial and Cross-Layer Adaptation for Machine Vision Compression

CrossMambaTuning：机器视觉压缩的协同空间与跨层适应
Xiong, Haobo, Liu, Shaobo, Liu, Kai, Ding, Chongyang
Abstract
To reduce deployment cost and retraining overhead, adapting pretrained learned image compression (LIC) models to downstream machine vision tasks has attracted growing attention. However, existing methods typically insert fine-tuning modules independently into frozen backbones, lacking explicit mechanisms for cross-layer coordination. To address this limitation, we propose a novel framework named CrossMambaTuning, which integrates State Space Models with cross-layer interaction mechanisms for parameter-efficient fine-tuning. Specifically, we design an efficient Mamba adapter equipped with task-specific prompts and multi-scale branching to precisely capture both local features and global dependencies. Furthermore, we introduce a Scale-Invariant Cross-Layer Adapter (SICA) utilizing a parameter-sharing strategy to fuse task information across different scales and reduce redundancy. Extensive experiments demonstrate that CrossMambaTuning achieves state-of-the-art (SOTA) performance on multiple machine vision tasks, reducing parameter overhead by 72\% compared to SOTA methods. Code is available at https://github.com/rsr1123/CrossMambaTuning.
Chinese Translation
为了降低部署成本和再训练开销，将预训练的学习图像压缩（LIC）模型适应于下游机器视觉任务引起了越来越多的关注。然而，现有方法通常将微调模块独立插入到冻结的主干网络中，缺乏跨层协调的明确机制。为了解决这一限制，我们提出了一种新颖的框架，命名为CrossMambaTuning，该框架将状态空间模型与跨层交互机制相结合，以实现参数高效的微调。具体而言，我们设计了一种高效的Mamba适配器，配备任务特定的提示和多尺度分支，以精确捕捉局部特征和全局依赖。此外，我们引入了一种尺度不变跨层适配器（SICA），利用参数共享策略在不同尺度之间融合任务信息并减少冗余。大量实验表明，CrossMambaTuning在多个机器视觉任务上实现了最先进（SOTA）的性能，与SOTA方法相比，参数开销减少了72%。代码可在https://github.com/rsr1123/CrossMambaTuning获取。
View on arXiv Download PDF AI Translation
cs.CV / 45 / 2608.25575

MLLMCLIP: Feature-Level Distillation of MLLM for Robust Vision-Language Representations

MLLMCLIP：针对稳健视觉-语言表征的特征级蒸馏
Kim, Jongsuk, Wu, Qiyu, Mao, Zhuoyuan, Wakaki, Hiromi, Kim, Junmo, Mitsufuji, Yuki
Abstract
Pretrained vision-language models such as CLIP excel at zero-shot recognition but often fail at compositionality, particularly attribute-object and relational structures. Recent studies mitigate this issue by augmenting training with synthetic hard negatives generated by a cascade of large language models and text-to-image models, which incurs substantial pipeline overhead. We instead propose MLLMCLIP, a heterogeneous distillation framework that transfers multimodal knowledge directly from a generative Multimodal Large Language Model (MLLM) teacher into a discriminative CLIP student, bypassing synthetic data entirely. To bridge the architectural mismatch between the two paradigms, we introduce an attention-based per-layer token selection and a CKA-based distillation loss. Compared to prior CLIP-enhancement methods, MLLMCLIP achieves state-of-the-art compositional accuracy while delivering consistent gains on standard zero-shot classification and image-text retrieval, showing that feature-level distillation strengthens both compositional and general vision-language representation capability.
Chinese Translation
预训练的视觉-语言模型如CLIP在零-shot识别方面表现优异，但在组合性方面常常表现不佳，尤其是在属性-对象和关系结构方面。近期研究通过使用由大型语言模型和文本到图像模型生成的合成困难负样本来增强训练，从而缓解了这一问题，但这会导致显著的管道开销。我们提出了MLLMCLIP，一种异构蒸馏框架，直接将多模态知识从生成型多模态大型语言模型（MLLM）教师转移到判别型CLIP学生，完全绕过合成数据。为了弥合这两种范式之间的架构不匹配，我们引入了一种基于注意力的逐层令牌选择和基于CKA的蒸馏损失。与之前的CLIP增强方法相比，MLLMCLIP在组合准确性上实现了最先进的水平，同时在标准的零-shot分类和图像-文本检索上提供了一致的提升，显示出特征级蒸馏增强了组合性和一般视觉-语言表征能力。
View on arXiv Download PDF AI Translation
cs.CV / 46 / 2608.25580

V-Rubrics: Visual Faithfulness via Rubric-Based Reinforcement Learning

V-评分标准：通过基于评分标准的强化学习实现视觉忠实性
Tian, Shulin, Li, Minglun, Dong, Yuhao, Ding, Hao, Yao, Jiarui, Diao, Haiwen, Yang, Jingkang, Zhu, Hongyuan, Liu, Ziwei
Abstract
Vision-language models can produce fluent answers that are insufficiently grounded in the visual evidence: a single unsupported object, chart value, or intermediate inference can undermine an otherwise plausible response. We argue that this is a credit-assignment failure in multimodal post-training. Scalar outcome rewards indicate whether an answer is acceptable, but do not identify which visual facts are grounded, which reasoning steps are valid, or which instruction constraints are missed. We introduce Visual Rubrics-Based Reinforcement Learning, which decomposes reference responses into atomic propositions and scores generated answers along Visual Faithfulness (VF), Reasoning Consistency (RC), and Instruction Following (IF). The resulting rubric items provide structured partial credit and localize rubric credit when supporting evidence spans are available. We first obtain an SFT checkpoint by fine-tuning Qwen3-VL-8B-Instruct on the public OpenMMReasoner-SFT-874K corpus, adapting OpenMMReasoner's cold-start data recipe. We construct V-Rubrics 50K, a 50,248-example training set from 17 visually grounded sources, by applying rule-based filters before deriving example difficulty from rejection-sampling scores and then annotating every example with Gemini-3-Pro under the same structured prompt and protocol. We train our model based on the same SFT checkpoint using component-wise, prefix-localized rubric credit. Experiments show that our rubricbased GRPO improves over both the shared SFT baseline and answer-only GRPO, with the largest gains on knowledge-oriented and visually grounded reasoning benchmarks. The results show rubrics as a useful reward abstraction for visual post-training.
Chinese Translation
视觉语言模型能够生成流畅的回答，但这些回答往往缺乏对视觉证据的充分支持：单个不支持的物体、图表值或中间推理都可能削弱一个原本合理的响应。我们认为这是多模态后训练中的信用分配失败。标量结果奖励指示答案是否可接受，但并未识别哪些视觉事实是有依据的，哪些推理步骤是有效的，或哪些指令约束被遗漏。我们提出了基于视觉评分标准的强化学习（Visual Rubrics-Based Reinforcement Learning），该方法将参考响应分解为原子命题，并根据视觉忠实性（Visual Faithfulness, VF）、推理一致性（Reasoning Consistency, RC）和指令遵循（Instruction Following, IF）对生成的答案进行评分。生成的评分标准项目提供了结构化的部分信用，并在支持证据的跨度可用时定位评分信用。我们首先通过在公共的 OpenMMReasoner-SFT-874K 语料库上微调 Qwen3-VL-8B-Instruct 获得一个 SFT 检查点，适应 OpenMMReasoner 的冷启动数据配方。我们构建了 V-Rubrics 50K，这是一个包含 50,248 个示例的训练集，来源于 17 个视觉基础的来源，通过应用基于规则的过滤器，在从拒绝采样分数推导示例难度之前进行处理，然后在相同的结构化提示和协议下使用 Gemini-3-Pro 对每个示例进行标注。我们基于相同的 SFT 检查点，使用组件级、前缀局部化的评分标准信用训练我们的模型。实验表明，我们的基于评分标准的 GRPO 在共享 SFT 基线和仅答案的 GRPO 上都有所改善，尤其在知识导向和视觉基础推理基准上取得了最大的提升。结果显示，评分标准作为视觉后训练的有用奖励抽象。
View on arXiv Download PDF AI Translation
cs.CV / 47 / 2608.25601

A Dual-Transformer for Multi-Camera View Recommendation

用于多摄像头视角推荐的双重变换器
Cabacas-Maso, Josep, Ventura, Carles, Benito-Altamirano, Ismael
Abstract
Multi-camera systems are foundational to modern media production, and multi-camera editing is a critical task. This involves the proper selection of the appropriate camera view at each moment. In this paper, we propose a novel Dual-Transformer architecture with Cross-Attention that heavily outperformed the current SOTA models over the TVMCE dataset (TV Shows Multicamera Editing dataset). Our model decouples these tasks: (1) a dedicated temporal encoder first processes the sequence of past frames to build a rich memory of the recent history, and (2) the candidate camera views then act as queries to this memory via a cross-attention module, allowing each candidate to independently interrogate the historical context and find the most relevant information for its own evaluation. Our approach achieved 56.60% [email protected], representing a substantial improvement over the prior best result of 37.16%. We further conducted an ablation study exploring the use of lightweight backbone architectures, where the SwinV2 backbone yielded the best performance, achieving 69.65% [email protected]. Using this best-performing configuration, we then investigated the feasibility of adapting the model to replicate the editing style of a specific human editor. To this end, we fine-tuned the model using varying proportions of the initial segment of a target video. Our results demonstrate that even with only 20% of the video used for fine-tuning, the model exhibited measurable improvements in [email protected], indicating strong potential for data-efficient personalization of editing style adapted to each individual TV show or producer.
Chinese Translation
多摄像头系统是现代媒体制作的基础，而多摄像头编辑是一项关键任务。这涉及在每个时刻正确选择合适的摄像头视角。本文提出了一种新颖的双重变换器架构，结合了交叉注意力（Cross-Attention），在TVMCE数据集（电视节目多摄像头编辑数据集）上显著超越了当前的最先进模型。我们的模型将这些任务解耦：（1）一个专用的时间编码器首先处理过去帧的序列，以建立丰富的近期历史记忆；（2）候选摄像头视角通过交叉注意力模块作为查询来访问该记忆，使每个候选视角能够独立地询问历史上下文，并找到与其自身评估最相关的信息。我们的方法在[email protected]上达到了56.60%，相比之前最佳结果37.16%有了显著提升。我们进一步进行了消融研究，探索使用轻量级主干架构，其中SwinV2主干表现最佳，达到了69.65%的[email protected]。利用这一最佳配置，我们还研究了将模型调整为复制特定人类编辑者编辑风格的可行性。为此，我们使用目标视频初始片段的不同比例对模型进行了微调。我们的结果表明，即使仅使用20%的视频进行微调，模型在[email protected]上也表现出可测量的改善，显示出在数据高效个性化编辑风格方面的强大潜力，适应于每个特定的电视节目或制作人。
View on arXiv Download PDF AI Translation
cs.CV / 48 / 2608.25608

When Should a Network Emit Geometry, and When Should It Detect It? Readout, Reconciliation, and Representation in Floorplan Vectorization

网络何时应发出几何信息，何时应检测几何信息？楼层平面图矢量化中的读出、调和与表示
Zhang, He
Abstract
A network trained to recover the walls, openings, and rooms of a rasterized floorplan can produce its output in two ways: by emitting the geometry as an autoregressive coordinate sequence, or by detecting it on dense junction and centerline heatmaps and assembling a graph. We compare the two readouts on the same trained network. On real scans (CubiCasa5K) detection is better on every wall measure (+2.7 wall F1 at tolerance 0.05, +5.1 at 0.015; paired bootstrap intervals exclude zero), and reading an opening heatmap the decoder never used raises opening F1 by 2.6x without retraining. Within real scans the readout's advantage grows with plan size and reverses on small plans; on clean vector renders sequence decoding is better by 5 to 8 points where its training covered the render style, while under full domain shift the readout, given calibrated thresholds, stays ahead; neither ink density nor plan size explains the reversal. With matched data and recipe, a room-centric system with a reconciliation step and a wall-first sequence model reach comparable wall quality, so the output representation matters less than is usually assumed. A prior from the other family helps at the output but not at the input: deterministic fusion of the two outputs raises wall F1 by 7 points, whereas conditioning one model on the other's output gives no gain in three forms, including two ground-truth-content controls. We also provide an edit-cost metric that scores a draft by the human work needed to correct it, corrected CubiCasa5K annotations, and ResPlan-FP, a CC BY 4.0 benchmark of 16,998 plans with frozen splits and three baseline tracks. Code, the benchmark, and the corrected annotations are available at https://github.com/Cyprinus12138/fpvec-lab
Chinese Translation
一个经过训练的网络可以通过两种方式输出其结果，以恢复光栅化楼层平面图的墙壁、开口和房间：一种是以自回归坐标序列的形式发出几何信息，另一种是通过在密集的交汇点和中心线热图上检测几何信息并组装图形。我们比较了同一训练网络的两种读出方式。在真实扫描数据（CubiCasa5K）中，检测在每个墙壁度量上表现更好（在容差为0.05时，墙壁F1提高了2.7，在0.015时提高了5.1；配对自助法区间不包含零），而读取一个解码器从未使用的开口热图则在不重新训练的情况下将开口F1提高了2.6倍。在真实扫描中，读出的优势随着平面图大小的增加而增长，并在小平面图上逆转；在干净的矢量渲染中，当其训练覆盖渲染风格时，序列解码的表现比其他方法高出5到8分，而在完全领域转移下，给定校准阈值的读出仍然领先；墨水密度和平面图大小并不能解释这种逆转。在匹配数据和配方的情况下，采用调和步骤的以房间为中心的系统和以墙壁优先的序列模型达到了相当的墙壁质量，因此输出表示的重要性低于通常的假设。来自另一家族的先验在输出时有所帮助，但在输入时没有：两种输出的确定性融合将墙壁F1提高了7分，而将一个模型的条件设定为另一个模型的输出在三种形式中没有获得任何增益，包括两个真实内容控制。我们还提供了一种编辑成本指标，通过所需的人力工作来评分草稿，修正的CubiCasa5K注释，以及ResPlan-FP，一个包含16,998个平面图的CC BY 4.0基准，具有固定的分割和三个基线轨道。代码、基准和修正的注释可在 https://github.com/Cyprinus12138/fpvec-lab 获取。
View on arXiv Download PDF AI Translation
cs.CV / 49 / 2608.25609

On the Separation of Human and AI-Generated Images in CLIP Embedding Space

在CLIP嵌入空间中区分人类与AI生成图像
Asperti, Andrea
Abstract
We identify a previously unreported phenomenon in CLIP representations: human and AI-generated paintings spontaneously separate along the dominant principal directions of their joint embedding distribution, without any supervised objective designed to distinguish the two classes. Rather than exploiting this phenomenon for detection, our objective is to interpret it: we seek to identify the visual information underlying the separation and to trace it back from the embedding space to the image domain. We pursue this objective through a progressive investigation combining interpretable image representations with gradient-based inversion, used systematically as an experimental probe of the relationships identified in feature space. Robustness experiments and increasingly expressive statistical descriptors progressively rule out several intuitive explanations based on global image properties and simple local statistics, and point instead to distributed multiscale image structure. Multiscale scattering provides the most informative interpretable representation considered, but offers only a partial account of the phenomenon. Direct inversion provides a complementary and striking observation: substantial displacements along the dominant CLIP directions can be induced by image perturbations that remain nearly imperceptible to human observers, showing that the directions involved in the separation are highly sensitive to image variations with very low perceptual salience for humans. Taken together, these results reveal a significant difference between the visual evidence reflected in CLIP representations and that readily accessible to human perception, raising broader questions about the relationship between artificial and human vision and, ultimately, between artificial and human aesthetic judgment.
Chinese Translation
我们发现了CLIP表示中一个先前未报告的现象：人类与AI生成的绘画在其联合嵌入分布的主导主方向上自发分离，而没有任何旨在区分这两类的监督目标。我们的目标不是利用这一现象进行检测，而是对其进行解释：我们寻求识别导致分离的视觉信息，并将其从嵌入空间追溯到图像域。我们通过结合可解释的图像表示与基于梯度的反演进行渐进式调查，系统地作为特征空间中识别关系的实验探针。稳健性实验和日益丰富的统计描述逐步排除了基于全局图像属性和简单局部统计的几种直观解释，而是指向分布式多尺度图像结构。多尺度散射提供了最具信息性的可解释表示，但仅对这一现象提供了部分解释。直接反演提供了一个互补且引人注目的观察：沿主导CLIP方向的显著位移可以通过对人类观察者几乎不可察觉的图像扰动引起，这表明参与分离的方向对图像变化高度敏感，而这些变化对人类的感知几乎没有显著性。综合来看，这些结果揭示了CLIP表示中反映的视觉证据与人类感知中易于获取的视觉证据之间的显著差异，提出了关于人工视觉与人类视觉之间关系的更广泛问题，以及最终关于人工与人类审美判断之间的关系。
View on arXiv Download PDF AI Translation
cs.CV / 50 / 2608.25622

Plans You Can Check: Verifier-Grounded Learning of an Open-Weight Planner for Executable Video-Editing

可检查的计划：基于验证者的可执行视频编辑器开放权重规划学习
Wang, Haoyu, Feng, Cheng, Bian, Liuyang, Huang, Ruiyang, Wei, Lei, Wen, Yafei, Chen, Xiaoxin, Tang, Xiaoying
Abstract
Practical video editing is not only pixel generation: an editor must turn a brief, a clip pool, music metadata, and hard constraints into an executable timeline. We study this decision layer as \emph{executable video-editing planning} and introduce RefineCut, which, unlike workflow systems that wrap a prompted frontier model, trains a compact open-weight planner for it. The planner edits a typed timeline through structured patches covering clip selection, trimming, ordering, transitions, and duration and music alignment; a deterministic verifier applies each patch and checks it against an explicit constraint ledger. Because editing has no single ground-truth repair, we do not imitate teachers directly: RefineCut replays every multi-teacher branch through the verifier and keeps verifier-best repairs as supervision. A second stage, RefineCut-Evo, lets the student score its own repairs with the verifier and a task rubric and trains on high-margin preference pairs, so the final $8$B planner runs in a closed verifier loop with no teacher calls at inference. On RefineCut-Bench ($3{,}578$ tasks, $7{,}971$ captioned clips, $499$ music tracks, explicit ledgers), verifier-replayed distillation lifts the planner from $0.620$ to $0.858$ on the protocol-specific Video-Editing Score and RefineCut-Evo reaches $0.924$; the gain transfers to Llama-3.1-8B and GLM-4-9B, and in the same closed loop the $8$B planner matches or exceeds its frontier teachers. Code and RefineCut-Bench are publicly released; see the Data Availability statement.
Chinese Translation
实际的视频编辑不仅仅是像素生成：编辑者必须将简要说明、剪辑池、音乐元数据和硬性约束转化为可执行的时间线。我们将这一决策层研究为可执行视频编辑规划，并引入RefineCut，与包裹提示前沿模型的工作流系统不同，它为此训练了一个紧凑的开放权重规划器。该规划器通过覆盖剪辑选择、修剪、排序、过渡、持续时间和音乐对齐的结构化补丁来编辑一个类型化时间线；一个确定性验证器应用每个补丁并根据明确的约束账本进行检查。由于编辑没有单一的真实修复，我们并不直接模仿教师：RefineCut通过验证器重放每个多教师分支，并将验证器最佳修复作为监督。第二阶段RefineCut-Evo允许学生使用验证器和任务评分标准对自己的修复进行评分，并在高边际偏好对上进行训练，因此最终的$8$B规划器在推理时以闭环验证器循环运行，没有教师调用。在RefineCut-Bench（$3{,}578$个任务，$7{,}971$个带字幕剪辑，$499$个音乐轨道，明确账本）中，验证器重放的蒸馏将规划器在协议特定的视频编辑评分上从$0.620$提升至$0.858$，而RefineCut-Evo达到$0.924$；这一增益转移到Llama-3.1-8B和GLM-4-9B，并且在同一闭环中，$8$B规划器的表现与其前沿教师相匹配或超过。代码和RefineCut-Bench已公开发布；详见数据可用性声明。
View on arXiv Download PDF AI Translation
cs.CV / 51 / 2608.25630

SeVeR: Selective Visual Exposure and Retrieval for 3D Medical Image Question Answering

SeVeR：用于3D医学图像问答的选择性视觉暴露与检索
Hu, Yaojun, Tu, Danyang, Liu, Yang, Zhang, Jiajin, Fang, Wei, Liu, Zhiqiang, Dong, Chunlai, Xia, Yingda, Ying, Haochao, Wu, Jian, Zhang, Ling
Abstract
Volumetric medical VQA requires reasoning over long and redundant 3D visual token sequences, especially in multi-sequence MRI where complementary modalities provide diverse diagnostic cues but expose the decoder to many repeated anatomical regions. To investigate reasoning under multi-sequence visual redundancy, we first introduce BreMRIs-VQA, a clinically curated breast MRI benchmark with 1.19M QA pairs from 71.0K sequences and 12.9K patients, covering both free-text and multiple-choice questions. We further propose SeVeR, a selective visual exposure framework that compresses dense volumes into modality-wise prototypes and retrieves complementary multi-level evidence with change-aware gated attention during decoding, trained with a marginal-utility self-consistency objective that suppresses unhelpful retrieval. Experiments on BreMRIs-VQA and public benchmarks show that SeVeR improves both discriminative and generative performance while exposing substantially fewer visual tokens.
Chinese Translation
体积医学视觉问答（VQA）需要对冗长且重复的3D视觉标记序列进行推理，特别是在多序列MRI中，互补模态提供了多样的诊断线索，但使解码器暴露于许多重复的解剖区域。为了研究多序列视觉冗余下的推理，我们首先引入BreMRIs-VQA，这是一个临床策划的乳腺MRI基准数据集，包含来自71,000个序列和12,900名患者的119万个问答对，涵盖了自由文本和多项选择问题。我们进一步提出SeVeR，一个选择性视觉暴露框架，它将密集体积压缩为模态特征原型，并在解码过程中通过变化感知门控注意力检索互补的多层次证据，训练时采用边际效用自一致性目标，以抑制无效的检索。在BreMRIs-VQA和公共基准上的实验表明，SeVeR在显著减少视觉标记的同时，提高了区分性和生成性表现。
View on arXiv Download PDF AI Translation
cs.CV / 52 / 2608.25648

MAMA-FLUX.2: Image-to-Image Synthesis of Post-Contrast Breast DCE-MRI for the MAMA-SYNTH Challenge

MAMA-FLUX.2：用于MAMA-SYNTH挑战的对比增强乳腺DCE-MRI图像到图像合成
Kwarciak, Kamil, Wodzinski, Marek
Abstract
Dynamic contrast-enhanced breast MRI is central to cancer diagnosis and monitoring, but requires gadolinium-based contrast agents. In this work, we address pre-to-post contrast breast MRI synthesis for the MAMA-SYNTH challenge. We propose MAMA-FLUX.2, a conditional latent flow-matching approach based on FLUX.2-Klein-4B. The pre-contrast image is encoded as spatial conditioning, while the model predicts the flow field associated with the post-contrast target latent. To adapt the pretrained model efficiently, we use LoRA fine-tuning and introduce a regional training objective combining global flow matching, tumor-region supervision, and stable foreground regularization. We further investigate LoRA rank, intensity windowing, and regional loss weights on axial slices, prioritizing clinically relevant tumor-focused metrics. Our ablation study shows that moderate tumor and stable-foreground weighting improves the trade-off between image fidelity and tumor-region accuracy. The final model achieves the best overall balance with LoRA rank/$\alpha=64/64$, $\mathrm{MHA}_{\max}=25$, $\lambda_{\mathrm{tumor}}=0.25$, and $\lambda_{\mathrm{stable}}=0.1$. These results demonstrate that compact pretrained rectified-flow transformers can be adapted for contrast-enhanced MRI synthesis using parameter-efficient fine-tuning and task-aware regional losses.
Chinese Translation
动态对比增强乳腺MRI在癌症诊断和监测中至关重要，但需要基于钆的对比剂。在本研究中，我们针对MAMA-SYNTH挑战解决了从对比前到对比后的乳腺MRI合成问题。我们提出了MAMA-FLUX.2，这是一种基于FLUX.2-Klein-4B的条件潜在流匹配方法。对比前图像被编码为空间条件，而模型预测与对比后目标潜在相关的流场。为了有效地适应预训练模型，我们使用LoRA微调，并引入了一个区域训练目标，结合了全局流匹配、肿瘤区域监督和稳定前景正则化。我们进一步研究了LoRA秩、强度窗口和轴向切片上的区域损失权重，优先考虑临床相关的肿瘤聚焦指标。我们的消融研究表明，适度的肿瘤和稳定前景加权改善了图像保真度与肿瘤区域准确性之间的权衡。最终模型在LoRA秩/$eta=64/64$、$ ext{MHA}_{ ext{max}}=25$、$eta_{ ext{tumor}}=0.25$和$eta_{ ext{stable}}=0.1$的条件下实现了最佳整体平衡。这些结果表明，紧凑的预训练修正流变换器可以通过参数高效的微调和任务感知的区域损失适应于对比增强MRI合成。
View on arXiv Download PDF AI Translation
cs.CV / 53 / 2608.25652

Diffusion Transformers for Roof Graph Synthesis and Reconstruction

用于屋顶图合成与重建的扩散变换器
Panangian, Daniel, Bittner, Ksenia
Abstract
We present RoofDiT, a generative framework for 2D roof graph synthesis and reconstruction. Roofs are compactly described as planar graphs of junctions and structural edges, but existing methods often rely on fixed geometric rules or direct reconstruction objectives. RoofDiT instead models roof structures directly as vertex-edge graphs and learns a conditional generative prior over their geometry and connectivity. Our framework follows a two-stage design: a diffusion transformer generates roof vertices, and an edge prediction module infers the corresponding graph topology. To improve geometric fidelity, RoofDiT combines relative geometry-aware attention with footprint and aerial-image conditioning, while using an alignment regularizer to encourage common horizontal, vertical, and diagonal roof patterns. The same model supports unconditional generation, footprint-conditioned synthesis, and image-guided reconstruction by changing the conditioning signal. Experiments show improved graph generation quality over a diffusion baseline, favorable performance against a straight-skeleton prior in the footprint-conditioned setting, and the highest edge F1 among compared methods for image-guided reconstruction.
Chinese Translation
我们提出了RoofDiT，一个用于二维屋顶图合成与重建的生成框架。屋顶被紧凑地描述为由交点和结构边缘构成的平面图，但现有方法通常依赖于固定的几何规则或直接的重建目标。RoofDiT则直接将屋顶结构建模为顶点-边图，并学习其几何形状和连通性的条件生成先验。我们的框架遵循两阶段设计：一个扩散变换器生成屋顶顶点，而边预测模块推断相应的图拓扑。为了提高几何保真度，RoofDiT结合了相对几何感知注意力与足迹和航拍图像条件，同时使用对齐正则化器来鼓励常见的水平、垂直和对角屋顶模式。同一模型通过改变条件信号支持无条件生成、足迹条件合成和图像引导重建。实验表明，相较于扩散基线，图生成质量有所提高，在足迹条件设置下相较于直骨架先验表现良好，并且在图像引导重建中，在比较方法中具有最高的边F1分数。
View on arXiv Download PDF AI Translation
cs.CV / 54 / 2608.25653

Towards Purified Multi-Label Test-Time Adaptation of Vision-Language Models

面向视觉-语言模型的纯化多标签测试时适应
Liang, Yiwen, Chen, Hui, Xiong, Yizhe, Lyu, Mengyao, Cao, Yuhan, Lin, Zijia, Niu, Shuaicheng, Zhao, Sicheng, Han, Jungong, Ding, Guiguang
Abstract
Test-time adaptation (TTA) has been widely explored in single-label recognition, effectively mitigating distribution shifts, especially when combined with vision-language models. However, real-world images often contain multiple objects, while the more practical multi-label test-time adaptation (MLTTA) has received little attention so far. Recent cache-based TTA methods have shown promising efficiency and effectiveness, yet directly extending them to multi-label scenarios suffers from a one-to-many mapping problem: a shared global representation entangling co-occurring objects is stored as class-wise cache prototypes, inducing dominant-label bias and compromised cache calibration. While introducing region-level cues helps isolate class-specific evidence, such regional evidence can also be unreliable under distribution shifts, making its identification and utilization non-trivial. To address these issues, we introduce PuRF, a novel PuRiFication-driven cache-based method for multi-label test-time adaptation of vision-language models. Specifically, PuRF first performs region purification to identify reliable regions, providing comprehensive regional cues for multi-label recognition and enabling fine-grained alignment. Based on these purified regions, PuRF conducts cache purification to enhance cache representation and adaptability, where episodic purification builds a discriminative region-based cache, and temporal refreshing further promotes long-term cache adaptability. Experiments demonstrate that PuRF consistently outperforms state-of-the-art methods, achieving a notable 4.05% mAP improvement on ViT-B/32 across five datasets.
Chinese Translation
测试时适应（TTA）在单标签识别中得到了广泛探索，有效缓解了分布偏移，尤其是在与视觉-语言模型结合时。然而，现实世界中的图像通常包含多个对象，而更具实用性的多标签测试时适应（MLTTA）迄今为止受到的关注较少。近期基于缓存的TTA方法显示出良好的效率和有效性，但直接将其扩展到多标签场景时面临一对多映射问题：共享的全局表示将共现对象纠缠在一起，作为类别缓存原型存储，从而导致主导标签偏差和缓存校准受损。虽然引入区域级线索有助于隔离类别特定证据，但在分布偏移下，这些区域证据也可能不可靠，使其识别和利用变得复杂。为了解决这些问题，我们提出了PuRF，一种新颖的基于纯化的缓存方法，用于视觉-语言模型的多标签测试时适应。具体而言，PuRF首先进行区域纯化，以识别可靠区域，为多标签识别提供全面的区域线索，并实现细粒度对齐。在这些纯化区域的基础上，PuRF进行缓存纯化，以增强缓存表示和适应性，其中情景纯化构建了基于区域的判别缓存，时间刷新进一步促进了长期缓存适应性。实验表明，PuRF在五个数据集上始终优于最先进的方法，在ViT-B/32上实现了显著的4.05% mAP提升。
View on arXiv Download PDF AI Translation
cs.CV / 55 / 2608.25675

Deep Learning Segmentation of Diffusion-Weighted MRI Acute Ischaemic Stroke: A Pragmatic Evaluation Across Three Datasets

深度学习在扩散加权MRI急性缺血性中风分割中的应用：跨三个数据集的务实评估
Bjørnerud, Atle, Schellhorn, Till, Skattør, Thor H., Nome, Terje, Ottesen, Jon André, Aamodt, Anne Hege, MacIntosh, Bradley J
Abstract
Objective: Diffusion-weighted MRI (DWI-MRI) is the gold standard for visualizing and quantifying acute ischaemic stroke (AIS). Although deep learning methods can accurately segment AIS lesions, the optimal image inputs and model architecture remain uncertain. We evaluated whether accurate AIS lesion segmentation can be achieved using a pragmatic deep learning approach with minimal preprocessing and clinically feasible inference times. Materials and Methods: Self-configured nnU-Net models were trained on 1,744 DWI cases from local, national, and open-access datasets and tested on 436 cases. Four experimental conditions were evaluated using five-fold cross-validation: with or without brain extraction and using either DWI alone or DWI plus apparent diffusion coefficient (ADC) images as inputs. Two architectures were compared: the baseline nnU-Net (base) and a residual encoder nnU-Net (ResEnc). Performance was benchmarked against the DeepISLES ensemble model from the 2022 ISLES challenge. Results: In the test set (n=436), the base model achieved a median (IQR) Dice similarity coefficient (DSC) of 0.84 (0.19). For the base model, only two of six pairwise comparisons between input configurations showed significant differences. ResEnc produced small but significant improvements in DSC compared with the base model for DWI, DWI+brain extraction, and DWI+ADC inputs (all p<0.02), but not for DWI+ADC+brain extraction (p>0.50). The base model significantly outperformed DeepISLES, particularly in patients with smaller infarct volumes (signed-rank test, p<0.01). Conclusions: A baseline nnU-Net trained on DWI alone, without preprocessing, enabled fast and accurate AIS lesion segmentation. This streamlined approach may facilitate clinical research and support acute stroke imaging workflows
Chinese Translation
目的：扩散加权MRI（DWI-MRI）是可视化和量化急性缺血性中风（AIS）的金标准。尽管深度学习方法能够准确分割AIS病灶，但最佳图像输入和模型架构仍不确定。我们评估了是否可以通过一种务实的深度学习方法，在最小预处理和临床可行的推断时间内，实现准确的AIS病灶分割。材料与方法：自配置的nnU-Net模型在来自地方、国家和开放获取数据集的1,744个DWI案例上进行训练，并在436个案例上进行测试。使用五折交叉验证评估了四种实验条件：有或没有脑提取，以及使用DWI单独或DWI加表观扩散系数（ADC）图像作为输入。比较了两种架构：基线nnU-Net（base）和残差编码器nnU-Net（ResEnc）。性能与2022年ISLES挑战中的DeepISLES集成模型进行了基准比较。结果：在测试集（n=436）中，基线模型的中位数（IQR）Dice相似系数（DSC）为0.84（0.19）。对于基线模型，六个输入配置之间的配对比较中仅有两个显示出显著差异。与基线模型相比，ResEnc在DWI、DWI+脑提取和DWI+ADC输入下的DSC上产生了小但显著的改善（均p<0.02），但在DWI+ADC+脑提取下并未显著改善（p>0.50）。基线模型显著优于DeepISLES，尤其是在小梗死体积的患者中（符号秩检验，p<0.01）。结论：仅基于DWI训练的基线nnU-Net，无需预处理，能够快速且准确地进行AIS病灶分割。这种简化的方法可能促进临床研究并支持急性中风影像学工作流程。
View on arXiv Download PDF AI Translation
cs.CV / 56 / 2608.25692

CloSeR: Unified Relational Distillation from Closed-Set Teachers for Category Discovery

CloSeR：来自闭集教师的统一关系蒸馏用于类别发现
Liu, Yuanpei, He, Zhenqi, Tang, Jialu, Han, Kai
Abstract
Generalized Category Discovery (GCD) is an intriguing open-world problem that has garnered increasing attention: given partially labelled data, the goal is to correctly recognize known classes while discovering coherent novel categories from unlabelled samples. Recent GCD methods typically adapt foundation models by jointly optimizing supervised classification and unsupervised discovery objectives on mixed labelled and unlabelled data. While effective, this coupled training can entangle closed-set recognition and open-set discovery, leading to objective conflict and biased predictions, and may disturb the semantic geometry of pretrained representations under limited labels and noisy pseudo-labels. We propose CloSeR, a simple plug-and-play framework that injects Closed-Set Relational knowledge into GCD training. CloSeR first builds a domain-adapted closed-set teacher by tuning lightweight block-wise adapters on labelled known-class data while keeping the foundation model backbone frozen, thereby preserving pretrained priors at low training cost. It then transfers the teacher's knowledge to downstream GCD via Unified Relational Distillation (URD), which distills complementary global sample-to-prototype relations to anchor known-class semantics and local sample-to-sample relations to preserve neighborhood structure, using separate feature pathways to reduce optimization interference. CloSeR is head-agnostic and readily integrates with both parametric and non-parametric GCD methods. Extensive experiments with DINO and DINOv2 backbones on six benchmarks (CIFAR-10/100, ImageNet-100, CUB, Stanford-Cars, and FGVC-Aircraft) show consistent gains over GCD baselines, achieving state-of-the-art performance. Project page: https://visual-ai.github.io/closer/
Chinese Translation
广义类别发现（GCD）是一个引人注目的开放世界问题，近年来受到越来越多的关注：给定部分标记的数据，目标是在识别已知类别的同时，从未标记样本中发现一致的新类别。最近的GCD方法通常通过在混合标记和未标记数据上联合优化监督分类和无监督发现目标来适应基础模型。尽管有效，这种耦合训练可能会将闭集识别和开集发现纠缠在一起，导致目标冲突和偏差预测，并可能干扰在有限标签和噪声伪标签下的预训练表示的语义几何。我们提出了CloSeR，一个简单的即插即用框架，将闭集关系知识注入GCD训练。CloSeR首先通过在标记的已知类别数据上调整轻量级块级适配器来构建一个领域适应的闭集教师，同时保持基础模型主干不变，从而以低训练成本保留预训练先验。然后，它通过统一关系蒸馏（Unified Relational Distillation, URD）将教师的知识转移到下游GCD，该方法蒸馏互补的全局样本与原型关系以锚定已知类别的语义，并保留局部样本与样本之间的关系以保持邻域结构，使用独立的特征路径以减少优化干扰。CloSeR与参数化和非参数化GCD方法均无关，并且可以轻松集成。在六个基准（CIFAR-10/100、ImageNet-100、CUB、Stanford-Cars和FGVC-Aircraft）上使用DINO和DINOv2主干进行的大量实验显示，与GCD基线相比，CloSeR consistently achieves consistent gains，达到了最先进的性能。项目页面：https://visual-ai.github.io/closer/
View on arXiv Download PDF AI Translation
cs.CV / 57 / 2608.25693

Unsupervised Anatomical Feature Learning via Diffusion Models: Enhanced Medical Image Segmentation with Denoising Diffusion Probabilistic Models

通过扩散模型进行无监督解剖特征学习：利用去噪扩散概率模型增强医学图像分割
G, Akshat, Gupta, Divyansh, Bhatnagar, Shaleen, Ankalaki, Shilpa, Mishra, Tusar Kanti
Abstract
Acquiring pixel-level annotations for medical image segmentation is a severe bottleneck. Traditional U-Net architectures, while effective, learn local texture patterns and lack awareness of global anatomical structures, leading to boundary delineation failures in low-data regimes. This research paper proposes utilizing unsupervised Denoising Diffusion Probabilistic Models (DDPMs) to extract anatomical features. We train a DDPM on 21 unlabeled abdominal CT scans to learn structural representations, transferring the encoder weights to a downstream segmentation task evaluated on the BTCV multi-organ dataset. Diffusion pretraining significantly improved liver segmentation: Dice increased from $0.75\pm0.36$ to $0.93\pm0.16$ ($p < 5.33\times10^{-26}$, 0.529 Cohen's d), Average Surface Distance (ASD) decreased by 66%, and 95th-percentile Hausdorff Distance (HD95) reduced by 45%. For kidney segmentation, Dice improved from $0.90\pm0.19$ to $0.95\pm0.10$ ($p < 4.01\times10^{-11}$). Multi-organ pooled performance showed a 68% variance reduction and a 74% improvement in boundary precision (Dice $0.95\pm0.07$). Crucially, frozen encoder models retained > 80% of fine-tuned performance without exposure to segmentation labels, proving the existence of learned anatomical priors. In low-data scenarios, diffusion-pretrained models maintained robust performance with only 50% (Dice: 0.92 liver, 0.94 kidney), 25%, and even 10% (Dice: 0.89 liver, 0.71 kidney) of labeled data. Using unlabeled images for diffusion-based pretraining successfully embeds robust anatomical features prior to human supervision, transforming U-Nets into anatomy-aware systems.
Chinese Translation
获取医学图像分割的像素级标注是一个严重的瓶颈。传统的 U-Net 架构虽然有效，但主要学习局部纹理模式，缺乏对全局解剖结构的认知，导致在低数据环境下边界划分失败。本文提出利用无监督的去噪扩散概率模型（Denoising Diffusion Probabilistic Models, DDPMs）来提取解剖特征。我们在 21 个未标记的腹部 CT 扫描上训练 DDPM，以学习结构表示，并将编码器权重转移到下游分割任务中，评估数据集为 BTCV 多脏器数据集。扩散预训练显著改善了肝脏分割：Dice 从 $0.75 ext{±}0.36$ 提高到 $0.93 ext{±}0.16$ ($p < 5.33 imes10^{-26}$, 0.529 Cohen's d)，平均表面距离（Average Surface Distance, ASD）减少了 66%，95 百分位 Hausdorff 距离（HD95）降低了 45%。对于肾脏分割，Dice 从 $0.90 ext{±}0.19$ 提升至 $0.95 ext{±}0.10$ ($p < 4.01 imes10^{-11}$)。多脏器的汇总性能显示出 68% 的方差减少和 74% 的边界精度提升（Dice $0.95 ext{±}0.07$）。重要的是，冻结的编码器模型在未接触分割标签的情况下保留了超过 80% 的微调性能，证明了学习到的解剖先验的存在。在低数据场景下，扩散预训练模型在仅使用 50%（Dice: 0.92 肝脏, 0.94 肾脏）、25% 甚至 10%（Dice: 0.89 肝脏, 0.71 肾脏）标记数据的情况下保持了稳健的性能。利用未标记图像进行基于扩散的预训练成功嵌入了稳健的解剖特征，为人类监督之前转变了 U-Net 成为具有解剖意识的系统。
View on arXiv Download PDF AI Translation
cs.CV / 58 / 2608.25701

Skeleton-based Zero-Shot Spatio-Temporal Action Localization via Weakly-Supervised Pretraining

基于骨架的零样本时空动作定位的弱监督预训练新策略
Nagano, Koshiro, Sato, Fumiaki, Hachiuma, Ryo, Tsutsukawa, Kazuki, Sekii, Taiki
Abstract
We propose a novel pretraining strategy for skeleton-based zero-shot spatio-temporal action localization to estimate unseen actions for person instances while overcoming high annotation costs for training via new target actions and pretraining using large-scale action scenery datasets. Specifically, our approach, termed Skeleton-Language feature Pooling Switching, introduces a weakly-supervised vision-language pretraining mechanism. This mechanism transitions pooling kernels from pretraining, which aggregates skeleton features at the video level and aligns them with each video's known action text embeddings, to the inference phase that computes instance-level features without training via target actions. Furthermore, we propose Scene-Mixed Discriminative Contrastive Learning to distinguish actions at the instance level within the combined scene through the MIL framework. Our experiments on four public spatio-temporal action localization and classification datasets demonstrate that the proposed method effectively addresses annotation limitations.
Chinese Translation
我们提出了一种新颖的预训练策略，用于基于骨架的零样本时空动作定位，以估计未见过的动作，同时克服通过新目标动作和使用大规模动作场景数据集进行预训练所带来的高标注成本。具体而言，我们的方法称为Skeleton-Language特征池切换，引入了一种弱监督的视觉-语言预训练机制。该机制将预训练阶段的池化核转移到推理阶段，在推理阶段计算实例级特征，而无需通过目标动作进行训练，从而聚合视频级的骨架特征，并将其与每个视频已知的动作文本嵌入对齐。此外，我们提出了场景混合判别对比学习，以通过多实例学习（MIL）框架在组合场景中区分实例级动作。我们在四个公共时空动作定位和分类数据集上的实验表明，所提出的方法有效解决了标注限制问题。
View on arXiv Download PDF AI Translation
cs.CV / 59 / 2608.25710

Difficulty-Aware Sample Allocation for Adaptive Data Augmentation in Semantic Segmentation

基于难度感知的样本分配用于语义分割中的自适应数据增强
Arigbabu, Olasimbo Ayodeji, Arigbabu, Abimbola Ismail
Abstract
Data augmentation is a standard component of modern semantic segmentation pipelines, but most augmentation techniques allocate transformations uniformly across training samples or adapt to a single difficulty signal such as loss. This ignores the fact that segmentation difficulty is multi-factorial, since ambiguous predictions, persistent optimization errors, rare classes, and complex object boundaries can each make a sample informative in different ways. This paper introduces Difficulty-Aware Sample Allocation (DASA), an architecture-agnostic framework that assigns stronger augmentation to samples estimated to be more difficult. DASA combines prediction ambiguity, training loss, class rarity, and boundary complexity into a normalized difficulty score, then maps that score to sample-specific augmentation strength during iterative training. Experiments on Oxford-IIIT Pet and binary Pascal VOC segmentation with U-Net, DeepLabV3, and SegFormer-B0 show that DASA improves over standard training and is competitive with or stronger than single-signal adaptive baselines. On Oxford-IIIT Pet, DASA improves DeepLabV3 from 0.633 to 0.740 mIoU. On binary Pascal VOC, DASA obtains the best foreground IoU for all three evaluated architectures. These results attest to the value of multi-factor difficulty estimation as a practical mechanism for directing augmentation where it is most useful.
Chinese Translation
数据增强是现代语义分割流程中的标准组成部分，但大多数增强技术在训练样本中均匀分配变换或仅适应于单一的难度信号，如损失。这忽视了分割难度是多因素的事实，因为模糊预测、持续的优化错误、稀有类别和复杂的物体边界都可以以不同的方式使样本具有信息量。本文提出了难度感知样本分配（Difficulty-Aware Sample Allocation, DASA），这是一个与架构无关的框架，能够为被估计为更困难的样本分配更强的增强。DASA将预测模糊性、训练损失、类别稀有性和边界复杂性结合成一个标准化的难度评分，然后在迭代训练过程中将该评分映射到样本特定的增强强度。在Oxford-IIIT Pet和二元Pascal VOC分割的实验中，使用U-Net、DeepLabV3和SegFormer-B0表明，DASA优于标准训练，并且与单信号自适应基线相比具有竞争力或更强的表现。在Oxford-IIIT Pet上，DASA将DeepLabV3的mIoU从0.633提高到0.740。在二元Pascal VOC上，DASA为所有三种评估架构获得了最佳前景IoU。这些结果证明了多因素难度估计作为一种实用机制的价值，能够有效引导增强到最有用的地方。
View on arXiv Download PDF AI Translation
cs.CV / 60 / 2608.25729

LongVU-TTT: Causal Test-Time Training for Visual Resampling in Long Video Understanding

LongVU-TTT：用于长视频理解的因果测试时训练视觉重采样
Ahmed, Mahmoud, Abdulah, Sameh, Ruwase, Olatunji, Jacobs, Sam Ade, Bode, Mathis, Elhoseiny, Mohamed
Abstract
Long-video MLLMs must model temporal change before a limited visual-token budget removes most frame evidence. We introduce LongVU-TTT, which inserts a convolutional Test-Time Training (TTT) resampler with causal fast-weight updates between the vision encoder and the LLM. Its grouped 2D fast weights adapt to each video and contextualize frame features before compression, while a hybrid uniform-and-change-aware selector retains explicit visual evidence for downstream reasoning. Under controlled conditions, TTT-Conv improves over TTT-MLP by up to +2.12 and bidirectional Mamba2 by up to +3.04 on MLVU, and it is stronger than attention- and fixed-state recurrent resamplers across three benchmarks. Analysis shows that the fast weights behave as a temporal aggregation state rather than a reliable long-horizon episodic memory: their benefit attenuates as evidence becomes more distant, motivating explicit frame retention. LongVU-TTT processes up to 512 frames before reducing them to 128 LLM frames and achieves competitive performance across five video understanding benchmarks.
Chinese Translation
长视频多模态语言模型（MLLMs）必须在有限的视觉标记预算移除大部分帧证据之前建模时间变化。我们提出了LongVU-TTT，它在视觉编码器和大语言模型（LLM）之间插入了一个具有因果快速权重更新的卷积测试时训练（TTT）重采样器。其分组的二维快速权重适应每个视频，并在压缩之前对帧特征进行上下文化，同时混合的均匀和变化感知选择器保留了明确的视觉证据以供下游推理。在受控条件下，TTT-Conv在MLVU上相较于TTT-MLP提升了最多+2.12，相较于双向Mamba2提升了最多+3.04，并且在三个基准测试中优于注意力和固定状态的递归重采样器。分析表明，快速权重表现为一种时间聚合状态，而非可靠的长期情节记忆：随着证据变得更加遥远，其效益减弱，促使显式帧保留。LongVU-TTT在将帧减少到128个LLM帧之前处理最多512帧，并在五个视频理解基准测试中实现了具有竞争力的性能。
View on arXiv Download PDF AI Translation
cs.CV / 61 / 2608.25733

MIMONet: Multi-scale Input and Multi-scale Output Network for Salient Object Detection

MIMONet：用于显著目标检测的多尺度输入和多尺度输出网络
Yao, Zhaojian, Gao, Wei, Zhao, Tiesong, Yuan, Hui, Kwong, Sam
Abstract
The existing methods for saliency detection task focus on the application of multi-level features, aiming to take advantage of the respective strengths of high- and low-level features. However, because the inputs of these models are single-size images, their multi-level features have difficulty in learning the knowledge of size variations of salient objects. Object-scale variation learning has great potential for detecting multi-scale objects, which has not been fully explored by existing methods. To improve the recognition ability of a model for objects with different sizes, we are inspired by the image pyramid to propose a Multi-scale Input and Multi-scale Output Network (MIMONet). In MIMONet, we extract multi-level features for three images with different resolutions to form three encoder branches, and information will be exchanged between the branches. The advantage of this approach is that the features of one branch can learn the knowledge of target size variation from the features of the other two branches. In addition, we design a Multi-scale Perception (MSP) module, in which the input feature layer is divided into several sub-layers with different resolutions. Capturing the multi-level structure information of the objects in these sub-layers can make the objects more fully perceived. For network training, we propose a Joint Saliency Loss (JSL), which can constrain multiple saliency maps output by the network to identify the same foreground objects, and induce their boundaries to be preserved clearly. Experimental results show that MIMONet has stronger detection capabilities and harvests better evaluation scores on multiple datasets compared to existing models. The code of our model will be released.
Chinese Translation
现有的显著性检测方法侧重于多层次特征的应用，旨在利用高层次和低层次特征各自的优势。然而，由于这些模型的输入是单一尺寸的图像，它们的多层次特征在学习显著目标尺寸变化的知识时存在困难。对象尺度变化学习在检测多尺度对象方面具有巨大潜力，但现有方法尚未充分探索。为了提高模型对不同尺寸对象的识别能力，我们受到图像金字塔的启发，提出了一种多尺度输入和多尺度输出网络（MIMONet）。在MIMONet中，我们为三张不同分辨率的图像提取多层次特征，以形成三个编码器分支，并在分支之间交换信息。这种方法的优势在于，一个分支的特征可以从其他两个分支的特征中学习目标尺寸变化的知识。此外，我们设计了一个多尺度感知（MSP）模块，其中输入特征层被划分为几个具有不同分辨率的子层。在这些子层中捕捉对象的多层次结构信息，可以使对象被更充分地感知。为了进行网络训练，我们提出了一种联合显著性损失（JSL），该损失可以约束网络输出的多个显著性图识别相同的前景对象，并使其边界保持清晰。实验结果表明，与现有模型相比，MIMONet具有更强的检测能力，并在多个数据集上获得了更好的评估分数。我们的模型代码将会发布。
View on arXiv Download PDF AI Translation
cs.CV / 62 / 2608.25734

InteractGesture: Progressive Chunk Guidance for Continuous Streaming Co-Speech Gesture Control

InteractGesture：用于连续流式共语手势控制的渐进式块引导
Pinyoanuntapong, Ekkasit, Deogade, Ajinkya, Streli, Paul, Zhang, Wenjing, Materzynska, Joanna, Wang, Pu, Ferrari, Vittorio, Shen, Jie
Abstract
Co-speech gesture generation has made significant progress toward realistic full-body motion from speaker audio, yet existing models lack fine-grained spatial controllability of individual joints. To address this, we introduce \emph{InteractGesture}, a model-agnostic, inference-time method for spatially controllable gesture generation. \emph{InteractGesture} guides target latent estimates of a diffusion sampler through a differentiable RVQ-VAE decoder, backpropagating spatial control gradients to adjust motion latents during sampling. A primary challenge in streaming co-speech generation is chunk-wise dependency: standard sequential inference freezes prior chunks, preventing spatial constraints in future chunks from adjusting preceding trajectories and causing boundary inconsistencies. To overcome this limitation, we propose \emph{Progressive Chunk Guidance}, a chunk-window strategy that maintains an active set of editable chunk latents with staggered delays, enabling spatial constraints to propagate gradients backward across chunk boundaries during streaming generation. Experiments on the BEAT2 dataset show that \emph{InteractGesture} improves multi-joint spatial control while preserving overall gesture quality. Furthermore, our approach supports diverse applications, including sparse joint positioning, dense joint trajectory control, and directional pointing. Our project page is available at https://exitudio.github.io/interactgesture-page .
Chinese Translation
共语手势生成在从说话者音频中实现逼真的全身运动方面取得了显著进展，但现有模型缺乏对单个关节的细粒度空间可控性。为了解决这一问题，我们提出了 extit{InteractGesture}，一种模型无关的推理时空可控手势生成方法。 extit{InteractGesture} 通过可微分的 RVQ-VAE 解码器引导扩散采样器的目标潜在估计，反向传播空间控制梯度以在采样过程中调整运动潜在。流式共语生成中的一个主要挑战是块间依赖性：标准的顺序推理冻结先前的块，阻止未来块中的空间约束调整先前的轨迹，从而导致边界不一致。为克服这一限制，我们提出了 extit{渐进式块引导}，这是一种块窗口策略，保持一组可编辑的块潜在，并采用错开的延迟，使空间约束能够在流式生成过程中跨块边界反向传播梯度。对 BEAT2 数据集的实验表明， extit{InteractGesture} 在保持整体手势质量的同时改善了多关节空间控制。此外，我们的方法支持多种应用，包括稀疏关节定位、密集关节轨迹控制和方向指向。我们的项目页面可在 https://exitudio.github.io/interactgesture-page 查阅。
View on arXiv Download PDF AI Translation
cs.CV / 63 / 2608.25736

Moving Beyond More Views: Redundancy-Aware Ego-Exo Fusion for Proficiency Estimation

超越更多视角：基于冗余感知的自我-外部融合用于能力评估
Dong, Xu, Li, Wanqing, Adeyemi-Ejeye, Anthony, Gilbert, Andrew
Abstract
EgoExo proficiency estimation aims to assess action quality by integrating fine-grained motion cues from egocentric (1st-person) views with spatial context from multiple exocentric (3rd-person) views. Simply adding more exocentric views degrades EgoExo performance, as redundant or noisy perspectives dilute useful motion cues. Our analysis identifies two key causes: (1) Multiview redundancy - From the data perspective, certain views provide limited or noisy information, diluting discriminative cues; (2) Overfitting - From the feature perspective, conventional fusion increases representational complexity, causing the model to memorise view-specific patterns rather than learn generalisable representations. To address these issues, we propose two complementary modules: AdaMVS, which adaptively identifies and fuses the most informative view tokens under weak supervision from the data perspective, and VIB-GB, which combines Gradient Blending and Variational Information Bottleneck regularisation from the feature perspective to compress redundant signals and suppress overfitting during training. Experiments on EgoExo-4D and EgoExo-Fitness demonstrate that our method learns both which view to look at and how to fuse them, achieving new state-of-the-art results. Our source code is available at https://github.com/dx199771/AdaMVS
Chinese Translation
自我-外部能力评估旨在通过整合来自自我中心（第一人称）视角的细粒度运动线索与来自多个外部（第三人称）视角的空间上下文来评估动作质量。简单地增加更多外部视角会降低自我-外部的性能，因为冗余或噪声视角会稀释有用的运动线索。我们的分析确定了两个关键原因：（1）多视角冗余 - 从数据的角度来看，某些视角提供有限或噪声的信息，稀释了区分性线索；（2）过拟合 - 从特征的角度来看，传统的融合增加了表示复杂性，导致模型记忆特定视角的模式，而不是学习可泛化的表示。为了解决这些问题，我们提出了两个互补模块：AdaMVS，它在数据视角的弱监督下自适应地识别和融合最具信息量的视角标记；以及VIB-GB，它结合了梯度混合（Gradient Blending）和变分信息瓶颈（Variational Information Bottleneck）正则化，从特征的角度压缩冗余信号并在训练过程中抑制过拟合。我们在EgoExo-4D和EgoExo-Fitness上的实验表明，我们的方法能够学习观察哪些视角以及如何融合它们，达到了新的最先进结果。我们的源代码可在https://github.com/dx199771/AdaMVS获取。
View on arXiv Download PDF AI Translation
cs.CV / 64 / 2608.25808

TDFNet: Tri-projection Deformable Fusion Network for Panoramic Salient Object Detection

TDFNet：用于全景显著物体检测的三投影可变形融合网络
Zhou, Qiangqiang, Yu, Jiacong, Xu, Jiawei, Chen, Yong, Huang, Xin, Li, Ping
Abstract
Recent years have witnessed the growing potential of panoramic salient object detection in robotic vision, virtual reality, and related applications. However, projecting spherical scenes onto 2D planes inevitably introduces geometric distortions, which fundamentally limit the effectiveness of existing projection-based methods. Specifically, Equirectangular Projection (ERP) suffers from severe polar stretching distortions, while cube map projection introduces discontinuities across cube-face boundaries, resulting in degraded feature discriminability and compromised geometric consistency. To address these limitations, we propose TDFNet, the first Tri-projection Deformable Fusion Network for panoramic salient object detection, exploiting complementary projection representations to alleviate geometric distortions and improve detection performance.Specifically, we design a cross-projection deformable attention (CDA) module that leverages spatial correspondences between different projections to construct geometry-aware sampling locations, guiding deformable attention for cross-projection contextual aggregation and enhancing robustness against projection-induced deformations. Furthermore, we introduce a latitude-guided fusion module, which utilizes spherical latitude priors to construct geometric confidence weights for adaptively balancing ERP and CMP features. Meanwhile, LGF incorporates distortion-reduced semantic references from Tangent Projection to achieve cross-projection feature refinement and spatial alignment.By constructing a three-branch encoding architecture based on ERP, CMP, and Tangent Projection, TDFNet simultaneously preserves global spatial continuity, local geometric details, and fine-grained boundary information.
Chinese Translation
近年来，全景显著物体检测在机器人视觉、虚拟现实及相关应用中展现出日益增长的潜力。然而，将球面场景投影到二维平面不可避免地引入几何失真，这从根本上限制了现有基于投影的方法的有效性。具体而言，等矩形投影（Equirectangular Projection, ERP）遭受严重的极地拉伸失真，而立方体映射投影（Cube Map Projection, CMP）在立方体面边界处引入不连续性，导致特征可区分性下降和几何一致性受损。为了解决这些限制，我们提出了TDFNet，这是首个用于全景显著物体检测的三投影可变形融合网络，利用互补的投影表示来减轻几何失真并提高检测性能。具体而言，我们设计了一个跨投影可变形注意力（Cross-projection Deformable Attention, CDA）模块，利用不同投影之间的空间对应关系构建几何感知的采样位置，引导可变形注意力进行跨投影上下文聚合，并增强对投影引起的变形的鲁棒性。此外，我们引入了一个纬度引导融合模块，利用球面纬度先验构建几何置信权重，以自适应平衡ERP和CMP特征。同时，LGF结合来自切线投影（Tangent Projection）的失真减少语义参考，实现跨投影特征细化和空间对齐。通过构建基于ERP、CMP和切线投影的三分支编码架构，TDFNet同时保持全局空间连续性、局部几何细节和细粒度边界信息。
View on arXiv Download PDF AI Translation
cs.CV / 65 / 2608.25810

Label-Free Foundational Model Selection for Medical Image Classification under Distribution Shift via Pseudo Label Discrepancy

基于伪标签差异的无标签基础模型选择在分布转移下的医学图像分类
Larrea, Juan Iñaki, Mansilla, Lucas, Ferrante, Enzo
Abstract
Foundation models are increasingly deployed for medical image analysis. However, under the inter-institutional distribution shift typical of deployment, their performance varies widely and cannot be known without target-domain labels, which are rarely available. This leaves a practical question unresolved: given several candidate foundational models and labeled-data from a source domain, which one to deploy in an unlabeled target domain? We propose a label-free selection criterion built on SUDO, a framework for evaluating clinical AI systems without ground-truth annotations. SUDO partitions the unlabeled target data by predicted probability and, for each region, measures a pseudo-label discrepancy reflecting class contamination; aggregated across regions, this yields a score (AURCC) requiring neither target annotation nor fine-tuning. We show that AURCC can be used to rank a variety of vision-language models (BioMedCLIP, CXR-CLIP, CheXzero, MedCLIP, MedImageInsight, CLIP) on chest X-ray classification across three inter-hospital shift scenarios, under zero-shot and MLP-probe regimes. The AURCC ranking recovers the ground-truth ranking with Spearman rho up to 0.943 (p<0.05). Against the natural baseline of ranking by held-out source accuracy, AURCC is competitive when the labeled source is large and yields a more accurate ranking once it is small; the regime of interest in resource-constrained settings.
Chinese Translation
基础模型在医学图像分析中的应用日益增多。然而，在典型的跨机构分布转移情况下，它们的性能差异很大，且在没有目标领域标签的情况下无法得知，而目标领域标签通常很难获得。这留下了一个未解决的实际问题：在多个候选基础模型和来自源领域的标注数据的情况下，应该在无标签的目标领域中部署哪一个模型？我们提出了一种基于 SUDO 的无标签选择标准，SUDO 是一个用于评估临床人工智能系统的框架，无需真实标签。SUDO 通过预测概率对无标签目标数据进行分区，并对每个区域测量反映类别污染的伪标签差异；在各区域汇总后，这产生了一个分数（AURCC），该分数既不需要目标注释也不需要微调。我们展示了 AURCC 可以用于对多种视觉-语言模型（BioMedCLIP, CXR-CLIP, CheXzero, MedCLIP, MedImageInsight, CLIP）在胸部 X 光分类中的排名，涵盖了三种跨医院转移场景，适用于零样本和 MLP 探测模式。AURCC 排名恢复了真实排名，斯皮尔曼相关系数高达 0.943 (p<0.05)。与通过保留源准确率进行排名的自然基线相比，当标注源数据较大时，AURCC 具有竞争力，而当标注源数据较小时，AURCC 产生更准确的排名；这是在资源受限环境中关注的模式。
View on arXiv Download PDF AI Translation
cs.CV / 66 / 2608.25819

Steer the Sampling, Not the Kernel Grid: Geometry-Guided Sampling Operator for Volumetric Segmentation

引导采样而非核网格：几何引导的体积分割采样算子
Wang, Sizhe, Peiris, Himashi, Chen, Zhaolin
Abstract
Accurate 3D segmentation is central to quantitative lesion assessment and anatomy mapping for clinical planning and follow-up. Thin, elongated, and fine anatomical/pathological structures (e.g., vessels) are a particularly challenging case: a one-voxel boundary error can disconnect a branch and change clinically relevant topology. In encoder-decoder networks (e.g., U-Net), repeated downsampling and fixed-grid convolution blur or alias fine structures and weaken orientation cues, so early mistakes propagate across scales. We propose a geometry-guided local operator that steers where features are sampled, rather than deforming convolutional kernels, under a single formulation for both feature refinement (stride 1) and resolution reduction (stride > 1). At each voxel, it predicts a local orientation and bounded step sizes, samples symmetrically along these directions, and transforms paired samples into compact geometric and boundary cues with lightweight mixing; a cross-scale consensus aligns encoder and decoder features at skip connections to reduce geometric mismatch. Replacing all stride 1 and stride 2 operators in a 3D U-Net yields consistent improvements on BraTS, MSD Hepatic Vessel, and TDSC-ABUS, with notably better boundary metrics (e.g., BraTS Dice 86.1 to 88.9, HD95 7.1 to 6.2; TDSC-ABUS HD95 39.1 to 27.8) while reducing parameters from 2.3M to 0.8M. We further demonstrate that the operator can be integrated into other backbones (e.g., nnU-Net, Swin-UNETR, and MedNeXt) without changing their macro-architectures while providing consistent performance gains.
Chinese Translation
准确的三维分割对于临床规划和随访中的定量病变评估和解剖映射至关重要。细长且精细的解剖/病理结构（例如血管）尤其具有挑战性：一个体素的边界错误可能会断开分支并改变临床相关的拓扑结构。在编码器-解码器网络（例如 U-Net）中，重复的下采样和固定网格卷积会模糊或混叠细小结构并削弱方向线索，因此早期的错误会在不同尺度间传播。我们提出了一种几何引导的局部算子，该算子引导特征的采样位置，而不是变形卷积核，在单一公式下同时实现特征细化（步幅为1）和分辨率降低（步幅>1）。在每个体素处，它预测局部方向和有限步长，沿这些方向对称采样，并通过轻量级混合将配对样本转化为紧凑的几何和边界线索；跨尺度共识在跳跃连接处对齐编码器和解码器特征，以减少几何不匹配。在3D U-Net中替换所有步幅为1和步幅为2的算子，在BraTS、MSD Hepatic Vessel和TDSC-ABUS上均取得了一致的改进，边界指标显著提升（例如，BraTS Dice从86.1提高到88.9，HD95从7.1降低到6.2；TDSC-ABUS HD95从39.1降低到27.8），同时将参数从2.3M减少到0.8M。我们进一步证明，该算子可以集成到其他主干网络（例如nnU-Net、Swin-UNETR和MedNeXt）中，而不改变其宏观架构，同时提供一致的性能提升。
View on arXiv Download PDF AI Translation
cs.CV / 67 / 2608.25828

FlowMoDL: Model-Based Deep Learning with Conjugate-Gradient Data Consistency for Highly Accelerated 4D Flow MRI Reconstruction

FlowMoDL：基于模型的深度学习与共轭梯度数据一致性用于高度加速的4D流动MRI重建
Gottwald, Tristan, Bruch, Michelle, Hassan, Mubashir-Ul, Alickovic, Fatma, Kloiber, Milan, Tenbrinck, Daniel, Panholzer, Torsten, Schaller, Melanie, Hutter, Jana
Abstract
We present FlowMoDL, an unrolled neural network for highly accelerated 4D flow MRI reconstruction that directly optimizes for both anatomical magnitude and phase-derived velocity accuracy. Building on the MoDL framework, FlowMoDL alternates a learned (3+1)D spatiotemporal denoiser with conjugate-gradient data-consistency updates based on the SENSE forward model. A novel dual-pathway conditioning scheme adapts the denoiser features and data-consistency weighting, enabling a single model to handle varying acceleration factors ($10\times$ to $50\times$). To ensure physiological accuracy, the network is trained using a deep-supervision composite loss that explicitly penalizes velocity magnitude and angular errors, stabilized by a curriculum schedule. We evaluate FlowMoDL on the multi-center CMRx4DFlow dataset against classical and deep-learning baselines (CG-SENSE, MoDL, FlowVN, and FlowMRI-Net). A key advantage of FlowMoDL is its superior gradient step efficiency. When evaluated under an equivalent, limited budget of gradient steps, competing flow-specific networks degrade significantly. In contrast, FlowMoDL robustly converges and strictly outperforms all competitors across all acceleration factors in magnitude SSIM, nRMSE, relative velocity error, and angular error, successfully recovering sharp structural details and temporally coherent velocity fields.
Chinese Translation
我们提出了FlowMoDL，这是一种用于高度加速的4D流动MRI重建的展开神经网络，直接优化解剖幅度和基于相位的速度精度。FlowMoDL基于MoDL框架，交替使用学习的（3+1）D时空去噪器与基于SENSE前向模型的共轭梯度数据一致性更新。一种新颖的双通道条件方案调整去噪器特征和数据一致性权重，使单一模型能够处理不同的加速因子（$10 imes$到$50 imes$）。为了确保生理准确性，网络使用深度监督复合损失进行训练，明确惩罚速度幅度和角度误差，并通过课程调度进行稳定。我们在多中心CMRx4DFlow数据集上评估FlowMoDL，与经典和深度学习基线（CG-SENSE、MoDL、FlowVN和FlowMRI-Net）进行比较。FlowMoDL的一个关键优势是其卓越的梯度步长效率。在相同的有限梯度步长预算下，竞争的流动特定网络显著退化。相比之下，FlowMoDL稳健地收敛，并在所有加速因子的幅度SSIM、nRMSE、相对速度误差和角度误差上严格超越所有竞争对手，成功恢复清晰的结构细节和时间一致的速度场。
View on arXiv Download PDF AI Translation
cs.CV / 68 / 2608.25836

Socialized Detector Learning: Trajectory-Guided and Reciprocal Distillation for Heterogeneous Object Detectors

社会化检测器学习：基于轨迹引导和互惠蒸馏的异构目标检测器
Li, Weihao, Zhu, Yunqi, Fan, Zhihe, Zhao, Ruipu, Tao, Boan, Yao, Xinjie, Fan, Yan, Zhu, Pengfei
Abstract
Object detection knowledge is fragmented across independently trained, heterogeneous detectors with complementary category supports. In socialized learning, this knowledge resides in a society, and learning aims to evolve the society collectively through exchange. However, aggregation-based socialization does not explicitly plan transfer order, whereas progressive multi-teacher distillation considers order but remains a one-way student enhancement in a shared category space. Building on Socialized Learning, we formulate Socialized Detector Learning (SDL) for heterogeneous, category-specialized object detectors and propose Trajectory-Guided and Reciprocal Distillation (TGRD).TGRD estimates directed operational Inter-Detector Transfer Difficulty (IDTD) from held-out feature-alignment residuals, precomputes a fixed score table, and greedily constructs a carrier trajectory. Along the trajectory, knowledge is progressively consolidated into a union-category carrier and then returned to experts through reciprocal transfer. A conditional proxy-certificate analysis shows that, under stated assumptions, the progressive certificate is no larger than an aggregated-target counterpart. On MS COCO with four heterogeneous experts and two carrier initializations, final carriers outperform epoch-matched simultaneous aggregation controls by 2.6 AP in both settings. Reciprocal detectors attain 20.8--28.4 AP on previously unsupported categories while remaining within 1.3 AP of original expert-specific performance. These results support order-aware progressive consolidation followed by reciprocal transfer as a viable mechanism for detector-society evolution.
Chinese Translation
目标检测知识在独立训练的异构检测器之间是碎片化的，这些检测器具有互补的类别支持。在社会化学习中，这些知识存在于一个社会中，学习的目标是通过交流共同进化这个社会。然而，基于聚合的社会化并没有明确规划转移顺序，而渐进式多教师蒸馏则考虑了顺序，但仍然是在共享类别空间中进行单向的学生增强。基于社会化学习，我们提出了异构类别专用目标检测器的社会化检测器学习（Socialized Detector Learning, SDL），并提出了基于轨迹引导和互惠蒸馏（Trajectory-Guided and Reciprocal Distillation, TGRD）。TGRD从保留的特征对齐残差中估计定向操作的检测器间转移难度（Inter-Detector Transfer Difficulty, IDTD），预计算固定的评分表，并贪婪地构建载体轨迹。在轨迹上，知识逐步整合到一个联合类别载体中，然后通过互惠转移返回给专家。条件代理证书分析表明，在给定假设下，渐进证书不大于聚合目标的对应物。在MS COCO数据集上，使用四个异构专家和两个载体初始化，最终载体在两个设置中均比与之匹配的同时聚合控制提高了2.6 AP。互惠检测器在之前不支持的类别上获得了20.8-28.4 AP，同时与原始专家特定性能保持在1.3 AP之内。这些结果支持顺序感知的渐进整合，随后进行互惠转移，作为检测器社会进化的可行机制。
View on arXiv Download PDF AI Translation
cs.CV / 69 / 2608.25845

THA-Flow Generative Model: Prosthesis Geometry Prediction from Preoperative CT

THA-Flow 生成模型：基于术前 CT 的假体几何预测
Wang, Yiping, Li, Jie, Shen, Jingyu, Wang, Liao
Abstract
Preoperative planning for total hip arthroplasty (THA) is commonly framed as selecting a single prosthesis configuration and placement for a patient's osseous anatomy. In practice, however, the same anatomy may admit several clinically reasonable solutions, making planning inherently a one-to-many problem that is better represented by a conditional probability distribution. We present THA-Flow, a conditional flow-matching model that generates three-dimensional prosthesis geometry directly from preoperative CT. Separate AutoencoderKL models compress preoperative bone anatomy and prosthesis geometry, while a three-dimensional UNet learns a rectified flow from Gaussian noise to the prosthesis latent space under spatial bone conditioning and optional structured prosthesis parameters. The retrospective cohort comprised 1,355 hips from 1,149 patients undergoing primary THA. Following rigid registration of postoperative CT to preoperative CT, the actual postoperative prostheses were transformed independently according to the pelvic and femoral registrations and represented as a dual-channel truncated signed distance field. The prosthesis autoencoder achieved a peak signal-to-noise ratio of 47.11 dB and a structural similarity index of 0.9964 on the validation set. Complete acetabular and femoral geometries were generated across seven major stem models representing 93.4% of the cohort. Repeated bone-conditioned sampling preserved component position, alignment, and the principal bone-prosthesis interfaces while allowing limited local geometric variation. To our knowledge, THA-Flow represents the first application of generative AI to three-dimensional surgical planning for THA.
Chinese Translation
全髋关节置换术（THA）的术前规划通常被视为为患者的骨骼解剖选择单一的假体配置和放置。然而，在实践中，相同的解剖结构可能允许多种临床合理的解决方案，使得规划本质上成为一个一对多的问题，更适合用条件概率分布来表示。我们提出了 THA-Flow，这是一种条件流匹配模型，能够直接从术前 CT 生成三维假体几何。独立的 AutoencoderKL 模型压缩术前骨骼解剖和假体几何，而三维 UNet 则在空间骨骼条件和可选的结构化假体参数下，从高斯噪声学习到假体潜在空间的校正流。回顾性队列包括 1,355 个髋关节，来自 1,149 名接受初次 THA 的患者。在将术后 CT 进行刚性配准到术前 CT 后，实际的术后假体根据骨盆和股骨的配准独立转换，并表示为双通道截断符号距离场。假体自编码器在验证集上达到了 47.11 dB 的峰值信噪比和 0.9964 的结构相似性指数。生成了覆盖七个主要干模型的完整髋关节和股骨几何，代表了 93.4% 的队列。重复的骨骼条件采样保留了组件位置、对齐和主要的骨-假体接口，同时允许有限的局部几何变化。根据我们的了解，THA-Flow 代表了生成性人工智能在 THA 三维外科规划中的首次应用。
View on arXiv Download PDF AI Translation
cs.CV / 70 / 2608.25851

DEFUSE: Generalizable Backdoor Defense for Self-Supervised Encoders with Generative Priors

DEFUSE：具有生成先验的自监督编码器的可推广后门防御
Chen, Tuo, Gui, Jie, Dong, Minjing, Fang, Lanting, Jia, Ju, Cui, Benlei, Liu, Jian
Abstract
Self-supervised learning (SSL) encoders are vulnerable to backdoor attacks, posing threats to both visual SSL encoders and vision-language encoders. Existing defenses are typically designed for only one of these paradigms and rely on restrictive assumptions such as access to uninfected in-distribution data or precomputed pseudo-labels, which are difficult to satisfy in practice. To address these limitations, we propose DEFUSE, a generalizable backdoor detection framework for SSL encoders. Inspired by Bayesian posterior inference, we reformulate backdoor detection as a representation-conditioned image likelihood estimation problem parameterized by a conditional diffusion generative model. Uninfected representations tend to yield semantically consistent reconstructions, whereas backdoored ones are more likely to be mapped to the attacker's target class or semantically meaningless images, deviating from the original semantics and thereby exposing the backdoor. However, we find that the exact likelihood is intractable, because highly abstracted representations discard the low-level information necessary for pixel-faithful reconstruction. We therefore relax the objective to semantic reconstruction and evaluate it in a well-separated representation space provided by a reference encoder. Rather than training from scratch, we fine-tune a pretrained diffusion model, leveraging its generative prior to map data onto the natural image manifold while preserving semantic content. Extensive experiments demonstrate that DEFUSE substantially outperforms existing detectors across diverse attack settings, generalizing to both visual SSL and vision-language encoders. Notably, our method greatly reduces the reliance on prior knowledge about the victim encoder or the attack strategy. The source code is available at https://github.com/jsrdcht/DEFUSE .
Chinese Translation
自监督学习（SSL）编码器易受到后门攻击，给视觉SSL编码器和视觉语言编码器带来了威胁。现有的防御方法通常仅针对这两种范式中的一种设计，并依赖于诸如访问未感染的分布内数据或预计算伪标签等限制性假设，这在实践中难以满足。为了解决这些局限性，我们提出了DEFUSE，一个针对SSL编码器的可推广后门检测框架。受到贝叶斯后验推断的启发，我们将后门检测重新表述为一个以表示为条件的图像似然估计问题，该问题由条件扩散生成模型参数化。未感染的表示往往会产生语义一致的重建，而后门表示更可能映射到攻击者的目标类别或语义上毫无意义的图像，偏离原始语义，从而暴露后门。然而，我们发现确切的似然是不可处理的，因为高度抽象的表示丢弃了像素忠实重建所需的低级信息。因此，我们将目标放宽至语义重建，并在参考编码器提供的良好分离的表示空间中进行评估。我们不是从头开始训练，而是微调一个预训练的扩散模型，利用其生成先验将数据映射到自然图像流形，同时保留语义内容。大量实验表明，DEFUSE在各种攻击设置下显著优于现有检测器，能够推广到视觉SSL和视觉语言编码器。值得注意的是，我们的方法大大减少了对受害编码器或攻击策略的先验知识的依赖。源代码可在 https://github.com/jsrdcht/DEFUSE 获取。
View on arXiv Download PDF AI Translation
cs.CV / 71 / 2608.25858

Precipitation Downscaling Using Foundation Model-Conditioned Diffusion

基于基础模型条件的降水下采样
Ribeiro, Victor Nascimento, Guevara, Jorge, Moraga, Jorge Sebastian, Lucas, Chris, Lord, Natalie, Taylor, Andrew, Lockhart, Edward, Trojak, Will, Schmude, Johannes, Jones, Anne
Abstract
High-resolution precipitation fields are essential for hydrological impact assessment, yet global climate model outputs are too coarse and biased for direct use. AI-based statistical downscaling with diffusion models offers a promising approach, but the mechanism by which large-scale atmospheric predictors condition generation remains largely unexplored. We investigate three conditioning strategies for a denoising diffusion probabilistic model applied to daily precipitation downscaling: channel concatenation of upsampled coarse predictors, cross-attention conditioning with a learned convolutional encoder, and cross-attention conditioning with the frozen encoder of the pretrained Prithvi WxC weather foundation model. All strategies are evaluated against an unconditioned baseline under identical conditions using probabilistic, distributional, spectral, and extreme-event metrics for the Colorado River Basin. Concatenation conditioning achieves the lowest point-wise CRPS and MSE, but tends to produce over-smoothed fields that suppress high-intensity events. In contrast, cross-attention conditioning provides substantially better distributional realism and modest improvements in spectral fidelity. Improvements are greatest for extremes: the Prithvi-WxC conditioned model retains over half of >100mm/day events, although estimates are uncertain due to limited samples. When trained on the full dataset, the learned convolutional model performs similarly to the foundation model-conditioned approach while requiring lower computational resources. However, the Prithvi-WxC-conditioned model achieves comparable performance with only five years of training data. These results indicate that cross-attention conditioning offers advantages over simple concatenation for probabilistic precipitation downscaling, and that pre-trained foundation model representations may offer benefits in data-limited settings.
Chinese Translation
高分辨率降水场对于水文影响评估至关重要，但全球气候模型的输出过于粗糙且存在偏差，无法直接使用。基于人工智能的统计下采样方法结合扩散模型提供了一种有前景的解决方案，但大规模大气预测因子如何影响生成的机制尚未得到充分探索。我们研究了三种针对日降水下采样的去噪扩散概率模型的条件策略：上采样粗糙预测因子的通道连接、使用学习的卷积编码器的交叉注意力条件，以及使用预训练的Prithvi WxC天气基础模型的冻结编码器的交叉注意力条件。在相同条件下，所有策略均与无条件基线进行评估，使用概率、分布、谱和极端事件指标对科罗拉多河流域进行评估。连接条件实现了最低的逐点CRPS和MSE，但往往会产生过于平滑的场，抑制高强度事件。相比之下，交叉注意力条件提供了显著更好的分布现实性和适度的谱保真度改善。对于极端事件的改善最大：Prithvi-WxC条件模型保留了超过一半的>100mm/天事件，尽管由于样本有限，估计存在不确定性。当在完整数据集上训练时，学习的卷积模型的表现与基础模型条件的方法相似，但所需的计算资源更低。然而，Prithvi-WxC条件模型在仅使用五年的训练数据时也能实现可比的性能。这些结果表明，交叉注意力条件在概率降水下采样中优于简单的连接方法，并且预训练的基础模型表示在数据有限的情况下可能提供优势。
View on arXiv Download PDF AI Translation
cs.CV / 72 / 2608.25862

Learning Late, Guiding Early: Timestep-Decoupled Semantic Guidance for Fair Face Generation

晚期学习，早期引导：时间步解耦的语义指导用于公平的人脸生成
Parida, Subir Kumar, Velmurugan, Rajbabu, Kotwal, Ketan, Sengar, R. S., Hiremath, Swati
Abstract
Demographic imbalance in synthetic face generation can propagate to downstream face recognition systems, making fairness an important consideration when diffusion models are used for data generation. Existing fairness-aware generation approaches often require model retraining, architectural modifications, or repeated guidance throughout the reverse diffusion process. In this work, we introduce Semantic Boundary Predictor (SBP), an inference-time framework that performs demographic guidance through a one-shot intervention during reverse denoising. Our approach is motivated by the observation that latent representations at different diffusion timesteps play distinct semantic roles: late-stage latents provide stronger demographic separability, whereas early-stage latents offer greater flexibility for semantic intervention. SBP leverages this timestep decoupling by learning linear semantic boundaries from late-stage latent representations while applying them only once at the initial noisy latent, allowing the remainder of the reverse denoising process to proceed unchanged. The method requires neither retraining nor fine-tuning of the underlying Latent Diffusion Model and operates without external balanced datasets. Experiments on CelebA-HQ demonstrate substantial improvements in demographic fairness, reducing fairness disparity by 98% for gender, 95% for binary race, and 15% for four-class race, while maintaining perceptual image quality across demographic groups. Owing to its one-shot inference strategy and model-agnostic design, SBP introduces only a small computational overhead and can be readily integrated with existing pre-trained latent diffusion models.
Chinese Translation
合成面孔生成中的人口统计不平衡可能会传播到下游的人脸识别系统，因此在使用扩散模型进行数据生成时，公平性成为一个重要的考虑因素。现有的公平性意识生成方法通常需要模型重训练、架构修改或在反向扩散过程中进行重复指导。在本研究中，我们引入了语义边界预测器（Semantic Boundary Predictor, SBP），这是一种在推理时通过一次性干预在反向去噪过程中执行人口统计指导的框架。我们的方法的动机是观察到不同扩散时间步的潜在表示扮演着不同的语义角色：晚期潜在表示提供更强的人口统计可分离性，而早期潜在表示则为语义干预提供更大的灵活性。SBP利用这种时间步解耦，通过从晚期潜在表示中学习线性语义边界，同时仅在初始噪声潜在表示上应用一次，从而使反向去噪过程的其余部分保持不变。该方法不需要重训练或微调基础的潜在扩散模型，并且在没有外部平衡数据集的情况下运行。在CelebA-HQ上的实验表明，人口统计公平性有显著改善，性别的公平差异减少了98%，二元种族减少了95%，四类种族减少了15%，同时在各人口统计组之间保持了感知图像质量。由于其一次性推理策略和模型无关的设计，SBP仅引入了少量计算开销，并且可以轻松与现有的预训练潜在扩散模型集成。
View on arXiv Download PDF AI Translation
cs.CV / 73 / 2608.25866

LUTSeg: A Longitudinal Multi-Expert Dataset for Ulcer Tissue Segmentation

LUTSeg：用于溃疡组织分割的纵向多专家数据集
Sanchez, Karen, Hinojosa, Carlos, Ávila, Albert A., Riano-Rojas, Andrea C., Romero, Diego H., Páez, Jenny C., Llinás, Martina, Ghanem, Bernard
Abstract
Quantifying wound tissue composition is essential for monitoring chronic ulcer progression and guiding treatment decisions. However, pixel-level annotations are costly, and multi-tissue wound datasets remain scarce, particularly for neglected diseases such as leprosy. We introduce LUTSeg, a longitudinal chronic ulcer dataset comprising 141 images from 39 patients with wound masks and five tissue categories annotated by five expert clinicians, including a multi-expert gold-standard subset for inter-rater agreement analysis. To establish an initial benchmark for LUTSeg, we further propose TiSage, a semi-supervised tissue segmentation framework that integrates multi-scale semantic priors from a frozen medical vision-language model within a teacher-student architecture. We evaluate TiSage on LUTSeg and DFUTissue, showing improvements over supervised and semi-supervised baselines in most low-label settings. Code & data: https://github.com/carlosh93/TiSage
Chinese Translation
量化伤口组织成分对于监测慢性溃疡的进展和指导治疗决策至关重要。然而，像素级标注成本高昂，且多组织伤口数据集仍然稀缺，尤其是在麻风等被忽视疾病中。我们引入了LUTSeg，这是一个纵向慢性溃疡数据集，包含来自39名患者的141幅图像，配有伤口掩膜和五类组织的标注，由五位专家临床医生进行标注，其中包括用于评估评审者一致性的多专家金标准子集。为了为LUTSeg建立初步基准，我们进一步提出了TiSage，这是一种半监督组织分割框架，结合了来自冻结的医学视觉-语言模型的多尺度语义先验，采用教师-学生架构。我们在LUTSeg和DFUTissue上评估了TiSage，结果显示在大多数低标签设置中，相较于监督和半监督基线有显著改善。代码与数据： https://github.com/carlosh93/TiSage
View on arXiv Download PDF AI Translation
cs.CV / 74 / 2608.25888

Embedding NDRE Trajectories into Contrastive Learning for Label-Free, Physiology-Aware Crop-Stress Staging and DSS Outputs

将 NDRE 轨迹嵌入对比学习中，以实现无标签、关注生理的作物压力分级和决策支持系统输出
Ahmad, Shafqaat
Abstract
Timely detection of crop stress is critical for sustaining yields under increasing drought frequency, yet conventional vegetation index thresholds or image-based clustering often fail to capture stress progression, limiting their value for farm decision-making. To address this gap, we present EigenCL, a physiology-guided contrastive learning framework that stages crop stress from Sentinel-2 NDRE trajectories, with the goal of providing interpretable and transferable stress diagnostics for decision support systems (DSS). EigenCL was trained on 10,000 maize NDRE patches from drought-affected Iowa fields in 2020 and tested on Nebraska fields in 2023 without retraining, with validation incorporating soil-moisture records, U.S. Drought Monitor maps, and county-level yield statistics. The model produced four physiologically coherent stress clusters (Healthy, Mild, Moderate, Severe), significantly outperforming baselines including K-Means, SimCLR, ProtoCLR, and an ablation model (Silhouette = 0.748, DBI = 0.35, CHI = 49,624). Clusters aligned with maize growth stages, with severe stress peaking around tasseling-silking (VT-R1), a stage known to drive yield loss; moreover, EigenCL clusters correlated with soil moisture at 0-14-day lags (rho up to 0.72) and matched yield anomalies in drought-affected counties. By embedding NDRE trajectory dynamics into contrastive learning, EigenCL enables early stress alerts and interpretable DSS outputs (e.g., heatmaps, scouting priorities, regional risk indices), extending beyond single-date NDRE thresholds and supporting scalable monitoring for climate-smart agronomy.
Chinese Translation
及时检测作物压力对于在日益频繁的干旱条件下维持产量至关重要，然而传统的植被指数阈值或基于图像的聚类方法往往无法捕捉压力的进展，限制了其在农场决策中的价值。为了解决这一问题，我们提出了 EigenCL，这是一种生理指导的对比学习框架，旨在通过 Sentinel-2 NDRE 轨迹对作物压力进行分级，提供可解释和可转移的压力诊断以支持决策支持系统（DSS）。EigenCL 在 2020 年对受干旱影响的爱荷华州田地的 10,000 个玉米 NDRE 片段上进行了训练，并在 2023 年对内布拉斯加州田地进行了测试，未进行重新训练，验证过程中结合了土壤湿度记录、美国干旱监测地图和县级产量统计数据。该模型生成了四个生理上连贯的压力聚类（健康、轻度、中度、重度），显著优于包括 K-Means、SimCLR、ProtoCLR 和一个消融模型在内的基线（Silhouette = 0.748，DBI = 0.35，CHI = 49,624）。聚类与玉米生长阶段一致，重度压力在抽雄-授粉（VT-R1）阶段达到峰值，该阶段已知会导致产量损失；此外，EigenCL 聚类与土壤湿度在 0-14 天滞后下的相关性达到 0.72，并与受干旱影响的县的产量异常相匹配。通过将 NDRE 轨迹动态嵌入对比学习，EigenCL 实现了早期压力预警和可解释的 DSS 输出（例如热图、侦查优先级、区域风险指数），超越了单一日期的 NDRE 阈值，并支持气候智能农业的可扩展监测。
View on arXiv Download PDF AI Translation
cs.CV / 75 / 2608.25924

Visual General Intelligence: A White Paper

视觉通用智能：白皮书
Kataoka, Hirokatsu, Fukuhara, Yoshihiro, Tian, Yonglong, Wu, Shangzhe, Deb, Oishi, Yamada, Ryousuke, Rupprecht, Christian, Wang, Jianyuan, Ide, Kohsuke, Namekata, Koichi, Ma, Xianzheng, Chen, Yiming, Geirhos, Robert, Raghunathan, Aditi, Asano, Yuki M., Ramanan, Deva, Fouhey, David, Davison, Andrew J., Du, Yilun, Wu, Jiajun, Liu, Zhuang
Abstract
This paper reconsiders intelligence from a vision-centered perspective and examines whether intelligence emerging from visual experience and learning may provide a pathway toward AGI. In the language domain, beginning with the introduction of the Transformer architecture, the GPT series has demonstrated transfer to unseen tasks through autoregressive language modeling on web-scale text combined with aggressive scaling. This raises a natural question, namely, what capabilities and forms of intelligence can emerge from visual modalities such as images, videos, and geometry? In this paper, we discuss whether visual intelligence can serve as a pathway toward AGI, referred to in this paper as visual general intelligence (VGI), by bringing together contributors from diverse standpoints and affiliations. Our aim is not to offer a single definition of visual intelligence, but to clarify the principles that computer vision should pursue in the AGI era, the visual input modalities, the benchmarks, the learning paradigms, and the relationship between vision, when taken as the core, and other modalities such as language.
Chinese Translation
本文从以视觉为中心的视角重新审视智能，并探讨从视觉经验和学习中产生的智能是否可以为通用人工智能（AGI）提供一条路径。在语言领域，自Transformer架构引入以来，GPT系列通过在网络规模文本上进行自回归语言建模和大规模扩展，展示了对未见任务的迁移能力。这引发了一个自然的问题，即，来自图像、视频和几何等视觉模态的智能和能力可以产生什么？本文讨论了视觉智能是否可以作为通用人工智能的路径，本文称之为视觉通用智能（VGI），并汇集了来自不同立场和机构的贡献者。我们的目标不是提供视觉智能的单一定义，而是澄清计算机视觉在AGI时代应追求的原则、视觉输入模态、基准测试、学习范式，以及以视觉为核心与语言等其他模态之间的关系。
View on arXiv Download PDF AI Translation
cs.CV / 76 / 2608.25927

Code World Model: Coding Agent as World Brain

代码世界模型：编码代理作为世界大脑
Chen, Yiwen, Lin, Guosheng, Zhang, Chi
Abstract
World models aim to simulate how complex environments evolve under actions and events, yet existing video-based world models primarily learn dynamics from visual observations, which reveal outcomes rather than the underlying knowledge, rules, and mechanisms governing world evolution. This makes it difficult to maintain persistent consequences and support coherent, open-ended evolution. We introduce Code World Model, a framework that separates world evolution from visual realization by combining the reasoning and coding capabilities of language models with the generative priors of video models. A coding agent serves as the world brain, reasoning about events and their consequences and generating executable code to maintain persistent world state and perform rule-consistent evolution. To connect executable state with visual generation, we introduce a proxy representation that encodes frame-wise spatiotemporal constraints and is compiled into a proxy video, which conditions a video model to render high-fidelity visual observations. We further develop data pipelines for constructing aligned proxy-observation pairs from gameplay and real-world videos. After fine-tuning on paired gameplay data, MiniMax-H3 follows proxy-based spatiotemporal specifications from simple interactive worlds built by the coding agent while preserving rich visual details and dynamics. These results demonstrate the potential of combining code for persistent world evolution with video models for flexible visual realization, providing a new path toward open-ended world models.
Chinese Translation
世界模型旨在模拟复杂环境在行动和事件下的演变，然而现有的基于视频的世界模型主要从视觉观察中学习动态，这些观察揭示了结果而非支配世界演变的基本知识、规则和机制。这使得维持持续的后果和支持连贯的开放式演变变得困难。我们提出了代码世界模型（Code World Model），一个通过结合语言模型的推理和编码能力与视频模型的生成先验，来将世界演变与视觉实现分离的框架。编码代理作为世界大脑，推理事件及其后果，并生成可执行代码以维持持久的世界状态并执行规则一致的演变。为了将可执行状态与视觉生成连接起来，我们引入了一种代理表示，它编码逐帧的时空约束，并被编译成代理视频，从而使视频模型能够渲染高保真的视觉观察。我们进一步开发了数据管道，以从游戏玩法和现实世界视频中构建对齐的代理观察对。在对配对的游戏数据进行微调后，MiniMax-H3遵循编码代理构建的简单交互世界中的基于代理的时空规范，同时保留丰富的视觉细节和动态。这些结果展示了将代码用于持久世界演变与视频模型用于灵活视觉实现相结合的潜力，为开放式世界模型提供了一条新的路径。
View on arXiv Download PDF AI Translation
cs.CV / 77 / 2608.25933

When Composition Doesn't Add Up: Humans Identifying Defects in AI-Generated Images

当组合不再成立：人类识别AI生成图像中的缺陷
Hu, Ruoqi, Zhao, Chulin, Chang, Jiashuo, Ruiz-Dolz, Ramon, Lin, Hanhe
Abstract
*Chulin Zhao and Ruoqi Hu contributed equally to this work. State-of-the-art text-to-image (T2I) models exhibit pronounced and systematic defects when prompts involve intricate compositional factors such as multiple entities and multiple attributes. In this paper, we investigate how humans identify such defects. Specifically, we manually select 651 reference images from the four categories of people, hand, object, and scene that exhibit complex compositional characteristics, from which prompts emphasizing compositional factors are derived by manually editing ChatGPT-generated prompts. We then feed the prompts into three selected T2I models to generate AI images and conduct a comprehensive subjective study to identify their defects. For each image, 29 participants provide multi-label assessments specifying defect types and locations. The study yields the compositional AI-generated image defect (CO-AID) dataset, including reference images, prompts, AI-generated images, and information on defect locations and types. Experimental results show that training a deep model on CO-AID can both predict defects in AI-generated images and optimize AI image generation, demonstrating its usability and effectiveness. The database and supplementary materials are available at: https://github.com/Future-IQA/CO-AID .
Chinese Translation
Chulin Zhao和Ruoqi Hu对本研究贡献相同。最先进的文本到图像（T2I）模型在涉及复杂组合因素（如多个实体和多个属性）时，表现出明显且系统性的缺陷。本文研究了人类如何识别这些缺陷。具体而言，我们从展示复杂组合特征的四个类别（人、手、物体和场景）中手动选择了651张参考图像，并通过手动编辑ChatGPT生成的提示，提取出强调组合因素的提示。然后，我们将这些提示输入到三个选定的T2I模型中生成AI图像，并进行全面的主观研究以识别其缺陷。对于每张图像，29名参与者提供多标签评估，指定缺陷类型和位置。研究结果生成了组合AI生成图像缺陷（CO-AID）数据集，包括参考图像、提示、AI生成的图像以及缺陷位置和类型的信息。实验结果表明，在CO-AID上训练深度模型可以预测AI生成图像中的缺陷并优化AI图像生成，展示了其可用性和有效性。数据库和补充材料可在以下网址获取：https://github.com/Future-IQA/CO-AID 。
View on arXiv Download PDF AI Translation
cs.CV / 78 / 2608.25935

TAU-Agent: An Agentic Retrieval-Augmented Framework for Traffic Anomaly Understanding

TAU-Agent：一种用于交通异常理解的代理检索增强框架
Lin, Yuqiang, Shi, Yan, Lockyer, Sam, Madabushi, Harish Tayyar, Evans, Adrian, Li, Wenbin, Wang, Yinhai, Zhang, Nic
Abstract
Traffic Anomaly Understanding (TAU) requires models and systems to detect, reason about, and explain anomalous events in transportation videos. To address this challenge, we propose TAU-Agent, an agentic retrieval-augmented framework for traffic anomaly understanding. Given a task query, a central retrieval agent orchestrates two visual perception tools, namely a Video Captioning Tool and an Open-Vocabulary Tracking Tool, to retrieve and select query-relevant evidence, including captions, temporal intervals, and object trajectories. The selected evidence, together with sampled video frames and the input query, is provided to a supervised fine-tuned vision-language model for final reasoning and answer generation. We evaluate TAU-Agent on both the in-domain and the out-of-domain benchmarks from the AI City Challenge 2026. TAU-Agent achieves scores of 0.6779 on Track 3, 0.3998 on Track 7, and 67.9275 on Track 8, ranking second, twelfth, and fifth, respectively. Code is available at: https://github.com/siri-rouser/TAU-Agent.
Chinese Translation
交通异常理解（TAU）需要模型和系统在交通视频中检测、推理和解释异常事件。为了解决这一挑战，我们提出了TAU-Agent，一种用于交通异常理解的代理检索增强框架。给定一个任务查询，一个中央检索代理协调两个视觉感知工具，即视频字幕工具（Video Captioning Tool）和开放词汇跟踪工具（Open-Vocabulary Tracking Tool），以检索和选择与查询相关的证据，包括字幕、时间间隔和物体轨迹。所选证据与采样的视频帧和输入查询一起提供给一个经过监督微调的视觉-语言模型，以进行最终推理和答案生成。我们在AI City Challenge 2026的领域内和领域外基准上评估了TAU-Agent。TAU-Agent在Track 3上获得0.6779的分数，在Track 7上获得0.3998，在Track 8上获得67.9275，分别排名第二、第十二和第五。代码可在以下链接获取：https://github.com/siri-rouser/TAU-Agent。
View on arXiv Download PDF AI Translation
cs.CV / 79 / 2608.25948

Auditable CT Phenotyping Through Report-derived Radiological Observations

可审计的CT表型通过报告衍生的放射学观察
Wu, Riga, Witschey, Walter, Li, Yicheng, Ordonez, Felix Barajas, Bressem, Keno K., Adams, Lisa C., Weissman, Gary E., Shen, Li, Davatzikos, Christos, Barbosa, Eduardo, Truhn, Daniel, Han, Tianyu
Abstract
Medical image foundation models can predict clinical phenotypes from computed tomography (CT), but strong performance leaves open whether they read disease-specific findings or shortcuts that correlate with the diagnosis. We tested this in 221 electronic-health-record (EHR) phenotypes using Auditable CT phenotyping (ACT), built on report-derived radiological observations. We trained ACT on 38,317 patients, mined 376,194 observations and evaluated it in 25,183 held-out patients. ACT exceeded five vision-language baselines on zero-shot annotation, and CT-CLIP across 221 phenotypes from unseen CT pulmonary angiography, both under zero-shot scoring (0.651 versus 0.572) and under linear probing (0.709 versus 0.662). Reading each probe exposes what accuracy conceals: only 97 observations occupy the 221 rank-1 positions, and one phrase describing aortic and coronary calcification ranks first for 20 phenotypes, including osteoporosis, urinary tract infection and major depressive disorder. Restricting the bank to clinician-specified evidence redirects those probes onto phenotype-related observations in 86 phenotypes at no accuracy cost (0.751 versus 0.741). Accurate CT-based EHR phenotyping can therefore rest on observations that are not valid evidence for the coded phenotype and that ACT can identify and intervene on.
Chinese Translation
医学图像基础模型可以从计算机断层扫描（CT）预测临床表型，但强大的性能使得我们无法确定它们是读取特定疾病的发现，还是与诊断相关的捷径。我们在221个电子健康记录（EHR）表型中测试了这一点，使用基于报告衍生放射学观察的可审计CT表型（ACT）。我们在38,317名患者上训练了ACT，挖掘了376,194个观察，并在25,183名保留患者中进行了评估。ACT在零样本注释中超越了五个视觉-语言基线，并在未见的CT肺动脉造影的221个表型中超越了CT-CLIP，无论是在零样本评分下（0.651对0.572）还是在线性探测下（0.709对0.662）。阅读每个探针揭示了准确性所隐藏的内容：只有97个观察占据了221个排名第一的位置，而描述主动脉和冠状动脉钙化的一个短语在包括骨质疏松症、尿路感染和重度抑郁症在内的20个表型中排名第一。将观察库限制为临床医生指定的证据，使这些探针转向86个表型相关的观察，而没有准确性损失（0.751对0.741）。因此，基于CT的EHR表型的准确性可以依赖于ACT能够识别和干预的并非有效证据的观察。
View on arXiv Download PDF AI Translation
cs.CV / 80 / 2608.25956

4DGS-WAM: Bridging Past and Future with an Object-Centric World Action Model based on 4D Gaussian Splatting

4DGS-WAM：基于4D高斯溅射的面向对象的世界行动模型，连接过去与未来
Ma, Yueen, Xu, Zenglin, King, Irwin
Abstract
Current world action models (WAMs) typically operate on 2D visual data. These models can achieve exceptional visual quality, but they lack explicit spatial structure for individual objects and repeatedly process redundant background content. Although point clouds can represent the world in 3D space, they can be difficult to align and accumulate across viewpoints. In this paper, we leverage an explicit 4D Gaussian Splatting (4DGS) representation that separately models dynamic objects and the static background of a scene. For dynamic objects, we use a policy model to predict future actor actions and a world model to predict transformations of their observed Gaussian splats. The static background need not be regenerated for future states, as much of it has already been observed in past frames. This forms an object-centric world action model, which we name 4DGS-WAM. It lifts 2D observations into a persistent 4D representation so that previously observed static content can be reused during future prediction. Future-state extrapolation can then focus on modeling the evolution of dynamic objects. Experiments on KITTI-MOT evaluate short-horizon prediction and past reconstruction.
Chinese Translation
当前的世界行动模型（WAMs）通常基于2D视觉数据进行操作。这些模型能够实现卓越的视觉质量，但缺乏对单个对象的明确空间结构，并且重复处理冗余的背景内容。尽管点云可以在3D空间中表示世界，但在不同视角之间对齐和累积可能会很困难。在本文中，我们利用一种明确的4D高斯溅射（4DGS）表示法，分别建模动态对象和场景的静态背景。对于动态对象，我们使用策略模型来预测未来行为者的动作，并使用世界模型来预测其观察到的高斯溅射的变换。静态背景不需要为未来状态重新生成，因为其中大部分内容已经在过去的帧中被观察到。这形成了一个面向对象的世界行动模型，我们将其命名为4DGS-WAM。它将2D观察提升为持久的4D表示，以便在未来预测中重用先前观察到的静态内容。未来状态的外推可以集中于建模动态对象的演变。在KITTI-MOT上的实验评估了短期预测和过去重建。
View on arXiv Download PDF AI Translation
cs.CV / 81 / 2608.25965

Less Contouring, More Accuracy: Lesion-Guided ROI Deep Learning for Ovarian Ultrasound Classification

减少轮廓描绘，提升准确性：基于病灶引导的区域兴趣深度学习用于卵巢超声分类
Ahmad, Mehran, Ardakani, Ali Abbasian, Mohammadi, Afshin, Mohebbi, Alisa, Kronreif, Gernot, Hatamikia, Sepideh
Abstract
Ovarian lesion classification using transvaginal ultrasound remains challenging due to overlapping imaging characteristics and the dependence on expert interpretation. This study investigates whether lesion-guided region-of-interest (ROI) deep learning can achieve competitive diagnostic performance while reducing the annotation burden associated with pixel-level lesion segmentation. Two publicly available ovarian ultrasound datasets were evaluated: the Multi-Modality Ovarian Tumor Ultrasound (MMOTU) dataset for eight-class classification and the Ovarian Ultrasound Dataset (OUD) for binary classification. Four strategies were compared under a unified framework: global image-based deep learning, lesion-guided ROI-based deep learning, lesion contour-based deep learning, and contour-based radiomics with machine learning classifiers. Four deep learning architectures, MaxViT-Tiny, Swin Transformer, EfficientNet-B7, and ResNet18, were evaluated. Radiomics models were developed using support vector machine, k-nearest neighbors, and artificial neural network classifiers, with ANOVA-based feature selection applied for the lower-sample OUD dataset. The lesion-guided ROI strategy achieved the strongest overall performance, with MaxViT-Tiny obtaining 93.10% accuracy and an AUC of 0.99 on MMOTU and 97.56% accuracy and an AUC of 0.99 on OUD. The contour-based approach achieved comparable accuracy but required substantially higher annotation effort. These findings demonstrate that lesion-guided ROI deep learning provides an effective balance between diagnostic performance and annotation efficiency, offering a practical approach for scalable AI-assisted ovarian ultrasound analysis
Chinese Translation
使用经阴道超声进行卵巢病灶分类仍然具有挑战性，因为影像特征重叠且依赖于专家解读。本研究探讨了病灶引导的区域兴趣（ROI）深度学习是否能够在减少与像素级病灶分割相关的标注负担的同时，实现具有竞争力的诊断性能。评估了两个公开可用的卵巢超声数据集：用于八类分类的多模态卵巢肿瘤超声（MMOTU）数据集和用于二元分类的卵巢超声数据集（OUD）。在统一框架下比较了四种策略：基于全局图像的深度学习、基于病灶引导的ROI深度学习、基于病灶轮廓的深度学习，以及基于轮廓的放射组学与机器学习分类器。评估了四种深度学习架构：MaxViT-Tiny、Swin Transformer、EfficientNet-B7和ResNet18。使用支持向量机、k近邻和人工神经网络分类器开发了放射组学模型，并对样本较少的OUD数据集应用了基于ANOVA的特征选择。病灶引导的ROI策略实现了最强的整体性能，MaxViT-Tiny在MMOTU上获得了93.10%的准确率和0.99的AUC，在OUD上获得了97.56%的准确率和0.99的AUC。基于轮廓的方法实现了可比的准确性，但需要显著更高的标注工作量。这些发现表明，病灶引导的ROI深度学习在诊断性能和标注效率之间提供了有效的平衡，为可扩展的AI辅助卵巢超声分析提供了一种实用的方法。
View on arXiv Download PDF AI Translation
cs.CV / 82 / 2608.25970

PANDA - Prototype-Anchored Alignment for Partially Unpaired Multimodal Learning, with Applications to Alzheimers MRI and TCGA Pathology

PANDA - 原型锚定对齐用于部分未配对的多模态学习，及其在阿尔茨海默病MRI和TCGA病理学中的应用
Bhat, Sheethal, Chowdhury, Mahfuzur Rahman, Perez-Toro, Paula Andrea, Wunderlich, Stephan, Bharat, Rose Dawn, Bayer, Siming, Maier, Andreas
Abstract
Multimodal medical prediction often faces incomplete pairing: auxiliary modalities with complementary signal are available for only a subset of subjects (or none) and cannot be assumed at deployment. We introduce PANDA (Prototype Anchored Data Alignment), a two-stage framework that transfers auxiliary information to a primary-modality model without auxiliary inputs at inference. Stage 1 learns a shared embedding from the paired subset and estimates class prototypes from auxiliary modalities; Stage 2 trains the primary encoder on all subjects using cross-entropy plus alignment to the frozen prototypes. Because supervision is defined at the class-prototype level, PANDA accommodates arbitrary pairing rates, including zero subject overlap. We evaluate PANDA on two applications. On a 1,021-subject multi-scanner ADNI cohort, we perform AD/CN classification with three auxiliary modalities at distinct pairing rates: tabular scores (44.8%), FDG-PET (18.7%), and external handwriting kinematics (0% overlap). Relative to the same-backbone MRI-only baseline, PANDA attains AUC 0.868 +-0.020 (+7.9pp) and reduces 1.5T CN false positives by 24.3pp; on a fully trainable Conv5-FC3 backbone it reaches AUC 0.893 (best overall). A pairing-rate ablation shows that the joint anchor remains within seed noise from 75% to 5% pairing. On TCGA-Lung survival prediction from whole-slide images with RNA-seq as auxiliary data, PANDA improves over WSI-only on 2-year OS (AUC +3.5pp) and Cox PH (C-index +9.0pts) and outperforms full-fusion training, which underperforms WSI-only, while requiring no RNA at inference; wide confidence intervals on this smaller cohort keep the gains below conventional significance. Overall, PANDA provides a deployment-oriented mechanism for leveraging incomplete auxiliary modalities to improve primary-modality prediction.
Chinese Translation
多模态医学预测常常面临不完整配对的问题：辅助模态的互补信号仅对部分受试者（或没有）可用，且在部署时无法假设。我们提出了PANDA（原型锚定数据对齐），这是一个两阶段框架，能够在推理时将辅助信息转移到主要模态模型，而无需辅助输入。第一阶段从配对子集中学习共享嵌入，并从辅助模态中估计类别原型；第二阶段在所有受试者上训练主要编码器，使用交叉熵加上对冻结原型的对齐。由于监督是在类别原型级别定义的，PANDA能够适应任意配对率，包括零受试者重叠。我们在两个应用上评估PANDA。在一个包含1,021名受试者的多扫描仪ADNI队列中，我们在不同配对率下使用三种辅助模态进行AD/CN分类：表格分数（44.8%）、FDG-PET（18.7%）和外部手写运动学（0%重叠）。相较于相同骨干的仅MRI基线，PANDA达到了AUC 0.868 ± 0.020（+7.9个百分点），并将1.5T CN假阳性降低了24.3个百分点；在一个完全可训练的Conv5-FC3骨干上，它达到了AUC 0.893（最佳结果）。配对率消融实验表明，联合锚定在75%到5%的配对中保持在种子噪声范围内。在使用RNA-seq作为辅助数据的全切片图像进行TCGA肺部生存预测中，PANDA在2年OS上相较于仅WSI有所改善（AUC +3.5个百分点）以及Cox PH（C-index +9.0分），并且在推理时不需要RNA，尽管全融合训练的表现不如仅WSI，且在这个较小的队列中，宽置信区间使得增益低于传统显著性。总体而言，PANDA提供了一种面向部署的机制，以利用不完整的辅助模态来改善主要模态的预测。
View on arXiv Download PDF AI Translation
cs.CV / 83 / 2608.25981

FRAME: separating sampling variation from representational cause in medical imaging fairness

FRAME：在医学影像公平性中分离采样变异与表征原因
Lotfinia, Mahshad, Truhn, Daniel, Maier, Andreas, Arasteh, Soroosh Tayebi
Abstract
Subgroup performance differences are the standard evidence for fairness bias in medical imaging, and the usual response removes the demographic information that a model encodes. Here we introduce Fair-model Reference And Mechanism Evaluation (FRAME), a two-step framework for auditing such a claim. The first step derives a fair-model reference, the distribution of the difference under exact fairness at the observed subgroup sizes. In the second step, we test the remainder with two operators in representation space. One operator cannot change a within-group ranking by construction. Across 702,206 images and 36 encoders, the reference accounts for a median 41% of the reported race difference and 22% of the age difference. Injecting demographic decodability leaves the remainder unchanged, while entangling the group with the disease direction raises the race difference from 0.077 to 0.118. No intervention we tested changes the remainder more than a change of random seed does. Those interventions reduce a difference at the operating point and leave the within-group ranking difference at a median of 0.000. Applied to 89 differences in 9 published studies across 6 medical imaging modalities, the reference accounts for a median 25% of a rate difference and 70% of a difference in the area under the receiver operating characteristic curve. Image-text pretraining instead raises worst-group performance by about 0.05. Applying FRAME before choosing an intervention could distinguish differences that need a mechanistic explanation from differences compatible with sampling variation at the current cohort sizes.
Chinese Translation
亚组性能差异是医学影像公平性偏见的标准证据，通常的应对措施是去除模型编码的人口统计信息。在此，我们引入公平模型参考与机制评估（Fair-model Reference And Mechanism Evaluation，FRAME），这是一个用于审计此类主张的两步框架。第一步推导出公平模型参考，即在观察到的亚组规模下，精确公平条件下差异的分布。第二步，我们在表征空间中使用两个算子测试剩余部分。其中一个算子在构造上无法改变组内排名。在702,206张图像和36个编码器中，参考模型解释了报告的种族差异中41%的中位数和年龄差异中22%的中位数。注入人口统计可解码性不会改变剩余部分，而将组与疾病方向纠缠则将种族差异从0.077提高到0.118。我们测试的任何干预措施都没有比随机种子变化更能改变剩余部分。这些干预在操作点上减少了差异，并使组内排名差异的中位数保持在0.000。应用于6种医学影像模式中9项已发表研究的89个差异，参考模型解释了比率差异中25%的中位数和接收者操作特征曲线下面积差异中70%的中位数。图像-文本预训练反而使最差组的性能提高约0.05。在选择干预措施之前应用FRAME，可以区分需要机制解释的差异与在当前队列规模下与采样变异兼容的差异。
View on arXiv Download PDF AI Translation
cs.CV / 84 / 2608.25998

Uncertainty-Guided Latent Diffusion Models for Faithful Super Resolution

基于不确定性引导的潜在扩散模型用于真实感超分辨率
Wang, Ren, Chuang, Yung-Yu
Abstract
The perception-distortion trade-off poses a fundamental challenge in single-image super-resolution (SR). Although diffusion-based SR methods excel at generating perceptually realistic images, achieving high fidelity remains a key limitation. Recent advances in diffusion-based SR have shown promise in improving fidelity, but these methods often compromise perceptual quality due to their high reliance on a high-fidelity image. To address this, we introduce UGDiff, a novel diffusion guidance paradigm designed to further improve the perception-distortion balance. In particular, we first estimate the reconstruction uncertainty of the latent features corresponding to a high-fidelity image. This uncertainty is then used to guide the diffusion process to selectively restore high-frequency details in high-uncertainty regions, while preserving fidelity elsewhere. Furthermore, our guidance method adaptively identifies the high-uncertainty regions by considering not only the estimated uncertainty but also the posterior variance of the diffusion sampler at each timestep. This relaxes the reliance on the high-fidelity image in the later stages of sampling, thereby achieving a better perception-distortion balance. Extensive experimental results demonstrate that our method performs favorably against state-of-the-art diffusion-based SR methods.
Chinese Translation
感知-失真权衡在单幅图像超分辨率（SR）中构成了一个基本挑战。尽管基于扩散的超分辨率方法在生成感知上真实的图像方面表现出色，但实现高保真度仍然是一个关键限制。最近在基于扩散的超分辨率中的进展显示出改善保真度的潜力，但这些方法往往由于对高保真图像的高度依赖而妥协了感知质量。为了解决这个问题，我们提出了UGDiff，一种新颖的扩散引导范式，旨在进一步改善感知-失真平衡。具体而言，我们首先估计与高保真图像对应的潜在特征的重建不确定性。然后利用这种不确定性引导扩散过程，在高不确定性区域选择性地恢复高频细节，同时在其他区域保持保真度。此外，我们的引导方法通过考虑每个时间步的估计不确定性和扩散采样器的后验方差，自适应地识别高不确定性区域。这减轻了在采样后期对高保真图像的依赖，从而实现了更好的感知-失真平衡。大量实验结果表明，我们的方法在与最先进的基于扩散的超分辨率方法的比较中表现良好。
View on arXiv Download PDF AI Translation
cs.CV / 85 / 2608.26033

UltraPIPS: Improving model perception in B-mode ultrasound with foundation models

UltraPIPS：利用基础模型改善B模式超声中的模型感知
Grutman, Tal, Ilovitsh, Tali
Abstract
In medical imaging, it is common to use learned perceptual image patch similarity (LPIPS) to compare images semantically in feature space. Although backbones pretrained on natural images are widely used for LPIPS computation, B-mode ultrasound images possess distinct speckle patterns and acoustic-specific image statistics that are fundamentally different from natural images and even from other images in radiology. Consequently, we propose that domain-specific models are needed to measure perceptual similarity in ultrasound data, a finding which is not necessarily the case for other imaging modalities. We compare LPIPS metrics across downstream tasks like classification, segmentation and reconstruction using natural image, medical generalist and ultrasound backbone models and show that selection of LPIPS backbone is a non-trivial design choice. In particular, the ultrasound backbone models were more correlated with downstream performance of supervised models than classical and natural image models, and optimization of the LPIPS loss with an ultrasound backbone achieved a strong balance between reconstruction quality and realism. Our code is available at https://github.com/talg2324/UltraPIPS and introduces the UltraPIPS library, a set of LPIPS metrics based on the open-source foundation models analyzed in this paper.
Chinese Translation
在医学成像中，通常使用学习的感知图像块相似性（LPIPS）在特征空间中对图像进行语义比较。尽管在自然图像上预训练的主干网络被广泛用于LPIPS计算，但B模式超声图像具有独特的散斑模式和声学特定图像统计特征，这些特征与自然图像甚至其他放射学图像根本不同。因此，我们提出需要领域特定的模型来测量超声数据中的感知相似性，这一发现并不一定适用于其他成像模式。我们比较了在分类、分割和重建等下游任务中使用自然图像、医学通用模型和超声主干模型的LPIPS指标，并表明LPIPS主干的选择是一个非平凡的设计选择。特别是，超声主干模型与监督模型的下游性能相关性更高，而使用超声主干优化LPIPS损失在重建质量和真实感之间达成了良好的平衡。我们的代码可在https://github.com/talg2324/UltraPIPS获取，并介绍了UltraPIPS库，这是一个基于本文分析的开源基础模型的LPIPS指标集合。
View on arXiv Download PDF AI Translation
cs.CV / 86 / 2608.26067

StreamPI: Streaming Multimodal Temporal Modeling for Vision-Language-Action Models

StreamPI：用于视觉-语言-动作模型的流式多模态时间建模
Liu, Zhe, Hou, Jinghua, Lu, Yuxiang, Yang, Zhenya, Fan, Xianzhe, Luo, Junwei, Li, Junyi, Han, Ruihua, Hou, Zhi, Zhao, Hengshuang
Abstract
Vision-Language-Action (VLA) models have demonstrated effectiveness in robot manipulation, yet state-of-the-art models such as pi0.5 operate under a single-frame paradigm, limiting their ability to retain past observations and develop precise spatial perception. In this paper, we propose StreamPI, a streaming multimodal temporal modeling framework that equips single-frame VLA with temporal reasoning capability without introducing any additional parameters. One core design is instruction-anchored temporal modeling. It treats each (visual observation, language instruction) pair as an atomic temporal unit: bidirectional attention within each pair enables cross-modal fusion, while causal attention across pairs preserves autoregressive streaming inference. This ensures the language instruction serves as a persistent semantic anchor throughout task execution. To bridge the gap between synchronous training and asynchronous real-robot deployment, we introduce a andom-interval streaming training strategy: a proper inter-frame interval (e.g., every 3 frames) enables faster and smoother action execution. Beyond this, randomizing the interval further improves robustness to frame-timing perturbations, supporting asynchronous deployment in practice. Furthermore, by leveraging the length extrapolation capability of the LLM backbone, StreamPI seamlessly inherits pretrained single-frame weights and supports flexible single-frame and multi-frame inference. Experiments on real-robot tasks spanning memory-dependent and precise perception scenarios, as well as the simulation benchmark LIBERO, demonstrate that StreamPI outperforms pi0.5 across diverse tasks.
Chinese Translation
视觉-语言-动作（VLA）模型在机器人操作中表现出色，但诸如 pi0.5 的最先进模型仍然在单帧范式下运行，这限制了它们保留过去观察和发展精确空间感知的能力。本文提出了 StreamPI，一种流式多模态时间建模框架，使单帧 VLA 具备时间推理能力，而无需引入任何额外参数。其核心设计之一是基于指令的时间建模。它将每个（视觉观察，语言指令）对视为一个原子时间单元：每对内的双向注意力实现跨模态融合，而对之间的因果注意力则保持自回归流式推理。这确保了语言指令在任务执行过程中作为持久的语义锚点。为了弥合同步训练与异步真实机器人部署之间的差距，我们引入了一种随机间隔流式训练策略：适当的帧间隔（例如，每3帧）能够实现更快更流畅的动作执行。此外，随机化间隔进一步提高了对帧时序扰动的鲁棒性，支持实际中的异步部署。此外，通过利用 LLM 主干的长度外推能力，StreamPI 无缝继承了预训练的单帧权重，并支持灵活的单帧和多帧推理。在涉及依赖记忆和精确感知场景的真实机器人任务以及模拟基准 LIBERO 上的实验表明，StreamPI 在各种任务中均优于 pi0.5。
View on arXiv Download PDF AI Translation
cs.CV / 87 / 2608.26094

MyoMechanix: Biomechanically-Grounded Compositional Skilled Activity Understanding and Coaching

MyoMechanix：基于生物力学的组合技能活动理解与指导
Yin, Hao, Parmar, Paritosh, Gu, Lijun, Xu, Lin, Guo, Tianxiao, Liu, Xiujin, Zheng, Tianyou, Zhang, Yang, Fu, Weiwei
Abstract
Existing action quality assessment (AQA) datasets and methods rely primarily on visual inputs such as RGB and pose, overlooking physiological dynamics such as muscle mechanics and often modeling actions as monolithic patterns. These limitations hinder fine-grained, biomechanically grounded feedback. We introduce MyoMechanix, a multimodal ecosystem for weight-loaded actions that aligns motion with muscle activity. Expert-annotated, it contains 7,500+ samples of 20 actions from 38 subjects, with synchronized multiview RGB video, 3D pose, sEMG, and additional physiological signals, forming the largest multimodal AQA benchmark to date. We further construct the Fitness Knowledge Graph (FKG), which organizes expert annotations into structured relationships among actions, phases, key steps, errors, and corrective feedback, enabling compositional scoring and interpretable assessment. Building on these representations, we develop CUBIST (Compositional Ontological Reasoning Engine), which performs decomposition-analysis-recomposition for fine-grained error attribution and feedback generation. We also establish MyoMechanix-AQA, MyoMechanix-VideoQA, and a novel MyoMechanix-Video2EMG task. Experiments show that multimodal sensing and structured representations improve performance, interpretability, and error attribution, with CUBIST achieving state-of-the-art results; VideoQA enhances language-grounded action understanding; and Video2EMG suggests video-based alternatives to costly EMG sensing. MyoMechanix advances skilled activity understanding toward biomechanically grounded, multimodal, and compositional reasoning for Physical AI applications in fitness, rehabilitation, healthcare, and machine learning. Project page: https://haoyin116.github.io/MyoMechanix/
Chinese Translation
现有的动作质量评估（AQA）数据集和方法主要依赖于RGB和姿态等视觉输入，忽视了肌肉力学等生理动态，通常将动作建模为单一模式。这些局限性妨碍了细粒度的、生物力学基础的反馈。我们提出了MyoMechanix，一个多模态生态系统，专注于负重动作，将运动与肌肉活动相结合。该系统经过专家标注，包含来自38名受试者的20个动作的7500多个样本，配有同步的多视角RGB视频、3D姿态、表面肌电图（sEMG）和其他生理信号，形成迄今为止最大的多模态AQA基准。我们进一步构建了健身知识图谱（Fitness Knowledge Graph, FKG），将专家标注组织成动作、阶段、关键步骤、错误和纠正反馈之间的结构化关系，从而实现组合评分和可解释的评估。在这些表示的基础上，我们开发了CUBIST（组合本体推理引擎），该引擎执行细粒度的错误归因和反馈生成的分解-分析-重组。我们还建立了MyoMechanix-AQA、MyoMechanix-VideoQA以及一种新颖的MyoMechanix-Video2EMG任务。实验表明，多模态传感和结构化表示提高了性能、可解释性和错误归因，CUBIST实现了最先进的结果；VideoQA增强了基于语言的动作理解；而Video2EMG则提供了基于视频的替代方案，以取代昂贵的肌电图传感。MyoMechanix推动了对技能活动理解的进展，朝着生物力学基础的、多模态的和组合推理的方向发展，适用于健身、康复、医疗保健和机器学习等领域的物理人工智能应用。项目页面：https://haoyin116.github.io/MyoMechanix/
View on arXiv Download PDF AI Translation
cs.CV / 88 / 2608.26095

A Visual Dependence-Aware Framework for Multimodal Unsupervised Continual Post-Training

一种视觉依赖感知框架用于多模态无监督持续后训练
Li, Kaichen, Zhu, Zhilin, Huang, Jianhao, Lai, Zhengqin, Xiong, Baochen, Shao, Zibo, Song, Yaguang, Xiao, Linhui, Yang, Xiaoshan, Xu, Changsheng
Abstract
In this paper, we explore a novel task of Multimodal Unsupervised Continual Post-Training (MU-CPT), enabling deployed MLLMs to continually evolve from streaming unlabeled data. Existing unsupervised post-training methods for MLLMs typically optimize target tokens uniformly, overlooking their heterogeneous visual dependence (VD). However, we reveal that token-level VD is crucial for MU-CPT. Specifically, its structural distortion serves as an indicator of cross-modal catastrophic forgetting, and its inherent heterogeneity acts as a compass to guide new-task learning. Leveraging this property, we propose a Visual Dependence-Aware (VDA) framework with two main components. First, Visually Constrained Optimal Transport (VC-OT) formulates the VD structural distortion of old-task VD during new-task learning as an optimal transport problem to mitigate cross-modal forgetting. By designing a region-aware ground cost and a dependence-stratified transport penalty, it prevents global shifts in visual focus while strictly prohibiting visual reliance from degenerating into language bias. Second, Visually Modulated Adaptation (VMA) exploits VD heterogeneity to emphasize visually grounded new-task learning, promoting new-task plasticity. Together, our method simultaneously maintains old-task stability and new-task plasticity during challenging MU-CPT. Extensive experiments under our MU-CPT setting validate the effectiveness of VDA.
Chinese Translation
在本文中，我们探索了一种新颖的多模态无监督持续后训练（Multimodal Unsupervised Continual Post-Training, MU-CPT）任务，使得已部署的多语言大模型（MLLMs）能够从流式无标签数据中持续演化。现有的多语言大模型无监督后训练方法通常均匀地优化目标标记，忽视了其异质的视觉依赖（Visual Dependence, VD）。然而，我们揭示了标记级的视觉依赖对于MU-CPT至关重要。具体而言，其结构扭曲作为跨模态灾难性遗忘的指示器，而其内在的异质性则作为指导新任务学习的指南。利用这一特性，我们提出了一种视觉依赖感知（Visual Dependence-Aware, VDA）框架，包含两个主要组成部分。首先，视觉约束最优传输（Visually Constrained Optimal Transport, VC-OT）将新任务学习中旧任务视觉依赖的结构扭曲形式化为一个最优传输问题，以减轻跨模态遗忘。通过设计区域感知的基础成本和依赖分层的传输惩罚，它防止了视觉焦点的全球性转移，同时严格禁止视觉依赖退化为语言偏见。其次，视觉调制适应（Visually Modulated Adaptation, VMA）利用视觉依赖的异质性来强调以视觉为基础的新任务学习，促进新任务的可塑性。综合来看，我们的方法在具有挑战性的MU-CPT过程中，同时保持了旧任务的稳定性和新任务的可塑性。在我们的MU-CPT设置下，广泛的实验验证了VDA的有效性。
View on arXiv Download PDF AI Translation
cs.CV / 89 / 2608.26101

RefVideo-6M: A Reliable Reference-Based Dataset for Instructional Video Editing

RefVideo-6M：一个可靠的基于参考的教学视频编辑数据集
Zi, Bojia, Yang, Xiaoyan, Zhou, Yu, Sun, Ruijie, Zhang, Lihan, Liang, Bin, Wong, Kam-Fai, Huang, Haibin, Zhang, Chi, Li, Xuelong
Abstract
Recent advances in video editing have been largely driven by large-scale instruction-based datasets. However, existing datasets still suffer from two critical limitations. First, target videos are commonly produced by automatic editing models, which may introduce visible artifacts and unreliable supervision signals. Second, most public datasets rely primarily on textual instructions, while lacking visual references that are crucial for precise, identity-preserving, and controllable editing. To address these limitations, we introduce RefVideo-6M, a large-scale reference-guided editing dataset containing 5 million video editing samples and 1 million image editing samples. To ensure reliable supervision, our dataset uses a construction pipeline that treats artifact-free real videos as editing targets and generates quality-filtered input conditions with multiple editing experts. In addition, it provides approximately 6 million visual references, covering diverse reference types and editing scenarios, thereby enabling models to learn fine-grained visual correspondence beyond text-only instructions. Based on RefVideo-6M, we further train a reference-guided video editing model, Ref-MoT, to evaluate the effectiveness and scalability of the proposed dataset. Extensive experiments demonstrate that RefVideo-6M provides substantially more reliable supervision than existing datasets and enables the training of powerful editing models with improved visual quality, controllability, and reference consistency. The open-source dataset is available at https://huggingface.co/datasets/RefVideo6M/RefVideo6M.
Chinese Translation
近年来，视频编辑的进展主要受到大规模基于指令的数据集的推动。然而，现有数据集仍然存在两个关键限制。首先，目标视频通常由自动编辑模型生成，这可能引入可见的伪影和不可靠的监督信号。其次，大多数公共数据集主要依赖文本指令，而缺乏对精确、保持身份和可控编辑至关重要的视觉参考。为了解决这些限制，我们引入了RefVideo-6M，一个大规模的参考引导编辑数据集，包含500万个视频编辑样本和100万个图像编辑样本。为了确保可靠的监督，我们的数据集使用了一种构建流程，将无伪影的真实视频视为编辑目标，并通过多个编辑专家生成质量过滤的输入条件。此外，它提供了约600万个视觉参考，涵盖多种参考类型和编辑场景，从而使模型能够学习超越仅文本指令的细粒度视觉对应关系。基于RefVideo-6M，我们进一步训练了一个参考引导的视频编辑模型Ref-MoT，以评估所提数据集的有效性和可扩展性。大量实验表明，RefVideo-6M提供的监督比现有数据集更为可靠，并能够训练出具有更高视觉质量、可控性和参考一致性的强大编辑模型。该开源数据集可在 https://huggingface.co/datasets/RefVideo6M/RefVideo6M 获取。
View on arXiv Download PDF AI Translation
cs.CV / 90 / 2608.26105

VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoning

VBVR-Pro：一个可扩展且可验证的原生视觉推理套件
Xu, Junxiang, Wang, Ruisi, Pu, Fanyi, Wang, Maijunxian, Ji, Ran, Zhou, Tongxi, Gu, Chenyang, Zuo, Jing, Xiao, Hongcan, Geng, Yimeng, Yin, Wanqi, Chen, Wei, Qian, Oscar, Yan, Zhengan, Huang, Ziqi, Diao, Haiwen, Pan, Liang, Li, Bo, Fan, Xiangyu, Luo, Dezhi, Yu, Fengyuan, Zhao, Zehong, Gao, Qingying, Zhu, Tinghui, Zhang, Yilan, Tong, Jingqi, Feng, Pinyuan, Jiang, Zhengze, Wang, Letian, Guo, Ziyu, Zhang, Renrui, Chen, Jieneng, Joseph, Sonia, Venhoff, Constantin, Motamed, Saman, Yang, Mengyue, Sripada, Chandra, Yuille, Alan, Torr, Philip, Zhang, Lvmin, Kumar, Vikash, Khashabi, Daniel, Kriegeskorte, Nikolaus, Millière, Raphaël, Müller, Vincent C., Rao, Anyi, Wang, Quan, Liu, Ziwei, Lin, Dahua, Yang, Lei, Deng, Hokin, Cai, Zhongang
Abstract
Native visual reasoning treats visual generation as the medium of reasoning itself: visual states (i.e. images and videos) are not merely inputs to be understood or outputs to be rendered, but first-class substrates for problem solving beyond language. Yet progress remains bottlenecked by the lack of scalable training tasks, reliable feedback, and controlled comparisons across generative substrates. In this work, we introduce VBVR-Pro, a closed-loop testbed that makes native visual reasoning through generation trainable, verifiable, optimizable, and experimentally controllable. 1) Task scaling. VBVR-Pro turns visual reasoning into a controlled task space of 300 procedurally generated tasks. Models trained on VBVR-Pro show strong transfer beyond the proposed suite across seven external visual reasoning benchmarks such as RISE-Video, MME-CoF-Pro, and BabyVision. 2) Verifiable rewards. VBVR-Pro provides verifiable reward scorers for task-grounded evaluation. Through a systematic study of leading MLLMs as judges, we identify recurring failure modes of the prevalent VLM-as-a-judge paradigm. In contrast, the proposed scorers are grounded in deterministic, task-specific rules, achieve fine-grained alignment with human judgments. Importantly, they serve as reliable reward signals for large-scale multi-task reinforcement learning and demonstrate stronger post-RL performance across visual reasoning tasks. 3) Mechanism study. VBVR-Pro enables controlled modality studies across more than 30 image, video, and interleaved generators. Our analysis shows that video generation remains strongest for tasks requiring persistent spatiotemporal state tracking, while interleaved generation provides a compute-efficient alternative. Critically, ablations and probing suggest the presence of vision-native trajectories that are crucial to visual reasoning. We release all data, models, scorers, and code.
Chinese Translation
原生视觉推理将视觉生成视为推理本身的媒介：视觉状态（即图像和视频）不仅仅是需要理解的输入或需要呈现的输出，而是超越语言的问题解决的第一类基底。然而，进展仍然受到可扩展训练任务、可靠反馈和跨生成基底的受控比较缺乏的瓶颈。在本研究中，我们介绍了VBVR-Pro，一个闭环测试平台，使得通过生成进行原生视觉推理变得可训练、可验证、可优化和可实验控制。1）任务扩展。VBVR-Pro将视觉推理转变为一个受控的任务空间，包含300个程序生成的任务。在VBVR-Pro上训练的模型在七个外部视觉推理基准（如RISE-Video、MME-CoF-Pro和BabyVision）上表现出强大的迁移能力。2）可验证的奖励。VBVR-Pro为任务基础的评估提供可验证的奖励评分器。通过对领先的多模态大模型（MLLMs）作为评判者的系统研究，我们识别出当前VLM作为评判者范式的反复失败模式。相比之下，所提出的评分器基于确定性的、任务特定的规则，能够与人类判断实现细粒度对齐。重要的是，它们为大规模多任务强化学习提供了可靠的奖励信号，并在视觉推理任务中展示了更强的后强化学习表现。3）机制研究。VBVR-Pro使得对30多个图像、视频和交错生成器的受控模态研究成为可能。我们的分析表明，视频生成在需要持续时空状态跟踪的任务中表现最强，而交错生成提供了一种计算高效的替代方案。关键的消融和探测结果表明，存在对视觉推理至关重要的视觉原生轨迹。我们发布所有数据、模型、评分器和代码。
View on arXiv Download PDF AI Translation
人工智能 (Artificial Intelligence)
55
cs.AI / 1 / 2608.24886

VLM-based automatic multi-granularity graph representation of building layouts for design informatics

基于视觉语言模型的建筑布局自动多粒度图表示用于设计信息学
Guo, Song, Chen, Zhuoshi, Li, Maosu, Zhuang, Weimin
Abstract
Architectural floorplan images encode rich relational knowledge among functional spaces, which underpins design retrieval, knowledge-based reasoning, and BIM enrichment through the building lifecycle. However, it remains challenging to automatically construct task-adaptive graph representations for public buildings. To address this gap, we first define a multi-granularity Level-of-Graphs (LoGs) for public building layouts. Methodologically, we present a Vision-Language Model (VLM)-based automatic LoG construction through node identification, edge inference, text parsing, and graph coarsening. VLM-generated representations are systematically evaluated and tested in real-world tasks, using 147 academic library floorplans worldwide as a case study. Experiments showed VLM-generated graphs were broadly consistent with human-labeled graphs (matched node ratio >= 92%; 509.3 s per floor plan for three-LoG graph generation). Meso-grained graphs yield the best node-level zone prediction (Macro F1 = 0.647, at 65% of fine-grained complexity), while coarse-grained graphs are most effective for graph-level layout quality evaluation (Spearman's \r{ho} = 0.610, at 16% of fine-grained complexity). By enabling scalable, annotation-free extraction of structured layout information from floorplan images, this study advances design informatics by converting plan images into knowledge representations, thereby enhancing the utilization of design information across the building life cycle.
Chinese Translation
建筑平面图像编码了功能空间之间丰富的关系知识，这为设计检索、基于知识的推理以及建筑生命周期中的建筑信息模型（BIM）丰富提供了基础。然而，自动构建适应任务的公共建筑图表示仍然具有挑战性。为了解决这一问题，我们首先为公共建筑布局定义了多粒度图层次（Level-of-Graphs, LoGs）。在方法论上，我们提出了一种基于视觉语言模型（Vision-Language Model, VLM）的自动LoG构建方法，通过节点识别、边推断、文本解析和图粗化实现。VLM生成的表示在真实世界任务中进行了系统评估和测试，以147个全球学术图书馆的平面图作为案例研究。实验表明，VLM生成的图与人工标注的图在广泛上是一致的（匹配节点比例 >= 92%；每个平面图生成三层LoG所需时间为509.3秒）。中粒度图在节点级区域预测中表现最佳（宏观F1 = 0.647，细粒度复杂度的65%时），而粗粒度图在图级布局质量评估中最为有效（斯皮尔曼相关系数 {ho} = 0.610，细粒度复杂度的16%时）。通过实现可扩展的、无注释的结构化布局信息提取，本研究通过将平面图像转换为知识表示，推动了设计信息学的发展，从而增强了建筑生命周期中设计信息的利用。
View on arXiv Download PDF AI Translation
cs.AI / 2 / 2608.24888

SIMGUIDE: Procedurally Grounded Multi-Context Representations for Personalized Agent Planning

SIMGUIDE：基于程序化的多上下文表示用于个性化代理规划
Shah, Chirag
Abstract
Personalized AI agents overwhelmingly treat users as single entities: a flat profile concatenated into a prompt. This fails when the same person holds different priorities across life contexts -- and fails catastrophically when those priorities conflict. The core problem is not that agents lack information about users; it is that the format of user representations determines whether an agent can act on that information at all. We introduce SIMGUIDE, a method that structures user context into typed, domain-specific blocks called Sims and grounds each constraint with procedural examples drawn from past decisions. To evaluate this, we construct SIMBENCH, a diagnostic suite of 47 preference-conditioned planning tasks where the correct plan depends on which user context is active -- a property no existing benchmark tests. Declarative Sim constraints alone do not outperform retrieval-based personalization (RAG). Procedurally grounded Sims outperform RAG on GPT-4o (+7.9 Preference Adherence points, $p = 0.013$), and this advantage replicates on 100 $\tau$-bench tasks across both GPT-4o and Claude Sonnet~4.5 ($p \leq 0.023$). At the parametric level, the same principle holds: training distribution dominates whether parametric adaptation succeeds at all. Task-matched LoRA fine-tuning improves generation quality by 12.8 ROUGE-L points over the unadapted base model, and routing adapters by Sim type rather than user identity adds a further 7.3 points, robust to 28% routing error. Representation format -- not representation content -- is the first-order design variable.
Chinese Translation
个性化人工智能代理通常将用户视为单一实体：一个扁平的个人资料被连接到提示中。当同一个人在不同生活上下文中持有不同优先事项时，这种方法就会失效——当这些优先事项发生冲突时，失效的后果更为严重。核心问题并不是代理缺乏关于用户的信息，而是用户表示的格式决定了代理是否能够对这些信息采取行动。我们提出了SIMGUIDE，这是一种将用户上下文结构化为称为Sims的类型化、领域特定的块的方法，并通过从过去决策中提取的程序示例为每个约束提供基础。为了评估这一方法，我们构建了SIMBENCH，这是一个包含47个偏好条件规划任务的诊断套件，其中正确的计划取决于哪个用户上下文处于活动状态——这是现有基准测试所不具备的属性。仅仅使用声明式的Sim约束并未优于基于检索的个性化（RAG）。基于程序化的Sims在GPT-4o上优于RAG（+7.9偏好遵循点，$p = 0.013$），这一优势在100个$ au$-bench任务中得到了复制，无论是在GPT-4o还是Claude Sonnet~4.5上（$p leq 0.023$）。在参数级别上，同样的原则适用：训练分布主导着参数适应是否成功。任务匹配的LoRA微调使生成质量比未适应的基础模型提高了12.8 ROUGE-L点，而按Sim类型而非用户身份进行路由适配又增加了7.3点，且对28%的路由错误具有鲁棒性。表示格式——而非表示内容——是第一优先的设计变量。
View on arXiv Download PDF AI Translation
cs.AI / 3 / 2608.24889

Reliable LLM-Powered Decision Engines for Large-Scale Supply Chain Operations: Architecture, Safety, and Performance Guarantees

可靠的基于大型语言模型的决策引擎在大规模供应链运营中的应用：架构、安全性与性能保障
Jingar, Nirmal Kumar
Abstract
Current large-scale supply chains are highly uncertain, dynamic, and disruption prone that are challenging to serve up timely and resilient decisions through traditional rule-based and optimization-only systems. The increasing supply of heterogeneous data sources, such as transactional demand signals and unstructured disruption report, presents a chance of intelligent systems, which could reason, adapt and optimize at the same time. A hybrid architecture that combines large language models (LLMs) with mathematical optimization, probabilistic forecasting, and safety-constrained decision filtering is proposed in this paper as a performance of a Decision Engine, which is called LLM-Powered Decision Engine (LLM-DE). In comparison to purely data-driven or heuristic solutions, LLM-DE integrates semantic reasoning with LLM with a set of performance and safety guarantees that allow safe decision-making in large-scale supply chain processes. The suggested framework enables the end-to-end decision making such as demand forecasting, inventory optimization, and transportation routing and disruption mitigation. The findings affirm that language-based reasoning combined with optimization and formal constraints can be used to come up with not only smarter but also safer and more scalable supply chain decisions. This research provides a new architecture, a complete pipeline of algorithm, and a formulation based on mathematical constructs of the operational decision systems incorporating LLM. The proposed model offers a pragmatic and theoretical basis of the next-generation intelligent supply chain infrastructures that can be implemented to work dependably in the face of uncertainty and massive complexity.
Chinese Translation
当前的大规模供应链面临高度的不确定性、动态性和易受干扰的挑战，传统的基于规则和仅优化的系统难以提供及时且具有韧性的决策。异构数据源的不断增加，例如交易需求信号和非结构化的干扰报告，为智能系统提供了机会，使其能够同时进行推理、适应和优化。本文提出了一种混合架构，将大型语言模型（LLMs）与数学优化、概率预测和安全约束决策过滤相结合，作为决策引擎的性能表现，称为LLM驱动的决策引擎（LLM-DE）。与纯粹的数据驱动或启发式解决方案相比，LLM-DE将语义推理与LLM整合，提供一系列性能和安全保障，允许在大规模供应链过程中进行安全决策。所建议的框架支持端到端的决策制定，如需求预测、库存优化和运输路线规划及干扰缓解。研究结果确认，基于语言的推理结合优化和形式约束不仅可以得出更智能的决策，还能确保供应链决策的安全性和可扩展性。本研究提供了一种新的架构、完整的算法管道以及基于数学构造的操作决策系统的公式，结合了LLM。所提出的模型为下一代智能供应链基础设施提供了务实和理论基础，能够在不确定性和巨大复杂性面前可靠地运作。
View on arXiv Download PDF AI Translation
cs.AI / 4 / 2608.24890

AI-Powered Mental Health Chatbots in Africa: A Systematic Review and Culturally Adaptive Framework

非洲的人工智能驱动心理健康聊天机器人：系统评估与文化适应框架
Lebese, Matshepo, Tsibolane, Pitso
Abstract
Mental health challenges in Africa remain under-addressed due to inadequate infrastructure, stigma, and a chronic shortage of professionals. Artificial Intelligence (AI)-powered chatbots are emerging globally as low-cost, accessible tools that can offer psychological support. This paper presents a systematic review of 52 empirical studies published between 2017 and 2025, critically analysing their cultural, linguistic, and infrastructural relevance to African contexts. The findings demonstrate the potential of AI chatbots to improve accessibility, reduce symptoms of anxiety and depression, and expand psychosocial support, yet reveal limited African-specific adaptation. Most systems remain rooted in Western models and English-language designs, leaving critical gaps in local relevance, inclusivity, and sustainability. Based on the findings, the authors develop a Culturally Adaptive Digital Mental Health (CADMH) framework that integrates African cultural values, multilingual design, mobile-first optimisation, and ethical safeguards. The study highlights opportunities and barriers for integrating AI chatbots into African healthcare, offering guidance for research, practice, and policy.
Chinese Translation
非洲的心理健康挑战由于基础设施不足、污名化和专业人员的长期短缺而未得到充分解决。人工智能（AI）驱动的聊天机器人作为低成本、可获取的工具，正在全球范围内崭露头角，能够提供心理支持。本文对2017年至2025年间发表的52项实证研究进行了系统评估，批判性地分析了它们在非洲背景下的文化、语言和基础设施相关性。研究结果表明，AI聊天机器人在改善可获取性、减轻焦虑和抑郁症状以及扩展心理社会支持方面具有潜力，但显示出对非洲特定适应的有限性。大多数系统仍然根植于西方模型和英语设计，导致在地方相关性、包容性和可持续性方面存在重要缺口。基于研究结果，作者提出了一个文化适应数字心理健康（Culturally Adaptive Digital Mental Health, CADMH）框架，该框架整合了非洲文化价值观、多语言设计、移动优先优化和伦理保障。该研究强调了将AI聊天机器人融入非洲医疗保健的机遇与障碍，为研究、实践和政策提供了指导。
View on arXiv Download PDF AI Translation
cs.AI / 5 / 2608.24891

Measurement-Budget Allocation in Quantum Learning with Finite-Shot Generalization Guarantees

具有有限样本泛化保证的量子学习中的测量预算分配
Catak, Ferhat Ozgur
Abstract
On near-term quantum hardware, estimating a Born probability requires repeated circuit executions. A quantum learning experiment with a fixed measurement budget $B$ must therefore decide how many distinct training states $n$ to use and how many shots $S$ to allocate to each state. We study this tradeoff for binary quantum classifiers with fixed or independently selected measurement operators $M$, where the ideal score is $\Tr(M\rho)$. We prove a distribution-free generalization bound that separates the finite-sample and finite-shot contributions. The sample term scales as $\sqrt{d/n}$, while the shot term scales as $\sqrt{(\log n)/S}$; under the constraint $B=nS$, these two terms move in opposite directions. Minimising a conservative closed-form surrogate of the bound gives the allocation rule $\nstar = 2\sqrt{2dB/\log(2B/\delta)}$ and $\Sstar = B/\nstar$. This surrogate has the same asymptotic scaling as the exact minimizer and yields a worst-case rate of $B^{-1/4}$. The guarantee is intentionally conservative, since it applies to the full class of binary quantum measurements. We complement the theory with PennyLane simulations using 2-qubit and 4-qubit variational quantum circuits on nine synthetic binary classification benchmarks. In all tested configurations, the one-sided empirical generalization gap remains below the theoretical bound. The result provides a conservative statistical guideline for allocating measurement budgets in finite-shot evaluation and pre-experimental planning for near-term quantum learning systems, complementing hardware-level scheduling and circuit-design considerations. Extending the guarantee to fully adaptive shot-noisy training remains an open problem.
Chinese Translation
在近期的量子硬件上，估计Born概率需要重复执行电路。因此，具有固定测量预算$B$的量子学习实验必须决定使用多少个不同的训练状态$n$以及为每个状态分配多少次测量$S$。我们研究了这一权衡，针对具有固定或独立选择测量算子$M$的二元量子分类器，其中理想得分为$ ext{Tr}(M ho)$。我们证明了一个无分布的泛化界限，该界限将有限样本和有限测量的贡献分开。样本项的规模为$ rac{ ext{sqrt}(d/n)}$，而测量项的规模为$ rac{ ext{sqrt}(( ext{log} n)/S)}$；在约束条件$B=nS$下，这两个项的变化方向相反。最小化保守的封闭形式界限的替代物给出了分配规则$ star = 2 ext{sqrt}(2dB/ ext{log}(2B/ ext{delta}))$和$ ext{Sstar} = B/ star$。该替代物具有与精确最小化器相同的渐近缩放，并且在最坏情况下的速率为$B^{-1/4}$。该保证是故意保守的，因为它适用于全类的二元量子测量。我们通过使用PennyLane在九个合成二元分类基准上进行2量子比特和4量子比特变分量子电路的模拟来补充理论。在所有测试的配置中，单侧经验泛化间隙始终低于理论界限。该结果为在有限测量评估和近期量子学习系统的预实验规划中分配测量预算提供了保守的统计指导，补充了硬件级调度和电路设计的考虑。将保证扩展到完全自适应的测量噪声训练仍然是一个未解决的问题。
View on arXiv Download PDF AI Translation
cs.AI / 6 / 2608.24892

Natural Language Input, Semantic Track Representation, and LLM Inference: Making the Maritime Information Exchange Model Tractable

自然语言输入、语义轨迹表示与大语言模型推理：使海事信息交换模型可操作化
Roth, Frederick
Abstract
We describe a practical architecture for making the Maritime Information Exchange Model (MIEM) and the broader Rich Semantic Track model tractable using current large language model (LLM) technology. The barrier to adoption of semantic track models in defense and law enforcement has been the requirement that operators learn formal ontology languages and manually encode observations as typed logical assertions. We propose eliminating this barrier entirely: operators contribute observations in natural language; an LLM translates these into typed Semantic Assertion Records (SARs), which are named case frames that capture n-ary relations in a single compact structure; a knowledge graph accumulates the SARs; and a second LLM pass performs inference, anomaly detection, and hypothesis ranking over the graph. We work through two detailed examples (a 9/11-era pre-attack indicator scenario and a maritime cargo inspection scenario) showing the full pipeline from natural language input to SAR representation to inference output. We argue that this architecture makes the Track Model and MIEM immediately deployable with current technology, establishes prior art against proprietary enclosure of the approach, and grounds the method in a theoretical framework connecting semantic track representations to neural manifold geometry.
Chinese Translation
我们描述了一种实用架构，以利用当前的大语言模型（LLM）技术，使海事信息交换模型（MIEM）和更广泛的丰富语义轨迹模型变得可操作。语义轨迹模型在国防和执法领域的采用障碍在于操作员需要学习正式本体语言并手动将观察结果编码为类型化的逻辑断言。我们提议完全消除这一障碍：操作员以自然语言贡献观察结果；LLM将这些转换为类型化的语义断言记录（SARs），这些记录是捕捉n元关系的命名案例框架，具有单一紧凑结构；知识图谱积累SARs；第二个LLM的处理对图谱进行推理、异常检测和假设排序。我们通过两个详细示例（一个是9/11时代的预攻击指示场景，另一个是海事货物检查场景）展示了从自然语言输入到SAR表示再到推理输出的完整流程。我们认为，这种架构使轨迹模型和MIEM能够立即在当前技术下部署，确立了针对该方法专有封闭的先前艺术，并将该方法基于一个理论框架，连接语义轨迹表示与神经流形几何。
View on arXiv Download PDF AI Translation
cs.AI / 7 / 2608.24893

Same-Player Verification for Account Consistency in Counter-Strike 2

反恐精英2中的同玩家验证以确保账户一致性
Zhang, Xuchen
Abstract
In competitive first-person shooter (FPS) games such as Counter-Strike 2 (CS2), account-integrity review often asks whether an account's recent behavior remains consistent with its historical operator. This consistency question arises in cases such as temporary substitution, rank boosting, and high-skill players using lower-ranked accounts, where manual review requires comparing a current match against multiple historical matches. We formulate this review task as same-player verification: we encode the behavioral trajectory of a single player in a match replay (demo) as a demo-player behavioral fingerprint, and train a model to judge whether two behavioral observations come from the same real player. Grounded in CS2 game understanding, the fingerprints cover crosshair control, movement-stop-fire coordination, economy/buy, combat/engagement, and temporal rhythm. From 1,330 CS2 demos we extract 13,300 demo-player observations, and sample 663,590 same/different pairs from an 88.4M candidate-pair space for supervised training and evaluation. The final pairwise model reaches an average ROC AUC of 0.931 and achieves 0.722 different-player recall at 95% precision. Feature analysis shows that the strongest identity signals come from low-level operations, especially crosshair control, firing rhythm, and movement-stop-fire coordination, indicating that stable low-level mechanical habits are more informative for this verification task than single-match performance outcomes. In the account-history aggregation evaluation, increasing history depth raises AUC from the K=1 single-pair baseline of 0.931 to 0.986 at K=10. These results show that CS2 demo behavior can support supervised same-player verification and account-level identity-consistency modeling through multi-demo history aggregation.
Chinese Translation
在反恐精英2（Counter-Strike 2, CS2）等竞争性第一人称射击（FPS）游戏中，账户完整性审查通常会询问一个账户的近期行为是否与其历史操作员一致。这种一致性问题出现在临时替换、段位提升以及高技能玩家使用低段位账户等情况中，手动审查需要将当前比赛与多个历史比赛进行比较。我们将这一审查任务形式化为同玩家验证：我们将单个玩家在比赛回放（demo）中的行为轨迹编码为demo玩家行为指纹，并训练模型判断两个行为观察是否来自同一真实玩家。基于对CS2游戏的理解，这些指纹涵盖了瞄准控制、移动-停止-射击协调、经济/购买、战斗/交战和时间节奏等方面。从1,330个CS2 demo中，我们提取了13,300个demo玩家观察，并从8840万个候选对中抽样663,590个同/不同对进行监督训练和评估。最终的成对模型达到了平均ROC AUC值为0.931，并在95%精度下实现了0.722的不同玩家召回率。特征分析表明，最强的身份信号来自低级操作，尤其是瞄准控制、射击节奏和移动-停止-射击协调，这表明稳定的低级机械习惯对于这一验证任务比单场比赛的表现结果更具信息量。在账户历史聚合评估中，增加历史深度将AUC从K=1的单对基线0.931提高到K=10的0.986。这些结果表明，CS2 demo行为可以通过多demo历史聚合支持监督的同玩家验证和账户级身份一致性建模。
View on arXiv Download PDF AI Translation
cs.AI / 8 / 2608.24918

Semantic Graph Unification for Industrial Digital Threads: Bridging 11 Heterogeneous Manufacturing Systems Through Ontology-Driven Knowledge Graphs

工业数字线程的语义图统一：通过本体驱动的知识图谱连接11个异构制造系统
Chethan, Grama
Abstract
Modern manufacturing enterprises operate heterogeneous systems -- ERP, MES, PLM, SCADA, QMS, SCM -- each with its own data model and API. The resulting silos prevent holistic analysis, delay root-cause investigation, and obstruct Industry 4.0 traceability. Point-to-point integration scales as O(n^2) and accumulates brittle dependencies. This paper presents an open framework for semantic graph unification of industrial digital threads. An ontology-driven RDF knowledge graph unifies data from 11 simulated sources across nine domains through a five-stage ETL pipeline with automated entity resolution spanning 97 owl:sameAs identity links. The ontology encompasses 78 RDFS classes, 108 object properties, and 243 data properties, drawing on ISA-95, OPC UA, eClass, the Asset Administration Shell, RAMI 4.0, and additional standards. An automated discovery engine applies nine strategy categories -- cross-station correlation, alarm coverage, ECN impact, CUSUM/EWMA drift detection -- to surface insights spanning system boundaries. The primary empirical result: blocking 24 cross-system tools reduces recall from 1.00 to 0.31 (F1 from 1.00 to 0.48), showing that 69% of discoverable signals require cross-system graph joins. Leave-one-out ablation confirms six of nine strategies contribute unique signals. Verification against a 65-signal manifest (16 positive, 49 null) yields F1 = 1.00 (95% Clopper-Pearson CI [0.79, 1.00]); as the manifest was author-constructed, this constitutes verification not independent validation. The graph is exposed to LLM agents via 287 Model Context Protocol tools as a SPARQL-native semantic layer. Five industry templates (aerospace, CPG, pharma, medical devices, turbine blades) demonstrate schema stability across manufacturing verticals.
Chinese Translation
现代制造企业运营着异构系统——ERP、MES、PLM、SCADA、QMS、SCM——每个系统都有其独特的数据模型和API。这些孤岛阻碍了整体分析，延迟了根本原因调查，并妨碍了工业4.0的可追溯性。点对点集成的复杂度为O(n^2)，并积累了脆弱的依赖关系。本文提出了一个开放框架，用于工业数字线程的语义图统一。一个基于本体的RDF知识图谱通过一个五阶段的ETL管道统一来自11个模拟源的数据，涵盖九个领域，并通过97个owl:sameAs身份链接实现自动实体解析。该本体包含78个RDFS类、108个对象属性和243个数据属性，借鉴了ISA-95、OPC UA、eClass、资产管理壳、RAMI 4.0及其他标准。一个自动发现引擎应用九种策略类别——跨站点关联、报警覆盖、ECN影响、CUSUM/EWMA漂移检测——以揭示跨系统边界的洞察。主要的实证结果是：阻止24个跨系统工具将召回率从1.00降低至0.31（F1从1.00降低至0.48），显示69%的可发现信号需要跨系统图连接。留一法消融验证确认九种策略中的六种贡献了独特信号。与65个信号清单（16个正信号，49个无信号）的验证结果为F1 = 1.00（95% Clopper-Pearson CI [0.79, 1.00]）；由于该清单由作者构建，因此这构成了验证而非独立验证。该图通过287个模型上下文协议工具以SPARQL原生语义层的形式暴露给LLM代理。五个行业模板（航空航天、消费品、制药、医疗设备、涡轮叶片）展示了制造垂直行业间的模式稳定性。
View on arXiv Download PDF AI Translation
cs.AI / 9 / 2608.24921

post-graph-rag: A PostgreSQL-Native Graph RAG Engine

post-graph-rag：一种原生于PostgreSQL的图形RAG引擎
Rajah, Chandan
Abstract
Graph-based retrieval-augmented generation connects facts that no single passage states, but current implementations pay for that three times: in infrastructure, requiring a vector store, graph database and document store to be kept consistent; in graph quality, because an extraction pipeline that never refuses output fills the graph with edges that assert nothing; and over time, because a graph that only accumulates treats superseded and current facts alike. post-graph-rag is an open-source engine addressing all three. Text chunks with embeddings, a canonical entity graph and community summaries live in one PostgreSQL database, with pgvector for search and edge tables for traversal. Extraction-time invariants run before anything is written: vague predicates, pronominal names and bare quantities are rejected; predicates are normalised onto an optional vocabulary; entities resolve to one vertex per canonical name via model-supplied aliases; and denied relations keep the positive predicate under a negation flag. A temporal layer lets relations carry a validity period from the prose, lets a later document supersede an earlier incompatible assertion from document order alone, and answers as-of queries. Against LightRAG on three corpora under identical extraction and embedding models, post-graph-rag builds a denser graph everywhere, up to $2.4\times$ the relations per entity, and a more queryable one: distinct edge labels run at 0.46 to 0.58 per relation, 0.11 under a controlled vocabulary, against 0.77 to 1.33. It answers comparably with lower query latency, and supports temporal evolution the baseline lacks: 13 and 8 relationships superseded on a novel sequence and a decade of filings, against zero. These are engineering measurements, not a benchmark result. Code: https://github.com/crajah/post-graph-rag, https://github.com/crajah/post-graph
Chinese Translation
基于图的检索增强生成（retrieval-augmented generation）连接了单一段落无法陈述的事实，但当前的实现方式在三个方面付出了代价：在基础设施方面，需要保持向量存储、图数据库和文档存储的一致性；在图的质量方面，由于提取管道从不拒绝输出，导致图中充满了没有任何断言的边；而随着时间的推移，单纯积累的图将过时的事实与当前事实视为相同。post-graph-rag是一个开源引擎，旨在解决这三个问题。文本块与嵌入、标准实体图和社区摘要存储在一个PostgreSQL数据库中，使用pgvector进行搜索，边表用于遍历。在写入任何内容之前，提取时的不变性会运行：模糊的谓词、代名词和裸数量会被拒绝；谓词会被规范化到一个可选的词汇表；实体通过模型提供的别名解析为每个标准名称的一个顶点；被拒绝的关系会将正谓词保持在否定标志下。时间层允许关系携带来自散文的有效期，让后续文档仅凭文档顺序取代早期不兼容的断言，并回答特定时间点的查询。在相同的提取和嵌入模型下，与LightRAG在三个语料库上的比较中，post-graph-rag在每个地方构建了更密集的图，实体的关系数量最高可达$2.4 imes$，并且查询性能更佳：不同的边标签在每个关系中运行在0.46到0.58之间，在受控词汇下为0.11，而对照组为0.77到1.33。它以较低的查询延迟提供了相似的回答，并支持基线缺乏的时间演变：在一个新序列和十年的文件中，分别取代了13个和8个关系，而对照组为零。这些是工程测量，而非基准测试结果。代码链接： https://github.com/crajah/post-graph-rag, https://github.com/crajah/post-graph
View on arXiv Download PDF AI Translation
cs.AI / 10 / 2608.24979

FrontierChallenge: Evaluating Scientific Workflow Completion

FrontierChallenge：评估科学工作流的完成情况
Su, Liangcai, Feng, Zhaopeng, Chen, Zhuo, Zhang, Zhen, Lin, Xiang, Li, Ruilin, Zhang, Handuo, Wang, Ning, Wen, Kailong, Guo, Yueqi, Xing, Feng, Guo, Yiling, Qian, Chenxiong, Du, Simon Shaolei, Bing, Lidong, Wang, Xinyu
Abstract
Scientific agents increasingly analyze data, execute code, and produce research artifacts, yet most benchmarks emphasize final answers, isolated programs, or a single domain. We introduce FrontierChallenge, a cross-domain benchmark comprising 300 end-to-end scientific workflows. In this paper, we release and evaluate 97 of these tasks, spanning quantum chemistry, molecular dynamics, materials characterization, analytical chemistry, life science, and electrochemistry/environment. Each task provides fixed inputs and specifies a bundle of required scientific deliverables. We evaluate twelve frontier models with three agent scaffolds. Pass Rate measures the fraction of tasks satisfying the full-completion criterion, while Avg. Score captures partial progress. Each of the best-performing configurations completed only 20 of the 97 released tasks, yielding a Pass Rate of 20.6%. Partial progress translated especially poorly into complete delivery in analytical chemistry and electrochemistry/environment: Avg. Scores reached 87.6 and 94.9, but the highest Pass Rates were only 4% and 0%. Among non-passing Claude Code trajectories, 75.5% still ended with language claiming completion. These findings show that neither high partial scores nor confident claims of completion reliably indicate that a scientific task has been fully delivered, highlighting the need to evaluate end-to-end workflow execution and the completeness of scientific deliverables together.
Chinese Translation
科学代理越来越多地分析数据、执行代码并生成研究成果，然而大多数基准测试强调最终答案、孤立程序或单一领域。我们引入了FrontierChallenge，这是一个跨领域基准，包含300个端到端的科学工作流。在本文中，我们发布并评估了其中97个任务，涵盖量子化学、分子动力学、材料表征、分析化学、生命科学以及电化学/环境。每个任务提供固定的输入，并指定一组所需的科学交付成果。我们评估了十二个前沿模型与三种代理框架。通过率（Pass Rate）衡量满足完全完成标准的任务比例，而平均得分（Avg. Score）则反映部分进展。每个表现最佳的配置仅完成了97个发布任务中的20个，导致通过率为20.6%。在分析化学和电化学/环境中，部分进展转化为完整交付的效果尤其差：平均得分分别达到87.6和94.9，但最高通过率仅为4%和0%。在未通过的Claude Code轨迹中，75.5%的轨迹仍以声称完成的语言结束。这些发现表明，高的部分得分或对完成的自信声明并不能可靠地指示科学任务已完全交付，强调了需要同时评估端到端工作流执行和科学交付成果的完整性。
View on arXiv Download PDF AI Translation
cs.AI / 11 / 2608.24986

Solving Robust POMDPs with Omega-regular Objectives via Partially Observable Stochastic Games

通过部分可观测随机博弈解决具有Omega-常规目标的鲁棒POMDPs
Latha, Durgam, Reji, Dion, Akshay, S., Zikelic, Djordje, Krishna, Shankaranarayanan
Abstract
Robust POMDPs (RPOMDPs) generalize classical POMDPs to the setting where exact transition probabilities are not known -- rather, they are only known to belong to some uncertainty set of values. In this work, we study the problem of solving RPOMDPs with general omega-regular objectives, which subsume a broad class of objectives such as reachability, safety, and linear temporal logic (LTL) objectives. We show that, for (s,a)-rectangular RPOMDPs with polytopic uncertainty sets, the problem of solving RPOMDPs under omega-regular objectives can be reduced to solving partially observable stochastic games (POSGs) under omega-regular objectives. Moreover, we show for the first time that reductions can be constructed in both directions, establishing the semantic equivalence between (s,a)-rectangular RPOMDPs with polytopic uncertainty sets and POSGs. This allows us to derive a range of new computational complexity results, including both upper and lower complexity bounds, on solving RPOMDPs with different omega-regular objectives. As a corollary, we also derive new computational complexity results for RMDPs.
Chinese Translation
鲁棒POMDPs（RPOMDPs）将经典POMDPs推广到一个设定，其中确切的转移概率并不为人所知，而仅仅知道它们属于某个不确定性值集。在本研究中，我们研究了具有一般Omega-常规目标的RPOMDPs的求解问题，这类目标涵盖了广泛的目标类别，如可达性、安全性和线性时序逻辑（LTL）目标。我们表明，对于具有多面体不确定性集的（s,a）-矩形RPOMDPs，解决具有Omega-常规目标的RPOMDPs的问题可以归约为解决具有Omega-常规目标的部分可观测随机博弈（POSGs）。此外，我们首次展示了可以在两个方向上构建归约，确立了具有多面体不确定性集的（s,a）-矩形RPOMDPs与POSGs之间的语义等价性。这使我们能够推导出一系列新的计算复杂性结果，包括解决具有不同Omega-常规目标的RPOMDPs的上下复杂性界限。作为推论，我们还推导出了RMDPs的新计算复杂性结果。
View on arXiv Download PDF AI Translation
cs.AI / 12 / 2608.25023

CVE-SAI: Counterfactual Visual Evidence-Guided Selective Attribute Indexing for Risk-Controlled E-commerce Search

CVE-SAI：基于反事实视觉证据的选择性属性索引在风险控制电子商务搜索中的应用
Sun, Xiaolong, Wang, Qichao, Li, Hangyu, Chen, Liang
Abstract
Multimodal product models can complete missing e-commerce attributes, yet current methods still optimize attribute-answer accuracy without verifying visual support, conflate transient prediction with persistent index admission, and lack explicit risk control over factually incorrect or visually unsupported values. We address these gaps with Counterfactual Visual Evidence-Guided Selective Attribute Indexing (CVE-SAI), which first infers and freezes an ontology-constrained candidate from the primary image and attribute question without catalog text, and then decides whether that candidate should enter the index. Focus-Zone Distortion (FZD) constructs an attribute-specific visual-dependence proxy through a controlled counterfactual intervention, and Evidence-Guided Attention Redistribution (EGAR) uses the proxy to refine ontology-constrained scoring. The canonical candidate is frozen before evidence necessity, evidence retention, nuisance-transformation stability, and candidate-specific catalog-text conflict audits; catalog text can only tighten admission and cannot revise the candidate. Independent family-level calibration selects one policy with a simultaneous one-sided finite-sample bound under a 5% unsafe-admission budget. Experiments on five visual attributes derived from Amazon Berkeley Objects show that CVE-SAI improves attribute inference and evidence localization, achieves the highest certified admission coverage under the shared risk protocol, and yields the strongest controlled retrieval performance with the lowest unsafe auto-induced exposure among automatic-admission systems. Separating inference from admission therefore enables visually supported attribute completion to improve retrieval while limiting persistent index contamination.
Chinese Translation
多模态产品模型能够补全缺失的电子商务属性，但当前的方法仍然优化属性答案的准确性而未验证视觉支持，将瞬时预测与持久索引接纳混为一谈，并且缺乏对事实不正确或视觉不支持值的明确风险控制。我们通过反事实视觉证据引导的选择性属性索引（CVE-SAI）来解决这些问题，该方法首先从主要图像和属性问题中推断并冻结一个受本体约束的候选项，而不依赖于目录文本，然后决定该候选项是否应进入索引。焦点区域失真（FZD）通过受控的反事实干预构建一个属性特定的视觉依赖代理，而证据引导的注意力重分配（EGAR）利用该代理来细化受本体约束的评分。在证据必要性、证据保留、干扰转化稳定性和候选特定目录文本冲突审计之前，规范候选项被冻结；目录文本只能收紧接纳标准，而不能修订候选项。独立的家庭级校准选择一个政策，在5%的不安全接纳预算下具有同时的一侧有限样本界限。对来自亚马逊伯克利物体的五个视觉属性的实验表明，CVE-SAI提高了属性推断和证据定位，在共享风险协议下实现了最高的认证接纳覆盖率，并在自动接纳系统中产生了最低的不安全自诱导暴露的最强控制检索性能。因此，将推断与接纳分离，使得视觉支持的属性补全能够改善检索，同时限制持久索引污染。
View on arXiv Download PDF AI Translation
cs.AI / 13 / 2608.25037

Retrieve, Match, Escalate: Accurate and Scalable Product Linking with VLM-Distilled Cross-Encoders and Agentic VLMs

检索、匹配、升级：基于VLM提炼的交叉编码器和自主VLM的准确且可扩展的产品链接
Wang, Jian, Xu, Steven, Thete, Sanjyot, Barouti, Maryam, Tang, Tom, Wu, Elaine, Sareen, Charu, MacDonald, Kyle
Abstract
Product linking, the entity-resolution task of mapping merchant product records to canonical catalog products, consolidates fragmented listings so downstream search, recommendation, and advertising see one clean entry per product. At marketplace scale, billions of noisy, multi-category records must be resolved against tens of millions of canonical products, where scoring every candidate with a single model is either too weak for the hard cases or too costly for the easy ones. We present a production retrieve-then-match cascade that spends computation in proportion to difficulty: retrieval surfaces plausible matches, a lightweight text cross-encoder auto-resolves the high-confidence majority, and an agentic multimodal vision-language model settles the ambiguous remainder by inspecting product images and issuing web searches for evidence that is in neither record. The cross-encoder is distilled from millions of dual-VLM-consensus labels, retiring human annotation from the training set, and is calibrated to auto-accept links at a 98% precision bar validated against a smaller operator-certified audit. The agent is a self-hosted open-weight model that reaches a closed frontier VLM's precision at a four-point recall cost (88% versus 92%) for roughly one-seventh the per-pair cost, with no fine-tuning. Per-pair cost spans nearly five orders of magnitude from the cheap cross-encoder to the frontier VLM, so escalating only the hard tail to the agent raises end-to-end link coverage from the cheap stage's 68% to 77%.
Chinese Translation
产品链接是将商家产品记录映射到标准目录产品的实体解析任务，旨在整合碎片化的列表，以便下游搜索、推荐和广告能够为每个产品提供一个干净的条目。在市场规模下，数十亿条嘈杂的多类别记录必须与数千万个标准产品进行解析，而用单一模型对每个候选项进行评分要么对困难案例过于薄弱，要么对简单案例成本过高。我们提出了一种生产级的检索-匹配级联方法，根据难度分配计算资源：检索阶段呈现出可能的匹配项，轻量级文本交叉编码器自动解析高置信度的多数情况，而自主的多模态视觉-语言模型通过检查产品图像并进行网络搜索以获取在两条记录中都不存在的证据来解决模糊的剩余情况。交叉编码器是从数百万个双VLM共识标签中提炼而来，取消了训练集中的人工标注，并经过校准以在经过较小的运营商认证审计验证的98%精度标准下自动接受链接。该代理是一个自托管的开放权重模型，其精度达到了闭合边界VLM的水平，但召回成本增加了四个点（88%对比92%），每对成本约为前者的七分之一，且无需微调。每对成本的范围几乎跨越了五个数量级，从便宜的交叉编码器到前沿VLM，因此仅将困难的尾部提升到代理，提高了端到端链接覆盖率，从便宜阶段的68%提升至77%。
View on arXiv Download PDF AI Translation
cs.AI / 14 / 2608.25039

LifePlanner: Evaluating LLM Agents for Geo-spatial Planning with Social Media Data

LifePlanner：评估用于地理空间规划的LLM代理与社交媒体数据
Dong, Zhen, Peng, Yuning, Shi, Yutao, Zhong, Lei, Mao, Yongsen, Liu, Yuan, Wang, Haiping
Abstract
Geo-spatial planning, like trip design, is a realistic testbed for LLM agents because it requires grounded tool use, noisy evidence retrieval, and multi-constraint reasoning. Most benchmarks, however, only provide clean geospatial data and tools, missing the open-ended social signals that people use in daily planning. We introduce LifePlanner, a benchmark that enriches map data with large-scale local social media posts and provides access through an MCP toolset. LifePlanner provides an evaluation suite spanning four task categories and three difficulty levels. Experiments show frontier LLMs perform well on simple retrieval but degrade sharply on complex planning, with the Pass Rate dropping to 40.2%. Results show that failures mainly stem from incomplete evidence acquisition from such a large multimodal database, imprecise tool use, and weak constraint integration rather than model size or reasoning length, suggesting that future progress requires effective grounded planning instead of scaling alone.
Chinese Translation
地理空间规划，如旅行设计，是评估LLM代理的一个现实测试平台，因为它需要基于工具的使用、嘈杂的证据检索和多约束推理。然而，大多数基准测试仅提供干净的地理空间数据和工具，缺乏人们在日常规划中使用的开放式社交信号。我们介绍了LifePlanner，一个通过大规模本地社交媒体帖子丰富地图数据的基准，并通过MCP工具集提供访问。LifePlanner提供了一个涵盖四个任务类别和三个难度级别的评估套件。实验表明，前沿的LLM在简单检索任务中表现良好，但在复杂规划任务中急剧下降，合格率降至40.2%。结果表明，失败主要源于从如此庞大的多模态数据库中获取证据的不完整、工具使用的不精确以及约束整合的薄弱，而不是模型规模或推理长度，这表明未来的进展需要有效的基于证据的规划，而不仅仅是规模的扩大。
View on arXiv Download PDF AI Translation
cs.AI / 15 / 2608.25067

SimVerity: When Does Simulated Agent Success Survive Physical Deployment?

SimVerity：模拟代理成功何时能够在物理部署中存活？
Zhan, Zhonghao, Zhang, Yefan, Li, Krinos, Haddadi, Hamed
Abstract
Simulated evaluation is widely used to benchmark AI agents, yet how much evidence a simulated pass provides about physical deployment has not been systematically quantified. We present SimVerity, a verdict-transfer assurance framework: it replays matched scenarios on target smart home deployments and cross-validates agent execution against independently qualified physical witnesses. Our evaluation highlights that deployment success is a real-world process, not a static property in simulation: completion, reported state, observable effect, and settled outcome diverged within the same execution. Although an advanced simulator cleared all 240 light trials, a camera caught 42 sub-second failures invisible to settled-state checks. False clearance was predictable: a risk profile learned from measured trials and locked before evaluation predicted failures on a path it never physically measured, beating a property-blind baseline in all eleven held-out sessions across two cohorts. Agent auditability was also measurable: switching one agent loop's model-client/serving configuration raised its scenario-matching share from 52-88% to 100%. Finally, a second qualified simulator added no independent cross-check: it never disagreed on any overlapping case, and only physical measurement exposed their shared blind spots. SimVerity turns verdict transfer into an explicit decision: clear, abstain, or escalate before deployment.
Chinese Translation
模拟评估广泛用于基准测试人工智能代理，但模拟通过所提供的关于物理部署的证据量尚未得到系统量化。我们提出了SimVerity，一个判决转移保障框架：它在目标智能家居部署上重放匹配场景，并对代理执行进行交叉验证，以独立合格的物理见证者为依据。我们的评估强调，部署成功是一个现实世界的过程，而不是模拟中的静态属性：在同一执行中，完成情况、报告状态、可观察效果和最终结果存在差异。尽管一个先进的模拟器通过了所有240个光源试验，但一台相机捕捉到了42个在稳定状态检查中不可见的亚秒失败。错误清除是可预测的：从测量试验中学习到的风险特征在评估前被锁定，预测了在其从未物理测量过的路径上的失败，在两个组的所有十一场保留会话中超越了盲属性基线。代理的可审计性也是可测量的：切换一个代理循环的模型-客户端/服务配置使其场景匹配比例从52-88%提升至100%。最后，第二个合格的模拟器没有提供独立的交叉检查：它在任何重叠案例上都没有不同意见，只有物理测量揭示了它们共同的盲点。SimVerity将判决转移转变为一个明确的决策：在部署前选择清除、弃权或升级。
View on arXiv Download PDF AI Translation
cs.AI / 16 / 2608.25091

Auto-Policy, not Auto-Skill: Compiled Agent Skills for the Physical World

自动策略，而非自动技能：为物理世界编译的代理技能
Zhan, Zhonghao, Haddadi, Hamed
Abstract
Self-evolving Skill harnesses (AutoSkills, Hermes Agent) generate more advisory orchestration automatically; their reported gains are efficiency, not safety. This misses the actual gap: a Skill describes how an agent should behave; a Policy decides which behavior is allowed to become an action. Today's format covers the first with markdown and scripts; the second is left to the model. Generating more Skills scales the gap, not the safety, especially when a wrong invocation can unlock a door or move money. Two adjacent attacks are documented: malicious skills compromising cloud software, and jailbroken LLM-controlled robots causing physical harm. Their intersection, malicious agent skills causing physical harm, follows directly but has not been reported. We name this class Borrowed Authority: Skills format gives the receiving agent no typed way to reject an inter-agent permission claim, so a malicious or misused Skill can drive actuation by attaching one. We propose Edge Skillguard, a typed authority layer that lives inside the Skill artifact rather than between tools as workflow engines do, with guards over world state and sensor evidence. On a live edge control-plane testbed, the guards reject 60/60 borrowed-authority requests across five attack variants without blocking benign requests, and the result holds at 5x scale and across hosts over a Tailscale mesh. These results suggest that high-risk Skills should co-package typed invocation policy with procedural knowledge, so that physical actions depend on machine-checkable evidence rather than peer-agent claims.
Chinese Translation
自我演化技能的利用（AutoSkills，Hermes Agent）能够自动生成更多的建议性协调；其报告的收益是效率，而非安全。这忽视了实际的差距：技能描述了代理应如何行为；而策略决定了哪些行为被允许转化为行动。当前的格式通过Markdown和脚本覆盖了前者，而后者则留给模型。生成更多技能扩大了差距，而非安全，尤其是在错误的调用可能解锁门或转移资金的情况下。我们记录了两种相邻的攻击：恶意技能危害云软件，以及越狱的LLM控制机器人造成物理伤害。它们的交集，即恶意代理技能造成物理伤害，直接跟随但尚未被报告。我们将这一类称为借用权威（Borrowed Authority）：技能格式使接收代理无法以类型化方式拒绝跨代理的权限请求，因此恶意或误用的技能可以通过附加一个来驱动执行。我们提出了边缘技能保护（Edge Skillguard），这是一种类型化的权威层，存在于技能工件内部，而不是像工作流引擎那样存在于工具之间，具有对世界状态和传感器证据的保护。在一个实时的边缘控制平面测试平台上，这些保护机制拒绝了60/60个借用权威请求，涵盖五种攻击变体，而没有阻止良性请求，并且这一结果在5倍规模和跨主机的Tailscale网络中保持有效。这些结果表明，高风险技能应与程序知识共同打包类型化调用策略，以便物理行动依赖于机器可验证的证据，而非同行代理的声明。
View on arXiv Download PDF AI Translation
cs.AI / 17 / 2608.25097

PhysElite: How Far Are LLMs from Solving Olympiad-Level Physics Problems?

PhysElite：大型语言模型距离解决奥林匹克级物理问题还有多远？
Xu, Ruoran, Gao, Wending, Chen, Liyunfeng, Shi, Aixin, Cheng, Haoyu, Fang, Zixiang, Zou, Yiqiang, Wang, Qiufeng
Abstract
Understanding how (multimodal) large language models perform on physics problems requires benchmarks that reflect the difficulty and breadth of expert-level physical reasoning. Existing physics benchmarks remain limited in the following two important ways: (1) short of high-difficulty datasets, and (2) lack of comprehensive coverage of visual forms, knowledge points, and step-by-step solution processes. As a result, model performance on current datasets may not be fully representative of their ability to solve complex physics problems. To address these issues, we present PhysElite, a large-scale bilingual multimodal benchmark for Olympiad-level physics reasoning. PhysElite contains 11,586 Olympiad-tier problems. For each problem, we provide corresponding visual diagrams, step-by-step bilingual Chinese-English solution derivations, and the final answer. We benchmark 18 open-source and closed-source MLLMs, and find that even the strongest model reaches only 33.7% answer accuracy. We additionally conduct step-level process evaluation to diagnose where models fail in the reasoning chain. Our datasets are released at https://huggingface.co/datasets/physelite/PhysElite.
Chinese Translation
理解（多模态）大型语言模型在物理问题上的表现需要反映专家级物理推理难度和广度的基准测试。现有的物理基准测试在以下两个重要方面仍然有限：（1）缺乏高难度数据集，以及（2）缺乏对视觉形式、知识点和逐步解决过程的全面覆盖。因此，当前数据集上的模型表现可能无法充分代表它们解决复杂物理问题的能力。为了解决这些问题，我们提出了PhysElite，这是一个针对奥林匹克级物理推理的大规模双语多模态基准测试。PhysElite包含11,586个奥林匹克级问题。对于每个问题，我们提供相应的视觉图示、逐步的双语中英文解决推导和最终答案。我们对18个开源和闭源的多语言大型语言模型（MLLMs）进行了基准测试，发现即使是最强的模型其答案准确率也仅为33.7%。此外，我们还进行了逐步过程评估，以诊断模型在推理链中失败的地方。我们的数据集已在https://huggingface.co/datasets/physelite/PhysElite发布。
View on arXiv Download PDF AI Translation
cs.AI / 18 / 2608.25100

Towards Reliable, Generalizable, and Specific In-Context Knowledge Editing via Multi-Objective Reinforcement Learning

通过多目标强化学习实现可靠、可推广和特定的上下文知识编辑
Wang, Xuzhong, Jiang, Maiqi, Nair, Tejal, Bhusal, Girija, Zhang, Yanfu, Chen, Haipeng
Abstract
Large Language Models (LLMs) are powerful but limited by static parametric knowledge that becomes outdated once pretraining ends. Knowledge editing addresses this problem by updating model behavior on target facts without full retraining. In particular, in-context knowledge editing has gained attention because it is training-free and readily applicable to black-box LLMs. Recent reinforcement learning (RL)-based approaches improve over fixed retrieval strategies by adapting prompt construction to the quantity-quality trade-off. Despite initial success, they fail to model the prompt as a structured entity under the distinct and often competing objectives of reliability, generality, and specificity. Previous methods largely optimize a single objective and make decisions over only part of the prompt construction process, thereby overlooking both the balance of different objectives and the global organization of demonstrations. We propose Multi-Objective In-context Knowledge Editing (MO-IKE), a multi-objective RL algorithm that formulates prompt construction for in-context knowledge editing as a Constrained Markov Decision Process. MO-IKE trains a dynamic retriever to optimize competing objectives in knowledge editing, enabling more balanced and globally coherent prompt construction. On Llama-3.2, MO-IKE improves edit success (reliability) from 85.0% to 92.0%, paraphrase consistency (generality) from 77% to 79%, while increasing retention rate (specificity) by 23.0% compared to prior RL-based methods.
Chinese Translation
大型语言模型（LLMs）功能强大，但受限于静态参数知识，一旦预训练结束便变得过时。知识编辑通过在不进行全面重训练的情况下更新模型对目标事实的行为来解决这一问题。特别是，上下文知识编辑因其无训练需求且可直接应用于黑箱LLM而受到关注。最近的基于强化学习（RL）的方法通过调整提示构建以应对数量与质量的权衡，改善了固定检索策略。尽管初步成功，但它们未能将提示建模为一个结构化实体，无法在可靠性、普适性和特异性这三个不同且常常相互竞争的目标之间进行有效平衡。以往的方法主要优化单一目标，并仅在提示构建过程的部分环节做出决策，从而忽视了不同目标之间的平衡以及示例的全局组织。我们提出了多目标上下文知识编辑（MO-IKE），这是一种将上下文知识编辑的提示构建形式化为约束马尔可夫决策过程的多目标RL算法。MO-IKE训练一个动态检索器，以优化知识编辑中的竞争目标，从而实现更平衡和全局一致的提示构建。在Llama-3.2上，MO-IKE将编辑成功率（可靠性）从85.0%提高到92.0%，释义一致性（普适性）从77%提高到79%，同时相比于以往的基于RL的方法，保留率（特异性）提高了23.0%。
View on arXiv Download PDF AI Translation
cs.AI / 19 / 2608.25158

FuzzingBrain-Bench V1: Evaluating Open-Ended Bug Discovery by LLMs

FuzzingBrain-Bench V1：评估大型语言模型在开放式漏洞发现中的能力
Sheng, Ze, Kezic, Aleksandar, Chen, Zhicheng, Huang, Jeff
Abstract
Evaluating the ability of large language models (LLMs) to discover software bugs is increasingly important. Existing benchmarks typically evaluate this capability by asking the model to generate a proof-of-concept input that triggers a predefined target vulnerability. However, this setup may overlook valid crashes discovered by the model when they do not match the predefined target. As a result, the evaluation may not reflect the model's real capability. We present FuzzingBrain-Bench, a benchmark for assessing AI models' ability to discover bugs in open-source software. Models are given an open-source project and a sanitizer-instrumented harness in a self-contained Docker image. Their goal is to generate inputs that trigger as many distinct crashes as possible through the harness. A model's performance on each challenge is scored based on the number of distinct crash signatures it produces, capped at a predefined maximum and weighted by a difficulty coefficient. FuzzingBrain-Bench V1 consists of 77 challenges drawn from 43 open-source projects, with 36 C, 32 C++, and 9 Java/JVM challenges. We evaluate Claude Haiku 4.5, Claude Sonnet 4.6, and Claude Opus 4.8 on the full benchmark. Claude Opus 4.8 performs best, triggering crashes in 60 of 77 challenges and achieving a score of 196 out of 579. None of the three models triggers a crash in 13 challenges. The FuzzingBrain-Bench corpus and harnesses are publicly available at https://github.com/fuzzingbrain/FuzzingBrain-Bench.
Chinese Translation
评估大型语言模型（LLMs）发现软件漏洞的能力变得越来越重要。现有基准通常通过要求模型生成触发预定义目标漏洞的概念验证输入来评估这一能力。然而，这种设置可能会忽视模型发现的有效崩溃，尤其是当这些崩溃不符合预定义目标时。因此，评估可能无法反映模型的真实能力。我们提出了FuzzingBrain-Bench，这是一个用于评估AI模型在开源软件中发现漏洞能力的基准。模型被提供一个开源项目和一个在自包含Docker镜像中的清理器仪器化的测试框架。它们的目标是生成输入，通过测试框架触发尽可能多的不同崩溃。模型在每个挑战中的表现根据其产生的不同崩溃签名的数量进行评分，评分上限为预定义的最大值，并根据难度系数加权。FuzzingBrain-Bench V1包含来自43个开源项目的77个挑战，其中包括36个C语言、32个C++和9个Java/JVM挑战。我们在完整基准上评估了Claude Haiku 4.5、Claude Sonnet 4.6和Claude Opus 4.8。Claude Opus 4.8表现最佳，在77个挑战中触发了60个崩溃，得分为196（满分579）。这三种模型在13个挑战中均未触发崩溃。FuzzingBrain-Bench语料库和测试框架可在https://github.com/fuzzingbrain/FuzzingBrain-Bench公开获取。
View on arXiv Download PDF AI Translation
cs.AI / 20 / 2608.25198

Tunable Tool-Call Rates in LLM Agents via Representation Steering

通过表示引导调节 LLM 代理中的工具调用率
Chen, Yuqi, Siu, Vincent, Liu, Yang, Song, Dawn, Wang, Chenguang
Abstract
Deciding whether to call a tool is a core competence of an LLM agent, and a costly one to get wrong: needless calls add latency, accrue cost, and may trigger irreversible side effects, while missing calls leave the model confidently wrong on questions it could only answer through tool-calls. Models manage this balance poorly, both over-using and under-using tools. Existing methods such as post-training and prompt engineering are expensive and difficult to modify at inference time. We show that whether an instruction-tuned model calls a tool can be controlled by a single linear direction in its residual stream, extracted without any training from the model's own tool-use preference signal and turned into an inference-time intervention with no prompt change. Adding the direction with strength $\alpha$ moves the call rate monotonically from near $0\% $ to over $90\%$ while keeping calls well-formed. The steering works in both directions: dialing it down suppresses calls, and dialing it up induces new calls that land precisely on the questions the model cannot answer from its own knowledge. We also show that the direction generalizes to unseen tools with strength comparable to each tool's own direction and without favoring any specific tool choice. With live tool execution, a single sweep of the steering traces a cost/accuracy Pareto frontier and nearly doubles open-domain QA accuracy ($0.29 \! \rightarrow \! 0.56$); the same recipe transfers across a diverse range of models spanning dense, MoE, and multimodal architectures, without any training. Our code is publicly available at https://github.com/YuqiChen4188/Steering-Tool-Use-Propensity.
Chinese Translation
决定是否调用工具是 LLM 代理的核心能力，同时也是一项容易出错的高成本决策：不必要的调用会增加延迟、产生成本，并可能引发不可逆的副作用，而漏掉调用则会使模型在只能通过工具调用回答的问题上自信地错误。模型在管理这种平衡方面表现不佳，既有过度使用工具的情况，也有不足使用的情况。现有的方法，如后训练和提示工程，成本高且在推理时难以修改。我们展示了指令调优模型是否调用工具可以通过其残差流中的一个线性方向来控制，该方向无需任何训练即可从模型自身的工具使用偏好信号中提取，并转化为推理时的干预，而无需更改提示。添加强度为 $eta$ 的方向可以使调用率单调变化，从接近 $0\%$ 增加到超过 $90\\%$，同时保持调用的良好格式。引导在两个方向上均有效：降低强度会抑制调用，而增加强度则会引发新的调用，恰好针对模型无法通过自身知识回答的问题。我们还展示了该方向能够推广到未见过的工具，其强度与每个工具自身的方向相当，并且不偏向任何特定的工具选择。在实时工具执行中，单次引导的操作描绘了成本/准确性帕累托前沿，并几乎将开放域问答的准确性翻倍（$0.29 ightarrow 0.56$）；相同的方法在包括密集型、MoE 和多模态架构在内的多种模型中均能有效转移，且无需任何训练。我们的代码已公开发布在 https://github.com/YuqiChen4188/Steering-Tool-Use-Propensity。
View on arXiv Download PDF AI Translation
cs.AI / 21 / 2608.25215

Federation Is Nearly Free, Reasoning Is Not: Tradeoffs for AI Co-Scientists in Protein Characterization Workflows

联邦几乎是免费的，推理却不是：AI共同科学家在蛋白质表征工作流程中的权衡
Kapur, Maia, Boe, Timothy, Jerger, Abby, Rigor, Paul
Abstract
Natural language driven autonomous co-scientist workflows involve a fundamental trade-off between flexibility and reasoning at the expense of determinism, reproducibility, and observability. Such agents increasingly must communicate across institutional boundaries, where federation topology can shape latency and cost. We systematically evaluated these tradeoffs using a controlled ablation on a production agentic platform for science. We use a verifiable task: given a protein sequence, we ask an agent to confidently characterize its function by routing across common tools. We compare federation topology, classic RL vs LLM-driven harnesses, language model, and prompt expertise. We also stratify results by protein novelty. We find that the choice of LLM dominated prediction quality far more than topology or prompting (Opus ~92%-94% vs o4-mini ~40%-50%). The PPO policy was nearly as accurate as the best LLM (88%) at zero token cost, fastest latency, and perfect consistency, but yields no reasoning trace. Expert prompted LLMs reached the highest accuracy but were high-cost and less consistent; prompt dependence was largest when the task was hardest. Federation imposed a negligible penalty on performance. These results offer actionable guidance for deploying agents for scientific workflows: for routine, verifiable tasks, a cheap deterministic policy delivers near-frontier accuracy with complete reproducibility, while flexible LLM reasoning is best reserved for open-ended discovery.
Chinese Translation
基于自然语言驱动的自主共同科学家工作流程在灵活性与推理之间存在根本的权衡，这种权衡以确定性、可重复性和可观察性为代价。这类代理越来越需要跨越机构边界进行沟通，其中联邦拓扑结构会影响延迟和成本。我们通过在一个科学生产代理平台上进行受控消融实验，系统地评估了这些权衡。我们使用一个可验证的任务：给定一个蛋白质序列，我们要求代理通过常用工具自信地表征其功能。我们比较了联邦拓扑、经典强化学习（RL）与大型语言模型（LLM）驱动的工具、语言模型和提示专业知识。我们还根据蛋白质的新颖性对结果进行了分层分析。我们发现，LLM的选择对预测质量的影响远大于拓扑或提示（Opus约92%-94%对比o4-mini约40%-50%）。PPO策略在零令牌成本、最快延迟和完美一致性下，其准确性几乎与最佳LLM（88%）相当，但未产生推理痕迹。专家提示的LLM达到了最高的准确性，但成本高且一致性较差；当任务最具挑战性时，提示依赖性最大。联邦对性能的影响微乎其微。这些结果为科学工作流程中代理的部署提供了可操作的指导：对于常规的、可验证的任务，廉价的确定性策略能够提供接近前沿的准确性和完全的可重复性，而灵活的LLM推理则最好保留用于开放式探索。
View on arXiv Download PDF AI Translation
cs.AI / 22 / 2608.25217

LLM-Driven, Datasheet-Aware Automated Hardware Compatibility Verification for Early-Stage, Pre-Schematic Embedded System Design

基于LLM驱动的、关注数据手册的自动化硬件兼容性验证框架，用于早期预原理图嵌入式系统设计
Qiao, Haotian, Dick, Robert P.
Abstract
We present an LLM-driven, datasheet-aware framework for early-stage hardware compatibility verification that identifies documentation-level interface incompatibilities based on hardware datasheets and high-level component connectivity descriptions. It does not require, and can therefore be used, before detailed schematic simulation and implementation. We view trustworthy LLM-assisted design automation not as directly generating answers from documents, but as transforming engineering information through traceable verification stages. Given hardware datasheets and high-level component connectivity descriptions, the framework constructs a design graph that captures device connectivity and shared interaction domains, retrieves only the engineering properties required by explicit, domain-oriented verification criteria , and generates deterministic scripts for compatibility evaluation. By decomposing compatibility analysis into modular stages and preserving intermediate results, the framework reduces context overhead, improves transparency and tractability, enables scaling, and avoids reliance on LLMs for numerical computation. Evaluated on seven embedded-system designs comprising 34 datasheets, our framework achieves 97.5% compatibility-verification accuracy and an 8.6 times reduction in input context size compared with ``upload-and-query'' workflows. These results demonstrate the feasibility of LLM-assisted, specification-based hardware compatibility verification at an early design stage, as well as the need for, and substantial benefits of, modular task decomposition, formalized verification criteria, and task-aware compact context construction.
Chinese Translation
我们提出了一种基于LLM驱动、关注数据手册的早期硬件兼容性验证框架，该框架根据硬件数据手册和高层组件连接描述识别文档级接口不兼容性。该框架不需要在详细原理图仿真和实现之前使用，因此可以在早期阶段应用。我们将可信赖的LLM辅助设计自动化视为通过可追溯的验证阶段转化工程信息，而不是直接从文档中生成答案。给定硬件数据手册和高层组件连接描述，该框架构建了一个设计图，捕捉设备连接性和共享交互域，仅检索由明确的、面向领域的验证标准所需的工程属性，并生成用于兼容性评估的确定性脚本。通过将兼容性分析分解为模块化阶段并保留中间结果，该框架减少了上下文开销，提高了透明度和可追溯性，支持扩展，并避免对LLM进行数值计算的依赖。在对包括34个数据手册的七个嵌入式系统设计进行评估时，我们的框架实现了97.5%的兼容性验证准确率，并将输入上下文大小减少了8.6倍，与“上传并查询”的工作流程相比。这些结果展示了在早期设计阶段进行LLM辅助、基于规范的硬件兼容性验证的可行性，以及模块化任务分解、形式化验证标准和任务感知紧凑上下文构建的必要性和显著优势。
View on arXiv Download PDF AI Translation
cs.AI / 23 / 2608.25220

FLARE: Verifying MILP Reformulations with LLM-Based Theorem Proving

FLARE：基于大型语言模型的定理证明验证混合整数线性规划重构
Robbins, Henry, Lawless, Connor, Udell, Madeleine, Vitercik, Ellen
Abstract
Mixed-Integer Linear Programming (MILP) is a fundamental tool for combinatorial optimization with extensive real-world applications. A central challenge is designing computationally efficient MILP formulations. Large Language Models (LLMs) offer new opportunities to automate the modeling process, from deriving formulations to strengthening them. Reliable automation requires robust methods for verifying that proposed formulations preserve the underlying optimization problem. However, existing approaches evaluate formulations numerically and fail to reason about general problem instances. We resolve this limitation by introducing a constructive definition of MILP reformulation that can be formalized in Lean and machine-checked. We develop FLARE (Formulation-Level Automated Reformulation Evaluation), a method that uses an LLM-based agent and the Lean proof assistant to verify proposed reformulations against a reference formulation. To evaluate our approach, we introduce FormulationBench, a challenging dataset of 20 problems and 109 formulations. FLARE outperforms existing methods, with 100% accuracy on the NP-hard subset of FormulationBench. Furthermore, FLARE produces a machine-checkable certificate for every reformulation it accepts. For cases where formal guarantees are not necessary, we introduce FLARE-NL, a fast and cheap LLM proxy that matches FLARE's accuracy but produces no certificate. These methods enable reliable verification in automated optimization modeling.
Chinese Translation
混合整数线性规划（MILP）是组合优化的基本工具，广泛应用于现实世界中。一个核心挑战是设计计算效率高的MILP公式。大型语言模型（LLMs）为自动化建模过程提供了新的机会，从推导公式到增强公式的过程。然而，可靠的自动化需要强大的方法来验证所提出的公式是否保留了基础优化问题。然而，现有的方法仅通过数值评估公式，无法对一般问题实例进行推理。我们通过引入MILP重构的构造性定义来解决这一限制，该定义可以在Lean中形式化并经过机器检查。我们开发了FLARE（公式级自动重构评估），这是一种利用基于LLM的代理和Lean证明助手来验证所提出的重构与参考公式的一致性的方法。为了评估我们的方法，我们引入了FormulationBench，这是一个包含20个问题和109个公式的挑战性数据集。FLARE在FormulationBench的NP难度子集上表现优异，准确率达到100%。此外，FLARE为每个接受的重构生成一个可机器检查的证书。对于不需要正式保证的情况，我们引入了FLARE-NL，这是一种快速且低成本的LLM代理，能够匹配FLARE的准确性，但不生成证书。这些方法使得在自动化优化建模中实现可靠的验证成为可能。
View on arXiv Download PDF AI Translation
cs.AI / 24 / 2608.25261

Hierarchical MoE for Multi-Modal ILD Diagnosis

用于多模态间质性肺病诊断的层次化专家混合模型
Peltekian, Alec K., Durak, Gorkem, Aktas, Halil Ertugrul, Richardson, Carrie Lynn, Carns, Mary, Aren, Kathleen, Budinger, GR Scott, Esposito, Anthony J., Misharin, Alexander, Choudhary, Alok Nidhi, Agrawal, Ankit, Bagci, Ulas
Abstract
Mixture-of-experts (MoE) models combine specialized predictors under learned routing, offering a principled mechanism for leveraging heterogeneity in medical data. We present a hierarchical multimodal MoE for interstitial lung disease (ILD) classification that integrates a frozen, pre-trained imaging expert with structured electronic health records (EHR) via two-stage gating. A modality-level gate assigns patient-specific weights to imaging and EHR predictions, while a sub-gating module decomposes the EHR branch into clinically defined feature groups with learned, group-specific contributions. This design preserves stable imaging representations while enabling input-dependent clinical weighting and explicit EHR specialization. Under strict patient-level cross-validation, the model achieved the highest mean AUC among the evaluated methods (0.8750 +- 0.0443), compared with 0.8646 for imaging-only REN and 0.7685 for SwinUNETR. The framework extends interpretability across anatomical regions, imaging--EHR utilization, and clinically defined EHR feature groups.
Chinese Translation
专家混合模型（Mixture-of-experts, MoE）通过学习的路由机制结合了专业预测器，为利用医疗数据的异质性提供了一种原则性的方法。我们提出了一种层次化多模态MoE用于间质性肺病（Interstitial Lung Disease, ILD）分类，该模型通过两阶段门控将冻结的预训练影像专家与结构化电子健康记录（Electronic Health Records, EHR）相结合。模态级门控为影像和EHR预测分配患者特定的权重，而子门控模块将EHR分支分解为具有学习到的、特定于组的贡献的临床定义特征组。这种设计在保持稳定的影像表示的同时，允许输入依赖的临床加权和明确的EHR专业化。在严格的患者级交叉验证下，该模型在评估的方法中实现了最高的平均曲线下面积（AUC）（0.8750 ± 0.0443），相比之下，仅影像的REN为0.8646，SwinUNETR为0.7685。该框架扩展了在解剖区域、影像与EHR的利用以及临床定义的EHR特征组之间的可解释性。
View on arXiv Download PDF AI Translation
cs.AI / 25 / 2608.25275

PhaseShift: Topology-Aware Data Harmonization and Model Consolidation Across Signalized Intersections

PhaseShift：基于拓扑的信号交叉口数据协调与模型整合
Ranjan, Yash, Kumik, Artur, Sengupta, Rahul, Rangarajan, Anand, Ranka, Sanjay
Abstract
Learned traffic-behavior models are commonly trained separately for each intersection, creating model portfolios that cannot share evidence across sites. We present PhaseShift, a topology-aware framework that harmonizes heterogeneous roadside trajectories into a shared actor-centric representation and trains one reusable backbone. Ego-relative coordinates, trajectory-induced movement paths, normalized signal context, and variable-cardinality interaction tokens remove site conventions while preserving behaviorally relevant topology. The backbone supports pooled operation, zero-shot at a held-out intersection, and low-data adaptation. We evaluate five intersections in two Florida regions on balanced field data, 100k training windows and equal-sized test sets per site under a replay-conditioned, best-of-sampled-trajectory protocol. At 10s, one pooled model lowers both minADE and minFDE relative to trained local models at all five sites, with median reductions of 36.8% and 22.0%. Leave-one-intersection-out deployment, including one cross-region fold, beats local training on both 10-s metrics at four of five sites, although short-horizon performance is less uniform. Fine-tuning with 1,000 target update windows improves on zero-shot at three sites and is the strongest regime at one. At site 7, every cross-site mixture sharply lowers long-horizon error under a fixed 100k-window budget; test-likelihood gains argue against a best-of-sample dispersion-only explanation. Local models fall behind calibrated IDM at the two highest-flow sites after long autoregressive rollouts; pretrained-backbone regimes do not. Within this five-site evaluation, PhaseShift demonstrates consolidation across heterogeneous physical control settings while identifying sites that still require adaptation. The protocol measures conditional single-vehicle generation under replayed context, not closed-loop traffic simulation.
Chinese Translation
学习的交通行为模型通常为每个交叉口单独训练，导致模型组合无法在不同地点共享证据。我们提出了PhaseShift，一个基于拓扑的框架，将异构的路边轨迹协调为共享的以参与者为中心的表示，并训练一个可重复使用的主干网络。自我相对坐标、轨迹引发的运动路径、标准化的信号上下文和可变基数的交互标记消除了地点约定，同时保留了行为相关的拓扑。该主干网络支持汇聚操作、在保留的交叉口进行零样本推理以及低数据适应。我们在佛罗里达州两个地区的五个交叉口上评估了平衡的现场数据，使用100,000个训练窗口和每个地点相等大小的测试集，采用重放条件的最佳样本轨迹协议。在10秒时，一个汇聚模型相较于在所有五个地点训练的本地模型降低了最小平均定位误差（minADE）和最小最终定位误差（minFDE），中位数减少分别为36.8%和22.0%。在五个地点中的四个，采用留一交叉口的部署（包括一个跨区域折叠）在两个10秒指标上优于本地训练，尽管短期表现不够均匀。使用1,000个目标更新窗口进行微调在三个地点的零样本推理上有所改善，并且在一个地点表现最佳。在第7个地点，每个跨地点组合在固定的100,000窗口预算下显著降低了长期误差；测试似然的提升反驳了仅基于样本分散的解释。在经过长时间自回归滚动后，本地模型在两个流量最高的地点落后于校准的IDM；而预训练主干网络的模式则没有。在这五个地点的评估中，PhaseShift展示了在异构物理控制设置之间的整合，同时识别出仍需适应的地点。该协议测量在重放上下文下的条件单车生成，而非闭环交通模拟。
View on arXiv Download PDF AI Translation
cs.AI / 26 / 2608.25286

BixBench3: Benchmarking AI agents on research-study-scale computational biology tasks

BixBench3：在研究规模的计算生物学任务中评估人工智能代理的性能
Koch, Zane, Wassie, Asmamaw T., Valdes-Aleman, Javier, Lee, Jason, Hinks, Michaela M., Rodriques, Samuel G., White, Andrew D., Laurent, Jon M.
Abstract
Artificial intelligence (AI) promises to accelerate biological research by automating computational analyses. Yet the ability of AI agents to carry out computational biology at the scale of complete research studies has not been systematically evaluated. Here we introduce BixBench3, a benchmark that measures the capacity of AI agents to process raw biological data through to scientific results. We designed BixBench3 tasks to mirror the delegation of work from a scientist to an agent: the scientist chooses the research question and high-level methods, then delegates implementation of all analyses to the agent. In each task, an agent receives a research objective, methodological guidance, and raw data derived from a published scientific study, and must execute a sequence of analyses to achieve the research objective. The data artifacts resulting from these analyses - such as peak call matrices or differential expression tables - are programmatically graded against the corresponding artifacts generated and reported in the original study. Across 20 BixBench3 tasks encompassing the generation of 138 unique artifacts, we find that 13 frontier models achieve scores ranging from 0.00 for Gemini 3.1 Flash Lite to 0.48 for GPT 5.6 Sol. Agents perform worse on tasks with larger raw datasets (0.36 on tasks with <100 GB versus 0.10 on tasks with >100 GB) and on analyses requiring more sequential steps (0.36 at 1-2 steps vs 0.24 at 3+). On average, agents use 6.8 hours, 102 million tokens, and $43 to complete each task, with the longest attempts consuming 24 hours, 1.07 billion tokens, and $525. Notably, the highest-scoring agents used fewer tokens and were cheaper than less performant options. These results reveal that LLMs vary substantially in their ability to (1) execute multiple sequential analysis steps coherently, (2) manage large quantities of raw data, and (3) work across scientific domains.
Chinese Translation
人工智能（AI）有望通过自动化计算分析来加速生物研究。然而，AI代理在完整研究规模的计算生物学任务中的能力尚未得到系统评估。在此，我们介绍了BixBench3，一个评估AI代理处理原始生物数据直至科学结果能力的基准。我们设计BixBench3任务以模拟科学家向代理委派工作的过程：科学家选择研究问题和高层次方法，然后将所有分析的实施委托给代理。在每个任务中，代理接收一个研究目标、方法指导以及来自已发表科学研究的原始数据，并必须执行一系列分析以实现研究目标。这些分析产生的数据工件——例如峰值调用矩阵或差异表达表——将与原始研究中生成和报告的相应工件进行程序化评分。在涵盖138个独特工件生成的20个BixBench3任务中，我们发现13个前沿模型的得分范围从Gemini 3.1 Flash Lite的0.00到GPT 5.6 Sol的0.48。代理在处理较大原始数据集的任务中表现较差（在<100 GB任务中得分为0.36，而在>100 GB任务中得分为0.10），以及在需要更多顺序步骤的分析中（在1-2步时得分为0.36，而在3步及以上时得分为0.24）。平均而言，代理完成每个任务需耗时6.8小时、使用1.02亿个标记，花费43美元，最长的尝试耗时24小时，使用10.7亿个标记，花费525美元。值得注意的是，得分最高的代理使用的标记更少，成本也低于表现较差的选项。这些结果揭示了大型语言模型（LLMs）在以下能力上存在显著差异：（1）连贯地执行多个顺序分析步骤，（2）管理大量原始数据，以及（3）跨科学领域工作。
View on arXiv Download PDF AI Translation
cs.AI / 27 / 2608.25325

FinRiskAtlas: Decision-Aligned Evaluation of Large Language Models for Financial Risk Review

FinRiskAtlas：与决策对齐的大型语言模型金融风险评估
Zhong, Suyang, Zhu, Jingzhe, Xu, Qi, Sun, Liyao, Wang, Yin, Sun, Qingqing, Chen, Shuai, Zhang, Tianyi
Abstract
Deploying large language models for professional financial review requires more than measuring general financial competence: models must perform the specific review operation required by a workflow and determine whether available evidence is sufficient for a defensible decision. Existing financial benchmarks cover knowledge, reasoning, compliance, and professional tasks, but their evaluation units are often organized around datasets or task formulations rather than the decisions that deployed systems support. We introduce FinRiskAtlas, a Chinese-language benchmark that evaluates financial LLMs along two complementary dimensions: operation execution under fixed evidence states and evidence-state control under evolving review conditions. The static benchmark contains 9,742 instances across 53 task families, including 42 Domain Knowledge families and eleven downstream review operations defined by explicit evaluation contracts. FinRisk-Ask extends this framework through offline replay of 680 pre-action states from 104 de-identified professional trajectories, withholding future evidence during inference and using it only to construct expert-verified evidence targets. Across 33 model configurations, operation-level evaluation yields non-redundant rankings (mean pairwise Spearman correlation 0.42 across downstream operations), and knowledge-based shortlisting can incur up to 18.01 points of regret on individual operations. FinRisk-Ask further shows that entering the Ask branch more frequently does not necessarily improve request targeting or end-to-end evidence acquisition. These results show that broad financial capability scores do not fully capture where models are reliable in professional workflows, motivating evaluation units aligned with the decisions and evidence states that deployed systems must support.
Chinese Translation
在专业金融审查中部署大型语言模型不仅需要衡量一般的金融能力：模型必须执行工作流程所需的特定审查操作，并确定可用证据是否足以支持一个有据可依的决策。现有的金融基准涵盖知识、推理、合规和专业任务，但它们的评估单位往往围绕数据集或任务表述组织，而不是围绕已部署系统所支持的决策。我们引入了FinRiskAtlas，这是一个中文基准，沿着两个互补维度评估金融大型语言模型：在固定证据状态下的操作执行和在不断变化的审查条件下的证据状态控制。该静态基准包含9,742个实例，涵盖53个任务家族，包括42个领域知识家族和由明确评估合同定义的11个下游审查操作。FinRisk-Ask通过离线重播来自104个去标识化专业轨迹的680个预行动状态扩展了这一框架，在推理过程中保留未来证据，仅用于构建专家验证的证据目标。在33种模型配置下，操作级评估产生了非冗余的排名（下游操作的平均成对Spearman相关系数为0.42），而基于知识的短名单在单个操作上可能导致高达18.01分的遗憾。FinRisk-Ask进一步表明，更频繁地进入Ask分支并不一定改善请求目标或端到端证据获取。这些结果表明，广泛的金融能力评分并不能完全捕捉模型在专业工作流程中可靠的地方，促使与已部署系统必须支持的决策和证据状态对齐的评估单位的必要性。
View on arXiv Download PDF AI Translation
cs.AI / 28 / 2608.25329

Learning What to Share and What to Personalize: Hierarchical Strategy Co-Evolution for Agent Memory

学习分享与个性化的内容：代理记忆的层次策略共同演化
Han, Yupeng, Liu, Shuochen, Zhang, Kai, Liu, Ze, Pan, Zhihong, Wang, Xianquan
Abstract
Memory-augmented agents maintain compact user profiles throughout extended conversations, enabling personalized and consistent responses without the need to process the entire dialogue history. The quality of these user profiles relies on the underlying memory management strategy: at each step, the agent must determine what to retain, compress, or discard. However, existing methods typically employ a static, one-size-fits-all strategy established before training. In practice, the optimal memory decision is inherently user-specific and dynamically evolves alongside policy optimization. To address this, we propose \textbf{HiPS} (\textbf{Hi}erarchical \textbf{P}ersonalized \textbf{S}trategy), a framework that decouples memory management into a globally shared foundation and a user-specific adaptive tier. Specifically, HiPS employs \textbf{Universal Strategy} to extract shared principles from cross-persona trajectories, alongside \textbf{Persona Delta Distillation} to generate tailored rules for users whose behaviors diverge from general patterns. \textbf{Cross-Level Rule Flow} dynamically calibrates their boundary by promoting broadly validated personal rules and demoting contradicted global ones. The architecture establishes a co-evolution loop where a mechanism guarantees that all strategy refinements are anchored to task outcomes. Extensive experiments demonstrate consistent improvements over memory-augmented baselines.
Chinese Translation
增强记忆的代理在长时间对话中维护紧凑的用户档案，从而实现个性化和一致性的响应，而无需处理整个对话历史。这些用户档案的质量依赖于基础的记忆管理策略：在每一步中，代理必须决定保留、压缩或丢弃什么。然而，现有方法通常采用在训练之前建立的静态、一刀切的策略。实际上，最佳的记忆决策本质上是用户特定的，并随着策略优化而动态演变。为了解决这个问题，我们提出了 extbf{HiPS}（ extbf{Hi}erarchical extbf{P}ersonalized extbf{S}trategy），一个将记忆管理解耦为全球共享基础和用户特定自适应层的框架。具体而言，HiPS采用 extbf{Universal Strategy}从跨人物轨迹中提取共享原则，同时利用 extbf{Persona Delta Distillation}为行为偏离一般模式的用户生成量身定制的规则。 extbf{Cross-Level Rule Flow}通过促进广泛验证的个人规则并降低与全球规则相矛盾的规则的优先级，动态调整其边界。该架构建立了一个共同演化循环，其中一个机制确保所有策略的改进都与任务结果相结合。大量实验表明，相较于增强记忆的基线方法，性能有了一致的提升。
View on arXiv Download PDF AI Translation
cs.AI / 29 / 2608.25358

Where vs What: Decomposing Structural and Content Failures in LLM-Generated Structured Outputs

何处与何物：分解大型语言模型生成结构化输出中的结构性和内容性错误
Zhang, Yiwei, Wu, Chengke, Wang, Li, Li, Jianqiang
Abstract
Structured outputs such as JSON and tables are central to modern LLM-based systems, yet generation failures are evaluated monolithically, conflating two distinct error modes: placement errors (correct values at wrong positions) and value errors (wrong values at intended positions). We introduce Structure-Content Decomposition (SCD), a framework that independently measures structural fidelity and content accuracy. Applying SCD to nested JSON and table tasks across six models (7B to frontier), we uncover a consistent phenomenon: structural fidelity degrades earlier and more sharply than content accuracy as complexity increases. At the highest complexity, even DeepSeek-V4-Flash (with reasoning) misplaces 35% of recalled values, while Qwen2.5-7B misplaces 74%. Controlled ablations suggest that this pattern is associated with reliance on semantic shortcuts rather than topological understanding of output structure. Based on these findings, we propose SA-RLVR, converting SCD metrics into verifiable rewards for reinforcement learning via GRPO. SA-RLVR successfully optimizes structural addressing across distinct topologies: it lifts JSON Value Placement Accuracy (VPA) from 26% to 63% while generalizing to held-out schemas; moreover, it consistently drives VPA improvements in the table domain, demonstrating that structure-aware rewards can directly enhance multi-domain structural positioning.
Chinese Translation
结构化输出如 JSON 和表格在现代基于大型语言模型（LLM）的系统中至关重要，但生成失败通常以整体方式进行评估，将两种不同的错误模式混为一谈：位置错误（正确的值放在错误的位置）和数值错误（在预期位置放置错误的值）。我们提出了结构-内容分解（Structure-Content Decomposition, SCD）框架，独立测量结构的保真度和内容的准确性。将 SCD 应用于六个模型（从 7B 到前沿）在嵌套 JSON 和表格任务中，我们发现一个一致的现象：随着复杂性的增加，结构保真度比内容准确性更早且更急剧地下降。在最高复杂度下，即使是 DeepSeek-V4-Flash（具备推理能力）也会将 35% 的回忆值放错位置，而 Qwen2.5-7B 则将 74% 的值放错。控制性消融实验表明，这一模式与对语义捷径的依赖而非对输出结构的拓扑理解有关。基于这些发现，我们提出了 SA-RLVR，将 SCD 指标转化为通过 GRPO 可验证的强化学习奖励。SA-RLVR 成功优化了不同拓扑结构下的结构性寻址：它将 JSON 值放置准确率（Value Placement Accuracy, VPA）从 26% 提升至 63%，同时对保留的模式具有良好的泛化能力；此外，它在表格领域持续推动 VPA 的提升，证明了结构感知奖励能够直接增强多领域的结构定位。
View on arXiv Download PDF AI Translation
cs.AI / 30 / 2608.25399

Can your AI agent be cheaper? Investigating the effects of task specifications on token spend in agentic coding tasks

你的人工智能代理能更便宜吗？任务规范对代理编码任务中令牌支出的影响研究
Smékal, Jakub
Abstract
Agentic coding workflows are now widely deployed in real-world systems. With long-horizon reasoning and tool use, token usage has become an important consideration for both cost and efficiency. Two engineers using AI will solve the same problem differently. How the specification of a task shapes an agent's token spend, and whether that spend can be predicted in advance, are open questions. Here, we study the effects of different task specifications on agentic token spend with the Kimi K3 model at three thinking efforts. Across $2,700$ runs, we show that reducing a full task specification to a bare user story raises token spend by $29.7\%$, while run-to-run variance remains unaffected by any prompt changes. We show that prompt-sensitivity is task-dependent, running from $13\%$ to $115\%$. We fit a simple predictor that can price a full distribution of task specifications and thinking effort configurations from a single cheap probe on an unseen task within $36\%$, improving over prior work in predicting token spend. Our work provides initial results quantifying the effects of task specification on agentic token spend and introduces a method that can be used to systematically evaluate the cost of AI coding workflows.
Chinese Translation
代理编码工作流程现已广泛应用于现实系统中。随着长时间推理和工具使用，令牌使用已成为成本和效率的重要考虑因素。两位工程师使用人工智能解决同一问题的方式不同。任务的规范如何影响代理的令牌支出，以及这种支出是否可以提前预测，仍然是未解的问题。在此，我们研究了不同任务规范对代理令牌支出的影响，使用 Kimi K3 模型在三种思考努力下进行实验。在 $2,700$ 次运行中，我们显示将完整任务规范简化为基本用户故事会使令牌支出增加 $29.7\%$，而运行间的方差不受任何提示变化的影响。我们表明提示敏感性依赖于任务，范围从 $13\\%$ 到 $115\\%$。我们拟合了一个简单的预测器，可以从一个便宜的探测器对未见任务的完整任务规范和思考努力配置进行定价，准确度在 $36\\%$ 以内，优于之前的令牌支出预测工作。我们的研究提供了初步结果，量化了任务规范对代理令牌支出的影响，并引入了一种可以系统评估人工智能编码工作流程成本的方法。
View on arXiv Download PDF AI Translation
cs.AI / 31 / 2608.25417

Paint What You See: Benchmarking Dexterous Visual Tool Use in Multimodal Agents

绘制你所见：多模态智能体灵巧视觉工具使用的基准测试
Liu, Shudong, Chen, Dongyang, Zhang, Enci, Liang, Jinwei, Ma, Zheng, Lu, Lewei
Abstract
Evaluation is shifting from static QA toward agentic settings where models act through external tools. We identify a critical yet underexplored capability within this space - dexterous visual tool use: fine-grained, closed-loop parameterized visual action in which models infer tool parameters from visual evidence, and those parameters directly govern the final result. Existing benchmarks cover web navigation, GUI operation, and software engineering, but rarely target this coupling between visual evidence and execution precision. We propose EASEL, a benchmark evaluating a controlled instance of dexterous visual tool use that adopts reference-guided visual reconstruction as its primary proxy task: the agent incrementally paints a canvas to match a reference image. EASEL additionally includes semantic tasks spanning region annotation, handwriting, and path planning. We further provide EASEL-Data, a 440k-sample two-stage curriculum dataset for trajectory supervision, and EASEL-9B to investigate its effect on this capability. Evaluation of 25 models reveals that current multimodal agents systematically struggle on EASEL. Reconstruction similarity bottlenecks at low levels (0.40-0.54), while trajectory diagnostics expose severe closed-loop instability - models typically saturate early or degrade post-peak. Semantic tasks reveal sharp capability boundaries in precision annotation and path planning. EASEL-9B, trained on EASEL-Data, surpasses the base model by a relative 6.3%, ranking third among all evaluated models.
Chinese Translation
评估正从静态问答转向智能体环境，在这些环境中，模型通过外部工具进行操作。我们识别出这一领域内一个关键但尚未深入探讨的能力——灵巧的视觉工具使用：一种细粒度的、闭环的参数化视觉动作，其中模型从视觉证据中推断工具参数，而这些参数直接决定最终结果。现有基准涵盖了网页导航、图形用户界面操作和软件工程，但很少针对视觉证据与执行精度之间的耦合进行测试。我们提出了EASEL，一个评估灵巧视觉工具使用的受控实例的基准，采用参考引导的视觉重建作为其主要代理任务：智能体逐步绘制画布以匹配参考图像。EASEL还包括跨区域注释、手写和路径规划的语义任务。我们进一步提供了EASEL-Data，一个包含440k样本的两阶段课程数据集用于轨迹监督，以及EASEL-9B以研究其对这一能力的影响。对25个模型的评估显示，当前的多模态智能体在EASEL上系统性地表现不佳。重建相似性在低水平（0.40-0.54）处瓶颈，而轨迹诊断揭示了严重的闭环不稳定性——模型通常在早期饱和或在峰值后退化。语义任务显示出在精确注释和路径规划方面的能力边界。经过EASEL-Data训练的EASEL-9B，相较于基础模型提高了6.3%，在所有评估模型中排名第三。
View on arXiv Download PDF AI Translation
cs.AI / 32 / 2608.25429

Distance Is Not Enough: Forget-Retain Alignment Gap Predicts LLM Relearning Robustness

距离不足：遗忘-保留对齐差距预测大型语言模型的重学习鲁棒性
Chen, Yi, Hsieh, Hanna, Liu, Shuhong, Hua, Chuanbo, Ma, Zihan, Wang, Kun, Kim, Joo-Young
Abstract
Machine unlearning aims to make a model forget specific data, yet unlearned LLMs often fail to stay unlearned: brief fine-tuning can revive removed knowledge. Existing robustness predictors rely on global weight-space displacement, but distance alone can be misleading when random or destructive updates collapse performance. We argue that relearning robustness depends on update structure: robust unlearning should affect forget-critical weights while sparing retain-critical ones. We introduce the Forget-Retain Alignment Gap (FRAG), a training-free predictor that scores an update's forget-retain alignment without running a relearning attack, and separates selective from dense updates more reliably than global distance. Building on the forget-critical, retain-sparing principle, Forget-Retain Pruning (FRP) improves relearning robustness. Our results suggest that weight selectivity better explains robustness than distance alone. Code is available at https://github.com/Yi1-Chen/FRAG.
Chinese Translation
机器遗忘旨在使模型忘记特定数据，但遗忘后的大型语言模型（LLMs）往往无法保持遗忘状态：短暂的微调可以恢复被移除的知识。现有的鲁棒性预测方法依赖于全局权重空间的位移，但仅靠距离在随机或破坏性更新导致性能崩溃时可能会产生误导。我们认为，重学习的鲁棒性依赖于更新结构：鲁棒的遗忘应影响遗忘关键权重，同时保护保留关键权重。我们引入了遗忘-保留对齐差距（Forget-Retain Alignment Gap, FRAG），这是一种无训练的预测器，可以在不进行重学习攻击的情况下评估更新的遗忘-保留对齐，并比全局距离更可靠地区分选择性更新和密集更新。基于遗忘关键、保留保护原则，遗忘-保留修剪（Forget-Retain Pruning, FRP）提高了重学习的鲁棒性。我们的结果表明，权重选择性比单纯的距离更好地解释了鲁棒性。代码可在 https://github.com/Yi1-Chen/FRAG 获取。
View on arXiv Download PDF AI Translation
cs.AI / 33 / 2608.25460

Training Alignment Auditors via Reinforcement Learning

通过强化学习训练对齐审计员
Rosu, Paul, Wang, Rowan
Abstract
Alignment auditing of frontier models increasingly relies on LLM auditors to surface undesirable behaviors at scale, but current automated auditors can struggle with coherent investigation and audit realism. In this work, we improve LLM auditors with reinforcement learning. In our best training environment, the policy investigates target models that potentially possess hidden behaviors planted via their system prompt. An LLM judge, which knows whether the target has a hidden behavior, holistically compares the policy's investigation to a reference investigation to determine the reward. With systematic ablations, we find that pairwise rewards yield more robust training compared to pointwise rewards, and that adding targets without planted behaviors helps maintain a low false positive rate. Training improves investigation quality against targets with planted behaviors, the rate of concerning behaviors surfaced in unmodified production models, and audit realism, while false-positive rates stay below 1%. Furthermore, auditing capabilities generalize across scaffolds: performance on AuditBench's adversarially fine-tuned targets substantially improves [Sheshadri et al., 2026].
Chinese Translation
前沿模型的对齐审计越来越依赖大型语言模型（LLM）审计员以大规模地揭示不良行为，但当前的自动化审计员在连贯调查和审计现实性方面可能面临困难。在本研究中，我们通过强化学习改进了LLM审计员。在我们最佳的训练环境中，策略调查潜在具有通过系统提示植入的隐藏行为的目标模型。一个LLM评审员知道目标是否具有隐藏行为，全面比较策略的调查与参考调查，以确定奖励。通过系统的消融实验，我们发现成对奖励相比于逐点奖励能产生更稳健的训练，而添加没有植入行为的目标有助于维持较低的假阳性率。训练提高了针对具有植入行为的目标的调查质量、在未修改的生产模型中揭示令人担忧的行为的比例，以及审计的现实性，同时假阳性率保持在1%以下。此外，审计能力在不同框架之间具有泛化性：在AuditBench的对抗性微调目标上的表现显著提高 [Sheshadri et al., 2026]。
View on arXiv Download PDF AI Translation
cs.AI / 34 / 2608.25486

PonsRAG: A Pons-Inspired RAG Bridging Cognitive Islands for Coordinated Long Narrative Reasoning

PonsRAG：一种受脑桥启发的RAG框架，连接认知岛屿以实现协调的长篇叙事推理
Zhao, Rongchen, Chen, Yu, Wang, Juyuan, Mo, Zhouting, Yu, Jianxing, Chen, Wenqing, Liu, Jingping
Abstract
Long Narrative Reasoning is an essential capability for processing and reasoning over complex narratives. While retrieval-augmented generation provides a promising framework, existing methods still face two critical challenges: cognitive islanding and cross-layer evidence disconnection. To address these issues, we propose PonsRAG, a coordinated RAG framework inspired by the biological pons. PonsRAG consists of two key components: Triple-Layer Indexing, which organizes documents into a connected knowledge structure to bridge cognitive islands, and Coordinated Reasoning, which retrieves evidence across distinct layers and integrates cross-layer information into a unified context. We evaluate PonsRAG on four long-context narrative benchmarks, and experimental results show that it outperforms the strongest baseline, achieving a 11.56% relative improvement in average accuracy on multi-choice tasks.
Chinese Translation
长篇叙事推理是处理和推理复杂叙事的重要能力。尽管检索增强生成（retrieval-augmented generation）提供了一个有前景的框架，但现有方法仍面临两个关键挑战：认知孤岛和跨层证据断连。为了解决这些问题，我们提出了PonsRAG，这是一种受生物脑桥启发的协调RAG框架。PonsRAG由两个关键组件组成：三层索引（Triple-Layer Indexing），它将文档组织成一个连接的知识结构，以桥接认知孤岛；协调推理（Coordinated Reasoning），它在不同层之间检索证据并将跨层信息整合到统一的上下文中。我们在四个长上下文叙事基准上评估了PonsRAG，实验结果表明，它超越了最强基线，在多选任务中实现了11.56%的平均准确率相对提升。
View on arXiv Download PDF AI Translation
cs.AI / 35 / 2608.25500

CaSKG: Counterfactual-Causal Skill Graphs for Scalable Agent Skill Retrieval

CaSKG：用于可扩展代理技能检索的反事实因果技能图
Li, Zhiyuan, Gao, Linyuan, Ding, Xuechun, Chen, Hongwei, Wu, Yuan, Chang, Yi
Abstract
Reusable skill libraries allow large language model (LLM) agents to reuse procedural knowledge across tasks, but they also turn memory access into a challenging retrieval problem. Full-library prompting preserves coverage at high context cost, vector retrieval returns compact neighborhoods but treats skills as independent text, and graph-based retrieval can recover workflow context only when the edges that carry relevance are reliable. We propose CaSKG, a counterfactual-causal skill graph framework that calibrates procedural relations before retrieval. CaSKG first builds a high-recall directed candidate graph from semantic, lexical, input/output, and structural evidence, with repair evidence and an optional LLM judge further refining candidate scores. It then applies direction-conditioned textual counterfactual probes that remove, substitute, and reorder skill pairs, aggregates the evidence with Bayesian smoothing, and publishes a state-filtered weighted graph for task-conditioned expansion. The graph is constructed offline and used without changing the downstream agent policy or task interface. Across six LLM backbones on ALFWorld ID-140 and ScienceWorld U211, CaSKG achieves the highest task score in all twelve combinations of model and benchmark. Relative to Graph-of-Skills (GoS), it improves the six-model macro-average ScienceWorld score from 72.62 to 80.50 and ALFWorld success from 80.01\% to 86.79\%, while reducing mean environment steps on both benchmarks. Qualitative and ablation analyses further show that calibrated edges help retrieval preserve prerequisites, state-changing actions, verification routines, and final completion steps. These results position edge-confidence calibration as an effective route to compact and executable skill retrieval at scale\footnote{Code is available at: https://github.com/ZhiyuanLi218/Caskg }.
Chinese Translation
可重用的技能库使大型语言模型（LLM）代理能够在任务之间重用程序知识，但这也将内存访问转变为一个具有挑战性的检索问题。全库提示在高上下文成本下保持覆盖率，向量检索返回紧凑的邻域，但将技能视为独立文本，而基于图的检索仅在携带相关性的边缘可靠时才能恢复工作流上下文。我们提出了CaSKG，一个反事实因果技能图框架，在检索之前校准程序关系。CaSKG首先从语义、词汇、输入/输出和结构证据构建一个高召回率的有向候选图，修复证据和可选的LLM评判进一步细化候选分数。然后，它应用方向条件的文本反事实探针，移除、替代和重新排序技能对，使用贝叶斯平滑聚合证据，并发布一个状态过滤的加权图以进行任务条件扩展。该图在离线构建并使用，而无需改变下游代理策略或任务接口。在ALFWorld ID-140和ScienceWorld U211的六个LLM骨干网络中，CaSKG在所有十二种模型和基准组合中实现了最高的任务得分。相较于技能图（Graph-of-Skills，GoS），它将六模型的宏平均ScienceWorld得分从72.62提高到80.50，将ALFWorld的成功率从80.01%提高到86.79%，同时减少了两个基准的平均环境步骤。定性和消融分析进一步表明，校准边缘有助于检索保留前提、状态改变动作、验证例程和最终完成步骤。这些结果将边缘置信度校准定位为实现大规模紧凑且可执行技能检索的有效途径。
View on arXiv Download PDF AI Translation
cs.AI / 36 / 2608.25518

Agentic Game Development as a Verifiable Trajectory Data Engine for Scaling World Models

作为可验证轨迹数据引擎的自主游戏开发以扩展世界模型
Zhou, Pengfei, Wang, Hexin, Zhang, Zhengfeiyang, Ma, Yixing, Wan, Zhenglin, Zhang, Kaipeng, Zhao, Wangbo, You, Yang
Abstract
A common strategy for scaling world models is to train on more crawled video with more compute. We argue that this strategy is inefficient: scaling world models also requires a recursive data engine that offers grounded reward signals. The success of code agents illustrates why this matters. As code is executable, compilers and runtimes can provide high-quality rewards for Reinforcement Learning (RL) post-training of LLMs. By contrast, spatial generation still relies largely on fuzzy proxies such as CLIP scores. These signals are fuzzy and biased, making them hard to support RL post-training. Compared with these, game development provides a missing reward environment for spatial world models. A scene encoded by a game engine is an executable world specification: the engine can efficiently check collision, physics, navigability and bounded playability, while the developer provides the global verification signal by judging whether the scene should be accepted. Game development also provides real-world long-horizon trajectory data for RL post-training. We therefore propose Reinforcement Learning with Human-Engine Verification (RLHEV), a post-training paradigm that combines dense engine signals with implicit human acceptance feedback from the development process.
Chinese Translation
扩展世界模型的一个常见策略是利用更多的爬取视频和计算资源进行训练。我们认为这一策略效率低下：扩展世界模型还需要一个递归数据引擎，以提供有根据的奖励信号。代码代理的成功说明了这一点的重要性。由于代码是可执行的，编译器和运行时可以为大规模语言模型（LLMs）的强化学习（RL）后训练提供高质量的奖励。相比之下，空间生成仍然在很大程度上依赖模糊的代理，例如 CLIP 分数。这些信号模糊且有偏差，使得它们难以支持 RL 后训练。与此相比，游戏开发为空间世界模型提供了缺失的奖励环境。由游戏引擎编码的场景是一个可执行的世界规范：引擎可以高效地检查碰撞、物理、可导航性和有限的可玩性，而开发者通过判断场景是否应该被接受来提供全局验证信号。游戏开发还为 RL 后训练提供了现实世界的长期轨迹数据。因此，我们提出了带有人机引擎验证的强化学习（RLHEV），这是一种将密集引擎信号与开发过程中的隐含人类接受反馈相结合的后训练范式。
View on arXiv Download PDF AI Translation
cs.AI / 37 / 2608.25623

Using profiles of cognitive capability to assess AI suitability for workplace tasks

利用认知能力档案评估人工智能在工作任务中的适用性
Prunty, Jonathan, Tešić, Marko, Quinn, Patrick, Hernández-Orallo, José, Cheke, Lucy
Abstract
Organisations deploying AI face a scoping problem: which tasks can be automated, which should remain with humans, and which are best shared between the two. Aggregate benchmark scores provide little insight into where systems will succeed or fail in practice, while human judgements of model capabilities quickly become outdated. We introduce a pipeline that profiles agents and tasks using a shared set of core cognitive capabilities. Cognitive capability profiling infers an agent's capabilities from performance on a benchmark battery annotated for the cognitive demands of each item. Task requirements weighting elicits from domain experts the relative importance of these same capabilities for their work. As both use a common set of cognitive dimensions, they can be updated independently as models and roles change, and combined to estimate AI suitability at the level of a domain, organisation, role, or individual duty. We validate capability recovery on synthetic agents, profile six AI systems, and elicit task requirements from 410 employees across six occupational domains. AI systems differed more across cognitive dimensions than across model families, while workplace activities converged on a shared cognitive core. The resulting scores provide a comparative scoping tool for identifying promising candidates for piloting and areas where current systems are unlikely to be well suited. We discuss extending the framework to profile human workers alongside AI systems, moving from AI suitability towards human-machine task allocation.
Chinese Translation
在部署人工智能的组织面临一个范围界定问题：哪些任务可以自动化，哪些任务应由人类完成，以及哪些任务最适合人机协作。综合基准分数对系统在实际应用中的成功与失败提供的洞察有限，而人类对模型能力的判断很快就会过时。我们引入了一种管道，通过一组共享的核心认知能力对代理和任务进行档案分析。认知能力档案通过在标注了每个项目认知需求的基准测试中的表现推断代理的能力。任务需求权重则从领域专家那里获取这些能力在其工作中的相对重要性。由于两者使用相同的认知维度集，因此可以在模型和角色变化时独立更新，并结合估算在特定领域、组织、角色或个别职责层面的人工智能适用性。我们在合成代理上验证了能力恢复，分析了六个人工智能系统的档案，并从六个职业领域的410名员工那里获取了任务需求。人工智能系统在认知维度上的差异大于模型家族之间的差异，而工作活动则趋向于共享的认知核心。由此产生的分数为识别适合试点的有前景候选者和当前系统不太适合的领域提供了一种比较范围界定工具。我们讨论了扩展该框架以同时对人类工人和人工智能系统进行档案分析，从而从人工智能适用性转向人机任务分配。
View on arXiv Download PDF AI Translation
cs.AI / 38 / 2608.25657

Narcissus: Program Synthesis Using Context-Aware LLM Approximations

水仙花：使用上下文感知的LLM近似进行程序合成
Hinnerichs, Tilman, Dumancic, Sebastijan, Yorke-Smith, Neil
Abstract
Large language models (LLMs) excel at programming, but not when the task fixes the target language: prompted with a grammar rare in their training data, their programs usually break the grammar or fail the given specification. Enumerative synthesizers search the space of syntactically correct programs systematically guided by LLMs; the state of the art guides them by approximating LLM proposals into rule frequencies, which loses where each construct belongs and prunes every rule the proposals miss, exactly when the proposals are wrong. We present Narcissus, a synthesizer that keeps the proposals as syntax trees and scores each expansion of a candidate program in its context: does a proposal with the same surrounding structure continue the same way, and does the expansion rebuild a fragment the proposals repeat? A regularization term keeps every rule reachable, so wrong proposals delay the solution but cannot hide it. Across five domains and two search backends, Narcissus beats static guidance at every budget and consistently outperforms re-prompting the LLM to fix its own proposals; it reaches proposal-like programs an order of magnitude sooner and solves $40\%$ of ARC tasks where the raw proposals solve $13\%$, all without a single LLM call during search.
Chinese Translation
大型语言模型（LLMs）在编程方面表现出色，但当任务固定目标语言时表现不佳：当提示的语法在其训练数据中较为稀少时，它们生成的程序通常会违反语法或未能满足给定的规范。枚举合成器通过LLMs系统性地搜索语法正确程序的空间；当前的技术通过将LLM的提议近似为规则频率来指导它们，这在提议错误时会丢失每个构造的归属，并修剪掉提议遗漏的每条规则。我们提出了水仙花（Narcissus），一种合成器，它将提议保留为语法树，并在其上下文中对候选程序的每次扩展进行评分：具有相同周围结构的提议是否以相同方式继续，扩展是否重建提议重复的片段？一个正则化项保持每条规则可达，因此错误的提议会延迟解决方案，但无法隐藏它。在五个领域和两个搜索后端中，水仙花在每个预算上都超越了静态指导，并且始终优于重新提示LLM以修正其自身提议；它在数量级上更快地达到了类似提议的程序，并解决了$40\%$的ARC任务，而原始提议仅解决了$13\\%$，所有这些在搜索过程中没有进行一次LLM调用。
View on arXiv Download PDF AI Translation
cs.AI / 39 / 2608.25775

ToST: A Tree-of-Thought Socratic Teaching Framework for Multi-Path Guidance and Parallel Thinking

ToST：一种树状思维苏格拉底教学框架，用于多路径指导和并行思维
Ling, Feng, Yu, Heng
Abstract
Large Language Models (LLMs) exhibit strong problem-solving abilities, positioning them as promising agents for Socratic teaching to guide students through step-by-step heuristic questioning. However, existing approaches typically adopt a one-problem-one-solution paradigm, restricting the teaching guidance to a single linear reasoning path. This design limits instructional flexibility, weakens error recovery, and restricts students' ability to engage in parallel thinking to explore multiple valid solutions. To overcome these, we propose ToST, a Tree-of-Thought Socratic Teaching framework that explicitly supports multi-path guidance under a one-problem-multiple-solutions paradigm. ToST employs Parallel Sowing, a parallel-thinking-oriented questioning strategy to encourage students to approach problems from diverse perspectives, and a Multi-Path Adaptive Guidance mechanism to provide more robust and non-linear instructions across alternative solution trajectories. Concurrently, to fill the void in systematically evaluating such non-linear instructional capabilities, we advance the task of multi-path Socratic guidance by establishing MPSG-Bench, a comprehensive benchmark that includes a dataset of 31K multi-path teaching dialogues and a five-dimensional evaluation framework grounded in the SOLO (Structure of Observed Learning Outcomes) theory to assess parallel-thinking guidance. Experimental results demonstrate that ToST significantly enhances guidance success rates while empowering students to navigate and explore multiple solution paths more effectively under both automatic and human metrics.
Chinese Translation
大型语言模型（LLMs）展现出强大的问题解决能力，使其成为有前景的苏格拉底教学代理，能够通过逐步的启发式提问引导学生。然而，现有的方法通常采用一问题一解决方案的范式，限制了教学指导仅限于单一线性推理路径。这种设计限制了教学灵活性，削弱了错误恢复能力，并限制了学生进行并行思维以探索多种有效解决方案的能力。为了解决这些问题，我们提出了ToST，一种树状思维苏格拉底教学框架，明确支持在一问题多解决方案的范式下进行多路径指导。ToST采用并行播种（Parallel Sowing），一种面向并行思维的提问策略，以鼓励学生从不同角度接近问题，并通过多路径自适应指导机制提供更强大且非线性的指导，涵盖替代解决方案轨迹。同时，为了填补系统评估这种非线性教学能力的空白，我们通过建立MPSG-Bench，推进多路径苏格拉底指导的任务，这是一项综合基准，包含31K个多路径教学对话的数据集，以及基于SOLO（观察学习结果结构）理论的五维评估框架，以评估并行思维指导。实验结果表明，ToST显著提高了指导成功率，同时使学生在自动和人工评估指标下更有效地导航和探索多种解决路径。
View on arXiv Download PDF AI Translation
cs.AI / 40 / 2608.25777

LocalLSTC: A Long Short-Term Control Architecture for Locally Deployed GUI Agents

LocalLSTC：一种用于本地部署GUI代理的长短期控制架构
Li, Weiming, Paik, Helen, Sui, Yulei
Abstract
Modern GUI-agent frameworks achieve strong desktop task performance with frontier API models, yet persistent control information often remains implicit in growing interaction trajectories. At each step, the planner reconstructs the active task stage, accumulated evidence, and runtime feedback before deciding the next action. This dependence becomes more pronounced under weaker local reasoning backbones. Across four representative state-of-the-art frameworks, replacing GPT-5 with Qwen3.5-9B reduces average OSWorld SR-100 from 60.9\% to 37.7\%. Trajectory annotation further identifies at least one control failure in 91.6\% of failed trajectories. To address this problem, we introduce LocalLSTC, a training-free architecture that organizes control by temporal scope, maintaining persistent cross-step state to guide short-term execution commitments. Long-Term Control maintains the active subgoal, subgoal-aligned evidence, and runtime feedback across interactions, while Short-Term Execution realizes bounded commitments for the current step. Long-to-Short Planning forms each commitment from persistent state, and Short-to-Long Control integrates execution outcomes back into that state for progress assessment, recovery, and termination. With Qwen3.6-27B, LocalLSTC reaches 64.7\% SR-100 on OSWorld and 65.3\% on WindowsAgentArena, outperforming the strongest prior local results on both benchmarks. Ablations further support contributions from mechanisms on both sides of execution. These findings identify temporal organization of control information as a distinct architectural dimension for locally deployed GUI agents.
Chinese Translation
现代GUI代理框架通过前沿API模型在桌面任务中实现了强大的性能，但持续的控制信息往往在不断增长的交互轨迹中保持隐性。在每一步中，规划者重建当前任务阶段、累积证据和运行时反馈，然后决定下一步行动。这种依赖在较弱的本地推理基础上变得更加明显。在四个具有代表性的最先进框架中，用Qwen3.5-9B替换GPT-5使得平均OSWorld SR-100从60.9\%降至37.7\%。轨迹注释进一步发现，在91.6\%的失败轨迹中至少存在一个控制失败。为了解决这个问题，我们提出了LocalLSTC，这是一种无训练架构，通过时间范围组织控制，保持持续的跨步骤状态以指导短期执行承诺。长期控制保持活跃的子目标、与子目标对齐的证据和跨交互的运行时反馈，而短期执行则实现当前步骤的有界承诺。长到短的规划从持续状态形成每个承诺，而短到长的控制将执行结果整合回该状态，以进行进度评估、恢复和终止。使用Qwen3.6-27B，LocalLSTC在OSWorld上达到了64.7\%的SR-100，在WindowsAgentArena上达到了65.3\%，在这两个基准上超越了之前最强的本地结果。消融实验进一步支持了执行两侧机制的贡献。这些发现将控制信息的时间组织识别为本地部署GUI代理的一个独特架构维度。
View on arXiv Download PDF AI Translation
cs.AI / 41 / 2608.25917

Choose Your Game Wisely: Measuring Game-Theoretic Structures in Real-World Vehicle Interactions

明智选择你的游戏：测量现实世界中车辆交互的博弈论结构
Li, Yueyuan, Nie, Rongcheng, Xi, Weijie, Jiang, Mingyang, Zhang, Songan, Zhuang, Hanyang, Yang, Ming
Abstract
Game-theoretic models provide principled frameworks for modeling vehicle interactions, but their underlying temporal assumptions have not been systematically examined against real-world driving behavior. In particular, it remains unclear how simultaneous, sequential, and asymmetric interaction structures can be measured from vehicle trajectories. This paper develops a trajectory-based interaction measurement framework to identify interaction events and quantify behavioral change onset, temporal organization, post-onset response dynamics, and ordering stability. The framework uses behavioral deviations to verify candidate interactions. We evaluate the framework on six real-world trajectory datasets, including INTERACTION, highD, inD, rounD, Waymo Open Motion, and nuPlan, covering diverse road geometries, traffic environments, and interaction types. The results show that concurrent and sequential behavioral changes both constitute substantial proportions of observed following, merging, and conflicting interactions. Among sequential interactions, stable ordering is more prevalent than alternating ordering, indicating that persistent asymmetric roles are a common interaction structure. Importantly, temporal precedence does not necessarily coincide with a measurable behavioral response, indicating that temporal ordering alone may not be sufficient to characterize behavioral dependence. These findings show that real-world interactions exhibit concurrent, sequential, and persistently ordered temporal structures. Different game-theoretic formulations are therefore better regarded as complementary modeling abstractions for different interaction regimes rather than as a universal structure governing all vehicle interactions.
Chinese Translation
博弈论模型为建模车辆交互提供了原则性框架，但其基础的时间假设尚未与现实世界的驾驶行为进行系统性检验。特别是，如何从车辆轨迹中测量同时、顺序和不对称的交互结构仍不清楚。本文开发了一种基于轨迹的交互测量框架，以识别交互事件并量化行为变化的开始、时间组织、开始后的响应动态和顺序稳定性。该框架利用行为偏差来验证候选交互。我们在六个现实世界的轨迹数据集上评估该框架，包括 INTERACTION、highD、inD、rounD、Waymo Open Motion 和 nuPlan，涵盖了多样的道路几何、交通环境和交互类型。结果表明，同时和顺序的行为变化都占据了观察到的跟随、并入和冲突交互的相当比例。在顺序交互中，稳定的顺序比交替的顺序更为普遍，表明持久的不对称角色是一种常见的交互结构。重要的是，时间优先并不一定与可测量的行为响应相吻合，这表明仅靠时间顺序可能不足以表征行为依赖性。这些发现表明，现实世界中的交互表现出同时、顺序和持久有序的时间结构。因此，不同的博弈论表述应被视为针对不同交互模式的互补建模抽象，而不是作为支配所有车辆交互的普遍结构。
View on arXiv Download PDF AI Translation
cs.AI / 42 / 2608.25920

Repair or Resample? Rethinking Failure Debugging in LLM Multi-Agent Systems

修复还是重采样？重新思考大型语言模型多智能体系统中的故障调试
Luan, Zhongwen, Zhang, Xiaoyu, Hu, Ming, Yang, Yue, Yu, Jiongchi, Chen, Xiaohong
Abstract
As large language model (LLM)-based multi-agent systems (MASs) are increasingly applied to long-horizon complex tasks, their reliability has emerged as the core bottleneck hindering their real-world deployment. Existing MAS debugging and repair methods typically rely on rerunning and resampling the entire execution trajectory. However, a fundamental question remains to be answered: do these methods causally repair MAS failures or merely stochastically repair by leveraging the randomness of LLM sampling? To evaluate the effectiveness of MAS repair methods, we introduce SymTrace, a controlled evaluation framework that records the MAS execution trajectory and establishes intervention anchors. During replay, it effectively reconstructs the execution before the anchor using recorded logs and only regenerates the downstream trajectory, thereby enabling the reliable reproduction of MAS failures. We further construct the dataset SymFail, comprising 536 human-annotated failure trajectories with graph-linked locations, categories, and trace evidence. Based on these foundations, we conduct a large-scale empirical study across three mainstream MAS frameworks. Our findings reveal that existing unguided rerun methods are highly unreliable, exhibiting low failure reproduction and repair rates (only 67.97% and 6.90%, respectively). Building upon these findings, we further explore the effectiveness of a symptom-driven intervention method, which successfully repairs 20.15% of the failed cases (a 191.89% improvement to state-of-the-art repair methods). This study aims to provide actionable insights for MAS debugging and repair research, paving the way for the robust deployment of multi-agent systems.
Chinese Translation
随着基于大型语言模型（LLM）的多智能体系统（MAS）在长期复杂任务中的应用日益增多，其可靠性已成为阻碍其在现实世界中部署的核心瓶颈。现有的MAS调试和修复方法通常依赖于重新运行和重采样整个执行轨迹。然而，一个根本性的问题仍需回答：这些方法是因果性地修复MAS故障，还是仅仅通过利用LLM采样的随机性进行随机修复？为了评估MAS修复方法的有效性，我们引入了SymTrace，一个控制评估框架，记录MAS执行轨迹并建立干预锚点。在重放过程中，它有效地使用记录的日志重构锚点之前的执行，并仅重新生成下游轨迹，从而实现MAS故障的可靠再现。我们进一步构建了数据集SymFail，包含536个带有图形链接位置、类别和追踪证据的人类标注故障轨迹。在这些基础上，我们在三个主流MAS框架中进行了大规模的实证研究。我们的研究发现，现有的无指导重跑方法极不可靠，故障再现率和修复率均较低（分别仅为67.97%和6.90%）。基于这些发现，我们进一步探索了一种以症状驱动的干预方法，该方法成功修复了20.15%的故障案例（相比于最先进的修复方法提高了191.89%）。本研究旨在为MAS调试和修复研究提供可操作的见解，为多智能体系统的稳健部署铺平道路。
View on arXiv Download PDF AI Translation
cs.AI / 43 / 2608.25926

Formal, Executable and Explainable Runtime Monitoring of Spoken Air Traffic Control Operational Procedures

正式、可执行且可解释的语音空中交通管制操作程序运行时监控
Luvini, Roberto, Longo, Giacomo, Armando, Alessandro, Russo, Enrico
Abstract
Air traffic control procedures are executed through spoken exchanges between controllers and pilots. These interactions are essential to the safety of air transportation: failures in their execution can create severe operational hazards, as evidenced by past fatal accidents. Assessing whether an instruction has been followed requires relating what was said to the aircraft concerned, its state, and the obligations that pilots must meet. We present a runtime verification framework that monitors such procedures by checking controller-pilot exchanges, surveillance data, and onboard observations. The framework parses radio communications into events linked to the entities they concern and merges them with surveillance and onboard observations into a time-stamped trace. The ICAO-derived obligations as formalized as temporal formulas with explicit time bounds and evaluated over execution traces. Every violation is reported along with the breached obligations and the observations that support the verdict. With real traffic, the complete pipeline reaches an F1 of 0.85 against blind human-annotated violations; in 1,495 synthetic situations derived from two public corpora, the monitor logic returns the expected verdict in every case. In two historical accidents reconstructed from official investigation reports, the monitor identifies the same procedural deviations documented by the investigators.
Chinese Translation
空中交通管制程序通过管制员与飞行员之间的语音交流进行执行。这些互动对航空运输的安全至关重要：其执行中的失败可能导致严重的操作危险，过去的致命事故便是明证。评估指令是否被遵循需要将所说内容与相关飞机、其状态以及飞行员必须满足的义务联系起来。我们提出了一种运行时验证框架，通过检查管制员与飞行员的交流、监视数据和机载观察来监控这些程序。该框架将无线电通信解析为与其相关实体关联的事件，并将其与监视和机载观察合并为时间戳追踪。根据国际民航组织（ICAO）制定的义务，这些义务被形式化为具有明确时间界限的时间公式，并在执行追踪中进行评估。每次违规都会报告违规情况、被违反的义务以及支持该裁决的观察结果。在真实交通中，完整的管道在盲目人工标注的违规情况中达到了0.85的F1值；在从两个公共语料库派生的1,495个合成情况中，监控逻辑在每种情况下都返回了预期的裁决。在从官方调查报告重建的两起历史事故中，监控器识别出调查人员记录的相同程序偏差。
View on arXiv Download PDF AI Translation
cs.AI / 44 / 2608.25934

How Robust Are Automated Fact-Checking Systems? A Cross-Benchmark Evaluation

自动化事实核查系统的稳健性如何？跨基准评估
Usmanova, Aida, Iklassov, Zangir, Leippold, Markus, Usbeck, Ricardo
Abstract
Automated fact-checking (AFC) systems retrieve evidence and predict claim veracity, yet evaluations omit simple baselines, systems are developed for a single benchmark and cannot be trusted to generalise across domains. No prior work cross-evaluates the full two-stage retrieve-then-verify pipeline across diverse datasets, complementing retrieval-only studies (Thakur et al., 2021) and single-stage benchmarking studies (Calamai et al., 2025). We benchmark nine models, ranging from random and sparse baselines to fine-tuned transformers, zero-shot LLMs, and the two highest-ranked systems from the AVeriTeC 2025 shared task, across four datasets spanning scientific, open-web, and climate domains. Three findings stand out: (1) on ClimateCheck claim-only and fine-tuned models outperform zero-shot LLM and top-performing AVeriTeC 2025 systems, highlighting that noisy evidence can degrade veracity prediction; (2) system rankings are strongly domain- and metric-dependent: the best model on SciFact (macro-F1 0.70) drops to 0.31 on ClimateCheck, while the AVeriTeC 2025 winner and runner-up swap rankings based on evaluation metrics and datasets; (3) replacing retrieved evidence with gold annotations improves veracity accuracy by 14-22 points across models, confirming retrieval remains primary bottleneck. We release code, pre-processed datasets, and all results to support reproducible AFC research.
Chinese Translation
自动化事实核查（AFC）系统检索证据并预测声明的真实性，但评估中忽略了简单的基线，系统通常针对单一基准开发，无法保证跨领域的泛化能力。之前的研究没有对多样化数据集中的完整两阶段检索-验证流程进行交叉评估，这补充了仅检索的研究（Thakur et al., 2021）和单阶段基准研究（Calamai et al., 2025）。我们对九个模型进行了基准测试，这些模型涵盖了从随机和稀疏基线到微调的变换器、零样本大语言模型（LLMs）以及来自AVeriTeC 2025共享任务的两个最高排名系统，评估了四个数据集，涵盖科学、开放网络和气候领域。三个发现尤为突出：（1）在ClimateCheck声明数据集中，微调模型的表现优于零样本LLM和表现最佳的AVeriTeC 2025系统，突显了噪声证据可能降低真实性预测的能力；（2）系统排名强烈依赖于领域和评估指标：在SciFact上的最佳模型（宏F1 0.70）在ClimateCheck上的表现下降至0.31，而AVeriTeC 2025的冠军和亚军根据评估指标和数据集的不同而互换排名；（3）用黄金注释替换检索到的证据使各模型的真实性准确性提高了14-22个百分点，确认了检索仍然是主要瓶颈。我们发布了代码、预处理的数据集和所有结果，以支持可重复的AFC研究。
View on arXiv Download PDF AI Translation
cs.AI / 45 / 2608.25937

Candidate supply and answer selection shape the value of LLM judging in multi-agent systems

候选供应和答案选择塑造多智能体系统中大型语言模型判断的价值
Ji, Jia-Hao, Li, Sijie, Cheng, Jiabei, She, Zixi, Yu, Jin-Tai, Yuan, Zhiyuan
Abstract
Multi-agent systems (MAS) sometimes already have the potential to answer correctly, but still report a wrong answer. Explaining this outcome is difficult because generation, communication and final answer-selection rules usually change simultaneously. We conceptualize multi-agent reasoning as an evolutionary pipeline of candidate generation, peer communication and terminal selection, wherein consensus without quality control can exhibit patterns of memetic drift. We study two questions: (1) when an LLM judge provides effective selection pressure by supplying a signal of answer correctness for candidates generated in a multi-agent system, and (2) when using that signal improves the reported answer. To map judge reliability, we analysed 15,336 questions from MMLU-Pro, GPQA, MedXpertQA and MuSR, with Humanity's Last Exam analysed separately. To test these rules, we replayed 81,390 fixed candidate pools drawn from 16,278 questions across five benchmarks. We report three findings. (1) A correct answer is often already present among the generated candidates, but the system can still converge on and report a wrong answer. (2) Judge reliability is not a fixed trait of the model, but varies with the task, the generator and how rare the correct answer is. (3) Combining answer frequency with the judge's evaluation changed only the final answer-selection rule and raised accuracy from 63.82% to 70.82-70.95%, primarily by rescuing correct answers that were outnumbered by popular errors. In the systems studied here, the value of generating more candidates depends on whether those extra samples make correct answers present, frequent or recognisable. By isolating generation, recognition and selection, these findings establish a diagnostic basis for designing multi-agent architectures that protect generated correct answers from being lost.
Chinese Translation
多智能体系统（MAS）有时已经具备正确回答的潜力，但仍然报告错误的答案。解释这一结果是困难的，因为生成、沟通和最终答案选择规则通常是同时变化的。我们将多智能体推理概念化为一个进化管道，包括候选生成、同行沟通和终端选择，其中缺乏质量控制的共识可能表现出模因漂移的模式。我们研究了两个问题：（1）当大型语言模型（LLM）评判者通过为在多智能体系统中生成的候选提供答案正确性的信号来施加有效的选择压力时，以及（2）使用该信号是否改善了报告的答案。为了映射评判者的可靠性，我们分析了来自MMLU-Pro、GPQA、MedXpertQA和MuSR的15,336个问题，同时单独分析了人类最后的考试。为了测试这些规则，我们重放了从五个基准中抽取的16,278个问题的81,390个固定候选池。我们报告了三项发现：（1）在生成的候选中，正确答案通常已经存在，但系统仍然可能收敛并报告错误答案。（2）评判者的可靠性并不是模型的固定特征，而是随着任务、生成器和正确答案的稀有程度而变化。（3）将答案频率与评判者的评估结合，仅改变了最终答案选择规则，并将准确率从63.82%提高到70.82-70.95%，主要是通过挽救被流行错误所淹没的正确答案。在这里研究的系统中，生成更多候选的价值取决于这些额外样本是否使正确答案变得存在、频繁或可识别。通过隔离生成、识别和选择，这些发现为设计保护生成的正确答案不被丢失的多智能体架构建立了诊断基础。
View on arXiv Download PDF AI Translation
cs.AI / 46 / 2608.25960

LivingRAG: Augmenting Graph RAG with Experience

LivingRAG：通过经验增强图形RAG
Cui, Yuzhuo, Zhang, Zongye, Liu, Qingjie
Abstract
Graph-based RAG improves multi-hop question answering by organizing evidence as a knowledge graph. However, most existing RAG systems process each query in isolation and discard useful reasoning from the LLM's response after inference. As a result, later related queries need to retrieve evidence and reason from scratch. We propose LivingRAG, a Graph RAG framework with writable and reusable reasoning experience. LivingRAG adds a writable experience store to a graph-based retrieval backbone, enabling verified experiences to be reused during inference in two ways. Stored graph signals help retrieval find entities and passages that were useful in earlier related queries. Stored summaries provide a reference reasoning pattern for answer generation. We analyze online QA streams and find reusable signals from shared entities, graph neighborhoods, and question templates. Experiments on multi-hop QA benchmarks show that LivingRAG improves accuracy over strong RAG baselines and reduces completion-token use when relevant prior experience is reused.
Chinese Translation
基于图形的RAG通过将证据组织为知识图谱来改善多跳问答。然而，大多数现有的RAG系统在处理每个查询时都是孤立进行的，并在推理后丢弃来自大型语言模型（LLM）响应中的有用推理。因此，后续相关查询需要从头开始检索证据和进行推理。我们提出了LivingRAG，一个具有可写和可重用推理经验的图形RAG框架。LivingRAG在基于图形的检索骨干上添加了一个可写的经验存储，使得在推理过程中可以以两种方式重用经过验证的经验。存储的图形信号帮助检索找到在早期相关查询中有用的实体和段落。存储的摘要为答案生成提供了参考推理模式。我们分析了在线问答流，并发现来自共享实体、图形邻域和问题模板的可重用信号。对多跳问答基准的实验表明，LivingRAG在准确性上优于强大的RAG基线，并在重用相关先前经验时减少了完成令牌的使用。
View on arXiv Download PDF AI Translation
cs.AI / 47 / 2608.25968

Quantitative Analysis of $\omega$-Regular Robust MDPs

对$ ext{ω}$-正则鲁棒马尔可夫决策过程的定量分析
Asadi, Ali, Chatterjee, Krishnendu, Goharshady, Ehsan Kafshdar, Karrabi, Mehrdad, Montaseri, Alipasha, Shafiee, Ali
Abstract
Robust Markov Decision Processes (RMDPs) generalize classical MDPs by allowing uncertainty in transition probabilities and optimizing against their worst-case realization. We consider $(s,a)$-rectangular RMDPs with \emph{linearly defined} uncertainty sets and study parity objectives, which are a canonical representation of $\omega$-regular objectives. An uncertainty set is linearly defined if it is described by linear inequalities over the transition distribution together with auxiliary variables, which capture the standard $L_1$ and $L_\infty$ balls as well as general polytopic uncertainty sets. The quantitative value is the supremum, over all agent policies, of the satisfaction probability guaranteed against the adversarial environment. Previous work studied the qualitative analysis, namely the almost-sure (resp. positive) problem that asks whether a single agent policy guarantees satisfaction with probability one (resp. positive probability) against every environment policy. In this work, we solve the exact quantitative problem. Our contributions are threefold. First, we show that both the agent and the environment admit pure memoryless optimal policies. Second, we give a polynomial-time algorithm for quantitative parity on linearly defined robust Markov chains and use it as a subroutine in a policy-iteration algorithm for RMDPs. The algorithm combines quantitative one-step improvements with qualitative almost-sure improvements. Finally, we report experiments comparing our approach with the explicit reduction to stochastic games.
Chinese Translation
鲁棒马尔可夫决策过程（RMDPs）通过允许转移概率的不确定性并针对其最坏情况实现进行优化，从而推广了经典的马尔可夫决策过程（MDPs）。我们考虑具有 extit{线性定义}不确定性集的$(s,a)$-矩形RMDPs，并研究奇偶目标，这是一种$ ext{ω}$-正则目标的典型表示。不确定性集如果通过关于转移分布的线性不等式以及捕获标准$L_1$和$L_ ext{∞}$球体的辅助变量来描述，则称其为线性定义。不确定性的定量值是所有代理策略中，针对对抗环境所保证的满足概率的上确界。之前的研究主要集中在定性分析上，即几乎必然（或正）问题，询问单一代理策略是否能够以概率1（或正概率）保证在每个环境策略下的满足。在本研究中，我们解决了精确的定量问题。我们的贡献有三方面。首先，我们证明代理和环境均可接受纯记忆无关的最优策略。其次，我们为线性定义的鲁棒马尔可夫链上的定量奇偶问题提供了一种多项式时间算法，并将其作为RMDPs中策略迭代算法的子程序。该算法结合了定量的一步改进和定性的几乎必然改进。最后，我们报告了与显式减少到随机博弈的比较实验。
View on arXiv Download PDF AI Translation
cs.AI / 48 / 2608.25973

SciMIF: Understanding Multimodal Instruction Following in Scientific Domains

SciMIF：理解科学领域中的多模态指令遵循
Shen, Ye, Zheng, Yuting, Pei, Dun, Chen, Zijian, Zhang, Wenlong, Jia, Qi, Zhai, Guangtao
Abstract
Understanding instruction-following capabilities in scientific domains is essential for effectively leveraging Multimodal Large Language Models (MLLMs) to advance the development of scientific fields. In this work, we introduce SciMIF, a novel benchmark designed to evaluate the capability of MLLMs in following complex scientific instructions. Specifically, based on an extensive analysis of 22 distinct tasks across 5 representative scientific disciplines, we propose a comprehensive taxonomy comprising 10 constraint groups that captures both general functional requirements and discipline-specific characteristics. Guided by this taxonomy, we develop a high-fidelity instruction injection pipeline to systematically augment existing scientific datasets. We conduct comprehensive experiments on multiple state-of-the-art closed-source and open-source MLLMs. Our findings reveal significant performance disparities across different scientific disciplines, with chemistry posing greater challenges for current MLLMs. Furthermore, we observe that increasing the model scale does not yield corresponding improvements in constraint adherence, and current models still struggle severely with fine-grained constraints and instructions requiring the deep application of disciplinary knowledge. SciMIF fills the current void in evaluating multimodal instruction adherence within scientific domains, laying a crucial foundation for future enhancements of MLLMs in rigorous scientific applications. Data and code will be released at https://github.com/shenye7436/SciMIF .
Chinese Translation
理解科学领域中的指令遵循能力对于有效利用多模态大型语言模型（MLLMs）推动科学领域的发展至关重要。在本研究中，我们介绍了SciMIF，一个旨在评估MLLMs遵循复杂科学指令能力的新基准。具体而言，基于对5个代表性科学学科中22个不同任务的广泛分析，我们提出了一个包含10个约束组的综合分类法，捕捉了通用功能需求和学科特定特征。在这一分类法的指导下，我们开发了一个高保真指令注入管道，以系统性地增强现有的科学数据集。我们在多个最先进的闭源和开源MLLMs上进行了全面实验。我们的研究结果揭示了不同科学学科之间显著的性能差异，其中化学领域对当前MLLMs提出了更大的挑战。此外，我们观察到，增加模型规模并未带来相应的约束遵循改进，当前模型在细粒度约束和需要深度应用学科知识的指令上仍然面临严重困难。SciMIF填补了当前在科学领域中评估多模态指令遵循的空白，为未来在严格科学应用中增强MLLMs奠定了重要基础。数据和代码将发布在 https://github.com/shenye7436/SciMIF 。
View on arXiv Download PDF AI Translation
cs.AI / 49 / 2608.25986

Multi-Granularity Context-Enhanced RAG over Multimodal Knowledge Graphs

基于多粒度上下文增强的多模态知识图谱检索增强生成（RAG）
Wu, Zongyu, Wang, Yilong, Wang, Xiaochen, Lin, Minhua, Xu, Zhichao, Ma, Fenglong, Zhang, Xiang, Wang, Suhang
Abstract
Retrieval-augmented generation (RAG) is widely used to mitigate hallucination issues in large language models (LLMs) and multimodal large language models (MLLMs). In particular, knowledge graph (KG)-based RAG leverages structured knowledge to provide (M)LLMs with high-quality external information. Building on these works, recent studies have explored multimodal knowledge graphs (MMKGs) as knowledge bases for GraphRAG. This enables Graph RAG to integrate knowledge across multiple modalities, thereby further enhancing its performance. However, existing MMKG-based RAG methods generally follow a common pipeline in which different modalities are largely processed independently before being fusion. As a result, textual context is only used to a limited extent during visual information extraction and subsequent multimodal knowledge fusion. This brings a semantic gap between images and text which limits the multimodal GraphRAG performance. To address this issue, we propose a novel framework for constructing a Context-Enhanced MMKG (CEMMKG) to better support multimodal GraphRAG. The proposed CEMMKG enriches each image with complementary textual context at both local and global scopes. Local context goes beyond the surrounding text by incorporating sentences that are semantically related to the image, while global context provides a summary of the entire passage. We further introduce a multi-granularity design for the local context, allowing it to capture semantically relevant information at different levels of detail. Extensive experiments on the selected vision-centric dataset validate that CEMMKG is effective in leveraging contextual information to improve MMKG-based RAG performance. Moreover, its effectiveness across different MMKG-based RAG methods demonstrates its broad applicability.
Chinese Translation
检索增强生成（RAG）广泛应用于缓解大型语言模型（LLMs）和多模态大型语言模型（MLLMs）中的幻觉问题。特别是，基于知识图谱（KG）的RAG利用结构化知识为（M）LLMs提供高质量的外部信息。在这些研究的基础上，近期的研究探索了多模态知识图谱（MMKGs）作为GraphRAG的知识库。这使得Graph RAG能够整合来自多种模态的知识，从而进一步提升其性能。然而，现有的基于MMKG的RAG方法通常遵循一个共同的流程，其中不同模态在融合之前大多是独立处理的。因此，在视觉信息提取和后续的多模态知识融合过程中，文本上下文的使用受到限制。这在图像和文本之间造成了语义差距，限制了多模态GraphRAG的性能。为了解决这一问题，我们提出了一种新颖的框架，用于构建上下文增强的多模态知识图谱（CEMMKG），以更好地支持多模态GraphRAG。所提出的CEMMKG在局部和全局范围内为每个图像丰富了互补的文本上下文。局部上下文通过结合与图像语义相关的句子超越了周围文本，而全局上下文则提供了整个段落的摘要。我们进一步引入了局部上下文的多粒度设计，使其能够在不同的细节层次上捕捉语义相关的信息。在选定的视觉中心数据集上进行的广泛实验验证了CEMMKG在利用上下文信息改善基于MMKG的RAG性能方面的有效性。此外，其在不同基于MMKG的RAG方法中的有效性展示了其广泛的适用性。
View on arXiv Download PDF AI Translation
cs.AI / 50 / 2608.25992

ProgRouter: Online Progress-Guided Orchestration for Multi-Agent LLM Workflows under Quality-Cost Tradeoffs

ProgRouter：在质量-成本权衡下的多智能体大语言模型工作流的在线进度引导调度
Li, Somgyuan, Abdelmoniem, Ahmed M., Wang, Shiqiang
Abstract
Multi-agent large language model (LLM) workflows have emerged as a powerful paradigm for solving complex, open-ended tasks through collaborative reasoning among specialized LLM agents, but they incur substantial operating costs due to repeated LLM invocations and long-horizon context accumulation. Existing cascade routing methods make one-shot, query-level decisions and cannot adapt to the dynamic, state-dependent nature of multi-step workflows, in which the right LLM at each step depends on evolving task progress, remaining task difficulty, and cost-efficiency requirements. We present ProgRouter, an online progress-guided routing framework that adaptively selects LLM agents across workflow steps to preserve task-solving quality while adhering to time and cost budgets. ProgRouter introduces a multi-view task progress scorer that combines coarse workflow outcome regimes with fine-grained signals on subtask completion, progress trends, and workflow state quality. Then, a dual-path task progress predictor and an adaptive meta-gating mechanism estimate the progress gain for each candidate routed LLM. ProgRouter makes online step-wise routing decisions that balance progress gain, task time budgets, and long-term operating cost efficiency. Experiments on HumanEval Plus, MBPP, MATH-500, and ASQA, spanning agentic code generation, mathematical reasoning, and retrieval-augmented long-form question answering, demonstrate that ProgRouter reduces the operating cost relative to key baselines while maintaining strong task-solving performance.
Chinese Translation
多智能体大语言模型（LLM）工作流已成为通过专门的LLM智能体之间的协作推理解决复杂开放任务的强大范式，但由于重复调用LLM和长时间上下文积累，它们会产生可观的操作成本。现有的级联路由方法做出一次性、查询级别的决策，无法适应多步骤工作流的动态状态依赖特性，其中每一步所需的LLM取决于不断变化的任务进展、剩余任务难度和成本效率要求。我们提出了ProgRouter，这是一种在线进度引导的路由框架，能够在工作流步骤中自适应选择LLM智能体，以保持任务解决质量，同时遵循时间和成本预算。ProgRouter引入了一种多视角任务进度评分器，结合了粗略的工作流结果模式与关于子任务完成、进度趋势和工作流状态质量的细粒度信号。然后，双路径任务进度预测器和自适应元门控机制估计每个候选路由LLM的进度增益。ProgRouter进行在线逐步路由决策，平衡进度增益、任务时间预算和长期操作成本效率。在HumanEval Plus、MBPP、MATH-500和ASQA等实验中，涵盖了智能代码生成、数学推理和检索增强的长篇问答，结果表明ProgRouter在保持强大任务解决性能的同时，相较于关键基线降低了操作成本。
View on arXiv Download PDF AI Translation
cs.AI / 51 / 2608.26004

AsymSpec: Context-Asymmetric Speculative Decoding for Agentic LLMs

AsymSpec：面向智能大型语言模型的上下文非对称推测解码
Liang, Sheng, Zhang, Yongyue, Brian, Nathanael, Lv, Hang, Wang, Hao, Zhang, Chen, Liu, Yong
Abstract
Agentic LLM pipelines face escalating inference costs as context accumulates across retrieval, tool use, and multi-turn interactions. To control latency, deployments routinely compress inputs, but this degrades task accuracy. Speculative decoding (SD) accelerates generation losslessly, yet it assumes the drafter and verifier share an identical context, preventing SD from resolving the accuracy-overhead trade-off. We propose AsymSpec, an asymmetric speculative decoding framework that breaks this symmetry: a lightweight drafter reads the full input while the large verifier operates on the compressed view. The drafter steers the verifier via a contrastive $\delta$-fusion of logits, modulated by a divergence-aware acceptance gate that preserves verification stability and high draft acceptance rates. Evaluated across four agentic capabilities and two end-to-end agent benchmarks, AsymSpec reaches $\approx 90\%$ of full-context accuracy on average, delivering $1.3$--$1.7\times$ throughput speedups at $0.2$--$0.3\times$ the compute cost on isolated text capabilities. These results show that asymmetric context access yields substantial gains precisely when compression discards critical reasoning signals.
Chinese Translation
智能大型语言模型的推理管道在检索、工具使用和多轮交互中随着上下文的累积面临不断上升的推理成本。为了控制延迟，部署通常会压缩输入，但这会降低任务的准确性。推测解码（Speculative Decoding, SD）以无损方式加速生成，然而它假设起草者和验证者共享相同的上下文，这使得SD无法解决准确性与开销之间的权衡。我们提出了AsymSpec，一种打破这种对称性的非对称推测解码框架：轻量级的起草者读取完整输入，而大型验证者则在压缩视图上操作。起草者通过对数的对比$oldsymbol{ ext{δ}}$-融合引导验证者，该过程由一个关注发散的接受门调节，以保持验证的稳定性和高起草接受率。在四种智能能力和两个端到端智能代理基准测试中评估，AsymSpec在平均情况下达到了约90%的全上下文准确性，在隔离文本能力上实现了1.3到1.7倍的吞吐量提升，计算成本为0.2到0.3倍。这些结果表明，非对称上下文访问在压缩丢弃关键推理信号时带来了显著的收益。
View on arXiv Download PDF AI Translation
cs.AI / 52 / 2608.26009

Imitation Learning for Connection-Tableau Construction

用于连接表构建的模仿学习
Rømming, Fredrik, Bakšys, Mantas, Fixman, Martin S., Holden, Sean B.
Abstract
An automated theorem prover builds a proof step by step, choosing at each point what to add and what to remove. We cast this construction as a policy acting in a transition system induced by a formal calculus, which fixes which steps are sound: for clausal connection tableaux, leanCoP-style search and plCoP/rlCoP-style planning then become stateful policies over one interface, and policy-learning methods apply directly. We equip such policies with a graph neural network that scores proof edits from structure that transfers across problems, train it by imitation learning from found proofs, and measure how performance holds as we remove search scaffolding, from full symbolic backtracking to a policy the network drives alone. Within a fixed step budget on M2k, MPTP2078-bushy, and TPTP v9.2.1, learned policies solve up to 46% more problems than leanCoP, and reach proofs in an order of magnitude fewer steps.
Chinese Translation
自动定理证明器逐步构建证明，在每个步骤中选择添加或移除的内容。我们将这一构建过程视为在由形式演算引发的状态转移系统中进行的策略，其中固定了哪些步骤是有效的：对于子句连接表，leanCoP 风格的搜索和 plCoP/rlCoP 风格的规划则成为一个接口上的有状态策略，策略学习方法可以直接应用。我们为这些策略配备了一个图神经网络，该网络根据跨问题转移的结构对证明编辑进行评分，通过从已找到的证明中进行模仿学习来训练它，并测量在我们去除搜索支架时性能的保持情况，从完全符号回溯到网络单独驱动的策略。在 M2k、MPTP2078-bushy 和 TPTP v9.2.1 上，在固定的步骤预算内，学习到的策略解决的问题比 leanCoP 多出最多 46%，并且在数量级上减少了证明所需的步骤。
View on arXiv Download PDF AI Translation
cs.AI / 53 / 2608.26036

Trace Integrity for LLM Data Agents: A Vision for Auditable Structured Reasoning in Real-World Systems

大规模语言模型数据代理的追踪完整性：现实世界系统中可审计结构化推理的愿景
Dutta, Srimonti, Moharir, Akshata Kishore
Abstract
Answer accuracy is an insufficient reliability signal for LLM data agents. In structured-data tasks, a benchmark-correct answer can be produced by an invalid trace. This paper introduces Trace Integrity, a deployment reliability criterion for evaluating whether the computation recorded behind an answer is explicit, executable, schema-valid, operator-faithful, replayable, answer-consistent, and auditable. We identify the Structure Gap as the deployment failure mode that makes Trace Integrity necessary: natural-language reasoning and free-form rationales do not reliably specify the operator-level programs required by real-world systems. We operationalize Trace Integrity with execution contracts, structured artifacts that bind user intent to schema elements, operator plans, assumptions, executable queries, verification status, and final-answer linkage. We also introduce CAIT (Correct Answer / Invalid Trace) Rate, which measures how often answer-only evaluation counts computationally unsupported outputs as successes. In an empirical demonstration on BIRD Mini-Dev, Direct SQL, Operation Summary + SQL, and Contract-First SQL achieve answer accuracies of 20%, 22%, and 24%, while their Trace Integrity Pass Rates are 39%, 43%, and 40% and their CAIT Rates remain high at 55%, 59.1%, and 45.8%, showing that answer accuracy, trace validity, and silent-failure risk are distinct evaluation signals. Real-world LLM data agents should, therefore, be evaluated not only by whether their outputs match a reference answer, but by whether those outputs are backed by auditable computation.
Chinese Translation
答案的准确性不足以作为大规模语言模型（LLM）数据代理的可靠性信号。在结构化数据任务中，基准正确答案可能由无效的追踪产生。本文提出了追踪完整性（Trace Integrity），作为评估答案背后记录的计算是否明确、可执行、符合模式、忠实于操作、可重放、与答案一致且可审计的部署可靠性标准。我们将结构差距（Structure Gap）识别为使追踪完整性成为必要的部署失败模式：自然语言推理和自由形式的理由并不能可靠地指定现实世界系统所需的操作级程序。我们通过执行合同（execution contracts）来操作化追踪完整性，这些结构化文档将用户意图与模式元素、操作计划、假设、可执行查询、验证状态和最终答案链接绑定在一起。我们还引入了CAIT（Correct Answer / Invalid Trace）率，用于衡量仅依赖答案评估计算上不支持的输出被视为成功的频率。在对BIRD Mini-Dev的实证演示中，直接SQL、操作摘要 + SQL和合同优先SQL的答案准确率分别为20%、22%和24%，而它们的追踪完整性通过率分别为39%、43%和40%，CAIT率则保持在55%、59.1%和45.8%的高水平，显示答案准确性、追踪有效性和静默失败风险是不同的评估信号。因此，现实世界的大规模语言模型数据代理的评估不仅应考虑其输出是否与参考答案匹配，还应考虑这些输出是否得到可审计计算的支持。
View on arXiv Download PDF AI Translation
cs.AI / 54 / 2608.26081

SwarmWorld: Stigmergic technological evolution in societies of language-model agents

SwarmWorld：语言模型代理社会中的刺激性技术演化
Pal, Subhadeep, Wang, Fiona Y., Buehler, Markus J.
Abstract
Collective intelligence can emerge when individuals coordinate through a shared environment, allowing local actions to accumulate into durable social organization. Language-model agents offer a new substrate for this process, yet most multi-agent systems rely on direct conversation, predefined roles, or centralized workflows. It remains unclear whether decentralized agents can build functional technologies and outperform independent search. Here, initially homogeneous LLM agents in SwarmWorld self-organize without assigned roles or recipes into evolving technological societies. Agents explore a spatial environment, process resources, test materials, construct persistent artifacts, and write executable controllers evaluated by a deterministic simulator under unseen disturbances after the agents are removed. SwarmWorld splits cognition from consequence: agents propose architectures and controllers within fixed action and material schemas, while the simulated world determines function. Shared societies develop broader, more resilient technological portfolios than a strong best-of-N isolated-search baseline, although isolated search remains competitive for the strongest artifact. Agents differentiate into exploration, construction, maintenance, and coordination behaviors, transitioning as the world matures. Technologies accumulate through collaborative construction, executable inheritance, and persistent agent-artifact networks, with most reuse beginning through physical observation rather than communication. Explicit cultural mechanisms amplify collaboration and organization, but functional benefits depend on outcome and timescale. Physical stigmergy alone supports capable societies, while interaction drives persistent technological ecologies rather than universally superior individual inventions.
Chinese Translation
当个体通过共享环境进行协调时，集体智能可以出现，从而使局部行动积累成持久的社会组织。语言模型代理为这一过程提供了一种新的基础，但大多数多代理系统依赖于直接对话、预定义角色或集中工作流程。目前尚不清楚去中心化的代理是否能够构建功能性技术并超越独立搜索。在这里，最初同质的 LLM 代理在 SwarmWorld 中自组织，无需分配角色或配方，形成不断演变的技术社会。代理探索空间环境，处理资源，测试材料，构建持久的人工制品，并编写可执行的控制器，这些控制器在代理被移除后由确定性模拟器在未见干扰下进行评估。SwarmWorld 将认知与后果分离：代理在固定的行动和材料框架内提出架构和控制器，而模拟世界则决定功能。共享社会发展出比强大的最佳独立搜索基线更广泛、更具韧性的技术组合，尽管独立搜索在最强人工制品方面仍具竞争力。代理在探索、构建、维护和协调行为上进行区分，并随着世界的成熟而转变。技术通过协作构建、可执行继承和持久的代理-人工制品网络积累，其中大多数重用始于物理观察而非交流。明确的文化机制增强了协作和组织，但功能性收益依赖于结果和时间尺度。仅凭物理刺激就能支持有能力的社会，而互动则推动持久的技术生态，而不是普遍优越的个体发明。
View on arXiv Download PDF AI Translation
cs.AI / 55 / 2608.26088

Planetary Prediction Engine: Autonomous Geospatial Prediction via Intelligent Data Selection and Foundation Model Embeddings

行星预测引擎：通过智能数据选择和基础模型嵌入实现自主地理空间预测
Ma, Evelyn, Pasumarthi, Rama Kumar, Shafin, Kishwar, Sharma, Mandar, Sun, Mimi, Sadeghi, Hamed, Ebengo, Dav M., Onesime, Mbulayi, Solomakhin, Rouslan, Wamburu, John, Ogallo, William, Walcott-Bryant, Aisha, Chen, Sanxing, Muslim, Arbaaz, Mayer, Yael, Ho, Ronald, Lee, Roy, Alcantara, Ruth, Diack, Abdoulaye, Bharel, Monica, Rosique, Lambert, Amez-Droz, Jeremy, Haire, Christopher, Manyika, James, Matias, Yossi, Efron, Niv, Prasad, Gautam, Shetty, Shravya
Abstract
Addressing critical global challenges, from food security and disaster risk to disease outbreaks and socio-economic vulnerability, demands high-fidelity geospatial modeling. However, building predictive planetary models remains bottlenecked by a fragmented data ecosystem, requiring manual data retrieval, multimodal data curation and fusion along with iterative model selection. We present the Planetary Prediction Engine (PPE), an autonomous AI system that executes this end-to-end workflow directly from natural-language queries. PPE synthesizes multimodal datasets on the fly, retrieving spatiotemporally relevant covariates across open-web and Earth observation platforms (Data Commons, Google Earth Engine) and fusing them with geospatial foundation model embeddings (PDFM, AlphaEarth). Simultaneously, it searches over task-tailored model architecture families with automated overfitting guards. Across diverse tasks, geographies, and scientific domains, PPE consistently outperforms state-of-the-art or manually tuned expert baselines. For US spatial regression, PPE improves mean $R^2$ across 21 CDC health indicators (76.8% vs. 60.0%), FEMA national risk indices (64.9% vs. 60.0%), and the Social Vulnerability Index (66.2% vs. 58.6%). For spatial downscaling in data-scarce settings, PPE integrates localized proxies to double baseline accuracy in Nigerian food security indicators ($R^2$ of 66.1% vs. 31.5%). For epidemiological nowcasting of the 2026 DRC Bundibugyo Ebola outbreak, PPE achieves a Recall@10 of 83.3% (identifying 15 of 18 newly invaded health zones across five weekly forecasts), a +10.3 percentage-point improvement over the public state-of-the-art modeling (~73%). By combining autonomous multimodal planetary data discovery with targeted model optimization, PPE lowers the technical barrier to planetary-scale analytics, enabling rapid, customized, expert-level deployment.
Chinese Translation
应对全球关键挑战，包括粮食安全、灾害风险、疾病暴发和社会经济脆弱性，要求高保真度的地理空间建模。然而，构建预测性行星模型仍然受到碎片化数据生态系统的制约，需手动数据检索、多模态数据策划和融合，以及迭代模型选择。我们提出了行星预测引擎（Planetary Prediction Engine, PPE），这是一个自主的人工智能系统，能够直接从自然语言查询中执行这一端到端工作流程。PPE即时合成多模态数据集，从开放网络和地球观测平台（数据公共库、谷歌地球引擎）中检索时空相关的协变量，并将其与地理空间基础模型嵌入（PDFM、AlphaEarth）融合。同时，它在任务定制的模型架构家族中进行搜索，并具备自动过拟合防护。在多样化的任务、地理区域和科学领域中，PPE始终优于最先进的或手动调优的专家基线。在美国空间回归中，PPE在21个CDC健康指标上提高了平均 $R^2$（76.8% 对比 60.0%），在FEMA国家风险指数上（64.9% 对比 60.0%），以及在社会脆弱性指数上（66.2% 对比 58.6%）。在数据稀缺环境下的空间下采样中，PPE整合了本地化代理，使尼日利亚粮食安全指标的基线准确度翻倍（$R^2$ 为 66.1% 对比 31.5%）。在2026年刚果民主共和国Bundibugyo埃博拉疫情的流行病学即时预测中，PPE实现了83.3%的Recall@10（在五个每周预测中识别出18个新入侵健康区中的15个），比公共最先进模型提高了10.3个百分点（约73%）。通过将自主多模态行星数据发现与针对性模型优化相结合，PPE降低了行星级分析的技术门槛，实现了快速、定制化的专家级部署。
View on arXiv Download PDF AI Translation
计算语言学 (Computation and Language)
69
cs.CL / 1 / 2608.24901

Detection != Reliable Control: Decodable Empathy Directions Yield at Most Partial Shifts in Automated Empathy Scores

检测不等于可靠控制：可解码的同理心方向在自动化同理心评分中最多只产生部分变化
Jisun, Haoran
Abstract
A decodable "empathy" direction is routinely read as a causal lever, conflating decodability, automated-metric control, and human-perceived change. We test this for two EPITOME-derived facets -- Recognition (cognitive) and Resonance (affective) -- in three instruction-tuned LLMs, scoring every intervention with two LLM judges and a discriminative EPITOME classifier, each gated by an emotional-vs-neutral positive control. The control passes for the affective facet across all automated instruments, but cognitive range is inconsistent across them. Both facets remain decodable after residualizing against a sentence-embedding-derived surface score, and steering can substantially rewrite the text. Yet adding the Resonance direction raises the affective score only partially -- in Qwen by +0.29 (approximately 26% of the natural gap). A direct between-direction contrast confirms the shift is facet-specific in Qwen and Llama (not Gemma); we do not, however, establish a matching human-perceived change. Additive cognitive steering produces no measurable change, but a within-domain control shows the cognitive instrument is too coarse to resolve the differences such steering would produce -- unmeasurable, not a clean null. By contrast, Gemma Recognition ablation lowers the classifier's cognitive score even after adjusting for response length. Detection does not imply reliable control under global interventions, and cognitive-empathy claims warrant an explicit measurement-sensitivity check.
Chinese Translation
可解码的“同理心”方向通常被视为因果杠杆，将可解码性、自动化指标控制和人类感知变化混为一谈。我们针对两个源自EPITOME的方面——识别（认知）和共鸣（情感）——在三种经过指令调优的LLM中进行测试，使用两个LLM评审和一个区分性的EPITOME分类器对每个干预进行评分，每个评分都通过情感与中性正控制进行筛选。情感方面在所有自动化工具中均通过控制，但认知范围在这些工具中表现不一致。经过基于句子嵌入的表面评分的残差处理后，这两个方面仍然可解码，并且引导可以显著重写文本。然而，添加共鸣方向仅部分提高了情感评分——在Qwen中提高了0.29（约占自然差距的26%）。直接的方向间对比确认了在Qwen和Llama中变化是方面特定的（而Gemma则不是）；然而，我们并未建立与人类感知变化相匹配的结果。附加的认知引导没有产生可测量的变化，但领域内控制显示认知工具过于粗糙，无法解决这种引导所产生的差异——不可测量，而不是干净的零结果。相比之下，Gemma的识别消融在调整响应长度后仍降低了分类器的认知评分。在全球干预下，检测并不意味着可靠控制，认知同理心的主张需要明确的测量敏感性检查。
View on arXiv Download PDF AI Translation
cs.CL / 2 / 2608.24920

Semantic Variability of Replies Across LLMs: Implications for Designing Conversation-Based Assessment

大型语言模型回复的语义变异性：对基于对话评估设计的启示
Hao, Jiangang
Abstract
This study examines whether LLM-generated replies remain semantically consistent when the underlying LLM changes. Using messages from real collaborative conversations, we compared the semantic similarity of generated replies across LLMs under two conditions: with and without preceding chat history. Results show that model choice and conversational context both affect response similarity and alignment with human replies. These findings indicate that prompting and conversational context alone may not be sufficient to preserve response consistency across LLMs, highlighting the need for infrastructure and design strategies that can maintain stable and comparable responses amid the rapid and continuous evolution of LLMs.
Chinese Translation
本研究考察了当基础的大型语言模型（LLM）发生变化时，LLM生成的回复是否保持语义一致性。我们使用真实协作对话中的消息，在有和没有前置聊天记录的两种条件下比较了不同LLM生成回复的语义相似性。结果表明，模型选择和对话上下文均会影响回复的相似性及其与人类回复的一致性。这些发现表明，仅依靠提示和对话上下文可能不足以保持不同LLM之间回复的一致性，强调了在LLM快速持续演变的背景下，建立能够维持稳定且可比回复的基础设施和设计策略的必要性。
View on arXiv Download PDF AI Translation
cs.CL / 3 / 2608.24952

The Dialect Tax: Dialectal Biases Persist throughout the Language Modeling Pipeline

方言税：方言偏见在语言建模流程中持续存在
Elle
Abstract
Systematic dialectal performance gaps in language models (LMs) are well documented, but the source of these disparities within the modern language modeling pipeline remains unclear. Our study traces this "dialect tax" across the natural language processing pipeline. Using parallel English dialect corpora that hold meaning fixed while varying surface form, we first confirm that LMs recognize matched Standard American English (SAE) and dialectal texts as semantically equivalent. However, we discover further representational gaps corresponding to downstream performance gaps. Across model families and generations, modern LMs still encode dialectal texts unequally during tokenization, pre-training, post-training, and inference. Strikingly, bypassing traditional subword segmentation via a character-level counterfactual tokenizer removes neither input and output asymmetries nor dialectal accuracy gaps. During pre-training, dialect pairs induce more divergent gradient updates than pairs of entirely unrelated SAE documents, indicating that models find semantically equivalent dialectal content harder to learn from than unrelated SAE documents. During post-training, reward models show contextual, unstable dialect preferences, assigning higher values to isolated AAVE-exclusive tokens than to SAE-exclusive tokens, while full reasoning contexts receive task- and model-dependent dialect penalties. Overall, our findings suggest that the dialect tax is encoded and accumulated not by any one step in isolation, but at every step of the language modeling process.
Chinese Translation
语言模型（LMs）中的系统性方言表现差距已被充分记录，但这些差异在现代语言建模流程中的来源仍不清楚。我们的研究追踪了这一“方言税”在自然语言处理流程中的表现。通过使用在表面形式上变化但意义保持不变的平行英语方言语料库，我们首先确认语言模型将匹配的标准美式英语（SAE）和方言文本视为语义等价。然而，我们发现进一步的表征差距与下游表现差距相对应。在不同的模型家族和代际中，现代语言模型在标记化、预训练、后训练和推理过程中仍然对方言文本进行不平等的编码。值得注意的是，通过字符级反事实标记器绕过传统的子词分割并未消除输入和输出的不对称性，也未消除方言准确性差距。在预训练期间，方言对比引发的梯度更新比完全无关的SAE文档对更为分歧，表明模型发现语义等价的方言内容比无关的SAE文档更难学习。在后训练期间，奖励模型显示出上下文不稳定的方言偏好，给予孤立的仅限AAVE（非洲裔美国人英语）标记比SAE独占标记更高的值，而完整的推理上下文则受到任务和模型依赖的方言惩罚。总体而言，我们的研究结果表明，方言税并不是由某一步骤单独编码和累积的，而是在语言建模过程的每一步中都存在。
View on arXiv Download PDF AI Translation
cs.CL / 4 / 2608.24982

Unsupervised Post-Training of Foundation Models: A Survey

基础模型的无监督后训练：综述
Xu, Yijie, Cai, Qianyi, Yao, Huizai, Wang, Yili, Wang, Tianfu, Yang, Cehao, Yao, Xingbo, Guo, Zhiyu, Liu, Aiwei, Hu, Xuming, Guo, Weiyu, Xiong, Hui
Abstract
Foundation-model post-training usually relies on human labels, preference data, stronger teachers, or executable verifiers. We study Unsupervised Post-Training (UPT): update-bearing adaptation on unlabeled inputs whose learning signal is derived from same-lineage model artifacts rather than an external oracle. We catalog 80 strict UPT methods and organize them by the object that supplies the update signal: a prediction statistic, a sample relation, a self-generated target, or an internal evaluator. Beyond inventory, we show how the choice of internal signal and task structure determines whether post-training improves the model or recursively amplifies error. An orthogonal Input Visibility $\times$ Update Persistence view maps deployment regimes and defines a unified framework for UPT selection and evaluation.
Chinese Translation
基础模型的后训练通常依赖于人工标签、偏好数据、更强的教师模型或可执行的验证器。我们研究了无监督后训练（Unsupervised Post-Training, UPT）：在无标签输入上进行的更新适应，其学习信号来源于同源模型的产物，而非外部oracle。我们对80种严格的UPT方法进行了分类，并根据提供更新信号的对象进行组织：预测统计量、样本关系、自生成目标或内部评估器。除了清单，我们还展示了内部信号和任务结构的选择如何决定后训练是改善模型还是递归放大错误。一个正交的输入可见性 $ imes$ 更新持久性视角映射了部署模式，并定义了一个统一的UPT选择和评估框架。
View on arXiv Download PDF AI Translation
cs.CL / 5 / 2608.24988

Does Fine-Tuning Undo Activation Steering? Behavioural Recovery Without Weight-Edit Reversal

微调是否会撤销激活引导？无权重编辑反转的行为恢复
Glass, Philipp E., Tucker, Allan, Li, Yongmin, Miron, Alina
Abstract
Activation steering can be embedded directly into a language model's weights, shaping behaviour without inference-time intervention and offering a way to encode alignment prior to release. However, models are routinely fine-tuned after deployment, and it is unknown whether embedded interventions survive this. We study the stability of embedded steering for refusal suppression and brevity induction across five instruction-tuned models (3B-14B) under non-adversarial SFT and RLHF. Behaviourally, preservation tracks the training data: steering degrades when optimisation pressure contradicts the targeted behaviour and persists otherwise, with refusal ablation losing 64% of its effect on average under SFT. Mechanistically, however, the weight edit survives almost untouched even where behaviour reverts: mean vector recovery is $\rho = 0.004$, and the fine-tuning update along the steering direction is near-orthogonal to its pre-edit weight pattern (mean $\cos\theta = 0.074$). When steered behaviour degrades, fine-tuning does not achieve it by dismantling or reversing the steering mechanism itself. Embedded steering is therefore mechanistically durable but functionally vulnerable, and requires behavioural re-validation after downstream training.
Chinese Translation
激活引导可以直接嵌入到语言模型的权重中，在不进行推理时干预的情况下塑造行为，并提供了一种在发布前编码对齐的方法。然而，模型在部署后通常会进行微调，而嵌入的干预是否能在此过程中存活尚不清楚。我们研究了在非对抗性的监督微调（SFT）和强化学习微调（RLHF）下，五个指令调优模型（3B-14B）中嵌入引导对于拒绝抑制和简洁性诱导的稳定性。从行为上看，保留情况与训练数据相关：当优化压力与目标行为相矛盾时，引导效果下降，否则则持续存在，在SFT下拒绝消融平均损失64%的效果。然而，从机制上看，即使行为恢复，权重编辑几乎保持不变：均值向量恢复为$ ho = 0.004$，沿引导方向的微调更新与其编辑前的权重模式近乎正交（均值$ ext{cos} heta = 0.074$）。当引导行为下降时，微调并不是通过拆解或反转引导机制本身来实现的。因此，嵌入引导在机制上是持久的，但在功能上是脆弱的，并且在下游训练后需要进行行为重新验证。
View on arXiv Download PDF AI Translation
cs.CL / 6 / 2608.25005

The Imperfective Paradox Is Not Necessarily in Large Language Models: A Benchmark Failure Before a Model Failure

未必存在于大型语言模型中的未完成体悖论：模型失败之前的基准测试失败
Han, Kaiqiao, Sun, Yizhou
Abstract
The imperfective paradox provides a useful test of compositional semantic analysis. Recent work constructs an NLI benchmark and reports that models frequently infer completed telic events from progressive descriptions, attributing this behavior to a Teleological Bias. It further argues that prompting interventions cause a Calibration Crisis. We reexamine the benchmark and conclusions and show that it is substantially affected by conceptual and evaluation mis-specifications. We identify three conceptual mis-specifications. In particular, Aspectual Reduction affects the benchmark construction, analysis, experiments, and conclusions. Under a strict NLI standard, 76% of Group A instances do not explicitly rule out culmination. In our native-speaker annotation, 38% of Group A examples and 29% of the Group C examples were judged to permit an alternative interpretation. To control these issues and lexical variation, we construct Lexically Matched Minimal Pairs. At the evaluation level, we formulate event-semantic NLI as a Multi-step Reasoning Problem and assess both intermediate semantic decisions and final predictions. Our results show that models often do not affirm culmination but nevertheless accept the corresponding simple-past hypothesis, a pattern we characterize as Sufficiency Bias. We further show that prompting interventions produce a Decision Shift among labels without reliably improving the underlying semantic understanding and reasoning. Intermediate and oracle-guided analyses identify two additional failure modes: errors in compositional aspectual classification and Surface-form Attraction toward surface-associated answers. Our experiments on Qwen-7B with suitable prompts, GPT-5.4, and Qwen-72B provide initial evidence for the context sensitivity of aspectual classification and suggest that these models can achieve performance comparable to that of human annotators.
Chinese Translation
未完成体悖论为组合语义分析提供了一个有用的测试。近期的研究构建了一个自然语言推理（NLI）基准，并报告模型经常从进行时描述中推断出已完成的目的性事件，将这种行为归因于目的性偏见（Teleological Bias）。研究进一步认为，提示干预导致了校准危机（Calibration Crisis）。我们重新审视该基准及其结论，显示其受到概念和评估误规范的显著影响。我们识别出三种概念误规范。特别是，体的简化（Aspectual Reduction）影响了基准的构建、分析、实验和结论。在严格的NLI标准下，76%的A组实例并未明确排除结果的发生。在我们的母语者注释中，38%的A组示例和29%的C组示例被判断为允许替代解释。为了控制这些问题和词汇变异，我们构建了词汇匹配的最小对（Lexically Matched Minimal Pairs）。在评估层面，我们将事件语义NLI表述为多步骤推理问题，并评估中间语义决策和最终预测。我们的结果显示，模型通常不确认结果的发生，但仍然接受相应的简单过去时假设，这一模式我们称之为充分性偏见（Sufficiency Bias）。我们进一步表明，提示干预在标签之间产生了决策转变，但并未可靠地改善基础的语义理解和推理。中间和oracle引导分析识别出两种额外的失败模式：组合体分类中的错误和表面形式吸引（Surface-form Attraction）向表面相关答案的倾斜。我们在适当提示下对Qwen-7B、GPT-5.4和Qwen-72B的实验提供了初步证据，表明体分类具有上下文敏感性，并建议这些模型可以达到与人类注释者相当的性能。
View on arXiv Download PDF AI Translation
cs.CL / 7 / 2608.25022

A Primer on Computational Semantics for Artificial Intelligence Systems

人工智能系统的计算语义入门
Kennington, Casey
Abstract
As people adopt transformer-based language models (e.g., ChatGPT and Gemini) for an increasing number of use-cases, it is important to know how such models learn and represent the meaning of the language, and to be more informed about what language is. This document is an attempt to help the reader understand how linguistic meaning (i.e., semantics) is approached from different fields of scientific and philosophical examination. I also explain three primary semantic theories: formal semantics, grounded semantics, and distributional semantics then compare how transformer-based language models differ from how humans learn language.
Chinese Translation
随着人们在越来越多的应用场景中采用基于变换器的语言模型（如 ChatGPT 和 Gemini），了解这些模型如何学习和表示语言的意义，以及对语言本质的更深入理解变得尤为重要。本文旨在帮助读者理解语言意义（即语义）在不同科学和哲学领域中的研究方法。我还解释了三种主要的语义理论：形式语义学、基础语义学和分布式语义学，并比较了基于变换器的语言模型与人类学习语言的不同之处。
View on arXiv Download PDF AI Translation
cs.CL / 8 / 2608.25028

Behind the [MASK]: Disentangling Representation and Faithfulness in DAPF-Based Dementia Detection

揭示[MASK]背后的真相：解构基于DAPF的痴呆检测中的表示与忠实性
Ranjbar-Noiey, Pardis, Parde, Natalie
Abstract
Spoken-language analysis via prompt-based domain-adaptive models is a promising direction for low-resource, non-invasive dementia screening, but such models remain internally opaque. We study the interpretability of the Domain-Adapted models via Prompt-based Fine-tuning (DAPF) framework, which casts dementia detection as diagnosis-related masked-token prediction. We interpret DAPF and strong baselines using a variety of probing and analysis techniques, finding that DAPF achieved the best overall performance (accuracy=0.83 and macro-F1=0.83) with diagnosis most recoverable from its [MASK] representation. However, this representational advantage did not extend to token-level explanation faithfulness. DAPF attributions primarily reflected language task vocabulary, discourse markers, and transcription artifacts, with perturbation tests showing weak or negative effects. This suggests that its masked-token interface determines diagnosis information without producing faithful token-level explanations.
Chinese Translation
通过基于提示的领域自适应模型进行口语语言分析是低资源、非侵入性痴呆筛查的一个有前景的方向，但此类模型内部仍然不透明。我们研究了通过基于提示的微调（DAPF）框架的领域自适应模型的可解释性，该框架将痴呆检测视为与诊断相关的掩码标记预测。我们使用多种探测和分析技术对DAPF及其强基线进行了解释，发现DAPF在整体性能上表现最佳（准确率=0.83，宏观F1=0.83），其诊断信息最能从[MASK]表示中恢复。然而，这种表示优势并未延伸到标记级解释的忠实性。DAPF的归因主要反映了语言任务词汇、话语标记和转录伪影，扰动测试显示出微弱或负面的效果。这表明其掩码标记接口在确定诊断信息的同时，并未产生忠实的标记级解释。
View on arXiv Download PDF AI Translation
cs.CL / 9 / 2608.25038

Padamitra: Grounded Glossary Generation for Classical Sanskrit

Padamitra：经典梵语的基础词汇生成
Jagadeeshan, Manoj Balaji, Marala, Sai Pragnaan, Goyal, Pawan
Abstract
We introduce grounded glossary generation, a structured task requiring models to recover semantically meaningful Sanskrit phrases and produce translation-grounded meanings from a sloka-translation pair, formalizing the traditional patha commentary practice as an evaluable NLP objective. We construct a benchmark of 31,316 sloka-translation-glossary triples from the Valmiki Ramayana and Srimad Bhagavatam, paired with two metrics: Jaccard for phrase recovery and Meaning Faithfulness for semantic consistency. Across zero-shot, few-shot, and instruction fine-tuned variants of Gemma-3n-E4B, Gemma-3-12B, Phi-4, and Qwen3.5-9B, instruction fine-tuning substantially outperforms prompting, while explicit segmentation yields gains. Error analysis identifies over-segmentation of sandhi and samasa compounds as the dominant failure mode, pointing to morphological modeling as the key bottleneck for faithful Sanskrit lexical decomposition.
Chinese Translation
我们提出了基础词汇生成这一结构化任务，要求模型从一对梵文诗句和其翻译中恢复出具有语义意义的梵语短语，并生成与翻译相关的意义，从而将传统的路径评论实践形式化为可评估的自然语言处理（NLP）目标。我们从《瓦尔米基史诗》和《斯里玛德·巴哈伽瓦塔》中构建了一个包含31,316对诗句-翻译-词汇三元组的基准，并配备了两个评估指标：用于短语恢复的杰卡德（Jaccard）指标和用于语义一致性的意义忠实度（Meaning Faithfulness）指标。在Gemma-3n-E4B、Gemma-3-12B、Phi-4和Qwen3.5-9B的零样本、少样本和指令微调变体中，指令微调显著优于提示，而显式分割则带来了提升。错误分析表明，沙提（sandhi）和合成词（samasa）复合词的过度分割是主要的失败模式，指出形态建模是忠实梵语词汇分解的关键瓶颈。
View on arXiv Download PDF AI Translation
cs.CL / 10 / 2608.25061

DataKernelBench: Can LLMs Optimize Database Queries on GPUs?

DataKernelBench：大型语言模型能在GPU上优化数据库查询吗？
Kumar, Gokul Karthik, Perlitz, Yotam, Lammie, Corey, Giovannini, Andrea, Hose, Katja
Abstract
GPUs increasingly accelerate database systems, but query-specific peak performance still often relies on hand-written kernels. Existing LLM kernel benchmarks focus on machine learning operators, leaving irregular, heterogeneous, data-movement-heavy database-style operators untested. We introduce DataKernelBench, which translates SQL into validated PyTorch TorchPlan programs and evaluates LLMs that optimize either the core tensor-bounded snippet or the full query in CUDA or Triton through execution-guided repair. Across ten proprietary and open-weight models on TPC-H SF10 with an H100 GPU, the strongest full-query CUDA configuration achieves $2.11\times$ speedup over torch.compile at full pass rate. We find that higher-performing implementations commonly use kernel fusion and execution-strategy changes, stronger models benefit most from full-query specialization, and workload context matters more than hardware context. To handle data larger than GPU memory, we extend TorchPlan with Dask-cuDF for on-demand partition loading on TPC-H SF100 with four H100 GPUs, achieving $2.54\times$ speedup
Chinese Translation
GPU越来越多地加速数据库系统，但查询特定的峰值性能仍然往往依赖于手写内核。现有的LLM内核基准测试集中于机器学习操作符，未对不规则的、异构的、数据移动密集型的数据库风格操作符进行测试。我们提出了DataKernelBench，它将SQL转换为经过验证的PyTorch TorchPlan程序，并评估优化核心张量约束片段或通过执行引导修复优化完整查询的LLM。在使用H100 GPU的TPC-H SF10上，十个专有和开放权重模型中，最强的完整查询CUDA配置在全通过率下实现了$2.11 imes$的加速。我们发现，高性能实现通常使用内核融合和执行策略变化，性能更强的模型最能从完整查询专业化中受益，而工作负载上下文比硬件上下文更为重要。为了处理大于GPU内存的数据，我们扩展了TorchPlan，使用Dask-cuDF在TPC-H SF100上进行按需分区加载，使用四个H100 GPU实现了$2.54 imes$的加速。
View on arXiv Download PDF AI Translation
cs.CL / 11 / 2608.25071

HealthBench-Psych: A Mental Health Subset of OpenAI's HealthBench

HealthBench-Psych：OpenAI HealthBench 的心理健康子集
Flathers, Matthew, Nguyen, Phuong Anh, Noorily, Jill, Herpertz, Julian, Chen, Meiting, Multani, Jasreen, Powell, Samuel, Granof, Mason, Kalinch, Mark, Torous, John
Abstract
General-purpose health benchmarks increasingly anchor claims about LLM medical performance, but they are not always resolved by clinical specialty, making domain-specific performance hard to isolate. Mental health is of acute public-health concern as millions of people turn to LLMs for psychological support, and most existing evaluations are bespoke academic benchmarks that are difficult to integrate into developer workflows. We introduce HealthBench-Psych and HealthBench-Psych-Hard. We screened HealthBench's 5,000 physician-rubric conversations for mental-health relevance with a transparent LLM-applied rubric, then validated the subset through two rounds of blinded clinician review with concealed known-exclude controls, yielding 610 conversations (12.2% of the corpus). Evaluating 20 frontier and open models under a cross-vendor panel of three LLM judges, we find a statistically tied frontier cluster, measurable refusal behavior in two models, and near-identical rankings across judges ($\tau \ge 0.92$). We release the subset, pipeline, model responses, grades, and analysis code as a reusable resource.
Chinese Translation
通用健康基准越来越多地支撑关于大型语言模型（LLM）医疗表现的主张，但它们并不总是按临床专业进行划分，这使得领域特定的表现难以孤立。心理健康是一个急需关注的公共卫生问题，因为数百万人寻求 LLM 的心理支持，而现有的大多数评估都是定制的学术基准，难以融入开发者的工作流程。我们介绍了 HealthBench-Psych 和 HealthBench-Psych-Hard。我们使用透明的 LLM 应用标准筛选了 HealthBench 的 5000 个医生评分对话，以评估其心理健康相关性，然后通过两轮盲法临床医生评审（包含隐蔽的已知排除对照）验证了该子集，最终获得了 610 个对话（占语料库的 12.2%）。在一个由三位 LLM 评审组成的跨供应商小组下评估 20 个前沿和开放模型，我们发现一个统计上相互关联的前沿集群、两个模型中可测量的拒绝行为，以及评审之间几乎相同的排名（$ au ge 0.92$）。我们将该子集、流程、模型响应、评分和分析代码作为可重复使用的资源发布。
View on arXiv Download PDF AI Translation
cs.CL / 12 / 2608.25085

MTDiag: A Multi-Turn Diagnostic Dataset Towards Clinically Meaningful LLM Evaluation

MTDiag：面向临床意义的大型语言模型评估的多轮诊断数据集
Chouayfati, Pia, Fichtl, Alexander M., Anschütz, Miriam, Doumat, George, Groh, Georg
Abstract
Clinical diagnosis is fundamentally interactive and incremental, yet the dominant paradigm for evaluating Large Language Models (LLMs) in medicine remains static QA benchmarks or template-based dialogues. These benchmarks say little about whether a model can serve as a diagnostic agent in a dynamic clinical encounter, with LLMs showing significant accuracy and reliability degradation in multi-turn settings. To address this issue, we present MTDiag, a large multi-turn diagnostic dialogue dataset constructed from three heterogeneous sources: DDXPlus, MIMIC-IV, and published case reports (AJCR), covering common ED presentations as well as long-tail rare and atypical conditions. All cases are normalized into a canonical schema anchored in the most comprehensive and widely-adopted medical knowledge bases (UMLS concept identifiers, with ICD-10 diagnosis codes). We release the schema, a UserLM-8B-based utterance-generation pipeline, and the physician-validated dataset that converts structured clinical evidence into natural-language utterances. Importantly, we introduce and motivate clinical knowledge-grounded metrics for evaluating LLMs as diagnostic agents, beyond diagnostic accuracy, for the task of multi-turn differential diagnosis.
Chinese Translation
临床诊断本质上是互动和渐进的，但目前评估医学领域大型语言模型（LLMs）的主流方法仍然是静态的问答基准或基于模板的对话。这些基准对模型在动态临床接触中能否作为诊断代理几乎没有提供信息，而LLMs在多轮对话环境中的准确性和可靠性显著下降。为了解决这一问题，我们提出了MTDiag，这是一个大型多轮诊断对话数据集，构建于三个异构来源：DDXPlus、MIMIC-IV和已发布的病例报告（AJCR），涵盖了常见的急诊表现以及长尾稀有和非典型疾病。所有案例都被规范化为一个基于最全面和广泛采用的医学知识库（UMLS概念标识符，包含ICD-10诊断代码）的标准模式。我们发布了该模式、基于UserLM-8B的发言生成管道，以及经过医生验证的数据集，该数据集将结构化的临床证据转换为自然语言发言。重要的是，我们引入并激励了基于临床知识的指标，以评估LLMs作为诊断代理的能力，超越诊断准确性，专注于多轮鉴别诊断任务。
View on arXiv Download PDF AI Translation
cs.CL / 13 / 2608.25089

Apples to Apples? Towards Comparable Crosslingual Language Model Evaluation

苹果对苹果？迈向可比较的跨语言语言模型评估
Yang, Xiulin, Wilcox, Ethan Gotlieb, Arnett, Catherine
Abstract
Crosslingual evaluation of language models that enables fair comparisons remains a fundamental challenge in multilingual NLP. Existing studies adopt a variety of downstream tasks and intrinsic metrics with different theoretical justifications, yet there has been little empirical investigation into whether these approaches yield meaningful crosslingual conclusions. We systematically examine crosslingual evaluation approaches using controlled monolingual language models trained on parallel data with varying tokenizer vocabulary sizes and model sizes, and further validate our findings on multilingual LLMs. We further discuss challenges in achieving comparable downstream evaluation across languages. Our results show that several widely used normalized metrics introduce crosslinguistic biases rooted in tokenization, encoding, and orthographic differences. In contrast, sentence-level negative log-likelihood computed over semantically equivalent sequences provides more meaningful and consistent crosslingual comparisons.
Chinese Translation
跨语言语言模型的评估能够实现公平比较，仍然是多语言自然语言处理中的一个基本挑战。现有研究采用了多种下游任务和内在指标，具有不同的理论依据，但对于这些方法是否能够得出有意义的跨语言结论，实证研究仍然较少。我们系统地检验了跨语言评估方法，使用在平行数据上训练的受控单语语言模型，考虑了不同的分词器词汇大小和模型规模，并进一步在多语言大语言模型（LLMs）上验证我们的发现。我们还讨论了在不同语言之间实现可比较的下游评估所面临的挑战。我们的结果表明，几种广泛使用的归一化指标引入了根植于分词、编码和正字法差异的跨语言偏差。相比之下，在语义等价序列上计算的句子级负对数似然提供了更有意义和一致的跨语言比较。
View on arXiv Download PDF AI Translation
cs.CL / 14 / 2608.25115

Less can be More: Relieving RAG Bottlenecks via Evidence Frontloading and Pressure-Adaptive Budgeting

少即是多：通过证据前置和压力自适应预算缓解 RAG 瓶颈
Cai, Weibin, Zafarani, Reza
Abstract
Existing methods for improving Retrieval-Augmented Generation (RAG) efficiency mainly optimize downstream LLM generation, such as context compression or serving optimization. However, RAG is an end-to-end system, and its bottleneck can shift between upstream reranking and downstream generation under different serving loads and reranking budgets.In this paper, we first empirically characterize this shifting-bottleneck behavior and show that upstream reranking can become the dominant bottleneck under high query rates or large reranking budgets. Reducing the reranking budget can relieve this bottleneck, but it may also drop supporting evidence and degrade recall. To address this problem, we propose \textbf{\textsf{PACE}} (\textbf{P}rioritized \textbf{A}daptive \textbf{C}overage of \textbf{E}vidence), a training-free framework that combines \textit{evidence frontloading} with \textit{pressure-adaptive budgeting}. \textsf{PACE} first reorders candidates by marginal evidence coverage, prioritizing documents that are query-relevant, complementary, and useful for forming multi-hop evidence chains. We show that this objective is monotone submodular, giving greedy selection a $(1-1/e)$ approximation guarantee. \textsf{PACE} then dynamically adjusts the reranking budget according to the relative pressure of the reranker and the LLM. Experiments on three multi-hop QA datasets and online serving simulations show that \textsf{PACE} improves evidence recall, reduces p95 latency under ranking-heavy workloads. More importantly, the two components together reveal that \textit{less can be more}: an evidence-dense top-ranked candidates enable higher final recall with fewer reranked documents.
Chinese Translation
现有的提高检索增强生成（Retrieval-Augmented Generation, RAG）效率的方法主要优化下游大型语言模型（LLM）生成，例如上下文压缩或服务优化。然而，RAG 是一个端到端系统，其瓶颈可能在不同的服务负载和重新排序预算下在上游重新排序和下游生成之间转移。本文首先通过实证研究表征了这种转移瓶颈行为，并展示了在高查询率或大重新排序预算下，上游重新排序可能成为主要瓶颈。减少重新排序预算可以缓解这一瓶颈，但也可能导致支持证据的丢失并降低召回率。为了解决这个问题，我们提出了 extbf{ extsf{PACE}}（ extbf{P}rioritized extbf{A}daptive extbf{C}overage of extbf{E}vidence），这是一个无训练的框架，结合了 extit{证据前置} 和 extit{压力自适应预算}。 extsf{PACE} 首先通过边际证据覆盖对候选项进行重新排序，优先考虑与查询相关、互补且对形成多跳证据链有用的文档。我们证明了这一目标是单调子模的，从而为贪心选择提供了 $(1-1/e)$ 的近似保证。接着， extsf{PACE} 根据重新排序器和 LLM 的相对压力动态调整重新排序预算。在三个多跳问答数据集和在线服务模拟实验中， extsf{PACE} 提高了证据召回率，并在重排序负载下减少了 p95 延迟。更重要的是，这两个组件共同揭示了 extit{少即是多}：证据密集的高排名候选项能够以更少的重新排序文档实现更高的最终召回率。
View on arXiv Download PDF AI Translation
cs.CL / 15 / 2608.25123

SelfGraphRAG: Bridging the Supervision Gap in Graph-Based RAG with Synthetic QA Generation

SelfGraphRAG：通过合成问答生成弥合图基RAG中的监督差距
Lagnese, Ben, Gaur, Manas
Abstract
Retrieval-augmented generation (RAG) improves large language models by incorporating external knowledge without retraining, but existing methods often underuse the relational structure encoded in knowledge graphs. Graph-based RAG can capture entity relationships, yet supervised graph retrieval typically requires labeled question-answer data that may not be available for newly constructed graphs. We address this limitation with SelfGraphRAG, a framework that generates question-answer pairs directly from knowledge graph structure and uses them to train a query-conditioned graph retriever. The generated questions capture multi-hop paths and local neighborhoods, providing relational supervision without manual annotation. Experiments on multi-hop question answering and classification benchmarks show that SelfGraphRAG improves retrieval precision and downstream reasoning performance over embedding-based baselines. These results suggest that knowledge graph structure can provide useful supervision for training graph retrievers when labeled data are unavailable.
Chinese Translation
检索增强生成（RAG）通过引入外部知识来改善大型语言模型，而无需重新训练，但现有方法往往未充分利用知识图谱中编码的关系结构。基于图的RAG能够捕捉实体关系，但监督图检索通常需要标注的问题-答案数据，而这些数据可能在新构建的图中不可用。我们通过SelfGraphRAG解决了这一限制，该框架直接从知识图谱结构生成问题-答案对，并利用这些对训练查询条件图检索器。生成的问题捕捉多跳路径和局部邻域，提供关系监督而无需手动标注。在多跳问答和分类基准上的实验表明，SelfGraphRAG在检索精度和下游推理性能方面优于基于嵌入的基线。这些结果表明，当缺乏标注数据时，知识图谱结构可以为训练图检索器提供有用的监督。
View on arXiv Download PDF AI Translation
cs.CL / 16 / 2608.25152

Belief Cascades Drive Persuasion in LLM Agent Networks

信念级联驱动大语言模型代理网络中的劝说
Qiu, Haoyi, Liu, Genglin, Venkit, Pranav Narayanan, Huang, Kung-Hsiang, Gabriel, Saadia, Wu, Chien-Sheng, Peng, Nanyun
Abstract
Multi-agent LLM systems increasingly debate answers, coordinate research, simulate users, and mediate information flows, making agent-to-agent persuasion a basic but undermeasured capability. We introduce a controlled testbed for studying how goal-directed persuaders shift elicited stances in networks of LLM agents grounded in real-world ego-network topologies. Across four LLM backbones, five graphs, and 55 policy statements, we find that persuasion dynamics depend on the interaction between topology, competition, topic, and model prior. Additionally, we show that direct exposure reliably predicts next-round stance change in competing runs, and peer relays carry smaller but measurable influence, showing that agents not assigned to persuade can still transmit persuasive force. Finally, analyzing post text alone misses important movement: planned strategies are only partly realized in executed messages, action choices can diverge from message content, and persuadees rarely state the stance shifts detected by probes. These results argue for evaluating multi-agent persuasion as a trajectory- and exposure-level process, using belief probes, exposure provenance, and action logs to identify who influenced whom and whether visible language reflects underlying stance movement.
Chinese Translation
多智能体大语言模型（LLM）系统越来越多地进行答案辩论、协调研究、模拟用户和调解信息流，使得代理间的劝说成为一种基本但未被充分测量的能力。我们引入了一个受控测试平台，用于研究目标导向的劝说者如何在基于现实世界自我网络拓扑的LLM代理网络中改变引发的立场。在四个LLM基础模型、五个图形和55个政策声明的研究中，我们发现劝说动态依赖于拓扑、竞争、主题和模型先验之间的相互作用。此外，我们还表明，直接接触可靠地预测竞争运行中的下一轮立场变化，而同伴中继则带来较小但可测量的影响，显示未被指派进行劝说的代理仍然可以传递劝说力量。最后，仅分析后文本会错过重要的动态：计划策略在执行消息中仅部分实现，行动选择可能与消息内容偏离，而被劝说者很少陈述探测到的立场变化。这些结果主张将多代理劝说评估为一个轨迹和接触层面的过程，使用信念探测、接触来源和行动日志来识别谁影响了谁，以及可见语言是否反映了潜在的立场变化。
View on arXiv Download PDF AI Translation
cs.CL / 17 / 2608.25166

The Changing Geometry of Grammar: Dimensionality and Neighborhood Reorganization across Transformer Layers

语法的变化几何：变换器层中的维度与邻域重组
Vallisa, Samuele, Ravenda, Federico, Palominos, Claudio, He, Rui, Raballo, Andrea, Mira, Antonietta, Homan, Philipp, Hinzen, Wolfram
Abstract
Transformer representations describe trajectories through high-dimensional vector spaces, which are shaped dynamically as tokens incorporate relational context across layers. Such data tend to concentrate on lower-dimensional sub-manifolds, a form of compression quantified by the Intrinsic Dimensionality (ID), the minimum number of independent variables needed to represent them without significant information loss. In this work, we ask whether the grammatical role of tokens, as marked by their part-of-speech (PoS) tag, shapes the local geometry of this manifold. To this end: (1) We investigate the layer-wise evolution of ID, finding that closed-class items expand earlier and collapse sooner than open-class ones; (2) We show its expansion and contraction to be explained by changes in the neighborhood structure, and hence in the relations between words within a sentence; (3) We compare encoders (ModernBERT, bigbird-roberta-large) and decoders (gemma-2-2B, Llama-3.2-3B), finding that the two families evolve differently across layers, consistently with how each integrates context;(4) We show that geometric features alone recover a token's grammatical role, and use them to interpret how the semantic content of each PoS evolves across layers in a downstream classification task.
Chinese Translation
变换器表示描述了在高维向量空间中的轨迹，这些轨迹随着标记在各层中融入关系上下文而动态形成。这类数据往往集中在低维子流形上，这是一种通过内在维度（Intrinsic Dimensionality, ID）量化的压缩形式，ID是表示这些数据所需的独立变量的最小数量，且不会显著损失信息。在本研究中，我们探讨了标记的语法角色（由其词性标签（Part-of-Speech, PoS）标记）是否影响该流形的局部几何特征。为此：（1）我们研究了ID的层级演变，发现闭合类项目比开放类项目更早扩展且更早收缩；（2）我们表明其扩展和收缩可以通过邻域结构的变化来解释，因此也影响句子中词与词之间的关系；（3）我们比较了编码器（ModernBERT, bigbird-roberta-large）和解码器（gemma-2-2B, Llama-3.2-3B），发现这两类模型在各层中的演变方式不同，这与它们整合上下文的方式一致；（4）我们展示了仅凭几何特征就能恢复标记的语法角色，并利用这些特征解释每个词性在下游分类任务中如何随层次演变其语义内容。
View on arXiv Download PDF AI Translation
cs.CL / 18 / 2608.25190

BanglaMamba: Exploring State Space Models for Bangla Fake News Detection

BanglaMamba：探索用于孟加拉假新闻检测的状态空间模型
Siam, M. K. Khalidi
Abstract
Fake news detection has become an important Natural Language Processing (NLP) task due to the rapid spread of misinformation through online news platforms and social media. While transformer-based models such as BanglaBERT achieve strong performance for Bangla text classification, their quadratic computational complexity makes them less suitable for long-document processing in resource-constrained environments. This paper investigates Mamba-based State Space Models (SSMs) as an efficient alternative for Bangla fake news detection. We propose BanglaMamba and compare it with pre-trained BanglaBERT and a similarly configured BERT model trained from scratch. Experimental results show that BanglaBERT achieves the highest Macro-F1 score (0.9260), while BanglaMamba (0.9029) achieves performance comparable to the from-scratch CustomBERT (0.9057) despite using a different architecture. Meanwhile, BanglaMamba achieves approximately $2.2\times$ higher inference throughput and 49% lower inference peak GPU memory usage than the BERT-based models. Cross-dataset evaluation shows that BanglaBERT generalizes better to an external dataset, highlighting the importance of large-scale pretraining. These findings demonstrate that Mamba-based SSMs can provide a competitive and computationally efficient alternative to Transformer-based architectures for Bangla fake news detection, particularly in resource-constrained settings.
Chinese Translation
假新闻检测已成为一个重要的自然语言处理（NLP）任务，因为虚假信息通过在线新闻平台和社交媒体的快速传播。尽管基于变换器的模型如BanglaBERT在孟加拉文本分类中表现出色，但其二次计算复杂度使其在资源受限的环境中不太适合处理长文档。本文研究了基于Mamba的状态空间模型（SSMs）作为孟加拉假新闻检测的高效替代方案。我们提出了BanglaMamba，并将其与预训练的BanglaBERT和从头开始训练的类似配置的BERT模型进行了比较。实验结果表明，BanglaBERT达到了最高的宏观F1分数（0.9260），而BanglaMamba（0.9029）的性能与从头训练的CustomBERT（0.9057）相当，尽管使用了不同的架构。同时，BanglaMamba的推理吞吐量约为BERT模型的$2.2 imes$，推理峰值GPU内存使用量降低了49%。跨数据集评估表明，BanglaBERT在外部数据集上具有更好的泛化能力，突显了大规模预训练的重要性。这些发现表明，基于Mamba的SSMs可以为孟加拉假新闻检测提供一种具有竞争力且计算高效的替代方案，特别是在资源受限的环境中。
View on arXiv Download PDF AI Translation
cs.CL / 19 / 2608.25243

From Memorization to Absorption: Mixed-Policy RL for Continual Knowledge Injection

从记忆到吸收：用于持续知识注入的混合策略强化学习
Hou, Zhibo, Zhao, Fan, An, Zhiyu, Du, Wan
Abstract
Continual knowledge injection is essential for keeping large language models up-to-date in a fast-evolving world. Existing methods rely on supervised fine-tuning (SFT), which memorizes injected facts in their training format but fails to generalize across paraphrasing, document combinations, and reasoning. To address this, we propose Golden-GRPO Injection (GRIN), a three-stage self-learning framework for continual knowledge injection. Golden-GRPO is a mixed-policy reinforcement learning algorithm designed specifically for knowledge injection, which injects a golden answer to provide learning signal even when on-policy rollouts fail on novel facts. We further introduce Blank and Counter, two document-level benchmarks targeting novel acquisition and counterfactual overwrite respectively, each evaluating single-fact recall, multi-source retrieval, and inferential reasoning. Our experiments establish a clear empirical claim: mixed-policy reinforcement learning enables knowledge absorption beyond what supervised fine-tuning can achieve. GRIN substantially outperforms SFT and mixed-policy RL baselines on the harder question types while matching them on basic fact recall.
Chinese Translation
持续知识注入对于在快速发展的世界中保持大型语言模型的最新状态至关重要。现有方法依赖于监督微调（SFT），这种方法在训练格式中记忆注入的事实，但在改写、文档组合和推理方面缺乏泛化能力。为了解决这个问题，我们提出了Golden-GRPO注入（GRIN），这是一个用于持续知识注入的三阶段自学习框架。Golden-GRPO是一种专门为知识注入设计的混合策略强化学习算法，它注入一个黄金答案，即使在策略执行对新事实失败时也能提供学习信号。我们进一步引入了Blank和Counter两个文档级基准，分别针对新颖获取和反事实覆盖，每个基准评估单一事实回忆、多源检索和推理能力。我们的实验明确表明：混合策略强化学习使知识吸收超越了监督微调所能达到的水平。GRIN在更难的问题类型上显著优于SFT和混合策略强化学习基线，同时在基本事实回忆上与它们相当。
View on arXiv Download PDF AI Translation
cs.CL / 20 / 2608.25276

Groundhog Bit-Flip Attack: Seeding Infinite Generation Loops in Mixture-of-Experts LLMs through Bit Flips

土拨鼠位翻转攻击：通过位翻转在混合专家大语言模型中引发无限生成循环
Lin, Huakang, Zheng, Tiancheng, Sun, Mingxuan, Xu, Tianhong, Zhang, Fan, Fei, Yunsi, Ding, Ruyi
Abstract
Mixture-of-Experts (MoE) architectures enable scalable and efficient large language models (LLMs) by selectively activating expert sub-networks through a routing mechanism. However, this adaptive design introduces a new attack surface: specific experts become disproportionately correlated with certain tokens (e.g., end-of-sequence), allowing adversaries to manipulate model behavior via lightweight perturbations. In this work, we present \textbf{Groundhog Bit-Flip Attack (GBFA)}, the first bit-flip-based \textit{ Denial-of-Wallet availability attack} against MoE-based LLMs. By identifying and flipping routing-layer bits associated with related expert activations, we demonstrate that GBFA substantially extends the decoding token usage across three different LLM modes: conversational, reasoning, and agentic tasks, while largely preserving semantic fidelity. Across four main real-world MoE-based LLMs, manually deactivating on average fewer than \textbf{4 experts} drives average output inflation to $\mathbf{5912\%}$, with the majority of test samples reaching max tokens. These results reveal a robustness vulnerability of MoE architectures to bit flip, and highlight the potential of GBFA as an availability attack against LLMs.
Chinese Translation
混合专家（MoE）架构通过路由机制选择性地激活专家子网络，从而实现可扩展和高效的大语言模型（LLMs）。然而，这种自适应设计引入了新的攻击面：特定专家与某些令牌（例如，序列结束）之间的相关性不成比例，使得对手能够通过轻量扰动操控模型行为。在本研究中，我们提出了 extbf{土拨鼠位翻转攻击（GBFA）}，这是针对基于MoE的大语言模型的首个基于位翻转的 extit{拒绝服务攻击}。通过识别并翻转与相关专家激活相关的路由层位，我们证明GBFA显著扩展了在三种不同LLM模式下的解码令牌使用：对话、推理和代理任务，同时在很大程度上保持语义的准确性。在四个主要的基于MoE的真实世界LLM中，手动停用平均不到 extbf{4个专家}的情况下，平均输出膨胀达到$ extbf{5912 ext{%}}$，大多数测试样本达到最大令牌。这些结果揭示了MoE架构对位翻转的鲁棒性脆弱性，并突显了GBFA作为针对LLM的可用性攻击的潜力。
View on arXiv Download PDF AI Translation
cs.CL / 21 / 2608.25277

Routed Graph Handoff: Adaptive Format Selection for Multi-Agent LLM Delegation

路由图切换：多智能体大语言模型委托的自适应格式选择
Banerjee, Pratyay, Chadha, Ankit
Abstract
Multi-agent LLM systems coordinate through natural-language messages that consume 40--60\% of their token budget. Replacing these with structured graphs reduces cost but fails on tasks requiring adaptive reasoning. We propose \textbf{Routed Graph Handoff}, where a lightweight LLM router (155 tokens, 0.15\% overhead) selects between a typed dependency graph and natural language for each delegation. On four benchmarks (1,050+ trajectories), the routed system matches or exceeds NL-only on every task: \textbf{+12.7\,pp} on $\tau$-retail at 3.2$\times$ compression ($p{<}0.01$), \textbf{+8.7\,pp} on BrowseComp at 2.2$\times$ compression ($p{<}0.05$), and parity on BFCL and AppWorld. Without the router, graph-only delegation regresses 14.6\,pp on AppWorld; the router eliminates this at near-zero cost. A graph-aware executor prompt is required: the same schema without interpretation guidance yields no gain. An oracle analysis reveals 8.6\,pp of additional headroom, motivating execution-time adaptive routing as future work.
Chinese Translation
多智能体大语言模型（LLM）系统通过自然语言消息进行协调，这些消息消耗了其令牌预算的40%至60%。用结构化图替代这些消息可以降低成本，但在需要自适应推理的任务上表现不佳。我们提出了 extbf{路由图切换}（Routed Graph Handoff），其中一个轻量级的LLM路由器（155个令牌，0.15%的开销）为每个委托在类型依赖图和自然语言之间进行选择。在四个基准测试（超过1,050条轨迹）中，路由系统在每个任务上都与仅使用自然语言的系统相匹配或超越：在3.2倍压缩下，$ au$-零售任务上提高了 extbf{12.7个百分点}（p{<}0.01）；在2.2倍压缩下，BrowseComp任务上提高了 extbf{8.7个百分点}（p{<}0.05）；在BFCL和AppWorld上表现持平。没有路由器的情况下，仅使用图的委托在AppWorld上下降了14.6个百分点；而路由器以近乎零的成本消除了这一问题。需要一个图感知的执行器提示：相同的模式在没有解释指导的情况下不会带来任何收益。一个oracle分析显示了8.6个百分点的额外提升空间，激励未来的执行时间自适应路由研究。
View on arXiv Download PDF AI Translation
cs.CL / 22 / 2608.25336

Provenance Before Prose: Claim-Locked Reporting

叙述之前的来源：声明锁定报告
Fan, Xiao, Li, Jingyuan, Guo, Hongbin, Han, Yubo, Zhang, Yi
Abstract
Large language models (LLMs) can fluently verbalize statistical evidence, yet statistical reports can still drift numerical values, invert effect directions, or restate thresholded contrasts as categorical effects. We frame these failures as a control problem: the evidence-bearing content of a scientific report should be fixed by structured statistical results rather than sampled during prose generation. We therefore use cross-run reproducibility to stress-test whether report-visible numbers and claims are bound before prose generation. Existing controls operate at the text or slot level; a deterministic hybrid template reproduces only 61.1% of report-visible numerical content across seeds because the LLM still selects which findings and numbers the template renders. We propose claim-locked reporting, a provenance-before-prose protocol that fixes the evidence source, numbers, direction, and allowed language strength of each reportable claim before the LLM writes only connective prose. Across fMRI functional-connectivity reporting and randomized controlled trial reporting on Evidence Inference 2.0, claim-locked reporting improves reproducibility over the hybrid template by 37.4 and 20.5 points, respectively. Blinded human audits support the observed direction-preservation and governance trends. In an fMRI cost analysis with DeepSeek, claim-locked reporting also yields the lowest observed token use and median generation latency.
Chinese Translation
大型语言模型（LLMs）能够流畅地表达统计证据，但统计报告仍可能出现数值漂移、效应方向反转或将阈值对比重新表述为类别效应。我们将这些失败框架视为一个控制问题：科学报告中承载证据的内容应由结构化的统计结果固定，而不是在叙述生成过程中进行采样。因此，我们利用跨运行可重复性来压力测试报告中可见的数字和声明是否在叙述生成之前被锁定。现有的控制措施在文本或槽位级别运作；一种确定性的混合模板在不同种子间仅能再现61.1%的报告可见数值内容，因为LLM仍然选择模板呈现哪些发现和数字。我们提出了声明锁定报告（claim-locked reporting），这是一种在LLM仅撰写连接性叙述之前，固定每个可报告声明的证据来源、数字、方向和允许的语言强度的来源优先协议。在fMRI功能连接性报告和关于证据推理2.0的随机对照试验报告中，声明锁定报告分别提高了可重复性，较混合模板提高了37.4和20.5个百分点。盲审人类审核支持观察到的方向保持和治理趋势。在与DeepSeek的fMRI成本分析中，声明锁定报告还产生了最低的观察到的标记使用量和中位生成延迟。
View on arXiv Download PDF AI Translation
cs.CL / 23 / 2608.25343

GUIDE: Generative Unsupervised Chinese Query Correction via Phonetic and Visual Shared-ID Encoding

GUIDE：基于语音和视觉共享ID编码的生成式无监督中文查询纠正
Yang, Lei, Huang, Binbin, Tan, Jiwei, Sui, Xuhui, Tu, Chang, Wang, Yi, Li, Han
Abstract
Chinese query correction (CQC) is important for search and query recommendation on content platforms, but supervised methods rely on large annotated correction pairs that are costly to maintain as query vocabularies evolve. Unsupervised correction with language models is attractive, yet in the short-query setting, unconstrained generation often over-corrects ambiguous inputs toward high-frequency phrases, causing intent drift. We propose \textsc{GUIDE}, a generative unsupervised framework for CQC based on a confuse-then-clarify paradigm. \textsc{GUIDE} encodes phonetically or visually confusable characters with shared-IDs and reconstructs the original query with an encoder--decoder architecture, which constrains correction to plausible confusion neighborhoods while learning from unlabeled query streams. A time-decayed, query-frequency-weighted objective further supports adaptation to rapidly changing query vocabularies. Experiments on \textit{QSpell 250K} and a large-scale real-world dataset (\textit{KwaiSearch}) show that \textsc{GUIDE} consistently outperforms strong baselines, while online A/B testing further confirms gains in correction quality and downstream engagement.
Chinese Translation
中文查询纠正（CQC）在内容平台的搜索和查询推荐中至关重要，但监督方法依赖于大量注释的纠正对，而这些在查询词汇不断演变的情况下维护成本高昂。基于语言模型的无监督纠正具有吸引力，但在短查询设置中，无限制的生成往往会将模糊输入过度纠正为高频短语，导致意图漂移。我们提出了 extsc{GUIDE}，一种基于混淆-澄清范式的生成式无监督CQC框架。 extsc{GUIDE}通过共享ID编码语音或视觉上易混淆的字符，并使用编码器-解码器架构重构原始查询，从而将纠正限制在合理的混淆邻域，同时从未标记的查询流中学习。一个时间衰减、查询频率加权的目标进一步支持对快速变化的查询词汇的适应。在 extit{QSpell 250K}和一个大规模真实世界数据集（ extit{KwaiSearch}）上的实验表明， extsc{GUIDE}始终优于强基线，而在线A/B测试进一步确认了纠正质量和下游参与度的提升。
View on arXiv Download PDF AI Translation
cs.CL / 24 / 2608.25347

Short Horizons and Sparse Concepts: a Mathematical View of the Readout in the J-lens

短视野与稀疏概念：对 J-透镜读出过程的数学视角
Yan, Shi-Qi, Ding, Kai-Xuan, Tan, Chao-Hong, Chen, Qian, Wang, Wen, Li, Xiangang, Ling, Zhen-Hua
Abstract
The Jacobian lens (J-lens) has been proposed as a way to read verbalizable representations from language models. However, its principle and meaning lack a detailed and theoretical discussion. We provide a mathematical view of this interpretation and of its assumed causal structure. Besides treating the J-lens as a heuristic probe, we further regard it as a first-order causal transfer operator from intermediate activations to expected future readouts. We study the Jacobian matrix as the optimal local linear approximation of the downstream mapping, analyze its global approximation behavior and bias, and identify its mathematical meaning as an expectation over anticipated future readouts. Further analysis of the Jacobian energy distribution reveals that its causal geometry is highly sparse. The energy decays with depth, concentrates in an extremely small proportion, and decomposes into diagonal pathways and specific critical positions. This decomposition further resolves the expectation of the J-lens over future outputs into short-horizon and sparse concept predictions, providing a more intuitive attribution and explanation for the ability of the J-lens to visualize concepts during the thinking process. Based on the theory, we propose a simple but effective improvement strategy and decoupling method for the J-lens, which significantly enhances the ability of the J-lens to read out correct intermediate concepts.
Chinese Translation
雅可比透镜（J-透镜）被提出作为从语言模型中读取可言说表征的一种方式。然而，其原理和意义缺乏详细的理论讨论。我们提供了对这一解释及其假定因果结构的数学视角。除了将 J-透镜视为一种启发式探针外，我们进一步将其视为从中间激活到预期未来读出的一级因果转移算子。我们研究了雅可比矩阵作为下游映射的最佳局部线性近似，分析其全局近似行为和偏差，并识别其在预期未来读出上的数学意义。对雅可比能量分布的进一步分析揭示了其因果几何高度稀疏。能量随着深度衰减，集中在极小的比例上，并分解为对角路径和特定的关键位置。这种分解进一步将 J-透镜对未来输出的期望解析为短视野和稀疏概念预测，为 J-透镜在思维过程中可视化概念的能力提供了更直观的归因和解释。基于该理论，我们提出了一种简单但有效的改进策略和解耦方法，以显著增强 J-透镜读取正确中间概念的能力。
View on arXiv Download PDF AI Translation
cs.CL / 25 / 2608.25359

Leveraging Speech Acts for Low-Data and Cross-Domain Conversation Derailment Forecasting

利用言语行为进行低数据和跨领域对话偏离预测
Yuan, Angela Yifei, De Kock, Christine, Leckie, Christopher
Abstract
Conversational derailment forecasting aims to predict when online discussions will escalate into hostility, enabling proactive moderation. Existing approaches often struggle in low-data settings and to generalize across domains. This poses a challenge for new platforms and smaller communities where annotated data is limited. We propose modeling pragmatic representations of conversations to reduce lexical noise and improve generalizability. Specifically, speech act information is used as an auxiliary learning signal alongside textual semantics. Experimental results show improved performance across three datasets, particularly in low-data and cross-domain settings.
Chinese Translation
对话偏离预测旨在预测在线讨论何时会升级为敌对行为，从而实现主动管理。现有的方法在低数据环境下和跨领域推广方面常常面临挑战。这对新平台和较小社区来说尤为困难，因为它们的标注数据有限。我们提出通过建模对话的语用表示来减少词汇噪声并提高推广能力。具体而言，言语行为信息被用作文本语义的辅助学习信号。实验结果表明，在三个数据集上，尤其是在低数据和跨领域设置中，性能得到了改善。
View on arXiv Download PDF AI Translation
cs.CL / 26 / 2608.25379

Adaptive Triggering for Bias Correction in LLM Reasoning

用于大语言模型推理偏差修正的自适应触发机制
Kim, Nayoung, Mancenido, Mickey, Liu, Huan
Abstract
Chain-of-thought prompting can expose and amplify demographic stereotypes within an LLM's intermediate reasoning and create a failure mode that final-answer debiasing alone cannot address. Mitigating such bias during generation presents a fundamental timing problem: intervening too late allows biased reasoning to propagate, while unnecessarily intervening can disrupt otherwise correct reasoning. Existing approaches largely avoid this decision by either evaluating completed reasoning chains post hoc or intervening at predetermined steps, leaving open when a developing reasoning trajectory provides sufficient evidence to warrant correction. We formulate this decision as an online change-point detection problem. A per-step bias signal updates a CUSUM statistic and a targeted correction is injected only when accumulated evidence crosses a detector-specific threshold calibrated on held-out data. We instantiate the framework with a white-box signal derived from next-token probabilities and a black-box signal obtained from an LLM judge, enabling deployment with both open-weight and hosted models. On gpt-4o-mini adaptive black-box triggering recovers most of the disambiguated-context accuracy lost under fixed-interval intervention while requiring substantially fewer interventions. That result holds even with an independent judge. Across six open-weight models, the white-box signal improves ambiguous-item accuracy on all six but reduces disambiguated-item accuracy on five because it cannot distinguish unsupported stereotype reliance from correct, stereotype-congruent evidence.
Chinese Translation
思维链提示可能会暴露并放大大语言模型（LLM）中间推理中的人口统计学偏见，并产生一种仅通过最终答案去偏见无法解决的失败模式。在生成过程中减轻这种偏见呈现出一个根本的时机问题：过晚干预会导致偏见推理的传播，而不必要的干预则可能会干扰本来正确的推理。现有方法在很大程度上通过事后评估完成的推理链或在预定步骤进行干预来规避这一决策，未能明确在何时发展中的推理轨迹提供了足够的证据以进行修正。我们将这一决策形式化为在线变点检测问题。每一步的偏见信号更新CUSUM统计量，只有在累积证据超过在保留数据上校准的检测器特定阈值时，才会注入针对性的修正。我们用从下一个令牌概率派生的白盒信号和从LLM评判者获得的黑盒信号实例化该框架，使其能够在开放权重和托管模型中部署。在gpt-4o-mini自适应黑盒触发下，恢复了大部分在固定间隔干预下丧失的消歧上下文准确性，同时所需干预显著减少。即使在独立评判者的情况下，该结果仍然成立。在六个开放权重模型中，白盒信号提高了所有六个模型的模糊项准确性，但在五个模型上降低了消歧项准确性，因为它无法区分不支持的刻板印象依赖与正确的、与刻板印象一致的证据。
View on arXiv Download PDF AI Translation
cs.CL / 27 / 2608.25398

OmniPhys: A Unified Multimodal Benchmark for Physics Understanding and Generation from Chinese Educational Corpora

OmniPhys：一个统一的多模态基准，用于物理理解和生成，基于中国教育语料库
Chen, Hao, Lin, Yumin, Yushanjiang, Nadila, Lin, Xin, Zhang, Min
Abstract
Multimodal Large Language Models (MLLMs) have demonstrated strong abilities in solving diverse visual and textual reasoning tasks. However, their development in the physics domain is significantly hindered by the lack of a comprehensive benchmark. To fill this gap, we introduce OmniPhys, a large-scale benchmark for multimodal physics understanding and reasoning, covering middle school through university-level problems from Chinese Educational Corpora. OmniPhys consists of 15,246 questions and 19,850 images, accompanied by detailed annotations that support fine-grained analysis of reasoning processes and knowledge usage. Beyond conventional evaluation, OmniPhys is a benchmark that systematically evaluates multimodal outputs in the physics domain, including models' ability to generate structured physics diagrams, which constitute a fundamental component of authentic physics problem solving. Extensive evaluations reveal critical gaps in the capabilities of current MLLMs, especially in complex reasoning and visual generation. To address this, we release OmniPhys to serve as a foundational resource for advancing multimodal intelligence in physics and scientific domains. Codes and data are available at https://github.com/ECNU-RAIL/OmniPhys-EMNLP2026.
Chinese Translation
多模态大型语言模型（MLLMs）在解决多样的视觉和文本推理任务方面展现了强大的能力。然而，它们在物理领域的发展受到缺乏全面基准的显著制约。为填补这一空白，我们推出了OmniPhys，这是一个大规模的多模态物理理解和推理基准，涵盖了从中学到大学水平的中国教育语料库中的问题。OmniPhys包含15,246个问题和19,850张图像，并附有详细的注释，以支持对推理过程和知识使用的细致分析。除了传统评估，OmniPhys还是一个系统性评估物理领域多模态输出的基准，包括模型生成结构化物理图的能力，这构成了真实物理问题解决的基本组成部分。广泛的评估揭示了当前MLLMs在复杂推理和视觉生成能力方面的关键差距。为了解决这一问题，我们发布了OmniPhys，以作为推动物理和科学领域多模态智能发展的基础资源。代码和数据可在https://github.com/ECNU-RAIL/OmniPhys-EMNLP2026获取。
View on arXiv Download PDF AI Translation
cs.CL / 28 / 2608.25428

DCGC: Draft-Conditioned Global Correction for Complex Reasoning with Masked Diffusion Models

DCGC：基于草稿条件的全球修正用于复杂推理的掩蔽扩散模型
Oh, Minhae, Lee, Nakyung, Lee, Jungwoo
Abstract
Correcting flawed reasoning traces remains a significant challenge for Large Language Models (LLMs), whose autoregressive generation can propagate early mistakes into subsequent reasoning. We introduce DCGC, a Masked Diffusion Model (MDM) framework for global correction that uses an imperfect solution draft from an upstream solver as auxiliary context. DCGC combines task-specific Supervised Fine-Tuning (SFT) with a novel inference-time mechanism called Dynamic Dual-CFG. This mechanism separates problem-only and joint problem-draft branches and scales the draft-conditioned residual using a relative confidence gap. Across math, code, and knowledge reasoning benchmarks, DCGC outperforms standard sampling and simpler CFG variants, with additional results suggesting transfer to different diffusion backbones. In test-time setting where ground-truth failure labels are unavailable, DCGC improves full test set accuracy by correcting low-consensus upstream outputs, highlighting its utility as a verifier-free global correction module for difficult reasoning instances.
Chinese Translation
修正有缺陷的推理轨迹仍然是大型语言模型（LLMs）面临的重大挑战，因为其自回归生成可能将早期错误传播到后续推理中。我们提出了DCGC，一种掩蔽扩散模型（MDM）框架，用于全球修正，该框架使用来自上游求解器的不完美解决草稿作为辅助上下文。DCGC将特定任务的监督微调（SFT）与一种称为动态双CFG（Dynamic Dual-CFG）的新颖推理机制相结合。该机制将仅问题分支和联合问题-草稿分支分开，并使用相对置信度差距来缩放草稿条件的残差。在数学、代码和知识推理基准测试中，DCGC的表现优于标准采样和更简单的CFG变体，额外结果表明其可以迁移到不同的扩散骨干网络。在测试时设置中，当真实失败标签不可用时，DCGC通过修正低共识的上游输出提高了完整测试集的准确性，突显了其作为困难推理实例的无验证者全球修正模块的实用性。
View on arXiv Download PDF AI Translation
cs.CL / 29 / 2608.25449

MathAdv: What Theorem Provers Know, Reason, Formalize, and Generalize

MathAdv：定理证明者的知识、推理、形式化与概括能力
Yuan, Jiaxin, Lockhart, Connor Martinez, Liu, Xiaoyu, Wang, Jiaqi, Deng, Chenghao, Han, Xiayimei, Mastrantonis, Vlasios, Gudin, Dmitrii, Zhu, Shaopeng, Mohamed, Abdirisak Abdullahi, Aytekin, Bilal Hamdi, Lang, Jiewen, Song, Zezheng, Huang, Furong
Abstract
Formal theorem proving enables machine-verifiable evaluation of mathematical reasoning, yet existing benchmarks often emphasize aggregate proof accuracy, concentrate on a narrow range of mathematics, and provide limited evidence of robustness to equivalent reformulations. We introduce MathAdv, a diagnostic benchmark spanning 13 domains across undergraduate- and graduate-level mathematics. Alongside Lean 4 theorem proving, MathAdv provides up to three auxiliary tasks: multiple-choice questions that probe mathematical knowledge, fill-in-the-blank problems that isolate informal reasoning, and expert-crafted transformations that test robustness to problem presentation. Our evaluation of contemporary theorem provers yields four findings: formalization remains a major bottleneck; performance varies substantially across mathematical domains; natural-language guidance helps general-purpose LLMs but can hinder proof-specialized models; and mathematically equivalent reformulations expose substantial robustness limitations. Together, these results show how component-wise evaluation can reveal model capabilities and failure modes that aggregate theorem-proving accuracy obscures. The dataset and evaluation scripts are available at https://github.com/margotyjx/MathAdv.git.
Chinese Translation
形式化定理证明使得数学推理的机器可验证评估成为可能，但现有基准往往强调整体证明准确性，集中于狭窄的数学范围，并提供有限的等价重构的鲁棒性证据。我们提出了 MathAdv，这是一个涵盖本科和研究生水平数学的13个领域的诊断基准。除了 Lean 4 定理证明外，MathAdv 还提供多达三个辅助任务：探测数学知识的多项选择题、隔离非正式推理的填空题，以及测试问题呈现鲁棒性的专家设计转换。我们对当代定理证明者的评估得出了四个发现：形式化仍然是一个主要瓶颈；不同数学领域的表现差异显著；自然语言指导对通用大型语言模型（LLMs）有帮助，但可能会妨碍专门用于证明的模型；而数学上等价的重构暴露了显著的鲁棒性限制。这些结果共同表明，逐项评估能够揭示模型能力和失败模式，而这些在整体定理证明准确性中被掩盖。数据集和评估脚本可在 https://github.com/margotyjx/MathAdv.git 获取。
View on arXiv Download PDF AI Translation
cs.CL / 30 / 2608.25478

VietAIDetector: An Open-Source Zero-Shot Detector for Vietnamese AI-Generated Text

VietAIDetector：一个开源的零样本检测器，用于检测越南语AI生成文本
Nguyen, Trieu Hai, Hoang, Van-Dung
Abstract
In recent years, distinguishing between AI-generated text and human-written text has remained a challenge. In this paper, we introduce VietAIDetector, an open-source tool designed specifically for detecting Vietnamese AI-generated text. It allows users to interact through a Gradio web interface with inputs ranging from raw Vietnamese text to common text file formats, including scanned documents and exceptionally long texts that exceed the context size of the employed Large Language Models (LLMs). The core component of the tool employs a Zero-Shot approach to detect AI-generated text without requiring domain-specific training data, building upon the previous VietBinoculars and Binoculars research. The tool is built upon a Vietnamese-specific language model and has been evaluated on out-of-domain datasets, demonstrating superior performance compared to existing methods primarily developed for English. Additionally, users can select optimal detection thresholds based on F1 score, accuracy, or [email protected] requirements. The results are presented through the web interface, allowing users to easily review and verify suspicious texts or download them as a PDF report. The tool is publicly available at https://github.com/trieuntu/VietAIDetector
Chinese Translation
近年来，区分AI生成的文本与人类撰写的文本仍然是一项挑战。本文介绍了VietAIDetector，这是一个专门用于检测越南语AI生成文本的开源工具。该工具允许用户通过Gradio网络界面进行交互，输入范围包括原始越南语文本以及常见文本文件格式，如扫描文档和超出所用大型语言模型（Large Language Models, LLMs）上下文大小的异常长文本。该工具的核心组件采用零样本（Zero-Shot）方法来检测AI生成的文本，无需领域特定的训练数据，基于之前的VietBinoculars和Binoculars研究。该工具基于一个特定于越南语的语言模型构建，并在域外数据集上进行了评估，显示出比主要为英语开发的现有方法更优越的性能。此外，用户可以根据F1分数、准确率或[email protected]要求选择最佳检测阈值。结果通过网络界面呈现，便于用户轻松审查和验证可疑文本或将其下载为PDF报告。该工具已在https://github.com/trieuntu/VietAIDetector上公开提供。
View on arXiv Download PDF AI Translation
cs.CL / 31 / 2608.25487

ReliableRAG: Combating Misinformation in Retrieval-Augmented Generation via Reliability-Guided Reasoning Chains

ReliableRAG：通过可靠性引导推理链对抗检索增强生成中的虚假信息
Jiang, Jinpu, Wu, Xuan, Song, Wenhao, Yang, Bo, Zhou, You, Ge, Hongwei, Lee, Heow Pueh, Liang, Yanchun, Wu, Chunguo
Abstract
Retrieval-Augmented Generation (RAG) has emerged as a powerful architecture for Question Answering (QA) by integrating external information into Large Language Models (LLMs). However, false, inaccurate, and misleading information in news and social media poses a serious challenge to real-world RAG systems, especially in multi-hop QA, where complex multi-step reasoning can be misled by even a single deceptive misinformation segment in the retrieved documents. Existing approaches mainly rely on implicit alignment or explicit regulation, but their limited ability to assess fine-grained information reliability makes them vulnerable to deceptive misinformation that is semantically relevant to the question yet factually incorrect, leading to erroneous answers. To address this limitation, we propose ReliableRAG, which, to the best of our knowledge, is the first reliability-driven framework that mitigates deceptive misinformation in multi-hop QA through fine-grained evaluation of individual triples. ReliableRAG first extracts information segments from source documents and represents them as structured triples. It then quantifies triple reliability by combining query-triple semantic relevance with triple credibility, retaining only the top-$K$ reliable and non-redundant triples. Based on these refined triples, ReliableRAG autoregressively constructs robust reasoning chains to consolidate trustworthy evidence and filter deceptive misinformation, producing accurate answers faithful to reliable information. Experiments on three multi-hop QA datasets show that ReliableRAG outperforms existing methods, substantially improving the factual reliability and robustness of RAG systems under deceptive misinformation injection.
Chinese Translation
检索增强生成（RAG）作为一种强大的问答（QA）架构，通过将外部信息整合到大型语言模型（LLMs）中而崭露头角。然而，新闻和社交媒体中的虚假、不准确和误导性信息对现实世界的RAG系统构成了严重挑战，尤其是在多跳问答中，复杂的多步骤推理可能会因检索文档中的单个误导性虚假信息片段而受到影响。现有的方法主要依赖于隐式对齐或显式调节，但它们在评估细粒度信息可靠性方面的能力有限，使其易受语义相关但事实错误的误导性虚假信息的影响，导致错误答案。为了解决这一局限性，我们提出了ReliableRAG，尽我们所知，这是第一个通过对个体三元组进行细粒度评估来减轻多跳问答中的误导性虚假信息的可靠性驱动框架。ReliableRAG首先从源文档中提取信息片段，并将其表示为结构化三元组。然后，它通过结合查询-三元组的语义相关性与三元组的可信度来量化三元组的可靠性，仅保留前$K$个可靠且不冗余的三元组。基于这些精炼的三元组，ReliableRAG自回归地构建稳健的推理链，以巩固可信证据并过滤误导性虚假信息，从而生成忠实于可靠信息的准确答案。在三个多跳问答数据集上的实验表明，ReliableRAG优于现有方法，显著提高了RAG系统在误导性虚假信息注入下的事实可靠性和鲁棒性。
View on arXiv Download PDF AI Translation
cs.CL / 32 / 2608.25523

TOPAS: Workflow-Aware Prefix-State Scheduling for Multi-Agent LLM Serving

TOPAS：面向工作流的多智能体大语言模型服务前缀状态调度
Ni, Hongqiu, Tian, Han, Zhang, Chi, Li, Guopeng, Tan, Haisheng
Abstract
Prefix caching introduces a fundamental tradeoff in multi-agent large language model (LLM) serving: retaining a long system-prompt key-value (KV) cache for an agent accelerates future calls, yet it reduces the GPU memory available for batching concurrent requests. In multi-stage workflows, existing schedulers tend to prioritize either immediate prefix locality or overall workflow progress. However, under a shared KV cache budget, optimizing either objective in isolation can prolong tasklevel job completion time (JCT) through downstream delays or frequent prefix replacement. To strike a balance, we here propose TOPAS, a Task-Oriented Prefix-Aware Scheduler that jointly decides which agent prefixes to keep in the cache and which requests to schedule for execution. TOPAS scores candidate post-decision states by trading off the expected reduction in each task's longest remaining service path against the near-term benefit of downstream prefix reuse, accounting for the costs of prefix movement and preemption. A task-level aging mechanism is also incorporated to prevent starvation. We implement TOPAS within the SGLang framework and assess its performance on three synthetic DAGs and two MetaGPT software-development workflows. Compared with the best performing baseline for each workload and metric, TOPAS reduces the mean/p99 JCT by up to 39.8%/49.4% on the synthetic workloads, while lowering mean JCT by 9.8% on MetaGPT-SOP and mean/p99 JCT by 22.0%/26.6% on MetaGPT-TL.
Chinese Translation
前缀缓存在多智能体大语言模型（LLM）服务中引入了一个基本的权衡：为一个智能体保留较长的系统提示键值（KV）缓存可以加速未来的调用，但同时会减少可用于批处理并发请求的GPU内存。在多阶段工作流中，现有调度器往往优先考虑即时前缀局部性或整体工作流进展。然而，在共享的KV缓存预算下，单独优化任一目标可能会通过下游延迟或频繁的前缀替换来延长任务级作业完成时间（JCT）。为此，我们提出了TOPAS，一种任务导向的前缀感知调度器，它共同决定保留哪些智能体前缀在缓存中以及调度哪些请求进行执行。TOPAS通过权衡每个任务最长剩余服务路径的预期减少与下游前缀重用的短期收益，来对候选后决策状态进行评分，同时考虑前缀移动和抢占的成本。我们还引入了一种任务级老化机制以防止饥饿。我们在SGLang框架中实现了TOPAS，并在三个合成有向无环图（DAG）和两个MetaGPT软件开发工作流上评估其性能。与每个工作负载和指标的最佳基线相比，TOPAS在合成工作负载上将平均/p99 JCT分别降低了最多39.8%/49.4%，同时在MetaGPT-SOP上降低了平均JCT 9.8%，在MetaGPT-TL上降低了平均/p99 JCT 22.0%/26.6%。
View on arXiv Download PDF AI Translation
cs.CL / 33 / 2608.25531

ClueWeaver: Reward-Guided Dual-Agent Evidence Reasoning for Compact LLMs on Literary Long Narratives

ClueWeaver：基于奖励引导的双代理证据推理框架，用于紧凑型大语言模型在文学长篇叙事中的应用
Zhu, Jihao, Yang, Zhiwei, Zhang, Wenxiao, Zhao, Junqian, You, Qi, Wang, Fangqi, Deng, Zheyuan, Yang, Hanzhe, Liu, Yu, Hong, Jin B.
Abstract
Humanities and social science research requires close reading of long narrative materials such as novels, scripts, archives, and case reports, yet many users have limited access to costly proprietary long-context models. Compact, locally deployable language models are a practical alternative, but directly feeding them an entire long context remains costly, hard to inspect, and prone to missing sparse evidence. We present ClueWeaver, an evidence-aware dual-agent framework for long-narrative question answering with compact local models. A Finder identifies passages containing answer-critical clues through retrieval-guided segmentation, while an Interpreter derives the answer from the selected evidence, produces rationales with paragraph-ID citations, and applies an internal self-calibration pass for high-risk questions. Both agents are optimized with reward-guided reinforcement learning: Finder rewards emphasize evidence retention and faithful paragraph-ID references, and Interpreter rewards emphasize correctness, grounding, and concise explanations. This decomposition makes evidence selection and reasoning more inspectable than end-to-end prompting. Experiments across multiple long-context narrative question answering and claim verification settings show that ClueWeaver substantially improves local end-to-end language models while providing evidence coverage and paragraph-referenced reasoning traces. Code is available at https://github.com/Ameame1/ClueWeaver.
Chinese Translation
人文学科和社会科学研究需要对长篇叙事材料（如小说、剧本、档案和案例报告）进行细致阅读，但许多用户对昂贵的专有长上下文模型的访问有限。紧凑型、可本地部署的语言模型是一个实用的替代方案，但直接将整个长上下文输入仍然成本高昂、难以检查，并且容易遗漏稀疏证据。我们提出了ClueWeaver，一个针对长篇叙事问答的证据感知双代理框架，适用于紧凑型本地模型。Finder通过检索引导的分段识别包含答案关键线索的段落，而Interpreter则从所选证据中推导答案，生成带有段落ID引用的推理，并对高风险问题应用内部自我校准。两个代理均通过奖励引导的强化学习进行优化：Finder的奖励强调证据保留和忠实的段落ID引用，而Interpreter的奖励则强调正确性、基础性和简洁的解释。这种分解使得证据选择和推理比端到端提示更易于检查。在多个长上下文叙事问答和主张验证设置中的实验表明，ClueWeaver显著改善了本地端到端语言模型，同时提供了证据覆盖和段落引用的推理痕迹。代码可在 https://github.com/Ameame1/ClueWeaver 获取。
View on arXiv Download PDF AI Translation
cs.CL / 34 / 2608.25555

Virgil: Navigating Explainability for Transformer-based Language Models

Virgil：为基于变换器的语言模型导航可解释性
Ciaperoni, Martino, Kutluk, Sezer, Muscato, Benedetta, Manerba, Marta Marchiori, Giannotti, Fosca
Abstract
Explainability for transformer-based language models is becoming crucial as these systems are deployed in high-stakes applications. As a result, the ecosystem of explainability tools is rapidly evolving, becoming richer, but also more fragmented and harder to navigate. To address this challenge, we present Virgil, an interactive system that lets practitioners and researchers, including non-experts, navigate explainability tools for transformer language models. Supported by a curated knowledge base, the system enables users to discover and compare explainability tools within a unified interface.
Chinese Translation
随着基于变换器的语言模型在高风险应用中的部署，可解释性变得至关重要。因此，可解释性工具的生态系统正在迅速发展，变得更加丰富，但也更加分散，难以导航。为了解决这一挑战，我们提出了Virgil，一个互动系统，使从业者和研究人员，包括非专家，能够导航基于变换器的语言模型的可解释性工具。该系统由一个经过精心策划的知识库支持，使用户能够在统一的界面中发现和比较可解释性工具。
View on arXiv Download PDF AI Translation
cs.CL / 35 / 2608.25561

EgoArgus: Benchmarking VLMs as Situational Assistants for Modality-Grounded User Supports

EgoArgus：将视觉语言模型作为情境助手的基准测试
Tang, Yu-Chien, Liu, Yu-Hsiang, Yen, An-Zi
Abstract
VLMs are increasingly positioned as daily assistants that perceive first-person environments, follow user dialogue, and decide how to help. Existing egocentric benchmarks mainly evaluate visual understanding in isolation, leaving open whether models can arbitrate between visual evidence and user-provided language when the two are helpful, irrelevant, or conflicting. We introduce EgoArgus, a human-annotated dataset for evaluating egocentric assistants on understanding and decision tasks in five dialogue-video daily scenarios. Our results demonstrate that it is still challenging for current VLMs as reliable egocentric assistants, which requires identifying which modality is trustworthy and deciding when intervention is warranted. Deeper analysis also shows that existing modality bias mitigation methods are quite restricted to enhance performance, providing insights to aid practioners into the deployment of current VLMs as daily assistants.
Chinese Translation
视觉语言模型（VLMs）越来越多地被定位为能够感知第一人称环境、跟随用户对话并决定如何提供帮助的日常助手。现有的以自我为中心的基准主要评估视觉理解的孤立性，尚未解决模型在视觉证据与用户提供的语言之间进行裁决的能力，尤其是在这两者有助、无关或相互冲突的情况下。我们引入了EgoArgus，这是一个经过人工标注的数据集，用于评估自我中心助手在五个对话-视频日常场景中的理解和决策任务。我们的结果表明，当前的视觉语言模型作为可靠的自我中心助手仍面临挑战，这需要识别哪个模态是可信的，并决定何时需要干预。更深入的分析还表明，现有的模态偏见缓解方法在提升性能方面相当有限，为实践者在将当前视觉语言模型作为日常助手部署时提供了见解。
View on arXiv Download PDF AI Translation
cs.CL / 36 / 2608.25569

Controllable Affective Generation via Latent Vector Steering

通过潜在向量引导实现可控情感生成
Yong, Xixian, Chang, Siyuan, Zhang, Yingying, Wu, Xian, Zhou, Xiao
Abstract
Large Language Models (LLMs) often produce emotionally flattened responses after alignment, limiting their effectiveness in affect-sensitive applications. In this paper, we propose EmoVec, a lightweight framework for controllable affective generation via latent vector steering. EmoVec extracts emotion-specific directions from paired neutral and emotion-conditioned responses using contrastive activation addition, and further refines them through task-specific debiasing and principal subspace removal. During inference, these vectors are injected into the final residual stream with static or scenario-adaptive scaling, enabling continuous control over emotional intensity without updating model weights. Experiments across three LLMs and eight emotions show that EmoVec consistently improves emotional salience while largely preserving semantic content, fluency, and coherence. Ablation studies and human evaluation further confirm the effectiveness of vector purification and adaptive scaling, establishing EmoVec as a practical inference-time method for affective control in deployed LLMs.
Chinese Translation
大型语言模型（LLMs）在对齐后往往会产生情感平淡的响应，这限制了它们在情感敏感应用中的有效性。本文提出了EmoVec，一个通过潜在向量引导实现可控情感生成的轻量级框架。EmoVec通过对比激活加法从配对的中性和情感条件响应中提取情感特定方向，并通过任务特定的去偏见和主子空间去除进一步精炼这些方向。在推理过程中，这些向量被注入到最终的残差流中，采用静态或情境自适应的缩放方式，使得在不更新模型权重的情况下能够对情感强度进行持续控制。在对三种LLM和八种情感的实验中，EmoVec始终提高了情感显著性，同时在很大程度上保留了语义内容、流畅性和连贯性。消融研究和人类评估进一步确认了向量净化和自适应缩放的有效性，确立了EmoVec作为一种在部署的LLM中进行情感控制的实用推理时方法。
View on arXiv Download PDF AI Translation
cs.CL / 37 / 2608.25574

Generative vs. Encoder Large Language Models for ASR Evaluation: A Comparative Study

生成模型与编码器大型语言模型在自动语音识别评估中的比较研究
Bañeras-Roux, Thibault, Kumar, Shashi, Khalil, Driss, Burdisso, Sergio, Motlicek, Petr, Liu, Shiran, Rouvier, Mickael, Wottawa, Jane, Dufour, Richard
Abstract
Automatic Speech Recognition (ASR) is typically evaluated using Word Error Rate (WER), which poorly reflects semantic similarity. While embedding-based metrics correlate better with human judgments, the respective roles of encoder and decoder-based Large Language Models (LLMs) remain underexplored. This paper presents a comparative study of both families for ASR evaluation. We analyze BERTScore and SemDist across different LLMs, layers, and pooling strategies, showing that both metrics can achieve strong correlation with human judgments when properly configured. For decoder models, we investigate generative LLMs in two settings: pairwise hypothesis selection via prompting and direct qualitative error classification. Our results show that encoder-based metrics remain highly competitive, while generative LLMs perform strongly in hypothesis comparison and improve the interpretability of ASR evaluation.
Chinese Translation
自动语音识别（ASR）通常使用词错误率（WER）进行评估，但这一指标无法准确反映语义相似性。尽管基于嵌入的度量与人类判断的相关性更高，但编码器和解码器基础的大型语言模型（LLMs）的各自角色仍然未得到充分探索。本文对这两类模型在ASR评估中的表现进行了比较研究。我们分析了不同LLMs、层次和池化策略下的BERTScore和SemDist，结果表明，当配置得当时，这两种度量能够与人类判断实现强相关性。对于解码器模型，我们在两种设置下研究了生成型LLMs：通过提示进行成对假设选择和直接的定性错误分类。我们的结果显示，基于编码器的度量仍然具有很强的竞争力，而生成型LLMs在假设比较中表现出色，并提高了ASR评估的可解释性。
View on arXiv Download PDF AI Translation
cs.CL / 38 / 2608.25579

Cross-Dataset Stability of Expert-Informed Skill Prompting and Fine-Tuning for Chinese Metaphor Identification

专家指导的技能提示与微调在中文隐喻识别中的跨数据集稳定性
Wu, Yufeng, Liu, Meichun
Abstract
Metaphor-identification performance can change markedly across datasets that differ in text distribution and annotation policy. We examine whether a fixed expert-informed procedure produces a more even cross-dataset profile than task-specific parameter adaptation. Four prespecified conditions are compared for Chinese sentence-level metaphor identification: BERT fine-tuning (BERT-FT), QLoRA-based large language model fine-tuning (LLM-FT), direct zero-shot LLM prompting (LLM-ZS), and zero-shot prompting with a frozen procedural Skill (Skill-ZS). The Skill operationalizes established criteria involving contextual meaning, basic meaning, contrast, and comparison. Evaluation covers CMRE Test and two external datasets, CCIME and CMC. Fine-tuned scores are means over three seeds, whereas each zero-shot score comes from one deterministic configuration. Fine-tuning remains strongest on the native test set: BERT-FT reaches 91.76 Macro-F1. LLM-FT has the highest external mean (83.52), while Skill-ZS is close at 82.92 and has both the highest external floor (82.64) and the smallest observed range across all three datasets (4.08 points). In the matched zero-shot comparison, adding the Skill reduces metaphorical predictions on every dataset. This sharply lowers false positives on CCIME but increases false negatives on CMRE Test and CMC. The results position expert-informed Skill prompting as a complementary route to more even observed cross-dataset performance, while fine-tuning retains its advantage in native-data accuracy. To our knowledge, this is the first study to compare an expert-informed procedural Skill with task-specific fine-tuning in the same cross-dataset evaluation of Chinese sentence-level metaphor identification.
Chinese Translation
隐喻识别的性能在文本分布和标注政策不同的数据集之间可能会显著变化。我们研究了固定的专家指导程序是否能比特定任务的参数适应产生更均匀的跨数据集表现。针对中文句子级隐喻识别，我们比较了四个预设条件：BERT微调（BERT-FT）、基于QLoRA的大型语言模型微调（LLM-FT）、直接零-shot大型语言模型提示（LLM-ZS）以及使用冻结程序技能的零-shot提示（Skill-ZS）。该技能实现了涉及上下文意义、基本意义、对比和比较的既定标准。评估涵盖了CMRE测试和两个外部数据集，CCIME和CMC。微调得分是三个种子的平均值，而每个零-shot得分来自一个确定的配置。微调在本地测试集上的表现最强：BERT-FT达到了91.76的宏F1值。LLM-FT在外部数据集中的平均得分最高（83.52），而Skill-ZS则接近82.92，并且在所有三个数据集中具有最高的外部底线（82.64）和最小的观察范围（4.08分）。在匹配的零-shot比较中，添加技能在每个数据集上减少了隐喻预测。这显著降低了CCIME上的假阳性，但在CMRE测试和CMC上增加了假阴性。结果表明，专家指导的技能提示作为一种补充途径，有助于实现更均匀的跨数据集表现，而微调在本地数据准确性方面仍然保持优势。据我们所知，这是首次在中文句子级隐喻识别的同一跨数据集评估中比较专家指导的程序技能与特定任务微调的研究。
View on arXiv Download PDF AI Translation
cs.CL / 39 / 2608.25583

GRIP: Granular Reward-Guided Parameter Interpolation for Efficient Reasoning

GRIP：基于细粒度奖励引导的参数插值以实现高效推理
So, Lam, Wu, Canhui, Lin, Han
Abstract
Reasoning-oriented large language models often achieve strong problem-solving performance by generating long chains of thought, but this behavior substantially increases inference cost and latency. In contrast, instruction-tuned models tend to answer more concisely, yet often lack comparable reasoning ability. This accuracy-efficiency mismatch motivates a lightweight approach that combines the strengths of both models without full model retraining. In this paper, we propose GRIP (Granular Reward-guided Interpolation of Parameters), a reward-guided parameter interpolation framework for efficient reasoning. Given a reasoning model and an instruction model with identical architectures, GRIP assigns learnable interpolation ratios to individual modules and optimizes only these ratios while keeping both source models frozen. The interpolation ratios are trained with a reward signal that favors responses that are both correct and concise. Experiments show that GRIP achieves a better accuracy-efficiency trade-off than fixed or search-based merging baselines and further reveals module-wise fusion patterns associated with efficient reasoning.
Chinese Translation
面向推理的大型语言模型通常通过生成长链思维来实现强大的问题解决能力，但这种行为显著增加了推理成本和延迟。相比之下，经过指令调优的模型往往回答更为简洁，但通常缺乏可比的推理能力。这种准确性与效率的不匹配促使我们提出一种轻量级的方法，结合两种模型的优势，而无需对整个模型进行重新训练。本文提出了GRIP（Granular Reward-guided Interpolation of Parameters），一种用于高效推理的奖励引导参数插值框架。在给定具有相同架构的推理模型和指令模型的情况下，GRIP为各个模块分配可学习的插值比率，并仅优化这些比率，同时保持两个源模型不变。插值比率通过奖励信号进行训练，该信号偏向于既正确又简洁的响应。实验表明，GRIP在准确性与效率的权衡上优于固定或基于搜索的合并基线，并进一步揭示了与高效推理相关的模块级融合模式。
View on arXiv Download PDF AI Translation
cs.CL / 40 / 2608.25593

JIT-Agent: Scaling Harness Intelligence via Just-in-Time Harness Evolution

JIT-Agent：通过即时演化提升智能代理的能力
Zhang, Guibin, Lu, Leo, Xie, Fangzhou, Zhu, Kang, Wang, Junhao, Xie, Zhifei, Yu, Zhaochen, Liu, Zihang, Sun, Zhongxiang, Li, Qiankun, Liao, Yue, Chang, Heng, Hu, Xiaobin, Ren, Qibing, Zhou, Wangchunshu, Yan, Shuicheng
Abstract
Agent capability is not determined by the model alone. The agent harness, encompassing memory management, planning strategy, action protocol, and tool/skill orchestration, can dominate the contribution of the underlying foundation model. Yet harness design remains manual, task-specific, and fundamentally unscalable. We present JIT-Agent, a harness intelligence model trained to synthesize task-adaptive agent harnesses on the fly for arbitrary off-the-shelf agentic LLMs. We formalize the agent harness as a composable, machine-generatable artifact governed by a fixed four-module protocol, and train JIT-Agent to customize harnesses for a given task at hand, repair harnesses for stable and reliable execution, and self-evolve by distilling performance signals from an expanding archive of prior harness configurations. Equipped with JIT-Agent as a harness helper, DeepSeek-V4-Flash surpasses GPT-5.6 on DeepSearchQA (+9.1) and OdysseyBench (+4.3), while the already strong GLM-5.2 gains up to +20.2 points. Across controlled evaluations, JIT-Agent-generated harnesses are performance-competitive with mature agent runtimes such as OpenCode and Claude Code and consistently improve multi-scale model families of DeepSeek V4, Mimo-V2.5, and Qwen3.6. To our knowledge, JIT-Agent is the first model purpose-built for just-in-time harness generation, establishing harness intelligence as a trainable, transferable, and compounding dimension of agent capability orthogonal to model scaling.
Chinese Translation
代理的能力不仅仅由模型决定。代理的工具（harness）包括内存管理、规划策略、行动协议以及工具/技能协调，可能在很大程度上影响基础模型的贡献。然而，工具设计仍然是手动的、特定任务的，并且在本质上无法扩展。我们提出了JIT-Agent，这是一种训练用于即时合成任务自适应代理工具的智能模型，适用于任意现成的代理大型语言模型（LLMs）。我们将代理工具形式化为一个可组合的、机器可生成的工件，由固定的四模块协议控制，并训练JIT-Agent为当前任务定制工具，修复工具以实现稳定可靠的执行，并通过从不断扩展的先前工具配置档案中提取性能信号进行自我演化。在JIT-Agent作为工具助手的支持下，DeepSeek-V4-Flash在DeepSearchQA上超越了GPT-5.6（+9.1）和OdysseyBench（+4.3），而已经表现强劲的GLM-5.2则提升了多达+20.2分。在受控评估中，JIT-Agent生成的工具在性能上与成熟的代理运行时（如OpenCode和Claude Code）具有竞争力，并且持续改善DeepSeek V4、Mimo-V2.5和Qwen3.6的多尺度模型系列。根据我们的了解，JIT-Agent是第一个专门为即时工具生成而构建的模型，确立了工具智能作为一种可训练、可转移和累积的代理能力维度，与模型扩展正交。
View on arXiv Download PDF AI Translation
cs.CL / 41 / 2608.25605

From Specialization to Generalization: Instruction-tuned LLMs for Robust Harmful Content Mitigation

从专业化到泛化：针对有害内容缓解的指令调优大型语言模型
Edman, Lukas, Dementieva, Daryna, Fraser, Alexander
Abstract
Large language models (LLMs) demonstrate impressive performance across a wide range of general NLP tasks; however, their effectiveness in sensitive domains, such as hate speech detection, remains less clear. Prior studies comparing prompted LLMs with state-of-the-art encoder-based models (e.g., BERT variants (Roy et al., 2023; D\"onmez et al., 2024)) have shown only marginal gains, suggesting that LLMs may not excel in hate speech detection or mitigation. In this work, we revisit this question through the lens of instruction tuning. By thoroughly unifying 36 English hate speech datasets spanning multiple labeling schemes, we fine-tune a generalist LLM, based on Qwen3 (Qwen Team, 2025), specifically for hate speech mitigation. Our results demonstrate not only state-of-the-art performance on in-domain benchmarks but also substantial improvements in cross-domain and cross-lingual generalization--areas where encoder-based specialist classifiers often struggle.
Chinese Translation
大型语言模型（LLMs）在广泛的通用自然语言处理任务中表现出色；然而，它们在敏感领域（如仇恨言论检测）中的有效性仍不明确。先前的研究将提示型LLMs与最先进的基于编码器的模型（如BERT变体（Roy et al., 2023; Dönmez et al., 2024））进行比较，仅显示出边际增益，这表明LLMs在仇恨言论检测或缓解方面可能并不出色。在本研究中，我们通过指令调优的视角重新审视这个问题。通过全面统一36个涵盖多种标注方案的英语仇恨言论数据集，我们针对仇恨言论缓解对基于Qwen3（Qwen Team, 2025）的通用LLM进行了微调。我们的结果不仅在领域内基准测试中展示了最先进的性能，还在跨领域和跨语言泛化方面取得了显著改善——这些领域通常是基于编码器的专业分类器所面临的挑战。
View on arXiv Download PDF AI Translation
cs.CL / 42 / 2608.25618

AWM: Answerable Working Memory for Long-Document VQA Agents

AWM：用于长文档视觉问答代理的可回答工作记忆
Zhou, Dongzhuoran, Zhu, Yuqicheng, Liu, Yule, Yang, Zhen, Lu, Rui, Dong, Yuxiao, Tang, Jie, Kharlamov, Evgeny
Abstract
Long-document visual question answering increasingly relies on VLM agents that retrieve candidate pages, inspect page images, write findings to working memory, and synthesize answers. Working memory should carry answer-supporting evidence across page inspections for later grounded answering, yet existing evaluation mainly checks final-answer correctness and evidence-page access. This creates a memory-quality blind spot: an agent may reach the right page and answer correctly while leaving behind memory too generic or incomplete to support answering once page context is removed. We introduce \emph{memory-only answerability}, a diagnostic that asks whether a reader can answer from the question and terminal working memory alone. Building on this diagnostic, \emph{Answerable Working Memory} (AWM) treats terminal working memory as an answerable evidence artifact, and AWM-GRPO incorporates this signal into the GRPO reward while preserving final-answer priority. Under GRPO, this reward assigns higher advantages to answer-correct trajectories whose terminal working memory remains answerable. On \textsc{MMLongBench-Doc}, even when gold evidence pages are provided, 42.5\% of correct answers still cannot be answered from terminal working memory alone. AWM-GRPO improves final-answer accuracy over the RAG baseline by 8.1 and 11.9 points on \textsc{MMLongBench-Doc} and \textsc{LongDocURL} and reduces the memory-missing-correct rate by 2.7 points over answer-only GRPO.
Chinese Translation
长文档视觉问答越来越依赖于检索候选页面、检查页面图像、将发现写入工作记忆并综合答案的视觉语言模型（VLM）代理。工作记忆应在页面检查中携带支持答案的证据，以便在移除页面上下文后进行后续的基于证据的回答。然而，现有的评估主要检查最终答案的正确性和证据页面的访问，这造成了一个记忆质量的盲点：一个代理可能到达正确的页面并给出正确的答案，但留下的记忆却过于笼统或不完整，以至于在移除页面上下文后无法支持回答。我们引入了 extit{仅基于记忆的可回答性}，这是一个诊断指标，用于询问读者是否仅凭问题和终端工作记忆就能回答。基于这一诊断， extit{可回答工作记忆}（AWM）将终端工作记忆视为一个可回答的证据工件，而AWM-GRPO则将这一信号纳入GRPO奖励中，同时保留最终答案的优先级。在GRPO下，该奖励对终端工作记忆仍然可回答的正确答案轨迹赋予更高的优势。在 extsc{MMLongBench-Doc}上，即使提供了黄金证据页面，42.5 ext{%}的正确答案仍无法仅从终端工作记忆中回答。AWM-GRPO在 extsc{MMLongBench-Doc}和 extsc{LongDocURL}上分别比RAG基线提高了8.1和11.9个百分点的最终答案准确性，并将仅基于答案的GRPO的记忆缺失正确率降低了2.7个百分点。
View on arXiv Download PDF AI Translation
cs.CL / 43 / 2608.25637

AutoVerifier: Residual-Guided Non-Parametric Optimization for Reference-Based Answer Verification

AutoVerifier：基于残差引导的非参数优化用于基于参考的答案验证
Zhao, Zebei, Shi, Zhihao, Shi, Minqi
Abstract
Reference-based verifiers are important for evaluating reasoning models and providing accurate outcome rewards in reinforcement learning with verifiable rewards. To improve verification accuracy, prior work has explored rule-based, model-based, and tool-augmented verifiers for checking answer equivalence across diverse answer forms. However, the equivalence of answer forms such as $1+3.14$ and $1+\pi$ may depend on the question and scoring criterion. We frame such implicit assumptions as verifier inductive biases. To address this challenge, we propose AutoVerifier, a residual-guided non-parametric optimization method that learns these biases from recurring verifier errors. Specifically, AutoVerifier records these biases in rule cards and promotes them to code modules or prompt guidance only after replay validation detects no direct regressions, keeping accepted updates auditable, editable, and reusable. Experiments on four verifier benchmarks demonstrate that AutoVerifier outperforms state-of-the-art verifiers by a large margin.
Chinese Translation
基于参考的验证器对于评估推理模型和在具有可验证奖励的强化学习中提供准确的结果奖励至关重要。为了提高验证准确性，先前的研究探索了基于规则、基于模型和工具增强的验证器，以检查不同答案形式之间的等价性。然而，答案形式如 $1+3.14$ 和 $1+ ext{π}$ 的等价性可能依赖于问题和评分标准。我们将这种隐含假设框架视为验证器的归纳偏差。为了解决这一挑战，我们提出了 AutoVerifier，一种基于残差引导的非参数优化方法，能够从重复的验证器错误中学习这些偏差。具体而言，AutoVerifier 在规则卡中记录这些偏差，并仅在重放验证未检测到直接回归后，将其提升为代码模块或提示指导，从而保持已接受更新的可审计性、可编辑性和可重用性。在四个验证器基准上的实验表明，AutoVerifier 的表现显著优于现有的最先进验证器。
View on arXiv Download PDF AI Translation
cs.CL / 44 / 2608.25654

Unmatched Does Not Mean False: Incomplete Reference Sets Can Reverse Calibration Rankings in Open-Ended Theory-of-Mind Tracking

不匹配并不意味着错误：不完整参考集可以逆转开放式心智理论追踪中的校准排名
Feng, Zhexi, Chen, Wuxi, Zhang, Bingrui
Abstract
Open-ended Theory-of-Mind (ToM) trackers emit valid beliefs absent from finite references. A finite-reference-plus-matcher pipeline marks unmatched outputs false, creating proxy labels that can reverse proper-score model selection on fixed outputs. Holding 259 beliefs and paired scores fixed, reference recoding lowers weighted prevalence from 0.783 to 0.295 and reverses strictly proper Brier risk: a frozen source-prior rule leads native confidence by 0.227 under reference labels and trails by 0.152 under blinded adjudication, in all six authored scenarios. A reference-only Platt recalibrator reverses further. An ICE-specific reversal appears in a released 301-question NQ-open DPR-BERT pipeline: its average-confidence baseline improves instance-level calibration error by 0.045 under exact match but worsens it by 0.074 under human correctness, with both intervals excluding zero. On independently authored OpenToM narratives, 90-96% of audited unmatched beliefs are literally true and the paired direction again reverses. An exact decomposition attributes the distortion to omitted truths, and a closed-form criterion correctly classifies comparisons from twelve released systems. Frozen-audit retrospective replay shows 50 attempted annotations recover ranking direction with probability at least 0.996. TriSource-Restore anchors full-frame reference labels and frozen automatic judgments to a probability-sampled human pilot, maintains at least nominal coverage, narrows intervals, and repairs confidence subject to a base-rate deployment gate.
Chinese Translation
开放式心智理论（ToM）追踪器在缺乏有限参考的情况下发出有效信念。有限参考加匹配器管道将不匹配的输出标记为错误，从而创建代理标签，这可能会逆转固定输出上的适当得分模型选择。在保持259个信念和配对得分固定的情况下，参考重编码将加权流行度从0.783降低至0.295，并逆转严格适当的Brier风险：在参考标签下，冻结的源先验规则使本地信心领先0.227，而在盲审下则落后0.152，在所有六个作者场景中均如此。仅参考的Platt重校准器进一步逆转。在发布的301问题NQ-open DPR-BERT管道中出现了特定于ICE的逆转：其平均信心基线在精确匹配下改善了实例级校准误差0.045，但在人工正确性下恶化了0.074，且两个区间均不包括零。在独立创作的OpenToM叙事中，90-96%的审计不匹配信念在字面上是真实的，配对方向再次逆转。精确分解将失真归因于遗漏的真相，而封闭形式标准正确地对比了来自十二个发布系统的比较。冻结审计的回顾重放显示，50个尝试的注释以至少0.996的概率恢复排名方向。TriSource-Restore将全框架参考标签和冻结的自动判断锚定到概率抽样的人类先导，保持至少名义覆盖，缩小区间，并修复信心，受基础率部署门的限制。
View on arXiv Download PDF AI Translation
cs.CL / 45 / 2608.25655

Reconstructing the Right Episode: Evaluating Interleaved Conversational Memory Beyond Long Context

重构正确的情节：评估超长上下文下交错对话记忆
Feng, Zhexi, Zhang, Ruiyi, Yang, Yongbo, Xie, Pengtao
Abstract
Conversations with chat assistants increasingly span many topics in a single long-running thread, challenging memory systems. Existing long-context and memory benchmarks often expose session or topic boundaries, or probe direct personal-memory questions. These settings understate a harder assistant-memory regime: a flat mixed-topic thread where the system must infer which earlier episode makes a later task decision valid. We introduce SCALE-QA, a constraint-grounded task QA benchmark for flat unsegmented threads targeting episode integrity failure. The dataset contains 3,000 audited questions across 10 domains, uses deterministic four-way multiple-choice grading, and includes a deterministic runtime builder; experiments use all 3,000 questions through 128k and a stratified 400-question diagnostic at 1M. SCALE-QA questions are ordinary task-oriented requests whose correct answer depends on causally related evidence introduced earlier in the conversation. We also propose Temporal-Semantic Interleaved Memory Reconstruction (TSIM), which segments the turn stream into coherent episodes and indexes them through a hierarchical multi-view memory stack with deterministic episode-level summary and cluster-routing views. Experiments show that SCALE-QA challenges strong RAG baselines and long-context LLMs alike; across three open-source and proprietary LLM backends, TSIM achieves the highest accuracy in every backend setting, gaining 5.6-17.6 accuracy points over the strongest corresponding baseline.
Chinese Translation
与聊天助手的对话越来越多地跨越多个主题，形成单一的长时间线程，这对记忆系统提出了挑战。现有的长上下文和记忆基准通常暴露会话或主题边界，或探测直接的个人记忆问题。这些设置低估了更复杂的助手记忆机制：一个扁平的混合主题线程，其中系统必须推断出哪个早期情节使得后续任务决策有效。我们引入了SCALE-QA，这是一个基于约束的任务问答基准，针对情节完整性失败的扁平未分段线程。该数据集包含3,000个经过审核的问题，涵盖10个领域，采用确定性的四选一评分，并包括一个确定性的运行时构建器；实验使用所有3,000个问题，通过128k和分层的400个问题诊断在1M下进行。SCALE-QA的问题是普通的任务导向请求，其正确答案依赖于在对话中早期引入的因果相关证据。我们还提出了时间语义交错记忆重构（Temporal-Semantic Interleaved Memory Reconstruction, TSIM），该方法将轮次流分割为连贯的情节，并通过具有确定性情节级摘要和聚类路由视图的层次多视图记忆堆栈进行索引。实验表明，SCALE-QA对强大的RAG基线和长上下文LLM都构成了挑战；在三个开源和专有的LLM后端中，TSIM在每个后端设置中都达到了最高的准确率，相较于最强的相应基线提高了5.6-17.6个准确率点。
View on arXiv Download PDF AI Translation
cs.CL / 46 / 2608.25660

Think-Probe-Respond: Improving Large Language Models as Judges of Research Idea Novelty

思考-探测-回应：提升大型语言模型作为研究创意新颖性评判者的能力
Schopf, Tim, Schreieder, Tobias, Aizawa, Akiko
Abstract
Automated novelty judgment can accelerate scientific discovery by enabling efficient evaluation, refinement, and comparison of research ideas. While large language models are increasingly adopted for this task, we investigate a previously overlooked limitation in their judgment capabilities: despite generating reasoning rationales that closely mirror those of human experts, their final novelty judgments often diverge substantially. We demonstrate that this miscalibration stems from a systematic bias towards judging ideas as "medium novel". To mitigate this, we propose Think-Probe-Respond (TPR), a lightweight approach that probes latent novelty judgments from hidden states during the reasoning phase and uses the probed judgments to condition the final response. Across strong baselines, TPR improves novelty judgment performance by 22.30% and successfully mitigates the prevalent "medium novelty" bias.
Chinese Translation
自动化的新颖性判断能够通过高效评估、完善和比较研究创意来加速科学发现。尽管大型语言模型在这一任务中被越来越广泛地应用，我们却发现其判断能力中一个被忽视的局限性：尽管生成的推理理由与人类专家的相似，但它们的最终新颖性判断往往存在显著差异。我们证明这种误校准源于对创意的系统性偏见，倾向于将创意判断为“中等新颖”。为此，我们提出了思考-探测-回应（Think-Probe-Respond, TPR）方法，这是一种轻量级的方法，通过在推理阶段从隐藏状态中探测潜在的新颖性判断，并利用探测到的判断来调整最终回应。在强基线的对比中，TPR将新颖性判断性能提升了22.30%，并成功减轻了普遍存在的“中等新颖”偏见。
View on arXiv Download PDF AI Translation
cs.CL / 47 / 2608.25662

Overview of SHROOM-Visions 2026: A Shared Task on Hallucination Detection in Large Vision-Language Models

SHROOM-Visions 2026概述：大型视觉-语言模型中的幻觉检测共享任务
Vázquez, Raúl, Sinha, Aman, Li, Chuyuan, Savelli, Claudio, Calò, Eduardo, Raimond, Emilio, Frank, Stella, Luo, Hengyu, Giobergia, Flavio, Segonne, Vincent, Vaiani, Lorenzo, Tiedemann, Jörg, Mickus, Timothee
Abstract
In 2026, we held the fourth iteration of the SHROOM Shared Task series: SHROOM-Visions (\textbf{S}hared-task on \textbf{H}allucinations and \textbf{R}elated \textbf{O}bservable \textbf{O}vergeneration \textbf{M}istakes in \textbf{Vision} language model\textbf{s}), which is hosted at the UncertaiNLP Workshop co-located with EMNLP 2026. Following the success of the 2024 and 2025 tasks, this time we aim to tackle hallucinations through a model-agnostic detection task focused on large vision-language models. Building on the recently introduced SHEEP dataset, designed for long-term evaluation across model generations, the task invites participants to detect and classify fine-grained hallucination spans in image-conditioned text generation (VQA, image captioning, etc.). The evaluation uses a five-class taxonomy of hallucinations spanning four languages: Chinese, English, French, and Italian. The shared task generated strong interest in the NLP community worldwide, with 27 teams contributing 600+ system submissions. The best systems achieve average scores of 0.58 in character-level correlation, 0.46 in label-conditioned correlation, and 0.51 in intersection-over-union (IoU) across four languages, outperforming the baselines by 30-40 points.
Chinese Translation
在2026年，我们举办了SHROOM共享任务系列的第四次迭代：SHROOM-Visions（ extbf{S}hared-task on extbf{H}allucinations and extbf{R}elated extbf{O}bservable extbf{O}vergeneration extbf{M}istakes in extbf{Vision} language model extbf{s}），该活动在与EMNLP 2026共同举办的UncertaiNLP研讨会上进行。继2024年和2025年任务的成功之后，这一次我们旨在通过一个模型无关的检测任务来解决大型视觉-语言模型中的幻觉问题。基于最近推出的SHEEP数据集，该数据集旨在跨模型代进行长期评估，此次任务邀请参与者检测和分类图像条件下文本生成中的细粒度幻觉范围（如视觉问答、图像描述等）。评估使用涵盖中文、英文、法文和意大利文四种语言的五类幻觉分类体系。该共享任务在全球NLP社区引起了强烈的兴趣，共有27个团队提交了600多个系统提交。最佳系统在字符级相关性上平均得分为0.58，在标签条件相关性上为0.46，在交并比（IoU）上为0.51，均超过基线30-40分。
View on arXiv Download PDF AI Translation
cs.CL / 48 / 2608.25677

Learning New Facts with QLoRA: An Acquisition-Retention Frontier

通过QLoRA学习新事实：获取-保留前沿
Zheng, Estelle, Warichet, Sébastien, Helbert, Emmanuel, Cerisara, Christophe
Abstract
Parameter-efficient fine-tuning is often assumed to preserve pretrained capabilities because it updates only a small number of parameters. We show that this assumption depends strongly on adapter capacity. We study factual acquisition in a controlled OpenStreetMap-derived benchmark where Qwen3-4B must acquire anonymized geographic associations while retaining unrelated capabilities. Comparing full fine-tuning (FFT) with quantized low-rank adaptation (QLoRA) at ranks 8, 16, 32, and 64, we find that rank induces a clear acquisition--retention frontier. Low-rank QLoRA preserves out-of-domain (OOD) performance but acquires fewer facts, whereas higher ranks improve same-fact paraphrase generalization at an increasing cost in performance on unrelated benchmarks. FFT behaves as a conservative baseline: it retains general capabilities well, but does not reach the highest factual-acquisition regime. Distributional, weight-space, and spectral diagnostics mirror this behavioral trade-off, with higher-rank QLoRA moving farther from the pretrained model. A separate math adaptation experiment shows a weaker frontier, suggesting that the effect is most pronounced when adaptation must install new factual associations rather than reinforce skills already supported by pretraining. Code and data are available at https://github.com/zhngstl/new_facts_forgetting.
Chinese Translation
参数高效的微调通常被认为能够保留预训练的能力，因为它仅更新少量参数。我们表明这一假设在很大程度上依赖于适配器的容量。我们在一个受控的基于OpenStreetMap的基准测试中研究事实获取，其中Qwen3-4B必须获取匿名的地理关联，同时保留无关的能力。通过比较全量微调（FFT）与在秩为8、16、32和64的量化低秩适配（QLoRA），我们发现秩引入了明显的获取-保留前沿。低秩QLoRA保留了域外（OOD）性能，但获取的事实较少，而更高的秩则在获取同一事实的释义泛化方面有所改善，但在无关基准上的性能代价也随之增加。FFT表现为一个保守的基线：它能够很好地保留一般能力，但未能达到最高的事实获取水平。分布、权重空间和谱诊断反映了这种行为权衡，更高秩的QLoRA与预训练模型的距离更远。一个单独的数学适配实验显示出较弱的前沿，表明这一效应在适配必须安装新的事实关联而非强化预训练已支持的技能时最为明显。代码和数据可在 https://github.com/zhngstl/new_facts_forgetting 获取。
View on arXiv Download PDF AI Translation
cs.CL / 49 / 2608.25717

When RAG Fails to Equalize: Geo-bias in Factual Question Answering over Public Companies

当 RAG 无法实现均衡：公共公司事实问答中的地理偏见
Havaldar, Abhinav, Santus, Enrico
Abstract
Retrieval-augmented generation (RAG) is widely assumed to mitigate factual errors in large language models (LLMs), but it remains unclear whether retrieval uniformly compensates for missing knowledge. We study this question in a controlled factual QA setting over public companies, constructing a benchmark of approximately 2,000 firms across global equity indices. We evaluate six LLMs on four atomic attributes under four conditions: no-context, perfect context, misleading context, and distraction context. We find strong geographic disparities in no-context accuracy, indicating uneven parametric knowledge. While perfect context improves performance, it does not eliminate these gaps: gains are correlated with baseline accuracy, suggesting retrieval effectiveness is coupled to internal representations. Under misleading context, models frequently copy incorrect information. Larger models improve overall performance but do not remove these structural effects. These results challenge the view of RAG as a universal corrective and highlight the interaction between model knowledge, context quality, and entity representation.
Chinese Translation
检索增强生成（RAG）被广泛认为可以减轻大型语言模型（LLMs）中的事实错误，但尚不清楚检索是否能均匀补偿缺失的知识。我们在一个针对公共公司的受控事实问答环境中研究了这个问题，构建了一个涵盖全球股票指数的约2000家公司的基准。我们在四种条件下评估了六个LLMs在四个原子属性上的表现：无上下文、完美上下文、误导性上下文和干扰上下文。我们发现无上下文准确性存在显著的地理差异，表明参数知识的不均匀性。尽管完美上下文提高了性能，但并未消除这些差距：性能提升与基线准确性相关，表明检索的有效性与内部表征相结合。在误导性上下文下，模型经常复制不正确的信息。更大的模型提高了整体性能，但并未消除这些结构性影响。这些结果挑战了将 RAG 视为普遍修正工具的观点，并突显了模型知识、上下文质量和实体表征之间的相互作用。
View on arXiv Download PDF AI Translation
cs.CL / 50 / 2608.25761

Beam Search, Self-Consistency, and the Limits of Inference-Time Scaling for Grammar-Constrained Text-to-SQL in Small Language Models

束搜索、自我一致性与小型语言模型中受限语法的文本到SQL推理时间扩展的限制
Chermsirivatana, Ty, MacCormick, John
Abstract
One common trade-off in the use of large language models involves reducing the size of the model while increasing the amount of computation at inference time, for example by using a wider beam search. In this paper, we examine the constrained case of this "model size vs. inference compute" trade-off, in which the model outputs are constrained by a strict grammar at inference time. Our results demonstrate that the constrained trade-off behaves differently from the unconstrained trade-off. We investigate the task of converting a prose query into an equivalent SQL query (text-to-SQL). Performance is evaluated on the Spider text-to-SQL benchmark, using the Qwen2.5-Instruct model family ranging in size from 0.5B to 7B parameters, all at 4-bit precision. We experiment with two approaches to varying inference compute: (i) beam search with a variable number of beams; and (ii) sample+vote, i.e., sampling several constrained outputs and then voting on their execution results, where the number of samples is varied. On the 1034-example development set, we find that: (a) both beam search and sample+vote improve accuracy, especially on smaller model sizes; (b) the "model size vs.\ inference compute" trade-off is not advantageous in this experiment, because moving to a larger model size typically results in higher accuracy than increasing inference compute on the same model size; (c) beam search outperforms sample+vote at a matched inference budget. This latter result is of particular interest since it contrasts with the findings of the unconstrained trade-off.
Chinese Translation
在使用大型语言模型时，一个常见的权衡是减少模型的大小，同时增加推理时的计算量，例如通过使用更宽的束搜索。本文研究了这一“模型大小与推理计算”权衡的受限情况，其中模型输出在推理时受到严格语法的限制。我们的结果表明，受限权衡的表现与不受限权衡有所不同。我们探讨了将自然语言查询转换为等效SQL查询（文本到SQL）的任务。性能评估基于Spider文本到SQL基准，使用Qwen2.5-Instruct模型系列，参数规模从0.5B到7B，均为4位精度。我们实验了两种变化推理计算的方法：（i）具有可变束数的束搜索；（ii）采样+投票，即采样多个受限输出，然后对其执行结果进行投票，样本数量可变。在1034个示例的开发集上，我们发现：（a）束搜索和采样+投票均提高了准确性，尤其是在较小的模型规模上；（b）在本实验中，“模型大小与推理计算”的权衡并不具有优势，因为转向更大的模型规模通常会导致比在相同模型规模上增加推理计算更高的准确性；（c）在匹配的推理预算下，束搜索的表现优于采样+投票。这个结果尤其引人关注，因为它与不受限权衡的发现形成了对比。
View on arXiv Download PDF AI Translation
cs.CL / 51 / 2608.25768

MoganBert-TR: A Turkish Encoder Foundation Model Trained from Scratch with a CLM-to-MLM Curriculum

MoganBert-TR：一种从零开始训练的土耳其编码器基础模型，采用CLM到MLM的课程学习
Yilmaz, Furkan, Tasdemir, Habibe Aleyna, Gozay, Muhammed Faruk
Abstract
Turkish encoder models have adopted modern architectures while leaving the pretraining objective fixed at masked language modelling. This paper introduces MoganBert-TR, a 149M-parameter Turkish encoder foundation model trained from scratch on a language-specifically filtered corpus, together with an embedding model derived from it (MoganBert-Embed). MoganBert-TR is trained over 237.3B tokens with a two-stage CLM-to-MLM curriculum: causal language modelling first, masked language modelling for the remainder, with the transition made inside the stable phase of a WSD schedule. In a controlled ablation under an equal step budget, this design outperforms pure MLM by 2.7-3.7x on Turkish MS MARCO retrieval; the measured mechanism is embedding geometry, where a single direction absorbs 28.1% of the variance under pure MLM against 11.9% under the curriculum. Long-context extension and learning-rate decay are then split into two branches after a shared prefix: running the final portion of decay at 1024 context improves the TrGLUE average by 0.49 +/- 0.26 points across five paired seeds (p = 0.013) and beats a model-soup alternative by 0.75 points at ~4.3% additional cost. MoganBert-TR attains 78.41 on TrGLUE, the best among the Turkish ModernBERT models compared, and 77.73 on TabiBench, where it leads two of the eight categories with the largest margin on code retrieval (+3.62 points over TabiBERT). MoganBert-Embed, produced through teacher distillation and multi-signal contrastive fine-tuning, ranks first among student models on the MTEB(Turkish) overall average with 68.30 and reaches 99.5% of its 7.57B-parameter teacher's score with a 51x smaller backbone. The accompanying 50,048-token tokenizer outperforms all compared Turkish tokenizers on compression and fertility across two independent test sets. Weights, tokenizer, embedding model and evaluation code: https://huggingface.co/moganai
Chinese Translation
土耳其编码器模型采用了现代架构，但在预训练目标上仍然固定为掩码语言建模（MLM）。本文介绍了MoganBert-TR，这是一个149M参数的土耳其编码器基础模型，基于特定语言过滤的语料库从零开始训练，同时还推出了一个衍生的嵌入模型（MoganBert-Embed）。MoganBert-TR在237.3B个标记上进行训练，采用了两阶段的CLM到MLM课程：首先进行因果语言建模（CLM），然后进行掩码语言建模（MLM），在WSD调度的稳定阶段内进行过渡。在相同步长预算下的受控消融实验中，该设计在土耳其MS MARCO检索上比纯MLM表现优越2.7-3.7倍；测量机制为嵌入几何，其中在纯MLM下单一方向吸收28.1%的方差，而在课程学习下仅为11.9%。长上下文扩展和学习率衰减在共享前缀后分为两个分支：在1024上下文下运行最终的衰减部分使TrGLUE平均分提高了0.49 +/- 0.26分（p = 0.013），并在约4.3%的额外成本下超越了模型组合替代方案0.75分。MoganBert-TR在TrGLUE上获得78.41，是比较中土耳其ModernBERT模型中表现最佳的，并在TabiBench上获得77.73，在代码检索上以+3.62分的优势领先于TabiBERT，位列八个类别中的两个。通过教师蒸馏和多信号对比微调生成的MoganBert-Embed，在MTEB（土耳其）整体平均中在学生模型中排名第一，得分为68.30，达到了其7.57B参数教师模型得分的99.5%，而其骨干网络则小51倍。附带的50,048标记的分词器在两个独立测试集上在压缩和丰富性方面超越了所有比较的土耳其分词器。权重、分词器、嵌入模型和评估代码： https://huggingface.co/moganai
View on arXiv Download PDF AI Translation
cs.CL / 52 / 2608.25824

Localize-Then-Decide Guarantees for LLM Judgments

局部化-再决策的LLM判断保证
Li, Xinyu, Zhou, Yi, Cao, Guanqun, Fu, Zeyu, Huang, Tianjin, Jin, Gaojie
Abstract
Large language models (LLMs) are increasingly used as evaluators to assess output quality and preference alignment, yet providing reliable guarantees of agreement with human judgments remains challenging. Recent work introduces confidence-thresholding methods that provide such guarantees for pairwise comparisons, relying on the assumption that higher estimated confidence implies lower disagreement risk with humans. However, this assumption can break down when the number of candidate responses increases, since distributing probability mass across many alternatives can distort confidence estimates. To address this issue, we propose a Localize-Then-Decide framework. First, conformal prediction localizes a small shortlist that contains the human-preferred response with high probability. Then, a calibrated confidence-based rule selectively chooses a single response from this shortlist or abstains. This design restores the monotonic relationship between confidence and disagreement risk and enables high-probability agreement guarantees. Experiments with multiple candidate sizes across several datasets and judge LLMs demonstrate that our framework consistently achieves higher guarantee success rates and substantially higher coverage than single-stage baselines.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作评估者，以评估输出质量和偏好一致性，但提供与人类判断一致的可靠保证仍然具有挑战性。近期的研究引入了置信度阈值方法，为成对比较提供了这样的保证，依赖于更高的置信度估计意味着与人类的低不一致风险的假设。然而，当候选响应数量增加时，这一假设可能会失效，因为在多个备选项之间分配概率质量可能会扭曲置信度估计。为了解决这个问题，我们提出了一个局部化-再决策框架。首先，符合预测（conformal prediction）定位一个小的短名单，其中高概率包含人类偏好的响应。然后，基于校准的置信度规则从这个短名单中选择一个单一响应或放弃。这一设计恢复了置信度与不一致风险之间的单调关系，并实现了高概率的一致性保证。在多个数据集和评估LLM的实验中，我们的框架在不同候选大小下始终实现了更高的保证成功率和显著更高的覆盖率，相较于单阶段基线。
View on arXiv Download PDF AI Translation
cs.CL / 53 / 2608.25826

Unfolding Scientific Papers into Multi-Turn Generation Trajectories for Continued Pre-Training

将科学论文展开为多轮生成轨迹以进行持续预训练
Xu, Qiankai, Chen, Qiguang, Su, Zixin, Huang, Wenhao, Gao, Yue, Liu, Jiaheng, Zhang, Ge
Abstract
A recent line of synthetic-data work reconstructs the thinking behind existing text rather than rewriting the text itself, but it operates on short web passages, recovers only local thoughts, and leaves the structure of whole documents untouched. Scientific papers are written to a clear and largely uniform structure and make a natural substrate for lifting this paradigm to the document level. We present a pipeline that unfolds each paper into a multi-turn generation trajectory in which a teacher model reconstructs the writing process of the whole paper: a writing request, a global plan, and pre-writing deliberation for each section. All section texts and the abstract are kept verbatim from the source paper. We apply the pipeline to quality-filtered arXiv papers and obtain a corpus for continued pre-training (CPT) that is roughly twice the size of the source text. The same reverse construction extends to instruction data and evaluation. Treating real paper text as the answer yields an SFT dataset. Anchoring tasks in held-out papers yields PAW-Bench, an academic-writing benchmark whose tasks carry their own rubrics and checklists. In controlled experiments CPT on our corpus followed by supervised fine-tuning on public datasets improves writing benchmarks broadly while preserving general reasoning and improving long-document reading. The writing gain persists even when every model is fine-tuned on a dedicated writing SFT dataset. Mixing our SFT data into that recipe lifts academic writing further.
Chinese Translation
近期的一系列合成数据研究重构了现有文本背后的思维，而不是简单地重写文本本身，但其仅针对短小的网络段落，恢复的仅是局部思路，并未触及整个文档的结构。科学论文遵循明确且基本统一的结构，成为将这一范式提升至文档层面的自然基础。我们提出了一种流程，将每篇论文展开为多轮生成轨迹，其中教师模型重构整篇论文的写作过程：写作请求、全局计划以及每个部分的预写思考。所有部分文本和摘要均逐字保留自源论文。我们将该流程应用于经过质量筛选的arXiv论文，获得了一个用于持续预训练（CPT）的语料库，其规模大约是源文本的两倍。相同的逆向构建也适用于指令数据和评估。将真实论文文本视为答案生成了一个SFT数据集。将任务锚定在保留的论文中生成了PAW-Bench，这是一个学术写作基准，其任务具有各自的评分标准和检查表。在受控实验中，在我们的语料库上进行CPT，随后在公共数据集上进行监督微调，广泛提高了写作基准，同时保持了通用推理能力并改善了长文档阅读。即使每个模型都在专门的写作SFT数据集上进行微调，写作提升依然存在。将我们的SFT数据混入这一配方中进一步提升了学术写作。
View on arXiv Download PDF AI Translation
cs.CL / 54 / 2608.25832

Skill Issue: Are Skills Language-Invariant in LLMs?

技能问题：技能在大型语言模型中是语言不变的吗？
Cheng, Bobby, Gaber, Adam, Liu, Zhengyuan, Arnett, Catherine, Goldman, Omer, Tan, Cheston, Choshen, Leshem
Abstract
Large language models access knowledge inconsistently across languages, but to what extent do they differ in their skill sets when interacting with different languages? This work quantifies cross-lingual skill inconsistency orthogonally from knowledge and general benchmark performance. We do this via multilingual self-play: two instances of the same model compete in a text-based game, each interacting through a different language interface. Since the model, opponent, rules, state space, and available actions remain fixed, this setting isolates the effect of language on the model's realized behavior. We build a multilingual extension to TextArena and evaluate three open-weight models across eight languages and six games covering spatial reasoning, imperfect information, resource allocation, and repeated interaction. We find that the same model can exhibit markedly different playing strength across languages, with systematic variation in win--loss margins, invalid actions, and strategic tendencies. Detailed analyses reveal language-specific failures in spatial reasoning, card-conditioned decisions, and optimal move selection. In some settings, changing only the intermediate reasoning language recovers much of the lost performance, suggesting that language can affect different stages of the decision process. These results show that skill discrepancies are a measurable major roadblock in the development of truly multilingual models. Better understanding these discrepancies can help us design models that perform more equitably across languages.
Chinese Translation
大型语言模型在不同语言间获取知识的方式不一致，但它们在与不同语言互动时的技能组合有多大差异？本研究从知识和一般基准性能的正交角度量化跨语言技能不一致性。我们通过多语言自我对弈来实现这一目标：同一模型的两个实例在一个基于文本的游戏中竞争，各自通过不同的语言接口进行互动。由于模型、对手、规则、状态空间和可用动作保持不变，这一设置隔离了语言对模型实际行为的影响。我们构建了 TextArena 的多语言扩展，并在八种语言和六个游戏中评估了三种开放权重模型，这些游戏涵盖空间推理、不完全信息、资源分配和重复互动。我们发现同一模型在不同语言中表现出明显不同的游戏实力，胜负差距、无效动作和战略倾向存在系统性变化。详细分析揭示了在空间推理、基于卡片的决策和最佳移动选择中存在语言特定的失败。在某些情况下，仅改变中间推理语言就能恢复大部分丧失的性能，表明语言可能影响决策过程的不同阶段。这些结果表明，技能差异是开发真正多语言模型的一个可测量的主要障碍。更好地理解这些差异可以帮助我们设计在不同语言中表现更为公平的模型。
View on arXiv Download PDF AI Translation
cs.CL / 55 / 2608.25854

Key Point Analysis Needs Structure Recovery: Task Definition, Dataset Diagnosis, and a Structure-Aware Benchmark

关键点分析需要结构恢复：任务定义、数据集诊断与结构感知基准
Shi, Zhiqiang, Cocarascu, Oana
Abstract
Key Point Analysis (KPA) aims to identify a concise set of key points that summarize a collection of arguments together with their prevalence. We argue that KPA is fundamentally a structured prediction problem that requires recovering semantic groupings, generating representative key points, ensuring coverage, and estimating prevalence. Under this formulation, we show that existing KPA benchmarks suffer from limitations in grouping quality, redundancy, coverage, and argument-key point mappings, causing ceiling violation and selection failure in reference-based evaluation. To support future research on true KPA, we introduce a structure-aware, distribution-sensitive benchmark built via a human-in-the-loop re-annotation. Human and LLM evaluations consistently show that the resulting structures yield more coherent groupings, higher-quality key points, better coverage, and more reliable prevalence estimates than existing annotations. We further release several annotation resources to support research on KPA evaluation, argument-key point matching, explainable KPA, and LLM-as-a-judge methodologies, and outline a research agenda for true KPA.
Chinese Translation
关键点分析（Key Point Analysis, KPA）旨在识别一组简明的关键点，以总结一系列论点及其流行程度。我们认为，KPA 本质上是一个结构化预测问题，需要恢复语义分组、生成代表性关键点、确保覆盖率并估计流行程度。在这种表述下，我们展示了现有的 KPA 基准在分组质量、冗余性、覆盖率和论点-关键点映射方面存在的局限性，这导致了基于参考的评估中出现上限违背和选择失败。为了支持未来对真实 KPA 的研究，我们引入了一个通过人机协作重新标注构建的结构感知、分布敏感的基准。人类和大型语言模型（LLM）的评估一致表明，所得到的结构比现有标注产生了更连贯的分组、更高质量的关键点、更好的覆盖率和更可靠的流行程度估计。我们还发布了多个标注资源，以支持 KPA 评估、论点-关键点匹配、可解释 KPA 和 LLM 作为评判者的方法论的研究，并概述了真实 KPA 的研究议程。
View on arXiv Download PDF AI Translation
cs.CL / 56 / 2608.25869

Anchoring Bias in LLM-as-a-Judge Systems: Prior Scores Compromise Evaluation Independence

大型语言模型作为评判系统中的锚定偏差：先前评分妨碍评估独立性
Kapetanovic, Ante, Altwlkany, Kemal, Mercep, Andro, Duricic, Tomislav, Lacic, Emanuel
Abstract
Large language models (LLMs) increasingly assess generated content, giving rise to the LLM-as-a-Judge paradigm. These systems now score outputs, filter content, and gate iterative refinement in production pipelines, where each judgment is often assumed to be independent of earlier evaluations. We test this assumption using three prompt conditions: no metadata, revision framing, and anchored metadata containing revision, attempt, and prior-score fields. We show that prior scores, even when included only as context metadata, anchor judgments and systematically shift ratings toward their values. Across 192,000 attempted evaluations (185,271 successful), seven out of the eight evaluated models have 95% task-stratified bootstrap intervals below zero for the total anchored-metadata effect on 20 fixed texts. Cohen's $d$, a standardized measure of the difference between score distributions, reaches an absolute value of 0.71. Token-level analysis of selected model-task probes suggests a threshold-like response pattern: introducing anchored metadata produces a marked redistribution of output-score probabilities, while changing the anchor value within the tested below-threshold range produces comparatively little additional variation. On categorical industry data with human-labeled ground truth, anchored metadata blocks 48% of error corrections and flips 10.18% of correct judgments toward an assigned wrong label, demonstrating the bias extends beyond numerical scoring to categorical decisions. Neither Chain-of-Thought nor a metadata-disregard warning reduces the total effect, although the warning improves the paired accuracy effect relative to baseline in the industry experiment. Reliable LLM evaluation demands careful context engineering rather than an assumption of impartiality. Effective mitigation must be validated for the intended model and task or domain.
Chinese Translation
大型语言模型（LLMs）越来越多地用于评估生成内容，形成了LLM作为评判者的范式。这些系统现在对输出进行评分、过滤内容，并在生产流程中控制迭代优化，其中每个判断通常被假设为独立于早期评估。我们使用三种提示条件来检验这一假设：无元数据、修订框架和包含修订、尝试和先前评分字段的锚定元数据。我们表明，即使先前评分仅作为上下文元数据包含，也会锚定判断并系统性地将评分向其值偏移。在192,000次尝试评估（185,271次成功）中，八个评估模型中有七个在20个固定文本上，其总锚定元数据效应的95%任务分层自助区间低于零。Cohen's $d$，作为评分分布差异的标准化度量，达到绝对值0.71。对选定模型任务探针的逐词分析表明了一种阈值响应模式：引入锚定元数据显著重新分配了输出评分概率，而在测试的低于阈值范围内改变锚定值则产生相对较小的额外变化。在具有人工标注真实值的分类行业数据上，锚定元数据阻止了48%的错误修正，并将10.18%的正确判断翻转为分配的错误标签，表明这种偏差超出了数值评分，影响了分类决策。无论是思维链（Chain-of-Thought）还是忽略元数据的警告都未能减少总效应，尽管该警告在行业实验中相对于基线改善了配对准确度效应。可靠的LLM评估需要仔细的上下文工程，而不是对公正性的假设。有效的缓解措施必须针对特定模型和任务或领域进行验证。
View on arXiv Download PDF AI Translation
cs.CL / 57 / 2608.25881

Loss-Based Active Learning for Neural Abstractive Summarization

基于损失的神经抽象摘要主动学习
Ioannou, Michail, Passali, Tatiana, Michalopoulos, George, Tsoumakas, Grigorios
Abstract
Fine-tuning abstractive summarization models requires high-quality annotated data. However, obtaining such corpora is expensive and time-consuming, as it requires human annotators to read and comprehend long documents to create accurate summaries. Active learning mitigates this issue by selecting only the most informative instances for annotation, allowing models to achieve competitive results with significantly fewer labels. However, the application of active learning to summarization remains under-explored, and existing studies often suffer from instability and significant computational bottlenecks. To overcome these challenges, we propose LOBSTER (LOss-BaSed acTivE leaRning), a novel active learning framework designed specifically for abstractive summarization. LOBSTER improves performance by prioritizing unlabeled instances semantically similar to the model's current high-loss training examples, enabling the model to explicitly correct its specific weaknesses. Our empirical evaluation across three benchmark datasets and two summarization backbone models demonstrates that LOBSTER consistently matches or outperforms current state-of-the-art approaches while achieving a query selection speedup of up to 665x.
Chinese Translation
微调抽象摘要模型需要高质量的标注数据。然而，获取这样的语料库既昂贵又耗时，因为这需要人工标注者阅读和理解长篇文档，以创建准确的摘要。主动学习通过仅选择最具信息量的实例进行标注，从而缓解了这一问题，使得模型能够在显著减少标签数量的情况下实现竞争性的结果。然而，主动学习在摘要生成中的应用仍然未被充分探索，现有研究往往面临不稳定性和显著的计算瓶颈。为了解决这些挑战，我们提出了LOBSTER（LOss-BaSed acTivE leaRning），这是一个专门为抽象摘要设计的新型主动学习框架。LOBSTER通过优先选择与模型当前高损失训练示例语义相似的未标记实例来提高性能，从而使模型能够明确纠正其特定的弱点。我们在三个基准数据集和两个摘要主干模型上的实证评估表明，LOBSTER始终能够与当前最先进的方法相匹配或超越，同时实现高达665倍的查询选择加速。
View on arXiv Download PDF AI Translation
cs.CL / 58 / 2608.25894

From Passive Response to Proactive Correction: Enhancing LLM Robustness Against Input Fact Perturbations

从被动响应到主动纠正：增强大语言模型对输入事实扰动的鲁棒性
Wang, Ping, Sun, Xiangguo, Xu, Bingbing, Li, Guocong, Meng, Xiaofeng
Abstract
Large language models (LLMs) frequently produce confident yet factually incorrect responses when user inputs contain misleading premises, a phenomenon we attribute to fact perturbations in the input. Existing approaches to hallucination mitigation typically assume reliable user inputs, overlooking how such factual errors can actively mislead model reasoning. To address this vulnerability, we propose DEDUCE, a three-stage framework that transforms LLMs from passive responders into proactive error correctors. DEDUCE operates in three stages: (1) detect errors through fine-grained fact extraction and verification; (2) devise correction strategies via multi perspective deliberation; and (3) correct misconceptions while delivering reliable answers. We also present MisFactQA, a dataset containing factual errors of varying degrees, and propose new metrics for evaluating model robustness. Experiments on TruthfulQA, FalseQA, and our MisFactQA benchmark demonstrate that DEDUCE significantly improves both accuracy and error correction capability. Consistent gains across Qwen, LLaMA, and Gemma families confirm its effectiveness and scalability.
Chinese Translation
大型语言模型（LLMs）在用户输入包含误导性前提时，常常会产生自信但事实错误的响应，这一现象我们归因于输入中的事实扰动。现有的幻觉缓解方法通常假设用户输入是可靠的，忽视了这些事实错误如何主动误导模型推理。为了解决这一脆弱性，我们提出了 DEDUCE，一个将 LLM 从被动响应者转变为主动纠正者的三阶段框架。DEDUCE 的操作分为三个阶段：（1）通过细粒度的事实提取和验证来检测错误；（2）通过多角度的深思熟虑来制定纠正策略；（3）在提供可靠答案的同时纠正误解。我们还提出了 MisFactQA，一个包含不同程度事实错误的数据集，并提出了新的评估模型鲁棒性的指标。在 TruthfulQA、FalseQA 和我们的 MisFactQA 基准上的实验表明，DEDUCE 显著提高了准确性和纠错能力。在 Qwen、LLaMA 和 Gemma 系列模型上的一致性提升证实了其有效性和可扩展性。
View on arXiv Download PDF AI Translation
cs.CL / 59 / 2608.25904

One Form to Transfer Them All: Pretraining Multilingual Language Models Beyond Native Orthography

一种通用的转移方式：超越母语正字法的多语言模型预训练
Zhang, Muge, Jencks, Aaron, Badikela, Krishna, Tsvetkov, Yulia, Kumar, Sachin
Abstract
Multilingual language models transfer knowledge across languages through shared subword vocabulary, a mechanism that breaks down when related languages use different writing systems. Prior work addresses this via script equalization (romanization or IPA transcription), but direct comparisons are rare; the focus has been on encoder-only models, with most work adapting existing pretrained models. We systematically compare different input representations in autoregressive multilingual pretraining, comparing orthographic text, IPA, and romanization in a controlled setup across three scales (467M, 709M, and 1.03B) on eight languages in four typologically motivated pairs. Across a wide range of downstream tasks on seen and unseen languages, romanized pretraining yields the strongest cross-lingual transfer, and the advantage over text widens with scale. IPA improves over text in most settings but trails romanization. Surprisingly, finetuning a text-pretrained model on romanized data hurts performance on languages already covered by the base model, only marginally helping when the model lacks script coverage. Our results indicate that for multilingual models spanning typologically diverse scripts, to obtain maximum benefits, romanization should be treated as a core design choice applied at pretraining rather than a post hoc fix.
Chinese Translation
多语言模型通过共享的子词词汇在不同语言之间转移知识，但当相关语言使用不同的书写系统时，这一机制会失效。之前的研究通过脚本均衡（罗马化或国际音标转录）来解决这一问题，但直接比较的研究较为稀少；现有研究主要集中在仅使用编码器的模型，大多数工作是对现有预训练模型的适应。我们系统地比较了自回归多语言预训练中不同输入表示，比较了正字法文本、国际音标（IPA）和罗马化在控制设置下的表现，涵盖了三个规模（467M、709M 和 1.03B）以及四对具有类型学动机的八种语言。在对已见和未见语言的广泛下游任务中，罗马化预训练在跨语言转移中表现出最强的效果，并且随着规模的扩大，其相对于文本的优势也在加大。国际音标在大多数设置中优于文本，但仍落后于罗马化。令人惊讶的是，在罗马化数据上对文本预训练模型进行微调会降低在基础模型已覆盖语言上的性能，仅在模型缺乏脚本覆盖时才会有微弱改善。我们的结果表明，对于跨越类型学多样脚本的多语言模型，为了获得最大收益，罗马化应被视为在预训练阶段应用的核心设计选择，而非事后修正。
View on arXiv Download PDF AI Translation
cs.CL / 60 / 2608.25910

SAMpLE: A SystemC-AMS Machine LEarning-based Framework for Virtual Prototyping

SAMpLE：基于SystemC-AMS的机器学习框架用于虚拟原型设计
Albu, Andrei Mihai, Vinco, Sara
Abstract
Machine Learning (ML) is increasingly used in virtual prototypes of embedded systems to model behaviors that are difficult to capture analytically. However, integrating ML models into virtual platform simulation is still typically done through ad hoc solutions, which limits reuse, comparability, and reproducibility. This paper presents \textbf{\textit{SAMpLE}}, an open-source SystemC-AMS-based framework that integrates ML models as first-class Timed Dataflow (TDF) components through a standardized plug-and-play interface. SAMpLE provides two execution backends: a native C++ backend for online training of lightweight models, and an offline backend for executing externally developed models without requiring re-implementation in C++ or manual integration steps. The framework uses ONNX as a standard model exchange format to enable integration of externally trained ML models into SystemC-AMS simulations, and allows the evaluation of different ML-based solutions within the same testbench, dataset, and simulation workflow. The modular design and unified and reproducible environment will allow future extensions of SAMpLE to new models, without modifying the SystemC-AMS structure.
Chinese Translation
机器学习（ML）在嵌入式系统的虚拟原型中越来越多地被用于建模难以通过分析捕捉的行为。然而，将ML模型集成到虚拟平台仿真中通常仍然是通过临时解决方案来实现的，这限制了复用性、可比性和可重复性。本文提出了 extbf{ extit{SAMpLE}}，一个基于SystemC-AMS的开源框架，通过标准化的即插即用接口将ML模型集成为一流的定时数据流（TDF）组件。SAMpLE提供了两个执行后端：一个用于轻量级模型在线训练的本地C++后端，以及一个用于执行外部开发模型的离线后端，无需在C++中重新实现或手动集成步骤。该框架使用ONNX作为标准模型交换格式，以便将外部训练的ML模型集成到SystemC-AMS仿真中，并允许在同一测试平台、数据集和仿真工作流中评估不同的基于ML的解决方案。模块化设计和统一可重复的环境将允许未来对SAMpLE进行扩展，以支持新模型，而无需修改SystemC-AMS结构。
View on arXiv Download PDF AI Translation
cs.CL / 61 / 2608.25922

Query-Side Attacks on GNN-Based KGQA: Tracing Failures from Entity Linking to Answer Generation

基于图神经网络的知识图谱问答中的查询侧攻击：从实体链接到答案生成的失败追踪
Kumar, Pankaj, Mishra, Subhankar
Abstract
GNN-based Knowledge Graph Question Answering (KGQA) pipelines process queries through four discrete stages: entity linking, subgraph retrieval, GNN reasoning, and answer generation. Standard robustness evaluations conflate stage-level failures into a single end-to-end metric, obscuring both the source of brittleness and the appropriate mitigation target. We ask which stage fails, and why, when the pipeline is subjected to adversarial perturbations on the input question. We introduce a stage-isolation protocol with two answer-preserving adversarial perturbations verified against the knowledge graph: Compositional Restructuring (CR) and Relation Synonym Swap (RS) target distinct stages while leaving entity seeds intact. Evaluated across ComplexWebQuestions and WebQSP, the results run counter to prevailing assumptions: the GNN reasoning stage retains near-baseline accuracy when the subgraph is intact, while subgraph construction accounts for over 99\% of the end-to-end collapse under CR, occurring even when the gold answer is present in 74\% of retrieved subgraphs. This exposes a fundamental distinction between answer presence and answer reachability that end-to-end metrics cannot detect, and places the mitigation target firmly at the subgraph construction stage rather than the reasoning model. Perturbed datasets and evaluation infrastructure are released at https://anonymous.4open.science/r/atkgrag-E85C .
Chinese Translation
基于图神经网络的知识图谱问答（KGQA）流程通过四个离散阶段处理查询：实体链接、子图检索、图神经网络推理和答案生成。标准的鲁棒性评估将阶段级失败混合为单一的端到端指标，模糊了脆弱性的来源和适当的缓解目标。我们探讨在输入问题受到对抗性扰动时，哪个阶段失败，以及原因是什么。我们引入了一种阶段隔离协议，采用两种保留答案的对抗性扰动，并针对知识图谱进行了验证：组合重构（Compositional Restructuring, CR）和关系同义词交换（Relation Synonym Swap, RS），它们针对不同阶段，同时保持实体种子不变。在对 ComplexWebQuestions 和 WebQSP 的评估中，结果与普遍假设相悖：当子图完整时，图神经网络推理阶段保持接近基线的准确率，而子图构建在 CR 下占据了超过 99\% 的端到端崩溃，即使在 74\\% 的检索子图中存在黄金答案时也会发生。这揭示了答案存在与答案可达性之间的根本区别，而端到端指标无法检测到这一点，并将缓解目标明确定位在子图构建阶段，而非推理模型。扰动数据集和评估基础设施已发布在 https://anonymous.4open.science/r/atkgrag-E85C 。
View on arXiv Download PDF AI Translation
cs.CL / 62 / 2608.25944

Unveiling Spectral Mechanisms in Training-Free LLM Text Detection

揭示无训练大型语言模型文本检测中的谱机制
Luo, Haitong, Meng, Xuying, Zhang, Weiyao, Zou, Wenji, Lou, Shengfeng, Jiang, Xuefeng, Lin, Chungang, Zhang, Yujun
Abstract
The rapid advancement of Large Language Models (LLMs) makes it increasingly difficult to distinguish human writing from machine-generated text. Training-free detection offers a scalable solution, yet common confidence-based metrics mainly measure average token probabilities and often miss the signal fluctuations that characterize human writing, which we call "generative vitality". Spectral analysis offers a way to capture this vitality, but its mechanism and practical boundaries remain underexplored. In this paper, we analyze spectral detection from both theoretical and empirical perspectives. We connect spectral energy to variance in proxy log-probability trajectories and explain how broader human token choices create the fluctuations used by frequency-domain indicators. We further show that the strength of this signal depends on text length and sampling range: spectral evidence is clearest for long, continuous, constrained generation, while short, fragmented, mixed, and edited settings require complementary confidence and fluctuation views. These findings clarify when frequency-domain detection works and provide guidance for future multi-dimensional detector design.
Chinese Translation
大型语言模型（LLMs）的快速发展使得区分人类写作与机器生成文本变得愈加困难。无训练检测提供了一种可扩展的解决方案，但常见的基于置信度的度量主要测量平均标记概率，往往忽视了人类写作所特有的信号波动，我们称之为“生成活力”。谱分析提供了一种捕捉这种活力的方法，但其机制和实际边界仍未得到充分探讨。本文从理论和实证两个角度分析谱检测。我们将谱能量与代理对数概率轨迹的方差联系起来，并解释了更广泛的人类标记选择如何产生频域指标所使用的波动。我们进一步表明，这种信号的强度依赖于文本长度和采样范围：谱证据在长、连续、受限的生成中最为清晰，而短小、碎片化、混合和编辑的设置则需要互补的置信度和波动视角。这些发现阐明了频域检测的有效性，并为未来多维检测器的设计提供了指导。
View on arXiv Download PDF AI Translation
cs.CL / 63 / 2608.25976

Lost but not erased: Finding traces of a forgotten language in neural speech models

失落但未被抹去：在神经语音模型中寻找被遗忘语言的痕迹
Plantinga, Peter, Moore, Charlotte, Donhauser, Peter W., Byers-Heinlein, Krista, Klein, Denise
Abstract
International adoptees retain phonological traces of a birth language they can no longer speak or comprehend, a persistence typically attributed to a biologically-timed critical period. We asked whether it could instead reflect the ordinary dynamics of learning, using automatic speech recognition models that simulate the international adoptee experience without maturational confounds. Models were trained on one language and then abruptly switched to a second. We found that traces of the first language persisted throughout second-language training, but mainly in the lowest, pre-phonemic layers. These traces were functional, as models with early exposure re-learned their lost first language 14% faster than naive models; this advantage held even against models adopted early from a related language and disappeared when the earliest layers were substituted from a non-adopted model. We argue that these critical-period effects reflect entrenchment of foundational representations rather than a maturational loss of plasticity, and that experience plays a central role in critical periods in language acquisition.
Chinese Translation
国际收养者保留了他们已无法说或理解的母语的音位痕迹，这种持续性通常归因于生物学上定时的关键期。我们探讨这是否可能反映了学习的普通动态，使用自动语音识别模型来模拟国际收养者的经历，而不受发育混淆的影响。模型在一种语言上训练，然后突然切换到第二种语言。我们发现第一语言的痕迹在第二语言训练过程中持续存在，但主要集中在最低的、前音位层。这些痕迹是功能性的，因为早期接触的模型以比天真的模型快14%的速度重新学习了他们失去的第一语言；这一优势在与早期从相关语言收养的模型比较时依然存在，并且在最早的层被来自非收养模型的层替代时消失。我们认为这些关键期效应反映了基础表征的巩固，而不是发育塑性丧失的结果，并且经验在语言习得的关键期中发挥了核心作用。
View on arXiv Download PDF AI Translation
cs.CL / 64 / 2608.25977

When Personality Meets Quantization: A Layer-wise MBTI Analysis of Quantized LLMs

当个性遇上量化：对量化大型语言模型的逐层 MBTI 分析
Fu, Yao, Huang, Lijia, Li, Xiaomin, Li, Runchao, Yin, Yu, Loparo, Kenneth A.
Abstract
Personality is increasingly important in large language models (LLMs), as it shapes users' trust, engagement, and emotional experiences. While the Myers--Briggs Type Indicator (MBTI) has emerged as a common framework for assessing LLMs' personality, existing studies focus primarily on full-precision models and evaluate only final outputs. They overlook the widespread deployment of quantized LLMs requiring low memory footprints, whose personality traits remain underexplored. In this work, we present a systematic MBTI analysis of open-source LLMs across multiple precisions, including mainstream 4-bit methods (GPTQ, AWQ) and extreme 2-bit settings (AQLM variants). Beyond output-level evaluation, we examine how personality emerges across layers through option-level entropy and confidence-gap dynamics, and introduce Uncertainty-Amplified Layer Decoding (UALD) to study decoding-induced personality drift at inference time. Our results reveal a key insight: LLMs' personality is not a static property, but an emergent, layer-dependent decision process sensitive to quantization, prompting, and decoding. Specifically, we find that (1) ENFJ remains dominant across model families and precisions; (2) 4-bit quantization largely preserves coarse personality structure, while 2-bit quantization disrupts fine-grained prompt consistency and cross-precision agreement; (3) personality decisions emerges in upper layers, following substantial ambiguity in early layers; and (4) inference decoding can shift personality, while personality-aligned conditioning improves robustness. These findings provide a new perspective on the behavioral reliability of quantized LLMs and highlight the importance of considering internal dynamics and inference strategies in personality-sensitive chatbot applications.
Chinese Translation
个性在大型语言模型（LLMs）中变得越来越重要，因为它影响用户的信任、参与度和情感体验。虽然迈尔斯-布里格斯性格类型指标（MBTI）已成为评估 LLMs 个性的常用框架，但现有研究主要集中在全精度模型上，仅评估最终输出，忽视了对需要低内存占用的量化 LLMs 的广泛部署，这些模型的个性特征仍然未被充分探索。在本研究中，我们对多种精度的开源 LLMs 进行了系统的 MBTI 分析，包括主流的 4 位方法（GPTQ、AWQ）和极端的 2 位设置（AQLM 变体）。除了输出级评估外，我们还通过选项级熵和置信度差动态考察个性如何在各层中显现，并引入不确定性放大层解码（UALD）来研究推理时解码引起的个性漂移。我们的结果揭示了一个关键见解：LLMs 的个性并不是静态属性，而是一个依赖于层次的、对量化、提示和解码敏感的涌现决策过程。具体而言，我们发现：（1）ENFJ 在各模型家族和精度中保持主导地位；（2）4 位量化在很大程度上保留了粗略的个性结构，而 2 位量化则破坏了细粒度的提示一致性和跨精度一致性；（3）个性决策在上层显现，早期层次存在显著模糊性；（4）推理解码可以改变个性，而与个性一致的条件设置提高了鲁棒性。这些发现为量化 LLMs 的行为可靠性提供了新的视角，并强调在个性敏感的聊天机器人应用中考虑内部动态和推理策略的重要性。
View on arXiv Download PDF AI Translation
cs.CL / 65 / 2608.25999

Distinct dynamics of conceptual and referential disruptions in human reading and large language model processing

人类阅读与大型语言模型处理中的概念性和指称性干扰的不同动态
He, Rui, Altay, Nihal, Hinzen, Wolfram
Abstract
Linguistic meaning is grounded in conceptual content, from which reference to particular entities emerges as words enter discourse. To examine the processing dynamics associated with these two dimensions of meaning, we selectively disrupted conceptual or referential information in short narratives and traced the resulting effects in human self-paced reading and in the predictive and representational processing of large language models. In human reading, conceptual disruptions produced a strong but localized processing cost, emerging immediately after the distorted word, reaching an early maximum, and then declining rapidly. Referential disruptions produced weaker effects, which decreased more gradually across subsequent words, and were more strongly modulated by sentence boundaries. In the language model, both disruptions emerged immediately at the manipulated word. Contextual model surprisal showed a pattern closely paralleling human reading: conceptual disruption produced a larger, more locally concentrated effect that decayed rapidly, whereas referential disruption produced a smaller and more gradual downstream effect. Output-layer representations showed a different pattern: referential disruption produced a larger initial displacement, while both distortions were subsequently characterized by power-law decay. Together, these results provide convergent evidence for distinguishable processing dynamics of two types of meaning: conceptual information imposes a more locally concentrated integration cost, whereas referential information engages a more distributed process of maintaining discourse-level identity.
Chinese Translation
语言意义植根于概念内容，从中对特定实体的指称随着词语进入话语而产生。为了考察与这两个意义维度相关的处理动态，我们在短篇叙述中选择性地干扰了概念信息或指称信息，并追踪了在人类自我调节阅读和大型语言模型的预测与表征处理中的结果。在人类阅读中，概念性干扰产生了强烈但局部的处理成本，出现在扭曲词语之后，迅速达到早期最大值，然后迅速下降。指称性干扰产生了较弱的效果，这些效果在随后的词语中逐渐减弱，并且受到句子边界的更强调节。在语言模型中，这两种干扰都在被操控的词语处立即显现。上下文模型的惊讶度显示出与人类阅读相似的模式：概念性干扰产生了更大、局部集中效果，且迅速衰减，而指称性干扰则产生了较小且逐渐的下游效果。输出层表征显示出不同的模式：指称性干扰产生了更大的初始位移，而两种扭曲随后都表现出幂律衰减的特征。综合来看，这些结果为两种意义类型的可区分处理动态提供了趋同证据：概念信息施加了更局部集中的整合成本，而指称信息则涉及维持话语层面身份的更分散的过程。
View on arXiv Download PDF AI Translation
cs.CL / 66 / 2608.26013

VISA: Agentic Self-Evolving Data Synthesis for Multimodal Instruction Following

VISA：用于多模态指令跟随的自主自演数据合成
Zeng, Min, Tan, Guanxin, Cen, Libin, Wen, Yawei, Hu, Rui, Bian, Liuyang, Chen, Xiaolong, Chen, Xiaoxin
Abstract
Multimodal instruction-following models require training data that is accurate, diverse, verifiable, and challenging. Existing synthesis pipelines typically follow a one-pass generate-and-filter paradigm, discarding feedback from failed samples, verifier outcomes, and target-model errors. We present VISA (Visual Instruction Synthesis Agent), an agentic framework that reformulates multimodal instruction synthesis as a self-evolving loop. At each round, VISA analyzes an image to filter incompatible constraints and discover new verifiable ones, samples diversity- and difficulty-aware constraint sets from persistent memory, generates candidate instructions, and verifies the resulting samples with executable tools and structured large language model judges. Failed samples trigger diagnostic-guided recovery, while accepted samples are probed against the target model to estimate difficulty. The resulting verifier signals and target-model failure profiles are written back to memory, allowing subsequent rounds to adaptively expand the constraint space, reduce template repetition, and focus on unresolved model weaknesses. The same verifier contracts further provide reward signals for reinforcement learning without a separately trained reward model. Experiments on MM-IFEval show that VISA consistently improves multimodal instruction following over strong baselines, while preserving general multimodal capability across seven public benchmarks.
Chinese Translation
多模态指令跟随模型需要准确、多样、可验证且具有挑战性的训练数据。现有的合成管道通常遵循一次性生成和过滤的范式，忽略了来自失败样本、验证器结果和目标模型错误的反馈。我们提出了VISA（视觉指令合成代理），这是一个将多模态指令合成重新构建为自我演化循环的自主框架。在每一轮中，VISA分析图像以过滤不兼容的约束并发现新的可验证约束，从持久记忆中抽样多样性和难度感知的约束集，生成候选指令，并使用可执行工具和结构化大型语言模型评审者验证生成的样本。失败的样本触发诊断引导的恢复，而被接受的样本则针对目标模型进行探测以估计难度。生成的验证器信号和目标模型失败特征被写回记忆，使后续轮次能够自适应地扩展约束空间，减少模板重复，并集中于未解决的模型弱点。同一验证器合约进一步提供了强化学习的奖励信号，而无需单独训练的奖励模型。在MM-IFEval上的实验表明，VISA在强基线之上始终改善了多模态指令跟随，同时在七个公共基准测试中保持了通用的多模态能力。
View on arXiv Download PDF AI Translation
cs.CL / 67 / 2608.26035

Beyond Local Surprise: Grounded Dialogue as Selective Belief Revision under Referential Uncertainty

超越局部惊讶：在指称不确定性下的扎根对话作为选择性信念修正
Liu, Ziming, Jasti, Bhanu Chaitanya, Xu, Ziyang, Wu, Hongyu, Wu, Yi, Liu, Jiqun
Abstract
When a speaker refers to a scene that the listener cannot directly see, the listener must decide whether to preserve its current understanding or revise it as new utterances arrive. Many language systems treat local mismatch as a cue for updating: divergence from the current understanding encourages adjustment. Yet conversational understanding may be more conservative, interpreting mismatching evidence relative to prior understanding rather than immediately revising it. We introduce a controlled, data-driven framework for turn-by-turn preserve/revise decisions in dialogue, where competing revision policies are learned under otherwise identical conditions. We compare four theory-driven revision strategies, each reflecting a different assumption about when listeners should preserve or revise. Two findings stand out. First, a mismatch-driven policy that updates solely based on local divergence reacts strongly to mismatch but destabilizes grounding and degrades retrieval. Second, an uncertainty-sensitive policy extends mismatch-based updating with accumulated evidence, preserving coherent understanding while maintaining strong retrieval performance. Surprisingly, coherent understanding emerges from a counterintuitive pattern: local mismatch promotes preservation, whereas accumulated uncertainty promotes revision, suggesting that listeners maintain prior understanding despite local mismatch and revise only when uncertainty sufficiently accumulates. This pattern is consistent with conceptual pact theory.
Chinese Translation
当说话者提到听者无法直接看到的场景时，听者必须决定是保留当前理解还是在新话语到来时进行修正。许多语言系统将局部不匹配视为更新的线索：与当前理解的偏离促使调整。然而，交际理解可能更为保守，相对于先前理解来解释不匹配的证据，而不是立即进行修正。我们引入了一个受控的数据驱动框架，用于对话中的逐轮保留/修正决策，在其他条件相同的情况下学习竞争的修正策略。我们比较了四种理论驱动的修正策略，每种策略反映了关于听者何时应保留或修正的不同假设。有两个发现尤为突出。首先，基于不匹配驱动的策略仅根据局部偏离进行更新，对不匹配反应强烈，但会破坏基础并降低检索性能。其次，敏感于不确定性的策略通过累积证据扩展了基于不匹配的更新，保持了一致的理解，同时维持了强大的检索性能。令人惊讶的是，一致的理解源于一种反直觉的模式：局部不匹配促进保留，而累积不确定性促进修正，这表明听者在局部不匹配的情况下仍然保持先前理解，只有在不确定性足够累积时才进行修正。这一模式与概念契约理论一致。
View on arXiv Download PDF AI Translation
cs.CL / 68 / 2608.26060

Fine-Tuning Whisper for Automatic Speech Recognition in Baniwa: A Preliminary Study

针对巴尼瓦语的自动语音识别的Whisper微调：初步研究
Duart, Leonardo, Fonseca, Tiago, Chacón, Thiago
Abstract
Automatic Speech Recognition (ASR) technologies have achieved remarkable performance in recent years through the use of large multilingual foundation models. However, most advances remain concentrated on high-resource languages, while indigenous languages continue to suffer from a lack of speech resources and language technologies. This work presents a preliminary study on the adaptation of Whisper for Automatic Speech Recognition in Baniwa, an indigenous Arawakan language spoken in Brazil, Colombia, and Venezuela. The experiments were conducted using a corpus of 1,373 manually transcribed recordings obtained from a linguistic documentation project. The corpus contains approximately 0.54 hours of speech and consists primarily of isolated words and short elicited utterances. The Whisper Small model was fine-tuned using supervised learning and evaluated using Word Error Rate (WER) and Character Error Rate (CER). The best model achieved a WER of 37.5% and a CER of 7.45%, demonstrating that multilingual foundation models can be successfully adapted to extremely low-resource indigenous languages. The results establish an initial baseline for Baniwa Automatic Speech Recognition and provide a foundation for future research involving larger datasets, language-specific adaptation strategies, and post-processing techniques.
Chinese Translation
自动语音识别（ASR）技术近年来通过使用大型多语言基础模型取得了显著的性能。然而，大多数进展仍集中在高资源语言上，而土著语言在语音资源和语言技术方面仍然面临缺乏的问题。本研究呈现了对Whisper模型在巴尼瓦语（Baniwa）自动语音识别中的适应性初步研究，巴尼瓦语是一种在巴西、哥伦比亚和委内瑞拉使用的土著阿拉瓦克语。实验使用了来自语言文献项目的1,373个手动转录录音的语料库进行。该语料库包含约0.54小时的语音，主要由孤立词和短语音刺激语句组成。使用监督学习对Whisper Small模型进行了微调，并通过词错误率（WER）和字符错误率（CER）进行评估。最佳模型达到了37.5%的WER和7.45%的CER，证明多语言基础模型可以成功适应极低资源的土著语言。这些结果为巴尼瓦语自动语音识别建立了初步基准，并为未来涉及更大数据集、特定语言的适应策略和后处理技术的研究提供了基础。
View on arXiv Download PDF AI Translation
cs.CL / 69 / 2608.26070

Prefix Sliding for efficient test-time scaling

前缀滑动：高效的测试时间扩展
Muennighoff, Niklas, Wang, Zhengyang, Chen, Zeyi, Shi, Weijia, Hui, Binyuan, Yang, John, Jiang, Dapeng, Senghaas, Mika, Obeid, Fares, Hagemann, Johannes, Jaghouar, Sami, Schmidt, Ludwig, Liang, Percy, Wei, Jason, Ng, Andrew Y., Zettlemoyer, Luke, Choi, Yejin, Lewis, Mike
Abstract
Test-time scaling uses extra test-time compute to improve performance, such as letting language models reason longer when solving a problem. As models keep the entire reasoning trace in memory via full attention, hard tasks that need long thinking can be prohibitively expensive. However, we find most intermediate reasoning tokens lose importance as the model continues reasoning. This calls into question whether retaining them is worth the cost. Based on this insight, we propose Prefix Sliding, which discards tokens during reasoning that are not part of the prefix or the window of the last few thousand tokens. The prefix has key instructions and tools available to the model, while the most recent tokens are the current reasoning the model is working on. This caps the total memory requirement regardless of how long the model reasons, allowing for efficient long-horizon test-time scaling. Without training, Prefix Sliding can make existing models 3x faster while maintaining performance. Training with Prefix Sliding using reinforcement learning can achieve better performance by enabling scaling to reasoning traces beyond a hundred thousand tokens. Ablations show Prefix Sliding outperforms summarizing intermediate tokens or vanilla sliding window. Our code is at https://github.com/Muennighoff/prefix-sliding
Chinese Translation
测试时间扩展利用额外的测试时间计算来提高性能，例如让语言模型在解决问题时进行更长时间的推理。由于模型通过全注意力机制在内存中保留整个推理轨迹，处理需要长时间思考的困难任务可能会非常昂贵。然而，我们发现大多数中间推理标记在模型继续推理时失去了重要性。这使得保留它们的成本是否值得成为一个问题。基于这一见解，我们提出了前缀滑动（Prefix Sliding），在推理过程中丢弃那些不属于前缀或最近几千个标记窗口的标记。前缀包含模型可用的关键指令和工具，而最近的标记则是模型当前正在处理的推理。这限制了总内存需求，无论模型推理多长时间，从而实现高效的长时间测试时间扩展。在不进行训练的情况下，前缀滑动可以使现有模型速度提高三倍，同时保持性能。通过使用强化学习进行前缀滑动训练，可以通过扩展到超过十万个标记的推理轨迹来实现更好的性能。消融实验表明，前缀滑动优于总结中间标记或普通滑动窗口。我们的代码可在 https://github.com/Muennighoff/prefix-sliding 获取。
View on arXiv Download PDF AI Translation
arXiv Reader
Automatically generated on 2026-08-31 06:58:27
Powered by OpenAI GPT & arXiv