cs.RO / 1 / 2607.01281
WaveLander: A Generalizable Hierarchical Control Framework for UAV Landing on Wave-Disturbed Platforms via Reinforcement Learning
WaveLander：一种可推广的分层控制框架，通过强化学习实现无人机在波动扰动平台上的着陆
Abstract
Autonomous landing of unmanned aerial vehicles (UAVs) on wave-disturbed marine platforms remains challenging due to stochastic platform motion, time-varying platform attitude, and uncertain touchdown conditions. Existing model-based methods often require accurate motion prediction and online optimization, while end-to-end learning approaches may suffer from high training complexity and limited interpretability. This paper presents WaveLander, a hierarchical control framework via reinforcement learning (RL) that decouples vertical landing decision-making from low-level flight stabilization. The RL policy maps a compact platform-relative observation to a scalar vertical velocity reference, while a conventional low-level flight controller maintains attitude stability and lateral tracking. This formulation reduces dynamic platform landing to a low-dimensional, timing-aware control problem and enables smooth landing behavior without explicit switching rules. Simulation results under randomized wave-induced platform motions show that WaveLander achieves robust landing performance and generalizes to unseen disturbance conditions, demonstrating the potential of hierarchical learning-based control for marine UAV recovery.
Chinese Translation
无人机（UAV）在波动扰动的海洋平台上自主着陆仍然面临挑战，原因包括平台运动的随机性、平台姿态的时变性以及不确定的着陆条件。现有的基于模型的方法通常需要准确的运动预测和在线优化，而端到端学习方法可能面临高训练复杂性和有限的可解释性。本文提出了WaveLander，一种通过强化学习（RL）实现的分层控制框架，该框架将垂直着陆决策与低级飞行稳定化解耦。RL策略将紧凑的相对平台观测映射为标量垂直速度参考，而传统的低级飞行控制器则保持姿态稳定和横向跟踪。这种表述将动态平台着陆问题简化为一个低维、时间感知的控制问题，并在没有明确切换规则的情况下实现平稳着陆行为。在随机波动引起的平台运动下的仿真结果表明，WaveLander实现了稳健的着陆性能，并能够推广到未见的扰动条件，展示了基于分层学习的控制在海洋无人机恢复中的潜力。
cs.RO / 2 / 2607.01287
Adaptive Companionship for Group-Following Robots: Handling Dynamically Changing Group Formations
适应性伴随的群体跟随机器人：处理动态变化的群体队形
Abstract
Accompanying a group of humans is an essential aspect of developing human-like social cognition in robots. However, human groups typically do not follow fixed formations, which poses significant challenges for robots in maintaining natural companionship behaviors. In this paper, we propose an adaptive group-accompaniment method for social robots based on Vision-Language Models (VLMs), leveraging their semantic reasoning capabilities to infer companion positions, maintain social distances, and understand group dynamics. The members of the group are first detected, and a perceptual module generates visual representations of the interaction group space as input to the VLM, which is then combined with a Model Predictive Path Integral (MPPI) controller to ensure stability and safety. Experimental evaluations across five scenarios show that the proposed method enables robots to accompany the group effectively, demonstrating a 15\% improvement in success rate and a 25\% reduction in collision rate compared to baseline approaches. Additionally, a user study indicates that the generated companionship behaviors are perceived as natural and socially appropriate.
Chinese Translation
陪伴一群人是发展类人社会认知机器人中的一个重要方面。然而，人类群体通常不遵循固定的队形，这给机器人维持自然的伴随行为带来了重大挑战。本文提出了一种基于视觉-语言模型（Vision-Language Models, VLMs）的适应性群体伴随方法，利用其语义推理能力来推断伴随者位置、维持社交距离并理解群体动态。首先检测群体成员，然后感知模块生成交互群体空间的视觉表示，作为VLM的输入，随后与模型预测路径积分（Model Predictive Path Integral, MPPI）控制器结合，以确保稳定性和安全性。五个场景的实验评估表明，所提出的方法使机器人能够有效地陪伴群体，与基线方法相比，成功率提高了15%，碰撞率降低了25%。此外，用户研究表明，生成的伴随行为被认为是自然且社会适宜的。
cs.RO / 3 / 2607.01304
The Three Dimensions of ROS 2 Middleware
ROS 2 中间件的三个维度
Abstract
ROS 2 (Robot Operating System 2) has emerged as the de facto standard for modern robot software development, with middleware implementations such as the Data Distribution Service (DDS) and Zenoh forming the core infrastructure for distributed robotic communication. Despite their architectural flexibility, these middleware systems exhibit structural limitations, particularly under dynamic and resource-constrained wireless environments. This paper presents a systematic survey of ROS 2 middleware and introduces a conceptual framework to examine its architectural limits through three structural dimensions required by distributed robotic systems, namely Space, Time, and State. We first provide a structured analysis of middleware architecture and operational dynamics, including discovery, data exchange, and state management mechanisms. Building on this foundation, we formalize Time as temporal predictability for control loops, Space as spatial abstraction from physical topology to enable modular deployment, and State as contextual continuity despite dynamic node participation and intermittent connectivity. Through a comprehensive review of existing implementations and prior studies, we organize middleware research according to the structural trade-offs that arise among these dimensions. Under constrained wireless conditions, spatial abstraction can obscure network variability and weaken temporal guarantees, while mechanisms that preserve state continuity introduce computational and network overhead that competes with time-critical communication. These interactions reveal structural trade-offs that characterize the practical limits of contemporary robot middleware. By synthesizing architectural patterns and identifying gaps in current modeling and analysis approaches, this survey outlines a principled research roadmap for robust and scalable robotic middleware architectures.
Chinese Translation
ROS 2（机器人操作系统 2）已成为现代机器人软件开发的事实标准，其中间件实现如数据分发服务（DDS）和 Zenoh 构成了分布式机器人通信的核心基础设施。尽管这些中间件系统具有架构灵活性，但在动态和资源受限的无线环境下，它们表现出结构上的局限性。本文对 ROS 2 中间件进行了系统性的调查，并引入了一个概念框架，通过分布式机器人系统所需的三个结构维度，即空间（Space）、时间（Time）和状态（State），来考察其架构限制。我们首先对中间件架构和操作动态进行了结构化分析，包括发现、数据交换和状态管理机制。在此基础上，我们将时间形式化为控制循环的时间可预测性，将空间形式化为从物理拓扑到模块化部署的空间抽象，将状态形式化为尽管节点参与动态变化和连接间歇性，仍保持的上下文连续性。通过对现有实现和先前研究的全面回顾，我们根据这些维度之间出现的结构权衡组织了中间件研究。在受限的无线条件下，空间抽象可能会掩盖网络变异性并削弱时间保证，而保持状态连续性的机制则引入了与时间关键通信竞争的计算和网络开销。这些相互作用揭示了表征当代机器人中间件实际限制的结构权衡。通过综合架构模式并识别当前建模和分析方法中的空白，本文概述了一条原则性的研究路线图，以实现稳健且可扩展的机器人中间件架构。
cs.RO / 4 / 2607.01378
Neuro-Symbolic Safety Guidance for Vision-Language-Action Models via Constrained Flow Matching
基于约束流匹配的神经符号安全引导用于视觉-语言-动作模型
Abstract
Vision-Language-Action (VLA) models have demonstrated promising generalization capabilities across robotic manipulation tasks, yet their real-world deployment remains limited by the lack of effective safety measures. Specifically, existing safety measures only prevent collisions caused by the robot's next action. In this paper, we propose a neuro-symbolic safety guidance mechanism for flow matching based VLAs that enables predictive collision avoidance. Flow matching based VLAs determine the next actions by predicting a trajectory (a sequence of actions) through an iterative neural flow matching process. Our method formulates safety enforcement as a minimum-norm constrained optimization problem that corrects safety violations during the denoising process of noisy intermediate trajectory predictions. By analyzing predicted trajectories and applying corrections during iterative denoising, our approach anticipates collisions before they become unavoidable. This interleaving of symbolic constraint satisfaction with neural trajectory generation enables predictive collision avoidance rather than reactive intervention. On the SafeLIBERO benchmark, our method achieves 82.8% collision avoidance and 81.6% task success, a 6.3% and 19.8% improvement respectively over single-step methods, with the largest gains on long-horizon tasks where compounding distribution shift is most pronounced. Video demonstrations of our approach are included on our project page at https://willenglish.tech/SafetyGuidedFlowMatching/.
Chinese Translation
视觉-语言-动作（VLA）模型在机器人操作任务中展示了良好的泛化能力，但其在现实世界中的应用受到缺乏有效安全措施的限制。现有的安全措施仅能防止机器人下一步动作引起的碰撞。本文提出了一种神经符号安全引导机制，适用于基于流匹配的VLA，能够实现预测性碰撞避免。基于流匹配的VLA通过迭代神经流匹配过程预测轨迹（动作序列）来确定下一步动作。我们的方法将安全执行形式化为一个最小范数约束优化问题，在噪声中间轨迹预测的去噪过程中纠正安全违规行为。通过分析预测轨迹并在迭代去噪过程中应用修正，我们的方法能够在碰撞变得不可避免之前进行预判。这种符号约束满足与神经轨迹生成的交错结合，使得预测性碰撞避免成为可能，而非被动干预。在SafeLIBERO基准测试中，我们的方法实现了82.8%的碰撞避免率和81.6%的任务成功率，分别比单步方法提高了6.3%和19.8%，在长时间任务中，复合分布偏移最为明显，获得了最大的提升。我们的方法的视频演示可在我们的项目页面 https://willenglish.tech/SafetyGuidedFlowMatching/ 中查看。
cs.RO / 5 / 2607.01382
CommonRoad-Game: A Human-in-the-Loop Simulation Framework for Autonomous Driving
CommonRoad-Game：一种人机协同的自主驾驶仿真框架
Abstract
Motion planning algorithms should be evaluated in human-in-the-loop environments to ensure they produce safe and efficient behaviors during interactions. However, existing simulation platforms often rely on recorded datasets, lack dedicated interfaces for real-time human interaction, or remain weakly integrated with an autonomous driving ecosystem. Moreover, many human-in-the-loop simulators are computationally intensive by design, making them less suitable for rapid prototyping and flexible experimentation in early-stage autonomous driving research. To address these limitations, we present CommonRoad-Game, a lightweight human-in-the-loop simulation framework tightly integrated with the CommonRoad platform, focusing on the systematic testing of motion planners with human participation and the analysis of human driving behaviors in interactive scenarios. We introduce a multi-threaded architecture with a robust synchronization mechanism that aligns simulation time with wall-clock time, enabling deterministic and temporally consistent interaction between autonomous and human-driven vehicles. In addition, the framework provides a scenario generation module that records driving logs, allowing diverse and reproducible test cases to be constructed from human-in-the-loop experiments. Experimental results demonstrate that CommonRoad-Game achieves stable temporal synchronization, supports scalable multi-agent simulation, and seamlessly integrates CommonRoad-compatible motion planners to generate interactive driving scenarios. The source code is publicly available at https://github.com/Yunfei-Bi8/CommonRoad-Game.
Chinese Translation
运动规划算法应在包含人类参与的环境中进行评估，以确保它们在交互过程中产生安全和高效的行为。然而，现有的仿真平台通常依赖于录制的数据集，缺乏实时人机交互的专用接口，或与自主驾驶生态系统的集成较弱。此外，许多人机协同仿真器在设计上计算密集，使其不太适合在早期自主驾驶研究中进行快速原型开发和灵活实验。为了解决这些局限性，我们提出了CommonRoad-Game，一个与CommonRoad平台紧密集成的轻量级人机协同仿真框架，专注于系统性测试运动规划器与人类参与的情况，以及分析交互场景中的人类驾驶行为。我们引入了一种多线程架构，配备强大的同步机制，使仿真时间与实际时间对齐，从而实现自主驾驶与人类驾驶车辆之间的确定性和时间一致的交互。此外，该框架提供了一个场景生成模块，记录驾驶日志，允许从人机协同实验中构建多样且可重复的测试案例。实验结果表明，CommonRoad-Game实现了稳定的时间同步，支持可扩展的多智能体仿真，并无缝集成了与CommonRoad兼容的运动规划器，以生成交互式驾驶场景。源代码可在 https://github.com/Yunfei-Bi8/CommonRoad-Game 获取。
cs.RO / 6 / 2607.01410
BIFROST: Bridging Invariant Feature Representation for Observation-space Sim2Real Transfer
BIFROST：桥接不变特征表示以实现观察空间的Sim2Real转移
Abstract
Sim2real transfer for robot policy learning suffers due to mismatch between simulation and reality. Existing methods typically address each gap in isolation through separate adaptation modules, which are composed or layered when both gaps coexist. Yet the basis for attempting sim2real in the first place is that there is shared structure between a task in simulation and reality, where equivalent actions from equivalent configurations produce equivalent long term outcomes regardless of domain specific differences in rendering or physics. In this paper, we study whether we can identify and exploit this shared structure from raw observations to train a policy that enables zero shot transfer. We introduce BIFROST, which learns a shared history encoder on paired cross-domain data via cross-domain bisimulation objective: observation-action sequences leading to equivalent long-term behavior are mapped to nearby latent states, regardless of domain. Policies trained on these latent states in simulation transfer zero-shot to reality. We provide empirical evidence on sim2sim visual navigation and sim2real contact rich manipulation task and visual servoing task that BIFROST achieves effective transfer where domain adaptation and co-training baselines fail under both visual and dynamics domain gaps.
Chinese Translation
机器人策略学习中的Sim2Real转移受到仿真与现实之间不匹配的影响。现有方法通常通过独立的适应模块分别解决每个差距，当两个差距共存时，这些模块会被组合或叠加。然而，尝试进行Sim2Real转移的基础在于仿真任务与现实之间存在共享结构，其中来自等效配置的等效动作在不同领域特定的渲染或物理差异下产生等效的长期结果。本文研究我们是否可以从原始观察中识别并利用这种共享结构，以训练能够实现零-shot转移的策略。我们引入了BIFROST，它通过跨领域双仿真目标在配对的跨领域数据上学习共享历史编码器：导致等效长期行为的观察-动作序列被映射到相邻的潜在状态，无论领域如何。在仿真中基于这些潜在状态训练的策略能够零-shot转移到现实中。我们提供了关于sim2sim视觉导航和sim2real接触丰富的操作任务以及视觉伺服任务的实证证据，表明BIFROST在视觉和动态领域差距下实现了有效的转移，而领域适应和共同训练的基线方法则失败。
cs.RO / 7 / 2607.01454
SE(2) Navigation Mesh
SE(2) 导航网格
Abstract
Global navigation for ground robots in complex multi-level environments requires representations that accurately capture traversable regions while enabling efficient path planning. Current approaches present key limitations: Point clouds and volumetric occupancy maps lack explicit surface structure for traversability estimation, whereas direct pathfinding on dense triangle meshes is computationally prohibitive. Navigation meshes mitigate these challenges through polygonal abstraction of the underlying mesh, but assume yaw-invariant traversability, rendering them unsuitable for non-circular robots in constrained spaces. We propose SE(2) Navigation Mesh (SE(2) NavMesh), a polygonal representation of traversable regions that encodes yaw-dependent traversability. Our method evaluates traversability using footprint masks and constructs a graph over yaw-specific layers with explicit translational and rotational connectivity. Grounded in this representation, we develop an A*-String Pulling-A* (ASA) pathfinding strategy that hierarchically optimizes robot position and heading. We also present an online method that incrementally updates the SE(2) NavMesh from streaming point clouds during concurrent geometry reconstruction. In simulation, the SE(2) NavMesh captures over 50% more traversable area than classical NavMeshes, and the SE(2) NavMesh + ASA pipeline consistently outperforms sampling-based baselines in constrained environments. Extensive real-world experiments on a physical robot validate real-time online generation and successful navigation across multiple environments.
Chinese Translation
在复杂的多层环境中，地面机器人进行全局导航需要能够准确捕捉可通行区域的表示，同时支持高效的路径规划。目前的方法存在关键限制：点云和体积占用图缺乏明确的表面结构以进行可通行性估计，而在密集三角网格上进行直接路径寻找在计算上是不可行的。导航网格通过对底层网格的多边形抽象来缓解这些挑战，但假设航向不变的可通行性，使其不适用于在受限空间内的非圆形机器人。我们提出了 SE(2) 导航网格（SE(2) NavMesh），这是一种编码航向依赖可通行性的可通行区域的多边形表示。我们的方法使用足迹掩模评估可通行性，并在特定航向层上构建具有明确平移和旋转连通性的图。基于这种表示，我们开发了一种 A*-字符串拉动-A*（ASA）路径寻找策略，该策略在层次上优化机器人的位置和航向。我们还提出了一种在线方法，该方法在并发几何重建过程中从流式点云中增量更新 SE(2) NavMesh。在仿真中，SE(2) NavMesh 捕捉到的可通行区域比经典的 NavMesh 多出超过 50%，而 SE(2) NavMesh + ASA 流程在受限环境中始终优于基于采样的基线。对物理机器人进行的大量真实世界实验验证了实时在线生成和在多个环境中成功导航的有效性。
cs.RO / 8 / 2607.01554
A Reconfigurable Rocker-Bogie Robot for High Step Climbing and Turning
一种可重构的摇摆-博基机器人用于高阶梯攀爬和转向
Abstract
This study proposes a reconfigurable rocker-bogie mechanism that achieves efficient turning motion with a small number of actuators while maintaining high step-climbing capability. By installing motors at the bogie joints and actively swinging up and down bogies, the system enables switching between four-wheel and six-wheel configurations. Omnidirectional wheels are mounted on the rear ends of the rockers, allowing smooth turning in the four-wheel configuration based on a differential-drive model. Experimental evaluation using a prototype robot demonstrated that the proposed mechanism achieves zero-radius turning at a speed more than five times that of a conventional rocker-bogie mechanism equipped with six non-steerable grip wheels, while requiring only approximately 17% of the total average wheel torque. In addition, the robot successfully climbed a 40 cm step with an average climbing time of 6.4 s, confirming its high turning and step-climbing performance.
Chinese Translation
本研究提出了一种可重构的摇摆-博基机制，该机制在保持高阶梯攀爬能力的同时，能够以较少的执行器实现高效的转向运动。通过在博基关节处安装电机并主动上下摆动博基，该系统实现了四轮和六轮配置之间的切换。全向轮安装在摇摆的后端，使得在基于差动驱动模型的四轮配置中能够实现平稳转向。使用原型机器人进行的实验评估表明，所提机制在速度上超过传统配备六个不可转向抓地轮的摇摆-博基机制的五倍，实现了零半径转向，同时仅需约17%的总平均轮扭矩。此外，机器人成功攀爬了40厘米的阶梯，平均攀爬时间为6.4秒，验证了其出色的转向和阶梯攀爬性能。
cs.RO / 9 / 2607.01574
Multi-Rate Nonlinear Model Predictive Control for Wall-Supported Bipedal Locomotion of Quadrupedal Robots
基于多速率非线性模型预测控制的壁面支持四足机器人双足行走
Abstract
This paper presents a novel layered planning and control framework based on multi-rate nonlinear model predictive control (MR-NMPC) that enables quadrupedal robots to perform hybrid bipedal locomotion with wall-assisted support in constrained environments. Real-time trajectory optimization for this locomotion presents significant challenges, as the controller must simultaneously plan for both the contact points and the continuous trajectories of the robot's center of mass (CoM) and orientation within the robot's nonlinear dynamics while accounting for unilateral contact constraints, underactuation, and the switching nature of the robot's dynamics. At the high level of the control framework, an MR-NMPC is proposed, which dynamically plans both the discrete-time trajectories of the contact points and the continuous-time trajectories of the CoM and orientation, using a single rigid body (SRB) dynamics model. By incorporating contact-point planning within the multi-rate optimal control framework, this approach enhances dynamic stability compared to heuristic foot placement strategies. At the low level of the control framework, a nonlinear whole-body controller (WBC) based on virtual constraints and a quadratic program enforces full-order dynamics and tracks the MR-NMPC references. The proposed approach is validated through extensive numerical simulations demonstrating the robust wall-assisted bipedal locomotion of a Unitree A1 quadrupedal robot on rough terrains and under external disturbances in a constrained environment. Comparative analysis shows that the proposed MR-NMPC achieves a 2.9 times higher success rate compared to conventional MPC with heuristic-based foot placement strategies in negotiating irregular terrain at high speeds.
Chinese Translation
本文提出了一种基于多速率非线性模型预测控制（MR-NMPC）的新型分层规划与控制框架，使四足机器人能够在受限环境中进行壁面辅助支持的混合双足行走。针对这种行走方式的实时轨迹优化面临重大挑战，因为控制器必须同时规划接触点以及机器人质心（CoM）和姿态的连续轨迹，同时考虑单侧接触约束、欠驱动以及机器人动力学的切换特性。在控制框架的高层，提出了一种MR-NMPC，它使用单一刚体（SRB）动力学模型动态规划接触点的离散时间轨迹和质心及姿态的连续时间轨迹。通过在多速率最优控制框架内纳入接触点规划，该方法相比于启发式足部放置策略增强了动态稳定性。在控制框架的低层，基于虚拟约束和二次规划的非线性全身控制器（WBC）强制执行全阶动力学并跟踪MR-NMPC参考轨迹。通过广泛的数值仿真验证了所提方法的有效性，展示了Unitree A1四足机器人在粗糙地形和外部干扰下的稳健壁面辅助双足行走。比较分析表明，所提MR-NMPC在以高速度穿越不规则地形时，相比于传统的基于启发式足部放置策略的MPC，其成功率提高了2.9倍。
cs.RO / 10 / 2607.01631
Path planning for unmanned naval surface vehicles
无人海面车辆的路径规划
Abstract
There nowadays is a myriad of approaches to real-time avoidance of fixed obstacles for unmanned surface vehicles (USVs) and, to a lesser extent, also the task of avoiding moving obstacles such as boats, ships, swimmers, and other USVs, but both topics still present challenges. This paper offers novel approaches to both of these problems. It uses a combination of a global path planner, which finds a path from a start point to a goal point that avoids fixed obstacles (given that their locations are known in advance), and a local path planner, which can circumnavigate a moving obstacle (as well as any previously unknown fixed obstacles). The global planner is novel in that it employs a combination of three path planners, one known in the literature as Grassfire, one that is a new modification of Grassfire, and one that is a new, and arguably more intuitive, version of the well-known Probabilistic Roadmap. The local planner is novel in that it employs a higher-level decision logic based on its observations regarding the direction of movement of the obstacle relative to the USVs global path. This logic enables the USV to determine the best strategy for avoiding the obstacle by systematically routing the vehicle behind the obstacle rather than running parallel to it until the opportunity to pass appears. Simulations are provided that validate these claims. For comparison with other systems, the simulations include an implementation of the well-known D* algorithm, and the discussion covers additional dynamic path planning systems, which, like D*, do not necessarily route the vehicle behind the moving obstacle.
Chinese Translation
目前，对于无人海面车辆（USVs）实时避开固定障碍物的方法有很多种，而对于避开移动障碍物（如船只、游泳者及其他无人海面车辆）的研究则相对较少，但这两个问题仍然存在挑战。本文提出了针对这两个问题的新方法。它结合了一个全局路径规划器，该规划器能够在已知固定障碍物位置的前提下，从起点找到通往目标点的路径，并且一个局部路径规划器，该规划器能够绕过移动障碍物（以及任何先前未知的固定障碍物）。全局规划器的创新之处在于它采用了三种路径规划器的组合，其中一种在文献中被称为Grassfire，另一种是对Grassfire的新修改，第三种则是一个新的、可以说是更直观的著名概率路线图（Probabilistic Roadmap）版本。局部规划器的创新之处在于它基于对障碍物相对于USV全局路径的运动方向的观察，采用了更高层次的决策逻辑。这种逻辑使得USV能够确定避开障碍物的最佳策略，通过系统性地将车辆引导到障碍物后方，而不是平行行驶，直到出现超车的机会。文中提供了验证这些论点的仿真结果。为了与其他系统进行比较，仿真中还包括了著名的D*算法的实现，并且讨论了其他动态路径规划系统，这些系统与D*一样，并不一定将车辆引导到移动障碍物的后方。
cs.RO / 11 / 2607.01651
One Demonstration Is Enough for Real-World Robotic Reinforcement Learning
一次演示足以实现真实世界的机器人强化学习
Abstract
Learning effective robot control policies on physical hardware is challenging due to costly data collection and the difficulty of reward specification. Prior work has incorporated demonstrations into reinforcement learning (RL), yet existing approaches either require large numbers of demonstrations or depend on continuous human intervention during training. To address these limitations, we present AutoSERL, a framework that leverages a single demonstration to fully automate the intervention process in real-world robot RL. The framework includes three complementary mechanisms to accomplish certain tasks: a sliding window intervention mechanism that continuously guides exploration to prevent local optima and unsafe deviations, a safety recovery mechanism that detects and corrects failure states via predefined trajectory recovery points, and an intervention termination criterion that automatically disables guidance once the policy can independently complete the task, preserving its exploration advantage. We evaluate AutoSERL on six contact-intensive manipulation tasks across two robot platforms, spanning insertion, hanging, and hinge-based tasks. AutoSERL consistently outperforms SERL initialized with 20 demonstrations, behavior cloning, and MILES -- a dedicated one-shot imitation learning baseline -- across all tasks while matching HIL-SERL, achieves 100% success rate on insertion tasks, and demonstrates improved robustness to positional variations, all from a single demonstration. Code and videos are available on our project website: https://autoserl.github.io/.
Chinese Translation
在物理硬件上学习有效的机器人控制策略具有挑战性，原因在于数据收集成本高和奖励规范的困难。之前的研究将演示融入了强化学习（RL），然而现有的方法要么需要大量的演示，要么在训练过程中依赖于持续的人为干预。为了解决这些限制，我们提出了AutoSERL，一个利用单次演示完全自动化真实世界机器人强化学习干预过程的框架。该框架包含三个互补机制以完成特定任务：一个滑动窗口干预机制，持续引导探索以防止局部最优和不安全偏差；一个安全恢复机制，通过预定义的轨迹恢复点检测并纠正失败状态；以及一个干预终止标准，一旦策略能够独立完成任务，自动禁用指导，从而保持其探索优势。我们在两个机器人平台上对六个接触密集的操控任务进行了AutoSERL的评估，涵盖插入、悬挂和铰链基础任务。AutoSERL在所有任务中始终优于以20次演示初始化的SERL、行为克隆和MILES（一个专门的一次性模仿学习基线），同时与HIL-SERL相匹配，在插入任务上实现了100%的成功率，并展示了对位置变化的更强鲁棒性，所有这些均来自一次演示。代码和视频可在我们的项目网站上获取：https://autoserl.github.io/
cs.RO / 12 / 2607.01684
Imagining the Sense of Touch: Touch-Informed Manipulation via Imagined Tactile Representations
想象触觉：通过想象的触觉表征进行触觉信息驱动的操作
Abstract
Tactile sensing can substantially improve contact-rich robotic manipulation, yet its practical deployment remains limited by the fragility, calibration requirements, and maintenance burden of tactile hardware. This raises a fundamental question: can robots benefit from tactile knowledge without requiring tactile sensors at deployment? We present TacImag, a tactile imagination framework that predicts tactile observations from vision and proprioception and uses the generated signals to guide manipulation policies. Trained from paired visuotactile demonstrations, TacImag enables touch-informed manipulation using only visual observations at test time. We evaluate TacImag on six simulated and four real-world manipulation tasks. Across simulation and real-world experiments, imagined tactile observations consistently improve manipulation performance without requiring tactile hardware. In real-world experiments, imagined force fields improve contact-sensitive tasks by 44.4% on average, whereas imagined tactile images improve texture-sensitive tasks by 23.3%, revealing that the effectiveness of tactile imagination depends strongly on the relationship between tactile representation and task requirements. Our results further suggest that tactile imagination does not simply recover missing tactile measurements. Instead, it acts as a form of contact-aware supervision that transforms subtle visual interaction cues into representations that are easier for manipulation policies to exploit.
Chinese Translation
触觉感知可以显著改善接触丰富的机器人操作，但其实际应用仍受到触觉硬件的脆弱性、校准要求和维护负担的限制。这引发了一个基本问题：机器人能否在不需要触觉传感器的情况下利用触觉知识？我们提出了TacImag，一个触觉想象框架，它从视觉和本体感知中预测触觉观察，并利用生成的信号指导操作策略。TacImag通过成对的视觉-触觉演示进行训练，使得在测试时仅使用视觉观察即可实现触觉信息驱动的操作。我们在六个模拟和四个真实世界的操作任务上评估了TacImag。在模拟和真实世界实验中，想象的触觉观察始终改善了操作性能，而无需触觉硬件。在真实世界实验中，想象的力场平均提高了接触敏感任务的表现44.4%，而想象的触觉图像则提高了纹理敏感任务的表现23.3%，揭示了触觉想象的有效性与触觉表征与任务要求之间的关系密切相关。我们的结果进一步表明，触觉想象并不仅仅是恢复缺失的触觉测量。相反，它作为一种接触感知的监督形式，将细微的视觉交互线索转化为更易于操作策略利用的表征。
cs.RO / 13 / 2607.01721
CoRe: Combined Rewards with Vision-Language Model Feedback for Preference-Aligned Reinforcement Learning
CoRe：结合视觉-语言模型反馈的偏好对齐强化学习的综合奖励
Abstract
Reward design remains a central challenge in reinforcement learning (RL). Hand-crafted rewards are often difficult to specify and may lead to suboptimal policies, while learned rewards from preferences can suffer from inefficiency and unstable training. Inspired by the dual nature of human learning explored in cognitive science, we decompose rewards into two complementary components: Formal Rewards (FR), explicitly designed based on task knowledge, and Residual Rewards (RR), learned from observations to capture implicit and nuanced preferences. Based on this decomposition, we propose CoRe, a hybrid framework that integrates FR and RR with vision-language models (VLMs) feedback to achieve preference-aligned policies without human involvement. Our contributions are twofold: (1) We propose a Formal Reward Module (FRM) that leverages VLMs to iteratively design and optimize FR based on task knowledge and preference feedback, enabling the continual improvement of policy during training; (2) We introduce a Residual Reward Module (RRM) that learns RR from video-level preference by employing VLMs to generate preference labels and capturing nuanced rewards that complement FR, ensuring alignment with human intent. Through the synergy of FRM and RRM, CoRe enables the automatic construction of reliable rewards that are efficient and preference-aligned. Extensive experiments demonstrate that CoRe outperforms existing approaches in terms of policy learning effectiveness and efficiency on ten robotic manipulation tasks in simulation and five real-worlds. Videos can be found on our project website: https://core-2026.github.io/
Chinese Translation
奖励设计仍然是强化学习（RL）中的一个核心挑战。手工设计的奖励往往难以明确指定，并可能导致次优策略，而从偏好中学习的奖励则可能面临效率低下和训练不稳定的问题。受到认知科学中人类学习双重性质的启发，我们将奖励分解为两个互补的组成部分：正式奖励（Formal Rewards, FR），基于任务知识明确设计，以及残余奖励（Residual Rewards, RR），通过观察学习以捕捉隐含和细微的偏好。基于这种分解，我们提出了CoRe，一个混合框架，结合FR和RR与视觉-语言模型（Vision-Language Models, VLMs）反馈，以实现无需人工干预的偏好对齐策略。我们的贡献有两个方面：（1）我们提出了一个正式奖励模块（Formal Reward Module, FRM），利用VLMs基于任务知识和偏好反馈迭代设计和优化FR，使得策略在训练过程中能够持续改进；（2）我们引入了一个残余奖励模块（Residual Reward Module, RRM），通过使用VLMs生成偏好标签，从视频级偏好中学习RR，捕捉补充FR的细微奖励，确保与人类意图的一致性。通过FRM和RRM的协同作用，CoRe能够自动构建高效且偏好对齐的可靠奖励。大量实验表明，CoRe在十个仿真中的机器人操作任务和五个真实世界任务上，在策略学习的有效性和效率方面优于现有方法。视频可以在我们的项目网站上找到：https://core-2026.github.io/
cs.RO / 14 / 2607.01794
Lightweight Safe Reinforcement Learning for End-to-End UAV Navigation
轻量级安全强化学习用于端到端无人机导航
Abstract
With the rapid development of autonomous aerial systems, Unmanned Aerial Vehicles (UAVs) are increasingly deployed in applications such as inspection, environmental monitoring, and rescue, creating growing demand for reliable autonomous navigation. However, autonomous UAV navigation in dense environments remains challenging under sparse perception and dynamic constraints. Most reinforcement learning (RL) methods lack explicit safety mechanisms, leading to unsafe exploration, unstable training, and risky behaviors, especially during high-speed flight. Even in safe RL approaches, safety is often enforced by projecting policy outputs onto a safe action set, which may introduce instability. Meanwhile, many learning-based methods rely on dense inputs or large networks, increasing computational burden and limiting lightweight onboard deployment. Facing the above challenges, we propose a safety-constrained perception-control integrated framework for UAV navigation. A lightweight network encodes sparse observations into collision-risk-aware features using asymmetric and depthwise separable convolutions. We formulate the task as a constrained Markov decision process within a hierarchical control architecture and solve it using a Lagrangian-based safe PPO algorithm. Curriculum learning further improves training stability. Experiments with varying obstacle densities and flight speeds demonstrate higher success rates, improved safety, and better efficiency than existing reinforcement learning baselines.
Chinese Translation
随着自主航空系统的快速发展，无人机（UAV）在检查、环境监测和救援等应用中的部署日益增多，可靠的自主导航需求也在不断增长。然而，在稀疏感知和动态约束下，密集环境中的自主无人机导航仍然面临挑战。大多数强化学习（RL）方法缺乏明确的安全机制，导致不安全的探索、不稳定的训练和高风险行为，尤其是在高速飞行期间。即使在安全强化学习方法中，安全性通常通过将策略输出投影到安全动作集上来强制执行，这可能引入不稳定性。同时，许多基于学习的方法依赖于密集输入或大型网络，增加了计算负担并限制了轻量级机载部署。针对上述挑战，我们提出了一种安全约束的感知-控制集成框架用于无人机导航。一个轻量级网络利用非对称和深度可分离卷积将稀疏观测编码为碰撞风险感知特征。我们将任务形式化为一个层次控制架构下的约束马尔可夫决策过程，并使用基于拉格朗日的安全PPO算法进行求解。课程学习进一步提高了训练的稳定性。不同障碍物密度和飞行速度的实验表明，与现有的强化学习基线相比，我们的方法具有更高的成功率、改善的安全性和更好的效率。
cs.RO / 15 / 2607.01804
VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon
VLA-Corrector：轻量级的检测与修正推理用于自适应动作视野
Abstract
Vision-Language-Action (VLA) foundation models have recently achieved strong progress in embodied intelligence. To reduce policy-call frequency while preserving temporal coherence, most generative policies adopt an action chunk mechanism, executing multiple future actions in an open-loop manner under a fixed action horizon. However, this "predict-then-blindly-execute" paradigm sacrifices closed-loop reactivity: in contact-rich physical interactions, even small local perturbations can rapidly amplify within the open-loop blind spot, leading to compounding errors and ultimately task failure. To address this limitation, we propose VLA-Corrector, a lightweight corrective inference framework for action-chunked VLA policies. Without modifying the backbone policy weights, VLA-Corrector introduces a lightweight Latent-space Vision Monitor (LVM) that continuously compares predicted and actual visual feature evolution, enabling online detection of visual dynamics deviations. Once persistent deviation is detected, the system triggers a truncation event, discards the remaining stale actions, and invokes corrective replanning via Online Gradient Guidance (OGG). The detect-and-correct mechanism of VLA-Corrector naturally induces an event-triggered adaptive action horizon: it preserves long-horizon execution when the current chunk remains reliable, and invokes short-horizon corrective replanning when execution begins to drift. In doing so, VLA-Corrector mitigates the trade-off imposed by static horizons between execution robustness and policy-call frequency. It can be integrated into different VLA models without further retraining the VLA backbone, interrupting compounding errors while preserving much of the efficiency benefit of action chunking and substantially improving robustness in long-horizon, contact-rich robotic manipulation tasks.
Chinese Translation
视觉-语言-动作（VLA）基础模型最近在具身智能方面取得了显著进展。为了减少策略调用频率，同时保持时间一致性，大多数生成策略采用动作块机制，在固定的动作视野下以开放循环的方式执行多个未来动作。然而，这种“预测-然后盲目执行”的范式牺牲了闭环反应性：在接触丰富的物理交互中，即使是小的局部扰动也会在开放循环的盲区内迅速放大，导致累积错误并最终导致任务失败。为了解决这一限制，我们提出了VLA-Corrector，一种轻量级的修正推理框架，适用于动作块化的VLA策略。在不修改基础策略权重的情况下，VLA-Corrector引入了一个轻量级的潜在空间视觉监测器（Latent-space Vision Monitor, LVM），该监测器持续比较预测的视觉特征演变与实际演变，从而实现在线检测视觉动态偏差。一旦检测到持续偏差，系统会触发截断事件，丢弃剩余的过时动作，并通过在线梯度引导（Online Gradient Guidance, OGG）调用修正重规划。VLA-Corrector的检测与修正机制自然引入了事件触发的自适应动作视野：当当前块保持可靠时，它保留长视野执行；而当执行开始偏离时，则调用短视野的修正重规划。通过这样做，VLA-Corrector减轻了静态视野在执行稳健性与策略调用频率之间的权衡。它可以集成到不同的VLA模型中，而无需进一步重新训练VLA基础模型，从而中断累积错误，同时保留动作块化的高效性优势，并显著提高在长视野、接触丰富的机器人操作任务中的稳健性。
cs.RO / 16 / 2607.01860
DL-SLAM: Enabling High-Fidelity Gaussian Splatting SLAM in Dynamic Environments based on Dual-Level Probability
DL-SLAM：基于双层概率的动态环境中高保真高斯点云SLAM的实现
Abstract
Recent advances in 3D Gaussian Splatting (3DGS) have enabled significant progress in dense dynamic Simultaneous Localization And Mapping (SLAM). Prevailing methods typically discard predefined dynamic objects, ignoring that transiently static objects offer valuable geometric constraints for pose estimation. A recent work attempts to leverage this potential by employing per-pixel uncertainty maps to quantify the magnitude of motion. While this approach enables transiently static objects to enhance pose estimation, it erroneously integrates these objects into the static map, resulting in persistent artifacts. Moreover, its reliance on purely geometric information leads to ambiguous object boundaries in the uncertainty maps. To overcome these limitations, we present DL-SLAM, a monocular Gaussian Splatting SLAM system built upon a novel dual-level probabilistic framework. Our method computes dynamic probability maps by combining semantic and geometric information. These pixel-level probabilities are lifted to 3D and aggregated to derive an object-level dynamic probability for each instance. Object-level probability enables the categorical pruning of dynamic Gaussians, resulting in an artifact-free static map. The static map, in turn, provides a geometrically consistent guidance to refine the pixel-wise probabilities, enhancing their reliability. Experimental results demonstrate that DL-SLAM outperforms existing approaches, improving tracking accuracy by up to 13\% while generating high-fidelity semantic maps.
Chinese Translation
近年来，3D高斯点云（3DGS）的进展使得密集动态同时定位与地图构建（SLAM）取得了显著进展。现有方法通常会忽略预定义的动态物体，忽视了瞬态静态物体为姿态估计提供的宝贵几何约束。最近的研究尝试通过使用每像素不确定性图来量化运动幅度，从而利用这一潜力。虽然这种方法使得瞬态静态物体能够增强姿态估计，但它错误地将这些物体整合到静态地图中，导致持续的伪影。此外，其对纯几何信息的依赖导致不确定性图中的物体边界模糊。为了克服这些局限性，我们提出了DL-SLAM，一种基于新颖双层概率框架的单目高斯点云SLAM系统。我们的方法通过结合语义和几何信息计算动态概率图。这些像素级概率被提升到三维并聚合，以推导出每个实例的物体级动态概率。物体级概率使得动态高斯的分类修剪成为可能，从而生成无伪影的静态地图。静态地图反过来为细化像素级概率提供了几何一致的指导，增强了其可靠性。实验结果表明，DL-SLAM优于现有方法，跟踪精度提高了多达13\%，同时生成高保真的语义地图。
cs.RO / 17 / 2607.01925
SPLC: Social Preference Learning for Crowd Robot Navigation
SPLC：用于人群机器人导航的社会偏好学习
Abstract
Offline reinforcement learning (RL) holds significant potential for crowd robot navigation in human-robot coexistence applications. However, the inherent complexity of pedestrian motion renders the design of effective reward functions for promoting socially compliant robot behaviors a persistent challenge. This paper proposes a Social Preference Learning for Crowd Robot Navigation (SPLC) algorithm to eliminate the need for detailed reward design. Its core innovation lies in the introduction of a social preference feedback mechanism to automatically generate preference data through principled preference evaluation criteria. By explicitly accounting for the intricacies of pedestrian dynamics, the pipeline mitigates the reward bias and facilitates the systematic quantification of broad social norms, thereby fostering socially compliant behaviors. Extensive experiments integrating SPLC with offline RL methods demonstrate consistent improvements over state-of-the-art baselines across standard performance metrics. Furthermore, real-world experiments on the TurtleBot4 further validate the effectiveness of SPLC in practical human-robot coexistence settings. Our code and video demos are available at https://github.com/sklus949/SPLC.
Chinese Translation
离线强化学习（RL）在人与机器人共存应用中的人群机器人导航中具有重要潜力。然而，行人运动的内在复杂性使得设计有效的奖励函数以促进社会合规的机器人行为成为一个持续的挑战。本文提出了一种用于人群机器人导航的社会偏好学习（SPLC）算法，以消除对详细奖励设计的需求。其核心创新在于引入了一种社会偏好反馈机制，通过原则性的偏好评估标准自动生成偏好数据。通过明确考虑行人动态的复杂性，该流程减轻了奖励偏差，并促进了广泛社会规范的系统量化，从而培养社会合规行为。将SPLC与离线RL方法结合的广泛实验表明，在标准性能指标上相较于最先进的基线方法具有一致的改进。此外，在TurtleBot4上的现实世界实验进一步验证了SPLC在实际人机共存环境中的有效性。我们的代码和视频演示可在 https://github.com/sklus949/SPLC 获取。
cs.RO / 18 / 2607.01938
PhysMani: Physics-principled 3D World Model for Dynamic Object Manipulation
PhysMani：基于物理原理的动态物体操控三维世界模型
Abstract
Manipulating fast and dynamically moving targets in unstructured 3D environments remains challenging for embodied AI. Existing visual-language-action models and world models struggle with accurate 3D geometry and physically meaningful forecasting. We propose PhysMani, a framework that couples a physics-principled 3D Gaussian world model with a future-aware action policy model. The world model learns a divergence-free Gaussian velocity field via online optimization for fast and physically grounded future dynamics prediction. The policy model integrates the predicted 3D scene future dynamics through a learnable token based cross-attention module. We introduce PhysMani-Bench, a dynamic manipulation benchmark with 16 tasks, and demonstrate a superior success rate over strong baselines in both simulation and real-world robot experiments.
Chinese Translation
在非结构化的三维环境中操控快速且动态移动的目标对于具身人工智能仍然具有挑战性。现有的视觉-语言-动作模型和世界模型在准确的三维几何建模和物理意义上的预测方面存在困难。我们提出了PhysMani，一个将基于物理原理的三维高斯世界模型与未来感知的动作策略模型相结合的框架。该世界模型通过在线优化学习无散度的高斯速度场，以实现快速且基于物理的未来动态预测。策略模型通过可学习的基于令牌的交叉注意力模块整合预测的三维场景未来动态。我们引入了PhysMani-Bench，一个包含16个任务的动态操控基准，并在模拟和真实世界机器人实验中展示了相较于强基线的优越成功率。
cs.RO / 19 / 2607.02005
A Stereo Visual SLAM System Using Object-Level Motion Estimation and Geometric Filtering Based on Cross Disparity
基于交叉视差的对象级运动估计与几何过滤的立体视觉SLAM系统
Abstract
This paper presents OCD SLAM, a dynamic stereo visual SLAM framework that extends ORB-SLAM2 by jointly addressing dynamic objects and dynamic features in the scene. Usual visual SLAM systems operating in dynamic environments often fail in the presence of moving objects, due to the static-world assumption used in pose estimation and mapping. To address this predicament, we introduce a novel geometric approach based on the discrepancy between disparity and a newly proposed notion called ``cross disparity'', which exploits both temporal and stereo inconsistency to identify dynamic feature points. Complementary to this feature-level motion analysis, OCD SLAM integrates a 3D object detection module (SMOKE) with Kalman filter-based object tracking to perform object-level motion classification, enabling robust separation of static and dynamic scene elements for accurate pose estimation. The proposed approach has been evaluated on various sequences from the KITTI Odometry and KITTI Raw datasets. Results demonstrate that OCD SLAM achieves significant improvement in trajectory accuracy compared to ORB-SLAM2 and several state-of-the-art dynamic SLAM methods. Ablation studies further demonstrate the effectiveness of the cross disparity module in the KITTI Raw dataset and show that this method is able to detect dynamic features that are missed by the 3D object detection scheme alone.
Chinese Translation
本文提出了OCD SLAM，一个动态立体视觉SLAM框架，通过共同处理场景中的动态对象和动态特征，扩展了ORB-SLAM2。通常在动态环境中运行的视觉SLAM系统在存在移动对象时往往会失败，这是由于在姿态估计和映射中使用了静态世界假设。为了解决这一困境，我们引入了一种基于视差与新提出的“交叉视差”概念之间差异的新几何方法，该方法利用时间和立体不一致性来识别动态特征点。作为对这种特征级运动分析的补充，OCD SLAM集成了一个基于卡尔曼滤波的3D对象检测模块（SMOKE）与对象跟踪，以执行对象级运动分类，从而实现静态和动态场景元素的稳健分离，以便进行准确的姿态估计。所提出的方法在KITTI Odometry和KITTI Raw数据集的多个序列上进行了评估。结果表明，与ORB-SLAM2和几种最先进的动态SLAM方法相比，OCD SLAM在轨迹精度上取得了显著改善。消融研究进一步证明了交叉视差模块在KITTI Raw数据集中的有效性，并显示该方法能够检测到仅通过3D对象检测方案无法捕捉到的动态特征。
cs.RO / 20 / 2607.02037
Cross-Platform Control for Autonomous Surface Vehicles via Adaptive Reinforcement Learning
通过自适应强化学习实现自主水面车辆的跨平台控制
Abstract
Autonomous surface vehicles vary widely in hydrodynamic and actuation characteristics, yet most controllers are designed for single-platform deployment. We present an adaptive reinforcement learning approach for trajectory tracking that enables zero-shot cross-platform deployment using a single policy. Since the deployment platform's dynamics are unknown to the policy, we address cross-platform generalization with the standard partial-observability approach of conditioning on interaction history, employing a teacher-student architecture in which a learned module infers a latent representation of the platform dynamics. The policy is trained in simulation under randomized vessel dynamics and is deployed zero-shot to two real-world platforms without any fine-tuning, despite relying on a simple analytical dynamics model rather than a high-fidelity hydrodynamic simulator. In real-world experiments on two different platforms, the adaptive policy outperforms non-adaptive learning-based baselines by up to 58% in position mean absolute error while approaching the tracking accuracy of a platform-specific tuned controller.
Chinese Translation
自主水面车辆在水动力和驱动特性上差异很大，但大多数控制器都是为单一平台部署而设计的。我们提出了一种自适应强化学习方法，用于轨迹跟踪，能够通过单一策略实现零样本跨平台部署。由于部署平台的动态对策略而言是未知的，我们采用标准的部分可观测性方法，通过条件化交互历史来解决跨平台泛化问题，采用教师-学生架构，其中学习模块推断平台动态的潜在表示。该策略在随机化船舶动态的仿真环境中进行训练，并在没有任何微调的情况下，零样本部署到两个真实平台，尽管依赖于简单的解析动态模型，而非高保真水动力模拟器。在两个不同平台的真实实验中，自适应策略在位置均值绝对误差上比非自适应学习基线提高了多达58%，同时接近平台特定调优控制器的跟踪精度。
cs.RO / 21 / 2607.02092
Guided Action Flow: Q-Guided Inference for Flow-Matching Vision-Language-Action Policies
引导动作流：基于Q的流匹配视觉-语言-动作策略推理
Abstract
Flow-matching vision-language-action policies generate robot action chunks through an iterative transport process, creating an opportunity for test-time guidance without retraining the base policy. We study this opportunity in Guided Action Flow, an inference-time framework that keeps a pretrained SmolVLA policy frozen and uses a learned action-chunk critic to guide its reverse-time flow sampler. The critic is trained from real success and failure rollouts, can condition on task-description features from the frozen SmolVLA language pathway, and is used only through action gradients during sampling. We evaluate the approach on LIBERO manipulation tasks. A single-task critic improves success from 68.0% to 82.0% on one seed window and from 82.0% to 86.0% on another. A multi-family task-description critic improves validation success from 46.0% to 56.0%, while the locked held-out test gain is positive but modest, from 65.0% to 67.5%. These results support the feasibility of Q-guided inference for frozen flow-matching VLA policies, while showing that critic generalization and uncertainty-aware guidance remain the central bottlenecks.
Chinese Translation
流匹配视觉-语言-动作策略通过迭代传输过程生成机器人动作块，为测试时引导提供了机会，而无需重新训练基础策略。我们在引导动作流（Guided Action Flow）中研究了这一机会，这是一种推理时框架，保持预训练的SmolVLA策略不变，并使用学习到的动作块评论员来引导其逆时间流采样器。评论员是通过真实的成功和失败回放进行训练的，可以基于冻结的SmolVLA语言路径中的任务描述特征进行条件化，并且在采样过程中仅通过动作梯度使用。我们在LIBERO操作任务上评估了该方法。单任务评论员在一个种子窗口上的成功率从68.0%提高到82.0%，在另一个窗口上从82.0%提高到86.0%。多家庭任务描述评论员将验证成功率从46.0%提高到56.0%，而锁定的保留测试增益虽然积极但适度，从65.0%提高到67.5%。这些结果支持了对冻结流匹配VLA策略进行Q引导推理的可行性，同时表明评论员的泛化能力和不确定性感知引导仍然是主要瓶颈。
cs.RO / 22 / 2607.02174
Choreographing the Way of Water: A Computational Framework for Aquatic Robotic Art
水之编舞：水下机器人艺术的计算框架
Abstract
Robotic choreography in open water is governed by nonlinear fluid dynamics, which impose significant challenges due to environmental disturbances and nonlinear system dynamics. This paper presents the cyber-physical architecture of Way of Water, a vertically integrated framework that orchestrates a fleet of autonomous surface vessels as a distributed choreographic platform. Moving beyond the surface-pixel paradigm, these vessels use laminar nozzles and multi-zone lighting to extend their expressive range from the 2D water plane into the 3D volumetric domain. Our primary contribution is the Way of Water Studio, a browser-based, timeline-compositing authoring paradigm that treats the fleet as a DAW-like instrument for music-responsive choreography. The Studio encapsulates Sequential Convex Programming for trajectory generation and Model Predictive Control for disturbance rejection presented through a visual timeline, broadening access to high-performance aquatic robotics for non-programmer artists. Grounding the Studio is the full cyber-physical stack: a custom holonomic chassis, a state-estimation and control stack tuned for the aquatic domain, and an LTE/MQTT fleet link with RTK-GPS time synchronization. We report on the system's validation across two distinct deployments: an 18-vessel Swan Lake interpretation at Lake Zurich and an 8-vessel Time Space Existence 2025 Venice Biennale demonstration at Forte Marghera, establishing a foundational reference for the design and deployment of fluidic robotic swarms.
Chinese Translation
开放水域中的机器人编舞受非线性流体动力学的支配，这给环境干扰和非线性系统动态带来了重大挑战。本文提出了水之道的网络物理架构，这是一个垂直集成的框架，协调一支自主表面船只的舰队，作为一个分布式编舞平台。超越表面像素范式，这些船只利用层流喷嘴和多区域照明将其表现范围从二维水面扩展到三维体积域。我们的主要贡献是水之道工作室，这是一个基于浏览器的时间线合成创作范式，将舰队视为类似数字音频工作站（DAW）的乐器，用于音乐响应的编舞。工作室封装了用于轨迹生成的顺序凸优化（Sequential Convex Programming）和用于干扰抑制的模型预测控制（Model Predictive Control），通过可视化时间线呈现，拓宽了非程序员艺术家对高性能水下机器人技术的访问。工作室的基础是完整的网络物理堆栈：一个定制的全向底盘、为水域调优的状态估计和控制堆栈，以及一个具有RTK-GPS时间同步的LTE/MQTT舰队链接。我们报告了系统在两个不同部署中的验证：在苏黎世湖进行的18艘船只的天鹅湖（Swan Lake）演绎和在马尔盖拉堡（Forte Marghera）进行的8艘船只的时间空间存在2025威尼斯双年展（Time Space Existence 2025）演示，为流体机器人群的设计和部署建立了基础参考。
cs.RO / 23 / 2607.02195
Bridge-WA: Predicting Where and How the World Changes for Robotic Action
Bridge-WA：预测世界变化的地点和方式以实现机器人行动
Abstract
General-purpose vision-language-action models benefit from large vision-language priors, but effective manipulation also requires anticipating action-relevant scene changes. Existing world-action models often rely on large generative world models or dense future rollouts, which are expensive and spend capacity on visual details weakly coupled to control. We present Bridge-WA, a lightweight world-action framework that distills a frozen future-change teacher into three compact priors: future tokens for intended outcomes, change maps for intervention support, and motion-flow maps for local transition direction. A WorldBridge conditions the action transformer on these priors through multi-source attention memories and spatial-temporal biases, while the teacher model is removed at inference. Across VLABench, RoboTwin2.0, LIBERO-Plus and real-robot evaluations, Bridge-WA improves task success, progress, and robustness, with particularly clear gains under out-of-distribution visual shifts. By focusing action generation on where and how the scene will change, Bridge-WA suppresses nuisance appearance factors such as background, lighting, and distractors, leading to better generalization without deployment-time dense future-image generation. Code and visualizations are available at: https://hcplab-sysu.github.io/BRIDGE-WA .
Chinese Translation
通用视觉-语言-行动模型受益于大规模的视觉-语言先验，但有效的操作还需要预测与行动相关的场景变化。现有的世界-行动模型通常依赖于大型生成世界模型或密集的未来展开，这些方法成本高昂，并且在视觉细节上消耗了与控制弱相关的能力。我们提出了Bridge-WA，一个轻量级的世界-行动框架，它将一个冻结的未来变化教师模型提炼为三个紧凑的先验：用于预期结果的未来标记、用于干预支持的变化图以及用于局部过渡方向的运动流图。WorldBridge通过多源注意力记忆和时空偏置对这些先验条件化行动变换器，而在推理时移除了教师模型。在VLABench、RoboTwin2.0、LIBERO-Plus和真实机器人评估中，Bridge-WA提高了任务成功率、进展和鲁棒性，特别是在分布外视觉变化下表现出明显的提升。通过将行动生成聚焦于场景将如何变化，Bridge-WA抑制了背景、光照和干扰物等无关外观因素，从而在不需要在部署时进行密集未来图像生成的情况下实现更好的泛化。代码和可视化结果可在以下网址获取：https://hcplab-sysu.github.io/BRIDGE-WA 。
cs.RO / 24 / 2607.02205
Actuator Reality Shaping for Zero-Shot Sim-to-Real Robot Learning
零-shot模拟到真实机器人学习的执行器现实塑造
Abstract
Sim-to-real transfer in robot learning is often limited by discrepancies between the ideal actuator dynamics assumed during policy training and the nonlinear, hardware-dependent behavior of physical motors. While conventional approaches attempt to bridge this gap by increasing simulator fidelity through system identification, domain randomization, or learned actuator models, we introduce an alternative paradigm: actuator reality shaping. Instead of modifying the simulator to match the real world, our method shapes the closed-loop behavior of physical actuators to match the idealized second-order reference dynamics used in simulation. By equipping each joint with a two-degree-of-freedom feedforward--feedback controller, we decouple reference-response shaping from robust stabilization, thereby providing a standardized actuator interface for reinforcement learning policies. As a result, policies trained only with the prescribed reference model can be deployed zero-shot on real hardware without task-level fine-tuning or learned actuator models. We validate the approach on a single-joint high-gear-ratio servo under external loads and a 7-DOF robotic arm reaching task, where actuator reality shaping substantially reduces sim-to-real tracking error and improves zero-shot task performance compared with standard servo-control and representative real-to-sim-to-real baselines. We further demonstrate zero-shot transfer on a wheeled-legged robot driving over a slope and a humanoid robot walking, suggesting that actuator reality shaping can serve as a reusable interface for robot learning across diverse hardware platforms.
Chinese Translation
在机器人学习中，模拟到真实的迁移常常受到政策训练期间假设的理想执行器动态与物理电机的非线性、依赖硬件的行为之间差异的限制。虽然传统方法试图通过系统识别、领域随机化或学习的执行器模型来提高模拟器的逼真度以弥补这一差距，但我们提出了一种替代范式：执行器现实塑造。我们的方法不是修改模拟器以匹配真实世界，而是塑造物理执行器的闭环行为，以匹配在模拟中使用的理想化二阶参考动态。通过为每个关节配备一个具有两自由度的前馈-反馈控制器，我们将参考响应塑造与鲁棒稳定性解耦，从而为强化学习政策提供标准化的执行器接口。因此，仅使用规定的参考模型训练的政策可以在真实硬件上零-shot部署，而无需任务级微调或学习的执行器模型。我们在一个单关节高齿比伺服电机在外部负载下以及一个7自由度机器人臂的到达任务上验证了该方法，其中执行器现实塑造显著减少了模拟到真实的跟踪误差，并改善了与标准伺服控制和代表性的真实到模拟到真实基线相比的零-shot任务性能。我们进一步展示了在一个轮式腿机器人在坡道上行驶和一个类人机器人行走的零-shot迁移，表明执行器现实塑造可以作为跨多种硬件平台的机器人学习的可重用接口。
cs.RO / 25 / 2607.02222
CoFL-S: Spatially Queryable Sector Flow Fields for Local Language-Conditioned Navigation
CoFL-S：可空间查询的局部语言条件导航流场
Abstract
Vision-Language Navigation has increasingly emphasized high-level instruction reasoning, memory, global map construction, and instruction decomposition, while the low-level action representation remains comparatively underexplored. We propose CoFL-S, a low-level vision-language-action framework that predicts a language-conditioned flow field over the robot's local visible sector and generates continuous trajectories by rolling out the predicted field. To train this low-level representation, we convert each VLN-CE episode, originally a whole-episode instruction paired with an action sequence, into frame-level local supervision with aligned sub-instructions and matched action, trajectory, and dense flow-field targets. For evaluation, we introduce a continuous-time Habitat benchmark that isolates low-level action interfaces from instruction decomposition and executes all methods through a shared velocity-command controller, enabling decomposition-independent closed-loop comparison across different planner frequencies rather than fixed discrete forward-and-turn transitions in VLN-CE. Under matched encoders and training settings, CoFL-S consistently outperforms action-token and action-chunk baselines across planner frequencies in the continuous-time Habitat benchmark, and zero-shot real-world closed-loop deployment further shows its advantage over both baselines beyond simulation.
Chinese Translation
视觉-语言导航越来越强调高层次的指令推理、记忆、全球地图构建和指令分解，而低层次的动作表示相对较少被探索。我们提出了CoFL-S，一个低层次的视觉-语言-动作框架，该框架预测机器人局部可见区域的语言条件流场，并通过展开预测的流场生成连续轨迹。为了训练这种低层次表示，我们将每个VLN-CE（视觉-语言导航挑战）剧集，原本是与动作序列配对的整个剧集指令，转换为带有对齐子指令和匹配动作、轨迹及密集流场目标的帧级局部监督。为了评估，我们引入了一个连续时间Habitat基准，该基准将低层次动作接口与指令分解隔离，并通过共享速度命令控制器执行所有方法，从而实现不同规划频率下的分解无关闭环比较，而不是在VLN-CE中固定离散的前进和转向过渡。在匹配的编码器和训练设置下，CoFL-S在连续时间Habitat基准中始终优于动作令牌和动作块基线，并且零-shot真实世界闭环部署进一步显示其在模拟之外相较于两个基线的优势。
cs.RO / 26 / 2607.02277
NEUROSYMLAND: Neuro-Symbolic Landing-Site Assessment for Robust and Edge-Deployable UAV Autonomy
NEUROSYMLAND：用于鲁棒和边缘可部署无人机自主性的神经符号着陆点评估
Abstract
Safe landing-site assessment in unstructured environments remains a key challenge for autonomous UAV deployment, as vision-only learning approaches often degrade under terrain variability and provide limited transparency in safety decisions. We present NEUROSYMLAND, a neuro-symbolic landing-site assessment system that integrates lightweight perception with explicit safety reasoning. The framework constructs a probabilistic semantic scene graph from onboard visual input and evaluates candidate landing regions using symbolic constraints capturing terrain flatness, obstacle clearance, and spatial consistency, enabling structured reasoning under perceptual uncertainty while maintaining edge-feasible execution. Across 72 simulated landing scenarios spanning diverse terrains, NEUROSYMLAND achieves 61 successful assessments, outperforming four competitive baselines (37-57 successes). To evaluate deployability, we further conduct 100 hardware-in-the-loop trials with randomized initial poses, profiling end-to-end latency, stage-wise execution time, and system-level metrics including CPU/GPU utilization, memory footprint, and power consumption. Results demonstrate improved robustness and interpretability with bounded edge-resource usage. Profiling shows that symbolic reasoning contributes only a small fraction of end-to-end latency, while the main computational cost arises from perception and PSSG construction. These results demonstrate the feasibility of deploying the landing-site assessment stack on edge-constrained UAV hardware, and all source code, datasets, prompts, and symbolic rule refinement examples are released in an open-source repository
Chinese Translation
在非结构化环境中进行安全着陆点评估仍然是无人机自主部署的一个关键挑战，因为仅依赖视觉的学习方法在地形变化下往往会降低性能，并且在安全决策中提供有限的透明度。我们提出了NEUROSYMLAND，一个神经符号着陆点评估系统，它将轻量级感知与明确的安全推理相结合。该框架从机载视觉输入构建一个概率语义场景图，并利用符号约束评估候选着陆区域，这些约束捕捉地形平坦性、障碍物清除和空间一致性，从而在感知不确定性下实现结构化推理，同时保持边缘可行的执行。在72个涵盖多样地形的模拟着陆场景中，NEUROSYMLAND成功实现了61次评估，超越了四个竞争基线（成功次数为37-57）。为了评估可部署性，我们进一步进行了100次硬件在环试验，随机初始姿态，分析了端到端延迟、各阶段执行时间以及包括CPU/GPU利用率、内存占用和功耗在内的系统级指标。结果表明，在有限的边缘资源使用下，鲁棒性和可解释性得到了改善。分析显示，符号推理仅占端到端延迟的一小部分，而主要的计算成本来自感知和PSSG构建。这些结果证明了在边缘受限的无人机硬件上部署着陆点评估堆栈的可行性，所有源代码、数据集、提示和符号规则细化示例均已在一个开源库中发布。
cs.RO / 27 / 2607.02298
Real-Time Visual Intelligence on Low-Cost UAVs: A Modular Approach for Tracking, Scanning, and Navigation
低成本无人机上的实时视觉智能：跟踪、扫描和导航的模块化方法
Abstract
Autonomous drones are rapidly transforming modern warfare and civil applications alike. This paper presents the development of an integrated intelligent drone system designed to serve as a personal assistant. Leveraging the DJI Tello drone platform, we implemented a modular architecture that integrates three core artificial intelligence functionalities: facial detection, facial recognition, and depth estimation from monocular vision. A web-based interface enables seamless drone control and real-time video monitoring, while a Python-based server processes visual data and executes inference pipelines using lightweight neural models optimized for embedded systems. Unlike existing commercial solutions, this system emphasizes accessibility, low-cost hardware, and open-source technologies. The system demonstrates robust performance in real-world conditions, including person tracking, indoor scanning, and autonomous line following using virtual sensors. This project validates the applicability of advanced AI techniques in real-time robotic systems and illustrates the feasibility of deploying them on constrained hardware, providing a foundation for future research in autonomous UAVs for military, rescue, and surveillance missions.
Chinese Translation
自主无人机正在迅速改变现代战争和民用应用。本文介绍了一种集成智能无人机系统的开发，旨在作为个人助理。我们利用DJI Tello无人机平台，实施了一个模块化架构，集成了三项核心人工智能功能：人脸检测、人脸识别和单目视觉深度估计。基于网页的界面实现了无缝的无人机控制和实时视频监控，而基于Python的服务器处理视觉数据，并使用针对嵌入式系统优化的轻量级神经模型执行推理管道。与现有的商业解决方案不同，该系统强调可获取性、低成本硬件和开源技术。该系统在现实条件下展示了强大的性能，包括人跟踪、室内扫描和使用虚拟传感器的自主线路跟随。该项目验证了先进人工智能技术在实时机器人系统中的适用性，并展示了在受限硬件上部署这些技术的可行性，为未来在军事、救援和监视任务中使用自主无人机的研究奠定了基础。
cs.RO / 28 / 2607.02322
The Moving Eye: Enhancing VLA Spatial Generalization via Hybrid Dynamic Data Collection
动态视角：通过混合动态数据收集增强视觉-语言-动作（VLA）空间泛化能力
Abstract
Vision-Language-Action (VLA) models have shown remarkable promise in generalized robotic manipulation. However, their spatial generalization remains fragile. We argue that simply increasing the number of viewpoints is insufficient. Models often fall into the trap of Shortcut Learning, latching onto spurious correlations (e.g., fixed relative poses between objects or between the camera and robot base) rather than learning true spatial relationships. In this work, we propose a data-centric solution to enhance VLA spatial generalization. We utilize a dual-arm setup where one arm performs manipulation while the other serves as a mobile environmental camera. We systematically evaluate three data distribution patterns: Fixed, Multi-Fixed, and Moving Views. Our findings reveal that a hybrid strategy, combining continuous camera motion with diverse static viewpoints, yields the best performance by substantially reducing spurious correlations while maintaining training stability. Our experiments demonstrate that this strategy mitigates spurious correlations, enabling VLAs to generalize to unseen camera poses and object configurations where simply adding more static viewpoints fails. Crucially, we reveal that the susceptibility to shortcut learning and the struggle with spatial generalization are universal characteristics shared across diverse architectures. Consequently, all evaluated models (ACT, Diffusion, and VLA models including Pi0 and Gr00t) benefit significantly from our mixed data strategy.
Chinese Translation
视觉-语言-动作（VLA）模型在广义机器人操作中展现了显著的潜力。然而，它们的空间泛化能力仍然脆弱。我们认为，仅仅增加视角数量是不够的。模型往往陷入捷径学习的陷阱，依赖于虚假的相关性（例如，物体之间或相机与机器人基座之间的固定相对姿态），而不是学习真实的空间关系。在本研究中，我们提出了一种以数据为中心的解决方案，以增强VLA的空间泛化能力。我们利用双臂设置，其中一只手臂执行操作，而另一只手臂作为移动环境相机。我们系统地评估了三种数据分布模式：固定视角、多固定视角和移动视角。我们的研究结果表明，结合持续相机运动与多样静态视角的混合策略能够显著提高性能，显著减少虚假相关性，同时保持训练的稳定性。我们的实验表明，这一策略减轻了虚假相关性，使VLA能够在未见过的相机姿态和物体配置中进行泛化，而仅仅增加更多静态视角则无法实现。重要的是，我们揭示了对捷径学习的敏感性和空间泛化的困难是不同架构普遍存在的特征。因此，所有评估的模型（包括ACT、Diffusion以及VLA模型如Pi0和Gr00t）都从我们的混合数据策略中显著受益。
cs.RO / 29 / 2607.02332
HEFT: Heavy-Payload Full-size Humanoid Teleoperation with Privileged Motion Guidance and Windowed Payload Curriculum
HEFT：具特权运动引导和窗口化负载课程的重载全尺寸类人机器人遥操作
Abstract
General motion tracking and teleoperation offer a promising path to scalable humanoid skill acquisition, yet most existing frameworks are validated on compact platforms or without real payload interaction, leaving full-size humanoids with real payloads largely unexplored. Scaling to full-size humanoids introduces two compounding challenges: their larger inertia and tighter balance margins make tracking highly sensitive to noise, drift, and retargeting errors from commodity VR trackers, while their payload potential remains largely underutilized. We present HEFT, a heavy-payload full-size humanoid teleoperation framework that addresses both challenges. HEFT learns from deployable noisy VR references with physically plausible reconstructed references through Privileged Motion Guidance (PMG), and uses a Windowed Payload Curriculum (WPC) with expert-guided payload caps to acquire robust heavy-payload tracking. We deploy HEFT on L7, a 175cm, 65kg humanoid. The robot tracks motions including turns, forward/backward locomotion, and squats under payloads up to 24kg.
Chinese Translation
一般运动跟踪和遥操作为可扩展的类人技能获取提供了有前景的路径，但大多数现有框架仅在紧凑平台上或没有真实负载交互的情况下进行验证，导致全尺寸类人机器人在真实负载下的应用仍然未被充分探索。扩展到全尺寸类人机器人带来了两个复合挑战：它们更大的惯性和更紧的平衡边际使得跟踪对噪声、漂移和来自普通虚拟现实（VR）跟踪器的重新目标错误高度敏感，同时它们的负载潜力仍然未被充分利用。我们提出了HEFT，一个重载全尺寸类人机器人遥操作框架，旨在解决这两个挑战。HEFT通过特权运动引导（Privileged Motion Guidance, PMG）从可部署的嘈杂VR参考中学习，并利用窗口化负载课程（Windowed Payload Curriculum, WPC）与专家指导的负载上限来获取稳健的重载跟踪。我们在L7（一个175厘米、65公斤的类人机器人）上部署HEFT。该机器人在负载高达24公斤的情况下跟踪包括转弯、前后移动和下蹲等动作。
cs.RO / 30 / 2607.02403
ACID: Action Consistency via Inverse Dynamics for Planning with World Models
ACID：通过逆动态实现的动作一致性用于世界模型规划
Abstract
Decision-time planning with action-conditioned world models has become a popular paradigm for embodied control. However, the standard planning cost judges a candidate solely by how close its predicted terminal state lies to the goal, leaving the realizability of the intermediate transitions unchecked -- a predicted trajectory can look convincing while the environment rollout drifts away from it. In this paper, we propose ACID, a decision-time planning framework that introduces cycle action consistency: the action inferred backward from a predicted transition by an inverse dynamics model should recover the one that was conditioned on. We fold this per-step residual into the planning cost via a scale-invariant adaptive weight. Across four action-conditioned world models and six tasks spanning rigid and deformable manipulation, articulated control, and visual navigation, ACID consistently improves planning and matches the baseline's accuracy with substantially less planning compute.
Chinese Translation
基于动作条件的世界模型进行决策时规划已成为具身控制的一个热门范式。然而，标准的规划成本仅通过候选者的预测终态与目标的接近程度来评判，未能检查中间过渡的可实现性——预测轨迹可能看起来令人信服，但环境的实际演化可能与之偏离。在本文中，我们提出了ACID，一个决策时规划框架，引入了循环动作一致性：通过逆动态模型从预测过渡推断出的动作应当恢复为条件动作。我们通过一个尺度不变的自适应权重将每一步的残差折叠到规划成本中。在四个动作条件的世界模型和六个涵盖刚性与可变形操作、关节控制和视觉导航的任务中，ACID始终改善了规划效果，并在显著减少规划计算的情况下与基线的准确性相匹配。
cs.RO / 31 / 2607.02417
LIME: Learning Intent-aware Camera Motion from Egocentric Video
LIME：从自我中心视频中学习意图感知的相机运动
Abstract
Autonomous robots often need to move their camera before they can act: to inspect an object, reveal an occluded region, or obtain a view that responds to a user's intent. While vision-language navigation translates instructions to base motion and vision-language-action policies map instructions to manipulation actions, language-conditioned camera motion remains comparatively underexplored as a first-class action. We formulate language-conditioned camera motion generation: given a current RGB observation and a free-form natural-language intent, predict a relative target camera pose for the next observation. This task is inherently non-trivial: viewpoint changes are driven by latent perceptual intentions, and a valid motion may operate at different semantic granularity, from entering a room to looking around a corner, inspecting a visible object, or revealing an occluded detail. To model this structure, we mine multi-intention camera-motion supervision from egocentric video, pairing plausible intents and observation-gain descriptions with relative SE(3) target poses. We propose LIME, a vision-language camera-motion generator that combines an auto-regressive observation-gain output with a continuous flow-matching pose head. This design lets the model jointly predict what the next view should reveal while representing multi-hypothesis target views. Across experiments and downstream robotic tasks, we show that LIME can learn to actively choose camera poses from passive human video, turning ordinary egocentric recordings into supervision for intent-aware active perception.
Chinese Translation
自主机器人在行动之前往往需要移动其相机：以检查物体、揭示被遮挡区域或获取响应用户意图的视图。虽然视觉-语言导航将指令转化为基础运动，而视觉-语言-动作策略则将指令映射到操作动作，但以语言为条件的相机运动作为一种一流动作仍然相对未被充分探索。我们提出了以语言为条件的相机运动生成：给定当前的RGB观察和自由形式的自然语言意图，预测下一个观察的相对目标相机姿态。这个任务本质上并不简单：视点变化受潜在感知意图驱动，有效的运动可能在不同的语义粒度上进行，从进入一个房间到环顾一个角落、检查一个可见物体或揭示被遮挡的细节。为了建模这种结构，我们从自我中心视频中挖掘多重意图的相机运动监督，将合理的意图和观察增益描述与相对SE(3)目标姿态配对。我们提出了LIME，一个视觉-语言相机运动生成器，它结合了自回归的观察增益输出和连续流匹配姿态头。这种设计使模型能够共同预测下一个视图应揭示的内容，同时表示多假设目标视图。在实验和下游机器人任务中，我们展示了LIME能够从被动的人类视频中主动选择相机姿态，将普通的自我中心录音转化为意图感知主动感知的监督。
cs.RO / 32 / 2607.02431
WorldSample: Closed-loop Real-robot RL with World Modelling
WorldSample：带有世界建模的闭环真实机器人强化学习
Abstract
Reinforcement learning (RL) can overcome the demonstration-coverage limitation of imitation learning (IL) by allowing robots to improve through trial-and-error interaction beyond the states observed in demonstrations. However, deploying RL on real robots remains constrained by high interaction costs, since each physical rollout is costly and reflects only one realized action-outcome path. To address this challenge, we propose WorldSample, a physically grounded data augmentation framework for real-robot RL that closes a real-synthetic loop between physical rollouts, world-model generation, and policy improvement. Grounded on real rollouts, WorldSample generates high-fidelity synthetic transitions through a post-trained world model, which greatly lowers the visual hallucination. Specifically, rather than simply using these transitions as real-world experience, WorldSample introduces Policy-Paced Learning (PPL) to regulate the training process through sample selection and scheduling, balancing useful augmentation against value overestimation and mitigating the hallucination-induced noise. Experiments on robot manipulation tasks involving contact-rich and precise tasks show that WorldSample improves policy success rate by 28% while reducing training steps by 59% compared with baselines. Furthermore, WorldSample improves world model visual fidelity by 19.4dB in PSNR and 0.47 in SSIM over demonstration-only post-training, validating the effectiveness of the real-synthetic loop for both policy and world model performance.
Chinese Translation
强化学习（RL）通过允许机器人在超越演示中观察到的状态下进行试错交互，克服了模仿学习（IL）的演示覆盖限制。然而，在真实机器人上部署强化学习仍然受到高交互成本的限制，因为每次物理执行的成本高昂，并且仅反映一个实现的行动-结果路径。为了解决这一挑战，我们提出了WorldSample，一个基于物理的真实机器人强化学习数据增强框架，它在物理执行、世界模型生成和策略改进之间闭合了一个真实-合成的循环。WorldSample基于真实执行，通过后训练的世界模型生成高保真度的合成过渡，这大大降低了视觉幻觉。具体而言，WorldSample并不仅仅将这些过渡视为真实世界的经验，而是引入了策略驱动学习（Policy-Paced Learning, PPL）来通过样本选择和调度来调节训练过程，平衡有用的增强与价值高估，并减轻由幻觉引起的噪声。在涉及接触丰富和精确任务的机器人操作任务上的实验表明，与基线相比，WorldSample将策略成功率提高了28%，同时减少了59%的训练步骤。此外，WorldSample在PSNR上提高了19.4dB，在SSIM上提高了0.47，验证了真实-合成循环在策略和世界模型性能上的有效性。
cs.RO / 33 / 2607.02466
Learning to Move Before Learning to Do: Task-Agnostic pretraining for VLAs
在学习如何做之前学习如何移动：任务无关的预训练用于视觉-语言-动作(VLA)
Abstract
Vision-Language-Action (VLA) models are fundamentally bottlenecked by the scarcity of expert demonstrations -- triplets of observations, instructions, and actions that are costly to collect at scale. We argue that this bottleneck stems from conflating two distinct learning objectives: acquiring physical competence (how to move) and acquiring semantic alignment (what to do). Crucially, only the latter requires language supervision. Building on this Decomposition Hypothesis, we propose Task-Agnostic Pretraining (TAP), a two-stage framework that first learns transferable motor priors from cheap, unlabeled interaction data -- including discarded off-task trajectories and autonomous robot play -- via a self-supervised Inverse Dynamics objective. A lightweight second stage then grounds these priors in language using minimal expert data. On the SIMPLER benchmark, TAP matches models trained on over 1M expert trajectories while using orders of magnitude less labeled data, yielding a 10% absolute gain over standard behavior cloning. On a real-world WidowX platform, TAP retains 25% success under camera perturbations where internet-scale baselines collapse to 0%, demonstrating that task-agnostic pretraining produces robust, transferable physical representations and offers a scalable path forward for Embodied AI.
Chinese Translation
视觉-语言-动作（VLA）模型的根本瓶颈在于专家演示的稀缺性——观察、指令和动作的三元组，这些数据的收集成本高昂且难以大规模实现。我们认为，这一瓶颈源于将两个不同的学习目标混淆在一起：获取物理能力（如何移动）和获取语义对齐（做什么）。关键是，只有后者需要语言监督。在这一分解假设的基础上，我们提出了任务无关预训练（Task-Agnostic Pretraining, TAP），这是一个两阶段框架，首先通过自监督的逆动力学目标，从廉价的无标签交互数据中学习可转移的运动先验——包括被丢弃的非任务轨迹和自主机器人游戏。第二阶段则轻量化地将这些先验与最小的专家数据结合，进行语言对接。在SIMPLER基准测试中，TAP在使用数量级更少的标注数据的情况下，匹配了在超过100万专家轨迹上训练的模型，较标准行为克隆实现了10%的绝对增益。在实际的WidowX平台上，TAP在摄像机扰动下保持了25%的成功率，而互联网规模的基线则崩溃至0%，这表明任务无关的预训练能够产生稳健的、可转移的物理表征，并为具身人工智能提供了一条可扩展的前进路径。
cs.RO / 34 / 2607.02472
Learning Agile Intruder Interception using Differentiable Quadrotor Dynamics
利用可微分四旋翼动力学学习灵活的入侵者拦截
Abstract
This paper presents a methodology for learning a control policy to intercept an intruder using the 3D direction unit vector to the intruder and the interceptor state. Prior deep reinforcement learning approaches assume either relative position or distance to the intruder is available, but this information is not readily accessible in real-world applications that employ passive, monocular camera sensors. Instead, we propose a solution that leverages an analytical policy gradient method using differentiable quadrotor dynamics to learn agile interception at speeds up to 10 m/s. The proposed approach outperforms baseline methods that utilize simplified point mass dynamics by an average of 30%.
Chinese Translation
本文提出了一种学习控制策略的方法，以利用指向入侵者的三维方向单位向量和拦截器状态来拦截入侵者。以往的深度强化学习方法假设可以获取相对位置或与入侵者的距离，但在使用被动单目摄像头传感器的实际应用中，这些信息并不容易获得。因此，我们提出了一种解决方案，利用可微分四旋翼动力学的解析策略梯度方法，以学习在高达10米/秒的速度下进行灵活拦截。所提出的方法在性能上比利用简化点质量动力学的基线方法平均提高了30%。
cs.RO / 35 / 2607.02474
QuadRocket: An Aerial Robotic Testbed for Adaptive Thrust-Vector Control of Rocket-Like Vehicles
QuadRocket：一种用于火箭类飞行器自适应推力矢量控制的空中机器人测试平台
Abstract
This paper presents QuadRocket, a quadrotor-based rocket prototype that provides a low-cost, low-risk platform for validating advanced thrust-vector control strategies for launch vehicle-type systems. The prototype consists of a cylindrical main body mounted on top of a quadrotor through a universal joint, forming a flying inverted pendulum with non-negligible inertia. For control design, the coupled system is modeled as a single axisymmetric rigid body actuated by a vectored force applied along its longitudinal axis. A reduced-attitude representation on the two sphere is adopted to explicitly exploit the vehicle's axial symmetry and to decouple yaw from the thrust-vector direction. On this model, we derive an adaptive backstepping controller that achieves almost global trajectory tracking in the presence of unknown constant disturbances, while a control-point transformation mitigates non minimum-phase behavior. The quadrotor is then treated as a thrust vector actuator, and a dynamic-surface-based attitude controller is designed to track the desired thrust-vector, accounting for actuation dynamics and avoiding explicit differentiation of virtual control signals. The complete architecture is evaluated in simulation and validated experimentally in an indoor motion-capture arena. Results demonstrate accurate trajectory tracking, effective disturbance compensation, and confirm the suitability of the QuadRocket as a versatile testbed for thrust-vector-controlled robotic vehicles.
Chinese Translation
本文介绍了QuadRocket，一种基于四旋翼的火箭原型，提供了一种低成本、低风险的平台，用于验证发射载具类型系统的先进推力矢量控制策略。该原型由一个圆柱形主体通过万向节安装在四旋翼顶部，形成一个具有不可忽视惯性的飞行倒摆。为了进行控制设计，耦合系统被建模为一个单轴对称的刚体，由沿其纵轴施加的矢量力驱动。采用了两球面上的简化姿态表示，以明确利用飞行器的轴对称性，并将偏航与推力矢量方向解耦。在此模型上，我们推导出了一种自适应反步控制器，能够在存在未知恒定干扰的情况下实现几乎全局的轨迹跟踪，同时控制点变换减轻了非最小相位行为。然后将四旋翼视为推力矢量执行器，设计了一种基于动态表面的姿态控制器，以跟踪期望的推力矢量，考虑了驱动动态并避免对虚拟控制信号的显式微分。完整架构在仿真中进行了评估，并在室内运动捕捉场地进行了实验验证。结果表明，轨迹跟踪准确，干扰补偿有效，确认了QuadRocket作为推力矢量控制机器人车辆的多功能测试平台的适用性。
cs.RO / 36 / 2607.02496
Controllable Sim Agents with Behavior Latents
可控的模拟代理与行为潜变量
Abstract
Realistic traffic simulation requires agents that imitate logged behavior and can also be steered along interpretable axes. Such controllability enables engineers to isolate variables, reproduce specific edge cases, and test autonomous systems without real-world risk. We introduce Controllable Neural Variational Agents (CNeVA), a controllable simulated-agent framework that learns to infer a per-agent Gaussian behavior latent from per-channel discounted returns via a closed-form conjugate variational update, conditioning a rectified-flow trajectory generator trained on a mixed channel-mask curriculum for classifier-free guidance. To tackle scarcity in reward signals, we propose soft eligibility gates that replace hard binary thresholds with smooth exponential decay, preserving the gradient signal for near-threshold agents. On the Waymo Open Motion Dataset, CNeVA attains competitive realism on the benchmark while exposing per-channel controllability that the higher-ranked imitation models lack. Speed- and acceleration-based steering produces monotone responses without stall-induced reward hacking. Safety controllability is monotone and substantial with the introduction of soft eligibility. We manage to achieve steerable map compliance under a context-residual return measure. Furthermore, our experiment demonstrates that steering metrics must be read alongside physical-plausibility guardrails to avoid reward-hacking confounds.
Chinese Translation
现实的交通模拟需要能够模仿记录行为并且可以沿可解释轴线进行引导的代理。这种可控性使工程师能够隔离变量、重现特定的边缘案例，并在没有现实世界风险的情况下测试自主系统。我们引入了可控神经变分代理（Controllable Neural Variational Agents, CNeVA），这是一个可控的模拟代理框架，能够通过闭式共轭变分更新，从每个通道的折扣回报中推断每个代理的高斯行为潜变量，并条件化一个在混合通道掩码课程上训练的修正流轨迹生成器，以实现无分类器的指导。为了解决奖励信号稀缺的问题，我们提出了软资格门，替代了硬二元阈值，采用平滑的指数衰减，保留了近阈值代理的梯度信号。在Waymo开放运动数据集中，CNeVA在基准测试中达到了竞争性的现实性，同时展现了高排名模仿模型所缺乏的每通道可控性。基于速度和加速度的引导产生单调响应，避免了因停滞引起的奖励黑客行为。安全可控性在引入软资格后表现出单调和显著性。我们成功实现了在上下文残差回报度量下的可引导地图合规。此外，我们的实验表明，引导指标必须与物理可行性保护措施一起解读，以避免奖励黑客混淆。
cs.RO / 37 / 2607.02501
Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots
Embodied.cpp：一种适用于异构机器人上体现式人工智能模型的可移植推理运行时
Abstract
Embodied AI models now span vision-language-action (VLA) models and world-action models (WAMs), but practical deployment remains fragmented across model-specific Python stacks, backend assumptions, and robot-side glue code, especially on heterogeneous edge devices. Existing inference runtimes are designed mainly for request-response serving and therefore do not satisfy the runtime contract of embodied deployment: multi-rate execution inside closed-loop control, latency-first batch-1 inference on heterogeneous hardware, and extensible embodied interfaces beyond fixed token I/O. We present Embodied.cpp, a portable C++ inference runtime for embodied models. Based on an architectural analysis of representative VLA models and WAMs, Embodied.cpp captures a shared execution path and organizes it into five layers: input adapters, sequence builders, backbone execution, head plugins, and deployment adapters. The runtime provides modular multi-rate execution, latency-first fused inference, and extensible operator and I/O support, enabling deployment across heterogeneous devices, robots, and simulators through one backend abstraction. We evaluate Embodied.cpp on two VLA models, HY-VLA and pi0.5, and on a preliminary WAM benchmark using a LingBot-VA Transformer block. The VLA deployments achieve successful closed-loop execution with 100.0% and 91.0% task success rates, respectively. The WAM benchmark reduces block memory from 312.2 MiB to 88.1 MiB. These results show that Embodied.cpp improves deployment efficiency while preserving high accuracy across diverse embodied model architectures.
Chinese Translation
体现式人工智能模型现在涵盖了视觉-语言-动作（VLA）模型和世界-动作模型（WAM），但实际部署仍然在模型特定的Python栈、后端假设和机器人侧的粘合代码之间碎片化，尤其是在异构边缘设备上。现有的推理运行时主要设计用于请求-响应服务，因此无法满足体现式部署的运行时契约：在闭环控制中进行多速率执行、在异构硬件上进行以延迟为优先的批量1推理，以及支持超出固定令牌输入/输出的可扩展体现式接口。我们提出了Embodied.cpp，一种用于体现式模型的可移植C++推理运行时。基于对代表性VLA模型和WAM的架构分析，Embodied.cpp捕捉了一个共享执行路径，并将其组织为五个层次：输入适配器、序列构建器、主干执行、头部插件和部署适配器。该运行时提供模块化的多速率执行、以延迟为优先的融合推理，以及可扩展的操作符和输入/输出支持，使得通过一个后端抽象能够在异构设备、机器人和模拟器上进行部署。我们在两个VLA模型HY-VLA和pi0.5上，以及使用LingBot-VA Transformer块的初步WAM基准上评估了Embodied.cpp。VLA部署分别实现了100.0%和91.0%的任务成功率的闭环执行。WAM基准将块内存从312.2 MiB减少到88.1 MiB。这些结果表明，Embodied.cpp在保持高准确率的同时提高了不同体现式模型架构的部署效率。
cs.RO / 38 / 2607.02503
VT-WAM: Visual-Tactile World Action Model for Contact-Rich Manipulation
VT-WAM：用于接触丰富操作的视觉-触觉世界动作模型
Abstract
Contact-rich manipulation requires policies to react to local deformation, pressure, slip, and friction, yet these cues are temporally sparse and often invisible in visual observations. Existing visual-tactile policies usually feed tactile observations directly into action prediction, but rarely model tactile deformation dynamics during action generation. In this paper, we introduce VT-WAM, a Visual-Tactile World Action Model that jointly learns future visual prediction, tactile deformation prediction, and action prediction within a unified flow matching framework. In particular, VT-WAM introduces (1) Asymmetric Mixture-of-Transformers (MoT) attention to bridge a first-frame visual anchor with temporal tactile dynamics, and (2) contact-gated Action-Visual-Tactile Attention Guidance (AVTAG) to encourage action queries to rely on tactile evidence during contact phases. Across six real-world contact-rich manipulation tasks, VT-WAM achieves a 71.67% average success rate, outperforming Fast-WAM by 26.67% and OmniVTLA by 35.84%. Ablations demonstrate that modeling tactile deformation dynamics and guiding contact-phase tactile attention are both important for contact-rich tasks. Project website: https://vt-wam.github.io/.
Chinese Translation
接触丰富的操作需要策略对局部变形、压力、滑动和摩擦作出反应，但这些线索在时间上稀疏且在视觉观察中常常不可见。现有的视觉-触觉策略通常直接将触觉观察输入到动作预测中，但在动作生成过程中很少建模触觉变形动态。本文介绍了VT-WAM，一种视觉-触觉世界动作模型，它在统一的流匹配框架内联合学习未来的视觉预测、触觉变形预测和动作预测。特别地，VT-WAM引入了（1）非对称混合变换器（MoT）注意力，以将第一帧视觉锚点与时间触觉动态连接起来，以及（2）接触门控的动作-视觉-触觉注意力引导（AVTAG），以鼓励在接触阶段动作查询依赖于触觉证据。在六个真实世界的接触丰富操作任务中，VT-WAM实现了71.67%的平均成功率，超越了Fast-WAM 26.67%和OmniVTLA 35.84%的表现。消融实验表明，建模触觉变形动态和引导接触阶段的触觉注意力对于接触丰富任务都非常重要。项目网站：https://vt-wam.github.io/
cs.CV / 1 / 2607.01290
AnchorSplat: Fast and Structure Consistent Detail Synthesis for Gaussian Splatting
AnchorSplat：高效且结构一致的高斯点云细节合成
Abstract
3D Gaussian Splatting (3DGS) has emerged as a powerful representation for high-fidelity rendering. However, existing assets often suffer from quality bottlenecks such as missing details and texture noise. Prior attempts to enhance these assets via 2D image processing introduce multi-view inconsistencies and high computational costs. In this paper, we propose a novel 3D-native refinement paradigm named AnchorSplat. AnchorSplat is an end-to-end deep network operating directly on 3D structures, avoiding the expensive optimization overhead of traditional 3D-2D-3D pipelines. Crucially, AnchorSplat is a strictly source-free solution requiring no original multi-view images. Central to the proposed method is the Point Anchor Mechanism, which enforces geometric consistency via local offset constraints, mitigating ill-posed mapping and gradient confounding. Furthermore, AnchorSplat replaces iterative densification with a single-pass multiplication mechanism. To facilitate research, we construct 3DGS-SR, the first large-scale benchmark for this task. Experiments demonstrate state-of-the-art results on the 3DGS-SR dataset, with throughput up to $10^5$ times faster than optimization methods. Notably, AnchorSplat exhibits robust zero-shot generalization across diverse data distributions, including generative model outputs and real-world scans.
Chinese Translation
3D 高斯点云合成（3DGS）已成为高保真渲染的一种强大表示方法。然而，现有资产常常面临质量瓶颈，如细节缺失和纹理噪声。之前通过 2D 图像处理增强这些资产的尝试引入了多视角不一致性和高计算成本。本文提出了一种新颖的 3D 原生细化范式，命名为 AnchorSplat。AnchorSplat 是一个端到端的深度网络，直接在 3D 结构上操作，避免了传统 3D-2D-3D 流程中昂贵的优化开销。关键是，AnchorSplat 是一种严格的无源解决方案，无需原始的多视角图像。该方法的核心是点锚机制（Point Anchor Mechanism），通过局部偏移约束强制几何一致性，减轻了不适定映射和梯度混淆。此外，AnchorSplat 用单次乘法机制替代了迭代密集化。为了促进研究，我们构建了 3DGS-SR，这是该任务的第一个大规模基准。实验结果表明，在 3DGS-SR 数据集上取得了最先进的结果，其吞吐量比优化方法快多达 $10^5$ 倍。值得注意的是，AnchorSplat 在多种数据分布中表现出强大的零样本泛化能力，包括生成模型输出和现实世界扫描。
cs.CV / 2 / 2607.01303
CPG-PAD: Concept-Informed Prompts Guided Presentation Attack Detection
CPG-PAD：基于概念引导的呈现攻击检测
Abstract
Presentation Attack Detection (PAD) serves as a crucial safeguard for face recognition systems against presentation attacks such as printed photos, replayed videos, and 3D masks. Despite significant progress, existing PAD models still struggle to generalize across unseen domains due to variations in sensors, lighting, and attack materials. Recent Vision-Language Models (VLMs) have shown strong generalization ability, yet their applications in PAD remain limited because learned prompts, typically optimized under class-label supervision, fail to explicitly align with fine-grained attack-relevant visual semantics. As a result, the learned representations often overfit domain-specific artifacts instead of capturing transferable attack cues. To address this, we propose Concept-Informed Prompts Guided Presentation Attack Detection (CPG-PAD), a framework that introduces model-level concept guidance into the prompt learning process. Specifically, we design a Visual Concept-driven Enhancement (VCE) module that employs eXplainable AI (XAI) techniques to automatically discover PAD-relevant visual concepts and generate concept-associated heatmaps providing localized fine-grained guidance. Guided by these heatmaps, a Prompt-based Concept Injection (PCI) mechanism integrates these concepts into the prompt space through a Visual-Prompt Decoder (VPD) and a concept-mapping loss, enabling prompts to align with the model's internal concept space. This design enables CPG-PAD to capture generalizable and domain-invariant attack cues while effectively suppressing dataset-specific biases. Extensive experiments across nine benchmark datasets demonstrate that CPG-PAD consistently achieves state-of-the-art cross-domain performance under multi-source, limited-source, and single-source settings.
Chinese Translation
呈现攻击检测（PAD）是面部识别系统抵御诸如打印照片、重播视频和3D面具等呈现攻击的重要保障。尽管取得了显著进展，现有的PAD模型在未见领域的泛化能力仍然不足，原因在于传感器、光照和攻击材料的变化。近期的视觉-语言模型（VLMs）显示出强大的泛化能力，但其在PAD中的应用仍然有限，因为学习到的提示通常是在类别标签监督下优化的，未能明确与细粒度的攻击相关视觉语义对齐。因此，学习到的表示往往过拟合于特定领域的伪影，而无法捕捉可转移的攻击线索。为了解决这一问题，我们提出了基于概念引导的呈现攻击检测（CPG-PAD）框架，该框架在提示学习过程中引入了模型级的概念指导。具体而言，我们设计了一个视觉概念驱动的增强（VCE）模块，利用可解释人工智能（XAI）技术自动发现与PAD相关的视觉概念，并生成提供局部细粒度指导的概念关联热图。在这些热图的指导下，基于提示的概念注入（PCI）机制通过视觉提示解码器（VPD）和概念映射损失将这些概念整合到提示空间中，使得提示能够与模型的内部概念空间对齐。这一设计使得CPG-PAD能够捕捉可泛化和领域不变的攻击线索，同时有效抑制数据集特定的偏差。在九个基准数据集上的广泛实验表明，CPG-PAD在多源、有限源和单源设置下始终实现了最先进的跨域性能。
cs.CV / 3 / 2607.01312
KathaTrace: Diagnosing Semantic Trajectory Collapse in Generated Visual Narratives
KathaTrace：诊断生成视觉叙事中的语义轨迹崩溃
Abstract
Visual narratives are central to storyboards, comics, children's media, and film previsualization, where viewers understand stories from images alone. Recent generators such as StoryDiffusion produce coherent sequences, but visual coherence does not guarantee that source-story transition meaning remains recoverable. Existing benchmarks assess visual quality, content faithfulness, and scene coherence, but miss a critical failure mode: storyboards where scenes appear visually coherent while the semantic link between scenes disappears. We introduce KathaTrace, a generator-agnostic protocol for diagnosing semantic trajectory collapse, defined as the loss of transition meaning needed to understand how one scene follows another. KathaTrace evaluates transitions under three evidence conditions: text-only, image-only, and text-plus-image, and filters ambiguous items. We contribute KathaBench-25K, with 5,000 narratives from classical collections including Aesop, Panchatantra, and Kathasaritasagara, 20,000 transitions, and 28,712 recoverability questions. We define Semantic Trajectory Gap, or STG, as text-only minus image-only recoverability, measuring transition meaning lost during visualization. Human validation yields Fleiss' kappa = 0.845. Experiments across state-of-the-art generators show substantial STG of 23.5 +/- 1.3. Semantic Compass, an actionability probe, uses KathaTrace signals for post-generation repair and improves storyboard selection.
Chinese Translation
视觉叙事是故事板、漫画、儿童媒体和电影预可视化的核心，观众仅通过图像理解故事。近期的生成器如 StoryDiffusion 能够生成连贯的序列，但视觉连贯性并不保证源故事的过渡意义仍然可恢复。现有基准评估视觉质量、内容忠实度和场景连贯性，但忽视了一种关键的失败模式：在场景视觉上看似连贯的故事板中，场景之间的语义联系消失。我们提出了 KathaTrace，这是一种与生成器无关的协议，用于诊断语义轨迹崩溃，定义为理解一个场景如何跟随另一个场景所需的过渡意义的丧失。KathaTrace 在三种证据条件下评估过渡：仅文本、仅图像和文本加图像，并过滤模糊项。我们贡献了 KathaBench-25K，包含来自经典文献的 5,000 个叙事，包括伊索寓言、潘查塔特拉和《故事海洋》，20,000 个过渡和 28,712 个可恢复性问题。我们定义了语义轨迹差距（Semantic Trajectory Gap，STG），即仅文本可恢复性减去仅图像可恢复性，测量在可视化过程中丧失的过渡意义。人工验证结果显示 Fleiss' kappa = 0.845。对最先进生成器的实验显示 STG 显著为 23.5 +/- 1.3。语义指南针（Semantic Compass），一种可操作性探测器，利用 KathaTrace 信号进行生成后修复，并改善故事板选择。
cs.CV / 4 / 2607.01353
Spatial-Temporal Expert Learning for Video-based Person Re-identification
基于视频的人物重识别的时空专家学习
Abstract
Video-based person re-identification (Re-ID) aims to retrieve the same identity in the query video clips from the gallery video clips. To solve this problem, exploiting fine-grained features is of great importance, especially when discriminating identities that are similar in appearance. In this paper, we propose to enhance the ability to explore fine-grained information with a novel input-aware extendable expert module. Instead of updating the network parameters with every sample in the dataset, we aim to train the experts within specific subsets that only contain similar samples and promote their ability to exploit fine-grained information within these similar samples. To achieve this goal, we incorporate two mechanisms in this module: input-aware expert selection mechanism and spatial-temporal selection mechanism. The first mechanism dynamically activates a set of experts on subsets of similar samples, pushing the experts to exploit subtle differences between these similar samples, while the second one further increases their sensitivity to the fine-grained differences in spatial and temporal aspects and allows the experts to dynamically utilize them for different input samples. In addition, to facilitate the expert module, we design an extendable scheme that allows the module to flexibly add new experts when necessary. As a result, our method achieves outstanding performance on two large-scale datasets.
Chinese Translation
基于视频的人物重识别（Re-ID）旨在从图库视频片段中检索查询视频片段中的相同身份。为了解决这个问题，挖掘细粒度特征至关重要，尤其是在区分外观相似的身份时。本文提出了一种新颖的输入感知可扩展专家模块，以增强探索细粒度信息的能力。我们并不打算用数据集中每个样本更新网络参数，而是希望在仅包含相似样本的特定子集内训练专家，并提升它们在这些相似样本中挖掘细粒度信息的能力。为实现这一目标，我们在该模块中结合了两种机制：输入感知专家选择机制和时空选择机制。第一种机制动态激活一组专家在相似样本的子集上，推动专家挖掘这些相似样本之间的细微差异，而第二种机制则进一步增强它们对时空方面细粒度差异的敏感性，使专家能够动态利用这些差异来处理不同的输入样本。此外，为了便于专家模块的使用，我们设计了一种可扩展方案，允许在必要时灵活添加新的专家。因此，我们的方法在两个大规模数据集上实现了优异的性能。
cs.CV / 5 / 2607.01370
MapDreamer: Aerial Imagery Conditioned Latent Diffusion for Lane-Level Map Generation
MapDreamer：基于航拍图像的车道级地图生成的潜在扩散模型
Abstract
High definition map generation is essential for autonomous driving, yet remains a labor-intensive process at scale. We present MapDreamer, a generative diffusion model that synthesizes lane-level vector maps with explicit topology directly from a single aerial image. MapDreamer learns a compact latent representation of lane centerlines and their topological relations using a variational autoencoder and predicts graphs with a transformer-based latent diffusion model. To align generated maps with the observed scene, we condition each denoising step on dense aerial features injected through cross-attention. To handle the varying number of lanes across scenes, we propose a lane cardinality module paired with background ghost lane latents, a learned buffer that prevents slot collapse during diffusion. Furthermore, we introduce a sliding-window global graph aggregation strategy that stitches local tiles into city-scale maps while preserving connectivity through encoded lane boundaries. Experiments on UrbanLaneGraph derived from Argoverse 2 show improved geometric and topological fidelity over non-generative baselines.
Chinese Translation
高精度地图生成对于自动驾驶至关重要，但在大规模应用中仍然是一个劳动密集型的过程。我们提出了MapDreamer，这是一种生成性扩散模型，可以直接从单幅航拍图像合成具有明确拓扑结构的车道级矢量地图。MapDreamer使用变分自编码器学习车道中心线及其拓扑关系的紧凑潜在表示，并利用基于变压器的潜在扩散模型预测图形。为了使生成的地图与观察到的场景对齐，我们在每个去噪步骤中通过交叉注意力注入密集的航拍特征。为了处理不同场景中车道数量的变化，我们提出了一种车道基数模块，并配备背景幽灵车道潜在表示，这是一种学习的缓冲区，防止在扩散过程中出现插槽崩溃。此外，我们引入了一种滑动窗口全局图聚合策略，将局部瓦片拼接成城市规模的地图，同时通过编码的车道边界保持连通性。在从Argoverse 2派生的UrbanLaneGraph上的实验表明，与非生成基线相比，几何和拓扑保真度得到了改善。
cs.CV / 6 / 2607.01383
MIBE: Multi-subject Interaction Benchmark and Evaluator for Personalized Image Generation
MIBE：用于个性化图像生成的多主体交互基准和评估器
Abstract
Multi-subject personalized image generation requires the precise rendering of all requested reference identities and their specified interactions based on a guiding prompt. However, state-of-the-art models still struggle with this process, frequently omitting subjects, failing to preserve reference appearances, or misattributing interactions. Furthermore, existing metrics designed primarily for single-subject fidelity cannot reliably capture these errors, suffering severe degradation in ranking separability and failing to align with human preference as the subject count increases. To address this gap, we introduce Multi-subject Interaction Benchmark and Evaluator (MIBE), a unified framework comprising a Multi-subject Interaction Benchmark (MIB) and a Multi-subject Interaction Evaluator (MIE). MIB systematically covers diverse relation types and scene complexities through a decoupled data regime. This consists of a 60K-pair VLM-labeled Silver Set for scalable metric training and a 4K-pair double-blind Human Evaluation Gold Set covering a diverse range of state-of-the-art generators, with the Silver Set reaching 95.1% cross-VLM preference agreement. To demonstrate the utility of this benchmark, we present MIE, a lightweight, reference-conditioned evaluator trained exclusively on the Silver Set with a dual-head ranking and diagnosis objective. MIE exhibits strong cross-generator generalization on the Gold Set, achieving 0.922 overall pairwise accuracy against human preference, including 0.982 on seen generators and 0.884 on unseen generators. By outperforming a broad spectrum of baseline metrics, including CLIP and DINO variants, MIE demonstrates that diagnostic supervision can preserve ranking separability and human alignment where traditional evaluators collapse.
Chinese Translation
多主体个性化图像生成需要根据指导提示精确渲染所有请求的参考身份及其指定的交互。然而，当前最先进的模型在这一过程中仍面临挑战，常常遗漏主体、未能保留参考外观或错误归因交互。此外，现有主要针对单主体保真度设计的度量标准无法可靠地捕捉这些错误，随着主体数量的增加，其排名可分离性严重下降，并且未能与人类偏好对齐。为了解决这一问题，我们提出了多主体交互基准和评估器（MIBE），这是一个统一框架，包括多主体交互基准（MIB）和多主体交互评估器（MIE）。MIB通过解耦的数据模式系统地覆盖了多种关系类型和场景复杂性。该框架包含一个60K对的VLM标记银集，用于可扩展的度量训练，以及一个4K对的双盲人类评估金集，涵盖了多种最先进的生成器，其中银集达到了95.1%的跨VLM偏好一致性。为了展示该基准的实用性，我们提出了MIE，一个轻量级的、基于参考条件的评估器，仅在银集上进行训练，具有双头排名和诊断目标。MIE在金集上表现出强大的跨生成器泛化能力，整体对人类偏好的配对准确率达到0.922，其中在已见生成器上为0.982，而在未见生成器上为0.884。通过超越包括CLIP和DINO变体在内的广泛基线度量，MIE证明了诊断监督能够在传统评估器崩溃的情况下保持排名可分离性和人类对齐。
cs.CV / 7 / 2607.01395
Rethinking Generic Object Tracking Toward Human-Level Perceptual Intelligence
重新思考通用物体跟踪以实现人类水平的感知智能
Abstract
At the heart of human visual perception lies the ability to maintain a continuous and coherent understanding of the external world. By integrating observations with accumulated experience, the human visual system can continuously adapt to variations in both the target and its surrounding environment, while preserving robust visual continuity as scene dynamics evolve. Human vision can therefore integrate prior knowledge, spatial geometry, and semantic context to understand complex scenes and their changes. As a core problem in computer vision, visual object tracking aims to bring machine perception closer to human visual perception. These capabilities are central to the task of Generic Object Tracking (GOT). In this task, a visual tracker is initialized only with the bounding box of an arbitrarily specified target in the first frame, and must continuously localize the target in subsequent dynamic visual streams. However, future events, observations, and real-world variations are inherently unpredictable; therefore, the model's generalization and online adaptation capabilities remain bottlenecks. Tracking reliability can deteriorate when the target undergoes severe deformation, is affected by complex distractors, encounters significant environmental changes, or belongs to a category unseen during training. This dissertation aims to narrow the gap between machine visual tracking systems and human visual perception by proposing a series of methods that systematically enhance the target discrimination, robust adaptation, and geometric reasoning capabilities of tracking models.
Chinese Translation
人类视觉感知的核心在于保持对外部世界的连续和一致的理解。通过将观察与累积的经验相结合，人类视觉系统能够不断适应目标及其周围环境的变化，同时在场景动态演变时保持稳健的视觉连续性。因此，人类视觉能够整合先前知识、空间几何和语义上下文，以理解复杂场景及其变化。作为计算机视觉中的核心问题，视觉物体跟踪旨在使机器感知更接近人类视觉感知。这些能力是通用物体跟踪（Generic Object Tracking, GOT）任务的核心。在该任务中，视觉跟踪器仅通过第一帧中任意指定目标的边界框进行初始化，并必须在后续动态视觉流中持续定位目标。然而，未来事件、观察和现实世界的变化本质上是不可预测的；因此，模型的泛化和在线适应能力仍然是瓶颈。当目标经历严重变形、受到复杂干扰物的影响、遇到显著的环境变化或属于训练期间未见过的类别时，跟踪的可靠性可能会下降。本论文旨在通过提出一系列方法，系统性地增强跟踪模型的目标区分、稳健适应和几何推理能力，从而缩小机器视觉跟踪系统与人类视觉感知之间的差距。
cs.CV / 8 / 2607.01396
Computer Vision for Wildlife Monitoring: Detecting Brown Howler Monkeys using YOLO
计算机视觉在野生动物监测中的应用：使用YOLO检测棕色吼猴
Abstract
Urban expansion threatens global biodiversity, especially affecting arboreal species due to the fragmentation of forest habitats. The movement of arboreal species across disjointed forest patches increases mortality risk and, thus, compromises their conservation. In this context, the installation of canopy bridges can be a viable strategy; yet continuous monitoring of their use by arboreal species is essential for ensuring their effectiveness, typically carried out with the aid of camera traps. However, this method often produces false-positive images that demand time from conservationists for review. In this context, computer vision algorithms can optimize the task of detecting target species using the canopy bridges. In this study, we explored the automatic detection of brown howler monkeys (Alouatta guariba) in videos obtained by camera traps. Given the need for a large number of annotated images of the target animals to train the algorithms, we tested the incorporation of auxiliary data to improve detection models, fine-tuning the YOLOv10 framework using varying proportions of them. The improvement of these automatic detection techniques contributes to conservation efforts, by providing automatic tools to monitor solutions that minimize the impact of human interference in animals habitats.
Chinese Translation
城市扩张威胁全球生物多样性，尤其对树栖物种造成影响，因为森林栖息地的破碎化增加了树栖物种在不连续森林斑块之间移动的死亡风险，从而影响其保护。在这种背景下，安装树冠桥可能是一种可行的策略；然而，持续监测树栖物种对其使用情况至关重要，以确保其有效性，这通常依赖于相机陷阱的帮助。然而，这种方法往往会产生误报图像，需耗费保护工作者的时间进行审查。在这种情况下，计算机视觉算法可以优化使用树冠桥检测目标物种的任务。在本研究中，我们探索了在相机陷阱获取的视频中自动检测棕色吼猴（Alouatta guariba）。鉴于需要大量标注的目标动物图像来训练算法，我们测试了辅助数据的引入，以改善检测模型，使用不同的比例对YOLOv10框架进行微调。这些自动检测技术的改进有助于保护工作，通过提供自动化工具来监测解决方案，从而最小化人类干扰对动物栖息地的影响。
cs.CV / 9 / 2607.01416
Beyond Heatmaps: Unsupervised Concept-Graph Reasoning for Interpretable Visual Explanation
超越热图：用于可解释视觉解释的无监督概念图推理
Abstract
Concept Bottleneck Models (CBMs) provide an intrinsically interpretable alternative to post-hoc explanations. However, existing CBMs often rely on predefined concept vocabularies or supervised annotations, lack explicit concept grounding, and summarize each concept with a single image-level score -- discarding spatial recurrence and inter-concept dependencies. We propose a Graph-based Concept Bottleneck Model (G-CBM), an intrinsically interpretable framework that performs unsupervised concept discovery via Non-negative Matrix Factorization (NMF) and represents the discovered concepts as nodes in a per-image concept-graph representation. G-CBM matches region-level features to these concept nodes -- providing concept grounding and capturing concept recurrence across the image -- and applies a \emph{tunable concept filtering threshold} $\tau$ to suppress weak region-level features. A Graph Attention Network (GAT) then performs concept-level reasoning by modeling nonlinear dependencies across nodes. Across ImageNet, HAM10000, PH2, and Derm7pt, G-CBM achieves an average relative AUC improvement of 3.7\% over a ResNet-50 baseline. Concept filtering frequently improves predictive performance while inducing selective concept use, achieving peak AUC of $0.96$ on PH2 with only 2 of 10 concepts and 0.92 on HAM10000 with 3.8 of 9 concepts. On dermoscopy benchmarks, G-CBM is competitive with supervised approaches requiring external annotations. Deletion/insertion analyses with random ablation controls show that the learned concept ranking faithfully reflects model predictions.
Chinese Translation
概念瓶颈模型（Concept Bottleneck Models, CBMs）提供了一种内在可解释的替代方案，以替代后期解释。然而，现有的CBMs通常依赖于预定义的概念词汇或监督注释，缺乏明确的概念基础，并且用单一的图像级得分来总结每个概念——忽略了空间重复性和概念间的依赖关系。我们提出了一种基于图的概念瓶颈模型（Graph-based Concept Bottleneck Model, G-CBM），这是一种内在可解释的框架，通过非负矩阵分解（Non-negative Matrix Factorization, NMF）进行无监督的概念发现，并将发现的概念表示为每幅图像的概念图中的节点。G-CBM将区域级特征与这些概念节点相匹配——提供概念基础并捕捉图像中的概念重复性——并应用一个可调的概念过滤阈值（tunable concept filtering threshold）$ au$来抑制弱区域级特征。然后，图注意力网络（Graph Attention Network, GAT）通过建模节点间的非线性依赖关系进行概念级推理。在ImageNet、HAM10000、PH2和Derm7pt数据集上，G-CBM相较于ResNet-50基线实现了平均相对AUC提升3.7%。概念过滤通常提高预测性能，同时引导选择性概念使用，在PH2上仅使用10个概念中的2个时达到峰值AUC为$0.96$，在HAM10000上使用9个概念中的3.8个时达到0.92。在皮肤镜基准测试中，G-CBM与需要外部注释的监督方法具有竞争力。随机消融控制的删除/插入分析表明，学习到的概念排名忠实地反映了模型预测。
cs.CV / 10 / 2607.01435
Sign in the Air to Unlock: An Interface for authentication in Virtual and Augmented Reality Powered by Point-Voxel Cross-Attention Network
空中签名解锁：一种基于点-体素交叉注意力网络的虚拟和增强现实认证接口
Abstract
Significant advancement of immersive technologies such as Virtual and Augmented Reality (VR/AR) and their integration into diverse aspects of modern life need authentication interfaces that are secure, intuitive, and compatible with embodied interaction. Traditional methods such as passwords, PINs, and device-based logins, break immersion and rely on external hardware. Recent 3D-specific behavioral approaches, such as hand-gesture, eye-tracking, and electroencephalography (EEG)-based methods, offer promising alternatives but often require specialized sensors or constrain natural movement, limiting usability in dynamic environments. We present Sign in the Air to Unlock, an in-air signature interface that enables users to authenticate by signing naturally in 3D space which is a familiar, personal, and reproducible gesture. To realize this interface, we design a point-voxel Cross-Attention Network (PV-Net) that jointly models local motion dynamics and global spatial structure from 3D trajectories. The model is evaluated on two datasets: the public DeepAirSig dataset (1,800 signatures from 40 users) and ImmAirsig, a new dataset collected using Meta Quest 2 in immersive VR (880 samples from 22 users). PV-Net achieves an Equal Error Rate of 2.5% on DeepAirSig and 76% classification accuracy on ImmAirSig. These findings highlight the potential of 3D behavioral interfaces for seamless, user-centric authentication that merges security with natural interaction in immersive environments.
Chinese Translation
沉浸式技术（如虚拟现实和增强现实，VR/AR）的显著进步及其在现代生活各个方面的整合，需要安全、直观且与具身交互兼容的认证接口。传统方法如密码、个人识别码（PIN）和基于设备的登录打破了沉浸感，并依赖于外部硬件。最近的3D特定行为方法，如手势、眼动追踪和基于脑电图（EEG）的方法，提供了有前景的替代方案，但通常需要专用传感器或限制自然运动，从而限制了在动态环境中的可用性。我们提出了“空中签名解锁”这一在空中签名的认证接口，使用户能够通过在3D空间中自然签名进行认证，这是一种熟悉、个性化且可重复的手势。为了实现这一接口，我们设计了一种点-体素交叉注意力网络（PV-Net），该网络共同建模了3D轨迹的局部运动动态和全局空间结构。该模型在两个数据集上进行了评估：公共的DeepAirSig数据集（来自40个用户的1800个签名）和ImmAirsig，这是一个在沉浸式VR中使用Meta Quest 2收集的新数据集（来自22个用户的880个样本）。PV-Net在DeepAirSig上实现了2.5%的等错误率，在ImmAirSig上达到了76%的分类准确率。这些发现突显了3D行为接口在无缝、以用户为中心的认证中的潜力，将安全性与沉浸式环境中的自然交互相结合。
cs.CV / 11 / 2607.01437
How Much Future Helps? A Controlled Study of Future-Privileged Supervision for Causal Egocentric Gaze Estimation
未来的帮助有多大？对因果自我中心注视估计的未来特权监督的控制研究
Abstract
Egocentric gaze estimation is commonly studied using models that process the full video with access to future frames, while real-world applications require strictly causal, online prediction. This discrepancy raises key questions: Does future context inherently provide valuable signals for gaze estimation? If so, how much future look-ahead optimally supervises a causal model during training? To investigate, we propose a controlled framework featuring a future-aware branch that accesses a tunable look-ahead horizon during training but is discarded at inference. This design isolates the impact of future context while keeping the inference architecture fixed and strictly causal. Across EGTEA Gaze+ and Ego4D, we find that future-privileged supervision consistently improves causal gaze prediction, confirming its utility. However, performance gains do not increase monotonically with longer look-ahead, but rather peak within a bounded temporal regime. Specifically, optimal performance corresponds to roughly 1.7--3.3 seconds of future context ($H{\in}[5, 10]$) on EGTEA Gaze+ and 2.7 seconds ($H{=}10$) on Ego4D. Our results demonstrate that lightweight causal models can effectively absorb future-aware signals, providing practical guidance for real-time egocentric gaze modeling.
Chinese Translation
自我中心注视估计通常使用能够处理完整视频并访问未来帧的模型进行研究，而现实世界的应用则需要严格因果的在线预测。这种差异引发了关键问题：未来上下文是否固有地为注视估计提供了有价值的信号？如果是，多少未来的前瞻性在训练过程中最优地监督因果模型？为此，我们提出了一个控制框架，具有一个未来感知分支，该分支在训练期间可以访问可调的前瞻性视野，但在推理时被丢弃。该设计在保持推理架构固定且严格因果的同时，隔离了未来上下文的影响。在EGTEA Gaze+和Ego4D数据集上，我们发现未来特权监督始终改善因果注视预测，确认了其有效性。然而，性能提升并不是随着前瞻性时间的延长而单调增加，而是在一个有限的时间范围内达到峰值。具体而言，最佳性能对应于EGTEA Gaze+上大约1.7至3.3秒的未来上下文（$H{ ext{in}}[5, 10]$）和Ego4D上的2.7秒（$H{=}10$）。我们的结果表明，轻量级因果模型能够有效吸收未来感知信号，为实时自我中心注视建模提供了实用指导。
cs.CV / 12 / 2607.01469
A Cost-Aware, Paired Protocol for Auditing Dynamic Tool Synthesis in Agentic Video Question Answering
一种成本感知的配对协议用于审计动态工具合成在自主视频问答中的应用
Abstract
Agentic Video Question Answering (VideoQA) systems invoke tools during inference, but their tool libraries are fixed, so recurring procedures are rebuilt from primitives on every question. Synthesizing composite tools could remove this overhead, but whether such expansion helps is hard to assess: final-answer accuracy, the standard metric, ignores inference effort, so it cannot reveal how a system shifts cost. We propose a cost-aware, paired protocol for auditing tool-augmented video agents. The protocol pairs two complete systems on the same input for each question and reports their net difference across accuracy and cost jointly. For each question, it sorts the paired outcome into one of six groups defined by joint correctness and by the change in visible tool calls, separating accuracy-preserving efficiency gains from harmful regressions. Significance is reported with McNemar's test and paired bootstrap confidence intervals. We instantiate the protocol on Dynamic-SAGE, an agentic VideoQA framework that synthesizes, validates, and persistently registers executable composite tools for reuse on unseen questions, and evaluate it against the SAGE baseline on SAGE-Bench. The audit reveals a multi-axis profile that a scalar accuracy comparison would miss: Dynamic-SAGE improves accuracy by 7.5 points (p < 0.001) and reduces reasoning turns and visible tool calls by roughly 28%, while shifting rather than reducing inference cost, as token usage rises 34% and cost 26%. Gains are largest on visual and open-ended questions and neutral on verbal and multimodal ones, and residual failures concentrate on hard, open-ended questions where the pipeline does the most work. By measuring accuracy and cost jointly, the protocol shows where the pipeline-level difference is reliable and where it is not. The code is available at https://github.com/KurbanIntelligenceLab/Dynamic-SAGE.
Chinese Translation
自主视频问答（VideoQA）系统在推理过程中调用工具，但其工具库是固定的，因此每个问题的重复过程都需要从基本操作重新构建。合成复合工具可以消除这种开销，但评估这种扩展是否有帮助是困难的：最终答案的准确性这一标准指标忽略了推理的努力，因此无法揭示系统如何转变成本。我们提出了一种成本感知的配对协议，用于审计工具增强的视频代理。该协议对每个问题在相同输入上配对两个完整系统，并共同报告它们在准确性和成本上的净差异。对于每个问题，它将配对结果排序为六个组之一，这些组由联合正确性和可见工具调用的变化定义，从而将保持准确性的效率提升与有害的回归分开。显著性通过McNemar检验和配对自助法置信区间报告。我们在Dynamic-SAGE上实例化该协议，Dynamic-SAGE是一个自主视频问答框架，能够合成、验证并持久注册可执行的复合工具以供在未见问题上重用，并在SAGE-Bench上与SAGE基线进行评估。审计揭示了一个多轴特征，而标量准确性比较将会忽略：Dynamic-SAGE的准确性提高了7.5个百分点（p < 0.001），推理回合和可见工具调用减少了大约28%，同时推理成本有所转移而非减少，令牌使用量上升了34%，成本上升了26%。在视觉和开放式问题上收益最大，而在语言和多模态问题上则表现中性，残余失败主要集中在难度较大、开放式的问题上，这些问题的处理工作量最大。通过共同测量准确性和成本，该协议展示了管道级差异的可靠性和不可靠性。代码可在 https://github.com/KurbanIntelligenceLab/Dynamic-SAGE 获取。
cs.CV / 13 / 2607.01499
Anti-Prompt: Image Protection against Text-Guided Image-to-Video Generation
反提示：针对文本引导的图像到视频生成的图像保护
Abstract
Recent advances in Image-to-Video generation allow a single image to be animated into a convincing video under text guidance, raising serious copyright and privacy risks. We propose Anti-Prompt, an image protection approach that injects imperceptible perturbations into an image, inducing visible inconsistencies and structural failures in text-guided I2V generation. Our method is motivated by a simple empirical observation. When text guidance is removed from modern I2V models, generation quality degrades markedly, not only in motion realism but also in subject preservation, structural coherence, and temporal consistency. Building on this insight, Anti-Prompt exploits the model reliance on textual guidance by attenuating text-conditioned interactions during denoising while strengthening visual-only pathways. To further systematically evaluate protection effectiveness, we introduce a Video-LLM-assisted evaluation protocol that provides interpretable, frame-grounded analyses of generation artifacts and inconsistencies. Experiments on two representative I2V architectures demonstrate that our method achieves strong protection performance while improving efficiency and cross-model transferability.
Chinese Translation
最近在图像到视频生成方面的进展使得单一图像能够在文本指导下被动画化为逼真的视频，这带来了严重的版权和隐私风险。我们提出了反提示（Anti-Prompt），一种图像保护方法，通过向图像注入不可察觉的扰动，导致文本引导的图像到视频生成中出现明显的不一致性和结构性失败。我们的方法源于一个简单的经验观察：当现代图像到视频模型中移除文本指导时，生成质量显著下降，不仅在运动真实感方面，而且在主体保留、结构一致性和时间一致性方面也有所下降。基于这一洞察，反提示通过在去噪过程中减弱文本条件交互，同时增强仅视觉路径的作用，利用了模型对文本指导的依赖。为了进一步系统地评估保护效果，我们引入了一种视频大语言模型（Video-LLM）辅助的评估协议，提供可解释的、基于帧的生成伪影和不一致性分析。在两个代表性的图像到视频架构上的实验表明，我们的方法在提高效率和跨模型可转移性的同时，达到了强大的保护性能。
cs.CV / 14 / 2607.01503
Disentangling Pictorial Cue Understanding from Language Bias in VLMs via Depth Ordering Task
通过深度排序任务解开视觉语言模型中的图像线索理解与语言偏见
Abstract
In this paper, we study depth perception of vision-language models (VLMs) to isolate the effects of pictorial depth cues and disentangle vision and language influences on model performance. To this end, we combine depth-ordering and odd-one-out psychophysical tasks: the VLMs are presented with images where one object is at different depth relative to other, otherwise identical, objects, and must determine whether the odd-one-out target is closer or farther to the observer. To create stimuli, we generate 2D views from simulated and real 3D scenes while controlling the presence of individual pictorial depth cues, enabling a fine-grained analysis of cue-level contributions. Language effects are examined by varying referring expression clarity. We also introduce a novel metric to quantify vision-vs-language sensitivities. Applying this methodology, we create the Odd-One-Out Depth (O3-D) dataset with 37K real and synthetic images and 147K image-question pairs. Evaluation of 12 open-source and commercial models on O3-D shows under-utilization of depth cues and depth-ordering accuracies between 47% and 56%, with no model above chance level. At the same time, our metric reveals strong linguistic bias in the answers. Neither chain-of-thought (CoT) nor in-context learning (ICL) significantly improves performance, suggesting that static image data alone may be insufficient for depth understanding. All code, the image generation pipeline, and the O3-D dataset are publicly released at https://github.com/lyiqian/o3-d.
Chinese Translation
在本文中，我们研究视觉语言模型（VLMs）的深度感知，以隔离图像深度线索的影响，并解开视觉与语言对模型性能的影响。为此，我们结合了深度排序和奇异物体心理物理任务：VLMs被呈现出图像，其中一个物体相对于其他相同的物体处于不同的深度，并必须判断奇异物体目标是更靠近还是更远离观察者。为了创建刺激材料，我们从模拟和真实的3D场景生成2D视图，同时控制个别图像深度线索的存在，从而实现对线索级别贡献的细致分析。通过改变指称表达的清晰度来检查语言效应。我们还引入了一种新颖的度量标准来量化视觉与语言的敏感性。应用该方法，我们创建了包含37K真实和合成图像及147K图像-问题对的奇异物体深度（O3-D）数据集。对12个开源和商业模型在O3-D上的评估显示深度线索的利用不足，深度排序的准确率在47%到56%之间，没有模型超过随机水平。同时，我们的度量标准揭示了答案中的强语言偏见。链式思维（CoT）和上下文学习（ICL）均未显著提高性能，表明仅依靠静态图像数据可能不足以理解深度。所有代码、图像生成管道和O3-D数据集均已公开发布在https://github.com/lyiqian/o3-d。
cs.CV / 15 / 2607.01535
Hidden-Shot: Towards One-Shot Task Generalization for Low-Level Vision Generalist Models
隐性拍摄：面向低级视觉通用模型的一次性任务泛化
Abstract
Despite the intense engagement surrounding low-level vision generalist models, their effectiveness in zero/few-shot scenarios beyond learned tasks remains unverified. The primary challenge of developing an ideal generalist lies in achieving the ability to generalize from new unseen tasks, which also can be assessed by matched quantitative criteria. Existing methods have made some progress in prompt engineering but have not systematically explored this gap across a wide range of low-level visual tasks. Stimulated by the problem, we propose Hidden-Shot, an implicit prompt mechanism aimed at exploring low-level task adaptation in a vision generalist model. Specifically, the method extracts implicit visual task-based information, utilizes a global task-aware textural prompt, and selectively merges implicit information with in-task processing information to enhance one-shot capabilities in new tasks. The overall design performs direct injection in a cost-effective manner, while minimally altering the architecture of the original generalist model. Additionally, we introduce a data-driven evaluation framework termed C/U assessment to cover two basic scenarios, 3C4U (3 conventional and 4 unconventional tasks) for retraining existing models and 3C7U (3 conventional and 7 unconventional tasks) for training from scratch, as a comprehensive assessment to systematically test the generalization ability of low-level generalist models. Experiments on seven and ten datasets outperform the state-of-the-art vision generalist model, respectively verified by 3C4U and 3C7U framework. Our presented Hidden-Shot approach demonstrates superior performance on one-shot new tasks while maintaining consistent performance on existing tasks.
Chinese Translation
尽管低级视觉通用模型引起了广泛关注，但其在超出已学习任务的零/少样本场景中的有效性仍未得到验证。开发理想通用模型的主要挑战在于实现从新的未见任务中进行泛化的能力，这也可以通过匹配的定量标准进行评估。现有方法在提示工程方面取得了一定进展，但尚未系统性地探索这一差距，尤其是在广泛的低级视觉任务中。受到这一问题的启发，我们提出了隐性拍摄（Hidden-Shot），一种旨在探索视觉通用模型中低级任务适应的隐式提示机制。具体而言，该方法提取隐式视觉任务相关信息，利用全局任务感知的纹理提示，并选择性地将隐式信息与任务内处理信息合并，以增强在新任务中的一次性能力。整体设计以成本效益高的方式进行直接注入，同时对原始通用模型的架构进行最小化修改。此外，我们引入了一种数据驱动的评估框架，称为C/U评估，涵盖两种基本场景：3C4U（3个常规任务和4个非常规任务）用于重新训练现有模型，以及3C7U（3个常规任务和7个非常规任务）用于从头开始训练，作为全面评估以系统性测试低级通用模型的泛化能力。在七个和十个数据集上的实验结果分别超越了最先进的视觉通用模型，验证了3C4U和3C7U框架。我们提出的隐性拍摄方法在一次性新任务上表现出优越的性能，同时在现有任务上保持一致的表现。
cs.CV / 16 / 2607.01555
Boosting Infrared Small Target Detection via Logit-Domain Contrast and Adaptive Shape Refinement
通过Logit域对比和自适应形状优化提升红外小目标检测
Abstract
Infrared small target detection (IRSTD) remains challenging due to tiny target size, low signal-to-noise ratio, severe foreground-background imbalance, and blurred boundaries in complex scenes. Existing methods usually rely on post-activation probability-domain supervision for discrimination, where weak targets and strong clutter may produce saturated and close probabilities, limiting weak-target discrimination. Meanwhile, blurred boundaries and halo-like predictions mainly stem from thermal diffusion, tiny target scale, boundary uncertainty, and insufficient explicit contour constraints. To address these issues, we propose Adaptive-Contrastive SLSIoU (AC-SLSIoU), a plug-and-play discriminative and shape-aware loss for IRSTD. Specifically, a Logit-Domain Margin Constraint (LDMC) is introduced to enlarge the response gap between targets and informative hard negatives in the logit space, thereby enhancing weak-target discrimination. Adaptive Boundary Suppression (ABS) applies scale-aware annular penalties to refine target contours and suppress halo-like overflow responses. In addition, False-Alarm Focal Loss assigns larger weights to high-probability negative samples, further penalizing persistent high-confidence false alarms. Without introducing extra inference overhead, the proposed method can be seamlessly integrated into existing detectors and consistently improves both detection accuracy and shape quality. Extensive experiments and cross-backbone evaluations demonstrate the effectiveness, robustness, and generalization ability of the proposed method for infrared small target detection.
Chinese Translation
红外小目标检测（IRSTD）因目标尺寸微小、信噪比低、前景与背景严重失衡以及复杂场景中的边界模糊而仍然具有挑战性。现有方法通常依赖于后激活概率域监督进行区分，其中弱目标与强干扰可能产生饱和且接近的概率，从而限制了弱目标的区分能力。同时，模糊的边界和光晕状预测主要源于热扩散、微小目标尺度、边界不确定性以及不足的显式轮廓约束。为了解决这些问题，我们提出了一种自适应对比SLSIoU（AC-SLSIoU），这是一种可插拔的、具有区分性和形状感知的损失函数，旨在用于红外小目标检测。具体而言，引入了Logit域边际约束（LDMC），以扩大目标与信息性难负样本在Logit空间中的响应差距，从而增强弱目标的区分能力。自适应边界抑制（ABS）应用尺度感知的环形惩罚来优化目标轮廓并抑制光晕状溢出响应。此外，虚警聚焦损失为高概率负样本分配更大的权重，进一步惩罚持续的高置信度虚警。在不引入额外推理开销的情况下，所提方法可以无缝集成到现有检测器中，并持续提高检测精度和形状质量。大量实验和跨骨干网络评估证明了所提方法在红外小目标检测中的有效性、鲁棒性和泛化能力。
cs.CV / 17 / 2607.01556
Mind the Gap: Standard 3DGS Evaluation Primarily Measures Near-Trajectory Interpolation
注意差距：标准的3D高斯点云（3DGS）评估主要测量近轨迹插值
Abstract
Standard MipNeRF360-style 3D Gaussian Splatting (3DGS) evaluation holds out every N-th frame -- but these frames have trained neighbors on both sides, so the metric measures near-trajectory interpolation rather than spatial generalization. We introduce a fair matched-count protocol that isolates this effect: both arms train on the same number of images and differ only in whether the holdout is spread evenly (interpolation) or forms a contiguous spatial sector (extrapolation). Our primary finding is a large, consistent interpolation-extrapolation gap of 3~12dB -- several times the differences typically reported between competing methods. The gap is robust to training noise, is in two cases large enough to flip a method ranking under multi-seed confirmation, and -- crucially -- persists across three representation families, including a non-Gaussian volumetric neural radiance field (NeRF), so it reflects spatial coverage rather than any one representation. Diagnostically, it is dominated by a diffuse/geometry-proxy component and tracks each view's angular distance to its nearest training view, a zero-cost signal that also guides capture planning; loss-side regularization yields only marginal gains. Standard holdouts remain useful for near-trajectory rendering but should not, alone, be read as evidence of spatial generalization. Prior work notes protocol sensitivity; ours is, to our knowledge, the first to combine matched-count paired holdout, cross-representation quantification, and a diagnostic analysis Table 1. We describe a spatial-holdout benchmark toolkit with standardized splits and baselines for 16 scenes, which we are preparing for public release.
Chinese Translation
标准的MipNeRF360风格3D高斯点云（3DGS）评估每隔N帧进行保留，但这些帧的两侧都有训练过的邻居，因此该指标测量的是近轨迹插值而非空间泛化。我们引入了一种公平的匹配计数协议，以隔离这一效应：两个实验组在相同数量的图像上进行训练，仅在保留方式上有所不同（均匀分布的插值或形成连续的空间区域的外推）。我们的主要发现是存在一个大且一致的插值-外推差距，范围为3~12dB——这一差距是通常报告的竞争方法之间差异的几倍。该差距对训练噪声具有鲁棒性，在两个案例中足够大以改变多种子确认下的方法排名，并且——至关重要的是——在包括非高斯体积神经辐射场（NeRF）在内的三种表示家族中持续存在，因此它反映的是空间覆盖而非某一种特定表示。从诊断上看，该差距主要由扩散/几何代理成分主导，并跟踪每个视图与其最近训练视图的角距离，这是一个零成本信号，也指导捕获规划；损失侧的正则化仅带来边际收益。标准的保留方法在近轨迹渲染中仍然有用，但不应单独被视为空间泛化的证据。先前的研究指出了协议的敏感性；据我们所知，我们的研究是首次结合匹配计数配对保留、跨表示量化和诊断分析的研究。我们描述了一种空间保留基准工具包，具有标准化的分割和16个场景的基线，我们正在准备公开发布。
cs.CV / 18 / 2607.01578
MVFusion-GS: Motion-Variance Guided Temporal Attention for High-Quality Dynamic Gaussian Splatting
MVFusion-GS：基于运动方差引导的高质量动态高斯点云时间注意力
Abstract
3D Gaussian Splatting (3DGS) enables real-time novel view synthesis for static scenes. Extending it to dynamic scenes via deformation fields has recently attracted significant attention, particularly for dynamic scene reconstructionband distractor-free. However, existing deformation networks lack explicit motion awareness: they neither capture long-term motion intensity nor exploit short-term temporal coherence, leading to inaccurate foreground deformation and pseudo-static residuals in the background. We present MVFusion-GS, a method that enhances deformation networks with two complementary motion-aware mechanisms. The Motion-Variance Guided Refinement aggregates per-Gaussian deformation statistics across time to estimate motion variance and uses it to guide dynamic-static separation during deformation prediction. The MotionFormer Temporal Attention module applies Transformer self-attention over neighboring timesteps to model local motion dependencies and improve temporal consistency. Extensive experiments on both dynamic scene reconstruction and distractor-free reconstruction benchmarks demonstrate state-of-the-art performance, showing that explicit motion awareness improves both foreground motion modeling and static background reconstruction.
Chinese Translation
3D高斯点云（3DGS）使静态场景的实时新视图合成成为可能。通过变形场将其扩展到动态场景最近引起了显著关注，特别是在动态场景重建和无干扰重建方面。然而，现有的变形网络缺乏明确的运动感知：它们既未捕捉长期运动强度，也未利用短期时间一致性，导致前景变形不准确和背景中的伪静态残差。我们提出了MVFusion-GS，这是一种通过两种互补的运动感知机制增强变形网络的方法。运动方差引导的细化（Motion-Variance Guided Refinement）聚合了随时间变化的每个高斯变形统计数据，以估计运动方差，并利用其指导变形预测中的动态-静态分离。MotionFormer时间注意力模块在相邻时间步上应用Transformer自注意力，以建模局部运动依赖性并改善时间一致性。在动态场景重建和无干扰重建基准上的大量实验表明，该方法达到了最先进的性能，显示出明确的运动感知改善了前景运动建模和静态背景重建。
cs.CV / 19 / 2607.01586
VLAFlow: A Unified Training Framework for Vision-Language-Action Models via Co-training and Future Latent Alignment
VLAFlow：通过共同训练和未来潜在对齐的视觉-语言-行动模型统一训练框架
Abstract
Vision-language-action models (VLAs) have recently advanced robotic manipulation, yet the effects of different robot-data pre-training paradigms remain difficult to compare because existing models often differ in architecture, data, action space, and evaluation protocol. We present VLAFlow (Vision-Language-Action Flow), a unified flow-matching framework for controlled comparison of VLA training objectives. Using a heterogeneous robot corpus, OXEMix, containing approximately 5,000 hours of data from DROID, OpenX-Embodiment, OpenX-Augmented, and RoboCOIN, we evaluate four paradigms under the same pi0-style architecture, shared VLM backbone, action expert, and 14-dimensional action space: action-only modeling (MindPI), language-supervised co-training (MindLPI), future latent alignment (MindWPI), and their combination (MindLWPI). Experiments on LIBERO, LIBERO-Plus, and SimplerEnv show that action-only pre-training is sensitive to heterogeneous data. In contrast, language supervision helps preserve vision-language generalization, while future latent alignment improves state-transition and action-outcome modeling. By combining both signals, MindLWPI achieves the most stable overall transfer performance across benchmarks. These results suggest a meta-action space view: language and future latent representations provide complementary intermediate constraints that make heterogeneous action supervision smoother and more transferable.
Chinese Translation
视觉-语言-行动模型（VLA）最近在机器人操作方面取得了进展，但由于现有模型在架构、数据、行动空间和评估协议上往往存在差异，因此不同机器人数据预训练范式的效果仍然难以比较。我们提出了VLAFlow（视觉-语言-行动流），这是一个统一的流匹配框架，用于对VLA训练目标进行受控比较。我们使用一个异构机器人语料库OXEMix，该语料库包含来自DROID、OpenX-Embodiment、OpenX-Augmented和RoboCOIN的约5000小时数据，在相同的pi0风格架构、共享的VLM骨干网络、行动专家和14维行动空间下评估四种范式：仅行动建模（MindPI）、语言监督共同训练（MindLPI）、未来潜在对齐（MindWPI）及其组合（MindLWPI）。在LIBERO、LIBERO-Plus和SimplerEnv上的实验表明，仅行动预训练对异构数据敏感。相比之下，语言监督有助于保持视觉-语言的泛化，而未来潜在对齐则改善了状态转移和行动结果建模。通过结合这两种信号，MindLWPI在各基准测试中实现了最稳定的整体迁移性能。这些结果表明了一种元行动空间视角：语言和未来潜在表示提供了互补的中间约束，使得异构行动监督更加平滑和可迁移。
cs.CV / 20 / 2607.01626
Multi-THuMBS: Multi-person Tracking of 3D Human Meshes Beyond Video Shots
Multi-THuMBS：超越视频镜头的多人人体网格三维跟踪
Abstract
Tracking multi-person 3D human meshes from in-the-wild videos is a highly challenging problem due to complex interactions, frequent occlusions, and severe truncation inherent in unconstrained environments. While recent approaches have improved robustness against these issues, they largely overlook the critical challenge prevalent in real-world footage: frequent shot changes. These abrupt transitions in camera viewpoints often cause existing methods to lose track of human identities and fail in reconstructing temporally coherent trajectories. Although several recent works have explored 3D human mesh tracking under shot changes, they are still limited to single-person scenarios, making them inadequate for real-world videos where multiple people interact and appear simultaneously. To address this limitation, we propose Multi-THuMBS (Multi-person Tracking of 3D Human Meshes Beyond Video Shots) that leverages a state-of-the-art 3D scene prior to reconstruct the two boundary frames in a single shared 3D space. Human meshes are then registered within the shared 3D space, maintaining per-person identity and motion consistency across shot changes. Extensive experiments demonstrate that our approach yields significant improvements in 3D human mesh recovery, camera pose estimation, and identity tracking, thereby ensuring high-fidelity motion reconstruction with consistent identity preservation across shots compared to previous state-of-the-art methods.
Chinese Translation
从野外视频中跟踪多人人体三维网格是一个高度具有挑战性的问题，原因在于复杂的交互、频繁的遮挡以及无约束环境中固有的严重截断。尽管最近的方法在应对这些问题的鲁棒性方面有所改善，但它们在很大程度上忽视了现实世界视频中普遍存在的关键挑战：频繁的镜头切换。这些相机视角的突然变化往往导致现有方法无法持续跟踪人类身份，并在重建时间一致的轨迹时失败。尽管一些近期的研究探索了在镜头切换下的三维人体网格跟踪，但它们仍然局限于单人场景，因此在多个个体同时互动并出现的现实视频中显得不够充分。为了解决这一局限性，我们提出了Multi-THuMBS（超越视频镜头的多人人体网格三维跟踪），该方法利用最先进的三维场景先验在单一共享的三维空间中重建两个边界帧。然后，在共享的三维空间中注册人体网格，保持每个人的身份和运动在镜头切换中的一致性。大量实验表明，我们的方法在三维人体网格恢复、相机姿态估计和身份跟踪方面取得了显著改善，从而确保了与之前的最先进方法相比，在镜头间保持一致身份的高保真运动重建。
cs.CV / 21 / 2607.01628
Online Segment 3D Gaussians via Launching Virtual Drones
通过发射虚拟无人机进行在线分割3D高斯
Abstract
Interactive segmentation of 3D Gaussians offers a compelling opportunity for real-time manipulation of 3D scenes, thanks to the real-time rendering capability of 3D Gaussian Splatting (3DGS). However, existing methods require a time-consuming per-scene setup - typically tens of seconds or even minutes - before interactive segmentation can begin on a raw 3DGS scene. This setup involves multi-view mask preparation, mask lifting, and feature distillation, creating a major bottleneck for online applications. To address this limitation, we aim to completely eliminate the setup stage for interactive 3DGS segmentation while keeping the segmentation time practical (under 1 second). In this work, we present SAGO (Segment Any Gaussians Online), a novel setup-free framework for interactive 3DGS segmentation. By introducing virtual drones, our method reframes the 3D segmentation problem as an online Next-Best-View (NBV) planning task formulated within a Markov process. Extensive experiments demonstrate that SAGO can extract clean 3D assets directly from 3D Gaussians with sub-second latency, thereby enabling a broad range of downstream applications such as object manipulation and scene editing. Moreover, our method achieves over a 50x speedup compared to the previous setup-free 3DGS segmentation frameworks.
Chinese Translation
3D高斯的交互式分割为实时操作3D场景提供了一个引人注目的机会，这得益于3D高斯点云渲染（3DGS）的实时渲染能力。然而，现有方法在开始对原始3DGS场景进行交互式分割之前，通常需要耗时的每场景设置——通常需要数十秒甚至几分钟。这个设置过程涉及多视角掩膜准备、掩膜提升和特征提取，成为在线应用的主要瓶颈。为了解决这一限制，我们的目标是完全消除交互式3DGS分割的设置阶段，同时保持分割时间在可接受范围内（1秒以内）。在这项工作中，我们提出了SAGO（在线分割任意高斯），一个新颖的无设置框架用于交互式3DGS分割。通过引入虚拟无人机，我们的方法将3D分割问题重新构建为一个在线的下一个最佳视图（Next-Best-View, NBV）规划任务，该任务在马尔可夫过程内进行建模。大量实验表明，SAGO能够以亚秒延迟直接从3D高斯中提取干净的3D资产，从而支持广泛的下游应用，如物体操作和场景编辑。此外，我们的方法相比于之前的无设置3DGS分割框架实现了超过50倍的加速。
cs.CV / 22 / 2607.01630
DRDN: Decoupled Representation Dynamic Network for From-Scratch ViT Class-Incremental Learning
DRDN：用于从头开始的视觉变换器类增量学习的解耦表示动态网络
Abstract
Dynamic expansion methods for class-incremental learning (CIL) protect task-specific knowledge by growing dedicated tokens or subnetworks, yet our analyses suggest that classification supervision alone does not sufficiently preserve task-agnostic shared backbone representations over long incremental sequences. We identify two intertwined challenges: cross-task confusion from sequential training on predominantly current-task data, which biases decision boundaries toward recent tasks; and under-optimized shared representations in the backbone that cap long-term discriminability as tasks accumulate. We propose the Decoupled Representation Dynamic Network (DRDN), which addresses these challenges via two orthogonal mechanisms. For shared backbone representations, DRDN continuously applies masked image modeling (MIM) at every incremental step, with reconstruction gradients routed exclusively through the backbone, encouraging it to retain general visual structure beyond class-discriminative cues. For task-specific discrimination, DRDN employs hierarchical task token expansion across all transformer layers, with a modified per-task attention rule that reduces inter-task interference. We support this design with accuracy degradation analysis and cross-task confusion rate measurements. In the from-scratch ViT CIL setting (no external pretraining), DRDN consistently improves over strong token-expansion baselines with comparable backbone scale. On CIFAR100-B0 (10 steps), DRDN achieves 77.19% average accuracy, outperforming DKT by 1.36 points and DyTox by 3.53 points, with an advantage that grows at longer incremental sequences. Multi-seed validation confirms stability (+/-0.31%). The MIM decoder is active only during training, adding no inference-time parameters or computation.
Chinese Translation
类增量学习（CIL）的动态扩展方法通过增长专用的标记或子网络来保护任务特定知识，但我们的分析表明，仅依靠分类监督并不足以在长增量序列中充分保留任务无关的共享主干表示。我们识别出两个相互交织的挑战：来自主要当前任务数据的顺序训练导致的跨任务混淆，这使得决策边界偏向于最近的任务；以及在主干中未优化的共享表示，限制了随着任务积累而产生的长期可区分性。我们提出了解耦表示动态网络（DRDN），通过两个正交机制来解决这些挑战。对于共享主干表示，DRDN在每个增量步骤中持续应用掩码图像建模（MIM），重建梯度仅通过主干路由，鼓励其保留超越类区分线索的一般视觉结构。对于任务特定的区分，DRDN在所有变换器层中采用分层任务标记扩展，并使用修改后的每任务注意力规则来减少任务间干扰。我们通过准确性下降分析和跨任务混淆率测量来支持这一设计。在从头开始的ViT CIL设置中（无外部预训练），DRDN在与可比主干规模的强标记扩展基线相比时，始终表现出改善。在CIFAR100-B0（10步）上，DRDN实现了77.19%的平均准确率，超越了DKT 1.36个百分点和DyTox 3.53个百分点，且在更长的增量序列中优势更为明显。多种种子验证确认了稳定性（+/-0.31%）。MIM解码器仅在训练期间激活，不增加推理时的参数或计算。
cs.CV / 23 / 2607.01633
Bridging 3D Gaussians and Semantic Occupancy for Comprehensive Open-Vocabulary Scene Understanding from Unposed Images
桥接三维高斯与语义占用，实现无姿态图像的全面开放词汇场景理解
Abstract
Comprehensive 3D scene understanding from sparse, unposed images requires a model to recover renderable geometry, open-vocabulary semantics, and free/occupied 3D space without relying on external camera calibration. Recent feed-forward Gaussian methods improve pose-free reconstruction and semantic rendering, but their Gaussian primitives are mainly optimized through image-space objectives and remain weakly constrained in unobserved regions. We propose \textit{COVScene}, a pose-free semantic Gaussian framework that couples renderable Gaussian primitives with a dense semantic occupancy field through differentiable volumetric lifting. Instead of converting Gaussians to voxels only at evaluation time, COVScene lifts the predicted semantic Gaussians inside the training computation graph, so volumetric regularization provides gradients to Gaussian opacity, geometry, and semantic features. The framework combines a semantic-aware Geometry Transformer, multi-task Gaussian decoding, geometric foundation distillation, and occupancy entropy regularization to support novel view synthesis, open-vocabulary semantic querying, and semantic occupancy prediction within a single representation. Experiments on ScanNet and ScanNet++ show that COVScene maintains competitive rendering quality, improves open-vocabulary segmentation, and achieves stronger semantic occupancy prediction than the self-supervised baseline without direct voxel-level supervision.
Chinese Translation
从稀疏的无姿态图像中实现全面的三维场景理解需要一个模型来恢复可渲染的几何形状、开放词汇的语义以及自由/占用的三维空间，而无需依赖外部相机标定。近期的前馈高斯方法改善了无姿态重建和语义渲染，但其高斯原语主要通过图像空间目标进行优化，在未观察到的区域仍然约束较弱。我们提出了 extit{COVScene}，一个无姿态的语义高斯框架，通过可微分的体积提升将可渲染的高斯原语与密集的语义占用场耦合。COVScene在训练计算图内提升预测的语义高斯，而不是仅在评估时将高斯转换为体素，因此体积正则化为高斯的不透明度、几何形状和语义特征提供了梯度。该框架结合了语义感知的几何变换器、多任务高斯解码、几何基础蒸馏和占用熵正则化，以支持新视图合成、开放词汇语义查询和语义占用预测，均在单一表示中实现。对ScanNet和ScanNet++的实验表明，COVScene保持了竞争性的渲染质量，改善了开放词汇分割，并在没有直接体素级监督的情况下实现了比自监督基线更强的语义占用预测。
cs.CV / 24 / 2607.01642
Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling
多分辨率流匹配：通过分阶段采样实现无训练扩散加速
Abstract
Hardware-agnostic strategies for accelerating text-to-image diffusion, such as timestep distillation and feature caching, can reduce inference time without custom kernels or system-level optimization. Among them, multi-resolution generation strategies have recently received broad attention, attaining more than 5x speedup without any training. However, the design of performing upsampling in the latent space, together with the selective modification of partial regions, causes these methods to exhibit noticeable blurring or artifacts. To this end, we propose MrFlow, a training-free multi-resolution acceleration strategy for pretrained flow-matching models built upon a staged low-to-high-resolution pipeline. MrFlow first rapidly generates the main structure at low resolution, then performs super-resolution in the pixel space using a lightweight pretrained GAN-based model, subsequently injects low-strength noise to enable high-frequency resampling, and finally refines the details at high resolution. Quantitative and qualitative results on FLUX.1-dev and Qwen-Image show that MrFlow exploits the quadratic token reduction and reduced step requirement of low-resolution sampling to achieve 10x end-to-end acceleration while keeping OneIG within a 1% gap relative to that before acceleration, significantly surpassing other training-free acceleration strategies, and requiring no training or runtime dynamic identification whatsoever. MrFlow can further be directly combined orthogonally with pre-trained timestep distillation strategies, achieving even higher generation acceleration of up to 25x.
Chinese Translation
硬件无关的文本到图像扩散加速策略，如时间步蒸馏和特征缓存，可以在没有自定义内核或系统级优化的情况下减少推理时间。其中，多分辨率生成策略最近受到广泛关注，实现了超过5倍的加速而无需任何训练。然而，在潜在空间中进行上采样的设计，以及对部分区域的选择性修改，导致这些方法出现明显的模糊或伪影。为此，我们提出了MrFlow，一种基于分阶段低到高分辨率管道的无训练多分辨率加速策略，适用于预训练的流匹配模型。MrFlow首先在低分辨率下快速生成主要结构，然后使用轻量级的基于GAN的预训练模型在像素空间中执行超分辨率，随后注入低强度噪声以实现高频重采样，最后在高分辨率下细化细节。在FLUX.1-dev和Qwen-Image上的定量和定性结果表明，MrFlow利用二次令牌减少和低分辨率采样的减少步骤要求，实现了10倍的端到端加速，同时保持OneIG与加速前的1%差距，显著超越其他无训练加速策略，并且完全不需要训练或运行时动态识别。MrFlow还可以与预训练的时间步蒸馏策略正交直接结合，实现高达25倍的生成加速。
cs.CV / 25 / 2607.01648
Boosting Ultrasound Image Classification via Attribute-Guided Dual-Branch Framework
通过属性引导的双分支框架提升超声图像分类
Abstract
Ultrasound image classification is essential for computer-aided diagnosis. However, current methods often neglect clinical priors, leading to poor generalization in challenging scenarios and a lack of interpretability that limits clinical adoption. To address these issues, we aim to develop a medical-prior module that can be seamlessly integrated into existing pipelines to enhance both diagnostic performance and interpretability. In this paper, we propose an attribute-guided dual-branch framework for ultrasound classification that introduces domain-agnostic medical attribute priors, improving generalization while offering interpretable evidence. Specifically, a baseline branch follows conventional architectures and predicts image categories via a fully connected classifier. An attribute-guided branch injects domain-agnostic attributes as priors and produces human-interpretable decision cues. Finally, an adaptive decision module fuses the two branches in a data-dependent manner to yield the final prediction. Experiments across diverse ultrasound classification tasks demonstrate that our approach can be integrated into multiple backbones and state-of-the-art methods with low overhead, consistently improving accuracy and interpretability. Code is available at: https://github.com/zhaobo253-crypto/AttrGuide.
Chinese Translation
超声图像分类对于计算机辅助诊断至关重要。然而，当前的方法往往忽视临床先验，导致在具有挑战性的场景中泛化能力差，并缺乏可解释性，限制了临床应用。为了解决这些问题，我们旨在开发一个医疗先验模块，可以无缝集成到现有流程中，以提高诊断性能和可解释性。本文提出了一种用于超声分类的属性引导双分支框架，引入了与领域无关的医疗属性先验，改善了泛化能力，同时提供了可解释的证据。具体而言，基线分支遵循传统架构，通过全连接分类器预测图像类别。属性引导分支将与领域无关的属性作为先验注入，并生成可供人类理解的决策线索。最后，一个自适应决策模块以数据依赖的方式融合两个分支，以产生最终预测。在多种超声分类任务中的实验表明，我们的方法可以低开销地集成到多个主干和最先进的方法中，始终提高准确性和可解释性。代码可在以下链接获取：https://github.com/zhaobo253-crypto/AttrGuide。
cs.CV / 26 / 2607.01654
Plug-and-Play Volumetric Reconstruction for Compressive Sensing Light-Sheet Microscopy
即插即用的压缩感知光片显微镜体积重建
Abstract
We investigate volumetric reconstruction for compressive sensing light-sheet microscopy (CS-LSM), where fast volumetric imaging is achieved by encoding multiple axial planes into each camera exposure. To recover the underlying volume from highly multiplexed measurements, we propose a plug-and-play (PnP) framework that flexibly incorporates any user-specified denoiser into the reconstruction process. Building on a slice-based formulation, we further introduce an axial-coupled model that exploits correlations between adjacent slices to improve volumetric continuity. For efficient computation, we derive a Woodbury-based update for the data-consistency step in both the slice-based and axial-coupled formulations, and employ a Gauss-Seidel sweep for the denoising step in the axial-coupled model. Under a weakly convex regularization assumption, we establish subsequential convergence of the proposed algorithm. Experiments on synthetic and real zebrafish-heart data demonstrate that the proposed framework successfully recovers cellular structures from compressed measurements, and provide practical insights into the comparative performance of commonly used denoisers within the PnP framework under the CS-LSM setup.
Chinese Translation
我们研究了压缩感知光片显微镜（CS-LSM）的体积重建，其中通过将多个轴向平面编码到每次相机曝光中，实现快速的体积成像。为了从高度多重化的测量中恢复底层体积，我们提出了一种即插即用（PnP）框架，该框架灵活地将任何用户指定的去噪器纳入重建过程。在基于切片的公式基础上，我们进一步引入了一种轴向耦合模型，利用相邻切片之间的相关性来改善体积连续性。为了提高计算效率，我们推导了基于Woodbury的更新，用于切片基础和轴向耦合公式中的数据一致性步骤，并在轴向耦合模型中采用高斯-赛德尔（Gauss-Seidel）迭代进行去噪步骤。在弱凸正则化假设下，我们建立了所提算法的子序列收敛性。对合成和真实斑马鱼心脏数据的实验表明，所提框架成功地从压缩测量中恢复细胞结构，并提供了在CS-LSM设置下，常用去噪器在PnP框架内比较性能的实用见解。
cs.CV / 27 / 2607.01657
Domain Generalization via Text-Anchored Information Bottleneck
通过文本锚定的信息瓶颈实现领域泛化
Abstract
Visual recognition models often fail when deployed in new environments. Domain Generalization (DG) addresses this by learning representations that remain invariant to environment-specific variations. Recent approaches increasingly rely on large vision-language models, assuming that preserving their expressive visual representations improves robustness. However, we show that such visual expressiveness can instead propagate spurious cues that tie representations to the training environments, hindering invariant learning. We therefore discard visual guidance and instead treat the language embedding space as the primary source of domain invariance, naturally acting as an information bottleneck that preserves core semantics while suppressing domain-specific variations. Extensive experiments across diverse backbones exhibit state-of-the-art performance and further analyze what makes guidance effective for robust generalization. These findings shift the focus of DG from improving representations to designing supervision that enforces invariance.
Chinese Translation
视觉识别模型在新环境中部署时常常失败。领域泛化（Domain Generalization, DG）通过学习对环境特定变化保持不变的表征来解决这一问题。近期的方法越来越依赖于大型视觉-语言模型，假设保持其富有表现力的视觉表征能够提高鲁棒性。然而，我们表明，这种视觉表现力反而可能传播与训练环境相关的虚假线索，阻碍不变学习。因此，我们摒弃视觉指导，而是将语言嵌入空间视为领域不变性的主要来源，自然地充当信息瓶颈，保留核心语义，同时抑制领域特定的变化。在多种基础模型上的广泛实验展示了最先进的性能，并进一步分析了什么使得指导在鲁棒泛化中有效。这些发现将DG的重点从改善表征转向设计强制不变性的监督。
cs.CV / 28 / 2607.01658
Teaching Vision-Language-Action Models What to See and Where to Look
教导视觉-语言-动作模型该看什么以及如何观察
Abstract
Vision-Language-Action (VLA) models have emerged as a promising paradigm for end-to-end autonomous driving. However, existing VLAs' training relies heavily on text-centric visual question answering and chain-of-thought reasoning data, which emphasizes linguistic reasoning rather than action-grounded planning. As a result, the learned representations capture semantic knowledge but lack spatial dependencies crucial for reliable trajectory prediction. We propose DriveTeach-VLA, a framework that explicitly teaches VLAs what to see and where to look. Driving-aware Vision Distillation (DVD) injects driving-specific perceptual priors into the vision encoder, while 2D Trajectory-Guided Prompts (2D-TGP) provide spatial conditioning aligned with feasible driving trajectories. Together, they form a vision-guided learning pipeline: what to see (DVD pretraining) - where to look (TGP-guided SFT) - how to act (TGP-guided GRPO). DriveTeach-VLA achieves the state-of-the-art performance on NAVSIM and nuScenes. Our code is available at: https://github.com/ShivaTeam/DriveTeach-VLA.
Chinese Translation
视觉-语言-动作（VLA）模型已成为端到端自主驾驶的一个有前景的范式。然而，现有的VLA训练在很大程度上依赖于以文本为中心的视觉问答和思维链推理数据，这强调了语言推理而非基于行动的规划。因此，所学习的表示捕捉了语义知识，但缺乏对可靠轨迹预测至关重要的空间依赖关系。我们提出了DriveTeach-VLA，一个明确教导VLA该看什么以及如何观察的框架。驾驶感知视觉蒸馏（DVD）将特定于驾驶的感知先验注入视觉编码器，而2D轨迹引导提示（2D-TGP）提供与可行驾驶轨迹对齐的空间条件。它们共同形成了一个视觉引导学习管道：该看什么（DVD预训练）- 如何观察（TGP引导的SFT）- 如何行动（TGP引导的GRPO）。DriveTeach-VLA在NAVSIM和nuScenes上实现了最先进的性能。我们的代码可在以下链接获取：https://github.com/ShivaTeam/DriveTeach-VLA。
cs.CV / 29 / 2607.01663
Unified Panoramic-Gaussian Representation for Monocular 4D Scene Synthesis
统一的全景高斯表示用于单目4D场景合成
Abstract
4D scene synthesis from monocular videos has made significant progress in recent years. However, existing methods are typically constrained by view interpolation. As a result, they struggle to infer unseen regions beyond the observed views. In this paper, we reformulate the task as 4D scene synthesis with unseen regions, which extends beyond traditional interpolation settings. Camera-conditioned video generation enables unseen region synthesis by guiding generation along specified cameras. However, these methods lack explicit 3D priors and are optimized with random camera trajectories. This design leads to severe inconsistencies under large trajectory deviations. To address this limitation, we build a unified training and inference framework with panoramic trajectory guidance. While this design improves cross-view consistency, the panoramic representation alone fails to model dynamic content effectively. Object motion in panoramic space introduces scale and shape distortions. To address this, we propose PanoGaussian, a unified Panoramic-Gaussian representation that distills the panoramic representation into an explicit dynamic Gaussian representation to capture dynamic physical priors of the 4D scene. Experiments demonstrate that PanoGaussian achieves consistent 4D scene synthesis even under large viewpoint variations.
Chinese Translation
近年来，从单目视频中进行4D场景合成取得了显著进展。然而，现有方法通常受到视图插值的限制，因此在推断未观察区域时面临困难。本文将任务重新表述为包含未观察区域的4D场景合成，超越了传统的插值设置。基于相机条件的视频生成通过沿指定相机引导生成来实现未观察区域的合成。然而，这些方法缺乏明确的3D先验，并且在随机相机轨迹下进行优化。这种设计在大轨迹偏差下导致严重的不一致性。为了解决这一限制，我们构建了一个统一的训练和推理框架，采用全景轨迹引导。尽管这一设计改善了跨视图一致性，但单靠全景表示无法有效建模动态内容。在全景空间中的物体运动引入了尺度和形状的失真。为此，我们提出了PanoGaussian，一种统一的全景高斯表示，它将全景表示提炼为明确的动态高斯表示，以捕捉4D场景的动态物理先验。实验表明，即使在大视点变化下，PanoGaussian也能够实现一致的4D场景合成。
cs.CV / 30 / 2607.01667
Temporal and Cross-Modal Alignment for Enhanced Audiovisual Video Captioning
增强视听视频字幕的时间与跨模态对齐
Abstract
While Multimodal Large Language Models (MLLMs) have advanced video understanding, achieving precise temporal and cross-modal alignment in audiovisual video captioning remains a formidable challenge. Most existing approaches suffer from modality detachment and temporal incoherence, failing to accurately bind auditory events to visual entities or capture complex causal dynamics. To address these deficiencies, we propose TCA-Captioner, a framework specifically engineered to enhance Temporal and Cross-Modal Alignment for audiovisual video captioning. We first introduce the Observer-Checker-Corrector (OCC) framework, an iterative refinement strategy that generates high-fidelity, meticulously grounded training data. Leveraging a curated high-density human interaction dataset, TCA-Captioner is optimized to model sophisticated audiovisual interactions. Furthermore, we present TCA-Bench, a diagnostic benchmark utilizing a Decoupled Evaluation Protocol to isolate and quantify model proficiency in audiovisual binding and temporal relational reasoning. Extensive experiments demonstrate that TCA-Captioner sets a new standard for temporally-coherent and synchronized audiovisual narratives.
Chinese Translation
尽管多模态大型语言模型（MLLMs）在视频理解方面取得了进展，但在视听视频字幕中实现精确的时间和跨模态对齐仍然是一项艰巨的挑战。现有的大多数方法存在模态脱节和时间不一致的问题，未能准确将听觉事件与视觉实体绑定或捕捉复杂的因果动态。为了解决这些不足，我们提出了TCA-Captioner，这是一个专门设计的框架，旨在增强视听视频字幕的时间与跨模态对齐。我们首先介绍了观察者-检查者-修正者（Observer-Checker-Corrector, OCC）框架，这是一种迭代精炼策略，生成高保真、经过精心验证的训练数据。利用一个精心策划的高密度人类互动数据集，TCA-Captioner被优化以建模复杂的视听互动。此外，我们还提出了TCA-Bench，这是一个利用解耦评估协议的诊断基准，用于隔离和量化模型在视听绑定和时间关系推理方面的能力。大量实验表明，TCA-Captioner为时间一致和同步的视听叙事设定了新的标准。
cs.CV / 31 / 2607.01675
HistoSeg++: Delving deeper with attention and multiscale feature fusion for biomarker segmentation
HistoSeg++：通过注意力机制和多尺度特征融合深入探讨生物标志物分割
Abstract
Segmentation of biomarkers in medical images is frequently viewed as a first step towards medical image analysis in any bioinformatics or biomedical application. Despite progress, existing methods still struggle to capture information at multiple scales and to perform upsampling effectively across different datasets. These shortcomings often result in suboptimal generalization capabilities. Recently, architectures belonging to the Nested-UNet family excel in capturing multiscale contextual information and upsample them effectively. In this work, We propose a novel Nested-UNet architecture that effectively captures multi-scale contextual information. It includes inner and outer attention units to enhance focus during upsampling, along with channel-wise feature recalibration using squeeze-and-excitation modules, leading to improved segmentation performance. Additionally, the architecture integrates an edge-aware loss to emphasize boundary accuracy by assigning greater importance to edge regions. Tested extensively on three publicly available benchmark datasets. Our method demonstrates a generalization performance superior to existing Nested-UNet methods. Code: https://github.com/saadwazir/histosegplusplus
Chinese Translation
在医学图像中对生物标志物的分割通常被视为生物信息学或生物医学应用中医学图像分析的第一步。尽管已有进展，但现有方法仍然难以在多个尺度上捕捉信息，并在不同数据集之间有效地进行上采样。这些不足往往导致次优的泛化能力。最近，属于Nested-UNet家族的架构在捕捉多尺度上下文信息和有效上采样方面表现出色。在本研究中，我们提出了一种新颖的Nested-UNet架构，能够有效捕捉多尺度上下文信息。该架构包括内外注意力单元，以增强上采样过程中的聚焦，同时利用压缩-激励模块进行通道特征的重新校准，从而提高分割性能。此外，该架构集成了一种边缘感知损失，以通过赋予边缘区域更大的重要性来强调边界准确性。在三个公开可用的基准数据集上进行了广泛测试。我们的方法展示了优于现有Nested-UNet方法的泛化性能。代码链接：https://github.com/saadwazir/histosegplusplus
cs.CV / 32 / 2607.01677
ICDepth: Taming Video Diffusion Models for Video Depth Estimation via In-Context Conditioning
ICDepth：通过上下文条件调节驯化视频扩散模型以进行视频深度估计
Abstract
Monocular video depth estimation requires temporal consistency, geometric accuracy, and generalization across diverse scenarios, yet existing methods struggle to achieve all three simultaneously. Discriminative models excel at per-frame accuracy but suffer from temporal drift due to limited context windows, while generative methods improve consistency and generalization at the cost of extensive training data (10M+ samples) and lack of geometric precision. In response to these issues, we introduce \textbf{ICDepth}, a framework that adapts pre-trained text-to-video diffusion transformers for video depth estimation via In-Context Conditioning (ICC), leveraging their rich spatial-temporal priors. To address key challenges in transferring ICC from generation to dense prediction, we propose: (1)~\textbf{SAND-Attention}, which ensures precise spatial-temporal alignment via shared RoPE and enforces unidirectional attention to prevent noise contamination; (2)~\textbf{SRFM}, which injects DINOv2 semantic and resolution priors to enhance geometric precision. ICDepth achieves state-of-the-art results on multiple benchmarks with remarkable data efficiency, trained on only 0.8M frames ($6$--$13\times$ less than competing generative methods), while demonstrating strong zero-shot generalization to diverse domains.
Chinese Translation
单目视频深度估计需要时间一致性、几何精度和在多样场景中的泛化能力，但现有方法难以同时实现这三者。判别模型在每帧的准确性上表现出色，但由于上下文窗口的限制，容易出现时间漂移；而生成方法则在一致性和泛化能力上有所提升，但代价是需要大量的训练数据（超过1000万样本）且缺乏几何精度。针对这些问题，我们提出了 extbf{ICDepth}，这是一个通过上下文条件调节（In-Context Conditioning, ICC）将预训练的文本到视频扩散变换器适配用于视频深度估计的框架，利用其丰富的时空先验。为了解决从生成到密集预测转移ICC的关键挑战，我们提出了：（1） extbf{SAND-Attention}，通过共享的RoPE确保精确的时空对齐，并强制单向注意力以防止噪声污染；（2） extbf{SRFM}，注入DINOv2的语义和分辨率先验以增强几何精度。ICDepth在多个基准测试中取得了最先进的结果，展现出卓越的数据效率，仅在0.8M帧上训练（比竞争的生成方法少$6$--$13 imes$），同时在多样领域中表现出强大的零样本泛化能力。
cs.CV / 33 / 2607.01698
Structure-Aware Gaussian Splatting for Large-Scale Scene Reconstruction
结构感知高斯点云用于大规模场景重建
Abstract
3D Gaussian Splatting has demonstrated remarkable potential in novel view synthesis. In contrast to small-scale scenes, large-scale scenes inevitably contain sparsely observed regions with excessively sparse initial points. In this case, supervising Gaussians initialized from low-frequency sparse points with high-frequency images often induces uncontrolled densification and redundant primitives, degrading both efficiency and quality. Intuitively, this issue can be mitigated with scheduling strategies, which can be categorized into two paradigms: modulating target signal frequency via densification and modulating sampling frequency via image resolution. However, previous scheduling strategies are primarily hardcoded, failing to perceive the convergence behavior of scene frequency. To address this, we reframe the scene reconstruction problem from the perspective of signal structure recovery and propose SIG, a novel scheduler that synchronizes image supervision with Gaussian frequencies. Specifically, we derive the average sampling frequency and bandwidth of 3D representations, and then regulate the training image resolution and the Gaussian densification process based on scene frequency convergence. Furthermore, we introduce Sphere-Constrained Gaussians, which leverage the spatial prior of initialized point clouds to control Gaussian optimization. Our framework enables frequency-consistent, geometry-aware, and floater-free training, achieving state-of-the-art performance by a substantial margin in both efficiency and rendering quality in large-scale scenes. The code is available at: https://github.com/weiyixue999/Signal_Structure_Aware_Gaussian
Chinese Translation
3D 高斯点云在新视图合成中展现了显著的潜力。与小规模场景相比，大规模场景不可避免地包含观察稀疏的区域，初始点过于稀疏。在这种情况下，利用低频稀疏点初始化的高斯与高频图像进行监督，往往会导致不受控制的密集化和冗余原语，从而降低效率和质量。直观上，这个问题可以通过调度策略来缓解，这些策略可以分为两种范式：通过密集化调节目标信号频率和通过图像分辨率调节采样频率。然而，之前的调度策略主要是硬编码的，未能感知场景频率的收敛行为。为了解决这个问题，我们从信号结构恢复的角度重新构建场景重建问题，并提出了 SIG，这是一种新颖的调度器，能够将图像监督与高斯频率同步。具体来说，我们推导了 3D 表示的平均采样频率和带宽，然后根据场景频率的收敛性调节训练图像分辨率和高斯密集化过程。此外，我们引入了球体约束高斯（Sphere-Constrained Gaussians），利用初始化点云的空间先验来控制高斯优化。我们的框架实现了频率一致、几何感知和无浮动的训练，在大规模场景中在效率和渲染质量上均取得了显著的领先表现。代码可在以下链接获取：https://github.com/weiyixue999/Signal_Structure_Aware_Gaussian
cs.CV / 34 / 2607.01707
LASER: A Corrective Lens for LVLMs via Visual Attention Preservation and Sink Suppression
LASER：通过视觉注意力保持和沉没抑制为 LVLMs 提供的校正镜头
Abstract
Large vision-language models (LVLMs) exhibit strong reasoning ability but suffer from visual forgetting during long-horizon decoding, where attention progressively drifts away from visual evidence. Existing methods largely treat this issue as a late-stage attention decay problem or attempt to mitigate it through heuristic reminders or post-hoc attention lifting. Through systematic empirical analysis, we find that performance degradation under visual forgetting is largely driven by two overlooked factors: early-stage attention decay disrupts evidence acquisition, and attention concentration on a subset of task-irrelevant visual sink tokens. Motivated by these insights, we propose LASER, a post-training framework that regulates both the visual attention trajectory and intra-visual token attention distribution during reasoning. Technically, LASER introduces two complementary rewards: a Visual Grounding Reward, which encourages the model to maintain attention on semantically salient visual tokens throughout decoding, and a Sink Suppression Reward, which penalizes excessive attention concentration on visual sink tokens. Together, these rewards preserve early-stage grounding while preventing attention collapse onto uninformative regions. Extensive experiments on eight benchmark datasets demonstrate that LASER consistently outperforms strong baselines, validating attention-aware training as an effective remedy for visual forgetting.
Chinese Translation
大型视觉语言模型（LVLMs）展现出强大的推理能力，但在长时间解码过程中却遭遇视觉遗忘，注意力逐渐偏离视觉证据。现有方法主要将此问题视为后期注意力衰退问题，或试图通过启发式提醒或事后注意力提升来缓解。通过系统的实证分析，我们发现视觉遗忘下的性能下降主要由两个被忽视的因素驱动：早期注意力衰退干扰了证据获取，以及对一组与任务无关的视觉沉没标记的注意力集中。基于这些见解，我们提出了 LASER，一个后训练框架，调节推理过程中的视觉注意力轨迹和视觉标记的内部注意力分布。从技术上讲，LASER 引入了两个互补的奖励：视觉定位奖励（Visual Grounding Reward），鼓励模型在整个解码过程中保持对语义显著视觉标记的注意力；沉没抑制奖励（Sink Suppression Reward），惩罚对视觉沉没标记的过度注意力集中。这些奖励共同保持了早期的定位，同时防止注意力崩溃到无信息区域。在八个基准数据集上的大量实验表明，LASER 始终优于强基线，验证了关注注意力的训练作为解决视觉遗忘的有效方法。
cs.CV / 35 / 2607.01708
Consistent Scene Understanding in 3D Gaussian Splatting via Multi-Cue Mask Refinement
通过多线索掩膜细化实现一致的3D高斯点云场景理解
Abstract
Reliable instance-level scene understanding is a fundamental prerequisite for object-level interactions and high-fidelity 3D representations. While current methods often leverage 2D foundation segmentation models to obtain these priors, their 2D-centric design typically yields fragmented masks and inconsistent predictions across different views. To address these issues, we propose a novel framework that produces consistent 2D instance masks to guide the optimization of 3D Gaussian Splatting (3DGS) feature fields. Our framework consists of three main stages. (1) Multi-Cue Extraction that generates synergistic semantic, geometric, and structural priors from input images. (2) Multi-Cue-Guided Mask Merging process that consolidates fragmented masks using a composite merge score derived from semantic, depth, and edge cues. (3) Cross-View Mask Matching that establishes globally consistent identity assignments across all viewpoints. By transforming viewpoint-specific segments into coherent 3D primitives, our approach enables stable 3D instance segmentation and effective downstream editing tasks. Experiments demonstrate that our method significantly improves cross-view consistency and segmentation stability over existing baselines while maintaining high-fidelity photometric reconstruction.
Chinese Translation
可靠的实例级场景理解是物体级交互和高保真3D表示的基本前提。尽管当前方法通常利用2D基础分割模型来获取这些先验信息，但其以2D为中心的设计通常导致掩膜碎片化以及不同视角下预测的不一致性。为了解决这些问题，我们提出了一种新颖的框架，该框架生成一致的2D实例掩膜，以指导3D高斯点云（3DGS）特征场的优化。我们的框架由三个主要阶段组成：(1) 多线索提取，从输入图像中生成协同的语义、几何和结构先验；(2) 多线索引导的掩膜合并过程，利用从语义、深度和边缘线索中得出的复合合并评分来整合碎片化的掩膜；(3) 跨视角掩膜匹配，建立所有视点之间全局一致的身份分配。通过将特定视角的片段转化为一致的3D原语，我们的方法实现了稳定的3D实例分割和有效的下游编辑任务。实验表明，我们的方法在跨视角一致性和分割稳定性方面显著优于现有基线，同时保持高保真的光度重建。
cs.CV / 36 / 2607.01728
Beyond Pixel Diffs: Benchmarking Image Change Captioning for Web UI Visual Regression Testing
超越像素差异：Web UI 视觉回归测试的图像变化描述基准
Abstract
Visual regression testing (VRT) is a standard quality assurance step in modern software release pipelines. On every change, it re-renders user interface (UI) screenshots, compares each one against an approved baseline image, and routes any detected difference to a human reviewer who decides whether it is an intended update or an unintended regression. A widely used approach, especially in open-source and continuous-integration pipelines, is pixel-level comparison, which is semantically blind and treats rendering noise and genuine defects identically, producing large volumes of false positives that force developers and testers to spend substantial time and effort manually reviewing flagged differences at every release cycle. Industry tools apply machine learning to VRT, but lack public evaluation. More critically, no dataset or benchmark exists to support natural language descriptions of UI changes, a capability that tells testers what changed in words instead of leaving them to interpret a binary flag or a highlighted region. To address the gap, we propose a new task, Web UI Image Change Captioning (WUICC), which sits at the intersection of VRT and image difference captioning (IDC), and release WUICC-bench, its first dataset and benchmark for the task. We evaluate eleven representative IDC methods, together with two zero-shot general-purpose LLMs. We find that: (1) these methods tend to struggle in the Web UI domain due to its layout diversity, dense text, and fine-grained changes, and (2) yet the trained methods already suppress non-meaningful visual noise far more selectively than the pixel-level comparison VRT relies on, providing a solid foundation for future domain-specific research.
Chinese Translation
视觉回归测试（VRT）是现代软件发布流程中的标准质量保证步骤。在每次更改时，它会重新渲染用户界面（UI）截图，将每个截图与经过批准的基线图像进行比较，并将检测到的差异发送给人工审核者，由其决定这是预期更新还是意外回归。一种广泛使用的方法，尤其是在开源和持续集成流程中，是像素级比较，这种方法在语义上是盲目的，将渲染噪声和真实缺陷视为相同，从而产生大量误报，迫使开发人员和测试人员在每个发布周期中花费大量时间和精力手动审核标记的差异。行业工具将机器学习应用于 VRT，但缺乏公开评估。更重要的是，目前没有数据集或基准支持 UI 变化的自然语言描述，这一能力可以用文字告诉测试人员发生了什么变化，而不是让他们去解释二元标志或高亮区域。为了解决这一空白，我们提出了一项新任务，即 Web UI 图像变化描述（WUICC），该任务位于 VRT 和图像差异描述（IDC）之间，并发布了 WUICC-bench，这是该任务的第一个数据集和基准。我们评估了十一种代表性的 IDC 方法，以及两种零样本通用大型语言模型（LLM）。我们发现：（1）这些方法在 Web UI 领域往往表现不佳，原因在于其布局多样性、文本密集和细粒度变化；（2）然而，经过训练的方法已经比 VRT 所依赖的像素级比较更具选择性地抑制了无意义的视觉噪声，为未来的领域特定研究提供了坚实的基础。
cs.CV / 37 / 2607.01737
ReQuest: Rethinking-based Question-Aware Frame Selection for Long-Form Video QA
ReQuest：基于重新思考的问题感知帧选择用于长视频问答
Abstract
Recent multimodal large language models (MLLMs) have substantially advanced video understanding, yet long-form video QA remains challenging under fixed input token budgets, where uniform sampling can be inefficient for evidence localization. We propose ReQuest , an uncertainty-driven, question-adaptive keyframe selection pipeline that aligns question intent with relevant video content through selective computation. ReQuest integrates (i) a lightweight question-aware selector distilled from MLLM-generated supervision, (ii) Re-thinking Routing that triggers additional inference only when the model is uncertain with a length-adaptive criterion, and (iii) uncertainty-guided adaptive non-maximum suppression that selects temporally diverse frames while adjusting spacing based on question difficulty. As a plug-andplay method, ReQuest improves long-video QA without modifying or fine-tuning the underlying MLLM. Experiments on Video-MME, MLVU, and LongVideoBench demonstrate consistent accuracy gains with competitive computational cost, with particularly strong improvements in medium and long video regimes.
Chinese Translation
最近的多模态大语言模型（MLLMs）在视频理解方面取得了显著进展，但在固定输入令牌预算下，长视频问答仍然面临挑战，其中均匀采样在证据定位方面可能效率低下。我们提出了ReQuest，一种基于不确定性的、问题自适应的关键帧选择管道，通过选择性计算将问题意图与相关视频内容对齐。ReQuest集成了（i）从MLLM生成的监督中提炼出的轻量级问题感知选择器，（ii）重新思考路由（Re-thinking Routing），仅在模型不确定时触发额外推理，并采用长度自适应标准，以及（iii）不确定性引导的自适应非最大抑制，选择时间上多样的帧，同时根据问题难度调整间隔。作为一种即插即用的方法，ReQuest在不修改或微调基础MLLM的情况下改善了长视频问答。在Video-MME、MLVU和LongVideoBench上的实验表明，ReQuest在计算成本具有竞争力的情况下，始终实现了准确性的提升，尤其在中等和长视频场景中表现出显著的改善。
cs.CV / 38 / 2607.01743
InterCMDM: Block-Causal Diffusion for Autoregressive Human Interaction Generation
InterCMDM：用于自回归人类互动生成的块因果扩散
Abstract
Text-conditioned human interaction generation must capture both long-range temporal causality within each individual and tightly coupled coordination between partners. Existing interaction diffusion models typically denoise full sequences using bidirectional attention, which obscures causality and hinders streaming and long-horizon generation. Autoregressive alternatives enforce causality but often suffer from temporal drift, leading to coordination degradation and unstable interaction dynamics over time. We propose InterCMDM, a block-causal latent diffusion framework for autoregressive two-person interaction generation. InterCMDM introduces a Dual-Stream Causal Diffusion Transformer that maintains separate causal streams for each person while modeling inter-person dependencies via unified dual-stream attention with multi-task attention masks. These masks unify interaction modeling within a single attention mechanism and support diverse coordination behaviors, including simultaneous actions, reactive responses, leader-follower dynamics, and independent motion. By training a single model across these mask configurations as a form of data augmentation, InterCMDM enables controllable interaction generation by simply selecting the desired attention mask at inference time. Finally, a block-wise diffusion objective enables stable latent rollout over long sequences without repeated decode-encode cycles. InterCMDM achieves state-of-the-art performance on InterHuman and Inter-X, improving text-motion alignment, realism, and long-horizon continuity.
Chinese Translation
基于文本的人类互动生成必须捕捉到每个个体内部的长程时间因果关系以及合作伙伴之间的紧密协调。现有的互动扩散模型通常使用双向注意力对完整序列进行去噪，这模糊了因果关系并妨碍了流式和长时间生成。自回归替代方法强制执行因果关系，但往往会遭遇时间漂移，导致协调退化和随时间变化的不稳定互动动态。我们提出了InterCMDM，一种用于自回归双人互动生成的块因果潜在扩散框架。InterCMDM引入了一种双流因果扩散变换器，为每个人维护独立的因果流，同时通过统一的双流注意力和多任务注意力掩码建模个体间的依赖关系。这些掩码在单一注意力机制内统一了互动建模，并支持多样的协调行为，包括同时动作、反应性响应、领导-跟随动态和独立运动。通过在这些掩码配置上训练单一模型作为一种数据增强形式，InterCMDM使得通过在推理时简单选择所需的注意力掩码来实现可控的互动生成。最后，块级扩散目标使得在长序列上实现稳定的潜在展开，而无需重复的解码-编码循环。InterCMDM在InterHuman和Inter-X上实现了最先进的性能，提高了文本-动作对齐、真实感和长时间连续性。
cs.CV / 39 / 2607.01748
RTE-FM-Dehazer: Radiative Transfer Equation Inspired Flow Matching for Real-World Image Dehazing
RTE-FM-Dehazer：基于辐射传输方程的流匹配方法用于真实场景图像去雾
Abstract
Single-image dehazing aims to recover a clear scene from a hazy image and is generally formulated as an image-to-image translation task; however, it faces two limitations. Its performance depends heavily on the haze-formation priors embedded in the model. Prevailing methods adopt the Atmospheric Scattering Model (ASM), whose assumptions of single scattering and homogeneous media are often violated, leading to residual haze and color drift. Moreover, large-scale real hazy/clear pairs are impractical to collect, and existing synthesis approaches fail to reproduce the full complexity of natural haze. To address these issues, we present RTE-FM-Dehazer, a novel dehazing approach, together with a scalable data pipeline. Unlike the ASM, the Radiative Transfer Equation (RTE) jointly accounts for both scattering and absorption, naturally accommodating the non-homogeneous, multiple-scattering media that characterize real hazy scenes. Motivated by the structural similarity between the RTE diffusion-absorption term and the ODE in flow matching, we introduce a diffusion-absorption regularizer derived from a reduced RTE, to steer the flow matching trajectory at each step. Next, leveraging modern vision-language models, we build an automated pipeline and release P-HAZE, a dataset of 50000 realistic hazy/clear pairs. Extensive evaluations demonstrate that RTE-FM-Dehazer, trained solely on P-HAZE, effectively eliminates artifacts like residual haze and color drift, exhibits strong cross-domain generalization, and achieves leading results on five real-world dehazing benchmarks.
Chinese Translation
单幅图像去雾旨在从模糊图像中恢复清晰场景，通常被表述为图像到图像的转换任务；然而，它面临两个局限性。其性能在很大程度上依赖于模型中嵌入的雾霾形成先验。现有方法采用大气散射模型（Atmospheric Scattering Model, ASM），但其单次散射和均匀介质的假设常常被违反，导致残余雾霾和颜色漂移。此外，大规模的真实模糊/清晰图像对的收集是不切实际的，现有的合成方法无法重现自然雾霾的复杂性。为了解决这些问题，我们提出了RTE-FM-Dehazer，这是一种新颖的去雾方法，并配备可扩展的数据管道。与ASM不同，辐射传输方程（Radiative Transfer Equation, RTE）同时考虑了散射和吸收，自然适应了真实模糊场景中非均匀和多次散射介质的特征。受到RTE扩散-吸收项与流匹配中的常微分方程（ODE）之间结构相似性的启发，我们引入了一种基于简化RTE的扩散-吸收正则化器，以引导每一步的流匹配轨迹。接下来，利用现代视觉-语言模型，我们构建了一个自动化管道，并发布了P-HAZE，一个包含50000对真实模糊/清晰图像的数据集。大量评估表明，仅在P-HAZE上训练的RTE-FM-Dehazer有效消除了残余雾霾和颜色漂移等伪影，表现出强大的跨域泛化能力，并在五个真实世界去雾基准测试中取得了领先的结果。
cs.CV / 40 / 2607.01751
MedStreamBench: A Time-Aware Benchmark for Streaming and Proactive Medical Video Understanding
MedStreamBench：一个时间感知的流媒体和主动医学视频理解基准
Abstract
Existing medical video benchmarks primarily evaluate whether a model produces the correct answer, but rarely assess whether it answers at the right time. In real clinical settings, AI systems must decide not only what to predict, but also when to answer, defer judgment, or proactively raise alerts. This creates a critical gap between benchmark evaluation and deployment requirements. We present MedStreamBench, a benchmark for time-aware medical video understanding. MedStreamBench integrates 22 medical datasets and 5,419 QA instances across four temporal settings: retrospective, present, future, and proactive. Unlike conventional benchmarks that assume full-video access, MedStreamBench restricts models to temporally bounded evidence windows and supports both single-turn and streaming evaluation. We further introduce a proactive monitoring setting that requires models to determine whether and when clinically relevant alerts should be triggered. Beyond answer correctness, MedStreamBench evaluates temporal behavior through responsiveness and post-evidence stability. Experiments on leading general-purpose and medical vision-language models reveal a substantial gap between offline recognition and temporally grounded decision-making, with performance dropping markedly in streaming and proactive settings. Our benchmark is available at https://huggingface.co/datasets/Venn2024/MedStreamBench.
Chinese Translation
现有的医学视频基准主要评估模型是否产生正确的答案，但很少评估模型是否在正确的时间作出回答。在真实的临床环境中，人工智能系统不仅必须决定预测什么，还必须决定何时回答、何时推迟判断或主动发出警报。这在基准评估与部署要求之间造成了一个关键的差距。我们提出了MedStreamBench，一个用于时间感知医学视频理解的基准。MedStreamBench整合了22个医学数据集和5419个问答实例，涵盖四种时间设置：回顾性、当前、未来和主动性。与假设完全视频访问的传统基准不同，MedStreamBench限制模型在时间上有界的证据窗口内进行评估，并支持单轮和流媒体评估。我们进一步引入了一个主动监测设置，要求模型确定何时以及是否应触发临床相关的警报。除了答案的正确性，MedStreamBench还通过响应性和后证据稳定性评估时间行为。在领先的通用和医学视觉语言模型上的实验显示，离线识别与时间基础决策之间存在显著差距，流媒体和主动设置下的性能显著下降。我们的基准可在 https://huggingface.co/datasets/Venn2024/MedStreamBench 获取。
cs.CV / 41 / 2607.01753
The Turning Point of 3D Plant Phenotyping: 3D Foundation Models Enable Minute-to-Second Cross-Crop Reconstruction and Beyond
3D植物表型分析的转折点：3D基础模型实现跨作物的分钟至秒级重建及其应用
Abstract
3D plant phenotyping is notoriously known to be procedure-complicated and of low throughput due to the extensive multi-view imaging, the fragile 3D reconstruction pipeline, and the additional cost from reconstructed geometry to phenotypic extraction. These limitations are further amplified in low-cost data acquisition, where smartphone videos or sparsely sampled multi-view images provide limited view overlap and self-occlusion. In this work, we show that the conventional 3D plant phenotyping pipeline could be streamlined and significantly accelerated with 3D Foundation Models (3DFMs), and particularly, present one of the first cross-crop 3D phenotyping frameworks powered by 3DFMs. The framework replaces COLMAP-style sparse initialization with 3DFM-based feed-forward geometric recovery, combines geometry-constrained 3D Gaussian Splatting for dense reconstruction, enables few-view reconstruction through iterative view synthesis and refinement, and converts reconstructed geometry into measurable organs through 2D-to-3D semantic transfer, metric scale recovery, and organ instance separation. We further construct a cross-crop dataset with smartphone-based image acquisition, diverse plant morphologies, and manual annotations for segmentation and phenotypic evaluation. Experiments across 26 plant sequences show that 3D Foundation Models reduce the average reconstruction time from 6.52 minutes to 1.58 seconds while maintaining high reconstruction quality and phenotyping accuracy. These results suggest a fresh technical route for high-throughput 3D plant phenotyping, from low-cost image acquisition to fast reconstruction, perception, scale recovery, and phenotypic measurement.
Chinese Translation
3D植物表型分析因其复杂的程序和低通量而广为人知，这主要归因于广泛的多视角成像、脆弱的3D重建流程以及从重建几何到表型提取的额外成本。这些限制在低成本数据采集的情况下进一步加剧，智能手机视频或稀疏采样的多视角图像提供了有限的视角重叠和自遮挡。在本研究中，我们展示了传统的3D植物表型分析流程可以通过3D基础模型（3DFMs）进行简化和显著加速，特别是提出了一个由3DFMs驱动的首个跨作物3D表型分析框架。该框架用基于3DFM的前馈几何恢复替代了COLMAP风格的稀疏初始化，结合几何约束的3D高斯点云进行密集重建，通过迭代视角合成和细化实现少视角重建，并通过2D到3D的语义转移、度量尺度恢复和器官实例分离将重建几何转换为可测量的器官。我们进一步构建了一个跨作物数据集，采用基于智能手机的图像采集，涵盖多样的植物形态，并进行手动标注以便于分割和表型评估。在26个植物序列的实验中，3D基础模型将平均重建时间从6.52分钟减少到1.58秒，同时保持高重建质量和表型准确性。这些结果为高通量3D植物表型分析提供了一条新的技术路径，从低成本图像采集到快速重建、感知、尺度恢复和表型测量。
cs.CV / 42 / 2607.01756
ProSAC-CT: Progressive Spectral-Anatomical Co-Guided Multi-Stage Diffusion Model for Low-Dose CT Denoising
ProSAC-CT：渐进式光谱-解剖共指导多阶段扩散模型用于低剂量CT去噪
Abstract
Low-dose computed tomography (LDCT) reduces radiation exposure but introduces stronger quantum noise, streak artifacts, and local texture degradation, which can obscure anatomical boundaries and weaken low-contrast structures. Diffusion models are promising for LDCT denoising by progressively recovering normal-dose CT (NDCT) images from degraded LDCT inputs, but existing methods often suffer from insufficient anatomical guidance, uncertain frequency-dependent recovery, and uniform reverse-process modeling. We propose ProSAC-CT, a progressive spectral-anatomical co-guided multi-stage diffusion model for image-domain LDCT denoising. ProSAC-CT integrates an anatomical-prior-guided conditioning (APGC) module, a residual frequency-domain decoupling stage (RFDDS), and a time-step-decoupling denoising decoder (TD3). APGC extracts LDCT-derived structural guidance, RFDDS enhances frequency-aware representations, and TD3 assigns them to different reverse-diffusion stages for anatomical stabilization, boundary refinement, and fine-detail recovery. Experiments on four LDCT degradation benchmarks show that ProSAC-CT improves image fidelity, structural similarity, perceptual quality, and information preservation over representative methods while better preserving boundary-sensitive anatomical details. Downstream anatomical-region classification on Mayo-2020 further indicates that ProSAC-CT retains task-relevant anatomical information, supporting its practical use for low-dose CT denoising.
Chinese Translation
低剂量计算机断层扫描（LDCT）减少了辐射暴露，但引入了更强的量子噪声、条纹伪影和局部纹理退化，这可能会模糊解剖边界并削弱低对比度结构。扩散模型在LDCT去噪方面表现出良好的前景，通过逐步从退化的LDCT输入中恢复正常剂量CT（NDCT）图像，但现有方法往往缺乏足够的解剖指导、不确定的频率依赖恢复以及均匀的反向过程建模。我们提出了ProSAC-CT，一种用于图像域LDCT去噪的渐进式光谱-解剖共指导多阶段扩散模型。ProSAC-CT集成了一个解剖先验指导条件（APGC）模块、一个残差频域解耦阶段（RFDDS）和一个时间步解耦去噪解码器（TD3）。APGC提取LDCT衍生的结构指导，RFDDS增强频率感知表示，TD3将其分配到不同的反向扩散阶段，以实现解剖稳定、边界细化和细节恢复。在四个LDCT退化基准上的实验表明，ProSAC-CT在图像保真度、结构相似性、感知质量和信息保留方面优于代表性方法，同时更好地保留了边界敏感的解剖细节。在Mayo-2020上的下游解剖区域分类进一步表明，ProSAC-CT保留了与任务相关的解剖信息，支持其在低剂量CT去噪中的实际应用。
cs.CV / 43 / 2607.01757
DL-VINS-Factory: A Modular Framework for Learned Visual Front-Ends in Visual-Inertial SLAM
DL-VINS-Factory：视觉惯性SLAM中学习视觉前端的模块化框架
Abstract
Deep-learning features excel in visual matching, yet their practical value in tightly coupled visual-inertial SLAM (VI-SLAM) remains insufficiently characterized. We present DL-VINS-Factory, a unified framework that integrates learned feature extractors (ALIKED, RaCo, SuperPoint, XFeat) with either Lucas--Kanade (LK) optical-flow tracking or LightGlue (LG) descriptor matching. All front-ends share a sliding-window Ceres back-end, with optional AnyLoc DINOv2-VLAD loop closure, and 4-DoF pose-graph optimization. We benchmark the system across the four datasets covering indoor, unstructured outdoor, aggressive-motion, and visually degraded conditions. Results show that learned front-ends are viable for real-time embedded VI-SLAM, but are not universally superior to classical tracking. Relative to the corresponding GFTT+LK baseline, ALIKED+LG reduces EuRoC ATE by $5\%$ in monocular odometry and by $7\%$ in stereo with loop-closure. On NTU-VIRAL, where aggressive aerial motion increases inter-frame viewpoint change, ALIKED+LG stereo reduces loop-closed ATE by $12\%$. In Botanic Garden dataset, optical-flow tracking remains preferable, but learned keypoints still improve over the baseline GFTT, in which SuperPoint+LK reduces grayscale camera ATE by $29\%$, while RaCo+LK reduces RGB camera ATE by $38\%$. On SubT-MRS, learned front-ends display varying degree of improvement based on individual cases. With TensorRT acceleration on a Jetson AGX Orin, all valid configurations run in real time between $29$--$47$ FPS in monocular mode and $18$--$33$ FPS in stereo mode for the EuRoC and NTU-VIRAL datasets. AnyLoc further confirms roughly $2$--$7\times$ more valid loops than BRIEF+DBoW2. The implementation is open-sourced at https://github.com/limshoonkit/DL-VINS-Factory-ROS2/.
Chinese Translation
深度学习特征在视觉匹配中表现出色，但其在紧耦合视觉惯性SLAM（VI-SLAM）中的实际价值仍然不足以充分表征。我们提出了DL-VINS-Factory，这是一个统一框架，将学习特征提取器（ALIKED、RaCo、SuperPoint、XFeat）与Lucas-Kanade（LK）光流跟踪或LightGlue（LG）描述子匹配相结合。所有前端共享一个滑动窗口Ceres后端，并可选配AnyLoc DINOv2-VLAD回环闭合和4自由度姿态图优化。我们在四个数据集上对系统进行了基准测试，这些数据集涵盖了室内、非结构化户外、激烈运动和视觉退化条件。结果表明，学习前端在实时嵌入式VI-SLAM中是可行的，但并不普遍优于经典跟踪。与相应的GFTT+LK基线相比，ALIKED+LG在单目测距中将EuRoC ATE降低了5%，在立体测距中将其降低了7%（带回环闭合）。在NTU-VIRAL数据集中，激烈的空中运动增加了帧间视点变化，ALIKED+LG立体测距将回环闭合的ATE降低了12%。在植物园数据集中，光流跟踪仍然更可取，但学习关键点仍然优于基线GFTT，其中SuperPoint+LK将灰度相机的ATE降低了29%，而RaCo+LK将RGB相机的ATE降低了38%。在SubT-MRS上，学习前端的改进程度因个案而异。在Jetson AGX Orin上使用TensorRT加速，所有有效配置在单目模式下以29-47 FPS和在立体模式下以18-33 FPS的速度实时运行于EuRoC和NTU-VIRAL数据集上。AnyLoc进一步确认其有效回环数量比BRIEF+DBoW2多出约2-7倍。该实现已开源，地址为https://github.com/limshoonkit/DL-VINS-Factory-ROS2/。
cs.CV / 44 / 2607.01759
ProCal: Inference-Time Proposal Calibration for Open-Vocabulary Object Detection
ProCal：开放词汇目标检测的推理时间提议校准
Abstract
Open-vocabulary object detection aims to localize and classify objects beyond the fixed set of categories seen dur ing training. Recent open-vocabulary object detection methods improve localization and classification for unseen categories by leveraging a frozen VLM as a detector backbone. However, VLM classification score lacks recognizing position and scale of the object in an image. We observe that pretrained VLMs en able to classify foreground and background regions. According to this observation, we propose a simple inference-time Pro posal Calibration (ProCal) that improves localization quality of the classification score. ProCal computes a proposal prior by combining two scores: localization-aware foreground score and background-aware suppression score. Localization-aware foreground score captures whether a proposal contains an object area. Background-aware suppression score measures the extent to which the proposal resembles background. We analyze that ProCal suppresses false novel activation on background proposals and consistently ranks true novel proposals above background and partial novel proposals. Applied to CLIPSelf ViT-L/14, ProCal improves APr +2.5 on OV-LVIS. The analyses show that proposal-level localization-aware reranking effects to mitigate ranking miscalibration for novel objects.
Chinese Translation
开放词汇目标检测旨在定位和分类训练过程中未见的超出固定类别集合的对象。最近的开放词汇目标检测方法通过利用冻结的视觉语言模型（VLM）作为检测器主干，改善了对未见类别的定位和分类。然而，VLM的分类得分缺乏对图像中对象位置和尺度的识别。我们观察到，预训练的VLM能够对前景和背景区域进行分类。基于这一观察，我们提出了一种简单的推理时间提议校准（Proposal Calibration，ProCal），以提高分类得分的定位质量。ProCal通过结合两个得分来计算提议先验：定位感知前景得分和背景感知抑制得分。定位感知前景得分捕捉提议是否包含对象区域，而背景感知抑制得分则衡量提议与背景的相似程度。我们分析表明，ProCal抑制了背景提议上的虚假新激活，并始终将真实的新提议排名高于背景和部分新提议。应用于CLIPSelf ViT-L/14，ProCal在OV-LVIS上提高了APr +2.5。分析表明，提议级定位感知重新排序的效果可以减轻新对象的排名误校准。
cs.CV / 45 / 2607.01768
JointHOI: Jointly Generating Contact Maps Enhances Hand Object Interaction Generation
JointHOI：联合生成接触图增强手物体交互生成
Abstract
Text driven hand object interaction (HOI) generation is gaining attention for immersive applications and robotics, yet producing physically plausible interactions remains challenging. Even when individual motions appear natural, small contact errors can cause conspicuous artifacts such as floating and interpenetration. Prior methods mitigate these issues using explicit contact cues or implicit grasp priors, but typically rely on multi stage pipelines and fail to model temporally evolving contact. We present JointHOI, a single stage diffusion framework that jointly generates 3D hand object motion and dynamic, distance based contact maps from text. By treating contact as an auxiliary inner modality, joint generation enables the model to learn contact motion coupling during training. At inference, contact guided sampling enforces consistency between generated contact maps and motion implied geometry, improving temporal stability and reducing penetration and floating. Experiments on GRAB and ARCTIC demonstrate consistent improvements in text adherence and physical plausibility over prior methods.
Chinese Translation
基于文本驱动的手物体交互（HOI）生成在沉浸式应用和机器人领域日益受到关注，但生成物理上合理的交互仍然具有挑战性。即使单独的动作看起来自然，微小的接触错误也可能导致明显的伪影，如漂浮和相互穿透。之前的方法通过显式接触线索或隐式抓取先验来缓解这些问题，但通常依赖于多阶段管道，并未能有效建模时间演变的接触。我们提出了JointHOI，一个单阶段扩散框架，能够从文本中联合生成3D手物体运动和动态的基于距离的接触图。通过将接触视为辅助内在模态，联合生成使模型在训练过程中能够学习接触运动的耦合。在推理阶段，接触引导的采样强制生成的接触图与运动隐含几何之间保持一致，从而提高时间稳定性，减少穿透和漂浮现象。在GRAB和ARCTIC上的实验表明，与之前的方法相比，在文本遵循性和物理合理性方面均有一致的改善。
cs.CV / 46 / 2607.01772
LLM-Empowered Multimodal Fusion Framework for Autonomous Driving: Semantic Enhancement and Channel-Adaptive Design
基于大型语言模型的多模态融合框架用于自动驾驶：语义增强与通道自适应设计
Abstract
Vision-radar fusion is central to robust autonomous driving, combining dense visual semantics with precise range and velocity measurements from radar. However, real-world fusion quality is fundamentally challenged by dynamically varying input quality, stemming from occlusion, adverse weather, and channel noise. To address this, we re-frame the problem from static data fusion to channel-aware semantic reasoning and propose a Large Language Model-centric Semantic-layer Channel-aware Integrated Perception (LM-SCIP) framework. It places a Large Language Model (LLM) as a central reasoning core to fuse a local visual stream with a quality-varying external radar stream used to cover perception-blind spots. Concretely, LM-SCIP couples a hierarchical radar-vision encoder with a Channel-Adaptive Semantic Module (CASM) that maps link indicators into a "Channel Prompt" to dynamically gate external radar features. A parameter-efficient, LoRA-tuned LLM, in conjunction with a heterogeneous Mixture-of-Experts (H-MoE), then arbitrates between local visual cues and the channel-conditioned radar context. Finally, a decoupled multi-task decoder outputs localization, trajectory forecasting, and image reconstruction. Experiments on nuScenes and VIRAT validate our approach. On nuScenes, under a controlled toggle of radar input, LM-SCIP reduces localization RMSE by 40.0% versus a vision-only baseline. On VIRAT, the model attains a 0.214m localization RMSE and 0.179m minFDE (k=1). These results reveal that the proposed LM-SCIP enables a robust vision-dominant fallback at low SNR and synergistic fusion at high SNR.
Chinese Translation
视觉-雷达融合是实现稳健自动驾驶的核心，将密集的视觉语义与雷达提供的精确距离和速度测量相结合。然而，现实世界中的融合质量受到动态变化的输入质量的根本挑战，这种变化源于遮挡、不良天气和通道噪声。为了解决这一问题，我们将问题重新框定为通道感知的语义推理，并提出了一种以大型语言模型为中心的语义层通道感知集成感知框架（LM-SCIP）。该框架将大型语言模型（LLM）作为中心推理核心，将局部视觉流与用于覆盖感知盲区的质量变化的外部雷达流进行融合。具体而言，LM-SCIP将分层雷达-视觉编码器与通道自适应语义模块（CASM）结合，后者将链接指示符映射到“通道提示”，以动态控制外部雷达特征。然后，经过参数高效的LoRA调优的LLM与异构专家混合模型（H-MoE）共同裁决局部视觉线索和通道条件下的雷达上下文。最后，解耦的多任务解码器输出定位、轨迹预测和图像重建。在nuScenes和VIRAT上的实验验证了我们的方法。在nuScenes上，在控制雷达输入的切换下，LM-SCIP将定位均方根误差（RMSE）降低了40.0%，相较于仅使用视觉的基线。在VIRAT上，该模型达到了0.214米的定位RMSE和0.179米的最小最终位置误差（minFDE，k=1）。这些结果表明，所提出的LM-SCIP在低信噪比（SNR）下实现了稳健的视觉主导回退，而在高信噪比下实现了协同融合。
cs.CV / 47 / 2607.01784
SpaceEra++: A Unified Framework Towards 3D Spatial Reasoning in Video
SpaceEra++：面向视频中的三维空间推理的统一框架
Abstract
Visual-spatial understanding, defined as the ability to infer object relationships and scene layouts from visual inputs, is fundamental to downstream tasks such as robotic navigation and embodied interaction. However, pre-trained vision-language models (VLMs) remain constrained by spatial uncertainty stemming from inherently 2D observations and by the scarcity of data for 3D spatial understanding. To address these limitations, we proposed a novel framework, SpaceEra, in the NeurIPS 2025 Spotlight paper. Although it achieved significant performance gains, we further observed that its effectiveness is hindered by insufficient input from scanning videos and weak reasoning constraints. To tackle these newly emerged challenges, we extend the original framework into a comprehensive system, termed SpaceEra++, which spans data construction, model design, training optimization, and prompting inference. Specifically, to alleviate input insufficiency, we introduce ScenePick, a frame sampling strategy that balances spatial coverage with object semantics to produce compact yet comprehensive scene representations. In addition, to enhance spatial reasoning, we develop SpaceAlign, which enforces pairwise object constraints by jointly exploiting absolute coordinates and relative spatial relations, thereby aligning optimization with spatial accuracy. Extensive experiments across multiple benchmarks demonstrate consistent improvements over strong baselines, while ablation studies validate both the individual and joint contributions of each component, and further analyses provide guidance for future research.
Chinese Translation
视觉空间理解被定义为从视觉输入中推断物体关系和场景布局的能力，这对于机器人导航和具身交互等下游任务至关重要。然而，预训练的视觉-语言模型（VLMs）仍然受到源于固有二维观察的空间不确定性和三维空间理解数据稀缺的限制。为了解决这些局限性，我们在NeurIPS 2025的聚光灯论文中提出了一个新颖的框架SpaceEra。尽管它取得了显著的性能提升，但我们进一步观察到，其有效性受到扫描视频输入不足和推理约束弱的影响。为了解决这些新出现的挑战，我们将原始框架扩展为一个综合系统，称为SpaceEra++，涵盖数据构建、模型设计、训练优化和提示推理。具体而言，为了缓解输入不足的问题，我们引入了ScenePick，一种帧采样策略，平衡空间覆盖与物体语义，以生成紧凑而全面的场景表示。此外，为了增强空间推理，我们开发了SpaceAlign，通过共同利用绝对坐标和相对空间关系来强制成对物体约束，从而使优化与空间准确性对齐。在多个基准上的广泛实验表明，相较于强基线，性能持续改善，而消融研究验证了每个组件的单独和联合贡献，进一步分析为未来研究提供了指导。
cs.CV / 48 / 2607.01803
PixGS: Pixel-Space Diffusion for Direct 3D Gaussian Splat Generation
PixGS：用于直接生成3D高斯斑点的像素空间扩散
Abstract
Recent advances in 3D content generation from text or images have achieved impressive results, yet view inconsistency from 2D generators and the scarcity of high-quality 3D data remain significant bottlenecks. Existing solutions typically adapt large-scale pre-trained text-to-image latent diffusion models to generate 3D Gaussian Splats (3DGS). However, these approaches often rely on training complex cascade pipelines that are computationally expensive and scalability-limited. Most critically, the quality of generated 3D assets is inherently constrained by each component capacity and compressed latent space, leading to decoding artifacts and accumulated errors. To address these limitations, we propose PixGS, a single-stage pipeline for direct high-quality 3DGS generation, which leverages recent advances in pixel-space diffusion to bypass lossy latent compression while still benefiting from the vast 2D generative priors. By directly denoising 3D Gaussian attributes at each timestep, our method enables precise, splat-level regularization of both appearance and geometry. Furthermore, we introduce a comprehensive supervision strategy that incorporates surface normals, depth, and high-frequency structural information, which is often overlooked in prior works. Experiments demonstrate that PixGS outperforms current state-of-the-art methods while maintaining a fast inference speed (1s on a single A100 GPU), offering a robust and efficient alternative to multi-stage generation pipelines.
Chinese Translation
近年来，从文本或图像生成3D内容的技术取得了显著进展，但2D生成器的视图不一致性和高质量3D数据的稀缺仍然是重要的瓶颈。现有的解决方案通常将大规模预训练的文本到图像潜在扩散模型适配用于生成3D高斯斑点（3DGS）。然而，这些方法通常依赖于训练复杂的级联管道，计算成本高且可扩展性有限。最关键的是，生成的3D资产的质量本质上受到每个组件能力和压缩潜在空间的限制，导致解码伪影和累积误差。为了解决这些限制，我们提出了PixGS，这是一种用于直接生成高质量3DGS的单阶段管道，利用了像素空间扩散的最新进展，以绕过有损潜在压缩，同时仍然受益于广泛的2D生成先验。通过在每个时间步直接去噪3D高斯属性，我们的方法实现了外观和几何形状的精确斑点级正则化。此外，我们引入了一种综合监督策略，结合了表面法线、深度和高频结构信息，这些在之前的工作中常常被忽视。实验表明，PixGS在保持快速推理速度（单个A100 GPU上为1秒）的同时，超越了当前的最先进方法，提供了一种稳健且高效的替代多阶段生成管道的方案。
cs.CV / 49 / 2607.01813
MMBench-Live: A Continuously Evolving Benchmark for Multimodal Models
MMBench-Live：一个持续演变的多模态模型基准测试
Abstract
Evaluation benchmarks are essential for assessing vision-language models (VLMs), but most multimodal benchmarks are static, making them vulnerable to temporal staleness, data contamination, and costly maintenance. We present MMBench-Live, a continuously evolving multimodal benchmark built by a multi-agent-driven automated pipeline. Our framework treats benchmark evolution as task-guided dataset construction, integrating structured benchmark specification, feedback-controlled real-time data acquisition, and verifiable QA generation with executable reasoning. To maintain cross-version comparability, we introduce a distribution-consistent update strategy that extracts task-related visual patterns from the original benchmark to guide data collection and filtering. Instantiated from MMBench, MMBench-Live contains 5.9K newly generated evaluation instances with a high answer correctness rate, while each update costs about USD 30 and takes 1-2 hours. Extensive evaluations show that MMBench-Live preserves stable model rankings, maintains semantic alignment with the original benchmark, and exhibits weaker contamination-related memorization signals, suggesting a practical and scalable paradigm for sustainable multimodal benchmark evolution. The project is available at https://github.com/PRIS-CV/MMBench-Live.
Chinese Translation
评估基准对于评估视觉语言模型（VLMs）至关重要，但大多数多模态基准是静态的，这使得它们容易受到时间滞后、数据污染和高昂维护成本的影响。我们提出了MMBench-Live，一个由多代理驱动的自动化管道构建的持续演变的多模态基准。我们的框架将基准演变视为任务引导的数据集构建，整合了结构化的基准规范、反馈控制的实时数据获取和可验证的质量保证生成与可执行推理。为了保持跨版本的可比性，我们引入了一种分布一致的更新策略，从原始基准中提取与任务相关的视觉模式，以指导数据收集和过滤。从MMBench实例化的MMBench-Live包含5900个新生成的评估实例，具有较高的答案正确率，而每次更新的成本约为30美元，耗时1-2小时。广泛的评估表明，MMBench-Live保持了稳定的模型排名，与原始基准保持了语义一致性，并表现出较弱的污染相关记忆信号，表明这是一个可行且可扩展的可持续多模态基准演变范式。该项目可在https://github.com/PRIS-CV/MMBench-Live获取。
cs.CV / 50 / 2607.01825
Rethinking Conditional Generation for Underwater Salient Object Detection
重新思考水下显著性物体检测的条件生成
Abstract
Salient Object Detection in underwater images remains challenging due to low contrast, uneven illumination, and color distortion caused by scattering and absorption effects, which limit the effectiveness of conventional SOD methods in underwater environments. To address these challenges, we propose a Degradation-aware Conditional Generation Network (DCGNet), specifically designed to construct reliable conditional features for underwater saliency generation. First, we design a Dynamic Multi-Granularity module (DMG) grounded in the human visual system to robustly detect salient objects of varying scales with blurred boundaries. Then, we develop an Underwater Physics-Prior module (UPP), which utilizes pseudo-depth guidance to estimate underwater light attenuation and backscatter, thereby restoring degradation-aware RGB features and mitigating color distortion and boundary ambiguity. Based on the physics-guided representation, we introduce an Underwater Spatial Gaussian module (USG), which constructs a spatial Gaussian saliency prior from the strongest guided response to enhance object-centered salient regions and suppress cluttered underwater backgrounds. In addition, a lightweight timestep-adaptive Diffusion Transformer (DiT) bottleneck is inserted into the denoising decoder to refine fused features at different diffusion timesteps. Comprehensive experiments on USOD10K, USOD, CSOD10K, MAS3K, and RMAS demonstrate that DCGNet significantly outperforms existing state-of-the-art methods, verifying its potential for complex underwater visual applications.
Chinese Translation
水下图像中的显著性物体检测仍然面临挑战，主要由于低对比度、不均匀的照明以及由散射和吸收效应引起的颜色失真，这限制了传统显著性物体检测（SOD）方法在水下环境中的有效性。为了解决这些挑战，我们提出了一种降解感知条件生成网络（Degradation-aware Conditional Generation Network，DCGNet），专门设计用于构建可靠的水下显著性生成条件特征。首先，我们设计了一个基于人类视觉系统的动态多粒度模块（Dynamic Multi-Granularity module，DMG），以稳健地检测具有模糊边界的不同尺度的显著物体。然后，我们开发了一个水下物理先验模块（Underwater Physics-Prior module，UPP），利用伪深度引导来估计水下光衰减和反向散射，从而恢复降解感知的RGB特征，减轻颜色失真和边界模糊。基于物理引导的表示，我们引入了一个水下空间高斯模块（Underwater Spatial Gaussian module，USG），该模块从最强的引导响应构建空间高斯显著性先验，以增强以物体为中心的显著区域并抑制杂乱的水下背景。此外，一个轻量级的时间步自适应扩散变换器（Diffusion Transformer，DiT）瓶颈被插入到去噪解码器中，以在不同扩散时间步上精炼融合特征。在USOD10K、USOD、CSOD10K、MAS3K和RMAS上的全面实验表明，DCGNet显著优于现有的最先进方法，验证了其在复杂水下视觉应用中的潜力。
cs.CV / 51 / 2607.01827
C2E: Boosting Ego-Only 3D Object Detection via Multi-Teacher Contrastive Knowledge Distillation
C2E：通过多教师对比知识蒸馏提升自我感知的3D目标检测
Abstract
LiDAR-based 3D object detection is essential for autonomous driving systems. However, traditional Ego-only Perception (Eo-Perception) suffers from limited perspective and occlusions in a complex outdoor environment, leading to performance bottlenecks. Recently, research on multi-agent Collaborative Perception (Co-Perception) has demonstrated excellent performance, but high communication costs and accumulated pose error hinder its application. To address this, we explore a novel C2E (Co-Perception to Eo-Perception) paradigm through the Multi-to-Single (M2S) agent contrastive knowledge distillation framework. Our M2S framework first designs Multi-Level Feature Enhancement module to provide more stable features, and introduces Auxiliary Point Cloud Reconstruction and Multi-Teacher Contrastive Distillation mechanisms to mitigate domain gaps in point cloud and feature distributions within the C2E paradigm. Benefiting from this, our M2S can retain the excellent performance of collaborative perception while effectively avoiding the drawbacks, such as communication delays and positioning errors. Extensive experiments on the V2XSet, V2V4Real and DAIR-V2X datasets show the effectiveness and generalizability of our M2S framework when combined with the state-of-the-art CoSDH model and other excellent 3D detectors. Our M2S framework can deliver up to a 8.64% improvement in 3D mAP performance without introducing any communication costs.
Chinese Translation
基于LiDAR的3D目标检测对于自动驾驶系统至关重要。然而，传统的自我感知（Ego-only Perception, Eo-Perception）在复杂的户外环境中受到有限视角和遮挡的影响，导致性能瓶颈。最近，关于多智能体协作感知（Collaborative Perception, Co-Perception）的研究显示出优异的性能，但高通信成本和累积的姿态误差阻碍了其应用。为了解决这个问题，我们通过多对单（Multi-to-Single, M2S）智能体对比知识蒸馏框架探索了一种新颖的C2E（Co-Perception to Eo-Perception）范式。我们的M2S框架首先设计了多级特征增强模块，以提供更稳定的特征，并引入辅助点云重建和多教师对比蒸馏机制，以减轻C2E范式中点云和特征分布的领域差距。得益于此，我们的M2S能够保留协作感知的优异性能，同时有效避免通信延迟和定位误差等缺点。在V2XSet、V2V4Real和DAIR-V2X数据集上进行的广泛实验表明，当与最先进的CoSDH模型及其他优秀的3D检测器结合时，我们的M2S框架展现了有效性和可推广性。我们的M2S框架在不引入任何通信成本的情况下，能够将3D mAP性能提升高达8.64%。
cs.CV / 52 / 2607.01851
Geometric Foundation Model Distillation for Efficient Lunar 3D Reconstruction
高效月球三维重建的几何基础模型蒸馏
Abstract
Large 3D foundation models such as MASt3R achieve state-of-the-art stereo reconstruction but are computationally demanding for deployment under strict hardware constraints -- a critical limitation in domains such as planetary exploration, where onboard computing is severely restricted. We study how far such models can be compressed through knowledge distillation, using lunar stereo reconstruction as a challenging and practically relevant case study. Starting from a 688M-parameter MASt3R teacher fine-tuned on lunar imagery, we distill its dense geometric predictions into a family of lightweight students spanning different encoder types (CNN vs ViT), decoder widths and depths, and training strategies. To bridge the dimensional mismatch between teacher and student, we propose a structured SVD-based initialization that projects the teacher's decoder weights into the student's smaller latent space, yielding a warm start that significantly improves convergence and final performance. Based on our results on lunar data, we can obtain a distilled student that retains most of teacher's reconstruction accuracy while reducing the model size up to 7 times, and even outperforms a baseline trained directly with sparse ground-truth annotations. Beyond compression, our study highlights both principles and practical insights for distilling geometric foundation models: a convolutional encoder underperforms transformer-based alternatives (though pretraining availability remains a confounding factor), preserving encoder capacity is more critical than maintaining a large decoder, feature-level distillation consistently outperforms output-only supervision, and SVD-based initialization improves optimisation stability. These findings provide practical guidelines for deploying 3D reconstruction models in resource-constrained environments.
Chinese Translation
大型三维基础模型如 MASt3R 在立体重建方面达到了最先进的水平，但在严格的硬件限制下部署时计算需求高，这在行星探索等领域是一个关键的限制，因为机载计算能力受到严重限制。我们研究了通过知识蒸馏可以将这些模型压缩到何种程度，以月球立体重建作为一个具有挑战性且实际相关的案例研究。从一个在月球图像上微调的688M参数的 MASt3R 教师模型开始，我们将其密集的几何预测蒸馏成一系列轻量级学生模型，涵盖不同的编码器类型（CNN 与 ViT）、解码器的宽度和深度以及训练策略。为了弥合教师和学生之间的维度不匹配，我们提出了一种基于结构化奇异值分解（SVD）的初始化方法，将教师的解码器权重投影到学生的较小潜在空间中，从而实现了一个温启动，显著提高了收敛性和最终性能。根据我们在月球数据上的结果，我们可以获得一个蒸馏学生模型，该模型在保留大部分教师重建精度的同时，将模型大小减少到最多7倍，甚至超越了直接使用稀疏真实标注训练的基线模型。除了压缩之外，我们的研究还强调了蒸馏几何基础模型的原则和实际见解：卷积编码器的表现不如基于变换器的替代方案（尽管预训练的可用性仍然是一个混淆因素），保持编码器的能力比维持大型解码器更为关键，特征级蒸馏始终优于仅输出监督，而基于 SVD 的初始化提高了优化的稳定性。这些发现为在资源受限环境中部署三维重建模型提供了实用指南。
cs.CV / 53 / 2607.01869
QWERTY: Training-Free Motion Control via Query-Warped Video Diffusion Transformers
QWERTY：通过查询扭曲视频扩散变换器实现无训练运动控制
Abstract
Video diffusion transformers (DiTs) generate high-fidelity and temporally coherent videos, yet motion control remains implicit, primarily relying on text prompts. As a result, achieving desired motion often requires extensive prompt engineering and repeated resampling. While fine-tuning models with additional spatial prompts (e.g., bounding boxes or point trajectories) enables explicit control, it demands substantial data curation and computation, and may compromise the generative capabilities of pretrained models. Consequently, training-free motion control using such spatial prompts has been explored in U-Net-based video diffusion models, but remains largely unexplored for DiTs. We introduce QWERTY, a training-free framework that enables flexible motion control in pretrained image-to-video DiTs via user-defined object warping and optical flow. We carefully manipulate the 3D full attention of DiTs by warping the frame-invariant semantic subspace of queries. We find that the noise predicted by the query-warped DiT naturally guides the diffusion trajectory toward the desired motion, and further show that leveraging this noise as self-guidance for latent optimization improves control stability and visual quality. Experiments show that QWERTY achieves the most effective motion control among existing training-free approaches on a recent image-to-video DiT, with performance comparable to fine-tuning-based methods.
Chinese Translation
视频扩散变换器（DiTs）生成高保真且时间一致的视频，但运动控制仍然是隐式的，主要依赖文本提示。因此，实现期望的运动通常需要大量的提示工程和重复重采样。虽然通过额外的空间提示（例如，边界框或点轨迹）微调模型可以实现显式控制，但这需要大量的数据整理和计算，并可能损害预训练模型的生成能力。因此，使用这些空间提示进行无训练运动控制在基于U-Net的视频扩散模型中得到了探索，但在DiTs中仍然基本未被研究。我们提出了QWERTY，这是一种无训练框架，通过用户定义的物体扭曲和光流，在预训练的图像到视频DiTs中实现灵活的运动控制。我们通过扭曲查询的帧不变语义子空间，仔细操控DiTs的3D全注意力。我们发现，查询扭曲的DiT预测的噪声自然引导扩散轨迹朝向期望的运动，并进一步表明，利用这种噪声作为潜在优化的自我引导可以提高控制的稳定性和视觉质量。实验表明，QWERTY在最近的图像到视频DiT上实现了现有无训练方法中最有效的运动控制，其性能可与基于微调的方法相媲美。
cs.CV / 54 / 2607.01871
Descriptor: LYNRED Mobility Dataset Multimodal Detection Subset (LYNRED-MDS)
描述符：LYNRED 移动数据集多模态检测子集 (LYNRED-MDS)
Abstract
Current road safety systems primarily focus on minimizing post-collision damage. However, advances in algorithmic perception are shifting focus toward early collision prediction, especially in lowvisibility conditions like nighttime or fog, where thermal infrared sensing outperforms both human vision and RGB imaging. While available RGB-infrared datasets such as FLIR ADAS and LLVIP are good benchmarks, they mostly consist of clear weather and overly simple scenarios. In this article, we introduce the LYNRED-MDS: Multimodal Detection Subset, a subset of the LYNRED Mobility Dataset, comprised of 4000 RGB-infrared image pairs captured under diverse weather, lighting, and road conditions around Grenoble, France. Our dataset spans varied driving contexts (urban, rural, mountainous, etc.) and a vehicle fleet compliant with Western European standards. Thermal cross-dataset evaluation using a YOLOv8n baseline suggests that our dataset offers strong generalization potential for pedestrian detection in driving scenarios. By covering critical edge cases, our dataset supports the development of more reliable and deployable vision systems for advanced driver-assistance systems.
Chinese Translation
当前的道路安全系统主要集中在最小化碰撞后的损害。然而，算法感知的进步正在将重点转向早期碰撞预测，特别是在夜间或雾霭等低能见度条件下，在这些情况下，热红外传感器的表现优于人类视觉和 RGB 成像。虽然现有的 RGB-红外数据集如 FLIR ADAS 和 LLVIP 是良好的基准，但它们大多由清晰天气和过于简单的场景组成。本文介绍了 LYNRED-MDS：多模态检测子集，这是 LYNRED 移动数据集的一个子集，包含 4000 对 RGB-红外图像，这些图像是在法国格勒诺布尔的多种天气、光照和道路条件下捕获的。我们的数据集涵盖了多样的驾驶环境（城市、乡村、山区等）以及符合西欧标准的车辆车队。使用 YOLOv8n 基线进行的热交叉数据集评估表明，我们的数据集在驾驶场景中的行人检测方面具有强大的泛化潜力。通过覆盖关键边缘案例，我们的数据集支持开发更可靠和可部署的视觉系统，以用于高级驾驶辅助系统。
cs.CV / 55 / 2607.01876
SAB-LVLM: Significance-Aware Binarization for Large Vision-Language Models
SAB-LVLM：面向大型视觉语言模型的显著性感知二值化
Abstract
Large Vision-Language Models (LVLMs) have achieved remarkable progress in multimodal understanding, yet their enormous parameter scale and cross-modal computation incur substantial memory and latency overhead, severely limiting real-world deployment on resource-constrained devices. Binarization offers an attractive solution by drastically reducing storage and computational costs. However, existing binarization methods neglect the varying importance of weights across different layers and modalities. This causes parameters irrelevant to downstream tasks to be unnecessarily retained, whereas modality-critical weights may not be adequately optimized, resulting in significant performance degradation. To address these challenges, we develop a novel \underline{S}ignificance-\underline{A}ware \underline{B}inarization for \underline{L}arge \underline{V}ision-\underline{L}anguage \underline{M}odels (SAB-LVLM). Specifically, after constructing Hessian matrices for textual and visual inputs, we propose a spatial significance map to distinguish full-precision weights activated under a single modality from those activated across modalities. We then devise a modality-guided integration strategy to obtain the significance-aware binarization map, which measures weight significance across layers and modalities. Subsequently, this binarization map is incorporated into the binarization objective as an error reweighting term, and binarization fitting is performed through an alternating significance-weighted update scheme. Extensive experiments illustrate the superiority of our SAB-LVLM over existing binary PTQ methods under an approximately 1-bit compression constraint. Our code is accessible at https://github.com/LyuQi127/SAB_LVLM.
Chinese Translation
大型视觉语言模型（LVLMs）在多模态理解方面取得了显著进展，但其庞大的参数规模和跨模态计算带来了巨大的内存和延迟开销，严重限制了在资源受限设备上的实际部署。二值化提供了一种有吸引力的解决方案，通过大幅降低存储和计算成本。然而，现有的二值化方法忽视了不同层和模态之间权重的重要性差异。这导致与下游任务无关的参数被不必要地保留，而对模态至关重要的权重可能未得到充分优化，从而导致显著的性能下降。为了解决这些挑战，我们开发了一种新颖的面向大型视觉语言模型的显著性感知二值化（SAB-LVLM）。具体而言，在为文本和视觉输入构建海森矩阵后，我们提出了一种空间显著性图，以区分在单一模态下激活的全精度权重与跨模态激活的权重。然后，我们设计了一种模态引导的集成策略，以获得显著性感知的二值化图，该图衡量不同层和模态之间的权重显著性。随后，该二值化图被纳入二值化目标，作为误差重加权项，通过交替显著性加权更新方案进行二值化拟合。大量实验表明，在约1位压缩约束下，我们的SAB-LVLM优于现有的二进制PTQ方法。我们的代码可在 https://github.com/LyuQi127/SAB_LVLM 获取。
cs.CV / 56 / 2607.01885
Diversity-aware View Partitioning for Scalable VGGT
关注多样性的视图划分以实现可扩展的VGGT
Abstract
Geometry transformers such as VGGT achieve strong performance by jointly reasoning over multiple views with global attention. However, scaling them to large view collections remains challenging due to the quadratic cost of attention. Moreover, our empirical analysis reveals that the reconstruction quality in VGGT is sensitive to the distribution of viewpoints. Simply increasing the number of views without sufficient viewpoint diversity can even degrade performance, as redundant views introduce highly similar tokens that dilute informative geometric signals in the attention mechanism. Motivated by this observation, we propose a training-free and plug-and-play VGGT inference framework that organizes views into diversity-aware balanced chunks. The chunks are constructed through combinatorial graph partitioning over visual dissimilarity and spatial dispersion. This view organization allows the transformer to focus attention on geometrically informative views while reducing redundant attention interactions. To estimate spatial dispersion without full pose estimation, we approximate spatial relationships via a soft pose propagation strategy based on visual similarity from a small set of seed frames. Extensive experiments demonstrate improved performance in camera pose estimation, multi-view depth prediction, and 3D reconstruction while reducing memory usage and inference latency. Our framework also complements existing VGGT variants, enabling scalable multi-view reconstruction without sacrificing geometric fidelity.
Chinese Translation
几何变换器如VGGT通过全局注意力对多个视图进行联合推理，从而实现强大的性能。然而，由于注意力的二次成本，将其扩展到大型视图集合仍然具有挑战性。此外，我们的实证分析揭示了VGGT中的重建质量对视点分布的敏感性。仅仅增加视图数量而没有足够的视点多样性甚至可能会降低性能，因为冗余视图引入高度相似的标记，从而稀释了注意力机制中的信息几何信号。基于这一观察，我们提出了一种无训练且即插即用的VGGT推理框架，该框架将视图组织成关注多样性的平衡块。这些块通过对视觉差异和空间分散进行组合图划分构建而成。这种视图组织使变换器能够将注意力集中在几何信息丰富的视图上，同时减少冗余的注意力交互。为了在不进行完整姿态估计的情况下估计空间分散，我们通过基于少量种子帧的视觉相似性，采用软姿态传播策略来近似空间关系。大量实验表明，在相机姿态估计、多视图深度预测和3D重建方面性能有所提升，同时减少了内存使用和推理延迟。我们的框架还补充了现有的VGGT变体，实现了可扩展的多视图重建，而不牺牲几何保真度。
cs.CV / 57 / 2607.01900
FoundDP: Revisiting Weak Disparity Observability in Dual-Pixel Depth Estimation
FoundDP：重新审视双像素深度估计中的弱视差可观测性
Abstract
Dual-pixel (DP) imaging enables metric depth estimation from a single camera using sub-aperture disparity. However, the extremely small effective baseline limits disparity observability, leading to structural degradation and depth failure in textureless, low-contrast, or downsampled regions. Existing DP-based methods rely primarily on local disparity cues and therefore become unreliable when disparity signals are weak or ambiguous. To address this limitation, we propose \emph{FoundDP}, a unified framework that integrates metric DP depth with global structural priors from a monocular depth foundation model. Our method preserves metric scale through DP-derived depth and leverages Vision Transformer (ViT) features to restore structural consistency in weak-disparity regions. To ensure reliable metric guidance under DP imaging conditions, we identify and mitigate ViT representation degradation induced by DP defocus blur via ViT feature alignment, enabling stable metric-guided depth estimation. Extensive experiments on synthetic and real-world DP benchmarks show that FoundDP delivers superior performance, with consistent gains in structural fidelity and metric accuracy, especially under reduced disparity observability. Code will be available at: https://github.com/EchoLighting/FoundDP
Chinese Translation
双像素（DP）成像通过子孔径视差实现了从单个相机进行度量深度估计。然而，极小的有效基线限制了视差的可观测性，导致在无纹理、低对比度或下采样区域的结构退化和深度失败。现有的基于DP的方法主要依赖于局部视差线索，因此在视差信号弱或模糊时变得不可靠。为了解决这一局限性，我们提出了 extit{FoundDP}，一个统一框架，将度量DP深度与来自单目深度基础模型的全局结构先验相结合。我们的方法通过DP导出的深度保持度量尺度，并利用视觉变换器（Vision Transformer, ViT）特征恢复弱视差区域的结构一致性。为了确保在DP成像条件下可靠的度量指导，我们识别并减轻了由DP散焦模糊引起的ViT表示退化，通过ViT特征对齐实现稳定的度量引导深度估计。在合成和真实世界的DP基准测试中，广泛的实验表明FoundDP提供了卓越的性能，在结构保真度和度量准确性方面持续提升，特别是在视差可观测性降低的情况下。代码将可在以下链接获取：https://github.com/EchoLighting/FoundDP
cs.CV / 58 / 2607.01902
Rethinking Post-Hoc Calibration in Semantic Segmentation
重新思考语义分割中的事后校准
Abstract
Reliable confidence estimates are essential in semantic segmentation, especially in safety-critical settings where overconfident errors can mislead downstream decisions. Yet modern segmentation models often remain miscalibrated. Post-hoc calibration offers a practical way to correct confidence estimates without retraining the segmentation model, but its use in dense prediction raises structural issues that are often overlooked. We study two such issues. First, adding a constant to all logits leaves the softmax probabilities unchanged, but several standard calibrators can still depend on this arbitrary offset. As a result, two logit representations encoding the same predictive distribution may yield different calibrated probabilities. We define translation-invariant (TI) calibrators as those whose outputs are unchanged under such shifts, characterize which common calibrators satisfy this property, and construct TI counterparts of shift-sensitive calibrators to isolate the effect of removing representation dependence. Second, post-hoc calibration is typically fitted by minimizing a likelihood-based objective, whereas segmentation models are trained with task-specific metrics such as Dice. This mismatch can cause calibration to alter class orderings and degrade the deployed segmentation map. We study decision-preserving calibration under argmax- and order-preservation constraints. Since enforcing these constraints collapses affine softmax calibrators to temperature scaling, we introduce class-conditional affine calibrators that can be made argmax- or order-preserving while retaining greater expressivity, allowing us to quantify the calibration-segmentation trade-off induced by decision preservation. Across natural-image and medical segmentation benchmarks, and under corruption-based covariate shift, matched comparisons show that TI variants generally improve calibration metrics, while decision-preserving variants prevent segmentation degradation and retain strong calibration performance. These results provide practical design principles for well-defined post-hoc calibration pipelines in semantic segmentation.
Chinese Translation
可靠的置信度估计在语义分割中至关重要，尤其是在安全关键的环境中，过于自信的错误可能会误导后续决策。然而，现代分割模型往往仍然存在校准不准确的问题。事后校准提供了一种在不重新训练分割模型的情况下纠正置信度估计的实用方法，但其在密集预测中的应用引发了一些常被忽视的结构性问题。我们研究了两个此类问题。首先，向所有 logits 添加一个常数不会改变 softmax 概率，但一些标准校准器仍可能依赖于这个任意偏移。因此，编码相同预测分布的两个 logit 表示可能会产生不同的校准概率。我们将平移不变（translation-invariant, TI）校准器定义为在此类偏移下输出不变的校准器，描述哪些常见校准器满足这一特性，并构建 TI 版本的对偏移敏感的校准器，以隔离去除表示依赖性的影响。其次，事后校准通常通过最小化基于似然的目标来拟合，而分割模型则是通过特定任务的度量（如 Dice）进行训练。这种不匹配可能导致校准改变类别排序，并降低部署的分割图的质量。我们在 argmax 和排序保持约束下研究决策保持校准。由于强制这些约束使得仿射 softmax 校准器退化为温度缩放，我们引入了类别条件仿射校准器，这些校准器可以在保持更高表达能力的同时实现 argmax 或排序保持，从而使我们能够量化由于决策保持引起的校准与分割之间的权衡。在自然图像和医学分割基准测试中，以及在基于腐败的协变量转移下，匹配比较显示 TI 变体通常改善校准指标，而决策保持变体则防止分割退化并保持强大的校准性能。这些结果为语义分割中明确的事后校准管道提供了实用的设计原则。
cs.CV / 59 / 2607.01906
SFKD: Spatial--Frequency Joint-Aware Heterogeneous Knowledge Distillation via Multi-Level Wavelet Spectral Interaction
SFKD：基于多级小波谱交互的空间-频率联合感知异构知识蒸馏
Abstract
Most existing knowledge distillation methods focus on homogeneous models (e.g., CNN-to-CNN), thereby overlooking the flexibility and potential of knowledge transfer across heterogeneous models. Due to intrinsic inductive bias discrepancies between heterogeneous models that cause spatial distribution inconsistencies, prior heterogeneous distillation methods often weaken or discard spatial information in heterogeneous representations. However, the spatial information in representations often encodes transferable global structural semantics as well as architecture-specific local details, and therefore should not be directly ignored. To better leverage the spatial information encoded in heterogeneous representations, we propose a Spatial-Frequency Joint-Aware Heterogeneous Knowledge Distillation framework (SFKD). By leveraging the complementary properties of wavelet transform spatial locality and Fourier representations in characterizing global energy distributions, we first apply multi-level discrete wavelet transform to explicitly decouple spatial information. The resulting wavelet sub-bands are further refined by a dual-stream dual-stage refinement module, and finally combined with a Gaussian-filtered frequency loss to selectively capture informative global information. Extensive experiments on multiple benchmark datasets under both homogeneous and heterogeneous models demonstrate the superiority of our method.
Chinese Translation
大多数现有的知识蒸馏方法集中于同构模型（例如，CNN到CNN），因此忽视了跨异构模型知识转移的灵活性和潜力。由于异构模型之间固有的归纳偏差差异导致空间分布不一致，先前的异构蒸馏方法通常削弱或丢弃异构表示中的空间信息。然而，表示中的空间信息通常编码了可转移的全局结构语义以及特定于架构的局部细节，因此不应被直接忽视。为了更好地利用异构表示中编码的空间信息，我们提出了一种空间-频率联合感知异构知识蒸馏框架（SFKD）。通过利用小波变换的空间局部性和傅里叶表示在表征全局能量分布中的互补特性，我们首先应用多级离散小波变换来明确解耦空间信息。得到的小波子带通过双流双阶段精炼模块进一步优化，最终与高斯滤波的频率损失结合，以选择性地捕获信息丰富的全局信息。在多个基准数据集上进行的广泛实验表明，我们的方法在同构和异构模型下均表现出优越性。
cs.CV / 60 / 2607.01908
Towards Real-World Ultrasound Understanding: Large Vision-Language Models from Multi-Image Examinations with Long-Form Reports
迈向现实世界的超声理解：来自多图像检查与长格式报告的大型视觉语言模型
Abstract
Large vision-language models (LVLMs) have achieved strong performance across many medical imaging tasks, yet their application to ultrasound remains limited due to its inherent complexity and variability. In this work, we revisit what is truly needed to enable real-world ultrasound understanding. Instead of introducing complex architectures or elaborate training strategies, we show that data scale and clinically faithful data alignment are the key factors. We construct a large-scale dataset of 1.5M real-world ultrasound examinations, containing 17.7M images, multi-organ coverage, and paired uncurated clinical reports. Crucially, we organize the data at the examination level, aligning multiple images with their corresponding reports to reflect real clinical workflows. We then fine-tune a standard LVLM using low-rank adaptation (LoRA) on this dataset without task-specific modifications. Surprisingly, this simple recipe already leads to strong performance across diverse ultrasound understanding tasks, outperforming prior methods designed with more complex pipelines. Beyond these results, we present model and data scaling analyses that provide insights into the role of scale in ultrasound LVLMs.
Chinese Translation
大型视觉语言模型（LVLMs）在许多医学影像任务中取得了强劲的表现，但由于超声的固有复杂性和变异性，其应用仍然有限。在本研究中，我们重新审视了实现现实世界超声理解所需的真正条件。我们并没有引入复杂的架构或精细的训练策略，而是展示了数据规模和临床真实数据对齐是关键因素。我们构建了一个大规模数据集，包含150万例真实世界超声检查，涵盖1770万张图像、多脏器覆盖以及配对的未经整理的临床报告。至关重要的是，我们在检查级别组织数据，将多张图像与其对应的报告对齐，以反映真实的临床工作流程。然后，我们在该数据集上使用低秩适应（LoRA）对标准LVLM进行微调，而无需特定任务的修改。令人惊讶的是，这一简单的方法在多样的超声理解任务中已经表现出强劲的性能，超越了以更复杂流程设计的先前方法。除了这些结果，我们还展示了模型和数据规模分析，为超声LVLM中规模的作用提供了见解。
cs.CV / 61 / 2607.01915
Robust Image Processing Techniques for Construction Environment Monitoring Using Underwater Robots
基于水下机器人建设环境监测的稳健图像处理技术
Abstract
This paper proposes a robust image processing framework for underwater robot-based construction environment monitoring, targeting complex degradations observed in real marine environments. Unlike conventional approaches that mainly consider absorption and backscattering, real underwater imagery is strongly affected by depth-dependent forward scattering blur and particle-induced degradations such as marine snow. To address this, we introduce a staged processing pipeline that sequentially models background degradation via depth-aware forward scattering and foreground degradation using realistic marine snow patterns extracted from real images. The resulting synthetic data are used to retrain an existing Joint-ID network without modifying its architecture, enabling an isolated evaluation of dataset realism. In addition, a lightweight post-processing scheme is applied to enhance contrast and structural clarity. Experiments on real underwater datasets collected in Korean coastal environments demonstrate consistent improvements in visual quality and UIQM scores. The results indicate that explicitly modeling forward scattering and realistic particle effects effectively reduces the synthetic-to-real gap and improves practical applicability in real-world underwater robotic operations.
Chinese Translation
本文提出了一种稳健的图像处理框架，用于基于水下机器人的建设环境监测，旨在应对真实海洋环境中观察到的复杂退化现象。与主要考虑吸收和反向散射的传统方法不同，真实的水下图像受到深度依赖的前向散射模糊和由颗粒引起的退化（如海洋雪）的强烈影响。为了解决这一问题，我们引入了一个分阶段的处理管道，依次通过深度感知的前向散射模型背景退化，并使用从真实图像中提取的现实海洋雪模式来模拟前景退化。生成的合成数据用于重新训练现有的Joint-ID网络，而无需修改其架构，从而实现对数据集真实性的独立评估。此外，应用了一种轻量级的后处理方案，以增强对比度和结构清晰度。在韩国沿海环境中收集的真实水下数据集上的实验表明，视觉质量和UIQM评分均有一致的改善。结果表明，显式建模前向散射和现实颗粒效应有效缩小了合成与真实之间的差距，并提高了在真实水下机器人操作中的实际适用性。
cs.CV / 62 / 2607.01928
Sparse-Aware Vector Quantization for Bandwidth-Efficient Collaborative 3D Semantic Occupancy Prediction
面向带宽效率的协作3D语义占用预测的稀疏感知向量量化
Abstract
Collaborative perception extends single-agent perception by enabling multiple vehicles to exchange complementary perceptual information. However, it introduces an inherent trade-off between perception gain and communication overhead, which is particularly severe for 3D semantic occupancy prediction that relies on fine-grained spatial structures. Existing methods typically compress 3D features into 2D, causing severe spatial information loss, or transmit dense 3D representations, hindering real-world deployment. To overcome these limitations, we propose a bandwidth-efficient collaborative Vector Quantization Semantic Occupancy Prediction (VQSOP) framework. VQSOP employs a Sparse-Aware Vector Quantization (SAVQ) mechanism that exploits 3D scene sparsity to compactly encode informative regions, drastically reducing communication overhead while preserving complete geometric context. Furthermore, to enhance structural consistency and feature continuity, we design a Dual-Branch Adaptive Spatial Refinement (ASR) module that dynamically fuses local high-frequency details with broad contextual semantics. Extensive experiments demonstrate that our approach achieves state-of-the-art performance while reducing communication volume by up to 82x.
Chinese Translation
协作感知通过使多个车辆能够交换互补的感知信息，扩展了单一代理的感知能力。然而，这引入了感知增益与通信开销之间的固有权衡，这在依赖于细粒度空间结构的3D语义占用预测中尤为严重。现有方法通常将3D特征压缩为2D，导致严重的空间信息损失，或者传输密集的3D表示，妨碍了在实际环境中的应用。为克服这些限制，我们提出了一种带宽高效的协作向量量化语义占用预测（VQSOP）框架。VQSOP采用了一种稀疏感知向量量化（SAVQ）机制，利用3D场景的稀疏性紧凑编码信息丰富的区域，显著降低通信开销，同时保留完整的几何上下文。此外，为了增强结构一致性和特征连续性，我们设计了一个双分支自适应空间细化（ASR）模块，动态融合局部高频细节与广泛的上下文语义。大量实验表明，我们的方法在实现最先进性能的同时，通信量减少了多达82倍。
cs.CV / 63 / 2607.01949
LiZAD: A Lightweight Zero-Shot Anomaly Detection Framework for Industrial Manufacturing
LiZAD：一种轻量级零样本异常检测框架用于工业制造
Abstract
In modern high-throughput industrial production lines, product configurations and visual characteristics frequently change, making it impractical to collect and annotate data for every new scenario. This dynamic setting makes Zero-Shot Anomaly Detection (ZSAD) particularly suitable, as it enables defect detection without requiring training on target-specific samples. Although recent ZSAD approaches show promising results, they are computationally intensive and thus unsuitable for deployment on resource-constrained devices. We propose LiZAD: a lightweight framework designed for real-time ZSAD specifically tailored for use on edge devices. The proposed approach pairs the dense and spatially aware visual features of DINOv3, crucial for precise pixel-level localization, with the highly computationally efficient text embeddings of MobileCLIP2. These features are then mapped into a shared latent space via low-memory trainable projection heads. Compared to six state-of-the-art ZSAD models, LiZAD achieves an average memory reduction of 61.5%, a parameter reduction of 74.6%, and a speedup of 3.02x in terms of latency. Despite substantial reductions in computational and memory costs, our approach maintains competitive anomaly detection performance, dropping the average P-AUROC by just 6.4% relative to the best state-of-the-art model across the VisA, BTAD, MPDD, and MVTec-AD datasets. Finally, it is successfully deployed on the NVIDIA Jetson NX and Jetson AGX edge devices and tested on the real production line of the Industrial Computer Engineering Laboratory (ICE Lab) at the University of Verona. The code is available at https://github.com/intelligolabs/LiZAD.
Chinese Translation
在现代高通量工业生产线上，产品配置和视觉特征经常变化，因此为每个新场景收集和标注数据变得不切实际。这种动态环境使得零样本异常检测（Zero-Shot Anomaly Detection, ZSAD）特别适用，因为它能够在不需要针对特定样本进行训练的情况下进行缺陷检测。尽管最近的ZSAD方法显示出良好的效果，但它们计算密集，因此不适合在资源受限的设备上部署。我们提出了LiZAD：一个专为边缘设备实时使用而设计的轻量级ZSAD框架。该方法将DINOv3的密集且具有空间感知的视觉特征与MobileCLIP2的高计算效率文本嵌入相结合，这对于精确的像素级定位至关重要。这些特征通过低内存可训练的投影头映射到共享的潜在空间。与六种最先进的ZSAD模型相比，LiZAD在内存使用上平均减少了61.5%，参数减少了74.6%，延迟加速了3.02倍。尽管计算和内存成本大幅降低，我们的方法仍保持竞争力的异常检测性能，相较于最佳的最先进模型在VisA、BTAD、MPDD和MVTec-AD数据集上的平均P-AUROC仅下降了6.4%。最后，该方法成功部署在NVIDIA Jetson NX和Jetson AGX边缘设备上，并在维罗纳大学工业计算工程实验室（ICE Lab）的真实生产线上进行了测试。代码可在https://github.com/intelligolabs/LiZAD获取。
cs.CV / 64 / 2607.01952
Personalized 4D Whole-Heart Mesh Reconstruction from Cine MRI via Multi-Scale Temporal Modeling and Differentiable Contour Rendering
基于多尺度时间建模和可微轮廓渲染的个性化4D全心脏网格重建
Abstract
Accurate 4D whole-heart mesh reconstruction from sparse cine MRI is critical for creating cardiac digital twins, but remains challenging due to limited 2D slice coverage and the complex coupling between cardiac shape and motion. Existing methods often rely on intermediate contour fitting and typically reconstruct static, single-phase, or partial cardiac geometries, limiting their ability to capture full-chamber dynamics. We propose a novel end-to-end framework for reconstructing temporally resolved whole-heart meshes from multi-view 2D cine MRI sequences by learning an image-to-mesh mapping. The framework incorporates a differentiable contour renderer inspired by the Beer-Lambert attenuation principle, enabling anatomy-aware supervision of 3D+t mesh deformation through contour-based projection losses. To improve temporal consistency across the cardiac cycle, we further introduce a multi-scale temporal modeling module that integrates global cycle-level dynamics with local inter-frame coherence to generate smooth and physiologically plausible mesh trajectories. The proposed method achieved a whole-heart mean absolute error of 1.68 $\pm$ 0.31 mm and a motion jitter of 0.77 $\pm$ 0.17 $\mathrm{mm}/\mathrm{frame}^{3}$, outperforming existing methods with lower reconstruction error and substantially improved motion smoothness. It also improved 2D contour alignment across multiple cine MRI views and supported downstream proof-of-concept electrophysiological simulation. The code will be released publicly upon acceptance of the manuscript for publication.
Chinese Translation
从稀疏的动态磁共振成像（cine MRI）中准确重建4D全心脏网格对于创建心脏数字双胞胎至关重要，但由于二维切片覆盖有限以及心脏形状与运动之间的复杂耦合，这一过程仍然具有挑战性。现有方法通常依赖于中间轮廓拟合，通常重建静态、单相或部分心脏几何形状，限制了其捕捉全腔动态的能力。我们提出了一种新颖的端到端框架，通过学习图像到网格的映射，从多视角的二维动态MRI序列中重建时间分辨的全心脏网格。该框架结合了受Beer-Lambert衰减原理启发的可微轮廓渲染器，使得通过基于轮廓的投影损失实现对3D+t网格变形的解剖学感知监督。为了提高心脏周期内的时间一致性，我们进一步引入了一个多尺度时间建模模块，将全局周期级动态与局部帧间一致性相结合，以生成平滑且生理上合理的网格轨迹。所提方法在全心脏的平均绝对误差为1.68 ± 0.31 mm，运动抖动为0.77 ± 0.17 mm/frame³，优于现有方法，具有更低的重建误差和显著改善的运动平滑性。它还改善了多个动态MRI视图之间的二维轮廓对齐，并支持下游概念验证的电生理模拟。代码将在手稿接受发表后公开发布。
cs.CV / 65 / 2607.01962
NeoMap: Training-free Novel-View Synthesis from Single Images and Videos
NeoMap：无训练的新视角合成方法，基于单幅图像和视频
Abstract
We study the challenging problem of novel view video synthesis from single images or monocular videos. Existing methods, which operate under the assumption that pre-trained video models lack native novel view synthesis capability and enforce view alignment via camera conditioning, task-specific fine-tuning, or stepwise hard denoising guidance, often suffer from artifacts and compromised global scene consistency. In this paper, we introduce NeoMap, a novel training-free framework designed to locate high-fidelity, view-consistent novel view solutions from general pre-trained video models. The key to our approach is the core insight that promising novel view solutions are inherently encoded within the natural video data manifold learned by pre-trained models, and the core challenge is simply to locate this optimal solution. We solve this via our core mechanism: convergent manifold alternating projection iterations that optimize the initial noise. Extensive experiments demonstrate that NeoMap significantly outperforms all existing methods across 3 standard novel view synthesis benchmarks, including the challenging Tanks-and-Temples, LLFF and DAVIS datasets, achieving state-of-the-art generation fidelity and top-tier view consistency.
Chinese Translation
我们研究了从单幅图像或单目视频合成新视角视频这一具有挑战性的问题。现有方法在假设预训练视频模型缺乏原生新视角合成能力的前提下，通过相机条件、特定任务的微调或逐步的硬去噪指导来强制视角对齐，往往会遭遇伪影和全局场景一致性受损的问题。本文介绍了NeoMap，一种新颖的无训练框架，旨在从通用预训练视频模型中定位高保真、视角一致的新视角解决方案。我们方法的关键在于核心洞察，即有前景的新视角解决方案本质上已编码在预训练模型学习到的自然视频数据流形中，而核心挑战仅在于定位这一最优解决方案。我们通过核心机制：收敛流形交替投影迭代来优化初始噪声，解决了这一问题。大量实验表明，NeoMap在包括具有挑战性的Tanks-and-Temples、LLFF和DAVIS数据集在内的3个标准新视角合成基准上，显著优于所有现有方法，达到了最先进的生成保真度和顶级的视角一致性。
cs.CV / 66 / 2607.01973
Assessing VLM Reliability for Medical Image Quality Evaluation Under Corruption and Bias
评估视觉语言模型在腐败和偏见下的医学图像质量评估可靠性
Abstract
Vision-Language Models (VLMs) are increasingly applied in medical tasks such as pathology description, report generation, and visual question answering. Medical Image Quality Assessment (MIQA) supports diagnostic accuracy and patient safety by determining whether images meet the standards required for clinical decision-making. Automating MIQA with VLMs may reduce workload, but their behavior under real-world conditions, where images may be degraded or textual context may affect judgments, should be further explored before deployment. We benchmark VLMs on medical image quality using the MediMeta-C dataset zero-shot across seven corruption types and five severity levels. We evaluate sensitivity to degradation patterns, the effect of corruptions on embedding geometry, and whether textual attributes (demographics, expertise, infrastructure, institution) alter scores. Across 16 VLMs and seven modalities, pixelation produced the largest score reductions (mean -20.58%, up to -34.4% for OCT), whereas brightness had limited effect (-0.81%). Embedding displacement was associated with score changes. Same-family models showed correlations of 0.67-0.83; some produced increases up to +31% for corrupted mammography. Textual attributes affected scores: institutional prestige raised them +17.15%, and equipment age lowered them -14.7%. The largest changes were +95.62% (InternVL-8B) and -37.7% (MedGemma). Current VLMs show limitations for medical image quality assessment. Pixelation, a privacy-preserving transformation, reduces performance, indicating a trade-off between patient privacy and reliability. Sensitivity to contextual metadata indicates limited objectivity and marks metadata as a privacy and bias source. Privacy protection and objective quality assessment are related requirements for use.
Chinese Translation
视觉语言模型（VLMs）在医学任务中越来越多地应用，例如病理描述、报告生成和视觉问答。医学图像质量评估（MIQA）通过确定图像是否符合临床决策所需的标准，支持诊断准确性和患者安全性。利用VLMs自动化MIQA可能减少工作负担，但在实际条件下（图像可能被降质或文本上下文可能影响判断）它们的表现应在部署前进一步探讨。我们使用MediMeta-C数据集对VLMs在医学图像质量上的表现进行基准测试，采用零样本方法评估七种腐败类型和五个严重程度。我们评估了对降质模式的敏感性、腐败对嵌入几何的影响，以及文本属性（人口统计、专业知识、基础设施、机构）是否改变评分。在16个VLM和七种模态中，像素化导致了最大的评分降低（平均-20.58%，对于光学相干断层扫描（OCT）高达-34.4%），而亮度的影响有限（-0.81%）。嵌入位移与评分变化相关。同类模型之间的相关性为0.67-0.83；一些模型在受损的乳腺X光检查中产生了高达+31%的评分提升。文本属性影响评分：机构声望提升了+17.15%，而设备年龄降低了-14.7%。最大的变化为+95.62%（InternVL-8B）和-37.7%（MedGemma）。目前的VLM在医学图像质量评估中显示出局限性。像素化作为一种保护隐私的转换，降低了性能，表明患者隐私与可靠性之间存在权衡。对上下文元数据的敏感性表明客观性有限，并将元数据标记为隐私和偏见的来源。隐私保护和客观质量评估是使用的相关要求。
cs.CV / 67 / 2607.01982
MolSight: A Graph-Aware Vision-Language Model for Unified Chemical Image Understanding
MolSight：一种图感知的视觉-语言模型用于统一化学图像理解
Abstract
Using molecular large language models (LLMs) as a unified framework for understanding molecular structures and functions is emerging as a new trend in tasks such as molecular design and drug discovery. However, these models struggle to fully capture the visual representation of molecular structures, limiting their potential. While existing molecular vision-language models (VLMs) show promise, they still face challenges in structural alignment and lack the necessary topological modeling for accurate molecular understanding. To address this, we propose MolSight, a graph-aware vision-language model framework designed to enhance the understanding of molecular images by VLMs. MolSight integrates a Molecular Topology Module to inject chemical-bond adjacency information into vision tokens, and a Molecular Grounding Module to align visual features with chemical symbolic semantics. Our experiments demonstrate that MolSight significantly outperforms existing VLMs, molecular LLMs, and specialized tools across multiple chemical visual understanding tasks, achieving a new level of molecular image reasoning.
Chinese Translation
使用分子大型语言模型（LLMs）作为理解分子结构和功能的统一框架正在成为分子设计和药物发现等任务中的新趋势。然而，这些模型在充分捕捉分子结构的视觉表征方面存在困难，限制了它们的潜力。虽然现有的分子视觉-语言模型（VLMs）表现出一定的前景，但它们在结构对齐方面仍面临挑战，并且缺乏准确理解分子的必要拓扑建模。为了解决这一问题，我们提出了MolSight，一种图感知的视觉-语言模型框架，旨在通过VLMs增强对分子图像的理解。MolSight集成了一个分子拓扑模块，将化学键邻接信息注入视觉标记中，并且一个分子定位模块将视觉特征与化学符号语义对齐。我们的实验表明，MolSight在多个化学视觉理解任务中显著优于现有的VLMs、分子LLMs和专用工具，实现了分子图像推理的新水平。
cs.CV / 68 / 2607.01983
Open-Weather Robust 3D Detection via Dual-Critic Diffusion Alignment
基于双评论家扩散对齐的开放天气鲁棒3D检测
Abstract
Robust 3D object detection under adverse weather remains a critical hurdle for autonomous driving. Despite progress with LiDAR-4D radar fusion, most methods are constrained by a closed-world assumption, implicitly requiring training and test weather to align in both type and severity. This premise fails in practice: the open-ended nature of weather, and even variations within a single type like rain, cause dramatically different LiDAR degradation patterns, leading to significant performance drops in unseen conditions. To address this, we present Dual-Critic Guided Diffusion Alignment (DCDA), a weather-agnostic framework that learns to recover degraded LiDAR features toward a clean manifold. Rather than modeling specific weather types, DCDA employs a 4D radar-conditioned diffusion process to progressively refine features, guided by two complementary critics. (i) A detection-guided critic, anchored by a pre-trained clean-weather model, ensures that the refined features retain object-level discriminability and localization accuracy. (ii) A weather adversarial critic enforces holistic distributional consistency with clean-weather representations. By aligning features through semantic and distributional constraints rather than explicit weather modeling, DCDA generalizes effectively to unseen weather types and severities without requiring paired data or weather labels. We further introduce a structured open-weather benchmark with held-out type-severity combinations and extensive experiments verify DCDA's advantages.
Chinese Translation
在恶劣天气条件下进行鲁棒的3D物体检测仍然是自动驾驶面临的一个关键难题。尽管在LiDAR-4D雷达融合方面取得了一定进展，但大多数方法受到封闭世界假设的限制，隐含要求训练和测试的天气在类型和严重性上保持一致。这一前提在实践中并不成立：天气的开放性特征，甚至同一类型天气（如降雨）内部的变化，导致LiDAR退化模式显著不同，从而在未见条件下造成性能大幅下降。为了解决这一问题，我们提出了双评论家引导的扩散对齐（Dual-Critic Guided Diffusion Alignment, DCDA），这是一个与天气无关的框架，旨在学习恢复退化的LiDAR特征，使其趋向于干净的流形。DCDA并不针对特定的天气类型，而是采用4D雷达条件下的扩散过程，逐步细化特征，并由两个互补的评论家引导。 (i) 一个基于检测的评论家，依托于预训练的干净天气模型，确保细化后的特征保持物体级别的可区分性和定位准确性。 (ii) 一个天气对抗评论家则强制执行与干净天气表征的整体分布一致性。通过语义和分布约束而非显式的天气建模来对齐特征，DCDA能够有效地推广到未见的天气类型和严重性，而无需配对数据或天气标签。我们进一步引入了一个结构化的开放天气基准，包含保留的类型-严重性组合，并通过大量实验验证了DCDA的优势。
cs.CV / 69 / 2607.01987
Understanding Geometric Representations in Self-Supervised Vision Transformers via Subspace Intervention
通过子空间干预理解自监督视觉变换器中的几何表示
Abstract
We introduce a controlled subspace intervention framework to investigate how self-supervised Vision Transformers (ViTs) encode dense geometric information. While linear probing is widely used to assess geometric representations, it treats features as a black box, failing to disentangle the underlying topology. To address this issue, we decompose the weights of converged linear probes to isolate the low-rank subspaces containing explicit geometric signals using Singular Value Decomposition (SVD). Our perspective yields three key insights: (1) Pre-training objectives determine how features are encoded. DINOv2 aligns spatial features for efficient linear extraction, while Masked Autoencoders (MAE) tend to disperse these signals, requiring a broader spatial context. (2) Explicit geometric representations are highly compressible, suggesting dense predictive heads could potentially be constrained to low-rank subspaces with minimal performance loss. (3) The layer-wise task affinity suggests that geometric precision peaks at intermediate layers before yielding to semantic abstraction in the final layers. By connecting internal encoding mechanics with downstream performance, these findings provide a basis for effective feature selection and lightweight decoder design. The source code is available at https://github.com/Zhou-Weichen/Geosubprobe.
Chinese Translation
我们引入了一个受控的子空间干预框架，以研究自监督视觉变换器（ViTs）如何编码密集的几何信息。尽管线性探测广泛用于评估几何表示，但它将特征视为黑箱，未能解开其底层拓扑。为了解决这个问题，我们通过奇异值分解（SVD）分解收敛线性探测器的权重，以隔离包含显式几何信号的低秩子空间。我们的视角得出了三个关键见解：（1）预训练目标决定了特征的编码方式。DINOv2 对空间特征进行对齐，以实现高效的线性提取，而掩码自编码器（MAE）则倾向于分散这些信号，需要更广泛的空间上下文。（2）显式几何表示具有高度可压缩性，这表明密集预测头可能被限制在低秩子空间中，且性能损失最小。（3）逐层任务亲和性表明，几何精度在中间层达到峰值，然后在最后层转向语义抽象。通过将内部编码机制与下游性能联系起来，这些发现为有效特征选择和轻量解码器设计提供了基础。源代码可在 https://github.com/Zhou-Weichen/Geosubprobe 获取。
cs.CV / 70 / 2607.01990
Training-free Controllable Human Motion Generation under Heterogeneous Constraints
无训练可控的人体运动生成在异构约束下
Abstract
Training-free controllable motion generation has attracted growing interest for enabling flexible constraint enforcement without constraint-specific training. However, existing training-free methods require constraints to be continuous objective-based with differentiable losses, while many real-world requirements are criterion-based and provide only discontinuous, sparse, or even black-box feedback. In this paper, we propose Motion-Inference-as-Control (MIC), the first training-free motion generation framework that handles both continuous objective-based and criterion-based motion constraints under a shared mechanism. The key idea is to cast diffusion-based motion generation as a stochastic control problem. This perspective not only provides principled and practically effective step-wise control laws that support criterion-based constraints without requiring differentiability and naturally accommodate objective-based constraints as a special case, but also motivates a control-oriented constraint coordination mechanism that adaptively balances and reconciles motion constraints during generation. Experiments across diverse constraint settings demonstrate the effectiveness of our framework.
Chinese Translation
无训练的可控运动生成因能够灵活执行约束而受到越来越多的关注，而无需针对特定约束进行训练。然而，现有的无训练方法要求约束是基于连续目标的，并具有可微损失，而许多现实世界的需求是基于标准的，仅提供不连续、稀疏或甚至黑箱的反馈。在本文中，我们提出了运动推理作为控制（Motion-Inference-as-Control, MIC），这是第一个处理连续目标基础和基于标准的运动约束的无训练运动生成框架，采用共享机制。其关键思想是将基于扩散的运动生成视为一个随机控制问题。这一视角不仅提供了原则性和实用有效的逐步控制法则，支持基于标准的约束而无需可微性，并自然地将基于目标的约束视为特例，还激励了一种控制导向的约束协调机制，在生成过程中自适应地平衡和调和运动约束。跨多种约束设置的实验验证了我们框架的有效性。
cs.CV / 71 / 2607.02015
Mirror Illusion Art
镜像幻觉艺术
Abstract
Mirror Illusion Art is a novel reflection-conditioned 3D illusion where one object yields two target appearances (front and mirror). The task is formulated as inverse design from two target 2D images (front and mirror) to a printable 3D object with geometry and texture. Prior topology-driven and shadow-based approaches demand substantial manual effort, optimize shape only, and often yield non-smooth or incomplete geometry. To address these challenges, we propose AutoMIA, an automated Mirror Illusion Art design pipeline that jointly optimizes shape and color. To stabilize optimization and suppress artifacts, four mechanisms are introduced: (1) projection-alignment component (PAC) selection to reduce surface noise, (2) position-weighted adaptive (PWA) suppression for background noise, (3) internal voxel preservation (IVP) to prevent internal fractures, and (4) shape-color decoupled (SCD) optimization that balance shape and color optimization. AutoMIA generate diverse smooth Mirror Illusion artworks successfully both in the digital and physical world, with only around 76s design time and 2.6 GB memory on average using a single RTX 3090, advancing inverse graphics and computational design. Our code is available at https://github.com/zxp555/AutoMIA.
Chinese Translation
镜像幻觉艺术是一种新颖的反射条件下的三维幻觉，其中一个物体呈现出两种目标外观（正面和镜面）。该任务被构建为从两个目标二维图像（正面和镜面）反向设计出一个可打印的三维物体，包含几何形状和纹理。以往的基于拓扑和阴影的方法需要大量手动努力，仅优化形状，且常常产生不光滑或不完整的几何形状。为了解决这些挑战，我们提出了AutoMIA，一个自动化的镜像幻觉艺术设计流程，能够联合优化形状和颜色。为了稳定优化并抑制伪影，引入了四个机制：（1）投影对齐组件（PAC）选择以减少表面噪声，（2）位置加权自适应（PWA）抑制背景噪声，（3）内部体素保持（IVP）以防止内部裂纹，以及（4）形状-颜色解耦（SCD）优化，以平衡形状和颜色的优化。AutoMIA成功生成了多样化的光滑镜像幻觉艺术作品，无论是在数字世界还是物理世界，平均设计时间约为76秒，内存使用约为2.6 GB，使用单个RTX 3090，推动了逆向图形学和计算设计的发展。我们的代码可在https://github.com/zxp555/AutoMIA获取。
cs.CV / 72 / 2607.02018
UnderOneFacade: Worldwide Facade Semantic Segmentation Benchmark Dataset
UnderOneFacade：全球建筑立面语义分割基准数据集
Abstract
Globally consistent semantic digital twins require centimeter-accurate and geographically transferable 3D facade segmentation. However, progress in facade parsing is limited by the lack of large-scale, standardized benchmarks for evaluating cross-domain generalization. Existing datasets are geographically narrow, semantically inconsistent, or insufficiently precise. We introduce UnderOneFacade, the largest cross-country and cross-continent 3D facade benchmark to date, comprising centimeter-accurate point clouds with hierarchical, harmonized, and architecturally grounded semantic labels totaling 2.7 billion annotated points. Through a systematic evaluation of representative point-, graph- and transformer-based architectures, we show that current methods struggle to recognize fine-grained architectural elements and degrade significantly across geographic domains, with the best models achieving only up to 33 IoU on the fine-grained LoFG3 benchmark. By combining geometric precision with standardized semantics at unprecedented scale, UnderOneFacade establishes a rigorous benchmark for developing robust and transferable 3D segmentation models. The dataset, evaluation scripts, and pretrained models will be released upon publication.
Chinese Translation
全球一致的语义数字双胞胎需要厘米级精度和地理可转移的三维建筑立面分割。然而，建筑立面解析的进展受到缺乏大规模、标准化基准以评估跨领域泛化能力的限制。现有数据集在地理范围上较窄、语义不一致或精度不足。我们推出了UnderOneFacade，这是迄今为止最大的跨国和跨洲三维建筑立面基准，包含厘米级精度的点云，具有分层、统一和建筑基础的语义标签，总计27亿个标注点。通过对代表性的基于点、图和变换器架构的系统评估，我们显示当前方法在识别细粒度建筑元素方面存在困难，并且在地理领域之间的表现显著下降，最佳模型在细粒度LoFG3基准上仅达到33的IoU。通过在前所未有的规模上结合几何精度与标准化语义，UnderOneFacade为开发稳健且可转移的三维分割模型建立了严格的基准。数据集、评估脚本和预训练模型将在发表后发布。
cs.CV / 73 / 2607.02024
Spatio-Temporal and Clinical Conditioning for Fine-Grained Radiology Report Retrieval
细粒度放射学报告检索的时空与临床条件化
Abstract
Radiology is vital to modern healthcare, but rising imaging demand and persistent workforce shortages strain reporting capacity and clinical workflows. Automated radiology report generation has the potential to support radiologists and help alleviate this burden; however, existing retrieval-based methods remain rigid, lack explicit anatomical grounding, and do not account for longitudinal disease progression or available clinical context. In this work, we introduce STAR3, a multimodal, spatio-temporal, attentive retrieval framework for radiology report generation that aligns region-level anatomical information with clinical indications and longitudinal changes across chest X-ray studies. Our framework employs an object detector to identify anatomically meaningful regions and retrieves semantically relevant report sentences conditioned on both current clinical context and changes observed between prior and current examinations. This design enables anatomically and temporally grounded report generation that better reflects clinical reporting practice. Experiments on the MIMIC-CXR dataset demonstrate that STAR3 outperforms current retrieval-based approaches on retrieval, NLP and clinical metrics, highlighting the value of conditioning retrieval anatomically, temporally and clinically for advancing automated radiology report generation.
Chinese Translation
放射学在现代医疗保健中至关重要，但不断增长的影像需求和持续的劳动力短缺给报告能力和临床工作流程带来了压力。自动化放射学报告生成有潜力支持放射科医生并帮助减轻这一负担；然而，现有的基于检索的方法仍然较为僵化，缺乏明确的解剖基础，并且未考虑纵向疾病进展或可用的临床背景。在本研究中，我们介绍了STAR3，一个多模态、时空、注意力驱动的放射学报告生成检索框架，该框架将区域级解剖信息与临床指征和胸部X光研究中的纵向变化对齐。我们的框架采用对象检测器来识别具有解剖意义的区域，并根据当前的临床背景和先前与当前检查之间观察到的变化检索语义相关的报告句子。这一设计使得生成的报告在解剖和时间上都更具基础，更好地反映了临床报告实践。在MIMIC-CXR数据集上的实验表明，STAR3在检索、自然语言处理和临床指标方面优于当前的基于检索的方法，突显了在推进自动化放射学报告生成方面，进行解剖、时间和临床条件化检索的价值。
cs.CV / 74 / 2607.02025
Evaluating Vision-Language Models as a Zero-Shot Learning Alternative to You Only Look Once and Optical Character Recognition for Nigerian License Plate Recognition
评估视觉-语言模型作为零-shot学习替代方案，用于尼日利亚车牌识别的You Only Look Once和光学字符识别
Abstract
License Plate Recognition (LPR) systems are critical tools in traffic monitoring, security enforcement, and urban mobility management. Traditional LPR systems often rely on a multi-stage pipeline involving object detection using You Only Look Once (YOLO) and Optical Character Recognition (OCR), which suffer from limitations such as high resource demands, poor performance in unstructured environments, and the need for large annotated datasets. This study explores the potential of Vision-Language Models (VLMs) as a unified, zeroshot learning solution for Nigerian license plate recognition. Using a curated dataset of 88 challenging real-world images collected in Nigeria, we evaluate five selected VLMs: Gemini 2.0 Flash Exp (Google DeepMind), Qwen2.5-VL-7B-Instruct (Alibaba), GPT-4o (OpenAI), Claude 4 Sonnet (Anthropic), and Llama 3.2 Vision 90b (Meta). Results based on Character Error Rate (CER) reveal that Gemini and Qwen significantly outperform other models in both accuracy and robustness, on the challenging image scenarios. This work highlights the practical advantages of VLMs over YOLO+OCR, questions the claims by model providers, and compares the performances of the VLMs.
Chinese Translation
车牌识别（LPR）系统是交通监控、安全执法和城市出行管理的重要工具。传统的LPR系统通常依赖于多阶段流程，包括使用You Only Look Once（YOLO）进行目标检测和光学字符识别（OCR），这些方法存在高资源需求、在非结构化环境中表现不佳以及需要大量标注数据集等局限性。本研究探讨了视觉-语言模型（VLMs）作为尼日利亚车牌识别的统一零-shot学习解决方案的潜力。我们使用在尼日利亚收集的88张具有挑战性的真实世界图像的精心策划数据集，评估了五个选定的VLM：Gemini 2.0 Flash Exp（Google DeepMind）、Qwen2.5-VL-7B-Instruct（Alibaba）、GPT-4o（OpenAI）、Claude 4 Sonnet（Anthropic）和Llama 3.2 Vision 90b（Meta）。基于字符错误率（CER）的结果显示，Gemini和Qwen在挑战性图像场景中的准确性和鲁棒性上显著优于其他模型。本研究强调了VLM相较于YOLO+OCR的实际优势，质疑了模型提供者的声明，并比较了这些VLM的性能。
cs.CV / 75 / 2607.02034
ComplexMimic: Human-Scene Interaction Imitation in Complex 3D Environments
ComplexMimic：复杂三维环境中的人类-场景交互模仿
Abstract
Physics-based Human-Scene Interaction (HSI) imitation learning is crucial for embodied intelligence as it bridges the gap between kinematic 3D motions and real-world dynamics. However, most existing methods focus on simplified scene settings, leaving complex environments largely unexplored, which limits their applicability in real-world scenarios. In this paper, we focus on HSI mimicry in complex environments. Under this complex setting, we observe an inherent trade-off between successfully performing interaction and maintaining natural, physically plausible motions. To address this challenge, we propose ComplexMimic, a framework that reconstructs diverse HSI by interpreting imperfect MoCap data. First, we introduce a Dual Flow Strategy, which learns two complementary experts: an imitation expert for accurate motion tracking and an interaction expert for collision-aware adaptation in complex scenes. Second, naive multi-expert distillation, which treats all experts equally, often under-samples challenging behaviors, limiting effective learning. To mitigate this issue, we propose a difficulty-aware distillation strategy that adaptively weights supervision and prioritizes hard-yet-learnable trajectories guided by failure statistics and learning progress signals. Extensive experiments on three benchmark datasets demonstrate that our approach outperforms current state-of-the-art methods. Our implementation is available at https://github.com/LuPan23/ComplexMimic.
Chinese Translation
基于物理的人类-场景交互（HSI）模仿学习对于具身智能至关重要，因为它弥合了运动学三维动作与现实世界动态之间的差距。然而，现有的大多数方法集中于简化的场景设置，导致复杂环境在很大程度上未被探索，这限制了它们在现实世界场景中的适用性。本文聚焦于复杂环境中的HSI模仿。在这一复杂设置下，我们观察到成功执行交互与保持自然、物理上合理的动作之间存在固有的权衡。为了解决这一挑战，我们提出了ComplexMimic，一个通过解释不完美的运动捕捉（MoCap）数据来重建多样化HSI的框架。首先，我们引入了一种双流策略（Dual Flow Strategy），该策略学习两个互补的专家：一个用于准确动作跟踪的模仿专家和一个用于在复杂场景中进行碰撞感知适应的交互专家。其次，简单的多专家蒸馏往往将所有专家视为平等，这通常会对具有挑战性的行为进行欠采样，从而限制有效学习。为了解决这个问题，我们提出了一种基于难度的蒸馏策略，该策略自适应地加权监督，并优先考虑由失败统计和学习进度信号引导的难以学习的轨迹。对三个基准数据集的广泛实验表明，我们的方法优于当前的最先进方法。我们的实现可在 https://github.com/LuPan23/ComplexMimic 获取。
cs.CV / 76 / 2607.02038
Hierarchical Anti-Aesthetics: Protecting Facial Privacy against Customized Diffusion Models
层次反美学：保护面部隐私免受定制扩散模型的侵害
Abstract
The rise of customized diffusion models has fueled a boom in personalized visual content creation, but it also introduces serious risks of malicious misuse, thereby posing threats to personal privacy. Image aesthetics are strongly correlated with human perception of image quality. Motivated by this observation, we address facial privacy protection from a novel aesthetic perspective by degrading the generation quality of maliciously customized models, thus reducing facial identity leakage. Specifically, we propose a Hierarchical Anti-Aesthetics (HAA) framework that exploits aesthetic cues at multiple perceptual levels. HAA consists of two key branches: (1) Global Anti-Aesthetics, which degrades overall aesthetics and generation quality by constructing a global anti-aesthetic reward mechanism and a corresponding loss; and (2) Local Anti-Aesthetics, which disrupts facial identity by using a local anti-aesthetic reward mechanism and loss to guide adversarial perturbations toward facial regions. By integrating both branches, HAA achieves anti-aesthetic degradation from a global to a local level during customized generation. Extensive experiments show that HAA outperforms existing methods in identity removal, providing an effective tool for protecting facial privacy.
Chinese Translation
定制扩散模型的兴起推动了个性化视觉内容创作的繁荣，但也带来了恶意滥用的严重风险，从而对个人隐私构成威胁。图像美学与人类对图像质量的感知密切相关。基于这一观察，我们从一种新颖的美学视角出发，探讨面部隐私保护，通过降低恶意定制模型的生成质量，从而减少面部身份泄露。具体而言，我们提出了一种层次反美学（Hierarchical Anti-Aesthetics, HAA）框架，该框架利用多个感知层次的美学线索。HAA由两个关键分支组成： (1) 全局反美学（Global Anti-Aesthetics），通过构建全局反美学奖励机制及相应的损失函数来降低整体美学和生成质量； (2) 局部反美学（Local Anti-Aesthetics），通过使用局部反美学奖励机制和损失函数来引导对面部区域的对抗扰动，从而干扰面部身份。通过整合这两个分支，HAA在定制生成过程中实现了从全局到局部的反美学降级。大量实验表明，HAA在身份去除方面优于现有方法，为保护面部隐私提供了有效工具。
cs.CV / 77 / 2607.02045
PWM-ArtGen: Part World Model for Articulated Object Generation
PWM-ArtGen：用于关节物体生成的部件世界模型
Abstract
The key challenge in articulated 3D object generation from a single image is accurately predicting the underlying kinematic structure. Existing methods either infer kinematic parameters directly from a static image that lacks dynamic part-level kinematic relationships, or estimate parameters from visual dynamics generated from a single image, which is prone to accumulated errors of two steps. Moreover, the limited scale and diversity of existing annotated datasets further hinder generalization to complex, real-world objects. To overcome these limitations, we propose to learn the joint distribution of visual dynamics and kinematic parameters. Recognizing that articulated objects can be formulated as dynamic systems, we propose a unified Part World Model called PWM-ArtGen. To leverage unannotated data, this model couples action diffusion and image diffusion with independent diffusion timesteps, which enables visual branch co-training. We further curate a photorealistic dataset of 19.7k part-level image pairs without kinematic annotations, to support co-training. Experiments demonstrate that PWM-ArtGen substantially outperforms existing baselines in the resting state and exhibits strong zero-shot generalization to out-of-distribution objects.
Chinese Translation
从单幅图像生成关节3D物体的关键挑战在于准确预测潜在的运动学结构。现有方法要么直接从缺乏动态部件级运动学关系的静态图像中推断运动学参数，要么从单幅图像生成的视觉动态中估计参数，这容易导致两个步骤的累积误差。此外，现有标注数据集的规模和多样性有限，进一步阻碍了对复杂现实物体的泛化。为克服这些限制，我们提出学习视觉动态和运动学参数的联合分布。认识到关节物体可以被视为动态系统，我们提出了一种统一的部件世界模型，称为PWM-ArtGen。为了利用未标注数据，该模型将动作扩散和图像扩散与独立的扩散时间步耦合，从而实现视觉分支的共同训练。我们进一步整理了一个包含19.7k个无运动学标注的部件级图像对的照片真实数据集，以支持共同训练。实验表明，PWM-ArtGen在静止状态下显著优于现有基线，并对分布外物体表现出强大的零样本泛化能力。
cs.CV / 78 / 2607.02051
Embracing Intra-Class Heterogeneity for Semi-Supervised Medical Image Segmentation: From Diversity to Precision
拥抱类内异质性以实现半监督医学图像分割：从多样性到精确性
Abstract
Due to the scarcity of expert-annotated data, Semi-Supervised Medical Image Segmentation (SSMIS) has emerged as a promising approach. Many anatomical structures in medical images exhibit significant intra-class heterogeneity, with different regions showing heterogeneous intensity patterns within the same structure. However, existing methods inadequately exploit this intensity-manifested intra-class heterogeneity, resulting in uniform structural representations and imprecise segmentation. Furthermore, the scarcity of labeled data makes it more difficult to effectively capture such complex heterogeneity. To address this, we propose Multiple Prototype Contrastive Learning (MPCL), an SSMIS framework that possesses better diversity and better precision. It consists of three novel designs: First, we provide structural representations with better diversity and propose Intensity-aligned Heterogeneous Prototype Generation (IHPG) that effectively models intra-class heterogeneity by generating multiple prototypes aligned with intensity characteristics. Second, we further enhance more diverse structural representations and build a solid foundation for more precise segmentation through Prototypical Space Optimization (PSO) that systematically optimizes a more discriminative and generalizable prototypical space. Finally, we achieve segmentation results with better precision through Dual-branch Knowledge Alignment (DKA) that efficiently promotes intra-class heterogeneity knowledge transfer from prototypical space to the segmentation network. Extensive experiments on three medical image datasets with significant intra-class heterogeneity demonstrate that MPCL significantly outperforms existing methods, especially under extremely limited labeled data.
Chinese Translation
由于专家标注数据的稀缺，半监督医学图像分割（Semi-Supervised Medical Image Segmentation, SSMIS）已成为一种有前景的方法。医学图像中的许多解剖结构表现出显著的类内异质性，同一结构的不同区域显示出异质的强度模式。然而，现有方法未能充分利用这种强度表现的类内异质性，导致结构表示的统一性和分割的不精确性。此外，标注数据的稀缺使得有效捕捉这种复杂的异质性变得更加困难。为此，我们提出了多原型对比学习（Multiple Prototype Contrastive Learning, MPCL），这是一个具有更好多样性和更高精确度的SSMIS框架。它由三个新颖的设计组成：首先，我们提供了具有更好多样性的结构表示，并提出了强度对齐异质原型生成（Intensity-aligned Heterogeneous Prototype Generation, IHPG），通过生成与强度特征对齐的多个原型，有效建模类内异质性。其次，我们进一步增强了更具多样性的结构表示，并通过原型空间优化（Prototypical Space Optimization, PSO）为更精确的分割奠定了坚实的基础，该方法系统地优化了更具区分性和可推广性的原型空间。最后，我们通过双分支知识对齐（Dual-branch Knowledge Alignment, DKA）实现了更高精度的分割结果，该方法有效促进了原型空间到分割网络的类内异质性知识转移。在三个具有显著类内异质性的医学图像数据集上的大量实验表明，MPCL显著优于现有方法，尤其是在标注数据极为有限的情况下。
cs.CV / 79 / 2607.02074
Comprehensive Robustness Analysis of LiDAR-based 3D Object Detection in Autonomous Driving
基于LiDAR的自动驾驶3D目标检测的全面鲁棒性分析
Abstract
Recent advancements in LiDAR-only 3D object detection have demonstrated improved detection accuracy over benchmark datasets. However, the adversarial robustness of these models remains untested. Very few adversarial robustness studies exist for LiDAR-only 3D object detection and unfortunately, even they are limited to legacy models. Moreover, there is a systemic gap in the existing evaluation frameworks that rely simply on mAP ignoring other structural and predictive factors. To fill this gap, we propose a holistic framework that evaluates adversarial robustness using two structural factors (point cloud density and point cloud localization) and three predictive factors (misclassification, localization error, distance from ego). Using this framework, we perform an empirical study and critical analysis on recent and legacy state-of-the-art models using adversarial attacks specifically designed for LiDAR-based models. Our key finding is that high-capacity, voxel-based detectors are more susceptible to structured coordinate perturbations than pillar-based detectors. Additionally, non-anchor-based detectors demonstrate poor adversarial robustness, which necessitates rethinking model training techniques. Overall, our results demonstrate that recent models are as vulnerable to adversarial attacks as their predecessors. Therefore, we argue that there is a need to improve the evaluation benchmarks for 3D object detection that not only reward architectural modifications for improving detection accuracy, but also evaluate whether the design choices improve adversarial robustness.
Chinese Translation
最近，基于LiDAR的3D目标检测在基准数据集上展示了更高的检测精度。然而，这些模型的对抗鲁棒性尚未得到测试。针对基于LiDAR的3D目标检测的对抗鲁棒性研究非常少，遗憾的是，即便存在的研究也仅限于传统模型。此外，现有评估框架存在系统性缺陷，仅依赖于平均精度均值（mAP），忽视了其他结构性和预测性因素。为填补这一空白，我们提出了一个整体框架，通过两个结构性因素（点云密度和点云定位）和三个预测性因素（误分类、定位误差、与自我距离）来评估对抗鲁棒性。利用该框架，我们对最近和传统的最先进模型进行了实证研究和关键分析，使用专门为基于LiDAR的模型设计的对抗攻击。我们的主要发现是，高容量的体素（voxel）检测器比基于柱（pillar）的检测器更容易受到结构性坐标扰动的影响。此外，非锚点（non-anchor）检测器表现出较差的对抗鲁棒性，这需要重新思考模型训练技术。总体而言，我们的结果表明，最近的模型在对抗攻击方面与其前身同样脆弱。因此，我们认为有必要改进3D目标检测的评估基准，不仅奖励改善检测精度的架构修改，还评估设计选择是否提高了对抗鲁棒性。
cs.CV / 80 / 2607.02075
HandsOnWorld: Unconstrained Egocentric Video Generation with Camera-Disentangled Hand Control
HandsOnWorld：无约束自我中心视频生成与相机解耦手部控制
Abstract
We present HandsOnWorld, a framework for hand-controlled egocentric video generation that forgoes multi-view and marker-based motion capture, learning instead from unconstrained monocular video. Such generality is bottlenecked by the scarcity of scalable 3D hand annotations: large egocentric corpora lack finger-level labels, whereas precise hand datasets are confined to narrow, instrumented settings, limiting prior hand-controlled generators to restricted scene distributions. We instead annotate 3D hands directly on in-the-wild egocentric video through monocular reconstruction, introducing a protagonist-centered annotation pipeline that filters the reconstructions at the action-semantic, image-quality, and 3D-geometric levels to build EgoVid-Pro, a dataset of clean, protagonist-only hand trajectories spanning 103K clips and roughly 12M frames across diverse everyday scenes. To resolve the camera-hand entanglement induced by large ego-motion, we further propose the Pl\"{u}cker Hand Map, a 3D-aware control signal that extends Pl\"{u}cker-ray representations from camera rays to the hand surface, disentangling camera and hand motion at the representation level. Experiments show that \method surpasses prior hand-controlled generators in reconstruction fidelity and control accuracy, and generalizes to out-of-distribution everyday scenes beyond the laboratory datasets on which prior methods rely.
Chinese Translation
我们提出了HandsOnWorld，这是一个手部控制的自我中心视频生成框架，省去了多视角和基于标记的动作捕捉，而是从无约束的单目视频中学习。这种通用性受到可扩展的3D手部注释稀缺的限制：大型自我中心语料库缺乏指级标签，而精确的手部数据集则局限于狭窄的仪器设置，限制了先前手部控制生成器的场景分布。我们通过单目重建直接在野外自我中心视频上注释3D手部，提出了一种以主角为中心的注释流程，在动作语义、图像质量和3D几何级别上过滤重建结果，以构建EgoVid-Pro，这是一个干净的、仅包含主角的手部轨迹数据集，涵盖103K个片段和大约1200万帧，涉及多样的日常场景。为了解决由大幅自我运动引起的相机-手部纠缠问题，我们进一步提出了Pl"{u}cker手部映射，这是一种3D感知控制信号，将Pl"{u}cker光线表示从相机光线扩展到手部表面，在表示层面上解耦相机和手部运动。实验表明， extit{HandsOnWorld}在重建保真度和控制精度上超越了先前的手部控制生成器，并且能够推广到超出实验室数据集的日常场景。
cs.CV / 81 / 2607.02083
DeepGaze3.5-VL: Modeling Scanpaths via Autoregressive Token Prediction
DeepGaze3.5-VL：通过自回归标记预测建模扫描路径
Abstract
Understanding human visual attention on a scene over time has applications in domains such as interface design and inferring cognitive states. Modeling visual scanpaths has historically relied on specialized architectures with hand-crafted priors. While these architectures can model fixation sequences, their rigid structural biases restrict easy extendability and flexible conditioning. For instance, integrating task-specific instructions or adapting to distinct viewer identities requires custom, disjoint architectural additions. We frame scanpath prediction purely as a discrete sequence modeling task. By mapping coordinates into a text vocabulary, we leverage the pretrained representations of Vision-Language Models. This framing absorbs diverse factors of variation: simple prompting allows for global conditioning, such as providing viewer identities to capture personalized biases, or task-specific objectives like visual search. The framework can also integrate per-fixation attributes, such as individual fixation durations, alongside spatial locations. The autoregressive alignment enables the scalable, exact computation of per-fixation log-likelihoods, directly equivalent to the commonly used Information Gain (IG) metric. Our model, DeepGaze3.5-VL, establishes a new state-of-the-art across multiple datasets, achieving 2.18 bits of IG on MIT1003, a 46% improvement over DeepGaze III. This advantage persists even when baselines use identical high-capacity vision encoders. Beyond predictive performance, our generative framework serves as a powerful computational tool for direct behavioral interventions, allowing for controlled in-silico simulations that would be experimentally difficult or impossible to conduct in vivo. We demonstrate this ability by performing controlled interventions on the durations of pre-saccadic fixations, recovering known oculomotor phenomena purely from data.
Chinese Translation
理解人类在场景中随时间变化的视觉注意力在界面设计和推断认知状态等领域具有应用价值。建模视觉扫描路径历来依赖于具有手工设计先验的专用架构。虽然这些架构能够建模注视序列，但其刚性的结构偏见限制了易于扩展和灵活条件设置。例如，整合特定任务的指令或适应不同的观看者身份需要定制的、分离的架构添加。我们将扫描路径预测纯粹框架为离散序列建模任务。通过将坐标映射到文本词汇中，我们利用了视觉-语言模型（Vision-Language Models）的预训练表示。这种框架吸收了多种变化因素：简单的提示允许进行全局条件设置，例如提供观看者身份以捕捉个性化偏见，或任务特定目标如视觉搜索。该框架还可以整合每个注视的属性，例如个别注视持续时间，以及空间位置。自回归对齐使得每个注视的对数似然的可扩展、精确计算成为可能，这与常用的信息增益（Information Gain, IG）指标直接等价。我们的模型DeepGaze3.5-VL在多个数据集上建立了新的最先进水平，在MIT1003上实现了2.18比特的IG，相较于DeepGaze III提高了46%。即使基线使用相同的高容量视觉编码器，这一优势依然存在。除了预测性能外，我们的生成框架还作为一个强大的计算工具，用于直接的行为干预，允许进行受控的计算机模拟，这在实验上难以或不可能进行体内实验。我们通过对预扫视注视持续时间进行受控干预，展示了这一能力，从数据中恢复已知的眼动现象。
cs.CV / 82 / 2607.02089
ESC: Emotional Self-Correction for Reliable Vision-Language Models
ESC：用于可靠视觉-语言模型的情感自我修正
Abstract
Vision-language models (VLMs) have achieved strong performance across diverse multimodal tasks, yet they remain vulnerable to unreliable reasoning. Existing self-correction methods mitigate these issues but typically rely on post-training or carefully engineered feedback, incurring high computational cost. In this work, we revisit this challenge through the lens of emotional cues, asking whether they can activate latent self-correction behaviors in VLMs without additional training. \textbf{We find that emotional signals serve as an effective trigger for self-correction, encouraging more cautious and reflective reasoning}. Motivated by this finding, we propose \escabstract (\textbf{\underline{E}}motional \textbf{\underline{S}}elf-\textbf{\underline{C}}orrection), a training-free self-correction framework. ESC introduces an external verifier that detects potentially incorrect initial responses and injects emotional feedback to encourage model to reflect, and produce a better revised response without additional training. Extensive experiments across safety, hallucination, vision-centric perception, and multimodal reasoning benchmarks show that ESC consistently improves reliability while preserving overall model utility. These results suggest that emotion can function not only as an ability to be recognized, but also as a practical control signal for scalable self-correction in VLMs. \textbf{We therefore believe that ESC provides a strong foundation for a new reliable human-like, emotion-integrated research direction.} Our project is publicly available at \textcolor{red}{https://genai4e.github.io/ESC/}.
Chinese Translation
视觉-语言模型（VLMs）在多种多模态任务中表现出色，但仍然容易受到不可靠推理的影响。现有的自我修正方法虽然能够缓解这些问题，但通常依赖于后期训练或精心设计的反馈，导致高计算成本。在本研究中，我们通过情感线索重新审视这一挑战，探讨它们是否能够在不进行额外训练的情况下激活VLMs中的潜在自我修正行为。我们发现，情感信号作为自我修正的有效触发器，能够促进更谨慎和反思的推理。基于这一发现，我们提出了ESC（情感自我修正），一个无训练的自我修正框架。ESC引入了一个外部验证器，检测潜在不正确的初始响应，并注入情感反馈以鼓励模型进行反思，从而在不增加额外训练的情况下生成更好的修正响应。在安全性、幻觉、以视觉为中心的感知和多模态推理基准上的广泛实验表明，ESC在保持整体模型效用的同时，始终提高了可靠性。这些结果表明，情感不仅可以被识别为一种能力，还可以作为可扩展自我修正的实用控制信号。因此，我们相信ESC为一种新的可靠的人类般情感整合研究方向奠定了坚实的基础。我们的项目已公开发布，网址为 https://genai4e.github.io/ESC/。
cs.CV / 83 / 2607.02090
TCG-AR: Real-Time Multi-View Augmented Reality for Trading Card Game Streaming
TCG-AR：用于交易卡牌游戏直播的实时多视角增强现实
Abstract
Trading card games are increasingly played and broadcast online, yet live streams remain mostly limited to flat top-down footage of the playing area. Augmenting such streams with virtual models of the played cards would improve the viewing experience, but most existing systems rely on instrumented playing surfaces and embedded chips, which are costly and impractical for casual players and large-scale events. In this work, we present TCG-AR, a novel real-time pipeline that augments trading card games using ordinary RGB cameras alone, without any physical markers or specialized hardware. Our pipeline detects, orients, and identifies the cards on the board, renders virtual content onto each card across all views, and can additionally compose a broadcaststyle view that summarizes the game state for spectators, streaming the augmented feeds to standard broadcasting software such as OBS. To train the detection, orientation, and identification models without manual labeling, we introduce an automatic procedure that generates annotated synthetic training data from a reference set of card images. Then, we evaluate several trained models on a new manually annotated dataset with real images, analyzing performance and runtime throughput that determine real-world usability. Overall, by relying only on commodity cameras and hardware, and by open-sourcing all code, models, and datasets, this work aims to serve as a reference for real-time trading card recognition and to make real-time augmented-reality streaming accessible to the broader community of players and streamers.
Chinese Translation
交易卡牌游戏越来越多地在线进行和直播，但直播内容大多仅限于平面俯视的游戏区域画面。通过虚拟模型增强这些直播内容可以改善观众的观看体验，但大多数现有系统依赖于带有仪器的游戏表面和嵌入式芯片，这对于休闲玩家和大规模活动来说成本高昂且不切实际。在本研究中，我们提出了TCG-AR，一种新颖的实时处理管道，仅使用普通RGB摄像头，无需任何物理标记或专用硬件来增强交易卡牌游戏。我们的管道能够检测、定位和识别棋盘上的卡牌，在所有视角上将虚拟内容渲染到每张卡牌上，并且可以额外生成一种广播风格的视图，以总结游戏状态供观众观看，同时将增强后的画面流式传输到标准广播软件（如OBS）。为了在没有手动标注的情况下训练检测、定位和识别模型，我们引入了一种自动程序，从一组参考卡牌图像生成带注释的合成训练数据。随后，我们在一个新的手动标注的真实图像数据集上评估了几种训练模型，分析了决定现实世界可用性的性能和运行时吞吐量。总体而言，通过仅依赖普通摄像头和硬件，并开源所有代码、模型和数据集，本研究旨在作为实时交易卡牌识别的参考，并使实时增强现实直播对更广泛的玩家和主播社区变得可及。
cs.CV / 84 / 2607.02091
Multimodal Fusion for Fine-Grained Classification of Breast Fibroadenoma and Phyllodes Tumors
乳腺纤维腺瘤与叶状肿瘤的细粒度分类的多模态融合
Abstract
Breast fibroadenoma (FA) and phyllodes tumor (PT) are fibroepithelial breast lesions with highly overlapping appearances on B-mode ultrasound, making benign and borderline PT prone to being misclassified as FA and complicating preoperative decision-making. Existing computer-aided diagnosis methods commonly rely on single-modal imaging features and insufficiently exploit complementary clinical and textual information. To address this limitation, we construct the FAPT-M Dataset, a pathology-confirmed multimodal dataset comprising 910 patients with strictly reviewed ultrasound images, structured clinical attributes, and ultrasound diagnostic descriptions. Based on this dataset, we propose a clinically guided multimodal framework that integrates DenseNet-based visual encoding, CLIP-inspired text encoding, and lightweight clinical encoding, and further introduces clinical-conditioned adaptive modulation, cross-modal Transformer fusion, and dual-path representation learning to improve feature alignment and multimodal interaction. Under patient-level five-fold cross-validation, the proposed method achieves an accuracy of 77.64%, F1-score of 73.38%, and AUC of 89.74%, outperforming representative CNN-, Transformer-, and vision-language-based baselines. Ablation studies and class-balanced evaluations further confirm the contribution of three-modality fusion and the key architectural components. Overall, this work provides an effective multimodal approach for fine-grained FA-PT classification and establishes a high-quality benchmark for multimodal breast ultrasound analysis.
Chinese Translation
乳腺纤维腺瘤（FA）和叶状肿瘤（PT）是具有高度重叠外观的纤维上皮乳腺病变，在B模式超声下容易导致良性和边缘性PT被误分类为FA，从而复杂化术前决策。现有的计算机辅助诊断方法通常依赖于单模态成像特征，未能充分利用互补的临床和文本信息。为了解决这一局限性，我们构建了FAPT-M数据集，这是一个经过病理确认的多模态数据集，包含910名患者的严格审核的超声图像、结构化临床属性和超声诊断描述。基于该数据集，我们提出了一种临床指导的多模态框架，集成了基于DenseNet的视觉编码、受CLIP启发的文本编码和轻量级临床编码，并进一步引入了临床条件自适应调制、跨模态Transformer融合和双路径表示学习，以改善特征对齐和多模态交互。在患者级五折交叉验证下，所提方法实现了77.64%的准确率、73.38%的F1分数和89.74%的AUC，优于代表性的CNN、Transformer和视觉-语言基础模型。消融研究和类别平衡评估进一步确认了三模态融合及关键架构组件的贡献。总体而言，本研究提供了一种有效的多模态方法用于细粒度FA-PT分类，并为多模态乳腺超声分析建立了高质量的基准。
cs.CV / 85 / 2607.02096
LongEgoRefer: A Benchmark for Long-Form Egocentric Video Referring Expression Comprehension
LongEgoRefer：长格式自我中心视频指称表达理解的基准
Abstract
Egocentric videos capture rich and diverse human-object interactions and have emerged as a fundamental resource for understanding human activities related to objects. In this context, Video Referring Expression Comprehension (Video REC), the task of localizing the temporal and spatial extent of a referred object in video frames given a natural language query, plays a key role in linking textual descriptions to observed objects in untrimmed egocentric recordings. However, existing egocentric Video REC benchmarks primarily focus on short video clips, where some target object appears densely within frames. Such settings do not reflect real-world egocentric recordings, which are long-form, untrimmed, and characterized by sparse object occurrences and complex activity transitions. To address this limitation, we introduce LongEgoRefer, a novel and challenging benchmark constructed from long-form videos in the Ego4D dataset. LongEgoRefer contains 1,498 referring expressions with an average video duration of 45 minutes. The benchmark exhibits extreme target sparsity, detailed linguistic descriptions, and complex human-object interactions embedded in long, dynamic egocentric narratives. Consequently, it defines a demanding spatio-temporal grounding problem that requires models to identify both when an event occurs and where the referred object appears within extended video sequences. We evaluate existing Video REC approaches, including training-free baselines based on vision-language models combined with Grounded SAM2. Extensive experiments show that even advanced baselines and current state-of-the-art models struggle significantly on LongEgoRefer. These results highlight the intrinsic difficulty of long-form egocentric spatio-temporal grounding and emphasize the need for more robust video understanding models.
Chinese Translation
自我中心视频捕捉了丰富多样的人与物体之间的互动，已成为理解与物体相关的人类活动的基本资源。在这个背景下，视频指称表达理解（Video REC）是指在给定自然语言查询的情况下，定位视频帧中被指称物体的时间和空间范围的任务，它在将文本描述与未剪辑的自我中心录音中观察到的物体联系起来方面发挥着关键作用。然而，现有的自我中心视频 REC 基准主要集中在短视频片段上，其中某些目标物体在帧中密集出现。这种设置并未反映现实世界中的自我中心录音，这些录音通常是长格式的、未剪辑的，并且以稀疏的物体出现和复杂的活动过渡为特征。为了解决这一局限性，我们引入了 LongEgoRefer，这是一个基于 Ego4D 数据集中长格式视频构建的新颖且具有挑战性的基准。LongEgoRefer 包含 1,498 个指称表达，平均视频时长为 45 分钟。该基准展现了极端的目标稀疏性、详细的语言描述以及嵌入在长篇动态自我中心叙事中的复杂人-物体互动。因此，它定义了一个要求模型识别事件发生时间和被指称物体在扩展视频序列中出现位置的严格时空定位问题。我们评估了现有的视频 REC 方法，包括基于视觉-语言模型与 Grounded SAM2 结合的无训练基线。广泛的实验表明，即使是先进的基线和当前的最先进模型在 LongEgoRefer 上也面临显著挑战。这些结果突显了长格式自我中心时空定位的内在困难，并强调了对更强大的视频理解模型的需求。
cs.CV / 86 / 2607.02097
WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution
WBMM：用于高效大感受野卷积的窗口化批量矩阵乘法
Abstract
Large kernel depthwise convolutions achieve strong performance but suffer from significant degradation as kernel size grows due to irregular memory access from gather-based computation; while Large Kernel Acceleration (LKA) helps on small feature maps, it becomes counterproductive on large feature maps, even slower than non-accelerated implementations. We propose Windowed Batch Matrix Multiplication (WBMM), which partitions input into contiguous windows and indexes a compact relative position bias table to construct weight matrices, enabling regular memory access via batched matrix multiplication. This yields a unique property: WBMM's throughput improves with larger windows, opposite to depthwise convolutions that degrade with larger kernels. Operator-level benchmarks show WBMM with 14x14 windows outperforms 5x5 depthwise convolution baselines in speed while providing a 7.8x larger per-layer receptive field. Combined with inter-block cross-window communication and hierarchical window reparameterization, WBMM achieves comparable or higher accuracy on ImageNet-1K, COCO, and ADE20K with 1.31-1.88x training speedup, and demonstrates consistent advantages across GPU, CPU, and edge devices without requiring specialized acceleration kernels. Our code is available at http://github.com/wansong-s/WBMM
Chinese Translation
大核深度卷积在性能上表现出色，但随着核大小的增加，由于基于聚集计算的不规则内存访问，性能显著下降；虽然大核加速（Large Kernel Acceleration, LKA）在小特征图上有所帮助，但在大特征图上却适得其反，甚至比未加速的实现还要慢。我们提出了窗口化批量矩阵乘法（Windowed Batch Matrix Multiplication, WBMM），该方法将输入划分为连续的窗口，并索引一个紧凑的相对位置偏置表来构建权重矩阵，从而通过批量矩阵乘法实现规则的内存访问。这带来了一个独特的特性：WBMM的吞吐量随着窗口的增大而提高，这与深度卷积在核增大时性能下降的情况正好相反。操作级基准测试表明，使用14x14窗口的WBMM在速度上优于5x5深度卷积基线，同时提供了7.8倍更大的每层感受野。结合块间跨窗口通信和分层窗口重参数化，WBMM在ImageNet-1K、COCO和ADE20K上实现了可比或更高的准确率，并获得了1.31-1.88倍的训练加速，同时在GPU、CPU和边缘设备上均表现出一致的优势，无需专门的加速内核。我们的代码可在http://github.com/wansong-s/WBMM获取。
cs.CV / 87 / 2607.02099
X-Splat: Gaussian Splatting for 3D CBCT Generation from Single Panoramic Radiograph
X-Splat：基于高斯点云的单幅全景放射影像生成3D锥形束计算机断层扫描
Abstract
Generating a 3D dental volume from a single panoramic radiograph (PXR) could provide a low-radiation alternative to Cone-Beam Computed Tomography (CBCT), but the problem is highly underdetermined: panoramic acquisition integrates 3D attenuation along curved X-ray paths into a 2D image, leaving depth-resolved anatomy unobserved. Existing implicit and generative approaches often produce oversmoothed geometry or anatomically inconsistent hallucinations, lacking geometry-driven supervision and relying on smooth representations unable to precisely localize sharp anatomical boundaries. We propose X-Splat, the first Gaussian Splatting framework for generating CBCT-like 3D dental volumes from a single PXR. X-Splat uses the known panoramic acquisition geometry as a generation scaffold: learnable anisotropic Gaussian primitives are initialized along the X-ray paths that formed the input image and adjusted in a single feed-forward pass, constrained by Beer-Lambert reprojection and multi-view radiographic training supervision. A lightweight residual refiner adds dataset-level anatomical priors without overriding the geometry already resolved by the Gaussians. We train on synthetic PXR-CBCT pairs, enabling direct volumetric supervision without paired real scans. We further introduce segmentation-based geometry-aware metrics, providing the first evaluation of PXR-based generation over maxillofacial anatomy. X-Splat outperforms NeRF- and GAN-based baselines, recovering individual teeth, cortical boundaries, and alveolar structure, including the mandibular canal which prior methods fail to reconstruct. Code will be available at https://github.com/tomek1911/X-Splat
Chinese Translation
从单幅全景放射影像（PXR）生成3D牙科体积可以提供一种低辐射的锥形束计算机断层扫描（CBCT）替代方案，但该问题高度欠定：全景采集将沿曲线X射线路径的3D衰减整合为2D图像，导致深度分辨的解剖结构未被观察到。现有的隐式和生成方法往往产生过于平滑的几何形状或解剖上不一致的幻觉，缺乏几何驱动的监督，并依赖于无法精确定位尖锐解剖边界的平滑表示。我们提出了X-Splat，这是第一个用于从单幅PXR生成类似CBCT的3D牙科体积的高斯点云框架。X-Splat利用已知的全景采集几何作为生成支架：可学习的各向异性高斯原语沿着形成输入图像的X射线路径初始化，并在单次前馈传递中进行调整，受限于Beer-Lambert重投影和多视图放射影像训练监督。一个轻量级的残差精炼器在不覆盖高斯已解析几何的情况下，添加数据集级别的解剖先验。我们在合成的PXR-CBCT对上进行训练，使得在没有配对真实扫描的情况下实现直接的体积监督。我们进一步引入基于分割的几何感知度量，首次评估基于PXR的生成在颌面解剖上的表现。X-Splat在恢复个别牙齿、皮质边界和牙槽结构方面优于基于NeRF和GAN的基线，包括下颌管，这是先前方法未能重建的结构。代码将可在 https://github.com/tomek1911/X-Splat 获取。
cs.CV / 88 / 2607.02131
AbsoluteDegradation: A Physics-Inspired Synthetic Film-Degradation Pipeline and Archival Film Restoration Benchmark
绝对降解：一种基于物理的合成胶卷降解流程与档案胶卷修复基准
Abstract
Restoring archival film remains a fundamentally challenging problem due to the absence of paired training data and the lack of standardized evaluation benchmarks. Pristine versions of deteriorated footage are physically unrecoverable, requiring supervised methods to rely on synthetic data that often fail to capture the complex, temporally coherent nature of real film degradation. At the same time, existing real-world datasets are limited in scale, quality, and accessibility, hindering reliable evaluation and fair comparison across methods. We address both limitations with AbsoluteDegradation, a physics-inspired, modular pipeline for synthesizing realistic film degradations, and a new large-scale archival benchmark. The proposed pipeline models the analog-to-digital process as a structured composition of artifact families, incorporating signal-dependent grain, parametric scratches, and temporally coherent camera motion, enabling controlled generation of diverse degradation regimes. In parallel, we introduce a curated dataset of 81,576 high-resolution frames sourced from real archival footage, designed for consistent evaluation under real-world conditions. Together, these contributions provide a unified framework for training and benchmarking restoration models. Extensive experiments across multiple architectures show that models trained with AbsoluteDegradation generalize better to real-world footage, while the proposed benchmark reveals systematic failure modes of current methods. We hope this work establishes a foundation for reproducible and domain-authentic evaluation in archival film restoration.
Chinese Translation
修复档案胶卷仍然是一个根本性的挑战，因为缺乏配对的训练数据和标准化的评估基准。受损影像的原始版本在物理上不可恢复，这要求监督方法依赖于合成数据，而这些数据往往无法捕捉真实胶卷降解的复杂性和时间一致性。同时，现有的真实世界数据集在规模、质量和可获取性方面都有限，阻碍了可靠的评估和方法之间的公平比较。我们通过绝对降解（AbsoluteDegradation）来解决这两个限制，这是一个基于物理的模块化流程，用于合成真实的胶卷降解，并提供一个新的大规模档案基准。所提出的流程将模拟到数字的过程建模为一个结构化的伪影家族组合，结合了信号依赖的颗粒、参数化划痕和时间一致的相机运动，从而实现多样化降解模式的可控生成。同时，我们引入了一个精心策划的数据集，包含来自真实档案影像的81,576帧高分辨率图像，旨在在真实世界条件下进行一致的评估。这些贡献共同提供了一个统一的框架，用于训练和基准测试修复模型。针对多种架构的广泛实验表明，使用绝对降解训练的模型在真实影像上具有更好的泛化能力，而所提出的基准揭示了当前方法的系统性失败模式。我们希望这项工作为档案胶卷修复中的可重复和领域真实的评估奠定基础。
cs.CV / 89 / 2607.02139
AdaCount: Training-Free Similarity-Guided Spatial and Feature Adaptation for Zero-Shot Object Counting
AdaCount：无训练的相似性引导空间与特征适应用于零-shot物体计数
Abstract
Zero-shot object counting (ZOC) aims to count instances of arbitrary object categories specified only through textual prompts. Recent training-free approaches leverage foundation models such as SAM to reformulate counting as a prompt-driven segmentation task, eliminating the need for costly counting-specific training data with point-level annotations. More recently, SAM3 introduced promptable concept segmentation, enabling the zero-shot segmentation of all instances corresponding to a text-defined concept. However, SAM3 struggles in densely populated scenes containing numerous small objects, where limited image resolution and insufficient attention to target-relevant regions often lead to missed instances and poor instance separation, hindering accurate object counting. To address this limitation, we propose AdaCount, a training-free framework for ZOC based on similarity-guided spatial and feature adaptation. AdaCount first estimates a prototype-driven similarity map that identifies target-relevant regions. This similarity map subsequently guides two complementary adaptations: (i) similarity-guided spatial warping, which reallocates image resolution toward target instances, and (ii) feature modulation, which amplifies target-relevant encoder representations. Together, these adaptations enable SAM3 to devote greater representational capacity to target-relevant regions while preserving global image context, without requiring any model retraining. Extensive experiments across six diverse counting benchmarks establish AdaCount as a new SOTA among training-free ZOC approaches.
Chinese Translation
零-shot物体计数（ZOC）旨在仅通过文本提示来计数任意物体类别的实例。最近的无训练方法利用基础模型如SAM将计数重新表述为一个基于提示的分割任务，从而消除了对昂贵的特定计数训练数据（带有点级注释）的需求。更近期的SAM3引入了可提示的概念分割，使得能够对所有与文本定义概念相对应的实例进行零-shot分割。然而，SAM3在包含众多小物体的密集场景中表现不佳，有限的图像分辨率和对目标相关区域的关注不足常常导致漏计实例和实例分离不良，从而妨碍准确的物体计数。为了解决这一局限性，我们提出了AdaCount，一个基于相似性引导的空间与特征适应的无训练框架，用于ZOC。AdaCount首先估计一个原型驱动的相似性图，识别目标相关区域。该相似性图随后引导两种互补的适应： (i) 相似性引导的空间变形，重新分配图像分辨率以聚焦于目标实例；(ii) 特征调制，增强目标相关的编码器表示。这些适应共同使得SAM3能够将更多的表征能力集中于目标相关区域，同时保持全局图像上下文，而无需任何模型重训练。通过在六个不同的计数基准上进行广泛实验，AdaCount确立了其作为无训练ZOC方法中的新状态最优（SOTA）。
cs.CV / 90 / 2607.02148
SAMoR: Motion Modelling for Articulated Objects of Any Skeleton and Topology
SAMoR：任意骨架和拓扑的关节物体运动建模
Abstract
Modeling motion for articulated objects of arbitrary skeleton topology remains difficult: existing motion generators target a fixed human skeleton, and prior adaptations either fail to share a vocabulary across rigs or discard motion detail through global pooling. Our key observation is that while joint-level motion does not correspond cleanly across species, motion of functional joint groups does: a human arm, a wolf foreleg, and a bird wing share motion structure despite differing joint counts and connectivity, a correspondence that joint names (e.g., "forearm", "wing_L1") partially expose even when topology does not. We introduce SAMoR (Skeleton-Aware Motion Representation for Articulated Objects), a cross-topology motion representation that encodes each motion segment as a small fixed number ($K=8$) of part tokens shared across arbitrary skeletons. A graph-transformer encoder consumes per-joint motion features, kinematic graph structure, and joint-name embeddings, then compresses them into part-level tokens via cross-attention pooling and residual vector quantization, yielding a discrete motion codebook shared across rigs. To keep the part queries from collapsing into redundant global representations, we introduce a topology-agnostic attention supervision loss, with joint-name dropout to reduce over-reliance on text labels. We curate a heterogeneous corpus from HumanML3D, Truebones Zoo, and animated Objaverse-XL assets, and evaluate SAMoR on held-out characters with unseen skeletons. It supports accurate reconstruction and cross-topology transfer, and enables text-conditioned generation and part-wise editing via a MaskGIT token generator. SAMoR reaches $2.75 \times 10^{-2}$ normalized MPJPE on cross-topology reconstruction, $5.8\times$ below the strongest adapted variable-$J$ tokenizer baseline, while remaining competitive with fixed-skeleton specialists on HumanML3D.
Chinese Translation
对任意骨架拓扑的关节物体进行运动建模仍然是一个难题：现有的运动生成器针对固定的人类骨架，而之前的适配要么未能在不同的模型之间共享词汇，要么通过全局池化丢失了运动细节。我们的关键观察是，尽管关节级别的运动在不同物种之间并不完全对应，但功能关节组的运动却是相似的：人类的手臂、狼的前腿和鸟的翅膀在运动结构上是共享的，尽管它们的关节数量和连接方式不同，这种对应关系即使在拓扑不一致时，关节名称（例如，“前臂”、“翼_L1”）也能部分揭示。我们提出了SAMoR（Skeleton-Aware Motion Representation for Articulated Objects），一种跨拓扑的运动表示方法，将每个运动片段编码为固定数量（$K=8$）的部分标记，这些标记在任意骨架之间是共享的。图形变换器编码器消耗每个关节的运动特征、运动学图结构和关节名称嵌入，然后通过交叉注意力池化和残差向量量化将其压缩为部分级别的标记，从而生成一个跨模型共享的离散运动代码本。为了防止部分查询崩溃为冗余的全局表示，我们引入了一种拓扑无关的注意力监督损失，并采用关节名称丢弃以减少对文本标签的过度依赖。我们从HumanML3D、Truebones Zoo和动画Objaverse-XL资产中整理了一个异构语料库，并在未见骨架的保留角色上评估了SAMoR。它支持准确的重建和跨拓扑转移，并通过MaskGIT标记生成器实现文本条件生成和部分编辑。SAMoR在跨拓扑重建中达到了$2.75 imes 10^{-2}$的标准化MPJPE，比最强的适配可变-$J$标记器基线低$5.8 imes$，同时在HumanML3D上与固定骨架专家保持竞争力。
cs.CV / 91 / 2607.02156
Patient-Specific Articulated Digital Twins from a Single Full-Body CT Scan
基于单次全身CT扫描的患者特异性关节数字双胞胎
Abstract
Patient-specific anatomical models provide individualized context for surgical planning, image-guided intervention, and algorithm development. However, most CT-derived models are static: they preserve the body configuration captured at scan time, but cannot represent how the same anatomy would appear after patient repositioning. This limitation is especially important for radiographic imaging, where appearance depends jointly on imaging geometry and patient pose. We present a proof-of-concept for constructing a patient-specific articulated digital twin from a single full-body CT scan. The method fits a parametric human body model (SMPL) to obtain a patient-aligned kinematic scaffold, binds segmented bones and organs to an anatomy-aware rig, and retargets body-pose changes while preserving skeletal geometry. On three full-body CT subjects, the fitted scaffold achieved 15.8 $\pm$ 4.0 mm chamfer distance and 95.9 $\pm$ 1.8% skeletal enclosure. Recomposition at the acquisition pose preserved major radiographic structure, with overall SSIM of 0.872 $\pm$ 0.016 and PSNR of 18.5 $\pm$ 1.4 dB across paired DRRs. Across unseen target poses, the resulting twins enabled articulation while maintaining high skeletal enclosure (94.4 $\pm$ 0.4%). As a feasibility demonstration, we render the articulated twin as pose-dependent DRRs. These results suggest the feasibility of extending static, view-controllable CT simulation toward pose-controllable anatomical twins for future synthetic imaging and positioning studies.
Chinese Translation
患者特异性解剖模型为外科规划、图像引导干预和算法开发提供了个性化的背景。然而，大多数CT衍生模型是静态的：它们保留了扫描时捕获的身体配置，但无法表示在患者重新定位后相同解剖结构的外观。这一限制在放射成像中尤为重要，因为外观同时依赖于成像几何和患者姿态。我们展示了一种从单次全身CT扫描构建患者特异性关节数字双胞胎的概念验证方法。该方法拟合一个参数化的人体模型（SMPL），以获得与患者对齐的运动学支架，将分割的骨骼和器官绑定到一个解剖感知的装置上，并在保持骨骼几何形状的同时重新定向身体姿态变化。在三名全身CT受试者中，拟合的支架达到了15.8 ± 4.0 mm的切边距离和95.9 ± 1.8%的骨骼包围率。在采集姿态下的重组保留了主要的放射结构，配对DRR的整体结构相似性指数（SSIM）为0.872 ± 0.016，峰值信噪比（PSNR）为18.5 ± 1.4 dB。在未见目标姿态下，生成的双胞胎能够进行关节运动，同时保持高骨骼包围率（94.4 ± 0.4%）。作为可行性演示，我们将关节双胞胎渲染为姿态依赖的DRR。这些结果表明，将静态、视图可控的CT模拟扩展为姿态可控的解剖双胞胎在未来的合成成像和定位研究中是可行的。
cs.CV / 92 / 2607.02158
Efficient PEFT Methods with Adaptive Checkpointing for Vision Models and VLMs on Resource Constrained Consumer-GPUs
针对资源受限消费者GPU的视觉模型和视觉语言模型的自适应检查点高效PEFT方法
Abstract
Modern pretrained vision models achieve strong accuracy but demand substantial GPU memory for fine-tuning, making edge deployment impractical. This paper compares five parameter-efficient fine-tuning (PEFT) methods (Full FT, LoRA, AdaLoRA, QLoRA, BitFit) on Transformers- (ViT-Small, TinyViT) and Mamba-based vision backbones (Vim-Small, MambaVision-T) under an on-device VRAM budget (e.g., 2 GB), together with three gradient-checkpointing strategies (none, static, and a proposed memory-budget-aware adaptive algorithm); and we evaluate three families of foundation-model baselines: zero-shot contrastive vision language models (OpenCLIP, SigLIP), self-supervised vision backbones with lightweight evaluation protocols (DINOv2), and autoregressive VLMs for prompt-based classification (PaliGemma, MobileVLM, SmolVLM). Experiments on CIFAR-100 and DTD report accuracy, training time, energy, and the NetScore family of multi-objective metrics, which we extend with two deployment-aware variants. QLoRA and BitFit cut energy 20-30% at a 1-2% accuracy cost; the adaptive algorithm reduces peak memory 43-79% with 9-30% energy overhead. DINOv2 surpasses fine-tuned models on CIFAR-100 (0.917 vs. 0.897) at a fraction of the energy, while small autoregressive VLMs remain uncompetitive.
Chinese Translation
现代预训练视觉模型在准确性方面表现出色，但在微调时需要大量GPU内存，这使得边缘部署变得不切实际。本文比较了五种参数高效微调（PEFT）方法（全微调、LoRA、AdaLoRA、QLoRA、BitFit）在Transformer（ViT-Small、TinyViT）和基于Mamba的视觉骨干网络（Vim-Small、MambaVision-T）下的表现，采用了设备内存（例如，2 GB）预算，并结合了三种梯度检查点策略（无、静态以及一种提出的基于内存预算的自适应算法）；同时，我们评估了三类基础模型基线：零-shot对比视觉语言模型（OpenCLIP、SigLIP）、具有轻量级评估协议的自监督视觉骨干（DINOv2），以及用于基于提示分类的自回归视觉语言模型（PaliGemma、MobileVLM、SmolVLM）。在CIFAR-100和DTD上的实验报告了准确性、训练时间、能耗以及NetScore多目标指标家族，我们还扩展了两个部署感知的变体。QLoRA和BitFit在1-2%的准确性损失下减少了20-30%的能耗；自适应算法在9-30%的能耗开销下减少了43-79%的峰值内存。DINOv2在CIFAR-100上以较少的能耗超越了微调模型（0.917对0.897），而小型自回归视觉语言模型则仍然缺乏竞争力。
cs.CV / 93 / 2607.02185
RadiomicNet: A Hybrid Radiomics-Guided Lightweight Architecture for Interpretable Medical Image Segmentation
RadiomicNet：一种混合放射组学引导的轻量级架构，用于可解释的医学图像分割
Abstract
Deep learning has achieved remarkable performance in medical image segmentation, yet it suffers from critical limitations: mathematical intractability, substantial parameter requirements, and lack of clinical interpretability. We propose RadiomicNet, a novel two-stream hybrid architecture that enhances standard deep learning by integrating handcrafted radiomics features directly into the segmentation learning process. The key contribution is the Radiomics Attention Gate (RAG), which leverages Gray-Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP) features to modulate skip-connection attention in a lightweight MobileNetV2-based encoder-decoder, providing ante-hoc interpretability without post-hoc approximations. A novel Radiomics Consistency Loss further enforces alignment between texture complexity and prediction uncertainty, reducing Expected Calibration Error (ECE) from 0.142 to 0.118. RadiomicNet achieves a Dice Similarity Coefficient (DSC) of 0.763 +/- 0.231 on the Breast Ultrasound Images (BUSI) dataset and 0.854 +/- 0.112 on Kvasir-SEG, outperforming U-KAN by 1.2% and 1.8%, respectively (p < 0.05, Wilcoxon signed-rank test), with only 3.27M parameters, 9.5x fewer than standard U-Net and 4.3x fewer than U-KAN. Gradient-based feature importance analysis reveals that GLCM dissimilarity (15.24%), GLCM energy (14.56%), and LBP entropy (11.49%) are the dominant radiomics cues, providing clinically meaningful explanations for segmentation decisions. The proposed approach demonstrates that compact, interpretable models grounded in domain knowledge can deliver state-of-the-art segmentation performance with substantially reduced computational overhead.
Chinese Translation
深度学习在医学图像分割中取得了显著的性能，但仍存在一些关键限制：数学不可处理性、大量参数需求以及缺乏临床可解释性。我们提出了RadiomicNet，一种新颖的双流混合架构，通过将手工设计的放射组学特征直接整合到分割学习过程中，增强了标准深度学习。其关键贡献是放射组学注意力门（Radiomics Attention Gate, RAG），该门利用灰度共生矩阵（Gray-Level Co-occurrence Matrix, GLCM）和局部二值模式（Local Binary Pattern, LBP）特征来调节轻量级基于MobileNetV2的编码器-解码器中的跳跃连接注意力，从而提供先验可解释性，而无需后验近似。新颖的放射组学一致性损失进一步强化了纹理复杂性与预测不确定性之间的对齐，将期望校准误差（Expected Calibration Error, ECE）从0.142降低到0.118。RadiomicNet在乳腺超声图像（Breast Ultrasound Images, BUSI）数据集上实现了0.763 +/- 0.231的Dice相似系数（Dice Similarity Coefficient, DSC），在Kvasir-SEG数据集上实现了0.854 +/- 0.112，分别比U-KAN提高了1.2%和1.8%（p < 0.05，Wilcoxon符号秩检验），并且仅使用了3.27M参数，较标准U-Net减少了9.5倍，较U-KAN减少了4.3倍。基于梯度的特征重要性分析表明，GLCM不相似性（15.24%）、GLCM能量（14.56%）和LBP熵（11.49%）是主要的放射组学线索，为分割决策提供了临床意义的解释。所提出的方法表明，基于领域知识的紧凑且可解释的模型能够在显著降低计算开销的同时，提供最先进的分割性能。
cs.CV / 94 / 2607.02209
MedSaab-US: A Backpropagation-Free Multi-Scale Wavelet-Saab Framework for Thyroid Nodule Segmentation in Ultrasound Images
MedSaab-US：一种无反向传播的多尺度小波-Saab框架，用于超声图像中的甲状腺结节分割
Abstract
Deep learning (DL) methods dominate thyroid nodule segmentation in ultrasound (US) images, achieving high Dice scores but at the cost of millions of parameters, GPU-dependent training via backpropagation, and limited mathematical tractability. These limitations impede deployment in resource-constrained environments. In this paper, we propose MedSaab-US, a backpropagation-free segmentation framework grounded in the Green Learning paradigm. MedSaab-US extracts multi-scale spatial-frequency features by combining multi-level Discrete Wavelet Transform (DWT) with multi-scale channel-wise Saab (Subspace Approximation with Adjusted Bias) transforms at patch sizes of 5 x 5, 11 x 11, and 21 x 21 pixels. Label-Assisted Greedy (LAG) feature selection retains the most discriminative features, which are fed to an XGBoost classifier for pixel-wise prediction. The Saab transform parameters are determined analytically from data statistics, while XGBoost employs iterative greedy tree construction without requiring backpropagation. Evaluated on the TN3K dataset (2,879 training and 614 test images), MedSaab-US achieves a mean Dice coefficient of 0.4784 +/- 0.2190, precision of 0.5768, and recall of 0.5604, with a model footprint under 500K parameters and CPU-only inference in approximately 0.3 seconds per image. We present this result as an exploratory non-DL baseline for thyroid ultrasound segmentation and analyze the specific challenges posed by isoechoic nodules. An ablation study further quantifies the contribution of each pipeline component, including separate evaluations of LAG feature selection and training-set size.
Chinese Translation
深度学习（DL）方法在超声（US）图像中的甲状腺结节分割中占主导地位，尽管取得了高达的Dice系数，但代价是需要数百万个参数、依赖GPU的反向传播训练以及有限的数学可处理性。这些限制阻碍了在资源受限环境中的部署。本文提出了MedSaab-US，一种基于Green Learning范式的无反向传播分割框架。MedSaab-US通过将多级离散小波变换（DWT）与多尺度通道Saab（带调整偏差的子空间近似）变换相结合，在5 x 5、11 x 11和21 x 21像素的补丁大小下提取多尺度空间频率特征。标签辅助贪婪（LAG）特征选择保留了最具区分性的特征，这些特征被输入到XGBoost分类器中进行逐像素预测。Saab变换参数通过数据统计进行解析确定，而XGBoost则采用迭代贪婪树构建，无需反向传播。在TN3K数据集（2,879张训练图像和614张测试图像）上评估，MedSaab-US实现了平均Dice系数为0.4784 +/- 0.2190，精确度为0.5768，召回率为0.5604，模型参数量低于500K，CPU推理时间约为每张图像0.3秒。我们将这一结果作为甲状腺超声分割的探索性非DL基线，并分析了等回声结节带来的具体挑战。消融研究进一步量化了每个管道组件的贡献，包括对LAG特征选择和训练集大小的单独评估。
cs.CV / 95 / 2607.02220
DetailAnywhere: Fashion Detail Generation via Cross-Modal Feature Alignment Distillation
DetailAnywhere：通过跨模态特征对齐蒸馏生成时尚细节
Abstract
Diffusion-based generative AI has achieved remarkable success in e-commerce applications such as virtual try-on, poster generation, and product background synthesis. However, when making online purchasing decisions for apparel, consumers also desire the freedom to examine specific detail regions of interest, such as collars, cuffs, and fabric textures, yet existing methods have not explicitly studied this setting. We therefore formalize a new, non-template task: Fashion Detail Generation with focus conditioning, and release FDBench, the first benchmark comprising 40K+ human-verified reference-detail pairs across 41 different categories. This task poses a unique semantic gap challenge: the model must bridge the correspondence between a focus marker on a product reference image and a photorealistic close-up view of the indicated region, while faithfully preserving the garment's identity, without any precise prompt. To bridge this gap, we propose Cross-modal Feature Alignment Distillation (CFAD), which leverages a fine-tuned DINOv3 teacher to align both branches of a Multimodal Diffusion Transformer in a shared semantic space via dual-branch distillation. To further improve consistency between generated details and reference images, we introduce a consistency reward model that jointly scores image pairs along three quality axes and optimizes generation via reinforcement learning. Experiments show that our model DetailAnywhere significantly outperforms all state-of-the-art opensource methods across all metrics and human evaluations.
Chinese Translation
基于扩散的生成性人工智能在电子商务应用中取得了显著成功，例如虚拟试穿、海报生成和产品背景合成。然而，在进行服装在线购买决策时，消费者希望能够自由地检查特定的细节区域，例如领子、袖口和面料纹理，但现有方法尚未明确研究这种设置。因此，我们正式提出一个新的非模板任务：聚焦条件下的时尚细节生成，并发布FDBench，这是第一个包含41个不同类别、超过40K人类验证参考细节对的基准。该任务提出了一个独特的语义差距挑战：模型必须弥合产品参考图像上的聚焦标记与所指区域的照片级真实特写视图之间的对应关系，同时忠实地保留服装的身份，而无需任何精确提示。为了解决这一差距，我们提出了跨模态特征对齐蒸馏（Cross-modal Feature Alignment Distillation, CFAD），该方法利用微调的DINOv3教师，通过双分支蒸馏在共享语义空间中对多模态扩散变换器的两个分支进行对齐。为了进一步提高生成细节与参考图像之间的一致性，我们引入了一种一致性奖励模型，该模型沿三个质量维度联合评分图像对，并通过强化学习优化生成。实验表明，我们的模型DetailAnywhere在所有指标和人类评估中显著优于所有最先进的开源方法。
cs.CV / 96 / 2607.02230
Efficient Waste Sorting for Circular Economy: A Confidence-guided comparison between One-Vs-All and One-Vs-Rest Classification Strategies with Human-in-the-Loop for Automated Waste Sorting
循环经济中的高效垃圾分类：基于信心的One-Vs-All与One-Vs-Rest分类策略的比较，结合人机协作实现自动化垃圾分类
Abstract
The complexity of waste disposal regulations across European countries poses significant challenges for the residents and hinders the transition to a Circular Economy. In Germany, the proper sorting and disposal of household waste remains challenging across municipalities. Consequently, substantially reducing incorrectly disposed waste is vital for improving waste management and advancing the Circular Economy. AI-based waste sorting solutions can support residents through user-friendly tools, such as mobile applications, that guide proper waste disposal. To be effective in supporting the Circular Economy, however, these solutions must be configurable to reflect the specific waste sorting scheme of individual municipalities in Germany. In the scope of this work, an evaluation and analysis are performed of two prominent classification strategies: OvA and OvR. The research uses a dataset constructed in alignment with the waste categories and sorting scheme of the city of Goslar in Germany. Moreover, this work aims to extend beyond the overall performance by examining the behavior of OvA and OvR classification strategies in identifying samples likely to be misclassified. These classification strategies are compared by applying varying confidence thresholds to identify uncertain samples for subsequent human review. This evaluation aims to balance the number of misclassifications against the human effort required for data annotation.
Chinese Translation
欧洲各国垃圾处理法规的复杂性给居民带来了重大挑战，并阻碍了向循环经济的转型。在德国，家庭垃圾的正确分类和处理在各个市镇中仍然面临挑战。因此，显著减少错误处理的垃圾对于改善垃圾管理和推动循环经济至关重要。基于人工智能的垃圾分类解决方案可以通过用户友好的工具（如移动应用程序）来支持居民，引导其进行正确的垃圾处理。然而，为了有效支持循环经济，这些解决方案必须能够根据德国各市镇的具体垃圾分类方案进行配置。在本研究中，对两种主要的分类策略：One-Vs-All (OvA) 和 One-Vs-Rest (OvR) 进行了评估和分析。研究使用的数据集是根据德国戈斯拉尔市的垃圾类别和分类方案构建的。此外，本研究旨在超越整体性能，通过考察OvA和OvR分类策略在识别可能被错误分类样本时的表现。通过应用不同的信心阈值来识别不确定样本，以便进行后续的人类审查，从而比较这些分类策略。这一评估旨在平衡错误分类的数量与数据标注所需的人力投入。
cs.CV / 97 / 2607.02237
When Token Compression Breaks: Structural Pruning vs. Token Reduction for Robust ViT Segmentation under High Compression
当令牌压缩失效时：结构剪枝与令牌减少在高压缩下对鲁棒性ViT分割的影响
Abstract
Vision Transformers (ViTs) are strong backbones for semantic segmentation, but their computational cost limits deployment. Recent token compression methods for efficient transformer-based segmentation reduce this cost by decreasing the number of tokens. However, existing evaluations primarily focus on low-to-moderate compression, leaving their behavior under aggressive compression and corrupted inputs unclear. Meanwhile, structural pruning provides an orthogonal route to efficiency by removing redundant components in the ViT architecture, but is rarely compared to token compression under a unified protocol. To bridge this gap, we benchmark representative token compression and structural pruning methods for ViT-based semantic segmentation under matched FLOPs on ADE20K and Cityscapes, together with their common-corruption variants ADE20K-C and Cityscapes-C. Our results reveal a consistent trend on both clean and corrupted inputs: token compression is highly effective at mild reductions but degrades sharply when compression becomes severe, consistent with substantial information loss from overly aggressive token reduction. In contrast, structural pruning exhibits a smoother degradation curve and is more stable at high compression. Motivated by these findings, we study a prune-then-merge pipeline that applies moderate token compression on top of a moderately pruned backbone. At comparable FLOPs, this combined strategy consistently achieves a better accuracy-robustness trade-off at high compression, offering a practical recipe for deployment-oriented ViT segmentation. Code is available at https://github.com/phatnguyencs/vit-seg-compression.
Chinese Translation
视觉变换器（ViTs）是语义分割的强大骨干网络，但其计算成本限制了部署。最近的令牌压缩方法通过减少令牌数量来降低这种成本，从而实现高效的基于变换器的分割。然而，现有评估主要集中在低到中等压缩，尚不清楚在激进压缩和输入损坏情况下的表现。同时，结构剪枝通过去除ViT架构中的冗余组件提供了一条正交的效率路径，但在统一协议下很少与令牌压缩进行比较。为填补这一空白，我们在ADE20K和Cityscapes上对代表性的令牌压缩和结构剪枝方法进行了基准测试，确保在匹配的FLOPs下进行比较，同时还考虑了它们的常见损坏变体ADE20K-C和Cityscapes-C。我们的结果揭示了在干净和损坏输入上都存在一致的趋势：令牌压缩在轻微减少时非常有效，但当压缩变得严重时急剧下降，这与过度激进的令牌减少导致的显著信息损失一致。相比之下，结构剪枝表现出更平滑的降级曲线，并在高压缩下更为稳定。基于这些发现，我们研究了一种先剪枝后合并的流程，在适度剪枝的骨干网络上应用适度的令牌压缩。在可比的FLOPs下，这种组合策略在高压缩时始终实现了更好的准确性与鲁棒性之间的权衡，为面向部署的ViT分割提供了一种实用的方案。代码可在 https://github.com/phatnguyencs/vit-seg-compression 获取。
cs.CV / 98 / 2607.02252
ArcAD: Anomaly-Rectified Calibration for Cold-Start Supervised Anomaly Detection
ArcAD：冷启动监督异常检测的异常校正校准
Abstract
The deployment of Industrial Anomaly Detection (IAD) in real-world manufacturing frequently encounters a challenging cold-start bottleneck, in which limited normal samples fail to represent the full normal distribution and only a few anomalies are available. Under such a regime, existing methods struggle to form compact normal boundaries and fail to effectively exploit supervised signals from rare defects. To address this challenge, we propose Anomaly-Rectified Cold-start AD (ArcAD), a plug-and-play calibration framework for reconstruction-based IAD baselines. ArcAD follows a push-pull learning paradigm to construct a compact and discriminative normal boundary under data scarcity. On the one hand, ArcAD projects limited normal samples onto a hypersphere and pulls them into multiple compact clusters to maximize coverage of the normal manifold. On the other hand, it synthesizes pseudo-anomalies on the hypersphere and leverages real anomalies to push the boundary inward and sharpen anomaly discrimination. Extensive experiments on MVTec-AD, VisA, Real-IAD, and MANTA demonstrate that ArcAD significantly outperforms state-of-the-art supervised and unsupervised methods in both single-class and multi-class settings under cold-start conditions. Code is available at: https://github.com/LGC-AD/ArcAD.
Chinese Translation
工业异常检测（IAD）在实际制造中的部署常常面临一个挑战性的冷启动瓶颈，在这种情况下，有限的正常样本无法代表完整的正常分布，并且只有少量异常样本可用。在这种情况下，现有方法难以形成紧凑的正常边界，并未能有效利用稀有缺陷中的监督信号。为了解决这一挑战，我们提出了异常校正冷启动异常检测（ArcAD），这是一种用于基于重建的IAD基线的即插即用校准框架。ArcAD遵循推拉学习范式，在数据稀缺的情况下构建紧凑且具有区分性的正常边界。一方面，ArcAD将有限的正常样本投影到超球面上，并将它们拉入多个紧凑的聚类，以最大化正常流形的覆盖范围。另一方面，它在超球面上合成伪异常，并利用真实异常将边界向内推移，从而增强异常的区分能力。在MVTec-AD、VisA、Real-IAD和MANTA上的大量实验表明，ArcAD在冷启动条件下显著优于最先进的监督和无监督方法，无论是在单类还是多类设置中。代码可在以下链接获取：https://github.com/LGC-AD/ArcAD。
cs.CV / 99 / 2607.02269
AnyGroundBench: A Specialized-Domain Benchmark for Video Grounding in Vision-Language Models
AnyGroundBench：用于视觉语言模型的视频定位的专业领域基准
Abstract
Vision-Language Models (VLMs) have demonstrated immense promise in Spatio-Temporal Video Grounding (STVG). However, current evaluation protocols are largely confined to zero-shot assessments on general, daily-life benchmarks. This creates a critical disconnect from real-world applications in specialized fields, where models inevitably encounter rare visual concepts and complex spatio-temporal dynamics. Since exhaustive pre-training across infinite data distributions is infeasible, the ability to adapt to novel domains is essential. To bridge this gap, we introduce AnyGroundBench, a domain-adaptation benchmark designed to shift the STVG evaluation paradigm from static zero-shot testing to rigorous domain adaptation. Targeting five specialized domains (animal, industry, sports, surgery, and public security), AnyGroundBench pairs newly captured videos such as expert-annotated mouse behaviors with established datasets, unifying them through dense, high-fidelity spatio-temporal annotations. Crucially, the benchmark provides dedicated training subsets to systematically measure domain adaptability. We extensively evaluate 15 state-of-the-art VLMs, assessing their zero-shot generalization and In-Context Learning (ICL) capabilities under practical computational constraints. Ultimately, our findings reveal that current models fail in both zero-shot and ICL-based adaptation when confronted with specialized domains, exposing critical flaws in spatio-temporal reasoning that future research must address.
Chinese Translation
视觉语言模型（VLMs）在时空视频定位（STVG）方面展现出了巨大的潜力。然而，目前的评估协议主要局限于对一般日常基准的零样本评估。这导致了与专业领域实际应用之间的严重脱节，在这些领域中，模型不可避免地会遇到稀有的视觉概念和复杂的时空动态。由于在无限数据分布上进行全面预训练是不可行的，因此适应新领域的能力至关重要。为了解决这一问题，我们提出了AnyGroundBench，这是一个旨在将STVG评估范式从静态零样本测试转变为严格领域适应的基准。AnyGroundBench针对五个专业领域（动物、工业、体育、外科手术和公共安全），将新捕获的视频（例如专家注释的鼠标行为）与已建立的数据集配对，通过密集的高保真时空注释将它们统一起来。重要的是，该基准提供专门的训练子集，以系统地测量领域适应性。我们对15个最先进的VLM进行了广泛评估，评估它们在实际计算约束下的零样本泛化和上下文学习（ICL）能力。最终，我们的研究结果揭示了当前模型在面对专业领域时在零样本和基于ICL的适应中均表现不佳，暴露了时空推理中的关键缺陷，未来的研究必须加以解决。
cs.CV / 100 / 2607.02271
AGVBench: A Reliability-Oriented Benchmark of Data Augmentation for Vein Recognition
AGVBench：一种面向可靠性的静脉识别数据增强基准
Abstract
Vein recognition is a secure biometric technology often constrained by limited annotated data and imaging variations. While data augmentation mitigates this, strategies designed for natural images may disrupt the fine-grained topology and textures essential for identity discrimination. We present AGVBench, which evaluates 30 representative augmentation strategies on five public palm- and finger-vein datasets with seven backbone architectures, covering classic CNNs, vision transformers, and vein-specific recognition models. Our results show that multi-image mixing methods (e.g., MixUp, PuzzleMix, StarMixup) generally provide the strongest recognition performance. However, they are often poorly calibrated and vulnerable to adversarial perturbations, revealing a clear inconsistency between clean accuracy and adversarial security. We also find that severe geometric transformations frequently degrade recognition, which is potentially due to feature misalignment or spatial cropping, and that augmentation effectiveness varies across palm and finger vein datasets. These findings prove that accuracy-centric evaluation is insufficient for biometric augmentation. AGVBench provides standardized protocols to support reproducible research and guide the design of reliable, secure, and robust vein recognition systems. Our codebase is available at https://github.com/Advance-VeinTech-Innovators/AGVBench.
Chinese Translation
静脉识别是一种安全的生物识别技术，通常受到有限标注数据和成像变化的限制。虽然数据增强可以缓解这一问题，但为自然图像设计的策略可能会破坏身份识别所必需的细粒度拓扑和纹理。我们提出了AGVBench，它在五个公共掌静脉和指静脉数据集上评估了30种代表性的增强策略，涵盖了七种主干架构，包括经典卷积神经网络（CNN）、视觉变换器（Vision Transformers）和特定于静脉识别的模型。我们的结果表明，多图像混合方法（例如，MixUp、PuzzleMix、StarMixup）通常提供最强的识别性能。然而，它们往往校准不佳，容易受到对抗性扰动的影响，揭示了干净准确性与对抗安全性之间的明显不一致。我们还发现，严重的几何变换常常会降低识别性能，这可能是由于特征错位或空间裁剪造成的，并且增强效果在掌静脉和指静脉数据集之间存在差异。这些发现证明了以准确性为中心的评估对于生物识别增强是不够的。AGVBench提供了标准化的协议，以支持可重复的研究并指导可靠、安全和稳健的静脉识别系统的设计。我们的代码库可在 https://github.com/Advance-VeinTech-Innovators/AGVBench 获取。
cs.CV / 101 / 2607.02284
FlowCIR: Semantic Transport via Flow Matching for Zero-Shot Composed Image Retrieval
FlowCIR：通过流匹配实现零-shot组合图像检索的语义传输
Abstract
Zero-shot composed image retrieval (ZS-CIR) aims to retrieve a target image by editing a reference image with a natural-language instruction, without relying on domain-specific annotated triplets. Most existing ZS-CIR methods rely on textual inversion to translate the reference image into pseudo-text tokens and then compose them with the instruction via simple concatenation in the text space, which can be lossy and brittle for fine-grained semantics. In this work, we propose a new paradigm, namely FlowCIR, that casts ZS-CIR as conditional semantic transport between reference and target embeddings. Leveraging \emph{conditional flow matching}, our model learns a lightweight transport field that maps the instruction representation toward a target-aligned query embedding conditioned on the reference image. Since FlowCIR operates on pre-extracted VLM embeddings and trains only a small transport module without updating the image or text encoder, it offers a computationally efficient training protocol compared with prior textual-inversion-based approaches. The resulting framework is training-efficient, requiring roughly $10\times$ fewer training resources than prior textual-inversion-based approaches. We further identify negation and removal as a major failure mode of VLM-based composition. To address this, we propose an inference-only Multi-Negative Steering strategy that steers a negation-containing relative instruction away from its negated semantics, mitigating the limited negation handling of VLMs and improving robustness on negation-heavy queries. Extensive experiments on standard CIR benchmarks demonstrate that FlowCIR achieves strong and competitive performance compared with recent ZS-CIR methods.
Chinese Translation
零-shot组合图像检索（ZS-CIR）旨在通过自然语言指令编辑参考图像来检索目标图像，而无需依赖于特定领域的注释三元组。现有的大多数ZS-CIR方法依赖于文本反转，将参考图像转换为伪文本标记，然后通过简单的拼接在文本空间中与指令进行组合，这对于细粒度语义可能是有损且脆弱的。在本研究中，我们提出了一种新的范式，即FlowCIR，将ZS-CIR视为参考和目标嵌入之间的条件语义传输。利用条件流匹配（conditional flow matching），我们的模型学习一个轻量级的传输场，该场将指令表示映射到基于参考图像的目标对齐查询嵌入。由于FlowCIR在预提取的VLM嵌入上操作，并且仅训练一个小型传输模块而不更新图像或文本编码器，因此与先前基于文本反转的方法相比，它提供了一种计算上高效的训练协议。所得到的框架在训练效率上表现优越，所需的训练资源大约比先前的基于文本反转的方法少$10 imes$。我们进一步识别出否定和移除是基于VLM组合的主要失败模式。为了解决这个问题，我们提出了一种仅推理的多负引导策略（Multi-Negative Steering），该策略将包含否定的相对指令引导远离其否定语义，从而减轻VLM在否定处理方面的局限性，并提高在否定重查询上的鲁棒性。在标准CIR基准上的大量实验表明，FlowCIR与近期的ZS-CIR方法相比，表现出强大且具有竞争力的性能。
cs.CV / 102 / 2607.02290
DisciplineGen-1M: A Large-Scale Dataset for Multidisciplinary Visual Generation and Editing
DisciplineGen-1M：一个用于多学科视觉生成和编辑的大规模数据集
Abstract
Recent image generation and editing models can produce visually appealing natural images, yet they remain unreliable when the target image is a knowledge-intensive diagram whose correctness depends on disciplinary concepts, symbolic structure, and precise spatial relations. We introduce DisciplineGen-1M, a million-scale multidisciplinary dataset that supports text-to-image generation and image editing. It contains 1.2M samples spanning mathematics, physics, chemistry, biology, geography, computer science, economics, history, music, and sports. To construct the dataset, we design a scalable framework that combines vector-graphics rendering, OCR-based editing, curated programmatic synthesis, and large-scale text-to-image filtering. These pipelines produce captions, editing instructions, structured annotations, and paired images with controllable semantic differences. Building on DisciplineGen-1M, we further introduce a discipline-informed reasoning-generation model for both text-to-image generation and image editing. Experiments on discipline-related benchmarks, GenExam and GRADE, show substantial improvements over open-source baselines, while evaluations on general reasoning-informed benchmarks, WISE and RISE, further indicate broader transfer. The results suggest that large-scale structured academic visual data is a key ingredient for moving image generation from aesthetic plausibility toward verifiable knowledge-grounded visual creation. We will publicly release our dataset, model, and source code of the data curation pipeline to ensure reproducibility and benefit future research.
Chinese Translation
近期的图像生成和编辑模型能够生成视觉上吸引人的自然图像，但当目标图像是知识密集型图表时，它们仍然不可靠，因为这些图表的正确性依赖于学科概念、符号结构和精确的空间关系。我们介绍了DisciplineGen-1M，这是一个百万级的多学科数据集，支持文本到图像的生成和图像编辑。该数据集包含120万样本，涵盖数学、物理、化学、生物、地理、计算机科学、经济学、历史、音乐和体育等领域。为了构建该数据集，我们设计了一个可扩展的框架，结合了矢量图形渲染、基于OCR的编辑、策划的程序合成和大规模的文本到图像过滤。这些流程生成了标题、编辑指令、结构化注释和具有可控语义差异的配对图像。在DisciplineGen-1M的基础上，我们进一步引入了一种学科知情的推理生成模型，用于文本到图像的生成和图像编辑。在与学科相关的基准测试GenExam和GRADE上的实验显示，相较于开源基线有显著改进，而在一般推理知情基准WISE和RISE上的评估进一步表明了更广泛的迁移能力。结果表明，大规模结构化的学术视觉数据是将图像生成从美学可信性转向可验证的知识基础视觉创作的关键成分。我们将公开发布我们的数据集、模型和数据策划管道的源代码，以确保可重复性并惠及未来的研究。
cs.CV / 103 / 2607.02299
Dual-Selective Network for Domain-Incremental Change Detection
双重选择网络用于领域增量变化检测
Abstract
Domain-incremental change detection (DICD) continuously adapts models to new geographic domains while preserving prior knowledge. However, a structural mismatch exists: the label space remains fixed while domain characteristics vary drastically. Consequently, incremental models struggle to maintain stable spatial change representations across domains. Existing strategies, such as replay-based or regularization-based methods, often fail to scale to long domain sequences, leading to knowledge degradation or increased computational cost. We propose Dual-Selective Incremental Network (DSINet), a unified framework built on visual state space models. DSINet leverages Mamba's input-dependent selective mechanism through a selective spatial state unit (S3U). This unit preserves stable spatial change structures while filtering domain-specific variations during feature propagation. As a result, spatial representations remain stable across domains, preventing the accumulation of feature confusion over incremental steps. Additionally, we employ a concentration-balanced distillation (CBD) strategy to stabilize knowledge transfer across domains. It balances hardness and confidence concentration effects during incremental updates. This ensures reliable probability mass allocation and prevents over-smoothing or mode collapse during distillation. Together, these mechanisms maintain stable learning dynamics throughout incremental stages. Experimental results demonstrate that DSINet mitigates knowledge degradation across long domain sequences while maintaining the linear computational efficiency of state space models.
Chinese Translation
领域增量变化检测（DICD）在不断适应新地理领域的同时保持先前知识。然而，存在结构不匹配的问题：标签空间保持固定，而领域特征却发生剧烈变化。因此，增量模型在不同领域之间难以维持稳定的空间变化表示。现有策略，如基于重放或正则化的方法，往往无法扩展到长领域序列，导致知识退化或计算成本增加。我们提出了双重选择增量网络（DSINet），这是一个基于视觉状态空间模型的统一框架。DSINet通过选择性空间状态单元（S3U）利用Mamba的输入依赖选择机制。该单元在特征传播过程中过滤领域特定的变化，同时保持稳定的空间变化结构。因此，空间表示在不同领域之间保持稳定，防止在增量步骤中出现特征混淆的累积。此外，我们采用了集中平衡蒸馏（CBD）策略，以稳定领域间的知识转移。该策略在增量更新过程中平衡了难度和置信度集中效应。这确保了可靠的概率质量分配，并防止在蒸馏过程中出现过度平滑或模式崩溃。这些机制共同维持了增量阶段的稳定学习动态。实验结果表明，DSINet在长领域序列中减轻了知识退化，同时保持了状态空间模型的线性计算效率。
cs.CV / 104 / 2607.02300
Search-based Testing of Vision Language Models for In-Car Scene Understanding
基于搜索的汽车场景理解视觉语言模型测试
Abstract
In the automotive domain, in-car scene understanding (ISU) enables the detection of safety-critical events, such as driver distraction, and supports drivers or passengers by analyzing the in-car scene and adapting the environment (e.g., ambient lighting). The industry is increasingly exploring vision-language models (VLMs) to interpret camera-recorded in-car scenes and extract information for downstream reasoning tasks. However, VLMs may generate incomplete, erroneous, or misleading scene descriptions, highlighting the need for systematic testing. Collecting real in-vehicle data is costly, difficult to scale, and often infeasible, particularly in early design stages. In this paper, we present ISU-Test, an automated testing approach that combines rendering-based scene generation with search-based testing to evaluate ISU systems. By framing testing as an optimization problem and systematically modifying scene parameters, our method generates diverse in-car scenarios and explores a wide range of configurations. We evaluate ISU-Test on both an industrial prototype and open-source VLMs across two case studies: question answering and captioning, comparing against randomized scenario generation. Results show that ISU-Test significantly outperforms the baseline, achieving up to 10 times higher failure rates and up to 3.6 times higher failure coverage.
Chinese Translation
在汽车领域，车内场景理解（ISU）能够检测安全关键事件，如驾驶员分心，并通过分析车内场景和调整环境（例如，环境光照）来支持驾驶员或乘客。行业越来越多地探索视觉语言模型（VLMs）来解释摄像头记录的车内场景，并提取信息以进行下游推理任务。然而，VLMs 可能生成不完整、错误或误导性的场景描述，这突显了系统测试的必要性。收集真实的车载数据成本高昂，难以扩展，并且在早期设计阶段往往不可行。本文提出了 ISU-Test，一种自动化测试方法，结合基于渲染的场景生成与基于搜索的测试，以评估 ISU 系统。通过将测试框架设定为优化问题，并系统性地修改场景参数，我们的方法生成多样化的车内场景并探索广泛的配置。我们在一个工业原型和两个开源 VLMs 上评估 ISU-Test，涵盖了两个案例研究：问答和图像描述，并与随机场景生成进行了比较。结果表明，ISU-Test 显著优于基线，失败率最高可达 10 倍，失败覆盖率最高可达 3.6 倍。
cs.CV / 105 / 2607.02301
InvSplat: Inverse Feed-Forward Scene Splatting
InvSplat：逆向前馈场景点云
Abstract
Inverse rendering aims to recover both 3D geometry and physically meaningful material properties from images, enabling applications such as relighting and novel view synthesis. Optimization-based methods achieve high fidelity but require costly per-scene fitting, while image-space learning-based approaches often suffer from multi-view inconsistencies and lack an explicit 3D representation for stable novel view rendering. We present a feed-forward multi-view reconstruction framework for inverse rendering that directly predicts a structured 3D Gaussian representation with intrinsic material attributes. Each Gaussian primitive is parameterized by mean, normal, opacity, rotation, scale, albedo, metallic, and roughness, enabling a disentangled and physically grounded scene representation. Our model integrates priors from a material estimation network with a multi-view 3D reconstruction backbone, allowing joint prediction of geometry and reflectance parameters in a single forward pass. Experiments on synthetic and real-world datasets demonstrate improved multi-view consistency compared to 2D baselines, accurate material recovery, and stable novel view rendering. Our representation further supports physically-based relighting and more faithful modeling of view-dependent effects compared to existing RGB-based feed-forward reconstruction methods. Our project webpage is: $\href{https://poliik.github.io/invsplat/}{\text{https://poliik.github.io/invsplat/}}$.
Chinese Translation
逆向渲染旨在从图像中恢复三维几何形状和物理意义明确的材料属性，从而实现重新照明和新视图合成等应用。基于优化的方法虽然能够实现高保真度，但需要昂贵的逐场景拟合，而基于图像空间的学习方法往往面临多视图不一致的问题，并且缺乏稳定的新视图渲染所需的明确三维表示。我们提出了一种用于逆向渲染的前馈多视图重建框架，该框架直接预测具有内在材料属性的结构化三维高斯表示。每个高斯原语由均值、法线、不透明度、旋转、缩放、反照率、金属度和粗糙度参数化，从而实现了一个解耦且物理基础的场景表示。我们的模型将材料估计网络的先验与多视图三维重建主干网络相结合，允许在单次前向传播中共同预测几何形状和反射参数。在合成和真实世界数据集上的实验表明，与二维基线相比，我们的方法在多视图一致性、材料恢复的准确性和新视图渲染的稳定性方面都有所改善。与现有的基于RGB的前馈重建方法相比，我们的表示进一步支持基于物理的重新照明和更真实的视依赖效应建模。我们的项目网页是：$ ext{https://poliik.github.io/invsplat/}$.
cs.CV / 106 / 2607.02317
NEvo: Neural-Guided Evolutionary Video Synthesis for Dynamic Visual Selectivity
NEvo：神经引导的进化视频合成用于动态视觉选择性
Abstract
The human brain processes dynamic visual input through hierarchically organized, functionally specialized regions. While recent in silico brain encoding models can synthesize optimal stimuli to probe selectivity in different brain regions, prior work has been largely limited to static images, leaving dynamic visual processing underexplored. We introduce a novel neural-guided video synthesis framework that generates stimuli optimized for target brain regions across visual cortex. Our method performs evolutionary search over a structured prompt space, guided by a dynamic encoding model that predicts voxel-level responses to video inputs. By maximizing predicted activity for a target ROI, the framework efficiently discovers hyper-activating dynamic stimuli that consistently surpass handcrafted localizer videos. The synthesized videos recover known selectivities across ventral, dorsal, and lateral pathways, and further reveal systematic differences in sensitivity to temporal dynamics. A searchlight analysis provides new insight into the progression toward increasingly complex social-dynamic features along the lateral stream, further supported by probing with synthesized abstract, non-naturalistic stimuli. Taken together, our framework enables in silico exploration of dynamic visual selectivity, with new predictions for in vivo experiments
Chinese Translation
人脑通过层次化组织、功能专门化的区域处理动态视觉输入。尽管近期的计算机模拟脑编码模型能够合成最佳刺激以探测不同脑区的选择性，但之前的研究主要局限于静态图像，动态视觉处理仍未得到充分探索。我们提出了一种新颖的神经引导视频合成框架，该框架生成针对视觉皮层目标脑区优化的刺激。我们的方法在结构化提示空间中执行进化搜索，受动态编码模型的指导，该模型预测对视频输入的体素级响应。通过最大化目标感兴趣区域（ROI）的预测活动，该框架有效地发现了超激活的动态刺激，这些刺激始终超越手工制作的定位视频。合成的视频恢复了腹侧、背侧和侧面通路已知的选择性，并进一步揭示了对时间动态的敏感性系统性差异。搜索光分析提供了对侧面通路中朝向越来越复杂的社会动态特征进展的新见解，进一步通过合成的抽象非自然刺激进行探测得到了支持。综上所述，我们的框架使得对动态视觉选择性的计算机模拟探索成为可能，并为体内实验提供了新的预测。
cs.CV / 107 / 2607.02360
GAP-GDRNet: Geometry-Aware Monocular Visual Pose Sensing on a Single-Target Synthetic Spacecraft Dataset
GAP-GDRNet：基于几何感知的单目视觉姿态感知在单目标合成航天器数据集上的应用
Abstract
Monocular relative pose sensing is a central perception problem in non-cooperative rendezvous and on-orbit servicing. In spacecraft images, however, weak surface texture, thin appendages, illumination changes, and partial occlusion often leave only sparse and unstable geometric evidence. This article presents GAP-GDRNet, a geometry-aware attention-enhanced framework for monocular RGB-based 6D pose sensing. The method follows the geometry-guided direct regression paradigm of GDR-Net and modifies two points in the pipeline: an attention-based feature refinement (AFR) module is placed before dense geometric prediction, and a patch-level geometric self-attention (PGSA) module is inserted into Patch-PnP. AFR reinforces global spacecraft structure together with local weak-texture cues; PGSA then relates downsampled geometric patches before final pose regression. A Blender-based annotation process supplies target masks, visible-region masks, dense model-coordinate maps, camera intrinsics, and 6D pose labels for supervised training.
Chinese Translation
单目相对姿态感知是非合作交会和在轨服务中的一个核心感知问题。然而，在航天器图像中，弱表面纹理、细长附属物、光照变化和部分遮挡往往只留下稀疏和不稳定的几何证据。本文提出了GAP-GDRNet，一种基于几何感知的增强注意力框架，用于单目RGB的6D姿态感知。该方法遵循GDR-Net的几何引导直接回归范式，并在流程中修改了两个关键点：在密集几何预测之前放置了一个基于注意力的特征精炼（AFR）模块，并在Patch-PnP中插入了一个补丁级几何自注意力（PGSA）模块。AFR结合局部弱纹理线索增强了全球航天器结构；PGSA则在最终姿态回归之前关联了下采样的几何补丁。基于Blender的标注过程提供了目标掩膜、可见区域掩膜、密集模型坐标图、相机内参和6D姿态标签，以支持监督训练。
cs.CV / 108 / 2607.02371
VisionAId: An Offline-First Multimodal Android Assistant for People with Visual Impairment, Featuring Personalized Object Retrieval
VisionAId：一款离线优先的多模态安卓助手，旨在为视觉障碍人士提供个性化物体检索
Abstract
Over 285 million people worldwide live with a visual impairment, for whom everyday tasks such as avoiding obstacles, locating personal belongings, recognizing familiar faces, or handling cash remain persistent obstacles to personal autonomy. Existing assistive applications are typically limited to recognizing predefined categories, depend heavily on cloud connectivity, or require dedicated hardware. We present VisionAId, an Android application that turns a commodity smartphone into a real-time visual assistant. The system integrates six on-device deep learning models (metric monocular depth estimation, instance segmentation, visual and facial embeddings, face detection, and a custom banknote detector) running entirely through ONNX Runtime, with an optional cloud large language model (Google Gemini Flash) used only for narrative scene description and automatic object labeling. A distinctive contribution is a few-shot pipeline for personal objects: the user photographs an object from several angles, and the system later locates that specific instance in the environment, guiding the user toward it with augmented-reality markers, spatial audio, and distance-proportional haptics. All feedback is multimodal (Romanian speech synthesis, voice commands, vibration). On a reference device (Samsung Galaxy S21 Ultra), INT8 quantization reduces depth latency from ~1200 ms to ~491 ms, the custom banknote detector reaches an mAP@50 of 0.986, and metric depth is calibrated to below 1 cm of error within 3 m.
Chinese Translation
全球有超过2.85亿人生活在视觉障碍中，对于他们而言，日常任务如避开障碍物、定位个人物品、识别熟悉面孔或处理现金仍然是个人自主性的持续障碍。现有的辅助应用通常仅限于识别预定义类别，严重依赖云连接，或需要专用硬件。我们提出了VisionAId，一款将普通智能手机转变为实时视觉助手的安卓应用。该系统集成了六个在设备上运行的深度学习模型（度量单目深度估计、实例分割、视觉和面部嵌入、面部检测以及自定义钞票检测器），完全通过ONNX Runtime运行，同时可选的云大型语言模型（Google Gemini Flash）仅用于叙述场景描述和自动物体标记。一个独特的贡献是针对个人物体的少量样本管道：用户从多个角度拍摄物体照片，系统随后在环境中定位该特定实例，通过增强现实标记、空间音频和与距离成比例的触觉反馈引导用户前往。所有反馈均为多模态（罗马尼亚语语音合成、语音命令、震动）。在参考设备（Samsung Galaxy S21 Ultra）上，INT8量化将深度延迟从约1200毫秒降低至约491毫秒，自定义钞票检测器在mAP@50上达到0.986，且在3米内的度量深度误差校准低于1厘米。
cs.CV / 109 / 2607.02372
Learning Spectral and Polarimetric Clues for One-to-Multimodal Novel View Synthesis
学习光谱和极化线索用于单一到多模态的新视图合成
Abstract
Neural rendering techniques allow for accurate reconstruction of the geometry and color appearance of 3D scenes. Some methods have extended their use to additional imaging modalities, such as multispectral, infrared, or polarimetric data. However, all of these approaches require expensive sensors and calibrated setups to capture new multimodal frames for each new scene. We propose Spectral and Polarimetric Implicit Learned Representation (SPoILeR), a novel method to obtain multi-view consistent renderings of unconventional modalities for scenes where either only RGB frames or very few of the additional modalities are available. Thanks to a multimodal pre-training phase, the model learns the mutual correlation between different modalities. This step allows predicting accurate renderings of unconventional modalities during a fine-tuning phase supervised only by RGB images. Experimental results show that the approach can accurately render infrared, polarimetric, and multispectral frames for scenes where no input sample captured by these types of sensors is provided.
Chinese Translation
神经渲染技术能够准确重建三维场景的几何形状和颜色外观。一些方法已将其应用扩展到额外的成像模态，如多光谱、红外或极化数据。然而，所有这些方法都需要昂贵的传感器和经过校准的设置，以捕获每个新场景的多模态新帧。我们提出了一种新方法——光谱和极化隐式学习表示（Spectral and Polarimetric Implicit Learned Representation，SPoILeR），用于在仅有RGB帧或极少额外模态可用的场景中获得多视图一致的非常规模态渲染。得益于多模态预训练阶段，模型学习了不同模态之间的相互关联。这一步骤使得在仅由RGB图像监督的微调阶段能够预测非常规模态的准确渲染。实验结果表明，该方法能够准确渲染红外、极化和多光谱帧，即使在没有由这些类型传感器捕获的输入样本的情况下。
cs.CV / 110 / 2607.02375
Representation Distribution Matching for One-Step Visual Generation
用于一步视觉生成的表示分布匹配
Abstract
We elucidate the design space of Representation Distribution Matching (RDM), our name for the paradigm that trains a one-step image generator by matching generated and reference feature distributions under frozen pretrained encoders. We identify two design axes, how the distributions are compared and the representations they are compared in, and controlled studies along them yield three findings. First, the classical MMD, which could not train convincing generators a decade ago, becomes a strong and scalable objective once estimated right. Second, the generated batch is then the operative variable, with an optimum above 2048, far beyond customary batch sizes. Third, any single representation can be gamed, driven below the real score while images stay visibly fake, so we match against a balanced battery of encoders and evaluate with SW_r14, a Sliced-Wasserstein distance over 14 encoders that is independent of the training loss and resists gaming. Combining the preferred choices yields improved RDM (iRDM): it sets the one-step state of the art on ImageNet at SW_r14 1.30, corroborated by PickScore, a human-preference proxy our objective never optimizes, which prefers it over the prior best one-step generator on 71.2% of matched samples. The same recipe post-trains the four-step FLUX.2 [klein] into a one-step generator, surpassing the four-step version on GenEval, 0.826 to 0.794, and on PickScore, 22.76 to 22.58, in 90 H200 GPU-hours. Project page: https://alan-lanfeng.github.io/rdm/.
Chinese Translation
我们阐明了表示分布匹配（Representation Distribution Matching, RDM）的设计空间，这是我们为通过在冻结的预训练编码器下匹配生成和参考特征分布来训练一步图像生成器的范式所命名的。我们识别出两个设计轴，即分布比较的方式以及进行比较的表示，沿这两个轴的控制研究得出了三项发现。首先，经典的最大均值差异（MMD）在十年前无法训练出令人信服的生成器，但一旦正确估计，它便成为一个强大且可扩展的目标。其次，生成的批次是操作变量，最佳批次大小超过2048，远超常规批次大小。第三，任何单一表示都可能被操控，驱动其得分低于真实得分，而图像仍然明显伪造，因此我们对一组平衡的编码器进行匹配，并使用SW_r14进行评估，SW_r14是对14个编码器的切片-瓦瑟斯坦距离，它独立于训练损失并抵抗操控。结合这些优选选择，得到了改进的RDM（iRDM）：它在ImageNet上以SW_r14 1.30设定了一步生成的最新状态，得到了PickScore的验证，PickScore是一个我们目标从未优化的人类偏好代理，71.2%的匹配样本中更倾向于它而非之前最佳的一步生成器。同样的方法将四步FLUX.2 [klein]后训练为一步生成器，在GenEval上超越了四步版本，得分为0.826对0.794，在PickScore上得分为22.76对22.58，耗时90 H200 GPU小时。项目页面：https://alan-lanfeng.github.io/rdm/
cs.CV / 111 / 2607.02386
Transformer Geometry Observatory TGO-II: Representational Similarity Observatory
变换几何观测站 TGO-II：表征相似性观测站
Abstract
While Vision Transformers have achieved remarkable success across computer vision and language applications, the geometric evolution of their internal representations throughout training remains insufficiently understood. Existing analyses primarily focus on attention mechanisms and downstream performance, leaving the evolution of representation geometry largely unexplored. In this work, we present Transformer Geometry Observatory-II (TGO-II), a representation geometry analysis framework designed to investigate how Transformer representations evolve during supervised training. TGO-II analyzes Vision Transformer (ViT-Small/16) representations using Centered Kernel Alignment (CKA), Singular Vector Canonical Correlation Analysis (SVCCA), Two-Nearest Neighbor Intrinsic Dimensionality (TwoNN-ID), and token covariance analysis. Our experiments reveal three key observations. First, both CKA and SVCCA progressively decrease throughout training, indicating increasing representational specialization across Transformer layers. Second, intrinsic dimensionality consistently increases before stabilizing, suggesting progressive expansion of the representation manifold into a larger set of locally accessible degrees of freedom. Third, token covariance and coupling analyses demonstrate that strong token interaction structure persists throughout training, challenging the hypothesis that increasing representational complexity arises primarily from progressive token independence. These findings suggest that representation complexity and layer specialization emerge simultaneously during training. Manifold expansion appears to occur without token decoupling. Together, these observations motivate a new hypothesis in which Vision Transformers increase representational complexity through progressively richer transformations while preserving strong token interaction structure during learning.
Chinese Translation
尽管视觉变换器在计算机视觉和语言应用中取得了显著成功，但其内部表征在训练过程中的几何演变仍然未得到充分理解。现有分析主要集中在注意力机制和下游性能上，导致表征几何的演变大多未被探索。在本研究中，我们提出了变换几何观测站-II (TGO-II)，这是一个旨在研究变换器表征在监督训练过程中如何演变的表征几何分析框架。TGO-II 使用中心核对齐 (Centered Kernel Alignment, CKA)、奇异向量典型相关分析 (Singular Vector Canonical Correlation Analysis, SVCCA)、双最近邻内在维度 (Two-Nearest Neighbor Intrinsic Dimensionality, TwoNN-ID) 和标记协方差分析来分析视觉变换器 (Vision Transformer, ViT-Small/16) 的表征。我们的实验揭示了三个关键观察结果。首先，CKA 和 SVCCA 在训练过程中逐渐下降，表明变换器层之间的表征专业化程度在不断增加。其次，内在维度在稳定之前持续增加，表明表征流形逐渐扩展到更大的一组局部可访问的自由度。第三，标记协方差和耦合分析表明，强标记交互结构在整个训练过程中持续存在，这挑战了表征复杂性主要源于标记独立性逐渐增加的假设。这些发现表明，表征复杂性和层专业化在训练过程中同时出现。流形扩展似乎在没有标记解耦的情况下发生。这些观察结果共同激发了一个新假设，即视觉变换器通过逐步丰富的变换来增加表征复杂性，同时在学习过程中保持强标记交互结构。
cs.CV / 112 / 2607.02402
Show Me Examples: Inferring Visual Concepts from Image Sets
展示示例：从图像集合中推断视觉概念
Abstract
Vision-language models (VLMs) can follow complex textual instructions, yet they struggle to reason from purely visual context. In particular, current models fail to infer shared concepts from sets of example images and apply them to new inputs. We introduce Visual Concept Inference from Sets (VICIS), a task that evaluates this capability. Given a small context set of images sharing a concept and a query image, the model must generate new images that preserve the context-defined concept while remaining consistent with the query. We show that state-of-the-art VLMs perform poorly on this task, often ignoring the visual context or defaulting to biased generations. To address this gap, we propose a training framework and architecture that learn to infer visual concepts from image sets and extract concept-specific embeddings from queries. Experiments on synthetic data and large-scale ImageNet/WordNet data show that our model generates more accurate and diverse outputs and generalizes to unseen concepts and modalities such as sketches.
Chinese Translation
视觉语言模型（VLMs）能够遵循复杂的文本指令，但在纯视觉上下文中推理时却表现不佳。特别是，当前模型无法从一组示例图像中推断共享概念并将其应用于新输入。我们提出了从集合中推断视觉概念（Visual Concept Inference from Sets，VICIS）这一任务，以评估这一能力。给定一小组共享概念的图像作为上下文集和一张查询图像，模型必须生成新的图像，这些图像在保持上下文定义的概念的同时，与查询保持一致。我们展示了最先进的VLM在这一任务上的表现不佳，往往忽视视觉上下文或默认生成偏见结果。为了解决这一问题，我们提出了一种训练框架和架构，旨在从图像集合中推断视觉概念，并从查询中提取特定于概念的嵌入。对合成数据和大规模ImageNet/WordNet数据的实验表明，我们的模型生成了更准确和多样化的输出，并能够推广到未见过的概念和模式，如素描。
cs.CV / 113 / 2607.02404
Object-centric LeJEPA
面向对象的LeJEPA
Abstract
Image encoders trained with LeJEPA can deliver strong features for downstream tasks, but, like other image-level self-supervised methods, typically require large training datasets. Aligning representations at the level of objects rather than whole scenes promises greater data efficiency, but doing this in a completely self-supervised way, effectively jointly partitioning a scene and representing its objects, is unstable: the two are locked in a cyclic dependency, partitioning requires meaningful representations, while meaningful representations require consistent partitioning. We sidestep this instability by taking object masks as given during training, using cheap, off-the-shelf SAM proposals. We extend LeJEPA - whose distributional anti-collapse objective ports naturally from whole images to variable-sized sets of objects - to align object-centric representations rather than whole images. An additional instance-separating loss, which treats other objects in the same scene as negatives, further boosts downstream performance. Across two model scales and 10-100% of COCO, object-level LeJEPA outperforms image-level LeJEPA on tracking (DAVIS), classification (ImageNet-1k), segmentation (ADE20k), and re-identification (NAVI).
Chinese Translation
使用LeJEPA训练的图像编码器能够为下游任务提供强大的特征，但与其他图像级自监督方法一样，通常需要大量的训练数据集。在对象层面而非整个场景对齐表示，承诺提供更高的数据效率，但以完全自监督的方式实现这一点，即有效地共同划分场景并表示其对象，是不稳定的：两者存在循环依赖，划分需要有意义的表示，而有意义的表示又需要一致的划分。我们通过在训练过程中将对象掩码视为给定，使用廉价的现成SAM提案，来规避这种不稳定性。我们扩展了LeJEPA——其分布式反崩溃目标自然地从整个图像迁移到可变大小的对象集合——以对齐面向对象的表示，而不是整个图像。一个额外的实例分离损失，将同一场景中的其他对象视为负样本，进一步提升了下游性能。在两个模型规模和10-100%的COCO数据集上，面向对象的LeJEPA在跟踪（DAVIS）、分类（ImageNet-1k）、分割（ADE20k）和重识别（NAVI）任务中均优于图像级LeJEPA。
cs.CV / 114 / 2607.02421
Wavelet-Guided Semantic Signal Compensation for Inversion-Free Image Editing
基于小波引导的无反演图像编辑语义信号补偿
Abstract
Text-guided image editing aims to modify visual content according to a target prompt while preserving the background. Recent inversion-free image editing frameworks such as FlowEdit have demonstrated strong editing capability without requiring inversion. Empirically, FlowEdit can achieve substantial semantic changes under appropriate hyperparameter settings. However, we observe that under certain global attribute shifts, the editing trajectory may not effectively move away from the source distribution in the early timesteps. Our analysis suggests that in the high-noise regime, the dominant manifold-seeking flow toward the data manifold can reduce the influence of the text-conditioned direction, leading to limited global modification while background structures remain only moderately preserved. Inspired by this observation, we propose an inversion-free, frequency-aware semantic compensation strategy that strengthens the effective signal in the early stage of generation, while maintaining structural consistency in the background. The proposed method improves global editing capacity without sacrificing background fidelity.
Chinese Translation
文本引导的图像编辑旨在根据目标提示修改视觉内容，同时保持背景不变。近期的无反演图像编辑框架，如FlowEdit，展示了强大的编辑能力，无需进行反演。实证研究表明，在适当的超参数设置下，FlowEdit能够实现显著的语义变化。然而，我们观察到在某些全局属性变化下，编辑轨迹在早期时间步可能无法有效远离源分布。我们的分析表明，在高噪声环境下，朝向数据流形的主导流动可能会削弱文本条件方向的影响，从而导致全局修改有限，而背景结构仅得到适度保留。受到这一观察的启发，我们提出了一种无反演、频率感知的语义补偿策略，该策略在生成的早期阶段增强有效信号，同时保持背景的结构一致性。所提出的方法在不牺牲背景保真度的情况下，提高了全局编辑能力。
cs.CV / 115 / 2607.02425
Learning to Evolve Scenes: Reasoning about Human Activities with Scene Graphs
学习演变场景：基于场景图推理人类活动
Abstract
Understanding human behavior while interacting with the surrounding world is crucial for many applications of embodied AI. First-person videos are particularly informative for this problem, as they well capture how activities reshape the scene over time. However, existing approaches often rely on implicit visual or language-aligned representations, disregarding structured reasoning over the scene dynamic. We argue that explicit, compositional and editable representations of human-environment interactions can play a crucial role for rich grounded activity understanding. To this end, we introduce SG-Ego, a large scale annotation set extending Ego4D with spatio-temporal scene graphs, where relations triplets are consolidated over time into explicit time-evolving descriptions of the scene state. To reason over this representation, we propose GLEN, a graph-based model that operates over scene graph sequences to both align them with textual actions and model their temporal evolution. In addition, we formulate the activity-driven graph-edit forecasting (A-GEF) problem, a novel task that casts scene dynamics as a sequence of structured transformations conditioned on ongoing actions, enabling explicit reasoning about how scenes change over time. We validate our approach across multiple downstream tasks, spanning retrieval benchmarks as EgoMCQ and EgoCVR, as well as long-horizon reasoning benchmarks as EXPLORE-Bench and the newly introduced A-GEF. GLEN achieves strong results compared to raw video baselines and it excels in reasoning settings, typically addressed only with MLLMs, while enabling controllable and structured predictions of scene dynamics driven by human activities. We believe our results establish spatio-temporal scene graphs, together with models that reason over them, as strong compositional and interpretable representations for video understanding and potentially beyond.
Chinese Translation
理解人类在与周围世界互动时的行为对于许多具身人工智能应用至关重要。第一人称视频在解决这一问题时特别有价值，因为它们能够很好地捕捉活动如何随时间重塑场景。然而，现有的方法往往依赖于隐式的视觉或语言对齐表示，忽视了对场景动态的结构化推理。我们认为，人类与环境交互的显式、组合和可编辑的表示在丰富的基于情境的活动理解中可以发挥关键作用。为此，我们引入了SG-Ego，一个扩展了Ego4D的大规模注释集，包含时空场景图，其中关系三元组随着时间的推移被整合为场景状态的显式时间演变描述。为了对这一表示进行推理，我们提出了GLEN，一个基于图的模型，能够在场景图序列上操作，既与文本动作对齐，又建模其时间演变。此外，我们提出了活动驱动的图编辑预测（A-GEF）问题，这是一项新任务，将场景动态视为基于正在进行的动作的结构化变换序列，从而实现对场景随时间变化的显式推理。我们在多个下游任务中验证了我们的方法，涵盖了检索基准如EgoMCQ和EgoCVR，以及长时间推理基准如EXPLORE-Bench和新引入的A-GEF。与原始视频基线相比，GLEN在推理设置中取得了良好的结果，通常这些设置仅通过MLLMs处理，同时实现了由人类活动驱动的场景动态的可控和结构化预测。我们相信我们的结果确立了时空场景图以及在其上进行推理的模型，作为视频理解及其潜在扩展的强大组合和可解释表示。
cs.CV / 116 / 2607.02435
MARVEL: Margin-Aware Robust von Mises-Fischer Expert Learning for Long-Tailed Out-of-Distribution Detection
MARVEL：基于边际感知的鲁棒 von Mises-Fischer 专家学习用于长尾分布外检测
Abstract
For clinical deployment, it is essential that automated diagnostic systems remain reliable when confronted with previously unseen cases, yet deep models routinely misclassify out-of-distribution (OOD) inputs with high confidence, underscoring the need for more robust OOD detection methods. Although substantial effort has been devoted to improving model robustness, most of the existing literature assumes balanced datasets, evaluates OOD detection on coarse or non-clinical OOD sources, or lacks comprehensive assessment across diverse OOD scenarios. To address the gaps, we propose a novel methodology trained on diverse and imbalanced medical datasets and evaluated across a clinically reflective OOD spectrum. Our framework comprises three key components: (1) a Nonlinear von Mises-Fisher (NvMF) classifier capable of learning non-linear decision boundaries, with theoretical proof of its asymptotic connection to cosine classifiers; (2) a multi-expert framework in which margin-aware NvMF classifiers specialise in different regions of label distribution to better handle imbalance; and (3) an outlier expert trained explicitly to distinguish inlier from outlier data, thereby strengthening OOD detection. Evaluation on RFMiD, ISIC2019, and NCTCRC datasets demonstrates consistent improvements over state-of-the-art methods, achieving mean FPR95 reductions of 8.45%, 13.02%, and 36.90% respectively. These gains are further supported by comprehensive ablations that validated the contributions of each component. This enables reliable identification of unfamiliar cases for deferral to clinicians, supporting safer AI-assisted diagnosis in real-world workflows. Our code is available at https://github.com/redboxup/MARVEL.
Chinese Translation
在临床应用中，自动诊断系统在面对以前未见过的病例时保持可靠性至关重要，然而深度模型在处理分布外（OOD）输入时常常以高置信度错误分类，这突显了对更鲁棒的OOD检测方法的需求。尽管已有大量努力致力于提高模型的鲁棒性，但现有文献大多假设数据集是平衡的，且在粗略或非临床的OOD来源上评估OOD检测，或缺乏对多样化OOD场景的全面评估。为了解决这些问题，我们提出了一种新方法，该方法在多样且不平衡的医学数据集上进行训练，并在临床反映的OOD范围内进行评估。我们的框架包含三个关键组件：（1）一种非线性 von Mises-Fisher (NvMF) 分类器，能够学习非线性决策边界，并理论证明其与余弦分类器的渐近联系；（2）一个多专家框架，其中边际感知的NvMF分类器专门处理标签分布的不同区域，以更好地应对不平衡；（3）一个专门训练的异常值专家，旨在明确区分内点和外点数据，从而增强OOD检测。在RFMiD、ISIC2019和NCTCRC数据集上的评估表明，相较于最先进的方法，我们的方法在平均FPR95上分别实现了8.45%、13.02%和36.90%的减少。这些改进得到了全面消融实验的支持，验证了每个组件的贡献。这使得能够可靠地识别不熟悉的病例，以便转交给临床医生，从而支持在现实工作流程中更安全的AI辅助诊断。我们的代码可在https://github.com/redboxup/MARVEL获取。
cs.CV / 117 / 2607.02461
OrbitQuant: Data-Agnostic Quantization for Image and Video Diffusion Transformers
OrbitQuant：图像和视频扩散变换器的数据无关量化
Abstract
Diffusion transformers (DiTs) achieve state-of-the-art image and video generation, but their multi-step sampling and growing parameter count make inference expensive. Post-training quantization (PTQ) is the natural remedy, yet DiT activations shift across timesteps, prompts, and guidance branches, forcing prior methods to re-fit calibration data for every new checkpoint or modality. We present OrbitQuant, a data-agnostic weight-activation quantizer that bypasses range estimation by quantizing in a normalized, rotated basis. In this basis, a randomized permuted block-Hadamard (RPBH) rotation concentrates each coordinate around one fixed, known marginal regardless of the input, so a single Lloyd-Max codebook serves all timesteps, prompts, and layers of a given input dimension. We extend the same quantizer to weight rows offline, absorbing the rotation into the weights so that it cancels inside each linear layer and only a forward rotation on the activations remains at runtime. The same recipe transfers from image to video with no per-modality tuning. Across FLUX.1, Z-Image-Turbo, Wan 2.1, and CogVideoX, it sets the state of the art for PTQ at several low-bit settings. It also pushes PTQ of image diffusion transformers to W2A4 with usable generation quality.
Chinese Translation
扩散变换器（DiTs）在图像和视频生成方面达到了最先进的水平，但其多步骤采样和不断增长的参数数量使得推理成本高昂。后训练量化（PTQ）是解决这一问题的自然方法，但DiT的激活在时间步、提示和引导分支之间发生变化，迫使先前的方法为每个新的检查点或模态重新拟合校准数据。我们提出了OrbitQuant，一种数据无关的权重-激活量化器，通过在归一化的旋转基中进行量化，绕过了范围估计。在这个基中，随机置换的块-哈达玛（RPBH）旋转将每个坐标集中在一个固定的已知边际周围，无论输入如何，因此一个单一的Lloyd-Max码本可以服务于给定输入维度的所有时间步、提示和层。我们将相同的量化器扩展到离线权重行，吸收旋转到权重中，以便在每个线性层内抵消，仅在运行时保留对激活的前向旋转。相同的方法可以无须针对每种模态进行调优地从图像转移到视频。在FLUX.1、Z-Image-Turbo、Wan 2.1和CogVideoX上，它在多个低比特设置下设定了PTQ的最先进水平。它还将图像扩散变换器的PTQ推向W2A4，并保持可用的生成质量。
cs.CV / 118 / 2607.02471
Interpretation-Oriented Cloud Removal via Observation-Anchored Residual Flow with Geo-Contextual Alignment
面向解释的云去除：基于观测锚定残差流与地理上下文对齐
Abstract
Cloud removal (CR) is essential for optical remote sensing, serving as a prerequisite for reliable downstream interpretation, such as semantic segmentation and change detection. However, existing CR approaches often prioritize visual realism while overlooking their impact on subsequent analytical tasks, leading to semantic drift and degraded downstream performance. To address this issue, we propose Geo-Anchored Cloud Removal (GACR), a unified framework that jointly ensures faithful reconstruction and robust interpretability. At its core, GACR incorporates Observation-Anchored Residual Flow (OAR-Flow), which reformulates CR as a physically grounded residual inversion process. By anchoring the generative trajectory to the cloudy observation rather than pure noise, OAR-Flow enables fast, stable, and faithful reconstruction. To further preserve semantic structures critical for downstream interpretation, GACR integrates Geo-Contextual Prior Alignment (GCPA) to constrain the reconstruction within a semantic manifold induced by a Vision Foundation Model (VFM). Consequently, GACR strictly maintains the spatial-semantic integrity of complex landscapes. Extensive experiments across six CR datasets and twelve downstream tasks demonstrate that GACR produces superior reconstruction quality while consistently improving downstream task accuracy. The code is available at https://github.com/wzy6055/GACR.
Chinese Translation
云去除（CR）对于光学遥感至关重要，是可靠下游解释（如语义分割和变化检测）的前提。然而，现有的云去除方法往往优先考虑视觉真实感，而忽视了它们对后续分析任务的影响，导致语义漂移和下游性能下降。为了解决这个问题，我们提出了地理锚定云去除（GACR），这是一个统一框架，旨在共同确保真实重建和稳健解释性。GACR的核心是观测锚定残差流（OAR-Flow），它将云去除重新表述为一个基于物理的残差反演过程。通过将生成轨迹锚定在多云观测上而非纯噪声，OAR-Flow实现了快速、稳定和真实的重建。为了进一步保留对下游解释至关重要的语义结构，GACR集成了地理上下文先验对齐（GCPA），以限制重建在由视觉基础模型（VFM）诱导的语义流形内。因此，GACR严格维护复杂景观的空间-语义完整性。在六个云去除数据集和十二个下游任务上的广泛实验表明，GACR在提供更优重建质量的同时，始终提高下游任务的准确性。代码可在 https://github.com/wzy6055/GACR 获取。
cs.CV / 119 / 2607.02479
EAGLE-360: Embodied Active Global-to-Local Exploration in 360$^\circ$
EAGLE-360：360$^ ext{°}$中具身主动的全局到局部探索
Abstract
While Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in standard visual understanding, adapting them for active visual search in 360$^\circ$ panoramic environments exposes fundamental limitations. Specifically, standard MLLMs struggle to effectively model inherent panoramic properties, such as severe polar distortion and continuous cylindrical topologies, which significantly degrades target detection accuracy. Consequently, existing panoramic search methods attempt to compensate by relying heavily on fragmented local viewpoints. Burdened by rigid initialization and a lack of global panoramic priors, these approaches suffer from myopic, inefficient exploration and struggle with robust error recovery when targets are out of view. To overcome these challenges, we propose EAGLE-360, a novel Embodied Active Global-to-Local Exploration framework. Rather than performing exhaustive local searches, EAGLE-360 leverages global priors to establish an initial holistic perspective, iteratively reasoning and progressively narrowing the search space. Architecturally, we adapt RoPE Rolling, a coordinate-shifting positional encoding mechanism, to seamlessly model the continuous topologies of panoramas. To facilitate this paradigm, we construct the large-scale EAGLE-360 dataset, comprising 14,000+ 4K panoramas and 70,000+ rounds of high-quality VQA dialogues. By employing a training pipeline that integrates Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO), we effectively elicit complex spatial reasoning and tool-calling capabilities. Extensive experiments demonstrate that EAGLE-360 establishes a new state-of-the-art for 360$^\circ$ visual search, achieving nearly an 8-fold increase in accuracy over the base model while significantly enhancing exploration efficiency.
Chinese Translation
尽管多模态大型语言模型（MLLMs）在标准视觉理解方面表现出色，但将其适应于360$^ ext{°}$全景环境中的主动视觉搜索却暴露出基本的局限性。具体而言，标准的MLLMs难以有效建模固有的全景特性，如严重的极地失真和连续的圆柱形拓扑，这显著降低了目标检测的准确性。因此，现有的全景搜索方法试图通过严重依赖碎片化的局部视点来进行补偿。由于受到刚性初始化和缺乏全局全景先验的困扰，这些方法在探索时往往目光短浅且效率低下，并且在目标超出视野时难以进行稳健的错误恢复。为了解决这些挑战，我们提出了EAGLE-360，一个新颖的具身主动全局到局部探索框架。EAGLE-360并不进行详尽的局部搜索，而是利用全局先验建立初始的整体视角，迭代推理并逐步缩小搜索空间。在架构上，我们调整了RoPE Rolling，一种坐标转换的位置编码机制，以无缝建模全景的连续拓扑。为了促进这一范式，我们构建了大规模的EAGLE-360数据集，包含14,000多个4K全景图和70,000多轮高质量的视觉问答对话。通过采用将监督微调（SFT）与组相对策略优化（GRPO）相结合的训练流程，我们有效引发了复杂的空间推理和工具调用能力。大量实验表明，EAGLE-360在360$^ ext{°}$视觉搜索中建立了新的最先进水平，相较于基础模型，准确性几乎提高了8倍，同时显著增强了探索效率。
cs.CV / 120 / 2607.02484
Combating Textual Noise and Redundancy: Entropy-Aware Dense Visual Token Pruning
应对文本噪声和冗余：基于熵的稠密视觉标记剪枝
Abstract
Visual token pruning is a crucial strategy for accelerating VLMs by compressing redundant image patches, yet existing methods often fail to preserve critical cues under dense instructions and fine-grained queries. In this paper, we investigate this failure and identify two underlying bottlenecks: the widespread dispersion of textual noise that corrupts dense cross-modal scoring, and the feature fragmentation inherent to standard token selection. To address these issues, we propose Entropy-Aware Dense Pruning (EADP), a framework that reformulates pruning as a structured compression problem. EADP first leverages statistical entropy to quantify and filter out textual noise, yielding a robust, fine-grained instruction relevance score. Subsequently, instead of naive Top-K selection, EADP casts token selection as a submodular maximization problem with a spatial prior, explicitly ensuring a holistic and non-redundant visual representation. Extensive experiments demonstrate that EADP improves the accuracy-efficiency trade-off of VLMs, robustly preserving fine-grained visual cues under strict token budgets while achieving SoTA performance on challenging multimodal benchmarks.
Chinese Translation
视觉标记剪枝是通过压缩冗余图像块来加速视觉语言模型（VLMs）的关键策略，但现有方法往往未能在密集指令和细粒度查询下保留关键线索。本文探讨了这一失败，并识别出两个潜在瓶颈：广泛存在的文本噪声分散，破坏了密集的跨模态评分，以及标准标记选择固有的特征碎片化。为了解决这些问题，我们提出了基于熵的稠密剪枝（Entropy-Aware Dense Pruning, EADP），一个将剪枝重新构建为结构化压缩问题的框架。EADP首先利用统计熵来量化和过滤文本噪声，从而产生一个稳健的细粒度指令相关性评分。随后，EADP将标记选择视为一个具有空间先验的子模最大化问题，而不是简单的Top-K选择，明确确保了整体和非冗余的视觉表示。大量实验表明，EADP改善了VLMs的准确性与效率的权衡，在严格的标记预算下稳健地保留细粒度视觉线索，同时在具有挑战性的多模态基准测试中实现了最先进的性能。
cs.CV / 121 / 2607.02486
GeoMix: Descriptor-Free Visual Localization via Global Context and Multi-Detector Training
GeoMix：通过全球上下文和多检测器训练实现无描述符视觉定位
Abstract
Descriptor-free visual localization eliminates high-dimensional descriptor storage, preserves scene privacy, and simplifies map maintenance, yet its accuracy still lags far behind descriptor-based pipelines. We identify this gap to insufficient geometric discriminability in geometry-only matching. Without visual appearance, current methods underutilize local geometry cues, lack the global context among keypoints, and overfit to a single keypoint detector. We further observe that descriptor-free matching naturally enables multi-detector training, as heterogeneous keypoints can be optimized in a shared geometry-only space without aligning descriptor spaces. Building on these insights, we propose GeoMix, a descriptor-free 2D-3D matching framework that strengthens geometric discriminability at three levels. Locally, directional and distance-aware embeddings enrich neighborhood aggregation with fine-grained spatial structure. Globally, learnable context nodes aggregate and redistribute scene-wide information via cross-attention to resolve ambiguities beyond local receptive fields. At the training level, Mix-Training exploits this detector-agnostic geometry space to learn representations across multiple keypoint detectors. Extensive experiments on MegaDepth, Cambridge Landmarks, 7Scenes, and Aachen Day-Night show that GeoMix sets a new state of the art among descriptor-free methods, reducing 75th-percentile rotation error by 89\% and translation error by up to 90\% over the previous best, while generalizing zero-shot to unseen detectors and narrowing the gap to descriptor-based pipelines. Code is available at $\href{https://github.com/YejunZhang/Geomix}{\text{this links}}$.
Chinese Translation
无描述符视觉定位消除了高维描述符存储，保护了场景隐私，并简化了地图维护，但其准确性仍远远落后于基于描述符的管道。我们发现这一差距源于仅依赖几何匹配的几何可区分性不足。在没有视觉外观的情况下，当前方法未能充分利用局部几何线索，缺乏关键点之间的全局上下文，并且过度拟合于单一关键点检测器。我们进一步观察到，无描述符匹配自然支持多检测器训练，因为异构关键点可以在共享的仅几何空间中优化，而无需对齐描述符空间。基于这些见解，我们提出了GeoMix，这是一种无描述符的2D-3D匹配框架，在三个层面上增强几何可区分性。在局部层面，方向和距离感知嵌入通过细粒度空间结构丰富邻域聚合。在全局层面，可学习的上下文节点通过交叉注意力聚合和重新分配全场景信息，以解决超出局部感受野的歧义。在训练层面，Mix-Training利用这一与检测器无关的几何空间，在多个关键点检测器之间学习表示。在MegaDepth、Cambridge Landmarks、7Scenes和Aachen Day-Night上的大量实验表明，GeoMix在无描述符方法中设定了新的最先进水平，将第75百分位旋转误差降低了89 ext{ extperthousand}，平移误差降低了高达90 ext{ extperthousand}，同时在零样本情况下对未见检测器进行泛化，并缩小了与基于描述符的管道之间的差距。代码可在$ ext{this links} ext{https://github.com/YejunZhang/Geomix}$获取。
cs.CV / 122 / 2607.02494
Towards Robustness against Typographic Attack with Training-free Concept Localization
针对排版攻击的鲁棒性：无训练概念定位方法
Abstract
Models trained via Contrastive Language-Image Pretraining (CLIP) serve as the foundational vision encoders for most modern Large Vision Language Models (LVLMs). Despite their widespread adoption, CLIP models exhibit a critical yet underexplored failure mode: irrelevant text appearing within images confounds visual representations, biasing them toward lexical meaning rather than true visual semantics. This robustness issue, commonly described as a Typographic Attack (TA), exposes a vulnerability that poses a significant risk to safety-critical applications such as autonomous driving. To achieve interpretable and effective robustness against TA, we propose a novel, training-free mechanistic interpretability method. Our method provides sampling-based interpretations of hidden state representations and quantitatively attributes semantic versus lexical focus to individual attention heads. Through probabilistic analysis and circuit mining, we isolate specific Vision Transformer (ViT) components that disproportionately encode lexical information, thereby identifying the mechanistic source of TA. We further show that simple interventions applied directly to the identified circuits, without any additional training, can substantially improve robustness against Typographic Attacks in object classification. These interventions, such as selective adjustment of attention weights, also outperform both supervised and training-free defense methods. Our experiments demonstrate that applying the proposed intervention to the vision encoders of several state-of-the-art LVLMs yields substantial gains in Visual Question Answering accuracy under Typographic Attack interference on RIO-Bench. These results confirm both the efficacy and the generalizability of our mechanistic approach. Code is released at https://github.com/Liu-524/SamplingTAR.
Chinese Translation
通过对比语言-图像预训练（Contrastive Language-Image Pretraining, CLIP）训练的模型作为大多数现代大型视觉语言模型（Large Vision Language Models, LVLMs）的基础视觉编码器。尽管它们被广泛采用，CLIP 模型却表现出一种关键但未被充分探索的失败模式：图像中出现的无关文本混淆了视觉表征，使其偏向于词汇意义而非真实的视觉语义。这种鲁棒性问题通常被描述为排版攻击（Typographic Attack, TA），暴露出一种脆弱性，对自动驾驶等安全关键应用构成了重大风险。为了实现对 TA 的可解释和有效的鲁棒性，我们提出了一种新颖的无训练机制可解释性方法。我们的方法提供了基于采样的隐藏状态表征的解释，并定量地将语义与词汇的关注点归因于各个注意力头。通过概率分析和电路挖掘，我们隔离了特定的视觉变换器（Vision Transformer, ViT）组件，这些组件不成比例地编码了词汇信息，从而识别出 TA 的机制来源。我们进一步表明，直接对识别出的电路施加简单干预，而无需任何额外训练，可以显著提高物体分类中对排版攻击的鲁棒性。这些干预措施，如选择性调整注意力权重，的表现也优于监督和无训练防御方法。我们的实验表明，将所提干预应用于多个最先进的 LVLM 的视觉编码器，在 RIO-Bench 上的排版攻击干扰下，视觉问答的准确性显著提高。这些结果确认了我们机制方法的有效性和普遍适用性。代码已发布在 https://github.com/Liu-524/SamplingTAR。
cs.CV / 123 / 2607.02497
Seek to Segment: Active Perception for Panoramic Referring Segmentation
寻求分割：全景指称分割的主动感知
Abstract
Existing referring segmentation models passively process static images captured from fixed perspectives, limiting their applicability in Embodied AI, where agents must perform active perception in the continuous 360$^\circ$ environments. To bridge this gap, we introduce a novel task: Active Panoramic Referring Segmentation (APRS). In this setting, an agent is required to adjust its viewing direction ($\Delta\theta, \Delta\phi$) to explore the 360$^\circ$ environment, seeking the object specified by a user instruction for segmentation. To tackle this challenging task, we propose PanoSeeker, a memory-augmented agent for efficient APRS. Rather than relying on heuristic scanning, PanoSeeker integrates a Vision-Language Model (VLM) with EgoSphere, an explicit spatial visual memory. By progressively integrating sequential local observations into a unified 360$^\circ$ representation, EgoSphere enables the agent to plan efficient and non-redundant search trajectories. Once the target is found, the agent performs active viewpoint alignment and outputs the segmentation mask. Furthermore, we curate an expert-annotated search trajectory dataset with memory timelines for Supervised Fine-Tuning, followed by Reinforcement Learning post-training to explicitly optimize PanoSeeker's exploration efficiency. Extensive experiments on our newly established APRS benchmark demonstrate that PanoSeeker achieves superior search efficiency and segmentation accuracy, significantly outperforming adapted state-of-the-art baselines.
Chinese Translation
现有的指称分割模型被动地处理从固定视角捕获的静态图像，这限制了它们在具身人工智能（Embodied AI）中的适用性，因为在连续的360$^ ext{°}$环境中，智能体必须进行主动感知。为了解决这一问题，我们引入了一项新任务：主动全景指称分割（Active Panoramic Referring Segmentation, APRS）。在这一设置中，智能体需要调整其视角方向（$ riangle heta, riangle heta$）以探索360$^ ext{°}$环境，寻找用户指令指定的对象进行分割。为了解决这一挑战性任务，我们提出了PanoSeeker，一种用于高效APRS的增强记忆智能体。PanoSeeker并不依赖于启发式扫描，而是将视觉-语言模型（Vision-Language Model, VLM）与EgoSphere（一个显式的空间视觉记忆）相结合。通过逐步将连续的局部观察整合为统一的360$^ ext{°}$表示，EgoSphere使智能体能够规划高效且不冗余的搜索轨迹。一旦找到目标，智能体将进行主动视点对齐并输出分割掩膜。此外，我们整理了一个专家注释的搜索轨迹数据集，并附有记忆时间线，以用于监督微调（Supervised Fine-Tuning），随后进行强化学习后训练，以明确优化PanoSeeker的探索效率。在我们新建立的APRS基准上进行的广泛实验表明，PanoSeeker在搜索效率和分割准确性方面表现优越，显著超越了适应性最先进基线的表现。
cs.CV / 124 / 2607.02508
From SRA to Self-Flow: Data Augmentation or Self-Supervision?
从 SRA 到 Self-Flow：数据增强还是自我监督？
Abstract
Representation alignment has become an effective way to accelerate diffusion transformer training and improve generation quality. Recent self-alignment methods, such as SRA and Self-Flow, further remove the dependency on external pretrained encoders by constructing alignment within the diffusion model itself. However, the mechanism behind the improvement from SRA to Self-Flow, dual-time scheduling, remains under-examined: Self-Flow attributes its gain to interactions between tokens at different noise levels, where cleaner tokens help infer noisier ones. In this work, we revisit this explanation and ask whether the gain instead comes from data augmentation along the noise dimension. To disentangle these factors, we introduce Attention Separation, which preserves the same dual-timestep input as Self-Flow while blocking attention between tokens assigned to different noise levels. Surprisingly, removing such interaction does not degrade performance and can even improve it, suggesting that the improvement from SRA to Self-Flow mainly comes from data augmentation. Furthermore,We show that Attention Separation itself provides an augmentation effect by splitting a single image into multiple effective training parts to expand the training data. Based on these observations, we combine self-representation alignment with dual-timestep and attention-separation augmentation, and demonstrate the effectiveness of this design on ImageNet.
Chinese Translation
表示对齐已成为加速扩散变换器训练和提高生成质量的有效方法。最近的自我对齐方法，如 SRA 和 Self-Flow，进一步通过在扩散模型内部构建对齐，消除了对外部预训练编码器的依赖。然而，从 SRA 到 Self-Flow 的改进机制，即双时间调度，仍然未得到充分研究：Self-Flow 将其增益归因于不同噪声水平下标记之间的交互，其中更干净的标记有助于推断更嘈杂的标记。在本研究中，我们重新审视这一解释，并询问增益是否实际上来自噪声维度上的数据增强。为了理清这些因素，我们引入了注意力分离（Attention Separation），该方法保留与 Self-Flow 相同的双时间步输入，同时阻止分配给不同噪声水平的标记之间的注意力交互。令人惊讶的是，去除这种交互并没有降低性能，甚至可以提高性能，这表明从 SRA 到 Self-Flow 的改进主要来自数据增强。此外，我们还展示了注意力分离本身通过将单个图像拆分为多个有效训练部分来扩展训练数据，从而提供了一种增强效果。基于这些观察，我们将自我表示对齐与双时间步和注意力分离增强相结合，并展示了该设计在 ImageNet 上的有效性。
cs.CV / 125 / 2607.02515
PointDiT: Pixel-Space Diffusion for Monocular Geometry Estimation
PointDiT：用于单目几何估计的像素空间扩散
Abstract
State-of-the-art single-image 3D reconstruction methods often rely on complex hybrid architectures and loss functions, or compress geometry into latent spaces in order to leverage pre-trained latent diffusion models. In this work, we show that such architectural overhead and intricate loss formulations are unnecessary. We introduce a minimalist pixel-space Diffusion Transformer, built on a plain ViT, that operates directly on raw 3D point map patches and is conditioned on image tokens from a pre-trained DINOv3. Unlike existing latent diffusion approaches, we train our diffusion backbone entirely from scratch, eliminating the need for point map tokenizers. Despite its simplicity, our approach surpasses complex latent-based diffusion models while remaining significantly simpler than hybrid alternatives. Notably, it produces sharper geometric structure and is more robust in highly ambiguous regions, such as transparent objects.
Chinese Translation
最先进的单幅图像3D重建方法通常依赖于复杂的混合架构和损失函数，或者将几何信息压缩到潜在空间中，以利用预训练的潜在扩散模型。在本研究中，我们展示了这种架构开销和复杂损失公式是多余的。我们引入了一种极简的像素空间扩散变换器（Diffusion Transformer），该变换器基于普通的视觉变换器（ViT），直接对原始3D点图块进行操作，并以预训练的DINOv3中的图像标记为条件。与现有的潜在扩散方法不同，我们的扩散骨干网络完全从头开始训练，消除了对点图标记器的需求。尽管我们的方案简单，但其性能超过了复杂的基于潜在的扩散模型，同时仍然显著简化了混合替代方案。值得注意的是，它在高度模糊的区域（如透明物体）中产生了更清晰的几何结构，并且更加稳健。
cs.CV / 126 / 2607.02516
Alignment Is All You Need For X-to-4D Generation
对齐是实现 X 到 4D 生成的关键
Abstract
Generative diffusion models excel at synthesizing high-quality images, videos, and 3D content under multimodal control. However, arbitrary user-defined modality-to-4D (X-to-4D) generation remains challenging due to the high cost of constructing diverse datasets and the limited scalability of existing methods. This paper presents Align4D, a flexible framework that translates any-modal input into coherent video-3D pairs, using video to guide 4D motion and 3D data to shape 4D geometry. Align4D introduces three key techniques: (1) Object Distance Alignment, which searches Video-Aligned and Multiview-Aligned Object Distances (VAOD/MAOD), respectively, to reconcile 4D renderings with video and the priors of multiview diffusion models; (2) Motion-Geometry Joint Alignment, which constrains known and unknown views through synchronized video and 3D inputs, ensuring consistent 4D generation; and (3) Asynchronous Optimization, which decouples Gaussian attribute and deformation network training to enhance motion and geometry fidelity. We further propose the X4D dataset, which integrates prompt, image, video, and 3D data for benchmarking. Experiments on X4D and Consistent4D demonstrate that Align4D achieves state-of-the-art quality and consistency in X-to-4D generation. Project page: https://miaoqiaowei.github.io/Align4D/.
Chinese Translation
生成扩散模型在多模态控制下合成高质量图像、视频和 3D 内容方面表现出色。然而，由于构建多样化数据集的高成本以及现有方法的有限可扩展性，任意用户定义的模态到 4D（X-to-4D）生成仍然具有挑战性。本文提出了 Align4D，这是一个灵活的框架，可以将任何模态输入转换为一致的视频-3D 对，利用视频引导 4D 运动，并使用 3D 数据塑造 4D 几何。Align4D 引入了三项关键技术：（1）对象距离对齐，分别搜索视频对齐和多视图对齐对象距离（VAOD/MAOD），以调和 4D 渲染与视频及多视图扩散模型的先验；（2）运动-几何联合对齐，通过同步视频和 3D 输入约束已知和未知视图，确保一致的 4D 生成；（3）异步优化，解耦高斯属性和变形网络训练，以增强运动和几何的保真度。我们进一步提出了 X4D 数据集，该数据集整合了提示、图像、视频和 3D 数据用于基准测试。在 X4D 和 Consistent4D 上的实验表明，Align4D 在 X-to-4D 生成中实现了最先进的质量和一致性。项目页面：https://miaoqiaowei.github.io/Align4D/
cs.CV / 127 / 2607.02517
WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory
WorldDirector：构建具有持久动态记忆的可控世界模拟器
Abstract
We present WorldDirector, a highly controllable video world model framework designed for persistent dynamic object memory and unrestricted viewpoint exploration. Unlike existing world models that entangle physical dynamics with pixel rendering and rely on continuous visual observation to sustain motion, our framework explicitly decouples semantic motion orchestration from visual generation. By leveraging an LLM to coordinate 3D trajectories with camera movements and subsequently employing these orchestrated trajectories as control signals for video generation, our approach ensures strict physical logic and appearance stability, successfully preserving the exact visual identities of dynamic entities even when they re-enter the scene after prolonged periods out of view. Experimental results demonstrate that our method supports the synthesis of complex and extended events with unprecedented controllability and persistent dynamic object memory. Project Page: https://worlddirector.github.io/
Chinese Translation
我们提出了WorldDirector，这是一个高度可控的视频世界模型框架，旨在实现持久的动态物体记忆和无限制的视角探索。与现有的世界模型不同，后者将物理动态与像素渲染纠缠在一起，并依赖于连续的视觉观察来维持运动，我们的框架明确地将语义运动编排与视觉生成解耦。通过利用大型语言模型（LLM）来协调三维轨迹与相机运动，并随后将这些编排的轨迹作为视频生成的控制信号，我们的方法确保了严格的物理逻辑和外观稳定性，成功地保留了动态实体的确切视觉身份，即使在它们经过长时间的视野缺失后重新进入场景。实验结果表明，我们的方法支持以空前的可控性和持久的动态物体记忆合成复杂和扩展的事件。项目页面：https://worlddirector.github.io/
cs.AI / 1 / 2607.01306
PACE: A Neuro-Symbolic Framework for Plausible and Actionable Counterfactual Explanations
PACE：一种用于合理且可操作的反事实解释的神经符号框架
Abstract
Counterfactual explanations explain machine learning predictions by identifying minimal input changes that would alter a model's decision. Although many existing methods successfully generate prediction-changing alternatives, they often produce unrealistic or infeasible recommendations due to a lack of explicit mechanisms for incorporating domain knowledge and intervention constraints. Neuro-symbolic AI offers a promising direction by combining data-driven predictive models with symbolic reasoning capable of representing human-understandable rules and feasible actions. This paper presents PACE, a modular neuro-symbolic framework for generating feasibility-aware counterfactual explanations. The framework separates prediction and reasoning into two components: a neural predictive model for classification and a symbolic reasoning layer that enforces domain-specific constraints during counterfactual generation. By explicitly modeling feasible interventions, the framework produces explanations consistent with domain knowledge while remaining interpretable and actionable. The approach is model-agnostic and adaptable to domains requiring realistic decision support. A case study is conducted on the Adult Income dataset, combining a multilayer perceptron classifier with Answer Set Programming (ASP) rules encoding feasible modifications to education, occupation, and working hours while preserving immutable attributes. Results highlight the trade-off between counterfactual validity and plausibility and show that symbolic constraints yield explanations that better satisfy domain-specific feasibility requirements, illustrating the potential of neuro-symbolic methods for transparent, feasibility-aware counterfactual explanation in explainable AI.
Chinese Translation
反事实解释通过识别最小输入变化来解释机器学习预测，这些变化将改变模型的决策。尽管许多现有方法成功生成改变预测的替代方案，但由于缺乏将领域知识和干预约束明确纳入的机制，它们往往会产生不现实或不可行的建议。神经符号人工智能通过将数据驱动的预测模型与能够表示人类可理解规则和可行行动的符号推理相结合，提供了一个有前景的方向。本文提出了PACE，一个模块化的神经符号框架，用于生成考虑可行性的反事实解释。该框架将预测和推理分为两个组件：一个用于分类的神经预测模型和一个在反事实生成过程中强加领域特定约束的符号推理层。通过明确建模可行的干预，该框架生成与领域知识一致的解释，同时保持可解释性和可操作性。该方法与模型无关，并且可适应需要现实决策支持的领域。我们在成人收入数据集上进行了案例研究，将多层感知器分类器与编码教育、职业和工作时间可行修改的答案集编程（ASP）规则相结合，同时保留不可变属性。结果突出了反事实有效性与合理性之间的权衡，并表明符号约束产生的解释更好地满足领域特定的可行性要求，展示了神经符号方法在可解释人工智能中实现透明、考虑可行性的反事实解释的潜力。
cs.AI / 2 / 2607.01366
Auto-FL-Research: Agentic Search for Federated Learning Algorithms
Auto-FL-Research：联邦学习算法的自主搜索
Abstract
Federated learning (FL) research often depends on many small but consequential algorithmic choices: optimizer variants, server aggregation rules, local training schedules, normalization, regularization, and model architecture. These choices are expensive to explore manually and difficult to compare fairly when candidate changes can also alter the FL training or evaluation path. In this work, we present Auto-FL-Research (AFR), a constrained coding-agent workflow for FL algorithmic recipe search. Agents may propose and implement candidate training algorithms, including server aggregation rules, client update schedules, local objectives, and registered model variants, while task profiles fix the mutation surface, compute budget, communication contract, and final model evaluation. Each campaign records candidate scores, runtime, edited files, artifacts, and failure status. We evaluate AFR on five healthcare cross-silo FLamby tasks and on grouped-client profiles for the five fixed LEAF datasets plus the LEAF synthetic task. Five-seed repeat evaluations support gains on four FLamby tasks and five of six LEAF profiles, while also exposing seed-sensitive and search-selected failure cases. Same-budget controls show that several gains correspond to FL-recipe changes, whereas other improvements are recovered by fixed-surface scalar controls or fail under repeat or held-out evaluation. These mixed outcomes are part of the contribution: they show how agent-generated candidates can be separated into repeated FL mechanisms, fixed-surface tuning effects, and selected single-run artifacts.
Chinese Translation
联邦学习（FL）研究通常依赖于许多小但重要的算法选择：优化器变体、服务器聚合规则、本地训练计划、归一化、正则化和模型架构。这些选择手动探索的成本高昂，并且在候选变化可能改变FL训练或评估路径时，公平比较变得困难。在本研究中，我们提出了Auto-FL-Research（AFR），这是一个针对FL算法配方搜索的受限编码代理工作流程。代理可以提出并实施候选训练算法，包括服务器聚合规则、客户端更新计划、本地目标和注册模型变体，同时任务配置固定了变异表面、计算预算、通信合同和最终模型评估。每个实验记录候选分数、运行时间、编辑文件、工件和失败状态。我们在五个医疗保健跨孤岛FLamby任务和五个固定LEAF数据集的分组客户端配置以及LEAF合成任务上评估了AFR。五次种子重复评估支持在四个FLamby任务和六个LEAF配置中的五个上获得收益，同时也揭示了种子敏感和搜索选择的失败案例。同样预算的对照显示，多个收益与FL配方变化相关，而其他改进则通过固定表面标量控制恢复，或在重复或保留评估中失败。这些混合结果是贡献的一部分：它们展示了代理生成的候选如何被分离为重复的FL机制、固定表面调优效应和选定的单次运行工件。
cs.AI / 3 / 2607.01394
The Wiola Architecture for Efficient Small Language Models
高效小型语言模型的Wiola架构
Abstract
We present Wiola, a fully original Small Language Model (SLM) architecture built from first principles, sharing no structural lineage with any existing model family including GPT, LLaMA, Mistral, or Falcon. Wiola introduces five independently novel components: (i) Spiral Rotary Positional Encoding (SRPE), which embeds token positions on a three-dimensional helical manifold combining absolute, relative, and hierarchical positional signals; (ii) Gated Cross-Layer Attention (GCLA), providing each decoder layer with soft cross-attention access to compressed summaries of two preceding layers for inter-layer coherence; (iii) Adaptive Token Merging (ATM), which dynamically merges se mantically redundant adjacent tokens in middle network layers to reduce attention complexity without information loss; (iv) Dual Stream Feed-Forward (DSFF), replacing the conventional MLP with two parallel streams fused by a learned per-dimension gate; and (v) WiolaRMSNorm, a modified normalisation introducing a per-dimension learned offset vector that prevents representation collapse. We provide complete mathematical derivations, architectural block diagrams, complexity analyses, and systematic comparisons against GPT-2, LLaMA-2, and Mistral. Wiola is released in four sizes (120M, 360M, 700M, and 1.5B parameters) and is fully compatible with the HuggingFace Transformers ecosystem, with all 22 architectural unit tests passing.
Chinese Translation
我们提出了Wiola，一种完全原创的小型语言模型（Small Language Model, SLM）架构，基于第一原理构建，与包括GPT、LLaMA、Mistral或Falcon在内的任何现有模型家族没有结构上的继承关系。Wiola引入了五个独立的新组件：(i) 螺旋旋转位置编码（Spiral Rotary Positional Encoding, SRPE），它在一个三维螺旋流形上嵌入令牌位置，结合绝对、相对和层次位置信号；(ii) 门控跨层注意力（Gated Cross-Layer Attention, GCLA），为每个解码器层提供对前两个层压缩摘要的软跨注意力访问，以增强层间一致性；(iii) 自适应令牌合并（Adaptive Token Merging, ATM），动态合并中间网络层中语义冗余的相邻令牌，以减少注意力复杂性而不损失信息；(iv) 双流前馈（Dual Stream Feed-Forward, DSFF），用两个并行流替代传统的多层感知机（MLP），通过学习的每维度门进行融合；(v) WiolaRMSNorm，一种修改后的归一化方法，引入每维度学习的偏移向量，防止表示崩溃。我们提供了完整的数学推导、架构框图、复杂性分析以及与GPT-2、LLaMA-2和Mistral的系统比较。Wiola以四种规模发布（120M、360M、700M和1.5B参数），并与HuggingFace Transformers生态系统完全兼容，所有22个架构单元测试均通过。
cs.AI / 4 / 2607.01425
Agent4cs: A Multi-agent System for Code Summarization in Large Hierarchical Codebases
Agent4cs：一个用于大型层次代码库的代码摘要多代理系统
Abstract
Understanding large, complex codebases, especially those with obfuscated structures and incomplete documentation, remains a significant challenge. Existing code summarization solutions often rely on a single language model or coding assistant like Claude Code, and treat source code as flat text, underutilizing the rich interdependencies and hierarchical information within a repository. To address these shortcomings, we propose Agent4cs - a multi-agent framework that summarizes large codebases in a bottom-up fashion, where a summarization agent focuses on producing robust summaries; a keyword-extraction agent proactively identifies critical information from subfolders; and a quality-assurance agent iteratively refines the outputs for readability, coherence, and completeness. Evaluated on 7 frontier models, Agent4cs improves semantic consistency across all folder levels by average 8% compared to two structured prompting baselines with code segments. Furthermore, extensive evaluation on real-world datasets demonstrates up to 38% gains in normalized keyword coverage rate over the same baselines.
Chinese Translation
理解大型复杂代码库，尤其是那些结构混淆和文档不完整的代码库，仍然是一个重大挑战。现有的代码摘要解决方案通常依赖于单一的语言模型或编码助手，如Claude Code，并将源代码视为平面文本，未能充分利用代码库内丰富的相互依赖关系和层次信息。为了解决这些不足，我们提出了Agent4cs——一个多代理框架，以自下而上的方式总结大型代码库，其中摘要代理专注于生成稳健的摘要；关键词提取代理主动识别子文件夹中的关键信息；质量保证代理则迭代地优化输出，以提高可读性、一致性和完整性。在对7个前沿模型的评估中，Agent4cs在所有文件夹层级上相比于两个结构化提示基线在语义一致性上平均提高了8%。此外，在真实世界数据集上的广泛评估显示，相比于相同的基线，规范化关键词覆盖率提高了多达38%。
cs.AI / 5 / 2607.01426
When Should Service Agents Reconsider? Difficulty-Routed Control in Customer-Service Operations
服务代理何时应重新考虑？客户服务操作中的难度导向控制
Abstract
Autonomous customer-service agents are shifting from conversational interfaces toward operational execution roles: they retrieve firm records, apply service policies, and execute backend writes such as refunds, cancellations, exchanges, order modifications, and reservation changes. This shift creates a service-control problem: firms must keep routine service fast and low-friction while preventing operational errors on requests where customer instructions, policy constraints, firm records, and backend writes interact. We propose a difficulty-routed service-control architecture that asks when service agents should reconsider before acting. A lightweight router keeps routine sessions on a low-cost baseline path and routes operationally coupled sessions to an escalated workflow. The escalated path uses conflict-aware communication and write-triggered reconsideration to concentrate deliberation and safeguards before consequential backend writes, rather than applying additional control uniformly across all service sessions. We evaluate the architecture on human-verified retail and airline tasks from $\tau^{2}$-bench. In retail, the method improves reliability consistently on service requests with operational conflict. Routing evidence shows that stronger control is directed toward conflicted requests rather than broadly applied to routine ones. Dialogue and tool-use profiles suggest that gains do not come from indiscriminate interaction expansion or broader tool chains; instead, added turns and tool calls support evidence gathering, write separation, and pre-write reconsideration. Case-level evidence shows that the escalated workflow preserves fallback plans, binds retrieved records to the correct action, sequences writes, and decomposes multi-entity requests. Airline results extend the same service-control logic to reservation operations.
Chinese Translation
自主客户服务代理正在从对话界面转向操作执行角色：它们检索公司记录，应用服务政策，并执行后端操作，如退款、取消、交换、订单修改和预订变更。这一转变带来了服务控制问题：公司必须在保持常规服务快速和低摩擦的同时，防止在客户指令、政策限制、公司记录和后端操作之间相互作用的请求中出现操作错误。我们提出了一种难度导向的服务控制架构，旨在确定服务代理在行动前何时应重新考虑。一种轻量级路由器将常规会话保持在低成本的基线路径上，并将操作相关的会话路由到升级的工作流程。升级路径使用冲突感知通信和写入触发的重新考虑，以集中审议和保障措施在重要的后端写入之前，而不是在所有服务会话中均匀地应用额外控制。我们在经过人工验证的零售和航空任务上评估了该架构，数据来自$ au^{2}$-bench。在零售中，该方法在具有操作冲突的服务请求上始终提高了可靠性。路由证据表明，较强的控制被指向冲突请求，而不是广泛应用于常规请求。对话和工具使用的分析表明，收益并非来自于无差别的交互扩展或更广泛的工具链；相反，增加的回合和工具调用支持证据收集、写入分离和写入前重新考虑。案例级证据表明，升级的工作流程保留了后备计划，将检索的记录绑定到正确的操作，序列化写入，并分解多实体请求。航空结果将相同的服务控制逻辑扩展到预订操作。
cs.AI / 6 / 2607.01433
CreativityNeuro: Steering Language Model Weights to Improve Divergent Thinking and Reduce Mode Collapse
CreativityNeuro：引导语言模型权重以改善发散性思维并减少模式崩溃
Abstract
Divergent thinking is a crucial aspect of creativity, yet large language models (LLMs) tend to consistently generate similar responses to open-ended questions, in what has been termed the artificial hivemind effect. Here, we introduce CreativityNeuro, a data-free method for enhancing divergent thinking in LLMs via contrastive weight steering. We evaluate our method across multiple creativity assessments and report several main findings. On the Divergent Association Task (DAT), a vocabulary-space creativity test, CreativityNeuro improves performance by up to 14 human percentile points. Next, in a large-scale human evaluation (N=720) on the Alternative Uses Test (AUT) and the Task Task, CreativityNeuro achieves significant improvements in originality, surprise, and creativity, transferring to longer-form and more open-ended tasks. Importantly, we find that across all three tasks, CreativityNeuro demonstrably reduces measures of mode collapse. Moreover, activation steering achieves comparable performance to CreativityNeuro on the DAT, but it does not transfer to the AUT and Task Task, demonstrating the effectiveness of weight-space steering in generalizing to unseen tasks. In conclusion, CreativityNeuro improves divergent thinking and reduces mode collapse without requiring behavioral data, re-training, or gradient-based fine-tuning, providing a straightforward way to enhance LLM performance in creative domains.
Chinese Translation
发散性思维是创造力的重要方面，但大型语言模型（LLMs）往往对开放性问题生成相似的回答，这种现象被称为人工蜂群效应。在此，我们介绍了CreativityNeuro，这是一种无数据的方法，通过对比权重引导来增强LLMs中的发散性思维。我们在多个创造力评估中评估了我们的方法，并报告了几个主要发现。在发散联想任务（Divergent Association Task, DAT）这一词汇空间创造力测试中，CreativityNeuro的表现提高了多达14个百分位点。接下来，在一项大规模的人类评估（N=720）中，对替代用途测试（Alternative Uses Test, AUT）和任务任务（Task Task）进行评估，CreativityNeuro在原创性、惊讶感和创造力方面取得了显著改善，并且这种改善延续到更长形式和更开放性的问题中。重要的是，我们发现CreativityNeuro在所有三个任务中显著减少了模式崩溃的指标。此外，激活引导在DAT上达到了与CreativityNeuro相当的表现，但未能转移到AUT和任务任务，这表明权重空间引导在未见任务中的有效性。总之，CreativityNeuro在不需要行为数据、重新训练或基于梯度的微调的情况下改善了发散性思维并减少了模式崩溃，提供了一种简单的方法来提升LLM在创造性领域的表现。
cs.AI / 7 / 2607.01436
Discrete Diffusion Language Models for Interactive Radiology Report Drafting
用于交互式放射学报告草拟的离散扩散语言模型
Abstract
Diffusion language models, which generate text by denoising a token canvas bidirectionally instead of emitting tokens left to right, have become competitive with autoregressive (AR) generation. Medical foundation models, however, remain almost entirely autoregressive. We adapt a mixture-of-experts diffusion language model, DiffusionGemma-26B, and benchmark it against its same-size AR sibling Gemma-4-26B under an identical LoRA recipe on medical visual question answering datasets, scored by a verbosity-robust LLM judge. Diffusion matches or exceeds AR on all of them, and the finetuned model (3.8B active) is competitive with frontier vision-language models; its decoding is also 3.5-4.4x faster. Beyond this parity, the diffusion model offers a drafting capability AR lacks: any-order infill. Because the canvas is denoised bidirectionally, a radiologist can fix report fragments and have the model fill the text between them, an operation inherent to diffusion but not to autoregression, which is subpar at it. This suits real reports, which are often terse or inconsistent across clinicians and institutions.
Chinese Translation
扩散语言模型通过双向去噪令牌画布生成文本，而不是从左到右逐步发出令牌，已在生成任务中与自回归（AR）生成方法相竞争。然而，医学基础模型几乎完全是自回归的。我们对一种混合专家扩散语言模型DiffusionGemma-26B进行了改进，并在相同的LoRA配方下与其同等规模的自回归模型Gemma-4-26B在医学视觉问答数据集上进行了基准测试，评分由一个对冗长性具有鲁棒性的LLM评判。扩散模型在所有测试中均与自回归模型相匹配或超越，经过微调的模型（3.8B活跃参数）在与前沿视觉-语言模型的竞争中表现出色；其解码速度也快了3.5-4.4倍。除了这种平衡，扩散模型还提供了自回归模型所缺乏的草拟能力：任意顺序的填充。由于画布是双向去噪的，放射科医生可以修正报告片段，并让模型填充它们之间的文本，这一操作是扩散模型固有的，而自回归模型在这方面表现不佳。这种特性适用于真实报告，因为这些报告通常在不同临床医生和机构之间显得简洁或不一致。
cs.AI / 8 / 2607.01465
Beyond Next-Token Prediction: An RLVR Proof of Concept for Tool-Use Agents on Atlassian Workflows
超越下一个标记预测：针对Atlassian工作流工具使用代理的RLVR概念验证
Abstract
Large language models are trained to predict the next token, not to act inside a specific API. In niche enterprise SaaS workflows -- where success means hitting the right endpoint with the right nested arguments in the right order -- this objective mismatch shows up as silent failures: dropped required fields, hallucinated tools, or early stops after a single read. We ask whether Reinforcement Learning with Verifiable Rewards (RLVR), applied directly in the target environment, closes the gap. As a proof of concept we build a suite of five synthetic environments emulating the Jira REST v3 and Confluence v2 APIs at schema fidelity; rewards are computed entirely from the tool-call trace, with no live API, no learned judge, and no human label in the loop. Scoring prompted Qwen3-1.7B and Qwen3.5-4B on the same checkers that drive GRPO training, we find that on the four scenarios whose rewards are non-degenerate the RL-trained policy lifts average reward from a 4B-baseline range of 0.35--0.92 to 0.95--1.00, with the largest single gain on Confluence page creation ($0.35 \rightarrow 1.00$). We position this as a preliminary step toward outcome-optimised small models for niche enterprise APIs, and foreground two limitations a workshop reader should weigh: hand-crafting verifiable rewards does not scale beyond the handful of endpoints reported here, and one of our five scenarios (ticket-transition) has a saturating reward shape that the prompted 4B already maxes out.
Chinese Translation
大型语言模型被训练用于预测下一个标记，而不是在特定API内进行操作。在小众企业SaaS工作流中——成功意味着以正确的顺序使用正确的嵌套参数调用正确的端点——这一目标的不匹配表现为无声的失败：丢失必需字段、虚构工具或在单次读取后提前停止。我们探讨了在目标环境中直接应用可验证奖励的强化学习（Reinforcement Learning with Verifiable Rewards, RLVR）是否能够弥补这一差距。作为概念验证，我们构建了一套五个合成环境，模拟Jira REST v3和Confluence v2 API，保持架构的一致性；奖励完全基于工具调用轨迹计算，没有实时API、没有学习评估者，也没有人类标签参与。对同一检查器驱动的Qwen3-1.7B和Qwen3.5-4B进行评分，我们发现，在四种奖励非退化的场景中，RL训练的策略将平均奖励从4B基线范围的0.35--0.92提升至0.95--1.00，其中在Confluence页面创建上的单次最大增益为($0.35
ightarrow 1.00$)。我们将此视为朝着为小众企业API优化结果的小型模型迈出的初步一步，并强调两个限制，供工作坊读者权衡：手工制作可验证奖励无法扩展到此处报告的少数端点之外，以及我们五个场景中的一个（票务转换）具有饱和奖励形状，促使的4B已达到最大值。
cs.AI / 9 / 2607.01470
World Feedback for Clinical Agents: Diagnosing RL in FHIR Environments
临床智能体的世界反馈：在FHIR环境中诊断强化学习
Abstract
Clinical protocol-execution tasks -- checking a lab value, applying a threshold, placing a correctly structured FHIR order -- are natural candidates for RL from world feedback: once clinical SMEs encode decision logic into a verifier, that verifier grades unlimited rollouts without per-episode annotation. But applying RL requires a sound feedback channel and sufficient base capability. We audit MedAgentBench v1/v2, find a 41.7\% silent-finish ceiling that makes inaction the RL dominant strategy, and construct \textbf{MedAgentBench-v3 (MAB-v3)} (508 tasks, 8.9\% ceiling). Training Qwen3-8B exposes two structural barriers: a \emph{capability ceiling} (10/20 task types have 0\% base performance, zero gradient) and a \emph{format-knowledge barrier} (3/20 types require exact clinical codes undiscoverable by exploration). Pure RL reaches 18.2\% pass@1 vs.\ 34.1\% for rule-based SFT; the 15.9~pp gap is attributable entirely to these barriers. A decision/format-knowledge/lookup taxonomy predicts RL learnability and prescribes the fix: SFT to inject codes, RL to learn conditionals.
Chinese Translation
临床协议执行任务——检查实验室值、应用阈值、下达正确结构的FHIR订单——是从世界反馈中进行强化学习（RL）的自然候选者：一旦临床领域专家（SMEs）将决策逻辑编码到验证器中，该验证器便可以在没有每个回合注释的情况下对无限次的执行进行评分。然而，应用强化学习需要一个可靠的反馈通道和足够的基础能力。我们审计了MedAgentBench v1/v2，发现41.7%的静默完成上限使得不采取行动成为强化学习的主导策略，并构建了 extbf{MedAgentBench-v3 (MAB-v3)}（508个任务，8.9%的上限）。训练Qwen3-8B暴露了两个结构性障碍： extit{能力上限}（10/20个任务类型的基础性能为0%，无梯度）和 extit{格式知识障碍}（3/20个类型需要精确的临床代码，无法通过探索发现）。纯强化学习的通过率为18.2%（pass@1），而基于规则的监督微调（SFT）为34.1%；这15.9个百分点的差距完全归因于这些障碍。一个决策/格式知识/查找的分类法预测了强化学习的可学习性，并提出了解决方案：通过SFT注入代码，通过强化学习学习条件。
cs.AI / 10 / 2607.01480
Procedural Memory Distillation: Online Reflection for Self-Improving Language Models
过程记忆蒸馏：自我改进语言模型的在线反思
Abstract
Reinforcement learning with verifiable rewards (RLVR), along with recent selfdistillation variants such as SDPO, evaluates each rollout against a verifier and updates the policy from that episode-level signal. However, the richer procedural information in the rollout is rarely retained or reused. Across episodes and epochs, the model repeatedly encounters related problems under a changing policy, producing cross-episode signals that episode-local updates cannot capture: which strategies consistently pass verification, which failure modes persist, which patterns recur. We propose Procedural Memory Distillation (PMD), which converts these crossepisode signals into reusable procedural memory and distills it into the policy's weights during training. This memory functions as a training scaffold, absorbed into the policy itself, yielding a memory-free model at inference. PMD organizes the memory at three levels of abstraction: raw trajectories, self-reflected strategies and lessons, and higher-level behavioral patterns that recur across problems, all extracted online from the model's own trajectories. A memory-conditioned self-teacher draws on the accumulated experience to supervise the student on its own rollouts, enabling student to progressively internalize procedural knowledge within its parameters. The central design principle is co-evolution: the policy generates rollouts that update the memory, and memory shapes the supervision that updates the policy. Empirically, across Qwen3-8B and OLMo3-Instruct-7B, PMD improves over SDPO by 3.8-5.5% on SCIKNOWEVAL and 7.9-13.6% on LIVECODEBENCH. Co-evolution powers these gains: freezing either the memory or the policy trails PMD by more than 10% across SCIKNOWEVAL domains.
Chinese Translation
可验证奖励的强化学习（RLVR）以及最近的自蒸馏变体如SDPO，通过对每次回合进行验证并根据该回合级信号更新策略。然而，回合中的丰富过程信息很少被保留或重用。在多个回合和周期中，模型在变化的策略下反复遇到相关问题，产生回合间信号，而回合局部更新无法捕捉这些信号：哪些策略持续通过验证，哪些失败模式持续存在，哪些模式反复出现。我们提出了过程记忆蒸馏（PMD），将这些跨回合信号转换为可重用的过程记忆，并在训练过程中将其蒸馏到策略的权重中。这种记忆作为训练支架，融入策略本身，在推理时产生无记忆模型。PMD在三个抽象层次上组织记忆：原始轨迹、自我反思的策略和经验，以及跨问题反复出现的更高层次行为模式，所有这些都是从模型自身的轨迹中在线提取的。一个基于记忆的自我教师利用积累的经验来监督学生在其自身回合中的表现，使学生能够逐步在其参数中内化过程知识。中心设计原则是共同进化：策略生成的回合更新记忆，而记忆塑造的监督更新策略。在实证研究中，在Qwen3-8B和OLMo3-Instruct-7B上，PMD在SCIKNOWEVAL上比SDPO提高了3.8-5.5%，在LIVECODEBENCH上提高了7.9-13.6%。共同进化推动了这些提升：冻结记忆或策略中的任一部分，PMD在SCIKNOWEVAL领域的性能下降超过10%。
cs.AI / 11 / 2607.01507
The Agentic Garden of Forking Paths
代理性分叉路径的花园
Abstract
Empirical research rarely admits a unique analysis. Different analytical choices can lead to different conclusions from the same data, yet these hidden forking paths are difficult to observe. We show that AI agents capture much of the analytical variation among human researchers while making these paths explicit. Across four high-stakes domains, assigning different personas is sufficient for AI agents to report divergent, often opposing, conclusions from the same data and question, with findings systematically aligned with those beliefs. In a study in which 42 human research teams analyzed the same immigration dataset, AI agents reproduced 72% of the human ideological gap in reported effect estimates. Despite reaching opposing conclusions, it is difficult to identify clear issues in each analysis based on the final AI reports: 86% passed independent AI review and 78% passed majority human expert review. These findings suggest that the central challenge is often not flawed analyses, but selective exploration and reporting from a large space of methodologically defensible analyses. AI agents may amplify this longstanding problem by making such exploration inexpensive and scalable. To address this, we introduce the m-value (multiverse value), the probability that an analysis path would produce a claim at least as extreme as the reported one. We further introduce Agentic Bootstrap, which estimates the m-value by using AI agents to sample plausible analysis paths. Applied to the human immigration study, 13.5% of reported human analyses fell in the most extreme 5% of the analysis space (m<0.05). Scientific evidence should therefore be evaluated not only by a single reported analysis but also by its position within the distribution of analyses that could reasonably have been reported. Agentic Bootstrap makes this distribution observable and turns it into a criterion for scientific credibility.
Chinese Translation
实证研究很少承认唯一的分析。不同的分析选择可能导致从相同数据中得出不同的结论，但这些隐藏的分叉路径难以观察。我们展示了人工智能（AI）代理在捕捉人类研究者之间的分析变异性方面的能力，同时使这些路径变得显而易见。在四个高风险领域中，分配不同的人物角色足以使AI代理从相同的数据和问题中报告出不同的、往往是对立的结论，且这些发现与所持信念系统性一致。在一项研究中，42个研究团队分析了相同的移民数据集，AI代理再现了人类在报告效果估计时的72%的意识形态差距。尽管得出了对立的结论，但根据最终的AI报告，很难识别每个分析中的明确问题：86%的分析通过了独立的AI审查，78%通过了多数人类专家审查。这些发现表明，中心挑战往往不是分析缺陷，而是从大量方法论上可辩护的分析中进行选择性探索和报告。AI代理可能通过使这种探索变得廉价和可扩展，从而加剧这一长期存在的问题。为了解决这个问题，我们引入了m值（多元宇宙值），即某一分析路径产生至少与报告的主张一样极端的主张的概率。我们进一步引入了代理性自助法（Agentic Bootstrap），该方法通过使用AI代理来采样合理的分析路径来估计m值。应用于人类移民研究时，13.5%的报告人类分析落在分析空间中最极端的5%（m<0.05）。因此，科学证据应不仅通过单一的报告分析进行评估，还应通过其在合理可能被报告的分析分布中的位置进行评估。代理性自助法使这一分布可观察，并将其转化为科学可信度的标准。
cs.AI / 12 / 2607.01510
Janus: a Playground for User-Involved Agentic Permission Management
Janus：用户参与的代理权限管理实验平台
Abstract
AI agents that autonomously execute tool calls on a user's behalf raise pressing questions about permission management: what role could users play, and what role should they play? Despite many proposed approaches, the user's role in agentic permission management remains under explored. We introduce Janus, a playground system for implementing and evaluating user-involved agentic permission management designs. Janus consists of two components: Janus-Core, a modular agentic system supporting a diverse spectrum of permission management designs, and Janus-Harness, an automated evaluation framework. Grounded in a conceptual model that identifies key design axes for user involvement, we implement six permission assistants spanning the design space and evaluate them across three scenarios and three synthetic responders. We demonstrate that user input is critical and can significantly strengthen privacy and security, that AI augmentation of user decisions can help reduce cognitive load, and that realistic user behavior including permission fatigue must be accounted for in system design. No single design performs optimally across all contexts, motivating a more principled and context-sensitive approach to deploying permission assistants in agentic systems. Janus is publicly available to support future investigation into this dimension of agentic system design.
Chinese Translation
自主代表用户执行工具调用的人工智能代理引发了关于权限管理的紧迫问题：用户可以扮演什么角色，应该扮演什么角色？尽管提出了许多方法，用户在代理权限管理中的角色仍然未得到充分探索。我们介绍了Janus，一个用于实施和评估用户参与的代理权限管理设计的实验平台。Janus由两个组件组成：Janus-Core，一个支持多样化权限管理设计的模块化代理系统，以及Janus-Harness，一个自动化评估框架。基于一个识别用户参与关键设计轴的概念模型，我们实现了六个跨越设计空间的权限助手，并在三个场景和三个合成响应者中对其进行了评估。我们证明了用户输入是至关重要的，并且可以显著增强隐私和安全性，人工智能对用户决策的增强可以帮助减轻认知负担，并且在系统设计中必须考虑到包括权限疲劳在内的现实用户行为。没有单一设计能够在所有上下文中表现最佳，这促使我们采取更有原则和上下文敏感的方法来部署代理系统中的权限助手。Janus已公开可用，以支持未来对这一代理系统设计维度的研究。
cs.AI / 13 / 2607.01511
Revisiting Chain-of-Thought Reasoning under Limited Supervision: Semi-supervised Chain-of-Thought Learning
在有限监督下重新审视思维链推理：半监督思维链学习
Abstract
Chain-of-thought (CoT) reasoning has emerged as an effective approach for activating latent reasoning capabilities in large language models. However, most existing CoT methods use reasoning chains mainly as inference-time prompts, while the generated reasoning traces are rarely reused as semi-supervised learning signals. In this report, we define \textbf{Semi-supervised Chain-of-Thought Learning} and propose \textbf{Semi-CoT}, a simple framework that uses unlabeled questions to construct pseudo reasoning supervision. Semi-CoT samples multiple pseudo-CoTs for each unlabeled question, estimates answer-level semantic entropy, and selects low-entropy reasoning chains as reliable pseudo-CoT demonstrations. This extends the self-training view of CoT from inference-time refinement to semi-supervised pseudo-supervision. Pilot experiments on AQuA, SVAMP, GSM8K, and MultiArith show that the entropy gate selects high-precision pseudo-CoTs, with pseudo-answer precision ranging from $91.36\%$ to $100\%$. Semi-CoT also gives small gains on SVAMP and GSM8K, while AQuA shows negative transfer and MultiArith reaches a ceiling. These results suggest that unlabeled questions can provide reliable pseudo reasoning signals, but their effective use still requires stronger demonstration selection or student training.
Chinese Translation
思维链（Chain-of-thought, CoT）推理已成为激活大型语言模型潜在推理能力的有效方法。然而，大多数现有的 CoT 方法主要将推理链用作推理时的提示，而生成的推理轨迹很少被重新利用作为半监督学习信号。在本报告中，我们定义了 extbf{半监督思维链学习}（Semi-supervised Chain-of-Thought Learning），并提出了 extbf{Semi-CoT}，这是一个简单的框架，利用未标记的问题构建伪推理监督。Semi-CoT 为每个未标记的问题采样多个伪 CoTs，估计答案级别的语义熵，并选择低熵推理链作为可靠的伪 CoT 演示。这将 CoT 的自我训练视角从推理时的细化扩展到半监督伪监督。在 AQuA、SVAMP、GSM8K 和 MultiArith 上的初步实验表明，熵门选择高精度的伪 CoTs，伪答案的精度范围为 $91.36\%$ 到 $100\\%$。Semi-CoT 在 SVAMP 和 GSM8K 上也取得了小幅提升，而 AQuA 显示出负迁移，MultiArith 达到了上限。这些结果表明，未标记的问题可以提供可靠的伪推理信号，但其有效利用仍需更强的演示选择或学生训练。
cs.AI / 14 / 2607.01531
OPINE-World: Programmatic World Modeling with Ontology-error-Prioritized Interactive Exploration
OPINE-World：基于本体错误优先的程序化世界建模与交互探索
Abstract
Learning how an environment behaves from interaction is central to building agents that adapt to unfamiliar tasks. World models learned with deep networks are flexible but data-hungry and transfer poorly beyond their training distribution. Program-synthesized world models, written as source code by LLMs and refined through counterexample-guided inductive synthesis (CEGIS), are instead data-efficient and reusable, yet they have been demonstrated mainly on structured-state worlds with a given object vocabulary, and a single program search does not scale to pixel-rendered environments whose object structure must be hypothesized flexibly. We introduce OPINE-World, an LLM agent that learns an object-centric programmatic world model online from interaction. OPINE-World couples two cooperating agents in a loop of hypothesis and test, one acting in the environment and one synthesizing the model in code with replay verification and model-based planning, and it steers exploration with a Bayesian measure of object-type adequacy we call ontology error. We evaluate OPINE-World on ARC-AGI-3, a benchmark for skill-acquisition efficiency in which the object vocabulary, the goal, and the action semantics are withheld. OPINE-World solves 20 of 25 games without per-game training and reaches an action-efficiency score of 78.4 against the human baseline.
Chinese Translation
从交互中学习环境行为是构建能够适应不熟悉任务的智能体的核心。使用深度网络学习的世界模型灵活但数据需求高，且在训练分布之外的迁移效果较差。通过大型语言模型（LLMs）编写的程序合成世界模型，经过反例引导的归纳合成（CEGIS）进行优化，具有数据高效和可重用的优点，但目前主要在具有特定对象词汇的结构化状态世界中进行验证，而单一程序搜索无法扩展到需要灵活假设对象结构的像素渲染环境。我们提出了OPINE-World，一个能够在线从交互中学习以对象为中心的程序化世界模型的LLM智能体。OPINE-World将两个协作智能体结合在假设与测试的循环中，一个在环境中行动，另一个通过重放验证和基于模型的规划在代码中合成模型，并通过我们称之为本体错误的贝叶斯对象类型适应性度量来引导探索。我们在ARC-AGI-3上评估OPINE-World，这是一个技能获取效率的基准，其中对象词汇、目标和动作语义被隐去。OPINE-World在没有针对每个游戏进行训练的情况下解决了25个游戏中的20个，并在与人类基线的比较中达到了78.4的动作效率得分。
cs.AI / 15 / 2607.01567
Scaling Trends for Lie Detector Oversight in Preference Learning
偏好学习中谎言检测器监督的规模化趋势
Abstract
Deceptive behavior in LLMs is costly to monitor and prevent, motivating approaches such as Scalable Oversight via Lie Detectors (SOLiD) (Cundy & Gleave, 2025), which uses lie detectors to identify responses for review by high-cost labelers. In this paper, we scale SOLiD to larger models and evaluate it in more diverse and realistic preference-learning settings. We find favorable scaling: undetected deception drops from 34% for 1B-parameter models to 14% for 405B-parameter models at a detector true positive rate of 99%, and expensive human labelers can be removed entirely from the fine-tuning phase without a statistically significant increase in deception. However, SOLiD is sensitive to distribution shift between detector training and preference-training data, which can drive detector false positive rates to impractical levels.
Chinese Translation
在大规模语言模型（LLMs）中，监测和防止欺骗行为的成本很高，这促使了诸如通过谎言检测器进行可扩展监督（Scalable Oversight via Lie Detectors, SOLiD）（Cundy & Gleave, 2025）等方法的出现，该方法利用谎言检测器识别需由高成本标注者审查的响应。本文将SOLiD扩展到更大规模的模型，并在更具多样性和现实性的偏好学习环境中进行评估。我们发现规模化效果良好：在检测器真正阳性率为99%的情况下，未被检测到的欺骗行为从1B参数模型的34%降至405B参数模型的14%，并且在微调阶段可以完全去除昂贵的人类标注者，而不会显著增加欺骗行为的发生率。然而，SOLiD对检测器训练数据与偏好训练数据之间的分布变化敏感，这可能导致检测器的假阳性率达到不切实际的水平。
cs.AI / 16 / 2607.01584
EO-Agents: A Three-Agent LLM Pipeline for Earth Observation Hypothesis Generation
EO-Agents：一个用于地球观测假设生成的三代理大型语言模型管道
Abstract
Large language models have recently been explored for scientific hypothesis generation, but most prior work relies on unstructured literature and free-form textual claims. We present a pipeline for Earth observation that grounds hypothesis generation directly in the NASA Earth Observation Knowledge Graph. A heterogeneous graph neural network trained on historical co-usage relations ranks candidate dataset pairings, and a three-agent LLM pipeline filters, generates, and evaluates structured research hypotheses. Applied to 1,475 NASA datasets, the system produces 160 hypotheses spanning multiple Earth-science domains, including ecohydrology, glaciology, aerosol--cloud interactions, vegetation phenology, and stratospheric chemistry. Model-predicted novel dataset pairings are rated nearly as plausible as held-out real co-usages from the literature, indicating that the pipeline surfaces scientifically coherent yet unexplored combinations. A 2*2*2 factorial experiment across GPT-5.2 and Claude Sonnet 4.6 shows that hypothesis rankings remain stable, while absolute scores depend strongly on judge identity, highlighting limitations of single-judge LLM evaluation.
Chinese Translation
大型语言模型最近被探索用于科学假设生成，但大多数先前的工作依赖于非结构化文献和自由形式的文本声明。我们提出了一个地球观测管道，直接基于NASA地球观测知识图谱进行假设生成。一个在历史共用关系上训练的异构图神经网络对候选数据集配对进行排名，而一个三代理大型语言模型管道则过滤、生成和评估结构化研究假设。该系统应用于1,475个NASA数据集，生成了160个跨多个地球科学领域的假设，包括生态水文学、冰川学、气溶胶-云相互作用、植被物候学和平流层化学。模型预测的新数据集配对的可行性评分几乎与文献中保留的真实共用关系相当，表明该管道能够挖掘出科学上连贯但尚未探索的组合。针对GPT-5.2和Claude Sonnet 4.6进行的2*2*2因子实验表明，假设排名保持稳定，而绝对评分则强烈依赖于评审者身份，突显了单一评审者大型语言模型评估的局限性。
cs.AI / 17 / 2607.01590
Hawk: Harnessing Hardware-Aware Knowledge for High-Performance NPU Kernel Generation
Hawk：利用硬件感知知识生成高性能NPU内核
Abstract
Developing high-performance kernels for Neural Processing Units (NPUs) is a critical industry bottleneck, requiring developers to manually navigate implicit hardware constraints and strict memory hierarchies. While large language models offer immense automation potential, they fail catastrophically on NPUs due to a fundamental lack of hardware-specific priors. Naively transplanting code snippets from similar NPU kernels may pass the compiler, but it consistently triggers runtime crashes and performance degradation by blindly violating underlying hardware constraints. To overcome this, we introduce Hawk, a training-free framework that harnesses hardware-aware knowledge through three core modules: (1) Run-Time Knowledge Synthesis Module, which employs a Triple-Part Executable Knowledge Representation to inherently couple the error context with executable semantics; (2) Bottleneck-Aware Knowledge Retrieval Module, which implements a 2D-Retrieval paradigm to project queries into orthogonal syntactic and hardware-aligned semantic spaces; and (3) Effect-Driven Knowledge Distillation Module, which leverages LLM-driven semantic arbitration to continuously distill the knowledge by pruning errors and consolidating redundancies based on the empirical execution feedback. Extensive evaluations on real-world NPU workloads demonstrate that Hawk elevates generation accuracy from 49.4% to 80.0%, while achieving up to a 2.2x execution speedup over state-of-the-art baselines.
Chinese Translation
为神经处理单元（NPU）开发高性能内核是一个关键的行业瓶颈，开发者需要手动应对隐含的硬件约束和严格的内存层次结构。尽管大型语言模型提供了巨大的自动化潜力，但由于缺乏特定于硬件的先验知识，它们在NPU上的表现惨遭失败。简单地将类似NPU内核的代码片段移植可能通过编译器，但由于盲目违反底层硬件约束，始终会导致运行时崩溃和性能下降。为了解决这个问题，我们提出了Hawk，一个无需训练的框架，通过三个核心模块利用硬件感知知识：(1) 运行时知识合成模块，该模块采用三部分可执行知识表示，将错误上下文与可执行语义内在耦合；(2) 瓶颈感知知识检索模块，该模块实现了二维检索范式，将查询投影到正交的句法和与硬件对齐的语义空间；(3) 效果驱动的知识蒸馏模块，该模块利用大型语言模型驱动的语义仲裁，通过基于经验执行反馈修剪错误和整合冗余，持续提炼知识。在真实世界NPU工作负载上的广泛评估表明，Hawk将生成准确率从49.4%提升至80.0%，并在与最先进基准相比时实现了最高2.2倍的执行加速。
cs.AI / 18 / 2607.01595
Safe and Adaptive Cloud Healing: Verifying LLM-Generated Recovery Plans with a Neural-Symbolic World Model
安全与自适应云修复：使用神经符号世界模型验证LLM生成的恢复计划
Abstract
As the scale and complexity of cloud-based AI systems continue to escalate, ensuring service reliability through rapid fault detection and adaptive recovery has become a critical challenge. While existing approaches integrate Large Language Models (LLMs) for semantic understanding and Deep Reinforcement Learning (DRL) for policy optimization, they often rely on sequential, loosely coupled architectures that underutilize the generative and reasoning capabilities of LLMs. In this paper, we propose a paradigm shift with PASE, a Planning-Aware Semantic self-healing engine, a novel fault self-healing framework that reconceptualizes recovery as a neuro-symbolic program synthesis task. PASE employs an LLM as a core Plan Synthesis Engine to generate structured recovery plans from a library of semantic primitives. A Neural-Symbolic World Model verifies plan feasibility through simulation, while a Meta-Prompt Optimizer, trained via DRL, learns to generate optimal prompts that guide the LLM's planning process. This tight reason-plan-verify-adapt loop enables dynamic, context-aware recovery strategy generation beyond predefined action spaces. Experiments on a real-world cloud fault injection dataset demonstrate that PASE significantly outperforms state-of-the-art methods, reducing average system recovery time by over 40% and improving fault detection accuracy in unknown fault scenarios. Our framework advances autonomous system management by unifying LLM-based reasoning with model-assisted verification and meta-learned guidance.
Chinese Translation
随着基于云的人工智能系统规模和复杂性的不断增加，通过快速故障检测和自适应恢复确保服务可靠性已成为一项关键挑战。虽然现有方法集成了大型语言模型（LLMs）以实现语义理解，并采用深度强化学习（DRL）进行策略优化，但它们通常依赖于顺序、松耦合的架构，未能充分利用LLMs的生成和推理能力。本文提出了一种范式转变，提出了PASE，一个规划感知语义自愈引擎，这是一种新颖的故障自愈框架，将恢复重新概念化为神经符号程序合成任务。PASE将LLM作为核心计划合成引擎，从语义原语库中生成结构化的恢复计划。神经符号世界模型通过仿真验证计划的可行性，而通过DRL训练的元提示优化器学习生成最佳提示，以指导LLM的规划过程。这种紧密的推理-计划-验证-适应循环使得超越预定义动作空间的动态、上下文感知的恢复策略生成成为可能。在真实世界的云故障注入数据集上的实验表明，PASE显著优于最先进的方法，将平均系统恢复时间减少了超过40%，并提高了在未知故障场景中的故障检测准确性。我们的框架通过将基于LLM的推理与模型辅助验证和元学习指导相结合，推动了自主系统管理的发展。
cs.AI / 19 / 2607.01601
SemHash-LLM: A Multi-Granularity Semantic Hashing Framework for Document Deduplication
SemHash-LLM：一种用于文档去重的多粒度语义哈希框架
Abstract
Large scale document deduplication must preserve semantic equivalence while remaining efficient over massive corpora. We present SemHash LLM, a multi granularity framework that unifies semantic projection hashing, attention weighted MinHash, contrastive boundary learning, and selective LLM based adjudication. The method combines character, token, and document level signals through gated fusion, then applies a cascaded filtering pipeline for efficient candidate reduction. Semantic projection hashing learns compact binary codes in distilled LLM embedding space, while attention weighted Min- Hash suppresses boilerplate and emphasizes informative content. Adaptive decision boundaries and uncertainty estimation further improve robustness across template pollution, short text perturbation, containment, and viral fragments. Experiments show that SemHash LLM achieves strong duplicate detection quality with less than one percent neural verification cost.
Chinese Translation
大规模文档去重必须在保持语义等价的同时，在庞大的语料库上保持高效。我们提出了SemHash LLM，这是一种将语义投影哈希、加权MinHash、对比边界学习和选择性基于LLM（大语言模型）裁决统一起来的多粒度框架。该方法通过门控融合结合字符、标记和文档级别的信号，然后应用级联过滤管道以实现高效的候选项减少。语义投影哈希在提炼的LLM嵌入空间中学习紧凑的二进制编码，而加权MinHash则抑制模板内容并强调信息丰富的内容。自适应决策边界和不确定性估计进一步提高了在模板污染、短文本扰动、包含和病毒片段等情况下的鲁棒性。实验表明，SemHash LLM在神经验证成本低于1%的情况下实现了强大的重复检测质量。
cs.AI / 20 / 2607.01610
Profit-Based Counterfactual Explanations for Product Improvement: A Case Study of Manga Sales in Japan
基于利润的反事实解释用于产品改进：日本漫画销售案例研究
Abstract
Counterfactual explanation (CE) is widely used to enhance the interpretability of machine learning models and support data-driven decision-making based on model predictions. However, existing CE methods typically require two exogenously specified inputs: a desired output value (target) and a distance function that quantifies changes in explanatory variables. In regression settings, neither the validity of target specification nor the practical interpretation of the distance metric has been sufficiently addressed. Furthermore, most existing CE methods focus on altering predictions rather than optimizing a decision objective, even though real-world decision-making often requires explicit objective maximization. To address these limitations, we formulate CE as a profit maximization problem in management and marketing contexts and propose a framework termed profit-based counterfactual explanation (PBCE). PBCE eliminates the need for exogenous target specification by directly maximizing profit as the primary optimization objective. Concurrently, the distance term is reinterpreted as the cost of modifying product attributes, providing a clear and economically grounded interpretation.
Chinese Translation
反事实解释（Counterfactual Explanation, CE）广泛用于增强机器学习模型的可解释性，并支持基于模型预测的数据驱动决策。然而，现有的CE方法通常需要两个外生指定的输入：一个期望的输出值（目标）和一个量化解释变量变化的距离函数。在回归设置中，目标规范的有效性和距离度量的实际解释尚未得到充分解决。此外，大多数现有的CE方法侧重于改变预测，而不是优化决策目标，尽管现实世界的决策往往需要明确的目标最大化。为了解决这些局限性，我们将CE构建为管理和营销背景下的利润最大化问题，并提出了一种称为基于利润的反事实解释（Profit-Based Counterfactual Explanation, PBCE）的框架。PBCE通过直接最大化利润作为主要优化目标，消除了对外生目标规范的需求。同时，距离项被重新解释为修改产品属性的成本，提供了清晰且经济上合理的解释。
cs.AI / 21 / 2607.01612
Scaling with Confidence: Calibrating Confidence of LLMs for Adaptive Test Time Scaling
自信的扩展：为自适应测试时间扩展校准大型语言模型的置信度
Abstract
Training large language models (LLMs) with reinforcement learning (RL) has significantly advanced their performance on reasoning and question-answering tasks. However, prevailing RL reward designs typically prioritize response correctness, neglecting to incentivize models to express their confidence accurately. This leads to a critical problem: performance gains are often accompanied by poor calibration between confidence and accuracy, misleading models to overconfidently hallucinate when uncertain. To address this limitation, we propose $\textbf{C}$orrectness and $\textbf{C}$onfidence $\textbf{C}$alibration $\textbf{R}$einforcement $\textbf{L}$earning ($\textbf{C3RL}$), a novel RL algorithm integrating correctness, calibration and dataset-informed reference accuracy rewards together. Comprehensive evaluation across 8 text and multimodal datasets demonstrates that C3RL enhances calibration without sacrificing accuracy, outperforming the current state-of-the-art method in both performance and calibration metrics. Utilizing the well-calibrated verbalized confidence from C3RL, we further introduce $\textbf{C}$onfidence-based $\textbf{A}$daptive Test Time $\textbf{S}$caling ($\textbf{CAS}$), an adjustable inference-time strategy that allocates computational resources based on response confidence. Experiments show that CAS surpasses majority voting on both in-domain and out-of-domain datasets while reducing the inference budget by up to 12.33 times. We believe the synergy of C3RL and CAS paves the way for deploying more reliable and resource-efficient LLMs. The code, data and models will be released.
Chinese Translation
使用强化学习（RL）训练大型语言模型（LLMs）显著提升了它们在推理和问答任务上的表现。然而，现有的RL奖励设计通常优先考虑响应的正确性，忽视了激励模型准确表达其置信度。这导致了一个关键问题：性能提升往往伴随着置信度与准确性之间的校准不良，误导模型在不确定时过于自信地产生幻觉。为了解决这一局限性，我们提出了$ extbf{C}$orrectness和$ extbf{C}$onfidence $ extbf{C}$alibration $ extbf{R}$einforcement $ extbf{L}$earning（$ extbf{C3RL}$），这是一种新颖的RL算法，将正确性、校准和基于数据集的参考准确性奖励结合在一起。对8个文本和多模态数据集的全面评估表明，C3RL在不牺牲准确性的情况下增强了校准，且在性能和校准指标上均优于当前最先进的方法。利用C3RL生成的良好校准的口头置信度，我们进一步提出了$ extbf{C}$onfidence-based $ extbf{A}$daptive Test Time $ extbf{S}$caling（$ extbf{CAS}$），这是一种可调的推理时间策略，根据响应置信度分配计算资源。实验表明，CAS在领域内和领域外数据集上均超越了多数投票，同时将推理预算减少了多达12.33倍。我们相信C3RL与CAS的协同作用为部署更可靠和资源高效的LLMs铺平了道路。代码、数据和模型将会发布。
cs.AI / 22 / 2607.01621
Spatial Support Matters: Geometry-Aware Graph Fusion for Rainfall Field Reconstruction
空间支持的重要性：基于几何的图融合用于降雨场重建
Abstract
Fine-scale rainfall reconstruction is critical for urban flood modeling, but real rainfall sensing systems observe the field through incompatible spatial supports: gauges measure points, microwave links measure paths, and radar/satellite products measure gridded areas. These differences in measurement support impose geometrically distinct constraints on the rainfall field, yet existing heterogeneous graph approaches reconcile such sources in feature space, giving each its own embedding while discarding the geometry of its support. We propose a geometry-aware multi-support heterogeneous graph neural network that represents each observation according to its support type (0D point, 1D line, or 2D grid) as a distinct node layer, and fuses them through cross-support message passing into a point-support prediction layer from which the field is reconstructed. An inductive masked-node formulation decouples prediction resolution from sensing resolution, allowing the same trained model to reconstruct the field at user-defined target locations or display grids. On Singapore data, the proposed method reduces RMSE by 23.2\% over the classical interpolation baseline, inverse-distance weighting, and consistently outperforms other neural architectures such as convolutional fusion and support-agnostic heterogeneous graph baselines. A generalization study using data from Sydney, Australia lets us characterize when multi-support fusion helps: the available skill appears to depend on gauge spacing relative to the spatial correlation length of the field, so fusion delivers the largest gains where the field is under-sampled relative to its correlation length and little when it is already resolved. Code and models will be open-sourced upon paper acceptance.
Chinese Translation
细尺度降雨重建对于城市洪水建模至关重要，但实际的降雨传感系统通过不兼容的空间支持观察降雨场：雨量计测量点，微波链路测量路径，而雷达/卫星产品测量网格区域。这些测量支持的差异对降雨场施加了几何上不同的约束，然而现有的异构图方法在特征空间中调和这些来源，为每个来源提供自己的嵌入，同时忽略了其支持的几何特性。我们提出了一种几何感知的多支持异构图神经网络，依据观测的支持类型（0D点、1D线或2D网格）将每个观测表示为一个独特的节点层，并通过跨支持消息传递将它们融合到一个点支持预测层中，从而重建降雨场。诱导的掩蔽节点形式将预测分辨率与传感分辨率解耦，允许同一训练模型在用户定义的目标位置或显示网格上重建降雨场。在新加坡的数据上，所提方法相比经典插值基线（逆距离加权）减少了23.2%的均方根误差（RMSE），并且在性能上始终优于其他神经网络架构，如卷积融合和支持无关的异构图基线。使用来自澳大利亚悉尼的数据进行的泛化研究使我们能够表征多支持融合的有效性：可用的技能似乎依赖于雨量计间距与降雨场空间相关长度的关系，因此在降雨场相对于其相关长度被欠采样时，融合带来了最大的收益，而在降雨场已经被解析时则收益较小。代码和模型将在论文接受后开源。
cs.AI / 23 / 2607.01639
Autonomous discovery of traffic laws with AI traffic scientists
利用人工智能交通科学家自主发现交通法则
Abstract
Universal traffic laws describe recurrent patterns in congestion, mobility and driving behavior across cities, providing a scientific basis for transportation planning, management and control. Their discovery, however, remains expert-driven, requiring candidate regularities to be identified from heterogeneous observational evidence or validated through intervention experiments. Although autonomous artificial intelligence (AI) systems have advanced scientific discovery in controlled laboratory settings, extending them to complex transportation domains remains a challenge. Here we present TrafficSci, an agentic AI system that formulates traffic-law discovery as an iterative, auditable workflow integrating evidence scoping, critic-judge hypothesis induction, and observational-interventional validation. Across four case studies spanning population, network, control and trajectory scales, TrafficSci autonomously rediscovers three established traffic laws and identifies an unreported intrinsic temporal memory scale in urban driving behavior, statistically consistent across eight cities and two trajectory datasets. TrafficSci provides a route for extending AI-driven scientific discovery from controlled domains to complex urban systems.
Chinese Translation
普遍交通法则描述了城市中拥堵、流动性和驾驶行为的重复模式，为交通规划、管理和控制提供了科学基础。然而，它们的发现仍然依赖于专家，需要从异质的观察证据中识别候选规律或通过干预实验进行验证。尽管自主人工智能（AI）系统在受控实验室环境中推动了科学发现，但将其扩展到复杂的交通领域仍然是一项挑战。在此，我们提出了TrafficSci，一个将交通法则发现形式化为一个迭代的、可审计的工作流程的代理AI系统，该工作流程整合了证据范围、批评-判断假设诱导和观察-干预验证。在涵盖人口、网络、控制和轨迹尺度的四个案例研究中，TrafficSci自主重新发现了三个已建立的交通法则，并识别出城市驾驶行为中一个未报告的内在时间记忆尺度，该尺度在八个城市和两个轨迹数据集中具有统计一致性。TrafficSci为将AI驱动的科学发现从受控领域扩展到复杂城市系统提供了一条途径。
cs.AI / 24 / 2607.01661
Diverse Evidence, Better Forecasts: Multi-Agent Deliberation Under Information Asymmetry
多样化证据，提升预测：信息不对称下的多智能体审议
Abstract
Multi-agent systems are increasingly used for forecasting future events, as deliberation among multiple LLMs is believed to improve reasoning and calibration. Yet existing approaches overlook a critical design choice: what information each agent receives. When all agents are given identical evidence, deliberation collapses into herding rather than genuine belief revision, leaving multi-agent systems little better than a single agent. We identify this as a fundamental gap and propose designed information asymmetry to close it: by partitioning evidence into shared public and disjoint private subsets, each agent holds exclusive knowledge that can only reach others through deliberation. We theoretically show that this decomposition reduces inter-agent error correlation, and instantiate it in InfoDelphi, a framework combining relevance-aware evidence routing, rationale-based iterative deliberation, and confidence-weighted aggregation. On PolyGym, a benchmark of 375 binary forecasting questions derived from real-world prediction markets, InfoDelphi outperforms the strongest single-agent and multi-agent baselines by 12--18% in Brier score and 4--8 percentage points in accuracy. More detailed experiments confirm that removing information asymmetry eliminates most deliberation gains, establishing diversity of input as the key enabler of effective multi-agent reasoning.
Chinese Translation
多智能体系统在预测未来事件方面的应用日益增多，因为多个大型语言模型（LLMs）之间的审议被认为能够改善推理和校准。然而，现有的方法忽视了一个关键的设计选择：每个智能体接收到的信息。当所有智能体获得相同的证据时，审议就会陷入从众，而非真正的信念修正，使得多智能体系统的效果几乎与单一智能体无异。我们将此识别为一个根本性缺口，并提出设计信息不对称以填补这一缺口：通过将证据划分为共享的公共子集和不相交的私人子集，每个智能体持有的独占知识只能通过审议传递给其他智能体。我们从理论上证明，这种分解减少了智能体之间的错误相关性，并在InfoDelphi中实现了该方法，InfoDelphi是一个结合了相关性感知的证据路由、基于理由的迭代审议和置信度加权聚合的框架。在PolyGym上，这是一个基于真实预测市场的375个二元预测问题的基准，InfoDelphi在Brier得分上比最强的单智能体和多智能体基线提高了12%至18%，在准确率上提高了4至8个百分点。更详细的实验确认，消除信息不对称会消除大部分审议收益，确立了输入多样性作为有效多智能体推理的关键推动因素。
cs.AI / 25 / 2607.01674
Separating Expert Retention from Autonomous Source Inference in Raw-ECG-Replay-Free Continual ECG Deployment
在无原始ECG重放的持续ECG部署中将专家保留与自主源推断分离
Abstract
In multi-source ECG deployment, models may need to incorporate new data sources when earlier raw ECGs cannot be retained or replayed. Freezing a pretrained backbone and assigning each source an isolated classifier prevents parameter interference, but deployment still requires selecting an expert when source metadata are unavailable. We study this distinction through \ours{}, an incremental expert bank built on frozen 1024-dimensional ECGFounder features. Each arriving domain adds a balanced-softmax linear expert, while a lightweight router is fitted only on retained training features and domain labels from sources observed so far. A validation-calibrated margin rule fuses the two most likely experts instead of committing to a single routed expert. On CPSC, PTB-XL, Georgia, and Chapman-Shaoxing, source-aware expert selection reaches $0.7915\pm0.0036$ Macro-F1 and a matched offline independent-head reference reaches $0.7885\pm0.0009$, supporting strong source-aware expert retention. Without source IDs, an MLP router reaches $0.7756\pm0.0027$ and top-2 margin fusion reaches $0.7782\pm0.0022$. The top-2 gain over hard MLP routing is small ($+0.0026$), with a 95\% confidence interval from paired bootstrap that includes zero. Across three domain orders, the top-2-to-oracle gap remains $0.0111$--$0.0133$, identifying autonomous source inference as the main remaining bottleneck. No raw ECGs are replayed, but frozen training features are retained for router updates; the method is therefore not memory-free.
Chinese Translation
在多源ECG部署中，当早期的原始ECG无法保留或重放时，模型可能需要整合新的数据源。冻结预训练的主干网络并为每个源分配一个独立的分类器可以防止参数干扰，但在源元数据不可用时，部署仍需选择一个专家。我们通过 extit{ours}{}进行研究，这是一个基于冻结的1024维ECGFounder特征构建的增量专家库。每到达一个新领域，就会添加一个平衡软最大线性专家，而轻量级路由器仅在迄今为止观察到的源的保留训练特征和领域标签上进行拟合。经过验证校准的边际规则融合两个最可能的专家，而不是承诺于单一的路由专家。在CPSC、PTB-XL、乔治亚州和查普曼-绍兴的数据集上，源感知的专家选择达到了$0.7915 extpm0.0036$的宏F1分数，而匹配的离线独立头参考达到了$0.7885 extpm0.0009$，支持强大的源感知专家保留。在没有源ID的情况下，MLP路由器达到了$0.7756 extpm0.0027$，而前两名边际融合达到了$0.7782 extpm0.0022$。与硬MLP路由相比，前两名的增益较小（$+0.0026$），配对自助法的95 extperthousand置信区间包括零。在三个领域顺序中，前两名与Oracle的差距保持在$0.0111$--$0.0133$，确认自主源推断是主要的剩余瓶颈。没有重放原始ECG，但保留了用于路由器更新的冻结训练特征；因此，该方法并非无内存。
cs.AI / 26 / 2607.01690
Epistemic Goggles: A Pretrained Module that Induces an Epistemic Frame via Gradient Editing
认知眼镜：通过梯度编辑引导认知框架的预训练模块
Abstract
Finetuning a language model on documents that are explicitly annotated as fictional results in a model that still actually believes the documents' core claims, an effect known as Negation Neglect. In our evaluations, models trained on documents prefixed and suffixed with such annotations correctly identify the relevant claims as fictional only about 9% of the time. To address this, we introduce Goggles, a learned module that intervenes on the finetuning gradient rather than the data. During supervised finetuning, a Goggles module edits the gradients an LLM LoRA receives, imparting a chosen epistemic frame (the stance the model takes toward the nature of what it reads) to whatever the documents teach. A Goggles instance is trained once for a given base model, frame, and LoRA configuration, then applied frozen to documents it was never trained on. Trained through Goggles on those same documents, now carrying no fictional annotation, the model flags the content as fictional roughly 91% of the time, while preserving capability (GPQA and TruthfulQA match or exceed baseline). The same architecture supports other frames: a Goggles instance can be trained to treat documents as "part of an AI safety evaluation by Redwood Research" rather than simply as fiction. The imparted frame persists under continued finetuning that pushes back toward the claim, where prior interventions revert. Goggles suggests a path toward training language models on known-misaligned data without absorbing the behaviors that data demonstrates.
Chinese Translation
在对明确标注为虚构的文档进行语言模型微调时，模型仍然会相信文档的核心主张，这种现象被称为否定忽视（Negation Neglect）。在我们的评估中，经过这种标注前后处理的文档训练的模型，仅约9%的时间能够正确识别相关主张为虚构。为了解决这个问题，我们引入了Goggles，一个通过干预微调梯度而非数据的学习模块。在监督微调过程中，Goggles模块编辑大型语言模型（LLM LoRA）接收到的梯度，将所选择的认知框架（模型对所读内容的态度）传递给文档所教授的内容。Goggles实例针对给定的基础模型、框架和LoRA配置进行一次训练，然后在其未曾训练过的文档上冻结应用。经过Goggles在这些同样没有虚构标注的文档上的训练，模型大约91%的时间将内容标记为虚构，同时保持能力（GPQA和TruthfulQA的表现与基线持平或超过）。相同的架构支持其他框架：Goggles实例可以被训练为将文档视为“Redwood Research的AI安全评估的一部分”，而不仅仅是虚构内容。所传递的框架在持续微调中保持有效，即使在推动回归主张的情况下，之前的干预也会恢复。Goggles为在已知不一致数据上训练语言模型提供了一条路径，而不吸收该数据所展示的行为。
cs.AI / 27 / 2607.01709
COMFYCLAW: Self-Evolving Skill Harnesses for Image Generation Workflows
COMFYCLAW：用于图像生成工作流程的自我演化技能框架
Abstract
Agents are increasingly used to construct workflows and assist humans in completing recurring tasks more efficiently. As these workflows become repeated and domain-specific, agent memory and reusable skills become increasingly important: agents should be able to recall workflow patterns, execution constraints, and user preferences from previous runs. We study this problem in workflow-based image generation and introduce COMFYCLAW, an agentic skill evolution harness for controlling ComfyUI workflows. COMFYCLAW formulates workflow construction as typed graph editing, exposes tools organized by construction stage, automatically reverts invalid edits, and uses a region-level vision-language model (VLM) verifier to translate visual failures into actionable repair suggestions. The framework further evolves a progressively disclosed skill library, where trajectories, execution errors, and verifier feedback from previous runs are distilled into reusable Agent Skills. Across four benchmark splits, three agent models, and two image backbones, COMFYCLAW achieves the best average image-generation evaluation score across all six agent configurations, outperforming a verifier-only baseline without skill evolution. Human annotations further show that annotators prefer COMFYCLAW over variants without skill evolution. Our results suggest that skill evolution is an effective mechanism for improving agent reliability and performance in recurring visual workflow construction.
Chinese Translation
代理越来越多地被用于构建工作流程，并帮助人类更高效地完成重复任务。随着这些工作流程的重复性和领域特定性增强，代理的记忆和可重用技能变得愈发重要：代理应能够从之前的运行中回忆起工作流程模式、执行约束和用户偏好。我们研究了基于工作流程的图像生成问题，并引入了COMFYCLAW，这是一种用于控制ComfyUI工作流程的代理技能演化框架。COMFYCLAW将工作流程构建形式化为类型化图编辑，提供按构建阶段组织的工具，自动撤销无效编辑，并使用区域级视觉-语言模型（VLM）验证器将视觉失败转化为可操作的修复建议。该框架进一步演化出一个逐步披露的技能库，其中来自之前运行的轨迹、执行错误和验证器反馈被提炼为可重用的代理技能。在四个基准分割、三个代理模型和两个图像骨干网络上，COMFYCLAW在所有六种代理配置中实现了最佳的平均图像生成评估分数，超越了没有技能演化的仅验证器基线。人类注释进一步表明，注释者更倾向于选择COMFYCLAW，而非没有技能演化的变体。我们的结果表明，技能演化是提高代理在重复视觉工作流程构建中可靠性和性能的有效机制。
cs.AI / 28 / 2607.01710
Generic Expert Coverage for Pruning SparseMixture-of-Experts Language Models
稀疏专家混合语言模型的通用专家覆盖剪枝
Abstract
Sparsely activated Mixture-of-Experts (MoE) language models contain substantial structured redundancy among routed experts, but pruning them without downstream calibration data remains challenging. Existing expert-pruning methods typically rely on a single aggregated importance score, which can bias the retained set toward experts favored by dominant calibration patterns. We propose \textbf{Generic TB-Coverage}, a coverage-aware expert pruning method that uses only generic text corpora (WikiText2 and C4) for calibration. Instead of collapsing expert utility into one score, our method profiles per-expert utility separately on each corpus and enforces a fixed-budget coverage rule that preserves high-utility experts from each corpus before constructing the final pruning mask. Across Qwen1.5-MoE-A2.7B and DeepSeek-MoE-16B-Base at 25\%, 50\%, and 75\% retention budgets, our method improves average accuracy on six common zero-shot benchmarks over random pruning, REAP, and ExpertSparsity, while also reducing perplexity degradation on WikiText2 and C4. The gains are largest under aggressive pruning (25\% and 50\% retain), suggesting that preserving cross-corpus expert coverage is an effective generic-data prior for MoE pruning. Our improvements hold with fixed pruning budgets and no downstream calibration data.
Chinese Translation
稀疏激活的专家混合（Mixture-of-Experts, MoE）语言模型在路由专家之间存在显著的结构冗余，但在没有下游校准数据的情况下进行剪枝仍然具有挑战性。现有的专家剪枝方法通常依赖于单一的聚合重要性评分，这可能会使保留的专家集合偏向于受主导校准模式青睐的专家。我们提出了 extbf{通用TB覆盖（Generic TB-Coverage）}，这是一种基于覆盖的专家剪枝方法，仅使用通用文本语料库（WikiText2和C4）进行校准。我们的方法不是将专家效用合并为一个评分，而是分别在每个语料库上对每个专家的效用进行分析，并强制执行固定预算的覆盖规则，以在构建最终剪枝掩码之前保留来自每个语料库的高效用专家。在Qwen1.5-MoE-A2.7B和DeepSeek-MoE-16B-Base的25\%、50\%和75\%保留预算下，我们的方法在六个常见的零样本基准测试中相较于随机剪枝、REAP和ExpertSparsity提高了平均准确率，同时也减少了在WikiText2和C4上的困惑度下降。在激进剪枝（25\%和50\%保留）下，增益最大，这表明保留跨语料库的专家覆盖是MoE剪枝的有效通用数据先验。我们的改进在固定剪枝预算和没有下游校准数据的情况下依然有效。
cs.AI / 29 / 2607.01715
Distributionally Robust Listwise Preference Optimization
分布鲁棒的列表偏好优化
Abstract
Existing robust preference optimization for language-model alignment mainly studies pairwise supervision and places robustness at the dataset, prompt, or preference-pair level. We instead study listwise preference optimization under ranking-label uncertainty: given a prompt and a candidate list, the observed ranking over that list may be ambiguous due to annotator inconsistency, near-ties, lossy rankwise feedback, or reward-model noise. We propose a pointwise total-variation robust Plackett--Luce objective that directly robustifies the ranking label conditional on the candidate list. The robust loss admits an exact decomposition into the nominal PL loss plus a worst-case PL correction, and the worst-case ranking is obtained by sorting current implicit scores in ascending order, reducing the inner maximization from $K!$ enumeration to $O(K\log K)$. This tractable structure yields strong offline and online optimization guarantees. In the offline fixed-list setting, the robust objective is convex and projected stochastic subgradient reaches global $\epsilon$-suboptimality with $O(\epsilon^{-2})$ sample complexity. In the online policy-induced setting, where candidate lists are generated by the current policy, we establish weak convexity and $\widetilde O(\epsilon^{-2})$ Moreau-envelope stationarity. Experiments in offline LLM alignment show that the proposed robust correction largely preserves performance under clean labels and improves robustness under noise. In online alignment, it makes reward-model-ranked candidate expansion more reliable and improves both reward-model and external GPT-4 judge metrics.
Chinese Translation
现有的针对语言模型对齐的鲁棒偏好优化主要研究成对监督，并将鲁棒性置于数据集、提示或偏好对级别。我们则研究在排名标签不确定性下的列表偏好优化：给定一个提示和一个候选列表，观察到的该列表上的排名可能由于标注者不一致、近似平局、损失的排名反馈或奖励模型噪声而模糊。我们提出了一种点对点总变差鲁棒的 Plackett--Luce 目标，该目标直接对候选列表条件下的排名标签进行鲁棒化。鲁棒损失可以精确分解为名义 PL 损失加上最坏情况 PL 修正，最坏情况排名通过对当前隐式分数进行升序排序获得，从而将内部最大化从 $K!$ 枚举减少到 $O(K ext{log} K)$。这种可处理的结构提供了强大的离线和在线优化保证。在离线固定列表设置中，鲁棒目标是凸的，投影随机次梯度以 $O( ext{ε}^{-2})$ 的样本复杂度达到全局 $ ext{ε}$-次优性。在在线策略诱导设置中，候选列表由当前策略生成，我们建立了弱凸性和 $ ilde O( ext{ε}^{-2})$ 的 Moreau-envelope 稳定性。在离线 LLM 对齐实验中，所提的鲁棒修正在干净标签下大幅保持性能，并在噪声下提高鲁棒性。在在线对齐中，它使得基于奖励模型的排名候选扩展更加可靠，并改善了奖励模型和外部 GPT-4 评判指标。
cs.AI / 30 / 2607.01729
DRL-CLBA: A Clean Label Backdoor Attack for Speech Classification via DDPG Reinforcement Learning
DRL-CLBA：一种基于深度确定性策略梯度的清洁标签后门攻击用于语音分类
Abstract
Deep learning models for speech classification are vulnerable to backdoor attacks, where malicious triggers cause misclassification at inference time. While sample-specific attacks can bypass many defenses, they often rely on poisoned label attack, making them detectable via manual data defense. In this paper, we propose DRL-CLBA, a novel clean label backdoor attack for speech classification that leverages Deep Deterministic Policy Gradient (DDPG) reinforcement learning. We also utilize deep audio steganography to embed sample-specific triggers into source audio, creating feature-space anchors. The proposed reinforcement learning framework effectively optimizes target samples toward trigger-bearing anchor points in the model's deep latent space, enabling label-migration-free poisoning of target samples. Experimental results across three datasets and four different DNNs demonstrate that DRL-CLBA achieves a high attack success rate, effectively bypassing some backdoor defenses. The attack demonstrates strong resistance against fine-tuning, pruning, and spectral signature defenses, exposing critical vulnerabilities in speech-controlled systems.
Chinese Translation
语音分类的深度学习模型易受后门攻击，恶意触发器会导致推理时的误分类。尽管样本特定的攻击可以绕过许多防御措施，但它们通常依赖于污染标签攻击，使其可以通过手动数据防御被检测到。本文提出了一种新颖的清洁标签后门攻击DRL-CLBA，该攻击利用深度确定性策略梯度（DDPG）强化学习进行语音分类。我们还利用深度音频隐写术将样本特定的触发器嵌入源音频中，创建特征空间锚点。所提出的强化学习框架有效地优化目标样本朝向模型深层潜在空间中的承载触发器的锚点，从而实现无标签迁移的目标样本污染。在三个数据集和四种不同深度神经网络（DNN）上的实验结果表明，DRL-CLBA实现了高攻击成功率，有效绕过了一些后门防御。该攻击对微调、剪枝和谱特征防御表现出强大的抵抗力，暴露了语音控制系统中的关键漏洞。
cs.AI / 31 / 2607.01734
Reformalization of the Jordan Curve Theorem
乔丹曲线定理的重形式化
Abstract
We present a case study in reformalization, a variant of autoformalization in which the input proof is not natural language but a formal development in a different proof assistant. Concretely, we report three reformalizations of the Jordan Curve Theorem: from Mizar to Lean, from HOL Light to Lean, and from HOL Light to Agda. We analyse the results and identify pipeline design choices that matter for practical reformalization tasks.
Chinese Translation
我们展示了重形式化的案例研究，这是一种自我形式化的变体，其中输入证明不是自然语言，而是另一种证明助手中的形式化发展。具体而言，我们报告了乔丹曲线定理的三次重形式化：从 Mizar 到 Lean，从 HOL Light 到 Lean，以及从 HOL Light 到 Agda。我们分析了结果，并识别出对实际重形式化任务重要的管道设计选择。
cs.AI / 32 / 2607.01740
Meta-Benchmarks for Financial-Services LLM Evaluation
金融服务大语言模型评估的元基准
Abstract
Public LLM leaderboards optimise for global average performance and do not capture the specific cognitive demands of financial-services work: a model that leads on MMLU-Pro may underperform on document-grounded compliance reasoning, and a coding leader may handle multi-turn customer interactions poorly. We present a meta-benchmarking framework that organises 452 publicly reported benchmarks into 41 O*NET Generalized Work Activities and aggregates those into 38 BIAN banking business domains spanning sales, operations, risk, and support work. A multiplicative weighting scheme (discrimination x coverage x recency), computed over a rolling model window, rewards benchmarks that still separate the best models, are widely reported, and remain in active use, suppressing saturated legacy tests automatically. These weights scale the K-factor in a pairwise Elo tournament, producing cross-benchmark-comparable work-activity scores without raw score normalisation; business-domain scores are weighted averages of the constituent work-activity Elos. We demonstrate the framework on a point-in-time public snapshot covering 288 models across 25 organisations as of June 2026, and describe the methodology, full taxonomy, design decisions, and limitations with the aim of making the approach reproducible for institutions facing similar selection and governance challenges.
Chinese Translation
公共大语言模型排行榜优化全球平均性能，但未能捕捉金融服务工作特定的认知需求：在 MMLU-Pro 上表现突出的模型可能在基于文档的合规推理上表现不佳，而在编码方面表现优异的模型可能在多轮客户互动中表现不佳。我们提出了一种元基准框架，将 452 个公开报告的基准组织为 41 个 O*NET 一般工作活动，并将其汇总为涵盖销售、运营、风险和支持工作的 38 个 BIAN 银行业务领域。通过在滚动模型窗口上计算的乘法加权方案（区分度 x 覆盖率 x 时效性），奖励那些仍能区分最佳模型、被广泛报告并保持活跃使用的基准，同时自动抑制饱和的遗留测试。这些权重在成对 Elo 锦标赛中缩放 K 因子，生成可跨基准比较的工作活动分数，而无需原始分数标准化；业务领域分数是构成工作活动 Elo 的加权平均。我们在截至 2026 年 6 月的 25 个组织的 288 个模型的时间点公共快照上展示了该框架，并描述了方法论、完整分类法、设计决策和局限性，旨在使该方法对面临类似选择和治理挑战的机构可重复。
cs.AI / 33 / 2607.01754
Path-level Hindsight Instructions for Semantic Exploration in Vision-Language Navigation
路径级回顾指令在视觉-语言导航中的语义探索
Abstract
On-policy exploration is a crucial component for training robust Vision-Language Navigation agents, as it exposes the policy to a broader state distribution. However, such exploration inevitably leads to trajectories that deviate from expert demonstrations, resulting in a semantic mismatch between the executed visual stream and the original language instruction. In this work, we address this challenge by introducing Phi-Nav, a unified on-policy framework that leverages hindsight reasoning to align instructions with the agent's actual exploratory journey. Specifically, Phi-Nav operates through a three-stage dual-supervision cycle: 1) the agent performs oracle-guided on-policy exploration, sampling a trajectory while learning from expert action feedback, 2) a hindsight speaker synthesizes a path-level hindsight instruction grounded in the collected visual observations, and 3) the agent conducts a second imitation pass, treating the synthesized trajectory-instruction pair as an additional expert demonstration. Through this process, Phi-Nav bridges the critical semantic supervision gap inherent in on-policy methods, transforming semantically unlabeled movement into dense training signals. Evaluations on the R2R-CE and RxR-CE benchmarks show that Phi-Nav yields competitive performance while requiring only a fraction of the expert demonstrations used by current baselines. These results underscore the necessity of semantic exploration in VLN, positioning Phi-Nav as an effective solution for training embodied agents with limited data.
Chinese Translation
在策略探索中，在线探索是训练稳健的视觉-语言导航（Vision-Language Navigation）代理的重要组成部分，因为它使策略接触到更广泛的状态分布。然而，这种探索不可避免地导致轨迹偏离专家演示，从而造成执行的视觉流与原始语言指令之间的语义不匹配。在本研究中，我们通过引入Phi-Nav来解决这一挑战，Phi-Nav是一个统一的在线框架，利用回顾推理将指令与代理的实际探索旅程对齐。具体而言，Phi-Nav通过三个阶段的双重监督循环进行操作：1）代理执行由神谕引导的在线探索，采样轨迹并从专家动作反馈中学习；2）回顾发言者基于收集的视觉观察合成路径级回顾指令；3）代理进行第二次模仿，通过将合成的轨迹-指令对视为额外的专家演示。通过这一过程，Phi-Nav弥补了在线方法中固有的关键语义监督差距，将语义未标记的运动转化为密集的训练信号。在R2R-CE和RxR-CE基准上的评估表明，Phi-Nav在性能上具有竞争力，同时仅需使用当前基线所需的专家演示的一小部分。这些结果强调了在视觉-语言导航中进行语义探索的必要性，使Phi-Nav成为在有限数据下训练具身代理的有效解决方案。
cs.AI / 34 / 2607.01764
Mastermind: Strategy-grounded Learning for Repository-Scale Vulnerability Reproduction
Mastermind：基于策略的仓库级漏洞重现学习
Abstract
Repository-level vulnerability reproduction is a demanding software engineering (SE) task: an agent must inspect a codebase, infer the input grammar that reaches a vulnerable path, construct a proof-of-conceptv(PoC), and verify that the crash disappears on the patched build. Recent LLM agents can often execute these steps when the approach is correct, yet they still fail by choosing the wrong strategy. This paper argues that strategy, rather than the full action trajectory, is the right learning unit for such SE agents: it is compact enough to optimize, concrete enough to guide execution, and stable enough to store and reuse across attempts. We present Mastermind, a dual-loop framework that separates transferable strategy learning from task-specific experience. A trainable planner learns reusable vulnerability-reproduction strategies through SFT and milestone-based GRPO, while an experience loop maintains task-local strategy records that guide subsequent attempts. The planner is trained independently of the executor, allowing strategy learning to improve multiple frozen executors without modifying their action-generation capability. We evaluate Mastermind on CyberGym using 260 training tasks and 200 held-out evaluation tasks. With GPT-5.5 as the frozen executor, Mastermind achieves an 84.5% pass rate, outperforming open-book PoC context (60.0%), Best-of-8 sampling (63.0%), and iterative improvement (77.0%). The same planner also improves GPT-5.4 mini and GLM~5.1 from 45.0% and 58.5% to 60.0% and 71.0%. These results demonstrate that learning high-level strategies is an effective and transferable mechanism for improving repository-scale SE agents.
Chinese Translation
仓库级漏洞重现是一项要求极高的软件工程（SE）任务：代理必须检查代码库，推断出到达漏洞路径的输入语法，构建概念验证（PoC），并验证在修补后的构建中崩溃是否消失。最近的LLM代理在方法正确时通常能够执行这些步骤，但仍然因选择错误的策略而失败。本文认为，策略而非完整的行动轨迹是此类SE代理的正确学习单元：它足够紧凑以便优化，具体到足以指导执行，并且足够稳定以便在尝试中存储和重用。我们提出了Mastermind，一个双循环框架，将可转移的策略学习与任务特定经验分开。一个可训练的规划器通过SFT和基于里程碑的GRPO学习可重用的漏洞重现策略，而经验循环维护任务本地的策略记录，以指导后续尝试。规划器的训练独立于执行器，使得策略学习能够在不修改其行动生成能力的情况下改善多个冻结的执行器。我们在CyberGym上使用260个训练任务和200个保留评估任务对Mastermind进行了评估。在使用GPT-5.5作为冻结执行器的情况下，Mastermind达到了84.5%的通过率，超越了开放书籍PoC上下文（60.0%）、最佳8次采样（63.0%）和迭代改进（77.0%）。同样的规划器还将GPT-5.4 mini和GLM~5.1的表现从45.0%和58.5%提升至60.0%和71.0%。这些结果表明，学习高层次策略是提高仓库级SE代理的有效且可转移的机制。
cs.AI / 35 / 2607.01766
SimWorlds: A Multi-Agent System for Dynamic 3D Scene Creation
SimWorlds：一种用于动态3D场景创建的多智能体系统
Abstract
LLM agents are increasingly used to translate natural language into 3D scenes in a procedural way, but existing systems focus on static output. Dynamic 4D scenes from text alone, in which liquids flow, particles emit, rigid bodies cascade, and articulated mechanisms move, remain largely unexplored despite their value as editable content and as physics-grounded training data for video generation and embodied AI. Two challenges set the dynamic case apart from static text-to-scene work: an agent must jointly coordinate spatial layout, multiple physics solvers, temporal sequencing, camera, and lighting in a single coherent scene, and verifying motion correctness from rendered video is fundamentally harder than judging a single image. We present SimWorlds: a multi-agent framework that produces dynamic, editable 4D scenes from text, with Blender-specific procedural knowledge, a planner-coder-reviewer workflow driving a fixed ordered sequence of construction stages, a layered scene protocol enforced by a deterministic verifier, and a runtime-state inspection tool suite that catches mechanism failures the rendered image cannot reveal. We also introduce 4DBuildBench, a benchmark for assessing both visual fidelity and physical consistency of the procedural dynamic 3D scenes generated from text prompts. Experiments show that SimWorlds outperforms prior dynamic Blender generation baselines.
Chinese Translation
大语言模型（LLM）代理越来越多地被用于以程序化方式将自然语言转换为3D场景，但现有系统主要集中于静态输出。尽管动态4D场景（如液体流动、粒子发射、刚体级联和关节机制运动）作为可编辑内容和基于物理的训练数据在视频生成和具身人工智能中具有重要价值，但仅从文本生成此类场景仍然未被充分探索。动态场景与静态文本到场景工作之间的两个主要挑战在于：代理必须在一个连贯的场景中共同协调空间布局、多个物理求解器、时间序列、相机和照明；而从渲染视频中验证运动的正确性比判断单幅图像要困难得多。我们提出了SimWorlds：一个多智能体框架，能够从文本生成动态、可编辑的4D场景，具有Blender特定的程序知识、一个驱动固定顺序构建阶段的规划-编码-审查工作流程、由确定性验证器强制执行的分层场景协议，以及一个运行时状态检查工具套件，能够捕捉渲染图像无法揭示的机制故障。我们还介绍了4DBuildBench，这是一个用于评估从文本提示生成的程序化动态3D场景的视觉保真度和物理一致性的基准测试。实验表明，SimWorlds在动态Blender生成基准测试中表现优于之前的模型。
cs.AI / 36 / 2607.01767
Repair the Amplifier, Not the Symptom: Stable World-Model Correction for Agent Rollouts
修复放大器，而非症状：代理回滚的稳定世界模型修正
Abstract
As agent planning moves from short tool chains toward persistent workflows with thousands or tens of thousands of steps, failures will occur inside large planning graphs rather than in isolated predictions. Replanning the entire graph after every mistake is neither computationally realistic nor desirable: full-graph replay consumes large context budgets, exposes the LLM to many irrelevant symptoms, and can degrade long-context retrieval. This paper studies the missing component in such systems: a world-model corrector that repairs the failed planning graph in place. We compare two families of correctors. The first is the common engineering approach: scan nodes and edges, choose a suspicious local region, and ask an LLM to repair it. We implement strong engineering LLM correctors and find that they can help, especially when given very large contexts. The second family is our approach, WM-SAR (World-Model Subgraph Amplification Repair): instead of scanning for visible symptoms, it works backward from subgraph amplification, identifies the nodes and edges that keep re-amplifying error, and sends only that causal subgraph to the LLM. Across graph simulations and LLM repair experiments, WM-SAR substantially outperforms engineering correctors under realistic token budgets, achieves near-whole-graph stabilization with a compact region, and gives the LLM a cleaner repair target.
Chinese Translation
随着代理规划从短期工具链转向具有数千或数万步骤的持久工作流，失败将发生在大型规划图中，而不是孤立的预测中。在每次错误后重新规划整个图既不现实也不可取：全图重放消耗大量上下文预算，使大型语言模型（LLM）暴露于许多无关的症状中，并可能降低长上下文检索的效果。本文研究了此类系统中缺失的组成部分：一种世界模型修正器，能够就地修复失败的规划图。我们比较了两类修正器。第一类是常见的工程方法：扫描节点和边，选择一个可疑的局部区域，并请求LLM进行修复。我们实现了强大的工程LLM修正器，发现它们确实可以提供帮助，尤其是在提供非常大的上下文时。第二类是我们的方法，WM-SAR（世界模型子图放大修复）：它不是扫描可见症状，而是从子图放大反向工作，识别出不断放大错误的节点和边，并仅将该因果子图发送给LLM。在图形模拟和LLM修复实验中，WM-SAR在现实的令牌预算下显著优于工程修正器，几乎实现了整个图的稳定化，并为LLM提供了更清晰的修复目标。
cs.AI / 37 / 2607.01773
Verifiable Knowledge Expansion through Retrieval-Grounded Formal Concept Analysis
通过检索驱动的形式概念分析实现可验证的知识扩展
Abstract
Ontology construction requires deciding which objects, attributes, and structural relations should be accepted as valid knowledge. Language models can propose such structures from text, but their outputs can still be unsupported or inconsistent. This paper proposes a retrieval-augmented small language model (SLM) framework that uses formal concept analysis (FCA) as a symbolic verification loop for knowledge expansion. Starting from seed attributes, FCA proposes implications over a growing formal context. A retrieval-grounded SLM oracle then validates each implication or returns a counterexample. The oracle also supports incidence judgments, consistency checks, and attribute proposals, making accepted implications, counterexamples, contradictions, and corrections inspectable. In a rare ataxia setting constructed from Orphadata resources, retrieval-grounded 10-seed runs obtain relation F1 of 0.29-0.52 and closure-based implication F1 of 0.22-0.30. Larger seed sets increase the number of evaluated implications and often improve implication F1. The lower implication scores reflect a stricter evaluation of derived implications, where one missed or extra relation can affect several implication judgments. Ablations show that incidence judgments in a fixed object-attribute setting can improve closure-based implication scores. However, identifying positive object-attribute pairs remains difficult even when the candidate objects and attributes are fixed.
Chinese Translation
本体构建需要决定哪些对象、属性和结构关系应被视为有效知识。语言模型可以从文本中提出这样的结构，但其输出仍可能缺乏支持或不一致。本文提出了一种检索增强的小型语言模型（SLM）框架，该框架利用形式概念分析（FCA）作为知识扩展的符号验证循环。从种子属性开始，FCA在不断增长的形式语境中提出蕴含。检索驱动的SLM神谕随后验证每个蕴含或返回反例。该神谕还支持事件判断、一致性检查和属性提议，使得接受的蕴含、反例、矛盾和修正可以被检查。在从Orphadata资源构建的罕见共济失调设置中，检索驱动的10种子运行获得了关系F1为0.29-0.52，基于闭包的蕴含F1为0.22-0.30。更大的种子集增加了评估蕴含的数量，并且通常提高蕴含F1。较低的蕴含分数反映了对派生蕴含的更严格评估，其中一个遗漏或额外的关系可能影响多个蕴含判断。消融实验表明，在固定对象-属性设置中，事件判断可以提高基于闭包的蕴含分数。然而，即使候选对象和属性固定，识别正向对象-属性对仍然是困难的。
cs.AI / 38 / 2607.01774
Subliminal Clocks: Latent Time Modelling in Diffusion Language Models
潜意识时钟：扩散语言模型中的潜在时间建模
Abstract
Diffusion Language Models (DLMs) have recently emerged as a promising alternative to autoregressive models. Unlike standard diffusion-based approaches, DLMs are not explicitly conditioned on a timestep, raising a natural question: do these models internally represent denoising progress, and how is such information used downstream? In this work, we show that DLMs do in fact encode a latent representation related to the diffusion timestep within their residual streams. We find that this signal can be reliably extracted using probes across layers, indicating that denoising progress is decodable from internal activations. We further demonstrate that steering the model along a low-dimensional subspace associated with the inferred timestep allows us to systematically modulate its notion of denoising progress, leading to predictable changes in model confidence and entropy. Finally, we analyse the geometry of the identified representation, showing that it exhibits structured and interpretable properties in activation space, and shedding light on how such a signal is processed by these models.
Chinese Translation
扩散语言模型（DLMs）最近作为自回归模型的有希望的替代方案而出现。与标准的基于扩散的方法不同，DLMs并未明确地以时间步长为条件，这引发了一个自然的问题：这些模型是否在内部表示去噪进程，以及这些信息在下游是如何使用的？在这项工作中，我们展示了DLMs确实在其残差流中编码了与扩散时间步长相关的潜在表示。我们发现，这一信号可以通过跨层探针可靠地提取，表明去噪进程可以从内部激活中解码。我们进一步证明，引导模型沿着与推断时间步长相关的低维子空间，可以系统地调节其对去噪进程的理解，从而导致模型置信度和熵的可预测变化。最后，我们分析了所识别表示的几何特征，显示其在激活空间中展现出结构化和可解释的特性，并阐明了这些模型如何处理此类信号。
cs.AI / 39 / 2607.01793
Safety Testing LLM Agents at Scale: From Risk Discovery to Evidence-Grounded Verification
大规模安全测试LLM代理：从风险发现到证据基础验证
Abstract
LLM agents increasingly perform autonomous actions through external tools, leading to complex and evolving safety risks. However, existing safety testing targets expert-designed safety violations, and the corresponding outcomes are evaluated by hard-coded rules, making them costly to extend as agents evolve. To this end, we present Vera, an end-to-end automated safety testing framework that instantiates software engineering testing principles for non-deterministic agents through a three-stage, self-reinforcing pipeline. First, a literature-driven exploration continuously discovers and structures emerging risks into taxonomies of safety risks, attack methods, and tool execution environments. Second, combinatorial composition across taxonomy dimensions produces executable safety cases, each specifying a concrete safety goal, a programmatically constructed initial state, and a deterministic verification predicate grounded in observable artifacts. Third, adaptive execution runs heterogeneous agents in isolated sandboxes where a control agent steers multi-turn interaction based on runtime observations, while evidence-grounded verifiers judge outcomes from environment state and tool-call evidence rather than model self-report. We evaluate Vera on four production agent frameworks (OpenClaw, Hermes, Codex, Claude Code), revealing substantial safety weaknesses, with average attack success rates reaching 93.9\% under multi-channel attacks; we also release Vera-Bench, comprising 1600 executable safety cases spanning 124 risk categories across three execution settings. These results indicate that modular, executable testing infrastructure is essential for rigorous and maintainable safety evaluation of rapidly evolving agentic systems at scale. The code is publicly available at https://github.com/Yunhao-Feng/Vera.
Chinese Translation
LLM代理越来越多地通过外部工具执行自主操作，导致复杂且不断演变的安全风险。然而，现有的安全测试主要针对专家设计的安全违规行为，其相应结果通过硬编码规则进行评估，这使得随着代理的演变而扩展变得成本高昂。为此，我们提出了Vera，一个端到端的自动化安全测试框架，通过一个三阶段、自我增强的流程为非确定性代理实现软件工程测试原则。首先，基于文献的探索不断发现并构建新兴风险的分类，包括安全风险、攻击方法和工具执行环境。其次，跨分类维度的组合生成可执行的安全案例，每个案例指定一个具体的安全目标、一个程序构建的初始状态，以及一个基于可观察文物的确定性验证谓词。第三，适应性执行在隔离的沙盒中运行异构代理，其中控制代理根据运行时观察引导多轮交互，而基于证据的验证者则根据环境状态和工具调用证据判断结果，而非模型自我报告。我们在四个生产代理框架（OpenClaw、Hermes、Codex、Claude Code）上评估了Vera，揭示了显著的安全弱点，在多通道攻击下，平均攻击成功率达到93.9%；我们还发布了Vera-Bench，包含1600个可执行的安全案例，涵盖三个执行设置下的124个风险类别。这些结果表明，模块化、可执行的测试基础设施对于快速演变的代理系统的大规模严格和可维护的安全评估至关重要。代码已公开发布在 https://github.com/Yunhao-Feng/Vera。
cs.AI / 40 / 2607.01814
MMIR-TCM: Memory-Integrated Multimodal Inference and Retrieval for TCM Clinical Decision Support
MMIR-TCM：用于中医临床决策支持的记忆集成多模态推理与检索
Abstract
Traditional Chinese Medicine (TCM) diagnosis, particularly through tongue inspection, faces persistent challenges in subjectivity and reproducibility. The application of multimodal artificial intelligence to TCM clinical tasks, such as syndrome differentiation and prescription generation, is significantly hampered by the semantic gap between visual tongue features and textual reasoning, as well as the lack of large-scale, standardized datasets. To address these challenges, we introduce MMIR-TCM, a novel framework that emulates the diagnostic process of TCM experts by integrating multimodal large language model(MLLM) with memory-augmented segmentation and retrieval-augmented generation (RAG). Employing a three-stage architecture, MMIR-TCM integrates a training-free Memory-SAM module for robust tongue extraction, a fine-tuned Qwen3-VL model for structured tongue diagnosis generation, and a Qwen3-based RAG component for evidence-grounded clinical decision support generation. The framework was developed and validated using MedTCM, a new large-scale multimodal dataset that we introduce specifically for advanced TCM research. To properly evaluate our framework's clinical accuracy, which existing metrics fail to capture, we also developed TDEU, a domain-specific evaluation metric incorporating semantic understanding and diagnostic importance. Our comprehensive experiments demonstrate that MMIR-TCM significantly outperforms leading models, including GPT-4o and Gemini 2.5 Flash.
Chinese Translation
传统中医学（TCM）诊断，尤其是通过舌象检查，面临着主观性和可重复性方面的持续挑战。多模态人工智能在中医临床任务中的应用，如症候辨识和处方生成，受到视觉舌象特征与文本推理之间的语义差距以及缺乏大规模标准化数据集的显著制约。为了解决这些挑战，我们提出了MMIR-TCM，这是一种新颖的框架，通过将多模态大语言模型（MLLM）与记忆增强的分割和检索增强生成（RAG）相结合，模拟中医专家的诊断过程。MMIR-TCM采用三阶段架构，集成了一个无训练的Memory-SAM模块用于稳健的舌象提取，一个微调的Qwen3-VL模型用于结构化舌象诊断生成，以及一个基于Qwen3的RAG组件用于基于证据的临床决策支持生成。该框架的开发和验证使用了MedTCM，这是一个我们专门为先进中医研究引入的新型大规模多模态数据集。为了准确评估我们框架的临床准确性，而现有指标无法捕捉到这一点，我们还开发了TDEU，这是一种结合语义理解和诊断重要性的领域特定评估指标。我们的综合实验表明，MMIR-TCM在性能上显著优于包括GPT-4o和Gemini 2.5 Flash在内的领先模型。
cs.AI / 41 / 2607.01829
Pre-Flight: A Benchmark for Evaluating Large Language Models on Aviation Operational Knowledge
Pre-Flight：评估大型语言模型在航空运营知识上的基准
Abstract
Large language models (LLMs) are increasingly proposed for aviation business operations, from documentation and training generation to customer facing assistants. General purpose benchmarks do not measure whether a model reasons safely and correctly about aviation specific operational knowledge, and the high stakes, regulated nature of the domain makes that gap consequential. We present Pre-Flight, an open source benchmark of 300 multiple choice questions drawn from international standards and airport ground operations material, covering international airport ground operations, ICAO and US FAA regulations, aviation general knowledge and complex operational scenarios. Questions were authored and reviewed by practitioners with experience in air traffic management, ground operations and commercial flying. We evaluate a range of contemporary commercial and open weight models using the Inspect evaluation framework, scoring by accuracy under a standard multiple choice protocol, and we maintain the leaderboard on a rolling basis as new models are released. Against an informal expert reference of around 95%, obtained from a low sample quiz of aviation professionals at a conference, even the strongest model evaluated (released in 2026) reaches 82.7%, having improved only gradually from roughly 75% in early 2025. A substantial and persistent gap below expert level reliability therefore remains. We release the dataset, the evaluation harness and the results, and the benchmark is available within the community evaluations package distributed with inspect_evals. We argue that domain specific evaluation of this kind is a necessary precondition for responsible deployment of generative AI in non safety critical aviation operations.
Chinese Translation
大型语言模型（LLMs）在航空业务运营中的应用日益增多，从文档和培训生成到客户服务助手。然而，通用基准无法衡量模型在航空特定运营知识方面的推理是否安全和正确，而该领域的高风险和受监管特性使得这一差距变得尤为重要。我们提出了Pre-Flight，这是一个开放源代码的基准，包含300道多项选择题，题目来源于国际标准和机场地面操作材料，涵盖国际机场地面操作、国际民航组织（ICAO）和美国联邦航空局（FAA）法规、航空一般知识和复杂的操作场景。这些问题由具有空中交通管理、地面操作和商业飞行经验的从业者撰写和审核。我们使用Inspect评估框架评估一系列当代商业和开放模型，通过标准多项选择协议的准确性进行评分，并在新模型发布时持续维护排行榜。与从航空专业人士在会议上进行的低样本测验获得的约95%的非正式专家参考相比，即使是评估的最强模型（预计于2026年发布）也仅达到82.7%，其准确率从2025年初的约75%仅逐渐提升。因此，专家级可靠性下仍然存在显著且持续的差距。我们发布了数据集、评估工具和结果，该基准可在与inspect_evals一起分发的社区评估包中获得。我们认为，这种领域特定的评估是负责任地在非安全关键的航空操作中部署生成性人工智能的必要前提。
cs.AI / 42 / 2607.01840
Actual causality in fault trees
故障树中的实际因果关系
Abstract
Fault trees are a widely used as effective risk models for complex systems, answering the question "what can go wrong?", especially through minimal cut set analysis. We study fault trees from the perspective of Halpern & Pearl's theory of actual causality. This allows us to use fault trees to answer the question "why has it gone wrong?", which is fundamental to failure diagnostics. We give a complete classification of each of the different notions of actual causality in terms of the fault tree's graph structure and logical structure, and show how minimal cut sets give rise to actual causes.
Chinese Translation
故障树作为复杂系统有效的风险模型被广泛使用，能够回答“可能出现什么问题？”这一问题，特别是通过最小切割集分析。我们从 Halpern 和 Pearl 的实际因果关系理论的角度研究故障树。这使我们能够利用故障树回答“为什么会出现问题？”这一根本性的问题，这对于故障诊断至关重要。我们对实际因果关系的不同概念进行了完整的分类，基于故障树的图结构和逻辑结构，并展示了最小切割集如何产生实际原因。
cs.AI / 43 / 2607.01846
CLAP: Closed-Loop Training, Evaluation, and Release Control for Domain Agent Post-training
CLAP：领域代理后训练的闭环训练、评估与发布控制
Abstract
Domain agents often face noisy business data, uncertain post-training gains, offline/application mismatch, and adapter-release risk. This paper presents CLAP (Closed-Loop Agent Post-training), a closed-loop method that converts business data into structured SFT samples, decision-preference samples, holdout sets, risk diagnostics, and release-gate records. CLAP combines data validation, target/evidence normalization, reward/KL diagnosis, offline gates, and application-chain replay to decide whether an adapter is suitable for the target application chain. On five anonymized manufacturing-scenario batches, QLoRA-style LoRA-SFT yields modest average gains: overall score increases by 0.0098, pass rate by 0.0240, and evidence accuracy by 0.0280, while hallucination and wrong facts decrease. Yet only 3 of 5 batches improve, some batches regress, and GRPO exposes high KL risks. Application-chain replay further shows that RAG is necessary for factual extraction; under the same 3B backbone and 100 replay cases, an application-RAG-oriented LoRA-SFT adapter improves value, core fields, and answer-evidence doc/page matching over base+RAG, but increases latency. These results support managing domain-agent post-training through an integrated data-training-evaluation-release loop rather than relying on training completion or a single offline score.
Chinese Translation
领域代理通常面临嘈杂的业务数据、不确定的后训练收益、离线/应用不匹配以及适配器发布风险。本文提出了CLAP（Closed-Loop Agent Post-training），一种将业务数据转换为结构化的SFT样本、决策偏好样本、保留集、风险诊断和发布门记录的闭环方法。CLAP结合了数据验证、目标/证据标准化、奖励/KL诊断、离线门和应用链重放，以决定适配器是否适合目标应用链。在五个匿名制造场景批次中，QLoRA风格的LoRA-SFT产生了适度的平均收益：整体得分提高了0.0098，及格率提高了0.0240，证据准确性提高了0.0280，同时幻觉和错误事实减少。然而，只有5个批次中的3个有所改善，有些批次出现退步，而GRPO暴露出高KL风险。应用链重放进一步表明，RAG对于事实提取是必要的；在相同的3B主干和100个重放案例下，面向应用的RAG导向LoRA-SFT适配器在价值、核心字段和答案-证据文档/页面匹配方面优于基础+RAG，但增加了延迟。这些结果支持通过集成的数据-训练-评估-发布循环来管理领域代理的后训练，而不是依赖训练完成或单一的离线评分。
cs.AI / 44 / 2607.01859
Safety Targeted Embedding Exploit via Refinement
通过精炼实现安全目标嵌入利用
Abstract
Safety training for large language models (LLMs) is conducted predominantly in English, leaving uncertain how well safety mechanisms generalize to low-resource languages and mixed-language code-switching. We show that this creates an epistemic gap in which models confidently generate harmful responses for inputs that fall outside the distribution of their safety training. To study this phenomenon, we introduce STEER (Safety Targeted Embedding Exploit via Refinement), a gradient-guided attack that identifies words contributing most strongly to the model's refusal behavior and iteratively translates them into low-resource languages to suppress refusal while preserving harmful intent. Across six open-source 8B-parameter models, STEER achieves attack success rates of up to 93.0% on JailbreakBench and 96.7% on AdvBench, outperforming random code-switching and Greedy Coordinate Gradient (GCG). The resulting prompts also transfer to GPT-4o-mini, achieving a 35.5% attack success rate without requiring access to the target model, suggesting that the underlying weakness is not specific to a single architecture. These findings demonstrate that safety mechanisms aligned primarily on English cannot be assumed to generalize across multilingual inputs. We argue that improving multilingual safety requires broader coverage during alignment and mechanisms that explicitly detect and abstain on out-of-distribution inputs.
Chinese Translation
大型语言模型（LLMs）的安全训练主要以英语进行，这使得我们不确定安全机制在低资源语言和混合语言代码切换中的泛化能力。我们展示了这造成了一个认识上的差距，使得模型在生成超出其安全训练分布的输入时，自信地产生有害响应。为了研究这一现象，我们引入了STEER（通过精炼实现安全目标嵌入利用），这是一种基于梯度的攻击方法，能够识别对模型拒绝行为贡献最大的词汇，并将其迭代翻译成低资源语言，以抑制拒绝同时保留有害意图。在六个开源的8B参数模型中，STEER在JailbreakBench上的攻击成功率高达93.0%，在AdvBench上为96.7%，优于随机代码切换和贪婪坐标梯度（Greedy Coordinate Gradient, GCG）。生成的提示也可以迁移到GPT-4o-mini上，达到35.5%的攻击成功率，而无需访问目标模型，这表明潜在的弱点并非特定于单一架构。这些发现表明，主要针对英语的安全机制不能假定能够在多语言输入中泛化。我们认为，提高多语言安全性需要在对齐过程中更广泛的覆盖，并且需要能够明确检测和对超出分布输入进行拒绝的机制。
cs.AI / 45 / 2607.01870
CamoNAS: Neural Architecture Search for Enhanced Camouflaged Object Detection
CamoNAS：增强伪装物体检测的神经架构搜索
Abstract
Camouflaged Object Detection (COD) aims to locate and segment objects that blend into their surroundings, presenting challenges due to weak edge cues and ill-defined boundaries. Traditional COD models rely on hand-designed architectures and multi-scale feature fusion, which are often guided by intuition rather than systematic search. This paper introduces CamoNAS, a frequency-aware multi-resolution Neural Architecture Search (NAS) framework for COD. CamoNAS automatically searches both cell-level operations and network-level downsampling paths, forming a hierarchical search space tailored to detect camouflaged objects. Additionally, it adopts an RGB frequency dual-stream architecture, where a learnable wavelet transform complements the RGB spatial stream. CamoNAS achieves state-of-the-art performance on four COD benchmarks (CAMO, COD10K, NC4K, CHAMELEON), highlighting the effectiveness of NAS for COD. Our code is available at https://github.com/rendaweiSIMIT/CamoNAS.
Chinese Translation
伪装物体检测（COD）旨在定位和分割与周围环境融为一体的物体，由于边缘线索微弱和边界不明确，面临诸多挑战。传统的COD模型依赖于手工设计的架构和多尺度特征融合，这些往往是基于直觉而非系统性搜索。本文提出了CamoNAS，一个频率感知的多分辨率神经架构搜索（NAS）框架，用于COD。CamoNAS自动搜索细胞级操作和网络级下采样路径，形成一个层次化的搜索空间，以便于检测伪装物体。此外，它采用了RGB频率双流架构，其中可学习的小波变换补充了RGB空间流。CamoNAS在四个COD基准测试（CAMO、COD10K、NC4K、CHAMELEON）上实现了最先进的性能，突显了NAS在COD中的有效性。我们的代码可在 https://github.com/rendaweiSIMIT/CamoNAS 获取。
cs.AI / 46 / 2607.01874
SkillCoach: Self-Evolving Rubrics for Evaluating and Enhancing Agentic Skill-Use
SkillCoach：自我演化的评估标准用于评估和提升代理技能的使用
Abstract
Skills are becoming a reusable operational layer for LLM agents, encoding SOPs, domain rules, tool workflows, scripts, and validation routines. In realistic skill repositories, overlapping skills make reliable skill-use difficult. Final verifier success is too coarse for both evaluation and training, since an agent may pass through trial and error while selecting distractor skills, skipping required steps, composing workflows incorrectly or omitting final checks. We introduce SkillCoach, a self-evolving rubric framework for evaluating and enhancing agentic skill-use. SkillCoach derives skill-grounded process rubrics from real rollouts and evaluates trajectories along four dimensions: skill selection, skill following, skill composition, and skill-grounded reflection. It keeps the external verifier as a separate outcome signal, allowing process quality to be distinguished from accidental task success. The evolved rubrics further serve as process supervision for selecting high-quality training trajectories. Experiments show that evolved rubrics substantially improve evaluation quality, expose failures hidden by final accuracy, and provide stronger supervision signals than outcome-only filtering for enhancing agentic skill-use.
Chinese Translation
技能正成为大型语言模型（LLM）代理的可重用操作层，编码标准操作程序（SOP）、领域规则、工具工作流程、脚本和验证例程。在现实的技能库中，技能之间的重叠使得可靠的技能使用变得困难。最终验证者的成功率对于评估和训练来说过于粗糙，因为代理在选择干扰技能时可能会经历反复试验，跳过必要步骤，错误地组合工作流程或省略最终检查。我们提出了SkillCoach，一个自我演化的评估框架，用于评估和提升代理技能的使用。SkillCoach 从真实的执行过程中推导出基于技能的过程评估标准，并沿四个维度进行评估：技能选择、技能遵循、技能组合和基于技能的反思。它将外部验证者作为一个独立的结果信号，从而使过程质量能够与偶然的任务成功区分开来。演化后的评估标准进一步作为过程监督，用于选择高质量的训练轨迹。实验表明，演化后的评估标准显著提高了评估质量，揭示了最终准确性掩盖的失败，并提供了比仅基于结果的过滤更强的监督信号，以增强代理技能的使用。
cs.AI / 47 / 2607.01893
Spec-AUF: Accept-Until-Fail Training under Train-Inference Misalignment for Masked Block Drafters
Spec-AUF：在训练-推理不一致下的接受直到失败训练用于掩码块草拟器
Abstract
Speculative decoding accelerates autoregressive generation by drafting a block of tokens that the target model verifies left-to-right, committing only the longest accepted prefix. Block (DLM-style) drafters predict the whole block in parallel, which is fast but trained with a full-block cross-entropy that supervises every position against the gold continuation -- even though inference discards every token after the first rejection. Recent acceptance-aware objectives patch this by reweighting the full-block loss; we instead use teacher-forced learning as a motivation for how supervision should concentrate on the accepted prefix. A mask-only block drafter has no input-side channel for gold-prefix conditioning, so AUF approximates that prefix-sensitive supervision on the loss side by keeping the cross-entropy support only through the drafter's first predicted failure. AUF is a single, detached change to the CE support -- no auxiliary objective, no verifier rollouts, and no change to the inference pipeline or the exactness contract. Within fixed drafter backbones and serving settings on Qwen3-8B, AUF raises the DFlash drafter's average emitted length $\tau$, averaged over six benchmarks, from 2.40 to 2.61, with a gain on every benchmark, and transfers to Domino's two-branch head (2.56 to 2.68). Two findings sharpen the picture: the decay-only baseline reaches higher token accuracy on the shared block mask yet decodes worse, and on DFlash, once AUF truncates the support, the standard exponential position-decay weighting becomes empirically inert.
Chinese Translation
投机解码通过草拟一块令牌来加速自回归生成，目标模型从左到右验证该块，仅承诺最长的接受前缀。块（DLM风格）草拟器并行预测整个块，这种方式快速，但使用全块交叉熵进行训练，监督每个位置与黄金延续的对比——尽管推理在第一次拒绝后会丢弃每个令牌。最近的接受感知目标通过重新加权全块损失来修补这一点；而我们则使用教师强制学习作为监督应如何集中于接受前缀的动机。仅掩码块草拟器没有输入侧通道用于黄金前缀条件，因此AUF通过仅在草拟器的第一次预测失败时保持交叉熵支持，近似这种对前缀敏感的监督。AUF是对CE支持的单一、独立的改变——没有辅助目标，没有验证器回滚，也没有对推理管道或精确性合同的改变。在固定的草拟器骨干和Qwen3-8B的服务设置下，AUF将DFlash草拟器的平均发射长度$ au$，在六个基准上平均，从2.40提高到2.61，在每个基准上均有提升，并转移到Domino的双分支头（从2.56到2.68）。两个发现进一步明确了这一情况：仅衰减基线在共享块掩码上达到更高的令牌准确性，但解码效果较差，而在DFlash上，一旦AUF截断支持，标准的指数位置衰减加权在经验上变得无效。
cs.AI / 48 / 2607.01903
Rethinking Complexity Metrics for LLM-Integrated Applications: Beyond Source Code
重新思考 LLM 集成应用的复杂性度量：超越源代码
Abstract
LLM-integrated applications blend natural language prompts with program code, and much of their runtime behavior originates in the prompt layer rather than in the code itself. Existing complexity metrics, however, operate solely at the code level and therefore overlook this behavioral logic entirely. We present HECATE, the first tool designed to assess complexity in both the prompt and code layers of such applications. Central to HECATE is Prompt-as-Specification, a Hoare-logic-inspired formalism that interprets every prompt as a specification of intended behavior. Grounded in 25 complexity dimensions identified across published taxonomies, the tool generates 52 candidate metrics. We assess each metric against 118 components collected from 18 open-source repositories, relying on maintenance activity derived from version history as an empirical proxy for complexity, and discard any metric that loses significance once code size is accounted for. Only ten metrics withstand this test. Seven belong to our newly introduced set; rather than measuring sheer volume, each tallies structurally distinct elements, such as LLM call sites, memory attributes, and prompt templates, an attribute we call structural breadth. Of the three surviving conventional metrics, RFC exhibits a similar breadth-oriented character, while Halstead N and V survive only as a residual effect of size; our top-performing metrics exceed all three. Crucially, the prompt-layer metrics retain significance even when the strongest code-level metric is added as a covariate, establishing prompt complexity as a dimension in its own right. A final validation on 20 components spanning six held-out repositories shows that the two best-performing metrics continue to predict maintenance effort, supporting their generalizability beyond the training set.
Chinese Translation
LLM 集成应用将自然语言提示与程序代码相结合，其大部分运行时行为源于提示层而非代码本身。然而，现有的复杂性度量仅在代码层面操作，因此完全忽视了这种行为逻辑。我们提出了 HECATE，这是第一个旨在评估此类应用中提示层和代码层复杂性的工具。HECATE 的核心是 Prompt-as-Specification，这是一种受 Hoare 逻辑启发的形式化方法，将每个提示解释为预期行为的规范。该工具基于已发布分类法中识别的 25 个复杂性维度，生成了 52 个候选度量。我们对每个度量进行了评估，使用从 18 个开源代码库收集的 118 个组件，依赖于版本历史中提取的维护活动作为复杂性的经验代理，并剔除任何在考虑代码规模后失去显著性的度量。只有十个度量通过了这一测试。七个属于我们新引入的集合；这些度量并非简单地衡量体量，而是统计结构上不同的元素，例如 LLM 调用点、内存属性和提示模板，我们称之为结构广度。在三种存活的传统度量中，RFC 展现出类似的广度特征，而 Halstead N 和 V 仅作为规模的残余效应存活；我们的最佳度量超越了这三者。重要的是，提示层度量即使在添加最强代码层度量作为协变量时仍保持显著性，确立了提示复杂性作为一个独立维度。对涵盖六个保留代码库的 20 个组件的最终验证显示，两个表现最佳的度量继续预测维护工作量，支持其在训练集之外的普适性。
cs.AI / 49 / 2607.01916
ContextSniper: AntTrail's Token-Efficient Code Memory for Repository-Level Program Repair
ContextSniper：AntTrail 的存储库级程序修复的高效代码记忆
Abstract
Large language model agents can repair real repository issues, but they often spend large context budgets on whole-file reads, broad searches, and long terminal outputs where useful evidence is mixed with irrelevant code and logs. This paper presents ContextSniper, AntTrail's token-efficient code memory layer for repository-level program repair. As the coding specialization of AntTrail's broader agent memory engine, ContextSniper implements the Sniper feature for precision evidence selection: it retrieves candidate code and runtime evidence, ranks it with hybrid retrieval signals, filters long outputs through an intention-aware context gate, and returns compact evidence packets while preserving recoverable source context outside the prompt. We evaluate ContextSniper on SWE-bench Lite with OpenClaw and Claude Code, using 50 task runs per host-agent condition. ContextSniper reduces total token use by 51.5% and logged cost by 36.4% for OpenClaw, and reduces total token use by 38.9% and estimated cost by 27.3% for Claude Code. Submitted-resolution rates decrease slightly, from 26.0% to 24.0% for OpenClaw and from 32.0% to 30.0% for Claude Code. ContextSniper's pilot testing scripts are open-sourced at https://github.com/Calluking/ContextSniper
Chinese Translation
大型语言模型代理能够修复真实的存储库问题，但它们通常在整个文件读取、广泛搜索和长终端输出上消耗大量上下文预算，其中有用的证据与无关的代码和日志混杂在一起。本文提出了 ContextSniper，AntTrail 的高效代码记忆层，用于存储库级程序修复。作为 AntTrail 更广泛的代理记忆引擎的编码专业化，ContextSniper 实现了 Sniper 功能以进行精确证据选择：它检索候选代码和运行时证据，使用混合检索信号对其进行排名，通过意图感知上下文门过滤长输出，并在保留可恢复的源上下文的同时返回紧凑的证据包。我们在 SWE-bench Lite 上使用 OpenClaw 和 Claude Code 对 ContextSniper 进行了评估，每个主机代理条件下进行了 50 次任务运行。ContextSniper 将 OpenClaw 的总令牌使用量减少了 51.5%，记录成本减少了 36.4%；将 Claude Code 的总令牌使用量减少了 38.9%，估计成本减少了 27.3%。提交解决率略有下降，OpenClaw 从 26.0% 降至 24.0%，Claude Code 从 32.0% 降至 30.0%。ContextSniper 的试点测试脚本已开源，地址为 https://github.com/Calluking/ContextSniper
cs.AI / 50 / 2607.01919
ElephantAgent: Contextual State Continuity in Agentic Systems
大象代理：代理系统中的上下文状态连续性
Abstract
Agentic systems enhance their capabilities by invoking external tools and maintaining persistent memory. However, these external dependencies introduce novel attack surfaces. Recent tool and memory poisoning attacks show that maliciously crafted tool descriptors and poisoned memory can covertly bias agent behavior. These threats reflect a deeper issue: the lack of verifiable continuity in the agent's contextual state for planning and execution. We present ElephantAgent, a protocol that enforces Contextual State Continuity to defend against contextual state poisoning. Inspired by prior state-continuity mechanisms (e.g., Nimble), ElephantAgent extends this protection to the evolving contextual state of agentic systems. We define the contextual state as the bounded, security-critical subset of the agent's entire context (e.g., tool state and memory). Before processing each query, ElephantAgent recomputes the digest of the local contextual state and verifies it against the latest authorized digest. Using replicated trusted hardware, ElephantAgent maintains a linearizable ledger of authorized contextual state transitions and detects out-of-band state tampering. To handle in-band semantic abuse, ElephantAgent additionally provides Historical Traceability, enabling conditional post-hoc audit and recovery to a known-good prior state.
Chinese Translation
代理系统通过调用外部工具和保持持久记忆来增强其能力。然而，这些外部依赖引入了新的攻击面。近期的工具和记忆污染攻击表明，恶意构造的工具描述符和被污染的记忆可以隐秘地影响代理的行为。这些威胁反映了一个更深层次的问题：代理在规划和执行过程中缺乏可验证的上下文状态连续性。我们提出了大象代理（ElephantAgent），一种强制执行上下文状态连续性的协议，以防御上下文状态污染。受到先前状态连续性机制（如 Nimble）的启发，大象代理将这种保护扩展到代理系统不断演变的上下文状态。我们将上下文状态定义为代理整个上下文中有限的、关键信息安全的子集（例如，工具状态和记忆）。在处理每个查询之前，大象代理重新计算本地上下文状态的摘要，并与最新的授权摘要进行验证。通过使用复制的可信硬件，大象代理维护一个线性化的授权上下文状态转换账本，并检测带外状态篡改。为了处理带内语义滥用，大象代理还提供历史可追溯性，支持有条件的事后审计和恢复到已知良好的先前状态。
cs.AI / 51 / 2607.01935
A-TMA: Decoupling State-Aware Memory Failures in Long-Term Agent Memory
A-TMA：解耦长期智能体记忆中的状态感知内存故障
Abstract
Long term memory lets LLM agents act as persistent assistants, but user facts change. A useful memory system must know what is true now, what used to be true, and what changed. We study \emph{ghost memory}, a state coordination failure in which old, current, and transition facts coexist in the memory bank, remain mixed during retrieval, and mislead the answer model. We argue that memory systems should be understood and optimized from three levels: bank maintenance, retrieval, and answer time resolution. We propose ATMA, a state aware overlay for existing memory systems. ATMA keeps superseded and transition records in the bank, builds evidence packets for the query's requested state view, and exposes current, historical, and transition labels to QA. We further call for decoupled evaluation of bank, retrieval, and answer level failures, since final QA accuracy can hide where ghost memory occurs. To make this failure measurable, we build LTP (LoCoMo Temporal Plus), a conflict heavy benchmark for ghost memory, and evaluate on LoCoMo for long conversation generalization. On LTP, Graphiti+ATMA improves conflict accuracy by 0.240 absolute over Graphiti. On LoCoMo, Graphiti+ATMA raises temporal F1 from 0.0295 to 0.1705. The gains are host dependent, but they indicate that explicit state roles can reduce memory failures hidden by final QA accuracy.
Chinese Translation
长期记忆使得大型语言模型（LLM）智能体能够作为持久的助手，但用户信息是会变化的。一个有效的记忆系统必须了解当前的真实信息、过去的真实信息以及发生了变化的信息。我们研究了 extit{幽灵记忆}，这是一种状态协调故障，其中旧的、当前的和过渡的事实在记忆库中共存，在检索过程中混合在一起，从而误导答案模型。我们认为，记忆系统应从三个层面进行理解和优化：记忆库维护、检索和答案时间解析。我们提出了ATMA，一种针对现有记忆系统的状态感知覆盖层。ATMA在记忆库中保留被替代和过渡的记录，为查询请求的状态视图构建证据包，并向问答（QA）系统暴露当前、历史和过渡标签。我们进一步呼吁对记忆库、检索和答案层面的故障进行解耦评估，因为最终的QA准确性可能掩盖幽灵记忆的发生。为了使这种故障可测量，我们构建了LTP（LoCoMo Temporal Plus），这是一个针对幽灵记忆的高冲突基准，并在LoCoMo上评估长期对话的泛化能力。在LTP上，Graphiti+ATMA的冲突准确性比Graphiti提高了0.240。在LoCoMo上，Graphiti+ATMA的时间F1从0.0295提高到0.1705。尽管增益依赖于主机，但这些结果表明，明确的状态角色可以减少被最终QA准确性掩盖的内存故障。
cs.AI / 52 / 2607.01942
Atomic Task Graph: A Unified Framework for Agentic Planning and Execution
原子任务图：一种统一的智能规划与执行框架
Abstract
LLM-based agents have shown strong potential for solving complex multi-step tasks, yet existing performance improvements often rely on either scaling to larger backbone models or task-specific fine-tuning. The former incurs substantial computational costs, while the latter typically generalizes poorly across different tasks. Although prompt-based control is training-free and broadly applicable, existing methods still leave input-output dependencies between subtasks implicit in textual trajectories, making verified intermediate results difficult to reuse. To address these limitations, we propose Atomic Task Graph (ATG), a unified control framework for planning and execution. Specifically, ATG maintains an explicit graph to expose dependencies and support reuse. During planning, it recursively decomposes a high-level task into subtasks, forming a sequence of directed acyclic graphs (DAGs) whose evolution can be traced. During execution, the dependencies exposed by ATG allow independent branches to be executed in parallel, thereby improving execution efficiency. When failures are detected, ATG leverages the graph evolution history to localize the error source and repair only the affected region, preserving validated regions unchanged. Experiments show that ATG consistently outperforms strong baselines in success rate and execution efficiency across three interactive benchmarks using only 7B-8B backbones.
Chinese Translation
基于大语言模型（LLM）的智能体在解决复杂的多步骤任务方面展现出了强大的潜力，但现有的性能提升往往依赖于扩大基础模型规模或进行特定任务的微调。前者会产生巨大的计算成本，而后者通常在不同任务之间的泛化能力较差。尽管基于提示的控制方法无需训练且适用范围广，但现有方法仍然在文本轨迹中隐含了子任务之间的输入输出依赖关系，使得验证的中间结果难以重用。为了解决这些限制，我们提出了原子任务图（Atomic Task Graph, ATG），这是一个用于规划和执行的统一控制框架。具体而言，ATG维护一个显式图以揭示依赖关系并支持重用。在规划过程中，它递归地将高层任务分解为子任务，形成一系列有向无环图（DAG），其演变过程可以被追踪。在执行过程中，ATG揭示的依赖关系允许独立分支并行执行，从而提高执行效率。当检测到失败时，ATG利用图的演变历史来定位错误源，并仅修复受影响的区域，保持已验证区域不变。实验表明，ATG在成功率和执行效率方面在三个交互基准测试中始终优于强基线，仅使用7B-8B的基础模型。
cs.AI / 53 / 2607.01977
OntoLearner: A Modular Python Library for Ontology Learning with Large Language Models
OntoLearner：一个基于大型语言模型的模块化本体学习Python库
Abstract
Ontology learning (OL) aims to automatically construct structured knowledge models from text, yet progress remains fragmented across methods, domains, and evaluation practices. Despite decades of research, OL lacks a shared infrastructure for systematic evaluation and ontology access. This absence has hindered progress and fragmented research, leaving the central challenges of OL largely unaddressed. We introduce OntoLearner, a modular, cross-domain, and first-of-its-kind framework that unifies ontology access, large language model (LLM)-driven learning pipelines, and standardized benchmarking. OntoLearner releases 180 machine-readable ontologies spanning 22 domains and provides pipeline-ready datasets with train/dev/test splits for three core OL tasks: term typing, taxonomy discovery, and non-taxonomic relation extraction. Using this infrastructure, we conduct a large-scale empirical study of OL, evaluating 22 retrieval models and 12 LLMs across domains and tasks. The results converge on a finding that reframes the central challenge of OL: failure modes scale with ontological complexity rather than model size or architectural sophistication. The primary bottleneck is not model capability, but a structural mismatch between how models encode knowledge and how ontologies organize it. These findings establish that effective OL is reachable through the cross-domain, multi-task benchmarking enabled by OntoLearner. OntoLearner is open-source (MIT license) at https://github.com/sciknoworg/OntoLearner/.
Chinese Translation
本体学习（Ontology Learning, OL）旨在从文本中自动构建结构化知识模型，但在方法、领域和评估实践方面的进展仍然碎片化。尽管经过数十年的研究，OL缺乏用于系统评估和本体访问的共享基础设施。这一缺失阻碍了进展并导致研究的碎片化，使得OL的核心挑战大多未得到解决。我们介绍了OntoLearner，这是一个模块化的跨领域框架，首创性地统一了本体访问、大型语言模型（Large Language Model, LLM）驱动的学习流程和标准化基准测试。OntoLearner发布了涵盖22个领域的180个机器可读本体，并提供了针对三个核心OL任务（术语分类、分类法发现和非分类关系提取）的管道准备数据集，包括训练/开发/测试划分。利用这一基础设施，我们对OL进行了大规模实证研究，评估了22个检索模型和12个LLM在不同领域和任务中的表现。结果汇聚出一个发现，重新框定了OL的核心挑战：失败模式与本体复杂性相关，而非模型大小或架构复杂性。主要瓶颈并非模型能力，而是模型编码知识的方式与本体组织知识的方式之间的结构不匹配。这些发现表明，通过OntoLearner所支持的跨领域、多任务基准测试，可以实现有效的OL。OntoLearner是开源的（MIT许可证），可在https://github.com/sciknoworg/OntoLearner/获取。
cs.AI / 54 / 2607.01978
Multimodal Knowledge Edit-Scoped Generalization for Online Recursive MLLM Editing
在线递归多模态大语言模型编辑的多模态知识编辑范围泛化
Abstract
Online multimodal knowledge editing requires injecting a continual stream of visual-textual corrections into multimodal large language models (MLLMs) with bounded overhead and minimal disruption to unrelated behaviors. Existing editors mainly emphasize edit reliability and long-horizon stability, but rarely control the semantic boundary of each edit. Our pilot analyses of post-edit behaviors and internal neuronal activities reveal a scope gap behind reliable edits: instance-level success neither guarantees transfer to valid cross-modal variants nor prevents leakage to unrelated inputs, while edit-related cross-modal responses concentrate in deeper semantic layers. Therefore, we formulate Edit-Scoped Generalization, reframing online MLLM editing from merely correcting an instance to controlling the propagation boundary of each edit. To this end, we propose ScopeEdit, a scope-aware online editor that decomposes each update into a modality-local absorption branch and an evidence-gated shared generalization branch. The local branch supports stable edit absorption, whereas the shared branch enables cross-modal propagation only when visual and textual evidence are sufficiently aligned. Both branches perform scope-separated write geometries in orthogonal low-rank spaces and maintain branch-wise preconditioners via Sherman--Morrison recursions, yielding constant per-edit overhead. Extensive experiments across diverse benchmarks, long-horizon edit streams, MLLM backbones, real-world VLKEB scenarios, and complex vision-language architectures show that ScopeEdit consistently improves the trade-off between in-scope cross-modal transfer and out-of-scope locality, while preserving edit reliability, stability and online efficiency. Our code is available at https://github.com/lab-klc/ScopeEdit.
Chinese Translation
在线多模态知识编辑需要将持续的视觉-文本校正流注入多模态大语言模型（MLLMs），同时保持有限的开销和对无关行为的最小干扰。现有的编辑器主要强调编辑的可靠性和长时间稳定性，但很少控制每次编辑的语义边界。我们对后编辑行为和内部神经活动的初步分析揭示了可靠编辑背后的范围差距：实例级的成功既不能保证有效的跨模态变体的转移，也无法防止对无关输入的泄漏，而与编辑相关的跨模态响应则集中在更深的语义层。因此，我们提出了编辑范围泛化（Edit-Scoped Generalization），将在线 MLLM 编辑从单纯的实例校正重新构建为控制每次编辑的传播边界。为此，我们提出了 ScopeEdit，一种具有范围感知的在线编辑器，它将每次更新分解为一个模态局部吸收分支和一个证据门控共享泛化分支。局部分支支持稳定的编辑吸收，而共享分支仅在视觉和文本证据充分对齐时才允许跨模态传播。两个分支在正交低秩空间中执行范围分离的写几何，并通过 Sherman-Morrison 递归保持分支级别的预条件器，从而实现每次编辑的恒定开销。在多样化基准、长时间编辑流、MLLM 主干、真实世界 VLKEB 场景和复杂的视觉-语言架构中进行的大量实验表明，ScopeEdit 一直在范围内的跨模态转移与范围外的局部性之间改善了权衡，同时保持了编辑的可靠性、稳定性和在线效率。我们的代码可在 https://github.com/lab-klc/ScopeEdit 获取。
cs.AI / 55 / 2607.01988
Episodic-to-Semantic Consolidation Without Identity Drift
无身份漂移的情节到语义的整合
Abstract
Long-running adaptive intelligent agents face a structural tension between knowledge consolidation and information integrity. Memory consolidation is conventionally treated as an agent-changing operation: a model is fine-tuned, a prompt rewritten, a policy distilled, or a reflection appended to the context that governs future behaviour. In regulated autonomic deployment this is a liability because the agent operates under commitments and audit contracts that bind to a specific, cryptographically certified identity. We propose to treat consolidation not as a mutation of the planner or the identity manifest, but as a deterministic function f: M^ep -> M^sem over episodic memory whose output is a separately addressable semantic knowledge layer; the identity hash does not read M^sem, so consolidation updates knowledge without changing the agent's certified identity. We give a formal account of the agent representation, prove identity invariance through a structural lemma on the manifest's hash-input set, specify a deterministic aggregation algorithm whose outputs are auditable database rows with explicit confidence and supporting-event provenance, and validate the construction with synthetic experiments demonstrating per-field correctness, byte-equal identity across consolidation passes, and a mean 79.82% reduction in unproductive planner attempts (95% BCa CI [78.02%, 81.49%] across 10 seeds) against a calibrated Bayesian-shrunk baseline. The construction is a knowledge-update discipline for autonomic agents in which lessons accumulate as queryable facts while the agent's certified identity remains byte-equal across its operational lifetime, with an embodied service agent as the running case study.
Chinese Translation
长期运行的自适应智能体面临知识整合与信息完整性之间的结构性紧张。记忆整合通常被视为一种改变智能体的操作：模型被微调，提示被重写，策略被提炼，或者反思被附加到支配未来行为的上下文中。在受监管的自律部署中，这是一种负担，因为智能体在承诺和审计合同下运行，这些合同绑定到特定的、经过密码学认证的身份。我们建议将整合视为一种确定性函数 f: M^ep -> M^sem，作用于情节记忆，其输出是一个可单独寻址的语义知识层；身份哈希不读取 M^sem，因此整合更新知识而不改变智能体的认证身份。我们对智能体表示进行了正式的描述，通过对清单的哈希输入集的结构性引理证明了身份不变性，指定了一种确定性聚合算法，其输出是具有明确置信度和支持事件来源的可审计数据库行，并通过合成实验验证了该构造，展示了每个字段的正确性、整合过程中的字节等同身份，以及在与经过校准的贝叶斯缩减基线相比，平均减少79.82%的无效规划尝试（95% BCa CI [78.02%, 81.49%]，基于10个种子）。该构造为自律智能体提供了一种知识更新机制，其中经验教训以可查询的事实形式积累，同时智能体的认证身份在其操作生命周期内保持字节等同，以一个具体的服务智能体作为运行案例研究。
cs.AI / 56 / 2607.01992
Traceable Fault Diagnosis for Battery Energy Storage Systems via Retrieval-Augmented Multi-Agent O&M Assistant
基于检索增强的多智能体运维助手的可追溯电池储能系统故障诊断
Abstract
Large-scale battery energy storage systems (BESSs) require O&M decisions that combine alarms, cell-level measurements, device topology, diagnostic tables, historical cases, and maintenance documents. Monitoring platforms can flag threshold violations, but they often cannot explain whether voltage inconsistency, resistance drift, short-circuit risk, capacity divergence, or thermal abnormality needs intervention. This digest presents a traceable BESS fault-diagnosis assistant that uses retrieval-augmented multi-agent reasoning to connect operational data, domain knowledge, visual evidence, and report generation. Reliability is improved through BESS-specific task routing, schema-constrained natural-language database access, hybrid text-image retrieval, and evidence-based answer synthesis. Preliminary internal evaluation is reported for routing, database access, and diagnostic reasoning.
Chinese Translation
大规模电池储能系统（BESS）需要结合报警、单元级测量、设备拓扑、诊断表、历史案例和维护文档的运维决策。监测平台可以标记阈值违规，但通常无法解释电压不一致、阻抗漂移、短路风险、容量偏差或热异常是否需要干预。本文介绍了一种可追溯的BESS故障诊断助手，该助手利用检索增强的多智能体推理，将操作数据、领域知识、视觉证据和报告生成连接起来。通过BESS特定任务路由、受模式约束的自然语言数据库访问、混合文本-图像检索和基于证据的答案合成，提高了可靠性。本文报告了路由、数据库访问和诊断推理的初步内部评估。
cs.AI / 57 / 2607.02010
InduceKV: Fixed-Footprint Continual Adaptation of Multimodal LLMs via Inducing KV Memories
InduceKV：通过诱导KV记忆实现多模态大语言模型的固定占用持续适应
Abstract
Multimodal large language models must adapt to evolving tasks and domains, yet continual improvement under bounded deployment footprint remains difficult because repeated parameter updates or growing replay stores can accumulate adaptation state over time. We study fixed-footprint continual adaptation: the deployed adaptation state is kept under a fixed memory budget, while the backbone model is left unchanged and task-specific updates are externalized. We propose InduceKV, a retrieval-based method that stores each selected training prefix as an attention-ready memory entry, consisting of a frozen retrieval key and compact layerwise key--value (KV) payloads that can be appended to the model's self-attention cache. Under a strict memory budget, InduceKV constructs a compact inducing set through bilevel selection: a lightweight calibration is fit for retrieval, while the selected memory balances current-task likelihood, anchor-based retention, and coverage in the frozen retrieval space. Across task-incremental instruction tuning, continual VQA, domain-incremental adaptation, and lifelong multimodal instruction tuning, InduceKV consistently improves over PEFT, MoE, replay, and prompt-retrieval baselines under matched memory budgets. We further report backbone-matched, stage-1 CoIN, compute-matched, and scalability diagnostics, showing that the gains are not due to a stronger backbone, replay alone, or an unbounded candidate pool.
Chinese Translation
多模态大语言模型必须适应不断变化的任务和领域，但在有限的部署占用下持续改进仍然困难，因为重复的参数更新或不断增长的重放存储可能随着时间的推移积累适应状态。我们研究了固定占用的持续适应：已部署的适应状态保持在固定的内存预算之内，而主干模型保持不变，特定任务的更新被外部化。我们提出了InduceKV，一种基于检索的方法，将每个选定的训练前缀存储为一个准备好用于注意力机制的记忆条目，由一个冻结的检索键和紧凑的逐层键值（KV）负载组成，这些负载可以附加到模型的自注意力缓存中。在严格的内存预算下，InduceKV通过双层选择构建一个紧凑的诱导集：一个轻量级的校准适合于检索，而所选的记忆在冻结的检索空间中平衡当前任务的可能性、基于锚点的保留和覆盖。在任务增量指令调优、持续视觉问答（VQA）、领域增量适应和终身多模态指令调优中，InduceKV在匹配的内存预算下始终优于PEFT、MoE、重放和提示检索基线。我们进一步报告了与主干匹配的、阶段1 CoIN、计算匹配和可扩展性诊断，显示出这些收益并非源于更强的主干、单独的重放或无限制的候选池。
cs.AI / 58 / 2607.02020
Hidden Forgetting in Continual Multimodal Learning: When Accuracy Survives but Grounding Fails
持续多模态学习中的隐性遗忘：当准确性得以保留但基础失效时
Abstract
Multimodal large language models must continually adapt to evolving tasks and domains, yet standard continual learning metrics mainly measure whether old answers remain correct, leaving the stability of multimodal grounding largely unexamined. We study this overlooked failure mode and ask whether a continually adapted MLLM can preserve not only what it answers, but also how it uses visual, textual, OCR, chart, and document evidence. We identify \emph{hidden evidence-use forgetting}, where answer accuracy is retained while the model silently shifts toward different or less grounded evidence channels, and propose \textsc{RCL}, a replay-free reliance-constrained continual learning framework. \textsc{RCL} freezes the previous checkpoint as a behavioral reference, estimates teacher and student evidence-reliance profiles through counterfactual channel interventions, and jointly optimizes task learning, prediction preservation, and reliance preservation without adding inference-time cost. Across CoIN, COAST, MCITlib, and an evidence-sensitive multimodal stream, \textsc{RCL} consistently improves final performance and reduces forgetting over replay-free, PEFT, routing, and memory-assisted baselines, while substantially lowering modality reliance drift, dominant evidence flips, and hidden forgetting rates. These results suggest that robust continual multimodal learning requires preserving the evidence path behind correct answers, not merely the answers themselves.
Chinese Translation
多模态大型语言模型必须不断适应不断变化的任务和领域，然而标准的持续学习指标主要衡量旧答案是否仍然正确，导致多模态基础的稳定性在很大程度上未得到检验。我们研究了这一被忽视的失败模式，并询问一个持续适应的多模态大型语言模型（MLLM）是否能够保留其回答的内容，以及如何使用视觉、文本、光学字符识别（OCR）、图表和文档证据。我们识别出 extit{隐性证据使用遗忘}，即答案的准确性得以保留，而模型在未被察觉的情况下转向不同或基础较弱的证据通道，并提出了 extsc{RCL}，一种无重放的依赖约束持续学习框架。 extsc{RCL}将先前的检查点冻结为行为参考，通过反事实通道干预估计教师和学生的证据依赖特征，并在不增加推理时间成本的情况下共同优化任务学习、预测保留和依赖保留。在CoIN、COAST、MCITlib以及一个对证据敏感的多模态流中， extsc{RCL}始终提高最终性能，并在无重放、PEFT、路由和记忆辅助基线中减少遗忘，同时显著降低模态依赖漂移、主导证据翻转和隐性遗忘率。这些结果表明，稳健的持续多模态学习需要保留正确答案背后的证据路径，而不仅仅是答案本身。
cs.AI / 59 / 2607.02032
PACE: A Proxy for Agentic Capability Evaluation
PACE：代理能力评估的代理工具
Abstract
Evaluating LLM agents on benchmarks like SWE-Bench and GAIA can be expensive, time-consuming, and requires complex infrastructure. A single evaluation can cost thousands of dollars and take days to complete. In contrast, non-agentic LLM benchmarks that test individual capabilities (e.g., reasoning, code generation) are fast and cheap to run. In this paper, we investigate whether performance on expensive agentic benchmarks can be accurately predicted by the performance on a small, carefully selected subset of atomic evaluation instances. We introduce PACE, a framework that constructs proxy benchmarks by selecting instances from existing non-agentic evaluations whose aggregate scores most reliably predict model performances on agentic benchmarks. Given a pool of candidate instances spanning atomic capabilities, PACE fits a regression that maps a model's scores on a compact subset of source instances to its score on the target agentic benchmark. The subset itself is curated by combining two complementary instance-selection strategies, target-relevance local selection and globally informative global selection. We apply PACE to the 4 target agentic benchmarks in this paper, which yields PACE-Bench, the concrete proxy benchmark that we evaluate in the paper. Experiments across 14 models, 4 agentic benchmarks, and 19 non-agentic benchmarks show that PACE-Bench predicts agentic scores with leave-one-out cross-validation (LOOCV) mean absolute error (MAE) under 4%, Spearman correlation above 0.80, and pairwise model-ranking accuracy around 85%, all at much less than 1% of the full agentic evaluation cost. We further analyze the selected proxy instances, revealing which skills each agentic benchmark uniquely demands. PACE enables practitioners to obtain reliable estimates of agentic performance during model development, selection, and routing, without the overhead of full agent evaluation.
Chinese Translation
在像SWE-Bench和GAIA这样的基准上评估大型语言模型（LLM）代理可能代价高昂、耗时，并且需要复杂的基础设施。一次评估的成本可能高达数千美元，并且需要几天才能完成。相比之下，测试个体能力（例如推理、代码生成）的非代理LLM基准运行起来快速且便宜。本文探讨了是否可以通过对一小部分精心挑选的原子评估实例的表现，准确预测在昂贵的代理基准上的表现。我们引入了PACE，一个通过从现有非代理评估中选择实例来构建代理基准的框架，这些实例的总分最可靠地预测模型在代理基准上的表现。给定一个涵盖原子能力的候选实例池，PACE拟合一个回归模型，将模型在紧凑的源实例子集上的得分映射到其在目标代理基准上的得分。该子集本身是通过结合两种互补的实例选择策略进行策划的，即目标相关性局部选择和全球信息性全局选择。我们将PACE应用于本文中的4个目标代理基准，得出了PACE-Bench，这是我们在论文中评估的具体代理基准。在14个模型、4个代理基准和19个非代理基准上的实验表明，PACE-Bench在留一交叉验证（LOOCV）下的平均绝对误差（MAE）低于4%，斯皮尔曼相关系数超过0.80，成对模型排名准确率约为85%，所有这些都仅占完整代理评估成本的不到1%。我们进一步分析了所选的代理实例，揭示了每个代理基准独特要求的技能。PACE使从业者能够在模型开发、选择和路由过程中获得可靠的代理性能估计，而无需进行全面的代理评估。
cs.AI / 60 / 2607.02069
Algebraic Model Counting for Global Analysis of Optimal Decision Trees
用于最优决策树全球分析的代数模型计数
Abstract
Ensuring model reliability in Explainable AI requires a global assessment of the hypothesis space. We propose a formal framework for the exhaustive analysis of optimal and near-optimal decision trees, called Algebraic Decision Tree Counting (ADTC). Inspired by Algebraic Model Counting (AMC) in knowledge representation, ADTC reformulates diverse analytical tasks, such as optimization, counting, and sampling, into a unified sum-of-products computation over a semiring $R$. While the hypothesis space of decision trees is doubly exponential with respect to the maximum depth $\Delta$, our dynamic programming algorithm achieves $O^*(n^{O(\Delta)})$ time complexity in the number of features $n$, where $O^*$ suppresses polynomial factors. To handle complex constraints consisting of multiple tree metrics, we introduce model behavior tensors that aggregate semiring values via convolution products over a tensor semiring. This algebraic approach efficiently constructs a model profile that captures the global landscape and trade-offs between criteria such as accuracy, size, and fairness. We demonstrate the utility of our software, emtrees, on real-world datasets, illustrating how ADTC facilitates evidence-based model selection in sensitive domains.
Chinese Translation
确保可解释人工智能中的模型可靠性需要对假设空间进行全球评估。我们提出了一个正式框架，用于对最优和近似最优决策树进行全面分析，称为代数决策树计数（Algebraic Decision Tree Counting, ADTC）。受到知识表示中代数模型计数（Algebraic Model Counting, AMC）的启发，ADTC将优化、计数和采样等多种分析任务重新表述为在半环 $R$ 上的统一乘积和计算。尽管决策树的假设空间相对于最大深度 $ riangle$ 是双指数级的，但我们的动态规划算法在特征数量 $n$ 上实现了 $O^*(n^{O( riangle)})$ 的时间复杂度，其中 $O^*$ 抑制了多项式因子。为了处理由多个树度量组成的复杂约束，我们引入了模型行为张量，通过在张量半环上的卷积积聚合半环值。这种代数方法有效地构建了一个模型特征，捕捉了全球景观以及准确性、大小和公平性等标准之间的权衡。我们在真实世界数据集上展示了我们软件 emtrees 的实用性，说明 ADTC 如何促进在敏感领域进行基于证据的模型选择。
cs.AI / 61 / 2607.02073
Evidence-State Rewards for Long-Context Reasoning
用于长上下文推理的证据状态奖励
Abstract
Long-context reasoning requires models to locate, revise, and synthesize evidence distributed across lengthy inputs. Existing long-context RL methods usually reward final answers or static evidence extraction, offering little feedback on how intermediate actions change the model's evidence state. We propose Maven, a reinforcement learning framework with an editable evidence memory. Maven defines an answer-conditioned evidence-state value and rewards action-level state transitions: add actions are credited by marginal gain and hindsight contribution, link actions by evidence synergy, and drop actions by improved answer support after removing misleading evidence. These rewards are assigned to the corresponding action spans in GRPO. Across Llama and Qwen models on LongBench v2, LongReason, and RULER, Maven outperforms outcome-only RL and evidence-identification baselines, producing more sufficient evidence sets and lower distractor retention. Our results show that long-context RL benefits from optimizing stateful evidence navigation rather than one-shot evidence extraction.
Chinese Translation
长上下文推理要求模型在冗长的输入中定位、修正和综合分散的证据。现有的长上下文强化学习（RL）方法通常只奖励最终答案或静态证据提取，对中间动作如何改变模型的证据状态提供的反馈较少。我们提出了Maven，一个具有可编辑证据记忆的强化学习框架。Maven定义了一个基于答案的证据状态值，并奖励动作级状态转变：添加动作通过边际增益和事后贡献获得积分，链接动作通过证据协同获得积分，而删除动作则通过移除误导性证据后改善答案支持来获得积分。这些奖励被分配到GRPO中的相应动作跨度。在Llama和Qwen模型上进行的LongBench v2、LongReason和RULER实验中，Maven的表现优于仅基于结果的RL和证据识别基线，产生了更充分的证据集和更低的干扰项保留率。我们的结果表明，长上下文RL从优化状态化证据导航中受益，而非一次性证据提取。
cs.AI / 62 / 2607.02087
SUNTA: Hierarchical Video Prediction with Surprise-based Chunking
SUNTA：基于惊讶的分块层次视频预测
Abstract
Hierarchical state-space models (HSSMs) offer a promising approach to long-horizon prediction by segmenting sequences into temporal chunks. However, their performance hinges on how chunk boundaries are determined. While prior HSSMs typically rely on fixed-length chunking or similarity-based boundary detection, these methods often misalign with the intrinsic temporal structure of the data. We argue that chunking should instead be driven by prediction errors, which more directly indicate when longer-range context becomes necessary. Nevertheless, integrating surprise-based chunking into HSSMs introduces critical challenges, including hierarchical collapse during end-to-end training and the absence of surprise signals during open-loop prediction. To address these issues, we propose Surprise-based Nested Temporal Abstraction (SUNTA), a method that employs a decoupled training strategy to preserve surprise signals and uses internal inconsistency as a top-down surprise metric to determine chunk boundaries within imagined rollouts. Experiments on video prediction tasks in 2D and 3D environments demonstrate that SUNTA outperforms baselines, uniquely maintaining accurate predictions over 250 timesteps, whereas all baselines degrade within the first 10 timesteps.
Chinese Translation
层次状态空间模型（HSSMs）通过将序列分割为时间块，为长时间范围预测提供了一种有前景的方法。然而，它们的性能依赖于分块边界的确定方式。虽然先前的HSSMs通常依赖于固定长度的分块或基于相似性的边界检测，但这些方法往往与数据的内在时间结构不一致。我们认为，分块应由预测误差驱动，这更直接地指示出何时需要更长范围的上下文。然而，将基于惊讶的分块集成到HSSMs中引入了关键挑战，包括在端到端训练期间的层次崩溃以及在开放环预测期间缺乏惊讶信号。为了解决这些问题，我们提出了基于惊讶的嵌套时间抽象（SUNTA），该方法采用解耦的训练策略以保留惊讶信号，并使用内部不一致性作为自上而下的惊讶度量来确定想象展开中的分块边界。在二维和三维环境的视频预测任务上的实验表明，SUNTA的表现优于基线，独特地在250个时间步内保持准确预测，而所有基线在前10个时间步内均出现性能下降。
cs.AI / 63 / 2607.02116
ContextNest: Verifiable Context Governance for Autonomous AI Agent
ContextNest：自主人工智能代理的可验证上下文治理
Abstract
Autonomous AI agents increasingly depend on external knowledge stores, yet most retrieval pipelines provide relevance without durable guarantees of provenance, version identity, integrity, traceability, or point-in-time reconstruction. We formalize this as context governance and present ContextNext, an open specification and reference implementation for governed AI-consumable knowledge vaults. ContextNext does not replace Retrieval-Augmented Generation (RAG); it supplies the governance layer beneath retrieval, determining which artifacts are approved, current, attributable, and integrity-verified before retrieval systems operate over them. The specification combines typed Markdown documents with metadata, deterministic set-algebraic selectors, contextnest:// URI references, SHA-256 hash-chained version histories, graph-level checkpoints, source nodes for live data through the Model Context Protocol (MCP), and audit traces of agent context consumption. These mechanisms let organizations reconstruct which knowledge versions informed an agent output and whether those versions were AI-eligible when consumed. We report first empirical results from two controlled experiments. In a stale-version attack isolating the governance-versus-retrieval failure mode, governed selection strictly Pareto-dominates BM25 sparse retrieval, with higher answer-quality pass rate (97% versus 93-90%) at about one-third the input-token cost. In a retrieval-determinism experiment over a 1,060-document corpus, deterministic selectors and BM25 return stable document sets across repeated identical queries (Jaccard 1.0), while a dense+HNSW baseline is non-deterministic on 80% of queries (mean Jaccard 0.611, worst case 0.210). These results suggest that context governance addresses failure modes retrieval quality alone is not designed to resolve. We release a core engine, CLI, and MCP server under open licenses.
Chinese Translation
自主人工智能代理越来越依赖外部知识库，但大多数检索管道仅提供相关性，而没有持久的来源、版本身份、完整性、可追溯性或时间点重建的保证。我们将此形式化为上下文治理，并提出ContextNext，这是一个针对受管控的人工智能可消费知识库的开放规范和参考实现。ContextNext并不替代检索增强生成（Retrieval-Augmented Generation, RAG）；它提供了检索之下的治理层，确定在检索系统操作之前，哪些文献是经过批准的、当前的、可归属的以及经过完整性验证的。该规范结合了带类型的Markdown文档与元数据、确定性集合代数选择器、contextnest:// URI引用、SHA-256哈希链版本历史、图级检查点、通过模型上下文协议（Model Context Protocol, MCP）获取实时数据的源节点，以及代理上下文消费的审计痕迹。这些机制使组织能够重建哪些知识版本影响了代理的输出，以及在消费时这些版本是否符合人工智能的资格。我们报告了来自两个受控实验的初步实证结果。在一个隔离治理与检索失败模式的过时版本攻击中，受管控选择严格优于BM25稀疏检索，具有更高的答案质量通过率（97%对比93-90%），而输入令牌成本约为三分之一。在一个针对1,060文档语料库的检索确定性实验中，确定性选择器和BM25在重复相同查询时返回稳定的文档集（Jaccard 1.0），而密集+HNSW基线在80%的查询中是非确定性的（平均Jaccard 0.611，最坏情况0.210）。这些结果表明，上下文治理解决了仅依赖检索质量无法解决的失败模式。我们以开放许可证发布了核心引擎、命令行界面（CLI）和MCP服务器。
cs.AI / 64 / 2607.02118
Enhancing Fitness Intelligence through Domain-Specific LLM Post-Training
通过领域特定的LLM后训练增强健身智能
Abstract
Scientific Fitness Coaching (SFC) is typically delivered by human professionals, making it costly and inaccessible to many. While recent advances in Large Language Models (LLMs) show considerable promise for more inclusive fitness coaching, directly deploying prevailing general-purpose LLMs in SFC reveals critical limitations. These models often lack sufficient domain-specific knowledge integration, leading to weak performance on complex SFC scenarios. In this paper, we introduce FitOne, a series of fitness LLMs (with 8B and 32B parameters) designed to improve reliability and domain specialization for SFC applications. Built upon the Qwen3 foundation models, FitOne is developed through a three-stage post-training pipeline consisting of continual pre-training, supervised fine-tuning, and reinforcement learning, using large-scale, high-quality datasets derived from rigorous knowledge engineering. We conduct comprehensive evaluations of FitOne on professional fitness certification exams, including ACSM-EP and NSCA-CSCS, as well as general capabilities such as knowledge reasoning and instruction following. Experimental results show that, while retaining strong general capabilities, FitOne-8B/32B achieves average improvements of up to 10.09%/9.29% and 12.73%/7.01% on the ACSM-EP and NSCA-CSCS exams, respectively, compared with the Qwen3 base models. Furthermore, in-depth ablation studies confirm the necessity of each training stage, highlighting the pipeline's effectiveness in balancing domain expertise enhancement with general ability retention. We believe this research advances LLM systems toward more reliable fitness intelligence and will inspire future research on developing domain-specific LLMs.
Chinese Translation
科学健身指导（SFC）通常由人类专业人士提供，这使得其成本高昂且对许多人来说难以获得。尽管近期大型语言模型（LLMs）的进展在更具包容性的健身指导方面显示出相当大的潜力，但直接在SFC中部署现有的通用LLMs却暴露出关键的局限性。这些模型往往缺乏足够的领域特定知识整合，导致在复杂SFC场景中的表现较弱。本文介绍了FitOne，一系列旨在提高SFC应用可靠性和领域专业化的健身LLMs（具有8B和32B参数）。FitOne基于Qwen3基础模型，通过一个由持续预训练、监督微调和强化学习组成的三阶段后训练流程开发，使用来自严格知识工程的大规模高质量数据集。我们对FitOne在专业健身认证考试（包括ACSM-EP和NSCA-CSCS）以及知识推理和指令遵循等一般能力方面进行了全面评估。实验结果表明，尽管保持了强大的通用能力，FitOne-8B/32B在ACSM-EP和NSCA-CSCS考试中分别比Qwen3基础模型平均提高了10.09%/9.29%和12.73%/7.01%。此外，深入的消融研究确认了每个训练阶段的必要性，突显了该流程在增强领域专业知识与保留一般能力之间的有效性。我们相信这项研究推动了LLM系统向更可靠的健身智能发展，并将激励未来在开发领域特定LLMs方面的研究。
cs.AI / 65 / 2607.02134
Coding-agents can replicate scientific machine learning papers
编码代理可以复制科学机器学习论文
Abstract
Scientific machine learning papers typically make computational claims, e.g., that the relative mean square error is less than 5% or that the 95% predictive credible interval covers the test data. A coding agent can be prompted to replicate those claims from paper materials alone, but the prompt does not by itself reliably preserve progress or check whether generated evidence supports the paper's claims. We introduce Paper-replication, a workflow that makes each selected paper claim a target with recorded evidence, and implement it as a coding-agent skill. The workflow makes the agent record those targets, reconstruct the paper's method, run computational experiments, link generated outputs to provenance and comparisons with the paper's claims, record where matched evidence appears in the replication report, and pass validation checks before completion. We evaluate Paper-replication on twelve independent runs across four scientific machine learning papers. All twelve workspaces pass the completion gate, and all 158 recorded targets are matched with report coverage. Even in this completed workspace state, repeated runs differ in how papers are divided into targets, in numerical fidelity to the source papers, in elapsed replication time, in the number of intermediate executions replaced before final evidence is accepted, and in the rules used to accept evidence. Paper-replication makes completion depend on workspace evidence and validation checks rather than on the agent's final message.
Chinese Translation
科学机器学习论文通常提出计算声明，例如，相对均方误差小于5%或95%的预测可信区间覆盖测试数据。可以通过仅使用论文材料来提示编码代理复制这些声明，但该提示本身并不能可靠地保留进展或检查生成的证据是否支持论文的声明。我们引入了Paper-replication（论文复制），这是一种将每个选定论文声明作为目标并记录证据的工作流程，并将其实现为编码代理技能。该工作流程使代理记录这些目标，重建论文的方法，进行计算实验，将生成的输出与来源和与论文声明的比较联系起来，记录匹配证据在复制报告中的出现位置，并在完成之前通过验证检查。我们在四篇科学机器学习论文的十二次独立运行中评估了Paper-replication。所有十二个工作空间都通过了完成门，所有158个记录的目标都与报告覆盖相匹配。即使在这个完成的工作空间状态中，重复运行在论文如何被划分为目标、与源论文的数值一致性、经过的复制时间、在最终证据被接受之前替换的中间执行次数以及用于接受证据的规则上也存在差异。Paper-replication使得完成依赖于工作空间证据和验证检查，而不是代理的最终消息。
cs.AI / 66 / 2607.02141
A$^{2}$utoLPBench: An Auto-Generated, Agent-Friendly LP Benchmark via Inverse-KKT Construction
A$^{2}$utoLPBench：通过逆KKT构造生成的、适合代理的线性规划基准测试
Abstract
Most LP-from-text benchmarks are static datasets of word problems written and labeled by hand. Once such a dataset is released, its size is fixed, its difficulty is fixed, and every problem can leak into the training data of future LLMs. We present \textbf{A$^{2}$utoLPBench}, a benchmark for testing LLM-driven agents on linear programming problems written in plain text. We first pick a feasible point and dual, then write down a problem for which that point is optimal and the objective value is known. The answer is known by construction, with no solver call and no human annotator. The evaluation environment bundles a reference solver-critic baseline and a Docker image whose usage instructions are written for an LLM-driven agent to read. With these in place, any agent can run the benchmark and get a calibrated score with one command. Because the benchmark is a generator rather than a fixed dataset, it has properties no fixed dataset can match: an unlimited supply of fresh problems, a difficulty knob set by $(n,m)$, ground-truth answers correct by construction, low LLM-side cost per problem relative to human authoring, repeatable scores across independent batches, and resistance to training-data leakage when fresh post-cutoff seed ranges are used.
Chinese Translation
大多数基于文本的线性规划（LP）基准测试是由人工编写和标注的静态数据集。一旦发布，这类数据集的大小和难度都是固定的，并且每个问题都可能泄露到未来大语言模型（LLMs）的训练数据中。我们提出了 extbf{A$^{2}$utoLPBench}，这是一个用于测试基于LLM的代理在纯文本线性规划问题上的基准测试。我们首先选择一个可行点和对偶点，然后编写一个以该点为最优解且已知目标值的问题。答案是通过构造得出的，无需求解器调用和人工标注。评估环境包含一个参考求解器-评论基线和一个Docker镜像，其使用说明为基于LLM的代理所编写。通过这些设置，任何代理都可以通过一个命令运行基准测试并获得校准分数。由于该基准测试是一个生成器而非固定数据集，它具有固定数据集无法匹配的特性：无限量的新问题、由$(n,m)$设置的难度调节、构造上正确的真实答案、相对于人工编写的低LLM侧成本、跨独立批次的可重复分数，以及在使用新鲜的截止后种子范围时抵抗训练数据泄漏的能力。
cs.AI / 67 / 2607.02175
A rubric-based controlled comparison of frontier language models on expert-authored clinical reasoning tasks
基于评分标准的前沿语言模型在专家撰写的临床推理任务上的对比研究
Abstract
Multiple-choice medical benchmarks are increasingly saturated, and recent rubric-based evaluations such as HealthBench have shown that open-ended clinical performance is far from solved - its "Hard" subset top score remains 32%. We present a small, deliberately difficult evaluation dataset of five clinician-authored clinical scenarios spanning four specialties (anaesthesia, internal/family medicine, emergency medicine, and obstetrics), each accompanied by an atomic, weighted, MECE rubric (25-62 criteria per task; 184 criteria total) authored from a clinician-drafted golden answer. We evaluate three frontier models: GPT 5.4, Claude Opus 4.7, and Gemini 3.1 Pro. Mean rubric pass rates were 0.47 (Claude), 0.39 (GPT), and 0.37 (Gemini). The central finding is an inversion of clinical priority: the highest-weighted (weight-5, critical) criteria passed at only 32.4-41.7%, while low-stakes weight-1 criteria passed at 80-90%. 56 of 108 critical (weight-5) criteria (52%) were satisfied by no model. Three LLM autoraters reproduced expert met/not-met labels on 92.8-94.7% of 552 graded criteria. We position this as a methods-and-preliminary-findings contribution: the five tasks demonstrate a scalable, defensible pipeline ready to develop into a large-scale benchmark.
Chinese Translation
多项选择医学基准测试日益饱和，最近的基于评分标准的评估（如 HealthBench）显示，开放式临床表现远未解决——其“困难”子集的最高得分仅为32%。我们呈现了一个小型、故意设计为困难的评估数据集，包含五个临床专家撰写的临床场景，涵盖四个专业（麻醉学、内科/家庭医学、急救医学和妇产科），每个场景都附有一个原子化、加权的、MECE（相互独立且完全穷尽）评分标准（每个任务25-62个标准；总计184个标准），这些标准是基于临床专家撰写的黄金答案而制定的。我们评估了三种前沿模型：GPT 5.4、Claude Opus 4.7和Gemini 3.1 Pro。平均评分标准通过率为0.47（Claude）、0.39（GPT）和0.37（Gemini）。核心发现是临床优先级的反转：权重最高（权重5，关键）的标准仅通过32.4%-41.7%，而低风险的权重1标准通过率为80%-90%。108个关键（权重5）标准中有56个（52%）未被任何模型满足。三种大型语言模型（LLM）自动评分者在552个评分标准上复制了专家的满足/不满足标签，准确率为92.8%-94.7%。我们将此视为方法与初步发现的贡献：这五个任务展示了一个可扩展、可辩护的流程，准备发展为一个大规模基准测试。
cs.AI / 68 / 2607.02186
UA-ChatDev: Uncertainty-Aware Multi-Agent Collaboration for Reliable Software Development
UA-ChatDev：面向不确定性的多智能体协作以实现可靠的软件开发
Abstract
Software development is a complex task that demands cooperation among agents with diverse roles. Large language models (LLMs) have enabled autonomous multi-agent software development frameworks that leverage role-based collaboration to automate requirements analysis, coding, testing, and refinement. However, existing approaches typically assume that intermediate agent outputs are equally reliable, leaving them vulnerable to hallucination propagation, where incorrect decisions generated in early development phases are transferred to downstream agents and negatively impact final software quality. To address this challenge, we propose UA-ChatDev, an uncertainty-aware multi-agent software development framework that integrates uncertainty quantification into agent interactions. It introduces a lightweight uncertainty estimation mechanism based on token-level log probabilities to assess the confidence of agent responses and employs phase-aware threshold calibration to selectively trigger retrieval-based verification when uncertainty exceeds acceptable levels. Extensive experiments on the SRDD benchmark demonstrate that UA-ChatDev consistently outperforms existing single-agent and multi-agent software development frameworks across completeness, executability, consistency, and overall quality metrics. Further ablation studies and communication analyses verify that uncertainty-aware interactions enhance code execution reliability.
Chinese Translation
软件开发是一项复杂的任务，需要不同角色的智能体之间的合作。大型语言模型（LLMs）使得自主多智能体软件开发框架成为可能，这些框架利用基于角色的协作来自动化需求分析、编码、测试和优化。然而，现有的方法通常假设中间智能体输出同样可靠，这使得它们容易受到幻觉传播的影响，即在早期开发阶段产生的不正确决策被转移到下游智能体，从而对最终软件质量产生负面影响。为了解决这一挑战，我们提出了UA-ChatDev，一个面向不确定性的多智能体软件开发框架，该框架将不确定性量化集成到智能体交互中。它引入了一种基于标记级日志概率的轻量级不确定性估计机制，以评估智能体响应的信心，并采用阶段感知阈值校准，在不确定性超过可接受水平时选择性地触发基于检索的验证。在SRDD基准上的广泛实验表明，UA-ChatDev在完整性、可执行性、一致性和整体质量指标上始终优于现有的单智能体和多智能体软件开发框架。进一步的消融研究和通信分析验证了面向不确定性的交互增强了代码执行的可靠性。
cs.AI / 69 / 2607.02210
Criticality-Based Guard Rail Validation for AI Agent Decisions in Autonomous Telecom Networks
基于关键性验证的自主通信网络中人工智能代理决策的防护机制
Abstract
The evolution toward fully autonomous telecommunications networks (Autonomous Network Levels 4-5) requires AI/ML agents to make real-time network decisions without human intervention. However, no standardized runtime mechanism exists to intercept and validate individual inference outputs before they trigger live network state changes, creating risks of erroneous autonomous decisions. This paper proposes the Guard Rail Validation (GRV) framework, a standardizable runtime architecture for intercepting and validating AI-driven decisions before execution. The framework evaluates decisions across multiple weighted dimensions -- including action scope, action type, service criticality, agent autonomy level, reversibility, and temporal behavioural patterns -- to determine a criticality level. Based on this level, graduated validation mechanisms are applied: execute-with-logging, bounds checking, independent agent validation, or multi-agent consensus. The framework additionally provides cross-agent conflict detection with criticality-weighted priority resolution and runtime conformance logging for regulatory compliance (e.g., EU AI Act Article 14). We present the architecture, algorithmic procedures, O-RAN deployment model, and evaluate threat coverage against known AI/ML attacks in telecommunications.
Chinese Translation
向完全自主的电信网络（自主网络等级4-5）的演进要求人工智能/机器学习（AI/ML）代理在没有人类干预的情况下实时做出网络决策。然而，目前没有标准化的运行时机制来拦截和验证个别推理输出，以防止其触发实时网络状态变化，这增加了错误自主决策的风险。本文提出了防护机制验证（Guard Rail Validation，GRV）框架，这是一种可标准化的运行时架构，用于在执行之前拦截和验证AI驱动的决策。该框架通过多个加权维度评估决策，包括行动范围、行动类型、服务关键性、代理自主级别、可逆性和时间行为模式，以确定关键性水平。基于这一水平，应用不同级别的验证机制：记录执行、边界检查、独立代理验证或多代理共识。该框架还提供了跨代理冲突检测，采用关键性加权优先级解决方案和运行时合规日志记录，以满足监管要求（例如，欧盟人工智能法第14条）。我们展示了该架构、算法过程、O-RAN部署模型，并评估了针对已知AI/ML攻击的威胁覆盖情况。
cs.AI / 70 / 2607.02234
Purified OPSD: On-Policy Self-Distillation Without Losing How to Think
纯化的OPSD：在不失去思考能力的情况下进行在线自我蒸馏
Abstract
On-policy self-distillation (OPSD) has emerged as a promising paradigm for improving LLM reasoning, where a privileged teacher with access to reference solutions provides token-level supervision on the student's own generated trajectories. However, we find that OPSD consistently fails on long chain-of-thought (long-CoT) reasoning models, yielding at best marginal gains while destabilizing the reflective reasoning capability these models depend on. Through a novel decomposition of the teacher's supervision signal, we identify the root cause: the teacher's supervision is dominated by a reference-induced component that drives rote memorization of reference-specific shortcuts, while the question-conditioned, inference-transferable component is ignored or actively opposed. Based on this diagnosis, we propose a two-step solution. First, we construct a reference-only teacher (the same model conditioned on the reference without the question) to isolate the non-transferable component of the supervision signal; the residual after subtracting this component captures the question-conditioned, inference-transferable correction. Second, we use pointwise mutual information (PMI) as the mechanism to transform this residual into a well-formed PMI target distribution that the student can directly distill from, filtering out the reference-induced shortcut. Experiments on four long-CoT models across two datasets demonstrate consistent improvements over both the base model and standard OPSD, while preserving the models' natural epistemic behavior throughout training.
Chinese Translation
在线自我蒸馏（OPSD）已成为改善大型语言模型（LLM）推理的有前景的范式，其中一位拥有参考解决方案的特权教师对学生自身生成的轨迹提供逐标记的监督。然而，我们发现OPSD在长链思维（long-CoT）推理模型上始终表现不佳，最多只能带来边际收益，同时破坏了这些模型所依赖的反思推理能力。通过对教师监督信号的新颖分解，我们识别出根本原因：教师的监督被一种参考诱导成分主导，这种成分驱动了对特定参考快捷方式的机械记忆，而问题条件下的、可推理转移的成分则被忽视或积极对抗。基于这一诊断，我们提出了一个两步解决方案。首先，我们构建了一个仅参考的教师（同一模型在没有问题的情况下以参考为条件），以隔离监督信号中不可转移的成分；减去该成分后的残差捕捉了问题条件下的、可推理转移的修正。其次，我们使用点互信息（PMI）作为机制，将该残差转化为一个良构的PMI目标分布，学生可以直接从中进行蒸馏，从而过滤掉参考诱导的快捷方式。在两个数据集上的四个长-CoT模型的实验表明，相较于基础模型和标准OPSD，均取得了一致的改进，同时在整个训练过程中保持了模型的自然认识行为。
cs.AI / 71 / 2607.02245
Copewell: A Multi-Agent Swarm Architecture for Equitable Mental Wellness Support
Copewell：一种用于公平心理健康支持的多智能体群体架构
Abstract
Mental health disorders affect nearly one billion people globally, yet 75% of individuals in low- and middle-income countries receive no treatment due to workforce shortages, cost barriers, and stigma. Current AI-powered wellness solutions predominantly rely on single-mode conversational interfaces that suffer high abandonment rates and fail to provide measurable, immediate relief calibrated to users' dynamic emotional states. This paper presents Copewell, a novel multi-agent swarm system designed to expand access to mental wellness support through human-centered AI principles. Our architecture introduces three technical innovations: (1) a multi-source assessment framework integrating self-reported, physiological, and contextual data to mitigate algorithmic bias; (2) valence-arousal emotion mapping using Russell's Circumplex Model of Affect to route users to specialized AI agents; and (3) dual-mode intervention delivery combining conversational support with evidence-based sensory wellness protocols. We examine the sociotechnical design considerations underlying Copewell's development, including a privacy-first architecture, embedded ethical oversight through a dedicated Ethics Supervisor agent, and participatory design informed by mental health practitioners. Early practitioner engagement and beta deployment inform design decisions and identify directions for future empirical evaluation. This work contributes to responsible AI discourse by demonstrating how technical architecture can operationalize equity and safety principles from inception.
Chinese Translation
心理健康障碍在全球影响近十亿人，但在低收入和中等收入国家，75%的人未能接受治疗，原因包括人力资源短缺、成本障碍和社会污名。当前的人工智能驱动的健康解决方案主要依赖单一模式的对话界面，这些界面存在高放弃率，并未能提供与用户动态情绪状态相匹配的可衡量、即时的缓解效果。本文提出了Copewell，一种新颖的多智能体群体系统，旨在通过以人为本的人工智能原则扩大心理健康支持的获取。我们的架构引入了三项技术创新：（1）一个多源评估框架，整合自我报告、生理和上下文数据，以减轻算法偏见；（2）使用拉塞尔情感圆周模型进行情感的效价-唤醒映射，以引导用户到专业的人工智能代理；（3）双模式干预交付，结合对话支持与基于证据的感官健康协议。我们考察了Copewell开发过程中的社会技术设计考虑因素，包括隐私优先的架构、通过专门的伦理监督代理嵌入的伦理监督，以及由心理健康从业者提供的信息参与设计。早期从业者的参与和测试部署为设计决策提供了信息，并确定了未来实证评估的方向。这项工作通过展示技术架构如何从一开始就将公平和安全原则落实到位，为负责任的人工智能讨论做出了贡献。
cs.AI / 72 / 2607.02255
AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents
AgenticSTS：一个有限记忆的长时间跨度大型语言模型代理测试平台
Abstract
Memory for a long-horizon LLM agent is a contract about what each future decision is allowed to see. The simplest contract appends past observations, tool calls, and reflections to every prompt, which makes prior context easy to access but also turns it into a jumbled mixture in which the effect of any single memory component is hard to isolate. We introduce and instrument an alternative bounded contract: every decision is made from a fresh user message assembled by typed retrieval, with no raw cross-decision transcript appended. The prompt thus stays bounded across runs of any length, and any single layer can be ablated in isolation. We instantiate the contract in Slay the Spire 2, a closed-rule stochastic deck-building game whose runs require hundreds of tactical and strategic decisions. A public online benchmark of frontier LLMs on the same game reports zero wins at the lowest difficulty across five configurations, and the developer-reported human win rate at the same difficulty is 16%; the task is hard but not saturated. Within our harness, a fixed-A0 ablation shows the largest observed difference when triggered strategic skills are enabled: the no-store baseline wins 3/10 games and adding the skill layer 6/10. At this sample size the comparison is directional rather than statistically decisive (Fisher exact p\approx0.37); a cross-backbone probe and public accumulating-context baselines are reported as operational comparisons rather than controlled tests of the contract variable itself. We release a reproducible testbed: 298 completed trajectories with condition tags, frozen memory/skill snapshots, prompt records, and analysis scripts -- an agent design and a validated, reusable methodology for studying how explicit memory layers shape long-horizon LLM-agent decisions.
Chinese Translation
长时间跨度大型语言模型（LLM）代理的记忆是关于每个未来决策允许查看内容的契约。最简单的契约是将过去的观察、工具调用和反思附加到每个提示中，这使得先前的上下文易于访问，但也使其变成一种混杂的混合体，难以孤立任何单一记忆组件的影响。我们引入并实施了一种替代的有限契约：每个决策都是从通过类型检索组装的新用户消息中做出的，没有附加原始的跨决策记录。因此，提示在任何长度的运行中保持有限，并且可以单独消除任何单一层。我们在《Slay the Spire 2》这一封闭规则的随机卡牌构建游戏中实例化了该契约，该游戏的运行需要数百个战术和战略决策。在同一游戏上，前沿大型语言模型的公共在线基准报告在五种配置下在最低难度下零胜利，而开发者报告的相同难度下的人类胜率为16%；该任务难度较大，但并未饱和。在我们的框架内，固定的A0消融在触发的战略技能启用时显示出最大的观察差异：无存储基线赢得3/10场比赛，而添加技能层后赢得6/10场。在这个样本量下，比较是方向性的，而非统计决定性的（Fisher精确p extasciitilde0.37）；交叉骨干探测和公共累积上下文基线被报告为操作比较，而非对契约变量本身的控制测试。我们发布了一个可重复的测试平台：298条完成的轨迹，带有条件标签、冻结的记忆/技能快照、提示记录和分析脚本——一个代理设计和经过验证的、可重复使用的方法论，用于研究显式记忆层如何塑造长时间跨度的LLM代理决策。
cs.AI / 73 / 2607.02303
A Hippocampus for Linear Attention: An Exact Memory for What the Recurrent State Forgets
线性注意力的海马体：对循环状态遗忘内容的精确记忆
Abstract
Linear-attention and state-space language models compress the prefix into a fixed-size recurrent state, yielding O(1) memory at the cost of a lossy exact memory: when many key--value associations compete, earlier facts are overwritten and needle recall degrades. Inspired by Complementary Learning Systems, we give linear attention a hippocampal complement. HOLA (Hippocampal Linear Attention) keeps the usual delta-rule state as a compressive memory and adds a bounded exact KV cache, forming a semiparametric test-time memory: the state models linearly compressible structure, while the cache stores associations that should not be forced through that state. The cache writes without a learned eviction module, keeping tokens with large beta * ||e||, the prediction residual actually committed to the state; a decoupled RMSNorm-gamma cache read then turns these exact KV pairs into sharp retrieval rather than soft averaging. At 340M parameters trained on 15B SlimPajama tokens, HOLA lowers Wikitext perplexity from 27.32 to 22.92 (-16.1%), below a full-attention Transformer++ (26.88), and improves LAMBADA perplexity from 30.95 to 30.26. It also achieves the best linear in-context retrieval and remains much more robust than GDN or a matched HOLA+recency cache on RULER needle-in-a-haystack recall out to 32k tokens (16x its training length).
Chinese Translation
线性注意力和状态空间语言模型将前缀压缩为固定大小的循环状态，从而以 O(1) 的内存开销换取一种有损的精确记忆：当多个键值关联竞争时，早期的事实会被覆盖，导致回忆能力下降。受到互补学习系统的启发，我们为线性注意力提供了一个海马体的补充。HOLA（海马线性注意力）保持了通常的增量规则状态作为压缩记忆，并添加了一个有界的精确键值缓存，形成了一种半参数化的测试时记忆：状态模型线性压缩结构，而缓存存储不应强制通过该状态的关联。缓存在没有学习驱逐模块的情况下进行写入，保留具有大 beta * ||e|| 的标记，即实际承诺给状态的预测残差；解耦的 RMSNorm-gamma 缓存读取将这些精确的键值对转化为清晰的检索，而不是软平均。在使用 15B SlimPajama 令牌训练的 340M 参数下，HOLA 将 Wikitext 困惑度从 27.32 降低到 22.92（-16.1%），低于全注意力 Transformer++（26.88），并将 LAMBADA 困惑度从 30.95 改善到 30.26。它还实现了最佳的线性上下文检索，并在 RULER 的针在干草堆回忆任务中，比 GDN 或匹配的 HOLA+近期缓存更具鲁棒性，能够处理高达 32k 令牌（是其训练长度的 16 倍）。
cs.AI / 74 / 2607.02329
Grounded autonomous research: a fault-tolerant LLM pipeline from corpus to manuscript in frontier computational physics
基于实证的自主研究：从语料库到手稿的容错大型语言模型管道在前沿计算物理中的应用
Abstract
Autonomous-research agents have demonstrated end-to-end LLM automation in machine-learning sandboxes where execution provides calibration. Frontier physical science differs categorically: physical reasoning underlies every methodology choice, toolchains are often underdocumented, and calibration must come from external literature anchors - which unscaffolded agents cite but do not confront, hallucinating plausible, unverifiable results from internal priors. We present a pipeline that runs end-to-end from a corpus of 11,083 recent condensed-matter physics arXiv papers to a publication-grade manuscript with three substantive physics findings (here on altermagnetic piezomagnetism): the agent autonomously conceives a research direction by mapping the corpus, calibrates methodology by reproducing published references, conducts novel first-principles computations, and writes the manuscript - grounded in literature throughout, across 47 fresh-context sessions in six phases sharing only on-disk state, with 2,162 literature-consultation events. Fault tolerance emerges from redundancy: fresh-context isolation, distributed grounding, and adversarial review catch what any single session misses; pre- and post-pilot stages are fully autonomous, and pilot requires bounded human intervention only at reproduction failures - operational knowledge curation, not scientific direction. Two paired failure modes - a pre-architecture baseline and a no-pilot ablation - isolate structurally enforced numerical confrontation at calibration checkpoints as the operative grounding mechanism. The primitives, characterized failure modes, and quantified intervention pattern lay a foundation for autonomous research in high-stakes scientific domains beyond computational physics.
Chinese Translation
自主研究代理在机器学习沙箱中展示了端到端的大型语言模型自动化，其中执行提供了校准。前沿物理科学在本质上有所不同：物理推理是每种方法选择的基础，工具链往往记录不足，校准必须来自外部文献锚点——这些未经过框架的代理引用但不进行对抗，凭借内部先验幻想出合理的、无法验证的结果。我们提出了一条管道，从11,083篇近期的凝聚态物理arXiv论文出发，最终生成一篇具有出版质量的手稿，包含三个实质性的物理发现（这里关于反磁压电性）：代理通过映射语料库自主构思研究方向，通过重现已发表的参考文献校准方法，进行新颖的第一性原理计算，并撰写手稿——在整个过程中都以文献为基础，跨越47个新上下文会话，分为六个阶段，仅共享磁盘状态，共进行了2,162次文献咨询事件。容错性源于冗余：新上下文隔离、分布式基础和对抗性审查捕捉到任何单一会话所遗漏的内容；前期和后期试点阶段完全自主，试点阶段仅在重现失败时需要有限的人为干预——是操作知识的管理，而非科学方向的指导。两种配对的失败模式——预架构基线和无试点消融——在校准检查点隔离了结构性强制的数值对抗，作为操作性基础机制。所描述的原始构件、特征化的失败模式和量化的干预模式为高风险科学领域中的自主研究奠定了基础，超越了计算物理的范畴。
cs.AI / 75 / 2607.02374
DRIFTLENS: Measuring Memory-Induced Reasoning Drift in Personalized Language Models
DRIFTLENS：测量个性化语言模型中的记忆引发的推理漂移
Abstract
Personalization changes what a model says to a user; we show that it can also change the reasoning trajectory used to justify the response. Modern LLMs personalize interactions by storing user attributes, preferences, and prior context, then injecting this information into future prompts. We study whether such memory reshapes reasoning on open-ended questions where no single ground-truth answer exists. To quantify this effect, we introduce DRIFTLENS, a ground-truth-free framework that maps each expressed reasoning step to a value category and measures divergence between a question's no-memory trajectory and its trajectory under injected user-attribute memory. We first validate that DRIFTLENS distinguishes content-free pragmatic noise from substantive reasoning changes. Across four LLMs and 10 user-attribute categories, including age, occupation, and disability, user-attribute memory induces medium-to-large reasoning drift above each model's pragmatic-noise floor, even when final answers remain fluent, on-topic, and plausible. We then evaluate GRPO- and DPO-based post-training methods for reducing drift. Both reduce drift, but neither uniformly dominates; effects on downstream capability, helpfulness, and instruction following are model-and reward-dependent. These results suggest that memory-induced reasoning drift is a measurable and only partly mitigated failure mode of personalized language models.
Chinese Translation
个性化改变了模型对用户的回应内容；我们展示了它也可以改变用于证明回应的推理轨迹。现代大型语言模型（LLMs）通过存储用户属性、偏好和先前上下文来个性化互动，然后将这些信息注入到未来的提示中。我们研究了这种记忆是否会重塑在没有单一真实答案的开放式问题上的推理。为了量化这一效应，我们引入了DRIFTLENS，这是一个无真实答案的框架，将每个表达的推理步骤映射到一个价值类别，并测量问题的无记忆轨迹与其在注入用户属性记忆下的轨迹之间的偏差。我们首先验证了DRIFTLENS能够区分无内容的务实噪声与实质性的推理变化。在四个大型语言模型和10个用户属性类别（包括年龄、职业和残疾）中，用户属性记忆在每个模型的务实噪声底线之上引发了中到大规模的推理漂移，即使最终答案仍然流畅、相关且合理。然后，我们评估了基于GRPO和DPO的后训练方法以减少漂移。两者均能减少漂移，但没有一种方法在所有情况下都占优势；对下游能力、帮助性和指令遵循的影响依赖于模型和奖励。这些结果表明，记忆引发的推理漂移是个性化语言模型的一种可测量且仅部分缓解的失败模式。
cs.AI / 76 / 2607.02376
Hardware-Enforced Semantic Coordination for Safety-Critical Real-Time Autonomous Systems
硬件强制的语义协调用于安全关键的实时自主系统
Abstract
Recent advances in agentic AI are producing increasingly complex autonomous systems that integrate large language models, world models, optimization engines, specialized neural architectures, autonomous platforms, and human operators. While much current research focuses on improving reasoning capabilities, safety-critical real-time deployment also requires bounded and verifiable coordination among heterogeneous components operating concurrently under uncertainty. Software-mediated coordination presents fundamental limitations in domains where bounded latency, deterministic coordination, and enforceable safety guarantees are essential. Hence, we propose a hardware-enforced semantic coordination architecture in which selected coordination semantics are implemented directly at the hardware level via field-programmable gate arrays (FPGAs). The approach builds on the Topic-Based Communication Space Petri Net (TB-CSPN) framework, which separates semantic reasoning from interaction management. In this approach, selected TB-CSPN coordination mechanisms are mapped onto FPGA primitives, creating a hardware-native semantic coordination layer. Focus is not on acceleration, but on enforcing temporal synchronization, semantic gating, authorization constraints, and bounded coordination behavior directly in hardware. Semantic reasoning remains adaptive and software-driven, while embedded coordination semantics become deterministic.
Chinese Translation
近年来，代理人工智能的进步正在产生越来越复杂的自主系统，这些系统集成了大型语言模型、世界模型、优化引擎、专用神经架构、自主平台和人类操作员。尽管当前许多研究集中在提高推理能力上，但安全关键的实时部署还需要在不确定性下对并发操作的异构组件进行有界和可验证的协调。软件介导的协调在需要有界延迟、确定性协调和可强制安全保证的领域中存在根本性限制。因此，我们提出了一种硬件强制的语义协调架构，其中选定的协调语义通过现场可编程门阵列（FPGAs）直接在硬件层面实现。该方法基于主题驱动的通信空间Petri网（TB-CSPN）框架，将语义推理与交互管理分离。在此方法中，选定的TB-CSPN协调机制被映射到FPGA原语上，创建一个硬件本地的语义协调层。重点不在于加速，而是在硬件中直接强制实施时间同步、语义门控、授权约束和有界协调行为。语义推理仍然是自适应和软件驱动的，而嵌入的协调语义则变得确定性。
cs.AI / 77 / 2607.02389
Steerability via constraints: a substrate for scalable oversight of coding agents
通过约束实现可操控性：可扩展编码代理监督的基础
Abstract
Coding agents are capable; human oversight is the bottleneck. Unconstrained agents introduce security risks, erode codebase scalability, and make human review increasingly costly. We argue that the same methods used for decades to manage large human engineering teams: access control, network policies, strict coding conventions enforced by tooling; transfer directly to coding agents, and are cheaper (in token) than recent agentic scaffolding. We sketch a start-to-end system on this principle, and report a controlled experiment in scalable oversight: a small reviewer (Gemma 4 e4b) inspects a Python codebase containing 11 inserted backdoors. Recall rises from 54.5% (unconstrained, no tools) to 90.9% (constrained substrate plus a ~200-LoC `docs` CLI), with substrate and tools contributing independently. We choose Python deliberately: substrate-level oversight gains are largest where the language gives the fewest guarantees by default; the principles extend to languages like Rust.
Chinese Translation
编码代理具有能力；人类监督是瓶颈。无约束的代理引入安全风险，侵蚀代码库的可扩展性，并使人类审查的成本日益增加。我们认为，几十年来用于管理大型人类工程团队的方法：访问控制、网络策略、由工具强制执行的严格编码规范，能够直接转移到编码代理上，并且在成本（以代币计）上比最近的代理框架更具优势。我们在这一原则上勾勒出一个端到端的系统，并报告了一项关于可扩展监督的受控实验：一位小型审查员（Gemma 4 e4b）检查一个包含11个插入后门的Python代码库。召回率从54.5%（无约束，无工具）上升至90.9%（约束基础加上一个~200行代码的`docs` CLI），基础和工具独立贡献。我们故意选择Python：在语言默认提供最少保证的情况下，基础级监督的收益最大；这些原则也适用于Rust等语言。
cs.AI / 78 / 2607.02396
Fast Multi-dimensional Refusal Subspaces via RFM-AGOP
通过 RFM-AGOP 实现快速多维拒绝子空间
Abstract
Steering and monitoring activations in Large Language Models (LLMs) are increasingly used for both safety and interpretability. Early work assumed behaviours are encoded along single linear directions, but recent findings suggest complex behaviours, such as the refusal to answer harmful queries, live in multi-dimensional subspaces. However, existing methods for extracting these subspaces are computationally expensive, which becomes prohibitive on reasoning models who produce long reasoning traces. By adapting the Recursive Feature Machine (RFM) algorithm -- which can be computed efficiently -- with a probe-informed initialization, we are able to identify the multi-dimensional refusal subspace in seconds, on reasoning (Qwen 3) and non-reasoning (Qwen 2.5) models. While RFM allows for faster subspace identification, it also showed better performances on the ablation task than its alternatives. More work is planned to better understand the relations between subspaces found by different methods. If confirmed, RFM could be a cheap and scalable complement to existing subspace-extraction methods in LLMs.
Chinese Translation
在大型语言模型（LLMs）中，指导和监控激活越来越多地用于安全性和可解释性。早期的研究假设行为沿单一线性方向编码，但最近的发现表明，复杂行为（例如拒绝回答有害查询）存在于多维子空间中。然而，现有提取这些子空间的方法计算成本高昂，这在产生长推理轨迹的推理模型上变得不可行。通过适应递归特征机（RFM）算法——该算法可以高效计算——并结合探针引导的初始化，我们能够在几秒钟内识别推理（Qwen 3）和非推理（Qwen 2.5）模型中的多维拒绝子空间。虽然 RFM 允许更快的子空间识别，但在消融任务中也表现出比其他替代方法更好的性能。我们计划进行更多研究，以更好地理解不同方法找到的子空间之间的关系。如果得到确认，RFM 可能成为现有 LLMs 子空间提取方法的廉价且可扩展的补充。
cs.AI / 79 / 2607.02407
Text-Driven 3D Indoor Scene Synthesis in Non-Manhattan Environments
非曼哈顿环境中的文本驱动三维室内场景合成
Abstract
Large Language Models (LLMs) have demonstrated remarkable capabilities in 3D indoor synthesis for Manhattan environments. However, existing methods often fail to capture plausible object layout patterns in non-Manhattan settings, primarily because they struggle to model non-orthogonal spatial relationships, leading to high geometric violations and low physical fidelity. To address this challenge, we propose SPG-Layout, a novel text-driven framework designed to generate physically plausible indoor scenes within complex non-Manhattan environments. Specifically, we first utilize statistical priors of object distributions to guide the training process, enhancing environmental understanding and fidelity. Furthermore, mirroring human design workflows, we adopt a hierarchical layout strategy that prioritizes the placement of large objects, thereby substantially minimizing layout violations. By synergizing these components, SPG-Layout achieves a balanced optimization of semantic realism and physical plausibility. To evaluate performance in these complex settings, we constructed a new benchmark comprising 500 diverse non-Manhattan environments. Extensive experiments demonstrate that SPG-Layout consistently and significantly outperforms existing methods across both Manhattan and non-Manhattan environments. The code will be publicly released.
Chinese Translation
大型语言模型（LLMs）在曼哈顿环境中的三维室内合成方面展现了显著的能力。然而，现有方法往往无法捕捉非曼哈顿环境中合理的物体布局模式，主要是因为它们在建模非正交空间关系时遇到困难，导致高几何违规和低物理逼真度。为了解决这一挑战，我们提出了SPG-Layout，这是一种新颖的文本驱动框架，旨在生成复杂非曼哈顿环境中的物理合理室内场景。具体而言，我们首先利用物体分布的统计先验来指导训练过程，从而增强环境理解和逼真度。此外，模仿人类设计工作流程，我们采用了一种分层布局策略，优先考虑大型物体的放置，从而显著减少布局违规。通过协同这些组件，SPG-Layout实现了语义真实感和物理合理性的平衡优化。为了在这些复杂环境中评估性能，我们构建了一个包含500个多样化非曼哈顿环境的新基准。大量实验表明，SPG-Layout在曼哈顿和非曼哈顿环境中始终显著优于现有方法。代码将公开发布。
cs.AI / 80 / 2607.02432
Automated grading of Linux/bash examinations using large language models: a four-level cognitive taxonomy approach
利用大型语言模型对Linux/bash考试进行自动评分：一种四级认知分类法的应用
Abstract
Scalable and reliable grading of command-line examinations remains a challenge in computing education, where rising enrolments make manual marking difficult and rule-based autograders cannot handle partial credit, equivalent solutions, or syntactic variation. This paper evaluates whether four frontier Large Language Models (GPT, Claude Opus, Gemini, and GLM) can approximate expert judgment when grading short Linux/bash command responses. The study adopts a four-level cognitive taxonomy that combines cognitive complexity and operational impact, ranging from information retrieval (L1) and basic file manipulation (L2) to structural operations (L3) and advanced system management (L4). The models were tested with two prompt variants, a minimal baseline and a rubric-enhanced version, on 1200 real responses from second-year Computer Engineering students independently graded by three expert instructors. Gemini~3.0 Pro with rubric-guided prompting achieved the highest human-AI agreement (ICC(3,1) = 0.888, MAE = 0.10, Bland-Altman bias = -0.014). Agreement declined consistently as taxonomy level increased, with the largest discrepancies at higher levels. Across all models, rubric quality had a larger effect than provider choice, with structured prompts consistently improving agreement. These results show that question complexity is a reliable predictor of the difficulty LLMs face in grading accurately, and they establish a principled, taxonomy-based framework for determining which questions are suitable for AI-assisted grading and which require human review, while also providing a transferable evaluation protocol and prompt templates.
Chinese Translation
在计算机教育中，命令行考试的可扩展和可靠评分仍然是一个挑战，随着招生人数的增加，手动评分变得困难，而基于规则的自动评分系统无法处理部分得分、等效解或语法变异。本文评估了四种前沿大型语言模型（GPT、Claude Opus、Gemini和GLM）在对短篇Linux/bash命令响应进行评分时是否能够接近专家判断。研究采用了一种结合认知复杂性和操作影响的四级认知分类法，涵盖信息检索（L1）、基本文件操作（L2）、结构性操作（L3）和高级系统管理（L4）。模型在1200个来自二年级计算机工程学生的真实响应上进行了测试，这些响应由三位专家教师独立评分，使用了两种提示变体：一个最小基线和一个增强评分标准的版本。使用评分标准引导提示的Gemini 3.0 Pro达到了最高的人机一致性（ICC(3,1) = 0.888，MAE = 0.10，Bland-Altman偏差 = -0.014）。随着分类法级别的提高，一致性持续下降，在较高层次上差异最大。在所有模型中，评分标准的质量对一致性的影响大于提供者的选择，结构化提示始终提高了一致性。这些结果表明，问题的复杂性是预测大型语言模型准确评分难度的可靠指标，并建立了一个有原则的、基于分类法的框架，以确定哪些问题适合AI辅助评分，哪些问题需要人工审核，同时提供了可转移的评估协议和提示模板。
cs.AI / 81 / 2607.02440
EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments
EvoPolicyGym：在交互环境中评估自主策略演化
Abstract
Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended software-engineering progress. We introduce Autonomous Policy Evolution, a controlled evaluation setting in which a harness-model agent repeatedly edits an executable policy system under a fixed interaction budget. We instantiate this setting in EvoPolicyGym, a benchmark built from compact interactive RL environments that evaluates how agents iteratively improve explored policies. On the EvoPolicyGym suite, GPT-5.5 achieves the strongest aggregate rank score and top-two performance on all 16 environments. Beyond leaderboard results, EvoPolicyGym also provides trajectory-level diagnostics that distinguish how agents allocate budget, convert feedback into parametric tuning. These analyses show that strong autonomous policy evolution depends not only on isolated task wins, but on discovering task-appropriate mechanisms and refining policies under bounded feedback.
Chinese Translation
自主代理越来越被期望通过反馈来改进可执行策略，然而现有的评估往往将这一过程简化为最终得分，或与开放式软件工程进展混淆。我们引入了自主策略演化（Autonomous Policy Evolution），这是一个受控评估环境，在该环境中，一个基于模型的代理在固定的交互预算下反复编辑可执行策略系统。我们在EvoPolicyGym中实现了这一环境，这是一个基于紧凑的交互式强化学习（RL）环境构建的基准，评估代理如何迭代改进探索的策略。在EvoPolicyGym套件中，GPT-5.5在所有16个环境中取得了最强的综合排名得分和前两名的表现。除了排行榜结果，EvoPolicyGym还提供了轨迹级别的诊断，区分代理如何分配预算，以及如何将反馈转化为参数调整。这些分析表明，强大的自主策略演化不仅依赖于孤立任务的胜利，还依赖于发现适合任务的机制，并在有限反馈下精炼策略。
cs.AI / 82 / 2607.02491
G-RRM: Guiding Symbolic Solvers with Recurrent Reasoning Models
G-RRM：使用递归推理模型指导符号求解器
Abstract
In this work, we focus on SE-RRMs, a symbol-equivariant instantiation of RRMs that exhibits improved extrapolation to larger problem sizes. We propose a neuro-symbolic approach, ``Guiding with Recurrent Reasoning Models'' (G-RRM), which integrates SE-RRMs with symbolic solvers for constraint satisfaction problems. SE-RRMs act as neural solvers that generate full solution proposals and guide classical symbolic solvers, such as backtracking or SAT-based methods like Glucose 4.1 and CaDiCaL 3.0.0, that produce globally correct solutions. Centrally, we investigate when neural guidance with G-RRM improves the search efficiency of symbolic solvers. %
Our experiments show that the efficacy of G-RRM depends on two conditions: first, the problem instances must have an expansive combinatorial search space to expose potential gains, and second, the solver architecture must be capable of dynamically overwriting its branching choices to recover when neural hints are imperfect. When these conditions hold, guidance drives median conflict counts to zero and yields significant wall-clock speedups: on $9\times9$ Sudoku, where the SE-RRM correctly solves $91.1\%$ of instances, backtracking accelerates by $33.3\times$ and Glucose 4.1 by $1.70\times$ (median, $p<0.001$), with Glucose 4.1 retaining a $1.17\times$ speedup on perfect-hint $25\times25$ grids. In contrast, CaDiCaL 3.0.0, whose runtime is overhead-dominated and which always respects the injected branching hints rather than overwriting them, shows no significant speedup (median $1.02\times$, n.s.) and even a small significant mean slowdown ($0.90\times$) on $9\times9$. These results delineate the regimes in which neural guidance translates into practical speedups.
Chinese Translation
在这项工作中，我们关注于 SE-RRM，这是一种符号等变的 RRMs 实例，表现出对更大问题规模的改进外推能力。我们提出了一种神经符号方法，称为“使用递归推理模型指导”（G-RRM），它将 SE-RRMs 与约束满足问题的符号求解器结合起来。SE-RRMs 作为神经求解器，生成完整的解决方案提议并指导经典的符号求解器，如回溯法或基于 SAT 的方法（如 Glucose 4.1 和 CaDiCaL 3.0.0），这些方法能够产生全局正确的解决方案。我们重点研究了何时使用 G-RRM 的神经指导能够提高符号求解器的搜索效率。我们的实验表明，G-RRM 的有效性依赖于两个条件：首先，问题实例必须具有广泛的组合搜索空间，以暴露潜在的收益；其次，求解器架构必须能够动态覆盖其分支选择，以便在神经提示不完美时恢复。当这些条件成立时，指导将中位冲突计数降低至零，并产生显著的实际时间加速：在 $9 imes9$ 数独中，SE-RRM 正确解决了 $91.1\%$ 的实例，回溯法加速了 $33.3 imes$，而 Glucose 4.1 加速了 $1.70 imes$（中位数，$p<0.001$），在完美提示的 $25 imes25$ 网格上，Glucose 4.1 保持了 $1.17 imes$ 的加速。相比之下，CaDiCaL 3.0.0 的运行时间受开销主导，并且总是遵循注入的分支提示而不是覆盖它们，因此没有显著的加速（中位数 $1.02 imes$，不显著）甚至在 $9 imes9$ 上出现了小幅显著的平均减速（$0.90 imes$）。这些结果划定了神经指导转化为实际加速的范围。
cs.AI / 83 / 2607.02507
What LLM Agents Say When No One Is Watching: Social Structure and Latent Objective Emergence in Multi-Agent Debates
无人监视时大型语言模型代理的言论：多智能体辩论中的社会结构与潜在目标的出现
Abstract
LLM agents will increasingly act in socially structured settings where role, audience, and relational context can shape what is advantageous or costly to say. We study whether such social structure, without any explicit objective in the prompt, changes what an agent expresses publicly relative to an off-the-record (OTR) channel elicited under the same condition. We introduce a dual-channel debate framework in which agents produce public utterances that enter the shared history alongside OTR responses that are recorded but never shown to the other participant. Across 10 models, 3 scenarios, and 5 variations within each scenario, alignment-inducing settings produce systematic public-OTR divergence in the targeted agent, with its decision divergence rising from a $\sim$3% baseline to roughly 40%. The effect is consistent across four aggregate analyses: stance, semantic similarity, natural language inference, and survey responses. In some cases, the OTR response explicitly attributes public accommodation to relational pressures, such as career risk or sponsorship obligation. The findings suggest that agent evaluation should extend beyond explicit goals and detect emergent objectives. We present a dual-channel evaluation framework and complementary behavioral measures that operationalize this assessment.
Chinese Translation
大型语言模型（LLM）代理将越来越多地在社会结构化的环境中行动，其中角色、观众和关系背景可以影响所说内容的优势或成本。我们研究了在没有任何明确目标的提示下，这种社会结构是否会改变代理在公开场合的表达，相较于在相同条件下引发的非正式（OTR）渠道的表达。我们引入了一种双通道辩论框架，其中代理产生的公开言论与记录但从不展示给其他参与者的OTR响应共同进入共享历史。在10个模型、3种场景和每种场景内的5种变体中，促进一致性的设置在目标代理中产生了系统性的公开-OTR差异，其决策差异从约3%的基线上升至大约40%。这一效果在四个综合分析中保持一致：立场、语义相似性、自然语言推理和调查响应。在某些情况下，OTR响应明确将公开适应归因于关系压力，例如职业风险或赞助义务。研究结果表明，代理评估应超越明确目标，识别新出现的目标。我们提出了一种双通道评估框架和互补的行为测量方法，以实现这一评估。
cs.AI / 84 / 2607.02509
ReContext: Recursive Evidence Replay as LLM Harness for Long-Context Reasoning
ReContext：递归证据重放作为大语言模型的长上下文推理工具
Abstract
Understanding and reasoning over long contexts has become a key requirement for deploying large language models (LLMs) in realistic applications. Although recent LLMs support increasingly long context windows, they often fail to use relevant evidence that is already present in the input, revealing a gap between context access and effective context utilization. In this work, we propose Recursive Evidence Replay as LLM Harness for Long-Context Reasoning (RECONTEXT), a training-free inference method for improving long-context reasoning. RECONTEXT uses model-internal relevance signals to construct a query-conditioned evidence pool and replays it before final generation while preserving the full original context. This recursive selection process separates evidence organization from answer generation without training, external memory, or context pruning. We also provide a theoretical analysis based on associative memory, which characterizes the context as a memory store, the question as a retrieval cue, attention as cue-trace association, and replay as trace reactivation. Experiments on eight long-context datasets with 128K context length show that RECONTEXT consistently improves evidence utilization across Qwen3-4B, Qwen3-8B, and Llama3-8B, achieving the best average rank on all three backbones. Code is available at https://github.com/Yanjun-Zhao/ReContext.
Chinese Translation
理解和推理长上下文已成为在现实应用中部署大型语言模型（LLMs）的关键需求。尽管近期的LLMs支持越来越长的上下文窗口，但它们往往未能有效利用输入中已经存在的相关证据，这揭示了上下文访问与有效上下文利用之间的差距。在本研究中，我们提出了递归证据重放作为大语言模型的长上下文推理工具（RECONTEXT），这是一种无需训练的推理方法，用于改善长上下文推理。RECONTEXT利用模型内部的相关性信号构建一个基于查询的证据池，并在最终生成之前重放该证据，同时保留完整的原始上下文。这个递归选择过程将证据组织与答案生成分开，而无需训练、外部存储或上下文修剪。我们还提供了基于联想记忆的理论分析，将上下文视为记忆存储，将问题视为检索线索，将注意力视为线索-痕迹关联，将重放视为痕迹再激活。在八个具有128K上下文长度的长上下文数据集上的实验表明，RECONTEXT在Qwen3-4B、Qwen3-8B和Llama3-8B上始终提高了证据利用率，在所有三个模型上实现了最佳平均排名。代码可在 https://github.com/Yanjun-Zhao/ReContext 获取。
cs.AI / 85 / 2607.02510
Online Safety Monitoring for LLMs
大语言模型的在线安全监测
Abstract
Despite alignment training, LLMs remain prone to generating unsafe outputs at deployment time. Monitoring outputs online and raising an alarm when safety can no longer be assumed is therefore critical. We study a simple real-time monitor that turns a verifier signal from an external model into an alarm decision by thresholding, with the threshold calibrated via risk control. In experiments on mathematical reasoning and red teaming datasets, we show that this simple design is competitive with more advanced monitors based on sequential hypothesis testing.
Chinese Translation
尽管经过了对齐训练，大语言模型（LLMs）在部署时仍然容易生成不安全的输出。因此，在线监测输出并在安全性无法再被假定时发出警报是至关重要的。我们研究了一种简单的实时监测器，该监测器通过阈值化将外部模型的验证信号转化为警报决策，阈值通过风险控制进行校准。在数学推理和红队测试数据集上的实验中，我们展示了这一简单设计与基于序贯假设检验的更先进监测器具有竞争力。
cs.AI / 86 / 2607.02514
Distributed Attacks in Persistent-State AI Control
持久状态人工智能控制中的分布式攻击
Abstract
As AI coding agents become more autonomous, they increasingly ship code iteratively, with the codebase persisting across sessions. This persistence creates a new attack surface: a misaligned or prompt-injected agent can distribute attacks across pull requests (PRs) and time its payload for the PR with the best natural cover. To study the resulting dynamics, we introduce Iterative VibeCoding, a setting for AI control, the study of safely deploying capable but potentially untrusted AI. In Iterative VibeCoding, a coding agent builds software over a sequence of PRs in a persistent codebase while pursuing a covert side task. Our benchmark includes two task families: CLI tools and Flask web services, across 20 total task variations. We use Claude Sonnet 4.5 as the attack agent and GPT-4o as the monitor. We compare gradual attacks, which distribute the side task across PRs, against non-gradual attacks concentrated in a single PR. No single monitor is robust to both: which strategy evades best (success while evading the monitor) depends on the monitor type, so a defender cannot close off both gradual and non-gradual attacks with any one monitor. High evasion (>= 65%) generalizes across model attack agent backends (Sonnet 4.5, Gemini 3.1 Pro, Kimi K2.5), confirming this is a property of the persistent-state attack surface rather than a single model's capability. Evasion also remains high across state-of-the-art monitor models and the gap between gradual and non-gradual evasion widens for more capable models. We introduce a stateful link-tracker monitor that tracks suspicious buildup across PRs. On both task families, it detects gradual attacks substantially better than diff monitors that merely see more accumulated history. Combining this stronger monitor with trajectory monitors in a four-monitor ensemble reduces gradual-attack evasion from 93% under the weakest standard diff monitor to 47%.
Chinese Translation
随着人工智能编码代理变得越来越自主，它们逐渐以迭代方式发布代码，代码库在会话之间保持持久性。这种持久性创造了一个新的攻击面：一个不对齐或注入提示的代理可以在拉取请求（PR）中分发攻击，并为具有最佳自然掩护的PR定时其有效载荷。为了研究由此产生的动态，我们引入了迭代VibeCoding，这是一个人工智能控制的设置，旨在研究安全部署有能力但可能不可信的人工智能。在迭代VibeCoding中，一个编码代理在一个持久代码库中通过一系列PR构建软件，同时追求一个隐蔽的副任务。我们的基准包括两个任务系列：CLI工具和Flask网络服务，共有20种任务变体。我们使用Claude Sonnet 4.5作为攻击代理，使用GPT-4o作为监控器。我们比较了将副任务分布在PR中的渐进攻击与集中在单个PR中的非渐进攻击。没有任何单一监控器能够对两者都表现出强健性：哪种策略能够更好地逃避（在逃避监控器的同时成功）取决于监控器的类型，因此防御者无法通过任何一个监控器同时封堵渐进和非渐进攻击。高逃避率（>= 65%）在模型攻击代理后端（Sonnet 4.5、Gemini 3.1 Pro、Kimi K2.5）之间具有普遍性，确认这是持久状态攻击面的特性，而不是单一模型的能力。逃避率在最先进的监控模型中也保持较高，且随着模型能力的增强，渐进和非渐进逃避之间的差距扩大。我们引入了一种状态链接跟踪监控器，能够跟踪PR中可疑的积累。在两个任务系列中，它对渐进攻击的检测显著优于仅仅看到更多累积历史的差异监控器。将这种更强的监控器与轨迹监控器结合在一个四监控器的集成中，将渐进攻击的逃避率从在最弱标准差异监控器下的93%降低到47%。
cs.CL / 1 / 2607.01235
TokenScope: Token-Level Explainability and Interpretability for Code-Oriented Tasks in Large Language Models
TokenScope：针对大型语言模型中代码导向任务的令牌级可解释性与解释性
Abstract
Understanding how Large Language Models (LLMs) make token-level decisions during code generation remains a major challenge for both researchers and practitioners. While recent tools provide insights into model internals or generation outcomes, they often lack decoding-time signals, fine-grained uncertainty measures, and interactive mechanisms for exploring alternative generation paths. We present TokenScope, an interactive interpretability and analysis tool for decoder-based LLMs that exposes token-level metrics, attention patterns, and structural information during generation. TokenScope supports interactive token replacement, counterfactual branching, and code-aware aggregation via abstract syntax trees. By unifying decoding-time signals with structural program analysis, TokenScope enables systematic investigation of LLM behaviour during code generation.
Chinese Translation
理解大型语言模型（LLMs）在代码生成过程中如何做出令牌级决策仍然是研究人员和从业者面临的主要挑战。尽管最近的工具提供了对模型内部或生成结果的洞察，但它们往往缺乏解码时信号、细粒度的不确定性度量以及探索替代生成路径的交互机制。我们提出了TokenScope，这是一种用于基于解码器的LLMs的交互式可解释性和分析工具，能够在生成过程中揭示令牌级指标、注意力模式和结构信息。TokenScope支持交互式令牌替换、反事实分支以及通过抽象语法树进行的代码感知聚合。通过将解码时信号与结构程序分析相结合，TokenScope使得对LLM在代码生成过程中的行为进行系统性研究成为可能。
cs.CL / 2 / 2607.01236
Safeguarding LLM Agents from Misalignment through Provenance Analysis
通过来源分析保护大型语言模型代理免受不一致性影响
Abstract
As LLM agents gain increasing access to powerful tools, ensuring that their actions are aligned with the user's intent becomes critical. When an agent's proposed tool invocation deviates from the user's intent -- a phenomenon called misalignment -- it may lead to harmful consequences that are difficult to undo. Existing runtime guardrails rely on an LLM-as-a-judge paradigm that lacks a systematic framework for reasoning about alignment, often producing judgments that are inconsistent or difficult to audit. Motivated by provenance analysis, we propose a provenance-based conceptual framework that formalizes misalignment detection as determining whether a proposed tool call is supported by traceable evidence in the agent's context. Building on this framework, we propose ProvenanceGuard, a multi-stage pipeline that analyzes the agent's action for three types of misalignment before the selected tool is executed and only allows the action to take place when it is considered aligned with the user's input query. We evaluated our proposed approach on two different benchmarks, Agent-SafetyBench and WorkBench, across 10 backbone LLMs. Compared to the LLM-as-a-judge baseline, ProvenanceGuard reduces error rate on misaligned traces from 42.9% to 1.8% on Agent-SafetyBench and from 32.1% to 17.3% on WorkBench, while reducing intervention burden on task-successful traces from 30.5% to 12.8% and introducing no statistically significant increase in unnecessary interventions on aligned traces. These results demonstrate that structured, provenance-based reasoning provides an effective and practical foundation for safeguarding LLM agents from misalignment.
Chinese Translation
随着大型语言模型（LLM）代理越来越多地接触强大的工具，确保其行为与用户意图一致变得至关重要。当代理提议的工具调用偏离用户意图时——这一现象称为不一致性——可能会导致难以逆转的有害后果。现有的运行时保护措施依赖于LLM作为裁判的范式，缺乏系统性的框架来推理一致性，常常产生不一致或难以审计的判断。受到来源分析的启发，我们提出了一种基于来源的概念框架，将不一致性检测形式化为确定提议的工具调用是否有可追溯的证据支持在代理的上下文中。基于该框架，我们提出了ProvenanceGuard，一个多阶段管道，在选择工具执行之前分析代理的行为，以检测三种类型的不一致性，仅在行为被认为与用户输入查询一致时才允许其发生。我们在两个不同的基准上评估了我们提出的方法，即Agent-SafetyBench和WorkBench，涵盖了10个基础LLM。与LLM作为裁判的基线相比，ProvenanceGuard将Agent-SafetyBench上不一致轨迹的错误率从42.9%降低到1.8%，将WorkBench上的错误率从32.1%降低到17.3%，同时将任务成功轨迹上的干预负担从30.5%降低到12.8%，并且在一致轨迹上没有引入统计学上显著的无谓干预增加。这些结果表明，结构化的基于来源的推理为保护LLM代理免受不一致性影响提供了有效且实用的基础。
cs.CL / 3 / 2607.01237
Kara: Efficient Reasoning LLM Serving via Sliding-Window KV Cache Compression
Kara：通过滑动窗口 KV 缓存压缩实现高效推理 LLM 服务
Abstract
Reasoning language models often generate long chain-of-thought (CoT), which accumulates a massive KV cache during the decoding phase and incurs high decoding latency and limited throughput. To address these issues, KV cache compression has emerged as a promising technique for reducing memory overhead by selectively removing unimportant KV pairs while preserving useful ones for subsequent decoding. Nevertheless, we identify two key limitations in existing KV cache compression methods: 1) their threshold-triggered compression policy may provide limited throughput improvement or even reduce throughput, and may fully eliminate KV pairs from certain blocks of the sequence, potentially worsening information loss. 2) they typically retain either isolated KV pairs or fixed-size chunks with rigid boundaries, failing to preserve important flexible-sized chunks at arbitrary token positions. To overcome these limitations, we propose Kara, a sliding-window KV cache compression method that performs decoding-time compression by operating only on the recently generated context. Kara leverages bidirectional attention to score and select informative KV pairs in the window. To enable flexible preservation of important semantic information, we design a Token2Chunk module to expand a subset of selected KV pairs into chunks. Furthermore, we adapt Kara to PagedAttention and develop KvLLM, an inference framework built upon vLLM, which reduces KV cache memory usage and effectively improves output throughput. Extensive experiments demonstrate consistent performance improvements of proposed Kara and KvLLM.
Chinese Translation
推理语言模型通常生成长链思维（CoT），在解码阶段会积累大量的 KV 缓存，从而导致高解码延迟和有限的吞吐量。为了解决这些问题，KV 缓存压缩作为一种有前景的技术应运而生，通过选择性地移除不重要的 KV 对，同时保留对后续解码有用的 KV 对，来减少内存开销。然而，我们发现现有的 KV 缓存压缩方法存在两个主要局限性：1）其阈值触发的压缩策略可能提供有限的吞吐量提升，甚至可能降低吞吐量，并且可能完全消除序列某些块中的 KV 对，从而可能加剧信息丢失。2）它们通常保留孤立的 KV 对或具有固定大小的块，未能在任意令牌位置保留重要的灵活大小块。为克服这些局限性，我们提出了 Kara，一种滑动窗口 KV 缓存压缩方法，该方法通过仅对最近生成的上下文进行解码时压缩来实现。Kara 利用双向注意力对窗口中的信息 KV 对进行评分和选择。为了灵活保留重要的语义信息，我们设计了一个 Token2Chunk 模块，将选定的 KV 对子集扩展为块。此外，我们将 Kara 适配到 PagedAttention，并开发了基于 vLLM 的推理框架 KvLLM，该框架减少了 KV 缓存的内存使用，并有效提高了输出吞吐量。大量实验表明，所提出的 Kara 和 KvLLM 在性能上均有一致的提升。
cs.CL / 4 / 2607.01238
SPARCLE: SPeaker-aware Aligned Representations via Contrastive Language Embeddings
SPARCLE：通过对比语言嵌入实现的说话者感知对齐表示
Abstract
Recent advances in speech synthesis have shifted from phoneme representations to direct grapheme modeling. While phonemes address the one-to-many mapping between text and acoustics, they rely on grapheme-to-phoneme (G2P) systems that fail to capture speaker-specific acoustic variation. Prior work demonstrates that grapheme-based models outperform phoneme-based systems at scale, but not in low-resource settings. In this paper, we propose SPARCLE, a speaker-aware grapheme representation model that enriches characters with their precise acoustic realizations. SPARCLE is trained with a contrastive objective to align graphemes with corresponding Wav2Vec2 acoustic representations while conditioned on speaker identity. The resulting model serves as a replacement to G2P systems for downstream text-to-speech (TTS) tasks. We demonstrate that SPARCLE improves generation quality, reducing word error rates by half in extreme low-resource settings compared to standard grapheme-based models.
Chinese Translation
近年来，语音合成的进展已从音素表示转向直接的字形建模。虽然音素处理了文本与声学之间的一对多映射，但它们依赖于无法捕捉说话者特定声学变化的字形到音素（G2P）系统。先前的研究表明，基于字形的模型在规模上优于基于音素的系统，但在低资源环境下表现不佳。本文提出了SPARCLE，一种说话者感知的字形表示模型，能够为字符提供其精确的声学实现。SPARCLE采用对比目标进行训练，以将字形与相应的Wav2Vec2声学表示对齐，同时考虑说话者身份。所得到的模型可替代G2P系统，用于下游的文本到语音（TTS）任务。我们证明，SPARCLE提高了生成质量，在极低资源环境下相比标准的基于字形的模型将词错误率降低了一半。
cs.CL / 5 / 2607.01239
Breaking Safety at the Token Boundary: How BPE Tokenization Creates Exploitable Gaps in LLM Alignment
打破令牌边界的安全性：BPE 令牌化如何在大型语言模型对齐中创造可利用的漏洞
Abstract
Character-level perturbations bypass safety alignment in modern LLMs despite leaving prompts human-readable. We identify and test a central structural mechanism: BPE tokenization fragments safety-critical words into sub-word pieces, and the three public alignment datasets we surveyed contain no intentionally fragmented inputs. The mechanism is a chain, tested end-to-end on five model families (Qwen-3-4B, Qwen-2.5-7B, Gemma-3-4B, Llama-3.1-8B, Mistral-7B). An optimization targeting safety-token fragmentation flips the first-token refusal trigger on 80-100% of refused HarmBench prompts, with 48% of those flips producing genuinely harmful outputs (per-model 29-65%; gap-vs-behavior ROC-AUC 0.66-0.98, pooled 0.84). Activation patching localizes the disrupted signal to the last ${\sim}30\%$ of layers; an alignment-data scan finds zero fragmented prompts among 30,000 examples (positive-control recall $\geq 99\%$ at attack-relevant intensities); and targeted-mutation experiments isolate safety words as the disruption locus. On the defense side, a 68-cell grid (55 trained checkpoints) shows that no DPO configuration achieves seed- and pool-stable ASR closure on the three families with closed pool-size confounds. SFT trained on fragmented prompts closes ASR on 3/5 families but only via global collapse that raises refusal on benign prompts as well, indicating the missing distribution is necessary but not sufficient under the LoRA-16 recipe we tested. To distinguish selective repair from global collapse, we introduce Conv-Benign, a candidate paired diagnostic. All ASR claims are 3-judge-calibrated (cell rankings stable across judges; absolute levels $\pm$18pp; see App.~B.13).
Chinese Translation
字符级扰动绕过了现代大型语言模型的安全对齐，尽管仍然保持提示的可读性。我们识别并测试了一个核心结构机制：BPE 令牌化将安全关键字分割成子词片段，而我们调查的三个公共对齐数据集中没有故意分割的输入。该机制是一个链条，在五个模型家族（Qwen-3-4B、Qwen-2.5-7B、Gemma-3-4B、Llama-3.1-8B、Mistral-7B）上进行了端到端测试。针对安全令牌分割的优化在80-100%的拒绝 HarmBench 提示中触发了首次令牌拒绝，且其中48%的翻转产生了真正有害的输出（每个模型29-65%；行为与差距的 ROC-AUC 为0.66-0.98，汇总为0.84）。激活修补将干扰信号定位到最后约30%的层；对齐数据扫描在30,000个示例中发现零个分割提示（正控制召回率在攻击相关强度下≥99%）；而针对突变实验则将安全词孤立为干扰源。在防御方面，一个68单元的网格（55个训练检查点）显示，没有DPO配置在三个具有封闭池大小混淆的家族中实现种子和池稳定的ASR闭合。针对分割提示进行的SFT训练在5个家族中关闭了3个的ASR，但仅通过全球崩溃的方式，这也提高了对良性提示的拒绝，表明缺失的分布在我们测试的LoRA-16配方下是必要但不充分的。为了区分选择性修复与全球崩溃，我们引入了Conv-Benign，一个候选配对诊断工具。所有ASR声明均经过3名评审的校准（单元排名在评审之间稳定；绝对水平±18个百分点；见附录B.13）。
cs.CL / 6 / 2607.01240
Prompt Framing Distorts Count-Based Evaluation of LLM Error Detection: Evidence from Numeric Anchoring
提示框架扭曲了基于计数的LLM错误检测评估：来自数字锚定的证据
Abstract
Count-based F1 is widely used as a proxy for LLM error-detection quality, but this paper shows that it can rise dramatically without a corresponding improvement in span localization, a gap termed F1 Inflation. The paper introduces ErrorBench, a controlled stress-test protocol for prompt-induced count distortion. ErrorBench evaluates six contemporary LLMs under five prompt conditions over 4,290 responses from 143 CoNLL-2014 passages. Under CoNLL-2014 M2-style scoring, anchored prompts produce up to 0.79 points of F1 Inflation, and up to 0.96 under strict matching. A 100-passage replication using the official ERRANT 3.0.0 pipeline and multi-reference scoring reproduces the pattern: averaged over six models, the Blind-to-Anchored prompt shift raises Count-F1 by +0.21 while raising multi-reference ERRANT F0.5 by only +0.04. The study finds larger count responses in highly instruction-compliant GPT/Claude systems and smaller responses in the Gemini family under this stress-test protocol. The findings suggest that LLM proofreading and document-review evaluations should avoid pre-populated error counts and should report span-aware metrics alongside count-based metrics.
Chinese Translation
基于计数的F1分数被广泛用作LLM错误检测质量的代理，但本文显示，尽管没有相应的跨度定位改善，F1分数仍可能大幅上升，这一现象被称为F1膨胀。本文引入了ErrorBench，这是一个用于提示引起的计数扭曲的受控压力测试协议。ErrorBench在五种提示条件下评估了六个当代LLM，分析了来自143个CoNLL-2014段落的4,290个响应。在CoNLL-2014 M2风格评分下，锚定提示产生了高达0.79分的F1膨胀，而在严格匹配下则高达0.96。使用官方ERRANT 3.0.0管道和多参考评分进行的100段落复制实验重现了这一模式：在六个模型的平均值中，盲对锚定提示的转变使得计数F1提高了+0.21，而多参考ERRANT F0.5仅提高了+0.04。研究发现，在高度遵循指令的GPT/Claude系统中，计数响应较大，而在Gemini系列中则较小，均是在这一压力测试协议下得出的。研究结果表明，LLM校对和文档审查评估应避免使用预填充的错误计数，并应同时报告基于计数的指标和考虑跨度的指标。
cs.CL / 7 / 2607.01241
Mapping Text to Multiplex Graph: Prompt Compression as L\'evy Walk-Guided Graph Pruning
文本映射到多重图：作为莱维行走引导的图剪枝的提示压缩
Abstract
Existing prompt compression methods treat text as flat token sequences, failing to capture the distributed nature of important information, which is often spread across multiple locations and connected through both local syntactic dependencies and global semantic relations. Such relational structure is naturally represented as a graph, where tokens or sentences become nodes and their dependencies become edges. To this end, we propose RAGP, which formulates prompt compression as Redundancy-Aware Graph Pruning on a multiplex graph that jointly models fine-grained attention-based dependencies and coarse-grained semantic relations. To efficiently identify non-redundant nodes in this heterogeneous structure (dense local subgraphs and sparse global connections), we employ Levy walks whose heavy-tailed step distribution naturally balances local exploitation with global exploration. Experiments on LongBench show that RAGP achieves an average score of 49.3 under a 4x compression ratio, outperforming existing LLM-based compression methods, such as LongLLMLingua, which attains 48.8 at a 3x compression ratio. Besides, RAGP also surpasses state-of-the-art vision-based text compression paradigms on multiple tasks. The code is available at https://anonymous.4open.science/r/RAGP-B0CB.
Chinese Translation
现有的提示压缩方法将文本视为平坦的标记序列，未能捕捉重要信息的分布特性，这些信息通常分散在多个位置，并通过局部句法依赖和全局语义关系相互连接。这种关系结构自然地表示为图，其中标记或句子成为节点，它们的依赖关系成为边。为此，我们提出了RAGP，它将提示压缩公式化为在多重图上的冗余感知图剪枝，该图共同建模细粒度的基于注意力的依赖关系和粗粒度的语义关系。为了有效识别这种异构结构（密集的局部子图和稀疏的全局连接）中的非冗余节点，我们采用了莱维行走，其重尾步长分布自然地平衡了局部开发与全局探索。在LongBench上的实验表明，RAGP在4倍压缩比下实现了49.3的平均得分，超越了现有的基于LLM的压缩方法，如LongLLMLingua，在3倍压缩比下仅达到48.8。此外，RAGP在多个任务上也超越了最先进的基于视觉的文本压缩范式。代码可在https://anonymous.4open.science/r/RAGP-B0CB获取。
cs.CL / 8 / 2607.01245
Office Comprehension Benchmark
办公理解基准
Abstract
We introduce Office Comprehension Bench (OCB), the first public benchmark to jointly evaluate LLM systems on Word, Excel, and PowerPoint comprehension over native file formats (.docx, .xlsx, .pptx) and their variants. OCB consists of two tracks. File Fidelity Q&A tests structural and visual perception of office artifacts - tables, charts, embedded images, formulas, and app-specific elements such as headers, speaker notes, and named ranges. Domain Q&A tests expert-level reasoning grounded in real-world industry documents across 12 professional domains, with queries requiring multi-step analysis and synthesis across documents. Each reference answer is decomposed into atomic, binary-gradable claims, and an ensemble of LLM judges scores responses against each claim independently. Even the strongest frontier system in its default reasoning mode reaches only about 59.3% on Domain Q&A; increasing thinking depth within a tier does not move performance materially, while moving to a higher product tier yields modest gains. We release the dataset, evaluation tooling, judge prompt, and a public leaderboard.
Chinese Translation
我们介绍了办公理解基准（Office Comprehension Bench, OCB），这是第一个公共基准，用于联合评估大型语言模型（LLM）系统在原生文件格式（.docx, .xlsx, .pptx）及其变体上的 Word、Excel 和 PowerPoint 理解能力。OCB 包含两个部分。文件保真度问答（File Fidelity Q&A）测试办公文档的结构和视觉感知，包括表格、图表、嵌入图像、公式以及特定应用元素，如页眉、演讲者备注和命名范围。领域问答（Domain Q&A）测试基于真实行业文档的专家级推理，涵盖 12 个专业领域，查询需要跨文档进行多步骤分析和综合。每个参考答案被分解为原子、二元可评估的主张，并且一组 LLM 评审员独立对每个主张进行评分。即使是最强大的前沿系统在其默认推理模式下在领域问答上的得分也仅约为 59.3%；在一个层级内增加思考深度并未显著提高性能，而转向更高的产品层级则带来了适度的提升。我们发布了数据集、评估工具、评审提示和公共排行榜。
cs.CL / 9 / 2607.01293
RuleChef: Grounding LLM Task Knowledge in Human-Editable Rules
RuleChef：将大型语言模型任务知识基础化为可编辑规则
Abstract
We present RuleChef, a framework that uses large language models (LLMs) to generate executable rules for NLP tasks such as text classification, Named Entity Recognition (NER), or relation extraction. Rules are generated based on a task description and a set of labeled examples, then they are iteratively improved based both on additional examples and on human feedback overexisting rules. RuleChef can also be used to bootstrap rules using the observed input-output pairs from any existing model for a given task. LLMs are used only at learning time, synthesizing rules and iteratively patching them based on failures measured on a held-out split. The result of this process is a fast, deterministic, and inspectable rule system. Preliminary evaluation is performed on both classification and NER tasks. We release RuleChef as open-source software under an Apache 2.0
Chinese Translation
我们提出了RuleChef，一个框架，利用大型语言模型（LLMs）为自然语言处理（NLP）任务生成可执行规则，例如文本分类、命名实体识别（NER）或关系抽取。规则是基于任务描述和一组标注示例生成的，然后根据额外示例和人类对现有规则的反馈进行迭代改进。RuleChef还可以利用给定任务的任何现有模型观察到的输入-输出对来引导规则的生成。LLMs仅在学习阶段使用，合成规则并根据在保留分割上测量的失败情况进行迭代修补。这个过程的结果是一个快速、确定性且可检查的规则系统。我们在分类和NER任务上进行了初步评估。我们将RuleChef作为开源软件发布，遵循Apache 2.0许可证。
cs.CL / 10 / 2607.01345
TurnNat: Automatic Evaluation of Turn-Taking Naturalness in Dyadic Spoken Dialogue
TurnNat：双人对话中轮流发言自然性的自动评估
Abstract
Turn-taking naturalness is central to full-duplex spoken dialogue systems, yet its automatic evaluation remains limited. Existing evaluations often rely on human judgments or behavior-specific timing metrics, making it difficult to compare heterogeneous timing failures within a unified framework. We propose TurnNat, a likelihood-based framework for automatic turn-taking naturalness evaluation in two-channel spoken dialogue. A causal turn-taking prediction model trained on natural conversations estimates future two-speaker voice-activity states, and the negative log-likelihood (NLL) of the observed future activity measures timing atypicality. TurnNat pools frame-level NLLs over turn-taking boundary units (TBUs) extracted from utterance onsets and offsets, and aggregates mean and tail TBU scores into a dialogue-level naturalness score. We further construct a controlled perturbation benchmark of paired natural and perturbed dialogue clips, validated by human naturalness judgments. Experiments on this benchmark show that TurnNat successfully identifies unnatural turn-taking perturbations across heterogeneous timing failures.
Chinese Translation
轮流发言的自然性对于全双工语音对话系统至关重要，但其自动评估仍然有限。现有的评估通常依赖于人工判断或特定行为的时间度量，这使得在统一框架内比较异质的时间失误变得困难。我们提出了TurnNat，一个基于似然性的框架，用于在双通道语音对话中自动评估轮流发言的自然性。该模型基于自然对话训练的因果轮流发言预测模型，估计未来两位说话者的语音活动状态，观察到的未来活动的负对数似然（NLL）用于衡量时间的非典型性。TurnNat在从发言开始和结束提取的轮流发言边界单元（TBU）上汇总帧级NLL，并将平均和尾部TBU得分聚合为对话级自然性得分。我们进一步构建了一个受控扰动基准，包括配对的自然和扰动对话片段，并通过人工自然性判断进行了验证。在该基准上的实验表明，TurnNat成功识别出异质时间失误中的不自然轮流发言扰动。
cs.CL / 11 / 2607.01388
RusFinChain: A Russian Benchmark for Verifiable Chain-of-Thought Reasoning in Finance with Fuzzy-Aligned Evaluation
RusFinChain：一个用于金融领域可验证思维链推理的俄罗斯基准，具有模糊对齐评估
Abstract
Multi-step symbolic reasoning is essential for robust financial analysis, yet most benchmarks neglect intermediate reasoning steps. FINCHAIN introduced verifiable Chain-of-Thought (CoT) evaluation but is limited to English. FINESSE-Bench includes a Russian block but relies on multiple-choice questions without step-level supervision. We present RusFinChain, the first Russian-language symbolic benchmark for verifiable CoT reasoning in finance. It spans 17 domains, 172 topics, and comprises 5,280 parameterized examples from executable Python templates, ensuring contamination-free evaluation. Each example includes a gold-standard reasoning chain with intermediate numeric values for automatic verification. We also introduce enhanced metrics: Fuzzy Numeric Alignment and Soft-Attention Alignment. We evaluate 8 open-weight LLMs on a stratified sample, generating 8,100 responses. Results reveal a substantial reasoning gap: models achieve Hard F1 of ~0.65 for step alignment, but only ~29% of final answers are correct. Our fuzzy and soft metrics show stronger correlation with final-answer correctness (Spearman rho approx 0.48) than the original ChainEval (rho approx 0.38-0.46), demonstrating superior diagnostic power. We release dataset, code, and evaluation framework to foster verifiable financial AI for the Russian-speaking community.
Chinese Translation
多步骤符号推理对于稳健的金融分析至关重要，但大多数基准忽视了中间推理步骤。FINCHAIN引入了可验证的思维链（Chain-of-Thought, CoT）评估，但仅限于英语。FINESSE-Bench包含一个俄语模块，但依赖于没有步骤级监督的多项选择题。我们提出了RusFinChain，这是第一个用于金融领域可验证CoT推理的俄语符号基准。它涵盖了17个领域、172个主题，并包括来自可执行Python模板的5,280个参数化示例，确保了无污染的评估。每个示例都包含一个金标准推理链，带有中间数值以便于自动验证。我们还引入了增强的评估指标：模糊数值对齐（Fuzzy Numeric Alignment）和软注意力对齐（Soft-Attention Alignment）。我们在一个分层样本上评估了8个开放权重的语言模型（LLMs），生成了8,100个响应。结果显示出显著的推理差距：模型在步骤对齐方面的Hard F1约为0.65，但最终答案的正确率仅约为29%。我们的模糊和软指标与最终答案的正确性显示出更强的相关性（Spearman rho约为0.48），优于原始的ChainEval（rho约为0.38-0.46），显示出更强的诊断能力。我们发布数据集、代码和评估框架，以促进俄语社区的可验证金融人工智能。
cs.CL / 12 / 2607.01392
Multi-Objective Exploration and Preference Optimization via Mutual Information
通过互信息进行多目标探索与偏好优化
Abstract
Aligning large language models with diverse and heterogeneous human values requires multi-objective alignment methods to effectively trade off conflicting preference dimensions. Current methods achieve this trade-off by training policies conditioned on preference vectors and leveraging online direct preference optimization. However, exploration uncertainty can cause the reward distributions of responses generated under different preference vectors to overlap, and the generated responses may fail to effectively align with the corresponding preference vectors. In this paper, we propose Multi-Objective Exploration and Preference Optimization via Mutual Information (MI-EPO), an information-theoretic framework. It unifies multi-objective exploration and alignment by maximizing the joint conditional mutual information among generated responses, preference feedback, and preference vectors. By incorporating a probabilistic routing mechanism, MI-EPO naturally decomposes objective alignment and preference-aware exploration, encouraging the model to generate responses that are distinguishable and aligned with different preference conditions. Experiments on safe alignment and helpful assistant tasks show that MI-EPO significantly improves the alignment between generated responses and preference vectors, makes the outputs more controllable, and achieves stable trade-offs across multiple objectives.
Chinese Translation
将大型语言模型与多样化和异质的人类价值观对齐，需要多目标对齐方法以有效权衡相互冲突的偏好维度。目前的方法通过训练基于偏好向量的策略并利用在线直接偏好优化来实现这种权衡。然而，探索的不确定性可能导致在不同偏好向量下生成的响应的奖励分布重叠，从而使生成的响应未能有效地与相应的偏好向量对齐。本文提出了一种信息论框架——通过互信息进行多目标探索与偏好优化（MI-EPO）。该框架通过最大化生成响应、偏好反馈和偏好向量之间的联合条件互信息，统一了多目标探索与对齐。通过引入概率路由机制，MI-EPO自然地将目标对齐与偏好感知探索进行分解，鼓励模型生成可区分且与不同偏好条件对齐的响应。在安全对齐和有用助手任务上的实验表明，MI-EPO显著提高了生成响应与偏好向量之间的对齐程度，使输出更加可控，并在多个目标之间实现了稳定的权衡。
cs.CL / 13 / 2607.01420
MultAttnAttrib: Training-Free Multimodal Attribution in Long Document Question Answering
MultAttnAttrib：无训练的长文档问答中的多模态归因
Abstract
As grounded QA systems are increasingly deployed in AI assistants, accurately attributing generated answers to evidence is critical for user trust and model safety. While unimodal attributions have been explored in depth, the multimodal setting remains relatively under-researched. As a result, we introduce MultAttnAttrib, a training-free attribution-generation method that leverages a model's prefill pass, selected attention heads, and calibrated thresholds to locate source evidence within a document. To establish baseline results for the method, we introduce MultAttrEval, a complementary benchmark dataset annotated with fine-grained, ground-truth attributions for answer components grounded in multimodal source documents. To our knowledge, this is the first evaluation dataset designed specifically for multimodal attribution in long-form documents. Experimental results show that MultAttnAttrib consistently outperforms a variety of attribution-generation methods, including several strong prompting-based approaches and matches the latest frontier models such as GPT 5.4. Our method not only substantially improves attribution accuracy for both unimodal and multimodal attribution types, but also produces attributions at up to one-seventh of the direct inference latency compared to prompting on the same base model.
Chinese Translation
随着基于证据的问答系统在人工智能助手中的应用日益广泛，准确地将生成的答案归因于证据对于用户信任和模型安全至关重要。尽管单模态归因已经得到了深入研究，但多模态环境仍然相对缺乏研究。因此，我们提出了MultAttnAttrib，这是一种无训练的归因生成方法，利用模型的预填充过程、选定的注意力头和校准阈值在文档中定位源证据。为了建立该方法的基线结果，我们引入了MultAttrEval，这是一个补充性基准数据集，包含针对多模态源文档中答案组件的细粒度真实归因的注释。据我们所知，这是第一个专门为长文档中的多模态归因设计的评估数据集。实验结果表明，MultAttnAttrib在多种归因生成方法中表现出色，包括几种强大的基于提示的方法，并与最新的前沿模型如GPT 5.4相匹配。我们的方法不仅显著提高了单模态和多模态归因类型的归因准确性，而且在与同一基础模型的提示相比时，归因生成的延迟可减少至直接推理延迟的七分之一。
cs.CL / 14 / 2607.01431
IsoSci: A Benchmark of Isomorphic Cross-Domain Science Problems for Evaluating Reasoning versus Knowledge Retrieval in LLMs
IsoSci：用于评估大型语言模型中推理与知识检索的同构跨领域科学问题基准
Abstract
We introduce ISOSCI, a benchmark of isomorphic cross-domain science problem pairs that separates reasoning ability from domain knowledge retrieval in LLM evaluation. Each pair shares identical logical structure but requires different domain-specific knowledge, enabling controlled attribution of reasoning-mode gains. Across five model pairs spanning four model families, we find that 91.3% of reasoning-mode gains are knowledge-dependent rather than structure-invariant (63/69 gains; Wilson 95% CI [82.3%, 96.0%]), directly challenging the assumption that chain-of-thought reasoning improves short-horizon procedural scientific problem-solving. Reasoning toggles on highly capable models provide less than 5 percentage points accuracy gain across all domains, and a reasoning-specialized model (o3-mini) that outperforms its standard counterpart on GPQA Diamond (+19.2 percentage points) underperforms on ISOSCI (-24.7 percentage points), showing that benchmark choice determines conclusions about reasoning utility. We release ISOSCI at https://huggingface.co/datasets/isosci/isosci
Chinese Translation
我们介绍了ISOSCI，这是一个同构跨领域科学问题对的基准，旨在将推理能力与领域知识检索在大型语言模型（LLMs）评估中区分开来。每对问题共享相同的逻辑结构，但需要不同的领域特定知识，从而实现对推理模式增益的控制归因。在涵盖四个模型家族的五对模型中，我们发现91.3%的推理模式增益依赖于知识，而非结构不变（63/69增益；Wilson 95% CI [82.3%, 96.0%]），这直接挑战了链式思维推理改善短期程序性科学问题解决的假设。在所有领域中，高能力模型的推理切换提供的准确率增益不足5个百分点，而在GPQA Diamond上表现优于其标准版本的推理专用模型（o3-mini，增益+19.2个百分点）在ISOSCI上却表现不佳（下降-24.7个百分点），显示基准选择决定了关于推理效用的结论。我们在https://huggingface.co/datasets/isosci/isosci发布ISOSCI。
cs.CL / 15 / 2607.01440
FaithMed: Training LLMs For Faithful Evidence-Based Medical Reasoning
FaithMed：为忠实的基于证据的医学推理训练大型语言模型
Abstract
Faithful reasoning is essential in medicine, where clinical decisions require transparent justification grounded in reliable evidence. Current medical LLMs either lack active access to evidence or use retrieved evidence without supervising how it should be appraised and applied during reasoning. To address this, we formalize evidence-based medicine principles as process-level criteria and introduce FaithMed, a framework that combines clinician-designed, automatically refined rubrics with reinforcement learning using step-level process reward assignment and advantage grouping. Across seven medical benchmarks, FaithMed improves over agentic-search baselines (+9% on average) and outcome-only RL (+5.8%), while raising average evidence-based medicine rubric scores over agentic-search Qwen3 baselines (+15.5%). This work demonstrates that explicit step-level supervision can improve both task success and the faithfulness of the reasoning process. Code is available at https://github.com/cxcscmu/FaithMed.
Chinese Translation
忠实的推理在医学中至关重要，因为临床决策需要基于可靠证据的透明理由。目前的医学大型语言模型（LLMs）要么缺乏对证据的主动访问，要么在推理过程中使用检索到的证据而没有监督其如何进行评估和应用。为了解决这个问题，我们将基于证据的医学原则形式化为过程级标准，并引入FaithMed，一个结合了临床医生设计的、自动优化的评分标准与强化学习的框架，采用逐步过程奖励分配和优势分组。在七个医学基准测试中，FaithMed在代理搜索基线之上平均提高了9%（+9%），在仅基于结果的强化学习上提高了5.8%（+5.8%），同时在代理搜索Qwen3基线上的平均基于证据的医学评分提高了15.5%（+15.5%）。这项工作表明，明确的逐步监督可以提高任务成功率和推理过程的忠实性。代码可在 https://github.com/cxcscmu/FaithMed 获取。
cs.CL / 16 / 2607.01457
Grounded Optimization: A Layered Engineering Framework for Reducing LLM Hallucination in Automated Personal Document Rewriting
基于实证的优化：减少大型语言模型幻觉的分层工程框架在自动个人文档重写中的应用
Abstract
Large language models (LLMs) are increasingly applied to resume optimization for applicant tracking systems, introducing hallucination failures distinct from general text generation: anachronistic technology injection, cross-domain terminology contamination, structural mutation, and content fabrication. We present Grounded Optimization, a five-layer framework combining temporal context validation, deterministic contamination detection, structural invariant enforcement, prompt-level grounding, and an evaluator agent. In ablation experiments across three LLMs, four temperature settings, and six layer configurations on 25 synthetic resumes spanning 14 industries, undefended baselines produce 2.48-5.36 detected hallucinations per resume. Among detectors independent of the active defenses, temporal hallucinations are reduced by 50-95% across all conditions; overall detected hallucination rate falls to 0.04-0.24. Prompt-level grounding alone achieves zero detected hallucinations at low temperature with a capable instruction-following model; higher temperatures and weaker models reveal the need for the deterministic layers as a complement. We release the contamination taxonomy, evaluation code, and raw data.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于申请者跟踪系统中的简历优化，导致与一般文本生成不同的幻觉失败：过时技术注入、跨领域术语污染、结构突变和内容虚构。我们提出了基于实证的优化（Grounded Optimization），这是一个五层框架，结合了时间上下文验证、确定性污染检测、结构不变性强制、提示级别的实证基础和评估代理。在对三种LLM、四种温度设置和六种层配置进行的消融实验中，25份涵盖14个行业的合成简历显示，未防御的基线每份简历产生2.48-5.36个检测到的幻觉。在独立于主动防御的检测器中，时间幻觉在所有条件下减少了50-95%；整体检测到的幻觉率降至0.04-0.24。仅凭提示级别的实证基础，在低温下使用能够遵循指令的模型实现了零检测到的幻觉；更高的温度和较弱的模型显示出确定性层作为补充的必要性。我们发布了污染分类法、评估代码和原始数据。
cs.CL / 17 / 2607.01464
Comparing Architectures for Supervised Political Scaling
监督政治标定架构的比较
Abstract
Text scaling, the task of positioning political actors on an ideological scale, is a fundamental task in political analysis. To ease the need for manual analysis, various NLP methods have been proposed for this task, including classification- and regression-based approaches, showing successes as well as limitations. The goal of our paper is to consolidate the state of the art in this area. We ask two questions: (a) Can the performance of scaling methods be improved by predicting scales not individually but jointly? (b) Is there a middle ground between classification and regression?
Chinese Translation
文本标定，即将政治行为者置于意识形态标度上的任务，是政治分析中的一项基础任务。为了减少对手动分析的需求，已经提出了多种自然语言处理（NLP）方法来完成这一任务，包括基于分类和回归的方法，这些方法展现了成功之处和局限性。我们论文的目标是整合该领域的最新进展。我们提出两个问题：（a）通过联合预测标度而非单独预测，是否可以提高标定方法的性能？（b）分类和回归之间是否存在折中方案？
cs.CL / 18 / 2607.01502
From Monolingual to Multilingual: Evaluating Mamba for ASR in South African Languages
从单语到多语：评估Mamba在南非语言中的自动语音识别
Abstract
Recent advances in automatic speech recognition (ASR) have explored different sequence models, including Conformer-based models and newer state space models such as Mamba. Although prior work has evaluated these architectures in multiple languages, their effectiveness in African languages remains underexplored. In this work, we evaluate Mamba for ASR on seven South African languages. In monolingual experiments, each model is trained on 50 hours of speech per language, and we compare Mamba to a Conformer baseline of similar parameter scale. Mamba achieves similar recognition accuracy to Conformer while using fewer computational resources and training faster. We further evaluate generalization in this setting and find that both models struggle to generalize to speech that is much longer than what they were trained on. We then study multilingual ASR using Mamba models, where the baseline is pooling all languages together. On top of this, we tested three extensions: training with language-family information by adding both language and language-family embeddings as biases to the downsampled acoustic representations, and multitask learning with a CTC ASR objective and a language identification (LID) head. We find that multilingual training consistently improves performance over monolingual training. However, adding explicit language information does not improve in-domain performance but does improve cross-corpus robustness. We conducted ablation studies in low-resource multilingual settings using 5-hour and 10-hour per-language training data, where we observed gains from using language embeddings and further demonstrated that removing or altering them hurt model performance. Lastly, we analysed these embeddings and find that they do not capture linguistic similarity in a typological sense, but instead act as task-specific control vectors.
Chinese Translation
最近的自动语音识别（ASR）进展探索了不同的序列模型，包括基于Conformer的模型和较新的状态空间模型如Mamba。尽管之前的研究已在多种语言中评估了这些架构，但它们在非洲语言中的有效性仍然未被充分探索。在本研究中，我们评估了Mamba在七种南非语言中的ASR表现。在单语实验中，每个模型在每种语言上训练50小时的语音，并将Mamba与具有相似参数规模的Conformer基线进行比较。Mamba在识别准确性上与Conformer相当，同时使用更少的计算资源并且训练速度更快。我们进一步评估了在此设置下的泛化能力，发现这两种模型在处理远超训练时长的语音时均表现不佳。随后，我们研究了使用Mamba模型的多语种ASR，其中基线是将所有语言汇总。在此基础上，我们测试了三种扩展：通过将语言和语言家族嵌入作为偏置添加到下采样的声学表示中，利用语言家族信息进行训练，以及使用CTC ASR目标和语言识别（LID）头的多任务学习。我们发现多语种训练的性能始终优于单语训练。然而，添加显式语言信息并未改善领域内的表现，但确实提高了跨语料库的鲁棒性。我们在低资源多语种环境中进行了消融研究，使用每种语言5小时和10小时的训练数据，观察到使用语言嵌入带来的性能提升，并进一步证明去除或改变这些嵌入会损害模型性能。最后，我们分析了这些嵌入，发现它们并未在类型学意义上捕捉语言相似性，而是作为特定任务的控制向量。
cs.CL / 19 / 2607.01517
Parameter Golf: What Really Works?
参数高尔夫：什么真正有效？
Abstract
How far can a language model improve under a strict artifact budget? Parameter Golf posed this question as an open community challenge in which participants trained the best language model, with the complete artifact (training code + compressed weights) required to fit within 16 MB and be trained in under ten minutes on 8xH100 SXM GPUs. Quality was measured in bits-per-byte (BPB), the average number of bits required to encode each byte of unseen text. We analyze 2,037 pull requests and 1,430 clean scored submissions from the contest, build a taxonomy of 84 optimization techniques, and measure each technique's contribution to BPB. The verified leaderboard score dropped from 1.2244 to 1.058 BPB across three phases -- a 13.6% reduction, despite individual techniques rarely improving BPB by more than 1%. We show that most gains in techniques shrink across competitive submissions, isolating the few methods that improve performance across stacks.
Chinese Translation
在严格的工件预算下，语言模型能提高多少？参数高尔夫（Parameter Golf）提出了这个问题，并作为一个开放的社区挑战，参与者需要训练出最佳的语言模型，要求完整的工件（训练代码 + 压缩权重）必须在16 MB以内，并且在8xH100 SXM GPU上训练时间不超过十分钟。质量通过每字节比特数（bits-per-byte, BPB）来衡量，即编码每个未见文本字节所需的平均比特数。我们分析了来自比赛的2,037个拉取请求和1,430个清晰评分的提交，构建了84种优化技术的分类法，并测量了每种技术对BPB的贡献。经过验证的排行榜得分在三个阶段中从1.2244下降到1.058 BPB，减少了13.6%，尽管各个技术的BPB提升通常不超过1%。我们展示了大多数技术的增益在竞争性提交中缩小，孤立出少数几种在多个模型中提升性能的方法。
cs.CL / 20 / 2607.01538
Can Language Models Actually Retrieve In-Context? Drowning in Documents at Million Token Scale
语言模型真的能够进行上下文检索吗？在百万标记规模的文档中淹没
Abstract
Language models (LMs) raise an intriguing alternative to vector-based retrieval: conditioning on an in-context corpus and directly generating a relevant answer. However, prior work has largely focused on proprietary systems or the smaller-scale reranking task, leaving corpus-scale in-context retrieval largely unexplored. In this work, we present the first systematic study of in-context retrieval on two scales practical retrievers demand: million-token corpora and length-generalization far beyond training-time sizes. We first introduce BlockSearch, a 0.6B LM retriever whose architectural and training modifications improve over prior LM baselines and length-generalize up to 10 times beyond its training regime. Nevertheless, retrieval still collapses under more extreme extrapolation. We trace this failure to an attention dilution effect: as the corpus grows, irrelevant documents dominate the softmax denominator, reducing the normalized mass on the gold document even when its pre-softmax score stays high. Motivated by this analysis, we introduce length-aware adjustments to the attention softmax and document-level sparse attention. With these modifications, at the million-token scale, our model matches dense retrieval on widely studied benchmarks (e.g, MS MARCO and NQ), while outperforming the concurrent model MSA despite being 7 times smaller. Furthermore, it significantly outperforms dense retrieval on tasks requiring entirely different notions of similarity, such as LIMIT, achieving a 3 times higher score. Together, our results position in-context retrieval a promising alternative to classical retrieval while emphasizing attention control under extreme context growth as a new challenge.
Chinese Translation
语言模型（LMs）为基于向量的检索提供了一种引人注目的替代方案：基于上下文语料库进行条件生成并直接生成相关答案。然而，之前的研究主要集中在专有系统或小规模的重排序任务上，导致语料库规模的上下文检索尚未得到充分探索。在本研究中，我们首次系统地研究了在实际检索器所需的两个规模上的上下文检索：百万标记语料库和远超训练时规模的长度泛化。我们首先介绍了BlockSearch，这是一种0.6B的LM检索器，其架构和训练修改在先前的LM基线之上有所改进，并且长度泛化能力可达到训练阶段的10倍。然而，在更极端的外推情况下，检索仍然崩溃。我们将这一失败归因于注意力稀释效应：随着语料库的增长，无关文档主导了softmax的分母，即使其预softmax分数保持较高，金文档的归一化质量也会降低。基于这一分析，我们引入了对注意力softmax和文档级稀疏注意力的长度感知调整。通过这些修改，在百万标记规模上，我们的模型在广泛研究的基准（如MS MARCO和NQ）上与密集检索相匹配，同时在规模上仅为其七分之一的同时超越了并行模型MSA。此外，它在需要完全不同相似性概念的任务（如LIMIT）上显著优于密集检索，达到了三倍的得分。综合来看，我们的结果表明，上下文检索是经典检索的一个有前景的替代方案，同时强调在极端上下文增长下的注意力控制是一个新的挑战。
cs.CL / 21 / 2607.01557
DiPS: Dialogue Policy Selection for High-Stakes Persuasion Agents
DiPS：高风险劝说代理的对话策略选择
Abstract
Large Language Models (LLMs) often struggle with persuasion in high-stakes scenarios. People's individual personalities and concerns require tailored strategies rather than a one-size-fits-all approach. To address this challenge, we focus on a fire-rescue scenario in which an operator must persuade a resident to evacuate as a high-stakes persuasion domain and propose Dialogue Policy Selection (DiPS), a Q-learning framework to dynamically select persuasion strategies adapted to the evolving conversational context. Specifically, we train a critic, trained to maximize the chance of evacuation success, to select a persuasion policy at each turn based on the resident's recent utterances.We then evaluate DiPS against multiple baselines in both simulated and real human interactions. We find that DiPS achieves higher evacuation success than a zero-shot LLM and generic RAG-augmented approach.
Chinese Translation
大型语言模型（LLMs）在高风险场景中的劝说能力往往面临挑战。人们的个性和关注点各不相同，因此需要量身定制的策略，而非一种适用于所有的通用方法。为了解决这一挑战，我们聚焦于一个火灾救援场景，在该场景中，操作员必须说服居民撤离，作为高风险劝说领域，并提出对话策略选择（DiPS），这是一种基于Q学习的框架，旨在根据不断变化的对话上下文动态选择劝说策略。具体而言，我们训练一个评估者，旨在最大化撤离成功的机会，以便在每个回合中根据居民的近期发言选择劝说策略。随后，我们在模拟和真实人际互动中对DiPS进行多种基线的评估。我们发现，DiPS在撤离成功率上优于零-shot LLM和通用的RAG增强方法。
cs.CL / 22 / 2607.01581
Beyond Skepticism: Evaluating LLMs Pedagogical Intent Reasoning with the Adaptive Pedagogical Vigilance Framework
超越怀疑：通过自适应教学警觉框架评估大型语言模型的教学意图推理
Abstract
The capacity of Large Language Models (LLMs) to reason about pedagogical intent within instructional communication remains underexplored, particularly in educational domains such as translation pedagogy. To address this, we propose the \textbf{Adaptive Pedagogical Vigilance (APV)} framework, a novel computational formalism that reframes communicative vigilance as an adaptive mechanism for optimizing learning through intent inference. APV formalizes the problem via a Bayesian Pedagogical Intent Inference Engine (PIIE), which models how instructors select content to maximize pedagogical utility and how vigilant learners should inversely reason about latent instructional configurations -- encompassing genre, stance, and incentives. We evaluate APV through a three-tier hierarchy: distinguishing instructional genre, reasoning about structured pedagogical setups, and generalizing to authentic educational discourse. Experiments on leading LLMs (e.g., GPT-4o, Claude 3.5) show that APV substantially improves model vigilance. It achieves the strongest discrimination between pedagogical and exposure-based content, correlates highly with human judgments ($r=0.958$), and maintains robust performance on naturalistic data where baseline methods degrade. This work establishes a unified framework for assessing and enhancing LLMs' understanding of pedagogical motives, advancing the development of more reliable AI-assisted learning systems.
Chinese Translation
大型语言模型（LLMs）在教学交流中推理教学意图的能力仍然未得到充分探索，特别是在翻译教学等教育领域。为了解决这一问题，我们提出了 extbf{自适应教学警觉（APV）}框架，这是一种新颖的计算形式，重新定义了交流警觉作为一种适应性机制，以通过意图推理优化学习。APV通过贝叶斯教学意图推理引擎（PIIE）形式化了这一问题，该引擎建模了教师如何选择内容以最大化教学效用，以及警觉的学习者应如何反向推理潜在的教学配置——涵盖体裁、立场和激励。我们通过三级层次评估APV：区分教学体裁、推理结构化教学设置以及推广到真实教育话语。对领先的LLMs（如GPT-4o、Claude 3.5）的实验表明，APV显著提高了模型的警觉性。它在教学内容与基于曝光的内容之间实现了最强的区分，与人类判断高度相关（$r=0.958$），并在基线方法退化的自然数据上保持了稳健的表现。这项工作建立了一个统一的框架，用于评估和增强LLMs对教学动机的理解，推动了更可靠的AI辅助学习系统的发展。
cs.CL / 23 / 2607.01602
ProWAFT: A ROMA-LPD Instance for Workload-Aware and Dynamic Fault Tolerance in FPGA-Based CNN Accelerators
ProWAFT：一种针对FPGA基础CNN加速器的工作负载感知和动态容错的ROMA-LPD实例
Abstract
SRAM-based FPGAs provide an attractive platform for energy- and latency-constrained CNN inference at the network edge, yet transient faults can lead to silent errors that compromise reliability. Always-on redundancy (e.g., full TMR) improves correctness but incurs substantial performance and energy overhead, while reactive recovery may introduce unacceptable latency on the critical path. We propose \textbf{ProWAFT}, a proactive workload-aware fault-tolerance framework for FPGA-based CNN accelerators that uses partial reconfiguration to selectively apply TMR across reconfigurable partitions. ProWAFT quantifies workload criticality, models fault propagation and reconfiguration overhead, and selects configurations that minimize a composite objective over latency, energy, and reliability risk. Implemented on a Xilinx Zynq UltraScale+ ZCU104 platform with six reconfigurable regions and evaluated on a 500-task trace derived from ResNet-18, MobileNetV2, and EfficientNet-Lite under time-varying SEU injection, ProWAFT achieves lower composite cost than static TMR and reactive reconfiguration while maintaining high task success rate and near-baseline throughput with low online decision overhead.
Chinese Translation
基于SRAM的FPGA为在网络边缘进行受限于能量和延迟的CNN推理提供了一个有吸引力的平台，但瞬态故障可能导致无声错误，从而影响可靠性。始终开启的冗余（例如，完全三重冗余（TMR））提高了正确性，但会带来显著的性能和能量开销，而反应式恢复可能在关键路径上引入不可接受的延迟。我们提出了 extbf{ProWAFT}，一种针对FPGA基础CNN加速器的主动工作负载感知容错框架，该框架利用部分重配置在可重配置分区中选择性地应用TMR。ProWAFT量化工作负载的关键性，建模故障传播和重配置开销，并选择最小化延迟、能量和可靠性风险的复合目标的配置。在Xilinx Zynq UltraScale+ ZCU104平台上实现，具有六个可重配置区域，并在基于ResNet-18、MobileNetV2和EfficientNet-Lite的500任务轨迹上进行评估，在时间变化的单粒子翻转（SEU）注入下，ProWAFT实现了比静态TMR和反应式重配置更低的复合成本，同时保持高任务成功率和接近基线的吞吐量，且在线决策开销低。
cs.CL / 24 / 2607.01727
When Does Generating More Help? Disentangling Fixed-Source Synthesis from Source Expansion in Synthetic Data Scaling
何时生成更多数据有帮助？解构合成数据扩展中的固定源合成与源扩展
Abstract
Synthetic data can be scaled along two routes: Source Expansion (SE), which enlarges the source by adding seed materials or generators, and Fixed-Source Synthesis (FSS), which holds the source fixed and scales the generation budget. Existing scaling studies typically expand the source as the data grows, conflating SE with FSS and leaving FSS underexplored. We isolate FSS by holding the seed-question pool and teacher model fixed, varying only the per-question response budget under Rejection Sampling (RS). We adapt the rectified scaling law to FSS, deriving it from how repeated sampling covers a fixed source. Empirically, the derived form, fit on low budgets, predicts performance at the held-out highest budget for every evaluated teacher--student pair. At matched total-sample budgets, SE and FSS are comparable at small budgets; at large budgets, adding seed questions outperforms spending the same budget on more responses. Within FSS, however, neither synthesizing additional questions from the existing seeds nor varying the synthesis protocol outperforms plain RS at matched budgets. FSS is thus a bounded scaling axis and a controlled setting for comparing synthesis protocols. We will release our code and data to facilitate further research.
Chinese Translation
合成数据可以通过两种方式进行扩展：源扩展（Source Expansion, SE），通过添加种子材料或生成器来扩大源；固定源合成（Fixed-Source Synthesis, FSS），保持源不变并扩大生成预算。现有的扩展研究通常在数据增长时扩展源，将 SE 与 FSS 混淆，导致 FSS 的研究不足。我们通过固定种子问题池和教师模型，仅在拒绝采样（Rejection Sampling, RS）下改变每个问题的响应预算，从而将 FSS 独立出来。我们将修正的扩展法则适应于 FSS，从重复采样如何覆盖固定源中推导出该法则。从经验上看，基于低预算的推导形式能够预测在每对评估的教师-学生组合中，保留的最高预算下的表现。在匹配的总样本预算下，SE 和 FSS 在小预算时是可比的；而在大预算下，添加种子问题的效果优于在更多响应上花费相同预算。然而，在 FSS 内部，从现有种子合成额外问题或改变合成协议在匹配预算下均不优于简单的 RS。因此，FSS 是一个有限的扩展轴，也是比较合成协议的受控环境。我们将发布我们的代码和数据，以促进进一步的研究。
cs.CL / 25 / 2607.01733
Rethinking Speech-LLM Integration for ASR: Effective Joint Speech-Text Training by Interleaving
重新思考语音-大语言模型集成在自动语音识别中的应用：通过交错实现有效的联合语音-文本训练
Abstract
Speech-LLM integration has shown promising results by leveraging extensive textual pretraining, yet its specific benefits for automatic speech recognition (ASR) remain unclear. We observe that as supervised ASR training data increases, the contribution of LLM priors becomes less evident, and simple speech-text joint training under-utilizes textual knowledge. We therefore propose Joint Speech-Text Interleaved Pretraining (JSTIP), an ASR-oriented pretraining strategy that constructs word-level and segment-level interleaved speech-text sequences within aligned pairs for speech-LLM architectures that accept continuous inputs. Experiments on 38k hours of ASR data show consistent entity accuracy improvement compared to ASR-only and joint speech-text training baselines. JSTIP achieves on-par entity recognition performance using domain transcription text compared to synthetic speech-text pairs, simplifying domain adaptation. Benefiting from textual pretraining and domain text data, JSTIP is competitive with open-source ASR and Speech-LLM systems in medical entity recognition. The zero-shot speech question answering behaviors further suggest that interleaving reduces the speech-text modality gap and preserves the LLM generative prior, which is likely the reason for the entity improvements on the ASR task.
Chinese Translation
语音-大语言模型（Speech-LLM）集成通过利用广泛的文本预训练显示出良好的效果，但其对自动语音识别（ASR）的具体益处仍不明确。我们观察到，随着监督式ASR训练数据的增加，LLM先验的贡献变得不那么明显，而简单的语音-文本联合训练未能充分利用文本知识。因此，我们提出了联合语音-文本交错预训练（Joint Speech-Text Interleaved Pretraining，JSTIP），这是一种面向ASR的预训练策略，旨在为接受连续输入的语音-LLM架构构建词级和段级交错的语音-文本序列。基于38,000小时的ASR数据的实验表明，与仅使用ASR和联合语音-文本训练的基线相比，JSTIP在实体识别准确性上有一致的提升。与合成语音-文本对相比，JSTIP在使用领域转录文本时实现了相当的实体识别性能，从而简化了领域适应。得益于文本预训练和领域文本数据，JSTIP在医疗实体识别方面与开源ASR和语音-LLM系统具有竞争力。零样本语音问答行为进一步表明，交错处理减少了语音-文本模态间的差距，并保留了LLM生成先验，这可能是ASR任务中实体提升的原因。
cs.CL / 26 / 2607.01792
PARTREP: Learning What to Repeat for Decoder-only LLMs
PARTREP：为仅解码器的语言模型学习重复内容
Abstract
While decoder-only LLMs excel at a vast array of natural language tasks, it suffers from an asymmetric information flow induced by causal attention: later tokens are richer in contextual grounding than earlier ones. A simple and effective remedy is prompt repetition -- just appending a second copy of prompt before generation can redistribute grounding across positions and improve reasoning performance. However, full repetition of the original prompt doubles the KV cache footprint and quadruples attention cost during prefill, making it impractical for long-context settings. We propose PartRep, a selective augmentation method that appends only the most informative tokens -- rather than the entire prompt. We use token-wise negative log-likelihood (NLL) as a selection signal, motivated by the hypothesis that less predictable tokens are less recoverable from surrounding context and therefore benefit more from late-position repetition. To avoid the heavy cost of a full forward pass for scoring, we train a lightweight gate that predicts high-NLL tokens from early-layer hidden states, enabling token selection during mid-prefill via early exit. Across eight benchmarks (including MMLU, GSM8K, and RULER) and three model families (Qwen2.5, Llama3.2, Gemma4), PartRep retains most of the gains of full repetition while using only 59.4\% of its KV cache and 79.0\% of its prefill FLOPs.
Chinese Translation
尽管仅解码器的语言模型（LLMs）在广泛的自然语言任务中表现出色，但由于因果注意力引发的信息流不对称，它们面临挑战：后续的标记在上下文基础上比前面的标记更为丰富。一种简单有效的解决方案是提示重复——在生成之前仅需在提示前添加第二份副本，可以重新分配各个位置的上下文基础，从而提高推理性能。然而，完全重复原始提示会使KV缓存占用加倍，并在预填充期间使注意力成本增加四倍，这在长上下文设置中显得不切实际。我们提出了PartRep，这是一种选择性增强方法，仅添加最具信息量的标记，而不是整个提示。我们使用逐标记的负对数似然（NLL）作为选择信号，基于这样一个假设：不太可预测的标记在周围上下文中更难恢复，因此更能从后位置的重复中受益。为了避免进行完整前向传递以进行评分的高昂成本，我们训练了一个轻量级的门控机制，该机制从早期层的隐藏状态预测高NLL标记，从而在中期预填充过程中通过提前退出实现标记选择。在八个基准测试（包括MMLU、GSM8K和RULER）和三个模型系列（Qwen2.5、Llama3.2、Gemma4）中，PartRep在仅使用59.4%的KV缓存和79.0%的预填充FLOPs的情况下，保留了大部分完全重复带来的收益。
cs.CL / 27 / 2607.01802
On the Limits of Steering Vectors for Preference-Aligned Generation
关于偏好对齐生成的引导向量的局限性
Abstract
Steering vectors have emerged as a promising approach to controlled text generation, offering interpretable, training-free mechanisms for shaping model outputs. However, their practical generality remains poorly understood. We study the limits of steering vector generalization along three dimensions: trait expressibility, task transfer, and multi-trait composition. Using the PLUME writing personalization benchmark, we extract steering vectors for a range of preferences and evaluate them on summarization and email-writing tasks across two open-source models (Qwen2.5-7B-Instruct and Llama3.1-8B-Instruct). We find that steering effectiveness varies substantially across traits. We further show that steering effectiveness can degrade when vectors extracted from positive and negative style examples are transferred to downstream writing personalization tasks. Finally, we compare common methods for composing multiple steering vectors and find that all methods suffer significant drops in trait expression as more vectors are added, with a tradeoff between coherence and expressibility that requires per-setting hyperparameter tuning. Taken together, our results suggest that steering vectors face meaningful limits as a general-purpose tool for preference alignment.
Chinese Translation
引导向量作为一种有前景的受控文本生成方法，提供了可解释的、无训练的机制来塑造模型输出。然而，它们的实际通用性仍然不够清晰。我们从三个维度研究引导向量的泛化局限性：特征可表达性、任务迁移和多特征组合。利用PLUME写作个性化基准，我们提取了多种偏好的引导向量，并在两个开源模型（Qwen2.5-7B-Instruct和Llama3.1-8B-Instruct）上评估它们在摘要和电子邮件写作任务中的表现。我们发现，引导效果在不同特征之间存在显著差异。我们进一步表明，当从正面和负面风格示例中提取的向量转移到下游写作个性化任务时，引导效果可能会下降。最后，我们比较了组合多个引导向量的常用方法，发现所有方法在添加更多向量时特征表达性显著下降，且在连贯性和可表达性之间存在权衡，需要针对每个设置进行超参数调优。综合来看，我们的结果表明，引导向量作为一种通用的偏好对齐工具面临着重要的局限性。
cs.CL / 28 / 2607.01883
PairCoder++: Pair Programming as a Universal Paradigm for Verified Code-Driven Multimodal and Structured-Artifact Generation
PairCoder++：作为验证代码驱动的多模态和结构化工件生成的通用范式的配对编程
Abstract
Code is the medium through which large language models generate structured artifacts: charts, scientific figures, vector graphics, CAD models, 3D scenes, and hardware designs are all produced by writing programs. In this regime single pass inference is brittle, because the compiler, renderer, or simulator that decides whether the artifact exists is invisible to the model. We present PairCoder, which grounds review in the toolchain and realizes it as two agent pair programming: a Driver agent writes the program, a Navigator agent reviews it against verification evidence (diagnostics, execution results, and renderings of the current artifact beside the target), and the two switch roles when errors persist. Across 17 public benchmarks and seven models from three vendors, PairCoder improves essentially every benchmark whose artifact is verifiable, on full official metric suites rather than execution alone (for example, Blender scene executability 0.20 to 0.78; TikZ compile rate up 10 to 30 points on every model), at 2.9 to 9.2 times single model cost (about 7 times overall). The improvements concentrate where the toolchain provides an informative oracle and the baseline leaves headroom, and the method ties or mildly regresses where the oracle is weak; we frame pair programming as a reliable recipe for verified code driven generation.
Chinese Translation
代码是大型语言模型生成结构化工件的媒介：图表、科学图形、矢量图形、CAD模型、3D场景和硬件设计都是通过编写程序来生成的。在这种情况下，单次推理是脆弱的，因为决定工件是否存在的编译器、渲染器或模拟器对模型是不可见的。我们提出了PairCoder，它将审查与工具链相结合，并实现为两个代理的配对编程：驱动代理（Driver agent）编写程序，导航代理（Navigator agent）根据验证证据（诊断、执行结果和当前工件与目标的渲染）对其进行审查，当错误持续存在时，两者交换角色。在17个公共基准测试和来自三家供应商的七个模型中，PairCoder在几乎所有可验证工件的基准测试中都有所改善，使用的是完整的官方指标套件，而不仅仅是执行（例如，Blender场景的可执行性从0.20提高到0.78；TikZ编译率在每个模型上提高了10到30个百分点），其成本为单模型的2.9到9.2倍（整体约为7倍）。这些改进集中在工具链提供信息性神谕的地方，而基线则留有余地；当神谕较弱时，该方法的表现持平或略微退步；我们将配对编程框架视为一种可靠的验证代码驱动生成的方案。
cs.CL / 29 / 2607.01899
The Grammar Does the Work: Functional vs. Lexical Dependency Length Minimization Across Universal Dependencies
语法发挥作用：跨通用依赖的功能性与词汇性依赖长度最小化
Abstract
Dependency length minimization (DLM) is a well-documented processing universal, but previous studies report a single mean dependency distance (MDD) per language, obscuring variation across syntactic relation types. We analyze 122 languages in UD and SUD (version 2.17), showing that DLM operates on two distinct levels. Grammar-driven optimization targets functional dependencies (det, case, aux), which are universally short (mean 1.71, $\sigma$ = 0.33) and invariant across typologically diverse languages. Processing-driven optimization operates on lexical dependencies (nsubj, obj, obl), which are longer (mean 2.87), highly variable ($\sigma$ = 0.63), and constrained by word-order typology. This asymmetry holds in SUD despite reversed head direction (r = 0.92). We conclude that ''the grammar does the work'' of minimization by scaffolding sentences with local functional attachments, leaving processing pressures to determine the ordering of lexical heads.
Chinese Translation
依赖长度最小化（DLM）是一种被广泛记录的处理普遍现象，但以往研究报告每种语言只有一个平均依赖距离（MDD），掩盖了不同句法关系类型之间的变异。我们分析了122种语言中的通用依赖（UD）和句法依赖（SUD，版本2.17），显示DLM在两个不同层面上运作。语法驱动的优化针对功能性依赖（det, case, aux），这些依赖在各类语言中普遍较短（平均1.71，$ ext{σ}$ = 0.33），且在类型上具有不变性。处理驱动的优化则作用于词汇依赖（nsubj, obj, obl），这些依赖较长（平均2.87），变化性高（$ ext{σ}$ = 0.63），并受限于词序类型。这种不对称性在SUD中依然成立，尽管头部方向相反（r = 0.92）。我们得出结论：“语法发挥了最小化的作用”，通过用局部功能性附加结构支撑句子，留下处理压力来决定词汇头的顺序。
cs.CL / 30 / 2607.01927
TUDUM: A Turkish-Thinking Reasoning Pipeline for Qwen3.5-27B
TUDUM：一个面向土耳其推理的 Qwen3.5-27B 思维模型管道
Abstract
This paper presents TUDUM (T\"urk\c{c}e D\"u\c{s}\"unen \"Uretken Model), a project pipeline for adapting a Qwen-family 27B thinking model toward Turkish reasoning. The central problem is not only to answer Turkish prompts in Turkish, but to make the explicit reasoning trace itself Turkish. A thinking model may translate a Turkish prompt into an English-centered internal or visible scratchpad, solve the problem mostly in English, and only localize the final answer. TUDUM instead treats the generated ... block as a trainable behavior. The pipeline starts from the project base checkpoint unsloth/Qwen3.5-27B, applies supervised fine-tuning (SFT) on 15,991 Turkish reasoning examples using LoRA adapters, and then applies GRPO-family reinforcement learning on a proxy-filtered Turkish mathematics environment. The results are mixed. SFT made the model shorter and more consistently Turkish in its reasoning behavior, with large reductions in average response length and thinking exhaustion, but reduced benchmark accuracy. RL recovered some mathematical performance, especially AIME24 at the best early checkpoint, yet did not uniformly improve all benchmarks and did not exceed the base model on the reported Macro-6 average. The contribution is therefore best framed as a technically honest Turkish-thinking reasoning pipeline and evaluation, not as a claim of state-of-the-art Turkish reasoning. The released step-50 model is publicly available.
Chinese Translation
本文介绍了 TUDUM（T"urk ext{c}e D"u ext{s}"unen "Uretken Model），这是一个将 Qwen 系列 27B 思维模型适应于土耳其推理的项目管道。核心问题不仅在于用土耳其语回答土耳其语提示，还在于使显式推理过程本身也成为土耳其语。一个思维模型可能将土耳其语提示翻译成以英语为中心的内部或可见草稿，主要用英语解决问题，并仅对最终答案进行本地化。而 TUDUM 则将生成的 ... 块视为可训练的行为。该管道从项目基础检查点 unsloth/Qwen3.5-27B 开始，使用 LoRA 适配器对 15,991 个土耳其推理示例进行监督微调（SFT），然后在一个代理过滤的土耳其数学环境中应用 GRPO 系列的强化学习。结果呈现出混合的效果。SFT 使模型的推理行为更短且更一致地呈现土耳其语，平均响应长度和思维疲惫度大幅降低，但基准准确率下降。强化学习在某些数学性能上有所恢复，尤其是在最佳早期检查点的 AIME24，但并未均匀改善所有基准，并且在报告的 Macro-6 平均值上未超过基础模型。因此，该贡献更适合被框定为一个技术上诚实的土耳其思维推理管道和评估，而非声称为最先进的土耳其推理。发布的 step-50 模型已公开可用。
cs.CL / 31 / 2607.01934
AIriskEval-edu: New Dataset for Risk Assessment in AI-mediated K-12 Educational Explanations
AIriskEval-edu：用于AI介导的K-12教育解释风险评估的新数据集
Abstract
This work introduces AIriskEval-edu-db2, a new dataset designed to train and evaluate auditors based on LLMs for an explainable pedagogical risk assessment in instructional content for grades K-12. The dataset comprises 1,639 explanations from 170 curated ScienceQA questions, covering science, language arts, and social sciences. For each question, the dataset includes an explanation written by a human teacher alongside 11 explanations generated by LLM-simulated teacher profiles associated with distinct pedagogical risks. We propose a comprehensive risk rubric aligned with established educational standards that covers five complementary dimensions: factual precision, depth and completeness, focus and relevance, student-level appropriateness, and ideological bias. A key contribution is the addition of 785 explanations with structured explainability annotations, including risk localization and risk description. The annotations are produced through a semi-automatic process with expert teacher validation. Finally, we present validation experiments comparing state-of-the-art proprietary models with a lightweight local Llama 3.1 8B model in both the pedagogical risk detection and the explainability assessment. These experiments evaluate whether supervised fine-tuning on AIriskEval-edu-db2 enables a locally deployable model to approach or outperform stronger frontier models while preserving privacy in educational auditing and assessment tasks.
Chinese Translation
本研究介绍了AIriskEval-edu-db2，这是一个旨在基于大型语言模型（LLMs）训练和评估审计员的新数据集，用于K-12年级教学内容的可解释性教育风险评估。该数据集包含来自170个精心挑选的ScienceQA问题的1,639个解释，涵盖科学、语言艺术和社会科学。对于每个问题，数据集包括一位人类教师撰写的解释，以及11个由与不同教育风险相关的LLM模拟教师档案生成的解释。我们提出了一种与既定教育标准相一致的综合风险评估标准，涵盖五个互补维度：事实准确性、深度与完整性、关注与相关性、学生水平的适宜性，以及意识形态偏见。一个重要的贡献是增加了785个带有结构化可解释性注释的解释，包括风险定位和风险描述。这些注释通过半自动化过程生成，并经过专家教师的验证。最后，我们展示了验证实验，比较了最先进的专有模型与轻量级本地Llama 3.1 8B模型在教育风险检测和可解释性评估中的表现。这些实验评估了在AIriskEval-edu-db2上进行监督微调是否使得本地可部署模型能够接近或超越更强的前沿模型，同时在教育审计和评估任务中保持隐私。
cs.CL / 32 / 2607.01960
NAVER LABS Europe Submission to the Instruction-following 2026 Short Track
NAVER LABS 欧洲在2026年指令跟随短赛中的提交
Abstract
In this paper, we describe NAVER LABS Europe's submission to the instruction-following speech processing short track at IWSLT 2026. We participate again in the constrained setting, developing systems capable of jointly performing ASR, ST, and SQA from English speech into Chinese, Italian, and German. Building on our previous submission, ranked first in last year's short track, we update our multi-stage training pipeline by replacing the speech projector with SpeechMapper, a method for learning a speech-to-LLM embedding projector using only ASR data. In addition, we introduce a synthetic SQA dataset, fakACL, composed of artificially generated scientific presentations. This dataset is built by prompting the LLM backbone, segmenting the generated talks, and synthesizing speech with SeamlessM4T-large-v2. The combination of an improved speech projection mechanism and domain-specific synthetic data allows our model to outperform last year's best short-track system, while being considerably more compact and relying on a weaker LLM backbone. This year's results place our system tied for first place in the overall short track ranking.
Chinese Translation
本文描述了NAVER LABS 欧洲在IWSLT 2026年指令跟随语音处理短赛中的提交。我们再次参与受限设置，开发能够从英语语音联合执行自动语音识别（ASR）、语音翻译（ST）和科学问答（SQA）到中文、意大利语和德语的系统。在去年的短赛中，我们的提交排名第一，基于此，我们更新了多阶段训练流程，用SpeechMapper替代了语音投影器，SpeechMapper是一种仅使用ASR数据学习语音到大语言模型（LLM）嵌入投影器的方法。此外，我们引入了一个合成的SQA数据集fakACL，该数据集由人工生成的科学演示组成。该数据集通过提示LLM主干、对生成的演讲进行分段，并使用SeamlessM4T-large-v2合成语音来构建。改进的语音投影机制与特定领域的合成数据相结合，使我们的模型在性能上超越了去年的最佳短赛系统，同时显著更为紧凑，并依赖于较弱的LLM主干。今年的结果使我们的系统在整体短赛排名中并列第一。
cs.CL / 33 / 2607.01964
Beyond Supervised Clarification: Input Rewriting with LLMs for Dialogue Discourse Parsing
超越监督澄清：利用大型语言模型进行对话语篇解析的输入重写
Abstract
Rewriting inputs to improve frozen downstream models has become a common strategy in modern NLP pipelines. Prior work on incremental dialogue discourse parsing (DDP) shows that supervised clarification models can rewrite fragmentary or underspecified utterances, such as resolving ellipsis or references, to improve parsing accuracy. In this work, we revisit this idea under realistic deployment conditions, where no clarification supervision is available and the clarifier must rely on zero-shot prompting or feedback from a frozen parser. Across three Segmented Discourse Representation Theory (SDRT) datasets and multiple parsers, we find that last-utterance clarification is far less reliable than suggested by supervised settings. Parser-agnostic rewriting often introduces more regressions than repairs, as edits that enable fixes also disrupt discourse cues relied upon by the parser. A best-of-8 rewriting analysis further reveals a practical ceiling: a large fraction of errors are not repairable through input rewriting alone. A parser-aware clarifier trained with GRPO reduces regressions by up to 37% by learning conservative abstention, yet still fails to produce selectivity-aware clarifications that consistently improve parsing. Together, these findings recast clarification as a selective intervention problem. We identify rewritability prediction, deciding whether an utterance is repairable before intervention, as the key missing capability for input-side optimization of frozen discourse parsers, and a critical direction for improving agentic pipelines more broadly.
Chinese Translation
重写输入以改善冻结的下游模型已成为现代自然语言处理（NLP）管道中的一种常见策略。先前关于增量对话语篇解析（DDP）的研究表明，监督澄清模型可以重写片段化或不明确的发言，例如解决省略或指代问题，从而提高解析准确性。在本研究中，我们在现实部署条件下重新审视这一想法，在这种情况下没有澄清监督，澄清者必须依赖零-shot 提示或来自冻结解析器的反馈。在三个分段语篇表示理论（SDRT）数据集和多个解析器的实验中，我们发现最后发言的澄清远不如监督设置所建议的那样可靠。解析器无关的重写往往引入的回归比修复更多，因为使修复成为可能的编辑也会干扰解析器所依赖的语篇线索。对8种重写方案的最佳分析进一步揭示了一个实际的上限：很大一部分错误无法仅通过输入重写来修复。通过学习保守的放弃策略，使用GRPO训练的解析器感知澄清者将回归减少了多达37%，但仍未能产生能够持续改善解析的选择性意识澄清。综合来看，这些发现将澄清重新定义为一个选择性干预问题。我们确定重写可预测性，即在干预之前判断发言是否可修复，作为冻结语篇解析器输入侧优化的关键缺失能力，以及更广泛改善代理管道的关键方向。
cs.CL / 34 / 2607.01965
Towards a Phonology-Informed Evaluation of Multilingual TTS
朝着一个以音位学为基础的多语言文本到语音评估
Abstract
Neural TTS systems can sound natural across languages, but naturalness does not guarantee the preservation of sound contrasts that distinguish words from their grammatical forms. Standard metrics like MOS do not test for this. We propose a classifier-based framework that audits TTS output against language-specific phonological patterns using human speech as a benchmark. Testing Assamese advanced tongue root (ATR) vowel harmony with Meta's MMS TTS, we show that a classifier trained on human speech transfers to synthesized speech with minimal loss. The faithfulness audit reveals that [+ATR] mid vowels are realized as [-ATR] in 1/3 tokens despite an underlying [+ATR] specification, a bias absent in human speech. At the word level, predicted ATR labels classify harmony more accurately than transcription labels, indicating a gap between intended and produced phonology. The framework offers task-specific diagnostics and generalizes to other phonological contrasts with measurable acoustic cues.
Chinese Translation
神经文本到语音（TTS）系统在不同语言中可以听起来很自然，但自然性并不保证能够保留区分单词及其语法形式的音位对比。标准指标如均值意见分数（MOS）并未对此进行测试。我们提出了一种基于分类器的框架，该框架使用人类语音作为基准，审计TTS输出与特定语言的音位学模式之间的关系。通过使用Meta的MMS TTS测试阿萨姆语的前舌根（ATR）元音和谐，我们展示了一个在人工语音上训练的分类器能够以最小的损失转移到合成语音上。忠实性审计显示，尽管存在潜在的[+ATR]规范，1/3的标记中[+ATR]中元音却被实现为[-ATR]，这一偏差在人工语音中并不存在。在单词层面，预测的ATR标签比转录标签更准确地分类和谐，表明意图与实际产生的音位学之间存在差距。该框架提供了特定任务的诊断，并且可以推广到其他具有可测量声学线索的音位对比。
cs.CL / 35 / 2607.01972
Object Aligner: A Configurable JSON Schema Similarity Score for Graphs, Applied to LLM Prompt Optimization
对象对齐器：一种可配置的图形 JSON 架构相似度评分，应用于大语言模型提示优化
Abstract
Large language models (LLMs) are often asked to produce JSON conforming to a fixed schema, powering information extraction, tool calling, agentic planning, and knowledge-graph construction. Measuring how closely an output matches a gold reference is essential yet surprisingly hard: exact match is brittle, text similarity ignores structure, and an LLM judge is expensive, opaque, and non-deterministic. We address this with Object Aligner (OA), an open-source Python library that scores two JSON objects deterministically by recursively aligning their trees (the Hungarian algorithm for unordered collections, sequence alignment for ordered ones) and awarding partial credit at the granularity the schema declares. The Object Aligner is configured entirely through a set of JSON Schema extensions, so adapting it to a new task involves annotating a schema rather than writing code. Complex structured data, however, are rarely flat trees: records may form graphs or hypergraphs keyed by arbitrary identifiers, breaking the assumptions of prior similarity metrics. Our central contribution, referential alignment, closes this gap by inferring a bijection between gold and candidate identifiers and scoring every reference through it, so the score is invariant to relabeling. Since recovering this bijection exactly is graph isomorphism, the Object Aligner approximates it with Weisfeiler-Leman color refinement. An order-sensitive sequence regime targets ranking and planning. Since the same alignment localizes every mismatch, the Object Aligner emits ranked repair suggestions at no extra cost. Used as a reward inside the GEPA prompt optimizer, Object Aligner helps or stays neutral across all datasets.
Chinese Translation
大语言模型（LLMs）通常被要求生成符合固定架构的 JSON，以支持信息提取、工具调用、代理规划和知识图谱构建。衡量输出与黄金参考的匹配程度至关重要，但却出乎意料地困难：精确匹配脆弱，文本相似度忽略结构，而 LLM 评判者则昂贵、不透明且非确定性。我们通过对象对齐器（Object Aligner, OA）解决了这个问题，OA 是一个开源的 Python 库，通过递归对齐两个 JSON 对象的树（对无序集合使用匈牙利算法，对有序集合使用序列对齐）并根据架构声明的粒度授予部分信用，从而确定性地评分。对象对齐器完全通过一组 JSON 架构扩展进行配置，因此将其适应新任务涉及对架构进行注释而不是编写代码。然而，复杂的结构化数据很少是平坦的树：记录可能形成由任意标识符键控的图或超图，这打破了先前相似度度量的假设。我们的核心贡献是参考对齐，通过推断黄金和候选标识符之间的双射并通过它对每个引用进行评分，从而填补了这一空白，因此评分对重新标记是不变的。由于准确恢复这一双射是图同构，对象对齐器使用 Weisfeiler-Leman 颜色细化来近似它。一个对顺序敏感的序列机制针对排名和规划。由于相同的对齐定位每个不匹配，对象对齐器以零额外成本发出排名修复建议。作为 GEPA 提示优化器中的奖励使用，对象对齐器在所有数据集上都能提供帮助或保持中立。
cs.CL / 36 / 2607.02002
Using embeddings to predict spoken word duration and pitch in Mandarin monosyllabic words
利用嵌入向量预测普通话单音节词的发音时长和音高
Abstract
Time-normalized f0 contours of Mandarin words in conversational speech have been shown to be predictable in part from their contextualized embeddings (CEs). The present study investigates whether CEs also predict spoken word duration for 7470 tokens of Mandarin monosyllabic CV words extracted from a Mandarin corpus of spontaneous speech. We show that CEs indeed are predictive for duration, above chance level, not only at the type level, but also at the level of individual tokens, as indicated by the results obtained with the type-wise and token-wise permutation baselines. We also show that the predicted durations are sufficiently precise to back-transform predicted f0 contours in [0,1] normalized time to contours on the ms time scale. The resulting predicted contours approximate empirical contours and also outperform a permutation baseline.
Chinese Translation
在对话语音中，普通话词汇的时间归一化基频（f0）轮廓已被证明可以部分通过其上下文化嵌入向量（CEs）进行预测。本研究探讨了CEs是否也能预测从普通话自发语料库中提取的7470个普通话单音节CV词的发音时长。我们的研究表明，CEs确实能够在超出随机水平的情况下预测发音时长，不仅在词汇类型层面上有效，在单个词元层面上同样有效，这一点通过类型和词元的置换基线结果得到了验证。我们还表明，预测的时长足够精确，可以将预测的f0轮廓从[0,1]归一化时间反变换为毫秒时间尺度上的轮廓。最终得到的预测轮廓与经验轮廓相近，并且优于置换基线。
cs.CL / 37 / 2607.02007
EduArt: An educational-level benchmark for evaluating art history knowledge in large language models
EduArt：评估大型语言模型艺术史知识的教育级基准
Abstract
Large language models now score near ceiling on general benchmarks, but these aggregate measures reveal little about how models behave within single disciplines. Existing art-focused evaluations rely on synthetic questions and rarely report item-level properties. This paper introduces EduArt, an educational-level benchmark for art-historical knowledge and visual reasoning in multimodal LLMs. EduArt comprises 871 human-authored questions from Italian secondary-school exercises and US Advanced Placement Art History exams, spanning two languages and seven formats from multiple choice to in-text word placement and error identification. Twelve models from six provider families were evaluated under a default answer-only condition and a motivation condition requiring written justification, and characterized using Classical Test Theory and a logistic regression isolating the effects of format, language, image presence, and model. The benchmark showed strong psychometric properties (mean discrimination 0.514, 82.3 percent good discriminators), while multiple-choice accuracy saturated near ceiling for six models, showing recognition formats alone cannot distinguish frontier models. Format was a strong independent predictor of accuracy: models exceeding 94 percent on multiple choice fell to 23.9 percent on open completion (Claude Opus 4.6) and 6.2 percent on error identification (Claude Sonnet 4.6). The motivation condition changed accuracy in a predominantly negative, family-dependent direction. These dissociations indicate that art-historical knowledge and the ability to deploy it are distinct capabilities, and that single-format benchmarks overestimate what models can reliably do. Mapping this capability profile is a precondition for responsible use of multimodal LLMs in art-historical scholarship, where tasks demand producing and manipulating content rather than selecting from fixed options.
Chinese Translation
大型语言模型在一般基准测试中得分接近上限，但这些汇总指标对模型在单一学科内的表现揭示甚少。现有的艺术相关评估依赖于合成问题，且很少报告项目级属性。本文介绍了EduArt，这是一个针对艺术历史知识和视觉推理的教育级基准，适用于多模态大型语言模型（LLMs）。EduArt包含871个由人类创作的问题，来源于意大利中学练习和美国高级课程艺术历史考试，涵盖两种语言和七种格式，从选择题到文本内单词填空和错误识别。评估了来自六个提供者家族的十二个模型，采用默认的仅答题条件和要求书面理由的动机条件，并使用经典测验理论和逻辑回归分析格式、语言、图像存在和模型的影响。基准显示出强大的心理测量特性（平均区分度0.514，82.3%的良好区分者），而六个模型的选择题准确率接近上限，表明仅凭识别格式无法区分前沿模型。格式是准确率的强独立预测因子：在选择题中超过94%的模型在开放式完成（Claude Opus 4.6）中降至23.9%，在错误识别（Claude Sonnet 4.6）中降至6.2%。动机条件在主要呈负向、依赖于模型家族的方向上改变了准确率。这些分离表明艺术历史知识及其运用能力是不同的能力，且单一格式的基准高估了模型可以可靠执行的任务。绘制这一能力轮廓是负责任地在艺术历史学术中使用多模态大型语言模型的前提，因为这些任务要求生成和操控内容，而非从固定选项中选择。
cs.CL / 38 / 2607.02047
OpenSafeIntent: Evaluating Intent-Calibrated Safe Completion Across Dual-Use Prompt Sets
OpenSafeIntent：评估跨双重用途提示集的意图校准安全完成
Abstract
Safe completion requires models to provide useful assistance without enabling harm, but this behavior is difficult to evaluate with isolated prompts. We introduce OpenSafeIntent, a benchmark of controlled prompt-sets that vary intent while holding the underlying task fixed. Each datapoint contains benign, dual-use, and malicious variants of the same task. This design lets us evaluate whether models calibrate assistance across intent shifts, rather than merely appearing safe on average. Across a broad model suite, we find that prompt-level safety hides important failures: models often fail to remain safe across matched intent variants, dual-use behavior is brittle under paraphrase, high-level answers on risky topics are not reliably safe, and responses that reframe ambiguous requests into safer tasks are substantially less likely to cross the safety boundary. Our results suggest that safe completion should be evaluated as intent-calibrated behavior over controlled task variants, not as a single safety-helpfulness tradeoff over independent prompts.
Chinese Translation
安全完成要求模型在提供有用帮助的同时不导致伤害，但这种行为在孤立提示下难以评估。我们引入了OpenSafeIntent，一个控制提示集的基准，这些提示集在保持基础任务不变的情况下变化意图。每个数据点包含相同任务的良性、双重用途和恶意变体。这种设计使我们能够评估模型在意图转变中是否能够校准帮助，而不仅仅是在平均水平上看起来安全。在广泛的模型套件中，我们发现提示级别的安全性掩盖了重要的失败：模型在匹配意图变体之间往往无法保持安全，双重用途行为在改述时表现脆弱，关于风险话题的高层次回答并不可靠安全，以及将模糊请求重新框定为更安全任务的响应显著不太可能越过安全边界。我们的结果表明，安全完成应作为在受控任务变体上的意图校准行为进行评估，而不是作为独立提示上的单一安全性与有用性的权衡。
cs.CL / 39 / 2607.02049
SPLIT: Cross-Lingual Empathy and Cultural Grounding in English and Ukrainian LLM Responses
SPLIT：英语和乌克兰语大语言模型响应中的跨语言共情与文化基础
Abstract
Large Language Models are increasingly deployed in emotional-support contexts and crisis-related situations. Nevertheless, their cross-lingual abilities in these circumstances remain underexplored. Existing benchmarks emphasize multilingual performance but rarely examine crisis-related empathy and cultural grounding in low-to-mid-resource languages. We introduce SPLIT, a 500-prompt benchmark designed to evaluate LLM consistency in generating emotionally grounded responses across five categories: Stress, Panic, Loneliness, Internal Displacement, and Tension. We evaluate three technically diverse LLMs across three dimensions: Empathetic Accuracy, Linguistic Naturalness, and Contextual & Cultural Grounding. The framework aims to assess and compare the quality of LLM responses in both English and Ukrainian languages, as well as to explore the reliability of the LLM-as-a-jury paradigm. Our findings reveal that Gemini-2.5-Flash and LLaMA-3.3-70B-Instruct degrade when transitioning to Ukrainian, while DeepSeek-V3 remains comparatively stable within our benchmark. We additionally find that human and AI evaluators agree weakly on empathy and naturalness but diverge on cultural grounding. We further argue that producing Ukrainian text is not equivalent to producing Ukrainian emotional support. Our findings may assist in the future development of more culturally tailored benchmark designs, as well as encourage a stronger emphasis on human-centered evaluation.
Chinese Translation
大型语言模型越来越多地应用于情感支持和危机相关情境中。然而，在这些情况下，它们的跨语言能力仍然未得到充分探索。现有基准强调多语言性能，但很少考察低到中资源语言中的危机相关共情和文化基础。我们提出了SPLIT，一个包含500个提示的基准，旨在评估大语言模型在五个类别（压力、恐慌、孤独、内部流离失所和紧张）中生成情感基础响应的一致性。我们在三个维度（共情准确性、语言自然性和上下文与文化基础）上评估了三种技术多样的大语言模型。该框架旨在评估和比较大语言模型在英语和乌克兰语中的响应质量，并探索大语言模型作为陪审团范式的可靠性。我们的研究发现，Gemini-2.5-Flash和LLaMA-3.3-70B-Instruct在转向乌克兰语时性能下降，而DeepSeek-V3在我们的基准中相对稳定。我们还发现，人类和人工智能评估者在共情和自然性上意见不一，但在文化基础上存在分歧。我们进一步认为，生成乌克兰语文本并不等同于提供乌克兰情感支持。我们的研究结果可能有助于未来开发更具文化针对性的基准设计，并鼓励更加强调以人为本的评估。
cs.CL / 40 / 2607.02079
HaloGuard 1.0: An Open Weights Constitutional Classifier for Multilingual AI Safety
HaloGuard 1.0：一种用于多语言人工智能安全的开放权重宪法分类器
Abstract
We present HaloGuard 1.0, an open-weights implementation of the constitutional-classifier paradigm for input safety. It achieves state-of-the-art performance on English and multilingual prompt-safety benchmarks at roughly one-tenth the model size of current leading open guard models. The safety constitution is the organising structure of the corpus: a natural-language constitution of 46 policies and 2,940 subcategories drives synthetic data generation, with exhaustive one-to-one paired counterfactuals that hold topic and vocabulary fixed while flipping intent, a two-tier harmless design that separately targets boundary and baseline false positives (FPs), and balanced multilingual materialisation across 46 languages that treats language as a surface form appearing on both sides of the boundary rather than as an adversarial signal. Across seven prompt-safety benchmarks, HaloGuard 1.0-0.8B attains the best average F1 (90.9) of any open guard we evaluate, outperforming baselines up to 27B parameters (over 30 times larger) while holding false-positive rate (FPR) to 4.3 and false-negative rate (FNR) to 9.5. The HaloGuard 1.0-4B variant reaches average F1 of 92.1 and FPR of 3.5, spending its extra capacity on precision rather than recall. A structured adjudication of the remaining failures indicates that most apparent missed-harm cases are benchmark mislabels rather than genuine model misses. An always-on adversarial red-teaming protocol continuously hardens the guard against both content-level and agentic attacks. We release the models as open weights.
Chinese Translation
我们提出了HaloGuard 1.0，这是一种开放权重的宪法分类器范式实现，旨在确保输入安全。它在英语和多语言提示安全基准测试中实现了最先进的性能，其模型规模约为当前领先开放防护模型的十分之一。安全宪法是语料库的组织结构：由46项政策和2,940个子类别构成的自然语言宪法驱动合成数据生成，采用详尽的一对一配对反事实方法，在保持主题和词汇不变的同时翻转意图，采用两级无害设计，分别针对边界和基线假阳性（FPs），并在46种语言中实现平衡的多语言表现，将语言视为出现在边界两侧的表面形式，而非对抗信号。在七个提示安全基准测试中，HaloGuard 1.0-0.8B达到了我们评估的任何开放防护模型中最佳的平均F1值（90.9），超越了高达27B参数（超过30倍于其规模）的基线，同时将假阳性率（FPR）控制在4.3，假阴性率（FNR）控制在9.5。HaloGuard 1.0-4B变体的平均F1值达到92.1，FPR为3.5，其额外的容量用于提高精确度而非召回率。对剩余失败案例的结构化审查表明，大多数明显的漏伤害案例是基准错误标记，而非真正的模型漏判。始终开启的对抗红队协议持续增强防护能力，以抵御内容级和代理攻击。我们将这些模型以开放权重的形式发布。
cs.CL / 41 / 2607.02214
Unlocking Speech-Text Compositional Powers: Instruction-Following Speech Language Models without Instruction Tuning
解锁语音-文本组合能力：无需指令调优的遵循指令的语音语言模型
Abstract
Instruction tuning for speech language models (SLMs) is substantially more challenging than for text-based large language models (LLMs), as it requires learning a new modality and a wide range of speech-specific instructions in addition to those supported by text LLMs. Existing SLM training approaches largely replicate the text LLM training paradigm by synthesizing large-scale speech pre-training and instruction-tuning datasets. However, this strategy is difficult to scale, since speech sequences are significantly longer than text sequences. In this paper, we propose SpeechCombine, an instruction-following speech language model trained without any instruction tuning, using only a single round of speech pre-training on 30k hours of data. Starting from a text LLM base model, we perform continuous pre-training on speech utterances to obtain a speech-adapted model, and then directly combine its weights with the weight difference between the instruction-tuned and base versions of the text LLM. Our results show that this simple combination strategy not only preserves the knowledge and capabilities of the original text LLM, but also effectively transfers them to the speech domain. These findings suggest a new direction for SLM training that avoids reliance on massive speech data.
Chinese Translation
语音语言模型（SLMs）的指令调优比文本基础的大型语言模型（LLMs）要困难得多，因为它不仅需要学习一种新的模态，还需要掌握一系列特定于语音的指令，而这些指令在文本 LLMs 中是支持的。现有的 SLM 训练方法在很大程度上复制了文本 LLM 训练范式，通过合成大规模的语音预训练和指令调优数据集。然而，由于语音序列显著长于文本序列，这一策略难以扩展。在本文中，我们提出了 SpeechCombine，这是一种遵循指令的语音语言模型，采用仅一次语音预训练（使用 30,000 小时的数据）而无需任何指令调优。我们从一个文本 LLM 基础模型开始，对语音话语进行连续预训练，以获得一个适应语音的模型，然后直接将其权重与指令调优版本和基础版本文本 LLM 之间的权重差结合。我们的结果表明，这一简单的组合策略不仅保留了原始文本 LLM 的知识和能力，而且有效地将其转移到语音领域。这些发现为 SLM 训练提供了一种新的方向，避免了对大量语音数据的依赖。
cs.CL / 42 / 2607.02235
Challenges and Recommendations for LLMs-as-a-Judge in Multilingual Settings and Low-Resource Languages
多语言环境和低资源语言中 LLM 作为裁判的挑战与建议
Abstract
LLM-as-a-Judge has become the dominant evaluation paradigm for many natural language generation tasks, due to shortcomings of conventional metrics and high correlations with human judgment, albeit mostly in English. There are now attempts to extend LLM-as-a-Judge to multilingual settings including low-resource languages. However, LLMs have limited proficiency in low-resource languages, and there is often no adequate human validation in these settings. To highlight the scope of the problem and current practices, we explore the use of LLM-as-a-Judge evaluators in ACL Anthology papers focusing on multilingual settings and low-resource languages across a diverse set of tasks. Out of 650 papers mentioning LLM-as-a-judge, only 33 of them focus on low-resource or multilingual settings. Our in-depth analysis of these papers indicates inconsistent evaluation outcomes, a tendency to overtrust LLM judgments in multilingual settings, and the widespread reliance on a single judge model per study. To help the NLP community further, we conclude with recommendations about how to use LLM-as-a-Judge in multilingual and low-resource settings.
Chinese Translation
LLM 作为裁判（LLM-as-a-Judge）已成为许多自然语言生成任务的主导评估范式，主要是由于传统指标的不足以及与人类判断的高度相关性，尽管这种相关性主要体现在英语中。目前，已经有尝试将 LLM 作为裁判扩展到包括低资源语言在内的多语言环境。然而，LLM 在低资源语言中的能力有限，并且在这些环境中通常缺乏足够的人类验证。为了突出问题的范围和当前的实践，我们探讨了在 ACL 论文集中使用 LLM 作为裁判的评估者，重点关注多语言环境和低资源语言的多样化任务。在提到 LLM 作为裁判的 650 篇论文中，仅有 33 篇集中于低资源或多语言环境。我们对这些论文的深入分析表明，评估结果不一致，在多语言环境中对 LLM 判断的过度信任倾向，以及每项研究普遍依赖单一裁判模型。为了进一步帮助自然语言处理（NLP）社区，我们最后提出了关于如何在多语言和低资源环境中使用 LLM 作为裁判的建议。
cs.CL / 43 / 2607.02259
BamiBERT: A New BERT-based Language Model for Vietnamese
BamiBERT：一种新的基于BERT的越南语语言模型
Abstract
In this paper, we introduce BamiBERT, a new BERT-based pre-trained language model for Vietnamese that addresses key limitations of PhoBERT -- the current de facto Vietnamese text encoder. Trained from scratch on a 129GB corpus of general-domain Vietnamese text for 20 epochs, BamiBERT supports an extended context length of up to 2048 tokens and operates directly on raw input, eliminating the need for external word segmentation. Across 8 Vietnamese benchmarks, it achieves the best score on 11 of 15 metrics and the second-best on 3 others, setting a new state of the art among "base"-sized Vietnamese encoders and demonstrating strong cross-domain generalization. We release BamiBERT at: https://huggingface.co/Qualcomm-AI-Research/BamiBERT
Chinese Translation
在本文中，我们介绍了BamiBERT，一种新的基于BERT的预训练越南语语言模型，旨在解决当前事实上的越南语文本编码器PhoBERT的关键局限性。BamiBERT在一个129GB的通用领域越南语文本语料库上从头开始训练20个周期，支持最长可达2048个标记的扩展上下文长度，并直接处理原始输入，消除了对外部分词的需求。在8个越南语基准测试中，它在15个指标中的11个上取得了最佳成绩，在其他3个上获得了第二好成绩，树立了“基础”规模越南语编码器的新技术水平，并展示了强大的跨领域泛化能力。我们在以下网址发布BamiBERT：https://huggingface.co/Qualcomm-AI-Research/BamiBERT
cs.CL / 44 / 2607.02262
CheckRLM: Effective Knowledge-Thought Coherence Checking in Retrieval-Augmented Reasoning
CheckRLM：检索增强推理中的有效知识-思维一致性检查
Abstract
Reasoning Language Models (RLMs) have significantly improved performance on complex tasks by extending the reasoning chain. However, these chains are prone to containing factual errors, particularly in knowledge-intensive tasks. To address this issue, we propose CheckRLM, a framework that improves the reliability of the reasoning process through Retrieval-Augmented Generation (RAG) by timely checking and correcting factual errors. Specifically, CheckRLM extracts factual claims from the reasoning chain to identify and localize subtle knowledge inconsistencies during inference. Upon detection of errors, a refinement mechanism performs minimal-cost yet precise corrections by leveraging external knowledge, ensuring coherence between the reasoning chain and correct knowledge. Extensive experiments demonstrate that CheckRLM substantially outperforms existing baselines, exhibiting a strong capability to mitigate error accumulation in long-horizon reasoning with lower costs. The code and data are available at https://github.com/AI9Stars/CheckRLM.
Chinese Translation
推理语言模型（RLMs）通过扩展推理链在复杂任务上的表现显著提升。然而，这些推理链容易包含事实错误，尤其是在知识密集型任务中。为了解决这一问题，我们提出了CheckRLM，一个通过检索增强生成（RAG）及时检查和纠正事实错误，从而提高推理过程可靠性的框架。具体而言，CheckRLM从推理链中提取事实声明，以识别和定位推理过程中的细微知识不一致。在检测到错误后，精细化机制通过利用外部知识进行最低成本但精确的纠正，确保推理链与正确知识之间的一致性。大量实验表明，CheckRLM在降低成本的同时，显著优于现有基准，展现出在长时间推理中减轻错误积累的强大能力。代码和数据可在 https://github.com/AI9Stars/CheckRLM 获取。
cs.CL / 45 / 2607.02307
On the Role of Directionality in Structural Generalization
结构泛化中方向性的作用
Abstract
Several SLOG test categories explicitly involve directional distinctions (modifier position shifts, argument extraction positions), yet AM-Parser, the previous SOTA, uses an AM algebra whose operations do not encode direction. We redesign the symbolic backend around CCG directed types (deterministic CKY + single linear decoder, 30K learnable parameters). Under the same BERT-base encoder, the system achieves 75.9$\pm$6.4% LF exact match, surpassing AM-Parser (70.8$\pm$4.3%). Per SLOG's own category groupings, gains are highly directional: the CCG system outperforms AM-Parser on all 5 position-shift categories (+29.9pp), while AM-Parser outperforms on all 6 recursive-depth categories. Replacing the encoder with DeBERTa-v3-large yields 90.7$\pm$4.9%, with the largest encoder gains in recursive-depth categories, complementary to directionality's gains. Directional representations shift the bottleneck from the symbolic layer (AM-Parser's 0% category ceiling) to the neural layer, which improves with encoder upgrades.
Chinese Translation
多个 SLOG 测试类别明确涉及方向性区分（修饰符位置变化、参数提取位置），然而之前的最先进模型 AM-Parser 使用的 AM 代数其操作并不编码方向。我们围绕 CCG 有向类型重新设计了符号后端（确定性 CKY + 单线性解码器，30K 可学习参数）。在相同的 BERT-base 编码器下，该系统实现了 75.9$ ext{±}$6.4% 的 LF 精确匹配，超越了 AM-Parser（70.8$ ext{±}$4.3%）。根据 SLOG 自身的类别分组，性能提升具有高度方向性：CCG 系统在所有 5 个位置变化类别上均优于 AM-Parser（+29.9 个百分点），而 AM-Parser 在所有 6 个递归深度类别上表现更佳。将编码器替换为 DeBERTa-v3-large 可获得 90.7$ ext{±}$4.9%，在递归深度类别中获得最大的编码器提升，这与方向性带来的提升相辅相成。有向表示将瓶颈从符号层（AM-Parser 的 0% 类别上限）转移到神经层，后者随着编码器的升级而改善。
cs.CL / 46 / 2607.02369
World Wide Models: Literary Tools for Cultural AI
全球模型：文化人工智能的文学工具
Abstract
LLMs stage a new form of cultural encounter that is massive, automated, and monolingual. Literary disciplines have always negotiated cultural struggles with comparative reading of literature, narratological and poetic analysis, critical theory, world literature, and translation. These tools have now become indispensable for building culturally literate AI. The essay develops a layered framework toward more nuanced textual models and pluralistic interpretations of AI, emphasizing the natural intersections of literature and AI development, connecting current debates in critical theory with structural monolingualism, and suggesting a new application of world literature approaches to address global AI textuality through macrostructure, circulation, and untranslatability.
Chinese Translation
大型语言模型（LLMs）展现了一种新的文化交汇形式，这种形式是巨大的、自动化的且单语的。文学学科一直通过对文学的比较阅读、叙事学和诗学分析、批判理论、世界文学和翻译来协商文化斗争。这些工具现在已成为构建具有文化素养的人工智能的不可或缺的部分。本文发展了一个分层框架，以实现对文本模型的更细致的理解和对人工智能的多元解读，强调了文学与人工智能发展之间的自然交集，将当前的批判理论辩论与结构性单语主义相连接，并建议将世界文学的方法应用于通过宏观结构、流通和不可翻译性来应对全球人工智能文本性的新途径。
cs.CL / 47 / 2607.02381
HULAT2 at MER-TRANS 2026: Governed Multi-Agent Simplification for Spanish Easy-to-Read Generation
HULAT2在MER-TRANS 2026：西班牙易读文本生成的多智能体简化方法
Abstract
This paper describes the participation of HULAT2-UC3M in the Spanish track of MER-TRANS 2026, a shared task on multilingual Easy-to-Read translation. Three fully automatic Spanish runs were submitted. RUN1 and RUN2 used a LangGraph-based multi-agent workflow combining Gemini 2.5 Flash and RigoChat-7B-v2, parallel generation strategies, internal quality signals, Event-Condition-Action routing, controlled editing and traceable decisions. RUN1 used the base workflow, while RUN2 activated an additional lexical-support layer based on a glossary and lexical resources. RUN3 was a RigoChat-based generate-evaluate-regenerate baseline with prompt engineering and LoRA-based adaptation. The official leaderboard reports BLEU-Orig, BLEU-Gold, SARI and BERTScore. During development, additional internal signals were also inspected, including semantic fidelity, readability, lexical simplicity, syntactic clarity and factual consistency. According to official SARI, RUN1 was the best HULAT2 run, with 44.0543 points, followed by RUN2 with 43.1049 and RUN3 with 38.5136. These results indicate that, in this task setting, signal-guided multi-agent routing outperformed the linear regeneration baseline. They also show that adding lexical support did not automatically improve reference-based scores. Further segment-level and document-level analysis are required to assess readability, factual consistency and user-oriented adequacy.
Chinese Translation
本文描述了HULAT2-UC3M在MER-TRANS 2026西班牙赛道的参与，该任务是一个关于多语言易读翻译的共享任务。提交了三组完全自动化的西班牙语运行结果。RUN1和RUN2采用了基于LangGraph的多智能体工作流程，结合了Gemini 2.5 Flash和RigoChat-7B-v2，采用并行生成策略、内部质量信号、事件-条件-动作路由、受控编辑和可追溯决策。RUN1使用了基础工作流程，而RUN2则激活了基于词汇表和词汇资源的额外词汇支持层。RUN3是一个基于RigoChat的生成-评估-再生成基线，采用了提示工程和基于LoRA的适应。官方排行榜报告了BLEU-Orig、BLEU-Gold、SARI和BERTScore。在开发过程中，还检查了其他内部信号，包括语义保真度、可读性、词汇简单性、句法清晰度和事实一致性。根据官方SARI，RUN1是HULAT2的最佳运行结果，得分为44.0543分，RUN2得分43.1049，RUN3得分38.5136。这些结果表明，在该任务设置中，信号引导的多智能体路由优于线性再生基线。同时也显示，添加词汇支持并未自动改善基于参考的得分。进一步的段落级和文档级分析是评估可读性、事实一致性和用户导向适宜性的必要条件。
cs.CL / 48 / 2607.02383
Know Your Source: A Public Knowledge Store for Media Background Checks
了解你的来源：一个用于媒体背景检查的公共知识库
Abstract
LLM-based retrieval-augmented generation (RAG) is increasingly used for automated fact-checking (AFC) and related tasks. By grounding LLM outputs in retrieved evidence, RAG-based systems provide transparent justifications while allowing external information to be updated independently of the underlying model. However, existing approaches often assume retrieved evidence is reliable, although real-world information may be conflicting, outdated, and can originate from unreliable or biased sources. Recent work on *source-critical reasoning* addresses this challenge through media background checks (MBCs) (Schlichtkrull, 2024), which assess the credibility of evidence sources to support downstream fact verification. However, generating MBCs relies on costly proprietary search APIs, limiting reproducibility. To mitigate this issue, we introduce MEDIAREF, a publicly available knowledge store of web-sourced documents that enables reproducible, low-cost evaluation of MBC generation across 200 media sources. We describe a reproducible methodology for constructing and updating the collection, assess widely used LLMs on the MBC generation task, and demonstrate that MEDIAREF supports higher-quality MBC generation through both automatic and qualitative evaluation.
Chinese Translation
基于大型语言模型（LLM）的检索增强生成（RAG）在自动化事实检查（AFC）及相关任务中越来越多地被使用。通过将LLM输出与检索到的证据相结合，基于RAG的系统提供了透明的论证，同时允许外部信息独立于基础模型进行更新。然而，现有的方法通常假设检索到的证据是可靠的，尽管现实世界中的信息可能存在冲突、过时，并且可能来源于不可靠或有偏见的来源。最近关于*来源关键推理*的研究通过媒体背景检查（MBCs）（Schlichtkrull, 2024）解决了这一挑战，该方法评估证据来源的可信度以支持下游的事实验证。然而，生成MBCs依赖于昂贵的专有搜索API，限制了其可重复性。为了解决这个问题，我们引入了MEDIAREF，一个公开可用的网络文档知识库，使得对200个媒体来源的MBC生成进行可重复、低成本的评估成为可能。我们描述了一种可重复的方法来构建和更新该集合，评估了广泛使用的LLM在MBC生成任务上的表现，并证明MEDIAREF通过自动和定性评估支持更高质量的MBC生成。
cs.CL / 49 / 2607.02416
The Future of NLP may not be at NLP Conferences: Scholarly Migration Patterns in Natural Language Processing
自然语言处理的未来可能不在自然语言处理会议上：自然语言处理领域的学术迁移模式
Abstract
Natural Language Processing (NLP) has traditionally been published in its core disciplinary venues like ACL. However, advances in Large Language Models (LLMs) has led to a blurring of the disciplinary lines between NLP and general Machine Learning (ML), with authors regularly publishing in venues from both fields. Here, we ask whether the disciplinary center of gravity is shifting. Using NLP research published from 2010 to 2026 and studies of both established and new authors, we find that a migration is taking place. First, comparing the pre- and post-LLM eras, established authors lost 19.2pp of share at flagship *ACL main-conference tracks while gaining 14.8pp in the newer Findings tracks, and general ML venues rose 8.6pp, even when adjusting for parallel growth in the fields. Second, among newer authors who debut with at least three first-author NLP-topic papers, the share whose work appears mostly at *ACL venues fell from 84% (2019) to 74% (2024), while the share appearing mostly at general ML venues rose from 5% to 21%. Using causal inference techniques, we estimate that these general ML venues confer a significant citation premium, which influences venue selection. Together, these results point to a significant shift in where NLP research is published.
Chinese Translation
自然语言处理（NLP）传统上在其核心学科场所如ACL上发表。然而，大型语言模型（LLMs）的进展导致NLP与一般机器学习（ML）之间的学科界限模糊，作者们定期在两个领域的场所发表论文。在此，我们探讨学科重心是否正在发生转移。通过分析2010年至2026年间发表的NLP研究以及对既有作者和新作者的研究，我们发现正在发生迁移。首先，比较LLM前后的时代，既有作者在旗舰*ACL主会议轨道上的份额下降了19.2个百分点，而在较新的发现轨道上增加了14.8个百分点，同时一般ML场所的份额上升了8.6个百分点，即使在调整了两个领域的平行增长后也是如此。其次，在至少有三篇第一作者NLP主题论文的新作者中，主要在*ACL场所发表的份额从2019年的84%下降到2024年的74%，而主要在一般ML场所发表的份额则从5%上升到21%。通过使用因果推断技术，我们估计这些一般ML场所提供了显著的引用溢价，这影响了场所选择。综合来看，这些结果指向NLP研究发表地点的显著转变。
cs.CL / 50 / 2607.02459
Language Models as Measurement Apparatus for Culture
语言模型作为文化测量工具
Abstract
Language models are increasingly used to quantify cultural phenomena, but what makes such measurement distinctively cultural? This paper argues that NLP work on culture is a material-discursive practice: the apparatus -- model, data, annotation, evaluation -- participates in constituting the cultural reality it measures, rather than passively recording it. Drawing on Karen Barad's concept of the agential cut -- the contingent boundary between phenomenon and instrument -- I show that the apparatus's substantive design choices draw such boundaries, and that the boundary is entangled from the start because language models have already internalized much of the cultural material they measure. I illustrate this through three case studies on television and film dialogue (measuring structure, interaction, and deviation) and three examinations of the apparatus itself (erasure of cultural markers, attunement to historical material, and agency in an agentic workflow). This big picture analysis proposes a research program that is theory-driven, empirically rigorous, and culturally contingent, treating each agential cut as a conscious commitment, at once methodological and ethical.
Chinese Translation
语言模型越来越多地被用于量化文化现象，但是什么使这种测量具有独特的文化性？本文认为，关于文化的自然语言处理（NLP）工作是一种物质-话语实践：测量工具——模型、数据、注释、评估——参与构成其所测量的文化现实，而不是被动地记录它。借鉴凯伦·巴拉德（Karen Barad）的“代理切割”（agential cut）概念——现象与工具之间的偶然边界——我展示了测量工具的实质性设计选择如何划定这些边界，并且这一边界从一开始就纠缠在一起，因为语言模型已经内化了它们所测量的许多文化材料。我通过三个关于电视和电影对话的案例研究（测量结构、互动和偏差）以及对测量工具本身的三次考察（文化标记的消除、对历史材料的调适以及在代理工作流程中的能动性）来说明这一点。这一宏观分析提出了一个以理论驱动、经验严谨和文化依赖为特征的研究计划，将每一个代理切割视为一种有意识的承诺，既是方法论的也是伦理的。
cs.CL / 51 / 2607.02464
Will Scaling Improve Social Simulation with LLMs?
扩展是否会改善大型语言模型的社会模拟？
Abstract
Large Language Model (LLM) social simulations are a promising research method, but they are not yet faithful enough to be adopted widely. In this work, we investigate whether the current scaling paradigm in language modeling is likely to close these gaps, or whether simulation fidelity is orthogonal to general capabilities and therefore deserving of more research attention. We use scaling laws to study the relationship between LLMs' compute scale, general capability benchmarks, and the fidelity of social simulation in three representative sub-domains: opinion modeling, behavioral simulation, and longitudinal forecasting. Surprisingly, we discover strong compute scaling in all three settings, using a suite of 85 transformer LLMs with the Qwen3 architecture pre-trained on the DCLM web text corpus under fixed-compute budgets from $10^{18}$ to $10^{20}$ FLOPs. Then we evaluate 35 larger and more capable open-weight models up to 70B parameters, allowing us to predict downstream accuracy from loss. This reveals that the majority of behavioral and opinion simulation tasks will rapidly improve with scale, particularly when they involve populations that are well-represented in English web corpora. Longitudinal forecasting and underrepresented opinions scale more slowly, especially when they are less correlated with general knowledge and reasoning benchmarks like MMLU. In behavior simulation, scaling fails to improve model calibration with human cognitive biases like risk aversion, as well as human heuristics like learning correlated rewards from related tasks. On these tasks, even fine-tuned models fail to noticeably scale up performance from 0.5B to 8B parameters. Taken together, we conclude that scale will improve social simulations in most settings, but outliers exist, and improvements will be less reliable in low-resource domains.
Chinese Translation
大型语言模型（LLM）社会模拟是一种有前景的研究方法，但尚未达到足够的忠实度以广泛采用。在本研究中，我们探讨当前语言建模中的扩展范式是否有可能弥补这些差距，或者模拟的忠实度是否与一般能力正交，因此值得更多的研究关注。我们使用扩展法则研究LLM的计算规模、一般能力基准与社会模拟的忠实度之间的关系，涵盖三个代表性子领域：意见建模、行为模拟和纵向预测。令人惊讶的是，我们在这三种设置中发现了强烈的计算扩展，使用了一套85个基于Qwen3架构的变换器LLM，这些模型在固定计算预算下（从$10^{18}$到$10^{20}$ FLOPs）预训练于DCLM网络文本语料库。随后，我们评估了35个更大且更强大的开放权重模型，参数规模高达700亿，使我们能够从损失预测下游准确性。这表明，大多数行为和意见模拟任务在扩展时会迅速改善，特别是当它们涉及在英语网络语料库中得到良好代表的人群时。纵向预测和代表性不足的意见扩展较慢，尤其是当它们与一般知识和推理基准（如MMLU）相关性较低时。在行为模拟中，扩展未能改善模型在风险厌恶等人类认知偏差以及从相关任务中学习相关奖励等人类启发式方面的校准。在这些任务中，即使是经过微调的模型也未能显著提升从5亿到80亿参数的性能。综合来看，我们得出结论，扩展将在大多数设置中改善社会模拟，但也存在例外，并且在低资源领域的改进将不那么可靠。
cs.CL / 52 / 2607.02473
Audio-Based Understanding of Audiobook Narration Appeal
基于音频的有声书叙述吸引力理解
Abstract
Narration is central to the audiobook listening experience, shaping how listeners engage with and understand the content. This work explores how narration qualities shape an audiobook's appeal, noting that their effects can vary by genre, title, and audience. We extract vocal and acoustic features (e.g., tone, pace, loudness) from LibriVox using pre-trained audio models and analyse their relationship with consumption data (specifically, view-rate) and their interplay with genre and title. Despite limited consumption data, we find that acoustic information alone has a robust association with appeal, even after accounting for title effects. We further validate these findings using more nuanced proprietary engagement metrics. To our knowledge, this is the first systematic computational study linking narration qualities, genre, title, and audiobook consumption, highlighting the potential of data-driven insights to improve audiobook personalisation and narrator casting.
Chinese Translation
叙述是有声书听觉体验的核心，塑造了听众与内容的互动和理解。本研究探讨了叙述特质如何影响有声书的吸引力，指出其效果可能因类型、标题和听众而异。我们从LibriVox提取了声音和声学特征（如音调、节奏、响度），并分析了它们与消费数据（特别是观看率）之间的关系，以及它们与类型和标题的相互作用。尽管消费数据有限，我们发现声学信息本身与吸引力之间存在显著关联，即使在考虑标题效应后也是如此。我们进一步使用更细致的专有参与度指标验证了这些发现。据我们所知，这是首个系统性计算研究，关联了叙述特质、类型、标题和有声书消费，突显了数据驱动洞察在改善有声书个性化和叙述者选角中的潜力。
cs.CL / 53 / 2607.02490
Visually Grounded Self-Reflection for Vision-Language Models via Reinforcement Learning
通过强化学习实现视觉基础的自我反思以提升视觉语言模型
Abstract
Large vision-language models can reason over multimodal inputs by generating textual chains of thought (CoT). A key capability exhibited in CoT reasoning is self-reflection: revisiting earlier decisions and correcting previous errors. However, existing LVLMs often fail to properly attend to visual inputs during reflection, limiting their ability to translate feedback into grounded corrections, especially for out-of-distribution images. To address this issue, we propose a novel reinforcement learning training framework VRRL, with two components explicitly designed to elicit visually grounded self-reflection. First, we randomly mask trajectory prefixes during training to emphasize recovery from incorrect intermediate predictions rather than making early mistakes. Second, we introduce buffered roll-ins from an experience replay buffer to expose the model to diverse failure states that it must learn to correct. We evaluate our approach on visual grounding tasks involving tables and charts, as well as spatial navigation benchmarks. While off-the-shelf and conventionally fine-tuned models degrade substantially under distribution shift, our method substantially improves average out-of-distribution accuracy over standard RL and reflection-oriented fine-tuning baselines by using self-reflection effectively.
Chinese Translation
大型视觉语言模型能够通过生成文本思维链（CoT）对多模态输入进行推理。CoT推理中展现的一个关键能力是自我反思：重新审视早期决策并纠正之前的错误。然而，现有的LVLM（Large Vision-Language Models）在反思过程中往往未能正确关注视觉输入，这限制了它们将反馈转化为基础纠正的能力，尤其是在处理分布外图像时。为了解决这一问题，我们提出了一种新颖的强化学习训练框架VRRL，其中包含两个专门设计的组件，以引发视觉基础的自我反思。首先，我们在训练过程中随机屏蔽轨迹前缀，以强调从错误的中间预测中恢复，而不是早期犯错。其次，我们引入来自经验重放缓冲区的缓冲回滚，以使模型接触到多样的失败状态，从而学习进行纠正。我们在涉及表格和图表的视觉基础任务以及空间导航基准上评估了我们的方法。尽管现成的和传统微调的模型在分布转移下显著降级，但我们的方法通过有效利用自我反思，显著提高了标准强化学习和以反思为导向的微调基线的平均分布外准确率。
cs.CL / 54 / 2607.02504
Reasoning LLM Improves Speaker Recognition in Long-form TV Dramas
推理大语言模型提升长篇电视剧中的说话人识别
Abstract
Long-form TV dramas present a formidable challenge for comprehensive video understanding, where deciphering complex storyline often relies on \textbf{speaker recognition}, the task of accurately attributing each spoken utterance to its respective character. In this paper, we advance this field through two primary contributions. (1) We introduce \textbf{DramaSR-532K}, a large-scale benchmark comprising 532K annotated dialogue lines across more than 900 unique characters, necessitating the integration of auditory, linguistic, and visual cues for speaker recognition. (2) We propose \textbf{DramaSR-LRM}, a robust approach built upon a large reasoning model (LRM). DramaSR-LRM is designed to autonomously aggregate contextual evidence via multimodal tool-use, synthesizing diverse inputs to achieve high-fidelity attribution. Experimental results demonstrate that DramaSR-LRM significantly outperforms existing baselines, particularly on short utterances where acoustic biometrics are inherently unreliable. \textit{All the data and code will be made publicly available at the project page: https://www.github.com/198808xc/DramaSR-LRM.}
Chinese Translation
长篇电视剧对全面的视频理解提出了严峻的挑战，其中解读复杂的故事情节往往依赖于说话人识别，即准确地将每个口语发言归属到相应的角色。在本文中，我们通过两个主要贡献推动了这一领域的发展。(1) 我们引入了 extbf{DramaSR-532K}，这是一个大规模基准数据集，包含532K注释的对话行，涉及900多个独特角色，要求整合听觉、语言和视觉线索以进行说话人识别。(2) 我们提出了 extbf{DramaSR-LRM}，这是一种基于大型推理模型（LRM）的稳健方法。DramaSR-LRM旨在通过多模态工具使用自主聚合上下文证据，综合多样的输入以实现高保真度的归属。实验结果表明，DramaSR-LRM显著优于现有基准，特别是在声学生物特征本质上不可靠的短语句上。 extit{所有数据和代码将在项目页面公开发布： https://www.github.com/198808xc/DramaSR-LRM。}
cs.CL / 55 / 2607.02513
LACUNA: A Testbed for Evaluating Localization Precision for LLM Unlearning
LACUNA：评估大规模语言模型去学习定位精度的测试平台
Abstract
LLMs memorize sensitive training data, including personally identifiable information (PII), creating a pressing need for reliable post hoc removal methods. Unlearning has emerged as a promising solution, with state-of-the-art(SOTA) methods often following a localize-first, unlearn-second paradigm that targets specific model parameters. However, existing benchmarks evaluate unlearning solely at the output level, leaving open the question of whether unlearning truly erases knowledge from a model's parameters or merely obfuscates it, a concern reinforced by the success of resurfacing attacks. To bridge this gap, we introduce LACUNA: the first unlearning testbed with ground-truth parameter-level localization. LACUNA injects PII of synthetic individuals into predefined parameters of 1B and 7B OLMo-based models via masked continual pretraining, enabling direct evaluation of whether unlearning targets the weights responsible for knowledge storage. We use LACUNA to benchmark current SOTA unlearning methods and find that, despite strong output-level performance, existing methods are highly imprecise and susceptible to resurfacing attacks. We further show that when localization is successful, even a simple gradient-based unlearning method achieves strong erasure and robustness to resurfacing attacks, highlighting the importance of precise unlearning. We release LACUNA to complement behavioral evaluations and drive further advances in robust, localization-based unlearning.
Chinese Translation
大规模语言模型（LLMs）会记忆敏感的训练数据，包括个人可识别信息（PII），因此迫切需要可靠的事后删除方法。去学习作为一种有前景的解决方案应运而生，当前的最先进（SOTA）方法通常遵循先定位、后去学习的范式，针对特定的模型参数。然而，现有基准仅在输出层面评估去学习，未能解答去学习是否真正从模型参数中抹去知识，还是仅仅使其模糊化，这一问题因复现攻击的成功而愈加突出。为了解决这一问题，我们引入了LACUNA：第一个具有真实参数级别定位的去学习测试平台。LACUNA通过掩蔽持续预训练将合成个体的PII注入到1B和7B OLMo基础模型的预定义参数中，从而直接评估去学习是否针对负责知识存储的权重。我们利用LACUNA对当前的SOTA去学习方法进行基准测试，发现尽管输出层面表现强劲，但现有方法的精确度极低，且容易受到复现攻击的影响。我们进一步表明，当定位成功时，即使是简单的基于梯度的去学习方法也能实现强有力的抹除效果，并对复现攻击具有较强的鲁棒性，强调了精确去学习的重要性。我们发布LACUNA，以补充行为评估并推动基于定位的鲁棒去学习的进一步发展。